A biophysical marker of severity in sickle cell disease Please share

advertisement
A biophysical marker of severity in sickle cell disease
The MIT Faculty has made this article openly available. Please share
how this access benefits you. Your story matters.
Citation
Wood, D. K., A. Soriano, L. Mahadevan, J. M. Higgins, and S. N.
Bhatia. “A Biophysical Indicator of Vaso-Occlusive Risk in Sickle
Cell Disease.” Science Translational Medicine 4, no. 123
(February 29, 2012): 123ra26–123ra26.
As Published
http://dx.doi.org/10.1126/scitranslmed.3002738
Publisher
American Association for the Advancement of Science (AAAS)
Version
Author's final manuscript
Accessed
Thu May 26 20:49:41 EDT 2016
Citable Link
http://hdl.handle.net/1721.1/86116
Terms of Use
Creative Commons Attribution-Noncommercial-Share Alike
Detailed Terms
http://creativecommons.org/licenses/by-nc-sa/4.0/
NIH Public Access
Author Manuscript
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
NIH-PA Author Manuscript
Published in final edited form as:
Sci Transl Med. 2012 February 29; 4(123): 123ra26. doi:10.1126/scitranslmed.3002738.
A biophysical marker of severity in sickle cell disease
David K. Wood1, Alicia Soriano2, L. Mahadevan3,5,*, John M. Higgins4,5,*, and Sangeeta N.
Bhatia1,6,7
1Division of Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge,
MA 02139, USA
2Special
Hematology Laboratory, Brigham and Women’s Hospital, Boston, MA 02115, USA
3School
of Engineering and Applied Sciences, Department of Physics, Department of Organismic
and Evolutionary Biology, Wyss Institute for Biologically Inspired Engineering and Kavli Institute
for Nano bio Science and Technology, Harvard University, Cambridge, MA 02138, USA
4Center
for Systems Biology and Department of Pathology, Massachusetts General Hospital,
Boston, MA 02114, USA
NIH-PA Author Manuscript
5Department
of Systems Biology, Harvard Medical School, Boston, MA 02115, USA
6Howard
Hughes Medical Institute , Department of Electrical Engineering and Computer Science
and Koch Institute for Integrative Cancer Research, Massachusetts Institute of Technology,
Cambridge, MA 02139, USA
7Division
of Medicine, Brigham and Women’s Hospital, Boston, MA 02115, USA
Abstract
NIH-PA Author Manuscript
The search for predictive biomarkers of disease has largely focused on molecular indicators;
however, mechanical and biophysical markers, which can integrate multiple pathways across
length scales, may provide a more global picture of the underlying pathophysiology. Sickle cell
disease, the first disease to have its molecular origins decoded, affects millions of people
worldwide and has been studied intensively at the molecular, cellular, tissue and organismal level
for a century since its initial description. However, there are still few, if any, markers that allow us
to characterize the severity of this disease. Because the complications of sickle cell disease are
largely due to vaso-occlusive events, we hypothesized that a physical metric characterizing the
vaso-occlusive process could serve as a marker of disease severity. Here we use a simple
microfluidic device to characterize the dynamics of jamming in physiologically relevant
conditions, using the rate of change of the resistance to flow following a sudden deoxygenation
event. Our studies show that this single biophysical parameter could be used to distinguish
between patients with divergent clinical outcomes, unlike existing laboratory tests. Our assay
provides a biophysical marker of disease severity that could be used to guide timing of clinical
interventions, to monitor the progression of the disease, and to measure the efficacy of drug
response, transfusion, and novel small molecules in an in vitro setting.
Introduction
Biomarkers for diseases are critical to guide timing of clinical interventions, to monitor
treatment efficacy, and to speed the development of new therapies. The search for such
markers has yielded many new genetic, epigenetic, and proteomic indicators [1–5], but the
need for new markers continues to outpace their validation and use [6]. This problem is
*
To whom correspondence should be addressed (lm@seas.harvard.edu, john_higgins@hms.harvard.edu).
Wood et al.
Page 2
NIH-PA Author Manuscript
particularly acute for some diseases, such as sickle cell disease, where no molecular markers
have been found to correlate reliably with the clinical picture [7–10]. Markers that provide
integrative systemic readouts are complementary to molecular indicators; for example,
biophysical markers are already used clinically and range from long-standing measurements
such as blood pressure to newer imaging modalities such as Doppler ultrasound and cerebral
flow magnetic resonance imaging [11–13]. Because these biophysical metrics integrate
multiple length scales, molecular players, and pathways, they have the potential to correlate
better with disease state than their molecular counterparts. Even at the cellular scale, recent
studies suggest new possibilities for biophysical markers such as cell motility or cell
deformability as an important indicator of metastatic potential in cancer [14, 15] and other
diseases [16, 17].
NIH-PA Author Manuscript
Sickle cell disease, which affects more than 13 million people worldwide [18] and costs
more than US$1.1 billion per year in the United States alone [19], is an ideal candidate for a
biophysical marker. The molecular origin of the disease is a mutation in the β-globin gene
that results in a variant hemoglobin molecule (hemoglobin S, or HbS) that polymerizes into
long chains and extended gels upon deoxygenation [20, 21]. Sickle cells (erythrocytes or
RBCs containing HbS) stiffen when polymerization occurs, causing changes in sickle cell
blood flow, which can lead to clinical vaso-occlusive crisis events [22]. Although the
molecular origins of sickle cell disease have been known for more than 60 years [23], the
factors and processes that link these molecular and cellular events to clinical outcomes in
patients remain elusive [24–26]. This search for mechanisms is particularly challenging
because pathogenesis of vaso-occlusion involves a number of processes across multiple
length and time scales, including polymerization and melting of hemoglobin, stiffening and
morphological change of RBCs, and increasing apparent viscosity of blood. Each of these
processes has been examined separately in various in vitro systems [20, 21, 27–30].
However, studies that integrate all of these processes in physiologic regimes are rare, even
though their integrated effects are likely decisive in determining clinical outcomes [28, 31].
