Suppression of Ca²⁺ signaling enhances melanoma progression

Abstract

The role of store‐operated Ca²⁺ entry (SOCE) in melanoma metastasis is highly controversial. To address this, we here examined UV‐dependent metastasis, revealing a critical role for SOCE suppression in melanoma progression. UV‐induced cholesterol biosynthesis was critical for UV‐induced SOCE suppression and subsequent metastasis, although SOCE suppression alone was both necessary and sufficient for metastasis to occur. Further, SOCE suppression was responsible for UV‐dependent differences in gene expression associated with both increased invasion and reduced glucose metabolism. Functional analyses further established that increased glucose uptake leads to a metabolic shift towards biosynthetic pathways critical for melanoma metastasis. Finally, examination of fresh surgically isolated human melanoma explants revealed cholesterol biosynthesis‐dependent reduced SOCE. Invasiveness could be reversed with either cholesterol biosynthesis inhibitors or pharmacological SOCE potentiation. Collectively, we provide evidence that, contrary to current thinking, Ca²⁺ signals can block invasive behavior, and suppression of these signals promotes invasion and metastasis.

UV‐induced cholesterol production decreases Orai activation and SOCE in melanoma cells, promoting invasive phenotypes and anabolic pathways.

Introduction

Melanoma is a cancer of the cells of the melanocyte lineage, which are predominantly responsible for the melanin pigment production that defines skin and eye tone (Slominski et al, 2004). Melanoma incidence has progressively risen over the last 30 years (Schadendorf et al, 2018), and is now the fifth most common cancer among men and women in the US (Mea, 2019). Although melanoma is easily treated by surgical resection when detected early, if undetected, melanoma progresses from a radial to vertical growth phase to become an aggressive malignancy that easily metastasizes from the skin to the lymph nodes and other distal sites (Schadendorf et al, 2018). Currently, there are two major approaches to treating metastatic melanoma: targeted therapies and immunotherapies (Neves de Oliveira et al, 2018). Targeted therapies include inhibitors of targets in the Mitogen‐Activated Protein Kinase (MAPK) pathway, including oncogenic BRAF^V600E and MEK1/2 (trametinib) inhibitors. These drugs transiently increase progression‐free survival, although this is typically followed by drug resistance and relapse (Neves de Oliveira et al, 2018). Additionally, though promising, MAPK pathway inhibitors inhibit immune cell function, leading to increased clinical toxicity (Boni et al, 2010). Immunotherapies were first explored as first‐line melanoma treatments due to the observation that incidence of melanoma increased in immunosuppressed patients (Greene et al, 1981); targets of immune checkpoint inhibitors (ICIs) including Cytotoxic T‐Lymphocyte Antigen 4 (CTLA4) (ipilimumab) and Programmed Cell Death‐1 (PD1) (pembrolizumab). ICI therapies in conjunction with targeted therapies are currently in use. Continued efforts to identify new strategies are needed to control this deadly disease.

Ultraviolet (UV) radiation exposure is a major risk factor in melanoma, and the rising incidence of melanoma has been linked to increased sun exposure and artificial tanning (Matthews et al, 2017; Schadendorf et al, 2018). UV triggers melanin production in melanocytes, which protects against UV‐induced damage in skin (Brenner & Hearing, 2008). However, UV exposure affects multiple cellular processes ultimately promoting cancer initiation and progression; as such, UV is considered a “complete carcinogen,” (D'Orazio et al, 2013). The most immediate consequence of UV exposure is induction of pyrimidine dimers that can cause DNA damage and/or mutations. While the evidence that UV drives melanomagenesis is undeniable, the molecular pathways linking UV and driver mutations are unclear since the most common mutations associated with melanoma, such as BRAF and NRAS mutations, do not possess the UV signature mutations. Further, while UV has long been associated with melanomagenesis, UV also has been reported to drive melanoma progression (Bald et al, 2014; Arisi et al, 2018), marked by entry into the vertical growth phase (Guerry et al, 1993). The mechanism by which UV supports progression is not well understood and is thus the focus of the current study, which identifies a critical role for UV‐dependent suppression of store‐operated Ca²⁺ entry (SOCE) in regulating melanoma migration and invasion.

SOCE is mediated by the members of the Orai and STIM families. Orai channels are Ca²⁺‐selective ion channels located at the plasma membrane (PM) (Feske et al, 2006; Zhang et al, 2006; Vig et al, 2006b); STIM is a type 1A transmembrane protein primarily found in the endoplasmic reticulum (ER) membrane (Liou et al, 2005; Roos et al, 2005). STIM responds to ER Ca²⁺ depletion by translocating within the ER towards the PM, and physically associates with PM‐localized Orai channels to facilitate Ca²⁺ entry (i.e., SOCE (Putney, 1986); reviewed in (Soboloff et al, 2012)). SOCE has been linked to melanoma progression previously, although the nature of this relationship is somewhat unclear. Hence, a prior study from our group, revealed an inverse correlation between SOCE and invasiveness (Hooper et al, 2015), suggestive of an inverse relationship between SOCE and melanoma progression. Further, a randomized, unbiased ribozyme screen for metastasis related genes identified STIM1 as a repressor of melanoma metastasis (Suyama et al, 2004). In contrast, invasive melanoma has been shown to rely on enhanced SOCE and AKT activity for growth and survival (Feldman et al, 2010; Fedida‐Metula et al, 2012). Further, invasive melanoma exhibiting activated MAPKs were shown to exhibit higher STIM1 and Orai1 expression (Stanisz et al, 2014; Sun et al, 2014; Umemura et al, 2014). Given these seemingly contradictory findings, further effort is needed to understand the relationship between SOCE and melanoma progression, an important goal of the current investigation. Further, since UV exposure is the major driver of melanoma progression and the relationship between UV and SOCE has not previously been determined, this investigation focuses on UV‐dependent control of both SOCE and melanoma progression, using both in vitro and in vivo approaches.

Results

UV suppresses SOCE and enhances invasion in melanoma

We assessed the effect of UV on SOCE in melanoma in a panel of established non‐metastatic human melanoma cell lines containing a range of different common driver mutations common to melanoma (SKMEL5, UACC1273, SKMEL2, FS13, SKMEL28, and UACC257) (Bairoch, 2021) and (WM983, WM983BR) (Herlyn, 2021). Cells were exposed to UV (175 J/M²), then allowed to recover for 24 h. To induce SOCE, cells were treated with the Sarco/Endoplasmic Reticulum Calcium ATPase (SERCA) pump inhibitor thapsigargin (Tg; 2 μM) for 10 min to deplete the ER of Ca²⁺ prior to the addition of 1 mM Ca²⁺. SOCE could be detected following the addition of exogenous Ca²⁺ induced in all cell lines, although variable amplitudes were observed (Fig 1A). Prior exposure to UV suppressed SOCE in all cell lines tested 24 h after a single UV exposure (Fig 1A,B); no relationship to driver mutations in the different cell lines could be detected. To determine if this was a transient or stable effect, SOCE was tracked on a weekly basis for 5 weeks in WM983 and UACC257 cells after a single exposure to UV (Figs 1C and EV1A,B). WM983 cells exhibited significant SOCE suppression at 2–3 weeks after UV exposure followed by a second, steeper suppression of SOCE between 4 and 5 weeks whereas no significant changes in SOCE were observed in UACC257 cells in response to UV exposure. To determine if changes in UV‐dependent changes in SOCE were specific or reflective of a general shift in Ca²⁺ homeostasis, both basal and ER Ca²⁺ content were also measured in WM983 and UACC257 cells recovering from UV exposure (Fig EV1C,D). No significant changes were observed at any timepoint in either WM983 or UACC257 cells, leading us to conclude that UV exposure alters SOCE, but not basal or ER Ca²⁺ content.

Figure 1. Ultraviolet (UV) exposure suppresses store‐operated Ca²⁺ entry (SOCE) and enhances invasion.

