β-Galactosylceramidase Promotes Melanoma Growth via Modulation of Ceramide Metabolism

Melanoma is a skin tumor that arises from neural crest–derived melanocytes (1). Resistant to chemotherapy and radiotherapy, metastatic melanoma represents the most lethal form of skin cancer (2). Thus, a more detailed understanding of melanocyte development and differentiation may provide valuable information about the mechanisms that contribute to melanoma progression and for the development of novel therapeutic options (3).

Sphingolipids regulate various biological processes in cancer (4). In particular, ceramide acts as a tumor suppressor, and defects in ceramide metabolism contribute to tumor cell survival and resistance to chemotherapy (5).

In melanoma, the expression of acid sphingomyelinase, which catalyzes the breakdown of sphingomyelin to ceramide and phosphocholine, affects cell proliferation, survival, and dissemination and correlates inversely with tumor stage (6). In addition, downregulation of sphingomyelin synthase 1, which metabolizes ceramide into sphingomyelin, represents a worse prognosis biomarker in metastatic melanoma (7). Conversely, downregulation of the lactosylceramide synthase β4-galactosyltransferase 5 suppresses the tumorigenic potential of murine melanoma cells (8), whereas ceramide synthase 6 expression is related to human melanoma malignancy (9). Moreover, upregulation of neutral sphingomyelinase activity, with a subsequent increase of intracellular ceramide, elicits melanoma cell death (10). Thus, disturbance of ceramide metabolism exerts a deep impact on melanoma and may represent a target for anticancer therapy.

β-Galactosylceramidase (GALC; ref. 11) is a lysosomal hydrolase that catalyzes the removal of β-galactose from β-galactosylceramide and other sphingolipids, including galactosylsphingosine (psychosine). GALC deficiency causes globoid cell leukodystrophy (OMIM #245200), a sphingolipidosis characterized by degeneration of oligodendroglia and progressive demyelination (12). Two GALC co-orthologs (named galca and galcb) are coexpressed in zebrafish (Danio rerio) embryo (13). Both genes encode for lysosomal enzymes endowed with GALC activity, and double galca/galcb knockdown provides evidence for a role of GALC activity in brain development in zebrafish (13).

Here, we found that galcb is transiently expressed by melanoblasts/melanocytes in zebrafish embryos and that its downregulation affects the differentiation of these cells. In keeping with the correlation observed among genes that regulate melanocyte development and those that contribute to malignant melanoma (3, 14), we measured high levels of Galc expression in B16-F10 cells, a prototypic murine melanoma model. Notably, Galc-silenced B16-F10 cells showed a significant decrease in their tumorigenic and metastatic activity. This was paralleled by the accumulation of ceramide and upregulation of sphingomyelin phosphodiesterase 3 (Smpd3) that encodes for the ceramide-producing enzyme neutral sphingomyelinase 2. Consistent with these observations, modulation of GALC expression affected the tumorigenic activity of human melanoma A2058 cells. Accordingly, RNAscope analysis and tumor data mining revealed that the levels of GALC expression are related to human skin melanoma progression. Altogether, our observations show that GALC affects the tumorigenic activity of melanoma cells and may represent a novel target for melanoma therapy.

Wild-type AB, casper (15), and Tg(kdr:EGFP) (16) zebrafish lines were maintained at 28°C on a 14-hour light/10-hour dark cycle. Fertilized eggs were incubated at 28°C. Embryos were staged and maintained in 0.003% 1-phenyl-2-thiourea (Sigma) starting from 24 hours after fertilization (hpf) unless specified otherwise. Zebrafish experiments were performed as approved by the local animal ethics committee (OPBA, Organismo Preposto al Benessere degli Animali, Università degli Studi di Brescia, Italy).

Whole mount in situ hybridization (WISH) was carried out on 4% paraformaldehyde-fixed zebrafish embryos (13). Digoxigenin-labeled RNA probes were transcribed from linear cDNA constructs, and hybridizations were carried out overnight at 68°C (13). The expression of the genes under investigation following galcb knockdown was scored blindly as normal or downregulated by two independent observers, and data were analyzed by the χ² test.

