Prediction of Pregnancy-Induced Changes in Secretory and Total Renal Clearance of Drugs Transported by Organic Anion Transporters

Abstract

Pregnancy can significantly change the pharmacokinetics of drugs, including those renally secreted by organic anion transporters (OATs). Quantifying these changes in pregnant women is logistically and ethically challenging. Hence, predicting the in vivo plasma renal secretory clearance (CL_sec) and renal CL (CL_renal) of OAT drugs in pregnancy is important to design correct dosing regimens of OAT drugs. Here, we first quantified the fold-change in renal OAT activity in pregnant versus nonpregnant individual using available selective OAT probe drug CL_renal data (training dataset; OAT1: tenofovir, OAT2: acyclovir, OAT3: oseltamivir carboxylate). The fold-change in OAT1 activity during the 2^nd and 3^rd trimester was 2.9 and 1.0 compared with nonpregnant individual, respectively. OAT2 activity increased 3.1-fold during the 3^rd trimester. OAT3 activity increased 2.2, 1.7 and 1.3-fold during the 1^st, 2^nd, and 3^rd trimester, respectively. Based on these data, we predicted the CL_sec, CL_renal and total clearance ((CL_total) of drugs in pregnancy, which are secreted by multiple OATs (verification dataset; amoxicillin, pravastatin, cefazolin and ketorolac, R-ketorolac, S-ketorolac). Then, the predicted clearances (CLs) were compared with the observed values. The predicted/observed CL_sec, CL_renal, and CL_total of drugs in pregnancy of all verification drugs were within 0.80–1.25 fold except for CL_sec of amoxicillin in the 3^rd trimester (0.76-fold) and cefazolin in the 2^nd trimester (1.27-fold). Overall, we successfully predicted the CL_sec, CL_renal, and CL_total of drugs in pregnancy that are renally secreted by multiple OATs. This approach could be used in the future to adjust dosing regimens of renally secreted OAT drugs which are administered to pregnant women.

SIGNIFICANCE STATEMENT

To the authors’ knowledge, this is the first report to successfully predict renal secretory clearance and renal clearance of multiple OAT substrate drugs during pregnancy. The data presented here could be used in the future to adjust dosing regimens of renally secreted OAT drugs in pregnancy. In addition, the mechanistic approach used here could be extended to drugs transported by other renal transporters.

Introduction

Pregnant women commonly take drugs (medication) throughout their pregnancy, especially when afflicted with chronic diseases such as diabetes, hypertension, or infectious diseases (e.g., human immunodeficiency virus). About 97% take at least one drug, and about 31% take at least five drugs during pregnancy (Mitchell et al., 2011; Haas et al., 2018). Due to lack of clinical trials of drug candidates in pregnancy, a large proportion of approved drugs are prescribed off-label to pregnant women (Laroche et al., 2020). However, because pharmacokinetics of many drugs are modulated by pregnancy, such use may result in suboptimal dosing of pregnant women (Westin et al., 2018). To design optimal drug dosing regimens for pregnant women, it is important to quantify the pregnancy-induced changes in pharmacokinetics of drugs. However, such studies are logistically and ethically challenging. Therefore, an alternative is to predict these changes by first characterizing, using probe drugs, the magnitude of pregnancy-induced changes for a given drug clearance pathway. We have previously used this approach to predict the pregnancy-induced changes in pharmacokinetics of drugs that are metabolized by cytochrome P450 enzymes (Ke et al., 2012, 2013; 2014a). Here, we extended this approach to drugs that are renally cleared via organic anion transporters (OATs) using published data of renal secretory clearance (CL) of probe drugs that are selectively transported by a single OAT.

In clinical practice, 32% of the top 200 prescribed drugs are renally eliminated (i.e., ≥25% of the absorbed dose is excreted unchanged in urine) (Morrissey et al., 2013). Many of these drugs taken by pregnant women are secreted by single or multiple OAT (OAT1–3), such as antibiotics (e.g., amoxicillin, ampicillin) and antivirals (e.g., tenofovir, adefovir) (Mitchell et al., 2011; Leong et al., 2019). We have previously shown that P-glycoprotein- and/or OATP4C1-mediated secretory CL of digoxin is significantly increased during pregnancy (Hebert et al., 2008; Ke et al., 2014b). Likewise, the clearance of OAT-secreted drugs is also increased during pregnancy [e.g., amoxicillin and tenofovir (Andrew et al., 2007; Best et al., 2015)], suggesting that OAT activity is increased during pregnancy. To predict the magnitude of changes in OAT-mediated renal CL of drugs in pregnancy, we first quantified the magnitude of pregnancy-induced change in the activity of individual OATs (OAT1, 2, 3). This we did based on previously published renal secretory CL ( $\text{C}{\text{L}}_{\text{sec}})$ and total renal CL ( $\text{C}{\text{L}}_{\text{renal}})$ of OAT probe substrate drugs (training dataset; OAT1: tenofovir, OAT2: acyclovir, OAT3: oseltamivir carboxylate) that are known to be predominately renally cleared (>75%) by a single OAT. Then, these pregnancy-induced fold-changes in individual OAT activity were used to predict the pregnancy-induced changes in renal secretory CL ( $\text{C}{\text{L}}_{\text{sec}}^{\text{pregnant}}$), renal CL ( $\text{C}{\text{L}}_{\text{renal}}^{\text{pregnant}}$), and total CL ( $\text{C}{\text{L}}_{\text{total}}^{\text{pregnant}}$) of drugs transported by a single or multiple OATs (verification dataset; amoxicillin, pravastatin, cefazolin, and ketorolac, R-ketorolac, S-ketorolac). To do so, we estimated the fraction of drug transported by the individual OAT (1, 2, and 3) based on their in vitro transporter-mediated and passive clearance data in OAT- or mock-transfected cell lines. Predictions by any approach must be verified before it can be applied with confidence to other drugs. Therefore, to verify the above predictions, we compared the predicted and the observed $\text{C}{\text{L}}_{\text{sec}}^{\text{pregnant}}$, $\text{C}{\text{L}}_{\text{renal}}^{\text{pregnant}}$, and $\text{C}{\text{L}}_{\text{total}}^{\text{pregnant}}$ values of these drugs. We deemed our predictions successful if they fell within 80% to 125% of the observed values.

