Transcriptome Analysis of the Anti-TGFβ Effect of Schisandra chinensis Fruit Extract and Schisandrin B in A7r5 Vascular Smooth Muscle Cells

Schisandra chinensis fruit extract (SCE) has been used as a traditional medicine for treating vascular diseases. However, little is known about how SCE and schisandrin B (SchB) affect transcriptional output-a crucial factor for shaping the fibrotic responses of the transforming growth factor β (TGFβ) signaling pathways in in vascular smooth muscle cells (VSMC). In this study, to assess the pharmacological effect of SCE and SchB on TGFβ-induced transcriptional output, we performed DNA microarray experiments in A7r5 VSMCs. We found that TGFβ induced distinctive changes in the gene expression profile and that these changes were considerably reversed by SCE and SchB. Gene Set Enrichment Analysis (GSEA) with Hallmark signature suggested that SCE or SchB inhibits a range of fibrosis-associated biological processes, including inflammation, cell proliferation and migration. With our VSMC-specific transcriptional interactome network, master regulator analysis identified crucial transcription factors that regulate the expression of SCE- and SchB-effective genes (i.e., TGFβ-reactive genes whose expression are reversed by SCE and SchB). Our results provide novel perspective and insight into understanding the pharmacological action of SCE and SchB at the transcriptome level and will support further investigations to develop multitargeted strategies for the treatment of vascular fibrosis.

TGFβ signals through the type II receptor kinase (TβRII)-mediated activation of TβRI [12,13]. The activated TβRI propagates its downstream signaling through both the Smad-dependent canonical pathways and the Smad-independent non-canonical pathways. In the canonical pathways, TβRI phosphorylates Smad2 and Smad3 to form a heteromeric Life 2021, 11,163 2 of 14 complex with Smad4. The Smad complex enters the nucleus to initiate the transcription of its target genes [14,15]. In the non-canonical pathway, TβRI stimulates non-Smad signaling pathways, including extracellular signal-regulated kinase (ERK), phosphoinositide 3-kinases (PI3K), Rho-associated coiled-coil kinase (ROCK), and inhibitor of nuclear factor-κB kinase (IKK), which in turn, activate a range of transcription factors [16,17]. These results indicate that the transcriptional and phenotypical output of TGFβ signaling is determined by the cooperation between Smad-and non-Smad signaling cascades determines the transcriptional of TGFβ signaling, which leads to fibrotic changes.
Schisandra chinensis fruit (Figure 1a) as an oriental herbal medicine has been used traditionally to treat various human diseases [18]. Recently, accumulating evidence suggests that Schisandra chinensis fruit and its active ingredients have a potential role in the treatment of vascular fibrosis [19]. We have reported that Schisandra chinensis fruit extract (SCE) and its ingredient schisandrin B (SchB, Figure 1b) have a potent anti-fibrotic activity by suppressing the TGFβ signaling pathways in VSMCs [20][21][22]. SCE and SchB block TGFβ-induced phosphorylation and nuclear translocation of the Smad complex, which decreases the expression of ECM proteins [20]. In addition, SCE and SchB inhibit the phosphorylation of myosin light chain in a Smad-independent manner, which leads to suppressing in actin stress fiber formation and cell migration [21]. Moreover, SCE and SchB