Advertisement
Research ArticleArticle

In Vitro–to–In Vivo Extrapolation of Transporter-Mediated Renal Clearance: Relative Expression Factor Versus Relative Activity Factor Approach

Aditya R. Kumar, Bhagwat Prasad, Deepak Kumar Bhatt, Sumathy Mathialagan, Manthena V. S. Varma and Jashvant D. Unadkat
Drug Metabolism and Disposition June 2021, 49 (6) 470-478; DOI: https://doi.org/10.1124/dmd.121.000367

This article has a correction. Please see:

Abstract

About 30% of approved drugs are cleared predominantly by renal clearance (CLr). Of these, many are secreted by transporters. For these drugs, in vitro–to–in vivo extrapolation of transporter-mediated renal secretory clearance (CLsec,plasma) is important to prospectively predict their renal clearance and to assess the impact of drug-drug interactions and pharmacogenetics on their pharmacokinetics. Here we compared the ability of the relative expression factor (REF) and the relative activity factor (RAF) approaches to quantitatively predict the in vivo CLsec,plasma of 26 organic anion transporter (OAT) substrates assuming that OAT-mediated uptake is the rate-determining step in the CLsec,plasma of the drugs. The REF approach requires protein quantification of each transporter in the tissue (e.g., kidney) and transporter-expressing cells, whereas the RAF approach requires the use of a transporter-selective probe substrate (both in vitro and in vivo) for each transporter of interest. For the REF approach, 50% and 69% of the CLsec,plasma predictions were within 2- and 3-fold of the observed values, respectively; the corresponding values for the RAF approach were 65% and 81%. We found no significant difference between the two approaches in their predictive capability (as measured by accuracy and bias) of the CLsec,plasma or CLr of OAT drugs. We recommend that the REF and RAF approaches can be used interchangeably to predict OAT-mediated CLsec,plasma. Further research is warranted to evaluate the ability of the REF or RAF approach to predict CLsec,plasma of drugs when uptake is not the rate-determining step.

Significance Statement This is the first direct comparison of the relative expression factor (REF) and relative activity factor (RAF) approaches to predict transporter-mediated renal clearance (CLr). The RAF, but not REF, approach requires transporter-selective probes and that the basolateral uptake is the rate-determining step in the CLr of drugs. Given that there is no difference in predictive capability of the REF and RAF approach for organic anion transporter–mediated CLr, the REF approach should be explored further to assess its ability to predict CLr when basolateral uptake is not the sole rate-determining step.

Introduction

Accurate prediction of in vivo clearance (CL) is important to support drug candidate selection during early-stage development and to evaluate the impact of drug interactions and pharmacogenetics in clinical development. A comprehensive analysis of 391 drugs found that 31% of compounds were predominantly cleared by renal clearance (CLr) (i.e., CLr > 50% of total clearance) (Varma et al., 2009). Renal clearance is mediated by active secretion, filtration, and tubular reabsorption. Active secretion of drugs includes passive and transporter-mediated uptake and efflux CL, respectively, across the basal and apical membrane of the proximal renal tubule cells. Organic anion transporters (OATs 1–3), located on the basal membrane, are important contributors to the renal secretion of many renally cleared drugs including drugs such as antibiotics and antivirals (Feng et al., 2010). Filtration clearance is a passive process that depends on glomerular filtration rate and fraction of the drug unbound in the plasma (fu). Although tubular reabsorption (active or passive or both) can occur, it cannot be determined in vivo and is therefore assumed to be passive and minimal.

Common predictive preclinical methodologies used to estimate metabolic CL in humans are in vitro–to–in vivo extrapolation (IVIVE) using primary cells (e.g., hepatocytes) and physiologic or relative activity factor (RAF) scaling. Although IVIVE using RAF or physiologic scaling factors (PSFs) has been shown to be relatively successful in predicting metabolic clearance, such predictions for transporter-mediated clearance, including active secretion clearance, need to be verified (Rostami-Hodjegan and Tucker, 2007; Soars et al., 2007; Rowland et al., 2011; Ke et al., 2014). Moreover, unlike human hepatocytes, validated primary human kidney epithelial cells for transport studies are not routinely available. Although human CLr predictions can be conducted by allometric scaling of in vivo renal CL data in animals, due to interspecies differences in transporter abundance and activity, allometry can lead to inaccurate prediction of human CLr (Paine et al., 2011; Chu et al., 2013). Other methods for IVIVE of human CLr that have been used are human kidney slices (Watanabe et al., 2011; Scotcher et al., 2016a). However, kidney slices underestimated OAT3-mediated intrinsic renal secretory clearance of 7 OAT3-transported drugs, and IVIVE of their renal secretory CL required a scaling factor of 10.

