Sickle cell disease biochip: a functional red blood cell adhesion assay for monitoring sickle cell disease

Abstract

Sickle cell disease (SCD) afflicts millions of people worldwide and is associated with considerable morbidity and mortality. Chronic and acute vaso-occlusion are the clinical hallmarks of SCD and can result in pain crisis, widespread organ damage, and early movtality. Even though the molecular underpinnings of SCD were identified more than 60 years ago, there are no molecular or biophysical markers of disease severity that are feasibly measured in the clinic. Abnormal cellular adhesion to vascular endothelium is at the root of vaso-occlusion. However, cellular adhesion is not currently evaluated clinically. Here, we present a clinically applicable microfluidic device (SCD biochip) that allows serial quantitative evaluation of red blood cell (RBC) adhesion to endothelium-associated protein-immobilized microchannels, in a closed and preprocessing-free system. With the SCD biochip, we have analyzed blood samples from more than 100 subjects and have shown associations between the measured RBC adhesion to endothelium-associated proteins (fibronectin and laminin) and individual RBC characteristics, including hemoglobin content, fetal hemoglobin concentration, plasma lactate dehydrogenase level, and reticulocyte count. The SCD biochip is a functional adhesion assay, reflecting quantitative evaluation of RBC adhesion, which could be used at baseline, during crises, relative to various long-term complications, and before and after therapeutic interventions.

INTRODUCTION

In mammals, the red blood cell (RBC) has uniquely evolved to lose its nucleus and organelles to become remarkably flexible.¹ RBC's adherence to vascular wall and other cells is insignificant,² whereas most other cell types depend on adhesive interactions to survive.³ RBC repeatedly deforms and squeezes through narrow capillaries that can be as small as half of its diameter.^2,4,5 Its characteristic shape and exceptional mechanical deformability are determined by its membrane skeleton, which is located underneath the cell membrane and linked to adhesion receptors on the cell surface.^6-9 RBC's reduced deformability and increased adhesion have been associated with microcirculatory impairment in many diseases, including hemoglobin disorders,^10-14 sepsis,^15,16 malaria,^17-20 lupus,^21,22 heavy metal exposure,^23,24 blood transfusion complications,^25,26 diabetes,^27,28 cancer,^29,30 kidney diseases,^31-35 cardiovascular diseases,^36,37 obesity,^38,39 and neurologic disorders.^40-43 These diseases affect hundreds of millions of people globally with a socioeconomic burden of hundreds of billions of dollars annually.^44-51

In sickle cell disease (SCD), RBC adhesion has been associated with blood flow blockage,^52,53 disease severity,^10-13 and organ damage.⁵⁴ SCD arises from a point mutation in the β-globin gene resulting in production of hemoglobin S (HbS). Intracellular HbS molecules polymerize on deoxygenation, forming long fibers that lead to membrane damage and abnormal cellular stiffness. Membrane damage caused by HbS polymerization increases sickle RBC adhesion to vascular endothelium.^55,56 In addition, increased RBC stiffness impacts blood flow and, with abnormal cellular adhesion, results in blockage of blood vessels (vaso-occlusion).^52,53

SCD affects millions worldwide⁵⁷ and imposes significant physical, emotional, and financial burdens on its sufferers, their families and communities. Chronic and acute vaso-occlusion are the clinical hallmarks of SCD and can result in painful crises, cumulative organ damage, and early mortality.⁵⁶ SCD can cost >$8 million per patient over a 50-year life span (in the United States).⁵⁸ Even though abnormal RBC adhesion is the centerpiece of vaso-occlusion and vascular damage in SCD, there is no clinically relevant tool or method to evaluate cellular adhesion as a clinical biomarker for disease severity. Lack of such clinically applicable assays has slowed the development of new pharmaceutical and therapeutic approaches because there is no in vitro test for measuring the effects of these interventions on RBC adhesion. To address this clinically unmet need, we developed a versatile microfluidic platform for evaluation of RBC adhesion in whole blood samples. SCD may be an ideal disease with which to interrogate cellular adhesion as an indicator of disease activity and severity because point-of-care and real-time markers of disease activity are urgently needed.

In the 1980s, abnormal RBC adhesion in SCD was studied using flow chambers or ex vivo rat mesocecum.^52,59-61 However, cellular adhesion is not currently evaluated clinically because analyses of these cellular interactions are technically challenging and difficult to reproduce. Recently, microfluidic technologies have emerged as versatile platforms for diagnosing and monitoring diseases.^62-64 These devices allow simple and cost-efficient fabrication, short processing times, and minimal reagent use. Furthermore, microfluidic systems can be easily adapted to mimic biophysical microenvironment of cells for a more comprehensive and accurate analysis of the subject's pathophysiological state.^65-68 Microfluidic platforms have been used in diagnosis and/or monitoring of several life-threatening diseases such as cancer,^67,69,70 human immunodeficiency virus,^71,72 and thrombosis.^73,74

Here, we present a clinically relevant microfluidic device (SCD biochip) that allows serial quantitative measurements of RBC adhesion to endothelium-associated protein-immobilized microchannels (fibronectin [FN] or laminin [LN]) in a closed and preprocessing-free system. RBC deformability, when adhered, can also be evaluated using this technology. FN is a glycoprotein that circulates in plasma and is present in the endothelial cell membrane.^75,76 FN plays a role in SCD RBC adhesion, via RBC integrin α4β1 (also known as very late antigen-4 or VLA-4 integrin) interaction^75,77,78 with the endothelial wall.^75,77,78 LN is subendothelial and binds to an important RBC surface protein from the immunoglobulin superfamily, BCAM/LU (basal cell adhesion molecule, Lutheran antigen),^59,79-82 which is phosphorylated during beta-adrenergic stimulation.^59,80-82 We have analyzed more than 100 subject blood samples using the SCD biochip, showing significant associations between RBC adhesion on FN and LN with hemoglobin (Hb) phenotype (HbSS, HbSC, and HbAA), fetal hemoglobin (HbF) levels; lactate dehydrogenase (LDH); platelet and reticulocyte counts. The SCD biochip allowed us to analyze RBC adhesion in a sizable, adult, well-phenotyped, SCD population.

