Liver FGF21 mediates sex-specific adaptation to juvenile protein malnutrition

Liver FGF21 mediates sex-specific adaptation to juvenile protein malnutrition | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results Liver FGF21 mediates sex-specific adaptation to juvenile protein malnutrition View ORCID Profile Amélie Joly , View ORCID Profile Lucas Rebiffé , Anne Lambert , Estelle Caillon , View ORCID Profile Sandrine Hughes , View ORCID Profile Benjamin Gillet , View ORCID Profile Isabelle Rahioui , Pedro Da Silva , Justine Bruse , Anne Fougerat , View ORCID Profile Filipe De Vadder , View ORCID Profile François Leulier doi: https://doi.org/10.1101/2025.07.09.663702 Amélie Joly 1 Institut de Génomique Fonctionnelle de Lyon, École Normale Supérieure de Lyon, CNRS UMR 5242, Université Claude Bernard Lyon 1 , 69364 Lyon Cedex 07, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Amélie Joly Lucas Rebiffé 1 Institut de Génomique Fonctionnelle de Lyon, École Normale Supérieure de Lyon, CNRS UMR 5242, Université Claude Bernard Lyon 1 , 69364 Lyon Cedex 07, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Lucas Rebiffé Anne Lambert 1 Institut de Génomique Fonctionnelle de Lyon, École Normale Supérieure de Lyon, CNRS UMR 5242, Université Claude Bernard Lyon 1 , 69364 Lyon Cedex 07, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site Estelle Caillon 1 Institut de Génomique Fonctionnelle de Lyon, École Normale Supérieure de Lyon, CNRS UMR 5242, Université Claude Bernard Lyon 1 , 69364 Lyon Cedex 07, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site Sandrine Hughes 1 Institut de Génomique Fonctionnelle de Lyon, École Normale Supérieure de Lyon, CNRS UMR 5242, Université Claude Bernard Lyon 1 , 69364 Lyon Cedex 07, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Sandrine Hughes Benjamin Gillet 1 Institut de Génomique Fonctionnelle de Lyon, École Normale Supérieure de Lyon, CNRS UMR 5242, Université Claude Bernard Lyon 1 , 69364 Lyon Cedex 07, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Benjamin Gillet Isabelle Rahioui 2 INSA Lyon, INRAE, BF2I , UMR203, 69621 Villeurbanne, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Isabelle Rahioui Pedro Da Silva 2 INSA Lyon, INRAE, BF2I , UMR203, 69621 Villeurbanne, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site Justine Bruse 3 Toxalim (Research Center in Food Toxicology), INRAE, ENVT, INP-PURPAN, UMR1331, Université de Toulouse , Toulouse, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site Anne Fougerat 3 Toxalim (Research Center in Food Toxicology), INRAE, ENVT, INP-PURPAN, UMR1331, Université de Toulouse , Toulouse, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site Filipe De Vadder 1 Institut de Génomique Fonctionnelle de Lyon, École Normale Supérieure de Lyon, CNRS UMR 5242, Université Claude Bernard Lyon 1 , 69364 Lyon Cedex 07, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Filipe De Vadder François Leulier 1 Institut de Génomique Fonctionnelle de Lyon, École Normale Supérieure de Lyon, CNRS UMR 5242, Université Claude Bernard Lyon 1 , 69364 Lyon Cedex 07, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for François Leulier For correspondence: filipe.de_vadder{at}ens-lyon.fr francois.leulier{at}ens-lyon.fr Abstract Full Text Info/History Metrics Supplementary material Preview PDF Abstract Dietary protein sufficiency during childhood is essential for healthy growth and tissue development. Chronic protein deficiency leads to stunting, height-for-age more than two standard deviations below the median, and affects 149 million children under five ( https://www.who.int/publications/i/item/9789240073791 ). Linear growth is governed by the somatotropic axis, wherein pituitary growth hormone (GH) stimulates hepatic insulin-like growth factor 1 (IGF-1) to promote bone elongation 1 . Although boys are consistently more stunted than girls across populations (ranging from 18% to 45%) 2 , 3 , mechanisms underlying this apparent resilience remain unclear. Rodent studies suggest that the protein-to-carbohydrate ratio modulates life-history traits such as metabolism, lifespan, and reproduction, with evidence of sex-dependent sensitivity 4 , 5 . Whether such dimorphism reflects coordinated developmental adaptations in response to protein scarcity, particularly in females, has not been fully explored. Here, we show that juvenile dietary protein restriction causes sexually dimorphic adaptations with pronounced stunting in males and delayed reproductive maturation in females. Using liver-specific deletion, we identify fibroblast growth factor 21 (FGF21) as a sex-specific hepatic regulator of female somatic growth and reproductive development, revealing a dimorphic endocrine adaptation to early-life protein scarcity. Our findings position FGF21 as a critical physiological checkpoint coordinating growth and reproductive timing in response to nutritional stress. Main Text To model sustained protein restriction during juvenile development, we fed male and female C57BL/6N mice either a standard AIN-93G control diet 6 (CD; 20% kcal from protein) or an isocaloric low-protein diet 7 (LPD; 5% kcal from protein) from weaning onwards ( Extended Data Table 1 ). LPD markedly suppressed growth in males, with reduced body length by postnatal day (P) 28 and persistent stunting through P56 ( Fig. 1a ). In contrast, LPD females showed only transient early growth impairment and fully caught up by P56 ( Fig. 1b ). This sex difference was confirmed by large effect sizes in males (Cohen’s d > 5) versus minimal or transient effects in females ( Fig. 1c ). Body weight followed a similar pattern ( Fig. 1d–f ). Download figure Open in new tab Figure 1 Sex-specific effects of early-life protein restriction on somatic growth. a–b , Postweaning size curves in males ( a ) and females ( b ) fed control (CD) or low-protein diet (LPD) from P21 to P56. c , Effect size (Cohen’s d ) of LPD vs CD on body length in males and females over time. Dashed lines represent 95% confidence intervals. d–e , Postweaning weight curves in males ( d ) and females ( e ) from P21 to P56. f , Effect size (Cohen’s d ) of LPD vs CD on body weight in males and females over time. Dashed lines represent 95% confidence intervals. g–h , Growth rate (g/day) in males ( g ) and females ( h ) from P21 to P56. i , Femur length at P56. j , Circulating IGF1 concentrations (serum) in males and females, from P28 to P56. k , Hepatic Igf1 expression (log 2 relative to reference genes) in males and females, at P28, P42 and P70. l , Cumulative food intake per mouse (kcal) in males (top) and females (bottom), modelled using body weight as a covariate, over successive 7-day intervals. m–n , Feed efficiency ( m , weight gain per kcal) and protein-specific efficiency ( n , weight gain per protein-derived kcal) over successive 7-day intervals. Unless otherwise stated, data represent mean ± sd; symbols indicate post hoc significance of diet effect (* P □<□0.05, ** P □<□0.01, *** P □<□0.001), ns , not significant. The number of mice used in each panel is provided in Extended Data Table 2 . Growth rate analysis further showed sustained suppression in males starting at P21, whereas LPD females displayed a distinct trajectory, with peak growth rate delayed to around P30 ( Fig. 1g–h ). Femur length at P56 was reduced in LPD males but unaffected in females ( Fig. 1i ). Although circulating IGF-1 declined in both sexes under LPD, the reduction was more pronounced in males ( Fig. 1j ), despite similar hepatic Igf1 downregulation across sexes ( Fig. 1k ), suggesting post-transcriptional regulation or systemic buffering. Despite reduced growth, cumulative food intake normalized to body weight was comparable across groups ( Fig. 1l , Extended Data Fig. 1a–c), ruling out caloric intake as the cause of stunting. Instead, analysis of feed efficiency revealed that LPD males became less efficient overall, while females preserved total efficiency ( Fig. 1m ). However, only females increased efficiency specifically for protein-derived calories ( Fig. 1n ), with no similar gain for carbohydrate or lipid sources ( Extended Data Fig. 1d–e). In contrast, males maintained protein efficiency but showed decreased carbohydrate and lipid efficiency, contributing to their reduced total efficiency. In the C57BL/6N (B6N) background, these findings uncover a sex-specific adaptive response to protein scarcity: females prioritize somatic growth by enhancing protein utilization, whereas males fail to buffer growth under dietary stress. Under nutrient stress, organisms must balance growth and reproduction, a trade-off shaped by conserved