A pilot study of discovery and validation of peritoneal endometriosis biomarkers in peritoneal fluid and serum

To identify potential serum biomarkers in women with peritoneal endometriosis (PE) 33 by first looking at its source in the peritoneal fluid (PF). 34 Design: Case-control pilot studies, comprising independent discovery and validation sets. 35 Setting: KK Women’s and Children’s Hospital, Singapore. 36 Patient(s): Women with laparoscopically confirmed PE and absence of endometriosis (control). 37 Intervention(s): None. 38 Main Outcome Measure(s): In the discovery set, we used untargeted liquid chromatography-39 mass spectrometry (LC-MS/MS) metabolomics, multivariable and univariable analyses to 40 generate global metabolomic profiles of PF for endometriosis and to identify potential 41 metabolites that could distinguish PE (n=10) from controls (n=31). Using targeted 42 metabolomics, we validated the identified metabolites in PF and sera of cases (n=16 PE) and 43 controls (n=19). We performed the area under the receiver-operating characteristics curve 44 (AUC) analysis to evaluate the diagnostic performance of PE metabolites. 45 Result(s): In the discovery set, PF phosphatidylcholine (34:3) and phenylalanyl-isoleucine were 46 significantly increased in PE than controls groups, with AUC 0.77 (95% confidence interval 0.61-47 0.92; p=0.018) and AUC 0.98 (0.95-1.02; p<0.001), respectively. In the validation set, 48 phenylalanyl-isoleucine retained discriminatory performance to distinguish PE from controls in 49 both PF (AUC 0.77; 0.61-0.92; p=0.006) and serum samples (AUC 0.81; 0.64-0.99; p=0.004). 50 Conclusion(s): Our preliminary results propose phenylalanyl-isoleucine as a potential 51 biomarker of PE, which may be used as a minimally-invasive diagnostic biomarker of PE. 52

biomarker, diagnosis, endometriosis, LC-MS/MS, metabolomics 53 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 4

54 Endometriosis affects approximately 10% of reproductive aged women (1, 2) and is associated 55 with substantial morbidity, including chronic pelvic pain and infertility (3, 4). Endometriosis is 56 represented by three main subphenotypes: ovarian endometriosis (OE), superficial peritoneal 57 endometriosis (PE) and deep infiltrating endometriosis (DIE) (5). To diagnose endometriosis, 58 transvaginal ultrasonography can be used to detect OE and DIE (6, 7), with pelvic magnetic 59 resonance imaging to assess the extent of DIE (1). However, the detection of PE, characterized 60 by superficial endometrial lesions occurring on the peritoneum, remains challenging (8-11). 61 Laparoscopic visualization remains as the standard for definitive diagnosis of PE, the most 62 common subphenotype which accounts for ~80% of all endometriosis (11-13). 63 Using laparoscopic visualization as the first line diagnostic tool poses a number of 64 challenges, including its invasive nature, associated risks and potential complications of surgery 65 (11). Laparoscopy is appropriate when symptoms reach a level of severity to justify the surgical 66 risk (5), yet clinical symptoms has a poor correlation with disease burden (10). Indeed, accuracy 67 of diagnosis is dependent on practitioners’ laparoscopic skills due to the diversity of 68 endometriotic appearances and locations, insofar that endometriosis may be inadvertently 69 missed with less obvious or microscopic endometriotic lesions (14). Consequently, there is often 70 a delay with an average of eight years in the diagnosis of endometriosis (6). Thus, there is a 71 great need to identify a less invasive method for PE diagnosis, which would have a 72 groundbreaking impact in preventing or delaying disease progression, improving patients’ 73 quality of life and the efficacy of available treatments. This is particularly important in women for 74 whom fertility is a priority whereby hormonal treatment is not appropriate (13). 75 To date, despite the evaluation of numerous potential biomarkers, a reliable biomarker 76 specifically for the diagnosis of PE has yet to be identified (10). Advances of high-throughput 77 bioanalytical technologies in omics have made metabolomics a powerful tool for biomarker 78 discovery (15, 16). Metabolites are intermediates to a wide range of biological processes and 79 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 5 signaling axes such as mammalian target of rapamycin (mTOR), peroxisome proliferator-80 activated receptor (PPAR) and mitogen-activated protein kinase (MAPK) pathways (17-19). The 81 association of aberrant metabolism and endometriosis has emerged in recent years (20-22). 82 Differences in metabolism and metabolite levels reflecting endometriosis subphenotypes 83 pathophysiology potentially forms the basis for identifying novel biomarkers for PE. 84 The peritoneal fluid (PF) is notably rich in proteins and lipids including cytokines, 85 chemokines, growth factors and matrix metalloproteinases, and serves an important 86 environment where endometriotic lesions reside and communicate with surrounding tissues 87 including nerve cells and ovaries (23-25). The PF is therefore a source for the assessment of 88 the dysregulated peritoneal cavity, and for reflecting dysregulated metabolic state of the 89 subphenotypes. Thus, in this study, we aimed to identify potential biomarkers for the diagnosis 90 of peritoneal endometriosis by first looking at its source in the peritoneal fluid. Through 91 untargeted metabolomics, we characterized global metabolomics alterations in the peritoneal 92 fluid samples, and multivariable and univariable statistics were used to discover potential 93 biomarkers that resolve women with a laparoscopic diagnosed peritoneal endometriosis and 94 without endometriosis. To verify the identified potential biomarkers in peritoneal fluid, we 95 performed targeted metabolomics on peritoneal fluid and serum samples of women from 96 independent case-control sets. 97 98

