Association between Serum Glycerol-3-Phosphate and Fibroblast Growth Factor-23 in Nondiabetic Patients on Hemodialysis

Visual Abstract

Abstract

Key Points

• Serum glycerol-3-phosphate (G3P) levels were significantly associated with higher fibroblast growth factor 23 levels in ESKD.

• Pathway and mediation analyses identified the pentose phosphate pathway as the top mediator of the relationship between serum G3P and fibroblast growth factor 23.

• The role of G3P in ESKD provides new insights into potential therapeutic targets for CKD-mineral and bone disorder.

Background

Fibroblast growth factor 23 (FGF23) is a key regulator of mineral metabolism that is independently associated with mortality in patients with ESKD. Glycerol-3-phosphate (G3P), a byproduct of glycolysis that can be derived from injured kidneys, stimulates FGF23 production. We aimed to determine if serum G3P is associated with FGF23 levels in patients with ESKD and identify potential molecular pathways that mediate their relationship.

Methods

We conducted a cross-sectional study of 99 nondiabetic patients with ESKD on hemodialysis. We used linear regression to examine the association between G3P terciles and log-transformed C-terminal FGF23 levels, adjusting for demographics, coronary artery disease, serum calcium, phosphorus, and parathyroid hormone. Mann–Whitney U tests compared 247 serum metabolite levels between the first and third FGF23 terciles; significant metabolites (P < 0.01) were selected for pathway enrichment analyses. Top pathway scores were used in mediation analyses.

Results

The median age of participants was 54 years (interquartile range [IQR], 44–63), 38% were women, 71% self-identified as Black, and 27% had coronary artery disease. The median FGF23 level was 777 (interquartile range, 222–1310) RU/ml. In adjusted analyses, compared with participants with the lowest G3P tercile, those with the highest G3P tercile had a 95% higher FGF23 level (95% Confidence Interval, 6% to 260%; P = 0.04). Of the 27 metabolites significantly associated with FGF23 levels, pathway enrichment analysis identified the pentose phosphate pathway as the top hit (impact score=0.33, false discovery rate-adjusted P value = 0.01). The pentose phosphate pathway mediated the relationship between G3P and FGF23, resulting in a 62% change in the β coefficient.

Conclusions

In nondiabetic patients with ESKD on hemodialysis, serum G3P positively correlated with serum C-terminal FGF23, and this relationship was mediated by the pentose phosphate pathway. Exploring the pentose phosphate pathway could yield critical mechanistic insights into the regulation of FGF23, enhancing our understanding of its broader biologic functions.

Introduction

Cardiovascular disease is by far the most common cause of death in patients with CKD.^1,2 This is in part due to the association of kidney disease with traditional risk factors, such as diabetes and hypertension, as well as the dysregulation of calcium and phosphate metabolism.^3,4 CKD-mineral and bone disorder is a systemic disorder manifested by abnormalities of calcium, phosphorus, parathyroid hormone (PTH), or vitamin D metabolism; abnormalities of bone turnover; and ectopic soft tissue calcification including vascular calcification.^5,6

In the early 2000s, bone-derived hormone fibroblast growth factor 23 (FGF23) emerged as a central mediator of phosphate handling through binding and activating fibroblast growth factor receptor tyrosine kinases in an α-klotho coreceptor-dependent fashion.⁷ FGF23 is important in regulating mineral balance in healthy individuals and in disease states such as X-linked hypophosphatemia (XLH), oncogenic osteomalacia, and kidney disease.⁸ FGF23 downregulates 1,25-hydroxyvitamin D production and lowers serum phosphate levels. Notably, high circulating FGF23 is an independent risk factor for morbidity and mortality in the CKD population, partially due to its association with vascular dysfunction, atherosclerosis, and left ventricular hypertrophy.^9,10 These findings have prompted further investigation of lowering FGF23 levels in patients with CKD as a potential therapeutic target.

