FSH blockade improves cognition in mice with Alzheimer’s disease

Abstract

Alzheimer’s disease has a higher incidence in older women, with a spike in cognitive decline that tracks with visceral adiposity, dysregulated energy homeostasis and bone loss during the menopausal transition^1,2. Inhibiting the action of follicle-stimulating hormone (FSH) reduces body fat, enhances thermogenesis, increases bone mass and lowers serum cholesterol in mice^3–7. Here we show that FSH acts directly on hippocampal and cortical neurons to accelerate amyloid-β and Tau deposition and impair cognition in mice displaying features of Alzheimer’s disease. Blocking FSH action in these mice abrogates the Alzheimer’s disease-like phenotype by inhibiting the neuronal C/EBPβ–δ-secretase pathway. These data not only suggest a causal role for rising serum FSH levels in the exaggerated Alzheimer’s disease pathophysiology during menopause, but also reveal an opportunity for treating Alzheimer’s disease, obesity, osteoporosis and dyslipidaemia with a single FSH-blocking agent.

Alzheimer’s disease (AD) is a neurodegenerative disorder of the aging population that shows a higher incidence in women after the menopause⁸. With the paucity of disease-modifying agents, AD poses a major global health crisis, resulting in progressive dementia, profound disability and impaired quality of life. Neuropathological hall-marks include aggregated amyloid-β (Aβ) peptides and Tau proteins and chronic inflammation. Constituting around 70% of individuals with AD, women have a greater life-time risk for AD than men, and display approximately a threefold higher rate of disease progression with a broader spectrum of cognitive symptoms^9,10. The cause of this gender difference is unclear. A role for post-menopausal reductions in serum oestrogen remains controversial; improvement¹¹, no change^12,13 and worsening¹⁴ of cognition have been shown after oestrogen-replacement therapy. By contrast, high serum levels of pituitary gonadotropins, including FSH, are strongly associated with the onset of AD and have therefore been suggested as possible mediators^15,16. There is also a period in a woman’s life before the last menstrual period when oestrogen levels are relatively unperturbed, while serum FSH is rising sharply¹⁷. During this perimenopausal phase, which normally occurs between the ages of 42 and 52 years¹⁷, there is a transient decline in cognition, prominently verbal memory^18–20. This striking oestrogen-independent correlate prompted us to investigate whether FSH has a causal role in AD pathogenesis, and whether blocking FSH action may provide a means of ameliorating AD.

FSH-Ab prevents the AD phenotype

Our anti-FSHβ antibody (FSH-Ab) targets a 13-amino-acid mouse FSHβ sequence, LVYKDPARPNTQK, blocks FSH action on bone cells and adipocytes and, in doing so, increases bone mass, decreases body fat and elevates energy expenditure in mice^5,6,21. Here we show that FSH-Ab inhibits plaque and neurofibrillary tangle formation and reverses cognitive decline in ovariectomized 3xTg-AD (3xTg) mice (Fig. 1). These mice have a human transgene with mutated APP ^K670N/M671L and MAPT^P301L, and a Psen1^M146V knockin mutation, and display associative learning deficits at 3–5 months, Aβ plaques, impaired spatial working memory and retrieval in the Morris water maze test at around 6 months, and neurofibrillary tangles at around 12 months^22,23—all of these features are accelerated after ovariectomy²⁴. Four days after the ovariectomy or sham operation, mice (aged 3.5 months) received either FSH-Ab or goat IgG (200 μg per mouse, every 2 days, intraperitoneally (i.p.)) for 8 weeks. Besides displaying uterine atrophy and elevated FSH levels, ovariectomized mice showed high levels of Aβ and Tau in the hippocampus, and Aβ40 and Aβ42 in whole-brain extracts (Fig. 1a, b and Extended Data Fig. 1a, b). Brain pathology was substantially reduced in mice that were treated with FSH-Ab.

Fig. 1 |. FSH-blocking antibody reverses AD neuropathology and cognitive decline in Alzheimer’s mice.

a–e, The effects of FSH-blocking antibody (FSH-Ab) or goat IgG after ovariectomy (OVX) or sham operation (Sham) of 3xTg mice (aged 3 months) on hippocampal Aβ, pTau and proteinaceous deposits (a); Aβ40 and Aβ42 (b); C/EBPβ, AEP, cleaved APP, Tau and pTau (c); AEP activity (d); and parameters of spatial memory determined using Morris water maze testing (e). For a, scale bars, 50 μm. f, Separate experiments show the effect of FSH-Ab or goat IgG given to male APP/PS1 mice (aged 5 months) over 4 months on hippocampal and cortical Aβ40 and Aβ42. Data are mean ± s.e.m. n = 6 (b and d); n = 9 (e); and, from left to right, n = 9, n = 10, n = 4 and n = 4 (f) mice per group. Statistical analysis was performed using one-way analysis of variance (ANOVA) (b, d and e) or unpaired two-tailed Student’s t-tests (f). Gel source data are provided in Supplementary Fig. 1.

Ovariectomy also strongly induced the expression of the transcription factor C/EBPβ in hippocampal neurons (Fig. 1c and Extended Data Fig. 1c). C/EBPβ is known to activate arginine endopeptidase (AEP), a δ-secretase that cleaves amyloid precursor protein (APP) at residues Asn373 and Asn585 and Tau at Asn368 to promote Aβ and Tau aggregates, respectively. Deletion of AEP from 5xFAD or MAPT ^P301S mice ameliorates amyloid plaques and neurofibrillary tangles, respectively, and rescues the memory deficits in Morris water maze testing^25,26. Treatment with FSH-Ab caused an impressive reduction in ovariectomy-induced increases in expression and activation of the C/EBPβ–AEP/δ-secretase pathway, APP and Tau cleavage, and Tau phosphorylation in 3xTg mice (Fig. 1c, d and Extended Data Fig. 1b, c). FSH-Ab also reversed the ovariectomy-induced enhancement in neuronal apoptosis and reductions in dendritic spine and synapse numbers (Extended Data Fig. 1d–f). Morris water maze testing showed that ovariectomy induced a marked deficit in spatial learning and memory retrieval, evident from the increased latency of mice to mount a platform and the time spent in the platform quadrant, with no effect on motor activity (swim speed)—the altered variables were reversed in FSH-Ab-treated ovariectomized mice (Fig. 1e and Extended Data Fig. 1g).

We recognize that acute oestrogen withdrawal post-ovariectomy does not replicate the gradual menopausal transition in women, but does mirror the hormonal changes—notably reduced oestrogen and elevated FSH levels—that characterize the late menopause. The cognitive decline (spatial memory impairment) after ovariectomy also mimics that noted in women after surgical or chemical menopause²⁷, when FSH levels rise considerably—again providing evidence for direct translation of our findings to this group of women. Furthermore, with no other mammal, except for certain whale species that undergo menopause, the ovariectomized mouse, albeit imperfect, has been used widely for preclinical testing of almost all menopause-related drugs²⁸.

Following our own standards for ‘contemporaneous reproduction’ of datasets in another laboratory^6,29, we replicated the 3xTg data at Mount Sinai using APP/PS1 mice, which have a human transgene comprising APP ^K670N/M671L and PSEN1^ΔE9 (ref.³⁰). The mice present with amyloid plaques at around 6 months and overt cognitive impairment at around 12 months^30–32. We i.p. injected male APP/PS1 mice (aged 5 months) with FSH-Ab (escalating doses from 120 to 150 μg per mouse, 5 days per week) for 4 months. FSH-Ab prevented Aβ40 and Aβ42 accumulation in both the hippocampus and cortex (Fig. 1f). However, noting that APP/PS1 mice display a defect in recognition memory at around 15 months^23,31, as expected, we found no evidence of impaired novel-object discrimination or its alteration after treatment with FSH-Ab at 9 months (Extended Data Fig. 1h). Nonetheless, it was clear that the suppression of basal levels of serum FSH by FSH-Ab at very least prevented Aβ accumulation in male mice. Whether or not this translates to middle-aged female mice is unclear, but it seems to be highly probable as there is no known difference between male and female mice in serum FSH levels, in contrast to in humans³³.

Post-menopausal luteinizing hormone (LH) levels correlate with a higher incidence of AD^15,16 and LH has been found to impair cognition through a direct action on hippocampal LH receptors (LHCGRs)^34–36. However, although we cannot rule out a role for high LH in AD pathogenesis on its own, we find it very unlikely that changes in LH signalling contribute to the rescue of AD pathology by FSH-Ab in ovariectomized 3xTg mice. First, our highly FSH-specific antibody does not cross-react with LH in vitro (Extended Data Fig. 1i). Second, LHCGR content in the brain of FSH-Ab-treated ovariectomized 3xTg mice is identical to that in IgG-treated mice (Extended Data Fig. 1j). Third, serum LH levels remain unchanged after treatment with FSH-Ab or FSH (Extended Data Figs. 1a and 5b). Finally, in pharmacokinetic studies, blocking FSH action using our monoclonal humanized anti-FSHβ antibody (Hu6) did not alter serum LH, activin or inhibin levels³⁷.

