Trevor David McKee
INSiE
OF TECHNOLOY
at the
OCT 27 2005
LIBRARIES
A.
/
-
· · ...... · · · · · · .
· · · · · · .
Rakesh. · . · · .....
· Ja·
>--- ..
-
Accepted by ..........- .......................
7ha
ARCHIVES

This Doctoral Thesis has been examined by the following Thesis Committee:
Rakesh K. Jain, Ph.D.
Thesis Supervisor
Andrew Werk Cook Professor of Tumor Biology
Harvard Medical School
Massachusetts General Hospital
Boston, Massachsetts
.. , ; ... · o~ o. · · · · .
Peter T. C. So, Ph.D.
Thesis Supervisor
Professor of Mechanical and Biological Engineering
Massachusetts Institute of Technology
Cambridge, Massachusetts
William M. I)een, Ph.D.
Thesis Committee Chair
Carbon P. Dubbs Professor of Chemical and Biological Engineering
Massachusetts Institute of Technology
Cambridge, Massachusett an S .
.
.
Brian Seed, Ph.D.
Professor of Genetics and Health Sciences and Technology
Harvard Medical School
Massachusetts General Hospital
Boston, Massachusetts
Ioann~/ V. Yannas, Ph.D.
Profe or of Polymer Science and Engineering
Department ol Mechanical Engineering,
Materials Science and Engineering, and Biological Engineering
Massachusetts Institute of Technology
Cambridge, Massachusetts
3

Submitted to the Biological Engineering Division on May 3, 2005, in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Biological Engineering
Abstract
The extracellular matrix of tumors is a major barrier to the delivery of molecular medicine. We used fluorescence recovery after photobleaching combined with intrav-
found that the presence of fibrillar collagen correlated with hindered diffusion in vivo, and also in vitro, in collagen gels prepared to mimic tumor extracellular matrix. Modification of the tumor collagen matrix directly with purified bacterial collagenase, or indirectly with relaxin treatment, resulted in increased diffusion coefficients of macromolecules within tumors in vivo. In order to quantitate the changes in collagen content and structure induced by relaxin treatment, we adapted and further developed the imaging technique of intravital second harmonic generation microscopy. Using second harmonic generation imaging in combination with a fluorescently labeled gene therapeutic vector, we demonstrated that the spread of these viral vectors within tumors is limited by the fibrillar collagen in the extracellular matrix. Matrix modification via the introduction of bacterial collagenase along with the initial virus injection resulted in a significant improvement in the range of viral distribution within the tumor. This resulted in an extended range of infection of cells within the tumor, and improved virus propagation, ultimately leading to enhanced therapeutic outcome. Thus, we show that fibrillar collagen is an important barrier to the distribution of molecular medicine within tumors, and that matrix modifying treatments can significantly enhance both vector distribution, as well as ultimately therapeutic response.
Thesis Supervisor: Rakesh K. Jain
Title: Andrew Werk Cook Professor, Harvard Medical School
Thesis Supervisor: Peter T. C. So
Title: Professor of Mechanical and Biological Engineering
5

This thesis would not have been possible were it not for the support of many people within the laboratory, and my friends and family.
I would like to dedicate this thesis to my family: To my parents for their unlimited support and encouragement, for sacrificing much, personally and professionally, to move our family to the United States from South Africa in 1990 in search of better opportunities for me and my brother, and for teaching me the meaning of persistence by not giving up despite setbacks after the move. To my brother, for being a great friend and travel buddy, and for allowing me to occasionally take advantage of his gifts with medical illustration to help me with slides, diagrams and figures. To my grandfather,
Allan Trevor Montague McKee, for his continuous moral support, and for being an inspirational role model of someone who overcame tremendous obstacles to achieve success through hard work. I can only hope that I will be as adventurous, as healthy, and as eager to learn about new ideas and concepts when I am 88! To my grandmother, Mary
Joyce Parfit, who sadly passed away during my tenure at MIT, for being the true personification of grace, dignity, and courage, who showed tremendous inner strength despite health problems, and for all of her efforts to teach me correct manners, posture, and grammar (with varying degrees of success!). And last but not least, to my girlfriend
Jenn, who supported and encouraged me in countless ways during the writing of my thesis, for her countless hours of encouragement on the phone, for the large batches of frozen dinners prepared in advance of my thesis crunch, and for being there for me whenever I needed support.
Thanks also to Ali and Emilie, for being great friends through a tough time, to Paola,
Wilson, Ed, Mike, Peigen, Sergey, Leo, Yves, and the whole Steele Lab for being terrific people to work with, and for all of the late night companionship. Thanks to Coffee
Central at Mass General Hospital for feeding my co-addiction to caffeine and cinnamon raisin bagels, and being open until 11. Thanks to everyone in the Grad Student Council and at Fenway House for giving me many much-welcome distractions from lab work, and providing opportunities for me to contribute and help out my fellow graduate students.
Thanks to Mike Folkert for picking up my slack when I disappeared from the GSC
Housing and Community Affairs chairmanship to meet thesis deadlines, and for just being a superb advocate for graduate students in general. And thanks to MIT Masters
Swim Club for keeping me from being entirely out of shape, and coach Bill for being a great friend and always willing to chat and listen to my various random complaints. And finally, thanks to Bryan Hiles for being a great friend despite huge gaps in time and tremendous distances, and for always welcoming me back during my visits to SA.
7

9/99 to 6/05
Massachusetts Institute of Technology
Ph.D. in Bioengineering, June 2005
- 5 years experience in small animal (mouse) surgery; in the design, setup and operation of multiphoton laser scanning microscopes; and in the measurement of transport in vivo.
Active in the graduate student council, in teaching and grading assistantships, and as an undergraduate research mentor.
9/95 to 6/99
University at Buffalo (S.U.N.Y.)
B.S. in Chemical Engineering, with a minor in Biotechnology,
Graduated Magna Cum Laude
- Undergraduate research projects included the application of physiological models of the transport of inert gases within blood vessels to the understanding of decompression sickness; the modeling of dielectrophoresis to investigate whether it could be used for the separation of viral particles; and the programming of a cone-plate viscometer to mimic pulsatile blood flow. Active in the formation of a new Biomedical Engineering Society student chapter, and in the establishment of a community service organization.
6/98 to 8/98
GIBCO/I Life Technologies, Inc. (now Invitrogen) Grand Island, NY
Summer internship, Quality Engineering Dept.
6/99 to 8/99 Research internship, Cell Culture Research and Development Dept.
- Trained in FDA cGMP requirements, cell culture techniques, and bioreactor operation.
Manuscripts under review:
1.McKee TD*, Grandi P*, Mok W*, Alexandrakis G, Boucher Y, Breakefield XO, Jain
RK. Matrix modification combined with oncolytic herpesvirus gene therapy results in improved gene vector distribution and therapeutic efficacy. Submitted, May 2005.
* equal contribution
2.Demou ZN, Awad M, McKee TD, Wang X, Munn LL, Jain RK, Boucher Y. Lack of telopeptides in fibrillar collagen promotes membrane deformability and amoeboid invasion of breast adenocarcinoma. Cancer Research, in press.
3.Huang P, McKee TD, Fukumura D, Jain RK. A novel GFP expressing tumor model derived from a spontaneous osteosarcoma in a VEGF-GFP transgenic mouse.
Comparitive Medicine, in press.
Manuscripts published:
4.Alexandrakis G, Brown EB, Tong RT, McKee TD, Campbell RB, Boucher Y, Jain
RK. Two-photon fluorescence correlation microscopy reveals the two-phase nature of transport in tumors. Nature Medicine 10(2): 203-7, 2004.
5.Znati CA, Rosenstein M, McKee TD, Brown E, Turner D, Bloomer WD, Watkins S,
Jain RK, Boucher Y. Irradiation reduces interstitial fluid transport and increases the collagen content in tumors. Clinical Cancer Research 9(15): 5508-13, 2003.
9

6.McKee T*, Brown E*, diTomaso E, Pluen A, Seed B, Boucher Y, Jain RK. Dynamic imaging of collagen and its modulation in tumors in vivo using second-harmonic generation. Nature Medicine 9(6): 796-800, 2003. * published as Brown E, McKee
T et al., with equal contributions from the two first authors.
7.Ramanujan S, Pluen A, McKee TD, Brown EB, Boucher Y, Jain RK. Diffusion and convection in collagen gels: implications for transport in the tumor interstitium.
Biophysical Journal 83(3): 1650-60, 2002.
10.Koike C, McKee TD, Pluen A, Ramanujan S, Burton K, Munn LL, Boucher Y, Jain
RK. Solid stress facilitates spheroid formation: potential involvement of hyaluronan.
11.Pluen A, Boucher Y, Ramanujan S, McKee TD, Gohongi T, di Tomaso E, Brown EB,
Izumi
Y',
Campbell RB, Berk DA, Jain RK. Role of tumor-host interactions in interstitial diffusion of macromolecules: cranial vs. subcutaneous tumors.
Proceedings of the National Academy of Sciences USA 98(8): 4628-33, 2001.
l.McKee TI), Grandi P, Mok W, Boucher Y, Jain RK. Relaxin enhances drug delivery to tumors via cell-mediated modifications to the collagen matrix. Oral presentation,
International Conference on Relaxin and Related Peptides 2004, Jackson Hole, WY.
2.McKee TID, Brown EB, diTomaso E, Pluen A, Seed B, Boucher Y, Jain RK. Relaxin enhances drug delivery to tumors by permeabilizing the collagen matrix: Insight from second harmonic generation microscopy. Poster presentation, Gordon Research
Conference on Signal Transduction by Engineered Extracellular Matrices, June 27 -
July 2, 2004, Bates College, Lewiston, ME.
3.McKee TDI), Brown EB, diTomaso E, Pluen A, Seed B, Boucher Y, Jain RK. Dynamic imaging of collagen and it's modulation in tumors in vivo using second harmonic generation. Oral and Poster Presentation, Hot Topics from Selected Abstracts section of Gordon Research Conference on Collagen, July 27 -August 1, 2003, Colby-
Sawyer College, New London, NH.
4.McKee TD), Brown EB, diTomaso E, Pluen A, Seed B, Boucher Y, Jain RK. Dynamic imaging of collagen and it's modulation in tumors in vivo using second harmonic generation. Oral and Poster Presentation at the
9 4 th
Annual Meeting of the American
Association for Cancer Research, July 11-14, 2003, Washington, DC. Proceedings of the AACR 44(2):R984, 2003.
5.McKee TD, Pluen A, Boucher Y, Ramanujan S, Seed B and Jain RK. Relaxin improves the transport of large molecules within tumors. Poster presentation at the
Gordon Research Conference on Lasers in Medicine and Biology, July 12-15, 2002,
Kimball Union Academy, Meriden, NH.
6.McKee TD, Pluen A, Boucher Y, Ramanujan S, Seed B and Jain RK. Relaxin increases the transport of large molecules in high collagen content tumors. Poster presentation at the
92 nd
Annual Meeting of the American Association for Cancer
Research, March 2001, New Orleans, LA. Proceedings of AACR 42(1):158, 2001.
7.Van Liew HD, McKee TD. Effect of diffusion of inert gas in or out of arterial and venous blood vessels on washin and washout of tissues. Presentation at Gulf Coast
Chapter of the Undersea and Hyperbaric Medical Society, Galveston, TX, March 12-
15, 1998.
10

Title Page
Thesis Committee Page
Abstract
Acknowledgements
Biographical Sketch (Curriculum Vitae)
Table of Contents
List of Figures
List of Tables
1 Thesis Outline - Original contributions
1.1 Chapter 3: Tumor Host Interactions ............................
1.2 Chapter 4: Collagen Gel Transport .............................
1.3 Chapter 5: Second Harmonic Imaging ..........................
1.4 Chapter 6: Improving Gene Therapy ..........................
1.5 Chapter 7: Conclusions and future directions ...................
2 Motivation
2.1 Cancer therapy .........................................
2.2 Systemic therapy ...........................................
2.3 Novel targeted therapeutics ...................................
2.4 Transport of therapeutics within tumors .........................
2.5 The tumor extracellular matrix .................................
2.6 Matrix modification .........................................
2.7 Transport within the tumor extracellular matrix ...................
2.8 References ................................................
3 Tumor Host Interactions
3.2 Materials and Methods .......................................
3.2.1 Fluorescent Tracers ...................................
11
.
17
18
19
21
22
23
1
3
5
7
9
11
15
16
39
41
41
25
29
32
35
36
25
26
27
28

3.2.2 Animals and tumors ...................................
3.2.3 Diffusion measurements by FRAP .......................
3.2.4 Extracellular space organization .........................
3.2.5 Immunohistochemistry ................................
3.2.6 Electron microscopy ..................................
3.3 Results ..................................................
3.3.1 Interstitial diffusion decreases with increasing molecular size ..
3.3.2 Diffusion is faster in CW tumors than in DC tumors .........
3.3.3 The tumor capsule has a high density of fibroblast-like cells ...
3.3.4 DC tumors have high levels of fibrillar collagen type I .......
3.3.5 Decorin is restricted to the tumor periphery ................
3.3.6 Hyaluronan staining differences in CW and DC tumors ......
3.3.7 The ECM is of host origin .............................
3.4 Discussion ................................................
3.4.1 Importance of diffusion in drug design and selection ........
3.4.2 Contributions of tortuosity to diffusional hindrance .........
3.4.3 Role of ECM composition and organization in transport .....
3.5 Conclusions ...............................................
53
54
54
55
55
58
60
45
45
45
47
41
42
43
43
48
49
52
53
4 Collagen Gel Transport
4.1 Introduction ...............................................
4.2 Materials and Methods ......................................
4.2.1 Experimental Techniques ..............................
4.2.2 Theoretical Models ...................................
4.3 Results ...................................................
4.3.1 Collagen gel imaging reveals heterogeneous fibrillar assembly..
4.3.2 Collagen gels significantly hinder molecular diffusion .........
4.3.3 Gel diffusion data closely match measurements in tumors ......
4.3.4 Gelation of collagen solutions does not affect hindrance .......
4.3.5 Tracer diffusion in gels is influenced by method of preparation ...
4.3.6 The effective medium model underpredicts gel permeability ....
71
71
73
75
77
79
80
64
64
65
65
70
12

4.3.7 Measured gel permeability does not match tumor permeability .. 80
4.4 Discussion ................................................. 82
4.4.1 Collagen accounts for most of the tumor diffusional hindrance.. 82
4.4.2 Unassembled collagen is implicated in gel diffusive hindrance.. 83
4.4.3 Pure collagen offers more hindrance than pure hyaluronan ..... 85
4.4.4 Collagen gels pose a greater diffusive than hydraulic barrier ... 86
4.5 Conclusions ....... .. ......
4.6 References .. ............................................. .
88
5 Second Harmonic Imaging
5.2 Materials and Methods ......................................
91
92
5.2.1 Surgery and Imaging ..................................
5.2.2 In vitro SHG and autofluorescence signals .................
5.2.3 In vivo SHG signal of different tumor types .............
5.2.4 Collagen quantification with immunostaining ...............
5.2.5 Diffusion measurements ............. ..... .............
92
... 93
93
94
5.2.6 Enzyme dynamics ...................... ..............
92
94
5.2.7 Statistics .
....................................... 95
5.3 Results ..................... ..............................
5.3.1 Validation of collagen imaging by SHG in tumors ...........
96
96
5.3.2 Relationship of tumor SHG to diffusive transport ............
5.3.3 Dynamic imaging of collagen modification .........
101
........ 104
5.3.4 Relaxin enhances transport in tumors ...................
5.4 Discussion ...................
5.5 References ........
............. ................
.. 107
109
..................................... 112
6 Improving Gene Therapy
6.1 Introduction ............................ ....................
6.2 Materials and Methods ............. ..........................
6.2.1 Cell culture .. .
........
6.2.2 Viral vectors ...................................
...................... .........
.......
13
114
114
115
115
115

6.2.3 Dorsal skinfold window preparation ...................
6.2.4 Injection and imaging of labeled vectors ...................
6.2.5 Image analysis ......................................
... 116
116
.
117
6.2.6 Flank tumor growth delay .............................. 117
6.2.7 Immunostaining ...................................... 118
6.2.8 Statistical analysis .................................... 118
6.3 Results ......... ...... .... ............. .... 119
6.3.1 Virus distribution is hindered by collagen rich regions ....... 119
6.3.2 Collagenase improves virus distribution and gene expression .. 122
6.3.3 Collagenase enhances the efficacy of oncolytic viral therapy ... 122
6.3.4 Initial improved viral distribution improves efficacy ......... 127
6.4 Discussion .................. .............................. 128
6.5 References ................................................ 133
7 Conclusions / Future Directions
7.1 Introduction ........ .... ..............
136
... .. 136
7.2 Diffusive transport mechanisms within the tumor ECM ............. 137
7.3 Matrix modifying treatments and cancer therapy .................. 140
7.4 Conclusions ....... ............. ............ 142
7.5 References ........................................... 145
14

2.1 90 nm fluorescent nanoparticles within a tumor .................
2.2 Schematic of extracellular matrix composition ..................
2.3 Biochemical analysis of four tumor types .......................
3.1 Effective diffusion coefficients of tracers in vivo .................
3.2 Extracellular space in vivo ................................... 49
3.5 Electron microscopy of collagen fibril organization in vivo ........ 52
29
31
32
46
3.6 The effect of hyaluronidase treatment on IgG diffusion in vivo .....
4.1 Confocal reflectance microscopy of collagen gels .................
4.3 Effective diffusion coefficients of tracers in gels ..................
4.4 Schematic of tortuosity experienced by molecules in vivo ..........
58
72
74
76
4.5 Comparisons between gel and in vivo diffusion data ..............
4.6 Comparison between collagen solution and gel diffusion data ......
4.7 Comparison between experimental and theoretical permeability...
4.8 Comparison between gel and in vivo permeability ................
5.1 Second harmonic generation imaging of tumors in vivo ...........
5.2 SHG images of fibrillar collagen I in tumors ...................
5.3 Dependence of SHG signal on tumor type ......................
78
79
81
82
96
.
98
102
5.4 Effect of collagenase and relaxin on tumor collagen dynamics .....
6.2 Effect of collagenase on oncolytic viral therapy ..................
105
124
6.3 Effect of collagenase on MGH2-induced tumor growth delay ...... 126
6.4 Immunostaining analysis of tumor cell infection .................
6.5 A representative model of oncolytic viral distribution .............
128
130
15

5.1 Comparison of SHG with collagen and diffusion coefficients .......
103
16

In this thesis I will document the work I have done in quantitating the transport properties of the tumor interstitium in order to improve the delivery and efficacy of cancer therapeutics. The large size of many novel therapeutics impairs their transport through the tumor extracellular matrix and thus limits their therapeutic effectiveness. We propose that extracellular matrix composition, structure, and distribution determine the transport properties in tumors. Thus, the use of matrix modifying agents in combination with the delivery of therapeutically relevant macromolecules within tumors, should result ultimately in improved therapeutic efficacy. I will attempt to prove this hypothesis over the course of the work contained in this thesis. What follows is a brief summary of the work that has been performed for each chapter in my thesis, where each chapter fits in to the larger picture, and what I have contributed towards each chapter.
Chapter 2 is a brief introduction to cancer therapy and describes our motivation for investigating the movement of macromolecules and viral vectors within tumors. The goal of my work is to devise methods to modify the tumor extracellular matrix in order to improve the delivery of cancer therapeutics. To do this, we first need to understand the nature of transport within the tumor interstitium, we then need to devise methods to be
17

able to monitor transport within the tumor interstitium, and finally be able to test our predictions using a relevant therapeutic agent. Chapters 3 and 4 describe our work at trying to further the current understanding of the nature of transport within the tumor interstitium. Chapter 5 describes the application of second harmonic generation imaging to the tumor interstitium in order to monitor the changes in fibrillar collagen that are induced by relaxin, a matrix modifying agent. And finally Chapter 6 describes the use of a gene therapeutic agent, herpes simplex virus, in combination with an alternate matrix modifying agent, purified bacterial collagenase, to test whether matrix modification can have an effect in a clinically relevant model system.
Chapter 3 describes the work done to characterize the role of the host in determining the extracellular matrix content of tumors. In this chapter, we determined the diffusion coefficients of a number of different tracer molecules within the extracellular matrix of tumors grown in two different anatomical locations, the skin and the pial surface of the brain. We showed that diffusion of macromolecules and large particles in the skin tumors was significantly hindered in comparison to the brain tumors. We used immunohistochemistry and quantitative image analysis to show an increase in extracellular space, the number of stromal cells, and the amount of collagen type I within skin tumors in comparison to brain tumors. These results build upon a previous study[l ], which indicated that collagen content was an important determinant of macromolecular transport in vivo, and that modification with bacterial collagenase could improve transport. Our study in chapter 3 improved upon this result by finding that the tumor
18

extracellular matrix derived from the presence of host cells within the tumor, and by quantitating the transport properties within the tumor matrix of tracers spanning several orders of magnitude in size. We also quantitated the role of tortuosity due to the cellular components of the tumor in influencing transport.
This chapter was published in the Proceedings of the National Academy of Sciences,
U.S.A. in 2001[2], and was a collaborative study with the help of the following coauthors: A. Pluen, Y. Boucher, S. Ramanujan, T.D. McKee, T. Gohongi, E. di Tomaso,
E.B. Brown, Y. Izumi, R.B. Campbell, D.A. Berk, and R.K. Jain. My contribution to this work included the injection of fluorescent tracers within the tumors, and the acquisition of fluorescence photobleaching recovery (FRAP) data from these tumors for a number of the tracers used in the study. I also wrote computer programs for the quantification of immunohistochemical staining data, and the quantification of parameters such as extracellular space and the presence of host tumor cells within tissue sections. In summary, in chapter 3, we confirmed that the transport of macromolecules within tumors is limited by the presence of collagen I within tumors, and that this collagen is produced by the host cells within the tumor.
1.2 Chapter 4: Collagen Gel Transport
Building upon this result, we then undertook a study to determine the transport properties of collagen and hyaluronan gels prepared in vitro to match the content of these molecules present in the tumor extracellular matrix. The results of this work are described in chapter 4. We prepared these gels in vitro in order to more closely control the properties of the resulting gel, as the complexity of the in vivo microenvironment prevented a
19

detailed investigation of the transport properties of the tumor extracellular matrix. While other groups had investigated the transport properties of tracers within gels of extracellular matrix components[3, 4], we went further by relating the diffusion of tracers in these in -vitro preparations to measurements made in tumors of similar collagen content. We discovered that tracer diffusion data from these collagen type I gels prepared in vitro, when corrected for cell tortuosity, were in good agreement with data in
vivo in tumors of comparable collagen content. Hyaluronan gels prepared at similar concentrations to those present in vivo did not pose a significant diffusive barrier to tracer transport. We determined from gel imaging and transport studies that unassembled collagen was present in the gel void spaces and contributed to diffusive hindrance, which was validated through the use of alternate preparation techniques. We measured the permeability of collagen gels, and compared these results to predictions based on a fit of the diffusion data to an effective medium model. The experimental permeability measurements matched the model predictions only for the lower range of collagen content studied, indicating that high collagen concentration poses a greater barrier to diffusion than convection. This observation agrees with the in vivo permeability measurements of Netti et al.[ ], and the in vivo diffusion data described in chapter 3.
This chapter was published in the Biophysical Journal in 2002[5], and was a collaborative study with the help of the following co-authors: S. Ramanujan, A. Pluen,
T.D. McKee, E.B. Brown, Y. Boucher and R.K. Jain. My contribution to this work included assisting in the preparation of collagen and hyaluronan gels of the appropriate concentrations, and the acquisition of diffusion and permeability data within these gels. I prepared hyaluronan gels of greater concentrations, and showed more significant
20

diffusive hindrance, in agreement with a previous study[3]. I proposed and developed an alternate preparation method to test the role of unassembled collagen present between collagen fibers, and also performed imaging of the fibrillar nature of these gels. I assembled a device to test for leaching of collagen during the permeability measurements, and used it to show negligible leaching. I also quantitated the amount of extracellular space present in tissue sections. In summary, this study indicated that the mechanism of diffusive hindrance of macromolecular tracers within high collagen content tumors can be reproduced in large part by gels containing solely collagen of similar concentrations.
As described. in chapters 3 and 4, I used two photon microscopy for the quantitative imaging of tissue sections stained using immunohistochemistry, and for the imaging of collagen gels in vitro, using second harmonic generation. The ability to image collagen
in vivo in tumors using the same technique would be of great benefit to be able to quantitate changes induced in the collagen matrix by matrix modifying treatments. We therefore decided to build upon this work by developing the imaging methodology of second harmonic generation imaging of collagen in vivo. This work is described in chapter 5. We determined that the second harmonic signal coming from tumors grown in
vivo was proportional to the content of collagen in those tumors. Using collagen gels of known concentrations, we determined that second harmonic signal scaled linearly with collagen concentration. We determined that the second harmonic signal arises from fibrillar collagen type I in these tumors by comparing immunostaining for collagen type I
21

with second harmonic signal in tissue sections. We then used this technique to obtain quantitative information on the dynamics of collagen modification in vivo. We administered bacterial collagenase to the surface of tumors, and were able to show an exponential decay of second harmonic signal with time, which scaled linearily with the concentration of collagenase used. We treated the tumors with the hormone relaxin, and quantitated the resulting change in collagen content and structure with time.
Interestingly,, we were able to show using this technique that relaxin upregulated the rate of degradation of fibrillar collagen in these tumors, but did not change the overall collagen content. This indicated an increased turnover of collagen. Additionally, measurements of macromolecular diffusion coefficients within tumors treated with relaxin indicated increased molecular mobility in comparison to control tumors. We were thus able to show that collagen structure, as well as content, plays a critical role in the diffusion of molecules within the tumor interstitium.
This chapter was published in the journal Nature Medicine in 2003[6], and was a collaborative study with the help of the following co-authors: E.B. Brown, T.D. McKee,
E. di Tomaso, A. Pluen, B. Seed, Y. Boucher and R.K. Jain; with equal contributions from the first two authors. I was involved in all aspects of the preparation and implementation of this study.
While the diffusion of macromolecular tracers within tumors tells us a lot about how therapeutically relevant molecules might move within tumors, it would be useful to test our models using a therapeutically active agent. To this effect, we began studies using
22

herpes simplex virus particles, which have been in use clinically for the treatment of brain tumorsl[7]. These results are described in chapter 6. We obtained GFP labeled viral particles from our collaborators (Paola Grandi and Xandra Breakefield), and were able to show that, upon injection into tumors, these particles were only able to penetrate regions of the tumor that were devoid of fibrillar collagen, which was imaged using second harmonic generation. We showed that the addition of purified bacterial collagenase to the viral mixture allowed the virus to penetrate a greater area of tumor. Using oncolytic viral vectors., which selectively replicate within and destroy tumor cells, we were able to show an improved therapeutic response with the addition of bacterial collagenase. We used imaging of GFP expression and viral proteins labeled with immunohistochemistry in tissue sections to document the mechanism of this improved viral spread. This work has been submitted for publication to the journal Nature Biotechnology, and was a collaborative study with the help of the following co-authors: T.D. McKee, P. Grandi,
W. Mok, G. Alexandrakis, Y. Boucher, X.O. Breakefield and R.K. Jain, with equal contributions from the first three authors. I was involved in all aspects of the planning and implementation of this study.
Finally, in chapter 7 I summarize the conclusions we have made regarding the role of collagen in the transport of macromolecules within the tumor extracellular matrix. In conclusion, I have demonstrated that the fibrillar collagen present within tumors plays an important role in the distribution of macromolecules and therapeutic particles. Using optical methods such as FRAP I was able to quantitate differences in diffusion between
23

tumors grown in two different organs, and relate difference in diffusion to the structure and content of fibrillar collagen in the two sites of tumor implantation. We showed that gels of collagen type I prepared in vitro could match the diffusive properties of tumors, indicating that fibrillar collagen poses the main barrier to the transport of macromolecules and therapeutic particles within tumors. We developed the technology of second harmonic generation imaging of tumors in order to quantitate the changes induced in the collagen matrix by the hormone relaxin, and discovered that relaxin acts within tumors by upregulating both the destruction of old matrix as well as the production of new matrix, resulting in increased collagen turnover within the tumor, which also leads to increased diffusion of macromolecules. And finally, we show that the presence of fibrillar collagen within tumors limits the efficacy of a gene therapeutic agent through the use of fluorescently labeled herpes virus gene therapeutic particles. Modification of the tumor collagen using bacterial collagenase results in improved viral spread within tumors, resulting ultimately in improved viral efficacy.
24

