Polyamine-stimulation of arsenic-transformed keratinocytes

Abstract

Tumor promotion is strongly associated with inflammation and increased polyamine levels. Our understanding of relevant mechanisms responsible for arsenic-induced cancer remains limited. Previous studies suggest that arsenic targets and dysregulates stem cell populations that remain dormant in the skin until promoted to be recruited out of the bulge stem cell region, thus giving rise to skin tumors. In this study, we explored a possible mechanism by which increased keratinocyte polyamine biosynthesis promotes tumorsphere formation and invasiveness of arsenic-transformed HaCaT keratinocytes (As-HaCaT). Unlike parental HaCaT cells, As-HaCaT cells were tumorigenic in athymic nude mice, and the CD45^negative epithelial tumor cells had enriched expression of Toll-Like Receptor 4 (TLR4), CD34 and CXCR4 as did As-HaCaT tumorsphere cultures compared to As-HaCaT monolayer cultures. Ornithine decarboxylase (ODC) overexpressing keratinocytes (Ker/ODC) release increased levels of the alarmin high mobility group box 1 (HMGB1). Ker/ODC conditioned medium (CM) stimulated As-HaCaT but not parental HaCaT tumorsphere formation, and this was inhibited by glycyrrhizin, an inhibitor of HMGB1, and by TAK242, an inhibitor of the HMGB1 receptor TLR4. Compared to parental HaCaT cells, As-HaCaT cells demonstrated greater invasiveness across a Matrigel-coated filter using either fibroblast CM or SDF-1α as chemoattractants. Addition of Ker/ODC CM or HMGB1 dramatically increased As-HaCaT invasiveness. Glycyrrhizin and TAK242 inhibited this Ker/ODC CM-stimulated invasion of As-HaCaT cells but not HaCaT cells. These results show that polyamine-dependent release of HMGB1 promotes the expansion of stem cell-like subpopulations in arsenic-transformed keratinocytes while also increasing their invasiveness, suggesting that polyamines may be a potential therapeutic target for the prevention and treatment of arsenic-initiated skin cancers.

The polyamine-dependent release of HMGB1 and subsequent signaling through TLR4, promotes the expansion of stem cell-like subpopulations in arsenic-transformed keratinocytes while also increasing their invasiveness, suggesting that polyamines may be a potential target for the treatment of arsenic-initiated skin cancers.

Introduction

environmental exposure to naturally occurring arsenic in the drinking water poses a daunting global health issue, with approximately 150 million people exposed to toxic levels of arsenic (1,2). High concentrations of arsenic in underground water are also found in many parts of the United States. Arsenic is the most common worldwide contaminant in soil, groundwater, food and plants (2). Chronic exposure to arsenic in humans is causally associated with neoplasias of the skin and to a lesser extent, of the lung, liver, kidney and bladder. Epidemiological studies suggest that the population cancer risk from arsenic in water supplies in the United States may be comparable to that of environmental tobacco smoke and radon in homes with risk estimates of approximately 1 in 1000 (3). However, the mechanisms contributing to arsenic-induced cancer are complex and elusive, largely due to the lack of predictive animal models. The difficulty in inducing tumors in adult rodents following arsenic exposure as a single agent reflects that it often takes 10- to 100-fold higher doses of arsenic to manifest toxic effects in animals compared to that in humans (4). Most animal investigations of arsenic-induced carcinogenesis have included the co-administration of another carcinogen, UV irradiation or the presence of an activated oncogene (5).

Accumulating evidence suggests that arsenic is a transplacental carcinogen in both animals (6–8) and humans (9,10), and that it targets fetal stem cells leading to dysregulation of the normally tightly regulated process of stem cell self-renewal and differentiation (7,8). In addition, arsenic-induced in vitro transformation of human keratinocytes has been reported to lead to increased numbers of putative cancer stem cells (6). These observations suggest that arsenic targets and dysregulates stem cell populations that remain dormant in the skin until promoted (by TPA or wounding) to be recruited out of the bulge stem cell region, thus giving rise to skin tumors (7). Because carcinogen target cells are thought to be long lived, slowly cycling stem cells found in the hair follicle bulge region, it is essential to understand pathways that regulate stem cell recruitment in arsenic-induced skin carcinogenesis.

Using Cre recombinase-reporter mice, we have previously reported that elevated levels of polyamines stimulate the recruitment of bulge stem cells in quiescent skin (11). The polyamines putrescine, spermidine and spermine are some of the major cations present in all cells. Polyamines have long been known to be associated with cell proliferation in normal tissues, and polyamine levels are dramatically elevated in tumors (12). Polyamines are primarily bound to polyanionic macromolecules, particularly RNA, resulting in far-reaching effects upon cellular processes including DNA replication, transcription, and translation. A hallmark of tumor promoting activity involves the induction of ornithine decarboxylase (ODC), the initial rate-limiting enzyme in polyamine biosynthesis. Use of ODC transgenic mouse models has demonstrated that increased ODC activity is sufficient to promote tumor development following a single low dose exposure to a carcinogen (13). Our finding that elevated epidermal polyamine levels alone can stimulate the recruitment of bulge stem cells in quiescent skin (11) is significant with regard to the stem cell origin of skin cancer since initiated stem cells can remain dormant and never expand to tumors without a stimulus to recruit the initiated stem cells from their bulge stem cell niche.

