Modulating cell fate as a therapeutic strategy

Summary

In injured tissues, regeneration is often associated with cell fate plasticity, such that cells deviate from their normal lineage paths. It is becoming increasingly clear that this plasticity often creates alternative strategies to restore damaged or lost cells. Alternatively, cell fate plasticity is also part and parcel of pathologic tissue transformations that accompany disease. In this Perspective, we summarize a few illustrative examples of physiologic and aberrant cellular plasticity. Then we speculate on how one could enhance endogenous plasticity to promote regeneration and reverse pathologic plasticity, perhaps inspiring interest in a new class of therapies targeting cell fate modulation.

Introduction

In recent years, it has become evident that while cell fate conversions during homeostasis follow certain prescribed paths of differentiation, injury renders cell fate significantly more plastic. Early pioneering work in the late 1800s identified that newt retina, lens, and even limbs could regenerate, implying that some vertebrates retain extraordinary cell fate plasticity (reviewed in Del Rio-Tsonis and Tsonis, 2003). More modern work, which includes extensive genetic lineage tracing, has helped reveal that seemingly similar organisms can invoke different forms of plasticity while reaching phenomenologically similar forms of regeneration. There is no more spectacular example of regeneration than the restoration of a limb following its amputation. In both red spotted newt and axolotl larvae, regeneration following limb amputation involves the formation of an undifferentiated set of cells termed a blastema. In the adult newt, differentiated cells dedifferentiate and become proliferative while maintaining their original lineage commitment and form cells faithful to the fate of the original progenitor. That is to say, for example, myofibers become proliferative, migratory mononuclear cells that then generate more muscle. The red spotted newt’s close cousin, the axolotl, has comparable regenerative capabilities, and also does so through the formation of a seemingly undifferentiated set of blastema cells. However, in this case, the origin of muscle cells has been shown to be resident PAX7+ muscle stem cells, as opposed to the dedifferentiating myocytes of the red spotted newt (Sandoval-Guzman et al., 2013; Tanaka et al., 2016). Interestingly, large numbers of transcripts unique to axolotl blastema have been detected, suggesting that a mechanism distinctly evolved in these organisms (Haas and Whited, 2017). We emphasize this example to draw attention to the notion that even in the context of urodele amphibian regeneration, proliferation of lineage-restricted cells, stem cells, and dedifferentiating cells can be variously deployed to effect the same fundamental regenerative end. It seems as though evolution is capable of harnessing a deep homology concerning cell fate interconvertibility. It has become increasingly clear that plasticity is a cardinal mode of injury-response in myriad organs in mammals. Particularly prominent examples are found in epithelia where a focused effort has been made to identify plasticity phenomenon.

In addition to being a regenerative phenomenon, cell fate plasticity is also evident in disease states. Metaplastic transformation of tissues accompanies both malignant and non-malignant pathologies. Most notably cancer and its antecedents have been associated with arrested or aberrant differentiation. This conceptual piece points to the possibilities that (1) regeneration could be stimulated by enhancing effective plasticity and that (2) pathological plasticity could be reversed by modulating cell fate. We speculate how insights into cell plasticity may inspire next-generation therapeutics that move into the clinic.

Physiologic Cell Fate Plasticity

Dedifferentiation in Epithelia

While amphibians and teleost fish are known for their prodigious ability to regenerate, we will focus on a few illustrative examples of mammalian cell fate plasticity. Most of the well-documented examples of mammalian plasticity, proven using indelible lineage tracing in the mouse, occur in epithelia. These tissues have historically been described as having parental stem cells that appear “undifferentiated” and functional differentiated cells such as secretory, absorptive, ciliated, and sensory cells. Recent studies show that genetic ablation of specific cell types or physiologic damage to tissues can cause differentiated cells to revert into stem cells or assume a facultative stem cell function to repair tissues. In other cases, differentiated cells can assume alternative fates directly, despite the fact that these lineage paths do not exist normally in the steady state tissue.

In the epidermis, it has been reported that lineage restricted progenitors can acquire the potential for forming cells of other lineages following epidermal injury. For example, bulge stem cells are normally restricted to hair follicles, but extensive injury induces them to transiently contribute to the epidermal compartment during the acute phase of wound repair (Ito et al., 2005) (Figure 1A). On the other hand, the reverse has also been shown, where extensive injury induces fate plasticity in which epidermal cells of the interfollicular epidermis surrounding a wound generate hair follicles through a WNT dependent mechanism reminiscent of development (Ito et al., 2007)(Figure 1A). Similarly, the stem cells of the upper pilosebaceous unit are also lineage restricted, until punch biopsy injury, after which they contribute long term to the interfollicular epidermis (Page et al., 2013)(Figure 1A).

Figure 1.

