Dynamics of Organic Anion Transporter-Mediated Tubular Secretion during Postnatal Human Kidney Development and Maturation

Visual Abstract

Abstract

Background and objectives

The neonatal and juvenile human kidney can be exposed to a variety of potentially toxic drugs (e.g., nonsteroidal anti-inflammatory drugs, antibiotics, antivirals, diuretics), many of which are substrates of the kidney organic anion transporters, OAT1 (SLC22A6, originally NKT) and OAT3 (SLC22A8). Despite the immense concern about the consequences of drug toxicity in this vulnerable population, the developmental regulation of OATs in the immature postnatal kidney is poorly understood.

Design, setting, participants, & measurements

Recognizing that today it is difficult to obtain rich data on neonatal kidney handling of OAT probes due to technical, logistic, and ethical considerations, multiple older physiologic studies that used the prototypical organic anion substrate para-aminohippurate (PAH) were reanalyzed in order to provide a quantitative description of OAT-mediated tubular secretion across the pediatric age continuum. Parametric and semiparametric models were evaluated for kidney function outcome variables of interest (maximum tubular secretory capacity of PAH [Tm_PAH], effective renal plasma flow [ERPF], and GFR).

Results

Data from 119 neonates, infants, and children ranging in age from 1 day to 11.8 years were used to fit Tm_PAH, ERPF, and GFR as functions of postnatal age. Tm_PAH is low in the immediate postnatal period and increases markedly after birth, reaching 50% of the adult value (80 mg/min) at 8.3 years of age. During the first 2 years of life, Tm_PAH is lower than that of GFR when viewed as the fraction of the adult value.

Conclusions

During postnatal human kidney development, proximal tubule secretory function—as measured using PAH, a surrogate for OAT-mediated secretion of organic anion drugs, metabolites, and toxins—is low initially but increases rapidly. Despite developmental differences between species, this overall pattern is roughly consistent with animal studies. The human data raise the possibility that the acquisition of tubular secretory function may not closely parallel glomerular filtration.

Introduction

Membrane transport proteins facilitate the movement of endogenous metabolites and exogenous compounds (drugs, toxins, gut microbiome products) across biologic membranes (1,2). Multispecific transporters of the solute carrier (SLC) and the ATP-binding cassette (ABC) superfamilies localized to the basolateral and apical membranes of the kidney proximal tubules are involved in sequential uptake and efflux of drugs, metabolites, toxins, nutrients, and signaling molecules from blood into the urine (3,4).

From the standpoint of drug therapy, variable proximal tubule transporter activity due to drug interactions, genetics, and other factors is associated with altered drug disposition and consequent ineffective or toxic outcomes (2,5). Considerable effort is spent in the drug development setting to characterize the potential of a drug candidate to inhibit and/or undergo transport by these membrane proteins. For example, the US Food and Drug Administration (FDA) recommends that investigational drugs should be evaluated to determine whether they are substrates or inhibitors of three SLC22 transporters (6): OAT1 (SLC22A6), OAT3 (SLC22A8), and OCT2 (SLC22A2)—all of which are involved in the transport of organic solutes into the kidney tubules (7–9). Results of in vitro and in vivo studies are then presented in drug labeling to make prescribers aware of potential transporter-mediated drug interactions.

Despite the contribution of kidney proximal tubule drug transporters to drug clearance and dosing requirements, relatively little is known about the ontogeny of these transporters in humans (10–12). Much of what we know about the expression and regulation of OATs during pre- and postnatal kidney development comes from rodent studies (13–17).

The lack of information about the activity of OAT1 and OAT3 in humans as a function of age is troublesome in the current regulatory climate, in which planning for pediatric studies must begin at the end of phase 2—meaning that, despite potentially greater toxicity risks in the neonates and infants, pediatric studies now occur with far less adult data to inform the trials (18,19). Further, as stipulated in the 2012 FDA Safety and Innovation Act, drug studies in neonates are now required unless an appropriate rationale is given for not performing such studies (20). This creates a scenario whereby drugs are studied in the most vulnerable pediatric patients with little information on how drug handling in the kidney differs relative to adults. Thus, as emphasized in a recent whitepaper (10), there is a critical need for better understanding of the factors that regulate the expression and activity of kidney transporters in the newborn maturing kidney in order to provide safe and effective drug therapies for pediatric patients. Other recent calls for improved pharmacokinetic-pharmacodynamic models for antibiotics eliminated through the kidneys in neonates also highlight the importance of understanding developmental aspects of tubular secretion capacity (21).

