1. INTRODUCTION
Phenanthroindolizidine alkaloids (PAs) are a family of plant-derived small molecules that are isolated and characterized from three genera (Cynanchum, Pergularia, and Tylophora) [1]. PAs have several significant potential therapeutic effects, such as anti-arthritic, anti-viral, anti-inflammatory, and anti-lupus effects. The antitumor effect, especially in the treatment of gliomas, has attracted the most attention amongst clinicians [2]. It has been reported that some PA compounds decrease the expression of phosphorylated focal adhesion kinase, matrix metalloproteinase-2 (MMP-2), and MMP-9 proteins, and downregulate the activity of the phosphatidylinositol-3 kinase/AKT and Raf/extracellular signal-regulated kinase pathways, which regulate expression of Bcl-2 family proteins, activate caspase-3, caspase-9, and poly (ADP-ribose) polymerase, and initiate apoptosis in neuroblastoma cells [3]. More importantly, some compounds in this series, such as (+)-deoxytylophorinine (CAT), 13a-(S)-3-pivaloyloxyl-6,7-dimethoxyphenanthro(9,10-b)-indolizidine (CAT3), and 13a-(S)-3-hydroxyl-6,7-dimethoxyphenanthro(9,10-b)-indolizidine (PF403), exert cytotoxicity against SH-SY5Y neuroblastoma cells in vitro and in vivo, and inhibit invasiveness [4]. These findings suggest the potential of some PAs as new anticancer drugs for the treatment of neuroblastomas. In addition to suppression of normal protein and nucleic acid synthesis, some of these compounds exhibit antitumor activity via blockade of the Hedgehog pathway in vitro and enhance the cytotoxicity of temozolomide in vivo in temozolomide-resistant glioblastoma multiforme [5, 6].
However, some PA compounds, such as tylophorine and cryptopleurine, show central nervous system (CNS) toxicity that is too severe to be eligible for clinical trials [2, 7], which is the main obstacle preventing the use of PAs as drugs. Following the isolation of a series of other PA compounds from folk medicines that did not show CNS toxicity [3], interest in research concerning PAs has increased [2, 4, 7–9].
Indeed, new drug discovery is one of humanity’s most sophisticated and cutting-edge intellectual activities. Despite extensive investigations into the synthesis [4, 8], pharmacodynamics [3, 5, 6] and pharmacokinetics [10–12] of this series of compounds, little research has been conducted on the physical and chemical characteristics. For example, PA has a five-membered ring core structure and is highly hydrophobic, which affects its solubility in digestive juice and absorption rates [7], resulting in lower bioavailability and weakening anticancer activity when administered orally [13].
The biological properties of compounds are affected by physical and chemical constants [14–16], such as the acid dissociation constant (pK a) and oil-water partition coefficient (logP and logD). For example, the permeability, bioavailability, blood-brain drug concentration distribution, and accumulated toxicity in different organs are all related to the dissociation constant, hydrophilicity, and lipophilicity. Therefore, in the novel drug screening stage, the influence of physical and chemical constants on the biological properties of prescribed compounds should be recognized, and the values of these physicochemical and biochemical characteristics of compounds, known as drug-like properties, are more advantageous for discovering new drugs [17].
Drug-like properties are attributes of molecules that have sufficiently and highly acceptable absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties to proceed through the completion of phase I clinical trials [17]. However, it is impossible to assess these properties using traditional methods because many compounds need to be tested, each with a series of characteristics. Thus, to understand the drug-like properties exhibited by a large series of molecules, the easiest and least expensive method involves prediction in silico, and the best compounds can be chosen as candidates and compounds with poor drug-like properties can be removed [18, 19]. Thereafter, prediction software can also exhibit large deviations and interfere with reaching valid conclusions [20]. Thus, the best candidates need to be accurately measured before entering rigorous preclinical research [21]. In addition, the accurate determination of physical and chemical constants is important for animal and human physiologic-based pharmacokinetic models [22, 23].
In light of such an in-depth study, it would be valuable to clarify the role of the drug-like characteristics of PAs and extend the range of applications.
Research on the relationship between physicochemical constants and drug-like properties of PA compounds is currently incomplete, which hinders preclinical studies. Hence, the present study aimed to conduct in-depth research on drug-like properties, such as pK a, lipophilicity, permeability, solubility, blood-brain barrier effects, and P-glycoprotein (P -gp) transporters of 44 PA compounds. A systematic synthesis research method and biological evaluation testing did not show CNS toxicity [4] using ADMET Predictor™ in silico. Next, typical candidates were selected from the 44 compounds and data accuracy was validated. Moreover, the absolute bio-absorption of candidates was tested in vivo and compared to predicted values. In addition, the concentrations of compounds in the brain and plasma were tested to validate CNS targets. The potential shortcomings of the series of compounds, such as toxicity, were also considered and more directions for application were discussed. This research will provide a basis to choose appropriate molecular candidates during PA design, making the entire developmental process more convenient and relatively less expensive for researchers and research funding bodies, to suggest the best drug-like compounds for further research. Moreover, the results will provide assurance for the successful clinical evaluation of candidate drugs.
2. MATERIALS AND METHODS
2.1 Materials
CAT3 (purity > 99%), PF403 (purity > 99%), and CAT (purity > 99%) were prepared at the Institute of Materia Medica (IMM), Chinese Academy of Medical Sciences and Peking Union Medical College (CAMS&PUMC) [Beijing, China]. The structures were confirmed by 2-dimensional nuclear magnetic resonance, infrared, and Fourier transform mass spectrometry. Solvents for high-performance liquid chromatography (HPLC) were obtained from Thermo Fisher Scientific (Waltham, MA, USA). Hanks’ balanced salt solution (HBSS) was purchased from HyClone (GE Healthcare, city, UT USA). All other chemicals were of analytical grade.
2.1.1 Cell culture
Madin–Darby canine kidney (MDCK)-MDR1 cells expressing a high level of P-gp were generously provided by Professor Li (IMM, CAMS&PUMC). Cells were cultured in Dulbecco’s modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 1% penicillin/streptomycin (Gibco, Thermo Fisher Scientific, Grand Island, NY, USA) and 10% fetal bovine serum (Gibco) in 5% CO2 at 37°C.
