Chronic superphysiologic AMH promotes premature luteinization of antral follicles in human ovarian xenografts.

Anti-Müllerian hormone (AMH) is produced by growing ovarian follicles and provides a diagnostic measure of reproductive reserve in women; however, the impact of AMH on folliculogenesis is poorly understood. We cotransplanted human ovarian cortex with control or AMH-expressing endothelial cells in immunocompromised mice and recovered antral follicles for purification and downstream single-cell RNA sequencing of granulosa and theca/stroma cell fractions. A total of 38 antral follicles were observed (19 control and 19 AMH) at long-term intervals (>10 weeks). In the context of exogenous AMH, follicles exhibited a decreased ratio of primordial to growing follicles and antral follicles of increased diameter. Transcriptomic analysis and immunolabeling revealed a marked increase in factors typically noted at more advanced stages of follicle maturation, with granulosa and theca/stroma cells also displaying molecular hallmarks of luteinization. These results suggest that superphysiologic AMH alone may contribute to ovulatory dysfunction by accelerating maturation and/or luteinization of antral-stage follicles.

INTRODUCTION

Anti-Müllerian hormone (AMH), found in 1947 by Alfred Jost, was initially noted for its role in promoting male sexual development (). AMH secreted by Sertoli cells in the fetal testes drives regression of Müllerian ducts, thereby enabling establishment of the Wolffian ducts and their derivatives in response to testicular androgens (). Yet, AMH is also expressed by the correlate of Sertoli cells in the ovary, granulosa cells (GCs) that surround and foster the development of oocytes within ovarian follicles (, ). The exclusive production of AMH by GCs in growing follicles has made serum measurement of this protein a reliable diagnostic indicator of ovarian follicular reserve and a surrogate for antral follicle count (, ). While serum AMH measurement is routinely applied in reproductive medicine and assisted reproductive technologies, its physiological, and potentially therapeutic, function in the context of reproductive biology and fertility is not well understood.

AMH performs critical yet unclear functions during folliculogenesis. It is a glycoprotein belonging to the transforming growth factor–β (TGFβ) superfamily secreted by GCs of growing follicles beginning at primary stages (). AMH is believed to exert its activity in a paracrine fashion, with production by growing follicles providing negative feedback on activation to preserve the primordial follicle (PrF) pool for an extended reproductive life span (). AMH knockout mice exhibit precocious activation and growth of follicles and premature depletion of PrFs (), which can be partially reversed in vitro with supplementation of neonatal mouse ovary culture with AMH alone (). Previous studies have recapitulated this growth-suppressive effect in vitro in explanted ovaries of mouse (), bovine (), and human (), and continuous application of recombinant or adenovirus-encoded AMH has been shown to confer both contraceptive and fertoprotective effects in mice (, ). Work from our group (, ) has extended this growth-suppressive effect to human xenografts, demonstrating the capacity for exogenous paracrine AMH to improve the retention of PrFs. However, following extended treatment with recombinant AMH, mouse ovaries display compensatory rebound folliculogenesis, with a threefold increase in oocyte production (). Multiple studies have demonstrated a positive effect of AMH on follicle growth in rats, primates, and humans (–). For these reasons, the precise role and mechanistic function of AMH within the ovary remain controversial.

AMH activity has also been shown to intersect with other circulating factors that modulate the reproductive axis in women. Aromatase [cytochrome P450 family 19 subfamily A member 1 (CYP19A1)], the GC-specific enzyme that converts androgens to estrogens, is stimulated by follicle-stimulating hormone (FSH) derived from the pituitary. In cultured GCs, AMH has been shown to inhibit the expression of FSH receptor (FSHR) () and the catalytic activity of aromatase (). Conversely, FSH induces, and estradiol represses, the expression of AMH (). These relationships describe a feedback loop whereby stimulatory input from circulating FSH promotes increased estradiol production by GCs in growing follicles. At the same time, FSH also drives increased expression of AMH that tempers growth and aromatization until accelerated production of estradiol at later antral stages ultimately reaches a threshold that silences AMH expression. In parallel, androgens drive increased expression of FSHRs in the GCs of preantral follicles (), augmenting their growth response and potentially increasing their expression of AMH. While the intersection of these signaling factors is fundamental to the homeostasis of the reproductive axis, their imbalance can result in endocrine dysfunction and infertility.

The most common cause of anovulatory infertility is polycystic ovary syndrome (PCOS), affecting greater than 5% of reproductive-aged women (). The Rotterdam criteria for diagnosis of PCOS require two of the following three criteria: oligo-anovulation, clinical and/or biochemical signs of hyperandrogenism, and polycystic morphology of ovaries on ultrasound (). While the etiology remains poorly understood, increasing evidence points to developmental () and epigenetic () influences that take root in the ovary. An early treatment of PCOS known as ovarian wedge resection effectively addressed the condition (). Also, as patients approach menopause, they resume regular menstrual cycles with the waning of their ovarian reserve (). Increased serum AMH in women with PCOS is believed to be a by-product of the increased volume of growing follicles that occurs either in parallel to or downstream of a hyperandrogenic milieu. However, the specific contribution of elevated AMH to the molecular pathology of PCOS and its defining clinical features is unclear, as no study, to date, has examined the effect of chronically elevated AMH in an experimentally controlled in vivo model.

Here, we use ovarian cortical xenografts with cotransplantation of engineered endothelial cells (ECs) to test the effect of chronic paracrine AMH stimulus on human folliculogenesis. We show that long-term xenografts exhibit an accelerated growth rate in the context of chronically elevated AMH and exhibit a molecular signature indicative of more mature stages, including that of luteinization. GCs and theca/stroma from follicles grown in the context of chronic AMH exhibited increased expression of a broad spectrum of factors related to cholesterol biosynthesis and metabolism, as well as factors that have been linked to PCOS. These data decouple elevated AMH from the metabolic and hyperandrogenic conditions that define PCOS and suggest that chronically elevated AMH induces a molecular cascade that contributes, at least in part, to the anovulatory phenotype in these patients.