Until recently, upstream regulators of FGF23 were poorly understood. Some key regulators that have been identified include dietary phosphate, 1,25-hydroxyvitamin D, PTH, inflammation, iron deficiency, erythropoietin signaling, and calciprotein particles.^11–14 Kidney vein concentrations of glycerol-3-phosphate (G3P)—a metabolic product of glycolysis—were found to correlate with arterial FGF23 in patients undergoing cardiac catheterization, and the direct role of G3P in stimulating FGF23 production has been demonstrated in cell and animal models.¹⁵ Serum G3P is increased during AKI and correlates with serum FGF23 levels, suggesting that kidney-derived G3P may upregulate FGF23.¹⁶ However, the relationship between G3P and FGF23 in patients with ESKD is yet to be explored. This study aimed to determine if serum G3P is associated with FGF23 levels in patients with ESKD on hemodialysis as well as to identify molecular pathways that mediate their relationship.

Materials and Methods

Study Design

We conducted a cross-sectional analysis in a subset of 99 nondiabetic patients on hemodialysis from the Predictors of Arrhythmic and Cardiovascular Risk in ESKD (PACE) cohort. The PACE cohort is a prospective study comprising 568 patients who began regular, thrice-weekly, outpatient hemodialysis within 6 months of enrollment, recruited from November 2008 to August 2012. The study was conducted in accordance with the Declaration of Helsinki approved by the Institutional Review Boards of the Johns Hopkins School of Medicine and at the Hospital for Sick Children (PI Parekh). Details of the participant inclusion/exclusion criteria, procedures, and data collection process have been described previously.¹⁷ Patients with diabetes were excluded from this study to avoid potential confounding of diabetes-associated metabolic dysregulation.

Measurement of Serum Metabolites and FGF23

Serum was collected on a nondialysis day, at the baseline visit after approximately 8 hours of fasting, and stored at −80^°C. A total of 452 serum metabolites, including G3P (the primary independent variable), were measured using triple quadrupole mass spectrometry (AB Sciex 6500+ QTRAP; Sciex, Concord, Canada) coupled with a Waters (Milford, MA) Ultra Performance Liquid Chromatography.¹⁸ The chemical biologic pathway representation of these metabolites has been previously described, demonstrating broad coverage across diverse metabolic classes.¹⁹ Metabolites were included in the subsequent analyses if they had a coefficient of variation <30%, calculated from a pooled quality-control sample injected six times. 55% of metabolites (m=247) had CV <30% and were included in analyses.¹⁹ The coefficient of variation of G3P was 8.3%. The dependent variable, serum C-terminal FGF23 level, was measured using ELISA (CV=13%; Immutopics, San Clemente, CA).²⁰ The C-terminal assay recognizes epitopes on one side of the FGF23 cleavage site and therefore captures both biologically active, intact FGF23 and its cleaved form.²¹

Measurement of Covariates

Baseline demographic variables including age, sex, and race were self-reported by study participants. Medical comorbidities at baseline, including coronary artery disease (CAD), cardiovascular disease, and congestive heart failure, were assessed and confirmed by a physician committee. All other laboratory-based measurements were performed in the same serum samples used for the metabolomics panel, apart from serum PTH, total corrected calcium, and phosphorous levels, which were averaged from 3-months of laboratory values collected before the start of a dialysis session.

Statistical and Pathway Analyses

Participant characteristics were compared across serum G3P terciles using Fisher exact or Kruskal-Wallis tests. Covariates were selected based on clinical relevance and statistical significance based on their association with G3P and C-terminal FGF23 (P < 0.05). Correlations between continuous variables (namely, G3P and FGF23) were assessed using Spearman rank correlation coefficients. Linear regression was used to examine the association between G3P terciles with log-transformed C-terminal FGF23 while adjusting for demographics (i.e., age, race, and sex), history of CAD, total corrected serum calcium, phosphorus, and PTH. For sensitivity analyses, this model was further adjusted for hemoglobin and Kt/V. The percent difference in C-terminal FGF23 was calculated from the β coefficients generated from linear regression, as follows:

Percent difference in C-terminal FGF23 = (10^β−1)×100%

Mann–Whitney U tests were used to compare log-transformed serum metabolite levels in the first and third FGF23 terciles. Metabolites were deemed significant (“hits”) if associated with C-terminal FGF23 levels using a fold change threshold of ≥1 and an FDR-adjusted P value <0.01 and were subsequently selected for pathway analysis. Pathway-enrichment analyses using the Kyoto Encyclopedia of Genes and Genomes homo sapiens pathway library were performed using the metabolites significantly associated with FGF23. Pathways containing two or fewer significant metabolites were excluded from the subsequent mediation analysis. In addition, pathways that contained the independent variable, G3P, were excluded. The first principal component (PC1) score of the metabolite hits from each pathway was generated and used for pathway mediation analyses. A pathway was considered as a potential mediator if it was associated with both G3P and FGF23 levels. Baron and Kenny's method for mediation was used to assess pathway mediation and bootstrapping (500 replications) was used to assess the statistical significance of each pathway's indirect effects.^22,23 If the association between G3P and FGF23 completely disappeared after adjusting for the pathway, the pathway was considered to fully mediate the relationship. Descriptive statistics, linear regression, and mediation analysis were performed using STATA 17.0/BE (College Station, TX), with significance set as a two-sided P value of <0.05. Pathway and metabolomic analyses were performed using MetaboAnalyst 6.0 (Xia Lab, McGill University).²⁴

Results

Baseline Characteristics

The median age of participants was 54 years (interquartile range, 44–63), 38% were women, 71% were Black, and 27% had prevalent CAD (Table 1). The median G3P was 0.18 RU/ml (IQR, 0.15–0.22). Participants with higher G3P were younger, had higher serum phosphate, and had higher intact PTH (P = 0.02). The median FGF23 level among participants was 777 RU/ml (IQR, 222–1310 RU/ml). Higher FGF23 was associated with phosphorus (P < 0.001) and total corrected calcium (P = 0.02; Supplemental Table 1).

Table 1.

Baseline characteristics by serum glycerol-3-phosphate tercile among participants from the predictors of arrhythmic and cardiovascular risk in ESKD cohort

G3P Tercile	Overall (N=99)	1st tercile (n=33)	2nd tercile (n=33)	3rd tercile (n=33)
Age (yr)	54 [44–63]	57 [50–69]	55 [44–63]	50 [39–57]
Female	38 (38%)	11 (33%)	18 (55%)	9 (27%)
Race (black)	71 (71%)	24 (73%)	24 (73%)	23 (70%)
BMI (kg/m2)	25.8 [22.8–30.6]	25.7 [22.7–30.4]	25.6 [23.3–31.7]	27.3 [22.9–29.7]
Coronary artery disease	27 (27%)	10 (30%)	9 (27%)	8 (24%)
Cardiovascular disease	14 (14%)	4 (12%)	5 (15%)	5 (15%)
Congestive heart failure	29 (29%)	8 (24%)	11 (33%)	10 (30%)
Total corrected calcium (mg/dl)	9.0 [8.8–9.3]	9.3 [8.8–9.5]	8.9 [8.6–9.3]	8.9 [8.7–9.4]
Phosphorus (mg/dl)	5.1 [4.5–5.9]	4.5 [4.1–5.0]	5.0 [4.6–5.6]	6.0 [5.1–6.5]
PTH (RU/ml)	399.5 [269.6–581.3]	321.5 [247.0–465.0]	400.3 [257.1–605.8]	528.0 [349.3–724.7]
Hemoglobin (g/dl)	10.9 [9.5–11.8]	10.8 [9.8–11.7]	11.1 [10.2–11.8]	10.7 [9.4–11.5]
Albumin (g/dl)	3.8 [3.4–4.0]	3.7 [3.4–4.0]	3.8 [3.5–4.0]	3.7 [3.3–4.0]
LDL (mg/dl)	85.0 [60.2–104.2]	69.5 [47.8–106.8]	88.2 [68.3–100.8]	78.6 [64.8–98.9]
C-terminal FGF23	776.9 [218.3–1329.7]	324.0 [161.4–916.2]	402.16 [206.0–810.8]	1303.1 [884.0–1500.0]
C-reactive protein (µg/ml)	4.5 [1.7–11.4]	4.1 [1.5–13.4]	4.7 [2.1–13.3]	3.5 [1.8–9.1]
Ferritin (ng/ml)	328.9 [185.4–573.0]	374.0 [214.0–579.0]	290.5 [167.7–591.2]	302.7 [189.5–532.5]
Klotho (pg/ml)	355.7 [268.3–489.7]	355.7 [268.3–482.8]	379.0 [304.7–537.4]	338.3 [233.8–456.4]

