Osteoblasts contribute to a protective niche that supports melanoma cell proliferation and survival

Abstract Melanoma is the deadliest form of skin cancer; a primary driver of this high level of morbidity is the propensity of melanoma cells to metastasize. When malignant tumours develop distant metastatic lesions the new local tissue niche is known to impact on the biology of the cancer cells. However, little is known about how different metastatic tissue sites impact on frontline targeted therapies. Intriguingly, melanoma bone lesions have significantly lower response to BRAF or MEK inhibitor therapies. Here, we have investigated how the cellular niche of the bone can support melanoma cells by stimulating growth and survival via paracrine signalling between osteoblasts and cancer cells. Melanoma cells can enhance the differentiation of osteoblasts leading to increased production of secreted ligands, including RANKL. Differentiated osteoblasts in turn can support melanoma cell proliferation and survival via the secretion of RANKL that elevates the levels of the transcription factor MITF, even in the presence of BRAF inhibitor. By blocking RANKL signalling, either via neutralizing antibodies, genetic alterations or the RANKL receptor inhibitor SPD304, the survival advantage provided by osteoblasts could be overcome.


| INTRODUC TI ON
Melanoma is the deadliest form of skin cancer, and arises from the transformation of pigment cells in the skin called melanocytes. Poor prognostic outcomes in melanoma are associated with resistance to conventional therapy and the highly metastatic nature of melanoma (Sandru, Voinea, Panaitescu, & Blidaru, 2014), whereby multiple tissue-specific niches are capable of being populated by this transformed cell lineage. These niches provide unique signalling environments that are co-opted by the melanoma to promote growth and survival. Therefore, when studying the interface between therapies and basic melanoma biology, it is important to understand the cellular and tissue context of cancer cells. This is exemplified by the Paget-Seed-Soil theorem that states the tissue site (soil) that a cancer (seed) preferentially metastasizes to is pre-defined by mutualistic factors (Fidler, 2003).
One secondary site that appears to provide strong support for cancer cells is the bone (Bubendorf et al., 2000;Rahim et al., 2014). For instance, up to 90% of prostate cancers that metastasize will spread to the bone; an event that predicates poor survival independent of therapeutic intervention (Bubendorf et al., 2000). The stromal makeup of the bone is rich in endothelial and immune cells, though the deposition of extracellular matrix (ECM) is chiefly controlled by two cell lineages, osteoblasts and osteoclasts (Park, Keller, & Shiozawa, 2018). Osteoclasts breakdown and remodel the mineralized bone matrix, while osteoblasts function to produce new ECM in addition to controlling the number and activation of osteoclasts. The factors that drive the homing and aggressive growth of bone metastases are well appreciated, and include abundant supply of blood and oxygen, and a wealth of secreted factors that emanate from bone stroma, including IGFs, BMPs, FGFs, PDGF and RANKL (Park et al., 2018;Rahim et al., 2014;Sottnik & Keller, 2013;Zheng, Li, & Kang, 2016). RANKL is of particular interest because it stimulates osteoclastogenesis, which in turn drives the loss of bone matrix; an event often observed in patients with bone metastases (Hegemann, Bedke, Stenzl, & Todenhöfer, 2017). Indeed, the RANKL inhibitor Denosumab both decreases the rate of bone degeneration by inhibiting osteoclastogenesis, and diminishes the spread of prostate cancer (Hegemann et al., 2017). RANKL has also been shown to play a role in the ability of melanoma cells to home to, and grow within the bone niche (Jones et al., 2006).
Melanoma cell growth is typically driven by the ERK/MAPK pathway, which is hyper-activated in at least 80% of melanomas that occur through mutations in NRAS (~20%), BRAF (~50%) and NF1 (~14%) (Akbani et al., 2017). Therefore, monotherapies using a BRAF inhibitor (BRAFi) or combination therapies of BRAF and MEK inhibitors (MAPKi) are now considered a mainstay of melanoma treatment (Long et al., 2015). However, maintaining initial responses are problematic due to the development of resistance driven by a plethora of mechanisms (Arozarena & Wellbrock, 2017;. We have shown previously that the master regulator of survival, growth and differentiation in pigment cells, MITF, contributes to resistance by increasing tolerance to MAPKi during initial treatment (Smith et al., , 2017. This occurs in concert with alterations in surrounding tumour stroma that further decreases response to therapy (Smith et al., 2014;Wang et al., 2015;Young et al., 2017), and involves fibroblasts, macrophages and even the ECM (Hirata et al., 2015;Qin et al., 2016;Straussman et al., 2012). The variable composition of the stroma between potential metastatic sites suggests the possibility of differential responses to therapy. Indeed, melanomas located either in bone lesions or the Central Nervous System (CNS) have worse response rates to MAPKi therapy (16%) compared to all other sites (>70%) (Seifert et al., 2016). Additionally, mutations that drive resistance within a relapsed patient differ between metastatic sites (Kemper et al., 2015).
While secreted factors found in the cerebrospinal fluid are known to contribute to the CNS-induced therapy resistance of melanomas (Seifert et al., 2016), the contribution of the bone-specific stromal niche to resistance to targeted therapies is unknown. Thus, we examined signalling between melanoma and osteoblasts, and the role of this interplay in MAPKi resistance.  . Conditioned medium (CM) was generated by incubating cells for 24 hr with fresh culture medium containing FCS was then filtering (0.45 µm) to remove cells and debris.