2.1.2 Animals
Sprague–Dawley male rats (200–250 g) and Institute of Cancer Research (ICR) mice (18–20 g) were purchased from Beijing Huafukang Bioscience Co., Inc. (Beijing, China), and were raised at the IMM, CAMS&PUMC. The experimental protocol was approved by the Laboratory Animal Care and Use Committee of the PUMC on 11 April 2019, and the animal experiments lasted until 16 March 2020 (project identification code 5364). All animal experiments were performed in accordance with the Laboratory Animal Guidelines for Ethical Review of Animal Welfare (GB_T 35892.2018).
2.2. Methods
2.2.1 Prediction of the ADMET properties of 44 series compounds in silico
The 2-dimensional structures of phenanthroindolizidine compounds were generated using ChemDraw version 17 (PerkinElmer Inc., Waltham, MA, USA). Next, the main ADMET properties, including pK a, lipophilicity, permeability, solubility, blood-brain barrier activity, and P-gp transportation, were assessed using ADMET Predictor™ (version 8.0; Simulations Plus, Lancaster, CA, USA). The risks of absorption, metabolism, and toxicity were predicted in silico using the same software.
2.2.2 Determination of pK a, logP, and logD of CAT, CAT3, and PF403
CAT, CAT3, and PF403 were chosen as typical compounds to test the pK a, logP, and logD, and to validate the prediction results using ADMET Predictor™. The pK a was measured via ultraviolet spectrophotometry and the changes in the absorption spectra of the compounds at different pH values were calculated according to the Henderson–Hasselbalch equation [24–26]. The ionic and molecular states have different absorption coefficients at specific wavelengths and pH values. This approach was applied using equation (1):
where C u and C i are the molecular and ionic concentrations of the compounds, A is the absorbance of the compounds at the specified wavelength and pH value, A i is the absorbance in buffer (pH, 5.0) at the specified wavelength, which represents the total binding of the compounds with H+ (ion type), and A u is the absorbance of the buffer at the specified wavelength (pH 10.0), which represents the unionized (molecular) type. The pK a was calculated using equation (2):
Because the solubility of PAs in water is very low, it is difficult to conduct tests in pure water. Therefore, methanol was used as a co-solvent to test the pK a of PAs. In the binary mixed system of pure water and co-solvents, the pK a is considered to be the apparent dissociation constant (ps K a).
Methanol-water at concentrations (v/v) of 30%, 40%, and 50% were optimized as co-solvents and the ps K a was measured in a pH buffer solution (ionic strength, 0.1 mol·L−1) at different co-solvent ratios. Then, the Yasuda–Shedlovsky extrapolation procedure was used to obtain the pK a values of the phenanthroindolizidine compounds in aqueous solution.
By combining equations (1) and (2), we obtained equation (3):
Per equation (3), the ps K a at various pH values and wavelengths of compounds in 50%, 40%, and 30% (v/v) methanol as a co-solvent were calculated.
According to the Yasuda–Shedlovsky extrapolation method [27, 28], when the solution dielectric constant (D) [29, 30] is >50 F·m−1, the values of ps K a + log [H2O] and D at different co-solvent ratios and the reciprocal of ×1000 (1000/D) have a linear relationship, and the pKa value of compounds in pure water (D = 78.48 F·m−1 [30]) can be calculated using the least squares regression curve equation.
The specific experimental operations were as follows. First, the measurement wavelength was optimized by scanning the ultraviolet absorption (A) from 200−400 nm of 50% (v/v) methanol-buffer solution (pH 5.0) with a compound concentration of 5 mg·L−1 and a buffer solution with the same ratio at pH 10.0 and recording the spectra at the same time. Five wavelengths were chosen every 5 nm within 20 nm, with obvious changes in the A value as the measurement wavelengths.
In determining the ps K a50 in 50% (v/v) methanol-water solution, CAT, CAT3, and PF403 were dissolved in methanol and diluted to yield stock solutions with a concentration of 0.1 mg·mL−1. Methanol (4.5 mL) was added to 0.5 mL of stock solution and 5.0 mL of buffer solutions with pH values of 5.0, 7.0, 7.2, 7.4, 7.6, and 10.0 were added, mixed well, and used as the test solutions. The A, A i, and A u were measured at 260, 265, 270, 275, and 280 nm. We then calculated the ps K a50 using equation (3). The results are presented as the mean ± standard deviation (SD) for 20 experiments. We used a similar method to determine the psKa40 and ps K a30 in 40% or 30% (v/v) methanol-water systems.
The logP n-octanol and logD were determined via ultraviolet spectrophotometry. Specifically, an appropriate amount of CAT, CAT3, or PF403 was obtained and precision-weighed, then dissolved in n-octanol, which was saturated with water in advance, and ultrasonically dissolved to prepare a stock solution with a concentration of 0.1 mg·mL−1 for future use. Then, 0.5 mL of this stock solution was accurately removed, placed in a 10-mL centrifuge tube, and 5 mL of pure water saturated with n-octanol or buffer solutions with pH 1.2, 2, 3, 4, 5, 6, 7, 7.4, 8, 9, and 10, were added in advance. The buffer solutions were shaken carefully, placed in a 25°C constant temperature shaker, and shaken for 24 h at 150 rpm to complete diffusion and reach equilibrium. To separate the n-octanol layer from the water, the balanced system was allowed to stand and centrifuged at 3000 rpm for 10 min. Then, the n-octanol layer was diluted 20 times with methanol to meet the measuring concentration range of the spectrophotometer, and the absorbance was measured at 263 nm with methanol as the blank via ultraviolet spectrophotometry. The concentration of the compounds in the n-octanol layer were calculated using the external standard method. The apparent oil-water partition coefficient (P) was calculated using equation (4) and the logarithm with a base of 10 was calculated as logP n-octanol (in water) or logD (in series pH buffers).
where P is the apparent oil-water partition coefficient of CAT3, C is the initial concentration of compounds in n-octanol (0.1 mg·mL−1), C O is the concentration measured in the n-octanol phase (mg·mL−1), C w is the concentration (mg·mL−1) measured in the aqueous phase at distribution equilibrium, V O is the volume of the n-octanol layer (0.5 mL), and V W is the volume of the aqueous phase (5 mL).