attenuate TGFβ-mediated activation of IKKα/β, thereby inhibiting NF-κB activity [22]. However, little is known about how SCE and SchB affect the transcriptional output of the TGFβ signaling pathways. TGFβ signals through the type II receptor kinase (TβRII)-mediated activation of TβRI [12,13]. The activated TβRI propagates its downstream signaling through both the Smaddependent canonical pathways and the Smad-independent non-canonical pathways. In the canonical pathways, TβRI phosphorylates Smad2 and Smad3 to form a heteromeric complex with Smad4. The Smad complex enters the nucleus to initiate the transcription of its target genes [14,15]. In the non-canonical pathway, TβRI stimulates non-Smad signaling pathways, including extracellular signal-regulated kinase (ERK), phosphoinositide 3-kinases (PI3K), Rho-associated coiled-coil kinase (ROCK), and inhibitor of nuclear factor-κB kinase (IKK), which in turn, activate a range of transcription factors [16,17]. These results indicate that the transcriptional and phenotypical output of TGFβ signaling is determined by the cooperation between Smad-and non-Smad signaling cascades determines the transcriptional of TGFβ signaling, which leads to fibrotic changes.
Schisandra chinensis fruit (Figure 1a) as an oriental herbal medicine has been used traditionally to treat various human diseases [18]. Recently, accumulating evidence suggests that Schisandra chinensis fruit and its active ingredients have a potential role in the treatment of vascular fibrosis [19]. We have reported that Schisandra chinensis fruit extract (SCE) and its ingredient schisandrin B (SchB, Figure 1b) have a potent anti-fibrotic activity by suppressing the TGFβ signaling pathways in VSMCs [20][21][22]. SCE and SchB block TGFβ-induced phosphorylation and nuclear translocation of the Smad complex, which decreases the expression of ECM proteins [20]. In addition, SCE and SchB inhibit the phosphorylation of myosin light chain in a Smad-independent manner, which leads to suppressing in actin stress fiber formation and cell migration [21]. Moreover, SCE and SchB attenuate TGFβ-mediated activation of IKKα/β, thereby inhibiting NF-κB activity [22]. However, little is known about how SCE and SchB affect the transcriptional output of the TGFβ signaling pathways. In this study, we performed DNA microarray experiments in A7r5 VSMCs to examine the mechanism of action of SCE and SchB at the transcriptomic level. We found that TGFβ induced transcriptome changes in VSMCs and that these changes were significantly reversed by SCE and SchB. Further computational genome-wide analysis provided a global picture of the pharmacological effect of SCE and SchB on TGFβ-mediated biological processes and fibrotic changes. Our results provide novel perspective and insight into future translational research and into the development of healthcare strategies. In this study, we performed DNA microarray experiments in A7r5 VSMCs to examine the mechanism of action of SCE and SchB at the transcriptomic level. We found that TGFβ induced transcriptome changes in VSMCs and that these changes were significantly reversed by SCE and SchB. Further computational genome-wide analysis provided a global picture of the pharmacological effect of SCE and SchB on TGFβ-mediated biological processes and fibrotic changes. Our results provide novel perspective and insight into future translational research and into the development of healthcare strategies.

