Arsenic Exposure Transforms Human Epithelial Stem/Progenitor Cells into a Cancer Stem-like Phenotype

1Accumulating evidence indicates that cancer stem cells (CSCs) probably arise from normal stem cells (NSCs), or partially differentiated progenitor cells, and that CSCs are likely the driving force in tumor initiation, growth, and progression. This developing theory is compelling because CSCs and NSCs share key characteristics, including the capacity for limitless self-renewal, which normally allows pluridirectional replenishment via differentiation but during oncogenesis contributes to heterogeneous aberrant cell overgrowth. Moreover, NSCs are conditionally immortal and generally thought to be innately resistant to most toxic insults, likely facilitating survival selection and accumulation of molecular lesions required for acquired malignant phenotype, consistent with the multistep carcinogenesis model. Although CSCs share characteristics with tissue-concordant NSCs, they display dysregulated self-renewal. Like NSCs, CSCs often form free-floating “spheres” of viable cells in vitro. Although assumed, proof of the direct emergence of epithelial CSCs from an NSC population is not available, and the identity of the cells that acquire the molecular lesions initiating chemical carcinogenesis remains undefined. Typically, to date, CSCs have been isolated from advanced tumors based on stem cell (SC)-like characteristics, rather than contemporaneously with carcinogenic initiation. Defining the role of CSCs in this primary triggering event is critical to a complete understanding of oncogenesis.

2Inorganic arsenic is a widely distributed, naturally occurring environmental contaminant affecting tens of millions of people worldwide. Human or rodent arsenic exposure causes various urogenital system (UGS) cancers, including urinary bladder and kidney cancer. Evidence in human populations in parts of Asia, the United States, and Australia exposed to arsenic suggests that the metalloid can target the prostate during arsenic-induced carcinogenesis; in vitro, inorganic arsenic causes malignant transformation of various cells, including human prostate cells. Arsenic also has transplacental carcinogenic activity in mice and possibly in humans, with targets within the UGS. Because of relative fetal abundance and their role in organogenesis and differentiation, SCs appear to be key targets in transplacental carcinogenesis and are likely targets in transplacental arsenic carcinogenesis. Arsenic disrupts human and rodent skin SC dynamics in vivo and in vitro, eventually resulting in SC/CSC overabundance, an event probably significant to skin cancer development. Although arsenic is unequivocally carcinogenic, the mechanisms involved, including the precise target cells, remain enigmatic.

3Leukemia SCs (LSCs) can be forced to arise from hematopoietic SCs (HSCs) after molecular manipulations of leukemogenesis genes. However, it is unclear if human epithelial CSCs arise in a fashion similar to LSCs or if chemical carcinogens act through an attack on NSCs. Compared with other known carcinogens, a model prostate epithelial NSC population possesses a remarkable survival selection advantage toward arsenic over their differentiated isogenic counterparts (Tokar EJ, Waalkes MP, unpublished data), which would afford perpetuation but continued targeting during arsenic insult. Thus, we tested the hypothesis that arsenic targets prostate SCs as a central event in oncogenesis. For this, we selected the human SC/progenitor line WPE-stem derived from its parental, heterogeneous mature prostate line, RWPE-1, originally derived from normal prostate epithelium. WPE-stem cells have well-defined prostate SC/progenitor cell characteristics, including high p63 (tumor protein p63) and keratin 5/14 expression, and low androgen receptor, prostate-specific androgen, and K18 (keratin 18) expression. WPE-stem cells form free-floating spheres (prostaspheres) in culture and have a high proliferation rate and relatively high colony-forming ability in agar, typical NSC characteristics. WPE-stem cells also highly express the prostate SC markers ABCG2 [ATP-binding cassette, subfamily G (WHITE), member 2] and BMI-1 [BMI1 polycomb ring finger oncogene] [see Supplemental Material, Figure 1, available online (doi:10.1289/ehp.0901059.S1 via http://dx.doi.org/)]. We used arsenic at a concentration that approximates levels in drinking water in areas where arsenicosis is common; and cells were exposed continuously, consistent with human exposure.