LogP cyclohexane was determined using the same method in the equilibrium solution of cyclohexane and water.
2.2.3 Solubility of CAT, CAT3, and PF403
The solubilities of CAT, CAT3, and PF403 in water, simulated gastric fluid (SGF), and simulated intestinal fluid (SIF) were tested by placing 10 mg of the compounds in 200 mL of corresponding solutions and shaking for 24 h. The highest estimated effective human dose was 10 mg.
2.2.4 Transepithelial transport study
As described previously [12], the permeabilities of CAT, CAT3, and PF403 were investigated on an MDCK-MDR1 cell monolayer. MDCK-MDR1 cells were seeded at a density of 2 × 105 cells/well onto 12-well Transwell® filter inserts (pore size 0.4 μm, surface area 1.12 cm2; Corning Incorporated, Corning, NY, USA). The culture medium was changed every day for 7 days and cell monolayers with transepithelial electrical resistance (TEER) > 110 Ω·cm2 were used for the transport experiments. The culture medium was removed and the monolayer was pre-incubated with 0.5 mL of HBSS for 20 min at 37°C before performing the TEER test.
After measuring the TEER, we removed the HBSS and 0.5 mL of 100 ng/mL CAT, CAT3, or PF403 diluted with HBSS was added to the apical side (AP). Then, 1.5 mL of blank HBSS was added to the basolateral (BL) compartment as the adapted solvent. For P-gp inhibition studies under permeation with PF403, verapamil (100 μM) was added to the transport buffer on the apical and basolateral sides. The monolayer was incubated for 2 h at 37°C. Samples (50 μL) were obtained from the BL compartment every 30 min for 2 h. Next, the samples were stored at 4°C for analysis. The apparent permeability coefficient (P app) from the BL to AP side was determined by adding 1.5 mL samples to the BL side and 0.5 mL HBSS to the AP side.
The amount of CAT, CAT3, or PF403 permeated through the monolayer was determined using a liquid chromatography-tandem mass spectrometry (LC-MS/MS) system. The P app of CAT, CAT3, or PF403 was calculated according to the following equation: P app = dQ/dt × 1/(A × C 0), where dQ/dt indicates the linear rate of appearance of mass on the BL side (μmol/s), C 0 is the initial concentration of CAT, CAT3, or PF403 on the AP side (μmol/mL), and A is the surface area of the monolayer (cm2).
2.2.5 Determination of absolute bioavailability in vivo
As described previously [12], for pharmacokinetic studies, 10 rats were divided into 2 groups (n = 5 per group). The rats were fasted for 12 h prior to the day of administration, allowing water ad libitum. CAT3 (12 mg/kg) was administered orally in the first group, and 1 mg/kg was administered through the injection route (IV) in the second group. The drug content was pre-analyzed using spectrophotometric methods at 263 nm to ensure dose accuracy. Blood samples (approximately 150 μL) were collected into heparinized tubes from each rat through a retro-orbital sinus puncture, which was performed at various time points after administration. The blood samples were immediately processed to obtain plasma by centrifugation at 1500 rpm for 10 min. Plasma was collected and frozen at −80°C until LC-MS/MS analysis.
For in vivo analysis of CAT3 in rat plasma, 10 μL of the internal standard (ISTD; CAT, 10 ng/mL) was added to 50 μL of plasma. After adding 90 μL of acetonitrile for protein precipitation, the sample was vortex-mixed for 30 s, followed by centrifugation at 13,000 rpm for 10 min. The supernatant was transferred to a new tube and a 5-μL aliquot of the solution was injected into the LC-MS/MS mixture for analysis.
Triple-quadrupole mass spectroscopy (6410 B; Agilent, Santa Clara, CA, USA) was performed in the multiple reaction monitoring and electrospray ionization positive mode. Nitrogen (350°C) was used as the nebulizer gas at 40 psi and a flow rate of 10 L/min. The collision energy and voltage of the fragmentor potential were 20 and 135 eV, respectively. The ion reactions for the determination of CAT3 and ISTD were m/z 434.3→m/z 70.1 and 364.2→70.1, respectively [12].
The analyte concentrations were determined using the MassHunter software (Agilent). Chromatographic separation was performed on a 1200 series rapid resolution liquid chromatography system (Agilent) with a C18 column (50 mm × 2.1 mm, 1.8 μm; Agilent) and a corresponding guard column (octadecyl silica, 5 μm). Water and 0.1% formic acid acetonitrile (20:80, v/v) were used as the mobile phases for elution. The flow rate was 0.4 mL/min and the column temperature was 40°C.
The primary pharmacokinetic parameters were calculated based on the statistical moment method using Drug and Statistics (DAS) 3.0 software (BioGuider Medical Technology Co., Ltd., Shanghai, China).
2.2.6 Determination of brain-blood ratio of CAT3 in vivo
As described previously [31], to determine the brain:blood ratio of CAT3 in in vivo studies, 15 male rats were randomly divided according to 3 time points (n = 5) and CAT3 was administered orally at a dose of 10 mg/kg. Blood samples (approximately 150 μL) from each group were collected into heparinized tubes from each rat through a retro-orbital sinus puncture at predetermined time points (1, 4, and 8 h) before the rats were sacrificed and brain tissues were collected. The brain tissues were washed carefully immediately with normal saline and frozen at −80°C. Prior to HPLC-MS/MS, the brain samples were homogenized with normal saline (1:3). An aliquot (50 μL) of the homogenates was precipitated by adding 90 μL of acetonitrile and 10 μL of the ISTD, followed by vortexing for 30 s and centrifugation for 10 min at 13,000 rpm. The supernatant was transferred to a new tube and a 5-μL aliquot of the solution was injected into the LC-MS/MS system for analysis.
