FGF23: A Mature Renal and Cardiovascular Risk Factor?Zoccali C.a · Yilmaz M.I.b · Mallamaci F.a
aNephrology, Hypertension and Renal Transplantation Division and CNR-IBIM Clinical Epidemiology of Renal Diseases and Hypertension, Ospedali Riuniti, Reggio Calabria, Italy; bDepartment of Nephrology, Gulhane School of Medicine, Etlik-Ankara, Turkey
High FGF23 predicts renal function loss in chronic kidney disease (CKD) patients and graft failure in transplant patients. FGF23, parathyroid hormone and serum phosphate are all interrelated but among these CKD-MBD biomarkers only FGF23 is independently related with CKD progression. High FGF23 associates with endothelial dysfunction in CKD patients and in elderly individuals in the general population. Furthermore, independently of serum phosphate, high FGF23 associates with mortality and left ventricular hypertrophy in dialysis patients and with atherosclerosis in elderly individuals in the general population. FGF23 also predicts a high risk for death and cardiovascular events in predialysis CKD patients and in subjects with coronary artery disease. A recent trial in CKD patients showed that low phosphate intake associated with a phosphate binder produces a 35% decrease in plasma FGF23. Yet in this and in another trial testing several phosphate binders, FGF23 levels remained well beyond the upper limit of the normal range. Of note, in this latter study, calcification of the coronary arteries and abdominal aorta actually increased, rather than decreased, during treatment with these drugs in the face of evidence of negative phosphate balance and amelioration of hyperparathyroidism. Mechanistic studies are still needed before testing the hypothesis that FGF23 is implicated in a causal manner in cardiovascular and renal diseases. Given the modest effects of phosphate binders on serum FGF23 in CKD patients, pharmacologic interventions antagonizing the effects of this growth factor rather than phosphate-lowering interventions should be put in place to properly test this hypothesis in the clinical scenario in CKD.
Copyright © 2013 S. Karger AG, Basel
Reduced ability to dispose dietary phosphate load in chronic kidney disease (CKD) is an early consequence of nephron loss . Such an alteration demands complex adaptations aimed at increasing the excretion of filtered phosphate per glomerular filtration rate (GFR) unit (fractional phosphate excretion). These adaptations allow serum phosphate to be maintained in the normal range until a GFR of about 20 ml/min/1.73 m2 . Reduced synthesis of 1,25 vitamin D and secondary hyperparathyroidism are time-honored mechanisms whereby the kidney and the intestine adapt phosphate disposal in response to nephron loss. More recent studies now point to FGF23 as the earliest factor set into motion to increase fractional renal excretion of phosphate in CKD and as the main driver of low 1,25(OH)2vitamin D and high parathyroid hormone (PTH) in these patients  (fig. 1). Although the health implications of diets with a high phosphate content in persons with normal renal function are still unclear, greater dietary intake of this anion in healthy young people is associated with higher serum phosphate levels  as well as with sequential hormonal changes in FGF23 and 1,25(OH)2vitamin D levels directionally similar to those described in CKD patients . Maintaining phosphate balance in CKD patients, and perhaps also in individuals with a high dietary phosphate intake, imposes several important health trade-offs in various animal species and in man, a tenet supported by the observation that serum phosphate levels in mammals show a remarkable association with life span . Well beyond CKD, phosphate levels within the normal range was a strong predictor of cardiovascular complications in three fairly large population-based cohort studies [7,8,9].