life-history strategies 8 , 9 . Puberty represents a sensitive checkpoint in this balance, especially under dietary constraints 8 , 10 . To assess how this trait manifests under protein restriction, we examined sex-specific reproductive maturation in mice raised on LPD. Despite pronounced growth stunting, LPD males entered puberty on time, with no difference in balano-preputial separation (BPS) compared to controls ( Fig. 2a ), and preserved sperm counts and seminal vesicle size ( Fig. 2b–c ). However, testis weight was reduced ( Fig. 2d ), and major urinary proteins (MUPs), androgen-regulated sexual signals 11 , were markedly decreased ( Extended Data Fig. 2a), indicating reduced investment in secondary sexual traits while core reproductive function remained intact. Download figure Open in new tab Figure 2 Sex-specific effects of early-life protein restriction on reproductive maturation. a , Cumulative onset of balano-preputial separation (BPS) in males fed control (CD) or low-protein diet (LPD) from P21 to P56. b, Sperm count at P56 in adult males. c, Seminal vesicle weight at P56 plotted against body weight. Dashed line represents the fitted ANCOVA model without interaction (main effects only). d, Testis weight at P56. e, Cumulative onset of vaginal opening (VO) in females from P21 to P56. f, Cumulative onset of first oestrus in females from P21 to P56. g , Representative oestrus cycles of one CD-fed and one LPD-fed female, from P56 to P70. h, Number of complete oestrous cycles over 14 days (P56 to P70). i, Uterus weight at P56. j, Interval (days) between VO and first oestrus. Data are mean ± sd unless stated otherwise. Symbols indicate diet effect (* P □<□0.05, ** P □<□0.01, *** P □<□0.001), ns , not significant. Shaded areas ( a, e, f ) represent 95% confidence intervals from Kaplan-Meier survival models. “+” indicates right-censored observations. The number of mice used in each panel is provided in Extended Data Table 2 . Females, by contrast, showed clear reproductive delay: LPD shifted vaginal opening (VO) and first oestrus to later ages ( Fig. 2e–f ), reduced cycle frequency during early reproductive maturity ( Fig. 2g–h , Extended Data Fig. 2b), and decreased uterus weight at P56 ( Fig. 2i ). The interval between VO and first oestrus was extended ( Fig. 2j ), indicating delayed axis activation and oestrogen signalling. To understand the physiological logic behind these dimorphic responses, we modelled puberty using logistic regression, a method that estimates how likely an event (e.g., puberty onset) is to occur based on predictors such as age, weight, or diet. In males, BPS was tightly age-predicted, and diet had no explanatory value ( Extended Data Fig. 2c–e), suggesting a robust developmental clock. In females, body weight emerged as the primary predictor of VO timing ( Extended Data Fig. 2f – h ), and faster prepubertal growth was associated with earlier VO onset ( Extended Data Fig. 2i), consistent with a growth-dependent gating mechanism. However, this relationship collapsed under LPD: below a growth threshold of ∼0.48 g/day, VO timing became unpredictable, with a fivefold increase in residual variance ( Extended Data Fig. 2j). A permutation test confirmed that VO was no longer reliably gated by growth under this threshold ( P < 0.0001), highlighting disrupted coupling under protein restriction. This loss of predictability, or “decanalization” (sensu Waddington), suggests that LPD unmasks latent variation in developmental timing, a phenomenon common when environmental stress disrupts physiological homeostasis 12 . Together, these findings point to divergent sex-specific strategies: in females, delayed puberty tracks impaired early-phase growth, while males initiate reproductive maturation despite persistent stunting. This supports a model in which female puberty is tightly linked to somatic state during the early growth window, whereas male timing is less constrained. Having established distinct sex-specific growth and reproductive responses to protein restriction, we next asked which physiological adaptations support these divergent outcomes. Despite a fourfold drop in dietary protein, circulating amino acid concentrations remained remarkably stable in both sexes and across feeding states, except for selected essential amino acids such as branched-chain amino acids and phenylalanine ( Fig. 3a–c , Extended Data Fig. 3a–c). Intriguingly, portal–caval gradients pointed to active hepatic buffering: systemic amino acid levels exceeded portal concentrations postprandially, but the gradient reversed in the postabsorptive state ( Fig. 3a , Extended Data Fig. 3a), consistent with dynamic hepatic uptake and redistribution. Download figure Open in new tab Figure 3 Sex-specific hepatic adaptation to juvenile protein restriction. a–c, Total ( a ), essential ( b ), and branched-chain ( c ) amino acid concentrations in portal and cava veins at P56, in postprandial state. d, Heatmap of postabsorptive hepatic amino acid concentrations at P56, stratified by sex and diet. e, Enriched KEGG pathways from gene set enrichment analysis (GSEA; adjusted P < 0.05), in males (top) and females (bottom), curated using amino acid metabolism–related gene sets from Green et al. (2022). Pathways related to amino acid metabolism and nutrient stress are highlighted in red. f, Volcano plots of liver transcriptomes in males (top) and females (bottom), showing differentially expressed genes (DEGs; FDR□□1). Selected nutrient stress–responsive genes are highlighted in red. g , Heatmap of relative expression of BCAA-related catabolic genes, transporters (black box) and downstream targets. h–i, Heatmap of relative expression of canonical ATF4 ( h ) and PPARα ( i ) target genes in the liver. j , Hepatic Fgf21 expression (log 2 relative to reference genes) in males and females, at P28, P42 and P70. k , Circulating FGF21 concentrations (serum) in males and females, from P28 to P70. Unless otherwise stated, data are mean ± sd; symbols indicate post hoc significance of diet effect (* P □<□0.05, ** P □<□0.01, *** P □<□0.001), ns , not significant. The number of mice used in each panel is provided in Extended Data Table 2 . This paradoxical stability in circulating amino acids, even under a severe nutritional challenge, strongly suggests the existence of compensatory tissue-level mechanisms, most likely centred in the liver, the primary organ responsible for amino acid metabolism 13 . Liver metabolomics in the postabsorptive state confirmed marked sex- and diet-dependent shifts in hepatic amino acid content ( Fig. 3d ). In males, several non-essential amino acids, including glutamine, were elevated, consistent with nitrogen recycling and de novo synthesis ( Extended Data Fig. 3b,c). Females showed fewer changes, suggesting subtler hepatic remodelling. To interpret these differences across multiple interrelated metabolites, we applied sparse partial least squares discriminant analysis (sPLS-DA) restricted to amino acid–related KEGG pathways. This targeted analysis revealed distinct sex-and diet-dependent metabolic profiles ( Extended Data Fig. 3d), with separation driven by nitrogen metabolism, oxidative stress markers, and amino acid–urea cycle intermediates ( Extended Data Fig. 3e). To assess transcriptional coordination of these metabolic shifts, we performed liver bulk RNA-sequencing. To increase sensitivity and biological relevance, we used Gene Set Enrichment Analysis (GSEA) with a curated set of genes known to respond to protein restriction in C57BL/6 mice 5 . In males, amino acid biosynthesis and nitrogen recycling pathways were enriched ( Fig. 3e ), whereas these responses were absent in females. Global transcript analysis confirmed stronger induction of amino acid stress response genes in males, including Asns and Fgf21 ( Fig. 3f , Extended Data Table 2 ). Sparse PLS-DA of amino acid–related transcripts further supported this sex-specific pattern, revealing clear separation of male diet groups and strong contributions from genes involved in amino acid catabolism and nitrogen disposal ( Extended Data Fig. 3f–g). Although BCAAs are not taken up by the liver 13 , they are transported via LAT3 ( Slc43a1 ), a low-affinity system L transporter that mediates facilitated diffusion and is typically dominant in adult hepatocytes. In males, Slc43a1 was significantly downregulated under protein restriction, while Slc7a5 (LAT1), a high-affinity antiporter normally restricted to fetal liver or proliferative cells, was upregulated ( Fig. 3g , Extended Data Table 2 ). This male-specific switch in transporter expression suggests a shift