FSHR and LHR functional compensation reveals the mechanism and treatment of Ovarian Hyperstimulation Syndrome

Abstract

Gain-of-function mutations in the human follicle-stimulating hormone receptor (FSHR) cause spontaneous ovarian hyperstimulation syndrome (OHSS), a serious reproductive disorder. However, the molecular physiology and treatment options for OHSS remain elusive. Notably, estrildid finches naturally carry an FSHR variant (Thr449Ala) analogous to the pathogenic mutation in humans yet are resistant to OHSS. Here we show that this resistance stems from significantly reduced luteinizing hormone receptor expression in estrildid ovarian granulosa cells. Furthermore, treatment with the luteinizing hormone receptor antagonist alleviates OHSS symptoms in mouse models. Single-cell RNA transcriptomic reveals functional compensation of the two receptors to regulate estrogen production and vascular permeability, resembling the adaptive mechanisms observed in estrildid finches. Our study unravels the molecular mechanism underlying the physiological adaptation of estrildid ovaries to high FSHR constitutive activity and is a example of how the concept of Darwinian Medicine could be exploited to identify novel drug targets for ovarian hyperstimulation syndrome treatment.

Mutations in the follicle-stimulating hormone receptor (FSHR) cause spontaneous ovarian hyperstimulation syndrome (OHSS). Here, the authors report that estrildid finches naturally carry an analogous pathogenic mutation found in patients but avoid the syndrome by lowering a related receptor’s activity. Blocking this receptor successfully treats OHSS in mice.

Introduction

Ovarian hyperstimulation syndrome (OHSS) is a serious complication of ovarian stimulation, typically triggered by the excessive development of multiple follicles¹. The syndrome was first recognized in 1943, when ovulation was induced using exogenous gonadotropins². These early therapeutic agents were derived from pregnant mare serum, sheep pituitary glands, and the urine of pregnant women^2,3. Mild OHSS occurs in approximately 20-30% of in vitro fertilization (IVF) cycles, whereas intermediate and severe forms are observed in 3-6% and 0.1-2% of cases, respectively⁴. Clinically, OHSS manifests as abdominal distension and discomfort, which is manifested by, for example, bilateral, multiple ovarian cysts, ascites, and low circulating protein levels⁵. Although the exact mechanism remains incompletely understood, its pathophysiology is thought to proceed as follows: pharmacologically stimulation induces multi-follicular development during the follicular phase, followed by extensive luteinization of the retrieval follicles, leading to the formation of multiple corpora lutea. This, in turn, causes ovarian enlargement and an exaggerated release of ovarian hormones and vasoactive mediators, including cytokines, angiotensin, estrogen, and vascular endothelial growth factor (VEGF)^6,7. These factors collectively increase vascular permeability and promote fluid extravasation.

In female mammals, follicle development is tightly regulated by two pituitary gonadotropins: follicle-stimulating hormone (FSH) and luteinizing hormone (LH)⁸. Folliculogenesis typically culminates in the selection of a single dominant follicle, and the process is conventionally divided into gonadotropin-independent and gonadotropin-dependent phases. Early follicle development, encompassing the primordial, primary, and secondary stages, occurs independently of gonadotropins. The transition from the pre-antral to the early antral stage marks the onset of the gonadotropin-dependent phase, during which further growth and maturation require both FSH and LH^8,9. LH and FSH act through their specific receptors-FSH receptor (FSHR) and LH receptor (LHR)-both members of the G-protein coupled receptor (GPCR) family characterized by large extracellular domains responsible for hormone bindings¹⁰. In females, FSHR expression is restricted to granulosa cells throughout preovulatory follicular development, whereas LHR mRNAs is localized exclusively in the thecal cells of immature follicles and in both thecal and granulosa cells of mature antral follicles. Since LH and FSH cooperatively promote follicles growth, ovulation, and corpus luteum formation, they are routinely used during ovarian stimulation for in vitro fertilization/intracytoplasmic sperm injection (IVF/ICSI). FSH derives follicular growth and maturation and contributes to LH-induced ovulation and luteinization¹¹. Human chorionic gonadotropin (hCG), a natural analogue of LH that shares the same receptor but exhibits greater biological activity and a longer half-life, remains the gold standard for triggering ovulation¹².

OHSS typically arises following the administration of exogenous gonadotropins, such as FSH or human menopausal gonadotropin (hMG), in combination with hCG for ovulation induction, particularly when estradiol levels are elevated¹³. In rare instances, OHSS can occur in the absence of exogenous hormonal stimulation, a condition referred to as spontaneous OHSS (sOHSS)¹⁴. sOHSS is thought to result from unusually high circulating levels of endogenous hCG or thyroid-stimulating hormone (TSH), both of which can mimic gonadotropin activity and aberrantly activate FSHR on granulosa cells¹⁵. Up to now, seven distinct heterozygous activating FSHR variations have been identified: p.Asp567Asn/Gly, p.Thr449Ile/Ala, p.Ile545Thr, and p.Ser128Tyr/Thr, which can potentially trigger the development of severe sOHSS^16–23. The FSHR variants have a markedly increased sensitivity to normal TSH and/or hCG levels. Except for the p.Ser128Tyr/Thr mutation, the other amino acid exchanges are also known to induce constitutive activity of FSHR²⁴. Laboratory studies using a genetically engineered mouse model harboring a constitutively active FSHR mutation (FSHR^D580H) demonstrate that this mutation leads to a range of ovarian abnormalities including irregular ovarian structure, hemorrhagic cysts, accelerated deletion of primordial, primary, and secondary follicles, increased estradiol synthesis, and elevated prolactin (PRL) levels in adolescent mice²⁵. It is anticipated that humans with similar mutations may exhibit a comparable phenotype²⁵. Additionally, a positive correlation between the degree of constitutive activity and the severity of ovarian abnormalities has been suggested²⁵. Mice with the FHSR^D580Y mutation exhibit a milder ovarian phenotype, primarily characterized by hemorrhagic cysts²⁵.

The prevailing models on the functioning of GPCRs stipulate that they require the binding of at least one endogenous ligand to activate heterotrimeric G proteins, which regulate various physiological functions²⁵. However, research has shown that GPCR variants with constitutive activity are associated with different diseases^26,27. This indicates that the constitutive activities stimulated by these GPCRs are intricately linked to their corresponding physiological functions²⁸. In clinical patients, numerous gain-of-function variations have been discovered in GPCRs over the past few decades^27,29,30. Specifically, functionally activated variations in FSHR, accompanied by constitutive activation of Gs signaling, can facilitate OHSS³¹. However, the molecular physiology and treatment of OHSS remain elusive for a long time.

Here, the evolutionary analysis of vertebrate FSHR unexpectedly revealed that the gain-of-function variation T449A, previously discovered in clinical OHSS patients³², naturally occurs in the FSHR of estrildid finches examined. The Estrildidae family belongs to the songbird suborder Passeri, including species such as the zebra finch (Taeniopygia guttata, tg), a well-established model organism for birds, as well as the Gouldian finch (Erythrura gouldiae, eg), Bengalese finch (Lonchura striata domestica, ls), and white Java sparrow (Padda oryzivora, po), all examined in this study³³. In the present study, we elucidated the physiological adaptation of the estrildid ovary due to this natural FSHR variant and proposed an alternative therapeutic approach for treating OHSS. For a comprehensive list of abbreviations and compilation of animal species (Latin and common names) and their abbreviations, see Supplementary Data 1a (sheet1) and 1b (sheet2), respectively.