|Fig. 1. Relationship between fractional phosphate excretion (solid line), FGF23 (broken line) and GFR [redrawn from  ].|
Several mechanisms may underlie the excess risk for cardiovascular and renal disease by phosphate excess. In endothelial cells in vitro, high phosphate concentration increases reactive oxygen species production and decreases nitric oxide (NO) synthesis via inhibitory phosphorylation of endothelial NO synthase . This effect is not confined to in vitro models because the physiological hyperphosphatemia induced by a high phosphate meal impairs NO-dependent vasodilatation in the forearm of healthy subjects . Disturbed NO-dependent vasodilatation of the renal microcirculation is a hallmark of hypertension , diabetes  and chronic nephropathies , and this alteration is almost completely reversed by angiotensin-converting enzyme (ACE) inhibition [11,14]. The relevance of FGF23 in this hypothetic pathway is suggested by the fact that stimulation of the FGF receptor by another member of FGF family (FGF-2) increases the synthesis of ACE in human smooth muscle cells . High FGF23 per se  or the ensuing decrease in 1,25(OH)2vitamin D  impairs endothelium-dependent vasodilatation in CKD patients and predicts progression to end-stage renal disease (ESRD)  as well. Furthermore, FGF23 via 1,25(OH)2vitamin D suppression may in theory activate the renin-angiotensin system , a crucial mechanism in cardiovascular diseases and chronic nephropathies.
The relationship between phosphate and the risk for progression to ESRD in humans has been tested in several observational studies. Serum phosphate predicted the incidence of combined renal endpoints both in male US veterans with stage 3-4 CKD  and in patients of the African American Study of Hypertension and Kidney Disease (AASK) study cohort , a population with stage 3-4 CKD secondary to hypertension and nephrosclerosis. In the PREPARE study , which included only stage 4 CKD patients, phosphate was directly related with the slope of the GFR over time but no renal survival analyses were performed in this study. In a post hoc analysis of the REIN study , including CKD patients with proteinuric nephropathies of various severity from stage 2-5, serum phosphate predicted both progression to ESRD and a composite end-point including ESRD and creatinine doubling. These associations remained strong and largely independent of the GFR as measured by a state-of-the-art method like plasma iothalamate clearance, suggesting that it is unlikely that the predictive power of phosphate for these outcomes merely depends on its relationship with the GFR. The relationship between phosphate and renal outcomes resisted to adjustment for powerful risk factors for renal disease progression like proteinuria and arterial pressure implying that such a link may be causal in nature. In numerical terms, the magnitude of the risk for ESRD and/or creatinine doubling by 1 mg/dl was of the same order of that portended by a 5-g/24 h increase in proteinuria or by a 20-mm Hg increase in systolic pressure, and serum phosphate represented the strongest predictors of adverse renal outcomes. Overall, these data offer additional relevant support to the hypothesis that phosphate is causally implicated in the risk of renal disease progression.
The most stimulating new finding in this study is the observation that across the normal range of serum phosphate concentration relatively higher levels of this anion attenuated to an important extent the beneficial effect of ACE inhibition on renal disease progression. Such a phenomenon suggests that the interference of phosphate upon renal disease progression lies in a critical pathway conducive to renal damage. As alluded to before, in smooth muscle cells in vitro, stimulation of the receptor common to all members of the FGF family by a compound cognate to FGF23, i.e. FGF-2, increases the gene expression level and the gene product of ACE . Then, phosphate may increase ACE expression in tissue via FGF23 and the FGF receptor, and such increase may attenuate the response to standard doses of ACE inhibitors. The strong links between angiotensin II and NO-dependent renal vasodilatation offer another valid rationale to explain the attenuation by phosphate of nephroprotection by ACE inhibition. Due to a greater increase in efferent compared with afferent arteriolar resistance, inhibition of the NO system leads to glomerular hypertension , which is precisely the renal microcirculatory alteration corrected by chronic ACE inhibition . Thus, an interference of phosphate with NO-dependent renal vasodilatation, either by a direct mechanism and/or mediated by high FGF23, appears to be a plausible explanation for the attenuation by phosphate of the nephroprotective effect of ACE inhibition observed in the present study. ACE inhibition ameliorates the impaired renal vasodilatory response to the NO precursor (L-arginine) in patients with uncomplicated hypertension  and restores normal NO levels in patients along with an improvement in albuminuria in patients with IgA nephropathy . It is therefore plausible that such a beneficial effect be less pronounced in patients with relatively higher phosphate levels and more severe degrees of endothelial renal dysfunction by a phosphate-dependent mechanism. Augmented intrarenal activation of the renin-angiotensin system by a phosphate excess-mediated 1,25(OH)2vitamin D suppression is another potential pathway which may explain our findings. Experimental findings showing that nephroprotection by activated vitamin D via a TGF-β signaling pathway is amplified when blood pressure is controlled by ACE inhibition  would support this possibility. However, the issue clearly demands additional specific intervention studies based on diets with various phosphate content in suitable models of CKD.