Dedifferentiation in Epithelia. (A) In the epidermis, bulge, interfollicular epidermal, and upper pilosebaceous stem cells (PSU) normally produce differentiated cells restricted to their respective compartments. After injury, this restriction is lifted, allowing for bulge stem cells to contribute transiently to the epidermal compartment, epidermal stem cells to contribute to the generation of new hair follicles, and upper PSU stem cells to contribute to the interfollicular epidermis. (B) In the airway, differentiated secretory cells can dedifferentiate after ablation of the basal stem cell population. (C) In the stomach, Chief cells, which produce pepsinogen and chymosin, can generate all of the cell types of the tissue. This ability is enhanced significantly after injury. (D) In the intestine, differentiated cells are capable of broad levels of plasticity. After injury, reserve intestinal stem cells (rISC), along with early transit amplifying (TA) cells and secretory progenitors, are capable of gaining stem cell function. (E) In the cornea, ablation of the stem cells in the limbus induces dedifferentiation and migration of corneal cells to restore the limbus. (F) In the olfactory epithelium, severe injury involving loss of the Sus and olfactory sensory neurons (OSNs) induces dedifferentiation in neuronally specified progenitors (GBC_inp) to become multipotent progenitors (GBC_mpp).

In the airway, we have demonstrated that differentiated secretory cells can dedifferentiate after ablation of the pre-existing basal stem cell population, restoring the lost basal stem cells. Indeed, dedifferentiated cells that have adopted a basal fate are functionally identical to pre-existing stem cells, giving rise to long lived progeny that persisted for at least 2 months (Tata et al., 2013)(Figure 1B). Interestingly, there was an inverse relationship between the likelihood of dedifferentiation and the maturity of the differentiated cell, suggesting that cell fate becomes progressively “locked” with time. In the lung, it has also been shown using Hopx-CreER lineage tracing, that some type 1 lineage-traced cells can generate type 2 alveolar stem cells after pneumonectomy. Whether this is an example of dedifferentiation or labeling of a sub-population of type 1 cells remains to be elucidated (Jain et al., 2015).

In the gastrointestinal tract, several different populations of cells have demonstrated lineage plasticity. Most strikingly, differentiated TROY+ chief cells of the stomach that produce pepsinogen and chymosin, have also been shown to be capable of generating all of the cell types of the tissue, and this function is enhanced following injury (Stange et al., 2013)(Figure 1C). In the intestine itself, differentiated cells appear to have broadly “open” chromatin that allows their interconversion based on Notch target gene expression (Kim et al., 2014). Perhaps such permissive genomic architecture underlies the ability of Alpi+ enterocytes, Bmi1+ enteroendocrine, early transit-amplifying, and secretory progenitor cells to dedifferentiate into LGR5+ stem cells of the crypts (Buczacki et al., 2013; van Es et al., 2012; Jadhav et al., 2017; Tetteh et al., 2016; Tian et al., 2011; Yan et al., 2017)(Figure 1D).

More recently, it has been shown that depletion of ectodermally-derived corneal stem cells induces the dedifferentiation of committed corneal cells into stem cells to restore the lost stem cell compartment, remarkably reminiscent of the stem cell ablation in the airway. Interestingly, this study also provided a unique insight into mechanism by demonstrating that destruction of the limbal stroma prevented dedifferentiation, suggesting that some aspects of cell fate plasticity can be negotiated by adjacent mesenchyme (Nasser et al., 2018)(Figure 1E).

Finally, the murine olfactory epithelium, a pseudostratified neuroepithelial sensory tissue that is responsible for the sense of smell, utilizes dedifferentiation to accelerate tissue repair. The epithelium is comprised of sensory neurons that are unique in that they directly contact the external environment to bind odorants, which exposes them to damage (Shipley, 1985). To compensate for the continual turnover of damaged olfactory neurons, the epithelium is armed with several different pools of stem cells that act together throughout adult life to generate neurons and maintain olfaction (Fletcher et al., 2011; Huard et al., 1998; Schnittke et al., 2015). While the tissue contains a full complement of previously described conventional active and reserve stem cell pools, in times of significant injury, the tissue can also call upon neuronally specified progenitors to dedifferentiate into multipotent stem cells that then contribute to multi-lineage tissue repair using developmentally relevant pathways. These cells downregulate the epigenetic effector Ezh2, up-regulate Sox2, Klf4, and Pax6, and traverse the normal neurogenic path in reverse, until ultimately converting into a multipotent stem cell. This immediate injury response effectively increases the proliferative pool of multipotent stem cells available to the tissue during the initial week post-injury, accelerating overall tissue regeneration (Lin et al., 2017) (Figure 1F). Once again, the importance of epigenetic factors in cell plasticity phenomenon is evident.

These examples of injury responses suggest that dedifferentiation is ubiquitously deployed in varying degrees amongst many epithelia. Notably, all of the aforementioned tissues are exposed to constant environmental damage, and effective repair is crucial for organ homeostasis and survival. It remains to be seen whether these tissue-specific modes of regeneration share an underlying core mechanism, although in aggregate, it is likely that a common set of epigenetic principles determine which types of cell fate transitions are permitted (or favored) in a given injury in a given tissue. In the absence of direct genetic lineage tracing in humans, newer technologies will have to be employed to definitively establish plasticity phenomenon. These include clone determination using the tracking of somatic or mitochondrial mutations, or in the absence of an indelible genetic stamp, computational reconstruction (Bendall et al., 2014; Greaves et al., 2006; Horns et al., 2016; Treutlein et al., 2014; Woodworth et al., 2017).

