Pair your accounts.

Export articles to Mendeley

Get article recommendations from ACS based on references in your Mendeley library.

Pair your accounts.

Export articles to Mendeley

Get article recommendations from ACS based on references in your Mendeley library.

You’ve supercharged your research process with ACS and Mendeley!

STEP 1:
Click to create an ACS ID

Please note: If you switch to a different device, you may be asked to login again with only your ACS ID.

Please note: If you switch to a different device, you may be asked to login again with only your ACS ID.

Please note: If you switch to a different device, you may be asked to login again with only your ACS ID.

MENDELEY PAIRING EXPIRED
Your Mendeley pairing has expired. Please reconnect
ACS Publications. Most Trusted. Most Cited. Most Read
My Activity
CONTENT TYPES

Development of an In Silico Prediction Model for P-glycoprotein Efflux Potential in Brain Capillary Endothelial Cells toward the Prediction of Brain Penetration

  • Reiko Watanabe*
    Reiko Watanabe
    Laboratory of Bioinformatics, Artificial Intelligence Center for Health and Biomedical Research, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka 567-0085, Japan
    *Email: [email protected]
  • Tsuyoshi Esaki
    Tsuyoshi Esaki
    The Center for Data Science Education and Research, Shiga University, Hikone, Shiga 522-8522, Japan
  • Rikiya Ohashi*
    Rikiya Ohashi
    Laboratory of Bioinformatics, Artificial Intelligence Center for Health and Biomedical Research, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka 567-0085, Japan
    Discovery Technology Laboratories, Mitsubishi Tanabe Pharma Corporation, Muraoka-Higashi, Fujisawa, Kanagawa 251-8555, Japan
    *Email: [email protected]
  • Masataka Kuroda
    Masataka Kuroda
    Laboratory of Bioinformatics, Artificial Intelligence Center for Health and Biomedical Research, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka 567-0085, Japan
    Discovery Technology Laboratories, Mitsubishi Tanabe Pharma Corporation, Muraoka-Higashi, Fujisawa, Kanagawa 251-8555, Japan
  • Hitoshi Kawashima
    Hitoshi Kawashima
    Laboratory of Bioinformatics, Artificial Intelligence Center for Health and Biomedical Research, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka 567-0085, Japan
  • Hiroshi Komura
    Hiroshi Komura
    URA Center, Osaka City University, Osaka 545-0051, Japan
  • Yayoi Natsume-Kitatani
    Yayoi Natsume-Kitatani
    Laboratory of Bioinformatics, Artificial Intelligence Center for Health and Biomedical Research, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka 567-0085, Japan
  • , and 
  • Kenji Mizuguchi
    Kenji Mizuguchi
    Laboratory of Bioinformatics, Artificial Intelligence Center for Health and Biomedical Research, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka 567-0085, Japan
    Laboratory of In-Silico Drug Design, Center of Drug Design Research, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka 567-0085, Japan
Cite this: J. Med. Chem. 2021, 64, 5, 2725–2738
Publication Date (Web):February 23, 2021
https://doi.org/10.1021/acs.jmedchem.0c02011

Copyright © 2021 The Authors. Published by American Chemical Society. This publication is licensed under

CC-BY-NC-ND 4.0.
  • Open Access

Article Views

7566

Altmetric

-

Citations

LEARN ABOUT THESE METRICS
PDF (6 MB)
Supporting Info (2)»

Abstract

Developing in silico models to predict the brain penetration of drugs remains a challenge owing to the intricate involvement of multiple transport systems in the blood brain barrier, and the necessity to consider a combination of multiple pharmacokinetic parameters. P-glycoprotein (P-gp) is one of the most important transporters affecting the brain penetration of drugs. Here, we developed an in silico prediction model for P-gp efflux potential in brain capillary endothelial cells (BCEC). Using the representative values of P-gp net efflux ratio in BCEC, we proposed a novel prediction system for brain-to-plasma concentration ratio (Kp,brain) and unbound brain-to-plasma concentration ratio (Kp,uu,brain) of P-gp substrates. We validated the proposed prediction system using newly acquired experimental brain penetration data of 28 P-gp substrates. Our system improved the predictive accuracy of brain penetration of drugs using only chemical structure information compared with that of previous studies.

This publication is licensed under

CC-BY-NC-ND 4.0.
  • cc licence
  • by licence
  • nc licence
  • nd licence

Introduction

ARTICLE SECTIONS
Jump To

In the drug discovery of central nervous system (CNS), the design of pharmacologically active drug remains challenging for medicinal chemists. (1) To effectively act in the CNS, a drug must pass through the brain capillary endothelial cells (BCEC) which function as the blood–brain barrier (BBB). In contrast, it is important to limit the brain penetration of non-CNS drugs to avoid unexpected CNS adverse effects. The brain-to-plasma concentration ratio (Kp,brain) and the unbound brain-to-plasma concentration ratio (Kp,uu,brain) are generally used as in vivo pharmacokinetic (PK) parameters to assess the brain penetration of drugs. As unbound drugs are involved in the exertion of pharmacological and toxicological effects in CNS tissues, it is important to predict the in vivo PK parameters for the brain penetration of the drugs by medicinal chemists in the early stages of drug discovery, especially in the drug design process before chemical synthesis, to increase the success rate of drug development.
To predict Kp,uu,brain of compounds, it is necessary to consider the involvement of various transporters. P-glycoprotein (P-gp) is a well-characterized efflux transporter at the BBB; it strictly regulates the membrane permeability of substances between the brain tissue and blood by BCEC, pericytes, and astrocytes, as it maintains a low concentration of unbound molecules in the brain relative to that in plasma. (2) In addition, Kp,uu,brain is a parameter derived from Kp,brain. To determine Kp,uu,brain from Kp,brain, PK parameters such as the fraction unbound in brain homogenate (fu,brain) and the fraction unbound in plasma (fu,p) are required. Previously, we reported individual prediction models for fu,brain, (3)fu,p, (4) and P-gp net efflux ratio (NER). (5) Furthermore, Rodgers et al. proposed a formula to calculate Kp,brain using a differential phospholipid method, which mainly considers the physicochemical properties of compounds but does not take into account the involvement of transporters. (6,7) It is necessary to integrate the above-described prediction models to predict brain penetration, which involves a complex mechanism; there are several issues to resolve during the construction of models and the validation of results.
In vivo Kp,brain values in mice correlate linearly with the in vitro P-gp NER. In addition, Summerfield et al. reported a significant improvement in the correlation of Kp,brain values between wild-type (WT) and P-gp knock out (KO) mice when the Kp,brain value in the P-gp KO mice was corrected using in vitro P-gp NER. However, their calculations indicated that this correction was insufficient to allow extrapolation from WT mice to P-gp KO mice. (8) A possible explanation for the differences between the in vivo and in vitro results is that P-gp protein expression levels differ between P-gp-overexpressing cell lines and BCEC. (9) A previous study has reported differences in the expression levels of several transporters, not only between species but also between in vivo BCEC and in vitro cell lines, indicating the difficulty in the direct extrapolation of findings between in vivo and in vitro. (10)
To overcome these challenges, a strategy to determine transporter protein expression of in vitro and in vivo experimental compounds using the proteomics technique was proposed and its application in reconstructing in vivo P-gp function in both mice and human BCEC was demonstrated. (11−15) Their findings show that the reconstructed P-gp NER is similar to the in vivo P-gp NER at the BBB. Their study clarified the functional differences in BBB transport between in vitro and in vivo models and between human and animal models. However, it has not been verified whether the in vivo P-gp function can be reconstructed using the predicted value.
Unlike the predictive models for simple in vitro PK parameters such as fu,p and fu,brain, Kp,brain or Kp,uu,brain cannot be routinely verified in humans due to ethical constraints, a limited number of post-mortem determinations, and limited access to imaging analysis systems such as positron emission tomography. Thus, to develop CNS drugs, the Kp,brain or Kp,uu,brain values obtained using in vivo PK experiments in rodents are used as the substitute for experiments in humans, with the expectation that they would show similarities between rodents and humans. However, such expectations ignore potential problems associated with the assumptions that there is no species difference in the transport potential of substrates and that the Kp,brain and Kp,uu,brain values of rodents are similar to those of humans. Handling species differences remains an important issue.
In the present study, we developed prediction models for P-gp NER as well as for fu,p and fu,brain, and obtained the representative values of P-gp NER in human and rat BCEC based on the classification results of a human P-gp NER model. Herein, we propose a novel prediction system for Kp,brain and Kp,uu,brain using predicted fu,p and fu,brain and the representative value of in vivo P-gp NER in rat BCEC instead of the numerical value of P-gp NER in rat BCEC. Using this system, we demonstrated its potential to predict Kp,brain and Kp,uu,brain solely via chemical structure information and showed that the prediction accuracy of Kp,uu,brain was superior to that of the conventionally calculated Kp,uu,brain. The proposed prediction system is open to the public, and this method can have a beneficial effect on CNS drug design before chemical synthesis during the early stages of drug discovery.

Results and Discussion

ARTICLE SECTIONS
Jump To

Study Design Settings and an Overview of Analysis for Model Construction

The study design and an overview of analysis for model construction are presented in Scheme 1. Initially, three prediction models for in vitro P-gp NER, fu,p, and fu,brain were generated using the datasets obtained from in-house experiments and publicly available data (Model_pgpNER, Model_fup, and Model_fubrain). Next, we selected 50 diverse compounds from approved drugs that have been reported to be non-P-gp or P-gp substrates, and non-CNS or CNS drugs (Comp_50). The three constructed models were validated using fu,brain, fu,p and P-gp NER values (Ex-testset_fubrain, Ex-testset_fup, and Ex-testset_pgpNER), newly obtained from in-house experiments, for the compounds in Comp_50 (Scheme 1A). Finally, we proposed a prediction method for Kp,uu,brain considering the efflux transport by P-gp; from the Kp,brain predicted using Rodgers’ equation, Kp,uu,brain was calculated based on predicted P-gp NER, fu,p, and fu,brain. We experimentally obtained Kp,brain and Kp,uu,brain for 28 P-gp substrates in Comp_50 and validated the prediction results (Scheme 1B). Therefore, predicting the Kp,uu,brain value from the structural information is not accomplished in one-step, but it is a complex process with multiple steps using the four PK parameters.

