4-Methylumbelliferone Targets Revealed by Public Data Analysis and Liver Transcriptome Sequencing

4-methylumbelliferone (4MU) is a well-known hyaluronic acid synthesis inhibitor and an approved drug for the treatment of cholestasis. In animal models, 4MU decreases inflammation, reduces fibrosis, and lowers body weight, serum cholesterol, and insulin resistance. It also inhibits tumor progression and metastasis. The broad spectrum of effects suggests multiple and yet unknown targets of 4MU. Aiming at 4MU target deconvolution, we have analyzed publicly available data bases, including: 1. Small molecule library Bio Assay screening (PubChemBioAssay); 2. GO pathway databases screening; 3. Protein Atlas Database. We also performed comparative liver transcriptome analysis of mice on normal diet and mice fed with 4MU for two weeks. Potential targets of 4MU public data base analysis fall into two big groups, enzymes and transcription factors (TFs), including 13 members of the nuclear receptor superfamily regulating lipid and carbohydrate metabolism. Transcriptome analysis revealed changes in the expression of genes involved in bile acid metabolism, gluconeogenesis, and immune response. It was found that 4MU feeding decreased the accumulation of the glycogen granules in the liver. Thus, 4MU has multiple targets and can regulate cell metabolism by modulating signaling via nuclear receptors.


Introduction
In cell culture, 4-methylumbelliferone (4MU) is a coumarin derivative well-known for inhibiting the synthesis of hyaluronic acid (HA). Moreover, 4MU, under the trade name "Hymecromone", is a prescribed drug for the treatment of cholestasis. Studies in different animal models clearly show that 4MU treatment leads to a wide range of biological effects summarized in reviews [1,2]. Further, 4MU administration inhibits tumor growth and metastasis [3][4][5], fibrosis development [6,7], and inflammation [1,8,9]. A broad spectrum of 4MU biological effects can be explained by inhibiting hyaluronic acid production, but pharmacological targets of 4MU are still under debate. Direct interaction of 4MU and hyaluronan synthase enzyme was not demonstrated. Moreover, 4MU does not inhibit bacterial HAS activity in membrane preparations after the disintegration with ultrasound [10]. Cell-free hyaluronic acid synthesis by membrane preparations from human fibroblasts was not affected by 4MU, too [11]. These experimental facts point towards the possible indirect effect of 4MU on HA synthesis.
It was suggested that 4MU suppresses HA production by depleting the intracellular pool of UDP-glucuronic acid, a precursor of HA [12]. 4MU is a competitive substrate for UGT-glycosyltransferase (UGT), an enzyme which generates precursors for HA synthesis, UDP-glucuronic acid (UDP-GlcA) and UDP-N-acetylglucosamine (UDP-GlcNAc). Recently published data suggested that this interpretation may be insufficient. Nagy et al. demonstrated that in vivo administration of 4MU significantly decreased the level of UDP-GlcA but did not reduce the total concentration of HA in mouse pancreas, muscle, and liver [13]. The fact that 4MU does not inhibit chondroitin sulfate and heparan sulfate synthesis, where both HA precursors are involved, contradicts the UDP-GlcA depletion hypothesis [14,15]. According to published studies and the data from our laboratory, 4MU exerts a dual action on mammalian cells: at low micromolar concentration, it inhibits the HA synthesis, but in concentrations higher than 200 µm, it strongly affects multiple gene expressions, including HAS2 [6,15,16]. Our transcriptomic analysis demonstrated strong effects of 4MU on gene expression in hepatocytes, the cells which do not express HAS enzyme [6]. All these data point to multiple biological effects of 4MU.
Recently published data also indicates that 4MU administration induces a metabolic switch in different cell types, and this effect is not directly linked to HA inhibition. It was shown that 4MU treatment inhibits glycolysis in chondrocytes [8,9] and melanoma cell lines [27]. At present direct biological targets and mechanistic explanations of the biological effects of 4MU are missing. This hampers the development of new, more specific, and more potent drugs than hymecromone.