Materials and Methods

Pharmacokinetics Data Collection and Estimation of CL_int,sec of Drugs.

Briefly, renal total clearance ( $\text{C}{\text{L}}_{\text{renal},B}$) of training drugs (tenofovir, acyclovir, oseltamivir carboxylate) and verification drugs (amoxicillin, cefazolin, pravastatin and ketorolac, R- ketorolac, S-ketorolac) in nonpregnant and pregnant women were obtained from the literature or estimated from the reported data. These data were available mostly for the 2^nd and 3^rd trimester. Except for acyclovir, where the pharmacokinetics (PK) parameters were from two different cohort of women (pregnant and nonpregnant), for the remaining drugs, the pharmacokinetic parameters were from the same cohort of women studied, antenatal (oseltamivir carboxylate and ketorolac) and postpartum (≥6 weeks). We assumed pharmacokinetic parameters obtained from postpartum women had returned to levels of nonpregnant women prior to pregnancy. Observed intrinsic secretory CL ( $\text{C}{\text{L}}_{\text{int},\text{sec}}$), blood renal secretory CL ( $\text{C}{\text{L}}_{\text{sec},B}$) and blood renal total CL ( $\text{C}{\text{L}}_{\text{renal},B}$) of tenofovir, oseltamivir carboxylate, cefazolin, and ketorolac were calculated from their reported $\text{CL}/F$ (oral) or $\text{CL}$ (intravenous). Observed CL of acyclovir was estimated by digitizing the reported concentration-time profile using WebPlotDigitizer (https://apps.automeris.io/wpd/). Observed $\text{C}{\text{L}}_{\text{int},\text{sec}}\text{and C}{\text{L}}_{\text{sec},B}$ of amoxicillin and pravastatin were calculated from the reported $\text{C}{\text{L}}_{\text{renal},B}$. Details of how these parameters were estimated are provided below (Fig. 1, Step 1–3).

Fig. 1.

Workflow for predicting $\mathrm{C}{\mathrm{L}}_{\sec ,B}^{\mathrm{pregnant}}$ and $\mathrm{C}{\mathrm{L}}_{\mathrm{renal},B}^{\mathrm{pregnant}}$ of the verification drugs based on in vivo renal CL of the probe drugs during pregnancy. #: In vivo $\text{C}{\text{L}}_{\text{int},\text{PD}}$ of the drugs was estimated from previously reported in vitro $\text{C}{\text{L}}_{\text{int},\text{PD}}$, MPPGK, and kidney weight (eq. 7), MPPGK was 15 mg/g and kidney weight was 4.3 g/kg body weight (Mathialagan et al., 2017). $ - foldchange in OAT activity during pregnancy was calculated using eq. 8. $\text{f}{\text{t}}_{\text{OAT j}}$ is the fractional contribution of each OAT of the drug in nonpregnant women where $\text{OA}{\text{T}}_j$ indicates OAT1, OAT2, or OAT3. RAF for OAT1, OAT2, and OAT3 was 0.64, 7.32 and 4.09, respectively (Mathialagan et al., 2017).

First, $\text{C}{\text{L}}_{\text{renal},B}$ in both nonpregnant and pregnant women were estimated (if not already stated in the publications) using eqs. 1–4 (Choi et al., 2019):

where $\text{C}{\text{L}}_{\text{total},B}$ was the sum of the $\text{C}{\text{L}}_{\text{renal},B}$ and nonrenal clearance ( $\text{C}{\text{L}}_{\text{non}-\text{renal},B}$). $\text{AU}{\text{C}}_{\text{inf}}$ was the area under the plasma concentration-time profile from zero to infinity. Oral bioavailability (F) and fraction excreted in the urine as unchanged drug ( $f_e$) in nonpregnant women are provided in Table 1. F, $f_e$, and $\text{C}{\text{L}}_{\text{non}-\text{renal}}$ of tenofovir, acyclovir, oseltamivir carboxylate, cefazolin, and ketorolac were assumed to be unchanged by pregnancy. Blood to plasma ratio (BP) in nonpregnant or pregnant women was estimated using previous published methods (Uchimura et al., 2010; Zhang et al., 2017). Clearances were normalized by the nonpregnant women’s body weight (i.e., postpartum weight for those women who were also studied antenatal) to take into account the effect of body weight on CL.

