Skip to main content
NCBI home page
NPJ Precision Oncology logoLink to NPJ Precision Oncology
. 2025 Dec 5;10:7. doi: 10.1038/s41698-025-01210-2

CDKN2A homozygous deletions and TSC2 somatic mutations in metastatic pancreatic neuroendocrine tumors

Tito Teles Jesus 1,2, Lorenzo Ferrando 3, Lia Rodrigues 1,2,4, Rui Sousa Martins 1,2,4, Luís Cardoso 1,2,5, José Manuel Lopes 1,2,4, Paula Soares 1,2,4, Arnaud Da Cruz Paula 1,2,#, João Vinagre 1,2,4,✉,#
PMCID: PMC12775397  PMID: 41350422

Abstract

Despite improvements in the molecular profiling of pancreatic neuroendocrine tumors (PanNETs), predicting their clinical behavior and response to specific therapies remains challenging. We sought to elucidate the molecular basis underlying the broad phenotypic variations in these neoplasms through a genetic characterization of primary and metastatic PanNETs. Our findings revealed an enrichment of CDKN2A homozygous deletions and TSC2 somatic mutations in metastatic PanNETs when compared to non-metastatic lesions. Tumor evolution analysis further revealed the acquisition of such genetic alterations as late events in the progression of these neoplasms, conferring poor survival outcomes to the affected patients. Biallelic loss of DNA damage repair genes, ATRX and/or DAXX, was associated with a high fraction of the genome altered in PanNETs, with pathogenic alterations affecting those genes also being associated with a homologous recombination deficiency signature. These findings highlight molecular mechanisms driving PanNET progression and underscore the need for further molecular characterization and tumor evolution studies to evaluate targeted therapies for such a challenging disease.

Subject terms: Cancer genomics, Neuroendocrine cancer, Pancreatic cancer

Introduction

Pancreatic neuroendocrine tumors (PanNETs) are low-incidence diseases accounting for less than 3% of all pancreatic malignancies, but their prevalence is currently rising1. Patients with PanNET present metastases at diagnosis in 60–80% of cases1, and their prognosis differs widely, with some tumors having an indolent nature, with a reasonable length of survival even with a metastatic presentation, and others being extremely aggressive with a poor prognosis2. Recent advances in sequencing technologies have uncovered the molecular basis of numerous cancers and led to new prognostic classification systems and actionable targets3. Indeed, it has been demonstrated that increasing numbers of molecular pathways are involved in the biology and clinical behavior of PanNETs, such as DNA damage repair, chromatin remodeling, telomere alteration, PI3K/AKT/mTOR and p53/cell cycle signaling pathways3–5. DNA sequencing analysis of PanNETs has identified recurrent somatic mutations affecting MEN1, DAXX, and ATRX3–6. Loss of function of DAXX or ATRX, results in telomere dysfunction, due to the activation of homologous recombination in telomeric DNA, and consequently leads to genomic instability. The inactivation of these genes is also implicated in the activation of the alternative lengthening of telomeres (ALT) mechanism3,6,7. Additional mutations in the PI3K/AKT/mTOR pathway were also found in these neoplasms, such as those affecting TSC1/2 and PTEN tumor suppressor genes, and pronounced losses of these genes were correlated with liver metastases, shorter time to progression, and shorter disease-free and overall survival5,6,8. More recently, high frequencies of CDKN2A copy number (CN) losses were observed in metastatic PanNETs9. Despite improvements in molecular profiling and prognostic grading and staging systems, it remains a challenge to predict the clinical behavior of PanNETs and the response to specific therapies, given the high degree of heterogeneity of these tumors. While most PanNETs present as advanced disease, the available systemic therapies provide modest benefits. Therefore, there is a particular need to develop more effective systemic therapies based on the molecular profile of PanNETs. Specifically, determining the molecular basis of the wide phenotypic variations underlying this disease is crucial to develop future personalized therapies.

In an attempt to meticulously characterize the genetic repertoire and unravel clinically informative alterations affecting PanNETs, we analyzed the molecular features of 192 primary PanNETs and metastases from Nguyen et al.10, 6 primary tumors with matched metastases from Cowzer et al.11 and Jee et al.12, and 33 primary PanNETs from the Pan-Cancer Analysis of Whole Genomes (PCAWG)13.

Results

Clinical features of pancreatic neuroendocrine tumors

Tumors from 211 patients subjected to targeted massively parallel sequencing via MSK-IMPACT from 2014 to 2020 were identified10. Nineteen cases were excluded due to the absence of any genetic alterations, leaving 192 cases in the study cohort, including 87 primary tumors and 105 unmatched metastases. The median age of primary and metastatic tumors was of 58 and 59 years, respectively (range of primaries, 26-86 years; range of metastases, 28-82 years; Supplementary Table S1). Of the 87 primary PanNETs and 105 metastases, 14 (16%) and 50 patients (48%) died of the disease, respectively. The median follow-up time of the primary tumors was of 30.9 months (range 0.92–70.7 months), and of the metastases was 17.1 months (range 0.46-71.9 months; Supplementary Table S1). Of the 6 patients with primary tumors and matched metastases, 3 died of the disease11,12.