Here, we present an analysis of publicly available data bases: 1. Summary of 4MU targets identivied by small molecule libraries screening, depositted in PubChemBioAssay; 2. GO pathway analysis of identified 4MU target proteins; 3. Analysis of 4MU target protein expression across various human cell types (Single Cell and HPA datasets from Protein Atlas). Database analysis and our experimental data reveal that 4MU has multiple targets and can regulate cell metabolism through the modulation of nuclear receptor signaling.  [28]. The dataset for analysis was collected manually from the following queries: 4-methylumbelliferone (4MU, 1676 records), 4-methylumbelliferyl-beta-D-glucuronide (4MUG, 71 records) and 4-methylumbelliferyl sulfate (4MUS, 15 records), all henceforth jointly referred to as 4MU targets. Each record describes the effect of 4MU observed in small molecule libraries screening experiments targeting a particular protein or process activity. We excluded entries where 4MU did not have any activity, partially curated data records and experiments dedicated to the antibacterial properties of 4MU. After parsing the initial targets through the exclusion criteria, the 51 records obtained were sorted into either Active or Inconclusive sets based on PubChem BioAssay activity score. The final result after deduplication of 4MU and 4MUG common targets (HSD17B4, HSD17B10 for "Active" and GBA, HPGD for "Inconclusive") contained 22 "Active" targets and 39 "In-conclusive" targets 10 of which were common in both sets ( Figure 1B). Using the Gene Ontology database, we clustered the target proteins by function into three main groups namely: Enzymes (23 records), Transcription factors (TF; 20 records), and Others (nine records), mainly transport and scaffold proteins, as shown in Figure 1C,D and Table 1.  The key TF controls the protein quality in the endoplasmic reticulum. It interacts with the PPARA-RXR complex and regulates liver fatty acid oxidation and glucose metabolism.

Analysis of 4 MU Targets and Open Access Data Bases
[ [58][59][60] GLI3 Zinc finger protein GLI3 Mediates Sonic hedgehog signalling in many tissues. It represses lipid synthesis in the liver at a transcriptional level, and disruption of this pathway leads to steatosis. It is involved in immune cell development and maturation. [61,62]

HIF1A
Hypoxia-inducible factor 1-alpha The primary sensor of oxygen supply. Activation of the HIF-pathway shifts glucose metabolism from oxidative phosphorylation to glycolysis. [63]

NFE2L2
Nuclear factor erythroid 2-related factor 2 Involved in immediate reaction to oxidative stress, hypoxia, toxins and infections [64] NFkB Nuclear factor NF-kappa-B p105 subunit The key TF orchestrates inflammatory responses to different stimuli. [65] NR1H4 Farnesoid X receptor The primary receptor for bile acids, which is involved in the regulation of bile acid synthesis and degradation, glucose and fatty acid metabolism [66][67][68] NR1I3

Constitutive androstane receptor
The primary xenobiotic sensor. Natural ligands include androstanes and their derivatives. It regulates the expression of genes involved in drug and xenobiotic transformation and excretion. Its activation inhibits gluconeogenesis in hepatocytes and stimulates their proliferation. [69]

NR3C1
Glucocorticoid receptor Mediates cortisol and others glucocorticoids effects and controls genes involved in the development, metabolism and immune response. [70] PPARD PPARG Peroxisome proliferator-activated receptor delta Peroxisome proliferator-activated receptor gamma The nuclear receptors for free fatty acids, eicosanoids and vitamin B3.