Scheme 1

Scheme 1. (A) Overview for Model Building; (B) Validation of the Correction Method for the Prediction of Kp,uu,brain from Chemical Structure Information

Generation of Training Datasets for fu,brain, fu,p, and P-gp NER Prediction

Training datasets for fu,brain, fu,p, and P-gp NER prediction, containing 505, 462, and 397 compounds, respectively, were constructed from in-house experiments, publicly available data in ChEMBL, and previous study findings, as described in Supporting Information 2. Data were carefully curated to acquire the highest quality possible, as described previously. (4,16) The final number of compounds in datasets, range of parameters, and data sources are shown in Table 1; the datasets for, fu,brain, fu,p, and P-gp NER prediction were defined as Dataset_pgpNER, Dataset_fubrain, and Dataset_fup, respectively; final data are presented in Supporting Information 1.
Table 1. Basic Dataset Information for P-gp NER, fu,brain, and fu,p Predictions
parameterdatasetspeciesno. of compoundsrangesource
fu,brainDataset_fubrainrat5050.001–1in-house data Esaki et al. (3)
fu,pDataset_fuprat4620.001–1in-house data ChEMBL
P-gp NERDataset_pgpNERhuman3970.29–320in-house data

Construction of Three Prediction Models

We aimed to construct regression models with Dataset_pgpNER, Dataset_fubrain, and Dataset_fup; details of model construction are shown in Supporting Supporting Information 2, and an overview of the prediction model construction process is shown in Scheme S1 (Supporting Information 3). Initially, the conditions for prediction model construction were examined and then the model was finalized using all the data in the best condition during the initial step. All training steps were performed with 5-fold cross validation. To eliminate the influence of the split bias on the result, 10 patterns of dataset were created by random splitting, and the averages of mean square error (MSE) on the common test sets, along with their standard deviation, were compared as an evaluation metric in the regression among four types of algorithms. The best condition selected following condition examination was set with gradient boosting (GB) as a machine-learning algorithm with feature selection using Boruta, in the logarithmic scale. GB is one of the ensemble-learning methods using multiple weak prediction models. At each stage of parameter optimization, the parameters are updated using the gradient descent method to bring the predictions closer to the observed values. Boruta is an all-relevant feature selection wrapper algorithm designed to automatically perform feature selection on a dataset using random forest. The average of MSE and coefficient of determination (r2) in the best condition for fu,brain and fu,p in the validation and test sets in the initial step of condition examination are shown in Table 2, and the statistical results of individual models are shown in Table S1 in Supporting Information 3. The finalized models, using all the data in the dataset with the best condition showing the lowest MSE during validation, were selected as the best final prediction model of fu,brain, fu,p, and P-gp NER. For fu,brain and fu,p, the finalized models were defined as Model_fubrain and Model_fup. The regression model for P-gp NER did not show sufficient accuracy as shown in Tables 2 and S1; we also tried to construct a 3-class classification model to distinguish the transport potential of P-gp substrate as a substitute for the regression models based on a previously reported method. (5) The thresholds in Dataset_pgpNER were set to 1.4 and 9.8, as described in the Experimental Section. Splitting the dataset into three classes according to the transport potential of P-gp substrate (low, middle, and high) based on thresholds resulted in 187, 153, and 57 compounds, respectively. Quadratic weighted κ (κ) was used as the evaluation metric (Table 2), and the finalized model was defined as Model_pgpNER.
Table 2. Statistical Results of Prediction Models in the Condition Examination
parametertypevalidation/testMSEr2kappa
fu,brainreg.validation1.11 ± 0.290.65 ± 0.08 
  test1.48 ± 0.180.58 ± 0.06 
fu,preg.validation1.42 ± 0.270.58 ± 0.06 
  test1.70 ± 0.510.51 ± 0.11 
P-gp NERreg.validation0.80 ± 0.220.34 ± 0.19 
  test0.95 ± 0.220.26 ± 0.14 
 classvalidation  0.60 ± 0.12
  test  0.44 ± 0.17

Validation of the Finalized Prediction Models

The validation data for fu,brain, fu,p, and P-gp NER were experimentally obtained with 46, 45, and 50 compounds, respectively, from Comp_50 and defined as Ex-testset_fubrain, Ex-testset_fup, and Ex-testset_pgpNER, respectively. The finalized prediction models (Model_fubrain, Model_fup, and Model_PgpNER) were validated using these external datasets. The statistical results of the external test sets are shown in Table 3, and the plots are shown in Figure S1. The prediction accuracy of Ex_testset_fubrain was evaluated using Model_fubrain and our previously reported model (Model_fubrain_prev). (3) Out of the 46 compounds, 10 were excluded during evaluation using Model_fubrain_prev as they were included in the training set of Model_fubrain_prev. The MSE of Model_fubrain was lower than that of Model_fubrain_prev (1.10 vs 1.48), and the percentage of samples within 3-fold error increased from 46.7 to 71.7%, indicating that the versatility of our model for fu,brain prediction had improved. Forty-five compounds in Ex_testset_fup were then predicted using Model_fup and %rat_fup model in the ADMET Predictor (Simulations Plus, Lancaster, CA, U.S.A.), which is one of the most widely used commercial software. The MSEs were 1.66 and 2.07, and sample percentages within 3-fold error were 66.7 and 57.8%, for the Model_fup and %rat_fup, respectively. Model_fup presented a slightly better result than %rat_fup, and the prediction accuracy, especially in the lower range of fu,p, was higher in Model_fup, as shown in Figure S1B. When the degree of plasma protein binding is high, small differences in protein binding can have large effects on fu,p, and thus, the drug efficacy can change dramatically. Therefore, a model with a good prediction accuracy in a low range of fu,p may be more effective during drug development.
Table 3. Validation Results of the Prediction Models in the External Dataset of fu,brain and fu,p
parameterdatamodelMSE% within 3-fold error
fu,brainEx_testset_fubrainModel_fubrain1.1071.7
 Ex_testset_fubrainpreviously published model (3)1.4846.7
fu,pEx_testset_fupModel_fup1.6666.7
 Ex_testset_fup%rat_fup2.0757.8
The confusion matrix of Model_pgpNER with Ex_testset_pgpNER is shown in Table 4. The κ value for Ex_testset_pgpNER was 0.45. In the three-class classification model, misclassification across two categories should be avoided the most; specifically, a high-potential compound being predicted as a low-potential compound and vice versa, these judgment error rates were 7.7 and 13.3%, respectively, with the ex_testset_pgpNER.
Table 4. Confusion Matrix of Model_pgpNERa
   predictionmisprediction rate (pred. to exp.)
Datasetκexp.LMHL to H (%)H to L (%)L to M (%)H to M (%)M to H (%)M to L (%)
Ex_testset_pgpNER0.45L10217.713.315.446.718.231.2
  M7114
  H276
a

L, low; M, middle; and H, high.

Reconstruction from In Vitro P-gp NER in P-gp-Overexpressing Cells to P-gp NER in BCEC

We translated the in vitro P-gp NER in P-gp-overexpressing cells using in-house experiments data into P-gp NER in BCEC based on the translation method proposed by Uchida et al. (9) The details are shown in “Estimation of the P-gp NER in BCEC” of the Experimental Section. The experimentally obtained in vitro human and rat P-gp NER values with 48 compounds were used in this analysis. Equation 1 was used to translate in vitro human P-gp NER into P-gp NER in human BCEC and eq 2 was used to translate in vitro rat P-gp NER into P-gp NER in rat BCEC. The plot illustrating the relationship between in vivo P-gp NER in human and rat BCEC is shown in Figure 1. Compounds that showed lower P-gp NER in humans tend to show lower P-gp NER in rats, and compounds that are good substrates in humans had up to 10 times higher NER in rats. In addition, the experimental P-gp NER values in human and rat BCEC were significantly positively correlated (R = 0.72), which is in agreement with the finding of previous studies. (15,17) The regression equation [y = (x + 0.0119)/0.6533] was obtained and set as eq 3.

Figure 1

Figure 1. Correlation of the reconstructed P-gp NER between rat and human BCEC. The regression equation (eq 3) is shown in the top left corner; the number of compounds (n) and regression coefficient (R) are also shown. The solid line represents regression.

Setting of Representative Values of P-gp NER

To supplement the drawbacks that the numerical value cannot be predicted in the three-class classification model of P-gp NER, we used representative values of P-gp NER in rat BCEC for the prediction of brain penetration. The details are shown in Scheme 2. Initially, the median values of experimental in vitro human P-gp NER in the class of P-gp-mediated efflux potential (2.6 and 18.9 in middle and high classes) were converted into the representative value of in vivo P-gp NER in human BCEC using eq 1 (1.6 and 8.2 in middle and high classes). The representative values of in vivo P-gp NER in humans were then converted to those in rats using eq 3. Finally, the representative values of in vivo P-gp NER in rat BCEC were set to 2.2 and 26.3 for the middle and high classes, respectively. Furthermore, we used 1.0 as the representative value for the low class, as it is categorized as a P-gp non-substrate.

Scheme 2

Scheme 2. Calculation Process of Representative Values in P-gp NER through the Thresholds of the P-gp NER Prediction Model
To the best of our knowledge, we have demonstrated, for the first time, that the representative value of P-gp NER in human and rat BCEC from only compound structure information, can be predicted by unifying the predicted transport potential of the P-gp substrate using Model_pgpNER. We also proposed a method to translate from in vitro P-gp NER in P-gp overexpressing cells to P-gp NER in BCEC. In a previous study, the MDCK-MDR1 cell line from the National Institutes of Health showed higher P-gp expression than that obtained from the Netherlands Cancer Institute. (18) This illustrates the need for validating in vitro systems before applying it to CNS drug discovery. (19) Thus, the advantage of predicting the P-gp NER in BCEC is that it can identify in vivo P-gp NER directly without considering the differences in the P-gp expression levels among cell lines.

Relationship between Kp,uu,brain and Physicochemical/PK Properties

Experimental Kp,brain and Kp,uu,brain values in WT rats were obtained from experimentally acquired plasma and brain concentration data using eq 4 with 83 compounds, and the relationship between Kp,uu,brain and physicochemical properties as well as between PK parameters was analyzed. fu,p, fu,brain, P-gp NER, and nine physicochemical properties, all of which are generally considered to be important parameters for synthetic expansion, were compared in three parts: Kp,uu,brain < 0.1, 0.5 ≤ Kp,uu,brain, and in-between. The box plots are shown in Figure 2; the molecular weight (MW), topological polar surface area (TopoPSA), number of hydrogen bond acceptors (nHBAcc), number of hydrogen bond donors (nHBDon), log D pH 7.4, acidic pKa (apKa), and P-gp NER showed significant differences between compounds in Kp,uu,brain < 0.1 and 0.5 ≤ Kp,uu,brain. The median values for the group 0.5 ≤ Kp,uu,brain that has a tendency to achieve better brain penetration were MW = 299.1 Da, TopoPSA = 40.5 Å2, nHBAcc = 3, nHBDDon = 1, log D = 2.2, apKa = 13.8, and P-gp NER = 1.10.

Figure 2

Figure 2. Comparison of 12 physicochemical and PK properties between the degrees of Kp,uu,brain. The compounds that were 0.1 > Kp,uu,brain (n = 36), 0.1 ≤ Kp,uu,brain < 0.5 (n = 16), and 0.5 ≤ Kp,uu,brain (n = 31) are shown in blue, orange, and green boxes, respectively. *, p < 0.01; **, p < 0.05; cross mark, mean; the numbers above and below the bars indicate the median of the parameters that show significant differences.

Correction of Kp,brain Value in P-gp KO Rats Using the P-gp NER in Rat BCEC

To effectively validate whether the prediction method for Kp,brain and Kp,uu,brain that takes into account the influence of P-gp-mediated efflux transport using P-gp NER functions, we obtained experimental Kp,brain in P-gp KO rats for 46 compounds, which overlap with the 83 compounds’ Kp,brain value that were previously acquired in WT rats. Kp,brain values of P-gp KO rats were then corrected with three different types of P-gp NERs: (20) (1) numerical P-gp NER in rat BCEC derived from eq 5 (Figure 3B), (2) representative values of rat BCEC according to the classification of experimental P-gp NER using eq 6 (Figure 3C), and (3) representative values of rat BCEC according to the predicted class by Model_pgpNER (Figure 3D).