4RWPE-1 is a human papillomavirus (HPV)-18–immortalized, nontumorigenic, prostate epithelial cell line derived from normal adult human prostate. WPE-stem cells were isolated from RWPE-1 cells by single-cell dilution cloning, are nontumorigenic, and have extensive SC/progenitor cell characteristics [; see also Supplemental Material, Figure 1 and Table 1 (doi:10.1289/ehp.0901059.S1)], making them a good model for NSCs. Cells were maintained as previously described, passaged once weekly (preconfluence), continuously exposed to a nontoxic arsenic concentration (5 μM, as sodium arsenite; Sigma, St. Louis, MO) for up to 18 weeks, and compared with unexposed passage-matched controls. Three separate flasks for each treatment were maintained.

5We assessed in vitro biomarkers of carcinogenic transformation every 3 weeks. Secreted matrix metalloproteinase-9 (MMP-9) activity was examined using conditioned medium as previously described. Colony-forming efficiency (CFE) in soft agar, performed as previously described, was used to assess acquired malignant phenotype. CFE also correlates with NSC or CSC phenotype. We examined cellular invasiveness, another common characteristic of malignant cells, using a modified Boyden chamber invasion assay as described by. When in vitro biomarkers of acquired cancer phenotype became sufficiently positive, in vivo xenograft studies were performed to establish malignant transformation.

6Free-floating spheres of viable cells in culture are a characteristic of SCs and CSCs. To examine effects of arsenic on sphere formation, we plated cells (7.5 × 103) in T-75 flasks; cells were then grown for 10 days and fed every 48 hr. Total spheres were then counted.

7Free-floating spheres were collected from flasks and dissociated into single-cell suspensions by pipette trituration then passage through a 40-μm cell strainer (BD Biosciences, San Jose, CA). Cells were suspended in 250 μL Matrigel (BD Biosciences) in 48-well plates and placed in incubators for 24 hr. Matrigel was then covered with 300 μL growth medium, which was changed every 48 hr. Images were taken 2 weeks later.

8Total RNA was collected using TRIzol reagent (Invitrogen, Carlsbad, CA) and purified with RNeasy Mini Kit columns (Qiagen, Valencia, CA) according to the manufacturer’s protocol, and transcribed with MuLV (Moloney murine leukemia virus) reverse transcriptase and oligo-dT primers. Primers were designed with ABI Primer Express software (version 2.0; Applied Biosystems, Foster City, CA). We used SYBR Green Master Mix (ABgene, Rockford, IL) for real-time PCR (polymerase chain reaction) analysis. Cycle times were normalized with β-actin and glyceraldehyde-3-phosphate from the same sample and normalized to passage-matched controls. Genes examined included p63 [tumor protein p63 (TP63)], BMI-1, ABCG2, SHH (sonic hedgehog), OCT-4 [POU class 5 homeobox 1 (POU5F1)], NOTCH-1 [Notch homolog 1, translocation- associated (Drosophila)], K5 [keratin 5 (KRT5)], K18, and PTEN (phosphatase and tensin homolog). For gene and primer information, see Supplemental Material, Table 2 (doi:10.1289/ehp.0901059.S1). For Western blot analysis, protein extracts were collected using either NE-PER or M-PER extraction reagents (Pierce, Rockford, IL), separated by SDS-PAGE, transferred to polyvinyl difluoride membranes, and probed with anti-ΔNp63 (Santa Cruz Biotechnology, Santa Cruz, CA) or anti-cytokeratin-18 (Sigma) followed by horseradish peroxidase (HRP)-conjugated secondary antibodies (Santa Cruz Biotechnology). Membranes were stripped and reprobed with anti-β-actin antibody (Calbiochem, San Diego, CA) followed by peroxidase-conjugated anti-mouse (Santa Cruz Biotechnology). We used the ImageJ analysis program (version 1.24; ) for densitometric analysis.