They have a central role in the regulation of glucose and lipid homeostasis [71] RXRA Retinoid X receptor alpha Serves as the main heterodimerisation partner for type II nuclear receptors (LXR, PPARs, VDR, NR1H4 and NR1I2). RXRA is involved in regulating fatty acid and cholesterol metabolism, innate immunity and cellular senescence. [72,73] SMAD2 SMAD3 Mothers against decapentaplegic homolog 2 Mothers against decapentaplegic homolog 3 The Smad-signaling pathway, which operates downstream of the transforming growth factor-β (TGF-β) superfamily of ligands, regulates a diverse set of biological processes, including proliferation, differentiation and apoptosis, in many different organ systems [74][75][76]

4MU Target Proteins Revealed by Analysis of Affected Molecular Pathways
For the investigation of the involvement of 4MU target proteins in biological processes and pathways, we used the GO Database terms. TFs were mainly identified using DNAtemplated RNA transcription through cis-regulatory regions. Among identified TFs, 13 belong to the superfamily of nuclear receptors binding lipophilic ligands. The Enzymes group included enzymes primarily involved in steroid hormone metabolism, monoterpenoid metabolic processes, exogenous drug metabolism, long-chain fatty acid biosynthetic processes, and apoptosis. These enzymes possess hydroxylase, oxidoreductase, hydrolase and caspase activities "Others" group included proteins involved in nuclear membrane reassembly, organisation, and nuclear import with transmembrane transporter activity, RNA cap, and dynactin binding (Figure 2A,B). The type of quantitative High throughput screens performed (qHTS) are highlighted and were accessed in PubChem BioAssay using the AID numbers shown in Table 2.
Treatment with 4MU activates cellular stress response pathways mediated by TP53, HIF1A, NFE2L2, and ATF6 and initiates NFkB-dependent transcription ( Table 2). It also activates RXRA, the main dimerization partner for type I nuclear receptors, and mediates the biological effects of retinoic acid. Such pattern of TFs indicates activation of the defence program against different stress stimuli followed by a shift in fatty acid and glucose metabolism. 4MU has a dual effect (activation and inhibition) on 12 TFs (Table 2), mainly nuclear receptors. Furthermore, 4MU elicits different responses depending on the cell line and experimental conditions as is the case with VDR and NR3C1 (glucocorticoid receptor) which are well-known modulators of the immune response. In addition, 4MU stimulated the activity of enzymes, such as CASP3, GALC, KLK7, GAA, and GBA, which modulate protein processing, cell motility, membrane glycolipid composition and glycogen breakdown. In this Enzymes group, KLK7 peptidase was activated by the lowest concentration of 4MU (7 µM). KLK7 peptidase is involved in skin shedding, cancer and Alzheimer's disease progression [90][91][92].
It was found that 4MU inhibited 18 out of 24 enzymes (Table 2). These enzymes are involved in lipid and drug metabolism, oxidative stress response, cytokine activation, transcription and DNA repair processes. CA9 was inhibited by the lowest concentration of 4MU (0.56 µM). CA9 is abundantly expressed in many cancer types where it maintains the normal pH level in tumor cells in the hypoxic environment [93].
Transcription activity of GLI and NR1I3 and nuclear imports mediated by SNUPN and KPNB1 from the other groups was also inhibited by 4MU. Interestingly, we found three proteins in this group specifically involved in the development of inherited neurological disorders: In vitro, 4MU activated the phosphorylation of H2AX on Ser-139, a sensitive marker of DNA double strand break [95], and stimulated the expression of HSPB1, indicating the activation of the heat shock response pathway. These data correlate with ATF6α activation, which is the main sensor of unfolded protein stress in ER [58]. All these proteins from the "Others" group (HSPB1, LMNA, HTT, PMP22 and H2AX) can be considered indirect targets of 4MU.

4MU Target Proteins Revealed by Analysis of Gene Expression in Various Human Cell Types
The target gene expression of 4MU was evaluated in 51 human cell types from the Single Cell data of the Protein Atlas Database. The target genes 4MU fell into three clusters. Cluster 1 comprised of 11 genes; NR2H4, CYP1A2, KLK7, CA9, ESR2, PGR, HSD17B3, AR, CLI3, NR1I3, and ESR1. Expression of these genes varied significantly across cell types. Cluster 2 consisted of 26 genes, all highly expressed in almost all cell types. Cluster 3, composed of eight genes, namely ACHE, CA12, ABCG2, CASP1, VDR, TDP1, HPGD, and PPARG, showed a moderately high expression in some cell types. All gene clusters were detected in the cardiomyocytes in contrast to erythroid cells in which gene expression was undetectable ( Figure 3). These data indicate that 4MU alters metabolic pathways in a vast majority of cell types.
