Location, location, location: Melanoma cells “living at the edge”

Abnormal cell migration and invasion underlie metastatic dissemination, one of the major challenges for cancer treatment. Melanoma is one of the deadliest and most aggressive forms of skin cancer due in part to its migratory and metastatic potential. Cancer cells use a variety of migratory strategies regulated by cytoskeletal remodelling. In particular, we discuss the importance of amoeboid invasive melanoma strategies, since they have been identified at the edge of human melanomas. We hypothesize that the presence of amoeboid melanoma cells will favour tumor progression since they are invasive and metastatic; they support immunosuppression; they harbour cancer stem cell properties and they are involved in therapy resistance. The Rho‐ROCK‐Myosin II pathway is key to maintain amoeboid melanoma invasion but this pathway is further regulated by pro‐tumorigenic/pro‐metastatic/pro‐survival signalling pathways such as JAK‐STAT3, TGFβ‐SMAD, NF‐κB, Wnt11/5‐FDZ7 and BRAFV600E‐MEK‐ERK. These pathways support amoeboid behaviour and are actionable in the clinic. After melanoma wide surgical margin removal, we propose that possible remaining melanoma cells should be eradicated using anti‐amoeboid therapies.


| INTRODUC TORY OVERVIE W
Melanoma is one of the deadliest and most aggressive forms of skin cancer, with the occurrence of metastasis often being incurable and predictive of poor survival. 1 Local invasion and metastatic spread are responsible for the morbidity and mortality in melanoma. Patients with localized or regional disease have a relatively good prognosis with the 5-year relative survival rate of 98% and 64%, respectively.
In contrast, the 5-year survival rate is reduced to 23% in patients with metastatic melanoma (stage IV). 2 Clinically, the Breslow index or the absolute depth of local invasion, measured directly by histopathologic analysis, is the principal prognostic factor and primary criterion in melanoma staging. 3 Over the past years, several therapies have been approved for melanoma. 4 Depending on the features of the tumor (location, stage, and genetic profile), the therapeutic options may be surgical resection, chemotherapy, radiotherapy, photodynamic therapy (PDT), targeted therapy or immunotherapy. Despite the fact that surgery is the primary treatment in thin melanomas (up to 2 mm), adjuvant therapies are recommended. 5 Melanoma is highly metastatic and heterogeneous at early stages of disease, posing a great challenge in the clinic.
Melanocytes are derived from the embryonic neural crest which has undergone morphological changes accompanied by altered adhesion and migration abilities to colonize the epidermis. 6 Therefore, | 83 MAIQUES And SAnZ-MOREnO transformed melanocytes suffer a de-differentiation process that brings them closer to their former neural crest cell precursors. 7 Due to its plastic nature, melanoma cells can migrate individually or collectively as multicellular groups. 8 Melanoma cell migration requires both actin polymerization and Myosin II driven forces. Rho A/C-are crucial regulators of the cytoskeleton and activate Rho-associated protein kinases (ROCK1 and ROCK2). 9 ROCK1/2 promote actomyosin contractile force generation directly phosphorylating Myosin Light Chain 2 (MLC2) or indirectly by decreasing Myosin Phosphatase (MYPT) activity, therefore activating Myosin II complex. [10][11][12] Amoeboid cancer cell migration is propelled by very high levels of Rho-ROCK driven Myosin II activity and membrane blebs as functional protrusions. 11,13 We have found an enrichment of amoeboid invasive cells at the edge of human and mouse melanoma tumors 10,14-17 ( Figure 1A, B).
Such cells are therefore uniquely positioned to receive physicalchemical signals from the surrounding tissue; while they are capable of modifying that tissue more vigorously than cells at the tumor core.
Since these cells are at the edge of the tumor, they are well located to leave the tumor and access the vasculature ( Figure 1B Tyrosinase, Melan A, S100b and may not always detect undifferentiated melanoma. 25 Moreover, morphology-based classification methods do not provide relevant information for selecting treatments for patients whose tumors have metastasized. 26 It is therefore critical to improve clinical and histopathological criteria to predict metastasis risk. Apart from targeted therapies, immunotherapies have also significantly improved melanoma patient outcome 27 but partial responses or development of therapy resistance are a major challenge in the clinic, 28,29 therefore biomarkers of therapy response are also needed. We propose that detecting amoeboid melanoma cells (AMCs) could improve clinical practice since these cells are highly metastatic and involved in therapy resistance. Signalling pathways supporting amoeboid behaviour and possible markers for detecting AMCs are detailed below.

