Induction of cardiac FGF23/FGFR4 expression is associated with left ventricular hypertrophy in patients with chronic kidney disease

Abstract

Background

In chronic kidney disease (CKD), serum concentrations of fibroblast growth factor 23 (FGF23) increase progressively as glomerular filtration rate declines, while renal expression of the FGF23 coreceptor Klotho decreases. Elevated circulating FGF23 levels are strongly associated with mortality and with left ventricular hypertrophy (LVH), which is a major cause of cardiovascular death in CKD patients. The cardiac FGF23/FGF receptor (FGFR) system and its role in the development of LVH in humans have not been addressed previously.

Methods

We conducted a retrospective case–control study in 24 deceased patients with childhood-onset end-stage renal disease (dialysis: n = 17; transplanted: n = 7), and 24 age- and sex-matched control subjects. Myocardial autopsy samples of the left ventricle were evaluated for expression of endogenous FGF23, FGFR isoforms, Klotho, calcineurin and nuclear factor of activated T-cells (NFAT) by immunohistochemistry, immunofluorescence microscopy, qRT-PCR and western blotting.

Results

The majority of patients presented with LVH (67%). Human cardiomyocytes express full-length FGF23, and cardiac FGF23 is excessively high in patients with CKD. Enhanced myocardial expression of FGF23 in concert with Klotho deficiency strongly correlates with the presence of LVH. Cardiac FGF23 levels associate with time-averaged serum phosphate levels, up-regulation of FGFR4 and activation of the calcineurin-NFAT signaling pathway, an established mediator of cardiac remodelling and LVH. These changes are detected in patients on dialysis but not in those with a functioning kidney transplant.

Conclusions

Our results indicate a strong association between LVH and enhanced expression levels of FGF23, FGFR4 and calcineurin, activation of NFAT and reduced levels of soluble Klotho in the myocardium of patients with CKD. These alterations are not observed in kidney transplant patients.

INTRODUCTION

Cardiovascular mortality is excessively increased in patients with chronic kidney disease (CKD) compared with the general population [1–5], and left ventricular hypertrophy (LVH) is the most common cardiac abnormality in these patients with prevalence rates as high as 75% [6–10]. Hypertension, high body mass index and hyperparathyroidism were identified as independent predictors of LVH in CKD [11]. Fibroblast growth factor 23 (FGF23) is a bone-derived phosphaturic hormone that reduces the renal synthesis of 1,25-dihydroxy-vitamin D [1,25(OH)₂D₃] [12]. In CKD, circulating FGF23 levels increase progressively as glomerular filtration rate (GFR) and the renal capacity for phosphate excretion decline leading to hyperphosphataemia, 1,25(OH)₂D₃ deficiency, secondary hyperparathyroidism and potentially Klotho deficiency [13–15]. In end-stage renal disease (ESRD), FGF23 reaches levels that can be 1000-fold above the normal range [16–20]. In addition, elevated FGF23 levels have been associated with LVH, arterial stiffness and cardiovascular mortality in CKD patients before and after renal transplantation (KTx) [1, 6, 21–23].

FGF23 acts via binding to a receptor complex consisting of FGF receptors (FGFR) and the coreceptor α-Klotho (referred to as Klotho) [24, 25]. The Klotho protein exists in a soluble form, which can arise either from alternative splicing of Klotho mRNA or from ectodomain shedding of membrane-bound Klotho [26]. Reduced levels of circulating Klotho in rodents caused by genetic disruption or advanced CKD are associated with cardiac hypertrophy and dysfunction [27, 28]. As Klotho is not expressed in neonatal rat ventricular myocytes or murine hearts, and FGF23 fails to activate mitogen-activated protein kinase (MAPK) signalling in isolated cardiomyocytes [8], direct effects of FGF23 on the heart have not been expected. However, we and others have shown that FGF23 induces hypertrophy in cultured cardiomyocytes in the absence of Klotho, via phosphoinositide-specific phospholipase C gamma (PLCγ) dependent activation of the calcineurin–nuclear factor of activated T-cells (NFAT) pathway [8, 29]. In mice, systemic and intramyocardial administration of FGF23 lead to the development of LVH [8, 29, 30]. Moreover, as we have recently shown, FGF23 exclusively activates FGFR4 on cultured cardiomyocytes to stimulate PLCγ/calcineurin/NFAT signalling independent of Klotho, and mice lacking FGFR4 do not develop LVH in response to elevated FGF23 [31]. However, the cardiac FGF23/FGFR4 signalling system in the human heart and its potential role in the development of LVH in CKD patients have not been addressed so far.

