Fibroblastic reticular cells (FRC) are immunologically specialized myofibroblasts that control the elasticity of the lymph node, in part through their contractile properties. Swelling of tumor-draining lymph nodes is a hallmark of lymphophilic cancers such as cutaneous melanoma. Melanoma displays high intratumoral heterogeneity with the coexistence of melanoma cells with variable differentiation phenotypes from melanocytic to dedifferentiated states. Factors secreted by melanoma cells promote premetastatic lymph node reprograming and tumor spreading. Elucidating the impact of the melanoma secretome on FRC could help identify approaches to prevent metastasis. Here we show that melanocytic and dedifferentiated melanoma cells differentially impact the FRC contractile phenotype. Factors secreted by dedifferentiated cells, but not by melanocytic cells, strongly inhibited actomyosin-dependent contractile forces of FRC by decreasing the activity of the RHOA–RHO–kinase (ROCK) pathway and the mechano-responsive transcriptional coactivator Yes1 associated transcriptional regulator (YAP). Transcriptional profiling and biochemical analyses indicated that actomyosin cytoskeleton relaxation in FRC is driven by inhibition of the JAK1-STAT3 pathway. This FRC relaxation was associated with increased FRC proliferation and activation and with elevated tumor invasion in vitro. The secretome of dedifferentiated melanoma cells also modulated the biomechanical properties of distant lymph node in premetastatic mouse models. Finally, IL1 produced by dedifferentiated cells was involved in the inhibition of FRC contractility. These data highlight the role of the JAK1-STAT3 and YAP pathways in spontaneous contractility of resting FRC. They also suggest that dedifferentiated melanoma cells specifically target FRC biomechanical properties to favor tumor spreading in the premetastatic lymph node niche. Targeting this remote communication could be an effective strategy to prevent metastatic spread of the disease.
Communication between dedifferentiated melanoma cells and lymph node fibroblasts reprograms the biomechanical properties of the premetastatic lymph node niche to promote tumor invasion.
See related commentary by Lund, p. 1692
Melanoma is a very aggressive skin cancer due to its high propensity to metastasis and pronounced intratumor heterogeneity. Four genomic subtypes of cutaneous melanoma have been defined based on the mutational pattern in BRAF, RAS, NF1, or none of these (1). During metastatic progression, melanoma cells are highly plastic and dynamically switch between proliferative and invasive phenotypes associated with distinct differentiation states ranging from melanocytic to dedifferentiated (2–5). The proliferative melanocytic state is characterized by high expression of the melanocyte lineage-specific Microphthalmia-Associated Transcription Factor (MITF) and low expression of the tyrosine kinase receptor AXL, whereas the invasive dedifferentiated state shows low expression of MITF and high expression of AXL. Dedifferentiated melanoma cells display a mesenchymal-like phenotype associated with drug resistance (6–8). Whether the melanocytic and dedifferentiated melanoma cell populations differ in their ability to communicate with the metastatic stromal host niche remains poorly understood.
Cutaneous melanoma is a cancer with an inherent potential for lymph node colonization, an event contributing to systemic metastasis (9–11). Melanoma cells secrete extracellular vesicles (12) and soluble factors (13) migrating to the premetastatic lymph node and conditioning immune cells (14), lymphatic endothelial cells (13, 15), and fibroblastic reticular cells (FRC; ref. 16). This reprogramming of the lymph node microenvironment creates a favorable niche that supports metastatic development.
FRC are immunospecialized myofibroblasts of lymphoid organs characterized as CD31–Podoplanin (PDPN)+ mesenchymal cells. They form a network in close contact with T cells and dendritic cells (DC) in the lymph node paracortical area and regulate immune cell recruitment, survival, and activation (17). With a notable spontaneous contractibility (18–20) and the secretion and remodeling of a dense reticular network of conduits (17, 20), FRC control the lymph node elasticity and microarchitecture. PDPN was shown to drive the actomyosin contractility of FRC through its binding to proteins of the ezrin, radixin, and moesin (ERM) family, leading to RHOA activation (18, 19, 21, 22). During an immune response, migratory DC expressing high levels of C-type lectin-like receptor 2 (CLEC2) are recruited to the lymph node. CLEC2 binding to PDPN on FRC dismantles the PDPN–ERM interaction, inhibiting actomyosin contractility and resulting in FRC stretching allowing rapid lymph node expansion (18, 19).
In other tissues, like the skin, quiescent fibroblasts are not spontaneously contractile but can be converted into contractile myofibroblasts by factors secreted during wound healing or tumor progression, such as TGFβ or IL6 family cytokines (23). Factors secreted by tumor cells transform fibroblasts in the tumor vicinity into cancer-associated fibroblasts (CAF; ref. 24). CAF are characterized by a contractile phenotype and the high expression of markers like α-smooth muscle actin (αSMA, ACTA2 gene), platelet derived growth factor receptor (PDGFR)-α, and PDGFRβ, fibroblast activation protein (FAP) α, fibroblast-specific protein-1 (FSP1, S100A4 gene), fibronectin and a fibronectin isoform containing the EDA domain (EDA-FN), vimentin (VIM), or secreted protein acidic and rich in cysteine (SPARC; ref. 24). In CAF, actomyosin contractility is driven by RHO and RHO-kinase (ROCK) signaling, leading to an increased phosphorylation of the myosin light-chain 2 (MLC2; ref. 25), and by the activation of the mechano-responsive yes1 associated transcriptional regulator (YAP; refs. 26, 27). Cytokines from the IL6 family [IL6, leukemia inhibitory factor (LIF), Oncostatin M] have been shown to induce RHO-ROCK–dependent CAF contractility through the GP130 (IL6ST)–JAK1–STAT3 pathway (28, 29). TGFβ increases actomyosin contractility in fibroblasts by promoting LIF expression, which subsequently epigenetically activates JAK-STAT signaling (30). Actomyosin contraction in CAF is associated with cell shape remodeling, increased F-actin stress fibers and drives force-mediated extracellular matrix (ECM) remodeling (25, 26).
