Xinqidi,Your biological research partner
Xinqidi Biotech Co.,Ltd,Wuhan,China 2008-2021
R&D 13th year

Loss of Ambra1 promotes melanoma growth and invasion

Issuing time:2021-05-06 10:00


Melanoma is the deadliest skin cancer. Despite improvements in the understanding of the molecular mechanisms underlying melanoma biology and in defining new curative strategies, the therapeutic needs for this disease have not yet been fulfilled. Herein, we provide evidence that the Activating Molecule in Beclin-1-Regulated Autophagy (Ambra1) contributes to melanoma development. Indeed, we show that Ambra1 deficiency confers accelerated tumor growth and decreased overall survival in Braf/Pten-mutated mouse models of melanoma. Also, we demonstrate that Ambra1 deletion promotes melanoma aggressiveness and metastasis by increasing cell motility/invasion and activating an EMT-like process. Moreover, we show that Ambra1 deficiency in melanoma impacts extracellular matrix remodeling and induces hyperactivation of the focal adhesion kinase 1 (FAK1) signaling, whose inhibition is able to reduce cell invasion and melanoma growth. Overall, our findings identify a function for AMBRA1 as tumor suppressor in melanoma, proposing FAK1 inhibition as a therapeutic strategy for AMBRA1 low-expressing melanoma.


Melanoma is the most aggressive and deadly type of skin cancer. If not diagnosed early, melanoma becomes highly metastatic. Despite its heterogeneity, which is largely due to its high mutational burden, alterations of key genes of the mitogen-activated protein pinase (MAPK) and the phosphatidylinositol-3-kinase/AKT serine/threonine kinase 1 (PI3K/AKT) pathways have been identified as hallmarks of melanoma and have contributed to improve patient stratification1,2. Among these genes, v-raf murine sarcoma viral oncogene homolog (BRAF) mutations (e.g., the V600E substitution) are the most frequent and reported in >50% of patients with melanoma3. Additionally, further genetic modifications may alternatively or concomitantly affect other players of the MAPK and PI3K/AKT pathway, such as neuroblastoma RAS viral oncogene homolog (NRAS), Neurofibromin 1 (NF1) and phosphatase and tensin homolog (PTEN)3.

In recent years, efforts to understand the molecular mechanisms underlying melanoma ontogenesis and progression, as well as develop new therapeutic approaches, have been intensified4,5,6. Treatment of advanced melanoma has been improved considerably with the use of (i) targeted therapy, designed to target the most common oncogenic mutations (e.g., BRAF/MEK inhibitors); and (ii) immunotherapy, which aims at stimulating the anti-tumor immune response mainly by T-cell reactivation (e.g., immune checkpoint inhibitors)4,5. Despite these efforts, melanoma treatment remains challenging because of the strong metastatic propensity and the acquisition of resistance to therapy, which are distinctive features of this disease.

Ambra1 is a multifunctional scaffold protein whose functional deficiency induces neuroepithelial hyperplasia associated with autophagy impairment in mouse embryos7,8. Indeed, AMBRA1 is considered as a pro-autophagic protein, since it facilitates autophagy initiation by regulating Beclin1 and Unc-51 like autophagy activating kinase (ULK1) activity7,8,9. Furthermore, due to the great flexibility of its structure, which explains its ability to interact with many molecular partners, AMBRA1 can regulate a plethora of biological processes, spanning from apoptosis to cell proliferation10. Importantly, Ambra1 has been proposed to act as tumor suppressor, since Ambra1 absence and haploinsufficiency, respectively, lead to cellular hyperproliferation and onset of spontaneous tumors in animal models11. In particular, we discovered that AMBRA1 controls cell proliferation and cell cycle progression by regulating c-Myc stability through its interaction with the protein phosphatase 2A (PP2A)11 and, more recently, also the Cyclin D1 stability by interacting with the E3 ligase DDB1/Cullin412.

Recently, it was reported that concomitant loss of AMBRA1 and Loricrin (a major component of cornified cells) expression in the peritumoral epidermis represents a prognostic biomarker for early-stage melanoma and high-risk tumor subsets, independently from tumor invasion depth13,14. Although this can argue for AMBRA1 being prognostic of melanoma outcome, a direct role of AMBRA1 in this disease has not yet been investigated. Indeed, AMBRA1 expression levels in the tumor tissue have not yet been associated with any stage/feature of melanoma.

Here, by applying the preclinical BrafV600E/Pten-deleted mouse model of melanoma15combined with the conditional knock-out of Ambra116, we provide evidence that Ambra1impacts on melanoma development. In particular, we show that the absence of Ambra1promotes the formation of melanocytic nevi and accelerates melanoma growth, eventually enhancing metastatic potential.


Ambra1 deficiency promotes melanoma development in Braf V600E/+;Pten −/− mice

To assess the role of Ambra1 in Braf/Pten-driven melanoma, Ambra1flox/flox mice16 were crossed with mice carrying BrafV600E mutation and Pten deletion (BrafV600E/+;Ptenflox/flox)15, a genetic combination that induces melanoma with 100% penetrance in mice17,18. In this genetically engineered mouse model (GEMM), melanocyte-specific Cre recombination is controlled by the tyrosinase promoter (Tyr::CreERT2/+) and occurs only upon 4-hydroxytamoxifen (4-OHT) administration. Three-week-old Tyr::CreERT2/+;BrafV600E/+;Pten−/−;Ambra1+/+ (BPA+/+), Tyr::CreERT2/+;BrafV600E/+;Pten−/−;Ambra1+/− (BPA+/−) and Tyr::CreERT2/+;BrafV600E/+;Pten−/−;Ambra1−/− (BPA−/−) mice were locally treated with 4-OHT on the skin of the lower back to initiate melanoma (Supplementary Figs. 1a and 2a). Noticeably, BPA−/− mice developed larger tumors (Fig. 1a, b) with a higher growth rate (Fig. 1c) than BPA+/+ mice. The median survival of BPA−/− mice also significantly decreased from 56 to 42 days (Fig. 1d). However, the morphology of these tumors did not differ in terms of myxoid, S100+ and pigmented cells (Supplementary Fig. 2b, c). Notably, we observed an Ambra1 dose-dependent impact on melanoma development, as showed by tumor growth and median survival of Ambra1 heterozygous (BPA+/−) mice (Fig. 1a, c, d). Nevertheless, as the complete depletion of Ambra1 resulted in stronger effects, BPA+/− tumors were excluded from further analyses.