from constitutive BCAA efflux to a stress-adaptive mode that may promote hepatic uptake or intracellular redistribution of BCAAs or their ketoacid derivatives. This was accompanied by modest changes in catabolic gene expression, consistent with a strategy to conserve systemic BCAAs by limiting their hepatic breakdown. In contrast, females exhibited minimal transcriptional changes in both transporters and catabolic enzymes, suggesting a more buffered hepatic response to protein restriction. Prior studies show that essential amino acid restriction, particularly of leucine, activates the integrated stress response (ISR) through the GCN2 sensor, initiating ATF4-dependent transcription 14 , 15 . Gene expression analysis showed that males strongly increased the expression of several ATF4 targets such as Asns ( Fig. 3h ). This activation persisted across development, as confirmed by Asns qPCR at P28, P42, and P70 ( Extended Data Fig. 3h). In parallel, the nutrient-responsive transcription factor PPARα 16 was also activated. Males showed robust induction of canonical PPARα targets including Cyp4a14 and Acot3 ( Fig. 3i , Extended Data Fig. 3i–j), consistent with transcriptional engagement of nutrient stress pathways. These data suggest that males mount a coordinated hepatic stress response under protein restriction, coupling ISR and PPARα pathways, while females only have a modest activation of these pathways. These pathways converge on Fibroblast Growth Factor 21 ( Fgf21 ), a key hepatic hormone involved in nutrient sensing 16 – 18 . Consistent with their transcriptional activation, liver Fgf21 expression increased markedly under LPD, and was overall higher in males ( Fig. 3i–j ). Circulating FGF21 levels rose sharply in both sexes, reaching levels that surpass those induced by prolonged fasting 19 , 20 , with males consistently higher ( Fig. 3k ). Together, these data establish liver-derived FGF21 as an output of the hepatic response to protein restriction, positioned downstream of ATF4 and PPARα signalling, with sex-specific differences in magnitude ( Extended Data Fig. 3k). However, the physiological role of FGF21 remains complex and context-dependent. While global deletion of Fgf21 prevents growth attenuation during chronic undernutrition, implicating FGF21 in somatotropic suppression through growth hormone (GH) resistance 21 , other studies demonstrate that adipose-derived FGF21 can extend lifespan and improve metabolic health in adults without affecting growth 22 . Although transgenic or pharmacological models induce elevated FGF21, the similar concentrations observed under LPD arose via physiological regulation, emphasizing the role of nutritional context in modulating FGF21 signaling. In the gonadotropic axis, earlier studies show that excess FGF21 can delay female puberty by suppressing hypothalamic–pituitary–gonadal signalling 23 , though more recent evidence questions its necessity under physiological conditions 24 . Together, these findings underscore the nuanced, tissue- and context-specific roles of FGF21 in coordinating metabolic and developmental responses. This raises the question of whether endogenous, diet-induced FGF21 contributes causally to the sex-specific trade-offs between growth and reproductive timing observed under juvenile protein scarcity. To test whether hepatic FGF21 contributes to the compromise between growth and reproduction under protein scarcity, we used hepatocyte-specific Fgf21 knockout mice ( Fgf21 HepKO ) on a C57BL/6J (B6J) background, as described by Fougerat et al. 20 . We first validated that wild-type B6J mice replicate the key phenotypes observed in our original B6N dataset. Under low-protein diet (LPD), B6J mice showed marked post-weaning growth stunting, more severe than in B6N, with males disproportionately affected (−30% in males vs. −15% in females by P56; Extended Data Fig. 4a–b). In both strains, LPD delayed reproductive maturation in females but not in males ( Extended Data Fig. 4c–d). Serum FGF21 was strongly induced under LPD ( P = 0.004; Extended Data Fig. 4e). Serum FGF21 was nearly undetectable in Fgf21 HepKO mice regardless of diet ( Extended Data Fig. 4f–g), confirming the liver as its primary source. Expression of ATF4 and PPARα targets remained intact ( Extended Data Fig. 4h–j), indicating preserved nutrient-sensing upstream of FGF21. We first examined the contribution of hepatic FGF21 to somatic growth. In males, LPD impaired growth across genotypes ( Fig. 4a–b , Extended Data Fig. 4k), with no effect of Fgf21 deletion. Femur length and hepatic Igf1 expression were also similar between Fgf21 HepKO and floxed ( Fgf21 Hepfl/fl ) males ( Fig. 4c , Extended Data Fig. 4m), suggesting that hepatic FGF21 does not modulate growth in males under protein restriction. In contrast, Fgf21 HepKO females displayed increased growth from P42 onward ( Fig. 4d–e , Extended Data Fig. 4l), with longer femurs at P56 ( Fig. 4f ). This growth improvement was not accompanied by changes in hepatic Igf1 expression ( Extended Data Fig. 4m), consistent with the idea that protein-restricted growth is regulated beyond liver Igf1 transcription ( Fig. 1k ). These findings indicate that hepatic FGF21 contributes to early-phase growth suppression in females but is dispensable for growth regulation in males. Download figure Open in new tab Figure 4 Sex- and diet-specific role of hepatic FGF21 in juvenile adaptation to protein restriction. a–f , Postweaning growth indicators in males ( a–c ) and females ( d–f ) fed control (CD) or low-protein diet (LPD) from P21 to P56. Size curves in males ( a ) and females ( d ) from P21 to P56. Effect size (Cohen’s d ) of Fgf21 HepKO vs Fgf21 Hepfl/fl on body length in males ( b ) and females ( e ) over time. Dashed lines represent 95% confidence intervals. Femur length of males ( c ) and females ( f ) at P56. g , Cumulative onset of balano-preputial separation (BPS) in males from P21 to P56. h, Sperm count at P56 in adult males. i, Seminal vesicle weight at P56 plotted against body weight. Dashed line represents the fitted ANCOVA model without interaction (main effects only). j, Testis weight at P56. k, Cumulative onset of vaginal opening (VO) in females from P21 to P56. l, Cumulative onset of first oestrus in females from P21 to P56. m, Uterus weight at P56. Data represent mean□±□sd. Symbols indicate post hoc significance of diet or genotype effect, as highlighted in the figure (* P □<□0.05, ** P □<□0.01, *** P □<□0.001), ns , not significant. Shaded areas ( g, k, l ) represent 95% confidence intervals from Kaplan-Meier survival models. “+” indicates right-censored observations. The number of mice used in each panel is provided in Extended Data Table 2 . We then asked whether hepatic FGF21 modulates reproductive timing under LPD. In males, Cox regression revealed a significant diet effect on BPS but no effect of genotype ( Fig. 4g ). This slight delay, absent in wild-type B6N and B6J males, occurred equally in Fgf21 HepKO and Fgf21 Hepfl/fl littermates, indicating it is independent of FGF21. Logistic regression estimated that LPD modestly increased the age threshold for BPS regardless of genotype ( Extended Data Fig. 4n–p). Sperm count, seminal vesicle weight, and testis size were also unaffected by genotype ( Fig. 4h–j ). In females, diet strongly delayed vaginal opening (P = 0.0001; Fig. 4k ). Under LPD, Fgf21 HepKO females reached VO earlier than Fgf21 Hepfl/fl controls. Although the Cox model suggests they were, on average, 3.7 times more likely to reach puberty at any given time, the confidence interval is extremely wide (95% CI: 0.07–188), encompassing both negligible and very large differences ( P = 0.16) ( Fig. 4k ). However, logistic regression confirmed that the age threshold for VO decreased from 43.5 to 39.1 days in Fgf21 HepKO females under LPD ( Extended Data Fig. 4q–s). Under LPD, none of the Fgf21 Hepfl/fl females reached oestrus by P56, while several knockout littermates did ( Fig. 4l ). Time-to-event analysis confirmed that knockout females were 5 times more likely to reach oestrus before P56 compared to floxed littermates (95% CI: 1.7–18.2; P = 0.005). Uterine weight was also partially restored under LPD in Fgf21 HepKO mice ( Fig. 4m ), consistent with partial recovery of reproductive function. Together, these findings identify a female-specific endocrine checkpoint participating in juvenile adaptation to protein restriction. While FGF21 is dispensable in males, it fosters adaptation by limiting growth and delaying reproductive maturation in females. These effects originate from hepatocyte-derived FGF21 and are not solely