Recently, the RAF approach was successfully used by Mathialagan et al. (2017) to predict the in vivo human OAT-mediated renal secretory CL and total CLr of 31 drugs. Using cells expressing the transporter(s) of interest (e.g., OAT1-expressing cells), the RAF approach scales the in vitro transporter uptake CL of the drug of interest to its in vivo clearance. To do so, the RAF approach requires that the in vitro uptake CL of a probe drug be available in the transporter-selective (e.g., OAT1) cells as well as in vivo (Fig. 1). However, a shortcoming of the RAF approach is that such transporter-selective drugs are often not available for many transporters (e.g., breast cancer resistance protein, organic anion transporting polypeptides). An alternative approach, the relative expression factor (REF) approach, has recently begun to be explored for IVIVE of drug CL (Ishida et al., 2018; Kumar et al., 2018; Sachar et al., 2020). Unlike the RAF approach, the REF approach does not require a transporter-selective probe substrate. Instead, it requires information on the in vivo and in vitro abundance of the transporter in the tissue of interest (e.g., kidneys) and in the cells used to determine the drug’s in vitro transport CL (Fig. 1) (Kumar et al., 2018). Quantitative targeted proteomics, due to its selectivity, sensitivity, and lack of need for protein standards, has become the preferred method for quantification of abundance of transporters in both in vitro systems and tissue samples (Prasad et al., 2016). Then, this abundance is used to scale the in vitro transport CL in cells expressing the transporter of interest to that in vivo assuming that the maximal transporter-mediated fluxof each transporter is directly proportional to the abundance of the transporter and the Km of the drug for the transporter in vivo is identical to that in vitro. Another advantage of the REF over the RAF approach is that it can handle the involvement of multiple rate-determining transport steps in the CL of the drug, irrespective of whether the transporters are located at the basal, apical, or both membranes of the kidney epithelial cells (Patilea-Vrana and Unadkat, 2016). In this event, the RAF method would require multiple probe substrates, each reporting the individual rate-determining step, a scenario that is nearly impossible to achieve.

Fig. 1.
Fig. 1.

For IVIVE of renal CLsec,plasma of drugs, the REF approach scales the drug uptake CL into transporter-expressing cells using the REF (transporter abundance in transporter-expressing cells/transporter abundance in the human kidney). In contrast, the RAF approach scales the drug uptake CL into transporter-expressing cells using the RAF (uptake CL of the probe drug in transporter-expressing cells/in vivo probe CLint,sec,in vivo). Eq. X indicates the equation used in the REF approach or the RAF approach (Mathialagan et al., 2017). CLFiltration, filtration clearance; hOAT, human organic anion transporter.

Although both IVIVE scaling approaches (REF and RAF) have been successfully used to predict hepatic uptake clearance of drugs mediated by organic anion transporting polypeptides (Kunze et al., 2014a; Ishida et al., 2018), a direct comparison of the REF and RAF approach for IVIVE of renal secretory CL has never been reported. Therefore, the primary aim of this study was to compare the ability of the REF and RAF approaches to successfully predict the in vitro intrinsic renal secretory CL of several OAT-transported drugs. Our secondary aims were to test the ability of both these approaches to predict the total renal secretoryCL (active and passive) and the total renal CL of drugs. For these comparisons, the data for the RAF IVIVE of renal CL (including intrinsic, total secretory CL, and total renal CL) of these OAT drugs were obtained from a previous publication (Mathialagan et al., 2017).

Materials and Methods

The rationale for choosing the 26 OAT-transported drugs has been provided by Mathialagan et al. (2017). Briefly, these drugs were chosen because they are selectively transported by OATs. Detailed materials and methods used to conduct IVIVE of renal secretory CL of drugs using the RAF approach have been previously published (Mathialagan et al., 2017). Of note, tenofovir, acyclovir/ganciclovir, and oseltamivir/benzylpenicillin were used as probe substrates for OAT1, OAT2, and OAT3, respectively. Therefore, here we describe the materials and methods used for IVIVE of renal CL of drugs using only the REF approach.

Materials

Bovine serum albumin, ammonium bicarbonate (98% purity), iodoacetamide, dithiothreitol, and trypsin protease [mass spectrometry (MS ) grade] were obtained from Thermo Fisher Scientific (Rockford, IL). Stable isotope-labeled (heavy) peptides and synthetic unlabeled peptides were purchased from Thermo Fisher Scientific (Rockford, IL) and New England Peptides (Boston, MA), respectively. ProteoExtract native membrane protein extraction kit was purchased from Calbiochem (Temecula, CA). Optima MS-grade acetonitrile, methanol, chloroform, formic acid, and bicinchoninic acid (BCA) protein assay kit were purchased from Fisher Scientific (Fair Lawn, NJ). Human embryonic kidney (HEK) 293 cells were obtained from Pfizer Inc.

Determination of OAT-Mediated Uptake CL of Drugs Using OAT-Overexpressing HEK Cells

Since the OAT-mediated active uptake and passive diffusion CL of the 26 OAT-transported drugs was used for both the REF and RAF approach, the reader is referred to the previous publication as to how these values were experimentally obtained (Mathialagan et al., 2017).

Transporter Quantification

The values of protein abundance of renal transporters in the human renal cortex were previously generated by our laboratory (Prasad et al., 2016). The same quantification protocol was used to quantify the transporters in the OAT-overexpressing HEK cells and is detailed here. Briefly, membrane protein extraction of 3 to 5 million HEK293 cells overexpressing OAT1, OAT2, or OAT3 and three adult kidney cortex samples (∼50–100 mg) was performed as follows. Membrane proteins (2 mg/ml) were denatured (heating), reduced (dithiothreitol), alkylated (iodoacetamide), and digested using trypsin as per optimized conditions described previously (Prasad et al., 2016). The unlabeled synthetic surrogate peptides for each transporter (light peptides) were used as the calibrators. The corresponding peptides, labeled with [13C615N2]-lysine and [13C615N4]-arginine residues, were used as the internal standards (heavy peptides). Each trypsin digested sample (5 µl) was injected onto the column (ACQUITY UPLC HSS T3 1.8 μm, C18 100A; 100 × 2.1 mm, Waters, Milford, MA). Peptide quantification was performed using a triple-quadrupole MS instrument (Sciex Triple Quad 6500, Concord, ON) in electron spray ionization positive ionization mode coupled to an Acquity UPLC, I-class (Waters, Milford, MA) (Supplemental Table 1). The parent to product ion transitions for the light and heavy peptides were monitored using optimized liquid chromatography–tandem mass spectrometry parameters in electron spray ionization positive ionization mode as described previously. The liquid chromatography–tandem mass spectrometry data were processed using Analyst 1.6.2 version software (Sciex, Concord, ON, Canada) as described previously (Li et al., 2019).

