A novel germline NF1 splicing variant drives the onset of an anorectal mucosal melanoma in a patient with a stable and durable nivolumab response

Objective. Neurofibromatosis type-1 (NF1) patients rarely develop mucosal melanomas. We report a rare form of anorectal mucosal melanoma (ARMM) in an NF1 syndromic patient profiled for genomics and transcriptomics to assess the determinants of the response to nivolumab.

Methods. Primary melanoma and metastases were analyzed with targeted sequencing and gene expression profile (tGEP). We applied in silico (cBioPortal and predictor tools) and in vitro (hybrid minigene) approaches to confirm the variant pathogenicity.

Results. We detected the novel c.4269+2_4269+3delTG germline splicing variant in NF1, which proved to be pathogenic by the minigene assay showing an aberrant splicing. The tumor showed a copy-number (CN) neutral loss of heterozygosity for the WT allele, and both ARMM and metastases carried several CN gains associated with NF1-driven carcinogenesis and very low mutation burden. The tGEP analysis unveiled a macrophagic infiltration, with a pro-inflammatory M1-type polarization, in the context of lack of PD-L1 expression.

Conclusions. The response to nivolumab in a germline NF1-driven ARMM case seems independent from levels of TMB and PD-L1 expression and may be mediated by inflammatory response induced by M1-polarized macrophages.

Introduction

Somatic NF1 mutations in melanomas, mutually exclusive with BRAF or NRAS mutations, range from 13% to 40% and are related to sun exposure¹. NF1-driven malignancies typically harbor thousands of somatic mutations, triggered by UVs². Acral and mucosal melanomas show a similar spectrum of NF1 mutations, but UV-independent².

Germline NF1 gene variants cause the neurofibromatosis type-1 (NF1) syndrome, clinically marked by “café au lait” spots, lentiginosis, neurofibromas, and increased risk of rare neoplasias³. To date, there is not a clear correlation between NF1 syndrome and melanomas^1,3: NF1-related melanomas are rare and lethal NF1-associated malignancies³. In this scenario, anorectal mucosal melanoma (ARMM) is a rare neoplasm characterized by aggressive behavior and poor prognosis⁴. It is also typically diagnosed at a locally advanced stage⁴.

We performed genomic and transcriptomic analysis in a young NF1 patient with metastatic ARMM with long-standing clinical benefit from 4-years nivolumab therapy, to investigate a possible correlation between NF1 syndrome, melanoma insurgence and response to ICIs.

Materials and methods

The patient signed an informed consent at the Medical Genetics Unit of AOU Città della Salute e della Scienza in Turin for both germline from whole blood (WB), and tumoral molecular analyses from two formalin-fixed paraffin-embedded (FFPE) tissue blocks of the primary ARMM (ARMM#1 and ARMM#2), and from FFPE pericolic lymph nodes (LN_MTS) and adipose tissue (perirectal lipidic metastasis PRL_MTS) metastases

NEXT-GENERATION SEQUENCING

Samples were subjected the TruSightOncology 500 panel (Illumina, San Diego, U.S) and to the shallow whole genome sequencing for copy number profiling ^5,6. Raw data were processed with the DRAGEN server (v4.0, hg19 alignment, Illumina) and with a custom pipeline (https://github.com/ljwharbers/copynumber-calling).

CBIOPORTAL QUERY

We screened 218 manually curated studies with no overlapping samples, including TCGA and non-TCGA studies, for the NF1 mutations available on cBioPortal (https://www.cbioportal.org/, last access date 2023 December 12)⁷.

GENE EXPRESSION PROFILING

We applied the PanCancer IO 360™ Panel (NanoString Technologies, Seattle, WA, US) following the manufacturer’s protocol. mRNA counts were analyzed as previously described⁸.

Gene expression data were assessed to: (i) evaluate the intra-sample up-/down-regulated pathways using single sample GSEA (ss)GSEA, (ii) infer immune cell-type invasion using the nSolver Advanced Analysis algorithm to measure the abundance of various cell populations.

IMMUNOHISTOCHEMISTRY

We stained ARMM#1 FFPE tissue sections to assess biomarkers of immune cells with anti-CD8 (clone NCL-L-CD8-4B11, Leica), anti-CD68 (clone CD68 PG-M1, Agilent), and 22C3 (Dako) monoclonal antibodies to assess PDL1 protein expression on the Dako Autostainer instrument. PDL1 scoring was assessed based on the combined positive score (CPS).