Based on estimates, there will be 1.4 million new cases of cancer in the United States in
2005, and an estimated 570,000 deaths attributed to cancer[8]. Cancer is the second leading cause of death behind heart disease in the United States, with 90% of all cancer cases arising initially as solid tumors [8]. Solid tumors consist of abnormal cells that have evaded the normal controls on cell growth and division, and acquired other characteristics that provide for their continued survival and growth [9]. Carcinogenesis, or cancer formation, is a multistep process; genetically inherited mutations or exposure to mitogenic materials can accelerate this process.
Current methods for medical treatment of solid tumors involve generally one or more combinations of 3 classes of therapy: surgery, radiation and chemotherapy. Surgery and radiation are local therapies, while chemotherapy acts systemically. Surgery offers the possibility of a cure generally for tumors that are caught early in the progression of disease, and for those tumors growing in surgically accessible locations, where the
25

removal of the tumor will not compromise the function of the organ it is located in.
Radiation therapy kills cancer cells by causing lethal damage to DNA, and is focused using multiple beam paths to treat solid tumors, even those growing in surgically inaccessible locations. Radiation therapy offers the potent ability to affect logarithmic reductions in the number of viable cancer cells within tumors, better than any chemotherapeutic to date, but is generally limited by radiation-induced damage to normal tissue [10].
During the course of tumor progression, malignant cells often spread to other organs from the initial site of tumor growth via the blood or lymphatic vasculature, a process termed metastasis[11]. This metastatic spread necessitates the use of a systemically acting therapy, differing from the more localized treatments of surgery and radiotherapy.
Historically, chemotherapeutic agents have been small molecules that have been chosen for their ability to differentially affect cancerous cells while attempting to cause less harm to healthy cells within the body[12]. While an ideal chemotherapeutic drug would only affect cancer cells and have no effect on host cells, there is always some level of host cell toxicity that limits the amount of drug that can be used to treat a patient. In practice, cancer therapy often involves the combination of surgery, radiation and chemotherapy to most effectively treat the disease[13].
26

New classes of therapeutics are exploiting other aspects of tumor progression, for example the fact that the tumor has to recruit blood vessels to sustain itself, a process known as angiogenesis, has led to anti-angiogenic therapy [14]. The ability of molecular biology techniques to be able to identify specific molecules that are upregulated or pathways that are dysregulated in malignant cells has led to a much greater understanding of carcinogenesis and cancer progression. Based on these discoveries new therapies have been developed, often involving the creation of antibodies[ 15]., or even larger therapeutic particles such as liposomes[16, 17] and gene therapy vectors[7]. These new types of therapies have often been referred to as the agents of molecular medicine.
These therapies offer great promise, based on the ability to more selectively target cancer cells while sparing normal tissue, and based on novel mechanisms of tumor cell destruction, but they also pose new challenges in their distribution and delivery [18, 19].
Traditional chemotherapeutics are generally small molecules, less than a nanometer in diameter, but molecular medicine includes antibodies (-lOnm in diameter), liposomes
(-50-100nm) and gene therapeutic vectors (hundreds of nm), which are orders of magnitude larger in size. Compared to small chemotherapeutic agents there is a significantly greater transport hindrance of large molecules or particles through normal and tumor tissue.
27

Transport of molecules through tumors differs from transport in other tissues or organs, due to pathophysiologic characteristics underlying tumor formation and growth[19]. The process of angiogenesis, or neovascularization of tumor tissue, serves to provide tumor cells with adequate nutrients to grow beyond their initial size as a precancerous lesion
[20], and is a critical step in the progression of the disease. However, this process is dysregulated in comparison to the neovascularization of wound sites or healthy regenerating tissue (such as liver): an excess of angiogenic stimuli is produced both by tumor cells and host cells within the tumor, resulting in tumor blood vessels that are chronically hyperpermeable, and unable to maintain their vascular pressure[21]. This leads to a uniform, elevated interstitial fluid pressure within tumors, resulting in negligible convection within the majority of the tumor, except for the tumor periphery[221. Hyperpermeable tumor vessels can be advantageous to the selective delivery of molecular medicine, as the pore size cutoff of the vasculature within tumors is much larger than that for normal tissues[23, 24]. The selective extravasation of liposomes or gene vectors in tumors is sometimes referred to as the EPR effect, named for enhanced permeability and retention of liposomal and other macromolecular formulations within tumors [25].
While it is true that tumor vessels are permeable, making tumor targeting possible in the sense that liposomal formulations or gene therapy vectors will extravasate preferentially within tumors, as opposed to normal host tissues, there are still a number of physiological barriers to the effective delivery of these formulations beyond the perivascular space. For one, many blood vessels in central areas of tumors are compressed or
28

collapsed due to cellular proliferation, resulting in insufficient vascular supply, and thus molecular delivery, to large portions of the necrotic tumor core [26]. Additionally, even for functional vessels within tumors, in vivo imaging of nanoparticle delivery has shown that there is limited transport of liposomes beyond the peri-vascular space [27], as seen in figure 2.1. To be effective, these therapeutics must have the potential to reach all the cells within the tumor mass. The tumor cells in the hypoxic environments distant from blood vessels are more resistant to both radiation therapy [28] and chemotherapy[29], and are thus the ultimate target for cancer therapeutic agents. Therefore, a major barrier to overcome for molecular medicine is movement through the interstitial space separating the extravasated particles from the numerous tumor cells lying distant from blood vessels
[30].
Fimire 7-1 -- (1) i If, llhcWwq 9 nm fluorescent nanoparticles that have extravasated within the tumor after they have been circulating for 24 hours in the mouse. From Yuan et al. [24] bar = 100
[lm
The interstitial space between tumor cells is filled with an extracellular matrix that comprises on average approximately 20% of the tumor [31]. The tumor extracellular
29

matrix derives from a host of factors, including: i) the original matrix of the host organ within which the tumor is growing, ii) matrix that is produced by the tumor cells themselves, and iii) matrix produced by host cells present within the tumor [32], including that produced by inflammatory cells such as macrophages [33]. In fact, many cancers, particularly breast cancer, pass through a stage termed desmoplasia over the course of their development, a fibrotic reaction of the host cells in response to the tumor's presence [34, 35]. This reaction, while initiated by the host cells as a part of the innate immune system, can often in fact fuel the progression of the disease due to the release of growth and angiogenic factors [36].
The interstitial matrix of tumors in the most general sense consists of collagens, elastin, hyaluronan, proteoglycans and their associated glycosaminoglycans (GAGs) and other glycoproteins. The collagens are a large family of proteins, all sharing in common a tripeptide repeat motif containing glycine, proline or hydroxyproline, and a third amino acid, which assemble into a triple helix. In normal and tumor tissues collagen type IV forms the basement membrane of blood vessels in association with laminin, fibronectin and other glycoproteins. In several tumor types he space between the blood vessels and tumor cells is occupied by fibrillar collagen, which is composed primarily of collagen type I molecules that associate into fibrils, which then associate into larger collagen fibers
[37]. Proteoglycans consist of a protein core to which a number of GAG chains are attached. GAGs are sugar molecules composed of disaccharide repeats that extend into either linear chains or branched structures; a large number of growth factors including the important angiogenic molecules vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (bFGF) are known to bind to certain GAG motifs[38]. The
30

unsulfated G(AG hyaluronan in particular is a high molecular weight linear chain of the disaccharide repeat GlcUA and GlcNAc, and has been shown historically to play a role in tissue transport, mainly associated with fluid flow[39, 40]. And finally glycoprotens, such as laminin and fibronectin, in general form structural links bridging the individual units within the extracellular matrix, or provide a substrate upon which cells can attach and migrate along. A schematic of the extracellular matrix is shown in figure 2.2.
Figure 2.2: Schematic of extracellular matrix composition
Of these components, the basement membrane is generally the first extracellular matrix encountered in the delivery of systemically acting therapies to tumors, but the high levels
31

of angiogenesis taking place within tumors render the basement membrane highly permeable to therapeutics, as mentioned earlier [24]. Netti et al. (2000) [1] quantitated the amounts of collagen, sulfated and unsulfated GAGs within 4 tumor types grown subcutaneously within mice, their results are shown in figure 2.3. Thus, part of my research has been to attempt to quantitate the presence and amount of these matrix molecules within the tumor interstitium, in order to better understand the role of the content and structure of the tumor extracellular matrix in influencing the transport properties of cancer therapeutics.
I
I
0
0= i rg
I
.
t.
I aI
UaVIL*17 1T1* M*HI'
Figure 2.3: Biochemical analysis of four tumor types A) The tissue content of sulfated
GAG (proteoglycan, light shading) and unsulfated GAG (hyaluronan, dark shading), expressed in terms of equivalent mass of hexuronic acid/g wet tissue. B) total collagen content (hydroxyproline) in the four different tumor types. No significant differences were found between the two carcinomas (MCaIV and LS 174T) or between U87 and
HSTS 26T. The collagen content of U87 and HSTS 26T is significantly higher than in the two carcinomas (P < 0.007; ANOVA). Bars, SE. From Netti et al. [1]
The goal of my research is to attempt to alter the matrix in order to improve the transport of therapeutics through it. A number of matrix modifying therapies exist currently, which can be classified into three general categories: endogenous matrix modifying
32

enzymes, directly acting exogenous enzymes, and indirectly acting anti-fibrotics. The endogenous matrix modifying enzymes mainly fall under the class of molecules called matrix metalloproteinases (MMPs), a large family of enzymes so named because they contain a coordinated metal ion in the catalytic site. These enzymes come in both soluble and membrane-bound forms, and are secreted initially in an inactive proenzyme state, becoming active after the cleavage of an inhibitory N terminal propeptide[41]. The
MMPs act on a variety of extracellular matrix molecules, generally by hydrolyzing at specific location on a molecule, or along the triple helix of collagen [42]. Examples of directly acting exogenous enzymes include bacterial collagenase, which cleaves the collagen triple helix at multiple sites [43], and hyaluronidase, which hydrolyzes the large polysaccharide chain into individual disaccharide units [44]. The indirectly acting antifibrotics act via a different mechanism: instead of directly acting on the target molecules of interest, they instead target the tumor or host cells, causing the upregulation of MMP activity in these cells, or an alteration in matrix synthesis. An example is the hormone relaxin, which is naturally secreted during pregnancy and in reproductive organs, and acts to stimulate uterine growth and cervical dilation through the fibroblasts present in those organs[45].
Each of these therapies have benefits and risks associated with them. Many
MMPs have been implicated in tumor progression and metastasis, allowing tumor cells to escape from their primary site of growth and invading adjacent tissues. The exogenous enzyme bacterial collagenase degrades all types of collagen, including basement membrane collagen, and as such can compromise the structure of the tumor vasculature and induce hemorrhage; it also can evoke an immune response, since it is a bacterial
33

protein. And finally not a lot is known about the actions of anti-fibrotics such as relaxin in tumors, or in fact in many tissues, due to relaxin's complicated physiology, and the only recent discovery of the relaxin receptors. Nevertheless, these matrix modifying treatments give us some tools with which to modify the tumor extracellular matrix in order to determine the resulting change in transport properties. My work on the development of second harmonic generation imaging of collagen in combination with intravital microscopy was done in order to be able to quantitate the influence of these matrix modifying therapies on the fibrillar collagen present within the tumor interstitium, in order to further the knowledge of relaxin's mechanism of action.
Collagen triple helix - tropocollagen molecule
A Bactenal collagenase iv.xY
!C Antifibrotic Mechanism:
Promotes connective tissue turnover
|
ConXn
,Xs
.0
;~~~~~~~~~
010-61W
B Matnx metaloprotease co"O .0 0
I
<A; > _,<:
POG-4AGlGW i,4 l_ al(i)
4,_
A_ A_ CO
W.-" so
__-0 A d .W.Om-Sr~IYL
Figure 2.4: MIatrix modifying treatments A) mechanism of action of bacterial collagenase - nonspecific degradation of the triple helix at all repeating glycine residues along the triple helix. B) mechanism of action of MMP- 1, interstitial collagenase: specific cleavage at a particular site (PQG - IAGQRGVV on cl . chain), resulting in 3/4
& 1/4 fragments of tropocollagen molecule, C) putative mechanism of action of the antifibrotic therapy relaxin, from the website of Connetics Corp.
I
34

Hyaluronan has most often been associated with transport hindrance within tissues [40], although this has generally been associated with flow within the interstitium of the peritoneal cavity. Diffusion has been measured within hyaluronan solutions [3] and collagen solutions [4]. Chary and Jain (1989)[46] were the first to adapt fluorescence recovery after photobleaching to measure diffusion within tumors. Berk et al (1993)[47] improved upon the fluorescence recovery after photobleaching technique via the addition of spatial Fourier analysis, which allowed the measurement of diffusion within scattering media, and used it to quantify diffusion within tumors. To investigate the role of extracellular matrix in tissue transport, Netti et al. (2000)[1] described the transport of
BSA and IgG within 4 different tumor types, and showed high collagen content tumors exhibited greater diffusional hindrance of IgG in vivo. The addition of bacterial collagenase resulted in an increase of the diffusion coefficient of IgG by 100% in tumors with high collagen content. This formed the basis for my thesis, and in chapter 3 I will describe the steps we took to determine the influence of collagen on the transport of macromolecules within tumors.
35

2.8 References
1.
2.
3.
4.
5.
6.
7.
8.
Netti, P.A., D.A. Berk, M.A. Swartz, A.J. Grodzinsky, and R.K. Jain, 2000. "Role of extracellular matrix assembly in interstitial transport in solid tumors." Cancer
Res, 60(9): p. 2497-503.
Pluen, A., et al., 2001. "Role of tumor-host interactions in interstitial diffusion of macromolecules: cranial vs. subcutaneous tumors." Proc Natl Acad Sci USA,
98(8): p. 4628-33.
De Smedt, S.C., A. Lauwers, J. Demeester, Y. Engelborghs, G. De Mey, and M.
Du, 1994. "Structural information on hyaluronic acid solutions as studied by proble diffusion experiments." Macromolecules, 27: p. 141-146.
Shenoy, V. and J. Rosenblatt, 1995. "Diffusion of macromolecules in colagen and hyaluronic acid, rigid-rod - flexible polymer, composite matrices."
Macromolecules, 28: p. 8751-58.
Ramanujan, S., A. Pluen, T.D. McKee, E.B. Brown, Y. Boucher, and R.K. Jain,
2002. "Diffusion and convection in collagen gels: implications for transport in the tumor interstitium. " Biophys J, 83(3): p. 1650-60.
Brown, E., T. McKee, E. diTomaso, A. Pluen, B. Seed, Y. Boucher, and R.K.
Jain, 2003. "Dynamic imaging of collagen and its modulation in tumors in vivo using second-harmonic generation." Nat Med, 9(6): p. 796-800.
Chiocca, E.A., 2002. "Oncolytic viruses." Nat Rev Cancer, 2(12): p. 938-50.
American Cancer Society, 2005, "American Cancer Society: Cancer Facts and
Figures 2005": Atlanta, GA.
Hanahan, D. and R.A. Weinberg, 2000. "The hallmarks of cancer." Cell, 100(1): 9. p. 57-70.
10. Bernier, J., E.J. Hall, and A. Giaccia, 2004. "Radiation oncology: a century of achievements." Nat Rev Cancer, 4(9): p. 737-47.
1 1. Fidler, I.J., 2002. "Critical determinants of metastasis. " Semin Cancer Biol, 12(2): p. 89-96.
12. Chabner, B.A. and T.G. Roberts, Jr., 2005. "Timeline: Chemotherapy and the war on cancer." Nat Rev Cancer, 5(1): p. 65-72.
13. Zhu, A.X. and C.G. Willett, 2005. "Combined modality treatment for rectal cancer." Semin Oncol, 32(1): p. 103-12.
14. Carmeliet, P. and R.K. Jain, 2000. "Angiogenesis in cancer and other diseases."
Nature, 407(6801): p. 249-57.
15. Carter, P., 2001. "Improving the efficacy of antibody-based cancer therapies." Nat
Rev Cancer, 1(2): p. 118-29.
16. Park, J.W., C.C. Benz, and F.J. Martin, 2004. "Future directions of liposome- and immunoliposome-based cancer therapeutics." Semin Oncol, 31(6 Suppl 13): p.
196-205.
17. Torchilin, V.P., 2005. "Recent advances with liposomes as pharmaceutical carriers." Nat Rev Drug Discov, 4(2): p. 145-60.
36

1 8. Jain, R.K., 1998. "The next frontier of molecular medicine: delivery of therapeutics." Nat Med, 4(6): p. 655-7.
19. Jain, R.K., 1994. "Barriers to drug delivery in solid tumors." Sci Am, 271(1): p.
58-65.
20. Folkman, J., 1992. "The role of angiogenesis in tumor growth." Semin Cancer
Biol, 3(2): p. 65-71.
21. Boucher, Y. and R.K. Jain, 1992. "Microvascular pressure is the principal driving force for interstitial hypertension in solid tumors: implications for vascular collapse." Cancer Res, 52(18): p. 5110-4.
22. Boucher, Y., L.T. Baxter, and R.K. Jain, 1990. "Interstitial pressure gradients in tissue-isolated and subcutaneous tumors: implications for therapy." Cancer Res,
50(15): p. 4478-84.
23. Hobbs, S.K., W.L. Monsky, F. Yuan, W.G. Roberts, L. Griffith, V.P. Torchilin, and R.K. Jain, 1998. "Regulation of transport pathways in tumor vessels: role of tumor type and microenvironment. " Proc Natl Acad Sci U S A, 95(8): p. 4607-12.
24. Yuan, F., M. Dellian, D. Fukumura, M. Leunig, D.A. Berk, V.P. Torchilin, and
R.K. .Jain, 1995. "Vascular permeability in a human tumor xenograft: molecular size dependence and cutoff size." Cancer Res, 55(17): p. 3752-6.
25. Maeda, H., J. Wu, T. Sawa, Y. Matsumura, and K. Hori, 2000. "Tumor vascular permeability and the EPR effect in macromolecular therapeutics: a review." J
Control Release, 65(1-2): p. 271-84.
26. Padera, T.P., B.R. Stoll, J.B. Tooredman, D. Capen, E. di Tomaso, and R.K. Jain,
2004. "Pathology: cancer cells compress intratumour vessels." Nature, 427(6976): p. 695.
27. Yuan, F., M. Leunig, S.K. Huang, D.A. Berk, D. Papahadjopoulos, and R.K. Jain,
1994. "Microvascular permeability and interstitial penetration of sterically stabilized (stealth) liposomes in a human tumor xenograft." Cancer Res, 54(13): p. 3352-6.
28. Vaupel, P., 2004. "Tumor microenvironmental physiology and its implications for radiation oncology." Semin Radiat Oncol, 14(3): p. 198-206.
29. Yu, J.L., B.L. Coomber, and R.S. Kerbel, 2002. "A paradigm for therapy-induced microenvironmental changes in solid tumors leading to drug resistance."
Differentiation, 70(9-10): p. 599-609.
30. Jain, R.K., 2001. "Delivery of molecular medicine to solid tumors: lessons from in vivo imaging of gene expression and function." J Control Release, 74(1-3): p.
7-25.
31. Jain, R.K., 1987. "Transport of molecules in the tumor interstitium: a review."
Cancer Res, 47(12): p. 3039-51.
32. Davies Cde, L., D.A. Berk, A. Pluen, and R.K. Jain, 2002. "Comparison of IgG diffusion and extracellular matrix composition in rhabdomyosarcomas grown in mice versus in vitro as spheroids reveals the role of host stromal cells." Br J
Cancer, 86(10): p. 1639-44.
33. Ben-Baruch, A., 2003. "Host microenvironment in breast cancer development: inflammatory cells, cytokines and chemokines in breast cancer progression: reciprocal tumor-microenvironment interactions." Breast Cancer Res, 5(1): p. 31-
6.
37

34. Desmouliere, A., C. Guyot, and G. Gabbiani, 2004. "The stroma reaction myofibroblast: a key player in the control of tumor cell behavior." Int J Dev Biol,
48(5-6): p. 509-17.
35. Walker, R.A., 2001. "The complexities of breast cancer desmoplasia." Breast
Cancer Res, 3(3): p. 143-5.
36. Dvorak, H.F., 1986. "Tumors: wounds that do not heal. Similarities between tumor stroma generation and wound healing." N Engl J Med, 315(26): p. 1650-9.
37. Kadler, K.E., D.F. Holmes, J.A. Trotter, and J.A. Chapman, 1996. "Collagen fibril formation." Biochem J, 316 ( Pt 1): p. 1-11.
38. lozzo, R.V. and J.D. San Antonio, 2001. "Heparan sulfate proteoglycans: heavy hitters in the angiogenesis arena." J Clin Invest, 108(3): p. 349-55.
39. Fraser, J.R., T.C. Laurent, and U.B. Laurent, 1997. "Hyaluronan: its nature, distribution, functions and turnover." J Intern Med, 242(1): p. 27-33.
40. Flessner, M.F., 2001. "The role of extracellular matrix in transperitoneal transport of water and solutes." Perit Dial Int, 21 Suppl 3: p. S24-9.
41. Birkedal-Hansen, H., 1995. "Proteolytic remodeling of extracellular matrix." Curr
Opin Cell Biol, 7(5): p. 728-35.
42. Gross, J. and Y. Nagai, 1965. "Specific degradation of the collagen molecule by tadpole collagenolytic enzyme. " Proc Natl Acad Sci U S A, 54(4): p. 1197-204.
43. Watanabe, K., 2004. "Collagenolytic proteases from bacteria." Appl Microbiol
Biotechnol, 63(5): p. 520-6.
44. Jedrzejas, M.J., 2000. "Structural and functional comparison of polysaccharidedegrading enzymes. " Crit Rev Biochem Mol Biol, 35(3): p. 221-51.
45. Sherwood, O.D., 2004. "Relaxin's physiological roles and other diverse actions."
Endocr Rev, 25(2): p. 205-34.
46. Chary, S.R. and R.K. Jain, 1989. "Direct measurement of interstitial convection and diffusion of albumin in normal and neoplastic tissues by fluorescence photobleaching." Proc Natl Acad Sci U S A, 86(14): p. 5385-9.
47. Berk, D.A., F. Yuan, M. Leunig, and R.K. Jain, 1993. "Fluorescence photobleaching with spatial Fourier analysis: Measurement of diffusion in lightscattering media." Biophysical Journal, 62: p. 2428-36.
38