Genetic characterization of primary and metastatic pancreatic neuroendocrine tumors

We next assessed the genetic differences between primary tumors and metastases from PanNETs (Fig. 1). No statistically significant differences were observed regarding the tumor mutational burden (TMB) and fraction of the genome altered (FGA) between the metastases and primary tumors (2.59 vs. 2.94, p = 0.052; 37.3% vs. 47.7%, p = 0.316, respectively; Supplementary Fig. S1a, b). Genetic alterations were found to be significantly more frequent in the metastases when compared to primary tumors, such as those affecting CDKN2A (23% vs. 10%, p < 0.05), CDKN2B (21% vs. 8%, p < 0.05), KRAS, ATM (10% vs. 1%, p < 0.01, for both genes), LATS2 (8% vs. 0%, p < 0.01), BRAF (7% vs 0%, p < 0.05), FAT1 and APC (6% vs 0%, p < 0.05, for both genes) (Fig. 1a). As previously reported in PanNETs3,6, DAXX and ATRX genetic alterations were found to be mutually exclusive (p = 0.005; Fig. 1b). While the majority of the DAXX mutations were deemed clonal, a substantial proportion of mutations affecting ATRX were found to be subclonal (9/38, 24%), mainly in the primary tumors (6/22, 27%; Supplementary Fig. S1c, Supplementary Table S2). When looking at CN alterations, the frequency of gains of chromosome 22p and 22q was significantly higher in the metastases than in primary tumors (15.2% vs 2.3% and 12.4% vs 2.3%, respectively, p < 0.01; Fig. 1c). In addition, loss of chromosome 4q was found to be significantly more frequent in the metastases when compared to primary tumors (21% vs 9%, p < 0.05), as were the losses of chromosomes 9p and 9q (40% vs. 20%, and 24.8% vs. 9.2%, respectively, p < 0.01; Fig. 1c).

Fig. 1. Genomic features of primary neoplams and metastases from pancreatic neuroendocrine tumors.

Fig. 1

Open in a new tab

Comparisons between primary neoplams and metastases from pancreatic neuroendocrine tumors (PanNETs) for a, recurrent somatic alterations, b, mutual exclusivity analysis between DAXX and ATRX, and c, copy number gains and losses. Comparisons between liver metastases from PanNETs stratified according to the number of distant metastases for d, recurrent somatic alterations. Alteration types and number of metastases are color-coded according to the legend. Statistical significance was evaluated using Fisher’s exact test. CN copy number, PanNETs pancreatic neuroendocrine tumors.

We have also stratified the primary PanNETs according to the presence/absence of metastases. Such stratification revealed a significantly higher TMB in primary tumors with evidence of metastases than in primary tumors with no metastases (2.59 vs 0.98, p = 0.018; Supplementary Fig. S2a). No statistically significant differences regarding the FGA was observed between the former and the latter (45.6% vs 26.3%, p = 0.082; Supplementary Fig. S2b). When looking at the recurrent somatic mutations from both the MSK-IMPACT10 and Scarpa et al.6 data however, we observed a statistically significant enrichment of mutations affecting DAXX, ATRX, TSC2, TP53 and ARID1A in primary tumors with evidence of metastasis when compared to primary tumors with no metastasis (33% vs 19%, 25% vs 9%, 11% vs 3%, 10% vs 2%, and 8% vs 1%, respectively, all statistical differences at p < 0.05; Supplementary Fig. S2c). Furthermore, we observed in primary PanNETs a significantly higher frequency of gains of chromosomes 7 (50% vs. 17.6%, p < 0.05) and 14 (52.9% vs. 23.5%, p < 0.05), and losses of chromosomes 10p (50% vs. 17.6%, p < 0.05), 10q (52.9% vs 17.6%, p < 0.05), 21p (37.1% vs 11.8%, p < 0.05) and 21q (40% vs 11.8%, p < 0.05; Supplementary Fig. S2d).

Most of the metastases from PanNETs were found in the liver (84%; Supplementary Table S1). When stratifying the liver metastases according to the number of metastases each corresponding patient harbors, we noticed a significantly higher frequency of LATS2 genetic alterations in the metastases from patients harboring more than 2 metastases when compared with those harboring 2 metastases or less (15% vs 0%, p < 0.05; Fig. 1d).

Biallelic loss of DAXX or ATRX contributes to an increased fraction of the genome altered in pancreatic neuroendocrine tumors

When looking at the CN profiles of both primary tumors and metastases, we noticed two main sample groups: one containing extremely high numbers of CN alterations and another harboring samples with relatively quiet genomes (Supplementary Fig. S3a, b). Considering that PanNETs often present high levels of genetic instability at the chromosomal level6, we sought to compare the molecular features of primary and metastatic PanNETs with low FGA, with primary and metastatic PanNETs with high FGA, respectively (Fig. 2). In comparison to primary tumors with low FGA, primary tumors displaying high FGA were found to harbor significantly higher frequencies of MEN1 (40% vs. 79%, p < 0.001), DAXX (13% vs 45%, p < 0.01) and PTEN (0% vs 14%, p < 0.05) genetic alterations, with the latter harboring a substantial proportion of subclonal mutations (Fig. 2a, Supplementary Fig. S4a, Supplementary Table S2). The metastases with high FGA were found to have significantly higher frequencies of genetic alterations than metastases displaying low FGA, such as those affecting MEN1 (70% vs. 23%, p < 0.001), DAXX (42% vs. 12%, p < 0.001), TSC2 (36% vs. 4%, p < 0.001), ATRX (28% vs. 6%, p < 0.01), and CDKN1A (11% vs. 0%, p < 0.05; Fig. 2b). Of note, while only 6% (2/35) of the mutations affecting DAXX and ATRX were found to be subclonal in the metastases with high FGA, 44% (4/9) of mutations affecting these genes were deemed subclonal in the metastases with low FGA (Supplementary Fig. S4b, Supplementary Table S2).

Fig. 2. Genomic features of primary neoplams and metastases from pancreatic neuroendocrine tumors with low and high fraction of the genome altered.