DNA Microarray Experiments and Data Processing
DNA microarray experiments were performed using total RNA from A7r5 cells, following treatment with 100 mg/mL SCE or 10 µM SchB for 24 h in the presence or absence of 1 ng/mL TGFβ1 (R&D Systems) as described in our previous papers [22,23]. SCE were prepared by ethanol extraction using ultrasonic bath and SchB was purified from SCE using the HPLC system as described previously [20,24]. The microarray data are available through the Gene Expression Omnibus (GEO) database under the accession number GSE87439. The raw microarray data were normalized using single-channel array normalization (SCAN) method [25]. SCAN is a Bioconductor R package, and R version 3.6.0 was used for all analyses. Microarray probe sets were mapped to 14,065 genes using a custom mapping file, Rat2302_Rn_ENTREZG (version 23.0.0), which is provided by the BrainArray resource [26].

Collection of Public Microarray Data
To construct A7r5 VSMC-specific transcriptional interactome, public microarray datasets were collected from the GEO database (GSE15713 and GSE21573) and normalized using the SCAN method. These data were combined with our microarray data (GSE87439), which are available from GEO database (GSE134932). The batch effect among the data sets was removed as described in our previous papers [27][28][29].

Cluster Validation and Differentially Expressed Gene (DEG) Selection
Our microarray data consist of six experimental groups depending on the reagents used in the experiments: group 1 (vehicle), group 2 (SCE alone), group 3 (SchB alone), group 4 (TGFβ alone), group 5 (TGFβ + SCE), and group 6 (TGFβ + SchB). Feature selection across six groups was performed using the Linear Models for Microarray Data (Limma) Bioconductor R package with a multiclass statistical problem type [30]. The Benjamini and Hochberg (BH) procedure was used for the adaptive control of the false discovery rate (FDR) in multiple testing [31]. The significant features were determined by the threshold FDR q-value 0.2. Internal clusters were validated by hierarchical clustering and principal component analysis (PCA) using the selected features.

Functional Assessment for DEGs
For deep functional assessment of the enriched gene signatures, the DEGs identified by Limma (less strict threshold FDR q-value 0.3) across six groups were applied to Gene Signature Enrichment Analysis (GSEA) with Hallmark gene signatures (version 6.2 at Molecular Signatures Database (MSigDB), http://software.broadinstitute.org/gsea (accessed on 10 February 2021)) in a pairwise way (i.e., TGFβ versus vehicle, SCE versus vehicle, SchB versus vehicle, TGFβ + SCE versus TGFβ, and TGFβ + SchB versus TGFβ). Significantly enriched gene signatures were determined by the threshold FDR q-value 0.2.

Protein-Protein Interaction Network Analysis
From the functional assessment of DEGs, we identified SCE-and SchB-effective genes (details in Section 3). We used Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (ver. 11.0) [32] to perform protein-protein interaction analysis for the common genes between SCE-and SchB-effective. We set required interaction score 0 and edge color for interaction evidence.

Distinctive Changes in the Gene Expression Profile in A7r5 Cells
TGFβ stimulates the transcription of a range of its target genes through both Smaddependent and -independent pathways [15,17]. To better understand the pharmacological effect of SCE and SchB on vascular fibrosis, we performed DNA microarray experiments in A7r5 cells. Through feature selection, we identified 9549 DEGs across six experimental groups. Hierarchical clustering analysis and PCA demonstrated that these six groups were clustered into discrete ones (Figure 2a,b). These results indicate that each group shows distinctive changes in the gene expression profile.

