Genomic landscape of endometrial polyps.

The study was approved by the Ethics Review Board of the Hospital District of Helsinki and Uusimaa, Helsinki, Finland. The samples were collected with a written informed consent signed by the patients. The study material consisted of 23 fresh frozen endometrial polyps and corresponding formalin-fixed paraffin-embedded (FFPE) samples from 23 patients. Blood samples were available from 18 of the patients. The samples were collected during 2016–2019 at Jorvi, Helsinki University Hospital (HUH). The sample set included 18 nonfunctional and 5 functional polyps that were collected from 11 post- and 12 premenopausal women with a mean age of 52 (ranging from 31 to 78). A series of 54 archival FFPE endometrial polyps served as a validation set for targeted UBE2A screening. The pseudonymous FFPE samples were obtained from the Department of Pathology, HUH, with the authorization of the Ethics Review Board. The histopathology of all the samples was re-evaluated by a pathologist trained in gynecological pathology (RB, AP) at the Department of Pathology, HUH, Helsinki, Finland, following the histological criteria described in Blaustein’s Pathology of Female Genital Tract [ 23 ]. DNA was extracted from fresh frozen polyps using the GeneJet Genomic DNA Purification Kit (Thermo Scientific, Waltham, MA, USA), from FFPE samples using the phenol-chloroform method, and from blood using the Nucleospin DNA Blood XL kit (Macherey-Nigel, Düren, Germany). DNA samples from 22 polyps were prepared for whole-exome sequencing (WES) with the KAPA Hyper Prep kit (Roche NimbleGen, Madison, WI, USA), and the exomes were captured with the SeqCap EZ MedExome Kit (Roche NimbleGen). Sequencing was performed at the Institute for Molecular Medicine Finland (FIMM, Helsinki, Finland) on the Illumina HiSeq 2500 system with 101 bp paired-end reads with an average coverage of 170×. Whole-genome sequencing (WGS) of 23 polyps and 18 corresponding blood samples was conducted at the BGI Genomics (Copenhagen, Denmark), on the BGISEQ-500 platform using 150 bp paired-end reads with an average coverage of 81× for polyp and 38× for normal blood samples (Supplementary Table S10). WES and WGS raw read quality was evaluated with FastQC, and the data were preprocessed following Genome Analysis Toolkit (GATK) 4 best practices [ 24 , 25 ]. In brief, the data were trimmed with Trimmomatic v0.39 and aligned against the Genome Reference Consortium Human Build 38 genome using Burrows-Wheeler Aligner BWA-MEM [ 26 , 27 ]. Duplicate reads were removed using MarkDuplicates, and base quality score recalibration was performed using BaseRecalibrator and ApplyBQSR [ 25 ]. WES and WGS data (BAM files) were merged after the removal of duplicate reads to increase the sensitivity of detecting somatic mutations with low-allelic fraction. The quality of the merged data was evaluated with AlfredQC [ 28 ]. Mutect2 was used to perform somatic variant calling [ 25 ]. Paired variant calling was performed for polyps with a matching normal tissue (blood) sample, and non-paired variant calling was performed for polyps without a matching normal. Variants from all samples were filtered against a panel of normals generated from 48 exomes and 28 genomes available in-house and a panel of normals generated from the 1000 Genomes Project samples [ 29 ]. The in-house panel of normals includes 48 exomes (15 blood samples, 1 fallopian tube sample, and 32 myometrium samples including 20 samples from online archives [ 30 , 31 ]) and 28 genomes (18 blood samples and 10 myometrium samples). Polyps without a corresponding normal tissue sample were filtered against any variants present in the Genome Aggregation Database (gnomAD v2.0.1 and v3) [ 32 ]. Polyps with a corresponding normal were filtered against variants present in more than three individuals in gnomAD [ 32 ]. Variants with a sequencing depth of less than 20 reads were filtered out. Variants were annotated using Ensembl Variant Effect Predictor (VEP) and OncoKB annotator [ 33 , 34 ]. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis was performed with recurrently mutated genes using the WEB-based GEne SeT AnaLysis Toolkit (WebGestalt) [ 35 , 36 ]. Data visualization was done with Integrative Genomics Viewer (IGV) 2.5.0 and Maftools 2.10.0 [ 37 , 38 ]. OncodriveCLUST was used to identify genes with a significant bias towards mutation clustering within the protein sequence [ 39 ]. Combined Annotation-Dependent Depletion (CADD), Rare Exome Variant Ensemble Learner (REVEL), and AlphaMissense were utilized to predict the effects of missense mutations affecting UBE2A [ 40 – 42 ]. PhyloP100 values from the UCSC genome browser were obtained to investigate the predicted consequence of the in-frame mutation c.16_18del, p.(Arg6del) in UBE2A [ 43 , 44 ]. Splice-AI was applied to predict the impact of HMGA1 c.268C > G, p.(Leu90Val) on splicing [ 45 ]. DynaMut was used to study the potential effect of UBE2A c.16C > T, p.(Arg6Trp) mutation on protein stability by using PDB Accession 6CYO [ 46 , 47 ]. cBioPortal database and Leiden Open Variation Database were used to identify additional samples with mutations affecting the same region [ 48 – 50 ]. SigProfiler and MutationalPatterns 3.4.1 were used to characterize Single Base Substitution (SBS) signatures [ 51 , 52 ]. SigProfiler analysis was conducted with default parameters. Signature refitting with MutationalPatterns was done with strict mode and 1000 bootstraps. Tumor mutation burden and tumor heterogeneity analyses were performed using Maftools 2.10.0 [ 38 ]. Somatic copy number alterations were called using CNVkit with default parameters against a pooled normal consisting of the 18 corresponding blood samples [ 53 ]. Somatic chromosomal rearrangements were called using Delly with the parameters -q 0 -u 0 [ 54 ]. Samples were run against 18 corresponding blood samples and an in-house dataset of 11 FFPE normal tissue samples. Calls were filtered using Delly Filter with the