17-AAG inhibits vemurafenib-associated MAP kinase activation and is synergistic with cellular immunotherapy in a murine melanoma model

Heat shock protein 90 (HSP90) is a molecular chaperone which stabilizes client proteins with important roles in tumor growth. 17-allylamino-17-demethoxygeldanamycin (17-AAG), an inhibitor of HSP90 ATPase activity, occupies the ATP binding site of HSP90 causing a conformational change which destabilizes client proteins and directs them towards proteosomal degradation. Malignant melanomas have active RAF-MEK-ERK signaling which can occur either through an activating mutation in BRAF (BRAFV600E) or through activation of signal transduction upstream of BRAF. Prior work showed that 17-AAG inhibits cell growth in BRAFV600E and BRAF wildtype (BRAFWT) melanomas, although there were conflicting reports about the dependence of BRAFV600E and BRAFWT upon HSP90 activity for stability. Here, we demonstrate that BRAFWT and CRAF are bound by HSP90 in BRAFWT, NRAS mutant melanoma cells. HSP90 inhibition by 17-AAG inhibits ERK signaling and cell growth by destabilizing CRAF but not BRAFWT in the majority of NRAS mutant melanoma cells. The highly-selective BRAFV600E inhibitor, PLX4032 (vemurafenib), inhibits ERK signaling and cell growth in mutant BRAF melanoma cells, but paradoxically enhances signaling in cells with wild-type BRAF. In our study, we examined whether 17-AAG could inhibit PLX4032-enhanced ERK signaling in BRAFWT melanoma cells. As expected, PLX4032 alone enhanced ERK signaling in the BRAFWT melanoma cell lines Mel-Juso, SK-Mel-2, and SK-Mel-30, and inhibited signaling and cell growth in BRAFV600E A375 cells. However, HSP90 inhibition by 17-AAG inhibited PLX4032-enhanced ERK signaling and inhibited cell growth by destabilizing CRAF. Surprisingly, 17-AAG also stimulated melanin production in SK-Mel-30 cells and enhanced TYRP1 and DCT expression without stimulating TYR production in all three BRAFWT cell lines studied as well as in B16F10 mouse melanoma cells. In vivo, the combination of 17-AAG and cellular immunotherapy directed against Tyrp1 enhanced the inhibition of tumor growth compared to either therapy alone. Our studies support a role for 17-AAG and HSP90 inhibition in enhancing cellular immunotherapy for melanoma.


Introduction
Heat shock protein 90 (HSP90) is a molecular chaperone which stabilizes the activity of overexpressed and/or mutated signaling proteins commonly found in cancer. The stabilizing properties of HSP90 on client proteins have, in part, been discerned through an analysis of the effects of specific chemical inhibitors of the HSP90 ATPase activity. These include inhibitors of the benzoquinoid ansamycin class, such as geldanamycin (GA) and its semi-synthetic derivatives, 17-allylamino-17-demethoxygeldanamycin (17-AAG) and 17-dimethylaminoethylamino-17-demethoxygeldanamycin . ATP binding and hydrolysis in an Nterminal pocket of HSP90 [1] drives a conformational switch from an open configuration, where the client protein is bound initially in a complex with co-chaperones HSP70, HSP40, HOP, and HIP; to a closed state with Cdc37, p23, and immunophilins. In the closed state, HSP90 interacts with client proteins, such as ERBB2 (HER-2) [2], AKT [3], and CRAF (Raf-1) [4,5] to maintain their conformational stability. Inhibition of the closed state following occupancy of the ATP binding site by GA or one of its analogs causes degradation of the client protein through the proteasomal pathway [6][7][8].