Fig. 2

Open in a new tab

Comparisons between primary neoplasms and metastases from pancreatic neuroendocrine tumors (PanNETs) with low and high fraction of the genome altered for a and b, recurrent somatic alterations, and c, tumor mutational burden. Comparisons between PanNETs harboring single-hit alterations affecting DAXX and ATRX, and double genetic alterations affecting the same genes for d, fraction of the genome altered. Alteration types are color-coded according to the legend. Statistical significance was evaluated in (a) and (b) using Fisher’s exact test, and in (c) and (d) using the Mann-Whitney U test. FGA, fraction of the genome altered; PanNETs, pancreatic neuroendocrine tumors.

A significantly higher TMB was observed in both primary tumors and metastases with high FGA than in primary tumors and metastases with low FGA (primary tumors, 3.3 vs 1.72, p = 0.017; metastases: 3.5 vs 1.84, p = 0.01; Fig. 2c). We noticed that loss of heterozygosity (LOH) followed the majority of DAXX and ATRX genetic alterations in both primary tumors and metastases displaying high FGA (Fig. 2a, b). Compared to single alterations affecting DAXX or ATRX, samples with double genetic alterations affecting one of these two genes harbored a significantly higher FGA (15.3% vs 90%, p < 0.0001; Fig. 2d).

When looking at the survival curves, however (Fig. 3a, b), we noticed that metastatic patients harboring ATRX or DAXX genetic alterations had better survival outcomes than patients not harboring genetic alterations affecting these two genes, with this difference being statistically significant only for DAXX (p = 0.035; Fig. 3b). When stratifying the patients with biallelic loss of ATRX or DAXX according to their survival status, we found that 64% of patients who died from the disease harbored mutations in the TSC2 gene, being this difference statistically significant in comparison with the alive patients (64% vs 26%, p < 0.05; Supplementary Fig. S4c). Homozygous deletions of CDKN2A and CDKN2B genes were also more prevalent in deceased patients, though not statistically significant (50% vs 32%, p > 0.05; Supplementary Fig. S4c).

Fig. 3. Survival analysis of metastatic pancreatic neuroendocrine tumors harboring genetic alterations affecting ATRX, DAXX, MEN1, TSC2, CDKN2A and KRAS.

Fig. 3

Open in a new tab

Kaplan-Meier overall survival curves for 175 pancreatic neuroendocrine tumors stratified by altered gene status, namely a, ATRX, b, DAXX, c, MEN1, d, TSC2, e, CDKN2A (homozygous deletions), f, KRAS (G12D/V/R). Statistical significance was evaluated in (a), (b), (c), (d), (e) and (f) using the log-rank test. PanNETs, pancreatic neuroendocrine tumors.

CDKN2A homozygous deletions and TSC2 genetic alterations confer poor survival outcomes to the affected patients and are late events in the malignant progression of pancreatic neuroendocrine tumors

Considering the overall survival results for patients harboring ATRX or DAXX genetic alterations, we focused our attention on other altered cancer-related genes in metastatic PanNETs (Fig. 3c–f). Interestingly, we found that metastatic patients harboring TSC2 genetic alterations, as well as CDKN2A homozygous deletions and KRAS hotspot codon mutations (G12D/V/R) had worse survival outcomes when compared with patients not harboring TSC2, CDKN2A or KRAS genetic alterations, being this difference statistically significant for both CDKN2A and KRAS affected genes (CDKN2A homozygous deletions, p = 0.030, KRAS G12D/V/R, p = 0.001; Fig. 3e, f).

To further validate our previous findings, we retrieved the genomic data from six patients with primary PanNETs and matched metastases11,12. Except for one case, pathogenic mutations affecting MEN1, DAXX or ATRX were found to be clonal and at the trunk of the mutational phylogenetic trees (Fig. 4a–c, Supplementary Fig. 5a, c, Supplementary Table S3). Primary tumors and metastases with biallelic losses of ATRX or DAXX displayed extremely high FGA (Fig. 4a–c, Supplementary Fig. 5a). The sole case with one primary tumor and two matched metastases had the three components sharing three somatic mutations, including a clonal ATRX missense mutation, a MEN1 frameshift mutation, and a DAXX missense mutation (both mutations deemed subclonal in the primary tumor and becoming fully clonal in the metastases; Fig. 4c, Supplementary Table S3). Loss of heterozygosity affecting the three genes was found to occur only in the two metastases, with a marked increase of the FGA from the primary tumor to the first metastasis (Fig. 4c). The second metastasis was found to harbor 19 clonal and 154 subclonal mutations, including a subclonal TP53 hotspot mutation (Fig. 4c, Supplementary Table S3). Such high TMB can be explained by the treatment with alkylating agents and the corresponding dominant signature 11 observed in this component, as described by Cowzer et al.11 (Fig. 4c). Acquisition of CDKN2A homozygous deletions and TSC2 mutations was found to occur in three and in two metastatic components, respectively (Fig. 4a–c, Supplementary Table S3). Hotspot mutations affecting CDKN2A and TP53 were also found to affect the metastatic component of an additional two cases (Supplementary Fig. 5b, c, Supplementary Table S3).

Fig. 4. Clonal composition and phylogenetic analysis of primary pancreatic neuroendocrine tumors and matched metastases.

Fig. 4

Open in a new tab

Clonal composition and phylogenetic analysis of three cases (a–c). Heatmaps depicting the frequency of genetic alterations and cancer cell fractions of each somatic mutation in the primary pancreatic neuroendocrine tumors (PanNETs) and matched metastasis of a given case are shown ((a) and (b), top left; (c), top). The alteration types and cancer cell fraction are color-coded according to the legend. Copy number plots depicting segmented Log2 ratios (y-axis) according to genomic position (x-axis) of primary PanNETs and matched metastases are depicted ((a) and (b), top right; (c), bottom). Phylogenetic trees of primary PanNETs and matched metastases are shown ((a) and (b), bottom left; (c), middle). Trunk and branches are colored according to clusters, and the number of somatic mutations that result in the divergence of a clone/subclone from its ancestor is shown. Pathogenic mutations that define a given clone are depicted. Phylogenetic trees based on copy number alterations are shown ((a) and (b), bottom right; (c), bottom left). The numbers alongside the branches represent the number of copy number alterations. Gains and losses are shown in parentheses. Alkyl alkylating agent, dROS damage by reactive oxygen species, M metastasis, MSI microsatellite instability, NA not applicable, P primary, UV ultra-violet light, T trunk.

