SARS-CoV-2 infects human GnRH neurons and tanycytes, disrupting hypothalamic-pituitary hormonal axes


 Neuroinvasion by SARS-CoV-2 is now accepted. To investigate whether low testosterone levels observed in men with severe COVID-19 could be of central origin, we retrospectively analyzed blood samples from 60 male intensive-care patients and explored SARS-CoV-2 brain entry using animal and cellular models as well as adult COVID-19 patient and fetal human brains. Most hypotestosteronemic patients displayed hypogonadotropic hypogonadism or abnormal hypothalamic-pituitary-gonadal axis regulation. Neurons producing gonadotropin-releasing hormone (GnRH), the master molecule controlling fertility, expressed angiotensin-converting enzyme 2 and neuropilin-1, two host-cell factors mediating infection, and were infected and dying in all COVID-19 patient brains. Tanycytes - hypothalamic glia that regulate GnRH secretion - were also infected. Additionally, human fetal olfactory and vomeronasal epithelia, from which GnRH neurons arise, richly expressed both the above host-cell susceptibility factors and formyl peptide receptor 2, a putative vomeronasal receptor that also appeared involved in SARS-CoV-2 pathogenesis in humans and mice. Finally, a fetal human GnRH cell line expressing all these receptors could be infected by a SARS-CoV-2-like pseudovirus. Together, our findings suggest that GnRH neurons, which may be implicated in brain development and aging in addition to reproduction, are particularly vulnerable to SARS-CoV-2 in both adults and fetuses/newborns, with potentially devastating long-term consequences.


Introduction
Two years into the COVID-19 pandemic, it is increasingly evident that few if any tissues of the human body are resistant to infection by SARS-CoV-2. In addition to the respiratory tract, viral RNAs or proteins have been detected in several peripheral organs, and there are now multiple reports of neuroinvasion that could underlie the various neurological symptoms observed in COVID-19 patients. However, the dysfunction of peripheral organs and bodily processes could also stem from brain infection, with potentially worrying long-term consequences. This is particularly true of the hypothalamus, which lies at the crossroads of several putative infection routes, and whose diverse neuroglial populations participate in both hypothalamic-brainstem physiological circuits as well as hypothalamic-pituitary neuroendocrine circuits that regulate a vast range of functions, including reproduction and fertility. Indeed, a number of scienti c studies and popular reports have suggested that the virus could interfere with reproductive function 1 .
In humans, neurons secreting gonadotropin-releasing hormone (GnRH), the master regulator of the development of the reproductive axis and adult fertility, constitute a sparse population of barely a couple of thousand cells 2 , principally located in the infundibular nucleus of the tuberal region of the hypothalamus 2,3 . These neurons send projections to the external zone of the median eminence, where they release the neurohormone in the vicinity of the pituitary portal blood vessels, which are characterized by a "fenestrated" or highly permeable capillary wall instead of the usual tightly sealed endothelial layer that constitutes the blood-brain barrier (BBB). In the external zone of the median eminence, these GnRH neuronal terminals intermingle with and are ensheathed by the glial processes of tanycytes 3 , specialized radial-glia-like cells that form the oor and walls of the third ventricle. The growth or retraction of tanycytic end-feet, which are in contact with the capillary fenestrations, controls the access of GnRH nerve terminals to the pituitary portal circulation 4 . Blood-borne GnRH then reaches the anterior pituitary, where it acts on pituitary gonadotropes to modulate the secretion of luteinizing hormone (LH) and follicle stimulating hormone (FSH) into the general circulation, which in turn promote gametogenesis and gonadal steroid synthesis by the testes and ovaries 5 . Interestingly, the absence of the traditional BBB in the median eminence, as in certain other circumventricular organs, not only allows peptide neurohormones such as GnRH to easily reach their target cells in the pituitary, it also allows circulating peripheral signals to enter the brain by passive diffusion from the portal capillaries or active transport by tanycytes. In the context of a viral infection, however, this adaptation, necessary for the maintenance of bodily homeostasis by allowing the exchange of information between the brain and the periphery, could instead represent a breach in the brain's defenses against pathogens.