Para-aminohippurate (PAH) is the classic model organic anion, and the molecular mechanisms involved in PAH tubular secretion have been extensively studied (9). The kidney clearance of PAH has been used as a diagnostic tool to estimate renal plasma flow because its extraction ratio on a single passage through the kidney is approximately 90% due to extensive secretion by the tubules, free filtration from plasma by the glomerulus, low binding to plasma proteins, and lack of reabsorption by the tubules (2,22–24). Human OAT1 has been identified as an efficient PAH transporter. In hOAT1-expressing Xenopus laevis oocytes, the uptake of 1 µM PAH was 2.7±0.4 pmol/oocyte per 30 minutes with a K_m of 3.1±0.8 and 4.0±1.5 µM in the absence and presence of a chloride buffer, respectively (25). Moreover, in knockout animals, OAT1 appears to be the major PAH transporter in young and adult mice (13,17). Such knockouts also reveal major transporters of many organic anionic drugs, metabolites, and toxins (26–28). OAT3 can also transport PAH, but it is generally believed to have more limited or variable PAH transport activity relative to OAT1, although in some studies it has been shown to be comparable to OAT1 (25,29). It is, however, generally believed that the affinity of OAT1 for PAH is higher than for OAT3, although, even here, there are reports of comparable OAT1 and OAT3 affinity. Other SLC and ABC transporters expressed in the kidney, including other OAT isoforms and ABCC (MRP) isoforms, can transport PAH in vitro to varying degrees (30), but their in vivo significance with respect to PAH transport is unclear and likely to be small compared with OAT1 and OAT3.

In humans, PAH can be used to measure the effective renal plasma flow (ERPF) and the maximum tubular secretory capacity (Tm_PAH); the latter is the difference between the total rate of excretion and the quantity filtered by the glomeruli, which is primarily attributable to OAT activity (2,29,31). Clearance measurements using a single dose of PAH are generally inaccurate, necessitating continuous intravenous infusions to sustain the plasma PAH concentration (32).

Although PAH is the prototypical marker for organic anion tubular secretion capacity, today various technical, logistic, and ethical considerations make it difficult to rigorously study PAH secretion in neonates, infants, and children using the aforementioned methods. Even so, the clearance of PAH in pediatric patients was extensively studied in a number of separate studies more than half a century ago (33–36). This was before the understanding of the molecular mechanism of tubular secretion of PAH, before the identification of OAT1, originally identified as NKT in mice (14), and before the development of statistical software. Here, we have performed a re-evaluation of these data, pooled from several sound physiologic studies, which is perhaps the most detailed functional assessment of human organic anion handling during postnatal kidney development to date.

Materials and Methods

Data Abstraction

A literature search of MEDLINE/PubMed was conducted for articles published between January of 1940 and August of 2018 in English that contained individual-level data on PAH handling in children. Selected keywords included para-aminohippurate, aminohippuric acid, p-aminohippurate, PAH, tubular secretion, and Tm_PAH. The references of identified articles were reviewed to ensure that all relevant articles were included. Data were abstracted from previously published studies by Fawer et al. (33), Calcagno et al. (34), Dean et al. (35), and Rubin et al. (36) and subsequently pooled. Clinical study characteristics are shown in Table 1. Precise procedures for intravenous delivery of experimental substances by venous cannula and collection of urine and blood samples in pediatric patients are described in detail in these studies and their references. The clearance of PAH at low plasma levels (0.5–3.0 mg/dl) was used to measure the ERPF, whereas clearance of PAH at high plasma levels (50–100 mg/dl) was used to measure the maximum anionic secretory capacity (Tm_PAH). GFR was measured after single mannitol injections or after continuous infusions of inulin. Where multiple inulin clearance measurements were reported in the same subject, all available values were averaged.

Table 1.