99 Study design 100 In this case-control pilot study, investigation of biomarkers was conducted in two phases 101 (Supplemental Figure 1, available online). Phase I, defined as the discovery set, untargeted 102 metabolomics approach was employed to generate PF-specific global metabolomic maps of 103 endometriosis. Phase II, defined as the validation set, where two separate groups of women 104 were independently enrolled to verify the identified PF-metabolites for PE diagnosis. We 105 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 6 performed targeted metabolomics analysis using both PF (invasive) and serum samples 106 (minimally-invasive) among women with PE and controls. This study was conducted according 107 to the Helsinki Declaration, and all procedures were approved by the Singhealth Centralized 108 Institutional Review Board (reference 2010-167-D). 109 110 Clinical participants 111 Patients were recruited from the subfertility clinic in the KK Women’s and Children’s Hospital, 112 Singapore, between June 2010 and May 2013 for Phase I, and between June 2010 and August 113 2016 for Phase II. Laparoscopy was scheduled for suspected endometriosis, infertility, 114 sterilization procedures, and/or pelvic pain. Exclusion criteria included menstruating patients, 115 post-menopausal patients, patients on hormonal therapy (e.g. norethisterone, combined oral 116 contraceptive pill) for at least three months before laparoscopy, and other potentially 117 confounding diseases such as diabetes, adenomyosis or any other chronic inflammatory 118 diseases (rheumatoid arthritis, inflammatory bowel disease, systemic sclerosis). All eligible 119 patients provided written informed consent upon recruitment. 120 During diagnostic laparoscopy, a detailed inspection of the uterus, fallopian tubes, 121 ovaries, pouch of Douglas and the pelvic peritoneum was performed by senior gynecologists 122 subspecialized in reproductive endocrinology and infertility. Patients with laparoscopically 123 confirmed PE were defined as the cases, and staged according to the revised American Fertility 124 Society classification of endometriosis (AFS, 1985; ASRM, 1997). The possible overlapping of 125 the three lesion subphenotypes led us to classify the patients according to the worst lesion 126 found in each subject, based on endometriosis subphenotype grouping by Somigliana et al. (26) 127 and Chapron et al. (27). Patients from the same clinic but laparoscopically observed to be 128 without endometriosis were defined as the controls, including those with benign gynecological 129 presentations such as tubal occlusion, uterine fibroids, benign ovarian cysts and polycystic 130 ovary. 131 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 7 132 Sample collection and sample processing 133 For the collection of PF, clinical staff collected samples via aspiration with a syringe attached to 134 an irrigation/suction tube from the Pouch of Douglas, with 1% protease inhibitor added (Roche, 135 Switzerland) as previously described (28), and which is in line with Endometriosis Phenome and 136 Biobanking Harmonization Project Standard Operating Procedures (29). The aspirates were 137 then centrifuged (1,000×g, 4°C) for 10 min and the clear supernatants were transferred to 15 mL 138 tubes. For the collection of peripheral venous blood, samples were collected into the BD 139 Vacutainer® SST II tubes. After 10 min centrifugation (1,200×g, 4°C), the top yellowish layers 140 were transferred to 15 mL tubes, followed by centrifuging for another 10 min (3,600×g, 4°C) 141 where the supernatants were transferred to 1 ml aliquots. Both PF and serum samples were 142 stored at -80°C until further use (28). 143 Prior to LC-M/S analysis, 100 μ L from PF or serum sample was thawed at 4°C and 144 proteins were precipitated with 400 μ L ice-cold methanol. After vortexing for 1 min, the mixture 145 was centrifuged at 17,000×g for 10 min at 4 °C and the supernatant was collected and 146 evaporated to dryness in a vacuum concentrator. The dry extracts were then resuspended in 147 100 μ L of 98:2 water/methanol or in 100 µL of 0.1% formic acid in methanol for untargeted or 148 targeted metabolomics, respectively. Quality control (QC) samples were prepared by mixing 149 equal amounts of reconstituted extracts from all the samples and processed as per other 150 samples. All samples were kept at 4°C and analyzed in a random manner. QC samples are 151 interspersed through the analytical runs and ran after each 10th sample to monitor the stability of 152 the system. 