Application of specific ELISAs for BMP15 and GDF9 to cumulus cell extracts from infertile women

Bone morphogenetic protein 15 (BMP15) and growth differentiation factor 9 (GDF9) are oocyte-specific paracrine factors which regulate ovarian cumulus cell (CC) functions. This study aimed to investigate if BMP15 and GDF9 bound to CCs can be characterized, quantified


Introduction
Two transforming growth factor β (TGFβ) superfamily members, bone morphogenetic protein 15 (BMP15) and growth differentiation factor 9 (GDF9), are oocyte paracrine signaling factors essential to folliculogenesis.Studies in animal models and humans have led to the concept that these oocyte-secreted growth factors are key contributors to determining mammalian ovulation rate and fecundity (McNatty et al., 2004).Human studies suggest that different infertility pathologies are associated with mutations in BMP15 and GDF9.Inactivating mutations in BMP15 and GDF9 have been identified in women with premature ovarian insufficiency (Dixit et al., 2006;Laissue et al., 2006;Tiotiu et al., 2010), many of which affect protein expression and stability (Simpson et al., 2014;Patino et al., 2017).Rare mutations in GDF9, but not in BMP15, are associated with human dizygotic twinning (Palmer et al., 2006;Zhao et al., 2008).It has also been suggested that their expression and protein levels are altered in women with polycystic ovary syndrome (PCOS) (Teixeira Filho et al., 2002;Wei et al., 2011;de Resende et al., 2012;Riepsamen et al., 2019;Kristensen et al., 2022).When added exogenously to in vitro oocyte cultures, BMP15 and GDF9 improve oocyte quality as measured by subsequent preimplantation embryo development and pregnancy rates (Hussein et al., 2006;Sudiman et al., 2014;Sugimura et al., 2014Sugimura et al., , 2015)).
Intriguing insights into the regulation of fertility by BMP15 and GDF9 come from animal studies where mutations in the genes encoding BMP15 and GDF9 affect fertility potential in a dose-dependent and species-specific manner.Gdf9 null female mice are infertile due to a block in folliculogenesis at the primary follicle stage (Dong et al., 1996), while Bmp15 null female mice only have a mild fertility defect with reduced ovulation and fertilisation rates (Yan et al., 2001).By contrast, naturally occurring BMP15 mutations in ewes can lead to an increased ovulation rate or infertility, in a gene-dosage dependent manner (Galloway et al., 2000).Ewes with homozygous mutations in GDF9 or BMP15 are sterile while ewes heterozygous for mutations in both genes show greater fecundity due to an increased ovulation rate (Hanrahan et al., 2004).These and other studies have led to the hypothesis that the relative ratio of oocyte-secreted GDF9 and BMP15 determine mammalian ovulation rate and fecundity (Crawford and McNatty 2012; Christoforou and Pitman 2019), whereby GDF9 and BMP15 are both required for the low ovulation rate phenotype in mammals such as humans (McNatty et al., 2004).
Based on these observations that BMP15 and GDF9 have major physiological roles in regulating female reproductive potential, and that they are secreted predominantly or possibly exclusively by the oocyte (Aaltonen et al., 1999;Christoforou and Pitman 2019), it suggests that they represent important diagnostic and therapeutic targets in reproductive medicine.To explore their potential clinical utility, the logical biological samples to target are spent oocyte/embryo-conditioned media from IVF (Li et al., 2022) and/or discarded cumulus cells (CCs) from intracytoplasmic sperm injection (ICSI), which are known to capture oocyte-secreted BMP15 (Guéripel et al., 2006;Mester et al., 2015) and CCs are the principle site of biological action of the oocyte-secreted factors (Gilchrist et al., 2008).However, there are few methods available to accurately and reliably measuring these proteins.This is attributed to the complexity and low abundance of the biological samples of interest and a lack of both high affinity specific monoclonal antibodies and reliable commercial assays.This laboratory has developed BMP15 and GDF9 ELISAs for measurement in human serum (Riepsamen et al., 2019(Riepsamen et al., , 2021) ) and in this study these ELISAs were adapted to measure BMP15 and GDF9 in human CC extracts.
The aim of this study was to develop sensitive and specific ELISAs for the measurement of BMP15 and GDF9 on human CCs and examine if these growth factors are associated with clinical outcomes and pathologies associated with infertility.It was hypothesized that oocytesecreted BMP15 and GDF9 are altered in different infertility pathologies and is associated with embryo development.

Study approval
Approval of this study was granted by the IVF Australia Human Research Ethics Committee (R&D-109).This was a retrospective study with samples collected at a single centre at IVF Australia, Sydney, Australia.

Clinical cohorts
The samples from study cohort 1 (120 patients) were collected between September 2017 to April 2018, and study cohort 2 (n = 81) between February 2018 and October 2019.The collection of samples from both cohorts followed the same protocol.As this was a retrospective study, no exclusion criteria were applied.Participants in the study included intracytoplasmic sperm injection (ICSI) patients (n = 191), egg donors (n = 3) and patients having elective oocyte cryopreservation (n = 7), hence totaling 201 patient samples for most analyses, but 194 samples for the embryology data set after excluding oocyte cryopreservation cases.IVF patients were not included as their ovarian CCs were not available for collection on the day of OPU.The following patient information was obtained: age, pelvic ultrasound assessment with antral follicle count (AFC) noted in an unstimulated cycle, previous history of endometriosis diagnosed by laparoscopy, superovulation protocol (antagonist or agonist cycle), ovulatory trigger type (agonist or hCG), treating clinician at oocyte pick-up (OPU), number of oocytes retrieved and embryo assessment results.

