A distinct pattern of growth and RAC1 signaling in melanoma brain metastasis cells

Abstract

Background

Melanoma, the deadliest of skin cancers, has a high propensity to form brain metastases that are associated with a markedly worsened prognosis. In spite of recent therapeutic advances, melanoma brain lesions remain a clinical challenge, biomarkers predicting brain dissemination are not clear and differences with other metastatic sites are poorly understood.

Methods

We examined a genetically diverse panel of human-derived melanoma brain metastasis (MBM) and extracranial cell lines using targeted sequencing, a Reverse Phase Protein Array, protein expression analyses, and functional studies in vitro and in vivo.

Results

Brain-specific genetic alterations were not detected; however, MBM cells in vitro displayed lower proliferation rates and MBM-specific protein expression patterns associated with proliferation, DNA damage, adhesion, and migration. MBM lines displayed higher levels of RAC1 expression, involving a distinct RAC1-PAK1-JNK1 signaling network. RAC1 knockdown or treatment with small molecule inhibitors contributed to a less aggressive MBM phenotype in vitro, while RAC1 knockdown in vivo led to reduced tumor volumes and delayed tumor appearance. Proliferation, adhesion, and migration were higher in MBM vs nonMBM lines in the presence of insulin or brain-derived factors and were affected by RAC1 levels.

Conclusions

Our findings indicate that despite their genetic variability, MBM engage specific molecular processes such as RAC1 signaling to adapt to the brain microenvironment and this can be used for the molecular characterization and treatment of brain metastases.

Graphical Abstract

Graphical Abstract.

Key points.

• Genetically diverse brain-derived melanoma cells display slow growth and high RAC1 expression.

• RAC1/PAK1/JNK1 signaling is active in melanoma brain metastasis cells and can be targeted with inhibitors.

• Insulin and brain-derived factors stimulate RAC1 signaling and aggressive melanoma brain cell phenotypes.

Importance of the Study.

Brain metastasis is a deadly complication for cancer patients and remains poorly treated and understood. Our work shows that genetically diverse melanoma brain metastasis cells can acquire higher RAC1 levels and signaling while also displaying a slower proliferation rate compared to extracranial melanoma cells, thus unique brain-derived pathobiological properties. RAC1 confers a growth advantage to melanoma brain metastasis cells in vitro and in vivo and can be targeted therapeutically. While RAC1 mutations are known drivers of melanoma, our study suggests that RAC1 expression levels should also be considered to assess aggressive tumor behavior. Additionally, brain-derived melanoma cells respond more favorably to insulin and brain-derived factors; therefore, brain metastases studies are best conducted in the presence of brain-like environments.

Brain metastases constitute some of the most lethal complications in melanoma, occurring in over 50% of patients with metastatic disease.¹ Current therapeutic approaches against systemic melanoma include immunotherapy, targeted therapy, or combination treatments; brain lesions add surgery and radiotherapy to the list.^2,3 Although modern treatments have shown activity in the brain, there remains a high risk of mortality and resistance compared to extracranial lesions.^1,2,4,5 Multiple factors can contribute to brain metastasis therapy resistance including the blood-brain barrier, genetic and immunological heterogeneity, also the brain microenvironment.^4,6 However, our understanding of melanoma brain metastasis (MBM) is limited and new disease models are needed for this purpose.⁷

Hypotheses posit that brain metastases result from the seeding of circulating tumor cells in the brain microvasculature, where they colonize following adaptation.² The strong influence of the brain microenvironment was shown when human cancer cells were xenografted into different organs in mice and analyses revealed complete gene expression reprogramming, and even gain of neuronal characteristics.⁸ Given that melanocytes are derived from the neural crest, it is conceivable that they acquire and benefit from molecular mechanisms supported by the brain.⁹ Early studies demonstrated that cells isolated from mouse MBM displayed arrest or death and enhanced migration in the presence of mouse brain soluble factors, while cutaneous-based melanoma cells behaved differently, suggesting unique biological responses intrinsic to MBM.¹⁰

Brain tissue is structurally and metabolically distinct, it also harbors neurons, microglia, and astrocytes, which can contribute to MBM biology directly or through secreted factors.^11,12 One study showed that astrocytes activate the proliferator-activated receptor γ pathway in MBM, exerting a prometastatic effect.¹³ Analyses using patient-matched brain and extracranial metastases also identified immunosuppression and oxidative phosphorylation enrichment in MBM.^4,14 Altogether, increasing studies reveal intrinsic and extrinsic signals contributing to MBM-unique properties. In vivo studies using patient-derived xenografts (PDX) meanwhile, showed that human MBM-derived cells can cause mouse brain lesions starting from a subcutaneous injection, but this potential is uncertain after long-term in vitro culture, reaffirming the importance of the brain microenvironment in maintaining MBM properties.^7,15

