The plasma free fraction of 25-hydroxyvitamin D₃ is not strongly associated with 25-hydroxyvitamin D₃ clearance in kidney disease patients and controls

Abstract

Circulating 25-hydroxyvitamin D [25(OH)D] concentration is used to monitor vitamin D status. Plasma protein binding may influence the 25(OH)D dose-response to vitamin D treatment through a direct relationship between the plasma unbound (”free”) fraction and clearance of 25(OH)D. We previously evaluated 25(OH)D₃ clearance in relation to kidney function using intravenous administration of deuterium labeled 25(OH)D₃. In this follow up study, we determined the free fraction of 25(OH)D₃ in plasma (i.e., percent free 25(OH)D₃) and the serum concentration and haplotype of vitamin D binding protein in these participants. We hypothesized that the percent free 25(OH)D₃ would be positively associated with 25(OH)D₃ clearance and would mediate associations between clearance and vitamin D binding protein (GC) haplotypes. Participants were mean (SD) age 64 (10) years and included 42 individuals with normal kidney function (controls), 24 individuals with chronic kidney disease, and 19 individuals with kidney failure on hemodialysis. Free plasma 25(OH)D₂ and 25(OH)D₃ concentrations were quantified with a new liquid chromatography-tandem mass spectrometry (LC-MS/MS) method. Because there is no reference measurement procedure for free 25(OH)D, we compared the new method with a widely-used predictive equation and a commercial immunoassay. The percent free 25(OH)D₃ determined by predictive equation was weakly associated with 25(OH)D₃ clearance (R = 0.27; P = 0.01). However, this association was absent when percent free 25(OH)D₃ was determined using LC-MS/MS-measured free and total 25(OH)D₃ concentrations. Method comparison uncovered a negative bias in immunoassay-measured free 25(OH)D concentrations among participants with kidney failure, so immunoassay results were not used to evaluate the association between percent free 25(OH)D₃ and clearance. GC2 haplotype carriage was associated with 25(OH)D₃ clearance. Among individuals with 2 relative to no GC2 alleles, clearance was 87 (95% CI: 15 to 158) mL/d greater. However, in contrast with the literature, GC2 carriage was not significantly related to DBP concentration or the percent free 25(OH)D₃ (either predicted or measured). In conclusion, the free fraction of 25(OH)D₃ is not strongly associated with 25(OH)D₃ clearance but may explain small differences in clearance according to GC haplotype.

1. INTRODUCTION

Circulating 25-hydroxyvitamin D [25(OH)D] concentration is used to monitor vitamin D status and is largely a function of exposure to vitamin D. Another determinant of 25(OH)D concentration is clearance, which refers to the pharmacokinetic measurement of the volume of blood from which 25(OH)D is completely removed per unit time. The contributory processes of 25(OH)D clearance are all metabolic occurring in the kidney, liver, and possibly other organs [1, 2].

More than 99% of circulating 25(OH)D is bound to plasma proteins, mainly to vitamin D binding protein (DBP) and less so to albumin and lipoproteins. The remaining 25(OH)D (< 1%) is present in circulation in unbound (“free”) form and may be the fraction available to enter cells, which lack an endocytic receptor for DBP [3]. An important exception is proximal tubular cells of the kidneys, which reabsorb filtered 25(OH)D bound to DBP, a process that is important but not critical to vitamin D sufficiency if vitamin D supply is plentiful [4].

According to the well-stirred model, unbound clearance of slowly metabolized molecules like 25(OH)D depends upon the plasma unbound fraction and intrinsic clearance rather than rate of organ blood flow [5, 6]. Differences in plasma protein binding should theoretically influence clearance of 25(OH)D, particularly hepatic and other plasma clearance pathways thought to eliminate unbound 25(OH)D, but empirical evidence of this is limited. Complete loss of DBP in mice accelerates clearance of 25(OH)D and plasma appearance of 25(OH)D metabolites [7]. Jones at al. studied the associations of serum DBP concentration and DBP genotype (the GC gene) with 25(OH)D kinetics in 36 healthy men [8]. In that study, both DBP concentration and GC polymorphisms were shown to be related to 25(OH)D half-life. The latter association may be explained by genetic differences in either or both the DBP concentration and 25(OH)D-DBP binding affinity as plasma binding protein concentration and affinity, together, determine the plasma unbound fraction. Of interest, GC variants are associated with circulating 25(OH)D concentration in genome-wide association studies [9–12], possibly owing to genetic differences in DBP concentration [4, 13] and/or binding affinity [4].