Accumulating evidence suggests that inflammation is an important regulator of stem cells. Stem cells express receptors that detect pathogen-associated molecular patterns associated with microbes and alarmins (also known as danger-associated molecular patterns) that are released by damaged host cells. One of the best characterized danger-associated molecular pattern is high mobility group box 1 (HMGB1) which is released by stressed or dying cells and triggers sterile inflammation, innate and adaptive immunity, and tissue healing after damage (14,15). Physiologic polyamines with cytoprotective capabilities are also induced by a variety of stresses including reactive oxygen species, heat and ultraviolet irradiation, and polyamines levels are elevated in tissue remodeling events such as wound healing, tissue regeneration and tumor development. We previously reported that wound repair following abrasion in ODC transgenic mice with elevated epidermal polyamine biosynthesis leads to a prolonged inflammatory response and wound-induced tumor formation (16). This wound-induced tumor formation is dependent on both polyamines and inflammation since treatment with the anti-inflammatory agent dexamethasone prevents the wound-induced tumor growths in abraded ODC transgenic skin (16).

In this study, we explore mechanisms by which ODC-overexpressing keratinocytes stimulate accumulation of stem cell populations in arsenic-exposed human HaCaT keratinocytes. In particular, the role of the pro-inflammatory alarmin, HMGB1, in mediating the effect of polyamines on the expansion and invasiveness of transformed keratinocytes is investigated. We report that elevated biosynthesis of polyamines in normal keratinocytes, as occurs in wound healing, triggers keratinocyte-derived factors including HMGB1 that can expand and promote arsenic-transformed stem cell populations and invasiveness.

Materials and methods

Animals

Mice used for primary keratinocyte cultures and in wound healing experiments included K6/ODC transgenic mice and their normal littermates. A keratin 6 promoter constitutively directs ODC transgene expression to the outer root sheath cells of hair follicles in K6/ODC transgenic mice (13). Female athymic nude mice were obtained from Charles Rivers/NCI and used for tumor studies. Protocols for the use of animals in these studies were reviewed and approved by the Institutional Animal Care and Use Committee of the Lankenau Institute for Medical Research in accordance with the current US Department of Agriculture, Department of Health and Human Service regulations and standards.

Cell culture

HaCaT cells were purchased from the ATCC, and cells were expanded and frozen in liquid nitrogen upon delivery. Cell authentication and testing to ensure cells were mycoplasma negative were performed. To generate As-HaCaT cells, HaCaT cells were passaged twice weekly with media (DMEM, 10% FBS, 1% penicillin-streptomycin) containing 100 nM sodium arsenite for 40 weeks.

In vivo tumor experiments

Tumor models were established by subcutaneous injections of 5 × 10⁶ HaCaT and As-HaCaT cells suspended in a 50% growth factor-reduced Matrigel solution in the dorsal flank of athymic nude mice. Tumor growth was assessed morphometrically using calipers, and tumor volumes were calculated using the formula V (mm³) = π/6 × A × B² (A is the larger diameter and B is the smaller diameter).

Tumorsphere formation

Equal amounts of HaCaT and As-HaCaT cells (5 × 10³) were plated on ultralow adherence 24-well dishes in tumorsphere medium (equal parts DMEM and F12 media with final 1x B27, 20 ng/ml EGF, 20 ng/ml bFGF, 4 μg/ml heparin, 100 μg/ml gentamycin and 1x Pen/Strep. Tumorspheres were enzymatically disassociated to single cell suspension at least twice to enrich for self-renewing tumorsphere cells. These tertiary tumorsphere cells were plated again as single cell suspension on ultralow adherence dishes with or without any additional treatments. Approximately 4–5 days after plating, tumorspheres (with a diameter of at least three cells) were counted. Cells were cultured in triplicate for various treatment groups.

Primary keratinocyte culture

Primary cultures of epidermal cells were isolated from 3- to 4-day-old K6/ODC transgenic newborn pups and their normal littermates by a trypsin flotation procedure as previously described (17). K6/ODC transgenic pups were distinguished from their normal littermates by PCR genotyping for the K6/ODC transgene (13). Cells were cultured in low-calcium (0.05 mM) EMEM media (BioWhittaker Walkersville, MD), supplemented with 8% chelex-treated fetal bovine serum at 35°C. To generate conditioned medium, confluent cultures of keratinocytes were cultured in serum-free medium for 48 h and the conditioned medium was filtered and stored at −80°C until use.