parameters -a 0.05 -p. The calls were filtered further, requiring a quality score of at least 300 and at least one discordant read pair. Deletions, inversions, and duplications less than 1000 bps were filtered out. Filtered calls located 1000 bps upstream or downstream of the high confidence calls were returned using bedtools window. All calls were manually evaluated using IGV [ 37 ]. BEDtools merge was used to detect recurrent breakpoints located a maximum of 200,000 bps from each other [ 55 ]. BEDtools closest was used to detect Ensembl genes closest to the recurrently rearranged regions. Data visualization was done with IGV 2.5.0 and RCircos v1.2.1 [ 37 , 56 ]. For each sample with an HMGA1 and/or an HMGA2 rearrangement, at least one relevant rearrangement was validated by Sanger sequencing. All UBE2A codon 6 mutations observed by WES/WGS were validated, and an additional set of 54 FFPE-derived endometrial polyp samples were screened for UBE2A codon 6 mutation by Sanger sequencing. PCR was performed using DreamTaq DNA polymerase (Thermo Scientific). Sequencing was performed with the Applied Biosystems (ABI) 3730 DNA Sequencer at FIMM. Chromatograms were analyzed and visually inspected using FinchTV 1.4.0 [ 57 ]. Primer sequences are available upon request. Total RNA was extracted and purified from the FFPE polyp and endometrium samples using the RNeasy® FFPE Kit (QIAGEN, Hilden, Germany) and the deparaffinization solution (QIAGEN) according to the manufacturer’s protocol. The concentration and purity of the extracted RNA were analyzed using the LabChip GX Touch HT RNA Assay Reagent Kit (PerkinElmer, Waltham, MA, USA) and the Qubit RNA BR kit (Thermo Fisher Scientific, Waltham, MA, USA). Genomic DNA contamination was measured using the Qubit DNA BR kit (Thermo Fisher Scientific). DNase treatment was performed to eliminate genomic DNA fragments. 3′RNA sequencing was performed with 25 endometrial polyps (23 lesions with WGS data and two additional polyps with a UBE2A codon 6 mutation) and 6 normal endometrium tissue samples (Supplementary Table S10). Sequencing was performed as previously described [ 58 ]. In brief, dual-indexed mRNA libraries were created using the Lexogen Gmbh’s QuantSeq 3′mRNA-Seq Library Prep Kit FWD according to the manufacturer’s instructions. Sequencing was performed at FIMM using the NovaSeq 6000 System (Illumina) with a minimum target coverage of 15 M reads for each library and a read length of 2 × 101 base pairs. The Bluebee® Genomics analysis platform’s Integrated Data Analysis Pipeline version 2.3.1 FWD UMI (Lexogen Gmbh) was used to preprocess the FASTQ data. The reads were trimmed using BBDuk, aligned to the Genome Reference Consortium human construct 38 (GRCh38) reference genome by STAR, and counted using HTSeq [ 59 , 60 ]. Two technical replicates’ read counts were combined. DESeq2 was used to normalize the raw read counts [ 61 ]. Principal component analysis (PCA) and pairwise differential expression analysis were performed with DESeq2 on the Chipster platform [ 61 , 62 ]. Variance stabilizing transformation was done for the raw read counts on the Chipster platform, and Euclidean hierarchical clustering was performed with ComplexHeatmap version 2.14.0 using the top 1% of genes with the most variable expression [ 61 – 63 ]. Immunohistochemical stainings were performed on 5 µm FFPE tissue sections using an anti-HMGA1 antibody (1:200; HPA068442, Sigma-Aldrich Co. LLC, St. Louis, MO, USA) and an anti-HMGA2 antibody (1:2000; 59170AP, Biocheck Inc., South San Francisco, CA, USA). Heat-induced antigen retrieval was carried out in a microwave oven using citrate buffer, and endogenous peroxidase blocking was followed by overnight primary antibody incubation at 4 °C. Stainings were detected with the BrightVision system (Immunologic, Duiven, Netherlands) and visualized by the DAB Quanto system (Thermo Fisher Scientific). Each staining set included a positive and negative control sample with a previously determined HMGA1/2 status [ 64 ], as well as a normal endometrium control. Staining of stromal and glandular tissue compartments was analyzed and scored by a pathologist (AP) as follows: strong and diffuse (++), moderate and heterogenous (+), weak ((+)), or not detected (−). Strong stromal staining was considered overexpression. HMGA1 break-apart FISH was performed on nine FFPE samples harboring HMGA1 rearrangements and three FFPE samples that were wild-type based on WGS. Break-apart FISH was performed on 4 µm FFPE tissue sections using break-apart FISH probes (PanPath, Budel, the Netherlands) for HMGA1 (C822K.2030.05, C822K.2030.10). Break-apart FISH was performed according to the probe manufacturer’s protocol.

We first performed WES, which revealed a high number of low-allelic fraction mutations (VAF < 0.1). To increase the coverage and enhance the sensitivity, we generated WGS data and integrated it with the WES data. Analysis on the merged data revealed a relatively high tumor mutation burden (TMB) with an average of 2149 mutations per sample (range 290–5463) and an average of 25 nonsynonymous mutations in the coding regions (range 2–64). We found no significant differences in the mutational spectrum or TMB between functional and nonfunctional polyps ( p  = 0.06 for all mutations, p  = 0.08 for nonsynonymous mutations in the coding genome, Student’s t -test). Mutational signature analysis with SigProfiler revealed that the most prominent signatures in endometrial polyps were SBS1, SBS5, and SBS40 (Fig.  1 a). MutationalPatterns corroborated these findings and identified additional signatures SBS8 and SBS89. The mutational signatures were largely similar in both functional and nonfunctional polyps. Signature SBS1 is characterized by C > T mutations, and SBS8 carries a pattern of C > A and T > A mutations, while SBS5, SBS40, and SBS89 are relatively featureless signatures. Fig. 1 Spectrum of somatic mutations in endometrial polyps. a Mutational signature analysis using SigProfiler identified SBS1, SBS5, and