Successful Prediction of Fetal Exposure to Dual BCRP/P‐gp Drug Substrates Using the Efflux Ratio‐Relative Expression Factor Approach and PBPK M&S

To inform fetal drug safety, it is important to determine or predict fetal drug exposure throughout pregnancy. The former is not possible in the first or second trimester. In contrast, at the time of birth, fetal drug exposure, relative to maternal exposure, can be estimated as Kp,uu (unbound fetal umbilical venous (UV) plasma area under the curve (AUC)/unbound maternal plasma (MP) AUC), provided the observed UV/MP values, spanning the dosing interval, are available from multiple maternal‐fetal dyads. However, this fetal Kp,uu cannot be extrapolated to other drugs. To overcome the above limitations, we have used an efflux ratio‐relative expression factor (ER‐REF) approach to successfully predict the fetal Kp,uu of P‐gp substrates. Because many drugs taken by pregnant people are also BCRP substrates, here, we extend this approach to drugs that are effluxed by both placental BCRP and P‐gp or P‐gp alone. To verify our predictions, we chose drugs for which UV/MP data were available at term: glyburide and imatinib (both BCRP and P‐gp substrates) and nelfinavir (only P‐gp substrate). First, the ER of the drugs was determined using Transwells and MDCKII cells expressing either BCRP or P‐gp. Then, the ER was scaled using the proteomics‐informed REF value to predict the fetal Kp,uu of the drug at term. The ER‐REF predicted fetal Kp,uu of glyburide (0.43), imatinib (0.42), and nelfinavir (0.40) fell within two‐fold of the corresponding in vivo fetal Kp,uu (0.44, 0.37, and 0.46, respectively). These data confirm that the ER‐REF approach can successfully predict fetal drug exposure to BCRP/P‐gp and P‐gp substrates, at term.

