Paligenosis: Cellular Remodeling During Tissue Repair

Abstract

Complex multicellular organisms have evolved specific mechanisms to replenish cells in homeostasis and during repair. Here, we discuss how emerging technologies (e.g., single-cell RNA sequencing) challenge the concept that tissue renewal is fueled by unidirectional differentiation from a resident stem cell. We now understand that cell plasticity, i.e., cells adaptively changing differentiation state or identity, is a central tissue renewal mechanism. For example, mature cells can access an evolutionarily conserved program (paligenosis) to reenter the cell cycle and regenerate damaged tissue. Most tissues lack dedicated stem cells and rely on plasticity to regenerate lost cells. Plasticity benefits multicellular organisms, yet it also carries risks. For one, when long-lived cells undergo paligenotic, cyclical proliferation and redifferentiation, they can accumulate and propagate acquired mutations that activate oncogenes and increase the potential for developing cancer. Lastly, we propose a new framework for classifying patterns of cell proliferation in homeostasis and regeneration, with stem cells representing just one of the diverse methods that adult tissues employ.

1. INTRODUCTION

For most of the last half of the twentieth and early part of the twenty-first centuries, the canonical model for proliferation in adult tissue was the epigenetic landscape proposed initially by the geneticist Conrad Waddington (1), which implied that a stem cell is constantly present in every tissue. The stem cell self-renews and, when necessary, spawns differentiated cells in a unidirectional flow (Figure 1a). Such a professional stem cell is by definition self-replicative as well as multipotent (i.e., it can give rise to multiple cell types) and thus relatively undifferentiated (i.e., it lacks both specific markers attributable to differentiated cell types as well as specific mature cell architectural features like secretory granules).

Figure 1.

Schematic representation of homeostatic behavior and injury-induced cellular plasticity in the exocrine pancreas as well as epithelia of the gastric corpus and small intestine. (a) Illustration of cell fate decisions and plasticity within an adult tissue. (Left) Waddington landscape in which a stem cell at the top of the hill has the potential to roll down one of several grooves to a specific, differentiated state at the bottom of the landscape implying irreversible terminal differentiation. (Upper right) Examples of cell plasticity in adult tissue: Dedifferentiation (differentiated cells at the bottom to roll uphill) and transdifferentiation (differentiated cells located at a specific groove move to another one, overcoming the hurdle). (Lower right) Single-cell analyses such as RNA-seq suggest that cell state may be a fluid and not deterministic process as has been implied by traditional fixed tissues and histological approaches. (b) Homeostatic lineage differentiation from the professional stem cell (LGR5⁺ CBC) at homeostasis and injury-induced dedifferentiation. (c) Juxtaposition of the two proliferative zones within the gastric corpus at homeostasis and depiction of the divergent response to injury. (d) Scaled schematic representation demonstrating the dramatic differences in cellular differentiation between both different lineages and changes following injury. (Left panels) Schematic illustration of size-matched representative TEM images (right panels) of mitotic cells in the murine stomach corpus and pancreas at homeostasis and after injury: (i) isthmal stem cell (type I), (ii) mucous neck cell (type II), (iii) pancreas acinar cell (type III), and (iv) dedifferentiated gastric chief cell during paligenosis (type III). Colchicine was used to halt progression through mitosis and increase the likelihood of catching rare, homeostatic autoduplication events in TEM. Condensed chromosome (blue) and mitochondria (orange) are colorized. Right panels are the original TEM images of cells depicted in the left panels. Abbreviations: CBC, crypt base columnar cell; RNA-seq, RNA sequencing; TEM, transmission electron microscopy.

However, with tools like genetic and chemical lineage tracing and single-cell RNA sequencing (scRNA-seq), and with the advent of cellular reprogramming to generate induced pluripotent stem cells (2, 3), it has become clear that cell behavior in tissue can be far more complicated (see, e.g., 4, 5). For example, single-cell expression profiling shows that there might not even be clear grooves on the Waddington landscape, but instead, cell identity might be fluid, with the behavior and phenotype of a single differentiated cell being better modeled by a stochastic electron-cloud-like representation (Figure 1a). In other words, cell identity may be loosely constrained by a genetic/epigenetic envelope but perpetually restless in response to environmental, hormonal, and paracrine signals. We have likely missed such heterogeneity of adult cell identity because the techniques we traditionally employ to assess cell identity sample the cell at a single time point, using limited biomarkers. Moreover, in the setting of injury, the constraining envelope (i.e., Waddington groove) for a cell is perturbed, with mature cells able to access states not available during homeostasis. For example, they can undergo dedifferentiation to a stem cell–like state, or they can change phenotype to replace other adult cell lineages via transdifferentiation (6, 7) (Figure 1a).

An individual cell can be injured in numerous ways: chemical, thermal, inflammatory, hypoxic/ischemic, infectious, autoimmune, or physical or by starvation or being crushed. However, the fundamental response decision matrix is the same: (a) repair the damage (e.g., via induction and resolution of cell stress pathways like the unfolded protein response), (b) die (e.g., via apoptosis, pyroptosis, necroptosis, or anoikis, a noninjury-associated process), or (c) enter the cell cycle. Additionally, extracellular signals from immune cells and the mesenchyme can further modify cellular responses, such as coaxing cells to change identity (transdifferentiate) to replace other cell types lost by injury. Here, we focus not on extracellular signals but on the intrinsic aspects of a cell’s response either to its own injury or to injury in neighboring cells. Cell repair is the optimal choice if the insults are minor. Cell death benefits the organism if the cell itself is irreparably damaged, has acquired mutations that would predispose toward oncogenesis, or is harboring a pathogen. Cell proliferation may be necessary to regenerate tissue if other cells are dead or injured beyond repair. For example, after liver cell loss (e.g., chemical or partial hepatectomy) the liver must maintain metabolic and synthetic functions while it increases cell mass (7).

