Developing & Applying a Miniaturized Active
Microchip Device
ARCHIVES
AS SACHUSETTS
By
Byron C. Masi
INSTRTUE
JULRAR2E
B.S. Chemical & Biomolecular Engineering
Johns Hopkins University, 2007
IES
Submitted to the Department of Chemical Engineering in
Partial Fulfillment of the Requirements for the Degree of
Doctor of Philosophy in Chemical Engineering
at the
Massachusetts Institute of Technology
June 2012
© Massachusetts Institute of Technology
All rights reserved
Signature of Author...............................
..................
Byron C. Masi
Department of Chemical Engineering
April 18 th 2012
Ce rtifie d by...........................................................................
Robert S. Langer
Institute Professor
'v A esis
e5
or
Ce rtified by........................................................................
4richael J. Cima
Professor of Material Science and Engineering
Thesis Supervisor
A ccepted by ...............................................
. . .................
. . .....
Patrick S. Doyle
Professor of Chemical Engineering
Chairman, Committee for Graduate Students
..................
Developing & Applying a Miniaturized Active Microchip Device
By
Byron C. Masi
Submitted to the Department of Chemical Engineering
On April 18 th, 2012 in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Chemical Engineering
ABSTRACT
Glioblastoma multiforme (GBM) is the most common and aggressive malignant brain tumor.
Treatment of GBM is a daunting task with median survival just at 21 months. Methods of
localized delivery have achieved moderate success in treating GBM. Depot devices have been
limited due to the relatively narrow drug distribution profile they achieve. Convection enhanced
delivery has demonstrated that broad distribution is key, but is limited due to uncertain spatial
distribution and serious side effects. Miniaturized depot devices, implanted into the tissue
surrounding the tumor resection site, could achieve a broad aggregate distribution profile. The
capabilities of localized delivery can be enhanced by utilizing mircoelectromechanical systems
(MEMS) technology to deliver drugs with precise temporal control over release kinetics. An
intracranial MEMS based device was developed to deliver the clinically utilized
chemotherapeutic temozolomide (TMZ) in a 9L rodent glioma model. An activation mechanism
based on thermally induced membrane failure was developed and incorporated. The kinetics of
TMZ release were validated and quantified in vitro. The safety of implanting the device
intracranially was confirmed. The impact of TMZ release kinetics on survival was investigated by
comparing the effects of drug release rates and timing. TMZ delivered from the device
prolonged animal survival. The results from the in vivo efficacy studies indicate that early, rapid
delivery of TMZ from the device results in the most prolonged animal survival. This miniaturized
MEMS device holds tremendous potential for the treatment of GBM and related diseases.
Circuit diseases are neurological disorders that arise from the dynamic miscommunication
within a neural circuit. Anxiety, mood disorders, and the chronic effects of traumatic brain injury
(e.g. Parkinsonism) are prevalent, and are circuit diseases. Circuit diseases could be clinically
addressed by a technology capable of electrical, and chemical neuro-modulation. A catheter
based device capable of simultaneous infusion of multiple fluids and electrical stimulation was
designed and fabricated. Preliminary in vitro infusion studies indicate that the reliable and
reproducible infusion of multiple fluids is possible. Future work will focus on improving the
biocompatibility of the device and studying the performance of the device in non-human primate
models of neurological disorders.
Thesis Supervisors:
Thesis Committee:
Professor Michael J. Cima (MIT)
Institute Professor Robert S. Langer (MIT)
Professor Paula T. Hammond (MIT)
Dr. John T. Santini, Jr. (On Demand Therapeutics, Inc.)
2
Acknowledgements
Michael, you have taught me many lessons throughout the past 5 years, many of which will
remain with me for the rest of my life. Your influence has made me a better researcher and
thinker. Despite your busiest schedules you always make time for your students and never
cease to have insightful and instructive input. Thank you for mentoring me throughout my time
here.
Bob, I'd like to thank you for your never ending enthusiasm and support throughout my time
here. It has been an inspiration to work with you.
I would like to thank my thesis committee members John Santini and Paula Hammond for their
time, attention and guidance. Paula, you are consummately positive and motivating. Thank you
for being a part of my thesis. John, you have been tremendously generous with your time and
advice. Even when you moved to another state, you never ceased to be available for phone
calls and meetings. Thank you for your guidance and words of wisdom.
I would like to thank the members of the Cima lab, past and present, for their influence and
camaraderie. Karen, Grace, Heejin, Irene, Yibo, Yoda, Qunya, Maple, Vincent, Syed, Lenny,
Dan, Joan, Agata, Kevin, Negar, Jen, Anna R., Anna T., Yuan, Ollie, Matt, Laura and Jay, you
have all made for a rich, fun and welcoming group. Alex, thank you for working with me
throughout the early years. It was great to have a companion in putting together all those early
slide decks and figuring how on earth an 'HPLC' works. Hong Linh and Noel, I would like to
especially thank you two for being so helpful and giving with your expertise in MEMS and device
manufacturing. You both made my transition into an unfamiliar field very smooth, and relatively
painless. I wish you both the best with all your endeavors and growing families.
Very special thanks are reserved for Chris and Urvashi. Urvashi, your enthusiasm, humor,
support and astonishing intellect were crucial to my perseverance during the middle years of my
thesis. Thank you for guiding, and bolstering me throughout your time here. Chris, you are one
of the most remarkable people I have ever met. Your intellect, wit, patience and mentorship
have influenced me in countless ways. Thank you for the discussions and pep-talks when my
experiments weren't going well, and thank you for the constant good company and humor.
Steve, Justin, Josh and Adel: You were all a constant source of support, laughs and fond
memories. You're all the reason I survived our first year, and my time here at MIT was better for
having you four as close friends.
Mom, Dad, Dylan and Elizabeth: you have all helped make me the person I am today. Thank
you for your support during the hard times, enthusiasm during the successful times, and your
unerring advice throughout. You should all take pride in this document and the work it
describes. It would not have been possible without you.
3
Liz, you have been my most significant source of support and inspiration throughout my Ph.D.
You are one of the most thoughtful, enthusiastic, supportive, and sage people I have ever met.
Your strength, infectious work ethic, and dedication have bettered me in many ways. You have
made the last several years very special, and meeting you is the most significant event of my
time here at MIT.
4
Table of Contents
List o f Figu re s ................................................................................................................................................
8
List o f T a b le s ...............................................................................................................................................
13
1.
14
Introduction ........................................................................................................................................
1.1 M otivation.........................................................................................................................................14
1.2 Problem Statem ent...........................................................................................................................14
1.2.1 Methods of Local Delivery: Convection Enhanced Delivery & Depot Devices ......................
1.3 M iniaturized depot devices for im proved therapy ........................................................................
1.3.1 M icro-Electrical-M echanical-System s Based Depot Devices .................................................
14
17
18
1.4 Thesis Objectives...............................................................................................................................21
2.
1.5 Co n c lu sio n .........................................................................................................................................
22
Device Design ......................................................................................................................................
23
2 .1 . M ic ro c hip .........................................................................................................................................
23
2.1.1 Chip Release Kinetics .................................................................................................................
23
2.1.2 Device Activation M echanism : A New Approach ...................................................................
27
2.1.3 Energy Transm ission: Reducing Overall Device Invasiveness ...............................................
29
2 .2 Re se rvo ir ...........................................................................................................................................
3.
31
2.2.1 M aterial Selection ......................................................................................................................
31
2.2.2 Reservoir Architecture ...............................................................................................................
31
2.3 Activation Hardware .........................................................................................................................
33
M e th o d s..............................................................................................................................................
35
3.1. Clean room fabrication techniques.............................................................................................
35
3.1.1 Fuse Study Devices.....................................................................................................................
35
3.1.2 Three m em brane devices ..............................................................................................
36
3.2 TM Z analytics (HPLC).........................................................................................................................
37
3.2.1 High Pressure Liquid Chrom atography .................................................................................
37
3 .2 .2 Sta b ility ......................................................................................................................................
37
3.2.3 Solubility.....................................................................................................................................
38
3.3 Device Filling .....................................................................................................................................
5
38
3.3.1 Fuse Study Devices.....................................................................................................................
38
3.3.2 T M Z Devices ...............................................................................................................................
39
3.4 Device Assem bly ...............................................................................................................................
39
3.5 Release procedures...........................................................................................................................40
3.6 In vivo Studies ...................................................................................................................................
3.6.1 Biocom patibility study ...............................................................................................................
40
41
3.6.2 Efficacy Studies..........................................................................................................................41
3.7 Im m unohistological analyses............................................................................................................42
4.
Fuse Activation M echanism ................................................................................................................
43
4.1 Device M anufacture..........................................................................................................................
43
4.2 Activation Experim ents.....................................................................................................................45
4.2.180 pm mem branes.....................................................................................................................45
4.2.2 300 pm m em branes ...................................................................................................................
51
4.3 Release Experiments.........................................................................................................................52
5.
4.4 M echanism Discussion ......................................................................................................................
53
4 .5 C o n c lu sio n .........................................................................................................................................
56
In Vitro Characterization of Device Perfom ance............................................................................
5.1 T M Z Form ulation ..............................................................................................................................
57
57
5.1.1 Packaging in C02.........................................................................................................................57
5.1.2 Co-Form ulation of TMZ and
6.
...........................................................................................
59
5.2 In vitro Release Studies .....................................................................................................................
63
5 .3 Co n c lu sio n .........................................................................................................................................
67
In Vivo Studies.....................................................................................................................................69
6.1 Pilot in vivo studies ...........................................................................................................................
69
6.1.1 Potential toxicity and potential side effects ..........................................................................
69
6.1.2 Prelim inary survival studies ..................................................................................................
71
6.2 Large scal e in vivo .............................................................................................................................
73
6.2.1 Effect of TMZ delivery rate on efficacy.................................................................................
73
6.2.2 Effect of TMZ delivery time on efficacy .................................................................................
74
6.2.3 Com parison of the device to a polym er-based delivery system .............................................
75
6.2.4 Im m unohistological analyses..........................................................................................
6
76
6.3 Discussion of in vivo results ..............................................................................................................
6.3.1 Prelim inary in vivo studies .....................................................................................................
78
6.3.2 Efficacy Studies ..........................................................................................................................
78
6.3.3 Im m unohistological analyses.................................................................................................
80
6.4 Conclusions .......................................................................................................................................
7.
8.
78
Future W ork........................................................................................................................................
81
82
7.1 Co-delivery of synergistic m olecules for the treatm ent of GBM .................................................
82
7.1.2 Prelim inary tw o com partm ent device work ..........................................................................
84
7.1.3 Im pact of drug release kinetics on in vivo efficacy studies....................................................
86
7.2 Investigation of m ass transfer m echanism s in depot devices ......................................................
87
7.2.1 Varying Orifice Size and Num ber ..........................................................................................
88
7.2.2 Changing the Drug Payload ...................................................................................................
89
7.2.3 Including Excipient .....................................................................................................................
90
7.2.4 Plausible m echanism s and directions of further research....................................................
91
A new device: The 'Injectrode'...........................................................................................................
92
8.1 Circuit Diseases .................................................................................................................................
93
8.2 Clinical Rationale: Traum atic Brain Injury, Anxiety and M ood Disorders ....................................
94
8.2.1 Intractable Anxiety and M ood Disorders...............................................................................
94
8.2.2 Traum atic Brain Injury and Associated Chronic Neurological Disorders ...............................
95
8.3 Device Design and M anufacture...................................................................................................
96
8.3.1 The Prior Art for M icro-cannula Devices...............................................................................
96
8.3.2 The First Generation Injectrode ............................................................................................
98
8.4 In Vitro Infusion Results ..................................................................................................................
100
8.5 Future steps in the developm ent of this device .............................................................................
103
8.5.1 In vitro tissue phantom infusion studies.............................................................................
103
8.5.2 Device Coating for Structural Integrity and Biocom patibility ..................................................
104
8.5.3 In vivo experim entation to confirm neural circuit and behavior m odification ....................... 106
8.6 Conclusion.......................................................................................................................................107
9.
References ........................................................................................................................................
7
108
List of Figures
Figure 1.1 MRI images of a GBM patient. A coronal MRI image depicting a GBM lesion
(indicated by the white arrow) (A). Post resection surgery MRI depicting the void (white
arrow) left by the procedure (B). MRI image showing the same patient 6 months after
resection surgery, full chemotherapeutic, and radiation therapy regimen. Two recurrent
lesions are indicated by white arrows (C). Image post second resection surgery showing
two resection sites (D). Scan 3 months after second resection surgery depicting recurrent
15
lesions at the immediate periphery of the resection site (E). Image from Deorah et al.
Figure 1.2 Diagram of the envisioned device implementation. The primary tumor burden is
Infiltrative neoplastic cells
removed, to the extent possible, by surgical resection.
(cartooned here as the black shaded region surrounding the resection site) remain, and are
free to proliferate unless addressed by radiation and/or chemotherapy. Devices can be
packed within the resection site (not shown here for simplicity) and implanted into the tissue
immediately surrounding the resection site. Upon device activation the aggregate release
profile of many diffusion driven devices could be equivalent to the pressure driven release
18
profile of CED and hopefully address the majority of remaining glioma cells.
Figure 1.3 Schematic of the original MEMS based micro-chip drug delivery device
developed by John Santini, Ph.D.. In this manifestation the drug (active substance) is
loaded into micro-machined pyramidal reservoirs that are each capped by a gold
membrane. Activation induces the electrochemical dissolution (formation of gold chloride)
20
of the anode (membrane). Figure from Santini et al.
Figure 2.1 A CAD rendering of the active device in its current form. A liquid crystalline
polymer reservoir (gray) is capped by a silicon chip (purple and gold) that contains 3 nitride
membranes (green). Energy is delivered to the device via polyimide coated copper leads
23
(brown and gold).
Figure 2.2 Survival plots and in vitro release kinetics for the TMZ laden polymer study
conducted by Brem et al. Survival is depicted as the percentage of animals still alive
within a group as a function of time. Both polymer wafer groups displayed improved
median survival and number of long term survivors (animals surviving until the protocol
mandated stop date) when compared to oral treatment and control groups. Figure copied
from Brem et al. (A) The release kinetics of TMZ are first order and result in a final delivery
of approximately 70% of wafer payload in 75 hours. (B)
24
Figure 2.3 Sketch of the assumed spatial concentration profile as drug releases from the
device. The reservoir is assumed to be at saturation, Co. The exterior of the device is
assumed to be at infinite sink conditions, Ci = 0 mg/ml. Steady state, uni-directional
8
diffusion is assumed, allowing the gradient in Fick's first law to be one dimensionalized (Eq.
25
2) and linearized (Eq. 3).
Figure 2.4 Color photographs of the chip:lead assembly (A) and entire device and on a United
States penny (B). A close up photograph of the chip:lead assembly. Gold wire bonds are
visible between the gold coated copper pads of the flexible PCB and the patterned gold on
the chip (A). Photograph of the fully assembled device. The chip rests on the internal
shoulder of the reservoir, leaving a small space for the UV curable epoxy to wet into and be
cured within.
30
Figure 2.5 CAD rendering of the injection molded liquid crystalline polymer reservoir. The
reservoir dimensions are 3.7 by 3.2 x 2.2 mm. The total drug payload is 10 milligrams of
TMZ. The 200 pm shelf is visible on the interior face of the reservoir walls. This shelf
serves a seat for the chip and as an upper boundary for drug during the loading process. A
lead-way was designed in the top perimeter of the chip to allow the polyimide leads to
project out from the device.
32
Figure 2.6 Circuit diagram for the activation hardware. The switchbox (dashed box) contains a
matching resistor, shunt resistor and a single pole, triple throw switch. The head piece
contains each membrane on independent, parallel circuits. The matching resistor allows the
voltage source to be impedance matched to the load. The shunt resistor and membrane are
in parallel such that, when the membrane ruptures, current is directed through the shunt
resistor.
34
Figure 3.1 Fabrication process sequence for creating nitride membranes in the microchip. 36
Figure 4.1 Digital renderings of the suspended nitride membranes and overlying fuses. A.
Simple rectangular fuses (yellow) were deposited over the nitride (green) membrane in the
small membrane study. B. More intricate fuse geometries were studied in the large, 300
pm, membrane devices. Two 20 pm fuses, one 40 pm fuse, two 40 pm 'v' shaped fuses,
three 10 pm fuses, and one 40 pm diagonal fuse.
44
Figure 4.2 SEM microscope images showing the structure before and after activation for
each fuse width (250 nm Au layer). A. 20 pm wide. B. 40 pm wide fuse. C. 60 pm wide. 47
Figure 4.3 Confocal microscope images for 250 nm thick fuses. A. 20 pm wide. B. 40 pm
wide. C. 60 pm wide.
48
Figure 4.4 Oscilloscope capturing images showing activation pulse voltage values (blue)
and current (red) values for 250 nm thick fuses. A. 2 pm wide. B. 40 pm wide. C. 60 pm
wide.
49
9
Figure 4.5 Minimum energy to burst fuse as a function of fuse dimensions. Please refer to
Table 4.1 for values.
50
Figure 4.6 Light microscope images of ruptured 300 pm membranes.
ruptured via x fuses. B. Membranes ruptured via diagonal fuses.
A. Membranes
51
Figure 4.7 In vitro release profiles, n = 5, for activated devices, n = 5 for control devices.
Error bars represent standard deviation.
53
Figure 4.8 FEA Analysis of the suspended structure showing an
Misses stress on the surface area of the membrane,
displacement showing structural deformation. The mesh
approximately 10E5 elements. A. 20 pm wide fuse (B), 40 pm
fuse (D).
isometric view of the Von
and a lateral view for
of the analysis included
wide fuse (C), 60 pm wide
54
Figure 5.1 An example of inconsistent release profiles that result from packaging in C02.
The rate of release was consistent across the three devices and compares well to the
theoretical release. The onset of release and overall extent of release vary greatly across
the 3 devices. The theoretical line is based on zero order approximations of the release rate
utilizing Fick's first law. The controls demonstrate excellent sealing and drug retention
throughout the time scale for release.
58
Figure 5.2 Renders of the device in the three stages of assembly. A. An empty reservoir.
The shoulder that the chip must rest on is visible around the interior perimeter of the
reservoir. B. A filled reservoir. Drug and any excipients must be loaded into the device
without accumulating on the shoulder in order to ensure proper alignment during
subsequent assembly steps. C. An assembled device. The chip is placed in the reservoir,
resting on the shoulder. Biomedical grade epoxy is then used to secure the chip to the
reservoir and seal the reservoir.
60
Figure 5.3 A series of color photogrpahs depicting the steps in device filling. An empty
reservoir is placed in a fixture that allows a pre-weighed amount of drug to be added up the
the level of the shoulder (A). Drug is loaded into the reservoir and partially compacted to
create a small well within the drug powder. This well allows molten PEG to be added to the
reservoir (B). After typically three repititions, the majority of the air has been displaced from
the TMZ powder and replaced by PEG. The vacuum fixture ensures that TMZ and PEG do
not escape the reservoir or accumulate on the shoulder (C).
61
Figure 5.4 Schematic of the vacuum co-formulation process. The vacuum fixture clamps a
high air permeability Teflon AF membrane over the surface of the PEG:TMZ mixture. The
PEG:TMZ mixture is trapped within the reservoir, but any air is free to flux out when a
vacuum is pulled. Once the air is removed, PEG is free to wet within the interstices of the
TMZ powder.
62
10
Figure 5.5 Release curves for TMZ filled devices releasing into 37 0 C water. Devices (n=3)
with 3, 2, 1 or 0 membranes activated were placed in agitated baths and sampled for TMZ
content. The errors bars are the standard error. Release is a function of the number of
activated membranes and release does not occur unless the device is activated.
63
Figure 6.1 Graphical presentation of normalized animal weight as a function of time. All
animals displayed minimal weight loss during the acute phase and animal 5 displayed
robust weight gain over the chronic time scale.
70
Figure 6.2 Survival curves for the preliminary efficacy trial. Animals with activated devices
demonstrated the most prolonged survival. Un-activated devices had an identical impact
on animal survival as no treatment at all. Animals receiving aCSF had a slightly shorter
median survival than no treatment.
72
Figure 6.3 Impact of drug release rate on survival. Animals that received activated devices on
day 0 had median survivals of 40 (42.8 % LTS), 28 (28.5 % LTS), and 21 (12.5 % LTS)
days for 3, 2, and 1 membranes activated respectively.
