Regional Activation of Myosin II in Cancer Cells Drives Tumor Progression via a Secretory Cross-Talk with the Immune Microenvironment

1Metastasis accounts for >90% of cancer-related deaths, indicating an urgent need for clinical management. To leave the primary tumor, cancer cells disseminate using different migration modes. Although collective cell migration is important for tissue remodeling, single-cell migration (rounded-amoeboid or elongated-mesenchymal) allows transport, both locally and to distant sites along with invasion through basement membranes. Actomyosin contractility driven by Myosin II controls cytoskeletal remodeling and tumor dissemination. ROCK can directly phosphorylate myosin light chain 2 (MLC2) or indirectly decrease Myosin phosphatase (MYPT) activity increasing MLC2 phosphorylation. ROCK can also activate LIMK, which phosphorylates and inactivates cofilin resulting in F-actin stabilization. High levels of actomyosin contractility driven by Myosin II are key to sustain amoeboid bleb-based migration. Intravital imaging in melanoma and breast cancer mouse xenografts revealed amoeboid migration is favored in the tumor invasive fronts (IFs).

2On the other hand, cancer-associated inflammation promotes tumorigenesis at many levels. Inflammation is enabled by the secretion of multiple factors and the recruitment of immune cells, like monocytes. Monocytes differentiate to macrophages, which can change their phenotypes responding to microenvironmental signals. Classically activated macrophages are induced in response to pro-inflammatory stimuli, such as lipopolysaccharides (LPS) or interferon gamma (IFN-γ) and exhibit cytotoxic functions. Cytokines such as interleukin (IL)-4, IL-13, and IL-10, transforming growth factor beta (TGF-β), and/or glucocorticoids can support alternatively activated macrophages (AAMs) that promote tissue repair and tumor progression.

3Here, we investigated how Myosin II activity in cancer cells controls the secretion of factors regulating the tumor microenvironment (TME) via the establishment of a cross-talk with pro-inflammatory nuclear factor κB (NF-κB).

4Using intravital imaging in xenograft melanoma models, we have previously reported an enrichment in amoeboid migration in the invasive edge of tumors. To test whether this was recapitulated in human patient tissues, we evaluated human melanoma biopsies using a tissue microarray (40 human melanoma lesions, cohort A) and a smaller cohort (7 melanoma lesions, cohort B). For both cohorts, matched tumor body (TB) and IF were included. IFs of human primary melanomas were found enriched in rounded cancer cells independently of the morphology in the TB (Figure 1A). Importantly, we found a regional increase in phosphorylated MLC2 (p-MLC2) levels in the IF (Figure 1B) indicative of high Myosin II activity. Our data suggest that the combination of roundness and increased p-MLC2 can define the amoeboid contractile cancer phenotype accurately in patients’ biopsies.

5Interestingly, in both of our patient tissue cohorts, macrophages were a prominent population, in accordance with previous studies reporting that macrophages constitute up to 30% of the immune infiltrate in melanoma. Furthermore, high tumor-associated macrophage (TAM) infiltration has been associated to poor prognosis. We thus sought to identify regional differences in macrophage composition in patient cohort A. CD68 is a pan-macrophage marker, while CD163 and CD206 typify AAMs. We found no regional difference in CD68+ cells (Figure 1C). However, both CD163+ and CD206+ TAMs were enriched in the IFs of tumors (Figures 1D and 1E). TAMs support angiogenesis, while they are located in perivascular tumor areas. We observed elevated vessel density in the same tumor regions where macrophages were abundant (Figure 1F). Similar results were observed in cohort B (Figures S1A–S1D).

6The TME in aggressive tumors may favor the formation of pre-metastatic “invasive niches” composed of cancer cells, endothelial cells, and macrophages. Indeed, metastatic melanoma lesions were enriched in both rounded melanoma cells (Figure 1G) and CD206+ TAMs (Figures 1H, S1E, and S1F) in proximity to blood vessels (Figure 1I). Using the Cancer Genome Atlas (TCGA) database, we found increased CD206 mRNA levels in metastatic versus primary human melanomas (n = 384) (Figure 1J). Furthermore, using Gene Expression Omnibus (GEO) database, we found a positive correlation between CD206 and CD31 mRNA levels in melanoma patients (n = 322) (Figure S1G). These data support the notion that these non-cancerous cellular components are upregulated in human melanoma.

7Overall, the IFs of human melanomas are enriched in amoeboid melanoma cells, which are associated with a specific TME, the amoeboid-associated TME (AATME). Importantly, the TME found in metastatic sites mirrors the TME found in the IFs of melanomas, that is, the AATME (Figure 1K).

8Tumor cell-normal cell communication can be mediated by secreted factors. A375M2 are highly metastatic rounded melanoma cells (∼90% rounded) with higher Myosin II activity (Figure 2A). A375M2 cells are derived from poorly metastatic A375P more elongated melanoma cells (50% rounded, 50% elongated) with lower Myosin II activity compared to A375M2 cells (Figure 2A). Using a protein array consisting of 274 human chemokines, cytokines, growth factors, and matrix metalloproteinases, we found that 155 proteins were highly secreted by A375M2 cells compared to A375P cells (Figure 2B). These factors were sub-divided into 3 groups based on their fold change (Figure 2B). A375M2 cells were shown to secrete high levels of cytokines, such as IL-3, IL-4, IL-5, and IL-13. The amoeboid-melanoma secretome appears to be skewed toward a pro-inflammatory signature typically associated with tumor progression (Figure S2A). We confirmed by ELISA that A375M2 cells secreted high levels of pro-inflammatory IL-1α, IL-8, and immunosuppressive IL-10 and TGF-β (Figure 2C). To expand our observations to the clinical setting, GEO (n = 421) and TCGA (n = 354) databases were used to evaluate mRNA levels of some highly secreted factors by A375M2 cells. IL-1α, IL-10, TGF-β, IL-8, and IL-4 mRNA were all upregulated during melanoma progression with a significant increase in metastatic compared to primary human melanomas (Figure S2B) suggesting transcriptional regulation.

9We confirmed that protein secretion was Myosin II dependent since A375M2 cells depleted from MLC2 secreted significantly less cytokines and chemokines (Figures 2D and S2C). ROCK is a key regulator of Myosin II activity, and, as such, A375M2 cells depleted from ROCK1/2 via RNAi had decreased protein secretion (Figures 2E and S2D). These results were confirmed using 3 ROCK inhibitors (ROCKi H1152, Y27632, GSK269962A) (Figures 2F and S2E). Moreover, inhibition of LIMK downstream of ROCK in A375M2 cells resulted in reduced: p-cofilin levels (Figure S2F), cell roundness (Figure S2G), p-MLC2 levels (Figure S2H), and IL-8 secretion (Figures S2I). These data indicate that perturbing actin and myosin dynamics has an impact on secretion.

10Our observations were further expanded to the matched melanoma paired cell lines WM983B (metastatic) /WM983A (primary) derived from the same patient. WM983B cells were more rounded with higher p-MLC2 levels compared to WM983A cells (Figure 2G, upper panels) and were found to be more secretory (Figure 2G, lower panels). On the other hand, WM88 elongated melanoma cells did not secrete any of the cytokines measured (Figure 2G). To further asses the role of ROCK-Myosin II, WM983B cells were depleted from MLC2 (Figure 2H, upper panel) or treated with a ROCKi (Figure 2I, upper panel) and were found to secrete significantly less cytokines (Figures 2H and 2I, lower panels). Similar results were observed when ROCK was inhibited in an additional melanoma cell line WM793B (Figure S2J).

