{"paper_id":"9f764c5d-1f98-4e39-ad89-c59b5683e69c","body_text":"Endometriosis is a chronic condition defined by the presence of endometrial-like epithelial and stromal tissue outside the uterus and affects approximately 5–10% of reproductive-aged women ( Zondervan  et al. , 2018 ;  Tomassetti  et al. , 2021 ). Its clinical manifestation varies including dysmenorrhea, dyspareunia, dyschezia, and chronic pelvic pain ( Zondervan  et al. , 2018 ). There are three anatomical subtypes: superficial peritoneal endometriosis (SUP), ovarian endometrioma (OMA), and deep endometriosis (DE). The revised American Society for Reproductive Medicine (rASRM) categorizes endometriosis into four stages (I–IV; IV being most severe) based on anatomic subtype extent, location, and adhesions ( Canis  et al. , 1997 ).\nOur team and others have proposed that somatic genomic events/alterations, such as ‘cancer-driver’ mutations, may drive endometriosis severity and progression ( Vestergaard  et al. , 2011 ;  Govatati  et al. , 2014 ;  Anglesio  et al. , 2017 ;  Lac  et al. , 2019b ). Malignant transformation in ovarian endometriosis is rare (∼1% of cases), and somatic mutations have been found in endometriosis tissues adjacent to these endometriosis-associated ovarian cancers (clear cell and endometrioid subtypes) ( Wei  et al. , 2011 ;  Vercellini  et al. , 2014 ;  Anglesio  et al. , 2015 ). Nonetheless, somatic mutations like  KRAS ,  PIK3CA ,  CTNNB1 , and others have also been reported in endometriosis without cancer, confined to the endometriosis epithelium ( Vestergaard  et al. , 2011 ;  Fung  et al. , 2015 ;  Anglesio  et al. , 2017 ;  Yong  et al. , 2021 ). Their distribution spans all anatomic subtypes of endometriosis (SUP, DE, and OMA), and even iatrogenic endometriosis ( Lac  et al. , 2019b ;  Yong  et al. , 2021 ). These classic oncogenic mutations may influence the behavior of endometriosis which exhibits tumor-like behavior such as local invasion ( Anglesio  et al. , 2017 ;  Lac  et al. , 2019b ;  Yong  et al. , 2021 ).\nSomatic mutations involving tumor suppressors such as phosphatase and tensin homolog ( PTEN ) and AT-rich interaction domain 1A ( ARID1A ) genes have also been found in endometriosis without cancer; however, their clinical impact on endometriosis severity and progression remains unclear. PTEN is involved in several biological processes such as regulating genomic stability, cell survival, and metabolism ( Lee  et al. , 2018 ). It functions as a negative regulator of the phosphoinositide-3 kinase (PI3K) pathway ( Govatati  et al. , 2014 ;  Lee  et al. , 2018 ).  PTEN  mutations occur in 20% of ovarian endometrioid carcinomas, and PTEN protein loss is reported in endometriosis and normal endometrial tissue ( Govatati  et al. , 2014 ;  Yang  et al. , 2015 ). ARID1A, a member of the SWItch/sucrose non-fermentable (SWI/SNF) complexes, is pivotal for chromatin remodeling and DNA transcription ( Mittal and Roberts, 2020 ;  Fontana  et al. , 2023 ). SWI/SNF complexes are mutated in about 25% of human cancers, with  ARID1A  mutations occurring in about 30% of ovarian endometrioid carcinomas including up to half of ovarian clear cell carcinomas ( Jones  et al. , 2010 ;  Wiegand  et al. , 2010 ;  Xu and Tang, 2021 ;  Heinze  et al. , 2022 ). Several endometriosis studies have also reported on occasional heterogenous mutation/loss of ARID1A protein expression ( Samartzis  et al. , 2012 ;  Anglesio  et al. , 2017 ;  Yachida  et al. , 2020 ;  Praetorius  et al. , 2022 ).\nSomatic genomic events such as nonsense mutations, frameshift mutations, some missense mutations, deletions, and epigenetic silencing can result in loss of function and loss of expression of PTEN and ARID1A ( Perren  et al. , 1999 ;  Kang  et al. , 2002 ;  Lotan  et al. , 2011 ;  Djordjevic  et al. , 2012 ;  Samartzis  et al. , 2013 ;  Khalique  et al. , 2018 ;  Mittal and Roberts, 2020 ). Therefore, loss of expression in endometriosis epithelium on immunohistochemistry (IHC) serves as a proxy for these genomic events ( Khalique  et al. , 2018 ;  Köbel  et al. , 2018 ;  Yachida  et al. , 2020 ). In endometrial cancer, Wang et al. found that PTEN IHC had 75.4% sensitivity and 84.6% specificity for detecting PTEN mutations using deep sequencing of the entire coding region ( Wang  et al. , 2022 ). Similarly, Khalique et al. developed an ARID1A IHC scoring system that achieved 100% sensitivity and 97–100% specificity for  ARID1A  mutations in ovarian clear cell carcinomas, using a targeted capture panel with 99%  ARID1A  gene coverage ( Khalique  et al. , 2018 ). The effectiveness of IHC proxies for mutation status is further supported by several other key studies ( Perren  et al. , 1999 ;  Monte  et al. , 2010 ;  Lac  et al. , 2019a ;  Martins  et al. , 2020 ).