Activation of the Raf family kinases BRAF and CRAF [9][10][11] is important for the pathogenesis of malignant melanoma [12,13]. Activation can occur either through a gain of function somatic mutation in BRAF, generating the BRAF V600E substitution mutation found commonly in melanomas arising on non-chronically sun-exposed skin [9,14,15]. Alternatively, activation of wild-type BRAF (BRAF WT ) and CRAF can occur through activating mutations in NRAS, or KIT, or and through NF1 deletion [16], which occur in approximately 20% of melanomas [14]. These findings have prompted investigations of the efficacy of GA derivatives for inhibiting RAS-and RAF-dependent signaling and reducing the tumorigenicity of melanoma cells. It was reported earlier BRAF V600E stability was reduced by 17-AAG in A375 human melanoma cells, whereas BRAF WT in CHL melanoma cells was less affected. However, CRAF was degraded, and phosphorylation of ERK inhibited, in each of these cell lines. Only BRAF V600E , but not BRAF WT or CRAF, was associated with HSP90 [17]. In a similar study, sensitivity of BRAF to 17-AAG also extended selectively to melanoma cell lines with BRAF V600E mutations. In 4/4 human melanoma cell lines with BRAF WT (SK-Mel-2, SK-Mel-31, SK-Mel-147, and SK-Mel-103), no degradation occurred with 17-AAG concentrations up to 2.5 μM, whereas 17-AAG induced degradation in 5/5 cell lines with BRAF V600E (or BRAF V600D ) mutations (SK-Mel-1, SK-Mel-5, SK-Mel-19, SK-Mel-28, and WM 266.4). Nonetheless, 17-AAG inhibited melanoma cell proliferation regardless of BRAF mutational status [18], perhaps due to these cells' dependence upon CRAF signaling in melanomas with activating NRAS mutations [11] as well as activation of CRAF by BRAF WT under these conditions [19]. These data suggest that BRAF V600E in melanoma is an HSP90 client protein whose degradation induced by 17-AAG is potentially very important for its inhibitory effects upon melanoma cell growth. However, stabilization of disease course noted in a metastatic melanoma patient with an NRAS activating mutation, but BRAF WT , during a phase I clinical trial [20] suggests that the effect of 17-AAG on the NRAS mutant subset of melanomas requires further consideration.
In this report, we demonstrate that melanoma cells that either harbor activating NRAS mutations with BRAF WT or harbor the BRAF V600E mutation with wild-type NRAS, the BRAF is bound by HSP90 and under certain conditions is subject to degradation by HSP90 inhibition with 17-AAG. We confirm that 17-AAG inhibits melanoma cell proliferation with each mutational subtype. We also show that 17-AAG can suppress both cell proliferation and the paradoxical activation of ERK [21,22] in NRAS melanoma cells treated with BRAF inhibitor N-(3-(5-(4-chlorophenyl)-1H-pyrrolo[2;3-b] pyridine-3-carbonyl)-2,4-difluorophenyl) propane-1-sulfonamide (PLX4032 or vemurafenib). Since one mechanism of resistance to the anti-tumor effects of vemurafenib is the mutational or growth factor-induced activation of RAS [23,24], concurrent administration of BRAF inhibitors and 17-AAG might conceivably delay the growth of resistant clones of melanoma cells in patients with BRAF V600E melanomas. Delays in the development of resistance or therapeutics synergy with inhibitors of HSP90 and BRAF V600E have been shown in melanoma with the HSP90 inhibitors XL888 and AT13387 [25][26][27]. Here, in addition, we describe a novel activity of the HSP90 inhibitor 17-AAG which induces melanization in BRAF WT melanoma cells in a tyrosinase-independent manner while increasing expression of accessory melanogenic enzymes and tumor antigens [28,29] TYRP1 and DCT. Correspondingly, we show that 17-AAG and cellular immunotherapy directed against Tyrp1 are synergistic; revealing a new mechanism whereby 17-AAG can potentiate the effects of an established melanoma therapy. These findings extend the range of client proteins of HSP90 susceptible to BRAF degradation and reveal a novel activity of 17-AAG.

Results
We were initially interested in characterizing binding partners of the 85 kDa isoform of BRAF in human melanoma cells detected by different anti-BRAF antibodies. Immunoprecipitation of BRAF from Mel-Juso human melanoma cell lysate resulted in the isolation of a strong~85 kDa band on a Coomassie-stained gel near the area of an immunoreactive Western blot band from an identically-loaded gel lane (Fig 1A). Mass spectroscopic analysis of the excised gel band revealed peptides corresponding to the HSP90-α (HSP 86) and HSP90-β(HSP 84) isoforms of HSP90 (Table 1 and S1 Appendix). This provided initial evidence that BRAF interacts with HSP90 in human melanoma cells.