Kidney disease can disrupt plasma protein binding leading to increased drug clearance [14]. However, we recently characterized 25(OH)D₃ clearance using gold-standard pharmacokinetic methods and showed that kidney disease is a state of reduced clearance stemming from impaired renal metabolism of 25(OH)D₃, albeit clearance was still substantial among individuals with kidney failure, which underscores the contribution of non-renal pathways to overall clearance [15]. In this follow up study, we directly measured the free plasma 25(OH)D₃ concentration and quantified the associations of the percent free 25(OH)D₃ and GC haplotypes with 25(OH)D₃ clearance. We hypothesized the percent free 25(OH)D₃ would be positively associated with clearance and would explain differences in clearance according to GC haplotype. As a secondary objective, we evaluated the free fraction and concentration of 25(OH)D in plasma and the association between percent free 25(OH)D₃ and clearance according to kidney disease status.

2. MATERIALS AND METHODS

2.1. Study Participants

The Clearance of 25-hydroxyvitamin D in Chronic Kidney Disease (CLEAR) study (ClinicalTrials.gov identifier: NCT02937350) evaluated clearance of 25-hydroxyvitamin D₃ according to kidney function and race [15]. CLEAR included 87 adults of self-described Black or White race with a wide range of kidney function who were recruited from the Seattle metropolitan area. Participants were excluded if they were using vitamin D₂ supplements, 1,25(OH)₂D₃ or an analogue, or vitamin D₃ supplements > 400 IU/d. Other exclusion criteria were reported previously [15]. The CLEAR study was approved by the Institutional Review Board at the University of Washington, and all participants provided informed consent.

2.2. Pharmacokinetic Study Design

Blood was collected at baseline in plasma EDTA and serum tubes. Then participants received a single intravenous dose of hexadeuterated 25(OH)D₃ (d₆-25(OH)D₃). The dosage was scaled to participant body size to achieve a peak d₆-25(OH)D₃ near 5 ng/mL [15]. Blood was collected at 5 minutes, 4 hours, and 1, 4, 7, 14, 21, 28, 42, and 56 days post-dosing for determination of the d₆-25(OH)D₃ concentration-time curves. Total (protein-bound plus free) 25(OH)D₂ and total (protein-bound plus free) 25(OH)D₃ concentrations were quantified in plasma collected at baseline (prior to isotope administration) using liquid-liquid extraction (LLE) and LC-MS/MS as described previously [16]. The total 25(OH)D₂ and total 25(OH)D₃ concentrations were summed to obtain total (protein-bound plus free) 25(OH)D concentration. The d₆-25(OH)D₃ concentration in each post-dosing serum sample was quantified using LLE and LC-MS/MS methodology described previously [16], with the exception that trideuterated 25(OH)D₃ (Medical Isotopes Inc, Pelham, NH) was used as the internal standard. Analytical variability of the assays is shown in Supplementary Table 1.

2.3. Determination of 25-hydroxyvitamin D₃ Clearance

Serum concentration-time profiles of d₆-25(OH)D₃ were analyzed using non-compartmental analysis with Phoenix WinNonlin software (version 8.0.0). Area under the concentration-time curve (AUC_0–∞) was calculated using the trapezoidal rule to calculate AUC_0-last (to the last observed timepoint) and extrapolated to time = infinity. The systemic clearance of d₆-25(OH)D₃ was calculated as dose/AUC_0–∞.

2.4. Determination of Free (Unbound) Plasma 25(OH)D Concentration and Percent Free

Free (unbound) plasma 25(OH)D concentration and percent free 25(OH)D at baseline were determined with 3 different approaches. The first was a new LC-MS/MS method (developed at the ARUP Institute for Clinical and Experimental Pathology) that quantified the free 25(OH)D₂ and free 25(OH)D₃ concentrations in plasma [17]. The free 25(OH)D LC-MS/MS method and its performance characteristics are detailed in the Supplementary Methods. In brief, free 25OHD₂ and free 25OHD₃ were extracted from 100 μL of plasma and separated from bound 25(OH)D using Zeba desalting 96-well plates (Thermo Scientific, Waltham, MA). Free 25(OH)D₂ and free 25(OH)D₃ were eluted from the adsorbent using diisopropyl ether containing internal standards ¹³C₃-25OHD₂ and d₃-25OHD₃ (IsoSciences, King of Prussia, PA). Solvent was evaporated, the free 25(OH)D_2, free 25(OH)D_3, and internal standards were derivatized with Ampliflex Diene reagent (AB Sciex, Framingham, MA), and the samples were analyzed using LC-MS/MS. The percent free 25(OH)D₃ based on LC-MS/MS-measured values was calculated as the ratio of LC-MS/MS-measured free 25(OH)D₃ concentration to LC-MS/MS-measured total 25(OH)D₃ concentration multiplied by 100. To compare the new LC-MS/MS method with established methods for determination of free 25(OH)D, LC-MS/MS-measured free 25(OH)D₂ and free 25(OH)D₃ concentrations were summed to obtain the free 25(OH)D concentration. The percent free 25(OH)D was then calculated as the ratio of LC-MS/MS-measured free 25(OH)D concentration to LC-MS/MS-measured total 25(OH)D concentration multiplied by 100.

