The fibroblast growth factor 23 (FGF23) hormone, produced mainly by bone osteocytes, participates in phosphate (Pi) and vitamin D metabolism (ADHR Consortium, 2000). FGF23 was initially identified in studies on autosomal dominant hypophosphatemic rickets (ADHR), in which mutations in the FGF23 gene lead to excessive serum FGF23 concentration by preventing cleavage of the active intact FGF23 (iFGF23) into inactive N-terminal and C-terminal FGF23 (cFGF23). This results in higher iFGF23 concentrations and consequently in increased urinary phosphate loss (ADHR Consortium, 2000; White et al., 2001; Shimada et al., 2004; Takashi et al., 2018). Increased iFGF23 concentrations have, thereafter, been shown to be the underlying cause in several other genetic forms of hypophosphatemia (Shimada et al., 2001; Araya et al., 2005; Benet-Pagès et al., 2005; Martin et al., 2012; Gohil and Imel, 2019).

Phosphate is essential for bone mineralization and is mainly found in bone and intracellular compartments, with approximately 1% of total phosphate circulating in the blood stream. Circulating phosphate concentrations are tightly regulated by FGF23, 1,25 dihydroxyvitamin D (1,25-(OH)₂D), and parathyroid hormone (PTH) (ADHR Consortium, 2000; Shimada et al., 2001; Erben and Andrukhova, 2017; Arnold et al., 2021). FGF23, in its intact form, directly reduces reabsorption of phosphate in the renal proximal tubules by suppressing the expression of sodium phosphate co-transporters, of which type IIa is encoded by the SLC34A1 gene. FGF23 also affects phosphate concentrations indirectly by participating in the regulation of 1,25-(OH)₂D concentration by, for example, increasing the expression of 25-hydroxyvitamin D-24-hydroxylase, a catabolic enzyme of vitamin D, encoded by the CYP24A1 gene. Increased FGF23 leads to decreasing 1,25-(OH)₂D concentrations and reduced intestinal phosphate uptake (Shimada, 2007; Martin et al., 2012; Quarles, 2012; Lederer, 2014; Takashi et al., 2018).

Circulating concentrations of phosphate and FGF23 show individual variation. Genome-wide association studies (GWAS) have identified several single-nucleotide polymorphisms (SNPs) associated with serum phosphate and FGF23 concentrations. SNPs near the genes FGF23 and SLC34A1 and CASR, the gene encoding the calcium sensing receptor, are associated with serum phosphate (Kestenbaum et al., 2010), while SNPs in genes involved in vitamin D metabolism, namely, CYP24A1, are associated with differences in circulating FGF23 concentrations (Robinson-Cohen et al., 2018).

The FGF23 polymorphism rs7955866 (SNP1) is a missense variant located in the third exon of FGF23 (c.716C>T, p.239T > M). Based on its location and the differing properties of the amino acid variants, rs7955866 is thought to significantly influence interactions of FGF23 with FGF receptors and Klotho, and thereby to have a role in phosphate homeostasis (Yamazaki et al., 2008; Rendina et al., 2012; Agoro et al., 2020). In adults with calcium nephrolithiasis, rs7955866 is associated with lower phosphate concentrations and increased urinary phosphate excretion (Rendina et al., 2012). FGF23 variation has also been described to relate to cardiovascular outcomes, particularly in patients with chronic kidney disease. The 3′UTR variant rs11063112 (c.2185A>T) (SNP2) and intronic variant rs13312770 (c.211 + 3287A>G) (SNP3) have been associated with elevated risk of cardiovascular mortality in hemodialysis patients with chronic kidney disease and end-stage chronic kidney disease, respectively (Rothe et al., 2017; Schwantes-An et al., 2019).

We have previously reported that in infants aged 3–24 months, the concentrations and modifying factors of FGF23 depend on sex and age, iFGF23 being higher in girls and both iFGF23 and cFGF23 decreasing with age (Holmlund-Suila et al., 2016; Holmlund-Suila et al., 2017; Enlund-Cerullo et al., 2020). Furthermore, in a cohort of 183 school-aged children and adolescents, we have previously observed that common variants of FGF23 (SNP1, SNP2, and rs3832879) show associations with PTH concentrations and urinary phosphate excretion, as well as with total hip bone mineral density (BMD) Z-score (Pekkinen et al., 2015), thereby linking FGF23 variants to bone mass.

It remains unknown whether genetic variation in FGF23 influences phosphate and vitamin D metabolism and bone strength during infancy. In this longitudinal cohort study, we examined associations of three FGF23 polymorphisms (SNP1–3), with FGF23, phosphate, and vitamin D concentrations and with bone strength, determined by peripheral quantitative computed tomography (pQCT), from age 12–24 months in 622 healthy infants participating in a vitamin D intervention trial.