View this table:
TABLE 1

The predicted precision, bias, and percent of data within 2- or 3-fold of the observed value for CLint,sec,in vivo, CLsec,plasma, or CLr,plasma when using the REF or RAF approach

Passive diffusion secretory CL of the drug was assumed to be negligible.

Prediction of Renal Clearance

The REF value for each transporter was calculated using eq. 1. Embedded Image

where x is 1, 2, or 3. The renal cortex OATx abundance was that in the pooled sample of three kidneys that was assayed simultaneously with the OATx abundance in the HEK cells (Supplemental Table 2). When the in vivo intrinsic secretory clearance (CLint,sec,in vivo) of the drug was calculated by scaling only active uptake, eq. 2 was used. Embedded Image

View this table:
TABLE 2

The predicted precision, bias, and percent of data within 2- or 3-fold of the observed value for CLint,sec,in vivo, CLsec,plasma, or CLr,plasma when passive diffusion secretory CL was included in the RAF or REF approaches

where the active uptake CL represents the transporter-mediated uptake CL of the drug, the milligrams of protein per gram cortex (MPPGC) was 300 mg/g, and the grams of cortex per kilogram body weight (BW) was 3 g/kg (Bouchet et al., 2003; Kumar et al., 2018). However, when passive diffusion secretory was take into consideration in addition to the active uptake for CLint,sec,in vivo calculations, eq. 3 was used. Embedded Imagewhere passive diffusion CLin vitro was obtained from Mathialagan et al. (2017). Once the CLint,sec,in vivo predicted by the REF approach was calculated by either eq. 2 or eq. 3, it was used to predict the total renal secretory clearance (CLsec,plasma) of the drug using eq. 4. Embedded Imagewhere Qr (15.7 ml/min/kg) is the renal blood flow, fu,blood is the unbound fraction in blood, and (B/P) is the blood to plasma ratio. Then, the total renal plasma CL (CLr,plasma) was calculated using unbound fraction in plasma (fu,plasma), glomerular filtration rate (GFR; 1.78 ml/min/kg) (Varma et al., 2009), CLsec,plasma, and the fraction reabsorbed in kidney tubules (Freabs) (eq. 5). Because the passive reabsorption fraction (Freabs) of drugs cannot be determined in vivo, it was assumed to be zero (Mathialagan et al., 2017). Of note, the probe substrates are also multidrug resistance protein and/or multi-antimicrobial extrusion protein (MATE) substrates. However, in using eq. 5, we assumed, as did Mathialagan et al. (2017), that the transporter CL across the basal membrane (mediated by OATs) was the rate-determining step in the renal secretory CL of the drug. Embedded Image

Comparison of the Ability of REF and RAF to Predict CLint,sec,in vivo, CLsec,plasma, and CLr,plasma

Two approaches were used to compare the ability of REF and RAF to predict CLint,sec,in vivo, CLsec,plasma, and CLr,plasma. First, we determined the number of drugs for which the predicted values fell within 2-fold or 3-fold of the observed values. Second, to determine if the two approaches were significantly different from each other, we determined the precision [root mean squared error (RMSE); eq. 6] and bias [mean error (ME); eq. 7] of each approach, where n is the number of drugs tested. If the 95% confidence intervals of precision and bias of each approach overlapped, we concluded that the two approaches were not statistically different. The above statistics were computed with and without including passive diffusion secretory CL of the drugs. Also, of note, the probe drugs were not included when computing these statistics because they were used to derive the RAF values. Embedded Image Embedded Image

Results

REF and RAF Values

The REF values determined by quantifying the OAT1, 2, and 3 transporters in HEK293 and renal cortex were 0.16, 0.15, and 0.37, respectively (Supplemental Table 2) (Prasad et al., 2016). The RAF values for OAT1 (tenofovir), OAT2 (acyclovir and ganciclovir), and OAT3 (oseltamivir acid and benzylpenicillin) were previously reported as 0.64, 7.3, and 4.1, respectively (Mathialagan et al., 2017). As indicated in a previous publication, the chosen probe substrates (i.e., tenofovir, acyclovir/ganciclovir, and oseltamivir/benzylpenicillin) are selective for the specified transporter and have no significant uptake by the other OAT transporters located on the basal membrane of the kidney epithelial cells (Mathialagan et al., 2017).

