The citrate transporter SLC13A5 as a therapeutic target for kidney disease: evidence from Mendelian randomization to inform drug development

The kidney is a highly metabolically active organ. The provision of citric acid cycle intermediates has ameliorated renal tubular damage and progression of chronic renal injury in animal models [1‐3]. Solute carrier family 13 member 5 (SLC13A5), which is primarily expressed in hepatocytes of the liver, is a membrane-bound Na⁺-coupled co-transporter responsible for moving extracellular citrate into the cytosol [4, 5]. Systemic SLC13A5 inhibition could augment citrate flux to the kidney, thereby supporting renal energy metabolism. Moreover, given the role of citrate in regulating glucose metabolism, lipid metabolism, and inflammation, SLC13A5 inhibition could indirectly affect renal health [6‐13]. However, no SLC13A5 inhibitor has yet entered clinical study, and so evidence of its efficacy for ameliorating progression of kidney disease in humans is limited. Additionally, SLC13A5 shows species-specific effects [11, 14‐16].

Insights from human genetic data offer a powerful opportunity to prioritize and inform the design of clinical research. Given that learnings from such genetic analyses relate to the target organism, namely humans, they are well-placed to overcome some of the limitations of translating findings from animal models [17]. As genes code for proteins and proteins make up the majority of drug targets, it follows that naturally occurring variation in the genes coding for drug target proteins can be leveraged to inform on the effect of their pharmacological perturbation [18]. The random allocation of genetic variants at conception means that their association with clinical traits and phenotypes are less susceptible to the confounding factors and reverse causation bias that can hinder causal inference in traditional epidemiological study designs. The current availability of publicly accessible large-scale genetic association data also means that such a drug target Mendelian randomization paradigm is a relatively fast and cost-effective approach for inferring the potential clinical effects of perturbing drug targets [19].

Given the pre-clinical evidence supporting potential therapeutic applications of SLC13A5 inhibition [10, 12, 13, 20, 21], we sought to investigate its effects using drug target Mendelian randomization. By utilizing the known effect of SLC13A5 inhibition on increasing plasma citrate levels to identify genetic instruments [10, 22, 23], our primary objective was to investigate its effects on parameters of kidney function. In secondary analyses, we aimed to unravel potential mechanisms of action, as well as evidence of broader effects on glucose and lipid metabolism, and inflammation. Such insights would thus serve to inform and prioritize clinical development efforts for SLC13A5 inhibitors.

A total of 13 SNPs were identified as instruments for SLC13A5 inhibition (Additional file 2: Table S1), all of which had F-statistics > 10, consistent with low risk of weak instrument bias that might affect the conclusions of Mendelian randomization analyses. Functional annotations of the SNPs obtained using PhenoScanner version 2 are provided in Additional file 2: Table S2. There were only potential pleiotropic associations of one variant (rs75448233) with smoking traits; however, these associations did not reach genome-wide significance (Additional file 2: Table S3).

The main IVW Mendelian randomization identified positive associations of genetically proxied SLC13A5 inhibition with plasma calcium levels. Every 1-SD higher genetically proxied plasma citrate through SLC13A5 inhibition was associated with a 0.132 SD units higher plasma calcium level (95% CI 0.107 to 0.157, p = 6 × 10⁻¹³).

Figure 2 depicts the associations of genetically proxied SLC13A5 inhibition with kidney function parameters from the main IVW Mendelian randomization analysis. There were statistically significant associations of genetically proxied SLC13A5 inhibition with higher creatinine and cystatin C-based measures of eGFR (FDR adjusted p values = 0.006 and 0.010 respectively), and lower BUN (FDR adjusted p value = 0.002). There were no strong associations of genetically proxied SLC13A5 inhibition with uACR (FDR adjusted p value = 0.682), or risk of CKD (FDR adjusted p value = 0.516) or microalbuminuria (FDR adjusted p value = 0.682). Additional file 3: Fig. S1-S6 show scatter plots of the SNP-exposure and SNP-outcome associations. Similar Mendelian randomization point estimates were obtained when using a pruning threshold of r² < 0.01 (Additional file 2: Table S4).

Results of the colocalisation analysis for primary outcomes showing significant Mendelian randomization associations are presented in Additional file 2: Table S5. For all these analyses, there was only evidence of a causal variant in the locus affecting plasma citrate levels, suggesting that there was inadequate statistical power to test whether the Mendelian randomization associations were attributable to shared or distinct causal variants [46].

The results of the metabolome-wide Mendelian randomization analysis investigating the effects of SLC13A5 inhibition on plasma and urine metabolites are presented in Additional file 2: Table S6 and S7. As expected, there was genetic evidence supporting associations of SLC13A5 inhibition with higher plasma citrate. Additionally, there was evidence supporting associations of SLC13A5 inhibition with higher plasma aconitate and plasma N-acetylspermidine.