\nUtilizing IHC, the reported loss of PTEN and ARID1A in endometriosis varies. PTEN loss has been observed in 14–18% of endometriosis cases of various subtypes ( Lac  et al. , 2019b ;  Praetorius  et al. , 2022 ); ARID1A loss has been observed in less than 4% of endometriosis cases and 8–19% of OMAs specifically ( Samartzis  et al. , 2012 ;  Xiao  et al. , 2012 ;  Chene  et al. , 2015 ;  Yong  et al. , 2021 ;  Praetorius  et al. , 2022 ). However, clinical phenotypes elicited by these somatic genomic events in endometriosis detected by IHC are still not well understood. Challenges include the fact that endometriosis epithelium forms a small fraction of endometriosis surgical samples, multiple subtypes and multiple lesions per subtype are frequently found in affected individuals, lesions may be oligoclonal, and there is complexity in clinical outcomes ( Yong  et al. , 2021 ). We recently reported that  KRAS  codon 12 hotspot mutations are associated with a greater anatomic burden of endometriosis disease ( Orr  et al. , 2023 ). However, studies evaluating the clinical associations and phenotypes related to loss of tumor suppressors, such as PTEN and ARID1A, in endometriosis are lacking.\nIn this prospective longitudinal cohort, we evaluated the clinical phenotype of PTEN and ARID1A somatic loss in endometriosis epithelium. We hypothesized that, like  KRAS  mutation, somatic ARID1A and PTEN loss would be associated with greater endometriosis anatomic disease burden (i.e. more severe subtypes and higher stage) at the index surgery. We also hypothesized that ARID1A and PTEN loss may be associated with reoperation risk after the index surgery, as the presence of residual microscopic endometriosis with PTEN or ARID1A loss after the index surgery may drive disease recurrence. We additionally integrated somatic PTEN and ARID1A loss with previous data on  KRAS  codon 12 mutations to assess the impact of multiple somatic genomic events in this cohort.\n\nThis was a prospective longitudinal study conducted at the BC Women’s Center for Pelvic Pain and Endometriosis, a tertiary referral center for endometriosis involving gynecologic surgeons fellowship trained in endometriosis surgery. Prospective biobanking of surgically excised endometriosis began in 2013 (ENDOONC study; REB H11-00536 and H14-03040, University of British Columbia) with standardized clinical data collected as part of a prospective registry, the Endometriosis Pelvic Pain Interdisciplinary Cohort (EPPIC) (Clinicaltrials.gov  NCT02911090 ; REB H16-00264, University of British Columbia) ( Yosef  et al. , 2016 ;  Allaire  et al. , 2018 ). For EPPIC, baseline/preoperative data is collected through patient-reported questionnaires and clinician-reported variables from physical examination and point-of-care ultrasound that are prospectively reported into a standardized form at the initial visit. At the time of surgery following the initial visit, surgical findings are also prospectively reported into a standardized form by surgeons at the time of surgery. Follow-up patient-reported questionnaires are sent annually after baseline.