TABLE 1.

Summary of PK data of training and verification drugs

	$f_e$ a	Fa	fua	BPa	$$\text{C}{\text{L}}_{\text{int},\text{sec}}$$				$$\text{C}{\text{L}}_{\text{int},\text{PD}}$$		Reference
					Nonpregnant	1st Trimesterb	2nd Trimester	3rd Trimester	In Vitro	In Vivoc
					ml/min/kg				ul/min/mg protein	ml/min/kg
Tenofovir	0.81	0.25	0.99	0.58	0.49, 0.57	NA	1.39	0.43,0.65	0.10	0.01	(Best et al., 2015) (Colbers et al., 2013)
Acyclovir	0.76	0.22	0.85	1.00	2.80	NA	NA	8.66	0.27	0.02	De Miranda and Blum, 1983) (Kimberlin et al., 1998)
Oseltamivir carboxylate	0.93	0.79	0.97	0.67	3.43	7.45	5.70	4.39	0.10	0.01	(Pillai et al., 2015)
Amoxicillin	0.58	0.93	0.85	0.67	3.68	NA	7.29	6.85	4.60	0.30	(Andrew et al., 2007)
Cefazolin	0.80	1.00	0.18	0.55	4.40	NA	5.63	NA	1.70	0.11	(Philipson et al., 1987)
Pravastatin	0.45	0.18	0.47	0.56	13.51	NA	19.82	22.94	0.16	0.01	(Costantine et al., 2016)
Ketorolac	0.60	1.00	0.01	1.00	46.94	NA	NA	64.30	15.80	1.02	(Kulo et al., 2012)
S-ketorolac	0.60	1.00	0.01	1.00	62.51	NA	NA	103.68	15.80	1.02	(Kulo et al., 2017)
R-ketorolac	0.60	1.00	0.01	1.00	38.30	NA	NA	49.63	15.80	1.02	(Kulo et al., 2017)

^aF, $f_e$, fu and BP of the drugs are from nonpregnant women, except acyclovir, where the data are from healthy individuals (men and women).

^b1^st trimester: gestational age 1–12 weeks, 2^nd trimester: gestational age 13–28 weeks, 3^rd trimester: gestational age 29–40 weeks.

^cIn vivo $\text{C}{\text{L}}_{\text{int},\text{PD}}$ were estimated from in vitro $\text{C}{\text{L}}_{\text{int},\text{PD}}$ (Mathialagan et al., 2017).

NA, not available.

Then, the observed $\text{C}{\text{L}}_{\text{sec},B}$ and $\text{C}{\text{L}}_{\text{int},\text{sec}}$ in nonpregnant and pregnant women was calculated from $\text{C}{\text{L}}_{\text{renal},B}$ using the well stirred model (eqs. 5 and 6) (Ladumor et al., 2019):

where fraction unbound in blood ( $\text{f}{\text{u}}_B$) was the ratio of fraction unbound in plasma (fu) and blood to plasma ratio in nonpregnant or pregnant women estimated using previous published methods (Dallmann et al., 2017b; Zhang et al., 2017). Renal blood flow ( $Q_{\text{renal}}$) were determined from gestational age, dependent renal plasma flow, and hematocrit (Hct, %) (Odutayo and Hladunewich, 2012; Dallmann et al., 2017a). Glomerular filtration rate ( $\text{GF}{\text{R}}_{\text{inulin}}$) was based on inulin blood clearance in different trimesters (Koetje et al., 2011; Odutayo and Hladunewich, 2012). As justified by others, ${\text{F}}_{\text{reabs}}$ of the drugs was assumed to be negligible (Mathialagan et al., 2017).

In addition, gestational stage in our study was defined per US Department of Health and Human Services recommendations: 1–12 weeks as the 1^st trimester, 13–28 weeks as the 2^nd trimester, and 29–40 weeks as the 3^rd trimester. Except for oseltamivir carboxylate, drug CL data for the remaining probe drugs were not available for the 1^st trimester. Likewise, CL data for acyclovir and ketorolac in the 2^nd trimester and cefazolin in the 3^rd trimester were not available.

Determination of Modulation of OAT Activity during Pregnancy (Steps 4 and 5, Fig. 1).

As reported in our previous study (Kumar et al., 2021), the in vivo passive diffusion clearance ( $\text{C}{\text{L}}_{\text{int},\text{PD}})$ of the drugs was estimated (eq. 7) from previously reported in vitro $\text{C}{\text{L}}_{\text{int},\text{PD},in\mathrm{vitro}}$ in transfected HEK293 cells (Mathialagan et al., 2017)

where MPPGK (i.e. milligram of total proteins per gram of kidney) is 15 mg/g, and kidney weight is 4.3 g/kg body weight (Mathialagan et al., 2017).

Then, the fold-change in OAT activity during pregnancy was calculated by eq. 8.

where $\text{OA}{\text{T}}_j$ indicates OAT1, OAT2, or OAT3 (training dataset; OAT1: tenofovir, OAT2: acyclovir, OAT3: oseltamivir carboxylate; these drugs were assumed to be solely secreted by the listed OAT). $\text{C}{\text{L}}_{\text{PD}}$ was assumed to be unchanged by pregnancy.