Results

Natural p.Thr449Ala variant confers enhanced activity to estrildid FSHR

To evaluate whether FSHR underwent positive selection in estrildids, we retrieved the coding sequence of 57 vertebrate species from the NCBI database and used PCR amplification and sequencing to obtain the sequence information of E. gouldiae and P. oryzivora (Supplementary Data 2), resulting in a dataset of 59 species for branch-site selection analysis. A phylogenetic tree was generated using TimeTree to represent the evolutionary relationships among these 59 species, and two additional species, namely common wombats (Vombatus ursinus, vu) and Megaleporinus macrocephalus (a headstander fish), were included in the tree (Fig. 1a) because they also carry the p.Thr449Ala variant, thus providing a broader visual context for this rare substitution. The branch‑site test showed that only estrildid FSHR orthologs had a large nonsynonymous (dN)/synonymous (dS) substitution rate ratio (branch-site dN/dS of ω » 1) that was highly significant (likelihood ratio tests [LRTs], P < 0.05) (Fig. 1b). Two amino acid sites (167 and 449) underwent positive selection in estrildid birds, while the ancestral positions 167 and 449 were conserved in most vertebrates (Fig. 1a-c). Notably, at site 449, the amino acid threonine (T) is ubiquitously observed across nearly 3000 vertebrate species, with the exceptions of estrildid finches, Vombatus ursinus, and headstanders (Anostomidae), which exhibit a T to alanine (A) variation at this site (Supplementary Fig. 1). Additionally, species in the Mormyridae family and Greenland flounder (Reinhardtius hippoglossoides, rh) exhibit a T to isoleucine (I) variation at the same site (Supplementary Fig. 1). To explore the functional ramifications of this variation, the 449 sites were swapped between human and estrildid FSHR. The constitutive activity was evaluated using a CRE (cAMP response element)-luciferase assay in HEK293 cells. Our findings revealed a consistent enhancement in the constitutive activity of FSHR across a diverse array of species including mammals, birds, amphibians, reptiles, and fish when the 449th amino acid was replaced from T to A (Fig. 1d and Supplementary Fig. 2). Similarly, a substitution from T to I at this position enhanced FSHR constitutive activity as well (Supplementary Fig. 3). Consistently, in species like the estrildid finches, which naturally possess the 449 A variant, swapping to T at this position resulted in diminished FSHR constitutive activity and cellular cAMP concentration (Fig. 1d and Supplementary Fig. 2b, 4a, b). Notably, in V. ursinus, this variation led to an abolition of constitutive activity (Fig. 1d and Supplementary Fig. 2). Estrildid finch FSHR exhibited vigorous constitutive activity between 2.9-14.1 times higher than that observed in the non-estrildid Serinus canaria (sc) and chicken (Gallus gallus, gg) when expression was adjusted to levels similar to endogenous FSHR in estrildid ovaries (Supplementary Fig. 4c,d). These results also indicated that physiological levels of FSHR expression are sufficient to elicit constitutive activity (Supplementary Fig. 4c,d). Hence, the T449A variation universally augmented the Gs pathway of all vertebrate FSHRs and was specific to the Gs pathway (Supplementary Fig. 5).

Fig. 1. In estrildid finches FSHR conserved amino acid threonine 449 is substituted by alanine.

a A species tree was constructed using TimeTree (http://timetree.temple.edu/). Magenta branches represent avian species, blue branches indicate mammals. Species harboring positive selection sites p.Gly167Asn and/or pThr449Ala highlighted in red (e.g., Estrildidae) or orange letters. All animal silhouettes were sourced from phylopic.org. b FSHR underwent positive selection in estrildid finches. The dN/dS ratio (ω) was calculated across the entire protein-coding region using branch-site models. Likelihood ratio tests (LRTs) were performed to compare models allowing positive selection (ω > 1) on the foreground branch against null models constraining ω = 1, with significance assessed by χ² distribution. Statistical significance was defined as p < 0.01. c Schematic diagram showing the locations of positive selection sites on FSHR, Gly167 at the N-terminus and Thr449 in transmembrane domain 3 (TM3). The figure was created with biogdp.com. d Slope plots to illustrate the relationship between FSHR constitutive activity (y-axis) and the corresponding exogenous FSHR protein levels (x-axis). The CRE-luciferase signal intensity for both wild-type and mutants (swapping T to A and vice versa) groups was normalized to the baseline (0 ng). A higher value of slope indicates a higher level of Gs constitutive activity. e Concentration-response curves of hCG to mammalian and avian FSHR in HEK293 cells. The hsFSHR449 site corresponds to the orthologous sites 448 and 451 in M. musculus and S. canaria, respectively. For simplicity, “449” is used throughout. Position 449 as alanine (A) is indicated in red (dots), threonine (T) in blue (squares). pcDNA represents empty vector (black triangles). For a list of species abbreviations and image authorization and attribution, see Supplementary Data 1b and 1c. Species symbols in red or blue denote major naturally occurring A or T alleles, respectively. Data are represented as mean ± SD from at least three independent experiments, performed in technical triplicate. Source data are provided as a Source Data file.

T449A substitution enables FSHR activation by LHR ligands

The FSHR^T449A mutation was initially identified in patients with OHSS, and experiments have confirmed that it expands the ligand range of FSHR³². Based on this finding, we assessed the impact of the T449A variation on the ligand-induced response of avian and mammalian FSHR in HEK293 cells. Our results showed that mammalian FSHR and LHR (including those from humans, rats, mice, and V. ursinus) exhibited a high specificity for the FSH and LH ligands (Supplementary Fig. 6–8). In contrast, receptor-ligand specificity was markedly weaker in birds. While mammalian systems maintain strict FSH-FSHR and LH-LHR pairing, we found that in several avian species (such as T. guttata, L. striata domestica, G. gallus, and S. canaria), FSH can cross-activate LHR (Supplementary Fig. 8b). Moreover, both mammalian and avian FSHR and LHR could be activated by hCG (Fig. 1e and Supplementary Fig. 8a, b). The T449A variation increased FSHR sensitivity to hCG in both mammalian and avian species, consistent with previous findings³² (Fig. 1e). Notably, LH, which originally could only specifically activate LHR but not FSHR in humans and non-estrildid birds, was found to non-specifically activate FSHR in estrildid finches, as well as non-estrildid FSHR carrying the p.Thr449Ala variant (Supplementary Fig. 7). This result indicated that T449A enhanced the response of FSHR to LHR ligands (LH and hCG) and thus broadened its ligand range. Furthermore, the differential sensitivity of the 449 T and 449 A variant receptors to FSH varied among vertebrates. E.g., in the estrildid birds, T. guttata and L. striata domestica, the T449A mutation enhanced the stimulation by both recombinant human FSH (r-hFSH) and chicken FSH (ggFSH) (Supplementary Fig. 6). In contrast, amino acid exchanges at position 167 had no impact on the response of FSHR, nor did it affect the constitutive activity of the receptors, except for a slight decrease in T. guttata (tg)FSHR (Supplementary Fig. 9). These findings revealed that during the evolutionary process, estrildid finch FSHR had gained enhanced constitutive activity and a broader ligand range due to the T449A variation, similar to the Homo sapiens (hs)FSHR^T449A mutation, making it a valuable animal model for studying the mechanistic impacts of the natural FSHR^T449A variant.

FSHRT449A knock-in mice, unlike estrildids, are susceptible to OHSS

To investigate the impact of the FSHR^T449A variation on ovarian function, we generated a knock-in mouse model harboring the T449A mutation (FSHR^T449A-KI) (Fig. 2a, and Supplementary Fig. 10a). Our findings revealed that the T449A mutation did not affect ovarian histological architecture, ovary weight, ovarian cAMP levels, and the mice remained fertile (Supplementary Fig. 10). However, serum estradiol and FSH concentrations were significantly elevated in the homozygous FSHR^T449A-KI (HO) mice (Supplementary Fig. 10). Excessive doses of pregnant mare serum gonadotropin (PMSG), followed by administration of 5-30 IU hCG, are commonly used to induce OHSS in rodent^6,34. In the control group (standard superovulation protocol), animals received a single injection of PMSG followed by 5 IU of hCG 48 h later. Because the FSHR^T449A mutation enhances FSHR responsiveness to both FSH and hCG and expands its ligand spectrum, we applied a milder OHSS-induction protocol using a reduced hCG dose (1 IU) to assess whether FSHR^T449A-KI mice are predisposed to OHSS (Fig. 2b). We found that following induction with the same doses of PMSG and hCG, homozygous KI mice exhibited more severe OHSS symptoms compared to wild-type (WT) mice, as evidenced by a significantly greater increase in ovary size, serum and ovarian VEGF levels, serum estradiol concentrations and elevated numbers of corpora lutea (Fig. 2c–g, Supplementary Fig. 11g). Additionally, ovarian cAMP levels and genes related to the estrogen pathway, such as steroidogenic acute regulatory protein (Star), cytochrome P450 family 19 subfamily a member 1 (Cyp19a1), as well as genes regulating vascular permeability, including bone morphogenetic protein 2 (Bmp2)³⁵ and early growth response 1 (Egr1)³⁶ were significantly elevated in homozygous OHSS mice (Fig. 2h, i). Consistent with these findings, under the standard superovulation protocol, homozygous KI mice displayed clear OHSS-related phenotypes compared to WT mice. The KI animals showed significantly elevated ovarian cAMP levels, ovarian enlargement, and increased estrogen and VEGF production (Supplementary Fig. 11a–f), indicating an inherent susceptibility to OHSS even during routine superovulation procedures. Since the ovary/body weight ratio, serum estradiol and VEGF levels were related to the dose of hCG in WT mice (Supplementary Fig. 12), and higher hCG easily triggered OHSS, we hypothesized that the difference might be due to the enhanced responsiveness of FSHR to hCG induced by the T449A mutation, as detected at the cellular level (Fig. 1e).

