Tripartite collaboration of blood‐derived endothelial cells, next generation RNA sequencing and bioengineered vessel‐chip may distinguish vasculopathy and thrombosis among sickle cell disease patients

Abstract Sickle cell disease (SCD) is the most prevalent inherited blood disorder in the world. But the clinical manifestations of the disease are highly variable. In particular, it is currently difficult to predict the adverse outcomes within patients with SCD, such as, vasculopathy, thrombosis, and stroke. Therefore, for most effective and timely interventions, a predictive analytic strategy is desirable. In this study, we evaluate the endothelial and prothrombotic characteristics of blood outgrowth endothelial cells (BOECs) generated from blood samples of SCD patients with known differences in clinical severity of the disease. We present a method to evaluate patient‐specific vaso‐occlusive risk by combining novel RNA‐seq and organ‐on‐chip approaches. Through differential gene expression (DGE) and pathway analysis we find that BOECs from SCD patients exhibit an activated state through cell adhesion molecule (CAM) and cytokine signaling pathways among many others. In agreement with clinical symptoms of patients, DGE analyses reveal that patient with severe SCD had a greater extent of endothelial activation compared to patient with milder symptoms. This difference is confirmed by performing qRT‐PCR of endothelial adhesion markers like E‐selectin, P‐selectin, tissue factor, and Von Willebrand factor. Finally, the differential regulation of the proinflammatory phenotype is confirmed through platelet adhesion readouts in our BOEC vessel‐chip. Taken together, we hypothesize that these easily blood‐derived endothelial cells evaluated through RNA‐seq and organ‐on‐chips may serve as a biotechnique to predict vaso‐occlusive episodes in SCD patients and will ultimately allow better therapeutic interventions.


| INTRODUCTION
Sickle cell anemia (SCA) along with its other clinical subtypes (sickle cell β thalassemia, hemoglobin SC, etc.) is the most prevalent rare disease in the United States and most common genetic disease in the world. 1 Roughly 100,000 people are affected in the United States, out of which the African-American population has a particularly higher incidence of the disease, with at least one individual out of 13 carrying the autosomal recessive mutation. 2,3 Sickle cell disease (SCD) is characterized by a complex gamut of hematological and vascular complications. 4 Within the vessels, the unusual vaso-occlusive cascade involves endothelial activation, platelet adhesion and red cell binding, that can differ among patients. A hypercoagulable state of SCD blood further exacerbates the endothelial-blood interactions and can lead to vaso-occlusion. 5 The acute and chronic manifestations of vasculopathy in SCD are multifactorial as they are dependent on the relative hemoglobin distribution, extent of red cell hemolysis, presence of cell-free hemoglobin and heme, hypercoagulability of blood and endothelial activation. 6 Also, nearly a quarter of SCD patients encounter a stroke by the age of 45 years, 7 and the risk of stroke is associated with inherent vasculopathy.
The complications contributing to the vasculopathy in SCD result from a combination of proinflammatory phenotype of the native endothelium and a hypercoagulable state of blood. [8][9][10] Development of relevant animal models and advancements in the field of in vitro tissue engineered models, like organ-on-chip, have greatly enhanced our understanding of the disease. [11][12][13] However, there is still a considerable knowledge gap in understanding the clinical heterogeneity within the SCD population as these models cannot recapitulate populationspecific outcomes of the disease. It has been observed clinically that different patients show different extents and frequencies of vasoocclusive crises, 14 which ultimately necessitates the need of a predictive model that can differentiate patients and can aid clinicians as a risk evaluation methodology.
An essential requirement for developing a model that mimics patient pathophysiology is to identify autologous cell sources that can recapitulate patient-specific readouts in vitro. 15 In our recent work, we have identified blood outgrowth endothelial cells (BOECs) isolated from circulation as a disease-specific primary cell source to analyze endothelial activation and thromboinflammation in vitro. 16 We further hypothesize that they can potentially mimic patient-specific responses in disease. BOECs exhibit classical endothelial characteristics similar to primary cells and can reveal disease-specific differences in endothelial activation, oxidative stress and metabolic activity relative to control cells, once incorporated in the microfluidic vessel-chips. 16 Increased presence of circulating endothelial cells in vascular disorders also makes them a viable cell model. [17][18][19] Advancements in next-generation sequencing (NGS) like RNA-seq has further enabled assessment of differential gene expression in health and disease with high fidelity. Combining the predictive power of autologous, patient-derived cells like BOECs with tools like RNA-seq can allow investigation of patient-specific genome signature.
Incorporating BOECs in organ-or vessel-chips can further help in functional validation of patient-specific phenotype as predicted by RNA-seq and ultimately lead to development of a patient assessment pipeline.
In this report, we test the aforementioned methodology by isolating BOECs from two patients with known differences in their clinical SCD severity. We explored if easily derived BOECs taken from these patients may serve as: (1) a biomarker to validate the distinct clinical difference between the two patients; and (2)through RNA-seq analysis to diagnose a potentially differential molecular pathophysiology related to endotheliopathy and thrombosis. Through RNA-seq and differential gene expression (DGE) studies of these cells, as well as phenotypic assessment through vessel-chip blood perfusion experiments, we provide a proof-of-feasibility of using this integrative approach to assess endotheliopathy and thrombotic potential among SCD patients from tissue-to-molecular scale.