Determination of Fraction of the Verification Drug Transported by $\mathrm{OA}{T}_j$.

Fraction transported by $\text{OA}{\text{T}}_j$ ( $\text{f}{\text{t}}_{{\text{OAT}}_{\text{j}}}$) of the verification drugs (amoxicillin, pravastatin, cefazolin and ketorolac, R-ketorolac, S-ketorolac) was calculated by eq. 9.

where in vitro $\text{C}{\text{L}}_{\text{int},\text{OAT j}}$ and relative activity factor ( $\text{RA}{\text{F}}_{\text{OAT j}}$, 0.64, 7.3, and 4.1 for OAT1, OAT2, and OAT3, respectively) were obtained from a previous publication (Mathialagan et al., 2017). Then, $\text{f}{\text{t}}_{\text{OAT j}}$ was used to calculate the individual OAT-mediated $\text{C}{\text{L}}_{\text{int},\text{OA}{\text{T}}_j}^{\text{non}-\text{pregnant}}$ of verification drugs in nonpregnant individuals by eq. 10.

Prediction of $C{L}_{\sec }$, $C{L}_{\mathrm{renal}}$, and $C{L}_{\mathrm{total}}$ of the Verification Drugs in Pregnancy.

Based on the above data, $\text{C}{\text{L}}_{\text{int},\text{OA}{\text{T}}_j}^{\text{pregnant}}$ of the verification drugs was calculated by multiplying fold-change in activity of each OAT at different gestational ages with $\text{C}{\text{L}}_{\text{int},\text{OA}{\text{T}}_j}^{\text{non}-\text{pregnant}}$ in nonpregnant individuals (eq. 11). $\text{C}{\text{L}}_{\text{int},\text{sec}}^{\text{pregnant}}$ was calculated from the total of individual $\text{C}{\text{L}}_{\text{int},\text{OA}{\text{T}}_j}^{\text{pregnant}}$ and the predicted in vivo $\text{C}{\text{L}}_{\text{int},\text{PD}}$ (eq. 12). Further, $\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$ was estimated using well stirred model based on $Q_{\text{renal}}$ and $\text{f}{\text{u}}_B$ during pregnancy (eq. 13). Finally, $\text{C}{\text{L}}_{\text{renal},B}^{\text{pregnant}}$ was determined from $\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$, $\text{f}{\text{u}}_B$, and GFR_inulin during pregnancy (eq. 14). $F_{\text{reabs}}$ was assumed to be negligible (Mathialagan et al., 2017). Total clearance (based on blood concentrations) in pregnancy ( $\text{C}{\text{L}}_{\text{total},B}^{\text{pregnant}}$) was predicted using eq. 15, where nonrenal CL was assumed to be unaffected by pregnancy and calculated using eq. 16.

Prediction of Change in $C{L}_{\sec }$ and $C{L}_{\mathrm{renal}}$ or $C{L}_{\mathrm{total}}$ of the Training and Verification Drugs in Pregnancy.

The ratio of the predicted $\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$, $\text{C}{\text{L}}_{\text{renal},B}^{\text{pregnant}}$, or $\text{C}{\text{L}}_{\text{total},B}^{\text{pregnant}}$ and the corresponding clearance in nonpregnant women (i.e., fold-change in different trimesters) was estimated for the verification and training drugs. The latter was estimated as indicated above for the verification drugs (eqs. 11–16). That is, the small contribution of other OAT contributing to their secretion was taken into consideration.

Data Analysis.

To assess the accuracy of our predictions, the ratio of the predicted and the observed data and absolute average fold error (AAFE) in the predictions of verification drugs were calculated as follows:

Our a priori acceptable range for the ratio was 0.8–1.25, i.e., the bioequivalence criteria.

Results

Pharmacokinetic Data Collection.

Where literature data were available, we estimated $\text{C}{\text{L}}_{\text{int},\text{sec}}$ of tenofovir, acyclovir, oseltamivir carboxylate, amoxicillin, cefazolin, pravastatin and ketorolac, R-ketorolac, S-ketorolac in nonpregnant and pregnant women at various gestational ages (Table 1).

Fraction Transported by $\mathrm{OA}{T}_j$ in the Nonpregnant Population.

As expected, the training drugs tenofovir, acyclovir, and oseltamivir carboxylate were found to be selectively transported by OAT1, OAT2, and OAT3: $\text{f}{\text{t}}_{\text{OAT}1}$ of tenofovir, $\text{f}{\text{t}}_{\text{OAT}2}$ of acyclovir, and $\text{f}{\text{t}}_{\text{OAT}3}$ of oseltamivir carboxylate were 94%, 99%, and 100%, respectively (Table 2, Supplemental Table 1). Regarding the verification dataset drugs, pravastatin was transported by OAT3, with a $\text{f}{\text{t}}_{\text{OAT}3}$ of 100%. Cefazolin was transported by OAT1 (3%) and OAT3 (97%), ketorolac and its isomers by OAT1 (72%) and OAT2 (29%), and amoxicillin by OAT1 (9%) and OAT3 (91%), respectively.

TABLE 2.