Malignant lymph node colonization is preceded by premetastatic lymph node swelling (15, 16, 31) but it remains incompletely understood whether (and how) tumor-derived cues alter FRC-specific contractile properties during premetastatic niche reprogramming. In this study, we characterized the impact of factors secreted by melanocytic or dedifferentiated melanoma cells on the actomyosin contractility of isolated human FRC and in murine premetastatic lymph node models, and identified the underlying signaling pathways, the responsible tumor factors, as well as the impact of this process on the metastatic properties of tumor cells.
Materials and Methods
Isolation and culture of primary fibroblasts and CAF
Primary human lymph node fibroblasts (LN-F; #2530, ScienCell) were amplified in Fibroblast Medium (#2301, ScienCell) supplemented with 1% Fibroblast Growth Supplement (#2352, ScienCell), 10% FBS (Gibco), and 100 µg/mL penicillin/streptomycin solution (P/S; Gibco). Skin fibroblasts (skin-F) from healthy donors were isolated as described previously (32). CAF were isolated as described previously (33) from melanoma skin or lymph node clinical specimens (n = 9) obtained with written informed consent from patients, in accordance with the Declaration of Helsinki. The study was approved by local ethic committees (Nice Hospital Center and University Côte d'Azur, Nice, France). Skin-F and CAF were cultured in DMEM with 100 µg/mL P/S and 10% FBS. Fibroblasts were starved for 5 to 7 days in DMEM 0.5% FBS (control medium) before any experiment and were then stimulated every 2 days with control medium or melanoma conditioned media (CM) supplemented with 0.5% FBS. They were used until passage 10. In some experiments, cells were treated with 2 ng/mL TGFβ1 (#11343160, ImmunoTools), 2 ng/mL LIF (#300–05, PeproTech), 2 ng/mL IL1α (#200–01A, PeproTech), 2 ng/mL IL1β (#200–01B, PeproTech), or 10 µmol/L ruxolitinib (#S1378, Selleckchem), 10 µmol/L Y-27632 (#S1049, Selleckchem), or 10 µmol/L SB431542 (#S1067, Selleckchem). All experiments were performed on plastic, except collagen gel contraction assays and immunostainings.
Melanoma cell culture
Human melanoma cell lines were obtained as previously described (6, 34–36). They were authenticated by short tandem repeat DNA profile genotyping (Eurofins Genomics). Fluorescent 1205Lu Red cells were previously described (33). Short-term cultures of patient melanoma cells MM001, MM029, MM074, and MM099 were kindly provided by Jean-Christophe Marine (VIB KU Leuven Center for Cancer Biology, Leuven, Belgium; ref. 8) and MNC1 were described previously (36). Human melanoma cells were cultured in DMEM supplemented with 7% FBS and used until passage 30. Mouse YUMM1.7 cells (RRID: CVCL_JK16) were kindly provided by Marcus Bosenberg (Yale School of Medicine, New Haven, CT; ref. 37). They were cultured in Opti-MEM supplemented with 3% FBS. All cells were routinely tested for the absence of Mycoplasma by PCR.
Melanoma cells at 80% confluence were cultured for 24 hours in FBS-free medium. The culture supernatant was filtered (0.45 µmol/L) to remove cellular debris. CM prepared from 501Mel and 1205Lu melanoma cells were analyzed by mass spectrometry (MS) and antibody array as described (35, 36). CM injected to mice were previously concentrated 10-fold on 3 kDa MW cut-off membrane (Amicon Ultra-4, Merck Millipore). The resulting CM, or extracellular vesicles (EV), were aliquoted and frozen at −80°C until used.
Collagen gel remodeling assays
Fibroblasts (7 × 103 cells) were embedded in 30 µL of a 3.5-mg/mL collagen I (#354249, Corning) and 2-mg/mL Matrigel (#E1270, Sigma-Aldrich) mix and seeded in 5-mm Glass Diameter coverslip 96-well plate (#P96G-1.5–5-F, MatTek). Once the gel was set (30 minutes at 37°C), it was overlaid with 100 µL medium. Gels were photographed every 1 to 3 days to measure their area with ImageJ software (RRID: SCR_003070). The percentage of gel contraction was calculated as followed: 100 - 100 × (gel area/well area).
Atomic force microscopy
Mechanical properties of LN-F–remodeled collagen gels and unfixed 10-µm frozen lymph node sections were analyzed by atomic force microscopy (AFM; see Supplementary Data for details).
LN-F proliferation was measured using the CellTiter 96 Aqueous Non-Radioactive Cell Proliferation kit (#G5421, Promega) according to the manufacturer's instructions, or by cell counting, and was normalized to the control medium condition.
Cells were washed in PBS and incubated at 4°C for 30 minutes in PBS 2% FBS, 2 mmol/L EDTA with antibodies, and control isotypes listed in Supplementary Table S1. Then, cells were washed in PBS and analyzed with BD FACSCANTO II cytometer (BD Biosciences) and the FlowJo software (RRID: SCR_008520). A minimum number of 5 × 103 relevant LN-F were analyzed after exclusion of dead cells (SSC-H/FSC-H gate) and doublets (FCS-A/FSC-H gate).
Transfection of siRNAs listed in Supplementary Table S2 was carried out using Lipofectamine RNAiMAX (#13778150, Thermo Fisher Scientific) at a final concentration of 50 nmol/L. Cells were assayed at 2 days posttransfection.
Total RNAs were extracted using NucleoSpin RNA Plus kit (#740984.50, Macherey-Nagel). Reverse transcription was performed with the High-capacity cDNA Reverse Transcription Kit (#4368814, Applied Biosystems). Quantitative PCR was performed using the Platinum SYBR Green qPCR Supermix (#11558656, Thermo Fisher Scientific) with the StepOnePlus System (Applied Biosystems). Relative mRNA levels were determined using the 2ΔΔCt method and ACTB, GAPDH, HPRT, and PPIA as housekeeping genes.