Here, we performed a comprehensive expression analysis of the cardiac FGF23/FGFR4 system and we studied its association with cardiac remodelling and LVH in a cohort of deceased patients with childhood-onset ESRD. Characterizing the molecular components of FGF23-induced cardiac injury should help to develop targeting therapies that aim to protect the heart from undergoing pathological changes in patients with CKD.

MATERIALS AND METHODS

Detailed methodology is provided in Supplementary data.

Human study population and ethical standards

This retrospective case–control study included 24 deceased CKD patients, who received renal replacement therapy (RRT) before the age of 15 years, as well as 24 age- and sex-matched controls. It was performed in compliance with the Declaration of Helsinki and its later amendments, and has been approved by the ethical committee of the Hannover Medical School. All patients and controls underwent detailed autopsy. Exclusion criteria for control subjects were the presence of heart or kidney disorders. LVH was diagnosed by postmortem autopsy analysis and histological evaluation of cardiomyocyte cross-sectional area. Sixteen and eight patients were classified as LVH-positive (LVH⁺) and LVH-negative (LVH⁻), respectively. In addition, patients were classified according to RRT at the time of death (dialysis versus KTx) (Table 1). Nine out of 17 patients in the dialysis group previously underwent KTx.

Table 1.

Patients characteristic by mode of RRT at time of death

Characteristics	Mode of RRT at time of death		P-value
	Dialysis (n = 17)	KTx (n = 7)
Age at death (years)	10.9 ± 9.3	12.5 ± 6.7	0.69
Age at ESRD (years)	6.7 ± 5.4	6.6 ± 3.2	0.93
Duration of RRT (years)	4.4 ± 6.1	5.9 ± 4.5	0.14
Cumulative time on dialysis (years)	3.0 ± 4.6	0.9 ± 0.8	0.42
Cumulative time on dialysis (%)	99.5 (66, 100)	8 (2, 35)	<0.001
Time-averaged
Serum calcium (mmol/L)	2.22 ± 0.48	2.27 ± 0.16	0.79
Serum phosphate (mmol/L)	1.99 (1.40, 2.74)	1.74 (1.20, 4.13)	<0.05
iPTH (pg/mL)	896 (267, 1308)	275 (128, 778)	0.22
At time of death
Serum creatinine (µmol/L)	558 ± 407	250 ± 176	<0.05
eGFRa (mL/min/1.73 m2)	5	50.4 ± 48.5	<0.0001
Serum calcium (mmol/L)	2.07 (1.54, 2.30)	2.20 (2.15, 2.50)	0.26
Serum phosphate (mmol/L)	2.08 (1.70, 3.73)	1.47 (1.16, 3.53)	0.74
iPTH (pg/mL)	482 (81, 1967)	95 (38, 306)	0.10

Values are presented as mean ± SD, or median (interquartile range).

ESRD, end-stage renal disease; PTH, parathyroid hormone.

^aThe estimated glomerular filtration rate (eGFR) of kidney transplanted (KTx) patients was calculated by Schwartz formula, and dialysis patients were classified as eGFR of 5 mL/min/1.73 m².

Histological analysis, morphometry and quantification of protein expression and localization

Human formalin-fixed paraffin-embedded (FFPE) myocardial autopsy samples of the left ventricle were stained with hematoxylin and eosin (HE) and wheat germ agglutinin (WGA) to quantify cardiomyocyte cross-sectional area. Results were given as fold increase of cross-sectional area from patients compared with respective age and sex-matched controls. For immunohistochemistry, human myocardial autopsy sections were stained for FGF23, Klotho, FGFR1, FGFR4 or NFAT. The percentage protein expression per image was measured, and fold increase of protein in patients compared with respective age- and sex-matched controls were calculated. For immunofluorescence analysis, human sections were incubated with mouse monoclonal anti-sarcomeric α-actinin and rabbit polyclonal anti-FGF23 antibodies, followed by incubation with secondary antibodies coupled to Alexa488 or Atto555.

RNA isolation and qRT-PCR analysis

RNeasy FFPE Kit (Qiagen) was used to prepare total RNA from each human myocardial autopsy sample according to the manufacturer's protocol. For RNA isolation of snap-frozen human tissue, RNeasy Mini Kit (Qiagen) was used. For all samples, cDNA was generated using QuantiTect Reverse Transcription Kit (Qiagen), and qRT-PCR was performed in triplicates using QuantiFAST SYBR Green PCR Kit including ROX dye (Qiagen). Relative gene expression values were calculated with the 2^−ΔΔCt method [32] using β-actin (ACTB) as housekeeping gene.