• A–D
  Human melanoma cells were plated on glass coverslips overnight before exposure to 175 J/m² UV. (A) Representative traces showing the effect of UV on SOCE in eight human melanoma cell lines. Cells were incubated overnight before loading with Fura2‐AM; cells were treated with 2 μM Tg for 15 min in the absence of extracellular Ca²⁺ (not shown) before the addition of 1 mM Ca²⁺. Dashed lines indicate SEM. (B) UV‐dependent differences in SOCE were determined by comparing the maximal change in Fura2 ratio between UV‐treated and untreated cells after the addition of extracellular Ca². Each dot represents a biological replicate, representing 10–60 individual cells. Bars represent mean ± SEM. Data were analyzed by two‐way ANOVA with multiple comparisons (UV effect, P < 0.0001; effect of cell line, P > 0.05; interaction, P > 0.05). (C) UACC257 and WM983 cells were treated as described for panel A, except the incubation periods were extended as depicted (N ≥ 3 biological replicates); data displayed are mean ± SEM. Data were analyzed by two‐way ANOVA with multiple comparisons (UV effect, P < 0.01; Effect of cell line, P < 0.05; interaction, P > 0.05). (D) UV‐treated cells were placed in Matrigel transwell inserts in OptiMEM reduced serum media; DMEM with 10% FBS was used as a chemoattractant. Crystal violet‐stained invaded cells were measured 20 h after initiation of assay (N ≥ 4 biological replicates); data displayed are mean ± SEM. Sample images for each condition are shown below the corresponding bar. Data were analyzed by two‐way ANOVA with multiple comparisons (UV effect, P < 0.0001; Effect of cell line, P > 0.05; interaction, P < 0.01).
• E, F
  B16N murine melanoma cells were plated on glass coverslips overnight before exposure to 175 J/m² UV and then incubated under growth optimal conditions for 0, 2, or 5 weeks. (E; LEFT) SOCE was measured in B16N murine melanoma cells as in (A) and quantified as in (B) (RIGHT; N ≥ 15 biological replicates); data displayed are mean ± SEM. Data were analyzed by one‐way ANOVA with multiple comparisons (P < 0.05). (F) SOCE quantified as described in panel E is shown each week after UV.
• G
  B16N invasiveness was determined as in (D) (N ≥ 9 biological replicates); data displayed are mean ± SEM. Data were analyzed by one‐way ANOVA with multiple comparisons (P = 0.0001).
• H
  B16N cells were exposed to UV and incubated for the indicated time before subcutaneous injection into the right flank of syngeneic C57Bl6 mice. Shown, representative pictures of abdominal organs, peritoneal wall, and crude abdomen after 2 weeks of tumor cell in vivo growth (N ≥ 9).

Data information: Within all panels, differences established through multiple comparisons are labeled as *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure EV1. Extended calcium data from UVR‐treated human melanoma cells from Fig 1 .

WM983 and UACC257 human melanoma cells were plated overnight before exposure to 175 J/m² UVR. Cells were cultured under growth optimal conditions for the indicated time period before plating on glass coverslips and loading with Fura2‐AM.

• A, B
  Representative examples of Ca²⁺ measurements; recording began in the presence of Ca²⁺ (1 mM). Ca²⁺ was removed at least 1 min prior to the addition of Tg (2 μM) followed by the re‐addition of Ca²⁺ (1 mM). Each trace represents 40–60 individual cells; solid line represents the mean and the dotted line represents SEM.
• C
  Basal Ca²⁺ is the average Fura2 ratio during the first minute.
• D
  Tg‐induced Ca²⁺ release was quantified by calculating “area under curve” to determine total Ca²⁺ release after the addition of Tg, but prior to the re‐addition of Ca²⁺.

Data information: Data from panels C and D (n ≥ 3 biological replicates; data displayed are mean ± SEM) were analyzed by one‐way ANOVA (P > 0.05).

To determine if differences in the stability of UV‐induced SOCE suppression in WM983 and UACC257 cells correlated with invasiveness, transwell invasion assays were performed on non‐irradiated cells or cells exposed to UV and allowed to recover for either 2 or 5 weeks (hereafter termed UV2 or UV5, respectively; Fig 1D). Notably, whereas several‐fold increases in the number of invasive UV2 and UV5 WM983 were observed, no UV‐induced changes in UACC257 invasiveness were observed. To determine if increased invasiveness might reflect UV‐dependent effects on cell viability and/or proliferation, we performed WST‐1 proliferation assays (Appendix Fig S1A,B), with no UV‐dependent differences found in WM983 or UACC257 cells. These data demonstrate a direct correlation between SOCE suppression and invasive behavior in UV‐induced WM983, but not UACC257 melanoma progression, along with an intriguing delay between UV exposure and reprogramming of invasive responses.

To determine if UV‐induced SOCE suppression and invasiveness would also occur in vivo, we extended our study to B16N melanoma cells, with the ultimate goal of assessing metastasis in immunocompetent C57Bl/6 mice. Paralleling results in human melanoma cells, UV exposure led to SOCE suppression after a 5‐week incubation (Fig 1E,F), with a similar increase in invasion through Matrigel in a transwell invasion assay in vitro (Fig 1G), with no significant effect on cell proliferation (Appendix Fig S1C). In complementary work, untreated, UV2 or UV5 B16N cells were injected s.c. in syngeneic male C57BL/6 mice and allowed to grow for 2 weeks before being euthanized and assessing metastasis by pathology. Consistent with prior studies (Valle et al, 1992; Overwijk & Restifo, 2001), in 26 out of 28 mice injected with untreated B16N cells, no evidence of metastasis was found (Table EV1). In contrast, over half of the mice (8/15) injected with UV2 B16N cells had detectable metastases, as did 79% (19/24) of mice injected with UV5 B16N cells (Table EV1). Interestingly, for all B16N models, infiltration was observed throughout the peritoneum, but not in the lungs (Fig 1H; Table EV1). While it is more typical for melanoma to metastasize to the lungs, these data nevertheless demonstrate a potent UV‐dependent shift towards invasive behavior.

Melanoma progression mediated by SOCE suppression exhibits a “Goldilocks effect”

UV has a wide range of short‐ and long‐term effects on cell function (Coelho et al, 2009), only one of which is SOCE suppression. To more directly establish a relationship between suppressed SOCE and invasiveness, we suppressed SOCE in B16N cells using 3,5‐Bis(trifluoromethyl)pyrazole (BTP2), a well‐established inhibitor of both Orai1‐mediated SOCE and TRPC‐mediated Ca²⁺ entry (Ishikawa et al, 2003; Zitt et al, 2004; He et al, 2005; Schleifer et al, 2012). Titration of BTP2 at a concentration range of 20–100 nM suppressed SOCE at levels comparable to UV (Fig 2A,B) and additionally stimulated several‐fold increases in invasiveness through transwell invasion assays (Fig 2C). The effect on invasiveness was lost at higher BTP2 concentrations (e.g. 200 nM) suggesting an intriguing “Goldilocks effect.” Similar observations were made in WM983 cells (Fig EV2A,B). To further probe this concept, non‐invasive UV2 B16N melanoma cells were treated with BTP2 (20 nM), a concentration that had no effect on invasiveness in parental B16N cells (Fig 2D). Remarkably, increased invasiveness was observed, suggesting that SOCE was insufficiently suppressed in both cases to drive an increase in invasiveness through Matrigel. We then assessed the contribution of SOCE to the enhanced invasiveness of B16N UV5 cells by measuring invasiveness in the presence of BTP2 (200 nM; Fig 2E). This completely blocked invasiveness, indicating that a minimum amount of SOCE is needed to support invasive behavior, consistent with the proposed “Goldilocks Effect.” We note that these are very low concentrations of BTP2, well below toxic levels (Ishikawa et al, 2003; Zitt et al, 2004).

Figure 2. Partial store‐operated Ca²⁺ entry (SOCE) suppression increases invasiveness.

• A
  Representative traces of SOCE in Fura‐2 loaded B16N cells treated with the indicated concentrations of BTP2 for 15 min, in the presence of Tg. A representative trace of UV5 B16N is also shown for reference.
• B, C
  (B) SOCE from (A) was quantified; dose dependence was measured by non‐linear regression (n = 6 biological replicates); data displayed are mean ± SEM. UV5 SOCE is shown as a purple bar for comparison (C) BTP2‐treated B16N cells were plated on Matrigel transwell inserts in OptiMEM reduced serum media; DMEM with 10% FBS was used as a chemoattractant. Crystal violet‐stained invaded cells were measured 20 h after initiation of assay (N ≥ 3 biological replicates); data displayed are mean ± SEM. Sample images for each condition are shown above the corresponding bar. Data were analyzed by one‐way ANOVA with multiple comparisons (P < 0.01).
• D
  UV2 or (E) UV5 B16N cells treated with indicated concentrations of BTP2 followed by plating on matrigel transwell inserts as described (C).
• F
  Representative traces of SOCE in two Orai1‐CFP‐expressing cell lines exhibiting SOCE suppression similar to BTP2 (40 nM).
• G
  Relative SOCE suppression observed in Orai1‐expressing or UV5 B16N cells (N ≥ 5 biological replicates); data displayed are mean ± SEM. Data were analyzed by one‐way ANOVA with multiple comparisons (P < 0.01).
• H
  Transwell migration assays of Parental, UV2, UV5, O1C3, and O1C44 B16N cells as described in (C) (N ≥ 6 biological replicates); data displayed are mean ± SEM. Sample images for each condition are shown below the corresponding bar. Data were analyzed by one‐way ANOVA with multiple comparisons (P < 0.001).
• I
  Representative traces of SOCE in B16N cells stably expressing the dominant negative form of Orai1‐CFP (E106A; O1‐DN).
• J
  Transwell migration assays of B16N parental versus O1‐DN‐expressing cells as described in (C) (N ≥ 4). Data were analyzed by unpaired T test (P < 0.05).
• K
  Quantitation of B16N cell monolayers 20 h after being scratched (representative images in Appendix Fig S3). UV2, UV5 and O1C44 cells were incubated overnight in the presence (+) or absence (−) of 2 μM IA65. The cell‐free area was quantified by ImageJ (N ≥ 3 biological replicates); data displayed are mean ± SEM. Data were analyzed by two‐way ANOVA. In the absence of IA65, SOCE suppression effect was P < 0.05; effect of time was P < 0.0001; interaction was P > 0.05. In the presence of IA65, SOCE suppression effect was P > 0.05; effect of time was P < 0.0001; interaction was P < 0.01.
• L–N
  B16N parental, O1C44, and O1DN cells were subcutaneously injected in the right flank of syngeneic C57Bl6 mice and permitted to grow for 14 days. Histological analysis reveals the presence of O1C44 (but not B16N‐WT redisplayed from Fig 1H or O1DN) cells in abdominal organs and within the peritoneal wall (L). (M) H&E stains of the peritoneal wall revealing the presence of melanoma cells in the peritoneal wall. (N) Quantitative measurement of degree of invasion within and through the peritoneal wall. Data were analyzed using the Kruskal–Wallis Rank Sum test (P < 0.001).