Several studies to date have shown that male COVID-19 patients display lower circulating testosterone levels than their uninfected counterparts 6 , an effect that can last several months after recovery from infection 7 , and that lower testosterone levels are correlated with worse clinical outcomes 8, 9 . Here using human COVID-19 patient blood samples and post mortem brains, mouse models of SARS-CoV-2 brain infection as well as human fetal tissues and cell lines, we investigated whether this hypogonadism in COVID-19 patients could be due to the infection of hypothalamic GnRH neurons rather than purely peripheral, as assumed so far.

Results
We retrospectively measured plasma testosterone and gonadotropin concentrations, which serve as surrogates for GnRH release, in a cohort of 60 male COVID-19 patients, 35-82 years old, hospitalized in the intensive care unit (ICU) of the Lille Medical University Hospital (CHU Lille). Among these, around half the patients, of all ages, showed normal or near-normal testosterone levels (Group 4) whereas half showed evidence of hypogonadism or decreased testicular function as de ned by severely low total testosterone levels (< 0.9 ng/ml) during their rst week in the ICU (Figure 1a). However, contrary to previous reports, only 4 of these 31 hypogonadal patients displayed the compensatory elevation of LH and FSH levels expected of individuals with a properly functioning hypothalamic-pituitary gonadal (HPG) axis (Group 3), whereas 16 showed intermediate LH levels of 2-12 IU/L (Group 2), and 11 presented clear hypogonadotropic hypogonadism, with LH levels below 2 IU/L (Group 1) (Figure 1b,c). COVID-19 patients who underwent an extended ICU stay were also sampled at weeks 2 and 4 in the ICU. In these, the functioning of the HPG was seen to switch between normal (Groups 3 and 4), i.e. where LH and testosterone were inversely correlated (log-linear model, r 2 = 0.389, p = 1.88 −11 ), and abnormal (Groups 1 and 2), i.e., where both LH and testosterone levels were low (Figure 1a (1) = 10.38, p = 0.0013). Since obesity, which is characterized by resistance to elevated circulating levels of the adipocyte-derived metabolic hormone, leptin 10 , is a known risk factor for severe COVID-19 11 , we also analyzed plasma leptin levels and body mass index (BMI) in these patients. While leptin was negatively correlated with testosterone levels, as expected 12 (Supplementary Figure 1c), neither BMI nor leptin levels appeared to be confounding factors for the phenotype (see Supplementary Table 1). Surprisingly however, given previous reports of a correlation between the intensity of the in ammatory response and low testosterone levels 13 , the level of C-reactive protein (CRP), an indicator of in ammation, was also not found to be a confounding factor for hypogonadism (Supplementary gure 1d). Interestingly, COVID-19 patients with low LH/FSH levels (Groups 1 and 2) also displayed low levels of thyroid stimulating hormone, TSH (Supplementary gure 1e), whose releasing hormone, TRH, is secreted by hypothalamic neuroendocrine terminals in the median eminence similarly to GnRH 14,15 . Together, these results suggest that the severely low total testosterone levels seen in the majority of COVID-19 patients in intensive care are not just a re ection of gonadal insu ciency but of impaired hypothalamic function, or hypogonadotropic hypogonadism.
A central impairment of the HPG axis could result from either an absence of mature GnRH neurons or de cient/abnormal GnRH release 16 . Unlike other hypothalamic neurons driving bodily functions, GnRH neurons are not born in the brain but originate in the olfactory placode and migrate from the nose to the brain during embryogenesis 17,18 , remaining in contact with the olfactory bulb (OB) via long dendrites 2 . Among the molecular cues key to GnRH neuronal development and function are the class 3 semaphorins, ligands of Neuropilin 1 (NRP1), which GnRH neurons themselves also express [19][20][21][22] . Together with angiotensin-converting enzyme 2 (ACE2), the known cell-surface receptor for SARS-CoV-2, and several proteases, including notably transmembrane protease, serine 2 (TMPRSS2), that cleave and prime the SARS-CoV-2 spike protein (S-protein) 23 , NRP1 is also suspected of enhancing SARS-CoV-2 viral entry into target cells 24,25 . In light of these facts, we then assessed whether GnRH neurons could themselves be susceptible to infection by SARS-CoV-2 by isolating these neurons from GnRH::Gfp mice with the aid of uorescence-activated cell sorting and quantifying the above molecules using RT-PCR. Although transcripts for TMPRSS2 were undetectable in these neurons, they expressed both Nrp1 as well as Ace2 (Supplementary Figure 2a), suggesting that in species such as humans in which the virus binds to the native ACE2 (unlike wild-type mice, which are infected by other coronaviruses but not SARS-CoV-2), adult GnRH neurons could indeed be a target of SARS-CoV-2.