Most tissues do not actually have professional stem cells, in which case homeostatic proliferation is via mature cells autoduplicating: cytokinesis without significant prior loss or the degradation of differentiated features. Indeed, many nonvertebrate organisms may lack adult professional stem cells entirely. Thus, in most cases where cell proliferation is needed to regenerate tissue after injury, mature cells that are performing a differentiated function must be recruited. It is likely, given the universality of this tissue need, that mature cells have evolved a conserved program (akin to mitosis or apoptosis) that interprets stress as a signal to retool their differentiated cell architecture and reenter the cell cycle. We have termed this program paligenosis from pali/n/m (meaning backward or recurrence) + genea (born of, producing) + osis (an action or process).

Here, we discuss both homeostatic and postinjury reparative processes and primarily focus on paligenosis, contrasting paligenosis to other reparative processes and discussing how paligenosis might have evolved. We also discuss risks of endowing cells with the capacity to undergo paligenosis. Our focus centers on—but is not exclusive to—tissues of the gastrointestinal tract, where cross-tissue comparison of cellular behavior during homeostasis and following injury is relatively straightforward.

2. RECONSIDERING STEMNESS IN HOMEOSTASIS AND INJURY

The tissue that pioneered a Waddington, unidirectional hierarchy of adult stem cell behavior was arguably the hematopoietic system of vertebrates. At homeostasis, blood cells largely develop in the bone marrow from multipotent stem cells that spawn various lineages of differentiated cells via sequential fate choices, each of which restricts lineage options along Waddington differentiation grooves (8). There is also evidence in blood for another classical notion of professional stem cell behavior: that stem cells can remain undifferentiated but dormant for days, weeks, and months, activated only when needed (9–11). Such a notion that true, professional stem cells at the top of a multipotent hierarchy would be largely quiescent originated with early nucleotide analog tracing experiments (e.g., with ³H-thymidine) showing that some cells could take up label (i.e., proliferate, which, in the thinking of the era meant they must be stem or progenitor cells) and then retain it for weeks or months after chase without label (12). Outside the hematopoietic system, the quiescent satellite cells in skeletal muscle (13), and potentially the occasionally proliferative +4 reserve stem cell in the small intestine (12), little evidence has emerged for such so-called label-retaining stem cells. Rather, professional stem cells largely devote their daily activity to stochastic decisions about whether to proliferate or differentiate and are not quiescent at all (14).

Although the long-held notion of slow-cycling stem cells may actually apply only to blood at homeostasis, other tissues with professional stem cells do seem to adhere more or less to the unidirectional stem cell differentiation behavior illustrated by the Waddington landscape (1, 15) (Figure 1a,b). However, even in such stem cell–maintaining adult tissues, injury seems to scramble this behavior, with genetic lineage tracing studies clearly demonstrating that injury can induce diverse examples of cell plasticity. Moreover, most tissues or tissue compartments (and many entire organisms) lack professional stem cells and regenerate damaged tissue entirely by recruitment of mature cells into the cell cycle. In this section, we examine examples of how tissues self-maintain in homeostasis and injury in agreement or in contrast to Waddington stem cell dynamic models.

2.1. Dynamics of Professional Stem Cell Behavior at Homeostasis

Outside the hematopoietic system (and possibly skin), the tissue that has arguably been best studied in terms of its classical, hierarchical stem cell differentiation behavior is the small intestine, wherein the professional stem cell, the crypt base columnar cell (CBC), sits at the base of the intestinal crypt nestled between Paneth cells (16–18). This cell is most commonly identified by the biomarker LGR5, a cell surface receptor for R-spondin that augments WNT signaling (19). The CBC maintains itself by a stochastic mix of self-replication and generation of progeny (14) that go on to differentiate into all the principal intestinal lineages (16) (Figure 1b). Stem cells of the skin, including those of the epidermis and hair follicle, also largely follow classical differentiation at homeostasis (20, 21).

In the body of the stomach, the multipotent cell sits at the isthmus between the deeper glandular cells and those cells that rise to the surface of the gastric lumen (i.e., foveolar, surface, or pit) cells (Figure 1c). The isthmal stem cell seems to follow classical stem cell hierarchical dynamics at homeostasis: giving rise to multiple cell lineages in the epithelium (22). However, the digestive enzyme–secreting chief cells at the base seem to largely maintain their own population at homeostasis (23, 24) (Figure 1c).

2.2. Cellular Plasticity in Tissue with a Professional Stem Cell Following Injury

Injury induces cell plasticity, even in tissues with professional stem cells. For example, in the small intestine, there is a so-called +4 reserve stem cell population whose activity is induced by injury. The +4 cell has always been largely functionally defined, as it was previously the site of maximal label-retaining activity (12). More recently, it has been defined by lineage tracing that contrasts +4 cell behavior with CBC behavior. LGR5-negative, non-CBC stem cell activity localizes approximately to the +4 position using lineage tracing studies with various promoters, e.g., Bmi1 (25, 26), Hopx (27), mTERT (28), and Lrig1 (29, 30). Although there might be some homeostatic stem-like behavior of such +4 cells, they are likely at least partially differentiated, with most studies suggesting they are physiologically secretory cells that serve as reserve stem cells following injury (16, 31). However, their ability to be recruited stem cells is far from unique when radiation or chemical injury has compromised or overwhelmed the professional CBC stem cells. For example, Paneth cells (32, 33), enterocytes (34–38), enteroendocrine cells (39), and even tuft cell lineages demonstrate significant plasticity (40, 41) Any of these cells seem capable of ultimately replacing a lost CBC, meaning all intestinal lineages may have some capacity for complete dedifferentiation back to the multipotent state. In tissues with a professional stem cell, such cellular plasticity is necessary to restore the homeostatic cell hierarchy when the stem cells are lost. In the absence of such plasticity, ablation of the stem cell population would result in collapse and complete loss of these rapidly cycling tissues.

