Stem cells and their unique properties play a fundamental role in development, transforming a single fertilized egg into hundreds of cell types comprising the complex tissues and organs of the human body. The key properties which characterize a stem cell are self- renewal, the ability to reproduce cells with the same potency as themselves, and differentiation potential, the ability to divide and give rise to differentiated offspring (Gupta, Pastushenko, Skibinski, Blanpain, & Kuperwasser, 2019). Potency is a stem cell’s potential to differentiate into various cell lineages. Over the course of embryonic development cells lose potency the further down the hierarchy they differentiate (Figure 1).

The top of the hierarchy is occupied by totipotent stem cells, capable of differentiating into all cell types, both embryonic and extraembryonic. Totipotent stem cells give rise to pluripotent stem cells with the ability to differentiate into all cell types arising from the three germ layers, the endoderm, ectoderm, and mesoderm (Singh, Saini, Kalsan, Kumar, & Chandra, 2016). Further progressing down the hierarchy, multipotent, oligopotent, unipotent, and stem cell progenitors all maintain self-renewal capabilities, but have significantly limited differentiation potential compared to their predecessors. Multipotent stem cells can differentiate into cell types within a close cell family, and oligopotent, unipotent, and progenitors are able to produce progeny within the same lineage (Hayes, Curley, Ansari, & Laffey, 2012). Traditionally, the unilateral differentiation down the hierarchy was viewed as the dogma of embryonic development, however research in the last decade challenges this, proposing more fluid and multi-directional differentiation behavior along the hierarchy.

Long after embryonic development, stem cells play a crucial role in the body, primarily through adult tissue homeostasis, the maintenance of a constant number of healthy cells in an organ, and wound repair, the dynamic process of restoring damaged tissue (Goichberg, 2016; Shaw & Martin, 2016). Adult tissue homeostasis is necessary to ensure optimal organ function, and tissue stem cells differentiate into the specific cell types needed to replace the cells lost and damaged from general “wear and tear”. This is especially important in tissues such as the epidermis and its appendages and the intestinal epithelia, both of which are highly vulnerable to environmental stressors, resulting in rapid turnover (Ojeh, Pastar, Tomic-Canic, & Stojadinovic, 2015; Visvader & Clevers, 2016). In a number of tissues, including the well characterized epidermis and small intestine, stem cells reside in stem cell niches, localized microenvironments where stem cell populations are maintained and stored until activation (Morrison & Spradling, 2008). The location of niches exemplifies the importance of stem cells in homeostasis as they are functionally housed in deep, well protected areas within a tissue. Examples of this include the aptly named crypt base columnar cells (CBCs) located at the bottom of the intestinal crypt, their highly debated neighbors at the “+4” position above the base, stem cells located in the basal layer of the epidermis, and hair follicle stem cells (hfSCs) located in the “bulge” at the base of the hair follicle (Nick Barker, 2014; Morrison & Spradling, 2008; Ojeh et al., 2015; Sakthianandeswaren et al., 2011). During homeostasis, these stem cells migrate and differentiate up and outwards to replenish lost and damaged cells. Instead of passively waiting in the niche until they are needed for replacement, stem cells are highly active in sensing and signaling to their surroundings and neighboring cells as well as maintaining their own population through self-renewal. Key developmental pathways Wnt and Notch have been implicated in these homeostatic processes (Clevers, Loh, & Nusse, 2014; Morrison & Spradling, 2008; Sancho, Cremona, & Behrens, 2015). These events display the dynamic process of homeostasis and the high levels of signaling and regulation, involved in ensuring proliferation and differentiation of stem cells at the appropriate time, number, and location (Biteau, Hochmuth, & Jasper, 2011).

In addition to homeostasis, adult tissue stem cells are crucial in regenerating tissues after injury. Stem cells are crucial to wound repair due to their capability to migrate to the site of tissue injury, rapidly proliferate, and to differentiate into the necessary cell types to restore the damaged tissue (Goichberg, 2016; Shaw & Martin, 2016). Following injury, stem cells in the hair follicle bulge, which do not migrate to the epidermis under normal conditions, migrate to epidermal sites of injury and aid in re-epithelializing wounds (Chou et al., 2013; Ito et al., 2005). Additionally, after migration and re-epithelialization, hfSCs expressed an epidermal phenotype but were eliminated in the tissue after several weeks, displaying their targeted migration to aid in wound repair. In contrast, muscle stem cells which are primarily quiescent compared to hfSCs, have also been shown to rapidly migrate and differentiate into regenerative myofibers following tissue injury, highlighting the importance of stem cell involvement in wound repair across multiple tissue types (Dumont, Bentzinger, Sincennes, & Rudnicki, 2015).