\nWe included participants who had an index surgery at our center between 2013 and 2017 and consented to both prospective biobanking and the EPPIC registry; the same cohort was previously used to interrogate somatic  KRAS  mutations ( Orr  et al. , 2023 ). Surgeries were divided into fertility-sparing conservative surgery (excision of endometriosis) or hysterectomy (with excision of endometriosis). The study flowchart with study exclusion criteria is displayed in  Fig. 1 . Tissue with suspected endometriosis was excised during surgery, and samples included only if sufficient endometriosis epithelium/stroma were confirmed by H&E staining (before PTEN and ARID1A IHC staining). Samples for IHC were selected using the same selective sampling strategy as our study on  KRAS  mutations ( Orr  et al. , 2023 ). For each participant: all DE samples were included, if present; unilateral or bilateral OMA samples were included, if present; and one SUP lesion was included, with a second SUP lesion from a separate location, if present ( Orr  et al. , 2023 ).\nStudy flow chart.  Included and excluded participants throughout the study.\nIHC staining for PTEN and ARID1A was carried out at University of Calgary, using formalin-fixed, paraffin-embedded tissue and primary antibody for PTEN (Cell Signaling, Danvers, MA, USA; PTEN-Clone 138G6), and ARID1A (Abcam, Cambridge, UK; ab182560-clone  EPR13501 ). IHC for PTEN (dilution: 1/60) and ARID1A (dilution: 1/2000) was completed using the DAKO Omnis protocol (H30-R10-30).\nWe used our previously published IHC protocols to assess PTEN or ARID1A loss ( Lac  et al. , 2019a ;  Praetorius  et al. , 2022 ). For PTEN, scoring was based on cytoplasmic staining in endometriosis epithelial cells, while ARID1A scoring was based on nuclear endometriosis epithelial staining. IHC scoring for both PTEN and ARID1A was completed using an ordinal scale with four categories: (i) retained expression or 0% loss among epithelial cells (i.e. completely retained); (ii) 1–10% loss; (iii) 11–49% loss; and (iv) 50–100% loss. Loss in at least 10 adjacent endometriosis epithelial cells was required to qualify as PTEN or ARID1A loss ( Praetorius  et al. , 2022 ). The endometriosis stromal cells and other surrounding cells were utilized as an internal control (normal expression). IHC was independently scored by two observers (A.F.L. and D.R.T.) before measuring agreement and inter-rater reliability. Where independent scores differed, both observers jointly reviewed to reach a consensus. IHC was scored from scanned slides using Aperio ImageScope v12.4.0.5043 software (Leica Microsystems Pty Ltd, Wetzlar, Germany).\nFor analyses in this study, each participant’s PTEN and ARID1A status was defined as: retained (0% loss in all samples from the participant) or loss (any percentage loss in at least one sample from the participant). PTEN and ARID1A status was also combined with available  KRAS  codon 12 mutation data for this cohort to identify participants with two or more somatic events.\nThe primary analyses examined the association between PTEN loss, ARID1A loss, and having two or more somatic events with (i) anatomic subtype and (ii) rASRM stage at the index surgery, as indicators of endometriosis disease burden. Anatomical subtypes in each participant were grouped as previously reported: (a) SUP only; (b) DE only or OMA only; and (c) mixed subtypes with two or more anatomic subtypes, as confirmed by clinical pathology reports ( Orr  et al. , 2023 ). Fisher’s exact test or chi-square test assessed these associations.\nSecondary analyses examined associations between PTEN loss, ARID1A loss, and having two or more somatic events with the following clinical variables: demographics, other anatomic findings at the index surgery, markers of surgical difficulty, and baseline pain severity variables. Mann–Whitney  U  test or independent samples  t -test were performed for continuous variables, and Fisher’s exact test for categorical variables.\nAdditionally, we assessed whether PTEN loss, ARID1A loss, and having two or more somatic events was associated with pain severity after the index surgery (at 1- to 2-year follow-up after baseline), by performing linear regression while controlling for baseline pain scores and surgery type (hysterectomy versus conservative-only surgery). Analyses for dysmenorrhea severity excluded those having a hysterectomy. For the 5- to 9-year follow-up period, reoperation-free survival was analyzed using Kaplan–Meier analysis and the log-rank test. To account for the impact of surgery type on reoperation risk, we also analyzed risk after excluding those with a prior hysterectomy or who underwent hysterectomy during the index surgery.