2.3. Cellular Plasticity in Tissues without Professional Stem Cell or with Mixed Stem Cell Activity at Homeostasis

The epithelium of the gastric body (corpus) has a mixed phenotype of stem cell activity (Figure 1c,d). The professional, isthmal stem cell can repopulate the pit (surface/foveolar), neck, and parietal cell lineages (23, 24). However, using elegant lineage tracing, Han et al. (23) have demonstrated that the chief cells are responsible for maintaining their own population at homeostasis. Our work using BrdU/EdU pulse-chase labeling both at homeostasis and during injury corroborates their study (24). Thus, at homeostasis, chief cell dynamics resemble those of acinar salivary gland cells (42), exocrine pancreatic acinar cells (43), and hepatocytes (44): cell populations that maintain their own census at homeostasis through autoduplication. The self-duplicating quality of acinar cells has not been extensively studied, but it may be that only subpopulations of cells have proliferative potential with either mutational burden or local concentrations of growth factors likely playing a role in promoting this behavior (45). Unlike induced proliferation following injury that is discussed below, it seems that the homeostatic autoduplication that differentiated cells undergo occurs without large-scale cell architecture remodeling or dedifferentiation.

2.4. Tissues without Professional Stem Cells after Injury

The cells in the parenchyma of most organs at homeostasis seem to be devoted to carrying out their differentiated function and are mitotically quiescent (46), so when injury is severe enough to merit proliferation to replace lost cells, there have to be mechanisms to license mature, differentiated cells into the cell cycle that do not involve professional stem cells. For example, other than the rare autoduplicating cells, the exocrine pancreas is largely mitotically quiescent. However, following injury that causes cell loss, the remaining cells demonstrate remarkable plasticity. Injury either transiently with cerulein, a cholecystokinin analog, or permanently, via pancreatic duct ligation, causes acinar cells to revert to a more embryonic-like phenotype, with proliferating ductular morphology characterized by a relative depletion of zymogenic granules and a more cuboidal morphology resembling ductal precursors (recently reviewed in 47). Because of the duct-like morphology and because acinar cells activate the expression of genes characteristic of mature ducts (e.g., CK19, SOX9), the changes that occur are known as acinar-toductal metaplasia (ADM). As a metaplasia, ADM cells are morphologically distinct from acinar cells at homeostasis, yet they are not dysplastic. However, the term acinar-to-ductal metaplasia is misleading, because ADM cells do not become mature pancreatic duct-like cells (e.g., a transdifferentiation event), as they continue to express genes normally present only in the acinar cells (e.g., amylase) (31).

The importance of ADM is underscored by lineage tracing experiments demonstrating that this metaplasia can progress to pancreatic intraepithelial neoplasia (PanIN) (48–52). PanINs are precursor lesions to pancreatic ductal adenocarcinoma harboring mutations characteristic of invasive cancer (53–55).

The liver is similar to the pancreas in being largely mitotically quiescent at homeostasis, with hepatocytes performing their various roles in regulating metabolism and secreting bile and blood proteins. Hepatocytes, however, are exceedingly plastic with hepatocytes in all histoanatomical zones and of all functions able to enter the cell cycle after injury (56).

In the body of the stomach, the professional isthmal stem cell increases proliferation in response to injury; however, the chief cells in the base undergo a dramatic reprogramming akin to pancreatic ADM (57) (Figure 1c). Injury induces chief cells to express markers traditionally restricted to mucous neck cells or embryonic glandular cells [e.g., the epitope for the Griffonia simplicifolia lectin II (GSII), trefoil factor family 2 (TFF2), and mucin 6 (MUC6)] (58). This metaplastic response is referred to as SPEM (spasmolytic polypeptide–expressing metaplasia) because TFF2 was originally called spasmolytic polypeptide (59). But like the pancreatic acinar cells, these metaplastic zymogenic cells retain markers restricted to zymogenic cells [e.g., gastric intrinsic factor (GIF) and pepsinogen C (PGC)].

The innate similarity of SPEM and ADM (large, mitotically quiescent cells reprogramming to express embryonic and wound-healing-associated genes) leads to an obvious hypothesis that the cellular reprogramming program may be conserved across diverse organs. Indeed, the SPEM of chief cells always seems to occur in a broader gastric response that includes loss of acid-secreting parietal cells and increased proliferation of isthmal cells (mentioned above) and often hyperplasia of pit cells (foveolar hyperplasia). This overall pattern has been termed pyloric (or pseudopyloric) metaplasia and is so named because the gastric units with pyloric metaplasia resemble normal units of the gastric pylorus or antrum (57) (Figure 1c). It has been proposed that SPEM and ADM as well as gastric metaplasia in inflammatory bowel disease, autoimmune gastritis, and portions of the metaplasia in the esophagus known as Barrett’s esophagus are all manifestations of a common wound-induced pyloric metaplasia pattern that increases wound-healing factors such as trefoil factors and mucins and increases embryonic gene expression patterns that provide the right context for cells to enter the cell cycle and serve as progenitors to regenerate lost issue (57–62).

3. PALIGENOSIS: A CONSERVED CELLULAR-MOLECULAR PROGRAM MATURE CELLS USE TO REENTER THE CELL CYCLE

We hypothesized that ADM and SPEM within pyloric metaplasia may be instances of an even broader cellular-molecular program that evolution has bestowed upon mature, differentiated cells of a tissue. As mentioned above, such a program, paligenosis, affords mature cells that need to regenerate lost tissue a route to reenter the cell cycle (Figure 2a). However, allowing mature cells to cycle between proliferative and long-lived quiescent, functional states has the downside that it increases the chance for accumulation, storage, and unmasking of oncogenic mutations (60, 63) (see below for more discussion).