Wound repair is a highly dynamic process involving the coordination of chemical signaling, cell migration, and regulated growth and differentiation. The wound healing process is often broken up into three major phases—inflammation, proliferation, and remodeling. Mesenchymal stem cells (MSCs) are involved in all three stages, migrating to damaged tissue and interacting with the stroma and macrophage signaling. (Maxson, Lopez, Yoo, Danilkovitch-Miagkova, & LeRoux, 2012). As the first phase of the wound repair process, inflammation is an especially critical step as it guides the signaling and recruitment of effector cells necessary to execute downstream reconstruction of the extracellular matrix, angiogenesis, and remodeling to normal tissue. Inflammation is a self-limiting process, reliant on a highly sensitive balance of pro-inflammatory and anti-inflammatory cytokines, chemokines, and growth factors, where disruption of this balance has the potential to result in pathologies such as chronic inflammation and neoplasia (Coussens & Werb, 2002). MSCs and macrophages are especially important in helping wounds progress past the inflammation phase as MSCs decrease pro-inflammatory cytokines (TNF-α and IFN-γ) and increase production of anti-inflammatory cytokines (IL-10 and IL-4) and macrophages clear damaged and apoptotic leukocytes, resulting in the resolution of inflammation and allowing the wound to progress to the proliferation phase(Guo & DiPietro, 2010; Maxson et al., 2012). During proliferation and remodeling, MSCs release growth factors FGF, VEGF, PDGF, TGF-ß, and others to recruit fibroblasts, keratinocytes, and host stem cells to reconstruct the extracellular matrix and re-epithelialize the wound (Shi et al., 2012).

Chronic inflammation is a characteristic of the tumor microenvironment and is becoming increasingly implicated as a major mechanism involved in the initiation and progression of tumorigenesis. In the tumor microenvironment, inflammatory molecules such as IL-6, IL-8, TGF-β1, NFκβ, TNFα, and HIF-1α regulate proliferation and metastasis (Cabarcas, Mathews, & Farrar, 2011; Plaks, Kong, & Werb, 2015). The tumor microenvironment also works to actively immunosuppress natural killer (NK) cells and CD8+ T cells in order to allow continuous inflammation and tumor progression, which would otherwise be stopped by these cytotoxic cells (Plaks et al., 2015). The inflammatory infiltrate in tumors is maintained by a precise balance of pro-inflammatory and anti-inflammatory cytokines that stimulate angiogenesis and neoplastic growth (Coussens & Werb, 2002). Where normal tissues exhibit a highly organized structure utilizing angiogenesis and differentiation in wound repair to restore tissue organization, angiogenesis and neoplastic growth in invasive carcinomas occur in a highly disorganized manner reflecting a chaotic remodeling of the extracellular matrix. Indeed, the tissue reorganization field theory of cancer posits that alterations of the stroma and disrupted stromal/epithelial interactions are likely major drivers of carcinogenesis (Soto & Sonnenschein, 2011). More specifically, Soto and Sonnenschein show in multiple in vitro studies utilizing various cell lines that the dynamic interaction between mammary epithelial cells and the breast stroma is crucial to proper mammary gland morphogenesis and that this development is highly sensitive to alterations in structure and composition of the breast extracellular matrix (Barnes et al., 2014; Dhimolea, Soto, & Sonnenschein, 2012; Krause, Maffini, Soto, & Sonnenschein, 2010). Additionally, chronic inflammation results in the accumulation of damaging radicals such as reactive nitrogen and oxygen species secreted by macrophages and neutrophils during phagocytosis(Dedon & Tannenbaum, 2004; Kvietys & Granger, 2012). In a state of chronic tissue damage and regeneration, these reactive species have prolonged opportunity to interact with DNA and cause permanent damage and mutations, the accumulation of which could potentially lead to tumorigenesis. The phagocytic production of reactive radicals in response to foreign agents has also been proposed as a mechanism of initiation in cancers linked to infectious bacterial and viral agents such as human papilloma virus (HPV), hepatitis B and C viruses (HBV/HCV), H. pylori, and others(Morales-Sánchez & Fuentes-Pananá, 2014; Porta, Riboldi, & Sica, 2011).

These prior examples highlight the ability of stem cells to differentiate into cell types lost or damaged due to injury, but what happens when the stem cell populations themselves are depleted? Studies of the intestinal epithelium have identified two key mechanisms of recovery following depletion of CBCs, the highly proliferative stem cell population at the base of the intestinal crypt. CBCs have been characterized by high expression of the leucine-rich repeat containing G-protein coupled receptor 5 gene (Lgr5) and are responsible for generating secretory and enterocyte precursors which terminally differentiate and divide upwards to populate the gut (Tetteh et al., 2016). Using a transgenic mouse model knock in of the diptheria toxin receptor under the Lgr5 locus, Tian and colleagues showed that upon complete depletion of Lgr5+ CBCs, Bmi1 expressing quiescent stem cells at the +4 position underwent increased proliferation. Furthermore, lineage tracing showed that these Bmi1 expressing cells gave rise to Lgr5+ cells providing evidence that this population replenishes CBCs under injury and even normal conditions (Tian et al., 2011). An alternative mechanism to combat loss of CBCs is “dedifferentiation”, the process of a differentiated cell reverting to a less differentiated cell type within the same lineage (Tetteh, Farin, & Clevers, 2015a). Following depletion of CBCs by sub-lethal doses of radiation or ablation using the diptheria toxin receptor-transgenic mouse model, early secretory progenitors (Dll1^hi) and enterocyte progenitors (Alpi+) revert to stem cells (Tetteh et al., 2016; van Es et al., 2012). Tetteh and colleagues showed that upon depletion of both Lgr5+ CBCs and Bmi1 expressing +4 stem cell populations, Alpi+ enterocyte progenitors dedifferentiate into stem cells but that this dedifferentiation potential is limited to the 2–3 day window before Alpi+ cells have migrated out of the crypt and are no longer proliferative (Tetteh et al., 2016). This limited capacity for dedifferentiation highlights the intricacy of stem cell regulatory mechanisms and their dynamic interactions with extrinsic signals.

