Body weights, organ weights and transaminases in blood
No differences in body weight development of treated animals compared to the vehicle control were observed. Individual data are summarized in supplementary data Figure S1. Absolute liver weights did also not show relevant differences between vehicle control and treatment groups. The high-dose senkirkine treatment group showed a slightly but significantly increased relative liver weight (111% of vehicle control, p < 0.05) (Supplementary data, Fig. S2).
The levels of the hepatic enzymes aspartate aminotransferase (AST) and alanine aminotransferase (ALT) were determined in blood to identify hepatotoxicity. All groups did not show statistically significant (p < 0.05) changes in neither ALT nor AST levels, except the high-dose group of lasiocarpine (3.3 mg/kg bw) which showed a statistically significant (p < 0.05) increase of ALT from 85.8 ± 22.5 U/l in control animals to 160.5 ± 19.4 U/l (Supplementary data Fig. S3). However, no morphological changes were detected in this group during histopathological examination (see “Histopathology”).
The livers of all animals were examined histopathologically. A (multi)focal random mixed inflammatory cell infiltration as well as (multi)focal periportal mononuclear cell infiltration were seen to a very slight extent in up to two animals per group in most groups (Supplementary data Fig. S4). Within the senecionine high-dose group (3.3 mg/kg bw), four animals showed a (multi)focal random mixed inflammatory cell infiltration (3/5 very slight and 1/5 slight). However, these lesions occur occasionally in rats and are considered to be unrelated to the treatment.
Taken together, results of biochemical analysis showed only weak effects for lasiocarpine in the highest dose group by inducing a twofold increase in ALT activity. This suggests that the chosen dose range represents doses in the borderline range for classic toxicity and, thus, may be in the range of or below the no-observed adverse effect levels (NOAEL). Therefore, it is assumed that the changes in the following gene expression analysis are PA-specific effects instead of unspecific effects resulting from massive injury of the whole organism.
Whole-genome microarray analysis
A whole-genome microarray analysis was performed to identify PA-affected pathways in dependence of the PA structure type. For statistical evaluation, the whole dataset including 70,580 probe sets was reduced to 26,480 annotated probe sets and an ANOVA between all high-dose groups and vehicle control was performed. This resulted in 284 significantly altered (q < 0.05) probe sets (see Supplementary data, Table S1). Hierarchical clustering of these transcripts according to the PA exposure groups is shown in Fig. 2a.
Platyphylline as a non-hepatotoxic PA showed a gene expression pattern very similar to the vehicle control up to the highest tested dose of 3.3 mg/kg bw. Senkirkine showed the strongest alterations compared to the vehicle control, followed by senecionine. Heliotrine and lasiocarpine showed alterations of moderate magnitude between platyphylline and senecionine/senkirkine. Echimidine was the treatment group with the highest variability in the magnitude of alterations between the different individuals. To identify the direction of regulation, significantly differentially regulated (q < 0.05, ANOVA) probe sets were analyzed further performing two-group comparisons of each PA against the vehicle control. The number of up- and downregulated probe sets is shown in Fig. 2b. Most of the dysregulated genes were upregulated. Again, senecionine and senkirkine showed the highest numbers of significantly dysregulated probe sets. Platyphylline, a PA representative assumed to be non-hepatotoxic, provoked significant regulation of only 11 probe sets in total. The Venn diagram in Fig. 2c shows the overlap of similar dysregulated probe sets for the different PA. In total, a signature of 36 dysregulated probe sets corresponding to 35 individual genes was identified to be dysregulated by all hepatotoxic PA. This gene signature is summarized in Fig. 3. The genes regulated by individual PA are summarized in Table S2 in the supplemental material.
