Glibenclamide, ATP and metformin increases the expression of human bile salt export pump ABCB11

Background: Bile salt export pump (BSEP/ABCB11) is important in the maintenance of the enterohepatic circulation of bile acids and drugs. Drugs such as rifampicin and glibenclamide inhibit BSEP. Progressive familial intrahepatic cholestasis type-2, a lethal pediatric disease, some forms of intrahepatic cholestasis of pregnancy, and drug-induced cholestasis are associated with BSEP dysfunction. Methods: We started with a bioinformatic approach to identify the relationship between ABCB11 and other proteins, microRNAs, and drugs. A microarray data set of the liver samples from ABCB11 knockout mice was analyzed using GEO2R. Differentially expressed gene pathway enrichment analysis was conducted using ClueGo. A protein-protein interaction network was constructed using STRING application in Cytoscape. Networks were analyzed using Cytoscape. CyTargetLinker was used to screen the transcription factors, microRNAs and drugs. Predicted drugs were validated on human liver cell line, HepG2. BSEP expression was quantified by real-time PCR and western blotting. Results: ABCB11 knockout in mice was associated with a predominant upregulation and downregulation of genes associated with cellular component movement and sterol metabolism, respectively. We further identified the hub genes in the network. Genes related to immune activity, cell signaling, and fatty acid metabolism were dysregulated. We further identified drugs (glibenclamide and ATP) and a total of 14 microRNAs targeting the gene. Western blot and real-time PCR analysis confirmed the upregulation of BSEP on the treatment of HepG2 cells with glibenclamide, ATP, and metformin. Conclusions: The differential expression of cell signaling genes and those related to immune activity in ABCB11 KO animals may be secondary to cell injury. We have found glibenclamide, ATP, and metformin upregulates BSEP. The mechanisms involved and the clinical relevance of these findings need to be investigated.


Introduction
The bile salt export pump (BSEP), the major bile salt transporter in the liver canalicular membrane, is coded by ABCB11 gene, and mutations in this gene cause progressive familial intrahepatic cholestasis type-2 (PFIC-2) 1,2 . Besides PFIC-2, mutations or insufficiency of BSEP is associated with a variety of diseases such as drug-induced cholestasis, pregnancy induced cholestasis, cryptogenic cholestasis, cholangiocarcinoma and hepatocellular carcinoma, which are cancers of the liver 3-7 . Naturally, ABCB11 expression is induced by bile salts and is mediated by FXR-RXR heterodimer 8 . Here in this pilot study we explored in silico the interactions/networks around ABCB11. We wanted to identify the genes, drugs, microRNAs which might influence the expression of ABCB11. Drugs which could upregulate ABCB11 expression may be useful in ABCB11 haploinsufficiency and inhibition of the pump could result in the accumulation of toxic bile salts inside hepatocytes. Modulation of ABCB11 expression could be clinically beneficial in a variety of medical conditions.

Identification of differentially expressed genes
We analyzed the microarray data set of the liver samples from ABCB11 knockout mice (GEO accession GSE70179) using GEO2R online tool from NCBI 9 . All differentially expressed genes (DEGs) were filtered with two criteria: -1> log 2 FC >+1 and adj. p-value <0.05.

Pathway enrichment analysis
To identify DEGs which are significant, pathway enrichment analysis was conducted using the ClueGo v2.5.5 app from Cytoscape 10 . ClueGo constructed and compared networks of functionally related GO terms with kappa statistics, which was adjusted at >0.4 in this study.

Identification of hub genes and subnetwork analysis
The protein-protein interaction (PPI) networks were built by the Search Tool for the Retrieval of Interacting Genes (STRING v11.0) 11 and Cytoscape v3.7.1 software. The Molecular Complex Detection (MCODE v1.6), app from Cytoscape was used to screen modules of the PPI network with degree cut-off = 2, node score cut-off = 0.2, k-core = 2, and maximum depth = 100. The hub genes were identified by the CytoHubba v0.1 app. The top 10 nodes were considered as notable hub genes and displayed.

