Previous studies have shown that TGF-β is a key molecule that induces lung fibrosis. Inhibition of TGF-β can prevent the differentiation of fibroblasts, thereby slowing down the process of lung fibrosis. We hope to design a truncated receptor to block the activity of TGF-β, that is, only retain the binding domain of the TGF-β receptor and remove the domains related to signal transduction.
According to previous reports, we have found the following four truncated receptors:
（1） P144® (Anti-TGF-β)
（2） TMED10
（3） M7824
（4） Unnamed
We use the Z-DOCK model to obtain the parameters of these four truncated receptors:
| Variable | Declaration |
|---|---|
| iNat | Number of interface atoms |
| iNres | Number of interface residues |
| Surface A2 | Total surface area releasable by solvent |
| Interface area A2 | Area of interaction between two proteins |
| △iG kcal/mol | Free energy when the interface is formed (positive value) |
| △iG P-value | The hydrophobic properties of the interface (P>0.05) |
| NHB | Number of hydrogen bonds on the interface |
| NSB | Number of salt bridges on the interface |
| NDS | Number of disulfide bonds on the interface |
Perform Z-score normalization on these parameters:
A positive value in the figure represents an above average level, which means that it has a relatively strong ability to bind to TGF-β. M7824 and Unnamed have strong binding ability to TGF-β. Considering various parameters, we chose M7824 as the truncated receptor.
The accumulation of type 1 collagen is one of the pathological manifestations of pulmonary fibrosis. We thus screened for a series of probes targeting type I collagen to localize our protein drug to the fibrotic area6.
Next, we designed a series of probes targeting type I collagen (a characteristic molecule of fibrosis) and used Z-DOCK and I-TASSER for molecular docking and stability experiments. After that, three types of collagen-binding targeting probes with KD values of 0.01, 0.1, 10 were selected. The truncated receptor and the probe are connected using HL5 rigid linker. The sequence is as follows:
The truncated receptor and the probe are connected using HL5 rigid linker. The sequence is as follows:
1.M7824+HL5+Original CILY:
MGRGLLRGLW PLHIVLWTRI ASTIPPHVQK SVNNDMIVTD NNGAVKFPQL CKFCDVRFST CDNQKSCMSN CSITSICEKP QEVCVAVWRK NDENITLETV CHDPKLPYHD FILEDAASPK CIMKEKKKPG ETFFMC LAEAAAKEAAAKEAAAKEAAAKEAAAKAAA RRANAALKAGELYKCILY
2. M7824+HL5+SILY:
MGRGLLRGLW PLHIVLWTRI ASTIPPHVQK SVNNDMIVTD NNGAVKFPQL CKFCDVRFST CDNQKSCMSN CSITSICEKP QEVCVAVWRK NDENITLETV CHDPKLPYHD FILEDAASPK CIMKEKKKPG ETFFMC LAEAAAKEAAAKEAAAKEAAAKEAAAKAAA RRANAALKAGELYKCILY
3. M7824+HL5+I type 4:
MGRGLLRGLW PLHIVLWTRI ASTIPPHVQK SVNNDMIVTD NNGAVKFPQL CKFCDVRFST CDNQKSCMSN CSITSICEKP QEVCVAVWRK NDENITLETV CHDPKLPYHD FILEDAASPK CIMKEKKKPG ETFFMC LAEAAAKEAAAKEAAAKEAAAKEAAAKAAA SLACFSPERW
2. M7824+HL5+SILY:
MGRGLLRGLW PLHIVLWTRI ASTIPPHVQK SVNNDMIVTD NNGAVKFPQL CKFCDVRFST CDNQKSCMSN CSITSICEKP QEVCVAVWRK NDENITLETV CHDPKLPYHD FILEDAASPK CIMKEKKKPG ETFFMC LAEAAAKEAAAKEAAAKEAAAKEAAAKAAA RRANAALKAGELYKCILY
We use I-TASSER for simulation. The upper panel shows the binding of the fusion protein to its target, and the lower panel shows the enzyme activity of the fusion protein. The results indicated that no redundant enzyme activity was produced after the combination of the truncated receptor and the probe, that is, the fusion protein would not bind to proteins other than the targeted receptor. We will also combine the results of subsequent wet experiments to find the most suitable combination to optimize our model.
