Animal experiments were approved by the State Office Committee for Animal Welfare Lower Saxony (LAVES, #15/1912) in accordance with national animal protection guidelines from Directive 2016/63/EU of the European Parliament on the protection of animals used for scientific purposes. Mice were housed with a 12:12 h light-dark cycle, had free access to water and fed ad libitum with standard rodent chow. The Myh7-Cre mouse line expressing Cre recombinase linked to a Myh7 promotor (B6;129S1-Tg(Myh7-Cre)1Jmk) was kindly provided by Prof. Joerg Heineke (Department of Cardiovascular Physiology, European Center for Angioscience (ECAS), Medical Faculty Mannheim, University of Heidelberg, Mannheim, Germany) and described elsewhere (Heineke et al., 2007). Fgf23 floxed mice were generated based on CRISPR/Cas9 technology by InGenious Targeting, Inc. (Ronkonkoma, NY, United States). To obtain mice with cardiac myocyte-specific Fgf23 deletion, homozygous Fgf23^fl/fl mice (B6-Fgf23^tm1Mln) were interbred with B6;129S1-Tg(Myh7-Cre)1^Jmk. Fgf23^fl/fl/cre- littermates were used as controls.

For genotyping, earmarks were lysed in proteinase K buffer overnight. After deactivation of proteinase K, multiplex PCR was performed to amplify genomic DNA using primer for Fgf23 floxed allele (forward: AGA TTC CAT TTA CAG TGC CCC TTG G; reverse: ACT TCA GTT ACC TGA AGT CCC AGT TGG) and Myh7-Cre (forward: GGC GTT TTC TGA GCA TAC CT; reverse: CTA CAC CAG AGA CGG AAA TCC). PCR was run for 3 min at 94°C and 31 cycles of 30 s at 94°C, 45 s at 58°C, and 1 min at 72°C, followed by 10 min at 72°C. Samples were separate on a 3% agarose gel including GelRed Nucleic Acid Gel Stain (Biotrend) at 120 V for 60 min and analyzed with a Gel DocTMXR+ (Bio-Rad Laboratories).

Respective male and female Fgf23^fl/fl/cre⁻ and Fgf23^fl/fl/cre⁺ mice with the age of 0 (neonatal), 1, 3 and 6 months were analyzed. Femur, tibia, and calvaria were isolated and snap-frozen in liquid nitrogen. Isolated mouse hearts were washed in ice-cold PBS followed by 0.5% potassium chloride and cut into three cross-sections. For histological evaluation, the mid cross-section was fixated in 4% RotiHistofix (Carl Roth) and embedded in paraffin. The other two cross-sections were snap-frozen in liquid nitrogen and stored at −80°C for molecular and biochemical analyses.

Echocardiography was performed using the Vevo 2100 Imaging System (FUJIFILM VisualSonics Inc.) as described previously (Zacchigna et al., 2021). Anesthesia was induced with 3% isoflurane and maintained with 0.5–1% isoflurane over a breathing mask. Heart views were acquired in parasternal long axis in B- and M-mode, and short axis in M-mode. All echocardiographic images were analyzed using the Vevo Lab 3.2.0 software.

Cardiac catheterization was performed with PowerLab 16/35 (Millar Instruments). Anesthesia was initialized with 3% isoflurane and maintained with 0.8% isoflurane. The measurements were performed on the opened thorax using the polyamide catheter PVR 1045 (Millar Instruments) and analyzed with the software LabChart 8 (AD Instruments).

Transverse aortic constriction (TAC) or sham surgery was performed in eight-week-old male and female Fgf23^fl/fl/cre⁻ and Fgf23^fl/fl/cre⁺ mice by subjecting the aorta to a defined constriction as previously described (Grund et al., 2019a). In brief, a 7–0 silk suture was tied around the aortic arch between the branches of the brachiocephalic trunk and the left common carotid artery and a blunted 27-gauge needle, after which the needle was removed to create a defined constriction. The procedure for sham-operated animals was identical, but the aorta was not ligated. Two weeks after surgery, mice were sacrificed as described above.