Figure 3

Figure 3. Plot showing the experimental Kp,brain values in WT and (A) experimental Kp,brain value in P-gp KO rat without correction; (B) corrected Kp,brain using the experimental P-gp NER numerical values in rat BCEC reconstructed from in vitro rat P-gp NER; (C) corrected Kp,brain using the representative value of P-gp NER in rat BCEC based on experimental P-gp NER numerical values; (D) corrected Kp,brain using the representative value of P-gp NER in rat BCEC based on the predicted class with Model_pgpNER. P-gp substrates and non-substrates are indicated in gray and white circles, respectively. The MSE is shown in top left corners. The percentage of samples with a 3- or 5-fold error is shown at the bottom. Straight, dashed, and dotted lines indicate the lines of unity, 3-fold and 5-fold errors, respectively (n = 46).

The Kp,brain values in WT and P-gp KO rats are plotted in Figure 3A, with higher Kp,brain values observed in P-gp KO rats, demonstrating the influence of P-gp-mediated efflux transport. The samples were classified as either P-gp substrate (P-gp NER ≥ 1.4) or non-substrate (P-gp NER < 1.4) according to the results of our in vitro rat P-gp NER study, which showed that 33 and 13 of the compounds were P-gp substrates and non-substrates, respectively. As shown in Figure 3B, the Kp,brain value in the P-gp KO rat was corrected using P-gp NER numerical values in rat BCEC with eq 5. The MSE decreased from 0.79 to 0.37, and the percentage of P-gp substrates that fell within a 5-fold error increased from 52.2 to 78.3%. These results indicate that the P-gp NER in rat BCEC reconstructed using in vitro P-gp NER could be used quantitatively to account for the brain penetration observed in vivo relative to that in P-gp KO models, as reported previously. (9)
As shown in Figure 3C, the Kp,brain correlated with the representative values of rat BCEC according to the classification of experimental P-gp NER using eq 6. The percentage of compounds that fell within a 5-fold error increased from 52.2 to 76.1%, accompanied by a decrease in the MSE from 0.79 to 0.40 in all compounds. The effectiveness of correction was equivalent to that observed using experimental values (0.37 vs 0.40 in MSE and 78.3 vs 76.1% in compounds that fell within a 5-fold error, in the experimental values and the representative values, respectively). These results indicate that the Kp,brain value was well corrected by the representative value of P-gp NER.
We then corrected the Kp,brain value in P-gp KO rats using the representative values of rat BCEC according to the predicted class using Model_pgpNER (Figure 3D). The percentage of compound that fell within a 5-fold error was 69.6%, the MSE decreased to 0.46, and the effectiveness of correction was slightly lower than that achieved when the experimental values were used. However, approximately 70% of samples fall within a 5-fold error. These results indicate that the Kp,brain value could be corrected using the representative values based on the predicted classification of P-gp NER using Model_pgpNER.

Calculation of Kp,uu,brain Value Using Kp,brain, fu,p, and fu,brain

We calculated Kp,uu,brain from the corrected Kp,brain in P-gp KO rats, as shown in Figure 3C,D, using the experimental or predicted values of fu,p and fu,brain of 42 compounds using eq 7. The predicted fu,p and fu,brain values were derived from the prediction results of Model_fup with Ex_testset_fup and Model_fubrain with Ex_testset_fubrain. The prediction of Kp,uu,brain was evaluated within a 5- or 10-fold variation, as 5-fold experimental variation has been previously reported in the dataset of Kp,uu,brain resulting from combining three end points to derive Kp,uu,brain; Kp,brain, fu,p, and fu,brain. (21) Thus, we cannot definitively conclude that there is a significant difference with a 5-fold variation. (22) Obtaining experimental data for fu,p, fu,brain, and P-gp NER for compounds that are highly adsorbed onto plasma proteins or lipid membrane is technically difficult and might cause a huge experimental error. As these experimental errors affect Kp,uu,brain prediction, we considered the standard range to be 5-fold and set 10-fold as the acceptable error range for the evaluation of Kp,uu,brain prediction.
Figure 4 shows the plots of Kp,uu,brain value corrected using the representative values based on experimental and predicted P-gp NER in rat BCEC, and experimental and predicted fu,p and fu,brain data. With the experimental value, MSE and the percentage of compounds that fell within 5-fold and 10-fold errors were 0.31, 76.2, and 92.9%, respectively, indicating that Kp,uu,brain can be predicted using the correction of Kp,brain by P-gp NER in rat BCEC. The predicted data revealed that the MSE and the percentage of compounds fell within 5- and 10-fold errors were 0.63, 66.7, and 73.8%, respectively. We concluded that rat Kp,uu,brain could be calculated using our proposed method using representative P-gp NER values in rat BECE. Moreover, the experimental values of P-gp NER, fu,brain, and fu,p can be replaced by the predicted results of Model_pgpNER, Model_fubrain, and Model_fup, respectively.

Figure 4

Figure 4. (A) Plot of Kp,uu,brain in WT rats and p-gp KO rats calculated using experimental fu,p and fu,brain values. (B) Plot of Kp,uu,brain in WT rats and p-gp KO rats calculated using predicted fu,p and fu,brain values. In plots (A,B), P-gp substrates and non-substrates are shown in gray and white circles, respectively. The MSE and number of compounds (n) are shown in the top left and top right corners, respectively. Straight, dashed, and dotted lines indicate the line of unity, 5-fold, and 10-fold errors, respectively.

Validation of the Correction Method for the Prediction of Kp,uu,brain Using the Calculated Kp,brain Value with Rodgers’ Equation

For the practical application of the proposed correction method using P-gp NER in BCEC, it is necessary to predict Kp,brain according to simple measurements or using only structural information. The result revealed that Kp,brain in P-gp KO rats cannot be practically used. Therefore, Kp,brain was calculated using a differential phospholipid method proposed by Rodgers et al., (6,7) and the proposed correction method that we evaluated in P-gp KO rats was applied to Kp,brain calculated using Rodgers’ equation. As the prediction values of fu,p were required for Kp,brain calculation with Rodgers’ equation, the difference between the experimental values and the calculated Kp,brain using the predicted values was analyzed before validating the correction method (Figure S2). We concluded that the observed value can be replaced with predicted values.
Finally, our proposed correction method, using the representative value of P-gp NER in BCEC based on the predicted class with Model_pgpNER, Model_fubrain, and Model_fup, was applied to Kp,brain calculated using Rodgers’ equation. The results were validated using experimentally acquired Kp,brain and Kp,uu,brain values of 28 P-gp substrates in Comp_50, as shown in Scheme 1. The MSE decreased from 2.12 to 1.21 for Kp,brain, and from 2.34 to 1.55 for Kp,uu,brain following correction using the predicted P-gp NER in BCEC. The percentage of samples within a 10-fold error increased from 42.9 to 64.3% for Kp,brain, and from 50.0 to 64.3 for Kp,uu,brain (Figure 5A,B). These results indicate that the Kp,brain and Kp,uu,brain values were predicted only from structural information with better accuracy than the conventional method. Ten compounds, namely, atenolol, amprenavir, bepotastine, clarithromycin, dasatinib, digoxin, domperidone, imatinib, labetalol, and zolmitriptan, had shown substantial error of more than 10-fold (Table 5), and this could be attributed to the accumulation of gaps in the predicted values. The difference between the Kp,brain value using Rodgers’ equation and the experimental Kp,brain in P-gp KO rats appeared to be one of the most influenced factors, because compounds with a larger fold error in calculated Kp,brain presented a larger difference in Kp,uu,brain prediction. The Kp,brain value of atenolol, bepotastine, dasatinib, domperidone, imatinib, saquinavir, and zolmitriptan differed by 14.9–539 times (Table 5). Atenolol, apixaban, dasatinib, imatinib, and zolmitriptan are known as substrates of not only P-gp but also other transporters such as breast cancer resistance protein (BCRP) and organic anion-transporting polypeptide (OATP). (23−27) Although Rodgers’ equation is currently the best method to predict tissue-to-plasma partition coefficient, it does not consider the transporters’ involvement and prediction accuracy in tissues with a larger contribution of transporters, such as the brain, kidney, and liver, insufficient, especially when the compounds were substrates for multiple transporters. The results also illustrate the difficulty in predicting compounds such as digoxin. Digoxin binds slowly and specifically to its target molecule Na+/K+ ATPase, which is ubiquitously expressed in the body. Therefore, digoxin is extensively distributed in tissues and equilibrates slowly. (28,29) In addition, the prediction error in P-gp NER classification also influenced the accuracy of Kp,uu,brain prediction. 8 out of 10 compounds were misclassified in Model_pgpNER, suggesting that further improvement in the prediction accuracy of Model_pgpNER is required.

Figure 5

Figure 5. Plot of Kp,brain and Kp,uu,brain in the P-gp substrate before and after correlation. Experimental Kp,brain in WT rats vs Kp,brain calculated using Rodgers’ formula (A), vs Kp,brain corrected using the representative value based on the predicted P-gp NER (B). Experimental Kp,uu,brain in WT rats vs Kp,uu,brain calculated from Kp,brain without correction (C) and vs corrected Kp,uu,brain (D) with the predicted fu,p and fu,brain values. P-gp substrates (n = 28) are plotted. Straight, long-dashed, and dotted lines indicate the line of unity, 5-fold and 10-fold errors, respectively.

Table 5. Difference between the Experimental and Predicted Values of P-gp NER, fu,p, fu,brain, Kp,brain, and Kp,uu,brain in 28 P-gp Substrates for Validation
a

10 P-gp substrates whose fold error in Kp,uu,brain was more than 10-fold; exp., experimental; pred., predicted; calc., calculated; and fold error, the fold error between the experimental and the predicted/calculated values.

Using the predicted representative value of P-gp NER in BCEC and the predicted fu,brain and fu,p values, we have demonstrated the prediction system for Kp,brain and Kp,uu,brain from chemical information alone. Although the method proposed in this study is not applicable to all compounds, given that several transporters are involved in brain penetration of drugs, we showed that at least for P-gp substrates, its predictive accuracy for Kp,brain and Kp,uu,brain by correction using predicted representative P-gp NER increased compared with that for prediction of Kp,brain using Rodgers’ equation. Other studies have also reported species-specific differences in substrate recognition and transportability, as well as differences in absolute protein expression levels of BCRP and OATP in human and animal brain capillaries. (11−15) This method can be applied to other transporters such as BCRP and OATP when predicting Kp,brain and Kp,uu,brain for their substrates, making it possible to construct a prediction model that is intended for more compounds.

