A total of 48 male Wistar rats [age, 8–12 weeks; body weight (BW), 180 to 220 g] were purchased from the Pasteur Institute of Iran. In a polycarbonate cage, rats were subjected to a 12-h cycle of light and darkness. The humidity level was 65% ± 5%, and the temperature was adjusted to 25°C. Food (rat chow) and water were provided on ad libitum basis. The Regional Research Ethics Committee of Semnan University of Medical Sciences and Health Services approved all experimental protocols (IR.SEMUMS.REC.1399.291). A guide for the care and uses of laboratory animals published by the US National Institutes of Health was followed during the conduct of the study.

Hydro-alcoholic extracts of nettle leaf were prepared according to previous reports (43, 44). Dried leaves of nettle were purchased from an herbal medicine shop. The extraction method was cold maceration with an ethanol solvent (70%). Briefly, leaves were finely ground, and 100 g of the produced powder was macerated in 1,000 ml of 70% ethanol (herb/solvent ratio of 1:10 w/v) for 72 h with intermittent stirring. After filtering, the solvent was evaporated in a rotary evaporator at 50°C. The powder was kept at −20°C until use. The powder was dissolved in distilled water at the time of administration. The herb/extract ratio was 8:1 (w/w). Phytochemical analysis of the hydro-alcoholic extract of nettle leaf was done using the gas chromatography method by a model HP-6890N Gas chromatography–mass spectrometry (GC-MS) coupled to a 5975 mass spectrometer (Agilent Technologies, Palo Alto, CA, USA) combined with a TDU (Thermal Desorption Unit) and cooling injector system (CIS4, Gerstel). Chromatographic separation was done in an HP5MS-UI column (Agilent Technologies) with helium as the gas carrier in constant pressure mode and split ratio 1:50. Initial temperature was 50°C, increasing at a ratio of 5°C min⁻¹ until 70°C held for 1 min. In the next step, the temperature was increased to 240°C at 10°C min⁻¹ and held for 15 min.

The mass spectrometer operated at 70-eV ionization voltage. Source and quadrupole temperatures were 230°C and 150°C, respectively. The mass range was 30.0 to 450.0 uma at 4 scan/s. Mass Selective Detector (MSD) transfer line was maintained at 280°C. We used ChemStation software (version E.02.02 SP1, Agilent Technologies) to acquire chromatograms. Compounds were qualitatively identified by comparison using the Wiley10th-NIST11b mass spectral database (Agilent Technologies, Wilmington, DE, USA).

PubChem (https://pubchem.ncbi.nlm.nih.gov/) was searched to find SMILES notations of the detected compounds. As an alternative to PubChem, we used the OPSIN tool (https://opsin.ch.cam.ac.uk/), which provides an algorithm for interpreting most organic chemical nomenclature (45) to generate the appropriate SMILES notations for those chemicals without information available in PubChem. To perform further in silico target analysis, the SMILES notations were uploaded to the following servers. Using SwissADME (http://www.swissadme.ch), a tool that assesses the physicochemical properties, ADME (Absorption, Distribution, Metabolism, and Excretion) parameters, pharmacokinetic properties, drug-like characteristics, and medicinal chemistry friendliness of small molecules, we determined the absorption level, water solubility, and bioavailability scores of the chemicals. ProTox-II (https://tox-new.charite.de/protox_II/), which combines molecular similarity, fragment propensities, most frequent features, and (fragment similarity-based CLUSTER cross-validation) machine learning, based on 33 models predicting various toxicity endpoints (46) was applied to predict the LD50 of the chemicals and their toxicity class. For the purpose of identifying the target genes of chemicals, we used the SuperPred 3.0 server (http://prediction.charite.de). Using a machine learning algorithm and Morgan fingerprints of 2048 bits in length, SuperPred determines genes that may be targets of small molecules (47). Each chemical was assigned a specific target gene, as well as a combined target gene list for all chemical compounds ( Supplementary File 1 ). An analysis of 15 lists was conducted using the ToppFun enrichment tool of the ToppGene Suite (http://toppgene.cchmc.org/enrichment.jsp). The ToppFun program performs functional enrichment of a gene list based on a wide range of omics and ontological data (48). A ranked list of diseases associated with each target gene list has been developed. The rank of “hyperglycemia” was considered to be an indicator of the direct relatedness of the target genes to diabetes and thus the corresponding chemical.