RESULTS

Continuous paracrine stimulus of ovarian xenografts increases follicular activation and growth

We have previously demonstrated a benefit conferred by cotransplantation of ECs with thawed ovarian cortex and used these ECs as a vector to convey a continuous paracrine source of superphysiological AMH that aids in the retention of PrFs in short-term xenografts (, ). Here, we applied the same approach (Fig. 1A) in long-term (8 and 14 weeks) xenografts; exogenous ECs transduced with lentivirus encoding AMH complementary DNA (cDNA) secreted high levels of AMH in vitro (Fig. 1B) and were observed in a broad distribution within the fibrin matrix surrounding the graft (Fig. 1C) while also exhibiting continued expression of AMH from ECs at late stages (Fig. 1D). As observed in short-term grafts previously, the ratio of primordial to growing follicles was increased in the context of AMH ECs at the short-term interval (2 weeks; Fig. 1E and table S1). However, long-term xenografts at 8 and 14 weeks showed the reverse trend (Fig. 1E and table S1), with the diameter of antral follicles observed in the AMH condition threefold greater than controls (1.5 mm versus 0.5 mm, P = 0.01; Fig. 1F).

Fig. 1.

Long-term paracrine AMH stimulus accelerates growth in ovarian xenografts.

(A) Schematic representation of the experimental framework for cotransplantation of ovarian cortical strips along with either control or AMH-expressing ECs in immune-compromised mice. (B) Cell culture supernatant from ECs transduced with lentiviral particles encoding AMH cDNA shows >100-fold increased expression relative to supernatant from COV434 GCs. (C) ECs, labeled by mCherry fluorescent protein, persist in the margins of long-term xenografts. (D) Immunolabeling sections from 14-week xenografts transplanted with AMH ECs identify AMH protein produced by EC in the periphery of xenografts. (E) Recovery and quantification of primordial to growing follicle ratio in xenografts following 2, 8, and 14 weeks. (F) The diameter of antral follicles observed within xenografts recovered after 14 weeks. The arrowheads in (C) show primary and primordial stage follicles; the arrows in (D) show exogenous mCherry ECs forming vessel-like structures; the asterisks in (D) show isolated mCherry ECs. Error bars in (E) and (F) represent median ratios of primordial to growing follicles with each dot representing one replicate and bars indicating the median with 95% confidence interval. P values are shown. Scale bars, 1 mm.

Single-cell RNA sequencing of AMH-EC–conditioned antral follicles

Accelerated antral follicle growth in long-term xenografts with AMH (Fig. 1) suggested a role for this factor in regulating follicular homeostasis. To elaborate on the influence of chronic AMH on folliculogenesis, we performed single-cell RNA sequencing (scRNA-seq) on granulosa and stroma cells isolated from antral-stage follicles (Fig. 2A). Ovarian cortical fragments were transplanted under three conditions: (i) with control ECs (CTL ECs; n = 7 follicles), (ii) with a >95% purity population of ECs constitutively expressing AMH (high-AMH ECs, n = 7 follicles; Fig. 1B), or with a 1:9 ratio of high-AMH ECs to CTL ECs (low-AMH ECs, n = 3 follicles). We monitored xenograft-bearing mice via weekly magnetic resonance imaging (MRI; Fig. 2B), targeting follicles between 1 and 4 mm for recovery (Fig. 2C). Seven CTL EC follicles (1.93 ± 1.24 mm) were isolated from four mice at 136 ± 30 days after transplant, and 10 AMH EC (including 7 AMH-high and 3 AMH-low) follicles (2.28 ± 0.73 mm) were isolated from seven mice after 115 ± 18 days (P = 0.07; Fig. 2D).

Fig. 2.

Recovery of AMH-conditioned xenografts for scRNA-seq.

(A) Schematic representation of experimental approach for recovery and scRNA-seq of xenograft-resident follicles from control and AMH conditions. (B) Xenograft-bearing mice were subjected to weekly MRI to monitor emergence and progress of antral follicles. (C and D) Within 1 week of primary observation on MRI, antral follicles were recovered from the gluteus muscle (C); the term of xenograft and size upon recovery is shown in (D). (E to G) Following recovery, antral follicles were bisected and resident cells were enzymatically dissociated, labeled for FACS isolation, and submitted for single-cell cDNA library preparation and sequencing; uniform manifold approximation and projection analysis of 107,429 cells submitted for scRNA-seq identifies at least four distinct clusters (G) identified by expression of canonical markers (F) of EC (CDH5), SMC (RGS5), GC (FSHR and CYP19A1), theca/stroma (ANPEP and CYP17A1), and corpus luteum (CL) (LGALS3, BLVRB, and TIMP2) identity. White arrowheads in (B) designate incipient antral follicles; gray arrowhead in (C) identifies superficial antral follicle visible through the fascia of the gluteus maximus muscle. Scaling for average expression in each feature plot in (F) is shown at the right of each panel.

Following isolation, antral follicles were bisected to make the antral cavity accessible and subjected to enzymatic dissociation, as we have previously described (), and GCs (CD99) or theca/stroma (CD39/CD55) were isolated by fluorescence-activated cell sorting (FACS) for scRNA-seq. Following routine normalization and quality control in Seurat (RStudio) (), 107,429 cells remained for downstream analysis. Cells could be grouped by library (fig. S1) or transcript expression (table S2) into clusters (Fig. 2E) representing ECs (CDH5), smooth muscle cells (RGS5), GCs (FSHR and CYP19A1), theca/stroma [ANPEP (Aminopeptidase N) and CYP17A1], and an additional cluster of cells exhibiting markers evident in corpora lutea [LGALS3 (Galectin 3), BLVRB (Biliverdin reductase B), and TIMP2 (Tissue inhibitor of metalloproteinases 2)] (Fig. 2, F and G).