[³H] thymidine uptake

Parental HaCaT and As-HaCaT cells were grown in 60-mm tissue culture dishes, and ³H-thymidine incorporation assay was performed as described with some minor modifications (18). Briefly, when approximately 40% confluent, cells were pulsed with [³H] thymidine (1.5 μCi/ml) for 1 h at 37°C, washed twice with cold PBS containing unlabeled thymidine, and lysed in 0.5-ml cold 1 N NaOH. The lysates were put into glass conical tubes containing 0.5 ml of cold 1 N perchloric acid and centrifuged at 10 000 rpm for 10 min. The radioactivity in the supernatant was measured using a liquid scintillation counter. [³H] thymidine counts were normalized to the amount of DNA in each sample.

EdU incorporation

Cell proliferation was measured using Click-iT EdU incorporation flow cytometry kit (Invitrogen). Briefly, cells were grown in triplicate in 24-well dishes. When cells were approximately 40% confluent, the media was replaced with growth media containing the thymidine analog (5-ethynyl-2′-deoxyuridine [EdU], 10 μM) conjugated with Alexa fluor 488. Cells were incubated for 2 h at 37°C, then trypsinized, washed, fixed and permeabilized. The Click-iT reaction cocktail was prepared as per the manufacturer’s instruction and incubated with the cells for 30 min at RT in the dark. Cells were analyzed by flow cytometry.

Immunoblot analyses

Conditioned medium from normal littermate keratinocytes (Ker/Norm) and ornithine decarboxylase over expressing keratinocytes (Ker/ODC) was concentrated using an Amicon concentrator, separated by SDS-PAGE, transferred to polyvinylidene difluoride membranes (Millipore, Temecula, CA), and briefly stained with Ponceau S (Sigma, St. Louis, MO) to verify efficient transfer. Immunoblots were blocked with 5% non-fat dry milk followed by overnight incubation at 4°C with a rabbit polyclonal antibody against HMGB1 (Cell Signaling, Danvers, MA) and then a secondary antibody conjugated to horseradish peroxidase. Antibody binding was detected by chemiluminescence (ECL Plus Western Blotting Detection System, Amersham/GE Healthcare, Piscataway, NJ). Corresponding cell lysates were assayed by immunoblot analysis and probed with a mouse monoclonal antibody against β-actin (Sigma, St. Louis, MO) to verify equal numbers of cells.

Zymography

Cells were grown to 80% confluency, washed twice with PBS and then changed to serum-free DMEM. After 48 h the conditioned media was collected and centrifuged at 600× g to remove cell debris. A 500 μl aliquot was concentrated in an Amicon concentrator at 4°C to 100 μl. A total of 1 μg of protein per sample was loaded on a 10% zymography gel containing 0.1% gelatin (Novex). Matrix metalloproteinase 2 and matrix metalloproteinase 9 activity were detected by incubating the gel in 1× zymography renaturing buffer (Novex) for 30 min at room temperature followed by 1× developing buffer (Novex) overnight at 37°C. After Coomassie staining, the clear bands of proteolytic degradation were quantified.

Invasion assay

Boyden chamber inserts (for a 24-well plate) with 8 µm porous filters were coated with 50 μg of growth factor-reduced Matrigel and incubated at 37°C for 3 h. HaCaT and As-HaCaT cells were plated at 1 × 10⁶ cells in serum-free DMEM media to the interior of each insert. After 1 h, a gradient was initiated with the addition of serum-free DMEM media supplemented with the chemoattractant (conditioned medium from NIH3T3 fibroblasts or 200 ng/ml SDF-1α) to the lower compartment of the invasion chamber. Conditioned medium collected from NIH3T3 fibroblasts was used as the chemoattractant since it is a rich supply of growth factors and other chemoattractants for tumor cells (19). After 24-h incubation at 37°C, cells that had not migrated through the Matrigel-coated filter were removed with a cotton-tipped applicator and migrated cells on the underside of the filter were fixed in 10% neutral buffered formalin, stained with 1% crystal violet, and then mounted on a slide and counted using a microscope. In some cases, HaCaT and As-HaCaT cells were cultured with conditioned medium from primary cultures of Ker/ODC or Ker/Norm or with recombinant HMGB1, the HMGB1 inhibitor glycyrrhizin, or the TLR4 inhibitor TAK242 added to the medium in the upper chamber. Each condition was performed in triplicate.

Statistics

All in vivo experiments were carried out using multiple animals (n = 10 per experimental group). All in vitro experiments were performed in at least triplicate, and data compiled from 2 to 3 separate experiments. All analyses were done using a 1-way analysis of variance (ANOVA) with a Tukey test for statistical significance.