After adaptation to the laboratory environment, the diabetic model was induced in rats by intraperitoneal injection of STZ (Sigma-Aldrich, St. Louis, MO, USA) at a dose of 50 mg/kg BW. The STZ was dissolved in 0.1 ml of sodium citrate buffer (pH 4.5). On day 3 after the STZ injection, BGL from the fasted rats was measured, and rats with levels >150 mg/dl were considered to have diabetes. STZ-induced animals with diabetes murine models develop type 1 diabetes, due to the cytotoxic Glc analog STZ that is toxic to pancreatic β cells and causes insulin deficiency (49).

Following the induction of diabetes, rats underwent a 2-week adaptation period. In addition, rats were divided into six groups: (1) Ct (control animals, n =8), (2) EndTr (rats treated with EndTr, n = 8), (3) D (rats with STZ-induced diabetes, n = 8), (4) D + NLE (rats with STZ-induced diabetes and treated with nettle leaves extract, n = 8), (5) D + EndTr (rats with STZ-induced diabetes and treated with EndTr, n = 8), and (6) D + EndTr + NLE (rats with STZ-induced diabetes and treated with EndTr and NLE, n = 8). Because of the numerous previous publications indicating that nettle consumption has hepatoprotective and antioxidative effects in the liver, as well as the beneficial effects of combining NLE and EndTr in healthy subjects (50–56), we did not include separate groups for investigating the NLE effects individually and in combination with EndTr in rats without diabetes. Animals in the exercise groups performed the designed protocol of training for 8 weeks ( Table 1A ) as described previously (59). The rats in those groups that were supposed to receive the extract were administered one dose per day of NLE (150 mg/kg BW) intragastrically 5 days a week for 8 weeks in accordance with previous reports (60). To get acquainted with the training protocol, animals in the exercise groups were trained on the treadmill by walking slowly for 5 min at a speed of 5 to 10 m/min. Subsequently, the rats in the exercise groups performed the EndTr protocol for 8 weeks and 5 days per week at 8–10 a.m. The indirect protocol was used to measure the aerobic capacity of rats. Training intensity was determined by VO2max and its relationship to treadmill speed and incline (61). After 2 weeks of initial familiarization with the treadmill, the aerobic capacity of the rats was assessed. Upon completing the initial warm-up, the test was initiated at a speed of 6 m/min. The treadmill speed was increased every 2 min by 1.8–2 m/min until the rats reached exhaustion. A protocol of training was designed on the basis of the obtained average of test speed and time. The intensity of training increased gradually during the training period.

The intensity of training from 10 m/min and 23 min (including 5 min of warm-up and 3 min of cool-down) in the first week increased to 26 m/min and 59 min in each session in the eighth week. To apply the principle of overload, an average of 6 min per week (about 1 min per day) was added to the duration of the exercise and 2 m/min to the intensity of the exercise. Each training session, after warming up (at a speed of 5 to 10 m/min), started at a speed of 10 m/min, and, every 3 min, 2 m/min was added to the speed of the treadmill to reach the desired weekly speed ( Table 1A ).

To measure the BGL, fresh blood samples were collected from the rat tail vein, and Glc levels were then determined using a blood Glc meter (ACCU-CHEK^® Active, Roche Diagnostics, Mannheim, Germany) that uses test strips to assess a Glc oxidoreductase-mediated dye reaction, according to the manufacturer’s instructions. The BGL was measured on day 3 after STZ treatment and at the end of the eighth week of the training course. The measurements were performed in duplicate.

Tissue resection was performed at the end of the eighth week and 72 h after the last training session. This was after anesthesia with CO₂ gas and blood sampling from the heart. Then, liver tissue was extracted and, after washing in physiological serum, placed in microtubes and transferred to −70°C.

Kynurenine ELISA Kit, ZellBio GmbH (cat. no. ZB-11203C-R9648, Germany), which applies quantitative assay of KYN based on the biotin double antibody sandwich technology, was used to determine KYN level in liver tissue lysate following the manufacturer’s instructions. In addition, the malondialdehyde (MDA) assay kit (cat. no. ZB-MDA-96A, Germany), which works based on the colorimetric (535-nm) method, was used for MDA assay in liver homogenate following the instructions. The determinations were performed in duplicate.