Possibility of Practical Application for the Prediction of Brain Penetration of Drugs in Humans

As Kp,brain and Kp,uu,brain values in rodents can be acquired with comparative ease from the drug concentrations in the brain and plasma, fu,p and fu,brain, previous studies have reported several predictive models for Kp,brain and Kp,uu,brain values in rodents using machine-learning techniques. (30−32) However, it is not the best option to directly apply the prediction results in animals to humans; predictive models targeting transporter substrates must consider species-specific differences in transporter protein expression levels and their functions at the BBB.
The rat Kp,uu,brain value predicted using machine learning is not directly applicable to humans, whereas the human Kp,uu,brain value can be predicted using our proposed method by replacing the parameters used in rats with those used in humans. We believe that our proposed method can be expanded to include humans; however, the lack of human validation data remains an obstacle. Drug concentration in the cerebrospinal fluid (CSF) is used as a surrogate for cerebral interstitial fluid, as CSF is the only accessible sample in human CNS tissues. Friden et al. (30) compared human and rat unbound CSF with the plasma concentration ratio for 32 structurally diverse compounds and demonstrated an association with the brain exposure of species. However, as it is difficult to collect the human CSF data under unified conditions such as health status, time points of CSF sampling, and different CSF sources, it is difficult to distinguish whether the observed difference reflects a true species difference or observational bias. Moreover, several-fold differences between the concentration in the cerebral interstitial fluid and CSF have been reported. (33,34) The lack of reliable concentration data for human cerebral interstitial fluid, which can be used to verify the predicted results, is a major obstacle for future verification of the in silico prediction results.

Conclusions

ARTICLE SECTIONS
Jump To

In this study, we initially constructed the regression models of fu,p and fu,brain, and the 3-class classification model to distinguish the transport potential of the P-gp substrate accurately based on structural information; then, the representative values of P-gp NER in human and rat BCEC were set. Second, we proposed a new prediction system via correction using the representative values of P-gp NER to predict Kp,brain and Kp,uu,brain in P-gp KO rats, and the system was successfully validated. Finally, we demonstrated that our prediction system can be applied to the calculated Kp,brain, and after applying our prediction method, the predictive accuracy of Kp,brain and Kp,uu,brain in P-gp substrates in rats increased. A new web resource (http://adme.nibiohn.go.jp/brain-penetration) has been established to access the online system for the prediction of brain penetration, as described in this study. We believe that our proposed method can be potentially used in CNS drug discovery to estimate the pharmacological effects and optimize the drug candidates at an early stage.

Experimental Section

ARTICLE SECTIONS
Jump To

Materials

Test compounds were of analytical grade (purity >95%) and purchased from Tokyo Chemical Industry (Tokyo, Japan), MedChemExpress (Monmouth Junction, NJ, U.S.A.), AdooQ Bioscience (Irvine, CA, U.S.A.), Sigma-Aldrich (St. Louis, MO, U.S.A.), ChemScene (Monmouth Junction, NJ, U.S.A.), Fujifilm Wako Pure Chemical Corporation (Osaka, Japan), and Toronto Research Chemicals (North York, Canada). A total of 447 compounds were dissolved at a concentration of 10 mM in 100% DMSO and 1 mg/mL (0.5%, w/v) methyl cellulose for in vitro and in vivo experiments, respectively. The solutions were stored at 4 °C and used within 2 weeks following preparation.

Acquisition of Experimental Data

P-gp efflux transport in rat and human P-gp-overexpressing cells was measured to evaluate the P-gp substrate potential. LLC-PK1 and LLC-GA5-CoL150 cells were purchased from RIKEN BioResource Research Center (Cell IDs: RCB0558 and RCB0871; Tsukuba, Japan). LLC-GA5-CoL150 cells had been stably transfected with the human MDR1 gene to overexpress P-gp. (35,36) Rat Mdr1a-overexpressing LLC-PK1 cells were prepared by Sekisui Medical Co. Ltd. (Tokyo, Japan). The P-gp expression level in LLC-PK1, LLC-GA5-CoL150, and rat Mdr1a-overexpressing LLC-PK1 cells was determined by Sekisui Medical Co. Ltd. (Tokyo, Japan) as previously described. (7) Plasma and brain homogenate binding were determined using equilibrium dialysis by Sekisui Medical Co. Ltd. (Tokyo, Japan). The measurement of drug concentration in the brain and plasma in WT and P-pg KO rats was performed in Shin Nippon Biomedical Laboratories (SNBL; Tokyo, Japan), and studies were performed in accordance with all institutional guidelines, and the animal study protocol was approved by the Animal Care and Use Committee SNBL (approval numbers: IACUC819-010, IACUC819-011, IACUC819-015, and IACUC819-016). The details of experimental procedures in each experiment are shown in Supporting Information 4.

Threshold of Classification Model for P-gp NER

Thresholds of P-gp NER were established, and the process is shown in Scheme 2. Based on the method proposed by Ohashi et al., (5) propranolol (a non-P-gp substrate; geometric mean NER = 1.04) and prazosin (a weak P-gp substrate; geometric mean NER = 3.20) were included in each experiment as a control compound. The first threshold was determined as the border between the P-gp non-substrate and the P-gp weak substrate using the median of the data between 1.04 and 3.20, which was 1.4. The second threshold was 9.8, determined using the median of compounds to be >3.20 (geometric mean NER of weak substrate). The compounds with P-gp NER less than 1.4 (P-gp NER < 1.4), between 1.4 and 9.8 (1.4 ≤ P-gp NER ≤ 9.8), and more than 9.8 (9.8 < P-gp NER) were defined as low-, middle-, and high-class compounds, respectively.

Estimation of the P-gp NER in BCEC

Previous studies have reported the methodology to reconstruct in vivo P-gp function in the BBB (BCEC) from in vitro P-gp function in P-gp-expressing cells. (9,10) They demonstrated that the in vivo P-gp transport activity (NER) could be extrapolated from in vitro P-gp transport activity in P-gp-expressing cells by correction of the P-gp protein expression amount. We measured the P-gp protein expression amount of several cell lines and estimated the in vivo P-gp NER in rats from in vitro P-gp NER according to these reports. The measured in vitro P-gp protein expression amount in rat and human P-gp-overexpressing cells according to a previous protocol (37) were 15.0 and 7.72 fmol/μg protein in human and rat P-gp-overexpressing cells, respectively. Human and rat P-gp protein expressions in BCEC have previously been reported to be 6.06 (12) and 19.1 (15) fmol/μg protein, respectively. Accordingly, we calculated the human and rat P-gp NER in BCEC using eqs 1 and 2, respectively
(1)
(2)
We translated the P-gp NER in human BCEC from in vitro human P-gp NER according to eq 1 and then estimated P-gp NER in rat BCEC from P-gp NER in human BCEC according to the correlation, eq 3, in Figure 1
(3)
The median of in vitro human P-gp NER in middle- and high-class compounds (2.6 and 18.9 in the middle and migh classes, respectively), which were set in “Threshold of Classification Model for P-gp NER” section, were converted to the representative values of P-gp NER in rat BCEC according to eq 3 following the conversion to human P-gp NER in BCEC according to eq 1. The representative value of P-gp NER rat in BCEC for the compounds classified into middle and high classes by the classification model was set to 2.2 and 26.3 in middle and high classes, respectively. Furthermore, we used 1.0 as the final representative value for the compounds classified as low class using the classification model as it does not influence P-gp in low-class compounds.

Correction of Kp,brain and Calculation of Kp,uu,brain

Kp,brain values were calculated using eq 4
(4)
where Cbrain and Cplasma are concentrations of a compound in the brain and plasma, respectively.
The Kp,brain value in P-gp KO rats can be corrected to that of WT rat using the following equation described by Summerfield et al. (8)
(5)
Using eq 5, Kp,brain that takes into account the influence of P-gp can be estimated. However, the numerical value cannot be substituted in P-gp NER in BCEC in the prediction process as the quantitative prediction method of P-gp NER could not be constructed. Therefore, in this report, we propose the substitution to the representative values of P-gp NER based on the classification model for P-gp NER instead of the numerical value of P-gp NER in rat BCEC when the corrected Kp,brain is calculated using eq 5.
Thus, the Kp,brain value was corrected using each representative value calculated using eq 6 in Figure 4
(6)
The representative values (low = 1, middle = 2.2, or high = 26.3) were set.
Finally, the Kp,uu,brain value was calculated as follows
(7)

Kp,brain Value Calculated Using Rodgers’ Equation

Kp,brain was calculated for 28 compounds that are P-gp substrates in the external test set using a differential phospholipid method proposed by Rodgers et al. (6,7) The values of four physicochemical/PK parameters, namely, log P, pKa, fu,p, and blood to plasma concentration ratio (Rb) were required for the Kp,brain calculation using Rodgers’ equation. Log P and pKa were acquired from ChEMBL27. As predicted parameters, the Rb value was set to 1.0 and the predicted fu,p value with Model_fup was used. To compare the predicted accuracy of Kp,brain when using the predicted values, the experimental fu,p value and the Rb value in rats were obtained.

Supporting Information

ARTICLE SECTIONS
Jump To

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jmedchem.0c02011.

  • Experimental data and datasets; model construction; process of prediction model construction in fu,brain, fu,p, and P-gp NER; individual statistical results during condition examination in the construction of prediction models for fu,brain, fu,p, and P-gp NER; plots of prediction results in Model_fubrain and Model_fup; difference between the experimental values and the calculated Kp,brain, corrected Kp,brain, or Kp,uu,brain values using predicted values; plot of Kp,uu,brain in WT rats versus calculated Kp,uu,brain using predicted results; and experimental data acquisition (PDF)

  • Dataset_pgpNER, Dataset_fu,brain, Dataset_fu,p, Ex-testset_pgpNER, Ex-testset_fu,brain, and Ex-testset_fu,p (XLSX)

Terms & Conditions

Most electronic Supporting Information files are available without a subscription to ACS Web Editions. Such files may be downloaded by article for research use (if there is a public use license linked to the relevant article, that license may permit other uses). Permission may be obtained from ACS for other uses through requests via the RightsLink permission system: http://pubs.acs.org/page/copyright/permissions.html.

Author Information

ARTICLE SECTIONS
Jump To

  • Corresponding Authors
    • Reiko Watanabe - Laboratory of Bioinformatics, Artificial Intelligence Center for Health and Biomedical Research, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka 567-0085, JapanOrcidhttp://orcid.org/0000-0001-9359-8731 Email: [email protected]
    • Rikiya Ohashi - Laboratory of Bioinformatics, Artificial Intelligence Center for Health and Biomedical Research, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka 567-0085, JapanDiscovery Technology Laboratories, Mitsubishi Tanabe Pharma Corporation, Muraoka-Higashi, Fujisawa, Kanagawa 251-8555, Japan Email: [email protected]
  • Authors
    • Tsuyoshi Esaki - The Center for Data Science Education and Research, Shiga University, Hikone, Shiga 522-8522, Japan
    • Masataka Kuroda - Laboratory of Bioinformatics, Artificial Intelligence Center for Health and Biomedical Research, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka 567-0085, JapanDiscovery Technology Laboratories, Mitsubishi Tanabe Pharma Corporation, Muraoka-Higashi, Fujisawa, Kanagawa 251-8555, Japan
    • Hitoshi Kawashima - Laboratory of Bioinformatics, Artificial Intelligence Center for Health and Biomedical Research, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka 567-0085, Japan
    • Hiroshi Komura - URA Center, Osaka City University, Osaka 545-0051, Japan
    • Yayoi Natsume-Kitatani - Laboratory of Bioinformatics, Artificial Intelligence Center for Health and Biomedical Research, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka 567-0085, Japan
    • Kenji Mizuguchi - Laboratory of Bioinformatics, Artificial Intelligence Center for Health and Biomedical Research, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka 567-0085, JapanLaboratory of In-Silico Drug Design, Center of Drug Design Research, National Institutes of Biomedical Innovation, Health and Nutrition, Osaka 567-0085, JapanOrcidhttp://orcid.org/0000-0003-3021-7078
  • Notes
    The authors declare no competing financial interest.