2.2.7 Preliminary toxicologic study of CAT3 in vivo
To assess the safety and adverse effects of CAT3, the body weight and stool status of the mice were recorded at different doses. Gastric and intestinal tissue samples of the mice were collected and analyzed at the end of the study. Fifteen male ICR mice were randomly divided into 3 groups (n = 5 each). Saline (control group) and CAT3 (15 and 25 mg/kg/day) were then administered orally for 15 days. Fifteen female ICR mice were randomly divided into 3 groups (n = 5 each) and treated with the same methods as the male mice for the same length of time. The body weight of each mouse was recorded every 2 days over the entire treatment period. The stool status was observed every day. At the end of the 16th day, the male mice were sacrificed via cervical dislocation and the entire stomach and intestine contents of each mouse was immediately collected, flushed gently with saline, cleaned of fat and mesentery on ice-cold plates, and patted dry using filter paper. Next, the washed stomach, jejunum, ileum, colon, and rectum samples were fixed for 48 h in 4% paraformaldehyde, washed in phosphate buffer (pH 6.4), and embedded in paraffin blocks. The 4-μm slices were sectioned and stained with hematoxylin and eosin (H&E) or Alcian-Blue periodic acid-Schiff (AB-PAS) for histologic evaluation.
2.2.8 Statistical analysis
The statistical difference between the treatments was evaluated via analysis of variance. Data are reported as the mean ± SD. The statistical analyses were performed using SPSS (version 17.0; SPSS Inc., Chicago, IL, USA). The statistical significance of any differences was evaluated using Duncan’s multiple range test with a cut-off P<0.05.
3. RESULTS
3.1 Prediction of ADMET properties of 44 compounds in silico
The pentacyclic structure, which is the basic molecular structure of PAs, is shown in Figure 1 . The ADMET properties predicted results of CAT, CAT3, and PF403 in silico are shown in Table 1 . The detailed data of the 44 compounds are listed in Figure 2 and Table S1 . The pK a values of PAs are related to the nitrogen atom at position 20, which is the main ionization group of this molecule and the most important group for drug efficacy. The pK a value of the nitrogen atom at position 20 varied between 4.19 and 8.92, which was clearly affected by the substituent at position 16. In the absence of a substituent at position 16, the pK a was between 8.39 and 8.92. In contrast, when oxygen was substituted at the same position, the pK a value decreased to approximately 8.02. Similarly, when replaced by a secondary nitrogen atom, the pK a decreased significantly to 4.19–5.57. However, when an amide group was placed at position 21, there was no pK a value because dissociation was difficult.
Predicted ADMET properties of 44 phenanthroindolizidine compounds determined using ADMET Predictor™.
(A) Formula weight, (B) ADMET_risk, (C) Absn_risk, (D) CYP_risk, (E) TOX_risk, (F) S+logP, (G) S+logD, (H) S+MDCK, (I) Perm_Skin, (J) S+Sw, (K) logBB.
Predicted results for CAT, CAT3, and PF403 determined using ADMET Predictor™.
| ADMET properties | Identifier | CAT | CAT3 | PF403 |
|---|---|---|---|---|
| Molecular weight | AP_FWeight | 363.459 | 433.551 | 349.432 |
| Drug-like properties | ADMET_Risk a | 5.22 | 6.673 | 3 |
| pK a | S+Basic_pK a | 8.92 | 8.66 | 8.64 |
| Lipophilicity | S+logP | 4.821 | 5.696 | 4.26 |
| S+logD | 3.29 | 4.412 | 3.001 | |
| Permeability | S+MDCK (cm·s−1 × 107) | 772.382 | 784.235 | 583.274 |
| Perm_Skin (cm·s−1 × 107) | 6.307 | 19.741 | 1.974 | |
| Solubility | S+S w (mg/mL) | 0.019 | 0.007 | 0.097 |
| S+FaSSGF (mg/mL) | 0.859 | 0.293 | 1.531 | |
| S+FaSSIF (mg/mL) | 0.087 | 0.033 | 0.213 | |
| S+FeSSIF (mg/mL) | 0.215 | 0.093 | 0.396 | |
| BBB properties | BBB_Filter | High | High | High |
| LogBB | 0.924 | 0.952 | 0.758 | |
| Transporters | P -gp_Substr | Yes (82%) | Yes (97%) | Yes (86%) |
| P -gp_Inh | No (67%) | Yes (70%) | No (94%) | |
| Oral absorption properties | Absn_Risk b | 0.321 | 1.545 | 0 |
| Absn_Code | Kow | Size, Kow, Sw | / | |
| Fa, rat | 99.98% | 98.99% | 98.94% | |
| Fb, rat | 54.29% | 61.94% | 63.24% | |
| CYP substrate | CYP1A2_Substr | Yes (80%) | Yes (63%) | Yes (58%) |
| CYP2B6_Substr | Yes (79%) | Yes (79%) | Yes (52%) | |
| CYP2C8_Substr | Yes (50%) | Yes (54%) | No (98%) | |
| CYP2C9_Substr | Yes (53%) | No (77%) | Yes (53%) | |
| CYP2C19_Substr | Yes (50%) | No (72%) | No (68%) | |
| CYP2D6_Substr | Yes (81%) | Yes (81%) | Yes (81%) | |
| CYP3A4_Substr | Yes (98%) | Yes (98%) | Yes (98%) | |
| UGT1A1 | No (66%) | Yes (63%) | Yes (60%) | |
| UGT1A3 | No (47%) | No (51%) | Yes (53%) | |
| UGT1A9 | No (91%) | No (91%) | No (65%) | |
| Pharmacokinetic properties | CYP_Risk c | 1.899 | 1.858 | 1 |
| CYP_Code | 1A2, 2D6, CL | 1A2, CL | 1A2 | |
| Toxicity properties | TOX_Risk d | 3 | 2.634 | 2 |
| TOX_Code | hERG, Xm, MUT | hERG, Xm, MUT | hERG, MUT |
Note [32]: aThe best molecule has a value <7.5.
bThe best molecule has a value <3.5.
cThe best molecule has a value <2.5.
dThe best molecule has a value <2.0.