\nInter-rater reliability of ordinal PTEN and ARID1A sample scores was evaluated using Cohen’s weighted kappa, with interpretation by Landis and Koch’s recommendations ( Landis and Koch, 1977 ). Statistical analyses were performed using R Statistical Software v4.2.1 (Boston, MA, USA). Statistical significance was set to <0.05. Cases with missing data were excluded without imputation.\n\nThis study included 126 participants total ( Fig. 1 ): 231 samples from 117 participants for PTEN and 231 samples from 120 participants for ARID1A were assessed by IHC ( Table 1 ). The mean age was 34.5 years, and most participants were White (74.6%; 94/126) and nulliparous (72.2%; 91/126) ( Table 1 ). Almost all index surgeries were performed laparoscopically (98.4%; 124/126) with most participants receiving fertility-sparing conservative surgery (73%; 92/126) rather than hysterectomy (27%; 34/126). At the index surgery, just over half of the cohort had Stage III/IV endometriosis (55.6%; 70/126), and participants affected by more than one anatomical subtype of endometriosis (mixed) were the most common group (53.1%; 67/126) compared to SUP only (29.4%; 37/126) or DE or OMA only (17.5%; 22/126).\nClinical description of study cohort.\nExcludes those not sexually active.\nIHC scoring inter-rater reliability was substantial for PTEN (kappa = 0.69; 95% CI: 0.617, 0.768) and ARID1A (kappa = 0.64; 95% CI: 0.51, 0.77) based on Landis and Koch’s recommendations ( Landis and Koch, 1977 ). Images of PTEN and ARID1A IHC are displayed in  Supplementary Fig. S1 . PTEN loss was prevalent in the cohort: 68.3% (86/126 participants) had PTEN loss, 24.6% (31/126) had PTEN retained, and in 7.1% (9/126) PTEN scoring could not be completed due to technical factors (e.g. no evaluable cells on the slide or failed control staining) ( Table 2 ). ARID1A loss was comparatively less frequent, with 24.6% (31/126 participants) characterized as ARID1A loss, 70.6% (89/126) had ARID1A retained, and in 4.8% (6/126) non-scorable (due to the same technical factors as for PTEN) ( Table 2 ).\nCohort description by somatic events.\nMissing due to technical factors (e.g. no evaluable cells on the slide or failed control staining).\nFrom previously published study completed at our center ( Orr  et al. , 2023 ).\nOnly reported for participants with data on all three somatic events, i.e. PTEN, ARID1A, and  KRAS  mutation. The variable is coded as ‘yes’ for participants who had at least two of these three somatic events in our study, and ‘no’ if they had one or none.\nAmong individuals with data for PTEN, ARID1A and the previously reported  KRAS  mutation, the most common co-event (per participant assessment) was having PTEN loss and  KRAS  mutation (with ARID1A retained) (28.3%; 32/113) ( Fig. 2 ). Interestingly, <1% (1/113) of participants exhibited ARID1A loss as the sole somatic event, indicating that ARID1A loss more commonly coincided with other somatic events, especially when compared to the proportion of PTEN loss alone (21.2%; 24/113 participants) ( Table 2 ;  Fig. 2 ). The frequencies for each of the combinations of somatic events ( Table 2 ;  Fig. 2 ) were not significantly different than expected by chance.\nSomatic phenotype of the cohort.  Description and distribution of the study’s somatic phenotype based on participants with data for  PTEN ,  ARID1A , and  KRAS  (n = 113). The most frequent somatic phenotype per participant assessment was PTEN loss and  KRAS  mutation (with ARID1A retained) (28.3%; 32/113), followed by PTEN loss only (21.2%; 24/113), and the detection of all three somatic events ( PTEN  +  KRAS  +  ARID1A ) (15.9%; 18/113). Additionally, 16.8% of this cohort had none of the somatic events, and 0.9% of participants had ARID1A as the only event detected.