Estimates of the fraction of drug transported by each OAT, in vivo, in the nonpregnant population

$\text{Fraction of drug transported}in\mathrm{vivo}\text{by OATj}({\text{ft}}_{\text{OAT j}})=\frac{{\text{RAF}}_{\text{OATx}}{\text{*CL}}_{\text{int},\text{OATj}}}{{\sum }_3^1{\text{RAF}}_{\text{OATx}}{\text{*CL}}_{\text{int},\text{OATj}}}$ , where RAF is 0.64, 7.32, and 4.09 for OAT1, OAT2, and OAT3, respectively ( Mathialagan et al., 2017 ). The in vitro clearances ( $\text{C}{\text{L}}_{\text{int},\text{OATj}}$ ) were obtained from a previous study ( Mathialagan et al., 2017 ). $\text{f}{\text{t}}_{{\text{OAT}}_{\text{j}}}$ of R-ketorolac and S-ketorolac are assumed to be the same as the racemic mixture of ketorolac

Dataset	Drugs	$$\text{f}{\text{t}}_{\text{OAT}1}$$	$$\text{f}{\text{t}}_{\text{OAT}2}$$	$$\text{f}{\text{t}}_{\text{OAT}3}$$
		%	%	%
Training dataset	Tenofovir	94.00	0.00	6.00
	Acyclovir	0.00	99.00	1.00
	Oseltamivir carboxylate	0.00	0.00	100
Verification dataset	Pravastatin	0.00	0.00	100
	Cefazolin	3.00	0.00	97.00
	Amoxicillin	9.00	0.00	91.00
	Ketorolac (R-, S- and racemic mixture)	71.00	29.00	0.00

Fold-Change in OAT Activity during Pregnancy.

Compared with postpartum, OAT1 activity increased 2.9- and 1.0-fold in the 2^nd and 3^rd trimester, respectively, whereas OAT3 activity increased 2.2-, 1.7- and 1.3-fold in the 1^st, 2^nd, and 3^rd trimester, respectively. Compared with healthy individual, OAT2 activity increased 3.1 -fold in the 3^rd trimester (Fig. 2).

Fig. 2.

Fold-change in in vivo $\mathrm{OA}{\mathrm{T}}_{\mathrm{j}}$ activity at various stages of pregnancy. Fold-change in in vivo OAT activity due to pregnancy (pregnant/postpartum or pregnant/nonpregnant population) as measured by the change in vivo $\text{C}{\text{L}}_{\text{int},\text{OATj}}$ of tenofovir (OAT1), acyclovir (OAT2), and oseltamivir carboxylate (OAT3), respectively.

Predicted and Observed Renal and Secretory Clearance Values of the Verification Drugs.

The predicted/observed ratio of $\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$ of amoxicillin (2^nd trimester), pravastatin and ketorolac, R-ketorolac, and S-ketorolac fell within our a priori acceptance criteria (0.8–1.25 fold of the observed data) (Fig. 3, Table 3). The predicted/observed ratio of $\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$ of amoxicillin in the 3^rd trimester and cefazolin in the 2^nd trimester did not. They were 0.76 and 1.27 (but within 1.5-fold error), respectively. $\text{C}{\text{L}}_{\text{renal},B}^{\text{pregnant}}$ and $\text{C}{\text{L}}_{\text{total},B}^{\text{pregnant}}$ of verification dataset were all well predicted (within 0.80–1.25 fold, Fig. 3). The AAFEs of $\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$, $\text{C}{\text{L}}_{\text{renal},B}^{\text{pregnant}}$, and $\text{C}{\text{L}}_{\text{total},B}^{\text{pregnant}}$ of the verification drugs for the 2^nd trimester were 1.07, 1.02, and 1.04 respectively, and in the 3^rd trimester were 1.05, 1.02, and 1.05, respectively.

Fig. 3.

Predicted in vivo (A) $\mathrm{C}{\mathrm{L}}_{\sec ,B}^{\mathrm{pregnant}}$, (B) $\mathrm{C}{\mathrm{L}}_{\mathrm{renal},B}^{\mathrm{pregnant}}$ or (C) $\mathrm{C}{\mathrm{L}}_{\mathrm{total},B}^{\mathrm{pregnant}}$ of the verification drugs show good to excellent agreement with the observed data. The predicted $\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$ of amoxicillin (2^nd trimester), pravastatin, and ketorolac, R-ketorolac, and S-ketorolac fell within our a priori acceptance criteria (0.8–1.25 fold of observed values; dotted blue lines). However, the predicted/observed ratio of $\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$ of amoxicillin (3^rd trimester) and cefazolin (2^nd trimester) did not. They were 0.76 and 1.27 of the observed values, respectively. In contrast, the predicted $\text{C}{\text{L}}_{\text{renal},B}^{\text{pregnant}}$ and $\text{C}{\text{L}}_{\text{total},B}^{\text{pregnant}}$ of the verification drug all fell within our a priori acceptance criteria (within 0.8–1.25 fold of observed values).

TABLE 3.

