NEW YORK – Researchers at the University of Pennsylvania and the University of Delaware have developed a new method for identifying persistent fluctuations in gene expression across the transcriptome, and have found rare cell subpopulations that are associated with biologically distinct behaviors such as resistance to cancer drugs.
In a new study published on Thursday in Cell, the researchers noted that non-genetic factors can cause gene expression levels to fluctuate substantially in individual cells over time, but that it's unclear whether these fluctuations persist for longer than one cell division. In order to better measure these shifts, the investigators developed a method called MemorySeq, in which they combined established fluctuation analysis methods with population-based RNA sequencing and used it to identify genes across the transcriptome whose fluctuations persisted for several cell divisions.
They found multiple gene modules that were expressed together in rare cells within otherwise homogeneous clonal populations, and observed that these rare subpopulations had distinct behaviors.
"The identification of non-genetic, multigenerational fluctuations can reveal new forms of biological memory in single cells and suggests that non-genetic heritability of cellular state may be a quantitative property," the authors wrote.
In their experiment, the researchers grew 48 of what they called "MemorySeq clones" of isogenic melanoma cells in individual wells, eventually growing them to around 100,000 cells per clone. If a fluctuating gene transitioned in and out of the "high" expression state relatively rapidly compared to the cell division rate, then the investigators considered that a fairly constant proportion of those 100,000 cells would be in the high expression state for that gene.
At the opposite end, if the high expression state was long-lived compared to the cell division time and a cell occasionally moved into the high expression state early in the family tree, all of its progeny would remain in the that state, leading to a very high proportion of the final 100,000 cells being in the state.
Therefore, the researchers found that there was a high variance in the proportion of cells in the high expression state in the final population across multiple MemorySeq clones, where most clones would have low expression of that gene and a few clones would have high expression, depending on exactly how far up in the family trees the cells transitioned into the high expression state.
They then used bulk RNA sequencing to measure variability in the proportion of cells in the high expression state for any particular gene, by analyzing the transcription of all genes in each expanded clone and then measuring the variability in the expression of all genes across these clones.
The investigators first applied MemorySeq to the melanoma cell line WM989-A6, chosen because it contained rare cells expressing a particular subset of genes such as EGFR, NGFR, and AXL that were strongly associated with resistance to the targeted melanoma drug vemurafenib (Genetech's Zelboraf). When they analyzed expression variance across clones for all genes using MemorySeq, they were able to generate a panel of 227 genes as potentially having high heritability in WM989-A6.
The researchers were able to tag the NGFR gene and track it over a period of nearly nine days. As predicted, they found that occasional rare cells within the wider population displayed high levels of the gene. They then tracked 222 cell lineages through several cell divisions and found that cells would occasionally initiate high levels of expression of NGFR — once initiated, that high level of expression could be maintained through multiple cell divisions, thus confirming the presence of memory.
Importantly, cells expressing high levels of NGFR were much more likely to continue to proliferate when the researchers added vemurafenib to the cell lines.
The investigators noted that these rare subpopulations of cells could also be important in the context of healthy tissues or cellular reprogramming, such as the induction of induced pluripotent stem cells.
"The mechanisms underlying these fluctuations remain mysterious. We identified some transcription factors that can affect expression of these genes and found patterns of histone acetylation that are associated with these genes, as well," the authors concluded. "Further work will be required to test other potential models of gene regulation that could lead to fluctuations, such as methylation patterns or other regulatory mechanisms that may operate on the relevant timescales."