This list includes a high number of genes associated with DNA damage response; cell cycle control/integrity, or apoptosis: Rad51 recombinase (Rad51) plays an important role during homologous recombination and is upregulated. O-6-methylguanine-DNA methyltransferase (Mgmt) is involved in the repair of methylated DNA and is also upregulated. Cyclin-dependent kinase inhibitor 1A (Cdkn1a) and Cyclin G1 (Ccng1) function as regulators of the cell cycle. Cdkn1a regulates cell cycle progression and leads to cell cycle G1 phase arrest. Ccgn1 is associated with DNA damage-induced G2/M phase arrest. Both genes are regulated by the tumor suppressor protein p53 and were upregulated in this study. Upregulation of these genes suggests that PA may provoke DNA damage and, thus, induce expression of genes important for DNA repair and subsequent cell cycle control. The transcript for DNA damage-induced apoptosis suppressor (Ddias) was found to be upregulated; this may indicate a possible counter-regulation elicited to protect against DNA damage-induced cell death. MYB proto-oncogene like 2 (Mybl2) is a transcription factor that regulates cell cycle progression. Deregulation of Mybl2 expression is associated with cancer initiation and progression (Musa et al. 2017). Polo-like kinase 4 (Plk4) mediates duplication of centrioles during cell cycle and was also upregulated in this study. Minichromosome maintenance complex component 5 (Mcm5) is important for DNA replication during G1 and S phases of the cell cycle and was identified to be upregulated. Ribosomal protein L19 (Rpl19) is part of the 60S subunit of the ribosome and, thus, is essential for protein synthesis. Rpl19 was also among the upregulated common dysregulated genes. Upregulation of these three genes may point to replacement proliferation as a consequence of cell death resulting from DNA damage. Induction of cell death is furthermore emphasized by upregulation of genes related to cell death: the Fas cell surface death receptor (Fas) is involved in initiation of extrinsic apoptosis. An induction of apoptosis was also reported to be mediated by the ectodysplasin A2 receptor (Eda2r) (Sinha and Chaudhary 2004). BCL2-associated X (Bax) protein is another pro-apoptotic factor. Furthermore, several genes associated with transcriptional activity were identified as commonly dysregulated genes: ring finger proteins makorin ring finger protein 1 (Mkrn1) and ring finger protein 144A (Rnf144a) were downregulated, and ubiquitin like with PHD and ring finger domains 1 (Uhrf1) was upregulated. These three genes are associated with a novel class of zinc finger proteins that may act as transcriptional regulators. Putative homeodomain transcription factor 1 (Phtrf1) was downregulated. Finally, genes related to xenobiotic metabolism showed an altered expression upon treatment with PA: aldehyde dehydrogenase family 1 member A7 (Aldh1a7), aldehyde dehydrogenase 1 family member A1 (Aldh1a1) and glutathione S-transferase alpha 1 (Gsta1) encode enzymes for xenobiotic metabolism and were all upregulated in this study. A total of 48 genes were individually regulated by only one single PA (listed in Supplementary Table S2). While the low number of individual genes per PA does not allow concluding on specific mechanisms activated by only one of the test compounds, it is obvious that several genes are related to cell cycle, DNA repair or genotoxic responses, for example cyclin D1 (Ccnd1), platelet-derived growth factor C (Pdgfc), KiSS-1 metastasis suppressor (Kiss1), CD244 molecule (Cd244), caspase 12 (Casp12), or non-homologous end-joining factor 1 (Nhej1), respectively.
Figure 4 depicts the dose dependence of significantly deregulated (q < 0.05) transcripts for each PA in a principal component analysis (PCA) plot. The PCA shows that the highest dose (3.3 mg/kg bw) differs clearly from vehicle control among the main discriminate axis (PC1). A tendency of gene expression alterations compared to vehicle control was also observed for the second highest dose (1.0 mg/kg bw). These results suggest that 3.3 mg/kg bw is a dose resulting in clear effects on the transcriptome of the liver, and that 1.0 mg/kg bw was in the borderline dose range near the NOEL for transcriptomic changes.
In case of platyphylline exposure, only the highest dose was analyzed. The PCA analysis of this group, therefore, displays a more pronounced discrimination along PC1, but of opposite direction compared to the other PA. Since unlike for the other PA no lower doses were analyzed, no conclusions concerning dose response can be drawn for this PA.
Verification of microarray data
The expression levels of 4 upregulated and 3 downregulated genes were additionally assessed by qPCR. The transcripts for qPCR validation were randomly chosen from the set of significantly dysregulated genes (q < 0.05) which were directly affected by at least 3 of all toxic PA in the highest-dose group. The gene expression pattern as analyzed by qPCR was similar to that of the microarray assay (see Supplementary data Fig. S5), thus verifying the microarray data.