\nPTEN loss was significantly more frequent among participants with mixed anatomic subtypes (85.1%; 57/67) and DE only or OMA only (72.7%; 16/23), in comparison to SUP only (46.4%; 13/28) ( P  < 0.001) ( Table 3 ). PTEN loss was also significantly associated with rASRM stage ( P  = 0.024), being present in 47.8% (11/23) of Stage I, 73.7% (14/17) of Stage II, 80.8% (21/26) of Stage III, and 81.0% (34/42) of Stage IV ( Table 3 ). Additionally, the results remained significant ( P  = 0.031), when rASRM stage was binarized, with PTEN loss observed in 60.5% of participants with Stage I–II (26/43) and 80.0% of those with Stage III–IV (56/70). Similar trends were observed for ARID1A but did not reach statistical significance ( Table 3 ). Incorporating  KRAS  codon 12 mutation data, participants with ≥2 somatic events were also significantly more likely to have mixed anatomic subtypes and DE only or OMA only ( P  < 0.001), and to have Stage II–IV disease ( P  = 0.001) ( Table 3 ;  Fig. 3 ). Notably, co-occurrence of all three events (PTEN, ARID1A, and  KRAS ) within an individual participant was exclusive to Stages II, III, and IV ( Fig. 3 ).\nDistribution of somatic phenotypes stratified by anatomic subtype and stage.  ( A ) Anatomic subtype: There is a substantial difference in the proportion of participants with co-events in PTEN, ARID1A, and  KRAS  among the ovarian endometrioma (OMA) or deep endometriosis (DE) only, and mixed anatomic subtypes, compared to those with superficial peritoneal endometriosis (SUP) only. Additionally, having no event or only one event ( PTEN ,  ARID1A , or  KRAS  mutation) was most common among participants with SUP only. ( B ) Stage: Co-occurrence of the three events in a participant was observed exclusively in advanced stages (Stage II–IV). Stage I had the largest proportion of individuals with no events or those with only one event.\nBivariate relationship between somatic events and endometriosis anatomic subtypes and stage.\nDid not meet the minimum expected counts assumption for chi-square test.\nFrequencies are determined by row percentages.\nSUP, superficial peritoneal endometriosis; DE, deep endometriosis; OMA, ovarian endometrioma; rASRM, revised American Society for Reproductive Medicine.\nWe also examined co-events frequency at the lesion level, that is, the highest number of somatic genomic events in a lesion among all samples in a participant (considering ARID1A, PTEN, and  KRAS  data) ( Supplementary Table S1 ). We observed that detecting two or more events in the same lesion was more common in individuals with OMA or DE only (73.7%; 14/19) and Mixed subtypes (55.4%; 36/65) compared to SUP only (15.4%; 4/26). Similarly, this was also more common in higher rASRM stages: Stage I—13.6% (3/22); Stage II—52.6% (10/19); Stage III—72.0% (18/25); Stage IV—53.7% (22/41).\nDue to the prevalence of PTEN loss and  KRAS  mutation in the cohort, we also conducted a sub-analysis using multivariable logistic regression to assess whether they had independent effects on anatomic subtype (SUP only versus DE or OMA only/Mixed) and rASRM Stage (Stage I versus Stage II–IV) ( Supplementary Table S2 ). The results revealed that PTEN loss ( P  = 0.026; OR = 3.16; 95% CI: 1.15–8.70) and  KRAS  mutation ( P  = 0.016; OR = 3.67; 95% CI: 1.27–10.62) were independently associated with higher Stage. In the model for anatomic subtype, only PTEN loss showed an independent association with more severe anatomic subtype ( P  = 0.002; OR = 4.14; 95% CI: 1.68—10.21), while  KRAS  did not ( P  = 0.339; OR = 1.48; 95% CI: 0.66–3.33) ( Supplementary Table S2 ).\nAssociations between somatic events and other variables related to demographic, anatomic findings, surgical difficulty, and pain severity, were measured and reported in  Table 4 . PTEN loss was associated with self-reported non-White race/ethnicity (90%; 28/31) ( P  = 0.017), longer mean surgical times ( P  = 0.033), and higher frequency of ureterolysis (57.7%; 30/52) ( P  = 0.02) at the index surgery. No other significant associations were noted including none with baseline pain scores ( Table 4 ). Importantly, at our center, higher disease severity is associated with non-White ethnicities. In this study, non-White individuals experienced a higher proportion of Stage III–IV cases (82.8%; 24/29) compared to White individuals (51.6%; 48/93) ( P  = 0.004). Likewise, non-White individuals exhibited a significantly higher proportion of Mixed