Fig. 1: Ambra1 deletion promotes melanoma development in BrafV600E/+;Pten−/−mice.

a Representative pictures of tumors from BPA+/+, BPA+/− and BPA−/− mice at 32 days after treatment. b Weight of BPA+/+ (n = 12) and BPA−/− (n = 10) tumors at endpoint (42 days) (**p = 0.0053, two-tailed unpaired t-test, BPA−/− vs. BPA+/+). c Kinetics of tumor growth in BPA+/+ (n = 12), BPA+/− (n = 10) and BPA−/− (n = 10) mice. Each data point corresponds to the average tumor volume ± SEM for each experimental group (*p = 0.0128; ***p = 0.001, one-way ANOVA). d Kaplan–Meier overall survival curves of BPA+/+ (n = 11), BPA+/− (n = 8) and BPA−/− (n = 8) mice (p < 0.0001, Two-sided log-rank test, BPA−/− vs. BPA+/+) and e of mice subcutaneously injected with 2×106 BPA+/+- (sBPA+/+; n = 16) or BPA−/−- (sBPA−/−; n = 16) tumor-derived cells. Tumors appearance was monitored for 60 days (p < 0.0001, Two-sided log-rank test, sBPA−/− vs. sBPA+/+). f Areas of hyperplasia (black arrows) and of pigmentation in skin sections of BA+/+ (n = 9) and BA−/− (n = 7) mice are shown and g quantified. Each data point represents one mouse and corresponds to the average percentage ± SEM of the pigmented area (at least four regions analyzed for each mouse) (**p = 0.0033, two-tailed unpaired t-test, BA−/− vs. BA+/+). h Representative images of Ki67 immunostaining (red) in skin sections of BA+/+(n = 7) and BA−/− (n = 7) mice. Nuclei are shown in blue (Hoechst). White arrows indicate hair bulbs. iQuantification of the signal shown in h. Each data point represents one mouse and corresponds to the average ratio ± SEM of Ki67+ cells on the nuclei count (at least three fields analyzed for each mouse) (**p = 0.006, two-tailed unpaired t-test, BA−/− vs. BA+/+). Box-and-whisker plots shown in (b, g, i) represent values from minimum to maximum. Top and bottom whiskers denote upper (Q3) and lower (Q1) quartiles, respectively. Boxes refer to interquartile ranges. Medians are denoted as horizontal line in the middle of the boxes.

Additionally, the faster kinetics of growth was confirmed in a syngeneic mouse model, in which mice were subcutaneously injected with two different amounts of cells derived from BPA+/+ (sBPA+/+) and BPA−/− (sBPA−/−) tumors and followed up for 60 days (Fig. 1e) and 120 days (Supplementary Fig. 2d). In this timeframe, 87.5% and 80% of sBPA−/− mice developed tumors, respectively, whereas no tumors were observed in sBPA+/+ mice.

To test whether the loss of Ambra1 was sufficient to lead to melanoma formation in BrafV600E-mutated melanocytes18, Tyr::CreERT2/+;BrafV600E/+;Ambra1+/+ (BA+/+) and Tyr::CreERT2/+;BrafV600E/+;Ambra1−/− (BA−/−) mice were treated with 4-OHT on the dorsal skin at postnatal days 1, 3, and 5 (Supplementary Fig. 1b). Thirteen weeks after 4-OHT induction, skin sections from BA−/− mice displayed a significant increase in pigmentation (melanocytic nevi) and hyperplastic regions at higher magnification (Fig. 1f, g). These phenotypes were associated with a marked increase in Ki67 immunostaining, suggesting that BA−/−melanocytes proliferated more than BA+/+ counterparts (Fig. 1h, i). Ki67+ cells were mainly located at the highly pigmented areas in BA−/− skin sections, whereas Ki67 positivity was mostly detected at the hair bulbs of BA+/+ mice (Supplementary Fig. 2e). Considering the role of Ambra1 on c-Myc degradation via interaction with PP2A11 and its newly discovered function in controlling cell cycle progression by regulating Cyclin D1 stability12, we sought to investigate whether the hyperproliferation observed in Braf-mutated melanocytes in the absence of Ambra1 could be related to c-Myc and Cyclin D1 regulation. Interestingly, Ambra1-deficient nevi displayed a significant increase in Cyclin D1 levels, and a slight not significant upregulation of phosphorylated c-Myc (p-c-Myc-S62) (Supplementary Fig. 2f–h). BA+/+ and BA−/− mice were observed over a long period of time (up to 350 days) to monitor for melanoma onset. Tumor penetrance was low in BA−/− mice (12.5%) (Supplementary Fig. 2k, Supplementary Table 1), and no tumors were observed in BA+/+ mice (Supplementary Table 1).

These results suggest that Ambra1 deficiency promotes melanoma growth in a BrafV600E/+;Pten−/− genetic background. Also, it affects pigmentation and induces hyperplasia in the skin in combination with the BrafV600E mutation alone.