Fig. 2. Estrildid finches do not exhibit increased susceptibility to OHSS compared to mice with the FSHRT449A mutation.

a Murine Fshr gene diagram. Protein coding exons in yellow, UTRs in green and sequence confirming the A→G nucleotide exchange resulting in p.Thr449Ala in FSHR^T449A -KI mice (KI). b Wild-type (WT) and homozygous KI mice were treated with 20 IU PMSG daily for four consecutive days, followed by intraperitoneal injections of 1 IU hCG. Ovaries and serum were collected 24 h post-hCG administration (n = 4/group). c, d Representative images of ovaries (c) and ovary weight (d). e Serum estradiol concentration. f Ovarian and serum VEGF levels. g Representative images of H&E-stained ovaries (scale bar=200 μm) and corpora lutea counts. h Ovarian cAMP levels. i mRNA levels of Egr1, Star, Bmp2 and Cyp19a1 in WT and HO-KI OHSS ovaries. j Experimental scheme for inducing an avian OHSS-like phenotype using FSH or hCG. Adult female T. guttata (tg), L. striata domestica (ls), S. canaria (sc), and M. undulatus (mu) were treated with intraperitoneal injections of 20 IU PMSG daily for four consecutive days. Twenty-four hours after the final PMSG injection, a single dose of either 5 IU hCG or hFSH was administered to induce an OHSS-like phenotype. In the control group, only a single PMSG injection was given on day-2, followed by hCG administration 48 hours later. k Representative ovarian images from Estrildidae and non-Estrildidae species. l Changes in ovary/body ratio (control group: ls, n = 10; tg, n = 7; sc, n = 10; mu, n = 4. OHSS-like group: ls, n = 7; tg, n = 8; sc, n = 6; mu, n = 5). m serum estradiol concentration following hCG administration (control group: ls, n = 5; tg, n = 4; sc, n = 3; mu, n = 4. OHSS-like group: ls, n = 5; tg, n = 3; sc, n = 4; mu, n = 6). n Alterations in ovary/body ratio (control group: ls, n = 11; tg, n = 9; sc, n = 11; mu, n = 7. OHSS-like group: ls, n = 12; tg, n = 8; sc, n = 5; mu, n = 5). o Serum estradiol concentration (control group: ls, n = 7; tg, n = 5; sc, n = 5; mu, n = 4. OHSS-like group: ls, n = 4; tg, n = 5; sc, n = 4; mu, n = 5). p Representative images of H&E-stained bird ovaries after PMSG/hFSH stimulation (scale bar=100 μm). q mRNA levels of Egr1 and Bmp2 in control and OHSS-like ovaries (all normalized with GAPDH mRNA in percent). Data are represented as mean ± SD, statistical analysis was performed using two-tailed Student’s t-test (d–i, l–o, and q). Source data are provided as a Source Data file.

Similar to mammals, avian FSHR was mainly expressed in ovary tissue, followed by the lung, and at low levels in non-gonadal tissues^37,38 (Supplementary Fig. 13). Despite the gain-of-function variation in estrildid FSHR, no significant differences were observed in ovary weight or serum hormone concentrations (estradiol and FSH) between estrildid finches and other bird species without the FSHR^T449A variation (Supplementary Fig. 13c, e). Further histological analysis revealed no abnormal ovarian structures or hemorrhagic cysts, unlike those observed in FSHR^D580H-KI mice²⁵ (Supplementary Fig. 13d). Estrildid finches serve as natural models for the FSHR^T449A variation, yet it remained uncertain whether they would be susceptible at all to OHSS, as seen in mice and patients. To further investigate, we attempted to trigger OHSS in birds using PMSG and hCG or hFSH, since hCG does not naturally occur in birds (Fig. 2j). We found that hCG/hFSH could increase ovary weight ratio and estrogen levels in non-estrildid birds and induce symptoms similar to OHSS, but had a negligible effect on estrildid finches (Fig. 2k–o). In addition, histological examination revealed a greater number of enlarged follicles after PMSG/hFSH stimulation, especially in non-estrildid birds (Fig. 2p). Expression of VEGFA significantly increased in all four species (Supplementary Fig. 14). However, EGR1, BMP2, adhesion G Protein-Coupled Receptor G1 (ADGRG1, a gene that can regulate VEGF secretion³⁹), and CCAAT/enhancer-binding protein beta (CEBPB, which increased in OHSS mice (Fig. 5)) showed significant increases in non-estrildid species but not in estrildid finches (Fig. 2q, and Supplementary Fig. 14). These results indicated that although estrildid finches harbor the FSHR^T449A variation, they did not appear to be more susceptible to OHSS mice, suggesting that Estrildidae had evolved physiological adaptations throughout evolution to protect against OHSS.

Fig. 5. Functional compensation of Gs signaling between FSHR and LHR in OHSS.

a Experimental scheme for inducing OHSS in homozygous FSHR^T449A-KI mice. Solvent or antagonists (LUF5771 or hFSH-β-(33-53), 20 mg/kg) were administered 24 h after hCG. b Representative images of ovaries. c, Comparisons of ovary weight (n = 6). d Effect of LUF5771 and hFSH-β-(33-53) on serum estradiol concentration (n = 6). e Effect on serum and ovarian VEGF levels (n = 6). f Effect on ovarian cAMP concentration (n = 4). g Representative images of H&E-stained ovaries (scale bar=500 μm). h Effect on corpora lutea counts (n = 4). i UMAP plots of single-cell ovarian transcriptomes from OHSS mice (n = 1, generated by pooling both ovaries from one mouse) and LUF5771/hFSH-β-(33-53)-treated mice (n = 2, generated by pooling both ovaries from each of two mice) in the FSHR^T449A-KI mouse model, induced by intraperitoneal injection of PMSG and hCG. j FeaturePlot of cell type-specific markers in the integrated single-cell atlas of OHSS mice and LUF5771/hFSH-β-(33-53)-treated ovaries. k Proportions of distinct cell types identified in the single-cell ovarian transcriptomes. l Bar plot illustrating the number of DEGs across various cell types in the single-cell ovarian transcriptomes, comparing the OHSS vs. LUF5771 groups and OHSS vs. hFSH-β-(33-53) groups. m Heatmap and pathway enrichment analysis of shared DEGs in granulosa cells and theca cells between the OHSS group and the LUF5771/hFSH-β-(33-53) groups. Significance was determined by two-tailed Wilcoxon rank-sum test. n Comparison of signature scores for the angiogenesis (GO term: GOBP1) and estrogen synthesis (GO term: GOBP2) pathways between the OHSS group and the LUF5771- or hFSH‑β‑(33-53)-treated groups. Significance was determined by two-tailed Wilcoxon rank-sum test. o–s qRT-PCR validation of mRNAs corresponding to DEGs involved in estrogen signaling pathway (Jun, Akt3, Hbegf, Hspa1a, Star, Fos) (o), VEGF signaling pathway (Vegfa, ppp3ca) (p), vascular permeability (Bmp2, Bmpr2, Id1, Klf6) (q), TNF signaling pathway (Cebpb) (r)., cAMP signaling pathway (Nfkbia) (s) in the ovaries of FSHR^T449A-KI mice after OHSS induction and treatment with LUF5771 and hFSH-β-(33-53) using real-time PCR (n = 5). Data are represented as mean ± SD, analyzed using one-way ANOVA with multiple comparisons (c–f, o–s). Source data are provided as a Source Data file.

LHR significantly decreased in FSHRT449A-bearing estrildid ovaries

To further investigate the intrinsic adjustment mechanisms of estrildid finches to avoid OHSS, we performed single-cell RNA sequencing (scRNA-seq) on the ovaries of two estrildid species (T. guttata, tg, and L. striata domestica, ls) as well as one non-estrildid bird (S. canaria, sc) under steady-state physiological conditions without exogenous stimulation This experimental design enables the identification of baseline transcriptional adaptations in estrildid finches. After integration based on the canonical correlation analysis (CCA) strategy⁴⁰, all high-quality cells were divided into ten main homologous cell types based on the canonical cell markers (mesenchymal cells: DCN, MMP2, COL6A1; granulosa cells: FSHR, FOXL2, EMX2; endothelial cells: S1PR1, ADGRL4, KDR; immune cells: IL16, PTPRC, SPI1; glial cells: HEBP2, FRMD4A; HSD11B2⁺ cells: HSD11B2, SPP1; oocytes: DAZL, DDX4, ZP2; red blood cells: AK2, TAL1, CREG1, epithelial cells: KRT18, KRT6, MIOX⁺ cells: HSD17B4, PSPH) (Fig. 3a–c, and Supplementary Fig. 15a). By analyzing the distribution of each population in all species, the granulosa cells, immune cells, and mesenchymal cells constitute a relatively substantial proportion across ovarian atlases in all species (Fig. 3b). Principal component analysis (PCA) was performed to explore cell-type relationships across species. Correlation analysis and PCA showed tight clustering of cell types in each species, indicating that the inter-species difference was more significant than intra-species. The L. striata domestica was phylogenetically closest to T. guttata, and S. canaria was distant to both, reflecting their evolutionary relationships (Fig. 3d, e). Feature plots demonstrated that FSHR expression was primarily localized to granulosa and pre-granulosa cells across all three species, with additional expression observed in mesenchymal and epithelial cells (Fig. 3f). A cross-species comparison of granulosa cells revealed a significant downregulation in estrildid finches of pathways related to cell cycle regulation, blood vessel development and morphogenesis, and steroid metabolic process (Fig. 3g, and Supplementary Fig.15b). These findings were based on the intersection of differentially expressed genes (DEGs) from pairwise comparisons between T. guttata and S. canaria, as well as L. striata domestica and S. canaria, representing consistent transcriptional changes in estrildids relative to the non‑estrildid control. The expression of VEGFA and ADGRG1, which enriched in angiogenesis/vasculature development significantly increased in estrildid finches. In contrast, the expression of EGR1, BMP2, angiotensin I converting enzyme (ACE)⁴¹, and inhibitor of DNA binding 1 (ID1)^42,43 in the angiogenesis/vasculature development pathways that regulate vascular permeability were significantly decreased (Fig. 3g and Supplementary Data 3). Additionally, the expression of STAR and 17β-hydroxysteroid dehydrogenase type 7 (HSD17B7), which is enriched in the estrogen signaling pathway, was significantly decreased (Fig. 3g, and Supplementary Data 3). These differentially expressed genes identified in scRNA-seq were further validated by quantitative PCR analysis (Fig. 3h–j). Among these genes, EGR1 has been reported to promote the development of OHSS⁴⁴ and ACE inhibition can mitigate OHSS symptoms⁴⁵. This suggests that the observed changes in gene expression within granulosa cells reflect an adaptive evolutionary mechanism in estrildid finches, potentially serving as a strategy to mitigate the risk of OHSS.