Western blot analysis
Total proteins from HepG2 cells were prepared and run on 10% SDS-PAGE and transferred to a PVDF membrane using a transfer apparatus following the standard protocols (Bio-Rad). After incubation with 5% nonfat milk in TBST (10 mM Tris, pH 8.0, 150 mM NaCl, 0.5% Tween 20) for 1 h the membrane was washed once with TBST and incubated overnight at 4°C with rabbit antibodies against human ABCB11 (Affinity, Catalog #DF 9278) 1: 2000 dilution; mouse anti-human β-actin (Santa Cruz Cat.# SC4778), dilution 1:1000. The membrane was washed three times (TBST) and incubated with a 1:5000 dilution of horseradish peroxidase-conjugated anti-rabbit (Santa Cruz Cat# SC-2004)/anti-mouse antibodies (Cat.#SC-2005) for 2 h. Blots were washed with TBST four times and developed with the ECL system (Bio-Rad, US Cat.#170-5060) according to the manufacturer's protocol. The western blot images were acquired using iBright CL1000 (Invitrogen, Thermo Fisher Scientific).

Real-time PCR
Total RNA was isolated using NucleoZOL (Takara Cat. No. 740404.200) following manufacturer's instruction. cDNA was prepared from (deoxyribonuclease treated) total RNA using RevertAid Reverse Transcriptase (Thermo Cat. No. EP0441) following the manufacturer's instructions. Real Time PCR was done with unique oligonucleotide primers targeting ABCB11 and GAPDH, Ta=60°C, in triplicates and two repeats, using GoTaq® qPCR Master Mix (Promega Cat. No. A6001) following 'manufacturer's instructions on a Veriti Thermo Cycler from Applied Biosystems Waltham, Massachusetts, USA and data was acquired using the software associated with the same machine (ViiA7 V1.2) and relative quantification was calculated using the by 2 (-ΔΔCt) method. Oligonucleotide primer sequences are listed in Table 3.
An earlier version of this article can be found on biorxiv.org (DOI: https://doi.org/10.1101/2020.09.01.277434). genes (DEG) from the GSE dataset were classified in two groups -upregulated (375 genes) and downregulated (185 genes) (Extended data, Supp. Table-1) 36 . Gene ontology analysis was performed for functional analysis of DEGs by using ClueGo app from Cytoscape. PPIs of DEGs were constructed using STRING database showed an upregulation of genes related to cellular transport (pink colored nodes), and these nodes were also shared by Toll-like receptor (TLR) signalling ( Figure 1). Downregulated genes were involved in metabolic pathways (sterol, carbohydrate, alcohol, etc.) (Extended data, Supp. Table-2) 36 . We next identified top hub genes in PPI network using CytoHubba app from Cytoscape (Table 1). Immunologically important genes were among the top ranked upregulated hub genes ( Figure 2a) downregulated group majorly represents cell signaling and fatty acid metabolism ( Figure 2b). Epidermal growth factor receptor (EGFR) ranked first among the genes involved in signaling pathways. Kinases play a role in the transcription, activity, or intracellular localization of ABC transporters as do protein interactions 16 . Proteins interacting with ABCB11 are represented in Figure 3 which includes nuclear receptors NR1H4 and NR0B2. Most proteins were associated with bile acid metabolism and transport.
As described, sub-network analysis was performed using MCODE ( Figure 4), and CMPK2, ACTG1, and SSTR2 emerged as seed nodes among upregulated genes (Table 2). Among downregulated gene groups, only one subnetwork was found to be significant which had three genes: MIA3 (which codes a protein which is important in the transport of cargos that are too large to fit into COPII-coated vesicles such as collagen VII), IGFBP4 (encoding a protein that binds to both insulinlike growth factors and modifies their functions) and NOTUM (encoding a carboxylesterase that acts as a key negative regulator of the Wnt signaling pathway by specifically mediating depalmitoleoylation of WNT proteins).
Using CyTargetLinker identified two drugs, glibenclamide, and ATP, directly targeting ABCB11. We subsequently looked for microRNAs [Target-scan database] 17 that were associated with ABCB11, and a total of 14 microRNAs were identified targeting the gene ( Figure 5). Transcription factors and microRNAs targeting ABCB11 and interacting partners are represented in Figure 6.

