Assessment of Proximal Tubular Function by Tubular Maximum Phosphate Reabsorption Capacity in Heart Failure

Visual Abstract Background and objectives The estimated glomerular filtration rate (eGFR) is a crucial parameter in heart failure. Much less is known about the importance of tubular function. We addressed the effect of tubular maximum phosphate reabsorption capacity (TmP/GFR), a parameter of proximal tubular function, in patients with heart failure. Design, setting, participants, & measurements We established TmP/GFR (Bijvoet formula) in 2085 patients with heart failure and studied its association with deterioration of kidney function (>25% eGFR decrease from baseline) and plasma neutrophil gelatinase–associated lipocalin (NGAL) doubling (baseline to 9 months) using logistic regression analysis and clinical outcomes using Cox proportional hazards regression. Additionally, we evaluated the effect of sodium-glucose transport protein 2 (SGLT2) inhibition by empagliflozin on tubular maximum phosphate reabsorption capacity in 78 patients with acute heart failure using analysis of covariance. Results Low TmP/GFR (<0.80 mmol/L) was observed in 1392 (67%) and 21 (27%) patients. Patients with lower TmP/GFR had more advanced heart failure, lower eGFR, and higher levels of tubular damage markers. The main determinant of lower TmP/GFR was higher fractional excretion of urea (P<0.001). Lower TmP/GFR was independently associated with higher risk of plasma NGAL doubling (odds ratio, 2.20; 95% confidence interval, 1.05 to 4.66; P=0.04) but not with deterioration of kidney function. Lower TmP/GFR was associated with higher risk of all-cause mortality (hazard ratio, 2.80; 95% confidence interval, 1.37 to 5.73; P=0.005), heart failure hospitalization (hazard ratio, 2.29; 95% confidence interval, 1.08 to 4.88; P=0.03), and their combination (hazard ratio, 1.89; 95% confidence interval, 1.07 to 3.36; P=0.03) after multivariable adjustment. Empagliflozin significantly increased TmP/GFR compared with placebo after 1 day (P=0.004) but not after adjustment for eGFR change. Conclusions TmP/GFR, a measure of proximal tubular function, is frequently reduced in heart failure, especially in patients with more advanced heart failure. Lower TmP/GFR is furthermore associated with future risk of plasma NGAL doubling and worse clinical outcomes, independent of glomerular function.


Introduction
Impaired kidney function is common and associated with greater risk of poor outcomes in patients with heart failure (1). Many studies have addressed the cardiorenal syndrome but almost exclusively focused on glomerular filtration rate (GFR). Much less is known about prevalence, predictors, and clinical outcome of tubular function in patients with heart failure (2,3). Because proximal tubular function is of vital importance for sodium handling of the kidney, sodium retention, and diuretic response, it is of great relevance to patients with heart failure (3,4). Recent studies have drawn attention to the proximal tubule by showing that sodium-glucose cotransporter 2 (SGLT2) inhibitors improve outcomes in patients with heart failure with reduced ejection fraction (5,6).
The tubular maximum phosphate reabsorption capacity (TmP/GFR) indicates the maximum capacity of the kidney to reabsorb phosphate in the proximal tubule independent of GFR and net inflow of phosphate (7,8). As such, TmP/GFR can be considered a parameter of proximal tubular function. In experimental studies, expression of sodium phosphate cotransporters decreased in response to kidney injury, ranging from injury due to ischemia-reperfusion to hereditary kidney diseases, thereby decreasing tubular phosphate reabsorption (9)(10)(11)(12)(13)(14). Because of its high rates of oxygen consumption and relatively sparse endogenous antioxidant defenses, the proximal tubule as a whole is particularly vulnerable to injury (15). TmP/GFR of living kidney donors predicts recipient measured GFR independent of donor measured GFR, confirming TmP/GFR as a functional tubular parameter not primarily driven by GFR (16). As such, TmP/GFR might be useful to detect damage and dysfunction of the proximal tubule and investigate its consequences. We therefore aimed to study the clinical value of TmP/GFR and investigate effects of SGLT2 inhibition on TmP/GFR in heart failure.