NIH-PA Author Manuscript
In the absence of objective biomarkers of disease severity [7–10], it is difficult to predict the
frequency of painful and damaging vaso-occlusive crises and overall severity of disease
from one patient to the next. Common laboratory tests measure HbS fraction and fetal
hemoglobin (HbF) fraction, neither of which is predictive of clinical outcome for individual
patients [9, 32]. Thus we have no method to predict which patients will require urgent or
frequent treatment [8, 10], and it is very difficult to assess efficacy of different treatments.
This lack of clear predictors also confounds the development of new therapeutics because
we have no reliable in vitro methods to test which molecules will work in vivo [22]. Because
the morbidity and mortality are caused by vaso-occlusions [33], biophysical markers might
provide a link to clinical phenotype. For example, in vivo cerebral blood flow velocity
measurements have shown value in guiding transfusion therapy for pediatric sickle cell
patients [34–36], suggesting that it may be possible to translate the extensive research on
biophysical aspects of sickle cell disease [37] into in vitro biophysical markers.
Recent studies have shown that it is possible to halt the flow of sickle cell blood in a
capillary-sized microchannel under constant pressure simply by lowering the oxygen
concentration, thereby simulating the most basic features of a vaso-occlusive event [38, 39].
Here we show that a measurement of the dynamics of the integrated physical vaso-occlusive
process in vitro serves as a biophysical marker of pathophysiology in a physiologically
relevant range of hemoglobin concentration, hemoglobin composition, RBC volume
fraction, and local oxygen concentration [28]. This measurement characterizes the dynamics
of blood rheological properties during in vitro vaso-occlusion (Scheme 1, Movie S1). Our
experiments evaluate this measurement as a biophysical marker of disease severity in
patients by benchmarking it against currently available clinical laboratory tests for sickle
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
Wood et al.
Page 3
cell disease, and as an in vitro assessment of the efficacy of existing and experimental
treatments.
NIH-PA Author Manuscript
Results
Quantifying Blood Conductance Dynamics
NIH-PA Author Manuscript
We developed a microfluidic system to measure how the flow of sickle cell blood changes
after deoxygenation. The system includes a microfluidic device with a capillary-sized
channel that is diffusively coupled to a gas reservoir (Scheme 1). This system allows control
over many parameters important to vaso-occlusion, including channel size, blood pressure,
and oxygen concentration. Whole EDTA-anticoagulated human blood is driven through the
channel under constant pressure, while gas with prescribed oxygen concentration is driven
through the gas reservoir. RBCs are imaged in the Soret band [40] allowing determination of
hemoglobin oxygenation state based on a change in optical absorption (Fig. S1). As sickle
cells become deoxygenated, hemoglobin polymerizes, leading to cell stiffening and sickling
and an increase in the blood’s resistance to flow with a resulting drop in flow velocity
despite being subject to a constant driving pressure head, i.e. there is a drop in the blood
conductance defined as velocity per unit pressure drop and inversely proportional to the
effective viscosity of the suspension. By measuring the changes in flow velocity for a given
constant applied pressure head, we can quantify the kinetics of the change in conductance
for a blood sample following deoxygenation, as shown in Figure 1 for a blood sample from a
sickle cell patient (Movie 1 shows a full deoxygenation-reoxygenation cycle with vasoocclusion and recovery). When the oxygen concentration in the gas reservoir is reduced
rapidly (≈20 s, Fig. 1a) from 21% to 0%, it causes hemoglobin desaturation (see Fig. S2 for
oxygen measurements in blood channel), cell deoxygenation, stiffening and the overall
blood flow velocity decrease though the pressure drop across the device remains unchanged
(Fig. 1b, right y-axis). In contrast, blood samples from individuals with either normal
hemoglobin A or sickle cell trait show no response (see Fig. S3), and the steady-state
velocity at 21% oxygen depends linearly on the pressure drop, consistent with Poiseuille’s
Law[41], with the maximum shear rates in the middle of the channel in the range 100–400/s,
similar to shear rates found in the in vivo microcirculation [42]. For sickle cell blood
deoxygenation is followed by a velocity that also varies linearly with applied pressure, i.e.
the rate of decrease in conductance is independent of pressure as shown in Fig. 1c. We thus
use the rates of conductance decrease in blood samples from different sickle cell patients as
measures of disease severity.
Conductance Dynamics Correlate with Clinical Outcomes
NIH-PA Author Manuscript
If the rate of conductance decrease captures critical aspects of in vivo sickle cell
pathophysiology, then we would expect blood samples from the most severely affected
patients to show a more rapid reduction in conductance than samples from the least severely
affected patients. To test this hypothesis, we used simple criteria to distinguish “benign” and
“severe” blood samples. A “benign” sample is one collected from a sickle cell patient who
had not been treated with hydroxyurea, had not received any transfusions, and had no
unscheduled medical visits for a sickle cell disease-related indication in the 12 months
before or 2 months after the date of sample collection (see Methods); a “benign” blood
sample shows little or no drop in conductance following deoxygenation (Fig. S4). “Severe”
samples were those from sickle cell patients who had either received a transfusion or made
an unscheduled hospital visit for a sickle cell disease-related indication in the 12 months
before or 2 months after the date of sample collection; Figure 1 shows a large and rapid
conductance change following deoxygenation for a “severe” blood sample. Tests were
carried out with 23 severe blood samples and 6 benign blood samples, each from a unique
patient; all showed similar oxygenated conductances after adjustment to a hematocrit of
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
Wood et al.
Page 4
NIH-PA Author Manuscript
25%. (See Methods for details on conductance calculation). In Figure 2a, we show that
deoxygenation tests indicate that the rate of conductance decrease correlates strongly with
benign and severe classification for this set of blood samples, averaged over at least 3
deoxygenation experiments. Figure S5 shows de-oxygenation trials for 4 additional severe
blood samples, and Table S1 gives raw data for all samples. Blood samples from patients
with more benign disease showed a significantly (p < 0.01) slower decrease in the
conductance following deoxygenation. The conductance of the median severe sample fell
more than twice as fast as that of the median benign sample. We also calculated the receiver
operating characteristic (ROC), which characterizes the sensitivity and specificity for a
binary test as a function of the discrimination threshold, and the area under the curve (AUC),
which quantifies the ability of the test to discriminate between states [6, 43]. In contrast to a
random prediction (Fig. 2b, dashed line), which has no information content, our
measurement (Fig. 2b, solid line) demonstrates strong diagnostic potential with an area AUC
of 0.85 (95% confidence level = 0.65 – 0.95).