Successful Prediction of Fetal Exposure to Dual BCRP/P-gp Drug Substrates Using the Efflux Ratio-Relative Expression Factor Approach and PBPK M&S Ankit Balhara 1 , Mengyue Yin 1 and Jashvant D. Unadkat 1, * To inform fetal drug safety, it is important to determine or predict fetal drug exposure throughout pregnancy.The former is not possible in the first or second trimester.In contrast, at the time of birth, fetal drug exposure, relative to maternal exposure, can be estimated as K p,uu (unbound fetal umbilical venous (UV) plasma area under the curve (AUC)/unbound maternal plasma (MP) AUC), provided the observed UV/MP values, spanning the dosing interval, are available from multiple maternal-fetal dyads.However, this fetal K p,uu cannot be extrapolated to other drugs.To overcome the above limitations, we have used an efflux ratio-relative expression factor (ER-REF) approach to successfully predict the fetal K p,uu of P-gp substrates.Because many drugs taken by pregnant people are also BCRP substrates, here, we extend this approach to drugs that are effluxed by both placental BCRP and P-gp or P-gp alone.To verify our predictions, we chose drugs for which UV/MP data were available at term: glyburide and imatinib (both BCRP and P-gp substrates) and nelfinavir (only P-gp substrate).First, the ER of the drugs was determined using Transwells and MDCKII cells expressing either BCRP or P-gp.Then, the ER was scaled using the proteomicsinformed REF value to predict the fetal K p,uu of the drug at term.The ER-REF predicted fetal K p,uu of glyburide (0.43), imatinib (0.42), and nelfinavir (0.40) fell within two-fold of the corresponding in vivo fetal K p,uu (0.44, 0.37, and 0.46, respectively).These data confirm that the ER-REF approach can successfully predict fetal drug exposure to BCRP/ P-gp and P-gp substrates, at term.More than 80% of pregnant people take medication. 1,2However, most US Food and Drug Administration (FDA)-approved drugs taken by pregnant people (prescribed or over-the-counter) are used "off-label" and lack information on their pharmacokinetics (PK), safety, and efficacy during pregnancy. 2This is because pregnant people are often excluded from clinical trials out of concerns of potential toxicity to the fetus. 3hysiological changes during pregnancy can significantly affect the PK of drugs in the maternal-fetal unit and therefore, maternalfetal drug safety and efficacy. 4Because the rate and extent of fetal exposure to a drug likely drives the drug's fetal safety and efficacy, these parameters, which are easier to measure or predict, can be used to inform fetal drug safety and efficacy.The latter is relevant when the fetus is the target of drug therapy (e.g., to prevent vertical HIV transmission, treatment of fetal arrythmia, or maturation of fetal lungs due to premature labor). 5he rate and extent of fetal drug exposure is driven by three factors: (i) rate and extent of maternal exposure, determined by maternal PK; (ii) the rate and extent of drug passage through the placenta determined by transport, passive diffusion, and metabolism of the drug; and (iii) rate and extent of drug elimination from the fetus. 6Determining a drug's fetal PK is challenging, even at the time of birth when fetal drug plasma concentrations can be quantified by simultaneously collecting umbilical vein (UV) and maternal plasma (MP) samples.However, a single UV/MP value does not provide estimates of the rate and extent of fetal drug exposure.Therefore, UV and MP samples, pooled from multiple maternal-fetal dyads over multiple time-points (spread out over the dosing interval), are needed to allow such estimation.6][7][8][9] K p,uu is the ratio of fetal to maternal unbound plasma area under the curves (AUCs) after single-or multiple-dose drug administration or the corresponding average steady-state plasma concentrations after multiple-dose administration.However, the above approach cannot be generalized; that is, estimation of a single drug's K p,uu cannot be extrapolated to other drugs.Moreover, rich UV/MP data are not available for most drugs taken by pregnant people.In addition, the above approach is ethically not possible to implement at earlier gestational ages. 6To overcome these limitations, we have devised an in vitro to in vivo extrapolation approach to predict the rate and extent of fetal drug exposure to a drug, independent of the availability of UV/MP data.This approach called the efflux ratio-relative expression factor (ER-REF) approach relies on determining (i) net transport clearance of the drug, estimated as the ER, in Transwell experiments, using cells overexpressing the transporter of interest; and (ii) the ratio of transporter abundance ex vivo in the tissue of interest (in this case, the placenta) and transporter-overexpressing cells (REF), determined using quantitative targeted proteomics (Figure 1).The efflux transporteroverexpressing cell monolayer in the in vitro Transwell system mimics the placental syncytiotrophoblast layer in vivo, and the apical and basal chambers in the in vitro system, mimics the in vivo maternal and fetal blood compartments, respectively, enabling the ER-REF approach to predict the in vivo fetal K p,uu . 6The REF approach assumes that k cat of the transporters is identical in vivo and in the in vitro system.Based on this assumption, the activity of the transporters can be scaled from in vitro to in vivo based solely on their abundances in the human placenta and transporteroverexpressing cells.The ER-REF approach can also predict fetal exposure to drugs at earlier gestational age provided the abundance of the placental transporter of interest and fetal clearance, at these earlier gestational ages, is available. 10We have verified the ability of the ER-REF approach to predict the K p,uu of four P-gp substrates, namely dexamethasone, betamethasone, darunavir, and lopinavir. 8This approach needs to be verified for substrates of other placental drug transporters, abundant in the human placenta, such as the breast cancer resistance protein (BCRP). 10,11lthough very few drugs are solely transported by BCRP, 12 many drugs taken by pregnant people are transported by both BCRP and P-gp.3][24][25] In addition, to determine if our previous success in predicting K p,uu could be extended to other P-gp substrates, 8 we also included a drug that is transported by P-gp and not by BCRP, namely nelfinavir. 26,27To verify our predictions, rich UV/MP data (at term) were available for all these drugs from which the in vivo K p,uu of the drugs could be estimated.