73
Figure 6.4 Impact of drug release time on survival. Animals that had all 3 membranes activated
day 0, 3, or 5 had median survivals of 40 (42.8 % LTS), 24 (12.5 % LTS) and 23 days. 74
Figure 6.5 Comparison between microchip and polymer-based delivery methods. Those animals
that received two TMZ:polymer wafers on day 5 had a median survival of 34 days, while
those that had all 3 membranes opened on day 5 had median survival of 23 days.
75
Figure 6.6 Immunohistological results of Ki67 and caspace-3 staining. (A) Ki67 positive cells are
green and cell nuclei are blue. Each panel is a representative image from each efficacy
study group. The 3MOD (3 membranes activated on day 0) panel contains the fewest
number of Ki67 positive cells. (B) Quantitative results obtained by averaging 3
representative images from each group. The two longest surviving groups, wafer and
3MOD, have the lowest levels of Ki67 positive cells. (C) Caspase-3 positive cells are green
and cell nuclei are blue. Each panel is a representative image from each efficacy study
group. (D) Quantitative results obtained by averaging 3 representative images from each
group. The 3MOD (3 membranes activated on day 0) panel contains the second highest
number of caspace-3 positive cells.
77
Figure 7.1 CAD renderings of cross sectional views of the proposed two compartment
device. The reservoir architecture is essentially the same as the single compartment
device only mirrored around and internal plane. Each reservoir can be capped by the
normal chip without any alteration to the chip design or manufacturing.
84
11
Figure 7.2 Release of TMZ from a two compartment reservoir device. Each compartment was
loaded with 5 milligrams of TMZ and co-formulated with PEG. The first chip had 3
membranes activated at t = 0 hours and the second chip was activated at t = 162 hours. 85
Figure 8.1 Digital renders and SEM images of the core components of the injectrode. A CAD
rendering of the manifold. Four ports are visible. The port on the circular face is a through
hole through with the introducing needle and glass tube are threaded. The center port on
the cylinder surface is for filling the epoxy chamber (panel B). Pump tubing is inserted into
the remain two ports which service the infusion chambers (Panel B)( A). A cross sectional
view of the manifold depicting the epoxy chamber, two infusion chambers, and the silicone
septa (blue) (B). CAD rendering of the multi-lumen glass tube. The large lumen (90 pm)
houses the electrode. The two small (38pm) lumens conduct fluid from manifold down to
the distal tip of the device. Each 38 pm lumen has an access port cut into it to allow for
interfacing with the manifold (C). Scanning electron microscopy image of several glass
tubes viewed end on.
99
12
List of Tables
Table
Table
Table
Table
Table
2.1: Device dimensions for a 10 mg payload ............................................................
4.1: Activation energy values depicted in Figure 4.5. n = 5......................................
6.1: Survival data for the preliminary efficacy study .................................................
6.2: Survival data for the large scale efficacy study .................................................
8.1: Infusion results for the first two lumen device ......................................................
13
27
48
71
75
100
1. Introduction
1.1 Motivation
A glioma is a cancerous tumor arising from the glial cells of the brain.
The most
common and aggressive form of glioma that afflicts humans is glioblastoma multiforme (GBM)
[1]. GBM accounts for 12-15 percent of all primary brain tumors and afflicts 5000 Americans per
year [2, 3]. Glioblastoma multiforme is nearly invariably fatal and the average patient will survive
between a few months (minimal treatment) and 12 to 18 months (maximal treatment) [4, 5], and
those patients surviving 2-3 years are deemed 'long-term survivors'. Deorah et al, concluded
that GBM patients have not shown improved survival since the late 1980's based on a study
conducted in 2006 of brain cancer incidence and a review of survival rates in the United States
from 1973 to 2001 [1].
Glioblastoma multiforme and related gliomas present devastating
challenges to patient survival and demand improved treatment modalities.
1.2 Problem Statement
1.2.1 Methods of Local Delivery: Convection Enhanced Delivery & Depot Devices
Current treatment methods generally combine surgical resection (Figure 1.1) of the
primary tumor, radiation therapy and a chemotherapeutic regimen [5]. Chemotherapy is most
often administered systemically via intravenous injection or oral formulations. One of the major
limitations to the development of more effective brain tumor therapies is the presence of the
blood-brain barrier, which prohibits the transfer of molecules that are larger than 500 daltons or
are non lipid-soluble. The presence of the blood brain barrier causes only a fraction of
administered systemic dose to ever reach the brain. To achieve therapeutics levels within the
brain, excessive amounts of drug must be administered to account for the poor partitioning,
resulting in systemic toxicities. Dosing regimens are frequently designed to remain just under
14
toxicity thresholds, as opposed to achieving therapeutic tissue concentrations. Limitations in
drug exposure and systemic toxicity have spurred on the development of localized delivery
methods [6-12].
One method of circumventing the blood brain barrier is to pack the resection site with
drug depots which can then release drug into the afflicted tissue in a time-controlled manner.
These depot devices have shown promise in treatment, but achieve limited spatial distribution
[7, 8, 10, 13-15]. The concern is that the limited distribution of drug leaves wide spread, unresected, glioma cells unaddressed and free to proliferate (Figure 1.1 C,E) [7].
Efforts have
been made to find methods that improve spatial drug distribution.
One method for achieving increased drug distribution is convection-enhanced delivery
Figure 1.1 MRI images of a GBM patient. A coronal MRI image depicting a GBM lesion (indicated by
the white arrow) (A). Post resection surgery MRI depicting the void (white arrow) left by the procedure
(B). MRI image showing the same patient 6 months after resection surgery, full chemotherapeutic, and
radiation therapy regimen. Two recurrent lesions are indicated by white arrows (C). Image post
second resection surgery showing two resection sites (D). Scan 3 months after second resection
surgery depicting recurrent lesions at the immediate periphery of the resection site (E). Image from
Deorah et al.
15
(CED). A catheter is surgically placed in the target tissue under image guidance during CED.
An external pump is then used to infuse drug solution, and to drive subsequent fluid flow in the
tissue. CED has been shown to be efficacious both in achieving broad distribution of drug and
in retarding tumor progression [6, 7, 9, 12, 16-18]. The results of CED have demonstrated that
broad distribution is possible and is in fact critical to improved treatment, but as a treatment
modality CED suffers from several considerable drawbacks.
The medical procedure surrounding CED can be extremely inconvenient for the patient.
Catheter placement and infusion requires that the skull be immobilized in a stereotaxic device
and the catheter must be connected to an external pump and fluid reservoir during infusion.
Infusion times in the literature have ranged from hours to weeks and commonly last around 7
days [12, 15].
The greatest issue confronting CED is the uncertainty in fluid flow.
The physical
parameters and characteristics of the brain greatly dictate the direction of fluid flow. White
matter presents less resistance to fluid flow causing the infused fluid to preferentially flow
through white matter [7]. The natural structure of the brain can form conduits for fluid flow.
These effects are largely unpredictable and unavoidable.
Infusion itself changes the tissue
properties of the brain. The initial path of fluid flow alters the tissue, for example, by dilating the
ECM and causing it to present less resistance to fluid flow. Any subsequent administrations of
treatment, therefore, are more likely to follow the same path, possibly leaving portions of
diseased tissue completely unaddressed. CED's effect on surrounding tissue is so extensive
that it can cause edema that is in many cases indistinguishable from peri-tumoral edema [7].
Edema resulting from the disrupted vasculature of GBM tumors is already one of the leading
causes of morbidity in patients [19]. A treatment that exacerbates the level of edema, and
therefore morbidity, is severely limited.
16
The benefits and drawbacks of CED can best be understood in the context of the
disease. Glial tumors have been shown to have heterogeneous cellular populations within a
given tumor [4, 20, 21].
Two basic and easily defined cohorts are proliferative cells and
migratory cells. The presence of migratory cells is thought to be a leading cause of tumor
recurrence [20]. Migratory cells are able to infiltrate healthy tissue surrounding the primary
tumor. Their departure from the primary tumor allows them to avoid physical removal by
resection, and therapeutic insult because of the limited drug distribution profile discussed above
(Figure 1.1). A large distribution profile, therefore, is necessary for efficient treatment. Migratory
cells also tend to follow certain anatomical structures during infiltration [20]. These concepts
demonstrate that regional control of delivery of multiple agents is critical to disease treatment.
1.3 Miniaturized depot devices for improved therapy
Improved drug distribution is clearly necessary to improve treatment, but the migratory
nature of the disease exposes the fact that regional control of drug delivery is critical. The
uncertainty in fluid flow present in CED is unacceptable, but its increased distribution profile is
crucial. Depot devices could be employed in such a way as to both achieve an improved
distribution and capitalize on disease characteristics. Miniaturized depots could be packed into
the resection site as well as injected to a perimeter beyond the resection site (Figure 1.2). The
ability to place devices in specific locations facilitates delivery to specific anatomical features
while avoiding the uncertainty present in CED. Each device has its own reservoir allowing for
regional control of the delivery for each therapeutic. The majority of recurrent lesions occur
within 2-3 centimeters of the resection site [5]. Implanting devices within the tissue immediately
surrounding the resection site, therefore presents a promising approach to preventing tumor
recurrence. The use of miniaturized active devices allows for the heterogeneous nature of the
disease to be treated in a comparably heterogeneous fashion.
17
Figure 1.2 Diagram of the envisioned device implementation. The primary tumor burden is
removed, to the extent possible, by surgical resection. Infiltrative neoplastic cells (cartooned here as
the black shaded region surrounding the resection site) remain, and are free to proliferate unless
addressed by radiation and/or chemotherapy. Devices can be packed within the resection site (not
shown here for simplicity) and implanted into the tissue immediately surrounding the resection site.
Upon device activation the aggregate release profile of many diffusion driven devices could be
equivalent to the pressure driven release profile of CED and hopefully address the majority of
remaining glioma cells.
1.3.1 Micro-Electrical-Mechanical-Systems Based Depot Devices
Depot devices are excellent vehicles for drug delivery because they localize treatment,
avoiding systemic toxicity, and protect the drug from degradation and clearance until release.
Drug filled or impregnated depots have demonstrated success in both laboratory and clinical
settings (Gliadel @) [22-29]. Temozolomide loaded polymer microcapsules developed by Scott
et al achieved marked improvement in median survival over systemic administration in primary
18
brain tumor model [30]. These results motivate the development of miniaturized depot devices
for the local delivery of chemotherapeutic to the brain.
The versatility in function of depot devices can be enhanced by utilizing micro-electromechanical system (MEMS) technology. MEMS based, or 'active', devices offer the exquisite
advantage of being able to actively control the function of a device via minute electrical signals
[31-34].
Such communication offers the ability to control drug release rate and when drug
release begins, allowing for the creation of complex temporal profiles of one or multiple
therapeutics [32, 33, 35-41]. The physical manifestation of the devices is generally a macromachined drug reservoir that is capped by a micro-machined microchip such that drug release
is gated by the microchip. The microchip contains etched through-holes that are capped by a
ceramic or metallic membrane (Figure 1.3).
Patterned metallic layers serve as electrical
conduits for the purposes of device activation. The precise patterning of metal and membranes
dictates which, and how many membranes are activated (ruptured, dissolved or ablated) for a
given signal. The mechanism of activation depends on device design, but viable mechanisms
have included electrochemical dissolution', thermal ablation, thermally induced mechanical
failure and pressure gradient induced mechanical failure [40, 42-44]. The fundamental principle
of activation is that it removes the membrane(s) as a barrier to drug diffusion, thus allowing
release to begin.
1Application of this mechanism in vivo was limited by the adsorption of proteins to the electrochemically
active surfaces.
19
a
r-
Silicon
Cathode
Active
substance
Small reservoir opening
(usually covered by gold membrane)
b
-
Silicon side wall
Large reservoir
opening (for reservoir filling)
Figure 1.3 Schematic of the original MEMS based micro-chip drug delivery device
developed by John Santini, Ph.D.. The drug (active substance) is loaded into micro-machined
pyramidal reservoirs that are each capped by a gold membrane.
Activation induces the
electrochemical dissolution (formation of gold chloride) of the anode (membrane). Figure from Santini
et al.
The benefits of temporal control are more pronounced when viewed in the context of the
disease. Certain therapeutics have been delivered in tandem and demonstrated efficacy that is
greater than additive [21, 45, 46]. One of the agents, frequently is not itself efficacious, but
'potentiates' the primary agent, commonly by inhibiting natural biological methods of resistance
to the primary agent [46]. The efficacy of co-delivery can be temporally dependent such that a
pair of drugs will be much more effective when delivered at separate, specific times. There are
also two specific clinical scenarios where temporal control presents a pronounced advantage
over conventional depot devices. First, having precise control over when release begins allows
20
the clinician to implant multiple doses of drug during the initial procedure and to administer
those doses, at will, any time after implantation via activation. Polymer based devices either
begin releasing immediately or on a time scale dictated by polymer degradation (e.g. in the case
where drug reservoirs are capped by biodegradable polymers of varying molecular weight). The
clinician, therefore, can only implant a single dose or is bound to the dosing regimen dictated by
the degradation of the polymer itself. Second, the cellular composition of a tumor can vary with
time, either as a result of therapy or by natural means. The clinician, therefore, would benefit
from being able to implant devices containing an array of therapeutics, and to decide which
therapeutics are delivered during the course of therapy. This is the popular 'pharmacy on a chip'
idea. These concepts demonstrate that precise temporal control over delivery of multiple agents
would be a superb tool in treating glioblastoma multiforme.
Pulsatile release of drugs in vitro and in vivo has been achieved with a previous
generation of the active device [13, 47]. Controlled release devices utilizing MEMS technology
have been used for in vivo delivery of small molecules and polypeptides and have demonstrated
efficacy in reducing disease progression in a rodent flank glial tumor model [44, 47, 48]. The
first human trials, in fact, with a microchip drug delivery device were just completed with
stunning success[41]. These devices, however, have previously been restricted to the
subcutaneous space due to prohibitively large structural components. This work details the
design, manufacture and testing of a miniaturized MEMS based device aimed at treating GBM.
1.4 Thesis Objectives
The overarching objective of this work was to design and fabricate a miniature active
device that is capable of intracranial implantation and drug delivery in a rodent glioma model.
Sub-aims within this objective were to:
21
-
Conduct a redesign of the device structure and manufacturing techniques to reduce
overall size , while improving reliability
-
Introduce a new activation mechanism aimed at reducing the energy requirements of
activation and improving device reliability
-
Conduct in vitro release experiments to quantify and validate the reliable release of drug
from the device
-
Demonstrate device efficacy in an intracranial rodent model of glioma
1.5 Conclusion
Advances in fabrication techniques and a revamped structural design allowed for the
creation of an active device with characteristic dimensions similar to previously used polymer
based intracranial depot devices [28, 30]. This miniaturized device can now be implanted
intracranially and its efficacy studied in a rodent glioma model. A new activation mechanism
was adopted into the device design. The popular chemotherapeutic temozolomide (TMZ) was
chosen as our active agent and methods of loading and formulating were developed. In vitro
release studies quantified and validated the reliability and kinetics of TMZ release. The device
was designed and manufactured such that 3 different release rates are possible. Several in vivo
studies were conducted to determine the in vivo reliability in function of the device as well as the
efficacy of TMZ delivery against a rodent glioma model. In vivo survival studies in a 9L
gliosarcoma rodent study demonstrated that temozolomide delivery from the active device is
capable of prolonging animal survival. Subsets of these in vivo studies began to investigate the
impact of device release rate and release timing on animal survival.
22
2. Device Design
2.1. Microchip
Figure 2.1 A CAD rendering of the active device in its current form. A liquid crystalline
polymer reservoir (gray) is capped by a silicon chip (purple and gold) that contains 3 nitride
membranes (green). Energy is delivered to the device via polyimide coated copper leads (brown and
gold).
The active microchip device (Figure 2.1) has been an ongoing project in the Cima and
Langer Laboratories for many years. The device as a whole underwent a major revision
throughout this thesis work. This chapter is a summary of the major revisions that were made to
the device as well a brief discussion of the rationale utilized in each revision. The overarching
objective of each modification was to reduce the device size and to improve reliability and
efficacy.
2.1.1 Chip Release Kinetics
23
100
so
80
60
40
20
A
0
150
100
50
0
Time (Days)
4.0
o
3.5
f
__3.0Wi
0O
~2.5
20
M 1.5
1.0
B
0.5
0.0
0
10
20
.-
.-
.
.-
.
30
40
50
a.
.-
.
60
70
80
Time (hrs)
Figure 2.2 Survival plots and in vitro release kinetics for the TMZ laden polymer study
conducted by Brem et al. Survival is depicted as the percentage of animals still alive within a
group as a function of time. Both polymer wafer groups displayed improved median survival and
number of long term survivors (animals surviving until the protocol mandated stop date) when
compared to oral treatment and control groups. Figure copied from Brem et al. (A) The release
kinetics of TMZ are first order and result in a final delivery of approximately 70% of wafer payload in
75 hours. (B)
TMZ laden polymer wafers were chosen as a gold standard for the design of the active
device release kinetics. The wafers had proven efficacious in a 9L gliosarcoma rodent study,
prolonging survival over oral treatment (Figure 2.2 A) [28].
These wafers, when loaded with 5
mg of TMZ, release their payload with first order diffusion kinetics, achieving 70% release in
roughly 75 hours (Figure 2.2 B). This release rate was chosen as target rate to be achieved by
the active device. Zero order approximations of Fick's first law allowed for the total area for flux
required to be determined.
Where J is flux (mg/(s*cm 2)), D is the diffusivity, C is the
concentration, x is the vector normal to the surface of the chip, Ci is concentration outside the
device, C, is the concentration inside the device, A is the surface area for flux and m is the
mass transfer rate.
0
Orifice
Reservoir
Exterior
C
C
0
U
C1
x dimension
Figure 2.3 Sketch of the assumed spatial concentration profile as drug releases from the
device. The reservoir is assumed to be at saturation, Co. The exterior of the device is assumed to be
at infinite sink conditions, Ci = 0 mg/ml. Steady state, uni-directional diffusion is assumed, allowing the
gradient in Fick's first law to be one dimensionalized (Eq. 2) and linearized (Eq. 3).
1= -DVC
Eq. 1
I
Eq. 2
= -D
dC
2x
25
=
-D
Eq. 3
"o
(1-0)
Eq. 4
J= D
D
10 -=; Co =8.8 ';
C, = 0
=; 0.0003 cm
J -A = r
A =Eq.
Eq. 5-7
Eq. 8
9
The total area for flux is divided among several through-holes that are each capped by a
nitride membrane. Two constrictions were utilized to determine the size and number of nitride
membranes that would be incorporated into the device. First, a maximum membrane length
and width was set at 300 micrometers (pm). This dimension was set as a balance between
large membranes for rapid release and robust membranes that would be capable of enduring
the fabrication, assembly and implantation process 2. The suspended nitride membranes are
fabricated by a self-terminating 20% potassium hydroxide (KOH) wet etch. KOH etches silicon
100 anisotropically with an angle of 57.740 from the wafer plane, in this case resulting in a
truncated pyramidal shape with the suspended nitride membrane at the top. Each membrane,
therefore, requires a -720 by 720 pm square on the backside of the wafer to initiate the etch.
This larger feature determined the spacing between membranes, and the chip size per number
of membranes. The membranes we separated and staggered to increase the amount of bulk
silicon. The resulting membrane layout is depicted in Figure 2.4 A.
The second constriction is that the final device dimensions must allow for it to be
implanted within the cranium of a rat. Device dimensions for a payload of 10 milligrams (mg)
MIT Microsystems Technology Laboratory technicians served as knowledgeable sources for the upper bound of
robust nitride membranes. A test wafer was fabricated with 300 by 300 pm membranes and run through the more
vigorous processing steps (piranha, spin coating and dicing) to verify membrane survival before proceeding with
full mask design and purchase.
2
26
Table 2.1: Device dimensions for a 10 mg payload
Length
Width
Depth
Payload
(mm)
(mm)
(mm)
(mg)
5
4.2
4.2
1.6
10
3
3.7
3.2
2.2
10
2
3.7
2.6
2.6
10
# of Membranes
were calculated for devices containing 5 membranes, 3 membranes and 2 membranes (Table
2.1). Varying the number of membranes changes the dimensions of the chip and therefore the
dimensions of the reservoir. The wafers were again used as a standard. Two wafers implanted
simultaneously resulted in the best animal survival. The overall dimensions of two polymer
wafers are approximately 3 millimeters cubed. The final device design incorporates 3
membranes and has overall dimensions of, 3.7 by 3.2 by 2.2 mm3 . Approximations of Fick's law
predict a release rate of 3.5 milligrams in -120 hours.