Results

Chronic arsenic exposure of HaCaT cells induces a stem cell-like phenotype

To evaluate stem cell characteristics in arsenic-transformed human keratinocytes, HaCaT cells were continuously exposed for 40 weeks to low, nontoxic levels of inorganic arsenite (100 nM). The human HaCaT keratinocyte cell line was originally derived from normal human adult skin and is nontumorigenic in vivo (20). Previous reports have shown that chronic exposure to low-dose arsenic malignantly transforms HaCaT keratinocytes and expands its stem cell-like subpopulation (6). Surprisingly, arsenic-treated cells (As-HaCaT) had a slower cell proliferation rate compared to parental HaCaT cells (matched for passage number), as measured by incorporation of either [³H] thymidine or the thymidine analog EdU (Figure 1A and B). In addition, As-HaCaT cells formed significantly more tumorspheres than passage-matched parental HaCaT cells when grown in suspension on ultralow adherence plates (Figure 1C), suggesting that chronic arsenic treatment of keratinocytes increases the number of potential cancer stem/progenitor cells (21). It is important to note that these tumorspheres were self-renewing in that they formed tumorspheres following passaging on ultralow adherence plates for more than three passages. Because epidermal bulge stem cells express the cell surface glycoprotein CD34 (22), both monolayer and self-renewing tumorsphere cultures of parental HaCaT and As-HaCaT cells were analyzed for expression of this stem cell marker. As-HaCaT cells expressed markedly higher levels of CD34 in both monolayer and tumorsphere cultures compared to parental HaCaT cells (Figure 1D).

Figure 1.

Chronic arsenic exposure changes HaCaT cell phenotype. HaCaT cells were chronically exposed to 100 nM sodium arsenite for 40 weeks. Proliferation in cells exposed to arsenic (As-HaCaT cells) and in parental HaCaT cells was measured by (A) [³H] thymidine incorporation ([³H] cpm/μg DNA ± SD) and (B) EdU incorporation (% EDU incorporation). (C) The ability to form tertiary tumorspheres was assessed using ultralow adherence plates (number of tumorspheres ± SD). (D–F) Surface markers on HaCaT and As-HaCaT cells isolated from a monolayer or from tumorspheres were analyzed by flow cytometery for (D) CD34, (E) CXCR4 and (F) TLR4. Data are representative of multiple experiments assayed in triplicate. *P ≤ 0.05 and #P ≤ 0.01.

Stem cell populations have also been shown to express higher levels of the innate immune response receptor Toll-Like Receptor 4 (TLR4) (23–26) and CXCR4 (27–29). TLR4 is expressed in embryonic stem cells and adult progenitor cells from mammary and intestine with its activation leading to proliferation and stem cell expansion (30–32). As-HaCaT tumorspheres expressed elevated levels of both CXCR4 and TLR4 compared to tumorspheres formed by parental HaCaT cells or monolayer cultures of As-HaCaT cells (Figure 1E and F). Taken together, these data suggest that As-HaCaT cells have acquired a more stem cell-like phenotype that is slow growing.

To verify malignant transformation of HaCaT cells following chronic arsenic exposure, passage-matched HaCaT and As-HaCaT cells were intradermally injected into athymic nude mice. As expected, none of the mice injected with the parental HaCaT cells developed tumors at the injection site. However, all 10 mice injected with As-HaCaT cells developed measurable tumors (Figure 2). Interestingly, flow cytometry analyses of surface markers revealed that between 15 and 20% of the tumor epithelial cells (CD45^negative) expressed the stem cell markers, CD34, CXCR4 and TLR4, at higher levels compared to that in the cultured As-HaCaT cells which were initially injected into the mice (Figure 1).

Figure 2.

Arsenic-transformed As-HaCaT cells form tumors in vivo. (A) Passage-matched HaCaT and As-HaCaT cells (5 × 10⁶) suspended in PBS were diluted 1:1 with Matrigel and injected subcutaneously into athymic nude mice. Graph shows tumor progression (mean ± SEM). Upon sacrifice, single cell suspensions of tumor and spleen were stained and analyzed by flow cytometry for CD45^negative cells expressing (B) CD34, (C) CXCR4 and (D) TLR4.

Polyamine-stimulation of HMGB1 release by keratinocytes

Arsenic-initiated genetic changes may remain dormant in stem cell subpopulations in the skin for many years until a tumor promotional event triggers the development of skin tumors. Interestingly, most arsenic-induced skin lesions are found in areas of frictional contact or wounding in the skin rather than in sun-exposed areas of the skin (33). This suggests that inflammation associated with tissue remodeling is necessary to promote arsenic-induced skin cancer. In particular, both TLR4 and CXCR4 serve as receptors for the pro-inflammatory alarmin, HMGB1 (14,34,35). HMGB1 is required for skin regeneration, and it is transiently increased 3 days following wounding (36,37). Moreover, HMGB1 has been shown to promote tumorigenesis via many pathways and mechanisms including by directly promoting tumor growth and invasion (38). Since polyamine levels are also elevated following wounding, we measured HMGB1 release in primary cultures of keratinocytes isolated from the skin of K6/ODC transgenic mice and normal littermates. Figure 3A shows that elevated levels of polyamine biosynthesis in primary keratinocyte cultures isolated from ODC transgenic mouse skin (Ker/ODC) stimulated greater release of HMGB1 in the culture medium compared to that in keratinocytes cultured from their wildtype normal littermates (Ker/Norm). In contrast to quiescent epidermal cells in intact skin, cultured epidermal keratinocytes express markers of hyperproliferative, activated keratinocytes, such as IL-1 and keratin 6, as are seen in epidermal cells following wound healing (39). Interestingly, the prolonged inflammation and wound-induced skin tumors that we reported in ODC transgenic mice (16) was accompanied by elevated levels of HMGB1 epidermal staining in the skin of ODC transgenic mice even 14 days following skin abrasion when HMGB1 staining was no longer elevated in normal littermate wounded skin (Supplementary Figure S1, available at Carcinogenesis Online). This HMGB1 release following wounding was polyamine-dependent since it was prevented with inhibition of ODC activity with α-difluoromethylornithine (data not shown).