Cardiac specific cell types were isolated from mice after sham and TAC as previously described (Appari et al., 2017). In brief, ventricular cardiac myocytes were isolated using a Langendorff system, cardiac endothelial cells and fibroblasts were isolated using MACS technology with CD146 microbeads followed by feeder removal microbeads (Miltenyi Biotec). All isolated cell types were directly used for RNA extraction according to standard protocol followed by deep-sequencing analysis as previously described (Grund et al., 2019b).

Cardiac magnetic resonance imaging (MRI) was performed for TAC or sham-operated Fgf23^fl/fl/cre⁻ and Fgf23^fl/fl/cre⁺ mice. Cardiac MRI images were recorded with a Bruker 7T Pharmascan 70/16 (Bruker). Anesthesia was induced with 3% isoflurane and maintained with 0.5% isoflurane. First, a positioning scan was performed, followed by images in 9–10 stacks. For every stack, images were made to encompass an entire heart cycle. All cardiac MRI images were analyzed with Mass4Mice 7.1 software (MEDIS Medical Technologies). Epicardium and endocardium were marked manually to ascertain left ventricular (LV) mass, end-systolic (ES) and end-diastolic (ED) volume, stroke volume (SV), cardiac output, and ejection fraction (EF). In addition, LV wall geometry was determined at the level of papillary muscles. With regard to slice thickness, papillary muscles were included in the LV cavity.

Plasma was analyzed for total FGF23 (#60-6300) and intact FGF23 (#60-6800) by ELISA (both from Immutopics). Urine and serum biochemistry were measured on a cobas c111 system (Roche Diagnostics) using Calcium Gen.2, Phosphate ver.2 and creatinine Creatinine Jaffé Gen.2 (all from Roche Diagnostics). No urine was available for one month old Fgf23^fl/fl/cre⁺ and Fgf23^fl/fl/cre⁻ mice.

Three µm thick cardiac mid-chamber sections were deparaffinized in xylene and hydrated using a descending alcohol series. Hematoxylin and eosin (HE) staining was performed using Meyers hematoxylin and eosin G (Merck Millipore). To visualize interstitial cardiac fibrosis, sections were incubated using picrosirius red solution (Merck Millipore) in 1.2% picric acid. For immunohistochemical staining of FGF23 in heart tissue, antigen retrieval was done in citrate buffer pH 6.0, followed by inhibition of endogenous peroxidase activity by hydrogen peroxidase and blocking with 10% normal goat serum in PBS. After incubation with anti-FGF23 primary antibody (Bioss #bs-5768R, dilution 1:200), sections were incubated with horseradish peroxidase-conjugated goat anti-rabbit (Jackson ImmunoResearch #111-035-144, dilution 1:100), followed by incubation with diaminobenzidine substrate (Thermo Scientific #34065) and counterstaining with hematoxylin. Images were taken with either 10x (HE) or 20x objective using the Axio Observer Z1 microscope (Carl Zeiss). To quantify interstitial cardiac fibrosis, five pictures per mouse of the left ventricle were taken using 20x magnification followed by quantification using ImageJ (Böckmann et al., 2019).

For visualization of cell type-specific expression of FGF23, cardiac tissue was stained for anti-CD31 (R&D systems #AF3628; dilution 1:100 in 5% milk in TBST), followed by donkey-anti-goat Alexa Fluor 480 secondary antibody (Invitrogen #A11055, dilution 1:500), anti-α-Actinin (Sigma-Aldrich #A7732, dilution 1:100), followed by goat-anti-mouse Alexa Fluor 660 secondary antibody (Invitrogen #A21054, dilution 1:500), or anti-PDGFRα (R&D Systems #AF1062, dilution 1:100), followed by donkey-anti-goat Alexa Fluor 480 secondary antibody. Afterwards, tissues were stained with anti-FGF23 (Bioss #bs-5768R, dilution 1:200), followed by goat-anti-rabbit Alexa Fluor 555 secondary antibody (Invitrogen #A21428, dilution 1:500). All primary antibodies were incubated over night at 4°C, and secondary antibodies were incubated for 90 min at room temperature. For measurement of capillarization and cardiac myocyte size, the staining with CD31 was followed by incubation with 5 µg/ml wheat germ agglutinin (WGA)Alexa Fluor 555 (Invitrogen #W32464) for one hour. Nuclear counterstaining in immunofluorescence was performed with DAPI. All images were taken with a 20x objective on a Zeiss Axio Observer Z1 microscope (Carl Zeiss) and quantification of cardiac myocyte area was performed with 100 cells per animal using the Zen 2.3 lite software (Carl Zeiss). For quantification of capillarization, five images in 20x magnification per animal were taken and number of capillaries were divided by number of cardiac myocytes per power field.