Participant characteristics are compared between glycerol-3-phosphate categorical groups using the Fisher exact test for categorical variables and the Kruskal-Wallis test for continuous variables. For continuous variables, values with normal distribution are as median [interquartile range]. Categorical variables are presented as absolute number (percentage). BMI, body mass index; FGF-23, fibroblast factor-23; G3P, glycerol-3-phosphate; LDL, low-density lipoprotein; PACE, predictors of arrhythmic and cardiovascular risk in ESKD; PTH, parathyroid hormone.

Association of G3P with FGF23

Serum G3P was associated with C-terminal FGF23 when G3P was examined as a continuous variable in unadjusted (higher by 101% per 0.1 RU/ml P < 0.001) models (Table 2). This was slightly attenuated but remained statistically significant (P = 0.002) after adjustment for age, race, sex, CAD, serum calcium, phosphorus, and PTH. Serum G3P and FGF23 levels were positively correlated (Spearman ρ=0.49, P < 0.001) when analyzed as continuous variables. Figure 1 shows the distribution of FGF23 by G3P terciles (P for trend < 0.001). In unadjusted analyses, those with the highest G3P tercile had 162.5% (95% CI, 57.4% to 338.0%) higher C-terminal FGF23 level compared with participants with the lowest G3P tercile (Figure 1). In adjusted analysis, those within the highest G3P tercile had a 95% higher level of FGF23 (95% CI, 6% to 260%) compared with participants within the lowest G3P tercile (Table 2). A positive but nonsignificant association was found between G3P and FGF23 between the first and second terciles. To account for potential confounding by dialysis clearance, we performed a sensitivity analysis adjusting for dialysis adequacy. The association remained significant after adjustment for single-pool Kt/V in the fully adjusted model (P = 0.04; Supplemental Table 4). In an additional sensitivity analysis, the association between G3P and C-terminal FGF23 remained significant after further adjustment for hemoglobin as a covariate (P = 0.03).

Table 2.

Linear regression of fibroblast factor-23 with glycerol-3-phosphate

G3P	Unadjusteda		Adjustedb
	% Difference (95% CI)	P Value	% Difference (95% CI)	P Value
Continuous (per 0.1 RU/ml)	100.6 (50.8 to 166.8)	P < 0.001	74.5 (24.0 to 145.6)	0.002
Tercile 1 (0.09–0.16 RU/ml)	Ref.	Ref.	Ref.	Ref.
Tercile 2 (0.16–0.21 RU/ml)	17.5 (−29.5% to 96.0)	P = 0.53	1.5% (−42.0 to 77.7)	0.96
Tercile 3 (range: 0.21–0.44 RU/ml)	162.5 (57.4 to 338.0)	P < 0.001	95.0% (5.8 to 259.6)	0.04

“Range” represents the minimum and maximum values of glycerol-3-phosphate in each tercile. CI, confidence interval; FGF23, fibroblast growth factor-23; G3P, glycerol-3-phosphate.

^an=98.

^bn=95. Adjusted for age, sex, race, coronary artery disease, total corrected serum calcium, phosphorus, and parathyroid hormone.