To confirm this interaction, an enhanced HSP90 signal was observed in an anti-BRAF coimmunoprecipitate in comparison to the signal observed with an immunoprecipitate prepared from a non-specific, isotype-matched antibody ( Fig 1B). We evaluated a set of conditions utilized previously to evaluate interactions between RAF/Raf kinases and HSP90 and its orthologues. Of these, we found that conditions minimizing incubation time with the antibody at room temperature enhanced detection of the interaction ( Fig 1C). As described previously, HSP90 co-immunoprecipated with CRAF in melanoma cell lysate as well. CRAF also coimmunoprecipitates with BRAF (Fig 1D), implying that there is a triad of interactions between BRAF, CRAF, and HSP90 in human melanoma cells. This is in contrast to interactions studied previously in PC12 cells, where BRAF interacts with HSP90 but CRAF does not [30].
Previous work with CRAF has demonstrated that disruption of the interaction between CRAF and HSP90 [31] with the benzoquinone ansamycin geldanamycin leads to destabilization of CRAF [4] and loss of downstream MAP kinase signaling [5]. Similar findings have been described with BRAF V600E in human cells [17,18]. We were interested in determining whether BRAF WT in melanoma cell lines such as Mel-Juso where we have demonstrated an interaction with HSP90 were destabilized by HSP90 inhibition, since this has not been previously reported. In addition to Mel-Juso, we selected 2 additional cell lines with activating mutations in NRAS (SK-Mel-30 and SK-Mel-2) as well as 2 established melanoma cell lines with a BRAF V600E mutation (A375 and SK-Mel-28) for comparison. SK-Mel-2 melanoma cells have been reported to contain either an NRAS Q61R [9,17] or a KRAS Q61R [11] mutation, but   previously by one group [17] but not by another where BRAF WT was stable following incubation with 17-AAG [18]. We examined how 17-AAG, with consequent degradation of BRAF and CRAF, affected signaling downstream from the RAF kinases. Western blots were examined for relative expression of phosphorylated MEK and ERK kinases. Relative MEK and ERK phosphorylations were diminished at increasing concentrations of 17-AAG in BRAF V600E and NRAS mutated human melanoma cell lines. Inhibition occurs even when BRAF WT (Mel-Juso and SK-MEL-30 cells) was not degraded (Fig 2), confirming the notion that signaling is dependent upon CRAF in this cellular subset [11].
To determine whether loss of RAF integrity or inhibition of MAP kinase signaling were correlated with a functional effect of 17-AAG upon cell proliferation, we used the MTT assay to measure the relative rates of cell proliferation of these 5 melanoma cell lines with increasing concentrations of 17-AAG. In a preliminary experiment, cells were incubated with increasing concentrations of 17-AAG (0.1, 0.3, and 1.0 μM). Although partial inhibition of cell viability was observed with 100 nM 17-AAG, for all 5 cell lines, a concentration of 300 nM 17-AAG was sufficient to confer the maximal inhibitory effect (Fig 3A). In a time-dependent manner, 300 nM 17-AAG conferred inhibition of cell proliferation of all 5 melanoma cell lines. With 3/5 cell lines (SK-Mel-2, Mel-Juso, and SK-Mel-28), growth was completely suppressed, whereas with 2/5 cell lines (SK-Mel-30 and A375), growth was partially suppressed at this concentration ( Fig 3B).