SCE and SchB Reverse TGFβ-Induced Changes in the Gene Expression Profile
To identify DEGs among experimental groups, we performed Limma analysis in a pairwise way and presented the results as Venn diagrams (Figure 3a,b,d,e) or heatmaps (Figure 3c,f). We found 5521 DEGs (1969 up-regulated and 3552 down-regulated) in TGFβ against vehicle (Tables S1 and S2) and we defined these genes TGFβ-reactive genes. Additionally, we found 3838 DEGs (2894 up-regulated and 944 down-regulated) in TGFβ + SCE against TGFβ (Tables S3 and S4), and 851 DEGs (200 up-regulated and 651 down-regulated) in TGFβ + SchB against TGFβ (Tables S5 and S6). On the other hand, we found 6132 DEGs (3999 up-regulated and 2133 down-regulated) in SCE versus vehicle (Tables S7 and S8) and 3012 DEGs (1275 up-regulated and 1737 down-regulated) in SchB versus vehicle (Tables S9 and S10).    To assess the pharmacologic action of SCE and SchB at the transcriptome level, we identified the TGFβ-reactive genes but reversed by SCE. We counted the DEGs up-regulated by TGFβ, but down-regulated by TGFβ+SCE or vice versa (red circles, inter-sections between two bottom Venn diagrams in Figure 3a,b). Then, we excluded the DEGs by SCE solely (top Venn diagrams in Figure 3a,b) to cancel the effect of SCE irrespective of TGFβ. Therefore, we found that SCE reversed the expression of 855 out of 5521 TGFβ-reactive genes (199 out of 1969 up-regulated genes and 656 out of 3552 down-regulated genes, highlighted in light blue in Figure 3a-c, Table S11). We defined these genes SCE-effective genes. Likewise, SchB reversed the expression of 157 out of 5521 TGFβ-reactive genes (124 out of 1969 up-regulated genes and 33 out of 3552 down-regulated genes, highlighted in light blue in Figure 3d,e, Table S12). We defined these genes SchB-effective genes. The number of SCE-effective genes, 855 (199 + 656) is much larger than the number of SchBeffective genes, 157 (124 + 33). These results indicate that SCE has a broader pharmacologic effect than its active component SchB in terms of the TGFβ-induced transcriptome.
Among SCE-and SchB-effective genes (i.e., TGFβ-reactive genes whose expression levels are reversed by SCE and SchB), 10 genes are commonly up-regulated, and 35 genes are down-regulated by TGFβ + SCE or TGFβ + SchB versus TGFβ (Figure 4a-c). Among these 45 genes, SGK1, and CAMK2D have been known to play crucial roles in vascular fibrosis [33,34]. In addition, many other genes, such as NRG4, SCYLI, MYH2, FXYD5, NCR1, LCN1, and FCGR2B, have been reported to be associated with fibrosis in various tissues, including liver, heart, and lung [35][36][37][38][39][40][41]. In protein-protein interaction network analysis on the common DEGs between SCE-and SchB-effective genes, a total of 29 connections (edges) were identified among 45 proteins (nodes) with average node degree of 1.35 (Figure 4d). Gene Ontology cellular component identified that the identified nodes are involved in vesicle coat-and plasma membrane-associated functions with FDR q-value < 0.05. These results suggest that our computational analysis can be useful for discovering novel therapeutic targets or predictive markers for TGFβ-mediated vascular fibrosis. analysis on the common DEGs between SCE-and SchB-effective genes, a total of 29 connections (edges) were identified among 45 proteins (nodes) with average node degree of 1.35 (Figure 4d). Gene Ontology cellular component identified that the identified nodes are involved in vesicle coat-and plasma membrane-associated functions with FDR qvalue < 0.05. These results suggest that our computational analysis can be useful for discovering novel therapeutic targets or predictive markers for TGFβ-mediated vascular fibrosis.

Functional Assessment of the Biological Effect of SCE and SchB by GSEA
To assess the molecular signatures and biological effects of SCE and SchB on TGFβinduced phenotypic changes, we performed GSEA using Hallmark gene signatures on the 9549 DEGs that were identified through feature selection. We found that SCE and SchB reversed TGFβ-induced up-regulated Hallmark gene signatures, such as EPITHE-LIAL_MESENCHYMAL_TRANSITION, TGF_BETA_SIGNALLING, TNFA_SIGNALING_ VIA_NFKB, and MTORC1_SIGNALING signatures, but that TGFβ-induced down-regulated Hallmark signatures, such as E2F_TARGETS, G2M_CHECKPOINT, and FATTY_ACID_ME TABOLISM signatures, compared to vehicle (Figure 5a-d, Tables S13 and S14). The Hallmark signature 'TGF_BETA_SIGNALLING', as an internal control, was readily reversed by SCE or SchB (Table S13). These results demonstrate that SCE and SchB potently suppress the changes in a range of TGFβ-induced biological processes, which can contribute to their anti-fibrotic activity.
regulated Hallmark signatures, such as E2F_TARGETS, G2M_CHECKPOINT, and FATTY_ACID_METABOLISM signatures, compared to vehicle (Figure 5a-d, Tables S13 and S14). The Hallmark signature 'TGF_BETA_SIGNALLING', as an internal control, was readily reversed by SCE or SchB (Table S13). These results demonstrate that SCE and SchB potently suppress the changes in a range of TGFβ-induced biological processes, which can contribute to their anti-fibrotic activity.