attributable to differences in somatic size. Thus, FGF21 extends its role as a metabolic regulator to mediate sex-specific physiological outcomes in early life. This sex difference supports the idea that female puberty is an energy-sensitive checkpoint integrating peripheral metabolic cues with central neuroendocrine circuits 25 . Although Fgf21 overexpression in transgenic models can delay female puberty by inhibiting hypothalamic–pituitary–gonadal signalling 23 , the relevance of this phenotype under physiological levels of FGF21 remains unclear 24 . Our loss-of-function data show that liver-derived FGF21 actively regulates female reproductive timing under protein restriction. The strong sexual dimorphism in hypothalamic kisspeptin neurons 26 , 27 offers a plausible substrate for how a shared hepatic stress signal yields such sex-specific outcomes. Whether FGF21 acts centrally or modulates puberty indirectly by limiting nutrient availability remains unclear. In males, Fgf21 deletion did not influence growth or advance puberty, suggesting that somatic investment is constrained by intrinsic nutritional limits rather than reproductive trade-offs. Given the low energetic cost of male reproduction, diverting these resources is unlikely to rescue growth under protein restriction. More broadly, the robust male induction of FGF21 despite its limited role raises the possibility that some sex-biased gene regulation reflects shared stress responses or unresolved intralocus conflict 28 rather than adaptive tuning. Our model specifically tested sustained protein dilution as a tractable nutritional stressor without manipulating other macronutrients or caloric intake. Nutritional geometry studies have shown that protein-to-carbohydrate ratio is a dominant determinant of reproductive timing and lifespan in mice 4 . Whether the FGF21-mediated mechanism observed here generalizes to other nutritional contexts, including fat or sugar restriction, remains to be determined. Recent studies implicating FGF21 in protein appetite, sweet preference, and energy balance 29 , 30 support its broader role as a macronutrient-sensitive endocrine integrator. Our findings identify liver-derived FGF21 as a sex-specific endocrine mediator of juvenile adaptation to protein malnutrition, revealing how a shared metabolic signal can differentially regulate growth and reproductive timing in line with sex-dependent life-history priorities. METHODS Animals All animal experiments have been approved by a local animal care and use committee (CECCAPP) and subsequently authorized by the French Ministry of Research (APAFIS #38677-2022040517218744 v6 and APAFIS #44718-2023051708064873 v7). All experiments were carried out in accordance with the European Community Council Directive of September 22 nd , 2010 (2010/63/EU) regarding the protection of animals used for experimental and other scientific purposes. Animal experiments were carried out at IGFL, Lyon, France. 21-day-old C57BL/6N or C57BL/6J conventional mice or 14-day-old C57BL/6N mice with their mothers were purchased from Charles River (France). Fgf21 hepatocyte-specific knockouts ( Fgf21 HepKO ) were generated as previously described 20 . Mice were kept in The Innorack® IVC Mouse 3.5 disposable cages (Inovive, USA), exposed to 12:12 hours light-dark cycles (lights on: 8 AM; lights off: 8 PM), supplied with tap water and fed ad libitum . Experimental design At postnatal day 21 (P21), male and female mice weighing more than 8 g were weaned onto either standard AIN93G control diet (CD) or an isocaloric low-protein variant (LPD). Full diet compositions are listed in Extended Data Table 1 . Except for Experiment 4, animals were weighed and measured weekly from P21 to P56 under brief isoflurane anesthesia (Isoflurin, VetCare). Food intake was assessed weekly per cage by subtracting uneaten pellets from the initial quantity provided. Unless otherwise stated, mice were maintained on CD or LPD ad libitum throughout this period, and tissue and serum were collected at P56. Full sample sizes and raw data are provided in Extended Data Table 3 . Experiment 1: C57BL/6N mice At P36, mice were transferred to the ANIPHY Platform (Lyon) and acclimated for 3 days. Body composition was measured at P38 by non-invasive low-field nuclear magnetic resonance (Minispec LF90 II, Bruker, France). Mice were returned to IGFL at P42. At P56, animals were fasted for 6 h starting at 8 AM before receiving a subcutaneous injection of buprenorphine (0.1 mg/kg, Buprecare 0.3 mg/mL, Axience) 30 minutes prior to sampling, followed by oral gavage of glucose (1 g/kg, D-(+)-glucose, Sigma G7021) 15 minutes before terminal anesthesia (ketamine 100 mg/kg i.p., Imalgene 1000, Merial; plus subcutaneous lidocaine, Lurocaine, Vetoquinol). A midline laparotomy was performed for portal vein blood collection. Blood was transferred to Z-gel microtubes (Sarstedt, 41.1500.005), kept on ice, centrifuged at 10,000 g for 5 min at 4 °C, and serum was stored at –80 °C. Femur and tibia were dissected and measured with calipers. Experiment 2: C57BL/6N mice At P56, animals were fasted from 8:00 AM and killed by cervical dislocation between 2:00 and 4:00 PM. Blood was collected via submandibular vein and processed as described above. Experiment 3: C57BL/6N mice At P35, P42, and P49, mice were fasted for 6 h (starting at 8:00 AM), and blood was collected from the tail vein using CB 300 Z Microvette® tubes (Sarstedt, 16.440). Samples were centrifuged at 10,000 g for 5 min at 4 °C, and serum was stored at – 20 °C. At P56, portal and caval blood were collected between 8:00 and 10:00 AM without prior fasting, followed by cervical dislocation. Experiment 4: C57BL/6N mice Mice were killed by cervical dislocation at P28, P42, or P70 between 8:00 and 10:00 AM without prior fasting. Experiment 5: C57BL/6J mice Mice were weighed weekly (no body length measurements). At P56, they were killed by cervical dislocation without fasting. Blood was collected from the cava vein, as described above. Experiment 6: Fgf21 HepKO mice Alb-Cre +/- Fgf21 fl/fl ( Fgf21 HepKO ) mice were crossed with Alb-Cre -/- Fgf21 fl/fl (control Fgf21 Hepfl/fl ) animals to generate littermates. Mice were weaned at P21 onto CD or LPD as described above and housed by sex and diet to minimize cage effects. At P56, animals were killed by cervical dislocation between 8:00 and 10:00 AM without fasting and blood was collected from the cava vein. Assessment of reproductive development Females: From weaning onward, mice were examined daily to monitor reproductive maturation. Puberty onset was defined by the day of vaginal opening. Following this, oestrous cycle stage was assessed daily via vaginal cytology, as previously described 31 . Males: As in females, daily monitoring began at weaning. Puberty onset was assessed by balano-preputial separation (BPS) 32 , 33 . Testis and seminal vesicle weights were used as indicators of testicular growth and androgenic status 34 – 38 . Epididymal sperm count was performed as previously described 39 . Briefly, whole epididymides, collected at dissection and stored at −80 °C, were thawed, weighed, finely minced in 1 mL PBS, and vigorously shaken. The sperm suspension was examined under brightfield microscopy, and sperm count was determined by counting intact sperm heads. Measurement of circulating blood hormones Circulating hormones were quantified using enzyme-linked immunosorbent assays (ELISAs) following the manufacturers’ protocols. IGF-1 levels were measured with the Mouse/Rat IGF-1 Quantikine ELISA Kit (R&D Systems, #MG100), and FGF21 levels with the Mouse/Rat FGF21 ELISA Kit (BioVendor, #RD291108200R). RT-qPCR Total RNA was extracted and purified from liver samples using the NucleoSpin RNA kit (Macherey-Nagel, #740955.250). RNA concentration was measured with a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific). Reverse transcription was performed on 1 µg of total RNA using the SensiFAST cDNA Synthesis Kit (Meridian Biosciences, #BIO-65053). Quantitative PCR was then carried out using Takyon™ No ROX SYBR MasterMix blue dTTP (Eurogentec, #UF-NSMT-B0701) on a Bio-Rad CFX96 real-time PCR system. Gene expression analysis followed the MIQE guidelines 40 . Relative expression levels were normalized using two validated reference genes ( Tbp and Actb ). Gene stability was assessed using the geNorm module integrated in the CFX96 software. Primer sequences for all control and target genes are listed in Extended Data Table 4 . Western blotting Mouse urine samples were collected and stored at –20 °C until analysis. For each sample, 3 µL of urine was mixed with 8.1 µL