Acknowledgments

ARTICLE SECTIONS
Jump To

This work was conducted as part of the “Development of a Drug Discovery Informatics System” supported by the Japan Agency for Medical Research and Development (AMED). We thank Dr. Chioko Nagao at Osaka University for extending cooperation in facilitating data collection. We also thank Daisuke Sato in Lifematics Inc. for helping with the creation of the web interface. Finally, we would like to thank Dr. Lokesh Tripathi and Editage (www.editage.jp) for English language editing.

Abbreviations

ARTICLE SECTIONS
Jump To

ANN

artificial neural network

apKa

the most acidic pKa

BBB

blood–brain barrier

BCEC

brain capillary endothelial cells

BCRP

breast cancer resistance protein

bpKa

the most basic pKa

CSF

cerebrospinal fluid

CNS

central nervous system

fu,brain

fraction unbound in brain homogenate

fu,p

fraction unbound in plasma

GB

gradient boosting

KO

knock out

Kp,brain

brain-to-plasma concentration ratio

Kp,uu,brain

unbound brain-to-plasma concentration ratio

MDR1

multidrug resistance 1

MSE

mean square error

nAromAtoms

the number of aromatic atoms

nAromBonds

the number of aromatic bonds

NER

net efflux ratio

nRots

the number of rotatable bonds

OATP

organic anion-transporting polypeptide

P-gp

P-glycoprotein

Rb

blood to plasma concentration ratio

RF

random forest

SDF

structure data file

SVM

support vector machine

WT

wild-type

References

ARTICLE SECTIONS
Jump To

This article references 37 other publications.

  1. 1
    Palmer, A. M.; Alavijeh, M. S. Translational CNS Medicines Research. Drug Discovery Today 2012, 17, 10681078,  DOI: 10.1016/j.drudis.2012.05.001
  2. 2
    Lin, J. H.; Yamazaki, M. Role of P-glycoprotein in Pharmacokinetics: Clinical Implications. Clin. Pharmacokinet. 2003, 42, 5998,  DOI: 10.2165/00003088-200342010-00003
  3. 3
    Esaki, T.; Ohashi, R.; Watanabe, R.; Natsume-Kitatani, Y.; Kawashima, H.; Nagao, C.; Mizuguchi, K. Computational Model to Predict the Fraction of Unbound Drug in the Brain. J. Chem. Inf. Model. 2019, 59, 32513261,  DOI: 10.1021/acs.jcim.9b00180
  4. 4
    Watanabe, R.; Esaki, T.; Kawashima, H.; Natsume-Kitatani, Y.; Nagao, C.; Ohashi, R.; Mizuguchi, K. Predicting Fraction Unbound in Human Plasma from Chemical Structure: Improved Accuracy in the Low Value Ranges. Mol. Pharm. 2018, 15, 53025311,  DOI: 10.1021/acs.molpharmaceut.8b00785
  5. 5
    Ohashi, R.; Watanabe, R.; Esaki, T.; Taniguchi, T.; Torimoto-Katori, N.; Watanabe, T.; Ogasawara, Y.; Takahashi, T.; Tsukimoto, M.; Mizuguchi, K. Development of Simplified in Vitro P-glycoprotein Substrate Assay and in Silico Prediction Models to Evaluate Transport Potential of P-glycoprotein. Mol. Pharm. 2019, 16, 18511863,  DOI: 10.1021/acs.molpharmaceut.8b01143
  6. 6
    Rodgers, T.; Leahy, D.; Rowland, M. Physiologically Based Pharmacokinetic Modeling 1: Predicting the Tissue Distribution of Moderate-to-Strong Bases. J. Pharm. Sci. 2005, 94, 12591276,  DOI: 10.1002/jps.20322
  7. 7
    Rodgers, T.; Rowland, M. Physiologically Based Pharmacokinetic Modelling 2: Predicting the Tissue Distribution of Acids, Very Weak Bases, Neutrals and Zwitterions. J. Pharm. Sci. 2006, 95, 12381257,  DOI: 10.1002/jps.20502
  8. 8
    Summerfield, S. G.; Stevens, A. J.; Cutler, L.; del Carmen Osuna, M.; Hammond, B.; Tang, S.-P.; Hersey, A.; Spalding, D. J.; Jeffrey, P. Improving the in Vitro Prediction of in Vivo Central Nervous System Penetration: Integrating Permeability, P-glycoprotein Efflux, and Free Fractions in Blood and Brain. J. Pharmacol. Exp. Ther. 2006, 316, 12821290,  DOI: 10.1124/jpet.105.092916
  9. 9
    Uchida, Y.; Ohtsuki, S.; Kamiie, J.; Terasaki, T. Blood-Brain Barrier (BBB) Pharmacoproteomics: Reconstruction of in Vivo Brain Distribution of 11 P-glycoprotein Substrates Based on the BBB Transporter Protein Concentration, in Vitro Intrinsic Transport Activity, and Unbound Fraction in Plasma and Brain in Mice. J. Pharmacol. Exp. Ther. 2011, 339, 579588,  DOI: 10.1124/jpet.111.184200
  10. 10
    Ohtsuki, S.; Uchida, Y.; Kubo, Y.; Terasaki, T. Quantitative Targeted Absolute Proteomics-Based Adme Research as a New Path to Drug Discovery and Development: Methodology, Advantages, Strategy, and Prospects. J. Pharm. Sci. 2011, 100, 35473559,  DOI: 10.1002/jps.22612
  11. 11
    Kamiie, J.; Ohtsuki, S.; Iwase, R.; Ohmine, K.; Katsukura, Y.; Yanai, K.; Sekine, Y.; Uchida, Y.; Ito, S.; Terasaki, T. Quantitative Atlas of Membrane Transporter Proteins: Development and Application of a Highly Sensitive Simultaneous LC/MS/MS Method Combined with Novel in-Silico Peptide Selection Criteria. Pharm. Res. 2008, 25, 14691483,  DOI: 10.1007/s11095-008-9532-4
  12. 12
    Uchida, Y.; Ohtsuki, S.; Katsukura, Y.; Ikeda, C.; Suzuki, T.; Kamiie, J.; Terasaki, T. Quantitative Targeted Absolute Proteomics of Human Blood-Brain Barrier Transporters and Receptors. J. Neurochem. 2011, 117, 333345,  DOI: 10.1111/j.1471-4159.2011.07208.x
  13. 13
    Ito, K.; Uchida, Y.; Ohtsuki, S.; Aizawa, S.; Kawakami, H.; Katsukura, Y.; Kamiie, J.; Terasaki, T. Quantitative Membrane Protein Expression at the Blood-Brain Barrier of Adult and Younger Cynomolgus Monkeys. J. Pharm. Sci. 2011, 100, 39393950,  DOI: 10.1002/jps.22487
  14. 14
    Agarwal, S.; Uchida, Y.; Mittapalli, R. K.; Sane, R.; Terasaki, T.; Elmquist, W. F. Quantitative Proteomics of Transporter Expression in Brain Capillary Endothelial Cells Isolated from P-glycoprotein (P-gp), Breast Cancer Resistance Protein (BCRP), and P-gp/BCRP Knockout Mice. Drug Metab. Dispos. 2012, 40, 11641169,  DOI: 10.1124/dmd.112.044719
  15. 15
    Hoshi, Y.; Uchida, Y.; Tachikawa, M.; Inoue, T.; Ohtsuki, S.; Terasaki, T. Quantitative Atlas of Blood-Brain Barrier Transporters, Receptors, and Tight Junction Proteins in Rats and Common Marmoset. J. Pharm. Sci. 2013, 102, 33433355,  DOI: 10.1002/jps.23575
  16. 16
    Esaki, T.; Watanabe, R.; Kawashima, H.; Ohashi, R.; Natsume-Kitatani, Y.; Nagao, C.; Mizuguchi, K. Data Curation Can Improve the Prediction Accuracy of Metabolic Intrinsic Clearance. Mol. Inf. 2019, 38, 1800086,  DOI: 10.1002/minf.201800086
  17. 17
    Uchida, Y.; Yagi, Y.; Takao, M.; Tano, M.; Umetsu, M.; Hirano, S.; Usui, T.; Tachikawa, M.; Terasaki, T. Comparison of Absolute Protein Abundances of Transporters and Receptors among Blood-Brain Barriers at Different Cerebral Regions and the Blood-Spinal Cord Barrier in Humans and Rats. Mol. Pharm. 2020, 17, 20062020,  DOI: 10.1021/acs.molpharmaceut.0c00178
  18. 18
    Kikuchi, R.; de Morais, S. M.; Kalvass, J. C. In Vitro P-Glycoprotein Efflux Ratio Can Predict the in Vivo Brain Penetration Regardless of Biopharmaceutics Drug Disposition Classification System Class. Drug Metab. Dispos. 2013, 41, 20122017,  DOI: 10.1124/dmd.113.053868
  19. 19
    Takeuchi, T.; Yoshitomi, S.; Higuchi, T.; Ikemoto, K.; Niwa, S.-I.; Ebihara, T.; Katoh, M.; Yokoi, T.; Asahi, S. Establishment and Characterization of the Transformants Stably-Expressing MDR1 Derived from Various Animal Species in LLC-PK1. Pharm. Res. 2006, 23, 14601472,  DOI: 10.1007/s11095-006-0285-7
  20. 20
    Wang, Y.-H.; Li, Y.; Yang, S.-L.; Yang, L. Classification of Substrates and Inhibitors of P-glycoprotein Using Unsupervised Machine Learning Approach. J. Chem. Inf. Model. 2005, 45, 750757,  DOI: 10.1021/ci050041k
  21. 21
    Dolgikh, E.; Watson, I. A.; Desai, P. V.; Sawada, G. A.; Morton, S.; Jones, T. M.; Raub, T. J. Qsar Model of Unbound Brain-to-Plasma Partition Coefficient, Kp,Uu,Brain: Incorporating P-glycoprotein Efflux as a Variable. J. Chem. Inf. Model. 2016, 56, 22252233,  DOI: 10.1021/acs.jcim.6b00229
  22. 22
    Liu, H.; Dong, K.; Zhang, W.; Summerfield, S. G.; Terstappen, G. C. Prediction of Brain:Blood Unbound Concentration Ratios in CNS Drug Discovery Employing in Silico and in Vitro Model Systems. Drug Discovery Today 2018, 23, 13571372,  DOI: 10.1016/j.drudis.2018.03.002
  23. 23
    Eechoute, K.; Sparreboom, A.; Burger, H.; Franke, R. M.; Schiavon, G.; Verweij, J.; Loos, W. J.; Wiemer, E. A. C.; Mathijssen, R. H. J. Drug Transporters and Imatinib Treatment: Implications for Clinical Practice. Clin. Cancer Res. 2011, 17, 406415,  DOI: 10.1158/1078-0432.ccr-10-2250
  24. 24
    Chen, Y.; Agarwal, S.; Shaik, N. M.; Chen, C.; Yang, Z.; Elmquist, W. F. P-glycoprotein and Breast Cancer Resistance Protein Influence Brain Distribution of Dasatinib. J. Pharmacol. Exp. Ther. 2009, 330, 956963,  DOI: 10.1124/jpet.109.154781
  25. 25
    Liu, H.; Yu, N.; Lu, S.; Ito, S.; Zhang, X.; Prasad, B.; He, E.; Lu, X.; Li, Y.; Wang, F.; Xu, H.; An, G.; Unadkat, J. D.; Kusuhara, H.; Sugiyama, Y.; Sahi, J. Solute Carrier Family of the Organic Anion-Transporting Polypeptides 1A2- Madin-Darby Canine Kidney II: A Promising in Vitro System to Understand the Role of Organic Anion-Transporting Polypeptide 1A2 in Blood-Brain Barrier Drug Penetration. Drug Metab. Dispos. 2015, 43, 10081018,  DOI: 10.1124/dmd.115.064170
  26. 26
    Zhang, D.; He, K.; Herbst, J. J.; Kolb, J.; Shou, W.; Wang, L.; Balimane, P. V.; Han, Y.-H.; Gan, J.; Frost, C. E.; Humphreys, W. G. Characterization of Efflux Transporters Involved in Distribution and Disposition of Apixaban. Drug Metab. Dispos. 2013, 41, 827835,  DOI: 10.1124/dmd.112.050260
  27. 27
    Chen, X.; Slattengren, T.; de Lange, E. C. M.; Smith, D. E.; Hammarlund-Udenaes, M. Revisiting Atenolol as a Low Passive Permeability Marker. Fluids Barriers CNS 2017, 14, 30,  DOI: 10.1186/s12987-017-0078-x
  28. 28
    Iisalo, E. Clinical Pharmacokinetics of Digoxin. Clin. Pharmacokinet. 1977, 2, 116,  DOI: 10.2165/00003088-197702010-00001
  29. 29
    Noël, F.; Azalim, P.; do Monte, F. M.; Quintas, L. E. M.; Katz, A.; Karlish, S. J. D. Revisiting the Binding Kinetics and Inhibitory Potency of Cardiac Glycosides on Na(+),K(+)-ATPase (a1b1): Methodological Considerations. J. Pharmacol. Toxicol. Methods 2018, 94, 6472,  DOI: 10.1016/j.vascn.2018.09.001
  30. 30
    Friden, M.; Winiwarter, S.; Jerndal, G.; Bengtsson, O.; Wan, H.; Bredberg, U.; Hammarlund-Udenaes, M.; Antonsson, M. Structure-Brain Exposure Relationships in Rat and Human Using a Novel Data Set of Unbound Drug Concentrations in Brain Interstitial and Cerebrospinal Fluids. J. Med. Chem. 2009, 52, 62336243,  DOI: 10.1021/jm901036q
  31. 31
    Spreafico, M.; Jacobson, M. P. In Silico Prediction of Brain Exposure: Drug Free Fraction, Unbound Brain to Plasma Concentration Ratio and Equilibrium Half-Life. Curr. Top. Med. Chem. 2013, 13, 813820,  DOI: 10.2174/1568026611313070004
  32. 32
    Loryan, I.; Sinha, V.; Mackie, C.; Van Peer, A.; Drinkenburg, W. H.; Vermeulen, A.; Heald, D.; Hammarlund-Udenaes, M.; Wassvik, C. M. Molecular Properties Determining Unbound Intracellular and Extracellular Brain Exposure of CNS Drug Candidates. Mol. Pharm. 2015, 12, 520532,  DOI: 10.1021/mp5005965
  33. 33
    Kodaira, H.; Kusuhara, H.; Fujita, T.; Ushiki, J.; Fuse, E.; Sugiyama, Y. Quantitative Evaluation of the Impact of Active Efflux by P-glycoprotein and Breast Cancer Resistance Protein at the Blood-Brain Barrier on the Predictability of the Unbound Concentrations of Drugs in the Brain Using Cerebrospinal Fluid Concentration as a Surrogate. J. Pharmacol. Exp. Ther. 2011, 339, 935944,  DOI: 10.1124/jpet.111.180398
  34. 34
    Liu, X.; Smith, B. J.; Chen, C.; Callegari, E.; Becker, S. L.; Chen, X.; Cianfrogna, J.; Doran, A. C.; Doran, S. D.; Gibbs, J. P.; Hosea, N.; Liu, J.; Nelson, F. R.; Szewc, M. A.; Van Deusen, J. Evaluation of Cerebrospinal Fluid Concentration and Plasma Free Concentration as a Surrogate Measurement for Brain Free Concentration. Drug Metab. Dispos. 2006, 34, 14431447,  DOI: 10.1124/dmd.105.008201
  35. 35
    Ueda, K.; Okamura, N.; Hirai, M.; Tanigawara, Y.; Saeki, T.; Kioka, N.; Komano, T.; Hori, R. Human P-Glycoprotein Transports Cortisol, Aldosterone, and Dexamethasone, but Not Progesterone. J. Biol. Chem. 1992, 267, 2424824252,  DOI: 10.1016/s0021-9258(18)35757-0
  36. 36
    Tanigawara, Y.; Okamura, N.; Hirai, M.; Yasuhara, M.; Ueda, K.; Kioka, N.; Komano, T.; Hori, R. Transport of Digoxin by Human P-glycoprotein Expressed in a Porcine Kidney Epithelial Cell Line (LLC-PK1). J. Pharmacol. Exp. Ther. 1992, 263, 840845
  37. 37
    Uchida, Y.; Tachikawa, M.; Obuchi, W.; Hoshi, Y.; Tomioka, Y.; Ohtsuki, S.; Terasaki, T. A Study Protocol for Quantitative Targeted Absolute Proteomics (QTAP) by LC-MS/MS: Application for Inter-Strain Differences in Protein Expression Levels of Transporters, Receptors, Claudin-5, and Marker Proteins at the Blood–Brain Barrier in ddY, FVB, and C57BL/6J Mice. Fluids Barriers CNS 2013, 10, 21,  DOI: 10.1186/2045-8118-10-21