Ambra1 deficiency affects proliferation in Braf V600E/+ ;Pten −/− mice

Despite the faster growth kinetics of Ambra1-deficient tumors, the proliferation rate of BPA−/− mice analyzed at the final endpoint (i.e., after 42 days, when maximum tolerated tumor size was reached) appeared slightly, but significantly, decreased (Fig. 2a, b). The relative amount of Ki67+ cells in Ambra1-deficient tumors was, indeed, reduced if compared to their wild-type counterparts at the same time point (42 days) (Fig. 2a, b). To assess whether this seemingly reduced cell proliferation was due to increased cell death, we performed TUNEL assays and western blot (WB) analyses of the apoptotic marker Casp3 in BPA+/+ and BPA−/− tumor sections and extracts. No significant difference between the two genotypes was observed (Supplementary Fig. 2l, m), except for a lower number of total nuclei present in BPA−/− tumors (Supplementary Fig. 2n), which was in line with the reduced proliferation rate. However, when Ki67 analysis was performed in BPA+/+ and BPA−/− tumors of same size (e.g., Breslow thickness)—either after tumor appearance (≤2 mm), or when tumors reached maximum allowed size (≥ 4 mm)—we did not observe any significant difference between the two groups (Fig. 2c, d), suggesting that BPA+/+ and BPA−/− tumors actually showed similar proliferation rates. Nevertheless, it should be noted that, independently on the genotype, bigger tumors (i.e., ≥4 mm) were characterized by an overall reduction of proliferation rate than smaller tumors (≤2 mm) (Fig. 2c, d). These results clearly indicate that the number of proliferating melanoma cells do not change if normalized on tumor size. Therefore, the larger volume of BPA−/− tumors (Fig. 1c) can be only ascribed to a proliferative advantage conferred by the loss of Ambra1, that allows them to reach the maximum size in a shorter timeframe. In support to this hypothesis, we observed that BPA−/− tumors showed increased c-Myc phosphorylation and Cyclin D1 levels (Fig. 2e–g). This additional evidence is in agreement with results provided elsewhere11,12, and indicates that loss of Ambra1 likely promotes melanoma growth by increasing cell proliferation.

Fig. 2: Cell proliferation upon Ambra1 deletion.

a Representative IHC staining of Ki67 on FFPE-tumor sections of BPA+/+ (n = 7) and BPA−/− (n = 7) mice. b Quantification of the Ki67+ cells shown in a. Each data point represents one mouse and corresponds to the average ratio ± SEM of Ki67+ cells on nuclei count (at least three sections analyzed for each mouse) (*p = 0.0262, two-tailed unpaired t-test, BPA−/− vs. BPA+/+). cRepresentative IHC staining of Ki67 on tumors collected at Breslow ≤2 mm (average = 1.79 ± 0.28 mm; BPA+/+ n = 3, BPA−/− n = 3) and ≥4 mm (average = 4.24 ± 0.69 mm; BPA+/+ n = 7, BPA−/− n = 7). dQuantification of Ki67+ cells shown in c. Each data point represents one mouse and corresponds to the average ratio ± SEM of Ki67+ cells on nuclei count for each mouse (**p = 0.0045 BPA+/+ ≥ 4 mm vs. BPA+/+ ≤ 2 mm; **p = 0.0015 BPA−/− ≥ 4 mm vs. BPA−/− ≤ 2 mm, two-way ANOVA). e Representative WB analysis and quantification (box plot) of Cyclin D1 in BPA+/+ (n = 4) and BPA−/− (n = 4) bulk tumor lysates. ß-tubulin was used as loading control (****p < 0.0001, two-tailed unpaired t-test, BPA−/− vs. BPA+/+). f Representative IHC staining of Cyclin D1 and p-c-Myc-S62 in FFPE-tumor sections of BPA+/+(n = 6) and BPA−/− (n = 6) mice. g Quantification of Cyclin D1 and p-c-Myc-S62 signals shown in f (*p = 0.0117; ***p = 0.0002, two-tailed unpaired t-test, BPA−/− vs. BPA+/+). Box-and-whisker plots shown in b, e, g represent values from minimum to maximum. Top and bottom whiskers denote upper (Q3) and lower (Q1) quartiles, respectively. Boxes refer to interquartile ranges. Medians are denoted as horizontal line in the middle of the boxes.

Ambra1 deficiency promotes an invasive phenotype in Braf V600E/+;Pten−/− mice

Next, to gain further insights into the processes affected by Ambra1 deficiency, we performed RNA sequencing of BPA+/+ and BPA−/− tumors at the 42-day endpoint. Gene ontology analysis revealed that a number of processes, namely extracellular matrix organization (ECM), integrin signaling, nervous system development and focal adhesion genes were significantly upregulated in BPA−/− tumors, as confirmed by various ontology libraries (p < 0.01) (Fig. 3a; Supplementary Fig. 3a–d and Supplementary Data 1). Gene set enrichment analysis (GSEA) also displayed an enrichment of gene sets related to invasiveness (invasive melanoma signature from Verfaillie et al.19), ECM organization (GO:0030198), epithelial-to-mesenchymal transition (EMT) (Hallmark gene set) and assembly of focal adhesions (GO:0005925) in BPA−/− melanomas (Fig. 3b). Prompted by these findings, we evaluated the ECM structure of BPA+/+ and BPA−/− tumor sections by Picrosirius Red (PR) staining. In general, tumors variably present areas with thick confluent collagen fibers, similar to the normal reticular dermis, areas where the thick collagen fibers were isolated and separated by a thin and fibrillary stroma, and areas with only thin fibrillary stroma. Histopathological analyses revealed that BPA−/− melanomas had a predominant fine fibrillar stroma, whereas BPA+/+ tumors more often displayed longer and thicker collagen fibers (Fig. 3c, d). An upregulation of ECM-related genes, including ECM-degrading proteases, was also evident from RNAseq analyses of BPA−/− melanomas (Fig. 3e), as further validated by quantitative reverse transcription PCR (RT-qPCR) for a selection of genes (Supplementary Fig. 4a). Notably, gelatin invadopodia assay showed that human SK-Mel-5 melanoma cells silenced for AMBRA1 (siAMBRA1) were characterized by a higher ability to degrade gelatin matrix, when compared to control (siScr) cells (Fig. 3f), thus highlighting possible implications of AMBRA1 deficiency in increased invasive and migratory capacity of melanoma cells. To further unravel this aspect, we performed wound-healing and transwell migration assays both ex vivo, i.e., in melanoma cells isolated from BPA+/+ and BPA−/− mice (Bdmc+/+ and Bdmc−/−, respectively) (Supplementary Fig. 4b, c), and in vitro, i.e., in human melanoma cell lines upon siAMBRA1-mediated knockdown (Fig. 3g, h and Supplementary Fig. 4d–g). In all different setting combinations, we observed that AMBRA1-deficiency elicited faster kinetics of wound closure as well as increased migratory capacity. The gain of invasive capability was also associated with an increased expression of EMT genes (e.g., CDH2, FN1, VIM, SNAI1), as revealed by both RT-qPCR and WB analyses of bulk tumors (Fig. 3i, j; Supplementary Fig. 4h), Bdmc+/+ and Bdmc−/− cells (Supplementary Fig. 4i) and siAMBRA1human melanoma SK-Mel-5 (Fig. 3k, l and Supplementary Fig. 4j), MeWo and SK-Mel-2 cells (Supplementary Fig. 4k–m).