SPLICING VARIANT PREDICTION AND MINIGENE ASSAY

We assessed the pathogenicity of the NF1 c.4269+2_4269+3delTG variant (RefSeq: NM_000267.3, NF1 exons: LRG_214) by interrogating the Human Splicing Finder (HSF, v3.1, http://www.umd.be/HSF/), Mutation taster (https://www.genecascade.org/MutationTaster2021/), and CADD (v1.7) tools (https://cadd.gs.washington.edu/snv), and by applying the hybrid minigene assay, as previously described⁹.

Extended Methods are reported in the Supplementary Material.

Results

A rectal lesion in a 34-year-old woman, with a clinical diagnosis of NF1, was diagnosed as BRAF and NRAS wild-type (WT) anorectal mucosal melanoma (ARMM) (Fig. 1A). Since CT-imaging revealed CNS and lung metastases, interdisciplinary consultation agreed on nivolumab treatment and surgical debulking, which confirmed the mucosal melanoma of the anorectal junction (Fig. 1A, ARMM#1 and ARMM#2). The patient displayed response for lung nodules (complete) and for the CNS lesions (partial) but she locally relapsed (lymph-node metastases, LN_MTS, and perirectal fat tissue deposits, PRL_MTS, surgically removed). Nivolumab therapy is ongoing as of May 2024. According to RECIST criteria v.1.1, the patient exhibited a partial radiological response followed by disease stabilization.

WB DNA sequencing identified a new heterozygous (VAF = 0.44) splicing variant affecting the GT-donor site downstream exon32 of the NF1 gene (NM_000267.3:c.4269+2_4269+3delTG, chr17:29585521-29585522), which has never been reported before (absent in GnomADv4.1.0 and in ClinVar, LOVD, HGMD mutation databases). Since the in silico tools predicted a potential pathogenic effect of this variant over the normal splicing of NF1 gene (Suppl. Fig.1A), we applied a minigene assay⁹, by transfecting a hybrid β-globin minigene containing WT and mutated NF1 exon32 GT-donor site into HEK293 cells (Fig.1B). Triplicated results showed that the mutant construct showed an aberrant splicing pattern, with skipping of exon 32 of NF1, which, although inframe, is crucial to protein function (Fig. 1C and Suppl. Fig. 1B). Based on these findings, we classified the variant as “likely pathogenic” (class 4) according to the ACMG criteria¹⁰ (PVS1, PM2), with a presumably de novo condition (no familial cause declared).

Next, we assessed the prevalence of somatic NF1 splicing variants in the cBioPortal database (70,655 samples, 67,030 patients). A total of 2,161 NF1 driver variants (1,814 samples, 1,746 patients), and 379 splicing variants were found in 363 patients (Suppl. Tab. I). Only five variants mapped to the same location of the proband alteration, all annotated as “pathogenic” (2) or “likely pathogenic” (3). Considering the cancer type, cBioPortal reported 23 ARMMs out of 67,030 tumors. Of these 23 ARMM cases annotated in cBioPortal, 5 carried a pathogenic NF1 variant and 2 had mutations affecting splice-sites, involving other nucleotide residues (Suppl. Tab. II).

Next, we sequenced the patient’s lesions (ARMM#1/2 and the metastatic deposit) Focusing on the NF1 splicing variant identified in the blood sample, all the neoplastic tissues showed a VAF higher than the expected heterozigosity (ARMM#1: 68%, ARMM#2: 65%, PRL_MTS: 62% and LN_MTS: 70%). By reviewing the H&Es of each lesion, we identified a tumor purity around 65%-70%, in line with a suggestive clonal loss of heterozygosity (LOH) of the WT allele. Paired comparison between WB and ARMM#1 demonstrated a shift from heterozygosity for most of chr17 SNPs (Fig. 2A). Moreover, the FACETS algorithm identified a copy number neutral LOH at chr17 NF1 locus, with a gain of the distal part of chr17 long arm (Fig. 2B).

Besides confirming the absence of somatic BRAF, NRAS and c-KIT alterations, we identified a low TMB (0.83-1.66 mut/Mb) and a microsatellite stable profile for both ARMM samples and metastases. The MAP3K1 p.P278S variant was shared by all lesions, whereas ARMM#1 and LN_MTS harbored the PIK3C2G p.Q22H variant. Tumors were characterized by a higher CN score compared to the CN-quiet profile of WB DNA (Shallow-WGS, Suppl.Fig.2A). Neoplastic samples showed superimposable CN profiles, distinct from the normal control (Suppl. Fig. 2B). At the gene level, we identified conserved CN gains (LRR > 1.5) for several cancer related genes, including TERT, MDM2, GLI1, CDK4, LGR5, and PPM1D (Fig. 2C).

