Profiling single-cell level phagocytic activity distribution with blood lactate levels†
The ability to kill infecting microbes is an essential facet of our immune response to an infection. However, phagocytic ability is often overlooked as a part of immunological profile in infected patients' diagnosis, as the understanding of phagocytic capabilities in disease states is incomplete. In this work, we have evaluated for the first time the relationship between blood lactate level and the neutrophil phagocytic activity at a single-cell level. Blood samples (N = 19) were grouped on the basis of their blood lactate levels i.e., below (control) or above 2 mmol L−1 (high-risk) (i.e., 2 mmol L−1 is a common clinical lactate threshold used for patients' triage). Neutrophils were isolated from whole blood and then incubated with fluorescent IgG coated beads for 40 minutes, and the ability of each neutrophil to internalize beads was quantified. Single-cell phagocytic activity analysis has shown interesting findings such as: (i) a single neutrophil was able to internalize up to 7 beads, (ii) for a control group, 39.76% cells didn't internalize any beads, while for a high-risk group, 30.65% cells didn't show any phagocytic activity, (iii) similarly, 30.46% cells internalize only 1 bead in a control group, while for a high-risk group the activity is slightly higher with only 31.73% cells showing single bead internalization, and (iv) 7 bead internalization activity was much higher for samples in a high-risk group (0.6% cells) compared to a control group (0.17% cells). We used multiple statistical tests to compare these differences. For a two-tailed T-test, we used the mean phagocytic activity of the cells (i.e., the average number of beads internalized by cells) isolated from the blood samples in the two groups (1.14 vs. 1.35) and found the p-value to be 0.08. We also used principal component analysis (PCA) on this high dimensional phagocytic activity distribution data and performed dimension reduction. However, the first 3 principal components didn't show a clear distinction between groups. Next, we developed machine learning models using artificial neural networks (ANNs) to differentiate between the distribution of phagocytic activity in neutrophil populations of the two groups. Our models yielded area under curve (AUC) values below 0.7 for receiver operator characteristic curves. Although our study highlighted interesting phagocytic activity findings at a single cell level, it further highlights the need for integration of an individual patient's medical record to get more personalized insights into individual phagocytic activity in the future.