Moreover, all 44 PAs were highly lipophilic compounds, the estimated logP values of which were between 2.5 and 5.85 ( Figure 2F ). The logD values changed after the nitrogen atom in the PA molecular structure ionized in different pH environments. Therefore, the logD values at pH 7.4 were all lower than the logP values ( Figure 2G ). PF403 had the strongest antitumor activity in vitro among the 44 PAs. However, PF403 was not the best compound with good activity in vivo [5]. After the 4-hydroxyl group of PF403 was substituted in CAT or CAT3, the logP values increased significantly compared to PF403 and a higher in vivo antitumor activity was obtained after oral administration [5, 12]. Therefore, it appears that a higher lipophilicity results in higher in vivo activity. Furthermore, the predicted aqueous solubility of the 44 PAs was extremely low (0.001–0.151 mg/mL).
The predicted blood-brain barrier (BBB) permeability properties of the 44 PAs were high; therefore, the PAs easily penetrated the BBB ( Tables 1 and S1, Figure 2K ). There were 29 of 44 PAs with a decimal logarithm of the brain-blood partition coefficient (logBB) >0.7; therefore, the concentration in the brain would be >5-fold higher than the blood. Moreover, 43 of 44 PAs had a logBB >0, indicating that the concentration in the brain was higher than that the blood.
The predicted permeability (P app) in MDCK cell monolayer models of the 44 PAs were between 260 and 1159 ( Tables 1 and S1, Figure 2H ), indicating that the 44 PAs had medium-to-higher permeability when administered orally.
All 44 PAs were predicted to be P -gp substrates ( Tables 1 and S1 ). However, only one-half of the PAs (19 of 44) were P -gp inhibitors. This finding may explain why CAT3 has better anti-glioma activity than PF403 [31].
Oral absorption properties, named as the absorption risk or Absn_Risk, were calculated based on eight physicochemical parameters, such as molecular size and charge, rotatable bonds, hydrogen bound donors and receptors, lipophilicity, permeability, and solubility in water [32]. Typically, molecules with an absorption risk score <3.5 are considered potential candidates [33]. The Absn_Risk values of all 44 PAs were 0−3.0; therefore, the PAs can be absorbed after oral administration ( Tables 1 and S1, Figure 2C ). However, there were still several risk factors that affect absorption, such as larger molecular size, low water solubility, and higher lipophilicity, which were marked as size, K ow, and S w in the column labeled Absn_Code. The absorbed ability (Fa) and absolute bioavailability (Fb) of CAT3 in rats were 98.99% and 61.94%, respectively; therefore, the bio-absorption was good.
Pharmacokinetic properties were characterized by the CYP_Risk model, which was comprised of the metabolism of the eight main cytochrome P450 proteins. Typically, a CYP_Risk score <2.5 is considered indicative of a potential candidate. The CYP_Risk values of the 44 PAs were 0−2.4; therefore, all PAs had acceptable metabolite properties after oral administration ( Tables 1 and S1, Figure 2D ). The high-risk items were CYP 1A2, 2D6, and 3A4 and human liver microsomes, respectively.
The toxicity properties were characterized using the TOX_Risk model, which comprises human Ether-a-go-go related gene (hERG), acute toxicity in rats, and carcinogenicity in chronic rat or mouse studies. Typically, molecules with a TOX_Risk score <2.0 are considered safe. The TOX_Risk values of the 44 PAs were 1−3.0 and 36/44 of the PAs were higher than the threshold ( Tables 1 and S1, Figure 2E ). The main risk factors of PA were hERG status, a TD50 of carcinogenicity in chronic mouse studies (Mouse_TD50 , Xm), and mutagenicity (MUT). This may be related to the outstanding anticancer activities of PAs by destroying the DNA structure and interfering with the DNA function of cancer cells [3].
Drug-like properties were characterized by ADMET_Risk, which comprised not only Absn_Risk, CYP_Risk, and TOX_Risk, but also the fraction unbound to protein in human plasma and human volume of distribution in the steady state. Typically, an ADMET_Risk score <7.5 is considered indicative of a potential candidate [33]. The ADMET_Risk values of the 44 PAs were 1−9, and 32 of 44 PAs had values <7.5, indicating that most of the PAs had good drug-like properties ( Tables 1 and S1, Figure 2B ). Considering the drug-like properties of the 44 PAs, 3 compounds (CAT, CAT3, and FP403) were the most promising candidate compounds. Therefore, these three compounds were studied in detail in subsequent experiments to verify the accuracy of the predicted values.
3.2 Dissociation constant (pK a), logP, and logD of CAT, CAT3, and PF403
3.2.1 Determination of pK a using ultraviolet spectrophotometry
PAs have similar ultraviolet absorption characteristics because PAs have the same phenanthroindolizidine-fused aromatic ring. As shown in Figure 3A , the maximum absorption peaks were observed at 263 and 285 nm.
(A) Ultraviolet spectrum of CAT3 in 50% (v/v) methanol-water in buffers at pH 5.0 and 10.0. (B) ps K a+log[H2O] of CAT, CAT3, and PF403 versus 1000/D.
The results showed that the absorption spectrum of CAT3 with different ionization levels was clearly different in the wavelength range of 260–280 nm. As the pH value increased, the absorbance decreased significantly and met the ps K a measurement requirements. Then, according to equations 1−3, the ps K a50, ps K a40, and ps K a30 of CAT3 were calculated from the absorbance values of CAT3 at various wavelengths under different pH values. The results are shown in Table 2 .
ps K a values of CAT, CAT3, and PF403 in serial methanol solutions.
| Methanol (%, v/v) | Dielectric constant (D, F·m−1) [30] | 1000/D | ps
K
a
| ||
|---|---|---|---|---|---|
| CAT | CAT3 | PF403 | |||
| 0 | 78.48 | 12.74 | 7.91 | 7.30 | 7.67 |
| 30 | 65.55 | 15.26 | 7.63 | 7.23 | 7.50 |
| 40 | 60.94 | 16.41 | 7.54 | 7.19 | 7.35 |
| 50 | 56.28 | 17.77 | 7.37 | 7.16 | 7.30 |
We performed linear regression analyses with ps K a+log[H2O] values in different concentrations of methanol and 1000/D. The regression curve is shown in Figure 3B .
