Author response:
Public Reviews:
Reviewer #1 (Public review):
Summary:
Naina Gour and colleagues provide a detailed observational study in which they demonstrate that MRGPRX4, a human G-protein coupled receptor (GPCR), is expressed exclusively in human melanomas and, when expressed in mouse melanocytes, drives the development of melanomas in mice. These findings provide evidence that MRGPRX4 has the properties of an oncogene, at least in certain cellular environments.
Strengths:
A strength of this work is the nice historical note in which overexpression of MAS1, a GPCR, led to classic studies of transformed fibroblasts in culture and tumors in nude mice. Cloning of MAS1 led to the identification of the MRGPR family of receptors, now known to be key players in neuroimmune and neurosensory phenomena. Here, the story comes full circle with a member of the MRGPR family being linked to a tumor, specifically melanoma. Perhaps the story is not entirely surprising given that the neural crest serves as a precursor for both nerves and melanocytes. But it is nice to see.
Additional strengths include the vast array of tools and techniques employed, from public databases to engineered mice, to establish firmly that MRGPRX4 is expressed in melanomas, although not in every malignant cell.
We thank the reviewer for the positive assessment of our work and for noting the arc back to the original MAS1 studies, which we agree makes for a satisfying full-circle story. We would add one nuance: while the shared neural crest origin of melanocytes and nociceptors offers a plausible explanation for why an MRGPR family member could be co-opted in melanoma, we were nonetheless surprised that this property is specific to MRGPRX4. When we tested the other MRGPRX family members, some of which are also known to be expressed in sensory neurons, using the same genetic strategy (MRGPRX1, MRGPRX2, MRGPRX3; Supplementary Fig. 2B), none induced melanoma, despite arising from the same family of receptors. This suggests the oncogenic property is not simply a generic consequence of neural crest ancestry, but reflects a biology specific to MRGPRX4 itself, which we find is one of the more intriguing open questions this work raises.
Weaknesses:
Given the power of the strengths of the data and story, the following comment is only sort of a weakness, as the topic is addressed while being saved for future studies. Specifically, what leads to the expression of MRGPRX4? The authors posit that it is an epigenetic phenomenon, look briefly at methylation, and rather than going down the proverbial rabbit hole of what comes first, have reasonably decided to punt.
We agree with the reviewer that understanding the precise mechanisms through which MRGPRX4 gets activated during oncogenesis is important, and we had noted this as a limitation of current findings. As we mentioned in the discussion, we hypothesize that one or more environmental modulators - UV exposure, inflammatory cues, or an aging skin microenvironment could initiate the epigenetic reprogramming underlying this expression. This will require dedicated studies and is suited for future work.
Another concern is that given what comes across as the initial observation of MRGPRX4 being expressed in melanoma, what do all of the additional studies add?
The expression data in human melanoma are correlative and serve only as the entry point. The subsequent studies establish MRGPRX4 as a causal, druggable driver of melanoma: it is sufficient to induce 100% penetrant metastatic melanoma in vivo, required for melanoma proliferation and invasion, signals through ligand-independent basal PI3K-AKT/MAPK activity, and is pharmacologically targetable.
For the non-cognoscenti, and to make the manuscript more accessible, the abbreviation NC/EMT, which is also inverted to EMT/NC, should be spelled out periodically as neural crest/epithelial-mesenchymal transition.
This is corrected in the revised version.
Please explain how this study came about. Was it a result of someone deciding to look at expression in the GTEx project and compare it to a tumor database?
This was a serendipitous discovery. As MRGPRs can modulate immune cell function, we were driving expression of various human MRGPRX genes in immune lineages in mice. We noticed that mice overexpressing MRGPRX4 developed spontaneous melanoma-like growths on the ear and tail. Upon investigation, we found that while the Cre driver we had used is appropriate for immune cells, it is also expressed at low levels in melanocytes. We hypothesized that our initial "immune-knock-in" strategy likely resulted in low-level knock-in in mouse melanocytes, thereby allowing MRGPRX4 to directly drive melanocytic transformation. This led us to explore MRGPRX4 expression in human malignancies, where we found MRGPRX4 was highly expressed in melanoma. Next, to directly test our hypothesis that MRGPRX4 transforms melanocytes, we overexpressed MRGPRX4 using Tyrosinase-CreER, the standard Cre driver for melanocytes. This led to a fully penetrant spontaneous melanoma phenotype (Fig. 2).