Data information: Within all panels, differences established through multiple comparisons are labeled as *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure EV2. Store‐operated Ca²⁺ entry (SOCE) Suppression in Human Melanoma Cells Exhibits a “Goldilocks Effect”.

• A
  Representative traces of SOCE in Fura‐2 loaded WM983 cells treated with the indicated concentrations of BTP2 for 15 min in the presence of Tg. Data are representative of five experiments. UV2 and UV5 WM983 cells are also shown for comparison.
• B
  BTP2‐treated WM983 cells were plated on Matrigel transwell inserts in OptiMEM reduced serum media; DMEM with 10% FBS was used as a chemoattractant. Invaded cells were measured after 20 h after staining with crystal violet (N ≥ 3 biological replicates; data displayed are mean ± SEM). Sample images for each condition are shown above the corresponding bar. UV2 and UV5 WM983 cells are also shown for comparison; data were analyzed by one‐way ANOVA (P < 0.0001).
• C
  Representative traces of SOCE in parental and an Orai1‐CFP‐expressing WM983 cells named O1C78 exhibiting SOCE suppression similar to BTP2 (80 nM).
• D
  Invasion assays performed as in panel B comparing the number of invaded cells in O1C78 cells to either untreated parental WM983 cells or WM983 cells exposed to UV irradiation and allowed to recover for 2 weeks (N ≥ 4). Sample images for each condition are shown below the corresponding bar. Data were analyzed by one‐way ANOVA (P < 0.0001).

Data information: Within all panels, differences established through multiple comparisons are labeled as *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

As another approach, we asked if genetic manipulation of SOCE would lead to a similar outcome. Overexpression of Orai1 has been reported to suppress SOCE through stoichiometric imbalance with STIM1 (Soboloff et al, 2006; Hoover & Lewis, 2011). Therefore, we generated a series of B16N cell lines stably expressing Orai1 and screened them to determine the level of SOCE suppression. Two Orai1‐overexpressing clones that exhibited SOCE suppression at levels similar to UV5 B16N cells were identified and designated O1C3 and O1C44 (Fig 2F,G). In both cases, these cells demonstrated enhanced invasion through Matrigel in a transwell invasion assay, to levels similar to those seen with UV5 B16N cells (Fig 2H). Similar observations were made in an Orai1‐overexpressing clone of WM983 cells (Fig EV2C,D).

Finally, we assessed the impact of a dominant negative Orai1 mutant (Orai1‐E106A‐CFP; Orai1‐DN; Prakriya et al, 2006; Yeromin et al, 2006; Vig et al, 2006a) previously shown to block invasion (Sun et al, 2014; Umemura et al, 2014). B16N cells stably expressing Orai1‐DN (O1DN) exhibited both ablated SOCE (Fig 2I) and diminished invasion in a transwell invasion assay (Fig 2J). These observations reveal a non‐linear relationship between SOCE and invasiveness in melanoma, in which partial suppression enhances invasion, but SOCE blockade abrogates it.

An important property of invasion is migration, which is regulated in part by transient changes in Ca²⁺ levels that facilitate filipodia formation, lamellipodia attachment, and activation of myosin light chain kinases (MLCKs) that control movement (reviewed in Gross et al, 2020). To assess the effect of SOCE on migratory potential, O1C44 and UV5 B16N cells were seeded as monolayers, grown overnight, scratched to create a wound, then imaged every 20 min for 20 h. Both O1C44 and UV5 B16N cells exhibited greater migratory behavior compared to control cells (Fig 2K; Appendix Fig S2). Interestingly, SOCE suppression in the absence of UV had a more profound effect on migration, with Orai1‐overexpressing cells nearly completely filling the wound 10 h after initial scratch. Considered together, these data demonstrate SOCE suppression as a driving force behind migration and a mediator of UV‐dependent enhancement of melanoma invasion.

To determine if SOCE suppression is sufficient to drive invasiveness in syngeneic mice, Orai1‐overexpressing B16N cells were injected s.c. and metastasis was assessed by pathology. Similar to mice injected with UV‐irradiated cells, extensive metastasis was observed throughout the peritoneal cavity in 13 of 14 mice (Table EV1; Fig 2L). Although less common for melanoma, there is evidence showing that several metastatic cancers have been known to penetrate the peritoneal cavity through the peritoneal wall, a process commonly referred to as intraperitoneal carcinomatosis (IC) (Lee et al, 2014; Mikuła‐Pietrasik et al, 2018; McBride & Calhoun, 2019; Desai & Moustarah, 2020). Additionally, when O1DN cells were injected s.c., three of 10 mice exhibited contact with the peritoneal wall (Table EV1; Fig 2L,M). Further, high resolution images of the peritoneal wall reveal no untreated B16N cells, but significant penetration of UV‐irradiated or Orai1‐expressing B16N cells (Fig 2M,N). The three mice containing O1DN tumors showed less than 0.1 mm² of total invasion, even less than the two control mice, exhibiting peritoneal wall invasion (Fig 2N). Considered collectively, these data reveal that modest SOCE suppression promotes metastasis from an s.c. melanoma, while SOCE ablation has minimal or no effect.

SOCE recovery inhibits invasion

To further establish the concept that SOCE suppression contributes to UV‐induced invasiveness, we used 4‐((5‐Phenyl‐4‐(trifluoromethyl)thiazol‐2‐yl)amino)benzoic acid (IA65), a potentiator of Orai1 and SOCE (Azimi et al, 2020; Zhang et al, 2020). IA65 had no effect on cell viability (Appendix Fig S3A) or on SOCE in untreated WM983 cells, but strikingly reversed UV‐induced SOCE suppression (Fig 3A,B). Notably, WM983 cells treated with IA65 failed to invade through Matrigel in a transwell assay even after exposure to UV (Fig 3C) and did not have UV‐induced enhanced cell migration (Fig 3D–F). Similar observations of the effect of IA65 on viability (Appendix Fig S3A,B), SOCE (Appendix Fig S4A,B,D,E), and invasion (Appendix Fig S4C,F) were made in B16N cells. Collectively, these observations indicate that SOCE suppression is a required component of UV‐induced increases in invasiveness in melanoma cells.

Figure 3. Enhancing store‐operated Ca²⁺ entry (SOCE) suppresses ultraviolet (UV)‐induced invasiveness.

WM983 cells were plated on glass coverslips overnight before exposure to 175 J/m² UV and then incubated under growth optimal conditions for 0 (control), 2 (UV2) or 5 (UV5) weeks.

• A, B
  (A) Representative traces of SOCE in Fura2‐loaded cells treated with 2 μM IA65 or vehicle (Control) in the presence of Tg in nominally Ca²⁺‐free media for 15 min prior to the addition of Ca²⁺ (1 mM) (B) Quantitation of SOCE from panel A (N ≥ 4 biological replicates); data displayed are mean ± SEM and analyzed by two‐way ANOVA (effect of UV, P > 0.05; IA65, P < 0.001; interaction was P < 0.01).
• C
  Control, UV2 and UV5 cells were treated with IA65 (0 or 2 μM) for 15 min before being placed in Matrigel transwell inserts in OptiMEM reduced serum media; DMEM with 10% FBS was used as a chemoattractant. Crystal violet‐stained invaded cells were measured 20 h after initiation of assay (N ≥ 4 biological replicates); data displayed are mean ± SEM. Sample images for each condition are shown above the corresponding bar. Data were analyzed by two‐way ANOVA (effect of UV, P < 0.01; IA65, P < 0.0001; interaction was P < 0.01).
• D–F
  Confluent layers of WM983 cells treated as indicated were ‘scratched’ followed by a 20 h incubation at 37°C with 5% CO₂. (D) Representative images of cells after being scratched after being incubated for 0, 10 or 20 h in the absence (−) or presence (+) of 2 μM IA65. (E, F) The cell‐free area was quantified by ImageJ (N ≥ 8 biological replicates); data displayed are mean ± SEM. Data were analyzed by two‐way ANOVA. In the absence of IA65, SOCE suppression effect was P < 0.05; effect of time was P < 0.0001; interaction was P < 0.05. In the presence of IA65, SOCE suppression effect was P > 0.05; effect of time was P < 0.0001; interaction was P > 0.05.