The dedifferentiation potential of intestinal progenitors demonstrates another key characteristic of stem cells - plasticity. Plasticity is the ability of stem cells to delineate from homeostatic behavior and adopt an alternate cell fate in response to cellular and environmental needs such as injury or stress (Ge & Fuchs, 2018; Tetteh, Farin, & Clevers, 2015b) (Figure 2). The epithelial to mesenchymal transition (EMT) and its reverse, the mesenchymal to epithelial transition (MET), serve as prototypical examples of plasticity, where cells undergo a phenotypic transition in response to environmental cues. EMT involves the modification of adhesive proteins and loss of apico-basal polarity allowing epithelial cells to acquire a more migratory and invasive phenotype with self-renewal capabilities (Scheel et al., 2011). During the gastrulation portion of embryogenesis, epithelial-like cells of the epiblast primary streak undergo EMT forming embryonic mesoderm which later gives rise to the primary mesenchyme (Kalluri & Weinberg, 2009). This process is tightly regulated by a combination of EMT transcription factors (Twist 1, Snail 1/2, Zeb 1/2;, and FOXC2), pro-inflammatory signals, cytokines, and signaling pathways Wnt and TGF-ß (Chaffer, San Juan, Lim, & Weinberg, 2016; Tsai, Donaher, Murphy, Chau, & Yang, 2012).

EMT and MET occur along a bidirectional continuum rather than a switch between two discrete states. Both immortalized human mammary epithelial cells induced for EMT and tumorigenic mammary cells have been found to co-express E-cadherin, an epithelial marker, and vimentin (VIM), a mesenchymal marker, indicating the presence of hybrid EMT/MET states and showcasing the gradient nature of these transitions (Mani et al., 2008; Yagasaki, Noguchi, Minami, & Earashi, 1996). Hybrid EMT/MET cells showed an increase in mammary stem cells compared to control (mammosphere assay) and the co-expression of E-cadherin and VIM in tumorigenic cells was associated with axillary metastases. These hybrid EMT/MET cells are also detectable in the normal mammary gland, and have a gene expression signature consistent with aggressive triple negative breast cancers (Colacino et al., 2018). “Partial EMT” has been observed in cancer epithelial cells, where partial transitions along the continuum are potentially more favorable to tumorigenesis than complete state shifts. The migratory characteristics acquired through EMT are favorable when carcinogenic cells leave the primary tumor site to metastasize, however upon arrival to a secondary tumor site, MET is necessary for cells to regain epithelial qualities in order to establish a metastatic colony and the outgrowths to support it (Chaffer et al., 2016). Indeed, carcinoma cells which have undergone a complete EMT program and lost all epithelial characteristics have been shown to be ineffective at proliferating and establishing metastases at secondary sites following dissemination. (Tran et al., 2014; Tsai et al., 2012). The necessity for EMT and plasticity in both embryogenesis and carcinogenesis highlights another striking similarity between these two processes.

The continuous nature of stemness observed during the EMT of early embryogenesis is also reflected in tissue specific differentiation. Studies using scRNA-seq have allowed us to observe these changes across the cellular landscape at single cell resolution (Giraddi et al., 2018; Macaulay et al., 2016; Nestorowa et al., 2016; Pal et al., 2017). These studies and others have pioneered a paradigm shift from the traditional view of differentiation as a unidirectional hierarchy of discrete stages to a dynamic and multi-directional continuum lacking clear boundaries between cell populations at varying stages of development (Y. Zhang, Gao, Xia, & Liu, 2018). This updated model emphasizes the intricate relationship between stem cells and environmental cues which guide the dynamic behavior of cells along the continuum (Fig1B). Using single cell transcriptomic expression to computationally order cells in a differentiation timeline (pseudotime), hematopoietic stem and progenitor cells isolated from mouse bone marrow and hematopoietic cells isolated from zebrafish kidneys both followed a continuous trajectory of differentiation directed by gradual changes in transcriptomic expression (Macaulay et al., 2016; Nestorowa et al., 2016). Single cell RNA sequencing analyses of bulk mouse mammary epithelial cells sampled at varying stages of the life course followed a continuous sequential trajectory from most immature to most adult after single cell sequencing and mapping, rather than grouping in distinct cellular states (Giraddi et al., 2018; Pal et al., 2017). While the datasets from these two studies were generated separately and sampled different developmental time points (Giraddi: embryonic day 16, embryonic day 18, post-natal day 4, adult) (Pal: post-natal 2 weeks, post-natal 5 weeks, adult), the single cell mapping of both datasets clearly depict a gradual differentiation continuum rather than distinct clusters from the different developmental stages. These computational methods have shed major insight into the differences in transitional transcriptomic changes that varying cell types undergo as they progress through the differentiation landscape, opening the door for future research to use these transcriptional profiles to further study these intermediate states. The continuous nature of stemness and plasticity in development and in carcinogenesis has made it challenging to identify a clear and universal transcriptomic signature for stem cells, likely due to tissue specific differences in stem cell regulation. Significant progress, however, has been made in identifying markers for specific stem cell populations within tissues.