A comment could be made to explain that while NSG and normal mice were used, the former are immunocompromised, and drawing conclusions without specifying these differences is a weakness.
This is a fair point, and we agree the distinction deserves clarification. The two mouse systems in this study serve different purposes and are not interchangeable. The transgenic TyrCreER+; MRGPRX4LSL+/- model, in which melanoma arises spontaneously from endogenous mouse melanocytes, was studied in immunocompetent mice, allowing us to fully characterize the tumour, including the tumour immune microenvironment (Fig. 2L–O), in an intact immune setting. In contrast, the human A2058 melanoma cell studies (proliferation and metastatic seeding; Fig. 4B–J) required an immunodeficient host, since human cells would otherwise be rejected by a competent murine immune system. For these experiments, we used NSG mice, the standard immunodeficient strain used in human xenograft studies. We have updated the text to reflect this.
In Figure 1A, the p-value of -145 begs for a little explanation. I don't recall seeing such a p-value.
The small p-value reflects two features of the comparison: (1) both GTEx normal skin and TCGA-SKCM contain large sample sizes (hundreds of samples per group), which gives the Mann-Whitney U test (also known as the Wilcoxon rank-sum test) statistical power, and (2) MRGPRX4 expression shows minimal overlap between the two groups; it's essentially undetectable in normal skin but broadly expressed across melanoma samples. Together, these yield such a p-value. This is expected for rank-based tests applied to large, cleanly separated datasets.
Have you considered treating the murine melanomas with murine via PD-L1? I appreciate that this comment is somewhat superfluous given the inhibition of MRGPRX4 with compound 31-2, but given the human therapeutics combined with the fact that you have done 'everything else', I wonder what might happen.
We agree with the reviewer that testing checkpoint blockade in the MRGPRX4-driven model is a relevant direction. Our finding that MRGPRX4-driven tumours develop a PD-L1hi, neutrophil-rich microenvironment (Fig. 2M–O) makes anti-PD-L1 treatment a natural next experiment, and we would predict it to be informative both on its own and in combination with an MRGPRX4 inhibitor, given that the two target distinct compartments- the tumor-immune microenvironment versus tumour-intrinsic proliferative/invasive signaling. We consider it an important direction for future work.
Given the basal ligand-independent signaling, might engineering variants of MRGPRX4 that do not signal be of value?
This is a great point. A variant that disrupts basal (ligand-independent) signaling specifically would be informative. Identifying and validating such a basal-activity-disrupting variant, including confirming normal receptor expression and trafficking, is an important direction for future work.
Reviewer #2 (Public review):
Summary:
This study presents a fundamental new finding - the identification of a sensory-neuron itch receptor, MRGPRX4, as an unexpected melanoma oncogene through a mechanism of lineage-inappropriate expression rather than mutation. The evidence supporting the core observation (tumor-specific upregulation, restriction to invasive transcriptional states, and sufficiency to drive fully penetrant metastatic melanoma in vivo) is compelling, drawing on convergent human genomic datasets and a well-controlled genetic mouse model. However, several of the mechanistic and translational claims - particularly regarding causal drivers of invasion, the immunosuppressive tumor microenvironment, and in vivo pharmacological efficacy - remain incomplete, relying on correlative evidence.
Strengths
The authors propose that MRGPRX4, normally restricted to a subset of peripheral sensory neurons, is aberrantly re-expressed in melanoma rather than through mutational mechanisms, and that this re-expression is sufficient to drive tumorigenesis through basal, ligand-independent GPCR signaling. This is a genuinely novel model for oncogenesis, and the manuscript deploys an impressive range of approaches - bulk and single-cell transcriptomics, spatial transcriptomics, proteomics, phosphoproteomics, and pharmacology - to support it.