Currently, few studies demonstrate the presence of brain-specific genetic alterations in MBM; however, insights are emerging.^16,17 For example, RAS mutations and genetic alterations associated with sensitivity to PI3K/AKT/mTOR, cyclin-dependent kinases, and epidermal growth factor receptor (EGFR) inhibitors were highlighted.^18–20 PI3K pathway inhibitors in particular showed efficacy in preclinical MBM models and this pathway is one of the most frequently activated in MBM.^12,18,21,22 PI3K signaling can increase through mutations of its effectors (e.g. AKT1^E17K), by upstream activators (e.g. RAS), or through regulator loss (e.g. the phosphatase and tensin homolog [PTEN]).^20,23 However, the MBM molecular targets mentioned so far are by no means comprehensive.²⁴

Here, we characterized a new panel of genetically diverse human-derived MBM cell lines, investigated new brain-specific biological processes, and we explored anti-MBM therapeutic approaches. To do so, we used genetic and protein analyses, functional studies in standard 2-dimensional (2D) and 3-dimensional (3D) cultures, and in vivo observations.

Materials and Methods

Cell Culture and Reagents

The human MBM cell lines M230, M331, M450, WM4237, WM4265-2, and the nonMBM lines WM793, 1205Lu, WM983B, and WM3918 were established by our group and that of Dr. M. Lotem (Hadassah Medical Center, Israel); patient samples were collected under Institutional Review Board approval and were confirmed as melanomas (Supplementary Figure 1A).^15,25 Human astrocytes (ScienCell #1800, Carlsbad, USA) were a gift from Dr. C. Berndt. Reagents were purchased from Tocris Bioscience (Bristol, United Kingdom), Sigma-Aldrich (Munich, Germany), or Selleckchem (Houston, USA). Details in Supplementary Information.

Genetic Analyses

Melanoma cell lines were genetically characterized using a custom-targeted massively parallel sequencing panel consisting of 109 melanomagenesis-relevant genes (Supplementary Table 1). Detailed methods, analysis, mutation, and copy number calling were performed as previously described.²⁵

Proliferation and Viability Assays

Proliferation and viability were assessed by manual cell counting using a Neubauer hemocytometer, by the CellTiter-Blue Cell Viability Assay (Promega, Mannheim, Germany), or by cell staining with crystal violet then imaging. Details in Supplementary Information.

Adhesion, Migration, and Invasion Assays

Adhesion was assessed by serum deprivation, short-term cell incubation and washing, and 4 hours recovery of remaining cells prior to CellTiter-Blue addition and measurement. For migration, cells in serum-free media were seeded in transwell inserts (Corning, New York, USA) and were allowed to migrate 24 hours towards preconditioned media with 10% FBS, neurobasal media, or human astrocyte media, prior to Hoechst staining and imaging. Invasion was measured similarly, with Matrigel (Corning) precoated inserts. Melanoma spheroids were generated over 96 hours and were evaluated with the Live/Dead Viability/Cytotoxicity Kit (Invitrogen, Carlsbad, USA). Details are available in Supplementary Information.

Reverse Phase Protein Array (RPPA)

Reverse Phase Protein Array (RPPA) and data processing were performed by the MD Anderson Center RPPA Core Facility (Houston, USA). Briefly, cell lysates (and controls) were printed on nitrocellulose-coated slides using a 2470-Microarray printer (Aushon Biosystems, Billerica, USA). Slides were probed with primary antibodies plus biotin-conjugated secondary antibodies followed by signal detection. Data were normalized using median-centering across antibodies.²⁶ RStudio (Boston, USA) was used for data visualization.

Western Blots (WB)

Protein extraction was performed using a Triton-based Lysis Buffer, then 25–100 µg of protein/sample were resolved on 10% SDS-polyacrylamide gels. A Trans-Blot Turbo Transfer System (Bio-Rad, Hercules, USA) was used to transfer proteins to nitrocellulose membranes. Membranes were blocked with 5% BSA, incubated with primary antibodies overnight, then 1 hour with secondary antibodies (antibodies listed in Supplementary Table 2). WB was imaged and analyzed using an Odyssey Infrared Imaging System and Image StudioLite software (LI-COR, Lincoln, USA). Details are available in Supplementary Information.

Immunohistochemistry (IHC) and immunofluorescence (IF) assays

Human melanoma brain tumor specimens were collected and used under the Abramson Cancer Center’s melanoma research program tissue collection protocol UPCC08607 in accordance with the IRB of the University of Pennsylvania. IHC was performed on 5-μm-thick, formalin-fixed, paraffin-embedded sections. A 3,3′-diaminobenzidine-peroxidase substrate system produced the brown-staining product. Mouse brain specimens containing GFP-tagged WM4265-2 MBM cells were generated via intracardiac injection.¹³ IF was performed on 60-μm-thick whole mount sections and images were acquired with a LeicaSP5 confocal microscope. Antibodies are in Supplementary Table 2.