NIH-PA Author Manuscript
We explored the possibility that this effect could be explained by other blood parameters,
and we found no systematic difference between the benign and severe groups for HbS
fraction (Fig. 2c), unadjusted hematocrit (HCT, Fig. S6a), mean hemoglobin concentration
(MCHC, Fig. S6b), sample storage time (Fig. S6c) or steady-state oxygenated conductance
(Fig. S6d). We also considered the role of white blood cells (WBCs) and found no
correlation between sample severity and WBC count (Fig. S6e). We expect that most WBCs
are not viable at the time of testing and therefore that WBCs are not responsible for the
differences in rheodynamics. We further explored this hypothesis by depleting WBCs from a
severe sample and showing that the rheodynamics were not significantly altered (Fig. S6f),
i.e. none of these can serve as a predictive biomarker for disease severity.
NIH-PA Author Manuscript
Another potential explanation of our results is that the benign group has a higher fraction of
HbF, which inhibits HbS polymerization [44]. In order to compare HbF fractions between
the benign and severe samples, we first subdivided the severe samples into those from
patients who had been or had not been treated with hydroxyurea (“treated severe” or
“untreated severe”), a drug that works in part by increasing HbF expression [45, 44]. As
shown in Fig. 2d, benign blood samples had higher HbF fractions than untreated severe
samples. However, the HbF fractions for treated severe samples were comparable to those of
the benign samples. Despite elevated HbF fraction, the hydroxyurea-treated patients had a
more severe clinical course, and their blood samples showed rates of conductance decrease
comparable to those of untreated severe samples (Fig. 2e). While increased HbF fraction
may account for some of the difference in rates of conductance decrease between benign
samples and untreated severe samples, HbF fraction cannot account for the difference
between the benign samples and the treated severe samples. Further, if HbF fraction alone
determined the rate of conductance decrease, then one would not expect the untreated and
treated severe samples to show such similar rates of conductance decrease. We also assessed
the role of cell volume by comparing the mean cell volumes (MCV) in the groups.
Hydroxyurea is known to raise MCV, and we find a slight elevation in MCV in the
hydroxyurea-treated group as expected (see Fig. S6g). We find no significant different
between the benign and untreated severe groups, however, making it unlikely that
differences in RBC volume are responsible for the differences in rheodynamics.
Simulated Transfusion Reduces Rates of Conductance Change
We also tested whether a simulated transfusion of normal (hemoglobin A, or HbA) blood to
lower HbS fraction would lead to slower declines in conductance analogous to the way that
it improves patient clinical status [45]. We mixed ABO- and RhD-compatible HbA blood
with blood samples from sickle cell patients. Figure 2f shows how the flow dynamics of 4
severe samples changed following a simulated transfusion that reduced the HbS fraction by
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
Wood et al.
Page 5
NIH-PA Author Manuscript
about one third. As expected, lower HbS fraction is associated with slower conductance
decreases for each sample. The magnitude of the effect is quite variable between patients.
Neither the initial HbS fraction nor the fractional decrease predicts how much slower the
conductance would fall.
Small Molecule Treatment Modulates Rate of Conductance Change
If the rate of conductance decrease correlates with clinical outcome, small molecules with
demonstrated clinical potential should reduce the rate of conductance decrease. Recent work
has shown that a small molecule that increases hemoglobin oxygen affinity (5hydroxymethyl furfural, or 5HMF) can significantly improve phenotypes in a mouse model
of sickle cell disease [46]. We used our system to assess the in vitro effect of this
experimental therapy on the rates of conductance decrease in patient blood samples. Figure
3 compares a sample that was treated with either 0 mM (Fig. 3a) or 10 mM 5HMF (Fig. 3b).
The results are dramatic, with a large and rapid conductance decrease in response to
deoxygenation for the untreated sample but very little conductance change for the treated
sample. Treatment with 5HMF causes a 5-fold slower conductance decrease (Fig. 3c). This
marked rheodynamic response is associated with a slower change in the optical absorption
measurement (Fig. 3b) as the treated sample unloads oxygen much more slowly than the
untreated sample.
NIH-PA Author Manuscript
Discussion
NIH-PA Author Manuscript
Our microfluidic system enables us to make a novel measurement of the speed of the
rheological response of patient blood samples to deoxygenation by integrating the
molecular, cellular, and rheological processes that likely occur during in vivo vasoocclusion, in sharp contrast with previous approaches that have typically focused on just one
aspect of the vaso-occlusive process, such as HbS polymerization, RBC sickling, or blood
rheology. We have shown that the rate of change in the rheological properties of blood
samples, characterized by the rate of change in conductance during an in vitro vasoocclusive event, is strongly correlated with overall patient disease severity for the set of
patients in our study. Our finding in this patient cohort is statistically robust (p < 0.01) with
> 90% statistical power (for p = 0.05). From a diagnostic perspective, our biophysical
marker, with an AUC = 0.85, compares well against other biomarkers in development, such
as a micro-NMR based cancer diagnostic (AUC = 0.44 – 0.88) and urinary biomarkers for
kidney injury (AUC = 0.73 – 0.91) [47, 48]. This diagnostic efficiency contrasts sharply
with existing laboratory tests such as HbS and HbF fractions which are much more weakly
correlated with patient outcomes [10, 49]. Because we find no benign samples with fast
rheodynamics, this novel measurement may be most useful for identifying patients at
increased risk of complications. Although the sudden and seemingly unpredictable onset of
vaso-occlusive crises suggests that it may never be possible to predict the exact timing of
individual occlusive events, our measurement may provide the basis for an objective and
accurate tool to identify those patients who have not responded as intended to hydroxyurea
or transfusion or who may benefit most from initiation such treatment.