Predicted and observed values of $\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$, $\text{C}{\text{L}}_{\text{renal},B}^{\text{pregnant}}$, and their corresponding ratios for the verification drugs

Drug (Transporters)	Parameters	$$\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$$		$$\text{C}{\text{L}}_{\text{renal},B}^{\text{pregnant}}$$
		2nd Trimesterc	3rd Trimesterc	2nd Trimester	3rd Trimester	Reference
		ml/min/kg
Ketorolac (OAT1,2)	Observed	NA	0.59	NA	0.61	(Kulo et al., 2012)
	Predicted	NA	0.69	NA	0.71
	Ratioa	NA	1.17	NA	1.16
S-Ketorolac (OAT1,2)	Observed	NA	1.14	NA	1.17	(Kulo et al., 2017)
	Predicted	NA	1.11	NA	1.14
	Ratio	NA	0.97	NA	0.97
R-Ketorolac (OAT1,2)	Observed	NA	0.35	NA	0.36	(Kulo et al., 2017)
	Predicted	NA	0.43	NA	0.44
	Ratio	NA	1.23	NA	1.22
Pravastatin (OAT3)	Observed	8.65	8.49	10.54	10.42	(Costantine et al., 2016)
	Predicted	9.19	7.48	11.08	9.40
	Ratio	1.06	0.88	1.05	0.90
Amoxicillin (OAT1,3)	Observed	5.96	5.37	8.63	8.05	(Andrew et al., 2007)
	Predicted	5.41	4.07	8.09	6.75
	Ratio	0.91	0.76	0.94	0.84
Cefazolin (OAT1,3)	Observed	1.85	NA	2.64	NA	(Philipson et al., 1987)
	Predicted	2.35	NA	3.14	NA
	Ratio	1.27	NA	1.19	NA
	AAFEb	1.07	1.02	1.05	1.02

^a $\mathrm{Ratio}=\frac{\mathrm{Predicted}}{\mathrm{Observed}}$

^bAbsolute average fold error ( $\text{AAFE})={10}^{|\frac{1}{\text{N}}\sum \text{log}\frac{\mathrm{predicted}}{\mathrm{observed}}|}$.

^c2^nd trimester: gestational age 13–28 weeks, 3^rd trimester: gestational age 29–40 weeks

NA, not available.

Predicted Change in $C{L}_{\sec }$ and $C{L}_{\mathrm{renal}}$ or $C{L}_{\mathrm{total}}$ of the Training and Verification Drugs in Pregnancy.

As expected, the $\text{total CL}(\text{C}{\text{L}}_{\text{total},B}^{\text{pregnant}})$ of an OAT-secreted drug is most affected by OAT induction in pregnancy when $f_e$ of the drug is large, and when $\text{C}{\text{L}}_{\text{sec}}$ is a large fraction of the total $\text{C}{\text{L}}_{\text{renal}}$ of the drug (e.g., oseltamivir carboxylate, Fig. 4, Supplemental Table 2).

Fig. 4.

Predicted fold-change in $\mathrm{C}{\mathrm{L}}_{\sec ,B}^{\mathrm{pregnant}}$, $\mathrm{C}{\mathrm{L}}_{\mathrm{renal},B}^{\mathrm{pregnant}}$, and $\mathrm{C}{\mathrm{L}}_{\mathrm{total},B}^{\mathrm{pregnant}}$ of the training and verification drugs when compared with that in the nonpregnant population. The $\text{C}{\text{L}}_{\text{total}}$ of a drug was most affected by induction of OAT when $f_e$ and $\text{C}{\text{L}}_{\text{sec}}$/ $\text{C}{\text{L}}_{\text{renal}}$ of the drug are large (e.g., oseltamivir carboxylate).

Discussion

Here, we predicted the $\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$, $\text{C}{\text{L}}_{\text{renal},B}^{\text{pregnant}}$, and $\text{C}{\text{L}}_{\text{total},B}^{\text{pregnant}}$ of OAT drugs transported by single or multiple OATs. In doing so, we employed an experimental design not adopted by others when predicting renal CL of drugs (van Hasselt et al., 2014; Wang et al., 2019). First, this is the first time that probe drug data, in combination with the RAF approach, have been used to predict $\text{C}{\text{L}}_{\text{sec}}$ and $\text{C}{\text{L}}_{\text{renal}}$ of drugs during pregnancy. Others have reported using physiologically-based PK models to predict renal CL of OAT drugs during pregnancy. However, they have either ignored pregnancy-induced changes in renal secretion of these drugs or have concluded that such changes are absent (De Sousa Mendes et al., 2016; Dallmann et al., 2017b); second, we focused on predicting $\text{C}{\text{L}}_{\text{sec}}$ rather than $\text{C}{\text{L}}_{\text{renal}}$ or $\text{C}{\text{L}}_{\text{total}}$, as the latter two can be predicted well even when $\text{C}{\text{L}}_{\text{sec}}$ is poorly predicted, such as for drugs where $\text{C}{\text{L}}_{\text{sec}}$ is a small or a minor contributor to $\text{C}{\text{L}}_{\text{renal}}$; third, most publications use creatinine renal CL as a measure of glomerular filtration CL (Banfi et al., 2009; Garner et al.,2019; Wiles et al., 2020). However, creatinine is also secreted via OCT2 and OAT2 (Gutiérrez et al., 2014; Lepist et al., 2014; Chu et al., 2016), which could confound interpretation of changes in OAT/OCT activity during pregnancy. Therefore, we used inulin renal CL, a gold-standard measure of glomerular filtration CL, because it is not actively secreted (; Smith et al., 2008; Koetje et al., 2011).

