Past,Present,and Future of Direct Cell Reprogramming

Introduction

Cellular identity changes occur naturally during development and in adult organisms in response to stress, injury, and disease (Rajagopal and Stanger, 2016). For instance in cancers, where it is common that cells change to a less differentiated state, a process sometimes referred to as dedifferentiation. Change of cell fate is most prominent during normal differentiation in various stem cell niches in the embryo and adult tissues and during the remarkable regeneration in response to injury in lower vertebrates such as fish and salamanders.

For a period, it was also believed that bone marrow-derived cells and certain stem and progenitor cells could give rise to cells of other lineages, for instance neurons and cardiomyocytes (Mezey et al., 2000; Orlic et al., 2001), in a process termed transdifferentiation. This, however, turned out, for the vast majority of cases, to be mediated by cell fusion events (Nygren et al., 2004; Wagers et al., 2002).

Under normal physiological conditions in adult mammal cell fate changes, outside of stem cell harboring or spontaneously regenerating organs are, however, rather rare events. Nonetheless, we now know that somatic cells retain cellular plasticity (Blau et al., 1985) that can be experimentally manipulated.

The nomenclature around cellular plasticity is somewhat blurred and researchers use terms such as dedifferentiation, transdifferentiation, conversion, and reprogramming, sometimes with prefixes such as nuclear, somatic, direct, and/or lineage, interchangeably and with their own definitions (Mills et al., 2019). Nuclear reprogramming refers to the process where the environment in enucleated oocytes or pluripotent cells reprogram the epigenetic profile of a donor nucleus. Transdifferentiation, lineage conversion, and direct reprogramming are commonly used to describe the same phenomenon where a cell changes fate without passing through a progenitor population, while dedifferentiation describes the reversion to an earlier developmental stage, sometimes in the cancer field also referred to as an undifferentiated state. In addition, newer terms such as forward programming or transcription factor programming have appeared to describe the forced or instructed differentiation of stem and progenitor populations using lineage transcription factor activation.

For the purpose of this perspective, we define direct cellular reprogramming as the experimentally induced in vitro or in vivo change from one cell fate to another, within or across lineages, without the passage through a pluripotent stage.

In this perspective, we outline the history of direct cell reprogramming, highlight ongoing research, and discuss the challenges and future developments in this fast paced area. This text is by no means intended as a complete assessment of the field but rather an overview, highlighting some of the key findings, implications, and how we see the future.

We make our inferences mainly through the lens of reprogramming toward neural lineages, which have parallels in other lineages.

Early Days

In 2012, Sir John B. Gurdon and Shinya Yamanaka received the Nobel prize in physiology or medicine “for the discovery that mature cells can be reprogrammed to become pluripotent.”

The traditional view of cellular differentiation postulates a gradual loss of potency along development. Popularly visualized by the Waddington landscape (Waddington, 1957), where a marble on the top of a hill represents a pluripotent cell. As development begins, the marble rolls down the hill and encounters bifurcations or valleys where choices are made to what lineage and cell fate to commit. At the bottom of the hill the marble rests in a valley in a fully committed cell fate. Because of gravity, the marble cannot go back up the hill or into another valley and the fate of that cell is irreversible set.

This notion was, however, challenged by early experiments using somatic cell nuclear transfer. Results from these studies showed that nuclei from later developmental stages or even adult animals transplanted to an enucleated oocyte could sustain pluripotency and differentiation potential to give rise to adult animals (Briggs and King, 1952; Gurdon et al., 1958; Gurdon, 1962). Thereby, the nuclei had become reprogrammed, hence the term nuclear reprogramming. These seminal findings taught us that it is not permanent changes to the cell or genome that restricts differentiation potential during development and that the environment in oocytes can reinstate pluripotency.

These findings together with studies of cell fusion between somatic cells and teratocarcinoma cells showing that hybrid cells displayed varying levels of pluripotency (Miller and Ruddle, 1976) eventually led to cloning of Dolly the sheep (Campbell et al., 1996).

Collectively, these studies implied that environmental and epigenetic mechanisms control differentiation and that this process could be manipulated.

Supporting this idea, treatment with a demethylating agent, 5-azacytidine, and thereby altering gene expression induced differentiation of fibroblasts to adipocytes and myoblasts (Taylor and Jones, 1979). Later sets of studies showed that transcription factors could drive transdifferentiation and identified that a single transcription factor, MyoD, was mainly responsible for and sufficient to induce a myocyte fate (Davis et al., 1987; Weintraub et al., 1989).