Our study highlights the potential diagnostic power of a biophysical marker to integrate
multiple interdependent pathophysiological processes, in contrast to previous studies at the
molecular or cellular level [20, 21, 27–30]. Furthermore, we see that a previously
unappreciated difference between patients is found in the dynamics of blood rheological
properties, reflecting the pathophysiological transition of a soft suspension that flows
steadily to a jammed one when the cells stiffen in response to deoxygenation. Thus, it should
not be surprising that steady-state values, such as hematocrit, mean corpuscular volume,
mean hemoglobin concentration, WBC counts, and oxygenated conductance do not correlate
with clinical outcome, while the dynamics of flow do correlate with the occlusive events
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
Wood et al.
Page 6
NIH-PA Author Manuscript
ultimately responsible for disease severity. Indeed, more generally, stasis and hyperviscosity
lead to thrombosis in vivo and thus imply a lower limit on steady-state velocity in the
microcirculation [50]. It is therefore not surprising that the key determinant of vasoocclusion is the rate at which flow approaches this threshold. In our measurements, the
patients most at risk are those whose blood shows a rapid conductance change, indicating a
high probability that it would slow sufficiently quickly in the microcirculation to produce a
vaso-occlusion. Understanding the occurrence of these extreme events requires a dynamical
assay such as that characterized here.
In summary, the strong correlation between blood rheodynamics and the clinical outcomes
in our study population provides a valuable tool for scientific discovery, drug development,
and possibly for patient monitoring and clinical decision-making in sickle cell disease. For
example, although transfusion of HbA blood to lower the HbS fraction is the most common
treatment for sickle cell disease in the US and provides short-term improvement [45], it is
not clear what hematocrit and HbS fraction targets are optimal for each patient; efficacy in
one cohort can lead to complications in another [8]. Our quantitative measurements of the
effect of simulated transfusion on rates of conductance change (Fig 2f) may provide
additional guidance.
NIH-PA Author Manuscript
Our approach may provide a better way to assess the efficacy of experimental
pharmaceuticals in humans. Animal models of sickle cell disease are limited by significant
phenotypic differences from human patients and the complexity of experimental protocols
[7]. In contrast, our in vitro system can be used to identify potential drugs and quantify their
efficacy much more easily. As an example, a reduced rate of conductance change following
treatment of a blood sample with 5HMF (Fig. 3), is consistent with the known effects of this
small molecule on increasing hemoglobin oxygen affinity, and thus slowing the kinetics of
deoxygenation and subsequent polymerization of HbS [46]. However, these increases in
hemoglobin oxygen affinity are large, and it is possible that a much milder increase in
hemoglobin oxygen affinity may be sufficient to alter flow dynamics in vitro and possibly in
vivo.
NIH-PA Author Manuscript
Our results also help focus the question of why hydroxyurea treatment has such variable
efficacy [44, 51]. Hydroxyurea is currently the only approved pharmaceutical treatment for
sickle cell disease, and it works in part by increasing the HbF fraction [10, 44, 52].
However, many patients treated with hydroxyurea show large increases in HbF fraction
without significant clinical benefits. HbF is known to slow HbS polymerization and should
therefore change the kinetics of RBC stiffening and blood conductance change [44].
Surprisingly, we find no consistent effect of total HbF fraction in a blood sample on the rate
of conductance decrease, which is consistent with previous studies showing that the fraction
of HbF in treated patients is not a useful predictor of clinical outcomes in individual patients
[32]. One existing hypothesis to explain this result is that the uniformity of the HbF
distribution across all RBCs may be more important than the total fraction of HbF [8, 44,
53]. This hypothesis is consistent with our findings, and our work focuses the question on
analyzing the effects of HbF distribution on rate of conductance decrease and its correlation
with patient clinical outcomes.
More generally, although our study is limited to a small cohort of patients and uses simple
criteria for the “severity” of the disease, our results are robust and motivate further work to
predict clinical course in larger populations over long periods of time. Our work also
provides an impetus for the development of integrative markers of disease that link multiple
molecular and cellular pathways across diverse length and time scales as tools to
complement the growing biomolecular armamentarium in the quest to predict clinical
outcomes and reveal the mechanisms underlying for human disease.
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
Wood et al.
Page 7
Methods
Blood Specimens
NIH-PA Author Manuscript
Blood specimens were collected during the normal course of patient care at two tertiary care
hospitals and used in experiments in accordance with a research protocol approved by the
Partners Healthcare Institutional Review Board. Blood samples were collected in 5-ml
EDTA vacutainers and stored at 4°C for up to 14 days. Hematocrit was determined using a
Siemens ADVIA 2120 automated analyzer. Hemoglobin fractions were determined using
cellulose agar electrophoresis and confirmed by high performance liquid chromatography
(HPLC) with a Tosoh G7 column. Blood samples used in simulated transfusion experiments
were tested to exclude the presence of any RBC alloantibodies.
For all measurements, hematocrit was adjusted to 25% by fractionating RBCs and plasma in
a centrifuge at 3500 rpm for 10min. Separated plasma was then removed or added to adjust
total hematocrit. For transfusion studies, hematocrit-adjusted sickle blood was mixed with
type-matched HbA blood to achieve the desired sickle cell fraction. In order to treat with 5hydroxymethyl furfural (5HMF, Sigma), we first dissolved 1M of drug in DMSO. Blood
samples were treated with appropriate dose and incubated at 37°C for 1hr. After incubation,
samples were transferred to 4°C until ready for use.
NIH-PA Author Manuscript
Patient blood samples were classified as “benign” if all of the following criteria were met
during the 12 months prior to sample collection and the 2 months following: no patient
hospital admissions or emergency room visits for sickle cell disease-related complications,
no blood transfusions for sickle cell disease-related indications, and no hydroxyurea use.
Patient blood samples were classified as “severe” if the patient had had at least one hospital
admission or emergency room visit for a sickle cell disease-related complication or a blood
transfusion for a sickle cell disease-related indication in the 12 months prior to sample
collection or the 2 months following.