9We conducted immunocytochemistry for p63 protein as described previously. Chamber slides (BD Biosciences) were coated with matrix, and cells were plated and grown to subconfluence, washed with phosphate-buffered saline, and fixed with a 1:1 solution of methanol:acetone for 2 min. For staining, cells were blocked with normal horse serum (1 hr), incubated with primary antibody (ΔNp63, 1:100, 1 hr), rinsed, incubated with HRP-conjugated anti-mouse secondary antibody (1:500, 1 hr), and rinsed again. Staining was developed with ImmPACT DAB (diaminobenzidine) substrate (Vector Labs, Burlingame, CA), which produces a brown reaction product. Incubations were done at room temperature.

10Once biomarkers of carcinogenic transformation suggested that arsenic-induced carcinogenic conversion had occurred, arsenic-treated and control cells were injected into male nude mice (NCr-nu; Charles River Laboratory, Wilmington, MA). Mice were housed at the NCI-Frederick animal facility (Frederick, MD), and animal care was provided in accordance with the Public Health Service policy on the care and use of animals. For the first study, 1 × 106 total cells (floating spheres and adherent cells) were collected and injected under the renal capsules of mice. In the second study, arsenite-treated floating cells were first separated from adherent cells and 1 × 106 cells from each subgroup was injected subcutaneously (dorsal thoracic midline) into separate groups of mice. Animals were observed for tumor formation over a 6-month period.

11All data except tumor incidence represent mean and 95% confidence interval (CI). We used an unpaired Student’s t-test to compare arsenic-treated cells with untreated time-matched controls at individual time points. A Fisher’s exact test was used for tumor incidence data. In all cases, we consider a two-sided p < 0.05 significant. Sample sizes are given in figure legends.

12MMP-9, an enzyme that when secreted digests extracellular matrix to aid in invasion and metastasis typical of cancer cells, including those produced by arsenic, showed marked increases between 16 and 18 weeks of arsenic exposure in WPE-stem cells (Figure 1A). Marked increases in cellular invasion and CFE, which are additional characteristics of malignant cells, occurred concurrently (Figure 1B,C). In Matrigel, the WPE-stem cells formed ductal-like structures from single cells (Figure 1D), which is typical for NSCs from various tissues, but the arsenite-treated structures grew much faster, were much more branched, and hence appeared much more “aggressive.” Together, these characteristics show that WPE-stem cells rapidly acquired a cancer phenotype with arsenite exposure.

13To establish malignant transformation, we inoculated cells under the renal capsules of mice. Arsenite-treated cells (As-CSCs) rapidly developed into highly pleomorphic tumors, with regional invasion and distant metastases that often dictated euthanasia in as little as 2–3 weeks. Tumors were highly undifferentiated, highly malignant, and composed of immature epithelial- and mesenchymal-like cells (Figure 2A,B). Strong heterogeneous staining for human K5 (Figure 2C), a marker of putative prostate epithelial SC/progenitor cells, supports a CSC-like nature and pluripotent cell of origination for these cancers. Tumors were very aggressive, showing invasion into renal parenchyma, abdominal muscle, adrenals, stomach, and small intestine, and very frequently metastasized to the lungs. During the 6-month observation period, no tumors occurred in mice inoculated with untreated cells, whereas tumor incidence was > 40% after inoculation with As-CSCs (Figure 2D).

14A common characteristic of SCs in culture is the formation of floating “spheres” of viable cells. Malignantly transformed As-CSCs showed a 230% increase in free-floating spheres over control (Figure 3A). After separating floating and adherent cells, secreted MMP-9 activity from arsenite-exposed sphere cells was approximately 6 times that of either control (sphere or adherent) or arsenic-exposed adherent cells (Figure 3B). CFE of arsenite-exposed sphere cells was markedly elevated compared with adherent arsenite-exposed cells or either adherent or sphere control cells (Figure 3C). The cells driving tumor formation were present only in arsenite-exposed spheres (Figure 3D), because when sphere cells were first separated from adherent cells and then inoculated into mice, only arsenite-exposed sphere cells rapidly produced highly pleomorphic cancers (Figure 2D). Mice inoculated with arsenite-exposed adherent cells or pooled control cells showed no tumor formation when euthanized 6 months after inoculation.