The main IVW Mendelian randomization analysis results measuring the association of genetically proxied SLC13A5 inhibition with parameters of glucose and lipid metabolism, and inflammation are presented in Fig. 3. In this hypothesis-generating analysis, there were no significant associations after adjusting for multiple testing, but some evidence to support possible associations of SLC13A5 inhibition with lower fasting glucose levels (− 0.026 mmol/l per 1-SD higher plasma citrate through SCL13A5 inhibition, 95% CI − 0.047 to 0.004, raw p = 0.022) [10, 21]. Neither proteome-wide Mendelian randomization (Additional file 2: Table S8) nor PheWAS (Additional file 2: Table S9) identified any significant associations for genetically proxied SLC13A5 inhibition after correcting for multiple testing.

All analyses were also performed considering higher plasma citrate through any mechanism as the exposure, instead of genetically proxied SLC13A5 inhibition (i.e., SNPs associated with higher plasma citrate selected as in the analysis for SLC13A5, but not restricted to 200kB within the SLC13A5 gene). The SNPs used as instruments for plasma citrate are presented in Additional file 2: Table S10, and the analysis results are presented in Additional file 2: Tables S11-S14. Although the main IVW analysis did not identify evidence supporting associations of genetically predicted plasma citrate levels with eGFR or BUN, the Egger and weighted median sensitivity analyses that are more robust to the inclusion of pleiotropic SNPs were more supportive of such an association (Additional file 2: Table S15). This suggests that the genetic variants that affect plasma citrate are heterogeneous and pleiotropic in the mechanisms by which they achieve effects on eGFR and BUN. Consistent with this, there was strong evidence of heterogeneity in all the IVW MR analyses undertaken with plasma citrate as the exposure (Additional file 2: Table S15). This implies that increasing plasma citrate through SLC13A5 inhibition would potentially improve eGFR and BUN, but that increasing plasma citrate by other mechanism not necessarily achieve this same result.

Using large-scale genetic association data, we identified a genetic instrument for SLC13A5 inhibition that we leveraged in the Mendelian randomization paradigm to provide clinically translatable insights. Our main findings generated human genetic evidence that supports higher plasma citrate and calcium as biologically plausible biomarkers of target engagement, higher plasma citrate as a potential mechanism of action biomarker, and favourable effects on parameters of kidney function, namely creatinine and cystatin C-based measures of eGFR, and BUN. Of note, we did not identify statistically significant evidence of effects of SLC13A5 inhibition on uACR nor risk of microalbuminuria. This discrepancy may be explained by these biomarkers measuring different domains of kidney function [52]. Of relevance, calcium citrate treatment in an acute kidney damage animal model resulted in significantly reduced levels of proteinuria [1]. The primary Mendelian randomization analysis also did not identify evidence of SLC13A5 inhibition having effects on CKD risk, which may in part be explained by the pathophysiological heterogeneity underlying CKD [53], and SLC13A5 inhibition potentially only being relevant to some of these mechanisms.

Secondary Mendelian randomization analyses supported potential effects of SLC13A5 inhibition on reducing fasting glucose. Given that SLC13A5 is predominantly expressed in the liver [26], these findings are consistent with direct hepatic effects. The lack of strong associations of genetically proxied SLC13A5 inhibition with clinical outcomes in PheWAS is reassuring regarding safety profile, although the limited statistical power also means that findings may warrant further investigation. This suggests smaller reductions in SLC13A5 activity are unlikely to recapitulate the autosomal-recessive epileptic encephalopathy phenotype observed with rare mutations affecting critical regions of the SLC13A5 gene [54].

SLC13A5 is a membrane-bound citrate transporter, and its inhibition results in reduced entry of plasma citrate into the cytosol [4, 5]. This would be expected to have implications for cellular metabolism in cells expressing SLC13A5, with the higher circulating plasma citrate levels also potentially impacting metabolism systemically. Consistent with the results of our main genetic analyses, previous work has identified protective effects of citrate in various animal models of kidney disease [1, 6, 55, 56]. Although the main IVW Mendelian randomization analysis investigating effects of plasma citrate levels did not identify evidence supporting effects on parameters of kidney function, there was strong evidence of heterogeneity in estimates generated by individual SNPs, consistent with the presence of pleiotropic effects that could bias the main IVW analyses [43]. Furthermore, the Egger and weighted median Mendelian randomization statistical sensitivity analyses, which can be more robust to the inclusion of pleiotropic variants, produced estimates more suggestive of favourable effects of higher plasma citrate on measures of kidney function. This observation may suggest that plasma citrate is a heterogeneous trait that is affected through several distinct pathways, and that at least some of these may favourably impact kidney function. This would be in-keeping with the favourable effects of SLC13A5 on kidney function occurring through its effects on plasma citrate. It has been reported that low citrate levels in urine are associated with various forms of kidney disease, and low urinary citrate is a biomarker for kidney disease progression [55, 57‐62]. Furthermore, pre-clinical studies with SLC13A5 knockout mice have supported effects on sympathoadrenal mechanisms [63], which may also contribute to nephroprotective effects of genetically proxied SLC13A5 inhibition.

