One of the most well known and widely used cell lines originated from the eponymous Henrietta Lacks. However HeLa cells are just one of a multitude of cell lines that have since been generated and cultured, providing a vital resource for scientific progress and often forming the foundation of preclinical studies. A clear understanding of the genetic makeup of these cell lines is therefore essential, particularly in light of the oft made assumption that these lines remain clonal. Mutagenesis and transformation are just two of the processes that can destabilise the genomes of these cells over time, leading to experimental variability, issues around reproducibility, and hampering attempts at building upon findings from these cells. In a Genome Biology study, Matthew Porteus from Stanford Medical Center, USA, and colleagues present a cellular barcoding method for tracking the clonal dynamics of cultured cells, in the hopes that adoption of this technique will improve experimental design and the interpretation of results. Here Porteus reveals what they discovered in HeLa, HEK-293T, and K562 cells, the particular utility of this method in stem cell and cancer biology, and the scope for using ZFNs and CRISPRs.
What motivated you to develop a reproducible means of tracking the clonal dynamics of cultured cells within a population?
We were initially motivated by the observation that in clinical gene therapy trials using retroviral vectors, the preclinical assays had not identified the risk of leukaemia that was discovered during the clinical trials. We felt that we needed an assay that could pick up small proliferative changes in individual clones over a prolonged period of time and thought that a barcode marking system that labelled tens of thousands to millions of cells might be such an assay. As we developed the assay, however, we recognised that it had tremendous power in revealing the dynamics of populations of cells in general.
The problem of cell lines changing with multiple passages is well known, however your study now brings quantitative insights to the scale of this problem. Were you surprised by any of your results?
We were pleased that we provided a quantitative measure for what has been well known qualitatively for some time. I don’t think we were particularly surprised by the extent or rate that cell lines change over time in vitro. I was personally surprised, however, that while there were certainly large clonal changes in the xenogeneic experiments, the total clone number was better maintained in vivo than in vitro. I expected that the selective pressure of the mouse would be greater than that found in a tissue culture dish with cells grown in plastic in artificial media at 21 percent oxygen. I guess as I write this, I should not have been surprised.
I was also relatively surprised that the same clone became dominant in both the in vivo and in vitro experiments as I expected that each environment would exert unique selective pressures that would result in different clones becoming dominant. That may turn out to be true as this system is used to study other human tumours. The barcode system will also be a quantitative measure of how much better it is to put tumours into their appropriate organ in a xenogeneic model. That is, is there a difference when a primary lung cancer, for example, is grown subcutaneously rather than propagated through the lungs of a mouse.
Your study largely used a lentiviral approach to barcode cells at random sites, however you also adapted this method to use zinc finger nucleases (ZFNs). Why did you do this and how did the ZFN approach affect clonal dynamics?
My lab has had a long standing interest in using engineered nucleases as a method to stimulate homologous recombination and as an approach to gene therapy. One of the assumptions of this strategy is that the functional effects on clonal dynamics of targeted integration of transgenes would be less using engineered nucleases than with lentiviral integrations. We were surprised to find the opposite, that integration using ZFNs created greater clonal dynamics with higher clonal dropout and more clonal dominance. With lentiviral integration in K562 cells, for example, we did not see clones reproducibly becoming dominant in the population, suggesting that there was a stochastic effect to that phenomenon. In contrast, using the ZFN mediated targeted integration we saw many of the same clones growing out in different samples suggesting that there was a heritable and non-stochastic event that gave those clones a proliferative advantage.
Are there plans to adapt your method to CRISPR as well? Would you expect CRISPR to impact the cell population in a similar way as ZFN?
We are in the active process of comparing ZFNs, TALENs, and CRISPRs for their effects on clonal dynamics. My expectation is that CRISPRs will show less clonal dropout and clonal dominance than ZFNs but the barcode system has continually surprised me and may do so again. One of the important features of this system, is that it is a functional assay. Significant discussion has been going on about identifying the off-target sites for different nucleases. These studies have been and will continue to be important in understanding aspects of nuclease specificity. But it is probably impossible to identify all off-target sites; for example, those sites at which the nuclease only cuts at 1:10000 the frequency as its on-target site. So as we think about the safety of using cells clinically that have been modified using engineered nucleases, it is important to develop functional assays that have a high degree of sensitivity – how does the modification process affect the way cells behave and can you identify processes that might only create deleterious events at low frequencies? We believe that the barcode system is an important step in that direction.
Do you see scope for combining your cellular barcoding approach with single-cell sequencing, in order to further probe the nature of cellular heterogeneity?
One of the questions that we regularly get asked and regularly talk about is what is causing the cellular heterogeneity. We know that there are ongoing genetic changes in the cell and agree that using single cell sequencing may help determine if the functional heterogeneity we see is the result of changes in the nucleotide composition of the genome (single base pair mutations, copy number changes, gross chromosomal rearrangements). The relatively rapid pace of diversity we observed when we started with a single cell/clone, however, suggests that in these cell lines epigenetic variations may be a more important contributor. Thus, technologies that are able to interrogate the epigenome at a single cell level, not just sequencing, will have powerful synergy with the barcoding system.
The implications of your study stretch across all disciplines using cell lines, however do you see any specific areas of research where this method would be particularly useful?
We agree that the utility of barcoding goes beyond the analysis of cell lines and are beginning to pursue some of these ourselves. As mentioned before, we are using this system to explore the functional effects of different nuclease platforms. Systems similar to this have already been used to track the reconstitution of the haematopoietic system following transplantation. In the future, I would expect that such systems will be used to understand the clonal dynamics in expansion of primary cells, including primary stem cells. It is possible, and has important ramifications, that when primary cells expand in vitro that the process ends up selecting just a handful clones over time. I would also expect that this barcode system will be a powerful tool in cancer biology: what is the clonal composition of primary tumours? what is the clonal composition after recurrence? what is the clonal composition of metastases? what is the clonal composition as it relates to the development of resistance to standard cytotoxic chemotherapy and to targeted therapies?
Finally, the barcode system will be a powerful approach to understanding how cells might behave differently when they are isolated and alone as compared when they are part of a large population (either seemingly homogeneous or overtly heterogeneous). By barcode marking, one can track how different, seemingly identical, clones respond to the signals in their environment. And given the cleverness and innovativeness of researchers, I expect that the barcode system will be used in many other ways in the future.
The validity of findings based exclusively on cultured cell lines has been subject to debate, raising the question of whether research should be so reliant on these cell lines. In light of your recent findings, where do you stand?
Cell lines continue to serve an incredibly valuable purpose and will continue to do so. But our studies do caution about over-interpreting any result that is based on experiments in cell lines. Our results should certainly serve as a strong reminder that the cell line one is using may have little relevance to the process one thinks one might be studying because these lines change so rapidly. So one model would be that cell lines serve an important role in exploration and discovery and in some proof-of-concept studies, but that critical validation in primary cells and in vivo is essential. Along similar lines, for lines of experiments that are focused on potential medical applications in humans, a model could be that any potential discovery made in mouse or non-human systems is validated in a human system.