15We examined expression of several NSC self-renewal genes during arsenite-induced CSC phenotype acquisition. A remarkably consistent temporal pattern emerged in which NSC self-renewal gene expression was first lost and then reactivated with arsenite-induced acquired CSC-like phenotype (Figures 4 and 5). As arsenite transformed NSCs into As-CSCs, the expression of self-renewal– associated genes p63 (Figure 4), BMI-1, ABCG2, SHH, OCT-4, and NOTCH-1 (Figure 5A–E) was first markedly suppressed (~ weeks 1–9) then reactivated (~ weeks 9–18). Additionally, K5, a marker of undifferentiated prostate SC/progenitor cells, showed the same pattern (Figure 5F). Conversely, K18, a marker of differentiated cells, showed the opposite trend of NSC self-renewal genes, first increasing and then returning to basal levels with transformation (Figure 6). This indicates that SC self-renewal was initially lost during early transformation, potentially due to aberrant differentiation, and then regained with malignant transformation, consistent with distorted self-renewal common to CSCs. This is consistent with CSC overproduction induced by arsenite and the aggressive, highly pluripotent tumors formed by As-CSC cell inoculation. WPE-stem cells normally show homogeneous expression of prostate SC marker p63, and immunocytochemical assessment showed that, although exposure-duration–related expression changes occurred, expression remained homogeneously distributed throughout arsenite-induced transformation (Figure 4C), indicating that subpopulation selection did not occur.

16PTEN, a tumor suppressor gene that modulates cell signaling pathways, has critical roles in SC differentiation and is often inactivated in malignancies. Starting at 9 weeks of arsenite exposure, a rapid, progressive decrease in PTEN expression occurred (Figure 7). This decrease coincided with the reactivation of the SC self-renewal genes, indicating that these events are likely linked.

17Human arsenic exposure is linked to various UGS tumors, potentially including prostate cancers, and emerging data indicate that arsenic may have human transplacental carcinogenic activity. Similarly, prenatal arsenic exposure in mice induces or initiates neoplastic lesions throughout the UGS in adulthood, and because SCs may be an important target in transplacental carcinogenesis, this suggests that UGS NSCs are key targets of arsenic. Using a model prostate epithelial NSC population, we found that the SC phenotype appears to have a strong survival selection advantage toward arsenite (Tokar EJ, Waalkes MP, unpublished data), likely allowing for survival in the face of continued carcinogenic insult. Based on these observations, in the present study an established human prostate SC/progenitor cell line was chronically exposed to a nontoxic, environmentally relevant concentration of arsenite and rapidly (18 weeks) produced malignant transformant cells that showed multiple characteristics of CSCs, proved highly aggressive, and produced xenograft tumors indicative of a remarkably pluripotent cell of origin. This is compelling evidence that the metalloid can act directly upon a human SC/progenitor population to initiate formation of CSC-like cells and provides unique insights into the very earliest period of arsenic carcinogenesis. In contrast, the same level of arsenite will induce transformation in its isogenic, mature heterogeneous parental RWPE-1 cell line, but this takes 30 weeks, and the cells form much more differentiated tumors of epithelial origin with mature prostate qualities [; see Supplemental Material, Table 1 (doi:10.1289/ehp.0901059.S1)]. As further examples, in vitro arsenic-transformed RWPE-1 cells show relatively modest increases in MMP-9 secretion and relatively low invasiveness and colony formation (Tokar EJ, Waalkes MP, unpublished data) compared with As-CSC cells. Thus, the WPE-stem cells transform much faster and show a much more aggressive and totipotent phenotype (present study) compared with mature heterogeneous isogenic parental transformed cells. In accord with the CSC theory and the multistep carcinogenesis theory, the stage of differentiation determines not only the number of genetic alterations necessary for carcinogenic transformation but also dictates the aggressiveness of resulting tumors. NSCs, with their permissive state of differentiation, would need much fewer molecular transformational events, whereas mature cell populations would require more numerous alterations. The much more rapid transformation and resultant aggressive, pluripotent phenotype seen in arsenite-transformed As-CSC cells compared with the transformed mature, parental cells is consistent with this “permissive plasticity” concept of SCs in chemical carcinogenesis theory.