The second approach was a commercial immunoassay (DIAsource ImmunoAssays SA, Louvain-la-Neuve, Belgium) that quantified the free plasma 25(OH)D concentration (hereafter referred to as the immunoassay-measured free plasma 25(OH)D concentration) [18].The percent free 25(OH)D based on immunoassay-measured values was then calculated as the ratio of immunoassay-measured free 25(OH)D concentration to LC-MS/MS-measured total 25(OH)D concentration multiplied by 100. This immunoassay does not separately quantify free 25(OH)D₂ and free 25(OH)D₃ concentrations, so results were reserved for method comparison and were not used to test hypotheses surrounding 25(OH)D₃ clearance. The interassay CV of the immunoassay in the performing laboratory is 14.2% at concentrations from 3.1–16.5 pg/mL.

The third approach predicted the percent free 25(OH)D using an equation developed by Bikle et al. [19], which relies on measured serum DBP concentration [DBP] and albumin concentration [ALB] and published 25(OH)D₃ binding constants, which are 7 × 10⁸ M⁻¹ and 6 × 10⁵ M⁻¹ for DBP and albumin, respectively:

$$\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{d}\mathrm{p}\mathrm{e}\mathrm{r}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{f}\mathrm{r}\mathrm{e}\mathrm{e}25(\mathrm{O}\mathrm{H})\mathrm{D}=\frac{1}{\left(1+\left({\mathrm{K}}_{\mathrm{a}\mathrm{A}\mathrm{L}\mathrm{B}}\cdot \left[\mathrm{A}\mathrm{L}\mathrm{B}\right]\right)+\left({\mathrm{K}}_{\mathrm{a}\text{DBP}}\cdot \left[\mathrm{D}\mathrm{B}\mathrm{P}\right]\right)\right)}$$

The free 25(OH)D₃ concentration at baseline was then predicted by multiplying the predicted percent free by the LC-MS/MS-measured total 25(OH)D₃ concentration. To enable comparison with the immunoassay, we used the same equation and the LC-MS/MS-measured total 25(OH)D concentration (sum of total 25(OH)D₂ and total 25(OH)D₃ concentrations) to predict the free 25(OH)D concentration. This approach has been used widely in prior research despite recognition that the above binding constants were obtained from experiments with ³H-25(OH)D₃ [19]. To our knowledge, equivalent binding constants for 25(OH)D₂ have not been published.

2.5. Determination of Serum Vitamin D Binding Protein Concentration and GC Haplotype

The DBP concentration was quantified in baseline serum samples using LC-MS/MS as described previously [20]. Total imprecision of the assay is 7.3–9.0% CV at approximately 250 mg/L. The LC-MS/MS method was also used to identify the presence/absence of the 3 common isoforms of DBP (GC1f, GC1s, and GC2) according to quantitation of isoform-specific peptides [20]. These isoforms reflect the genotypes at chromosomal positions chr4:71752617 and chr4:71752606 (GRCh38.p13) within the GC gene. Carrying the reference allele at both positions results in GC1f haplotype/isoform. SNP rs7041 (A>C) at chr4:71752617 results in substitution of glutamate for aspartate at amino acid residue 416 of the mature protein, which confers GC1s haplotype/isoform in place of GC1f. SNP rs4588 (G>T) at chr4:71752606 results in substitution of lysine for threonine at residue 420, which confers GC2 haplotype/isoform in place of GC1f. These SNPs have not been shown to co-occur, so the genotypes at these chromosomal positions occur in 1 of 3 possible haplotypes (GC1f, GC1s, and GC2) inherited to yield 1 of 6 possible diplotypes. The LC-MS/MS method identifies the presence and absence of the haplotypes with ≥ 98% and ≥ 97% accuracy, respectively, compared with genotyping [20]. Participant GC haplotype/diplotype was inferred from our LC-MS/MS data to facilitate comparison of study results with results of large genetic studies that have analyzed the associations of rs7041, rs4588, and consequent GC haplotype with vitamin D status, vitamin D metabolism, and health outcomes.

2.6. Statistical Analysis

The 3 methods used to determine free plasma 25(OH)D concentrations (LC-MS/MS, predictive equation, and immunoassay) were compared using Pearson’s correlation coefficients and Deming regression assuming equal variance across methods. The statistical comparison of methods was based upon formal comparison of the slopes of the Deming regression lines. To examine whether the association between predicted and LC-MS/MS-measured free 25(OH)D concentrations varied according to GC diplotype, we used linear regression of the predicted free 25(OH)D concentration on the independent variables LC-MS/MS-measured free 25(OH)D concentration, GC diplotype, and the interaction of the two. Statistical significance of a Wald test of the interaction terms was taken as evidence of differing association by GC diplotype. The same approach was used to examine whether total and free 25(OH)D concentrations varied according to kidney disease status (control, CKD, and kidney failure).