We also analysed 4MU target gene expression in immune cells using the data from the HPA dataset deposited in the Protein atlas database. Immune cells express 43 out of 45 genes which form three clusters. PGR and NR1H4 expression was undetectable ( Figure 4).
Cluster 1 includes eight genes whose expression is generally low and is specific to peripheral blood mononuclear cell (PBMC), non-classical, classical, and intermediate monocytes. The co-expression of ALDH1A1, NR1I3, and PPARG is characteristic of PMBC, myeloid dendritic cells (DC), macrophages, and memory CD4+ T cells. CA12, CA9, AR, ALDH1A1, and NR1I3 co expression is unique to plasmacytoid DC.
Earlier on, we highlighted that the type of response to 4MU elicited is dependant type of cell line and experimental conditions as is the case with VDR and NR3C1. NR3C1 is highly expressed in all immune cells and regulates innate and adaptive immunity. Generally, NR3C1 suppresses proinflammatory activation of macrophages, inhibits the production of cytokines (IL-1, IL-6, IL-8) and chemokines (Ccl2, Ccl3, Ccl4, Cxcl9, and Cxcl11) which promotes the acquisition of inflammation resolving phenotype [70]. Functional activity of NR3C1 is essential for T-cell polyclonal activation and maturation of either Th1-and Th2cell types [101]. It is also involved in the modulation of B-cell apoptosis [102].
VDR is a potent immune stimulator, upon activation boosts the innate immune cells' chemotactic and phagocytic capabilities. Moreover, VDR induces the transcription of cathelicidin and defensin β2 [78]. It was shown that VDR activation prevents nuclear translocation of p65/p50 subunits of NFkB and degrades IκBα protein, thereby inhibiting the inflammatory response [103]. Its activation suppresses autoimmunity driven by Th1 and Th17 cells [78]. The mild immunosuppressive properties of 4MU can be explained by the synergistic action of both VDR and NR3C1, which occurred in the 30-50 µM range according to PubChem BioAssay data analysis (Table 2).

Effect of 4MU Treatment on the Gene Expression Profile of Normal Mouse Liver
According to the previously published data, the liver is one of the most prominent target organs of 4MU action [13]. Our previously published results demonstrated a profound effect of 4MU on gene expression after two weeks of treatment and its cessation after four weeks in CCL4 induced murine model of hepatic fibrosis [6].
In this study, we present a detailed transcriptomic analysis of 4MU-dependent changes in a healthy liver. Bulk RNA-Seq data analysis showed a high degree of similarity between control and 4MU treated liver samples, as reflected by Spearman correlation coefficients ( Figure 5A). Principal component analysis (PCA) showed that 40.6% of sample variance originated from the first two components, which clearly separated the experimental groups into two clusters ( Figure 5A). Oral 4MU treatment of healthy animals has very mild effects on gene expression. We found 62 genes which were significantly up-(26 genes) and downregulated by 4MU (36 genes), ( Figure 5B). The complete list of DEGs with their relative expression levels is presented in Figure 5C,D and Table 3. Affected genes were annotated according to GO Biological Pathways database terms ( Figure 5E,F). Most of the downregulated genes (29 out of 36) are involved in different aspects of immune response ( Figure 5E). These genes encode the surface markers of different immune cells: Ly6d is a marker of B and dendritic cells, Ly6c1, Themis2, and Ms4a4b, are markers of T and NK cells; whilst Fcgr1 is characteristic of liver macrophages. Treatment with 4MU significantly decreases the expression level of 14 genes involved in interferon signalling and response (Tgtp1, Mx1, Epsti1, Ifit1, Ifit3, Ifit3b, Gbp3, Isg15, Il18bp, Usp18, Irf7, Oasl2, Ifi27l2a, Oas1a). The expression level of Naip1 was also significantly diminished by 4MU. Naip 1, whose role is to recognise bacterial proteins, is abundantly expressed in dendritic cells and macrophages. In the 4-MU treated mice, we observed a decreased expression of proteins (Cdc20, Mki67, Ccn5, Racgap1) involved in cycle progression, and maintenance of DNA integrity and stability (Top2a, Blm). Among other downregulated genes, we identified two transcriptional factors (Nfe2, Snai2), haemoglobin's chains genes (Hba-a2, Hbb-b1, Hbb-bt), retinol-binding protein 1 (Rbp1), Ig delta chain C region (Ighd) and mitochondrial Mg 2+ transporter (Slc41a3). Surprisingly, Cyp7a1 one of the critical enzymes of bile acid production was downregulated four-fold [104].  