Gene knockdown experiments were performed as described (13). Specific galcb-morpholino (MO; 5′-GCAGAGTTTACCTGTAGTCTG-3′) targeting the exon 3 of zebrafish transcript was injected (1.4 pmol) into the yolk of 1- to 4-cell stage embryos. Std-MO (5′-CCTCTTACCTCAGTTACAATTTATA-3′) was used as control. Sense mRNA encoding full-length galcb was transcribed in vitro from pGEM-T Easy/galcb vector using mMESSAGE mMACHINE kit and Poly(A) Tailing Kit (Ambion).

B16-F10 and A2058 melanoma cells were obtained from ATCC-LGC Standards Repository (ATCC number CRL-6475 and CRL-2731, respectively), maintained at low passage, tested regularly for Mycoplasma negativity, and authenticated by Power-Plex Fusion System (Promega). B16-F10 cell subcloning was performed as described in Supplementary Information. B16-F10 and A2058 cell silencing was carried out by transduction of lentiviral particles encoding mouse Galc or human GALC targeting short hairpin RNA, respectively. For GALC overexpression, A2058 cells were infected with a lentivirus harboring the murine Galc cDNA. See Supplementary Information for more details.

B16-F10 cells were seeded at 10⁴ cells/cm² in medium supplemented with 0.4% FBS. After 24 hours (T₀), fresh medium was added and cells were counted 24 to 72 hours thereafter. When described, cells were treated with GW4869 (Sigma-Aldrich; dissolved in ethanol) in medium containing 2.0% FBS and counted 24 hours thereafter. A2058 cells were seeded in 24-well plates at 10⁴ cells/cm² in medium added with 1.0% FBS. After 24 hours (T₀), cells were added with fresh medium and counted 24 and 48 hours thereafter.

Note that 500 or 5,000 cells were seeded in 35 mm culture dishes or embedded in Matrigel (Cultrex BME) and seeded in 24-well plates. After 15 days, cell colonies were observed under a phase contrast microscope and photographed.

Confluent B16-F10 cells were scraped with a 1,000 μL tip to obtain a mechanical wound and maintained in DMEM, supplemented with 0.4% FBS. Cells at the leading edge of the wound were observed by time-lapse microscopy as described in Supplementary Information.

B16-F10 cells stained with CellTracker Red CMTPX Dye were seeded at 10⁴ cells/well on a murine 1G11 endothelial cell monolayer stained with CellTracker Green CMFDA Dye (Thermo Fisher Scientific). Time-lapse microscopy was performed under an inverted Zeiss Axiovert 200 M photomicroscope (one snap photograph every 15 minutes), and the percentage of invading cells was calculated after 12 hours.

GALC-mediated hydrolysis of the fluorescent GALC substrate LRh--6-GalCer (N-lissamine rhodaminyl-6-aminohexanoylgalactosyl ceramide) following its incubation with 30 μg of cell extract was quantified by thin-layer chromatography (TLC; ref. 13).

B16-F10 cells were analyzed by RT-qPCR using the primers listed in Supplementary Table S1, and data were normalized for Gapdh expression (17). See Supplementary Information for more details.

Lipids were extracted and analyzed as previously described with minor modifications (18, 19). Separation of GlcCer and GalCer was attempted as reported by Nakajima and colleagues (20) with minor modifications.

shNT-B16-F10 and shGALC-B16-F10 cells were stained with Cell Tracker Red CMTPX Dye or CellTracker Blue CMF2HC Dye (Thermo Fisher Scientific), respectively, and mixed at 1:1 ratio (5 × 10³ cells/μL). Then, 4 nL containing 20 cells for each cell type were injected into the blood circulation of transgenic Tg(kdr:EGFP) zebrafish embryos at 48 hpf. Transendothelial migration in intersomitic vessels was evaluated at different hours post injection (hpi; ref. 21).

B16-F10 cells were stained with CellTracker Green CMFDA Dye or CellTracker Red CMTPX Dye. Then, 80 to 100 cells/embryo were injected at 48 hpf into the blood circulation of wild-type AB or Tg(kdr:EGFP) zebrafish embryos, respectively. During the next 3 days, the growth of metastases in the tail vascular plexus was quantified by fluorescence microscopy followed by computerized image analysis (22). Metastatic area data were analyzed by two-way ANOVA corrected by the Bonferroni test.

C57BL/6 and NOD/Scid mice (Charles River) and breeder twitcher heterozygous mice (C57BL/6J, twi/+; Jackson Laboratories) were maintained under standard conditions. Experiments were performed according to the Italian laws (D.L. 116/92 and following additions) that enforce the EU 86/109 Directive and were approved by the local animal ethics committee (OPBA, Università degli Studi di Brescia, Italy).