The D value of pure water is 78.48 F·m−1, and we calculated that in pure water the ps K a+log[H2O] values of CAT, CAT3, and PF403 were 9.05, 9.42, and 9.66, respectively. Thus, the pK a values were 7.91, 7.30, and 7.67, respectively.
3.2.2 Determination of logP n-octanol and logP cyclohexane
The maximum absorption wavelength of CAT3 in methanol solution was 263 nm and the absorbance values were 0.2−0.8 when the concentration was between 2 and 7 mg·L−1. There was a significant linear relationship between the concentration and absorbance, and the absorption coefficient was stable and reliable. The logP determination results for CAT, CAT3, and PF403 in n-octanol-water and cyclohexane-water are shown in Table 3 . ΔlogP represents the difference between the logP n-octanol and logP cyclohexane.
LogP values of CAT, CAT3, and PF403 in n-octanol-water and cyclohexane-water.
| logP | CAT | CAT3 | PF403 |
|---|---|---|---|
| n-octanol-water | 3.68 | 4.34 | 3.26 |
| Cyclohexane-water | 2.35 | 2.96 | 2.07 |
| ΔlogP | 1.33 | 1.38 | 1.19 |
CAT, CAT3, and PF403 had higher logP n-octanol values of up to 3.68, 4.34, and 3.26, respectively. The logP cyclohexane values were 2.35, 2.96, and 2.07, respectively. The ΔlogP values were all >1.0.
3.2.3 Determination of logD at serial pH values
Figure 4 shows the logD values of CAT, CAT3, and PF403 at different pH values.
The logD values of CAT and CAT3 both increased significantly with increasing pH ( Figure 4 ). PF403 showed a similar trend; however, after the pH exceeded 9, the logD value decreased. Among the three compounds, the logD value of CAT3 was significantly higher than the other two, indicating that it had a greater lipophilicity. The logD values of CAT, CAT3, and PF403 were 2.57, 3.86, and 2.39, respectively, at pH 7.4 ( Figure 4 ).
3.3 Solubility of CAT, CAT3, and PF403
The solubility of CAT, CAT3, and PF403 in water, SGF, and SIF were tested by placing 10 mg in 200 mL of the corresponding solutions. The compounds were recognized as insoluble if the samples were not dissolved in the solution for 24 h. The results showed that none of the PAs could be dissolved completely in any solution. Thus, CAT, CAT3, and PF403 were insoluble compounds in water, SGF, and SIF.
3.4 Bidirectional transport in MDCK cell monolayer
As described previously [31], the transepithelial transport of CAT, CAT3, and PF403 across the MDCK-MDR1 cell monolayer was determined using LC-MS/MS. The results are summarized in Table 4 . CAT, CAT3, and PF403 exhibited a very low P app(AP-BL) [(0.1272±0.02352) × 10−6, (0.1139±0.01827) × 10−6, and (0.1880±0.03736) × 10−6 cm/s, respectively] in the MDCK-MDR1 cell model at 100 ng/mL. Neither CAT nor CAT3 showed an observable efflux profile; the efflux ratios (ERs) were 1.04 and 1.09, respectively. However, PF403 exhibited strong efflux in vitro, with an ER of 2.991. Moreover, when incubated with verapamil, the ER decreased to 1.56, indicating that PF403 may be the substrate of P -gp.
3.5 Determination of absolute bioavailability in vivo
The plasma concentration-time curve after the oral administration of CAT3 at a dose of 12 mg/kg or after intravenous injection at a dose of 1 mg/kg is shown in Figure 5 and the main pharmacokinetic parameters are listed in Table 5 . The AUC0∼t, Cmax, and MRT0∼t of CAT3 in the group treated with oral administration at a dose of 12 mg/kg were 25.98±6.23 h·ng/mL, 1.2±0.19 ng·mL−1, and 43.71 h, respectively. The respective values in the group treated with intravenous injection at a dose of 1 mg/kg were 43.68±4.28 h·ng/mL, 256.43±26.68 ng·mL−1, and 0.26±0.1 h.
Pharmacokinetic parameters of CAT3 in mice after oral or intravenous injection of CAT3 at a dose of 12 or 1 mg·kg−1 (n = 5).
| Parameters | Unit | p.o. 12 mg·kg−1 | i.v. 1 mg·kg−1 |
|---|---|---|---|
| AUC(0−t) | h·ng·mL−1 | 25.98±6.23 | 43.68±4.28 |
| AUC(0−∞) | h·ng·mL−1 | 44.19±4.22 | 43.87±4.22 |
| MRT(0−t) | H | 43.71±0.1 | 0.26±0.1 |
| MRT(0−∞) | H | 43.38±0.11 | 0.27±0.11 |
| t1/2 | H | 28.4±0.17 | 0.42±0.17 |
| Tmax | H | 4.5±4.44 | 0.04±0.02 |
| Cmax | ng·mL−1 | 1.2±0.19 | 256.43±26.68 |
| V | L·kg−1 | 14403.76±4747.53 | 13.57±4.96 |
| CL | L·h−1·kg−1 | 136.79±187.79 | 22.97±2.3 |
The absolute bioavailability of CAT3 was determined as the ratio of the dose-normalized area under the curve (AUC) after oral administration to intravenous administration, and the result was 4.96%.
3.6 Determination of the brain-blood ratio of CAT3 in vivo
Plasma and brain tissue concentrations among rats after oral administration of CAT3 were determined by HPLC-MS/MS. The drug concentrations in the brain tissue were 2.257±0.415, 0.602±0.370, and 0.156±0.117 ng/mL at 1, 4, and 8 h, respectively, after administration ( Figure 6 ). The plasma concentrations were 2.482±0.218, 0.312±0.039, and 0.145±0.015 ng/mL at the corresponding time points. Therefore, the ratio of the CAT3 concentration between blood tissue and plasma was 0.909, 1.927, and 1.081 at 1, 4, and 8 h, respectively, and the logarithmic values (logBB) were −0.0414, 0.2850, and 0.0340, respectively. Therefore, CAT3 crosses the BBB readily and quickly distributes from the blood to the brain tissue [34]. Moreover, the CAT3 concentration in the brain was close to or significantly higher than plasma.
