A point-of-care microfluidic biochip for quantification of CD64 expression from whole blood for sepsis stratification

Abstract

Sepsis, a potentially life-threatening complication of an infection, has the highest burden of death and medical expenses in hospitals worldwide. Leukocyte count and CD64 expression on neutrophils (nCD64) are known to correlate strongly with improved sensitivity and specificity of sepsis diagnosis at its onset. A major challenge is the lack of a rapid and accurate point-of-care (PoC) device that can perform these measurements from a minute blood sample. Here, we report a PoC microfluidic biochip to enumerate leukocytes and quantify nCD64 levels from 10 μl of whole blood without any manual processing. Biochip measurements have shown excellent correlation with the results from flow cytometer. In clinical studies, we have used PoC biochip to monitor leukocyte counts and nCD64 levels from patients' blood at different times of their stay in the hospital. Furthermore, we have shown the biochip's utility for improved sepsis diagnosis by combining these measurements with electronic medical record (EMR).

Figure 1. Overview of the electrical cell-counting technology.

(a) Process schematic of the differential expression-based cell-counting technology. Whole blood (10 μl) is infused in the biochip along with lysing and quenching buffers, to preferentially lyse erythrocytes. Cells are electrically counted and differentiated based on their size using microfabricated electrodes. Anti-CD64 (clone 10.1) antibody is initially adsorbed in the chamber. The CD64+ cells get captured based on their CD64 expression level. The difference in the cell counts from cell counters is linearly correlated with the nCD64 expression level. (b) The resulting pulse amplitude histogram representing lymphocytes and granulocytes + monocytes as two distinct populations. (c) Correlation (coefficient of determination: R²=0.89, P<0.0001) in between biochip total leukocytes versus control leukocyte counts obtained from haematology analyzer using n=181 blood samples. (d) The correlation (coefficient of determination: R²=0.88, P<0.0001) in between biochip total granulocytes + monocytes versus control cell counts obtained from haematology analyzer.

Figure 2. Clinical validation of nCD64 for sepsis stratification and prognostication.

(a) The CD64 expression histograms for monocytes, neutrophils and lymphocytes for n=450 blood samples. The inset histogram represents nCD64 expression for all the blood samples. (b) The plot represents the ratio of mCD64/ nCD64 versus nCD64. (c) ROC curves showing sepsis predictability using Quick SIRS (AUC=0.7) and SIRS + CD64 (AUC=0.77). (d) The box plots showing the control nCD64 expression value from 316 blood samples collected from 68 patients (who recovered later) at different time windows of their hospital stay. (e) The box plots showing the control nCD64 expression value from 94 blood samples collected from six patients who are at different times of their hospital stay. Unfortunately, these patients did not recover. (f) The nCD64 and WBC counts are used for sepsis prognosis for all the six time windows using ANN model. ROC curves for TW1 (left) and TW5 (right) show the highest AUC>0.9 for sepsis prognosis by predicting the recovery of the patients. P-values are reported in Methods section.

Figure 3. Expression-based CD64 cell capture in the capture chamber immobilized with anti-CD64 (clone 10.1) antibody.

(a) Antibody adsorption characterization of the capture chamber. BSA-blocked chamber shows no fluorescence; however, PE-conjugated anti-CD64 antibody-filled chamber shows high fluorescence. In antibody-adsorbed chamber, white circular patterns at the periphery of the pillars show the adsorbed antibody after the unadsorbed antibody is flushed away. Intensity plot along A–B, with intensity peaks represent the white regions around pillars’ periphery and valleys represent pillars. (b) CD64 expression-based capture of monocytes using Sample A (mCD64=6.03), with inset CD64 histograms of monocytes before capture (Blue) and after capture (Red) shows almost complete capture. The red and blue curves show the exit and the entrance normalized percent monocyte count versus mCD64 expression,respectively, with green bars representing difference of entrance minus exit cell counts. (c) CD64 expression-based capture of neutrophils using Sample A (nCD64=2.43). (d) CD64 expression-based capture of monocytes using Sample B (mCD64=3.9). (e) CD64 expression-based capture of neutrophils using Sample B (nCD64=1.08). (f) The amplitude histogram of the cell pulses from the entrance counter (green) and exit counter (red). (g) False-coloured fluorescent image showing the captured CD64 cells around the pillars. Cells are captured specifically with antigen–antibody interaction as they lie alongside the streamlines (scale bar: 40 μm). (h) The plot shows a linear correlation between percent granulocytes + monocytes capture versus nCD64 expression ratio with (coefficient of determination: R²=0.87, P<0.0001). (i) The percent granulocytes + monocytes capture averaged from all samples in a bin versus 0.25 nCD64 bins. The plot shows a linear correlation between them with (coefficient of determination: R²=0.96, P<0.0001) with error bars representing s.d in cell capture. The green bars of secondary y axis represent the number of samples in each bin.

Figure 4. Patient stratification and time course measurements.

(a) The comparison plot between nCD64 value obtained from biochip versus flow cytometry, showing the good linear correlation of coefficient of determination: R²=0.87, P<0.0001. (b) The Bland–Altman analysis comparing the nCD64 values from biochip versus flow cytometer (control). It shows −0.0002 average difference of nCD64 value in between biochip and flow cytometer (control) and 0.49 value as limits of agreement. (c) ROC curves to predict the three bins used for sepsis diagnosis (Fig. 2h). AUC for Bins 1, 2 and 3 are 0.98, 0.85 and 0.9, respectively, showing the high predictability accuracy for all the bins. (d) The plot represents the nCD64 prediction accuracy from the biochip. The accuracy is > 85% for the bin size of 0.6 nCD64 value, which corresponds to ten bins. (e) The box plots showing the nCD64 expression value obtained from biochip from 91 blood samples collected from patients at different times of their hospital stay. (f) The box plots showing the total leukocyte counts obtained from biochip from 91 blood samples from the same patients as in e. (g) The plot representing the nCD64 value and total WBC counts obtained from biochip using 91 blood samples collected at different times of patient hospital stay. (h) Total leukocyte count and nCD64 value of the patients are tracked over time. (i) ROC curves showing high sensitivity and specificity of leukocytosis and neutrophilia diagnosis from the biochip.