Summary of pediatric physiologic studies using para-aminohippurate

Variable	Study
	Rubin et al. (36)	Fawer et al. (33)	Fawer et al. (33)	Calcagno et al. (34)	Dean et al. (35)
Method
GFR	Mannitol	Inulin	Inulin	Inulin	Inulin
ERPF	PAH (0.5–3 mg/dl)	PAH	PAH	PAH (<4 mg/dl)
TmPAH	PAH (50–100 mg/dl)
Number of participants	63	16	28	4	8
Median postmenstrual age (range)	84.7 wk (40.3 wk–12.6 yr)a	32.7 wk (28.3–35.0 wk)	37.9 wk (35.2–43.3 wk)	41.9 wk (41.1–74.7 wk)a	40.4 wk (40.3–41.1 wk)
Median gestational age (range)	NR	31.5 wk (28.0–34.0 wk)	37.5 wk (35.0–43.0 wk)	NR	40 wkb
Median postnatal age (range)	1.1 yr (7 d–11.8 yr)	2.3 d (1–16 d)	2.0 d (1–19 d)	13.0 d (8–243 d)	2.5 d (2–8 d)
Median weight (range), kg	9.6 (2.7–35.5)	1.5 (1.1–2.1)	2.3 (1.4–5.4)	3.5 (2.5–7.6)	3.0 (2.2–4.0)
Median GFR (range), ml/min	31.2 (5.3–86.0)	1.7 (0.93–4.0)	2.5 (1.1–7.3)	6.3 (2.9–29.8)	3.4 (1.7–5.5)
Median TmPAH (range), mg/min	15.6 (0.43–68.1)	NR	NR	NR	NR
Sex reported	Yes	No	No	No	Yes
More than one observation per subject	No	No	No	No	No
Reported pathology	Normal, well children	Prematurity, hyperbilirubinemia, intrauterine growth restriction	Hyperbilirubinemia, intrauterine growth restriction	Mixed: Down syndrome, oral facial cleft, intellectual disability, septal defect, tetralogy of Fallot, essential hypertension, pulmonary stenosis	Inoperable meningomyeloceles

ERPF, effective renal plasma flow; PAH, para-aminohippurate; Tm_PAH, maximum tubular secretory capacity of PAH; NR, not reported.

^aGestational age of 40 wk was imputed in order to estimate postmenstrual age.

^bStudy reported full-term birth, which was estimated to be 40 wk.

Statistical Analyses

Statistical and graphic analyses of extracted data were performed using R Studio (Version 0.99.489) with base R and dplyr (0.4.3) and ggplot2 (2.2.1.9000) packages (R Core Team, Hadley Wickham). For modeling each outcome variable (Tm_PAH, ERPF, and GFR) as a function of postnatal age, several parametric statistical models were considered, including linear, polynomial, and Gompertz models, with the response on the natural or log-transformed scale. Separate models were developed for each outcome variable with and without normalization for body size (i.e., per 1.73 m² of body surface area). In addition, semiparametric (smoothing splines) models were used in exploratory analyses. For each response (and for the transformed and untransformed response separately), the final model was chosen using the Akaike information criterion. The choice of model on the transformed versus untransformed response scale was made on the basis of proportion of variability explained and residual plots. All models were evaluated for biologic plausibility, and compared visually with the smoothing spline model.

Results

We performed a detailed examination of human studies of neonates, infants, and children that measured organic anion transport using PAH in the context of measurements of kidney function. The types of data available from each of these studies are summarized in Table 1. We then pooled appropriate data to analyze, using the statistical approaches and software described in the Materials and Methods section, ERPF, Tm_PAH, and GFR.

ERPF

The measured ERPF using PAH in 111 neonates, infants, and older children ranging in age from 1 day to 11.8 years is shown in Figure 1. For non–body size–adjusted data, a linear model fitted to logarithmic transformations of age and ERPF best described the data:

For body size–adjusted data, a Gompertz model fitted to logarithmic transformations of age and ERPF best described the data:

Figure 1.

Effective renal plasma flow (ERPF) as predicted by postnatal age using a generalized linear regression model. (A) Scatter plot of observed ERPF values (open circles) versus postnatal age. A regression line predicted by the model (solid line) and a smoothing spline curve fit to observed data (dashed line) are displayed. (B) Proportion of asymptotic ERPF during the first 2 years of life shown as model-predicted (solid line) and observed (dashed line) values.