C57BL/6 or NOD/Scid mice were injected s.c. into the dorsolateral flank with 3 × 10⁴ B16-F10 cells or 3 × 10⁶ A2058 cells, respectively, and tumors were measured with calipers. When indicated, tumors were harvested and processed for RT-qPCR analysis or GALC, SMPD3, and ceramide immunostaining as described in Supplementary Information. Tumor volume data were analyzed by two-way ANOVA corrected by the Bonferroni test.

B16-F10 cells (5 × 10⁴ cells in 100 μL of PBS) were injected into the tail vein of 7-week-old C57BL/6 mice. After 3 weeks, lungs were harvested and metastases were counted under a dissecting microscope.

Tissue samples from 60 patients who underwent curative resection were collected from the archive of the Unit of Pathology of the University of Bari, Hospital Policlinico, Bari, Italy. The study was approved by the Institutional Review Board at Hospital Policlinico (Bari) and conducted in accordance with Good Clinical Practice and the Declaration of Helsinki. Written informed consent was obtained from all patients. Tumors were divided into five histologic subgroups: common nevi (n = 10), dysplastic nevi (n = 10), melanoma stages I–II (n = 20), melanoma stage III (n = 10), and melanoma stage IV (n = 10).

RNA ISH was performed using RNAscope 2.5 HD Reagent Kit (RED 322360, Advanced Cell Diagnostics) using an Hs-GALC probe (ref. 566801), positive control probe Hs-PPIB (ref. 313901), or negative control probe DapB (ref. 310043). IHC was performed on the same 60 samples used for RNAscope analysis using mouse monoclonal anti-MITF (ref. M3621, Agilent Dako), anti-tyrosinase (ref. M3623, Agilent Dako), anti-SMPD3 (ref. sc-166637, Santa Cruz Biotechnology), and anti-ceramide (Enzo Life Science) antibodies. RNAscope measurements were defined as number of probes per cell nucleus. MITF, tyrosinase, SMPD3, and ceramide IHC staining were quantified by morphometric analysis. See Supplementary Information for more details.

The zebrafish genes galca and galcb, co-orthologs of human GALC, are coexpressed in embryonic brain during development (13). A detailed WISH analysis of galca and galcb expression in zebrafish embryos at 28 and 40 hpf revealed that galcb, but not galca, is expressed also by melanoblasts/melanocytes alongside the embryo trunk (Fig. 1A). No galcb-positive cells were instead detected along the trunk of casper zebrafish mutants that lack all melanophores (15), thus confirming the identity of these cells as melanoblasts/melanocytes (Fig. 1A). Notably, galcb expression in embryonic melanocytes is transient, being lost at 48 hpf.

To assess the impact of galcb expression on zebrafish embryo melanocytes, loss-of-function studies were performed using a specific galcb-MO that induces skipping of exon 3 of the immature galcb mRNA, leading to the formation of an enzymatically inactive protein (13). When compared with controls, galcb morphants showed a significant delay in the appearance of melanin-positive melanocytes migrating laterally onto the yolk and dorsally to the embryo tail (Fig. 1B and C). Coinjection of the galcb-MO with an excess of galcb mRNA fully rescued this defective phenotype (Fig. 1B and C). No defects in melanocyte appearance were observed in embryos injected with an anti-galca MO (Supplementary Fig. S1A; ref. 13), thus confirming the specificity of galcb-MO effects.

Zebrafish melanocytes develop from crestin-positive neural crest cells where the SRY-related HMG-box 10 (sox10) transcription factor plays a primary role in pigment cell lineage fate specification. In melanoblasts, sox10 activates the microphthalmia transcription factor (mitfa), involved in melanocyte differentiation. Finally, dopachrome tautomerase (dct) and tyrosinase (tyr) are markers of differentiating and terminally differentiated melanocytes, respectively (23). As shown in Fig. 1D and E, galcb knockdown caused a significant inhibition (P < 0.001, χ² test) of the expression of the melanoblast/melanocyte markers mitfa, dct, and tyr, with no effect on crestin and sox10 expression. No defects in mitfa and dct expression were instead observed in galca morphants (Supplementary Fig. S1B). These data point to a role for galcb in the differentiation, but not specification, of embryonic melanocytes in zebrafish.