Figure 3.

ODC-overexpressing keratinocytes secrete elevated levels of HMGB1. (A) Primary keratinocytes were isolated from the skin of newborn K6/ODC transgenic mice (Ker/ODC) and their normal littermates (Ker/Norm). Following 2 days in culture, conditioned medium from these cells was analyzed in duplicate by immunoblot analysis for HMGB1 protein and the cell lysates analyzed for β-actin. (B) Working hypothesis of effect of skin wounding on arsenic-initiated stem cell-like populations. Wounding induces polyamine biosynthesis in epidermal cells that is mimicked by primary keratinocyte cultures isolated from K6/ODC transgenic mouse skin (Ker/ODC). HMGB1 released from Ker/ODC acts as a ligand with its receptors TLR4 and in conjunction with SDF-1α with CXCR4 that are enriched in As-HaCaT cells to promote tumorsphere formation, invasiveness, and tumor formation. HMGB1-stimulation of As-HaCaT cells can be inhibited by the HMGB1 inhibitor glycyrrhizin or by the TLR4 inhibitor TAK242.

Polyamine-dependent release of HMGB1 stimulates As-HaCaT tumorsphere formation

The polyamine-stimulated release of HMGB1 from keratinocytes suggests that HMGB1 release from keratinocytes may act as a ligand for TLR4 receptors expressed in arsenic-transformed HaCaT tumorspheres and in As-HaCaT tumors (Figure 3B). To evaluate if factors, such as HMGB1 that are released from keratinocytes with elevated polyamine biosynthesis, can differentially affect tumorsphere formation by As-HaCaT cells, we tested the effect of secreted factors from Ker/Norm and Ker/ODC cultures on As-HaCaT and parental HaCaT tumorsphere formation. Ker/Norm and Ker/ODC primary cultures were cultured in serum-free medium for 48 h, and this conditioned medium was added to trypsinized parental HaCaT and As-HaCaT tertiary tumorsphere cells and seeded again in low adherence plates for tumorsphere growth. Incubation of As-HaCaT cells with Ker/ODC conditioned media significantly increased tumorsphere formation while incubation with Ker/Norm media had no significant effect on As-HaCaT tumorsphere formation (Figure 4A). In contrast, incubation of parental HaCaT cells with Ker/ODC conditioned media resulted in only a small but non-significant increase in tumorsphere formation, indicating that chronic arsenic treatment of HaCaT cells not only enriches for a stem-cell-like subpopulation but also increases their responsiveness to factors released by keratinocytes with elevated polyamine biosynthesis.

Figure 4.

HMGB1 promotes tumorsphere formation in As-HaCaT cells. (A) Tumorsphere formation was assessed for HaCaT and As-HaCaT cells in which Ker/Norm or Ker/ODC conditioned media was added to tumorsphere growth media (1:5 ratio). (B) Tumorsphere formation was assayed with the addition of either Ker/Norm conditioned media or Ker/ODC conditioned media ± 5 ng/ml of recombinant human HMGB1 and ±100 μM of the HMGB1 inhibitor glycyrrhizin. (C) The effect of HMGB1 (5 ng/ml) and glycyrrhizin (100 μM) on As-HaCaT tumorsphere formation was assayed and compared to the effect of Ker/ODC conditioned media. (D) Tumorsphere formation was assayed with the addition of either Ker/Norm conditioned media or Ker/ODC conditioned media ± the TLR4 inhibitor TAK242 (10 μM). Glycyrrhizin abbreviated as G in panel C. Data are representative of multiple experiments assayed in triplicate. *P ≤ 0.05 and #P ≤ 0.01.