Ovarian stimulation and collection of CCs and media
IVF treatment drugs, doses and duration of stimulation as well as patient monitoring were chosen by the treating physician as per standard clinical practice.Ovarian stimulation was performed using one of a range of commercially available FSH preparations: Menopur (Ferring Pharmaceuticals, Switzerland), Gonal F (Merck Serono, Germany), Puregon (MSD, NJ, USA), Bemfola (Gedeon Richter, Hungary) or Elnova (MSD), with doses ranging from 75 to 450 IU/day.Gonadotrophinreleasing hormone (GnRH) agonist or antagonist (Synarel, Pfizer, NY, USA, or Orgalutran, MSD, or Cetrotide, Merck Serono) was coadministered to prevent the endogenous luteinising hormone (LH) surge.Follicular growth was monitored with regular ultrasound and serum estradiol measurements.Final oocyte maturation was triggered with recombinant chorionic gonadotrophin (Ovidrel, Merck Serono), or recombinant LH (Luveris, Merck Serono) or synthetic GnRH agonist (Lucrin, Abbott, Il, USA; or Decapeptyl, Ipsen, France or Pregnyl, MSD).Cumulus-oocyte-complexes (COCs) were retrieved by transvaginal ultrasound-guided puncture of follicles 34-36 h post trigger depending on the trigger type used.COCs were collected and kept prior to denudation in GT-L medium (2-5 mL, Vitrolife, Sweden) pre-equilibrated prior to use (6% CO2 and 37 • C).Oocytes were denuded of CCs 2-4 h post oocyte retrieval by mechanical pipetting in Hyase (40 IU/ml, Vitrolife) in GT-L medium (Vitrolife) pre-equilibrated prior to use.The CCs including corona cells were collected and pooled from all of an individual patient's oocytes.The cells were dislodged from the bottom of the denuding well by vigorous pipetting.Cells and media were transferred to a microcentrifuge tube and centrifuged at 5000 g for 2 min at room temperature.The media was removed, and the cell pellet was snap frozen in liquid nitrogen.Spent oocyte conditioned media (OCM) samples were collected for all individual patients and snap frozen.All samples were initially stored at − 30 • C and then moved within 3 weeks for storage at − 80 • C for up to 12 months.Effect of storage duration on sample stability was assessed.

Cumulus cell extraction
A cell extraction method was developed to simultaneously extract protein and DNA from GCs/CCs to enable measurement of proteins and DNA from the same sample for relative quantitation.The frozen CC samples were thawed and treated with extraction buffer (50 mM phosphate buffer pH 7.5, 0.2-2M NaCl, 1 mM PMSF) with maximum extraction of DNA at 1.5M NaCl (Fig. 1D).For routine CC extractions, 1.5M NaCl in extraction buffer was employed.The samples were vortexed three times for 10 s every 10 min at room temperature.The samples were diluted with extraction buffer, with volumes based on the number of oocytes collected.Samples were centrifuged for 30 min at 16,900 g at 4 • C. The supernatants were immediately used for the measurement of DNA and BMP15 (cohorts 1 and 2) as well as GDF9 (cohort 2).

Granulosa cell isolation
For use in assay validation and as quality control (QC) samples and reference preparations, granulosa cells (GCs) were obtained from pooled follicular fluid (FF) of patients following transvaginal oocyte retrieval and isolated based on the methods of (Ferrero et al., 2012;Chang et al., 2013).In brief, FF was filtered through 70 μm cell strainers (Corning, NY, USA).Cells were then fractionated on a Ficoll-Paque column (Premium 1.084, GE Healthcare, IL, USA) and centrifuged at 1000 g for 25 min.The cell interphase containing GCs was collected, washed and centrifuged at 5000 g for 2 min.Supernatant was removed, cell pellet snap-frozen and stored at − 80 • C. A single large GC pool was obtained by combining GC pellets from multiple patients that were not necessarily the same patients contributing CC samples but who were also undergoing infertility treatment at the same clinic over the same period.The GC pellets were extracted with high salt buffer as described above for CC extracts, snap frozen and stored at − 80C in aliquots.This preparation, designated as hGC-RP, was used as reference preparation for BMP15 and GDF9 ELISAs.Use of hGC-RP as a reference preparation was based on the information that the Ficoll-Paque gradient method removes the bulk of contaminating cells, and that the ratio of GDF9:BMP15 in CC extracts using GC extracts as the ELISA standard (hGC-RP) resulted in a ratio of ~1, which means that the dual composition of GDF9 and BMP15 in CC extracts was similar to that of the hGC-RP.Hence the test and reference preparations were similar in composition.QC samples were prepared from CC and GC samples as above, aliquoted, snap frozen and kept at − 80 • C.

DNA fluorometric assay
As a measure of number of CCs present per patient extract, the DNA concentration in CC extracts (CC DNA) was determined using a fluorometric assay based on the method of (Calafiore et al., 1991) using the fluorescent probe Hoechst-33342.The DNA reference preparation (calf thymus DNA (Invitrogen)) was prepared in the extraction buffer, snap frozen and stored at − 80 • C in aliquots.A separate GC extract was prepared and used as quality control (QC) samples.The DNA assay was undertaken in 96 well polystyrene flat bottom microplate (Corning, New York, USA).The DNA standard, QC and patient samples (150 μl/well), followed by Hoechst 33342 (50 μl/well, 4 μg/mL) were added to the plate.Fluorescence was read immediately at 350 nm and 455 nm excitation and emission wavelengths, respectively.CC DNA concentrations were calculated using a third order polynomial (cubic) transformation of the DNA standard curve (GraphPad Prism).The sensitivity of the DNA assay (23 ng/ml) was defined as the dose of the reference preparation corresponding to 2 SD above that of the assay blank value.The intra-assay variation was the average CV for clinical samples assayed in duplicate from five experiments (0.8%).The inter-assay variation was calculated from five experiments using the GC QC samples (mean CV = 4.6%).