As expected, the fraction of probe drugs transported by the OATs (training dataset) was more than 90% (Table 2). Further, there is no evidence of transporter-mediated reabsorption of tenofovir, acyclovir, or oseltamivir carboxylate, and the $\text{C}{\text{L}}_{\text{PD}}$ of these drugs is relatively small (Table 1). Therefore, tenofovir, acyclovir, and oseltamivir carboxylate can be used with confidence as selective in vivo probes of pregnancy-induced changes in OAT1, OAT2, and OAT3 activity, respectively.

We found that each OAT activity was induced by pregnancy to a different extent, with maximum induction of about 3-fold for OAT1 and OAT2 and about 2-fold for OAT3 (Fig. 2). The time course of this induction also varied, with OAT3 activity peaking earlier (1^st trimester) than OAT1 (2^nd trimester), whereas OAT2 activity peaked in the 3^rd trimester. However, additional data at various gestational ages are needed to confirm this conclusion. This pattern of change in OAT activity during pregnancy may be driven by the pattern of changes in plasma concentration of pregnancy hormones as we have shown previously for induction of hepatic CYP3A4 (Zhang et al., 2015). For example, the plasma concentration of estradiol, progesterone, human chorionic gonadotropin, and relaxin peaks during 1^st, 2^nd, and 3^rd trimester of human pregnancy, respectively (Steroid Endocrinology of Pregnancy, 2009; Cheung and Lafayette, 2013; Papageorgiou et al., 2013). Indeed, experiments with opossum kidney cells show that 17β-estradiol induces human renal OAT1 likely via the estrogen receptor α (Euteneuer et al., 2019). Expression of renal Oat2 mRNA level (slc22a7) increases 1.19- to 1.88-fold during pregnancy in mice (Yacovino et al., 2013), and mRNA expression of rat Oat2 is modestly upregulated by estradiol and progesterone (1.25- and 1.18-fold) but significantly downregulated by testosterone (0.27-fold) (Ljubojević et al., 2007). Last, the activity of OAT3 increased during pregnancy, and it was the highest in the 1^st trimester. This trend was similar to the change in plasma concentration of human chorionic gonadotropin and relaxin (Cheung and Lafayette, 2013). These data of pregnancy-induced OAT induction are hypothesis-driving observations and should be followed up by experimental studies to determine the mechanism(s) of regulation of OAT transporters. An alternative explanation is that this OAT induction is a nonspecific change in the size and volume of kidneys and/or the length of proximal tubule during pregnancy (Cheung and Lafayette, 2013). However, such a change would likely result in the same extent and time-frame of induction of the OATs. Our data suggest that this is not the case (Fig. 2).

Based on the changes in OAT activity during pregnancy, we successfully predicted the pregnancy-induced changes in $\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$, $\text{C}{\text{L}}_{\text{renal},B}^{\text{pregnant}}$, and $\text{C}{\text{L}}_{\text{total},B}^{\text{pregnant}}$ of the verification drugs [amoxicillin, cefazolin, pravastatin, and ketorolac (R-ketorolac, S-ketorolac, and the racemic mixture)] (Table 3, Fig. 3). These predictions were within our a priori acceptance criteria (i.e., 0.80–1.25 fold of the observed value), except for $\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$ of amoxicillin in the 3^rd trimester and cefazolin in the 2^nd trimester, which were marginally outside (0.76 and 1.27, respectively) our acceptance criteria (Fig. 3). This highlights the importance of predicting the $\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$ rather than $\text{C}{\text{L}}_{\text{renal},B}^{\text{pregnant}}$ or $\text{C}{\text{L}}_{\text{total},B}^{\text{pregnant}}$ of drugs.

There are a few limitations to our study. First, we assumed that the rate-determining step in renal secretory CL of the drugs was their OAT-mediated transport. That is, even though some of the probe and verification drugs are substrates of other transporters (basal and apical; see Supplemental Table 1), we assumed that their secretory CL was determined only by OATs. Given the low CL_PD of the drugs, and assuming this value applies to the basal efflux CL of the drugs based on the extended CL model, this is a reasonable assumption (Patilea-Vrana and Unadkat, 2016). If this assumption is incorrect, then the proposed approach will work only if probe drugs are available that selectively report the activity of each of these alternative transporters. Second, we assumed that $\text{C}{\text{L}}_{\text{non}-\text{renal}}$ of tenofovir, acyclovir, oseltamivir carboxylate, cefazolin, and ketorolac is not affected by pregnancy. About 20% of ketorolac is metabolized by UGT enzyme and 9%–14% of acyclovir is metabolized by alcohol and aldehyde dehydrogenases. However, these enzymes may not be affected by human pregnancy (Mroszczak et al., 1987; De Miranda and Good, 1992; Anderson, 2005; Jelski et al., 2020). Third, the pharmacokinetics of the drugs in postpartum women were assumed to have returned to those in nonpregnant women prior to pregnancy. Indeed, comparison of their postpartum pharmacokinetics with those in nonpregnant individuals supported this conclusion (data not shown). Fourth, $F_{\text{reabs}}$ was assumed to be negligible as previously justified (Mathialagan et al., 2017). If this assumption is substantially incorrect, our predictions for the verification drugs would not have met our acceptance criterion. Last, but not least, in vivo data for OAT1 in the 1^st trimester or for OAT2 in the 1^st and 2^nd trimester are, as yet, not available, limiting the application of our approach to OAT1 and OAT2 substrates in these trimesters.