Device Fabrication
NIH-PA Author Manuscript
The device comprises 3 layers: the blood capillary, the gas reservoir, and a reservoir for
hydration (Scheme 1). Each of these layers is separated from the adjacent layers by a 100
µm thick polydimethylsiloxane (PDMS) membrane, which allows for gas transport between
the layers (Scheme 1). The blood capillary is 7 cm long with cross section 15 µm by 10 µm.
The hydration layer, which is required to prevent dehydration of the blood through the
membrane [54], is 30 cm long with cross section 1.5 mm by 100 µm. The gas reservoir
completely overlaps the hydration layer but is 150 µm tall. The layers were fabricated in
PDMS using soft lithography. The blood layer master was made by spinning SU-8 2015
(MicroChem Corp.) onto a silicon wafer at 3500 rpm for 30 s and then photocrosslinking the
SU-8 through a 50,000 dpi transparency mask. The hydration and gas layer molds were
made by spinning SU-8 2050 at 1500 rpm and 1000 rpm, respectively, for 30 s. The PDMS
layers were made and assembled using thermal bonding. For the hydration layer and
membrane, we mixed PDMS with a 20:1 elastomer:curing agent ratio and spun this onto the
mold at 500 rpm for 30s. The gas layer was cast at 1 cm thickness using a 5:1
elastomer:curing agent ratio and bonded to the hydration layer at 80°C for 2h. The blood
layer and corresponding membrane was made by mixing PDMS with a 5:1 elastomer:curing
agent ratio and spin casting using the same conditions as for the hydration layer. The gas and
hydration reservoirs were then removed from the hydration layer mold and bonded to the
blood layer at 80°C overnight. The final structure was then plasma bonded to a glass
microscope slide.
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
Wood et al.
Page 8
Device Operation
NIH-PA Author Manuscript
The experimental setup is shown in Scheme 1. Gas was flowed through the top chamber of
the device (Scheme 1) under constant pressure (1 psi) from precision pressure regulators
(PRG200-25, Omega). We switched between 0% oxygen (N2) and 21% oxygen (21% O2,
5% CO2, balance N2) using solenoid valves downstream of the regulators. We flowed 1X
phosphate buffered saline (PBS) in the hydration layer at 1 ml/hr using a syringe pump. This
aqueous layer prevented dehydration of the blood. Whole blood (EDTA-anti-coagulated)
from human patients was flowed through the capillary under constant pressure, provided by
compressed N2, using a digital pressure regulator (PCD, Alicat Scientific).
Oxygen Measurement
NIH-PA Author Manuscript
Oxygen concentration of the gas phase was measured using a fiber optic sensor (FOXY,
Ocean Optics) connected to the outlet of the gas reservoir. We also measured the oxygen
saturation of the RBCs directly using an optical technique. An optical filter (Brightline 434
+/− 17 nm, Semrock) was placed between the light source and the device (Scheme 1). The
transmission band of this filter is centered on an absorption peak for deoxy-hemoglobin
(Fig. S1). When the hemoglobin is oxygenated, the absorption peak shifts (20 nm) out of the
transmission band of the filter (Fig. S1), causing deoxygenated cells to absorb more light
and appear dark, while oxygenated cells remain transparent (Fig. S1 inset). A measurement
of the RBC optical density through the filter revealed the oxygenation state of the cells (Fig.
S1). For calibration of the measurement, oxygen concentration in the blood channel was
measured directly using an oxygen-sensitive luminescence probe (Fig. S2). Tris(4,7diphenyl-1,10-phenanthroline)ruthenium(II) dichloride complex [55] was dissolved in PBS
and flowed through the blood channel under the same conditions as the blood measurement.
Luminescence was measured under illumination at 488 nm.
Blood Conductance and Dynamics
NIH-PA Author Manuscript
Blood rheology was measured by capturing high speed videos in real time and performing
video tracking of individual RBCs offline. Using this method we were able to compute
instantaneous blood velocities over very short time intervals and track changes in blood
velocity as a function of RBC oxygenation state. Video processing methods were described
in detail previously [39]. Briefly, cells in each video frame were identified computationally
based on morphologic criteria. Cell locations in subsequent frames were linked to form
trajectories using heuristics and machine learning techniques. We defined the velocity at
each point in time as the median cell velocity calculated over a 32-frame video captured at
higher than 200fps. We excluded any noisy videos where the inter-quartile range of
calculated velocities was greater than the larger of the median velocity or 40 um/s. All
rheology data is reported as device conductance, which was calculated by dividing the
median blood velocity by the applied pressure bias. Stretches of constant oxygen and
pressure were identified automatically, and the beginning and ending of conductance
transients were identified by examining a 5-point moving average of the conductance. Once
the time range was defined for a conductance transient, rates of conductance change were
computed and fit to a linear function using the robust least-squares method available in the
MATLAB curve fitting toolbox (MathWorks). Average rates of conductance change for
each sample were defined by taking the mean of calculated transients from at least 3
independent deoxygenation cycles. Nonparametric analysis to compare groups of samples
was performed using the Mann-Whitney non-parametric test in the MATLAB statistics
toolbox [56]. Power analysis was performed using bootstrapping methods (resampling with
replacement from our data set) along with numerically simulated trials (N=100,000) [57].
Experiments flowing whole blood in microfluidic channels are prone to fouling,
delamination, and other artifacts. We assessed the integrity of each deoxygenation/
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
Wood et al.
Page 9
reoxygenation cycle by measuring the conductance before and after. Any significant (>20%)
change likely reflects device compromise, and the results of such transients were excluded.
NIH-PA Author Manuscript
Supplementary Material
Refer to Web version on PubMed Central for supplementary material.