The FSH-blocking action of FSH-Ab appears to be exerted both centrally and peripherally, in the latter case, making less free FSH available to act on brain FSH receptors (FSHRs). We found that FSH and FSH-Ab, when injected peripherally, both cross the blood–brain barrier and localize in brain tissue. AlexaFluor750-labelled FSH, injected intravenously (i.v.), was detected in the brain by IVIS imaging, proving that FSH permeates the blood–brain barrier (Extended Data Fig. 1k). Similarly, i.p. FSH injection (5 IU) led to increased brain FSH levels (Extended Data Fig. 1l). Using four complementary methods, we found that peripherally injected FSH-Ab also localized to brain tissue. First, biotinylated FSH-Ab (or goat IgG) injected i.p. displayed a non-cellular localization on MAP2 co-staining (Extended Data Fig. 1m). Second, we detected ⁸⁹Zr-labelled monoclonal FSH-Ab (injected i.v.) in brain tissue using positron emission tomography (PET) imaging and γ-counting (Extended Data Fig. 1n and Supplementary Video 1). Third, using IVIS, we show that AlexaFluor750-labelled monoclonal FSH-Ab delivered i.v. localized to the brain (Extended Data Fig. 1o). Finally, specific localization in the hippocampus was confirmed by immunohistochemistry analysis of perfused brain, establishing the presence of FSH-Ab in brain tissue, external to CD31-positive cells (Extended Data Fig. 1p).

FSHR signalling in neurons

End-point and quantitative PCR (qPCR) analyses show FSHR mRNA expression in human cortex, human neuroblastoma (SH-SY5Y) cells, mouse cortex and hippocampus, and rat neurons (Fig. 2a, b). Western immunoblotting further showed an ~85 kDa FSHR protein in both male and female mouse brains, but not in brains from Fshr^−/− mice (Fig. 2c). In a separate experiment, knockdown of hippocampal Fshr by stereotactic injection of AAV2 expressing short interfering RNA (siRNA) targeting Fshr (AAV2-siFshr) led to a significant decrease in band intensity, establishing the molecular identity of the detected FSHR protein (Fig. 3a). Furthermore, RNAscope analysis of whole-brain sections revealed Fshr transcripts in cortex and hippocampus of wild-type mice, but not Fshr^−/− mice (Fig. 2d, e). The fact that abundant Fshr transcripts were present in Sertoli cells, but not Leydig cells in the same testis sections conclusively established probe specificity (Fig. 2d). Fshr expression was also detected in other brain regions and subregions, with the highest Fshr density noted in the granular cell layer of the dentate gyrus, where transcripts localized primarily to Nissl-stained neurons (Fig. 2d, e). Using immunofluorescence analysis, FSHRs were found to co-localize with NeuN in hippocampal neurons, with little-to-no expression in glial cells (Fig. 2f). Importantly, FSHR immunostaining was abrogated after stereotactic hippocampal injection of AAV2-siFshr, and was absent in similarly stained sections from Fshr^−/− mouse brains (Fig. 2f). Consistent with the mouse brain, a ViewRNA analysis showed that FSHR transcripts co-localized with the non-coding RNA MALAT1 in the granular cell layer of the hippocampal dentate gyrus and in the parahippocampal cortex (Fig. 2g).

Fig. 2 |. Neuronal FSH receptors in mouse and human brain.

a–c, Fshr expression in human cortex, neuroblastoma cells (SH-SY5Y), mouse cortex and hippocampus, rat cortical neurons and/or mouse ovaries, determined by end-point PCR (a), qPCR (b) and/or western blotting (c). d, RNAscope signals (red arrows) in haematoxylin-stained cells from testes, and Nissl-stained neurons in the hippocampal dentate gyrus and entorhinal cortex of Fshr^+/+ and Fshr^−/− mice. Scale bars, 50 μm. e, Transcript counts and density in multiple brain regions from 34 sections. f, Co-staining of FSHR, NeuN, GFAP and IBA1 in the hippocampus after stereotactic siControl or siFshr injection in wild-type mice or in uninjected Fshr^−/− mice. Scale bars, 100 μm (main images), 10 μm (magnified images). g, ViewRNA signals (FSHR, dark blue; MALAT1, red) in cells from the hippocampal dentate gyrus (granular layer) (left) and parahippocampal cortex (layers V–VI) (right) from post-mortem human brain. Scale bars, 500 μm (left, main image), 50 μm (left, inset), 200 μm (right, main image), 20 μm (right, inset). For b, data are mean ± s.e.m. The number of mice per group is shown.

Fig. 3 |. Targeted Fshr knockdown in the hippocampus ameliorates AD neuropathology and impaired spatial memory.

a–e, The effect of stereotactic injection of siFshr versus siControl into ovariectomized 3xTg mice on total C/EBPβ, AEP, cleaved APP and Tau, and FSHR levels in the whole brain (a); AEP activity (b); synapse number (transmission electron micrographs, red arrows) (c); cell viability (NeuN-positivity and TUNEL) (d); and parameters of spatial memory determined using Morris water maze testing (e). Scale bars, 1 μm (c), 20 μm (d). Data are mean ± s.e.m. n = 3 (a), n = 5 (b), n = 3 (10 sections) (c), and n = 7 or n = 8 (e) mice per group. Statistical analysis was performed using unpaired two-tailed Student’s t-tests.

We studied a role for the C/EBPβ–AEP/δ-secretase pathway in mediating FSH action using human neuroblastoma SH-SY5Y cells and primary rat cortical neurons. Treatment with FSH (30 ng ml⁻¹) increased C/EBPβ and activated AEP in a time-dependent manner in both cell types, and the activated AEP cleaved both APP and Tau (Extended Data Fig. 2a). qPCR showed a time-dependent increase in CEBPB, LGMN, APP and MAPT expression, as well as enhanced AEP activity after FSH treatment (Extended Data Fig. 2b, c). Consistent with the known proinflammatory responses to C/EBPβ activation³⁸, FSH also induced the expression of the pro-inflammatory cytokines IL-1β and IL-6 (Extended Data Fig. 2d). Notably, the FSH-induced increases in C/EBPβ expression, AEP activity, and APP and Tau cleavage were decreased in the two cell types after infection with human siFSHR and rat AAV2-siFshr, respectively (Extended Data Fig. 2e). Similarly, in SH-SY5Y cells, siFSHR attenuated FSH-induced increases in CEBPB, LGMN, APP and MAPT mRNA, AEP activity, and secreted IL-1β and IL-6 (Extended Data Fig. 2f–h). We also infected both cell types with a lentivirus (LV) containing short hairpin RNA against Cebpb (shCebpb) and AAV-AEP^C189S, an AAV-expressing inactive AEP. Knocking down C/EBPβ or inactivating AEP substantially reduced the FSH-induced cleavage of APP and Tau (Extended Data Fig. 3a). Co-staining of cortical neurons confirmed that the FSH-induced increases in Aβ, phosphorylated Tau (pTau), cleaved APP and Tau, and AEP activity were reduced in cells infected with both viruses (Extended Data Fig. 3b, c). Collectively, these data confirm that FSH acts on FSHR to activate the C/EBPβ–AEP/δ-secretase pathway.

To examine the signalling mechanisms, we studied the effect of FSH on ERK1/2, SRPK2 and AKT in both cell types. Notably, MAPK-induced phosphorylation of residue Thr188 of C/EBPβ regulates its transcriptional activity³⁹. Similarly, AKT phosphorylates SRPK2, which in turn activates AEP by phosphorylating residue Ser226 (ref.⁴⁰). We found that total C/EBPβ and pC/EBPβ and active AEP progressively increased from 5 min after FSH exposure, whereas ERK1/2, AKT and SFRP2 phosphorylation peaked at ~30 min, with a delayed rise in NF-κB p65 (Extended Data Fig. 3d). Pertussis toxin, which decouples FSHRs from Gα_i protein^6,41, inhibited FSH-induced C/EBPβ and AEP activation and AKT phosphorylation (Extended Data Fig. 3e). Consistent with maximal inhibition of the Gα_i pathway, the adenylate cyclase inhibitor SQ22536 had no effect, whereas the AKT inhibitor AKTi-1/2 attenuated FSH-induced C/EBPβ expression, AEP cleavage and SRPK2 phosphorylation, without an effect on ERK1/2 (Extended Data Fig. 3e). Similarly, whereas the MEK1 inhibitor PD98059 blocked FSH-induced C/EBPβ, AEP and SRPK2 activation, AKT was unaffected (Extended Data Fig. 3e). Together, these data suggest that, in both human and rat neurons, FSH phosphorylates AKT, ERK1/2 and SRPK2 leading to the activation of C/EBPβ–AEP/δ-secretase and subsequent proteolytic cleavage of APP and Tau.