The distributions of free plasma 25(OH)D concentration and percent free 25(OH)D according to kidney disease status were assessed for normality using the Shapiro-Wilk test and then compared using ANOVA with Tukey’s post hoc tests or, in the case of skewed data, Kruskal-Wallis with Dunn’s post hoc tests. We evaluated associations of 25(OH)D₃ clearance with the serum DBP concentration and the percent free 25(OH)D₃ using linear regression, Pearson’s correlation coefficients, and partial correlation coefficients. The partial correlations were adjusted for kidney disease status and race, which were identified as significant determinants of clearance in the CLEAR study. Of note, baseline total plasma 25(OH)D₃ concentration, BMI, age, sex, and diabetes status were not associated with clearance in the CLEAR study and were not evaluated further in this follow up study [15].

Serum DBP concentration, percent free 25(OH)D₃, and 25(OH)D₃ clearance according to GC1s haplotype carriage (0, 1, or 2 alleles) and GC2 haplotype carriage (0, 1, or 2 alleles) were evaluated using linear regression models adjusted for kidney disease status and race. Regression diagnostics indicated assumptions were met for all regression models. In accordance with the primary analyses of the parent study, we excluded from analyses two participants whose 25(OH)D₃ clearances were extreme outliers (919 and 801 mL/d). All analyses were performed using R software (version 4.2.1) within RStudio software (version 1.1.5033). A 2-sided alpha level of 0.05 was used to indicate statistical significance.

3. RESULTS

3.1. Participant Characteristics

The CLEAR study population included 42 adult controls with normal kidney function, 24 adults with chronic kidney disease, and 19 adults with kidney failure receiving maintenance hemodialysis (Table 1). Self-identified race was either White (69%) or Black (31%), and the population was 59% male and 41% female. Mean total plasma 25(OH)D concentration at baseline was 24 ng/mL overall and did not vary according to kidney disease status (Table 1). However, mean total 25(OH)D₂ was higher in the kidney failure group, despite supplementation with vitamin D₂ being a participant exclusion criterion (Table 1). One-third of study participants had prevalent diabetes. Baseline characteristics according to kidney disease status were reported previously in greater detail [15].

Table 1.

CLEAR participant characteristics

	Overall (n=85)	Control (n=42)	CKD (n=24)	Kidney failure (n=19)
Age, mean (SD)	64 (10)	65 (10)	67 (11)	59 (9)
Gender
Female, n (%)	35 (41)	23 (55)	8 (33)	4 (21)
Male, n (%)	50 (59)	19 (45)	16 (67)	15 (79)
Self-reported race
Black, n (%)	26 (31)	12 (29)	8 (33)	6 (32)
White, n (%)	59 (69)	30 (71)	16 (67)	13 (68)
Diabetes, n (%)	28 (33)	7 (17)	13 (54)	8 (42)
BMI (kg/m2), mean (SD)	29 (6)	29 (6)	29 (7)	30 (7)
eGFR (mL/min/1.73 m2), mean (SD)	-	85 (14)	42 (13)	N/Aa
UACR (mg/g), median (IQR)	-	3 (0–12)	28 (7–178)	N/Aa
Total plasma 25(OH)D (ng/mL), mean (SD)	24 (8)	23 (8)	25 (9)	25 (9)
Total 25(OH)D3 (ng/mL), mean (SD)	22 (8)	22 (8)	24 (9)	21 (9)
Total 25(OH)D2 (ng/mL), mean (SD)	1.4 (3.1)	0.7 (1.0)	0.8 (0.9)	3.6 (6.0)
Plasma 1,25(OH)2D (pg/mL), mean (SD)	34.3 (21.7)	47.6 (16.7)	32.3 (16.2)	7.4 (5.7)
Plasma 24,25(OH)2D3 (ng/mL), mean (SD)	0.9 (0.7)	1.3 (0.8)	0.8 (0.6)	0.3 (0.2)
Serum albumin (g/dL), mean (SD)	4.0 (0.3)	4.1 (0.3)	4.0 (0.3)	4.0 (0.3)
Serum DBP (mg/L), mean (SD)	222 (34)	216 (30)	217 (30)	240 (40)
Serum PTH (pg/mL), median (IQR)	73 (48–169)	60 (43–70)	89 (48–156)	474 (305–745)
Serum FGF23 (pg/mL), median (IQR)	76 (57–204)	59 (53–71)	106 (75–140)	2796 (786–8727)
Serum calcium (mg/dL), mean (SD)	9.1 (0.4)	9.2 (0.3)	9.2 (0.3)	9.0 (0.6)
25(OH)D3 clearance (mL/d), mean (SD)	312 (92)	350 (85)	313 (86)	228 (52)

^aParticipants with KF had little to no urine output and a median dialysis vintage of 2 years. Abbreviations: BMI, body mass index; CKD, chronic kidney disease; CLEAR, Clearance of 25-hydroxyvitamin D in Chronic Kidney Disease Study; DBP, vitamin D binding protein; eGFR, estimated glomerular filtration rate; FGF23, fibroblast growth factor 23; N/A, not applicable; 1,25(OH)₂D, 1,25-dihydroxyvitamin D; PTH, parathyroid hormone; 25(OH)D, 25-hydroxyvitamin D; 24,25(OH)₂D, 24,25-dihydroxyvitamin D.