Chronic AMH augments GC transcripts related to cholesterol and sex steroid metabolism

We performed an analysis of cells contained within the GC cluster (82,542 cells) comprising the aggregate GC population (fig. S2 and Fig. 3A). Cells in the AMH condition showed an increasing trend in the proportion of cells in the G2-M and S phases (Fig. 3B), although this did not reach statistical significance. Discrimination of differentially expressed transcripts between the control and AMH conditions revealed a broad increase in enzymes driving cholesterol biosynthesis and sex steroid metabolism (table S3 and Fig. 3C). Transcripts for apolipoprotein E (APOE), scavenger receptor class B member 1 (SCARB1), low-density lipoprotein receptor (LDLR), 3-hydroxy-3-methylglutaryl–coenzyme A reductase (HMGCR), cytochrome P450 family 11 subfamily A member 1 (CYP11A1), and CYP19A1 (aromatase) were increased in low-AMH EC and high-AMH EC conditions (Fig. 3C). Additional transcripts related to general cytochrome P450 activity [POR (Cytochrome P450 reductase), FDX1 (Adrenal ferredoxin), and FDXR (Ferredoxin reductase)], fatty acid/cholesterol biosynthesis [FASN (Fatty acid synthase), SCD (Stearoyl-CoA desaturase), SCAP (SREBP cleavage-activating protein), SREBF1/2 (Sterol regulatory element-binding transcription factor1/2), and INSIG1 (Insulin induced gene 1)], and aromatase activity [FOS, FOSB, JUN, and JUNB ()] were also differentially expressed between conditions (table S3).

Fig. 3.

GCs from chronic AMH xenografts exhibit altered metabolism.

(A) A total of 82,542 cells comprising the GC populations isolated from control and low- and high-AMH xenografts were subclustered according to their origin. (B) Percentage of cells at G1, G2-M, or S phase was determined for GCs derived from control or AMH xenografts. (C to H) Enzymes that mediate cholesterol/steroid biosynthesis and metabolism (C) were differentially expressed to a significant degree between control and low-/high-AMH conditions; (D to H) immunostaining highlights the differential expression and localization of LDLR and SCARB1 (D), CYP17A1 and APOE (E), HMGCR (F), CYP11A1 (G), and CYP19A1 (H) in antral follicles from control and AMH conditions. Arrowheads in (panels D and E) indicate LDLR + and APOE + cells. Asterisks in (panels D, E, G and H) indicate the inner boundary of the GC layer. (I to L) Serum from xenograft-bearing mice (n = 4 control at 2 weeks, 2 AMH at 2 weeks, and 4 control long-term and 6 AMH long-term) was drawn, and levels of AMH (I), progesterone (J), testosterone (K), and estradiol (L) were determined. Lines in (C) indicate the median value. Scale bars, 1 mm (D to H). Bars in (B) and (I) to (L) represent SD of values; P values are shown in (B) and (L) with only the significantly different value shown in (L).

To elaborate on increased expression of steroidogenic factors in AMH xenografts, we performed immunolabeling and localized differentially expressed factors (Fig. 3, D to H). While control xenografts exhibited LDLR cells exclusively within the theca, LDLR cells were also observed among GCs under the AMH condition (Fig. 3D). Similarly, although SCARB1 was present in GCs under both conditions, signal intensity was increased in GCs within AMH xenografts (Fig. 3D). APOE, which was expressed in the theca/stroma layer of control xenografts, was present in the GC layer along with CYP17A1 in the AMH condition (Fig. 3E). HMGCR protein was observed in GCs under both control and AMH conditions; however, the range of cells in which it was present was increased in AMH xenografts (Fig. 3F). Signal intensities of CYP11A1 (Fig. 3G) and CYP19A1 (Fig. 3H) were also increased under the AMH condition, concordant with transcriptomic data. These results suggest that xenografts under the AMH condition are stimulated to increase cholesterol and sex steroid biosynthesis.

To determine whether the follicle-specific increase in mediators of cholesterol and sex steroid biosynthesis/metabolism has an effect on circulating factors, we conducted serum measurements of AMH, progesterone, testosterone, and estradiol in xenograft-bearing mice (n = 8 control and 8 high/low AMH; Fig. 3, I to L). Relative to xenograft-bearing mice that were euthanized at 2 weeks, mice bearing long-term xenografts showed elevated serum AMH and sex steroid levels. Mice bearing AMH xenografts that were euthanized after 2 weeks showed no detectable serum AMH (Fig. 3I), suggesting that AMH observed in other xenograft-bearing mice is not derived from exogenous ECs. Mice bearing long-term xenografts under the control condition exhibited increased AMH and estradiol; however, AMH xenografts showed very low or undetectable estradiol (Fig. 3, I to L).

