Integrating deep learning with microfluidics for biophysical classification of sickle red blood cells adhered to laminin

Sickle cell disease, a genetic disorder affecting a sizeable global demographic, manifests in sickle red blood cells (sRBCs) with altered shape and biomechanics. sRBCs show heightened adhesive interactions with inflamed endothelium, triggering painful vascular occlusion events. Numerous studies employ microfluidic-assay-based monitoring tools to quantify characteristics of adhered sRBCs from high resolution channel images. The current image analysis workflow relies on detailed morphological characterization and cell counting by a specially trained worker. This is time and labor intensive, and prone to user bias artifacts. Here we establish a morphology based classification scheme to identify two naturally arising sRBC subpopulations—deformable and non-deformable sRBCs—utilizing novel visual markers that link to underlying cell biomechanical properties and hold promise for clinically relevant insights. We then set up a standardized, reproducible, and fully automated image analysis workflow designed to carry out this classification. This relies on a two part deep neural network architecture that works in tandem for segmentation of channel images and classification of adhered cells into subtypes. Network training utilized an extensive data set of images generated by the SCD BioChip, a microfluidic assay which injects clinical whole blood samples into protein-functionalized microchannels, mimicking physiological conditions in the microvasculature. Here we carried out the assay with the sub-endothelial protein laminin. The machine learning approach segmented the resulting channel images with 99.1±0.3% mean IoU on the validation set across 5 k-folds, classified detected sRBCs with 96.0±0.3% mean accuracy on the validation set across 5 k-folds, and matched trained personnel in overall characterization of whole channel images with R2 = 0.992, 0.987 and 0.834 for total, deformable and non-deformable sRBC counts respectively. Average analysis time per channel image was also improved by two orders of magnitude (∼ 2 minutes vs ∼ 2-3 hours) over manual characterization. Finally, the network results show an order of magnitude less variance in counts on repeat trials than humans. This kind of standardization is a prerequisite for the viability of any diagnostic technology, making our system suitable for affordable and high throughput disease monitoring.

Integrating deep learning with microfluidics for biophysical classification of sickle red blood cells

Sickle cell disease (SCD), a group of inherited blood disorders with significant morbidity and early mortality, affects a sizeable global demographic largely of African and Indian descent. It is manifested in a mutated form of hemoglobin that distorts the red blood cells into a characteristic sickle shape with altered biophysical properties. Sickle red blood cells (sRBCs) show heightened adhesive interactions with inflamed endothelium, triggering obstruction of blood vessels and painful vaso-occlusive crisis events. Numerous studies have reported microfluidic-assay-based disease monitoring tools which rely on quantifying adhesion characteristics of adhered sRBCs from high resolution channel images. The current workflow for analyzing images from these assays relies on manual cell counting and detailed morphological characterization by a specially trained worker, which is time and labor intensive. Moreover manual counts by different individuals are prone to artifacts due to user bias. We present here a standardized and reproducible image analysis workflow designed to tackle these issues, using a two part deep neural network architecture that works in tandem for automatic, fast and reliable segmentation and classification into subtypes of adhered cell images. Our training utilized an exhaustive data set of images generated by the SCD BioChip, a microfluidic assay which injects clinical whole blood samples into protein-functionalized microchannels, mimicking physiological conditions in the microvasculature. The automated image analysis performs robustly in comparison to human classification: accuracies were similar to or better than those of the trained personnel, while the overall analysis time was improved by two orders of magnitude.

Confined filaments in soft vesicles – the case of sickle red blood cells

Abnormal shapes of red blood cells (RBC) have been associated with various diseases.