Next, in order to investigate potential routes of infection of GnRH neurons, we used K18-hACE2 mice, a widely studied mouse model of coronavirus brain infection in which human ACE2 (hACE2) is expressed under the control of the keratin 18 promoter 26 . We found that 7 days after intranasal infection of these mice with SARS-CoV-2, transcripts for the nucleocapsid or N-protein were detected throughout the brain by RT-PCR using a set of Food and Drug Administration (FDA)-approved primers used to diagnose COVID-  Figure 2c). Immuno uorescence labeling for the S-and N-proteins of SARS-CoV-2 showed massive viral infection in the whole brain in 4 out of 6 mice and viral infection restricted to the hypothalamus in 2 mice, recapitulating the RT-PCR data obtained from the same animals ( Figure 2c). Our observations strengthen the possibility of a hematogenic route of viral infection through hypothalamic circumventricular organs lacking a BBB, which we have also recently shown to be targeted by SARS-CoV-2 27 , in addition to the more commonly accepted olfactory route 28, 29 . Notably, several GnRH neurons were indeed seen to contain viral proteins (Figure 2c). The uorescent signal for viral markers was completely absent in the brain of mock-infected K18-hACE2 mice (Figure 2d), validating the speci city of our tools to detect SARS-CoV-2 infection.
Given the putative susceptibility of GnRH neurons to SARS-CoV-2 and the existence of at least two potential infection routes revealed by our mouse models above, we next asked whether these suppositions were borne out in COVID-19 patients. We looked for the presence of viral proteins and RNA in the brains of four patients who died of COVID-19, including one who displayed viremia at the time of death, and compared them with the brains of four age-matched uninfected patients (negative nasal swab PCR or deceased before the pandemic). All patients died in the ICU (Supplementary Table 1). Using the same 3 pairs of primers used to diagnose patients in the USA and used to detect viral infection in K18-hACE2 mouse brains above (Figure 2a), we observed that three of the four COVID-19 patients, including the one who had viremia at the time of death, had readily detectable levels of N-protein transcripts in the hypothalamus ( Figure 3a). Next, we succeeded in performing multiplex uorescent in situ hybridization (RNAscope) in the hypothalamus in human brain tissue xed in paraformaldehyde for longer than a week, and detected S-protein transcripts in vessels, some neuron-like cells and cells of the ependymal wall, while such labeling was absent in control patients despite the strong visualization of positive-control U6 mRNA (Figure 3b).
Immuno uorescence labeling also revealed abundant N-protein and double-stranded RNA (dsRNA) in numerous cells of the median eminence/infundibular nucleus of COVID-19 patients (Figure 3c,d), unlike uninfected controls (Figure 3c), indicating robust SARS-CoV-2 entry and replication. However, while Nprotein was often colocalized with tanycytic processes, dsRNA labeling was fainter in tanycytic cell bodies lining the ventricular wall than in the nuclei of non-tanycytic cells morphologically associated with vimentin-immunoreactive tanycytic processes (Figure 3c,d). Interestingly, immunolabeling for S-protein, which mediates host-cell entry 23 , was extremely high in ACE2-and TMPRSS2-coexpressing tanycytic end-feet, which contact fenestrated capillaries at the external pial surface of the median eminence ( Figure   3e,f) and where they are known to interact morphologically with GnRH and TRH axon terminals 3,14,15 . In addition, despite the extreme paucity of GnRH neurons and their scattered distribution in the hypothalamus, we identi ed NRP1-or ACE2-positive GnRH neurons (Figure 4a,b), including several positive for the viral S-protein, in all COVID-19 patient brains ( Figure 4b). Unexpectedly, more than onethird of GnRH neurons in COVID-19 patient brains displayed a bloated or abnormal morphology rather than the typical fusiform morphology, suggesting that they were sick or dying, while the number of GnRH neurons with an abnormal morphology was negligible in control brains (Figure 4c,d). In keeping with the view that GnRH neurons were being killed by SARS-CoV-2 infection, GnRH neurons in all infected patient brains, but none in control subjects, were immunoreactive for cleaved caspase-3, and a majority of these dying neurons also displayed an abnormal morphology (Figure 4c,e). Further supporting the putative death or dysfunction of GnRH neurons on a massive scale, RT-PCR analysis of the hypothalamus of four COVID-19 patient brains revealed an almost complete disappearance of GnRH transcripts as compared to ve control brains, whereas NRP1, which is expressed by other cell populations in the hypothalamus, was not signi cantly affected (Figure 4f).