11Overall, our data show that the secretion of immunomodulatory factors, which are important during melanoma progression, is regulated by ROCK-Myosin II activity in melanoma cells.

12Enrichment in amoeboid melanoma cells and TAMs in the IF of human melanomas (Figure 1) suggests a potential communication between cancer cells and macrophages. On the other hand, the secretome of amoeboid melanoma cells is rich in over 20 chemotactic factors (Table S1) known to trigger chemotaxis in monocytes, which are the precursors of macrophages. Indeed, migration of human peripheral blood mononuclear cell (PBMC)-derived monocytes and monocytic cell lines (THP-1 and U-937) was increased toward secreted factors (conditioned media [CM]) derived from amoeboid A375M2 versus more elongated A375P cells (Figure 3A). The timescales of this assay (2–4 h) suggest that increased chemotaxis was responsible for increased monocytic migration.

13We next addressed whether monocytes could be differentiated into macrophages in response to CM A375M2, as this media was rich in factors (macrophage colony stimulating factor [M-CSF], IL-4, IL-10, IL-13, TGF-β) that affect monocyte-macrophage commitment. We developed a spectrum of in-vitro-polarized macrophages (Figure 3B). M-CSF, IL-4, or IL-10 treatment showed increased CD163+CD206+ expression in macrophages, while HLA-DR+CD86+ macrophages were induced after IFN-γ&LPS (Figure 3B). Specifically, IL-4 induced CD206, IL-10 induced CD163, IFN-γ&LPS induced CD86 expression, while HLA-DR expression was similar across treatments (Figure S3A). Importantly, CM A375M2 induced CD163+CD206+macrophages more efficiently than CM A375P (Figures 3C–3E and S3B). CD206 expression was increased in macrophages induced by CM A375M2 (Figure 3E) and comparable to that after IL-4 stimulation (Figure S3A). Neither CM A375M2- nor CM A375P-treated macrophages showed difference in expression levels of classic activation markers: HLA-DR and CD86 (Figures 3F and S3C). Macrophages induced by CM A375M2, IL-4, M-CSF, or IL-10 retained a mixed morphology (Figures 3G and S3D) indicative of differentiation. Cell elongation, characteristic of AAMs, was profound in CM A375M2-treated macrophages, while minimal after IFN-γ&LPS treatment or in untreated monocytes (Figure 3G).

14To characterize the functional role of macrophages-induced by amoeboid melanoma cells, a tumor cell-killing assay was performed (Figure 3H). IFN-γ&LPS-treated macrophages killed both amoeboid A375M2 and elongated WM88 cells (Figure 3I). In contrast, macrophages induced by CM A375M2 supported tumor cell viability (Figure 3I). Such observations were expanded to a wider panel of melanoma cells (WM1366, WM793B, WM3854, and WM983A) (Figure S3E). To further validate the tumor-promoting role of macrophages-induced by amoeboid melanoma cells, in vivo experiments were performed using A375M2 and WM983B xenografts in SCID mice. Macrophage depletion after clodronate liposome delivery resulted in impaired in vivo tumor growth in both A375M2 and WM983B tumors (Figures S3F and S3G). Depletion of F4/80+CD206+ cells was confirmed (Figure S3H). These data suggest a key role for macrophages in mediating melanoma tumor growth.

15We next used serum derived from melanoma patients or healthy donors to treat PBMC-derived monocytes (Figure 3J). Melanoma-patient-derived sera induced higher levels of CD163+CD206+ macrophages compared to healthy-donor-derived sera, while HLA-DR and CD86 expression remained unchanged (Figures 3J–3L). Thus, melanoma-patient-derived secreted factors induce macrophage polarization comparable to amoeboid melanoma cells.

16Overall, our data show that amoeboid melanoma cells with high Myosin II activity can recruit monocytes, differentiate them into macrophages, and functionally educate them to support tumor growth.

17We next assessed whether intrinsically high Myosin II activity in melanoma cells has an impact on macrophage recruitment in vivo. A375M2-EGFP cells (amoeboid and higher Myosin II activity) or A375P-EGFP cells (more elongated and lower Myosin II activity) (Figures 4A and S4A) were injected subcutaneously into SCID mice. A375M2 tumors grew faster compared to A375P ones (Figure 4B). There was an increase in melanoma cell roundness at the IFs of all tumors (Figure S4B). We assigned scores from 0 (low intensity) to 4 (very high intensity) of phospho-MLC2 (Figures 4C, 4D, and S4C). Overall, Myosin II levels were higher in A375M2 tumors compared to A375P tumors in all areas (Figures 4C, 4D, upper panel, and S4C). Importantly, Myosin II activity was significantly increased in all the IFs (Figures 4C and 4D), while we detected the highest levels of Myosin II activity at the IF of A375M2 tumors (Figures 4C, 4D, and S4C) accompanied by the highest infiltration of F4/80+CD206+ TAMs in conjunction with vessel density (Figures 4E–4H). Similarly, WM983B-EGFP cells (amoeboid, higher Myosin II) and WM983A-EGFP cells (more elongated, lower Myosin II) (Figures 4I and S4D) were injected into SCID mice. WM983B tumors grew faster compared to WM983A ones (Figures 4J). Melanoma cell rounding was observed at the IF (Figure S4E), while WM983B tumors had the highest levels of Myosin II (Figures 4K, S4F, and S4G). Similarly, WM983B tumors had the highest F4/80+CD206+ TAM infiltration at the IFs, compared to WM983A tumors (Figures 4L and 4M). Notably, A375P versus A375M2 and WM983A versus WM983B showed similar proliferation rates in vitro (Figure S4H). Since SCID mice are deficient in B and T cells, amoeboid melanoma cells could induce macrophage polarization in vivo. Altogether, these data suggest that invasive amoeboid melanoma cells with high Myosin II activity educate macrophages to jointly support tumor growth.

18Next, we investigated whether amoeboid invasive melanoma cells could retain their secretory memory. Melanoma cells were isolated from the IFs and TBs of A375M2 tumors and cultured ex vivo (Figure S4I). Melanoma cells isolated from the IF of A375M2 tumors were more secretory (Figure S4J) and induced CD163+CD206+ macrophages more efficiently ex vivo when compared to their TB counterparts (Figure S4K).

19Moreover, to ensure that AATME can be generated even in the presence of all immune components, highly metastatic and amoeboid B16F10 cells were injected into immunocompetent C57BL/6J mice. We observed rounded melanoma cells (Figure S4L) with high Myosin II levels (Figure 4N) and increased F4/80+CD206+ TAMs at the IFs (Figures 4O and S4M). To closely recapitulate the physiological development of melanoma TME, we used an orthotopic model in which amoeboid 5555 cells—derived from a BRAFV600E mouse model—were injected intradermally into C57BL/6J mice. As in previous models, IFs of 5555 tumors were rich in rounded cancer cells (Figure S4N) with increased Myosin II activity (Figure 4P) and F4/80+CD206+ TAM infiltration (Figures 4Q and S4O). Overall, these results suggest that AATME is a conserved feature in several in vivo melanoma tumor models.

20We next investigated whether ROCK-Myosin II axis in amoeboid cancer cells is responsible for polarizing macrophages. Monocytes were treated with CM from ROCK-inhibited A375M2 cells using 3 different ROCK inhibitors (Figure 5A). ROCK-inhibited A375M2 cells could not induce CD163+CD206+macrophages (specifically CD206 expression) compared to control (Figures 5B, S5A, and S5B). HLA-DR and CD86 expression did not change (Figures S5C and S5D). In all cases, ROCK-inhibited A375M2 cells had lowered p-MLC2 levels (Figure S5E). Similarly, A375M2 cells were less efficient in inducing CD163+CD206+macrophages after MLC2 RNAi transfection (Figure 5C), while HLA-DR+CD86+ macrophages did not change (Figure S5F).