We applied the (ss)GSEA algorithm to rank enriched pathways (Human MSigDB hallmark (H) genesets, Suppl.Table 3). We observed an overexpression of IFN-γ and IFN-α signaling in the ARMM sample, and all the lesions showed TNFα and JAK-STAT3 pathway enhancement (Fig.3A). By decoding the immune cell-type from bulk GEP, both primary and metastatic lesions were found to be enriched for macrophages and CD8 T-cells, whereas Dendritic Cells (DC) were present only in the primary ARMM (Fig. 3B). H&E section review and IHC staining corroborated the infiltration by macrophages (CD68⁺) and T-cells (CD8⁺), with a low PDL1 expression (Fig.3C). Finally, the heatmap reporting literature-based biomarkers¹¹ showed an overexpression of the proinflammatory and M1 status macrophage markers, together with a low count of CD163 and MRC1 mRNA, typically related to anti-inflammatory M2 status (Fig. 3D).

Discussion

We report a novel NF1 splice variant in a NF1 syndromic patient, confirmed as pathogenic, driving the onset of a mucosal melanoma through NF1^WT allele LOH. The patient’s response to ICIs was potentially enhanced by a macrophagic-induced proinflammatory microenvironment.

The impact of the NF1 variant over splicing mechanisms⁹, the CN neutral LOH¹², the lack of mutations in other driver genes² and the co-occurrence of CN gains of TERT, MDM2, and CDK4 genes¹² corroborate the role of NF1 in ARMM development for this patient. Four examples of genotype-to-phenotype correlation between NF1 syndrome and the development of ARMMs have been previously reported^13-16. Ben-Izhak and colleagues reported a de novo, NF1 proband with an anal melanoma¹⁵, whereas Ishii et al. associated NF1 syndrome and ARMM in a patient with somatic NF1 loss¹⁴, similarly to Soon and colleagues who reported a rectal melanoma in a NF1 syndromic patient¹³. However, these manuscripts lack comprehensive molecular analysis, firstly reported by our report in terms of somatic mutational landscape (SNVs and CNVs) and gene expression patterns in both primary and metastatic lesions.

ARMMs are aggressive lesions (10-20% 5-year OS) and without alternative treatments other than chemotherapy⁴. Overall, mucosal melanomas treated with nivolumab alone, or in combination with ipilimumab show an objective response rate of 23.3% and 37.1%, respectively¹⁷, whereas NF1-mutated responders usually carry a high TMB². This contrasts with the low TMB and minimal PDL1 expression detected in our patient. Moreover, Buchbinder and colleagues reported a TMB-independent response to ICIs, potentially influenced by the tumor micro-environment¹⁸. NF1-associated tumors harbor a strong immune infiltration, with a predominance of CD68+ macrophages, confirmed to be recruited by NF1 mutation in SW10 cells¹⁹. We observed an abundant M1-macrophage infiltration and an anti-tumoral microenvironment, with overexpression of IFN pathways and secretion of pro-inflammatory cytokines. Interestingly, Duan and Luo proposed that the M1 phenotype leads to the activation of an inflammatory response, enhancing the anticancer effects in patients treated with ICIs, reinforced by an induction of M1 polarization by both anti-PD-1/PD-L1 immune checkpoint blockade²⁰.

Our study has some limitations, the main being the lack of WES and RNAseq for a larger comprehensive genomic evaluation, unfeasible because of the low quality of the archival, FFPE-purified nucleic acids.

Conclusions

Based on genomic and transcriptomic analyses, we comprehensively elucidated the durable response to nivolumab in a patient affected by ARMM carrying a novel germline NF1 gene splicing variant. DNA sequencing suggested that somatic CN neutral LOH likely drove ARMM development in this patient, with a somatic mutational landscape characterized by low TMB and several gene amplification events, mostly related to NF1 carcinogenesis (MDM2, CDK4 and TERT genes were affected). Based on our gene expression data we propose that in this specific patient NF1 driven carcinogenesis was strictly associated with macrophagic infiltration favored a prolonged response to nivolumab.

CONFLICTS OF INTEREST STATEMENT

CM report personal consultancy fees from Menarini and Roche and speaker fees from Illumina, Veracyte and Daiichi Sankyo outside the scope of the present work. All the other authors have no conflict of interest to disclose.