Glibenclamide ATP and Metformin upregulates ABCB11
We evaluated in vitro, the effect of three drugs, two of which were bioinformatically predicted (Glibenclamide, ATP) and one based on literature 18 . We found all the three compounds upregulating ABCB11 expression based on qPCR, and this was confirmed by western blot (Figure 7). Unannotated western blot images and raw qPCR Ct values are available as Underlying data 36 .

Discussion
We identified several immunologically important genes being upregulated during ABCB11 deficiency. The reason could be liver cell injury secondary to bile salt accumulation, which triggers the sterile immune response 19,20 and the downregulation of transport proteins and metabolically important genes could be because of decreased liver function following damage. A regenerative response follows cell injury, and a host of genes involved in regeneration are upregulated 21-23 ; however, it appears that bile salts in the absence of BSEP hamper the regenerative response reflected by dysregulated collagen transporting protein MIA3 and NOTUM a protein involved in Wnt signaling. It's also possible that EGFR is dysregulated via accumulating bile salts mediated by STAT3 24 . We have observed an upregulation of ABCB11 in a liver cell line (HepG2) on treatment with glibenclamide, metformin, and ATP. This expression is upregulation may be a compensatory mechanism in the case of glibenclamide and metformin because these drugs are known to inhibit ABCB11 25 . Metformin is known to interfere with ABCB11 function, mediated through AMPK-FXR crosstalk 18 involving metformin induced FXR phosphorylation. ATP acts through ATP receptors on hepatocytes 26,27 . ATP is known to cross the plasma membrane 28 and this can act via AMPK. However, ATP has a very short half-life 29 , and it may be converted to ADP, which can activate AMPK 30 . In a recent report, metformin was shown to suppress ABCB11 expression, which is not in agreement with our observation, however, they performed their experiment on primary human hepatocytes, Table 1. Genes with the greatest changes in expression. We observed that the top ranked hub genes in PPI network which were upregulated were associated with immune activity while those downregulated are associated with cell signaling and fatty acid metabolism. EGFR came first in the ranking which is a critical receptor in several cell signaling pathways.

Rank Gene
UniportKB/Swiss-Prot Function  This network was constructed to analyze the relationship between ABCB11 and other proteins. Cytohubba app was used to calculate centrality of each node by MCC method. Node colour (red to yellow) represents the significance of the centrality in the group. In this analysis, we counted 11 nodes and 42 edges. These proteins majorly involved in bile acid metabolism and transport. Most of these genes are participant of more than one pathway which was expected because these pathways intersect and coregulated. We also mapped the NR0B2 protein, which is participate in sterol metabolism.

Figure 2. We identified top hub genes in PPI network of both upregulated and downregulated genes by using CytoHubba app from Cytoscape.
We observed that the top ranked hub genes in the upregulated group were mainly related to immune activity (a). The top hub genes in downregulated group were associated with cell signaling and fatty acid metabolism. EGFR emerged as the top hub gene, a growth factor receptor which is crucial factor several cell signaling pathways (b).

Figure 4. Sub-network analysis was conducted by using the Molecular Complex Detection (MCODE) app from Cytoscape.
Top sub-networks on the basis of MCODE score (Degree cut-off= 2, node score cut-off = 0.2, k-core = 2 and max. depth = 100). Upregulated gene group clusters, we identified seed nodes (CMPK2, ACTG1 and SSTR2) in the network (green and blue). In downregulated gene group, we identified only one subnetwork which qualified cut off criteria. Three genes in this sub-network was identified: MIA3, IGFBP4 and NOTUM (red).

