miR-378a-5p expression correlates with melanoma progression, drug resistance and response to therapy
We firstly evaluated the expression levels of miR-378a-5p and miR-378a-3p in a large melanoma patient cohort from The Cancer Genome Atlas (TCGA) database. As shown in Fig. 1a, b, both miRNAs were significantly upregulated in metastatic melanoma when compared to primary melanoma. Motivated by the lack of any data on the role of miR-378a-5p in melanoma pathobiology, we focused our attention on the biological functions of this miRNA in human melanoma. As reported in Fig. 1c, the expression of miR-378a-5p was significantly higher in melanoma specimens from our Institute compared to melanoma in situ. We also found an increased miR-378a-5p expression in the A375 melanoma cells resistant to Vemurafenib (BRAF inhibitor) or to Dabrafenib/Trametinib (BRAF/MEK inhibitors) compared to sensitive ones (Fig. 1d). Interestingly, while exposure to Dabrafenib did not affect miR-378a-5p expression (data not shown), the treatment with Trametinib significantly reduced its expression in both M14 and A375 cells (Fig. 1e).
Next, we evaluated the effect of miR-378a-5p on the expression of some target genes (Supplementary Table 1) including putative, such as STAMBP and SP1, or validated, such as FUS-1/TUSC2, SUFU and KLF9. To this purpose we modulated miR-378a-5p expression transfecting melanoma cells with hsa-miR-378a-5p miRNA mimic or antisense sequence against miR-378 or the relative scrambled controls. Upregulation of miR-378a-5p in M14 melanoma cells reduced both mRNA (Supplementary Fig. 1a) and protein (Supplementary Fig. 1b) levels of STAMBP, SP1, KLF9, FUS-1, and SUFU when compared with control transfected cells, while miR-378a-5p inhibition led to an opposite effect, upregulating the expression of target genes (Supplementary Fig. 1c). As reported in Supplementary Fig. 1d, luciferase assay confirmed the binding of miR-378a-5p to the 3'UTR region of STAMBP mRNA.
We next assessed the functional relevance of miR-378a-5p on in vitro cell proliferation, migration, invasion, and clonogenic ability. Contrary to the results demonstrating miR-378a-5p ability to affect proliferation of several tumor histotypes, we did not observe any effect of miR-378a-5p either on proliferation or clonogenic ability of M14 melanoma cells (Supplementary Fig. 2).
As observed in other tumor histotypes, miR-378a-5p overexpression induced a significant increase of both migratory and invasive capacity of M14, A375, and SBCL1 melanoma cells (Fig. 2a, Supplementary Fig. 3), as well as an increased expression of Metalloprotease-2 (MMP2) (Fig. 2b), which is a key metalloprotease involved in melanoma progression.
We further analyzed the involvement of miR-378a-5p in vasculogenic mimicry (VM), an alternative way to provide tumor blood perfusion. As reported in Fig. 2c and Supplementary Fig. 4, the formation of channel-like structures, evaluated in terms of number of intersections, was augmented in M14 and SBCL1 melanoma cells overexpressing miR-378a-5p, compared with control cells.
We next investigated on the possible factors involved in miR-378a-5p-induced VM. VM depends on the expression of several factors including VEGF and interleukin-8 (IL-8) and both these factors are modulated by miR-378a-5p in lung carcinoma cells. Thus, we evaluated whether VEGF and IL-8 are regulated by miR-378a-5p in melanoma models and their impact on miR-378a-5p-induced VM. In agreement with previously reported data, miR-378a-5p induced a significant increase of VEGF secretion in M14 and A375 cells (Fig. 2d). Of note, neutralizing antibodies directed versus VEGF were able to strongly reduce miR-378a-5p-induced VM in M14 and SBCL1 cells (Fig. 2e, Supplementary Fig. 4a, b). On the contrary, IL-8 was found to be equally secreted in control and miR-378a-5p overexpressing cells (data not shown), and IL-8-neutralizing antibodies did not affect miR-378a-5p-induced VM (Fig. 2e, Supplementary Fig. 4).
We also evaluated whether miR-378a-5p was able to affect the expression of uPAR a very critical regulator of migration, invasion and VM. The modulation of miR-378a-5p in melanoma cells shows a significant regulation of uPAR expression at transcriptional (Fig. 2f) and protein (Fig. 2g) level.