Fig. 3: Ambra1 deletion induces an invasive phenotype in BrafV600E/+;Pten−/− mice.

a Upregulated processes (p < 0.01) and b GSEA (NES, normalized enrichment score; FDR, false discovery rate) in BPA−/− (n = 3) vs. BPA+/+ (n = 3) tumors. c Representative images and dhistopathological evaluation of Picrosirius Red staining in BPA+/+ (n = 6) and BPA−/− (n = 6) tumors. eDifferential expression of selected ECM-related genes in tumors (n = 3 each group, FC > 2, padj < 0.05.) f Gelatin degradation in AMBRA1-silenced SK-Mel-5 cells. F-Actin and nuclei were counterstained using Phalloidin and Hoechst. Images are representative. Each dot represents the average count per field ± SD (n = 3; **p = 0.0022; ****p < 0.0001, one-way ANOVA). g Wound closure area in AMBRA1-silenced SK-Mel-5. Quantification is shown as percentage ± SD vs. T0 at indicated times. White and yellow lines outline wound edge at T0 and at the times indicated. Images are representative (n = 4; **p = 0.0021 6 h; *p = 0.026 9 h; ****p < 0.0001 12 h, 18 h and 24 h, siAMBRA1#1vs. siScr. n = 3; °p = 0.0234 6 h; °°°°p < 0.0001 9 h, 12 h, 18 h and 24 h, siAMBRA1#2 vs. siScr, two-way ANOVA). h Count of migrating cells in AMBRA1-silenced SK-Mel-5. Data are expressed as mean ± SD, images representative (n = 3; *p = 0.0289 siAMBRA1#1 vs. siScr, *p = 0.0454 siAMBRA1#2 vs. siScr, one-way ANOVA). WB in fh indicate silencing efficiency. i RT-qPCR analyses of Cdh1, Cdh2, Zeb-1, Snai1, Vim, Fn1 and Ctnnb1 in BPA−/− (n = 48) and BPA+/+ (n = 4–8). Data normalized to L34 are expressed as fold-change (*p = 0.0158 Chd1; *p = 0.0417 Chd2; **p = 0.0096 Zeb-1; ***p = 0.0006 Snai1; *p = 0.0239 Vim; *p = 0.0458 Fn1, two-tailed unpaired t-test, BPA−/− vs. BPA+/+). j WB analyses of Cdh1, Cdh2, ß-Catenin, Vimentin and Snai1 in bulk tumors (n = 3 each group). k RT-qPCR analyses of CDH2, FN1, VIM and SNAI1 in AMBRA1-silenced SK-Mel-5. Data normalized to L34 are expressed as fold-change ± SD (n = 4; **p = 0.0045 CDH2; **p = 0.0022 FN1; *p = 0.0353 VIM; ***p = 0.0005 SNAI1, two-tailed unpaired t-test, siAMBRA1 vs. siScr, dashed line). l Representative (n = 4) WB analyses of CDH1, CDH2, ß-Catenin, VIM and SNAI1 in AMBRA1-silenced SK-Mel-5. Box-and-whisker plots in f, irepresent minimum-to-maximum values. Top and bottom whiskers denote upper (Q3) and lower (Q1) quartiles, respectively. Boxes refer to interquartile ranges. Medians are denoted as horizontal line in the middle of the boxes.

Altogether, these results indicate that loss of Ambra1 promotes melanoma invasion by inducing ECM remodeling and an EMT-like phenotype.

Ambra1 deficiency promotes melanoma metastasis in mouse models

In order to address the relevance of these findings in vivo, we analyzed the ability of Ambra1-deficient melanoma cells to colonize other organs. Macroscopic observations and histological analyses revealed that inguinal lymph nodes (iLNs) of BPA−/− mice harbored larger pigmented areas compared to BPA+/+ mice at 42-day endpoint (Fig. 4a–d), and showed a higher number of S100+ melanoma cells (Fig. 4c, d). Interestingly, histological analyses of iLNs of BPA+/+ mice at maximum tolerated tumor size (59 days versus 42 days of BPA−/−) showed a trend of increase in the pigmented areas of BPA+/+ similar, yet lower than that observed in BPA−/− mice (Fig. 4a–d). This evidence supports the hypothesis that Ambra1deletion accelerates melanoma progression and promotes the invasive state of melanoma. However, when the maximum tolerated size of the primary tumors was reached, no lung metastases could be observed in either of the two groups. Therefore, to assess whether distant organs were more efficiently colonized by Ambra1-deficient melanoma cells, we took advantage of a syngeneic mouse model, i.e., C57Bl/6 mice injected with freshly dissociated cells from BPA+/+ and BPA−/− tumors through the tail vein (sBPA+/+ and sBPA−/−, respectively) (Fig. 4e). The health status of sBPA+/+ and sBPA−/− mice was monitored for 3 months, after which they were euthanized and lungs collected. Although metastases were not macroscopically visible, histological analyses revealed that lungs from sBPA−/− mice exhibited increased homing and larger nodules than sBPA+/+ mice (Fig. 4f–h), suggesting that Ambra1 depletion promotes melanoma metastasis.