Figure 2.

Schematic representation of the dynamic modulation mTOR and autophagy involved in paligenosis. (a) Temporal model of cellular architectural changes through the three stages of paligenosis overlain on the modulation of mTOR and autophagic activity as well as the expression of metaplastic genes. (b) The molecular and cytologic consequences of IFRD1 and DDIT4 knockout on mTOR and autophagy in paligenosis. (c) Molecular schema of how IFRD1, DDIT4, and p53 orchestrate paligenosis. Abbreviations: DDIT4, DNA damage-inducible transcript 4; mTORC1, mechanistic target of rapamycin complex 1.

3.1. Stages of Paligenosis

The pathologist George Adami described over a century ago what a mature cell would need to become proliferative again: It would need to switch energy use from fueling functional activity (e.g., secreting digestive enzymes) to fueling proliferation (64). We now know that paligenosis is just such a switch revolving around the cellular energy hub mTOR, mechanistic target of rapamycin complex 1 (mTORC1) (65), with mTORC1 elevated in homeostatic, differentiated cells to fuel protein production for secretion, mTORC1 switched off as the cell activates autophagy and lysosomes to downscale the large cellular architecture of a differentiated cell, and then mTORC1 switched back on to fuel cell cycle reentry (65) (Figure 2a).

It is not yet clear how the mixture of cell-intrinsic and -extrinsic signals dictates whether a cell will successfully undergo paligenosis. Expression of ATF3 is one of the earliest events in paligenosis, which transcriptionally regulates autophagy via induction of RAB7B (66). We and others have found that reactive oxygen species (ROS) are clearly an upstream event, as ROS-handling programs are induced and critical for normal paligenosis progression (67). Management of ROS depends on the xCT cysteine/glutamate transporter, which ensures progression through paligenosis, at least in gastric chief cells (67). Despite these cell-intrinsic phenomena, it has also been shown that macrophages (68), group 2 lymphoid cells (69), and a variety of cytokines are involved in orchestrating paligenotic metaplasia (70–72). Clearly, much future work is needed to understand how all of these aforementioned and other signals are assimilated by the cell to drive the reciprocal mTORC1/autophagy cycle characteristic of paligenosis. Here, we review our current understanding of the orderly series of cellular-molecular events in paligenosis by which a mature, differentiated cell undergoes paligenosis to become a primitive proliferative cell.

Paligenosis begins when a differentiated cell tunes down mTORC1 activity and upregulates autophagy in order to down scale the mature cellular machinery (Stage 1). This results in a less differentiated, more plastic cell that later begins to express metaplastic genes (e.g., SOX9) coincident with tuning mTORC1 activity (Stage 2) back up to enter the cell cycle (Stage 3) (65).

If lysosomal function is compromised as in the Gnptab^−/− mouse (65), which lacks the ability to transport lysosomal hydrolases, or by treatment with hydroxychloroquine, (66) the cell’s ability to undergo even the initial stages of paligenosis is compromised. Without functional lysosomes, most paligenotic cells do not scale down differentiated cell architecture, fail to express metaplasia-associated genes, and reenter the cell cycle. In addition, failure at this stage (which also occurs when cells lack ATF3 to induce lysosomes) causes increased cell death (66, 73). If we inhibit mTORC1 function with rapamycin, cells downscale and express metaplastic genes but do not proliferate. Thus, we conclude that there are at least three stages of paligenosis: Stage 1 with autophagy/lysosomal degradation of architecture, Stage 2 with induced expression of embryonic/wound-healing/metaplastic genes (e.g., CD44v, SOX9, nuclear YAP1, and others mentioned above), and Stage 3 with cell cycle reentry (Figure 2a).

In accordance with the concept of paligenosis, various forms of autophagy have been implicated as important in zebrafish muscle reprogramming (74), pancreatic duct cell transdifferentiation into endocrine cells (4), beige-to-white fat conversion (75–77), hepatocyte proliferation posthepatectomy (78), and skin repair (79). mTORC1 has also been shown to be a key driver of emergence from quiescence (reviewed in 80, 81), as muscle satellite cells reenter the cell cycle during regeneration (82), as so-called +4 intestinal cells are recruited following crypt injury (83), and in reprogramming of differentiated cells into induced pluripotent stem cells (84).

3.1. Genes Essential to Paligenosis

We hypothesized that there would be a core set of conserved, shared genes/proteins responsible for the precise orchestration of paligenosis the way caspases and BCL family members are critical for apoptosis. These genes would be dispensable at homeostasis and thus there would be no overt phenotype in knockout mice. Further, they would be upregulated following a diverse range of injury models, species and tissues. Accordingly, we screened genes upregulated during injury and known to modulate mTORC signaling. In a recent in silico screen, only two genes fulfilled these criteria: IFRD1 and DDIT4 [DNA damage-inducible transcript 4 (85)].

3.2. IFRD1, a Paligenosis Gene

IFRD1 [interferon-related developmental regulator, also known as PC4 (86) and Tis7 (87)] is a stress-induced protein conserved even in Schizosaccharomyces pombe (85). There is no apparent ortholog in budding yeast. The importance of this protein in regeneration and disease has been demonstrated by its activity as a disease modifier in models of short bowel syndrome (88) and cystic fibrosis (89). IFRD1 has also been demonstrated to alter bone repair and axonal regrowth (85) in addition to the injury models discussed here. A puzzlingly diverse array of cellular functions have been ascribed to IFRD1, including modulating histone deacetylase activity (89, 90), inhibiting NF-κB/RelA acetylation (90), repressing PGC-1α expression (91), and negatively regulating BMP-2 signaling (92).