153 154 Untargeted mass spectrometry analysis 155 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 8 We performed metabolomics analysis as previously described with modifications (30). 156 Reversed-phase liquid chromatography-MS analyses were performed using the Agilent 1290 157 ultrahigh pressure liquid chromatography system (Waldbronn, Germany) equipped with a 6520 158 Q-TOF mass detector managed by a MassHunter workstation. The column used for the 159 separation was rapid resolution HT Zorbax SB-C18 (2.1×100 mm, 1.8 μ m; Agilent Technologies, 160 USA), and the mobile phase was (A) 0.1% formic acid in water and (B) 0.1% formic acid in 161 methanol. The initial condition of the gradient elution was set at 2% B for 2 min with a flow rate 162 of 0.4 ml/min. A 7 min linear gradient to 70% B was then applied, followed by a 5 min gradient to 163 100% B which was held for 3 min. The sample injection volume was 2 μ L and the oven 164 temperature was set at 40°C. The electrospray ionization mass spectra were acquired in both 165 positive and negative ion mode. Mass data were collected between m/z 100 and 1000 at a rate 166 of two scans per second. The ion spray voltage was set at 4,000 V for positive mode and 3,500 167 V for negative mode. The heated capillary temperature was maintained at 350°C. The drying 168 gas and nebulizer nitrogen gas flow rates were 12.0 L/min and 50 psi, respectively. Two 169

masses were continuously infused to the system to allow constant mass correction 170 during the run: m/z 121.0509 (C5H4N4) and m/z 922.0098 (C18H18O6N3P3F24). 171 172 Targeted mass spectrometry analysis 173 The targeted LC-MS/MS analysis was performed in multiple reaction monitoring mode via Triple 174 Quadrupole 6460 mass spectrometer with Jet Stream (Agilent Technologies). Chromatographic 175 separation was achieved by using Eclipse Plus column C18 (2.1×50 mm; Agilent, US) with 176 mobile phases (A) 10 mM ammonium formate and 0.1% formic acid in water and (B) 0.1% 177 formic acid in methanol. The initial condition was set at 100% A for 3 min at a flow rate of 178 0.3ml/min. A 3 min linear gradient to 100% B was then applied and held for 3 min. Then the 179 gradient returned to starting conditions over 0.1 min and maintain at the initial condition for 3 180 min. The column was kept at 45°C and the flow rate was 0.3 mL/min. The auto-sampler was 181 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 9 cooled at 4°C and an injection volume of 2 μ L was applied. Mass transition and collision energy 182 were optimized for each compound by direct infusion of individual standard solutions. Both 183 positive and negative electrospray ionization modes were performed with the following source 184 parameters: drying gas temperature at 250°C with a flow of 5 L/min, nebulizer gas pressure at 185 40 psi, sheath gas temperature at 400°C with a flow of 11 L/min, capillary voltage 4,000 and 186 3,500 V for positive and negative mode respectively, and nozzle voltage 500 V for both positive 187 and negative modes. Data acquisition and processing were performed using MassHunter 188 software (Agilent Technologies, US). 189 190 Data analysis and Compound identification 191 For metabolomics analysis, raw spectrometric data were converted to mzData (LC-MS/MS) and 192 NetCDF (GC-MS) formats via Masshunter (Agilent, US) and input to open-source software 193 MZmine 2.0 for peak finding, peak alignment and peak normalization across all samples. The 194 structure identification of the differential metabolites was based on a previously described 195 strategy (31). First, the element composition C13H25NO4 of the m/z 260.18 ion was calculated 196 based on the exact mass, the nitrogen rule and the isotope pattern by Masshunter software 197 from Agilent. Then, the elemental composition and exact mass were used for open source 198 database searching, including LIPIDMAPS (http://www.lipidmaps.org/), HMDB 199 (http://www.hmdb.ca/), METLIN (http://metlin.scripps.edu/) and MassBank 200 (http://www.massband.jp/). Next, MS/MS experiments were performed to obtain structural 201 information via the interpretation of the fragmentation pattern of the metabolite. 202 203 Statistical analysis 204 We used IBM SPSS statistics, version 19 (USA), the Unscrambler and GraphPad Prism, version 205 7.0 (USA) for statistical analyses. For the untargeted metabolomics, we used partial least 206 squares analysis and volcano plots to preliminarily select significant metabolites. Mann-Whitney 207 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 10 U-test and Fisher’s exact test were used to compare continuous and categorical subject 208 characteristics respectively between PE versus controls. Significantly different metabolites 209 (SDMs) were defined by a fold change of >1.50 for increased metabolites and <0.67 for 210 decreased metabolites. We performed the area under the receiver-operating characteristics 211 (ROC) curve (AUC) analysis to evaluate the diagnostic performance of PE metabolites. 212 AUC>0.70 was defined as optimal diagnostic performance value. Sensitivity and specificity were 213 determined at maximum Youden Index. For the discovery datasets, a two-sided p<0.05 was 214 considered statistical significance. For the validation datasets, a two-sided p<0.025 (0.05/2 215 outcomes) to account for multiplicity was considered statistically significant. 216 217