Patrolling Monocytes Scavenge Endothelial Adherent Sickle Red Blood Cells in Sickle Cell Disease

Sickle cell disease (SCD) is characterized by hemolytic anemia and increased entrapment of sickle red blood cells (RBCs) via attachment to the underlying activated vascular endothelium, resulting in vaso-occlusive crisis (VOC), marked by severe pain. The endothelial scavenging patrolling monocytes (PMos) expressing high levels of the heme oxygenase 1 (HO-1), a heme degrading enzyme, were recently shown to protect against vaso-occlusion in SCD, although their ability to scavenge endothelial-attached sickle RBCs was not tested. Here, we found that circulating PMos from SCD patients showed roughly 5% ± 0.5% engulfed GPA+ or Band3+ RBC specific material as compared to 0.7% ± 0.04% in healthy donor (HD) PMos or 0.85% ± 0.07% in SCD classical monocytes (CMos) as detected by flow as well as imagining flow cytometry, suggesting that PMos uptake RBCs in SCD. To further investigate this, RBCs purified from HDs (to mimic transfused cells) or SCD patients were labelled with CFSE, and co-cultured with purified monocytes without or with human microvascular endothelial cells (HMVEC). We found 11% ± 0.5% CFSE+ PMos in co-cultures with SCD RBCs in presence of HMVEC as compared to 2.7% ± 0.4% when cultured with HD RBCs, indicating that PMos engulf sickle RBCs, but not HD RBCs. Low levels of CFSE+ PMos (2-3%) were detected in cocultures with either sickle or control RBCs in the absence of HMVEC, implicating that PMos preferentially uptake endothelial cell (EC)-attached sickle RBCs. In contrast to PMos, CMos always had minimal CFSE+ reactivity (2-3%) when cultured with sickle RBCs or HD RBCs with or without HMVEC, supporting a role for PMos, but not CMos, as scavengers of sickle RBCs in the vasculature. Further analysis revealed significantly increased (two-fold) levels of annexin V+ apoptotic marker on sickle RBC engulfed PMos (CFSE+PMos), but also higher levels of HO-1 ((50% ± 3%) as compared to non-engulfed (CFSE-) PMos (1.3% ± 0.4%), suggesting that induction of HO-1 upon uptake of sickle RBCs may counteract the cytotoxic effects of engulfed RBC breakdown products in PMos. To test this hypothesis, monocytes were pre-treated with tin protoporphyrin IX (SnPPIX) to inhibit HO-1 activity prior to coculture with sickle RBCs and HMVEC. We found increased apoptosis in PMos from SnPPIX-treated (18% ± 0.8%) as compared to untreated cultures (6.6% ± 0.6%), consistent with a cytoprotective role of HO-1 induction in sickle RBC-engulfed PMos. Antibody blocking studies identified ICAM-1, VCAM-1, CD11a, CD18 as well as CD16 as key molecules involved in PMo-HMVEC-sickle RBC interactions, but not CD11b, CD11c, CD31, PSGL-1, CD32, CD64, Fc a/m receptor and phosphatidylserine. Interestingly, HMVEC activation induced by heme treatment resulted in significantly higher CFSE+ PMos (17.8% ± 0.9%) when cultured with sickle RBCs, but not HD RBCs (3.5% ± 0.6%), indicating that PMos preferentially uptake sickle RBCs bound to heme-treated ECs. To formally test this, DiI labelled mouse sickle or control RBCs were transfused to heme-treated Nr4a1-GFP mice which express GFP on their circulating PMos, followed by perfusion to remove non-EC attached cells. Using confocal microscopy, we detected DiI+ sickle but not control non-sickle RBCs, engulfed by GFP+ PMos in the vasculature of perfused recipients, confirming uptake of EC-attached sickle RBCs by PMos. Consistent with increased EC-attached sickle RBCs during VOC, SCD patients experiencing acute VOC (n=12) showed increased frequency of GPA+ PMos (13% ± 0.1% vs 5% ± 0.2%, p<0.01) but still only 1.8% ± 0.3% GPA+ CMos. However, total numbers of PMos including HO-1hi PMos were more than two-fold lower in these patients as compared to patients at steady state. Altogether, these data demonstrate for the first time that in SCD, PMos scavenge EC-attached sickle RBCs, but not transfused HD RBCs, through interactions involving manly integrins, resulting in HO-1 upregulation to counteract the cytotoxic effects of engulfed RBC breakdown products. With increased adherence of sickle RBCs to vasculature as a result of damage to ECs such as during VOC, PMos scavenge more RBCs, but their numbers become limiting with implications for beneficial effects of transfusions and potential for PMos as therapeutic targets against VOC in SCD. Disclosures No relevant conflicts of interest to declare.