An increasing evidence correlates genes that regulate melanocyte development with those that contribute to malignant melanoma (3, 14). These findings prompted us to assess Galc expression in B16-F10 cells, a prototypic murine melanoma model.

As shown in Fig. 2A, B16-F10 cells express high levels of Galc, similar to those observed in the adult murine brain. Accordingly, a significant GALC activity was detected in B16-F10 cell extract, and a strong GALC immunoreactivity signal is present in B16-F10 tumor grafts when compared with the peritumoral tissue.

To assess the role of Galc in this melanoma model, B16-F10 cells were infected with lentiviral particles containing a shRNA target–specific construct designed to knockdown the Galc gene. The puromycin-resistant cell population (shGALC-B16-F10 cells) showed a significant GALC downregulation in terms of Galc transcript, protein levels, and enzymatic activity when compared with cells infected with a nontargeting control shRNA (shNT-B16-F10 cells; Fig. 2B).

GALC downregulation caused the decrease in the growth rate of Galc-silenced cells when compared with shNT-B16-F10 and uninfected cells [cell duplication time (mean ± SEM) equal to 32.3 ± 1.9 vs. 17.5 ± 2.2 and 20.8 ± 1.5 hours, respectively; Fig. 2C). Consistently, shGALC-B16-F10 cells showed a reduced colony formation capacity on tissue culture plastic or in a 3D Matrigel layer (Fig. 2D). At variance, no differences in cell-cycle distribution and apoptosis were observed among uninfected, shNT-B16-F10 and shGALC-B16-F10 cells (Supplementary Fig. S2A and S2B).

To confirm the impact of Galc silencing on the growth capacity of B16-F10 cells, single-cell–derived clones were obtained from shNT-B16-F10 and shGALC-B16-F10 cell populations, and the days required by each clone to reach a cell number equal to 5.0 × 10⁶ was calculated. As shown in Fig. 2E, only 16 of 28 shGALC-B16-F10 clones showed a rapid rate of growth when compared with 25 of 27 shNT-B16-F10 clones (P < 0.01, χ² test). RT-qPCR analysis confirmed that Galc expression in shGALC-B16-F10 clones was lower than in shNT-B16-F10 clones. In addition, a significant inverse correlation was found between the levels of Galc expression in each clone and its proliferative capacity (Fig. 2F).

Next, time-lapse videomicroscopy was used to assess the effect of Galc downregulation on the motility of B16-F10 cells in a mechanical wound healing assay. When compared with uninfected and shNT-B16-F10 cells, shGALC-B16-F10 cells showed a significant decrease in their capacity to repair the wounded cell monolayer with a reduction of their velocity and accumulated distance (Fig. 2G). In addition, the capacity of B16-F10 cells to invade a murine microvascular endothelial cell monolayer was assessed in vitro (24). As shown in Fig. 2H, shNT-B16-F10 cells adhere on the top of endothelial cells, cross the endothelium, and attach to the substratum. This process was significantly impaired in shGALC-B16-F10 cells. For both cell populations, melanoma cell invasion was hampered by neutralizing antibodies against N-cadherin, a major mediator of endothelium-melanoma cell adhesion (25).

The sphingolipid, phospholipid, and neutral lipid composition of B16-F10, shNT-B16-F10, and shGALC-B16-F10 cells was investigated. Among the sphingolipids tested (see Supplementary Table S2 for a complete list of all the analyzed lipids), shGALC-B16-F10 cells were characterized by a significant increase of the levels of ceramide (Fig. 3A). This was paralleled by a decrease of sphingomyelins that occurred concomitantly to the reduction of phosphatidylethanolamines/lyso-phosphatidylethanolamines, cholesteryl esters, and phosphatidylcholines, in parallel with an increased concentration of diacylglycerols (Fig. 3A). At variance, no alterations of the levels of hexosyl- and lactosyl-ceramides, as well as of sphingosine-1-phosphate were observed following GALC downregulation. It must be mentioned that the separation of GalCer and GlcCer could not be achieved to background level. Because GlcCer is more abundant than GalCer, GlcCer levels may mask a possible difference in amounts of GalCer between control and Galc-silenced cells.

These findings prompted us to assess the levels of expression of a series of enzymes of the sphingolipid metabolism (Fig. 3B; Supplementary Table S3). In keeping with the observed alterations of the ceramide/sphingomyelin ratio, shGALC-B16-F10 cells displayed a remarkable upregulation of Smpd3 expression. Smpd3 encodes for neutral sphingomyelinase 2 that catalyzes the hydrolysis of sphingomyelin to ceramide and phosphocholine. No alteration was observed for the expression levels of the acid lysosomal sphingomyelinase Smpd1, as well as of all the other enzymes tested.