SBS40 as the most prominent mutational signatures in endometrial polyps (left). MutationalPatterns confirmed these signatures and additionally detected signatures SBS8 and SBS89 (right). The size of the circles represents the relative contribution of each signature. b Analysis of somatic mutations revealed the presence of at least three mutations in seven genes, including the well-known cancer genes KRAS , PIK3CA , PIK3R1 , and PTEN . c Among the most frequently mutated genes, UBE2A exhibited the highest average variant allelic fraction (VAF) Spectrum of somatic mutations in endometrial polyps. a Mutational signature analysis using SigProfiler identified SBS1, SBS5, and SBS40 as the most prominent mutational signatures in endometrial polyps (left). MutationalPatterns confirmed these signatures and additionally detected signatures SBS8 and SBS89 (right). The size of the circles represents the relative contribution of each signature. b Analysis of somatic mutations revealed the presence of at least three mutations in seven genes, including the well-known cancer genes KRAS , PIK3CA , PIK3R1 , and PTEN . c Among the most frequently mutated genes, UBE2A exhibited the highest average variant allelic fraction (VAF) Somatic variant calling revealed 22 genes that were recurrently mutated in endometrial polyps, with most mutations having a low-allelic fraction (Supplementary Table S1). OncoKB annotator revealed oncogenic and likely oncogenic mutations in 14 cancer driver genes, including well-established hotspot mutations in KRAS , PIK3CA , ERBB2 , PIK3R1 , PPP2R1A , PTEN , and FBXW7 (Supplementary Table S2). The most frequently mutated genes were KRAS , ARHGAP35 , PIK3CA , PIK3R1 , PTEN , and UBE2A , each showing at least three mutations (Fig.  1 b, Supplementary Table S3). KRAS was the most frequently mutated gene, with 6 out of 23 samples (26%) harboring a mutation. One sample displayed two KRAS hotspot mutations, c.35G > T, p.(Gly12Val) and c.38G > A, p.(Gly13Val), which were present at low-allelic fraction and on different reads (Supplementary Fig. S1, Supplementary Table S3). We identified possible biallelic mutations in ARHGAP35 and PIEZO2 —a combination of a missense mutation and a loss-of-function mutation in both genes. Pathway enrichment analysis using recurrently mutated genes revealed endometrial cancer as the most significantly mutated KEGG pathway, with an FDR-adjusted p value of 2.824 × 10 −5 (Supplementary Fig. S2). The most frequently mutated genes harbored very low-allelic fraction mutations (Fig.  1 c). The only exception was ubiquitin-conjugating enzyme E2 A ( UBE2A ), which harbored three mutations with an average allelic fraction of 0.28 (range 0.1–0.38). Tumor heterogeneity analysis demonstrated that in all three samples, the UBE2A mutations were present in a more clonal population than the smallest clone (Supplementary Fig. S3). Two samples harbored the same missense mutation c.16C > T, p.(Arg6Trp), and the third an in-frame deletion c.16_18del, p.(Arg6del) (Table  1 ). None of the mutations was present in the Genome Aggregation Database (gnomAD), and all mutations were confirmed somatic by direct Sanger sequencing (Supplementary Fig. S4A). Targeted screening of an independent validation set of 54 FFPE polyp samples revealed two additional polyps with the c.16C > T, p.(Arg6Trp) mutation (Supplementary Fig. S4B). Table 1 Somatic and germline mutations affecting codons 6–8 of UBE2A Sample ID HGVSc HGVSp Condition Prediction score Reference Endometrial polyps  1086_1_S1 c.16C > T p.(Arg6Trp) Endometrial polyp CADD: 24.4 This study  1126_1_S1 c.16C > T p.(Arg6Trp) Endometrial polyp CADD: 24.4 This study  1852_1_S1 c.16C > T p.(Arg6Trp) Endometrial polyp CADD: 24.4 This study  1857_1_S1 c.16C > T p.(Arg6Trp) Endometrial polyp CADD: 24.4 This study  1096_1_S1 c.16_18del p.(Arg6del) Endometrial polyp phyloP100: 7.9 This study Cancers  CRUK0028-R1 c.17G > T p.(Arg6Leu) Non-small cell lung cancer CADD: 24.5 [ 65 ]  TCGA-EI-6510-01 c.17G > A p.(Arg6Gln) Colorectal cancer CADD: 23.2 [ 66 ]  C3N-00386 c.16C > T p.(Arg6Trp) Endometrial cancer CADD: 24.4 [ 67 ]  GARV_0661 c.23G > A p.(Arg8His) Pancreatic cancer CADD: 25.2 [ 68 ]  TCGA-CK-6746-01 c.23G > A p.(Arg8His) Colorectal cancer CADD: 25.2 [ 66 ] Germline  25644381-FamL67 c.19C > T p.(Arg7Trp) Intellectual disability CADD: 27.7 [ 69 ] Somatic and germline mutations affecting codons 6–8 of UBE2A The UBE2A mutations identified in this study were located in a highly conserved region consisting of three codons (6–8) coding for three consecutive arginines (Fig.  2 a-b). The mutations were significantly clustered (OncodriveCLUST score of 0.83, q  = 1.51 × 10 −5 ), suggesting a potential gain-of-function effect. The missense mutation c.16C > T, p.(Arg6Trp) was predicted to be pathogenic with a CADD value of 24, REVEL score of 0.56, and AlphaMissense score of 0.96. The 3 bp deletion c.16_18del, p.(Arg6del) has a phyloP100 conservation score of 7.9, suggesting that this position does not tolerate variation. Fig. 2 Location of UBE2A mutations identified in endometrial polyps and the potential impact of the p.(Arg6Trp) mutation on the structure of UBE2A. a Genome sequencing of 23 endometrial polyps revealed an UBE2A mutation in three lesions and targeted Sanger sequencing of 54 endometrial polyps found two additional mutations affecting codon 6. b The mutations occurred in a highly conserved region (codons 6–8) of UBE2A that encodes for three consecutive arginine residues. c The location of the c.16C > T, p.(Arg6Trp) mutation in the predicted 3D structure of UBE2A . The mutation has a predicted stabilizing effect on protein folding, with a ΔΔG prediction outcome of 1.650 kcal/mol. Cartoon visualization depicts predicted interactions in the wild-type and mutant versions of the UBE2A protein Location of UBE2A mutations identified in endometrial polyps and the potential impact of the p.