21TAMs can secrete immunosuppressive factors, such as TGF-β and IL-10. Macrophages induced by CM A375M2 secreted high levels of TGF-β and IL-10 (Figures S5G and S5H). Tumor necrosis factor (TNF)-α, a key feature of classically activated macrophages, was undetectable (data not shown). Furthermore, monocytes treated with melanoma-patient-derived sera differentiated into macrophages, which also secreted higher levels of IL-10 compared to monocytes treated with sera from healthy donors (Figure S5I), showing that this is a feature of melanoma-associated macrophages. Moreover, monocytes treated with CM from ROCK-inhibited A375M2 cells differentiated into macrophages that secreted significantly less TGF-β and IL-10 (Figures S5G and S5H). Of note, macrophages in all experimental conditions were competent for phagocytosis and sensitive to cytochalasin D (Figure S5J). These data confirm that ROCK-Myosin II activity in melanoma cells influences macrophage functional polarization.

22We have shown in Figure 1 that AAMs are found in proximity of blood vessels and AAMs can promote endothelial cell survival. Indeed, macrophages induced by CM A375M2 sustained endothelial cell growth in HMVECs (human microvascular endothelial cells) and HUVECs (human umbilical vessel endothelial cells) more efficiently than macrophages treated with CM from ROCK-inhibited A375M2 cells (Figures 5D). Overall, our data show that decreasing Myosin II activity in cancer cells leads to phenotypic and functional macrophage reprograming.

23To confirm the role of ROCK-Myosin II activity in controlling AATME in vivo, 5555 cells were pre-treated with ROCKi ex vivo and subsequently injected in the dermis of C57BL/6J mice without any further ROCKi treatment (Figure 5E). 8 days post-injection, we observed that control tumors generated AATME (Figures S5K and S5L). Importantly, ROCKi pre-treated 5555 cells were not able to increase Myosin II levels or recruit F4/80+CD206+ TAMs in the IFs of tumors as efficiently as controls (Figures S5K and S5L). 14 days post-injection, melanoma cells had invaded the dermis using amoeboid invasion (Figure 5F). At this time point, tumors in the ROCKi pre-treated group displayed pronounced loss of invading amoeboid melanoma cells (Figures 5G–5I) and a reduction in F4/80+CD206+ TAMs (Figure 5J). The two groups grew similarly in vitro, but ROCKi pre-treated group displayed a clear growth disadvantage in vivo (Figures S5M and S5N). These data suggest that Myosin II activity triggers AATME early on in tumorigenesis, while AATME is further supported during tumor development by amoeboid invasive melanoma behavior.

24To extend our observations to the clinical setting, in which drugs are administered systemically in established tumors, mice harboring A375M2-EGFP tumors were treated with ROCKi (Figures 5K–5N). Similar to the pre-treatment setting, reduced tumor growth after systemic ROCKi treatment (Figure 5K) was associated with loss of melanoma cell roundness (Figure 5L), decreased F4/80+CD206+ TAM infiltration (Figure 5M), and decreased vasculature (Figure 5N) in the IFs.

25Overall, we suggest that AATME composition in vivo is dependent on ROCK-Myosin II activity in melanoma cells located in the IF. Inhibition of ROCK-Myosin II in melanoma cells hinders the generation of tumor-supportive microenvironments.

26We have identified high levels of protein secretion as a characteristic of amoeboid melanoma cells. To assess whether protein secretion could affect Myosin II activity itself, A375P cells were treated with CM A375M2 and displayed increased: cell roundness (Figure 6A, upper panel), p-MLC2 levels (Figures 6A, lower panel, and S6A), and migration (Figure S6B). Thus, amoeboid melanoma cells exert paracrine effects toward other melanoma cells inducing amoeboid features via secretion. Next, A375M2 cells were treated with brefeldin A (BFA) to block trafficking of pre-stored soluble factors and receptors and displayed loss of cell roundness (Figure 6B, upper panel) and p-MLC2 levels (Figures 6B, lower panel, and S6C). Thus, protein secretion sustains amoeboid features in an autocrine manner.

27To understand which signaling pathways were linked to the amoeboid secretory phenotype, GeneGo MetaCore analysis was performed using the protein array data (Figure 2B). Amoeboid melanoma cells secrete factors that are part of a network centered on NF-κB (Figure 6C), while the second enriched network was centered on STAT3 (data not shown). NF-κB family members RELA (p65), RELB, and REL (c-Rel) contain transactivation domains. NFKB1 (p105) and NFKB2 (p100) encode longer proteins processed to the shorter DNA-binding forms p50 and p52. Phosphorylation of IκBα by the IKK complex enables dissociation from the NF-κB complex (p65/p50) with nuclear translocation of the latter. To test whether protein secretion mediates NF-κB activity, A375P cells were treated with CM from A375M2 cells and displayed increased p-IκBα levels (Figure 6D). Conversely, depletion of MLC2 in A375M2 cells led to decreased p-IκBα levels (Figures 6E and 6F) suggesting that Myosin II regulates NF-κB activity. Moreover, amoeboid A375M2 melanoma cells have intrinsically high NF-κB activity (p65 nuclear translocation) compared to more elongated A375P cells (Figure 6G). These data show that amoeboid cells sustain NF-κB activity in an autocrine and paracrine manner.

28We have shown that IL-1α and IL-8 secretion is regulated by ROCK-Myosin II (Figure 2), while NF-κB can be activated by either of these factors. We hypothesized that ROCK-Myosin II activity could regulate NF-κB activation via either IL-1α or IL-8. CM from A375M2 cells induced p65 nuclear translocation in A375P cells (Figures 6H and S6D, upper panel). However, blocking IL-1α, but not IL-8, in CM A375M2 abolished these effects in A375P cells (Figures 6H and S6D, upper panel). To further understand whether secreted IL-1α itself had a role in controlling amoeboid features, A375P cells were treated with CM A375M2 in which IL-1α had been blocked. Blocking secreted IL-1α led to loss of cell roundness induced by CM A375M2, whereas blocking IL-8 had no effect (Figures 6I and S6D, lower panel). These data show that Myosin II activity in amoeboid melanoma cells regulates NF-κB activation via secreted IL-1α and conversely, IL-1α perpetuates the amoeboid phenotype.

29We have shown that IL-1α supports both NF-κB activity and amoeboid phenotype. We hypothesized that NF-κB itself transcriptionally controls expression of secreted factors in CM A375M2 (Figure 6C) that induce an amoeboid phenotype (Figure 6A). A375M2 cells depleted from NFKB1 lost cell roundness and p-MLC2 cortical levels (Figures 6J, 6K, and S6E). Blockade of IKKβ activity led to inhibition of IκBα phosphorylation in A375M2 cells (Figure S6F) and an IKKβ inhibitor (IKKβi) yielded similar results as NFKB1 RNAi depletion (Figures S6G and S6H). Thus, IKKβ/NF-κB supports Myosin II activity in amoeboid melanoma cells generating a positive feedback loop. To further understand this signaling network, we studied the kinetics of this process. Cell roundness and p-MLC2 levels were decreased in a time-dependent manner using either ROCKi (Figures S6I–S6K) or IKKβi (Figures S6L–S6N). However, ROCK inhibition resulted in earlier changes compared to IKKβ inhibition, suggesting that NF-κB affects Myosin II activity indirectly at later time points via transcriptional mechanisms controlling cytokine expression, while ROCK regulates Myosin II activity directly.