Microarray experiment and analysis
LN-F were cultured for 48 hours with control medium or 1205Lu CM in 4 different experiments. RNAs were then extracted as described above and analyzed on SurePrint G3 Human Gene Expression 8 × 60K v2 Microarrays (#G4851B, Agilent Technologies) as previously described (see Supplementary Data for details; ref. 38).
Immunoblots were realized as previously described (36) with antibodies listed in Supplementary Table S1.
RHOA activity was measured on LN-F lysates using the RHOA G-LISA Activation Assay kit (#BK124, Cytoskeleton). Data are the mean ± SEM of arbitrary units of RHOA activity.
LN-F were grown 4 days on coverslips coated with 20 µg/cm2 collagen or on synthetic hydrogels of 2.8 kPa or 0.2 kPa stiffness coated with 20 µg/cm2 collagen. Hydrogels were prepared as previously described (see Supplementary Data for details; ref. 39). Then, cells were fixed in 4% paraformaldehyde, permeabilized in 0.3% Triton, blocked in 0.1% Triton, 5% goat serum, and stained overnight at 4°C with primary antibodies (Supplementary Table S1) diluted in 0.1% Triton, 2% Goat Serum. Following incubation with Alexa Fluor-conjugated secondary antibodies (1:1000, Thermo Fisher Scientific) and Texas Red- or Alexa Fluor 488-conjugated phalloidin (1:200, Thermo Fisher Scientific), nuclei were stained with Hoechst (#H1399, Thermo Fisher Scientific) and coverslips were mounted in ProLong diamond antifade (#P36961, Thermo Fisher Scientific). Images were acquired on a wide-field microscope (Leica DM5500B, ×40 magnification) or a confocal microscope (Nikon Eclipse Ti, ×20 or ×40 magnification). Images were analyzed with the ImageJ software to quantify the cell shape index, the mean fluorescence per cell (integrated density) and the nuclear/cytosolic ratio of YAP.
Mouse model of premetastatic draining lymph node
Experiments with mice were approved by an Institutional Animal Care and Use Committee under project license APAFIS#21820–2019070916066283v4. Female athymic nude mice (Charles River, RRID:IMSR_CRL:490) or immune competent 8-week-old C57BL/6j mice (Janvier, RRID: IMSR_JAX:000664) were injected every other day with 15 µL of 1205Lu or YUMM1.7 CM in the dermis of the right ear and 15 µL of 501Mel CM or control medium in the left ear. At day 7, mice were sacrificed and the superficial parotid draining lymph nodes were harvested for subsequent analysis. Lymph nodes of the same mouse were compared for paired statistical analysis. Lymph nodes were fixed overnight at 4°C with 3% paraformaldehyde (except lymph nodes prepared for AFM), soaked in 30% sucrose, and embedded in Tissue-Tek OCT Compound (#4583–01, Gentaur). Frozen sections (10 µm) were then treated as described in the immunofluorescence (IF) section. Z-stack images were captured every 0.5 µm on a confocal microscope (Nikon Eclipse Ti, ×60 magnification). Three-dimensional (3D)-reconstituted images were analyzed with the ImageJ software to quantify the nuclear/cytosolic ratio of YAP.
Collagen gel invasion assays
LN-F (2.3 × 104) were embedded in 100 µL of 2.5 mg/mL collagen I and 2 mg/mL Matrigel mix and seeded inside 24-well cell culture inserts (8 µm pore size, #3422, Corning). Once the gels were polymerized, the chambers were filled with control medium (with IL1β or Y-27632) or melanoma CM. The medium was changed on day 2. All conditions were performed in triplicates. On day 5, the LN-F remodeled collagen gels were washed twice, and gaps around contracted gels were filled with collagen/Matrigel mix. Then, 105 1205Lu Red cells were seeded on top of the gels in DMEM and the lower chamber was filled with DMEM 20% FBS. After 3 days, the collagen gels were placed upside down on glass-bottom dishes (#80136, Ibidi). Z-stack images were captured every 10 µm over 200 µm, on an inverted confocal microscope (Nikon Eclipse Ti, ×20 magnification). The number of cells in each 10-µm–step image was analyzed with ImageJ software and normalized to the number of cells detected in the stack. Maximum distance of invasion was defined as the z value above which 5% of cells in the stack was found. 3D projections were obtained with the NIS-Elements software (Nikon).
Statistical analysis was performed using Prism (v8, GraphPad, RRID: SCR_002798). Unpaired two-tailed Mann–Whitney tests were used for statistical comparisons between two groups and Kruskal–Wallis tests with Dunn posttests or two-way ANOVA tests with Sidak posttests to compare three or more groups. Histogram plots and curves represent mean ± SEM and violin plots represent median ± quartiles. P values ≤ 0.05 were considered statistically significant.
The experimental data from microarray have been deposited in the NCBI Gene Expression Omnibus (GEO) database (RRID: SCR_005012) under the series record GSE157355.
Lymph node fibroblasts harbor a CAF-like phenotype associated with spontaneous cell contractility
To investigate the interactions between melanoma cells and FRC, we employed primary fibroblasts isolated from human lymph nodes. Because several mesenchymal subsets coexist in the lymph node (20, 40), we first characterized these primary LN-F using microarray profiling, qRT-PCR, and flow cytometry (Supplementary Fig. S1A–S1C). Markers expressed by LN-F were typical of FRC (PECAM1–, PDPN+, CR2–, MADCAM1–), the most abundant lymph node mesenchymal subset.