Master Regulators (MRs) That Regulate SCE-and SchB-Effective Genes
For a mechanistic understanding of the transcriptional regulation by SCE and SchB, we constructed an A7r5-specific transcriptional interactome using ARACNe algorithm and performed MRA. ARACNe inferred a small consensus network of 17,645 interactions among the 891 Rattus norvegicus TF hub markers (genes) and the 14,004 genes. From this

Master Regulators (MRs) That Regulate SCE-and SchB-Effective Genes
For a mechanistic understanding of the transcriptional regulation by SCE and SchB, we constructed an A7r5-specific transcriptional interactome using ARACNe algorithm and performed MRA. ARACNe inferred a small consensus network of 17,645 interactions among the 891 Rattus norvegicus TF hub markers (genes) and the 14,004 genes. From this transcriptional interactome, the MRA predicted 38 or 2 TFs as MR candidates that regulate the expression of SCE-or SchB-effective genes, respectively (Figure 6a,b, Tables 1 and 2).

Discussion
We have reported the effect of SCE and its ingredient SchB on the Smad-dependent and -independent TGFβ signaling cascades in VSMCs [20][21][22]. However, little is known about their effect on TGFβ-induced transcriptional output, which is crucial for shaping fibrotic responses. In this study, our computational analysis demonstrated that SCE and SchB considerably reverse TGFβ-induced changes in terms of transcriptional output. In addition, we aggregated the public microarray data obtained from the experiments using VSMCs to identify TFs that act as MRs to regulate SCE-and SchB-effective genes. Therefore, the current paper will provide the basis of future research for understanding the pharmacologic actions of SCE and SchB in terms of gene expression regulation.
It has been known that inflammation and VSMC proliferation and migration are crucial features for vascular fibrosis [2,3]. Our computational analysis indicates that SCE and SchB attenuates inflammation processes (e.g., TNFA_SIGNALING_VIA_NFKB, IL2_STAT5_SIGNALING, INFLAMMATORY_RESPONSE, INTERFERON_GAMMA_RESP ONSE) and cell migration and proliferation processes (e.g., EPITHELIAL_MESENCHYMAL _TRANSITION, MTORC1_SIGNALING, KRAS_SIGNALING) ( Figure 5). In fact, SCE and SchB has been known to inhibit the inflammatory responses and synthetic phenotypes of VSMCs [19], confirming the usefulness of our transcriptomic approach. These results suggest that our work can contribute to developing predictive markers of efficacy of anti-TGFβ or anti-fibrosis therapies. In addition, our analysis raised a new possibility that SCE and SchB regulate the unfolded protein response and hedgehog signaling, which can pave a way to understand their novel pharmacological actions ( Figure 5).
Our bioinformatic analysis found that SCE and SchB partially reversed TGFβ-induced changes in the gene expression profile. On the other hand, we have confirmed that SCE and SchB almost completely inhibited TGFβ-induced Smad phosphorylation and reporter gene activity [20,21]. These results suggest that SCE and SchB primarily reverse the Smad-dependent transcriptional program in TGFβ-treated cells, and partly affect the non-canonical pathway-dependent transcriptional program. Therefore, our results provide a crucial clue for dissecting specific signaling pathways that are affected by SCE and SchB.
In traditional medicine, the crude extracts of plants have been preferentially used over their isolated ingredients. Because the crude extracts consist of various active ingredients and these ingredients can produce synergistic effects, it has been considered that the crude extracts usually exert greater pharmacological activity than their active ingredients. In line with this notion, we obtained evidence that SCE has a broader pharmacologic effect than its active component SchB at the transcriptome level. These results provide insight into future research for developing therapeutic strategies.
In summary, this study showed that SCE and SchB effectively reversed TGFβ-induced transcriptome changes in VSMCs. These results provide novel insight into future translational research for clinical application and for the development of healthcare strategies.  Table S11: List of 855 SCE-effective genes (red, up-regulation; blue, down-regulation); Table S12: List of 157 SchB-effective genes (red, up-regulation; blue, down-regulation); Table S13: List of TGFβ-induced upregulated Hallmark gene signatures that are reversed by SCE or SchB; Table S14: List of TGFβ-induced down-regulated Hallmark gene signatures that are reversed by SCE or SchB; Table S15: List of RelA target genes predicted by MRA.