CONSENT FOR PUBLICATION

The patient provided authorization for publication of images and deidentified information reported here.

AVAILABILITY OF DATA AND MATERIALS

The datasets generated and/or analyzed in the current study are available in the current manuscript and Suppl.material, and available from the corresponding author upon reasonable request. Methods are reported in the Suppl.methods.

FUNDING

This work was supported by Italian Ministry of Health, Ricerca Corrente 2024 and by grants from the Italian Ministry of University and Research (grant numbers 2017BJJ5EE and PNRR-TR1-2023-12377161 to E.T.). EB was the recipient of a PhD fellowship under the funding of Dipartimenti di Eccellenza 2018–2022 (Project No. D15D18000410001).

AUTHORS’ CONTRIBUTIONS

Conceptualization: SR, LM, CM. Data curation: EB, SEB. Formal Analysis: EB, SEB, VM, ET. Funding acquisition: CM, ET. Investigation: EB, SEB, LM, VM, AG, DZ, IS, RS, NC. Methodology: EB, SEB, LM. Resources: CM, PQ, AS, SR. Software: SEB. Supervision: CM,SR. Validation: EB, SEB. Visualization: EB. Writing – original draft: EB, CM,SR. Writing – review & editing: all authors.

ETHICAL CONSIDERATION

The patient signed an informed consent at the Medical Genetics Unit of AOU Città della Salute e della Scienza in Turin.

Supplementary material

Methods

The patient signed an informed consent for both germline and tumoral molecular analyses at the Medical Genetics Unit of AOU Città della Salute e della Scienza in Turin. This consent allowed us to sequence both germline and tumoral DNA for genomic purposes.

TISSUE SAMPLE REVIEW, DNA AND RNA EXTRACTION

To investigate the potential NF-1 gene germline alteration, we first sequenced DNA purified from whole blood (WB), purified following the manufacturer protocol of the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany). We then extracted DNA from two independent Formalin-fixed paraffin-embedded (FFPE) blocks corresponding to the primary ARMM (ARMM#1 and ARMM#2), and to the metastatic deposit in both pericolic lymph nodes (LN_MTS) ln and adipose tissue (APL_MTS). Two independent pathologists evaluated the tumor cellularity for all the blocks on Hematoxylin and eosin (H&E) stained sections, and six-micron thick sections were prepared and meso-dissected to enrich tumor cell content. DNA was purified using the Maxwell® RSC DNA FFPE Kit (Promega, Madison, USA) and RNA was extracted with Maxwell® RSC RNA FFPE Kit (Promega, Madison, USA). Both nucleic acids were quantified with spectrophotometer (Nanodrop 1000, ThermoFisher Scientific, Waltham, MA, US) and fluorometer (Qubit, ThermoFisher Scientific, Waltham, MA, US), and checked for integrity using the Agilent 4150 TapeStation System, with Genomic DNA ScreenTape Assay for TapeStation Systems or Agilent RNA ScreenTape Assay (Agilent Technologies, Santa Clara, US).

NEXT-GENERATION SEQUENCING METHODS

DNA and RNA extracted from the archival FFPE slices showed TapeStatation (Agilent Technologies, Inc., Santa Clara, U.S) DIN and RIN of 1.2 respectively, not allowing Whole Exome and RNAseq sequencing.

Comprehensive genomic profiling

Both germline and somatic DNA were subjected to deep sequencing, using the TruSight Oncology 500 panel (Illumina, San Diego, California, U.S), which comprises 523 gene spanned on a total of 1.2 Mb of coding region (1.94 Mb of total coverage). Libraries were generated following the manufacturer’s protocol, with a total of 80 ng of DNA as input. All libraries pass the quality controls and were sequenced on the Illumina Next-Seq 550 instrument (Illumina, San Diego, California, U.S).

Raw data were transferred on the DRAGEN on-site server (v4.0, Illumina), loaded with the TSO500 specific “tumour-only “pipeline, and aligned to the human reference sequence GRCh37 (hg19). We also endorse the germline variants by comparing sequencing results derived from whole blood DNA and the tumoral lesions. All variants with a VAF >10% were annotated using the Intervar tool (doi:10.1016/j.ajhg.2017.01.004), which predicted the in-silico level of pathogenicity (12 predictors proper of the ANNOVAR tool) and reported the clinical significance by ClinVar (version 20221231 hg19 build). Finally, we also assess the potential actionability of the somatic variants using the python tool oncokb/oncokb-annotator, which reported the OncoKB¹⁵ annotation. Internal TSO500 pipeline also identified the tumor mutation burden (TMB) and the microsatellite instability status (MSI) of 120 mictosatellite loci.