(Arg6Trp) mutation on the structure of UBE2A. a Genome sequencing of 23 endometrial polyps revealed an UBE2A mutation in three lesions and targeted Sanger sequencing of 54 endometrial polyps found two additional mutations affecting codon 6. b The mutations occurred in a highly conserved region (codons 6–8) of UBE2A that encodes for three consecutive arginine residues. c The location of the c.16C > T, p.(Arg6Trp) mutation in the predicted 3D structure of UBE2A . The mutation has a predicted stabilizing effect on protein folding, with a ΔΔG prediction outcome of 1.650 kcal/mol. Cartoon visualization depicts predicted interactions in the wild-type and mutant versions of the UBE2A protein We then explored the putative effect of p.Arg6Trp on the protein’s stability and function with DynaMut. The ΔΔG prediction outcome was 1.650 kcal/mol, indicating a stabilizing effect of the mutation (Fig.  2 c). To explore whether similar mutations occur in other tumor types, we searched the cBioPortal database and identified four mutations in five samples, including one endometrial cancer carrying the missense mutation c.16C > T, p.(Arg6Trp) (Table  1 ). Additionally, analysis of germline mutations in the Leiden Open Variation Database unveiled a pathogenic mutation within the same region, c.19C > T, p.(Arg7Trp). The mutation was detected in two family members affected with an X-linked mental retardation syndrome (Variant #0000408054). All identified mutations were predicted to be deleterious (Table  1 ). Somatic copy number analysis revealed very few copy number changes. We identified no signs of high-level amplifications, homozygous deletions, or aneuploidy (Supplementary Fig. S5). Conversely, we identified simple and complex chromosomal rearrangements involving two to four chromosomes in 17 of 23 (74%) endometrial polyps (Supplementary Fig. S6). These rearrangements rarely resulted in large-scale losses or gains of adjacent regions. In total, we identified 420 chromosomal breakpoints (Supplementary Table S4), with recurrent breakpoint regions in 5q13.1, 6p21.31 ( HMGA1 ), 6q21, 7p15.2, 10p15.1, 10q22.3, 12q14.3 ( HMGA2 ), 14q24.1, Xq22.1, and Xq22.2 (Supplementary Table S5). Six polyps (6/23; 26%) displayed no chromosomal rearrangements, and two of these polyps harbored a UBE2A hotspot mutation. In addition to chromosomal rearrangements, structural variant analysis identified gene retrocopy insertions of MATR3 in one endometrial polyp. The MATR3 retrocopy was inserted into at least three positions in the genome (chromosomes 1p.21.1, 3q26.33, and 12q24.12, Supplementary Fig. S7). It remains unclear whether the insertions were somatic or in the germline due to the unavailability of matching normal tissue from this patient. Analysis of recurrent breakpoint regions revealed simple and complex chromosomal rearrangements involving 6p21.31 ( HMGA1 ) and/or 12q14.3 ( HMGA2 ) in 17 of 23 (74%) polyps (Fig.  3 a-–c, Table  2 ). One sample displayed both an HMGA1 and an HMGA2 rearrangement (Fig.  3 c). We identified an HMGA1 rearrangement in 14 polyps, of which 12 showed breakpoints upstream of HMGA1 (Supplementary Fig. S8A). One polyp harbored breakpoints downstream of HMGA1 , whereas two had breakpoints in the 3′UTR of HMGA1 (Supplementary Fig. S8A). In one sample with a 3′UTR breakpoint, we identified a somatic missense mutation c.268C > G, p.(Leu90Val) in the penultimate exon of HMGA1 . The variant was weakly predicted to result in loss of the donor site by splice-AI with a delta score of 0.25. The variant had a relatively high-allelic fraction (VAF = 0.34) and was present in a more clonal cell population compared to the low-allelic fraction mutations (Supplementary Fig. S9). We identified four HMGA2 rearrangements, two of which displayed intragenic breakpoints (Supplementary Fig. S8B). Occasionally, polyps with HMGA1 or HMGA2 rearrangements displayed other independent chromosomal rearrangements not involving any recurrently rearranged regions (Supplementary Table S5). The average VAF for HMGA1 alterations was 0.24 (range 0.06–0.43), and for HMGA2 alterations 0.24 (range 0.14–0.45). Fig. 3 Most endometrial polyps show chromosomal alterations involving HMGA1 or HMGA2 . a One third of endometrial polyps (9/23; 39%) displayed a balanced translocation involving HMGA1 or HMGA2 . b Five samples (5/23; 22%) displayed a 3-way translocation involving HMGA1 or HMGA2 . c Six samples (6/23; 26%) with an HMGA1/2 rearrangement harbored additional independent chromosomal alterations. These rearrangements were occasionally more complex, involving several clustered breakpoint regions and more than two chromosomes. One sample (1092_1_S1) displayed both an HMGA1 and an HMGA2 rearrangement. Interchromosomal rearrangements are shown in blue and intrachromosomal inversions in red. In the inner circle, deletions called as structural variants are marked with red and duplications called as structural variants in blue. The schematic figures are not in scale Table 2 Chromosomal rearrangements affecting HMGA1 and HMGA2 in endometrial polyps Sample HMGA rearrangement Candidate partner Type of change Involved chromosomes 1097_1_S1 HMGA1 upstream ENSR00001272277 (7p15.2) Balanced translocation 6, 7 1124_1_S1 HMGA1 upstream ENSR00001272277 (7p15.2) Balanced translocation 6, 7 1098_1_S1 HMGA2 upstream ENSR00001272277 (7p15.2) Complex translocation 7, 12 1100_1_S1 HMGA1 upstream LRMDA (10q22.3) intragenic Balanced translocation 6, 10 1088_1_S1 HMGA2 intragenic LRMDA (10q22.3) intragenic Balanced translocation 10, 12 1094_1_S1 HMGA1 upstream TRAF3IP2 (6q21) downstream Balanced translocation and inversion 5, 6 1093_1_S1 HMGA1 downstream TRAF3IP2 (6q21) downstream Complex intrachromosomal 5, 6 1085_1_S1 HMGA1 upstream RAD51B (14q24.1) downstream 3-way balanced translocation 6, 14, 15 1092_1_S1 HMGA2 upstream RAD51B (14q24.1) downstream Balanced translocation 12, 14 1092_1_S1 HMGA1 upstream COL6A3 (2q37.3) intragenic Complex translocation 2, 6, 13, X 1090_1_S1 HMGA1 upstream LINC00629 (Xq26.3) intragenic Balanced translocation 6, X 1120_1_S1 HMGA1 upstream PIK3R1 (5q13.1) 3′UTR Balanced translocation 5, 6 1095_1_S1 HMGA1 upstream No clear candidate (2q35) Balanced translocation 2, 6 1126_1_S1 HMGA1 upstream No clear candidate (18p11.22) 3-way complex translocation 6, 11, 18 1123_1_S1 HMGA1 upstream and 3′UTR NEB (2q23.3) intragenic and PGR (11q22.1) intragenic 3-way complex translocation 2, 6, 11 1121_1_S1 HMGA1 3′UTR ZFP41 (8q24.3) intragenic 3-way complex translocation 2, 6, 8 1516_1_S1 HMGA2 upstream and intragenic ARL15 (5q11.2) intragenic and SETBP1 (18q12.3) intragenic 3-way complex translocation 5, 2, 18 1087_1_S1 HMGA1 upstream No clear candidate (6q21) Complex intrachromosomal 6 1086_1_S1 – – – – 1091_1_S1 – – – – 1096_1_S1 – – – – 1101_1_S1 – – – – 1103_1_S1 – – – – 1119_1_S1 – – – – Most endometrial polyps show chromosomal alterations involving HMGA1 or HMGA2 . a One third of endometrial polyps (9/23; 39%) displayed a balanced translocation involving HMGA1 or HMGA2 . b Five samples (5/23; 22%) displayed a 3-way translocation involving HMGA1 or HMGA2 . c Six samples (6/23; 26%) with an HMGA1/2 rearrangement harbored additional independent chromosomal alterations. These rearrangements were occasionally more complex, involving several clustered breakpoint regions and more than two chromosomes. One sample (1092_1_S1) displayed both an HMGA1 and an HMGA2 rearrangement. Interchromosomal rearrangements are shown in blue and intrachromosomal inversions in red. In the inner circle, deletions called as structural variants are marked with red and duplications called as structural variants in blue. The schematic figures are not in scale Chromosomal rearrangements affecting HMGA1 and HMGA2 in endometrial polyps We identified four regions that were recurrently targeted by HMGA1 and/or HMGA2 rearrangements. A small region within 7p15.2 (size 12 kb) was the most frequently observed rearrangement partner, with two HMGA1 rearrangements and one HMGA2 rearrangement involving this region (Fig.  4 a, Supplementary Fig. S10A). The 7p15.2 region contained no nearby genes, with the closest regulatory element being an enhancer (ENSR00001272277). We also identified LRMDA as a translocation partner for HMGA1 in one sample and for HMGA2 in another (Fig.  4 b, Supplementary Fig. S10B). The breakpoints were located upstream of HMGA1 and intragenic of HMGA2 and LRMDA . RAD51B was identified as a translocation partner for HMGA1 and HMGA2 in one sample each, with breakpoints located downstream of RAD51B (Fig.  4 c, Supplementary Fig. S10C). Two samples displayed an upstream region of TRAF3IP2 as a recurrent rearrangement partner (Fig.  4 d, Supplementary Fig. S10D). Fig. 4 Chromosome region 7p15.2, LRMDA , RAD51B , and TRAF3IP2 were identified as recurrent rearrangement partners for HMGA1 and HMGA2 . a We detected three endometrial polyps with 7p15.2 as a rearrangement partner for HMGA1 or HMGA2 . We found no candidate partner gene in this region and the closest regulatory element was an enhancer (ENSR00001272277). b We identified LRMDA as a translocation partner for HMGA1 or HMGA2 in two samples. c RAD51B was a translocation partner for HMGA1 and HMGA2 in two samples. d Two samples displayed an upstream region of TRAF3IP2 as a rearrangement partner for HMGA1 . The schematic figures are not in scale. In the circos plots, interchromosomal translocations are shown in blue and intrachromosomal inversions are in red. In the inner circle, deletions called as structural variants are marked with red and duplications called as structural variants in blue Chromosome region 7p15.2, LRMDA , RAD51B , and TRAF3IP2 were identified as recurrent rearrangement partners for HMGA1 and HMGA2 . a We detected three endometrial polyps with 7p15.2 as a rearrangement partner for HMGA1 or HMGA2 . We found no candidate partner gene in this region and the closest regulatory element was an enhancer (ENSR00001272277). b We identified LRMDA as a translocation partner for HMGA1 or HMGA2 in two samples. c RAD51B was a translocation partner for HMGA1 and HMGA2 in two samples. d Two samples displayed an upstream region of TRAF3IP2 as a rearrangement partner for HMGA1 . The schematic figures are not in scale. In the circos plots, interchromosomal translocations are shown in blue and intrachromosomal inversions are in red. In the inner circle, deletions called as structural variants are marked with red and duplications called as structural variants in blue We performed 3′RNA sequencing on 25 endometrial polyps and 6 normal endometrium tissue samples. This set consisted of the 23 polyps that were included in the WES/WGS analysis and the two additional samples that were found to carry a UBE2A mutation in the subsequent targeted screening. We confirmed upregulation of HMGA1 in polyps harboring an HMGA1 rearrangement when compared to normal endometrium ( p  = 0.02, Student’s t -test, log2FC = 1.06). Some polyps harboring an HMGA2 rearrangement displayed increased expression, but the result did not reach statistical significance ( p  = 0.37, Student’s t -test, log2FC = 1.90) (Fig.  5 a). HMGA1 and HMGA2 have both been proposed to regulate the expression of PLAG1 , and we identified significant upregulation of PLAG1 in endometrial polyps harboring either an HMGA1 ( p  = 3.91 × 10 −6 , Student’s t -test, log2FC = 3.22) or an HMGA2 ( p  = 0.02, Student’s t -test, log2FC = 3.38) rearrangement when compared to endometrium (Fig.  5 a). Endometrial polyps did not form distinct clusters based on the presence of alterations affecting HMGA1 , HMGA2 , or UBE2A neither according to principal component analysis using all genes nor according to hierarchical clustering analysis using genes with the most variable expression (Fig.  5 b, c). Fig. 5 Endometrial polyps with HMGA1 and HMGA2 rearrangements show upregulation of HMGA1 and/or PLAG1 . a Expression profiling confirmed upregulation of HMGA1 in endometrial polyps carrying an HMGA1 rearrangement (left) and upregulation of PLAG1 in polyps with an HMGA1 and/or HMGA2 rearrangement (right). Expression of HMGA2 