Within all panels, differences established through multiple comparisons are labeled as *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Identification of the mechanism of UV‐induced SOCE suppression

To determine the mechanism of UV‐induced SOCE suppression, we first measured the level of expression of the five members of the STIM and Orai families by RT‐qPCR (Appendix Fig S5). No significant UV‐induced changes in the expression of Orai genes were observed in WM983 or B16N cells (Appendix Fig S5A,C), although significant UV‐induced increases in Orai1 expression did occur in UACC257 (P = 0.0441; Appendix Fig S5B). Since no significant UV‐induced changes in SOCE were observed in UACC257 cells, the biological significance of this observation is unclear. Both O1C78 WM983 and O1C44 B16N cells exhibited a modest, but statistically significant increase in Orai1 expression (Appendix Fig S5A,C). No significant changes in the expression of either STIM1 or STIM2 were observed in WM983, UACC257 or B16N cells (Appendix Fig S5D–F). Since no decreases in STIM or Orai expression were observed in UV5 cells, these data do not provide an explanation for UV‐induced SOCE suppression. Therefore, we performed RNAseq, taking advantage of the observed differences in SOCE suppression in WM983 and UACC257 cells 2 weeks after UV exposure, comparing each cell model to non‐irradiated cells (see Figs 1C and 4A–C). Principal component analysis (PCA) revealed both minimal replicate variability and the existence of four discrete groups, with UV having distinct effects in WM983 and UACC257 cells (Fig 4D). Comparison of the genes changed in these cell types reveals marked expression differences in the effect of UV on UACC257 and WM983 cells (Fig 4E). To identify those genes whose expression was altered by UV that could contribute to SOCE suppression, we then performed a Venn analysis comparing the genes whose expression was significantly altered (P < 0.01, q < 0.01) by UV exposure in WM983, but not UACC257 cells (Fig 4F), performing Ingenuity Pathway Analysis (IPA) on this group of 421 genes. Interestingly, the pathway whose expression was most altered was the superpathway of cholesterol (CHL) biosynthesis (also known as the mevalonate pathway; Appendix Table S1) with eight of 28 genes upregulated (Table EV2). CHL has been previously identified as an endogenous inhibitor of SOCE due to a direct interaction with Orai1 (Derler et al, 2016). Further, the mevalonate pathway has been to shown to contribute to melanoma progression in vivo (Pencheva et al, 2014). Finally, we performed Cancer Genome Atlas (TCGA) analysis to determine if the mevalonate pathway was similarly altered in clinical tumor samples (Fig 4G). Notably, melanoma was the second most likely tumor‐type to exhibit mevalonate pathway dysregulation (3.4%); the most likely tumor type to exhibit mevalonate pathway dysregulation was anal cancer; however, since this was based on only one tumor (out of only 21 total), this is likely irrelevant. Therefore, we assessed the possibility that enhanced CHL biosynthetic pathways contributes to SOCE suppression and UV‐induced melanoma invasiveness and metastasis.

Figure 4. Mevalonate pathway upregulation correlates with ultraviolet (UV)‐induced store‐operated Ca²⁺ entry (SOCE) suppression.

• A–C
  WM983 and UACC257 cells were plated on glass coverslips overnight before exposure to 175 J/m² UV and then incubated under growth optimal conditions for 0 (control) or 2 (UV) weeks followed by plating on glass coverslips, Fura2 loading and administration of Tg in nominally Ca²⁺‐free media for 15 min prior to the addition of Ca²⁺ (1 mM), as in Fig 1A–C (N ≥ 4 biological replicates); data displayed are mean ± SEM.
• D–F
  RNA was extracted from WM983 or UACC257 cells 2 weeks after UV exposure (0 or 175 J/m²) followed by cDNA production and analysis (three biological replicates per condition). (D) Principal component (PC) analysis showing the variability between the replicates. (E) Heatmap comparing genes significantly differing in WM983 and UACC257 cells 2 weeks after UV exposure (P < 0.01, q < 0.01). (F) Venn diagram showing the genes significantly changed due to UV in WM983 but not significantly changed in UACC257 cells.
• G
  The percentage of tumors in which the mevalonate pathway by tumor type; data were obtained from TCGA. The numbers indicate the total number of samples available for each tumor type; the.
• H–K
  WM983 and UACC257 human melanoma cells were transfected with Perfringolysin O (PFO)‐mCherry D4H cholesterol sensor and plated on glass coverslips. (H–J) Cells were exposed to UV (0 or 175 J/m²) and incubated for 0, 2 or 5 weeks as indicated. (H) Representative cell images. (I) Fluorescence intensity from cells analyzed in (H) was measured and quantified using LASX analysis software (N ≥ 8 biological replicates); data displayed are mean ± SEM. Data were analyzed using 1‐way ANOVA with multiple comparisons (WM983, P < 0.001; UACC257 P > 0.05). **P < 0.01, ***P < 0.001, ****P < 0.0001. (J, K) O1C78 WM983 cells expressing the PFO‐mCherry D4H cholesterol sensor were examined as in panels H and I (N ≥ 8 biological replicates); data displayed are mean ± SEM.

To determine if UV treatment increased CHL content, we designed CHL sensors based on Perfringolysin O (PFO) toxin, a toxin with CHL‐binding activity (Maekawa, 2017). Briefly, the CHL‐binding portion of the PFO toxin was mutated to increase affinity with a D434S point mutation within the D4 domain (D4H), tagged with mCherry and transfected into WM983 and UACC257 cells (Fig 4H,I). UV markedly increased mCherry expression in WM983 in a time‐dependent manner, but did not do so in UACC257 cells (Fig 4H,I), consistent with the effect of UV on SOCE on these two cell types (Fig 1C). Analogous experiments were done in an Orai1 overexpressing clone of WM983 cells (O1C78) and showed no significant change in mCherry expression, revealing that cholesterol upregulating is upstream of SOCE suppression and is likely not bidirectional (Fig 4J,K). To assess if UV‐induced CHL production contributes to UV‐induced SOCE suppression, cells were treated with either the HMG‐CoA reductase inhibitor compactin (a statin drug) or the squalene synthase (SS) inhibitor TAK475, which targets a downstream step in the CHL biosynthesis pathway. While neither agent significantly affected SOCE in untreated WM983 cells, SOCE in UV5 WM983 cells was markedly enhanced after a 24 h treatment with either agent (Fig 5A,B). Further, when these same CHL inhibitor drugs were given to non‐SOCE suppressed UACC257 UV5 cells, no change occurred in SOCE (Appendix Fig S6A,B). Additionally, both drugs abrogated UV‐induced increases in WM983 melanoma invasiveness through Matrigel on a transwell assay, while having no effect on invasiveness in untreated cells (Fig 5C).

Figure 5. Cholesterol production is a key mediator of ultraviolet (UV)‐induced metastasis.

WM983 cells were plated overnight before exposure to 175 J/m² UV and then incubated under growth optimal conditions for 0 (control), 2 (UV2) or 5 (UV5) weeks.

• A
  Following UV exposure and incubation, WM983 human melanoma cells were plated on glass coverslips and treated with or without the cholesterol synthesis inhibitors TAK475 or compactin overnight followed by Fura2 loading and administration of Tg in nominally Ca²⁺‐free media for 15 min prior to the addition of Ca²⁺ (1 mM).
• B
  Store‐operated Ca²⁺ entry (SOCE) from panel A (N ≥ 5 biological replicates; data displayed are mean ± SEM) was quantified and compared by two‐way ANOVA (cholesterol inhibition was P = 0.0014; incubation time, P > 0.05; interaction was P < 0.0001).
• C
  UV5 or control WM983 cells were placed in Matrigel transwell inserts in OptiMEM reduced serum media with or without TAK475 or compactin; DMEM with 10% FBS was used as a chemoattractant. Crystal violet‐stained invaded cells were measured 20 h after initiation of assay (N ≥ 4 biological replicates); data displayed are mean ± SEM. Sample images for each condition are shown above the corresponding bar. Data were analyzed by two‐way ANOVA (cholesterol inhibition was P = 0.0004; incubation time, P > 0.05; interaction was P = 0.0002).
• D
  B16N murine melanoma cells were treated with or without UV and incubated for 5 weeks. Ldlr ^−/− C57Bl6 mice were fed with regular chow or high fat diet for 2 weeks prior to administration of syngeneic B16N cells (N = 10 biological replicates; data displayed are mean ± SEM). The area of cell invasion into and through peritoneal wall measured using Aperio Imagescope software and analyzed by two‐way ANOVA (high fat diet, P > 0.05; UV exposure, P > 0.05; interaction was P < 0.05).
• E
  Representative images of invasion into the peritoneal wall.