Fig. 3. Single-cell transcriptomic comparative analysis of estrildid and non-estrildid ovaries.

a UMAP plots of single-cell transcriptomes of ovaries from adult T. guttata (tg, n = 3), L. striata domestica (ls, n = 2), and S. canaria (sc, n = 3) under normal physiological conditions. b Proportions of distinct cell types identified in the single-cell ovarian transcriptomes. c Dot plots showing the expression of key marker genes (x-axis) of major cell types (y-axis). d PCA analysis of cell types across these three birds. e Similarity of these cell types between tg, ls, and sc. f, k FeaturePlots depicting FSHR (f) and LHR (k) expression across the integrated single-cell atlas of tg, ls, and sc ovaries. g Heatmap and GO pathway enrichment analysis of shared differentially expressed genes (DEGs) in granulosa cells. Significance was determined by two-tailed Wilcoxon rank-sum test. h mRNA levels of genes involved in angiogenesis/vasculature development/vascular permeability (EGR1, ADGRG1, ACE, ID1, BMP2, VEGFA and KLF6) (n = 5). i Estrogen signaling pathway (STAR, HSD17B7, JUN) (n = 5). j TNF pathway (CEBPB mRNA) in ovaries from adult tg, ls, and sc (n = 5). l Violin plot showing LHR expression in granulosa cells of the three species. Significance was determined by two-tailed Wilcoxon rank-sum test (sc vs. tg/ls; p < 0.0001). m Lower LHR expression in estrildid ovaries (n = 4). All normalized with GAPDH mRNA in percent. Data are represented as mean ± SD, analyzed using one-way ANOVA with multiple comparisons (h–j, l). Source data are provided as a Source Data file.

In addition to scRNA-seq, a total of 12 ovaries were collected from two estrildid species (T. guttata, n = 3; L. striata domestica, n = 3), as well as two non-estrildid species (S. canaria, n = 3; G. gallus, n = 3), and their RNA was deep-sequenced (with 78.5 ± 17.6 SD million reads per sample) (Supplementary Data 3). Comparative analysis of transcriptomes between estrildid and non-estrildid ovaries revealed that 291 genes were expressed at higher and 284 genes at lower levels (Supplementary Data 4). Notably, the LHR gene, which was significantly decreased in granulosa cells of estrildid finches according to scRNA-seq analysis of avian ovaries, also consistently exhibited lower expression in estrildid ovaries as confirmed by RNA-seq analysis (Fig. 3g, and Supplementary Fig. 16). scRNA-seq revealed that LHR was highly expressed in the granulosa and theca cells of S. canaria but was nearly undetectable in the granulosa cells of the T. guttata and L. striata domestica (Fig. 3k, l). Real-time PCR (qRT-PCR) further confirmed the lower expression of LHR in the ovaries of Estrildidae species (Fig. 3m). This suggested that the diminished LHR expression could be a crucial mechanism by which estrildids adapted to the FSHR variation.

LHR as a therapeutic target for OHSS intervention

To investigate whether the reduction of LHR signaling is a key adaptive mechanism for estrildid finches to avoid OHSS, we induced OHSS in C57BL/6 mice using PMSG and hCG, and administered the allosteric LHR inhibitor LUF5771⁴⁶ to investigate the effects (Fig. 4a). RU486 was used as a positive control, as it has been shown to reduce LHR expression and OHSS symptoms⁴⁷. As expected, PMSG and hCG significantly induced ovary enlargement, increased levels of estradiol and VEGF, and elevated numbers of corpora lutea (Fig. 4b–h, and Supplementary Fig. 17a). After treatment with LUF5771, there was a significant reduction in cAMP concentration in the mouse ovaries, as well as a decrease in ovary weight, estradiol, and VEGF levels (Fig. 4b–h, and Supplementary Fig. 17a), indicating that LUF5771 could ameliorate the OHSS phenotype in mice. Additionally, we explored the therapeutic effects of two newly discovered LHR antagonists, BAY-298 and BAY-899, on rat models of OHSS, as these inhibitors selectively antagonize LHR in humans, rhesus macaques, and rats⁴⁸ (Fig. 4i). Compared to the untreated OHSS group, ovaries treated with BAY-298/899 exhibited significantly lower cAMP concentrations, reduced ovary weight, and decreased levels of estradiol and VEGF, effectively treating OHSS in the rat model (Fig. 4j–p, and Supplementary Fig. 17b). These findings suggested that inhibition of LHR signaling is a physiological adaptation used by estrildid finches in response to the FSHR^T449A variation, which could also serve as a therapeutic strategy for OHSS.

Fig. 4. LHR antagonists alleviate OHSS symptoms in rodent models.

a Experimental scheme for inducing mouse OHSS using PMSG and hCG. Solvent or antagonists (LUF5771, 20 mg/kg or RU486, 20 mg/kg) were administrated 24 h after hCG injection. b Representative images of ovaries. c Comparisons of ovary weight (n = 8). d, Ovarian cAMP concentration after treatment with LUF5771 (n = 5). e Effect of LUF5771 and RU486 treatment on VEGF levels in serum (n = 7) and ovarian (n = 6). f Serum estradiol concentration (n = 5). g Corpora lutea count (n = 6). h Representative images of H&E-stained ovaries (scale bar=500 μm). i Experimental scheme for inducing a rat OHSS model with PMSG and hCG. Solvent or drugs were administered 24 h after hCG injection. j Representative images of ovaries. k Comparison of ovary weight (n = 5). l Effect of BAY-298/BAY-899 on VEGF levels in serum and ovary tissue (n = 5). m Serum estradiol concentration (n = 6). n Ovarian cAMP concentration after treatment with BAY-298 and BAY-899 (n = 5). o, p Representative images of H&E-stained ovaries (p scale bar=1 mm) and Corpora lutea counts (o) (n = 4). Data are represented as mean ± SD, analyzed using one-way ANOVA with multiple comparisons (c, e–g, k–o), two-tailed student’s t-test (d). Source data are provided as a Source Data file.

Functional compensation of evolutionarily related FSHR and LHR

LHR and FSHR are evolutionarily related and conserved GPCRs across the entire vertebrate spectrum (Supplementary Fig. 18). As members of the glycoprotein hormone GPCR family, they are predominantly expressed in the ovary tissue of females (Supplementary Fig. 19). To elucidate, why estrildid finches have evolved to reduce LHR as an adaptation to the FSHR^T449A variation, we treated homozygous FSHR^T449A- KI mice with LHR and FSHR antagonists after administration of PMSG and hCG, followed by comparative analyses to investigate the downstream signaling pathways of LHR and FSHR in the OHSS model (Fig. 5a). As expected, LHR antagonist LUF5771 and FSHR antagonists hFSH-β-(33-53)⁴⁹ significantly countered the OHSS-induced ovary weight increase, serum estradiol and serum and ovarian VEGF levels (Fig. 5b–e, and Supplementary Fig. 17c). Additionally, these antagonists reduced cAMP concentrations and the number of corpora lutea in FSHR^T449A-KI mice (Fig. 5f–h). The therapeutic effect of LUF5771 was superior, with all indicators almost returning to the control levels, in OHSS model of FSHR^T449A-KI mice, comparable to OHSS model of WT mice (Fig. 5c–f and Fig. 4c–g). Ovaries from the OHSS and antagonists treated groups were collected for scRNA-seq analysis. After integration using CCA, all high-quality cells were classified into eight distinct cell types based on canonical markers (Fig. 5i, j, Supplementary Fig. 20). Granulosa cells represented a relatively substantial proportion of cells in the OHSS ovaries (Fig. 5k). Both Fshr and Lhr were expressed in granulosa and theca cells of mouse ovaries (Fig. 5j). After treatment with LUF5771 and hFSH-β-(33-53), the differentially expressed genes in the ovaries when compared to those in OHSS ovaries were predominantly found in theca cells, followed by granulosa cells (Fig. 5l). Furthermore, the ratio of granulosa to theca cell numbers was significantly reduced after treatment of LUF5771 or hFSH-β-(33-53) (Fig. 5k).

