Melanoma cells repress Desmoglein 1 in keratinocytes to promote tumor cell migration

Abstract

Melanoma is an aggressive cancer typically arising from transformation of melanocytes residing in the basal layer of the epidermis, where they are in direct contact with surrounding keratinocytes. The role of keratinocytes in shaping the melanoma tumor microenvironment remains understudied. We previously showed that temporary loss of the keratinocyte-specific cadherin, Desmoglein 1 (Dsg1), controls paracrine signaling between normal melanocytes and keratinocytes to stimulate the protective tanning response. Here, we provide evidence that melanoma cells hijack this intercellular communication by secreting factors that keep Dsg1 expression low in the surrounding keratinocytes, which in turn generate their own paracrine signals that enhance melanoma spread through CXCL1/CXCR2 signaling. Evidence suggests a model whereby paracrine signaling from melanoma cells increases levels of the transcriptional repressor Slug, and consequently decreases expression of the Dsg1 transcriptional activator Grhl1. Together, these data support the idea that paracrine crosstalk between melanoma cells and keratinocytes resulting in chronic keratinocyte Dsg1 reduction contributes to melanoma cell movement associated with tumor progression.

Figure 1.

Desmoglein 1 is reduced in melanoma-adjacent keratinocytes in human epidermis. (A) Paraffin-embedded sections of benign nevi and primary-stage melanoma were costained for Dsg1 or Ecad and Mel-A to highlight cadherins at cell–cell interfaces of keratinocytes surrounding pigmented cells. The dashed line indicates basement membrane. (B–E) Both distant (two to four cells away) and proximal keratinocytes were examined for cadherin staining for quantification. Pixel intensities were determined for four benign and 12 melanoma samples and the ratio of proximal to distant intensities plotted. A significant decrease in Dsg1, but not Ecad, intensity proximal to the melanoma lesions was observed. (Note that the melanoma sample in the bottom panel was also used in Fig. 5 A where E-cadherin fluorescence intensity was correlated with the extent of melanoma cell spread.) Mean ± SEM depicted. **P < 0.01. Student’s t test.

Figure S1.

Dsg1 expression in sun-exposed and non-sun-exposed skin; transcriptomic analysis reveals increased activation of keratinocyte Slug and reduced keratinocyte differentiation signaling by melanoma cells. (A) GTEx (Genotype-Tissue Expression) data showing Dsg1 mRNA levels were not altered in chronically sun-exposed skin. FPKM, bars represent minimum and maximum values. (B and C) Gene set enrichment analysis (GSEA) of RNA Seq run 2 comparing differentially expressed gene sets to published signatures of primary human keratinocyte differentiation expression patterns show a decrease in the enrichment of granular (GRN) and spinous II (SPNII) layer genes. (D–G) mRNA sequencing was performed on primary human keratinocytes (KC) treated with conditioned media from melanocytes or melanoma cells. Volcano plots depict significantly changed genes in keratinocytes treated with melanoma-conditioned media compared with those treated with melanocyte-conditioned media. Highlighted genes are those involved in pathways of differentiation and the identified upstream regulators for keratinocytes treated with 1341D media (run 1, D; run 2, E) or 501mel conditioned media (run 1, F; run 2, G). (H) Heatmap showing upstream regulators in keratinocytes predicted to be activated by melanoma conditioned media (run 2) using Ingenuity Pathway Analysis (IPA) including RAC1, Snai2, and MAPK family signaling members.

Figure 2.

Melanoma cells downregulate keratinocyte Dsg1 through paracrine signaling. Primary human keratinocytes were treated for 24 h with conditioned media from melanocytes (MC) or the melanoma cell lines WM1341D (labeled as 1341D in figures), WM1366 (labeled as 1366 in figures), WM3211 (labeled as 3211 in figures), YUMAC, and 501mel, and RNA or protein was collected. (A and B) RT-PCR (A) and Western blots (B) were performed for Dsg1, Ecad, and Dsg3. Significantly lower mRNA and protein expression of Dsg1, but not Ecad or Dsg3, was observed. n = 3. Mean ± SEM depicted. *P < 0.05; **P < 0.01; ***P < 0.001. One-way ANOVA. (C) Melanocytes or WM1341D melanoma cells were seeded with normal primary keratinocytes and grown as 3D organotypic raft cocultures for up to 10 d. Paraffin sections were costained for Dsg1 or Ecad and Mel-A. The dashed line indicates the basement membrane. (D) Pixel intensities were determined and the ratio of proximal to distant intensities plotted. A significant decrease in Dsg1, but not Ecad, intensity proximal to the WM1341D lesions was observed. Mean ± SEM depicted. n = 6. *P < 0.05, Student’s t test. Source data are available for this figure: SourceData F2.

Figure 3.