Fig. 4: Ambra1 deletion promotes melanoma metastases.

a Macroscopic evaluation of inguinal lymph nodes (iLNs) collected when BPA−/− tumors reached maximum tolerated size (42 days, BPA+/+, n = 4; BPA−/−, n = 4) or at maximum allowed volume of BPA+/+ tumors (59 days, n = 4). Three representative images are shown for each group. b Average area of iLNs was calculated at 42 (BPA+/+, n = 4; BPA−/−, n = 4) and 59 (BPA+/+, n = 4) days. Each point represents the average area of the iLNs for each mouse. Data are shown as mean ± SEM (*p = 0.0172; **p = 0.0063, one-way ANOVA). c FFPE-iLNs sections were stained with H&E (left panels) and anti-S100 (IHC, middle panels; IF, rightmost panels). Areas of metastases are indicated by black arrows. Nuclei were counterstained with Hoechst in IF staining. Images are representative of each group (n= 5). d Quantification of S100 IHC. Each data point represents the average percentage of S100+ area of the iLNs of each mouse ± SEM (n = 5, **p = 0.0079; ***p = 0.0002, one-way ANOVA). e Schematic representation of the syngeneic model for lung homing of melanoma cells; 0.25 × 106 cells/mouse were intravenously injected. f Perfused FFPE-lungs were stained with H&E and areas of cells homing indicated by black arrows. Images are representative of each group (n = 4). g Total area (**p = 0.0075) and h number of masses (**p = 0.0025) in lungs of sBPA+/+ (n = 4) and sBPA−/− (n = 4) mice were quantified (two-tailed unpaired t-test, sBPA−/− vs. sBPA+/+). Box-and-whisker plots shown in b, d, g, h) represent values from minimum to maximum. Top and bottom whiskers denote upper (Q3) and lower (Q1) quartiles, respectively. Boxes refer to interquartile ranges. Medians are denoted as horizontal line in the middle of the boxes.

Ambra1 deficiency induces hyperactivation of FAK1 signaling in melanoma

Bulk RNAseq and GSEA analyses showed that focal adhesion assembly genes (GO:0005925) were upregulated in Ambra1-depleted tumors (Fig. 3a, b and Supplementary Fig. 3d). Recently, it has been demonstrated that AMBRA1 controls the focal adhesion kinase 1 (FAK1)-mediated signaling pathway by interacting with FAK1 and SRC in mouse squamous cell carcinoma cells20. Therefore, we investigated whether the invasive phenotype observed in Ambra1-deficent tumors correlated with the phospho-activation of FAK1 signaling axis, which was already shown to contribute to a more aggressive phenotype in melanoma21,22,23,24. WB analyses revealed a marked increase of the phospho-active forms of FAK1 (pFAK-Y397 and -Y576) and SRC (pSRC-Y416) in BPA−/− tumors (Fig. 5a and Supplementary Fig. 5a) and in human melanoma cells silenced for AMBRA1 (Fig. 5b and Supplementary Fig. 5b–d), as well as in iLN metastasis of BPA−/− (Supplementary Fig. 5e) and in melanocytic nevi of BA−/− mice (Supplementary Fig. 2f, i). Immunofluorescence analyses further confirmed the hyperactivation of FAK1 in siAMBRA1 SK-Mel-5 cells (Fig. 5c and Supplementary Fig. 5f) and revealed that pFAK-Y397 signal colocalized at the focal adhesion (FA) with FA-associated vinculin at the edge of F-Actin fibers (Fig. 5c and Supplementary Fig. 5f). Of note, AMBRA1 deficiency was associated with an increased size and length and a decreased shape factor of FAs, a feature of cancer cells characterized by a faster motility (Fig. 5d–g and Supplementary Fig. 5g–i)25,26.

Fig. 5: Ambra1-deficient melanoma displays FAK1 signaling hyperactivation.

a WB analyses of pFak-Y576, pFak-Y397, Fak1, pSrc-Y416 and Src in 42-day BPA+/+/BPA−/− and in 59-day BPA+/+ tumors (n = 3 each group) and b in AMBRA1-silenced SK-Mel-5 (images are representative of n = 4). c Representative (n = 3) IF staining of pFAK-Y397 in AMBRA1-silenced SK-Mel-5. Phalloidin, Vinculin, and Hoechst were used to counterstain F-Actin fibers, cytoskeleton and nuclei. Rightmost panels display thresholds used for quantifying d average size (***p = 0.001), e length (Feret) (***p = 0.0003), f perimeter (***p = 0.0005) and g shape factor (***p = 0.0098) of FAs (n = 3, two-tailed unpaired t-test, siAMBRA1 vs. siScr). h Representative (n = 3) WB analyses of cleaved PARP-1 (CL-PARP-1) and CASP-3 (CL-CASP3) (Long exposure: red bands identify signal saturation) in AMBRA1/FAK1 co-silenced SK-Mel-5. i Wound closure was assessed in the conditions specified in hand shown as percentage ± SD vs. T0 at indicated times. White and yellow lines outline wound edge at T0 and at 24 h. Statistics and representative images are provided at 24 h (n = 3; *p = 0.037; ****p < 0.0001, two-way ANOVA). j Count of migrating cells in AMBRA1/FAK1 co-silenced SK-Mel-5. Data are expressed as mean ± SD, images representative (n = 3; *p = 0.0246 siScr:siAMBRA1 vs. siScr:siScr; *p = 0.0116 siFAK1:siAMBRA1 vs. siScr:siAMBRA1; *p = 0.0304 siScr:siFAK1 vs. siScr:siScr, two-way ANOVA). k Schematic representation of FAKi treatment. l Tumor growth kinetics in Vehicle- (n = 5) or FAKi- (n= 5) treated sBPA−/− mice. Data points represent average volume vs. day 0 ± SEM (**p = 0.0022; ****p < 0.0001, two-way ANOVA). m Weight of sBPA+/+:Vehicle (n = 8), sBPA+/+:FAKi (n = 5), sBPA−/−:Vehicle(n = 10) and sBPA−/−:FAKi (n = 9) tumors at sacrifice (***p = 0.0007; ****p < 0.0001, two-way ANOVA). n Wound closure of Bdmc+/+ (n = 3) and Bdmc−/− (n = 3) cells treated with 1 µM FAKi for 24 h. Lines are defined as in i, data shown as percentage ± SEM vs. T0 at 24 h, images representative (***p = 0.0003; ****p < 0.0001, two-way ANOVA). o Count of migrating Bdmc+/+ (n = 3) and Bdmc−/− (n = 3) cells treated as in n. Data are expressed as mean ± SEM, images representative (**p = 0.0047; ****p < 0.0001, two-way ANOVA). Box-and-whisker plots in dg, m represent minimum-to-maximum values. Top/bottom whiskers denote upper (Q3)/lower (Q1) quartiles. Boxes refer to interquartile ranges. Medians are denoted as horizontal line in the middle of the boxes.