Total RNA was isolated from liver samples using TRIzol Reagent (Invitrogen, USA), treated with DNase I, and quantified by NanoDrop (Thermo Fisher Scientific). RNA quality was determined by examining the 260/280 ratio >1.8. A total of 1 µg of RNA was then reverse-transcribed to cDNA using the RevertAid First Strand cDNA Synthesis kit (Thermo Fisher Scientific) according to the manufacturer’s protocol. Expression of Ahr, Cyp1a1, and Ido1 was measured using specific primers purchased from SinaColon, Iran ( Table 1B ). CYP1A1 upregulation is a hallmark of AHR activation (62). Amplification was performed in duplicate in ABI Prism 7500 sequence detection system (Life Technologies real-time RT-PCR device). All data were analyzed with the ΔΔCt method, and the mRNA of glyceraldehyde-3-phosphate dehydrogenase (Gapdh) was used as the internal standard. The primer concentration was 10 μl. Amplification conditions were as follows: initial denaturation (95°C for 15 min for all genes); start of the cycle with denaturation (95°C for 30 s for Gapdh, and 95°C for 5 min for Ahr, Cyp1a1, and Ido1); annealing (55°C for 30 s for Gapdh, 48°C for 105 s for Ido1, 56°C for 90 s for Cyp1a1, and 60°C for 90 s for Ahr); and extension (60°C for 30 min for all genes). At the end of the amplification cycles, reactions were given a final extension step at 60°C–95°C. The relative percentage expressions of candidate genes in STZ-induced rats with diabetes were computed by considering the expression of Gapdh in Ct rats as 100%.

ELISA kits from ZellBio GmbH (Germany) were utilized for determining AHR (cat. no. ZB-16349C-R; assay range, 0.313 to 10 ng/ml; sensitivity, 0.04 ng/mL), CYP1A1 (cat. no. ZB-11059C-R9648; assay range, 3 to 96 ng/ml; sensitivity, 0.35 ng/ml), and IDO1 (cat. no. ZB-10730C-R9648; assay range, 30 to 960 pg/ml; sensitivity, 3.5 pg/ml) protein levels in liver lysates. Enzyme-linked immunosorbent assays were performed using kits based on biotin double antibody sandwich technology. Reagents, samples, and standards were prepared following the manufacturer’s instructions. Final calculations were done using the standard curves provided by the kits for each protein. The determinations were performed in duplicate.

Data are presented as means ± standard deviations (SD) in the text and tables. In addition, for each variable, the percent change in each group compared with that in the Ct group was calculated (for example, percent change of the D group versus the Ct group was calculated as following formula: m e a n   o f   D   g r o u p − m e a n   o f   C t   g r o u p m e a n   o f   C t   g r o u p × 100 ), and its statistical significance was assessed using a one-sample t-test. Pearson and partial correlation coefficients were calculated to measure the relationship between AHR, CYP1A1, IDO1, KYN, and MDA, and protein levels.

A three-way multivariate analysis of variance (MANOVA) with all main and interaction terms was first performed to determine the effect of three factors (having diabetes, yes/no; exercise, yes/no; and nettle use, yes/no) on the latent variable. If the interaction term was not statistically significant, then it was removed from the model and the analysis rerun without the interaction term. Failure to remove an interaction term that was not statistically significant will lead to an incorrect conclusion.

In the next step, if the result of the multivariate analysis was statistically significant, then univariate three-way analysis of variance (ANOVA) were done to discover the effect of a significant factor on each indicator variable (AHR, CYP1A1, IDO1, KYN, and MDA for first latent variable; Ahr, Cyp1a1, and Ido1 for second latent variable), which was followed by the Tukey’s post hoc test. Assumptions of three-way ANOVA were verified with the Shapiro–Wilk test, Levene’s test, Doornik–Hansen test, Box’s M-test, and evaluation of the homogeneity of covariance matrices. All statistical computations were done using Stata version 16 (StataCorp LLC, College Station, TX, 2019) and GraphPad Prism 8 (GraphPad Software Inc., La Jolla, California, USA). The results were considered significant with P ≤ 0.05.

As shown in Figure 1A , the GC-MS chromatogram of hydro-alcoholic extract of nettle leaf revealed peaks, which corresponded to 18 bioactive compounds identified by comparison with the known chemicals described in Wiley10th-NIST11b mass spectral database based on the peak retention time, peak area (%), peak height (%), and mass spectral fragmentation patterns. The following in silico target analysis includes chemicals (14 molecules) that have been found listed in the PubChem database with CID identifiers and those chemicals (four molecules) that we produced their SMILES notations using OPSIN ( Figure 1B ).