The RNeasy Mini Kit was used according to manufacturer’s protocol for total RNA isolation from murine heart and cardiac specific cell types. For murine bone (either calvaria or femur), the RNeasy Lipid Tissue Kit was used. 500 ng RNA was transcribed into cDNA according to the data sheet of the QuantiTect Reverse Transcription Kit. Real-time polymerase chain reaction (PCR) was performed in triplets using the QuantiFAST SYBR Green PCR Kit (all kits from Qiagen) on a 7900 HT Fast RT-PCR System (Thermo Fisher Scientific) and analyzed using the SDS Software v2.4 (Thermo Fisher Scientific). For murine primer sequences see Supplementary Table S1. Relative gene expression levels of Fgf23^fl/fl/cre⁺ mice compared to control were calculated using the 2^−ΔΔCt method, with Gapdh and 18s (in bone) serving as housekeeping genes. For Fgf23 mRNA quantification in Figures 1, 4, 2^−ΔCt values were used.

Protein isolation of murine heart tissue was performed with the TissueLyser LT (Qiagen) using RIPA buffer with proteinase and phosphatase inhibitors (Sigma-Aldrich). Samples were incubated at 4°C for 30 min followed by sonification. Protein concentrations were measured via BCA assay (Thermo Fisher Scientific) according to manufacturer’s protocol.

100 µg total protein was separated on a 12% SDS-gel and transferred on nitrocellulose membrane in transfer buffer containing 18% methanol for 1 h. Membranes were blocked in 5% milk in TBS containing 0.05% Tween-20 for 1 h, afterwards primary antibodies were incubated over-night at 4°C (FGF23-6310 Goat mAb, dilution 1:500, Immutopics; GAPDH 14C10 Rabbit mAb, dilution 1:1,000, Cell Signaling Technology). After washing, secondary antibodies were incubated for 1 h at room temperature (Goat anti-rabbit IgG HRP, dilution 1:1,000, Santa Cruz; Donkey anti-goat IgG HRP, dilution 1:2000, R and D Systems). ECL development was done with the SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific) and signals were detected with the Odyssey FC Imaging System (LI-COR Bioscience). Expression levels were quantified with the Image Studio Lite 5.2 software (LI-COR Bioscience).

All statistical analyses were performed using GraphPad Prism software 7 (GraphPad Software) Gaussian’s distribution was analyzed by Shapiro-Wilk test. Differences between two groups were determined using 2-tailed Students t-test or Mann Whitney U test, respectively. For time-dependent analyses, one-way ANOVA or Kruskal-Wallis test followed by Dunnett or Dunn’s multiple comparison tests were used. TAC-experiments were analyzed by two-way ANOVA followed by Tukey’s multiple comparison test. p values < 0.05 were considered as statistically significant.