| AMOEBOID MEL ANOMA CELL S AND THEIR TR ANSCRIP TIONAL RE WIRING , MICROENVIRONMENT REMODELLING , AND CLINIC AL RELE VAN CE
Cancer cells can use different molecular mechanisms to migrate away from the primary tumor. Cell migration is a well-organized biological phenomenon which is modulated by multiple intrinsic (adhesion, actomyosin contractility, nuclear deformability) and extrinsic factors (matrix organization and composition). 30,31 Cancer cells switch between different modes of migration depending on environmental signals. Indeed, cells turn on amoeboid migration when cell-Extracellular Matrix (ECM) adhesion is reduced in pliable matrices. [32][33][34] However, under high confinement and low adhesive conditions, distinct amoeboid motility modes have been described (A1 and A2). 31,35,36 Indeed, conditions of high contractility generated by myosin II favour the A2 mode. Interestingly, tumor cells, as well as leukocytes, prefer the A2 mode. 31,35,36 We have also observed that AMCs present high levels of cortical Myosin II activity driven by Rho-ROCK signalling in pliable complex matrices 10,[14][15][16][17]37,38 suggesting that amoeboid cells retain a mechanical or a chemical memory (or both). In the past decade, we have reported several signalling pathways that re-inforce and sustain an amoeboid cancer cell state via secreted factors and transcriptional reprogram-  Figure 2A). This indicates that amoeboid cancer cells are part of the EMT spectrum. 41,42 During "Phenotype switching" 41,42 melanoma cells switch from a proliferative to a migratory state, all orchestrated by MITF and AXL. 43,44 However, a population of AXL high -MITF high melanoma cells was compatible with increased invasiveness and proliferation. 45 In agreement with this, we observed that AMC at the edge of mouse and human tumors were both proliferative and invasive 16 and in our transcriptome studies we observed that they expressed MITF mRNA. 16 As a result of high transcriptional TGFβ signalling activation, a subpopulation of melanoma cells was identified that simultaneously displayed proliferative and invasive properties. 45 Since their gene ontology analysis showed an enrichment in amoeboid features, 45 it is possible that this the same population of AMC. Since MITF plays a mechanosensitive role in melanoma 46 how ROCK driven contractility and MITF functions are balanced in amoeboid cancer cells remains to be fully understood.

| Tumor microenvironment of AMCs
Within the tumor, a variety of normal cells interact with the cancer cells promoting tumorigenesis such as stromal, endothelial and immune cells. 47 AMCs secrete a complex set of proteins that are crucially controlled by IL1α-NF-κB 14 (Figure 2A). Such ROCK-Myosin II-NF-κB driven secretion attracts monocytes and polarizes them into CD163+CD206+ pro-tumorigenic macrophages ( Figure 2C). Therefore, amoeboid behaviour is sustained via a positive feedback loop between ROCK-Myosin-II-driven secretion and IL-1α/ NF-κB, generating a strong circuit of signal amplification. While AMC-associated macrophages support melanoma cell growth, 14 mural cells secrete factors that support MAPK-ROCK2-Myosin IIdependent growth. 48 Moreover, AMCs disrupt endothelial junctions and increase endothelial cell permeability via secreted factors which aids during lung metastatic colonization 14 ( Figure 2C). These data suggest a complex set of feedback secretory loops between the TME and melanoma cells to support contractility, cancer cell growth and dissemination. In line with this, therapy-resistant melanoma tumors with high Myosin II levels recruit/polarize pro-tumorigenic macrophages and immunosuppressive FoxP3+ T cells 15 ( Figure 2C).
Using pre-clinical mouse models, ROCK inhibitor improved the efficacy of BRAF inhibitors or immunotherapy. 15 The ECM is a key component of the TME. Cancer cells switch migratory modes, depending on their environment. 49 Interestingly, physical confinement can trigger amoeboid behaviours 35,36 and Myosin II isoforms regulate actomyosin contractility levels under high levels of confinement in several tumor types. 50 Increasing evidence suggests that the nucleus can act as cellular mechano-sensor while chromatin organization and gene expression can be affected by mechanical forces. 51 The nucleus is capable of sensing physical confinement inducing actomyosin-dependent migratory behaviour. 52,53 Nuclear tension recruits cytosolic phospholipase A 2 (cPLA 2 ) to the inner nuclear membrane (INM) and produces arachidonic acid (AA), activating Myosin II in a calcium-dependent manner. Future studies should elucidate how confinement imposed by the growing tumor and the opposing extracellular matrix regulates Myosin II dependent behaviours at the edge of melanomas.