Twitcher mice are characterized by a spontaneous Galc mutation that results in a dramatic reduction of GALC activity in homozygous twi/twi animals (26). In accordance with the results obtained for shGALC-B16-F10 cells, a significant Smpd3 upregulation was observed in postnatal fibroblasts, lung, liver, and kidney of twi/twi mice when compared with wild-type animals (Fig. 3C), thus supporting the hypothesis that an inverse correlation may exist between Galc downregulation and Smpd3 upregulation.

In order to assess whether Smpd3 upregulation may contribute to the oncosuppressive activity exerted by Galc downregulation, shGALC-B16-F10 and shNT-B16-F10 cells were treated with the neutral sphingomyelinase inhibitor GW4869 (27). As shown in Fig. 3D and E, GW4869 caused a significant stimulation of shGALC-B16-F10 cell proliferation paralleled by a decrease of intracellular ceramide content, with no effect on the proliferation of shNT-B16-F10 cells. Together, these data point to a role for Smpd3 upregulation, with a consequent increase of ceramide content, in mediating the inhibitory effect exerted by Galc silencing in B16-F10 cells.

To evaluate the impact of Galc downregulation on the tumorigenic activity of B16-F10 cells, shNT-B16-F10 and shGALC-B16-F10 cells were implanted subcutaneously in syngeneic C57BL/6 mice. As shown in Fig. 4A, Galc downregulation caused a significant delay in shGALC-B16-F10 tumor growth when compared with shNT-B16-F10 lesions. Accordingly, the average weight of shGALC-B16-F10 tumors harvested 26 days after grafting was significantly smaller than that of shNT-B16-F10 lesions that expressed higher levels of Galc transcript and of the cell proliferation marker Cyclin D1 (Fig. 4A). In keeping with the in vitro observations, IHC analysis showed an increase of SMPD3 protein levels in shGALC-B16-F10 tumors when compared with controls, which was paralleled by a significant increase of ceramide levels (Fig. 4B). At variance, shGALC-B16-F10 and shNT-B16-F10 grafts did not differ in the expression levels of various immune checkpoint genes and tumor cell infiltrate markers (including monocyte/macrophage, B and T lymphocyte, natural killer cell, and granulocyte markers; Supplementary Table S4). Notably, consistent with in vitro data, in vivo administration of the neutral sphingomyelinase inhibitor GW4869 increased the rate of growth of shGALC-B16-F10 tumor grafts, further supporting the role of neutral sphingomyelinase activity in mediating the antitumorigenic effects exerted by Galc downregulation in B16-F10 melanoma cells (Supplementary Fig. S3A and S3B).

In a second set of experiments, shNT-B16-F10 and shGALC-B16-F10 cells were injected into the tail vein of C57BL/6 mice, and experimental lung metastases were counted 3 weeks after. As illustrated in Fig. 4C, Galc downregulation caused a significant decrease in the metastatic potential of melanoma cells when compared with controls.

Tumor cell extravasation represents an important step in the metastatic process. To assess whether Galc downregulation affects the capacity of B16-F10 cells to extravasate in an in vivo zebrafish embryo model (21), shNT-B16-F10 and shGALC-B16-F10 cells were stained with the red fluorescent CellTracker Red CMTPX Dye and the blue fluorescent CellTracker Blue CMF2HC Dye, respectively. Then, cells were coinjected in the blood circulation of transgenic Tg(kdr:EGFP) zebrafish embryos, and the number of cells that migrated across the fluorescent EGFP-positive endothelium of intersomitic vessels was counted at different hpi (21). In keeping with in vitro data, blue fluorescent shGALC-B16-F10 cells showed a reduced ability to extravasate when compared with red fluorescent, control shNT-B16-F10 cells (Fig. 4D). On this basis, CellTracker Red CMTPX Dye–stained B16-F10 cells were injected into the blood stream of Tg(kdr:EGFP) zebrafish embryos at 48 hpf, and the growth of micrometastases in the hematopoietic tissue of embryo tail was evaluated (22). As shown in Fig. 4E, metastatic lesions generated by shGALC-B16-F10 cells showed a reduced rate of growth when compared with shNT-B16-F10 lesions.