BMP15 and GDF9 ELISAs
BMP15 and GDF9 ELISAs were based on our previous study (Riepsamen et al., 2019) and adapted to measure BMP15 and GDF9 in CC and GC extracts.Both ELISAs are based on a dual antibody system using one antibody as capture and the second biotinylated antibody as tracer, and streptavidin-horse radish peroxidase as detection system.Buffers, reagents and incubation times were similar to that reported (Riepsamen et al., 2019).The extraction of GDF9/BMP15 immunoactivity from CC samples was initially assessed using a range of salt concentrations (0.2-2M NaCl), showing maximal extraction of DNA in 50 mM phosphate buffer pH 7.5, 1.5M NaCl, 1 mM PMSF.The CC extract was then diluted with assay buffer +0.5M NaCl for a final salt concentration of 1.0M NaCl prior to assay.Immunoactivity of CC samples peaked at 1.0M NaCl (Fig. 1C) and this salt concentration was routinely used in both ELISAs.Absorbance was determined at 450 nm.BMP15 and GDF9 concentrations were calculated using cubic spline and polynomial transformations of the hGC standard curves (GraphPad Prism) depending on the ELISA type.An assessment of non-specific binding was undertaken by investigating the binding of CC and GC extracts in the ELISAs in the absence of a coated antibody.
Specific and non-specific binding were investigated in a study of CC from 20 patients covering a large range (6-31) of oocytes collected.Regression analysis of these samples in the BMP15 ELISA showed a close relationship (correlation coefficient r = 0.95) with a slope of 12.4 ± 3.4% indicating a constant level of non-specific binding per sample.The corresponding response of non-specific vs total binding in the GDF9 ELISA showed a close relationship (r = 0.97) with a non-specific value of 8.5%.Exploring other blocking agents (different preparations of BSA, skim milk, casein hydrolysate, casein from bovine milk, gelatin from cold water fish skin, a commercial blocking agent (Pierce Protein-Free Blocking Buffer, Thermo Fisher Scientific)) showed no improvement.As the proportion of non-specific signal was constant for all CC extracts, the CC sample BMP15/CC DNA and GDF9/CC DNA values were not corrected for these non-specific (12.4% and 8.5%, respectively) values.This background is constant across all test samples so does not affect the validity of measurements.
For stability studies, a high salt GC extract, prepared in a similar manner to the hGC-RP, was also used as a QC in both ELISAs and showed no loss of immunoactivity covering the duration of evaluation of the two cohorts (~1 year).The hGC-RP was expressed with common assigned arbitrary unitage (au) in both ELISAs.The hGC-RP was used as reference preparation with a common assigned arbitrary unitage in both ELISAs.A.E. Krysta-Matter et al.Calculated BMP15/GDF9 levels from patient CC samples were normalized to CC DNA concentration and expressed as arbitrary units per μg DNA (au BMP15/μg CC DNA, au GDF9/μg CC DNA).

BMP15/GDF9 ELISA validation
The ELISA sensitivity was defined as the dose of the reference preparation corresponding to 2 SD above that of the assay blank value.Samples which measured less than the sensitivity value were considered undetectable and given BMP15 and GDF9 values equal to the sensitivity value.The specificities of the assays were examined using recombinant preparations of human pro-BMP15, GDF9, BMP5, BMP6, BMP7 and mouse GDF9.The BMP15 and GDF9 ELISA validation criteria in terms of sensitivity, specificity, intra-and inter-assay variation are summarized in Table 1.
The BMP15 specific (28A) and non-specific (OVA) immunoneutralising peptides were added in increasing molar excess amounts (0-56 μg/mL) to biotinylated mAb 28A and preincubated 2.5 h at room temperature prior to inclusion in the ELISA.Neutralisation was assessed using pro-BMP15 and a GC extract.A corresponding set of samples was assessed in the ELISA without the coating antibody for assessment of non-specific binding.Similar studies using GDF9 specific peptides to neutralise GDF9 antibodies in the GDF9 ELISA were unsuccessful owing to elevated backgrounds observed at high peptide concentrations.

BMP15 in oocyte conditioned media (OCM)
Because oocyte-secreted BMP15 and GDF9 could be secreted by cumulus-oocyte complexes (COCs) into the oocyte collection medium, levels of BMP15 in OCM were investigated in spent collection media, as described above (see Collection of CCs and media).Given the high ratio of collection medium to COC and in an attempt to obtain sufficient material, OCM from 4 to 5 separate patients were pooled and then concentrated 7-22 fold (given the high medium:COC ratio in the OCM) using the Vivaspin 20 concentration columns (GE Healthcare, IL, USA) prior to assay in the BMP15 ELISA, using the ELISA protocol described above.GDF9 levels were not assessed.As this was a proof-of-principal experiment that did not detect BMP15 (see Results), there was no prior patient selection.

Western blot PAGE
Prior to Western blotting, GC extracts were processed to reduce protein load and reduce the presence of biotinylated proteins in order to minimize non-specific binding from use of the streptavidin detection system used in the Western blots.BMP15 present in GC extracts was partially purified by hydrophobic interaction chromatography (HIC), as previously undertaken for the partial purification of GDF9 (Hickey et al., 2005).GC extracts were filtered (0.45 μm filter; Merck Millipore, Ireland) to remove particulate matter and passed through HIC Phenyl Sepharose cartridges (HiTrap Phenyl FF 1 mL, GE Healthcare, Sweden).Proteins were eluted in 50 mM potassium phosphate buffer pH 7.5, 0.154 M NaCl and precipitated with trichloroacetic acid.The pellet was washed with cold acetone, centrifuged, resuspended in RIPA buffer and stored at − 80 • C. For GDF9, the biotinylated proteins in the GC extracts were removed by incubating the GC extract with streptavidin-coated magnetic beads (Myone Streptavidin C1 Dynabeads, ThermoFisher Scientific, Vilnius, Lithuania).
For BMP15 and GDF9 Western blots, recombinant proteins and processed GC extracts were mixed with Bolt LDS sample buffer and Bolt sample reducing agent (Life Technologies), then boiled at 95 • C for 5 min.Samples were loaded onto Bolt 4-12% Bis-Tris Plus Gel in Bolt MES SDS running buffer in the presence of Bolt Antioxidant (Life Technologies) and proteins separated by electrophoresis at 100V for 2h.Proteins were then transferred overnight onto 0.45 μm nitrocellulose membrane (Thermo Fisher Scientific) and blocked using a blocking buffer (3% BSA in Tris-buffered saline buffer with 0.05% Tween-20) for 3.5 h with shaking at room temperature.The detection antibodies (BMP15, biotinylated mAb#28A; GDF9, biotinylated mAb#72B) were used in blocking buffer, with or without pre-incubation with 150 μg of their respective specific immuno-neutralising peptides for 2.5 h at room temperature.The membrane was incubated overnight at 4 • C with biotinylated-mAb with or without prior peptide neutralisation, washed, and incubated for 1 h at room temperature with streptavidin-HRP (SNN2004, Life Technologies).The bands were detected with ECL Prime (Amersham, GE Healthcare) and imaged with the ImageQuant LAS 4000 (GE Healthcare).The lanes were then scanned using ImageJ 1.53c (a public domain Java image processing program).Molecular size of bands was compared to the molecular weight marker SeeBlue Plus2 pre-stained standard (Invitrogen, Carlsbad, CA, USA).The molecular weights of the immunoactive bands were determined by comparing their relative membrane location with the relative membrane location of the protein marker standards.A cubic spline-fitting transformation of the relative membrane location of the protein markers in relation to their known molecular weights was used to derive a calibration curve from which the molecular weights of the immunoactive bands were determined (GraphPad Prism version 7, GraphPad Software, La Jolla, CA).