The final models for unadjusted and body surface area (BSA)-adjusted ERPF described 87% and 88% of the variability, respectively (Table 2). At birth (day 1), the model-predicted ERPF is 2.9 ml/min, representing approximately 0.62% of the typical adult value. The model-predicted ERPF values (ml/min) at 3, 6, 12, and 24 months are 46.2, 70.6, 109.0, and 166.8, respectively. The predicted ages at which childhood ERPF reaches 25% and 50% of adult ERPF (600 ml/min) are 1.7 and 5.6 years, respectively.

Table 2.

Final derived models

Statistic or Estimate	Dependent Variable
	BSA-Unadjusted			BSA-Adjusted
	ERPF (ml/min)	TmPAH (mg/min)	GFR (ml/min)	ERPF (ml/min per 1.73 m2)	TmPAH (mg/min per 1.73 m2)	GFR (ml/min per 1.73 m2)
Equation	ln(ERPF)=1.06998 + [0.61380*ln(PNA)]	ln(TmPAH)=−0.34731 + [0.50361*ln(PNA)]	ln(GFR)=−0.08774 + [0.56491*ln(PNA)]	ln(ERPF)=8.56933−e[1.73111−0.13918*ln(PNA)]	ln(TmPAH)=5.4188−e[1.2441−0.1480*ln(PNA)]	ln(GFR)=5.83855−e[1.32901−0.19257*ln(PNA)]
R2	0.87	0.70	0.91	0.88	0.38	0.87
Root mean square error	0.65	0.55	0.47	0.48	0.50	0.38
Percentage within 10%	55	52	40	70	65	67
Percentage within 30%	73	79	59	94	94	94
Estimate at age
10 d	12.0	2.3	3.4	87.4	19.1	30.4
1 mo	23.5	3.9	6.3	156.3	27.7	48.3
3 mo	46.2	6.8	11.6	257.4	37.9	70.1
6 mo	70.6	9.7	17.2	339.8	45.1	85.5
1 yr	109.0	13.8	25.7	439.3	53.0	102.1
2 yr	166.8	19.6	38.0	552.0	61.0	118.8
6 yr	327.4	34.0	70.6	760.1	74.2	145.4

ERPF, effective renal plasma flow; Tm_PAH, maximum tubular secretory capacity of PAH; PAH, para-aminohippurate; PNA, postnatal age in days.

Tm_PAH

The measured Tm_PAH in 52 neonates, infants, and older children ranging in age from 7 days to 11.8 years is shown in Figure 2. For non–body size–adjusted data, a linear model fitted to logarithmic transformations of age and Tm_PAH best described the data:

For body size–adjusted data, a Gompertz model fitted to logarithmic transformations of age and Tm_PAH best described the data:

Figure 2.

Maximum tubular secretory capacity of para-aminohippurate (Tm_PAH) as predicted by postnatal age using a generalized linear regression model. (A) Scatter plot of observed Tm_PAH values (open circles) versus postnatal age. A regression line predicted by the model (solid line) and a smoothing spline curve fit to observed data (dashed line) are displayed. (B) Proportion of asymptotic Tm_PAH during the first 2 years of life shown as model-predicted (solid line) and observed (dashed line) values.

The final models for unadjusted and BSA-adjusted Tm_PAH described 70% and 38% of the variability, respectively (Table 2). At 7 days, the model-predicted Tm_PAH is 1.88 mg/min, representing approximately 2.4% of the typical adult value. The model-predicted Tm_PAH values (mg/min) at 3, 6, 12, and 24 months are 6.8, 9.7, 13.8, and 19.6, respectively. The predicted ages at which childhood Tm_PAH reaches 25% and 50% of adult Tm_PAH (80 mg/min) are 2 and 8.3 years, respectively.

GFR

The measured GFR using either mannitol or inulin in 119 children ranging in age from 1 day to 11.8 years is shown in Figure 3. A linear model fitted to logarithmic transformations of age and GFR best described the data:

For body size–adjusted data, a Gompertz model fitted to logarithmic transformations of age and GFR best described the data:

Figure 3.