GTPase Activity Assays

RAC1 activity was measured using an Active RAC1 Pull-Down and Detection Kit (Thermo Fisher Scientific, Schwerte, Germany) according to the manufacturer’s instructions. RhoA activity was measured using a similar pull-down assay; details in Supplementary Information.

Melanoma/Primary Hippocampal Neuron Coculture

Primary hippocampal cultures were prepared from P0 Wistar rats.²⁷ Melanoma cells were prestained with CellTracker-Green-CMFDA (Molecular Probes #C2925, Eugene, USA). Cocultures were imaged (Cytation5, BioTek) and analyzed with Matlab (Mathworks). The melanoma Circularity Score = Cell Perimeter²/(4xπxarea), where, a perfect circle = 1. The Mann-Whitney U test was applied as a statistical test. Details in Supplementary Information.

Knockdown and Expression Experiments

Stable RAC1 downregulation was performed by shRNA transfection into melanoma cells with a lentiviral vector (pLKO.1). Lentiviral production was performed in HEK293TN (#LV900A-1-GVO-SBI, Heidelberg, Germany) over 48 hours and melanoma cell infection occurred 24 hours prior to puromycin selection. MISSION shRNAs were purchased from Sigma; Supplementary Table 3. NADPH oxidase (NOX) expression studies are in Supplementary Information and Supplementary Table 4.

In Vivo Studies

Experiments were performed in accordance with The Wistar IACUC protocol 201227 in NOD/LtSscidIL2Rγnull mice (NSG). The WM4237 MBM model was previously described.^7,15 Female and male mice (randomized) were implanted s.c. with 80 000 cells embedded in reduced growth factor BD-Matrigel Basement Membrane Matrix (Becton Dickinson, Bedford, USA); tumors were measured every 3–5 days. Details are available in Supplementary Information.

Statistical Analyses

Two-sample t-tests (2-sided, alpha = 0.05) evaluated differences between nonMBM and MBM cells, and compared control and shRAC1 groups. RPPA proteins were identified where the FDR associated with the t-test comparing protein expression of nonMBM and MBM cells was less than 0.05 or 0.1. A nested analysis of variance (ANOVA) was used to evaluate the difference between nonMBM and MBM cells when there were multiple independent experiments for each cell line. An ANOVA test was used for more than 2 groups; Tukey’s procedure for pairwise comparisons; Dunnett’s procedure to compare all groups to controls. A Wilcoxon rank sum test was used to compare the number of days to measurable tumors. With evidence of unequal variances, Welch’s t-test or ANOVA was used. Details in Supplementary Information.

Results

Melanoma Brain Metastasis Cells Display Slow Proliferation and Distinct Signaling Patterns In Vitro

We generated human-derived melanoma cell lines from different patients and metastatic sites including brain lesions.^15,25 Targeted sequencing of these lines featured common melanoma-altered genes (Supplementary Table 1) and a cell line panel was selected to compare genetically diverse intracranial and extracranial melanoma cells (featuring mutant BRAF, mutant N-RAS, PTEN, NF1 loss). We observed no unique genetic patterns specific to our brain-derived lines (Table 1). Although the genetics of the melanoma cells varied, in standard culture our MBM cell lines displayed significantly reduced proliferation compared to nonMBM lines, shown by cell number (Figure 1A) and proliferation rate (Figure 1B). We also examined metastasis-associated properties, but MBM cells did not display significantly altered adhesion (Figure 1C, P = .440), migration (Figure 1D, P = .425), invasion (Figure 1E, P = .630), or spheroid collagen invasion (Figure 1F) compared to extracranial cells (statistics, Supplementary Table 5). Our observations indicated that genetically diverse MBM cells display reduced growth in standard cell culture but adhesive or invasive properties are similar to nonMBM cells.

Table 1.

Melanoma Cell Lines Established from Extracranial or Intracranial Lesions and Their Genetic Profile. Cell lines were selected to include wild type or mutant BRAF, wild type or mutant N-RAS, wild type or mutated PTEN, and NF1 loss in at least 2 lines.