3.2. Free plasma 25(OH)D method comparison

Compared with predicted concentrations, free 25(OH)D concentrations measured by immunoassay were systematically lower, whereas free 25(OH)D concentrations measured by LC-MS/MS were systematically higher (Supplementary Figure 1). There was a strong correlation between LC-MS/MS-measured and predicted free 25(OH)D concentrations that was slightly stronger when comparing free 25(OH)D₃ concentrations (Figure 1). Immunoassay-measured free 25(OH)D and predicted free 25(OH)D concentrations were also highly correlated (R = 0.82; P < 0.001), whereas the correlation between immunoassay-measured and LC-MS/MS-measured concentrations was weaker (R = 0.61; P < 0.001). However, these associations with immunoassay-measured free 25(OH)D concentrations were attenuated, and the slope of the Deming regression line was significantly weaker among individuals with kidney failure, suggesting the immunoassay performed differently in this patient group (Figures 2A and 2C). We speculated this may be due to the immunoassay’s lower cross-reactivity with free 25(OH)D₂ relative to free 25(OH)D₃, a known characteristic of this immunoassay [18]. We therefore visually assessed the contribution of free 25(OH)D₂, as measured directly by LC-MS/MS, to the observed bias. As shown in 2B and 2D, the negative bias in immunoassay-measured free 25(OH)D concentrations among individuals with kidney failure did not appear to be due to lower sensitivity to free 25(OH)D₂. Additionally, the associations between predicted and LC-MS/MS-measured free 25(OH)D concentrations did not depend on GC haplotype, which support use of a predictive equation with a generic (rather than isoform-specific) binding constant for 25(OH)D₃ binding to DBP.

Figure 1.

The LC-MS/MS method and predictive equation for free plasma 25(OH)D concentration were compared using Deming regression (lines) and Pearson’s correlation coefficients (R) in the full study population (A) and within each kidney function group (B). “Free 25(OH)D LC-MS/MS” refers to the sum of LC-MS/MS-measured free 25(OH)D₂ and free 25(OH)D₃ concentrations. “Predicted free 25(OH)D” was obtained using a predictive equation and LC-MS/MS-measured total 25(OH)D concentration (sum of total 25(OH)D₂ and total 25(OH)D₃ concentrations). Agreement between methods was slightly improved when the comparison was restricted to free plasma 25(OH)D₃ concentrations (C and D). Abbreviations: CKD, chronic kidney disease; KF, kidney failure; LC-MS/MS, liquid chromatography-tandem mass spectrometry; 25(OH)D, 25-hydroxyvitamin D.

Figure 2.

Free plasma 25(OH)D concentrations determined by predictive equation and immunoassay (A and B) and LC-MS/MS and immunoassay (C and D) were compared using Deming regression (lines) and Pearson’s correlation coefficients (R). Panels B and D are excerpts of the results corresponding to the kidney failure group from panels A and C, respectively. The slope of the Deming regression of immunoassay-measured values on predicted values was lower among participants with kidney failure (0.25; 95% CI: −0.08, 0.49) compared with controls (0.65; 95% CI: 0.56, 0.77) and participants with CKD (0.67; 95% CI: 0.49, 0.87) (A). The slope of the Deming regression of immunoassay-measured values on LC-MS/MS-measured values was lower among participants with kidney failure (0.013; 95% CI: −0.007, 0.047) compared with controls (0.077; 95% CI: 0.059, 0.10) and participants with CKD (0.010; 95% CI: 0.059, 0.14) (C). Free 25(OH)D₂ concentration was measured by LC-MS/MS, and data corresponding to “Free 25(OH)D LC-MS/MS” are the summed concentrations of free 25(OH)D₂ and free 25(OH)D₃. There was no obvious contribution of high free 25(OH)D₂ to the negative bias in immunoassay-measured free 25(OH)D concentrations among individuals with kidney failure (B and D). Abbreviations: CKD, chronic kidney disease; KF, kidney failure; LC-MS/MS, liquid chromatography-tandem mass spectrometry; 25(OH)D, 25-hydroxyvitamin D.