Figure 1.

Serum C-terminal FGF23 by G3P tercile. Serum G3P, shown as terciles, was measured using triple quadrupole mass spectrometry in nondiabetic patients on hemodialysis (N=99). The dependent variable, serum C-terminal FGF23 level, was measured using ELISA. Unadjusted linear regression was performed to evaluate for trend. FGF23, fibroblast growth factor-23; G3P, glycerol-3-phosphate.

Pathway Enrichment and Mediation Analyses

Twenty-seven metabolites were significantly associated with FGF23 levels (P < 0.05; Supplemental Table 1). The top three pathways yielded from enrichment analysis were the pentose phosphate pathway) (impact=0.22, P = 0.002), glycerophospholipid metabolism (impact=0.21, P = 0.006), and the tricarboxylic acid cycle (impact=0.09, P = 0.02) (Figure 2 and Supplemental Table 2). The latter two pathways were excluded as they had two or less significant metabolites after the removal of G3P. After adjustment for the PC1 score generated from the four metabolites included in the pentose phosphate pathway (D-gluconic acid, 6-Phospho-D-gluconate, D-Ribose 5-phosphate, and Erythrose-4-phosphate), the magnitude of association between first versus third G3P tercile and FGF23 was greatly attenuated (β coefficient for G3P was lower by 62 E%) and no longer reached statistical significance (P = 0.7). The mediation effect was significant with P < 0.001 by bootstrapped indirect effect. Moreover, the PC1 score of the pentose phosphate pathway predicted G3P tercile in linear regression analysis (β=0.8, P < 0.001). These findings suggest that the pentose phosphate pathway mediates the relationship between G3P and FGF23 (Figure 3).

Figure 2.

Pathway analysis. Pathway-enrichment analyses using the KEGG homo sapiens pathway library were performed using the 27 metabolites associated with FGF23 (shown in Supplemental Table 2). Compounds without an HMDB ID were excluded from analysis. The top four pathways with two or more metabolites are indicated in black text. Pentose phosphate pathway metabolites included in analysis were D-gluconic acid, 6-Phospho-D-gluconate, D-Ribose 5-phosphate, and Erythrose-4-phosphate. Courtesy of metaboanalyst.ca (MetaboAnalyst 6.0, Xia Lab, McGill University). HMDB, human metabolome database; ID, identification number; KEGG, Kyoto Encyclopedia of Genes and Genomes.

Figure 3.

Mediation of the relationship between G3P and FGF23 by the pentose phosphate pathway. Adjusted for age, sex, race, CAD, total corrected serum calcium, phosphorus, and log-transformed PTH. ^†n=95. CAD, coronary artery disease; PTH, parathyroid hormone.

Discussion

In alignment with previous studies in cell culture, animal models, healthy patients, and those with AKI,¹⁵ this study demonstrates a positive association of G3P with FGF23 in patients with ESKD on dialysis. This finding appears to be novel in the medical literature and suggests that there is a role of G3P as a regulator for FGF23 even in ESKD.

Studying the regulators of FGF23 is of paramount significance due to its established role as an independent predictor of mortality in patients with ESKD.^25,26 Specifically, a recent cohort study of patients with advanced CKD documented a robust, independent association of FGF23 with all-cause mortality and cardiovascular death with a hazard ratio of 1.76 (95% CI, 1.28 to 2.44) when comparing FGF23 in the fourth versus the first quartile.²⁷ Moreover, preclinical studies have demonstrated direct deleterious effects of FGF23 on vascular physiology, fibrosis, and immune dysregulation—both in the presence or absence of FGF23's α-klotho coreceptor signaling.^28–31 Beyond ESKD, excessive FGF23 is also implicated in diseases such as oncogenic osteomalacia, XLH, and fibrous dysplasia/McCune-Albright syndrome.^32,33 Although monoclonal antibodies against FGF23 (i.e., burosumab) have proved promising in the treatment of bone remodeling and clinical outcomes in XLH and fibrous dysplasia/McCune-Albright syndrome,^34,35 their application in CKD remains limited by safety concerns, high cost, and the potential to undermine FGF23's role in maintaining phosphate balance in earlier stages of kidney disease.^36–38 Developing a better mechanistic understanding of upstream regulators of FGF23 production is essential to identify novel modifiable targets that could improve FGF23-associated cardiovascular mortality.