Cell Line	Lesion Type	BRAF Status	N-RAS Status	Additional Gene Mutations	High Gain Copy Number	Homozygous Copy Loss
WM793	Vertical growth phase melanoma	V600E	Wild type	CDK4, EGFR, MAP3K5, PTEN, TERT	BRAF, CDK6, DYNC1I1, EGFR, EZH2, GNAS, MET, RAC1, SMO, SRC, TRRAP
1205Lu	Lung metastasis	V600E	Wild type	CDK4, GRM3, MAP3K5, PREX2, PTEN, TERT	KRAS, MYC, NF1, RAC1, SNX31	CDKN2A/B
WM3918	Metastatic lesion	Wild type	Wild type	DCC, DYNC1I1, GRIN2A, KDR, PDGFRA, PREX2, PTPRD, PP6C, SOX10, TERT	MYC	CDKN2A, NF1
WM983B	Inguinal node metastasis	V600E	Wild type	ATM, CDKN2A, KDR, TERT, TP53	BRAF, EGFR, EZH2, MITF, RAC1	MMP8
M230	Brain metastasis	V600K	Q61R	GRM3, PTCH1, SETD2, TERT, TRRAP	AKT3, MDM4, MITF, MMP8, STK19, TERT	CDKN2A/B
M331	Brain metastasis	V600E	Wild type	ALK, BRCA2, CDKN2A, DDX3X, JAK3, MAP2K1, MAP3K9, PREX2, RB1, RET, RHOT1, STK11, TERT, TP53, TRRAP	STK19	NF1
M450	Brain metastasis	Wild type	Wild type	BAP1, CDKN2A, CTNNB1, DCC, DDX3X, FGFR3, GRIN2A, GRM3, MAP3K9, NF1, PTEN, PTPRK, ROS1, TP53, TRRAP, WT1		TP53
WM4237	Brain metastasis	V600E	Wild type	ATM, MET, PDGFRA, RB1, TERT, TP53,	BRAF, CDK6, DYNC1I1, EIF2AK3, EZH2, GRM3, MET,
WM4265-2	Brain metastasis	Wild type	Q61K	ARID1B, CDKN2A, EGFR, NF1, PDGFRB, PTEN, TERT, TP53, TRRAP	MITF, STK19

Fig. 1.

Melanoma Brain Metastasis Cells Display Distinct Proliferation and Signaling In Vitro. (A) Proliferation of MBM (warm colors) vs nonMBM (cool colors) cells over 96 hours. Triangles represent the mean±SEM (n = 12 measurements; 4 wells/3 experiments/cell line). (B) Proliferation rates (estimated slopes) for MBM vs nonMBM cells. (C) Adhesion of MBM vs nonMBM cells. Data presented as mean±SEM (n = 12 as [A]). (D) Migration of MBM vs nonMBM cells over 24 hours. Data presented as mean±SEM (n = 3). (E) Invasion of MBM vs nonMBM cells through Matrigel (24 hours). Data presented as mean±SEM (n = 3). (F) Invasion of 3D melanoma spheroids through collagen (72 hours). Representative images (n = 3). Live cells (green); dead cells (red); scale bar: 100 µm. (A–E) Suppl.Table 5: statistics. (G–I) Heatmap generated from RPPA analyses featuring all hits with differential expression (FDR = 0.1); (G) proteins related to PI3K signaling, (H) migration/adhesion-related proteins, (I) DNA damage response elements.

We next investigated signaling pathways that could be uniquely activated in MBM. This was initiated by interrogating MAPK, PI3K, STAT3, and PD-L1 pathways.^20,28–30 The most notable observation, also reported by others, was the activation of AKT (but not predicted by PTEN levels; Supplementary Figure 1B, C). To further examine brain-specific signaling, we used RPPA to determine the expression of 297 proteins in our cell line panel (Supplementary File 1). Proteins differentially expressed between MBM and nonMBM groups using a false discovery rate (FDR) of 0.05 were: 4E-BP1, TP53BP1, androgen receptor (AR), D-alpha tubulin, EGFR [pY1173], N-RAS, poly-(ADP-ribose) polymer (PAR), and TFRC (coefficient of variation, Supplementary Table 5). With the FDR of 0.1, we identified c-Abl, NF-κBp65 [pS536], PLCG2 [pY759], PREX1, and RICTOR. Categorizing these hits and including associated proteins differentially expressed in MBM, we observed expression differences related to PI3K/mTOR, N-RAS, and NF-κB signaling in our MBM lines (Figure 1G); these are linked signaling-wise (Supplementary Figure 1D). Distinct expression patterns were also found in migration/adhesion (Figure 1H) and DNA damage-related proteins (Figure 1I).

The importance of PI3K/AKT signaling in MBM has been demonstrated by multiple groups.²⁰ The involvement of RAS was also reported.¹⁹ Meanwhile, the expression of mTOR and NF-κB effectors (Supplementary Figures 1E, F) was inconsistent. However, we observed a potential link with ras-related C3 botulinum toxin substrate 1 (RAC1) and this has not been extensively studied in MBM despite ras-related C3 botulinum toxin substrate 1 (RAC1) being among the top ten functionally mutated proteins in melanoma.³¹ RAC1 is involved in PI3K and RAS signaling, it controls phosphorylation of the myosin-II heavy-chain and actinomyosin disassembly, as well as DNA damage.^32,33 PREX1 is a RAC1 guanine nucleotide exchange factor (GEF) and Merlin (low in MBM) is a tumor suppressor involved in RAC1-PAK1 signaling.³⁴ We, therefore, focused on further investigating the role of RAC1 in MBM.