Patterning of the metal layer was designed such that each membrane can be activated
independently.
The final device, therefore, is capable of 3 different release rates.
This
versatility is a valuable tool in ascertaining what release rate or range of release rates is
efficacious in vivo.
2.1.2 Device Activation Mechanism: A New Approach
A new device activation mechanism was proposed and reduced to practice in an effort to
improve device reliability, namely improved device sealing, as well as to reduce the energy
requirements of the device.
The opening mechanism is a thin metallic fuse that spans a
Betty Tyler of the Brem lab was a priceless resource for discussing the impact of implantation on device design.
Mock-up devices machined from polypropylene were sent to the Brem lab and implanted as a preliminary trial of
the device architecture.
3
27
suspended membrane structure that isolates the contents of the device reservoir from the
environment. The fuse is geometrically laid out on the structurally weakest point of the
membrane. Activation causes rapid resistive heating of the fuse material, creating large stresses
on the suspended structure leading to membrane rupture and reservoir exposure.
Previous device generations did not have a complete, uninterrupted nitride layer that
spanned the entire device. These devices, instead, had titanium and gold membranes that
spanned the orifices created by the KOH etching [13]. Sealing, therefore, depended strongly on
the degree of alignment between the patterned membranes and the etched orifices.
The
fabrication process for the fuse mechanism, however, leaves the original conformal nitride layer
intact. Sealing, therefore, is purely a function of the nitride remaining patent.
The previous activation mechanism relied on the thermal ablation of the gold and
titanium membrane. The fuse mechanism reduces the amount of gold and titanium for a given
orifice size. The energy required, therefore, to heat the metal is greatly reduced. The question
remained, however, whether a thin strip of metal was capable of rupturing a much larger nitride
membrane when heated. The role of fuse width, thickness and geometry was closely studied in
this work (Chapter 4). The primary metrics for activation were the extent to which the membrane
was ruptured and energy requirements for achieving rupture.
The fuse mechanism was ultimately adopted into the final device design. Two fuses, 40
pm wide by 250 nm thick, traverse each membrane. The fuses are patterned such that they
form two Vs with the vertex of each 'V near the center of the membrane. The details of how
these dimensions and geometries were finalized are discussed in detail in Chapter 4.
28
2.1.3 Energy Transmission: Reducing Overall Device Invasiveness
Previous generations of the active device utilized hand soldered connectors and bundled
copper wires to conduct energy to the device. The resulting structure was large, time consuming
to construct, and awkward to implement in vivo. A flexible printed circuit board (PCB) was
designed and outsourced to Flexible Circuit Technologies, Inc. (Minneapolis, MN). The PCB
contains 4 copper leads that are 0.012" by 0.0014" by 1.2 inches. One end the PCB (0.197" x
0.118) is designed to insert into a zero insertion force connector (Tyco electronics, Berwyn, PA)
built into the activation set up. The 4 copper leads splay out into gold coated copper pads
(0.027" x 0.118") spaced to mate with the copper heads of the connector and are reinforced with
a stiffening layer. The other end of the PCB (0.110" x 0.039") is designed to match up to the
gold patterning on the chip (Figure 2.4 A). The copper again splays out into gold coated copper
pads (0.016" x 0.039") that are spaced to allow for gold wire ball bonding between the PCB and
chip. The intervening lengths of copper are coated in an insulating, biocompatible layer of
polyimide. The leads are fixed to the chip with a small amount of biocompatible cyanoacrylate.
The leads are now biocompatible, miniaturized, and can be rapidly and reliably connected to the
chip and allow quick and repeatable device connection to a multi-meter or function generation.
29
Figure 2.4 Color photographs of the chip:lead assembly (A) and entire device and on a United
States penny (B). A close up photograph of the chip:lead assembly. Gold wire bonds are visible
between the gold coated copper pads of the flexible PCB and the patterned gold on the chip (A).
Photograph of the fully assembled device. The chip rests on the internal shoulder of the reservoir,
leaving a small space for the UV curable epoxy to wet into and be cured within.
2.2 Reservoir
The previous reservoir design utilized sand blasted Pyrex@ sheets. These sheets were
coated in epoxy, and stacked to form the reservoir. The result was a large, labor intensive,
structure that had 5 total sealing interfaces. A single component reservoir was designed and
injection molded. The new reservoir is a fraction of the size of the Pyrex @ reservoir and has
only one sealing interface when assembled with the chip.
2.2.1 Material Selection
The reservoir is injection molded (Matrix, Inc. Providence, RI) liquid crystalline polymer
(Vectra 1300 @). Vectra 1300 LCP is a biocompatible, chemically inert, thermoplastic with high
tensile strength and modulus and excellent structural stability. Its low melt viscosity makes it an
excellent candidate for injection molding thin walled parts.
Injection molding the reservoirs
allows for much thinner walls (down to 200 pm) and tighter tolerances to be achieved resulting
in smaller device volume while improving device reliability.
2.2.2 Reservoir Architecture
31
Figure 2.5 is computer aided design (CAD) rendering of the reservoir.
The inner
dimensions of the reservoir were sized such that it would contain a 10 mg payload of TMZ when
loaded in powder form and could be capped by a 3 membrane chip 4 . The overall reservoir
dimensions are 3.7 by 3.2 by 2.2 mm. The side and bottom walls are 400 pm thick. A 200 pm
wide shelf was designed into the inner wall of the reservoir. This shelf and the bordering walls
serve as a seat for the chip, ensuring reproducible alignment and sealing between the chip and
reservoir. The shelf is recessed 400 pm from the top of the reservoir such that when the 300
Shoulder
Lead way
3 mm
Figure 2.5 CAD rendering of the injection molded liquid crystalline polymer reservoir. The
reservoir dimensions are 3.7 by 3.2 x 2.2 mm. The total drug payload is 10 milligrams of TMZ. The
200 pm shelf is visible on the interior face of the reservoir walls. This shelf serves a seat for the chip
and as an upper boundary for drug during the loading process. A lead-way was designed in the top
perimeter of the chip to allow the polyimide leads to project out from the device.
4 A sister, polymer based, device in the lab developed by Alex Scott was used to approximate the packing density of
TMZ.
32
pm thick chip is placed in the reservoir and rests on the shoulder a thin bead of epoxy can be
run around the perimeter of the chip thus securing the chip and isolating the contents of the
reservoir from the environment. One of the short (2.2 mm) bordering walls is only 350 pm tall
(from the shelf) to allow for the previously described flexible PCB to protrude out from the chip
and reservoir assembly.
2.3 Activation Hardware
Peripheral hardware was designed and manufactured to facilitate the rapid and
reproducible activation of each device. The activation hardware consists of two primary units:
the switchbox, and the head piece, which together result in the circuit diagrammed in Figure 2.6.
The switch box contains a printed circuit board designed to accept electrical input from via BNC
connector. Electrical signal is first routed through a 10 ohm resistor. This resistor is in series
with each electrical fuse of the device and is present so that the power source can be
impedance matched to the load (i.e. the device and activation set up). A switch is in series with
the matching resistor. The purpose of this switch is to route current to each membrane
selectively. A 10 ohm shunt resistor is in parallel with the device. Each fuse on the device is
approximately 2 ohms. When the fuse is intact, therefore, the majority of current will flow
through the device and not through the shunt resistor. The resistance of the device increases
dramatically (i.e. >>10 ohms) once the membrane ruptures, causing the majority of the current
to flow through the shunt resistor.
This feature was designed in so that the instant the
membrane ruptures, current would stop flowing to the device, thus minimizing the exposure of
the tissue to heating and potential electrocution.
The head piece contains a printed circuit board that receives current and routes it toward
the zif connector. The head was made as small as possible and separated from the switch box
by a length of copper cabling. This was done so that the ZIF connector can be brought into
33
close proximity of the animal in a variety of orientations. In vivo activations can now occur,
rapidly, and reproducibly, with very little manipulation of the animal or device.
r-
~' -
|
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~ ~ ~ ~ ~~~
~ ~ ~ ~ ~ ~ ~ ~l--
Rinat<.i
1A-
AIcin biratic
Figure 2.6 Circuit diagram for the activation hardware. The switchbox (dashed box) contains a
matching resistor, shunt resistor and a single pole, triple throw switch. The head piece contains each
membrane on independent, parallel circuits. The matching resistor allows the voltage source to be
impedance matched to the load. The shunt resistor and membrane are in parallel such that, when the
membrane ruptures, current is directed through the shunt resistor.
34
3. Methods
3.1. Clean room fabrication techniques
3.1.1 Fuse Study Devices
The fabrication process involves standard bulk micro-machining, shown in sequence in
Figure 3.1. The first step involves the deposition of low-stress, low pressure chemical vapor
deposition (LPCVD) Si3 N4 on 4 inch, 300 pm thick single-crystal-silicon (SCS) wafers
(orientation 100). One side of the wafer was patterned via standard photolithography and etched
via reactive ion etching (RIE) to define -200 pm by 200 pm regions of bare silicon. Exposure to
20% KOH solution resulted in a self-terminating etch that created 80 by 80 pm suspended Si3N4
membranes. Titanium and gold layers were sputtered after standard organic cleaning. The Au
thickness was varied at 100, 250 nm, 500 nm in order to vary the current density required to
experimentally burst the membranes.
The fuses were then defined using standard
photolithography, followed by sequential timed wet etching steps using gold and titanium
etchants. The lateral dimensions of the fuses were defined by the mask design as 80 pm long,
whereas the width was varied per design as 20, 40, 60 pm.
35
1. SCS Wafers coated
with 200 nm of LPCVD
Si3N4
2. Backside Photolith to
define membranes
3. KOH Etch defining
suspended membranes
4. Front-side Ti, Au
deposition
5. Front-side photolith to
define fuses and
electrodes
6. Au and Ti etching,
followed by organic
cleaning
Figure 3.1 Fabrication process sequence for creating nitride membranes in the microchip.
3.1.2 Three membrane devices
The same general fabrication process as outlined in 3.1.1.1 was used to generate the 3
membrane devices, except with minor edits to particular dimensions and geometries. One side
of the wafer was patterned via standard photolithography and etched via reactive ion etching
(RIE) to define -720 pm by 720 pm regions of bare silicon. Exposure to 20% KOH solution
resulted in a self-terminating etch that created 300 by 300 pm suspended Si 3N4 membranes.
Titanium and gold layers were sputtered after standard organic cleaning. The Ti thickness was
set at 30 nm and the AU thickness at 250nm. The fuses were then defined using standard
photolithography, followed by sequential timed wet etching steps using gold and titanium
etchants. The fuse dimensions were set at 250 nm thick by 40 pm. Fuse shape was varied to
36
determine which geometry resulted in the greatest area of membrane rupture. Eventually, the
fuse deposition step was reduced from a 2 step process to a single step deposition as lack of
conformal deposition on the second step led to variability in device activation reliability.
3.2 TMZ analytics (HPLC)
3.2.1 High Pressure Liquid Chromatography
Two methods were used to quantify TMZ using high pressure liquid chromatography
(HPLC). The first method was employed for samples of TMZ in water, ACSF, and pH buffers.
Briefly, 20uL of sample was quantified at 37C on Agilent 1200 SeriesHPLC using a Synchropak
SCD-100, 5 um, 150x4.6 mm column (Synchrom, Lafayette, IN, USA), a flow-rate of 1 ml/min,
0.01 M Ammonium Acetate (aq):Acetonitrile (92:8) mobile phase, and UV absorption at 316 nm.
A second method was used for quantification of TMZ in fetal bovine serum (FBS). The same
chromatographic conditions were used with the addition of minor sample preparation. 200 uL of
sample was added to 100 uL of 100ug/mL Hydrochlorothiazide. The resulting solution was
vortexed and then spun at 4500 G using a MiniSpin centrifuge (Eppendorf) at room temperature
for 1 minute. Samples of the supernatant were analyzed on the HPLC (These methods were
adapted from Kim et al 1997) [49].
3.2.2 Stability
Stability studies were conducted in HPLC-grade water (Sigma Aldrich), phosphate
buffered saline (PBS) (1x), artificial cerebrospinal fluid (aCSF) and pH7 buffer. Solutions of
approximately 0.5, 0.25 and 0.1 mg/ml were made for each solvent. The resulting solutions
were stored at 37 *C and sampled periodically. Samples were analyzed for TMZ content by the
HPLC methods described above.
37
3.2.3 Solubility
Saturated solutions of TMZ were prepared in pH 1,2,3,4, and 5 buffers and in pH 7
HPLC/MS grade water. Prior to mixing with TMZ, each buffer was analyzed using the HPLC
method described above to determine the presence of any interfering endogenous peaks. No
interfering peaks were found.
buffer solution at 37 OC.
Approximately 10 mg of TMZ were added to 500 pL of each
Each solution was left for 20 minutes at 37
0C
with intermittent
vortexing. Solutions were then spun at 10,000 RPM for 5 minutes. Precipitate was present in
all samples. Samples for HPLC analysis were prepared from the supernatant at the following
dilutions; pure supernatant, 1:5, 1:10, and 1:20. All dilutions were made with the appropriate
buffer pre-heated to 37
0C.
Samples that read in the linear range of the standard curve were
used to calculate solubility by adjusting the read concentration by the appropriate dilution factor.
3.3 Device Filling
3.3.1 Fuse Study Devices
Release study devices were assembled in the following fashion. The wafers were diced,
creating chips that each contained one membrane. The chips were affixed to Pyrex@ reservoir
pieces via UV cured epoxy. The Pyrex reservoir pieces are 2.2 mm thick, 6.5 mm long, 5 mm
wide, and drilled to define a cylindrical reservoir with a 3.5 mm diameter. The base layer was
affixed to the backside of the reservoir piece via UV cured epoxy. The devices were filled with a
motorized syringe unit (World Precision Instruments, USA) to provide a precise payload volume.
The input port was sealed with UV cured epoxy (Dymax, Inc) after filling the devices. The error
for volume loading was estimated as 1 pL. Copper wires were soldered directly on the electric
pads, and covered with UV-epoxy to prevent any contamination, and guarantee electrical
isolation.
38
3.3.2 TMZ Devices
Reservoirs were loaded with TMZ in solid form in order to maximize payload and drug
stability [37]. Polyethylene glycol (PEG) was added to displace air trapped within the packed
TMZ powder, therefore reducing air bubble formation within the reservoir. Molecular weight
1450 PEG was used because its melting temperature is between 43-46
0C.
The PEG,
therefore, is solid when implanted at body temperature, but the TMZ:PEG mixture can be
melted at moderate temperatures [37].
The process begins by packing the reservoir with TMZ powder. The reservoir is then
placed in a fixture that allows molten PEG to be pipetted onto the TMZ in each reservoir. The
second half of the fixture is then joined with the first half such that a high air permeability Teflon
AF membrane (Biogeneral Inc, San Diego, CA) is fixed over the PEG:TMZ surface. The
assembled fixture is placed in a vacuum oven and vacuum is pulled at 55
0C
for 20 minutes.
When the two halves of the fixture are joined the only path for air to travel is through the
membrane and out of the fixture. The TMZ and PEG are therefore trapped within the reservoir,
but air is free to flux out. Under vacuum and 55
0C
conditions, the air within the powder drug
fluxes out of the reservoir through the Teflon membrane. The molten PEG is then free to wet
the TMZ powder and fill in the interstices of the packed TMZ, therefore creating a homogeneous
mixture of TMZ and PEG throughout the reservoir. This process is repeated (typically 3 times)
until all of the air has been removed from the PEG:TMZ mixture.
3.4 Device Assembly
Polyimide coated copper leads were attached to each chip by biomedical grade
cyanoacrylate (Loctite Intl.).
A precision machined fixture was designed and fabricated to
ensure that the gold coated copper leads aligned precisely with the copper pads of the chip and
39
that a robust bond was made. Each copper lead was gold wire bonded to its corresponding
gold pad on the chip thus achieving electrical connectivity.
The chip:lead assembly was then placed on the shoulder structure of the reservoir such
that the upper most walls of the reservoir surround the chip. A bead of biomedical grade UV
curable epoxy (Dymax Corp., Torrington, CT) is then run around the perimeter of the chip such
that it fills the gap between chip and reservoir wall. The epoxy was then cured in place with a
compatible UV light source (Dymax Corp., Torrington, CT).
3.5 Release procedures
Fuse study devices were filled with known amounts of radioactive (C-14) mannitol
(Moravek Biochemicals Inc., Brea, CA) and activated to initiate release. Once activated, each
device was placed in a 4 milliliter (ml) stirred water bath at room temperature. Samples were
taken periodically to quantify the amount of radioactive material in the bath. Fresh water was
added to maintain the bath volume at 4 ml. Samples were mixed with scintillation fluid and
analyzed for radioactive content using a Packard Tri-Carb 2810 TR liquid scintillation counter
(Perkin Elmer, Waltham, MA).
TMZ devices were filled with known amounts of drug and then activated and placed in
agitated baths at 370C.
Release studies were conducted in both water and artificial
cerebrospinal fluid (Adapted from Alzet). The bath volume and frequency of sampling was
varied to ensure that samples contained quantifiable amounts of drug, but maintained
approximate sink conditions.
3.6 In vivo Studies
Female Fisher 344 rats weighing 125-175 grams were purchased from Charles River
Laboratories (Wilmington, DE). All animals were given free access to food and water at all
40
times. All animals were housed in accordance with the Johns Hopkins University Care and Use
Committee rules and regulations. Animals were intracranially implanted with 9L gliosarcoma
obtained from the UCSF Tumor Bank (San Francisco, CA) that has been passaged
continuously in carrier flanks of F344 rats.
The intracranial tumor and/or device implantation method used, as previously detailed by
Brem et al.[28], is briefly described here. Animals were anesthetized with an intraperitoneal (i.p.)
injection of 3-5 ml/kg of a stock solution containing ketamine hydrochloride 25 mg/ml (Ketlar;
Parke-Davis Corporation Morris Plains, NJ), xylazine 2.5 mg/ml (Rompun; Mobay Corp.,
Shawnee, Kansas), and 14.25% ethyl alcohol in 0.9% NaCl.
All surgical procedures were
carried out using standard sterile surgical technique. The head was shaved and prepared with
alcohol and prepodyne solution. A midline scalp incision was made, exposing the sagittal and
coronal sutures. A small burr hole was drilled, centered 3 mm lateral to the sagittal suture
(avoiding the sagittal sinus) and 5 mm posterior to the coronal suture. The 9L tumor and/or
device was implanted and animals were then randomized into various treatment groups. The
incision was then closed with staples and the animal allowed to recover.
3.6.1 Biocompatibility study
All techniques described above were used in the biocompatibility study. The animals,
however, received only implantation of the unactivated device (no tumor). This study involved 5
rats, and weight was used as a proxy for health. The rats were sacrificed on predetermined
days 1, 2, 3, and 7 post-implantation and autopsies performed to determine whether the chips
caused toxic side effects.
3.6.2 Efficacy Studies
A 9L tumor piece and the TMZ-filled device were implanted on day 0 and animals were
assigned to various treatment groups. Devices were activated on the specified days (either day
41
0, 3 or 5) post-tumor implantation by anesthetizing the animal, removing the staples, locating
the electrical leads, and applying a brief electrical pulse. The incision was then closed with
staples and the animal allowed to recover. Polymer wafers were implanted on day 5 after
anesthetizing the animal and removing the staples. The wafers were implanted through the burr
hole, the incision was stapled closed and the animal was allowed to recover.
Animals were observed daily for behavior, gait, grooming, weight loss, and mobility.
Animals were euthanized when they became moribund and the MEMS device was removed for
analysis and the brain was placed in formalin or flash frozen for subsequent histological and
immunohistochemical analysis.
3.7 Immunohistological analyses
Rat brain tissues were fixed with 4% paraformaldehyde at 4'C overnight. Tissue
samples were washed with phosphate-buffered saline (PBS), and then embedded in paraffin.