Considering that GnRH neurons, which migrate through the OB to the hypothalamus, continue to maintain dendritic contact with the former, we also examined this tissue as an alternative route of infection. While we did not detect any residual GnRH neurons in the OB, immuno uorescence labeling showed abundant S-protein expression along with ACE2 and TMPRSS2 in the olfactory nerve layer (ONL), where olfactory marker protein (OMP)-expressing axons from sensory neurons of the olfactory epithelium enter the OB The existence of an olfactory route combined with the vulnerability of GnRH neurons to SARS-CoV-2 also raises another specter: that of an infection of these neurons by vertical transmission from the mother during embryonic development, which could impinge on the establishment and function of the HPG axis postnatally. Indeed, the migration, maturation and correct adult function of GnRH neurons, which are born in the nose during embryonic life, may be important for other aspects of brain development, such as the regulation of metabolism 19,31 . We therefore analyzed the expression of SARS-CoV-2 vulnerability factors in the olfactory epithelium of 7-, 11-and 14-week-old human fetuses. We found abundant expression of ACE2 and TMPRSS2 in neurons in the olfactory epithelium as well as their TAG-1-immunoreactive axons, which extend into the OB (Supplementary Figure 4), adding to previous reports of ACE2 and TMPRSS2 expression in other cell populations of the olfactory epithelium 32 . Serendipitously, we noticed on exploring the fetal nasal epithelium that neurons composing the vomeronasal organ, which disappears in adults 33 , also abundantly expressed ACE2 and TMPRSS2 (Figure 5a-e), rmly establishing the susceptibility to viral infection of the birthplace of GnRH neurons (Figure 5f,g) as well as the axonal tracts along which they migrate into the brain (Figure 5h). Both GnRH neurons and the axonal scaffold composed of the olfactory and vomeronasal nerves also abundantly expressed NRP1 (Figure 5h), as we have previously shown 20 , potentially further favoring SARS-CoV-2 cell entry 24,25 .
Intriguingly, one of the molecules found to be differentially expressed in the lungs of COVID-19 patients is the formyl peptide receptor, FPR2/FPRL1/ALX 34 , a G protein-coupled receptor and homolog of the Fpr family of vomeronasal receptors in mice. While FPR2 and its mouse equivalent are not considered to act as vomeronasal receptors 35,36 , it is active both in the brain and in immune cells, and is best known for its proin ammatory or in ammation-resolving actions depending on the activating ligand and signaling pathway triggered 37 . It also binds a variety of viral peptides and regulates viral RNA replication 38- 40 . In light of all these facts, we investigated the presence of FPR2 protein by immunolabeling in the fetal nasal epithelium. To our surprise, FPR2 was not only richly expressed by neurons of the olfactory and vomeronasal epithelia and their axonal tracts, which constitute the migratory scaffold of GnRH neurons, it was also expressed by several GnRH neurons themselves (Figure 6a,b). Given the implication of FPR2 in the binding and replication of other RNA viruses, we reexamined the brains of infected K18-hACE2 mice and human COVID-19 patient brains for the expression of FPR2 transcripts and proteins. Interestingly, Fpr2, whose expression in control animals was quite low, was highly induced in the brains of K18-hACE2 mice following 7 days of SARS-CoV-2 infection (Figure 6c), and Fpr2 immunolabeling showed a strong colocalization with the distribution of SARS-CoV-2 S-protein in infected mice (Figure 6d). Immunolabeling studies also revealed the presence of FPR2 protein in the ONL and various hypothalamic cells in human COVID-19 patient brains (Figure 6e,f), with a remarkable increase in its expression in COVID-19 patient OBs as compared to control brains, reminiscent of our observations in infected mice and suggestive of its known upregulation by other viral dsRNAs 41 . Together, these results provide a striking indication that the combination of ACE2, NRP1 and FPR2 expression in adult and fetal human GnRH neurons and the human olfactory and vomeronasal epithelia during development could render these structures particularly vulnerable to SARS-CoV-2 infection.