MBM Cells Express High RAC1 Protein Levels

To examine the involvement of RAC1 in MBM, we explored RAC1 and Cdc42 expression in our panel of cell lines. We consistently detected higher total RAC1 protein levels in all our MBM compared to nonMBM lines, but not Cdc42 (Figure 2A) or RAC3 (Supplementary Figure 2A). Since standard 2D culture conditions are not fully representative of in vivo conditions, we also interrogated 3-dimensional (3D) spheroids. Under these conditions, we still observed higher MBM RAC1 levels (Figure 2B). RAC1 activity, however, was not increased (Figure 2C). Confirming in vitro results, RAC1 was also found highly expressed in human MBM patient samples, but variability exists (Figure 2D).

Fig. 2.

MBM Cells Express High RAC1 Protein Expression. (A–C, E–F) NonMBM cell lines (blue); MBM cell lines (red). (A) WB analysis for RAC1 and Cdc42; band quantitation is the ratio between evaluated protein and controls (calnexin, GAPDH). (B) WB analysis of RAC1 in lysates from 3D collagen-embedded spheroids. Quantitation is the ratio between RAC1 and GAPDH. (C) Active RAC1 pull-down assay. Active RAC1 band quantitation is relative to total protein. Control: β-actin. (D) IHC staining of RAC1 (brown signal) in human melanoma brain tumor lesions (n = 3 patients). Scale bar: 100 µm. (E–F) WB analysis of PAK1 and JNK1 in (E) 2D lysates and (F) 3D lysates. Band quantitation is shown as the ratio between phosphorylated and total protein. Controls: calnexin, GAPDH.

RhoA is in a regulatory feedback loop with RAC1.³⁵ However, our analyses showed no consistent increase in RhoA levels or activity in MBM (Supplementary Figure 2B–C). The RAC1 effector p21-activated kinase 1/serine/threonine-protein kinase 1 (PAK1) and downstream effector c-Jun N-terminal kinase 1 (JNK1) on the other hand, showed higher activation in 2D and 3D MBM cultures (Figure 2E–F).^36,37 These findings suggested a potential RAC1/PAK1/JNK signaling axis at play in our MBM cells, which is of interest given the important role of PAKs in drug resistance and RAC/PAK/JNK signaling in neurons.^38,39 Using IF staining, we detected diffuse cytoplasmic RAC1 expression in our MBM cell lines in 2D and 3D cultures, similar to our patient samples (Supplementary Figure 2D–E). Together, our findings suggested that high RAC1 levels played a unique role in MBM.

RAC1 Inhibition Impairs Aggressive MBM Cell Phenotypes

Given previous reports of RAC1 GTPase-independent activity, we next investigated the importance of higher RAC1 levels in our MBM cell lines.⁴⁰ With shRNA knockdown (KD) of RAC1 (Supplementary Figure 3A), our melanoma cells had a slight reduction in proliferation, but this was not unique to MBM (Figure 3A, Supplementary Table 6). RAC1 KD also showed cell line-dependent effects on adhesion (Figure 3B), migration (Figure 3C), and invasion (Figure 3D–E). We next examined how RAC1-associated signaling was affected by its levels in standard culture. As Figure 3F shows, phosphorylated p-AKT and p-ERK levels were decreased in all lines. However, in MBM we observed higher p-PAK1 and p-JNK1, possibly to maintain signal activity (Figure 3F). These data further supported distinct RAC1/PAK1/JNK1 signaling in MBM but warranted more examination of the contribution of RAC1 to MBM.

Fig. 3.

RAC1 Inhibition By Knockdown Or Targeted Inhibitors Impairs Aggressive MBM Phenotypes. (A) Proliferation, (B) adhesion, (C) migration of MBM vs nonMBM cell lines with RAC1 KD over 72 hours. Data presented as mean±SEM (n = 12 measurements; 4 wells/3 experiments/condition). (D) Invasion of melanoma cells with RAC1 KD through Matrigel (24 hours). Data presented as mean±SEM (n = 12 as [A]). (E) Invasion of 3D spheroids through collagen (72 hours). Representative images (n = 3). Live cells (green), dead cells (red); scale bar: 100 µm. (F) WB analyses of adherent cells following RAC1 KD. Band quantitation is relative to total protein. Control: calnexin. (G) Inhibitors of RAC (EHop-016, 2.5 µM) and BRAF^V600E (vemurafenib, 5 µM) in MBM cell culture over 72 hours. Representative images of crystal violet stained cells (n = 3). Scale bar: 50 µm. (Supplementary Figure 3B: quantitation). (H) Inhibitor treatment of 3D spheroids over 72 hours. Representative images (n = 3). Live cells (green), dead cells (red); scale bar: 100 µm. (Supplementary Figure 3C: quantitation). (I) Small molecule inhibitors targeting PAK and (J) JNK for 72 hours were assessed in MBM (M331, M450, WM4237, WM4265-2) and nonMBM cells (WM793, 1205Lu, WM983B, WM3918). Proliferation changes are shown as a mean of n = 4 cell lines/condition; each line contributed n = 4 wells/condition ±SEM from 3 experiments. (A–D, I–J) Supplementary Table 6: statistics; significance: *P ≤ .05, **P ≤ .01, ***P ≤ .001.