MATERIALS AND METHODS
For Materials, please see Supplementary Material.
Quantification of the in vitro ER of the drugs as well as BCRP and P-gp abundance in MDCK-hBCRP cMDR1-KO and MDCK-hMDR1 cMDR1-KO cells MDCKII cells expressing human BCRP or P-gp, with the endogenous canine P-gp knocked-out (MDCK-hBCRP cMDR1-KO and MDCK-hMDR1 cMDR1-KO ), 28,29 (generously provided by Drs.Karlgren and Artursson, Uppsala University) were cultured in DMEM medium containing 10% fetal bovine serum, 1% penicillin (10,000 U/mL)/streptomycin (10,000 g/mL), and 1% Glutamax.These cells will henceforth be referred to as BCRP and P-gp expressing cells, respectively.Geneticin (670 μg/mL) and hygromycin B (375 μg/mL) were used as selective antibiotics for BCRP and P-gp expressing cells, respectively.The culture conditions were maintained at 37°C, 5% CO 2 , and 95% humidity.The cells were harvested using trypsin and sub-cultured twice a week.
The ER of efavirenz (1 μM), glyburide (0.1, 1 μM), imatinib (0.5, 5 μM), or nelfinavir (1 μM) was determined in BCRP or P-gp expressing cells in three independent experiments, each conducted in triplicate.The limits of quantification of our liquid-chromatography tandem mass spectrometry assay did not allow us to go beyond the lowest concentrations listed above.The integrity of the tight junctions was assessed by lucifer yellow (50 μM) for all Transwell plates.Concurrently, separate Transwell plates were used to determine the ER of prazosin (3 μM) and quinidine (3 μM), markers of BCRP and P-gp activity, respectively.Where included, Ko143 (5 μM) and tariquidar (5 μM) were used as inhibitors of BCRP and P-gp, respectively.The stock solution of test and marker compounds as well as inhibitors were prepared in DMSO and concentration of DMSO was < 0.2% in the final solution.
On day 0, 0.6 × 10 6 cells/well were plated on the apical side of the 12-well Transwell insert.The medium was changed on days 2 and 3; for BCRP cells, on day 3, the medium was supplemented with 10 mM sodium butyrate.On day 4, cells were washed 3 times with transport buffer (10 mM HEPES in HBSS, pH 7.4, maintained at 37°C) and incubated for 15 minutes in an orbital shaker at 120 rpm.Then, bi-directional transport of the drugs was determined using different Transwells, that is, either in the A → B or the B → A direction, where the apical (A) and the basal (B) compartment volumes were 0.5 and 1 mL, respectively.Transport was initiated by adding the donor solution (transport buffer containing the test or marker drugs and lucifer yellow (± inhibitor)) to the donor compartment (A or B).The receiver compartment was sampled (100 μL) at 60, 120, 180, and 240 minutes (for efavirenz, glyburide, imatinib, and nelfinavir), or 30, 60, and 90 minutes (for quinidine), or 30, 60, 90, and 120 minutes (for prazosin).After sampling, the receiver compartment volume was replenished with an equivalent volume of the transport buffer ± inhibitor (and this dilution was accounted for in our ER calculation).The donor compartment was sampled (5 μL) at 0 minutes and at the end of the transport experiment when the cells were immediately washed 3 times with the ice-cold transport buffer and lysed for total protein content and targeted proteomics quantification (see Supplementary Material).
The ER of each test or marker drug by a given transporter, in the presence (ER +inh ) and absence of the inhibitor (ER −inh ), was determined as we have described before. 8,30Data from Transwells were considered acceptable only if the permeability of lucifer yellow was < 2 × 10 −6 cm/s and the ER of the marker compounds was > 7. 8 Prediction of fetal K p,uu from in vitro studies using the ER-

REF approach
The ER values were scaled using the REF to predict the in vivo fetal K p,uu (Eq.1): where transporter refers to either BCRP or P-gp, HP is the total protein in the homogenate of the placenta, BCRP, or P-gp expressing cells.
Where multiple transporters efflux the drug, the in vivo fractional contribution to efflux (f t ) by each transporter can be estimated as: (1)