3.3. Association between total and free 25(OH)D according to kidney disease status

We examined whether the relationship between total and free 25(OH)D concentrations varied according to kidney disease status. There was a positive association between total 25(OH)D and predicted free 25(OH)D (R = 0.94; P < 0.001) that did not differ according to kidney disease status (Figure 3A). Likewise, there was a positive association between total 25(OH)D and LC-MS/MS-measured free 25(OH)D (R = 0.85; P < 0.001) that did not differ according to kidney disease status (Figure 3B). By contrast, the association between total 25(OH)D and immunoassay-measured free 25(OH)D was attenuated among participants with kidney failure (P-for-interaction < 0.001; Figure 3C). As a result, the percent free 25(OH)D, when determined by immunoassay-measured concentrations of free 25(OH)D, was significantly lower among the kidney failure group (Supplementary Figure 2). Similar to total 25(OH)D concentration (Table 1), the distribution of free 25(OH)D concentration did not differ according to kidney disease status (Supplementary Figure 2).

Figure 3.

LC-MS/MS-measured total plasma 25(OH)D concentration was compared with free plasma 25(OH)D concentration as determined by predictive equation (A), LC-MS/MS (B), and immunoassay (C) among healthy controls (squares), participants with chronic kidney disease (CKD) (circles), and participants with kidney failure (KF) (triangles) using linear regression (lines) and Pearson’s correlation coefficients (R). The association between free and total 25(OH)D did not vary according to kidney disease status except when free 25(OH)D was determined by immunoassay in that the association was attenuated in the kidney failure group (C). Abbreviations: CKD, chronic kidney disease; KF, kidney failure; LC-MS/MS, liquid chromatography-tandem mass spectrometry; 25(OH)D, 25-hydroxyvitamin D.

3.4. Association between percent free 25(OH)D₃ and clearance

25(OH)D₃ clearance was weakly, inversely associated with serum DBP concentration and weakly, positively associated with the predicted percent free 25(OH)D₃ (Figures 4A and 4C). By contrast, there was no association between 25(OH)D₃ clearance and the percent free 25(OH)D₃ based on LC-MS/MS-measured free 25(OH)D₃ concentrations (Figure 4E). These relationships did not vary according to kidney disease status, although small sample sizes may have limited our capacity to detect interaction (Figures 4B, 4D, and 4F). Associations were slightly attenuated by adjustment for kidney disease status and race (Figures 4A, 4C, and 4E).

Figure 4.

Comparisons of 25(OH)D₃ clearance with serum DBP concentration (A and B), the predicted percent free 25(OH)D₃ (C and D), and the percent free 25(OH)D₃ based on LC-MS/MS-measured concentrations (E and F). The associations were evaluated with linear regression (lines), Pearson’s correlation coefficients (R), and partial correlation coefficients adjusted for kidney disease status and race (R_Partial). The associations did not vary significantly according to kidney disease status (B, D, and F). Abbreviations: CKD, chronic kidney disease; DBP, vitamin D binding protein; KF, kidney failure; LC-MS/MS, liquid chromatography-tandem mass spectrometry; 25(OH)D, 25-hydroxyvitamin D.

3.5. DBP concentration, percent free 25(OH)D₃, and clearance according to GC haplotype

The distribution of DBP haplotypes/diplotypes differed according to race (Supplementary Table 2). Among Black participants, the most common haplotype was GC1f (88% prevalence), followed by GC1s (35%), then GC2 (8%). Among White participants, the most common haplotype was GC1s (69% prevalence), followed by GC2 (47%), then GC1f (37%).

With adjustment for kidney disease status and race, mean 25(OH)D₃ clearance did not differ according to GC1s haplotype carriage (Table 2). Mean serum DBP concentration was significantly greater with each additional GC1s allele, whereas mean percent free 25(OH)D₃ (based on LC-MS/MS-measured values) did not vary according to GC1s carriage (Table 2). Regarding GC2 haplotype, mean clearance increased with each additional GC2 allele and was significantly greater among participants with 2 compared with no GC2 alleles. Mean serum DBP concentration and percent free 25(OH)D₃, however, did not vary significantly according to GC2 haplotype carriage (Table 2).

Table 2.

25(OH)D₃ clearance, serum DBP concentration, and percent free 25(OH)D₃ according to GC1s and GC2 haplotype carriage^a

		25(OH)D3 clearance (mL/d)		Serum DBP (mg/L)		Measured percent free 25(OH)D3 (%)b
	N	Mean (95% CI)	Difference (95% CI)	Mean (95% CI)	Difference (95% CI)	Mean (95% CI)	Difference (95% CI)
No. of GC1s
0	35	306 (274, 338)	Reference	207 (194, 220)	Reference	0.195 (0.177, 0.213)	Reference
1	29	277 (246, 307)	−29 (−69, 10)	228 (215, 240)	21 (5, 37)c	0.192 (0.175, 0.209)	−0.003 (−0.025, 0.019)
2	21	271 (238, 305)	−35 (−79, 9)	231 (217, 244)	24 (6, 42)c	0.191 (0.173, 0.210)	−0.004 (−0.029, 0.021)
No. of GC2
0	55	272 (246, 299)	Reference	226 (215, 238)	Reference	0.192 (0.177, 0.206)	Reference
1	25	285 (254, 315)	12 (−26, 50)	220 (207, 233)	−6 (−23, 10)	0.190 (0.173, 0.207)	−0.001 (−0.023, 0.020)
2	5	359 (291, 427)	87 (15, 158)c	208 (178, 237)	−18 (−50, 13)	0.214 (0.176, 0.252)	0.023 (−0.018, 0.063)

