The PV-1 was prepared by Longhua Hospital, Shanghai University Traditional Chinese Medicine (China). Herbal components were provided by Shanghai Traditional Chinese Medicine Technology Co. Ltd. (SPH Hua Yu Chinese Herbs Co. Ltd., Shanghai, China), the formula was prepared in a GMP facility (National Traditional Chinese Medicine Pharmaceutical Engineering Technology Research Center, Building 4, No. 200, Newton Road, Zhangjiang Hi-Tech Park, Pudong, Shanghai, China), allowing for standardized preparation of the herbal materials. The composition of PV-1 includes: Polygonum bistorta, Prunella vulgaris, Sonchus brachyotus, and Dictamnus dasycarpus 2:1:1:1. The main components of PV-1 were characterized by LC-MS. The resulting powder was stored at −80°C and freshly dissolved into water before each use. Vinyl carbamate (VC) was purchased from Toronto Research Chemicals (Canada). VC was dissolved in saline and freshly prepared immediately before each use. Benzo[a]pyrene [B(a)P] , 4NQO and tricaprylin were obtained from Sigma (St. Louis, MO). Female A/J, C57BL/6 and Rag 2^-/- mice were purchased from Jackson Laboratories (Bar Harbor, ME). Lewis lung carcinoma (LLC) cells were obtained from American Type Culture Collection (ATCC). B16-BL6 (B16) melanoma cells were obtained from Dr. Li Wang (Lerner Research Institute) (20). Both LLC and B16 cells were cultured in RPMI-1640 (Thermofisher), which was supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. The animal studies were approved by the Houston Methodist Research Institute Institutional Animal Care and Use Committee (approval number: IS00006363).

Eight-week-old female A/J mice were used in the PV-1 toxicity study. Mice were treated with different doses of PV-1 by oral gavage, the duration was five times per week for eight weeks. Body weights were measured bi-weekly. One hour after the last treatment, serum was collected to test liver function (alanine transaminase (ALT) and aspartate aminotransferase (AST)) and kidney function (blood urea nitrogen (BUN)) at Marshfield Labs (Marshfield, WI) (21). In the meantime, plasma was collected in EDTA-treated tubes for the quantitative analysis of PV-1. The samples were analyzed with UPLC-MS/MS. Baohuoside I was used as an internal standard.

Animal studies with two different carcinogens were carried out. For the first experiment, A/J mice were given a single intraperitoneal injection of VC (16 mg/kg body weight) in saline. One week after the VC injection, mice were randomly divided into two groups, the vehicle control group and the PV-1 treatment group (n=12 per group). PV-1 (3000 mg/kg body weight) was given by oral gavage (0.3 ml once daily), five times per week, for 20 weeks; control mice were given the same volume of PBS. Body weights were recorded every week. Mice were analyzed for the number and/or size of tumors and imaged by MRI before being euthanized. A 9.4 T MRI (Bruker) with a custom birdcage style Quadrature coil (Doty Scientific) was used to monitor the tumors as described previously (22). Five lungs and spleens from each group were collected for flow cytometry analysis.

In the second experiment, A/J mice were given intraperitoneal injection of 100 mg/kg body weight B(a)P. B(a)P was dissolved in tricaprylin. Mice were randomly divided into two groups, the vehicle control group and PV-1-treated group (n=10 per group). Two weeks after B(a)P injection, mice were treated for 18 weeks as described above. Body weights were recorded every week. Tumor numbers and sizes were assessed at the experimental endpoint.

For both experiments, lungs from each mouse were fixed in zinc formalin for 24 hours and then stored in 70% ethanol. The lungs were evaluated with a dissecting microscope as previously described (23). The following formula: V= (4/3) πr³ was used to calculate the tumor volume, and the tumor load was determined by averaging the total tumor volumes of individual mouse in each group.

The 4NQO-induced mouse oral cancer model has been commonly used to assess the effect of natural compounds on oral carcinogenesis (24). Six-week-old C57BL/6 J mice were randomized into two groups: vehicle control group and PV-1-treated group. To drinking water, 4NQO (50 μg/mL) was provided to mice for 16 consecutive weeks. Mice were treated with PV-1 or vehicle one week after 4NQO treatment and continued for 20 weeks. Mice were treated five times per week. The body weights were measured weekly. Twenty weeks after the start of PV-1 treatment, mice were euthanized, and tongues were collected to measure oral tumor size.

The tumors out of three pairs of lungs from each group (control and PV-1-treated) of B(a)P-treated mice were harvested and pooled by individual mouse for RNA-seq analysis. The Qiagen (Valencia, CA) RNeasy^® Mini Kit was used to extract total RNA from the lung tissues. Agilent 2100 Bioanalyzer (Santa Clara, CA) was used for measuring the quality of the RNA samples. The HiSeq 2500 sequencing platform (Illumina, San Diego, CA) was used for whole transcriptome analysis of RNA-seq library samples, and the quality of the RNA-seq reads was analyzed using the FastQC program (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). The coverage ranged from 15 million to 32 million reads per sample. The quality scores of >95.3% of all the bases of each sample were at least 30, and averaged about 40, greatly exceeding the gold standard threshold of 20.