Following injury, IFRD1 mRNA levels increase within hours and, unlike most other mRNAs whose translation is suppressed due to low mTORC1 activity, IFRD1 is translated due to an upstream open reading frame that augments eIF2-mediated translation (93, 94). Despite mRNA and protein levels of IFRD1 increasing early, loss of IFRD1 manifests later in paligenosis. In Ifrd1^−/− mice following injury we observed that paligenotic cells had the expected initial decrease in mTORC1 activity and increase in autophagy as well as the induction of metaplastic genes. However, paligenotic cells lacking IFRD1 do not efficiently reactivate mTORC1 and favor cell death over proliferation (85) (Figure 2b). This effect is at least partially due to stabilization of p53, as the phenotype of Ifrd1 null mice was ameliorated in the double knockout [Ifrd1^−/−, Trp53^−/− (85)] (Figure 2c). However, other mechanisms are likely involved given the multiple pathways IFRD1 affects (89–92). P53 itself seems to have a critical role in paligenosis, as it suppresses mTORC1 until DNA damage is cleared, thereby performing a key role in preventing mature cells (that may have accrued mutations in previous rounds of paligenosis or via recent ROS) from entering the cell cycle until their genome is intact, an aspect of paligenosis we detail below.

3.3. DDIT4, Another Paligenosis Gene

DDIT4, also known as REDD1 (regulated in development and DNA damage 1), was identified in our screen for paligenosis-related genes. DDIT4 appears to have evolved after IFRD1 because there is no orthologous protein identified in yeast. DDIT4 is best known to suppress mTORC1 indirectly via the TSC2 protein (95, 96), which is believed to occur via DDIT4 releasing TSC2 from associating with its inhibitory 14–3-3 protein (97–100). Although it was originally proposed that DDIT4 directly binds 14–3-3, structural evidence for this interaction is lacking (101). With 14–3-3 inhibition, TSC2 is released, facilitating the hydrolytic activity of RHEB (102), which is required for mTORC1 activity (103–108).

Like others, we found that suppression of mTORC depends on DDIT4, as Ddit4^−/− mice fail to tune down mTORC activity following injury (85). Accordingly, the autophagic response in the presence of persistent mTORC1 is not as robust, and consequently paligenotic cells proceed earlier to mitosis, often prior to fully scaling down cell architecture (85) (Figure 2b,c). The phenotype was similar to that observed in the Trp53 null mice following injury (85).

3.4. Paligenosis-Like Processes

Paligenosis has been defined as a cellular process whereby a differentiated cell modulates mTORC1 and autophagy to augment plasticity and access a proliferative embryonic-like cell state (Figure 2a). We hypothesize that similar, but perhaps not identical, reactions to injury likely occur elsewhere and potentially in tissues that do not have the ability to proliferate. Moreover, it is clear that the decision to proliferate after injury can be separated from the downscaling and gene expression changes. For example, as discussed above, rapamycin-mediated inhibition of mTORC1 can block cells in a downscaled state expressing metaplastic genes yet not proliferating. In human tissue and in other, more chronic SPEM-inducing situations in mice, cells can reprogram into the SPEM phenotype without proliferating (109–111). Indeed, most SPEM in humans and in chronic mouse models (e.g., in mice infected with Helicobacter spp.) is actually not proliferative, so it is possible the first two stages of paligenosis may happen frequently while entry into the third stage is differentially regulated. Alternatively, we speculate that after a brief period of proliferation, redifferentiation may be required prior to another proliferative burst, and this may be an infrequent event (e.g., there is an additional unrecognized checkpoint that prevents dedifferentiated, metaplastic cells from continuously proliferating).

Another example of a paligenosis-like process is the regeneration after axotomy of neurons. It had been previously established that induction of autophagy promotes axonal regeneration (112) and similarly demonstrated that high mTORC1 activity is required for the process to occur (113, 114). Accordingly, we have observed that IFRD1 was also necessary (85), and others have shown that ATF3 is critical for axon regeneration (115). Thus, even in tissues without the ability to proliferate (thus not paligenosis as it was originally defined), similar, conserved molecular networks control reaction to injury. We look forward to having a much more extensive understanding of the conserved interaction of mTORC1, p53, DDIT4, IFRD1, and ATF3 across tissues and species.

4. PALIGENOSIS IS ADAPTIVE FOR MULTICELLULARITY BUT CARRIES RISKS FOR TUMORIGENESIS

Summarizing what we have discussed so far, we can say that the patterns of cell self-renewal and regeneration of tissues at homeostasis and after injury are diverse, and the Waddington, unidirectional model with a professional stem cell in a strict hierarchy is the exception, not the rule. In this section, we examine multicellularity and paligenosis in an evolutionary context, beginning with how paligenosis might have evolved from stress responses in unicellular eukaryotes.

4.1. The Foundation of Paligenosis May Date Back to Yeast

The foundation of paligenosis, a process whereby differentiated cells in a multicellular organism change their differentiation state and are licensed to proliferate, may have originated in unicellular organisms. For example, following injury, S. pombe change from division via fission/mitosis to a meiotic state, which involves changes in differentiation including the generation of a shmoo. Mating via meiosis increases genetic diversity and provides the species a greater likelihood of surviving death or injury. To proceed to the meiotic phenotype following stress, S. pombe must downregulate mTOR (116–120) and induce autophagy (121, 122), completely analogous to stages 1 and 2 of paligenosis. Interestingly, mutant screens in S. pombe implicate its IFRD1 ortholog in stress-induced meiosis and shmoo formation.