was not significantly different between the sample groups (middle). p values are indicated in the figure. b According to the principal component analysis, endometrial polyps did not cluster based on the presence of HMGA1 and HMGA2 rearrangements or UBE2A mutations. c Similarly, hierarchical clustering of the genes with the most variable expression displayed no clear clustering based on HMGA1 , HMGA2 , or UBE2A mutational status Endometrial polyps with HMGA1 and HMGA2 rearrangements show upregulation of HMGA1 and/or PLAG1 . a Expression profiling confirmed upregulation of HMGA1 in endometrial polyps carrying an HMGA1 rearrangement (left) and upregulation of PLAG1 in polyps with an HMGA1 and/or HMGA2 rearrangement (right). Expression of HMGA2 was not significantly different between the sample groups (middle). p values are indicated in the figure. b According to the principal component analysis, endometrial polyps did not cluster based on the presence of HMGA1 and HMGA2 rearrangements or UBE2A mutations. c Similarly, hierarchical clustering of the genes with the most variable expression displayed no clear clustering based on HMGA1 , HMGA2 , or UBE2A mutational status We compared 25 endometrial polyps against 6 normal endometrium tissue samples, revealing 488 differentially expressed genes ( q   1, Supplementary Table S6). We also compared 17 endometrial polyps harboring an HMGA1 or HMGA2 rearrangement against the endometrium tissue samples and the remaining polyps, revealing 57 uniquely expressed genes ( q   1, Supplementary Table S7). ZMAT3 was the most significantly upregulated gene in these polyps. PLAG1 was the sixth most significantly upregulated gene, and C19orf38 , another HMGA subtype biomarker, was the 13th most significantly upregulated gene (Table  3 ). We also compared 5 endometrial polyps with a UBE2A mutation against the endometrium tissue samples and the remaining polyps, which revealed 39 uniquely expressed genes ( q   1, Supplementary Table S8). Table 3 The most significantly differentially expressed genes in endometrial polyps Polyps vs endometrium Unique for HMGA1 and HMGA2 Unique for UBE2A Gene log2FC q value Gene log2FC q value Gene log2FC q value RNU6-70P 2.09 7.97E−14 ZMAT3 1.65 6.98E−13 CFD 2.72 1.86E−07 SCARNA3 2.79 1.22E−13 GDF15 2.13 2.51E−06 HAND2 1.45 5.98E−07 IGFBP6 3.37 1.22E−13 HSPB6 1.53 9.71E−06 MBNL1-AS1 2.01 1.48E−06 SLIT3 2.15 4.04E−13 FABP4 3.02 9.71E−06 BLOC1S2 −1.2 3.14E−06 RNU6-865P 2.1 4.04E−13 SYNE3 1.23 9.71E−06 ZNF154 2.48 8.07E−06 NFIX 1.53 1.33E−12 PLAG1 1.82 9.71E−06 ISOC1 −1.52 4.05E−05 LMOD1 2.84 1.33E−12 TRHDE-AS1 2.47 9.71E−06 CXCL14 3.34 1.74E−04 RNU6-826P 1.82 3.86E−11 ABCD2 1.72 9.48E−05 MCTP2 1.1 3.33E−04 PLA2G7 −1.49 6.84E−11 GRIA3 2.02 1.74E−04 SERPINH1 −1.2 3.60E−04 RNU6-1147P 2.37 8.57E−11 MIR497HG −1.76 1.90E−04 TMOD1 1.46 5.39E−04 DRGX −4.11 1.61E−10 FXYD6 1.13 2.06E−04 TTYH3 −1.22 5.57E−04 PI15 3.16 1.65E−10 CDKN1A 1.24 3.49E−04 LPIN3 1.11 6.35E−04 SNORA31 1.79 5.58E−10 C19orf38 2.03 3.49E−04 MMP11 −2.31 6.35E−04 RNU6-618P 2.08 6.03E−10 CHRNA3 −2.17 3.91E−04 TCEAL7 −1.62 7.39E−04 ARHGAP4 1.39 7.88E−10 MYOCD 1.66 4.45E−04 PPFIBP2 1.12 8.30E−04 PDE5A 1.28 3.34E−09 MAMSTR 1.28 6.21E−04 ATP1B1 −1.12 8.30E−04 CYS1 3.65 3.34E−09 TRHDE 2.26 6.32E−04 PYGM 1.48 1.18E−03 RPS7 −1.86 3.34E−09 TAF7L 1.14 6.38E−04 RIMS4 −1.89 1.36E−03 SNED1 2.06 4.04E−09 PTGES 1.49 9.36E−04 ENSG00000232855 1.68 1.42E−03 SNORA75 2.34 4.04E−09 BCHE 1.81 9.99E−04 MEX3A −1.05 1.57E−03 The most significantly differentially expressed genes in endometrial polyps To evaluate the impact of HMGA1/2 alterations at the protein level, we conducted HMGA1 and HMGA2 immunohistochemistry (IHC) on the 23 polyp and 3 normal endometrium samples (Fig.  6 , Supplementary Table S9). We observed positive staining for HMGA1 and HMGA2 in epithelial cells of the glandular tissue compartment across both the polyp and normal endometrium samples. Therefore, we only considered strong stromal staining as indicative of overexpression. After excluding two samples that failed in all IHC analyses, we observed strong stromal staining for HMGA1 in 9 out of 12 polyp samples with an HMGA1 alteration and strong stromal staining for HMGA2 in all three polyps carrying only an HMGA2 rearrangement. The one sample with both an HMGA1 and an HMGA2 rearrangement exhibited positive staining for HMGA1 but negative staining for HMGA2. While most samples showed positive concordance between the genomic HMGA1/2 alterations and corresponding strong stromal protein expression, some displayed discrepant results. One sample showed strong staining for HMGA1, and one sample showed strong staining for HMGA2 even without corresponding genomic changes. Finally, three samples with an HMGA1 alteration showed only moderate staining for HMGA1. Overall, Fisher’s exact test revealed a significant association between the IHC and WGS results (HMGA1, p  = 0.008, HMGA2, p  = 0.012). Fig. 6 Endometrial polyps with an HMGA1 or an HMGA2 rearrangement show corresponding stromal protein overexpression. Immunohistochemistry revealed HMGA1 or HMGA2 protein overexpression in the stromal cells of polyps with corresponding genomic HMGA1 or HMGA2 alterations. The epithelial cells of the glandular tissue compartment displayed strong positive staining in both the polyp and normal endometrium samples. Magnification 100× Endometrial polyps with an HMGA1 or an HMGA2 rearrangement show corresponding stromal protein overexpression. Immunohistochemistry revealed HMGA1 or HMGA2 protein overexpression in the stromal cells of polyps with corresponding genomic HMGA1 or HMGA2 alterations. The epithelial cells of the glandular tissue compartment displayed strong positive staining in both the polyp and normal endometrium samples. Magnification 100× Break-apart FISH was performed to validate the localization of the observed HMGA1 rearrangements, but the FFPE sample quality was suboptimal and did not allow for formal analysis with appropriate cell counting and scoring (Supplementary Fig. S11).