Patient Populations
We used the index cohort of a systems Biology Study to Tailored Treatment in Chronic Heart Failure (BIOSTAT-CHF), an investigator-driven, multicenter clinical study consisting of 2516 patients with the aim of identifying patients with a poor outcome despite currently recommended treatment with angiotensin-converting enzyme inhibitors/angiotensin receptor blockers and b-blockers (17).
Additionally, we evaluated TmP/GFR in the randomized, double-blind, placebo-controlled, multicenter pilot study on the effects of empagliflozin on clinical outcomes in patients with acute decompensated heart failure (EMPA-RESPONSE-AHF) cohort. The main outcomes have been published (18). In brief, patients hospitalized for acute heart failure treated with (intravenous) loop diuretics were included. Patients were randomly assigned to empagliflozin 10 mg/d or placebo within 24 hours of presentation and treated for 30 days.
Both studies were conducted in accordance with the Declaration of Helsinki, national ethics and legal requirements, and relevant European Union legislation. All participants provided written informed consent for study participation.

Calculations and Biomarker Measurements
TmP/GFR could be calculated in 2085 patients in BIOSTAT-CHF and 78 patients in EMPA-RESPONSE-AHF using the formula described by Payne (7) and originally devised by Bijvoet (19). The study subset did not differ substantially from patients without available TmP/GFR (Supplemental Table 1 Urinary phosphate and other urinary measurements were measured in nonfasting spot urine using standardized methods in both cohorts. The normal reference range of 0.80-1.35 mmol/L for TmP/GFR was derived from data in healthy individuals 65-75 years old (7).
The association between TmP/GFR, doubling of plasma neutrophil gelatinase-associated lipocalin (NGAL) levels, and deterioration of kidney function (.25% eGFR decrease) was analyzed between baseline and 9 months (20,21). Sensitivity analyses were conducted for .30% and .40% eGFR decrease from baseline. eGFR was calculated using the Chronic Kidney Disease Epidemiology Collaboration creatinine formula (22). Presence of CKD was defined as eGFR,60 ml/min per 1.73 m 2 . After blood was drawn by venipuncture, samples were stored at 280 C. If possible, analyses were performed directly locally with standardized international methods; otherwise, they were performed in a central laboratory. Urinary kidney injury molecule-1 (KIM-1) and NGAL were measured using an in-house developed and validated multiplex immunoassay (xMAP; Luminex, Austin, TX) (23). Measurement of additional biomarkers was performed as previously described (24)(25)(26). Endogenous fractional excretion of lithium was measured prior to empagliflozin administration using inductively coupled plasma mass spectrometry in EMPA-RESPONSE-AHF (Supplemental Methods).

Study End Points
The relation of TmP/GFR with all-cause mortality, heart failure hospitalization, and those combined was evaluated. End points were adjusted for BIOSTAT-CHF risk models created for each specific outcome (27).

Statistical Analyses
Data are presented as means6SD when normally distributed, as medians (quartiles 1 to quartile 3) when skewed, and as frequencies (percentages) when categorical, and statistically tested using ANOVA, the Kruskal-Wallis H test, and the chi-squared test, respectively. Trends were tested with the Cochran-Armitage trend test, the Jonckheere-Terpstra test, or linear regression for categorical, skewed variables, and normally distributed variables, respectively.
Determinants of TmP/GFR were analyzed using linear regression. All variables with P50.10 univariably were included in multivariable analysis and subjected to backward elimination. Variables with P50.05 were retained in the final multivariable regression model. Prior to linear regression, normal distribution of residuals was checked as well as the presence of outliers. If necessary, variables were transformed using natural logarithm, including TmP/GFR. The association between log TmP/GFR, risk of plasma NGAL doubling, and deterioration of kidney function was evaluated using logistic regression. Results were adjusted for log plasma and urinary NGAL, eGFR, log urine albumin-creatinine ratio (UACR), log KIM-1, log fractional excretion of sodium (FE Na ), and log fractional excretion of urea (FE Urea ). Results are expressed as odds ratios (ORs) with 95% confidence intervals (95% CIs). Cox proportional hazard models were constructed to evaluate the prognostic value of log TmP/GFR and adjusted for BIOSTAT-CHF risk models, phosphate, eGFR, and New York Heart Association class. Results are expressed as hazard ratios (HRs) with 95% CIs. Kaplan-Meier plots were constructed using the survminer package. The dendrogram was constructed using the Hmisc package. Analysis of covariance was used to study the effects of empagliflozin versus placebo on TmP/GFR, controlling for baseline TmP/GFR. A twotailed P50.05 was considered statistically significant. Statistical analyses were performed with R (version 1.3.1073).