Our findings align with previous human and experimental studies linking G3P production to FGF23 regulation. Simic et al.¹⁵ measured metabolites in kidney veinous blood obtained from patients undergoing cardiac catheterization and found G3P had the highest correlation with serum intact arterial FGF23. In the same article, a clinical case-control study involving patients who developed AKI after surgery, researchers observed that circulating G3P levels increased rapidly postoperatively.¹⁵ Moreover, higher G3P concentrations correlated with elevated FGF23 levels 24 hours after surgery, supporting a direct link between kidney injury, G3P production, and FGF23 regulation. Additional studies were conducted to better characterize the mechanisms and sources of increased G3P in states of kidney insufficiency, looking specifically at kidney ischemia and phosphate loading. Although G3P is produced ubiquitously throughout the body by glycolysis, the authors demonstrated that G3P produced specifically from kidney artery clamping and regulated FGF23 production in bone through a glycerophospholipid intermediate—lysophosphatidic acid.^39,40 In a mouse model of CKD induced by an adenine-rich diet, the same group found that phosphate loading stimulated kidney production of G3P via G3P-dehydrogenase 1 (GPD1). GPD1 is a cytosolic enzyme that can be found in the proximal tubule cells that synthesizes G3P from dihydroxyacetone phosphate while oxidizing nicotinamide adenine dinucleotide hydride to nicotinamide adenine dinucleotide.⁴¹ Although there are other routes to synthesizing G3P, GPD1 was found to be the predominant pathway for kidney G3P production.⁴² In these experiments, G3P was found to be acutely elevated in response to phosphate loading in both mice and healthy human participants.⁴¹ This finding aligns with our cohort, which—despite having significantly greater, chronic impairments in kidney function—also shows a robust association between serum phosphate and G3P.

In GPD1-deficient mice, both G3P and FGF23 were only partially suppressed, suggesting that there are alternative pathways linking CKD to G3P-mediated FGF23 regulation.⁴² The role of G3P production from nonkidney tissues on FGF23 is yet to be fully explored. We built on this work by demonstrating a positive association between G3P and FGF23 in patients with ESKD. Our findings suggest that there may be a significant role for kidney-independent sources of G3P in FGF23 regulation, given the patients in this study were dialysis-dependent with variable residual kidney function in the first 6 months of dialysis. However, since only a third of participants had available data on residual kidney function, we could not assess its potential effect modification.⁴³ The observed associations might differ in a cohort on long standing dialysis with negligible residual kidney function. In the chronic dialysis population, G3P may be elevated due to passive accumulation through decreased kidney clearance, the persistence of G3P in circulation, or increased synthesis through nonkidney tissues. Future studies using lipidomics and targeted metabolic profiling would help delineate the tissue-specific sources contributing to circulating G3P and downstream mediators of the molecule on FGF23 production.