PLX4032, (vemurafenib) an ATP-competitive RAF inhibitor, inhibits ERK signaling and cell proliferation in BRAF V600E melanomas, but enhances the signaling in BRAF WT cells with  17-AAG inhibits MAPK activation and promote cellular immunotherapy against melanoma mutated NRAS [21,22]. Yet, the ability of 17-AAG to inhibit MAP kinase activity and cell proliferation was retained in NRAS mutant melanoma cell lines (Figs 2 and 3). Therefore we wondered whether 17-AAG could inhibit the PLX4032-induced ERK activation in NRAS mutated human melanoma cells. To achieve this goal, first we examined the effect of PLX4720 and PLX4032 on ERK activation on candidate NRAS mutated (Mel-Juso) and BRAF V600E (A375) cells. As expected, relative MEK phosphorylation and ERK phosphorylation were enhanced in Mel-Juso and inhibited in A375 cells, confirming prior observations [21,22]. CRAF expression was enhanced in Mel-Juso and slightly diminished in A375 cells, while BRAF V600E and BRAF WT expression was invariant to PLX treatment in each cell lines. PLX4720, a less potent inhibitor of BRAF V600E than PLX4032, showed weaker inhibition of MEK phosphorylation and ERK phosphorylation in A375 cells relative to PLX4032. However, PLX4720 showed similar paradoxical activation of MAPK signaling as indicated by enhanced activation of MEK and ERK phosphorylation similar to PLX4032 in NRAS mutated Mel-Juso cells (Fig 4A). To determine whether altered MAPK signaling was correlated with a functional effect upon cell proliferation, we used Trypan blue exclusion to measure the relative rates of cell proliferation. PLX4720 (1μM) and PLX4032 (1μM and 2μM) suppressed completely the growth of A375 cells (Fig 4B), whereas there was no significant effect on Mel-Juso cells (Fig 4C). Despite the enhanced ERK signaling observed in PLX4032 and PLX4720-treated Mel-Juso cells no additional effect on cell proliferation rate was detected [21]. Since PLX4720 had similar effects on activation of MAPK signaling and cell growth as PLX4032 in BRAF WT melanoma cells, but showed a weaker inhibition of MAPK signaling in BRAF V600E mutant melanoma cells, we decided to used PLX4032, a widely use compound, for further experiments.
These experiments were extended to include 17-AAG. As before, relative MEK and ERK phosphorylation was inhibited in BRAF V600E A375 melanoma cells but enhanced by PLX4032 in all 3/3 BRAF WT melanoma cell lines evaluated (Mel-Juso, SK-Mel-30, SK-Mel-2). However, relative MEK and ERK phosphorylation was inhibited in both BRAF V600E and BRAF WT cell lines when 17-AAG was added after PLX4032. The addition of 17-AAG to PLX4032-treated cells led to a significant degradation of CRAF at 24h in 3/3 BRAF WT cell lines tested (Fig 5A). In addition, to determine the minimum concentration of 17-AAG required to inhibit PLX 4032 enhanced ERK signaling, Mel-Juso cells were incubated with increasing concentrations of 17-AAG (.03, 0.1, 0.3, and 1.0μM) with or without PLX4032 pretreatment. MEK and ERK phosphorylation were inhibited to their minimum values by 0.1μM of 17-AAG in control cells but required the higher 0.3μM concentration of 17-AAG in PLX4032 pre-treated cells ( Fig  5B). Cell growth studies showed that 2μM PLX4032 by itself conferred no positive or negative effect on cell proliferation in all 3/3 BRAF WT melanoma cell lines. However, 2μM PLX4032 pretreatment in combination with 1μM 17-AAG significantly inhibited cell proliferation over the period of time in 3/3 (Mel-Juso, SK-Mel-30 and SK-Mel-2) BRAF WT cells (Fig 5C). In addition, the cell growth inhibition in PLX4032-pretreated BRAF WT melanoma cells by 17-AAG was further studied for cell cycle inhibition and apoptosis. PLX4032 by itself conferred no positive or negative effect on cell cycle and apoptosis in all 3/3 BRAF WT melanoma cell lines. However, PLX4032-pretreatment in combination with 17-AAG and 17-AAG by itself induced cell cycle arrest in all 3/3 BRAF WT melanoma cell lines. In all three, 17-AAG by itself and in combination with PLX4032 pretreated BRAF WT melanoma cell lines showed highly significant decreases in S phase and increase in G0/G1 arrest. In SK-Mel-2 cells there was significant increase in sub G1 phase in both 17-AAG by itself and in combination with PLX4032-pretreated cells (Fig 6A). In addition, PLX4032 + 17-AAG and 17-AAG alone showed higher Annexin V-positivity in 3/3 BRAF WT melanoma cell lines in comparison to their respective controls. The addition of PLX4032 to 17-AAG also, increased apoptosis in Mel-Juso and SK-Mel-2 cells (Fig 6B). These observations demonstrate that 17-AAG, an Hsp90 inhibitor, can inhibit RAF inhibitor enhanced MAP Kinase signaling through destabilizing CRAF resulting in BRAF WT melanoma cell arrest and apoptosis.