REF transporter =
Abundance of the transporter in the human placenta pmol∕mg HP Abundance of trasnporter in transporter overexpressing cells pmol ∕mg HP (4) Estimation of the in vivo fetal K p,uu of the drugs The in vivo fetal K p,uu of imatinib (0.37) and nelfinavir (0.46), previously estimated using m-f PBPK M&S and their respective UV and MP concentrations at term, were used with a slight modification. 9In this previous publication, 9 the in vivo fetal K p,uu of these drugs were estimated by PBPK M&S without incorporating the variability in placental efflux transporters.Here, using the Simcyp simulator, we re-estimated these in vivo fetal K p,uu values after incorporating the percent coefficient of variation in the placental transporter abundance, as reported by Anoshchenko et al., 2020. 10 This modification resulted in minor changes in the fetal K p,uu values.The same approach was used to estimate the in vivo fetal K p,uu of glyburide at term.Briefly, we first developed and verified a glyburide PBPK model for the nonpregnant population (see Supplementary Material).Then, without changing the drug-specific parameters (except as described below), this model was extended to the pregnant population using the default Simcyp pregnant population model.Only a single glyburide UV/MP term dataset is available in the literature and this dataset demonstrates large interindividual variability in glyburide PK, especially in the absorption phase. 31Fixing the absorption parameters to those verified in the non-pregnant population resulted in underprediction of the glyburide absorption phase profile at term.Thus, only the glyburide absorption parameters ( f a , k a , and lag time) were optimized to match the observed data (Table S3).K p,uu of efavirenz was not estimated, as our data showed that it is not a substrate of BCRP or P-gp (see "Results" section below).
The pregnant Simcyp population was modified to include 2-fold induction of hepatic CYP3A4 enzyme throughout the pregnancy as compared with the non-pregnant population based on our previous study. 32dditionally, the induction of CYP2C9 by 1.4-, 1.5-, and 1.6-fold during the first, second, and third trimesters was incorporated in the pregnant population. 33Data on changes in hepatic CYP2C19 and transporter activity are not available or questionable 34,35 and thus were not incorporated in the pregnant population model.Afterwards, we optimized the passive and active placental clearances of glyburide to match the observed UV/ MP data, as described previously. 8,9Briefly, intrinsic placental passive diffusion clearance (CL int,PD,placenta ) of the glyburide was estimated as: where P app,glyburide and P app,midazolam are apparent membrane permeability values of glyburide (P app,glyburide = 1.46 × 10 −5 cm/s; mean value in Caco-2 36 and MDCKII-LE cells 37 ); and midazolam (P app,midazolam = 4.9 × 10 −5 cm/s).The estimated in vivo CL int,PD,placenta of midazolam was 500 L/h. 8,9,38The resulting CL int,PD,placenta of glyburide was 149 L/h, which is much greater than the placental blood flow at term (~ 45 L/h).Therefore, the in vivo CL int,PD,Placenta of glyburide was considered to be perfusionlimited (45 L/h).
After incorporating the drug's CL int,PD,palcenta in the permeabilitylimited placenta model of the Simcyp pregnancy PBPK model, the drug's CL int,efflux,placenta was adjusted until the predicted UV/MP values best matched the observed data (absolute average fold error (AAFE) = 1.0).The AAFE in the predictions of UV/MP values was calculated as per Eq.6: Then, the fetal K p,uu was estimated (Eq. 7), where AUC fetal,u and AUC maternal,u are the area under the curve of the unbound umbilical vein plasma concentration-time profile, and unbound maternal plasma concentration-time profile, and N is the number of observed and predicted UV/MP values.The overall in vivo f t of drug by placental transporters can be determined as 1 − K p,uu .The fraction unbound of the drug was predicted by Simcyp taking into consideration the change in fetal and maternal plasma protein concentrations during pregnancy. 6,39A virtual population of 100 individuals was simulated to generate the predicted mean as well as the 5 th and the 95 th percentile profiles (90% confidence interval (90% CI)).
PBPK model predicted fetal K p,uu of the glyburide and imatinib at earlier gestational ages (GW15 and GW25) The fetal K p,uu of nelfinavir at earlier gestational ages was reported in our earlier publication 9 but without incorporating the variability in placental transporters abundance.Thus, its fetal K p,uu was re-estimated (Figure S2) using the same trial design as previously reported, 9 but after incorporating the variability in placental transporters abundance values reported by Anoshchenko et al. 10 The early gestational age K p,uu values for imatinib were not estimated in that publication as the relative contribution of BCRP and P-gp to its placental efflux was unknown.Now that we have these contributions for both imatinib and glyburide, their fetal K p,uu were predicted at earlier gestational ages (GW15 and GW25), by incorporating the previously quantified total placental P-gp and BCRP abundance (and their variability) at different gestational ages 10 into the Simcyp pregnancy module.To do so, the CL int,efflux,placenta was segregated into clearances mediated by BCRP and P-gp (CL int,BCRP,placenta and CL int,P-gp,placenta ) based on the relative abundances of these transporters at different gestation ages.The CL int,BCRP,placenta and CL int,P-gp,placenta values, as well as CL int,PD,placenta were incorporated into the model (Table S6).Then, the maternal-fetal PK profiles and fetal K p,uu of the glyburide and imatinib at GW15 and GW25 were predicted using the same trial design as for term.

Statistical analyses and verification of predictions
The 90% CI of the predicted fetal K p,uu was generated using pooled variance approach, where the variability in ER and REF were included.The predicted fetal K p,uu (using the ER-REF approach) was considered verified if the mean predicted fetal K p,uu fell within 2-fold of the observed fetal K p,uu , or the predicted/observed K p,uu was within 0.5-2.0.We acknowledge that this acceptance criterion should be more stringent for drugs with narrow therapeutic index.

ER of glyburide, imatinib, nelfinavir, and efavirenz in Transwell assays using BCRP and P-gp expressing cells
As expected, glyburide and imatinib were transported by both BCRP and P-gp, as evidenced by their BCRP-and P-gpmediated ER in BCRP and P-gp expressing cells (Figure 2, Table 1).The transporter-mediated ER (normalized to that of the marker drug) of glyburide, for both BCRP and P-gp, was not significantly different between the two concentrations (0.1 and 1 μM; p > 0.05, unpaired ttest).In contrast, the markernormalized transport-mediated ER of imatinib for BCRP (pvalue = 0.0148, unpaired ttest) was significantly larger at 0.5 than at 5 μM concentrations, whereas that for P-gp was not significantly different between the two concentrations (p > 0.05, unpaired ttest; Figure S1).Nelfinavir was transported by P-gp (Figure 2) and not by BCRP (data not shown).Unexpectedly, efavirenz showed no efflux by either BCRP or P-gp as evidence (5)   CL int,PD,placenta of glyburide = P app,glyburide P app,midazolam × CL int,PD,placenta of midazolam Fetal K p,uu = AUC fetal,u by an ER ~ 1 in both BCRP and P-gp expressing cells.Therefore, further work on this drug was not pursued.Consistent with previously published data, the ER of the marker drugs, prazosin (for BCRP) or quinidine (for P-gp), were greater than our acceptable in-house cutoff value (> 7). 8Based on the above data, the ER of the test compound, determined at the lowest possible concentration (Figure 2, Table 1), was used in the analyses below.