| RESULTS AND DISCUSSION
We initiated the study by selecting two age-matched patients who represented significantly different clinical manifestations of the sickle cell disease ( Table 1). The critical distinction between the two was that one patient had hemoglobin SC disease (SCD-SC) with a relatively milder disease severity, while the other patient had hemoglobin SS (SCD-SS) and had a confirmed history of stroke and transfusion therapy, very likely susceptible to endothelial dysfunction and thrombosis. 20,21 Hemoglobin SC (HbSC) disease is clinically considered a milder variant of SCA although the treatments available to patients are largely derived from studies performed on hemoglobin SS patients. 22 Although the two subtypes constitute the majority of SCD population with~30% of patients having the HbSC mutation, the clinical manifestation and phenotype are very different. 23 Being the less severe phenotype, patient morbidity and mortality are lower among the HbSC patients. On the other hand, patients with sickle cell anemia (1) have more exaggerated inflammatory profiles in blood, (2) have a higher incidence of irreversible RBC sickling, (3) have shortened RBC lifespans compared to hemoglobin SC patients, (4) witness more vasoocclusive episodes, and (5) are more susceptible to infections. 24,25 Reports suggest that HbSC disease patients have lower levels of fetal hemoglobin (HbF) compared to SCA counterpart and the same is witnessed in our findings (Table 1). Hence it is of utmost importance that we gain knowledge of the clinical distinction and possible manifestations to develop better disease management strategies and targeted therapies for the two SCD variants. After selecting the patients, we isolated mRNA from respective patient BOECs and processed them for next generation RNA sequencing (Figure 1a).
Post-sequencing and alignment of sequence reads, we investigated differential gene expression among the SCD patients with respect to GO:0045321) using Cytoscape. 27,28 As expected, the severe SCD-SS case had more genes regulating these processes compared to SCD-SC and exhibited stronger interactions between the regulating genes ( Figure 1f,g). This broad categorization of biological processes into the GO terms listed above in fact encompassed few critically suspected endothelial activation and thromboinflammation pathways as predicted by KEGG analysis ( Figure S4). Specifically, the family of genes encoding for cell adhesion molecules was upregulated in the patients and contributes to the thromboinflammatory phenotype of these blood derived cells. 29,30 Taken together, these results support that the SCD patient who had a history of stroke and was clinically diagnosed with severe SCD symptoms, had a transcriptomic upregulation of endothelial activation and thrombosis.
To further identify the extent of endothelial activation among the patients, we performed a KEGG pathway clustering of the conserved genes (~400, Figure 1b To investigate the differential expression of genes belonging to the aforementioned KEGG pathways, we generated heatmaps for comparison among the two patients relative to controls ( Figure 2b).
Interestingly, BOECs from severe SCD-SS patient expressed genes contributing to endothelial activation to a higher extent relative to control and SCD-SC implying that BOECs from SCD-SS were in a severely thromboinflammation state. In contrast, BOECs from patient SCD-SC exhibited signs of endothelial dysfunction that were intermediate between that of controls and SCD-SS (Figure 2b). Such widespread comparison between patients not only revealed the differential presence of these pathways, but also the extent to which they were differentially expressed; SCD-SS had a much diverse expression profile with more upregulated/downregulated genes, while F I G U R E 1 Qualitative assessment of differential gene expression among sickle cell disease patients through autologous BOECs and RNAseq. (a) Schematic of the BOEC isolation, expansion and subsequent mRNA extraction process followed in this study. Isolated mRNA from control and SCD patient (SCD-SC and SCD-SS). BOECs were then assessed for quality and processed for sequencing. (b) Post-RNA sequencing and subsequent alignment, differential gene expression analysis revealed significant differences between patient genetic signatures; SCD-SS BOECs had significantly more differentially expressed genes compared to SCD-SC. Out of the~2000 genes analyzed, roughly 400 genes were conserved in SCD-SC and SCD-SS. (c and d) Volcano plots of the differentially expressed genes for SCD-SC and SCD-SS, respectively, relative to heathy controls. In agreement with (b), SCD-SS BOECs exhibit relatively higher and statistically stronger fold change differences compared to SCD-SC (black: excluded genes with −2 < log 2 [FC] < 2, red: differentially expressed genes). (e) Gene ontology (GO) based clustering of differentially expressed genes indicate differences within regulating biological processes and key cellular components between SCD-SC and SCD-SS (p < 0.05).
(f and g) Gene cluster networks exhibiting complex interactions between the most prominent biological processes regulating vascular tone (cell adhesion: GO:0007155; cell-cell signaling: GO:0007267; chemotaxis: GO:0006935, and leukocyte activation: GO:0045321). Compared to SCD-SC, patient SCD-SS expressed significantly more genes and hence exhibited more complex gene interactions among the aforementioned biological processes (red, up-regulated; blue, down-regulated; size increases with significance) SCD-SC had fewer genes being differentially regulated (Figure 2c,d).
These results agree with the qualitative gene expression profiles described earlier (Figure 1) as well as the clinical histories of the two patients (Table 1).
In order to support the results obtained through the RNA-seq and DGE studies, we also analyzed common endothelial activation and vaso-protective markers like E-selectin, P-selectin, ICAM-1, VCAM-1, tissue factor (TF), thrombomodulin, and von Willebrand Factor (VWF).
Selectins, specifically P-selectin, have been implicated in SCD causing endothelial-RBC interactions and subsequent thrombosis and ischemia. 12,15 Tissue factor expression by endothelial cell in SCD physiology has also been postulated to contribute to the ensuing vaso-occlusive crises. 31 In agreement with these findings, our results reveal that among the common adhesion proteins expressed by the endothelium, both SCD patients had an upregulation of E-selectin, P-selectin, tissue factor, and VWF while other markers like ICAM-1 and VCAM-1 were moderately upregulated (Figure 3a). Additionally, these genes were differentially regulated between the two patients with SCD-SS exhibiting a higher fold change expression compared to SCD-SC and both patients having more expression than control (Figure 3a). Taken together, these results suggest that RNA-seq of BOECs from SCD patients may serve as a model to assess SCD patient severity.
Finally, we set out to investigate phenotypic differences that the BOECs exhibit between the SCD patients and predict microvascular thromboinflammatory consequences due to disease severity within the patients. Our prior work has repeatedly shown that in vitro blood vessel organ-on-a-chip is a platform technology to visualize blood-endothelial interactions in real-time. 32 arterioles" were ready, we perfused them with autologous blood samples at arteriolar flow conditions and examined real-time plateletendothelial adhesion and coagulation using fluorescence microscopy ( Figure 3b, Movie S1). We observed that BOEC-vessel-chip of the SCD patients were both more adhesive than normal controls. However, severe SCD-SS patient had a significantly higher platelet adhesion to the BOEC endothelium, relative to the mild SCD-SC patient, demonstrating that BOECs of a severe SCD case are hyperactivated and prothrombotic (Figure 3c,d and Movie S1). These functional blood perfusion studies also correlate to the DGE results obtained through RNA-seq (Figures 1 and 2) and suggest that harnessing BOECs from patient blood samples, and analyzing them through RNA-seq and vessel-chips may provide a genotype and phenotype signature potentially valuable in assessing disease severity in SCD.