Immunoprecipitation analyses of both endogenous (Supplementary Fig. 5j) and AMBRA1-myc-tagged-expressing (Supplementary Fig. 5k) extracts of SK-Mel-5 cells confirmed the previously reported interaction of FAK1 and AMBRA120. Intriguingly, when we expressed the FAK1P876A/P882A mutant (HA-FAK1.miResAA)—previously reported to abrogate the interaction with AMBRA120— in FAK1 depleted SK-Mel-5 melanoma cells (siFAK1), we observed a reduction in AMBRA1/FAK1 interaction (Supplementary Fig. 5l). This phenomenon was accompanied by an increased phospho-activation of FAK1 signaling (Supplementary Fig. 5m), and resembled what we previously observed in AMBRA1-silenced cells (Fig. 5b), suggesting that the abrogation of the FAK1/AMBRA1 interaction is responsible for FAK1 hyperphosphorylation.

It was previously reported that PP2A interacts with AMBRA111 and regulates focal adhesion27,28. Based on this evidence, we sought to investigate whether PP2A activity could contribute to the increased phosphorylation of both FAK1 and SRC via AMBRA1. To this end, we knocked-down AMBRA1 levels in SK-Mel-5 by siRNA selectively targeting the 5’-UTR region of AMBRA1 messenger RNA (mRNA) (siAMBRA1 #2) and, next, induced the expression of the PXP mutant of AMBRA1 (AMBRA1PXP), which has already been reported to abrogate the binding of PP2A to AMBRA111,16. The expression of AMBRA1PXP, as well as the re-expression of WT AMBRA1 (AMBRA1WT), did not affect the phosphorylation state of either FAK1 or SRC (Supplementary Fig. 5n), thus excluding any involvement of AMBRA1-related PP2A activity in the regulation of FAK1 phosphorylation upon AMBRA1 depletion.

Finally, to evaluate whether the AMBRA1 pro-autophagic function participates to the hyperactivation of FAK1, we silenced the key autophagy gene ATG7 (siATG7) in melanoma cells. Of note, neither the activation of FAK1 signaling (Supplementary Fig. 6a–c) nor an increase of the invasive and migratory propensity were observed in siATG7 cells (Supplementary Fig. 6d–i), suggesting that these phenomena were not related to autophagy.

Overall, these results indicate that AMBRA1 regulates FAK1 signaling in melanoma cells independently from its PP2A- and autophagy-related functions.

Ambra1-mediated regulation of FAK1 affects melanoma progression

In order to understand whether or not FAK1 played a role in the invasive and migratory ability observed upon AMBRA1 deficiency (Fig. 3g, h), we performed wound healing and transwell migration assays in siAMBRA1 melanoma cells concomitantly downregulating FAK1 (siAMBRA1:siFAK1 double knocked down) (Fig. 5h and Supplementary Fig. 7a). Optic microscopy analyses indicated that FAK1 silencing caused a rescue of the phenotype distinctive of siAMBRA1 cells (Fig. 5i, j and Supplementary Fig. 7b, c), which, however, was unrelated to any apoptotic events (Fig. 5h and Supplementary Fig. 7a). On the other hand, the re-expression of the HA-FAK1.miResAA mutant resulted in a higher migratory capacity than that induced by WT FAK (HA-FAK1.miRes) (Supplementary Fig. 7d), supporting the relevance of the AMBRA1-FAK1 interaction in determining the invasive ability of melanoma cells in our settings.

Next, in order to address the therapeutic relevance of this finding, we tested whether the pharmacological inhibition of FAK1 (FAKi) could specifically affect the tumor growth in mice subcutaneously injected with melanoma cells derived from BPA−/− (sBPA−/−) tumors (Fig. 5k). Since sBPA+/+ mice were refractory to develop tumors (Fig. 1e and Supplementary Fig. 2d), we decided to increase the numbers of cells (up to 5 × 106) to produce detectable tumors in both genetic backgrounds. Remarkably, the kinetics of tumor growth was dramatically reduced upon FAKi administration in sBPA−/− mice (Fig. 5l and Supplementary Fig. 7e–g), as also demonstrated by reduced tumor weight at the time of collection (Fig. 5m), underlining the key role of the FAK1 signaling axis in sustaining Ambra1-deficient tumors.

To determine whether the inhibition of FAK1 could also affect the invasive/migratory capacity ex vivo, Bdmc+/+ and Bdmc−/− cells were treated with the maximum non-lethal dose of FAKi (1 µM) for 24 h (Supplementary Fig. 7h). Consistent with our hypothesis, wound-healing and transwell migration assays revealed a statistically significant rescue of the migrative and invasive phenotypes only in Bdmc−/− cells treated with FAKi (Fig. 5n, o).

Overall, these findings suggest that AMBRA1-mediated regulation of FAK1 signaling affects the invasive capacity of melanoma cells and impacts tumor growth in vivo.

Low expression of AMBRA1 in human melanoma is associated with an invasive phenotype and sensitivity to FAK inhibition

To strengthen the relevance of our findings, 17 human melanoma cell lines were examined for AMBRA1 expression. Nine of these cell lines were characterized by reduced expression of AMBRA1 (Fig. 6a, b). Transwell migration and wound-healing assay of selected AMBRA1 High- and Low-expressing cell lines revealed a clear inverse correlation between AMBRA1 expression and invasive/migratory capacity (Fig. 6c, d). Also, AMBRA1 expression negatively correlated with the expression of the mesenchymal genes CDH2, FN1 and VIM, whereas a positive correlation was evident for the epithelial marker CDH1 (Fig. 6e–h), hence suggesting that an active EMT-like process is associated with low AMBRA1 expression in human melanoma cells.

