The mechanisms that underlie enforced transitions between mature cell lineages are poorly understood. Profiling single skin cells that are induced to become neurons reveals that, unexpectedly, they often become muscle. See Letter p.391
Differentiated cells maintain their identity after development, ensuring specialized tissue function throughout adult life. However, they can be experimentally forced to change their identity — for instance, skin cells called fibroblasts can be reprogrammed to become more-primitive, embryonic-like stem cells1, or transdifferentiated into other specialized cell types such as muscle2, blood3 or neural cells4. These techniques are invaluable for studying cell plasticity and hold promise as possible tools for treating degenerative diseases, but the process is typically slow and inefficient. On page 391, Treutlein et al.5 address these shortcomings by presenting a molecular road map of fibroblasts as they convert to neurons, and provide intriguing evidence that such neuronal transdifferentiation often entails an unanticipated tug-of-war between alternative outcomes.
Reprogramming and transdifferentiation experiments typically involve overexpression of regulatory transcription factors that bind to DNA and induce gene-expression patterns characteristic of a specific cell type6. One of the groups that carried out the current study previously identified4 three brain-specific transcription factors — Brn2, Ascl1 and Myt1l, collectively known as BAM factors — whose overexpression in vitro converts fibroblasts into cells that resemble brain-derived neurons. This group also demonstrated7 that overexpression of Ascl1 alone can induce transdifferentiation into neuron-like cells, albeit with lower efficiency than the BAM cocktail. With either approach, most cells resist transdifferentiation, but the reasons for this remain unclear.
Single-cell RNA sequencing is a useful tool for assessing global gene-expression patterns in rare cell types within mixed cell populations8. Treutlein et al. apply this technology to neuronal transdifferentiation mediated by either BAM or Ascl1 alone. They measure total RNA levels in 405 single cells after 0, 2, 5 and 22 days of culture. They then use sophisticated computational tools to visualize data obtained from the cultures as a two-dimensional representation and thus reconstruct the progression of the fibroblasts into neurons over time.
These experiments reveal that transdifferentiating fibroblasts transit through two discernible stages, which the authors dub initiation and maturation (Fig. 1). During initiation, the cells cease to express genes that are characteristic of fibroblasts, stop proliferating and transiently activate genes whose expression marks neuronal progenitor cells. These changes take place in most cells and are orchestrated by Ascl1. By contrast, only a subset of cells progresses to maturation. This phase is characterized by the activation of genes that establish, and subsequently maintain, a mature neuronal lineage, and involves all three BAM factors. Thus, the transition from initiation to maturation represents a bottleneck during fibroblast-to-neuron conversion, correlating with the low efficiency of transdifferentiation.
One plausible explanation for this bottleneck is that rare cell types in the mixed cultures are uniquely susceptible to transdifferentiation. This would predict the existence of a transcriptionally distinct subset of fibroblasts. However, when examining gene expression in 73 individual fibroblasts, Treutlein et al. find no evidence for distinct subsets, and cultures seem remarkably homogeneous, making this possibility unlikely.
Another possibility is that the viral vector used by the authors to deliver the transcription factors into cells is silenced. To address this point, Treutlein et al. compare the expression of the introduced Ascl1 transgene in dozens of single cells. Although Ascl1 is initially expressed in most fibroblasts, the transgene is frequently silenced as cells transition to the maturation phase. These data suggest that viral silencing accounts, at least in part, for the low efficiency of neuronal transdifferentiation.
Perhaps the most surprising result emerges from a comparison of gene-expression data in single cells that activate the neuronal marker gene Tau after receiving either Ascl1 alone or the BAM factors. This analysis reveals that cells that receive only Ascl1 adopt a muscle-like (myogenic) gene-expression program despite activating Tau, whereas Tau-expressing cells that received BAM assume the expected neuronal fate. Unfortunately, the frequency with which Ascl1 expression alone gives rise to bona fide neurons and the functionality of the myogenic cells remain unclear. Nonetheless, these data suggest that Myt1l and Brn2 not only promote neuronal identity, but also prevent acquisition of a competing myogenic fate (Fig. 1).
Treutlein and colleagues' study underscores the power of single-cell technology to deconstruct complex cell-fate transitions across different lineages and time. Key questions that remain are why and how the Ascl1 transgene is silenced in fibroblasts, and whether an understanding of this inhibition would help to reveal the mechanisms that normally safeguard cell identity. Of note, recent data9,10 suggest that the mechanisms that facilitate viral-transgene silencing also act as barriers to reprogramming and transdifferentiation, suggesting a functional connection between these processes.
Another unresolved question is how Ascl1 supports transdifferentiation into myogenic cells, given that it is not normally expressed in muscle. Ascl1 belongs to a class of 'pioneer' factors11 that can associate with and regulate dormant regions of the genome that are inaccessible to other transcription factors. As such, its forced expression in fibroblasts might trigger nonspecific binding to hundreds of DNA sequences, including inactive muscle genes. Alternatively, Ascl1 might partner with another transcription factor that redirects it to muscle genes. A potential candidate partner is the protein MEF2, which is highly expressed both in fibroblasts and during muscle development, and has been shown12 to interact with Ascl1.
Regardless of the mechanisms that give rise to myogenic cells, Treutlein and colleagues' results have several practical implications. For example, the data suggest that identifying transdifferentiated neurons using a single marker such as Tau is insufficient, because the gene is also activated in cells that acquire a myogenic fate. The authors' findings further imply that the general trend of reducing the set of transdifferentiation-inducing transcriptional regulators to a minimum, as was the case here and in previous studies, may have unintended consequences.
This is particularly relevant in a therapeutic setting, in which it is crucial to transplant well-defined, homogeneous cell populations. A case in point is the finding13 that liver-like cells transdifferentiated from fibroblasts seem to be more similar to progenitors of the intestinal lining than to liver cells, and accordingly give rise to functional intestinal cells in mice. It is tempting to speculate that removing, adding or exchanging transcription factors from other transdifferentiation cocktails, combined with a more in-depth molecular analysis of the generated cell types, might uncover other similar shifts in cell fate or maturity.