Most clinical and experimental studies over the past decade focused on FGF23 in the context of LV pathologies (Faul et al., 2011; Grabner et al., 2015; Matsui et al., 2018). It has not been investigated so far in which areas of the whole heart Fgf23 is locally expressed. Therefore, we first performed immunohistochemical staining for Fgf23 using longitudinal sections of healthy adult C57BL/6N wild-type mice. The expression of Fgf23 was comparable between the left and the right ventricle (Figure 1A). In direct comparison, the synthesis of Fgf23 was higher in the left atrium than in the right atrium (Figure 1A). Fgf23 was clearly expressed in all regions of endocardium and endothelium (Figure 1A), while it appeared to be only minimal in the tunica media of the depicted vessels, which is predominantly composed of vascular smooth muscle cells (Figure 1A). To further elucidate the expression pattern of murine Fgf23 within specific cardiac cell types, we performed co-immunofluorescence staining of Fgf23 with cell type specific markers for cardiac myocytes, fibroblasts and endothelial cells in formalin-fixed paraffin-embedded cardiac cross-sections of adult C57BL/6N wild-type mice. Fgf23 was co-localized with the cardiac myocyte-specific marker α-Actinin, verifying, that healthy cardiac myocytes express Fgf23 (Figure 1B). Fgf23 was further expressed in platelet-derived growth factor receptor A (PDGFRα) positive cardiac fibroblast (Figure 1C) and co-localized with CD31, known as platelet endothelial cell adhesion molecule-1 (PECAM-1), suggesting the expression of Fgf23 in endothelial cells of the murine heart, too (Figure 1D).

Studies in humans showed that circulating total FGF23 is highest in infants and adolescents and lowest at adult age (Fischer et al., 2012), but the age-dependent synthesis of FGF23 in the heart and bone has not been studied so far. Therefore, we next investigated the Fgf23 expression in heart and bone tissue of C57BL/6N mice from neonatal (0) to 6 months of age. Although the individual variability of mice was high, cardiac Fgf23 mRNA levels were significantly lower in 1 month old mice compared to neonates and increased again with increasing age (Figure 1E), whereas no age-related changes in bone Fgf23 mRNA expression were noted (Figure 1F). Of note, osseous expression of Fgf23 was approximately 20-fold higher than in the heart. Immunoblot analysis of total cardiac tissue confirmed an age-dependent increase of Fgf23 protein synthesis (Figure 1G), which was verified by immunofluorescence staining of Fgf23 in cardiac mid-chamber cross-sections (Figure 1H). It appears that the increasing total amount of cardiac Fgf23 originates mainly from cardiac myocytes and that expression in endothelial cells and fibroblasts remains approximately the same with increasing age.

Taken together, Fgf23 is not only expressed in cardiac myocytes, but also in cardiac fibroblasts and endothelial cells, suggesting that the heart is a source of Fgf23.

The physiological role of bone-derived endocrine-acting FGF23 in renal phosphate homeostasis is well-established (Shimada et al., 2004), whereas the role of intra-cardiac FGF23 synthesis has been poorly studied. To investigate cardiac-derived FGF23 in more detail, we generated a cardiac myocyte-specific Fgf23 knockout mouse using Cre-LoxP system (Figure 2A). The first LoxP site was inserted 222 bp upstream of exon 2 and the Neo cassette, containing the second LoxP site, was inserted 342 bp downstream of exon 2. The size of the target region was 668 bp containing exon 2. The respective Fgf23-LoxP construct was delivered to iTL BF1 (C57BL/6 FLP) embryonic stem cells, which were microinjected into Balb/c blastocysts using standard protocols. Resulting chimeras with high percentage black fur color were mated to C57BL/6 wild-type mice to generate germline floxed Fgf23 mice. After intercrossing Fgf23^fl/+ mice, homozygous non-recombined Fgf23-LoxP mice (Fgf23^fl/fl) were mated with transgenic cardiac myocyte-specific B6;129S1-Tg(Myh7-Cre)^1Jmk mice by standard breeding regimen. As published elsewhere, in B6;129S1-Tg(Myh7-Cre)^1Jmk mice, the Cre enzyme has a significant activity at embryonic day 17.5 (Oka et al., 2006). Finally, female Fgf23^fl/fl/Myh7-cre⁻ mice were bred with males fully recombined for Fgf23 (Fgf23^fl/fl/Myh7-cre⁺). To verify genotypes, PCR products were amplified from genomic DNA using primer pairs spanning the floxed exon 2 region of murine Fgf23 and Cre recombinase in a semi-quantitative multiplex PCR (see Methods). Thereby a 253 bp product represented the wild-type Fgf23 allele, a 367 bp product the floxed Fgf23 allele, and a 550 bp product the Cre expression (Figure 2B). Male and female Fgf23^fl/fl/Myh7-cre⁺ mice (in the following termed as Fgf23^fl/fl/cre⁺) were analyzed in all experiments and respective non-recombined Fgf23^fl/fl/Myh7-cre⁻ littermates (in the following termed as Fgf23^fl/fl/cre⁻) were used as controls. First, we verified the cardiac myocyte-specific Fgf23 knockout. Thereby, Fgf23^fl/fl/cre⁺ mice showed significantly reduced cardiac Fgf23 mRNA expression compared to controls (Figure 2C), while their osseous Fgf23 synthesis remained unaffected (Figure 2D). Western blot analysis showed significantly reduced full-length cardiac Fgf23 protein in Fgf23^fl/fl/cre⁺ compared to control mice (Figure 2E). Using anti-α-Actinin antibody, immunofluorescence microscopy confirmed the cardiac myocytes-specific Fgf23 knockout (Figure 2F). However, circulating total and intact Fgf23 (iFgf23) levels in Fgf23^fl/fl/cre⁺ mice were comparable to controls (Figures 2G,H), and phosphate and calcium metabolism remained unchanged up to six months of age (Supplementary Table S2).