METHODS

Study design

All attainment of samples and clinical information, such as laboratory tests, treatment courses, and medical history, was performed with informed consent of the subjects and under the institutional review board approval. It is standard of care for patients to get complete blood counts (CBCs) during routine (baseline) clinic visits. Blood samples from patients with HbSS, HbSC, HbSC/Sβ⁺, and healthy controls (Hb AA) were de-identified and collected during the regular course of the Adult Sickle Cell Clinic at the University Hospitals Case Medical Center in Cleveland. Surplus blood samples, drawn in an ethylenediaminetetraacetic acid (EDTA)–containing Vacutainer (“purple top”) tube, were obtained. A volume of 0.25-mL whole blood is used in the SCD biochip. All samples were run within 24 hours of acquisition. Collected data included Hb phenotype, CBC (including white blood cell [WBC, 10⁹/L], red blood cell [10¹²/L], absolute neutrophil count [10⁶/L], platelet count [10⁹/L], and reticulocyte count [10⁹/L]), plasma LDH (U/L), and pain level (by FACES or numeric pain rating scale, between 0 [no pain] and 10 [worst possible pain]). Sample attainment relative to any treatments, such as hydroxyurea use or blood transfusions, was noted. Hb identification of Hb S, Hb F, Hb A, and Hb A2 was conducted via high-performance liquid chromatography (HPLC) at the Core Laboratory of University Hospitals Case Medical Center, using the Bio-Rad Variant II Instrument (Bio-Rad, Montreal, QC, Canada). Clinical laboratory values for subjects’ blood samples are listed in Table I. Adults who were documented with SCD, including HbSS or compound heterozygous HbSC- or HbSβ-thalassemia diagnosis as evidenced by one or more clinical features, were assessed for eligibility (Fig S1). Written informed consent was obtained from the subjects who were able to comply with the requirements of the study. Presence of a condition or abnormality that, in the opinion of the investigator, would compromise the safety of the patient or researcher performing the experiment (such as infection) was regarded as an exclusion criterion. Control samples (HbAA) were obtained from Research Blood Components (Boston, MA) or from healthy volunteers. We recorded pain scores of patients in our study but did not rigorously assess temporal proximity to painful crises. Furthermore, any acutely ill patient was sent from our clinic to the day hospital for treatment, and those samples were not available for processing. We estimate that 5%–10% of patient visits were handled in this way. A subset of early samples was obtained anonymously without correlative clinical data beyond hemoglobin phenotype.

Table I.

Clinical phenotype of the consented SCD patient population

	Mean ± SE	Range
Age	37.6 ± 2.5	19–72
WBC (109/L)	9.7 ± 0.6	4.1–22.6
Reticulocyte count (109/L)	324.1 ± 29.0	89–798
Hemoglobin (g/dL)	8.8 ± 0.3	5.5–12
Platelet count (109/L)	361.1 ± 22.7	155–703
Absolute neutrophil count (106/L)	4899.1 ± 459.7	720–14,240
Lactate dehydrogenase (U/L)	408.3 ± 32.7	157–934
Pain*	3.4 ± 0.5	0–10
Hemoglobin F (%)	9.8 ± 1.3	0.4–23.9
Hemoglobin S (%)	72.3 ± 3.2	17.6–91.1
Hemoglobin A (%)	10.9 ± 3.2	0–66.1

Abbreviations: SE, Standard error; WBC, white blood cell.

Data presented is based on 32 blood samples (n) from 27 consented SCD patients (male = 11, female = 16).

^*“n” for pain score was 31.

SCD biochip device fabrication

SCD biochip is designed and developed to mimic the physiological conditions of the microvasculature. The microfluidic devices are designed and fabricated to be single use to prevent cross-contamination between samples. Microfluidic channels are composed of a glass surface functionalized with FN or LN, a poly(methyl methacrylate) plastic top (encompassing inlets and outlets), and sandwiched 50-μm-thick double-sided adhesive tape that defined the height and shape of the microchannels (Fig S2). RBC adhesion to FN or LN was measured in SCD biochip, without preprocessing, at or above physiological flow shear stresses of postcapillary venules (1–5 dyne/cm²),^53,83-85 ranging from 1 to 50 dyne/cm². Briefly, SCD biochip was fabricated by assembling poly(methyl methacrylate) top parts, cut by VersaLaser system (Universal Laser Systems Inc., Scottsdale, Ariz), and double-sided adhesive film (iTapestore, Scotch Plains, NJ) which have an inlet and outlet (0.61 mm in diameter and 26 mm apart).⁶⁸ They were placed onto Gold Seal glass slide (adhesion coating: APTES, 3-Aminopropyl Triethoxysilane, Electron Microscopy Sciences [Hatfield, Pa]). Surface chemistry inside microchannels was achieved by flushing the channels with N-g-Maleimidobutyryloxy succinimide ester (0.28% vol/vol) followed by incubation in FN (fibronectin from human plasma, 1:10; Sigma Aldrich, St Louis, Mo) or LN (laminin from Engelbreth-Holm-Swarm murine sarcoma basement membrane, LN 1, 1:10; Sigma Aldrich) for 1.5 hours at room temperature. Then, channels were incubated in bovine serum albumin solution (20 mg/mL) overnight at 4°C for blocking nonspecific binding events. After assembly of inlet and outlet tubing, undiluted whole blood samples from subjects were injected on to FN-or LN-immobilized microchannels at 18.5 μL/min to 25.8 μL/min to fill channels, then 1.85 μL/min (1 dyne/cm²) to 2.58 μL/min (1.4 dyne/cm²) until 15 μL of the sample was introduced. After the blood flow, 180 μL of flow cytometry staining buffer (FCSB, 1X, with albumin 0.5%–1%, without fluoresceinated antibodies) was run at 10 μL/min (1 dyne/cm²) for 18 minutes in the channels, to remove non-adhered RBCs. All flow steps were conducted using New Era NE-300 syringe pump system (Farmingdale, NY). An inverted microscope (Olympus IX83) and microscopy camera (EXi Blue EXI-BLU-R-F-M-14-C) was used to obtain high-resolution images of whole channels. All experiments were conducted at room temperature.