Not all tissues employ dedifferentiation as the mechanism of choice for tissue regeneration, although it appears widespread in epithelia containing resident stem cells. In some tissues that experience low levels of turnover, the primary mode of regeneration may be transdifferentiation. Upon ablation of insulin-producing β cells of the pancreas, glucagon-producing α cells can transdifferentiate into the lost β cells to supplement those cells produced by the replication of existing β cells. Oddly, juvenile mice do not employ this α cell transdifferentiation. Instead, somatostatin-positive δ cells engage a developmentally relevant pathway to dedifferentiate and regenerate lost β cells (Chera et al., 2014; Thorel et al., 2010). This suggests that lineage plasticity, even within one mammalian organ, can follow different courses with age and maturity. As in the trachea, developmental maturity, either with reference to the age of an organism or the age of a particular cell, seems to have a profound influence of cell plasticity (Tata et al., 2013).

Metaplasia and Pathologic Cell Fate Plasticity

Epithelial metaplasia

Metaplasia has been classically defined as a tissue type transformation, and this is how the term is used by pathologists to grossly describe the phenotypes of tissues. In this context, the term does not directly imply a particular form of cell plasticity and merely refers to the presence of novel cells or disproportionate numbers of normal cells within a tissue compartment. The best examples arise from human tissues. The pathologist’s definition of metaplasia also does not explicitly denote whether the tissue type transformation is beneficial or harmful. It has been speculated that metaplasias are associated with a beneficial temporary benefit, but can become the harbingers of pathologic metaplasia and cancer. The exact mechanisms governing the loss of metaplasia reversibility remain to be elucidated. Similarly, there is little understanding of why metaplasia leads to transformation and frank cancer. As an example, airway exposure to noxious stimuli like cigarette smoke leads to an increase in the numbers of squamous cells. These squamous cells are presumed to subtend a protective epithelial barrier function. Since squamous cells are not present in the unperturbed airway, a tissue transformation is grossly evident, but the source of new cells is still mysterious. It has been shown that in patients that have stopped smoking, squamous differentiation is less evident, suggesting that this form of metaplasia can be physiologically reversible (Lapperre et al., 2007; Rigden et al., 2016; Schamberger et al., 2015). At the same time, squamous metaplasia is thought to presage lesions with cytologic atypia that are on their way to becoming frankly dysplastic and eventually giving rise to squamous cell cancer. Little is known about the possible conversion of physiologic protective squamous metaplastic into pre-cancerous metaplasia. Indeed, it remains a formal possibility that these are distinct states that resemble one another histologically.

Although the appearance of a novel cell type may represent the emergence of a new cell fate from a stem cell or a transdifferentiation of a pre-existing mature cell type into a novel cell type, it is also possible that unidentified rare progenitor cell populations expand and provide a source of novel cell types. Thus, tissue level metaplasia may arise from cellular plasticity or involve the expansion of cell types that are not evident in the steady state. Indeed, the latter appears to be the case in Barrett’s esophagus, perhaps the most classic example of human metaplasia, in which continuous acid and bile reflux cause an adaptive change in tissue composition. In this case, the distal esophageal epithelium adopts an intestinal phenotype presumed to protect the mucosa from injury. Many hypotheses have been raised for the origin of this metaplasia; however, recently a novel transitional basal cell population has been identified. These transitional basal cells, which are TRP63+, KRT5+, and KRT7+, are distinct from normal squamous basal cells of the esophagus or stem cells of the adjacent stomach. This transitional epithelium expands following injury and is particularly receptive to CDX2-mediated intestinal metaplasia. Interestingly, similar cells are found at other squamocolumnar junctions, such as those of the anus and cervix (Jiang et al., 2017)(Figure 2A), implying that many epithelial metaplasias may be a result of the expansion of a rare cell population (with or without subsequent cell plasticity events).

Figure 2.

Pathologic Cell Fate Plasticity. (A) In Barrett’s esophagus, rare transitional basal cells in the squamocolumnar junction (SCJ) expand to generate intestinal metaplasia. (B) In mucous metaplasia of the airway, chronic inflammation induces secretory cells to differentiate into goblet cells that produce excess mucous. (C) In progressive osseous heteroplasia, activating mutations in the BMP pathway induce adiopocyte progenitors to induce bone forming transcriptional programs and deposit calcium. (D) in calciphylaxis, vascular smooth muscle cells (VSMCs) in the tunica media induce ectopic Runx1 and begin to deposit calcium. (E) In normal hematopoiesis, hematopoietic stem cells (HSCs) generate the common myeloid progenitor (CMP) that then give rise to myeloblasts, which can then generate monocytes or granulocytes. In acute myeloid leukemia, this differentiation step in myeloblasts is blocked, and large amounts of myeloblasts are generated.

