STZ was purchased from Sigma-Aldrich (Shanghai, China). KM [purity >99.5%, as determined by high-performance liquid chromatography (HPLC)] and gelsemine (GM) (purity >92.8%, as determined by HPLC) –used as internal standards in PK studies–were isolated from Gelsemium elegans according to our previously reported method (Su et al., 2011). HPLC-grade methanol was purchased from Merck (Darmstadt, Germany), and analytical grade formic acid was obtained from Anaqua™ Chemicals Supply (Houston, TX, USA). Deionized water was produced using a Millipore Direct-Q® water purification system (Billerica, MA, USA). KM was dissolved in 0.9% sterile saline (NS) and orally administered to rats at doses (0.28, 1.4, and 7.0 mg kg⁻¹ for KM) that previously exhibited anti-allodynic activity in a STZ-induced diabetic model of neuropathic pain (Ling et al., 2014). All other chemicals were of analytical grade.

Male Sprague–Dawley (SD) rats (6–8 weeks old; 180–200 g; specific pathogen-free) were obtained from Charles River (Zhejiang, China). The animals, housed in clear plastic cages, were acclimatized for at least 5 days before the experiments. The specific pathogen-free environmental conditions were maintained as follows: 12/12 h light/dark cycle, 40–70% relative humidity, and 20–26°C. The rats were provided free access to water and food ad libitum, but fasted overnight before drug administration with free access to water. All experimental procedures and protocols were reviewed and approved by the Animal Ethics Committee of Fujian Medical University (permit number: 2017-01). All experiments in this study were performed with the aim of minimizing animal suffering. All methods and results were reported according to the ARRIVE guidelines (Kilkenny et al., 2010).

The MWT test protocol was established according to previous studies (Ling et al., 2014). Before the test, SD rats were placed in a transparent organic glass box (approximately 30 × 20 × 20 cm) with a glass plate on top and metal mesh at the bottom and habituated for 30 min before the measurement. Then, the MWT was assessed using an electronic von Frey pain measurement instrument (Model 2391; IITC Life Science, Woodland Hills, CA, United States) using the central plantar surface of the right hind paw. The duration of each stimulation was <1 s, and the stimulation interval was 30 s. The stimulation force (maximum: 55 g) was automatically recorded using the pain meter when the rats shrank, lifted, or licked their paws. The MWT was calculated as the mean of triplicate measurements of minimal force required to observe a response. The mean MWT of triplicate evaluations (once a day for three consecutive days) prior to STZ injection was defined as the baseline MWT.

The STZ-induced diabetic rat model was established according to a previous study with minor modifications (Sun et al., 2012; Ling et al., 2014). Briefly, overnight-fasted SD rats (198–235 g body weight) were intraperitoneally injected with a single dose of STZ (70 mg kg⁻¹); age-matched naïve rats were administered vehicle (sterile 0.9% NS). Blood samples collected from the tail vein were tested using a portable glucose meter (maximum: 33 mmol L⁻¹; One-Touch Ultra, Life Scan, Shanghai, China). Rats with fasting blood glucose levels above 16.6 mmol L⁻¹ (300 mg dl⁻¹) were considered diabetic and included in the study. Fasting blood glucose level and body weight were checked every week for 21 days after STZ injection. The fasting blood glucose level prior to STZ injection was defined as the baseline blood glucose. Twenty one days after intraperitoneal STZ administration, 42 diabetic rats with a >20% decline in MWT compared with the baseline value were classified as rats with DNP, of which 24 rats were used in the PD study and 18 rats were used in the PK study on day 22.

Twenty-four STZ-induced DNP rats were evenly assigned to four groups: the vehicle-treated group (vehicle), and low-dose, medium-dose, and high-dose KM groups. Six aged-matched naïve SD rats comprised the naïve group (naïve) and received oral gavage of NS. Rats in the low-dose, medium-dose, and high-dose KM groups received 0.28, 1.4, and 7.0 mg kg⁻¹ KM, respectively, by oral gavage. The MWT was measured at 0, 0.083, 0.167, 0.333, 0.667, 1, 2, 4, 6, 8, and 12 h after administration. Curves of the threshold for paw withdrawal over time were plotted. Thereafter, the area under the curve (AUC) of MWT against time was calculated.