30In summary, ROCK-Myosin II activity and NF-κB establish a positive feedback loop initiated by ROCK regulation of IL-1α and perpetuated and amplified by IL-1α/IKKβ/NF-κB supporting amoeboid behavior in return (Figure 6L).

31We have shown that melanoma cells acquire amoeboid features that can be perpetuated by secretion via a cross-talk with NF-κB. Over 20 chemokines (Table S1) and several cytokines secreted by amoeboid cells were connected to NF-κB (Figure 6C) pointing at this transcription factor as a master regulator of the effects exerted by amoeboid cells on TME. In fact, NFKB1 depletion in A375M2 cells resulted in CM with decreased chemotactic potential toward monocytic cells (Figure S7A). Moreover, depletion of NFKB1 in amoeboid melanoma cells reduced Myosin II activity (Figure 7A, left panels), secretory potential and macrophage polarization (Figure 7A, right panel). These effects were comparable to MLC2 depletion via RNAi (Figure 7A). These data show that ROCK-Myosin II regulates macrophage polarization mainly via cytokines connected to NF-κB network (Figure 6C). Importantly, MLC2 and NFKB1 depletion in melanoma cells resulted in decreased CD206 in macrophages (Figure S7B). Among the cytokines in our panel, IL-4 induced the highest levels of CD206 expression (Figure S3A). Importantly, we could rescue the defects in macrophage polarization after NFKB1 or MLC2 depletion upon addition of IL-4 (Figure 7A, right panel). Therefore, the cross-talk between ROCK-Myosin II and NF-κB helps melanoma cells at the IF polarize macrophages via secreted factors.

32Once amoeboid cells acquire high levels of Myosin II at the IF, they invade the dermis (Figures 5F and 5G). Such cancer cells could successfully colonize distant metastatic sites, where they need to extravasate. At this stage, endothelial cell-cancer cell communication is crucial. Since the amoeboid secretome is rich in factors controlling vascular permeability (Figure 2B), we hypothesized that amoeboid cells could regulate the endothelium via secretion. In fact, abnormal endothelium and vessel leakiness are characterized by inter-endothelial gaps. Endothelial cell monolayers recapitulating established blood vessels were treated with CM A375M2 or CM A375P (Figure S7C, upper panel), and inter-endothelial gap formation was favored when HMVECs were treated with CM A375M2 compared to CM A375P (Figures S7C and S7D). This was accompanied by decreased vascular endothelial (VE)-cadherin junctional index (Figures S7C and S7D, lower panel). CM A375M2-induced effects were lost if HMVECs were treated with CM derived from ROCK1/2-depleted or NFKB1-depleted A375M2 cells (Figures 7B and 7C, right panel). In these conditions, A375M2 had lowered Myosin II levels (Figure 7C, left panel).

33Cytoskeletal changes in endothelial cells affect endothelial integrity. A phospho-antibody array (141 cytoskeletal regulators) was used to measure changes in endothelial cells (Figures 7D and S7E). HMVECs treated with CM A375M2 ROCK1/2-depleted cells showed decreased levels of phospho-proteins (Figures 7D and S7E) regulating vascular permeability (MEK, Src, PKA, and MLC2) were measured (Figures 7D and S7E). Furthermore, we found reduced endothelial permeability in HMVECs treated with CM from A375P cells compared to A375M2 cells (Figure S7F, left panel). Similar effects were observed in HMVECs when treated with CM from A375M2- ROCK1/2 or NFKB1-depleted (Figure 7E), or CM from A375M2 cells treated with ROCKi (Figure S7F, right panel). In all cases, Myosin II levels in melanoma cells were decreased (Figures 7C, 7F, and S7G). Similar effects were observed in HUVECs (Figure S7H). Therefore, cross-talk between ROCK and NF-κB in melanoma cells facilitates endothelial cell cytoskeletal remodeling.

34Lung is one of the main sites where melanoma metastasizes, and lung retention assays have been used to measure metastatic colonization abilities. Importantly, melanoma cells with decreased ROCK-Myosin II (Figure 7F) could not alter the permeability of lung capillaries (Figure 7G) and were less efficient in colonizing the lung compared to controls (Figure 7H).

35Thus, amoeboid melanoma cells with high Myosin II activity and high secretory potential have an advantage once they reach the metastatic site; they remodel the cytoskeleton of endothelial cells and increase vascular permeability to colonize the lung.

36ROCK-Myosin II contributes to amoeboid tumor invasion and metastasis in a cell-autonomous manner. Here, we show that amoeboid melanoma cells educate both the myeloid and the endothelial compartment via secretion, regionally, in the IFs of tumors and later at distant metastatic sites. Thus, we define a new regulatory role for Myosin II dynamics in cancer cells beyond intrinsic control of cell motility.

37AAMs compose the bulk of TAMs and are considered to be tumor promoting. Amoeboid cancer cells induce macrophages that are CD206highCD163highHLA-DR+CD86low and produce IL-10 and TGF-β. Such macrophages support both melanoma cell growth and endothelial cell growth. The ratio of “amoeboid melanoma-induced macrophages” over “tumor-killing macrophages” will have a positive impact on tumorigenesis.

38We show that amoeboid cancer cells retain their secretory potential at later metastatic stages, since they have an advantage during metastatic colonization via disrupting endothelial junctions and increasing endothelial cell permeability. Therefore, amoeboid cancer cells hijack these efficient mechanisms—typical of immune cells—to modify the vasculature.

39The secretome of amoeboid melanoma cells is complex and rich in immunomodulatory cytokines, chemokines, and growth factors. We show that amoeboid behavior is sustained via a positive feedback loop between ROCK-Myosin-II-driven secretion and IL-1α/NF-κB signaling, generating a strong circuit of signal amplification. Depletion or inhibition of ROCK-Myosin II activity reduces (1) amoeboid behavior, (2) secreted IL-1α, and (3) NF-κB activation in melanoma cells. Conversely, targeting NF-κB in melanoma cells abolishes amoeboid behavior and their secretory profile resulting in defective macrophage polarization and vascular permeability. While targeting NF-κB in the clinic is challenging, ROCK inhibitors are used to treat cerebral vasospasm and intraocular pressure. ROCK inhibitors decrease invasion in pancreatic cancer and improve efficacy of chemotherapy. Moreover, ROCKi AT13148 is in clinical trials for solid tumors (ClinicalTrials.gov identifier: NCT01585701). In light of our findings, ROCK-Myosin II inhibition in melanoma cells could also be used to reprogram the innate immune microenvironment.

40On the other hand, blocking secreted IL-1α in melanoma cells is sufficient to diminish NF-κB activation and amoeboid features. An FDA-approved antagonist of IL-1R for rheumatoid arthritis has been evaluated for the clinical management of cancer metastasis (ClinicalTrials.gov identifier: NCT00072111) and a monoclonal antibody against IL-1α is in phase III of clinical trials for colorectal cancer (ClinicalTrials.gov identifier: NCT01767857).

41In conclusion, regional distribution of amoeboid melanoma cells with high Myosin II activity can contribute to lack of therapy responses by establishing a tumor-supportive AATME. We propose that, after surgical removal of the primary melanoma lesion, the amoeboid phenotype should be targeted using either ROCKi or IL-1α blocking antibodies as therapies to restrict immunosuppressive microenvironments and metastatic dissemination.

42Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Victoria Sanz-Moreno (v.sanz-moreno@qmul.ac.uk)

43Patients were staged and classified according to the American Joint Committee on Cancer Melanoma Staging and Classification criteria. Human samples were collected with informed written consent, in accordance with the Helsinki Declaration, and the study design was approved by the Guy’s Research Ethics Committee and Ethics Committee of Guy’s and St Thomas’ NHS Foundation Trust and the Ethics Committee of the IRBLleida Biobanc, in accordance with the Human Tissue Act, 2004. This study was approved by the Guy’s Research Ethics Committee, study number 08/H0804/139. Tables S2–S4 show clinical information from human melanoma patients.

44A375P, A375M2, WM88, WM1366, 5555 and B16F10 cells were maintained in DMEM, containing pyruvate, 4.5 g/ml D-Glucose, supplemented with 10% fetal calf serum (FCS), 100 units/ml penicillin, 100 μg/ml streptomycin and 2 mM L-glutamine and incubated at 37°C, 10% CO2. WM983A, WM983B, WM783B and WM8354 cells were cultured in RPMI supplemented with 10% FCS, 2mM L-glutamine, 100 units/ml penicillin and 100 μg/ml streptomycin and incubated at 37°C, 10% CO2. HUVECs were cultured in EBM2 media containing 2% fetal bovine serum (FBS), 0.1% hydrocortisone, 0.4% hFGF-b, 0.1% VEGF, 0.1% R3-IGF, 0.1% ascorbic acid, 0.1% hEGF, 0.1% GA-1000 and 0.1% heparin and cultivated on fibronectin coated flasks. HMVECs from a single neonatal donor were cultured in EGM-2MV bullet kit media comprised of EBM-2 Basal Medium, 5% FBS and EGM-2 MV growth supplements and were cultivated on gelatin-coated flasks. Both HUVEC and HMVEC were maintained in a 37°C, 5% CO2 incubator. THP-1 and U937 were cultured in RPMI supplemented with 10% FCS, 2mM L-glutamine, 100 units/ml penicillin and 100 μg/ml streptomycin and incubated at 37°C, 5% CO2. Primary monocytes were cultured in RPMI supplemented with 10% FCS, 2mM L-glutamine, 100 units/ml penicillin and 100 μg/ml streptomycin and incubated at 37°C, 5% CO2.

45Fibrillar bovine dermal collagen (no. 5005-B; PureCol, Advanced BioMatrix) was prepared at 1.7 mg/ml in DMEM; 100 μl/well in 96-well plates; 700 μl/well in 12-well plates. After collagen gel polymerization (4 h), cells were seeded on top of collagen in medium containing 10% FCS, allowed to adhere for 24 h, treatments added (where appropriate), imaged and fixed or lysates collected. All assays were performed with melanoma cells seeded on top of a thick layer of collagen unless otherwise mentioned.

46pLNT/SFFV EGFP lentivector (1 μg) was transfected into HEK293T along with packaging vectors (p-MD2.VSVg (0.4 μg) and pΔ8.91 (1 μg)) using Lipofectamine 2000 (7.5 μl/well of a 6-well dish). Media was replaced 6 h after transfection. Media with lentiviruses were collected 48 h after transfection, spun down, filtered (0.45 μm) and added to recipient cell lines (A375P, A375M2, WM983A, WM983B and B16F10). EGFP-positive cells were FACS-sorted and used for subsequent experiments.

47HEK293T cells on 10 cm2 dishes were transfected using 2M CaCl2, and 2x HBS (51558, Sigma) with lentiviral Venus vector (15 μg), pMD2-VSVg (6 μg), pRSVrev (6 μg) and pMDL-g/p-RRE (6 μg). Supernatants were collected 48 h and 72 h after transfection. For lentiviral transduction, 105 5555 cells/well were seeded in 6-well tissue culture dish and infected with VENUS reporter lentiviruses added in suspension using 10 μg/ml Polybrene (107689, Sigma). After 48 h, successfully transduced cells were trypsinised and FACS sorted according to their VENUS expression.

48For ROCK inhibition, 1%-FCS media with inhibitors (5 μM H1152 or 10 μM Y27632) was added to cells for 4, 24 or 48 h with re-addition at 24 h. For IKKβ inhibition, IKKβ inhibitor III (0.5 μM) was used for 24 h. DMSO was used a vehicle in the same concentrations as the inhibitors. For the time-course experiments, the inhibitors were used at the same concentration for all the indicated time-points.

49A375P, A375M2, WM983A, WM983B, WM88 and WM793B (2.5x105 cells/well) were seeded in complete DMEM or complete RPMI (+10% FCS) in 6-well plates. Next day cells were washed with PBS (with calcium and magnesium) and were cultured in serum-free (SF) media for 48 h. For inhibitor treatments, H1152 (5 μM), Y27632 (5 μM) or GSK269962A (5 μM) was added in serum-free media and replenished after 24 h. Then media was collected, spin down to eliminate debris and used fresh in subsequent experiments. Recipient cells on collagen matrices were treated with secreted media for 24 h and then cell morphology and p-MLC2 levels were assessed. For the experiments with CM derived from melanoma cells isolated from TB or IF of A375M2 tumors (see also Diagram 1 in Methods S1), CM was generated similarly to above. For experiments using blocking antibodies, A375M2-derived media was pre-incubated with blocking antibodies (anti-IL1α, anti-IL8 and their respective isotype controls (IgG2A and IgG1), 0.75μg/ml) for 1 h at 37°C. Then, media was added to A375P cells seeded on coverslips (1 h treatment, for p65 immunofluorescence) or to A375P cells seeded on top of collagen (24 h treatment, for cell morphology analysis). For macrophage polarization experiments, the secreted media was concentrated using Amicon Ultra-4 Centrifugal Filter Unit with Ultracel-3 membrane (Millipore, Watford, UK). After centrifugation for 45 min at 4000 rpm at 4°C, protein concentration was measured by BCA Protein Assay Kit (Life Technologies).

502x105 melanoma cells/well were seeded on 6-well plates and transfected the next day with 20-40 nM SmartPool or individual OTs (On Target) siRNA oligonucleotides, using Optimem-I and Lipofectamine 2000 (Invitrogen). Non-targeting siRNA was used as control. In case of transfection for siRNA to MYL12B and ROCK1/2, transfected cells were incubated for 24 h and 48 h, respectively, after which they were harvested and re-seeded at 25x104 cells/well for conditioned media experiments and serum starved for 48 h.

51Peripheral blood mononuclear cells (PBMC) from healthy donors were obtained from anonymized human buffy coats supplied by the NHS Blood and Transplant (Tooting, London, UK). Buffy coat was diluted with PBS (GIBCO) and PBMC isolation was performed by Lymphoprep density gradient separation (Axis-Shield, Oslo, Norway). MACS technology was used to isolate CD14+ monocytes.

52For melanoma conditioned macrophage differentiation, 106 CD14+ monocytes per well (6-well dish) were seeded in complete RPMI and incubated in 5% CO2 at 37°C. On day 3, 50% of the media was replenished with fresh media containing 35 μg/ml of CM derived from A375P, A375M2, H1152- or GSK269962A-treated A375M2 melanoma cells or CM derived from A375M2 depleted from MLC2 (MYL12B) or NFKB1 and monocytes were incubated for 3 additional days. A panel of known stimuli was used as controls of different types of differentiated/polarized macrophages. M-CSF (50 ng/ml), IL-4 (20 ng/ml) or IL-10 (20 ng/ml), or IFN-γ (20 ng/ml) plus LPS (100 ng/ml) were added on day 3 to monocytes after media replenishment. CD14+ cells with media only served as control. All cytokines were from Peprotech (London, UK) while LPS (derived from Escherichia coli 0111:B4) was from Sigma (Dorset, UK).