There are several hypotheses for how citrate may exert its beneficial effects on kidney function. Generally, elevated plasma citrate levels are proposed to lead to increased excretion of citrate through the urine. Higher urinary citrate may in turn provide an additional energy supply for kidney cells [1‐3], potentially enhancing their resilience and function. Furthermore, citrate may contribute to supporting the biosynthesis of essential molecules such as glucose, lipids, and amino acids. Additionally, it is known, due to its alkalinizing function, citrate contributes to improved acid–base balance within the kidneys, promoting a favourable environment for cellular processes. Finally, urinary citrate’s solubilizing properties might offer protection against vascular calcification and the formation of kidney stones, thus safeguarding kidney health.

This work has several strengths. By leveraging large-scale genetic association data within the Mendelian randomization paradigm, we were able to rapidly perform several hypothesis-driven and hypothesis-free genetic analyses relevant to humans, and consequently generate insight related to drug development. Our findings represent numerous potential learnings, including highlighting higher plasma citrate and calcium as biomarkers of target engagement, higher plasma citrate as a biomarker of mechanism of action, efficacy for improving kidney function, general safety profile, and potential effects on reducing fasting glucose. Further, the random allocation of genetic variants at conception means that these findings are less vulnerable to the confounding and reverse causation bias that can hinder causal inference in traditional epidemiological study designs [64].

This work should also be interpreted in the context of its limitations. The Mendelian randomization paradigm considers small lifelong effects of germline genetic variation. This is not the same as a discrete clinical intervention in later life, which may be of shorter duration but larger magnitude. Thus, it is possible that some of the findings of these genetic analyses may not predict what is observed with inhibition of SLC13A5 in clinical practice, particularly in quantitative terms. Further, some of our current analyses may also be limited by phenotypic definition and statistical power. For example, we note that kidney traits were not prioritised in the results of PheWAS analysis. In addition, the colocalization analyses did not support a single causal variant underlying the observed Mendelian randomization associations, and this may be attributable to either low statistical power, or confounding due to variant in linkage disequilibrium. In the Mendelian randomization analyses, we used a combination of European ancestry and multi-ancestry genetic association data, creating the potential for bias related to population substructure. Similarly, there may be interactions of the genetic variants employed as instruments with dietary intake of citrate, which may vary across population groups, potentially also introducing bias. While challenges in estimating the variance of drug target perturbation predicted by genetic variants means that it is not possible to perform conventional power calculations for drug target Mendelian randomization [65], some indication of the relative statistical power available for each analysis is apparent from the size of the 95% CIs. From this, it is clear that the analyses considering CKD had much less power than those considering measures of eGFR or BUN, for example. Of relevance, numerous preclinical mechanistic studies have demonstrated a role of SLC13A5 in favourably affecting hepatic lipid and glucose metabolism [10‐13, 20, 21, 66, 67], and the limitations of our current analyses may explain the discrepancy in findings. The plasma and urine metabolome-wide Mendelian randomization analyses were undertaken using outcome data obtained from a CKD population and may thus be vulnerable to collider bias [68]. This is because stratifying on a collider can introduce associations with confounding factors. Our Mendelian randomization evidence supports effects of SLC13A5 inhibition on parameters of kidney function, in-keeping with CKD therefore representing a potential collider. While there are strategies available to help explore potential bias from this [69], they either require individual participant data, which were not available for the current Mendelian randomization analyses, or have other prohibitive limitations [70]. Finally, the genetic analyses undertaken in this work provide insight into the on-target effects of perturbing SLC13A5 [19], but cannot directly inform on the effects of any pharmacological agents used to inhibit SLC13A5, including their pharmacokinetic and pharmacodynamic properties, or off-target effects.

In summary, this Mendelian randomization analysis provides novel human-centric insight to guide the clinical development of an SLC13A5 inhibitor. We identify biologically plausible biomarkers of target engagement and mode of action, as well as genetic evidence supporting potential effects in improving kidney function, with higher plasma citrate levels a possible underlying mechanism. The null PheWAS findings are reassuring in that they do not identify any safety signals. Further study in the form of early-stage clinical trials may now be warranted to help translate these findings towards improving patient care.