The coordinated gene expression necessary for stem cell maintenance, differentiation, cellular dedifferentiation, and transdifferentiation are established by epigenetic changes, including histone modifications, DNA methylation, and non-coding RNAs. Embryonic stem cells are characterized by an overall open chromatin state (Meshorer & Misteli, 2006). In these cells, key developmental genes, which are transcriptionally silent, are marked by bivalent chromatin domains, containing both the activating mark H3 lysine 4 methylation and the repressive mark H3 lysine 27 methylation (Bernstein et al., 2006). Acquisition of this bivalency state requires simultaneous histone modifications by the MLL2 and Polycomb complexes (G. Mas et al., 2018) Accompanying the open chromatin state is a reduced amount of global DNA methylation in naïve stem cells (Bibikova et al., 2006) but an increased amount of DNA hydroxymethylation specifically at the locations of bivalent chromatin domains (Pastor et al., 2011). As stem cells differentiate, these epigenetic patterns are altered, characterized by a gain of DNA methylation in pluripotency and germline-specific genes (Mohn et al., 2008) and an accumulation of transcriptionally inactive heterochromatin (Francastel, Schübeler, Martin, & Groudine, 2000). The entire 3D genomic architecture shifts during differentiation, as the interactions between different chromatin domains is altered to allow for the expression of concurrently regulated gene networks essential for the functioning of the differentiated cell (Dixon et al., 2015). Throughout the differentiation process, there is a simultaneous coordinated regulation of non-coding RNAs, which also dynamically shift to regulate the expression of key developmental pathways (Dinger et al., 2008). The epigenetic processes which regulate stem cell pluripotency and differentiation are tightly regulated and require the coordination of the various known mechanisms.

The dynamic epigenetic changes that occur during dedifferentiation in the induced pluripotency process (Takahashi et al., 2007) have provided important insights into developmental programming and epigenetic dysregulation in cancer. The temporal epigenetic alterations that accompany embryonic stem cell differentiation are reversed during the induced pluripotency process, including DNA demethylation and acquisition of an open chromatin state (Takahashi & Yamanaka, 2015). Pluripotent stem cells are not, however, epigenetically equivalent to embryonic stem cells. Instead, these cells maintain an “epigenetic memory” of their tissue of origin, which survives through the reprogramming process (K. Kim et al., 2010). The epigenetic alterations during the reprogramming process have also shed light into CSC biology. Early genome-wide profiles of DNA methylation across multiple tumor types identified abundant enrichment of DNA methylation at sites required for differentiation in embryonic stem cells, marked by the Polycomb repressive complex, which helps lock tumor cells in a state of self-renewal (Widschwendter et al., 2007). Like embryonic stem cells, cancers are also typically characterized by a state of global DNA hypomethylation (Feinberg & Vogelstein, 1983). Extraction of epigenomic and transcriptomic feature sets from pluripotent stem cells and tumors using machine learning found that the most aggressive cancers have an epigenetic signature that resembles a stem cell (Malta et al., 2018). Further, metastasizing cells from breast cancer patients and mouse models of breast cancer in circulating tumor cell clusters are characterized by hypomethylation of key pluripotency related factors, such as OCT4, NANOG, and SOX2, reflecting a reacquisition of an embryonic stem-like phenotype in the most aggressive cancer cells (Gkountela et al., 2019). These results point to epigenetic alterations consistent with reprogramming to a stem cell state as a mechanistic driver of the cancer cell invasion and metastasis.

While the evidence for a subpopulation of cancer cells with stem cell like properties is growing, the cell of origin for CSCs within the normal tissue hierarchy has not been formally defined. We hypothesize that there are three potential origins for CSCs. First, normal stem cells could acquire genetic and epigenetic changes that confer the ability to inappropriately undergo symmetric self-renewal and initiate tumorigenesis. A second possibility is that CSCs derive from lineage committed rapidly cycling progenitor cells that undergo mutations that reconfer stem-like properties. A final possibility is that a series of mutations in fully differentiated cells can lead to dedifferentiation to a tumorigenic stem state. For example, expression of a mutated form of the oncogene PIK3CA in luminal and basal mammary gland cells leads to the reacquisition of a multipotent state (Van Keymeulen et al., 2015). Conversely, transient expression of the stem cell pluripotency factors Oct4, Sox2, Klf4, and Myc in genetically modified mice, absent any other genetic mutation, is sufficient to induce cancers of epithelial origin in various tissues with an embryonic stem cell-like gene expression pattern (Ohnishi et al., 2014). As the body of research surrounding both normal and CSCs grows, evidence for multiple pathways of CSC generation is accumulating.