Protein isolation and immunoblotting

For protein extraction from snap-frozen human autopsy samples, tissue was homogenized in RIPA extraction buffer, and total protein was analysed by SDS-PAGE followed by immunoblotting. Membranes were incubated with primary antibodies against FGF23, FGFR1, FGFR4, Klotho or GAPDH followed by incubation with IRDye^® secondary antibodies (LI-COR Biosciences). The Odyssey Imager (LI-COR Biosciences) was used for protein detection and quantification.

Statistical analysis

Data are presented as mean ± SD if not indicated otherwise. Comparison between patients and matched controls were done by paired t-test in the case of normal distributed data or the Wilcoxon matched-pairs signed-rank test in the case of non-Gaussian distributions (GraphPad Prism Software version 6.0). Correlation analysis was done by a two-sided Pearson's or Spearman's rank correlation after testing for Gaussian distribution with Kolmogorov–Smirnov test (SPSS Software version 22/23). The following variables were assessed for possible associations with cardiomyocyte cross-sectional area, cardiac FGF23, FGFR4 and Klotho % area: sex, age, primary renal disease (congenital CKD versus others), age at onset and duration of CKD, mode of RRT (dialysis versus KTx), standard deviation score values for mean blood pressure (BP), height and weight, time-averaged values for eGFR, calcium, phosphorus, parathyroid hormone (PTH), c-reactive protein, mean cumulative intake of calcium-containing phosphate binders and calcitriol prescribed during the total time of therapy, and the mean number of BP medications prescribed per day. Variables associated with P < 0.1 on univariate analysis were entered into the multivariable regression model. Two-tailed P-values of <0.05 were considered statistically significant.

RESULTS

LVH is associated with age, time on RRT and dialysis treatment in CKD patients

The majority of patients (67%) with childhood-onset ESRD presented with LVH at the time of death. Patients diagnosed with LVH are significantly older, and spent longer time on RRT and dialysis treatment, respectively (Table 2). Mean standardized height and weight are low but relative heart weight, cardiomyocyte cross-sectional area and expression of brain natriuretic peptide (BNP), an established marker of pathological cardiac hypertrophy [33], are significantly higher in the whole-patient cohort, and the LVH⁺ group (Table 3 and Figure 1A–C and E). Furthermore, cardiomyocyte cross-sectional area correlates positively with duration of ESRD (Figure 1D).

Table 2.

Characteristics of patients with ESRD according to presence of LVH

Characteristics	All patients (n = 24)	Patients with LVH (n = 16)	Patients without LVH (n = 8)	P-value
Age at death (years)	11.4 ± 8.5	14.1 ± 8.0	6.5 ± 7.4	<0.05
Age at ESRD (years)	6.7 ± 4.8	7.8 ± 4.2	4.8 ± 5.4	0.09
Duration of RRT (years)	4.9 ± 5.6	6.3 ± 5.7	2.0 ± 4.3	<0.01
Cum. time on dialysis (years)	2.5 ± 4.2	3.3 ± 4.8	0.6 ± 0.8	<0.05
Cum. time on dialysis (%)	76 (14, 100)	58.3 ± 38.6	68.6 ± 47.2	0.43
Time-averaged
Systolic BP (SDS)	1.76 ± 1.90	2.13 ± 1.99	0.98 ± 1.53	0.19
Diastolic BP (SDS)	1.43 ± 1.74	1.54 ± 1.87	1.18 ± 1.53	0.66
Serum calcium (mmol/L)	2.23 ± 0.41	2.16 ± 0.32	2.36 ± 0.54	0.25
Serum phosphate (mmol/L)	1.92 (1.32, 3.06)	2.10 ± 1.11	2.62 ± 1.38	0.38
iPTH (pg/mL)	482 (211, 1087)	811 (198, 1198)	269 (155, 1003)	0.51
Medications
Calcium-containing phosphate binders (g)a	591 ± 589	814 ± 644	341 ± 326	0.28
Calcitriol (µg)a	451 ± 447	548 ± 492	387 ± 304	0.43
No. of antihypertensives/day	2.2 ± 1.3	2.5 ± 1.5	1.5 ± 0.6	0.18
ACE inhibitors (yes/no)	5/19	5/11	0/8	0.22

Values are presented as mean ± SD, or median (interquartile range).

RRT, renal replacement therapy; BP, blood pressure; PTH, parathyroid hormone; ACE, angiotensin converting enzyme; SDS, standard deviation score.

^aMean cumulative intake.

Table 3.

Characteristics of patients and age- and sex-matched controls

Characteristics	All patients (n = 24)	Controls (n = 24)	P-value
Male/female (n)	15/9	15/9
Age at death (years)	11.4 ± 8.5	11.3 ± 8.2	0.97
Height (SDS)	−2.75 ± 1.79	−0.66 ± 1.61	<0.001
Weight (SDS)	−3.07 ± 2.73	−0.54 ± 1.78	<0.001
BMI (SDS)	−1.35 ± 3.54	−0.24 ± 1.42	0.19
Relative heart weight (%)	0.71 ± 0.28	0.48 ± 0.07	<0.01

Values are presented as mean ± SD. SDS, standard deviation score; BMI, body mass index.