Raw RNA-seq reads were aligned to the mm9 mouse genome (UCSC version, July 2007) using Bowtie-TopHat (version 2.0.4) (25, 26). HTseq was used to obtain read counts (27) and the R package RUVSeq was used to adjust for batch effects (28). Data normalization and differential expression analysis were conducted using edgeR and limma (29, 30).

To determine marker genes for different immune cell populations, the tumor genes that were differentially expressed between the PV-1-treated and vehicle control mice were analyzed using gene set variation analysis (GSVA) (31). The R package heatmap3 (https://cran.r-project.org/web/packages/heatmap3/) was used to generate gene expression heatmaps. Bar plots were generated using GraphPad Prism 9.0 software (https://www.graphpad.com/company/).

Mouse lungs tumors were harvested and pooled from each mouse, processed into a single-cell suspension using a tumor dissociation kit (Miltenyi Biotec, CA). Spleens were collected and processed to single-cell suspensions. Cells were first stained with fixable live/dead dye in PBS, and after a PBS wash, they were stained with surface T cell-staining buffer containing fluorochrome-conjugated anti-CD45, anti-CD4, anti-CD8a, anti-CD44, and/or anti-CD62L antibodies; or with myeloid-staining buffer containing anti-CD11c, anti-CD11b, anti-Ly6G, anti-Ly6c, anti-F4/80, NKp46, and/or anti-MHCII antibodies.

To measure intracellular cytokines, after T cell surface staining, cells were washed, fixed, permeabilized, and then stained with intracellular cytokine staining buffer containing anti-granzyme B, anti-IFN-γ, and/or anti-TNF-α antibody. Cells were evaluated by an LSR-II flow cytometer (BD). FlowJo software was used for the analysis of the data (32).

To generate syngrafts in C57BL/6 mice, 3 × 10⁴ B16 melanoma cells or 1×10⁵ LLC cells (in 0.1 ml of PBS) were inoculated subcutaneously into the right flanks of mice. Animals were randomly assigned to one of four groups: 1) vehicle control; 2) PV-1 (3000 mg/kg/day); 3) 200 µg/mouse anti-PD-1 antibody (Biolegend, San Diego, CA) given by the intraperitoneal injection of every other day; or 4) combined PV-1 and anti-PD-1 antibody. For the tumor syngrafts in B16 Rag2^-/- mice, 6-week animals were given: 1) vehicle control; 2) PV-1 (3000 mg/kg/day); 3) anti-PD-1 antibody. Mice in LLC model were treated 16 days, and mice in B16 model were treated 3 weeks. Tumor sizes were measured with a digital caliper. The following formula (D × d2)/2, was used for tumor size, in which D represents the large diameter and d represents the small diameter of the tumor.

GraphPad Prism 9.0 (GraphPad Software) was used for general statistical analyses between treatments. Student’s t-test was applied for pairwise comparisons. For assessing multiple comparisons, ANOVA was used with Tukey’s post-hoc test. P-values < 0.05 were considered significant. All data were presented as means ± standard error of means (SEM). Sample sizes (n) and replicates were mentioned on each figure legend. For all figures, *p < 0.05,**p <0.01, and ***p < 0.001 were considered statistically significant.

To monitor the toxicity of PV-1, an eight-week study was conducted in A/J mice given either 1× (3000 mg/kg.bw), 2× (6000 mg/kg.bw), or 3× (9000 mg/kg.bw) five times per week of PV-1. Body weights were measured biweekly. Compared with the control group, treatment with different doses of PV-1 did not cause changes in body weight ( Figure 1A ). The general appearance of the animals (skin, eyes, breathing, and hair loss) and clinical signs for potential toxicities such as bleeding or diarrhea were also evaluated. No differences were observed between the PV-1-treated and control mice. Serum liver enzymes (ALT, AST) and BUN were also measured, and PV-1 did not change any of these liver or kidney function indicators ( Figures 1B-D ). These results suggest that PV-1 is safe and suitable for future clinical studies.

UPLC-MS/MS was used to analyze the components of PV-1 based on previous publications (33). The key components of PV-1 are identified as dictamine, maackiain, and matrine. Dictamine was selected to be used as a marker compound for PV-1 due to its bioactivity against cancer cells and its reasonable abundance in PV-1 ( Figure 1E ). Mice were treated with different doses of PV-1 for 8 weeks. One hour after the final treatment, dictamine plasma concentrations were assessed. Dictamine levels exhibited time- and dose-dependent increases in blood samples ( Figure 1F ).