53For macrophage differentiation with melanoma patient-derived serum (MPS), CD14+ monocytes (106/ml) from healthy PBMC were plated in SF-RPMI and 10% of melanoma patient serum was added per well. Serum from either healthy volunteers (HVS) or human AB pooled serum (Sigma, Dorset, UK), were used as controls in the same volumes. Cells were incubated for 6 days in 5% CO2 at 37°C. HVS samples were obtained from King’s College London and National Institute for Health Research Biomedical Research Centre at Guy’s and St Thomas’ Hospitals. All sera samples used in these experiments were allogeneic to CD14+ monocytes. See also Table S2 for clinical information.

54On day 6, cell supernatants were collected for the detection of cytokines. PBS −/− was added on cells and plates were placed on ice for 20-30 min. Cells were then scrapped using a p1000 pipette and used for phenotypic staining.

55Bright-field images from day 6 of cell culture (see also ‘In vitro differentiation of human CD14+ monocytes to macrophages’ section) were used to analyze macrophage morphology using ImageJ software. Cells were divided into three categories according to their shape: rounded-immature, ‘fried-egg’ and spindle and one field of view was evaluated per condition. Data from 3 independent experiments with cells from 3 different healthy donors were used.

56Zymosan A S. cerevisiae fluorescein-conjugated BioParticles (Life Technologies, Paisley, UK) were sonicated (3x20s; 90 ultrasonic watts) to obtain a homogeneous population. Zymosan particles were re-suspended in PBS−/− (108 particles/ml) and opsonized with an equal volume of human AB pooled serum (Sigma) for 1h at 37°C. Differentiated macrophages were serum-starved for 2 h before starting the assay. Particles were thoroughly washed and added in a ratio of 10:1 (particles:macrophages) to macrophages and incubated for 1 h at 4°C to allow particles to bind on the cells and synchronize the onset of phagocytosis. Then cells were washed to remove unbound particles and incubated for 1 h at 37°C, 5%CO2, after which cells were harvested with PBS−/−, washed with FACS buffer and fixed in 1% para-formaldehyde to quench phagocytosis. Cells were acquired on a BD FACS CANTO II. As a negative control of phagocytosis, macrophages were treated with 5 μM cytochalasin D (Insight Biotechnology, Wembley UK), which is an inhibitor of actin polymerization-dependent phagocytosis. The inhibitor was added to the relevant wells 1 h before adding the zymosan particles and maintained throughout the assay.

57Briefly, melanoma target cells (A375M2, WM88, WM1366, WM793B, WM3854 and WM983A) were trypsinized, counted and labeled with CFSE (5 μM) using the CFSE cell division tracker unit from Biolegend following manufacturer’s instructions. Both CFSE-labeled melanoma cells and macrophages were washed twice in complete RPMI before the co-culture. Targets (104 melanoma cells) were added on macrophages (105 macrophages) in a ratio of 1:10 (melanoma targets:macrophages) in complete RPMI and incubated for 48 h in 48-well plates in duplicates. Co-cultured cells were harvested with PBS −/− and washed with FACS buffer. Cells were re-suspended in DAPI solution (5 μg/ml) for viability staining and immediately acquired on a BD FACS CANTO II.

58Tumors were dissected from mice, transferred in falcon tubes with PBS −/− CaCl2/ MgCl2 and kept in ice before subjecting to tumor cell isolation. Tumors were sliced in the middle into two parts using a scalpel; one part was kept in 4% formaldehyde solution for 48 h for paraffinization while the second part was used for tumor cell isolation. For the latter, the tumor half was peeled with the use of forceps, a pair scissors and scalpel to collect the peritumoral area which was more transparent, while the tumor core was a solid, darker area. Tumor core and peritumoral area were both collected and chopped into small pieces and incubated with digestion solution (1ml/sample); (Digestion solution: 90μl Liberase TM (Roche), (5 mg/ml), 90μl Liberase TH (Roche), (5 mg/ml), 30μl DNase I (Sigma), (5 mg/ml) and 6 mL HBSS (GIBCO)) at 37°C with shaking. See also Diagram 1 in Methods S1.

59Chemotaxis of human primary peripheral blood monocytes, THP1, U937 was assessed using 6.5-mm Costar Transwell cell culture chamber with polycarbonate membrane (5.0 μm pore) following the manufacturer’s protocol. Cell suspensions of 3x105 cells/0.1ml in SF-RPMI were loaded in the upper chamber compartment. CM from A375M2 or A375P were loaded in the lower chamber. Migrated cells were counted using a Neubauer Chamber. For chemotaxis of A375M2 or A375P cells, 8.0 μm-pore transwells were used instead, and the cell suspension was of 2x105 cells/0.1ml in SF-DMEM.

60CM from A375P, A375M2 control, or ROCK1/2 or NFKB1-depleted A375M2, or A375M2 cells treated with ROCK inhibitors H1152 (5 μM) or Y27632 (10 μM) for 48 h were collected and added to confluent monolayers of HMVECs on gelatin-coated Transwell filters (Costar). FITC-dextran (0.1 mg/ml) was added to the top chamber. Samples from the lower chamber were removed after 1 h incubation and added to a black 96-well plate. Fluorescence was measured using a microplate analyzer. Each condition was performed in triplicates.

61HMVECs or HUVECs cells were seeded in 6 replicates (104/well) on 1% gelatin (Sigma, Dorset, UK)-coated 96-well plates in complete EGM-2MV medium. The following day, cells were washed with PBS (+/+) and treated with 50% supernatant derived from monocytes treated with melanoma-derived CM (A375M2 +H1152 or GSK269962A), from untreated monocytes or with EGM-2MV medium (as control) for 72 h. Proliferation was measured by an MTT assay (5 mg/ml; Sigma, Dorset, UK). The absorbance of the samples at 570 nm and 695 nm was measured on a Perkin Elmer/Packard fusion Alpha-FP microplate analyzer. The results are presented as the average values of 6 replicates after background subtraction (O.D. 570 nm - O.D. 695nm).

62Secreted media from A375M2 and A375P cells seeded on a thick layer of collagen I were collected after 48 h and incubated with Human Cytokine Antibody Array (RayBiotech, Inc., C4000) following the manufacturer’s protocol. Membranes were incubated with biotinylated detection antibody cocktail, with HRP-conjugated streptavidin and with detection buffers. Images were obtained with a chemoluminiscent imaging system and densitometry analysis was performed using the Protein Array Analyzer plugin for ImageJ (http://image.bio.methods.free.fr/ImageJ/?Protein-Array-Analyzer-for-ImageJ.html). Enrichment maps and networks and process networks were obtained using MetaCore software from Thomson Reuters (https://thomsonreuters.com/metacore/).

63HMVECs were incubated with CM derived from A375M2 or ROCK1/2-depleted A375M2 (48 h) and after 1 h, cells were lysed using a protein extraction buffer and lysis beads from the Cytoskeleton Phospho Antibody Array (Full Moon Biosystems, CP141). Extracted protein was purified through columns and quantified measuring UV absorption. Protein lysate (40 μg) was biotinylated and conjugated to the antibody array. This array contains antibodies against 141 proteins involved in cytoskeletal pathways.

64Cells were lysed in Laemmli Buffer and lysates were resolved by 10 or 12% SDS-polyacrylamide (PAGE) gels or pre-made Nu-PAGETM 4%–12% Bis-Tris gels (Invitrogen) and transferred to PVDF filters (0.45 μm, Immobilon). The ECL Plus or Prime ECL detection systems (GE Healthcare) with HRP-conjugated secondary antibodies (GE Healthcare) were used for detection. Bands were quantified using ImageJ (https://imagej.nih.gov/ij/).