Acknowledgments
We thank the many patients who generously donated blood for this study. We thank Elyse Mandell, Rosy Jaromin,
Hasmukh Patel, Svetal Patel, Maureen Hames, Patricia Guglietta, Mary Huang, David Ebb, Lorna Lampert,
Christine Markopoulos, and Amanda Griffin for help acquiring and testing blood samples. We thank Frank Bunn,
Carlo Brugnara, and Maureen Okam for helpful discussions. All video processing was performed on the Harvard
Medical School Orchestra Computing Cluster. We thank the Stanford Microfluidics foundry and the MIT
Microtechnology Laboratory for fabrication support. DKW acknowledges support from the National Institute of
Biomedical Imaging and Bioengineering for an NRSA fellowship. LM acknowledges the Kavli Institute for Nanbio
Science and Technology at Harvard and the MacArthur Foundation for partial support. JMH acknowledges support
from National Institute of Diabetes and Digestive and Kidney Diseases Grant DK083242. SNB is a Howard Hughes
Investigator.
References
NIH-PA Author Manuscript
NIH-PA Author Manuscript
1. Small EM, Olson EN. Pervasive roles of microRNAs in cardiovascular biology. Nature. 2011;
469:336–342. [PubMed: 21248840]
2. Brower V. Biomarkers: Portents of malignancy. Nature. 2011; 471:S19–S21. [PubMed: 21430715]
3. Miller G. Alzheimer’s biomarker initiative hits its stride. Science. 2009; 326:386–389. [PubMed:
19833956]
4. Bonventre JV, Vaidya VS, Schmouder R, Feig P, Dieterle F. Next-generation biomarkers for
detecting kidney toxicity. Nat. Biotech. 2010; 28:436–440.
5. Whitfield ML, George LK, Grant GD, Perou CM. Common markers of proliferation. Nat. Rev.
Cancer. 2006; 6:99–106. [PubMed: 16491069]
6. Warnock DG, Peck CC. A roadmap for biomarker qualification. Nat. Biotech. 2010; 28:444–445.
7. Schechter AN. Hemoglobin research and the origins of molecular medicine. Blood. 2008;
112:3927–3938. [PubMed: 18988877]
8. Steinberg MH, Hebbel RP. Clinical diversity of sickle cell anemia: Genetic and cellular modulation
of disease severity. Am. J. Hematol. 1983; 14:405–416. [PubMed: 6190397]
9. Brittenham GM, Schechter AN, Noguchi CT. Hemoglobin S polymerization: primary determinant
of the hemolytic and clinical severity of the sickling syndrome. Blood. 1985; 65:183–189.
[PubMed: 3965046]
10. Platt OS, Thorington BD, Brambilla DJ, Milner PF, Rosse WF, Vichinsky E, Kinney TR. Pain in
sickle cell disease. Rates and risk factors. N. Engl. J. Med. 1991; 325:11–16. [PubMed: 1710777]
11. Jain RK, Schlenger K, Hockel M, Yuan F. Quantitative angiogenesis assays: Progress and
problems. Nat. Med. 1997; 3:1203–1208. [PubMed: 9359693]
12. Cohen LS, Escobar PF, Scharm C, Glimco B, Fishman DA. Three-dimensional power Doppler
ultrasound improves the diagnostic accuracy for ovarian cancer prediction. Gynaecol. Oncol.
2001; 82:40–48.
13. Wolf RC, Gron G, Sambataro F, Vasic N, Wolf ND, Thomann PA, Saft C, Landwehrmeyer GB,
Orth M. Magnetic resonance perfusion imaging of resting-state cerebral blood flow in preclinical
Huntington’s disease. J. Cereb. Blood Flow Metab. 2011
14. Karnoub AE, Dash AB, Vo AP, Sullivan A, Brooks MW, Bell GW, Richardson AL, Polyak K,
Tubo R, Weinberg RA. Mesenchymal stem cells within tumour stroma promote breast cancer
metastasis. Nature. 2007; 449:557–563. [PubMed: 17914389]
15. Chaffer CL, Weinberg RA. A perspective on cancer cell metastasis. Science. 2011; 331:1559–
1564. [PubMed: 21436443]
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
Wood et al.
Page 10
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript
16. Cross SE, Jin Y-S, Rao J, Gimzewski JK. Nanomechanical analysis of cells from cancer patients.
Nat. Nano. 2007; 2:780–783.
17. Lee GY, Lim CT. Biomechanics approaches to studying human diseases. Trends in Biotechnology.
2007; 25:111–118. [PubMed: 17257698]
18. Weatherall DJ. The inherited diseases of hemoglobin are an emerging global health burden. Blood.
2010; 115:4331–4336. [PubMed: 20233970]
19. Ballas SK. The cost of health care for patients with sickle cell disease. Am. J. Hematol. 2009;
84:320–322. [PubMed: 19415728]
20. Noguchi CT, Torchi DA, Schechter AN. Determination of deoxyhemoglobin S polymer in sickle
erythrocytes upon deoxygenation. Proc. Natl. Acad. Sci. U.S.A. 1980; 77:5487–5491. [PubMed:
6933568]
21. Eaton WA, Hofrichter J. Sickle cell hemoglobin polymerization. Adv. Protein Chem. 1990; 40:63–
279. [PubMed: 2195851]
22. Wood AJ, Steinberg MH. Management of sickle cell disease. N. Engl. J. Med. 1999; 340:1021–
1030. [PubMed: 10099145]
23. Orkin SH, Higgs DR. Sickle cell disease at 100 years. Science. 2010; 329:291–292. [PubMed:
20647458]
24. Gladwin MT, Barst RJ, Castro OL, Gordeuk VR, Hillery CA, Kato GJ, Kim-Shapiro DB, Machado
R, Morris CR, Steinberg MH, Vichinsky EP. Pulmonary hypertension and NO in sickle cell.
Blood. 2010; 116:852–854. [PubMed: 20688967]
25. Bunn HF. Pathogenesis and treatment of sickle cell disease. N. Engl. J. Med. 1997; 337:762–769.
[PubMed: 9287233]
26. Steinberg MH. Sickle cell anemia, the first molecular disease: overview of molecular etiology,
pathophysiology, and therapeutic approaches. ScientificWorldJournal. 2008; 8:1295–1324.