By comparing gene expression in granulosa cells between the treatment groups and the untreated OHSS group, we identified 482 differentially expressed genes in the LUF5771-treated mice and 460 DEGs in the hFSH-β-(33-53)-treated mice (Supplementary Fig. 20 and Supplementary Data 5). Venn analysis revealed 227 shared DEGs following treatment with either antagonist, including genes involved in the cAMP pathway, such as AKT serine/threonine kinase 3 (Akt3), proto-oncogene (Jun and Fos), ATPase, Na⁺/K⁺ transporting subunit beta 2 (Atp1b2), Lhr, etc (Supplementary Data 5). The decreased expression of Akt3, Jun and Fos may influence several signaling pathways, including the VEGF, MAPK, TNF, and estrogen signaling pathway (Fig. 5m, and Supplementary Data 5). In addition, the Hippo and TGF-β pathways were also downregulated (Fig. 5m). Using the R package GSEABase and the AddModuleScore function, we further assessed signature scores for angiogenesis and estrogen synthesis pathways. LUF5771 treatment significantly lowered the estrogen synthesis pathway score in both granulosa and theca cells, and reduced the angiogenesis pathway score in granulosa cells. Similarly, TFA treatment decreased the estrogen synthesis pathway score in granulosa cells (Fig. 5n). In the OHSS mouse model, the estrogen synthesis-related gene Star was highly expressed, and treatment with LUF5771 and hFSH-β-(33-53) did not alter its expression (Fig. 5o, and Supplementary Data 5). However, several other genes related to estrogen signaling, such as Akt3, Fos, heparin-binding epidermal growth factor-like growth factor (Hbegf), heat shock protein family A member 1 A (Hspa1a), and Jun, were significantly downregulated following treatment (Fig. 5o, and Supplementary Data 5). Among VEGF pathway genes, the expression of Akt3 and protein phosphatase 3 catalytic A (Ppp3ca) was reduced (Fig. 5p, and Supplementary Data 5). Genes associated with vascular permeability, such as Bmp2 and its receptor Bmpr2, Id1, and Klf6⁵⁰, were all downregulated (Fig. 5q, and Supplementary Data 5). Additionally, the TNF pathway gene Cebpb and the cAMP signaling-related gene Nfkbia showed decreased expression (Fig. 5r, s and Supplementary Data 5). In theca cells, treatment with LUF5771 and hFSH-β-(33-53) resulted in 462 overlapping DEGs, which were mainly enriched in the metabolism of xenobiotics by cytochrome P450, cAMP signaling and HIF-1 pathways, and more (Fig. 5m, and Supplementary Data 5). This suggests that FSHR and LHR share similarities in their downstream signaling pathways.

To assess changes in granulosa cell and theca cell subtypes, we performed re-clustering and annotation using high-quality marker genes reported in the literature. Granulosa cells were classified into distinct subtypes, including an Antral-Mural state (Supplementary Fig. 21a, b). Compared with the OHSS group, treatment with LUF5771 and hFSH-β-(33-53) reduced the number of granulosa cells in atretic follicles, while cells in the preantral follicle state showed signs of recovery (Supplementary Fig. 21c), suggesting that these treatments may help restore normal granulosa cell composition and limit follicle atresia. The expression patterns of differentially expressed genes related to estrogen signaling and vascular permeability further pointed to their potential role in this process (Supplementary Fig. 21d). Theca cells were primarily classified into early-stage cells (involved in androgen synthesis) and late-stage cells (associated with signal transduction, chemotaxis, and cell growth) (Supplementary Fig. 21e–g). Following treatment with LUF5771 or hFSH‑β‑(33‑53), the proportion of early‑stage theca cells decreased relative to the OHSS group (Supplementary Fig. 21f), indicating a possible downregulation of the androgen synthesis pathway that serves as the precursor for estrogen production. Moreover, expression of Bmp2 and Bmpr2 was reduced in both early‑ and late‑stage theca cells after treatment (Supplementary Fig. 21h).

Together, these findings suggest that enhanced function caused by the FSHR variation could be balanced by reducing the activity of the LHR pathway. This represents an important adaptive mechanism in estrildid finches in response to the FSHRT449A variation.

Discussion

Human reproductive processes are primarily regulated by LH and FSH, both acting through their specific GPCRs. Mutations in these receptors can either enhance (activating) or impair (inactivating) receptor function⁵¹. Such alterations have significant clinical consequences: Inactivating mutations in FSHR are associated with primary or secondary amenorrhea, infertility, and premature ovarian failure (POF), whereas activating mutations predispose individuals to OHSS¹³. Notably, the FSHR^T449A/I mutation has been identified in two families presenting with spontaneous OHSS^16,32. In both cases, OHSS occurred during pregnancy, and functional studies revealed that the mutant receptors exhibit markedly increased sensitivity to hCG, providing a mechanistic explanation for the observed phenotype^16,32. Consistently, homozygous FSHR^T449A-KI mice developed more severe OHSS phenotypes than wild-type controls, both under conventional superovulation protocol (Supplementary Fig. 11) and following low-dose hCG to induce ovulation (Fig. 2). Interestingly, this same FSHR-T449A substitution has become fixed in estrildid finches. Cross-species scRNA-seq analysis revealed that genes involved in angiogenesis/vascular development/permeability, such as VEGFA and ADGRG1, were expressed at higher levels in estrildid ovaries compared with non-estrildid species, suggesting that the FSHR-T449A variant may confer an increased risk of ovarian hyperstimulation in these birds (Fig. 3h). Notably, FSHR^T449A-KI mice also exhibited an increased fertility rate (our unpublished observations), consistent with clinical findings that women with multiple mature follicles are more prone to OHSS^52,53 and to multiple pregnancies once conception occurs⁵⁴. This reproductive “trade-off” may explain why the mutation was evolutionarily retained in estrildid finches.

LHR is evolutionarily related to FSHR, both belonging to the glycoprotein hormone receptor family⁵⁵. In our study, we found that LHR expression was reduced in estrildid finches carrying the FSHR^T449A mutation, without impairing their reproductive capabilities, suggesting the presence of compensatory mechanisms to counteract deficiency of LHR function (Fig. 3). In mammals gonadotropin-receptor interactions are highly specific, ensuring precise endocrine regulation under physiological hormone levels⁵⁶. By contrast, studies in non-mammalian species, including teleosts⁵⁷, sea turtles⁵⁸ and chickens⁵⁹, indicate a less stringent specificity of gonadotropin receptors. In teleosts, for example, the FSH receptor displays ligand promiscuity and can be cross-activated by LH, whereas the LH receptor remains highly selective, being activated only by its cognate ligand⁵⁷. In situ hybridization studies further revealed overlapping expression of FSHR and LHR in granulosa cells of vitellogenic follicles in tilapia and in full-grown follicles of amago salmon¹⁵. Functional analyses also support the notion of partial redundancy between LH and FSH signaling: both hormones stimulate estradiol (E₂) production with equal potency during vitellogenesis in several teleosts, and in goldfish, both FSH and LH are equally effective in inducing germinal vesicle breakdown (meiotic maturation)⁶⁰. Consistently, Xie et al. demonstrated in mutant zebrafish that LH and FSH signaling can functionally compensate for each other via receptor cross-activation⁶¹. These findings collectively support our hypothesis that the p.Thr449Ala substitution in estrildid finche FSHR confers enhanced receptor promiscuity, allowing activation by both FSH and LH, thereby increasing sensitivity to hCG stimulation. This gain-of-function effect likely amplifies downstream FSH pathway activity and could, in theory, predispose to OHSS-like responses. However, the concurrent reduction of LHR expression in estrildids may serve as a compensatory mechanism, maintaining elevated local cAMP levels while dampening excessive downstream signaling triggered by the FSHR-T449A variant. This balance likely prevents pathological ovarian overstimulation despite the heightened receptor sensitivity (Fig. 6). Given previous reports linking glycoprotein hormone glycosylation patterns to receptor specificity⁶², we further analyzed the N-glycosylation profiles of LH and FSH in estrildid finches. No species-specific glycosylation differences were detected, indicating that the observed promiscuity of estrildid FSHR is primarily attributable to the p.Thr449Ala substitution rather than to changes in hormone glycosylation (Supplementary Fig. 22).

Fig. 6. Functional cAMP pathway compensation between evolutionarily related FSHR and LHR illuminates the molecular mechanism and OHSS treatment options.

LHR and FSHR are glycoprotein hormone receptors expressed in ovarian granulosa cells. In non-mammalian species, FSH can activate both receptors, while LH is more specific, it can also activate FSHR in Estrildidae and mammalian variants. In mammals (phylogenetic tree not drawn to scale), FSHR responds exclusively to FSH, while LHR is activated by LH. The p.Thr449Ala (T449A) variation in FSHR reduces ligand specificity, allowing activation also by LH and increasing sensitivity to hCG. Avian FSHR exhibits constitutive activity, which is further enhanced by the T449A variation, enabling dual activation by FSH and LH, thereby increasing the risk of OHSS. Reduced LHR expression helps to maintain ovarian cAMP levels, offering insights into adaptive strategies and potential therapeutic targets for OHSS. All animal silhouettes were sourced from phylopic.org.

