Regulation of C-Terminal and Intact FGF-23 by Dietary Phosphate in Men and Women

Authors


  • The authors state that they have no conflict of interest.

Abstract

FGF-23 is a novel regulator of phosphate metabolism. We studied the regulation of FGF-23 by dietary phosphate in 66 men and women using two assays. Dietary phosphate restriction decreased FGF-23 and loading increased FGF-23 significantly. An assay that measured intact FGF-23 showed the effects of dietary phosphate much more clearly than an assay that also measures presumed biologically inactive fragments. Dietary phosphate is a key regulator of circulating FGF-23; choice of assay is critical when studying FGF-23 physiology.

Introduction: Fibroblast growth factor 23 (FGF-23) is a novel phosphaturic factor discovered through genetic studies of patients with renal phosphate wasting disorders. Ablation of the FGF-23 gene in mice reduces renal phosphate excretion and increases serum phosphate, suggesting that FGF-23 is critical for normal phosphate homeostasis. We examined the role of dietary phosphate in the regulation of FGF-23 in humans.

Materials and Methods: Sixty-six healthy males and females were randomized to either phosphate-depleted or -loaded diets for 5 days, after a 4-day run-in diet. FGF-23 was measured using an “intact” assay that only detects intact FGF-23 peptide and with a “C-terminal” assay that measures both intact FGF-23 peptide and presumed biologically inactive carboxyl terminal fragments. The main outcome was the within group change in FGF-23 with either phosphate depletion or loading.

Results: Using the intact FGF-23 assay, mean FGF-23 area under the curve (AUC) decreased by 9 ± 16% with phosphate depletion (p = 0.0041) and increased by 35 ± 29% with loading (p < 0.0001). Using the C-terminal FGF-23 assay, mean FGF-23 AUC decreased by 8 ± 12% with phosphate depletion (p = 0.0003) and increased by 13 ± 20% with loading (p = 0.0016). Increases in FGF-23 with phosphate loading were greater with the intact assay than with the C-terminal assay (p = 0.0003). Using the intact assay only, FGF-23 was significantly associated with serum phosphate (r = 0.39, p < 0.01), 24-h urinary phosphate (r = 0.47, p < 0.01), fractional excretion of phosphate (r = 0.29, p < 0.01), and 1,25-dihydroxyvitamin D (r = −0.30, p < 0.01). The association between the assays was weak (r = 0.26, p < 0.01).

Conclusions: Dietary phosphate is a key regulator of circulating FGF-23 levels in humans. Additionally, choice of assay is critical when performing physiologic investigations of FGF-23.

INTRODUCTION

Phosphate is essential for bone mineralization, muscle function, encoding of genetic material, protein modification, signal transduction, and energy use. The kidney is the main organ regulator of phosphate homeostasis; PTH and 1,25-dihydroxyvitamin D are established hormonal determinants of phosphate homeostasis. In vitro and in vivo data suggest that fibroblast growth factor 23 (FGF-23) is a novel and critical regulator of phosphate metabolism.

Dramatic phosphate wasting can occur in patients with autosomal dominant hypophosphatemic rickets (ADHR), X-linked hypophosphatemia (XLH), and tumor induced osteomalacia (TIO). These patients develop rickets/osteomalacia because of both urinary phosphate wasting that results in hypophosphatemia and inappropriately low serum 1,25-dihydroxyvitamin D levels. These findings are most likely caused by elevated circulating FGF-23 in these conditions that have different underlying molecular defects.(1–6) When FGF-23 levels are restored to normal, as by removal of the responsible tumors in TIO, renal phosphate wasting ceases, and serum phosphate, 1,25-dihydroxyvitamin D, and bone mineralization normalize.(5–8) These observations strongly suggest that elevated FGF-23 levels are responsible for the phosphate wasting in these patients. Furthermore, animal studies show that FGF-23 has powerful effects on phosphate homeostasis. Administration of recombinant FGF-23 to mice or transplantation of cell lines expressing FGF-23 into mice induces phosphaturia, hypophosphatemia, and inappropriately low serum 1,25-dihydroxyvitamin D.(9,10) More importantly, mice in which the FGF-23 gene has been ablated are hyperphosphatemic and have elevated serum 1,25-dihydroxyvitamin D, findings that strongly suggest that FGF-23 plays a role in normal phosphate homeostasis.(11,12)

Other factors such as serum frizzled related protein 4 (sFRP4) and matrix extracellular protein (MEPE) are overexpressed in TIO,(13,14) and both sFRP4 and MEPE are phosphaturic.(15–17) Recent data suggest that MEPE may be overexpressed in XLH.(18) Thus, it is possible that FGF-23, sFRP4, and MEPE are all important for phosphate regulation. Currently, there are commercially available assays to measure FGF-23 only.