Differential activity of paracrine AMH on mural versus oophorus GCs

Integration of paracrine signaling, in particular in the case of morphogens from the TGFβ superfamily, is dependent on the concentration and source (direction) of the ligand. Given the localization of AMH-expressing ECs to the margins of antral follicles, mural GCs are expected to exhibit an increased response to exogenous AMH (Fig. 1D). We previously demonstrated the progressive restriction of Nectin1 protein (PVRL1 transcript) to oophorous GCs with increasing antral follicle size (Fig. 4A) (). To examine whether mural versus oophorous GCs were differentially influenced by exogenous AMH, we fractionated the GC population via threshold expression of Nectin1 and Nectin1 subsets (CTL mural, CTL ooph, AMH mural, and AMH ooph in Fig. 4, B and C). As we previously showed (), oophorous GCs exhibited increased entry into G2-M and S phases; however, this was augmented under the AMH condition (Fig. 4B). Focusing on a subset of transcripts related to steroidogenesis (CYP11A1, FDX1, and CYP19A1), mural and oophorous GCs from the control condition showed similar expression levels (Fig. 4C). In the context of AMH, the median expression level was globally increased, and the increase was more pronounced in mural GCs. In addition, expression of SOX4 (SRY-box transcription factor 4), a regulator of developmental gonadogenesis () and TGFβ-responsive gene (), was relatively higher under the control condition, specifically in mural GCs, but was decreased in the context of AMH ECs and was specifically lower in mural GCs from the high-AMH condition (Fig. 4C). To determine whether these trends were also observed in cells that were prospectively isolated on the basis of surface expression of Nectin1, we sorted Nectin1 and Nectin1 cells from one similarly sized (3 mm) follicle of each condition (sort CTL mural, sort CTL ooph, sort AMH mural, and sort AMH ooph in Fig. 4, D and E). Cells in G2-M and S phases were enriched under the AMH condition, and CYP11A1, FDX1, and CYP19A1 transcripts were increased and SOX4 transcripts were decreased in sorted cells from the AMH condition, with a more pronounced influence exhibited among the mural (Nectin1) cells. Last, immunolabeling corroborated a reduction of nuclear-localized Sox4 protein in the GCs of AMH xenografts (Fig. 4F).

Fig. 4.

Mural GCs display enhanced phenotypic response to exogenous AMH.

(A) Representative image of a ~2-mm antral follicle showing specific staining of Nectin1 on GCs in the oophorous compartment. (B to E) The Nectin1/Nectin1 transcript-expressing (B and C) or Nectin1/Nectin1 surface protein–containing (D and E) fractions were segregated and compared; the percent of each population at G1, G2-M, and S phases was determined for each population (B and D), and ridge plots comparing expression levels of differentially regulated transcripts are shown (C and E). (F) Immunolabeling for SOX4 identifies protein localization antral follicles isolated from control and AMH conditions. The shaded circles in each ridge plot in (C) and (E) represent the median. Scale bars, 100 μm (A) and 1 mm (F).

GCs in AMH xenografts exhibit a precocious maturation signature

Increased rate of antrum growth (Fig. 1F) and increased activation of factors that mediate steroid and cholesterol metabolism (Fig. 3, C to H) are attributes that are typically evident at a more advanced stage of follicle development (). On the basis of this, we speculated that chronic AMH stimulus may induce precocious maturation. To place AMH and control transcriptional signatures within the developmental hierarchy of human GCs, we compared bulk RNA-seq transcriptomes of GCs from antral-stage follicles of a single pair of donor ovaries (n = 4 follicles, 5.0 ± 2.16 mm) to those from the pooled cumulus-oocyte complex (COC) and follicular fluid of four in vitro fertilization (IVF) patients following ovarian hyperstimulation and trigger (Fig. 5, A and B). Numerous transcripts that were up- or down-regulated in posttrigger GCs relative to antral follicle GCs (Fig. 5C) were similarly expressed under the AMH condition relative to control (CTL ECs; Fig. 5D). Pregnancy-associated plasma protein A (PAPPA) and insulin-like growth factor binding protein 5 (IGFBP5), which have been implicated in both IGF signal integration and ovulation, were increased in the context of AMH, and immunolabeling revealed an increase in PAPPA cells in the GC and theca of these follicles (Fig. 5E). Phosphorylated AKT was also increased in the GC and theca region of follicles from the AMH condition (Fig. 5F). Together, these results suggest that antral follicles that develop in the context of chronic AMH are endowed with a molecular signature that approximates follicles of a more advanced, periovulatory phase.

Fig. 5.

GCs in the context of chronic AMH resemble posttrigger GCs.

(A) Schematic for isolation of uterine epithelium, preovulatory GCs and postovulatory GCs from patients. EPCAM, Epithelial cell adhesion molecule. (B) Heatmap showing relative expression levels of marker genes that identify each cell type. (C) Volcano map showing transcripts that are up-regulated (green) and down-regulated (blue) in posttrigger GCs relative to GCs isolated from antral follicles. FC, fold change. (D) Dotplot showing transcripts that are similarly regulated in the scRNA comparison (D) as they are in the post- and preovulatory GC comparison (C). (E and F) Immunolabeling for IGFBP5 and PAPPA (E) and phosphorylated AKT (F) in antral follicle GCs from control and AMH conditions. For the dotplot (D), measurement of average expression and the percent of cells expressing each transcript are shown in the legend. Scale bars, 1 mm (E) and 100 μm (F).

Xenografts in the presence of chronic AMH stimulus enrich for cells of luteinized phenotype