It is puzzling in some ways that the Waddington, professional stem cell theory of proliferation regulation grew to such hegemony given that the single-celled common ancestors of metazoans were able to (or had to) balance maintaining all their key organelles and special architecture while also being able to divide. In other words, being able to maintain cells with limited specialized subcellular architecture, whose only function is to divide and fuel other cells, is a special luxury of multicellular organisms. Studies of choanoflagellates, the closest known relative to animals, have been particularly helpful in understanding how multicellularity (and stem cells and paligenosis) might have evolved (123–125). These protists are generally a unicellular organism, but some species (e.g., Salpingoeca rosetta) also show facultative multicellularity, thereby modeling the transition from unicellularity to multicellularity (126, 127). Choanoflagellates are equipped with not only a nucleus/nucleolus and cell membrane, but also mitochondria, endoplasmic reticulum, a Golgi apparatus and associated vesicles, food vacuoles, and flagellum. Thus, a single cell is capable of accomplishing multiple functions that, by division of labor, would be shared by multiple specialized cell types in multicellular organisms (125, 128, 129). In terms of overall organismal complexity, such redundancy is inherently inefficient.

In another protist, the facultatively multicellular Dictyostelium discoideum, some additional division of cell labor has evolved with these organisms undergoing aggregative multicellularity versus the clonal multicellularity of choanoflagellates (125, 130). In Dictyostelium, mTOR plays a central role in cell dedifferentiation (80, 131), suggesting that paligenosis may be a universal feature of cell reprogramming.

Teleologically, multicellular organisms evolved in part because having a large number of cells allows division of labor, with cells differentiating to each be uniquely suited for a specific function. In the process of differentiation, cells scale up some subcellular features at the expense of others (132). However, when a multicellular organism relies on division of labor among multiple different cell types executing a different function, survival requires the ability to regulate cell numbers and health at homeostasis and replace damaged and lost cell populations following injury. In more complex multicellular organisms, cellular differentiation occurs within specialized organs comprising long-lived differentiated cells, with specific and diverse mechanisms required for cell renewal in homeostasis and after injury (133). Paligenosis likely evolved as one way for organisms to harmonize maximal, differentiated division of labor while retaining a contingency plan for regeneration after large-scale tissue damage.

Specialization also permits diverse organismal behavior, protecting the organism from predators and permitting adaptation into new environmental niches. Only in the context of a complex, organ-containing organism does a professional stem cell make adaptive sense. Even in this context, the specialized professional stem cell, whose sole function is to proliferate, would be useful only to replace specific tissue compartments that are continuously lost. Accordingly, the somatic tissues in vertebrates with such professional stem cells are largely those subject to environmental insult: the skin, the lumen of the gastrointestinal tract, portions of the genitourinary tract, and the hematopoietic lineages. Tissues that interact daily with the external environment have a programmed loss of cell mass via shedding (i.e., anoikis) that creates a need for continuous renewal. Blood cells similarly undergo programmed loss via death/pruning (e.g., erythrocytes in the spleen or the kamikaze nature of neutrophils). In the vast majority of organ systems, the parenchyma comprises long-lived specialized cells that only rarely undergo programmed death. In such organs, existence of a cell maintained in a constantly less-specialized (undifferentiated) state, whose sole function is to divide and differentiate, would be as superfluous as it is in less complex organisms lacking dedicated organs.

Although large-scale regeneration of the type seen in more primitive metazoans or in plants (e.g., of limbs, heads, or branches) does not occur in many higher metazoans (134, 135), paligenosis may be a unique vestige of these ancestral injury-induced regenerative pathways. There is limited information on whether paligenosis functions in more primitive multicellular organisms, but that is potentially only because this question has not yet been asked.

4.2. Paligenosis Is a Double-Edged Sword: Critical for Repair but Risky for Cancer

In some ways, cancer is a unique phenomenon of complex multicellular organisms. The invasion, metastases, and death caused by cancer are possible only if there is a complex multicellular organism able to provide support and a niche for the high metabolic demands of these invasive rogue cells. Interestingly, in addition to the infrastructure provided by complex multicellular organisms, it appears that the genetic regulatory pathways that allow cancer to occur are also unique to complex multicellular organisms (136, 137). Trigos et al. (136) have suggested that cancer represents a breakdown of the molecular pathways evolved in higher organisms to restrain unicellular proliferative pathways. Accordingly, when TP53 is exogenously expressed in unicellular organisms (S. pombe) that lack an ortholog, it restrains proliferation (138), corroborating the hypothesis that TP53 suppresses ancestral proliferative pathways.

DDIT4 exemplifies the risks inherent to paligenosis. As discussed above, its best-described role is suppressing mTOR in the setting of DNA damage. Loss of Ddit4 in mice causes cell cycle reentry in paligenosis despite DNA damage, which eventually, through cycles of paligenosis, increases tumorigenesis (139). The checkpoint that halts cells with DNA damage depends on TP53 suppressing the mTORC1 reactivation required for cell cycle reentry (Figure 2). Thus, it appears that the modulation of mTOR and autophagy are used by higher organisms as a pause/checkpoint to assess the state of the cell, allowing only healthy cells to proliferate. Another group has described a similar p53/Ddit4/mTOR checkpoint in tumor suppression (140). Thus, although cell plasticity may be essential to maintain cell populations that lack a stem cell after injury, it must be tightly regulated because the breakdown of licensing checkpoints can be fatal for the organism.