Genetic alterations affecting ATRX and DAXX are associated with homologous-recombination deficiency features in primary pancreatic neuroendocrine tumors

To further characterize which PanNET display a homologous recombination deficiency (HRD) feature, we next looked at the mutational signatures inferred from somatic mutations of 33 primary PanNETs subjected to whole-genome sequencing (WGS)13 and performed a genetic comparison between PanNETs with no evidence of HRD features and with PanNETs with patterns of HRD (Fig. 5). Mutational signatures inferred from somatic mutations of these 33 primary PanNETs revealed the presence of 16 PanNETs with no mutational signature 3 exposure (“non-HRD” group), 12 PanNETs harboring a second dominant mutational signature 3 (“HRD-ambiguous” group), and 5 PanNETs with a dominant mutational signature 3 (“HRD-like” group; Fig. 5a). When looking at the genetic repertoire of these three groups, we observed a step-wise increment of the frequency of genetic alterations affecting ATRX and DAXX from the “non-HRD” to the “HRD-like” group, with the latter having a higher frequency of these alterations when compared to the “HRD-ambiguous” (100% vs 42%, p = 0.029) and “non-HRD” (100% vs 37%, p = 0.035) groups (Fig. 5b). As expected, the ALT probability observed in the HRD dominant cases was extremely high (>95%; Fig. 5b). Of note, the “HRD-like” group also harbored TSC2 pathogenic mutations affecting 40% of patients, with these alterations being absent in both the “HRD-ambiguous” and “non-HRD” groups (Fig. 5b).

Fig. 5. Genomic features of primary pancreatic neuroendocrine tumors with and without homologous recombination deficiency features.

Fig. 5

Open in a new tab

Comparisons between primary pancreatic neuroendocrine tumors stratified according to the presence/absence of homologous recombination deficiency features assessed by mutational signatures (a) for b, recurrent somatic mutations and c, copy number alterations (gains, losses, and amplifications). Alteration types are color-coded according to the legend. Statistical significance was evaluated using Fisher’s exact test. ALT, alternative lengthening of telemores; BER base-exchange repair, CN copy number, HRD homologous recombination deficiency, MSI microsatellite instability, PanNETs pancreatic neuroendocrine tumors.

When looking at CN alterations, we observed that between 25-80% of all 33 PanNETs harbored amplifications affecting chromosomes 4, 5, 7, 9, 12, 13q, 14q, 17, 18, 19, and 20, with chromosome 5 being one of the the most affected chromosome, as seen by the frequency of TERT gene amplifications present in 36% of all cases (Fig. 5b, c). In addition, a significantly higher frequency of amplifications of chromosome 9p was observed in the “HRD-like” tumors when compared to “non-HRD” PanNETs (80% vs 25%, p < 0.05; Fig. 5c).

Pancreatic neuroendocrine tumors harbor recurrent somatic genetic alterations affecting genes in the PI3K/AKT/mTOR, p53, and DNA damage repair pathways

Given the fact that the most affected genes in PanNETs are located in distinct signaling pathways, we performed a pathway analysis based on the somatic mutations, structural variants, CN amplifications, and homozygous deletions present in the 192 PanNETs included in this study. Such analysis revealed an enrichment of genetic alterations targeting the PI3K/AKT/mTOR, DNA damage repair, and p53/cell cycle pathways (Fig. 6a, b). The most affected gene was the tumor suppressor gene MEN1, with half of the patients harboring MEN1 genetic alterations. The frequency of genetic alterations affecting TSC2, an additional tumor suppressor gene, was also found to be high, mainly in the metastases (20%; Fig. 6a). Both tumor suppressor genes, DAXX and ATRX were found to affect 46% of all cases, with a higher prevalence in primary tumors with evidence of metastases (Fig. 6b). PanNETs harboring such genetic alterations have potentially a defective DNA damage repair mechanism/signaling pathway. When looking downstream of such pathway, ATM and TP53 genetic alterations affected 10% and 20% of the metastases, respectively (Fig. 6b). In addition, and as observed above, a step-wise increment in the frequency of homozygous deletions affecting CDKN2A was observed from primary tumors to the metastases. Neoplasms with such alterations have a defective cell cycle pathway (Fig. 6b).

Fig. 6. Signaling pathways most affected by genetic alterations present in pancreatic neuroendocrine tumors.

Fig. 6

Open in a new tab

Frequency of activating (red) or loss-of-function (blue) somatic genetic alterations affecting genes in pancreatic neuroendocrine tumors (PanNETs) and present in the canonical a PI3K/AKT/mTOR, and b p53/DNA damage repair/cell cycle signaling pathways. The frequency of primary PanNETs with no evidence of metastases, primary tumors with evidence of metastases, and metastatic samples is depicted under the gene name. Statistical significance was evaluated using using Mann-Whitney U test.