Model ---- PBPK Modeling
We analyzed the metabolism of peptide drugs in animal models and humans in the two treatment modules, and used the physiological pharmacokinetic (PBPK) model prediction software GastroPlusTM to predict all peak concentration-time (Cp-time) curves. We focus on the metabolism of peptide drugs through intravenous administration and pulmonary administration, as well as the influence of the presence or absence of liposomes on drug metabolism.
We first analyzed the drug metabolism of AP26 and SAK (two activation peptides) in rats and humans. Both SAK and AP26 activation peptides have certain side effects. The pharmacokinetic model will guide us in drug screening and delivery methods.
1. Modeling parameters and model formulas
1.1 Modeling parameters
We use the ADMET Predictor module of GastroPlusTM to obtain related modeling parameters based on the prediction of the structural formula of peptide drug molecules. This module is based on the prediction model established by the relationship between compound structural formula and property (QSAR), which can directly predict the respective ADME property parameters through the compound structural formula; the structural formulas of the two peptides are as follows：
Fig. Structural formulas of AP26
Fig. Structural formulas of SAK
The relevant ADME parameters predicted based on the molecular structural formulas of two peptide drugs are shown in the following table：
The prediction results show that AP26 has a lower fat solubility (logP) than SAK; at the same time, the solubility in water is also lower than SAK, but both show higher solubility characteristics; Due to the larger molecular weight, both drug molecules have performance poor intestinal penetration characteristics, suggesting that it is not conducive to the absorption process of oral preparations;
The whole blood/plasma drug concentration ratio of the drug and the plasma free fraction showed similar results, suggesting that the two have similar properties in binding to plasma proteins. In the aspect of the distribution characteristics in the body, AP26 shows less distribution behavior in rats and humans, and SAK can distribute more in tissue cells;
The prediction results also found that the two drugs are less likely to be eliminated by liver metabolism, and there is no relevant liver metabolism value. In terms of renal clearance, the model’s glomerular filtration clearance is used to calculate the free drug fraction multiplied by the glomerular filtration rate. As AP26 exhibits a higher plasma free behavior, a higher renal clearance value is obtained. Taken together, it reflects that AP26 shows a smaller plasma elimination half-life due to its higher clearance ability and smaller plasma distribution value, and its elimination rate in the body will be faster than SAK.
1.2 Plasma volume of distribution
For the systemic circulation process in the body after administration, the volume of distribution (Vc) and clearance rate are the two main influencing factors, which determine the characteristics of the change of drug concentration in plasma. Therefore, in the PBPK model, it is particularly important to accurately examine the clearance and distribution characteristics of AP26 and SAK.
In the work of this project, based on the GastroPlusTM PBPK model and combining the physical and chemical properties of the drug and the physiological parameters of the body, the plasma distribution behavior of the two peptide drugs is calculated through the built-in formula of the software. The corresponding calculation formula is as follows：
In formula，Vp refers to the volume of plasma，Ve is the volume of red blood cells，E:P is the ratio of drug concentration in red blood cells to plasma，Vt is the volume of the organ/tissue, Kpt is the partition coefficient of the drug in tissues and plasma. This parameter can be calculated by combining the properties of the drug and physiological parameters，ERt is the extraction rate of the drug in a specific tissue.
Vnlt is the percentage of neutral lipids in the tissue，Vpht is the percentage of phosphate in the tissue，Vewt is the percentage of extracellular water in the tissue;
Viwt is the percentage of intracellular water in the tissue;
Ka is the combination constant of basic compound and acidic phosphate
[AP]T is the concentration of acid phosphate in each tissue;
X[D], IW, X[D],P are the percentage of neutral drug in the cell (pH=7); and the percentage in the plasma (pH=7.4)；
P is the partition coefficient of the drug in solvent/water；
fup is the free fraction of the drug;
RAtp is the concentration ratio of albumin in tissue and plasma;
Fn+Fa is the percentage of the drug when there is no positive charge in the plasma;
Fc is the percentage of the drug when there is a positive charge in the plasma;
Kpu is the concentration ratio of free tissue to plasma.