Neuronal Fshr knockdown decreases AD pathology

We examined whether the selective ablation of the Fshr in the hippocampus mimics the effect of systemic FSH-Ab. In brief, 3xTg mice underwent ovariectomy 7 days after stereotactic injection of AAV2-siFshr or scrambled control viral suspension (2 μl at 0.25 μl min⁻¹, 2 × 10⁹ vector genomes per μl) into the hippocampus. AAV2-siFshr-injected mice showed substantial decreases in FSHR protein and mRNA; total C/EBPβ and AEP; cleaved APP; cleaved and phosphorylated Tau; Aβ and its isoforms Aβ40 and Aβ42; and AEP activity (Fig. 3a, b and Extended Data Fig. 4a–c). The inhibition was recapitulated at the mRNA level for Cebpb, Lgmn, App and Mapt (Extended Data Fig. 4a). Similarly, compared with control-siRNA-injected mice, AAV2-siFshr-injected mice showed increases in dendritic spines and synapse number, and dampened neuronal apoptosis (Fig. 3c, d and Extended Data Fig. 4d). Testing using the Morris water maze revealed significant enhancements in spatial memory, notably, reduced latency to mount a platform and increased time spent in the platform quadrant in AAV2-siFshr-injected mice (Fig. 3e). Together, these data provide clear evidence for a role of hippocampus-resident, mainly neuronal, FSHRs in mediating, at least in part, the effects of FSH on AD pathogenesis. However, despite the preferential infection of neurons by AAV2 (serotype 2), we cannot exclude the possibility that non-neuronal glial cells are also infected by AAV siRNA^42,43. However, the fact that we almost exclusively find FSHRs on NeuN-positive neurons and not in glial cells (Fig. 2f) establishes that the effects of Fshr knockdown on the AD phenotype are mediated largely through neurons.

Systemic FSH triggers AD pathology

The i.p. injection of female 3xTg mice (aged 2.5 months) with a single dose of 2 IU, 5 IU or 10 IU of recombinant human FSH showed a dose-dependent increase in serum human FSH levels at 24 h (Extended Data Fig. 5a). However, as expected, endogenous mouse FSH levels remained stable at around 10 mIU ml⁻¹, and serum LH levels did not change (Extended Data Fig. 5a, b). The elevation in serum FSH at 5 IU was similar to that observed post-ovariectomy (compare with Extended Data Fig. 1a), as well as during the human menopausal transition¹⁷. Notably, we also detected an increase in FSH in brain lysates after i.p. injection, suggesting that FSH crosses the blood–brain barrier (Extended Data Fig. 1l). FSH (5 IU per mouse, i.p.) injected daily for 3 months caused a marked increase in brain C/EBPβ and AEP, and a robust activation of AEP to cleave APP and Tau (Fig. 4a, b and Extended Data Fig. 5h, i). Similarly, FSH induced pTau, Aβ40 and Aβ42 isoforms, and plaque formation was confirmed in hippocampal sections by thioflavin-S staining, and in the cortex and the cornu ammonis 1 (CA1) and dentate gyrus regions of the hippocampus by silver staining (Fig. 4c and Extended Data Fig. 5c, d, j). Importantly, FSH-induced enhancements in C/EBPβ and AEP occurred in Aβ- and pTau-bearing NeuN-positive neurons (Extended Data Fig. 5l). This, together with the selective expression of FSHR in human and mouse neurons (Fig. 2), further confirm a primary neuronal action of FSH. Furthermore, FSH triggered Cebpb, Lgmn, App and Mapt expression (Extended Data Fig. 5e), and caused marked apoptosis in hippocampal and cortical neurons, with reduced dendritic spine and synapse numbers (Extended Data Fig. 5f, g, k). Testing using the Morris water maze showed that FSH-injected mice had impaired spatial memory associated with reduced long-term potentiation (field excitatory post-synaptic potentials (fEPSPs)) in the CA1 hippocampal region (Fig. 4d, e). Taken together, systemic FSH over 3 months caused a marked acceleration of AD pathology in female 3xTg mice.

Fig. 4 |. Recombinant FSH induces AD pathologies and cognitive decline in 3xTg Mice.

a–d, The effect of injecting female 3xTg mice with recombinant human FSH on whole-brain C/EBPβ, AEP, cleaved APP and Tau proteins (a); AEP activity (b); Aβ40 and Aβ42 (c); and parameters of spatial memory determined using Morris water maze testing (d). e, The effect of FSH on long-term potentiation (LTP) shown as fEPSPs before (grey) and 60 min after (red) theta-burst stimulation. Data are mean ± s.e.m. n = 5 (b and c), n = 7 (d) and n = 4 (e) mice per group. Statistical analysis was performed using unpaired two-tailed Student’s t-tests.

To mimic the mid- and late-menopausal transition, during which FSH levels rise in the face of unaltered serum oestrogen⁴⁴, we replaced oestrogen in ovariectomized 3xTg mice. Serum oestrogen was clamped close to basal levels (~75 pg ml⁻¹) by s.c. implanting 17β-estradiol 90-day-release pellets (0.36 mg) before i.p. FSH injections (5 IU daily) given for 3 months (Extended Data Fig. 6a). Even in this oestrogen-replete state, FSH induced C/EBPβ and AEP expression, AEP activation, Aβ and pTau accumulation, APP and Tau cleavage, and Tau phosphorylation (compare Extended Data Fig. 6b–d with Fig. 4a, b and Extended Data Fig. 5c, d). These data provide unequivocal evidence that the effect of FSH is independent of oestrogen in female mice. Moreover, i.p. FSH injections (5 IU per mouse) for 3 months triggered identical neuropathology and cognitive impairment in male 3xTg mice (compare Extended Data Fig. 7 with Fig. 4 and Extended Data Fig. 5). Thus, while FSH triggers overt AD-like features in both female and male 3xTg mice, the phenotype is exacerbated after ovariectomy (high FSH) in female mice and, by implication, after the menopause in women.

As mouse Aβ barely oligomerizes to form plaques, aging wild-type mice do not display human-like AD features, despite increases in mouse Aβ production⁴⁵. To provide a non-transgenic control for 3xTg mice (Fig. 4), we studied the effect of high FSH in wild-type mice aged 3 months. Although i.p. treatment of FSH (5 IU per day for 3 months) increased C/EBPβ and AEP expression, AEP activity and the cleavage of mouse APP and Tau, it did not yield Aβ plaques or trigger memory impairment (Extended Data Fig. 8a–e). Thus, mice must express human APP to form Aβ plaques and induce cognitive decline in response to FSH. However, both 3xTg and APP/PS1 mice overexpress human Aβ transgenically at levels that are potentially neurotoxic. By contrast, in APP-KI mice, three amino acid substitutions (G601R, F606Y and R609H) are knocked into exon 14 of the Aβ-encoding App gene resulting in the expression of oligomerizable human Aβ at basal levels. FSH (5 IU per day for 3 months, i.p.) injected into female APP-KI mice (aged around 3 months) increased C/EBPβ and AEP expression and APP cleavage, and induced Aβ plaques in the hippocampus and/or cortex, Aβ40 and Aβ42 accumulation in the whole brain, and apoptosis in the hippocampus (Extended Data Fig. 8f–h, i, k). Although Tau is not mutated in APP-KI mice, there was an increase in the cleavage of endogenous Tau (Extended Data Fig. 8f). Morris water maze testing of FSH-treated APP-KI mice revealed impaired spatial memory (Extended Data Fig. 8j). Collectively, the responsiveness of APP-KI mice to FSH further reaffirms a role for FSH as a driver of AD in a model that closely resembles the human disease. It also rules out confounding effects of toxic levels of transgenic Aβ overexpression in 3xTg and APP/PS1 mice.

FSH induces AD pathology through C/EPBβ

We generated compound Cebpb^+/− 3xTg mutant mice on a C57BL/6 background by crossing 3xTg with Cebpb^+/− mice. We i.p. injected 2.5-to 3-month-old compound mutant mice with FSH (5 IU per mouse) daily for 3 months. Cebpb^−/− mice were not used due to a potentially confounding metabolic phenotype⁴⁶. Haploinsufficiency of Cebpb led to attenuated baseline events—notably, lower AEP activation and APP and Tau cleavage (Fig. 5a, c). In 3xTg mice, FSH induced AEP activity, APP and Tau cleavage, and Aβ and pTau accumulation, whereas the responses to FSH were attenuated in Cebpb^+/− 3xTg mice (Fig. 5a and Extended Data Fig. 9b–d). FSH-induced increases in Cebpb, Lgmn, App and Mapt expression were similarly attenuated in Cebpb^+/− 3xTg mice (Extended Data Fig. 9a). C/EBPβ haploinsufficiency also reversed the decreases in dendritic spine numbers, as well as the cognitive deficit induced by FSH (Fig. 5e and Extended Data Fig. 9e, f). As FSH is elevated after ovariectomy (Extended Data Fig. 1a), we predicted that the effects of ovariectomy in inducing AD pathology and cognitive decline would also be attenuated in compound Cebpb^+/− 3xTg mice. Indeed, FSH-induced Cebpb, Lgmn, App and Mapt expression, AEP activation, APP and Tau cleavage, Aβ and pTau accumulation, dendritic spine deficits and cognition defects were all lower in Cebpb^+/− 3xTg mice compared with 3xTg mice (Fig. 5b, d, f and Extended Data Fig. 10a–f). Taken together, these data provide strong genetic evidence for a primary role for C/EBPβ in mediating the AD pathology induced by FSH.