Baseline Characteristics
In BIOSTAT-CHF, median TmP/GFR was 0.67 (0.46-0.88) mmol/L. Low TmP/GFR (,0.80 mmol/L) was observed in 1392 (67%) patients. Patients with lower TmP/ GFR were older; had lower levels of serum phosphate and higher fractional excretion of phosphate; had lower eGFR; had higher doses of loop diuretics; and had higher levels of N-terminal probrain natriuretic peptide (NT-proBNP), FE Na , FE Urea , plasma and urinary NGAL, and UACR (Table  1). Baseline characteristics by median TmP/GFR and presence/absence of CKD are shown in Supplemental Table 2 and according to low, normal, and high TmP/GFR in Supplemental Table 3, the latter yielding similar trends to Table 1.

Association between Tubular Maximum Phosphate Reabsorption Capacity and Clinical Variables
Multivariable linear regression for log TmP/GFR including serum phosphate is shown in Supplemental Table 4 (adjusted R 2 50.83). Upon exclusion of phosphate, the strongest predictors of log TmP/GFR were log FE Urea , log urea, hemoglobin, log urinary creatinine, and log loop diuretic dose (Table 2) (adjusted R 2 50.24).

Correlation Plots and Hierarchical Cluster Analysis of Tubular Maximum Phosphate Reabsorption Capacity
Correlation plots showing the association between TmP/ GFR and eGFR, plasma NGAL, urea, and FE Urea are displayed in Figure 1. In hierarchical cluster analysis ( Figure  2), TmP/GFR clustered closely with urinary creatinine, FE Na , age, and hemoglobin. TmP/GFR did not directly cluster with serum phosphate or eGFR.

Tubular Maximum Phosphate Reabsorption Capacity Predicts Doubling of Plasma Neutrophil Gelatinase-Associated Lipocalin over Time
Deterioration of kidney function occurred in 221 (19%) of 1168 patients with available measurements. Log decrease of TmP/GFR was neither a significant predictor for deterioration of kidney function overall (OR, 0.53; 95% CI, 0.27 to 1.04; P50.06) nor a significant predictor in a subgroup of patients without CKD (n5569; OR, 0.65; 95% CI, 0.24 to 1.78; P50.40). Sensitivity analyses for .30% and .40% decrease in eGFR yielded consistent results.
Doubling of plasma NGAL between baseline and 9 months occurred in 413 (35%) of 1178 patients with available measurements. Log TmP/GFR decrease was a significant predictor for doubling of plasma NGAL after adjustment for baseline plasma and urinary NGAL, eGFR, urinary KIM-1, UACR, FE Na , and FE Urea (
The combination of low TmP/GFR and CKD was significantly associated with higher risk of poor outcomes after adjustment (Supplemental Tables 6 and 7). Kaplan-Meier curves also revealed worst survival for this combination (Supplemental Figure 2).

Effect of Empagliflozin on Tubular Maximum Phosphate Reabsorption Capacity
Median TmP/GFR in EMPA-RESPONSE-AHF was 0.97 (0.78-1.13) mmol/L. Low TmP/GFR was observed in 21 (27%) patients. Patients with TmP/GFR below the median had higher levels of NT-proBNP, FE Urea , and fractional excretion of phosphate and lower levels of hemoglobin, urinary creatinine, and serum phosphate (Supplemental Table  8) (all P,0.05).
Only 1 day after administration, patients randomized to empagliflozin had higher TmP/GFR compared with placebo (1.17 [1.00-1.33] mmol/L versus 0.92 [0.70-1.10] mmol/L; F58.983; P50.004, adjusted for baseline TmP/ GFR) as shown in Figure 3. However, after additional adjustment for eGFR change between baseline and day 1, this difference was no longer statistically significant. Urinary phosphate and urinary glucose were not correlated at day 1 in patients allocated to empagliflozin.