Statistical analysis
Statistical analyses were performed using GraphPad Prism or SPSS version 25 (IBM Corp., Armonk, NY, USA).All data were assessed for normality using the Shapiro-Wilcox normality test.Pearson correlation and independent samples t-tests were applied to normally distributed data.For data that were not normally distributed, Spearman correlations were applied.Continuous variables were described using means and standard deviations.In cases where data was not normally distributed, non-parameteric analyses were performed and the data are presented as medians and interquartile range.Categorical variables are described using frequencies and percentages.Univariate linear regression was used to determine the relationship between DNA, BMP15, GDF9, the number of oocytes and age.Multivariate linear regression was used to assess the relationship between BMP15/CC DNA, GDF9/CC DNA and demographic and clinical predictors.Backwards stepwise regression was used, with a univariate threshold of p < 0.25 for inclusion in the multivariate model.Statistical significance was determined by p < 0.05.All standard assumptions of linear regression were tested.

BMP15 and GDF9 ELISA validation
BMP15 and GDF9 ELISAs were applied to 1.5M NaCl extracts of ovarian cumulus and granulosa cell preparations, demonstrating dosedependent increases in GC and CC extracts and respective recombinant proteins (Fig. 1A and B).However, a lack of parallelism was noted between dose-response curves of recombinant pro-BMP15 (as well as E. coli mature BMP15, data not shown) and both CC and GC extracts (Fig. 1A), making either BMP15 preparation unsuitable as reference preparations.Similarly, non-parallelism was observed between mature GDF9 and both CC and GC extracts in the GDF9 ELISA (Fig. 1B), whereas dose response curves of CC and GC extracts were parallel to each other over most of the dose range in both ELISAs (Fig. 1A  and B).Simultaneous extraction of BMP15 and DNA from CC samples was performed using increasing NaCl concentrations (Fig. 1C and D) with maximum extraction for both BMP15 and DNA achieved at 1.5M NaCl (1M NaCl final concentration in the ELISA).Given the close molecular similarities of BMP15 and GDF9, these same conditions were applied in the assay of GDF9, using the same samples as BMP15.
In the BMP15 ELISA, neutralisation of biotinylated 28A with specific peptide (28A) at the highest molar excess, resulted in a signal suppression of the recombinant pro-BMP15 and GC extract by 100% and 92%, respectively, while the non-specific peptide (OVA) had no significant effect.In the GDF9 ELISA, neutralisation of mAb 72B, or biotinylated mAb 53/1 or both mAbs, was problematic in the GDF9 ELISA of CC extracts as these blocking peptides alone contributed to significant background signals attributed to the interaction of the blocking peptide with the CC extract.Decrease in signal was expressed as percentage reduction from sample signal in the absence of an antigen peptide.Signal of recombinant GDF9 was reduced in a dose dependent manner with specific 72B and 53/1 peptides resulting in signal reductions of 99.7% and 100%, respectively.However, the greatest reduction in the signal of GC extract was only 17.3% or 50%, respectively, as high concentrations of the blocking peptides caused excessive background signal, interfering in the ELISAs.The non-specific peptide (OVA) had no effect of the signal of recombinant GDF9 or GCs in the GDF9 ELISA.
BMP15 levels detected in concentrated OCM pools were less than the limit of quantification of the BMP15 ELISA.Based on the most concentrated pool of OCM, the proportion of BMP15 in OCM was <3% of the total BMP15 in its matching CC sample.Hence, the presence of BMP15 in the OCM was considered negligible.The measurement of GDF9 in OCM pools has not been assessed.BMP15 was undetectable in 16/201 CC samples across both study cohorts and for GDF9, 3/81 CC samples were undetectable.To examine the potential effect of CC sample storage duration on loss of immunoactivity, the 120 CC BMP15/μg DNA values (cohort 1) were plotted against storage time and showed no change in the slope of the regression line of average BMP15/DNA versus time (data not shown).These results indicate that there is little evidence of storage losses.

BMP15 antibody specificity and western blot analysis
The peptide sequence of pro-mature human BMP15 and the location of the peptide used to raise mAb 28A are presented in Fig. 2A.The RRTR cleavage site generates the 16-19 kDa mature domain and 33 kDa prodomain.Also shown are two other putative cleavage sites in the prodomain, as well as three potential (Saito et al., 2008) N-linked glycosylation sites in the pro-domain and one in the mature domain, and an O-linked glycosylation site in the region of the putative mAb 28A antibody binding epitope (Fig. 2A).Fig. 2B and C shows Western blot analyses using biotinylated mAb#28A which detects the 19 kDa mature domain of rec-hBMP15 (R&D) and a 30 kDa band, which is believed to be an aggregate of mature BMP15, as GDF9 and BMP15 are known to aggregate (McIntosh et al., 2008).Neutralisation of biotinylated mAb#28A with a 28A-specific peptide eliminated detection of the 19k and 30k forms of recombinant BMP15 (Fig. 2B and C).Western blot of GC extracts post-HIC showed multiple immunoreactive bands at 19 kDa, 41 kDa, 63 kDa, 83 kDa and 105 kDa, of which the 39 and 43 kDa bands (indicated in Fig. 2C as red stars) were fully neutralised and immunoactivity at 19 kDa was heavily suppressed (Fig. 2C).The 60-70 kDa region (Fig. 2B and C) was not neutralised, which is attributed to the presence of biotinylated proteins, most likely biotinylated carboxylases present in the GC extracts.