| Osteoblast differentiation and co-culture
Osteoblast precursor cells hFOB 1.19 were acquired from ATCC (CRL-11372). hFOB 1.19 cells were cultured at 34°C in HAMs F12 medium and DMEM/10% FCS (PAA, Yeovil, UK) at a ratio of 1:1 in a humidified 5% CO2 incubator. Differentiation was performed by transferring cells to 39°C in a humidified 5% CO2 incubator and supplementing media with either filtered CM from melanoma cells or spiked with recombinant PTH. For co-culture assays, hFOB 1.19 cells were differentiated in transwell inserts (BD Biosciences) and washed 3x with DMEM before they were incubated with melanoma cells. For direct co-culture experiments individual cultures of 0.2 × 105 osteoblasts and 0.5 × 105 A375 cells, respectively were stained and quantified and compared to a co-culture of 0.2 × 105 osteoblasts and 0.5 × 105 A375 cells.

| RNA isolation and RT-qPCR analysis
RNA from cell lines was isolated with TRIZOL® and selected genes were amplified by quantitative real time PCR (RT-qPCR) using

Significance
Understanding how specific tissue niches leads to resistance to melanoma therapies is essential to generating robust and sustainable patient responses. By looking at the signalling between melanoma cells and osteoblasts we have characterized a mechanism with the potential to provide resistance to BRAF inhibitors that is specific to the bone niche. The output of this investigation has identified a RANKL-MITF signalling axis that can be targeted to antagonize osteoblast contributions towards resistance.
SYBR green (Qiagen, Valencia, CA, USA). Relative expression was calculated using the delta-delta CT methodology and beta-actin was used as reference housekeeping gene (Livak & Schmittgen, 2001).

| IncuCyte caspase analysis
To assess apoptosis induction, 5,000 cells per well were seeded in a black 96 well tissue culture plate (BD Falcon, SLS). IncuCyte Kinetic Caspase-3/7 Apoptosis Assay Reagent (Essen BioScience) was added at a dilution of 1/10,000. If caspase 3/7 had been activated apoptotic cells could be detected by a fluorescence signal. Cells were kept at 37°C in a humidified 5% CO2 incubator and imaged using an IncuCyte ZOOM (Essen BioScience) and a 20 × lens.

| Cell lysis and immunoblotting
Cells were lysed in SDS sample buffer and analysed by standard

| Statistical analysis
If not indicated otherwise, data represent the results for assays performed in triplicate, with error bars to represent SEM. Statistics used were: predominately Student t test and one-way ANOVA with Tukeys's post hoc test performed using GraphPad Prism version 7.00 for Mac OS, GraphPad Software (San Diego California USA).

| Differentiated osteoblasts enhance melanoma growth and differentiation
To investigate whether a niche of bone stromal cells interacts with melanoma cells we utilized a model of differentiating osteoblast precursor cells hFOB 1.19 as previously described (Harris, Enger, Riggs, & Spelsberg, 2009). Transfer of hFOB 1.19 cells to 39°C and subsequent treatment with PTH over 7 days led to increased expression of differentiated osteoblast markers such as PTHrP, SPP1 and RANKL ( Figure 1a). The pro-growth effect in response to osteoblast-secreted factors was linked to enhanced expression of genes related to cell-cycle progression and pigment cell differentiation ( Figure 1d, Figure S1c).
Many of these genes are MITF targets and accordingly, MITF expression was increased ( Figure 1d). When the protein levels of MITF were analysed, only CM from differentiated osteoblasts was able to enhance MITF expression; CM from precursor hFOB 1.19 cells failed to elicit any change in MITF protein expression ( Figure 1e).
Together, these data suggest that MITF may be orchestrating the effects on melanoma cell differentiation and proliferation induced by osteoblasts.
Osteoblast cells are known to secrete many pro-inflammatory cytokines that activate the NFκB family, and activation of RelA/ p65 has previously been linked to MITF expression downstream of TNFα (Chu et al., 2014;Smith et al., 2014). Indeed, we observed that osteoblast-CM induced phosphorylation of p65 in melanoma cells within hours and this correlated with increased MITF protein levels ( Figure 1f).