Comparison of the REF and RAF Approaches to Predict Secretory and Total Renal Clearance of Drugs

For the REF approach, 46% and 62% of the CLint,sec,in vivo predictions were within 2- and 3-fold of the observed values, respectively, whereas the corresponding values for the RAF approach were 62% and 73%, respectively (Table 1; Fig. 2, A and B). For the REF approach, 50% and 69% of the CLsec,plasma predictions were within 2- and 3-fold of the observed values, respectively, whereas the corresponding values for the RAF approach were 65% and 81%, respectively (Table 1; Fig. 2, C and D). Finally, for the REF approach, 65% and 92% of the CLr,plasma predictions were within 2- and 3-fold of the observed values, whereas the corresponding values for the RAF approach were 81% and 88%, respectively (Table 1; Fig. 2 E and F). The 95% confidence intervals for the precision (RMSE) and bias (ME) of the REF-predicted CLint,sec,in vivo, CLsec,plasma, and CLr,plasma overlapped with those of the RAF approach (Table 1; Fig. 3). Precision and bias calculations were identical for the predicted CLsec,plasma and CLr,plasma, as addition of filtration clearance (the same constant value in both approaches) to CLsec,plasma did not alter the predictive power of the two approaches. The REF and RAF CLint,sec,in vivo, CLsec,plasma, and CLr,plasma prediction, observed value, and fold error for each drug are listed in Supplemental Table 3.

Fig. 2.
Fig. 2.

Observed and predicted values of CLint,sec,in vivo (A and B), CLsec,plasma (C and D), or CLr,plasma (E and F) when using the REF and RAF approaches. The solid line is the line of identity. Passive diffusion secretory CL of the drugs was assumed to be negligible.

Fig. 3.
Fig. 3.

Both the REF and RAF approaches were equally precise (RMSE) and unbiased (ME) in predicting the CLint,sec,in vivo (A and B) as demonstrated by the overlapping 95% confidence intervals (lines). This conclusion remained the same for precision and bias of CLsec,plasma and CLr,plasma predictions (C and D) by the two approaches. Passive diffusion secretory CL of the drugs was assumed to be negligible.

View this table:
TABLE 3

Total protein concentration in whole kidney and kidney cortex tissue

Theoretically, the REF approach should be highly sensitive to the value of the PSF used, whereas the RAF approach should not be, as it does not necessarily need to use a PSF. However, although Mathialagan et al. (2017) did use PSFs, the RAF values are independent of the PSF used, as this value cancels out when predicting CLint,sec,in vivo (see eq. 9 in Mathialagan et al., 2017). Nevertheless, their PSF values were 15 mg of protein per gram of kidney (0.25 mg of protein per million HEK cells, 60 million HEK cells per gram of kidney) and 4.3 g of kidney per kilogram of body weight (hereafter referred to as the kidney PSF) (Mathialagan et al., 2017). Therefore, we examined the sensitivity of the REF and RAF approach to the value of PSF used. The PSF for the REF approach that we used was the value that we have previously determined in kidney tissue where the aforementioned transporters were quantified, i.e., 210 mg of protein per gram of kidney (hereafter referred to as the cortex PSF). Since these approaches used different PSFs, we compared the predictive power of the two approaches using the same PSF. As expected (since the RAF approach is independent of the PSF used), when the cortex PSF was used, the predicted CLsec,plasma by the RAF differed from that by kidney PSF by only 1.6% (Supplemental Fig. 1). In contrast, the predicted CLsec,plasma by the REF using the kidney PSF considerably underpredicted the observed values by an average of about 10-fold (Supplemental Fig. 1).

In the above analyses, for both the REF and RAF approach, passive diffusion secretory CL of the drug was not taken into consideration. Therefore, we compared the predictive capability of the REF and RAF approaches with inclusion of passive diffusion secretory clearance. To be consistent across both approaches, the cortex PSF was used to scale the passive diffusion secretory CL. In doing so, for the REF approach, 23% and 50% of the CLint,sec,in vivo predictions were within 2- and 3-fold of the observed values, respectively, whereas the corresponding values for the RAF approach were 35% and 54%, respectively (Table 2; Fig. 4 A and B). For the REF approach, 31% and 65% of the CLsec,plasma predictions were within 2- and 3-fold of the observed values, respectively, whereas the corresponding values for the RAF approach were 46% and 58%, respectively (Table 2; Fig. 4 C and D). For the REF approach, 38% and 88% of the CLr,plasma predictions were within 2- and 3-fold of the observed values, respectively, whereas the corresponding values for the RAF approach were 65% and 85%, respectively (Table 2; Fig. 4 E and F). The 95% confidence intervals for the precision (RMSE) and bias (ME) of the REF-predicted CLint,sec,in vivo, CLsec,plasma, and CLr,plasma overlapped with those of the RAF approach. However, the REF approach demonstrated a positive bias for the CLint,sec,in vivo predictions (Table 2; Fig. 5), whereas the RAF approach did not. Nevertheless, even with the addition of passive diffusion secretory CL, the CLint,sec,in vivo, CLsec,plasma, and CLr,plasma predictions for both the REF and RAF approaches were equally as precise and unbiased as the predictions when passive diffusion secretory CL was not taken into consideration (Supplemental Fig. 2). The REF and RAF CLint,sec,in vivo, CLsec,plasma,and CLr,plasma prediction, observed value, and fold error after the inclusion of passive diffusion secretory CL for each drug are listed in Supplemental Table 4.

Fig. 4.
Fig. 4.

Observed and predicted values of CLint,sec,in vivo (A and B), CLsec,plasma (C and D), or CLr,plasma (E and F) when passive diffusion secretory CL was included in the REF and RAF approaches. The solid line is the line of identity.

Fig. 5.
Fig. 5.