Other forms of epithelial metaplasia are also considered beneficial injury responses. In addition to the squamous metaplasias alluded to above, mucous metaplasia arises in tissues such as the airway or nasal epithelium. In this setting, the epithelia are characterized by increased numbers of goblet cells and basal stem cells. Thus, the tissue is characterized as metaplastic based on these gross changes in cell proportionality of normally resident cell populations. The excess of goblet cells, stimulated by inflammatory cells following infection or allergy, are thought to be needed to produce mucous to increase barrier function alongside the squamous metaplasia that often accompanies the same injury (Curran and Cohn, 2010)(Figure 2B). The source of new goblet cells is debated, but likely includes the differentiation of pre-existing secretory cells. Goblet cell metaplasia is reversible at the histologic tissue level, but once again it is unclear whether the goblet cells die, are extruded, undergo apoptosis, or whether they revert into secretory cells. Thus, even this simple form of beneficial metaplasia is poorly understood at a cellular level. Additionally, why squamous and mucous metaplasia arise in the same setting in the airway, and what dictates which metaplasia arises in a given locale within the airway remains mysterious. In aggregate, these examples illustrate that the basis of tissue level metaplasias as defined by histologic tissue transformation are not understood at the cellular level. At the end of the day, all we can say is that metaplasias may arise from plasticity phenomena, but they may also arise from the expansion of unknown cell populations.

Mesenchymal metaplasia: Transdifferentiation Yielding Ectopic Calcification and Bone Tissue

Some striking examples of tissue metaplasia can occur in mesenchymal tissues where fibrosis or, in extreme situations, ectopic mineralization and bone formation can occur. Fibrosis, at times may be associated with epithelial-to-mesenchymal transition (EMT), but often also occurs because of an expansion of non-epithelial cell populations, including fibroblasts of various sources, smooth muscle cells, and pericytes (Nieto et al., 2016). Recent studies suggest some forms of fibrosis result from the expansion and over-activity of a cell type present in or near the tissue, rather than a cell fate change (Dulauroy et al., 2012; Zepp et al., 2017). In addition to fibrosis, forms of metaplasia such as calcification and bone formation can occur outside the skeletal system as hereditary or non-hereditary. Calcification can arise from systemic mineral imbalance, where elevated levels of phosphorous and calcium can form mineralized deposits without known transdifferentiation events (Block et al., 1998; Giachelli, 1999). However, in hereditary and non-hereditary heterotopic ossification and in vascular calcification, it is becoming clear that specific populations of tissue resident cells are adopting a bone regulatory program, which includes activation of BMP or Hedgehog signaling or of bone associated transcription factors such as Osterix or Runx2 (Lin et al., 2015; Regard et al., 2013; Zhang et al., 2017). The hereditary forms of heterotopic ossification include fibrodysplasia ossificans progressiva (FOP) and progressive osseous heteroplasia (POH), and they are caused by gain-of-function mutations in ACVR1 resulting in activated BMP signaling, and inactivating mutations in GNAS which result in activated Hedgehog signaling (Regard et al., 2013; reviewed in Shore and Kaplan, 2010). Although both diseases are characterized by the progressive formation of bone in extraskeletal connective tissues, they have distinguishing features. POH can occur more often in the dermis near adipose tissue and moves to deeper regions of connective tissue and muscle, whereas ossification in FOP is found in tendons, ligaments, muscle, and fascia and can be compounded by injury involving the immune system. Lineage tracing experiments suggest non-endothelial interstitial Tie2⁺/Pdgfrα⁺ fibro-adipogenic progenitors, rather than smooth muscle cells or muscle cells, contribute to heterotopic bone (Dey et al., 2016; Lees-Shepard et al., 2018; Lounev et al., 2009; Wosczyna et al., 2012). In addition, expression of activating ACVR1 mutations in Scleraxis-derived tendon cells rather than in other endothelial or myofibroblast cell types causes ectopic expression of Sox9 and Osterix and subsequent bone formation (Agarwal et al., 2017) (Figure 2C).

The ability of tendon and ligament cells to form bone is not limited to these hereditary ectopic bone disorders. Cartilage and bone can form in tendons, ligaments, and their attachment sites because of overuse conditions, genetic defects in matrix components, or X-linked hypophosphatemia (AMEYE et al., 2002; Archambault et al., 2007; Benjamin and Ralphs, 1998; Liang et al., 2009). Interestingly, spinal cord, traumatic brain, and burn injuries as well as major surgical events such as hip arthroplasty can initiate heterotopic ossification in soft connective tissue in joints unrelated to the injury site (Ranganathan et al., 2015). The mechanisms underlying these lesions are not well understood, but are thought to involve the inflammatory response and changes in the local microenvironment. For example, ossification of the ligamentum flavum, which can lead to spinal stenosis, was linked to increased expression of the pro-inflammatory cytokine, TNFα. Culture with TNFα upregulated Osx and promoted osteoblast differentiation in primary cells cultured from the ligamentum flavum (Zhang et al., 2017). Changes in mechanical signals also have effects on gene expression in connective tissues (Archambault et al., 2007), consistent with the idea that the mechanical environment can shift cell identity (Engler et al., 2006). Tendon, ligament, and joint tissues appear particularly susceptible to shifting fates towards cartilage and bone programs. Whether this stems from their close developmental relationship (Blitz et al., 2013; Soeda et al., 2010; Sugimoto et al., 2013) or an inherent connection between skeletal and connective tissue regulatory programs is currently unknown, but it is interesting to note that there is specificity in the cell types that can be acted upon to form bone. This observation could provide important clues as to the underlying biology of the cell types capable of transdifferentiation, as well as to the local environmental signals eliciting such dramatic cell fate changes.