Development and verification of glyburide PBPK model for the non-pregnant and pregnant populations
The developed PBPK model of glyburide successfully captured the observed data of both the training and verification datasets (Figure 3) as evidenced by the predicted vs. observed PK parameters falling within 0.8-1.25 (Table S4).The model was further verified by confirming that it could capture the magnitude of DDI with rifampin, ciprofloxacin, clarithromycin, and erythromycin (Table S5), drugs that interact with OATPs, CYP2C8, and CYP3A4. 36,40The verified non-pregnant glyburide PBPK was successfully extrapolated to pregnancy; the predicted PK profile and parameters fell within 0.8-and 1.25 of the observed values (Figure 3, Table S4).
Estimates of in vivo fetal K p,uu obtained using a m-f PBPK model The fetal K p,uu values of imatinib and nelfinavir, estimated after incorporation of variability in placental transporter abundance, were 0.37 and 0.46, respectively. 9In our previous publication, their estimates were 0.35 and 0.41. 9The optimized in vivo glyburide placental efflux clearance, CL int,efflux,placenta (63.59 L/h or 1,545 μL/min/mL placental volume), best captured the observed UV/MP data at term (AAFE = 1.0 and observed UV/MP data fell within the 5 th and 95 th percentiles of the simulated data; Figure 4), yielding glyburide K p,uu of 0.44.When glyburide placental efflux was eliminated, the K p,uu, as expected, was 1 (AAFE = 3.1).

Prediction and verification of fetal K p,uu using the ER-REF approach
The ER-REF predicted in vivo fetal K p,uu (mean and 90% CI; Eq. 1), fell within two-fold of the mean observed in vivo K p,uu (estimated by m-f PBPK modeling; Table 1, Figure 5).These K p,uu values translated into mean in vivo total f t of 0.56, 0.63, and 0.54 for glyburide, imatinib, and nelfinavir, respectively.Of these f t values, the fractional contribution by BCRP and P-gp (Eq.4; Table 1) was about the same for glyburide and imatinib, while for nelfinavir it was 1.0 for P-gp.
Prediction of glyburide and imatinib K p,uu at earlier gestational ages (GW15 and GW25) Like nelfinavir, 9 the predicted fetal K p,uu of both glyburide and imatinib decreased with gestational age (Figure 4, Figure S2).
Correspondingly, the predicted CL int,BCRP,placenta , CL int,P-gp,placenta , and CL int,PD,placenta values of both drugs increased with gestational age (Table S6).

