Genomic profiling of dedifferentiated endometrial carcinomas arising in the background of high‐grade carcinoma: a targeted next‐generation sequencing study

Our understanding of dedifferentiated endometrial carcinoma (DEC), a rare and aggressive malignancy, mainly reflects undifferentiated carcinomas (UC) arising in the setting of low‐grade endometrial cancer (DEC‐LG). However, cases of UC arising in the setting of high‐grade EC (DEC‐HG) have been noted in the literature. Our knowledge of the genomics of DEC‐HG is limited. To characterise the molecular landscape of DEC‐HC, targeted genomic sequencing and immunohistochemical analysis was carried out on seven DEC‐HG and four DEC‐LG.


Introduction
Dedifferentiated endometrial carcinoma (DEC) is an aggressive subtype of endometrial cancer, which up until the 2020 WHO classification of female gynaecologic tumours was strictly defined as composed of undifferentiated carcinoma (UC) and a low-grade endometrioid carcinoma (FIGO grade 1 or 2) components (DEC-LG). [1][2][3][4] However, cases of UC arising in the setting of high-grade endometrial cancer have been noted in the literature. [5][6][7][8]10 To further understand this phenomenon, a systematic characterisation of the morphologic, immunohistochemical, and clinicopathologic characteristics of UC in the context of DEC-HG was recently reported by our group. 11 Our previously published findings indicate that DEC is not restricted to tumours with low-grade endometrioid morphology and supports the revised diagnostic criteria for DEC. Moreover, the lack of previous formal recognition by the WHO underscores the need for better awareness of this phenomenon.
The undifferentiated component of DEC is characterised by the absence of cell cohesion and glandular differentiation, which imparts a patternless solid growth of tumour cells microscopically. P53 is often wildtype (~70% of the cases 6,7 ), DNA mismatch repair (MMR) protein expression is lost in about 45-70% of cases, 6,7,12,13 and switch/sucrose nonfermenting chromatin-remodelling complex (SWI/SNF) ATPase subunit members (SMARCA4/BRG1 or SMARCB1/INI1 or ARID1A and ARID1B) exhibit mutually exclusive loss in approximately two-thirds of cases. 12,14 To date, the genomic landscape of DEC was predominantly investigated in the context of undifferentiated carcinoma arising in the setting of low-grade endometrial cancer. Such genetic analyses have demonstrated that the undifferentiated component of DEC is clonally related to the concurrent differentiated component; thus, the two are often regarded as a single entity in molecular studies. 4,9 Accumulating data demonstrates that undifferentiated/dedifferentiated carcinomas may develop through any of the four molecular pathways described by The Cancer Genome Atlas for endometrial cancer. 6,13 Several reports suggest that undifferentiated/dedifferentiated carcinomas exhibit frequent microsatellite instability and mutations in genes commonly altered in endometrial cancer such as PTEN, PIK3CA, and CTNNB1. 7,9,[11][12][13][14] Further, the SWI/SNF has been investigated in the context of undifferentiated/dedifferentiated carcinoma due to its established role in tumour progression and dedifferentiation. [14][15][16][17] However, whether the molecular landscape of genetic alterations in DEC-LG and DEC-HG are distinct is currently unknown.
In the current study, we investigated the genomic landscape of seven DEC-HG to gain insight into their molecular pathogenesis. It was then compared to four DEC-LG cases from our institutional cohort and DEC-LG cases from the literature. In addition, undifferentiated and concurrent differentiated endometrial cancer components for each DEC case were subjected to comparative genetic analysis to interrogate their clonal relationship.