Eighteen STZ-induced DNP rats were randomly assigned to three groups (n = 6 per group); they received a single dose of 0.28, 1.4, or 7.0 mg kg⁻¹ KM by oral gavage. Eighteen body-weight-matched naïve SD rats were used as controls and randomly assigned to three groups (n = 6 per group) and received a single dose of 0.28, 1.4, or 7.0 mg kg⁻¹ KM by oral gavage. The rats were anesthetized with 3% isoflurane (RWD Life Science, Shenzhen, China) before blood collection. Approximately 0.3 ml of blood samples was collected from the postorbital venous plexus of anesthetized rats in heparinized tubes at 0.083, 0.167, 0.333, 0.667, 1, 2, 4, 6, 8, and 12 h after the administration of KM. The samples were immediately centrifuged at 2,000 × g for 5 min to obtain plasma, which was stored at −80°C until further analysis.

The preparation and quantification of KM in the plasma were performed as previously reported (Ye et al., 2020). In brief, 50 μL of plasma was spiked with GM at a final concentration of 80 ng ml⁻¹ and evaporated to dryness. The dry sample was mixed with 400 μL of ethyl acetate by vortexing and then centrifuged at 12,000 × g for 10 min at 4°C. The supernatant was carefully transferred into fresh tubes and evaporated to dryness. The residues were redissolved with 100 μL of 50% methanol solution, of which 5 μL was injected for ultra-performance liquid chromatography (UPLC)–tandem mass spectrometry (MS/MS) analysis. The assay was performed on an Agilent 1290 Infinity UPLC^® system (Agilent Technologies, Santa Clara, CA, United States). The mobile phase comprised methanol and water with 0.1% formic acid. The flow rate was 0.4 ml min⁻¹ at 40°C. The gradient program is shown in SupplementaryTable S1. Only the eluent between 2.0 and 5.0 min was passed into the MS/MS system, operated in the positive ionization mode. The multiple reaction monitoring transition of KM was at m/z 307.2/180.1 and GM was at m/z 323.1/70.1.

PK data were first estimated using a non-compartmental analysis with Phoenix® WinNonLin™ 8.2 software (Pharsight Corporation, Certara, CA, United States). The time (T_max) to maximum observed plasma concentration was determined directly from the observed values. The AUC from time zero to infinity (AUC_0-∞) was calculated using the log-linear trapezoidal method. The t_1/2 was calculated as ln (2)/(slope of the terminal log-linear phase). Systemic clearance (CL) was calculated as dose/AUC_0-∞. The apparent volume of distribution (V_d) was determined as CL/(slope of the terminal log-linear phase) (Kim and Baek, 2018; Ye et al., 2020).

The data are presented as mean ± standard error of the mean (SEM) unless otherwise stated, and plotted and analyzed using Prism 8.0 (GraphPad Software, San Diego, CA, USA). All normally distributed data were statistically analyzed with parametric tests, including Student’s t-test and one-way or two-way analysis of variance (analysis of variance, repeated measures) followed by Tukey–Sidak post hoc analysis. All non-normally distributed data were analyzed with nonparametric statistics, including Friedman analysis of variance test, followed by Mann–Whitney U test for post hoc comparison. Unless otherwise specified, the significance level was set at p < 0.05.

Intraperitoneally injected STZ considerably elevated the blood glucose level (all >16.6 mmol L⁻¹) from 3 to 22 days after administration (p < 0.001, Figure 1A). STZ-induced diabetic rats exhibited a >20% reduction in MWT starting from day 7 and further decreased till day 21 after STZ injection (p < 0.001, Figure 1B). Furthermore, compared with naïve rats, STZ-induced diabetic rats exhibited a significant loss of body weight from 7 to 22 days (p < 0.001, Figure 1C). However, vehicle-treated rats maintained their baseline blood glucose level and MWT along with approximate 100 g body weight gain during the 22 days (Figures 1A–C).

Each dose of orally administered KM (0.28, 1.4, and 7.0 mg kg⁻¹) significantly increased the MWT compared with the vehicle treatment in STZ-induced DNP rats (all p < 0.001, Figure 2A). Furthermore, the effect of 7.0 mg kg⁻¹ KM was significantly stronger than that of KM at 0.28 mg kg⁻¹ (p < 0.05, Figure 2A). The AUCs for MWT after the oral administration of KM (0.28, 1.4, and 7.0 mg kg⁻¹) were significantly higher than after vehicle treatment in STZ-induced DNP rats (all p < 0.001, Figure 2B). The AUC for MWT after 7.0 mg kg⁻¹ KM administration was 1.34- and 2.88-fold higher than that after 0.28 mg kg⁻¹ KM and vehicle administration, respectively (147.4 ± 8.0 vs. 109.8 ± 5.3 and 51.2 ± 8.1, respectively).