FRC exhibit spontaneous contractility (18–20). When embedded in a collagen gel, LN-F isolated from 4 different donors demonstrated basal force-mediated gel remodeling, driven by the ROCK-actomyosin pathway, and inhibited by the ROCK inhibitor Y-27632 and the myosin inhibitor blebbistatin (Fig. 1A; Supplementary Fig. S1D–S1E). LN-F displayed the same high propensity to drive collagen gel remodeling as primary CAF isolated from skin or lymph node melanomas (Fig. 1B). However, primary skin-F were not able to contract the collagen gel unless activated with TGFβ to a CAF-like phenotype. In contrast, LN-F spontaneous contraction was not modulated by TGFβ. As for LN-F, contractile properties of TGFβ–activated skin-F and CAF were dependent of the ROCK–actomyosin pathway and inhibited by Y-27632. In agreement with the collagen gel remodeling results, LN-F contractile activity was associated with higher basal levels of F-actin stress fibers and YAP nuclear localization compared to skin-F (Fig. 1C and D). However, in contrast to skin-F, F-actin fibers and YAP nuclear localization were not increased by TGFβ treatment in LN-F. qRT-PCR analysis revealed that several CAF markers like ACTA2 (αSMA), FAP, EDA-FN, or SPARC were highly expressed in LN-F compared with skin-F, suggesting that resting LN-F share many properties with CAF (Fig. 1E). Immunoblot analysis also showed TGFβ–independent high expression of FN, PDGFR-β, FAP, and αSMA in resting LN-F, to levels equivalent or higher than TGFβ–activated skin-F (Fig. 1F). This is consistent with the notion that high αSMA expression is associated with fibroblast contractility (41). Collectively, these results validate the FRC signature of LN-F and unveil that quiescent LN-F display some phenotypic and functional properties of CAF. Our aim was then to investigate whether, and how, melanoma cells modulate the contractile properties of fibroblasts in the lymph node metastatic niche.
Factors secreted by dedifferentiated melanoma cells with a MITFlow AXLhigh signature inhibit LN-F contractility
Melanoma cells harbor transcriptional states ranging from melanocytic (MITFhigh AXLlow) to dedifferentiated (MITFlow AXLhigh; refs. 2–5). To compare the ability of these melanoma subpopulations to modulate LN-F contractility, we selected cell lines as well as short-term cultures of melanoma patients, either melanocytic (501Mel, MeWo, SK-MEL-28, WM164 and MM001, MM074) or dedifferentiated (1205Lu, WM793, WM2032, SBcl2 and MNC1, MM029, MM099), whose expression of MITF and AXL genes was analyzed by qRT-PCR (Supplementary Fig. S2A). Then, to model melanoma distant reprogramming of FRC in the premetastatic lymph node, we treated LN-F with CM harvested from melanoma cell cultures and containing the factors secreted by melanoma cells. Tumor-cell–derived factors, like TGFβ, are known to induce fibroblast actomyosin contraction, as during CAF differentiation (24). Strikingly, CM secreted by dedifferentiated melanoma cells drastically inhibited the spontaneous ability of LN-F to contract collagen gels, while CM from melanocytic melanoma cells had no effect (Fig. 2A). The ability of melanoma cells to inhibit LN-F contractility was indeed strongly inversely correlated to their MITF expression (Fig. 2B). To get a more comprehensive view of the effect of factors secreted by melanocytic or dedifferentiated melanomas on LN-F reprogramming, we carried out the study with two representative melanoma cell lines displaying the melanocytic MITFhigh AXLlow (501Mel) or the dedifferentiated MITFlow AXLhigh (1205Lu) phenotypes. Consistently, LN-F contractility was inhibited by CM of 501Mel cells silenced for MITF expression by siRNA and pushed toward a dedifferentiated phenotype with increased AXL expression, compared with control 501Mel cells (Fig. 2C; Supplementary Fig. S2B). These results were further validated by AFM analysis of collagen gels remodeled by LN-F. Indeed, gels treated with 1205Lu CM were softer than those treated with 501Mel CM, with similar stiffness to those treated with Y-27632 (Fig. 2D). Interestingly, LN-F embedded in collagen gels recovered their spontaneous contractility upon 1205Lu CM withdrawal, demonstrating that melanoma cell-induced inhibition of LN-F contraction was a reversible process (Fig. 2E).
During an immune challenge, FRC activation is associated with proliferation, upregulation of markers of fibroblast activation such as PDPN, and disruption of the RHOA-ROCK–mediated actomyosin cytoskeleton contraction due to CLEC2 interaction with PDPN (18, 19, 42). Similarly, the 1205Lu CM stimulated LN-F proliferation and increased PDPN cell surface expression whereas 501Mel CM had no effect (Fig. 2F and G). qRT-PCR analysis showed that expression of several markers of fibroblast activation like PDPN, FAP, CXCL12, LIF, and TNC was also upregulated by 1205Lu CM while S100A4 (FSP1) and ACTA2 (αSMA) levels were not affected (Fig. 2H). MS analysis also revealed that 1205Lu-activated LN-F secreted more ECM proteins (TNC, FN) and cytokines (IL6), validating the functional activation of LN-F (Fig. 2I). These results indicate that the inhibition of FRC contractility by dedifferentiated melanoma CM is a functional response of activated LN-F, like the FRC behavior observed during immunization. Importantly, our observations suggest that factors secreted by dedifferentiated and melanocytic melanoma cells do not share the same abilities to reprogram LN-F.
The propensity of fibroblasts to contract a collagen gel is regulated by forces generated by the actomyosin cytoskeleton (25, 26). We next examined by IF the F-actin organization and phosphorylation of the ROCK substrate MLC2 in LN-F treated with 501Mel CM or 1205Lu CM. The ROCK inhibitor Y-27632 was used as a control (Fig. 3A and B). We observed similar changes related to inhibition of actomyosin contraction (26) in LN-F treated with Y-27632 or 1205Lu CM, like decreased MLC2 S19 phosphorylation, less F-actin fibers, and the remodeling of the cell morphology from a stellate to a fusiform shape.
To identify the molecular pathways of LN-F contractility regulated by dedifferentiated melanoma CM, we performed microarray-based gene expression profiling on LN-F cultured in control conditions or exposed to 1205Lu CM (Fig. 3C). From the 25,000 most expressed LN-F genes, we selected differentially expressed genes (DEG) with an absolute log—fold change (logFC) ≥ 0.5. We identified 1,113 DEGs upregulated in 1205Lu CM-treated LN-F and 1,110 DEGs downregulated, validating LN-F transcriptional reprogramming by dedifferentiated melanoma cells. Gene Set Enrichment Analysis (GSEA) of the 1,110 downregulated genes validated the inhibition of pathways related to actin cytoskeleton polymerization, RHOA GTPase activity, and pointed towards the regulation of YAP and its cotranscription factor TEAD2 in 1205Lu-reprogrammed LN-F (Fig. 3D; Supplementary Fig. S3). GSEA plots also revealed that 1205Lu-treated LN-F converted from a myofibroblastic CAF signature to an inflammatory CAF signature (Fig. 3E).