Materials and Methods
This study was approved by the Research Ethics Board at Sunnybrook Health Sciences Centre (SUN-2197). A subset of 11 DECs ( Table 1) that underwent hysterectomy from 2008 to 2019 were identified from our larger cohort of 18 DEC cases through a retrospective search of the laboratory information system, Sunquest CoPath, as previously described. 11 Briefly, original haematoxylin and eosin (H&E) slides for each case were retrieved and reviewed by three gynaecologic pathologists (J.M., C.P.H., A.B.). The presence of an undifferentiated component was confirmed using the criteria proposed by Silva et al., 1 including the presence of sheets of noncohesive cells without glandular formation or obvious nests and trabecular architecture. The grade of the admixed differentiated carcinoma component was assessed and classified as low (n = 4, endometrioid FIGO grade 1 or 2) or high (n = 7, endometrioid FIGO grade 3, serous, mixed, or with ambiguous features/unclassifiable, Figure 1). Only cases with a full consensus regarding histotype were included. Cases of pure undifferentiated carcinoma, carcinosarcoma, and poorly differentiated tumours showing specific lines of differentiation, such as solid growth of serous carcinoma, FIGO grade 3 endometrioid carcinoma or neuroendocrine carcinoma were excluded.
The following pathologic and demographic features were recorded for each case: tumour size, stage, the presence of lymphovascular space invasion, age at presentation, presenting symptoms, date of surgery, date of recurrence, date of death, date of last followup, and status at follow-up. INI1, and SMARCA4/BRG1 proteins, as previously described (antibodies listed in Table S1). 11 Briefly, staining for p53 was graded as wildtype (patchy) or mutation-pattern, including overexpression, (≥80% of tumour cells show strong nuclear staining), null staining (complete absence of expression), and cytoplasmic staining patterns. Any nuclear staining for MMR, SMARCB1/INI1, or SMARCA4/BRG1 proteins was considered positive. ARID1A and ARID1B immunohistochemistry (IHC) was not available for this study.
Genomic DNA was extracted from representative formalin-fixed, paraffin-embedded (FFPE) tissue blocks as previously described. 18 Distinctive tumour areas representative of the undifferentiated and background endometrial carcinoma component were macrodissected separately. Targeted exome sequencing was performed with the Ion S5XL next-generation sequencing system with the AmpliSeq Comprehensive Cancer Panel (CCP, ThermoFisher, Waltham, MA, USA). The panel surveys all exon regions of 409 oncogenes and tumour suppressor genes (excluding ARID1B and POLE). The amplicon library was constructed using the Ion Ampliseq Library Kit Plus and Ion Xpress barcode adapters according to the manufacturer's instructions. Sequencing template preparation was done using Ion Chef with Ion 540 Chef Kits. Sequencing was performed for 500 flows on an Ion S5XL Sequencer with the Ion 540 chip. For detection of somatic mutations, the sequencing was performed to at least a 400-fold mean depth per sample.

S E Q U E N C I N G D A T A A N A L Y S I S
Variants were called by the Ion Torrent platformspecific pipeline software, Torrent Suite v. 5.10.1 (ThermoFisher), and Ion Reporter v. 5.16 (Thermo-Fisher). Subsequently, a custom filter pipeline was used to filter out known common single-nucleotide polymorphisms (SNPs) in the UCSC database and SNPs with global minor allele frequency >0.01 in the 1000 Genomes and 5000 Exomes databases. Filtered variants were then annotated based on the COSMIC, dbSNP, SIFT, PolyPhen, and ClinVar databases. Alignments were visually verified with the Integrative Genomics Viewer v. 2.8.9.

G E N O M I C L A N D S C A P E O F U N D I F F E R E N T I A T E D A N D D I F F E R E N T I A T E D C O M P O N E N T S O F D E D I F F E R E N T I A T E D E N D O M E T R I A L C A R C I N O M A S
Next-generation sequencing analyses of DEC identified 398 and 320 mutations in the undifferentiated and differentiated components of DEC-HG, respectively. In the undifferentiated and differentiated components of DEC-LG, 187 and 202 mutations were detected, respectively. The percent of mutations shared between concurrent undifferentiated and differentiated components ranged from 22.2% to 64.1% in DEC-HG and 39.6% to 64.0% in DEC-LG. There were no statistically significant differences in mutation frequencies between undifferentiated and differentiated components of DEC-HG or DEC-LG (P > 0.05).
SMARCA4 mutations were detected in 57% DEC-HG and in 25% DEC-LG (undifferentiated component, Figure 2). No genomic alterations in SMARCB1/INI1 were observed. ARID1A mutations were identified in 86% DEC-HG and 100% DEC-LG ( Figure 2). In DEC-HG, ARID1A mutations were found in both undifferentiated and differentiated components in three cases, solely in the differentiated component of two cases, and in the undifferentiated component of one case. In DEC-LG, mutations were solely found in the differentiated component of one case and in both components of three cases. TP53 mutations were detected in 57% DEC-HG and 50% DEC-LG ( Figure 2).
SMARCA4 and ARID1A mutations were mutually exclusive in DEC-LG but not DEC-HG (Figure 2). SMARCA4 and TP53 alternations were mutually exclusive in DEC-LG and showed overlap in one DEC-HG (case #7). ARID1A and TP53 mutations were not mutually exclusive in either DEC-HG or DEC-LG.