FIGURE 1:

Patients with CKD develop LVH. (A) The heart weight/body weight ratio of each patient in relation to its respective control is significantly higher. (B) Representative sections from the left ventricle stained with HE or WGA are shown (both magnification, ×20; scale bar, 100 μm). (C) Quantification of cardiomyocyte cross-sectional area of patients in relation to respective controls (n = 100 cells per section) demonstrate increased surface area of individual cardiomyocytes. (D) Correlation of cardiomyocyte cross-sectional area with duration of ESRD. (E) Quantification of BNP mRNA expression in the myocardium of patients in relation to matched controls. Values are presented as mean ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001.

Full-length FGF23 is expressed in human cardiomyocytes, and cardiac FGF23 levels are elevated in CKD patients with LVH

Next, we performed immunoblot analysis of snap-frozen heart autopsy samples of healthy subjects to investigate whether FGF23 is expressed in human heart tissue. FGF23 is detected in human cardiac tissue lysates as a protein band co-migrating with a positive control for intact FGF23, i.e. plasma from a 16-year-old patient with CKD Stage 5d (Figure 2A). Human recombinant intact FGF23 protein, fused at its C-terminus to a 6 His-tag, serves as a second positive control for the specificity of the used anti-FGF23 antibody. Consistent with the protein band and known proteolytic cleavage of FGF23 [12], cardiac FGF23 comprises the full-length form of FGF23. The cleaved inactive N- and C-terminal fragments of ∼16 and ∼10 kDa, respectively, are not detected. Immunofluorescence co-staining for FGF23 and sarcomeric α-actinin revealed predominant expression of FGF23 in cardiomyocytes (Figure 2B). Both, cardiac FGF23 mRNA and protein levels are higher in the whole-patient cohort, and associate significantly with LVH (Figure 2C–E). Most notably, high cardiac FGF23 levels correlate with enlarged cardiomyocyte cross-sectional area, and with the induction of BNP expression (Figure 2F and G) indicating that local FGF23 expression is related to cardiac hypertrophy.

FIGURE 2:

Cardiac FGF23 is expressed in human left ventricular myocardium, and correlates with LVH in CKD patients. (A) Representative immunoblots for cardiac FGF23 protein and GAPDH (loading control) in myocardial tissue of healthy subjects. Plasma and recombinant human FGF23 (rhFGF23) reveals full-length O-glycosylated species, whereas cleaved FGF23 products are not detected. (B) Immunofluorescence co-staining for FGF23 and sarcomeric α-actinin demonstrating predominantly expression of FGF23 in human cardiomyocytes (magnification, ×63; scale bar, 50 µm). (C) Cardiac FGF23 mRNA expression in myocardium of patients in relation to matched controls is elevated, and associates with LVH. (D) Representative immunohistochemistries of human myocardial sections stained for FGF23 (brown) with hematoxylin counterstain (blue) (magnification, ×63; scale bar, 50 µm). (E) Quantification of FGF23% area in human myocardial tissue is enhanced in the LVH⁺ group. (F) Correlation of cardiac FGF23 with cardiomyocyte cross-sectional area, and (G) with the expression of BNP, indicating that local FGF23 expression is related to cardiac hypertrophy. Values are presented as mean ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001.

In CKD patients, LVH correlates with a reduction of cardiac Klotho levels, and an increase in cardiac FGFR4 expression

To study FGF23 signalling in human heart tissue, gene and protein expression of FGFRs and Klotho were analysed in snap-frozen heart autopsy samples of healthy subjects. In line with our previous studies in rodents [8], mainly FGFR1 and FGFR4 are detected in heart autopsy samples. In contrast, KLOTHO mRNA is absent from the heart (Figure 3A). Interestingly, with kidney protein lysates used as positive control, we could detect both forms of soluble Klotho, i.e. cleaved (∼120 kDa) and secreted (∼65 kDa) Klotho protein [26], in human myocardial tissue by immunoblot analysis (Figure 3B). These findings indicate that Klotho is not expressed in the human heart and that the detected soluble Klotho protein most likely derives from the circulation.