Mice with cardiac myocyte specific Fgf23 deletion showed normal body length and heart size compared to age-matched controls (Figure 3A). The body weight related organ weights including heart, lung, kidney and liver increased with age, but did not differ between both genotypes (Supplementary Table S3). However, the mortality of Fgf23^fl/fl/cre⁺ mice during the first 6 months of life was slightly but significantly increased compared to controls (Figure 3B). Next, we performed echocardiography and Millar catheter to determine whether cardiac function was affected in Fgf23^fl/fl/cre⁺ mice, which in turn may promote their increased mortality. In echocardiography, no differences between Fgf23^fl/fl/cre⁻ and Fgf23^fl/fl/cre⁺ mice were observed, irrespective of age (Figure 3C, Supplementary Table S4), except for a significantly reduced end-systolic volume (ESV) and slightly lower end-diastolic volume (EDV) in six months old Fgf23^fl/fl/cre⁺ mice compared to control (Figures 3D,E). Using Millar catheter, no differences were found in end-systolic or end-diastolic LV pressure or stroke work (Figures 3F,G). Overall, these data indicate no pathologic cardiac phenotype in Fgf23^fl/fl/cre⁺ mice.

Next, we performed histological analyses to determine alterations in the cardiac structure of Fgf23^fl/fl/cre⁺ and control mice. Quantified by wheat germ agglutinin (WGA) staining and fluorescence microscopy, cardiac myocyte cross-sectional area did not differ between the genotypes (Figures 4A,B). This was in line with the equal expression of the pro-hypertrophic markers atrial natriuretic peptide (ANP), brain natriuretic peptide (BNP) and beta myosin heavy chain (bMHC) in Fgf23^fl/fl/cre⁺ mice compared to controls, irrespective of age (Supplementary Figures S1A–C). Neither the capillary density (Figures 4C,D) nor the accumulation of interstitial collagen fibers (Figures 4E,F) were affected by the conditional cell-specific disruption of Fgf23 in cardiac myocytes over time. The mRNA expression of the pro-fibrotic marker transforming growth factor beta 1 (Tgfb1) and its downstream target connective tissue growth factor (Ctgf) was equal between both groups (Supplementary Figures S1D,E). Consistently, collagen type 1, alpha 1, recently identified as a marker in human heart failure progression related to tissue fibrosis (Hua et al., 2020) and encoded by the gene Col1a1, was not differentially expressed in Fgf23^fl/fl/cre⁺ mice compared to age-matched controls (Supplementary Figure S1F). Thus, in line with normal heart function in echocardiography and Millar catheter, conditional deletion of Fgf23 in cardiac myocytes does not alter cardiac structure.