1.3 Drug clearance rate
The clearance rate is another important parameter that characterizes the circulation process in the body after administration. This work is based on the molecular structure of the drug to predict the corresponding liver metabolic clearance and renal excretion clearance. Predicted by the QSAR model, it is found that the possibility of two drug molecules being metabolized by the liver is small, and the corresponding metabolic clearance rate is 0. The renal excretion and clearance were further investigated. It is believed that the free drug is excreted and cleared by glomerular filtration. The corresponding clearance rate is characterized by multiplying the fraction of free drug by the glomerular filtration rate (fup*GFR). We ignore the possible bile excretion and clearance of the drug by the transporter, and the renal tubular secretion and reabsorption mediated by the transporter (the corresponding transport kinetic data has not yet been predicted by a mature QSAR model).
Based on the comprehensive effects and investigation of the PBPK model and the QSAR model, the relevant data on the distribution, clearance, and plasma half-life of AP26 and SAK in rats or humans are as follows (part of the data is also listed in the above "modeling parameters" section):
2. PBPK model by intravenous administration
This part aims to use the PBPK model and predicted drug ADME modeling parameters to build and predict the pharmacokinetic curve of AP26 and SAK in rats or humans after intravenous administration, and to view the PK of two drug molecules in the corresponding species Differences in behavior and exposure, and briefly evaluate the PK characteristics of the two.
2.1 Setting of model parameters
2.1.1 Setting of dosing regimen
2.1.2 Setting of prediction model parameters
The basic physical and chemical property parameters of the modeled drugs and the corresponding dosing schedule are as described above; the Gut Physiology interface is mainly used to set the physiological parameters of the gastrointestinal tract of the simulated species. We select the default parameters of the physiological model of the corresponding species; Pharmacokinetics interface It is mainly used to set the physiological parameters of the system circulation process of the simulated species and the modeling parameters that characterize drug clearance and distribution. The PBPK model parameters of tissues and organs are set as the default physiological data of the model.
2.2 Prediction results of PK
According to the above-mentioned related modeling parameters and the built-in PBPK model that characterizes the physiological pharmacokinetics of the drug entering the system circulation and the plasma circulation process, the corresponding PK curve prediction is performed, and the relevant prediction results are shown as follows：
The prediction results found that after intravenous administration, the plasma concentrations of AP26 and SAK in rats and humans showed an initial rapid decline, and then showed a relatively slow elimination process;
Compared with AP26, SAK shows a slower elimination process in rats, and the end of its PK curve will decline more slowly. Since the clearance of the two drugs in rats is relatively similar, it can be found that the plasma exposure AUC0-t of the two drugs predicted by the model is basically similar；但But because SAK presents a slower distribution process, its AUC0-inf is larger than AP26;
In the human body model, because SAK has a greater clearance value and greater distribution behavior, the downward trend of its PK curve is slower than that of AP26; Simultaneous drug exposure in the body (AUC0-t以及AUC0-inf) are all larger than AP26；
For the same drug molecule, due to the species differences between rats and humans (the differences in drug clearance and distribution behavior caused by physiological factors), it will be found that both AP26 and SAK exhibit different PK curves in different species. But the overall curve trend is somewhat similar, that is, SAK shows a slower half-life in rats and humans, while AP26 clears faster.
3. PBPK model by lung administration
This part aims to use the PBPK model and predicted drug ADME modeling parameters to build and predict the pharmacokinetic curve of AP26 and SAK in rats or humans after pulmonary administration, and to view the two drug molecules in the corresponding species PK behavior and exposure differences, and briefly evaluate the PK characteristics of the two.