Unlike mammals, which possess a pair of ovaries and oviducts, most female birds have only a single functional left ovary and oviduct. In birds, follicular growth begins at puberty and proceeds in an orderly, sequential manner during each subsequent breeding season. At the onset of egg laying, follicles at various developmental stages coexist, and follicular atresia is rarely observed. By contrast, in placental mammals, once a dominant follicle is selected during each estrous or menstrual cycle, all subordinate follicles within the same cohort rapidly undergo atresia⁶³. Despite these structural and physiological differences, the final stages of follicular growth and differentiation in both mammals and birds are highly dependent on GPCR-mediated signaling, particularly through the cAMP-dependent FSHR and LHR pathways⁶⁴. In chickens, FSHR expression begins in granulosa cells of small follicles (1-2 mm in diameter), but FSH stimulation does not significantly increase intracellular cAMP levels until after follicular selection⁶⁵. As follicles mature, FSHR mRNA levels rise, enabling granulosa cells to acquire the capacity for robust progesterone secretion, which triggers the LH surge and initiates ovulation^66,67. Moreover, experimental studies have shown that FSH facilitates follicle selection in hens, and treatment with PMSG or FSH increases the number of developing follicles⁶⁸, consistent with our observations in Fig. 2p. These findings indicate that the FSH-LH system plays comparable roles in follicle development, selection, and ovulation in both birds and mammals. Therefore, studying ovarian adaptive mechanisms in birds may provide valuable insights into human reproductive disorders.

In conclusion, our findings suggest that estrildid finches have evolved adaptive mechanisms to mitigate the risk of OHSS arising from FSHR gain-of-function mutations in ovarian granulosa cells. Specifically, this adaptation involves the downregulation of genes associated with OHSS, notably LHR, which serves to balance excessive activation of the FSHR signaling pathway. The study presented here is a new example of Darwinian medicine exploiting mechanisms underlying the adaptive evolution in Estrildidae for potential treatment of the human genetic disease.

Our results establish a conceptual and experimental foundation for targeting LHR as a therapeutic strategy to prevent or mitigate OHSS. Importantly, this approach appears effective not only under constitutively active FSHR conditions but also in clinically induced OHSS during standard IVF protocols, suggesting broad translational relevance. While current evidence supports reduced LHR expression as one mechanism conferring protection against OHSS, additional, yet unidentified mechanisms may also contribute, potentially even more significant in certain contexts. These possibilities warrant further investigation in future studies.

Methods

Animals

zebra finches (Taeniopygia guttata), Bengalese finches (Lonchura striata domestica), Gouldian finches (Erythrura gouldiae), common canaries (Serinus canaria), budgerigars (Melopsittacus undulatus), Japanese white-eyes (Zosterops japonicus), white Java sparrows (Padda oryzivora) and chickens (Gallus gallus) were bought from Flower and Bird Market (Nanjing, China). Birds were maintained in a temperature-controlled (21-23 °C) and humidity-controlled (48% average) vivarium under a 12/12 h light/dark cycle. Animals were housed at a density of five per cage (160 × 50 × 40 cm). Female C57BL/6 mice were obtained from GemPharmatech Co., Ltd., Jiangsu, China. FSHR^T449A-KI mice (The hsFSHR 449 site corresponds to the homologous site 448 in mice, for consistency, we refer to it as 449A throughout the article) were generated on a C57BL/6 background using CRISPR-Cas9-mediated gene editing. Two pairs of gRNAs were designed to induce double-strand breaks, and a donor vector containing the “KI region-p.T448A (ACT to GCT)” cassette, along with Cas9 mRNA, was co-injected into fertilized mouse zygotes to produce the desired knock-in offspring. Founder (F0) animals were identified through PCR and confirmed by sequence analysis. These F0 mice were then bred with wild-type mice to evaluate germline transmission and generate F1 offspring. Details of the gRNA sequences and PCR primers used for sequencing are provided in Supplementary Data 6. The homozygous and WT littermates used in this study were produced by breeding heterozygous parents. Mice were housed in plastic cages in specific pathogen free (SPF) grade animal rooms. They were kept under a 12 h light/dark cycle-controlled atmosphere, with ambient temperature of 22 ± 2 °C and relative humidity of 50 ± 10%. Standard chow and water were available ad libitum. All animal experimental and care procedures were performed in accordance with the guidelines of Animal Management Regulations of the Ministry of Health, China (Document No. 55, 2001) and were approved by the Nanjing Normal University Animal Faculty [IACUC- 20210233].

Reagents

The in vivo and in vitro experiments utilized hCG (Anhui Fengyuan Pharmaceutical Co., Ltd), recombinant human FSH (Gonal-f, Serono Pharma S.p.A., Bari, Italy) and recombinant human LH (Luveris, Serono Pharma S.p.A., Bari, Italy) were generously provided as gifts by Dr. Chen Li of Jinling Hospital (Nanjing, China). PMSG were purchased from Ningbo Second Hormone Factory (Ningbo, China). Chicken FSH and estradiol (JL15972) ELISA kit was purchased from Shanghai Jianglai Biotechnology Co., Ltd, China. Mouse and rat VEGF (YJ002076, YJ064294) and estradiol ELISA (YJ001962, YJ002871) kit were purchased from Shanghai Enzyme-linked Biotechnology Co., Ltd, China. LHR allosteric inhibitor LUF5771, and antagonist BAY-298 was synthesized by Yanshen Technology Co., Ltd., (Jilin, China) and WuXi AppTec Co., Ltd., (Wuxi, China), respectively. Progesterone antagonist RU486 (84371-65-3), LHR antagonist BAY-899 (2471967-92-5) were purchased from Sigma. FSHR antagonist hFSH-β-(33-53) (HY-P3343A) was obtained from MedChemExpress (MCE, Shanghai, China).

Phylogenetic Analysis of FSHR in Vertebrates

The FSHR coding sequences were downloaded from NCBI, translated into amino acid (aa) sequences, and aligned with MAFFT (v7.515) using default settings. A maximum-likelihood (ML) phylogenetic tree was constructed by IQ-TREE (v2.2.2.7) based on multiple sequence alignments with the options -m MFP -B 1000 --bnni -T AUTO. Positive selection amino acid positions were detected using a branch-site model of PAML (Phylogenetic Analysis by Maximum Likelihood) version 4.9. The estrildid branch was set as the foreground branch, and other species were set as the background branch. Specifically, we used Model A to test for positive selection on the foreground branch. This analysis compares the alternative hypothesis, in which sites on the foreground branch are allowed to evolve with ω > 1, against the null hypothesis, in which ω is fixed at 1. Sites under positive selection were identified using the Bayes Empirical Bayes (BEB) method with a posterior probability ≥ 0.8. For the FSHR coding sequences of E. gouldiae and P. oryzivora, we performed PCR and tblastn on the RNAseq de novo assembled using trinity.

To investigate the distribution of amino acid at the FSHR 449 sites across all species with genomes cataloged in the NCBI database, we utilized the Human FSHR protein sequence as a query. Subsequently, we performed both blastp and tblastn. From the blast results, we manually curated and confirmed the orthologous FSHR genes. Further, we conducted sequence alignments using the MAFFT (v7.515) software, extracting the corresponding amino acid sequences for alignment. The construction of the phylogenetic tree encompassing all species was carried out using the NCBI taxonomy database (https://www.ncbi.nlm.nih.gov/taxonomy/), and then visualized by iTOL (https://itol.embl.de/). All animal silhouettes were sourced from phylopic.org.

Prediction of N-linked Glycosylation Sites on FSH and LH

The amino-acid sequences of the target ligands (CGA, FSHB, and LHB) were used to generate three-dimensional structural models with the AlphaFold 3 server (http s://alphafoldserver.com/), resulting in high-confidence protein structures. These models were subsequently analyzed using the GlycoShape online platform (https://glycoshape.org/). Following the methodology described in the associated article⁶⁹, automated glycosylation analysis was performed using its core Re-Glyco algorithm. To cross-validate and complement the structure-based predictions, we additionally employed the sequence-based prediction tool NetNGlyc 1.0 (https://services.healthtech.dtu.dk/services/NetNGlyc-1.0/). Amino acid sequences in FASTA format were submitted to the NetNGlyc server using default parameters to independently assess potential N-glycosylation sites based on primary-sequence features.

Plasmids used for luciferase assay

Full-length FSHR (including point-mutations) and LHR of various species, G. gallus (gg)CGA, ggLHB and ggFSHB, were synthesized by General Bio Co., Ltd (Anhui, China) based on NCBI sequences, and then cloned into pcDNA3.1-V5/His vector with C-terminal V5-His tag. The serum response element (SRE), cAMP response element (CRE), T-lymphocyte activation of nuclear factor-responsive element (NFAT-RE) and serum response factor response element (SRF-RE) luciferase reporter plasmids were obtained from Promega.