We performed a subanalysis of cells that comprised the theca/stroma populations (19,013 cells) of control and AMH xenografts (fig. S3 and Fig. 6A). Identification of transcripts that were uniquely expressed in discrete Seurat clusters (Fig. 6B and table S2) revealed canonical markers identifying smooth muscle cells (RGS5), stroma (ENG and OGN), and theca interna (CYP17A1) (Fig. 6C). In addition, a collection of transcripts that are enriched in corpus luteum (–) (TIMP2, CEBPB (CCAAT/enhancer-binding protein beta), BLVRB, and LGALS3; Fig. 6, D and E) were expressed in clusters composed almost exclusively of cells from the AMH condition (Fig. 6F). Comparison of aggregate GC and stroma cells from control and AMH conditions revealed a significant increase in transcripts that are observed in periovulatory [CYP11A1 (3beta-hydroxysteroid dehydrogenase/delta(5)-delta(4)isomerase type II), HSD3B2, LDLR, SCARB1, INHBA (Inhibin beta A), INHA (Inhibin alpha), and APOE; yellow box in Fig. 6G) and/or luteal (BLVRB, TIMP2, and LGALS3; blue box in Fig. 6G) cells. Immunolabeling of follicles from control and AMH conditions localized LGALS3, a marker of luteinized follicles (), to GCs (Fig. 6, H to K); while staining was prominent in the stroma in both conditions, LGALS3 was moderately enriched in GCs of antral follicles at early stages (~1 mm) in the context of AMH (Fig. 6J), with robust signal present in CD99/ANPEP cells distributed throughout the GC layer of larger (>2.5 mm) antral follicles (Fig. 6K). Last, transcripts for factors that are differentially regulated in patients with PCOS [MRO (Maestro), CFP, PF4V1 (Properdin, Platelet factor 4 variant 1), and THBS1 (Thrombospondin 1)]; gray box in Fig. 6G) (–) were similarly modulated in the context of exogenous AMH (Fig. 6G).

Fig. 6.

Chronic AMH promotes a luteal phenotype.

(A and B) Cells comprising the theca/stroma cluster (total 19,013 cells) were isolated for subanalysis and plotted according to origin (A) and Seurat cluster (B). (C) Feature plots identifying cells of SMC (RGS5), general stromal (ENG and OGN) and theca interna (CYP17A1) identity. (D) Feature plots of transcripts that have been observed in the context of corpus luteum and/or GC luteinization. (E) Violin plots show the relative expression of corpora luteal transcripts between stromal populations of control versus AMH conditions. (F) Percent of all theca/stroma cells from control and AMH conditions occupying the corpus lutea–like cluster. (G) Dotplot showing the aggregate transcript values of all cells from the control or AMH condition grouped into GC, stroma, or the corpus lutea–like cluster; transcripts that are specific for corpus luteum (blue) or luteinization (yellow) or otherwise related to PCOS (gray) are shown. (H to K) Immunolabeling of antral follicles from control (H and I) and AMH (J and K) xenografts reveals localization and levels of LGALS3 protein in stroma and GCs. Scaling for average expression in each feature plot in (C) and (D) is shown at the right of each panel. Lines in (E) show the median. For the dotplot (G), measurement of average expression and the percent of cells expressing each transcript are shown in the legend below the plot. Stroke boxes in (H) to (K) are magnified in the panels to the right. Arrowheads in (panels H to K) indicate the GC basement membrane; asterisks in (panels H to K) indicate the inner boundary of the GC layer. Scale bars, 1 mm.

DISCUSSION

The function of AMH in the context of normal ovarian physiology and its potential therapeutic or pathologic impact on folliculogenesis remain poorly understood. Here, we have used human cortical xenografts to model the influence of chronically elevated AMH on follicular development independent of the complex endocrine environment present in the ovary. We showed that chronic AMH has a profound influence on the growth and transcriptomic signature of antral-stage follicles, with a global increase in factors related to the synthesis and/or metabolism of cholesterol and sex steroid hormones, as well as precocious expression of factors typically observed at more advanced stages of folliculogenesis. These experiments highlight the cascade of molecular pathways that are modulated by AMH alone and elucidate a potential mechanistic contribution to the anovulatory phenotype observed in the context of PCOS.

In the late follicular phase of a healthy reproductive cycle, follicle-resident cells undertake a radical shift in their expression profile as steroidogenesis shifts from generating androgens/estrogen to producing progesterone (). Hence, global increases in transcripts related to cholesterol and steroid biosynthesis/metabolism (Fig. 3) are consistent with precocious maturation of antral follicles under the AMH condition. Increased expression of SCARB1 and LDLR and a resultant increase in scavenging of substrate for cholesterol biosynthesis have previously been observed in luteinizing macaque GCs (). Similarly, APOE, which provides reverse cholesterol transport to serve as a substrate for increased steroid production (), was observed among GCs under the AMH condition, and the range of expression of HMGCR, the rate-limiting enzyme of cholesterol synthesis (), was expanded. While increased expression of these factors, as well as CYP11A1 and STAR, provides an enzymatic pathway to increased progesterone production to support the luteal phase (), CYP17A1 and CYP19A1 drive production of androgen and estrogen, respectively, and are reduced following ovulation trigger. Yet, the expression of CYP19A1 increases in antral follicles at more advanced preovulatory stages (), and although typically exclusive to theca cells in preovulatory follicles, CYP17A1 expression is driven by luteinizing hormone (LH) (), and GCs are endowed with LH receptor approaching ovulation and can be induced to express CYP17A1 and produce androgens via inhibition of FOS (). Hence, the drastic reduction of FOS, FOSB, JUN, and/or JUNB (table S3) in GCs under the AMH condition may account for increased expression of LHCGR and CYP17A1 (table S3 and Fig. 3E). Meinsohn et al. () showed that administration of AMH suppresses follicle development and ovulation and used scRNA-seq to demonstrate that increased activity of AP1 transcription factors coincides with this phenotype; our observation that FOS, FOSB, JUN, and JUNB are decreased may suggest that modulation of these immediate early genes underlies a switch from inhibition of follicle growth to differentiation of GCs in antral follicles.