A key feature of normal stem cells that has proven essential to isolate and characterize these cells is the expression of specific surface and enzymatic markers of stemness. Early studies of the cellular hierarchy of the breast identified that the expression of the cell surface markers MUC-1- to ±/CD-10 ± to +/ESA+ isolated cells with the ability to develop colonies with both luminal and myoepithelial features (John Stingl, Eaves, Kuusk, & Emerman, 1998). A follow-up study identified that normal bipotent mammary progenitors are enriched in a cellular subfraction expressing both CD49f (α6 integrin) and EpCAM (J. Stingl, Eaves, Zandieh, & Emerman, 2001). Others have identified that normal bipotent mammary stem cells are further enriched in the CD44+/CD24-fraction of CD49f+/EpCAM+ cells (Ghebeh et al., 2013). LGR5 is a well characterized marker of normal and cancer stem cells in multiple tissues (Leung, Tan, & Barker, 2018), and methods were recently published for the antibody-based isolation of LGR5-expressing normal and cancerous cells from human primary intestine and colonic organoids (Dame, Attili, McClintock, Dedhia, Ouillette, et al., 2018). In addition to stem cell enriching cell surface protein markers, enzymatic markers of stemness have also been identified. Normal breast stem and progenitor cells express high levels of aldehyde dehydrogenase 1 (ALDH1) (Ginestier et al., 2007). Similar results for enrichment of ALDH1 expression in stem cells were also found in normal and cancerous colon (E. H. Huang et al., 2009), head and neck cancers (Clay et al., 2010), and pancreatic cancer (M. P. Kim et al., 2011). ALDH1 expressing cells can be identified using the non-immunological Aldefluor assay, where the substrate, Bodipy-aminoacetaldehyde, is converted intercellularly to fluorescent Bodipy-aminoacetate. Stem cells can also be isolated by exploiting their increased expression of ATP-binding cassette drug transporters (Bunting, 2002). By staining cells with Hoechst, a DNA-binding dye that is effluxed by ATP-binding cassette transporters, one can discriminate populations of cells that are high and low Hoechst staining, with the low-Hoechst stained, stem cell enriched, fraction termed the “side population”.

Model organisms genetically engineered to express fluorescent reporters of stem cell regulators have allowed for in depth analyses of stem cells in development, tissue homeostasis, and disease. In particular, genetically engineered mice have been widely used to trace stem cell differentiation and understand the impact of alterations to the stem cell niche. In a seminal study, Barker and colleagues developed the Lgr5-EGFP-IRES-creERT2 knock-in mouse line, which allows for immunofluorescent identification of Lgr5+ stem cells and was used to provide substantial evidence of the essential role these cells have in generating the intestinal epithelium (N Barker et al., 2007). These mice have been used to assess the function of Lgr5+ cells in development and tissue homeostasis across a range of organ systems including mammary gland (Plaks et al., 2013), stomach (Nick Barker et al., 2010), liver (Planas-Paz et al., 2016), lung (Lee et al., 2017), prostate (B. E. Wang et al., 2015), and ovary (Ng et al., 2014). Directed knockout of the Wnt pathway activator APC in these intestinal Lgr5+ cells led to the rapid development of adenomas, suggesting that these cells are a likely cell of origin in APC driven colon cancers (Nick Barker et al., 2009). Directed knockout of the tumor suppressor gene p53 in Lgr5+ cells in mouse treated with a DNA damaging agent, azoxymethane, and an inflammation inducing agent, dextran sodium sulfate, showed that stem cell specific deletion of p53 significantly increases colon tumor size and incidence in this inflammation and DNA-damage driven model (Davidson et al., 2015). Additionally, the differentiation hierarchy of the hematopoietic system is very well characterized, and transgenic mice with various reporters for cell types across this hierarchy have been developed (Vacaru, Vitale, Nieves, & Baron, 2014). These models hold significant promise to provide fundamental understanding of how environmental factors perturb stem cell biology en route to cancer formation. Specifically, they allow for the interrogation of stem cells in situ through microscopy or flow cytometry, are amenable to further genetic manipulation to either drive Lgr5 reporter expression in specific tissues or knock out cancer-related genes, and allow for the assessment of stem cell function in the presence of the entire stem cell niche.

A number of functional assays for the identification and classification of both normal and cancer stem cells have been established. The first series of experiments to identify the presence of tumor initiating cells utilized a transplantation assay of human acute myeloid leukemia cells into severe combined immunodeficient (SCID) mice (Lapidot et al., 1994). The SCID mice were examined for the presence of human leukemia cells, with limiting dilution experiments identifying that approximately 1 in 250,000 cells have the ability to engraft, with the cells most likely to engraft possessing the CD34+CD38-hematopoietic progenitor signature. In the human breast, cancer stem cells were first identified by injecting single cell suspensions of dissociated human breast tumor tissue into NOD/SCID mice (Al-Hajj, Wicha, Benito-Hernandez, Morrison, & Clarke, 2003). Tumor cells were sorted based on cell surface markers (CD44, CD24, EpCAM) and injected into the mammary fat pads of mice at limiting dilutions, revealing that as few as 200 ESA+/CD44+/CD24-cells were consistently able to form tumors, while injection with 20,000 CD44+/CD24+ cells failed to grow tumors (Al-Hajj et al., 2003). Furthermore, the ESA+CD44+CD24-cell fractions were able to recapitulate the original tumor phenotype after serial transplantations, showing that these cells possess the ability to both proliferate and differentiate into the different cell types that comprised the original tumor. In general, the serial reimplantation assay is considered the “gold standard” assay in the field to identify cancer stem cells from human tumors, although the assay has a number of potential weaknesses, including the injection of dissociated single cell suspensions rather than complete microenvironments and the lack of a competent immune system in the host animal (Rycaj & Tang, 2015).