[PubMed: 19112541]
27. Ballas SK, Mohandas N. Sickle red cell microrheology and sickle blood rheology.
Microcirculation. 2004; 11:209–225. [PubMed: 15280093]
28. Hiruma H, Noguchi CT, Uyesaka N, Schechter AN, Rodgers GP. Contributions of sickle
hemoglobin polymer and sickle cell membranes to impaired filterability. Am. J. Hematol. 1995;
268:H2003–H2008.
29. Chien S, Kaperonis AA, King RG, Lipowsky HH, Schmalzer EA, Sung LA, Sung KL, Usami S.
Rheology of sickle cells and its role in microcirculatory dynamics. Prog. Clin. Biol. Res. 1987;
240:151–165. [PubMed: 3615484]
30. Nash G, Johnson C, Meiselman H. Rheologic impairment of sickle RBCs induced by repetitive
cycles of deoxygenation-reoxygenation. Blood. 1988; 72:539–545. [PubMed: 3401593]
31. Eaton WA, Hofrichter J. Hemoglobin S gelation and sickle cell disease. Blood. 1987; 70:1245–
1266. [PubMed: 3311198]
32. McDonagh KT, Dover GJ, Donahue RE, Nathan DG, Agricola B, Byrne E, Nienhuis AW.
Hydroxyureainduced HbF production in anemic primates: augmentation by erythropoietin,
hematopoietic growth factors, and sodium butyrate. Exp. Hematol. 1992; 20:1156–1164.
[PubMed: 1385194]
33. Embury SH. The not-so-simple process of sickle cell vasoocclusion. Microcirculation. 2004;
11:101–113. [PubMed: 15280086]
34. Adams R, McKie V, Nichols F, Carl E, Zhang DL, McKie K, Figueroa R, Litaker M, Thompson
W, Hess D. The use of transcranial ultrasonography to predict stroke in sickle cell disease. N.
Engl. J. Med. 1992; 326:605–610. [PubMed: 1734251]
35. Adams RJ, McKie VC, Hsu L, Files B, Vichinsky E, Pegelow C, Abboud M, Gallagher D, Kutlar
A, Nichols FT, Bonds DR, Brambilla D, Woods G, Olivieri N, Driscoll C, Miller S, Wang W,
Hurlett A, Scher C, Berman B, Carl E, Jones AM, Roach ES, Wright E, Zimmerman RA,
Waclawiw M. Prevention of a first stroke by transfusions in children with sickle cell anemia and
abnormal results on transcranial doppler ultrasonography. N. Engl. J. Med. 1998; 339:5–11.
[PubMed: 9647873]
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
Wood et al.
Page 11
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript
36. Adams RJ. D. Brambilla & Optimizing Primary Stoke Prevention in Sickle Cell Anemia (STOP 2)
Trial Investigators, Discontinuing prophylactic transfusions used to prevent stroke in sickle cell
disease. N. Engl. J. Med. 2005; 353:2769–2778. [PubMed: 16382063]
37. Barabino GA, Platt MO, Kaul DK. Sickle cell biomechanics. Annu. Rev. Biomed. Eng. 2010;
12:345–367. [PubMed: 20455701]
38. Higgins JM, Eddington DT, Bhatia SN, Mahadevan L. Sickle cell vasoocclusion and rescue in a
microfluidic device. Proc. Natl. Acad. Sci. U.S.A. 2007; 104:20496–20500. [PubMed: 18077341]
39. Higgins JM, Eddington DT, Bhatia SN, Mahadevan L. Statistical dynamics of flowing red blood
cells by morphological image processing. PLoS Comput. Biol. 2009; 5:e1000288. [PubMed:
19214200]
40. Zhujun Z, Seitz WR. Optical sensor for oxygen based on immobilized hemoglobin. Anal. Chem.
1986; 58:220–222.
41. Kundu, PK.; Cohen, IM. Fluid Mechanics. third edition. Elsevier; New York: 2004. p. 279-285.
42. Lipowsky HH. Microvascular rheology and hemodynamics. Microcirculation. 2005; 12:5–15.
[PubMed: 15804970]
43. Zweig MH, Campbell G. Receiver-operating characteristic (ROC) plots: a fundamental evaluation
tool in clinical medicine. Clin. Chem. 1993; 39:561–577. [PubMed: 8472349]
44. Akinsheye I, Alsultan A, Solovieff N, Ngo D, Baldwin CT, Sebastiani P, Chui DHK, Steinberg
MH. Fetal hemoglobin in sickle cell anemia. Blood. 2011
45. Verduzco LA, Nathan DG. Sickle cell disease and stroke. Blood. 2009; 114:5117–5125. [PubMed:
19797523]
46. Abdulmalik O, Safo MK, Chen Q, Yang J, Brugnara C, Ohene-Frempong K, Abraham DJ,
Asakura T. 5-hydroxymethyl-2-furfural modifies intracellular sickle haemoglobin and inhibits
sickling of red blood cells. Brit. J. Haematol. 2005; 128:552–561. [PubMed: 15686467]
47. Haun JB, Castro CM, Wang R, Peterson VM, Marinelli BS, Lee H, Weissleder R. Micro-NMR for
rapid molecular analysis of human tumor samples. Sci. Transl. Med. 2011; 3:71ra16.
48. Vaidya VS, Ozer JS, Dieterle F, Collings FB, Ramirez V, Troth S, Muniappa N, Thudium D,
Gerhold D, Holder DJ, Bobadilla NA, Marrer E, Perentes E, Cordier A, Vonderscher J, Maurer G,
Goering PL, Sistare FD, Bonventre JV. Kidney injury molecule-1 outperforms traditional
biomarkers of kidney injury in preclinical biomarker qualification studies. Nat. Biotechnol. 2010;
28:478–485. [PubMed: 20458318]
49. van den Tweel XW, van der Lee JH, Heijboer H, Peters M, Fijnvandraat K. Development and
validation of a pediatric severity index for sickle cell patients. Am. J. Hematol. 2010; 85:746–751.
[PubMed: 20806231]
50. Klatt, EC.; Kumar, V. Robbins and Cotran Review of Pathology. third edition. Saunders/Elsevier;
New York: 2009.