Fig. 5 |. FSH-induced AD pathology is dampened in Cebpb^+/− 3xTg mice.

a–f, The effect of injecting recombinant human FSH (a, c and e) or ovariectomy (b, d, f) in 3xTg mice or compound mutant Cebpb^+/− 3xTg mice on C/EBPβ, AEP, cleaved APP, cleaved Tau, pTau and FSHR (a, b); AEP activity (c, d); and parameters of spatial memory determined using Morris water maze testing (e, f). Data are mean ± s.e.m. n = 6 (c), n = 5 (d), n = 9 (e), and n = 7 or n = 8 (f) mice per group. Statistical analysis was performed using one-way ANOVA.

Discussion

The idea that the brain is a target for FSH is consistent with research that has established broad ubiquity for pituitary hormone action—a stark departure from the long-held view that pituitary glycoproteins act solely on endocrine targets⁴⁷. We first discovered an extragonadal action of FSH by demonstrating high bone mass in FSHβ haploinsufficient mice that were eugonadal⁴¹. Studies have since documented the existence of functional FSHRs on bone cells, adipocytes and hepatocytes, and have established that selective FSH blockade—such as by FSH-Ab used here—stimulates new bone synthesis, reduces body fat, increases thermogenesis and lowers serum cholesterol^3–7. In gonadal cells, FSH is also known to activate C/EBPβ^48,49, which has been associated with AD pathology, particularly as it transactivates pro-inflammatory genes that mediate the chronic inflammation of AD³⁸. We previously showed that the spatiotemporal dysregulation of the C/EBPβ–AEP/δ-secretase pathway mediates AD pathology⁵⁰.

Thus, our observations that link FSH and AD through the C/EBPβ–AEP/δ-secretase pathway, taken together with the strong clinical association of AD with rising serum FSH levels, provide the basis for an array of actionable targets for AD in post-menopausal women. Furthermore, as FSH levels also rise in aging men³³, our finding that systemic FSH induces and FSH-Ab rescues AD pathology in both sexes provides an opportune window for testing an anti-FSH agent in humans. Given that blocking FSH not only reduces bone loss, body fat and serum cholesterol^3,5,6,21, but also dampens AD pathology in mice as shown here, an agent, such as a highly targeted anti-FSH antibody³⁷, could be tested in the future for the co-therapy of osteoporosis, obesity, dyslipidaemia and Alzheimer’s disease.

Methods

Transgenic mice

3xTg mice were obtained from Jackson Laboratory (034830). The mice have a transgene with mutated human APP ^K670N/M671L and MAPT ^P301L, and a knockin mutation Psen1^M146V and display AD neuropathology and a decline in long-term memory at around 3 to 4 months²². Mutant Cebpb mice⁵¹ were maintained at the Emory University School of Medicine as heterozygotes on C57BL/6 and 129Sv backgrounds. The two strains were crossed to generate viable F₁ hybrid wild-type and Cebpb^+/− littermates; the latter were then crossed with 3xTg mice to generate compound Cebpb^+/+ 3xTg and Cebpb^+/− 3xTg mutants. APP/PS1 mice, which were obtained from Jackson Laboratory (034829) and bred at the Icahn School of Medicine at Mount Sinai (ISMMS), have a human transgene comprising APP ^K670N/M671L and PSEN1^ΔE9 (ref.³⁰). APP/PS1 mice develop amyloid plaques at 6 months and overt cognitive impairment at a later age of 15 months; they do not display neurofibrillary tangles^30–32. These mice, when bred on a pure C57BL/6J background, show evidence of sudden death (27%) due to seizure activity (65%)⁵². Our losses were mostly in the 5 month, drug-naive period (13 deaths out of 28), followed by 5 deaths during the 4 month treatment with IgG (3 deaths) or FSH-Ab (2 deaths). APP/PS1 mice were genotyped and regenotyped for human PSEN1, followed by qPCR for human APP. In contrast to 3xTg and APP/PS1 transgenic mice, in APP-KI mice, three amino acid substitutions (G601R, F606Y and R609H) are knocked into Aβ-coding exon 14 of the App gene—this results in the non-transgenic expression at basal levels of readily oligomerizable human Aβ (Jackson Laboratory, 030898). For studies using 3xTg and APP-KI mice, sample sizes were determined on the basis of previous studies by K.Y. and other groups. Mice were randomly selected to be assigned into separate groups—namely, ovariectomy or sham-operation, PBS or FSH, IgG or FSH-Ab, and siControl or siFshr. Key behavioural studies at Emory University were conducted in the Rodent Behavioral Core by technicians who were unaware of the mouse groups. Animal care and handling were performed according to NIH animal care guidelines at both the Emory University School of Medicine and the ISMMS. The protocols were reviewed and approved by the respective Institutional Animal Care and Use Committees.

Ovariectomy

Each mouse underwent shaving and prepping of the lower back using 70% ethanol and sterile PBS, followed by washing with povidone iodine solution before the surgery. A ~0.5 cm skin incision and a further incision through the muscle layer allowed access to the peritoneal cavity. The ovaries were visualized through the muscle layer and, one at a time, extracted through the incision, tied off and removed. The muscle layer was then sutured followed by closure of the external incision with wound clips. Each mouse was placed in a clean cage and allowed to recover from the anaesthesia, and returned to home cage after observation of normal behaviour and ambulation.

Stereotactic injection

AAV2-siFshr virus (iAAV04355302) and control virus (iAAV01502) from Applied Biological Materials were injected stereotactically into female 3xTg mice (aged 3 months) under isoflurane anaesthesia. We used the following coordinates for bilateral intracerebral injections: −2.1 mm anteroposterior and −1.8 mm mediolateral from the bregma, and −1.5 mm dorsoventral from the dural surface. Viral suspension (2 μl) containing 2 × 10⁹ vector genomes per μl was placed into each site at a rate of 0.25 μl min⁻¹ using a 10 μl glass syringe with a fixed needle. The needle remained in place for 5 min and was removed slowly over 2 min. Mice were placed on a heating pad until they turned from a supine to a prone position, indicative of recovery from anaesthesia. The mice underwent ovariectomy 7 days after stereotactic injection.

Antibodies and reagents

Please refer to the Reporting Summary and Supplementary Table 1 for details, including prior validation of each antibody. Antibodies against C/EBPβ (H-7, sc-7962, H-7, 1:1,000 dilution for western blotting and 1:200 for immunofluorescence), C/EBPβ (rabbit polyclonal, sc-150, 1:1,000 dilution) and FSHβ (sc-374452, C12, 1:1,000 dilution for western blotting) were from Santa Cruz; antibodies against AEP (6E3) were a gift from C. Watts (1:1,000 dilution for western blotting); antibodies against FSHR (PA5–50963, 1:1,000 dilution for western blotting and 1:200 for immunofluorescence), pTau(Ser202–Thr205) (MN1020, AT8, 1:1,000 dilution for western blotting and 1:200 for immunofluorescence and immunohistochemistry) and IBA1 (PA5–18039, 1:500 dilution for immunofluorescence) were from Thermo Fisher Scientific; antibodies against Legumain (LGMN, D6S4H) (93627, 1:2,000 dilution for western blotting and 1:500 for immunofluorescence), AKT (4691s, 1:1,000 dilution for western blotting), pAKT^S473 (rabbit polyclonal, 9271s, 1:1,000 dilution for western blotting), ERK1/2 (9102s, 1:1,000 dilution for western blotting), pERK1/2 (9106s, 1:2,000 dilution for western blotting) were purchased from Cell Signaling Technology; antibodies against NeuN (MAB377, 1:300 dilution for immunofluorescence), NeuN (ABN90, 1:600 dilution for immunofluorescence), Tau(210–241)(Tau-5, MAB361, 1:2,000 dilution for western blotting), β-actin (A5316, 1:3,000 dilution for western blotting) and GFAP (MAB360, GA5, 1:400 dilution for immunofluorescence) were from Sigma-Aldrich; antibodies against Aβ (800701, 4G8, 1:400 dilution for immunohistochemistry and immunofluorescence) were obtained from BioLegend; and antibodies against SRPK2 (611118, 23/SRPK2, 1:2,000 for western blotting) were from BD Biosciences. Antibodies against pAEP^S226, Tau(1–368), APP(1–585), APP(1–373) and pSRPK2(T492) were developed in the K.Y.’s lab (1:1,000 for western blotting; 1:600 and 1:200 for immunofluorescence with Tau(1–368) and APP^C586 antibodies, respectively). The FSH-blocking polyclonal antibody (FSH-Ab) and humanized monoclonal antibody (Hu6) were generated by Genscript, and developed and characterized in the Zaidi laboratory^5,21. siFshr (sc-35415) was obtained from Santa Cruz, and the TUNEL In Situ Cell Death Detection Kit (11684817910) was obtained from Roche. Human Aβ40 (KHB3481), Aβ42 (KHB3544) and inflammatory cytokine ELISA kits (BMS224–2 and KHC0061) were purchased from Invitrogen. The human FSH ELISA kit was obtained from Abcam (ab108641). The mouse FSH (CSB-E06871m-96) and LH ELISA (CSB-E12770m-96) kits were from Cusabio. The 17-β estradiol ELISA kit was obtained from Abcam (ab108667). The AEP substrate Z-Ala-Ala-Asn-AMC (4033201) was obtained from Bachem, and the EZ-Link Sulfo-NHS-LC-Biotinylation Kit was obtained from Thermo Fisher Scientific (21435). Recombinant human FSH used in vitro and in vivo experiments was obtained from Sigma-Aldrich (F4021) and EastCoast Bio (LA252), respectively. The 90-day-release pellets containing 0.36 mg 17β-estradiol (NE121) were purchased from Innovative Research of America. All chemicals not mentioned above were purchased from Sigma-Aldrich.