After including passive diffusion secretory clearance when predicting CLint,sec,in vivo, the REF approach demonstrated a positive bias (ME), whereas the RAF approach did not, but both approaches were equally precise (RMSE) as demonstrated by the overlapping 95% confidence intervals [lines; (A) and (B)]. In addition, both approaches were equally precise (RMSE) and unbiased (ME) in predicting CLsec,plasma and CLr,plasma of the drugs (C and D).

Discussion

The goal of this work was to evaluate the capability of the REF approach to predict CLint,sec,in vivo, CLsec,plasma, and CLr,plasma from in vitro OAT uptake studies in OAT-overexpressing HEK cells and compare the predictability to the previously reported RAF approach (Mathialagan et al., 2017). Our primary endpoint was comparison of the predicted CLsec,plasma as opposed to CLr,plasma. The latter is often used as the prediction endpoint, but this is misleading because CLr,plasma can be predicted well even when CLsec,plasma is not. This will occur when CLsec,plasmais a relatively small fraction of the total renal CL of the drug (Kunze et al., 2014b).

Three main assumptions were made in our study. First, of the three parts that compose renal clearance, we accounted for filtration and tubular secretion, but reabsorption was assumed to be negligible based on a sigmoidal permeability–tubular reabsorption model (Scotcher et al., 2016b; Mathialagan et al., 2017). Second, a well stirred model was used to predict the CLsec,plasma. The well stirred model postulates instantaneous and complete mixing of the unbound drug between the interstitial space of the renal cortex and blood (Malcolm Rowland, 2011). Third, we assumed that the rate-determining step for CLsec,plasma of the 26 drugs studied was uptake via the designated OAT transporters; other apical or basolateral transporters, if any, were considered insignificant contributors to their CLr,plasma (Watanabe et al., 2011; Mathialagan et al., 2017).

When passive diffusion secretory CL was assumed to be negligible, the predictive power of the two approaches was not statistically different as indicated by the overlapping 95% confidence intervals of the precision and bias of the predictions of CLint,sec,in vivo, CLsec,plasma, and CLr,plasma (Table 1; Fig. 3). As expected, precision in the prediction of CLsec,plasma or CLr,plasma was respectively greater than CLint,sec,in vivo for the following reasons. First, incorporation of renal blood flow (eq. 4) in predicting CLsec,plasma dampened the contribution of CLint.sec to the total CLsec,plasma. Second, when predicting CLr,plasma, due to contribution of filtration CL, the contribution of CLint.sec,in vivo to CLr,plasma further diminishes.

Although the in vivo passive diffusion secretory CL may be negligible for some drugs, this may not be the case for other drugs. Thus, the correct approach is to compare the observed CLint,sec,in vivo with that predicted using the REF and RAF approaches after including the predicted passive diffusion secretory CL. In doing so, the predictions of all CL parameters by both the REF and RAF approaches worsened as indicated by the greater number of drugs that fell outside the 2- and 3-fold error windows (CLsec,plasma: n = 9) (Table 2; Fig. 4). For the two approaches, those drugs for which the predicted values of CLsec,plasma were worse and fell out of the 2- or 3-fold window had notable contribution from passive diffusion secretory CL. The drugs that fell out of the 2- or 3-fold window for the REF approach had a 20%–50% contribution of passive diffusion secretory CL of the total CLint,sec,in vivo. The corresponding contribution was 39%–99% for the RAF approach. Interestingly, where available, this predicted passive diffusion secretory CL was corroborated by the in vivo passive diffusion secretory CL as measured by the change in the secretory renal CL of the drug when the drug was co-administered with probenecid to inhibit OATs (Mathialagan et al., 2017). In the presence of probenecid, data were available for renal CL of three of the nine drugs that fell out of either the 2- or 3-fold windows for CLsec,plasma predictions. For those drugs, the in vivo passive diffusion secretory CL was estimated to be 32%–67% of the total renal CL, whereas our predictions for these three drugs ranged from 23% to 50%. Nevertheless, even with this deterioration in the precision of predictions, the precision and bias in CLint,sec,in vivo, CLsec,plasma, or CLr,plasma predictions by the REF and RAF approaches did not differ significantly (Supplemental Fig. 2). Irrespective of whether passive diffusion CL was included or not, there were no trends identified regarding the drugs that were outside the 2- and 3-fold error windows when drug-dependent characteristics such as molecular weight, ionization state, LogP, fu, and magnitude of CLsec,plasma were evaluated. However, there was a high degree of overlap between the REF and RAF approaches (9/13 and 14/18 drugs overlapped for the CLsec,plasma predictions with only active uptake CL and with the inclusion of passive diffusion CL, respectively). The overlap is likely due to the similar assumptions made in clearance predictions as discussed earlier. These results do point to an interesting finding that suggests that our approach to predict passive diffusion secretory CL of OAT-transported drugs needs refinement. Until that refinement has been accomplished, because predictions using only the active uptake clearance were more precise, we would suggest IVIVE of OAT-mediated renal CL of drugs based on active uptake clearance alone.