Longitudinal brain sections of 5 pm thickness were prepared for further immunohistological
analyses. Tissue slides were stained as previously described [50]. Briefly, tissue slides were
blocked with 3% goat serum in PBS for 30 min, and were incubated with polyclonal rabbit antiCaspase-3 (Abcam, UK, 1:200 dilution) or polyclonal anti-Ki67 (Abcam, UK, 1:200 dilution)
antibodies at 4'C overnight. After washing with PBS, a secondary anti-rabbit Alexa-488
(Invitrogen, US, 1:500 dilution) antibody was added onto the tissue sections and incubated for
45 min. Sections were washed with PBS, followed by counter-staining with DAPI. The stained
samples were mounted with the Vectashield mounting medium (Vector Laboratories, US) and
stored at 4'C until further analyses. Images were captured with 20x objective by Zeiss
AxioPlan2 fluorescent microscopy. Three representative images were taken for each sample.
Adobe Photoshop was used to quantify the positive signal.
42
4. Fuse Activation Mechanism
Reducing the energy requirements of the device is a crucial step toward miniaturizing
the device and ensuring its efficacious function in vivo. Peripheral device components, such as
the copper leads that conduct energy to the device, can be reduced in size as the activation
energy is reduced, contributing to the overall device miniaturization.
Tantamount to device
miniaturization is reducing the impact of activation on the host tissue and active therapeutic.
Activation will result in some extent of thermal energy transfer to the surrounding tissue and
payload.
Reduction of the total energy supplied to the device upon activation will result in
decreased exposure of tissue to elevated temperatures, therefore reducing the possibility of
tissue damage. A new activation mechanism aimed at reducing the energy requirements of
device activation was developed. The premise of the mechanism is utilizing a stark difference in
the thermal expansion of two materials to rupture a sealing membrane, thus initiating release.
This premise was reduced to practice by depositing thin gold fuses over suspended silicon
nitride membranes. Activation causes rapid resistive heating of the fuse material, creating large
stresses on the suspended structure leading to membrane rupture and reservoir exposure.
4.1 Device Manufacture
Silicon wafers containing 80 by 80 pm silicon nitride membranes were manufactured
(3.1.1). Gold fuses were then patterned over the membranes, traversing the center of the
suspended structure. The fuse dimensions were varied: thicknesses of 100, 250 and 500 nm;
and widths of 20, 40 and 60 pm.
Devices with 300 by 300 pm membranes were manufactured to achieve relevant release
rates of temozolomide. The results of the small membrane study were used to design fuses for
43
A
B
Figure 4.1 Digital renderings of the suspended nitride membranes and overlying fuses. A.
Simple rectangular fuses (yellow) were deposited over the nitride (green) membrane in the small
membrane study. B. More intricate fuse geometries were studied in the large, 300 pm, membrane
devices. Two 20 pm fuses, one 40 pm fuse, two 40 pm 'v' shaped fuses, three 10 pm fuses, and one
40 pm diagonal fuse.
the large membrane devices. The fuses in this study were set at 250 nm thick, but the geometry
was varied to determine which configuration resulted in the largest area of membrane rupture.
44
Figure 4.1 B depicts the five fuse layouts that were studied. They are: a single 40 pm wide fuse
traversing the center of the membrane, two 20 pm wide fuses that divide the membrane into
three equal areas, three 10 pm wide fuses dividing the membrane into quarters, one 40 pm fuse
traversing the membrane diagonal, and two 40 pm fuses patterned in 'V' shapes across the
membrane (these fuses will be referred to as 'x' fuses for the remainder of this document).
4.2 Activation Experiments
4.2.1 80 pm membranes
The fuse devices were placed inside of a transparent plastic container to simulate device
activation during release. The membrane was fully immersed in water preventing any bubbles
from affecting the experimental results. A probe station (Signatone, USA) was used to connect
devices to a high power pulse generator (Hewlett Packard, USA). The current was monitored as
the voltage drop across an in series resistor and was captured using an oscilloscope (Tektronix,
USA) as the pulse was applied. The membrane structures were visually inspected during the
experiment using the probe-station stereoscope in order to check if the applied pulse provided
enough energy to burst the membrane. The optimization step took place by reducing the voltage
necessary to burst a membrane, and then by reducing the width (ps) of the pulse. Open circuit
conditions were verified by measuring the electrical resistance using a standard ohm-meter.
Table 4.1 provides a summary of the experimental results, showing the average values
per fuse type. Figures 4.2-4.3 show a series of SEM, and confocal microscope images of
membranes per fuse type for a 250 nm thick gold layer prior to, and post activation. The
effective area for the 20 pm wide fuse is critically smaller than for the 40 pm and 60 pm wide
fuses. The images clearly show that the membranes ruptured, and that the fuses melted. Figure
4.3 also shows 3D confocal images of the membranes after fuse activation. The 20 pm wide
fuse was partially burst and deformed; whereas for the 40 pm and 60 pm wide fuses, the
45
membranes were fully burst. Figure 4.4 shows the voltage and current traces for a 250 nm thick
gold layer for each fuse width, captured using the oscilloscope. The pulse height and width of
the current decrease as a function of width. Figure 4.5 shows a three dimensional plot of the
energy required to burst a membrane as function of fuse dimensions. It is possible to observe
the energy required to burst the membrane scales down as function of width and thickness. The
voltage and time was also indicated in the 3D plot. 40 pm wide fuses were chosen as the
optimal for the release characterization, they are capable of bursting the entire membrane
consistently, and maximize the flux area.
46
Figure 4.2 SEM microscope images showing the structure before and after activation for
each fuse width (250 nm Au layer). A. 20 pm wide. B. 40 pm wide fuse. C. 60 pm wide.
47
-~
C.
Figure 4.3 Confocal microscope images for 250 nm thick fuses. A. 20 pm wide. B. 40 pm wide.
C. 60 pm wide.
Tek
J1
e Acq
owteste
M Pomo s00
Tek
8V
JU
acqfCempMet.
PU.
am ,
e
10 V
0.5
OA
_12
V
m
0075 A
70 ps
0 eq
A Ctwe, MPw1044os
J
.9A 0
180 ps
480 p
A.
B.
C.
Figure 4.4 Oscilloscope capturing images showing activation pulse voltage values (blue)
and current (red) values for 250 nm thick fuses. A. 2 pm wide. B. 40 pm wide. C. 60 pm wide.
Table 4.1: Values depicted in Figure 4.5. n = 5.
Number
Width
Thickness
Rmeasured
(PM)
(nm)
(Q)
500
Time(ps)
Energy (pJ)
0.5
970
1409
250
0.68
250
347
3
100
1.7
58
40
4
500
0.52
500
512
250
1.01
170
134
6
100
1.7
65
30
7
500
1.27
94
78
250
1.72
66
40
100
3.95
18
14
1
2
5
8
9
20
40
60
49
6
10
500
0
250
100
Figure 4.5 Minimum energy to burst fuse as a function of fuse dimensions. Please refer to
Table 4.1 for values.
50
4.2.2 300 pm membranes
The procedure for testing the activation of 300 pm membranes was the same as above.
The two fuse configurations that resulted in the largest amount of ruptured membrane area were
diagonal and x fuses. Characteristic photos of the resulting membrane rupture are depicted in
Figure 4.6. The x fuse configuration resulted in the largest ruptured membrane area. The
diagonal fuses appear to have been incapable of creating large enough stresses in the corners
of the membrane that were not traversed by the fuse. The x fuse, however, appears to have
created more uniform stresses that were capable of rupturing the membrane. The energy
required to rupture the 300 by 300 pm membranes with an x fuse is approximately 200 pJ.
B.
-
Figure 4.6 Light microscope images of ruptured 300 pm membranes. A. Membranes ruptured
via x fuses. B. Membranes ruptured via diagonal fuses.
51
4.3 Release Experiments
The viability of the fuse mechanism as a mechanism for controlled release was tested by
assembling 80 pm membrane devices with Pyrex@ reservoirs. The assembled devices were
filled with a radio-labeled mannitol-C1 4 (Moravek, USA) water solution with a syringe microinjector. The devices were activated using the high pulse voltage generator at 2 V and pulse
width of 250 ps, introducing a factor of safety of approximately 400 percent from the
experimental optimization results. Immediately after the devices were activated, the resistance
of the fuse was measured to verify open circuit condition, which was verified for all the activated
devices. The release experiments consisted of using 10 devices, 5 served as un-activated
controls and the remaining 5 were activated for release characterization. Each device was
placed into a stirred 4 milliliter de-ionized water bath at room temperature. The baths were
sampled regularly and the extent of release was quantified using a liquid scintillation counter
(Perkin Elmer, Waltham, MA). Figure 4.7 is a plot of the average extent of release versus time
of the activated and control devices. The amount released by each device was normalized to its
original payload and plotted as a function of time. The control devices remain at baseline levels
for the duration of the study. The activated devices all demonstrated reproducible, diffusion
driven release. The relatively small amount of error present in the release curves demonstrates
that gold fuses are capable of rupturing underlying nitride membranes reliably and in a manner
that produces a consistent aperture, providing a very repeatable flux.
52
1.0
-
Control
-- Release
0.8
0.6
0.4
LL0.2
0.0
0
50
100
150
200
250
Time (hrs)
Figure 4.7 In vitro release profiles, n = 5, for activated devices, n
Error bars represent standard deviation.
=
5 for control devices.
4.4 Mechanism Discussion
The fuse mechanism was simulated using a FEA model to provide an analysis of the
mechanical behavior of the suspended membrane. Figure 4.8 provides a unique insight to
understand how the mechanical dimensions of the fuse makes an impact on the mechanical
deformation of the membrane. The Von Misses stresses and displacements were numerically
calculated and plotted. The fuse acts as an electro-thermal actuator that allows for rapid
mechanical expansion of the silicon nitride membrane. Figure 4.8 also shows the lateral view of
the displacement per fuse type as the fuse temperature is defined as the melting point of gold.
Three displacement peaks are clearly noticed. The first peak shows how the bending stresses
introduce deformation in outward direction. The second and third peaks show how the bending
stresses introduce deformation in the inward direction. An interesting aspect of the simulation
53
shows how the magnitudes of the upper peak displacement and the magnitude of the lower
peak displacement increase as a function of the increasing width. It is important to notice the
distance between peaks is proportional to fuse width, forcing the peaks closer to the edges as
the width increases. Figure 4.8 also reveals that the peak sharpness decreases as a function of
the increasing width. The simulations provide an understanding for optimization of the fuse
width: if the fuse is too narrow, the displacement peak will be sharp but the stress magnitude will
not be as high as the widest fuse. If the fuse width is too wide, too much energy will be
inefficiently used.
A.
B.
C.
Figure 4.8 FEA Analysis of the suspended structure showing an isometric view of the Von
Misses stress on the surface area of the membrane, and a lateral view for displacement
showing structural deformation. The mesh of the analysis included approximately 10E5 elements.
A. 20 pm wide fuse (B), 40 pm wide fuse (C), 60 pm wide fuse (D).
It is important to notice that the quasi-static analysis was implemented defining a
constant temperature along the surface area of the fuse. Figure 4.5 shows a perspective on the
experimental optimization providing a selection criterion for the device dimensions for reliable in
vitro and in vivo experiments. The thickness plays a key role as the resistance increases with
decreasing cross-sectional area, resulting in an increase of the current density. Although less
energy is required to melt the fuse, more energy is required to achieve plastic deformation. As a
result of this optimization, the chosen fuse dimensions for the release experiments were the
intermediate values, 40 pm wide and 250 nm thick, in order achieve a wide and sharp
deformation peak that would provide enough energy to completely burst the membrane with the
lowest energy.
The previous device generation required approximately 1 Joule for activation. The
energy requirements for activating the fuse mechanism range from 9 pJ to 640 pJ depending on
the fuse geometry (width and thickness), which represents at least a 4 order of magnitude
improvement. We utilized the geometry of 40 pm by 250 nm as the exact middle point to
achieve low activation energy as well as satisfying the requirement of complete membrane area
opening.
The very tight tolerances during manufacturing process assured consistent resistor
values that provided repeatable results for the release experiment. The release curves shown in
Figure 4.7 provide valuable information on the process that follow diffusion curves from a point
source. Figure 4.7 shows the release experiments during 200 hours of delivery. The control
curves demonstrated that the devices were sealed, and did not provide any significant leakage.
The tight variations within profiles guarantee a controlled release for drug delivery devices that
could be used as novel therapeutic modalities for a wide number of applications.
55
4.5 Conclusion
The motivation for this work was to explore a fuse activation mechanism that operates
via electro-thermally induced structural failure for miniaturized implantable drug delivery devices
capable of on demand activation for controlled release at low energy consumption. The role of
fuse dimensions was closely investigated to optimize energy consumption. Two metrics: extent
of membrane rupture, and energy consumption, were used to evaluate the performance of each
fuse type. The fuse mechanism has been optimized to reduce the energy consumption, while
maximizing the aperture area for flux. Drug delivery microdevices implemented with this
mechanism released C14 mannitol in a controlled, reproducible fashion.
56
5. In Vitro Characterization of Device
Performance
In vitro release studies were conducted to characterize the release kinetics of TMZ from
microchip devices. These studies performed the crucial function of validating and quantifying
several factors affecting the function of the device. The most important factors are: verifying
that drug releases reliably, drug only releases from devices that have been activated, and that
the release rate can be varied by changing the number of ruptured membranes. Preliminary in
vitro studies revealed that drug co-formulation is necessary. A method for co-formulating TMZ
with polyethylene glycol within the device was developed. Devices containing co-formulated
TMZ are capable of reliable release in water and artificial CSF and the release rate can be
varied by varying the number of ruptured membranes.
5.1 TMZ Formulation
Devices that were packaged (filled with drug and sealed) in atmospheric conditions
released minimal amounts of drug in 370C water. Drug release was not only slow, but erratic
with prolonged (hours to days) periods of no release. The thought is that air trapped in the
interstices of the packed powder forms a bubble in the reservoir, occluding the orifice and
preventing release. Loading drug in solid form is desired because it maximizes the mass that
can loaded into a given volume and is the most stable form of the drug [37].
A method,
therefore, must be developed for co-formulating TMZ with another material such that the air
within the interstices of the drug is displaced.
5.1.1 Packaging in CO2
Preliminary efforts focused on altering the composition of the atmosphere in which the
devices were packaged such that the atmosphere was primarily CO2. Packaging devices in CO2
57
12
11
10
'~'tk
-<a- Releasel
--4- Release2
--<- Release3
Control1
S-+-4-- Control2
8
-- Theoretical
7 -
E
6-
IN5
3
2
1
0
0
50
100 150 200 250 300 350 400 450
Time (hrs)
Figure 5.1 An example of inconsistent release profiles that result from packaging in C02.
The rate of release was consistent across the three devices and compares well to the theoretical
release. The onset of release and overall extent of release vary greatly across the 3 devices. The
theoretical line is based on zero order approximations of the release rate utilizing Fick's first law. The
controls demonstrate excellent sealing and drug retention throughout the time scale for release.
lead to extensive drug release, but the release curves were still not reproducible. Figure 5.1 is a
plot of TMZ release from devices packaged in CO 2 and with a small amount of PEG melted onto
the backside of each membrane5 . Two of the three devices required a 24 hour induction period
before TMZ began releasing whereas the third device began releasing immediately. The overall
extent of release also varied greatly between the devices with a range of 4 to 8 milligrams being
s PEG was melted onto the backside of each membrane to serve as an impetus for water influx.
58
delivered. Co-formulation with CO2 succeeded in increasing the overall extent of release, but the
method proved unreliable.
5.1.2 Co-Formulation of TMZ and PEG
Packaging in CO2 was deemed too variable, likely due to the difficulty inherent in
controlling the atmospheric composition during filling, assembly and sealing. Co-formulation of
TMZ with solid PEG was the next method attempted.
The critical factors to the PEG co-
formulation process are: the PEG must wet and fill the interstices of the packed TMZ powder
thus displacing any trapped air, the process must be reproducible, TMZ payload cannot be lost
or destroyed, the process must be such that the resulting PEG:TMZ filled reservoir is still
compatible with subsequent device assembly steps (Figure 5.2).
59
Initial efforts focused on creating mixtures of PEG and TMZ in bulk and then transferring
the mixture to each reservoir. Difficulties inherent in handling and transferring the mixtures
resulted in variable device payloads and uncertainty in the precise amount of drug in each
reservoir. A method for co-formulating PEG and TMZ within the reservoir was developed to
reduce payload variability and improve the device packaging and assembly process.
B
A
C
Figure 5.2 Renders of the device in the three stages of assembly. A. An empty reservoir. The
shoulder that the chip must rest on is visible around the interior perimeter of the reservoir. B. A filled
reservoir. Drug and any excipients must be loaded into the device without accumulating on the
shoulder in order to ensure proper alignment during subsequent assembly steps. C. An assembled
device. The chip is placed in the reservoir, resting on the shoulder. Biomedical grade epoxy is then
used to secure the chip to the reservoir and seal the reservoir.
60
AE
Figure 5.3 A series of color photographs depicting the steps in device filling. An empty
reservoir is placed in a fixture that allows a pre-weighed amount of drug to be added up the the level
of the shoulder (A). Drug is loaded into the reservoir and partially compacted to create a small well
within the drug powder. This well allows molten PEG to be added to the reservoir (B). After typically
three repititions, the majority of the air has been displaced from the TMZ powder and replaced by
PEG. The vacuum fixture ensures that TMZ and PEG do not escape the reservoir or accumulate on
the shoulder (C).
The final process begins by packing the reservoir with a known mass of TMZ powder.
The TMZ filled reservoir is then placed in a fixture that allows molten PEG to be added to each
reservoir (Figure 5.3 B). The second half of the fixture is then joined with the first half such that
a high air permeability Teflon@ membrane is fixed over the TMZ:PEG surface. When the two
halves of the fixture are joined, the only path for air to travel is through the membrane and out of
the fixture (Figure 5.4). The drug and PEG are therefore trapped within the reservoir, but air is
free to flux out.
61
TMZ:PEG
LCP Reservoir
Teflon AF Membrane
Fixture
Figure 5.4 Schematic of the vacuum co-formulation process. The vacuum fixture clamps a high
air permeability Teflon AF membrane over the surface of the PEG:TMZ mixture. The PEG:TMZ
mixture is trapped within the reservoir, but any air is free to flux out when a vacuum is pulled. Once
the air is removed, PEG is free to wet within the interstices of the TMZ powder.
The assembled fixture is placed in a vacuum oven and vacuum is pulled at 55 *C for 20
minutes.
55 *C is above the melting point of PEG molecular weight 1450, but a moderate
enough temperature that TMZ is still stable [37].
The mechanism, therefore, is that under
vacuum the air within the powder drug fluxes out of the reservoir through the Teflon membrane.
The molten PEG is free to wet the TMZ powder and fill the interstices after the air has been
removed. The fixture is chilled in the freezer for 5 minutes after each 20 minute session in the
vacuum oven to ensure that the PEG solidifies before disassembling the fixture.
Three
iterations of adding PEG and heating under vacuum are usually necessary before all of the
trapped air has been displaced by PEG (Figure 5.3 C). Stability studies confirmed that no TMZ
was lost or degraded during this process.
procedure.
Briefly, several devices underwent the above
Devices were removed at different points of the procedure and after varying
numbers of iterations.
Each device was then assayed for TMZ content via HPLC.
The
recovered amount of TMZ as determined by HPLC was compared to the initial TMZ loading.
Minimal loss of TMZ occurred.
62
5.2 In vitro Release Studies
In vitro release studies were conducted to characterize the release kinetics of coformulated TMZ from the active device. Briefly, devices were placed in agitated baths and
activated. Periodically the devices were moved to new baths to allow for the quantification of
TMZ in the old baths. The bath volume and frequency of sampling was varied depending on the
number of membranes activated in an effort to maintain approximate sink conditions within the
bath, but still achieve quantifiable levels of TMZ. These studies were conducted on groups of
devices with either 3, 2, 1 or 0 membranes ruptured to quantify the effect of varying total orifice
area.
M--
3 Membrane
2 Membrane
-+- 1 Membrane
-+- Control
-A-
10
8
0>
6
E
N
4
20
0
10
20
30
40
50
60
70
80
Time (hrs)
Figure 5.5 Release curves for TMZ filled devices releasing into 37 0 C water. Devices (n=3)
with 3, 2, 1 or 0 membranes activated were placed in agitated baths and sampled for TMZ content.
The errors bars are the standard error. Release is a function of the number of activated membranes
and release does not occur unless the device is activated.