In addition to the transplantation assays described above, a number of other assays have been utilized to enrich and characterize normal and cancer stem cells. Neural stem cells were first discovered to grow in anchorage independent, serum free culture conditions forming free floating spheroids of neural cells termed neurospheres (Reynolds & Weiss, 1996). These culture conditions were later adapted for mammary tissue, where both tumor and normal breast stem/progenitor cells were found to propagate under these conditions, termed mammosphere formation conditions (Dontu et al., 2003). Spheroid conditions have now been applied to study stemness and 3D structure in a range of different normal and tumor types (Colacino, 2016). Tumorsphere formation is useful for characterizing three key aspects of stem cell biology: (1) proliferation potential; (2) the ability to self-renew; and (3) the ability to differentiate into downstream progeny. Since each tumorsphere is initiated by a single cell, proliferation capacity can be assessed by tumorsphere size. Tumorspheres can then be serially passaged to assess stem cell self-renewal capacity over time. Finally, tumorspheres can be stained for known markers of lineage differentiation, or plated into differentiating culture conditions to assay the potency of the stem cell population.

Human organoid cultures afford advantages over spheroid models as they allow for the long-term propagation of stem cell enriched patient derived samples in a physiologically relevant extracellular matrix. By varying the composition of the matrix, different properties relevant to cancer stem cells can be altered, including proliferation, invasion, and dissemination (Shamir et al., 2012). Biobanks of normal and cancerous patient derived organoids have been established for breast (Sachs et al., 2018), colon (Dame, Attili, McClintock, Dedhia, Ouilette, et al., 2018; Fujii et al., 2016; van Sluis et al., 2015), and prostate (Beshiri et al., 2018; Gao et al., 2014). These organoid biobanks are comprised of tissues which span the spectrum from normal, to early stage, to advanced cancers, which provides unique resources to understand alterations in stem cell biology through carcinogenesis. New experimental techniques are being developed to improve the long term survival, cellular diversity, and physiological relevance of these 3D culture systems (Fujii et al., 2018; Neal et al., 2018). While experimental analyses of environmental effects on stem cell biology in organoid systems are still in their infancy, these models have been used to assess the impact of dietary fat, essential nutrients, and toxic heavy metals on stem cell regulation relevant to tumorigenesis (Beyaz et al., 2016; McClintock et al., 2018; Rocco et al., 2018).

While genetic risk factors for cancer have been well characterized and described, an estimated 70–90% of lifetime cancer risk is estimated to derive from exposures to extrinsic factors (Wu, Powers, Zhu, & Hannun, 2016). Similar etiologic contributions of environmental risk factors have been estimated for other chronic diseases (Rappaport, 2011). Although there is debate within the literature, high profile studies have also identified that the number of stem cells, and stem cell divisions, in a tissue is a strong predictor of the probability of developing cancer in that tissue (Albini, Cavuto, Apolone, & Noonan, 2015; Tomasetti, Li, & Vogelstein, 2017; Tomasetti & Vogelstein, 2015). The ‘bad luck’ theory of carcinogenesis by Tomasetti and Vogelstein is still controversial due to potential limitations in the study including the cancers studied and the fact that association does not mean causation (Albini et al., 2015). Additionally, emerging evidence points to the number or proportion of stem cells in a given normal tissue between different individuals is highly variable (Colacino et al., 2018; Kumar et al., 2018; Nakshatri, Anjanappa, & Bhat-Nakshatri, 2015). Studies of patient tissues have identified some of the predictors of tissue stem cell number. Women who inherit BRCA1 mutations have a higher proportion of stem cells in their non-cancerous breast tissue (Honeth et al., 2015; Merajver et al., 2008). Obesity is associated with an increase in the number of functional mammary epithelial progenitor cells in breast tissue from healthy women (Chamberlin, D’Amato, & Arendt, 2017). Quantifying the impacts of both intrinsic and extrinsic risk factors on stem cell function will be essential to understanding why people get cancer.

A growing body of experimental literature is showing that environmental exposures during development can have lasting impacts on stem cells, with important implications for cancer risk. The concept of windows of susceptibility proposes that exposures during critical time points, such as early life development, can result in life-long effects due to the vulnerability of developing organ systems (Birnbaum & Fenton, 2003). Young children and developing fetuses are especially vulnerable due to incomplete or the lack of development of protective mechanisms such as DNA repair, immune system function, detoxifying enzymes and liver metabolism, and the blood/brain barrier (Schug, Janesick, Blumberg, & Heindel, 2011). The in utero period of development is a period of heightened susceptibility due to the rapid metabolism and proliferation of developing organs. Embryonic development is a stage of extreme vulnerability to environmental exposures which may signal and interfere with epigenetic programming, which has been shown to persist in daughter cells and even trans-generationally (Leslie, 2013; Skinner, Manikkam, & Guerrero-Bosagna, 2010; Zeybel et al., 2012). Impaired organ development early on in life has been shown to alter development at critical downstream time points such as puberty and pregnancy (Rudel, Fenton, Ackerman, Euling, & Makris, 2011). Understanding that timing is key in the damaging potential of environmental exposures on stem cell populations will ultimately lead to efforts to prevent adverse exposures and the resultant effects in the most vulnerable populations—pregnant women, developing fetuses, and young children.