Image processing and quantification

Whole phase-contrast images of microchannels with adhered RBCs are obtained using Olympus (20x/0.45 ph2 and 40x/0.75 ph3) long working distance objective lenses. Images were then processed using Adobe Photoshop software (San Jose, CA) for quantification of adhered RBCs per unit area (32 mm²). Here, we categorized RBCs based on morphology as deformable RBCs with the characteristic biconcave shape and nondeformable RBCs lacking biconcave morphology, based on earlier work.⁶⁶

Statistical analysis

The relationship between individual components of the CBC and serum tests and the number of adhered RBCs was analyzed using K-means clustering. The patients with HbSS were clustered into 2 groups, and the resultant groups were evaluated for differences in adhesion. Single and multiple components of the CBC and serum analyses were used to identify the 2 subgroups. Once the subgroups were identified by K-means clustering, the difference between the numbers of adhered RBCs between these groups was tested for statistical significance using the one-way analysis of variance (ANOVA) test. The testing level (alpha) was set as 0.05 (two-sided). The component or components that lead to the significant subgroups in terms of the differences between the numbers of adhered RBCs were reported in this article. The K-means clustering was performed using Matlab (The MathWorks, Inc, Natick, Mass). Receiver-operating curves were used to determine the SCD biochip's accuracy of differentiation between hemoglobin phenotypes. The curves were generated using Matlab (The MathWorks, Inc). In addition to the area under the curve, sensitivity, specificity, positive and negative likelihood ratios, and positive and negative predictive values were calculated as follows: Sensitivity was calculated as # true positives/(# true positives + # false negatives), specificity as # true negatives/(# true negatives + # false positives). Positive likelihood ratio was defined as sensitivity/(1-specificity). Negative likelihood ratio was (1-sensitivity)/specificity. Positive predictive value was # true positives/(# true positives + # false positives), and negative predictive value was # true negatives/(# true negatives + # false negatives).

RESULTS

Probing abnormal RBC adhesion in a microfluidic channel

The SCD biochip affords quantitative analysis of RBCs with abnormal membrane and adhesive properties following attachment to endothelium-associated proteins (Fig 1, A). The number of adhered RBCs is quantified inside the FN- or LN-immobilized microfluidic channels. We observed abnormal adhesion of RBCs to the SCD biochip in blood samples from subjects with SCD (Fig 1, B). In contrast, adhesion of RBCs in blood samples from normal subjects was negligible (not shown). Different levels of RBC adhesion were observed in patients with various clinical phenotypes (Fig 1, C and D).

Fig 1.

Sickle cell disease biochip (SCD biochip) probes red blood cell (RBC) adhesion in a closed system using minuscule amounts of whole blood samples. (A) Flow and RBC cell adhesion are illustrated inside the SCD biochip. Shown are varying levels of sickling and adherence to endothelium-associated protein-coated microchannel surfaces. (Inset) SCD biochip consists of multiple parallel microchannels. Scale bar represents a length of 10 mm. (B) Number of adhered RBCs are quantified inside microfluidic channels that are functionalized with FN or LN. We observed abnormal RBC adhesion in blood samples from SCD subjects. Adhered RBCs are marked with red dots. (C and D) High-resolution phase-contrast images of adhered RBCs with heterogeneous sickle morphologies inside the SCD biochip are shown. Different levels of RBC adhesion were observed in blood samples from patients with various clinical phenotypes, such as high or low hemoglobin F (HbF) levels. FN, Fibronectin; LN, laminin; RBC, red blood cell; SCD, sickle cell disease.