Discussion

Despite improvements in the molecular profiling of PanNETs over the last decade, predicting the clinical behavior of these tumors and their response to specific therapies remains challenging. Such difficulties are mainly due to the high degree of heterogeneity observed in these neoplasms. With this in mind, we sought to meticulously characterize the genetic repertoire of primary PanNETs and metastases10–12 and unravel clinically informative alterations affecting these tumors. When looking at the molecular features of 87 primary PanNETs and 105 unmatched metastases, we noticed a significantly higher frequency of CDKN2A homozygous deletions in the metastases when compared to primary tumors, as well as poor survival outcomes in patients harboring these deletions. In addition, tumor evolution analysis of 6 primary tumors with matched metastases revealed that 4 metastases harbored private genetic alterations affecting CDKN2A. Previous studies have also observed an association of CDKN2A homozygous deletions with metastases from PanNETs9,14. Guo et al. also observed these deletions occurring exclusively in the metastases (4/10 patients) compared to matched primary tumors9. Beyond CN alterations, other CDKN2A inactivating mechanisms were also linked with survival outcomes in PanNETs, or with the risk of developing the disease. For instance, CDKN2A promoter hypermethylation has been associated with poor prognosis in pancreatic cancer15, and more specifically, in pancreatic endocrine neoplasms16. Pathogenic germline variants affecting CDKN2A have also been shown to be associated with the risk of developing such disease17. All of these observations, combined with the clinical outcome obtained in the present study, suggests that CDKN2A should be regarded as a diagnostic and prognostic biomarker in PanNETs. Applying genotyping analysis of this gene in clinical practice could greatly improve patient risk stratification.

Akin to patients harboring CDKN2A homozygous deletions, those harboring TSC2 genetic alterations were also found to have poor survival outcomes, and a subset of metastases was found to harbor private TSC2 somatic mutations coupled with LOH when compared to their matched primary tumors. Loss of TSC2 by immunohistochemistry in PanNETs was previously demonstrated to be associated with significantly shorter overall survival18. In the study from Guo et al., 3 out of 10 metastases were found to harbor private TSC2 somatic mutations when compared with their matched primary tumors9. While CDKN2A belongs to the cell cycle signaling pathway, TSC2 is a negative regulator of the mTOR signaling pathway. Hence, although patients with advanced PanNETs showed relatively high response rates when treated with capecitabine and temozolomide19, it would be interesting to determine if mTOR pathway inhibitors such as everolimus and its analog sirolimus could be promising drugs against PanNETs genotyped for the somatic loss of TSC2.

We also determined the CN alterations present in the 192 PanNETs and noticed that biallelic losses of DAXX or ATRX were associated with high frequencies of the FGA in both primary tumors and metastases. Such associations were previously described in PanNETs: Marinoni et al. have demonstrated that immunohistochemical loss of DAXX and ATRX was associated with chromosomal instability in 39 out of 142 PanNETs20; Scarpa et al. have also shown that biallelic inactivation of DAXX or ATRX were strongly associated with increased telomere length in 26 out of 86 PanNETs6. Because both tumor suppressor genes, DAXX and ATRX, regulate p53 chromatin binding and DNA damage response21, complete loss of one of these two genes would inactivate/disrupt the DNA damage repair signaling pathway and consequently compromise the genomic stability of PanNETs. Loss of ATRX or DAXX were reported as an independent prognostic factor for poor disease-free survival in these tumors22. Nonetheless, ATRX/DAXX mutations remain a topic of debate as prognostic markers. In fact, there have been conflicting studies, raising the possibility that ATRX/DAXX mutations may be associated with poor overall survival in low stage tumors, but with better survival outcomes in advanced stage or metastatic tumors23. Our findings suggest a clear and better survival outcome in metastatic patients harboring DAXX mutations, and no association between survival outcomes and ATRX genetic alterations in advanced PanNETs.

With the notion that a subset of PanNETs displays high levels of HRD, we also looked at the mutational signatures previously inferred in 33 primary PanNETs subjected to WGS from PCAWG13. We found that five tumors had a dominant signature 3 (“HRD-like”) associated with HRD features. When comparing the genetic repertoire of these tumors with those harboring a second dominant signature 3 (“HRD-ambiguous”) or with those with no evidence of signature 3 (“non-HRD”), we also found that pathogenic genetic alterations affecting ATRX and DAXX were enriched in the “HRD-like” group. An elegant study from Juhász et al. demonstrated that ATRX depletion in HeLa cells abolishes DNA repair synthesis and prevents the formation of sister chromatid exchanges at exogenously induced DNA double-stranded breaks (DSBs)24. The authors also described that ATRX partners with DAXX to incorporate histone H3.3 during homologous recombination-mediated repair of such exogenously induced DSBs24. Such observations definitely warrants additional studies regarding the association between HRD features and pathogenic mutations affecting ATRX and DAXX in PanNETs. To further investigate the function of ATRX in the DNA damage repair signaling pathway, Garbarino et al. created isogenic wild-type and ATRX knockout glioma cell lines using CRISPR-based gene targeting. This study revealed that loss of ATRX confers sensitivity to poly(ADP)- ribose polymerase (PARP) inhibitors, which was linked to an increase in replication stress25. Investigating whether PARP inhibitors could also be effective in ATRX- or DAXX-depleted PanNETs would be of great clinical interest to the affected patients.

Our study has significant limitations. As we could only retrieve the molecular and clinical features from pan-cancer studies, we could not assess the telomere lengths in the MSK-IMPACT-derived cohort, nor could we perform clinical associations according to tumor grade. The results from the tumor evolution analysis were limited by the reduced number of matched PanNETs included, and restricted by the use of in silico bioinformatics tools only. Finally, the mutational re-analyses performed in the metastatic cohort were restricted to cancer-related genes only, and we cannot rule out that other genetic alterations could play a role in these tumors studied here.