In addition to ectopic bone, two primary types of vascular calcification have been described in adult disease: (1) intimal calcification associated with atherosclerosis, myocardial infarction, and stroke; and (2) medial calcification associated with chronic kidney disease, and diabetes (Amann, 2008; Otsuka et al., 2014). Calciphylaxis is a rare, but well documented syndrome, characterized by the calcification of microvessels (40–600 micrometers) in dermal and subcutaneous adipose tissue and distinctive painful skin lesions (Chen et al., 2014; Nigwekar et al., 2015). Analyses of biopsies reveal medial wall calcification leading to ischemia with further occlusion by thrombosis resulting in necrosis and the hallmark skin lesions (Magro et al., 2010; Nigwekar et al., 2008, 2015; Weenig, 2008).

The predominant cell type within the tunica media, the site of medial vascular calcification and calciphylaxis, is the vascular smooth muscle cell (VSMC). Under normal homeostasis, the function of this contractile cell is to maintain vessel wall integrity and regulate arterial tone. However, VSMCs exhibit remarkable plasticity even in adulthood. Evidence suggests contractile VSMCs can adopt an osteogenic phenotype causing vascular calcification (Goiko et al., 2013; Gomez and Owens, 2012; Malhotra et al., 2015; Shanahan et al., 1994; Shankman et al., 2015; Speer et al., 2010). This phenotypic switch is characterized by loss of smooth muscle cell markers (e.g. SM22, α-smooth muscle actin, calponin), and induction of osteogenic cell markers (e.g. RUNX2 and osteopontin)(Speer et al., 2002, 2010; Steitz et al., 2001). The triggers for this phenotypic switch are myriad, including aging, oxidative stress, inflammatory cytokines, high phosphate concentrations, and paracrine signaling (Chang et al., 2008; Chen et al., 2014; Kapustin and Shanahan, 2016). Unfortunately, the specific trigger for calciphylaxis remains unknown, though it has been hypothesized that it may be acute tissue trauma in the context of risk factors like inflammation and high phosphate. It is known, however, that expression of the transcription factor RUNX2 is associated with the development of calciphylaxis (Kramann et al., 2013; Nigwekar et al., 2017; Speer et al., 2010). Subsequent to osteogenic phenotype switching, VSMCs secrete matrix vesicles containing calcium phosphate that serve as a nidus for medial vascular calcification (Kapustin et al., 2015). Altogether, VSMC transdifferentiation – likely mediated by Runx2 – is implicated in the pathophysiology of calciphylaxis (Figure 2D).

Interestingly, pathological cues can induce heart valves, which contain both endothelial and interstitial cells, to undergo valvular calcification. Interstitial cells can undergo a phenotypic transition to an osteoblast-like cell that produces bone matrix and signaling molecules, ultimately promoting valvular calcification. This process mirrors vascular calcification involving BMP-induced Runx2 (Leopold, 2012; Rajamannan et al., 2003). On the other hand, endothelial cells lining the valve surface can also undergo a phenotypic transition to an osteoblast-like cell. To do this, they first pass through an endothelial-to-mesenchymal transformation, mimicking a developmentally critical process for normal valvulogenesis (Hjortnaes et al., 2015; Wylie-Sears et al., 2011). The mechanism underlying this phenomenon has been well characterized—it is known that calcification is normally inhibited in endothelial cells by shear stress, which activates Notch signaling and induces expression of calcification-inhibiting Matrix Gla protein (White et al., 2015). Accordingly, loss-of-function mutations in Notch1 have been shown to result in severe aortic valve calcification in adults (Theodoris et al., 2015). Altogether, these studies and others demonstrate the importance of cell fate plasticity in a wide range of calcification pathologies.

Differentiation Arrest and Cancer

The concept of differentiation arrest pre-dates our molecular and genetic understanding of cancer. Since the advent of microscopy, it has been recognized that malignant cells appear morphologically less-differentiated, or aberrantly-differentiated, and have been termed dysplastic. Furthermore, in many circumstances the degree of visible differentiation correlates with the clinical aggressiveness of disease, forming the basis of current tumor grading schemas. Indeed, the most concerning tumors are referred to as anaplastic, denoting cells that have lost many or all the characteristics of their tissue-type of origin. In the most undifferentiated cases, these anaplastic malignant cells lose all identifiable features, and appear only as “small-blue-round-cell-tumors”, a histopathological category unto themselves.