Since the prediction capability of the REF and RAF approaches were not significantly different, our results indicate that the two approaches can be used interchangeably to predict the renal secretory CL of OAT substates. When applied to clearance (renal or hepatic) via other transporters, both approaches have their pros and cons. Since the RAF approach requires a probe substrate, the REF approach is more useful when a transporter-selective probe substrate is not available for any one of the transporters of interest. For example, the REF approach can potentially be used to determine the clearance of drugs that are not solely rate-determined by a single apical (MATE1/2K, multidrug resistance protein 4, etc.) or basal (OAT1/2/3, organic cation transporter 2/3, etc.) transporter. Abundance of the basal and apical transporters in HEK293 cells and renal cortex would need to be measured to include them in the REF approach. Verification of the REF approach to test the predictive power of CLsec,plasma and CLr,plasma for the drugs with multiple clearance pathways as the rate-determining steps remains to be tested. In contrast, when a transporter-selective probe substrate is available in vivo and in vitro (if using primary cells, e.g., kidney epithelial cells), the RAF approach will likely perform better. Unfortunately, such probe substrates are rarely available. Even for the OAT substrates studied here, many are multiple OAT transporter substrates. In that event, data on multiple probe substrates, each selective for a given OAT, are needed. In addition, the RAF approach assumes that the in vivo renal secretory CL is the only rate-determining step in the systemic renal CL of the drug. If the apical transporters are involved (e.g., MATEs or P-glycoprotein), this assumption will not hold, and therefore the estimation of the renal CL using the RAF approach will be inaccurate. On the other hand, assuming no passive diffusion secretory CL, the REF approach (but not the RAF approach) is highly dependent on the PSF used as demonstrated by the 10-fold difference in the CLsec,plasma predictions when the cortex versus kidney PSF was used (Supplemental Fig. 1). However, it is important to note that when the passive diffusion secretory clearance is included to predict the CLsec,plasma, both approaches, REF and RAF, need to use PSF, and therefore estimation of this parameter will be highly dependent on the PSF value. Therefore, we gathered literature values on the various PSFs determined by us and others (Table 3). The kidney PSF used by Mathialagan et al. (2017) is the lowest, whereas the one used here is the highest reported (15–210 mg protein/g kidney; Table 3) (Mitchell et al., 1945; Forbes et al., 1953; Cooper et al., 1956; Snyder, 1979; Pacifici et al., 1988; Knights et al., 2016; Mathialagan et al., 2017; Scotcher et al., 2017; Kumar et al., 2018). Thus, it is imperative that the correct PSF value be used when using both approaches, and, going forward, a consensus is needed on the PSF value that should be used.

Human renal clearance predictions are often based on preclinical animal data (i.e., rat and dog CLr) due to lack of reliable in vitro based approaches (Paine et al., 2011). The REF or RAF approaches, which were demonstrated to provide reasonable IVIVE in our study, can be employed to project renal clearance in drug discovery setting and thus enable dose predictions. Additionally, this approach allows for quantitating individual transporter contribution (fraction transported) to the overall renal secretion, which allows for drug-drug interaction predictions in drug development.

In conclusion, using the same in vitro and in vivo data set, we showed that the REF and RAF approaches were not significantly different in their ability to predict CLr of OAT substrates. However, for drugs that have renal (or hepatic) CL rate-determined by both basal and apical transporters, the REF approach has an advantage over the RAF approach. This is because the latter is dependent on the availability of in vivo renal CL data for a probe drug. It is highly unlikely that such data are possible to obtain with currently approved drugs. Despite the theoretical advantage of the REF approach (i.e., it does not require data on probe drugs), its ability to simultaneously and accurately predict renal clearance when multiple rate-determining steps (and therefore transporters) are involved remains to be tested.

Acknowledgments

The authors would like to thank A. David Rodrigues (Pfizer, Inc.) and Emi Kimoto (Pfizer, Inc.) for their input during this work. Matthew Karasu supported the proteomics sample preparation.

Authorship Contributions

Participated in research design: Prasad, Varma, Unadkat

Conducted experiments: Prasad, Bhatt, Mathialagan

Performed data analysis: Kumar, Bhatt, Mathialagan

Wrote or contributed to the writing of the manuscript: Kumar, Prasad, Bhatt, Mathialagan Varma, Unadkat

Footnotes

    • Received January 10, 2021.
    • Accepted March 26, 2021.

  • This work was supported in part by funding from Pfizer Inc. A.R.K. was supported by National Institutes of Health National Institute of General Medical Sciences [Grant GM007750].

  • https://dx.doi.org/10.1124/dmd.121.000367.

  • ↵Embedded ImageThis article has supplemental material available at dmd.aspetjournals.org.

Abbreviations

BCA
bicinchoninic acid
CL
clearance
CLint,sec,in vivo
in vivo intrinsic secretory clearance
CLr
renal clearance
CLr,plasma
renal plasma clearance
CLsec,plasma
renal secretory clearance
fu
unbound fraction
HEK
human embryonic kidney
IVIVE
in vitro–to–in vivo extrapolation
MATE
multi-antimicrobial extrusion protein
ME
mean error
MS
mass spectrometry
OAT
organic anion transporter
PSF
physiological scaling factor
RAF
relative activity factor
REF
relative expression factor
RMSE
root mean squared error

