What we do

The goal of our lab is the technological development of microfluidic platforms to enable the study of tissues at single-cell resolution. Single-cell genomics technologies enable the sequencing of tissues cell by cell to reveal the presence of genetically distinct cell subpopulations, such as emerging clones responsible for the growth and spread of tumors. Early diagnosis of diseases such as cancer is paramount because early intervention is associated with a much better outcome. However, this requires identification of subpopulations that can represent a very small fraction of the cell population. This challenge highlights the need for large scale single cell preparation and enrichment to efficiently capture this clinically relevant small fraction.

My work

My dissertation explores the magnetic manipulation of microfluidic droplets to enable high-throughput single-cell genomics that will address this challenge. It develops 3 main thrusts: 1) the use of magnetic particles to purify mRNA molecules from droplets, 2) a widely accessible high-resolution 3D microflow mapping tool that can be applied to study the recirculation flow inside microfluidic droplets, and 3) the use of magnetic particles to manipulate and select specific single-cell laden hydrogel beads formed using droplet microfluidics.

Research thrust 1: Physical dynamics of magnetic droplet purification

Droplet microfluidics, which encapsulates a sample within tiny (10-100μm) droplets surrounded by oil, is a technology of choice for single-cell genomics. It enables manipulation of single-cell-filled microdroplets at high-throughput (> 1000 droplets per second). Droplets can encapsulate cells, be fused, injected, mixed, incubated and even sorted using dedicated processing modules. However, this toolbox lacks a high-throughput module for droplet purification or enrichment of target molecules. This hinders droplet-based single cell analysis of mRNAs because the cell lysate concentration in a typical microfluidic droplet is high enough to inhibit the reverse-transcription reaction, an essential step of the workflow. Our lab developed a module to purify mRNAs from droplets using magnetic microparticles surface-functionalized to capture mRNAs. Beads are marginalized within the droplet before their enrichment via droplet splitting (Fig. 1a). However, the module throughput was limited. My first thrust consists of understanding the interplay between the magnetic and hydrodynamic forces that limits the module throughput.

In our magnetic purification module, there is an antagonistic relationship between throughput and purification efficiency. Beads aggregate under magnetic attraction (Fig. 1b) but are dispersed by viscous drag from the droplet’s internal recirculation. Dispersed aggregates end up in waste droplets, lowering purification efficiency. Higher throughput increases the internal flow and thus the dispersion forces (Fig. 1c). To inform optimization of this module’s throughput and efficiency, I am investigating the relationship between magnetic attraction and viscous dispersion forces in flowing microfluidic droplets. I comprehensively map the aggregates’ characteristics to develop an explanatory physical model. From droplet images captured across a wide range of throughputs and droplet sizes, I extract aggregate geometry, position, and distribution data to build large datasets characterizing aggregate populations. I analyze these datasets to understand how aggregate characteristics evolve and develop a model to describe these observations. The new knowledge will drive the design of a droplet purification module with enhanced throughput.

Research thrust 2: High-resolution 3D microflow mapping

Understanding the interplay between the magnetic and hydrodynamic forces requires a detailed 3D map of the droplet’s internal recirculation flow. 3D microflow mapping techniques are typically either limited in depth resolution and/or require complicated and expensive experimental setups such as multiple cameras. To address these gaps, I am developing a high-resolution 3D flow mapping technique using only a single high-speed camera, common laboratory equipment, and open-source software. First, flow tracer particles are located and tracked in the XY plane across video frames. Second, their Z-depth is obtained by comparing their out-of-focus diffraction patterns against reference images of known depth (Fig. 2). I perform depth classification via cross-correlation and deep learning. Deep learning is a promising approach because determining particle depth is an image classification problem. Through a collaboration with the lab of Prof. Jun Kong at Georgia State University, I investigate deep learning for improved classification accuracy over cross-correlation. I apply this technique to map the internal recirculation of microdroplets and correlate the map to Chapter 1’s results. The novel technique will prove broadly useful as a development tool in microfluidics, which increasingly uses engineered 3D flow patterns.

 

Fig. 2: (a) Particles located in XY via purple “spots”. (b) Particle images extracted and diffraction patterns Z-classified vs. reference images

Research thrust 3: Magnetic manipulation of hydrogel beads enmeshed with single-cell DNA/RNA

Single-cell genomics relies on barcodes, unique pieces of DNA that are appended to the genetic material of single cells before sequencing. Barcodes identify the cell of origin and are typically delivered by microbeads surface-functionalized with DNA pieces generated by a split-pooling method. Thus, current strategies rely on cell-bead coupling, which is highly inefficient and renders the technique expensive. My PI collaborated with Prof. Wigler at Cold Spring Harbor Laboratory (CSHL) to design an alternative workflow. Here, single cells are first transformed into individual Balls of Acrylamide Gel (BAGs) using droplet microfluidics. The single cell DNA/RNA is bound to the hydrogel matrix. Barcoding is performed directly on the BAGs by split-pooling. My second thrust builds on my experience with magnetic particles to magnetize single-cell-laden hydrogel beads to facilitate their manipulation and isolation.

Split-pooling involves several cycles of the following steps: 1) split beads across a 96-well plate, 2) add a unique tag by ligation, 3) pool beads. n such cycles yield 96n unique barcodes and enable large-scale, albeit very labor-intensive, barcoding. The magnetic manipulation of BAGs would greatly reduce the labor cost of split-pooling; therefore, I will investigate magnetizing BAGs through physical and chemical trapping of magnetic microparticles within the gel matrix. First, I will modify the polymerization parameters to physically trap microparticles by reducing the mean pore size and increasing the stiffness of the hydrogel. Second, I will use linker molecules to chemically bind beads to the matrix via surface functional groups. This novel functionality will greatly reduce the labor associated with the BAG approach and will support its widespread adoption.