Despite the broad increase of factors governing cholesterol/sex steroid metabolism in the context of AMH, serum measurements of these mice revealed a significant decrease in estradiol (Fig. 3L). Serum may not provide an accurate measure of sex steroids present at the follicle level, in particular in the xenograft model; however, reduced estradiol output is a hallmark of luteinization (), and lower E2/oocyte ratio is associated with poorer IVF outcomes (). Reduced estradiol could also account for the marked reduction of SOX4, as this transcription factor has been shown to be under the control of estrogen and progesterone in female reproductive tissues (). SOX4 is also activated by TGFβ signaling via the Smad2/3 branch (), and although AMH signal integration remains unclear, it is thought to act in opposition to Smad2/3 signaling via the AMH receptor AMHR2 (). In aggregate or as individual influences, disruption of TGFβ signaling and discordance of steroidogenic processes in the context of chronic superphysiologic AMH may contribute to the fertility challenges observed in patients with PCOS.

Comparison of posttrigger GCs to GCs isolated from 2- to 6-mm antral follicles highlighted differences in transcript levels that were consistent with the comparison of GCs from xenografts in the context of control versus AMH ECs (Fig. 5). Among the transcripts that were increased under the AMH condition, the presence of IGFBP5 and PAPPA were corroborated at the protein level, with PAPPA markedly increased (Fig. 5E). PAPPA is predominantly expressed in theca cells of small follicles, becoming localized to GCs at preovulatory stages (), and acts to increase the bioavailability of IGF ligands via cleavage of IGFBP4/5 (). Knockout of PAPPA compromises fertility in mice, and a common variant that reduces cleavage efficiency of IGFBP5 () has been associated with recurrent pregnancy loss (). IGF signaling is critical for follicle progression and has been implicated in the onset of ovulation (), and GCs in the context of AMH showed an increase in phosphorylated AKT (Fig. 5F). Hence, increased PAPPA under the AMH condition and augmented bioavailability of IGF ligands may increase signaling and promote accelerated growth and/or maturation. In support of this, we have previously shown in xenografts of ovarian cortical tissue that chronic paracrine expression of exogenous IGF1 increases the pace of follicle growth at preantral stages but accelerates luteinization at antral stages (). If chronic AMH drives a similar phenotype, then targeted inhibitors of PAPPA that have been developed by multiple groups for limiting IGF bioavailability (–) could mitigate this negative influence.

The first-line therapy to address anovulation in PCOS women of normal body mass index is ovulation induction, using either clomiphene citrate (CC) or letrozole (). While CC restores ovulation in ~80% of patients with PCOS, pregnancy is achieved in about ~35 to 40% (). This can be attributed in part to the antiestrogenic effects of CC on the endometrium. Nonetheless, LH is also relatively elevated in the follicular phase of women with PCOS (), and CC further increases these levels (). While many studies have linked increased LH to hyperandrogenic action on the theca, and AMH is thought to directly regulate gonadotrophs to increase LH pulsatility (), our results suggest that chronic high AMH can induce expression of the LH receptor at earlier stages of folliculogenesis, thereby exacerbating the disruptive effect of elevated LH from the pituitary. These data, combined with numerous randomized controlled trials suggesting improved outcomes with the use of letrozole versus CC (), may suggest against CC in treating anovulatory PCOS. Alternatively, gonadotropin-releasing hormone antagonists (e.g., ganirelix) have been used to mitigate premature luteinization in ovulation induction cycles that employ CC. Antagonists have become favored for treating women with PCOS who undergo IVF () because they not only prevent LH surge but also mitigate the increased probability for these patients to develop ovarian hyperstimulation syndrome. Nevertheless, oocytes from these cycles exhibit lower fertilization rates () and impaired developmental competence (). While this negative correlation has been attributed to the influence of abnormal metabolic and androgenic influences on oocyte development (), our results support an alternative mechanism based on the deleterious influence of chronically elevated AMH on the synchronization of follicle/oocyte maturity.

Apart from anovulation and polycystic ovaries, women with PCOS often present with hyperandrogenemia and hyperinsulinemia/insulin resistance (). As these pathologies typically coincide, this study is the first to isolate the influence of superphysiological AMH on follicular development. Although CYP17A1 expression was expanded in GCs, testosterone was undetectable in the serum of mice bearing AMH xenografts (Fig. 2D), suggesting that AMH alone does not drive increased androgenesis within the ovary. Chronic exposure to exogenous AMH alone does not reinforce a hyperandrogenic state in graft-bearing mice; however, the model fails to capture the complex endocrine and metabolic milieu present in healthy or PCOS ovaries. Hence, this system is limited in its capacity to assess whether observed differences are relevant to the diversity of metabolic phenotypes observed in patients with PCOS. Obesity and metabolic syndrome are commonly associated with PCOS, and weight loss and lifestyle changes in obese patients with PCOS can often resolve anovulation and hyperandrogenemia (). Consistent with this, a study using laparoscopic ovarian cautery (ovarian drilling) demonstrated reduced androgens in women with PCOS but did not result in significant changes to insulin sensitivity or serum lipid profile (). While these clinical results illustrate the multifactorial nature of PCOS symptoms and etiology, the profound influence of chronic AMH on the transcriptomic and phenotypic profiles of xenografted antral follicles underscores the relevance of this single factor to follicle development and infertility.

We have applied a xenograft platform to isolate the influence of long-term superphysiologic AMH on the development of human antral follicles, a stage of development that is both difficult to experimentally address and vitally important to the pathology of PCOS. Our findings underscore the broad influence of AMH on transcriptional activity and maturation state of follicles and support an independent role for dysregulation of AMH signaling in driving anovulation in women with PCOS. Conclusions from this study must be drawn with caution, however, as elevated AMH is almost always observed in combination with one or more symptomatic hallmarks in PCOS. Moreover, oophorectomy of xenograft-bearing mice may eliminate their endogenous source of AMH and steroid hormones but the rest of the axis remains intact, and cross-talk between host and graft may be affected by species differences and/or blood volume, graft site, etc. We previously demonstrated subtle differences in transcriptional signature between cells from xenografted versus native ovarian antral follicles (). Last, the use of cotransplanted ECs as a source of paracrine AMH, while robust, does not recapitulate the dynamic expression of AMH within the ovary. Despite these limitations, the analysis provides a deep and high-resolution examination of AMH action on human folliculogenesis and suggests a prominent effect on antral follicle maturation.