Data information: Within all panels, differences established through multiple comparisons are labeled as *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Finally, to assess the contribution of CHL on melanoma metastasis in vivo, we analyzed the metastasis of parental and UV5 B16N cells grown for 5 weeks post‐irradiation before implantation s.c. in Ldlr ^−/− mice (Fig 5D,E). Ldlr ^−/− mice exhibit relatively modest 50–300% increases in CHL levels when fed regular chow and a >800% increase in CHL when fed high fat diet in comparison to WT C57Bl/6 mice (Hartvigsen et al, 2007; Franciosi et al, 2009). Interestingly, very modest invasion of untreated B16N cells into the peritoneal wall was observed in mice fed regular chow that was dramatically increased by high fat chow (Fig 5D,E). UV5 B16N cells exhibited substantial invasiveness in Ldlr ^−/− mice fed either regular chow or high fat diet. These data demonstrate that elevated CHL levels are sufficient to mimic UV‐induced melanoma metastasis and support a model in which UV drives melanoma metastasis due to SOCE suppression mediated by increased CHL biosynthesis.

SOCE suppression drives a metabolic shift towards anabolic pathways

In an effort to determine how SOCE suppression might contribute to invasive behavior in melanoma, we performed RNAseq analysis of untreated, UV2, UV5, and O1C44 B16N cells (Fig 6A,B). PCA revealed minimal replicate variability and significant differences from control for each experimental condition (Appendix Fig S7A–C) and marked differences in expression patterns between each of the 3 groups (Fig 6A; Appendix Fig S7D). To identify genes associated with both SOCE suppression and invasiveness, we then performed a Venn analysis comparing the genes whose expression was significantly altered (P < 0.01, q < 0.01) in UV5, O1C44, and/or UV2. The exclusion of UV2 in one group was performed because these cells did not exhibit SOCE suppression (Fig 1E,F), were not invasive through a transwell assay (Fig 1G) but did exhibit metastatic properties in a small subset of mice (Fig 2M,N; Table EV1). There were 187 genes changed that were common to all three groups, including multiple genes associated with epithelial–mesenchymal transition (EMT), melanocyte dedifferentiation, metabolism, and cell survival (Table EV3). If the non‐SOCE suppressed UV2 cells are excluded, there were 472 genes changed. Interestingly, with the exception of dedifferentiation, the number of pathways changed in each category increased when UV2 was excluded, particularly as related to control of glucose uptake and metabolism (Table EV3). Dysregulation of both Ca²⁺ signals and metabolism in the context of cancer has been extensively reported (Reviewed in Prevarskaya et al, 2018; Dejos et al, 2020); since all gene changes were observed in O1C44 cells in which SOCE is suppressed, we assessed metabolic function.

Figure 6. RNAseq reveals store‐operated Ca²⁺ entry (SOCE) suppression drives invasion.

• A, B
  RNA was extracted from UV2, UV5, O1C44 or control B16N murine melanoma cells followed by RNAseq analysis (three biological replicates per condition). (A) Heatmap containing gene expression differences. (B) Venn diagram comparing genes significant changed in B16N cells 2 weeks (UV2) or 5 weeks (UV5) after exposure to those changed by Orai1 overexpression (O1C44) (P < 0.01).
• C
  Glucose uptake measured by luciferase‐based assay (B16N data based on 6 biological replicates each including four technical replicates; WM983 and UACC257 data based on three biological replicates each including four technical replicates; data displayed are mean ± SEM). Data were analyzed by one‐way ANOVA (B16N, P < 0.0001; WM983, P < 0.0001; UACC257, P > 0.05).
• D, E
  Oxygen consumption rate (OCR; D) extracellular acidification rate (ECAR; E), were measured in Control, UV5 and O1C44 B16N melanoma cells (n ≥ 15), Control, UV5 and O1C78 WM983 melanoma cells (n ≥ 10) and Control, UV2 and UV5 UACC257 melanoma cells (B16N data based on six biological replicates each including four technical replicates; WM983 and UACC257 data based on three biological replicates each including four technical replicates; data displayed are mean ± SEM) using an Agilent Seahorse XF96.
• F
  Basal respiration and basal glycolysis data were determined based on data in panels D and E, respectively, prior to the addition of oligomycin and glucose, respectively. Data displayed are mean ± SEM was analyzed by one‐way ANOVA (P < 0.0001).
• G
  5 × 10⁵ WM983 cells were transfected with 25 pmol siOGT (Invitrogen) or a scrambled control RNA (Ambion). Cells were split into two groups and incubated under growth optimal conditions for 2 days. LEFT: Cells were lysed and qPCR for OGT1 was performed using 18S ribosomal gene as a control. Data were analyzed by two‐way ANOVA (siRNA effect was P < 0.0001; no change in OGT1 expression between control, UV5 or O1C78), RIGHT: Cells were plated in Matrigel transwell inserts in OptiMEM reduced serum media. Crystal violet‐stained invaded cells were measured 20 h after initiation of assay (N ≥ 3 biological replicates; data displayed are mean ± SEM). Invasion data were analyzed by two‐way ANOVA (siOGT effect was P = 0.0002; interaction of treatment and siOGT was P = 0.0449). Sample images for each condition are shown below the corresponding bar.
• H
  B16N or WM983 cells were placed in Matrigel transwell inserts in OptiMEM reduced‐serum media, with or without OSMI1 (40 μM); DMEM with 10% FBS was used as a chemoattractant. Crystal violet‐stained invaded cells were measured 20 h after initiation of assay (N ≥ 4 biological replicates; data displayed are mean ± SEM). Sample images for each condition are shown below the corresponding bar. Data were analyzed by two‐way ANOVA (OSMI1 effect was P < 0.0001 in B16N and WM983 cells).

Data information: Within all panels, differences established through multiple comparisons are labeled as *P < 0.05, **P < 0.01, ****P < 0.0001.

Analysis of glucose uptake revealed significant many‐fold increases in the highly invasive UV5 and O1C44 B16N cells, UV5 and O1C78 WM983 cells, although not in UV5 UACC257 cells (Fig 6C). Further, analysis of oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) revealed only modest UV or Orai1‐dependent changes in OCR in B16N, WM983, and UACC257 cells (Fig 6D) and either decreased or unchanged ECAR (Fig 6E). Measurements of Basal Respiration and Basal Glycolysis show either decreased or unchanged aerobic respiration of glycolytic activity across both human and murine cells (Fig 6F). Importantly, these changes in ECAR and OCR did not correlate well with the invasiveness phenotypes of these cells, implying the highly elevated levels of imported glucose were being used for alternative purposes than glycolysis.

Although most glucose is normally processed through glycolysis, glucose can also be shunted down an alternative biosynthetic route, the hexosamine biosynthetic pathway (HBP). In normal cells, this represents roughly 2–5% of glucose consumption, however, cancer cells have been shown upregulate use of the HBP (Marshall et al, 1991; Akella et al, 2019). Elevated HBP increases uridine diphosphate‐N‐acetylglucosamine (UDP‐GlcNAc) production, which ultimately drives post‐transcriptional O‐GlcNAcylation mediated by the enzyme O‐linked N‐acetylglucosamine (GlcNAc) transferase (OGT), promoting EMT, proliferation, cell survival, and transcription (Akella et al, 2019). Interestingly, RT‐qPCR analysis of OGT gene expression revealed either unchanged or decreased levels of OGT in response to UV exposure or SOCE suppression (Fig 6G) which implies that changes in HBP activity are likely caused by increased substrate and OGT activity. To determine whether elevated OGT activity contributes to the increased invasiveness of cells with depressed SOCE, cells were either transfected with a siRNA targeting OGT (siOGT; Fig 6G) or incubated with specific and selective pharmacological inhibitor of OGT, OSMI1 (Barkovskaya et al, 2019) (Fig 6H). OGT knockdown decreased OGT expression between 50–70% across all conditions relative to scrambled RNA control (Fig 6G). Further, both OGT knockdown and OGT inhibition using OSMI1 blocked invasiveness for both WM983 and B16N cells (Fig 6H). These observations reveal a critical role for OGT activity in the observed increase in invasiveness due to SOCE suppression in melanoma cells.

Patient surgical explants have low SOCE and a similar pharmacological profile to melanoma cell lines

To determine the extent to which the observations reported here can be extended to human melanoma, we used a series of freshly isolated primary patient surgical explants (Fig 7). The majority of patient samples exhibited a “low SOCE” profile (Fig 7A), similar to the SOCE suppressed samples reported elsewhere in this study. Given this profile, it is not entirely surprising that UV‐irradiation failed to further suppress SOCE across all 27 samples (Fig 7B) although the top 15% of SOCE responders (4 highest of 27 tested) tended to respond more to UV exposure (Fig 7A). We were able to obtain both primary and metastatic lymph nodes (LN) tumor samples from two of the 27 patients (P4 and P8; Fig 7C); the LN tumors in both patients exhibited substantially lower SOCE than the primary tumor (Fig 7C) consistent with the concept that SOCE suppression is a component of metastasis. Finally, since all samples were genetically profiled for somatic mutations (Appendix Table S2), we attempted to determine if SOCE levels had any link to mutation status. Consistent with prior studies (Esteves et al, 2020), BRAF mutations were associated with significantly elevated SOCE (Fig 7D); although, inconsistent with this previous study, statistically significant changes in SOCE due to NRAS mutations were not observed (P = 0.09; Fig 7D), although, we attribute this discrepancy primarily to sample size. No other relationships between SOCE and mutations were observed within the statistical power of our sample size.

Figure 7. Analysis of patient surgical explants reveals clinical relevance of low SOCE and cholesterol.