Fig. 6: Low AMBRA1 expression in human melanoma correlates with invasive gene signature and sensitivity to FAKi.

a WB analyses of AMBRA1 in lysates from 17 melanoma cell lines. Actin was used as a loading control. Image is representative of n = 3 gels. b AMBRA1 High and Low groups were determined by densitometric analyses of AMBRA1, expressed as AMBRA1/Actin ratio ± SD (n = 3; ****p < 0.0001, two-tailed unpaired t-test). c Count of migrating AMBRA1 High (n = 3) and Low (n = 3) cells. Data are expressed as mean ± SD and images representative (**p = 0.0026, two-tailed unpaired t-test, AMBRA1 Low vs. AMBRA1 High). d Wound closure of AMBRA1 High (n = 3) and Low (n = 3) cells is shown as percentage ± SD vs. T0 at 12 h. White and yellow lines outline wound edge at T0 and at 24 h. Statistics and representative images are provided at 12 h (*p = 0.0387, two-tailed unpaired t-test, AMBRA1 Low vs. AMBRA1 High). eh Correlative RT-qPCR analyses of AMBRA1 and e CDH1 (n = 17; r = 0.5188; p = 0.0329, two-tailed Pearson correlation), f CDH2 (n = 17; r = −0.5129; p = 0.0353, two-tailed Pearson correlation), g FN1 (n = 17; r = −0.5045; p = 0.0389, two-tailed Pearson correlation) and hVIM (n = 17; r = 0.6812; p = 0.0026, two-tailed Pearson correlation). Data were normalized to L34. iGSEA of transcriptomic data from melanoma samples of The Cancer Genome Atlas (TCGA-SKCM), melanoma cell lines of The Cancer Cell Line Encyclopedia (CCLE) datasets and melanoma samples of the Leeds Melanoma Cohort (LMC). ECM organization (GO:0030198), Invasive signature19, EMT (Hallmark gene set) and Focal Adhesion Assembly (GO:0005925) gene sets in AMBRA1_Low vs. AMBRA1_High in TCGA-SKCM (n = 70 for each subgroup), CCLE (n = 7 for each subgroup) and LMC (n = 105 for each subgroup). NES, normalized enrichment score; FDR, false discovery rate. jCorrelative dose-response analyses between AMBRA1 expression and sensitivity to FAKi (IC50, µM) in n = 15 melanoma cell lines (r = 0.7752; p = 0.0007, two-tailed Pearson correlation). k Area of wound closure of two selected AMBRA1 High and Low cells upon FAKi, shown as percentage ± SD with respect to T0 at 24 h. (n = 3; ***p = 0.0003 for ED043; ***p = 0.0001 for ED129, two-tailed unpaired t-test, FAKi vs. Vehicle).

Transcriptomic data from melanoma samples of The Cancer Genome Atlas (TCGA-SKCM) dataset (n = 473), melanoma cell lines of The Cancer Cell Line Encyclopedia (CCLE) dataset (n = 49) and primary melanoma samples of the Leeds Melanoma Cohort (LMC, n = 703)29were also analyzed. In all datasets, samples were ranked according to their AMBRA1 mRNA levels. AMBRA1_High and AMBRA1_Low subgroups were defined considering the top 15% and the bottom 15% of the population in terms of AMBRA1 mRNA expression, respectively. Intriguingly, GSEA of all datasets revealed that AMBRA1_Low subgroups were enriched for ECM, EMT and FA genes, as well as for genes linked to invasiveness in melanoma19 (Fig. 6i).

Finally, we sought to assess whether the expression levels of AMBRA1 could differently affect the sensitivity of human melanoma cell lines to FAKi. To this aim, we treated a panel of 15 melanoma cell lines with increasing doses of FAKi for 24 h (Supplementary Fig. 8a) and revealed a striking correlation between AMBRA1 protein expression levels and sensitivity to FAKi (Fig. 6j). Moreover, we also observed that non-lethal doses of FAKi were effective (Supplementary Fig. 8b, c) and distinctly altered the invasive/migratory capacity of selected AMBRA1-low-expressing human melanoma cell lines (Fig. 6k and Supplementary Fig. 8d).

Overall, this evidence indicates that low expression of AMBRA1 correlates with an invasive phenotype and increased sensitivity to FAKi in human melanoma cells.


The results herein reported argue for a major role of Ambra1 in the different phases of melanoma development, as demonstrated in the Braf/Pten mouse models of melanoma, which recapitulate key pathophysiological aspects of the human disease15. In particular, we found that Ambra1 deficiency accelerates tumor growth and leads to an earlier invasive and aggressive phenotype of Braf/Pten melanoma. Specifically, we have shown that Ambra1-deficient melanomas are larger and display faster growth kinetics, with reduced overall mice survival, and melanoma cells showing higher capability to affect ECM architecture, migrate and invade, eventually enhancing melanoma metastasis (Fig. 7).

Fig. 7: Ambra1 impacts on melanoma development.

Ambra1 deficiency increases the formation of nevi by affecting melanocytes proliferation in BrafV600E mutated mice. When melanoma is induced upon BrafV600E and Pten ablation, the absence of Ambra1 promotes tumor growth and metastasis by increasing cell migration and invasion, activating the EMT process, remodeling the ECM and triggering the FAK1 signaling pathway.

AMBRA1 exerts most of its functions as a scaffold protein, interacting with different partners and modulating various cellular processes, including apoptosis, autophagy and cell proliferation10,12. Consistent with AMBRA1 involvement in many aspects of a cell life, it has emerged that mouse haploinsufficiency of Ambra1 promotes spontaneous tumorigenesis in several organs. Indeed, Ambra1’s role as tumor suppressor gene has been directly linked to its interaction with PP2A and DDB1/Cullin4 with their respective regulation of c-Myc11 and Cyclin D112, thus indicating Ambra1 as a negative regulator of cell proliferation. Considering that we observed: (i) increased hyperpigmentation in the murine skin of mice with BrafV600Ebackground, and (ii) large-sized and early tumors in Braf/Pten-driven mouse models of melanoma upon Ambra1 deficiency, we investigated and related the proliferative advantage of Ambra1-deficient melanocytes and melanoma to both c-Myc and Cyclin D1.