Suppression of LN-F contractility by factors secreted by dedifferentiated melanoma cells is associated with impaired YAP activity
To validate the inhibition of the YAP pathway identified by GSEA (Figs. 3D and 4A), we investigated the effect of melanocytic and dedifferentiated melanoma cell CM on YAP nuclear localization and activation. Because YAP is sensitive to the cell microenvironment stiffness (27), its modulation by melanoma CM was investigated on hydrogels recapitulating the range of stiffness of LN-F–remodeled collagen gels (0.25 to 2.5 kPa; Fig. 2E). In agreement with the inhibition of LN-F contractility induced by dedifferentiated melanoma cues, IF analysis on 2.8 kPa hydrogels revealed that LN-F incubated with CM from dedifferentiated 1205Lu, WM793, or WM2032 cells exhibit an elongated shape, with more cytosolic YAP compared with control LN-F (Fig. 4B–D). In contrast, CM of 501Mel or SK-MEL-28 melanocytic cells did not induce any significant changes. Consistently, the expression of YAP target genes CYR61, CTGF, and SDPR was decreased in LN-F treated with 1205Lu CM (Fig. 4E), as was the expression of genes regulated by the YAP cotranscription factor TEAD2 (Fig. 4F). Although less pronounced, similar results were obtained on 0.2 kPa hydrogels (Fig. 4G–I).
To address the contribution of YAP in LN-F contractility, YAP expression was silenced using a siRNA approach (Fig. 4J). The spontaneous LN-F–mediated collagen gel remodeling was suppressed by YAP knockdown, revealing that YAP is not only a marker of cell contraction, but actively controlled LN-F contractility. However, YAP depletion had no effect on LN-F proliferation (Supplementary Fig. S4A). Together, our data suggest that relaxation of the FRC actomyosin network induced by factors secreted by dedifferentiated melanoma cells is associated with inhibition of YAP-TEAD2 transcriptional activity.
The JAK1-STAT3 pathway is inhibited by secreted factors from dedifferentiated melanoma cells and is involved in basal LN-F contraction
Next, we sought to determine signaling pathways linking actomyosin cytoskeleton relaxation and YAP inhibition induced by melanoma secreted factors in LN-F. Previous studies demonstrated that actomyosin contraction of FRC was driven by the interaction of PDPN with proteins from the ERM family (21, 22). Binding of PDPN to its ligand CLEC2 dismantled the PDPN–ERM interaction and inhibited ERM phosphorylation and PDPN-driven actomyosin contraction (18, 19). However, our data excluded the possible contribution of the CLEC2–PDPN–ERM pathway in melanoma-induced LN-F relaxation. Indeed, CLEC2 was not detected in 1205Lu cells and ERM phosphorylation was not affected in 1205Lu CM-treated LN-F (Fig. 5A). In agreement with LN-F activation (Fig. 2G) and the concomitant inhibition of F-actin stress fiber formation observed in presence of 1205Lu CM (Fig. 3A and B), 1205Lu CM increased PDPN expression, and reduced αSMA protein levels compared with 501Mel CM treatment or control LN-F.
Because TGFβ and IL6 family cytokines are known to regulate ROCK-mediated actomyosin contraction in myofibroblasts and CAF (24, 28, 29), we next investigated if the TGFβR1–SMAD and GP130—JAK–STAT pathways were also involved in LN-F contractility. The JAK1/2 inhibitor ruxolitinib strongly inhibited the spontaneous contraction of LN-F isolated from 4 different donors while the TGFβR1 inhibitor SB431542 had no effect (Fig. 5B). However, ruxolitinib, SB431542, or Y-27632 had no effect on LN-F proliferation (Supplementary Fig. S4B). Importantly, immunoblot analysis revealed that ruxolitinib inhibited both STAT3 and MLC2 phosphorylation, linking the JAK-STAT pathway to actomyosin contraction in LN-F (Fig. 5C). These results demonstrated that, beyond the CLEC2—PDPN–ERM pathway, LN-F contractility could also be controlled by the JAK-STAT pathway, but not by the TGFβR1–SMAD pathway. Indeed, although STAT3 phosphorylation was increased early by 1205Lu CM, probably because of the high level of cytokines secreted by 1205Lu cells compared with 501Mel cells (36), it was maintained inhibited over the following days (Fig. 5D). Sustained inhibition of STAT3 phosphorylation was also observed after a 96-hour treatment with CM from the dedifferentiated WM2032 cell line or the short-term MM099 cells, but not after treatment with CM from the melanocytic short-term MM074 cells (Fig. 5E). The JAK-STAT3 pathway was thus specifically and consistently inhibited by factors secreted by dedifferentiated melanoma cells. Furthermore, both 1205Lu CM and ruxolitinib inhibited RHOA activity in LN-F, suggesting that suppression of actomyosin contraction induced by 1205Lu CM was mediated through RHOA inhibition by the JAK-STAT pathway (Fig. 5F). Next, we identified JAK1 as the main JAK involved in LN-F contraction by showing that siRNA depletion of JAK1 (siJAK1), but not JAK2, inhibited LN-F–mediated collagen gel contraction (Fig. 5G). Then, we validated that STAT3 depletion by siRNA inhibited LN-F collagen gel contraction, suggesting that JAK1 controlled LN-F contractility through STAT3 (Fig. 5H). JAK1 or STAT3 siRNAs had no effect on LN-F proliferation (Supplementary Fig. S4A). GSEA analysis of the 2,223 most regulated transcripts from the microarray revealed that several transcripts strongly upregulated in 1205Lu CM-treated LN-F were indeed related to the negative regulation of the JAK-STAT pathway, such as MIR146A, Suppressor of Cytokine Signaling (SOCS) family members, and several phosphatases (Fig. 5I). Together, our results highlight the contribution of the JAK1-STAT3 pathway in the basal LN-F contraction and its inhibition by melanoma dedifferentiated cells.