Patient surgical explants were received from the biosample repository of FCCC. Cells were isolated from tumors using 50 μM Medicons and the Medimachine system and grown for 3 days before experimentation to eliminate non‐melanoma cells.

• A, B
  Plated control or UV irradiated (175 J/m²) cells were then incubated overnight on glass coverslips prior to Fura2 loading and administration of Tg in Ca²⁺‐free media for 15 min. (A) Representative examples of SOCE responses in patient samples following the addition of Ca²⁺ (1 mM). Each trace represents five to 10 individual cells; solid line represents the mean and the dotted line represents SEM. (B) Each dot represents SOCE in cells obtained from a different patient sample (n = 20; data displayed as mean ± SEM). The effect of UV on SOCE was determined by one‐way ANOVA (P > 0.05).
• C
  Comparison of SOCE in cells collected from both primary tumors and metastatic lymph nodes from the same patient (from patients analyzed in panels A, B).
• D
  SOCE in primary patient samples exhibiting NRAS (8 of 27 samples), BRAF (6 of 27 samples), TERT promoter (6 of 27 samples), NF1 (5 of 27 samples), and CDKN2A (5 of 27 samples) alterations were compared to SOCE in samples that expressed WT or the unaltered gene. Data displayed as mean ± SEM.
• E
  Single cells from patient samples were treated with or without the indicated drug in the presence of Tg in nominally Ca²⁺‐free media for 15 min prior to the addition of Ca²⁺ (1 mM). The number on each bar represents the number of patient specimens tested. Data were compared against their matched control analyzed by the unpaired T‐test (Tak475, P = 0.015; compactin, P = 0.0049; IA65, P = 0.0390).
• F
  Patient cells were placed in Matrigel transwell inserts in OptiMEM reduced serum media with or without TAK475 (10 μM), compactin (2 μM), IA65 (2 μM) or OSMI1 (40 μM); DMEM with 10% FBS was used as a chemoattractant. Crystal violet‐stained invaded cells were measured 20 h after initiation of assay. Each dot represents 1 patient sample. Data analyzed by two‐way ANOVA (cholesterol inhibition was P < 0.0001; patient samples were P < 0.001; interaction was P < 0.01). For IA65, (inhibition was P < 0.0001; patient samples were P < 0.01; interaction was P < 0.0001). For OSMI1, (inhibition was P < 0.05; patient samples were P > 0.05; interaction was P < 0.05).

Data information: *P < 0.05, **P < 0.01, ****P < 0.0001.

To assess the effect of inhibition of CHL biosynthesis and SOCE augmentation on human melanomas, several of the patient samples were treated with the mevalonate pathway inhibitors compactin and TAK475 and the Orai potentiator IA65 (Fig 7E). Both TAK475 and compactin significantly enhanced SOCE, indicating SOCE was being suppressed by CHL biosynthesis in the clinical samples. No significant effect of IA65 was observed, although a smaller pool was tested due to limited sample availability. Finally, treatment with mevalonate pathway inhibitors compactin and TAK475, the Orai1 potentiator IA65 or the OGT inhibitor OSMI1 all led to diminished invasion in vitro (Fig 7F). Overall, these observations reveal the clinical relevance of CHL‐mediated SOCE suppression and subsequent O‐GlcNAcylation in melanoma.

Discussion

In the current study, we have demonstrated that a single exposure to UV suppresses SOCE, which promotes melanoma migration, invasion, and metastasis. This effect is both rapid, occurring within 24 h after UV exposure, and durable for at least 5 weeks after exposure. This effect is reversible, in that an Orai1 potentiator blocked both migration and invasion. We further established UV‐induced SOCE suppression was mediated by induction of CHL, an established inhibitor of endogenous Orai1 channel activity (Derler et al, 2016). Finally, we established that UV‐induced SOCE suppression drives a phenotypic shift towards invasive behavior due to a combination of transcriptional and post‐transcriptional changes.

This study demonstrates that suppression of Ca²⁺ signals through UV exposure or inhibition of Orai1‐STIM signaling drives invasion and metastasis; however, numerous prior investigations have observed that increased Ca²⁺ signals drive invasive behavior (Umemura et al, 2014; Bong & Monteith, 2018; Gross et al, 2020). While the conclusions of these investigations would seem to be contradictory, it is likely a reflection of different contexts. Importantly, using both pharmacological and genetic approaches, we established that the anti‐invasive properties of Ca²⁺ are limited to a very narrow range; if SOCE is abrogated rather than attenuated, invasion did not occur. The major pathway identified in other works as responsible for pro‐invasive Ca²⁺ functions is invadopodia formation (Leslie, 2014; Sun et al, 2015; Pourfarhangi et al, 2018). In the current study, we show that SOCE suppression drives changes in the expression of a large number of genes associated with EMT, cell survival, contact inhibition and metabolism. We further establish that post‐translational O‐GlcNAcylation drives invasive behavior in multiple cell lines and surgically collected patient explants. Given that Ca²⁺‐dependent signaling is highly dependent upon concentration, context and intracellular location, these data are consistent with a model in which Ca²⁺ signals can be either pro‐ or anti‐invasive depending on cell type, degree of change and the signaling environment.

RNA‐seq analysis of UV‐ and SOCE suppression sensitive genes revealed Aryl Hydrocarbon Signaling, Glucose Metabolism Disorder and Impaired Glucose Metabolism as 3 of the top hits (Table EV3). It is interesting to note that these pathways are associated with altered insulin sensitivity leading to increased glucose uptake (Wang et al, 2011; Natividad et al, 2018), which ultimately does occur in these cells. Further, within glucose metabolism disorder pathways, many genes were altered including those associated with glucose uptake, insulin sensitivity, glucose homeostasis, GLUT4 translocation and regulation of pyruvate dehydrogenase (Table EV3). We were also intrigued by the observation that seven genes associated with EMT were upregulated, since the activity of some EMT genes (e.g., Snai1, c‐myc, YAP, and NF‐κB) is driven by O‐GlcNAcylation (Akella et al, 2019). Further, a shift to HBP could drive other pro‐EMT post‐translational modifications such as TGFβ glycosylation (Akella et al, 2019), although this was not demonstrated in this study. Considered collectively, these observations demonstrate that SOCE suppression drives a shift to EMT through coordinated changes in both transcriptional and post‐translational events.

In considering the lack of effect of UV exposure on SOCE in patient melanoma samples, it is important to recognize that the series of human cell lines exhibiting UV‐induced SOCE suppression in Fig 1A were all non‐invasive. This is because we established that invasive melanoma cells exhibit lower SOCE in a prior study (Hooper et al, 2015); the goal of this investigation was to examine the relationship between SOCE suppression and invasiveness within the same cell lines. Since a selection process like this was not possible with freshly isolated surgical explants, the likely explanation for the lack of effect of UV on SOCE is that the majority of the patients that donated melanoma samples for this study were afflicted with an invasive form of this disease.

The connection between UV and CHL production established here is also of clinical relevance. The introduction and wide‐spread use of statins for the control of clinical CHL levels has provided an opportunity to assess the relationship between CHL and cancer progression. Statins have been shown to significantly decrease melanomagenesis (Lee et al, 2016) and slow progression (Pich et al, 2013), although, they cannot decrease the number or clinical profile of pre‐existing dyplastic nevi (Linden et al, 2014). Further, it was found that melanoma metastasis could be suppressed in mice using therapeutics including LXRb (a CHL sensor/transcription factor) agonists GW3965 or T0901317 (Zhao & Dahlman‐Wright, 2010; Pencheva et al, 2014), and statins (Favero et al, 2010; Tsubaki et al, 2015) which target CHL biosynthetic pathways. In the current investigation, we show that UV exposure leads to upregulation of the mevalonate pathway and increases cellular CHL levels. Since blocking the mevalonate pathway blocked invasion in vitro and metastasis and high CHL in vivo was sufficient to promote metastasis independent of UV, our findings support the concept that CHL serves a critical role as a mediator of UV‐induced melanoma metastasis and reveals a new mechanism through which CHL promotes metastasis; SOCE inhibition. Hence, blockade of the mevalonate pathway completely reversed UV‐induced SOCE inhibition and invasiveness in vitro. Since SOCE inhibition was sufficient to drive invasiveness both in vitro and in vivo, these data reveal SOCE suppression as a new mechanism for CHL‐induced metastasis. While we have not investigated how CHL inhibits SOCE in this study, CHL has previously been shown to bind directly to Orai1 where it interferes with the STIM‐mediated transition to the open state (Derler et al, 2016; Hooper et al, 2016). Interestingly, the effect of CHL on SOCE appears to be unidirectional with CHL serving as an endogenous upstream inhibitor SOCE. Our data are highly consistent with these findings and provides a new pathophysiological context for this observation.