GFR as predicted by postnatal age using a generalized linear regression model. (A) Scatter plot of observed GFR values (open circles) versus postnatal age. A regression line predicted by the model (solid line) and a smoothing spline curve fit to observed data (dashed line) are displayed. (B) Proportion of asymptotic GFR during the first 2 years of life shown as model-predicted (solid line) and observed (dashed line) values.

The final models for unadjusted and BSA-adjusted GFR described 91% and 87% of the variability, respectively (Table 2). At birth (day 1), the model-predicted GFR is 0.92 ml/min, representing approximately 1.3% of the typical adult value. The model-predicted GFR values (ml/min) at 3, 6, 12, and 24 months are 11.6, 17.2, 25.7, and 38.0, respectively. The predicted ages at which childhood GFR reaches 25% and 50% of adult GFR (120 ml/min) are 1.3 and 5.5 years, respectively.

Discussion

In humans, limited information is available regarding the developmental aspects of mRNA and/or protein expression of transporters in the kidney (10). In rodents, these data exist for transcript level and OAT function as measured by PAH handling (17). Our reanalysis of older studies, including those >70 years old, now provides important functional information on OAT-mediated tubular secretion in humans as a function of postnatal developmental time. By pooling abstracted data from multiple studies and reanalyzing, we have provided a detailed portrait of organic anion handling in neonates, infants, and children.

It is known that the developmental expression patterns and/or function of kidney organic anion transporters are roughly comparable in a variety of mammals, including mouse, rat, and sheep (reviewed in (10), Table 3, and the Supplemental Material [14–17, 37]). Although postnatal nephrogenesis occurs in rodents, in mice this appears to be complete between several days to 1 week after birth (38,39). Mouse functional data indicate that at least a four-fold increase in PAH clearance occurs after the first postnatal week—i.e., after nephrogenesis is complete. Thus, during a roughly comparable developmental span (after completion of nephrogenesis in mice and humans to the point of mature kidney function), the overall increase in PAH handling seems to be, very generally, comparable. Although there are, and should be, concerns about directly extrapolating from rodent to human postnatal development, at least from the perspective of functional development of organic anion transport capacity after completion of nephrogenesis, the data from both species suggest that murine models may have relevance for humans (17). Although appropriate caution must be exercised in comparing rodent and human physiology, the data nevertheless suggest that rodent studies aimed at understanding the regulation of OAT expression and function during postnatal kidney development may have relevance for understanding these processes in humans, where such studies may be very difficult or impossible to perform.

Table 3.

Summary of model estimates

Dependent Variable	Coefficient	95% CI
Linear regression models for non–body size–adjusted parameters
ERPF, ml/min
Intercept, β0	1.07	0.85 to 1.29
Slope, β1	0.61	0.57 to 0.66
TmPAH, mg/min
Intercept, β0	−0.35	−0.90 to 0.21
Slope, β1	0.50	0.41 to 0.60
GFR, ml/min
Intercept, β0	−0.09	−0.23 to 0.06
Slope, β1	0.57	0.53 to 0.60
Gompertz exponential growth models for BSA-adjusted parameters
NormERPF, ml/min per 1.73 m2
β1	8.57	6.60 to 10.54
β2	−1.73	2.06 to −1.40
β3	0.14	0.06 to 0.22
NormTmPAH, mg/min per 1.73 m2
β1	5.42	1.24 to 9.60
β2	−1.24	−2.02 to −0.47
β3	0.15	−0.23 to 0.53
NormGFR, ml/min per 1.73 m2
β1	5.84	5.07 to 6.61
β2	−1.33	−1.51 to −1.15
β3	0.19	0.12 to 0.27

95% CI, 95% confidence interval; ERPF, effective renal plasma flow; Tm_PAH, maximum tubular secretory capacity of para-aminohippurate; NormERPF, BSA-normalized ERPF; NormTm_PAH, BSA-normalized Tm_PAH; NormGFR, BSA-normalized GFR.