In conclusion, and considering the extremely high levels of heterogeneity in PanNETs, we highlighted relevant clinically informative alterations in advanced PanNETs, especially those affecting CDKN2A, TSC2 (both conferring poor survival outcomes), ATRX and DAXX (both inducing chromosomal instability/homologous recombination deficiencies), that may be explored therapeutically or included in a daily routine for stratification of patients with a poorer prognosis. This knowledge may contribute to the development of targeted therapies tailored to the unique molecular vulnerabilities of these tumors. Understanding the molecular landscape of PanNETs may improve patient stratification for personalized treatments, potentially increasing efficacy while minimizing side effects, leading to innovative diagnostic tools, prognostic markers, and transforming the management and prognosis of this complex disease.

Methods

Case selection

All patients with PanNETs under an institutional review board–approved protocol at Memorial Sloan Kettering Cancer Center and whose tumors were subjected to targeted massively parallel sequencing via Memorial Sloan Kettering–Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT)10 from 2014 to 2020 were identified (n = 211 cases sequenced) and accessed through cBioPortal. Cases not harboring any genetic alteration were excluded from the analysis (n = 19), leaving 192 PanNETs selected (Supplementary Table S1). Clinicopathological data, including age at diagnosis, sample type (87 primary tumors and 105 unmatched metastases), location and number of the metastases, and past cancer history, were also retrieved10 (Supplementary Table S1). Of the 87 primary PanNETs, 70 patients had evidence of metastases. In addition, 6 patients with primary PanNETs and matched metastases whose tumors were also subjected to targeted massively parallel sequencing via MSK-IMPACT were identified11,12. Besides this, 33 additional primary PanNETs subjected to WGS were identified and retrieved from the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium13.

Massively Parallel Sequencing Analysis

Tumor and matched normal DNA samples were subjected to targeted massively parallel sequencing using MSK-IMPACT, which targets all exons and selected introns of 341-468 cancer genes26,27. Relevant genomic data derived from MSK-IMPACT included somatic mutations, structural variants, copy number (CN) alterations (gains and losses, gene amplifications, and homozygous deletions), TMB, and FGA10,11. The CN segment files were also retrieved to determine whether genes harboring somatic mutations were targeted by LOH28. CN alterations were estimated by taking the median of the log ratios per chromosomal arm. The absolute CN values were then calculated according to the tumor purity of a given sample, as previously described28. The cancer cell fractions (CCFs) of all somatic mutations were computed using ABSOLUTE29, taking into account the variant allele frequency (VAF) and the ploidy status of each somatic variant. Solutions from ABSOLUTE were manually reviewed. A mutation was classified as clonal if its probability of being clonal was >50% or if the lower bound of the 95% confidence interval of its CCF was >90%, as previously described30. Mutations that did not meet the above criteria were considered subclonal. Relevant genomic data from PCAWG included somatic mutations, structural variants, CN alterations, TMB, and mutational signatures13. CN alterations were inferred based on the median of the absolute CN values of a given gene, clustered by chromosomal arm. The probability of the ALT mechanism previously calculated for all the 33 PCAWG samples was also retrieved31.

Genetic comparisons between primary and metastatic pancreatic neuroendocrine tumors, and with low and high fraction of the genome altered

The frequencies of somatic genetic alterations of 87 primary PanNETs were compared with those of 105 unmatched metastases (Supplementary Table S1). In addition, the frequencies of these alterations were also compared between the number of distant metastases each patient harbored. We used a cutoff of 2 distant metastases to stratify these patients for increased statistical power. We have also compared the frequencies of somatic genetic alterations between PanNETs with low and high FGA. For this, we used the median of the FGA as a cutoff (primary tumors: 37.3%; metastases: 47.7%) to stratify the samples into low and high FGA. All of these genetic comparisons (including the TMB), were performed using MSK-IMPACT-derived samples only (n = 192). The sole exception regards the mutation comparisons between primary tumors with evidence of metastases and primary tumors with no evidence of metastases. For these, we also added the mutation information of the 98 PanNETs retrieved from Scarpa et al.6. In that way, we ended up accruing 97 primary tumors with no evidence of metastases (17 primary tumors subjected to MSK-IMPACT and 80 primary tumors subjected to WGS (Scarpa et al.6)), and 88 primary tumors with evidence of metastases (70 primary tumors subjected to MSK-IMPACT and 18 primary tumors subjected to WGS (Scarpa et al.6)). For all of the genetic comparisons, only the 341 genes targeted by the smallest MSK-IMPACT panel were used.

Tumor evolution analysis of primary pancreatic neuroendocrine tumors with matched metastasis

The clonal composition of each somatic mutation was determined using ABSOLUTE29, as described above. A given genetic alteration (somatic mutations and homozygous deletions) was considered “shared” if it was present in both the primary and matched metastatic lesion. We defined alterations “private to the primary lesion” and “private to the metastatic lesion” as those present only in the primary tumor or in the metastasis, respectively. Mutational signatures were defined by deconstructSigs using all single-nucleotide variants (SNVs)32 at default parameters, as previously described33, for samples with ≥10 somatic SNVs. To reconstruct the phylogeny of the primary and matched metastatic PanNETs, we used Treeomics34 based on all genetic alterations identified and the CCF values calculated above, as previously described35,36. For the construction of phylogenetic trees based on CN alterations, major and minor CNs were modeled through transducer-based pairwise comparison functions using MEDICC37, assuming a diploid state with no CN alterations to root the phylogenies.

Genetic comparisons between pancreatic neuroendocrine tumors according to the presence or absence of homologous recombination deficiency features

For each of the 33 PanNETs subjected to WGS, we retrieved the frequency of the mutational signatures from the PCAWG study13. Each sample was stratified according to the frequency of the mutational signature associated with homologous recombination deficiency (HRD, Signature 3)38. PanNETs with a dominant signature 3 were considered “HRD-like” tumors (n = 5), those with a second dominant signature 3 were considered “HRD-ambiguous” tumors (n = 12), and those with no evidence of signature 3 were considered “non-HRD” tumors (n = 16). The frequencies of somatic genetic alterations were compared between these three groups.