3.1 Setting of model parameters
3.1.1 Setting of dosing regimen
3.1.2 Setting of prediction model parameters
The basic drug model parameter settings are similar to the above. In order to show the difference between pulmonary drug delivery and intravenous drug delivery, we have made the following improvements in the software: In the Compound interface, due to the change of the dosage form, this part is adjusted to PL: Soln in the Dosage Form option. In addition, after pulmonary administration, the drug will be deposited (Deposition) in various parts of the lung tissue (external chest cavity, chest cavity, bronchi, alveoli, etc.), and further absorbed by the pulmonary blood vessels into the plasma circulation. Therefore, the accurate characterization of the deposition of peptide drugs in various parts of the lung is essential for the reasonable prediction of the drug concentration changes in plasma and lung tissue after pulmonary administration. Here we combine the ADME data of the drug with the physiological parameters of lung tissue, and calculate the deposition of AP26 and SAK in the lungs of rats and humans through the built-in mathematical formula. The corresponding deposition behavior is as follows：
The deposition ratio of AP26 in the lungs of rats: The built-in model predicts that after the administration of AP26 lung absorption solution, it is mainly deposited in the chest cavity of the lung, and the corresponding deposition ratio is 100%
The deposition ratio of SAK in the lungs of rats: The built-in model predicts that after the administration of the SAK lung absorption solution, it is mainly deposited in the outer thoracic cavity. The corresponding deposition ratio is 50.51% and 0.34% in the outer thoracic cavity. Deposited, the rest is not deposited; showing an incomplete deposition process.
3.2 Prediction result of PK
According to the above-mentioned related modeling parameters and the built-in PBPK model that characterizes the physiological pharmacokinetics of the drug entering the system circulation and the plasma circulation process, the corresponding PK curve prediction is performed, and the relevant prediction results are shown as follows：
The prediction results found that, due to the difference in the physiological behavior of the lung tissues between rats and humans, AP26 and SAK after pulmonary administration were found to show a faster plasma concentration peaking process in humans than rats;
Due to the effect of drug deposition after pulmonary administration and the transport process of larger molecules across lung cells, it was found that AP26 and SAK showed low bioavailability in both rats and humans, and the corresponding values were basically distributed in 5~ Within 10%; among them, the bioavailability prediction result of AP26 in human body is better, and the corresponding result is close to 70%;
The difference in PK behavior and bioavailability of the two drug molecules may be caused by differences in drug deposition in lung tissue and transport across cell membranes. Among them, AP26 shows nearly 100% deposition in the thoracic cavity in human lung tissue, which can make the drug in the human body be better absorbed through the pulmonary blood vessels and enter the blood circulation, thereby obtaining higher bioavailability. The SAK is mainly distributed in the outer thoracic cavity, and only about 50% can be deposited in the lungs; at the same time, the outer thoracic cavity is directly connected to the gastrointestinal tract, which may partially return to the gastrointestinal tract without being absorbed into the blood. Bioavailability (about 10%);
Therefore, further investigation of the deposition behavior of the evaluated drug preparations in rats and human lung tissues is important for accurately predicting the PK behavior of these two peptide drug molecules in rats or humans.
4. Release curve and PK curve prediction of sustained-release preparations
In order to predict the effect of liposome delivery mode on drug metabolism mode, we constructed the release curve model of sustained-release preparations based on transdermal administration and predicted the PK curve.
We used GastroPlusTM to build a release profile model for sustained-release preparations. The results showed that the release began almost linearly, and finally reached a plateau within 2-4 hours, with a plateau value at 240 min (T 240 min), with an average release range of 97.1%-104.7%, and at 30 minutes (release midpoint value, The average release range is 45.3%~64.8%) and the release rate at 40 minutes is equal to the slope of the curve.