This study finds that, in healthy young children, common FGF23 SNPs rs7955866 (SNP1), rs11063112 (SNP2), and rs13312770 (SNP3) are associated with circulating FGF23 and phosphate concentrations and with pQCT-derived bone strength parameters, as well as with change in bone strength from 12 to 24 months. In all participants, SNP1 minor allele homozygosity was related to lower cFGF23 concentrations at both 12 and 24 months of age. SNP3 minor allele carriers had greater total BMC and showed greater increase in total BMC, but smaller increase in total CSA and PMI, from 12 to 24 months. In addition, the SNP2 genotype was associated with change in phosphate from 12 to 24 months, with the greatest decrease in concentrations observed in minor allele carriers. Further genotype associations were observed only in participants receiving standard vitamin D supplementation, although vitamin D intervention did not affect the studied bone strength or biochemical parameters, apart from 25-OHD and PTH. Concentrations of 25-OHD were not modified by the studied FGF23 SNP genotypes.

Previous GWAS studies in adults have found SNPs near FGF23, SLC34A1, and CASR to relate to differences in phosphate concentrations (Kestenbaum et al., 2010). While direct associations of FGF23 SNP2 or SNP3 with phosphate metabolism have not previously been described, minor allele homozygosity of SNP1 is related to lower phosphate concentrations in adult nephrolithiasis patients and a smaller cohort of school-aged children (Rendina et al., 2012; Pekkinen et al., 2015). SNPs linked to genes involved in vitamin D metabolism; cleavage and inactivation of FGF23 (PCSK9); and cell proliferation, inflammation, and apoptosis (HLA-DQA1, HLA-DQB1, and TGFB2, respectively) have, in turn, are associated with circulating FGF23 concentrations (White et al., 2001; Benet-Pagès et al., 2004; Robinson-Cohen et al., 2018; Chuang et al., 2020). Apart from confirming the associations of FGF23 genetic variation with phosphate metabolism in the studied age group, our study is, to our knowledge, the first to show a direct association between variation in the FGF23 gene and FGF23 concentrations.

In all participants, minor alleles were related to lower cFGF23 throughout follow-up (SNP1) and positive allelic effects on phosphate at 12 months of age (SNP1-2), but the greatest decrease in phosphate was observed from 12 to 24 months (SNP2 in all, SNP3 in boys). In analyses by sex, FGF23 minor allele carriers showed lowest iFGF23 in boys (SNP2) and negative allelic effects in girls (SNP3). The FGF23 genotype also notably affected bone strength and age-dependent change thereof. Minor alleles related to greater increase of total BMC, but smaller increase of total CSA and PMI, compared with major allele homozygotes, with heterozygotes having highest total BMC, CSA, and PMI at 24 months (SNP3). In contrast to SNP3, in boys, minor allele homozygosity of SNP2 was observed to be associated with lower total BMC at both 12 and 24 months, and lower total vBMD at 24 months.

The genotype-related differences in phosphate concentrations at 12 months differ from those previously reported in adult nephrolithiasis patients and a smaller cohort of school-aged children (Rendina et al., 2012; Pekkinen et al., 2015), with minor allele homozygosity relating to higher, instead of lower, phosphate concentration. We have previously found phosphate concentrations to markedly decrease from 12 to 24 months of age (Koljonen et al., 2021). In light of the age-related decrease in phosphate, which is greater in FGF23 minor allele carriers (SNP2-3), the different direction of genotype associations observed at 12 vs. 24 months of age, and the simultaneously decreasing FGF23 and PTH concentrations, the differing findings could be explained by age-specific regulation.

Differences in the direction of genotype associations at 12 and 24 months are similarly observed for several bone parameters. Significant differences in temporal change, but not in cross-sectional adjusted means, as observed, for example., for SNP3 and total BMC, CSA, and PMI, are indicative of differences in genotype effects at the two studied time points. Growth velocity in infancy is greater in boys and decreases after 12 months of age (Kiviranta et al., 2016), while physical activity increases, including standing and onset of walking, that previously have been shown to be associated with bone strength in early childhood (Ireland et al., 2014). Observed differences between the sexes possibly reflect differences in growth velocity, as well as differences in FGF23 regulation previously observed between girls and boys in the studied age group (Holmlund-Suila et al., 2017; Enlund-Cerullo et al., 2020). We hypothesize that the differences in genotype effects, observed both for studied biochemical and bone parameters, potentially relate to age-specific differences in regulation, as well as to differences in growth phase and physical activity at 12 and 24 months.

The observed associations of the FGF23 genotype and cFGF23 concentrations in all participants could potentially indicate reduced expression of iFGF23, with correspondingly higher phosphate concentration as observed at 12 months, in minor allele carriers (SNP1), potentially related to altered binding to the FGF receptor and Klotho complex (Yamazaki et al., 2008; Rendina et al., 2012). The reduced expression of iFGF23 could be balanced by a decrease in cleavage, resulting in non-significant differences in iFGF23 by the genotype, but subsequently observed as a decrease in cFGF23 in minor allele carriers. Findings of minor allele genotype associations with low iFGF23 at 12 months observed in analyses by sex potentially support this hypothesis.