Differentiation arrest has been best exemplified in disorders of blood cells, mostly owing to years of research contributing to a deep understanding of hematopoietic stem cell biology and the normal, step-wise developmental process of hematopoiesis. Even prior to the identification of the first translocations and mutations contributing to acute myeloid leukemia (AML), there was the recognition that the leukemic “blast” cells from different patients appeared variably mature under the microscope (Bennett et al., 1976) (Figure 2E). Originally, this sub-division of AML was merely academic, but is now recognized as therapeutically relevant.

Therapeutic Modulation: Forms of Therapeutic Cellular Conversions

Differentiation Therapy for Cancer

Differentiation therapy promises to target and to overcome the differentiation roadblock present in disorders such as AML, prompting the corrupted cancer cell to resume its normal process of maturation, development, and death. Furthermore, because the tumor-initiating cells are believed to be particularly stem-like, differentiation therapy might preferentially target this critical reservoir, with curative potential. Originally, all patients with AML received the same chemotherapy regardless of differentiation state. However, it was recognized that patients with acute promyelocytic leukemia (M3) presented with a clinically distinct disease.

Upon microscopic examination, the AML M3 leukemic blasts were typically packed with darkly staining primary granules, apparently frozen at a developmental state half-way between a non-descript myeloid blast and a terminally-differentiated neutrophil. Ultimately, advances in karyotyping would demonstrate that every M3 leukemia was marked by a recurrent translocation involving the retinoic acid receptor (RAR), typically t(15;17), which results in production of the PML/RAR-alpha chimeric oncoprotein.

This recurrent translocation also predicted sensitivity to treatment with high doses of all-trans retinoic acid (ATRA). Remarkably, ATRA relieved the transcriptionally repressive program of PML/RAR-alpha, and the leukemic cells were freed from differentiation arrest, and rapidly progressed down the normal developmental course to mature neutrophils, despite the fact that they still carried the abnormal translocation (Wang and Chen, 2008). Therapy with ATRA has since been combined with arsenic trioxide, which seems to target the PML/RAR-alpha fusion for degradation. This combination of molecularly-targeted pro-differentiation agents has resulted in curative therapy for the majority of patients, obviating the need for conventional chemotherapy. Where acute promyelocytic leukemia once carried the worst prognosis of all the forms of AML, it now carries the best, and this complete inversion of the survival curve underscores the most successful prototype for differentiation therapy as a whole (de The, 2017) (Figure 3A).

Figure 3.

Therapeutic Modulation of Cell Fate. (A) An example of differentiation therapy in which all-trans retinoic acid (ATRA) and arsenic trioxide (As₂O₃) are used in acute myeloid leukemia (AML), to force the differentiation of aberrant myeloblasts into monocytes and granulocytes. (B) Localized administration of Notch inhibitors could prevent the differentiation of secretory cells into goblet cells, removing the source of excess mucous in airways diseases. (C) Localized TNF-α or dorsomorphin could block Runx2 or BMP, respectively, and reverse calcification. (D) Restoring the ability of Muller glia to express Ascl1, Sox2, and KLF4 could yield cells capable of dedifferentiation and restoration of lost rod and cone cells in the retina. (E) Viral infection of the airway selectively targets short-lived luminal cells, preventing long lasting treatment. Subsequent basal cell ablation or outright induction of dedifferentiation of virally transduced secretory cells could generate long-lasting basal stem cells carrying the proper gene therapy. (F) Cardiac infarct generates fibroblast scarring in heart tissue alongside the death of cardiomyocytes. Directed reprogramming of scar fibroblasts into cardiomyocytes could be used to treat both the scar and restore the myocardium.

Given the successes in acute promyelocytic leukemia, investigators have turned to other subtypes of AML. AML is a genetically diverse malignancy, and categorizing recurrent mutations has confirmed that many of these disrupt the function of proteins known to be involved in normal maturation, suggesting that directed small-molecule therapy may be able to overcome differentiation arrest. This has prompted the development of recently-approved therapies targeting mutant IDH (Stein et al., 2017), as well as new investigational targets including LSD1 (Schenk et al., 2012), DOT1l, and EZH2 in clinical development.

Other investigators, have taken an unbiased approach to identifying new differentiation therapies in AML, relying on phenotypic screens rather than target-directed screens (Banerji et al., 2012; Pikman et al., 2016; Radomska et al., 2015; Stegmaier et al., 2004; Sykes et al., 2016). This approach presupposes that different leukemias may share a common node of differentiation arrest despite their genetic heterogeneity. New targets have emerged from these screens, and the clinical utility of small molecules targeting SYK, GSK3-alpha, MTHFD2, and DHODH are all under investigation.