FIGURE 3:

FGFR1 is not regulated in human myocardial tissue of CKD patients independent of LVH. (A) Quantitative RT-PCR analysis, calculated according to the 2^−ΔCt method with normalization to β-actin (ACTB), reveals predominant mRNA expression of FGFR1 and FGFR4, but not KLOTHO (n.d., not detectable) in the left ventricle of human myocardial tissue of healthy subjects. (B) Immunoblots of lysates from human myocardial autopsy samples confirm protein expression of FGFR1 and FGFR4, and demonstrates clearly the presence of Klotho protein, with kidney protein lysates as positive control. (C) Quantitative RT-PCR analysis of FGFR1 mRNA expression in heart tissue of our patient cohort is unaffected in the presence of CKD or LVH. (D) Representative immunohistochemistries of human myocardial sections and aorta as positive control stained for FGFR1 (brown) with hematoxylin counterstain (blue) (magnification, ×63; scale bar, 50 µm). (E) Quantification of FGFR1% area of human myocardial tissue demonstrates no changes between patients and controls. Values are presented as mean ± SEM.

Next, we investigated the expression of FGFR1, FGFR4 and Klotho in myocardial autopsy samples of CKD patients. FGFR1 is the most abundant FGFR isoform in myocardial tissue independent of the presence of CKD or LVH (Figure 3C–E). The amount of Klotho in cardiac tissue is significantly lower in LVH⁺ CKD patients as demonstrated by immunohistochemistry (Figure 4A and B). Human kidney sections serve as a positive control for antibody specificities. Moreover, the decline in Klotho correlates with the duration of ESRD, and cardiomyocyte cross-sectional area (Figure 4C and D). In contrast, cardiac expression of FGFR4 is enhanced in CKD, and associates with LVH (Figure 4F). Subsequent immunohistochemical analysis confirmed the up-regulation of FGFR4 in the myocardium of LVH⁺ CKD patients (Figure 4E), with liver tissue used as positive control. In addition, cardiac FGFR4 expression levels correlate positively with cardiomyocyte cross-sectional area (Figure 4G). These data indicate that mainly cardiac FGFR4 expression is regulated in CKD, where enhanced FGFR4 levels coincide with cardiac hypertrophy.

FIGURE 4:

LVH correlates with declining Klotho, and induction of cardiac FGFR4 expression in CKD patients. (A) Detection of Klotho (brown) with hematoxylin counterstain (blue) in cardiac tissue sections of patients and controls by immunohistochemistry, with human kidney sections as positive control (magnification, ×63; scale bar, 50 µm). (B) Quantification of Klotho % area demonstrates an association of Klotho deficiency with LVH in CKD. (C) Fold increase of Klotho % area correlates negatively with duration of ESRD, and (D) with cardiomyocytes cross-sectional area in CKD patients. (E) Representative immunohistochemistries of human myocardial sections stained for FGFR4 (brown) with hematoxylin counterstain (blue), and human liver tissue as positive control (magnification, ×63; scale bar, 50 µm). (F) FGFR4 mRNA expression is induced in the myocardium of CKD patients in relation to matched controls, and is higher in individuals with LVH. (G) Cardiac FGFR4 correlates positively with cardiomyocyte cross-sectional area in CKD. Values are presented as mean ± SEM; *P < 0.05, ***P < 0.001, ****P < 0.0001.

The pro-hypertrophic calcineurin-NFAT signalling pathway is activated in the myocardium of CKD patients with LVH

It is known that PLCγ-induced elevation of cytoplasmic Ca²⁺-levels activates the protein phosphatase calcineurin. After activation, calcineurin dephosphorylates members of the NFAT family in the cytoplasm. Once dephosphorylated, NFAT translocates into the nucleus and triggers expression of target genes that are involved in the development of cardiac hypertrophy, including BNP [34].

As we have previously shown in neonatal rat cardiomyocytes, FGF23-mediated hypertrophy can be blocked by pharmacological inhibition of PLCγ and calcineurin [8]. Therefore, we determined whether the pro-hypertrophic calcineurin–NFAT signalling pathway is activated in myocardial tissue of CKD patients. In LVH⁺ CKD patients, mRNA levels of the calcineurin subunit B (CNB), and NFAT are induced, while no changes in calcineurin–NFAT expression are observed in LVH⁻ patients (Figure 5A and B). The expression of CNB correlates positively with the duration of CKD (R = 0.579; P < 0.05; data not shown). Furthermore, immunohistochemical analysis of myocardial sections reveals nuclear localization of the activated NFAT only in LVH⁺ CKD patients (Figure 5C).

FIGURE 5:

The calcineurin–NFAT-associated signalling pathway and pro-hypertrophic gene programs are induced in LVH⁺ CKD patients. (A) mRNA expression of the regulatory subunit of calcineurin (CNB) is up-regulated in myocardial autopsy samples of LVH⁺ CKD patients. (B) Likewise, the mRNA of NFAT is higher in LVH⁺ group. (C) Representative immunohistochemistries of human myocardial sections stained for NFAT (brown) with hematoxylin counterstain (blue) (magnification, ×63; scale bar, 50 µm). Note nuclear localization of NFAT protein in LVH⁺ but not in LVH⁻ CKD patients. (D) Indicating cardiac remodelling, the ratio of ACTA to ACTC mRNA is significantly higher in the LVH⁺ group. Values are presented as mean ± SEM; *P < 0.05, **P < 0.01.