To confirm these findings and in order to rule out possible off target effects of the anti-Galc shRNA used to generate shGALC-B16-F10 cells, we performed a second infection of B16-F10 cells with a lentivirus harboring a different anti-Galc shRNA, generating a novel, independent shGALC-B16-F10 cell population (shGALC-bis-B16-F10 cells). As observed for shGALC-B16-F10 cells, Galc downregulation caused a significant decrease of the proliferation, colony formation, and motility of shGALC-bis-B16-F10 cells (Supplementary Fig. S4A–S4D). This resulted in a significant decrease of the tumorigenic and metastatic activity of these cells that was paralleled by an increase of SMPD3 and ceramide levels in shGALC-bis-B16-F10 tumor xenografts when compared with controls (Supplementary Fig. S4E–S4H). Collectively, these data demonstrate that Galc downregulation significantly hampers the tumorigenic activity of murine B16-F10 cells.

To further assess the impact of GALC expression in melanoma, human melanoma A2058 cells harboring the V600E B-RAF mutation (28) were infected with lentiviral particles containing a shRNA target–specific construct designed to knockdown the human GALC gene or a nontargeting control shRNA, thus generating shGALC-A2058 and shNT-A2058 cells, respectively. This resulted in a significant decrease of GALC activity in shGALC-A2058 cells when compared with controls (Fig. 5A). In parallel, a second aliquot of A2058 cells was infected with a lentivirus harboring the murine Galc cDNA, thus generating upGALC-A2058 cells that exhibited a higher GALC activity when compared with A2058 cells transduced with an empty vector (mock-A2058 cells; Fig. 5A).

Again, an inverse correlation was observed between GALC and SMPD3 expression in these A2058 cell populations (Fig. 5B). In addition, upGALC-A2058 cells showed a higher proliferative capacity in vitro or when grafted in NOD/Scid mice, whereas GALC downregulation resulted in a decreased rate of growth of shGALC-A2058 cells both in vitro and in vivo (Fig. 5C and D). Again, ceramide and SMPD3 protein levels in tumor grafts were inversely related to GALC expression in the different A2058 cell populations (Fig. 5E).

In keeping with a protumorigenic role of GALC in melanoma, Human Protein Atlas (www.proteinatlas.org) and Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo) data mining indicates that GALC expression levels in human tumor specimens are related to melanoma progression (Supplementary Fig. S5A and S5B). To confirm these observations, RNAscope analysis was performed on 60 human melanoma samples to evaluate GALC expression at different stages of tumor progression. In keeping with public data sets, morphometric analysis demonstrates a progressive increase of GALC expression starting from common nevi to stage IV melanomas that was paralleled by a progressive increase of MITF and tyrosinase immunoreactivity in the same samples (Fig. 6A and B). Independently of tumor stage, a direct correlation was observed between GALC expression levels and MITF and tyrosinase immunoreactivity (Fig. 6C). Similar results were obtained by the analysis of the gene expression levels of GALC, MITF, and TYR in 103 human primary melanoma samples downloaded from The Cancer Genome Atlas (TCGA) GDC Data Portal (Supplementary Fig. S5C). Finally, a progressive decrease of SMPD3 and ceramide immunoreactivity occurs in the human skin melanoma samples investigated starting from common nevi to stage IV melanomas (Fig. 7A and B). Remarkably, both SMPD3 and ceramide immunoreactivity levels were inversely correlated to GALC expression in these samples, ceramide levels decreasing as a function of the GALC mRNA/SMPD3 protein ratio (Fig. 7C). Accordingly, public data mining indicates that an inverse correlation might indeed exist between GALC and SMPD3 mRNA levels in human melanoma and that SMPD3 expression appears to be higher in human primary tumors when compared with melanoma metastases (Supplementary Fig. S6A–S6C).

Alterations in the metabolism of sphingolipids, including the tumor-suppressor ceramide, exert a deep impact in melanoma (6–10). GALC catalyzes the removal of galactose from terminal β-galactose–containing sphingolipids (12). The present work demonstrates that GALC critically modulates the oncogenic activity of melanoma cells in vitro and in vivo.