218 Characteristics of participants 219 For the discovery set, PF samples were analyzed for 10 women with PE (mean age 33.3 years 220 old) and 31 women who served as controls (mean age 33.9 years old). Majority of women with 221 PE were at minimal-mild stage of endometriosis (rAFS stage I-II). No differences were observed 222 in terms of age, ethnicity and cycle phase between women with PE and controls (Supplemental 223 Table 1, available online). 224 For the validation set, women with PE and controls were compared and analyzed. These 225 included PF samples for 19 PE (mean age 34.0 years old) and 20 controls (mean age 35.9 226 years old). Serum samples were available for 16 PE and 19 controls, as illustrated in 227 Supplemental Figure 1 (available online). Similar to the discovery set, majority of women with 228 PE (89.5%) were at minimal-mild stage of endometriosis. No differences in age, ethnicity and 229 cycle phase were shown between women with PE and controls. 230 231 Different PF metabolomic profiles between women with PE and controls 232 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 11 Multivariable partial least squares regression (PLSR) model was constructed to reveal the PF 233 metabolome differences between women with PE and controls from the discovery set (Phase I). 234 The PLSR score plots unbiasedly yielded good separation between metabolomic profiles of 235 women with PE and controls (Figure 1A). Next, we projected the metabolomic profiles on the 236 endometriosis subphenotypes to identify metabolites that drive the subphenotype separation. 237 The PLSR loadings plot yielded biochemically diverse metabolites for the separation of women 238 with PE and controls. This discrimination was primarily driven by the top 20 metabolites shown 239 in Figure 1B. These metabolites comprised of 7 phospholipids, 5 free fatty acids, 4 amino acids, 240 1 carnitine, 1 sphingolipid, 1 dipeptide and 1 tricarboxylic acid cycle intermediate. The list of PF 241 metabolites for PE and controls are shown in Supplemental Table 2 (available online). To 242 further ascertain the relationship of key metabolites to the subphenotypes, fold change was 243 applied to determine metabolites that are significantly different in PE compared with controls. 244 245 Discovery of PF metabolites that distinguished PE from controls 246 Using volcano plots, 13 PF metabolites showed significant fold change differences (p1.50 or <0.67. Compared with the controls, women with PE exhibited higher levels 249 of 5-tetradecenoylcarnitine (2.33-fold), phosphatidylcholine C34:3 (1.66-fold), phenylalanyl-250 isoleucine (1.79-fold) and tetracosahexaenoic acid (1.72-fold), but lower level of ceramide d34:0 251 (0.44-fold) (Figure 2). When diagnostic performance was evaluated by ROC analysis, 252 phenylalanyl-isoleucine and phosphatidylcholine C34:3 demonstrated diagnostic potential in 253 distinguishing women with PE from controls, as indicated by AUC 0.98 (95% confidence interval 254 (CI) 0.95-1.02; p<0.001; sensitivity 100%; specificity 95.8%) and by AUC 0.77 (95% CI 0.61-255 0.92; p=0.018; sensitivity 77.8%; specificity 71.4%), respectively (Figure 3). 256 257 Validation of potential PF and serum metabolites that distinguished PE from controls 258 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 12 In Phase II, to more precisely quantify the relationship of the candidate biomarkers in PE 259 relative to controls, we next developed a targeted metabolomics assay to validate biomarker 260 candidates phenylalanyl-isoleucine and phosphatidylcholine C34:3. Analyzing the PF samples 261 from an independent set of women using our developed targeted metabolomics assay, 262 phenylalanyl-isoleucine retained its diagnostic potential in distinguishing PE from controls (AUC 263 0.77, 95% CI 0.61-0.92; p=0.006; sensitivity 73.7%; specificity 72.2%) (Figure 4A). By contrast, 264 phosphatidylcholine C34:3 did not hold up to its initial diagnostic value (AUC 0.65, 95% CI 0.47-265 0.83; p=0.121) and hence not considered for further analysis. Next, we investigated the utility of 266 phenylalanyl-isoleucine as a minimally-invasive biomarkers using serum samples. We found 267 that circulating phenylalanyl-isoleucine consistently distinguished PE from controls with AUC 268 0.81 (95% CI 0.64-0.99; p=0.004; sensitivity 71.4%; specificity 100%) (Figure 4B). 269 270

271 In this pilot study, we identified PF metabolites that differentiated women with PE from controls. 272 Using untargeted metabolomic LC-MS/MS that characterized the global metabolomic alterations 273 in the PF, a diverse PF metabolomes was shown in these women. The PF metabolomic profiles 274 pertaining to the PE subphenotype were then linked the pathophysiological changes to the 275 serum for biomarker discovery using a combination of untargeted and targeted metabolomics. 