Tobacco smoking is associated with death from multiple cancers, including lung, colorectal, stomach and liver, and tobacco smoking is estimated to cause approximately 10 million deaths per year (Proctor, 2001; Torre et al., 2015). The genotoxic effects of exposure to the complex mixture of chemicals in tobacco smoke are well characterized (DeMarini, 2004). There is emerging evidence that tobacco smoke and nicotine can modify stem cell associated pathways relevant for carcinogenesis across multiple organ systems. In mice exposed to tobacco smoke with total particulate matter of 85 mg/m³ for 12 months using a smoking apparatus, histopathological analyses of the liver tissues identified changes consistent with an EMT and increased Oct4 and Nanog expression, which was linked to an increase in IL-33 expression and phosphorylated p38 (Xie et al., 2019). Co-exposure of mice to tobacco smoke and the p38 inhibitor SB203580 (1 mg/kg body weight) for 12 weeks attenuated the tobacco smoke-induced upregulation of stemness factors CD133, Nanog, and Oct4 (Xie et al., 2019). Exposure of non-tumorigenic (MFC10A and MCF12A) and tumorigenic (MCF7) breast cell lines to tobacco smoke extract for 72 weeks or cigarette smoke condensate for 40 weeks led to dose-dependent acquisition of an EMT-like phenotype, increased anchorage independent growth, and increased metastatic dissemination when implanted into immunocompromised mice (Di Cello et al., 2013). Exposure of two head and neck squamous carcinoma cell lines (UMSCC10B and HN-1) to 0.1, 1, or 3mM nicotine for 6 weeks similarly caused a dose-dependent increase in the expression of EMT-associated markers Snail, Twist, and Vimentin (Yu et al., 2012). These changes were accompanied by the induction of stem cell markers Oct4, Nanog, CD44, and BMI-1, although the highest induction of these factors was observed at the 1mM nicotine exposure(Yu et al., 2012). In the lung, long term exposure of immortalized human bronchial epithelial cells to 2% cigarette smoke extract for 25 weeks caused an increase in expression of cancer stem cell associated genes, including CD133 and ALDH1 and an increase in spheroid formation and upregulation of Wnt pathway members p-GSK3β, β-Catenin, and C-Myc (J. Wang et al., 2018). Nicotine (2μM) or e-cigarette extract exposure for around 24 hours in non-small cell lung cancer cell lines A549 and H1650 caused upregulation of Sox2 and an EMT-like phenotype through activation of nicotinic acetylcholine receptors, Yap1, and E2F1 (Schaal, Bora-Singhal, Kumar, & Chellappan, 2018). Early life tobacco smoke exposure through second hand smoke has also been associated with an increased risk of later life lung cancer, and that this was modified by a polymorphism in the dopamine receptor gene DRD1 (Ryan et al., 2013). Developmental tobacco smoke exposure is also known to cause very consistent DNA methylation alterations in blood (Joubert et al., 2016), which persist into adulthood (Richmond, Suderman, Langdon, Relton, & Smith, 2018). These same DNA methylation alterations were also enriched in lung cancer tissues isolated from smokers from the Cancer Genome Atlas (Bakulski, Dou, Lin, London, & Colacino, 2019). Finally, exposure of immortalized human urothelial cells to cigarette smoke extract for 20 weeks led to an increase in anchorage independent growth, an increase in growth in immunocompromised mice, increased expression of EMT markers and upregulation of Wnt/β-catenin signaling (Liang et al., 2017). Treatment with curcumin, a polyphenol derived from turmeric, led to a downregulation of Wnt signaling and a reversal of the EMT phenotype associated with the cigarette smoke extract (Liang et al., 2017). Curcumin has been shown to have similar stem cell targeting effects in other tissues as well (Colacino, McDermott, Sartor, Wicha, & Rozek, 2016; Kakarala et al., 2010; Zang, Liu, Shi, & Qiao, 2014), suggesting that this compound may have value as a stem cell differentiation agent for cancer prevention and treatment.

Numerous epidemiological studies have previously connected endocrine disruption with an increase in cell stemness. In a study assessing the epigenetic effects of hormonal exposures, 120 (0.5%) of all CpG sites analyzed were hypermethylated in epithelial cells exposed to estrogen (Rodriguez et al., 2008). Of these 120 loci, 111 were methylated in the transcription start sites of genes (Rodriguez et al., 2008). In this same study 23% of the methylated targets were also Polycomb group proteins which direct pluripotency of stem and progenitor cells (Rodriguez et al., 2008). Key developmental regulators appeared to also be epigenetically targeted by estrogen since 8 of the loci with hypermethylation were also tumor suppressor genes that are down-regulated in cancers (EGR2, FANCF, MXI1, PTPRG, RPRM, RUNX3, TFAP2C, and WNT5A) (Rodriguez et al., 2008). The results highlight the impact of estrogen signaling, and its potential for dysregulation by EDCs, in stem cell programming.