To determine the physiological role of dietary phosphate in the regulation of FGF-23 in humans, we assessed the effects of short-term manipulations of dietary phosphate on circulating FGF-23 levels in normal adult volunteers. Moreover, to determine whether the choice of FGF-23 assay affects interpretation of its physiological role, we measured FGF-23 using one assay that detects both biologically active intact FGF-23 peptide and carboxyl terminal fragments that are presumed to be biologically inactive, and a second assay that detects only intact FGF-23 peptide.

MATERIALS AND METHODS

Study subjects

We recruited healthy subjects through advertisements that were approved by our local institutional review board. To be eligible, subjects were required to be 18–45 years old, have a body mass index within 15% of ideal, and be capable of complying with the study diet. Female subjects were required to have regular menses; oral contraceptive use was allowed. Subjects with disorders known to affect phosphate metabolism (e.g., primary hyperparathyroidism or renal insufficiency) or with a history of nephrolithiasis, diabetes mellitus, malabsorption, recent ethanol abuse, or clinically significant cardiac, hepatic, oncologic, or psychiatric disease were excluded. Subjects using medications that are known to affect phosphate metabolism (e.g., high dose vitamin D [>1000 units daily], high dose vitamin A [>20,000 units daily], phosphate binding antacids, calcitonin, calcitriol, etidronate, growth hormone, or anticonvulsants) were also excluded. Subjects were required to have a serum inorganic phosphate level of at least 2.5 mg/dl (0.80 mM), serum 25-hydroxyvitamin D level of at least 15 ng/ml (37 nM),(19) serum PTH of 60 pg/ml or less, serum creatinine of <1.5 mg/dl (133 μM), serum glucose level of <126 mg/dl (7.0 mM), plasma hemoglobin of at least 11 g/dl (110 g/liter), serum aspartate and alanine aminotransferase levels that were less than twice the upper limit of normal, and normal serum levels of thyroid-stimulating hormone and testosterone (males only).

Study protocol

Subjects were randomly assigned by a computerized system to one of two groups. During the first 4 days of the study, both groups consumed a run-in diet that contained 900 mg of phosphate daily. This diet was used to increase the homogeneity of the study participants at the time of the study intervention, given the individual variation in dietary phosphate intake. During the last 5 days of the study, the phosphate-depleted group consumed 500 mg of phosphate daily plus aluminum and magnesium hydroxide to bind dietary phosphate, whereas the phosphate-loaded group consumed 2500 mg of phosphate daily. Randomization was stratified by gender.

During the study, subjects only consumed food and beverages prepared by the General Clinical Research Center (GCRC) dietitians. The 2500-kcal diet was designed to contain stable amounts of nutrients known to affect phosphate metabolism. On days 1–4, the diet contained ∼380 g of carbohydrate, 76 g of fat, 77 g of protein, 1000 mg of calcium, 900 mg of phosphate, and 3300 mg of sodium. On days 5–9, the protein content was decreased for both groups to facilitate the decrease in dietary phosphate in the phosphate-depleted group. To maintain calories, the carbohydrate content was increased. Thus, during the final 5 days, the diet contained ∼440 g of carbohydrate, 77 g of fat, 39 g of protein, 1000 mg of calcium, 500 mg of phosphate, and 3500 mg of sodium. We decreased the dietary protein similarly in both groups on days 5–9 to ensure that the two groups only differed with respect to phosphate intake. Starting on day 5, the phosphate-depleted group received 2.2 g of aluminum and magnesium hydroxide four times a day with meals to decrease absorption of dietary phosphate, and the phosphate-loaded group received phosphate supplements (either Neutra-Phos or Phos-NaK) four times a day with meals such that the total daily phosphate intake was 2500 mg (500 mg from the diet and 2000 mg from the supplements).