To assess allele-specific copy number and evaluate loss of heterozygosity (LOH) involving NF1 gene, we also applied FACETS algorithm v0.6.2. First, reference and variant allele read counts files were generated from tumor/normal BAM files for dbSNP build 151 (GRCh37p13 genome) using snp-pileup with the parameters “-g -q15 -Q20 -P100 -r25,0” according to manual. Next, copy numbers were estimated using FACETS with the following parameters: snp.nbhd = 150, Cval = 100 and ndepthmax=1000.

CUTseq method for shallow WGS

CUTseq

All samples (WB and neoplastic lesions) were also subjected to the CUTseq assay, a cost-effective, quick, and robust library prep method to assess genome wide copy-number alteration (PMID: 31628304). Briefly, using I-DOT One robot, we dispense 0.35 ul of each sample with an initial concentration of 20 ng/ul, for a total input of about 7 ng. After NlaIII-HF enzyme digestion, we ligated a 33 nM-CUTseq adapter for each sample, using the T4 rapid DNA ligase (Thermo Fisher Scientific). After ligation, we pooled the sample and we fragmented DNA using Covaris ME220 Focused-ultrasonicator with the target peak of 200 bp. Sheared DNA was purified using the Agencourt Ampure XP beads (Beckman Coulter, Cat. No. A63881) and eight of the pooled library were used the IVT step following the protocol described in the CUTseq manuscript. The produced library was sequenced on an Illumina NextSeq 550 platform with high output 75 bp single-end kit (Illumina). The global .fastq file, containing all the CUTseq adapted samples, was demultiplexed using bcl2fastq2 software (version 2.20). This .fastq was pre-processed using a previously developed custom pipeline (https://github.com/ljwharbers/copynumber-calling). Briefly, we scanned for sequencing reads retaining the CUTseq adapter (8 nt of UMI + 8 nt of sample barcode) using SAMtools ((v1.9)26), producing sample specific .fastq files. After trimming, reads were aligned to the GRCh37/hg19 reference genome using BWA-MEM (v0.7.17 release) with default options. Using the python umi_tools (v1.1.4) 29 package (default parameters), we removed PCR duplicates. After pre-processing, read counts were calculated for each bin (window size of 100 kb) and then the segmentation was performed using the circular binary segmentation algorithm implemented in DNAcopy R package (v1.74.1) to determine segment copy number levels (expressed as log2 ratio of the read counts). We also inferred gene copy number overlapping detected segments based on genomic coordinates using GenomicRanges R package (v1.52.0).

CBIOPORTAL QUERY

We screened 218 studies that are manually curated, including TCGA and non-TCGA studies, with no overlapping samples for the NF1 of gene mutations available on cBioPortal (last access date 2023 December 12).

GENE EXPRESSION PROFILING

To evaluate gene expression data, we applied the PanCancer IO 360™ Panel (NanoString Technologies, Seattle, WA, US) on the neoplastic lesions. Two hundred and fifty ng of total RNA were captured by probes encompassing a total of 770 genes, as previously described¹. The files containing mRNA counts were analyzed with the nSolver Software (NanoString Technologies, Seattle, WA, US) and with the nSolver Advanced Analyses, based on robust R statistics.

Gene expression data were assessed to evaluate the intra-sample up-down regulated pathways. To do this, we applied single sampleGSEA (ssGSEA), a variant of the GSEA algorithm which provided a score for each gene set in a single sample. The nicolerg/ssGSEA2 R package computed the NES score and the relative FDR-adjusted p-value for each geneset reported in the Human MSigDB hallmark (H) set. Since we applied a targeted GEP approach, we considered only those genesets with at least the 25% of covered genes, with a p-adjusted <0.01 as significance cut-off. GEP data were also exploited to infer immune cell-type invasion using the nSolver Advanced Analysis algorithm of immune cell count, based on the Danaher et al. (2017) to measure the abundance of various cell populations.

IMMUNOHISTOCHEMISTRY

We stained ARMM#1 FFPE tissue sections to assess biomarker immune cells expression with the monoclonal antibodies anti-CD8 (clone NCL-L-CD8-4B11, Leica), anti-CD68 (clone CD68 PG-M1, Agilent), and with the 22C3 (Dako) clone to assessed PDL1 protein expression on the Dako Autostainer instrument. PDL1 scoring was assessed based on the combined positive score (CPS).