18The arsenic biomethylation capacity of WPE-stem cells is unknown. However, the parental RWPE-1 cells only very poorly methylate arsenic. Whether the transformative agent of WPE-stem was inorganic arsenic, a methylated metabolite, or some combination remains to be determined.

19CSCs isolated from advanced cancers typically reinitiate their tumor of origin. The remarkably low level of differentiation of As-CSC–produced xenograft tumors is perhaps a result of being transformed in the absence of surrounding mature cell signals directing them to a particular cancer based on tissue of origin. Such a minimally differentiated status is rare in adulthood tumors but common in some childhood tumors (e.g., nephroblastomas, teratomas) that are derived from immature embryonal-like cells. This primordial epithelial/mesenchymal histopathology, coupled with strong heterogeneous staining for epithelial SC markers (K5; ), suggests that the As-CSC hetero-transplantation tumors developed from highly pluripotent cells, allowing divergent differentiation toward both epithelial and mesenchymal components. The rapid progression and high aggressiveness suggest that arsenite quickly triggered acquisition of a malignant CSC-like phenotype consistent with a loss of self-renewal and differentiation control. Similarly, pleomorphic uterine carcinosarcomas, although not associated with arsenic exposure in humans, were once thought to be “collision” tumors but are now thought to derive from a single SC population that has acquired the ability to undergo aberrant, divergent differentiation, producing both mesenchymal and epithelial cancerous components. Our data indicate that arsenite has instilled a similar potential into As-CSC cells. Furthermore, very recently we found that when injected intravenously into mice, As-CSC cells formed highly pleomorphic malignancies in lung, liver, and skin (Tokar EJ, Waalkes MP, unpublished data). This provides significant support to the concept that arsenic has transformed human NSC/progenitor cells into CSC-like cells and adds credence to this as a plausible component of the mechanism of arsenic carcinogenesis.

20Many cultured NSC or CSC lines form free-floating spherical clusters of viable cells containing a preponderance of the NSCs or CSCs. Only sphere cells from arsenite-transformed WPE-stem cells formed xenograft tumors. This strongly supports our contention that arsenite has triggered CSC-like formation from phenotypically normal SCs because tumor-forming CSCs derived from advanced cancers typically reside in these spheres. Furthermore, arsenite produced an overabundance of CSC-like cells, primarily in the spheres, potentially accounting for the rapid formation, totipotent nature, and remarkable aggressiveness of the resulting xenograft tumors. In vivo arsenic alters skin SC population dynamics, eventually enhancing skin carcinogenesis coinciding with increased CSCs in epidermal carcinomas. Thus, arsenic appears to target SC populations both in vitro (present study; ) and in vivo, distorting SC dynamics and precipitating CSC overproduction. Whether other carcinogens act in a similar fashion is an important question that requires further investigation.