The full-scan precursor ions and products of KM and GM are shown in Supplementary Figures S1A,B. The representative chromatograms presented in Supplementary Figures S1C–F are for blank rat plasma, blank rat plasma spiked with KM (final concentration: 20.0 ng ml⁻¹) and GM (final concentration: 80 ng ml⁻¹), naïve rat plasma collected 0.17 h after a single oral administration of 7.0 mg kg⁻¹ KM and spiked with GM, and STZ-induced DNP rat plasma collected 0.17 h after a single oral administration of 7.0 mg kg⁻¹ KM and spiked with GM. The calibration curves for KM (0.2–200 ng ml⁻¹) were fitted to a 1/x ² weighted least squares linear regression model. The intra-day and inter-day precisions for the four levels of quality control (QC) samples (lower limit of quantitation [LLOQ], low, medium, and high) ranged from 1.6 to 12.7% and 3.5% to 13.8%, respectively (see detailed method validation data in Supplementary Table S2). At the LLOQ of 0.2 ng mL⁻¹ KM, the average inter-day deviations of the predicted concentrations from the nominal values were within 2%. The matrix effects for the QC samples were 100.0–108.0% for KM (%CV: < 6%). The mean extraction recoveries of KM ranged from 59.0 to 67.2% among QC samples (Supplementary Table S3). The detailed stability KM data are shown in Supplementary Table S4.

Figure 3 displays the mean plasma concentration-time profiles on a linear (Figure 3A) or log10-linear (Figure 3C) scales of KM (0.28, 1.4, and 7.0 mg kg⁻¹) after oral gavage in STZ-induced DNP rats. The PK parameters calculated using the non-compartmental model are summarized in Table 1. In accordance with our previously published studies (Ye et al., 2020), the absorption of KM in the gastrointestinal tract was rapid (T_max: 0.21–0.36 h) for all doses in DNP rats. KM was detected at 0.083 h in the majority of DNP rats. The AUC_0–∞ of the 7.0 mg kg⁻¹ dose was over 27-fold higher than that of the 0.28 mg kg⁻¹ dose (220.66 ± 23.64 vs. 8.10 ± 2.58 ng h ml⁻¹; p < 0.001; Table 1) and nearly 8-fold higher than that of the 1.4 mg kg⁻¹ dose (220.66 ± 23.64 vs. 27.69 ± 6.68 ng h ml⁻¹; p < 0.001; Table 1). The t_1/2 of the 7.0 mg kg⁻¹ dose was 3.6-fold longer than that of the 0.28 mg kg⁻¹ KM dose in DNP rats (3.55 ± 1.36 vs. 0.97 ± 0.35 h; p < 0.01; Table 1) and over 2-fold longer than that of the 1.4 mg kg⁻¹ dose (3.55 ± 1.36 vs. 1.63 ± 0.44 h; p < 0.05; Table 1). The V_d of KM was significantly lower for the 0.28 mg kg⁻¹ dose than for the 7.0 mg kg⁻¹ dose (p < 0.01; Table 1). There was a slight decrease in CL for the 7.0 mg kg⁻¹ dose, but this did not significantly differ among the three doses (p > 0.05).

The PK parameters of orally administered KM in naïve and DNP rats are summarized in Table 1. Figure 3 displays the mean plasma concentration-time curves on a linear or log10-linear scales of KM (0.28, 1.4, and 7.0 mg kg⁻¹) after oral administration to DNP rats (Figures 3A,C) or naïve (Figures 3B,D). Similarly, KM was rapidly absorbed in both naïve and DNP rats. The T_max of KM in naïve and DNP rats ranged from 0.14 to 0.36 h (p > 0.05). The AUC_0–∞ for KM in DNP rats was > 3-fold higher than that in naïve rats (220.66 ± 23.64 vs. 49.49 ± 8.53 ng h ml⁻¹ for 7.0 mg kg⁻¹; 27.69 ± 6.68 vs. 7.35 ± 1.72 ng h ml⁻¹ for 1.4 mg kg⁻¹; and 8.10 ± 2.58 vs. 2.46 ± 0.85 ng h ml⁻¹ for 0.28 mg kg⁻¹; p < 0.001; Table 1). The t_1/2 was comparable in the naïve and DNP groups. The CL of KM was significantly lower (by 21.2% for 7.0 mg kg⁻¹, 26.7% for 1.4 mg kg⁻¹, and 31.8% for 0.28 mg kg⁻¹, respectively) in DNP rats than in naïve rats (p < 0.001; Table 1). The V_d values of KM were significantly lower (by 33.0% for 7.0 mg kg⁻¹, 49.5% for 1.4 mg kg⁻¹, and 37.1% for 0.28 mg kg⁻¹, respectively) in DNP rats than in naïve rats (p < 0.001; Table 1).