^aValues are grand means (95% CI) adjusted for kidney disease status and race. GC haplotype was inferred from results of an LC-MS/MS method that uses isoform-specific peptides to identify the GC1s, GC1f, and GC2 isoforms of vitamin D binding protein in serum [20].

^bPercent free 25(OH)D₃ based on free 25(OH)D₃ and total 25(OH)D₃ concentrations quantified by LC-MS/MS.

^cP < 0.05 for t-test of regression coefficient testing the difference compared with the reference level.

Abbreviations: DBP, vitamin D binding protein; LC-MS/MS, liquid chromatography-tandem mass spectrometry; No., number; 25(OH)D, 25-hydroxyvitamin D.

4. DISCUSSION

In this follow up to the Clearance of 25-hydroxyvitamin D in Chronic Kidney Disease (CLEAR) study, we detected little to no relationship between the free fraction of 25(OH)D₃ in plasma (i.e., percent free 25(OH)D₃) and the clearance of 25(OH)D₃ from the body, across a wide range of kidney function. There was a weak positive association between clearance and the percent free 25(OH)D₃ when percent free was estimated with a predictive equation that relies heavily on measured serum DBP concentration. However, this association was absent when the percent free was determined directly from LC-MS/MS-measured total and free 25(OH)D₃ concentrations. Overall, our data indicate the plasma free fraction of 25(OH)D₃ explains little of the variance in systemic clearance of 25(OH)D₃, which means that the variance not explained by kidney function [15] is largely explained by unbound intrinsic clearance (i.e., metabolic enzyme function) or bound clearance. Nevertheless, vitamin D binding protein (GC) haplotype was associated with clearance, as hypothesized. Our data suggest this may be mediated by small haplotype-related differences in serum DBP concentration, but the variability in our data precludes us from making firm conclusions on this issue.

A study of 36 healthy men who ingested a dose of deuterium labeled (deuterated) 25(OH)D₃ showed that lower serum DBP concentration was associated with shorter 25(OH)D₃ plasma half-life [8]. In another study of pregnant and nonpregnant women, serum DBP concentration was the most significant correlate of the deuterated 25(OH)D₃ area under the concentration-time curve after oral dosing of deuterated vitamin D₃ [21]. Presumably, these observations reflect the influence of DBP concentration on 25(OH)D₃ clearance due to an inverse relationship between DBP concentration and the plasma unbound fraction of 25(OH)D₃ and a theoretical direct relationship between clearance and the unbound fraction. Although systemic clearance of 25(OH)D is likely a complex combination of bound (kidney) and unbound (liver) clearances, at least because of reabsorption of bound 25(OH)D in the proximal tubules and the dominant role of the kidneys in metabolism of 25(OH)D, we expected a discernible positive association between the free fraction and clearance of 25(OH)D₃. Moreover, we anticipated greater power to detect this association when directly measuring the percent free 25(OH)D₃ rather than predicting it from an equation that incorporates a single binding constant for 25(OH)D₃ binding to DBP. For example, affinity for DBP has been alleged to differ by isoform [22]. We found that clearance of 25(OH)D₃ was positively related to the predicted percent free 25(OH)D₃, consistent with our prior observation that DBP concentration is directly associated with 25(OH)D₃ clearance [15]. However, contrary to our expectation, 25(OH)D₃ clearance was unrelated to the percent free based on LC-MS/MS-measured concentrations. This may indicate that our understanding of 25(OH)D transport and metabolism is flawed or incomplete. Alternatively, the percent free 25(OH)D₃ as determined by measured free 25(OH)D₃ concentrations may have lacked the precision needed to detect a real but very weak relationship between percent free 25(OH)D₃ and clearance, as free 25(OH)D₃ circulates in the pM range, whereas DBP circulates in μM.

Also to our surprise, the association between serum DBP concentration (or predicted percent free 25(OH)D₃) and 25(OH)D₃ clearance was less (not more) evident among the participants with kidney failure who had near to complete absence of nephron function. Although tentative, this suggests that the association observed among participants with normal to modestly impaired kidney function was related to renal metabolism of free 25(OH)D₃ that entered tubular cells through the basolateral membrane or to loss of very small amounts of free 25(OH)D₃ in the urine. Of note, we do not suspect that urinary excretion of bound 25(OH)D₃ contributed significantly to clearance, as participants with kidney failure were anuric, and none of the other study participants exhibited severe proteinuria [23].