According to the characteristics of the in vitro release profile, determine the release amount at each time point. In combination with the built-in Weibull equation in the software, a complete release curve is obtained based on the expected release at each point and used for the prediction of the PK curve. The results are as follows:
For the dosage form, we set it as a sustained-release transdermal preparation and select the default SQ: Cont Rel for simulation. Other parameter settings are consistent with the intravenous administration model; the dosage setting is the same as the previous model, and the rat is 1.0 mpk, the human body is 1.5 mpk; the physiological parameters of the simulated object (rat or human) are also consistent. The obtained PK curve prediction results are as follows:
The simulation results show that the sustained-release preparations of AP26 and SAK by transdermal administered have significantly increased exposure in rats and humans, and their bioavailability has basically reached 100%. Due to the difference in the release effect of sustained-release preparations, it may bring about different PK characteristics, so determining the appropriate release behavior of the target preparation has a significant impact on the prediction results of its PK curve.
5. Conclusion of PBPK modeling
We use the molecular structural formulas of AP26 and SAK to predict various ADME modeling parameters through the QSAR model. The relevant modeling parameters are further input into the PBPK model to predict the PK curve of rats and humans after intravenous or pulmonary administration.
（1）Drug clearance rate. The model predicts that after intravenous administration, the plasma concentrations of AP26 and SAK in rats and humans both show an initial rapid decline, and then show a relatively slow elimination process. In addition, AP26 showed a slower elimination process than SAK in rats, and the plasma AUC0-inf of SAK was greater than that of AP26. In the human body model, SAK has a higher clearance rate and a larger distribution behavior, so the downward trend of its PK curve is slower than that of AP26;；
（2）Bioavailability and drug deposition in lung tissue。In the lung administration model, due to the differences in the physiological behavior of the lung tissues between rats and humans, both AP26 and SAK after pulmonary administration were found to show a faster plasma concentration peaking process in humans than in rats. No matter in the lung administration model of rats or humans, both AP26 and SAK showed low bioavailability (in the range of 5-10%, but AP26 has a bioavailability close to 70% in humans). The difference in PK behavior and bioavailability of the two drug molecules may be caused by differences in drug deposition in lung tissue and transport across cell membranes. Therefore, further investigation of the deposition behavior of the evaluated drug preparations in rats and human lung tissues is more important for accurately predicting the PK behavior of these two peptide drug molecules in rats or humans;
（3）In the administration model of sustained-release preparations, the exposure of AP26 and SAK in rats and humans was significantly increased, and the bioavailability basically reached 100%, suggesting that sustained-release reagents can significantly improve the bioavailability of drugs
In summary, through PBPK modeling, we clarified the advantages of lung administration over intravenous administration. After pulmonary administration, the deposition of drugs in the lesion area increased significantly; Drugs embedded in liposomes can be designed into sustained-release preparations to have higher bioavailability to achieve better therapeutic effects. For the two activating peptides (AP26 and SAK), AP26 showed a higher drug deposition rate in lung, higher bioavailability and faster clearance rate. Therefore, AP26 is the final choice due to its potential smaller side effects.
Model ---- Future plan
We want to provide more options for probes loaded on liposomes. The first step in designing a probe is to choose a suitable target which can localize the liposomes to fibrotic area. We selected three targets after investigation: TGFβ, type 1 collagen and macrophages related to fibrosis7. Then screen for proteins that can bind to these targets and use I-TASSER to find the alpha helix in recognition domain of the selected proteins8,9. These sequences composed of alpha helix may become potential probes. We will use wet experiments to verify the affinity and other properties of these probes in the future.