Overall, these data provide coherent support to the contention that phosphate is a risk factor for CKD in patients with proteinuric nephropathies, and generate the hypothesis that reducing excessive phosphate load by phosphate binders and/or by low phosphate diets may potentiate the renoprotective effect of ACE inhibition in CKD. A remarkable stabilization in renal function in a case series of patients maintained on very low phosphate diet (7 mg/kg) was reported in 1983 by Barsotti et al. . In the 1980s, low-phosphate diets were considered a promising treatment strategy to attenuate calcitriol deficiency and secondary hyperparathyroidism  and progression to ESRD , but interest in the possibility of modifying the evolution of renal diseases by this intervention faded away thereafter. A recent study by Di Iorio et al.  is fully in line with the possibility that low phosphate intake may exert favorable effects on proteinuria in progressive nephropathies.
The study by Zoccali et al.  is a post hoc analysis of a clinical trial and the study by Di Iorio et al.  is an analysis of clinical data collected in regular clinical practice and therefore these studies are inherently limited and do not allow strong conclusions about the risk of phosphate excess and adverse renal outcomes. In these studies, FGF23, 1,25(OH)2vitamin D and PTH were not measured and therefore mechanism(s) of nephroprotection by ACE inhibition and low phosphate intake remain purely speculative. On the other hand, observational studies suggest that phosphate-driven endocrine adaptations to high phosphate load per functioning renal unit may amplify the risk of CKD progression. Indeed, in a study by Fliser et al. , FG23 proved to be a powerful predictor of the risk of progression to end-stage kidney disease. Likewise, high FGF23 predicted renal graft loss in transplant patients in a more recent study by Wolf et al. . In this study, FGF23, PTH and serum phosphate were all interrelated. However, only FGF23 was an independent risk factor for graft loss pointing to this biomarker as a key player in the adverse long-term renal outcomes in transplant patients. Similar results have also been reported in subsequent studies in CKD patients [33,34].
FGF23 and Cardiovascular Risk in CKD
Evidence has been provided that high FGF23 associates with compromised endothelium-dependent vasoregulation in CKD patients  and that endothelial dysfunction in dialysis patients reverts in strict parallelism with FGF23 normalization after renal transplantation . This evidence in humans is fully in line with experiments in heterozygous (kl -/+) mice , a model which mirrors the effects of high FGF23 because low or absent Klotho evokes a sustained counter-regulatory increase in FGF23. Accordingly, in these mice the contractile response to norepinephrine is amplified and the vasodilator responses to acetylcholine attenuated as compared to kl +/+ mice . Increased FGF23 may contribute to the adverse outcomes in CKD and in old people. Indeed, independently of serum phosphate, high FGF23 associates with mortality  and left ventricular hypertrophy  in dialysis patients, and with atherosclerosis in elderly individuals in the general population . FGF23 also predicts a high risk for death [33,34] and cardiovascular events  in predialysis CKD patients and in subjects with coronary artery disease . As mentioned, FGF23 is an important factor in endothelial cell biology . Accordingly, a population-based study showed that high FGF23 is associated with reduced endothelial function (as measured by vasoreactivity in the forearm) in individuals with normal renal function . In a study by Yilmaz et al. , in a large series of stage 3 and 4 CKD patients, most of whom had moderate to severe proteinuria, FGF23 was coherently associated in an inverse fashion with flow-mediated vasodilatation in the forearm both on categorical and noncategorical analyses. An intriguing finding in this study is that FGF23 was directly associated with a biomarker of endothelial dysfunction like ADMA, an endogenous inhibitor of NO synthase. As mentioned, the effects of the Klotho-FGF23 complex in endothelial cells are mediated by the NO system . This biological connection may be of relevance in patients with CKD because in the Yilmaz et al.  study adjustment for ADMA attenuated to an important extent the strength of the FGF23-FMD link, a phenomenon suggesting that FGF23 and ADMA are in the same pathway conducive to vascular dysfunction. Preliminary analyses in a large cohort of stage 3-5 CKD patients confirm that FGF23 and ADMA may be in the same pathway conducive to adverse clinical outcomes .