The JAK1-STAT3 pathway controls YAP activity and the actin cytoskeleton polymerization
We next questioned whether JAK1-STAT3 inhibition impacted on YAP function and actomyosin cytoskeleton remodeling in LN-F. JAK1 silencing by siRNA (siJAK1) induced YAP cytosolic relocation and inhibited the formation of actin filaments (Fig. 5J–L). Interestingly, similar modifications were observed after STAT3 depletion (siSTAT3), suggesting that JAK1 effects on YAP and actomyosin were mediated by STAT3. Accordingly, expression of the YAP target genes CTGF and CYR61 was also inhibited after STAT3 silencing (Fig. 5M). Similar findings were observed with ruxolitinib treatment, suggesting that the JAK1-STAT3 pathway is essential for maintaining nuclear YAP and actomyosin network tension (Supplementary Fig. S4C–S4F). After deciphering the signaling pathway modulated by dedifferentiated melanoma cells to inhibit LN-F contraction, we turned our attention towards the premetastatic reprogramming of LN-F contractile properties in vivo.
Factors secreted by dedifferentiated melanoma cells inhibit murine LN-F contractility in vivo and decrease lymph node stiffness
The premetastatic modulation of LN-F contractile properties was analyzed in two mouse models by tracking the cellular localization of YAP in LN-F of draining lymph nodes. On the one hand, CM of dedifferentiated 1205Lu cells and melanocytic 501Mel cells were respectively injected into each ear of nude mice (Fig. 6A). In this premetastatic model, CM from dedifferentiated 1205Lu cells induced cytosolic translocation of YAP in LN-F compared with CM from melanocytic 501Mel cells (Fig. 6B and C). On the other hand, CM of dedifferentiated murine melanoma cells YUMM1.7 and control medium were respectively injected into each ear of syngeneic C57Bl/6j mice (Fig. 6D). Analysis of Mitf and Axl expression by YUMM1.7 cells is shown in Supplementary Fig. S5. In this syngeneic premetastatic model, CM from dedifferentiated YUMM1.7 melanoma cells induced cytosolic translocation of YAP in LN-F compared with control LN-F (Fig. 6E and F). In addition, YUMM1.7 CM induced swelling of the draining lymph nodes compared with the contralateral lymph nodes, and AFM analysis revealed that YUMM1.7 CM-draining lymph nodes were softer than control lymph nodes (Fig. 6G). These findings suggest that YAP-dependent inhibition of LN-F contractility is accompanied by changes in lymph node biomechanical properties.
Cytokines IL1α and IL1β secreted by dedifferentiated melanoma cells inhibit LN-F contractility
The next step was to identify the nature of the tumor cues inhibiting LN-F contractility. The CM contains all factors secreted by melanoma cells, including soluble proteins and lipids and EVs such as exosomes. Although EVs are known to promote lymph node niche formation and distal metastatic tumor growth (12), we identified that LN-F contractility was not modulated by melanoma EVs but rather by soluble proteins (Supplementary Fig. S6A–S6C). We thus focused on proteins oversecreted by 1205Lu cells compared with 501Mel cells, as previously identified by MS and antibody array analysis (Fig. 7A; refs. 35, 36). Candidate factors tested in gel contraction assays were restricted to secreted proteins present in the signature of dedifferentiated melanoma cells (Supplementary Fig. S6D; ref. 8), and found overexpressed by qRT-PCR in dedifferentiated cell lines 1205Lu, WM793, WM2032 compared with melanocytic cell lines 501Mel, MeWo, SK-MEL-28 (Supplementary Fig. S6E). Following this strategy, we identified that IL1α and IL1β effectively inhibited LN-F contractility (Fig. 7B), whereas IL6 or IL8 had no effect (Supplementary Fig. S6F). Both inflammatory cytokines bind to the IL1R1 receptor that is highly expressed by LN-F (43). ELISA analysis of melanoma CM confirmed that IL1α and IL1β were more secreted by dedifferentiated cells than by melanocytic cells (Fig. 7C) and that the ability of melanoma cells to inhibit LN-F contractility was strongly correlated with the amount of IL1α and IL1β secreted (Fig. 7D). As CM from dedifferentiated melanoma cells, IL1α or IL1β inhibited STAT3 phosphorylation in LN-F (Fig. 7E). Furthermore, IL1A and IL1B expressions were significantly inversely associated with MITF expression in patients with skin cutaneous melanoma (The Cancer Genome Atlas dataset; Fig. 7F). Our results thus show that IL1α and IL1β cytokines secreted by dedifferentiated melanoma cells inhibit LN-F contractility.
LN-F reprogrammed by dedifferentiated melanoma cells promote melanoma cell invasiveness
To understand the role played by inhibition of LN-F contractility by dedifferentiated melanoma cells in metastatic progression, melanoma cells were first cultured on confluent monolayers of contracted or relaxed LN-F. The proliferation and 2D migration of 1205Lu fluorescent cells (1205Lu Red), monitored by real-time microscopy, were not affected by reprograming of LN-F with 1205Lu CM, IL1β, or Y-27632 (Supplementary Fig. S7A and S7B), indicating that LN-F relaxation did not modulate cancer cell proliferation or migration. Next, 3D invasion of 1205Lu Red cells into LN-F remodeled collagen gels was analyzed by confocal microscopy (Fig. 8A and B). Whereas 40% to 60% of the cancer cells browsed 30 µm in gels remodeled by relaxed LN-F treated with 1205Lu CM, IL1β, or Y-27632, less than 20% did so in gels remodeled by contracted LN-F treated with control medium or 501Mel CM (Fig. 8C). In parallel, the z-distance covered by the 5% more invasive 1205Lu Red cells was increased from 40 μm–70 µm to 100 μm–130 µm between collagen gels remodeled by relaxed LN-F compared with resting LN-F (Fig. 8D). These results indicate that the inhibition of LN-F contractility driven by dedifferentiated melanoma cells strongly enhances 3D cancer cell invasion.