C O R R E L A T I O N B E T W E E N I M M U N O H I S T O C H E M I S T R Y A N D G E N O M I C P R O F I L E S O F D E D I F F E R E N T I A T E D E N D O M E T R I A L C A R C I N O M A S
In our cohort, concomitant SMARCA4/BRG1 inactivating gene mutations and protein expression loss was identified solely in the undifferentiated component of 3/7 (43%) DEC-HG and 1/4 (25%) DEC-LG (Table 2). Additionally, one DEC-HG case (case #5) of concurrent frameshift and missense mutations in SMARCA4 in both undifferentiated and differentiated components was without associated protein loss.
In our cohort, only missense MLH1 variants of uncertain significance were detected, which did not correlate with protein loss ( Table 2). We also identified a pathogenic nonsense MLH2 mutation (case #7) and a missense MSH6 variant of uncertain significance (case #2); neither correlated with expression loss. No PMS2 mutations were detected. Based on targeted gene panel analysis, there was no significant difference in the mean number of mutations between MMR-deficient versus MMR-proficient cases (average 91 versus 117 mutations/cases, P = 0.425).
Two DEC-HG cases exhibited both p53 mutationpattern IHC and TP53 missense mutations (Table 2). An additional DEC-HG case (case #2) had a p53 mutation-pattern IHC in the differentiated component and concomitant TP53 missense mutations in undifferentiated and differentiated components. Other TP53 missense mutations were detected in one DEC-HG (case #3) and two DEC-LG (cases #10, 11), all p53 IHC wildtype.

G E N E T I C A L T E R A T I O N S I N U N D I F F E R E N T I A T E D C A R C I N O M A S A S S O C I A T E D W I T H H I G H -G R A D E V E R S U S L O W -G R A D E E N D O M E T R I A L C A R C I N O M A S
Next, we compared the genomic profiles of an undifferentiated component arising in high-grade versus low-grade endometrial cancer to determine the mechanisms underlying their dedifferentiation. Among the seven undifferentiated components of DEC-HG, 398 mutations (SNV/MNV/INDELS) in 206 genes were identified ( Figure 3). Additionally, 187 mutations in 133 genes were identified in four undifferentiated components from DEC-LG. Of these, 89 genes were shared between DEC-HG and DEC-LG, although only 14 mutated loci were identical.

M U T A T I O N A L P R O F I L E S O F C O N C U R R E N T U N D I F F E R E N T I A T E D A N D D I F F E R E N T I A T E D C O M P O N E N T S O F D E D I F F E R E N T I A T E D E N D O M E T R I A L
Mutational profiles of the undifferentiated and concurrent differentiated components of DEC-HG and DEC-LG were compared to define their clonal relationship. In DEC-HG, a total of 158 identical mutations were detected in both the undifferentiated and concurrent differentiated components of the same tumour. Each pair shared 14-27 (average 23 mutations/case) identical genetic variants. Additionally, 11-115 (average 34 mutations/case) mutations were identified only in the undifferentiated component, and 7-47 (average 23 mutations/case) were exclusively found in the high-grade differentiated component of all pairs analysed.
In DEC-LG, a total of 103 identical mutations were detected in both the undifferentiated and concurrent differentiated components of the same tumour. Each pair shared 22-30 (average 26 mutations/case) identical mutations. Additionally, 12-29 (average 21 mutations/case) mutations were identified only in the undifferentiated component, and 13-46 (average 25 mutations/case) were exclusively found in the low-grade differentiated component. These findings demonstrate a clonal relationship between the undifferentiated and concurrent differentiated components of DEC with a period of common ancestry represented by the identical mutations followed by independent cell division of each component that gave rise to heterogeneous subclones with unique mutations. Focused analysis of genes commonly altered in endometrioid endometrial cancer, the SWI/SNF genes, chromatin remodelling, and MMR genes (commonly documented in DEC) is summarised in Figure 2.