GDF9 antibody specificity and western blot analysis
In Fig. 2D, the peptide sequence of pro-mature human GDF9 and the location of the peptides used to raise mAbs #72B and #53/1 are presented along with the RHRR cleavage site separating the 16-17 kDa mature domain and the 33 kDa pro-domain, and a putative cleavage site at amino acid residue 229.Also shown are five predicted N-linked glycosylation sites in the pro-domain, as well as one in the mature domain in the region of the putative mAb 72B antibody binding epitope (Fig. 2D).Western blot analysis of GDF9, pro-GDF9 and a GC extract and following incubation of mAb#72B with a neutralising peptide are presented (Fig. 2E).The GC extract was partially purified by removal of biotinylated native proteins using streptavidin-coated beads (see Methods Section).Molecular weight bands were observed (Fig. 2E) and their size and abundance quantified (Fig. 2F), demonstrating the prominent 21 kDa mature domain of GDF9 (R&D), and both the mature 18 kDa domain of the processed protein and the 62 kDa unprocessed pro-protein of recombinant pro-GDF9 (Fig. 2E and F).The 40 kDa band in the R&D GDF9 sample is believed to be the aggregate of two mature GDF9 domains (McIntosh et al., 2008).In terms of analysis of the GC extract, prominent GDF9 molecular mass forms were identified at 40 kDa, 147 kDa and 163 kDa, with notably less abundant bands at 44 and 63 kDa and little evidence of mature peptide (Fig. 2E and F).The GC 63 kDa band is consistent with the unprocessed pro-protein form of GDF9.The 40 kDa, 45 kDa bands are either dimer aggregates of the mature GDF9 or more likely a processed form following cleavage at amino acid residue 229 as postulated.Pre-absorption of biot-mAB#72B with the 72B specific peptide resulted in a marked suppression of all band intensities (Fig. 2E, red stars; Fig. 2F), in the three samples examined.The residual binding is attributed to the presence of remaining biotinylated proteins in the extract.The large molecular weight species (147k, 163 kDa) are unknown, but are neutralised by the neutralising peptide, and hence are likely to be aggregates or multimers of GDF9 pro-and mature domains, as previously reported (McIntosh et al., 2008).

Application of BMP15 and GDF9 ELISAs to patient CC samples
BMP15 and DNA concentrations were determined in two study cohorts (cohorts 1 and 2) consisting of 120 and 81 patients, respectively, using pooled CCs from individual patients.The median age at OPU across both cohorts was 38.0 with 9.0 oocytes retrieved/patient (Supplementary Table 1).The two cohorts were collected on average 6 months apart, analysed separately and then combined.Identical protocols of CC collection, processing and BMP15 and DNA measurements were undertaken in both cohorts.Descriptive characteristics and comparisons between the two cohorts are provided in Supplementary Table 1.Significant differences in BMP15/oocyte (198 vs 419 au/ oocyte) but not DNA/oocyte (5.5 vs 5.4 μg/oocyte) were noted between cohorts 1 and 2, respectively (Supplementary Table 1).This difference could not be attributed to technical limitations in the BMP15 ELISA (e.g.stability of the hGC-RP) or to COC collection and processing protocols (note the similarity in DNA/oocyte between cohorts) or to clinical treatment protocols (e.g.stimulation or trigger hormones used; Supplementary Table 1).Nonetheless, correlations between BMP15 and oocyte numbers (and CC DNA vs oocyte number, and BMP15 vs CC DNA) were highly comparable and equally significant, between cohorts 1 and 2, permitting a combined analysis of the cohorts (Supplementary Table 2).
When total CC BMP15 per patient for the combined cohorts were matched with corresponding oocyte numbers (Fig. 3A, Supplementary Table 2) a significant correlation (r = 0.61, p < 0.001, Fig. 3A) was observed, with a wide spread of BMP15 values which is likely due to the inherent relationship between total number of CCs collected and oocyte Each membrane was probed without or with immuno-neutralised mAb#28A using a specific neutralising peptide.Panels C,F: Quantification of WB grey scale intensity and molecular weight calibrated to standard curves, without (solid lines) and with (dotted lines) antibody peptide neutralisation.Numbers on peaks correspond to the calculated molecular weights of bands.Panel E: Western blots using biotinylated mAb#72B for detection of human GDF9, pro-GDF9 and GC extracts.Each membrane was probed without or with immuno-neutralised mAb#72B using a specific neutralising peptide.Red stars depict the bands which were fully neutralised in the GC extract samples.number.To account for variable number of CCs/oocyte, DNA levels were measured by a fluorescence assay in the salt extracts.Although a relatively strong relationship was observed between CC DNA and number of oocytes (r = 0.69, p < 0.0001, Fig. 3E), there was nonetheless a large variance between patients in the number of pooled CCs collected.However, when BMP15 levels in CC pellets were matched to their corresponding DNA levels (i.e.cell content) a notably tighter relationship (r = 0.92, p < 0.0001, Fig. 3C) between BMP15 and CC DNA content was noted.Thereafter expressing BMP15 levels in CC extracts as BMP15/CC DNA was employed across the clinical study.
CC GDF9 levels were determined in cohort 2 only.As observed with BMP15, a limited correlation (0.67, p < 0.001, Fig. 3B) was observed between CC GDF9 and oocytes.The correlation improved notably (r = 0.96, p < 0.001, Fig. 3D) when CC GDF9 was correlated with CC DNA levels.Hence similarly, GDF9 levels in CC extracts were expressed in terms of its DNA content (GDF9/CC DNA).
An interesting observation was made when cohort 2 GDF9/CC DNA and BMP15/CC DNA values were plotted as a regression line (Fig. 3F).A highly significant correlation coefficient (r = 0.90, p < 0.001) was observed with an equation of the regression line y = 1.02x + 24.9.The slope of this regression line (1.02) indicates that, on average the GDF9: BMP15 ratio in the CC samples is similar to that found in the hGC-RP.To re-express this relationship in terms of BMP15 and GDF9 levels in the hGC-RP, the levels of BMP15 and GDF9 in the hGC-RP were estimated in their respective ELISAs based on a comparison of average values equivalent to their ED50 values, as hGC-RP and BMP15 and GDF9 protein standards are not parallel in their respective ELISAs.For the GDF9 ELISA, one au hGC-RP = 3.33 ± 10.4 pg GDF9 (R&D) (n = 5 expts; mean ± SD, CV 12%) and in the BMP15 ELISA, one au hGC-RP = 1.70 ± 0.27 pg pro-BMP15 (n = 3; CV 16%).Thus, when adjusting the slope of the regression line between GDF9/CC DNA and BMP15/CC DNA in terms of BMP15 and GDF9 levels in the hGC-RP, the slope value is 2.0.Hence the GDF9:BMP15 ratio on CCs is ~2:1.