Human tissue

Post-mortem brain tissue for western blotting and PCR was obtained from the Emory Alzheimer’s Disease Research Center and was approved both by the Biospecimen Committee (approved on 10 February 2021, renewal 1 April 2022), and the Institutional Review Board (IRB) (approved on 12 February 2019; IRB00045782). For ViewRNA studies, we used a flash-frozen, never-thawed brain sample from a 85-year-old white male (de-identified) from Mount Sinai’s NIH Brain Tissue Repository (NBTR) (IRB Exempt Status: HS#:13–00709 PS).

Human brain transcriptomics analysis using ViewRNA

We used two AD-vulnerable regions—the hippocampus and parahippocampal cortex. Serial, cryostat sections (thickness, 12 μm) in the coronal plane were treated with proteinase K followed by in situ hybridization and amplification with specific probe sets in combination: FSHR-T6 and MALAT1-T1 using the QuantiGene ViewRNA Assay (Invitrogen). The sections were counterstained with Gill’s hematoxylin. This enabled us to map the cellular expression of FSHR in MALAT1-positive neurons. Images were acquired using our Zeiss Axio-Imager Z1 and Pannoramic 250 (3DHistech) high-resolution scanner.

Quantitative mouse brain transcriptomics by RNAscope

RNAscope allowed the detection of single transcripts in wild type mouse brain. In brief, 20 pairs of double-Z probes specific to the Fshr transcript (ACD Biotech, 400468) were used to hybridize sections. Preamplifiers were allowed to hybridize to the 28 bp binding site formed by each double-Z probe. The amplifiers then further bound to the multiple binding sites on each preamplifier and, finally, labelled probes containing a chromogenic molecule were allowed to bind to multiple sites of each amplifier. Brain tissue and testes were collected from wild-type and Fshr^−/− male mice. In brief, mice were anaesthetized with isoflurane (2–3% in oxygen; Baxter Healthcare) and perfused transcardially with 0.9% heparinized saline followed by 4% paraformaldehyde (PFA). Brains and testes were extracted, sectioned into 0.5-cm-thick slices and post-fixed in 4% PFA for 12 h before being transferred to 70% ethanol for 24 h, and embedded in paraffin after gradual ethanol dehydration. Coronal sections were cut at 5 μm. For brain, every tenth section was mounted onto ~20 slides, with two sections on each slide. This method enabled us to cover all regions of the brain and to eliminate the likelihood of counting the same transcript twice. Sections were stored at −80 °C. RNAscope was performed using the ACD RNAscope HD Brown LS Reagent Kit and RNAscope LS 2.5 Probe for Mm-Fshr. The slides were thawed at room temperature for 10 min before baking at 60 °C for 60 min. The slides were deparaffinized, air-dried for 5 min, blocked with hydrogen peroxide for 10 min at room temperature and pretreated using Target Retrieval Solution at 100 °C for 20 min and Protease Plus at 40 °C for 30 min. Probe hybridization and signal amplification were performed according to the manufacturer’s instructions for chromogenic assays. The slides were imaged on an Aperio CS2 Scanner (Leica) and CaseViewer v.2.4 (3DHistech). The Allen Mouse Brain Atlas (https://mouse.brain-map.org/static/atlas) was used to identify and manually map Fshr-positive cells with precision at each neuroanatomical level. Thereafter, transcripts were counted under high magnification (×40) in each relevant region and subregion (shown in Fig. 2e).

Cells

The human neuroblastoma cell line SH-SY5Y was obtained from ATCC and was not authenticated or tested for mycoplasma contamination. SH-SY5Y was cultured in DMEM/F12 with 10% fetal bovine serum (v/v), penicillin (100 U ml⁻¹, w/v) and streptomycin (100 μg ml⁻¹, w/v). SH-SY5Y transfection was performed using Lipofectamine 3000 (Invitrogen). Primary culture of rat cortical neurons was described previously⁵³. AAV2-siFshr virus (iAAV05724202) was purchased from Applied Biological Materials. LV-shCebpb-GFP, LV-GFP and AAV-AEP^C189S were packaged by the Viral Vector Core (VVC) of Emory University. Neurons cultured 7 days in vitro (DIV 7) were transfected with LV-shCebpb-GFP, LV-GFP, AAV-AEP ^C189S or AAV2-siFshr. Virus solution (2 μl) was added to 1 ml culture medium and applied to primary neuron cultures. Seven days later (DIV 14), the neurons were treated with FSH (30 ng ml⁻¹) for 48 h. The cells were incubated at 37 °C in a humidified atmosphere of 5% CO₂.

AEP activity assay

Tissue homogenates or cell lysates (10 μg) were incubated in 200 μl assay buffer (20 mM citric acid, 60 mM Na₂HPO₄, 1 mM EDTA, 0.1% CHAPS and 1 mM dithiothreitol, pH 6.0) containing 20 μM δ-secretase substrate, Z-Ala-Ala-Asn-AMC (Bachem). AMC released by substrate cleavage was quantified by measuring at 460 nm using a fluorescence plate reader at 37 °C in kinetic mode.

qPCR analysis

mRNA levels were analysed using qPCR. In brief, RNA was isolated by TRIzol (Life Technologies). Reverse transcription was performed using SuperScript III reverse transcriptase (Life Technologies). Gene-specific primers and Taqman probes were designed and bought from Applied Biosystems (human GAPDH (Hs02758991), mouse Gapdh (Mm99999915), rat Gapdh (Rn01775763); human CEBPB (Hs00942496), mouse Cebpb (Mm00843434); human LGMN (Hs00271599), mouse Lgmn (Mm01325350); human APP (Hs00169098), mouse App (Mm01344172); human MAPT (Hs00902194), mouse Mapt (Mm00521988); and human FSHR (Hs01019695), mouse Fshr (Mm00442819) and rat Fshr (Rn01648507). All qPCR reactions were performed using the ABI 7500-Fast Real-Time PCR System (SDS v.2.3) and the Taqman Universal Master Mix Kit (Life Technologies). The relative quantification of gene expression was calculated using the ΔΔC_t method.

Western immunoblotting

Cells and brain tissue were washed with ice-cold PBS and lysed in 50 mM Tris-HCl, pH 7.4, 40 mM NaCl, 1 mM EDTA, 0.5% Triton X-100, 1.5 mM Na₃VO₄, 50 mM NaF, 10 mM sodium pyrophosphate and 10 mM sodium β-glycerophosphate, supplemented with protease inhibitor cocktail at 4 °C for 0.5 h, and centrifuged for 25 min at 15,000 rpm. The supernatant was boiled in SDS loading buffer. After SDS–PAGE, the samples were transferred to a nitrocellulose membrane. The membrane was blocked with TBS containing 5% non-fat milk and 0.1% Tween 20 (TBST) at room temperature for 2 h, followed by incubation with primary antibodies at 4 °C overnight, and with secondary antibodies at room temperature for 2 h (the dilutions are described in the ‘Antibodies and reagents’ section). After washing with TBST, the membrane was developed using the enhanced chemiluminescent detection system.