Discussion
In this targeted genomic analysis of 11 DECs, we provide insights into the genomic landscape of DEC-HG versus DEC-LG and their disease biology. Our analysis identified that overall mutational frequency and spectrum is relatively similar for DEC-HG and DEC-LG as well as their undifferentiated and differentiated components. The clonal mutations characteristic of endometrial carcinoma likely arise early in tumour formation, suggesting similar pathogenesis for DEC within the two groups and a similar origin of the undifferentiated component from the associated differentiated carcinoma. Conversely, the genomic aberrations exclusive to the different components within DEC-HG and DEC-LG likely occur late in tumour development and account for tumour heterogeneity. We found that ARID1A mutations and MMR IHC deficiency were common alterations in the undifferentiated and/or differentiated components, detected in 91% and 64% of our cohort, respectively. P53 mutation-pattern IHC was observed in 18% of cases, while TP53 missense mutations were detected in 55% of cases in the undifferentiated and/or differentiated components. The discordance between p53 IHC and TP53 genomic alterations results can be attributed to the fact that many of the TP53 mutations identified in our study are of unknown significance. Therefore, these mutations may not alter p53 function or do not lead to aberrant protein expression. Further, it has been shown that p53 IHC is not a perfect surrogate for all TP53 missense alterations, with discordance between IHC and mutation status reported in up to 25% of cases. 19,20 Similarly, many of the mutations in MMR genes identified in our study are of unknown significance. These missense and nonsense mutations may not interfere with protein function or may not alter the antigenic portion of the protein; thus, no loss of immunoreactivity is observed. 21 Additionally, loss of MLH1 expression is usually the result of gene promoter methylation rather than mutation. 22 The current findings add to a growing body of literature on DEC. A study by Rosa-Rosa et al. 6 of 10 DEC-LG and 1 DEC-HG found MMR deficiency in all cases and TP53 missense mutation with concurrent aberrant p53 expression in the undifferentiated component of one case. They also found that 36% of DEC had loss of ARID1A expression in the undifferentiated, differentiated, and/or both components of DEC. Further, in their series of 13 DEC-LG, Espinosa et al. 23 found 46% to be MMR deficient and 54% had aberrant p53 expression, in the undifferentiated or both components of DEC. Loss of ARID1A expression was found in 62%. Lastly, analysis of eight DEC-LG by Karnezis et al. 14 found MMR deficiency in 75% cases, a TP53 missense mutation in the undifferentiated component of 13% cases, and ARID1A mutations in 75%, in both components. None of the above studies investigated ARID1B. SMARCA4/BRG1 deficiency was identified solely in the undifferentiated component of 50% of cases. ARID1A mutations and SMARCA4/BRG1 protein loss were not mutually exclusive. However, the differentiated component of four DEC cases was not analysed in each study and the later studies did not include any DEC-HG cases.
Our targeted sequencing and subsequent IHC analysis revealed that the loss of SMARCA4/BRG1 expression is associated with histologic dedifferentiation in 43% of DEC-HG and 25% of DEC-LG. These findings further support previous studies suggesting that deficiency in the SWI/SNF complex, including the loss of SMARCA4, could represent the molecular mechanism underlying dedifferentiation in endometrial and other carcinomas. 14,24 It is important to note that SWI/ SNF deficiency is primarily found in the undifferentiated component of DEC, which may have important practical implications for pathologists when choosing the proper block for molecular testing if/when targeted therapy of SWI/SNF-deficient cancers is developed and incorporated into clinical practice. Surprisingly, we identified a missense SNV and two frameshift mutations in the differentiated components of one DEC-HG case, resulting in a truncated protein that was immunolabelled. This can be explained by the fact that the SMARCA4/BRG1 antibody binds to the N-terminal of the protein, which might have not been affected by these mutations. Thus, although SMARCA4/BRG1 IHC deficiency may be predictive of a SMARCA4 mutation, it is not a perfect surrogate for the direct assessment of mutation status.
Further evaluation of the mutational landscape of DEC identified some distinct pathogenic pathways within DEC. This is in agreement with previous findings regarding the molecular heterogeneity of this malignancy. 6,[12][13][14] In our study, SMARCA4/BRG1 protein loss and ARID1A mutations were mutually exclusive in DEC-LG, but not DEC-HG, suggesting potential functional redundancy in EC dedifferentiation. Aberrant SMARCA4/BRG1, MMR, and p53 protein expression in both undifferentiated and background differentiated carcinoma components of DEC-HG and DEC-LG were not mutually exclusive, indicating a possibility for a combined effect of cell growth/division, MMR, and SWI/SNF complex pathways in promoting dedifferentiation. However, our cohort is too small to draw conclusions from these observations. Notably, there was no loss of SMARCB1/INI1 in any cases in our cohort, likely due to the low frequency of this alteration, mutually exclusive to SMARCA4/BRG1 loss. 14 Our analysis is limited by the small sample size. An additional limitation of our results might lie in the retrospective design and use of material from a single institution. Further, lack of ARID1B protein expression data to investigate the concurrent ARID1A and ARID1B inactivation previously reported in DEC, as well as POLE mutations, which were previously found in a subset of DEC, was not available for our cohort. 13,15 Further genomic analysis of larger cohorts is needed to verify the generalisability of our findings.
To conclude, in this study we systematically describe the genomic profiles of the undifferentiated and differentiated components of DEC-HG and DEC-LG. The study documented that DEC arising in the background of either low-grade or high-grade endometrial cancer occur mainly in the setting of MMR deficiency and accumulation of common mutations characteristic of endometrioid endometrial carcinoma, such as ARID1A, PTEN, PIK3CA, and CTNNB1. Further, our data support the link between SWI/SNFdeficiency and development of DEC. Comparing DEC-HG and DEC-LG, we observed some similarities in the spectrum of mutations, although several genomic alterations were more frequent in the undifferentiated component of DEC-HG and others in the undifferentiated component of DEC-LG. Taken together, our results suggest that the molecular mechanisms of dedifferentiation are heterogeneous in DEC arising in the background of either low-grade or high-grade and support the inclusion of DEC-HG in the histologic definition of DEC. 10