Fasting blood samples were collected between 7:00 and 9:00 a.m. on days 1, 4, 5, 6, 7, 8, and 9. Fasting double voided urine samples were collected for calculation of the fractional excretion of phosphate (FePO4) on the same days. Twenty-four-hour urine samples were collected on days 4 and 8 for phosphate, creatinine, sodium, and calcium measurements. Height was measured with a stadiometer, and weight was measured on a calibrated digital scale.

Compliance with the study protocol was assessed by review of the change in 24-h urinary measurements and by daily review of returned meal containers, supplement, and antacid use.

The study was approved by the institutional review board of Partners HealthCare Systems. All subjects provided written informed consent.

Measurement of FGF-23

Plasma or serum FGF-23 was measured using two different assays. Initially, we used a human FGF-23 immunometric assay that measures both intact peptide and carboxyl terminal (C-terminal) fragments (Immutopics, San Clemente, CA, USA). We refer to this as the “C-terminal” FGF-23 assay. This assay has a lower limit of detection of 3 reference units (RU)/ml; however, we used the first nonzero standard of 15 RU/ml as our lower limit of detection. The intra- and interassay CVs were ≤5% and ≤7%, respectively.(6) Using this assay, in 141 healthy individuals, the mean FGF-23 level was 55 ± 50 (SD) RU/ml.(6)

We also measured FGF-23 levels using another immunometric assay that only detects the intact FGF-23 peptide (Kainos Co., Tokyo, Japan). We refer to this as the “intact” FGF-23 assay. This assay has a lower limit of detection of 3 pg/ml; however, we used the first nonzero standard of 10 pg/ml as our lower limit of detection. The intra- and interassay CVs were ≤3% and ≤4%, respectively (Kainos Co.). Using this assay, in 104 healthy individuals, the mean FGF-23 level was 29 ± 11 pg/ml.(5)

All FGF-23 measurements were performed in duplicate in our GCRC core laboratory, and all samples for a given subject were analyzed in the same assay. We used quality controls generated by the GCRC core laboratory as well as the manufacturer's quality controls to assure comparability of the assays.

Other laboratory measurements

Serum PTH was measured with the Nichols Advantage chemiluminescence assay (Nichols Institute Diagnostics, San Clemente, CA, USA). Serum 25-hydroxyvitamin D and 1,25-dihydroxyvitamin D were measured by radioimmunoassay (DiaSorin, Stillwater, MN, USA). Serum and urine phosphates were measured by colorimetric method (Roche Diagnostics, Indianapolis, IN, USA).

Study endpoints

The primary study endpoint was the change in circulating FGF-23 with dietary phosphate depletion or loading. Secondary endpoints included the change in serum phosphate, PTH, 1,25-dihydroxyvitamin D, and FePO4. FePO4 is a measure of phosphate excretion performed on fasting blood and urine samples and calculated using the following formula: FePO4 = {(urine [PO4] × blood [creat])/(blood [PO4] × urine [creat]) × 100}. The normal range for FePO4 is 0–20%.(20)

Statistical analysis

Our primary endpoint was the within-group change in mean FGF-23 area under the curve (AUC). By design, values from days 4 and 5 represent the run-in period and values from days 6–9 represent the intervention period. We compared the net AUC change in FGF-23 during the intervention between the two study groups using Student's unpaired t-test. For the other endpoints, representative values for the run-in and intervention diets were obtained by averaging the measurements on days 4 and 5 for the run-in period and on days 8 and 9 for the intervention. A longitudinal mixed effects models approach was used to examine the effect of the intervention on the longitudinal FGF-23 changes after adjusting for the effects of multiple covariates. The predictor variables in the model were study group, time, study group × time interaction, serum phosphate, PTH, body mass index, age, and gender, where the subject level intercept and regression coefficient of the time effect were random effects and the assumed longitudinal correlation structure was exchangeable. The relationship between the two FGF-23 assays was examined by correlational analysis.

Unpaired t-test was used to determine the effect of gender on serum phosphate, FePO4, and FGF-23, and paired t-test was used to determine the difference in the change in mean FGF-23 AUC between assays for both the phosphate-depleted and -loading groups. All data are expressed as mean ± SD unless specified otherwise. All comparisons were performed by two-sided tests, and resulting p < 0.05 was considered statistically significant. We performed all analyses using SAS V8.2.