Interestingly, when SK-Mel-30 cells were treated with 17-AAG with or without PLX4032 for 72h, the cells demonstrated a higher visible level of pigmentation ( Fig 7A). This observation suggests that 17-AAG, in addition to inhibiting MAPK signaling and human melanoma cell growth, can also stimulate melanin formation. This observation led us to determine whether 17-AAG induced melanin synthesis is correlated with the expression of the canonical melanogenic proteins dopachrome tautomerase (DCT), tyrosinase-related protein-1 (TYRP1), and tyrosinase (TYR). PLX4032 by itself conferred no effect on the expression of DCT, TYRP1 and TYR in all 3/3 BRAF WT melanoma cell lines. However, 17-AAG, in combination with and without PLX4032, increased the expression of DCT and TYRP1, but not TYR, in all 3/3 BRAF WT melanoma cell lines (Fig 7B). It was surprising that the enhanced melanization observed in SK-Mel-30 cells occurred in the absence of an induction of tyrosinase, the ratelimiting enzyme in melanin biosynthesis [32]. To determine whether DCT and/or TYRP1 contribute to 17-AAG-induced melanization, we used siRNAs to deplete these proteins. 17-AAGinduced DCT and TYRP1 induction was significantly blunted in DCT and TYRP1-depleted cells in comparison to control siRNA (Fig 7C and 7E). Depletion of DCT and TYRP1 also reduced total melanin content compared to siRNA-transfected controls (Fig 7D and 7F). Hence 17-AAG stimulates melanization in SK-Mel-30 cells via a tyrosinase-independent mechanism that is dependent upon DCT and TYRP1.
In previous studies, DCT and TYRP1 were identified as melanoma tumor antigens [28,29]. Their induction by 17-AAG treatment of melanoma cells could render cells more immunogenic against adoptive, cell-based therapy. To test this hypothesis, we utilized a preclinical mouse model whereby the adoptive transfer of naive CD4+ T cells specific for Tyrp1, a melanoma differentiation antigen, results in conversion to cytotoxic T cells in vivo and inhibition of tumor growth in lymphopenic murine hosts by direct tumor cell killing [33]. However, murine hosts only exhibited a transient antitumor response, with tumor recurring locally [33,34]. Hence we asked whether the combination of 17-AAG, responsible for induction of Tyrp1, and adoptive transfer of cytotoxic CD4+ T cells specific for Tyrp1 can enhance the anti-tumor immune response by delaying recurrence of B16 melanomas.
First, we determined whether HSP90 inhibition by 17-AAG exhibited similar effects upon Tyrp1 and Dct expression in cultured B16 murine melanoma cells. These experiments confirmed that 17-AAG inhibited MAPK signaling shown by absence of phosphorylated MEK and ERK, and induced expression of Dct and Tyrp1 melanogenic proteins (Fig 8A). To determine if 17-AAG treatment can enhance the immune response of adoptively transferred cytotoxic CD4+ T cells against established melanoma tumors, we subcutaneously injected C57BL6 mice with B16.F10 mouse melanoma cells. On the fifth day after xenografting, mice received either 17-AAG or vehicle for 5 days. 7 days after xenografting, all C57BL6 mice were irradiated and 2 x 10 5 Tyrp1-specific CD4+ T cells were injected intravenously into selected 17-AAG and vehicle-treated tumor-bearing mice (Fig 8B). Tumor measurements were made twice weekly until 40 days after xenografting or until the experimental endpoint was reached. Adoptive transfer of Tyrp1-specific CD4+ T cells by itself caused significant regression of established tumor relative to vehicle treated melanoma bearing mice (Fig 8C, upper panels). However, similar to prior studies [33,34] we observed recurrence of melanoma tumor in 6/6 mice receiving CD4+ T cells alone (Fig 8C, upper right panel). In mice receiving the combination of 17-AAG and antigen-specific CD4+ T cells, tumor regrowth was significantly delayed, with only 2/6 mice reaching the experimental endpoints (Fig 8C, right bottom panel). The combination of 17-AAG and Tyrp1-specific CD4+ T cells treated mice showed significantly higher 17-AAG inhibits MAPK activation and promote cellular immunotherapy against melanoma survival at Day 40 after tumor challenge when compared with Tyrp1-specific CD4+ T cells alone (P<0.05, Fig 8D).