The upregulated genes are involved in the modulation of transcriptional activity (Bcl3, Spen), histone acetylation (Dot1l, Kmt2d), protein phosphatase activity (Dusp6, Ppp1r10), mitochondrial biogenesis and function (Lars2, Pprc1), extracellular matrix remodelling (Egfl7, Ntn3, Papln), muscle cells differentiation (Myoz2, Tmod4), and cancer progression (Sez6l2). Moreover, 4MU induced a nearly two-fold increase in the expression levels of transcriptional factors Hhex, Jun, Nfkbiz, and Zbtb16 (Figure 5C,D; Table 3). Hhex is an essential transcription factor for hepatoblast differentiation and intrahepatic bile duct morphogenesis [105]. Embryonic knockout of Hhex is lethal because it causes abnormal liver development. In adult mice, transcriptional activation of Hhex expression depends on Farnesoid X receptor (NR1H4) activation by bile acid. Overstimulation of this signalling pathway leads to liver hypertrophy [106]. Jun is involved in hepatocyte proliferation and liver regeneration [107]. It has been demonstrated that c-Jun is involved in the bile-acidinduced downregulation of Cyp7a1 expression [108]. Nfkbiz is a transcriptional regulator that augments the inflammatory responses from Toll-like receptors or interleukin signalling. The hepatocyte-specific knockout of this gene significantly accelerates the progression of non-alcoholic fatty liver disease in mice, while its overexpression protects the liver from steatosis. Microarray analysis revealed that the overexpression of Nfkbiz downregulates the expression of genes involved in the triglyceride metabolism pathway [109]. Zbtb16, also known as promyelocytic leukaemia zinc finger protein (PLZF), is a transcription repressor involved in energy metabolism maintenance and pathogenesis of metabolic diseases. In the adult liver, Zbtb16 is an essential regulator of gluconeogenesis. The activation of Zbtb16 expression induces the expression of vital gluconeogenic genes in vitro and in vivo [110]. Zbtb16 expression is activated by the glucocorticoid receptor (NR3C1) in many cell types, including hepatocytes [110], breast cancer cell line [111], endometrial stromal cells, and myometrial smooth muscle cells [112].
This finding is in concert with the observed depletion of glycogen granules in mice after treatment with 4MU, as shown in Figure 6. Figure 6. Glycogen granules in mouse hepatocytes after 4MU treatment. Gomori-Grocott methenamine silver staining of glycogen granules in mice hepatocytes after 2 (2w) and 4 weeks(4w) of 4MU treatment. In the figure, hepatocytes are light green and dark grey while glycogen granules are black. Scale bar-100 µm. The plot represents the percentage of glycogen-positive area to total tissue area. Statistical difference was estimated by one-way ANOVA with Tukey post-hoc test (****-p.val < 0.0001; ns-not significant).

Effect of 4MU on Glycogen Storage in the Liver
Public data analysis (Pub Chem Bio Assay, Table 2 assay id 2112) shows 4MU activation of acid alpha-glucosidase (GAA), a key enzyme involved in lysosomal glycogen metabolism. The half-maximal efficacy concentration of 4MU for GAA was 25.1 µM in vitro (Tables 1  and 2). We performed glycogen granules staining with Gomori-Grocott methenamine silver stain in healthy liver at two or four weeks after 4MU treatment n = 3 controls, n = 4 in 4 MU treated group. As anticipated, GAA activation by 4MU depleted hepatocyte glycogen granules after two weeks ( Figure 6). However, the glycogen granules increased after four weeks.