RBC adhesion varies among hemoglobin phenotypes

We analyzed the number of adhered RBCs per unit area (32 mm²) in FN- and LN-functionalized parallel microchannels (Fig S2) in blood samples from subjects without SCD (HbAA), or with compound heterozygous (HbSC or HbSβ⁺-thalassemia, HbSC/Sβ⁺) or homozygous SCD (HbSS), using high-resolution images from microfluidic channels (Fig 2, A and B). RBC adhesion to FN- or LN-coated microchannels was highest in HbSS relative to HbSC/Sβ⁺, and higher in HbSC/Sβ⁺ relative to HbAA containing blood samples (Fig 2, C and D). Close-up views of adhered RBCs in functionalized microchannels are shown in Fig S3. For selected thresholds of adhered RBC numbers, we observed 0.93 true-positive rate and 0.00 false-positive rate for differentiating between HbSS and HbAA phenotypes (Fig 2, E and F). Area under the curve for differentiating between HbSS and HbAA was >0.85 both for FN and LN (Table II). The ROC were strongest in discriminating between AA and SS or SC, compared with discrimination between SS and SC. These results demonstrate the ability of the SCD biochip to discriminate among hemoglobin phenotypes solely based on cellular adhesion. These data support the role that abnormal cellular adhesion plays in accounting for differences among SCD phenotypes. High levels of HbF are associated with improved clinical outcomes in SCD.⁸⁶ Therefore, we investigated RBC adhesion in HbSS blood samples with low (<%8) and high (>%8) HbF levels (Fig 3), based on estimates of what is physiologically relevant.⁸⁶ The mean number of adhered RBCs was significantly higher in blood samples from HbSS subjects with low HbF compared with those from subjects with high HbF levels, in both FN- (Fig 3, A) and LN-functionalized (Fig 3, B) microchannels. These results demonstrate that SCD biochip is a plausible in vitro adhesion assay for functional phenotypes of SCD.

Fig 2.

Adhesion of RBCs in FN- and LN-functionalized microchannels varies among SCD hemoglobin pheno-types and are greatest in HbSS. Shown are high resolution images of microchannels in (A) FN or (B) LN. The number of adhered RBCs was significantly higher in samples from subjects with HbSS > HbSC/Sβ⁺> HbAA in both (C) FN and (D) LN immobilized microchannels. The horizontal lines between individual groups represent a statistically significant difference based on a one-way ANOVA test (P < 0.05). Data point cross bars represent the mean. “N” represents the number of subjects. (E and F) Receiver operating-characteristic (ROC) curves display a true-positive rate (sensitivity) and a false-positive rate (1-specificity) for differentiation between SS-AA, SC-AA, and SS-SC hemoglobin phenotypes based on adhesion of RBCs to (E) FN and (F) LN. Defined thresholds for adhered RBC numbers on the ROC are as shown (◇ = 9 (SS-AA), □ = 9 (SC-AA), and ○ = 30 (SS-SC) for FN; ◇ = 16 (SS-AA), □ = 16 (SC-AA), and ○ = 170 (SS-SC) for LN). ANOVA, Analysis of variance; FN, fibronectin; LN, laminin; RBC, red blood cell; SCD, sickle cell disease.

Table II.

Ability of SCD biochip in differentiating between hemoglobin phenotypes based on RBC adhesion to FN and LN microchannels

	FN			LN
	SS-AA	SC-AA	SS-SC	SS-AA	SC-AA	SS-SC
Threshold	9	9	30	16	16	170
Sensitivity	0.93	0.80	0.75	0.93	0.89	0.87
Specificity	1.00	1.00	0.70	1.00	1.00	0.78
PLR	Inf	Inf	2.50	Inf	Inf	3.90
NLR	0.07	0.20	0.36	0.07	0.11	0.17
PPV	1.00	1.00	0.88	1.00	1.00	0.87
NPV	0.83	0.83	0.50	0.89	0.89	0.78
Accuracy	0.95	0.90	0.74	0.96	0.94	0.83
AUC	0.88	0.88	0.77	0.86	0.99	0.89

Abbreviations: AUC, Area under the receiver operating characteristic curve; FN, fibronectin; LN, laminin; PLR, positive likelihood ratio; PPV, positive predictive value; NLR, negative likelihood ratio; NPV, negative predictive value.

Threshold unit is number of adhered RBCs per unit area.

Fig 3.

Adhesion of RBCs is greater in HbSS subjects with a low (<%8) HbF, compared with high (>%8) HbF. (A and B) RBC adhesion was quantified in blood samples of HbSS patients with high and low HbF levels, from temporally nearest clinical measurement (not uniformly contemporaneous). Number of adhered RBCs was significantly higher in blood samples from subjects with low HbF levels compared with blood samples from subjects with high HbF levels in both (A) FN and (B) LN immobilized microchannels. The horizontal lines between individual groups represent a statistically significant difference based on a one-way ANOVA test (P < 0.05). Data point cross bars represent the mean. “N” represents the number of subjects. ANOVA, Analysis of variance; FN, fibronectin; HbF, fetal hemoglobin; LN, laminin; RBC, red blood cell.