DISCUSSION
Using the ER-REF approach, combined with m-f PBPK modeling, we extended our previous research 8 and successfully predicted the in vivo term fetal K p,uu of drugs that are effluxed by both placental BCRP and P-gp or by only P-gp.Several factors were incorporated in our research design to enhance the fidelity of our K p,uu predictions.First, we used transfected cell lines that lack endogenous canine P-gp activity, ensuring that the measured ER values were not confounded by such activity.Second, rather than use our in-house historical data, we quantified the abundance of the transporters, BCRP and P-gp, in cells that were plated to determine the ER of the drugs.Third, we confirmed that the ER of the drugs (glyburide) was independent of the concentration used or we used the lowest possible concentration allowed by the sensitivity of our assay (imatinib or nelfinavir).Fourth, we always included in our experiments a positive transporter marker, namely prazosin for BCRP and  1 for a detailed summary of the data).ER, efflux ratio.
quinidine for P-gp, to confirm robust transport activity in the cells.Finally, the K p,uu was predicted using rich UV/MP data, obtained from multiple maternal-fetal dyads.This was important to mitigate the significant interindividual variability in the UV concentration that can arise from significant interindividual variability in MP.
Our results are consistent with previous findings that glyburide [16][17][18][19][20][21] and imatinib [22][23][24][25] are excellent substrates of both BCRP and P-gp, whereas nelfinavir is an excellent substrate of P-gp, but not of BCRP (Figure 2, Table 1). 26,27Consistent with the previous reports, 13,14 efavirenz was found not to be a P-gp substrate.However, contrary to a previous report, 15 we found that it is not a substrate of BCRP either.Hence, this drug was not studied further.Nevertheless, it is intriguing that the estimated K p,uu of efavirenz is 0.39 9 suggesting the involvement of other placental transporters or extensive placental/fetal metabolism of the drug.These potential drug transporters and metabolizing enzymes in the placenta and the fetus are discussed in detail in our previous publication. 6ur ER-REF approach resulted in excellent prediction of the observed in vivo K p,uu values of the drugs at term (predicted/ observed values ranged 0.87-1.13; Figure 5, Table 1), demonstrating the success of this approach.The predicted and observed K p,uu of these drugs (~ 0.4) are consistent with our findings that these drugs are excellent substrates of the efflux transporter(s) highly abundant in the human placenta. 10,11Consequently, these transporter(s) restrict fetal exposure to drugs.Hence, drugs coadministered with glyburide, imatinib, or nelfinavir, that are inhibitors or inducers of these placental transporters, have the potential to respectively enhance or further reduce fetal exposure to these drugs.In addition, genetic polymorphism of BCRP or P-gp can result in reduced function 41 and therefore enhance fetal exposure to these drugs.Understanding the mechanisms and extent of fetal exposure to drugs can help predict potential changes in fetal drug exposure in the presence of drug-drug interaction or transporter genetic polymorphism.
Previously, we demonstrated the success of the ER-REF approach to predict fetal K p,uu at term of four P-gp substrate drugs. 8ere, we showed that the ER-REF approach has the capability to successfully predict K p,uu of drugs that are dual substrates of the efflux transporters, BCRP and P-gp.In addition, once again, we confirmed the ability of the ER-REF to predict the in vivo fetal K p,uu of a drug solely transported by P-gp (nelfinavir).Combined with our previous study, this success enhances our confidence in using the ER-REF approach to predict in vivo fetal K p,uu of any drug transported across the placenta, including those that are substrates of placental influx transporters.We note here that we are not aware of any drug that is transported into the fetal compartment by placental influx transporters.
The versatility of the ER-REF approach goes beyond the ability to predict fetal K p,uu of drugs when multiple transporters are involved.This approach, combined with an m-f PBPK model, can predict fetal K p,uu of drugs, and the dynamic changes in their fetal plasma concentrations, at earlier gestational ages provided the placental abundance of the relevant transporter(s) at these earlier gestational ages is available.Such predictions are significant as it is ethically and impossible to obtain UV concentrations other than at the time of delivery.Even in the case of elective pregnancy termination, all one can obtain is fetal tissue drug concentration at a single timepoint, not sufficient to determine the K p,uu of the drug.Prediction of fetal K p,uu of a drug (and the dynamic changes in the drug's fetal plasma concentrations) at earlier gestational ages is necessary to inform potential fetal drug toxicity and efficacy to enable choice of drug or adjustment of drug dosing regimens to minimize fetal drug toxicity and maximize its efficacy.Therefore, we utilized the ER-REF approach to predict the fetal K p,uu of glyburide and imatinib (and the dynamic changes in their fetal plasma concentrations) at earlier gestational ages.The predicted fetal K p,uu of glyburide and imatinib at GW15 (0.24, 0.21) and GW25 (0.34, 0.29) were lower than at term (0.44, 0.37; Figure 4, Figure S2).This is because both CL int,efflux,placenta and CL int,PD,placenta decreased with gestational age due to the decrease in placental volume.In addition, CL int,PD,placenta decreased more than CL int,efflux,placenta because BCRP and P-gp transporter abundance per mL or gram of placenta is greater at earlier gestational ages than at term. 10 Our study has several limitations.First, we had only one glyburide pregnancy dataset (demonstrating high interindividual variability) to verify our glyburide m-f PBPK.Although this led us to use glyburide absorption parameters different from those in the non-pregnant population, this did not have any impact on our estimate of its K p,uu because K p,uu is determined only by maternal plasma concentration-time profile, transplacental drug clearance, and metabolic clearance in the placenta or in the fetal compartment.Second, we assumed that there was minimal metabolism of the three drugs in the placenta or in the fetal compartment.These assumptions are reasonable as the placenta has few CYP-metabolizing enzymes and the fetal liver has small (mostly CYP3A7) metabolic capacity, likely much less than the transplacental clearance of the drugs. 6Third, as to whether pregnancy induces changes in hepatic transporters (OATP1B1, 1B3, and 2B1) activity or abundance is unknown; one study has shown that pregnancy reduces hepatic OATP1B1 and 1B3 activity, 34 whereas another study with human hepatocytes, exposed to pregnancyrelated hormones, did not observe any significant changes in OATP transporters abundance. 35Fourth, because the fraction of the drug unbound in fetal plasma was not available to us, we used the Simcyp-predicted value.Errors in the prediction of this fraction will result in changes in estimates of the drug's fetal plasma concentrations.This is because our m-f PBPK model first estimated the K p value of the drug which was then converted into the K p,uu value using the predicted fraction unbound in both UV and MP.Fifth, the Simcyp pregnancy module does not allow K p,uu predictions earlier than at GW15 because fetal physiological data at < GW15 are currently not available.Finally, pooling UV/MP data from multiple maternal-fetal dyads results in large variability in the UV/MP data and therefore reduced confidence in the estimate of the in vivo K p,uu .Observed (mean) and predicted glyburide plasma C-T profile after 2.41 mg intravenous bolus dose to non-pregnant adults. 43(c) Dosenormalized observed (mean) and predicted glyburide plasma C-T profiles after oral doses of 0.875, 44 1.75, 45 and 3.5 mg 43 to non-pregnant adults.(d) Observed (mean) and predicted glyburide plasma C-T profile after oral dose of 1 mg twice daily administered for more than 1 week to the pregnant population (28-38 gestational weeks). 31The observed data (filled circles) fell within the 5 th and 95 th percentiles (dashed lines) of the predicted data (continuous black line).The predicted pharmacokinetic end points (AUC and C max ) also fell within 0.80-to 1.25-fold of the observed data (Table S4).Insets show the log-transformed C-T profiles.
In summary, because obtaining term UV/MP data of prescribed to pregnant people is challenging, our ER-REF approach, combined with an m-f PBPK model, provides a valuable tool to predict fetal drug exposure using only in vitro data, even when the drug is transported by multiple transporters.Moreover, this approach can be used to predict the drug K p,uu at gestational ages from 15 W to term.Parenthetically, we note here that our ER-REF approach has been successfully applied to predict the K p,uu of drugs in other tissues, such as the brain, where multiple transporters are expressed (e.g., BCRP and P-gp at the blood brain barrier). 30Our study offers a tool to predict fetal drug exposure at different stages of pregnancy to inform evaluation of potential fetal risks and benefits associated with maternal drug administration.for imatinib).Glyburide (1 mg, twice daily) was orally administered for more than 1 week between 28 and 38 weeks of gestation. 31Imatinib (400 mg daily) was administered between 35 and 41 weeks of gestation. 46Dashed lines represent the 5 th and 95 th percentiles of the predicted data.The fetal K p,uu value for glyburide and imatinib, estimated from the UV/MP data using PBPK M&S, were 0.44 and 0.37, respectively.(b, d) Simulated steady-state UV/MP profiles of glyburide (1 mg twice daily for 7 days) and imatinib (400 mg once daily for 15 days) at varying gestational ages.The predicted fetal K p,uu values for glyburide were 0.44, 0.34, and 0.24 at GWs 40, 25, and 15 and those for imatinib were 0.37, 0.29, and 0.21, respectively.The x-axis indicates the time since the last dose (GW40 solid line; GW25 dotted line, and GW15 dashed line).The UV/MP-time profiles of nelfinavir at different GW are provided in our previous publication 9 and showed the same trend.M&S, modeling and simulation; PBPK, physiologically-based pharmacokinetic; UV/MP, umbilical venous/maternal plasma.