Cell culture

HEK293 cells were obtained from Fuxiang biotechnology (Shanghai, China) and maintained in Dulbecco’s modified Eagle’s medium (DMEM, Thermo-Fisher, 11965092), supplemented with 10% fetal bovine serum (FBS, Gibco, A5256701), 1% penicillin-streptomycin (Gibco, 15140122), in a 37 °C incubator in 5% CO₂ atmosphere.

Luciferase assay

HEK293 cells were placed into 24-well dishes and transfected with the indicated amounts of receptors together with 50 ng of CRE-luciferase reporter plasmid using lipofectamine 6000 regent (Beyotime, C0526, Shanghai, China). Six h after transfection, the medium was replaced with fresh serum-containing medium and the constitutive activity was detected using the assay kit (Beyotime, RG005, Shanghai, China) 48 h after transfection by a EnVision™ Multilabel Plate Reader (PerkinElmer, USA). CRE, NFAT-RE, SRF-RE and SRE were utilized for testing Gs, Gq, G12 and potential Gi pathway respectively.

For ligand stimulation (hCG, hLH and hFSH), the culture medium was replaced with serum-free medium 36 h after transfection and, at the same time, cells were given the corresponding doses of ligands. Luciferase activity was measured 12 h after ligand stimulation.

For ggFSH and ggLH stimulation assays, HEK293 cells were seeded into 48-well plates and co-transfected in serum-free medium using a lipofection reagent (Yeasen, 40802ES03, Shanghai, China). Equal amounts (50 ng each) of receptor plasmid and cAMP response element (CRE)-luciferase reporter plasmid were used, with the empty vector pcDNA3.1 serving as a negative control. Six h after transfection, the medium was replaced with fresh serum-containing medium, and the cells were further cultured. At 24 h post-transfection, ggFSHB/ggLHB and ggCGA plasmids were co-transfected at a 1:1 ratio in serum-free medium at gradient concentrations (2, 5, 10, 20, 50, 100, and 200 ng) to assemble and stimulate the receptors. After 6 h, the medium was once more replaced with fresh medium, and the cells were cultured for an additional 18 h. Cells were then collected, and the supernatant was discarded. Luciferase activity was measured using a luciferase assay kit (Beyotime, RG005, Shanghai, China). All experiments were performed independently at least three times.

cAMP detection

For in vitro studies, HEK293 cells were seeded into 24-well plates and transfected with indicated amount of plasmid. After 48 h of transfection, the culture medium was changed to 200 μl serum-free medium and incubated for 30 min, followed by washing with 100 μl of PBS and digestion with 50 μl trypsin. Cells were resuspended in serum-free medium containing 0.3% BSA and 3-isobutyl-1-methylxanthine (IBMX, MCE, HY-12318, Shanghai, China), and 5 μl of cell suspension transferred to 384-well plate for analysis.

For in vivo studies, 0.1 g of ovary tissue was homogenized on ice in RIPA buffer containing IBMX, followed by centrifugation at 4 °C, 4700 × g for 30 seconds. The supernatant was placed on ice for 5 min, followed by centrifugation at 4 °C, 20100 × g for 10 min. Five μl of the supernatant was taken for analysis in a 384-well plate. The detection procedure was done in accordance with the instructions of the cAMP Gs dynamic kit (CSBIO, 62AM4PEB).

Rat OHSS induction and pharmacological intervention

All animal protocols were approved by the committee on the Ethics of Animal Experiments at Nanjing Normal University. To induce OHSS, three-week-old female Wistar rats were housed under a 12-h light/dark cycle at 22°C with free access to food and water. The OHSS group was treated with daily intraperitoneal injections of 50 IU PMSG for four consecutive days, followed by a single injection of 20 IU hCG on the fifth day. The test compounds (BAY-298 and BAY-899) were dissolved in a vehicle consisting of 15% DMSO, 15% Tween-80, and 70% saline, and administered intraperitoneally at doses of 4.5 mg/kg and 14 mg/kg, respectively. Drug treatment began 24 h after hCG injection and continued for two consecutive days. Rats receiving an equal volume of the vehicle alone served as the solvent control group. An additional conventional superovulation group (non-OHSS control) received a single intraperitoneal injection of 10 IU PMSG, followed 48 h later by 10 IU hCG. All rats were humanely euthanized by decapitation under isoflurane anesthesia 24 h after the last drug or solvent administration.

Mouse OHSS induction and pharmacological intervention

The OHSS model in C57BL/6 mice was established using a protocol similar to that for rats. Five-week-old female C57BL/6 mice were used. The OHSS group received daily intraperitoneal injections of 20 IU PMSG for four consecutive days, followed by a single injection of 5 IU hCG on the fifth day. For adult (8-week-old) female mice, the induction protocol was modified by increasing the PMSG dose to 30 IU per day to account for increased body weight, while the hCG dose remained unchanged at 5 IU. LUF5771 and RU486 were dissolved in a vehicle consisting of 15% DMSO, 15% Tween-80, and 70% saline, and administered intraperitoneally at doses of 20 mg/kg. Drug treatment began 24 h after hCG injection and continued for two consecutive days. Mice receiving an equal volume of the vehicle alone served as the solvent control group. An additional conventional superovulation group (non-OHSS control) received a single intraperitoneal injection of 5 IU PMSG, followed 48 h later by 5 IU hCG⁷⁰. All mice were humanely euthanized by decapitation under isoflurane anesthesia 24 h after the last drug or solvent administration.

For FSHR^T449A-KI mice, which have an elevated risk of developing OHSS due to the point mutation, the hormone dosages were adjusted accordingly. The OHSS group (five-week-old) was injected with 20 IU of PMSG for four consecutive days, followed by a single injection of 1 IU hCG on the fifth day. For adult (8-week-old) female mice, the induction protocol was modified by increasing the PMSG dose to 30 IU per day, while the hCG dose remained unchanged at 1 IU. LUF5771 and hFSH-β-(33-53) were administered intraperitoneally at doses of 20 mg/kg. Drug treatment began 24 h after hCG injection and continued for two consecutive days. Mice receiving an equal volume of the vehicle alone served as the solvent control group. An additional conventional superovulation group (non-OHSS control) received a single intraperitoneal injection of 5 IU PMSG, followed 48 h later by 1 IU hCG. All mice were humanely euthanized by decapitation under isoflurane anesthesia 24 h after the last drug or solvent administration.

Avian ovarian hyperstimulation induction

Adult female birds were intraperitoneally injected with PMSG at a dose of 20 IU for four consecutive days, followed by a single intraperitoneal injection of 5 IU hCG or hFSH on the fifth day to induce OHSS. The control birds were treated with 5 IU PMSG once and intraperitoneally injected with 5 IU hCG or FSH 48 h later. All birds were humanely euthanized by decapitation under isoflurane anesthesia 24 h after hCG administration.

Hematoxylin-eosin staining

The ovaries were fixed overnight in 4% paraformaldehyde (PFA) and subsequently embedded in paraffin. Serial sections (5 μm in thickness) were obtained from the largest cross-sectional region, yielding nine consecutive sections per ovary. The sections were stained with haematoxylin and eosin (H&E) for morphological assessment and quantification of corpora lutea. For corpus luteum counting, the nine consecutive H&E-stained sections were analysed as a continuous series. Each section was examined in reference to its preceding and succeeding section to trace the structural continuity of individual corpora lutea and to prevent redundant counting. The average number of corpora lutea per ovary was used for comparison among different treatment groups.

ELISA

For hormone detection, blood was collected and centrifuged at 1260 ×g for 10 min. Serum concentrations of LH, FSH, and estradiol were determined using the enzyme-linked immunosorbent assay (ELISA) following the instructions of the suppliers.

For total FSHR protein expression detection, HEK293 cells were placed into 24-wells dishes and transfected with the indicated amounts of receptors using lipofectamine 6000 reagent. The supernatant was removed 48 h after transfection and the cells fixed with 150 μl of 4% PFA for 15 min, and then treated with 150 μl of 0.1% Triton-X-100 for 15 min. After PBS washing, cells were blocked with 1% BSA for 1 h, incubated with V5 antibody (CST, 13202S, 1:1000 dilution, USA) for 2 h and then with HRP- conjugated secondary antibody (CST, 7076, 1:1000 dilution, USA) for 1 h, and finally 200 μl of TMB (Beyotime, ST746, Shanghai, China) were added. After incubation for 30 min at room temperature, the absorbance value was measured at 370 nm using BioTek Synergy H1.

Real-time PCR (qRT-PCR)

Total RNA was extracted using TRIzon reagent (CWBIO, CW0580S, Beijing, China) and 1 μg RNA was reverse-transcribed into cDNA using HiScript IV All-in-One Ultra RT SuperMix for qPCR (Vazyme, R433-01, Nanjing, China) as per manufacturer’s instructions. For real-time PCR, cDNAs were amplified using SYBR Green mixture (Vazyme, Q312-02, Nanjing, China), and each primer pair was amplified in an ABI 7500 real-time PCR system (Applied Biosciences). Results were semi-quantified with GAPDH gene expression levels as an internal reference. Primers used in this study are shown in Supplementary Data 6.