Finally, to verify that human fetal GnRH neurons can indeed be infected by SARS-CoV-2, we tested the ability of pseudotyped viral particles expressing the full-length SARS-CoV-2 S-protein and the ZsGreen reporter gene 42 to infect a human GnRH-expressing cell line isolated from the fetal vomeronasal organ, FNC-B4 43 . Interestingly, cells differentiating into GnRH neurons expressed ACE2 as well as NRP1 protein, and ACE2, NRP1 and FPR2 mRNAs (Figure 7a,b). Flow cytometry experiments further con rmed that a fraction of these cells were infected by the pseudotyped virus particles and expressed green uorescence 48h later (p=0.0235, two-tailed unpaired t-test) (Figure 7c,d), strongly suggesting that at least some GnRH neurons in human fetuses could indeed be infected by SARS-CoV-2 in case of vertical transmission from infected mothers (see for example 44,45 ).

Discussion
To summarize, our work demonstrates that GnRH neurons both in adult humans and in fetuses express multiple susceptibility factors for SARS-CoV-2 and can be infected by the virus, with potentially devastating long-term effects on later fertility as well as essential cognitive and metabolic processes. Multiple routes of infection of GnRH neurons in COVID-19 patients are possible. While we have previously shown that brain endothelial cells are infected by SARS-CoV-2, suggesting a hematogenic route for viral entry across the BBB 27 , our current ndings also provide support for an olfactory route both in fetuses and in adults. In addition, blood-borne SARS-CoV-2 viral particles extravasating from fenestrated vessels of the pituitary portal blood system may also directly attack GnRH neuroendocrine terminals that dip into the perivascular space of these vessels, or infect ACE2-, TMPRSS2-and possibly FPR2-expressing tanycytes, whose end-feet surround GnRH neuronal terminals. While an alteration of GnRH secretion due to the infection of either the neurons themselves or of tanycytes could cause transient hypogonadotropic hypogonadism, in the worst-case scenario, the death of GnRH neurons that we observed in this normally stable population and the dramatic decrease in GnRH transcripts in the hypothalamus of COVID-19 patient brains, may have long-lasting effects on fertility in patients who survive infection. In addition, since tanycytes also tightly control the release of other neurohormones such as TRH into the pituitary portal circulation 14,15 , and the external zone of the median eminence also harbors the neurosecretory terminals of TRH neurons, the infection of tanycytes or these other neurons associated with them may durably impact other physiological processes, and lead to endocrine-triggered neurological or psychiatric disorders. For instance, depression is known to occur in at least 33% of patients diagnosed with COVID-19 46 and in up to 80% of hospitalized COVID-19 patients 47 . Strengthening this idea, the group of COVID-19 patients with hypogonadotropic hypogonadism also showed impaired TSH levels in our study. In addition, SARS-CoV-2 viral infection was also visualized in the paraventricular nucleus of the hypothalamus, where TRH neuronal cell bodies are located (Supplementary Figure 5), providing further support for a hematogenic route of infection in addition to the olfactory one, and emphasizing the need for the long-term monitoring of hormone levels in COVID-19 survivors.