We hope to use in vitro and animal experiments to build a more accurate and comprehensive PBPK model. The disadvantage of our current PBPK model is that we only use the molecular structures of AP26 and SAK to predict the corresponding ADME parameters, and combine the physiological PBPK model to predict the PK behavior of rats and humans after intravenous or lung administration. However, since the experimental data of the drug preparation parameters and the drug's own ADME are not currently available, it may affect the accuracy of the results; Therefore, in order to obtain a more accurate PBPK model, we need to obtain drug-related parameters through further in vivo and in vitro experiments to optimize and verify the PBPK model. In in vitro experiments, in vitro metabolism and transport data are usually scaled up in vivo based on the physiological ratio (ie IVIVE method) to obtain tissue intrinsic clearance (CLint) and other data; In in vivo experiments, we expect to construct IPF diseases model in rats. Through the dosing experiment, we can obtain the plasma drug concentration time data, tissue drug concentration time data, excretion data, etc. After calculation, the tissue plasma distribution ratio (Kp, tissue), absorption rate constant (Ka), Renal clearance rate (CLr), excretion fraction and other parameters can be obtained. Finally, we use the parameters of preclinical animals to predict the relevant parameters of the human body. The relevant experimental design is described in the following table：
Furthermore, we currently only construct the PBPK model of the second treatment module’s drugs. For the first treatment module, we not only need to consider the therapeutic effects of the four candidate drugs in in vitro experiments, but also need to combine the PBPK model to analyze their metabolism in animals in order to select the most suitable peptide drug and dosage, Route of administration, etc.
For the modeling of liposome delivery systems, because we need to optimize the size of liposomes, we need more suitable kinetic models to predict the delivery systems of liposomes of different sizes. On the basis of the Korsmeyer-Peppas and Weibull models, Lu et al.10 proposed using Laplace pressure as the driving factor of release, and proposed a new equation and proved its improvement on the traditional formula. We next consider using this model to optimize the PBPK model of our sustained-release preparations.
Reference
[1]. Yang, J. & Zhang, Y. I-TASSER server: new development for protein structure and function predictions. Nucleic Acids Res 43, W174-181, doi:10.1093/nar/gkv342 (2015).
[2]. Pierce, B. G. et al. ZDOCK server: interactive docking prediction of protein-protein complexes and symmetric multimers. Bioinformatics 30, 1771-1773, doi:10.1093/bioinformatics/btu097 (2014).
[3]. Zarranz-Ventura, J. et al. Transforming growth factor-beta inhibition reduces progression of early choroidal neovascularization lesions in rats: P17 and P144 peptides. PLoS One 8, e65434, doi:10.1371/journal.pone.0065434 (2013).
[4]. Nakano, N. et al. TMED10 Protein Interferes with Transforming Growth Factor (TGF)-β Signaling by Disrupting TGF-β Receptor Complex Formation. J Biol Chem 292, 4099-4112, doi:10.1074/jbc.M116.769109 (2017).
[5]. Lan, Y. et al. Enhanced preclinical antitumor activity of M7824, a bifunctional fusion protein simultaneously targeting PD-L1 and TGF-β. Sci Transl Med 10, doi:10.1126/scitranslmed.aan5488 (2018).
[6]. Wahyudi, H., Reynolds, A. A., Li, Y., Owen, S. C. & Yu, S. M. Targeting collagen for diagnostic imaging and therapeutic delivery. J Control Release 240, 323-331, doi:10.1016/j.jconrel.2016.01.007 (2016).
[7]. Morse, C. et al. Proliferating SPP1/MERTK-expressing macrophages in idiopathic pulmonary fibrosis. Eur Respir J 54, doi:10.1183/13993003.02441-2018 (2019).
[8]. Dong, X., Hudson, N. E., Lu, C. & Springer, T. A. Structural determinants of integrin beta-subunit specificity for latent TGF-beta. Nat Struct Mol Biol 21, 1091-1096, doi:10.1038/nsmb.2905 (2014).
[9]. Liu, S. S. et al. Targeting Degradation of the Transcription Factor C/EBPbeta Reduces Lung Fibrosis by Restoring Activity of the Ubiquitin-Editing Enzyme A20 in Macrophages. Immunity 51, 522-534 e527, doi:10.1016/j.immuni.2019.06.014 (2019).
[10]. Lu, T. & Ten Hagen, T. L. M. A novel kinetic model to describe the ultra-fast triggered release of thermosensitive liposomal drug delivery systems. J Control Release 324, 669-678, doi:10.1016/j.jconrel.2020.05.047 (2020).