ABCB11
CCTCCATCCGGCAACGCT CACTGAATTTCAGAATCCTCCTAACTGGG Figure 5. CyTargetlinker was used to screen microRNAs using the Target-scan database. We identified microRNAs that were associated with ABCB11. In total 14 microRNA identified targeting the gene. Others can be investigated in future studies. We counted 55 nodes and 89 edges in the search of microRNA targeting the ABCB11 network. Four genes (ABCB11, ATP8B1, SLC10A2 and NR1H4) targeted by multiple microRNAs also some microRNA such as has-miR-203a-3p.2 and has-miR-203a-3p.2 target more than one gene. By nature, a microRNA can regulate several pathways therefore it would be interesting to study in future the dysregulation of these microRNAs and interaction with Identified transcription factors. and they have also treated their cells with dimethylsulfoxide (DMSO) 31 .
There are many reports stating the influence of DMSO on human gene expression. For example, Verheijen et al. "exposed 3D cardiac and hepatic microtissues to medium with or without 0.1% DMSO and analyzed the transcriptome, proteome and DNA methylation profiles". They found that "in both tissue types, transcriptome analysis detected >2000 differentially expressed genes affecting similar biological processes, thereby indicating consistent cross-organ actions of DMSO". In both tissue types, the transcriptome analysis detected over 2000 differentially expressed genes affecting similar biological processes 32 . Moskot et al. reported alterations of lysosomal ultrastructure upon DMSO treatment 33 . Alizadeh et al. reported that DMSO catalyzes hepatic differentiation of adipose tissuederived mesenchymal stem cells 34 . It has been observed that "culturing pluripotent stem cells in DMSO activates the retinoblastoma protein, increases the proportion of cells in the early G1 phase of the cell cycle, and subsequently improves their competency for directed differentiation into multiple lineages in more than 25 stem cell lines" 35 . However, we are not sure whether the observed difference is attributed to DMSO.
In conclusion, we need more experiments to determine the mechanisms of action of these drugs on the upregulation of ABCB11. Many changes in gene expression following ABCB11 knockout could be secondary to stress, immune and regenerative responses following hepatocyte injury in mice liver.  • realtime and western blottt (1) In this manuscript the authors have used the data on hepatic gene expression submitted to NCBI by Zhang Y, Neale G and Schuetz to determine the effect on the haplodeficiency of ABCB11 (BSEP) in mouse liver. The 375 upregulated genes included those involved in cellular transport and innate immunity (e.g. IFN signaling), whereas the 185 downregulated genes included signal transduction proteins, e.g. epidermal growth factor receptor and those involved with metabolic pathways. In addition, they have examined the effect of Glibenclamide, metformin and ATP on the expression of human ABCB11 in the human cell line HepG2. All three chemicals were found to increase ABCB11 expression.

Data availability
Comments: The information gene pathway analysis provides useful information as it has revealed "nodes" and "hubs" that mediate the up regulation and down regulation of the genes the expression of which is dysregulated in deficiency of ABCB11 function. However, as the original data set had been derived from whole mouse liver, it comprises gene expression by hepatocytes as well as nonparenchymal cells. It is known that bile acid accumulation can affect gene expression in various cell types in addition to hepatocytes (such as lung cells). Perhaps, the authors can mention this in the Discussion section.
A second concern is that HepG2 cells are not the most appropriate cell line for modeling the induction of ABCB11 in human hepatocytes, because, unlike some other human hepatoma cell lines, these cells lack the most abundant hepatocyte microRNA, miR-122. miR-122 is known to target various genes, thereby enhancing IFN signaling. On the other hand IFN downregulates the expression of miR-122. Thus, transcriptional induction of ABCB11 in HepG2 cell may not parallel that in primary human hepatocytes. Therefore, the drug induction study should be validated in a different human hepatoma cell line, or at least, the authors should discuss this complexity in interpreting the results.

Is the work clearly and accurately presented and does it cite the current literature? Yes
Is the study design appropriate and is the work technically sound? Partly

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Partly