3.7 Preliminary toxicologic study of CAT3 in vivo
The body weight and stool status of the mice after oral administration of CAT3 were observed. The body weight of animals receiving oral administration of CAT3 was significantly lower than the control group in a dose-dependent manner ( Figure 7 ). Furthermore, the sensitivity of body weight to CAT3 was related to the sex of mice. In short, at the same dosage, the body weight of male mice was more sensitive to CAT3 than female mice and showed a greater decrease. The mice in the control group had brown granular stool and no diarrhea from the beginning to the end of the experiment, while the mice in the CAT3 groups experienced diarrhea 3−16 days after administration, and gradually increased, with light yellow soft stools or light yellow loose stools. Severe diarrhea was more commonly observed in the higher dose group than in the lower dose group.
The toxicity of the gastrointestinal organs to CAT3 could be more clearly observed in pathologic sections. H&E staining showed that the general structure of the gastrointestinal tissue was not significantly altered in the control group, and the intestinal villi and muscular layer were generally normal ( Figure 8 ). However, inflammatory cell infiltration and bleeding spots were observed in the stomachs of the animals after administration of CAT3, indicating that it had a stimulating effect on the stomach. The pathologic results of the colon showed obvious infiltration of inflammatory cells accompanied by a significant increase in the connective tissue gap in the CAT3 group. More importantly, compared with other gastrointestinal sites, the infiltration of inflammatory cells in the rectal area of animals in the CAT3 group was more severe. The above stimulatory effects all showed an obvious dose dependence. However, there was no obvious infiltration of inflammatory cells in the jejunum or ileum.
Histopathologic features of gastrointestinal organs after oral administration of CAT3 (H&E stain ×100). Black bar: 100 μm.
The goblet cells of the small intestine were stained with AB-PAS ( Figure 9 ). The results showed that oral administration of CAT3 had no significant effects on the jejunum and ileum. However, the number of goblet cells and the amount of the secretions decreased significantly in the colon, showing a typical dose-dependent phenomenon in the late stage of the inflammatory response. In the rectum, the number and amount of secretion of goblet cells in the high-dose group decreased; however, the low-dose group showed no significant effect.
Histopathologic features of gastrointestinal organs after oral administration of CAT3 (AB-PAS stain ×100). Black bar: 100 μm.
Therefore, the main organs that showed toxicity to CAT3 were the rectum and colon, and the infiltration of inflammatory cells may be one of the causes of diarrhea.
4. DISCUSSION
A large number of compounds are produced during the research and development of novel PA drugs [4]. It is not appropriate to use the activity in vitro as the only basis for selection; drug-like properties are also very important. Early information about drug-like properties can help teams make more accurate decisions and avoid wasting precious resources on candidates that are unlikely to pass clinical trials [35]. Thus, in parallel with activity screening, high-throughput ADMET assays have been implemented and are being widely used to drive drug discovery projects [17]. To gain a deeper understanding of the structural properties of PAs, a series of key ADMET characteristics need to be predicted and assayed, such as solubility, pK a, log P, permeability, bioavailability, brain:blood ratio, and toxicity. Many commercial or free software are available for the prediction of drug-like properties [36]. Among the software, ADMET Predictor™ provides the best physicochemical (pK a [37], logP, logD, and aqueous solubility [38]) and biological (permeability of the MDCK monolayer or human models, BBB, plasma protein binding, and bioavailability) properties. ADMET Predictor™ includes built-in models based on statistical methodology and ADMET Predictor™ is also capable of predicting toxicologic endpoints, such as chronic toxicity, genotoxicity, carcinogenicity, endocrine disruption, cardiac toxicity, phospholipidosis, hepatotoxicity, and ecotoxicity.
As a dissociation parameter, the pK a reflects the relative ratio of ionic and molecular types under the influence of a physiologic pH environment [39–41] and represents the mutual conversion of solubility and permeability. Similarly, the oil-water partition coefficient, the logarithmic value of which is referred to as logP or logD, reflects the relative characteristics of drug hydrophilicity and lipophilicity. The oil-water partition coefficient is the ratio of the drug distribution concentration in the n-octanol oil phase to that in the water phase. Oral drugs with suitable logP values will have perfect lipid solubility across biofilms and have good hydrophilicity to be transported smoothly in body fluids [42]. Therefore, the pK a and logP are considered to be the most important parameters that affect gastrointestinal absorption, bioactivity, and toxicity [43].
The nitrogen atom at position 20 of the 44 serial compounds, as the main ionizing group, exists in the same position as the phenanthroindolizidine-fused aromatic ring and has similar pK a values. However, owing to the different substitution groups, the hydrophilic and lipophilic properties of the 44 compounds are quite different. PAs have extremely low solubility in water and cannot be determined by commonly used titration methods. An ultraviolet method with high accuracy and good sensitivity was established to determine the pK a of three typical compounds. The results showed that the measured pK a values were approximately 7.6 and the logP values ranged from 3.26–4.34, which were close to the predicted values. This proves the accuracy of the ADMET Predictor™, the usefulness for rapid screening of PA candidates, and optimization of the molecular structures in the future. In addition, PA has suitable pK a values; therefore, under normal physiologic conditions, a proper ratio of ionized and non-ionized species can exist in the small intestine. ADMET Predictor™ considers the solubility and permeability, which are beneficial for biological absorption.
However, the high logP and logD values of PAs are not conducive to dissolution in water. Therefore, an effective dose of approximately 10 mg is difficult to dissolve in 200 mL of aqueous medium, which has become the main factor limiting biological absorption. In addition, it was reported that the apparent partition coefficient of the non-ionized forms is expected to be at least three orders of magnitude higher than the ionized forms; therefore, logD provides more important estimates of bioavailability than logP at a given pH [44]. Due to the higher logP and logD 7.4 values of CAT3, which were 4.34 and 3.86, respectively, the absolute bioavailability was only 4.96% in our study. Therefore, not only CAT3, but also other PAs belong to the Biopharmaceutical Classification System (BCS) class IV; therefore, the PAs are very poorly soluble and poorly permeating compounds.