A summary of the gastrointestinal (GI) absorption, water solubility, and bioavailability scores calculated by SwissADME, as well as their LD50s and toxicity classes predicted by ProTox-II, can be found in Table 2 for the compounds with CID identifiers. In addition, SuperPred predicted that the chemicals had a different number of target genes ( Table 2 ; Supplementary File 1 ). The greatest number of predicted target genes, 194, was associated with chemical 10129, and the smallest number, 80, was associated with chemical 8379, as shown in Table 2 . Furthermore, ToppFun analyzed the predicted lists and identified the 50 most related diseases and abnormalities for each group of genes and ranked them ( Supplementary File 2 ). The predictions include complications related to diabetes and associated diseases such as diabetic nephropathy, diabetic cardiomyopathies, metabolic diseases, non-alcoholic fatty liver disease, cardiovascular diseases, and hypertensive disease. In particular, hyperglycemia, which is directly related to diabetes, was predicted for 18 gene lists with varying rankings ( Table 2 ; Supplementary File 3 ). The link between hyperglycemia and the target genes for chemicals 10129, 5280435, and 550234 came in third place among 50 diseases. On the other hand, hyperglycemia was not predicted to be associated with the target genes of chemicals 10914, 60161, and (tetrahydroxycyclopentadienone)-tricarbonyliron(0) chemicals. There is therefore a possibility that among the compounds analyzed, chemicals 10129, 5280435, and 550234 may play a more significant role in diabetes physiology. As far as GI absorption is concerned, these three chemicals are the same; however, 10129 has a greater water solubility and is the least toxic (LD50, 15,380 mg/kg) of the three, compared with 5280435 (LD50, 5,000 mg/kg) and 550234 (LD50, 800 mg/kg). It may therefore be worthwhile to investigate the effects of chemical 10129 on diabetes pathways through further analysis. Interestingly, IDO1 was listed among the target genes for chemicals 10129 and 550234 ( Supplementary File 1 ). When all predicted target genes for all chemicals were combined, there were 462 unique genes, with hyperglycemia ranking fifth among the abnormalities assigned to them. The first four abnormalities were hypertension, pain, myocardial infarction, and cardiovascular diseases, which are conditions related to diabetes.

As shown in Figure 2A , Glc levels were determined in the blood of rats in different study groups. The Glc level in the D group was 252.5 ± 27.10 mg/dl, which was significantly higher than the Glc level in the Ct group (86.75 ± 4.30 mg/dl). Therefore, diabetes was successfully induced in rats using STZ. In Figure 2A , it can be seen that the BGLs of the study groups differed significantly. Both the D + EndTr and D + NLE groups showed a significant decrease in BGL (P< 0.001). Nevertheless, the combination of EndTr and NLE in the D + EndTr + NLE group reduced BGL more than their individual applications. As a result, BGL was reduced to 126.25 ± 5.40 mg/dl, which was significantly lower than in D (P< 0.001), D + EndTr (P< 0.001), and D + NLE (P = 0.002) groups. In addition, the difference in BGL between the D + NLE and D + EndTr groups was significant (P = 0.011). However, although BGL was significantly lower in the D + EndTr, D + NLE, and D + EndTr + NLE groups than that in the D group, it was still significantly higher than that in the Ct group. Comparison of EndTr and Ct groups, indicated that EndTr had no effect on BGL of healthy rats.

According to Figure 2B , MDA, a biomarker of oxidative stress, was significantly higher in the liver tissues of the D group in comparison with controls (P< 0.001). However, its level was significantly (P< 0.001) reduced by EndTr (in the D + EndTr group) and NLE (in the D + NLE group). In the D + EndTr + NLE group, MDA levels were significantly lower than in the D (P< 0.001), D + EndTr (P = 0.001), and D + NLE (P< 0.032) groups. It should be noted that there was no significant difference between the MDA levels in the D + NLE and D + EndTr groups, and compared with the Ct group, EndTr did not result in an increase in MDA levels in healthy rats’ livers.

In the D group ( Figure 2C ), KYN levels significantly elevated as compared with controls (P = 0.001). However, a significant reduction in KYN levels was observed in the D + EndTr group (P = 0.006). Despite the lower KYN level recorded in the D + NLE (8.15 ± 0.32 ng/ml) and D + EndTr (7.67 ± 0.32 ng/ml) groups in comparison with the D group (9.41 ± 0.32 ng/ml), the differences were not statistically significant. However, the combined application of EndTr and NLE (in the D + EndTr + NLE group) resulted in significantly (P = 0.046) lower levels of KYN (6.73 ± 0.87 ng/ml) in comparison with using NLE alone. In addition, because there was no significant difference between the EndTr and control groups, it can be concluded that exercise did not affect the levels of KYN in the liver of healthy rats.