To further characterize cardiac remodelling and isoform switches leading to pathologic cardiac hypertrophy, mRNA expression of cytoskeletal markers, cardiac (ACTC) and skeletal (ACTA) α-actin, were investigated [35]. A significantly higher ACTA/ACTC ratio is observed in LVH⁺ CKD patients suggesting FGF23-mediated induction of cardiac remodelling (Figure 5D).

Cardiac FGF23 expression correlates with time-averaged serum phosphate levels, cardiac FGFR4 levels and renal function in CKD patients, and is low after kidney transplantation

Next, we addressed potential predictors of cardiac FGF23 expression in our patient cohort. While cardiac FGF23 levels correlate positively with time-averaged serum phosphate levels in CKD patients (Figure 6A), no correlation is observed for calcitriol dosage and for intake of calcium-containing phosphate binders (data not shown). Most importantly, cardiac expression of FGF23 correlates positively with high expression levels of FGFR4 (Figure 6B), and negatively with eGFR (Figure 6C) indicating a potentially functional interaction between FGF23 and FGFR4 in cardiac tissue of patients with CKD.

FIGURE 6:

Cardiac FGF23 mRNA expression correlates with time-averaged serum phosphate, and FGFR4 expression in patients with CKD, and is low after kidney transplantation. (A) Cardiac FGF23 mRNA correlates positively with time-averaged serum phosphate levels, and (B) FGFR4 expression, and (C) cardiac FGF23 correlates negatively with eGFR in the whole-patient cohort. (D) Cardiac FGF23 mRNA expression is high in dialysis but not in transplant patients. (E) Representative immunohistochemistries of human myocardial sections stained for FGF23 (brown) with hematoxylin counterstain (blue) (magnification, ×63; scale bar, 50 µm), and quantification of FGF23% area demonstrates higher FGF23 protein in dialysis but not in transplant patients. (F) Cardiac FGFR4 mRNA expression is elevated significantly in dialysis but not in transplant patients. (G) Representative immunohistochemistries of human myocardial sections stained for Klotho (brown) with hematoxylin counterstain (blue) (magnification, ×63; scale bar, 50 µm). Klotho protein is significantly lower in dialysis patients, but not in KTx patients. (H) In myocardial autopsy samples, mRNA expression of calcineurin subunits B (CNB) is up-regulated in dialysis patients compared with respective controls, and slightly down-regulates after KTx. (I) The expression of the transcription factor NFAT is up-regulated significantly in patients on dialysis, and reduced in KTx patients. (J) Myocardial hypertrophy is induced in dialysis patients as demonstrated by marked up-regulated BNP mRNA expression, whereas no significant difference is observed in KTx group. Values are presented as mean ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Therefore, we performed subgroup analyses relative to the mode of RRT. Cardiac levels of FGF23 mRNA and protein are higher in patients on dialysis but lower in KTx patients (Figure 6D and E). Likewise, cardiac FGFR4 expression is significantly induced in CKD patients on dialysis but not with functioning renal transplants (Figure 6F). Moreover, the amount of Klotho protein in cardiac tissue is reduced in dialysed patients (Figure 6G). Of note, Klotho levels in myocardial tissue of KTx patients do not significantly differ from controls. Analysis of the calcineurin–NFAT signalling pathway reveals an up-regulation of CNB and NFAT expression levels predominantly in dialysis but not in KTx patients (Figure 6H and I). Furthermore, BNP expression is significantly higher in dialysis but not in KTx patients (Figure 6J). These results suggest that renal transplantation causes a reduction of enhanced cardiac FGF23 and FGFR4 expression as well as a normalization of Klotho protein levels following the down-regulation of calcineurin–NFAT signalling, and consequently reversal of cardiac remodelling and LVH in CKD patients.

Expression analysis of the cardiac FGF23/FGFR4/Klotho system provides independent predictors of myocardial hypertrophy in CKD patients

A multivariable linear regression analysis reveals cardiac FGF23 expression and the amount of Klotho protein in cardiac tissue as independent predictors of cardiomyocyte cross-sectional area (R² = 0.465; P = 0.009; Table 4). The mode of RRT at time of death is the only independent predictor of cardiac FGF23 (R² = 0.351; P = 0.026). The amount of Klotho protein in cardiac tissue of CKD patients correlates with the duration of ESRD, and the mode of RRT at time of death (R² = 0.459; P = 0.002). Finally, cardiac FGFR4 expression correlates with eGFR, mode of RRT at time of death, cumulative time spent on dialysis and cardiac FGF23 expression levels (R² = 0.742; P = 0.003).