The results of permeability measurement using the MDCK-MDR1 cell monolayer model showed that the P app values of CAT, CAT3, and PF403 were all <2 × 10−6 cm/s. Thus, the PAs are low-permeability compounds. However, the S+MDCK values were approximately 6−8 × 10−6 cm/s when determined using ADMET Predictor™. The S+MDCK was significantly higher than the measured values. A possible reason may be that the extremely high lipophilicity of PAs makes it easier to combine with the cell membrane in the in vitro permeation model; therefore, the measured P app was lower than the predicted values. Notably, the predicted S+MDCK can only be used as a reference and PAs still belongs to BCS class IV.
The predicted results showed that all PA compounds were P -gp substrates. In addition, 19 of 44 PAs are P -gp inhibitors, including CAT and CAT3 but not PF403. CAT and CAT3 are substrates of P -gp and its inhibitors, which may explain why CAT and CAT3 did not show ER in the in vitro MDCK-MDR1 cell monolayer model. However, PF403 is only a substrate of P -gp and not an inhibitor. Therefore, this may explain why PF403 is easily effluxed by P -gp and has a higher ER. It has been reported that PF403 cannot be absorbed orally, which may be related to the above explanation. For PA compounds, when the above two indicators of whether they are P -gp substrates and whether they are P -gp inhibitors are analyzed simultaneously, the permeability prediction accuracy of the compound will be better.
In addition, the gap in logP values between n-octanol and cyclohexane solvents reflect the distribution ratio of the drug in the brain and blood [45, 46]. The brain:blood ratio has great significance for drugs that target brain tissue to treat CNS diseases. The results of the present study showed that in the tissue distribution experiment among rats, the brain concentration of CAT3 was significantly higher than plasma. The logBB prediction results of 44 PAs also confirmed that the PAs easily penetrated the BBB. Therefore, the application prospects of PAs as CNS therapeutic agents are very broad [10].
A comparison of the ADMET Predictor™ results with CAT3 have been reported [12] and showed that ADMET Predictor™ can predict CYP 1A2, 3A4, 2B6, 2C8, and 2D6 correctly but predicts CYP 2C9 and 2C19 incorrectly. ADMET Predictor™ predicts UGT 1A1 correctly but UGT 1A3 and 1A9 incorrectly. In general, the accuracy of the prediction results is relatively high and can be used as a reference for other PA compounds.
The toxic and adverse effects of CAT3 and PA compounds mainly affect the gastrointestinal tract, and manifest as diarrhea and loose stools [5, 31]; however, no detailed research results have been reported. The results of H&E staining and AB-PAS staining showed that the gastrointestinal toxicity of CAT3 mainly occurred in the rectum and colon, and showed inflammatory cell infiltration. This may be related to the higher pH value in the rectum and colon and because PAs are in an undissociated molecular state under these conditions, PAs have a higher logD value, which will be more conducive to accumulation in the tissues, thereby increasing damage to the rectum and colon. Moreover, according to the ADMET Predictor™ results, the risk of PA compound toxicity may be attributable to hERG, MUT, and Xm, which require further experimental verification. In summary, the above-mentioned toxicity factors should be considered when selecting PA candidates. More choices may arise from developing drug delivery systems (DDSs) to avoid or reduce related gastrointestinal adverse effects [31]. DDSs have been widely used to protect drugs against conjugation and metabolic inactivation, as well as to enhance aqueous solubility and hence to ameliorate the oral bioavailability of sparingly soluble drug molecules. It has been reported that the solid lipid nanoparticles and self-microemulsifying drug delivery system were used to delivery CAT3 to solve the above problems [13, 32].
Drug-like agents play an important role in guiding the rational selection of pharmaceutical preparations. PA compounds have a suitable pK a that facilitates dissolution and bio-absorption and a high logP value that facilitates BBB penetration. PAs also have a high brain:blood concentration ratio and can be taken orally to reach the brain tissue as a unique feature, making PAs suitable for the treatment of neurologic diseases. In addition, the skin permeability prediction results (Perm_skin) of ADMET Predictor™ showed that PA has a strong ability to penetrate the skin and can be considered for administration through the skin to reduce the gastrointestinal adverse effects caused by oral administration. Next, it is important to pay attention to the biological absorption of PAs and to apply DDSs to promote absorption [31, 47].
5. CONCLUSIONS
PAs are characterized by low solubility and poor permeability. PAs are BCS class IV compounds with low absolute bioavailability. These compounds have high lipophilicity, distribute easily and fast into the brain after oral administration, and can be used to treat brain tumors and other diseases. However, PAs also show obvious gastrointestinal toxicity; specifically, the colon and rectum are main organ targets of toxicity. ADMET Predictor™ predicts physicochemical and biological properties quickly and accurately for PAs, which is suitable as a reference in the discovery stage of candidate compounds.
ABBREVIATIONS
PAs, phenanthroindolizidine alkaloids; CAT, (+)-Deoxytylophorinine; CAT3, 13a-(S)-3-pivaloyloxyl-6,7-dimethoxyphenanthro(9,10-b)-indolizidine; PF403, 13a-(S)-3-hydroxyl-6,7-dimethoxyphenanthro(9,10-b)-indolizidine; ADMET, absorption, distributeon, metabolism, excretion and toxicity; CNS, central nervous system; BBB, blood-brain-barrier; MDCK, Madin-Darby canine kidney cells; SGF, simulated gastric fluid; SIF, simulated intestinal fluid; D, dielectric constant; pKa, dissociation constant; psKa, the pKa in co-solvents; PK, pharmacokinetic; P, apparent oil-water partition coefficient; logP, partition coefficient, logarithm of P in water with a base of 10; logD, distribution coefficient, logarithm of P in series pH buffers with a base of 10; TEER, transepithelial electrical resistance; HBSS, Hanks’ balanced salt solution; Papp, apparent permeability coefficient; AP, apical side; BL, basolateral; LC-MS/MS, liquid chromatography-mass spectrometry; H&E, hematoxylin and eosin; AB-PAS, Alcian-Blue Periodic acid-Schiff; P-gp, Pglycoprotein; DDS, drug delivery systems; hERG, human Ether-a-go-go Related Gene; Xm, TD50 of carcinogenicity in chronic mouse studies; MUT, mutagenicity.