Discussion
There are two major processes of drug discovery, namely target-based drug discovery (TBDD) and phenotype-based drug discovery (PBD). TBDD is a predominant route taken by pharmaceutical companies developing drugs interacting with well-characterised targets. In contrast to TBDB, PBD evaluates a compound's ability to modulate a certain trait of a given biological system, the effect of a biologically active compound is based on the analysis of cell culture and whole-body phenotypes. Initially, the biological target and molecular mechanism of action are unknown. Nonetheless, effective drug development demands we identify the drug targets of the biologically active compounds.
In this study we identified 4MU targets by analysing data from publicly available databases coupled with our RNA-Seq transcriptomic profiling of bulk liver. We show that 4MU modulates the functional activity of at least 45 proteins, mainly TFs and enzymes. These proteins are expressed in 51 different human cell types, including immune cells, which explains the wide range of biological effects upon 4MU treatment. According to the data, 4MU modulates the activity of at least 13 nuclear receptors, namely the transcription factors which regulate various metabolic pathways and immune responses. Our analysis of publicly available data indicates that 4MU treatment leads to activation of cell stress response pathways via HIF1A, NFkB, NFE2L2 and ATF6 transcription factors. In addition, 4MU exposure activates several lysosomal enzymes involved in the degradation of glycolipids and glycogen (GAA, GBA, GALC and GLA).
In our previously published work [6], we showed that 4MU inhibits HAS2 expression and hyaluronan deposition in liver parenchyma during the development of liver fibrosis in mice. In the current study, we evaluate the effect of 4MU treatment in normal liver, which expresses very low levels of HAS2 [113]. Our data indicate that 4MU influences energy metabolism in a healthy liver by activating glycogen utilisation. Exposure of mice to 4MU for two weeks led to the depletion of glycogen granules in the liver. These findings are consistent with publicly available data of high content screening experiments targeting GAA, the main glycogen catabolic enzyme. The exhaustion of glycogen indicates an alteration in the liver's energy and carbohydrate metabolic signature. This finding is in correlation with the conclusion of the recently published extensive research on the effects of 4MU on energy metabolism in mice [20].
Liver transcriptome analysis of mice on normal and 4MU diet confirms this finding: 4MU strongly affected the expression of Zbtb16, one of the key regulators of hepatic gluconeogenesis [110]. Concurrently, 4MU treatment significantly affected the expression of genes involved in bile acid metabolism. Even though 4MU was introduced in some countries as a drug for cholestasis treatment more than 30 years ago, the precise mechanism of its action remains unknown. Our findings reveal an upregulation of Hhex and Jun, the transcription factors regulating the conversion of cholesterol into bile acids and a downregulation of Cyp7a1 which is the rate-limiting enzyme in bile acid synthesis via the classical pathway [104,106,108]. The expression of Cyp7a1 is tightly regulated; it is induced by feeding and suppressed by high levels of bile acid in the serum through activation of NR1H4 (Farnesoid X receptor). Such an autoregulatory loop controls the enterohepatic circulation of bile acids and maintains a constant circulating level of bile acid [114]. It has been shown that bile acid regulates the expression of c-Jun via nuclear receptor SHP [115], and c-Jun is involved in NR1H4-dependent downregulation of Cyp7a1 [108]. Our study revealed that 4MU treatment significantly modified the expression of two NR1H4-dependent genes, namely Cyp7a1 and Hhex. This finding is consistent with the PubChem BioAssay data, which clearly states that 4MU is a non-specific ligand for nuclear receptors, including NR1H4. We also observed an indubitable downregulation of immune-associated genes involved in interferon signalling.