anatomic subtype (71.9%; 23/32) compared to White individuals (46.8%; 44/94) ( P  = 0.004). Therefore, to further examine the relationship between PTEN loss and ethnicity (focusing on non-White individuals), we performed two multivariable logistic regressions adjusting for anatomic subtype and rASRM stage. Adjusting for rASRM stage (OR = 1.934; 95% CI: 1.193–3.136,  P  = 0.007), PTEN loss was more common in non-White individuals though not statistically significant (OR = 3.094; 95% CI: 0.821–11.654,  P  = 0.095). Similarly, adjusting for anatomic subtype (OR = 2.121; 95% CI: 1.090–4.129,  P  = 0.027), PTEN loss was more common in non-White individuals but not statistically significant (OR = 3.056; 95% CI: 0.814–11.466,  P  = 0.098). For ARID1A loss, we did not observe associations with analyzed variables ( Table 5 ). For participants harboring at least two somatic events, significant associations with non-White race/ethnicity (75.9%; 22/29) ( P  = 0.016) and need for ureterolysis at the index surgery were observed (66.1%; 39/59) ( P  = 0.014) ( Table 6 ).\nBivariate analyses between PTEN expression and demographic, clinical, and surgical variables.\nFrequencies are determined by row percentages.\nPOD, pouch of Douglas.\nBivariate analyses between ARID1A expression and demographics, clinical and surgical variables.\nDid not meet the minimum expected counts assumption for chi-square test.\nFrequencies are determined by row percentages.\nPOD, pouch of Douglas.\nBivariate analyses between having two or more somatic events and demographic, clinical, and surgical variables.\nFrequencies are determined by row percentages.\nPOD, pouch of Douglas.\nThere was no association observed between post-operative follow-up pain scores and individuals affected by endometriosis harboring somatic alterations in PTEN, ARID1A, or having two or more somatic events (PTEN, ARID1A or  KRAS  mutation) ( Supplementary Tables S3 ,  S4 , and  S5 ). Overall, reoperation was uncommon in this cohort, occurring in 13.5% (17/126) of individuals over the 5- to 9-year prospective follow-up period. For participants with complete  PTEN ,  ARID1A , and  KRAS  data, we explored reoperation-free survival among different combinations of these somatic events ( Fig. 4 ). The reoperation-free survival curves showed the shortest reoperation-free survival (i.e. higher risk of reoperation) in participants with PTEN loss and  KRAS  mutation (without ARID1A loss); however, no statistical significance based on log-rank testing ( P  = 0.44) ( Fig. 4 ). Similar results were observed in a sub-analysis of reoperation-free survival excluding participants with previous hysterectomy and those who underwent a hysterectomy at index surgery at our site ( P  = 0.77) ( Fig. 5 ).\nKaplan–Meier survival analysis for reoperation risk for the follow-up period of 5–9 years.  Kaplan–Meier curve depicting reoperation-free survival based on somatic phenotype for individuals with complete data available for  PTEN ,  ARID1A , and  KRAS  mutation (n = 113); reoperation occurred in (11.3%; 15/113) of these cases. There was no significant association between reoperation at follow-up and somatic phenotype in the cohort ( P  = 0.44). However, participants with PTEN loss and  KRAS  mutation (without ARID1A loss) showed the lowest reoperation-free survival.\nKaplan–Meier survival analysis of a sub-analysis excluding participants with a prior hysterectomy before the index surgery or who underwent a hysterectomy at our site during the index surgery.  The graph includes data for individuals within this sub-group with complete data available on  PTEN ,  ARID1A , and  KRAS  mutation (n = 79); reoperation occurred in 16.5% (13/79) of these cases. The sub-analysis also showed no significant association between reoperation at follow-up and somatic phenotype in the cohort ( P  = 0.77), but participants with PTEN loss and  KRAS  mutation (without ARID1A loss) showed the lowest reoperation-free survival.