Figure 1
Figure 1 Workflow for the prediction of fetal K p,uu by the ER-REF approach and its subsequent verification by comparison with the observed in vivo K p,uu estimated by m-f PBPK M&S.Left panel: The efflux ratio of the drug, determined using the in vitro Transwell system and efflux transporter-overexpressing cells (e.g., BCRP or P-gp), was scaled by the relative expression factor (REF) to predict the in vivo fetal K p,uu .The predicted fetal K p,uu was then verified by comparing with that observed in vivo as estimated by m-f PBPK M&S.Right panel: The PBPK model of the drug, developed and verified for the non-pregnant population, was extended to the pregnant population.First, the placental passive diffusion clearance of the drug was estimated from the in vitro drug and midazolam permeability data.Second, the in vivo placental efflux clearance (CL efflux ) was adjusted until the m-f PBPK model best predicted the observed UV/MP data.Finally, the in vivo K p,uu was estimated by the ratio of the fetal to maternal unbound plasma AUCs.ER-REF: efflux ratio-relative expression factor approach; P app(B→A) and P app(A→B) : apparent permeability of the drug in the indicated direction; CL int(B→A) and CL int(A→B) : apparent intrinsic clearances of the drug in the indicated direction; ER Transporter = ER (−Inh) − ER (+Inh) , where ER (−Inh) and ER (+Inh) is the efflux ratio of the drug mediated by either BCRP or P-gp in the absence and presence of their specific inhibitor; REF Transporter : relative expression factor of either BCRP or P-gp; CL PD,placenta : intrinsic placental passive diffusion clearance; CL efflux : intrinsic placental efflux clearance.m-f PBPK, maternal-fetal physiologically-based pharmacokinetic; UV/MP, umbilical venous/maternal plasma.