Discussion
Previous reports have considered the activity of 17-AAG on the RAF kinases CRAF, and BRAF, in melanoma cells. A study including two human melanoma xenograft lines showed that CRAF was destabilized by the geldanamycin analogs 17-allylamino-17-demethoxygeldanamycin (17-AAG) and 17-dimethylaminoethylamino-17-demethoxy-geldanamycin (17-DMAG) [35], although this effect was variable when another xenograft line was examined [36]. Subsequent studies with cutaneous and uveal melanoma cells [17,18,37,38] have suggested that BRAF V600E is more sensitive to 17-AAG-induced degradation than BRAF WT , a suggestion that has been attributed to preferential or exclusive association of BRAF V600E rather than BRAF WT with HSP90. Here we show that BRAF WT , at least in the context of activated NRAS in melanoma cells, binds to HSP90 in melanoma cells in a complex with CRAF. Our ability to detect this interaction was enhanced with a shorter incubation time and inclusion of sodium molybdate that were originally described for studying the interaction between HSP90 and CRAF [31]. Altering experimental conditions to resemble these conditions may favor the ability to detect weaker interactions between HSP90 and other client proteins.
Despite the fact that BRAF WT is bound by HSP90 in Mel-Juso cells (Fig 1), it was not degraded by 17-AAG (Fig 2). However, melanoma cell proliferation was inhibited in these and other cell lines (Fig 3) regardless of whether BRAF underwent degradation. In both our study and those reported previously [17,18], BRAF proteins have been found to be less sensitive to HSP90 inhibition than CRAF. Importantly, cell proliferation and MAP kinase signaling were inhibited in three melanoma cell lines with NRAS activating mutations, Mel-Juso, SK-Mel-30, and SK-Mel-2, where BRAF WT was variably insensitive to 17-AAG. These findings support the notion that CRAF is the effector RAF kinase for signaling to MEK in melanoma cells harboring NRAS mutations [11].
Currently, no targeted therapy is available to treat advanced BRAF WT melanomas. Treatment of BRAF V600E melanomas invokes resistance in BRAF V600E melanoma cells, with mechanisms of resistance including the development of upstream mutations in NRAS or KRAS [24]. One possible strategy to delay the clinical onset of resistance is to co-administer drugs that inhibit the ability of initial NRAS or KRAS-mutant clones to proliferate, as has been suggested previously with 17-AAG and melanoma [39]. Here we show that co-administration of 17-AAG inhibits the paradoxical activation of MAP kinase signaling observed in PLX-treated BRAF WT , NRAS-mutant human melanoma cells, providing experimental support for the notion that co-administration may block the expansion of nascent, RAS-mutant clones in PLX-treated melanomas. Recent reports with alternative HSP90 inhibitors support this idea [25][26][27].
17-AAG also induced melanin formation via a tyrosinase-independent mechanism in SK-Mel-30 cells. MEK inhibition has previously been shown to increase melanogenic gene expression, including tyrosinase expression, in immortalized murine melanocytes [40], and tyrosinase gene and protein expression in B16 murine melanoma cells [41]. Here the 17-AAG dependent, tyrosinase-independent induction of melanin formation in human SK-Mel-30 melanoma cells implicates DCT and TYRP1 in facultative melanization, showing that these accessory melanogenic enzymes [42,43] can modulate pigmentation without changes in tyrosinase. The prior identification of DCT and TYRP1 as melanoma tumor antigens [28,29] prompted us to investigate whether melanoma cells became more immunogenic following 17-AAG exposure (Fig 8). Our results confirm that this is the case and represent the first 17-AAG inhibits MAPK activation and promote cellular immunotherapy against melanoma demonstration of a synergistic therapeutic effect of an HSP90 inhibitor with cellular immunotherapy in malignant melanoma. Additional work will be required to determine whether this synergistic effect depends upon tumor antigen specificity or on a broad enhancement of tumor inflammation as has been recently suggested [44].  [17], or a 50:50 mixture of DMEM/Ham's F12 medium [18]. B16.F10 (H-2b), hereafter called B16, is a Tyrp1+ spontaneous murine melanoma cell line and maintained in culture conditions as previously described [45].