The combination of BRAF^V600E inhibitors with other targeted therapies showed positive clinical results.⁴¹ We, therefore, tested vemurafenib with the RAC inhibitor EHop-016 against MBM cells and observed that this combination was more effective than either single agent (Figure 3G, Supplementary Figure 3B, and Supplementary Table 7). Similar effects were observed with 3D spheroids (Figure 3H, Supplementary Figure 3C, Supplementary Table 7). As a single agent, EHop-016 also inhibited melanoma proliferation and induced cell death (Supplementary Figure 3Di–ii; Supplementary Table 7), but the effects were strongest in combinations. Vemurafenib and the MEK inhibitor trametinib as single agents had only partial inhibitory effects on MBM cells (Supplementary Figure 3Ei–ii, Supplementary Table 7). Meanwhile, the PAK inhibitor FRAX597 (Figure 3I) and JNK inhibitor SU3327 (Figure 3J) both decreased cell proliferation, but FRAX597 required higher concentrations in MBM (Supplementary Table 7), possibly due to higher target levels or compensatory signaling. Summarized, our results showed that targeting RAC/PAK/JNK in vitro was effective against melanoma cells, including brain-derived cells, and combination therapies were most robust to eradicate MBM cells. These data also suggested a biological role for high RAC1 in MBM, especially since RAC1 can affect multiple functional mechanisms and pathways by binding to different molecular partners.

Insulin Enhances MBM Proliferation, Adhesion, Migration, and Invasion In Vitro

In standard culture, RAC1 function or reliance may not be fully engaged in MBM despite high protein levels being present. To understand the plasticity of MBM in different microenvironments, we stimulated MBM cells using insulin, IGF-1, VEGF, glucose, or cell density. Only human insulin significantly stimulated the proliferation of all MBM lines compared to extracranial lines (Figure 4A, Supplementary Figure 4A–D, and Supplementary Tables 89). Adhesion, migration, and invasion were also altered in MBM compared to nonMBM lines following insulin treatment (Figure 4B–D, Supplementary Figure 4E, Supplementary Table 9). Additionally, insulin increased RAC1 protein expression (Figure 4E) while RAC1 KD dampened insulin-stimulated proliferation (Figure 4F, Supplementary Table 8). These MBM results parallel the important role of insulin for neuronal maintenance.⁴²

Fig. 4.

Insulin Enhances MBM Growth, Adhesion, and Migration. (A) Proliferation, (B) adhesion, of MBM (warm colors) vs nonMBM (cool colors) cells following insulin treatment for 72 hours Data presented as mean±SEM (n = 12 measurements; 4 wells/3 experiments/cell line). (C) Migration of melanoma cells over 24 hours following 72 hours of insulin treatment. Data are presented as mean ± SEM (n = 3). (D) Invasion of 3D spheroids through collagen in the presence of insulin (72 hours). Representative images (n ≥ 4). Live cells (green); dead cells (red); scale bar: 100 µm. (E) WB analysis of melanoma cells treated with insulin (100 µg/ml, 72 hours). RAC1 quantitation was normalized to the control GAPDH. (F) RAC1 KD melanoma cell proliferation in 100 µg/ml insulin (72 hours). Data presented as mean±SEM (n = 12 as [A]). Significance: *P ≤ .05, **P ≤ .01, ***P ≤ .001. Supplementary Table 8: statistics.