Clinical correlates for RBC adhesion to FN and LN

We used K-means clustering analysis to identify subgroups of HbSS patients based on individual components of standard blood tests. We analyzed HbSS RBC adhesion to FN microchannels in blood samples from patients with dichotomized laboratory values, including LDH levels and percent HbS. The mean adherence was directly associated with LDH levels (FN, Fig 4, A). K-means clustering analyses were then applied to determine univariate models with greater significance in correlation with RBC adhesion to FN. Levels of HbS varied due to recent transfusions (HbA ≥10%) or due to increased levels of HbF. High levels of HbS were associated with increased LDH levels (group 1 relative to group 2, Fig 4, B). We observed that patients with high HbF tended to have low serum LDH levels (14/16 patients with >%8 HbF have <500 U/L LDH; Fig S4). Adherence to FN was significantly greater in blood samples with higher LDH and higher HbS levels (group 1 compared with group 2, Fig 4, C). Most transfused subjects (7/9) were in the lower LDH subgroup (LDH <500 U/L, Fig S5). In addition, subjects with a lower platelet count (<320 10⁹/L) consistently had fewer than 400 adherent RBCs per unit area of FN (Fig S6). These methods were also applied to determine correlations with RBC adhesion to LN (Fig 5). Adherence to LN was significantly greater in HbSS blood samples with higher LDH (Fig 5, A) or higher reticulocyte counts (Fig 5, B). High levels of LDH were associated with elevated reticulocyte counts (group 1 relative to group 2, Fig 5, C). In univariate models, we observed significantly greater adhesion to LN in HbSS blood samples with higher LDH and higher reticulocyte levels (group 1 compared with group 2, Fig 5, D), independent of HbS (Fig S5, C and D). Moreover, we analyzed correlation between RBC adhesion to FN and LN, and reticulocyte counts for SCD subjects (Fig S7). We found a statistically significant correlation between reticulocyte counts and RBC adhesion to LN (Pearson product-moment correlation coefficient, PCC = 0.53, P = 0.007) and to FN (PCC = 0.46, P = 0.035), and this effect was increased in the absence of transfused subjects for both LN (PCC = 0.72, P < 0.0001) and FN (PCC = 0.66, P = 0.018). Furthermore, we analyzed RBC counts of the patient blood samples used in our study and found an average number of 2.65 ± 0.66 × 10⁹ RBCs/mL (mean ± standard deviation, Fig S8, A), where 21 of 31 blood samples had RBC counts within 1.9–2.9 × 10⁹ RBCs/mL. There was no statistically significant correlation between RBC counts of the blood samples and number of adhered RBCs to either FN or LN (Fig S8, B and C).

Fig 4.

RBC adhesion to FN is associated with lactate dehydrogenase (LDH) and HbS percentage. (A) Number of adhered RBCs in FN microchannels was significantly higher in blood samples with high LDH (>500 U/L). Seven of 9 samples are from subjects who had recently been transfused (Fig S4) and who had low LDH and low adhesion. (B) In blood samples with higher LDH and higher HbS (group 1), determined by k-means clustering analysis, (C) adherence to FN was significantly greater compared with lower LDH and lower HbS (group 2). Samples from recently transfused subjects are shown with triangle markers. The horizontal lines between individual groups represent a statistically significant difference based on a one-way ANOVA test (P < 0.05). Data point cross bars represent the mean. “N” represents the number of subjects. ANOVA, Analysis of variance; FN, fibronectin; RBC, red blood cell.

Fig 5.

RBC adhesion to LN is associated with high LDH and high absolute reticulocyte counts. (A and B) RBCs in blood samples with (A) high LDH (>500 U/L) and (B) higher reticulocyte counts (>320 10⁹/L) showed a significantly higher adherence to LN-immobilized microchannels compared with RBCs from samples with low LDH (<500 U/L) and low reticulocyte counts (<320 10⁹/L), respectively. (C) Blood samples with higher LDH and higher reticulocyte counts (group 1), determined by k-means clustering analysis, (D) showed significantly greater adhesion to LN compared with blood samples with lower LDH and lower reticulocyte counts (group 2). The horizontal lines between individual groups represent a statistically significant difference based on a one-way ANOVA test (P < 0.05). Data point cross bars represent the mean. “N” represents the number of subjects. ANOVA, Analysis of variance; LDH, lactate dehydrogenase; LN, laminin; RBC, red blood cell.

These results show clinical associations between RBC adhesion to FN or LN with LDH, platelet counts, or reticulocyte counts. Nonetheless, heterogeneity was present in all analyses, for example, low adhesion in some patients with a high LDH or elevated reticulocyte count (Figs 4 and 5); this was also seen in a subset of chronically transfused patients; the converse was less ambiguous, for example, low adhesion was present in most patients with a low LDH (Fig 4).

Identifying subpopulations of adherent RBCs

The morphology and number of adherent RBCs in HbSS blood samples were examined after controlled detachment of cells at step-wise increased flow shear stresses of 1, 4, and 50 dyne/cm² (Fig 6, A–C). Based on morphologic characterization, adhered RBCs were categorized as deformable (Fig 6, D) or nondeformable (Fig 6, E), based on our earlier work.⁶⁶ The percent of deformable and nondeformable RBCs, relative to total adhered RBCs at 1 dyne/cm² shear stress, was calculated in each experiment, and the mean and standard error were determined (Fig 6, F). These analyses were repeated at higher shear stresses of 4 and 50 dyne/cm². The total number of adhered RBCs decreased with higher shear stresses, whereas the proportion of nondeformable RBCs increased in these same experiments (Fig 6, F). The percent of nondeformable RBCs, determined at 1 dyne/cm², correlated significantly with serum LDH levels in test subjects (Fig 6, G; PCC = 0.74; P < 0.0001). These results indicate a morphologic heterogeneity in adhered RBCs and suggest an association between LDH (an indicator of hemolysis) and adherent nondeformable RBCs. A detailed description of RBC adhesion at differing flow rates is shown in Fig S9, with close-up views of adhesion to FN and LN, respectively, in Figs S10 and S11.

Fig 6.

Heterogeneity in adhered RBCs in FN-functionalized microchannels is associated with serum LDH levels. (A–C) Number and morphology of adhered RBCs were analyzed in HbSS blood samples at step-wise increased flow shear stresses; (A) 1 dyne/cm², (B) 4 dyne/cm², and (C) 50 dyne/cm². (D) Deformable and (E) nondeformable RBCs were determined morphologically. Scale bars represent a length of 5 μm. (F) Nondeformable RBCs (% of total adhered RBCs) at 1, 4, and 50 dyne/cm² flow shear stress were calculated (columns). Total number of RBCs at each flow velocity is shown. The horizontal lines between individual groups represent a statistically significant difference based on a one-way ANOVA test (P < 0.05). Error bars represent the standard error of the mean. “N” represents the number of subjects. (G) Shown is adhered nondeformable RBCs (% of total) and serum LDH (U/L) at 1 dyne/cm² (Pearson correlation coefficient of 0.74, P < 0.0001, N = 21). ANOVA, Analysis of variance; FN, fibronectin; LDH, lactate dehydrogenase; RBC, red blood cell.