of denaturing buffer (containing 20% glycerol, 10% β-mercaptoethanol, 10% SDS, and 62.5 mM Tris), 18.9 µL of water, and denatured at 95 °C for 10 minutes. A total of 10 µL of each sample was loaded onto 4–20% Mini-PROTEAN® TGX™ precast protein gels (Bio-Rad) and electrophoresed at 200 V for 30 minutes. Proteins were transferred to nitrocellulose membranes using the Trans-Blot® Turbo™ Transfer System (Bio-Rad), and total protein content was verified. Membranes were blocked for 1 hour in TBS-T (1× TBS + 0.1% Tween 20) containing 5% milk, then incubated for 1 hour with a 1:500 dilution of MUP (C-7) mouse monoclonal IgG1 antibody (Santa Cruz Biotechnology, #sc-166429). After three washes with TBS-T, membranes were incubated for 20 minutes with a 1:2,500 dilution of StarBright Blue 700 Goat Anti-Mouse IgG secondary antibody (Bio-Rad, #12004159), followed by three washes in PBS-T. Blots were imaged at 660 nm using the Amersham ImageQuant™ 800 system (Cytiva), and band quantification was performed with Image Lab software (Bio-Rad). The uncropped Western blot image is provided in Extended Data Figure 5. Amino acid analysis and quantification Amino acids were quantified from serum samples collected in Experiments 1 and 3 using high-performance liquid chromatography (HPLC) on an Agilent 1100 system (Agilent Technologies, Massy, France), equipped with a guard cartridge and a reverse-phase C18 column (Zorbax Eclipse-AAA, 3.5 µm, 150 × 4.6 mm). The method was specifically developed for this setup. Samples were buffered with borate at pH 10.2 prior to injection, and amino acids were derivatized at room temperature using ortho-phthalaldehyde (OPA) for primary amines and 9-fluorenylmethyl chloroformate (FMOC) for secondary amines. Derivatization was automated using the Agilent 1313A autosampler. Chromatographic separation was performed at 40 °C with a 2 mL/min flow rate, using 40 mM NaH ₂ PO ₄ (eluent A, pH 7.8, adjusted with NaOH) and a non-polar phase consisting of an acetonitrile/methanol/water mixture (45:45:10, v/v/v; eluent B). A gradient was applied, increasing from 20% to 80% B. Detection was performed by fluorescence with excitation/emission wavelengths of 340/450 nm (or 266/305 nm for proline). Due to derivatization limitations, cystine was not detected, and 19 amino acids were quantified. Norvaline was used as an internal standard, and response factors were determined using a 250 pmol/µL amino acid standard mix. Data acquisition and analysis were performed using ChemStation for LC 3D Systems (Agilent Technologies). Liver RNA-Seq analysis The concentration and integrity of DNase-treated liver RNA samples (n = 20) were assessed using a Qubit 4.0 fluorometer (Thermo Fisher Scientific) and a TapeStation 4150 system (Agilent Technologies), respectively. mRNA was purified from 1 µg of total RNA per sample using the Poly(A) mRNA Magnetic Isolation Module (New England Biolabs), following the manufacturer’s protocol. RNA-seq libraries were prepared using the CORALL Total RNA-Seq Library Prep Kit (Lexogen). This fragmentation-free protocol ensures uniform transcript coverage and incorporates 12 bp Unique Molecular Identifiers (UMIs) during linker ligation to reduce amplification bias and improve quantification accuracy. Following Qubit and TapeStation quality control, libraries were pooled equimolarly and sequenced on two high-output runs using a NextSeq500 system (Illumina) in single-end mode (1 × 86 bp). Sequencing yielded 1.08 billion total reads, with 31–44 million reads per sample (mean: 36 million). Read processing followed a Lexogen-derived pipeline. Adapters were trimmed with Cutadapt (v3.5) 41 , and read quality was assessed using FastQC. Reads were mapped to the Mus musculus reference genome (GRCm39, Ensembl release 110) using STAR (v2.7.11a) 42 . UMI-based deduplication was performed using UMI-tools dedup (v1.1.2) 43 to eliminate PCR duplicates. Gene-level quantification was conducted with htseq-count (v2.0.3) 44 , producing count tables for downstream analysis. Liver transcriptome analysis Raw read counts for each sample were obtained by merging gene-level output files from featureCounts. Count matrices were constructed using Ensembl gene identifiers and matched to corresponding gene symbols using the biomaRt R package (release 110, GRCm39). Genes with low counts (fewer than 5 reads in fewer than 3 samples) were excluded. Differential expression analysis was conducted using DESeq2 with a linear model including the main effects of Sex, Diet, and their interaction (Sex × Diet). Wald tests were performed for the main effect of Diet (LPD vs CD), for Sex, and for the interaction term. Stratified models were used to compute diet-specific effects in males and females. Differentially expressed genes were defined as those with |log ₂ FC| ≥ 1 and Benjamini-Hochberg adjusted P < 0.05. MA plots were generated for each contrast. Normalized counts and full differential expression results, including main effects and sex-stratified contrasts, are provided in Extended Data Table 4 . Pathway enrichment and focused gene filtering Gene-level KEGG pathway enrichment was performed using clusterProfiler (v4.10.1) based on Entrez Gene IDs for differentially expressed genes in males and females, using all expressed genes with valid Entrez mappings as background. For focused mechanistic analysis, we intersected our DE gene lists with amino acid metabolism genes curated by Green et al. (2022) 5 , specifically those showing significant and robust transcriptional responses to protein dilution in C57BL/6 mice (adjusted P 1). Enriched KEGG pathways from these filtered sets, such as “ Biosynthesis of amino acids” and “ Arginine biosynthesis” , were visualized using dot plots. Volcano plots highlighting selected amino acid–related transcripts were generated with globally aligned axes to enable direct comparison between sexes. To further explore sex-specific hepatic adaptations, we visualized normalized expression profiles for three curated gene sets: (1) enzymes and transporters of the BCAA catabolic pathway, (2) ATF4 target genes involved in amino acid stress response, and (3) PPARα-regulated genes involved in lipid metabolism and peroxisomal signaling. Expression values were extracted from normalized DESeq2 outputs, z-score transformed across samples, and displayed as heatmaps with group annotations by sex and diet. Targeted liver metabolomics of polar metabolites Liver polar metabolites were extracted from ∼30 mg of frozen tissue using 80% methanol followed by sonication and homogenization in Precellys Cryolys tubes. Supernatants were clarified by centrifugation (15,000 rpm, 15 min, 4°C), evaporated to dryness, and reconstituted in methanol according to total protein content (measured on pellets by BCA assay). Targeted metabolite profiling was performed using ultra-high performance liquid chromatography coupled to tandem mass spectrometry (UHPLC-MS/MS) in multiple reaction monitoring (MRM) mode on a 6495 iFunnel Triple Quadrupole mass spectrometer (Agilent). Two complementary chromatographic modes with both positive and negative electrospray ionization were used to maximize detection of central carbon metabolites, including amino acids, TCA cycle intermediates, purine/pyrimidine species, and glycolytic derivatives, enabling relative quantification of over 400 polar compounds 45 – 47 . Data were processed using Agilent MassHunter (version B.07.00); extracted ion chromatogram (EIC) areas were normalized and filtered to exclude features with >30% coefficient of variation across QC samples. Metabolomics data curation and normalization Raw metabolite peak areas were filtered to exclude redundant features detected in both positive and negative ionization modes. Redundancy was identified by matching column names differing only by a numeric suffix (e.g., ophthalmic.acid vs ophthalmic.acid.1 ), and in such cases, the ionization mode with the highest integrated peak area was retained. This filtering reduced the dataset from 189 to 164 unique metabolites. Compound names were harmonized to match the MetaboAnalyst 6.0 database ( https://www.metaboanalyst.ca ) using the online compound name matcher, with manual curation of ambiguous entries based on empirical formulae (e.g., UDP-hexose inferred to be UDP-glucose ). Normalization was performed using MetaboAnalyst 6.0, applying probabilistic quotient normalization (PQN) using male samples as the reference group, followed by log 10 transformation and autoscaling (mean-centering and division by standard deviation). Normalized counts are provided in Extended Data Table 5 . Liver amino acid quantification