| CONCLUSIONS
In this proof-of-concept study, we present a patient vaso-occlusive risk assessment methodology utilizing a novel combination of autologous endothelial progenitors from cardiovascular patients as an alternative cell model, RNA-sequencing and organ-on-chip technology.
F I G U R E 3 Functional assessment of patient-derived BOECs and vessel-on-chip assembly. (a) Quantification of expression of common endothelial surface markers through qRT-PCR reveals a significant upregulation in E-selectin, P-selectin, tissue factor, and VWF in SCD-SS BOECs relative to control. In agreement with the sequencing results, SCD-SC BOECs exhibited lower expression of the aforementioned markers relative to SCD-SS although more than that of controls. (b) Schematic of the thromboinflammation analysis performed with patient BOECs vessel-chips. After an overnight culture in rectangular microchannels under constant laminar media perfusion, autologous BOECs were exposed to healthy whole blood in vitro and subsequent platelet adhesion and clotting events were monitored through real-time fluorescent microscopy.  Current in vitro microfluidic models of SCD have put primary focus on red blood sickling and hemolysis in SCD and the endothelial activation in SCD has been relatively understudied. 39,40 As a result, there is a knowledge gap in understanding the interactions between native endothelium and blood components in SCD microcirculation. Inability to study the convoluted transformation from a healthy, to an "activated" state and ultimately acquiring a "dysfunctional" endothelial phenotype has added additional burden over existing disease management strategies. Previously published studies have reported endothelial-blood interactions in SCD, they however utilize primary cells isolated from healthy individuals that are exogenously stimulated to mimic an activated endothelium and hence cannot elicit differences in endothelial-blood crosstalk among patients. 41,42 Consequently, this is a first-of-its-kind study utilizing autologous SCD patient cells to characterize differential vascular dysfunction between two clinically diverse patients.
We compare the gene expression profiles of these patients and categorize the differentially expressed genes into biological processes and molecular pathways using widely used pathway annotation tools F I G U R E 4 Future scope of the BOEC-RNA-seq-organ-chip pipeline to evaluate patient-specificity within SCD population. (a) Schematic demonstrating application of the study pursued in this article to a wider, more diverse population of SCD patients. Profiling personalized genetic signatures through RNA-seq and phenotyping microvascular behavior using organ-chip technology can allow clinicians and pharmaceutical researchers correlate patient clinical outcomes and eventually improve treatment prospects via personalized therapy. (b) We hypothesize that exploiting the predictive power of BOECs, harnessed via RNA-seq and organ-chip technology, can also allow clustering/grouping of patients into different categories based on disease severity. This will enable clinicians to prescribe treatments specific to patient categories, improve the drug discovery and screening pipeline in the pharmaceutical industry, improve therapeutic outcomes and ameliorate patient conditions, and ultimately progress the current state of healthcare like DAVID and Cytoscape that offer gene ontology (GO) and KEGG pathways-based clustering. Although these annotation methodologies have their caveats as clustering is often broad, specificity can be low and matching of pathways is limited to the current annotations present in the database, 43,44 these can still provide holistic differences between patient genetic profiles.
Although we have limited the scope of this study to characterize two patients only, this proof-of-feasibility study further lays the groundwork for assessment of a much diverse and extensive SCD patient cohort (Figure 4a).