Endometrial polyps are localized overgrowths of endometrial glands, altered stroma, and blood vessels that protrude into the uterine cavity [ 1 ]. These tumor-like lesions occur in both pre- and postmenopausal women [ 1 , 2 ]. Approximately 10% of adult women have endometrial polyps, although the exact prevalence remains unknown as the majority of polyps are incidentally detected during a routine gynecological examination [ 2 , 3 ]. Endometrial polyps can cause abnormal uterine bleeding and infertility, but are most often asymptomatic [ 1 , 4 ]. Histologically, endometrial polyps can be divided into functional and nonfunctional polyps based on the activity of endometrial glands; functional polyps exhibit active endometrial glands that undergo cyclic histological changes, whereas nonfunctional polyps have inactive glands that are unresponsive to circulating hormones [ 1 , 5 ]. Primary treatment for symptomatic endometrial polyps is hysteroscopic polypectomy, a minimally invasive procedure involving the removal of polyps using a hysteroscope [ 6 ]. Endometrial polyps are non-cancerous tumor-like lesions, although a subset of them may undergo malignant transformation [ 7 , 8 ]. Hyperplasia can be seen in 3 to 30% of the lesions, with up to 3% of polyps eventually transforming into endometrial cancer [ 8 – 10 ]. Polyps that result in abnormal bleeding in postmenopausal women have a higher propensity for malignant progression [ 7 , 8 ]. Tamoxifen treatment has also been associated with an increased risk for malignant transformation [ 11 ]. The etiology and pathogenesis of endometrial polyps remain elusive. Several risk factors have been associated with polyp development, including genetic changes, overexpression of endometrial aromatase, unbalanced activity of estrogen and progestin, inhibition of apoptosis, inflammation, tamoxifen treatment, and obesity [ 5 , 11 – 15 ]. Estrogen has been shown to promote the development and growth of endometrial polyps [ 5 , 12 ]. Cytogenetic studies conducted during the 1990s revealed chromosomal aberrations in approximately half of endometrial polyps [ 13 ]. These studies divided endometrial polyps into four subtypes based on cytogenetic features: polyps with a normal karyotype, polyps with a rearrangement in 6p21-22, polyps with a rearrangement in 12q13-15, and polyps with a 7q22 deletion. Similar chromosomal alterations have been reported in other benign tumors, like uterine leiomyomas and lipomas [ 16 , 17 ]. The rearrangements in 6p21-22 and 12q13-15 result in upregulation of high mobility group AT-hook 1 ( HMGA1 ) and high mobility group AT-hook 2 ( HMGA2 ), respectively [ 18 ]. Recently, the role of HMGA1/2 alterations as drivers in polyp etiology has been questioned, as such rearrangements were not identified in a study that used targeted sequencing [ 19 – 21 ]. Instead of chromosomal alterations, targeted sequencing efforts have reported low-allelic fraction mutations in known cancer genes like KRAS , PIK3CA , and FBXW [ 19 , 22 ]. Despite the high prevalence of endometrial polyps, very little research has been directed towards defining the molecular mechanisms underlying these lesions, and a thorough understanding of their genomic background is still lacking. Here, we performed genomic sequencing together with global gene expression profiling to elucidate the molecular landscape of endometrial polyps.

In this study, we conducted a comprehensive analysis on the genomic landscape of endometrial polyps, providing insights into their molecular characteristics. We identified chromosome-level alterations in 74% of the lesions, which is a much higher frequency than previously reported [ 13 , 19 ]. Rearrangements were predominantly balanced and did not result in significant losses or gains of large chromosomal regions. Importantly, we did not identify any high-level amplifications, homozygous deletions, or aneuploidy, suggesting a milder pattern of genomic instability in endometrial polyps compared to most cancers. Interestingly, a subset of chromosomal rearrangements was more complex in nature, involving multiple breakpoints and chromosomes. Contrary to previous studies, we did not observe aberrations in the 7q22 region in our dataset, challenging the previous notion that 7q22 aberrations represent a common molecular subtype of endometrial polyps [ 13 ]. The most common alterations in endometrial polyps were chromosomal rearrangements affecting HMGA1 and HMGA2 . These changes were observed in 74% of the lesions; indeed, all samples with chromosomal alterations harbored an HMGA1 or an HMGA2 rearrangement as a likely driver event. The most frequent partner for HMGA1/2 alterations was a highly confined region on chromosome 7p15.2. Previously, t(7;17)(p15;q21) and t(6;7)(p21;p15) translocations have been reported as the most frequent chromosomal abnormalities in endometrial stromal tumors [ 70 – 72 ]. These rearrangements have been proposed to generate fusion genes involving JAZF1 at 7p15 [ 70 ]. The breakpoints in our dataset were confined to a very specific region 2 million bps upstream of JAZF1 indicating JAZF1 as an unlikely target gene in endometrial polyps. This region contained no obvious candidate genes, suggesting that the rearrangements target a tissue-specific enhancer. We identified a putative enhancer, ENSR00001272277, as the closest regulatory element, suggesting enhancer hijacking as a mechanism to upregulate the involved genes [ 73 ]. Other chromosomal regions recurrently targeted by HMGA1/2 translocations included previously reported partner genes LRMDA , RAD51B , and TRAF3IP2 [ 74 – 76 ]. RAD51B and TRAF3IP2 have been reported as translocation partners for HMGA2 in uterine leiomyomas and the 10q22 region, which contains LRMDA , has been reported to be recurrently rearranged in endometrial polyps and uterine leiomyomas [ 13 , 75 – 77 ]. While intragenic breakpoints were occasionally observed in HMGA2 , no intragenic breakpoints were identified in HMGA1 , suggesting that fusion genes may not be the primary mechanism through which these genes contribute to tumorigenesis. This likely explains why HMGA1/2 alterations were not detected in a recent study, which utilized targeted methods that are more optimal for detecting fusion genes[ 19 ]. Immunohistochemistry revealed that most polyps carrying HMGA1/2 alterations exhibit strong positive staining of the corresponding protein in stromal cells. This indicates that the chromosomal alterations are located in the stromal tissue compartment, where they result in corresponding protein overexpression. This finding is consistent with early 1990s reports, where clonal 6p21 rearrangements were confined to the mesenchymal components of endometrial polyps [ 78 , 79 ]. Epithelial cells in the glandular tissue compartment displayed positive staining across the polyp samples as well as in normal endometrium control samples, suggesting that HMGA1/2 expression is characteristic for endometrial glands. This prevalent expression in epithelial cells may have masked signals originating from stromal cells, thus potentially affecting the clustering analysis of the 3′RNAseq data. Limitations arising from the polyp’s cellular composition and technical challenges in detecting structural variants from short-read data might account for the undetected rearrangement in two samples exhibiting strong protein expression without corresponding genomic alterations. Furthermore, stained FFPE samples often exhibited cautery artifact, which is caused by heat changes during polypectomy. This may have resulted in reduced staining sensitivity in the three polyp samples with an HMGA1 alteration but only moderate protein expression. Overall, the genomic alterations and IHC results demonstrated strong concordance, with statistically significant correlations for both HMGA1 and HMGA2. These findings also support the interpretation of polyps as neoplasms of stromal origin [ 19 – 21 ]. HMGA1 and HMGA2 modulate the transcription of various genes by altering the accessibility of regulatory factors to DNA [ 80 ]. HMGA1 and HMGA2 are generally downregulated in adult differentiated tissues, and their upregulation