Cited By

ARTICLE SECTIONS
Jump To

This article is cited by 22 publications.

  1. Yulong Gou, Suyong Re, Kenji Mizuguchi, Chioko Nagao. Impact of Hydrogen Bonding on P-Glycoprotein Efflux Transport as Revealed by Evaluation of a De Novo Prediction Model. ACS Medicinal Chemistry Letters 2024, 15 (1) , 54-59. https://doi.org/10.1021/acsmedchemlett.3c00376
  2. Hideaki Mamada, Mari Takahashi, Mizuki Ogino, Yukihiro Nomura, Yoshihiro Uesawa. Predictive Models Based on Molecular Images and Molecular Descriptors for Drug Screening. ACS Omega 2023, 8 (40) , 37186-37195. https://doi.org/10.1021/acsomega.3c04073
  3. Hitoshi Kawashima, Reiko Watanabe, Tsuyoshi Esaki, Masataka Kuroda, Chioko Nagao, Yayoi Natsume-Kitatani, Rikiya Ohashi, Hiroshi Komura, Kenji Mizuguchi. DruMAP: A Novel Drug Metabolism and Pharmacokinetics Analysis Platform. Journal of Medicinal Chemistry 2023, 66 (14) , 9697-9709. https://doi.org/10.1021/acs.jmedchem.3c00481
  4. Morgan Lawrenz, Mats Svensson, Mitsunori Kato, Karen H. Dingley, Jackson Chief Elk, Zhe Nie, Yefen Zou, Zachary Kaplan, H. Rachel Lagiakos, Hideyuki Igawa, Eric Therrien. A Computational Physics-based Approach to Predict Unbound Brain-to-Plasma Partition Coefficient, Kp,uu. Journal of Chemical Information and Modeling 2023, 63 (12) , 3786-3798. https://doi.org/10.1021/acs.jcim.3c00150
  5. Isaac M. Jackson, E. William Webb, Peter J.H. Scott, Michelle L. James. In Silico Approaches for Addressing Challenges in CNS Radiopharmaceutical Design. ACS Chemical Neuroscience 2022, 13 (12) , 1675-1683. https://doi.org/10.1021/acschemneuro.2c00269
  6. Bharti Devi, Sumukh Satyanarayana Vasishta, Bhanuranjan Das, Anurag T. K. Baidya, Rahul Salmon Rampa, Manoj Kumar Mahapatra, Rajnish Kumar. Integrated use of ligand and structure-based virtual screening, molecular dynamics, free energy calculation and ADME prediction for the identification of potential PTP1B inhibitors. Molecular Diversity 2024, 28 (2) , 649-669. https://doi.org/10.1007/s11030-023-10608-8
  7. Mayuri Gupta, Jun Feng, Govinda Bhisetti. Experimental and Computational Methods to Assess Central Nervous System Penetration of Small Molecules. Molecules 2024, 29 (6) , 1264. https://doi.org/10.3390/molecules29061264
  8. Soné Kotze, Andrea Ebert, Kai-Uwe Goss. Effects of Aqueous Boundary Layers and Paracellular Transport on the Efflux Ratio as a Measure of Active Transport Across Cell Layers. Pharmaceutics 2024, 16 (1) , 132. https://doi.org/10.3390/pharmaceutics16010132
  9. Yongfen Ma, Mengrong Jiang, Huma Javeria, Dingwei Tian, Zhenxia Du. Accurate prediction of Kp,uu,brain based on experimental measurement of Kp,brain and computed physicochemical properties of candidate compounds in CNS drug discovery. Heliyon 2024, 10 (2) , e24304. https://doi.org/10.1016/j.heliyon.2024.e24304
  10. Wanessa S. Mota, Simone S.C. Oliveira, Matheus M. Pereira, Damião P. Souza, Mayara Castro, Pollyanna S. Gomes, Herbert L.M. Guedes, Vinícius F. Souza, André L.S. Santos, Ricardo L.C. Albuquerque-Junior, Juliana C. Cardoso, Cristina Blanco-Llamero, Sona Jain, Eliana B. Souto, Patrícia Severino. Isopentyl caffeate as a promising drug for the treatment of leishmaniasis: An in silico and in vivo study. Current Research in Biotechnology 2024, 7 , 100209. https://doi.org/10.1016/j.crbiot.2024.100209
  11. Jéssica Veiga-Matos, Ana I. Morales, Marta Prieto, Fernando Remião, Renata Silva. Study Models of Drug–Drug Interactions Involving P-Glycoprotein: The Potential Benefit of P-Glycoprotein Modulation at the Kidney and Intestinal Levels. Molecules 2023, 28 (22) , 7532. https://doi.org/10.3390/molecules28227532
  12. Hiroshi Komura, Reiko Watanabe, Kenji Mizuguchi. The Trends and Future Prospective of In Silico Models from the Viewpoint of ADME Evaluation in Drug Discovery. Pharmaceutics 2023, 15 (11) , 2619. https://doi.org/10.3390/pharmaceutics15112619
  13. Siyu Liu, Yohei Kosugi. Human Brain Penetration Prediction Using Scaling Approach from Animal Machine Learning Models. The AAPS Journal 2023, 25 (5) https://doi.org/10.1208/s12248-023-00850-1
  14. Yanling Zhao, Han Yan, Xue Liang, Zhenyu Zhang, Xuan Wang, Nianwei Shi, Weihong Bian, Qing Di, He Huang. Hydrogen Sulfide Attenuates Lipopolysaccharide-Induced Inflammation via the P-glycoprotein and NF-κB Pathway in Astrocytes. Neurochemical Research 2023, 48 (5) , 1424-1437. https://doi.org/10.1007/s11064-022-03840-5
  15. Bhanuranjan Das, Alen T. Mathew, Anurag T. K. Baidya, Bharti Devi, Rahul Rampa Salmon, Rajnish Kumar. Artificial intelligence assisted identification of potential tau aggregation inhibitors: ligand- and structure-based virtual screening, in silico ADME, and molecular dynamics study. Molecular Diversity 2023, 13 https://doi.org/10.1007/s11030-023-10645-3
  16. Liadys Mora Lagares, Marjana Novič. Recent Advances on P-Glycoprotein (ABCB1) Transporter Modelling with In Silico Methods. International Journal of Molecular Sciences 2022, 23 (23) , 14804. https://doi.org/10.3390/ijms232314804
  17. Ben B. Ma, Andrew P. Montgomery, Biling Chen, Michael Kassiou, Jonathan J. Danon. Strategies for targeting the P2Y12 receptor in the central nervous system. Bioorganic & Medicinal Chemistry Letters 2022, 71 , 128837. https://doi.org/10.1016/j.bmcl.2022.128837
  18. Louise Breuil, Sébastien Goutal, Solène Marie, Antonio Del Vecchio, Davide Audisio, Amélie Soyer, Maud Goislard, Wadad Saba, Nicolas Tournier, Fabien Caillé. Comparison of the Blood–Brain Barrier Transport and Vulnerability to P-Glycoprotein-Mediated Drug–Drug Interaction of Domperidone versus Metoclopramide Assessed Using In Vitro Assay and PET Imaging. Pharmaceutics 2022, 14 (8) , 1658. https://doi.org/10.3390/pharmaceutics14081658
  19. Matteo Pavan, Davide Bassani, Giovanni Bolcato, Maicol Bissaro, Mattia Sturlese, Stefano Moro. Computational Strategies to Identify New Drug Candidates against Neuroinflammation. Current Medicinal Chemistry 2022, 29 (27) , 4756-4775. https://doi.org/10.2174/0929867329666220208095122
  20. E. Johanna L. Stéen, Danielle J. Vugts, Albert D. Windhorst. The Application of in silico Methods for Prediction of Blood-Brain Barrier Permeability of Small Molecule PET Tracers. Frontiers in Nuclear Medicine 2022, 2 https://doi.org/10.3389/fnume.2022.853475
  21. Sabina Quader, Kazunori Kataoka, Horacio Cabral. Nanomedicine for brain cancer. Advanced Drug Delivery Reviews 2022, 182 , 114115. https://doi.org/10.1016/j.addr.2022.114115
  22. Chad R. Johnson, Brian D. Kangas, Emily M. Jutkiewicz, Jack Bergman, Andrew Coop. Drug Design Targeting the Muscarinic Receptors and the Implications in Central Nervous System Disorders. Biomedicines 2022, 10 (2) , 398. https://doi.org/10.3390/biomedicines10020398
  • Abstract