Longitudinal analysis of RBC adhesion

We conducted longitudinal analyses on a limited number of individual subjects, performed ≥1 month apart using the same biomolecule, FN or LN (Fig S12). Stable SCD RBC adhesion to FN was seen in 2 patients being treated for SCD, including unique patient number (UPN) 21 (Fig S12, A), examined when HbA levels were stable on transfusions, and UPN 118, examined when HbF levels were stably elevated on hydroxyurea (Fig S12, A). Noted, but of unknown significance, since in 2 patients, is the overall lower adhesion seen in the subject on hydroxyurea, given the pancellular effects of hydroxyurea compared with transfusions. Adhesion to FN dropped after each transfusion in UPN 67, monitored over 6 months, after 2 episodes of transfusion (Fig S12, B). Adhesion to LN was stable in UPN 20, being managed with supportive care, and in UPN 35, stably on hydroxyurea (Fig S12, C). Adhesion to LN in a single patient dropped by >80% from baseline after initiation of hydroxyurea, as HbF rose from 1% to 20% (Fig S12, D).

DISCUSSION

The sickle hemoglobin mutation afflicts many millions of people worldwide and is associated with considerable morbidity and mortality.⁸⁶ Even though the molecular underpinnings of SCD were recognized more than 60 years ago,⁸⁷ there is no clear-cut molecular or biophysical markers through which to evaluate disease severity in the clinic. Biophysical phenomena, such as cell adhesion and deformability, reflect the multi-scale dimensions of SCD including molecular alterations, perturbed regulatory pathways, and cellular abnormalities. These are plausible biomarkers of dis ease activity and progression. RBC adhesion has been shown to be a critical biophysical factor involved in vaso-occlusion^52,53 and to correlate with disease severity.^10-12 However, studies of RBC adhesion in SCD have been primarily proof-of-concept rather than clinically applicable, due to requirements for complicated custom-designed systems, highly trained personnel, specially obtained blood samples, and extensive sample manipulation.^52,59-61 These significant technological barriers have hindered the widespread evaluation of RBC adhesion clinically or as a research tool. This has restricted our ability to integrate complex RBC adhesion phenomena into our understanding of SCD.

Here, we introduce a novel microfluidic biochip technology, SCD biochip, which provides an easy-to-use platform for probing RBC adhesion quantitatively in clinically available blood samples. The SCD biochip allows rapid, fully enclosed, and preprocessing-free analysis of blood samples, which affords simultaneous interrogation of both cellular behavior and plasma characteristics.⁶⁶ Cumulative membrane damage and aberrant activation of multiple adhesion receptors are acquired over the life of a HbS-containing RBC and cause abnormal adhesion. This integrates pathophysiological processes, that are both intracellular (eg, hemoglobin concentration, composition,^88-90 and polymerization^55,91), and extracellular/plasma (eg, cytokines^92,93 or beta adrenergic stimulation^94,95). The SCD biochip measures RBC adhesion quantitatively in patients’ whole blood samples without any preprocessing.

Blood flow in microcirculation is dependent on the intrinsic RBC biomechanical properties, namely deformability^96,97 and adhesiveness.^11,35,94 RBC deformability and adhesion are compromised in many disease states, resulting in blockages of blood flow in microcapillary networks. Abnormal RBC adhesion to endothelium has been implicated in other multi-system diseases, such as β-thalassemia, diabetes mellitus, hereditary spherocytosis, polycythemia vera, and malaria.^13,98,99 The precise contribution of different cell types to initiation and propagation of the vaso-occlusive crisis, which is pathophysiologically central to SCD, is unknown. We observed adhesion of different RBC subpopulations to FN and LN, reminiscent of RBC subpopulations defined in the literature for vaso-occlusion.¹⁰⁰ Furthermore, although deformable sickle RBCs adhere in higher numbers (especially to LN), nondeformable RBCs display greater adhesion strength, suggesting a critical contribution by both populations to vaso-occlusion.

RBC adhesion in SCD is governed by the complex interactions among RBC receptors, endothelial cell ligands, and plasma factors.^77,101-103 In this study, we functionalized microfluidic channels with FN or LN, which are known to be critical in RBC adhesion to endothelium.⁵⁶ FN is an adhesive glycoprotein that circulates in plasma and is present on the endothelial cell membrane.^78,102 FN is reported to interact with multiple RBC adhesion molecules, including CD44, CD147 (basigin/neurothelin), and α4β1 integrin (VLA-4),¹⁰² and with other plasma proteins, such as thrombospondin.^104,105 Furthermore, a primary role was suggested for FN in RBC adhesion during vascular injury or inflammation.^78,106 In our study, we used FN as an adhesion molecule based on multiple earlier in vitro^60,77,101 and ex vivo animal experiments,^106,107 suggesting a critical role of FN in mediating RBC adhesion.