Figure 5.5 is a plot of the release of TMZ into water as a function of time. Release
curves are shown for devices with 3, 2, 1 and zero membranes activated. Devices with zero
membranes ruptured serve as controls to verify that TMZ release is dependent on device
activation. These 'leak test' devices remained robustly sealed over the time scale required for
the 3 and 2 membrane activated devices to release their payload, and an equivalent time scale
for the 1 membrane activated devices to release their payload. The leak test devices ultimately
began leaking after approximately 700 hours (-30 days) in 37 *C water. It is important to note
that all of the membranes remained intact in these devices and that leaking therefore reflects a
degradation and failure of the sealing epoxy6.
The activated devices display reproducible release kinetics that vary with the number of
membranes activated. Three membrane activated devices release with an average rate of 0.3
milligrams per hour (mg/hr) achieving a final release of 90 ± 3.2 % in approximately 30 hours.
Two membrane activated devices release with an average rate of 0.136 mg/hr, achieving a final
release of 82 ± 1.9 % in about 60 hours. One membrane activated devices release with a rate
of 0.007 mg/hr achieving an overall release of 60 ± 12 % over roughly 800 hours (-34 days).
It is interesting to note that introducing PEG appears to introduce a different mechanism
or driving force for release. Zero order approximations, like those done in Chapter 2, predict
that 10 milligrams of TMZ should diffuse out of a 3 membrane activated device in approximately
350 hours. The data in Figure 5.5, however, show that the duration of release is approximately
30 hours. When PEG is not present, zero order release rate approximations generally agree
with experimental release results as evidenced by the comparable slopes in Figure 5.1.
The epoxy used to seal these devices was UV curable epoxy 1180-M (Dymax Corp.). This epoxy was
selected
because it is capable of binding the relevant materials, it is biomedical grade and has lowest water absorption
properties available.
64
6
The
introduction of PEG, therefore, likely introduces a new mechanism of release, drastically
increasing TMZ release rate.
It is possible that differences in release rates between PEG co-formulation and C02 coformulation is a result of a change in the rate of water influx and the results of water influx. The
following are the posited fundamental steps of TMZ release from the device in the absence of
PEG:
1.) Activation allows water to diffuse into the device
2.) The water wets and dissolves the TMZ
3.) Dissolved TMZ diffuses from the device at a rate dictated by the concentration
gradient and area available for flux.
The release rate in this mechanism is not dependent on the rate of water influx as long as it is
faster than TMZ dissolution and diffusion.
The introduction of PEG, however, may introduce another mechanism of TMZ release.
The fundamental steps of TMZ release in the presence of PEG are thought to be:
1.) Activation allows water to diffuse into the device
2.) The water wets the TMZ:PEG formulation allowing TMZ to dissolve and causing
PEG to swell
3.) The swelling of PEG within the confinements of the device introduces a mechanical
driving force (forced convention) for TMZ release from the device.
The rate of water influx in this mechanism would dictate the rate at which the PEG swells and
therefore partially determine the rate of TMZ release from the device 7. The relatively high
solubility of PEG 1450 (-100 mg/ml) will cause the rapid influx of water and therefore rapid
The number of membranes that have been activated will not only affect the rate at of water influx, but also affect
the rate at which TMZ:PEG is released.
65
swelling of PEG. The resulting convection of dissolved TMZ and possibly TMZ:PEG particulate
results in the very rapid release of TMZ.
Several observations support this mechanism. First, devices with payloads of 8, 9 and
10 milligrams all release their payload in approximately 30 hours when 3 membranes are
activated. If simple diffusion of TMZ from the device was rate controlling then the 8, 9, and 10
mg devices would release with the same rate, but with varying overall durations of release.
Experimental results, however, display the opposite trend, where the overall duration of release
is consistent and the rate varies. If the rate of water influx was rate determining then the
release rate of TMZ should be largely independent of TMZ payload and depend predominantly
on the number of membranes activated and the presence of PEG.
The second observation is the non-linear dependence of release rate on the number of
membranes that have been activated. The 3, 2, and 1 membrane activated devices release
with average rates of 0.3, 0.13 and 0.007 milligrams per hour respectively. Theoretically three
membrane activated devices would release 1.5 times faster than two membrane devices and 3
times faster than one membrane devices. The three and two membrane devices follow this
general trend when compared to each other, but both the three and two membrane devices
release much faster than would be predicted based on the one membrane device. The 1
membrane devices release with a rate of 0.007 mg/hr (60 % of their payload in
-
800 hours).
Zero order diffusion approximations predict a comparable release rate of 0.011 (80 % over 800
hours). Release from one membrane devices, therefore, appears to be governed by simple
diffusion of TMZ from the device whereas two and three membranes devices are
mechanistically different. It is possible the area for flux between 2 activated membranes and 1
activated membrane bridge some threshold value above which water influx is dominating, and
below which simple diffusion controls release rate. It is possible that the number of membranes,
as opposed to area for flux, is dominating in switching between mechanisms. The author at this
66
time, however, cannot reason out a plausible physical explanation for this scenario. The exact
origins of the release mechanism are not fully understood. The effects, however, are beneficial
as animal studies in Chapter 6 indicate that more rapid release results in improved survival.
In vitro release studies verified the reliable function of the active device. There are three
criteria for 'reliable function.' First, TMZ release should occur only when the device has been
activated. The leak test devices satisfied this criterion by remaining robustly sealed over the full
time scale of the release study. It is important to note that all of the membranes remained intact
in these devices and that the eventual leaking of these devices, therefore, reflects a degradation
and failure of the sealing epoxy. Second, the rate of TMZ release should vary with the number
of activated membranes. The release curves presented in Figure 5.5 demonstrate that TMZ
release rate is indeed a function of the number of membranes that have been activated and that
a broad range of release rates can be achieved. Third, the kinetics of release should be
reproducible for each scenario of device activation.
Devices from each group displayed
reproducible average rates for release, and most importantly were very consistent in overall
extent and duration of release.
5.3 Conclusion
The aim of the work presented in this chapter was to develop methods of loading each
device with TMZ in powder form and to assay the function of the device via in vitro release
experiments. Methods of co-formulating TMZ had to be developed once 'neat' loading of TMZ
proved to be subject to variable rates and extents of release. Packaging the device in a CO2
environment improved the uniformity in release rate, but was still subject to variability in overall
extent of release and lag time before release began. A method for co-formulating TMZ with PEG
within the reservoir was developed. This method proved effective at removing any lag times and
in reducing the variability in release rate and extent of release. The inclusion of PEG increases
67
the rate of release from the device, and may create a mechanistic difference in the driving
forces for release. The release rate from the device can be varied by varying the number
ruptured membranes and the resulting range of release rates presents an interesting
opportunity for in vivo studies investigating the role of release rate in disease progression.
68
6. In Vivo Studies
This chapter is a description of the in vivo studies that were conducted with the
miniaturized microchip device. The overarching aim was to demonstrate the safety and efficacy
of the device in a rodent model of glioma. The impact of implanting and activating a device in
the brain of a healthy rat was assessed by monitoring the health of that rat. Pilot scale efficacy
studies were conducted prior to large scale efficacy studies. The effect of whether a faster rate
of drug delivery, as opposed to longer duration, resulted in better median survival was
investigated in a large scale in vivo study. The impact of the timing of drug delivery on survival
was evaluated. The microchip-based delivery method was compared to polymer-based delivery
of TMZ. Histological analysis of tissue samples obtained from the efficacy study was conducted
to confirm that TMZ retains its cytotoxic potential throughout the formulation, packaging and
release process.
6.1 Pilot in vivo studies
Several pilot scale in vivo studies were conducted prior to the full scale efficacy study.
The aim of these studies was to investigate different components of the device function and
experimental procedure on a smaller scale than would be required of a full scale efficacy study.
6.1.1 Potential toxicity and potential side effects
6.1.1.1 Effects of the intracranial presence of the device
The toxicity of implanting an unactivated microchip inside the brain of a rat was
evaluated. This study involved 5 rats, and used weight as a proxy for health. The rats were
sacrificed on predetermined days 1, 2, 3, and 7 post-implantation and autopsies were performed
to determine whether the chips caused toxic side effects. No gross abnormalities were noted on
the autopsies, and as apparent from Figure 6.1, the weights of the rats did not vary in the given
69
1.4 ,
-*-Rat1
1.3
-4-Rat 2
-0-Rat 3
0 1.2
-o-Rat 4
-0-Rat 5
0
1.1
N
1
0.9
0.8.
0
1
.
.
.
.
.
.
.
2
3
4
5
6
7
156
Day Post Microchip Implantation
Figure 6.1 Graphical presentation of normalized animal weight as a function of time. All
animals displayed minimal weight loss during the acute phase and animal 5 displayed robust weight
gain over the chronic time scale.
time period. Rat 5 was still alive at 156 days post-implantation and has gained a substantial
amount of weight.
6.1.1.2 Device activation in vivo
This pilot study had two animal groups (n=5) to test two primary objectives: first,
determine if the device is robust enough to survive the implantation procedure and long term
implantation; second, verify that fuse activation in vivo ruptures the membranes and does not
damage the surrounding tissue. The surgical and implantation procedure explained in 3.6 was
used for all animals except no tumors were implanted.
The first group of animals had TMZ filled devices implanted, but not activated.
The
electrical connectivity of each membrane was monitored during implantation with a standard
ohm-meter.
All membranes remained intact during implantation.
After the devices were in
place the polyimide leads were snipped to allow the incision to be stapled closed. Animals were
monitored for 5 days to determine if the presence of the device resulted in poor health or death.
70
All animals remained healthy for the duration of the study. The devices were explanted on day 5
and inspected for gross abnormalities (e.g. extensive clotting or adherent tissue) and membrane
patency. All membranes were intact upon explantation and no gross abnormalities were
observed. Explanted devices were opened and assayed for TMZ content. The full payload was
present in each device.
The second group of animals had TMZ filled devices implanted and activated. The
electrical connectivity was again monitored during implantation to determine if the membranes
survived implantation. Activation occurred after the devices were in place. The animals were
observed during activation for seizure or altered respiration rate.
None of the animals
demonstrated any gross physiological response to activation. Animals were again monitored for
the next 5 days. All animals remained healthy during the observation period. The devices were
explanted on day 5 and inspected. No gross abnormalities were observed and all membranes
were ruptured. Explanted devices were opened and assayed for TMZ content. Approximately
60% of the payload had released.
6.1.2 Preliminary survival studies
The second pilot in vivo study had 3 objectives: first, in a diseased animal, does the
presence of an un-activated device affect animal survival; second, does the activation of a
device alone affect survival; finally, does activation and drug release from the active device
improve survival over no treatment? This study contained four animal groups (n=5). All animal
groups received a tumor burden on day 0. One group served as a no treatment control and only
received tumor. The second group had artificial CSF filled devices implanted and activated.
The third and fourth groups had TMZ filled (8mg) device implanted, but only the group 4 devices
were activated. After device implantation and activations the incisions were stapled closed and
the animals were returned their cages for observation. The animals were checked daily for
signs of disease progression, specifically for lethargy, loss of motor skills, cachexia and death.
71
Figure 6.2 depicts the survival curves for each group.
Survival is depicted as the
percentage of animals that are still alive in each group as a function of time. Tumor control and
un-activated device animals all expired by day 19, with a median survival of 16 days. The
artificial CSF group all expired on day 12. Animals that received activated TMZ devices all
expired by day 51, with a median survival of 44 days. Table 6.1 summarizes the groups, n and
median survival for this study. Statistical analysis was performed using Prism GraphPad
software package and included Kaplan-Meier survival analysis tools.
-9L
---
100
Control
Unactivated Device
ACSF
Activated Device
90o-80
70-
60S50'
S40W 30"
20'10ef.
0
20
10
40
30
Time (Days)
50
60
Figure 6.2 Survival curves for the preliminary efficacy trial. Animals with activated devices
demonstrated the most prolonged survival. Un-activated devices had an identical impact on animal
survival as no treatment at all. Animals receiving aCSF had a slightly shorter median survival than no
treatment.
Table 6.1: Survival Data for Preliminary Efficacy Study
Median Survival (days)
Treatment Condition
n
No treatment
5
16
Artificial CSF
5
12*
Un-activated Chip
5
16
Activated Chip
5
44*
* p<0.05 compared to no treatment control
72
6.2 Large scale in vivo
A large scale in vivo study comprised of 9 animals groups was conducted to study the
impact of TMZ release kinetics on animal survival and to compare the microchip device to
polymer based delivery of TMZ.
6.2.1 Effect of TMZ delivery rate on efficacy
There were a total of 5 groups in this experiment: 9L control (tumor only), unactivated
device (tumor + device), 1 membrane opened on day 0, 2 membranes open on day 0, and 3
membranes open on day 0. This experiment examines whether a faster rate of drug delivery, as
opposed to a longer duration, results in increased median survival.
The survival curves are depicted in Figure 6.3. Long term survivors are those animals
surviving until the protocol mandated 120 day stop date. The no treatment controls and unactivated device groups had median survivals of 13 and 16 days, respectively (Table 6.2). The
100
90
.
80-706050403020100
0
-9L
-
Control
Unactivated Device
3 Membranes, Day 0
80--2 Membranes, Day 0
1 Membranes, Day 0
,
10
20
30 40 50
Time (days)
60
110 120
Figure 6.3 Impact of drug release rate on survival. Animals that received activated devices on day
0 had median survivals of 40 (42.8 % LTS), 28 (28.5 % LTS), and 21 (12.5 % LTS) days for 3, 2, and
1 membranes activated respectively.
73
0
.
100
--
9L Control
90
-
Unactivated Device
3 Membranes, Day 0
3 Membranes, Day 3
3 Membranes, Day 5
--
8070605040302010-
/
0
0
10
20
30
40
50
60
110 120
Time (days)
Figure 6.4 Impact of drug release time on survival. Animals that had all 3 membranes activated
day 0, 3, or 5 had median survivals of 40 (42.8 % LTS), 24 (12.5 % LTS) and 23 days.
3 membrane activated group had a median survival of 40 days with 42.8% long term survivors
(LTS). The 2 membrane activated group had a median survival of 28 days and 28.5% LTS. The
1 membrane activated group had a median survival of 21 days with 12.5% LTS.
6.2.2 Effect of TMZ delivery time on efficacy
There were a total of 5 groups in this experiment: 9L control (tumor only), unactivated
device (tumor + device), 3 membrane opened on day 0, 3 membranes open on day 3, and 3
membranes open on day 5. This experiment examines and quantifies the effect of earlier drug
delivery.
The survival curves are depicted in Figure 6.4. Again, the no treatment controls and unactivated device groups had median survivals of 13 and 16 days, respectively (Table 6.2). The 3
membrane day 0 group had a median survival of 40 days with 42.8% LTS. The 3 membrane
day 3 group had a median survival of 24 days and 12.5% LTS. The 3 membrane day 5 group
had a median survival of 23 days and no LTS.
74
6.2.3 Comparison of the device to a polymer-based delivery system
Lastly, the efficacy of microchip-based delivery is compared with polymer-based drug
delivery. There were a total of 5 groups in this experiment: 9L control (tumor only), unactivated
device (tumor + device), 3 membranes opened on day 5, and 2 TMZ polymer-based wafers
implanted on day 5. Both the device and the wafers contain 10 mg of TMZ and started to
release drug on the same day.
The survival curves are depicted in Figure 6.5. Again, the no treatment controls and unactivated device groups had median survivals of 13 and 16 days, respectively (Table 6.2). The 3
membrane day 5 group had a median survival of 23 days and no LTS, and the 2 TMZ wafer
group had a median survival of 34 days and LTS.
100
90
807060
504030201000
-
--
10
20
30
40
9L Control
Unactivated Device
2 TMZ Wafers
3 Membranes, Day 5
50
60
Time (days)
Figure 6.5 Comparison between microchip and polymer-based delivery methods. Those animals
that received two TMZ:polymer wafers on day 5 had a median survival of 34 days, while those that
had all 3 membranes opened on day 5 had median survival of 23 days.
75
Table 6.2: Survival data for the large scale efficacy studies
Treatment Condition
n
Median
% Long Term
Survival (days)
Survivors (n)
No Treatment
8
13
0
Un-activated Device
8
16*
0
Two TMZ Wafers
8
34*
0
3 Membranes Activated, Day 0
7
40*
42.8 (3)
2 Membranes Activated
7
28*
28.5 (2)
1 Membrane Activated
8
21*
12.5 (1)
3 Membranes Activated, Day 3
8
24*
12.5(1)
3 Membranes Activated, Day 5 8
23
*p < 0.05 when compared to no treatment control
0
6.2.4 Immunohistological analyses
Tissue samples were collected from animals that expired at or near the median survival
date of that group. All treatment groups have elevated levels of cleaved Caspase-3 positive
cells with the longest surviving group, 3 membranes activated on day 0, showing the second
highest numbers of positive cells. The two longest surviving groups, 3 membrane day 0 and 2
TMZ wafer groups, have the fewest number Ki67 positive cells.
76
Device
A
B
Device
10
0O0
4
Figure 6.6 lmmunohistological results of Ki67 and caspace-3 staining. (A) Ki67 positive cells are
green and cell nuclei are blue. Each panel is a representative image from each efficacy study group.
The 3MOD (3 membranes activated on day 0) panel contains the fewest number of Ki67 positive cells.
(B) Quantitative results obtained by averaging 3 representative images from each group. The two
longest surviving groups, wafer and 3MOD, have the lowest levels of Ki67 positive cells. (C) Caspase3 positive cells are green and cell nuclei are blue. Each panel is a representative image from each
efficacy study group. (D) Quantitative results obtained by averaging 3 representative images from
each group. The 3MOD (3 membranes activated on day 0) panel contains the second highest number
of caspace-3 positive cells.
77
6.3 Discussion of in vivo results
Previous studies using the MEMS device had been restricted to rodent flank tumor
models, due to large structural components. The adoption of injection molding technology, solid
drug loading, and a redesign of the device geometry allowed for an 80 percent reduction in
device volume for the same payload [47]. This is the first study to successfully demonstrate the
efficacy of an active, intracranial microchip-based drug delivery system.
6.3.1 Preliminary in vivo studies
The device was demonstrated to be robust enough to survive the implantation procedure
and remain in the animal long term. The electrical resistance across the fuses was measured
after implantation and verified that the fuses were intact. It was also verified that fuse activation
in vivo neither adversely or beneficially affects the survival of animals simultaneously implanted
with a tumor. The active device therefore can be used in vivo without diminishing the precise
control over the timing of TMZ release or introducing competing effects in the analysis of
survival in disease studies.
6.3.2 Efficacy Studies
The objective of this study was to determine the efficacy of TMZ delivered from the active
device under different conditions. All treatment groups except the 3 membrane, day 5 group had
significantly improved survival over the no treatment controls (p < 0.05, Table 6.2). The overarching objective of designing a MEMS based active device capable of intracranial implantation
and efficacy in rodent gliosarcoma model was achieved.
The ability to activate different
numbers of membranes at different times leads to some interesting insights into the function of
this device and how disease progression may be deterred.
78
6.3.2.1 Delivery rate
Each treatment group activated on day 0 had improved animal survival over the no
treatment control. The p value when each day 0 treatment group is compared head to head is
greater than 0.05, but there are 3 strongly indicative trends in the survival data that imply that a
dose dependent type response is present within the day 0 groups. The survival curves (Figure
6.3) are all nested in order of device release rate with minimal overlap. The median survival of
each group also trends with release rate. The day 0 groups have median survivals of 44, 28
and 21 days for 3, 2 and 1 membranes activated respectively. Finally, the percentage of long
term survivors increases with the number of activated membranes (1 membrane: 12.5 %LTS, 2
membranes: 28.5 %LTS, 3 membranes: 42.8 % LTS,). These three trends when considered
together indicate that rapid release, as opposed to longer duration, is more effective at retarding
disease progression in this model.
6.3.2.2 Delivery time
Head to head comparisons within the 3 membrane activated groups, again yields p
values of greater 0.05, but similar trends emerge with date of activation as with number of
activated membranes. The survival curves (Figure 6.4) are nested in order of activation date
with groups receiving earlier activation faring better. Both median survival and the number of
long term survivors increase with earlier activation: Day 5: 23 days, 0 % LTS, Day 3; 24 days,
12.5 % LTS, Day 0; 40 days, 42.8 % LTS.