It is well known that melanoma is characterized by a high tumor heterogeneity and plasticity and that a phenotype rewiring from a high-proliferative/low-invasive to a low-proliferative/high-invasive state may occur30,31. These two states are associated with different phases of melanoma progression and response to therapy32,33. In particular, it has been shown that phenotype switching towards a slow-cycling/high-invasive state correlates with increased melanoma aggressiveness, metastasis formation and resistance to BRAF inhibitors (BRAFi)33,34,35,36,37. Of the highest importance, our data clearly show that loss of Ambra1 not only accelerates the tumor growth, but also promotes the invasive state of melanoma. Indeed, Ambra1-deficient tumors displayed increased metastatic dissemination to local lymph nodes and increased homing of melanoma cells to the lungs. Consistently, AMBRA1-silenced cells were more prone to invade and migrate. Phenotype switching in melanoma is regulated by key transcriptional factors, e.g., the melanoma transcriptional master regulator microphthalmia-associated transcription factor (Mitf)38,39 and the nerve growth factor receptor CD271 (Ngfr)40, the latter promoting an invasive and metastatic cell behavior accompanied by upregulation of neural crest stem cell genes32,34,35,39. Consistent with the anti-invasive role of Ambra1, Ambra1-deficient tumors show an increase in the expression of Ngfr and other neural crest-related genes, which was confirmed by the upregulation of the nervous system development GO process (Fig. 3a and Supplementary Fig. 3e). This suggests that, at least in the case of melanoma, the pro-invasive phenotype of Ambra1-null cells may associate with tumors in a more de-differentiated state.

The more invasive state of Ambra1-deficient tumors was accompanied by other key processes specific to high-invasive melanomas, such as ECM proteolysis/remodeling, EMT activation and FA activation19,23,41. Schoenherr and colleagues have recently shown that AMBRA1 is involved in SRC/FAK1-dependent polarization and chemotactic invasion of cancer cells through its interaction with FAK1 and several trafficking proteins, thus regulating adhesion and migration20. In agreement with these data, here we have shown the in vitro and in vivo relevance of Ambra1-mediated regulation of FAK1 in melanoma development. Our results reveal that FAK1 signaling is hyperactivated in BPA−/− melanomas, either directly or indirectly contributing to most of the phenotypes observed. Specifically, FAK1 signaling activation in cancer can modulate a plethora of oncogenic processes including EMT, cell motility and metalloproteases (MMPs) expression42. Consistently, we have shown that Ambra1 loss deeply affects the melanoma microenvironment, marked by the upregulation of ECM-related genes, such as Mmps and lysyl-oxidases (Loxl). The role of these enzymes in modifying ECM architecture and promoting cancer progression has already largely been demonstrated for various tumor types41,43,44. Along with this, an upregulation of several types of collagen and an altered ECM architecture were observed in Ambra1-null melanomas, thus arguing for a more tumor-permissive environment44,45. As for this aspect, we have provided evidence that AMBRA1-deficient melanoma cells efficiently degrade gelatin matrix, thus pointing to their direct involvement in ECM remodeling. Although the exact mechanism(s) by which Ambra1 may regulate the ECM structure are still elusive, the massive modulation of gene expression observed in bulk RNAseq of BPA−/−tumors strongly suggests that a transcriptional rewiring occurs upon Ambra1 knock out. Of note, though these changes could be largely related to the stromal composition of the tumor (e.g., fibroblasts), the evidence that increased expression of EMT markers occurs both in human melanoma cells silenced for AMBRA1 and in Ambra1 KO tumor-derived primary cells argues for a contribution of melanoma cells. In line with this, it is worth mentioning that a novel transcriptional function of Ambra1 in the nucleus has recently been reported46. Indeed, its association to chromatin and interaction with transcriptional factors may likely hint at a direct contribution of Ambra1 to the transcriptional reprogramming observed in Ambra1-deficient tumors.

Autophagy was suggested to have a role during melanoma development and progression47,48. However, we were not able to correlate either the pro-invasive state of BPA−/− melanoma or the hyperactivation of FAK1 signaling to the AMBRA1 pro-autophagy functions7. In fact, while autophagy flux was affected in AMBRA1-silenced cells (Supplementary Fig. 6j, k), bulk autophagy was not altered in Ambra1-null Braf-mutated melanocytes (Supplementary Fig. 2f, j) and melanomas (Supplementary Fig. 6l, m), thus implying compensatory events taking place in the tumor.

On the other hand, we demonstrated that the pharmacological inhibition of FAK1 signaling was sufficient to revert the growth of Ambra1-deficient Braf/Pten-driven melanomas in vivo. Moreover, a reduced invasiveness as well as an increased sensitivity to FAKi was observed in null or low-expressing AMBRA1 melanoma cells treated with FAKi in vitro. Although this evidence should be validated in vivo through FAK1 genetic ablation in order to circumvent possible off-target effects of pharmacological inhibition, concomitant silencing of FAK1 and AMBRA1 in vitro confirmed the reduced propensity of melanoma cells to migrate and invade. This discovery paves the way for a possible therapeutic application in AMBRA1-deficient melanomas of FAK1 inhibitors, which are already exploited in preclinical models of melanoma with acquired resistance to BRAFi49, and in clinical trials for solid tumors42,50,51. In support to this, we have also observed that AMBRA1-low-expressing human melanoma cell lines and tumors (patient data from the TCGA database and LMC cohort) exhibit upregulation of FAK1 signaling, ECM remodeling, EMT and invasiveness genes.

Taken together, our discoveries pinpoint AMBRA1 as tumor suppressor in melanoma, and suggest the potential use of FAK1 inhibitors in current melanoma therapy for AMBRA1-low tumors.

Article classification: Biological abstract
Share to:
Add:Room A11-329, 1st Floor, No.1, SBI Venture Street, Optics Valley, East Lake
New Technology Development Zone, Wuhan, China.
Certificate NO.:U18Q28010569R0S