RESULTS

Subject characteristics

The baseline characteristics of the 66 subjects who completed the protocol are shown in Table 1. There were no significant differences between groups in baseline characteristics. Male and female subjects had similar measures of creatinine clearance as calculated by the Cockcroft and Gault method (119 ± 20 versus 113 ± 24 ml/minute in the phosphate-depleted group [p = 0.41] and 113 ± 22 versus 109 ± 24 ml/minute in the phosphate-loaded group [p = 0.66]).

Table Table 1.. Demographics of Subjects at Baseline
original image

Adherence to the study protocol and changes in body weight

Another six subjects in the phosphate-loading group did not complete the study protocol: two subjects were intolerant of the diet during the run-in period; three subjects were not compliant with taking the phosphate supplements or the time schedule; and one subject was withdrawn for anemia. Their data were excluded. All subjects randomized to the phosphate depletion group completed the protocol. Body weight varied by ≤5% over the course of the study in 60 of 66 of the subjects.

24-h urinary phosphate, sodium, and calcium

The mean 24-h urinary phosphate excretion was similar in both groups after the dietary run-in period. Mean 24-h urinary phosphate excretion decreased from 500 ± 192 (16 ± 6 mmol) to 79 ± 85 mg (3 ± 3 mmol) with dietary phosphate depletion (p < 0.01) and increased from 447 ± 155 (14 ± 5 mmol) to 1044 ± 318 mg (34 ± 10 mmol) with loading (p < 0.01). Mean 24-h urinary sodium excretion was similar in both groups at baseline and increased in both groups with the intervention from 105 ± 40 to 129 ± 44 mmol with phosphate depletion (p = 0.04) and from 98 ± 35 to 137 ± 36 mmol with loading (p < 0.01). Mean 24-h urinary calcium excretion was similar in both groups at baseline. Mean 24-h urinary calcium excretion increased from 170 ± 62 (4.2 ± 1.5 mmol) to 210 ± 87 mg (5.2 ± 2.2 mmol) with phosphate depletion (p < 0.01) and decreased from 174 ± 102 (4.3 ± 2.5 mmol) to 93 ± 73 mg (2.3 ± 1.8 mmol) with loading (p < 0.01).

Fractional excretion of phosphate and serum phosphate

Mean FePO4 was similar in both groups after the dietary run-in period. Mean FePO4 decreased from 9 ± 4% to 2 ± 3% with dietary phosphate depletion (p < 0.01) and increased from 10 ± 6% to 16 ± 4% with loading (p < 0.01; Fig. 1A). Mean serum phosphate levels decreased from 3.2 ± 0.5 (1.03 ± 0.16 mM) to 2.8 ± 0.6 mg/dl (0.90 ± 0.19 mM) with phosphate depletion (p < 0.01) but did not change with loading (Fig. 1B). There was no difference in the FePO4 between male and female subjects in the phosphate depleted (p = 0.18–0.64) or loaded (p = 0.35–0.72) groups. Similarly, there was no difference in serum phosphate levels between male and female subjects in the phosphate depleted group (p = 0.64–0.90). Female subjects in the phosphate loading group had significantly higher phosphate levels than males on day 9 only (3.7 ± 0.4 [1.19 ± 0.13 mM] versus 3.1 ± 0.5 mg/dL [1.00 ± 0.16 mM], p = 0.01).

Figure Figure 1.

Serial measurements of (A) fractional excretion of phosphate, (B) serum phosphate, (C) serum PTH, and (d) serum calcium. Days 4 and 5 represent the end of the run-in period. The intervention, dietary phosphate depletion (○) or dietary phosphate loading (•), was started after the day 5 samples were collected. Values are presented as mean ± SE. Systeme International conversion factors: phosphate (mM), 0.3229; PTH (ng/liter), 1; calcium (mM), 0.2495.

1,25-dihydroxyvitamin D, PTH, and calcium

Mean serum 1,25-dihyroxyvitamin D levels increased from 49 ± 22 (118 ± 53 pM) to 62 ± 29 pg/ml (149 ± 70 pM) with dietary phosphate depletion (p < 0.01) but did not change with loading (p = 0.16; data not shown). Serum PTH levels did not change with phosphate depletion (p = 0.81) but increased from 35 ± 14 (35 ± 14 ng/liter) to 43 ± 15 pg/ml (43 ± 15 ng/liter) with loading (p < 0.01; Fig. 1C). Blood calcium levels were stable with both interventions (Fig. 1D).