3.1 Introduction
Blood-borne therapeutics must extravasate and penetrate the interstitial matrix to reach cancer cells in a tumor [1]. We recently have shown that tumor-host interactions regulate transvascular transport in tumors [2], but how they affect tumor interstitial transport is not known. Because of uniformly elevated interstitial fluid pressure in solid tumors, convection in the tumor interstitium is negligible [3], and drug delivery through the extracellular matrix (ECM) relies on passive diffusive transport [4]. Unfortunately, passive delivery becomes increasingly inefficient for larger particles. The success of novel cancer therapies that rely on large agents such as proteins, liposomes, nanoparticles, or gene vectors will hinge on their ability to penetrate the tumor interstitium [1, 5-7]. It is thus vital to identify the ECM constituents and characteristics that restrict diffusion and to determine how these are affected by tumor type and site.
Different ECM components, including collagen, glycosaminoglycans, and proteoglycans such as decorin, form a complex structured gel [8]. Resistance to interstitial flow has been strongly linked to glycosaminoglycans and especially
39

hyaluronan (HA) [8-10]. However, a recent in vivo study from our lab found an inverse correlation between collagen content of tumors and diffusion of IgG [ 11]. Furthermore, in vitro experiments found that diffusion of albumin is weakly hindered in HA gels [10] but significantly hindered in collagen gels [12]. Thus, we expect that tumor interstitial transport properties will depend on the volume, interaction, structure, and distribution of the matrix molecules and not simply on their overall levels [13]. Furthermore, because the bulk of the matrix in many tumors is produced by stromal cells [14, 15], we hypothesize that the diffusion of macromolecules will depend on tumor-host interactions.
Here we present analysis of the combined effect of the ECM composition, structure, and distribution and the role of tumor-host interaction on diffusion in the tumor interstitium. Using the fluorescence recovery after photobleaching (FRAP) technique [ 11,
16, 17], we measured the diffusion coefficients of proteins, dextrans, and liposomes in two different human tumor xenografts implanted either in the dorsal chamber (DC) or cranial window (CW) in mice. Diffusion coefficients were related to the distribution and relative levels of collagen type I, decorin, and HA as determined from stained tissue sections. Collagen organization was characterized by transmission electron microscopy.
We also estimated the effect of cellular geometry (tortuosity) on transport. The results provide critical data on the delivery of molecular medicine in solid tumors.
40

3.2 Materials and Methods
FITC-conjugated particles/molecules of various sizes were studied. In order of increasing size, these included lactalbumin and BSA (Molecular Probes), nonspecific IgG (Jackson
ImmunoResearch), nonspecific IgM (Sigma), FITC-dextran 2,000,000 MW (Sigma), and liposomes. IgM was purchased unlabeled and then conjugated to FITC by using the Fluo
EX-protein labeling kit (Molecular Probes). All other molecules were purchased in FITClabeled form. Liposomes (150 nm in diameter --- determined from the diffusion coefficients in solution by using Eq. 1) were prepared from dipamitoylphosphatidylcholine with 1 mol % of the fluorescent phospholipid carboxyfluoresceindioleoyl phosphatidylethanolamine [18].
Human glioblastoma (U87) and melanoma (Mu89) were implanted in two different sites in severe combined immunodeficient mice as described: (i) on the s.c. tissue of the skin
(DC) [19], and (ii) on the pial surface (CW) [2]. The pial surface approximates an orthotopic site for U87 tumors whereas skin is orthotopic for Mu89. Tumors can be visualized directly in these preparations. Animals were used for experiments 3-4 weeks after tumor implantation.
41

Injection of tracer. Small molecules (lactalbumin, BSA, and IgG) were injected i.v. via the tail vein., To ensure sufficient fluorescence and homogeneous distribution, molecules larger than IgG were introduced by direct intratumoral injection: 1 il of fluorescent solution was infused through thin micropipettes (25-30 jtm inner diameter) at constant pressure using a syringe pump (Harvard Apparatus) for 15-20 min. Diffusion was measured by FRAP 30 min after the end of micropipette injection. In preliminary studies, no statistical difference in the diffusion of IgG was found between i.v. or micropipette injections in the human sarcoma HSTS26T (high collagen content tumors; ref. [11]) implanted in DC (D = (8.85 + 0.8) 10
8 vs. (9.3 + 0.7) 10
-8 cm
2 s - 1 for micropipette and i.v.
injection, respectively).
FRAP measurements. The FRAP technique and method of analysis are described fully elsewhere [20]. In brief, redistribution of fluorescent molecules in bleached tissue yields the effective diffusion coefficient, independently of convection [16]. Unlike multiphoton
FRAP [21], FRAP measurements are restricted to less than 100 [tm from the tumor surface due to light scatter.
Hydrodynamic radius determination. The hydrodynamic radius of the fluorescent molecules, RH, was determined from the diffusion coefficient, Do, in PBS solution at
T=26°C (299K) using the Stokes-Einstein equation:
Do = kBT/(6nlrRH )
42

in which kB is Boltzmann constant, T is the temperature in Kelvin (K), and · is the viscosity of water (0.8705 cP at T=299K). Diffusion coefficients in solution were then scaled to T=37
0
C by correcting for the effect of temperature on the viscosity.
Extracellular space organization was characterized in tissue sections embedded in the hydrophilic resin, LR White (Ted Pella, Redding, CA). Tumors were fixed in 2.5% glutaraldehyde and 2.0% paraformaldehyde in PBS and embedded in the LR White resin
[22]. Toluidine blue stained sections were photographed using a color CCD camera mounted on a Nikon microscope.
Rabbit antiserum against type I collagen (LF-67) [23] and against human (LF-136) [24,
25] and mouse (LF-113) [26] decorin were generously provided by Larry Fisher
(National Institute of Dental Research, Bethesda, MD). LF-67, LF-136, and LF-1 13 were used at dilutions of 1:50, 1:500, and 1:1,000, respectively. Mouse anti-human collagen type IV (Dako) and rabbit anti-mouse collagen type IV (Chemicon) were used at dilutions 1:100 and 1:30, respectively. HA was detected with a HA biotinylated proteoglycan fragment (8 tpg/ml), a generous gift of Charles Underhill (Georgetown
University, Washington, DC).
Tumors were perfusion-fixed through the heart with 4% paraformaldehyde in
PBS. The tissue was infiltrated with sucrose and embedded in OCT. For immunostaining, sections were blocked with rabbit or goat serum, incubated with the antibody overnight at
43

4°C and then with the appropriate secondary antibodies conjugated to Cy-5 (Jackson
ImmunoResearch Laboratories). For HA staining, the sections were stained for 1 h with the biotinylated proteoglycan fragment diluted in 10% calf serum and incubated with
Texas red-conjugated streptavidin (Jackson ImmunoResearch Laboratories). The cell nuclei were stained with the Alexis nuclear stain (Molecular Probes). Sections were photographed with a Leica TCS-NT4D confocal microscope. For quantification of the fraction of tissue occupied by collagen type I staining, photographs were taken with a custom-built two photon microscope based on a MRC 600 platform (Bio-Rad). Using a constant 10 mW of 720-nm light through a 0.9 numerical aperture water immersion lens, we generated image stacks of the histological slices, with 10-20 images per stack. A maximum intensity projection was performed on the image stacks to form a single image of the section, thereby ensuring that each pixel value represents the best colocalization of the excitation volume with the slice. Using a series of threshold pixel values we automatically segmented images into regions corresponding to 1) tissue section versus not, 2) cell nuclei versus extracellular space, 3) nonspecific staining, and 4) specific staining for fibrillar collagen. The fraction of the pixels in a 50 x 100 ILm window
(oriented perpendicular to the tumor surface) that were stained for collagen was quantified using this technique. The average pixel value of the collagen pixels was calculated as an indicator of collagen type I staining.
Organization of collagen bundles and interfibrillar spacing were characterized by electron microscopy. Tumors were fixed by immersion in 2.5% glutaraldehyde and 2.0%
44

paraformaldehyde in PBS for 4-6 h. Small tumor pieces were washed overnight in PBS, dehydrated in ethanol, fixed in 1% osmium, and embedded in Polybed 812. Thin sections were stained with uranyl acetate and lead citrate and examined with a Phillips CM10 transmission electron microscope (Phillips Electronic Instruments, Mahwah, NJ) operating at 80 kV.
3.3 Results
Figure 3.1.a presents diffusion coefficients obtained in Mu89 and U87 in both implantation sites. In the two tumors, the diffusion of larger molecules is significantly slower than that of smaller molecules. The decrease in diffusion with particle size in tumors is even greater than one would predict from pure solution data, due to the presence of cellular obstacles and matrix molecules. To examine these contributions, we introduce the concept of tortuosity.
45

a
'0)
C
10) b nt-5 lU -
II 104 aF 10-
7
0
._4
Q 10-' o1,,4o a 109
0.1 1 10 100
Experimental hydrodynamic radius, RH, nm
W 10-9
0-lo
1laQ
0.1 1 10 100
Exp. hydrodynamic radius, R , nm
Figure 3.1. Effective diffusion coefficients of tracers in vivo (a) Effective diffusion coefficients in PBS solution were measured at T = 26
0
C and scaled to 37C according to the Stokes-Einstein equation. Diffusion coefficients were measured in DC (filled symbols and dotted lines) and CW (open symbols and continuous line) tumors. (b) Interstitial diffusion coefficients in tumors (Dint = g
2 Deff) as a function of hydrodynamic radius, RH, using the experimentally obtained value g =
1.19. The diffusion coefficients in solution
(Io) are pictured (black-triangle) to illustrate the ECM influence on retardation.
The increase in path length induced by physical obstacles and ECS connectivity is described by the tortuosity. The effective diffusion measured in tissues [27] is related to the tortuosity by Dejj= (1/
2
)Do [28]. Geometric effects imposed by the organization of cells are likely to be the major hindrance to long-range diffusion of small molecules.
Frictional effects assume greater importance as the size of diffusing particles increases to become comparable to the dimensions of channels through which they move. On this basis, we separate tortuosity into viscous (v) and geometric (-g) contributions according to =gv [291 so that:
(Defj/Dint)
=
(I/g 2 ) (1]/vj) Do
46

where Dint is the interstitial diffusion coefficient and Do the diffusion coefficient in solution. The ratio Defj/Dint=l/g
2 measures hindrance due to cellular obstacles. The ratio
2 measures hindrance within the ECM. The geometric tortuosity may be estimated using a sufficiently small molecule for which viscous hindrance is negligible
(-v=1) so that Defg(1/ g
2
)Do. From diffusion measurements of fluorescein (RH= 0.4 nm) in U87 DC, the geometric tortuosity was estimated at g=1. 19. Figure 3.1.b presents the interstitial diffusion coefficients in tumors (Dint= g
2
Deff) as a function of the hydrodynamic radius, illustrating that the reduction of diffusion coefficient with particle size is greater in the ECM than in solution.
No statistical difference was observed in the diffusion coefficients of small molecules such as lactalbumin and BSA between the two tumor types and sites of implantation.
However, the diffusion coefficients of larger molecules (particles equal in size and larger than IgG: RH > 5.5 nm) were significantly decreased (p<0.05) in DC as compared to CW tumors (Figure 3.1). Figure 3.1.a illustrates the distinction between a "fast diffusion group" (CW tumors) and a "slow diffusion group" (DC tumors). The difference in diffusion increases with particle size and is striking for molecules such as dextran
2,000,000 MW. Figure 3.1.a also shows diffusion coefficients for liposomes (RH = 75 nm) in CW tumors. Diffusion coefficients of liposomes in DC tumors could not be assessed by FRAP due to prohibitively slow diffusion and inhomogeneous distribution of the particles.
47

Significant differences in cellular content and ECM organization were found between DC and CW tumors. Typically, DC tumors were separated from the glass cover slip by a fibrous capsule composed of several layers of fibroblast-like cells, separated by ECM
(Figure 3.2.a & b). The ECM of the capsule was continuous with that of the underlying tumor cells. In Mu89, cellular nodules were surrounded by a thin layer of ECM and stromal cells, whereas in U87 single tumor cells or groups of tumor cells were separated by larger ECM spaces (Figure 3.2.a & b). In contrast to DC tumors, at the outer edge of
CW tumors, only one layer of fibroblast-like cells was observed in contact with the underlying tumor cells, which were separated from each other by narrow ECM spaces
(Figure 3.2.c & d).
48

Figure 3.2. Extracellular space in vivo Light microscopy (LR White sections) of the peripheral region of DC and CW tumors. The capsule of U87 (a) and Mu89 (b) DC tumors is composed of several layers of fibroblast-like cells separated by ECM. Note the large intercellular spaces in U87 and the narrow space that separates two cellular nodules in Mu89. The connective tissue at the edge of U87 (c) and Mu89 (d) in the CW is composed of one fibroblast cell layer; the tumor cells are separated by narrow intercellular spaces. C, capsule; T, tumor; black arrows, ECM; white arrows, fibroblastlike cells. (Bar = 10 jm.)
To compare the influence of tumor implantation site on the ECM, the distribution and staining intensity of collagen types I and IV, decorin and HA were characterized.
Collagen type I staining was abundant in DC tumors, approaching levels found in normal skin. In these tumors, collagen type I fibers were identified between the layers of fibroblast-like cells (Figure 3.4.a). In central regions of Mu89, tumor cell clusters were surrounded by thin layers of type I collagen, whereas in U87, tumor cell clusters or single cells were separated by wider spaces occupied by type I collagen. In comparison to DC
49

tumors the staining occupied a smaller area in CW tumors (Figure 3.4.a & b).
Quantitative image analysis within the superficial 100 pm of Mu89 tumors revealed 36 ±
11% tissue area stained for collagen I in the DC, as opposed to 12 i 5% in the CW
(Figure 3.3). The collagen type I staining also occupied a greater proportion of the ECM in DC than in CW tumors (Fig 3.4.a & b). In CW tumors collagen type I was predominantly localized at the tumor edge with scattered staining between tumor cells
(Figure 3.4.b). As expected, staining for collagen type IV was associated with tumor
_ vessels in both sites (data not shown).
4i 1U U/o
C
90%
0
8
0%
E 70%
0 u 60)%
U 50%
) 40% o 30%
C 20%
0E 1%
E o0%
CW 1 CW 2 CW 3
Tissue section
DC 1 DC 2 DC 3
Figure 3.3. Quantification of the fractional tissue area stained specifically for
fibrillar collagen I As described in the methods, we used a series of threshold pixel values to automatically segment images into regions corresponding to 1) tissue section versus not
('holes' in the tissue section), 2) cell nuclei versus extracellular space, 3) nonspecific staining
(determined using negative controls), and 4) specific staining for fibrillar collagen. The fraction of the pixels in a 50 x 100 jim window (oriented perpendicular to the tumor surface) that were stained for collagen was quantified using this technique. Each bar corresponds to the results for one tumor, from a total of between 9 and 18 images per tumor. Bar = SE.
50

MuS9 Cranial Window
Mu89 Cranial Window
Collagen type I
Mouse decorin
HA l
Figure 3.4. Extracellular matrix composition Immunostaining for collagen type I (a and b) and decorin (c and d), and labeling for HA (e and f) in DC (a, c, and e) and CW (b, d, and f) tumors. Collagen type I occupies a greater area of the periphery in DC than in
CW tumors. In both DC and CW tumors the decorin staining is restricted to the periphery of the tumor. HA staining is intense in the center of Mu89 in the DC, whereas in the periphery the staining is weak. (Bar = 50 pm.)
The collagen organization was characterized by electron microscopy. Fibrillar collagen was abundant in the capsule of DC tumors. Bundles of aligned and compact fibrils (interfibrillar spacing 20-42 nm) were found adjacent to bundles that were poorly
51

organized with larger interfibrillar spaces (75-130 nm) (Figure 3.5.a & b). In the center of
U87 especially, fibrillar collagen was less abundant and poorly organized. This finding, coupled with the extensive collagen type I staining in the center of U87, suggests that the deposited collagen is poorly assembled. In CW tumors, collagen fibrils had no specific organization and appeared as isolated fibrils.
Figure 3.5. Electron microscopy of the organization of collagen fibrils in the capsule of
U87 tumors in the DC. (a) The longitudinally oriented fibrils are parallel to one another with an interfibrillar spacing that varies from 20 to 42 nm. (b) The fibrils are poorly organized. The interfibrillar spacing varies between 75 and 130 nm. (Bar = 200 nm.)
Because decorin participates in the organization of fibrillar collagen, we characterized its distribution. Decorin was present between fibroblast-like cells in the capsule of DC tumors. However, in contrast to type I collagen, decorin immunostaining was not detected in the extracellular space separating tumor cells (Figure 3.4.c). In CW tumors, decorin staining was almost exclusively restricted to the tumor edge (Figure 3.4.d).
52

In comparison to the tumor center, HA staining was absent or significantly reduced in the capsule of DC tumors (Figure 3.4.e). In the center of Mu89 especially, tumor nodules were separated by intense HA staining. The staining intensity for HA was greater in skin than in DC tumors. In U87 and Mu89 in the CW, HA staining was distributed diffusely throughout the tumor. No obvious difference in the relative levels of HA was detected between tumor implantation sites.
The origin (tumor vs. host) of ECM components in the human tumor xenografts implanted in mice was determined by immunostaining. Staining of the ECM by antibodies against human decorin and collagen type IV was significantly weaker than for corresponding murine antibodies
(data not shown), indicating that the ECM observed was primarily of host origin.
3.4 Discussion
Our diffusion measurements provide necessary data for prediction of transport properties of therapeutic molecules over a wide range of molecular weights. Although no significant difference in diffusion coefficients was observed for small proteins (lactalbumin, albumin) between implantation sites, diffusion of larger molecules (IgM and dextran
2,000,000 MW) was 5 to 10-fold faster in CW tumors than in DC tumors.
53

Depending on tumor site, tumors fell into slow-diffusing (DC) and fast-diffusing
(CW) groups, characterized by high and low collagen type I levels, respectively. The hindrance to diffusion of dextran 2,000,000 MW (R = 19 nm) in DC tumors was comparable to that of liposomes (RH = 75 nm) in CW tumors (Figure 3.1.a). In DC tumors, diffusion of the same liposomes was prohibitively slow for measurement. A rough estimate based on extrapolation of the measured diffusion coefficients suggests that the diffusion of liposomes in DC tumors would be 1-2 orders of magnitude slower than in CW tumors. Thus, passive delivery of liposomes might be more feasible in lowcollagen brain tumors than in high collagen tumors. Our results emphasize that the delivery of larger particles will be highly influenced by the tumor site and possibly by other factors that influence ECM composition/structure
We estimated the geometric tortuosity in U87 DC tumors as g=1.19+0.10. Our results compare well with previous Monte-Carlo simulations, which predicted a tortuosity of 1.4
for 3-dimensional radial diffusion through an array of evenly spaced cells [29, 30]. The geometric tortuosity could vary with cellular arrangement and ECS connectivity.
Complex cellular arrangements may differentially affect the transport of large vs. small particles, restricting large particles to wider intercellular paths. The matrix, its
54

distribution and organization further compound the hindrance via the viscous tortuosity which, as shown in Figure 3.1.b, increases significantly for larger molecules and at higher collagen type I levels. Although the true tortuosity of long-range motion may indeed increase with particle size, the most likely explanation for the increased hindrance for larger particles is the increased viscous drag from solid obstacles (cells, matrix fibers) as the size of the diffusing particles becomes significant compared to intercellular or interfibrillar spacing [29, 31].
Role of collagen. Expanding on the results of Netti et al. (2000) [11], we find that collagen type I and its organization into fibrils have a significant role in limiting the diffusion of large molecules (e.g. IgG, IgM and dextran 2,000,000 MW). Fibrillar collagen occupied a greater portion of the ECM in DC than in CW tumors. The narrow spacing (20 - 40 nm) between collagen fibrils will exclude or hinder (frictional interaction, steric hindrance) the migration of larger particles. The tortuous paths around compact collagen bundles or within loose bundles (interfibrillar spacing = 75 - 130 nm) will also hinder the diffusion of large molecules. Interestingly, the 5 - 10 fold difference in diffusion between CW and DC tumors was found for molecules with diameters approaching the interfibrillar spacing.
Role of proteoglycans. The alignment and spacing of collagen fibrils is modulated by proteoglycans. The protein core of decorin binds to fibrils, and the dermatan/chondroitin
55

sulfate side chains form complexes that bridge the interfibrillar space at intervals of 60-
65 nm [32]. Decorin knockout mice exhibit wider interfibrillar spaces in the skin, and inhibition of decorin synthesis by B-D xyloside induces large separations in the fibrillar collagen network of the corneal stroma [33, 34]. Thus, the wider interfibrillar spaces in the center of U87 in the DC could be due to the reduced expression of decorin. However, the presence of occasional compact collagen bundles in the center of U87 and the tightly organized fibrillar collagen in the center of Mu89 where decorin expression is reduced suggest that other proteoglycans, possibly lumican [35, 36], may participate in the organization of fibrillar collagen in these tumors. It remains to be established whether the interaction between proteoglycans and fibrillar collagen limits the diffusion of macromolecules in tumors.
Role of' hyaluronan. The low levels or absence of HA staining in the capsule of DC tumors suggest that HA was not a contributor to transport hindrance in these tumors. In the tumor center, the higher levels of HA could potentially influence interstitial transport.
Several studies have clearly demonstrated that HA impedes fluid flow in tissues [37, 38], whereas the effect of HA on the diffusion of macromolecules in normal or tumor tissues is yet to be determined. The degradation of HA in normal tissues with hyaluronidase either does not modify or even decreases the diffusion of albumin [38, 39]. Indeed, we have observed similar effects of hyaluronidase administration on the diffusion of IgG in tumors (Figure 3.6, unpublished data). Based on these results, it is possible that the swelling potential of intact HA increases the pore size between ECM molecules and thus actually facilitates diffusion.
56

Role of tumor site and tumor-host interactions. Differences in the levels of collagen type
I and decorin between DC and CW tumors reflect the greater recruitment of stromal cells
(e.g. fibroblasts) in DC tumors. The greater accumulation of collagen type I and decorin in DC tumors was associated with a higher density of stromal cells. In general, stromal cells, and not neoplastic cells, synthesize these molecules in carcinomas [15, 40].
Immunostaining also showed that decorin and collagen IV in Mu89 and U87 were produced by host (murine) cells. In contrast, previous studies have shown that HA is produced by neoplastic cells as well as by stromal cells [41-43]. In vitro, paracrine interactions and direct cell-cell contact between tumor cells and fibroblasts can increase the fibroblast synthesis of collagen type I, HA, and decorin [40, 43, 44].
57

3 1 n-7
5. I .f
U)
04-
C 1.5107
e)
0
I
(5 5107 ol_
1 2 3 4 5
6 7
Figure 3.6. The effect of hyaluronidase treatment on the diffusion coefficient of IgG
within HSTS tumors Effective diffusion coefficients of IgG administered systemically
(IgG) measured by FRAP in HSTS tumors, compared to measurements taken within the same tumor 24 hours after hyaluronidase administration (IgG+HAse). Effective diffusion coefficients show either no change or a slight decrease following hyaluronidase treatment. (unpublished data)
The present study provides critical data on the diffusive transport of particles in tumors for a wide range of particle sizes. Tumors studied fell into slow vs. fast diffusion groups, corresponding to high vs. low collagen type I content respectively, supporting a central role for fibrillar collagen in determining interstitial hindrance. The results demonstrate that diffusion of large molecules (IgG. IgM, dextran 2M and liposomes) is much faster in
CW than in DC tumors. The greater hindrance to diffusion in DC tumors was associated
58

with a higher density of host stromal cells, which synthesize and organize collagen type I.
These results also point to the necessity of site-specific drug carriers to improve drug delivery. Finally, our results underscore that efficient gene therapy will require a better integration of drug design and in vivo experimentation.
59