Pathway analyses

A DAVID pathway analysis was conducted based on genes affected by somatic genetic alterations39. Pathways found to be statistically significantly enriched (p < 0.01) in PanNETs and previously curated and reported in Sanchez-Vega et al. were selected40. Additionally, a mutual exclusivity analysis was performed using combinations of mutually exclusive alterations (CoMET) with a pairwise Fisher’s exact test41.

Statistical analysis

Statistical analyses were performed using R v3.1.2. Fisher’s exact tests were employed for comparisons between categorical variables, whereas Mann–Whitney U tests were used for continuous variables. All tests were two-sided, and a p-value < 0.05 was considered statistically significant. Survival analyses were performed using univariate Cox regressions, and Kaplan–Meier curves were displayed using the R package survival42. For the survival analyses comprising the most affected genes in PanNETs, only the follow-up time from metastatic patients (n = 175) was considered. For a proper visualization, only recurrently altered genes per sample type are represented in the heatmaps.

Supplementary information

Supplementary information (4.2MB, pdf)

Acknowledgements

This study was funded by national funds by FCT—Fundação para a Ciência e Tecnologia, I.P., through a research contract to AdCP (2022.06547.CEECIND), to JV (2022.00276.CEECIND), and to the project PTDC/MED-ONC/0531/2021—CTRL+ALT+CEL: how ATRX controls an alternative program in the β-cell. Further funding was obtained from the project “2022-C05IO101-02—Agenda Illiance (Bosch, project nº 46)—PPS4—OLI health”, with reference C644919832-00000035, funded by PRR—Plano de Recuperação e Resiliência e pelos Fundos Europeus NextGenerationEU, through «Agendas para a Inovação Empresarial». Further funding was obtained from Consortium PORTO.CCC—Porto Comprehensive Cancer Center, supported by Norte Portugal Regional Operational Programme (NORTE 2020), under the PORTUGAL 2020 Partnership Agreement, through the European Regional Development Fund (ERDF) (NORTE-01-0145-FEDER-072678).

Author contributions

Concept and design: A.D.C.P. and J.V. Acquisition, analysis, or interpretation of data: all authors. Drafting the manuscript: T.T.J., A.D.C.P,. and J.V. Critical revision of the manuscript for important intellectual content: T.T.J., L.C., J.M.L., P.S., A.D.C.P,. and J.V. Bioinformatics and statistical analysis: L.R., R.S.M., L.F., and A.D.C.P. Supervision: A.D.C.P. and J.V. Final approval of completed version of manuscript: all authors.

Data availability

All clinical and molecular data was retrieved from the cBioPortal database ([https://www.cbioportal.org/)](https://www.cbioportal.org/)).

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Arnaud Da Cruz Paula, João Vinagre.

Supplementary information

The online version contains supplementary material available at 10.1038/s41698-025-01210-2.