FGF-23

The effect of dietary phosphate manipulation on blood FGF-23 is shown in Figs. 2 and 3: the percent change in FGF-23 in the combined cohort is shown in Fig. 2 and the percent change in FGF-23 for female and male subjects separately is shown in Fig. 3.

Figure Figure 2.

Percent change in FGF-23 during the intervention period for the combined cohort. ○, phosphate depletion intervention as measured with the intact FGF-23 assay; •, phosphate loading intervention as measured with the intact FGF-23 assay. □, phosphate depletion intervention as measured with the C-terminal FGF-23 assay; ▪, phosphate loading intervention as measured with the C-terminal FGF-23 assay. The intervention, dietary phosphate depletion or loading, was started after the day 5 samples were collected. Values are presented as mean ± SE.

Figure Figure 3.

Percent change in FGF-23 during the intervention period in (A) women and (B) men. ○, phosphate depletion intervention as measured with the intact FGF-23 assay; •, phosphate loading intervention as measured with the intact FGF-23 assay. □, phosphate depletion intervention as measured with the C-terminal FGF-23 assay; ▪, phosphate loading intervention as measured with the C-terminal FGF-23 assay. The intervention, dietary phosphate depletion or loading, was started after the day 5 samples were collected. Values are presented as mean ± SE.

Using the C-terminal assay, mean FGF-23 AUC decreased by 8 ± 12% with dietary phosphate depletion (p = 0.0003). Levels decreased maximally by day 7 and returned to baseline by day 9 (Fig. 2). Using the intact assay, mean FGF-23 AUC decreased by 9 ± 16% with dietary phosphate depletion (p = 0.0041); this decrease was sustained throughout the intervention. Dietary phosphate loading resulted in greater changes in circulating FGF-23 than did phosphate depletion. Using the C-terminal assay, mean FGF-23 AUC increased by 13 ± 20% with dietary phosphate loading (p = 0.0016). Using the intact assay, mean FGF-23 AUC increased by 35 ± 29% with dietary phosphate loading (p < 0.0001). The change in FGF-23 with phosphate loading was greater when intact FGF-23 was measured than when C-terminal FGF-23 was measured (p = 0.0003).

As a secondary analysis, we assessed the within-group change in mean FGF-23 AUC in men and women separately (Fig. 3). Using the C-terminal assay, in females phosphate depletion decreased FGF-23 AUC by 7 ± 12% (p = 0.0195) and phosphate loading increased FGF-23 AUC by 12 ± 17% (p = 0.0065). In males, phosphate depletion decreased FGF-23 AUC by 13 ± 7% (p < 0.0001); the increase with phosphate loading (14 ± 23%) was not significant (p = 0.07). There was no effect of gender on the change in FGF-23 across both groups, however (p = 0.8). Using the intact assay, in females phosphate depletion decreased FGF-23 AUC by 11 ± 13% (p = 0.0029) and phosphate loading increased FGF-23 AUC by 35 ± 29% (p < 0.0001). In males, phosphate loading increased FGF-23 AUC by 34 ± 31% (p = 0.0042); the change with phosphate depletion (6 ± 19%) was not significant (p = 0.21). As with the C-terminal assay, using the intact assay there was no effect of gender on the change in FGF-23 across both groups, however (p = 0.64).

The differences in FGF-23 levels between the phosphate depletion and loading groups were significant using either assay. Using the C-terminal assay, the between group difference in mean FGF-23 AUC with dietary phosphate manipulation was 21 ± 16% (p < 0.0001). Using the intact assay, the between group difference in FGF-23 with dietary phosphate manipulation was 43 ± 23% (p < 0.0001). Overall, the association between circulating FGF-23 measured using both assays was weak (r = 0.26, p < 0.0001; Fig. 4).

Figure Figure 4.

Univariate association between FGF-23 measured using the intact FGF-23 assay and the C-terminal FGF-23 assay. The Pearson correlation coefficient, r, is included.