has been associated with tumorigenesis [ 80 ]. We have previously shown that HMGA1 and HMGA2 alterations are associated with upregulation of PLAG1 and that uterine leiomyomas harboring rearrangements involving either HMGA1 , HMGA2 , or PLAG1 form a distinct molecular subtype characterized by similar gene expression profiles [ 64 , 81 ]. Although we could not see evidence of transcriptional clustering by mutation status, polyps with HMGA1 rearrangements displayed upregulation of HMGA1 , and polyps with either HMGA1 or HMGA2 rearrangements displayed upregulation of PLAG1 . In addition, we observed C19orf38 as one of the most significantly upregulated genes in HMGA1/2 aberrant endometrial polyps, a gene we recently highlighted as a biomarker for leiomyomas with HMGA1 , HMGA2 , or PLAG1 alterations [ 64 ]. Taken together, all these observations support the previous cytogenetic findings and suggest that a distinct molecular subtype of endometrial polyps is driven by chromosomal rearrangements affecting HMGA1 and HMGA2 . It remains to be seen whether also PLAG1 alterations are found in endometrial polyps. Similar HMGA1 and HMGA2 alterations that were observed in endometrial polyps have been identified in other tumors, including uterine leiomyomas [ 75 ]. Uterine leiomyomas are common benign smooth muscle tumors arising from the uterine wall and affecting up to 70% of women [ 82 , 83 ]. Most uterine leiomyomas (70–80%) are driven by MED12 mutations, while a smaller proportion (10–20%) harbor HMGA2 aberrations [ 84 ]. In contrast to uterine leiomyomas, which rarely display HMGA1 rearrangements, we identified HMGA1 rearrangements as the most common driver alteration in endometrial polyps. ZMAT3 was the most significantly overexpressed gene in HMGA1/2 aberrant endometrial polyps. Interestingly, we previously reported ZMAT3 as the most significantly upregulated gene in leiomyomas irrespective of the underlying driver mutation [ 58 ]. ZMAT3 is an RNA-binding protein that is induced by the tumor suppressor protein p53 [ 85 ]. It has the ability to bind to numerous mRNA precursors leading to an impact on their splicing patterns [ 85 ]. Overall, these findings emphasize the molecular similarities between endometrial polyps and uterine leiomyomas, two common female lesions that originate in the same organ but from different tissue types. This indicates intriguing opportunities for targeted treatment options that may be effective in both tumor types. In addition, HMGA2 is frequently upregulated in ovarian carcinomas, where it has been proposed as a promising target for gene silencing therapy [ 86 ]. In addition to chromosomal alterations, we identified small, low-allelic fraction mutations in half of the polyps. This finding supports a recent study that used a more targeted approach and identified similar mutations in endometrial polyps [ 19 ]. Mutations included well-characterized hotspot mutations in the known cancer genes like KRAS , PIK3CA , PIK3R1 , and PTEN. Such mutations are characteristically present in other lesions arising from the endometrium, including endometriosis [ 87 ]. Interestingly, similar mutations have also been observed in normal healthy endometrial glands [ 88 ]. It has been suggested that while the actual tumor development requires additional somatic alterations, these low-allelic fraction mutations represent early stages of neoplastic changes [ 88 ]. Indeed, mutations in these same genes are present in endometrioid ovarian and endometrial cancers, where they serve as actionable targets for treatment [ 89 , 90 ]. The low-allelic fraction nature of mutations in endometrial polyps suggests a confined localization within a specific tissue compartment, most likely the glandular compartment, similarly to healthy endometrium. However, one study reported the presence of KRAS and NRAS mutations in both stromal and glandular components [ 22 ]. KRAS mutations in endometrial polyps have also been associated with tamoxifen [ 91 ], but according to the patient data, none of our patients had been using tamoxifen. While our results confirm the presence of low-allelic fraction cancer gene mutations in endometrial polyps, the role of these mutations in polyp development and their potential for promoting malignant transformation remain obscure. Interestingly, we identified recurrent and highly specific mutations in UBE2A . All these UBE2A mutations were located in a highly conserved region in exon 1 (codons 6–8 encoding for three consecutive arginines) and displayed a higher allelic fraction than most other recurrently mutated genes. It is worth emphasizing that all in silico tools used predicted that these mutations are likely pathogenic. Although these tools do not offer definitive proof of pathogenicity, the cumulative evidence suggests that the mutations likely disrupt the normal function of the encoded protein. UBE2A encodes for a ubiquitin-conjugating enzyme that is required for ubiquitination of mitochondrial proteins upon depolarization [ 92 ]. Previously, germline mutations in UBE2A have been shown to cause an X-linked intellectual disability syndrome [ 47 , 92 ]. Interestingly, one germline mutation causing the syndrome p.(Arg7Trp) affects the same stretch of three arginines that was somatically mutated in the polyps. This germline mutation was observed in a family with two carrier sisters and their three affected sons [ 92 ]. Both sisters showed complete skewing of X-inactivation in blood and no symptoms. The mutation resulted in reduced UBE2A expression and in mitochondrial dysfunction in mouse- and patient-derived cells. Moreover, in a yeast model, mutations in codon 6 of Rad6 (an ortholog of human UBE2A and UBE2B ) lead to impaired ubiquitination of a conserved lysine in H2B, which broadly impacts the chromatin structure, transcription, and a histone modification cascade on active genes [ 93 ]. Results from these functional experiments indicate that the localized UBE2A mutations observed in endometrial polyps are functionally relevant and support the notion that these mutations represent early driver alterations in a subset of polyps. Recurrent somatic UBE2A mutations that lead to decreased UBE2A activity have been reported in progressed myeloid leukemia, but these mutations are not located in the same region as the mutations observed in polyps [ 94 ]. Overall, further studies are needed to elucidate the functional effect of the observed UBE2A mutations in endometrial polyp etiology. Endometrial polyps displayed a mutational burden similar to that observed in cancers. Mutational burden and patterns were similar in functional and nonfunctional polyps. We identified five previously characterized mutational signatures (SBS1, SBS5, SBS8, SBS40, and SBS89) in the polyps, indicating the involvement of similar mutational processes as seen in cancer development. Signatures SBS1, SBS5, and SBS40 typically correlate with patients’ age, and these signatures can also be seen in normal healthy endometrium [ 88 , 95 ]. Interestingly, SBS8 is common in cancer, but rare in benign tissues [ 96 ]. The etiology of SBS8 is not well understood, but it may arise due to late replication errors during cancer progression [ 96 ]. SBS89 is not as well established and validated as the other extracted signatures and has an unknown etiology.

Our results show that chromosomal rearrangements affecting HMGA1 or HMGA2 are the major cause underlying endometrial polyp development. We also present UBE2A as a novel candidate driver gene for human tumorigenesis . In addition, we show a high mutational burden of low-allelic fraction mutations in endometrial polyps, including well-known driver gene mutations frequently observed in endometrium-derived malignancies. While endometrial polyps are benign lesions, the presence of cancer-associated mutations supports the notion that some polyps can act as precursor lesions for endometrial cancer. Taken together, these findings bring out new knowledge on the molecular background of endometrial polyps and provide new tools for the development of targeted non-invasive treatment options.

Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2.