The decrease in median survival and long term survivors with delayed activation can be
explained by two basic arguments concerning tumor growth and device function. The longer the
delay between tumor implantation and the initiation of treatment the more time the tumor has to
grow. The TMZ distribution achieved by the active device in vivo is biased towards the face of
the device that contains the membranes. The tumor is not only growing larger, but as it grows
79
away from, and around the device it can also grow out of the 'therapeutic reach' of the device.
Portions of the tumor, therefore, would be left unaddressed and free to proliferate.
The
conclusion from this set of experiments is that releasing drug sooner results in better median
survival. This conclusion is consistent with experiences in the clinic, where earlier detection and
treatment results in better survival for many cancers.
6.3.2.3 Comparison to polymer-based drug delivery
Both the 3 membrane day 5 group and the wafer group released the same amount of
TMZ (10 mg). The wafer group, however, had a median survival of 34 days, while the 3
membrane day 5 group had median survival of 23 days. Similar to the reasoning presented in
the previous section, the wafers likely achieved better median survival because they release
drug isotropically, while the microchips release drug from only 1 of 6 surfaces.
6.3.3 Immunohistological analyses
Immunohistological analysis was conducted on tissue samples from the efficacy study
with markers for cleaved caspase-3 (a marker for apoptosis) and Ki67 (a marker for
proliferation). TMZ causes DNA adducts, arresting the replication process and initiating
apoptosis [37].
Delivery of TMZ to a tumor mass should increase the number of cells
undergoing apoptosis and commensurately reduce the number of actively proliferating cells,
resulting in an increased number of cleaved caspase-3 positive cells and reduced number of
Ki67 positive cells. Animals receiving devices with 3 membranes activated on day 0 showed the
second highest number of caspase-3 positive cells, and lowest number of Ki67 positive cells.
This result confirms that TMZ is released in a viable, cytotoxic form, from the microchip device
and is consistent with the results from the efficacy studies where the three membrane, day 0
group demonstrated the most prolonged survival.
80
6.4 Conclusions
In vivo survival studies in a 9L gliosarcoma rodent study demonstrated that
temozolomide delivery from the active device is capable of prolonging animal survival versus no
treatment controls. The capabilities of this device introduce interesting opportunities for studying
and understanding the interplay between drug delivery rate, the timing of release and disease
progression. The ability to vary the release rate from the device with all other device properties
remaining the same is an important tool for studies aimed at determining optimal release rates
for certain drug:disease pairings. Delayed activation allows for studies to be conducted where
tumor and device are implanted simultaneously, but drug release is only initiated at a
predetermined time sometime after implantation. This ability can be a powerful tool in studies
where multiple therapeutics are to be delivered and the role of relative releasing timing is of
interest (e.g. localized delivery of temozolomide and a potentiating factor such as 06Benzylguanine).
Implications of this work include the ability to implant a microchip containing a variety of
drugs during intracranial tumor resection surgery. Based on the particular genetic abnormalities
that develop, tailored combinations of drugs would be able to be locally delivered without the
necessity of additional surgeries.
This is the first successful demonstration of intracranial microchip-based drug delivery.
The safety, kinetics, and efficacy of this method of drug-delivery have been demonstrated and
compared with one of the commonly implemented methods of intracranial drug delivery.
81
7. Future Work
This chapter is a discussion of meaningful future directions for the microchip project. The
first section is a brief proposal to co-deliver TMZ with a synergistic molecule 0 6-Benzylguanine
from a two compartment active device. The second section is a discussion of observed drug
release phenomena that motivate a mechanistic study into the mass transfer processes
involved in drug release from a depot device. Both proposed efforts represent important
directions towards the development of efficacious drug delivery devices for the improved
treatment of diseases.
7.1 Co-delivery of synergistic molecules for the treatment of GBM
The current generation of the active microchip device has been proven reliable and
efficacious in a rodent glioma model when delivering the chemotherapeutic temozolomide. The
ability to actively control how many membranes are ruptured, and when, proved to be an
interesting method of studying the interplay of release rate, release timing and disease
progression. A natural extension of this capacity is to now deliver multiple therapeutics from a
single device. A myriad of therapeutic combinations are possible, but combinations of particular
interest are those where the molecules being delivered are synergistic in their cytotoxicity.
Synergy is often achieved when one of the molecules inhibits a natural mechanism of resistance
to the primary agent, thus increasing the cytotoxicity of the primary agent. A promising
candidate molecule to be paired with TMZ is the molecule 0 6 -Benzylguanine.
The mechanism of action for TMZ is as follows: TMZ spontaneously degrades into its
active metabolite MTIC, MTIC methylates adenine and guanine residues in the cellular DNA,
this methylation causes a mismatch between paired DNA strands, native repair mechanisms fail
at repairing the mismatch due to the presence of a non-native residue (methylated adenine or
82
guanine), the result of the failed repair attempts is long lived 'nicks' in the DNA, which ultimately
initiate apoptosis [37]. A native adduct repair mechanism, and therefore mode of TMZ
resistance, is a protein called AGT [51]. AGT is capable of repairing guanine residues that have
been methylated at the 0-6 position. The act of repairing the DNA adduct irreversibly inactivates
AGT. The molecule 0 6 -Benzylguanine (06BG) is an effective 'fraudulent' substrate for AGT
[51]. Pretreatment of tumor cell lines with 06BG can increase the cytotoxicity of TMZ several
fold, presumably by reducing the number of DNA adducts that are repaired [37]. The exact
scheduling of 06BG pretreatment can affect the cytotoxicity of TMZ, indicating that AGT is
naturally regenerated and must be continuously depleted in order to realize any improved
efficacy of TMZ [37].
The potency of 06BG increasing TMZ cytotoxicity is schedule dependent. This pair of
drugs, therefore, is an excellent candidate to be co-delivered from the active microchip device.
To achieve the best results it appears necessary to deliver 06BG prior to and during TMZ
administration, in order to reduce AGT levels and maintain them at a minimum. 06BG and TMZ,
therefore, should be loaded into separate, independent chambers of the device (Figure 7.1).
Each chamber could be capped by a 3 membrane microchip, allowing for independent
activation of each reservoir. Independent control over the release rate and timing of each drug
would allow for in vivo tumor studies investigating the most efficacious kinetics of 06BG and
TMZ delivery. This pairing of synergistic molecules and device capabilities will demonstrate the
academic (e.g. using this device as a tool to learn the interplay between kinetics and disease
progression for new drug(s):disease pairings) and clinical merit of developing this technology.
The knowledge and technology gained from these experiments can be expanded to other drug
pairings and, hopefully, improved clinical possibilities.
83
Figure 7.1 CAD renderings of cross sectional views of the proposed two compartment
device. The reservoir architecture is essentially the same as the single compartment device only
mirrored around and internal plane. Each reservoir can be capped by the normal chip without any
alteration to the chip design or manufacturing.
7.1.2 Preliminary two compartment device work
Two compartment reservoirs were machined from acrylonitrile butadiene styrene (ABS)
plastic8 . The two compartment reservoir is a mirror image of the single compartment device
around the base. The overall reservoir size has to be maintained to remain implantable inside a
rat cranium. The resulting structure is 3.9 x 3.4 x 2.7 mm and has two reservoirs that can each
be capped by the 3 membrane chip and have a capacity of 5 mg of TMZ.
A preliminary release experiment was conducted with a two compartment device. The
purpose of this experiment was to demonstrate that co-formulated drug could be released
independently from each chamber. Each compartment of the device was loaded with 5
milligrams of TMZ and co-formulated with PEG. A standard 3 membrane chip was used to cap
each reservoir.
The device was placed in a stirred water bath and the first chip had all 3
8 ABS was selected because it is available, machinable, compatible with our UV epoxies and is well tolerated in
rodents.
84
10
86 -
2
2 -
0
0
25
50
75
100 125 150
175 200
Time (hrs)
Figure 7.2 Release of TMZ from a two compartment reservoir device. Each compartment was
loaded with 5 milligrams of TMZ and co-formulated with PEG. The first chip had 3 membranes
activated at t = 0 hours and the second chip was activated at t = 162 hours.
membranes activated at time equals zero. The release bath was replaced periodically and
assayed for TMZ content.
The activation of the second chip was delayed to sufficiently
demonstrate the completion of release from the first compartment, and the subsequent plateau
in release until the second chip was activated. The second chip had 3 membranes ruptured at
approximately 160 hours. The overall recovery of TMZ was excellent (- 90 %). There was,
however, a difference in release rates from the two compartments. It is not perfectly clear what
the cause of this disparity is. Two likely causes are either the presence of an air bubble in the
first reservoir, or differing amounts of PEG added to each reservoir. This was the first two
compartment device to go through the co-formulation and assembly process, which included
first generation vacuum fixtures for two compartment devices. Optimization of this process and
85
the handling of two compartment devices should quickly remove any substantial variability in the
loading, co-formulation and assembly process.
This preliminary experiment demonstrates that independent delivery from each
compartment can be achieved with this reservoir and chip design. The next step is to co-deliver
06BG and TMZ from the same device. Co-formulation of 06BG with PEG has been proven to
produce stable, high solubility formulations that produce reliable release kinetics [52]. The coformulation process that was developed for TMZ should, therefore, be a viable means for coformulating 06BG. HPLC methods exist for the quantification of 06BG and its metabolites [53].
In vitro release studies can therefore be conducted to quantify and validate the reliable release
of both TMZ and 06BG from the new two compartment device.
7.1.3 Impact of drug release kinetics on in vivo efficacy studies
The ultimate target for this project will be to demonstrate that varying the conditions
(release rate and relative timing) of 06BG delivery impacts the cytotoxicity of TMZ and,
therefore, the efficacy of TMZ delivery in in vivo tumor studies. These studies would be very
similar in execution to those presented in Chapter 6, except now there would be a second set of
kinetic parameters to investigate. The number of permutations possible with this device exceeds
those that can be feasibly investigated in vivo. In vitro cell culture and in vivo flank tumor model
studies could be conducted to cull down the number experimental groups that would ultimately
be examined in the intracranial tumor studies. The tumor cell line used in the efficacy studies
would have to be switched from 9L (low endogenous levels of AGT) to F98 (higher endogenous
levels of AGT) in order to demonstrate the impact of 06BG delivery [45]. The tissue levels of
total and inactivated AGT could be quantified by Western blot and radiolabelled stoichiometric
reaction assays [54].
86
There are several possible pitfalls. One of the most prominent is simply that the 3
membrane chip in its current form does not achieve a useful range of 06BG release rates. A
new chip would have to be designed, fabricated and validated after identifying the appropriate
range of release rates. Another possible major pitfall is the issue of anisotropy. Theoretically,
the best results would be achieved when the distribution profiles of 06BG and TMZ overlap.
The current design of the two compartment device, however, would have 06BG and TMZ
releasing from opposite sides of the device, which may reduce the efficacy of 06BG at
potentiating TMZ cytotoxicity. A two compartment device with both drugs releasing from the
same face can be made, but it would require the design and fabrication of a new chip and
reservoir.
The co-delivery of 06BG and TMZ is an exciting natural extension to the in vivo efficacy
studies conducted in this work. The active device is a uniquely capable tool to study the
interplay between 06BG and TMZ release kinetics and disease progression. The development
of the two compartment device is by no means trivial, and would involve the extensive
characterization of device performance in vitro and in vivo. The result, however, would be an
important advance in the use of active microchip devices for the localized therapy of GBM.
7.2 Investigation of mass transfer mechanisms in depot devices
The passive diffusion of drug from a depot device is not quite as simple as one might
initially intuit. Some baffling trends can emerge even from the simplest experiments. These
trends and behaviors can generally be designed around, but the intelligent design of passive9
devices would greatly benefit from an exhaustive study into the mechanisms at play. Three
particular phenomena have occurred across several projects within the Cima lab. This section is
Here I refer to passive devices as those that rely on primarily on dissolution and diffusion as the mechanism of
mass transport, as opposed to devices that rely on pressure gradient induced convective transport.
9
87
a brief description of those phenomena followed by a discussion of what key parameters and
mechanisms likely need to be investigated further to begin creating a more robust set of design
rules.
7.2.1 Varying Orifice Size and Number
Depot devices created in the Cima lab generally utilize micro or macro machined holes
to allow for mass transfer from the device. A natural extension of this design principle is to vary
the size and number of these through holes as a means for varying the release rate. The
principles of diffusion and Fick's laws would predict that mass transfer will vary linearly with area
given the same driving force and diffusivity. More often than not, however, the release rate does
not vary linearly with area. Frequently, when going from small total area to larger total area, the
release rate will be many times faster than predicted. This phenomenon can be designed
around, but makes a priori targeting of a particular release rate difficult. Valuable time and
materials could be spared if the root cause of this behavior could be incorporated into design
principles.
Plausible explanations for this phenomenon center around two key questions. What is
the role of water influx and what is the impact of device structure. Water influx is generally
assumed to occur readily, and rapidly enough to not be rate determining. The non-linear
dependence of release rate on orifice area may actually be a result of the fact that the rate of
water influx also increases with increasing orifice area. If the rate of water influx is, in part, rate
limiting, then varying the number of holes will non-linearly increase rate. The physical structure
of the device, and specifically the holes, may be the key factor. Is it possible that the actual
number of holes, as opposed to total area, is contributing to the change in release rate? Would
it be equivalent to have two holes of area A/2 or one hole of area A? Could changing from a
88
single orifice to multiple orifices change the physical processes at play, and if so, what are those
processes and how are they controlled?
7.2.2 Changing the Drug Payload
Whenever possible we load drug into our depot devices in dry, powder form. This
approach generally maximizes payload and drug stability within the reservoir prior to
implantation. The basic premise for release is that, upon implantation, extra cellular fluid will
flux into the device, wetting the reservoir and dissolving the drug. This approach results in a
saturated solution within the reservoir with some amount of the payload remaining in solid form
until late in the release 0 . The release rate should, therefore, be determined by the solubility limit
and diffusivity of the drug, and the area for flux. Under this premise, one would not expect the
release rate to depend on the device payload, as long as it was sufficient to maintain a
saturated solution within the reservoir over a prolonged period of time. The experimental results
of several projects in the Cima lab, however, indicate that release rate is indeed a function
payload.
Students have found that the release rate will increase with increasing payload,
despite the fact that all payloads used should be well above the solubility limit. These results
indicate that the simple mechanism presented above does not completely recapitulate the
mechanisms of mass transfer at play.
Several considerations need to be included to begin creating a more accurate proposed
release mechanism. First, what are the dissolution kinetics for the drug of interest? If the rate of
drug dissolution is sufficiently slow as to be release rate controlling, then the drug release rate
should actually scale with payload. Second, what is the impact of device structure? Is it actually
the fact that the void fraction of the device is changing with increasing payload? Would the
1 This is a result of large
payloads, small reservoirs and the generally low solubility of therapeutic molecules.
89
same rate be achieved if the reservoir volume was scaled with the payload? Finally, does the
drug powder contain a co-factor or excipient that can alter the physical environment within the
reservoir? If the drug is actually co-formulated with, for example, an acid, then increasing the
payload will decrease the pH within the reservoir which may alter the solubility or stability of the
drug.
7.2.3 Including Excipient
Including an excipient introduces a degree of uncertainty in the interactions between the
drug, excipient, dissolving media and ultimately the release rate. For example, an excipient
aimed at improving drug stability may affect the drug's solubility thus altering the release rate.
An example from this work is including PEG as a means of displacing trapped air. The inclusion
of PEG increased the rate of drug release significantly. It was posited in Chapter 5 that including
PEG, a highly water soluble molecule, introduced water influx and PEG swelling as critical steps
in the release mechanism. Creating design rules by studying the precise interaction of every
drug:excipient pairing is not feasible, but this phenomena introduces intriguing questions about
the role of water influx and drug swelling in drug release kinetics.
Excipients are generally included to improve drug stability, solubility and packaging (e.g
binders in tablets, or the usage of PEG to displace air). The results discussed above indicate
that excipients can result in secondary effects that can have a drastic impact on the core
function of the device. There are a huge number of potential excipients. Cataloging each of their
secondary, or even tertiary interactions is not feasible. It would probably be best to categorize
the key set of physical processes involved in drug release (e.g. water influx) and understand
how a potential excipient might vary the relative role of each process (e.g. high solubility PEG
imbibing water rapidly).
90
7.2.4 Plausible mechanisms and directions of further research
The plausible explanations for these phenomena are varied and frequently only lead to
more questions. What emerges from this set of questions and hypotheses, however, is that
there are two key components that require further investigation: reservoir structure and material
properties. The crucial aim is to understand how these components influence the mechanisms
of mass transfer. Specific aspects of interest in device structure are: the number and size of
holes, the aspect ratio or shape of the reservoir, and the ratio of reservoir volume to drug
payload. The material properties of all components should be considered. It would be
particularly beneficial to study impact of solubility, wettability and diffusion characteristics. The
focus, again, should be on how these factors affect which mass transfer mechanisms are
involved in drug release. The most prevalent mass transfer processes appear to be: water
influx, drug dissolution, drug diffusion, and drug/excipient swelling. Developing experimental
protocols to differentiate these effects will lie and the crux of this project. The ultimate objective
will be to distill these complex and varied effects into broadly applicable design rules. These
rules will, by no means, be comprehensive or foolproof, but will provide valuable guidance in the
selection of materials and design of the device reservoir.
91
8.
A new device: The 'Injectrode'
Starting in the Spring of 2009 a collaboration developed between the laboratories of Ann
Graybiel, Ph.D. (Institute Professor, Department of Brain and Cognitive Science) and Michael
Cima, Ph.D. (my co-advisor, Professor, Department of Material Science & Engineering). The
collaboration centered on the development of micro-cannula devices that would be capable of
furthering both, our understanding of fundamental neuroscience and technology for clinical
intervention in neurological disorders and dysfunction.
Initially Ann reached out to Michael,
proposing that they co-write a small proposal for the McGovern Institute Neurotechnology
Program (MINT). This program supplies small seed grants to McGovern Institute Researchers
and collaborators outside of the neuroscience field towards developing 'technical innovations
that will help to transform the future of neuroscience and medicine.' The award was granted and
Michael asked that I head up the collaboration.
We all quickly learned that there was
tremendous promise in developing this technology, and also that significantly more funding
would be required to give it the attention it deserved.
The device, or 'injectrode', is 12 cm long, approximately 200 pm in diameter and has
three bores that run axially along the length of the device (Figure 8.1). Two of these bores are
conduits for fluid flow and the third bore is meant to house an electrode. This configuration
allows for multiple therapeutics to be infused independently with precise temporal control, and
for simultaneous electrical stimulation. This device would be a chronic intracranial implant and
would facilitate the administration of both chemical and electrical therapy to specific anatomical
nuclei. The following chapter is a synopsis of the efforts that have been made so far on this new
project as well as a discussion of future efforts. First, the concept of circuit based diseases as it
relates to neurologic disorders is introduced. Second, excerpts from two proposals written
between the Cima and Graybiel groups are presented. These excerpts are included to provide
92
clinical context and motivation to the development of these micro-cannula based devices. Third,
the prior art and the first generation injectrode device are introduced, detailing the components
and aspects of manufacture. Fourth, preliminary in vitro infusion data are presented. Finally,
future aims for improving the device design and fabrication are discussed".
8.1 Circuit Diseases
There is increasing evidence that the pathology underlying many neurologic disorders is
a failure in the dynamic communication between multiple brain areas as opposed to a deficit in a
single brain area. This has led to the notion of circuit-based disorders [55-60]. The conceptual
change here is more than just semantics. Instead of focusing treatment on the suspected
pathological neural tissue, the new therapeutic target is normalizing activity across a circuit.
Normalization can be achieved by intervention at any level within the circuit and not just in the
brain area where the original insult occurred. The treatment of circuit-based disorders, therefore,
is significantly improved by tools that allow for modulating the precise tuning of circuit activity
(i.e. reducing activity of an overactive node in the.circuit or conversely increasing activity in a
node that is pathologically silent). The brain uses both chemical as well as electrical methods of
communication. Combining chemically-based and electrically-based therapeutic approaches,
therefore, has great potential advantages for facilitating the fine-tuning of activity across such
circuits. A single device that is capable of delivering drugs while performing simultaneous
electrical stimulation is exactly what is needed to precisely tune activity. The greater precision in
regulating neural circuit activity introduced by this combined control will reduce the number and
frequency of debilitating side effects common when systemic drugs are administered as well as
prolong the efficacy of deep brain stimulation alone. For example, a drug can be administered to
" Dr. Patrick Tierney, Staff Scientist in the Graybiel Laboratory, contributed significantly to the composition of this
chapter.