3.6 References
1. Jain, R.K., 1998. "The next frontier of molecular medicine: delivery of therapeutics." Nat Med, 4(6): p. 655-7.
2.
3.
Hobbs, S.K., W.L. Monsky, F. Yuan, W.G. Roberts, L. Griffith, V.P. Torchilin, and R.K. Jain, 1998. "Regulation of transport pathways in tumor vessels: role of tumor type and microenvironment." Proc Natl Acad Sci USA, 95(8): p. 4607-12.
Boucher, Y., L.T. Baxter, and R.K. Jain, 1990. "Interstitial pressure gradients in tissue-isolated and subcutaneous tumors: implications for therapy." Cancer Res,
50(15): p. 4478-84.
4. Netti., P.A., et al., 1999. "Enhancement of fluid filtration across tumor vessels: implication for delivery of macromolecules." Proc Natl Acad Sci U S A, 96(6): p.
3137--42.
5.
6.
7.
8.
9.
Yuan, F., M. Dellian, D. Fukumura, M. Leunig, D.A. Berk, V.P. Torchilin, and
R.K. Jain, 1995. "Vascular permeability in a human tumor xenograft: molecular size dependence and cutoff size." Cancer Res, 55(17): p. 3752-6.
Jacobs, A., X.O. Breakefield, and C. Fraefel, 1999. "HSV-1-based vectors for gene therapy of neurological diseases and brain tumors: part II. Vector systems and applications." Neoplasia, 1(5): p. 402-16.
Weyerbrock, A. and E.H. Oldfield, 1999. "Gene transfer technologies for malignant gliomas." Curr Opin Oncol, 11(3): p. 168-73.
Gribbon, P.M., A. Maroudas, K.H. Parker, and C.P. Winlove, Water and solute
determinants, in Connective tissue biology . integration and reductionism, R.K.
Reed and K. Rubin, Editors. 1998, Portland: London; Miami. p. 95-124.
Levick, J.R., 1987. "Flow through interstitium and other fibrous matrices." Q J
Exp Physiol, 72(4): p. 409-37.
10. Comper, W.D. and T.C. Laurent, 1978. "Physiological function of connective tissue polysaccharides." Physiol Rev, 58(1): p. 255-315.
11. Netti.. P.A., D.A. Berk, M.A. Swartz, A.J. Grodzinsky, and R.K. Jain, 2000. "Role of extracellular matrix assembly in interstitial transport in solid tumors." Cancer
Res, 60(9): p. 2497-503.
12. Shenoy, V. and J. Rosenblatt, 1995. "Diffusion of macromolecules in collagen and hyaluronic acid, rigid-rod-flexible polymer, composite matrices."
Macromolecules, 28(26): p. 8751-8758.
13. Ogston, A.G., B.N. Preston, and J.D. Wells, 1973. "On the Transport of Compact
Particles Through Solutions of Chain-Polymers." Proc. R. Soc. Lond. A.,
333(1594): p. 297-316.
14. Toole, B.P., C. Biswas, and J. Gross, 1979. "Hyaluronate and invasiveness of the rabbit V2 carcinoma." Proc Natl Acad Sci USA, 76(12): p. 6299-303.
15. Brown, L.F., et al., 1999. "Vascular stroma formation in carcinoma in situ, invasive carcinoma, and metastatic carcinoma of the breast." Clin Cancer Res,
5(5): p. 1041-56.
16. Berk, D.A., F. Yuan, M. Leunig, and R.K. Jain, 1997. "Direct in vivo measurement of targeted binding in a human tumor xenograft." Proc Natl Acad
Sci USA, 94(5): p. 1785-90.
60

17. Chary, S.R. and R.K. Jain, 1989. "Direct measurement of interstitial convection and diffusion of albumin in normal and neoplastic tissues by fluorescence photobleaching." Proc Natl Acad Sci USA, 86(14): p. 5385-9.
18. Szoka, F., Jr. and D. Papahadjopoulos, 1980. "Comparative properties and methods of preparation of lipid vesicles (liposomes)." Annu Rev Biophys Bioeng,
9: p. 467-508.
19. Leunig, M., A.E. Goetz, M. Dellian, G. Zetterer, F. Gamarra, R.K. Jain, and K.
Messmer, 1992. "Interstitial fluid pressure in solid tumors following hyperthermia: possible correlation with therapeutic response." Cancer Res, 52(2): p. 487-90.
20. Berk., D.A., F. Yuan, M. Leunig, and R.K. Jain, 1993. "Fluorescence photobleaching with spatial Fourier analysis: measurement of diffusion in lightscattering media." Biophys J, 65(6): p. 2428-36.
21. Brown, E.B., E.S. Wu, W. Zipfel, and W.W. Webb, 1999. "Measurement of molecular diffusion in solution by multiphoton fluorescence photobleaching recovery." Biophys J, 77(5): p. 2837-49.
22. Newman, G.R., 1987. "Use and abuse of LR White." Histochem J, 19(2): p. 118-
20.
23. Bernstein, E.F., L.W. Fisher, K. Li, R.G. LeBaron, E.M. Tan, and J. Uitto, 1995.
"Differential expression of the versican and decorin genes in photoaged and sunprotected skin. Comparison by immunohistochemical and northern analyses." Lab
Invest, 72(6): p. 662-9.
24. Bianco, P., L.W. Fisher, M.F. Young, J.D. Termine, and P.G. Robey, 1990.
"Expression and localization of the two small proteoglycans biglycan and decorin in developing human skeletal and non-skeletal tissues." JHistochem Cytochem,
38(11): p. 1549-63.
25. Fisher, L.W., J.D. Termine, and M.F. Young, 1989. "Deduced protein sequence of bone small proteoglycan I (biglycan) shows homology with proteoglycan II
(decorin) and several nonconnective tissue proteins in a variety of species." J Biol
Chem, 264(8): p. 4571-6.
26. Fisher, L.W., J.T. Stubbs, 3rd, and M.F. Young, 1995. "Antisera and cDNA probes to human and certain animal model bone matrix noncollagenous proteins."
Acta Orthop Scand Suppl, 266: p.
6 1
-
5
.
27. el-Kareh, A.W., S.L. Braunstein, and T.W. Secomb, 1993. "Effect of cell arrangement and interstitial volume fraction on the diffusivity of monoclonal antibodies in tissue." Biophys J, 64(5): p. 1638-46.
28. Nicholson, C. and J.M. Phillips, 1981. "Ion diffusion modified by tortuosity and volume fraction in the extracellular microenvironment of the rat cerebellum." J
Physiol (Lond), 321: p. 225-57.
29. Rusakov, D.A. and D.M. Kullmann, 1998. "Geometric and viscous components of the tortuosity of the extracellular space in the brain." Proc Natl Acad Sci U SA,
95(15): p. 8975-80.
30. Chen, K.C. and C. Nicholson, 2000. "Changes in brain cell shape create residual extracellular space volume and explain tortuosity behavior during osmotic challenge." Proc Natl Acad Sci U SA, 97(15): p. 8306-11.
61

31. Phillips, R.J., 2000. "A hydrodynamic model for hindered diffusion of proteins and micelles in hydrogels." Biophys J, 79(6): p. 3350-3.
32. Scott, J.E., K.M. Dyne, A.M. Thomlinson, M. Ritchie, J. Bateman, G. Cetta, and
M. Valli, 1998. "Human cells unable to express decoron produced disorganized extracellular matrix lacking "shape modules" (interfibrillar proteoglycan bridges)." Exp Cell Res, 243(1): p. 59-66.
33. Hahn, R.A. and D.E. Birk, 1992. "beta-D xyloside alters dermatan sulfate proteoglycan synthesis and the organization of the developing avian corneal stroma." Development, 115(2): p. 383-93.
34. Danielson, K.G., H. Baribault, D.F. Holmes, H. Graham, K.E. Kadler, and R.V.
Iozzo, 1997. "Targeted disruption of decorin leads to abnormal collagen fibril morphology and skin fragility." J Cell Biol, 136(3): p. 729-43.
35. Chakravarti, S., T. Magnuson, J.H. Lass, K.J. Jepsen, C. LaMantia, and H.
Carroll, 1998. "Lumican regulates collagen fibril assembly: skin fragility and corneal opacity in the absence of lumican." J Cell Biol, 141(5): p. 1277-86.
36. Leygue, E., et al., 2000. "Lumican and decorin are differentially expressed in human breast carcinoma." JPathol, 192(3): p. 313-320.
37. Bert, J.L. and R.H. Pearce, The interstitium and microvascular exchange, in
Handbook of Physiology: The Cardiovascular System., R.M. Berne and N.
Sperelakis, Editors. 1990, American Physiological Society: Bethesda, Md. p. 521-
547.
38. Parameswaran, S., L.V. Brown, G.S. Ibbott, and S.J. Lai-Fook, 1999. "Hydraulic conductivity, albumin reflection and diffusion coefficients of pig mediastinal pleura." Microvasc Res, 58(2): p. 114-27.
39. Qiu, X.L., L.V. Brown, S. Parameswaran, V.W. Marek, G.S. Ibbott, and S.J. Lai-
Fook, 1999. "Effect of hyaluronidase on albumin diffusion in lung interstitium."
Lung, 177(5): p. 273-88.
40. lozzo., R.V. and I. Cohen, 1994. "Altered proteoglycan gene expression and the tumor stroma." Exs, 70: p. 199-214.
41. Kimata, K., M. Takeda, S. Suzuki, J.P. Pennypacker, H.J. Barrach, and K.S.
Brown, 1983. "Presence of link protein in cartilage from cmd/cmd (cartilage matrix deficiency) mice." Arch Biochem Biophys, 226(2): p. 506-16.
42. Turley, E.A., C.A. Erickson, and R.P. Tucker, 1985. "The retention and ultrastructural appearances of various extracellular matrix molecules incorporated into three-dimensional hydrated collagen lattices." Dev Biol, 109(2): p. 347-69.
43. Knudson, W., 1996. "Tumor-associated hyaluronan. Providing an extracellular matrix that facilitates invasion." Am JPathol, 148(6): p. 1721-6.
44. Noel, A., C. Munaut, B. Nusgens, J.M. Foidart, and C.M. Lapiere, 1992. "The stimulation of fibroblasts' collagen synthesis by neoplastic cells is modulated by the extracellular matrix." Matrix, 12(3): p. 213-20.
62

63

4.1 Introduction
Optimal therapy of tumors requires delivery of sufficient amounts of therapeutic agents to the target cancer cells. Thus, the agent must penetrate the tumor interstitial matrix (IM), a complex assembly of collagen, glycosaminoglycans, and proteoglycans [1]. Convection through the tumor IM is poor due to interstitial hypertension, leaving diffusion as the major mode of drug transport. As anti-cancer therapy focuses increasingly on larger therapeutics such as liposomes, which are typically at least 90 nm in diameter [2, 3]and gene vectors, which range in diameter from 20 to 300 nm [4], diffusion within the tumor IM becomes a greater barrier to delivery [5-7].
Glycosaminoglycans (GAGs), and particularly hyaluronan (HA), are believed to play a primary but not exclusive role in regulating fluid movement in the IM [8, 9]. However, diffusion of large molecules in tumors has been correlated to collagen content and organization, but not to HA content [10, 11]. These in vivo studies correlated matrix composition to diffusive hindrance, but the biological complexity prohibited detailed analysis of the mechanisms of transport hindrance within the tumor IM. For example, even within a
64

given tumor, Pluen et al. (2001) [11] found varying degrees of collagen organization and heterogeneous distribution of different matrix molecules.
To overcome these problems, we measured diffusion and hydraulic conductivity in pure collagen type I gels and compared these results directly with previously published results for tumors of comparable collagen concentration. Furthermore, we compared the structure of the gels with that seen in tumors. To investigate the role of collagen structure, we compared diffusion in collagen gels and solutions of the same concentrations. The findings presented here are important to the development of improved drug delivery strategies [12] and to pharmaceutical applications of collagen matrices, including the design of tissue substitutes and controlled release devices [13].
4.2 Materials and Methods
Preparation of Collagen Gels. Vitrogen 100 collagen type I solution was purchased from
Collagen Corp. (Cohesion Technologies, Palo Alto, CA) at a concentration of approximately
3 mg/ml. The pH and ionic strength were adjusted by addition of NaOH (pH 7.4) and 10X phosphate buffered saline (PBS). To concentrate the solution, the collagen was ultracentrifuged (Beckman LC-300) at 10°C for 26-48 hours for preparation of 10-45 mg/ml gels. Supernatant was extracted and pellets were maintained at 4
0
C. Collagen concentration in the pellet was determined from the difference between pre-centrifugation and supernatant collagen content as determined by UV spectrophotometery. Pellet concentration was adjusted by dilution with PBS. The polymerization of highly concentrated collagen solutions leads to the formation of fibers and filaments. To obtain a collagen gel formed predominantly
65

of fibers, 30 ml of neutralized collagen type I (0.4 mg/ml) was polymerized at 32 0
C for 48 h.
The collagen was centrifuged at 11,000 or 25,000 RPM for 12 or 30 min, respectively. The collagen gel was collected on a plastic coverslip that was attached to the bottom of the centrifuge tube. To determine the organization of the fibers and the dimensions of the gel, second harmonic images of the collagen were obtained with a multiphoton microscope [14].
The collagen concentration estimates were based on the unpolymerized collagen volume and the final gel volume after centrifugation.
For fluorescence recovery after photobleaching (FRAP) experiments at low collagen concentration (2.4 mg/ml), capillary tubes were partially filled with unconcentrated collagen solution and kept for 2 hours in a 37°C incubator. After gelation, an aqueous solution of tracer molecules (2 mg/ml) was added to the capillary, which was then sealed and maintained overnight at 37C to allow tracer penetration of the gel. For FRAP experiments with more concentrated gels, the appropriate tracer molecule solution was added during adjustment of the pellet concentration. The samples were then prepared on concave microscope slides under coverslips, and sealed with silicone grease. Samples for permeability and visualization experiments were prepared in Transwell (24 mm diameter; for 0.24% gels) or Snapwell (12 mm diameter, for [I1% gels) membrane-bottomed cell culture chambers (Corning Costar
Corp., Cambridge, MA) and maintained in a 37°C incubator for at least 1 hour to allow gelation. PBS was then added to chambers to maintain hydration.
Preparation of Hyaluronan Solutions. Hyaluronic acid sodium salt isolated from rooster comb (Sigma Chemical Co., St. Louis, MO) was dissolved by slow addition of 1X PBS (pH
7.4) for a final concentration of 4 mg/ml. Fluorescent markers at a concentration of 2 mg/ml
66

were added to the solution. The solution was stirred at 4
0
C for 10 hours and subsequently stored at 4
0
C overnight. Samples were prepared and sealed in capillary tubes as described above for low concentration collagen gels.
Measurement of Diffusion Coefficients. Diffusion coefficients were measured using the
FRAP with spatial Fourier analysis technique described previously [ 15, 16]. Briefly, samples permeated with FITC-conjugated tracer molecules were placed on a microscope stage. Each sample was subjected to brief localized 488 nm irradiation from a krypton-argon laser, resulting in bleaching of fluorescence in the irradiated spot (radius -20 tm). Images were recorded by CCD camera as the bleached spot recovered fluorescence. The diffusion coefficient was extracted from the exponential time decay of the spatial Fourier transform of fluorescence intensity. The diffusion coefficient for a given sample represents the average of
5 - 10 FRAP measurements in the sample. When not specified otherwise, three samples were used to determine the diffusion coefficient of each molecule-gel combination. Tracer molecules including lactalbumin (LA), bovine serum albumin (BSA), and dextrans of molecular weights 4,400 - 2,000,000 were purchased in FITC-labeled form (Sigma).
Nonspecific IgG was purchased unlabeled (Sigma) and subsequently conjugated to FITC using the Fluo-EX labeling kit (Molecular Probes, Portland OR).
Collagen gel samples were prepared as described above and maintained at 37°C throughout diffusion measurements. For measurements in unassembled collagen solutions, samples were maintained at temperatures between 12 and 17°C by supporting the sample on a metal plate in contact with an ice pack. Gelation did not occur at these temperatures, as detected by a lack of OD450 absorbance, indicating no turbidity or light-scattering in these
67

samples. For both gels and solutions, temperature was continuously monitored using a thermocouple and maintained within ±1°C for all measurements on a given sample.
Measurements were also made in solutions of HA (Sigma) at pH 7.4 and 37°C.
Measurement of Darcy Permeability. Permeability was measured by monitoring flow rate through collagen gels under hydrostatic pressure in an apparatus described previously [17].
Briefly, Transwell or Snapwell cell-culture chambers containing gel samples supported on a highly porous membrane were fit snugly into a sample holder and maintained at 37°C. By adjusting the height of the downstream reservoir, a constant hydrostatic pressure was applied to force flow through the gels. The flow was directed through a thin capillary into which one small air-bubble had been injected. Air-bubble motion was visually undetectable due to the low flow rates through the samples. Thus, the linear velocity of the air-bubble was monitored by a photodiode attached to a servo-null motor, which tracked the bubble for 30 min - 1 hour and was used to determine volumetric flow rates. Hydrostatic pressures (,uP) of 5-15 cm H
2
0
(depending on sample concentration) were imposed to create flow that resulted in the lowest measurable bubble velocity. Low concentration (0.24%) gels were not tested as they were not sufficiently viscous/solid. Gels at 1% were cast in Transwell chambers that fit directly into the apparatus sample holder. Higher concentration gels (i1%) were cast in Snapwell inserts, and a silicone ring was used to seal the space between the insert and the outer
Transwell support. All junctions between plastic and gels (collagen or silicone) were sealed with Krazy Glue to prevent leakage. Leaky samples were quickly detected due to immediate, rapid movement of the air-bubble and were discarded. The surface area (A) and thickness (L) of each sample were measured. The Darcy permeability (K) of the sample was then
68

determined from the time-averaged volumetric flow rate (Q) and viscosity () using Darcy's
Law:
L
Measurement of gel permeability by this method was validated using agarose gels prepared and sealed in identical holders. Results at uP=10 cm H
2
0 matched the values obtained by extrapolating agarose permeability data of Johnson and Deen (1996)[ 18] to zero pressure drop (data not shown). To determine whether the hydrostatic pressure used in these experiments actually compacted the gels and hence produced erroneous results, permeability was measured at two different pressures (10 cm and 5 cm H
2
0). The ratio of the two flow rates was 2.37+0.86 (N = 12), approximately equal to the expected value of 2, suggesting that compaction was not significant.
Visualization by laser scanning microscopy using either confocal reflectance or second
Samples were prepared as described in Transwell inserts and sealed under a coverslip. Confocal reflectance microscopy was performed using a modified Bio-Rad MRC600 (Bio-Rad
Laboratories, Hercules, CA), an Olympus 1OOX 1.4 NA objective (Olympus America Inc.,
Melville, NY), and 488 nm light from a Kr-Ar laser (American Laser Corp., Salt Lake City, UT).
Reflected light from the back surfaces of the objective was attenuated using a quarter wave plate and an analyzer at the detector [19-211. Gels were also imaged using second harmonic generation 1141; 810 nm laser light from a mode-locked Ti:Sapphire was scanned through a sample using a modified Bio-Rad MRC600, and second harmonic light was collected using a
405DF33 bandpass filter and an HC125-02 photomultiplier tube (Hammamatsu, Bridgewater,
NJ).
69

Effective Medium Model. To account for hydrodynamic interactions and relate the permeability of a matrix to its diffusive hindrance, Phillips et al. (1989)[22] proposed the Brinkman (or effective medium) model for a stationary sphere in imposed flow. This model was later modified slightly to account for hindered diffusion in a medium of interest [23, 24]:
D
Do j j+ X + I Rh
The model relates D and K in an immobile, rigid, and homogeneous medium under the assumption that the ratio of a molecule's diffusion coefficient in the medium and solution
(D/Do) is related to its partition coefficient between the phases. The factor alpha is a constant of proportionality introduced to improve the quality of curve fits to this equation. The effective medium model, when used in combination with the Carman-Kozeny model [25] below, was found by Pluen et al. (1999)[26] to give the best correlation with pore size in agarose gel experiments.
Carman-Kozeny Model. We estimated pore size in gels using the Carman-Kozeny model to relate permeability, K, and pore size, a, for a gel of porosity P: a
2
K=
4k
This model treats the gel as an array of cylinders characterized by a geometric factor, k. If the cylinders are assumed to be randomly oriented in three dimensions, the geometric factor is given by:
k = (2k+ + k )
70

where:
2/3
1 1+4(1 /)I / (1 /)21n
The porosity of the gel is related to the volume fraction, r, of collagen by the equation
4 = 1 , where r is the product of the collagen concentration and the effective specific volume of collagen (protein + bound water), previously reported as 1.89 ml/g [9].
4.3 Results
The organization of gels was visualized using a laser-scanning microscope (in confocal reflectance or 2HG mode). Confocal reflectance microscopy and second harmonic generation are both performed in unfixed, hydrated samples, and are useful techniques for the visualization of the collagen network with a spatial resolution of -0.5 m, including distribution and bulk organization of fibers [14, 20]. No structure was detected in collagen solutions at 12-17°C (data not shown). Figure 4.1 shows the isotropic, three-dimensional nature of collagen gels of concentrations 0.24% and 4.5%. After gelation, low-concentration gels (0.24%, Figure 4.1.a) show a highly fibrillar organization as seen previously in gels of comparable concentration [20, 21]. Unlike the long fibers oriented primarily in two dimensions seen by Friedl et al. (1997)[20], our gels show more 3-dimensionally oriented
71

fibers. At higher collagen concentrations studied (1, 3, 4.5%, Figure 4.1.b), CLSM revealed poorly organized collagen with denser arrays of shorter fibers replacing the long fibers seen at lower concentrations. Inhomogeneous organization of collagen gels prepared from highconcentration solutions was also seen by transmission electron microscopy (Figure 4.2) as dense, short-banded structures alongside unbanded filamentous structures. These observations agree with previous reports that at concentrations higher than 0.5%, collagen gels in vitro are formed of a mixture of banded fibrils and filamentous structures [27]. All these gels had an apparent pore size roughly equal to or greater than the -0.5 tm spatial resolution of the microscope.
Figure 4.1 (a and b) Confocal reflectance microscopy of 0.24% (a) and 4.5% (b) collagen gels. Note the long collagen fibers in a in comparison to the shorter collagen fibers in b
(bar = 10 m). (c) Second harmonic image of 0.04% collagen gel subsequently centrifuged to high concentration. Note the retention of long fibers as in (a) (bar = 10 gim).
72

Figure 4.2. Electron microscopy of 4.5% collagen gels (a), and organization of fibrillar collagen in the periphery of the U87 tumor (b) with a high collagen content (estimated at
4.5%). In the collagen gel, compact banded collagen fibrils are find adjacent to filamentous structures that are less organized. In the tumor periphery, compact collagen fibrils are also associated with less organized fibrils separated by larger interfibrillar
When low-concentration collagen solutions were gelated and then centrifuged to a higher concentration, a dense mat of highly fibrillar collagen was formed (Figure 4.1.c) with many long fibers compressed close together, with an interfibrillar spacing close to or smaller than the -0.5 m resolution of the microscope. Note that the presence of organized structures does not preclude the existence of unpolymerized collagen in what appear to be void spaces.
Diffusion data obtained in collagen gels prepared from solutions of various concentrations are shown in Figure 4.3.a, along with data for diffusion in saline and in HA solution. Results of a one-sample t-test on slopes of diffusion coefficient vs. collagen concentration for representative tracer molecules (dextran 4K, BSA, dextran 2M) verified that the diffusion coefficients decrease significantly (p<0.05) with increasing collagen content. The
73

hydrodynamic radius,
Rh, of each molecule was determined from its diffusion coefficient in solution, Do, and the Stokes-Einstein relation, under the assumption that the molecule assumes a spherical configuration: kBT
6:qRl/, where kB is Boltzmann's constant, kB=1.38 x 10-23 J/deg; T is temperature in K, and It is the viscosity of water (0.8705 cP at T = 299 K).
A
Ini
5
B
I
E
* 10-
7 o
:3
,-X
1
0
3
0
§ 10
' a 0o
.
0.8
0.6
Z0
*" 0.4
0
0.2
o
.
0 1
0' 9
1 10 100 10 100
Hydrodynamic radius, RH, nm Hydrodynamic radius, RH, nm
Figure 4.3 FRAP data for diffusion coefficients of tracer molecules at 37 0
C in saline, 0.4%
HA, and 0.24. 1, 3, and 4.5% collagen gels. (a) Diffusion coefficients (D) as a function of tracer molecule hydrodynamic radius (Rh). Lines represent linear fits to data. (b) Diffusional hindrance
(D/Do. where Do is diffusion coefficient in saline) as a function of tracer molecule hydrodynamic radius. Dotted lines represent least-square-error fits to effective medium model (mean ± SD).
For reference., correction to 37°C of the diffusion data of Shenoy and Rosenblatt [28]in 30
10-7 cm2/s for BSA (Rh=4nm). and D37-c=2.0 x 10 7 cm
2
/s for 69kD dextran (Rh=6nm). The linearity of the data sets indicates that the different classes of tracer particles (globular proteins, dextrans, liposomes) behave similarly in our experiments, so that particle conformation and interaction
74

with the matrix do not introduce experimental confounds. In Figure 4.3.b, we plot the ratio of the diffusion coefficients obtained in gels to those in free solution as a function of the experimental hydrodynamic radius, to more clearly illustrate the hindrance presented by the gels. The data clearly indicate that at physiologically relevant concentrations (1-4.5%), collagen poses a significant barrier to diffusive transport. HA solutions at 0.05% (0.5 mg/ml) showed statistically significantly less diffusive hindrance relative to the >1% collagen physiological gels studied here (p < 0.001 for BSA). This HA concentration used was chosen to correspond to the HA content of the four tumors under consideration (see below). At much higher HA concentrations (0.4%), we found significant diffusive hindrance (D/DO -0.56+0.11
for IgG, D/DO -0.27+0.04 for 2,000,000 MW Dextran), equivalent to that found in previous studies [29] (data not shown).
We studied gels prepared from 1% (10 mg/ml), 3% (30 mg/ml), and 4.5% (45 mg/ml) solutions specifically to allow comparison with diffusion data obtained by Netti et al. (2000)
[ 0]and Pluen et al. (2001)[10] in the following tumors implanted in mouse dorsal chambers: human colon adenocarcinoma LS174T, mammary carcinoma MCAIV, human soft tissue sarcoma HSTS-26T, and human glioblastoma U87. Measurements by Netti et al. of collagen and HA content in tumors are given in Table 1. IM concentrations in these tumors are estimated by approximating the interstitial volume fraction of the tumor as f=0.20 [30] and assuming that (1) matrix components are distributed throughout the interstitial volume, and
(2) tissue density is -1 g/ml. Although the interstitial volume fraction will vary between tumors, reaching up to 50% (unpublished data) and matrix component distribution is not uniform within a given tumor, these approximations provide a rough basis for comparison.
75