PFAS are a class of surfactants used industrially and commercially and are becoming of increasing concern to human health due to their environmental persistence even after being phased out of production. Most commonly used are perfluorooctane sulfonate (PFOS) and perfluorooctanoic acid (PFOA) whose endocrine disrupting capabilities have recently been studied both in vitro and in vivo. Exposure of T47D hormone-dependent breast cancer cells to PFOS or PFOA (10⁻¹² to 10⁻⁴ M) alone showed no activation of estrogenic activity and no effect on cell proliferation (Sonthithai et al., 2016). Interestingly, co-incubation of T47D cells with PFOS (10⁻¹⁰ to 10⁻⁷ M) or PFOA (10⁻⁹to 10⁻⁷ M) and 1 nM of 17β-estradiol (E2), resulted in enhanced E2-induced estrogen response element transcriptional activation, E2-induced proliferation, ERK 1/2 activation, and upregulation of E2 responsive gene pS2 (Sonthithai et al., 2016). Similarly, human breast epithelial cells (MCF10A) exposed to PFOA (100μM) or PFOS (10μM) showed no change in ERα or ERß protein expression relative to control (Pierozan & Karlsson, 2018). However, PFOA (50 and 100 μM) and PFOS (1 and 10 μM) enhanced cell proliferation through cell cycle acceleration of the G₀/G_1-to-S phase transition and stimulated cell migration and invasion (Pierozan & Karlsson, 2018), demonstrating the potential for these chemicals as inducers of neoplastic transformation. Additionally, 30 minutes pre-incubation of PFOA treated cells with estrogen receptor antagonist ICI 182,780(100nM) showed no effect but pre-incubation with PPARα antagonist GW 6471 (1μM) prevented MCF-10A proliferation, indicating that PFOA likely works through a PPARα mechanism instead of an estrogenic pathway (Pierozan & Karlsson, 2018). Female offspring of CD-1 and C57Bl/6 mice exposed to low dose PFOA (0.1–1mg/kg) during gestational days 1–17 experienced delays in pubertal mammary gland development exhibited by defects in terminal end budding and branching (Tucker et al., 2015). White and colleagues also showed that gestational PFOA exposure of 5 mg PFOA/kg BW/day exposure by oral gavage (GD 1–17, 8–17, or 12–17) alters mammary epithelial differentiation in CD-1 dams in addition to female pups (White et al., 2006). In dams, PFOA exposure resulted in delayed mammary differentiation and epithelial involution as well as alterations in milk protein gene expression. PFAS exposure poses significant risk to human health, particularly during crucial stages in development, and this body of work displays the need to prevent and regulate PFAS exposure.

Based on the US National Health and Nutrition Examination Survey (NHANES) data, oral HPV has a prevalence of 6.9% in men and women from 14 to 69 years of age (95% CI: 5.7–8.3%) (Gillison et al., 2012). HPV is the primary cause of a subset of oropharyngeal squamous cell carcinomas (OSCCs) (Gillison et al., 2012). There are over 160 strains of HPV, and HPV16 and HPV18 are considered the most oncogenic with 50 to 90% of HPV-positive head and neck squamous cell carcinomas (HNSCCs) being due to HPV16 (Gillison et al., 2012; Pullos, Castilho, & Squarize, 2015). Using the ALDH as a biomarker of cancer cell stemness in OSCCs related to HPV16, there was a 62.5 fold greater frequency of CSCs in HPV16-positive OSCCs than HPV negative OSCCs (M. Zhang et al., 2014). When isolating CSCs in culture from HPV-positive and HPV-negative HNSCCs, there was, however no statistically significant difference in CSC proportions (Tang et al., 2013). In general there is still some uncertainty within the field of HPV-positive HNSCCs and CSC occurrence and further research is needed (Pullos et al., 2015).

In HPV-associated cervical cancers, there are mutations that are associated with CSCs and HPV (Pullos et al., 2015; Tyagi et al., 2016). High-risk HPV strains HPV16 and HPV18 code for viral oncoproteins E6 and E7 that interfere with the regulatory proteins p53 and retinoblastoma (Münger et al., 1989; Tyagi et al., 2016; Werness, Levine, & Howley, 1990). The HPV oncoprotein E6 has been found to control stemness and self-renewal through upregulation of HES1 in cervical cancer cells from both primary tissue and cell line (HeLa) xenografts into mice (Tyagi et al., 2016). HES1 is a downstream gene of Notch1 which is responsible for stemness, and through blocking E6 in both primary tissue and cell line xenografts, HES1 became blocked (Tyagi et al., 2016). Other factors associated with stemness of cervical CSCs, like ABCG2 receptor and cervicosphere formation in culture, were also increased in HPV-positive cervical cancers (Tyagi et al., 2016).

Hepatocellular carcinomas (HCC) are cancers primarily due to Hepatitis B or C infection, but are also associated with liver disease and alcohol abuse (Torre et al., 2015; Xiong et al., 2019). Overall in developing countries 32% of infection-related cancers are due to hepatitis B (HBV) and hepatitis C (HCV), whereas in more developed countries, HBV and HCV accounts for 19% of all infection related cancers (Xiong et al., 2019). In a study aimed at determining HCV and HBV mechanisms for creating aberrant transcriptional enhancers, researchers measured hypomethylation among HCC patients in Hong Kong at the Prince of Wales Hospital (Xiong et al., 2019). When observing the methylome of HCC patients, specific enhancer C/EBPβ hypomethylation was correlated with a shorter survival rate (HR: 4.4, p <0.005) (Xiong et al., 2019). When C/EBPβ was deleted, the global enhancer activity reduced causing less invasion and colony formation (Xiong et al., 2019). Clinicopathological assessment of human liver tumors has also consistently identified an expansion of stem-like oval cells in hepatitis-associated cancers (Hsia, Evarts, Nakatsukasa, Marsden, & Thorgeirsson, 1992). Oval cell activation is an important first step in liver regeneration, as oval cells are bipotent cells that can differentiate into both bile duct epithelial cells and hepatocytes (Erker & Grompe, 2008). A major hypothesized mechanism for this expansion is the increase in inflammatory signaling within the stem cell niche, driven by infiltrating immune cells (Libbrecht, Desmet, Van Damme, & Roskams, 2000)