To further explore the involvement of HMGB1 release from Ker/ODC cells in As-HaCaT tumorsphere formation, HaCaT and As-HaCaT self-renewing tumorspheres were incubated with either Ker/Norm or Ker/ODC conditioned media supplemented with either exogenous HMGB1 or the HMGB1 inhibitor glycyrrhizin (40,41). Incubation of As-HaCaT cells with Ker/Norm-conditioned medium supplemented with HMGB1 significantly increased As-HaCaT tumorsphere formation to levels comparable to incubation with Ker/ODC conditioned medium without added HMGB1 (Figure 4B). On the other hand, the HMGB1 inhibitor glycyrrhizin blocked As-HaCaT tumorsphere formation even when As-HaCaT cells were incubated with Ker/ODC conditioned medium that has high levels of endogenous HMGB1 (Figure 4B). The addition of only HMGB1 (5 ng/ml) also increased As-HaCaT tumorsphere formation (Figure 4C), but not to the same degree as adding Ker/ODC conditioned media, indicating that HMGB1 contributes to As-HaCaT tumorsphere formation but other Ker/ODC secreted factors may also play a role. In contrast, the addition of HMGB1 or glycyrrhizin to either Ker/Norm or Ker/ODC conditioned medium had no effect on parental HaCaT tumorsphere formation. To investigate the involvement of TLR4, which is elevated in As-HaCaT tumorspheres, on HMGB1 dependent tumorsphere formation we used the TLR4 inhibitor TAK242 to block signaling through TLR4 (42). The addition of TAK242 had no effect on tumorsphere formation when As-HaCaT cells were grown in Ker/Norm conditioned media, but significantly inhibited tumorsphere formation when As-HaCaT cells were incubated with Ker/ODC conditioned media containing high endogenous levels of HMGB1 (Figure 4D). These data taken together suggest that malignant transformation of As-HaCaT cells by arsenic enriches for stem cell populations that are responsive to secreted factors in keratinocytes with elevated polyamine biosynthesis, specifically HMGB1 which can potentially signal through TLR4.

Polyamine-induced secreted factors and HMGB1 increase As-HaCaT invasiveness

Many cancer deaths are caused by tumor invasiveness and metastasis rather than the primary tumor itself. Matrix metalloproteinase 2 and matrix metalloproteinase 9 degrade extracellular matrix proteins and play a major role in tumor invasion and metastasis (43). As-HaCaT cells secreted higher levels of matrix metalloproteinase 2 and matrix metalloproteinase 9 compared to parental HaCaT cells (Figure 5). Boyden chamber inserts coated with Matrigel were used as an in vitro invasion assay to compare the ability of HaCaT and As-HaCaT cells to degrade extracellular matrix and invade. Using NIH3T3 fibroblast conditioned media as a growth-factor-enriched chemoattractant (19), we observed significantly more As-HaCaT cells invading across the Matrigel-coated filter compared to that of parental HaCaT cells (Figure 6A).

Figure 5.

Matrix metalloproteinase 2 levels are elevated in As-HaCaT cells. HaCaT and As-HaCaT cells were cultured in serum-free DMEM medium for 48 h. (A) Secreted MMPs in the conditioned medium were analyzed using a 10% zymography gel containing 0.1% gelatin. Bands for (B) matrix metalloproteinase 9 and (C) matrix metalloproteinase 2 were quantified. #P ≤ 0.01.

Figure 6.

Invasiveness of As-HaCaT cells is enhanced by Ker/ODC conditioned media. Invasiveness of HaCaT and As-HaCaT cells was assessed using Matrigel-coated Boyden chambers. NIH3T3 fibroblast conditioned media was used as a chemoattractant to promote movement across the Matrigel-coated membrane of the Boyden chamber. HaCaT and As-HaCaT cells in the Boyden chamber were incubated with either Ker/Norm or Ker/ODC conditioned media (A) ±10 ng/ml HMGB1 and ± the HMBG1 inhibitor glycyrrhizin (1 mM) or (B) ±10 μM of the TLR4 inhibitor TAK242. (C) SDF-1α (200 ng/ml), the ligand for CXCR4, was used as a chemoattractant instead of NIH3T3 conditioned media. (D) HaCaT and As-HaCaT cells in the Boyden chamber were incubated with Ker/Norm conditioned media ± 1 μM or 5 μM spermidine (Spd) with NIH3T3 conditioned media as a chemoattractant. Data are representative of multiple experiments assayed in at least triplicate. *P ≤ 0.05 and #P ≤ 0.01.

To explore the effect of keratinocyte polyamine biosynthesis and HMGB1 on As-HaCaT invasiveness, HaCaT and As-HaCaT keratinocytes were cultured with either Ker/Norm or Ker/ODC conditioned media in the Boyden chamber insert. Incubation with either Ker/Norm or Ker/ODC conditioned media increased the invasiveness of both parental HaCaT and As-HaCaT cells, but Ker/ODC conditioned media significantly increased invasiveness more than Ker/Norm conditioned media (Figure 6A). Furthermore, As-HaCaT cells were more invasive following incubation with Ker/Norm or Ker/ODC conditioned media compared to parental HaCaT cells. The addition of exogenous HMGB1 to Ker/Norm conditioned media increased the degree of invasiveness seen with As-HaCaT cells to levels similar to that seen when As-HaCaT cells were incubated with Ker/ODC conditioned media alone (Figure 6A). Conversely, when the HMGB1 inhibitor glycyrrhizin was added to Ker/ODC conditioned media, invasiveness was dramatically reduced compared to Ker/ODC conditioned media alone. Indeed, glycyrrhizin reduced As-HaCaT invasiveness to levels similar to basal invasive levels of parental HaCaT cells not incubated with either Ker/Norm or Ker/ODC conditioned medium (Figure 6A), indicating that HMGB1 secreted from Ker/ODC plays a critical role in stimulating the invasiveness of As-HaCaT cells.