Another key finding of our study is that the pentose phosphate pathway represented by the PC1 score of four significant compounds within it (i.e., D-gluconic acid, 6-Phospho-D-gluconate, D-Ribose 5-phosphate, and Erythrose-4-phosphate), statistically mediated the relationship between G3P and FGF23. The pentose phosphate pathway is an anabolic pathway that plays a key role in nucleotide synthesis (through ribose 5-phosphate) and the generation of reducing equivalents in the form of NADPH, which is essential for redox balance.⁴⁴ The relationship between G3P and the pentose phosphate pathway is dynamic and multifaceted. Glycolysis and the pentose phosphate pathway share several intermediates, including glucose-6-phosphate and glyceraldehyde-3-phosphate. Glucose-6-phosphate, through the action of its dehydrogenase, bridges the gap between glycolysis and the oxidative branch of the pentose phosphate pathway.⁴⁵ High levels of G3P have been found to serve as an inhibitor of glucose-6-phosphate isomerase and therefore could theoretically modulate the flux of glycolytic carbon into the pentose phosphate pathway.⁴⁶ Alternatively, FGF23's effect on phosphate metabolism through its effects on bone and the kidney may indirectly affect the availability of phosphate for metabolic processes like the pentose phosphate pathway.⁴⁷ Interestingly, a recent study by members of our group found that the genes associated with the pentose phosphate pathway were highly enriched in the genetic loci associated with circulating FGF23.⁴⁸ In that study, the top single nucleotide polymorphism from their multitrait analysis of genome-wide association studies was adjacent to transketolase, an enzyme within the nonoxidative branch of the pentose phosphate pathway. Given these findings, it is possible that enzymes within the pentose phosphate pathway may play a central role in connecting FGF23 to vital cellular functions including glucose, lipid, and energy metabolism. A particularly significant link is the generation of glyceraldehyde-3-phosphate by transketolase from ribose-5-phosphate. This intermediate feeds into a key glycolytic step catalyzed by glyceraldehyde-3-phosphate dehydrogenase, which is coupled through nicotinamide adenine dinucleotide/nicotinamide adenine dinucleotide hydride recycling by GPD1 to produce G3P, providing a direct biochemical route between the pentose phosphate pathway and G3P synthesis (Figure 4). In this study, the observed association between the pentose phosphate pathway and FGF23 may be confounded by unmeasured metabolic disturbances associated with ESKD. Thus, mechanistic studies are needed to validate this relationship, clarifying causality and directionality.

Figure 4.

Proposed metabolic pathways linking glycolysis, the pentose phosphate pathway, and FGF23 regulation. Large arrows indicate metabolites that were found to be significantly associated with FGF23 in the current observational study. Dashed lines represent pathways with multiple intermediate steps. Dashed lines with “?” symbol indicate proposed mediation pathways. Created courtesy of BioRender.com. GPD1, glycerol-3-phosphate dehydrogenase; G6PD, glucose-6-phosphate dehydrogenase; RPI, ribose-5-phosphate isomerase; TPI, Triose phosphate isomerase; 6GPL, 6-phosphogluconolactonase; 6PGD, 6-phosphogluconate dehydrogenase.

This study has several limitations. First, the observational design prevents the determination of the causation or directionality between G3P and FGF23 levels. There is still a complex interplay of metabolic factors and comorbidities in ESKD that may confound the relationship.⁴⁹ Second, although limiting the study to nondiabetic patients removed a potential confounder, the results may be less generalizable to the broader population on long-term dialysis.⁵⁰ Given that insulin resistance and glycolytic flux are altered in diabetes,⁵¹ investigating whether similar metabolic-FGF23 relationships exist in diabetic patients in ESKD may be warranted, since FGF23 is elevated and associated with cardiovascular outcomes and mortality in diabetic nephropathy.^52,53 Similarly, our study participants were in a fasting state at the time of blood collection. Since previous animal studies have demonstrated a more pronounced phosphate-dependent response of G3P in the fed state,⁴¹ we may have underestimated the full potential of this metabolic axis. Future studies including nonfasted patients may be warranted. Third, this study did not assess key cardiovascular outcomes associated with elevated G3P and FGF23 levels. Understanding the implications of G3P metabolism in patient prognosis would be crucial for clinical guidelines and future research.⁵⁴ Finally, a key limitation is the lack of a replication cohort to validate our findings, and future studies in larger and more diverse cohorts are needed to confirm these associations in other ESKD populations. Despite these limitations, our study has several strengths. Most notably, the study consists of a well-defined cohort and uses rigorous statistical methods to adjust for important covariates and multiple comparisons to provide a robust evaluation of the relationship between G3P and FGF23. The study explores underlying molecular mechanisms supporting the relationship and limits metabolomic analysis to compounds with low CVs.