There have been no prior publications demonstrating the impact of PV-1 on the development of lung tumors. Here, the chemopreventive effect of oral gavage PV-1 in VC- or B(a)P-induced lung adenocarcinoma in female A/J mice was investigated according to the experimental designs shown in Figures 2A, D . In both lung cancer models, tumor incidence was 100% for all treatment groups. There were no obvious abnormal manifestations such as hair loss, icteric sclera, or changes in body weight (data not shown). The absence of any clinical toxicity is consistent with the data in Figure 1 .

In the VC-induced lung adenocarcinoma model, MRI imaging was used to evaluate tumor size. Observable differences were identified between control and PV-1-treated mice ( Figure 2B ). At the experimental endpoint, tumor loads in the VC-induced control group averaged 16.4 ± 0.3 mm³ in size, while tumor loads in the PV-1 groups were significantly reduced to 7.6 ± 0.5 mm³ and 5.6 ± 0.1 mm³, respectively ( Figure 2C ). In the B(a)P model, tumor load in the control group was 1.3 ± 0.3 mm³, while mice treated with PV-1 showed a significantly decreased tumor load of 0.5 ± 0.1 mm³ (62.0% inhibition) compared with the controls ( Figure 2E ).

Mouse tumors were collected from control and PV-1-treated mice in the B(a)P model for RNA-seq and assessed for changes that would indicate effects on immune cell populations. PV-1 caused an upregulation of six marker genes known to be expressed in effector-memory CD8+ T cells ( Figure 3A ). GSVA levels indicated that the effector-memory CD8+ T cell population was significantly more activated in tumors of mice treated with PV-1 than in vehicle controls ( Figure 3B ). These results suggest that the anti-tumor efficacy of PV-1 could involve the modulation of immune cells in the tumor microenvironment, specifically the activation of effector T cells.

To further explore whether PV-1 modulates immune function, flow cytometry was conducted on tumors and splenocytes isolated from VC-induced mice using various immune cell markers. In the lung tumors from VC-induced mice, the percentage of CD8+ tumor-infiltrating lymphocytes (TILs) was significantly increased upon PV-1 treatment ( Figure 4A ), and the production of granzyme B, TNF-α and IFN-γ by CD8+ cells were all significantly increased in these mice ( Figures 4B-D ). The g-MDSCs (CD11b+Ly6G+) were significantly decreased in the spleens and tumors of PV-1-treated mice compared with controls ( Figures 5A, B ). In addition, significant increases in the percentages of natural killer (NK) cells and antigen-presenting cells (APC) were found in mouse spleens after treatment with PV-1 ( Figures 5C, D ). Because MDSCs can markedly suppress the T cell response, PV-1-mediated suppression of MDSCs could enhance the function of T cells that target tumor cells. These results suggest that the cytotoxic function of these CD8+ cells is increased by PV-1, which could facilitate their migration to tumor sites and increase their potential to kill tumor cells.

Clinical studies have demonstrated that PD-1/PD-L1 antibodies/inhibitors induce anti-tumor efficacy in non-small cell lung cancer (34, 35). Therefore, an LLC syngraft mouse model was used to determine whether PV-1 could enhance the efficacy of the anti-PD-1 checkpoint blockade. Subcutaneous injection of LLC tumor cells was done on the flanks of C57BL/6 mice, and mice were treated with PV-1, anti-PD-1 antibody, or both in combination, five times/week for 16 days ( Figure 6A ). Treatment with PV-1 alone inhibited tumor growth by 53.8%, while treatment with anti-PD1 antibody alone inhibited tumor growth by 56.7%, when compared with the control ( Figure 6B ). Notably, a significantly greater anti-tumor effect was observed with the combination of PV-1 and anti-PD1 antibody (75.4% inhibition vs. control).

To confirm that the anticancer effects of PV-1 are mediated through adaptive immune mechanisms, the effects of PV-1 and anti-PD-1 antibody were measured in B16-inoculated Rag2^-/- mice ( Figure 7A ), which do not contain mature B or T lymphocytes (36). Like the LLC model, inhibition of B16 tumor volumes in wild-type mice by PV-1 or anti-PD1 antibody alone were 45.0% and 52.1%, respectively ( Figure 7B ). The combination of PV-1 and anti-PD1 antibody induced better anti-tumor efficacy than either agent alone (75.0% inhibition) ( Figure 7B ). In Rag2^-/- mice, neither PV-1 nor anti-PD-1 antibody treatment induced significant tumor inhibition ( Figure 7C ). These results indicate that the anti-tumor effects of PV-1 are heavily dependent on adaptive immunity.

To test the impact of PV-1 on 4NQO-induced oral carcinogenesis, C57BL/6 mice were treated as illustrated in Figure 8A . Compared with the control group, mice treated with PV-1 showed no difference in body weight (data not shown), and consumption of water was similar between the control and treated groups. Tongues of 4NQO-treated mice showed white spots or exogenous papillary lumps ( Figure 8B ). PV-1 treatment induced significant inhibition of tumor lesions when compared with controls; the mean tumor area decreased from 8.1 mm² in the 4NQO control group to 3.8 mm² (53.1% inhibition) in the PV-1 group ( Figure 8C ).