Our study provides evidence that human FRC are contractile cells like murine FRC (18–20). Quiescent FRC display a myofibroblast-like phenotype along with high expression of PDPN, FAP, and αSMA and thus share many properties with CAF. These CAF hallmarks suggest that FRC could play a tumor supportive role in the lymph node metastatic microenvironment. Indeed, PDPN was identified as a CAF marker in a variety of malignancies and is associated with metastasis and poor prognosis (44, 45). PDPN expression on CAF favors force-mediated matrix remodeling through the activation of the RHO‐ROCK pathway and promotes cancer cell invasion (45). In patients with breast cancer, metastatic lymph nodes are enriched in a CAF subpopulation inducing cancer cell invasion and exhibiting similar markers (PDPNhigh, FAPhigh, αSMAhigh, PDGFR-βhigh) to FRC (46), suggesting that lymph node CAF mostly originated from resident FRC.
Focusing on the signaling pathway(s) driving FRC actomyosin contractility, we provide evidence that the spontaneous contractility of quiescent FRC relies on a basal level of YAP and JAK1-STAT3 activation, and not only on the PDPN-ERM pathway as previously identified in murine FRC. YAP activation is known to reflect the actomyosin contractile state of FRC (19) and CAF (26), and to regulate FRC differentiation during lymph node development (47). We show here that YAP is not only a marker but also controls actomyosin contractility of differentiated human FRC. YAP and the RHO-ROCK pathway are intimately linked, regulating each other, and controlling the actomyosin cytoskeleton plasticity (26, 27). We also disclose that inhibition of JAK1 or STAT3 reduces FRC contractility and leads to inhibition of RHOA and YAP. Interestingly, our MS analysis suggests that basal JAK1-STAT3 signaling in resting human FRC could be due to autocrine secretion of IL6 (Fig. 2I). Previous studies have shown that JAK1-STAT3 signaling increases ROCK-mediated actomyosin contractility in CAF, and that ROCK signaling induces STAT3 phosphorylation and transcriptional responses. Thus, JAK1-STAT3 and RHO-ROCK are interdependent and cross-regulate each other (29). Collectively, our results support the notion that the RHO-ROCK–driven contractility of human FRC is not only controlled by the PDPN-ERM pathway, but also by signaling pathways shared with CAF.
Tumor cells secrete growth factors and inflammatory factors, such as TGFβ or IL6 family cytokines, which trigger fibroblast activation and CAF transformation, both of which are associated with increased actomyosin contractility. Strikingly, dedifferentiated melanoma cells, but not melanocytic melanoma cells, secrete factors drastically suppressing force-driven collagen gel remodeling by FRC. Factors secreted by dedifferentiated melanoma cells inhibit JAK1-STAT3 signaling, which decreases RHO–ROCK–MLC2 signaling and YAP activity. During an immune challenge, FRC relaxation is associated with FRC activation (18, 19). Similarly, FRC stimulated with CM from dedifferentiated melanoma cells show an activated phenotype: they proliferate more, upregulate markers of activation, and increase the secretion of extracellular matrix proteins and cytokines. Interestingly, both BRAF-mutated (1205Lu, WM793, MM099) and NRAS-mutated (WM2032, SBcl2) dedifferentiated melanoma cells were able to inhibit FRC contractility. The question remains whether tumor-derived factors from other cancers that spread to the lymph node, such as breast cancer, can exert a similar task on FRC.
Culture of FRC with factors secreted by dedifferentiated melanoma cells strikingly phenocopies the prevention of PDPN signaling by the provision of CLEC2, antibody blockade, or genetic deficiency (18, 19). It induces similar morphologic changes with an elongated cell shape, decreased stress fibers, less YAP nuclear activation, but also increased proliferation and PDPN expression. So, while we have excluded the modulation of ERM phosphorylation by melanoma CM, the implication of PDPN in the inhibition of FRC contractility mediated by melanoma CM is still an open question. Interestingly, PDPN can also activate the FRC actomyosin machinery independently of its binding to ERM proteins, by engaging a neighboring transmembrane protein (19). This unknown transmembrane protein may drive JAK1 signaling and binding to its ligand secreted by dedifferentiated melanoma cells could dissociate such interaction. In light of our results, the IL1 receptor IL1R1 might be an interesting candidate as it is highly expressed in FRC (43). Alternatively, JAK1 may directly bind to the cytoplasmic tail of PDPN through its N-terminal FERM domain (48). PDPN controls a wide range of physiologic effects, such as contractility, migration, proliferation, or differentiation. Multiple molecular mechanisms of PDPN regulation are therefore likely in various cell types (45).
It is established that factors secreted by melanoma cells, including TGFβ, growth factors, or proinflammatory molecules promote tumorigenesis (49) and lymph node metastasis (13). Melanoma EVs have been particularly described as potent inducers of lymph node premetastatic niches (12), but did not modulate FRC contractility in our experimental setting. Analyses of factors preferentially secreted by dedifferentiated melanoma cells (35, 36) and inhibiting LN-F contractility pointed towards the IL1 inflammatory cytokine, although the participation of other tumoral factors cannot be excluded. IL1 plays an important role in tumor progression and metastasis (50) and is detected at high levels in the serum of patients with melanoma (51). Interestingly, FRC express high levels of IL1R1 and respond rapidly to systemic IL1β stimulation in vivo (43). Furthermore, IL1 inhibits the contractility of lung fibroblasts (52–54). The link between IL1 and inhibition of the JAK1-STAT3 pathway needs further investigation, but it may involve activation of the p38 MAPK, as in synovial fibroblasts (55).