Ingenuity Pathway Analysis
IPA analysis was based on a subset of 284 significantly (q < 0.05, ANOVA) dysregulated probe sets. For identification of a dose-dependent up- or downregulation of the transcripts, the correlation coefficient of a rank regression performed across all dose levels was uploaded to IPA. An IPA comparison analysis between all PA predicted affected canonical pathways, diseases and functions, and upstream regulators for the toxic PA echimidine, heliotrine, lasiocarpine, senecionine and senkirkine (Fig. 5). Platyphylline as a non-toxic PA representative showed no predicted activation or inhibition of any of the analyzed functions. The canonical pathway analysis predominantly predicted interaction with pathways related to DNA damage and cell cycle control/integrity (Fig. 5a).
IPA analysis found that ataxia telangiectasia mutated (ATM) signaling and tumor suppressor protein p53 signaling show the highest activation z-score. ATM signaling was predicted to be activated by heliotrine, senecionine and senkirkine; p53 signaling was predicted to be activated by all PA except platyphylline. According to the IPA knowledgebase (QIAGEN Inc., https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis), both of these pathways are involved in initiation of DNA repair and able to lead to cell cycle arrest in the case of massive DNA damage. P53 furthermore regulates as a transcription factor the expression of several genes. Additionally, estrogen-mediated S-phase entry (activated by heliotrine, senkirkine and senecionine), G1/S checkpoint regulation (inactivated by heliotrine, senkirkine and senecionine), cyclins and cell cycle regulation (activated by heliotrine, senkirkine and senecionine), G2/M DNA damage checkpoint regulation (activated by senkirkine and senecionine), cell cycle regulation by BTG family proteins (activated by senkirkine), and mitotic roles of polo-like kinase (activated by senkirkine and senecionine) were predicted to be regulated. All of these pathways are related to cell cycle regulation. PI3K/AKT signaling was predicted to be activated by senkirkine. It regulates cell proliferation and cell death and is associated with cancer. Signaling by Rho family GTPases was predicted to be activated by heliotrine, senkirkine and senecionine. Rho family GTPases are important for regulating signal transduction and, thus, involved in several cellular processes. Finally, aryl hydrocarbon receptor signaling was predicted to be inactivated. The aryl hydrocarbon receptor is a transcription factor which regulates the expression of several genes, e.g., relevant for xenobiotic metabolism, cell proliferation and, thus, cell cycle regulation.
Most predictions in the area of diseases and functions are associated with cell viability/cell death/cell cycle, DNA damage or carcinogenesis (Fig. 5b). These predictions point to a genotoxic effect accompanied by replacement proliferation. Most of the predicted diseases and functions were affected by all PA except platyphylline. Platyphylline shows no impairment of any disease or biofunction.
Limiting the diseases and biofunctions predictions to hepatotoxicity only, resulted in predictions of functions related to liver cancer (Fig. 5c), such as stimulation of formation of liver tumor, hepatocellular carcinoma, liver cancer and liver carcinoma. Again, for platyphylline, no activation or inactivation of any hepatotoxic function was predicted.
IPA upstream analysis filtered for transcription regulators identified several regulators associated with cell cycle control (Fig. 6). Cyclin D1 (Ccnd1) is involved in G1/S transition during cell cycle control and was activated by all analyzed PA, except platyphylline. Ccnd1 interacts with retinoblastoma protein (Rb/Rb1). Rb/Rb1 inhibits cell cycle progression and functions as a tumor suppressor. Rb/Rb1 was downregulated on mRNA level by all PA (except platyphylline) pointing again to proliferation and possibly a promotion of cancer development. Rb/Rb1 binds also to transcription factors of the E2f family leading to its inactivation. Transcription factors of the E2f family are also important for cell cycle progression. E2f1, E2f and E2f3 were predicted to be activated by all analyzed toxic PA. E2f6 was predicted to be inactivated by all PA, except lasiocarpine and platyphylline. Further activated transcription regulators, such as β-catenin, Foxm1, Irf1, Myc, Mitf, Tbx2 with a somewhat lower z-score did also point to cell cycle regulation, DNA damage and cancer development.