In summary, among patients with ESKD on dialysis, serum G3P is positively and robustly associated with serum C-terminal FGF23. The findings support that there is a kidney-independent mechanism by which G3P regulates FGF23 in ESKD. Exploring the mediation role of the pentose phosphate pathway in the relationship between G3P and FGF23 could yield critical mechanistic insights into the regulation of FGF23, enhancing our understanding of its broader biologic functions. Further exploration of these relationships will be vital for developing therapies that address both kidney and cardiovascular health in this high-risk population.

Supplementary Material

cjasn-21-211-s001.pdf ^{(2.6MB, pdf)}

cjasn-21-211-s002.pdf ^{(267.6KB, pdf)}

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/CJN/C484.

Author Contributions

Conceptualization: Wei Chen, Michelle M. Estrella, Bernard G. Jaar, Aline Martin, Rulan S. Parekh, Stephen M. Sozio, Scott M. Wilson, Shari Zaslow.

Data curation: Wei Chen, Bernard G. Jaar, Cassiane Robinson-Cohen, Stephen M. Sozio, Scott M. Wilson, Shari Zaslow.

Formal analysis: Wei Chen, Farzana Perwad, Scott M. Wilson, Shari Zaslow.

Funding acquisition: Wei Chen, Bernard G. Jaar, Stephen M. Sozio.

Investigation: Wei Chen, Michelle M. Estrella, Bernard G. Jaar, Aline Martin, Rulan S. Parekh, Stephen M. Sozio, Scott M. Wilson.

Methodology: Wei Chen, Michelle M. Estrella, Bernard G. Jaar, Aline Martin, Stephen M. Sozio, Scott M. Wilson, Shari Zaslow.

Resources: Bernard G. Jaar, Stephen M. Sozio.

Supervision: Rulan S. Parekh.

Validation: Cassiane Robinson-Cohen.

Visualization: Farzana Perwad, Scott M. Wilson.

Writing – original draft: Scott M. Wilson, Shari Zaslow.

Writing – review & editing: Wei Chen, Michelle M. Estrella, Bernard G. Jaar, Aline Martin, Rulan S. Parekh, Farzana Perwad, Cassiane Robinson-Cohen, Stephen M. Sozio, Shari Zaslow.

Funding

R.S. Parekh: National Institute of Diabetes and Digestive and Kidney Diseases (R01 DK072367 and R01 DK090070). W. Chen: National Heart, Lung, and Blood Institute (R01 HL172073), National Institute of Diabetes and Digestive and Kidney Diseases (R03 DK131246), Irma T. Hirschl Trust (Irma T. Hirschl Career Scientist Award). A. Martin: National Institute of Diabetes and Digestive and Kidney Diseases (R01 DK132342 and R01DK131046).

Declarative Statements

This study includes clinical experimentation and received Institutional Review Board or Ethics Committee approval. All patients provided written informed consent. This study includes clinical experimentation and complies with the Declaration of Helsinki. Large dataset (e.g., omics data, health care data, clinical trial data, imaging data).

Data Availability Statements

Original data generated for the study will be made available upon reasonable request to the corresponding author. Data Type: Observational Data. Reason for Restricted Access: The datasets generated and/or analyzed during the current study are available upon request but may require a data sharing agreement. Requests for access will be reviewed on a case-by-case basis.

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/CJN/C485.

Supplemental Table 1. Baseline characteristics by C-terminal FGF23 tercile among participants.

Supplemental Table 2. Metabolites significantly associated with dichotomized C-terminal FGF23 (third versus first tercile).

Supplemental Table 3. Sensitivity analyses of the relationship of FGF23 with G3P tercile.

Supplemental Figure 1. Volcano plot of metabolites significantly associated with dichotomized C-terminal FGF23 (third versus first tercile).