Discussion
In this study, we showed that TmP/GFR is frequently reduced in patients with heart failure, especially in patients with more advanced heart failure, worse kidney function, and increased signs of tubular damage. TmP/GFR predicted plasma NGAL doubling over time after adjustment for several kidney parameters but not deterioration of kidney function. TmP/GFR was independently associated with both all-cause mortality and heart failure hospitalization. Finally, empagliflozin increased TmP/GFR compared with placebo 1 day after the start of administration but not after adjustment for eGFR change.
Symptoms of heart failure are primarily driven by congestion, for which loop diuretics are the cornerstone of treatment (28). However, a large number of patients with heart failure do not appropriately respond to loop diuretics (28). Because approximately 65% of sodium handling occurs at the proximal tubule, this kidney segment is of particular relevance to patients with heart failure (3). However, the proximal tubule is one of the most vulnerable segments; for this reason, proximal tubule cells are early and central sensors, effectors, and injury recipients (15,29), leading to multiple features of CKD (15,30,31). Proximal tubule cells are as such suggested as a primary player in the progression of acute kidney injury and CKD (15). Following damage, proximal tubules also become evidently less functional due to pathologic changes impairing reabsorption and secretion (32). In heart failure, hypoperfusion of the kidney and congestion (1) might cause proximal tubular dysfunction and damage. This, vice versa, might in theory (counterbalancing natriuretic peptide release) cause congestion through increased sodium and water reabsorption downstream, similar to mechanisms involved in diuretic resistance (28), and lead to progression of heart failure. This has, for example, been shown in a study by Rao et al. (33), which addressed mechanisms of diuretic resistance and showed that compensatory distal tubular sodium reabsorption makes the largest relative contribution (79%) to diuretic-induced increase in FE Na . Additionally, sympathetic nervous system activation in heart failure with formation of reactive oxygen species may also have detrimental effects on proximal tubular cells (34).
For all of these reasons, a relevant metric of proximal tubular function that allows heart failure specialists to look beyond eGFR will be highly relevant. Several heart failure studies have focused on tubular damage markers, such as NGAL or KIM-1 (35), yet these do not assess tubular function. As such, TmP/GFR might be a clinically relevant metric of proximal tubular function in heart failure.
Originally, TmP/GFR was invented to help distinguish hypercalcemia due to hyperparathyroidism from other causes (19). We now know that expression of (sodium) phosphate (co-)transporters (i.e., NaPi-IIa, NaPi-IIc, and PiT-2 that determine TmP/GFR) decreases in response to injury (9)(10)(11)(12)(13)(14)32). In studies conducted in rats, expression of phosphate (co-)transporters also decreased in the setting of acute hypertension (36). TmP/GFR is thought to be independent of glomerular function but instead represents the functioning of phosphate reabsorption in the proximal tubule (7,16). TmP/GFR thus shows promise as a functional proximal tubular parameter. TmP/GFR was already suggested to be useful for monitoring kidney tubular response to therapy in tubular damage over two decades ago (7), and it has since been reported as a marker of tubular dysfunction (13,14). Additionally, TmP/GFR of living kidney donors was independently associated with recipient measured GFR independent of donor measured GFR but not NGAL or KIM-1 (16).
In this study, we observed that TmP/GFR is frequently reduced in patients with heart failure, more often so in BIOSTAT-CHF (67%) than in EMPA-RESPONSE-AHF (27%). The difference in the proportion of the occurrence of low TmP/GFR between the two cohorts might be explained by EMPA-RESPONSE-AHF including a higher percentage of de novo heart failure compared with BIOSTAT-CHF (47% versus 28%, respectively) and less frequently of an ischemic etiology (29% versus 61%, respectively). Longer existence of heart failure and ischemic injury might result in more proximal tubular dysfunction. Indeed, in patients with heart failure and proven renovascular disease, an ischemic etiology was more often present compared with in patients without renovascular disease (37).
Second, lower TmP/GFR was indeed associated with higher doses of loop diuretics, higher levels of bioactive adrenomedullin (a congestion marker [38]) and NT-proBNP, and higher fractional excretion of lithium, providing hints toward patients with lower TmP/GFR having more congestion and/or diuretic resistance. Lower TmP/GFR was also associated with higher levels of tubular damage markers, such as NGAL, but less strongly than with higher FE Na and FE Urea , which may also be considered functional parameters (32,39). This seems logical precisely because NGAL levels do not necessarily indicate tubular function, and furthermore originate from multiple tubular segments. The association between TmP/ GFR and FE Urea might well be explained by the fact that a large proportion of urea is already absorbed in the proximal tubule and is rather a reflection of proximal peritubular forces than distal forces (39). In addition, TmP/GFR was independently associated with the future doubling of plasma NGAL and poor clinical outcomes, further underlining its clinical relevance, independent of glomerular function. Of course, its association with clinical outcomes might partly be influenced by other forces driving phosphate metabolism, but our findings that (1) TmP/GFR did not cluster with fibroblast growth factor 23 (FGF23; neither was FGF23 a significant determinant of TmP/GFR) and that (2) serum phosphate, by far the largest determinant of TmP/GFR, was not associated with any outcome in this study make it less likely that these effects were of a large magnitude. Lastly, one might postulate that TmP/GFR decline could also be a physiologic compensatory mechanism in preventing hyperphosphatemia, but values below the lower reference limit of serum phosphate (the lowest two quartiles of TmP/GFR) imply an "overshoot" of