and statistical analysis We extracted liver concentrations of 19 amino acids from normalized targeted metabolomics data (excluding cysteine, which was not reliably detected). Sex- and diet-specific differences were assessed using linear models with an interaction term (AminoAcid ∼ Diet × Sex), followed by estimated marginal means (EMMs) and pairwise contrasts (Tukey correction, Benjamini-Hochberg FDR). Amino acids were annotated as essential or non-essential, and values were displayed in a heatmap using fixed axis scaling across all groups to enable direct comparison. Sparse Partial Least Squares Discriminant Analysis (sPLS-DA) To assess sex- and diet-dependent variation in transcriptomic and metabolomic profiles related to amino acid metabolism, we performed sPLS-DA using the mixOmics R package. Separate models were fit for transcriptomic and metabolite datasets filtered to KEGG-defined amino acid metabolism pathways. Optimal numbers of components and selected variables were determined via 5-fold cross-validation with repeated tuning (10 iterations, Balanced Error Rate). Resulting models retained the top discriminatory features across experimental groups (M-CD, M-LPD, F-CD, F-LPD). Projections onto the first two components were visualized with 95% confidence ellipses, and variance explained was reported per axis. Loadings were thresholded (|loading| > 0.1) to identify top contributing genes and metabolites for Components 1 and 2 ( Extended Data Fig. 3). Ensembl gene identifiers were mapped to gene symbols using the org.Mm.eg.db package. Final plots were generated in ggplot2 using a fixed axis scale and a consistent group color scheme. Statistical analyses Longitudinal and Growth Analyses Body weight and length trajectories were modelled using linear mixed-effects models (LMMs), incorporating cubic splines (bs(Age, df = 4)) to capture non-linear age-dependent growth. Fixed effects included Sex, Diet, and Age, with random intercepts for Mouse_ID and Cohort. Individual growth rates were computed as the first derivative of the spline-fitted weight curve per mouse. Feed efficiency (weight gain per kilocalorie) and macronutrient intake were modelled using LMMs with fixed effects of Sex, Diet, and developmental Interval, with mean body weight included as a spline-adjusted covariate (bs(mean_BW, df = 4)). To reflect the hierarchical design, we included random intercepts for Cage nested within Protocol. To dissect macronutrient-specific effects, energy intake was partitioned into protein, carbohydrate, and fat kilocalories, scaled to reflect each macronutrient’s relative contribution (%) to total energy during each interval. Separate models were fit for total energy and for each macronutrient. Growth-Associated and Reproductive Phenotypes Femur weights were analysed using two- or three-way ANOVA. Seminal vesicle weights were analyzed using ANCOVA with body weight as a covariate. As interaction terms for Diet × Body Weight and Diet × Genotype were not significant ( P > 0.05), we adopted a simplified additive model without interactions, which provided a good fit to the data. In contrast, uterus and testis weights showed poor model fit in ANCOVA (both raw and log-transformed), with low correlation to body weight (Spearman’s ρ and Kendall’s τ < 0.3; P < 0.05); these were therefore analysed using factorial ANOVA on raw values. Sperm count was log 10 -transformed prior to analysis. Comparisons across groups used two-way ANOVA or Welch’s t-tests based on variance homogeneity. Serum IGF-1 and FGF21 concentrations were modelled using LMMs with fixed effects of Sex, Diet, and Age, and random intercepts for Mouse_ID and Cohort. FGF21 was log-transformed to correct for right-skewed residuals. In datasets with left-censored values (i.e., below the assay detection limit of 7 pg/mL), log-transformed FGF21 values were analysed using accelerated failure time (AFT) models to account for censoring. Liver gene expression (RT-qPCR) data were log 2 -transformed prior to analysis. Two- or three-way ANOVA models were used to assess the effects of Sex, Diet, and Genotype. When residual heteroscedasticity was detected, permutation-based ANOVA ( permuco ) was applied. Puberty Timing and Logistic Models Pubertal onset (VO and BPS) was analysed using time-to-event methods. When a single categorical variable (Diet) was assessed, Kaplan–Meier survival curves and log-rank tests ( survdiff ) were used for group comparisons. For factorial designs involving multiple covariates (Diet × Genotype), Cox proportional hazards models ( coxph ) were applied. In cases with complete or near-complete right-censoring (e.g., VO in LPD Fgf21 Hepfl/fl females), Firth’s penalized Cox regression ( coxphf ) was used to correct for bias. Receiver Operating Characteristic (ROC) curves and Area Under the Curve (AUC) values were computed using the pROC package to assess model discrimination. The best model was defined as that with the highest AUC. Likelihood ratio tests were used to test whether adding Diet or Genotype significantly improved model fit. To assess the relationship between growth and pubertal onset, we first used segmented linear regression ( segmented ) to identify an inflection point in the association between Age at VO and inverse growth rate (1/Growth Rate). This breakpoint (0.48 g/day) was then used to stratify mice into high- and low-growth groups. Reciprocal models (Age at VO ∼ 1 / Growth Rate) were fitted using non-linear least squares ( nls ), and 95% confidence intervals were obtained by bootstrap resampling (5,000 iterations). To assess growth-independent heterogeneity in timing, model residuals were compared between growth strata using a two-sided permutation test (10,000 iterations) on absolute residuals. Serum amino acid profiling Serum amino acid concentrations were analysed using linear mixed-effects models (LMMs) with fixed effects of Diet, Sex, and Compartment (cava vs. porta), and a random intercept for Mouse ID to account for repeated measures. Raw concentrations were log-transformed to correct right-skewed distributions and improve model fit. For heatmap visualization, values were Z-scored within each amino acid to facilitate relative comparison across groups. General Notes All analyses were performed in R version 4.4.3 using core and contributed packages, including tidyverse and emmeans . Data were assessed for normality and homoscedasticity using visual diagnostics and residual simulation ( DHARMa ). When assumptions were violated, non-parametric permutation tests ( permuco , 5,000 iterations) were applied. Statistical significance was defined at α = 0.05, with false discovery rate (FDR) correction applied for multiple comparisons where appropriate. Graphical outputs were generated using ggplot2 (v3.5.2), ggprism (v1.0.6), or GraphPad Prism v10.5.0 to ensure consistency across plots. A summary of the models used is available in Extended Data Table 6 . The authors declare no conflict of interest related to this work. USE OF AI The authors used OpenAI’s ChatGPT (model GPT-4o) to assist with improving English phrasing and grammar during manuscript preparation, and to refine and optimize R scripts used for data analysis and figure generation. All scientific content, interpretation, and conclusions were developed and verified by the authors. AUTHOR CONTRIBUTIONS AJ, FDV and FL conceived and designed the research. AJ, LR, AL and EC performed mouse experiments. AJ, LR, EC also performed bench experiments. BG and SH performed liver RNA sequencing experiments and generated the count tables. IR and PDS performed mass spectrometry analyses of circulating amino acids. JB and AF generated the Fgf21 fl/fl and Alb-Cre +/- mouse colony. JB assisted with Fgf21 -related data analysis. AF also assisted with transcriptomic pipeline development. AJ and FDV analysed data. AJ, LR, FDV and FL interpreted results. AJ and FDV prepared figures and drafted the manuscript. AJ, FDV and FL edited and revised the manuscript. Download figure Open in new tab Extended Data Figure 1 Sex-specific effects of early-life protein restriction on macronutrient intake, feed efficiency, and body composition. (Related to Figure 1 ) a–c , Cumulative protein ( a ), lipid ( b ) and carbohydrate ( c ) intake per mouse (kcal) in males (top) and females (bottom), modelled using body weight as a covariate, over successive 7-day intervals. d–e , Lipid- ( d ) and carbohydrate- ( e ) specific feed efficiency (weight gain per macronutrient-derived kcal) over successive 7-day intervals. f–i , Relationship between log-transformed body weight and log-transformed lean mass ( f , males; g , females) or fat mass ( h , males; i , females) at P42. Unless otherwise stated, data represent mean ± sd. Symbols indicate post hoc significance of diet effect (* P □<□0.05, ** P □<□0.01, *** P □<□0.001), ns , not significant. The number of mice used in each panel is provided in Extended Data Table 2 . Download figure Open in new tab Extended Data Figure 2 Markers of sexual maturation and modelling of pubertal onset. (Related to Figure 2 ) a, Western blot of major urinary proteins (MUPs) in urine at P56. b , Representative oestrus cycles of six CD-fed and six LPD-fed females, from P56 to P70. c–h, Logistic regression modelling of puberty in males ( c–e ) and females ( f–h ). Receiver operating characteristic (ROC) curves for logistic models predicting balano-preputial separation (BPS) ( c ) or vaginal opening (VO) ( f ) based on age, diet or combined predictors including body weight. AUC values are shown for each model. P-values indicate likelihood ratio tests comparing nested models with and without body weight. Logistic regression fits predicting BPS ( d–e ) or VO ( g–h ) as a function of age ( d,g ) or body weight ( e,h ). Thresholds represent a predicted probability of 50%. Shaded areas represent 95% confidence intervals of the logistic models. i–j, Relationship between age at VO and prepubertal growth rate in females. A reciprocal model (Age ∼ 1 / Growth Rate) was fitted ( i ), and residuals plotted against Growth Rate ( j ). The shaded blue region denotes values below a threshold derived from segmented regression. Residual spread was compared above vs below this threshold using a permutation test ( P < 0.0001). The number of mice used in each panel is provided in Extended Data Table 2 . Download figure Open in new tab Extended Data Figure 3 Sex-specific amino acid responses to protein restriction. (Related to Figure 3 ) a, Total amino acid concentrations in portal and cava veins at P56, in postabsorptive state. b–c, Heatmap of circulating amino acid concentrations at P56, stratified by sex and diet, in postprandial ( b ) and postabsorptive ( c ) states. d–g, Sparse partial least squares discriminant analysis (sPLS-DA) projections of metabolomic ( d ) and transcriptomic ( f ) profiles restricted to amino acid metabolism pathways. Ellipses indicate 95% confidence intervals. Top metabolites ( e ) and genes ( g ) contributing to Component 1 and 2 of the sPLS-DA, based on loading values exceeding a threshold of |0.1|. h–i , Hepatic expression (log 2 relative to reference genes) of Asns ( h ), Cyp4a14 ( i ) and Acot3 ( j ) in males and females at P28, P42 and P70. k, Putative hepatic mechanism leading from amino acid scarcity to FGF21 secretion in males and females. Created in BioRender. Joly, A. (2025) https://BioRender.com/o3k6389 Unless otherwise stated, data are mean ± sd; symbols indicate post hoc significance of diet effect (* P □<□0.05, ** P □<□0.01, *** P □<□0.001), ns , not significant. The number of mice used in each panel is provided in Extended Data Table 2 . Download figure Open in new tab Extended Data Figure 4 Additional phenotyping in Fgf21 Hepfl/fl and Fgf21 HepKO mice. (Related to Figure 4 ) a–b , Postweaning weight curves in male ( a ) and female ( b ) C57BL/6J mice fed control (CD) or low-protein (LPD) diet. c–d , Cumulative pubertal onset in C57BL/6J males ( c , BPS) and females ( d , VO) from P21 to P56. e , Serum FGF21 concentrations in C57BL/6J mice at P56. f–g , Serum FGF21 concentrations in male ( f ) and female ( g ) Fgf21 Hepfl/fl and Fgf21 HepKO mice at P56. h–j , Hepatic expression (log 2 relative to reference genes) of nutrient-responsive genes Asns ( h ), Acot3 ( i ), and Cyp4a14 ( j ) Fgf21 Hepfl/fl and Fgf21 HepKO mice at P56, analysed by RT-qPCR. k–l , Postweaning weight curves in male ( k ) and female ( l ) Fgf21 Hepfl/fl and Fgf21 HepKO mice. m , Hepatic Igf1 expression (log 2 relative to reference genes) in Fgf21 Hepfl/fl and Fgf21 HepKO males and females at P56. n–s, Logistic regression modelling of puberty in males ( n–p ) and females ( q–s ). Receiver operating characteristic (ROC) curves for logistic models predicting balano-preputial separation (BPS) ( n ) or vaginal opening (VO) ( q ) based on age, diet or combined predictors including body weight and genotype. AUC values are shown for each model. P-values indicate likelihood ratio tests comparing nested models with and without body weight or body weight + genotype. Logistic regression fits predicting BPS ( o–p ) or VO ( r–s ) as a function of age ( o,r ) or body weight ( p,s ). Thresholds represent a predicted probability of 50%. Shaded areas represent 95% confidence intervals of the logistic models. Unless otherwise stated, data represent mean□±□sd. Symbols indicate post hoc significance effect (* P □<□0.05, ** P □<□0.01, *** P □<□0.001), ns , not significant. The number of mice used in each panel is provided in Extended Data Table 2 . Download figure Open in new tab ACKNOWLEDGEMENTS This work was supported by funding from Fondation pour la Recherche Médicale (Équipe FRM EQU202203014629), and the Agence Nationale de la Recherche (grant ANR PRC 2023 CE14 “GutStunting”). AJ was supported by a grant from Fondation pour la Recherche Médicale (FDT202304016501). AJ and LR were supported by a doctoral grant from the French Ministry of Research. We thank Jean-Louis Thoumas for assistance with mouse handling and strain management. We are grateful to Raghav Galgali (Master’s intern, IISER Bhopal) for his contributions to the liver transcriptome analysis and sperm count assays. We also acknowledge Manuela Forero (Master’s intern, ENS de Lyon) for her work on the Fgf21 knockout mouse experiments. We thank Daniela Fernandois and Marialetizia Rastelli (Lille Neuroendocrinology, Université de Lille) for their help in setting up the puberty-related assays and their contributions to the interpretation of the data. We thank Julijana Ivanisevic, Hector Gallart Ayala, and Tony Teav (Metabolomics Unit, University of Lausanne) for performing targeted liver metabolomics and for helpful discussions regarding the analysis and interpretation of metabolomics data. Funder Information Declared Agence Nationale de la Recherche , ANR PRC 2023 CE14 GutStunting Fondation pour la Recherche Médicale, https://ror.org/04w6kn183 , Équipe FRM EQU202203014629 , FDT202304016501 REFERENCES 1. ↵ Yakar , S. & Isaksson , O . Regulation of skeletal growth and mineral acquisition by the GH/IGF-1 axis: Lessons from mouse models . Growth Horm. IGF Res . 28 , 26 – 42 ( 2016 ). OpenUrl CrossRef PubMed 2. ↵ Mertens , A. et al. Causes and consequences of child growth faltering in low-resource settings . Nature 621 , 568 – 576 ( 2023 ). OpenUrl CrossRef PubMed 3. ↵ Thurstans , S. et al. Boys are more likely to be undernourished than girls: a systematic review and meta-analysis of sex differences in undernutrition . BMJ Glob Health 5 , e004030 ( 2020 ). OpenUrl Abstract / FREE Full Text 4. ↵ Solon-Biet , S. M. et al. Macronutrient balance, reproductive function, and lifespan in aging mice . Proc Natl Acad Sci U S A 112 , 3481 – 3486 ( 2015 ). OpenUrl Abstract / FREE Full Text 5. ↵ Green , C. L. et al. Sex and genetic background define the metabolic, physiologic, and molecular response to protein restriction . Cell Metab 34 , 209 – 226 .e5 ( 2022 ). OpenUrl CrossRef PubMed 6. ↵ Reeves , P. G. , Nielsen , F. H. & Fahey , G. C . AIN-93 purified diets for laboratory rodents: final report of the American Institute of Nutrition ad hoc writing committee on the reformulation of the AIN-76A rodent diet . J Nutr 123 , 1939 – 1951 ( 1993 ). OpenUrl Abstract / FREE Full Text 7. ↵ Thoumas , J.-L. , Cavaroc , A. , Sery , D. , Leulier , F. & De Vadder , F . Suboptimal refeeding compensates stunting in a mouse model of juvenile malnutrition . J Nutr 155 , 849 – 861 ( 2025 ). OpenUrl PubMed 8. ↵ Zera , A. J. & Harshman , L. G . The Physiology of Life History Trade-Offs in Animals . Annu. Rev. Ecol. Syst . 32 , 95 – 126 ( 2001 ). OpenUrl CrossRef Web of Science 9. ↵ Flatt , T. , Tu , M.-P. & Tatar , M . Hormonal pleiotropy and the juvenile hormone regulation ofDrosophila development and life history . Bioessays 27 , 999 – 1010 ( 2005 ). OpenUrl CrossRef PubMed Web of Science 10. ↵ Ellison, P. T. Energetics and reproductive effort. American J Hum Biol 15 , 342 – 351 ( 2003 ). 11. ↵ Giller , K. , Huebbe , P. , Doering , F. , Pallauf , K. & Rimbach , G . Major urinary protein 5, a scent communication protein, is regulated by dietary restriction and subsequent re-feeding in mice . Proc Biol Sci 280 , 20130101 ( 2013 ). 