Univariate relationships between the absolute levels of FGF-23 and serum phosphate, FePO4, 24-h urinary phosphate excretion, 1,25-dihydroxyvitamin D, and PTH differed depending on the FGF-23 assay. Using the intact assay, significant associations were detected between FGF-23 and serum phosphate, FePO4, 24-h urinary phosphate, PTH, and 1,25-dihydroxyvitamin D (Table 2). No significant associations were detected using the C-terminal assay.

Table Table 2.. Univariate Associations Between Absolute Levels of FGF-23 and Serum Phosphate, Fractional Excretion of Phosphate, 24-h Urinary Phosphate, Serum 1,25 Dihydroxyvitamin D, Serum PTH, and Serum Calcium for the Combined Groups
original image

We created a multivariate regression model with intact FGF-23 as the response variable, and age, gender, study group, time, a study group × time interaction term, PTH, and serum phosphate as the predictor variables (Table 3). We excluded FePO4 and serum 1,25-dihydroxyvitamin D as predictor variables because FGF-23 changes the FePO4 and the production of 1,25-dihydroxyvitamin D.(10,12,21–23) After adjusting for other covariates, both dietary phosphate (p = 0.0015) and serum phosphate (p = 0.0059) were important determinants of circulating FGF-23; however, age, gender, BMI, and PTH were not predictors of FGF-23.

Table Table 3.. Multivariate Analysis of Factors That Predict FGF-23 Levels
original image

DISCUSSION

In this study, we showed that dietary phosphate loading increases both FePO4 and circulating FGF-23 but has no effect on serum phosphate. In contrast, dietary phosphate deprivation reduces FePO4, serum phosphate, and circulating FGF-23. Notably, the magnitude of change in FGF-23 with dietary phosphate depletion was smaller than with dietary phosphate loading, a finding that suggests that FGF-23 is a hormone that primarily promotes phosphate wasting. Importantly, we showed for the first time that the two currently available assays differ in their ability to detect physiologic changes in FGF-23, and that the correlation between assays is weak. Univariate analysis showed that circulating FGF-23 levels were positively associated with serum phosphate, FePO4, and 24-h urinary phosphate excretion and negatively associated with serum 1,25-dihydroxyvitamin D. Similar to recently published animal studies,(24,25) multivariate analysis showed that circulating FGF-23 levels were predicted by both dietary and serum phosphate. The relationship between PTH and FGF-23 is controversial and deserves special mention. Although there was a weak univariate association between PTH and FGF-23, this association disappeared with multivariate analysis. Finally, in contrast to the findings of a small, earlier study,(26) changes in serum phosphate and FePO4 during phosphate depletion were similar in males and females. Using the C-terminal assay, phosphate depletion had similar effects on men and women separately; phosphate loading increased FGF-23 significantly in women only. Using the intact assay, phosphate loading had similar effects on men and women separately; phosphate depletion decreased FGF-23 significantly in women only. However, when we tested for an effect of gender by multivariate analysis or unpaired t-tests, we found that FGF-23 was regulated by dietary phosphate similarly in men and women.

The kidneys are the main regulators of phosphate homeostasis. Phosphate in the glomerular filtrate is actively reabsorbed by two distinct sodium–phosphate co-transporters (NaPi2a and NaPi2c) in the proximal tubules. Both NaPi2 proteins can be rapidly degraded or newly synthesized to promote phosphate wasting or reabsorption, respectively.(27,28) Changes in fasting serum phosphate with dietary phosphate manipulation are caused mostly by changes in renal tubular clearance, or FePO4, and by a lesser degree to the decrease in phosphate in glomerular filtrate.(29,30) PTH promotes phosphate wasting by causing the internalization of NaPi2a from the cell membrane,(31) and preliminary findings suggest that PTH has a similar effect on NaPi2c.(32) Administration of FGF-23 also induces internalization of NaPi2a and 2c.(21,33) PTH also regulates phosphate by promoting the 1α hydroxylation of 25 hydroxyvitamin D(34); 1,25-dihydroxyvitamin D then promotes reabsorption of phosphate from the gut and release of phosphate from bone.(20) Like PTH, FGF-23 also regulates production of 1,25-dihydroxyvitamin D but FGF-23 inhibits rather than enhances 1α hydroxylation of 25-hydroxyvitamin D.(35) Thus, 1,25-dihydroxyvitamin D levels are lower than expected in subjects who overproduce FGF-23(5–7) and are elevated in mice that lack the FGF-23 gene.(11,12) 1,25-dihydroxyvitamin D increased in subjects who underwent phosphate depletion, as expected, but did not change with phosphate loading. The latter finding may be caused by the increase in PTH, which could offset any anticipated decrease in 1,25-dihydroxyvitamin D because of phosphate loading.