The present study was conducted using blood samples from patients hospitalized in the ICU with severe COVID-19. However, while ICU hospitalization is itself known to be correlated with a transient drop in testosterone levels in male patients [7][8][9] , the fact that more than half of these COVID-19 patients clearly displayed hypogonadism of hypothalamic origin points to the viral infection and not merely the ICU hospitalization as the cause of this alteration. This is especially preoccupying considering the current global decline in human fertility, including in Europe (https://www.euro.who.int/__data/assets/pdf_ le/0010/73954/EN63.pdf). Whether similar phenomena could occur in patients with less severe infection remains to be investigated, perhaps in patients with "long COVID" to begin with. In addition, as a signi cant proportion of COVID-19 patients display anosmia or dysosmia 48 and defects in GnRH neuronal function may also be associated with alterations in olfactory perception 16,19,49 , it would be interesting to investigate the long-term consequences of SARS-CoV-2 infection on fertility in anosmic and normosmic COVID-19 patients. In this regard, mutations in NRP1, a molecule that can potentiate SARS-CoV-2 host cell infection 24,25 , causes Kallmann syndrome in humans, associating hypogonadotropic hypogonadism with anosmia 22 , suggesting that NRP1expressing cells like GnRH neurons and their migratory scaffold (17) might be especial targets of the virus. GnRH is also involved in regulating normal energy metabolism (29) and age-related changes in cognitive function 50 , raising the possibility that the death or long-term dysfunction of these neurons following their infection may lead to metabolic and cognitive disorders in addition to reproductive ones.
In addition to ACE2 and NRP1, already known to mediate the host-cell binding and internalization of SARS-CoV-2, we identi ed a novel endogenous viral target in the brain and nasal epithelia -FPR2. Intriguingly, FPR1, which shares a number of ligands with FPR2, though not always with the same effect, binds a highly conserved sequence of the SARS S-protein C-terminal region that is involved in membrane fusion 51 and is also present in the SARS-CoV-2 S-protein (https://www.uniprot.org/uniprot/P0DTC2#sequences). Together with our observations regarding the dramatic upregulation of FPR2 in the OB of COVID-19 patients as well as the hypothalamus of infected K18-hACE2 mice, this suggests that FPR2 is not only involved in SARS-CoV-2 pathogenesis but could act as an alternative receptor or co-receptor for the virus in both the brain and potentially the lung and other tissues. Its presence in GnRH neurons and their migratory scaffold, along with ACE2 and NRP1, could thus represent a triple whammy for these neurons. Interestingly, the modulation of systemic aging by GnRH neurons appears to be mediated by its regulation of the NFκB pathway 50 . The fact that FPR2 is known to modulate in ammation through the same pathway 52 , and that moreover, the SARS-CoV-2 main protease, Mpro, cleaves NEMO, an essential modulator of NFκB that also plays a role in the survival of certain cells 27 , suggests that the presence of FPR2 in GnRH neurons and its eventual upregulation by the virus may have far-reaching implications for neuroin ammation and neurodegeneration 36 . In addition, the unexpected expression of FPR2 in human fetal olfactory and vomeronasal tissues is intriguing leading one to question the traditional view that it is a uniquely "immune/in ammatory" FPR rather than a pheromone receptor in humans 35,36 .
In light of the unusual vulnerability of fetal GnRH neurons, particular attention must also be paid to the consequences of maternal or perinatal COVID-19 infection in neonates 53 , since the rst postnatal activation of the HPG axis, i.e., minipuberty, a phenomenon that plays a key role in the later maturation of the reproductive system 54,55 and likely also in brain development in a broader sense, occurs during the infantile period. The impairment of minipuberty 56,57 , for example by premature birth 58,59 , may be correlated with the incidence of a range of non-communicable diseases or metabolic dysfunction later in life 58,59 . Studies following cohorts of babies born during the pandemic, and possibly treatments aimed at normalizing GnRH secretion, are therefore essential from a public health point of view to avoid a second pandemic of such diseases in the decades to come.

Ethics authorizations
All human tissues were obtained in accordance with French laws (Good Practice Concerning the Conservation, Transformation, and Transportation of Human Tissue to Be Used Therapeutically, published on December 29, 1998).