    Scheme 1

    Scheme 1. (A) Overview for Model Building; (B) Validation of the Correction Method for the Prediction of Kp,uu,brain from Chemical Structure Information

    Figure 1

    Figure 1. Correlation of the reconstructed P-gp NER between rat and human BCEC. The regression equation (eq 3) is shown in the top left corner; the number of compounds (n) and regression coefficient (R) are also shown. The solid line represents regression.

    Scheme 2

    Scheme 2. Calculation Process of Representative Values in P-gp NER through the Thresholds of the P-gp NER Prediction Model

    Figure 2

    Figure 2. Comparison of 12 physicochemical and PK properties between the degrees of Kp,uu,brain. The compounds that were 0.1 > Kp,uu,brain (n = 36), 0.1 ≤ Kp,uu,brain < 0.5 (n = 16), and 0.5 ≤ Kp,uu,brain (n = 31) are shown in blue, orange, and green boxes, respectively. *, p < 0.01; **, p < 0.05; cross mark, mean; the numbers above and below the bars indicate the median of the parameters that show significant differences.

    Figure 3

    Figure 3. Plot showing the experimental Kp,brain values in WT and (A) experimental Kp,brain value in P-gp KO rat without correction; (B) corrected Kp,brain using the experimental P-gp NER numerical values in rat BCEC reconstructed from in vitro rat P-gp NER; (C) corrected Kp,brain using the representative value of P-gp NER in rat BCEC based on experimental P-gp NER numerical values; (D) corrected Kp,brain using the representative value of P-gp NER in rat BCEC based on the predicted class with Model_pgpNER. P-gp substrates and non-substrates are indicated in gray and white circles, respectively. The MSE is shown in top left corners. The percentage of samples with a 3- or 5-fold error is shown at the bottom. Straight, dashed, and dotted lines indicate the lines of unity, 3-fold and 5-fold errors, respectively (n = 46).

    Figure 4

    Figure 4. (A) Plot of Kp,uu,brain in WT rats and p-gp KO rats calculated using experimental fu,p and fu,brain values. (B) Plot of Kp,uu,brain in WT rats and p-gp KO rats calculated using predicted fu,p and fu,brain values. In plots (A,B), P-gp substrates and non-substrates are shown in gray and white circles, respectively. The MSE and number of compounds (n) are shown in the top left and top right corners, respectively. Straight, dashed, and dotted lines indicate the line of unity, 5-fold, and 10-fold errors, respectively.

    Figure 5

    Figure 5. Plot of Kp,brain and Kp,uu,brain in the P-gp substrate before and after correlation. Experimental Kp,brain in WT rats vs Kp,brain calculated using Rodgers’ formula (A), vs Kp,brain corrected using the representative value based on the predicted P-gp NER (B). Experimental Kp,uu,brain in WT rats vs Kp,uu,brain calculated from Kp,brain without correction (C) and vs corrected Kp,uu,brain (D) with the predicted fu,p and fu,brain values. P-gp substrates (n = 28) are plotted. Straight, long-dashed, and dotted lines indicate the line of unity, 5-fold and 10-fold errors, respectively.

  • References

    ARTICLE SECTIONS
    Jump To

    This article references 37 other publications.