Table 4.

Multivariable linear regression analysis

Constant	Predictor	β coefficient	P-value	Cum. R2
Cardiomyocyte cross-sectional areaa				0.465
	Cardiac FGF23a	0.583	0.008
	Klotho % areaa	−0.434	0.038
Cardiac FGF23a				0.351
	RRT at time of death	0.593	0.026
Klotho % areaa				0.459
	Duration of ESRD (years)	−0.483	0.007
	RRT at time of death	0.418	0.017
Cardiac FGFR4a				0.742
	eGFR at time of death	−0.793	0.003
	RRT at time of death	−0.630	0.018
	Cum. time on dialysis (years)	0.437	0.022
	Cardiac FGF23a	0.385	0.049

^aLog-transformed fold increase.

FGF, fibroblast growth factor; FGFR, FGF receptor; RRT, renal replacement therapy (dialysis versus KTx); ESRD, end-stage renal disease; eGFR, estimated glomerular filtration rate.

DISCUSSION

By investigating endogenous FGF23 in myocardial tissue, we demonstrate that FGF23 is expressed in human cardiomyocytes and that myocardial expression of FGF23 in concert with Klotho deficiency strongly correlates with the presence of LVH in CKD patients. Together, both factors explain 47% of the overall variability of cardiomyocyte hypertrophy in our patient cohort. These findings indicate that in CKD, enhanced cardiac FGF23 expression is associated with time-averaged serum phosphate levels, up-regulation of cardiac FGFR4 expression, and activation of calcineurin–NFAT signalling, an established inducer of cardiac remodelling and LVH. Furthermore, these molecular changes are not observed in patients with a functioning kidney transplant.

Major stimuli for the elevation of circulating FGF23 levels are phosphate load [36], calcitriol [37] and PTH [38]. In this study, time-averaged serum phosphate levels reflecting chronic phosphate load are associated with cardiac FGF23 expression. Moreover, in a recent study by Hu et al. [27] phosphate was identified as a modulator of cardiac remodelling in Klotho deficient mice. In CKD patients, phosphate and FGF23 are proposed to represent distinct cardiovascular risk factors supporting vascular calcification and LVH, respectively [39]. While calcification is not detected in myocardial tissue of our patient cohort (data not shown), FGF23-mediated LVH strongly associates with Klotho deficiency.

In previous studies of CKD patients, elevated serum FGF23 concentrations were independently associated with increased ventricular mass index and LVH [6, 40]. In the present study, the majority of CKD patients (67%) presented with LVH, and cardiac FGF23 as well as FGFR4 expression levels and the amount of Klotho in myocardial tissue correlate significantly with LVH and BNP, a biomarker of ventricular stress and hypertrophy [41]. Our data suggest that rising levels of cardiac FGF23 may induce LVH via paracrine mechanisms, which stimulate FGFR4 resulting in activation of the calcineurin–NFAT signalling pathway that causes LVH in the context of Klotho deficiency, i.e. CKD. However, the respective contributions of systemic versus cardiac FGF23 in the development of LVH remain to be elucidated.

The membrane-bound form of Klotho is expressed primarily in distal renal tubules, parathyroid glands and choroid plexus, and can be cleaved by ADAM10 and ADAM17 to generate a soluble endocrine-acting factor [42]. Animal studies indicate that the kidney is the principal source of soluble Klotho and that other Klotho-expressing organs are incapable of correcting systemic Klotho deficiency [43]. Accordingly, a swift reduction in serum Klotho levels by 30% was noted in healthy kidney donors after unilateral nephrectomy [44]. In CKD, renal Klotho expression decreases while circulating FGF23 serum levels rise excessively with declining GFR [42]. While FGF23 can suppress Klotho expression in the kidney, soluble Klotho increases FGF23 mRNA and protein levels in osteocytes [42, 45]. Consistent with previous results from animal models [8], we observe that Klotho is not expressed in human myocardial samples of the left ventricle. In contrast, we can clearly detect soluble Klotho in human heart tissue indicating that circulating soluble Klotho can target the heart. Calcitriol treatment and use of angiotensin converting enzyme (ACE) inhibitors were shown to enhance renal Klotho expression in animals with preserved renal function [46]. However, in the present study, no association is observed between soluble Klotho in myocardial samples and calcitriol or ACE inhibitor treatment, which might be at least partly due to advanced stage of CKD, i.e. ESRD. In contrast, duration of ESRD and mode of RRT (dialysis versus KTx) appear to be independent predictors of soluble Klotho levels in the heart of CKD patients.