\nPTEN and ARID1A status for participants are further described using three-tier IHC scoring criteria: retained in all samples in a participant; highest proportion of loss in glandular epithelium across any individual lesion = 1–49% in a participant; and highest proportion of loss across samples ≥50% in a participant ( Supplementary Table S6 ). According to the three-tier scoring criteria, 32.5% (38/117) of participants had ≥50% loss in at least one sample for PTEN, and 1.7% (2/120) for ARID1A. Furthermore, based on the three-tier scoring system, PTEN continued to have associations with anatomic subtype and with rASRM stage ( Supplementary Table S7 ).\n\nIn this prospective study, PTEN somatic loss was associated with both more severe anatomic subtypes and higher stages of endometriosis. PTEN loss was also associated with extended surgical times and a higher need for ureterolysis as markers of surgical difficulty, reflecting pronounced anatomical disruption observed during surgery due to more severe disease. These findings suggest a potential influence of PTEN loss (a somatic ‘cancer-driver’) on lesion survival, growth, proliferation, and progression in endometriosis without cancer, through subsequent dysregulation of the PI3K/AKT/mTOR pathway that is known to also be frequently altered in endometriosis-associated ovarian carcinomas ( Álvarez-Garcia  et al. , 2019 ;  Driva  et al. , 2023 ). Furthermore, PTEN loss was more prevalent in non-White individuals, consistent with previous associations between  KRAS  codon 12 mutation and race/ethnicity ( Orr  et al. , 2023 ). However, ARDI1A loss was less frequently observed (24.6%), which limited power for associations with the clinical variables. Unsurprisingly, participants with endometriosis having multiple somatic events (PTEN loss, ARID1A loss,  KRAS  mutation) exhibited more severe anatomic subtypes, increased stage, and greater surgical difficulty.\nInterestingly, the prevalence of PTEN loss (68.3% of participants) and ARID1A loss (24.6% of participants) in this cohort is higher than what has been reported previously for studies utilizing an IHC approach, with PTEN loss in 14–18% of endometriosis cases of various subtypes and ARID1A loss observed in <4% of endometriosis cases and 8–19% of OMA specifically ( Samartzis  et al. , 2012 ;  Xiao  et al. , 2012 ;  Chene  et al. , 2015 ;  Lac  et al. , 2019a , b ;  Yong  et al. , 2021 ). This discrepancy is likely attributed to how prevalence was calculated in this cohort: loss was determined per participant, defined as any loss in at least one lesion within a participant, with multiple lesions and anatomic subtypes sampled in each participant allowing for a comprehensive assessment of mutation status. When we applied a more stringent three-tier IHC definition specifically considering cases with ≥50% loss in at least one lesion, the observed loss rates aligned more closely with published data (32.5% for PTEN and 1.7% for ARID1A). For the time being, we recommend research studies attempt to gather as granular data as is reasonable such that clinically useful cutoffs for any of these biomarkers can be reviewed when appropriate.\nSequencing studies with potential to detect  PTEN  mutations in endometriosis have shown variability. Earlier studies utilizing Sanger sequencing have found  PTEN  mutations in 34% of DE but 0% of OMA ( Govatati  et al. , 2014 ;  Zou  et al. , 2018 ).  Anglesio  et al.  (2017)  did not report confirmed  PTEN  mutations on either exome or targeted panel sequencing, while  Suda  et al.  (2018)  identified  PTEN  mutations on sequencing in 4% of OMA cases. Praetorius et al. found PTEN loss in 18.5% of endometriosis cases (with at least two of the three anatomic subtypes) by IHC, however, none had confirmed mutations though only a very limited set of  PTEN  mutation hotspots were analyzed ( Praetorius  et al. , 2022 ). In the study by  Anglesio  et al.  (2017) , 8.3% (2/24) DE cases exhibited an  ARID1A  frameshift inactivating mutation with IHC-based validation (showing loss). Suda et al. found  ARID1A  mutations in 10% of OMA cases (present in 16% of subjects), post-laser capture enrichment, all of which were nonsense mutations or insertion–deletion frameshift mutations ( Suda  et al. , 2018 ).