Loss of Grhl1 is associated with melanoma-mediated Dsg1 loss. (A) Model depicting the proposed signaling pathway leading to Dsg1 loss in keratinocytes. (B and C) Primary human keratinocytes were treated for 24 h with conditioned media from melanocytes (MC) or the melanoma cell lines WM1341D and 501mel and RNA or protein was collected. RT-PCR and Western blots were performed for Grhl1. Significantly lower mRNA and protein expression of Grhl1 was observed in the keratinocytes treated with melanoma-conditioned media when compared with melanocyte control–conditioned media. RNA levels and quantification for blots represent average fold change. Mean ± SEM depicted. n = 3. *P < 0.05, **P < 0.01, ***P < 0.001. One-way ANOVA. (D) Melanocytes or WM1341D melanoma cells were seeded with normal primary keratinocytes and grown as 3D organotypic raft cocultures for 6 d. Paraffin sections were costained for Grhl1 and Mel-A. Pixel intensities were determined and the ratio of proximal to distant intensities plotted. A significant decrease in Grhl1 intensity proximal to the WM1341D lesions was observed. Mean ± SEM *P < 0.05. Student’s t test. (E) Paraffin-embedded sections of benign nevi and melanomas were costained for Grhl1 and Mel-A and nuclear and cytoplasmic pixel intensities were measured in cells proximal to Mel-A stained cells and plotted as nuclear/cytoplasmic ratios. A significant decrease in keratinocyte Grhl1 intensity was observed in keratinocytes adjacent to melanoma compared with those in benign nevi. Mean ± SEM depicted. n > 3. **P < 0.01. Student’s t test. (F) Melanocytes or WM1341D melanoma cells were seeded with primary keratinocytes and grown as 3D organotypic raft cocultures for 6–10 d. Paraffin sections were costained for Slug and Mel-A. Pixel intensities were determined, and the ratio of proximal to distant intensities was plotted. A significant increase in Slug intensity proximal to the WM1341D lesions was observed. n = 3 ±SEM; ***P < 0.001. Student’s t test. (G) Paraffin-embedded sections of benign nevi and melanomas were costained for Slug and Mel-A. Nuclear and cytoplasmic pixel intensities were measured in cells proximal to Mel-A stained cells and the nuclear/cytoplasmic ratios were plotted. A significant increase in keratinocyte Slug intensity was observed in keratinocytes adjacent to melanoma compared with those in benign nevi. Mean ± SEM depicted. n > 3. *P < 0.05. Student’s t test. (H) Primary human keratinocytes exogenously expressing GFP or Grhl1 were treated for 24 h with conditioned media from melanocytes (MC) or the melanoma cell lines WM1341D, and 501mel and RNA was collected. RT-PCR was performed for Dsg1. Introduction of Grhl1 into the keratinocytes rescued Dsg1 loss caused by melanoma conditioned media. n = 3. Mean ± SEM depicted. *P < 0.05; **P < 0.01; ***P < 0.001. One-way ANOVA. (I and J) Similarly, primary human keratinocytes exogenously expressing siCTL or siSNAI2 were treated for 24 h with conditioned media from melanocytes (MC) or the melanoma cell lines WM1341D, and 501mel and RNA was collected. RT-PCR was performed for Dsg1 and Grhl1. siSNAI2 transfected keratinocytes rescue Dsg1 and Grhl1 loss caused by melanoma-conditioned media. n = 3. Mean ± SEM depicted. *P < 0.05; **P < 0.01; ***P < 0.001. One-way ANOVA. Source data are available for this figure: SourceData F3.

Figure S2.

Grhl1 is regulated upstream of Dsg1. (A and B) Retroviral transduction of shNT (non-targeting control) or shDsg1 knockdown vectors was performed in primary human keratinocytes and RNA or protein was collected for RT-PCR and Western blot validation of knockdown. No significant difference in Grhl1 mRNA expression or protein level was observed in response to Dsg1 knockdown. (C) Retroviral transduction of GFP or GRHL1 (Grhl1 OE) was performed in primary human keratinocytes and protein was collected for Western blot validation of expression (see Fig. 3 H). (D) 3D organotypic raft cultures comprises shNT or shDsg1 expressing keratinocytes were grown for 6 d and stained for Grhl1. Nuclear to cytoplasmic staining intensity of Grhl1 was not statistically different between the two cultures. Mean ± SEM depicted. n = 3, Student’s t test. (E) Similarly, 3D organotypic raft cultures composed of shNT or shDsg1 expressing keratinocytes were grown for 6 d and stained for Slug and the number of suprabasal Slug expressing cells was counted. The number of suprabasal Slug expressing cells was not different between the two cultures. n = 3. Student’s t test. Source data are available for this figure: SourceData FS2.

Figure S3.

Slug loss in keratinocytes results in decreased Dsg1. (A) Microarray data shows an increase in Dsg1 expression in primary human keratinocytes upon loss of Slug expression, as well as a concurrent decrease in CXCL1 expression. (B) siRNA knockdown of Slug causes an increase in Dsg1 protein levels as seen by Western blot. n = 3. *P < 0.05, **P < 0.01. Student’s t test. (C) Slug knockdown levels assessed by Western blot, corresponding to experiment in Fig. 3, I and J. Source data are available for this figure: SourceData FS3.

Figure 4.

