Where a cell comes from—its lineage—can be critical to its function, but few methods allow us to study the effect of this intrinsic property on cell state or behavior. At the Massachusetts Institute of Technology, Scott Manalis and his graduate student Robert Kimmerling began thinking about the problem by chance. The Manalis lab excels at measuring the properties of single cells, and the researchers were working on long-term culture to measure cell growth. While prototyping a microfluidic device using arrays of hydrodynamic traps, “we ran into the idea of tracking this with time-lapse imaging,” says Kimmerling. “We could actually reconstruct lineages.”

The lineage project came to life in the face of a biological question. Activated CD8+ T cells give rise to both effectors that clear an infection and memory cells that persist to fight subsequent infections. How this happens is “hotly debated,” says Kimmerling. According to competing models, the two subtypes could arise immediately from an early asymmetric division in the activated cell's lineage, or a memory phenotype might develop gradually over time.

To address such questions, Manalis, Kimmerling and their colleagues teamed up with the lab of Alex Shalek to link lineage with gene expression data across cell dynasties. They designed a device made of winding channels studded with hydrodynamic traps, pockets that confine cells on the basis of differential fluid flow. After division, displaced daughter cells flow into the next available trap, while the process is tracked with time-lapse imaging. In designing the chip, the researchers had to find a sweet spot between high-efficiency capture and individual cell release for sequencing, as fluidic resistance is the same in both directions. Controlled release with minimal shear stress takes finesse—“Rob has to watch the cells and ... turn the knobs in such a way to catch them,” says Manalis. They are working to automate the process.

The researchers cultured lymphocytic leukemia cells and primary CD8+ T cells from mice for two generations on the chip and then analyzed gene expression from individual cells. Both cell types showed less transcriptional variation within clonally related cells than between clones. This was true even for genes such as granzyme B, which is involved in targeted killing by cytotoxic T cells. “We were looking at the list of biologically relevant genes, ones that were related to T cell function, and we still saw this distinct lineage effect,” says Kimmerling.

They also measured the time since cell division and used a regression model to find the genes that differed most with respect to this variable. The top 300 genes were very different for the two cell types, providing potential insight into cell cycle drivers.

Each module on the chip currently contains 40 traps to study up to five generations, and cells must be somewhat uniform in size and grow without anchorage. Adherent cells, including human cancer cells, have been grown on the device for up to eight days. The researchers are also coupling serial suspended microchannel resonators with the platform, enabling the measurement of cell mass accumulation. Other measures are in the works. “Taking all physical things we can measure about the cell and linking that to RNA-seq—it's exciting,” says Manalis.

Although the work is in vitro, Kimmerling and Manalis believe that the highly controlled environment offers advantages; “We are collecting a set of blueprints for what things are intrinsic to the cell and not dependent on the microenvironment,” says Kimmerling. Insights from the chip could inform models of in vivo data and might ultimately help solve the question of CD8+ T cell differentiation.