The results of real-time RT-PCR tests revealed that Ahr, Cyp1a1, and Ido1 expression levels significantly increased in the livers of rats in the D group ( Figure 2D ). It is notable that, although EndTr (P<0.001) and NLE (P< 0.001) have a substantial effect on the downregulation of Ahr expression, there were still significant differences between controls and the treated D + EndTr and D + NLE groups ( Figure 2D ). Despite this, their combination reduced Ahr expression to a level near that of controls ( Figure 2D ). Although EndTr and NLE reduced Cyp1a1 expression in the liver, there were no significant differences between the D group and the D + EndTr and D + NLE groups ( Figure 2E ). The combination of EndTr and NLE, however, resulted in a significant decrease in Cyp1a1 expression (P = 0.018). The expression of the Ido1 gene was also downregulated in the D + EndTr, D + NLE, and D + EndTr + NLE groups; however, the differences were not statistically significant ( Figure 2F ). In addition, exercise did not significantly alter the expression of these genes in the liver of healthy rats.

AHR, CYP1A1, and IDO1 levels significantly increased in the liver of the D group as determined by ELISA tests ( Figure 2G ). There was a significant effect of EndTr (P< 0.001) and NLE (P = 0.021) on the downregulation of AHR protein expression ( Figure 2G ). In addition, in the D + EndTr + NLE group, AHR levels significantly reduced (P = 0.001), although there were still significant differences between the control and treated groups (D + EndTr, D + NLE, and D + EndTr + NLE). Neither EndTr nor NLE or their combination significantly reduced the increased level of CYP1A1 ( Figure 2I ). An analysis of the liver CYP1A1 levels of healthy rats in the EndTr and Ct groups showed that EndTr did not significantly elevate liver CYP1A1 levels. In terms of IDO1, its level was significantly lower in the D + EndTr + NLE group compared with the D and D + NLE groups, as shown in Figure 2I (P< 0.001).

In Table 3 , descriptive statistics and percent change of assessed variables in the liver are presented for each of the six groups. On the basis of this table, AHR, CYP1A1, KYN, and MDA showed the greatest percent change in the D group, whereas IDO1 showed the greatest percent change in both the D and D + NLE groups. Combining EndTr and NLE may reduce elevated levels of AHR, CYP1A1, KYN, IDO1, and MDA in diabetic conditions in a synergistic manner, as outlined in the table. There was no significant violation of the assumption of multivariate normality [χ2 (10) = 16.23, P = 0.093]. Furthermore, Box’s M-test showed that the observed covariance matrices of the dependent variables were equal across groups (P = 0.064).

On the basis of the results of Table 4 , a significant effect of exercise × NLE × diabetes (Wilks’ λ = 0.35, partial eta square = 0.41, P< 0.001) was observed on the latent variables (AHR, CYP1A1, KYN, IDO1, and MDA simultaneously). This impact was considered in the individual univariate analysis, which indicated significant effects on AHR [F(2,42) = 16.46, P< 0.001], CYP1A1 [F(2,42) = 5.50, P = 0.008], KYN [F(2,42) = 4.53, P = 0.010], IDO1 [F(2,42) = 5.59, P = 0.007], and MDA [F(2,42) = 4.89, P = 0.012]. Although according to this table, the main effects of exercise, NLE, and diabetes, as well as the interaction effect of exercise and diabetes, were significant, but in cases where the three-way interaction effect is significant, these main and two-way effects should not be interpreted. The significance of three-way interaction of exercise, NLE, and diabetes means that the interaction among the two factors (exercise × NLE) is different across the levels of the third factor (diabetes) for AHR, CYP1A1, KYN, IDO1, and MDA. Similar findings were obtained for gene expression of Ahr, Ido1, and Cyp1a1 ( Table 5 ).

It should be noted that, in addition to Figure 2 , which shows the significant comparison of all groups, to clarify the comparisons, two groups that had a significant difference in only one factor were given in Figure 1s of Supplemental section.

Table 6A indicates that there was a highly positive significant correlation between the studied factors based on the r coefficient. Moreover, in the bivariate analysis, there was a moderately positive correlation between factors (based on r_p coefficient), but it was significant only for some of them ( Table 6A ). Figure 2 presents the results of the Tukey’s post hoc test and pairwise comparisons of groups. In addition, Table 6B presents the parameter estimates of variables in the six study groups. The control group was considered as a reference group in this analysis, and the other groups were compared with it. Resulting from the regression model, diabetes elevated all assessed variables in the liver, and the combination of EndTr and NLE significantly reduced these factors. Comparing the effects of EndTr and NLE (considering the β coefficient) suggests that EndTr affects all variables except MDA more strongly than NLE ( Table 6B ).