TmP/ GFR decline that is likely attributable to proximal tubular dysfunction. Finally, SGLT2 inhibition significantly increased TmP/ GFR 1 day after administration after adjustment for baseline TmP/GFR but not after additional adjustment for eGFR change (40). This was contrary to our expectation: first because of the anatomic proximity of SGLT2 to NaPi-IIa (believed to be responsible for 70% of phosphate reabsorption), with both primarily located in the S1 segment (41,42). Second, SGLT2 inhibitors have an effect on phosphate homeostasis by increasing serum phosphate in both the short and long terms (43,44). Third, prevention of sodium reabsorption by SGLT2 inhibition causes the sodium gradient to be preserved for, among others, sodium-dependent phosphate transporters (43). Alternatively, changes in urine osmolality induced by empagliflozin (although no correlation between urinary phosphate and urinary glucose was observed) (40) or neurohormonal influences might also influence phosphate transport.
Certain associations of TmP/GFR deserve additional discussion. First, low TmP/GFR was associated with higher albuminuria. This might be explained by the fact that decreased albumin reabsorption in tubules due to tubular damage can contribute to albuminuria (1). Second, FGF23 not clustering with or being a determinant of TmP/GFR   was unexpected because FGF23 regulates urinary phosphate excretion (45). Likely, FGF23 is more strongly affiliated with different pathways in heart failure, like development of ventricular hypertrophy (46), sodium homeostasis (47), volume overload, and more pronounced RAAS activation (24,45). Third, low TmP/GFR was not associated with eGFR decline. This might be explained by eGFR changes not necessarily reflecting changes in intrinsic kidney (tubular) damage (2). We showed that low TmP/ GFR and low eGFR often occur together and confer a very poor survival and cardiorenal profile if they do, but they might not necessarily determine one another, consistent with regression and cluster analyses.
To our knowledge, this is the first study to pose TmP/ GFR as a potentially clinically relevant parameter of proximal tubular function in heart failure ( Figure 4). Looking beyond glomerular function with a relevant metric of proximal tubular function will provide valuable cardiorenal insights because it might ultimately contribute to therapies that modulate proximal tubular function and improve sodium homeostasis in heart failure. Further research is warranted to study dynamics of TmP/GFR and its value in prospectively predicting (long-term) future CKD onset. Additionally, because we could only include derivatives of sodium homeostasis and diuretic response and because EMPA-RESPONSE-AHF had a small sample size, extended exploration of the role of TmP/GFR in this sense will be important. Finally, exploration of other parameters reflecting tubular (sodium) transport will also be highly valuable to further increase our knowledge of proximal tubular function in heart failure.
We investigated TmP/GFR in relation to a wide panel of variables in a large, heterogeneous, multinational heart failure population, which is a strength of our study. Furthermore, this study included a good balance of patients with and without CKD, allowing reliable subgroup analyses. Additional insights were provided by studying the effects of empagliflozin on TmP/GFR in a small but very welldefined cohort in patients with acute heart failure. However, we are mainly able to describe associations, not causality. We also realize that this study does not provide definitive proof of the relationship of TmP/GFR with actual proximal tubular damage or overall reduced sodium reabsorption of the proximal tubule, as previous experimental studies in rats have detected upregulation of the NHE3 transporter in response to heart failure (48)(49)(50). Necessity for collecting urinary samples is a disadvantage. Also, the formula is quite complex, but this can easily be solved by implementation in electronic patient systems. Furthermore, urinary NGAL would have been preferred over plasma NGAL, but unfortunately, no measurements were available at 9 months. We were only able to show dynamic changes of TmP/GFR in EMPA-RESPONSE-AHF because no repeat measurements were performed in  BIOSTAT-CHF. Finally, no true GFR or parathyroid hormone levels were measured, and patients were not fasted and did not have consistent sample collection times, which would have provided more consistency in TmP/GFR readouts. TmP/GFR, a metric of proximal tubular function, is frequently reduced in patients with heart failure, especially in patients with more advanced heart failure. Lower TmP/ GFR is associated with plasma NGAL doubling over time and poor outcomes, independent of glomerular function. SGLT2 inhibition increased TmP/GFR but not after adjustment for eGFR change.   The sodium phosphate cotransporters NaPi-IIa, NaPi-IIc, and PiT-2 determine the maximum capacity of the kidney to reabsorb phosphate in the proximal tubule, which is reflected by TmP/GFR. TmP/ GFR is calculated by first determining the fractional tubular reabsorption of phosphate (TRP) using Equation (1). When TRP is #0.86, indicating maximum phosphate reabsorption, there is a linear relationship between serum phosphate concentration and excretion, and TmP/ GFR is subsequently calculated by Equation (2). When TRP is .0.86, TmP/GFR is calculated using a different second formula, which was not included in this figure for purposes of comprehensibility. This resulted in TmP/GFR values, a parameter of proximal tubular function. (B) In heart failure, hypoxia, congestion, and formation of reactive oxygen species can lead to the development of tubular damage, notably of the proximal tubule. Damage to proximal tubular cells triggers multiple features of CKD, such as fibrosis, inflammation, and capillary rarefaction, and might lead to impaired sodium handling of the kidney. In response to injury, expression of the sodium phosphate cotransporters declines, resulting in decreased tubular phosphate reabsorption and, thus, a decreased TmP/GFR. Decreased TmP/GFR is associated with increased fractional excretion of sodium and urea; increased levels of KIM-1, NGAL, and NT-proBNP; higher doses of loop diuretics; and a poor prognosis.