In limited adult studies, iFGF23 concentrations have been found to be negatively associated with BMD (Bouksila et al., 2019; Wang et al., 2020). Apart from its effects on phosphate metabolism, in vitro studies have suggested FGF23 expression to independently regulate bone formation (Wang et al., 2008). Inverse association of iFGF23 with BMD parameters has also recently been observed in a Mendelian randomization study exploring causal effects of FGF23 (Yokomoto-Umakoshi et al., 2021). As our study setting does not allow us to establish causality, the mechanisms underlying the observed differences between FGF23 genotype and bone strength and temporal change, thereof, require further studies. Based on earlier reports (Bouksila et al., 2019; Wang et al., 2020; Yokomoto-Umakoshi et al., 2021), the observed genotype effects on bone parameters could potentially correspond with effects on iFGF23 expression and concentrations. Our previous study in a smaller cohort of school-aged children observed associations of FGF23 diplotype and total hip BMD Z-scores (Pekkinen et al., 2015). The present study finds variation in FGF23 in infants to be primarily associated with differences in total BMC, total CSA, and PMI, i.e., measures of bone size, mass, and resistance to torsion, rather than with volumetric bone mineral density (total vBMD), for which associations are observed only in Group₁₀ (SNP3).

Considering the tight interplay between vitamin D, phosphate, and bone metabolism, our study uniquely explores the effects of genetic variation in FGF23 in a vitamin D intervention setting. In our largely vitamin D sufficient study population, vitamin D intervention affected 25-OHD and PTH levels, as previously reported, but did not directly modify other biochemical or bone parameters (Rosendahl et al., 2018; Valkama et al., 2020). Some of our findings were, however, specific for participants receiving standard vitamin D supplementation (Group₁₀). In contrast to our earlier study in school-aged children, FGF23 genotype did not modify 25-OHD, but in Group₁₀, minor alleles associated with lower PTH (SNP1), a finding in line with our previous report (Pekkinen et al., 2015). As vitamin D intervention had a significant effect on PTH and 25-OHD concentration, it is plausible that the impact of higher vitamin D supplementation in Group₃₀ did not allow detection of possible smaller effects of the SNP1 genotype on PTH and 25-OHD concentrations. We have previously found 25-OHD to modify iFGF23 in girls, while seasonal differences, possibly reflecting changes in vitamin D status, are associated with iFGF23 concentrations in boys (Enlund-Cerullo et al., 2020). In the present study, 25-OHD modified adjusted iFGF23 concentrations by the SNP1 genotype. Indirect effects of 25-OHD are possibly also reflected in the seasonal differences observed in bone parameters and iFGF23 and cFGF23 concentrations.

It is interesting to note that we observe positive effects on bone strength parameters for SNP1 only in participants receiving standard, but not high-dose, vitamin D supplementation. In Group₁₀, minor allele homozygosity was related to higher total BMC, CSA, and PMI at 12 months and higher total BMC but lower CSA at 24 months, and smaller increase of total CSA and PMI from 12 to 24 months (SNP1), largely mirroring the findings in all participants (SNP3). As effects of SNP1 include potential differences in binding of FGF23 to FGF receptors and Klotho (Yamazaki et al., 2008; Rendina et al., 2012), it can be speculated that interaction with vitamin D and its metabolites is modified in minor allele carriers, possibly depending on 25-OHD and vitamin D supplementation dose. Further studies are needed to explore the interplay of vitamin D, FGF23, and genetic variation in FGF23 with bone strength and modeling in young children.

When considering observed associations between singular SNPs and phenotype, it is to be noted that associations may, in part, be due to linkage disequilibrium between studied and other genetic variants. In our study, potential differences in expression and post-translational modification of FGF23 through altered binding to the FGF receptor and Klotho and consequent differences in signaling (SNP1) (Yamazaki et al., 2008; Rendina et al., 2012) and possible differences in susceptibility to O-glycosylation (SNP2) may directly contribute to the observed differences in biochemical and bone parameters. The associations of SNP3 variation with bone parameters could possibly reflect minute effects on FGF23 function and PTH and phosphate concentrations, that were not statistically detectable.

The large cohort of healthy infants with uniform and extensive follow-up is a notable strength of this study, uniquely allowing us to examine genotype effects on temporal changes, in addition to cross-sectional associations, of the studied biochemical and bone parameters. Despite the cohort size, minor allele homozygotes of the studied SNPs, and consequently their haplotypes, were, however, rare, making detection of significant genotype effects challenging and arguably inconclusive, for example, for variants where heterozygotes show greatest means. It is possible that some genotype associations may have escaped detection, and larger studies are warranted to further examine the associations of FGF23 variation and mineral and bone metabolism. Although the intervention setting allowed us to examine the potential impact of different vitamin D supplementation doses, direct supplementation effects on, for example, 25-OHD and PTH concentrations may have made detection of smaller, FGF23 genotype-derived associations more challenging, especially in the higher-dose supplementation group. Analyses of 1,25-(OH)₂D, urinary phosphate, creatinine, or calcium were unfortunately not conducted in this study.

Our study presents novel associations of FGF23 and phosphate concentrations and bone strength with common SNPs of FGF23 in healthy infants. Our findings highlight the role of FGF23 in bone modeling, in particular effects on bone mineral content, cross-sectional area, and PMI, in growing young children. As simple measurements of FGF23 concentrations are subject to many confounding factors, our findings can help understand the role of FGF23 in bone metabolism and its temporal changes during early childhood.