Brain-Derived Factors Enhance MBM Aggressive Behavior In Vitro And In Vivo

The brain microenvironment could further stimulate MBM cells compared to nonMBM cells; therefore, we tested cell behavior in the presence of conditioned media from rat neurons and glia cocultures. Under these conditions, MBM cell migration was significantly higher compared to nonMBM migration (P < .001; Figure 5A, Supplementary Figure 5A). In addition, MBM cells grown among rat neural cells showed significant (P < .001) spreading (Figure 5B). Exposure of melanoma cells to human-derived astrocyte conditioned media (HACM) only increased proliferation in 2 out of 5 MBM lines (Figure 5C, Supplementary Figure 5B). However, most interesting was the significant difference in adhesion (P < .001; Figure 5D, Supplementary Figure 5C), migration (P < .001; Figure 5E, Supplementary Figure 5D), and invasion in the presence of HACM (Figure 5F, Supplementary Figure 5E–F, Supplementary Table 10). WB analyses showed that total-RAC1 and phosphorylated-PAK expression were also elevated under these conditions but not RAC1 activity (Figure 5G, Supplementary Figure 5G), suggesting the important role of the brain microenvironment in promoting RAC1/PAK signaling but not necessarily through higher GTP-RAC1. In the presence of RAC1 KD (Figure 5H) or FRAX597 (Figure 5I), proliferation was significantly decreased, even in the presence of HACM; decreases in adhesion and migration were also observed (Supplementary Figure 5H, I, and J, Supplementary Table 10). These data demonstrated the contribution of RAC1 to MBM in the presence of brain-derived factors.

Fig. 5.

Brain-Associated Factors Enhance MBM Growth, Adhesion, Migration, and Invasion. (A) Melanoma migration in conditioned media from rat glia and neuron cocultures. Data presented as mean±SEM (n = 3). (B) Melanoma cell spreading in rat glia and neuron cocultures (72 hours). Representative images (n = 3). Scale bar: 50 µm. Circularity score applies to melanoma cells only; a higher score = greater circularity. Data shown as mean±SEM (WM983B, n = 32; WM4237, n = 82). (C) Proliferation in HACM (72 hours). Data presented as mean±SEM (n = 12 from 4 wells/3 experiments/cell line). (D) Adhesion in HACM. Data presented as mean±SEM (n = 12 as [C]). (E) Migration (24 hours) in HACM. Data presented as mean±SEM (n = 3). (F) Invasion of 3D spheroids through collagen, with HACM (72 hours). Representative images (n ≥ 6). Live cells (green); dead cells (red); scale bar: 100 µm. (A–F) Experimental schematics are available in Supplementary Figure 5. (G) WB analysis of melanoma cells in HACM (72 hours). RAC1 quantitation was normalized to β-actin; for PAK/JNK it is the ratio between phosphorylated and total protein. (H) Proliferation of RAC1 KD melanoma cells in HACM (72 hours). Data presented as mean±SEM (n = 12 as [C]). (I) Effects of FRAX597 (1.5µM-72hours) on shRAC1 KD cell proliferation in the presence of HACM. Comparisons focus on untreated vs FRAX597, and HACM vs HACM+FRAX597 treatments. Data presented as mean ± SEM (n = 6 from 2 wells/3 separate experiments/condition). (J) IF staining of whole mount slides featuring mouse brains with GFP-tagged WM4265-2 human MBM cells (representative images). The mouse MBM model was generated over 4 weeks following intracardiac injection. DAPI (blue), GFP (green), RAC1 (red). Scale bar: 100 µm. Inset: white arrows point to focused RAC1 signals within WM4265-2 cells. (K) In vivo tumor growth of WM4237 with RAC1 KD in NSG mice. A summary of tumor size per group is shown over 58 days; tumors were resected for 2 mice in the control group (day 53, black dot). Control (n = 9), shRAC1_1 (n = 9), shRAC1_2 (n = 4). (L) (i) Median time to first measurable tumor (days); (ii) mean tumor size with or without RAC1 KD on day 50. Significance: *P ≤ .05, **P ≤ .01, ***P ≤ .001. Supplementary Table 8: statistics.

An intracardiac-induced in vivo mouse model of human MBM showed single melanoma cells in the brain displaying focused RAC1 areas (Figure 5J); this is in contrast to the diffuse RAC1 signal observed in patient tumors and spheroids. Since single cells neighbor different cell types, RAC1 may have different functions and localization depending on the environment and proliferation status. To confirm that RAC1 levels are essential to MBM in vivo, we used a MBM-PDX model (WM4237) that recapitulates human disease, starting from a s.c. injection site and disseminating to the brain.^7,15 We injected WM4237 cells with RAC1 shRNA in NSG mice and monitored primary tumors and metastasis using a luciferase signal (Figure 5K). RAC1 KD significantly delayed tumor appearance (P-value = .026; Figure 5Li) and reduced tumor volumes (P-value = .012; Figure 5Lii). Due to aggressive primary disease in the control group following our ex vivo MBM handling, studies were halted prior to brain-lesion detection (data not shown); however, our findings still support the important role of RAC1 for MBM tumor latency and growth in vivo.