Immunostaining

We use free-floating 25 μm brain sections for immunostaining. For immunohistochemistry, brain sections were treated with 0.3% H₂O₂ (v/v) for 10 min. The sections were washed three times in PBS and blocked in 1% BSA (w/v) and 0.3% Triton X-100 (v/v) for 30 min, followed by overnight incubation with anti-Aβ (1:400) and anti-AT8 (1:200) antibodies at 4 °C. The signal was developed using the Mouse and Rabbit Specific HRP/DAB (ABC) Detection IHC kit (Abcam). For immunofluorescence staining, sections were incubated overnight at 4 °C with primary anti-Aβ (1:400), anti-APP(C586) (1:200), anti-AT8 (1:200) or anti-Tau(1–368) (1:600) antibodies. After washing with Tris-buffered saline, the sections were incubated with a mixture of labelled secondary antibodies for detection. Details of antibody combinations are provided in the Reporting Summary. DAPI (1 μg ml⁻¹) (Sigma-Aldrich) was used for staining nuclei. For thioflavin-S and Aβ double staining, slides were first stained with anti-Aβ antibodies, rinsed in PBS and then stained with freshly prepared 0.0125% thioflavin-S in 50% ethanol for 10 min. The sections were washed with 50% ethanol and placed in distilled water. Images were acquired with an Olympus Confocal FV1000 Imaging System.

Gallyas silver staining

Brain sections (25 μm) were incubated in 5% periodic acid for 5 min, washed in water and then placed in alkaline silver iodide solution (containing 1% (v/v) silver nitrate) for 1 min. The sections were then washed in 0.5% acetic acid (v/v) for 10 min, placed in developer solution for 15 min, and washed with 0.5% acetic acid (v/v) and then with water. The sections were treated with 0.1% gold chloride for 5 min, washed in water and incubated in 1% sodium thiosulfate (v/v) for 5 min before a final washing.

Golgi staining

Mouse brains were fixed in 10% formalin (v/v) for 24 h and then immersed in 3% potassium bichromate (w/v) for 3 days in the dark. The solution was changed each day and the brains were transferred into 2% silver nitrate (w/v) solution and incubated for 7 days in the dark. The solution was changed each day. Vibratome sections were cut at 30 μm, air dried for 10 min, dehydrated through 95% and 100% ethanol, cleared in xylene and coverslips were applied. Spine numbers were counted as described previously⁵⁴.

Transmission electron microscopy of synapses

After deep anaesthesia, mice were perfused transcardially with 4% paraformaldehyde (v/v) in PBS. Hippocampal slices were post-fixed in cold 1% OsO₄ (w/v) for 1 h. Samples were prepared and examined using standard procedures. Ultrathin sections (90 nm) were stained with uranyl acetate and lead acetate and viewed at 100 kV under a JEOL 200CX Electron Microscope. Synapses were identified by the presence of synaptic vesicles and postsynaptic densities. Synapse number was quantified as described previously^25,26.

Morris water maze

Mice were trained in a round, water-filled tub (diameter, 52 inches) in an environment rich with extra maze cues, as described previously. Each subject was given 4 trials per day for 5 consecutive days with a 15 min intertrial interval. The maximum trial length was 60 s and, if mice did not reach the platform in the allotted time, they were manually guided to it. After the 5 days of task acquisition, a probe trial was presented during which time the platform was removed and the percentage of time spent in the quadrant that previously contained the escape platform during task acquisition was measured over 60 s. All trials were analysed for latency by means of MazeScan (TopScan v.3.0, Clever Sys).

Novel-object recognition test

Mice were presented with two identical objects during the first session, and then one of the two objects was replaced by a novel object during a second session⁵⁵. On day 1, a habituation phase in an empty arena (for 5 min), was followed 24 h later by the training phase, which allows for a 5 min exploration in the habituated arena in which two identical objects are placed in opposite quadrants. The testing phase followed a gap of 20 min. For testing, one object was replaced with a novel object followed by 5 min of exploration. Data were collected using the ANY-maze Video Tracking System v.3.3 (Stoelting).

Electrophysiology

Mice were anaesthetized with isoflurane, decapitated and their brains were dropped in ice-cold artificial cerebrospinal fluid (a-CSF) containing 124 mM NaCl, 3 mM KCl, 1.25 mM NaH₂PO₄, 6.0 mM MgCl₂, 26 mM NaHCO₃, 2.0 mM CaCl₂ and 10 mM glucose. Hippocampi were dissected and cut into 400-μm-thick transverse slices with a vibratome. After incubation at 23–24 °C in a-CSF for 60–90 min, the slices were placed in a recording chamber (RC-22C, Warner Instruments) on the stage of an upright microscope (Olympus CX-31) and perfused at a rate of 3 ml per min with a-CSF (containing 1 mM MgCl₂) at 23–24 °C. A 0.1 MΩ tungsten monopolar electrode was used to stimulate the Schaffer collaterals. fEPSPs were recorded under current-clamp mode in CA1 stratum radiatum by a glass microelectrode filled with a-CSF with a resistance of 3–4 MΩ. The stimulation output (Master-8; AMPI) was controlled by the trigger function of an EPC9 amplifier (HEKA Elektronik). Data were filtered at 3 kHz and digitized at sampling rates of 20 kHz using PatchMaster v.2×90.1 (HEKA Elektronik). Stimulus intensity (0.1 ms duration, 10–30 mA) was set to evoke 40% of the maximum fEPSP and the test pulse was applied at a rate of 0.033 Hz. The LTP of fEPSPs was induced by 3 theta-burst-stimulation (4 pulses at 100 Hz, repeated 3 times with a 200 ms interval). The magnitudes of LTP are expressed as the mean percentage of the baseline fEPSP initial slope.

Distribution studies

The fluorescent signal following the injection of AlexaFluor750-tagged FSH or Hu6 was quantified in dissected tissues using the IVIS Spectrum In Vivo Imaging System (Perkin Elmer). For ⁸⁹Zr-distribution studies, a solution of ⁸⁹Zr oxalate in 1 M oxalic acid was neutralized using a 1 M Na₂CO₃ solution until a pH of 6.8–7.4 was reached. The ⁸⁹Zr solution was added to the DFO-containing Hu6 and incubated at 37 °C using a thermomixer (600 rpm) for 60 min. The resulting solution was purified using a PD-10 column with PBS as eluent. C57BL/6 mice were injected with ⁸⁹Zr-DFO-Hu6 (100 μCi) through the tail vein. The mice were imaged by PET, and were euthanized after extensive perfusion with PBS (24–48 h after injection). Brains were collected and weighed before counting radioactivity on a 2480 Wizard² automatic γ-counter (Perkin Elmer). Radioactivity data were corrected for decay and normalized to tissue weight to express radioactivity concentration as the percentage injected dose per gram.

Statistical analysis

Statistical analyses were performed using GraphPad Prism v.8. The tests were either unpaired two-tailed Student’s t-test (two-group comparison) or one-way ANOVA followed by Fisher’s least significant difference post hoc test (more than two groups). Differences with P< 0.05 were considered to be significant. P values are annotated in the figures and extended data figures, and are provided in the source data.

Ensuring rigour and reproducibility

There is a nascent movement to ensure that preclinical data is true and accurate^56–61. M.Z. and C.J.R. coined the phrase ‘contemporaneous reproducibility,’ which refers to the synchronous reproduction of data in more than one laboratory. As Zaidi’s discovery of the effects of FSH on body fat were new and unexpected, he reached out to C.J.R. for help in the form of a reproducibility study. Key datasets were reproduced by C.J.R. in a process that lasted for more than three years, as other validation studies were added by both laboratories. The term replicability refers to the ability of one or more independent groups to replicate a finding using a different technology or method—replicability is a measure of truth or significance of a given finding²⁹. Here we replicated the key finding, namely the effect of the FSH-Ab, in the K.Y. and M.Z. laboratories using two mouse models of AD and separate protocols to examine FSH-Ab action in both sexes. We have also used three transcriptomic technologies, qPCR (M.Z. laboratory), RNAscope (Alamak Biosciences) and ViewRNA (V.H. laboratory) to reproduce the findings on FSHR expression in the mouse and human brain. To further enhance rigour and transparency, we have exchanged Excel spread sheets to cross-check primary datasets and reviewed each figure independently to determine accuracy. Everything from simple immunoblots to microscopy images has been vetted by the research groups. Such validation practices, we believe, require unfettered transparency and remain fundamental to ensuring rigour.

Ensuring integrity of microscopy images

Images were visualized by the three independent observers in several steps (V.R., T.Y. and V.M.). First, raw camera images were loaded onto a Cloud by S.S.K. Second, individual images were downloaded and scrutinized against the corresponding panel in the manuscript to ensure that the published images matched the source dataset. Third, we compared merged immunofluorescence images against their original single images to be certain that the merge arose from the specific single images. Fourth, we overexposed or dimmed the background to verify the integrity of both single and merged images. Fifth, to make sure that there were no accidental duplications, we compared each image against all other images in the manuscript. Finally, during the latter test, we also rotated and flipped each image to validate integrity.

