{"paper_id":"c9845376-6f65-44c7-ad3c-23f4a54858b7","body_text":"Controlled ovarian stimulation (COS) is a key step of  in vitro \nfertilization used to stimulate the development of multiple follicles, enable the\nretrieval of multiple oocytes, and thus allow the selection of the best embryo for\ntransfer. The retrieval of 5 -14 oocytes has been defined as a clinically\nappropriate ovarian response to COS ( Kyrou\n et al ., 2009 ), but poor and excessive response to\nCOS are not rare.\nThe balance between retrieving too few or too many oocytes is difficult to strike,\nand may affect the chances of taking a baby back home and even increase the risk of\npatients suffering negative health outcomes through events such as the ovarian\nhyperstimulation syndrome (OHSS) ( Kumbak  et\nal ., 2009 ).\nNevertheless, it is nearly impossible to accurately predict ovarian response and\ntailor stimulation protocols based on parameters such as maternal age, antral\nfollicle count, anti-Müllerian hormone level, or prior poor/excessive\nresponse to COS. The development of noninvasive techniques to predict response to\nCOS would allow individualized treatment, significantly increase treatment success\nrates, and alleviate the physical, emotional, and economic burden faced by\npatients.\nMetabolomics is a great tool for the comprehensive study of the dynamic changes of\nthe metabolome, and provides a powerful platform to discover biomarkers and improve\ndiagnostic and therapeutic monitoring ( Eckhart\n et al ., 2012 ;  Wang\n et al. , 2013 ;  Hertel\n et al ., 2016 ). In the field of assisted\nreproduction, most of the studies have focused on the embryonic metabolome through\nthe analysis of spent culture media ( Botros\n et al ., 2008 ;  Cortezzi  et al ., 2013 ). In addition, follicular fluid\nmetabolites have also been studied to better understand oocyte competence ( O'Gorman  et al. , 2013 ),\novarian aging ( de la Barca  et al. ,\n2017 ), endometriosis ( Santonastaso\n et al. , 2017 ), and polycystic ovarian syndrome\n(PCOS) ( Zhang  et al. , 2017 ).\nTo date, there are no studies investigating the correlation between metabolomic\nprofiling and response to COS.\nThis pilot study aimed to look into the use of serum metabolites as potential\nbiomarkers of response to controlled ovarian stimulation (COS) in patients\nundergoing intracytoplasmic sperm injection (ICSI) cycles.\n\nThis case-control study analyzed serum samples from 30 patients aged <36 years\nundergoing COS for ICSI in a university-affiliated assisted reproduction center\nfrom January 2017 to August 2017. The samples were split into three groups based\non response to COS as follows: poor responders: <4 retrieved oocytes (PR\ngroup, n=10); normal responders: ≥8 and ≤12 retrieved oocytes (NR\ngroup, n=10); and hyper-responders: >25 retrieved oocytes (HR, n=10). The\nmetabolic profiles of the serum samples were compared between the groups.\nThe patients gave informed consent prior to joining the study. The local\ninstitutional review board approved the study.\nThe patients were prescribed recombinant FSH (Gonal-F ® , Merck\nKGaA, Darmstadt, Germany) for controlled ovarian stimulation and a GnRH\nantagonist (GnRH - Cetrotide ®  Merck KGaA, Darmstadt, Germany)\nfor pituitary suppression. Follicular growth was monitored using transvaginal\nultrasound examination starting on day 4 of gonadotropin administration. When\nadequate follicular growth and serum E2 levels were observed, recombinant hCG\n(Ovidrel ® , Merck KGaA, Darmstadt, Germany) was\nadministered to trigger the final follicular maturation. The oocytes were\ncollected 35 hours after hCG administration through transvaginal ultrasound ovum\npick-up.\nThe retrieved oocytes were assessed to determine their nuclear status, and the\nones in metaphase II were submitted to ICSI following routine procedures ( Palermo  et al ., 1997 ).\nMetabolites were extracted on ice for protein precipitation using\nMethanol/Chloroform (2:1 v/v). The mixture was centrifuged at\n8,000× g  for 15 min. After centrifugation, the\nsupernatant was collected and transferred to a 96-well plate, which was placed\ninside a microTOF-QII ™  mass spectrometer equipped with an\nApollo II electrospray ion source (Bruker, Billerica, USA) coupled to a UFLC\nProminence binary liquid chromatograph (Shimadzu, Kyoto, Japan).\nThe plate was stored into the SIL-30AC autosampler at 10ºC prior to\nanalysis. The samples were directly injected into the analyzer in a 1 uL volume,\nby a 20 mmol/L ammonium formate solution in acetonitrile/2-propanol (4:1, v/v)\nat a flow rate of 200 µL/min.\nSpectra were acquired in the positive mode using a range of  m/z \n50-1200 Da. Sodium formate clusters in isopropyl alcohol within the\n m/z  50-1200 Da range were used as the calibration\nstandard.\nPatient and cycle characteristics were analyzed on SPSS Statistics 21 (IBM, New\nYork, NY, USA). Variables were tested for normality and group homogeneity using\nthe Shapiro-Wilk and Levenne tests, respectively. When needed, the samples were\nstandardized using the z-score.\nThe variables were compared between groups through one-way ANOVA, followed by the\nBonferroni post-hoc test. Variables were described as mean values ±\nstandard deviation and significance was attributed when α was 5%.