Additional studies further hinted at the power of lineage-specific transcription factors in altering cell fate. For instance GATA1 (Kulessa et al., 1995) and C/EBP alone or together with PU.1 (Laiosa et al., 2006; Xie et al., 2004) could reprogram cells within the hematopoietic lineage and a combination of Neurog3, Pdx1, and Mafa reprogrammed exocrine cells to insulin secreting cells in the adult mouse pancreas (Zhou et al., 2008). Similarly, astrocytes could be coaxed to generate neurons by overexpression of Pax6, Ngn2, or Ascl1 (Berninger et al., 2007; Heins et al., 2002).

Decades after the initial findings, Yamanaka and Takahashi presented their seminal findings that only a handful of transcription factors, Oct3/4, Sox2, Klf4, and c-Myc, enriched in embryonic stem (ES) cells could reprogram mouse (Takahashi and Yamanaka, 2006) and human (Takahashi et al., 2007) fibroblasts into induced pluripotent stem cells (iPSCs). These findings were indeed paradigm shifting and opened completely new avenues for regenerative medicine, as it made human pluripotent stem cells and differentiated cell derivatives, readily available to the research community. Previously, access was restricted to few human ES (hES) cell lines and to those with appropriate technology and with stringent ethical oversight.

A Field Emerges

Discovery of transcription factor mediated reprogramming to pluripotency reignited and kick-started the field of direct cell reprogramming.

The initial studies had only managed to reprogram between closely related, or cells within the same germ lineage. However, just a few years after the discovery of iPS technology, it was shown that mouse fibroblasts of mesodermal lineage could be reprogrammed to neurons of ectodermal fate by a cocktail of Brn2, Ascl1, and Myt1l transcription factors (Vierbuchen et al., 2010) and the addition of NeuroD1 allowed for reprogramming also of human fibroblasts (Pang et al., 2011).

In parallel, progress was made in generating cardiomyocytes from cardiac fibroblasts using a combination of Gata4, Mef2c, and Tbx5 in vitro (Ieda et al., 2010) and in vivo (Qian et al., 2012; Song et al., 2012).

It is perhaps not surprising that neurons and cardiomyocytes were the first target cells for direct cellular reprogramming as the heart and brain are notoriously poor at regenerating, and at the same time, neurological and cardiovascular diseases are leading causes of disability and death.

Since the reawakening of the field, a wide variety of cells of all germ layers have been not only generated typically starting from mouse or human fibroblasts but also from other cell types. Many different types of somatic and various progenitor cells, for example but not limited to, hepatocytes, adipocytes, dendritic, and endothelial cells have been generated (Basu and Tiwari, 2021; Wang et al., 2021a). In the neural field, it is now possible to generate all main types of cells; excitatory and inhibitory as well as a number of subtype-specific neurons (Bocchi et al., 2022), oligodendroglial cells (Chanoumidou et al., 2021; Tanabe et al., 2022; Yang et al., 2013), astrocytes (Caiazzo et al., 2015; Quist et al., 2022), and also neural stem and progenitor cells (Lujan et al., 2012; Thier et al., 2012).

Although the main method used in reprogramming studies is the forced expression, using viral vectors, of lineage-specific transcription factors, a variety of other approaches have also been used. Use of miRNAs, small molecules, modified mRNA, chemicals, siRNAs, and CRISPR-mediated gene activation has been applied either alone or in combination.

While we most likely do not know all the cellular changes that occur during direct reprogramming between different fates, it at least involves the activation and establishment of a core transcriptional program of the target cell and a loss of the starting cell program (Wang et al., 2021a). How extensive the epigenetic remodeling is that occurs during direct reprogramming is not entirely clear, but most likely, it is not as extensive as when reprogramming to pluripotency. Direct reprogramming also in many cases involves a change in metabolism, which can act as a barrier to efficient reprogramming, such as the shift from glycolysis to oxidative phosphorylation during neuronal reprogramming (Gascón et al., 2016).

Direct cell reprogramming have made cell types previously hard to access readily available for studies and opened up for many different applications.

Screening for transcription factors able to reprogram cells also informs about their role in normal differentiation and has become a useful complementary tool in developmental biology. Studying these transcription factors have revealed pioneering activity for factors such as Ascl1, Foxa3, and GATA4 (Wang et al., 2021a), and novel roles such as the promiscuity in fate determination by MyoD (Lee et al., 2020) and the ability of Myt1l to suppress multiple lineages except the neuronal during reprogramming of fibroblasts to neurons as well as during normal neurogenesis (Mall et al., 2017).