To determine if HMGB1 in Ker/ODC conditioned media may stimulate invasiveness through the TLR4 receptor, we used the TLR4 inhibitor TAK242 to block TLR4 signaling (Figure 6B). TAK242 had no effect on HaCaT or As-HaCaT invasiveness stimulated with Ker/Norm conditioned medium; however, TAK242 significantly reduced As-HaCaT invasiveness stimulated with Ker/ODC conditioned medium (Figure 6B). Interestingly, whereas the HMGB1 inhibitor glycyrrhizin inhibited the invasiveness of both HaCaT and As-HaCaT cells cultured with Ker/ODC conditioned medium to varying degrees (Figure 6A), inhibition of TLR4 signaling with TAK242 had no effect on HaCaT cell invasiveness cultured with Ker/ODC conditioned media (Figure 6B), indicating that HMGB1 signaling through TLR4 is particularly important in the arsenic-transformed As-HaCaT cells.

To assess if the CXCR4 receptor played a role in Ker/ODC stimulation of invasiveness we used the CXCR4 ligand SDF-1α as the chemoattractant instead of NIH3T3 conditioned media. When SDF-1α was used as the chemoattractant, conditioned medium from Ker/ODC, but not Ker/Norm, stimulated increased invasiveness with both HaCaT and As-HaCaT cells with the magnitude of the increase greater for the As-HaCaT cells (Figure 6C). Interestingly, incubation with the HMGB1 inhibitor glycyrrhizin significantly inhibited As-HaCaT invasiveness cultured with Ker/ODC conditioned media, but not HaCaT cells.

Another possible mechanism by which Ker/ODC secreted factors may stimulate invasiveness in As-HaCaT cells may be due to increased polyamine levels themselves in Ker/ODC conditioned media. To test this hypothesis, we added physiological levels of spermidine (44) to serum-free, Ker/Norm conditioned media and then assessed its effect on the invasiveness of parental HaCaT and As-HaCaT cells. We found no effect of increasing spermidine levels on invasiveness (Figure 6D), indicating increased polyamine levels in keratinocytes trigger secretion of downstream factors such as HMGB1 that stimulate As-HaCaT invasiveness.

Discussion

Polyamines are intrinsically important to tumors, providing biomass through increased protein translation and enhancing survival pathways for tumor cells, but they also influence cancer stem cells by modulating the pro-inflammatory tumor microenvironment. Because polyamine-dependent release of the alarmin HMGB1 is pro-inflammatory, our results suggest that polyamines stimulate the expansion of arsenic-transformed keratinocytes by activating a chemotactic inflammatory response, particularly in stem cell subpopulations that form tumorspheres and express increased levels of the HMGB1 receptors TLR4 and CXCR4. We show that the polyamine-dependent release of HMGB1 from keratinocytes promotes the expansion of stem cell-like tumorsphere growth of malignantly transformed arsenic-treated keratinocytes while also increasing their invasiveness, suggesting that polyamines may be an enticing potential therapeutic target for the prevention and treatment of arsenic-derived skin cancers.

It is well established that inflammation is an essential part of all phases of tissue repair and tumorigenesis (45,46). In addition to inflammation that acts as a defense against pathogens, a sterile inflammation in the absence of infection can be triggered in response to chemical or physical damage (15). Most known tumor-promoting agents are potent inducers of sterile inflammation, leading to the activation of previously dormant premalignant cells (46). TLR4 activation in keratinocytes is a major driver of cutaneous inflammation and is required for tumor development and progression in skin cancers (38,42,47,48). In a two-step inflammation-induced skin tumorigenesis model, treatment with the tumor promoter, croton oil, triggered HMGB1 release from the skin and TLR4 signaling that led to the recruitment of inflammatory cells and inflammatory cytokines (48). HMGB1 release from damaged or stressed cells can also prime tissue regeneration and wound healing via the proliferation of resident quiescent adult stem cells (49). In this study, we have shown that increased polyamine biosynthesis in keratinocytes leads to HMGB1 release that stimulates expansion of stem cell-like subpopulations of arsenic-transformed keratinocytes as well as their invasiveness.