Mass spectrometric identification of HSP90 as BRAF interactor
Cell lysate was prepared from Mel-Juso cells cultured as above by lysing cell pellets in RIPA buffer with protease and phosphorylation inhibitors (20 μg/ml aprotinin, 20 μg/ml leupeptin, 1 mM sodium orthovanadate, 2 mM phenylmethylsulfonylfluoride). BRAF was immunoprecipated from cell lysate by incubating 100 μl lysate with 10 μl antibody solution and 20 μl Protein A-agarose (Pierce) in TENSV buffer [5] with the protease and phosphorylation inhibitors described above at room temperature for 30 min with gentle rolling. The antigen-antibody-Protein A-agarose complex was collected by centrifugation at 3000 g for 1 min. The resulting complex was washed twice with 0.5 ml TENSV buffer with inhibitors and loaded onto lanes of a 10% SDS-PAGE gel for electrophoresis. The gel was Coomassie-stained and the indicated band ( Fig 1A) visualized, cut with a sharp blade, and submitted for mass-spectroscopic analysis.

Co-immunoprecipitation of BRAF, CRAF, and HSP90 proteins
Immunoprecipitation of BRAF was performed as described above. The resulting complex was washed twice with TENSV buffer containing inhibitors and loaded onto lanes of a 10% SDS-PAGE gel for electrophoresis. Following electrophoresis, the gel was subjected to Western transfer and incubation with antibodies. Similar procedures were used for immunoprecipitation of HSP90 and CRAF, except that protein G-agarose, instead of protein A-agarose, was used.

Inhibition of MAP kinase signaling and evaluation of BRAF and CRAF stability by 17-AAG, PLX4032 and PLX4720
Cells (Mel-Juso, SK-Mel-2, SK-Mel-28, SK-Mel-30, A375 and B16) were grown in 10 cm tissue culture plates under the conditions described previously. 17-AAG (Cancer Therapy and Evaluation Program, National Cancer Institute) at a stock concentration of 10 mM in DMSO was added at final concentrations of 0, 100, 300, and 1000 nM for 24 h or PLX4032 or PLX4720 (ChemieTek. Indianapolis, IN) at a concentrations of 1 and 2 μM for 24h or 48h. Cell pellets were lysed in RIPA buffer with protease and phosphorylation inhibitors as above and protein concentration determined using the Bio-Rad protein assay (Bio-Rad Laboratories, #500-0006).Lysates were loaded onto a 10% SDS-PAGE gel, subjected to electrophoresis, and transferred to a PVDF (Invitrogen) membrane for Western blotting. The following primary antibodies were used: monoclonal anti-BRAF ( Following the incubation period, 10 μl of MTT labeling reagent was added to each well and incubated at 37 o C for an additional 4 hours. 100 μl of solubilization reagent was added to each well and the colorimetric reaction was allowed to develop overnight in the 37 o C incubator. An ELISA plate reader was used to record the optical densities at 550 nm and 690 nm. To assess the effect of PLX4032 and PLX4720 on human melanoma cell proliferation, the Trypan blue exclusion method was used. Melanoma cells (Mel-Juso and A375) were seeded overnight in individual wells of 6 well plate with 2ml of the indicated cell culture medium. The initial medium was replaced with 2ml of medium containing 1μM of PLX4720, 1 or 2 μM PLX4032 or vehicle for 24, 48, and 72 hours. Time-dependent inhibition of human melanoma cells with PLX was determined by counting Trypan blue-excluded cells.
To determine the effect of 17-AAG on PLX resistant WT BRAF melanoma cells, trypan blue exclusion method was used. Melanoma cells (Mel-Juso, SK-Mel30, SK-Mel2) were treated with 2 μM PLX4032 or vehicle for 24h as described above and were further treated with or without 17-AAG for 24, 48 and 72 hours. Time-dependent inhibition of human melanoma cells with PLX4032 and 17-AAG was determined by counting Trypan blue-excluded cells.

Flow cytometry assay
To determine the effect of 17-AAG on cell cycle arrest of PLX-resistant WT BRAF melanoma cells (Mel-Juso, SK-Mel-30, SK-Mel-2), cells were pre-incubated with PLX4032 (2μM) or vehicle (DMSO) for 24h and these cells were incubated subsequently with or without 17-AAG (1μM) for an additional 72h. Post treatments, cells were fixed in chilled ethanol (70%), washed with PBS, and then treated with DNase-free RNase (200 μg/ml) for 1 hour at 37˚C, stained with PI (40 μg/ml) and kept in the dark for 15 minutes at 20-25˚C. Staining was measured on a FACSCanto II flow cytometer (BD Biosciences, San Jose, CA), and percentages of cells in different cell cycle phases were determined using FlowJo Software (Tree Star, Inc., Ashland, OR).