PubChem BioAssay database revealed high-affinity specific inhibition of HSD17b3, HSD17b4, and HSD17b10 belonging to the HSD17B family of NAD(P)H/NAD(P)+dependent short-chain oxidoreductases. HSD17B1, HSD17B2, HSD17B3, HSD17B5, and HSD17B6 catalyse the interconversion between less potent 17-ketosteroids and more potent 17-hydroxysteroids to maintain the balance between 17-keto/17β-hydroxy forms of estrogens and androgens [116]. Other members of this family are involved in fatty acid metabolism, cholesterol biosynthesis, and bile acid production [117]. One member of this family, HSD17B13, is of paramount interest as a potential target for drug development. HSD17B13 was first cloned from the human liver cDNA library [118]. It was later shown that the HSD17B13 genetic polymorphism rs72613567, causing splicing alteration and loss of HSD17B13 activity, is associated with a substantial decrease in serum alanine aminotransferase and aspartate aminotransferase levels in a European population study. This polymorphism was also associated with a lower risk of NAFLD and non-alcoholic cirrho-sis [119]. These findings were later confirmed in an independent study by Pirola, C.J. et al. [120]. Although data on 4MU inhibition of HSD17B13 are missing from the PubChem BioAssay database, the high homology of HSD17 family members and inhibition of b3, b4 and b10 by low micromolar concentration of 4MU allow us to infer the possible inhibition of HSD17b13 activity by 4MU, which may contribute to its protective effect in liver fibrosis [6].
Inhibition of KAT2A by 4MU in low micromolar concentrations (Table 2) underpins the recent discovery of its involvement in biliary fibrosis in mice. KAT2A, also known as GCN5, is a histone acetyltransferase (HAT) that functions primarily as a transcription activator. It also represses NFkB activity by promoting the ubiquitination of the NFkB subunit RelA (p65) in a HAT-independent manner [121]. Both pharmacologic inhibition of KAT2A lysine acetyltransferase activity and cholangiocyte-specific deletion of KAT2A were protective in mouse models of biliary fibrosis [122].
TGF-β ligands that activate the SMAD-2/3 intracellular pathway are considered a major driver of human fibrotic pathologies [123][124][125]. The involvement of the TGF-β/SMAD3 pathway is well documented in hepatic stellate cell activation and collagen production [124]. SMAD3 signalling was mechanistically linked to the liver fibrosis development in mice. Specifically, the anti-fibrotic action of Umbelliferon, a close analogue of 4MU, was associated with inhibition of TGF-β SMAD3 signalling [126]. SMAD3 inhibition by 4MU was reported in four independent bioassays (Table 2). Together, these data justify the validity of our bioinformatic approach, which independently identified SMAD3 as a potential 4MU target.
By its nature the big data analysis presented here only suggested potential targets of 4MU which have to be individually validated in our ongoing research using in vitro and in vivo techniques, including gene knockout and knockdown.

Materials and Methods
Data collection. Publicly available data were collected from the PubChem BioAssay (U.S. National Library of Medicine, Bethesda, MD, USA) [28] database for the next CIDs: 5280567 (Hymecromone, 4MU), 84843 (4-methylumbelliferone sulfate, MUS), 91553 (4-methylumbelliferone glucuronide, MUG). Data were downloaded manually, and all subsequent data manipulations were performed with a custom-made R script in RStudio version 1.4.1106, R version 3.6.3 (Posid, Boston, MA, USA). Briefly, the collected dataset containing 1762 rows was filtered according to substance activity. Individual observations without target names (calibration controls) were excluded from further analysis. In the final step, we identified common 4MU targets and its derivatives based on protein name. Expression profiles of 4MU target proteins in different cell types were manually downloaded from the Human protein atlas database [127] (https://www.proteinatlas.org/; accessed on 3 June 2022) from Single Cell and HPA datasets.
Mice experiments. All animal procedures were conducted following the Russian Academy of Science Guidelines for Animal Experimentation and were approved by the Institute of Developmental Biology RAS Ethics committee. 8-week-old female Balb/c mice (18-20 g.) were obtained from "Scientific Centre for Biomedical Technologies" of the Federal Medical and Biological Agency, Moscow, Russia. Mice were fed a diet of natural ingredients and housed in a 12-h light/dark cycle with an ad libitum access to food and water at a room temperature of 22 • C and relative humidity of 55-65%. Mice were allowed to acclimatise to the conditions for 2-3 days prior to the experiment. The 4MU was purchased from Sigma-Aldrich (Cat # M1381, St. Louis, MO, USA), mixed with 0.5% methylcellulose, and administered per os through gavage at a concentration of 600 mg/kg daily for 2 or 4 weeks in total. To the control mice, 5% methylcellulose slurry was administered via the same route. At the end of the experiment, animals were terminally anaesthetised with 5% isoflurane and humanely sacrificed by rapid exsanguination followed by cervical dislocation. Liver samples were collected for RNA isolation and histological examination.