51. de Montalembert M. Hydroxyurea treatment in patients affected with sickle cell anemia: efficacy
and safety. Transfus. Clin. Biol. 2008; 15:34–38. [PubMed: 18499496]
52. Serjeant GR, Serjeant BE, Mason KP, Hambleton IR, Fisher C, Higgs DR. The changing face of
homozygous sickle cell disease: 102 patients over 60 years. Int. J. Lab. Hematol. 2009; 31:585–
596. [PubMed: 18644042]
53. Nathan DG. Search for improved therapy of sickle cell anemia. J. Pediatr. Hematol./Oncol. 2002;
24:700–703.
54. Noblin X, Mahadevan L, Coomaraswamy IA, Weitz DA, Holbrook NM, Zwieniecki MA. Optimal
vein density in artificial and real leaves. Proc. Natl. Acad. Sci. U.S.A. 2008; 105:9140–9144.
[PubMed: 18599446]
55. Klimant I, Wolfbeis OS. Oxygen-sensitive luminescent materials based on silicone-soluble
ruthenium diimine complexes. Anal. Chem. 1995; 67:3160–3166.
56. Hollander, M.; Wolfe, DA. Nonparametric Statistical Methods. John Wiley & Sons; Hoboken, NJ:
1999.
57. Hesterberg, T.; Moore, DS.; Monaghan, S.; Clipson, A.; Epstein, R., editors. Introduction to the
Practice of Statistics. fifth edition. W. H. Freeman & Co.; New York: 2005. chap. 14: Bootstrap
Methods and Permutation Tests.
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
Wood et al.
Page 12
NIH-PA Author Manuscript
NIH-PA Author Manuscript
Figure 1. Measuring rate of conductance decrease after deoxygenation
NIH-PA Author Manuscript
Time course of measurement showing oxygen concentration (a) in the gas reservoir and
corresponding median velocity (b, left y-scale and circles) of RBCs by video tracking as
well as applied pressure bias (right y-scale and dot-dash line). Corresponding channel
conductance (c) is measured by dividing instantaneous velocity by the pressure. We see that
the rate of change of the conductance as the blood flow stops is effectively constant and
independent of both the velocity and pressure. All conductance values are scaled by the
mean HbA blood conductance (C*HbA ~ 1.7 µm/s/mmHg (0.013 µm/s/Pa)), and time is
scaled by the time scale for hemoglobin deoxygenation (τdeox ~ 10s).
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
Wood et al.
Page 13
NIH-PA Author Manuscript
NIH-PA Author Manuscript
Figure 2. Rate of conductance decrease correlates with patient clinical course
NIH-PA Author Manuscript
(a) Box plot comparing rates of conductance decrease for benign and severe samples.
Receiver operating characteristics (b) for rate of conductance decrease (solid line) and a
theoretical random prediction (dashed line). The areas under the ROC curves are 0.85 and
0.5 respectively. (c) Box plot comparing HbS fraction for benign and severe samples. HbF
fractions (d) and rates of conductance change (e) for benign and severe samples further
subdivided by hydroxyurea use. Rates of conductance decrease (f) for patient blood samples
before and after addition of ABOand RhD-compatible HbA blood, simulating a blood
transfusion. For each sample, data is normalized to unmodified sample. *p < 0.01 as
determined by Mann-Whitney non-parametric analysis [56]. Bar heights in (f) represent
means, and error bars represent standard deviations of at least 5 deoxygenation cycles. In
box plots (a,b-e) red line is the median, blue box shows interquartile range (IQR), and
dashed lines show extent of data within 1.5 times IQR. Rates of conductance decrease are
scaled by mean HbA blood conductance (C*HbA ~ 1.7 µm/s/mmHg (0.013 µm/s/Pa))
divided by the time scale for hemoglobin deoxygenation (τdeox ~ 10s).
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
Wood et al.
Page 14
NIH-PA Author Manuscript
Figure 3. Rate of conductance decrease is modulated by a small molecule
NIH-PA Author Manuscript
Oxygen data as measured in the gas reservoir (top graph) and conductance data (bottom
graph) are shown for an untreated severe sample (a) and the same sample treated (b) with
10mM 5-hydroxymethyl furfural (5HMF). Oxygen data (top graph) are shown as measured
in the gas reservoir (dashed line) and in the blood channel (open circles, as measured by
RBC intensity). Rates of conductance decrease (open circles) are quantified in (c). *p<0.05
as determined by Mann-Whitney non-parametric analysis [56]. Bar heights in (c) represent
means, and error bars represent standard deviations of at least 5 independent deoxygenation
cycles. All conductance values are scaled by the mean HbA blood conductance (C*HbA ~
1.7 µm/s/mmHg (0.013 µm/s/Pa)), and time is scaled by the time scale for hemoglobin
deoxygenation (τdeox ~ 10s). Rates of conductance decrease are scaled by mean HbA blood
conductance divided by the time scale for hemoglobin deoxygenation (C*HbA /τdeox).
NIH-PA Author Manuscript
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
Wood et al.
Page 15
NIH-PA Author Manuscript
NIH-PA Author Manuscript
Scheme 1. Microfluidic device for studying sickle cell blood flow conductance
NIH-PA Author Manuscript
The device comprises 3 layers (inset): artificial capillary for blood flow, hydration layer
with PBS, and gas reservoir. Blood is flowed under constant pressure bias, controlled by a
digital pressure regulator. Two solenoid valves control the gas in the top chamber. A fiber
optic probe is used to measure the oxygen concentration in the gas reservoir. The device is
illuminated through an optical filter whose transmission band (434+/−17 nm) is centered on
an absorption peak for deoxy-hemoglobin (Fig. S1). The absorption peak of oxy-hemoglobin
shifts making deoxygenated RBCs appear dark and oxygenated RBCs transparent.
Qualitative measurements of hemoglobin oxygen saturation are made using the intensity of
transmitted light.
Sci Transl Med. Author manuscript; available in PMC 2013 April 23.
Download