Tumor Type Collagen
Content (mg/g wet tissue)
HA Content
(mg/g wet tissue)
IM Collagen
(mg/ml IM)
IM HA
(mg/ml IM)
MCaIV
LS174T
U87
HSTS26T
1.8 ± 0.5
1.8± 0.5
8.9 i 4.2
5.8 ± 1.1
0.16 0.03
0.11 0.02
0.11 i 0.03
0.16 ± 0.02
9.0 2.5
9.0 2.5
44.5 i 21
29 ± 5.5
0.80 0.15
0.55 0.10
0.55 0.15
0.80 ± 0.10
Table 4.1 Interstitial matrix composition of human and murine tumors grown in mouse dorsal chambers (based on data of Netti et al.. 2000[ 10])
To compare diffusion in gels and tumors, we also account for the tortuosity of the interstitial space resulting from cellular obstacles, as illustrated in Figure 4.4. Diffusion along an interstitial path with tortuosity r is reduced according to: Dl,=Dge/ - ' [31, 32].
Tortuosity is difficult to measure and exhibits inter and intratumor variation. In the absence of detailed data on the tortuosity of the tumor types in question, the tortuosity of a wellpacked system of cells can be estimated theoretically, although such a theoretical estimate is a possible source of error. Analytical and numerical calculations have yielded the value
=2X1/2 for two-dimensional diffusion in arrays of cells with negligible intercellular spacing, and for diffusion in a two-dimensional isotropic pore network [33, 34]. We use this value to adjust gel data for comparison with tumor tissue data, because the FRAP technique measures two-dimensional radial diffusion.
Figure 4.4 Schematic of the tortuous path encountered by molecules diffusing in the interstitial matrix between tumor cells. Tortuosity is defined as the ratio of effective path length to linear path length (L/Lo).
76

In Figure 4.5.a-c, we compare the adjusted gel diffusion coefficients to the data of
Pluen et al. (200 1)[1 1] in tumors of comparable collagen content. Overall, the gel and tumor data match well, especially considering the absence of other matrix components in the gel and the likely differences in collagen organization and distribution between tumors and gels.
The absence of other matrix components may explain the faster decrease of D with
Rh in tumors than in gels. The difference in slopes is reflected in Figure 4.5.d, which shows an increase in the effective tortuosity,r* =
JD1M/D8el with particle size. The effective tortuosity, r*, is the value of the tortuosity necessary to completely account for the difference between the gel and tumor diffusion coefficients, and reflects effects beyond the geometric considerations discussed above.
Diffusion coefficients were measured in collagen samples pre- and post-gelation.
Measurements were obtained pre-gelation at 12-17°C and corrected to 37 0
C using the
Stokes-Einstein equation. Confocal reflectance images of collagen solutions verified a lack of observable structure in pregelation samples (figure not shown), which was further confirmed by optical density measurements, which were equivalent to those obtained in water. Pre- and postgelation diffusion coefficients were determined for collagen concentrations of 0-4.5%, from multiple measurements within the same sample pre- and post-gelation and are shown in
Figure 4.6. No significant difference was detected between diffusion coefficients pre- and
77

post-gelation at any of the concentrations of collagen studied, after correction for temperature and viscosity using the Stokes-Einstein relation.
a _ if1-
'' -
LSI74T.1ICANIV,U7,:u b10 4
to ,u-
10;10,1
1i
0
AdSTSA
0 .
.
10
4 lo -lO radiusRn
0..4
-
·
,,;d hlgh
'; )
(45i )
MLnt-Carto cstimaw
-c~------
-
I *
1 0
4
I
· -' f i
'"'111 " .
· .
.
.
.
=I i
I
1 10
Hydrodynamic radius, R, nm
100 1 10
Hydrodynamic radius, R,. nm
100
Figure 4.5 (a-c) Comparison of tortuosity-corrected diffusion data in gels to diffusion data in tumors from Netti et al., 2000 and Pluen et al.. 2001 [10, 11]. Corrected diffusion coefficient is calculated as D/T, using estimate I = sqrt(2). Comparisons are show between: 1% gels (open circle ) and data for LS 1 74T, MCAIV, and U87cw (closed
(triangle ) and U87dc (black-triangle) (c). (d) Effective tortuosity necessary to account for discrepancy between uncorrected gel data (Dgel) and tumor data (DIM) as a function of tracer molecule hydrodynamic radius. Values are calculated as T
= (Dg,,/DlM)
1
/
2 from
78

a
10
4
-4
E
%
P~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ i
10
4
-;
E
.t
4104
.
U;W t
'aU o
C
·
~
.
~
10
4
1 10
Hydrodynamic radius, R., nm
--lrag
104.
u
Hydrodynamic radius, R, nm
IM
10
4 l
*# %.
v\\* t i i~~~~~~~~~~~~~~~~~~~~~~~ m
"e
_l a ·
¾*
,1
'O .
14
4.
C
.
4 IL
:'
I
IE
10'
1 10
Hydrodnmamic radius. R,, nm
100 1 10
Hydrodynamic radius, R., nm
Figure 4.6 Comparison of diffusion data before and after incubation (gelation) at 37 0
C for free solution (a), 0.24% collagen gel (b), 1% collagen gel (c), and 4.5% collagen gel
(d). Preincubation data (open circles) obtained at 12-17°C and corrected to 37°C using the Stokes-Einstein equation, and postincubation data (closed circles) obtained at 37°C represent multiple measurements in the same sample. Averaged data (closed squares) obtained from multiple samples and presented earlier in Figure 4.3 are provided for reference.
The diffusion coefficient of 2,000,000 MW dextran was measured in -20% collagen gels
V'"V"Y VVI Y VU·II· · LUn U ULI 1 ·
±
79

1.7 x 10
-
8 cm 2 s
-1 was significantly faster (p < 0.01) than the value of 7.8 ± 5.3 x 10
- 9 cm
2 s
- 1 measured in collagen gels prepared by direct gelation of 4.5% collagen solutions as discussed above.
The Darcy permeability of 1%, 3%, and 4.5% collagen gels was determined experimentally and also estimated from diffusion data using the effective medium model. Curve-fits of the diffusion data to the model are shown in Figure 4.3.b and the experimental measurements and model estimates of the Darcy permeability are compared in Figure 4.7.a. The experimental values and model estimates agree only for the 1% gels. Above this concentration, the experimental measurements are increasingly greater than the model estimates, with an order of magnitude difference for the 4.5% gels. This difference in permeability values translates into a difference in pore size as estimated by the Carman-
Kozeny model, as shown in Figure 4.7.b.
The permeability of gels correlated inversely with collagen content, whereas the permeability of tumors with corresponding collagen content did not (Figure 4.8.a). To compare the permeability measurements in collagen gels with the published measurements in tumors, the gel measurements must be adjusted by the area fraction (fA) in a tumor slice and the tortuosity, or increased length of the fluid path through the slice. Adjusting the gel data by
80

Ktumor=Keil f/T, where the interstitial area fraction is estimated at fA=0.
2 and the theoretical estimate =21/2 is used for the tortuosity, we obtained the data shown in Figure 4.8.b
alongside published tumor measurements. Collagen gels corrected for the absence of cells are significantly less permeable than tumors of comparable collagen content, although direct comparison may be complicated by the use of different permeability measurement techniques for gels and tumors, and by intratumoral ECM heterogeneity.
10' : .
1000
100 .. 100 .
s-
-
3
Percent collagen, g/l00ml
4
10i
1 2 3
Percent collagen, g/100ml
4
Figure 4.7 (a) Comparison of experimental measurements (closed circle) and modelbased predictions (open circles) of Darcy permeability as a function of collagen gel concentration,. Model predictions are obtained from application of the effective medium model to diffusion data (see curve fits. Figure 4.3.b). (b) Comparison of theoretical pore size predicted by Carman-Kozeny model from experimental measurements of Darcy permeability (closed circle), and from effective medium estimates of permeability from diffusion data (open circle).
81

E fc1
-)4.
1W
I
'i
1000
't'.
0 0 0 too
10
0
1% 3% 4.i%
Percent collagen, glOOml
1% 3% 4.5%
Percent collagen, g/lOnml
Figure 4.8 (a) Comparison of Darcy permeability measurements in tumors and gels of measured collagen concentrations. Confined compression measurements of K in MCaIV,
LS174T (1), HSTS26T (1), and U87 tumors [10], micropipette measurements of K in
LS 174T (2) tumors [5], and pressure gradients across clamped HSTS26T (2) tumor tissue sections [35]. (b) Comparison of Darcy permeability measured in tumors to those in gels when corrected for area fraction and tortuosity.
Collagen significantly impedes diffusion, and the extent to which it does so, when corrected for the tortuosity of the interstitium, is consistent with diffusion data obtained in tumors of comparable collagen content (Figure 4.5.a-c). Note that the slope of the diffusion data differs between gel and tumor data sets. This phenomenon is also seen as an increase with molecular size of the effective tortuosity in tissue (Figure 4.5.d), and has been observed in studies of diffusion in the brain 321. In tumors, matrix components other than collagen could affect this slope by differentially affecting the diffusion of small versus large molecules. Heterogeneity of collagen structure and distribution in tumors, as shown by Pluen et al. (2001)[1 may also
82

differentially affect particles of different sizes. Thus the effective tortuosity in a tumor scales with particle size and is heterogeneous, depending on the local tissue composition and structure.
Our results suggest that diffusion in pure collagen gels mimics that in the tumor IM over the wide range of particle sizes studied. However, extrapolating these results to particles with a hydrodynamic radius larger than 2,000,000 MW dextrans may not be justified. The
Carman-Kozeny estimates of pore size and the linearity of the diffusion data sets suggest that the particles we used are smaller than the effective pore sizes of the gels studied. As particle sizes approach the effective pore size of the media, the fine structure of the matrix is expected to critically influence transport hindrance, and in vitro gels may no longer capture the in vivo behavior. Rusakov and Kullmann (1998)[36] argued that large molecules comparable to the pore size experience greater hindrance due to viscous interactions unaccounted for in tortuosity corrections. Matrix pore size is expected to be different in gels than in tumors, where factors such as compaction of collagen fibrils by fibroblasts [20, 37, 38] and additional IM molecules such as decorin 139] play a role. Thus, although the agreement between the gel and tumor measurements is surprisingly good, these results should not be extrapolated to larger particle sizes.
After correction for the effect of temperature on viscosity and molecular motion, there was no significant difference in diffusion between collagen solutions and collagen gels gelated from equivalent concentrations. These data are consistent with those of Shenoy and
Rosenblatt (1995)128], where solutions of succinylated collagen at room temperature were capable of significantly slowing diffusion. This fact, combined with imaged pore sizes that appear too large (several hundreds of nanometers) to significantly hinder diffusion, suggests that unassembled collagen in the void spaces of these gels plays a role in hindering
83

diffusion.
Note that gels formed by gelation of different concentrations of collagen are not simply more or less concentrated versions of the same structure. The highly fibrillar network formed from the gelation of low-concentration collagen solutions is qualitatively different from the dense array of short fibers and partially formed structures generated upon gelation of high-concentration collagen solutions. When gels formed from lowconcentration collagen solutions (-0.04%) are subsequently centrifuged to higher concentrations (-20%) than the gels formed by direct gelation of high-concentration solutions (-4.5%), the resultant gel retains its original highly fibrillar structure, but the long fibers are significantly compacted, forming a dense mat. Not surprisingly, these qualitatively different gels prepared by centrifugation postgelation do not reproduce the diffusive hindrance of gels prepared by simple gelation, exhibiting a significantly higher diffusion coefficient for 2,000,000 MW dextran. The compaction of the array of long fibers initially formed at low concentrations could be markedly inferior to that of the dense array of short fibers and partially formed structures generated by gelating a high-concentration solution.
Additionally, it is known that the partitioning of collagen between assembled and unassembled states varies with the concentration at which the gel is polymerized [27]. We conclude that the poorly assembled gels formed by simple polymerization of collagen solutions and containing that proportion of unassembled collagen dictated by the concentration at time of gelation are the gels that quantitatively mimic the diffusive hindrance of tumor interstitium of equivalent collagen concentration.
Although these gels quantitatively mimic the diffusive hindrance of the tumor interstitium, this does not mean that these gels completely reproduce the interstitial matrix at a molecular level. Other matrix molecules are certainly present in vivo, and the structure of collagen assembled in vivo is likely to differ from that assembled in vitro. However, the poorly assembled gels studied here do have structural similarities to the collagen of the tumor interstitium, which is poorly organized in comparison to normal tissue. Pluen et al.
84

(2001)1111 reported that subcutaneous U87 tumors stain positively for collagen type I in the tumor center where only few fibrils were detected by EM visualization, whereas the periphery of U87 and other tumor types showed a high density of collagen fibrils. These results suggest that unassembled molecules between the fibers of the interstitial matrix can influence the diffusion of macromolecules in vivo just as they seem to do in vitro. In pure collagen type I gels, these unassembled molecules can only be collagen type I, while in vivo, these unassembled molecules may include other matrix molecules, such as nonfibrillar collagen type I, other collagen types, or HA.
The diffusion data attest to the ability of collagen gels at concentrations comparable to those of the tumor IM to significantly hinder diffusive transport (Figure 4.3). In contrast, HA solutions at concentrations comparable to the tumors analyzed here (0.05%) pose a weaker barrier to diffusion. For 3% and 4.5% collagen gels, the diffusive barrier offered by HA
(i.e., D/Do) is far less than that offered by collagen, suggesting that in tumors with these collagen concentrations (e.g., HSTS26T and U87), collagen alone can account for the diffusive hindrance in the tumor. For the lowest collagen concentration gels (1%), the barrier offered by HA is over half the barrier offered by collagen, suggesting that in tumors with this collagen concentration (e.g., LS 174T) HA may have some influence on diffusive hindrance.
This finding does not apply to tissues with higher HA content, including the tumor spheroids studied by Davies et al. (2002)[40] and other GAG-rich tissues, such as cartilage.
Furthermore, the pure HA solutions do not replicate possible in vivo interactions between different species of matrix molecules (e.g., Turley et al., 1985[41]), which may affect transport properties.
85

Data collected from several organs have indicated that permeability is inversely correlated to collagen content [91. We have found the same trend in collagen gels.
However, the permeability values in tumors did not match the data in collagen gels quantitatively, nor did they show the qualitative inverse correlation with collagen content.
Furthermore, when the data for collagen gels were adjusted for area fraction and tortuosity in tumors, the permeability was higher in tumors than in gels of comparable concentration. The differences in permeability could be due partially to measurement techniques. Even within tumors, the confined compression technique used by Netti et al.
(2000)1 101 predicted significantly higher hydraulic conductivity compared to the micropipette approach [5] and clamp methods [35]. The lack of correlation between collagen and permeability observed by Netti et al. in tumors suggests a more important contribution from other matrix molecules.
Estimates of gel permeability based on the effective medium model matched experimental measurements of permeability only for 1% collagen gels (Figure 4.7). At greater concentrations, the diffusion-based effective medium model values increasingly underestimated the true permeability. In contrast, the model was reported to be accurate for agarose gels 126], and underestimated diffusion coefficients in various other gels, a deviation qualitatively opposite to that observed here [24]. In general, discrepancies between gel measurements and effective medium model predictions may result from model assumptions of fiber rigidity, immobility, and homogeneity. Furthermore, the effective medium model empirically relates two fundamentally different modes of transport (convection and
86

diffusion), which can be differentially regulated. The accuracy of the effective medium prediction at low collagen concentration and the increasing discrepancy at higher collagen concentrations may also indicate that high concentrations of poorly organized collagen pose a greater barrier to diffusion than to convection. This argument is also supported by the observation that diffusional hindrance in tumors correlates with collagen content 111], whereas the measured permeability of tumors does not [10].
4.5 Conclusions
In conclusion, our data show that collagen at physiological concentrations presents a major barrier to molecular diffusion, especially for larger particles. Furthermore, theoretical correction of gel diffusion data for the effects of in vivo tortuosity yielded good agreement with in vivo measurements in tumors of comparable collagen concentration. The diffusive hindrance data combined with imaging of the gels and permeability measurements suggest that unassembled collagen in the void spaces of the gel plays a role in hindering diffusion. In vivo, this role may be played by unassembled collagen or other matrix molecules. These findings support our hypothesis that collagen is a major contributor to diffusive hindrance in tumors. In addition, it suggests that in vitro gel models can be used to investigate diffusion in tissues, with theoretical correction for issues such as tortuosity providing the necessary bridge between the in vivo and in vitro measurements. This work has important implications for drug delivery in tumors and for tissue engineering, where transport in collagen-based tissue replacements or scaffolds is an important design consideration. Furthermore, interfering with collagen synthesis or reducing collagen content may improve drug delivery to tumors 1421.
87

4.6 References
I.
Alberts, B., D. Bray, J. Lewis, M. Raff, K. Roberts, and J. Watson, Extracellular
Matrix of Animals, in Molecular Biology of the Cell, B. Alberts, D. Bray, J.
2.
Lewis, M. Raff, K. Roberts, and J. Watson, Editor. 1994, Garland Publishing:
New York. p. 971-995.
Gabizon, A., D. Goren, R. Cohen, and Y. Barenholz, 1998. "Development of liposomal anthracyclines: from basics to clinical applications." J Control Release,
53(1-3): p. 275-9.
3.
4.
5.
6.
7.
8.
Kulkarni, S.B., G.V. Betageri, and M. Singh, 1995. "Factors affecting microencapsulation of drugs in liposomes." J Microencapsul, 12(3): p. 229-46.
Costantini, L.C., J.C. Bakowska, X.O. Breakefield, and 0. Isacson, 2000. "Gene therapy in the CNS." Gene Ther, 7(2): p. 93-109.
Boucher, Y., C. Brekken, P.A. Netti, L.T. Baxter, and R.K. Jain, 1998.
"Intratumoral infusion of fluid: estimation of hydraulic conductivity and implications for the delivery of therapeutic agents." British Journal of Cancer,
78(11): p. 1442-14448.
Jain, R.K., 1999. "Transport of molecules, particles, and cells in solid tumors."
Annual Review of Biomedical Engineering, 01: p. 241-263.
Netti, P.A., et al., 1999. "Enhancement of fluid filtration across tumor vessels: implication for delivery of macromolecules." Proc Natl Acad Sci U S A, 96(6): p.
3137--42.
Gribbon, P.M., A. Maroudas, K.H. Parker, and C.P. Winlove, Water and solute
determinants, in Connective tissue biology : integration and reductionism, R.K.
Reed and K. Rubin, Editors. 1998, Portland: London; Miami. p. 95-124.
9. Levick, J.R., 1987. "Flow through interstitium and other fibrous matrices." Q J
Exp Physiol, 72(4): p. 409-37.
10. Netti, P.A., D.A. Berk, M.A. Swartz, A.J. Grodzinsky, and R.K. Jain, 2000. "Role of extracellular matrix assembly in interstitial transport in solid tumors." Cancer
Res, 60(9): p. 2497-503.
11. Pluen, A., et al., 2001. "Role of tumor-host interactions in interstitial diffusion of macromolecules: cranial vs. subcutaneous tumors." Proceedings of the National
Academy of Sciences of the United States of America, 98(8): p. 4628-33.
12. Jain, R.K., 1998. "The next frontier of molecular medicine: delivery of therapeutics." Nat Med, 4(6): p. 655-7.
13. Sano, A., T. Hojo, M. Maeda, and K. Fujioka, 1998. "Protein release from collagen matrices." Adv Drug Deliv Rev, 31(3): p. 247-266.
14. Williams, R.M., W.R. Zipfel, and W.W. Webb, 2001. "Multiphoton microscopy in biological research." Curr Opin Chem Biol, 5(5): p. 603-8.
15. Berk, D.A., F. Yuan, M. Leunig, and R.K. Jain, 1993. "Fluorescence photobleaching with spatial Fourier analysis: Measurement of diffusion in lightscattering media." Biophysical Journal, 62: p. 2428-36.
88

16. Berk, D.A., F. Yuan, M. Leunig, and R.K. Jain, 1997. "Direct in vivo measurement of targeted binding in a human tumor xenograft." Proceedings of
the National Academy of Sciences, USA, 95: p. 1785-90.
17. Chang, Y.S., et al., 2000. "Effect of vascular endothelial growth factor on cultured endothelial cell monolayer transport properties." Microvasc Res, 59(2): p. 265-77.
18. Johnson, E.M. and W.M. Deen, 1996. "Hydraulic permeability of agarose gels."
AIChE Journal, 42(5): p. 1220-1224.
19. Cheng, P.C., and R. G. Summers, Image contrast in confocal microscopy, in
Handbook of Biological Confocal Microscopy, J.B. Pawley, Editor. 1990, Plenum
Press: New York. p. 170-196.
20. Friedl, P., K. Maaser, C.E. Klein, B. Niggemann, G. Krohne, and K.S. Zanker,
1997. "Migration of highly aggressive MV3 melanoma cells in 3-dimensional collagen lattices results in local matrix reorganization and shedding of 2 and 131 integrins and CD44.." Cancer Research, 57: p. 2061-2070.
21. Brightman, A.O., B.P. Rajwa, J.E. Sturgis, M.E. McCallister, J.P. Robinson, and
S.L. Voytik-Harbin, 2000. "Time-lapse confocal reflection microscopy of collagen fibrillogenesis and extracellular matrix assembly in vitro." Biopolymers,
54(3):: p. 222-34.
22. Phillips, R.J., W.M. Deen, and J.F. Brady, 1989. "Hindered transport of spherical macro-molecules in fibrous membranes and gels." AIChE Journal, 35(11): p.
1761- 1769.
23. Solomentsev, Y.E. and J.L. Anderson, 1996. "Rotation of a sphere in Brinkman fluids." Physics of Fluids, 8(4): p. 1119.
24. Phillips, R.J., 2000. "A hydrodynamic model for hindered diffusion of proteins and micelles in hydrogels." Biophys J, 79(6): p. 3350-3.
25. Carman, P.C., 1937. "Fluid flow through granular beds." Trans. Inst. Chem. Eng.,
15: p. 150-166.
26. Pluen, A., P.A. Netti, R.K. Jain, and D.A. Berk, 1999. "Diffusion of macromolecules in agarose gels: comparison of linear and globular configurations." Biophys J, 77(1): p. 542-52.
27. Williams, B.R., R.A. Gelman, D.C. Poppke, and K.A. Piez, 1978. "Collagen fibril formation. Optimal in vitro conditions and preliminary kinetic results." JBiol
Chem, 253(18): p. 6578-85.
28. Shenoy, V. and J. Rosenblatt, 1995. "Diffusion of macromolecules in colagen and hyaluronic acid, rigid-rod - flexible polymer, composite matrices."
Macromolecules, 28: p. 8751-58.
29. De Smedt, S.C., A. Lauwers, J. Demeester, Y. Engelborghs, G. De Mey, and M.
Du, 1994. "Structural information on hyaluronic acid solutions as studied by proble diffusion experiments." Macromolecules, 27: p. 141-146.
30. Jain, R.K., 1987. "Transport of molecules in the tumor interstitium: a review."
Cancer Res, 47(12): p. 3039-51.
31. Nicholson, C. and J.M. Phillips, 1981. "Ion diffusion modified by tortuosity and volume fraction in the extracellular microenvironment of the rat cerebellum." J
Physiol (Lond), 321: p. 225-57.
32. Nicholson, C. and E. Sykova, 1998. "Extracellular space structure revealed by diffusion analysis." Trends in Neuroscience, 21: p. 207-215.
89