21NSCs and CSCs share fundamental properties, including self-renewal capacity, which is typically dysregulated in oncogenesis. In this regard, a remarkably consistent “U-shaped” temporal response occurred with several SC self-renewal genes (p63, BMI-1, ABCG2, SHH, OCT-4, and NOTCH-1) during arsenic-induced malignant transformation, with an initial suppression followed by reactivation. p63 is essential for SC self-renewal and proliferation within the prostate and other tissues. BMI-1 is required for self-renewal of both NSCs and CSCs in the prostate. In several tissues SHH promotes SC self-renewal, whereas aberrant activation is implicated in carcinogenesis. ABCG2 plays a clear role in maintaining SC populations and is a prostate NSC or CSC marker. OCT-4 is a pluripotency gene required for NSC self-renewal and maintenance that is often activated in carcinogenesis. The Notch signaling pathway regulates NSC self-renewal and differentiation and is frequently dysregulated in oncogenesis. In the prostate, Notch-1 is a marker for SC/progenitor cells, is required for normal development, and has been shown to function as both an oncogene and tumor suppressor during prostate carcinogenesis. Additionally, for keratins, K5 is a marker for undifferentiated basal SCs, whereas K18 is a marker of differentiated cells, and K5 expression increases during carcinogenesis. The remarkably consistent trend in these critical self-renewal/differentiation genes during WPE-stem transformation suggests that arsenite instilled aberrant self-renewal and differentiation capabilities as these cells acquired a malignant CSC phenotype. Intriguingly, this trend is consistent with that seen when LSCs are formed by introduction of the mixed lineage leukemia (MLL-AF9) gene into hematopoietic committed progenitor cells. A group of self-renewal genes highly expressed in HSCs is lost in committed progenitors but reactivated in LSCs initiated by introduction of the MLL-AF9 fusion protein. Likewise, reactivation of an SC-like genetic program in adult hematopoietic cells can introduce aberrant self-renewal characteristics of CSCs. Thus, distortion of an NSC self-renewal-associated signature drives these hematopoietic cells toward LSC phenotype and clearly shows that this trend is not arsenic specific, but potentially common during the process of CSC formation. Together with prior work, the present data suggest that sequential distortion of self-renewal–associated genes is critical to CSC formation.

22During arsenite-induced transformation and formation of As-CSCs, the dramatic decrease in PTEN expression coincided with the reactivation of self-renewal genes and acquisition of malignant phenotype. Poor prostatic PTEN expression increases SC-like cells and causes cancer initiation and in prostate cancer cells increases sphere-forming ability and xenograft sphere cell tumorigenicity. Furthermore, loss of PTEN enhances self-renewal capacity of SCs without dramatically altering pluripotency. PTEN expression maintains HSCs in a quiescent state, and its deletion leads to HSC depletion and leukemia formation enriched for LSCs. The loss of PTEN provides a plausible mechanism through which As-CSC cells reactivate self-renewal, although dysregulated, and yet maintain and potentially distort multipotential differentiation capacity.

23Our data support the hypothesis that invokes emergence of CSCs from NSCs as a primal initiating event in oncogenesis, at least for arsenic. Indeed, arsenite directly targeted the human WPE-stem line, transforming them into a highly aggressive CSC-like phenotype. Once transformed, these cells possess multiple CSC-like characteristics, including enhanced self-renewal capacity, competency to proliferate, clonogenicity, and, perhaps most important, the ability to initiate highly aggressive, immature, pluripotent tumors in vivo. The truly stunning capacity to rapidly produce highly pleomorphic tumors in xenograft studies and the fact that only cells from the arsenite-treated spheres formed tumors, together with strong evidence that CSCs reside in similar spheres derived from advanced cancers, support a direct acquisition of malignant CSC-like phenotype from NSC/progenitor cells triggered by arsenic. Additionally, our data fortify the emerging concept that the loss and subsequent reactivation of SC self-renewal are critical during malignant transformation of SCs, as appears to be the case with leukemias.

24Overall, this study expands our understanding of the carcinogenic potential of arsenic, a common environmental contaminant, by indicating that it can directly target SCs for carcinogenic transformation. Millions of people are exposed, likely throughout their entire lifetime, to unhealthy levels of arsenic in drinking water. Arsenic is clearly a transplacental carcinogen in rodents and probably in humans, and SCs are likely key targets in in utero life-stage susceptibility to chemical carcinogenesis. Therefore, the ability to malignantly transform NSCs could be of major mechanistic significance in arsenic-exposed human populations.