MATERIALS AND METHODS

Procurement of ovarian tissue

Ovarian cortex was isolated from whole bilateral ovaries obtained from four brain-dead organ donors (Table 1), aged 25 (donor ovary 1), 29 (donor ovary 2), 28 (donor ovary 3), and 35 (donor ovary 4) years, as a part of a collaboration with the International Institute for the Advancement of Medicine. Ovarian cortical fragments were also obtained from four patients undergoing fertility preservation (Fert Pres) in advance of treatment for malignancy (Table 1), aged 26 (Fert Pres patient 1) 18 (Fert Pres patient 2), 5 (Fert Pres patient 3), and 19 (Fert Pres patient 4) years. Ovary donors had no history of chemotherapy and no apparent history of endocrine or reproductive abnormalities. A chart detailing vital statistics for all individuals used in the study is shown in Table 1.

Table 1.

Origin of ovarian tissue used for the study and patient/donor condition.

Fertility preservation, Fert Pres.

Source	Age	Condition/cause of death
Fert Pres patient 1	26	Hodgkin’s lymphoma
Fert Pres patient 2	18	Lymphoblastic lymphoma
Fert Pres patient 3	5	Beta thalassemia
Fert Pres patient 4	19	Ewing’s sarcoma
Donor ovary 1	25	Head trauma
Donor ovary 2	29	Anoxia
Donor ovary 3	28	Anoxia
Donor ovary 4	35	Head trauma

Ovaries were processed as previously described (). Briefly, ovaries were resected and placed in sterile Leibowitz medium (Gibco) on ice for transport to the research laboratory for processing following a cold ischemic interval of ≤4 hours. Cortical tissue was separated from medulla and further processed to remove medulla tissue, and thinned cortex was cut into fragments of roughly 3 mm by 12 mm by 1 mm. Each cortical strip was then equilibrated in cryoprotectant solution in cryovials, slow-frozen, and stored in liquid nitrogen until thaw and transplantation. All tissue was obtained from donors and patients with informed consent and with approval of the Weill Cornell Institutional Review Board (IRB) (study title: Low temperature preservation of ovarian tissue; study number, 0803009702).

Ovarian cortical xenografts

All procedures were approved and experiments were performed in accordance with the guidelines and regulations of the Institutional Animal Care and Use Committee (IACUC) of Weill Cornell Medicine (IACUC protocol no. 2014-0008, Assessment of angiogenic and hematopoietic tissue in mouse). These procedures were previously described in (, ). Briefly, cryopreserved tissue was thawed rapidly, washed of cryoprotectant, and encapsulated in fibrin that was premixed with fluorescently labeled single-cell suspension of ECs. Fibrin-embedded tissue was then transplanted into oophorectomized mice bilaterally under the fascia of the gluteus maximus. The fascia, dorsal wall, and skin were closed with sutures. Xenografted animals were maintained under sterile conditions until animals were euthanized and xenografts recovered for fixation, cryosectioning, and immunohistochemical staining, or for excision of antral follicles and isolation of follicle-resident cells. For quantification of follicle volume, distribution, and size, a third of the slides of each xenograft were stained with hematoxylin and eosin, enabling identification and quantification of follicles of all developmental stages. All unique follicles (excluding those that were large enough to traverse multiple sections) were counted once; antral follicles were captured using a motorized stage and mosaic tiling (Zeiss LSM 710), and a digital scale bar was used to precisely measure antral cavity in a section represented the midpoint with greatest diameter.

Isolation of ovarian follicles and GCs from xenografts

These procedures were previously described in (). Cells isolated from a total of 21 antral follicles (17 in xenografted tissue originally isolated from three organ donors and 4 primary antral follicles isolated from a fourth organ donor for bulk RNA-seq) were analyzed for this study. A breakdown of total antral follicles and their use is included in table S5. Individual follicles ≤4 mm at the superficial cortical layer of ovaries were identified, and surgical instruments were used to excise cortical tissue encompassing them. Each follicle was bisected and placed in a 60-mm dish with 6 ml of Accutase (Invitrogen), followed by incubation in a humidified incubator at 5% CO2 and 37°C. After 10 min, 6 ml of serum-containing medium was added to follicles, and cells were flushed from the tissue via repeated recycling of medium over bisected follicles with a P1000 micropipette. After flushing, follicle tissue was put aside, and the supernatant was collected and passed through a 100-μm filter (Corning) before centrifugation. Following centrifugation, the supernatant was aspirated, leaving the cell pellet. Following flushing, the remnant tissue was placed in collagenase (100 U/ml) and dispase (1 U/ml) in Hanks’ balanced salt solution and incubated for 30 min in a humidified incubator at 5% CO2 and 37°C. The tissue was then recovered and triturated via pipetting through a P1000 micropipette tip and filtered through a 100-μm filter before centrifugation, supernatant aspiration, and labeling of cell pellets.

Labeling and FACS of cells from antral follicles

These procedures were previously described in (). For cells isolated from an ovary for bulk RNA-seq, pellets were resuspended in blocking solution and mouse anti-human FSHR for 10 min at 4°C. After washing in phosphate-buffered saline (PBS), centrifugation, and aspiration of supernatant, cell pellets were resuspended in blocking solution containing Alexa Fluor 488–conjugated secondary antibody and incubated at 4°C for 10 min. After washing and centrifugation, cells were resuspended in FACS buffer containing 4′,6-diamidino-2-phenylindole (DAPI) and run on a FACSJazz (BD Biosciences) with collection and validation of the FSHR fraction following initial sort. A total of at least 20,000 cells were obtained for each sample for RNA-seq. For isolation of follicle-derived cells from xenografts for scRNA-seq, antibodies that were directly conjugated to fluorophores specified (CD99, CD39, and CD55) were used after being resuspended in blocking solution.