Liver histology and staining. Freshly isolated liver samples were cut into 3 × 3 mm pieces and fixed in 10% buffered formalin for 24 h. After an extensive wash in diH2O, samples were dehydrated in isopropanol solutions with rising concentrations from 70% to 100%, followed by two immersions in xylene and then embedded in Histomix (Biovitrum, Russia) at 56 • C. Embedded tissue samples were sectioned by microtome at 5 µm slices and mounted onto SuperFrost glass slides. We used commercially available kits for Pass-Shiff and Gomori-Grocott staining (Biovitrum, Russia). The images of histological specimens were taken with microscope Keyence BZ9000 BioRevo (Yokogawa Electric Corporation, Tokyo, Japan).
RNA isolation and library preparation. TotRNA was isolated from liver tissue samples (60-100 mg) using TRI Reagent (Sigma-Aldrich, Merck KGaA, Darmstadt, Germany) according to the manufacture protocol. 3-5 µg of totRNA was used for isolated mRNA by NEBNext Poly(A) mRNA Magnetic Isolation Module (New England BioLabs, MA, USA) according to the manufacture protocol. RNA concentration was measured by Nanodrop (Thermo Fisher Scientific, MA, USA), and RNA samples were analysed by capillary electrophoresis using an Agilent capillary electrophoresis system (Agilent, CA, USA). cDNA libraries were constructed using NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (New England BioLabs, MA, USA) following the manufacturer's protocol. cDNA libraries were sequenced using the NextSeq500 (Illumina, San Diego, CA USA) instrument. Hence, 33-41 million raw reads were obtained for each sample with a 75-bp read length.
RNASeq data analysis. Raw reads were preprocessed as follows: trimmed with Trimmomatic (v0.39) [124] to remove the adapters. The processed reads were aligned to the Mus musculus genome (assembly GRCm39.105) using the HISAT2 algorithm [128]. Follow analysis was performed in R version 3.4.2. Gene read counts were calculated using Rsubread package, function feature Counts (with parameters countMultiMappingReads = F, isPairedEnd = TRUE) [129]. Genes with less than 10 reads on average for each sample were filtered out. To check the data for self-consistency, the following analysis was carried out: correlation analysis and MDS using the Spearman correlation and PCA using normalised gene counts as CPM (count per million). DEGs were identified using edgeR package version 3.20.9 [130] as follows. First, the read counts were normalised using the calcNormFactors function (RLE algorithm). Genes differentially expressed between each group of samples and appropriate control were identified using the estimateDisp, glmFit, and glmLRT functions with a 0.05 FDR significance threshold [131]. Hierarchical clustering was carried out to identify the modules of genes using hclust and cutree functions. Gene ontology analysis of DEGs modules was performed using the ClusterProfiler package with all significant genes as background [132]. Pathway enrichment analyses were performed across Bioplanet 2019, GO Molecular function [133,134], MSigDB_Hellmark_2020, and KEGG_Pathways_2019 CHEA 2016 databases to annotate differences in molecular functions, biological processes, and signalling pathways between experimental groups with Enrichr (http://amp.pharm.mssm.edu/Enrichr; New York, NY, USA).

Conclusions
Our study presents multiple biological targets of 4MU, including enzymes, nuclear receptors, and transcription factors. Detailed investigation of unravelled 4MU targets presented here will be pivotal in the discovery of new drugs for the efficient treatment of pathological processes associated with energy metabolism, inflammation, carcinogenesis, and tissue repair. Our transcriptomic data support the public data indicating nuclear receptors as the main targets of 4MU action. Future work will allow us to validate these targets using in vivo gene knockdown techniques in experimental models of fibrosis as well as immunological and neurodegenerative disorders.  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.