33. Blum, J.J., G. Lawler, M. Reed, and I. Shin, 1989. "Effect of cytoskeletal geometry on intracellular diffusion." Biophys J, 56(5): p. 995-1005.
34. Chen, K.C. and C. Nicholson, 2000. "Changes in brain cell shape create residual extracellular space volume and explain tortuosity behavior during osmotic challenge." Proc Natl Acad Sci U S A, 97(15): p. 8306-11.
35. Griffon-Etienne, G., Y. Boucher, C. Brekken, H.D. Suit, and R.K. Jain, 1999.
"Taxane-induced apoptosis decompresses blood vessels and lowers interstitial fluid pressure in solid tumors: clinical implications." Cancer Res, 59(15): p. 3776-
82.
36. Rusakov, D.A. and D.M. Kullmann, 1998. "Geometric and viscous components of the tortuosity of the extracellular space in the brain. " Proc Natl Acad Sci U S A,
95(15): p. 8975-80.
37. Guidry, C. and F. Grinnell, 1987. "Heparin modulates the organization of hydrated collagen gels and inhibits gel contraction by fibroblasts." Journal of Cell
Biology, 104: p. 1097-1103.
38. Huang-Lee, L.L.H., J.H. Wu, and M.E. Nimni, 1994. "Effects of hyaluronan on collagen fibrillar matrix contraction by fibroblasts." Journal of Biomedical
Materials Research, 28: p. 123-132.
39. Pins, G.D., D.L. Christiansen, R. Patel, and F.H. Silver, 1997. "Self-assembly of collagen fibers. Influence of fibrillar alignment and decorin on mechanical properties." Biophys J, 73(4): p. 2164-72.
40. Davies, C.d.L., D. Berk, A. Pluen, and R.K. Jain, 2000. "Correlation between diffusion of IgG and extracellular matrix in rhabdomyosarcomas growing as tumors in dorsal chambers or multicellular spheroids." in preparation.
41. Turley, E.A., C.A. Erickson, and R.P. Tucker, 1985. "The retention and ultrastructural appearances of various extracellular matrix molecules incorporated into three-dimensional hydrated collagen lattices." Dev Biol, 109(2): p. 347-69.
42. McKee, T.D., A. Pluen, Y. Boucher, S. Ramanujan, E. N. Unemori, B. Seed, and
R. K. Jain, 2001. "Relaxin increases the transport of large molecules in high collagen content tumors." Proceedings of the American Association for Cancer
Research, 42: p. 30.
90

5.1 Introduction
Collagen content and structure are key determinants of macromolecular transport in tumors[1-3]. We propose that the penetration of therapeutic molecules could be estimated based on the correlation between collagen content and diffusive transport[l]. We also propose that drug penetration in tumors could be improved by administering agents that modify the matrix and increase diffusion. Testing these hypotheses would require a routine, noninvasive technique to monitor the collagen content and structure of tumors in vivo.
Collagen is known to induce SHG[4-7]. Here, we obtain high-resolution images of fibrillar collagen in tumors in vivo using SHG. SHG imaging offers many advantages: SHG is an intrinsic signal and does not require the addition of extrinsic dyes; the signal and background are better than those of autofluorescence imaging; nonlinear excitation permits threedimensional resolution in vivo[8], and the SHG emission wavelength scales with the excitation wavelength, allowing spectral separation between signals from fibrillar collagen and other fluorophores.
91

Mouse mammary adenocarcinoma MCaIV and human colon adenocarcinoma LS 1 74T, melanoma MU89 and soft tissue sarcoma HSTS26T were grown in dorsal skinfold chambers in severe combined immunodeficient mice[ 1]. All experiments were done with approval of the Institutional Animal Care and Use Committee. Images were obtained using a custom-built multiphoton laser scanning microscope[9, 10]. Images of the SHG signal were obtained using 810 nm excitation and 405DF30 emission filters (Chroma).
Spectra were generated using a focal spectrum analyzer based on a multiphoton laser scanning microscope (12-nm spectral resolution). The short wavelength signal from the fibrillar structures in tumor in vivo was located at exactly half the excitation wavelength and had an extremely narrow bandwidth. However, at longer wavelengths, there was a broad autofluorescence peak generated mainly by punctate structures. The narrow bandwidth of the short wavelength peak, the fact that it was located at half the excitation wavelength and the fact that it shifted center wavelength when the excitation wavelength shifted were all consistent with SHG.
Collagen I gels (Cohesion Technologies) were prepared at concentrations that reproduced tumor matrix (15-65 mg/ml)[3]. Coverslips were coated with mouse ultrapure laminin and mouse collagen IV (BD Biosciences) at a concentration of 10 mug/cm2 and were
92

dried. Coverslips were coated with Matrigel (BD Biosciences) at a concentration of 12 mg/ml using the manufacturer's thick gel protocol. The mean SHG signal at 810 nm excitation was determined by generating a series of five image stacks of each gel with a times5 objective, each stack consisting of five 2-mm times 1.33-mm images spaced 20 mum apart, using identical imaging conditions (equal laser power, photomultiplier tube voltage and so on). Maximum intensity projections of each stack were generated and mean pixel counts were calculated. For the matrix materials that formed only thin layers on the coverslip, optical sections were generated at the optimum focus, determined using
770 nm excitation and a 400-nm to 500-nm filter, allowing detection of faint autofluorescence.
The mean SHG signal of five specimens of three tumor types was calculated by acquiring multiple three-dimensional image stacks as described above. Maximum intensity projections of each stack were generated and a two-dimensional montage of the visible tumor surface was formed (Figure 5.1 .a). An outline was drawn around the tumor area and the mean pixel count in that area was calculated.
Tumors (six LS 174T; five MU89) were grown in the dorsal skinfold chamber until they reached approximately 3-4 mm in diameter. Excised tumors were fixed in 4% paraformaldehyde, embedded in optimum cutting temperature compound (Sakura
Finetek) and cut into sections 10 mum in thickness oriented perpendicular to the surface.
93

Collagen I staining and quantification were done as described before[2]. This generated
41 images for LS 174T and 52 for MU89. To identify elastin, adjacent sections from an
MU89 tumor embedded in paraffin were prepared. One section was stained with the
Weigert stain and the adjacent section was treated with a bath of xylene and a series of ethanol baths at decreasing concentrations to remove paraffin and was used for spectral analysis.
Diffusion coefficients were measured by FRAP[1-3] seven times in each of four LS174T and five Mu89 samples, and four to eight times in control and relaxin-treated HSTS26T tumors.
The dynamic action of collagenase I was monitored in vivo by removal of the dorsal skinfold chamber coverslip and pipetting of -100 gll Clostridium collagenase (0%, 1% and 10% solutions in saline; Sigma). Consecutive SHG stacks were obtained every 5 min using a X20 objective, 0.5 numerical aperture H
2
0 lens, each stack consisting of 30 images spaced 5 glm apart, covering an area of 3.03 x10
5 im2. Relaxin was delivered by
Alzet 1002 osmotic pumps (Durect) loaded with 100 gl recombinant human relaxin (5 mg/ml) in mice bearing 14-day-old HSTS26T tumors. SHG stacks were generated every
2-3 d for 2 weeks. Maximum-intensity projections were generated and mean pixel counts in the area of the image were calculated to yield the average SHG signal. Average fiber lengths were determined by binarizing maximum-intensity projections of the first image
94

stack in a time series with a threshold chosen to set 90% of the image equal to zero. The resultant image was inspected to choose the five brightest fibers. The three-dimensional end-to-end length of each fiber was measured in the original image by determining the shortest distance between a fiber's ends using Scion Image software (Scion). An 'end' was defined as the point at which the fiber was no longer visible or bifurcated. A fiber's length therefore changed if it became cleaved, its ends shortened, its branching structure changed or it curled into a loop. Fiber lengths were measured in all stacks until their location could no longer be recognized. Hence, drastically changed morphology in a region prevented fiber tracking in that region, thereby underestimating average changes in fiber length of a highly dynamic population. Lines were drawn around regions of interest encompassing selected fibers to measure the average pixel counts of the individual fibers.
Statistical significance was determined using Student's t-test. Equality of variances was tested using an F-test or analysis of variance. All values were expressed as mean + s.e.m.
95

We validated the utility of SHG imaging in tumors by implanting Mu89 human melanomas in the dorsal skinfold chamber of immunodeficient mice. We generated highcontrast images of fibrillar structures (Figure 5.1 .a,b), whose emission spectra indicated that the fibrillar structures were imaged by SHG[4] (Fig. 5.1 .c).
Figure 5.1. SHG imaging of tumors in vivo a) Second-harmonic signal in a Mu89 melanoma grown in the dorsal skinfold chamber of a severe combined immunodeficient mouse, image to left. This image was a montage of 12 separate images, each of which was a maximum intensity projection of 5 images obtained at 20 [m steps. The image shown is 6.6 mm in width. (b) Second-harmonic signal with highlighted vessels. Vessels were highlighted with an intravenous injection of 0.1 ml tetramethylrhodamine-dextran
(10 mg/ml; red pseudocolor). SHG signal. green pseudocolor. There was no colocalization of SHG signal with the borders of blood vessels. The image shown is 275 pim in width.
96

en
0.5
0.4
I I I
I e-810 nm excitation
MMM"
7C"
- -I__
TM
._
c k 0.3
C
0.2
i
.74
Uj
0.1
i i i
II
1,P f
I
370 390 410 430 450 470 490 510 530
Wavelength, nm
Figure 5.1. SHG imaging of tumors in vivo (c) Average spectra of light generated with
810 nm excitation of an approximately 0.25-mm
2 region of a Mu89 melanoma in the dorsal skinfold chamber of an immunodeficient mouse.
To determine the origin of the SHG signal in tumors, we did four studies. First, we imaged gels made of collagen I or Matrigel (a mouse tumor basement membrane extract) and layers of collagen IV and laminin deposited on coverslips, all at tumor concentrations. Using identical SHG conditions, we obtained distinct SHG images of a fibrillar meshwork from the collagen I gel (data not shown), but no signal from the other preparations (average SHG pixel count/SHG pixel count of collagen fibers 9.3 x10
3
).
Second, we noted SHG structures colocalizing with fluorescently labeled antibody to collagen I in tumor sections. A highly fibrillar subpopulation of antibody-labeled structures produced SHG (Fig. 5.2.a).
97

._-.-__
Figure 5.2. SHG images of fibrillar collagen I in tumors Second-harmonic signal colocalized with staining with antibody to collagen I in a tumor section 10 pm in thickness. Left, antibody to collagen I conjugated with FITC (red pseudocolor) and 4,6diamidino-2-phenylindole (green pseudocolor); right, SHG (red pseudocolor) and 4,6diamidino-2-phenylindole (green pseudocolor). The SHG signal appeared as narrow fibers that formed a subpopulation of the antibody-stained structures. Each image was a maximum intensity projection of five images spaced 5 plm apart, and are 275 plm in width. The surface of the tumor facing the 'window' of the dorsal skinfold chamber was to the right in each image.
98

1
C
0.8
.
E
._
0.6
0.4
---- I
I
._
m
0
0.2
i i
I
I
I
I
I
I i i i
I
I i
I
I
11
I
II---- i
I
/
I
I
I
I i -
I i i i
i i
I
I -i i i i i
___I__J_.___-.__ i.__~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
E
W
I
370 390
I o Elastin I l
I I I I I
410 430 450 470 490 510 530
Wavelength, nm
-----
Figure 5.2 SHG images of fibrillar collagen I in tumors (b) Elastin did not generate significant SHG in tumors. Left, fixed section of a Mu89 mouse melanoma stained with
Weigert stain and imaged with transmitted light. Dark brown, elastin; red arrow, thin elastin band around the peritumor arteriole. Right, spectrum of the selected band in an adjacent unstained section. Elastin did not generate significant SHG at 405 nm, half the excita r lgure .zL. images o iDrlllar collagen 1 in tumors (c) SHG was better than autofluorescence in imaging fibrillar collagen in tumors. Left, image generated with SHG
(81 0-nm excitation), showing fibrillar collagen with high signal and low background.
Right. image generated at the same location with 770-nm excitation and 400- to 500-nm emission (optimum for autofluorescence from collagen), dominated by a bright punctate background from nonfibrillar autofluorescent molecules.
99

Third, we noted blood vessels in tumors and normal tissue in dorsal skinfold chambers but did not find SHG signal forming a sheath around blood vessels (Fig. 5.1 .b). A sheath of mainly nonfibrillar collagen IV forms the basement membrane of capillaries, postcapillary venules and arterioles[ 11]. Around peritumor arterioles in fixed sections, we noted an autofluorescent ring whose excitation and emission peaks and physiological distribution were consistent with those of the elastin layer. The elastin layer did not generate detectable SHG, however (Fig. 5.2.b).
Fourth, we demonstrated the advantage of SHG imaging in tumors in vivo by comparing
SHG and autofluorescence of collagen I (ref. [12]; Fig. 5.2.c). In collagen gels, distinct autofluorescence imaging of collagen is possible[5], but in tumors in vivo the situation is much more complicated. We noted bright punctate autofluorescence from other intrinsic fluorophores in the tissue, which overwhelmed the autofluorescence from collagen fibers
(Fig. 5.2.c), whereas collagen fibers were distinctly visible in the SHG image of the same region.
These observations indicated that SHG signal in tumors in vivo readily imaged a highly fibrillar subpopulation of collagen I and did not image elastin, collagen IV or other basement membrane components. In vivo images of collagen in tumors are detected using the same objective lens used to provide excitation light (epidetection mode). A relatively small fraction of SHG light is expected to be transmitted back to the objective lens[4]; most SHG light is expected to be transmitted in the direction of the excitation light[8].
100

Unfortunately, imaging with transmitted light in tumors in vivo is extremely inefficient because tumors are more than 1 mm in thickness and, depending on the surgical preparation, can have the animal's skin, other tissues or whole body on the other side of the tumor.
The total collagen content of tumors can serve as a predictor of diffusive transport[ 1]. We hypothesized that the SHG signal from fibrillar collagen could be used to easily and noninvasively assess the penetrability of tumors. SHG signal depends on the orientation of the fiber relative to the plane of polarization of the laser. It also depends on the local spacing of fibers, scaling linearly with fiber spacing for widely dispersed fibers and quadratically with fiber spacing when several fibers are contained in one focal volume[8,
13]. Tumor images showed a random distribution of fiber orientations and a wide variety of interfiber spacings in tumors (Fig. 5.1.a,b). Hence, it was difficult to predict how the average SHG signal of a tumor would scale with total collagen content. To determine this, we made collagen I gels with a range of concentrations corresponding to the tumor models studied here (see Methods)[1, 3]. We collected the average SHG signal over wide areas of the gel to obtain an average over the various fiber spacings and orientations (see
Methods and Fig. 5.3.a). The average SHG signal scaled with approximately the first power of the total collagen content (power-law exponent, 0.91). The relationship between gel SHG signal and concentration did not depend on the numerical apertures of the range of lenses used here (numerical apertures, 0.15-0.5; P > 0.2; n = 10).
101

An
-y = 0.84443* xA(0.91021) R= 0.99466
I I l I
.
-
C
,)
Ub
L.
- e -
*'i"t
_-I t ---':-"-
.zZ1Z11._
-
10-
-rl--L.-
-
, , i j .
I
I ·
_
_i l
~;
l
1
_ l
;i l
-CL lic
: l
75
_e i
I-c c
T
L2-4 _
I
,, I
-t
5.
1
--
0 30 40 50 60 7 D
Collagen Gel Concentration, mg/ml
Figure 5.3 Dependence of SHG signal on tumor type.
Mu89 LS1 74T MCalV
Second-harmonic signal scaled linearly with collagen gel concentration. Collagen gels were prepared that reproduced tumor diffusive hindrance over a range of concentrations equal to the tumors studied here. The average SHG signal generated from these gels scaled with the first power of the collagen concentration (1 8-24 measurements at each concentration; best-fit power law exponent = 0.91; Ri
=
0.995). (b) Histogram of average SHG signals, which varied with diffusion, had a high SHG signal. MCaIV mammary adenocarcinomas and LS174T colon adenocarcinornas, which have identical, low collagen content and identical, high macromolecular diffusion, had identical low SHG signal. Although the difference between the mean SHG signal of LS 1 74T and Mu89 was statistically significant (P < 0.05), the difference between MCalV and Mu89 was just below significance (P = 0.064), and the difference between
LS 174T and MCaIV was not statistically significant.
We next measured the SHG signal from tumors in an identical way. We examined three tumor types, MCaIV, LS174T and Mu89, because of known differences in their collagen content[l. 2]. The SHG signal varied with tumor type, with the signal of MCaIV being about equal to that of LS 174T. which was less than that of Mu89 (Fig. 5.3.b). We then compared this result to published values of collagen content or to measurements of collagen by immunohistochemical staining of tissue sections. We found diffusion
102

coefficients in published reports or measured using fluorescence recovery after photobleaching[ 1, 2] (Table 1). In both cases (LS174T:MCaIV and LS174T:Mu89), the ratio of SHG signals was not statistically different from the ratio of collagen quantification (P > 0.05) and correlated with the relative diffusive hindrance in these tumor types (MCaIV being about equal to LS 174T, which was less than Mu89). These results showed that in three diverse tumor types, our ability to detect tumor collagen with
SHG allowed us to simply and rapidly quantify relative collagen content and estimate drug delivery characteristics noninvasively, obviating the need for biopsies, sectioning or staining.
Tumor types SHG ratio Collagen ratio Diffusion coefficient ratio
LS174T:MCaIV 0.75 + 0.46 1.0 0.22
(hydroxyproline content from
0.96 ± 0.21
(FRAP measurements from published reports)' published reports)'
0.41 0.073 1.45 ± 0.33
LS174T:Mu89 0.29 ±0.16
(immunohistochemical staining
(FRAP measurements
Table 5.1. Comparison of SHG ratios with collagen and diffusion coefficient ratios
103

We hypothesized that SHG allowed dynamic imaging of collagen modification in vivo.
To test this, we applied bacterial collagenase to Mu89 melanomas and imaged the tumors
(Fig. 5.4.a). The SHG signal of collagen fibers disappeared with a single exponential decay consistent with Michaelis-Menten kinetics. The exponential decay times were 3.7 x
10
3 min, 1.2 x 102 min, and 9.9 min for 0%, 1% and 10% collagenase, respectively. The mean collagen fiber length did not change significantly (P > 0.05), consistent with the possibility that a free enzyme attacked multiple sites along a fiber, 'dissolving' it instead of cleaving the fiber at one location. Although the unknown constant of proportionality between local in vivo SHG signal and local in vivo collagen concentration prevented an exact calculation of the kinetic constants for the enzymatic reaction, the exponential decay time provided direct measurement of relative enzymatic efficacy. The exponential decay time for SHG loss scaled linearly with collagenase concentration, at least up to
10% collagenase, indicating that these concentrations were in the linear range of the dose-response curve. The considerable loss of SHG signal from tumors after collagenase application was consistent with the substantial (approximately twofold) increase in diffusion coefficient measured in tumors treated in an identical way[ 1].
104

M
-
I.
(w
(A
----
I i y 1.0016 *
I I I I
A(-0.00S
23
82x R 0.98432
-- y = 1.0038 eA(-0.10086x) R 0.99095
- y = 1.0466' e^(0.00027271x) R= 0.24034
~j--48.4 "-4- r
1
C3
0)
Ulm e!
1%
10%
0.0
, , , I I ,
Figure 5.4. Effect of collagenase and relaxin on tumor collagen dynamics.
(a) SHG noninvasively quantified enzymatic degradation of tumor collagen in vivo.
Different concentrations of collagenase were used to degrade collagen of Mu89 melanomas in the dorsal skinfold chamber. Loss of SHG signal, monitored after application, showed a single exponential decay time that varied linearly with collagenase concentration. No change in fiber length occurred while the fibers were visible.
105

_-
_!_
]
|1
L i _
J | , , ;
| .
|.
_
_
Figure 5.4. Effect of collagenase and relaxin on tumor collagen dynamics.
(b) SHG dynamically imaged the effects of chronic relaxin treatment. Five maximumintensity projections of the same region of collagen fibers in a Mu89 tumor were obtained every 3 d after the beginning of relaxin treatment, showing the effect of chronic relaxin treatment on the structure and brightness of preexisting fibers. The original width of each region was 70 pin. A representative region from a placebo treated mouse is shown below the relaxin treated image. for comparison.
106

4 1)
I .x
1
C 0.8
"O
0 0.6
z
E
0.4
L.
0.2
n
-2 0 2 4 6 8 10 12 14 16 18
Figure 5.4. Effect of collagenase and relaxin on tumor collagen dynamics.
(c) Chronic relaxin treatment altered the characteristic length of collagen fibers. The average lengths of collagen fibers in the relaxin group (8 mice, 77 fibers) and in controls
(8 mice, 103 fibers) were monitored for 12 d during relaxin treatment. The average fiber length decreased in each case, and relaxin treatment induced a significantly greater decrease than no relaxin (control; placebo; P < 0.05).
Application of bacterial collagenase was 'proof of principle' that SHG imaging allowed dynamic monitoring of collagen modification in vivo. We next used this technique to test relaxin, an agent that can be used in a clinical setting[14]. The nontoxic hormone relaxin
107

is secreted by women during pregnancy to induce upregulation of various matrixdegrading enzymes[ 15] known as matrix metalloproteinases. Consequently, we hypothesized that chronic relaxin treatment would degrade the tumor matrix and improve macromolecular diffusion in tumors.
To test this, we implanted relaxin-loaded osmotic pumps in immunodeficient mice bearing 2-week-old human HSTS26T sarcomas in dorsal skinfold chambers. We then obtained images of tumors for 12 d using SHG. The average SHG signal of both the relaxin-treated and control mice did not change significantly from the first to the last day of the experiment (P > 0.05; n = 8 for both). Furthermore, the average SHG signal of relaxin-treated mice on day 12 was not significantly different from that of control mice; the mean pixel count (normalized to 1 at day 0) was 1.77 + 0.28 for treated mice and 1.35
± 0.16 for control mice (P > 0.05; n = 8 for both).
Average lengths of preexisting collagen fibers underwent a statistically significant decrease from the first day to the last day of the experiment in both groups (P < 0.05; n =
8 for both; Fig. 5.4.b,c). This decrease was significantly greater in the relaxin-treated mice than in control mice (P < 0.05; n = 8 for both).
Finally, the SHG signal generated by individual pre-existing fibers decreased in the relaxin-treated mice (P < 0.05; n = 14), whereas in the control mice the SHG signal did not change significantly (P > 0.05, n = 14). The SHG signal in the relaxin-treated mice on
108

day 12 was significantly smaller than that of control mice (P < 0.05; 0.66 ± 0.07 versus
1.3 0.10; n = 14 fibers each, normalized to 1 at day 0).
We treated a cohort of HSTS26T-bearing mice with relaxin and assessed the evolution of diffusive transport with fluorescence recovery after photobleaching (FRAP) after treatment. We found statistically significant increases in the diffusion coefficients of IgG and dextran 2,000,000 MW: IgG, 13.5 ± 5.6 x 10
-8 cm
2
/s in relaxin-treated (n = 6) and 7.5
4:
2.5 x 10 -8 cm 2 /s in control (n = 4), P < 0.005; dextran 2,000,000 MW, 5.7 ± 1.5 x 10-
9 cm
2
/s in relaxin-treated (n = 6) and 2.0 ± 1.0 times 10-9 cm2/s in control (n = 5), P < 0.05.
5.4 Discussion
These data showed that over 2 weeks in control mice, neither the SHG signal averaged over large regions of the tumor nor the brightness of individual pre-existing fibers changed significantly. However, the average length of pre-existing fibers decreased slightly. This represented the normal dynamic equilibrium of collagen turnover in a tumor, in which old fibers are degraded through shortening and cleavage but the overall level of collagen remains the same because of the appearance of new fibers. When mice were treated with relaxin for 2 weeks, the average SHG signal remained the same as in control mice, but the brightness and average length of individual pre-existing fibers decreased significantly more than in control mice. Furthermore, the diffusion coefficients of probe molecules increased, indicating that at a scale below optical resolution, the matrix was loosened and hindrance decreased. In vivo, matrix-degrading enzymes (matrix
109

metalloproteinases) consist of both free and membrane-bound forms[ 16]. The membranebound forms are bound to cells such as fibroblasts. The decrease in brightness of preexisting fibers indicated an increase in activity of free matrix metalloproteinases, analogous to administration of free bacterial collagenase. The decrease in end-to-end length of individual fibers, which was caused in part by localized cleavage of entire fibers and loss of material from their visible ends, was consistent with fibroblast upregulation of matrix metalloproteinases, which would locally cleave the fiber adjacent to the fibroblasts. The decrease in end-to-end length was also caused by alterations in fiber curvature and branching, consistent with an overall increase in the activity of fibroblasts, which modify their surrounding matrix by tugging and repositioning fibers[ 17]. The fact that the dynamic processes of matrix degradation and alteration were apparently upregulated by relaxin administration, combined with the maintenance of the equilibrium
SHG signal, indicated that matrix production processes may also have been upregulated in tumors. This has been seen in the uteri of mice[18] and rats[19] during relaxin treatment. The new tumor matrix created by the acceleration of both degradation and production, while the same equilibrium level of collagen was maintained, had a more porous structure and hence weaker diffusive hindrance.
In conclusion, SHG imaging in tumors allowed visualization of fibrillar collagen structure in vivo; this may lead to simple noninvasive estimation of drug accessibility in tumors. Furthermore, SHG imaging allowed the noninvasive measurement of enzymatic modification of tumor collagen. Finally, relaxin chronically decreased diffusive hindrance in tumors, and we used SHG imaging to determine how this hormone altered
110

the tumor matrix, offering insight into the mechanisms by which it decreased diffusive hindrance. This could provide a useful new technology to evaluate strategies for altering the tumor extracellular matrix, to increase or decrease diffusive resistance, as studied here. Although the device at present is suitable for tumors that grow in optically accessible locations, with the availability of small hand-held laser-scanning microscopesl[20] this approach could be adapted for use in clinics in the near future.
111