Another aspect of paligenosis with respect to tumorigenesis and multicellularity is that differentiated, functional cells in a complex organism are often long-lived. It is energetically wasteful to elaborate all the complex architecture to make such cells perform their function efficiently and then have those cells constantly turn over. However, paligenosis affords such long-lived cells a route to reenter the cell cycle and then redifferentiate over multiple cycles. Thus, such cells can accumulate mutations over time, each of which may be relatively innocuous, but eventually, an oncogenic mutation may arise, and certain combinations of mutations can lead to cancer. Such a phenomenon has been illustrated in pancreatic acinar cells. Induction of constitutively active Kras^G12D GTPase promotes tumors; however, activated Kras is not sufficient to drive dedifferentiation in the absence of injury or inflammation (53, 141–143). Rather, mutant Kras^G12D must be expressed in cells that have undergone ADM for it to cause uncontrolled growth. Thus, mature differentiated cells can store an oncogenic mutation that becomes unmasked during paligenosis.

We have called this potentially dangerous, mutation-accumulating and -unmasking aspect of paligenosis the cyclical hit model (31, 60, 139). Given that professional stem cells have evolved to constantly generate cells that are quickly shed, we have argued that differentiated, paligenotic cells are much more likely the cells in which the multiple hits of mutations that cause tumors occur (31). Overall, although it is necessary for maintenance of many tissues that lack a professional stem cell, it would seem that licensing differentiated cells to proliferate is inherently dangerous. Moreover, because injury and/or inflammation induce paligenosis, it is possible that the long-understood correlation between inflammation and the development of adult cancer stems in part from this relationship.

5. A PROPOSED CLASSIFICATION OF TYPES OF CELLULAR PROLIFERATION IN ADULT TISSUES

As in most nascent fields, terminology and nomenclature are inconsistent, and this can lead to inappropriate interpretation of data (144). The take-home message from our review of the extant literature is that cell plasticity terminology is still a moving target and may need to be adjusted as future data and technologies become available.

Here, we propose a classification of different types of proliferative, regenerative cell behavior that occur in homeostasis and during injury. We do this in part to underscore that a unidirectional hierarchy model based on a multipotent professional stem cell is only one aspect of the complex patterns of proliferative cell behavior in multicellular organisms. Currently, we believe there is evidence for five different cell types (Table 1; see also Figure 1d). Type I is the professional stem cell. As discussed, many tissues and tissue compartments do not have such a cell and rely on the parenchyma for maintenance of cell population. This cell is constitutively proliferative and lacks differentiated cell features (note the small, granule-free isthmal gastric epithelial stem cell in Figure 1d, subpanel i). It is capable of self-replication and is multipotent, meaning it is able to generate multiple other distinct cell types in the tissue compartment. A subset of type I cells are highly proliferative progenitors that are more differentiated than the multipotent stem cells in a tissue and even more proliferative; this subset would include transit-amplifying cells in the small intestine and the various hematopoietic intermediate progenitor cell types like myeloid or lymphoid progenitors. Type II cells are less proliferative than the constitutively active stem cells but can be proliferative when activated. Like professional stem cells type II cells have some subcellular specialization. +4-Intestinal stem cells exemplify this behavior. The behavior of mucous neck cells in the gastric epithelium has been less well described owing to the lack of good lineage tracing promoters, but they may also eventually be classified as type II (note that in Figure 1d, subpanel ii, the homeostatic, dividing neck cell is small but does have characteristic mucous granules). Type III cells are highly differentiated cells that maintain their own population at homeostasis via autoduplication (see, e.g., Figure 1d, subpanel iii). They are typically (if not always) in tissue compartments without professional stem cells. At times of need (e.g., after injury) they can be recruited as more proliferative, progenitor-like cells via paligenosis. This process involves downscaling of elaborate mature cell architecture and change in gene expression (metaplasia) that allows for more rapid regeneration of cells than can occur via single autoduplication events. Compare the large, autoduplicating pancreatic acinar cell in Figure 1d, subpanel iii (with all of its elaborate secretory function-associated machinery like abundant rough endoplasmic reticulum and secretory granules) with the much smaller paligenotic cell (Figure 1d, subpanel iv), which maintains only scattered granules (consistent with its origin as a digestive enzyme–secreting cell) and trace rough endoplasmic reticulum after massive downscaling for proliferation. Type IV cells have largely been defined in the intestine, where there has been abundant lineage tracing to determine relatively subtle differences in cell behavior. They are differentiated cells, known to be mostly maintained in homeostasis by a professional stem cell (type I). Genetic lineage tracing has demonstrated that injury-induced ablation of the stem cell causes these cells to revert to the stem cell lineage that originally spawned them. The cellular-molecular mechanisms that govern this reversion have not been described: they could potentially be similar to paligenosis in modulating mTOR and autophagy; however, this has not been specifically studied. Finally, type V cells are terminally differentiated and unable to reenter the cell cycle after embryonic development.

Table 1.

Proposed cell type classification based on basal and injury-induced plasticity

Type	Nomenclature	Proliferation time	Description	Examples
I	Constitutively dividing professional stem cell	Homeostasis, postinjury	Constantly dividing Undifferentiated	Intestine: LGR5+ CBC Stomach: isthmal stem cell Skin: basal epithelial cell Bone marrow: stem cell
II	Inducible progenitor	Homeostasis, postinjury	Mildly proliferative Some differentiated features	Skeletal muscle: satellite cell Intestine: +4 cell
III	Autoduplication	Homeostasis, paligenosis	Infrequent division Differentiated	Pancreas: acinar cell Stomach: chief cell Liver: hepatocyte
IV	Differentiated	Postinjury	No basal proliferation Differentiated	Intestine: Paneth cell Intestine: enterocyte Intestine: enteroendocrine cell
V	Terminally differentiated	Not applicable	No proliferation Differentiated	Heart: cardiomyocyte Nervous system: neuron

Abbreviation: CBC, crypt base columnar cell.