Because RAC1 regulates reactive oxygen species (ROS) generation, we also investigated if NOX mRNA levels were altered in MBM.⁴³ No unique NOX expression patterns were observed (Supplementary Figure 6A, Supplementary Information) and the response to redox-regulating agents was also nonspecific (Supplementary Figure 6B–C). While the role of ROS in MBM needs further exploration, our preliminary studies suggest this work, as well as further studies on RAC1 in MBM, could all benefit from more in vivo-like experimental conditions.

Discussion

In this study, we aimed to identify MBM-specific molecular processes with therapeutic potential. Our cell lines displayed no clear or unique genetic alterations using targeted sequencing analyses, in line with what is reported so far.²⁰ However, our MBM cell lines displayed slower proliferation rates in vitro and this was only enhanced by insulin or HACM. This observation suggests remnants of a dormant phenotype, stimulated by specific conditions.² Our MBM lines did not display higher adhesion, migration, or invasion potential in vitro compared to nonMBM lines, suggesting these functions are not selected for, or these properties are lost ex vivo. Brain-derived microenvironment factors, on the other hand, played an important role in further teasing-out MBM differences.

Regarding distinct signaling, analyses of melanoma brain specimens previously highlighted pathways such as INK4a-CDK4/6-RB, ARF-p53-MDM2, RAS-RAF-MAPK, and PTEN-PI3K-AKT, among others.⁴⁴ These pathways were also found upregulated in our MBM cells (by RPPA); however, we also detected RAC1 effectors, later validated by high RAC1 protein expression analyses. Rac1 nonsilent mutations are found in approximately 5% of melanomas and the RAC1P29S mutation is the third most common hotspot mutation in melanoma.^31,45 Differences in RAC1 expression could therefore also contribute to aggressive cases.

RAC1 is a member of the Rho family of GTPases that regulates MAPK, mTOR, NF-κB signals, and multiple functions; it also plays a critical role in neuritogenesis and neuronal migration.⁴⁶ In our MBM lines, RAC1 appeared associated with PAK1/JNK1 signaling and to be stimulated by brain-derived factors. PAK1 is important for neuronal polarization and differentiation, while JNKs regulate cell death in the nervous system.^47,48 Therefore, there could be a brain-induced reprogramming of metastasized melanoma cells featuring a gain of neuronal characteristics. Interestingly, RPPA analyses conducted on other tumor types also reveal PAK1 and JNK as differentially expressed in brain samples.⁴⁹

Our observed increase in RAC1 levels in MBM without higher activity in vitro does not dismiss its important role.⁴⁰ Cells could be poised for activation in brain-like environments, a possibility given that we could stimulate MBM proliferation, invasion, and RAC1 expression with insulin or brain-derived factors. Interestingly, insulin is needed for neuronal survival, circuitry, and synaptic activity.⁵⁰ Additionally, RAC1 KD affected MBM growth and latency in vivo. The WM4237 MBM model requires long periods (>17 weeks) for brain metastasis detection.⁷ This study was the first attempt to use this model for biological studies and our results suggest that it is best used with limited ex vivo cell handling.

With respect to drug resistance, we observed partial response of our MBM cells to BRAF/MEK inhibitors, but combination approaches including a RAC inhibitor that eradicated almost all MBM cells, and PAK and JNK inhibitors were also effective, indicating the potential of co-targeting MBM-specific pathways to achieve a greater anti-MBM response.

We acknowledge that many of our findings are in vitro and could apply to only a subset of patients with MBM; nevertheless, our first attempt at characterizing a diverse panel of MBM cell lines unraveled unique properties. In sum, our findings highlight the importance of RAC1 signaling for MBM and the need to recapitulate the brain microenvironment for MBM-based studies.

Funding

Work supported by: the National Cancer Institute (NCI, PO1-CA114046, P01-CA025874, P30-CA010815, and R01-CA047159), the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation to M.H.; the German Research Foundation (DFG, SFB1027-C4, SFB1190-P17, BO3643/3-2, and IRTG1816) to I.B.; the American Cancer Society (ACS-IRG-96-153-10) to A.V.; NCI (R01-CA241490) to Q.C. The Wistar Institute core facilities are funded by the Cancer Center Support Grant P30-CA010815.

Conflict of Interest

The authors declare no potential conflicts of interest.

Authors’ Contribution

Conceptualization: A.V., I.S.-T., M.H. Experimentation: I.S.-T., A.V., A.W., P.B., C.K., T.G., H.L., Z.BdR., A.Z., S.M.S., B.G., S.S., A.Z., X.X., Q.C., J.H.O. Writing: A.V., I.S.-T. Resources: I.B., M.H., K.L.N., S.O.R., D.M.K., M.S., M.L., G.B.M., M.A.D., M.P.S. Analysis: I.S.-T., P.A.G., A.V., P.B., B.G., K.L.N., G.B.M., M.A.D. Supervision: A.V., I.B., M.H.; Funding: I.B., M.H., A.V. Proofreading: all authors.