Isolation and flow cytometry of GCs from follicular fluid

Cellular material was isolated from follicular fluid of two patients undergoing oocyte retrieval as previously described (). The IRB Committee of Weill Cornell Medical College approved the collection of follicular fluid, which is considered discarded material, upon the patients’ informed consent. Procedures for the collection of cells were previously described in (). Cells were enzymatically dissociated from COCs using Accutase, and separately, pooled follicular fluid was poured over the Ficoll-Paque Plus solution and centrifuged for 30 min at 1625g at 4°C. The interface containing the cellular fraction was collected and diluted in Dulbecco’s minimum modified Eagle’s (DMEM)/F12 media with 10% knockout serum replacement and 1× penicillin-streptomycin (Gibco), followed by centrifugation at 500g for 10 min. The cell pellet was resuspended in DMEM/F12 (Gibco) media to wash, then centrifuged, and resuspended in PBS with 0.1% bovine serum albumin. The COC and fluid-derived cells were labeled with an antibody against CD99 and sorted on a FACSJazz (BD).

Immunofluorescence

Cryosections were labeled as previously described (, ). Samples were permeabilized in PBS/0.1% Tween 20 with 5% donkey serum (Millipore) and incubated in 2 to 5 μg/ml concentrations of specified antibodies, washed, and counterstained with DAPI and mounted in ProLong Gold (Gibco). Images were captured using a Zeiss 710 confocal microscope. A list of antibodies used in the study is described in table S6.

Endothelial cells

Human ECs were obtained from Angiocrine Bioscience and were originally isolated from neonatal umbilical vein [human umbilical cord ECs (HUVECs)] as described () under an IRB-approved protocol for use of discarded biological material. HUVECs were isolated and expanded for three passages in EC growth medium before cryopreservation.

Lentiviral vectors and transduction of cells

Lentiviral particles (GeneCopoeia, LPP-L2481-Lv215-050-GS) encoding AMH cDNA connected via IRES (internal ribosome entry site) to mCherry fluorescent protein were added (multiplicity of infection = 10) to cultured ECs and incubated for 48 hours. Cells were then expanded in culture for 1 passage, sorted for mCherry to obtain a population of ≥95% purity, and expanded for an additional two to three passages before cryopreservation in DMEM/fetal bovine serum containing 10% dimethyl sulfoxide.

Bulk RNA-seq RNA isolation

These procedures were previously described in (). Cell pellets from FACS isolation of antral follicle or follicular fluid GCs were processed using the Arcturus PicoPure RNA isolation kit (Thermo Fisher Scientific), with the treatment of each sample using ribonuclease (RNase)–free deoxyribonuclease. RNA quality was checked by Agilent Technologies 2100 Bioanalyzer.

Bulk RNA-seq and analysis

These procedures were previously described in (). At least 100 ng of high-quality total RNA was used as input to convert the mRNA into a library of template molecules for subsequent cluster generation and sequencing using the reagents provided in the Illumina TruSeq RNA sample preparation kit. Following purification of the poly-A containing mRNA molecules using poly-T oligo-attached magnetic beads, the mRNA was fragmented into small pieces using divalent cations under elevated temperature. The cleaved RNA fragments were copied into first-strand cDNA using reverse transcriptase and random primers. This was followed by second-strand cDNA synthesis using DNA polymerase I and RNase H. These cDNA fragments then went through an end repair process, the addition of a single “A” base, and then ligation of the adapters. The products were then purified and enriched with polymerase chain reaction to create the final cDNA library. After quantifying and checking the size and purity of the product, multiplexed DNA libraries were normalized to 10 nM, and then two sample libraries were pooled together in equal volumes. Seven picomolar of each pooled DNA library template was amplified on Illumina cBot instrument involving immobilization and 3′ extension, bridge amplification, linearization, and hybridization and then sequenced on one lane of the Illumina HiSeq 2000 sequencer using the pair end module and generating 2 × 58–bp–long reads. The samples were aligned to the human GRCh38 reference assembly using STAR aligner, and subsequently, genes were counted in htseq.

scRNA-seq and analysis

These procedures were previously described in (). Library preparation was performed using the Chromium Single Cell 3′ Reagent Kit and sequenced on a NovaSeq 6000 sequencer with 300 cycles per run. Samples were aligned to the human GRCh38 reference assembly using STAR aligner, and downstream analysis was performed using the R package Seurat version 2.2. Details of the bioinformatic analysis are contained within the script accompanying the manuscript, but briefly, to exclude poor-quality cells, only cells with greater than 200 but less than 4000 features and less than 20% of reads mapping to mitochondrial genes were retained. Counts were normalized using default normalization (Function NormalizeData). Retained cells from all libraries were integrated using the FindIntegrationAnchors and IntegrateData functions in Seurat and were then scaled to regress cell cycle effects. To correct for patient effects, the mutual nearest-neighbor function was used. Function FindAllMarkers performed differential expression analysis between cells in a cluster relative to the remaining cells in the dataset, or alternatively, directly compared two populations. All transcripts with a P value of 0.005 or less were included in gene lists.

Statistical analyses

These procedures were previously described in (). The significance of differentially expressed genes between groups in Figs. 1 (E and F), 2D, and 3 (B and I to K) was calculated using the nonparametric Mann Whitney U test (Prism 8). A P value of less than 0.05 was considered significant. For the significance of scRNA-seq data, the P values for each comparison were calculated using the Wilcoxon rank sum test.