276 We provided evidence showing a novel PF metabolite, phenylalanyl-isoleucine, has the 277 potential as a biomarker of PE and was subsequently validated. Importantly, this metabolite was 278 reflected in circulation, with a diagnostic performance value of 81.4% in serum (sensitivity 279 71.4%; specificity 100%), suggesting its potential use as a minimally-invasive diagnostic 280 biomarker of PE. 281 PF is proximal to endometriotic lesions and thus, forms an environment that reciprocally 282 communicates with the lesions (32, 33). This indicates that PF can be a potential useful source 283 for biomarker discovery of PE. Intriguingly, the diagnostic characteristics of PF biomarkers for 284 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 13 PE using metabolomics approach have not been assessed. Majority of earlier studies have 285 focused on identifying potential biomarkers of endometriosis in general without subphenotypes 286 specificity (21, 28, 34-37), with few focused on OE (22, 38, 39) but limited on PE (40). 287 Importantly, validation of proposed biomarkers which is critical to demonstrate biomarker 288 robustness has been rarely conducted (41). In biomarker discovery studies, differential levels of 289 metabolites such as cancer antigen 125 (CA-125), carnitines, phosphatidylcholine and 290 sphingomyelin in PF were observed in women with endometriosis (28, 38, 42), but there is 291 seldom linkage of these biomarkers to the circulating levels. CA-125 is a commonly investigated 292 biomarker for endometriosis despite its undefined role in primary diagnosis (3, 10). As 293 suggested, we performed additional analysis for CA-125 in serum of women from the validation 294 set and compared with its performance with serum phenylalanyl-isoleucine. Serum CA-125 295 showed poor diagnostic performance (AUC 0.66; 95% CI 0.47-0.84; p=0.120) in differentiating 296 PE (n=19) from controls (n=16) (Supplemental Figure 2, available online). Our findings are 297 consistent with international guidelines, such as the National Institute for Health and Care 298 Excellence (NICE) and the European Society of Human Reproduction and Embryology 299 (ESHRE), which have made recommendations to not use serum CA-125 as biomarker for 300 endometriosis diagnosis due to its limited diagnostic performance (13, 43). 301 Our results demonstrate that phenylalanyl-isoleucine was increased in PF and serum of 302 PE women, with a high discriminatory ability to distinguish between women with PE and other 303 gynecological disorders requiring laparoscopic diagnosis (controls). Thus, data obtained from 304 this study have connected pathophysiology in the dysregulated peritoneal cavity to the 305 circulation in endometriosis, in line with reports showing correlations between biomarkers in PF 306 and serum (44, 45). Phenylalanyl-isoleucine (C15H22N2O3) is a dipeptide composed of 307 phenylalanine and isoleucine (46), which has been shown to play an essential role in 308 intracellular signal transduction (47). However, its functional role in endometriosis 309 pathophysiology remains unclear. To the best of our knowledge, phenylalanyl-isoleucine is a 310 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 14 novel metabolite which has not been reported for endometriosis. Increased phenylalanyl-311 isoleucine in PF might indicate an altered metabolic state that would potentially contribute to 312 further growth and development of peritoneal lesions (48). Compared to healthy controls, 313 women with early stage of endometriosis had previously found to exhibit decreased 314 phenylalanine and isoleucine in endometrial tissue and serum, respectively (20, 37). In contrast, 315 phenylalanyl-isoleucine appeared to be relatively abundant in PF of women with PE, which 316 could be explained by the proximity of PF with endometriotic lesions implanted in the peritoneal 317 cavity. It is possible that increased phenylalanyl-isoleucine reflected shedding into the PF due to 318 high levels of cell division, cell death and protein degradation. This is supported by Li et al. (49) 319 which found upregulations of various amino acids in the eutopic endometrium of women with 320 early endometriosis. Further research is required to elucidate the association of phenylalanyl-321 isoleucine with PE pathophysiology. 322 Several studies have classified women with endometriosis according to their disease 323 stage, where greater altered metabolomic profiles and metabolite levels were seen in women 324 with later stages (III-IV) of endometriosis as compared with the controls (44, 45). Importantly, 325 differences in the levels and types of metabolites were also shown for women with stage I and II 326 endometriosis in relation to the controls (37, 49), indicating their potential for early disease 327 diagnosis and prognosis. In this study, PE cases were mostly classified at minimal-mild stages 328 (stage I-II), suggesting the candidate metabolite in women with PE might be additionally useful 329 to reflect the early stage of the disease. 