Figure 2
Figure2Transport-mediated efflux ratios of the test compounds in the Transwell system using (a) BCRP or (b) P-gp expressing cells.Imatinib (0.5 μM) and glyburide (1 μM) were found to be excellent substrates of both BCRP and P-gp as evidenced by their respective BCRP-and P-gp-mediated efflux ratios, ER (-Inh) , ER (+Inh) indicates the ER in the absence and presence of the respective transporter inhibitor.Likewise, nelfinavir (1 μM) was found to be an excellent P-gp substrate (B) but not a BCRP substrate (ER BCRP = 0.9; data not shown).Data represents individual experiments; lines represent means and standard deviations of the data from three independent experiments, each conducted in triplicate (see Table1for a detailed summary of the data).ER, efflux ratio.

Figure 3
Figure3Predicted and observed glyburide plasma concentration-time (C-T) profiles in the non-pregnant and pregnant populations.(a) Observed (mean) and predicted glyburide plasma C-T profile after intravenous infusion of 2 mg (over 1 hour) to non-pregnant adults.42(b) Observed (mean) and predicted glyburide plasma C-T profile after 2.41 mg intravenous bolus dose to non-pregnant adults.43(c) Dosenormalized observed (mean) and predicted glyburide plasma C-T profiles after oral doses of 0.875,44 1.75, 45 and 3.5 mg43 to non-pregnant adults.(d) Observed (mean) and predicted glyburide plasma C-T profile after oral dose of 1 mg twice daily administered for more than 1 week to the pregnant population (28-38 gestational weeks).31The observed data (filled circles) fell within the 5 th and 95 th percentiles (dashed lines) of the predicted data (continuous black line).The predicted pharmacokinetic end points (AUC and C max ) also fell within 0.80-to 1.25-fold of the observed data (TableS4).Insets show the log-transformed C-T profiles.

Figure 5
Figure 5 The mean fetal K p,uu of glyburide, imatinib, and nelfinavir, predicted by the ER-REF approach (black and white boxed bars), is in excellent agreement with the observed in vivo K p,uu as estimated by m-f PBPK M&S (hatched bars).The ratio of predicted K p,uu and observed K p,uu (right axis, and gray dotted bars) was within 0.5-2 fold range for all 3 drugs (our pre-defined verification criterion).The dotted line represents the two-fold range of predicted and observed fetal K p,uu ratio.ER-REF, efflux ratio-relative expression factor; m-f PBPK, maternalfetal physiologically-based pharmacokinetic; M&S, modeling and simulation; P/O, predicted/observed.

Figure 4
Figure 4 Predicted (solid line) and observed (filled circles; pooled from multiple maternal-fetal dyads) UV/MP values at steady-state, during a dosing interval.(a, c) UV/MP-time profiles of glyburide and imatinib at the time of birth (predictions are at GW38 for glyburide and GW39 for imatinib).Glyburide (1 mg, twice daily) was orally administered for more than 1 week between 28 and 38 weeks of gestation.31Imatinib (400 mg daily) was administered between 35 and 41 weeks of gestation.46Dashed lines represent the 5 th and 95 th percentiles of the predicted data.The fetal K p,uu value for glyburide and imatinib, estimated from the UV/MP data using PBPK M&S, were 0.44 and 0.37, respectively.(b, d) Simulated steady-state UV/MP profiles of glyburide (1 mg twice daily for 7 days) and imatinib (400 mg once daily for 15 days) at varying gestational ages.The predicted fetal K p,uu values for glyburide were 0.44, 0.34, and 0.24 at GWs 40, 25, and 15 and those for imatinib were 0.37, 0.29, and 0.21, respectively.The x-axis indicates the time since the last dose (GW40 solid line; GW25 dotted line, and GW15 dashed line).The UV/MP-time profiles of nelfinavir at different GW are provided in our previous publication9 and showed the same trend.M&S, modeling and simulation; PBPK, physiologically-based pharmacokinetic; UV/MP, umbilical venous/maternal plasma.

Table 1
ER, REF, f t and the predicted plus observed fetal KThe average of two values of transporter abundance was used to calculate the REF.,uu value (two-fold range around the mean value) estimated from the in vivo UV/MP data obtained at term.
a c Observed K p