Adult COVID-19 patient brains and blood samples and control brains were obtained under authorization for the GonadoCOVID study (French protocol # 2-20-056 id8504) and authorized by the Lille Neurobiobank.
The studies on human fetal tissue were approved by the French agency for biomedical research (Agence de la Biomédecine, Saint-Denis la Plaine, France, protocol no.: PFS16-002). Non-pathological human fetuses were obtained at GW7, GW11 and GW14 from voluntarily terminated pregnancies after written informed consent from the donors (Gynecology Department, Jeanne de Flandre Hospital, Lille, France).

COVID-19 patient and control brains
The brains of 4 subjects (3 males and 1 female) who died of COVID-19 infection in the Lille University Hospital and 5 control subjects (4 males and 1 female) who did not test positive for COVID-19, including 2 who died before the pandemic began, were used for this study. COVID-19 and control subjects were matched for age, sex and comorbidities as far as possible. Their clinical characteristics are summarized in Table 1.
In keeping with strict protocols regarding the treatment of SARS-CoV-2-infected human tissues, human brains were immersion-xed in 10% formalin for 1 week at room temperature. The hypothalamus was then dissected out and immersion-xed in 4% paraformaldehyde in PBS 0.1M, pH7.4, for an additional 48h at 4°C, cryoprotected in 30% sucrose for an additional week at 4°C, embedded in Tissue-Tek and frozen in liquid nitrogen at the crystallization temperature of isopentane.

Human fetuses
Non-pathological human fetuses (7, 11 and 14 gestational weeks (GW), n = 1 per developmental stage) were obtained from voluntarily terminated pregnancies after written informed consent was obtained from the parents (Gynecology Department, Jeanne de Flandre Hospital, Lille, France). Fetuses were xed by immersion in 4% PFA at 4°C for 5 days. The tissues were then cryoprotected in PBS containing 30% sucrose at 4°C overnight, embedded in Tissue-Tek OCT compound (Sakura Finetek), frozen on dry ice, and stored at -80°C until sectioning. Frozen samples were cut serially at 20 mm intervals with a Leica CM 3050S cryostat (Leica Biosystems Nussloch GmbH) and immunolabeled, as described below. FNC-B4 human embryonic GnRH neuronal cell line FNC-B4 cells 60 were kept in culture in Coon's modi ed Ham's F12 medium complemented with 10% FBS at 37°C and 5% CO2 and medium was changed twice weekly. Cells were used for pseudovirus infection and gene expression analysis when they reached 70% con uency. Gene expression assays for GnRH, ACE2, NRP1 and FPR2 were carried out on uninfected cells by quantitative RT-PCR.  RNAscope hydrogen peroxide treatment and target retrieval), incubated with RNAscope protease IV for 10 min at room temperature, and the signal revealed using the RNAscope multiplex uorescent assay.

K18-hACE2 mice and SARS-CoV-2 infection
Immunohistochemistry and quanti cation for viral and host-cell markers in human and mouse tissues Immunolabeling in the human brain using the two antibodies to human ACE2 (R&D Systems, with tyramide ampli cation, and Abcam, without ampli cation), labeled similar cells.
For immunolabeling of human fetuses, 20 µm-thick sections of entire heads at GW 7, GW 11 and GW 14 were processed as follows. Slides rst underwent antigen retrieval for 20 minutes in a 5mM citrate buffer heated to 90°C, then were rinsed in TBS and blocked/permeabilized for 2 hours at room temperature in TBS + 0.3% Triton + 0.25% BSA + 5% Normal Donkey Serum ("Incubation solution", ICS). Sections were then incubated with primary antibodies (see Antibody   into 4 groups depending on whether total testosterone levels were severely decreased (hypogonadism; < 0.9 ng/ml) or not, and whether LH levels were appropriate (normal HPG axis function) or inappropriate (abnormal HPG axis function) for the observed testosterone level. A two-tailed unpaired t-test was used to estimate the signi cance of the difference between groups, where n=62 samples from 60 patients during rst week of sampling (n for Group 1 = 11, Group 2 = 16, Group 3 = 4, Group 4 = 31) and *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns = non-signi cant.