Supplemental Material
This article contains the following supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN. 03720321/-/DCSupplemental. Supplemental Figure 1. Kaplan-Meier curve for the combined end point for quartiles of TmP/GFR. Supplemental Figure 2. Kaplan-Meier curves according to low/ high TmP/GFR and presence of CKD predicting the combined end point.
Supplemental Figure 3. Correlation plot of TmP/GFR with fractional excretion of lithium.
Supplemental Material. Supplemental Methods. Supplemental Table 1. Baseline characteristics of study subset and patients without TmP/GFR measurements in BIOSTAT-CHF.
Supplemental Table 2. Baseline characteristics according to low/ high TmP/GFR and presence of CKD.
Supplemental Table 3. Baseline characteristics according to low, normal, and high TmP/GFR. Supplemental Table 4. Multivariable regression analysis for log TmP/GFR with phosphate included.
Supplemental Table 5. Cox regression analysis of serum phosphate.
Supplemental Table 6. Cox proportional hazards analysis according to low/high TmP/GFR and presence of CKD predicting all-cause mortality. Supplemental Table 7. Cox proportional hazards analysis according to low/high TmP/GFR and presence of CKD predicting the combined end point.
Supplemental Table 8. Baseline characteristics of the EMPA-RESPONSE-AHF cohort stratified by above or below median TmP/GFR.