The low values for Tm_PAH and ERPF in the immediate postnatal period point to poorly developed organic anion transport mechanisms in neonates. At birth, a combination of low arterial BP, low ultrafiltration pressure, and low glomerular capillary surface area lead to low glomerular filtration, which contributes to lower PAH extraction by filtration (40). Additionally, it is believed that all nephrons of the mature kidneys are formed by 36 weeks of gestation, even as hyperplasia continues until the sixth postnatal month after which cell hypertrophy is responsible for increases in the size of the kidney (40). However, our analysis of the developmental pattern of Tm_PAH—which accounts only for the quantity of PAH secreted by the tubules—strongly suggests, particularly in the context of largely concordant rodent studies described below (12,17,41), that upregulation of OAT1 and OAT3 in the proximal tubules during the postnatal period is a driving factor.

Physiologically based pharmacokinetic modeling (PBPK), which is commonly used to predict drug exposures in children, generally assumes that the ontogeny of active tubular secretion is the same as that for filtration (42). The current results suggest that the acquisition of filtration and secretion do not occur at identical rates (Figure 4). During the first months of life, the maximum tubular secretory capacity of organic anions is lower than that of GFR when viewed as the fraction of the asymptotic (i.e., adult) value. In addition, there is significant intersubject variability in the ratio of GFR/Tm_PAH during the first year of life (Figure 4). For a drug that is a substrate of OAT1/3 and that undergoes extensive tubular secretion, the kidney clearance in young children may be lower than would be predicted on the basis of GFR alone. This may be a very important point; PBPK models that incorporate the ontogeny of human OAT1- and OAT3-mediated anionic secretion may result in improved model fit for drugs that are extensively secreted by the kidneys.

Figure 4.

(A) Fraction of adult values of kidney function parameters as predicted by generalized linear regression models during the first 2 years of life. Effective renal plasma flow (ERPF, red), maximum tubular secretory capacity of para-aminohippurate (Tm_PAH, green), and GFR (blue) were each predicted as a function of postnatal age using logarithmically transformed linear regression models (B) GFR/Tm_PAH in 52 neonates, infants, and older children ranging in age from 7 days to 11.8 years.

This study has some limitations. Although kidney function in neonates will be influenced by kidney maturity, data on gestational age at birth were generally not available. Nonetheless, considering that the abstracted data used in this study were obtained in the 1940s and 1950s, it is likely that most subjects were at or near full term because it would seem that the survival of early preterm neonates would have been much lower than today. Accordingly, the present findings may not apply to preterm neonates. Next, pooled studies used for analyses included both healthy and sick children, with some reported pathologies that could potentially affect kidney function, including intrauterine growth restriction and essential hypertension. Further, our final model of Tm_PAH described only 70% of the variability. This points to the technical challenges of measuring Tm_PAH, where high plasma levels must be achieved with a continuous infusion in order to saturate the capacity of the tubules, and also possibly to heterogeneity in OAT1/3 expression and activity. Further, although OATs on the basolateral appear to mediate the rate-limiting step in PAH secretion, multiple apical transporters are involved in apical efflux, although this latter step is generally not viewed as rate limiting. Single nucleotide polymorphisms may also affect the expression or activity of OATs as well as apical PAH transporters, and the functional relevance of this to variability in PAH secretion can only be assessed with genetic analysis (43,44). Thus, a more robust model could theoretically be elucidated with the incorporation of additional patient covariates, such as body size, sex, gestational age, and genetic information.

In conclusion, we provide a detailed picture of PAH handling in neonates, infants, and children. Values for ERPF and Tm_PAH are very low in the immediate postnatal period relative to adults. ERPF and Tm_PAH increase markedly after birth, the latter measure presumably reflecting increased OAT expression and/or function. Many drugs administered in the neonatal intensive care unit (e.g., nonsteroidal anti-inflammatory drugs, diuretics, antivirals, β-lactam antibiotics), as well as those given to infants and children, are primarily eliminated into the urine via the OATs. The results of our analyses support the concern about the possible toxic effects of OAT-transported drugs, particularly immediately after birth and perhaps up to 2 years of age. On the other hand, hepatic metabolism is also limited in the newborn. Thus, drug dosing and potential toxicities in this vulnerable population need to be viewed more broadly from the perspective of the developmental expression and activity of all relevant enzymes and transporters involved in drug disposition. The analysis provides the clearest quantitative assessment to date of the dynamics of tubular secretion during human kidney development—at least in the context of organic anion drugs—and will presumably also facilitate development of more accurate pediatric PBPK models.

Disclosures

None.