The possibility to access human patient and disease-specific cells has opened up for a revolution in disease modeling and drug discovery. It has become clear that animal models are not ideal for studies of many diseases. Numerous mouse models do not fully recapitulate the human diseases they were designed for. There are cellular traits that are human specific and not conserved between species. Furthermore, rodent models do not age or develop age-related diseases in the same way as humans. In contrast, directly reprogrammed cells retain several age-related phenotypes and have been used successfully to model several neurological disorders (Legault et al., 2022), including age-related neurodegenerative diseases such as Alzheimer's (Mertens et al., 2021; Traxler et al., 2022) and Huntington's disease (Pircs et al., 2021).

Direct reprogramming also offers a mean for generating cells for cell replacement therapy in degenerative diseases and could in the future be a faster way, to obtain patient-specific cells for transplantation purposes, than first reprogramming to iPSCs followed by traditional differentiation. However, so far, cells generated from pluripotent stem cells are superior in terms of functional maturity and, therefore, seem a more logic source. With more research, perhaps directly reprogrammed cells will become a contender source for transplantation in degenerative diseases. Nonetheless, direct cell reprogramming have already inspired several protocols for forward programming, where transcription factors, often the same as those used for direct reprogramming, are used to efficiently drive differentiation of pluripotent stem cells into desired cell fates (Canals et al., 2018, 2021; Garcia-Leon et al., 2018; Ng et al., 2021; Yang et al., 2017; Zhang et al., 2013).

As was shown early on (Qian et al., 2012; Torper et al., 2013) and which has recently gained a lot of attention, direct reprogramming can also be achieved in vivo. The idea of repairing organs that normally do not readily regenerate such as the heart or brain by reprogramming abundant or spontaneously replenishing support cells, that is, cardiac fibroblasts in the heart or glial cells in the brain, is mind boggling. The possibility to reprogram cells in vivo represents a powerful opportunity for regenerative medicine applications.

In vivo reprogramming could potentially be utilized as a therapy not only for trauma and degenerating disorders but also for reprogramming of cancer cells into less aggressive cells or cells more susceptible to conventional therapy is an attractive approach that has gained some traction (Fehrenbach et al., 2016; Rapino et al., 2013; Su et al., 2014).

The Road Ahead

The field has progressed rapidly in a relatively short amount of time, but additional methods and protocols to generate the multitude number of cells in the body from various other cell types are needed and will move the field forward. An immediate contribution for the direct cell reprogramming studies is the identification of a minimal combination of transcription factors that has a critical instructive role in that lineage. Ideally, we will in the future have such minimal combinations for all the different cell types in the body. For instance, several different types of neuronal cells such as glutamatergic, GABAergic, serotonergic, and dopaminergic cells have been reported. However, far from all, the different subtypes and regionally specified neurons have been generated, but will be needed for accurate disease modeling and potential future therapeutic purposes.

As more methods to generate different types of cells emerge, there are also opportunities to create more complex models containing multiple cell types to study their interaction in homeostatic and disease conditions. For instance, cocultured neurons and astrocytes (Quist et al., 2022) and, eventually also, microglia generated by direct reprogramming could be used to study neuron-glia communication in neurodegenerative disorders such as Alzheimer's disease and ALS, as had been previously done using iPSCs (Guttikonda et al., 2021).

A lot of attention is given to the efficiency of various protocols, in generating the target cell of interest. Direct cell reprogramming is already more efficient than reprogramming to iPSCs, and therefore, efficiency might not necessarily be that important to improve, especially for in vitro applications or if the target cell can be isolated by, for instance, fluorescence-activated cell sorting (FACS). However, direct reprogramming often starts with cell populations that cannot be indefinitely propagated in vitro and yields postmitotic cells or cells with limited proliferation capacity, with the exception of certain stem and progenitor cells, which also have been generated by direct reprogramming. In addition, efficiency of direct reprogramming of adult human cells and reprogramming in vivo in mouse models remain for the most cases rather modest, limiting the potential for translation and needs to be improved.