In addition to the animal studies showing the phosphaturic effect of parenterally administered FGF-23, the elevated FGF-23 levels in patients with XLH and TIO, and the resolution of phosphate wasting with removal of tumors in patients with TIO suggest that FGF-23 promotes phosphate wasting. Furthermore, the observations that serum phosphate is elevated in mice lacking the FGF-23 gene and in patients with tumoral calcinosis who have inactivating FGF-23 mutations(36–39) strongly suggest that FGF-23 is central to normal phosphate homeostasis. Importantly, patients with tumoral calcinosis have elevated circulating levels of C-terminal FGF-23 fragments but low levels of intact FGF-23, a finding that shows that the C-terminal fragments of FGF-23 are biologically inactive. The increase in circulating FGF-23 in response to dietary phosphate loading in this study shows that FGF-23 is a key factor in the handling of dietary phosphate loads in normal human physiology. Additionally, the finding that changes in FGF-23 with phosphate loading were larger when assessed with the intact assay than with the C-terminal assay suggests that the intact peptide more accurately reflects physiologic changes in FGF-23 with dietary phosphate loading.

The regulation of FGF-23 by dietary phosphate in humans has been studied previously with conflicting results. In a group of six men, phosphate depletion and loading had no effect on FGF-23 levels.(40) The failure to find an effect of dietary phosphate in this study is likely caused by the small sample size. In a group of 29 men, Ferrari et al.(41) reported that phosphate depletion decreased and phosphate loading increased FGF-23 levels. Finally in eight men, preliminary data suggest that circulating FGF-23 levels increase significantly 8 h after a phosphate load.(42) Although the findings of Ferrari et al.(41) are similar to ours, interpretation of this study is limited by several factors. First, their study used a cross-over design (phosphate depletion followed by loading), and it is unknown whether effects of the first manipulation may have carried over into the second part of the protocol. Second, the study by Ferrari et al.(41) included only men so that the results may not be applicable to women. Our data show that dietary phosphate similarly regulates FGF-23 in both men and women. Our study thus extends their findings to be applicable to women. Third, dietary calcium, protein, carbohydrate, and sodium were not kept constant in the study by Ferrari et al.(41) The consequences of failing to rigorously control intake of other dietary factors that have important effects on phosphate metabolism are unknown. Fourth, and most importantly, FGF-23 was only measured with a C-terminal assay in the study by Ferrari et al.(41) Because the C-terminal assay measures both intact peptide and presumed biologically inactive C-terminal fragments, data from that assay may not accurately reflect physiologic changes in circulating FGF-23 with dietary phosphate manipulation. Although our findings were qualitatively similar to those of Ferrari et al.,(41) our study proves that, in retrospect, C-terminal FGF-23 measurements actually are a reasonable proxy for the biologically active hormone when studying the effects of dietary phosphate on FGF-23 regulation in normal men and women. If the regulation of FGF-23 were limited to studies using assays that measure presumed biologically inactive, C-terminal FGF-23, the true effect of dietary phosphate on the physiological regulation of FGF-23 would remain uncertain.

Because both PTH and FGF-23 seem to regulate blood phosphate levels, it is possible that the effects of either of these hormones on phosphate homeostasis may be caused, at least in part, by changes in the other hormone. In patients with primary hyperparathyroidism, one preliminary study reported significantly higher FGF-23 levels and a decrease after parathyroidectomy,(43) whereas others have reported that FGF-23 levels are similar to controls in such patients.(44–46) In patients with hypoparathyroidism, FGF-23 levels are elevated and correlate with phosphate levels.(47) The persistence of hyperphosphatemia in patients with hypoparathyroidism, despite elevated FGF-23 levels, suggests that both PTH and FGF-23 are needed for phosphate regulation. We found that PTH levels increased slightly with dietary phosphate loading(48) and did not change with phosphate depletion. We did not detect any significant association between PTH and FGF-23, however. The low to normal PTH levels in the hyperphosphatemic FGF-23 gene knockout mouse(11,12) and the lack of correlation between PTH and FGF-23 in this study suggest that FGF-23 affects phosphate regulation independently of PTH.