References

    1. Al-Jahdari WS,
    2. Yamamoto K,
    3. Hiraoka H,
    4. Nakamura K,
    5. Goto F, and , and
    6. Horiuchi R
    (2006) Prediction of total propofol clearance based on enzyme activities in microsomes from human kidney and liver. Eur J Clin Pharmacol 62:527–533.
    OpenUrlCrossRefPubMed
  2. ↵
    1. Bouchet LG,
    2. Bolch WE,
    3. Blanco HP,
    4. Wessels BW,
    5. Siegel JA,
    6. Rajon DA,
    7. Clairand I, and
    8. Sgouros G
    (2003) MIRD Pamphlet No 19: absorbed fractions and radionuclide S values for six age-dependent multiregion models of the kidney. J Nucl Med 44:1113–1147.
    OpenUrlAbstract/FREE Full Text
  3. ↵
    1. Chu X,
    2. Bleasby K, and
    3. Evers R
    (2013) Species differences in drug transporters and implications for translating preclinical findings to humans. Expert Opin Drug Metab Toxicol 9:237–252.
    OpenUrlCrossRefPubMed
  4. ↵
    1. Cooper AR,
    2. Forbes RM, and
    3. Mitchell HH
    (1956) Further studies on the gross composition and mineral elements of the adult human body. J Biol Chem 223:969–975.
    OpenUrlFREE Full Text
  5. ↵
    1. Feng B,
    2. LaPerle JL,
    3. Chang G, and
    4. Varma MV
    (2010) Renal clearance in drug discovery and development: molecular descriptors, drug transporters and disease state. Expert Opin Drug Metab Toxicol 6:939–952.
    OpenUrlCrossRefPubMed
  6. ↵
    1. Forbes RM,
    2. Cooper AR, and
    3. Mitchell HH
    (1953) The composition of the adult human body as determined by chemical analysis. J Biol Chem 203:359–366.
    OpenUrlFREE Full Text
  7. ↵
    1. Ishida K,
    2. Ullah M,
    3. Tóth B,
    4. Juhasz V, and
    5. Unadkat JD
    (2018) Successful prediction of in vivo hepatobiliary clearances and hepatic concentrations of rosuvastatin using sandwich-cultured rat hepatocytes, transporter-expressing cell lines, and quantitative proteomics. Drug Metab Dispos 46:66–74.
    OpenUrlAbstract/FREE Full Text
  8. ↵
    1. Ke AB,
    2. Nallani SC,
    3. Zhao P,
    4. Rostami-Hodjegan A, and
    5. Unadkat JD
    (2014) Expansion of a PBPK model to predict disposition in pregnant women of drugs cleared via multiple CYP enzymes, including CYP2B6, CYP2C9 and CYP2C19. Br J Clin Pharmacol 77:554–570.
    OpenUrlCrossRefPubMed
  9. ↵
    1. Knights KM,
    2. Spencer SM,
    3. Fallon JK,
    4. Chau N,
    5. Smith PC, and
    6. Miners JO
    (2016) Scaling factors for the in vitro-in vivo extrapolation (IV-IVE) of renal drug and xenobiotic glucuronidation clearance. Br J Clin Pharmacol 81:1153–1164.
    OpenUrl
  10. ↵
    1. Kumar V,
    2. Yin J,
    3. Billington S,
    4. Prasad B,
    5. Brown CDA,
    6. Wang J, and
    7. Unadkat JD
    (2018) The importance of incorporating OCT2 plasma membrane expression and membrane potential in IVIVE of metformin renal secretory clearance. Drug Metab Dispos 46:1441–1445.
    OpenUrlAbstract/FREE Full Text
  11. ↵
    1. Kunze A,
    2. Huwyler J,
    3. Camenisch G, and
    4. Poller B
    (2014a) Prediction of organic anion-transporting polypeptide 1B1- and 1B3-mediated hepatic uptake of statins based on transporter protein expression and activity data. Drug Metab Dispos 42:1514–1521.
    OpenUrlAbstract/FREE Full Text
  12. ↵
    1. Kunze A,
    2. Huwyler J,
    3. Poller B,
    4. Gutmann H, and
    5. Camenisch G
    (2014b) In vitro-in vivo extrapolation method to predict human renal clearance of drugs. J Pharm Sci 103:994–1001.
    OpenUrl
  13. ↵
    1. Li CY,
    2. Hosey-Cojocari C,
    3. Basit A,
    4. Unadkat JD,
    5. Leeder JS, and
    6. Prasad B
    (2019) Optimized renal transporter quantification by using aquaporin 1 and aquaporin 2 as anatomical markers: application in characterizing the ontogeny of renal transporters and its correlation with hepatic transporters in paired human samples. AAPS J 21:88.
    OpenUrl
  14. ↵
    1. Troy DB
    1. Malcolm Rowland TNT
    (2011) Well-stirred model of hepatic clearance, in Clinical Pharmacokinetics and Pharmacodynamics: Concepts and Applications (Troy DB, ed) p 4, Lippincott Williams & Wilkins, Philadelphia.
  15. ↵
    1. Mathialagan S,
    2. Piotrowski MA,
    3. Tess DA,
    4. Feng B,
    5. Litchfield J, and
    6. Varma MV
    (2017) Quantitative prediction of human renal clearance and drug-drug interactions of organic anion transporter substrates using in vitro transport data: a relative activity factor approach. Drug Metab Dispos 45:409–417.
    OpenUrlAbstract/FREE Full Text
  16. ↵
    1. Mitchell H,
    2. Hamilton T,
    3. Steggerda F, and