Our study stemmed from the observation that galcb, the zebrafish ortholog of human GALC (13), is transiently expressed in melanoblasts/melanocytes of zebrafish embryo and that its downregulation affects melanocyte differentiation in zebrafish embryos. In keeping with the correlation observed among genes that regulate melanocyte development with those that contribute to malignant melanoma (3, 14), we found that Galc is highly expressed in murine B16-F10 melanoma cells. Galc knockdown resulted in a significant decrease of the proliferative, migratory, and invasive capacity of shGALC-B16-F10 cells, paralleled by a reduced tumorigenic and metastatic activity in vivo. These findings were confirmed by a second infection of B16-F10 cells with a lentivirus harboring a different anti-Galc shRNA, generating a novel, independent shGALC-bis-B16-F10 cell population that showed a similar decrease in its tumorigenic and metastatic potential in vitro and in vivo. Similar results were observed following GALC silencing in human melanoma A2058 cells, whereas GALC upregulation conferred a more potent tumorigenic activity to these cells.

In keeping with the hypothesis that GALC may exert a pro-oncogenic activity in melanoma, cancer data mining indicates that GALC expression in human tumor specimens is related to skin melanoma progression. Accordingly, RNAscope analysis performed at different stages of human melanoma progression demonstrates a progressive increase of GALC expression starting from common nevi to stage IV melanoma.

Knockdown of galcb in zebrafish embryos causes the downregulation of genes involved in melanocyte differentiation, including mitfa and tyr. MITF plays an important role also in human melanoma progression by modulating tumor cell proliferation, survival, and metastatic capacity (29, 30). Highly variable MITF expression levels are found in melanoma cell subpopulations that confer tumor heterogeneity and plasticity (31). In addition, peripheral blood TYR mRNA and tissue tyrosinase immunoreactivity are melanoma biomarkers (32, 33). Here, we show that GALC expression correlates with MITF and tyrosinase immunoreactivity in a cohort of 60 human samples at different stages of melanoma progression. Accordingly, TCGA GDC Data Portal mining indicates that a direct correlation exists between GALC and MITF/TYR expression levels in a cohort of 103 human primary melanoma samples. Altogether, these data strongly argue for a role of GALC in human melanoma progression.

Genetic GALC deficiency causes globoid cell leukodystrophy (12), a neurodegenerative disease characterized by the accumulation of the cytotoxic sphingolipid metabolite psychosine (34). On the other hand, experimental evidence indicates that GALC downregulation may exert also psychosine-independent effects. For instance, an altered expression of the neuronal marker neuroD and increased apoptotic events are observed in double galca/galcb zebrafish embryo morphants in the absence of any significant accumulation of psychosine (13). Similarly, GALC downregulation causes a significant decrease in the mitogenic and motogenic responses of human endothelial cells to vascular endothelial growth factor with only minor modifications in intracellular psychosine levels (35). These results raise the possibility that changes in sphingolipid metabolism, other than psychosine accumulation, may explain the biological effects consequent to the modulation of GALC activity. For instance, Galc-null twitcher mice show a significant increase in the brain levels of the GALC substrate lactosyl-ceramide (36), a lipid raft component implicated in signaling events linked to cell differentiation, development, apoptosis, and oncogenesis (37). In addition, Galc silencing affects the sphingolipid profile and functionality of murine hematopoietic stem cells (38).

Here, we demonstrate that Galc silencing affects the sphingolipid, phospholipid, and neutral lipid profiles of melanoma B16-F10 cells causing a significant increase of intracellular ceramide. This was accompanied by a decrease in sphingomyelin, with no alterations of hexosyl- and lactosyl-ceramides. Various sphingolipid metabolizing enzymes, including acid sphingomyelinase (6), the lactosyl-ceramide synthase β4-galactosyltransferase 5 (8), and ceramide synthase 6 (9), modulate ceramide levels and affect the malignant behavior of melanoma cells. Our results indicate that alterations of the ceramide/sphingomyelin balance that occur in shGALC-B16-F10 cells are accompanied by the upregulation of the neutral sphingomyelinase Smpd3, an enzyme that hydrolyzes sphingomyelin to ceramide, with a consequent increase of SMPD3 and ceramide levels in the corresponding tumor grafts. Smpd3 upregulation was observed also in postnatal fibroblasts and different organs of Galc-null twitcher animals, pointing to an inverse correlation between Galc downmodulation and Smpd3 upregulation.