It was notable that the majority of samples tested exhibited a “low SOCE” phenotype, showing no change in SOCE in response to UV exposure. In our prior study (Hooper et al, 2015), we observed that invasive melanoma exhibits a “low SOCE” phenotype, while non‐invasive cells have substantially more SOCE. In the current study, only non‐invasive melanoma cell lines exhibiting “higher SOCE” were exposed to UV and tested, as the focus of the work was to determine if SOCE had a mediatory role in UV‐induced melanoma progression. However, this principle was not applied to obtaining patient surgical explants; here, we examined all samples that were made available to us. Further, since the clinic retains tissue samples for pathological analysis and genetic profiling, patients with relatively small lesions had insufficient material to be shared with us; as such we only received samples from relatively advanced patient samples, likely creating a selection bias towards more advanced melanoma.

Conclusions

Within the current investigation, we have defined a previously unknown relationship between UV, CHL biosynthesis, Ca²⁺ signaling and invasive behavior. While CHL biosynthetic pathways have been proposed to drive melanoma progression previously, the molecular context for why CHL biosynthesis might be dysregulated in melanoma had not previously been defined. Considered collectively, the relationship between Ca²⁺ signals and tumor progression is less linear than previously believed with Ca²⁺ signals serving anti‐invasive role in melanoma progression. It is interesting to note that, unlike most tissues, melanocytes grow within the low extracellular Ca²⁺ concentration of the basal layer of skin. If so, the suppression of Ca²⁺ entry may be an adaptation used by melanoma cells to tolerate the high extracellular Ca²⁺ content of non‐native tissues. However, whether the anti‐invasive properties of Ca²⁺ signals are a property unique to melanoma and/or other skin tumors is not known and may be the topic of future investigations.

Materials and Methods

Cell culture

SKMEL5, UACC1273, SKMEL2, FS13, WM983 B‐Raf‐inhibitor resistant (WM983BR), SKMEL28, WM983, UACC257, B16N, B2905A cells were grown in DMEM with 4.5 g/l glucose supplemented with 10% FBS and 1% Gentamycin (Full DMEM media; Thermo‐Fisher Scientific, Waltham, MA) at 37°C and 5% CO₂. Cells were grown in Falcon T75 flasks until ~70% confluency. Cells in conditions <70% confluency were removed of media, washed with PBS and then administered fresh Full DMEM media and incubated until confluency. Upon confluency, cells were removed of media, washed with PBS and treated with 0.25% Trypsin for 10 min at 37°C and 5% CO₂. Cells were then split 1:10 and allowed to incubate in 37°C and 5% CO₂ and while cells were checked daily, they typically grew for 3 days before reaching ~70% confluency.

UV exposure

Cells were plated on glass coverslips for 24 h before changing to PBS. Lids were replaced with a thin layer of taut plastic wrap before irradiating with a broadband UV spectrum containing 35% UVA and 65% UVB (with peak emission at 313 nm in the UVB range; De Fabo et al, 2004). Cells received a total UV dose of 175 J/M² over 90 s at a rate of 1.94 W/m². Approximately 80% of the melanoma cells survive this exposure. Cells were passaged 1:10 every 3 days and cultured under growth optimal conditions for the indicated time period (see Cell culture subsection) before subsequent experimentation.

Generating stable Orai1‐WT‐CFP and Orai1‐E106A‐CFP cell lines

B16N and WM983 cells were transfected with either pIRES‐Neo Orai1‐WT‐CFP or pIRES‐Neo Orai1‐E106A‐CFP vector via electroporation using Gene Pulser Xcell (Biorad, Hercules, CA). Cells were cultured for 1 week under growth optimal conditions before selection with G418 for 2 weeks. Cell sorting was used to isolate the top 5% of the CFP expressing cells CFP before growth and cloning. Clones were then expanded and screened for CFP fluorescence intensity and SOCE on Leica DMI 6000B fluorescence microscope controlled by Slidebook software.

siRNA transfection

5 × 10⁵ cells were plated in six‐well plates overnight prior to treatment. Cells were transfected with 25 pmol of scRNA kit (Invitrogen, Waltham, MA) or siOGT (Sense: 5′‐GCAGUUCGCUUGUAUCGUAtt‐3′; Antisense: 5′‐UACGAUACAAGCGAACUGct‐3′; Invitrogen) using a Lipofectamine‐based RNAimax kit (Invitrogen) according to kit protocol (DOI: 10.1016/j.cell.2010.05.017).

SOCE measurement

Cells grown on glass coverslips were incubated in a cation‐safe buffer (107 mM NaCl, 7.2 mM KCl, 1.2 mM MgCl₂, 11.5 mM Glucose, 20 mM HEPES‐NaOH, 1 mM CaCl₂, pH 7.2) and loaded with Fura2‐acetoxymethylester (Fura2‐AM; 2 μM) for 30 min at 24°C as previously described (Hooper et al, 2015; Go et al, 2019). Cells were washed and allowed to de‐esterify for a minimum of 30 min at 24°C. Following de‐esterification, cells were treated with 2 μM Thapsigargin (SERCA inhibitor) for 10 min prior to imaging. Ca²⁺ measurements were taken using a Leica DMI 6000B fluorescence microscope controlled by Slidebook software (Intelligent Imaging Innovations, Denver, CO). Fluorescence emission at 505 nm was monitored in response to excitation at alternating 340 and 380 nm wavelengths at a frequency of 0.67 Hz; intracellular Ca²⁺ measurements are shown at 340/380 nm ratios obtained from groups (2–45 for patient samples, 35–45 for established cell lines) of single cells.

RT‐qPCR

Cells were grown to 70% confluency and collected as cell pellets prior to RNA extraction using a kit (Ambion, Austin, TX). Following RNA extraction, 500 ng RNA was converted to cDNA using ezDNAse and Superscript IV Reverse Transcriptase Assay Kit (Invitrogen) and diluted 1:10 for working concentration of 50 ng per reaction. 500 nM housekeeping 18S ribosomal RNA (18S) and genes of interest were used for each reaction mixed with Powerhouse SYBR green (Applied Biosystems, Bedford, MA) and analyzed using QuantStudio8 software (Applied Biosystems).

Proliferation assays

500 cells/well were plated on a 96‐well plate for the indicated time point and treated with 1:10 WST‐1 for 1 h at 37°C, 5% CO₂. Absorbance was measured using GloMax plate reader at 450 nm.

Migration assay

Cells were plated on glass coverslips (Ibidi, Fitchburg, WI) overnight to form a complete monolayer before “wounding” with a p20 pipette tip. Coverslips were placed in an incubation chamber (37°C; 5% CO₂) and monitored by confocal microscopy (Leica SP8 Laser Scanning Microscope). Z‐stack brightfield images were obtained every 20 min for 20 h; the % of the wound that was refilled over time was calculated using ImageJ.

Transwell migration assays

Transwell migration assays were performed as previously described (Hooper et al, 2015). Briefly, Matrigel transwell chambers (Corning, Bedford, MA) were hydrated with serum‐free Opti‐MEM for 2 h (37°C, 5% CO₂) prior to the addition of 2,500 cells/well. Growth optimal media was placed on the bottom of the chamber prior to incubation (20 h; 37°C, 5% CO₂). After incubation, media was removed and edges were cleaned before methanol fixation and staining with Crystal Violet (0.5%, 2 min). Images were taken by EVOS cell imager and quantified on ImageJ.

Viability assays

500,000 cells were plated in 24‐well plates overnight in the absence or presence of 2 μM IA65. Cells were then trypsinized with 0.25% Trypsin–EDTA for 10 min at 37°C. Following, cells were treated with 0.4% Trypan Blue and placed in Countess II FL slidereader (Invitrogen) and comparing the number of stained versus unstained cells.

In vivo metastasis assay

Experiments were performed in C57Bl/6 mice (Jax) housed in standard cages (≤5/cage) at LKSOM in accordance with Temple University guidelines (Animal welfare # A3594‐01). B16N cells were counted using a hemocytometer and 1 × 10⁶ cells in 200 μl sterile PBS were injected subcutaneously on the right flank of syngeneic C57BL6 mice. Mice were monitored for 14 days and then sacrificed followed by full body dissection and histopathological analysis. Tissues were fixed in 10% Neutral‐Buffered Formalin (pH 6.8–7.2) followed by H&E staining and Aperio Imagescope software (Leica Biosystems, Buffalo Grove, IL) scan for peritoneal wall invasion.

RNA sequencing analysis

Raw RNA‐Seq data were processed using the subread algorithm established in (Liao et al, 2014). Preprocessed data files from the samples were subject to RNA sequencing data analysis using Rsubread package (Liao et al, 2019). Genome indices were built using the buildindex function in Rsubread package. RNA transcripts from each sample were mapped to the mouse genome reference consortium build 38 (GRCm38) genome. Alignment of the sample reads to the reference genome was performed using the align function within the Rsubread package. featureCounts function within the Rsubread package was used to summarize the data to integer‐based, gene‐level read counts. Read counts generated from the pipeline were annotated using mm10 annotation. Differential expression analysis of the read counts was performed using the DESeq2 package (Anders & Huber, 2010). Differentially expressed genes generated from the DESeq2 analysis were subjected to pathway analysis using Ingenuity Pathway Analysis tool (Ingenuity® Inc., Redwood city, CA) with default settings.