Multivariate analysis showed that dietary phosphate manipulation is a significant determinant of circulating FGF-23 independent of age, gender, and PTH. Additionally, multivariate analysis indicated that serum phosphate is an independent predictor of FGF-23, distinct from dietary phosphate. Together with the significant univariate associations of FGF-23 with serum phosphate, FePO4, 24-h urinary excretion of phosphate, and serum 1,25-dihydroxyvitamin D, these findings add further support to the hypothesis that FGF-23 is a physiologically important regulator of phosphate homeostasis in humans.

As noted above, our study is the first to compare the two currently available FGF-23 assays with respect to physiologic changes in FGF-23. Similar to the findings of Ito et al.,(49) who measured FGF-23 in patients with XLH, we found that the two assays correlate poorly in the physiologic range. This lack of correlation is not surprising given that the mechanism by which FGF-23 fragments are generated or metabolized is unknown. Nevertheless, both assays have been useful in describing changes in FGF-23 in disease states characterized by abnormal phosphate handling, such as XLH, TIO, or renal failure.(5,6,40) However, in physiologic states, the intact assay elucidates the physiology of FGF-23 more clearly than does the C-terminal assay in several respects. First, the magnitude of both within group and between group changes in FGF-23 with phosphate loading or depletion was greater when FGF-23 was measured using the intact assay. If we express the changes in FGF-23 like the data from Ferrari et al.,(41) phosphate loading increased FGF-23 by 52 ± 55% using the intact assay but by only 29 ± 39% using the C-terminal assay, a value that was very similar to the 31 ± 10% change reported by Ferrari et al.(41) Second, we only saw important associations between FGF-23 and serum phosphate, FeP04, 24-h urinary phosphate, and 1,25-dihydroxyvitamin D with the intact FGF-23 assay. The shown differences in detection of physiologic changes between the two FGF-23 assays are important because our data provide a rationale for the increasing choice of the intact FGF-23 assay, over the C-terminal assay, by many other investigators of FGF-23 physiology.(25,42,50–52) It will be important to consider assay methodologies when interpreting other studies involving FGF-23 measurements because our data suggest that choice of assay is key to the advancement of the understanding of FGF-23 physiology.

Our study has limitations. First, although we had objective measures to assess compliance, it is possible that lack of compliance with the study protocol led to underestimation of the effect of dietary phosphate on FGF-23. Second, whereas we excluded potential subjects whose baseline caloric intake greatly exceeded that which was provided by the study, it is possible that the effect of short-term dietary phosphate loading would be reduced in subjects whose chronic prestudy phosphate intake was very different from that provided in the protocol. Third, we provided a fixed caloric diet, as opposed to providing calories per kilogram body weight. We selected the former diet to ensure that phosphate intake was constant within each group. If we had provided calories per kg body weight changes in FGF-23 may have been confounded by differing phosphate intake between subjects. Fourth, we only assessed the effects of dietary phosphate on FGF-23 levels over a brief period, and it is unknown if these changes will be maintained during a longer observation. Finally, despite the lack of correlation between FGF-23 and PTH, it is still possible that PTH contributes to the response of FGF-23 to dietary phosphate.

In summary, this study shows that manipulation of dietary phosphate alters circulating FGF-23 levels within 48 h of the manipulation and that the two currently available assays differ in their ability to detect physiologic changes and associations of FGF-23. These findings suggest that FGF-23 is a physiologically important phosphate-regulating hormone in humans, particularly in response to dietary phosphate loading, and that choice of assay is critical when performing physiologic studies of FGF-23.

Acknowledgements

We are grateful to the staff of the Mallinckrodt General Clinical Research Center for implementation of the study protocol, and Gregory Neubauer and Carrie Whelan for the performance of the FGF-23 assays. This study was supported by National Institute of Health Grants F32 DK063761 (to SMB), K24 DK 02759 (to JSF), and M01-RR-01066 (to the Mallinckrodt General Clinical Research Center at Massachusetts General Hospital); Massachusetts General Hospital Physician-Scientist Development Award (to SMB); and the Massachusetts General Hospital Endocrine Gift and Bequest fund.

Ancillary