Based on the above observations, an important question that arises is: under what circumstances will pregnancy-induced changes in OAT-mediated drug CL result in recommendation of change in dosing regimen of a drug administered to pregnant women? Of course, the drug would have to be predominately renally cleared from the body, and $\text{C}{\text{L}}_{\text{sec}}$ would need to constitute a large fraction of $\text{C}{\text{L}}_{\text{renal}}$ of the drug (high $\text{C}{\text{L}}_{\text{sec}}/\text{C}{\text{L}}_{\text{renal}}$ and therefore low $\text{f}{\text{u}}_B\text{*GF}{\text{R}}_{\text{inulin}}$/ $\text{C}{\text{L}}_{\text{renal},B}$). A good example of these guidelines is acyclovir (Fig. 4, Supplemental Table 2). Acyclovir’s $f_e$ is 0.76 and $\text{C}{\text{L}}_{\text{sec}}/\text{C}{\text{L}}_{\text{renal}}$ is 0.57, and it is secreted predominately by OAT2 (97%), which is induced by 3.1-fold during the 3^rd trimester of pregnancy. Based on these data, we predict that its $\text{C}{\text{L}}_{\text{total}}$ will increase (compared with nonpregnant population) by 1.69-fold during the 3^rd trimester. However, if acyclovir’s $f_e$ and $\text{C}{\text{L}}_{\text{sec}}/\text{C}{\text{L}}_{\text{renal}}$ were much greater (e.g., 0.9), the change in the $\text{C}{\text{L}}_{\text{total}}$ of this hypothetical drug would be 2.4-fold. Such a change would warrant dosing regimen change for only those drugs that have a narrow therapeutic window. In contrast, a small $\text{C}{\text{L}}_{\text{sec}}/\text{C}{\text{L}}_{\text{renal}}$ or $f_e$ will reduce the impact of OAT induction on $\text{C}{\text{L}}_{\text{total}}$ of the drug in pregnant women (Fig. 4). For example, the predicted $\text{C}{\text{L}}_{\text{total}}$ of tenofovir in the 2^nd trimester was less affected by induction of OAT1 activity (2.9-fold), even though its $f_e$ is 0.81, because its $\text{C}{\text{L}}_{\text{sec}}/\text{C}{\text{L}}_{\text{renal}}$ is small (0.23). This is not surprising since contribution by filtration clearance modulates the change in $\text{C}{\text{L}}_{\text{renal}}$ produced by large and significant changes in $\text{C}{\text{L}}_{\text{sec}}$ of the drug. For example, the maximum reduction in in vivo $\text{C}{\text{L}}_{\text{total}}$ of an OAT substrate (bumetanide) in the presence of probenecid, a potent OAT inhibitor, is ∼85% (Mathialagan et al., 2017).

In summary, we showed for the first time that the changes in $\text{C}{\text{L}}_{\text{sec},B}^{\text{pregnant}}$ and $\text{C}{\text{L}}_{\text{renal},B}^{\text{pregnant}}$ of OAT-transported drugs can be predicted during pregnancy using probe drugs and the RAF approach. This approach could be used in the future to prospectively adjust dosing regimens of renally secreted OAT drugs (likely narrow therapeutic window drugs) administered to pregnant women without a need to conduct pharmacokinetic studies in this difficult to study and understudied population.

Abbreviations

AAFE

absolute average fold error

CL

clearance

CL_{int,OAT j}

individual OAT_j mediated clearance

CL_int,PD

intrinsic passive diffusion clearance

CL_int,sec

intrinsic renal secretory clearance

CL_non-renal,B

blood nonrenal clearance

CL_PD

passive diffusion clearance

CL^pregnant_renal

plasma renal clearance in pregnancy

CL^pregnant_sec

plasma renal secretory clearance in pregnancy

CL^pregnant_total,B

plasma total clearance in pregnancy

CL^pregnant_renal,B

blood renal clearance in pregnancy

CL^pregnant_sec,B

blood renal secretory clearance in pregnancy

CL^pregnant_int,sec

intrinsic renal secretory clearance in pregnancy

CL^pregnant_{int,OAT j}

individual OAT-mediated intrinsic clearance in pregnancy

CL_renal

plasma renal clearance

CL_renal,B

blood renal clearance

CL_sec

plasma renal secretory clearance

CL_sec,B

blood renal secretory clearance

CL_total

plasma total clearance

CL_total,B

blood total clearance

F

oral bioavailability

f_e

fraction of drug excreted unchanged in the urine

F_reabs

fraction of drug reabsorbed

ft_{OAT j}

fraction of drug transported by OAT_j

fu

fraction unbound of drug in plasma

MPPGK

milligram of total proteins per gram of kidney

OAT

organic anion transporter

PK

pharmacokinetics

RAF

relative activity factor

Authorship Contributions

Participated in research design: Peng, Ladumor, Unadkat.

Conducted experiments: Peng, Ladumor.

Performed data analysis: Peng, Ladumor, Unadkat.

Wrote or contributed to the writing of the manuscript: Peng, Ladumor, Unadkat.