65CM from melanoma cells or supernatants from macrophages were used for sandwich ELISA experiments using commercially available kits (see also the table for reagents above). 96-well NUNC clear flat-bottom plates were coated with antibodies (IL-10, IL-8, IL-1α and TNF-α) and incubated O/N at 4°C in a humified chamber. ELISA MAX™ standard sets (Biolegend) were used for the detection of IL-10, IL-8 and TNF-α; ELISA MAX Deluxe (Biolegend) for IL1-α; and for TGF-β, LEGEND MAX Total TGF-β1 ELISA Kit with pre-coated plates was used (Biolegend). Microwell absorbance was read at 450 nm on a Thermo Scientific Multiskan EX microplate reader. All samples were quantified based on a standard curve using Microsoft Excel.

66HMVECs seeded onto gelatin-coated coverslips were incubated in the presence of media derived from A375M2 cells, ROCK-depleted A375M2 cells or ROCK-inhibited A375M2 cells for 1 h. Cells were fixed with 4% p-formaldehyde, permeabilised with 0.3% Triton and stained with phalloidin for F-actin, with anti-VE-cadherin and with Hoechst 33258 for the nuclei. A375M2 or A375P cells were seeded on top of collagen gels and treated as indicated. Cells were fixed, permeabilized, incubated with primary antibody (pMLC Ser19, Cell Signaling) and stained with secondary Alexa Fluor-647 anti-rabbit (Life Technologies) and Alexa Fluor 546-phalloidin for F-actin detection (Life Technologies). For imaging, gels were inverted onto MatTek dishes, while coverslips were mounted with DAKO fluorescence mounting medium (Dako, Cambridgeshire). In both cases, images were taken with a Zeiss LSM 510 Meta confocal microscope (Carl Zeiss, Germany) with C-Apochromat 40 × /1.2 NA (water) objective lens and Zen software (Carl Zeiss).

67For cell morphology, the shape descriptor “roundness” in ImageJ was used after manually drawing around the cell shape using F-actin staining images. Phospho-MLC2 fluorescence signal was quantified calculating the pixel intensity in single cells relative to the cell area.

68Junctional index and gap area parameters were measured with ImageJ. Junctional index is calculated as described by.

69For p65 staining, cells seeded on glass coverslips were fixed with 4% formaldehyde for 15 min and washed with PBS (3 × 5 min). Then, cells were permeabilised for 20 min with 0.5% Triton X-100 in 4% BSA-PBS and washed 3 times with PBS. After blocking for 30 min, cells were incubated with anti-p65 (1:100) overnight at 4°C. Cells were then washed 3 times with PBS and incubated with anti-Rabbit Alexa Fluor 488 (1:350) and Alexa Fluor 546-phalloidin (1:350) for 2 h at room temperature. Finally, cells were washed with PBS 4 × 5 min, incubated with 5 μg/ml Hoechst 33258 in the 4th wash, and mounted on a slide with Fluoroshield mounting media. Blocking and antibodies were prepared with 0.2% Triton X-100 in 4% BSA-PBS.

70Cells were washed once with FACS buffer (PBS−/−, 1% BSA, 2 mM EDTA, 0.1% NaN3) and after FcR blocking (Human TruStain FcX, Biolegend) co-stained for: HLA-DR-FITC (1:50, clone: L243), CD86-PE-Cy7 (1:20, clone: IT2.2), CD163-APC (1:20, clone: GHI/61); all from eBiosciences (Hertfordshire, UK) and CD206-PE (1:20, clone: 15-2), (Biolegend, London, UK). After 30 min incubation at 4°C in the dark, cells were washed twice with FACS buffer (1500 rpm, 4°C, 5 min) and re-suspended in 500 μl DAPI solution (5ug/ml, Biolegend) for viability and immediately acquired on a BD FACS CANTO II flow cytometer and analyzed using FlowJo 7.6.5 software (Tree Star). Purity of isolated CD14+ cells was checked by staining for CD14-PerCP-Cy5.5 (1:50, clone: HCD14, Biolegend) and was routinely > 95% across all the experiments.

71Gene expression data of human melanoma samples from published microarray studies was used to analyze most of the cytokines enriched in media derived from A375M2 cells in melanoma progression. We only took into account studies with enough sample purity (> 95% melanocytic cells), enough patient samples to perform statistical comparisons (n > 40) and studies including normal tissue. From public database GEO we extracted the Avery (GEO Accession number GSE29359), Xu (GEO Accession number GSE8401), Talantov (GEO Accession number GSE3189), Kabbarah (GEO Accession number GSE46517) and Riker (GEO Accession number GSE7553) series. Samples from these studies were reported to have > 95% melanocytic/melanoma cells and no mixed histology. Data were normalized using Gene Pattern (https://www.broadinstitute.org/cancer/software/genepattern/) and analyzed as described in the ‘Statistical analysis’ section below.

72Gene expression data of human melanoma samples from The Cancer Genome Atlas (TCGA) database (https://cancergenome.nih.gov/) was also used to analyze cytokines (enriched in media from highly contractile A375M2 cells) expression in melanoma progression. We only took account patients who had not received neo-adjuvant treatment prior to the resection of the tumor that yielded the sample submitted for TCGA. Normalized expression data and z-scores for mRNA expression data were donwloaded from cBio-Portal.

73All animals were maintained under specific pathogen-free conditions and handled in accordance with the Institutional Committees on Animal Welfare of the UK Home Office (The Home Office Animals Scientific Procedures Act, 1986). All animal experiments were approved by the Ethical Review Process Committee at King’s College London and carried out under license from the Home Office, UK.

74All animal procedures were approved and carried out in accordance with the UK Home Office and an Ethical Review Panel. All mice were obtained from Charles River. Severe combined immunodeficient mice (SCID; CB17/Icr-Prkdcscid/IcrIcoCrl) were used for studies with human melanoma cell lines (for A375P-, A375M2-, WM983A- and WM983B-EGFP cell lines) while C57BL/6J mice were used with mouse melanoma cell lines (B16F10-EGFP and 5555-Venus). Mice were female 5-6 weeks old for all experiments except for experiments using 5555 cells (6-10 weeks old).

75Prior to injection, cells were counted and resuspended in PBS −/− CaCl2/ MgCl2. Mice were anaesthetized with isoflurane and cells (2 × 106 for A375P, A375M2 and WM983A, WM983B; and 1.5 × 105 for B16F10) in a volume of 50 μL were subcutaneously injected on the flank. Mice were continually monitored with tumor dimensions being determined by calliper measurements. Tumor volume (mm3) = length x width x height x 0.52.

76Visualization of tumor fluorescence was achieved using a Fluorescent Protein Flashlight (NIGHTSEA) post injection and upon tumor establishment. A375s and WM983s tumors were grown for 33 and 55 days, respectively, while B16F10 tumors were grown for 15 days. Prior to dissection all tumors were measured, dissected out, surrounding tissue removed, followed by fixation in 4% formaldehyde solution for 48 h. All dissected tumors were weighed, and fluorescence images taken using a Fluorescence labeled Organism Bioimaging Instrument (FOBI; NeoScience).

77For Y27632 ROCK inhibitor studies, drug administration was by intraperitoneal injection every other day once tumors had reached a mean volume of 160 mm3. Dosing was at 50 mg/kg (Y27632; Bio-Techne). Y27632 was prepared in PBS −/− CaCl2/ MgCl2, and sterile filtered prior to use.

78For experiments using 5555 cells, prior to injection 5555-Venus cells were pre-treated in vitro with GSK269962A (5 μM) or DMSO as control for 5 days, with fresh drug added every day. Then 2x105 cells were injected intradermally into C57BL/6J mice. Both in vitro and in vivo viability was measured after drug removal. Therefore, mice were never treated during the course of the in vivo experiment. Tumor volume was measured twice a week and tumors harvested after 8 and 14 days for further analyses.