Strengths:
The claim that MRGPRX4 is selectively upregulated in melanoma and confined to neural-crest-like/invasive transcriptional states is well supported, with consistent results across multiple independent human scRNA-seq datasets. The claim that ectopic MRGPRX4 is sufficient to drive melanoma is convincingly demonstrated by the fully penetrant, metastatic phenotype in the Tyr-CreER;MRGPRX4-LSL model, with appropriate specificity controls showing that MRGPRX1, MRGPRX2, and MRGPRX3 do not phenocopy this effect.
The claim that MRGPRX4 signals through basal, ligand-independent activity is reasonably well supported by bile-acid quantification showing endogenous ligand concentrations well below the EC50 required for activation.
We are thankful to the reviewer for the positive feedback.
Weaknesses:
The claim that MRGPRX4 remodels the tumor microenvironment toward an immunosuppressive state rests on flow cytometric frequency data (altered neutrophil/eosinophil ratios, increased PD-L1+ myeloid populations) but lacks any functional immune assay to demonstrate that this remodeling actually impairs anti-tumor immune responses.
We acknowledge the reviewer’s suggestion that functional immune assays will demonstrate that MRGPRX4-driven tumor remodeling impairs checkpoint-driven anti-tumour immune response. We consider this an important direction for future work.
The claim that the two MRGPRX4-enriched tumor subpopulations (ECM-rich and NC-like/invasive) underlie the observed invasive and metastatic phenotype is not directly tested; the authors appropriately acknowledge this as an open question, but it is worth noting explicitly that this leaves the mechanistic link between the identified cell states and the functional phenotype (proliferation, invasion, metastasis shown in Figure 4) unresolved.
We agree with the reviewer. To test the direct role of these subpopulations in driving metastasis in our model, we need to ablate them specifically. This is possible with an intersectional genetics strategy, wherein we could use a state-specific Dre or Flp driver (e.g., a Prrx1-DreER or Prrx1-FlpO knock-in) crossed to a dual-recombinase-dependent effector allele (e.g., a Frt- or Rox-gated diphtheria toxin receptor), and then crossed to our TyrCreER-MRGPRX4-LSL model. This quadruple-transgenic strategy would allow ablation restricted specifically to MRGPRX4-driven tumour cells occupying the target mesenchymal state, for example. These genetic lines can be created but require generating or sourcing new dual-recombinase-dependent alleles and a substantially longer breeding and validation timeline (approximately 18-24 months). Together, these represent possible, if long-term, strategies for directly testing the causal contribution of these MRGPRX4-enriched subpopulations to melanoma invasion and metastasis.
Finally, the comparison with BRAF- and NRAS-driven GEMMs (Figure 3K-L) establishes overlap in transcriptional cell states but does not report whether these canonical models themselves upregulate endogenous Mrgprx4. This omission leaves unclear whether MRGPRX4 acts as a convergent node downstream of canonical oncogenic signaling, or represents an independent, parallel route to a similar phenotypic endpoint - a distinction that matters considerably for how broadly the finding should be interpreted.
MRGPRX4 is a primate-specific receptor with no mouse ortholog in melanocytes; thus, we cannot assess endogenous expression of MRGPRX4 in BRAF/NRAS-driven GEMMs. Nevertheless, the following observations argue against MRGPRX4 functioning solely downstream of canonical oncogenes. First, TyrCreER<sup>+</sup>; MRGPRX4LSL mice develop fully penetrant melanoma without engineered BRAF or NRAS activation or tumour-suppressor loss. Second, MRGPRX4 loss in BRAF V600E mutant A2058 cells reduces pERK1/2, pAKT, and pS6K, showing that MRGPRX4 sustains these signaling outputs even in the presence of activated BRAF. Together, these findings support a model in which MRGPRX4 could provide a distinct oncogenic input.
Overall assessment:
The manuscript's central, most novel claim - that lineage-inappropriate expression of a sensory GPCR is sufficient to drive melanoma - is compellingly supported. The secondary mechanistic and translational claims built around this finding are convincing and consistent with the broader literature but are currently supported by correlative rather than causal or functional evidence.