\nMass spectrometry results were obtained using the DataAnalysis 4.1 sofware\n(Bruker Daltonics Bremen, Germany) and data analyses were performed on\nMetaboAnalyst 3.0 ( http://www.metaboanalyst.ca ).\nLog 2  was used to normalize intensity values followed by self-scaling.\nPrincipal component analysis (PCA), an unsupervised method, was applied to the\ndata set to detect intrinsic clusters based on metabolic profiles. From these\nanalyses, a list of ions responsible for group discrimination was obtained.\nThese ions were used to build a receiver operating characteristic (ROC) curve and\nto evaluate the strength of the model at predicting response to COS.\nMetabolite attribution was performed based on the Human metabolites database\n( http://www.hmdb.ca/ ), with a maximum mass tolerance of 0.01 Da.\nThe maximum mass error was 30 ppm. For the attribution, only molecules\ncontaining hydrogen (M+H + ), sodium (M+Na + ) and potassium\n(M+K + ) as adducts were considered.\n\nThe patient and cycle characteristics are described in  Table 1 . As expected, the level of estradiol on hCG trigger\nday, the number of aspirated follicles, the number of retrieved follicles, and\nthe number of mature follicles were higher in the HR group followed by the NR\ngroup, while the PR group presented the lowest results.\nPatient and cycle characteristics for the poor, normal, and\nhyper-responder groups\nBMI= body mass index.\n(one way ANOVA followed by Bonferroni post hoc test,\n p <0.05)\nConsidering components 1 and 2, PCA clearly distinguished between the PR, NR, and\nHR groups ( Figure 1 ).\nFigure 1 Variance among groups according to principal component analysis (PCA\nscore plot)\nVariance among groups according to principal component analysis (PCA\nscore plot)\nThe ion masses associated to the separation of the groups were obtained by the\nPCA loading plot, from which 10 ions were chosen as potential biomarkers ( Table 2 ).\nTentative biomarker attribution, loading values, and predictive values\nfor high and poor responders\nThe relative abundances of individual biomarkers are demonstrated in the box plot\ncharts ( Figure 2 ). The PCA loading values\nfor the selected ions based on principal component 1 are presented in  Table 2 .\nFigure 2 Boxes plot charts for the relative abundances of individual\nbiomarkers\nBoxes plot charts for the relative abundances of individual\nbiomarkers\nOur evidence demonstrated that all ions selected in the present study were more\nspecific for the PR group when compared to the NR group. The ROC curve\nconsidering the PR and NR groups presented an area under the curve (AUC) of\n99.6% (95% CI: 88.9 - 100%,  Figure 3 ).\nFigure 3 ROC curve and predicted class probabilities for normal and poor\nresponders\nROC curve and predicted class probabilities for normal and poor\nresponders\n\nThe success of assisted reproductive technology (ART) treatments is highly dependent\non the response to COS. Moreover, considering the physical, emotional, and economic\nburden faced by patients undergoing ART treatment, finding a method to predict\novarian response to COS would be a major step forward in the field of reproductive\nmedicine.\nThis study was able to identify metabolites that might be used as predictive\nmolecular markers of response to COS. PCA clearly distinguished between PR, NR and\nHR groups, and 10 ions were chosen as potential biomarkers of response to COS. To\ndate, this is the first study to investigate whether blood plasma metabolites might\npredict ovarian response to COS. The advantage of the method used in this study is\nhaving knowledge of the patients’ ability to respond to gonadotropins before the\nstart of stimulation. Previous metabolomic studies in assisted reproduction\ninvestigated the metabolic profile on seminal plasma ( Deepinder  et al. , 2007 ) and fluid ( Gupta  et al ., 2011 ), embryo\nspent culture media ( Cortezzi  et\nal ., 2013 ), follicular ( O'Gorman\n et al ., 2013 ) and endometrial fluid ( Braga  et al ., 2017 ).\nOMICS technologies study cellular events and interactions from deoxyribonucleic acid\n(DNA) and genes to metabolites in a global way. Although metabolomics was recognized\nas a separate area of science much later than the other “omics” - such as genomics,\ntranscriptomics, and proteomics - it provides a powerful platform for the discovery\nof novel biomarkers. The main advantages of metabolomics are its familiarity to the\nactual phenotype and the number of possible low molecular weight bio-compounds\n( Yoshida  et al .,\n2012 ).\nIn conclusion, our preliminary evidence suggests that serum metabolites might work as\npredictive molecular markers of ovarian response to controlled stimulation. We\nbelieve that the quality of emerging data will eventually allow the\nindividualization of COS, and further studies will validate new biomarkers of\novarian response. To date, the technology and software around metabolomics are still\ndeveloping as the human metabolome is being mapped. The integration of clinical and\n“omics” findings will eventually allow the migration toward an era of personalized\ntreatment in the field of reproductive medicine.","source_license":"CC-BY-4.0","license_restricted":false}