Reporting summary

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

Extended Data

Extended Data Fig. 1 |. Effects of Anti-FSHβ Antibody in Reversing Ovariectomy-Induced Neuropathology in 3xTg Mice.

a, Ovariectomized 3xTg mice displayed hypoplastic thread-like uteri and elevated serum FSH and LH levels. FSH-Ab (200 μg/mouse, every 2 days, i.p., 8 weeks) did not alter total serum FSH, LH or 17β-estradiol levels. Statistics: mean ± s.e.m., N = 8 mice per group, one-way ANOVA. b, Immunofluorescent micrographs showing enhanced labelling in the following pairs: amyloid β (Aβ, red) and thioflavin-S (Thio-S, green); Aβ (red) and cleaved APP^C586 (green); and pTau (red) and Tau^1–368 (green) in the hippocampus after OVX, and its amelioration with FSH-Ab (scale bar, 20 μm). c, Upregulation of Cebpb, Lgmn, App and Mapt in OVX mouse brains, with reversal to near-baseline with FSH-Ab. Statistics: mean ± s.e.m., N = 3 mice per group, one-way AVOVA. d, Immunofluorescence micrographs showing that OVX induces apoptosis (TUNEL, green) in hippocampal NeuN-positive neurons (red); this apoptosis is abolished by FSH-Ab (scale bar, 20 μm). e, Golgi staining on brain sections from CA1 region post-OVX shows a substantial reduction in spine numbers, which is corrected with the FSH-Ab (scale bar, 5 μm). Statistics: mean ± s.e.m., N = 3 mice per group (10 sections), one-way AVOVA. f, Transmission electron micrographic images and quantitative analysis of synapses in hippocampal sections post-OVX treated with IgG or FSH-Ab (scale bar, 1 μm). Statistics: mean ± s.e.m., N = 3 mice per group (8 sections), one-way AVOVA. g, Morris Water Maze testing shows no differences in swim speed. Statistics: mean ± s.e.m., N = 9 mice per group, one-way ANOVA. h, Cognitive testing using the Novel Object Recognition test revealed the absence of significant difference between APP/PS1 and non-transgenic mice in Discrimination Index [(Novel Object Head Entry – Familiar Object Head Entry)/Total Head Entry]; the result is expected at 9 months of age in APP/PS1 mice. Thus, no effect of FSH-Ab was noted at this age, despite the reduction in Aβ40 and Aβ42 accumulation shown in Fig. 1f. Statistics: mean ± s.e.m., mice per group 9, 10, 4 and 4 from left to right; Whisker plot, upper and lower ends of the whiskers show maxima and minima, line in box shows median, and upper and lower box boundaries show 75^th and 25^th percentile, respectively; unpaired two-tailed Student’s t-test; i, ELISA showing no cross-reactivity of FSH-Ab with LH. j, Western immunoblot showing no change in expression of the LHCGR in whole brain lysates upon OVX or FSH-Ab treatment (N = 2 mice per group). k, IVIS imaging of isolated tissues from mice injected with AlexaFluor750–FSH, i.v., showing localization of FSH in the brain (N = 3 mice per group). l, Western immunoblots of whole brain lysates showing that i.p. injection of human FSH (5 IU) causes an elevation of brain FSH (N = 3 mice per group). m, Immunofluorescence micrographs showing the detection of peripherally injected (i.p.) biotinylated FSH-Ab (red) and biotinylated goat IgG (red) in brain sections (scale bar, 20 μm). Note the absence of cellular or nuclear co-localization with MAP2 or DAPI, respectively. n, Representative PET image shows that ⁸⁹Zr-labelled humanized monoclonal FSH-Ab (⁸⁹Zr-Hu6), injected i.v., is localized to live brain (arrows). γ-counting in perfused tissue shows presence of ⁸⁹Zr-Hu6 in dissected brain tissue at 24 and 48 h post-injection (N = 4 mice). o, IVIS imaging and quantitation with AlexaFluor750-labelled Hu6, given i.v. shows localization in perfused whole brain tissue; N = 3 mice per group. Control (Ctrl): phosphate-buffered saline (PBS). p, Confirmatory immunofluorescence on the same mice (o) using anti-human IgG showing Hu6 localization (red) in proximity to CD31⁺ endothelial cells (green) (scale bar, 100 μm). For gel source data, see Supplementary Fig. 1.

Extended Data Fig. 2 |. FSHR Activation Triggers Amyloidogenic Protein Accumulation.

a, Western immunoblots showing the effect of activating neuronal FSHRs by FSH (30 ng/mL) in human SH-SY5Y and primary rat neuronal cells on the expression of C/EBPβ, AEP, as well as the cleavage of amyloid precursor protein (APP) and Tau using antibodies noted in ‘Methods’. FSH (30 ng/mL) likewise stimulated the expression of CEBPB, LGMN, APP and MAPT (qPCR) (b); AEP activity (c); and certain inflammatory cytokines (ELISA), namely IL-6 and IL-1β (d). Statistics: mean ± s.e.m.; Mice per group, (b) 3, (c) 6, and (d) 6; one-way ANOVA. e, Western immunoblotting showing C/EBPβ, AEP, APP, cleaved APP^1–585, total Tau, and cleaved Tau^1–368 in response to FSH or PBS following transfection with human FSHR siRNA (si-FSHR) for SH-SY5Y cells or rat Fshr siRNA (si-Fshr) for primary rat neurons, or appropriate scrambled siRNAs. f, mRNA levels of CEBPB, LGMN, APP and MAPT in SH-SY5Y cells incubated with FSH after control or si-FSHR transfection. g, AEP activity after incubation with FSH in control or si-FSHR-transfected SH-SY5Y cells. h, IL-6 and IL-1β levels (ELISAs) in SH-SY5Y cells incubated with FSH following control or si-FSHR infection. Statistics: mean ± s.e.m.; (f) 3 biological replicates; (g, h) 6 mice per group; one-way ANOVA.

Extended Data Fig. 3 |. FSH Induces APP and Tau Cleavage Through C/EBPβ and AEP/δ-Secretase Activation in Human SH-SY5Y Cells and Rat Cortical Neurons.

a, Western immunoblots showing the effect of FSH (30 ng/mL) on Tau, APP, AEP and FSHR of knocking down C/EBPβ expression by lentiviral infection with shRNA-Cebpb (sh-Cebpb) or reducing δ-secretase activity by adeno-associated virus infection of AEP^C189S in both human SH-SY5Y cells and rat cortical neurons. The stimulatory action of FSH was reversed at 48 h. b, c, Effect of FSH (30 ng/mL) on APP, APP^C586, pTau and Tau^1–368 accumulation (immunofluorescence, scale bar, 40 μm, b) and AEP activity (c) in rat cortical neurons infected with sh-Cebpb or AAV-AEP^C189S. Statistics: (c) mean ± s.e.m.; N = 6 mice per group; one-way ANOVA. d, Western immunoblots showing the time course of FSH effects on C/EBPβ, phosphorylated C/EBPβ (pC/EBPβ), AEP, pAEP^S226, total AKT, pAKT^S473, total ERK1/2, pERK1/2, total SRPK2, pSRPK2^T492 and pNFκB-p65. e, Western immunoblots showing the effect of a 30-minute incubation with FSH (30 ng/mL) on levels of C/EBPβ, AEP, pAEP^S226, total AKT and pAKT^S473, total ERK1/2 and pERK1/2, total SRPK2 and pSRPK2^T492 in the presence or absence of the cAMP inhibitor SQ22536 (100 μM), Gα_i inhibitor pertussis toxin (PTX, 50 ng/ml), AKTi-1/2 inhibitor (10 μM) and ERK1/2 inhibitor PD98059 (10 μM).

Extended Data Fig. 4 |. Targeted Knockdown of Fshr in the Hippocampus Diminishes AD Pathologies.

a, Quantitative PCR shows significantly reduced expression of Fshr, Cebpb, Lgmn, App and Mapt. b, Immunohistochemistry of the hippocampus shows reduced accumulation of Aβ and pTau, as well as of proteinaceous deposits (silver staining) in si-Fshr-injected OVX mice (scale bar, 50 μm). c, The two isoforms of Aβ, namely Aβ40 and Aβ42, were also reduced. d, Notable is the marked increase in dendritic spines (Golgi staining) (scale bar, 5 μm). Statistics: mean ± s.e.m., (a) 3 biological replicates; (c) 5 mice per group; (d) 10 sections from 3 mice per group; unpaired two-tailed Student’s t-test.