Patient factors contributing to differing CC BMP15 and GDF9 concentrations
To identify patient factors contributing to the variance (6.1-fold) in BMP15/CC DNA and GDF9/CC DNA levels between patients, correlations, univariate and multivariate linear regression analyses were conducted.Regression analysis showed a significant direct association between BMP15/CC DNA and oocyte number (Table 2, p = 0.011; Fig. 4A) and likewise GDF9/CC DNA and oocyte number (Table 2, p = 0.018, Fig. 4B).BMP15/CC DNA (Table 2, p = 0.051, Fig. 4D), but not GDF9/CC DNA (Table 2; Fig. 4E), showed a borderline decrease with age.The GDF9:BMP15 ratio did not change with increasing patient age (Fig. 4F), or number of oocytes collected (Fig. 4C), despite in the case of the latter both BMP15 and GDF9 levels increasing with higher oocyte number.Hence, patients with more oocytes had more CC-bound GDF9 and BMP15, such that the relationship between GDF9 and BMP15 expressed as a ratio remained unchanged (Fig. 3F).

Minimal correlation of BMP15/CC DNA and GDF9/CC DNA with other patient characteristics or embryo outcomes
BMP15/CC DNA levels were not associated with a history of endometriosis and BMP15/CC DNA and GDF9/CC DNA were not associated with whether patients received antagonist or agonist treatment or were given agonist or hCG triggers (Table 2 and Supplementary Table 1).Likewise, BMP15/CC DNA and GDF9/CC DNA were not associated with oocyte meiotic arrest (%GV, germinal vesicle) or the matching oocytes' capacities to progress to metaphase II (%MII), be fertilized (% 2 PN, 2 pronuclei), progress to day 3 or day 5 (blastocyst) of embryo development (Table 3 and Supplementary Figs. 1 and 2).There was a tendency for BMP15/CC DNA to predict the number of blastocysts (p = 0.065; Table 3, Supplementary Fig. 1), but not euploid blastocyst number or % euploid embryos (Table 3, Supplementary Fig. 1).