93
slightly increase the excitability of a brain structure. The result is that lower electrical stimulation
currents can be used to modulate activity of that area. Lower currents reduce the occurrence of
side effects by limiting the spread of electrical activation to neighboring brain nuclei that may not
be part of the targeted circuit as well as reducing the potential damage to neural tissue due to
heat generation at the electrode tip. Using lower currents will also prolong the battery life of
stimulators, reducing the number of invasive maintenance procedures. The success of multipronged approaches to disease treatment is repeatedly demonstrated across human medicine
(e.g. the success of drug "cocktails" for the treatment of HIV/AIDS, deep brain stimulation in
Parkinson's disease with systemic drug administration, combined radiation and chemotherapy
for cancers, treatment of depression with conventional talk therapy in combination with drug
therapy).
8.2 Clinical Rationale: Traumatic Brain Injury, Anxiety and Mood Disorders
8.2.1 Intractable Anxiety and Mood Disorders
Anxiety and mood disorders are among the most common mental disorders that afflict
Americans, and they can be severely debilitating [61]. Most individuals with these disorders are
initially treated with oral administration of pharmaceuticals and behavioral or counseling therapy.
However, for an important subset of these individuals, these treatment options have very limited
therapeutic efficacy. Furthermore, long-term treatment with oral medications frequently leads to
drug tolerance and dose escalation. Due to the large doses usually required, the side effects of
these drug treatments often exceed the therapeutic effects, thus eliminating pharmacological
treatments as a feasible strategy for patients. Large numbers of otherwise promising drugs have
been set aside for this reason. Alternative treatment strategies are currently being tested for
patients who do not respond to pharmacological interventions. Early trials with electrical deep
brain stimulation (DBS) have successfully been used to treat mood disorders. It stands to
94
reason that combining electrical stimulation with local drug delivery in the specific neural circuits
affected in the disorders could achieve greater and more flexible therapeutic strategies.
Suitable technology for precise drug delivery combined with electrical stimulation, however,
does not exist.
Mood and anxiety disorders share the common feature of repetitive and perseverative
thoughts and moods that do not adapt to changing situations [62, 63]. For example, individuals
with mood disorders tend to evaluate their situation either too negatively or positively. In the
same vein, individuals with obsessive-compulsive disorder (OCD) have difficulty inhibiting
behaviors that are based on biased evaluations of the situation. It seems that emotions hold a
disproportionate influence on the decision-making process [64]. Interestingly, experiments in
humans as well as awake, behaving non-human primates report that neuro-electrical as well as
neuro-chemical activity in the cortico-basal ganglia loop connecting the ACC and striatum is
abnormal in bipolar disorder [65, 66], mood disorder [67-69], OCD [70-74], and in addiction [75,
76]. These findings suggest that a dysfunction in the ACC-basal ganglia loop could be the cause
of the imbalance in the influence of emotion on behaviors.
Despite this new clue to the
pathophysiology of anxiety and mood disorders, there are few technologies available to develop
and implement effective treatment regimens
8.2.2 Traumatic Brain Injury and Associated Chronic Neurological Disorders
Approximately 1.5 million people suffer a traumatic brain injury (TBI) in the United States
each year [77]. The probability of developing chronic neurologic disorders correlates with the
severity of the initial neurologic injury. TBI patients are at high-risk of developing chronic
conditions such as posttraumatic epilepsy (PTE) and chronic traumatic encephalopathy (CTE).
Population-based studies of TBI patients have reported the 30-year cumulative incidence of
PTE of 17% [78]. Other studies of TBI patients report incidence of PTE between 2-50% [79].
95
Roughly thirteen percent of PTE patients will be refractory to multiple antiepileptic medications
and may then be referred for neurosurgical evaluation. Persons subjected to repeated
concussive injury or even mild TBI are at risk for development of CTE, characterized clinically
by parkinsonism, behavioral and cognitive impairment, and speech and gait abnormalities [80].
Much of the data on CTE have come from athletes who have experienced repetitive TBI, and
the findings are considered to apply to any individuals prone to head trauma, such as military
veterans. Studies report the incidence of CTE to be just under twenty percent. It appears that
CTE is a pathologically distinct entity that mimics and affects many of the same areas of the
brain implicated in Parkinson's disease and Alzheimer's disease. The same medications given
for Parkinson's disease are administered to CTE patients, who experience symptomatic relief,
but also suffer similar long-term side effects of chronic administration, including on/off
oscillations, dose failure, and disabling dyskinesias [81]. Both PTE and CTE represent circuitbased disorders in which some part (or multiple parts) of interacting neural circuits have been
perturbed.
8.3 Device Design and Manufacture
8.3.1 The Prior Art for Micro-cannula Devices
Current micro-cannula devices can be divided into two subsets.
The first subset
includes devices that have a single lumen. These devices are appealing due to their high
aspect ratio (length:diameter, typically 10 cm:250 pm) and structural integrity, which is generally
achieved by using stainless steel hypodermic tubing [82-85]. However, these devices have
limited functionality since they only administer one fluid at a time and do not contain onboard
electrodes or tetrodes. Switching solutions with single lumen devices would require expelling the
entire dead-volume of the device, which could lead to overdosing or to drug delivery beyond the
specifically targeted region of the brain. The second subset of devices includes those that have
96
more diverse functionality engineered into them, such as multiple lumens and on-board
electrodes [86-88].
The merit of these devices is the range of functions (multiple solution
infusion and electrophysiology) that are possible with a single device. The brittle materials and
methods of manufacture of these devices, however, limit the aspect ratios that can be achieved
(5 mm:200 pm). The current state-of-the-art devices of this type are limited to use in rodent
models and are not directly scalable to larger animal models or to clinical applications [89].
The device that we seek to develop in the proposed work will be more significantly
robust than these available alternatives, and the device will possess more diverse capabilities
than existing devices while still minimizing size, and therefore invasiveness. The device will be
12 cm long, approximately 200 pm in diameter and will be manufactured out of flexible
borosilicate glass tubing and coated in reinforcing metal layers and surface functionalized for
improved biocompatibility. The glass has three bores that run axially along the length of the
device allowing for numerous configurations of multiple drug solutions, electrodes, tetrodes, and
neurochemical sensing probes. The precise combinations of functions can be varied without
varying the overall geometry, materials, or method manufacture of the device. This versatility
will allow for a more efficient and cost effective production of the device as well as more efficient
treatment process. Each device can be tailored to the needs of each individual patient using the
same starting materials.
Neural probes are limited in chronic settings due to the natural formation of a glial scar
around the implant. The glial scar manifests as a dense sheath surrounding the implant, and is
a result of astrocytes and microglia reacting with the foreign material. Scar formation is more
robust immediately after the implantation due to tissue trauma and disrupted vasculature
immediately surrounding the implant, but persists beyond the initial injury due to the constitutive
presence of astrocytes and microglia within the brain. Neural probes fail because the sheath
eventually excludes neurons from the immediate vicinity of the implant, thus preventing
97
electrical interaction between the implant and neurons. Farra et al, determined that fibrous
encapsulation does not, however, prevent or diminish drug delivery from chronically implanted
devices [41]. Their work demonstrated that an implanted microchip based device can deliver
hPTH(1-34) after fibrous encapsulation with pharmacokinetic parameters that are similar to
subcutaneous injections. This result indicates that the chemical functionality of the injectrode
may persist throughout glial scar formation, whereas all functionality of conventional neural
probes would eventually be lost. Biochemical coatings incorporated into the injectrode will
diminish the incidence and extent of glial scar formation, thus prolonging even the electrical
functionality of the device and generally improving device:host tissue interactions.
8.3.2 The First Generation Injectrode
A first generation device was designed between the Graybiel and Cima laboratories.
Prototype devices have been manufactured in the Cima laboratory and are undergoing
preliminary in vitro testing in the Cima laboratory. The device is constructed of 5 basic
components: stainless steel guide tube, commercially available boroscilicate glass tubing, 3
compartment manifold, electrode and upstream pump and reservoirs. The boroscilicate glass
tubing forms the heart of the device. It is 12 centimeters long and 150 micrometers (pm) in
diameter with 3 lumens running its length (Figure 8.1C,D). Two lumens are 38 pms in diameter
and are intended for fluid flow. The third lumen is 90 pms in diameter and is meant to house the
electrode. The glass tube nests within the steel tube (200 pm OD). The steel tube offers
increased rigidity to the length of device allowing for more accurate placement of the distal tip of
the device within tissue. A 75 pm diameter tungsten electrode will used in these studies to listen
to the electrophysiological response of the tissue to chemical stimulus. The manifold (Figure
8.1A,B) establishes fluidic connections between the 38 pm lumens and the upstream pumps
and reservoirs.
98
Infusion
chamber
B
C
__
Figure 8.1 Digital renders and SEM images of the core components of the injectrode. A
CAD rendering of the manifold. Four ports are visible. The port on the circular face is a through hole
through with the introducing needle and glass tube are threaded. The center port on the cylinder
surface is for filling the epoxy chamber (panel B). Pump tubing is inserted into the remain two ports
which service the infusion chambers (Panel B)( A). A cross sectional view of the manifold depicting the
epoxy chamber, two infusion chambers, and the silicone septa (blue) (B). CAD rendering of the multilumen glass tube. The large lumen (90 pm) houses the electrode. The two small (38pm) lumens
conduct fluid from manifold down to the distal tip of the device. Each 38 pm lumen has an access port
cut into it to allow for interfacing with the manifold (C). Scanning electron microscopy image of several
glass tubes viewed end on.
The primary difficulty in constructing these devices is establishing clean, functioning
fluidic connections between the 38 pm lumens and the surgical grade tubing that is compatible
with the syringe pump. Two basic criteria must be achieved. First, access ports must be cut into
the side of the glass tube in precise locations such that each access port only intersects one 38
pm lumen (Figure 8.1C). Access ports have been successfully cut into glass tubes using a
precision micro-milling machine a diamond coated engraving tool. Second, the manifold must
contain chambers that align to the cut access ports and are sealed to prevent contamination
99
between the reservoirs and the environment. A prototype manifold has been designed and
fabricated.
The manifold is constructed of five components and when assembled contains 3
chambers (Figure 8.1A,B). Two chambers align to the access ports of the glass tube and are
meant for fluid flow. These chambers have been designed such that the pump tubing can simply
be inserted into their sidewall and fixed in place by biomedical grade epoxy. Each chamber is
separated from the environment and neighboring chambers by silicone septa. A needle is
threaded through the manifold, traversing all three chambers and piercing each septum. The cut
glass tubing is then threaded through the needle and, once threaded, the needle is retracted. As
the needle is retracted the elastomeric silicone collapses down onto the glass tubing. The septa
grip the glass tube such that the tube:manifold assembly can be manipulated without loss of
alignment, and form a preliminary seal around the glass tube between the chambers.
Permanent seals are achieved by curing biomedical grade epoxy on either end of the manifold
and within the middle chamber of the manifold.
8.4 In Vitro Infusion Results
Prototype devices have undergone in vitro infusion studies to determine the reliability in
function of these devices. The primary targets for reliable function are: that a range of volumes
can be infused accurately and reproducibly, cross contamination does not occur within the
manifold, and that the device continues to function over long periods of time and does not
become clogged. The protocol for infusion studies is as follows: load two syringe pumps with
dye solutions (one with red and one with blue), connect the pump tubing to the manifold of the
100
Table 8.1: Infusion results from the first two lumen device
Blue
Target
Volume
(ni)
Red
Standard Target
Deviation Volume
as %
(nl)
Average Standard
Measured Deviation
Volume
(nl)
n=3
(nl)
300
262
11.6
200
183
4.64
100
89.1
4.66
50
47.9
1.4
*The quoted accuracy of the syringe
Average Standard Standard
Measured Deviation Deviation
Volume
(nl)
as %
n=3
(nl)
4.43
300
305.6
20.2
6.6
2.53
200
212.5
6.1
2.9
5.23
100
102.5
3.0
2.9
2.96
50
59.1
3.0
5.1
pumps themselves is 3-5% of infusion volume.
device, flush dye through each lumen for 5-10 minutes to flush out any air bubbles and check
for leaks, after flushing place the tip of the device in a pre-filled cuvette, infuse the desired
volumes replacing the cuvette and each infusion. The amount of dye in each sample was
determined by measuring the peak absorbance at a given wavelength for each dye. Preliminary
studies using our first generation prototype have yielded delivered volumes with moderate
accuracy and reproducibility within about 5% of the desired volumes (Table 8.1).
The data in Table 8.1 are from a single device. Each infusion volume was repeated in
triplicate. The measured infusion volumes for each dye are consistent, with standard deviations
of approximately 5%.
The micro syringe pumps used to control the infusion have quoted
accuracies of 3-5% of infusion volume. The device itself, therefore, does not introduce a large
degree of variability. The measured volumes for red dye are relatively accurate to the desired
volume, whereas the measured volumes for blue dye are consistently low. A possible
explanation for this result is insufficient flushing of blue dye through the device, resulting in the
infusion of a slightly diluted dye solution.
There were no signs of leaking from the tubing
junctions or manifold itself. Repetitions of this experiment with other devices will help identify
key steps in the infusion protocol that will affect variability and accuracy.
101
A key point of concern in the design of this device is whether any cross contamination
occurs within the reservoir. There are two primary scenarios in which cross contamination
would occur. First, there is a crack or gap in the epoxy chamber, thus allowing the dyes to flow
or diffuse between chambers. Second, if the epoxy seal in the epoxy chamber is not rigid
enough, the increased pressure in one chamber, due to infusion of that dye, may induce an
increased pressure in the second chamber, thus causing a small degree of infusion of that dye.
Each sample in Table 8.1 contained levels of both dyes (e.g. a small amount of red when blue is
infused). This could be a result cross contamination or, simply diffusion of dye from the tip of the
device. Diffusion control samples were taken where neither pump was turned on. Analysis of
these samples determined that diffusion alone can result in quantifiable signal of both dyes.
Comparison of the signal from the diffusion samples and the infusion samples confirmed that no
cross contamination occurred and that passive diffusion caused a small amount of the noninfused dye to be detected in each sample. The manifold design, therefore, is capable of
isolating the two infusion chambers, and directing each dye to the correct lumen.
The data in Table 8.1 only reflect the performance of one device. Many other devices
failed immediately either due to clogging of one of the lumens or fracture of the glass tubing
upon assembly. Clogging may be due to the presence of large particulate in the dye solutions.
Another possible culprit is the potting epoxy itself. Failed devices have been disassembled and
epoxy was found within the infusion chambers. The assembly and epoxy potting protocol, and
choice of epoxy, will have to be refined to reduce the incidence of epoxy entering the infusion
chambers. The assembly protocol will similarly have to be improved to reduce the incidence of
glass tube fracture. The design of manufacture of assembly tooling will drastically reduce the
variability in assembly and, therefore, incidence of critical flaws.
102
8.5 Future steps in the development of this device
8.5.1 In vitro tissue phantom infusion studies
The end use of this device will be to infuse fluid into tissue. The physical environment of
infusing into tissue is a significant departure from infusing into water baths. Similar, albeit much
larger, devices used for the delivery of chemotherapeutics have encountered three primary
obstacles when infusing into tissue: catheter clogging with tissue as the device is advanced
through tissue, back flow (flow of solution back along the catheter: tissue interface), and
increased interstitial pressure. This device should be less susceptible to developing tissue plugs
because the cross sectional area of the lumens is much smaller (-40 pm vs. >>100 pm) and the
tip of the electrode will encounter the primary resistance to insertion. Backflow should be
diminished with this devices due to the reduced size (-200 pm vs >1mm) and reduced infusion
volume (100-200 nI vs >100 pl). The smaller the device, the less it interrupts the native structure
and integrity of the surrounding tissue leading to less native tissue damage, a more intimate
tissue:device interface and less backflow. Reducing the infusion volume also diminishes the
increase in interstitial pressure reducing the driving force for backflow and the deleterious
effects of high interstitial pressure.
Agarose tissue phantoms will be used as an in vitro proxy for tissue [87]. The device will
be inserted into the tissue and advanced under the same parameters as in vivo. Solutions of
dye will be infused, again under in vivo parameters. The trajectory of the dye can observed in
real time and a computer aided image analysis can be utilized to characterize the spatial
distribution of infused dye. In line pressure gauges will be able to detect if the tip of the device is
clogged by agarose as their will be an initial spike in upstream pressure before the plug is
ejected. The results of these studies will be used to critique the lumen, guide tube and infusion
parameter design.
103
8.5.2 Device Coating for Structural Integrity and Biocompatibility
Neural probes often fail after six months of implantation despite operating correctly
initially [90, 91]. The loss in operating capabilities is largely due to a process known as gliosis.
Resident inflammatory cells in the brain, astrocytes and microglia, react to foreign bodies and
form a fibrous capsule to isolate the implant. This fibrous capsule can reduce signal to noise
ratio for electrodes and reduce electrical stimulation efficacy. Trauma from probe implantation
activates astrocytes and microglia and stimulates gliosis [92, 93].
compounds
the
inflammatory
process
by
rupturing
Probe insertion further
microvasculature,
thus
allowing
extravasation of circulating inflammatory cells to the site of injury/implantation. Gliosis persists
beyond the initial implantation injury as astrocytes and microglia continue to react with the
foreign material. Microfractures in the brain resulting from compliance mismatch and movement
between the device and brain are also believed to increase the amount of reactive cells present
and stimulate gliosis [94]. It stands to reason that minimizing implant size concurrently reduces
tissue injury [95], while proper surface chemistry may reduce activation of resident inflammatory
cells. Furthermore, local delivery of anti-inflammatory (e.g., dexamethasone) drugs through one
of the lumens may reduce gliosis in vivo [96].
The current generation of the device utilizes a custom made, 33 gauge (200 pm OD),
stainless steel guide tube to improve the structural integrity of the device during insertion. This
approach is limited by several critical factors. First, stainless steel is not readily amenable to
direct surface functionalization. Second, threading the glass tubing through twelve centimeters
of steel tubing is a low yield and time consuming process. The overall diameter of the first
generation injectrode is limited to 33 gauge as this is the smallest commercially available
stainless steel hypodermic tubing.
Smaller, and therefore less invasive devices, are not
possible due to this manufacturing limitation.
104
We plan to employ a coating procedure in place of the stainless steel guide tube for our
device that will improve device rigidity while minimizing the chronic immune response to the
implant.
The coating will enable the injectrode to operate effectively for long periods of
implantation. The proposed coating consists of two layers: a metal structural layer and a
functionalized passivating layer suitable for additional conjugation of bioactive molecules. The
purpose of the structural layer is to provide sufficient strength throughout the implantation
procedure and to function as a substrate for the subsequent functionalization chemistry. The
passivating layer will consist of non-fouling, hydrophilic polymers with functional handles to
conjugate bioactive molecules, which may reduce gliosis. Gold will be used for the structural
layer.
Gold has been used in many medical applications and its biocompatibility is widely
accepted. It can be deposited on borosilicate glass through standard physical vapor deposition
(PVD) techniques. Conjugation chemistry on gold surfaces is well studied and will allow for
facile deposition of non-fouling self-assembled monolayers as well as reactive handles for
conjugation of bioactive molecules.
Promising coating and conjugation combinations will be first characterized in vitro.
Buckling tests will quantify the impact of varying the composition and thickness of the structural
layers. Cell culture experiments and subsequent immunoctyochemistry with primary neural cells
will quantify the extent of reactive gliosis for different conjugation chemistries. Those coating
and conjugation combinations that perform well in in vitro testing will go on to in vivo testing in
rodents.
Acute and chronic intracranial implantation and biocompatibility studies will be
conducted to validate and quantify the performance of each coating in actual tissue. The most
successful coating strategy will be adopted into the devices used for the non-human primate
behavior studies.
105
8.5.3 In vivo experimentation to confirm neural circuit and behavior modification
The Graybiel lab are the preeminent experts on behavior and mood disorders, and the
use of non-human primate (NHP) models for the investigation of root causes and methods of
clinical intervention. They have animal models in place that recapitulate the executive functions
at the core of mood disorders and neurological diseases. The ultimate performance of the
injectrode will be validated and quantified in Rhesus macaque monkeys. This model not only
captures meaningful aspects of human executive function and behavior, it also allows the
device to be tested on a physical scale that is comparable to the human brain. This is a
valuable step towards clinical adaptation of this technology.