| Vessel-chip design and fabrication
Microfluidics vessel-chips were designed in SolidWorks so that they resembled small arterioles (~100 μm).

| Vessel-chip functionalization and endothelial cell culture
The microfluidic devices were first treated oxygen plasma for 30 s at a power of 50 W prior to treatment with a 1% solution of (3-aminopropyl)-trimethoxysilane (APTES, Sigma-Aldrich) in 200 proof ethanol. After a 10 min silane treatment, the solution was rinsed out first with 70% ethanol followed by 100% ethanol. The devices were then stored in a 70 C oven for 2 h. The devices were then filled with a 100 μg/ml solution of type-1 rat tail collagen (Corning) and incubated at 37 C in a 5% CO 2 incubator for an hour. The collagen solution was then rinsed out with BOEC media. Patient BOECs were trypsinized from confluent cell culture flasks and resuspended BOEC growth media at a concentration of 10 million cells/ml and seeded into pretreated microfluidic devices. The BOEC seeded microfluidic devices were incubated at 37 C in a 5% CO 2 incubator for an hour.
After initial attachment of cells on one side of the microfluidic devices, the process was repeated by seeding a freshly prepared cell suspension at the aforementioned concentration in the devices and incubating again at 37 C in a 5% CO 2 incubator for an hour while upside-down to promote BOEC attachment on all sides of the microfluidic channel. To mimic the native vessel physiology, BOEC seeded microfluidic devices were constantly perfused with growth media overnight. The devices were then connected to a syringe pump that perfused BOEC growth media through the devices at flow rate of 1 μl/min (shear stress: 0.81 dynes/cm 2 ; shear rate: 81 s −1 ) overnight.
This flow rate was chosen to provide the cells with an arteriolar shear rate while optimizing growth media usage. 38,47,48 Perfusion of growth media ensured constant supply of nutrients to the cells and alignment of the cells along the direction of flow.

| Blood perfusion and microscopy
Blood from healthy donors was collected in 3.2% sodium citrate tubes (BD Biosciences) and used according to the Institutional Review Board and 75 mM MgCl 2 was mixed with blood in a 1:10 ratio prior to perfusion. 32 The devices were mounted on an automated microscope (Ziess Axio Observer) and real-time fluorescence imaging was performed for a duration of 15 min.

| Statistics
All data shown are mean ± SD. Statistical analysis has been performed using GraphPad Prism ver. 8. Comparisons between groups made using ANOVA or Student's t-test. Differences are considered statistically significant if p < 0.05.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.