330 Strengths of this study included quantitative and structured methodological framework to 331 identify the potential biomarker of PE, with the applications of untargeted, unbiased 332 metabolomics in the discovery process and targeted metabolomics with exquisite assay 333 sensitivity and precision in the validation process. Importantly, this study captured the most 334 comprehensive catalogued metabolome space of PF in women with laparoscopic diagnosed PE 335 to-date. By treating the identification of PF biomarker as a pre-selection procedure for 336 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 15 subsequent study for serum marker, we successfully demonstrated the presence of candidate 337 biomarker in sera of PE women from an independent group, with optimal diagnostic 338 performance for PE. The procedure of validating biomarker using separate, independent sets of 339 samples by applying same analytical approach has rarely been carried out in most previous 340 studies. However, an important limitation of this study was the small sample size. This has 341 restricted our capability to stratify women with PE into individual stage of disease for further 342 assessing the prognostic applications of phenylalanyl-isoleucine. However, preliminary findings 343 generated from this pilot study serves as important baseline supportive evidence for conducting 344 subsequent larger scale studies to confirm the reliability and validity of phenylalanyl-isoleucine 345 for PE diagnosis. 346 347

348 In conclusion, we have identified a signature metabolite known as phenylalanyl-isoleucine in PF 349 of women with PE through untargeted and targeted metabolomics in independent datasets, 350 suggesting its involvement in the pathophysiology of PE. The same metabolite was identified in 351 the sera of women with PE. Large scale metabolomics studies validating this biomarker across 352 different populations will further determine its diagnostic robustness. More studies are also 353 warranted to test for the robustness of this biomarker in distinguishing PE from other 354 subphenotypes of endometriosis and pelvic inflammatory diseases that exhibit similar clinical 355 symptoms. 356 357

358 We thank the participants of the study. This work was supported by the National Research 359 Foundation Singapore under its National Medical Research Council Centre Grant Program 360 (NMRC/CG/M003/2017) and administered by the Singapore Ministry of Health’s National 361 Medical Research Council, and Duke-NUS Office of Academic Medicine (Clinical and 362 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 16 Translational Research in Endometriosis). JCKY received salary support from Singapore’s 363 Ministry of Health’s National Medical Research Council (CSA-SI-008-2016). 364 365

366 1. Alimi Y, Iwanaga J, Loukas M, Tubbs RS. The clinical anatomy of endometriosis: a 367 review. Cureus 2018;10:e3361. doi: 10.7759/cureus.3361. 368 2. Shafrir AL, Farland LV, Shah DK, Harris HR, Kvaskoff M, Zondervan K, et al. Risk for 369 and consequences of endometriosis: a critical epidemiologic review. Best Pract Res Clin 370 Obstet Gynaecol 2018;51:1-15. doi: 10.1016/j.bpobgyn.2018.06.001. 371 3. Vercellini P, Viganò P, Somigliana E, Fedele L. Endometriosis: pathogenesis and 372 treatment. Nat Rev Endocrinol 2014;10:261-75. doi: 10.1038/nrendo.2013.255. 373 4. Prescott J, Farland LV, Tobias DK, Gaskins AJ, Spiegelman D, Chavarro JE, et al. A 374 prospective cohort study of endometriosis and subsequent risk of infertility. Hum Reprod 375 2016;31:1475-82. doi: 10.1093/humrep/dew085. 376 5. Zondervan KT, Becker CM, Missmer SA. Endometriosis. N Engl J Med 2020;382:1244-377 56. doi: 10.1056/NEJMra1810764. 378 6. Barbieri RL. Why are there delays in the diagnosis of endometriosis. OBG Manag 379 2017;29:8-11. 380 7. Ferrero S. Endometriosis: modern management of an ancient disease. Eur J Obstet 381 Gynecol Reprod Biol 2017;209:1-2. doi: 10.1016/j.ejogrb.2016.12.036. 382 8. Dunselman GA, Vermeulen N, Becker C, Calhaz-Jorge C, D'Hooghe T, De Bie B, et al. 383 ESHRE guideline: management of women with endometriosis. Hum Reprod 384 2014;29:400-12. doi: 10.1093/humrep/det457. 385 9. Nisenblat V, Bossuyt PM, Farquhar C, Johnson N, Hull ML. Imaging modalities for the 386 non-invasive diagnosis of endometriosis. Cochrane Database Syst Rev 387 2016;2:CD009591. doi:10.1002/14651858.CD009591.pub2. 388 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 17 10. Parasar P, Ozcan P, Terry KL. Endometriosis: epidemiology, diagnosis and clinical 389 management. Curr Obstet Gynecol Rep 2017;6:34-41. doi: 10.1007/s13669-017-0187-1. 390 11. Chapron C, Marcellin L, Borghese B, Santulli P. Rethinking mechanisms, diagnosis and 391 management of endometriosis. Nat Rev Endocrinol 2019;15:666-82. doi: 392 10.1038/s41574-019-0245-z. 393 12. Horne AW, Daniels J, Hummelshoj L, Cox E, Cooper KG. Surgical removal of superficial 394 peritoneal endometriosis for managing women with chronic pelvic pain: time for a 395 rethink? BJOG 2019;126:1414-6. doi: 10.1111/1471-0528.15894. 396 13. Kuznetsov L, Dworzynski K, Davies M, Overton C. Diagnosis and management of 397 endometriosis: summary of NICE guidance. BMJ 2017;358:j3935. doi: 398 10.1136/bmj.j3935. 399 14. Mettler L, Schollmeyer T, Lehmann-Willenbrock E, Schüppler U, Schmutzler A, Shukla 400 D, et al. Accuracy of laparoscopic diagnosis of endometriosis. JSLS 2003;7:15-8. 401 15. Cui L, Lu H, Lee YH. Challenges and emergent solutions for LC-MS/MS based 402 untargeted metabolomics in diseases. Mass Spectrom Rev 2018;37:772-92. doi: 403 10.1002/mas.21562. 404 16. Pang H, Jia W, Hu Z. Emerging applications of metabolomics in clinical pharmacology. 405 Clin Pharmacol Ther 2019;106:544-56. doi:10.1002/cpt.1538. 