12. ↵ Gibson , G . Decanalization and the origin of complex disease . Nat Rev Genet 10 , 134 – 140 ( 2009 ). OpenUrl CrossRef PubMed Web of Science 13. ↵ Paulusma , C. C. , Lamers , W. H. , Broer , S. & van de Graaf , S. F. J . Amino acid metabolism, transport and signalling in the liver revisited . Biochem Pharmacol 201 , 115074 ( 2022 ). OpenUrl CrossRef PubMed 14. ↵ Yap , Y. W. et al. Restriction of essential amino acids dictates the systemic metabolic response to dietary protein dilution . Nat Commun 11 , 2894 ( 2020 ). OpenUrl CrossRef PubMed 15. ↵ Chaveroux , C. et al. Regulating the expression of therapeutic transgenes by controlled intake of dietary essential amino acids . Nat Biotechnol 34 , 746 – 751 ( 2016 ). OpenUrl PubMed 16. ↵ Inagaki , T. et al. Endocrine regulation of the fasting response by PPARalpha-mediated induction of fibroblast growth factor 21 . Cell Metab 5 , 415 – 425 ( 2007 ). OpenUrl CrossRef PubMed Web of Science 17. Laeger , T. et al. FGF21 is an endocrine signal of protein restriction . J Clin Invest 124 , 3913 – 3922 ( 2014 ). OpenUrl CrossRef PubMed Web of Science 18. ↵ Laeger , T. et al. Metabolic Responses to Dietary Protein Restriction Require an Increase in FGF21 that Is Delayed by the Absence of GCN2 . Cell Reports 16 , 707 – 716 ( 2016 ). OpenUrl PubMed 19. ↵ Iroz , A. et al. A Specific ChREBP and PPARα Cross-Talk Is Required for the Glucose-Mediated FGF21 Response . Cell Rep 21 , 403 – 416 ( 2017 ). OpenUrl CrossRef PubMed 20. ↵ Fougerat , A. et al. ATGL-dependent white adipose tissue lipolysis controls hepatocyte PPARα activity . Cell Rep 39 , 110910 ( 2022 ). 21. ↵ Kubicky , R. A. , Wu , S. , Kharitonenkov , A. & De Luca , F . Role of fibroblast growth factor 21 (FGF21) in undernutrition-related attenuation of growth in mice . Endocrinology 153 , 2287 – 2295 ( 2012 ). OpenUrl CrossRef PubMed Web of Science 22. ↵ Gliniak , C. M. et al. FGF21 promotes longevity in diet-induced obesity through metabolic benefits independent of growth suppression . Cell Metab S1550-4131 ( 25 ) 00267 –0 ( 2025 ) doi: 10.1016/j.cmet.2025.05.011 . OpenUrl CrossRef 23. ↵ Owen , B. M. et al. FGF21 contributes to neuroendocrine control of female reproduction . Nat Med 19 , 1153 – 1156 ( 2013 ). OpenUrl CrossRef PubMed 24. ↵ Singhal , G. et al. Fibroblast growth factor 21 has no direct role in regulating fertility in female mice . Mol Metab 5 , 690 – 698 ( 2016 ). OpenUrl PubMed 25. ↵ Anderson , G. M. et al. Metabolic control of puberty: 60 years in the footsteps of Kennedy and Mitra’s seminal work . Nat Rev Endocrinol 20 , 111 – 123 ( 2024 ). OpenUrl CrossRef PubMed 26. ↵ Roa , J. et al. Metabolic regulation of female puberty via hypothalamic AMPK-kisspeptin signaling . Proc Natl Acad Sci U S A 115 , E10758 – E10767 ( 2018 ). OpenUrl Abstract / FREE Full Text 27. ↵ Wang , L. & Moenter , S. M . Differential Roles of Hypothalamic AVPV and Arcuate Kisspeptin Neurons in Estradiol Feedback Regulation of Female Reproduction . Neuroendocrinology 110 , 172 – 184 ( 2020 ). OpenUrl CrossRef PubMed 28. ↵ Bonduriansky , R. & Chenoweth , S. F. Intralocus sexual conflict . Trends Ecol Evol 24 , 280 – 288 ( 2009 ). OpenUrl CrossRef PubMed Web of Science 29. ↵ Solon-Biet , S. M. et al. Defining the Nutritional and Metabolic Context of FGF21 Using the Geometric Framework . Cell Metab 24 , 555 – 565 ( 2016 ). OpenUrl CrossRef PubMed 30. ↵ Solon-Biet , S. M. et al. Toward reconciling the roles of FGF21 in protein appetite, sweet preference, and energy expenditure . Cell Rep 42 , 113536 ( 2023 ). 31. ↵ Byers , S. L. , Wiles , M. V. , Dunn , S. L. & Taft , R. A . Mouse Estrous Cycle Identification Tool and Images . PLoS One 7 , e35538 ( 2012 ). OpenUrl CrossRef PubMed 32. ↵ Korenbrot , C. C. , Huhtaniemi , I. T. & Weiner , R. I . Preputial Separation as an External Sign of Pubertal Development in the Male Rat1 . Biology of Reproduction 17 , 298 – 303 ( 1977 ). OpenUrl CrossRef PubMed Web of Science 33. ↵ Ma , T. et al. Environmentally relevant perinatal exposure to DBP disturbs testicular development and puberty onset in male mice . Toxicology 459 , 152860 ( 2021 ). 34. ↵ Chai , C. K . Seminal Vesicle Growth as a Function of Androgen Stimulation in Mice . American Journal of Physiology-Legacy Content 186 , 463 – 467 ( 1956 ). OpenUrl 35. De Gendt , K. et al. Development and Function of the Adult Generation of Leydig Cells in Mice with Sertoli Cell-Selective or Total Ablation of the Androgen Receptor . Endocrinology 146 , 4117 – 4126 ( 2005 ). OpenUrl CrossRef PubMed Web of Science 36. Glass , A. R. , Herbert , D. C. & Anderson, J. Fertility Onset, Spermatogenesis, and Pubertal Development in Male Rats: Effect of Graded Underfeeding . Pediatr Res 20 , 1161 – 1167 ( 1986 ). OpenUrl PubMed Web of Science 37. Russell , L. D. , Ettlin , R. A. , Hikim , A. P. S. & Clegg , E. D . Histological and Histopathological Evaluation of the Testis . International Journal of Andrology 16 , 83 – 83 ( 1993 ). OpenUrl CrossRef 38. ↵ Spears , J. , Matthews , A. & Hartwig , S . Evaluation of noninvasive methods of predicting testicular weight and development in an inbred and outbred strain of laboratory mouse . Theriogenology 79 , 566 – 573 ( 2013 ). OpenUrl CrossRef PubMed 39. ↵ Solon-Biet , S. M. et al. Dietary Protein to Carbohydrate Ratio and Caloric Restriction: Comparing Metabolic Outcomes in Mice . Cell Rep 11 , 1529 – 1534 ( 2015 ). OpenUrl CrossRef PubMed 40. ↵ Bustin , S. A. et al. The MIQE guidelines: minimum information for publication of quantitative real-time PCR experiments . Clin Chem 55 , 611 – 622 ( 2009 ). OpenUrl Abstract / FREE Full Text 41. ↵ Martin , M . Cutadapt removes adapter sequences from high-throughput sequencing reads . EMBnet.journal 17 , 10 – 12 ( 2011 ). OpenUrl 42. ↵ Dobin , A. et al. STAR: ultrafast universal RNA-seq aligner . Bioinformatics 29 , 15 – 21 ( 2013 ). OpenUrl CrossRef PubMed Web of Science 43. ↵ Smith , T. , Heger , A. & Sudbery , I . UMI-tools: modeling sequencing errors in Unique Molecular Identifiers to improve quantification accuracy . Genome Res 27 , 491 – 499 ( 2017 ). OpenUrl Abstract / FREE Full Text 44. ↵ Anders , S. , Pyl , P. T. & Huber , W . HTSeq--a Python framework to work with high-throughput sequencing data . Bioinformatics 31 , 166 – 169 ( 2015 ). OpenUrl CrossRef PubMed Web of Science 45. ↵ Gallart-Ayala , H. et al. A global HILIC-MS approach to measure polar human cerebrospinal fluid metabolome: Exploring gender-associated variation in a cohort of elderly cognitively healthy subjects . Anal Chim Acta 1037 , 327 – 337 ( 2018 ). OpenUrl CrossRef PubMed 46. Medina , J. et al. Single-Step Extraction Coupled with Targeted HILIC-MS/MS Approach for Comprehensive Analysis of Human Plasma Lipidome and Polar Metabolome . Metabolites 10 , 495 ( 2020 ). 47. ↵ van der Velpen , V. et al. Systemic and central nervous system metabolic alterations in Alzheimer’s disease . Alzheimers Res Ther 11 , 93 ( 2019 ). View the discussion thread. Back to top Previous Next Posted July 11, 2025. Download PDF Supplementary Material Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Liver FGF21 mediates sex-specific adaptation to juvenile protein malnutrition Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Liver FGF21 mediates sex-specific adaptation to juvenile protein malnutrition Amélie Joly , Lucas Rebiffé , Anne Lambert , Estelle Caillon , Sandrine Hughes , Benjamin Gillet , Isabelle Rahioui , Pedro Da Silva , Justine Bruse , Anne Fougerat , Filipe De Vadder , François Leulier bioRxiv 2025.07.09.663702; doi: https://doi.org/10.1101/2025.07.09.663702 Share This Article: Copy Citation Tools Liver FGF21 mediates sex-specific adaptation to juvenile protein malnutrition Amélie Joly , Lucas Rebiffé , Anne Lambert , Estelle Caillon , Sandrine Hughes , Benjamin Gillet , Isabelle Rahioui , Pedro Da Silva , Justine Bruse , Anne Fougerat , Filipe De Vadder , François Leulier bioRxiv 2025.07.09.663702; doi: https://doi.org/10.1101/2025.07.09.663702 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Physiology Subject Areas All Articles Animal Behavior and Cognition (7635) Biochemistry (17690) Bioengineering (13892) Bioinformatics (41935) Biophysics (21451) Cancer Biology (18587) Cell Biology (25499) Clinical Trials (138) Developmental Biology (13377) Ecology (19899) Epidemiology (2067) Evolutionary Biology (24318) Genetics (15609) Genomics (22506) Immunology (17736) Microbiology (40394) Molecular Biology (17181) Neuroscience (88601) Paleontology (666) Pathology (2832) Pharmacology and Toxicology (4824) Physiology (7641) Plant Biology (15152) Scientific Communication and Education (2045) Synthetic Biology (4294) Systems Biology (9825) Zoology (2271)