\nIn our study, it was notable that PTEN loss and ARID1A loss were not associated with baseline pain scores nor with follow-up pain scores after the index surgery, consistent with previous findings for  KRAS  codon 12 mutations. This is perhaps unsurprising given the known lack of association between stage and pain symptoms ( Szendei  et al. , 2005 ;  Vercellini  et al. , 2007 ;  Zondervan  et al. , 2018 ) likely related to comorbid pain conditions that can be present in endometriosis ( Zondervan  et al. , 2018 ). The low rate of reoperation in this cohort (13.5% over 5–9 years) may reflect the surgical expertise at this tertiary referral center, but we cannot rule out participants perhaps undergoing additional surgeries elsewhere. This low rate limited the statistical power for analyzing reoperation-free survival, and at this point, larger studies are needed to examine these reoperation outcomes with sufficient power.\nWhile our study does not investigate the relationship of endometriosis to related malignancies, clear cell or endometrioid ovarian carcinoma, we should recognize again that alterations to ARID1A, the PIK3CA/PTEN signaling axis, and KRAS activation are not uncommon in these malignancies ( Anglesio and Yong, 2017 ;  Dawson  et al. , 2018 ). Furthermore, our data are consistent with endometriosis exhibiting tumor-like traits more frequently when affected by  KRAS  mutation or PTEN loss at least, which may be somatic ‘drivers’ of greater burden/spread of disease (stage) and invasive characteristics (DE and OMA).\nA principal strength of this study is its prospective longitudinal study design inclusive of reoperation surveillance over 5–9 years. The prospective design allowed for standardized and systematic collection of clinical variable data. Moreover, the study included sensitive IHC with spatial information on PTEN and ARID1A expression and the results were analyzed by two observers with substantial inter-rater observer reliability. The use of IHC is also more potentially translatable to clinical pathologic use of PTEN and ARID1A status in endometriosis in the future.\nDespite using validated IHC proxies for mutation status, the absence of  PTEN  and  ARID1A  mutation validation via sequencing is a limitation in this study. While some mutations can lead to altered protein expression, not all result in detectable changes in protein levels, particularly for PTEN missense mutations which may be missed by IHC ( Köbel  et al. , 2018 ;  Wang  et al. , 2022 ). Conversely, IHC may detect additional loss of function or expression at the protein level that may be missed by sequencing, such as epigenetic silencing or large deletions. Djordjevic et al. showed substantial correlation between  PTEN  sequencing abnormalities and PTEN IHC (89% of cases with abnormalities show protein loss on IHC) in endometrial carcinoma, and stated that IHC was the superior method as it revealed additional functional loss that was undetected by sequencing alone ( Djordjevic  et al. , 2012 ).\nFor future research, it is critical that larger sample sizes are accumulated for these studies involving clinical correlation of somatic events, with corroborating sequencing data. A larger sample size will allow stratification by more racial/ethnic categories, especially among non-White individuals. Previous work at our center suggested that more anatomic severe endometriosis may particularly be higher among individuals of East and South-East Asian descent ( Williams  et al. , 2019 ). A larger sample size will also allow for controlling for concurrent pain conditions that may confound any associations with pain scores, and more power to detect any significant associations with reoperation-free survival (risk of reoperation).\n\nThis study demonstrates a significant association between somatic PTEN loss with higher endometriosis stage and more severe anatomic subtypes of endometriosis, reflecting a link to greater disease burden and subsequent surgical complexity. These findings underscore the need for further research to comprehensively elucidate the roles of somatic genomic events, including in tumor suppressors such as PTEN and ARID1A, in a future molecular classification for endometriosis.","source_license":"CC0","license_restricted":false}