79A375M2 cells were transfected with either control siRNA or siROCK1/2 (60 nM total for both). Three days later, cells were labeled with 10 μM CMFDA-Green (C7025, Life Technologies) for 10 min and then they were trypsinized and counted. 1x106 labeled cells/0.1ml PBS were injected into tail vein of NOD/SCID/ IL2Rγ−/− mice (NSG, Charles River). 24 h later, TRITC-dextran (70 kDa, D1818, ThermoFisher) was intravenously injected and 10 min later mice were sacrificed. Lungs were extracted, washed with PBS (with calcium/magnesium) twice and fixed with 4% formaldehyde for 16 h at 4°C. Lungs were examined under a confocal microscope (see Immunofluorescence section). Data are presented as % field of area covered by fluorescence (green for cells, red for dextran which represents permeability), n = 5 mice/condition for each experiment, n = 2 independent experiments.

80Severe combined immunodeficient mice (SCID; CB17/Icr-Prkdcscid/IcrIcoCrl; Charles River) were injected intraperitoneally (i.p.) with 150 μl of clodronate liposomes or PBS liposomes as control (cat# CP-025-025, Liposoma B.V. the Netherlands). Next day, tumor cells (2x106 cell/ 0.1ml PBS) were injected subcutaneously along with 20 μl of clodronate or PBS liposomes into mice. Then, 150 μl clodronate or PBS liposomes was administered i.p. twice per week. Tumor volume was monitored every 3 days by caliper measurements as in “Melanoma tumour models” section. On day 27, mice were culled and tumors were harvested, fixed and embedded in paraffin using standard protocols (see “Immunohistochemistry” section). Macrophage depletion was confirmed by F4/80 and CD206 IHC staining at endpoint.

81Two cohorts of human melanoma samples were included in the case series. The ‘cohort A’ is a tissue microarray comprising of two slides with consecutive sections of individual tumor cores corresponding to either TB or IF for 40 different patients (n = 24 primary and n = 16 metastasis). Each patient is represented by 8 cores (4 replicates/tumor body and 4 replicates/IF) so 320 cores were totally analyzed for each marker. The ‘cohort B’ was comprised by whole-section tissues corresponding to 7 different patients.

82Whole sections from subcutaneous tumors (human A375P-, A375M2-, WM983A- and WM983B-EGFP; and murine B16F10-EGFP) and from intradermal tumors (5555-Venus) were included.

83All tissue samples were formalin-fixed paraffin-embedded (FFPE) and were sectioned (3 or 4 μm-thick) and dried for 1 h at 65°C. Next, tissue samples were subjected to deparaffinization, rehydration and heat-induced epitope retrieval using a Biocare Decloaking Chamber (DC2012) at 110°C for 6 min in Access Super Menarini Buffer (MP-606-PG1). Then endogenous peroxidase and phosphatase alkaline were blocked with Dual Endogenous Enzyme-Blocking Reagent (Dako, Agilent) for 10 min.

84Incubation with primary antibodies was performed O/N at 4°C in a humidified slide chamber. Human samples were subjected to IHC for pSer19-MLC2 (1:50, polyclonal, #3671), CD68 (1:2 dilution/pre-diluted; clone: KP-1, #ab74704), CD163 (1:100, clone: EDHu-1, MCA1853), CD206 (1:2000, ab64693) and CD31 (1:100, ab28364). Murine tumor samples were stained for GFP (1/2000, A-11122), S100β (1/400, ab52642), F4/80 (1/400, ab100790), CD206 (1/2500, ab64693), CD31 (1/100, ab28364) and pSer19-MLC2 (1:50, polyclonal, #3671).

85Relevant secondary antibodies IgG HRP (Horse-Radish Peroxidase) (1/100, anti-rabbit or anti-mouse, Dako) or AP (Alkaline Phosphatase) (1/100, anti-rabbit or anti-mouse, Dako) were incubated for 1 h at room temperature. Subsequently, samples were developed by incubation in DAB+ or Permanent Red chromogen solutions (Dako, Agilent). Samples were counter stained with hematoxylin. Human melanoma tissues and the adjacent normal skin were used as positive and negative controls for the above-mentioned antibodies.

86Human tissue samples were imaged and scored as follows. Cohort A was scanned on a ZEISS Axio Scan.Z1, and images were analyzed by ZEN 2012 (blue edition). Cohort B was scanned using a Hamamatsu Nanozoomer and images analyzed using NDP view2. For the analysis, each individual tumor core was evaluated for the total number of vessels or macrophages and then averages of the numbers per IF or TB were calculated for each patient. For the comparison between primary and metastatic melanoma patients, averages from all the replicates (including TB and IF) were calculated. For vessel and macrophage analysis, ≥ 2 fields of view were used per TB or IF for analysis. The latter was defined as the tumor area composed by melanoma cells only with at least 50% cell surface in contact with the matrix.

87Cell shape scoring was performed as previously described. Briefly, cell shape score = ((percentage of cells [%] shape 0 × 0) + (% shape 1 × 1) + (% shape 2 × 2) + (% shape 3 × 3)), with values ranging from 0 (all cells round) to 300 (all cells spindle). For quantification of p-MLC2 staining, samples were scored blind and a staining H-score was provided for each section, where H-score = Σ (% of cells with 4∗4) + (% of cells with 3∗3) +(% of cells with 2∗2) +(% of cells with 1∗1); where 0 = no staining, 1 = weak, 2 = moderate, 3 = intense and 4 = very intense staining.

88For A375P/A375M2 and WM983A/WM983B tumors, samples were scored 0-4 using the H-score method for p-MLC2, in addition to quantification of the percentage of cells with very intense (H-score = 4) p-MLC2 levels at distances 0-2 mm from the IF of tumors.

89For murine tumors, 6 consecutive sections from each sample were used. Two sections were used to detect F4/80+CD206+ macrophages, two to detect melanoma cells using the antigen S100β and GFP and two for p-MLC2 and CD31+. One area for IF and TB was selected for each sample and imaged at 20X magnification using iScope microscope (IS.1159EPLi, Euromex) and ImageFocus 4.0 (Euromex) software (Diagram 2A in Methods S1). Next, the images were run in ImageJ platform (https://imagej.nih.gov/ij/), scaled at 3.08 px/μm2. Color deconvolution H-DAB plug-in (https://imagej.net/Colour_Deconvolution) was applied to split the image in three main channels/colors. The area of interest was determined, and the number of positive cells was determined using “find maxima” tool. The process was iterated for all markers used in the analysis (F4/80, CD206, GFP and S100β) (Diagram 2B in Methods S1). All the values were introduced in an Excel file and were standardized to the same area (105 μm2). To avoid false positive macrophages (that is cancer cells that were CD206+) we applied a correction factor: subtraction of GFP or S100β signal from CD206 total signal.

90Statistical analysis was performed using GraphPad Prism (version 6, San Diego California USA). The following statistical tests were used: t test unpaired, t test with Welch’s correction, Mann-Whitney test, one-way ANOVA with Tukey post hoc, Kruskal-Wallis with Dunn’s multiple comparison test and Wilcoxon matched-pairs signed rank test and Two-way ANOVA- post with Bonferroni post hoc test. All experiments were analyzed with a minimum of three independent repeats. Outliers were excluded using the ROUT method. For column bar or scatter dot plots, error bars are the average ± SEM. Boxplots show min to max values or 10-90 percentile. ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, ∗∗∗∗ p < 0.0001.