LN is a subendothelial matrix protein that interacts with RBC BCAM/Lu.^56,108-110 This interaction has been attributed primarily to interactions with the α5 chain of LN subtypes 10/11,^102,111-113 although other investigators have described RBC- or BCAM/Lu-mediated interactions to human placental or non-10/11 LN subtypes,^{59,108,111,114} including LN 1.¹¹⁴ Human placental LN contains a mixture of LN isoforms, including LN 1, 2, 3, 4, and 10, and generally contaminated with FN, entactin, and collagen type IV.^111,115-117 Here, we observed adhesion of sickle RBCs to subtype LN 1 that is associated with clinical phenotype (increased LDH, reticulocytosis). Augmentation of sickle RBC binding in the presence of autologous plasma has been suggested for other RBC interactions,^103,118-120 and this effect may be at play in our system as well.

We used patient blood samples collected in Vacutainer tubes with EDTA. EDTA is the most widely used anticoagulant for blood collection.¹²¹ Thus, we tested our SCD biochip technology with a standard blood collection method that is applicable to most clinical settings. The blood used in our studies, surplus from clinically obtained complete blood cell analyses, is easily attainable, due to the frequent use of this test in the clinical care of patients with SCD. To date, EDTA anti-coagulated blood has not appreciably affected analyses of RBC adhesion to FN or LN. In the future, we plan to use alternate anti-coagulants (eg, sodium citrate or corn trypsin inhibitor) with specially drawn blood samples for detailed pathophysiological analyses of RBC and WBC adhesion.

We showed significant associations between RBC adhesion and individual features of red cell phenotype, including hemoglobin content (HbSS, HbSC, HbAA, %HbF, %HbA), reticulocyte count, LDH; and RBC adhesion to FN and LN. These factors have known clinical relevance.^56,122-124 However, the relative contribution made by each RBC characteristic is not yet known. Furthermore, a range of hemoglobin phenotypes can be distinguished through the use of the SCD biochip. Greater adhesion of RBCs to endothelial cells or endothelium-associated proteins in HbSS blood samples compared with HbAA blood samples has been shown abundantly in the literature by a number of eminent researchers.^{59,60,77,125-128} Moreover, RBC adhesion to FN and LN in patients undergoing hydroxyurea treatment was shown to decrease with increasing HbF levels.^110,126,129 Our results also showed greater adhesion of RBCs in HbSS blood samples compared with other hemoglobin phenotypes, and also in HbSS blood samples with lower HbF levels, which is in agreement with the previous studies.^{59,60,77,110,125-127,129} However, to the best of our knowledge, there is no study reporting a side-by-side comparison of RBC adhesion from HbAA, HbSC, and HbSS patient blood samples using a standard adhesion assay and multiple adhesive interactions, as we have shown here. It is likely that useful insights about hemoglobin phenotype and disease severity may be gleaned from the SCD biochip, if applied widely.

Hemolysis has been identified as a central pathophysiological trigger in SCD, due to the deleterious impact of free heme on endothelial health and inflammation.^123,130-135 The relationship between RBC biophysical properties and hemolysis in SCD, including RBC adhesion and deformability, is not fully understood. Increased rigidity and cyclic sickling of RBCs, resulting in cell membrane damage, has been suggested as the primary factor in intravascular hemolysis,^{55,134,136,137} and it is possible that red cell adhesion may associate more strongly with hemolysis than with pain. Hemoglobin and arginase released from RBCs during hemolysis result in decreased nitric oxide bioactivity, affecting vascular tone and endothelial activation.^134,138 Markers of hemolysis in blood, such as LDH, have been associated with elevated soluble adhesion molecules in plasma, including Vascular Cell Adhesion Molecule 1 (VCAM-1), Intercellular Adhesion Molecule 1 (ICAM-1), P-selectin, E-selectin, and Von Willebrand factor.^123,138,139 Furthermore, significant correlations between percentage of dense and nondeformable RBCs and hemolysis markers, including bilirubin, aspartate transaminase, and LDH, have been shown.^13,140 Although an association between hemolysis and elevated plasma adhesion molecules has been shown before,^123,138,139 there is no report, to the best of our knowledge, that links hemolysis to adhesion and deformability of RBCs simultaneously. In this work, we identified a unique adherent nondeformable RBC population, significantly correlated with serum LDH levels, which reflects a functional outcome in terms of both deformability and adhesion of RBCs and is a pathophysiologically plausible source for active hemolysis in SCD. Of note, no single clinical characteristic (high LDH, increased reticulocytes, or low HbF) definitively identified high and low adhesion subgroups; we are working to develop better predictive models in the future, using multivariate analyses and larger patient numbers. Pain scores vary significantly from patient to patient at baseline.^141,142 Subjectivity in pain perception and variations in pain scores in SCD patients might contribute to the fact that we did not observe associations between pain scores and our measurements.

Understanding the pathophysiology of abnormal cell adhesion has allowed development of new pharmaceutical approaches, such as agents targeting platelet adhesion to Von Willebrand factor in atherosclerosis,^143,144 and blocking therapies inhibiting malaria-infected RBC adhesion to endothelial cells.¹⁴⁵ Quantitative assessment of cellular adhesion is a promising indicator of disease activity due to its integration of multiple scales of interrogation, from molecular to cellular interactions. This concept has been used in cancer¹⁴⁶ and atherosclerosis.^74,147 In parallel, microfluidic technologies developed recently have paved the way for widespread utilization of cellular adhesion as a clinical marker due to their ease of use, short analysis time, and enhanced capacity to mimic physiological conditions.^65-67 Hydroxyurea, the only approved pharmaceutical treatment for SCD, increases HbF content in the RBCs in some, but not all, treated subjects.¹²⁴ Hydroxyurea may decrease RBC adhesion through both HbF-inducing and non–HbF-inducing mechanisms.^92,113,126 One quarter of all samples with high HbF (>8%), used in our studies, were obtained from patients with hereditary persistence of HbF. Furthermore, our description of an association between lesser RBC adhesion and higher HbF% is congruent with the observation that increased HbF content interferes with polymerization of sickle hemoglobin molecules and mitigates the effects of the disease. The contribution to this effect that is attributable to non–HbF-related effects of hydroxyurea cannot be discerned in our current analyses.