A recent community study  showed that high FGF23 may have implications well beyond CKD. Indeed, in this study FGF23 levels in the normal to high-normal range were directly associated with incident risk for cardiovascular death independently of classical risk factors (fig. 2). Detailed analyses testing for the confounding effect of the GFR showed that the FGF23 cardiovascular death link was confined to individuals with a GFR below 60 ml/min, suggesting that the risk portended by relatively higher FGF23 is restricted to CKD patients. Of note, in this study individuals with high FGF23, low GFR and albuminuria had the highest cardiovascular risk, emphasizing that these risk factors may exert independent and additive effects on cardiovascular outcomes.
|Fig. 2. Relationship between FGF23 in the normal and high-normal range and the risk of cardiovascular mortality in a community-based cohort study [redrawn from  ].|
FGF23 is an inhibitor of vitamin D hydroxylation and PTH secretion and therefore relatively higher FGF23 may be a surrogate of vitamin D deficiency and hyperparathyroidism, i.e. two CKD-MBD biomarkers commonly altered in the general population, particularly in the elderly. Even though no adjustment for these factors was performed in this study, the association of FGF23 with cardiovascular outcomes was not attenuated by statistical adjustment for PTH and 1,25(OH)2vitamin D in the previously discussed studies [33,38,45]. Thus, it appears likely that the vascular system represents a direct target for FGF23, a tenet in line with vascular function  and echocardiography studies . This intriguing cohort study does not solve the question of causality. FGF23 inhibits bone mineralization  and it may in theory have a similar effect at a vascular level, thus acting as a vasculoprotective factor in individuals with CKD, a hypothesis in line with recent experimental data presented by Lim et al. . However, FGF23 directly promotes myocardiocyte growth in vitro and triggers left ventricular hypertrophy in mice with normal kidney function , an experimental observation which goes along with the association between this growth factor and left ventricular mass in CKD patients  and in elderly subjects in the general population . Thus, we still ignore whether FGF23 is an endogenous toxin for the cardiovascular system or a vasculoprotective factor.
The fundamental question of whether FGF23 can be usefully modified in long-term clinical trials in normophosphatemic CKD, i.e. in the vast majority of stage 3-4 CKD patients, by the use of phosphate binders remains unanswered. A recent trial in this population showed that low phosphate intake associated with a phosphate binder produces a 35% decrease in plasma FGF23. Yet in this trial, like in another recent study by Block et al.  with various phosphate binders, FGF23 levels remained well beyond the upper limit of the normal range . Of note, in Block et al.'s study, calcification of the coronary arteries and abdominal aorta actually increased during treatment with these drugs in the face of evidence of negative phosphate balance and amelioration of hyperparathyroidism. This study had several limitations . Yet the problem remains that the efficacy of phosphate binders as FGF23-lowering agents in normophosphatemic CKD patients remains uncertain at best. Furthermore, this study raises the question of the safety of these drugs, an issue that has been largely overlooked so far. In conclusion, mechanistic studies are still needed before testing the hypothesis that FGF23 is implicated in a causal manner in cardiovascular and renal diseases. Given the modest effects of phosphate binders on serum FGF23 in CKD patients, pharmacologic interventions antagonizing the effects of this growth factor rather than phosphate-lowering interventions should be put in place to properly test this hypothesis in the clinical scenario of CKD.
Unità Operativa di Nefrologia e CNR-IBIM
IT-89124 Reggio Calabria (Italy)
Published online: May 25, 2013
Number of Print Pages : 6
Number of Figures : 2, Number of Tables : 0, Number of References : 52
Vol. 36, No. 1, Year 2013 (Cover Date: June 2013)
Journal Editor: Ronco C. (Vicenza)
ISSN: 0253-5068 (Print), eISSN: 1421-9735 (Online)
For additional information: http://www.karger.com/BPU