For both in vitro and in vivo applications, identity and functionality of induced cells is critical. So far, hES/iPSC-derived cells are superior in terms of functional maturity. It will be important to keep a stringent definition of successful reprogramming, both in terms of efficiency and fidelity (similarity to target cells). A number of criteria should be fulfilled. Most studies so far have used a combination of morphological, immunophenotypic, transcriptomic, and functional assays to claim successful reprogramming. As the field moves forward, these minimal criteria should be kept, revised, and strengthened. It will be necessary to benchmark directly reprogrammed cells to primary or PSC-derived cells and ideally to the bona fide in vivo counterpart whenever possible.

Comparison to cells in vivo is sometimes difficult and in certain cases, especially when it comes to human cells, not possible with available tools. However, such validation will be essential before directly reprogrammed cells could be considered for clinical translation.

Transplantation to animal models (Tanabe et al., 2022), integration into human PSC-derived organoids, or ex vivo tissue culture systems (Miskinyte et al., 2017), followed by careful assessment of functional integration, could serve as a surrogate and should help move the field closer to successful translation.

Directly reprogrammed cells will also have to pass safety concerns. Tumorigenic potential of directly reprogrammed cells is substantially lower as compared to PSC-derived cells. Nonetheless, concerns regarding insertional mutagenesis when using integrating viral vectors have to be addressed and will facilitate development of alternative safe delivery methods of reprogramming factors.

The possibility to reprogram cells in vivo has enormous potential for regenerative medicine and could potentially be used for degenerative diseases or trauma in organs such as the heart and brain that do not regenerate efficiently. In addition, generating immune cells for intratumoral transplantation or reprogramming cancer cells in situ to antigen presenting cells, able to present novel tumor antigens, could enable novel immunotherapeutic approaches (Zimmermannova et al., 2021).

Previous studies have shown that reprogramming can indeed be performed in vivo (Bocchi et al., 2022). However, there is ongoing controversy in the field when it comes to in vivo reprogramming of glial cells to neurons in the brain, where several studies have failed to reproduce some extraordinary findings (Wang and Zhang, 2022). Supposedly reprogrammed neurons have been shown to be endogenous neurons falsely labeled by suboptimal viral reporters. It has become clear that reprogramming transgenes can induce promoter activation resulting in false-positive reporting, which complicates lineage tracing using viral vectors alone (Wang et al., 2021b).

As the field moves forward, it will be important to reproduce the previous and design new studies using rigorous genetic lineage tracing and prelabeling of starting as well as target cells. Live imaging, whenever possible, and time course immunohistochemistry should help visualizing the process of reprogramming capturing intermediate stages and not only the intended final target. In addition, scRNAseq at different time points during reprogramming will increase confidence that reprogramming have occurred.

One of the promises of directly reprogrammed cells for use in disease modeling is the retention of age-associated signatures. Unlike cells derived from iPSCs, which are rejuvenated, studies have shown that, for instance, directly reprogrammed neurons and oligodendroglial cells retain the relative age of the starting cell population (Chanoumidou et al., 2021; Mertens et al., 2015). This property of directly reprogrammed cells might help the discovery of age-related disease mechanisms, which is substantially more difficult when working with iPSCs. However, this also poses a challenge for the use of directly reprogrammed cells for regenerative approaches for age-related diseases. It remains to be answered whether all directly reprogrammed cells retain features of aging and this probably will depend on the factors used.

Notably, some of the Yamanaka factors are also used in protocols for direct reprogramming. For instance, neural stem cells generated by reprogramming using Sox2 and c-MYC showed signs of rejuvenation, but without passing through a detectable pluripotent stage (Sheng et al., 2018).

The recent development of single cell omics and aging clocks based on epigenetic and transcriptomic data (Noroozi et al., 2021) will help answering whether retention of aging signatures is dependent on reprogramming factors, starting and target cells, or a combination of all these aspects.

Concluding Remarks

Recent progress has showed that it is indeed possible to generate a wide variety of cells in vitro and in vivo through direct cellular reprogramming. There are, in theory, no boundaries to what type of cells that could be generated, and further pursuit along those routes should be encouraged.

However, it is also important to take caution when defining successful reprogramming keeping criteria and validation strict, especially for human cells and in vivo applications. Careful phenotypic and functional comparison to the natural target cell and increasing functional maturation of directly reprogrammed cells is important both for disease modeling and potential therapeutic applications. Furthermore, development of safe in vivo delivery systems for reprogramming factors is essential for translation.

The future for direct cellular reprogramming is bright and the research community will hopefully achieve significant progress toward clinical translation in years to come.