Smith et al. [42] demonstrate that soluble Klotho selectively modulates FGF23 activity. In vitro, FGF23 elicits phosphorylation and thereby activation of FGFR1 only in the presence of soluble Klotho, supporting the concept of an FGF23/soluble Klotho/FGFR1 interaction. However, in the present study, soluble Klotho is diminished in myocardial tissue of CKD patients, and correlates negatively with LVH. In contrast, FGF23 is able to bind to and activate FGFR4 in the presence [25] and absence [47] of soluble Klotho. We recently reported that in cultured cardiomyocytes, FGF23 activates FGFR4 independent of Klotho mediating pro-hypertrophic effects, and that treatment of isolated cardiomyocytes with a FGFR4-specific blocking antibody inhibits FGF23-induced hypertrophy [31]. Our analysis of myocardial tissue from CKD patients indicates that FGFR4 is up-regulated, and that this effect is accompanied by a reduction of soluble Klotho levels and associated with cardiac hypertrophy. In contrast, FGFR1 is unaffected in our patient cohort, further indicating that in CKD, enhanced FGF23 levels induce LVH via FGFR4 activation in a Klotho-independent manner. The potentially anti-hypertrophic effects of soluble Klotho in the myocardium require further elucidation.

Molkentin et al. [33, 48] have described several pathways for pathological hypertrophy of the heart, which are classically associated with cardiac remodelling and the induction of fetal gene programs. Reactivation of the fetal gene programs includes the induction of ANP, BNP, ACTA and β-MHC transcription [49]. Cardiac expression of calcineurin or Nfatc4 in respective transgenic mice is sufficient to elicit a pathologic cardiac growth response, and is also indicator for cardiac remodelling [48, 50]. Touchberry et al. [29] discovered that BNP, β-MHC and ACTA are elevated in Col4a3^−/− mice, an animal model of CKD with elevated FGF23 serum levels. We also detected elevation of BNP and β-MHC expression in mice after intravenous injections of recombinant FGF23 protein [30]. In our patient cohort, we observe a similar rise in cardiac mRNA levels of BNP and ACTA, and an increase in FGF23 and calcineurin expression and NFAT activation in myocardial tissue of LVH⁺ CKD patients but not in LVH⁻ patients. Interestingly, CNB expression was strongly elevated in CKD patients with LVH. In combination with experimental studies conducted in rodents [8, 29, 51], the present analysis suggests a causal role of the calcineurin–NFAT signalling pathway in FGF23-induced uraemic cardiac hypertrophy in humans.

Of note, the activation of the calcineurin–NFAT signalling cascade observed in dialysis patients is not observed in KTx patients. In most [52] but not all studies [53], a rapid decline in circulating FGF23 serum levels was observed in CKD patients after KTx. In view of the strong association between cardiac FGF23 and FGFR4 expression, and low protein levels of soluble Klotho in LVH⁺ CKD patients, our data suggest that the reported increase in cardiovascular mortality in dialysis patients compared with KTx patients [3] may reflect an underlying activation of the FGF23/FGFR4-associated pathway promoting cardiac hypertrophy. In a previous study, we have shown in 5/6 nephrectomized rats that pharmacological inhibition of all FGFR-isoforms using a pan-inhibitor protects from the development of LVH [8]. Di Marco et al. [54] recently demonstrate that blocking FGFRs in the same animal model of CKD reverses established LVH if treatment starts early in the progression of CKD, i.e. 2 weeks after surgery. In contrast, pharmacological blockade of FGF23 6 weeks after 5/6 nephrectomy, in an advanced stage of CKD, has no impact on the incidence of LVH [55]. This discrepancy suggests that the formation of cardiac hypertrophy is treatable only in early stages of CKD to prevent uraemia-associated cardiovascular mortality. Our data implicate that the mode of RRT in patients with childhood-onset ESRD strongly affects cardiac FGF23/FGFR4 activation and modulation of LVH, and further support the need for early or preemptive kidney transplantation, particularly in paediatric CKD patients.

A limitation of this study is its retrospective design and its descriptive nature. Consequently, our conclusions are based solely on correlative data. The strengths of our study include the use of case-controlled ESRD patients with age- and sex-matched controls, and the correlation of clinical and molecular data. This approach enabled us to provide several novel aspects regarding the modulatory effects of FGF23/FGFR4 signalling and Klotho deficiency on LVH in patients with ESRD.

CONFLICT OF INTEREST STATEMENT

None declared.

(See related article by Gutierrez. Connecting the dots on fibroblast growth factor 23 and left ventricular hypertrophy. Nephrol Dial Transplant 2016; 31: 1031–1033)