Loss of keratinocyte Dsg1 drives melanoma cell migration. Melanoma cells were cultured for 24 h in conditioned media from keratinocytes transduced with shNT or shDsg1 knockdown vectors and collected for RNA-sequencing. (A and B) GSEA comparing differentially expressed gene sets in RNA Seq run 2 to published signatures of melanoma signaling and differentiation. Normalized enrichment score (NES). (B) Top summary terms from Metascape for overrepresented in genes upregulated by shDsg1 keratinocyte conditioned media. (C) Schematic of trans-well migration assay. Melanoma cells were cultured in conditioned media from shNT expressing keratinocytes or shDsg1 expressing keratinocytes for 24 h. 50,000 cells were then seeded in serum-free media in the upper chamber of a trans-well insert (8- μM pore size). Lower wells contained DMEM supplemented with 5% FBS. (D and E) After 24 h, migrated cells were fixed and stained for visualization. Bars represent normalized migration compared with shNT control conditioned media. Mean ± SEM depicted. n = 3; *P <0.05 Student’s t test.

Figure S4.

Modulation of melanoma migration is correlated with Dsg1 levels in keratinocytes. Conditioned media from primary keratinocytes derived from four separate donors was used to treat melanoma cells in migration assays. Dsg1 knockdown and rescue were not equal in each clone. Shown is a graph of the average migration increase in melanoma cells upon treatment with shDsg1 conditioned media or shDsg1 + FL conditioned media compared with control, correlated with the keratinocyte Dsg1 protein levels in the clones used for each in the shDsg1 cells (purple) and the shDsg1 + FL cells (blue).

Figure 5.

Dsg1 and its regulators are associated with intraepidermal spread of melanoma cells in vivo. (A) Paraffin-embedded sections of melanoma lesions were stained for Dsg1, Grhl1, Slug, or Ecad and Mel-A. (B and C) The number of individual melanoma cells spreading away from the basal layer, (B) as well as the total number of melanoma cells in each section (C) were counted and then correlated to the staining intensity of Dsg1, Grhl1, Slug, and Ecad. (Note that the E-cadherin/Mel-A stained tumor sample in the bottom panel was also used in Fig. 1 A for analysis of E-cadherin fluorescence intensity in proximal versus distal keratinocytes.) n > 10. Simple linear regression.

Figure 6.

Melanoma-induced loss of keratinocyte Dsg1 drives ERK1/2 dependent pro-migratory CXCL1 production. (A) Primary human keratinocytes were treated for 48 h with conditioned media from melanocytes (MC) or the melanoma cell lines WM1341D and 501mel and RNA or protein was collected. RT-PCR was performed for CXCL1 expression. Mean ± SEM depicted. n = 3. *P <0.05. One-way ANOVA. (B) Western blot was performed for Dsg1, ERK1/2, and phosphor-ERK1/2. Mean ± SEM depicted. n = 4, *P <0.05 **P < 0.01. One-way ANOVA. Increased ERK1/2 phosphorylation was observed in keratinocytes treated with melanoma-conditioned media. (C) Primary human keratinocytes were treated with melanocyte or melanoma-conditioned media with either DMSO or 5 μM U0126 (MEK1/2 inhibitor). Western blot was performed to confirm a decrease in ERK1/2 phosphorylation in U0126-treated cells. n = 3. (D) CXCL1 gene expression was no longer increased in keratinocytes treated with melanoma-conditioned media in the presence of U0126 n = 3. *P <0.05. One-way ANOVA. (E and F) Melanoma cells were treated with conditioned media from shNT, shDsg1, or shDsg1 plus a full-length (FL) wild-type Dsg1 rescue expressing keratinocytes for 24 h in the presence of DMSO or the CXCR2 inhibitor (500 nM SB22502) then plated for trans-well migration and collected after 24 h. Loss of keratinocyte Dsg1 no longer increased melanoma cell migration when the CXCL1 receptor, CXCR2, was inhibited. Mean ± SEM depicted. n = 3 *P < 0.05, **P < 0.01. One-way ANOVA. Source data are available for this figure: SourceData F6.

Figure S5.

Gene expression of candidate keratinocyte chemokines and cytokines released in response to melanoma-conditioned media. (A) Schematic depicting receptor-ligand binding partners for candidate chemokine targets. (B–H) Primary human keratinocytes were treated for 48 h with conditioned media from melanocytes (MC) or the melanoma cell lines WM1341D and 501mel and RNA was collected. RT-PCR was performed for chemokines and cytokines induced by shDsg1. Mean ± SEM depicted. n = 3. One-way ANOVA.

Figure 7.

Model: Bi-directional paracrine signaling between melanoma cells and keratinocytes potentiates melanoma cell movement. Our data support a model whereby factors secreted from melanoma cells activate keratinocyte Slug, which in turn decreases Grhl1, a transcriptional activator of Dsg1. The consequent loss of Dsg1 activates ERK1/2 to increase keratinocyte CXCL1 production, which in turn promotes melanoma cell migration through activation of CXCR2.