Single-cell capture, library construction and next generation sequencing

For avian ovaries, scRNA‑seq samples were processed as single organs (one ovary per bird). The T. guttata (tg) and S. canaria (sc) groups each contain n = 3 biological replicates, while the L. striata domestica (ls) group contains n = 2 biological replicates. Due to the limited cell yield from a single mouse ovary, ovaries were pooled prior to single‑cell suspension preparation. Specifically, the OHSS group consisted of n = 1 biological replicate, generated by pooling both ovaries from one mouse. The LUF5771‑treated and hFSH‑β‑(33-53)-treated groups each contained n = 2 biological replicates. Each replicate was prepared by pooling both ovaries from each of two mice, resulting in a total of four ovaries per replicate. All animal ovaries were dissected and trimmed of excess impurities in culture dishes containing DPBS (Gibco, 14287080). The digestion protocol was similar in all species. The single-cell suspension of ovaries was generated by 2 mg/ml collagenase Type II/IV (1:1) digestion for 30 min at 37 °C. The supernatant was removed by centrifugation at 11300 ×g for 10 seconds at 4 °C and immediately placed on ice, then 0.04% BSA (Sigma-Aldrich, 9048-46-8)-DPBS was added for resuspension. Finally, the single-cell suspension was stained with trypan blue to assess cell number and viability (>80%) and used to construct scRNA-seq libraries with the Chromium Single Cell 3ʹ Gel Bead-in-Emulsion (GEM), Library & Gel Bead Kit v3 (10× Genomics, Cat:1000121), according to the manufacturer’s instructions as briefly described: The single-cell suspension was loaded onto a 10× Chromium system (10× Genomics). Following gel bead-in-emulsion generation, cDNA synthesis proceeded using the Single Cell 3’ Reagent Kit v3 (10× Genomics). The cDNA libraries were PCR amplified followed by fragmenting and sequencing on the Illumina NovaSeq 6000 platform.

Processing of scRNA-seq data

CellRanger software (v7.0.1) was used to process, align, and summarize unique molecular identifier (UMI) counts against bTaeGut1.4 (T. guttata), cibio_Scana_2019 (S. canaria), lonStrDom2 (L. striata domestica) and GRCm39 (M. musculus). Ovary libraries were processed with SoupX (v0.3.1)⁷¹ to remove ambient RNA. Filtered UMI count matrices were imported into R for downstream analysis. To obtain high-quality data for all species, different quality control parameters have been set: S. canaria (sc), cells with over 20% UMIs of mitochondrial RNA, below 500 total UMIs and above 4000 total UMIs were removed; T. guttata (tg): cells with over 20% UMIs of mitochondrial RNA, below 200 total UMIs and above 7000 total UMIs were removed; L. striata domestica (ls): cells with over 10% UMIs of mitochondrial RNA, below 200 total UMIs and above 7000 total UMIs were removed; M. musculus (mm): cells with over 15% UMIs of mitochondrial RNA, below 200 total UMIs and above 7500 total UMIs were removed. Variable features were identified in Seurat before scaling data and running PCA. Gene counts were normalized using the NormalizeData method and 2000 highly variable genes were selected for downstream analysis. We applied the Canonical Correlation Analysis (CCA) method for dataset integration to correct batch effects. The union of the top 2000 genes in each dataset with the highest dispersion was taken, and anchors were determined using the FindIntegrationAnchors function. Then we applied the IntegrateData function to generate an integrated expression matrix. We scaled the data and applied the RunPCA function for dimensional reduction. Cells were clustered using the official Seurat pipeline and visualized using uniform manifold approximation and projection (UMAP). Cluster markers were found using FindAllMarkers and FindMarkers with the Wilcoxon rank sum test.

For the re-clustering of ovarian granulosa and theca cells, respective cell populations were extracted from corresponding clusters across the OHSS, LUF5771-, and hFSH-β-(33-53)-treated groups. Datasets were grouped by condition, re-normalized, and integrated using the FindIntegrationAnchors and IntegrateData functions in Seurat. Following integration, cells were re-clustered and visualized in UMAP using the top 2000 highly variable genes and the first 20 principal components. Positively expressed differentially expressed genes (DEGs) for each cluster were identified with the FindAllMarkers function. Pathway activity for estrogen response and angiogenesis was scored and visualized using the AddModuleScore function from Seurat, with gene sets obtained via the R package GSEABase (v.1.68) and referenced against the mouse database.

Bulk RNA-seq and analysis

The avian ovaries were processed and analysed as single organs (one ovary per bird). No pooling was performed. For the comparative analysis presented in Supplementary Fig. 16, each group contains n = 3 biological replicates. Total RNA was extracted from each sample using RNAiso Plus reagent according to the manufacturer’s instructions. We estimated the integrity and quality of the total RNA using a Bioanalyzer 2100 system (Agilent Technologies, Palo Alto, CA, USA) and an RNA 6000 Nano kit. Sixty poly (A) RNA-seq libraries were constructed and sequenced using the BGISEQ DNBSEQ-T7 platform (BGI lnc., Shenzhen, China) with a paired-end sequencing length of 150 bp (PE150) at Novogene Bioinformatics Technology Co., Ltd (Beijing, China). The quality of the raw data was determined using the FastQC software⁷² and the adapters were trimmed using Trimmomatic⁷³. The clean reads were aligned to corresponding reference genomes from Ensembl database (tg: bTaeGut1_v1; ls: LonStrDom1; sc: SCA1.107; gg: GRCg7b) by HISAT2 (v2.2.1) with default parameters by HISAT2 (v2.2.1)⁷⁴. SAMTools was used to sort the bam files of the aligned reads⁷⁴. The transcripts were assembled by StringTie (v1.3.6) with parameters to assess gene expression based on the TPM (Transcripts Per Million) values of each mRNA⁷⁵.

To compare transcription rates across species, we downloaded the homologous gene lists among three species from Ensemble BioMart (http://www.ensembl.org/biomart/martview/8031c97af5aea7bfb70ce3f47e39e344). As a Lonchura striata domestica dataset was not available in Ensembl Biomart, we performed reciprocal blast of the two proteomes (T. guttata and L. striata domestica) to identify orthologous genes. We combined the data with the homologous gene lists from Ensembl Biomart to form a one-to-one homologous gene set shared by the four bird species. The DEGs based on the read count data from these orthologous gene pairs were identified using DESeq2 package⁷⁶. The significant DEGs were screened with a false discovery rate <0.05 and |log2 fold change | >1 as cutoffs. To investigate the expression patterns of DEGs across species, we used transcripts per million 10 K (TPM10K), a metric which normalizes TPM to account for different sequencing depths among species⁷⁷.

For the RNA-seq de novo assembly, we used trimmomatic filters raw readings to remove adapters and low-quality sequences. The de novo assembly was performed by Trinity (v2.15.1) using the default parameters. Then BLAST software (v2.13.0 + ) was used to generate P. oryzivora FSHR sequences from the assembly results as a database, and L. striata domestica FSHR sequence as a reference.

Functional enrichment analyses were performed using Metascape (http://metascape.org) with default parameters⁷⁸. All genes were converted to human orthologs by Ensembl Biomart, and the target gene lists were uploaded as inputs for enrichment. We chose H. sapiens as the target species, and enrichment analysis was performed against all genes in the genome as the background set, with the biological process (BP) of Gene Ontology (GO) as the functional test set. Only GO terms with a P value < 0.01 and annotated to ≥3 genes were considered significant.

Statistics

All experiments were repeated independently at least three times. The data were analyzed using GraphPad Prism 8.3 and represented as the mean ± SD from at least three independent experiments, performed in triplicates. No statistical method was used to predetermine sample size. No data were excluded from the analyses. The experiments were randomized. The investigators were not blinded to allocation during experiments and outcome assessment. Differences between the two groups were compared using two-tailed Student’s t-test. Comparisons between multiple groups were performed using one-way ANOVA. Statistical significance was defined as a p < 0.05.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

C.D and S.L conceived and supervised the whole project; S.L., Y.H., A.D., H.H., X.Y., S.W., S.M., Q.R., and J.Y conducted experiments; Y.H., P.H., and A.D performed the RNA and single-cell sequencing analysis; S.L., Y.H., A.D., H.H., S.M., J.L., Y.Z., X.Z., and C.D performed data analysis and interpretation; S.L., Y.H., and C.D. wrote the paper; J.B., S.L., and C.D. revised the paper.

Peer review

Peer review information

Nature Communications thanks Adolfo Rivero-Muller and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The single-cell RNA-seq data of bird ovaries, the single-cell RNA-seq data of KI mouse ovaries, and the transcriptome data of bird ovaries generated in this study have been deposited in the NCBI database under accession codes PRJNA1432434, PRJNA1432435, and PRJNA1432433. The Differential gene expression and functional enrichment data generated in this study are provided in the Supplementary Information/Source Data file. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-71338-7.