Discussion
This study outlines the development of specific ELISAs for GDF9 and BMP15 applied to cumulus cell extracts from women with differing infertility pathologies.Sufficiently sensitive and specific ELISAs have only recently been developed to measure BMP15 and GDF9 in follicular fluid (Kristensen et al., 2019), serum (Riepsamen et al., 2019(Riepsamen et al., , 2021(Riepsamen et al., , 2023) ) and embryo culture medium (Cadenas et al., 2022;Li et al., 2022).As it is known that oocyte-secreted BMP15 and GDF9 are captured by CCs (Guéripel et al., 2006;Watson et al., 2012;Mester et al., 2015), in this study we adapted our ELISAs developed to measure BMP15 and GDF9 in serum (Riepsamen et al., 2019(Riepsamen et al., , 2021)), to detect these in pooled CCs from individual patients undergoing ICSI.It was found that CC-bound BMP15 and GDF9 are highly correlated, BMP15 is negatively associated with maternal age, and that neither BMP15 nor GDF9 are associated with embryology outcomes.It is recognized that the likelihood of pregnancy following ART is related to the decline in oocyte quality with maternal age and the decline in BMP15/CC with age parallels this decrease.This fall in BMP15/CC can be attributed to either a reduced production of BMP15 by the ageing oocyte and/or to a decrease in CC capacity to bind the oocyte-secreted BMP15.At this stage it is unclear which possibility applies.By contrast, CC-bound GDF9 did not decline with maternal age.This lack of correlation between GDF9 and maternal age may be due, in part, to the lower sample number in GDF9 measurements (n = 81) compared to BMP15 (n = 201), as the observed high correlation (r = 0.90) between GDF9/CC and BMP15/CC within patients, would suggest that GDF9/CC may also show an age-related decline with increased sample size.The notion of lower intra-oocyte BMP15 protein concentrations being associated with oocytes of poorer quality is consistent with established knowledge that the oocyte-secreted factors BMP15 and GDF9 are the principal regulators of cumulus cell development and function and hence are important to COC and oocyte integrity (reviewed (Gilchrist et al., 2008;Findlay et al., 2018);).
The deterioration in oocyte quality with age in ART manifests as impaired preimplantation embryo development, increased rates of embryo aneuploidy and lower pregnancy rates (Cimadomo et al., 2018).Thus, the decrease in oocyte output of BMP15 with increasing age may indicate that BMP15/GDF9 are markers of oocyte quality.No correlation was found between BMP15 or GDF9 levels and oocyte and embryo outcomes such as oocyte maturation, fertilization rate, embryo yield on day 3 or day 5, or percentage of euploid embryos.This demonstrates that there are no differences in average oocyte-secreted BMP15/GDF9 on CCs between women with poor and good prognosis, when examining the total pool of a woman's CCs from an OPU.This is the first study to quantify BMP15 and GDF9 by ELISA in CCs, but does not include measurement of BMP15 and GDF9 in CCs from individual COCs.These findings differ in some respects to previous studies, albeit studies using different detection methods (different ELISA or RT-PCR) and using different sample types (CC RNA or embryo conditioned media (Li et al., 2014;Li et al., 2022);).Li et al. (2014) found an association between higher CC mRNA expression levels of GDF9 and BMP15 and improved embryo and pregnancy outcomes.Human CCs appear to express GDF9 and BMP15 mRNA at low levels, likely at multiple orders of magnitude  A.E. Krysta-Matter et al. less than oocytes (Zhang et al., 2018;Fan et al., 2021).By contrast, Li et al. (2022) (Gilchrist et al., 2004;Lin et al., 2012), including potentially principally in monomeric form (Heath et al., 2017).Examination of the BMP15 amino acid sequence (Fig. 2A) predicts that, in addition to the canonical RRTR pro-mature processing site, there are two additional potential cleavage sites in the pro-domain, with the position 61 site scoring the highest of the three predicted cleavage sites.Cleavage at the two non-canonical sites would generate 39 and 34 kDa fragments of BMP15 when the mature domain is attached, which equates to the 39 kDa immunoreactive protein detected in the WB of the GCs, which was also immuno-depletable.If that protein is also glycosylated then it may account for the immuno-depletable 41 kDa protein detected.Likewise, non-canonical cleavage in the GDF9 pro-domain at position 229, with the mature domain attached, would generate an ~40 kDa protein, which equates to the immunoreactive band of that size which was immuno-depletable.It is noteworthy that no processed, mature domain GDF9 (i.e.18-21 kDa) was detected on GCs.These observations are consistent with the notion that a substantial proportion of oocyte-secreted BMP15 and GDF9 are in various precursor pro-forms (Gilchrist et al., 2004;Heath et al., 2017), and indicates that pro-BMP15 and pro-GDF9 are likely to be contributing to ELISA immunoreactivity in the CC samples.In addition, it is known that human BMP15 is glycosylated and phosphorylated (Hashimoto et al., 2005;Saito et al., 2008;Pulkki et al., 2011) and here we show that the O-linked glycosylation site in the mature domain is in the approximate mAb 28A binding region.Native BMP15 produced by oocytes in vivo is likely differentially glycosylated compared to recombinant pro-BMP15 produced in the HEK cells in vitro, and such glycosylation may interfere with mAb 28A binding interactions, which may partially explain the lack of parallelism between the recombinant reference preparation and the GC/CC samples.Future characterization of native forms of BMP15 in these samples will clarify this issue.
A notable finding from the current study was the tight correlation (r = 0.90) between GDF9/CC DNA and BMP15/CC DNA, in terms of the hGC preparation used as reference standard in both ELISAs (Fig. 3F), with a regression slope value of 1.02.When the BMP15 and GDF9 levels were re-expressed in terms of purified recombinant human BMP15 and GDF9, the slope of the resulting regression line was ~2.0, suggesting an ~2:1 relationship between GDF9 and BMP15.A recent study, using different ELISA assays and different types of clinical samples, report an ~10:1 relationship between GDF9 and BMP15 (Cadenas et al., 2022).At this stage it is unclear how exactly to interpret the tight correlation between GDF9 and BMP15 as the molecular compositions of native BMP15 and GDF9 bound to their CCs are unknown (see above).There are several possible explanations for this close correlation between BMP15/CC and GDF9/CC: 1) That GDF9 and BMP15 expression, production and secretion are tightly coupled, consistent with the known extraordinarily tight correlation (R 2 = 0.99) in mRNA expression of BMP15 and GDF9, within species and across a range of species (Crawford and McNatty 2012;Lin et al., 2012).
2) That GDF9 and BMP15 exist as a complex in the form of cumulin (the heterodimer of GDF9 and BMP15 chains).The formation of the cumulin heterodimer has been produced by recombinant methods (Mottershead et al., 2015) and shown to be hyperactive in a number of GDF9 and BMP15 in vitro bioassay systems (Mottershead et al., 2012;Richani et al., 2019;Stocker et al., 2020).However, native cumulin has not been purified and characterized and there is contradictory evidence as to whether cumulin exists per se in vivo (Heath et al., 2017;Kristensen et al., 2022).It is also possible that since BMP15/GDF9 have been extracted from isolated cumulus and granulosa cells that BMP15/GDF9 are extracted as a complex in conjunction with a binding or carrier protein; e.g. a proteoglycan-like molecule (Watson et al., 2012) or a receptor (e.g.BMPRII) known to bind BMP15 and cumulin (Mottershead et al., 2015;Heath et al., 2017), where fixed molar ratios of BMP15 and GDFD9 may be evident.
3) That the GDF9 and BMP15 ELISAs are detecting native forms of GDF9 and BMP15 which cross react in both ELISAs.To date both mature and precursor forms of recombinant GDF9 and BMP15 have been assessed and they are not detected in the opposing ELISAs, however it is conceivable that these ELISAs may exhibit different specificities to the as yet uncharacterized native forms.To partially assess this possibility, immuno-neutralisation of the biot-28A antibody with a blocking peptide in the ELISA resulted in a dose-dependent reduction of specific binding to pro-BMP15 standard and to native BMP15 in GC extract, demonstrating that the BMP15 signal is specific.Technical difficulties were experienced with GDF9 where the use of a neutralising GDF9 peptide led to elevated backgrounds.However, it should be noted that dose response curves of CC and hGC extracts give non-parallel responses with recognized GDF9 and BMP15 reference preparations.Collectively, these data suggest that the two ELISAs are detecting a complex of some sort, resulting in non-parallelism with recognized reference preparations.For this reason, a high salt extract of hGCs was employed as an ELISA reference preparation throughout the studies as it provides a parallel response with CC extracts.We are currently exploring the development of a cumulin ELISA consisting of mAbs to both GDF9 and BMP15 to further characterize the hypothesis that GDF9 and BMP15 are present in cumulus salt extracts as a complex.The above discussion points reiterate the previously noted complexity, and at times controversies, surrounding the nature of the interactions between BMP15 and GDF9 (Mottershead et al., 2013;Heath et al., 2017), notwithstanding which, are important to understand, as these interactions have profound effects on mammalian ovulation rate and fecundity (Galloway et al., 2000).
A potential significant limitation of this study relates to the notion that GDF9 and BMP15 exist as separate entities in COCs which can be quantitated separately using specific assay methods.However, as discussed above, these proteins may also exist as complexes with each other and with other carrier proteins, in a similar manner to that seen with other heterodimeric TGFβ members (e.g.inhibins).Identifying both the native forms of GDF9 and BMP15 as a dimer or as a complex is an important next step in these studies.A further limitation of the study is the inherent heterogeneity of the patient samples examined, whereby the patients had a range of differing infertility pathologies, who received differing treatment protocols by a number of different physicians.These factors no doubt contributed to the large variance observed between patient samples.
In conclusion, the approach used in the current study provides a novel method to quantify oocyte-secreted BMP15 and GDF9 production in ICSI patients.Further studies are required to determine if BMP15 alone or in combination with GDF9 can be used as molecular markers of oocyte and embryo quality in the clinical setting.This study demonstrates that BMP15 levels on CCs decline with maternal age, which might be associated with the decline in oocyte quality related to reproductive ageing.BMP15 and GDF9 showed lack of association with embryo outcomes, which could be due to the fact that BMP15/GDF9 was measured on patient's pooled CC as opposed to measuring it in CCs from individual oocytes.Hence, this study opens the opportunity for future research to measure oocyte-secreted factors on discarded CCs from single oocytes, to determine their diagnostic value in a clinical setting.