5.5 References
1. Netti, P.A., D.A. Berk, M.A. Swartz, A.J. Grodzinsky, and R.K. Jain, 2000. "Role of extracellular matrix assembly in interstitial transport in solid tumors." Cancer
Res, 60(9): p. 2497-503.
2.
3.
4.
5.
6.
7.
8.
Pluen, A., et al., 2001. "Role of tumor-host interactions in interstitial diffusion of macromolecules: cranial vs. subcutaneous tumors." Proceedings of the National
Academy of Sciences of the United States of America, 98(8): p. 4628-33.
Ramanujan, S., A. Pluen, T.D. McKee, E.B. Brown, Y. Boucher, and R.K. Jain,
2002. "Diffusion and convection in collagen gels: implications for transport in the tumor interstitium." Biophys J, 83(3): p. 1650-60.
Williams, R.M., W.R. Zipfel, and W.W. Webb, 2001. "Multiphoton microscopy in biological research." Curr Opin Chem Biol, 5(5): p. 603-8.
Zoumi, A., A. Yeh, and B.J. Tromberg, 2002. "Imaging cells and extracellular matrix in vivo by using second-harmonic generation and two-photon excited fluorescence." Proc Natl Acad Sci U SA, 99(17): p. 11014-9.
Campagnola, P.J., A.C. Millard, M. Terasaki, P.E. Hoppe, C.J. Malone, and W.A.
Mohler, 2002. "Three-dimensional high-resolution second-harmonic generation imaging of endogenous structural proteins in biological tissues." Biophys J, 82(1
Pt 1): p. 493-508.
Freund, I., M. Deutsch, and A. Sprecher, 1986. "Connective tissue polarity.
Optical second-harmonic microscopy, crossed-beam summation, and small-angle scattering in rat-tail tendon." Biophys J, 50(4): p. 693-712.
Moreaux, L., O. Sandre, and J. Mertz, 2000. "Membrane imaging by secondharmonic generation microscopy." Journal of the Optical Society of America B:
9.
Optical Physics, 17(10): p. 1685-1694.
Brown, E.B., R.B. Campbell, Y. Tsuzuki, L. Xu, P. Carmeliet, D. Fukumura, and
R.K. .lain, 2001. "In vivo measurement of gene expression, angiogenesis and physiological function in tumors using multiphoton laser scanning microscopy."
Nat Med, 7(7): p. 864-8.
10. Jain, R.K., L.L. Munn, and D. Fukumura, 2002. "Dissecting tumour pathophysiology using intravital microscopy." Nat Rev Cancer, 2(4): p. 266-76.
11. Fleischmajer, R., J.S. Perlish, E.D. MacDonald, 2nd, A. Schechter, A.D.
Murdoch, R.V. Iozzo, and Y. Yamada, 1998. "There is binding of collagen IV to beta 1 integrin during early skin basement membrane assembly." Ann N YAcad
Sci, 857: p. 212-27.
12. Agarwal, A., M.L. Coleno, V.P. Wallace, W.Y. Wu, C.H. Sun, B.J. Tromberg, and S.C. George, 2001. "Two-photon laser scanning microscopy of epithelial cellmodulated collagen density in engineered human lung tissue." Tissue Eng, 7(2): p.
191-202.
13. Stoller, P., K.M. Reiser, P.M. Celliers, and A.M. Rubenchik, 2002. "Polarizationmodulated second harmonic generation in collagen." Biophys J, 82(6): p. 3330-
42.
112

14. Seibold, J.R., et al., 2000. "Recombinant human relaxin in the treatment of scleroderma. A randomized, double-blind, placebo-controlled trial." Ann Intern
Med, 132(11): p. 871-9.
15. Unemori, E.N. and E.P. Amento, 1990. "Relaxin Modulates Synthesis and
Secretion of Procollagenase and Collagen By Human Dermal Fibroblasts."
Journal of Biological Chemistry, 265(18): p. 10681-10685.
16. Egeblad, M. and Z. Werb, 2002. "New functions for the matrix metalloproteinases in cancer progression." Nat Rev Cancer, 2(3): p. 161-74.
17. Grinnell, F., 2000. "Fibroblast-collagen-matrix contraction: growth-factor signalling and mechanical loading." Trends Cell Biol, 10(9): p. 362-5.
18. Bylander, J.E., E.H. Frieden, and W.C. Adams, 1987. "Effects of porcine relaxins upon uterine hypertrophy and protein metabolism in mice." Proc Soc Exp Biol
Med, 185(1): p. 76-80.
19. Frieden, E.H. and W.C. Adams, 1985. "Stimulation of rat uterine collagen synthesis by relaxin." Proc Soc Exp Biol Med, 180(1): p. 39-43.
20. Helmchen, F., M.S. Fee, D.W. Tank, and W. Denk, 2001. "A miniature headmounted two-photon microscope. high-resolution brain imaging in freely moving animals." Neuron, 31(6): p. 903-12.
113

6.1 Introduction
Oncolytic vectors are mutant viruses that replicate in tumor cells preferentially over normal cells and have shown promise in the treatment of various preclinical tumor modelsl 1, 2]. Oncolytic viral therapy employs a novel method of tumor destruction mediated by viral replication and selective lysis of cancer cells[3, 41. The creation of more oncolytic virus by infected tumor cells and infectious spread from one cell to the next result in improved performance over more passive forms of therapeutic delivery[5,
6]. Early phase human clinical trials of G207, an oncolytic HSV vector, for the treatment of recurrent malignant glioblastomas have demonstrated both safety and efficacy[7].
However, the inability to efficiently propagate throughout the tumor and infect cells distant from the injection site limits the capacity of oncolytic viruses to achieve consistent therapeutic responses[8].
Viral therapeutics are orders of magnitude larger than traditional chemotherapeutics, and thus may encounter transport limitations not associated with those drugs. Quantitative studies have yet to be performed to determine the interstitial barriers to viral transport. In this study, we show that fibrillar collagen, previously found
114

to be the major barrier to the transport of large molecules in the tumor interstitium[9- 11 , also limits viral distribution within tumors. Direct degradation of the fibrillar collagen network using collagenase improves viral distribution and leads to the improved oncolytic viral therapy of tumors.
E5 and E26 cells (from Dr. Neal DeLuca, University of Pittsburgh[ 12]) were maintained in DMEM growth medium supplemented with 200 tM L-glutamine (Invitrogen), 100 units/ml penicillin and 100 tg/ml streptomycin (Sigma, St. Louis, MO), and 10% fetal bovine serum (Sigma) under standard cell culture conditions.
The HSV-1 recombinant viruses used in this study were the replication defective mutant
Gal4 (ICP4-, lacZ+; from Dr. Neal DeLuca[13]) and MGH2 (ICP6-, Oc4.5', eGFP+; from
Dr. E. Antonio Chiocca and Dr. Yoshi Saeki; Tyminski et al., unpublished data). MGH2 is a replication conditional virus that is attenuated by deletion of two nonessential viral functions, the ICP6 gene encoding the large subunit of ribonucleotide reductase and
634.5, a product known to overcome impaired viral protein synthesis in neurons by inducing the dephosphorylation of eIf2c[14]. These deletions impair virus replication in non-dividing cells, but allow virus replication in tumor cells.
115

Gal4 and MGH2 stocks were propagated in E5 and E26 cells, respectively, which supply the HSV- 1 ICP4 protein (E5) or HSV ICP4 and ICP27 proteins (E26) in trans. To obtain GFP-labelled HSV particles, E5 and E26 cells were transfected with a plasmid
(pVP 16-GFP[15]) encoding the fusion protein VP16-GFP and infected with Gal4 and
MGH2, respectively. After purification and concentration the number of DNA-containing particles in each virus preparation was quantified by transduction assay counting lacZ- positive cells for Gal4 and GFP-positive cells for MGH2 virus.
Human melanoma Mu89 cells were grown in dorsal skinfold chambers in severe combined immunodeficient mice as described previously[16]. All animal experiments were done with the approval of the Institutional Animal Care and Use Committee.
For dorsal chamber tumor studies, HSV vectors labeled with VP16-GFP were mixed with either 0.2 Ftg/'tl bacterial collagenase (Sigma) or PBS, to a final titer of 106 t.u./l. One microliter of virus was loaded into a glass micropipette with a beveled tip measuring 25-
30 microns in diameter. The micropipette was connected to a Harvard syringe pump apparatus and the fluid was infused into the tumor at constant pressure (- I/l0min).
Images were obtained using a custom built multiphoton laser scanning microscope[17.
Images of the SHG signal[10] were obtained using a 435DF30 emission filter, and of the
GFP using a 525DF100 emission filter, with input excitation at 880 nm, and a high pass
475 dichroic filter to separate the two signals.
116

Multiphoton images were analyzed to determine the relative localization of collagen (red pixels) and injected particles (viral vectors, green pixels; dextran, blue pixels). Pixel intensities were spatially compared along lines drawn perpendicular to the periphery of virus containing regions. For both viral vectors and dextran, analysis was performed for 3 injections in separate tumors, and for 20 images taken at different depths in each tumor.
As the spatial variable, the pixel intensities were plotted as a function of the relative distance from the observed interface with fibrillar collagen.
The area of viral vector distribution following intratumoral injection was quantified as follows. A maximum intensity projection of 10 images was performed to create a single image for each injection site. An outline of the area of viral distribution was drawn on these images and the area was calculated with imaging software (ImageJ).
Human melanoma Mu89 cells were implanted subcutaneously in the flank of SCID mice, and allowed to reach an average volume of 100 mm
3
, at which point mice were of either PBS; 1.0 tg collagenase; 106 t.u. MGH2; or a mixture 106 t.u. MGH2 and 1.0 itg collagenase. A second equivalent injection was performed two days later. Mice were examined daily and tumor volume measured every 2-3 days. Tumor volume was calculated according to the formula volume = jTAB
2
/6, where A and B are the maximum and minimum diameters, respectively. Mice died from the natural progression of their
117

disease process, or were euthanized when (a) tumor mass exceeded a size of 2,000 mm
3 or (b) premorbid behavior (imminent death from lethargy, respiration depression, and/or severe weight loss) was noted.
Two days after viral and/or collagenase treatment of tumors, mice were euthanized and tumors were removed and snap frozen in liquid nitrogen. Frozen tissue was embedded in
OCT and sectioned such that a 10 tm section was kept every 300 tm throughout the tumor volume. Tissue sections were stained for tegument and envelope virion proteins using anti-HSV primary antibody (AB1125, Chemicon International) and as secondary anti-rabbit Alexa-conjugated antibody, counterstained for nuclei (DAPI), and imaged for
GFP expression and fluorescent markers using confocal microscopy.
Data are expressed as mean + SEM. Statistical significance between groups was determined by an unpaired Student t-test. Statistical analysis was performed using
StatView 4.51 software (Abacus Concepts Inc.). Differences were considered statistically significant for P < 0.05.
118

6.3 Results
In order to quantify virus distribution following direct injection, one microliter containing
106 viral transducing units (t.u.) of VP16-GFP labeled non-replicative HSV-1 (Gal4; kindly provided by Dr. Neal DeLuca[l131) virions (150nm in diameter) were directly injected into Mu89 human melanomas grown in dorsal skin windows in SCID mice.
Multiphoton imaging performed in vivo approximately 30 minutes following injection revealed that viral particles distributed mostly near the site of injection. Second harmonic generation (SHG) was used to simultaneously image fibrillar collagen at the injection site. Viral particles distributed primarily within collagen free areas of the tumor, with limited penetration into collagen rich regions (Figure 6.1.a). To quantify viral penetration, pixel counts of collagen (SHG) and virus (GFP) were measured in one image slice along lines drawn perpendicular to the periphery of virus containing regions (an example line is shown in Figure 6.1.a). Averaging over many lines revealed an inverse correlation between collagen and viral particles, such that a sharp decrease in virus signal corresponded to an increase in the amount of fibrillar collagen present (Figure 6.1.c).
While collagen has previously been shown to hinder the interstitial transport of macromolecules[9, 11], nearly complete exclusion to this extent has not been seen. In order to directly compare viral distribution in the interstitial matrix with another macromolecular tracer, Cascade-blue conjugated 2x10 6 molecular weight dextran tracer molecules (RH 20 nm) were co-injected with HSV vectors. While the dextran penetrated into collagen rich regions, viral particles were excluded (Figure 6. 1.b,d).
119

C
-n
150
.l d p,.tretIonl 9 copeaeun@In d t
' 9
"
15
2
I
I"e
'
5 sa
73 I
1
I,,
~135 a
-t .S-b -n a l diotsm ftom tce. pm o 0
-S 2S dmc cm Ittmo, m so
I
73
Figure 6.1. Viral vector distribution following intratumoral injection. (a,b) Multiphoton images of Mu89 melanomas 30 minutes after intratumoral injection of VP16-GFP labeled
Gal4 vectors (green), either alone (a) or with Cascade blue-conjugated dextran (blue) (b).
Second-harmonic generation (SHG) signal denotes fibrillar collagen (red pseudocolor). HSV vectors localized in extracellular spaces around individual tumor cells and were excluded by areas of intense SHG signal. In contrast, the smaller dextran tracer penetrated regions rich in fibrillar collagen.. (c,d) Relative localization of collagen and injected particles determined by pixel analysis. Spatial comparison of pixel intensities was performed for collagen (red pixels) and either viral particles (green pixels; c) or dextran (blue pixels; d). Analysis was performed along lines drawn perpendicular to the border of SHG signal, and mean values plotted. A representative image and line are shown for each case (a,b). Collagen and viral localization in the tumor are anti-correlative
120

r i
E 7-.+05 ua 6.E+05 -
F
*.;
L.
5.E+05 -
4.E+05 -
'. 3.E+05 -
C
._
_
,z :I
.,
C
1.E+05 -
_ O.E+00 -
X
T
Control injection
(n=10)
*
Collagenase injection
(n=4)
Figure 6.1 (e) Multiphoton image of viral vector distribution following co-injection with
collagenase. This area of distribution is outlined in blue for a representative injection of virus with collagnease, to the left. The image to the right is a representative image for vectors injected alone outlined in blue. (f) Comparison of the area of viral vector distribution following intratumoral injection. Areas measured from a maximum intensity projection of 10 images taken -3() minutes following injection. Collagenase co-injection resulted in a 3-fold increase in the area of viral distribution (P < 0.05).
121

In order to test whether disruption of the collagen network could improve the penetration of viral particles into the tumor, the viral vectors were co-injected with bacterial collagenase (0.2 tg/tl). Collagenase increased the area of viral distribution by 3-fold compared to control injections (Figure 6. 1.f). Rather than distributing into restricted regions bounded by fibrillar collagen, as in the case of injection without collagenase, vectors spread more uniformly from the injection site upon collagenase treatment (Figure
6. I .e, shown next to representative control injection).
The oncolytic virus MGH2 (kindly provided by Drs. E. Antonio Chiocca and Yoshi
Saeki, Ohio State University) has the same backbone as G207[14, 18], but carries GFP as a reporter gene instead of lacZ. MGH2 replicates in Mu89 melanoma cells in culture, resulting in GFP expression and cell lysis within 24-48 hours (data not shown). To test whether fibrillar collagen would also limit the spread of actively replicating viral particles within the tumor, 106 t.u. MGH2 were injected into Mu89 tumors grown in dorsal window chambers in SCID mice and GFP expression was imaged (Figure 6.2). As seen previously, the initial distribution of viral particles was limited by fibrillar collagen (data not shown). Twenty-four hours later, the area of infection was localized in only a small proportion (-15%) of the entire tumor mass, corresponding to the site of injection (Figure
6.2.b). Even 11 days following the initial injection of MGH2, viral vectors could not
122

penetrate sufficiently to infect the entire tumor mass (Figure 6.2.a). No significant treatment response was observed in any of the tumors injected with MGH2 alone.
In contrast, when the same amount of oncolytic virus was co-injected with collagenase (0.2 tg/tl), the initial viral distribution was greater relative to virus alone
(data not shown), and this translated into an improved area of tumor cell infection (Figure
6.2.b). Therapeutic response was observed in all four collagenase co-treated tumors, with nearly complete regression in two cases (Figure 6.2.b). Due to the time limitations in using this particular tumor window model, we were not able to monitor the mice for a longer period of time to follow tumor regression.
123

Figure 6.2. Effect of collagenase on oncolytic viral therapy. Mu89 melanomas implanted in the dorsal skinfold chamber of SCID mice were treated with the oncolytic vector MGH2 in combination with PBS (left panels) or collagenase (right panels). (a)
Fluorescent and b) brightfield and fluorescent microscopic images of tumors following a timecourse after injection with oncolytic virus. Infection of tumor cells was detected by expression of the reporter gene GFP (encoded by the virus). Co-injection of MGH2 and collagenase resulted in a greater distribution of infected cells, relative to injection of
MGH2 alone. At 11 days, nearly complete regression of the tumor (as evidenced by absence of tumor vasculature) was achieved with MGH2 and collagenase co-injection, while no significant change in volume was observed with MGH2 treatment alone.
Extent of tumor outlined in blue to guide the eye.
124

We then tested if the co-injection of collagenase and MGH2 would increase the therapeutic efficacy of MGH2 over longer time intervals. When Mu89 tumors growing in the flank of SCID mice reached 100 mm
3
, they were injected intratumorally with either 1 ytg collagenase, 106 t.u. MGH2, or both collagenase and MGH2, followed by similar injections two days later. As a control, tumors were injected with PBS alone. The time for the tumor to reach ten times the initial volume (mean ± SEM) was compared for each group (Figure 6.3). If the tumors failed to reach ten times the initial volume due to morbidity, the time to their last measurement was used as a conservative approximation of growth delay. Both collagenase treatment alone (19 ± 1 days) and MGH2 injection alone (27 ± 3 days) had no significant effect on tumor growth compared to PBS control
(24 ± 3 days) (P > 0.05, both cases). In the group treated with MGH2 alone, one tumor showed marked regression, but recurred after 10 days. However, co-injection of MGH2 with collagenase (50 ± 9 days) significantly delayed the growth of tumors compared to all other treatment groups (P < 0.05 for all cases). In this group two out of seven tumors failed to grow to 200 mm
3 even 60 days after treatment and apparently complete regression of the tumor was observed in another animal, although it recurred 20 days later.
125

4 A
12
.
10
)
8
U-
N
U)
0
I-
L 6
4
2
0 10 20 30 40 50 60
Days
Figure 6.3. Effect of collagenase on MGH2-induced tumor growth delay. Tumors were grown subcutaneously in the hind flank of SCID mice. When tumors reach -100 mm
3
, animals were divided into four groups (n = 6-7) and treated twice (day 0 and day 2) with 10 tl of PBS (green), collagenase (0. 1 g/tl) (black), MGH2 (106 t.u.) in PBS
(blue), or MGH2 (106 t.u.) and collagenase (0.1 Itg/Ltl) in PBS (red). Tumor volumes were measured every 2-3 days and the time to reach a given volume was expressed as mean ± SEM for each group. The time to reach ten times the initial volume was compared. There was no significant difference between PBS (23 ± 3 days) and either collagenase treatment alone (19 + 1 days) or MGH2 alone (27 + 3 days) (P > 0.05 for both cases). However, MGH2 and collagenase co-treatment induced a significant tumor growth delay (50 + 8 days) relative to all other groups (P < 0.05 for all cases).
126

In order to investigate the mechanism of improved efficacy, tumors were treated as before with MGH2, either alone or with collagenase, and analyzed two days after the second injection. To determine viral distribution, tissue sections were stained for structural virion proteins, counterstained for nuclei (DAPI), and imaged for GFP expression using confocal microscopy. As expected, the immunostaining revealed the presence of HSV virion particles within and surrounding cells expressing GFP (Figure
6.4.a,b). In tumors treated with MGH2 alone, virion particles and infected cells were distributed only along the 500 /m width needle track (data not shown). In contrast, for
MGH2 and collagenase treatment, a diffuse distribution of infected cells was observed throughout the entire tumor section, spanning an area of up to 3 x 7 mm (data not shown).
Imaging of sections at later times showed that virus was able to continue to spread within the tumor, but not within collagen containing areas at the edge of the tumor (data not shown).
127

Figure 6.4. Immunostaining analysis of tumor cell infection. Representative tissue sections of Mu89 flank tumors injected with either MGH2 alone (a) or MGH2 and collagenase (b), were labeled with anti-HSV antibodies. GFP expression from MGH2infected cells (green), HSV proteins detected with Alexa-conjugated secondary Ab (red), and nuclear stain DAPI (blue) are shown. Slightly more cells are stained for viral particles than those expressing GFP. This is probably due to either the delay between infection and expression or the overwhelming signal from new viral proteins synthesized in infected cells. In the absence of collagenase treatment, viral particles and infected cells are localized in dense clusters at the site of injection. Collagenase co-injection results in infection of tumor cells dispersed throughout the tumor.
The development of strategies to improve both the initial vector distribution within tumnors and the ability of these vectors to propagate through the entire tumor mass is critical to the success of oncolytic viral therapy[l 1. Our results demonstrate the important role that fibrillar collagen plays in regulating both of these processes, and ultimately in determining therapeutic efficacy. We have previously shown that fibrillar collagen is the major barrier to the transport of macromolecules through the extracellular matrix of tumors[9], an effect that increases with larger particle size[l 1. In the present study we
128

observed that whereas smaller tracers (2x106 MW dextran, RH-20 nm, as well as IgG,
RH-5 nm, data not shown) distributed relatively uniformly within the tumor following injection, the vast majority of HSV virions (150 nm in diameter) were located only in collagen-free areas. The absence of virus penetration into fibrillar collagen rich areas suggests that: the effective pore size cutoff of the collagen network is smaller than the size of viral particles. This finding has far-reaching implications: while many tumor models in rodents consist of fast growing tumors that lack a significant collagen network, many tumors in humans show extensive stromal infiltration with extracellular matrix and collagen deposition[19, 20]. Thus preclinical models must take in to account the deposition of extracellular matrix in order to properly mimic human disease.
Oncolytic vectors are thought to overcome some of the delivery issues faced by non-replicating viral vectors through their ability to propagate on site in tumors (thereby amplifying the input dose) and spread from tumor cell to tumor cell. However, we found that the collagen network restricted the distribution of the intratumorally injected oncolytic vector MGH2 and limited the area of tumor cell infection. Even several weeks after treatment, tumor cell infection remained confined to a small area and the tumor continued to grow (Figure 6.2.a, data not shown). Co-injection of MGH2 with collagenase resulted in a broad, uniform distribution of viral particles and infected cells
(Figure 6.2.b), with substantial tumor regression and improved efficacy in a flank tumor model. The dispersed distribution of virus following collagenase co-injection can lead to improved therapeutic outcome in several ways: (1) the increased initial virion distribution improves the chance that viral vectors can penetrate all regions of the tumor; (2) the occurrence of multiple infections of the same tumor cell decreases, while the number of
129

distinct tumor cell infections increases; and (3) once the virus replicates and lyses the cell it has infected, it has access to a greater number of previously uninfected neighboring cells. All together, these processes can lead to increased oncolytic activity, as shown schematically in Figure 6.5.
Control Injection
Initial virus distribution
See figure la, le
Initial cell infection
See figure 2c, 2d
First viral replication
See figure 4a, 4b
Secondary cell infection i ...
7 .
Figure 6.5. A representative model of improvement in oncolytic viral distribution
and tumor cell infection by collagenase treatment. Following direct intratumor injection, viral spread (red area) is limited by fibrillar collagen (red lines) and results in a cluster of infected cells (light green). The collagen network also restricts the distribution of subsequent viral progeny and tumor cell infection beyond the initial injection site is not achieved. In contrast, co-injection of virus with collagenase results in a more diffuse distribution of viral particles and a greater number of initially infected cells (light green).
Viral particles released by these cells have greater access to neighboring uninfected cells.
This process results in more widespread secondary infection (dark green) and ultimately greater therapeutic efficacy.
130