Lymph nodes are less stiff and more deformable following antibody blockade of PDPN and FRC contractility (19). Although other factors, such as lymph transport and immune cell recruitment, participate in modulating the biomechanical properties of the premetastatic lymph node (15, 16, 31), our results in vivo validate that FRC contractility impacts lymph node stiffness and show that it is controlled by factors secreted by dedifferentiated melanoma cells. Melanoma-induced stromal reprogramming in the tumor-draining lymph node has been analyzed in other studies using the murine cell line B16.F10 (16, 31). B16.F10 cells do not harbor classic human melanoma mutations (BRAF, NRAS, or NF1) and are known to express high levels of MITF despite their metastatic potential. In contrast to our results with melanocytic MITFhigh melanoma cells, these studies reported the enhanced capacity of FRC treated with B16.F10 CM to contract collagen gels, the increased RHO signaling in murine FRC isolated from B16.F10-draining lymph nodes and increased stiffness and intranodal pressure of B16.F10-draining lymph nodes. The reason behind this discrepancy remains unclear but suggests that B16.F10 cells may control FRC contractility differently from human melanocytic melanoma cells. These results underline that metastatic disease development is a highly complex process fueled by intratumoral heterogeneity and mutational background.
Plasticity of melanoma cells is considered a key driver in the development of the disease and cellular phenotypes drive individual steps of melanoma progression (2–5). The phenotype of dedifferentiated cells is associated with increased invasive and metastatic capabilities. Our findings further suggest that the dedifferentiated state, in contrast to the melanocytic state, allows melanoma cells to reprogram the fibroblastic stroma of the lymph node prior to metastatic dissemination, and that this contributes to lymph node tumor invasion. FRC contractile phenotype regulates lymph node swelling and tunes lymph node immunity (18, 19). Furthermore, microenvironment clues such as tissue stiffness regulate the properties of tumor cells. Our data therefore suggest that significant changes in lymph node stiffness and FRC network microstructure could modulate the metastatic capabilities of tumor cells, but also the antitumor immune response. Our data show that relaxation of the FRC three-dimensional network in collagen gels increases the invasive capabilities of melanoma cells. Thus, the reprogramming of the biomechanical and immune properties of the lymph node by the loss of FRC contractility could be essential for survival and development of metastatic cells in the lymph node niche.
In conclusion, we identified that factors secreted by dedifferentiated melanoma cells, such as IL1, reprogram the functions of FRC from the tumor-draining lymph node and we deciphered the underlying signaling pathways involved in human FRC cytoskeleton relaxation. Our work illustrates that FRC activation and actomyosin relaxation in the lymph node might be a prognostic marker of melanoma invasive potential, suggesting that the microarchitecture of the FRC reticular network and PDPN expression could be examined in the sentinel tumor-draining lymph node at diagnosis. Our data also reinforce the rational for clinical trials combining anti-IL1 strategies with checkpoint blockade immunotherapy (50). Blocking remote communication between dedifferentiated melanoma cells and the FRC may inhibit the formation of a permissive biomechanical and immunologic lymph node niche, thereby impinging on lymphatic metastasis.
I. Berestjuk reports personal fees from Ligue Contre le Cancer during the conduct of the study. S. Diazzi reports personal fees from Fondation pour la Recherche Médicale (FRM) during the conduct of the study. T. Passeron reports personal fees from Abbvie, Pfizer, Eli Lilly and Company, and personal fees from Incyte outside the submitted work; in addition, T. Passeron is the cofounder of YUKIN Therapeutics. V. Prod'homme reports grants from Canceropôle Provence Alpes Côte d'Azur, Fondation ARC, Ligue Contre le Cancer, Institut National du Cancer, and grants from National Research Agency during the conduct of the study. No disclosures were reported by the other authors.
C. Rovera: Conceptualization, data curation, formal analysis, validation, investigation, visualization. I. Berestjuk: Methodology. M. Lecacheur: Methodology. C. Tavernier: Formal analysis, validation, investigation. S. Diazzi: Data curation, investigation. S. Pisano: Data curation, formal analysis, investigation. M. Irondelle: Methodology. A. Mallavialle: Data curation, formal analysis, validation, investigation. J. Albrengues: Methodology. C. Gaggioli: Methodology. C.A. Girard: Conceptualization, writing–review and editing. T. Passeron: Resources. M. Deckert: Conceptualization, supervision, funding acquisition, validation, methodology, project administration, writing–review and editing. S. Tartare-Deckert: Conceptualization, supervision, funding acquisition, validation, methodology, writing–original draft, project administration, writing–review and editing. V. Prod'homme: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.
This work was funded by Institut National de la Santé et de la Recherche Médicale (Inserm), Université Côte d'Azur (Nice, France), Canceropôle Provence Alpes Côte d'Azur (emergence grant to V. Prod'homme), Fondation ARC (projet fondation ARC to V. Prod'homme), Ligue Contre le Cancer (equipe labellisée to S. Tartare-Deckert), and Institut National du Cancer (INCA_12673 to S. Tartare-Deckert). Funding from Agence Nationale de la Recherche (ANR-18-CE14-0019-01 to M. Deckert), ITMO Cancer Aviesan within the framework of the Cancer Plan and the French Government through the ‘’Investments for the Future’’ Labex SIGNALIFE (ANR-11-LABX- 0028-01) is also acknowledged. I. Berestjuk is a recipient of a doctoral fellowship from Ligue Contre le Cancer and S. Diazzi is a recipient of a doctoral fellowship from FRM. The authors thank J.-C. Marine and G. Ghanem for short-term cultures of melanoma cells, M. Bosenberg for YUMM1.7 cells, and the Microscopie Imagerie Côte d'Azur (MICA) platform supported by the GIS IBiSA, the Conseil Départemental 06, and the Région Provence-Alpes-Côte d'Azur. The authors also acknowledge the HistoC3M histology facility funded by the Canceropôle Provence-Alpes-Côte d'Azur and the AFM core facility supported by the Fondation ARC, the GIS IBiSA, and the Conseil Général 06 de la Région Provence-Alpes-Côte d'Azur.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.