Quantifying biophysical heterogeneity by single-cell microfluidic impedance cytometry
Chair: Gustavo Kunde Rohde (ECE, BME)
Advisor: Nathan Swami (ECE)
Todd W. Bauer (Surgery, School of Medicine)
Mohammad Fallahi-Sichani (BME)
Federica Caselli (CE & CS, University of Rome Tor Vergata)
Heterogeneity in biophysical properties, which is inherent to the functional and structural organization of biosystems, presents challenges to cell biologists and clinicians seeking to associate biological function and disease with particular markers. Current high throughput methods to quantify heterogeneity focus on biochemical properties, as quantified by single-cell flow cytometry after fluorescent staining for their characteristic cell surface proteins. Cellular biophysical metrics, on the other hand, have often been restricted to size-based differences that do not provide sufficient functional information on the biosystem. Frequency-resolved impedance cytometry in microfluidic systems is emerging as a tool for multiparametric and high-throughput biophysical stratification of phenotypes in a label-free manner. However, there is a need to standardize the metrics for enabling facile recognition and automated fitting to quantify subpopulations in heterogeneous biological samples. This will be explored per the following aims.
Aim 1 – Modified red blood cells as model particles with modulated electrophysiology:
Biophysical properties, such as cell membrane capacitance and its interior organization, cannot be determined by data normalization using common model particles like fixed-size insulating beads that lack the internal structure of live cells or yeast that have a wide range of size and shape distributions. We seek to validate the hypothesis that red blood cells (RBCs) that are tailored in biophysical properties, such as membrane capacitance and cytoplasm conductivity, can potentially be used to normalize single-cell impedance metrics to account for temporal variations during the measurements and for enabling facile comparison of data across biological samples.
Aim 2 – Machine learning based methods for automating quantification of subpopulations:
The lack of standardized methods to gate single-cell impedance data can lead to to user-related variations in subpopulation quantification. Hence, automated gating protocols that are informed by biophysical information from dielectric cell models will be developed to standardize the quantification of subpopulations after the treatment of patient-derived pancreatic cancer xenograft models with chemotherapeutic agents. Specifically, we will explore: (i) Supervised learning to quantify proportions of apoptotic (early and late) vs. viable and necrotic populations in cell lines of varying levels of drug resistance (gemcitabine); and (ii) Supervised learning to distinguish drug sensitivity of cancer versus fibroblast cells in samples of varying heterogeneity.
In this manner, standardized model particles (Aim 1) and automated data quantification protocols (Aim 2) that are developed as part of this work will enable improved quantification of biophysical heterogeneity in complex samples, for serving as label-free cellular markers to aid the study of disease progression, drug toxicity, and immune responses.