Fig. 1 .
Fig. 1.BMP15 and GDF9 ELISA development and validation.(A) Doseresponse curves of cumulus cell and granulosa cell extracts in the BMP15 ELISA are non-parallel with human pro-BMP15.(B) Similar in the GDF9 ELISA, human GDF9 is non-parallel with cumulus cell and granulosa cell.(C,D) Extraction of proteins and DNA from cumulus cells was optimised using a range of sodium chloride concentrations.Data are presented as means of duplicate measures (black dots).The horizontal dotted line refers to the level of detection.

Fig. 2 .
Fig. 2. BMP15 and GDF9 structure and Western blot analysis of cumulus and granulosa cell extracts.Sequences of human pro-BMP15 (Panel A) and pro-GDF9 (Panel D).The pro-(roman text) and mature-(bold text) domains of human BMP15 and human GDF9 are shown.The residues are numbered according to the first residue of the signal peptide.Known and putative cleavage sites are depicted above the sequence.N-and O-linked glycosylation sites are also indicated (shaded in pink).The peptide sequences used to raise the BMP15 mAb#28A (A) and the two GDF9 mAbs#72B and #53/1 (D) are presented, as well as the mapped binding epitope of mAB#53/1 (D).Panel B: Western blots using biotinylated mAb#28A for detection of human pro-BMP15 and granulosa cell (GC) extracts.Each membrane was probed without or with immuno-neutralised mAb#28A using a specific neutralising peptide.Panels C,F: Quantification of WB grey scale intensity and molecular weight calibrated to standard curves, without (solid lines) and with (dotted lines) antibody peptide neutralisation.Numbers on peaks correspond to the calculated molecular weights of bands.Panel E: Western blots using biotinylated mAb#72B for detection of human GDF9, pro-GDF9 and GC extracts.Each membrane was probed without or with immuno-neutralised mAb#72B using a specific neutralising peptide.Red stars depict the bands which were fully neutralised in the GC extract samples.

Fig. 3 .
Fig. 3. Relationships between number of oocytes, levels of BMP15, GDF9 and DNA from cumulus cells.The relationships between total BMP15 (A), GDF9 (B), cumulus cell DNA (E) and oocyte number/patient, and between total BMP15 (C), GDF9 (D) and cumulus cell DNA are presented, as is the association between GDF9/CC DNA and BMP15/CC DNA (F).Each dot represents an individual patient, with closed circles (patient cohort 1) and open circles (patient cohort 2).Data represent linear regression analyses with regression lines plotted and corresponding p-values.

Fig. 4 .
Fig. 4. Associations between oocyte number and maternal age with oocyte BMP15 and GDF9 production.Associations between BMP15/CC DNA (A, D), GDF9/CC DNA (B,E) and their ratio (C,F) with oocyte number (A-C) and maternal age (D-F) are presented.Each dot represents an individual patient, with closed circles (patient cohort 1) and open circles (patient cohort 2).Data represent linear regression analyses with regression lines plotted and corresponding p-values.

Table 1
Assay validation of BMP15 and GDF9 ELISAs for human cumulus cells.

Table 2
Univariate linear regression results for predictors of change in BMP15/CC DNA and GDF9/CC DNA.

Table 3
Univariate linear regression results for predictors of change in oocyte and embryo outcomes.
measured GDF9 by ELISA in medium from single embryos, and found that GDF9 negatively correlates with embryo quality and probability of live birth.Whether measurement of oocyte-secreted GDF9 (on CCs) versus embryo-secreted GDF9 (in medium) represent comparable cellular processes reflective of function, requires further investigation.Nonetheless, these findings open the future possibility of measuring BMP15/GDF9 during ART treatment, on discarded CCs or media samples from individual oocytes/embryos of ICSI patients, as potential non-invasive biomarkers of female reproductive function.It is worth noting that native BMP15 and GDF9 have not yet been isolated and characterized.Previous studies indicate that oocytes secrete BMP15 and GDF9 in various forms, particularly as high molecular weight unprocessed proteins