406 17. Gehart H, Kumpf S, Ittner A, Ricci R. MAPK signalling in cellular metabolism: stress or 407 wellness? EMBO Rep 2010;11:834-40. doi:10.1038/embor.2010.160. 408 18. Ahmadian M, Suh JM, Hah N, Liddle C, Atkins AR, Downes M, et al. PPARγ signaling 409 and metabolism: the good, the bad and the future. Nat Med 2013;19:557-66. doi: 410 10.1038/nm.3159. 411 19. Saxton RA, Sabatini DM. mTOR signaling in growth, metabolism, and disease. Cell. 412 2017;169:361-71. doi: 10.1016/j.cell.2017.03.035. 413 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 18 20. Dutta M, Joshi M, Srivastava S, Lodh I, Chakravarty B, Chaudhury K. A metabonomics 414 approach as a means for identification of potential biomarkers for early diagnosis of 415 endometriosis. Mol Biosyst 2012;8:3281-7. doi:10.1039/c2mb25353d. 416 21. Braga DPAF, Montani DA, Setti AS, Turco EGL, Oliveira-Silva D, Borges E Jr. 417 Metabolomic profile as a noninvasive adjunct tool for the diagnosis of Grades III and IV 418 endometriosis-related infertility. Mol Reprod Dev 2019;86:1044-52. 419 doi:10.1002/mrd.23221. 420 22. Karaer A, Tuncay G, Mumcu A, Dogan B. Metabolomics analysis of follicular fluid in 421 women with ovarian endometriosis undergoing in vitro fertilization. Syst Biol Reprod Med 422 2019;65:39-47. doi: 10.1080/19396368.2018. 423 23. Zhou J, Chern BS, Barton-Smith P, Phoon JW, Tan TY, Viardot-Foucault V, et al. 424 Peritoneal fluid cytokines reveal new insights of endometriosis subphenotypes. Int J Mol 425 Sci 2020;21:3515. doi:10.3390/ijms21103515. 426 24. Lee YH, Yang JX, Allen JC, Tan CS, Chern BSM, Tan TY, et al. Elevated peritoneal fluid 427 ceramides in human endometriosis-associated infertility and their effects on mouse 428 oocyte maturation. Fertil Steril 2018;110:767-77.e5. doi: 429 10.1016/j.fertnstert.2018.05.003. 430 25. Laganà AS, Garzon S, Götte M, Viganò P, Franchi M, Ghezzi F, et al. The pathogenesis 431 of endometriosis: molecular and cell biology insights. Int J Mol Sci 2019;20:5615. 432 doi:10.3390/ijms20225615. 433 26. Somigliana E, Infantino M, Candiani M, Vignali M, Chiodini A, Busacca M, et al. 434 Association rate between deep peritoneal endometriosis and other forms of the disease: 435 pathogenetic implications. Hum Reprod 2004;19:168-71. doi:10.1093/humrep/deg513. 436 27. Chapron C, Souza C, de Ziegler D, Lafay-Pillet MC, Ngô C, Bijaoui G, et al. Smoking 437 habits of 411 women with histologically proven endometriosis and 567 unaffected 438 women. Fertil Steril 2010;94:2353-5. doi: 10.1016/j.fertnstert.2010.04.020. 439 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 19 28. Lee YH, Tan CW, Venkatratnam A, Tan CS, Cui L, Loh SF, et al. Dysregulated 440 sphingolipid metabolism in endometriosis. J Clin Endocrinol Metab 2014;99:E1913-21. 441 doi: 10.1210/jc.2014-1340. 442 29. Rahmioglu N, Fassbender A, Vitonis AF, Tworoger SS, Hummelshoj L, D'Hooghe TM, et 443 al. World Endometriosis Research Foundation Endometriosis Phenome and Biobanking 444 Harmonization Project: III. Fluid biospecimen collection, processing, and storage in 445 endometriosis research. Fertil Steril 2014;102:1233-43. doi: 446 10.1016/j.fertnstert.2014.07.1208. 447 30. Cui L, Lee YH, Thein TL, Fang J, Pang J, Ooi EE, et al. Serum metabolomics reveals 448 serotonin as a predictor of severe dengue in the early phase of dengue fever. PLoS Negl 449 Trop Dis 2016;10:e0004607. doi:10.1371/journal.pntd.0004607. 450 31. Chan KR, Gan ES, Chan CYY, Liang C, Low JZH, Zhang SL, et al. Metabolic 451 perturbations and cellular stress underpin susceptibility to symptomatic live-attenuated 452 yellow fever infection. Nat Med 2019;25:1218-24. doi: 10.1038/s41591-019-0510-7. 453 32. Koninckx PR, Kennedy SH, Barlow DH. Endometriotic disease: the role of peritoneal 454 fluid. Human Reprod Update 1998;4:741-51. 455 33. Yang H, Lau WB, Lau B, Xuan Y, Zhou S, Zhao L, et al. A mass spectrometric insight 456 into the origins of benign gynecological disorders. Mass Spectrom Rev 2017;36:450-70. 457 doi: 10.1002/mas.21484. 458 34. Lee YH, Cui L, Fang J, Chern BS, Tan HH, Chan JK. Limited value of pro-inflammatory 459 oxylipins and cytokines as circulating biomarkers in endometriosis - a targeted 'omics 460 study. Sci Rep 2016;6:26117. doi: 10.1038/srep26117. 461 35. Vicente-Muñoz S, Morcillo I, Puchades-Carrasco L, Payá V, Pellicer A, Pineda-Lucena 462 A. Pathophysiologic processes have an impact on the plasma metabolomic signature of 463 endometriosis patients. Fertil Steril 2016;106:1733-41.e1. doi: 464 10.1016/j.fertnstert.2016.09.014. 465 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 20 36. Letsiou S, Peterse DP, Fassbender A, Hendriks MM, van den Broek NJ, Berger R, O et 466 al. Endometriosis is associated with aberrant metabolite profiles in plasma. Fertil Steril 467 2017;107:699-706.e6. doi: 10.1016/j.fertnstert.2016.12.032. 468 37. Dutta M, Singh B, Joshi M, Das D, Subramani E, Maan M, et al. Metabolomics reveals 469 perturbations in endometrium and serum of minimal and mild endometriosis. Sci Rep 470 2018;8:6466. doi: 10.1038/s41598-018-23954-7. 471 38. Vouk K, Ribič -Pucelj M, Adamski J, Rižner TL. Altered levels of acylcarnitines, 472 phosphatidylcholines, and sphingomyelins in peritoneal fluid from ovarian endometriosis 473 patients. J Steroid Biochem Mol Biol 2016;159:60-9. doi: 10.1016/j.jsbmb.2016.02.023. 