| CLINI C AL IMPAC T OF AM C s
Using a large set of human melanoma tissues we have reported that the edge-or the invasive front (IF)-of such tumors is enriched in cells with a rounded morphology that harbour very high levels of Myosin II activity as measured by phospho-MLC2 ( Figure 1A-C). 10,14,16,17 We have measured an increase in the levels of several amoeboid markers (active STAT3, MMP-9, CITED1) at the edge of such melanomas 10,17,38 and an association with pro-tumorigenic macrophages and vasculature. 14 Furthermore, non-canonical Wnt markers and CSC markers were also enriched at the IF. 16 Importantly, high levels of CITED1, 17 high levels of ALDH1A1-a CSC marker 16 and high levels of ROCK-Myosin II pathway regulators 15 conferred worse prognosis to patients. These data suggest that amoeboid cell content could predict worse outcomes for melanoma patients. On the other hand, therapy resistant melanoma cells were more sensitive to ROCK inhibitors and combination treatments improved targeted therapy and immunotherapy responses in tumors with high Myosin II levels. 15 We speculate that the presence of AMCs could indicate a better response to ROCK inhibitors.

| CON CLUS I ON S AND PER S PEC TIVE S
Our studies in hepatocellular carcinoma 54,55 suggest that amoeboid cancer cells (ACCs) may be induced at the IF of other solid cancers. Identification of universal biomarkers of ACCs at the edge of tumors will be crucial for clinical pathologists to predict ROCK inhibitor sensitivity. Moreover whether all amoeboid cells are the same or there is heterogeneity within patient samples, also needs to be determined. Recently, the dual ROCK-AKT inhibitor, AT13148 was used in a phase I clinical trial for solid tumors, mainly colorectal. 56 Dose-limiting toxicity due to hypotension was reported and AT13148 pharmacodynamics was also a limitation for its efficacy, while lack of response was associated with no reduction of Myosin II activity in the biopsies analysed. 56 After surgical removal, we suggest eradicating AMCs at the edge of melanomas. Since AMCs rely on ROCK for their aggressive behaviour, we suggest that AMC detection could improve clinical responses to ROCK inhibitors. Future research into this class of drugs will involve testing ROCK II isoform specific inhibitors or soft ROCK inhibitors with less toxicities. 57 For melanoma in particular, it will be important to consider the possibility of topical applications to reduce systemic side effects. Moreover, we have identified key signalling pathways that AMCs rely on and that are actionable in the clinic. JAK inhibitors, 58 IKK inhibitors, 59 TGFBR inhibitors 60 or IL-1alpha blocking antibodies 61 are all used for clinical applications and could be tested for abrogation of the amoeboid cell state. 62,63

ACK N OWLED G EM ENTS
This work was supported by Cancer Research UK (CRUK) C33043/ A24478 and Barts Charity.

AUTH O R CO NTR I B UTI O N S
OM and VSM wrote the manuscript. OM performed IHC and image analysis in Figure 1.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing not applicable to this article as no datasets were generated or analysed during the current study.