GC haplotype was associated with 25(OH)D₃ clearance as hypothesized. Specifically, we hypothesized clearance would be increased with GC2 carriage and decreased with GC1s carriage due to genetic differences in serum DBP concentration and (possibly) 25(OH)D₃ affinity. Indeed, clearance was 87 mL/d higher among individuals carrying 2 relative to no GC2 alleles. erum DBP concentration and LC-MS/MS-measured percent free 25(OH)D₃ were not significantly different with GC2/GC2 diplotype. However, in the literature, serum DBP concentration is consistently 10–15% lower in people carrying GC2, and some but not all studies have reported 25(OH)D binding affinity is lowest with GC2, intermediate with GC1s, and highest with GC1f [4]. GC2 carriage has also been associated with lower total circulating 25(OH)D concentration in several studies, albeit the effect is of small magnitude [3, 24, 25]. Our data suggest this effect may be the consequence of increased 25(OH)D clearance among individuals with GC2 haplotype, but we cannot rule out that our finding is due to confounding or coincidence.

Regarding GC1s carriage, serum DBP concentration was higher with each additional GC1s allele, but clearance and GC1s carriage were not associated. In the one prior tracer study relating GC genotype to 25(OH)D pharmacokinetics, rs7041 variant allele (which confers GC1s carriage) was associated with a longer 25(OH)D plasma half-life [8]. In that same study, rs4588 variant allele (which confers GC2 carriage) was unrelated to 25(OH)D half-life, but only 3 of 30 participants were heterozygous for the variant allele and none were homozygous. In summary, our current findings are suggestive and agree with prior data but do not strongly support our original hypothesis that the association between 25(OH)D₃ clearance and GC haplotypes is mediated by genetic differences in percent free 25(OH)D₃, possibly due to limited power to detect weak biological relationships.

To our knowledge, this is the first study that compared predicted and directly measured free 25(OH)D concentrations in individuals with and without kidney disease. The relationship between free and total 25(OH)D did not vary according to kidney disease status, except when free 25(OH)D was measured by immunoassay in that the relationship was attenuated among participants with kidney failure. A previous preliminary study applied the same immunoassay in 84 adults with CKD and reported the ratio of free to total 25(OH)D strongly decreased with the decline of kidney function [26]. Our current findings challenge the validity of that conclusion. We included a novel LC-MS/MS method in our method comparison, and the results suggest a negative bias in immunoassay-measured free 25(OH)D concentrations among individuals with kidney failure. Although not clearly supported by our data, this bias may be attributable in part to lower cross-reactivity with free 25(OH)D₂, a known characteristic of this immunoassay [18] that complicates its use in CKD patients given the prevalence of ergocalciferol supplementation in these patients and the long whole body half-life of vitamin D [27]. Alternatively, the bias may be attributable to substances present in blood of individuals with kidney failure receiving hemodialysis, which affected immunoassay performance. For example, some immunoassay methods for total 25(OH)D have exhibited negative bias (relative to LC-MS/MS) among dialysis patients, which was not ascribed to high 25(OH)D₂ concentration [28].

The main limitation of our study is the small sample size, which limited statistical power. As far as we know, the CLEAR study is the largest published stable isotope study of human vitamin D metabolism. CLEAR provided a unique opportunity to evaluate mechanistic hypotheses about human vitamin D metabolism but a small sample size relative to clinical trials and observational studies including genome-wide association studies. Although we observed notable differences in the concentration of free 25(OH)D between methods, unfortunately, there is no reference measurement procedure for free 25(OH)D by which to evaluate method trueness. More research is needed to evaluate free 25(OH)D assays in clinical populations.

In conclusion, using state-of-the art methods to measure free plasma 25(OH)D₃ concentration and 25(OH)D₃ systemic clearance, we established that the percent free 25(OH)D₃ is not strongly associated with 25(OH)D₃ clearance but may mediate small differences in clearance according to vitamin D binding protein (GC) haplotype. In addition, the percent free 25(OH)D is not altered among patients with kidney failure. However, when free plasma 25(OH)D concentration is determined by immunoassay, the percent free 25(OH)D may appear lower in individuals with kidney failure likely caused by interfering substances in the blood.

HIGHLIGHTS.

• We assessed systemic clearance of 25(OH)D₃ in humans

• Clearance of 25(OH)D₃ varied according to vitamin D binding protein (GC) haplotype

• We employed a novel liquid chromatography-mass spectrometry assay for free 25(OH)D

• We observed a negative bias in immunoassay-measured free 25(OH)D concentrations

Data Availability:

The data that support the findings of this study are available from the corresponding author upon reasonable request.