| Melanoma cells can differentiate osteoblasts and enhance RANKL expression
We have recently shown that melanoma cells can promote the differentiation of macrophages (Young et al., 2017), as well as stimulate fibroblasts to alter the stiffness of their deposited ECM (Miskolczi et al., 2018). Similarly, we tested whether melanoma cells directly impact osteoblast differentiation by exposing the latter to melanoma cell CM. As observed for PTH-differenti- So far we observed that CM from differentiated osteoblasts not only induces MITF expression, but also activates NFκB signalling in melanoma cells (see Figure 1f). A cell-type-restricted NFκB activator expressed at high levels in differentiated osteoblasts is RANKL (Leibbrandt & Penninger, 2008). Indeed, we found that RANKL expression was high in in vitro differentiated osteoblasts, but not significantly high in melanoma cells ( Figure 2c). Moreover, we found that RANKL expression was up-regulated at both RNA and protein level, in the melanoma CM-differentiated  (Figure 2d,e), suggesting that it could contribute to the induction of MITF expression in melanoma cells (see Figure 2b).
In order to respond to RANKL, melanoma cells need to express its receptor RANK. Indeed, RANK expression is significantly increased in transformed melanoma cell lines when compared to normal human melanocytes (Figure 2f, Figure S2a). Strikingly, higher expression of RANK in melanoma biopsies correlated with a significant decrease in overall survival, highlighting the potential importance of this signalling node (Figure 2g).

| RANKL enhances proliferation and differentiation via MITF induction
Our data suggested that although melanoma cells may not secrete To further characterize the RANKL-RANK-MITF signalling axis, we first used RNAi to deplete MITF in melanoma cells prior to RANKL stimulation (Figure 3e). EdU incorporation assays demonstrated that MITF
Co-treatment of RANKL and BRAFi had no effect on the loss of activated ERK compared to BRAFi alone, indicating that the increase in cell number was not mediated via re-activation of the MAPK pathway ( Figure S4a); a mechanism reported for other secreted factors such as HGF and EDN1 (Smith et al., 2017;Straussman et al., 2012).   (Figure 4c,d), which is lost when MITF is depleted (Figure 4e).

Co-treatment of MITF negative melanoma cells with RANKL and
BRAFi conferred no survival advantage ( Figure S4c). Altogether, these data support an MITF-dependent role for the RANKL-mediated increase in melanoma survival upon BRAF inhibition.

| Osteoblasts antagonize BRAF inhibition via RANKL secretion
Complex stromal and tumour cell interactions have previously been shown to be potent drivers of resistance during MAPK inhibitor therapies through the secretion of different growth factors such as TNFα, HGF and others (Hirata et al., 2015;Lito et al., 2012;Smith et al., 2014;Straussman et al., 2012). We had observed that CM from melanoma cells induced RANKL expression in differentiating  (Figure 2c-d), and so we next asked whether secreted RANKL from osteoblasts is sufficient to provide melanoma cells with a survival advantage.

BRAFi-treated melanoma cells co-cultured with differentiated
osteoblasts increased in cell number when compared to those cocultured with undifferentiated hFOB1.19 cells; an effect that was abolished by addition of a neutralizing antibody (nAb) to RANKL ( Figure 5a, Figure S5a). A concomitant increase in MITF expression was observed in the osteoblast melanoma co-cultures that was attenuated by addition of a RANKL nAb (Figure 5a, Figure S5a). An identical effect was observed when the RANK inhibitor SPD304, that blocks the trimerization of RANK and TNFα receptors (Douni et al., 2012), was substituted for the nAb (Figure 5b).
We next examined whether the increase in RANKL-mediated survival occurred through inhibition of apoptosis. Indeed, co-cul-

| D ISCUSS I ON
Characterization of the complex mechanisms that govern metastatic growth in distant organs has been a goal for decades, ever since the "seed and soil" hypothesis was first postulated in 1889. In this study, we examined the role of the bone in fostering the growth and survival of melanoma. Bone stroma is composed of a network of cells working in concert to balance matrix deposition with destruction; chief among these cells are the osteoblasts. The relationship we have uncovered between osteoblasts and melanoma cells is one of mutuality; melanoma cells increase their proliferation when co-cultured with osteoblasts, and osteoblasts show enhanced differentiation in the presence of a majority of melanoma cell lines. A similar relationship has been described for prostate and breast cancer cells that exhibit a predilection for forming bone metastases (Rahim et al., 2014).