Taking the paradigm one step further, one might hypothesize that many solid cancers arise in the setting of arrested differentiation. If so, what then are the barriers to developing differentiation therapy in non-hematologic malignancies? The clinical success in AML builds on a foundation of deep understanding of normal hematopoiesis including the predictable expression of transcription factors and cell-surface proteins effectively marking each stage of differentiation. This same roadmap is not available in all tissue or tumor-types, making it difficult to devise a similar differentiation screening strategy. Despite the technical challenges inherent to phenotypic screens, we feel strongly that these have the potential to not only identify new biology, but also to identify therapies with a clinically-relevant therapeutic window. Our current reliance on screens with a simple viability read-out must make way to more sophisticated screens that report on cell-fate decisions. This can facilitate the path towards identifying small molecules that can bias a malignant progenitor towards “proper” differentiation.

Indeed, large-scale genomic analysis of colorectal cancer has yielded a potential path to treat PTPRK-RSPO3-fusion positive tumors that is strikingly reminiscent of the path taken for AML. In this subset of tumors, it has been shown in xenograft models that anti-Rspondin treatment results in tumor growth inhibition. Furthermore, treatment resulted in a downregulation of stem-cell markers and an increase in mature crypt differentiation markers as seen through RNAseq, and also exhibited a differentiated appearance. This study and others have provided a proof-of-principle for differentiation therapy in non-hematologic malignancies (Chartier et al., 2016; Storm et al., 2016).

We can also imagine that an effective way for cancer to evade chemotherapy might involve a cell fate switch. Indeed, cellular plasticity in the form of EMT has long been recognized as a phenotype encountered in resistant cancers. As the mesenchymal state is being characterized, one could imagine blocking the transdifferentiation of epithelia into more mesenchymal cells. More shockingly, epithelial cancers have been demonstrated to undergo wholesale conversion from one tissue type to another. Indeed EGFR+ lung adenocarcinomas have converted into small cell cancers (Niederst et al., 2015; Oser et al., 2015) and prostate cancers can gain resistance to anti-androgen treatment by shifting from an androgen receptor (AR) dependent luminal state to AR-independent basal cell state mediated through loss of Trp53 and Rb1 whose loss is promoted by Sox2 (Ku et al., 2017; Mu et al., 2017). If these forms of cellular plasticity represent a common escape mechanism from targeted therapies, it follows that blocking transdifferentiation might prevent escape.

Reversing Transdifferentiation

Based on the initial success of differentiation therapies in the setting of leukemia, one can imagine similar strategies for reversing pathologic metaplasias. In mucous metaplasia, such as encountered in asthma, COPD, or cystic fibrosis, repetitive insults or infection invoke an inflammatory response, triggering goblet cell hyperplasia. Most strategies that have been designed for treating diseases associated with pathologic mucous metaplasia have aimed to halt the inciting immunoinflammatory stimuli. In the case of asthma, corticosteroids and antibiotics, along with more selective immunomodulatory agents such as leukotriene antagonists and anti-IL-13 monoclonal antibodies are often effective but carry significant side effect profiles (Curran and Cohn, 2010; Nguyen et al., 2017; Walker, 2003). We propose that targeting the actual goblet cell differentiation event, and thus directly blocking the cell fate transitions that result in excess mucous-producing goblet cells, might be of therapeutic value regardless of whether the underlying pathology is stimulated by an allergen in asthma or smoke in COPD-related coughing. Indeed, the Notch pathway is critical for cell fate decisions in the airway epithelium and in promoting airway mucous metaplasia (Chiba et al., 2009; Miklossy et al., 2013). Furthermore, inhibition using Notch antagonists has been shown to block IL13-induced mucous metaplasia. Together, this suggests that a local Notch inhibitor, potentially delivered as an aerosol, could specifically reverse mucous metaplasia by rebalancing the ratio of goblet cells to normal (Danahay et al., 2015; Guseh et al., 2009)(Figure 3B). Of note, since this form of differentiation blockade involves a balance of multiple cell types, one must be careful not to alter the normal differentiation of a necessary cell type that is not involved in the pathology. This stands in sharp contrast to the situation with cancer in which it is preferable that all pathologic cells are removed. Additionally, one must note that signaling pathway modifiers may have untoward effects in various tissues throughout the body. Some form of localized therapy is likely to be needed, including topical or aerosolized, or non-absorbable formulations for the skin, lungs, and GI tract respectively.

The lack of effective therapies for a disease such as calciphylaxis or the hereditary heterotopic bone disorders suggests the possibility that reversing the osteogenic phenotype might be used therapeutically. In this case, there is the potential to utilize the existing knowledge of the mechanisms underlying transdifferentiation and vascular calcification to develop therapies that could inhibit or even reverse pathologic vascular calcification. Existing research described above suggests that downregulating RUNX2 (such as through TNF-α) (Gilbert et al., 2002) or inhibiting BMP signaling in vessel walls would be effective at preventing VSMC transdifferentiation (Figure 3C). With FOP in particular, there is a focus on therapeutic agents that can interfere with BMP receptor activation (Luo et al., 2016; Yu et al., 2008). Finally, inhibition of paracrine signaling or cytokine-mediated fate changes or restoring normal mechanical signals in different forms of heterotopic bone formation could prevent or reverse the underlying osteogenic pathology. Ultimately, however, a greater understanding of the molecular mechanisms that induce altered pathologic cell fates are required to identify more specific targets.