The first in vivo NHP studies will be aimed at modifying activity across a neural circuit.
Briefly, the injectrode will be implanted into the primate brain and neuromodulatory drugs will be
infused with concurrent electrical stimulation. The injectrode tip will be implanted into the cortex
of the animal and separate electrodes will record the response in downstream structures.
Depending on the nature of the infused drug (stimulant or suppressant) activity across a neural
circuit will be either increased or attenuated.
Subsequent studies will aimed at actually modifying the behavior of animals.
The
monkey will be trained in two tasks that rely on the executive functions at the core of anxiety
and mood disorders, namely risk assessment and motivation. The two tasks center on
motivation and risk:reward decision. The success rate, number of task attempts, and tendency
towards risk or reward will be the primary metrics. The injectrode will used to modulate neural
activity, both electrically and chemically, while the animal performs these tasks. The kinetics and
magnitude of the behavior change will be used to evaluate the device performance.
106
8.6 Conclusion
The injectrode is an exciting new addition to the Cima lab's repertoire of devices. The
development of this device will involve some intriguing technical challenges, and is a crossroads
of neuroscience, drug delivery, micro-fabrication and chemistry. Preliminary designs and results
are promising, and indicate that the device is indeed feasible. Biochemical coatings that improve
biocompatibility will allow the device to function predictably and reliably for prolonged periods of
time. The capability to simultaneously deliver therapy via chemical and electrical means, on a
human length scale, presents an impactful advance in medical intervention for neurological
disorders.
107
9. References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
Deorah, S., et al., Trends in brain cancer incidence and survival in the United States: Surveillance,
Epidemiology, and End Results Program, 1973 to 2001. Neurosurgery Focus, 2006. 20(4):E1: p. 17.
Giglio, P., Chemotherapy for glioblastoma: Past, present and future, in Glioblastoma. p. 203-216.
Palu, G., et al., Gene therapy of glioblastoma multiforme via combined expression of suicide and
cytokine genes: a pilot study in humans. Gene Ther, 1999. 6(3): p. 330-7.
Holland, E.C., Glioblastoma multiforme: The Terminator. PNAS, 2000. 97(12): p. 6242-6244.
Hou, L.C., et al., Recurrent glioblastoma multiforme: a review of natural history and
management options. Neurosurgery Focus, 2006. 20(4):E3: p. 1-9.
Kawakami, K., et al., Distribution Kinetics of targeted cytotoxin in glioma by bolus or convectionenhanced delivery in a murine model. Journal of Neurosurgery, 2004. 101: p. 1004-1011.
Raghavan, R., et al., Convetion-enhanced delivery of therapeutics for brain disease, and its
optimization. Neurosurgery Focus, 2006. 20(3): E12: p. 1-13.
Heimberger, A.B., et al., Temozolomide Delivered by Intracerebral Microinfusion is Safe and
Efficacious Against Malignant Gliomas in Rats. Clinical Cancer Research, 2000. 6: p. 4148-4153.
Sampson, J.H., et al., Comparison of intratumoral bolus injection and convection-enhanced
delivery of radiolabeled antitenascin monoclonal antibodies. Neurosurgery Focus, 2006.
20(4):E14: p. 1-4.
Sawyer, A.J., J.M. Piepmeier, and W.M. Saltzman, New Methods for Direct Delivery of
Chemotherapy for Treating Brain Tumors. Yale Journal of Biology and Medicine, 2006. 79: p.
141-152.
Groothuis, D.R., et al., Comparison of cytosine arabinoside delivery to rat brain by intravenous,
intrathecal, intraventricular and intraparenchymal routes of administration. Brain Research
Interactive, 2000. 856: p. 281-290.
Lidar, Z., et al., Convection-enhanced delivery of paclitaxel for the treatment of recurrent
malignant glioma: a Phase I/i clinical study. Journal of Neurosurgery, 2004. 100: p. 472-479.
Li, Y., et al., In vivo release from a drug delivery MEMS device. Journla of Controlled Release,
2004. 100: p. 211-219.
Kim, G.Y., et al., Resorbable Polymer microchips releasing BCNU inhibit tumor growth in the rat
9L flank model. Journla of Controlled Release, 2007. 123: p. 172-178.
Bobo, R.H., et al., Convection-enhanced delivery of macromolecules in the brain PNas, 1994. 91:
p. 2076-2080.
Lieberman, D., et al., Convection-enhanced distribution of large molecules in grey matter during
interstitial drug infusion. Journal of Neurosurgery, 1995. 82: p. 1021-1029.
Laske, D., R. Youle, and R. Oldfield, Tumor Progression with regional distribution of the targeted
toxin TF-CRM 107 in patients with malignant brain tumors. Nature Medicine, 1997. 3: p. 13621368.
Neeves, K.B., et al., Dilation and degradation of the brain extracellular matrix enhances
penetration of infused polymer nanoparticles. Brain Res, 2007. 1180: p. 121-32.
Batchelor, T.T., et al., AZD2171, a pan-VEGF receptor tyrosine kinase inhibitor, normalizes tumor
vasculature and alleviates edema in glioblastoma patients. Cancer Cel, 2007. 11(1): p. 83-95.
Giese, A., et al., Cost of Migration: Invasion of Malignant Gliomas and Implications for
Treatment. Journal of Clinical Oncology, 2003. 21(8): p. 1624-1636.
108
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
Lefranc, F., et al., Combinded cimetidine and temozolomide, compared with temozolomide
alone: significant increases in survival in nude mice bearing U373 human glioblastoma
multiforme orthotopicxenografts. Journal of Neurosurgery, 2005. 102: p. 706-714.
Lesniak, M.S., et al., Local delivery of doxorubicin for the treatment of malignant brain tumors in
rats. Anticancer Res, 2005. 25(6B): p. 3825-31.
Hsu, W., et al., Local delivery of interleukin-2 and adriamycin is synergistic in the treatment of
experimental malignant glioma. J Neurooncol, 2005. 74(2): p. 135-40.
Valtonen, S., et al., Interstitial chemotherapy with carmustine-loaded polymers for high-grade
gliomas: a randomized double-blind study. Neurosurgery, 1997. 41(1): p. 44-8; discussion 48-9.
Westphal, M., et al., A phase 3 trial of local chemotherapy with biodegradable carmustine
(BCNU) wafers (Gliadel wafers) in patients with primary malignant glioma. Neuro Oncol, 2003.
5(2): p. 79-88.
Guerin, C., et al., Recent advances in brain tumor therapy: local intracerebral drug delivery by
polymers. Invest New Drugs, 2004. 22(1): p. 27-37.
Tyler, B., et al., A thermal gel depot for local delivery of paclitaxel to treat experimental brain
tumors in rats. J Neurosurg, 2010. 113(2): p. 210-7.
Brem, S., et al., Local Delivery of Temozolomide by biodegradable polymers is superior to oral
administration in a rodent glioma model. Cancer Chemotherapy Pharmacology, 2007.
Kim, G.Y., et al., Resorbable polymer microchips releasing BCNU inhibit tumor growth in the rat
9L flank model. J Control Release, 2007. 123(2): p. 172-8.
Scott, A.W., et al., Intracranial microcapsule drug delivery device for the treatment of an
experimental gliosarcoma model. Biomaterials, 2011. 32(10): p. 2532-9.
Grayson, A.C.R., et al., Electronic MEMSfor triggered delivery. Advanced Drug Delivery Reviews,
2004. 56(2): p. 173-184.
Elman, N.M. and U.M. Upadhyay, Medical applications of implantable drug delivery microdevices
based on MEMS (Micro-Electro-Mechanical-Systems). Curr Pharm Biotechnol, 2010. 11(4): p.
398-403.
Santini, J.T., Jr., et al., Microchip technology in drug delivery. Ann Med, 2000. 32(6): p. 377-9.
Santini, J.T., Jr., M.J. Cima, and R. Langer, A controlled-release microchip. Nature, 1999.
397(6717): p. 335-8.
Santini, J.T., et al., Microchips as Controlled Drug-Delivery Devices. Angew. Chem. Int. Ed, 2000.
39: p. 2396-2407.
Elman, N.M., et al., The next generation of drug-delivery microdevices. Clin Pharmacol Ther,
2009. 85(5): p. 544-7.
Newlands, E.S., et al., Temozolomide: a review of its discovery, chemical properties, pre-clinical
development and clinical trials. Cancer Treatment Reviews, 1997. 23: p. 35-61.
Moses, M.A., H. Brem, and R. Langer, Advancing the field of drug delivery: taking aim at cancer.
Cancer Cell, 2003. 4(5): p. 337-41.
Richards Grayson, A.C., et al., Multi-pulse drug delivery from a resorbable polymeric microchip
device. Nat Mater, 2003. 2(11): p. 767-72.
Santini, J.T., M.J. Cima, and R. Langer, A controlled-release microchip. Nature Biotechnology,
1999. 397: p. 335-338.
Farra, R., et al., First-in-Human Testing of a Wirelessly Controlled Drug Delivery Microchip. Sci
Transl Med, 2012. 4(122): p. 122ra21.
Maloney, J.M., et al., Electrothermally activated microchips for implantable drug delivery and
biosensing. Journal of Controlled Release, 2005. 109: p. 244-255.
109
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
Elman, N.M., et al., Electro-thermally induced structuralfailure actuator (ETISFA)for implantable
controlled drug delivery devices based on micro-electro-mechanical-systems. Lab Chip, 2010.
10(20): p. 2796-804.
Elman, N.M., H.L. Ho Duc, and M.J. Cima, An implantable MEMS drug delivery device for rapid
delivery in ambulatory emergency care. Biomed Microdevices, 2009. 11(3): p. 625-31.
Rhines, L.D., et al., 06-Benzylguanine Potentiates the Antitumor Effect of Locally Delivered
Carmutisine against an Intracranial Rat Glioma. Cancer Research, 2000. 60: p. 6307-6310.
Storm, P.B., et al., Systemic BCNU enhances the efficacy of local delivery of a topoisomerase I
inhibitor against malignant glioma. Cancer Chemotherapy Pharmacology, 2004. 54: p. 361-367.
Li, Y., et al., In vivo delivery of BCNU from a MEMS device to a tumor model. Journia of
Controlled Release, 2005. 106: p. 138-145.
Prescott, J.H., et al., Chronic, programmed polypeptide delivery from an implanted,
multireservoir microchip device. Nature Biotechnology, 2006. 24(4): p. 437-438.
Kim, H.K., et al., High-performance liquid chromatographic determination and stability of 5-(3methyltriazen-1-yl)-imidazo-4-carboximide, the biologically active products of the antitumor
agent temozolomide, in human plasma. Journal of Chromatography B, 1997. 703: p. 225-233.
Xue, Y., et al., PDGF-BB modulates hematopoiesis and tumor angiogenesis by inducing
erythropoietin production in stromal cells. Nat Med, 2011. 18(1): p. 100-10.
Berg, S.L., et al., Pharmacokinetics of O-6-benzylguanine and its active metabolite 8-oxo-O-6benzylguanine in plasma and cerebrospinal fluid after intrathecal administration of 0-6benzylguanine in the nonhuman primate. Clinical Cancer Research, 1998. 4(11): p. 2891-2894.
Bindra, D.S., T.D. Williams, and V.J. Stella, Degradation of 0-6-Benzylguanine in Aqueous
Polyethylene-Glycol-400 (Peg-400) Solutions - Concerns with Formaldehyde in Peg-400.
Pharmaceutical Research, 1994. 11(7): p. 1060-1064.
Stefan, T.L., et al., Simultaneous determination of O-6-benzylguanine and 8-oxo-0-6benzylguanine in human plasma by reversed-phase high-performance liquid chromatography.
Journal of Chromatography B, 1997. 704(1-2): p. 289-298.
Berg, S.L., et al., Plasma and Cerebrospinal-Fluid Pharmacokinetics of 0-6-Benzylguanine and
Time-Course of Peripheral-Blood Mononuclear Cell 0-6-Methylguanine-DNA Methyltransferase
Inhibition in the Nonhuman Primate. Cancer Research, 1995. 55(20): p. 4606-4610.
Uhlhaas PJ, S.W., Neural synchrony in brain disorders: relevance for cognitive dysfunctions and
pathophysiology. Neuron, 2006. 5:52((1)): p. 155-168.
DeLong MR, W.T., Circuits and circuit disorders of the basal ganglia. Arch Neurol, 2007. 64(1)(1): p.
20-24.
Zgaljardic DJ, B.J., Foldi NS, Mattis P., A review of the cognitive and behavioral sequelae of
Parkinson's disease: relationship to frontostriatalcircuitry. Cogn Behav Neurol, 2003. 16(4): p.
193-210.
Marsh R, M.T., Peterson BS., Functional disturbances within frontostriatal circuits across multiple
childhoodpsychopathologies. Am J Psychiatry, 2009. 166(6): p. 664-674.
Bonelli RM, C.J., Frontal-subcorticalcircuitry and behavior. Dialogues Clin Neurosci, 2007. 9(2): p.
141-151.
Brody AL, B.M., Bota RG, Saxena S., Prefrontal-subcorticaland limbic circuit mediation of major
depressive disorder. Semin Clin Neuropsychiatry, 2001. 6(2): p. 102-112.
Kessler, R.C., et al., Prevalence, severity, and comorbidity of 12-month DSM-IV disorders in the
National Comorbidity Survey Replication. Arch Gen Psychiatry, 2005. 62(6): p. 617-27.
Joseph, R., Frontal lobe psychopathology: mania, depression, confabulation, catatonia,
perseveration, obsessive compulsions, and schizophrenia. Psychiatry, 1999. 62(2): p. 138-72.
110
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
80.
81.
82.
83.
84.
Watkins, E.R., Constructive and unconstructive repetitive thought. Psychol Bull, 2008. 134(2): p.
163-206.
Bechara A Fau - Damasio, H., et al., The Iowa Gambling Task and the somatic marker hypothesis:
some questions and. Trends Cogn Sci, 2005. 9(4): p. 159-62; discussion 162-4.
Blumberg Hp Fau - Stern, E., et al., Increased anterior cingulate and caudate activity in bipolar
mania. Biol Psychiatry, 2000. 48(11): p. 1045-52.
Hwang J Fau - Lyoo, I.K., et al., Basal ganglia shape alterations in bipolar disorder. Am J
Psychiatry, 2006. 163(2): p. 276-85.
Baumann B Fau - Danos, P., et al., Reduced volume of limbic system-affiliated basal ganglia in
mood disorders. J Neuropsychiatry Clin Neurosci, 1999. 11(1): p. 71-8.
Drevets Wc Fau - Price, J.L., et al., Subgenual prefrontal cortex abnormalities in mood disorders.
Nature, 1997. 386(6627): p. 824-7.
Mayberg Hs Fau - Liotti, M., et al., Reciprocal limbic-cortical function and negative mood:
converging PETfindings in. Am J Psychiatry, 1999. 156(5): p. 675-82.
Cavedini P Fau - Ferri, S., et al., Frontal lobe dysfunction in obsessive-compulsive disorder and
major depression: a. Psychiatry Res, 1998. 78(1-2): p. 21-8.
Chacko Rc Fau - Corbin, M.A., R.G. Corbin Ma Fau - Harper, and R.G. Harper, Acquired obsessivecompulsive disorder associated with basal ganglia lesions. J Neuropsychiatry Clin Neurosci, 2000.
12(2): p. 269-72.
Graybiel, A.M. and S.L. Rauch, Toward a neurobiology of obsessive-compulsive disorder. Neuron,
2000. 28(2): p. 343-7.
Maltby N Fau - Tolin, D.F., et al., Dysfunctional action monitoring hyperactivatesfrontal-striatal
circuits in. Neuroimage, 2005. 24(2): p. 495-503.
Van Laere K Fau - Nuttin, B., et al., Metabolic imaging of anterior capsular stimulation in
refractory. J Nuc| Med, 2006. 47(5): p. 740-7.
Volkow Nd Fau - Fowler, J.S., et al., Role of dopamine, the frontal cortex and memory circuits in
drug addiction. Neurobiol Learn Mem, 2002. 78(3): p. 610-24.
Robbins Tw Fau - Everitt, B.J. and B.J. Everitt, Limbic-striatal memory systems and drug
addiction. Neurobiol Learn Mem, 2002. 78(3): p. 625-36.
Thurman D,A.C., Dunn K, Guerrero J, Sniezek J, Traumatic brain injury in the United States: a
public health perspective. Journal of Head Trauma Rehabilitation, 1999. 14(6): p. 602-15.
Chen JW, R.R., Eavey R, Wasterlain CG Posttraumatic epilepsy and treatment. J Rehabil Res Dev,
2009. 46(6): p. 685-696.
Tomkins 0, F.A., Benifla M, Cohen A, Friedman A, Shelef 1, Blood-Brain Barrier Breakdown
Following Traumatic Brain Injury: A Possible Role in Posttraumatic Epilepsy. Cardiovascular
Psychiatry and Neurology, 2011.
McKee AC, C.R., Nowinski CJ, Hedley-Whyte E, Gavett BE, Budson AB, Santini EB, Hyo-Soon Lee,
Kubilus C Stern RA, Chronic Traumatic Encephalopathy in Athletes: Progressive Tauopathy
following Repetitive Head Injury. J Neuropathol Exp Neurol, 2009. 68(7): p. 709-735.
Simuni T, S.M., Does levodopa accelerate Parkinson's disease? Drugs Aging, 1999. 14(6): p. 399408.
Heiss, J.D., et al., Image-guided convection-enhanced delivery of muscimol to the primate brain. J
Neurosurg, 2010. 112(4): p. 790-5.
Fiandaca, M.S., et al., Image-guided convection-enhanced delivery platform in the treatment of
neurological diseases. Neurotherapeutics, 2008. 5(1): p. 123-127.
Kawakami, K., et al., Distribution kinetics of targeted cytotoxin in glioma by bolus or convectionenhanced delivery in a murine model. J Neurosurg, 2004. 101(6): p. 1004-1011.
111
85.
86.
87.
88.
89.
90.
91.
92.
93.
94.
95.
96.
Raghavan, R., et al., Convection-enhanced delivery of therapeutics for brain disease, and its
optimization. Neurosurgical Focus, 2006. 20(4): p. E12.
Chen, J., et al., A multichannel neural probe for selective chemical delivery at the cellular level.
IEEE Trans Biomed Eng, 1997. 44(8): p. 760-9.
Neeves, K.B., et al., Fabrication and characterization of microfluidic probes for convection
enhanced drug delivery. Journal of Controlled Release, 2006. 111(3): p. 252-262.
Neeves, K.B., et al., Dilation and degradation of the brain extracellular matrix enhances
penetration of infused polymer nanoparticles. Brain Research, 2007. 1180: p. 121-132.
Foley, C., et al., Flexible microfluidic devices supported by biodegradable insertion scaffolds for
convection-enhanced neural drug delivery. Biomedical Microdevices, 2009. 11(4): p. 915-924.
Rousche, P.J. and R.A. Normann, Chronic recording capability of the Utah Intracortical Electrode
Array in cat sensory cortex. J Neurosci Methods, 1998. 82(1): p. 1-15.
Kipke, D.R., et al., Silicon-substrate intracortical microelectrode arrays for long-term recording of
neuronal spike activity in cerebral cortex. IEEE Trans Neural Syst Rehabil Eng, 2003. 11(2): p. 1515.
Szarowski, D.H., et al., Brain responses to micro-machined silicon devices. Brain Research, 2003.
983(1-2): p. 23-35.
Edell, D.J., et al., Factors influencing the biocompatibility of insertable silicon microshafts in
cerebral cortex. Biomedical Engineering, IEEE Transactions on, 1992. 39(6): p. 635-643.
Buchko, C.J., K.M. Kozloff, and D.C. Martin, Surface characterization of porous, biocompatible
protein polymer thin films. Biomaterials, 2001. 22(11): p. 1289-300.
Thelin, J., et al., Implant size and fixation mode strongly influence tissue reactions in the CNS.
PLoS One, 2011. 6(1): p. e16267.
Zhong, Y. and R.V. Bellamkonda, Dexamethasone-coated neural probes elicit attenuated
inflammatory response and neuronal loss compared to uncoated neural probes. Brain Research,
2007. 1148: p. 15-27.
112