Researchers have developed other methods to try to overcome the limited distribution of oncolytic vectors in tumors[21]. One such method is the use of multiple injections, either on successive days or with fractionation of the initial dose at multiple sitesl22, 23]. However, in the absence of extracellular matrix-modification, the viral distribution at each individual injection site would still be limited by collagen fibers.
Indeed, a phase II trial with an oncolytic adenoviral vector showed limited improvement in efficacy even with daily injections that included fractionation[241. As an alternative to increasing viral distribution, combination therapy with either radiation or chemotherapy is often employed to improve oncolytic activity[25, 26]. Collagenase treatment is compatible with combination therapy and could further improve efficacy. Indeed, this may be a complementary therapy for combination with radiation, which can induce fibrosis and lead to an increase in interstitial collagen[27].
In the present study, however, it was also noted that intratumoral haemorrhages occurred in many of the tumors treated with collagenase. While bleeding from collagenase treatment alone did not affect tumor growth in either the dorsal chamber
(data not shown) or flank models, this phenomenon demonstrates the complex interactions between the extracellular matrix and cells within the tumor, including both tumor cells and host endothelial cells. Furthermore, it is possible that collagenase treatment of tumors may increase the risk of metastasis. The development of this matrixmodulating technique for clinical applications may require the use of specific matrix proteases, such as MMP-8, which degrades collagen and decreases metastasis[28, 291.
131

In conclusion, we determined that even with the on-site generation of viral particles provided by the replication-competent nature of oncolytic viruses, there still exist barriers, namely the collagen network, that are sufficient to prevent viral spread throughout the entire tumor. Disruption of the collagen network within tumors leads to an increase in both initial vector distribution and subsequent propagation of virus through the tumor mass, resulting in significantly improved therapeutic outcome. This is a powerful result since it applies to all viral particles and gene delivery stratagies, as well as imaging systems involving the use of nano-particulates[30] - as all run into the problem of insufficient delivery to the target cells. Furthermore, the method of modification can be versatile: any technique that decreases the collagen content of tumors would be useful. These findings open the way for increasing the potency of gene therapy in cancer and other diseases.
132

6.5 References
1.
2.
3.
Everts, B. and H.G. van der Poel, 2005. "Replication-selective oncolytic viruses in the treatment of cancer." Cancer Gene Ther, 12(2): p. 141-61.
Martuza, R.L., A. Malick, J.M. Markert, K.L. Ruffner, and D.M. Coen, 1991.
"Experimental therapy of human glioma by means of a genetically engineered virus mutant." Science, 252(5007): p. 854-6.
Kirn, D., R.L. Martuza, and J. Zwiebel, 2001. "Replication-selective virotherapy for cancer: Biological principles, risk management and future directions." Nat
4.
5.
6.
7.
8.
Med, 7(7): p. 781-7.
Chiocca, E.A., 2002. "Oncolytic viruses." Nat Rev Cancer, 2(12): p. 938-50.
Lee, C.T., et al., 2004. "Combination therapy with conditionally replicating adenovirus and replication defective adenovirus." Cancer Res, 64(18): p. 6660-5.
Ichikawa, T. and E.A. Chiocca, 2001. "Comparative analyses of transgene delivery and expression in tumors inoculated with a replication-conditional or defective viral vector." Cancer Res, 61(14): p. 5336-9.
Markert, J.M., et al., 2000. "Conditionally replicating herpes simplex virus mutant, G207 for the treatment of malignant glioma: results of a phase I trial."
Gene Ther, 7(10): p. 867-74.
Harrison, D., H. Sauthoff, S. Heitner, J. Jagirdar, W.N. Rom, and J.G. Hay, 2001.
"Wild-type adenovirus decreases tumor xenograft growth, but despite viral persistence complete tumor responses are rarely achieved--deletion of the viral
Elb-19-kD gene increases the viral oncolytic effect." Hum Gene Ther, 12(10): p.
1323-32.
9. Netti, P.A., D.A. Berk, M.A. Swartz, A.J. Grodzinsky, and R.K. Jain, 2000. "Role of extracellular matrix assembly in interstitial transport in solid tumors." Cancer
Res, 60(9): p. 2497-503.
10. Brown, E., T. McKee, E. diTomaso, A. Pluen, B. Seed, Y. Boucher, and R.K.
Jain, 2003. "Dynamic imaging of collagen and its modulation in tumors in vivo using second-harmonic generation." Nat Med, 9(6): p. 796-800.
11. Pluen, A., et al., 2001. "Role of tumor-host interactions in interstitial diffusion of macromolecules: cranial vs. subcutaneous tumors." Proc Natl Acad Sci U S A,
98(8): p. 4628-33.
12. Samaniego, L.A., A.L. Webb, and N.A. DeLuca, 1995. "Functional interactions between herpes simplex virus immediate-early proteins during infection: gene expression as a consequence of ICP27 and different domains of ICP4. " J Virol,
69(9): p. 5705-15.
13. Grondin, B. and N. DeLuca, 2000. "Herpes simplex virus type 1 ICP4 promotes transcription preinitiation complex formation by enhancing the binding of TFIID to DNA. " J Virol, 74(24): p. 11504-10.
14. Kramm, C.M., et al., 1997. "Therapeutic efficiency and safety of a secondgeneration replication-conditional HSV 1 vector for brain tumor gene therapy."
Hum Gene Ther, 8(17): p. 2057-68.
15. Bearer, E.L., X.O. Breakefield, D. Schuback, T.S. Reese, and J.H. LaVail, 2000.
"Retrograde axonal transport of herpes simplex virus: evidence for a single
133

mechanism and a role for tegument. " Proc Natl Acad Sci U S A, 97(14): p. 8146-
50.
16. Leunig, M., F. Yuan, M.D. Menger, Y. Boucher, A.E. Goetz, K. Messmer, and
R.K. Jain, 1992. "Angiogenesis, microvascular architecture, microhemodynamics, and interstitial fluid pressure during early growth of human adenocarcinoma
LS 174T in SCID mice." Cancer Res, 52(23): p. 6553-60.
17. Brown, E.B., R.B. Campbell, Y. Tsuzuki, L. Xu, P. Carmeliet, D. Fukumura, and
R.K. Jain, 2001. "In vivo measurement of gene expression, angiogenesis and physiological function in tumors using multiphoton laser scanning microscopy."
Nat Med, 7(7): p. 864-8.
18. Mineta, T., S.D. Rabkin, T. Yazaki, W.D. Hunter, and R.L. Martuza, 1995.
"Attenuated multi-mutated herpes simplex virus-i for the treatment of malignant gliomas. " Nat Med, 1(9): p. 938-43.
19. Elenbaas, B. and R.A. Weinberg, 2001. "Heterotypic signaling between epithelial tumor cells and fibroblasts in carcinoma formation." Exp Cell Res, 264(1): p. 169-
84.
20. Martinez-Hernandez, A., 1988. "The extracellular matrix and neoplasia. " Lab
Invest, 58(6): p. 609-12.
21. Jia, W. and Q. Zhou, 2005. "Viral vectors for cancer gene therapy: viral dissemination and tumor targeting." Curr Gene Ther, 5(1): p. 133-42.
22. Kirn, D., 2001. "Clinical research results with d11520 (Onyx-015), a replicationselective adenovirus for the treatment of cancer: what have we learned?" Gene
Ther, 8(2): p. 89-98.
23. Currier, M.A., L.C. Adams, Y.Y. Mahller, and T.P. Cripe, 2005. "Widespread intratumoral virus distribution with fractionated injection enables local control of large human rhabdomyosarcoma xenografts by oncolytic herpes simplex viruses."
Cancer Gene Ther.
24. Nemunaitis, J., et al., 2001. "Phase II trial of intratumoral administration of
ONYX-015, a replication-selective adenovirus, in patients with refractory head and neck: cancer." J Clin Oncol, 19(2): p. 289-98.
25. Khuri, F.R., et al., 2000. "a controlled trial of intratumoral ONYX-015, a selectively-replicating adenovirus, in combination with cisplatin and 5fluorouracil in patients with recurrent head and neck cancer." Nat Med, 6(8): p.
879-85.
26. Kim, S.H., et al., 2005. "Combination of mutated herpes simplex virus type 1
(G207 virus) with radiation for the treatment of squamous cell carcinoma of the head and neck." Eur J Cancer, 41(2): p. 313-22.
27. Znati, C.A., et al., 2003. "Irradiation reduces interstitial fluid transport and increases the collagen content in tumors." Clin Cancer Res, 9(15): p. 5508-13.
28. Agarwal. D., S. Goodison, B. Nicholson, D. Tarin, and V. Urquidi, 2003.
"Expression of matrix metalloproteinase 8 (MMP-8) and tyrosinase-related protein-i (TYRP-1) correlates with the absence of metastasis in an isogenic human breast cancer model." Differentiation, 71(2): p. 114-25.
29. Montel, V., J. Kleeman, D. Agarwal, D. Spinella, K. Kawai, and D. Tarin, 2004.
"Altered metastatic behavior of human breast cancer cells after experimental
134

manipulation of matrix metalloproteinase 8 gene expression." Cancer Res, 64(5): p. 1687--94.
30. Ferrari, M., 2005. "Cancer nanotechnology: opportunities and challenges." Nat
Rev Cancer, 5(3): p. 161-171.
135

7.1 Introduction
In this thesis I have investigated the transport of macromolecules, liposomes and gene therapeutic particles within tumors, in order to determine how to improve delivery of these agents to tumor cells. Here I will summarize the conclusions and propose future directions for the two following areas of my thesis: 1) Understanding the composition and transport properties of the tumor extracellular matrix, and 2) A comparison of the matrix modifying treatments used in this study, with the goal of considering practical approaches to improving therapeutic transport within tumors.
136

The tumor interstitial matrix is a complex mixture of glycosaminoglycans, collagens, and associated proteoglycans whose formation depends on the interactions between the tumor cells and host cells present in or invading into the tumor[ 1-4]. The tumor matrix is highly heterogeneous in its composition, which varies depending on the interaction between the particular tumor cells and the host within which it is growing. For this reason, it is important to investigate tumor models that are orthotopic - implanted into the same host tissue from which the tumor initially arose[5]. The tumor models HSTS26T, a soft tissue sarcoma, and Mu89, a melanoma, have been used repeatedly in my work because they are both orthotopic to the subcutaneous tissue present in the dorsal chambers used in my studies. The tumor matrix is also heterogeneous due to the amorphous character of tumor cell growth, an uncontrolled and highly disordered process in which solid stresses imposed by the cancer cells has been shown to influence tumor blood flow[6], and extracellular matrix production[7]. As such, it can be difficult to relate analytical models of transport to the diffusion coefficients observed within the tumor interstitial matrix. We have attempted to quantify the composition of the interstitial matrix in this work in order to have a more accurate picture of at least the overall composition of the tumor extracellular matrix. The tumor interstitial matrix was analyzed using biochemical means by Netti et al.,[8] (see figure 1.1), indicating that collagen, detected by the presence of hydroxyproline within tissues, is more abundant on a per-mass basis than glycosaminoglycans. This does not necessarily point to any conclusions regarding collagen's influence on transport in vivo, however, since small amounts of
137

glycosaminoglycan chains (such as hyaluronan) have the capacity to adsorb large quantities of water, and as such have been shown to affect interstitial fluid flow within tissues[9- 11 ]. What is more indicative of the influence of collagen on interstitial diffusion is the observed correlation between higher collagen content, and reduced diffusion coefficients of IgG within the extracellular space[8]. This was confirmed for a broad range of macromolecular tracers in the work presented in chapter 3. The concept of cell tortuosity was introduced to be able to separate geometric effects (imposed by tumor cells on tracer molecules) from viscous effects (imposed by the tumor extracellular matrix on tracer molecules)[12, 13]. While correcting for the presence of cells is necessary to relate diffusion within model gels to diffusion within the tumor matrix, as mentioned in chapters 3 and 4, the division of transport into simply two components, namely geometric and viscous, may not be sufficient in some cases, for example for molecules that approach the size of the interfibrillar space between collagen fibrils. For molecules significantly below this size, the collagen fiber would appear as any other part of the matrix would, as a purely viscous barrier to diffusion. However, for large molecules that approach or exceed this size, the collagen fiber would then appear as an impenetrable (or highly impermeable) object, shifting its definition to a geometric barrier to diffusion. This was apparently the case when we were imaging the viral distribution within the tumor - the collagen network appeared to be significantly impermeable to the viral particles. The viral particle diameter was measured as approximately 300nm in size using laser scattering techniques, indeed larger than the interfibrillar space. To investigate the topic of the amount of extracellular matrix space available to tracer particles of different sizes, one potential study to perform would be the co-injection of a
138

number of different particles of varying sizes, labeled with separate distinguishable fluorophores. Quantum dots, which are semiconductor nanoparticles, offer great promise in this regard, as it is possible to tune the wavelength of these particles such that it is possible to distinguish many more individual populations of particles than are possible using conventional fluorophores. The use of these tracer particles would permit the simultaneous imaging of the exclusion of large molecules and the penetration of small molecules into collagen fibers, which could be imaged using second harmonic generation. This would allow a more direct determination of the "pore size" of the collagen within the tumor interstitial matrix. While the search for the description of a definitive "pore size" associated with the tumor interstitium, or components of the tumor interstitium such as collagen, is promising, the argument also exists that molecular motion will always be possible along certain preferential pathways within the tumor matrix. An alternate way to describe the reduction of movement through the tumor interstitium due to the gradual exclusion of accessible volume is to describe the diffusion as occurring on a fractal substrate, which is equivalent to describing the diffusion as falling into the regime of anomalous subdiffusion. This mechanism of transport has been described for an increasing number of gel[14, 15] and drug delivery systems[16], and is frequently mentioned when referring to the transport of molecules within the crowded cytoplasm of cells[17-19]. While anomalous subdiffusion has been characterized in these systems, it remains to be seen whether this is the case in the tumor interstitium, as we currently do not have enough data to be able to either confirm or deny this mechanism of transport.
139

Despite the complexities of understanding transport within the tumor interstitial matrix, the attempts we have made to modify the content or structure of collagen within the tumor interstitial matrix have resulted in increased transport of tracer molecules or viral particles. Direct application of bacterial collagenase to tumors by Netti et al.,[8] resulted in an increase in the diffusion coefficients of IgG in HSTS tumors of 100% after 24 hours. Relaxin treatment significantly increased the diffusion coefficients of both IgG and dextran 2,000,000 MW in HSTS and Mu89 tumors. And finally application of bacterial collagenase simultaneously with viral therapy resulted in improved penetration of viral particles within the tumor mass.
While these results are promising for the improved delivery of macromolecules, particles and gene vectors to tumors, at the same time care must be taken to consider the effects these matrix modifying treatments themselves will have on the tumor progression.
As mentioned previously, many matrix metalloproteases have been implicated in tumor invasion and metastatic progression[20-22]. In fact, a number of inhibitors of matrix metalloproteases have been developed with the hope of being able to use them clinically to prevent tumor invasion. While these inhibitors of matrix metalloproteases have generally done poorly in clinical trials[23], nevertheless the dogma in the field has been to inhibit rather than encourage matrix degradation. However, this is not universally the case - while there are some matrix metalloproteases, such as MMP-2 and MMP-9, that have been shown to increase the invasiveness of cancer cells[24], and have been implicated in increasing the metastatic progression of tumors, there are also others, such as MMP-8, that have been shown to both degrade collagen and decrease metastasis[25,
140

26]. The clinical application of this research should take care in choosing the appropriate matrix modifying therapy in order to improve penetration of a therapeutic agent, while not promoting increased invasiveness of the tumor. For example, relaxin has also been shown to increase angiogenesis and the expression of angiogenic factors in a number of different animal and cell culture models[27, 28]. Thus, care must be taken to ensure that the positive effects of relaxin, in improving the transport of macromolecular therapies to tumors, do enough good to justify the use of a potentially pro-angiogenic molecule in tumor therapy. While bacterial collagenase is a highly nonspecific enzyme, and is thus unlikely to increase metastasis in the same way that human-derived MMPs would, it would nevertheless be important to test whether bacterial collagenase would have an effect on the metastatic progression of cancer. Bacterial collagenase is also capable of inducing an immune response, which would reduce its effectiveness after administration of multiple doses. A future direction to more effectively improve the distribution of viral vectors, especially oncolytic viral vectors, within tumors, would be to insert a gene encoding for the matrix modifying treatment of choice, for example MMP-8, into the gene therapeutic vector. Thus, as the virus would infect cells and lyse them, it would simultaneously express the matrix modifying enzyme, which would reciprocally aid in the spread of the viral particles to more of the tumor mass.
141

7.4 Conclusions
In conclusion, in chapter 3, we measured diffusion coefficients of macromolecules and liposomes in tumors growing in cranial windows (CWs) and dorsal chambers (DCs) by fluorescence recovery after photobleaching. For the same tumor types, diffusion of large molecules was significantly faster in CW than in DC tumors. The greater diffusional hindrance in DC tumors was correlated with higher levels of collagen type I and its organization into fibrils. For molecules with diameters comparable to the interfibrillar space the diffusion was 5- to 10-fold slower in DC than in CW tumors. I. The slower diffusion in DC tumors was associated with a higher density of host stromal cells that synthesize and organize collagen type I.
In chapter 4, diffusion coefficients of tracer molecules in collagen type I gels prepared from 0-4.5% w/v solutions were measured by fluorescence recovery after photobleaching. When adjusted to account for in vivo tortuosity, diffusion coefficients in gels matched previous measurements in human tumor xenografts with equivalent collagen concentrations. In contrast, hyaluronan solutions hindered diffusion to a lesser extent when prepared at concentrations equivalent to those reported in these tumors.
Collagen permeability, determined from flow through gels under hydrostatic pressure, was compared with predictions obtained from application of the Brinkman effective medium model to diffusion data. Permeability predictions matched experimental results at low concentrations, but underestimated measured values at high concentrations.
Permeability measurements in gels did not match previous measurements in tumors.
Visualization of gels by transmission electron microscopy and light microscopy revealed
142

networks of long collagen fibers at lower concentrations along with shorter fibers at high concentrations. Negligible assembly was detected in collagen solutions pregelation.
However, diffusion was similarly hindered in pre and postgelation samples. Comparison of diffusion and convection data in these gels and tumors suggests that collagen may obstruct diffusion more than convection in tumors.
In chapter 5, we show that it is possible to optically image fibrillar collagen in tumors growing in mice using second-harmonic generation (SHG). Using this noninvasive technique, we estimated relative diffusive hindrance, quantified the dynamics of collagen modification after pharmacologic intervention and provided mechanistic insight into improved diffusive transport induced by the hormone relaxin.
And in chapter 6, we show that the spread of oncolytic viral vectors within tumors is limited by the fibrillar collagen in the extracellular matrix. Thus, tumor cells in inaccessible regions continue to grow, remaining out of the range of viral infection, and tumor eradication cannot be achieved. Matrix modification with bacterial collagenase upon initial virus injection results in a significant improvement in the range of viral distribution within the tumor. This results in an extended range of infected tumor cells, and improved virus propagation, ultimately leading to enhanced therapeutic outcome.
Thus, in this work we have shown that fibrillar collagen is an important barrier to the macrmolecular transport and viral distribution within tumors, and matrix modifying treatments that degrade the fibrillar collagen within tumors can significantly enhance the
143

penetration of large molecular therapeutics, ultimately resulting in an improved therapeutic response. These findings have significant implications for drug delivery in tumors and for tissue engineering applications.
144

7.5 References
1.
2.
3.
4.
5.
6.
Clarijs, R., D.J. Ruiter, and R.M. De Waal, 2003. "Pathophysiological implications of stroma pattern formation in uveal melanoma." J Cell Physiol,
194(3): p. 267-71.
Tuxhorn, J.A., G.E. Ayala, and D.R. Rowley, 2001. "Reactive stroma in prostate cancer progression." J Urol, 166(6): p. 2472-83.
Pluen, A., et al., 2001. "Role of tumor-host interactions in interstitial diffusion of macromolecules: cranial vs. subcutaneous tumors." Proceedings of the National
Academy of Sciences of the United States of America, 98(8): p. 4628-33.
Micke, P. and A. Ostman, 2004. "Tumour-stroma interaction: cancer-associated fibroblasts as novel targets in anti-cancer therapy?" Lung Cancer, 45 Suppl 2: p.
S 163 -75.
Killion, J.J., R. Radinsky, and I.J. Fidler, 1998. "Orthotopic models are necessary to predict therapy of transplantable tumors in mice." Cancer Metastasis Rev,
17(3): p. 279-84.
Padera, T.P., B.R. Stoll, J.B. Tooredman, D. Capen, E. di Tomaso, and R.K. Jain,
2004. "Pathology: cancer cells compress intratumour vessels." Nature, 427(6976):
7. p. 695.
Koike, C., et al., 2002. "Solid stress facilitates spheroid formation: potential involvement of hyaluronan." Br J Cancer, 86(6): p. 947-53.
8. Netti, P.A., D.A. Berk, M.A. Swartz, A.J. Grodzinsky, and R.K. Jain, 2000. "Role of extracellular matrix assembly in interstitial transport in solid tumors." Cancer
Res, 60(9): p. 2497-503.
9. Flessner, M.F., 2001. "The role of extracellular matrix in transperitoneal transport of water and solutes. " Perit Dial Int, 21 Suppl 3: p. S24-9.
10. Coleman, P.J., D. Scott, R.M. Mason, and J.R. Levick, 2000. "Role of hyaluronan chain length in buffering interstitial flow across synovium in rabbits." J Physiol,
526 Pt 2: p. 425-34.
11. Bert, J. and R.K. Reed, 1998. "Hyaluronan, hydration and flow conductivity of rat dermis." Biorheology, 35(3): p. 211-9.
12. Sykova, E., 2004. "Extrasynaptic volume transmission and diffusion parameters of the extracellular space." Neuroscience, 129(4): p. 861-76.
13. Nicholson, C., 2005. "Factors governing diffusing molecular signals in brain extracellular space." J Neural Transm, 112(1): p. 29-44.
14. Masuda, A., K. Ushida, and T. Okamoto, 2005. "New fluorescence correlation spectroscopy enabling direct observation of spatiotemporal dependence of diffusion constants as an evidence of anomalous transport in extracellular matrices." Biophys J, 88(5): p. 3584-91.
15. Fatin-Rouge, N., K. Starchev, and J. Buffle, 2004. "Size effects on diffusion processes within agarose gels." Biophys J, 86(5): p. 2710-9.
16. Saltzman, W.M. and R. Langer, 1989. "Transport rates of proteins in porous materials with known microgeometry. " Biophys J, 55(1): p. 163-71.
145

17. Tolic-Norrelykke, I.M., E.L. Munteanu, G. Thon, L. Oddershede, and K. Berg-
Sorensen, 2004. "Anomalous diffusion in living yeast cells." Phys Rev Lett, 93(7): p. 078102.
18. Wong, .Y., M.L. Gardel, D.R. Reichman, E.R. Weeks, M.T. Valentine, A.R.
Bausch, and D.A. Weitz, 2004. "Anomalous diffusion probes microstructure dynamics of entangled F-actin networks." Phys Rev Lett, 92(17): p. 178101.
19. Weiss, M., M. Elsner, F. Kartberg, and T. Nilsson, 2004. "Anomalous subdiffusion is a measure for cytoplasmic crowding in living cells." Biophys J,
87(5): p. 3518-24.
20. Hofmann, U.B., R. Houben, E.B. Brocker, and J.C. Becker, 2005. "Role of matrix metalloproteinases in melanoma cell invasion." Biochimie, 87(3-4): p. 307-14.
21. Egeblad, M. and Z. Werb, 2002. "New functions for the matrix metalloproteinases in cancer progression." Nat Rev Cancer, 2(3): p. 161-74.
22. Lynch, C.C. and L.M. Matrisian, 2002. "Matrix metalloproteinases in tumor-host cell communication." Differentiation, 70(9-10): p. 561-73.
23. Overall, C.M. and C. Lopez-Otin, 2002. "Strategies for MMP inhibition in cancer: innovations for the post-trial era." Nat Rev Cancer, 2(9): p. 657-72.
24. Rudek, M.A., J. Venitz, and W.D. Figg, 2002. "Matrix metalloproteinase inhibitors: do they have a place in anticancer therapy?" Pharmacotherapy, 22(6): p. 705-20.
25. Montel, V., J. Kleeman, D. Agarwal, D. Spinella, K. Kawai, and D. Tarin, 2004.
"Altered metastatic behavior of human breast cancer cells after experimental manipulation of matrix metalloproteinase 8 gene expression." Cancer Res, 64(5): p. 1687-94.
26. Agarwal, D., S. Goodison, B. Nicholson, D. Tarin, and V. Urquidi, 2003.
"Expression of matrix metalloproteinase 8 (MMP-8) and tyrosinase-related protein- I (TYRP-1) correlates with the absence of metastasis in an isogenic human breast cancer model." Differentiation, 71(2): p. 114-25.
27. Unemori, E.N., et al., 2000. "Relaxin induces vascular endothelial growth factor expression and angiogenesis selectively at wound sites." Wound Repair Regen,
8(5): p. 361-70.
28. Gavino, E.S. and D.E. Furst, 2001. "Recombinant relaxin: a review of pharmacology and potential therapeutic use." BioDrugs, 15(9): p. 609-14.
146

MITLibraries
Room 14-0551
77 Massachusetts Avenue
Cambridge, MA 02139
Ph: 617.253.5668 Fax: 617.253.1690
Email: docs@mit.edu
http://libraries. mit. edu/docs