Extended Data Fig. 5 |. Recombinant FSH Triggers AD Pathology in 3xTg Mice.

a, Serum FSH levels—both mouse (endogenous) and human (exogenous)—24 h after i.p. injection of 2, 5 or 10 IU human recombinant FSH. b, Serum LH levels also shown. Female 3xTg mice were injected with recombinant FSH (5 IU per mouse, daily, i.p., 3 months). c, Immunohistochemistry for Aβ or pTau in hippocampus post-FSH injection (scale bar, 50 μm). d, Silver staining of the prefrontal cortex, and hippocampal CA1 and dentate gyrus (DG) regions showing enhanced proteinaceous deposits in FSH-injected mice (scale bar, 50 μm). e, Brain mRNA levels of Cebpb, Lgmn, App and Mapt. f, Golgi staining of brain sections from the CA1 region shows reduced spine numbers in FSH-injected mice (scale bar, 5 μm). g, Transmission electron micrographs of hippocampal sections showing reduced synapse numbers post-FSH (scale bar, 1 μm). Immunofluorescence micrographs showing the following image pairs in the hippocampus and/or cortex post-FSH: (h) Aβ (red) and cleaved APP^C586 (green); (i) pTau (red) and cleaved Tau^1–368 (green); (j) Aβ (red) and thioflavin-S (green); and (k) NeuN (red) and TUNEL (green) (scale bar, 20 μm). l, Immunofluorescence showing co-localization of C/EBPβ, AEP, Aβ and pTau to NeuN-positive neurons upon FSH stimulation [10x (scale bar, 300 μm) and 40x (scale bar, 50 μm) magnifications]. Statistics: mean ± s.e.m., (a, b) 3 mice per group; (e) 3 biological replicates, (f, g) 10 sections from 3 mice per group; unpaired two-tailed Student’s t-test.

Extended Data Fig. 6 |. Effect of Recombinant FSH in Triggering AD Pathology in Ovariectomized 3xTg Mice With Oestrogen Replacement.

3xTg mice were ovariectomized at 3 months and supplemented with 17β-estradiol using 90-day-release pellets (E2, 0.36 mg) to render them biochemically eugonadal. The mice were randomly divided to be injected with PBS or recombinant human FSH (5 IU per mouse, daily, i.p., 3 months). a, Serum level of FSH and 17β-estradiol. b, Western immunoblotting showing increased C/EBPβ, AEP, cleaved APP^1–373 and APP^1–585, total Tau, cleaved Tau^1–368 and pTau in the brain after FSH injection. c, Brain AEP enzymatic activity also shown. d, Immunohistochemistry of the hippocampus shows increased expression of Aβ and pTau in the FSH group. Silver staining showed increased proteinaceous deposits in FSH-treated mice (scale bar, 50 μm). Statistics: mean ± s.e.m., mice per group; (a) 4 and (c) 5; unpaired two-tailed Student’s t-test.

Extended Data Fig. 7 |. Effect of Recombinant FSH in Triggering AD Pathology and Cognitive Decline in Male Mice.

Male 3xTg mice were injected with recombinant FSH at 5 IU per mouse daily, i.p. for 3 months. a, Western immunoblots showing increased C/EBPβ, AEP, cleaved APP^1–373 and APP^1–585, total Tau, cleaved Tau^1–368 and pTau in the brain (3 mice per group). b, c, Brain AEP activity (b) and Aβ isoforms, Aβ40 and Aβ42 (c) were also increased with FSH. d, Morris Water Maze test shows enhanced escape latency to mount the platform (seconds). Also shown are integrated escape latency (area under the curve, AUC) and percentage of time spent in the target quadrant (Probe Trial Test). e, Silver staining of the prefrontal cortex, and hippocampus CA1 and dentate gyrus (DG) regions showing enhanced proteinaceous deposits in FSH-injected mice (scale bar, 50 μm). f, Immunohistochemistry for Aβ or pTau in the hippocampus post-FSH injection (scale bar, 50 μm). g, Brain mRNA levels of Cebpb, Lgmn, App and Mapt. h, Golgi staining of brain sections from the CA1 region of the hippocampus showing reduced spine numbers in FSH-injected mice (scale bar, 5 μm). i, Transmission electron micrographs of hippocampal sections showing reduced synapse numbers post-FSH (scale bar, 1 μm). Immunofluorescence micrographs showing the following image pairs: (j) Aβ (red) and cleaved APP^C586 (green); (k) pTau (red) and cleaved Tau^1–368 (green); (l) Aβ (red) and thioflavin-S (green); and (m) NeuN (red) and TUNEL (green) in the hippocampus and/or cortex of male 3xTg mice after FSH (scale bar, 20 μm). Statistics: mean ± s.e.m., mice per group, (b, c) 5, (d) 7, (g) 3, (h, i), 3 (10 sections); unpaired two-tailed Student’s t-test.

Extended Data Fig. 8 |. Effect of FSH in Triggering AD Pathology and Cognitive Decline in Female Wild Type and APP-KI Mice.

In APP-KI mice, three amino acid substitutions (G601R;F606Y;R609H) are knocked into Aβ-coding exon 14 of the APP gene—this results in the non-transgenic expression at basal levels of oligomerizable human Aβ. Female wild type and APP-KI mice were injected with recombinant FSH (5 IU, daily, i.p. 3 months). a–d, In wild type mice, Western immunoblotting showing increased C/EBPβ, AEP, cleaved APP^1–373 and APP^1–585, total Tau, and cleaved Tau^1–368 in whole brain (3 mice per group) (a), as well as increased silver staining (b), elevated AEP activity (c) and increases Aβ isoforms, Aβ40 and Aβ42 (d) upon FSH treatment. e, Morris Water Maze test, however, showed no difference in escape latency to mount the platform (seconds). Also shown are no differences in integrated escape latency (area under the curve, AUC) and percentage of time spent in the target quadrant (Probe Trial Test). f–I, Western immunoblotting showing elevations in C/EBPβ, AEP, cleaved APP^1–373 and APP^1–585, and cleaved Tau^1–368 in whole brain (f), along with enhancements in silver staining (g), AEP activity (h), and Aβ isoforms (i) in APP-KI mice in response to FSH injection. j, There was also a significant spatial memory deficit on the Morris Water Maze test. k, Immunofluorescence micrographs showed increases in the hippocampus and/or cortex of female APP-KI mice post-FSH injected in the following pairs: Aβ (red) and cleaved APP^C586 (green); Aβ (red) and thioflavin-S (green); and NeuN (red) and TUNEL (green). Scale bar: (b, g) 50 μm, (k) 100 μm (magnified view, 10 μm). Statistics: mean ± s.e.m.; mice per group, (c, d, h, i) 5, (e, j) 8; unpaired two-tailed Student’s t-test.

Extended Data Fig. 9 |. C/EBPβ Mediates FSH-Induced AD Neuropathology and Cognitive Decline in 3xTg Mice.

a, Cebpb, Lgmn, App and Mapt mRNA expression following FSH injection to 3xTg or Cebpb^+/− 3xTg mice. Statistics: mean ± s.e.m., N = 3 biological replicates, one-way ANOVA. b, Immunohistochemistry for Aβ and pTau and silver staining for proteinaceous deposits (scale bar, 50 μm). c–e, Immunofluorescence staining for Aβ (red) and C/EBPβ (green) (c) and for pTau (red) and C/EBPβ (green) (d) (scale bar, 20 μm), and Golgi staining for dendritic spines (e) in the hippocampus in female 3xTg or Cebpb^+/− 3xTg mice, post-FSH (scale bar, 5 μm). Statistics: mean ± s.e.m., 10 sections from 3 mice per group, one-way ANOVA. f, Morris Water Maze testing showed no difference in swim speed. Statistics: 9 mice per group, one-way ANOVA.

Extended Data Fig. 10 |. C/EBPβ Mediates Ovariectomy-Induced AD Neuropathology and Cognitive Decline in 3xTg Mice.

a, Cebpb, Lgmn, App and Mapt mRNA expression following ovariectomy of 3xTg or Cebpb^+/− 3xTg mice. Statistics: mean ± s.e.m., 3 biological replicates, one-way ANOVA. b, Immunohistochemistry for Aβ and pTau and silver staining for proteinaceous deposits (scale bar, 50 μm). c–e, Immunofluorescence staining for Aβ (red) and C/EBPβ (green) (c) and for pTau (red) and C/EBPβ (green) (d) (scale bar, 20 μm), and Golgi staining for dendritic spines (e) in the hippocampus in female 3xTg or Cebpb^+/− 3xTg mice (scale bar, 5 μm), post-OVX. Statistics: mean ± s.e.m., 10 sections from 3 mice per group, one-way ANOVA. f, Morris Water Maze test showed no difference in the swim speed between 3xTg and Cebpb^+/− 3xTg mice. Statistics: left to right: 7, 8, 8 mice per group, one-way ANOVA.

Data availability

All original western blots are provided in the Supplementary Information. The original unedited camera images for histology and immunohistochemistry are available online (https://osf.io/9hp8r/). There are no restrictions on data availability. Unique biological material will be made available to other investigators on request. Source data are provided with this paper.