Seifert et al have observed variable efficacy of melanoma re-
sponses to MAPK inhibitor therapy between metastatic sites, suggesting fundamental signalling differences between these sites (Seifert et al., 2016). Another possible explanation for this observation is that different sites for metastasis have different levels of drug bioavailability. However, preclinical in vivo studies of the BRAF-V600E inhibitor vemurafenib find that drug accumulation occurs in both the liver and bones; niches that predict poor response in patients. As such, it is unlikely that bioavailability contributes to the poor response rates observed in melanoma bone metastases (Rissmann, Hessel, & Cohen, 2015). Thus, our finding that paracrine signalling via RANKL antagonizes BRAF inhibition is likely to have clinical significance.
RANKL does not drive tolerance to MAPK inhibition through pathway re-activation as has commonly been observed (Lito et al., 2012;Smith et al., 2017;Straussman et al., 2012), but rather via an increase in MITF expression similar to the TNF-α mediated increase in MITF that occurs when macrophages are co-cultured with melanoma cells (Smith et al., 2014). It is now well established that elevated MITF expression mediates resistance to MAPK inhibitor based therapies (Haq, Shoag, et al., 2013;Johannessen et al., 2013;Rose et al., 2016;Smith et al., , 2013Smith et al., , 2017. As such, RANKL signalling may be an "Achilles heel" of melanoma bone metastases that can be exploited therapeutically. A combinatorial approach using MAPK inhibition with the RANKL inhibitor Denosumab could, therefore, be employed specifically in patients displaying bone lesions. The importance of RANKL/RANK signalling for bone niche maintenance, and previous reports demonstrating that melanoma migration and metastasis is driven by RANKL (Jones et al., 2006;Peinado et al., 2012), identifies RANK signalling as a candidate pathway worthy of further investigation in melanoma biology. We find that RANKL accelerates cell-cycle progression, leading to enhanced proliferation and growth of melanoma cells. RANKL stimulation also leads to MITF up-regulation with a subsequent enhancement of differentiation and survival likely due to activation of a MITF transcriptional program known to mediate these outcomes (Garraway et al., 2005;Johannessen et al., 2013;Miskolczi et al., 2018;Wellbrock & Arozarena, 2015). We found that in the TCGA melanoma cohort, inhibitors becoming increasingly common in the clinic it is unsurprising that anti-RANKL therapies have been proposed for melanoma, both alone and in combination with anti-CTLA-4 therapies (Ahern et al., 2017;Smyth, Yagita, & McArthur, 2016). Anti-RANKL therapies may synergize not only with MAPKi, but also therapies that activate the adaptive immune system; the two standard of care approaches for treatment of disseminated melanoma.
How MAPK inhibition alters the bone microenvironment is poorly understood, and may contribute to the often-poor drug responses observed in patients with bone metastasis. Indeed, stromal fibroblasts in BRAF inhibitor treated tumours alter the ECM and produce a more regressive micro-environment for melanoma cells (Hirata et al., 2015). Further, MAPK inhibition may also alter osteoblast function. Although we have not addressed the role of MAPK signalling in osteoblasts specifically, we do find that co-culture of melanoma cells with differentiated osteoblasts results in sufficient production of RANKL to antagonize MAPK inhibition. and PTHrP are expressed in some melanoma cell lines, this may be a conserved mechanism worthy of further investigation (Sottnik & Keller, 2013).
In conclusion, we have demonstrated that the RANKL-RANK signalling that occurs between osteoblasts and melanoma cells drives the proliferation, differentiation and survival of melanoma cells.
Osteoblast-derived RANKL stimulates MITF-driven tolerance to MAPK inhibition, which may contribute to the increased resistance to targeted therapies observed in melanoma patients with metastatic bone lesions.

ACK N OWLED G EM ENTS
We thank Stephen Taylor University of Manchester (UoM) for the use of equipment and resources. We also thank Adam Hurlstone

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.