Recently, extraordinary efforts have been made to transform current knowledge about abnormal cellular adhesion in SCD into clinical benefit for patients using novel targeted agents. VLA-4 blocking antibodies have been proposed in SCD.¹⁴⁸ Beta-adrenergic receptor blockade, which targets epinephrine-mediated red cell adhesion,^95,149-152 has been used in patients, using an U.S. Food and Drug Administration (FDA) approved medication (propranolol).¹⁵³ Red cell adhesion to an activated endothelium has specifically been targeted with small molecules (α_Vβ₃),¹⁵⁴ low-molecular-weight heparin (P-Selectin),^155,156 and an oral agent in phase I/II studies in humans (P-selectin).^125,157,158 Abnormal interactions between leukocytes and activated endothelium and endothelial selectins have been targeted in vitro and in vivo with antibodies and small molecules^159-162 and are now the focus of phase II and III clinical studies in SCD. Although both P-selectin and E-selectin are important for adhesion of leukocytes to the endothelium, endothelial E-selectin appears crucial for generating secondary activating signals in the leukocyte.¹⁶³ The novel synthetic pan-selectin inhibitor, GMI-1070, with maximal activity against E-selectin has made it possible to test this paradigm in vivo in crisis.^162,164 Furthermore, a clinical trial using intravenous immunoglobulin infusions to block FcγRIII receptor mediation of secondary activity on neutrophils in children with SCD in crisis is in process (ClinicalTrials.gov identifier: NCT01757418).¹⁶⁵ A functional in-vitro adhesion test, such as the SCD biochip, could be a useful tool for measuring patients’ response to treatment. Moreover, the SCD biochip could be adopted as an in-vitro model in which to test therapeutics that target cellular adhesion.

Key clinical and experimental studies, using flow chambers or ex vivo rat mesocecum,^52,59-61 have shown that sickle RBC adhesion and deformability, WBC adhesion and activation,¹⁶⁶ and aberrant endothelial activation all plausibly contribute to the pathogenesis of vaso-occlusive crisis^{52,53,125,155} and may correlate with disease severity.^10-12 Besides FN and LN, sickle RBC adhesion to other endothelium receptors, such as p-selectin, α_vβ₃ integrin, and VCAM-1, has been presented in the literature.^53,125,151 Monocytes^167-170 and neutrophils^171-173 are often increased in number, abnormally activated, and associated with adverse outcomes in SCD.^{130,170,174-176} Quantitative abnormalities in lymphocytes and natural killer cells have been described^154,177 and are the focus of recent therapeutic interventions.^160,178 Perspectives on SCD pathophysiology must, in addition to RBC abnormalities, also integrate endothelial, WBC, and platelet activation and adhesion, inflammation, and thrombophilia.^{167,172,179-189} Our work is not meant to be an exhaustive survey of all potential abnormal cellular adhesion events in SCD. The new SCD biochip platform, introduced here, has the potential to informatively interrogate leukocyte adhesion, leukocyte activation, and RBC adhesion to other endothelial surface proteins, which we will incorporate in future studies.

Despite recent advances in identifying and targeting cellular adhesion in SCD, knowledge about biophysical properties of abnormal cellular adhesion has not been integrated into routine clinical care or trial design, due to a requirement for complicated custom-designed systems, highly trained personnel, specially obtained blood samples, and extensive sample manipulation. SCD biochip will provide a more precise characterization of abnormal adhesive events in a given individual and may allow a more accurate assessment of response to therapy, especially emerging adhesive therapies, overall. Our goal is to make the SCD biochip as feasible, and clinically meaningful in SCD as glucose testing is in diabetes.

The SCD biochip is an in vitro assay for quantitative assessment of RBC adhesion, which has potential for on-site monitoring of disease activity and vaso-occlusion. The SCD biochip is feasibly applicable to a range of “real-world” clinical scenarios, having been tested using surplus blood obtained from a busy urban SCD clinic, at baseline and during clinical flux (including promising initial studies after treatment and in association with long-term complications). Our expectation is that longitudinal interrogations of cellular adhesion with the SCD biochip will be informative about physiological changes in patients over time during the clinical course of SCD.

AT A GLANCE COMMENTARY.

Alapan Y, et al.

Background

Abnormal cellular adhesion to endothelium is at the root of vaso-occlusion in sickle cell disease (SCD). However, cellular adhesion is not currently evaluated clinically because these analyses are technically challenging and difficult to reproduce. This has limited the widespread applicability of adhesion assays clinically or as a research tool in SCD.

Translational Significance

We developed a microfluidic system that allows serial quantitative evaluation of red blood cell adhesion to endothelium proteins in a closed and preprocessing-free system. The presented in vitro adhesion assay would enable a comprehensive and integrated evaluation of red blood cell adhesive properties, which could be used at baseline, during crises, and with therapeutic interventions in SCD.

Abbreviations

CBC

complete blood count

EDTA

ethylenediaminetetraacetic acid

FCSB

flow cytometry staining buffer

FN

fibronectin

GMBS

N-g-Maleimidobutyryloxy succinimide ester

Hb

hemoglobin

LDH

lactate dehydrogenase

LN

laminin

RBC

red blood cell

SCD

sickle cell disease