A hallmark of tumor promoting activity is elevated ODC activity and polyamine levels. Use of ODC transgenic mice in which elevated ODC activity is directed to the epidermis has revealed that elevated epidermal polyamine biosynthesis stimulates epidermal proliferation, invasiveness, vascularization in the underlying dermis, and is also immunomodulatory, all of which drive skin tumor development in mice initiated with subthreshold levels of carcinogens (50–52). Interestingly, induction of genetic lesions is not required to initiate skin tumors in ODC transgenic mice. Wounding via skin abrasion in ODC transgenic mice leads to a prolonged inflammatory response and wound-induced tumor formation that is inflammation-dependent (16). Although alarmins such as HMGB1 have been shown to play an important role in wound healing in the skin (36,37), we have shown that levels of cytoplasmic HMGB1 staining remain elevated in abraded ODC transgenic mouse skin long after cytoplasmic HMGB1 is no longer detected in abraded normal littermate skin. Because HMGB1 promotes inflammation and wound healing and is secreted by primary keratinocytes with elevated polyamine biosynthesis, the tumor-promoting effects of increased polyamine biosynthesis is probably due, at least in part, to the pro-inflammatory effects of alarmins such as HMGB1 that are secreted by ODC-overexpressing keratinocytes. In addition to HMGB1, other alarmins, such as hyaluronan, S100 proteins, heparin sulfate, fibrinogen and heat-shock proteins, may be released by ODC-overexpressing keratinocytes. Moreover, HMGB1 can stimulate receptors other than TLR4, including TLR2, TLR3, TLR9 and RAGE, which may also be expressed on arsenic-transformed stem cell subpopulations. It is likely that wound-associated inflammation plays a role in arsenic-induced skin tumorigenesis, particularly since arsenic-induced keratoses and tumors tend to be located at sites of friction and trauma not receiving maximum sunlight exposure (i.e. palms, soles of feet, cloth-covered body parts) (33). Our data suggest that wounding-associated increases in keratinocyte polyamine levels and release of HMGB1 stimulates the recruitment and expansion of epidermal stem cell populations that harbor arsenic-initiated genetic lesions to promote the development of skin tumors.

Keratinocyte stem cells display an enriched expression profile of genes involved in innate immunity, and it has been suggested that inflammation plays a role in regulating and mobilizing keratinocyte stem cells in tissue remodeling events such as wound healing and tumorigenesis (23). Our results show that As-HaCaT keratinocytes are characterized by increased tumorsphere formation, reflecting a higher percentage of a stem-like subpopulation compared to that in the parental HaCaT cells. Importantly, As-HaCaT tumorsphere formation is stimulated by HMGB1/TLR4 signaling that is initiated by HMGB1 released from ODC-overexpressing keratinocytes. We show that tumorspheres formed from arsenic-transformed keratinocytes express increased levels of TLR4 and CXCR4, both of which are receptors for HMGB1 (14,34,35). Conversely, inhibitors of HMGB1 and TLR4 suppress the stimulation of As-HaCaT tumorsphere formation with conditioned medium from ODC-overexpressing keratinocytes. In addition, there is an enrichment of TLR4⁺ CXCR4⁺ CD34⁺ epithelial tumor cell subpopulation in As-HaCaT in vivo tumors compared to that in the As-HaCaT cells initially injected in the mice. The stem cell-like nature of this enriched tumor subpopulation is in agreement with the finding that the CD34⁺ tumor subpopulation is over 100-fold more potent in initiating secondary skin tumors than unsorted tumor cells (53).

Because cancer stem cells are resistant to anoikis and are self-sustaining, they can form tumorspheres in nonadherent conditions and they have greater motility (21). As expected, As-HaCaT cells were indeed more invasive than nontransformed HaCaT cells, particularly when stimulated with conditioned medium from Ker/ODC or with HMGB1 alone. Metastatic tumor cells have been shown to be attracted to tissue sites where SDF-1α is expressed (27–29). SDF-1α-chemoattracted invasiveness was increased only in As-HaCaT cells stimulated with Ker/ODC conditioned medium, and notably, this was inhibited with the HMGB1 inhibitor, glycyrrhizin. In all, these data suggest that increased epidermal polyamine biosynthesis promotes expansion, invasion and tumor growth of arsenic-transformed stem cell-like populations via HMGB1/TLR4 signaling. Because stem cell populations can harbor mutations and remain dormant for many years, it is important to identify and prevent signaling mechanisms that awaken these stem-like cells to grow into tumors. These studies elucidate a potentially novel mechanism by which polyamines influence tumor biology and progression. Targeting polyamine biosynthesis with inhibitors such as α-difluoromethylornithine or with inhibitors of HMGB1 and other alarmins has clinical relevance not only for preventing arsenic-induced cancer but also other cancers.

Funding

National Cancer Institute (CA190952 to S.G.).

Conflict of Interest Statement: None declared.

Glossary

Abbreviations

As-HaCaT

arsenic-transformed HaCaT cells

CM

conditioned medium

EdU

5-ethynyl-2′-deoxyuridine

HMGB1

high mobility group box 1

Ker/ODC

ornithine decarboxylase over expressing keratinocytes

Ker/Norm

normal littermate keratinocytes

ODC

ornithine decarboxylase

TLR4

Toll-like receptor 4