Cholesterol measurement

WM983 and UACC257 human melanoma cells were grown to 50% confluency before transfecting with PFO D4H‐mcherry via electroporation using Gene Pulser Xcell (Biorad). Cells were then plated on glass coverslips under growth optimal conditions before capturing images by confocal microscopy (Leica SP8 Laser Scanning Microscope). Fluorescence intensity was measured using Leica LASX software.

Glucose uptake assay

WM983 and UACC257 cells treated as described were cultured under growth optimal conditions parallel to unmanipulated control cells. Cells were seeded at density of 15,000 cells/well in 96‐well plates. Glucose uptake was measured using the Promega Glucose Uptake‐Glo Assay (Madison, WI); luminescence was recorded using 0.3–1 s integration using a Tecan INFINITE M1000 PRO monochromator‐based microplate reader.

Seahorse assays

Oxygen Consumption Rate (OCR) and Extracellular Acidification rate (ECAR) were measured using an Agilent Seahorse XF96 (Wilmington, DE). B16N, WM983 and UACC257 cells treated as described were cultured under growth optimal conditions parallel to unmanipulated control cells. 15,000 cells/well were seeded in Seahorse 96‐well cell culture plates for all experiments. Basal OCR was measured in media supplemented with 1 mM pyruvate, 2 mM glutamine, and 10 mM glucose followed by the sequential addition of Oligomycin (1.5 μM), FCCP (1 μM) and Rotenone/Antimycin‐A (1 μM) to perform the mitochondrial stress test. Basal ECAR was measured in media supplemented with 2 mM glutamine followed by the sequential addition of glucose (10 mM), oligomycin (1 μM) and 2‐Deoxy‐d‐glucose (2‐DG: 50 mM) to perform the glycolysis stress test. Results were quantified using Wave software (Agilent Technologies, Santa Clara, CA).

Freshly isolated patient surgical explants

Patient surgical explants were received from Dr. Jeffrey Farma at Fox Chase Cancer Center with all personal identification information being protected by HIPAA privacy. All patients had given informed consent to have their tissue used for research in accordance to protocols from Fox Chase Cancer Center.

Freshly isolated patient explants were incubated in DMEM containing Liberase TL (Roche Applied Science; 37°C, 30 min). Following incubation, tissue was cut using a sharp razor and dissociated through 50 μM Medicon screens (Becton Dickinson Biosciences, San Jose, CA; 2 min). Tissue samples were resuspended in DFD media (DMEM, 20% FBS; 500 μg/ml DNAse I, Sigma, St. Louis, MO) and passed through an 18‐gauge syringe to break up clumps of cells before filtration (70 μm, Corning, Glendale, AZ), centrifugation and resuspension in DMEM (2% FBS; 1% Gentamycin) for 3 days before use.

Primer sequences

18S:	Forward: CTT AGA GGG ACA AGT GGC G
	Reverse: ACG CTG AGC CAG TCA GTG TA
hSTIM1:	Forward: CAC TCT TTG GCA CCT TCC ACG T
	Reverse: CTG TCA CCT CGC TCA GTG CTT G
hSTIM2:	Forward: CAG TCT TTG GGA CTC TGC ACG T
	Reverse: GCC AGC GAA AAA GTC GTT CTC G
mSTIM1:	Forward: TGG ATG AGG AGA TTG TGT CGC C
	Reverse: GAC TCC GAA TCG GAA TGG GTC A
mSTIM2:	Forward: GCC AGT ATG CAG AGC AGG AAC T
	Reverse: GCT GAA GCC ATT TCT GTA GTG CG
hOrai1:	Forward: AGG TGA GCC TCA ACG AGC A
	Reverse: AGT CGT GGT CAG CGT CCA GCT
hOrai2:	Forward: CCT GTC GTG GCG GAA GCT CTA
	Reverse: ACT GGT ACT GCG TCT CCA GCT G
hOrai3:	Forward: TTA GCA AGG GTT GGG TAA GG
	Reverse: TTT CCA GGG CTA AGG ACT GG
mOrai1:	Forward: GTT ACT CCG AGG TGA GCC T
	Reverse: AGC TGG ACT TCC ACC ATC GCT A
mOrai2:	Forward: GAC ACA GAC GCT AGC CAC GA
	Reverse: ATG GGC ACA TTG AGC TCT GC
mOrai3:	Forward: TTA CCA CAT CAC AAC AGC CT
	Reverse: TGG TCC ATG AGC ACT ATC AC
hOGT:	Forward: CAG GAA GGC TAT TGC TGA GAG G
	Reverse: CGG AAC TCA CAT ATC CTA CAC GC

Statement about mouse sample size

Previous studies have done a minimum of five to six mice per condition. However, doing a dichotomous endpoint, two independent sample study with power of 80% and α = 0.05 would require 10 mice per condition which is the amount we have used for each condition in this study.

Statement about data inclusion

Any excluded data were based on human error that met pre‐established criteria including correct handling of reagents, correct administration of treatments to animals, and proper rigor and reproducibility of results.

Statement about statistical tests

All statistical analyses were done using suggested analyses by Graphpad Prism software.

Statement about blinding

No blinding was done in this study.

Statement about randomization

All mice were randomly selected from their pre‐determined conditions for experiments.

Ethics approval and consent to participate

Human samples were deidentified and were IRB exempt. All work performed in mice was IACUC approved.

Consent for publication

Not applicable.

Materials

All cell lines were acquired from American Type Culture Collection (ATCC; Manassas, VA). All Patient Melanoma Surgical Explants were acquired under IRB protocols from Fox Chase Cancer Center Biosample Repository. Media was purchased from Corning and Gibco (Waltham, MA). TAK475 and DNAse I were purchased from Sigma‐Aldrich (St. Louis). Compactin and BTP2 were purchased from Tocris Bio‐Techne Corporation (Minneapolis, MN). OSMI1 was purchased from Millipore Sigma (Burlington, MA). Matrigel Transwell Chambers were purchased from Corning (Waltham, MA). Liberase TL was purchased from Roche Applied Science (Indianapolis, IN). 50 μM Medicons and the medimachine were purchased from Becton Dickinson Biosciences. 8‐well u‐slides migration plates were purchased from Ibidi. Fura2‐AM was purchased from Thermo‐Scientific (Waltham, MA). Glucose Uptake Kit was purchased from Promega (Madison, WI), Lactate Production Kit was purchased from Sigma‐Aldrich (St. Louis). High fat mouse chow was purchased from Harlan labs (Frederick, MD). Orai1‐WT‐CFP and Orai1‐E106A‐CFP were generated as previously described (Zhou et al, 2009); the mCherry‐PFO‐D4 gene was synthesized by Integrated DNA Technologies (Coralville, IA) and combined via PCR into an mCherry‐C1 backbone (Addgene). PFO‐D4H was constructed via site‐directed mutagenesis. Scrambled RNA was purchased from Ambion (Silencer Negative Control No. 1 siRNA), RNAimax kit was purchased from Thermo‐Scientific (Waltham, MA). OGT‐targeting siRNA was purchased from Ambion.

Author contributions

Scott Gross: Formal analysis; investigation; methodology; writing – original draft; writing – review and editing. Robert Hooper: Conceptualization; formal analysis; investigation. Dhanendra Tomar: Formal analysis; investigation; methodology. Alexander P Armstead: Formal analysis; investigation. Pranava Mallu: Formal analysis; investigation; methodology. No'ad Shanas: Formal analysis; investigation; methodology. Hinal Joshi: Formal analysis; supervision; investigation; methodology. Suravi Ray: Formal analysis; supervision; investigation; methodology. Parkson Lee‐Gau Chong: Resources; supervision; investigation; methodology. Igor Astsaturov: Data curation; software; formal analysis; supervision; investigation; methodology. Jeffrey M Farma: Resources; formal analysis; supervision; investigation; writing – review and editing. Kathy Q Cai: Formal analysis; investigation; methodology. Kumaraswamy Naidu Chitrala: Conceptualization; data curation; software; formal analysis; supervision; funding acquisition; investigation; writing – original draft; project administration; writing – review and editing. John W Elrod: Formal analysis; supervision; investigation; methodology; writing – review and editing. M Raza Zaidi: Conceptualization; formal analysis; investigation; methodology. Jonathan Soboloff: Conceptualization; formal analysis; supervision; funding acquisition; investigation; methodology; writing – original draft; project administration; writing – review and editing.

In addition to the CRediT author contributions listed above, the contributions in detail are:

SG performed and designed experiments, analyzed data and contributed to the writing of the manuscript. RH and DT performed and designed experiments, analyzed data. APA, NS, SR, PM, and HJ contributed to some experiments. PL‐GC and IA contributed to experimental design. JMF provided clinical samples. JWE, KQC, and MRZ designed experiments and contributed to the writing of the manuscript. JS designed experiments, analyzed data and wrote the manuscript.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Table EV1

Table EV2

Table EV3

PDF+

The EMBO Journal (2022) 41: e110046.

Data availability

mRNA sequencing data are available in Gene Expression Omnibus (GEO) GSE205830. All source data are accessible through the following link https://figshare.com/projects/Supression_of_Ca2_Signaling_Enhances_Melanoma_Progression/141035.