siRNA inhibition of DCT and TYRP1
SK-Mel-30 cells seeded into 6 well plates at 2.0 × 10 5 were treated with either 40 nM DCT siRNA or TYRP1 siRNA #1 or #2 or 40 nM control non-targeting siRNA (Customized at Integrated DNA Technologies, Inc.) using the Lipofectamine RNAiMax transfection reagent (Invitrogen) as per the manufacturer's instructions. DCT and TYRP1 inhibition was confirmed by western blotting 120 h after transfection. There was significant decrease in DCT and TYRP1 protein relative to control siRNA treated SK-Mel-30 cells at the time point. The

Quantification of the melanin content
The cultured cells at the end of experiment were detached from the culture plate using trypsin-EDTA solution and counted in a hemocytometer. Cell suspensions equivalent to 200,000 cells were pelleted and lysed in 1ml of 1N NaOH and the melanin concentration is determined by measuring optical density at 475 nm and compared with a standard curve obtained using melanin isolated from sepia (Sigma). Data of melanin concentration is represented as melanin produced by each cell in a given time [46].

Mice and adoptive transfer of T cells
C57BL/6 mice were purchased from Veterinary Resources, Program in Comparative Medicine, University of Maryland School of Medicine. Tyrp1 TCR x tyrp-1 bw Rag1 -/transgenic mice express a transgenic TCR that recognizes the mtyrp1 (113-127) peptide in the context of MHC class II. These mice also are homozygous for the white-based brown mutation, Bw that has a defect in exon 1 of tyrosinase-related protein-1 gene [33,47].
The C57BL6 mice were injected subcutaneously in the abdomen with 2 x 10 5 B16 mouse melanoma cells with a 27-gauge needle. This dose is well above the dose required to establish tumors in mice. Tumors were allowed to become established and sizeable for injections. Animals were euthanized at signs of distress and/or cachexia and tumor sizes beyond 200 mm 2 .
Five days after B16 tumor cell injection, the mice received intraperitoneal injections of either 17-AAG or vehicle at a dose of 75 mg/Kg body weight x 5 consecutive days [48]. Seven days after tumor challenge, all mice were irradiated one time with 550 rads (5.5Gy). Nine to twelve days after tumor challenge, 2 x 10 5 Tyrp1-specific CD4+ T cells were injected intravenously into 50% of 17-AAG and 50% of vehicle treated tumor bearing mice. The 17-AAG dose of 75 mg/kg x 5 consecutive days were repeated every 2 weeks following initial administration of 17-AAG until the mice were euthanized. Tumors were measured with digital calipers every 3 to 4 days. The perpendicular diameters were determined and multiplied to generate the area (in mm 2 ), as described previously [33].
Tyrp1 CD4+ T cells were sorted from spleens of donor Tyrp1 TCR x Tyrp1 bw RAG-1 -/as described (Goding SR et al. 2013 J immuno). Spleens were harvested and made into single-cell suspensions. Cells were made devoid of red blood cells by ACK (Ammonium-Chloride-Potassium) lysis. Subsequently, cells were counted and enriched for CD4+ T cells by magnetic bead sorting using a CD4+ T cell enrichment kit (Miltenyl Biotech) as per manufactures instruction. Enriched Tyrp1 CD4+ T cells were counted and resuspended in PBS and used in adoptive transfer studies (2 x 10 5 cells/mouse).

Statistical analysis
All statistical analyses were performed using Graphpad Prism 6.0 software. One-way ANOVA was used to analyze data in experiments such as MTT assay, cell proliferation assay, apoptosis assay and quantification of melanin content. Two-way ANOVA was used to analyze data in cell cycle experiments. Survival curves were compared with a log-rank test. P values of 0.05 or less were considered significant.
Supporting information S1 Appendix. Mass spectroscopic analysis of the excised gel band of BRAF revealed peptides corresponding to isoforms of HSP90. Mass spectroscopic analysis of the excised gel band of BRAF from MelJuso melanoma cells showed 9 peptides corresponding to each α-and β-isoforms of HSP90. Nine peptides corresponding to full length sequences of α-and β-isoforms of HSP90 are highlighted in yellow.