6. PALIGENOSIS OUTSIDE NORMAL TISSUE: IN VITRO MODELS AND CANCER CELLS

As described in the previous section, the behavior of cells in tissue is complex within each cell, even in how they execute mitosis, displaying considerable variation. One reason why paligenosis has taken so long to manifest itself as an innate cell program (versus mitosis, which was discovered centuries ago and apoptosis a half century ago) may be because it is a program of differentiated cells in tissue. Tissue culture is designed to make cells undergo mitosis, and cellular pathways governing mitosis involve numerous checkpoints that lead to apoptosis if cells cannot divide. Both processes are thus easy to model in vitro. However, paligenosis occurs in cells that do not grow—cells that remain quiescent mitotically but are highly active when metabolically performing their functions. The whole purpose of tissue culture, even of normal cells (as opposed to transformed cell lines), has largely been to coax cells to grow, which, by definition, means cultured cells are not maintained in a long-lived quiescent, functionally active state.

As previously discussed, paligenosis is likely critical for inducing a regenerative program that is precancerous (i.e., is metaplastic). But what about tumors themselves? If tumor cells originate from metaplasias that emerged via paligenosis, do tumors undergo some sort of paligenosis? The questions about tumor paligenosis and tissue culture paligenosis are linked because most in vitro models involve cultures of cells derived originally from tumors, with genes required for normal error checking in paligenosis (e.g., TP53) frequently mutated in cancer cell lines.

Evidence for paligenosis in tumors stems from studies like those of Rehman et al. (145), who recently demonstrated that following chemotherapy, colorectal cancer decreased mTOR activity and increased autophagy to enter a dedifferentiated state that they likened to embryonic diapause. In this state (akin to the end of paligenosis Stage 1), they found the cells able to resist chemotherapy, slowly cycling, and able to emerge from relative quiescence after injury. They also found that autophagy was required to maintain viability as energy was shifted from anabolic to catabolic. All of these features are consistent with the molecular features of paligenosis that give differentiated cells the ability to become proliferative. Because cancer generally maintains some differentiated features from its parent tissue (e.g., pathologists grade them histologically as poorly, moderately, or highly differentiated), perhaps tumor cells have a low threshold for paligenosis that allows them to cycle between the mitotically quiescent (i.e., a quasi-differentiated state) and proliferative states. Clearly, much remains to be investigated along these lines.

Organoids may offer a unique opportunity to study paligenosis in normal or cancerous tissues in vitro. Organoids can be derived from biopsied or resected adult tissue as well as differentiated entirely in vitro from embryonic or induced pluripotent stem cells. In both cases, normal organoids would be presumably free of the types of mutations that permit the cancer cell lines to grow indefinitely in an unrestrained fashion. Unlike most cancer cell lines, organoids can be coaxed from a proliferative state that increases cell mass to differentiate by either reducing the concentration of specific media components or growing the organoids on a substrate, allowing media on one side and open atmospheric air on the other (the air-liquid interface) (146–149). Some investigators have referred to organoids that grow without substantial differentiation as stem cell organoids (150, 151), but such continuously proliferating organoids also exemplify a type of perpetual tissue plasticity because organoids have to continuously react to the stress of passaging, a sort of tissue injury. It will be interesting to see the behavior of epithelial cells in organoids as organoid culture becomes more sophisticated: adding mesenchyme and other cells that might mimic the cellular niche as opposed to growing organoids in a generic complex of extracellular proteins and with a cocktail of growth factors. Establishing such conditions that recapitulate the tonic stimulation present in vivo could continuously promote a differentiated state. Truly being able to switch organoids from an in vivo, quiescent, differentiating state to an injury-responsive state might help us study in vitro–specific mechanisms of paligenosis-like modulation of mTORC1 activity and autophagy along with specific paligenosis-related signaling cascades.

Clearly, much work needs to be done to better understand paligenosis in tumors and to develop in vitro models of paligenosis, but organoids appear to be a useful starting point.

7. CONCLUSIONS AND FUTURE DIRECTIONS

Considering how widespread tissue plasticity is following injury, in the formation of cancer, and potentially at homeostasis, the precise, conserved cellular-molecular events that govern switches in cellular identity should be a central focus of study. We have taken some forays into studying such mechanisms using rapid, drug-induced mouse injury models. However, there is huge potential for studying roles for autophagy and mTOR (and IFRD1 and DDIT4) in numerous systems where plasticity plays a critical role. For example, potential differences between our proposed type III and type IV cell proliferation behavior might be elucidated by delving into the sequence of cellular-molecular events within each intestinal epithelial cell type sequentially over time after initial injury. There is some indication that the process involves modulation of mTOR (83, 152); however, earlier time points are necessary to determine whether there is biphasic modulation, as we have demonstrated in the stomach and pancreas (65, 85).

Numerous other fundamental questions abound. Do some differentiated cells maintain this ancestral, cell cycle reentry mechanism, whereas others lose it? In the stomach, acid-secreting parietal cells never seem to reenter the cell cycle; they either survive injury or undergo programmed cell death. So, are chief cells the only cell type in the stomach capable of paligenosis? As alluded to above, are any mature cells and/or their partially differentiated progenitor in the small intestine capable of paligenosis? Following injury, they are clearly capable of dedifferentiating into a professional stem cell.

Clearly, we are just at the beginning of what we can learn about how cell plasticity informs our understanding of evolution, regeneration, and cancer. Moreover, plasticity-specific pathways, in particular, those governing paligenosis, might be exploited for therapy to encourage regeneration and discourage cancer in novel ways that differ from our current therapeutic strategies that largely target apoptosis and mitosis. Paligenosis-modulating drugs may have advantages, given that, unlike apoptosis and mitosis, paligenosis is largely a feature restricted to the injury response and so, in an otherwise normal person, paligenotic pathways may be specific to the regeneration and impose cell fidelity checkpoints that would discourage the formation of cancer.