Enhanced circulating and cardiac FGF23 levels are shown to be associated with LVH in CKD (Faul et al., 2011; Leifheit-Nestler et al., 2016). Furthermore, high plasma FGF23 is present in patients with infarction-related cardiogenic shock and acute decompensated heart failure, which is associated with disease severity and mortality (Pöss et al., 2013; Fuernau et al., 2014; Andersen et al., 2016). It was previously shown that TAC, a model for high blood pressure-induced cardiac hypertrophy and heart failure, induces the expression of cardiac Fgf23 in mice (Slavic et al., 2017). Thus, we next investigated whether the deletion of cardiac myocyte-specific Fgf23 is protective against pathological cardiac remodeling after aortic banding. Two weeks after TAC, cardiac function and geometry were determined by magnetic resonance imaging (MRI) (Figure 5A). TAC significantly enhanced LV mass and impaired ejection fraction (EF) in both genotypes (Figures 5B,C). It increased ESV, decreased stroke volume (SV) and enlarged systolic LV diameter in both groups (Figures 5D–F). Compared to respective TAC Fgf23^fl/fl/cre⁻ controls, heart function seemed to be even more impaired in TAC-operated Fgf23^fl/fl/cre⁺ mice. However, H&E-stained cardiac mid chamber cross-sections revealed increased LV wall thickness after TAC in both groups (Figure 6A). Concomitantly, similar increases in cardiac myocyte size and pro-hypertrophic BNP mRNA expression were seen in Fgf23^fl/fl/cre⁺ mice and controls after TAC (Figures 6B–D). The mean percentage of LV fibrosis after TAC was similar in Fgf23^fl/fl/cre⁺ and Fgf23^fl/fl/cre⁻ mice (Figures 6E,F). Taken together, our findings suggest that specific disruption of Fgf23 in cardiac myocytes does not essentially alter cardiac dysfunction in the TAC mouse model, with the exception of a modest alteration in ESV that however, does not translate to any changes in hypertrophy, fibrosis or chamber remodeling.

Next, we assessed the TAC-induced changes in Fgf23 synthesis. As shown by qPCR analysis, TAC stimulated the expression of Fgf23 in bone of both Fgf23^fl/fl/cre⁺ mice and Fgf23^fl/fl/cre⁻ mice, although the latter did not reach the level of statistical significance in Two-way ANOVA comparisons (Figure 7A). TAC further resulted in increased Fgf23 mRNA levels in Fgf23^fl/fl/cre⁻ mice using total cardiac tissue (Figure 7B). Interestingly, cardiac Fgf23 was also slightly increased after TAC in Fgf23^fl/fl/cre⁺ mice compared to sham, although Fgf23 mRNA levels were significantly lower than in TAC-operated controls. To elucidate the origin of TAC-induced cardiac Fgf23, we next performed co-immunofluorescence staining of Fgf23 with α-Actinin, PDGFRα, and CD31. TAC did not affect the synthesis of Fgf23 in cardiac myocytes (Figure 7C), while PDGFRα co-staining revealed enhanced accumulation of cardiac fibroblasts clearly expressing Fgf23 in both genotypes (Figure 7D), which was in line with severe fibrosis due to TAC as detected by picrosirius red staining (Figures 6E,F). Interestingly, TAC caused an enhanced Fgf23 synthesis in endothelial cells in both genotypes, as shown by increased Fgf23 detection in CD31-positive blood vessels (Figure 7E). The isolation of cardiac myocytes, cardiac fibroblasts and endothelial cells confirmed a specific induction of Fgf23 transcription in cardiac fibroblasts and endothelial cells after TAC compared to sham, while the absolute Fgf23 levels in cardiac myocytes remained unchanged (Figure 7F). Taken together, irrespective of the disruption of cardiac myocyte-specific Fgf23, TAC leads to increased cardiac Fgf23 levels that most likely results from a specific induction of Fgf23 synthesis in cardiac fibroblasts and endothelial cells.