    1. 1
      Palmer, A. M.; Alavijeh, M. S. Translational CNS Medicines Research. Drug Discovery Today 2012, 17, 10681078,  DOI: 10.1016/j.drudis.2012.05.001
    2. 2
      Lin, J. H.; Yamazaki, M. Role of P-glycoprotein in Pharmacokinetics: Clinical Implications. Clin. Pharmacokinet. 2003, 42, 5998,  DOI: 10.2165/00003088-200342010-00003
    3. 3
      Esaki, T.; Ohashi, R.; Watanabe, R.; Natsume-Kitatani, Y.; Kawashima, H.; Nagao, C.; Mizuguchi, K. Computational Model to Predict the Fraction of Unbound Drug in the Brain. J. Chem. Inf. Model. 2019, 59, 32513261,  DOI: 10.1021/acs.jcim.9b00180
    4. 4
      Watanabe, R.; Esaki, T.; Kawashima, H.; Natsume-Kitatani, Y.; Nagao, C.; Ohashi, R.; Mizuguchi, K. Predicting Fraction Unbound in Human Plasma from Chemical Structure: Improved Accuracy in the Low Value Ranges. Mol. Pharm. 2018, 15, 53025311,  DOI: 10.1021/acs.molpharmaceut.8b00785
    5. 5
      Ohashi, R.; Watanabe, R.; Esaki, T.; Taniguchi, T.; Torimoto-Katori, N.; Watanabe, T.; Ogasawara, Y.; Takahashi, T.; Tsukimoto, M.; Mizuguchi, K. Development of Simplified in Vitro P-glycoprotein Substrate Assay and in Silico Prediction Models to Evaluate Transport Potential of P-glycoprotein. Mol. Pharm. 2019, 16, 18511863,  DOI: 10.1021/acs.molpharmaceut.8b01143
    6. 6
      Rodgers, T.; Leahy, D.; Rowland, M. Physiologically Based Pharmacokinetic Modeling 1: Predicting the Tissue Distribution of Moderate-to-Strong Bases. J. Pharm. Sci. 2005, 94, 12591276,  DOI: 10.1002/jps.20322
    7. 7
      Rodgers, T.; Rowland, M. Physiologically Based Pharmacokinetic Modelling 2: Predicting the Tissue Distribution of Acids, Very Weak Bases, Neutrals and Zwitterions. J. Pharm. Sci. 2006, 95, 12381257,  DOI: 10.1002/jps.20502
    8. 8
      Summerfield, S. G.; Stevens, A. J.; Cutler, L.; del Carmen Osuna, M.; Hammond, B.; Tang, S.-P.; Hersey, A.; Spalding, D. J.; Jeffrey, P. Improving the in Vitro Prediction of in Vivo Central Nervous System Penetration: Integrating Permeability, P-glycoprotein Efflux, and Free Fractions in Blood and Brain. J. Pharmacol. Exp. Ther. 2006, 316, 12821290,  DOI: 10.1124/jpet.105.092916
    9. 9
      Uchida, Y.; Ohtsuki, S.; Kamiie, J.; Terasaki, T. Blood-Brain Barrier (BBB) Pharmacoproteomics: Reconstruction of in Vivo Brain Distribution of 11 P-glycoprotein Substrates Based on the BBB Transporter Protein Concentration, in Vitro Intrinsic Transport Activity, and Unbound Fraction in Plasma and Brain in Mice. J. Pharmacol. Exp. Ther. 2011, 339, 579588,  DOI: 10.1124/jpet.111.184200
    10. 10
      Ohtsuki, S.; Uchida, Y.; Kubo, Y.; Terasaki, T. Quantitative Targeted Absolute Proteomics-Based Adme Research as a New Path to Drug Discovery and Development: Methodology, Advantages, Strategy, and Prospects. J. Pharm. Sci. 2011, 100, 35473559,  DOI: 10.1002/jps.22612
    11. 11
      Kamiie, J.; Ohtsuki, S.; Iwase, R.; Ohmine, K.; Katsukura, Y.; Yanai, K.; Sekine, Y.; Uchida, Y.; Ito, S.; Terasaki, T. Quantitative Atlas of Membrane Transporter Proteins: Development and Application of a Highly Sensitive Simultaneous LC/MS/MS Method Combined with Novel in-Silico Peptide Selection Criteria. Pharm. Res. 2008, 25, 14691483,  DOI: 10.1007/s11095-008-9532-4
    12. 12
      Uchida, Y.; Ohtsuki, S.; Katsukura, Y.; Ikeda, C.; Suzuki, T.; Kamiie, J.; Terasaki, T. Quantitative Targeted Absolute Proteomics of Human Blood-Brain Barrier Transporters and Receptors. J. Neurochem. 2011, 117, 333345,  DOI: 10.1111/j.1471-4159.2011.07208.x
    13. 13
      Ito, K.; Uchida, Y.; Ohtsuki, S.; Aizawa, S.; Kawakami, H.; Katsukura, Y.; Kamiie, J.; Terasaki, T. Quantitative Membrane Protein Expression at the Blood-Brain Barrier of Adult and Younger Cynomolgus Monkeys. J. Pharm. Sci. 2011, 100, 39393950,  DOI: 10.1002/jps.22487
    14. 14
      Agarwal, S.; Uchida, Y.; Mittapalli, R. K.; Sane, R.; Terasaki, T.; Elmquist, W. F. Quantitative Proteomics of Transporter Expression in Brain Capillary Endothelial Cells Isolated from P-glycoprotein (P-gp), Breast Cancer Resistance Protein (BCRP), and P-gp/BCRP Knockout Mice. Drug Metab. Dispos. 2012, 40, 11641169,  DOI: 10.1124/dmd.112.044719
    15. 15
      Hoshi, Y.; Uchida, Y.; Tachikawa, M.; Inoue, T.; Ohtsuki, S.; Terasaki, T. Quantitative Atlas of Blood-Brain Barrier Transporters, Receptors, and Tight Junction Proteins in Rats and Common Marmoset. J. Pharm. Sci. 2013, 102, 33433355,  DOI: 10.1002/jps.23575
    16. 16
      Esaki, T.; Watanabe, R.; Kawashima, H.; Ohashi, R.; Natsume-Kitatani, Y.; Nagao, C.; Mizuguchi, K. Data Curation Can Improve the Prediction Accuracy of Metabolic Intrinsic Clearance. Mol. Inf. 2019, 38, 1800086,  DOI: 10.1002/minf.201800086
    17. 17
      Uchida, Y.; Yagi, Y.; Takao, M.; Tano, M.; Umetsu, M.; Hirano, S.; Usui, T.; Tachikawa, M.; Terasaki, T. Comparison of Absolute Protein Abundances of Transporters and Receptors among Blood-Brain Barriers at Different Cerebral Regions and the Blood-Spinal Cord Barrier in Humans and Rats. Mol. Pharm. 2020, 17, 20062020,  DOI: 10.1021/acs.molpharmaceut.0c00178
    18. 18
      Kikuchi, R.; de Morais, S. M.; Kalvass, J. C. In Vitro P-Glycoprotein Efflux Ratio Can Predict the in Vivo Brain Penetration Regardless of Biopharmaceutics Drug Disposition Classification System Class. Drug Metab. Dispos. 2013, 41, 20122017,  DOI: 10.1124/dmd.113.053868
    19. 19
      Takeuchi, T.; Yoshitomi, S.; Higuchi, T.; Ikemoto, K.; Niwa, S.-I.; Ebihara, T.; Katoh, M.; Yokoi, T.; Asahi, S. Establishment and Characterization of the Transformants Stably-Expressing MDR1 Derived from Various Animal Species in LLC-PK1. Pharm. Res. 2006, 23, 14601472,  DOI: 10.1007/s11095-006-0285-7
    20. 20
      Wang, Y.-H.; Li, Y.; Yang, S.-L.; Yang, L. Classification of Substrates and Inhibitors of P-glycoprotein Using Unsupervised Machine Learning Approach. J. Chem. Inf. Model. 2005, 45, 750757,  DOI: 10.1021/ci050041k
    21. 21
      Dolgikh, E.; Watson, I. A.; Desai, P. V.; Sawada, G. A.; Morton, S.; Jones, T. M.; Raub, T. J. Qsar Model of Unbound Brain-to-Plasma Partition Coefficient, Kp,Uu,Brain: Incorporating P-glycoprotein Efflux as a Variable. J. Chem. Inf. Model. 2016, 56, 22252233,  DOI: 10.1021/acs.jcim.6b00229
    22. 22
      Liu, H.; Dong, K.; Zhang, W.; Summerfield, S. G.; Terstappen, G. C. Prediction of Brain:Blood Unbound Concentration Ratios in CNS Drug Discovery Employing in Silico and in Vitro Model Systems. Drug Discovery Today 2018, 23, 13571372,  DOI: 10.1016/j.drudis.2018.03.002
    23. 23
      Eechoute, K.; Sparreboom, A.; Burger, H.; Franke, R. M.; Schiavon, G.; Verweij, J.; Loos, W. J.; Wiemer, E. A. C.; Mathijssen, R. H. J. Drug Transporters and Imatinib Treatment: Implications for Clinical Practice. Clin. Cancer Res. 2011, 17, 406415,  DOI: 10.1158/1078-0432.ccr-10-2250
    24. 24
      Chen, Y.; Agarwal, S.; Shaik, N. M.; Chen, C.; Yang, Z.; Elmquist, W. F. P-glycoprotein and Breast Cancer Resistance Protein Influence Brain Distribution of Dasatinib. J. Pharmacol. Exp. Ther. 2009, 330, 956963,  DOI: 10.1124/jpet.109.154781
    25. 25
      Liu, H.; Yu, N.; Lu, S.; Ito, S.; Zhang, X.; Prasad, B.; He, E.; Lu, X.; Li, Y.; Wang, F.; Xu, H.; An, G.; Unadkat, J. D.; Kusuhara, H.; Sugiyama, Y.; Sahi, J. Solute Carrier Family of the Organic Anion-Transporting Polypeptides 1A2- Madin-Darby Canine Kidney II: A Promising in Vitro System to Understand the Role of Organic Anion-Transporting Polypeptide 1A2 in Blood-Brain Barrier Drug Penetration. Drug Metab. Dispos. 2015, 43, 10081018,  DOI: 10.1124/dmd.115.064170
    26. 26
      Zhang, D.; He, K.; Herbst, J. J.; Kolb, J.; Shou, W.; Wang, L.; Balimane, P. V.; Han, Y.-H.; Gan, J.; Frost, C. E.; Humphreys, W. G. Characterization of Efflux Transporters Involved in Distribution and Disposition of Apixaban. Drug Metab. Dispos. 2013, 41, 827835,  DOI: 10.1124/dmd.112.050260
    27. 27
      Chen, X.; Slattengren, T.; de Lange, E. C. M.; Smith, D. E.; Hammarlund-Udenaes, M. Revisiting Atenolol as a Low Passive Permeability Marker. Fluids Barriers CNS 2017, 14, 30,  DOI: 10.1186/s12987-017-0078-x
    28. 28
      Iisalo, E. Clinical Pharmacokinetics of Digoxin. Clin. Pharmacokinet. 1977, 2, 116,  DOI: 10.2165/00003088-197702010-00001
    29. 29
      Noël, F.; Azalim, P.; do Monte, F. M.; Quintas, L. E. M.; Katz, A.; Karlish, S. J. D. Revisiting the Binding Kinetics and Inhibitory Potency of Cardiac Glycosides on Na(+),K(+)-ATPase (a1b1): Methodological Considerations. J. Pharmacol. Toxicol. Methods 2018, 94, 6472,  DOI: 10.1016/j.vascn.2018.09.001
    30. 30
      Friden, M.; Winiwarter, S.; Jerndal, G.; Bengtsson, O.; Wan, H.; Bredberg, U.; Hammarlund-Udenaes, M.; Antonsson, M. Structure-Brain Exposure Relationships in Rat and Human Using a Novel Data Set of Unbound Drug Concentrations in Brain Interstitial and Cerebrospinal Fluids. J. Med. Chem. 2009, 52, 62336243,  DOI: 10.1021/jm901036q
    31. 31
      Spreafico, M.; Jacobson, M. P. In Silico Prediction of Brain Exposure: Drug Free Fraction, Unbound Brain to Plasma Concentration Ratio and Equilibrium Half-Life. Curr. Top. Med. Chem. 2013, 13, 813820,  DOI: 10.2174/1568026611313070004
    32. 32
      Loryan, I.; Sinha, V.; Mackie, C.; Van Peer, A.; Drinkenburg, W. H.; Vermeulen, A.; Heald, D.; Hammarlund-Udenaes, M.; Wassvik, C. M. Molecular Properties Determining Unbound Intracellular and Extracellular Brain Exposure of CNS Drug Candidates. Mol. Pharm. 2015, 12, 520532,  DOI: 10.1021/mp5005965
    33. 33
      Kodaira, H.; Kusuhara, H.; Fujita, T.; Ushiki, J.; Fuse, E.; Sugiyama, Y. Quantitative Evaluation of the Impact of Active Efflux by P-glycoprotein and Breast Cancer Resistance Protein at the Blood-Brain Barrier on the Predictability of the Unbound Concentrations of Drugs in the Brain Using Cerebrospinal Fluid Concentration as a Surrogate. J. Pharmacol. Exp. Ther. 2011, 339, 935944,  DOI: 10.1124/jpet.111.180398
    34. 34
      Liu, X.; Smith, B. J.; Chen, C.; Callegari, E.; Becker, S. L.; Chen, X.; Cianfrogna, J.; Doran, A. C.; Doran, S. D.; Gibbs, J. P.; Hosea, N.; Liu, J.; Nelson, F. R.; Szewc, M. A.; Van Deusen, J. Evaluation of Cerebrospinal Fluid Concentration and Plasma Free Concentration as a Surrogate Measurement for Brain Free Concentration. Drug Metab. Dispos. 2006, 34, 14431447,  DOI: 10.1124/dmd.105.008201
    35. 35
      Ueda, K.; Okamura, N.; Hirai, M.; Tanigawara, Y.; Saeki, T.; Kioka, N.; Komano, T.; Hori, R. Human P-Glycoprotein Transports Cortisol, Aldosterone, and Dexamethasone, but Not Progesterone. J. Biol. Chem. 1992, 267, 2424824252,  DOI: 10.1016/s0021-9258(18)35757-0
    36. 36
      Tanigawara, Y.; Okamura, N.; Hirai, M.; Yasuhara, M.; Ueda, K.; Kioka, N.; Komano, T.; Hori, R. Transport of Digoxin by Human P-glycoprotein Expressed in a Porcine Kidney Epithelial Cell Line (LLC-PK1). J. Pharmacol. Exp. Ther. 1992, 263, 840845
    37. 37
      Uchida, Y.; Tachikawa, M.; Obuchi, W.; Hoshi, Y.; Tomioka, Y.; Ohtsuki, S.; Terasaki, T. A Study Protocol for Quantitative Targeted Absolute Proteomics (QTAP) by LC-MS/MS: Application for Inter-Strain Differences in Protein Expression Levels of Transporters, Receptors, Claudin-5, and Marker Proteins at the Blood–Brain Barrier in ddY, FVB, and C57BL/6J Mice. Fluids Barriers CNS 2013, 10, 21,  DOI: 10.1186/2045-8118-10-21
  • Supporting Information

    Supporting Information

    ARTICLE SECTIONS
    Jump To

    The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jmedchem.0c02011.

    • Experimental data and datasets; model construction; process of prediction model construction in fu,brain, fu,p, and P-gp NER; individual statistical results during condition examination in the construction of prediction models for fu,brain, fu,p, and P-gp NER; plots of prediction results in Model_fubrain and Model_fup; difference between the experimental values and the calculated Kp,brain, corrected Kp,brain, or Kp,uu,brain values using predicted values; plot of Kp,uu,brain in WT rats versus calculated Kp,uu,brain using predicted results; and experimental data acquisition (PDF)

    • Dataset_pgpNER, Dataset_fu,brain, Dataset_fu,p, Ex-testset_pgpNER, Ex-testset_fu,brain, and Ex-testset_fu,p (XLSX)


    Terms & Conditions

    Most electronic Supporting Information files are available without a subscription to ACS Web Editions. Such files may be downloaded by article for research use (if there is a public use license linked to the relevant article, that license may permit other uses). Permission may be obtained from ACS for other uses through requests via the RightsLink permission system: http://pubs.acs.org/page/copyright/permissions.html.