Experimental evidence points to SMPD3 as an oncosuppressor gene, neutral sphingomyelinase activity regulating tumor cell proliferation, survival, and response to chemotherapy (39). Accordingly, increased neutral sphingomyelinase activity, with a consequent increase of intracellular ceramide generation, affects melanoma cell survival (10). Our data demonstrate that an inverse correlation also exists between GALC and SMPD3 expression in human A2058 cells following GALC downregulation. In addition, we observed a progressive decrease of SMPD3 levels in human tumor specimens starting from common nevi to stage IV melanoma. This was paralleled by the reduction of ceramide immunoreactivity in the same lesions, ceramide levels decreasing as a function of the GALC mRNA/SMPD3 ratio. In keeping with these observations, public data mining revealed an inverse correlation between GALC and SMPD3 expression levels in human melanoma and a higher SMPD3 expression in human primary tumors when compared with melanoma metastases. In addition, our data demonstrate that the neutral sphingomyelinase inhibitor GW4869 (27) causes a significant stimulation of the tumorigenic activity of shGALC-B16-F10 cells paralleled by a decrease of intracellular ceramide levels. These data point to a role for Smpd3 upregulation, with a subsequent increase of intracellular ceramide, in mediating the inhibitory effect exerted by Galc downregulation in melanoma.

Different chemotherapeutics can regulate SMPD3 expression at the transcriptional levels via activation of Sp1 and Sp3 transcription factors (40). SMPD3 upregulation occurs in response to several activators, including TNFα (ref. 41 and references therein). In addition, SMPD3 has been identified as a p53 target gene downstream of the DNA damage pathway (42). Further experiments will be required to elucidate the molecular determinants of the inverse correlation existing between GALC and SMPD3 expression in melanoma cells.

Beside the observed increase in ceramide levels, we cannot rule out the possibility that other alterations of the lipid profile following Galc silencing may contribute to the decreased oncogenic potential of melanoma cells. For instance, the increased levels of diacylglycerols measured in shGALC-B16-F10 cells may cause the downregulation of the novel protein kinase PKCϵ and PKCδ (43), both involved in the oncogenic behavior of melanoma cells via activation of the activating transcription factor 2 (44, 45) or suppression of caspase-dependent apoptosis in NRAS/BRAF-mutated cells (46), respectively. Interestingly, recent observations have shown that SMPD3 deficiency disrupts the homeostasis of diacylglycerol, ceramide, and sphingomyelin in the Golgi compartment (47). Comprehensive lipidomic profiling may provide relevant information about the tumorigenic/metastatic potential of melanoma cells (48). In this frame, our findings set the basis for further investigations aimed at assessing the relationship between defined lipid changes and GALC expression levels and their impact on melanoma.

In summary, GALC appears to play a nonredundant role in melanoma progression by affecting the tumorigenic and metastatic potential of melanoma cells due to its impact on their sphingolipid metabolism. These observations indicate that GALC may represent a novel target for melanoma therapy.

E. Maiorano reports personal fees from ClabMeeting (honorarium for a lecture sponsored by Shire) outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

M. Belleri: Data curation, investigation, writing-review and editing, performed the experiments, analyzed the data and revised the paper. G. Paganini: Investigation, performed zebrafish experiments. D. Coltrini: Data curation, investigation, performed RT-qPCR analysis. R. Ronca: Investigation, analyzed the data and performed in vivo experiments. D. Zizioli: Investigation, performed zebrafish experiments. M. Corsini: Investigation, performed in vitro experiments. A. Barbieri: Investigation, performed zebrafish experiments. E. Grillo: Investigation, performed in vitro experiments. S. Calza: Data curation, performed bioinformatics analysis. R. Bresciani: Investigation, performed in vitro experiments. E. Maiorano: Investigation, collected clinical samples. M.G. Mastropasqua: Investigation, collected clinical samples. T. Annese: Investigation, performed RNAScope experiments and morphometric analysis. A. Giacomini: Investigation, performed in vitro experiments. D. Ribatti: Investigation, performed immunohistochemical analysis of clinical samples. J. Casas: Conceptualization, investigation, performed lipid analysis. T. Levade: Conceptualization, writing-review and editing, revised the paper. G. Fabrias: Conceptualization, investigation, performed lipid analysis. M. Presta: Conceptualization, resources, data curation, supervision, funding acquisition, writing-original draft, writing-review and editing, conceived the study and wrote the paper.

This work was supported in part by Associazione Italiana per la Ricerca sul Cancro (AIRC) IG grant n° 18493 and MFAG grant n° 18459 to M. Presta and R. Ronca, respectively.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.