Promoting Dedifferentiation

While we have speculated about reversing pathologic cell plasticity, one might consider harnessing cell plasticity as a direct form of therapy. There are several potential avenues for the future. It’s long been noted that “lower” organisms often possess greater regenerative potential than their mammalian counterparts. Leveraging what we know about cell plasticity in lower organisms might allow us to coax mammalian tissues in a similar vein. For example, in lower animals, the sensory epithelia of the olfactory system, retina, and inner ear are all known to possess regenerative capacity after injury. Interestingly, the olfactory epithelium of the mouse is the only sensory epithelium that has retained its regenerative potential, and does so through the induction of a subset of reprogramming factors (Sox2 and Klf4 along with c-Myc target genes). In contrast, although the zebrafish inner ear and retina are capable of regenerating after injury (Millimaki et al., 2010; Ramachandran et al., 2010), the mammalian cochlear epithelium and retina are incapable of significant regeneration of hair and photoreceptor cells, respectively. Zebrafish and embryonic chick Muller glia of the retina express sox2, along with the other reprogramming factors nanog, oct4, and c-mycA (Fischer et al., 2010; Gallina et al., 2014; Ramachandran et al., 2010). Perhaps the expression of these factors could facilitate adult chick retinal regeneration. Similarly, mammalian retina, expresses only Sox2 (Fischer et al., 2010) and perhaps re-expression of the missing reprogramming factors would permit retinal repair (Figure 3D). Indeed, there are suggestions that a brief overexpression of Oct4, Sox2, and Klf4 in fibroblasts, along with modulation of signaling cascades associated with the adult myocardial niche, results in effective reprogramming towards a cardiomyocyte lineage. However that has yet to be done successfully in vivo (Efe et al., 2011). An alternative strategy is also yielding fruit as recent work in the cochlea has demonstrated the feasibility of small molecule mediated expansion of LGR5+ supporting cells followed by their directed differentiation into hair cells (McLean et al., 2017).

We also speculate that dedifferentiation could be an attractive method of affecting long-lasting gene therapy in tissues with high turnover. In both the skin and the airway, it is difficult to virally transduce the underlying stem cell population, resulting in short-lived effects due to constant turnover of the targeted cells. However, efficient dedifferentiation of infected mature epithelial cells into basal stem cells would allow a durable repair of an epithelium (Figure 3E).

Directed Transdifferentiation

We would be remiss in discussing therapeutic approaches aimed at manipulating cell fate plasticity if we did not touch upon outright reprogramming or transdifferentiation in vivo (Srivastava and DeWitt, 2016). Of particular interest are efforts aimed at reducing the fibrosis that occurs after an infarct in the heart. If it were possible to convert cardiac fibroblasts directly into functional, beating cardiomyocytes, this would solve two problems at once. Indeed, direct reprogramming of cardiac fibroblasts using GATA4, MEF2C, and TBX5 after an infarct attenuated injury and reduced scar area (Qian et al., 2012). However, it appears that human cells require further factors for efficient repair (Patel et al., 2016). Similarly, research in reprogramming cardiomyocytes into functional pacemaker cells has yielded promising results in pig, where adenoviral delivery of TBX18 generated sufficient numbers of pacemaker cells to overcome complete heart block. Further research is required to refine this process, as the pacemaker function was transient, lasting only 2 weeks (Hu et al., 2014; Meyers et al., 2016). In the nervous system, the direct conversion of reactive glia into functional neurons would hold similar promise (reviewed in Li and Chen, 2016), and it is possible to imagine myriad opportunities in disease states where functional cells are lost, such as in type 1 Diabetes.

Conclusion

Differentiation therapy has already been effective in treating leukemias, but such approaches have only recently received attention in the treatment of solid tumors. Perhaps modulating plasticity may emerge as an important method to prevent EMT or recently identified gross cell fate conversions of one tumor type into another, which is increasingly recognized as a method of escape. We have laid out a map for imagining how encouraging or restraining cell fate plasticity could be deployed in pathologic settings, including those characterized by non-cancerous metaplasia or fibrosis. This points to the more general need to enumerate and understand what cell types exist in any given pathology, whether they are disproportionately represented by normal cell types or novel pathologic cell types that do not normally exist. Emerging technologies led by single cell transcriptomics should provide ample data in this regard. As with all therapies, encouraging beneficial plasticity may be associated with the risk of engendering pathologic cell plasticity, including cancer. Furthermore, any cell modulation therapy must be carefully deployed to act solely on a given tissue without altering the proportions of other necessary cell types of that tissue or other off -target tissues.

In this Perspective, Lin et al. highlight examples of plasticity during normal regeneration and in aberrant situations in a variety of tissues. The authors also discuss the merits of enhancing endogenous plasticity to promote regeneration as well as reversing pathologic plasticity as potential therapeutic strategies.