As miR-378a-5p negatively regulates the expression of SP1 (Supplementary Fig. 1), a transcription factor demonstrated to positively regulate the expression of uPAR, we searched for other possible transcription factors involved in miR-378a-5p-induced uPAR expression. To this purpose we performed qRT-PCR and western blot analyses to evaluate whether miR-378a-5p affects the expression of HOXD10, a transcription factor known for its ability to repress the expression uPAR and some MMPs in cancer. As reported in Fig. 3a, b, the modulation of miR-378a-5p affected the expression of HOXD10 both at the transcriptional and protein level, thus indicating that miR-378a-5p may target HOXD10. This hypothesis is supported by bioinformatics analysis predicting HOXD10 as a putative miR-378a-5p target gene (Supplementary Table 1) and confirmed by luciferase assay, showing the binding of miR-378a-5p to the wild-type 3'UTR region of HOXD10 mRNA. As shown in Fig. 3c, by co-transfecting the miR-378a-5p and the plasmid containing the 3'UTR region of HOXD10 mRNA, the signal of luciferase activity was significantly decreased due to the binding of the miRNA to its target sequence. The mutation of the binding sequence in the 3'UTR region of HOXD10 mRNA restored the basal luciferase activity (Fig. 3c). Moreover, the expression profiling analysis of HOXD10 in the same melanoma patient cohort from the TCGA database used to evaluate the expression levels of miR-378a-5p reported in Fig. 1a, showed negative correlation between HOXD10 and miR-378a-5p levels (R = -0.24, p-value = 2.7 × 10) and a lower level of HOXD10 transcript in metastatic melanoma samples compared to primary ones, although with a non-significant p-value (0.06) (data not shown). Interestingly, contingency table (Fig. 3d) demonstrated about 32% of high miR-378a-5p/low HOXD10 in metastatic samples while only about 17% was evidenced in primary samples (p-value = 0.0001; Pearson's R = -0.72).
Our data together with the ability of uPA/uPAR axis to function as a degrader of extracellular matrix and a regulator of migration, invasion and VM, are indicative of a possible involvement of uPA/uPAR axis in miR-378a-5p-induced in vitro tumor-promoting functions. To investigate the relevance of uPAR in the ability of miR-378a-5p to affect in vitro properties associated with melanoma aggressiveness, a specific small interference RNA smart pools (si-uPAR) able to reduce uPAR expression (Fig. 4a) was used after miR-378a-5p mimic transfection. As reported in Fig. 4b-d and Supplementary Fig. 5, 6, uPAR silencing strongly reduced miR-378a-5p ability to increase in vitro cell migration, invasion and VM, when compared to the relative control.
Through its interaction with integrins, we previously reported that uPAR is able to affect melanoma invasion, migration and response to therapy. We also demonstrated that the M25 linear peptide was able to uncouple uPAR from integrins thus affecting its functions. On the basis of these evidences, we reasoned that inhibition of uPAR functions with M25 peptide could produce functional effects similar to those obtained with uPAR silencing by means of si-uPAR. As reported in Fig. 4b, c and Supplementary Fig. 5, M25 peptide strongly reduced miR-378a-5p-induced migration and invasion, when compared to cells treated with the relative control peptide.
As reported in Fig. 2d, the secretion of the proangiogenic factor, VEGF, was significantly increased in melanoma cells after miR-378a-5p overexpression. Moreover, miR-378a-5p is exported from lung cancer cells in exosomes and the secretion of this miRNA has been reported to correlate with miR-378a-5p expression by the cells. Thus, we investigated whether miR-378a-5p was able to affect the in vitro and in vivo angiogenesis. As shown in Fig. 5a, b, human umbilical endothelial cells (HUVEC) seeded on Cultrex BME and exposed to conditioned medium (CM) derived from M14 cells overexpressing mimic miR-378a-5p, formed a significant increased number of tubular-like structures when compared to cells exposed to CM from control cells. In agreement with the in vitro results, matrigel plugs containing CM from mimic miR-378a-5p overexpressing M14 cells injected in C57Bl6 mice, showed a higher co-option of surrounding vessels and about 4-fold induction of the haemoglobin content when compared to the matrigel plugs containing CM from control cells (Fig. 5c, d).
We previously reported Bcl-2 ability to regulate the expression of miR-211 and miR-204, two miRNA involved in melanoma progression and resistance. To evaluate whether the expression of miR-378a-5p was modulated by Bcl-2, we used either Bcl-2 overexpressing clones previously obtained and characterized, or melanoma cells in which Bcl-2 expression was silenced by siRNA. Increased miR-378a-5p expression was found in M14 and A375 cells overexpressing Bcl-2 (Fig. 6a), while a reduced expression of miR-378a-5p was observed after Bcl-2 silencing both in M14 (Fig. 6b, c) and A375 (Fig. 6d, e) cells. Bcl-2 downregulation with siRNA also decreased pri-miR-378a-5p (pri-miR-378) expression (Fig. 6f). Interestingly, treatment with Venetoclax, a specific Bcl-2 inhibitor, significantly reduced miR-378a-5p expression in melanoma cells (Fig. 6g).