    4. Bean H
    (1945) The chemical composition of the adult human body and its bearing on the biochemistry of growth. J Biol Chem 158:625–637.
    OpenUrlFREE Full Text
  17. ↵
    1. Pacifici GM,
    2. Franchi M,
    3. Bencini C,
    4. Repetti F,
    5. Di Lascio N, and
    6. Muraro GB
    (1988) Tissue distribution of drug-metabolizing enzymes in humans. Xenobiotica 18:849–856.
    OpenUrlCrossRefPubMed
  18. ↵
    1. Paine SW,
    2. Ménochet K,
    3. Denton R,
    4. McGinnity DF, and
    5. Riley RJ
    (2011) Prediction of human renal clearance from preclinical species for a diverse set of drugs that exhibit both active secretion and net reabsorption. Drug Metab Dispos 39:1008–1013.
    OpenUrlAbstract/FREE Full Text
  19. ↵
    1. Patilea-Vrana G and
    2. Unadkat JD
    (2016) Transport vs. metabolism: what determines the pharmacokinetics and pharmacodynamics of drugs? Insights from the extended clearance model. Clin Pharmacol Ther 100:413–418.
    OpenUrl
  20. ↵
    1. Prasad B,
    2. Johnson K,
    3. Billington S,
    4. Lee C,
    5. Chung GW,
    6. Brown CD,
    7. Kelly EJ,
    8. Himmelfarb J, and
    9. Unadkat JD
    (2016) Abundance of drug transporters in the human kidney cortex as quantified by quantitative targeted proteomics. Drug Metab Dispos 44:1920–1924.
    OpenUrlAbstract/FREE Full Text
  21. ↵
    1. Rostami-Hodjegan A and
    2. Tucker GT
    (2007) Simulation and prediction of in vivo drug metabolism in human populations from in vitro data. Nat Rev Drug Discov 6:140–148.
    OpenUrlCrossRefPubMed
  22. ↵
    1. Rowland M,
    2. Peck C, and
    3. Tucker G
    (2011) Physiologically-based pharmacokinetics in drug development and regulatory science. Annu Rev Pharmacol Toxicol 51:45–73.
    OpenUrlCrossRefPubMed
  23. ↵
    1. Sachar M,
    2. Kumar V,
    3. Gormsen LC,
    4. Munk OL, and
    5. Unadkat JD
    (2020) Successful prediction of positron emission tomography-imaged metformin hepatic uptake clearance in humans using the quantitative proteomics-informed relative expression factor approach. Drug Metab Dispos 48:1210–1216.
    OpenUrlAbstract/FREE Full Text
  24. ↵
    1. Scotcher D,
    2. Billington S,
    3. Brown J,
    4. Jones CR,
    5. Brown CDA,
    6. Rostami-Hodjegan A, and
    7. Galetin A
    (2017) Microsomal and cytosolic scaling factors in dog and human kidney cortex and application for in vitro-in vivo extrapolation of renal metabolic clearance. Drug Metab Dispos 45:556–568.
    OpenUrlAbstract/FREE Full Text
  25. ↵
    1. Scotcher D,
    2. Jones C,
    3. Posada M,
    4. Rostami-Hodjegan A, and
    5. Galetin A
    (2016a) Key to opening kidney for in vitro-in vivo extrapolation entrance in health and disease: Part I: In vitro systems and physiological data. AAPS J 18:1067–1081.
    OpenUrlCrossRef
  26. ↵
    1. Scotcher D,
    2. Jones C,
    3. Rostami-Hodjegan A, and
    4. Galetin A
    (2016b) Novel minimal physiologically-based model for the prediction of passive tubular reabsorption and renal excretion clearance. Eur J Pharm Sci 94:59–71.
    OpenUrl
  27. ↵
    1. Snyder WS
    (1979) Report of the task group on reference man. Ann ICRP 3:iii.
    OpenUrlPubMed
  28. ↵
    1. Soars MG,
    2. McGinnity DF,
    3. Grime K, and
    4. Riley RJ
    (2007) The pivotal role of hepatocytes in drug discovery. Chem Biol Interact 168:2–15.
    OpenUrlCrossRefPubMed
  29. ↵
    1. Varma MV,
    2. Feng B,
    3. Obach RS,
    4. Troutman MD,
    5. Chupka J,
    6. Miller HR, and
    7. El-Kattan A
    (2009) Physicochemical determinants of human renal clearance. J Med Chem 52:4844–4852.
    OpenUrlCrossRefPubMed
  30. ↵
    1. Watanabe T,
    2. Kusuhara H,
    3. Watanabe T,
    4. Debori Y,
    5. Maeda K,
    6. Kondo T,
    7. Nakayama H,
    8. Horita S,
    9. Ogilvie BW,
    10. Parkinson A et al.
    (2011) Prediction of the overall renal tubular secretion and hepatic clearance of anionic drugs and a renal drug-drug interaction involving organic anion transporter 3 in humans by in vitro uptake experiments. Drug Metab Dispos 39:1031–1038.
    OpenUrlAbstract/FREE Full Text
Back to top

In this issue

Download PDF
Article Alerts
Email Article
Citation Tools

Research ArticleArticle

REF Versus RAF Prediction of Renal Clearance

Aditya R. Kumar, Bhagwat Prasad, Deepak Kumar Bhatt, Sumathy Mathialagan, Manthena V. S. Varma and Jashvant D. Unadkat
Drug Metabolism and Disposition June 1, 2021, 49 (6) 470-478; DOI: https://doi.org/10.1124/dmd.121.000367
Twitter logo Facebook logo Mendeley logo

Jump to section

Advertisement