A two-compartment model with first-order absorption and elimination best fits the plasma KM PK data as described by the differential Eqs. 6–8 (Supplementary Table S5): d X a d t = − K a × X a , (6) V 1 d C 1 ( t ) d t = K a × X a + C L 2 × C 2 ( t ) − C L 2 × C 1 ( t ) − C L 1 × C 1 ( t ) , (7) V 2 d C 2 ( t ) d t = C L 2 × C 1 ( t ) − C L 2 × C 2 ( t ) , (8) where X _a is the amount of KM at the administered compartment, K _a is the absorption rate constant, C ₁ and C ₂ are the KM concentrations in the central and peripheral compartments, respectively; V ₁ and V ₂ are the central and peripheral volumes of distribution, respectively; CL ₁ is the systemic clearance rate, and CL ₂ is the distribution clearance between the central and peripheral compartments. Among several residual models described in the method, the log-additive error model was found to best fit the residuals and make selections. The non-diagonal covariance matrix was used to establish the final PK model. The covariate blood glucose level prior to KM treatment (Glu) and dosage were included to obtain an acceptable goodness-of-fitting result, which affected the CL ₁ and V ₁ incorporated as a median-normalized covariates (Δ-2LL = −65 points) and well explained the BSV for these parameters. Figure 4A and Supplementary Figure S2 show the results of the VPC and the goodness-of-fit plots of the final population PK model, respectively. These plots suggested that the population PK model accurately described the IPRED and PRED of the KM plasma concentration. The final PK parameter estimates and the results of the bootstrap validation are presented in Table 2. The absolute value of the %RSE of each PK fixed-effect parameter estimate was <30%. The BSV could be estimated for all PK parameters, including the K _a , V ₁, V ₂, CL ₁, and CL ₂, with respect to the values of shrinkage estimates <40%. The %BSV for each BSV parameter was large (85.1–443.5%). However, the robustness of the population PK model was validated by bootstrap validation. Each parameter estimate obtained from the bootstrap procedure was similar to those obtained from the original dataset with a RSE <10%, indicating that the model adequately estimated the PK model parameters.

A linear model with an effect compartment best described the KM PD data as described by differential Eqs 9, 10: d C e d t = K e × [ C 1 ( t ) − C e ( t ) ] , (9) d E d t = E 0 + K eff × C e ( t ) , (10) where, C ₁ and C _e are the KM concentrations in the central and effect site compartments, respectively, E ₀ is the baseline of the MWT, and K _eff is the slope of the KM effect on MWT. The final graphical PK-PD models are shown in Figure 5, and describe the complete time course of KM and MWT changes. Figure 4B presents the results of the VPC for the final population PK-PD model for MWT changes. The final PD parameter estimates for the MWT and the results of the bootstrap validation are presented in Table 2. The %RSE of each fixed-effect parameter estimate and BSV parameter was acceptable. In several residual models described in the method, the log-additive error model was found to best describe the residuals and make selections. In a stepwise method, the covariate blood glucose level prior to KM treatment (Glu) and dosage were included to obtain a better-fitting result, which affected the E ₀ and K _eff, respectively (Δ-2LL = −148 points); this completely explains the BSV for this parameter. Figure 4B and Supplementary Figure S3 show the results of the VPC and goodness-of-fit plots of the final population PK-PD model, respectively. These plots show that the population PK-PD model adequately describes the IPRED and PRED of the KM plasma concentration and KM effect on MWT. The final PD parameter estimates and results of the bootstrap validation are shown in Table 2. Neither PD parameter could be estimated from BSV. The robustness of the population PD model was evaluated by bootstrap validation. Parameter estimates obtained from the bootstrap validation procedure was similar to those obtained from the original dataset with a RSE < 13%, suggesting that the model adequately estimated the PD model parameters.