474 39. Domínguez F, Ferrando M, Díaz-Gimeno P, Quintana F, Fernández G, Castells I, et al. 475 Lipidomic profiling of endometrial fluid in women with ovarian endometriosis†. Biol 476 Reprod 2017;96:772-9. doi: 10.1093/biolre/iox014. 477 40. Bilibio JP, Souza CAB, Rodlini GP, Andreali CG, Genro VK, de Conto E et al. Serum 478 prolactin and CA-125 levels as biomarkers of peritoneal endometriosis. Gynecol Obstet 479 Invest 2014;78:45-52. doi: 10.1159/000441790. 480 41. M D'Hooghe T, Fassbender A, F O D, Vanhie A. Endometriosis biomarkers: Will 481 codevelopment in academia-industry partnerships result in new and robust noninvasive 482 diagnostic tests? Biol Reprod 2019;101:1140-5. doi: 10.1093/biolre/ioz016. 483 42. Matalliotakis IM, Goumenou AG, Mulayim N, Karkavitsas N, Koumantakis EE. High 484 concentrations of the CA-125, CA 19-9 and CA 15-3 in the peritoneal fluid between 485 patients with and without endometriosis. Arch Gynecol Obstet. 2005;271:40-5. doi: 486 10.1007/s00404-004-0645-7. 487 43. Dunselman GA, Vermeulen N, Becker C, Calhaz-Jorge C, D'Hooghe T, De Bie B, 488 Heikinheimo O, Horne AW, Kiesel L, Nap A, Prentice A, Saridogan E, Soriano D, Nelen 489 W. ESHRE guideline: management of women with endometriosis. Hum Reprod. 490 2014;29:400-12. doi: 10.1093/humrep/det457. 491 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 21 44. Rižner TL. Diagnostic potential of peritoneal fluid biomarkers of endometriosis. Expert 492 Rev Mol Diagn 2015;15:557-80. doi: 10.1586/14737159.2015. 493 45. Jiang J, Jiang Z, Xue M. Serum and peritoneal fluid levels of interleukin-6 and 494 interleukin-37 as biomarkers for endometriosis. Gynecol Endocrinol 2019;35:571-5. doi: 495 10.1080/09513590.2018.1554034. 496 46. National Center for Biotechnology Information. PubChem Compound Summary for CID 497 18218232, Phenylalanylisoleucine. Available at: 498 https://pubchem.ncbi.nlm.nih.gov/compound/Phenylalanylisoleucine; 2020. Accessed 499 July 30, 2020. 500 47. Simmen T, Nobile M, Bonifacino JS, Hunziker W. Basolateral sorting of furin in MDCK 501 cells requires a phenylalanine-isoleucine motif together with an acidic amino acid cluster. 502 Mol Cell Biol 1999;19:3136-44. doi:10.1128/mcb.19.4.3136. 503 48. Koninckx PR, Ussia A, Adamyan L, Wattiez A, Gomel V, Martin DC. Heterogeneity of 504 endometriosis lesions requires individualisation of diagnosis and treatment and a 505 different approach to research and evidence based medicine. Facts Views Vis Obgyn 506 2019;11:57-61. 507 49. Li J, Guan L, Zhang H, Gao Y, Sun J, Gong X, et al. Endometrium metabolomic profiling 508 reveals potential biomarkers for diagnosis of endometriosis at minimal-mild stages. 509 Reprod Biol Endocrinol 2018;16:42. doi: 10.1186/s12958-018-0360-z. 510 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 22 Figure 1. PLSR analysis of peritoneal fluid metabolites from women with PE and 511 controls. (2A) Score plots of the discovery set samples. (2B) Loading plots of the 512 discovery set samples. CN, control; PE, peritoneal endometriosis; PLSR, partial least 513 squares regression. 514 515 Figure 2. Volcano plots showing peritoneal fluid differential metabolites in women with 516 PE relative to controls from the discovery set. Red and blue dots represent significantly 517 different metabolites with fold change >1.5 (increased) and <0.67 (decreased), 518 respectively (p<0.05). Metabolites were identified using untargeted LC-MS/MS 519 metabolomics. LC-MS, liquid chromatography-mass spectrometry; PE, peritoneal 520 endometriosis. 521 522 Figure 3. Receiver-operating characteristic curves of significantly different metabolites 523 for peritoneal endometriosis (versus controls). Metabolites were identified using 524 untargeted metabolomics in the discovery set. AUC of 0.5 suggests no discrimination. 525 526 Figure 4. Receiver-operating characteristic curves of phenylalanyl-isoleucine for 527 peritoneal endometriosis (versus controls) using peritoneal fluid (5A) and sera (5B). 528 Phenylalanyl-isoleucine was identified using targeted metabolomics in the validation set. 529 AUC of 0.5 suggests no discrimination.530 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint 23 Supplementary data 531 Supplemental Figure 1. Flow diagram illustrating the study design. Phase II comprised of an 532 independent set of patients from phase I which was used to validate the PE metabolites. LC-533 MS/MS, liquid chromatography-mass spectrometry; PE, peritoneal endometriosis; ROC, 534 receiver operating characteristics; SDMs, significantly different metabolites 535 536 Supplemental Figure 2. Receiver-operating characteristic curves of serum cancer 537 antigen 125 (CA-125) for peritoneal endometriosis (versus controls) in women from 538 validation dataset. 539 540 Supplemental Table 1. Participants’ characteristics according to types of endometriosis in the 541 discovery and validation sets. 542 543 Supplemental Table 2. Peritoneal fluid metabolites of women with peritoneal endometriosis 544 and controls, identified using untargeted metabolomics analysis in the discovery set. 545 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted October 14, 2020. ; https://doi.org/10.1101/2020.10.13.20211789doi: medRxiv preprint