References

  • 1.Dasari, A. et al. Trends in the Incidence, Prevalence, and Survival Outcomes in Patients With Neuroendocrine Tumors in the United States. JAMA Oncol.3, 1335–1342 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Dasari, A., Shen, C., Devabhaktuni, A., Nighot, R. & Sorbye, H. Survival According to Primary Tumor Location, Stage, and Treatment Patterns in Locoregional Gastroenteropancreatic High-grade Neuroendocrine Carcinomas. Oncologist27, 299–306 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Singhi, A. D. et al. Alternative Lengthening of Telomeres and Loss of DAXX/ATRX Expression Predicts Metastatic Disease and Poor Survival in Patients with Pancreatic Neuroendocrine Tumors. Clin. Cancer Res.23, 600–609 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Han, X. et al. Aberration of ARID1A Is Associated With the Tumorigenesis and Prognosis of Sporadic Nonfunctional Pancreatic Neuroendocrine Tumors. Pancreas49, 514–523 (2020). [DOI] [PubMed] [Google Scholar]
  • 5.Jiao, Y. et al. DAXX/ATRX, MEN1, and mTOR pathway genes are frequently altered in pancreatic neuroendocrine tumors. Science331, 1199–203 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Scarpa, A. et al. Whole-genome landscape of pancreatic neuroendocrine tumours. Nature543, 65–71 (2017). [DOI] [PubMed] [Google Scholar]
  • 7.Clynes, D. et al. Suppression of the alternative lengthening of telomere pathway by the chromatin remodelling factor ATRX. Nat. Commun.6, 7538 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Missiaglia, E. et al. Pancreatic endocrine tumors: expression profiling evidences a role for AKT-mTOR pathway. J. Clin. Oncol.28, 245–55 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Guo, Y. et al. Molecular and Functional Heterogeneity of Primary Pancreatic Neuroendocrine Tumors and Metastases. Neuroendocrinology113, 943–956 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Nguyen, B. et al. Genomic characterization of metastatic patterns from prospective clinical sequencing of 25,000 patients. Cell185, 563–575 e11 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Cowzer, D. et al. Clinical utility of plasma cell-free DNA in pancreatic neuroendocrine neoplasms. Endocr. Relat. Cancer31, e230292 (2024). [DOI] [PubMed] [Google Scholar]
  • 12.Jee, J. et al. Automated real-world data integration improves cancer outcome prediction. Nature636, 728–736 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.ICGC/TCGA Pan-cancer analysis of whole genomes consortium Pan-cancer analysis of whole genomes. Nature 578, 82–93 (2020).
  • 14.Roy, S. et al. Loss of Chromatin-Remodeling Proteins and/or CDKN2A Associates With Metastasis of Pancreatic Neuroendocrine Tumors and Reduced Patient Survival Times. Gastroenterology154, 2060–2063.e8 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Tang, B. et al. Clinicopathological significance of CDKN2A promoter hypermethylation frequency with pancreatic cancer. Sci. Rep.4, 13563 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.House, M. G. et al. Aberrant hypermethylation of tumor suppressor genes in pancreatic endocrine neoplams. Ann. Surg.238, 423–31 (2003). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Campa, D. et al. Common germline variants within the CDKNA/B region affect risk of pancreatic neuroendocrine tumors. Sci. Rep.23, 39565 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Uemura, J. et al. Immunohistochemically Detected Expression of ATRX, TSC2, and PTEN Predicts Clinical Outcomes in Patients With Grade 1 and 2 Pancreatic Neuroendocrine Tumors. Ann. Surg.274, e949–e956 (2021). [DOI] [PubMed] [Google Scholar]
  • 19.Kunz, P. L. et al. Randomized Study of Temozolomide or Temozolomide and Capecitabine in Patients With Advanced Pancreatic Neuroendocrine Tumors (ECOG-ACRIN E2211). J. Clin. Oncol.41, 1359–1369 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Marinoni, I. et al. Loss of DAXX and ATRX are associated with chromosome instability and reduced survival of patients with pancreatic neuroendocrine tumors. Gastroenterology146, 453–60 e5 (2014). [DOI] [PubMed] [Google Scholar]
  • 21.Gulve, N. et al. DAXX-ATRX regulation of p53 chromatin binding and DNA damage response. Nat. Commun.13, 5033 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Wang, F. et al. Prognostic significance of altered ATRX/DAXX gene in pancreatic neuroendocrine tumors: a meta-analysis. Front Endocrinol.12, 691557 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Veld, B. R., Hackeng, W. M., Luchini, C., Brosens, L. A. & Dreijerink, K. M. Clinical relevance of ATRX/DAXX gene mutations and ALT in functioning pancreatic neuroendocrine tumors. Endocr. Pathol.36, 3 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Juhasz, S., Elbakry, A., Mathes, A. & Lobrich, M. ATRX Promotes DNA Repair Synthesis and Sister Chromatid Exchange during Homologous Recombination. Mol. Cell71, 11–24.e7 (2018). [DOI] [PubMed] [Google Scholar]
  • 25.Garbarino, J., Eckroate, J., Sundaram, R. K., Jensen, R. B. & Bindra, R. S. Loss of ATRX confers DNA repair defects and PARP inhibitor sensitivity. Transl. Oncol.14, 101147 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Maximo, V., Melo, M., Sobrinho-Simões, M., Soares, P. & Da Cruz Paula, A. Genomic profiling of lymph node and distant metastases from papillary and poorly differentiated thyroid carcinomas. Endocrine86, 505–509 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Maximo, V. et al. Genomic profiling of primary and metastatic thyroid cancers. Endocr. Relat. Cancer31, e230144 (2024). [DOI] [PubMed] [Google Scholar]
  • 28.Smith, E. S. et al. Endometrial Cancers in BRCA1 or BRCA2 Germline Mutation Carriers: Assessment of Homologous Recombination DNA Repair Defects. JCO Precis. Oncol.3, PO.19.00103 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Carter, S. L. et al. Absolute quantification of somatic DNA alterations in human cancer. Nat. Biotechnol.30, 413–21 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Kim, S. H. et al. Identification of recurrent FHL2-GLI2 oncogenic fusion in sclerosing stromal tumors of the ovary. Nat. Commun.11, 44 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Barthel, F. P. et al. Systematic analysis of telomere length and somatic alterations in 31 cancer types. Nat. Genet.49, 349–357 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Rosenthal, R., McGranahan, N., Herrero, J., Taylor, B. S. & Swanton, C. DeconstructSigs: delineating mutational processes in single tumors distinguishes DNA repair deficiencies and patterns of carcinoma evolution. Genome Biol.17, 31 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Ashley, C. W. et al. Analysis of mutational signatures in primary and metastatic endometrial cancer reveals distinct patterns of DNA repair defects and shifts during tumor progression. Gynecol. Oncol.152, 11–19 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Reiter, J. G. et al. Reconstructing metastatic seeding patterns of human cancers. Nat. Commun.8, 14114 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Moukarzel, L. A. et al. Clonal relationship and directionality of progression of synchronous endometrial and ovarian carcinomas in patients with DNA mismatch repair-deficiency associated syndromes. Mod. Pathol.34, 994–1007 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Pareja, F. et al. Whole-exome sequencing analysis of the progression from non-low-grade ductal carcinoma in situ to invasive ductal carcinoma. Clin. Cancer Res.26, 3682–3693 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Schwarz, R. F. et al. Phylogenetic quantification of intra-tumour heterogeneity. PLoS Comput. Biol.10, e1003535 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Nik-Zainal, S. et al. Landscape of somatic mutations in 560 breast cancer whole-genome sequences. Nature534, 47–54 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Huang da, W., Sherman, B. T. & Lempicki, R. A. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat. Protoc.4, 44–57 (2009). [DOI] [PubMed] [Google Scholar]
  • 40.Sanchez-Vega, F. et al. Oncogenic signaling pathways in the Cancer Genome Atlas. Cell173, 321–337.e10 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Leiserson, M. D., Whu, H. T., Vandin, F. & Raphael, B. J. CoMEt: a statistical approach to identify combinations of mutually exclusive alterations in cancer. Genome Biol.16, 160 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Rizvi, A. A. et al. gwasurvivr: an R package for genome-wide survival analysis. Bioinformatics35, 1968–1970 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary information (4.2MB, pdf)

Data Availability Statement

All clinical and molecular data was retrieved from the cBioPortal database ([https://www.cbioportal.org/)](https://www.cbioportal.org/)).

Articles from NPJ Precision Oncology are provided here courtesy of Nature Publishing Group

RESOURCES