SPLICING VARIANT PREDICTION AND MINIGENE ASSAY

For mutation nomenclature, NM_000267.3 was used as reference sequence and NF1 exons were numbered as outlined in LRG_214.

The following online software for splicing prediction for the NF1 c.4269+2_4269+3delTG variant were employed: Human Splicing Finder (HSF) Version 3.1 (http://www.umd.be/HSF/), Mutation taster2021 (https://www.genecascade.org/MutationTaster2021/), and Combined Annotation Dependent Depletion (CADD) v1.7 (https://cadd.gs.washington.edu/snv).

Variant interpretation was based on ACMG criteria. Allele frequency was checked on GnomAD database v4.0 (https://gnomad.broadinstitute.org/)

A PCR fragment including NF1 exon 32 and at least 350 bp of the upstream and downstream introns was amplified from wild-type genomic DNA and cloned into the β-globin vector, as previously described². A schematic representation of the minigene construct is depicted in Figure 1B. The In-Fusion Cloning kit (Takara Bio Inc, Shiga, Japan) was employed to generate the vector harbouring the c.4269+2_4269+3delTG variant. The correctness of both wild-type and mutant clones was verified by Sanger sequencing.

6 x 10⁵ HEK293 cells were transfected with 1 μg of empty, wild type or mutated minigenes using Lipofectamine 2000 (ThermoFisher Scientific, Watham, MA, USA) following manufacturer’s instructions. Total RNA was extracted after 24 h with TRIzolTM (ThermoFisher Scientific, Watham, MA, USA), retrotranscribed using SuperScript II reverse transcriptase (ThermoFisher Scientific, Watham, MA, USA). Specific primers designed on exon 2 (forward) and exon 3 (reverse) of the β-globin gene were used to amplify the resulting cDNA. PCR products were then separated on a 2% agarose gel and individual bands were excised and sequenced.

History

Received: March 19, 2025

Accepted: March 20, 2025

Figures and tables

Figure 1. A) Clinical timeline of the proband. We report the most representative steps of the proband clinical evolution. CT: computed tomography, PR: Partial Response, CR: Complete Response, MRI: Magnetic Resonance Imaging. B) Schematic representation of the hybrid minigene construct used in the experiments. PCR fragments including exon 32 adjacent to NF1 c.4269+2_4269+3delTG variant and at least 350 bp of the upstream and downstream introns were amplified from wild-type genomic DNA and cloned into the β-globin vector, which was obtained by the cloning of the β-globin gene inside a pcDNA™3.1/Hygro(+) vector (ThermoFischer Scientific). Β: β-globin, Ex.: exon, Int.: intron. C) Sanger sequencing of the RT-PCR products obtained after cell transfection with the minigene constructs. HEK293 cells were transfected with the minigene vector harboring the c.4269+2_4269+3delTG variant (MUT) or with the corresponding wild-type one (WT). After RNA extraction and retrotranscription, the resulting cDNA was amplified using β-globin specific primers. On the left, a schematic representation of RT-PCR products obtained, while on the right the corresponding Sanger electropherograms. Junctions between exons are indicated above each sequence.

Figure 2. A) Chromosome 17 variant allele frequency (VAF) distribution. Paired t-test identified a VAF shift of germline-annotated variants between WB and ARMM#1. B) FACETS output for TSO500 targeted sequencing. Allelic phasing identified a copy number neutral LOH for the NF-1 gene locus. In the copy number (CN) plot, red lines represented the total CN, whereas the black ones the minor allele CN for each genomic segment. C) CUTseq trace for ARMM#1. Shallow whole genome sequencing identified several CN gains across genome, with CNA genes related to a NF-1 driven carcinogenesis annotated in the Fig..

Figure 3. A) Spider plot of (ss)GSEA significant genesets. TNFA and JAK-STAT3 pathway were enhanced in ARMM, whereas IFN-γ and INF-α signaling only in in all the lesions. The y-axis of the splider reported the NES by ssGSEA. B) Heatmap reporting immune cell-type profiles. Primary and metastatic lesions were highly infiltrated for macrophages and CD8 T-cells. C) ARMM#1 H&E section and IHC staining. Clockwise, we report H&E, CD8, PDL1 and CD68 10X pictures. D) Heatmap of macrophages M1 and M2 related genes. Proinflammatory and M1 markers where enriched in all the lesions.