Diverse MicroRNAs‐mRNA networks regulate the priming phase of mouse liver regeneration and of direct hyperplasia

Abstract Objectives Adult hepatocytes are quiescent cells that can be induced to proliferate in response to a reduction in liver mass (liver regeneration) or by agents endowed with mitogenic potency (primary hyperplasia). The latter condition is characterized by a more rapid entry of hepatocytes into the cell cycle, but the mechanisms responsible for the accelerated entry into the S phase are unknown. Materials and methods Next generation sequencing and Illumina microarray were used to profile microRNA and mRNA expression in CD‐1 mice livers 1, 3 and 6 h after 2/3 partial hepatectomy (PH) or a single dose of TCPOBOP, a ligand of the constitutive androstane receptor (CAR). Ingenuity pathway and DAVID analyses were performed to identify deregulated pathways. MultiMiR analysis was used to construct microRNA‐mRNA networks. Results Following PH or TCPOBOP we identified 810 and 527 genes, and 102 and 10 miRNAs, respectively, differentially expressed. Only 20 genes and 8 microRNAs were shared by the two conditions. Many miRNAs targeting negative regulators of cell cycle were downregulated early after PH, concomitantly with increased expression of their target genes. On the contrary, negative regulators were not modified after TCPOBOP, but Ccnd1 targeting miRNAs, such as miR‐106b‐5p, were downregulated. Conclusions While miRNAs targeting negative regulators of the cell cycle are downregulated after PH, TCPOBOP caused downregulation of miRNAs targeting genes required for cell cycle entry. The enhanced Ccnd1 expression may explain the more rapid entry into the S phase of mouse hepatocytes following TCPOBOP.

(PH). [4][5][6] Indeed, while miRs post-transcriptionally regulate genes that orchestrate proliferation in development and cancer, their role in the proliferation of fully differentiated hepatocytes is still largely unknown. In this context, the finding that hepatocyte-specific Dicer knockout transgenic mice developed normally, but exhibited enlarged livers compared to controls, strongly support the role for miR-NAs in the control of hepatocyte proliferation. 7 This study, together with the discovery that the association of miRNAs with different polysome fractions was altered during liver regeneration, 8 raised the intriguing possibility that miRNAs might regulate the regeneration of this organ.
As shown by Shu et al., 9 upregulation of a cluster of miRNAs takes place between 0 and 4 h after PH, a time corresponding to the so-called priming of hepatocytes, 10,11 characterized by refractoriness to DNA synthesis; conversely, downregulation of the vast majority of miRNAs associates with the transition G1-S of the cell cycle and the recovery of liver mass. Accordingly, the expression of most cell cycle-related genes is repressed for several hours after surgery.

Similar findings were reported by Yin et al., who identified in rat liver
transcription factors inhibiting the cell cycle, as early as 2 h after PH in rat liver. 12 After the priming phase, hepatocyte DNA synthesis peaks at 24 or 36 h in rats and mice, respectively. 11 Hepatocyte proliferation can be induced not only after cell death/loss but also following treatment with several xenobiotics or endogenous molecules (direct/primary mitogens), able to induce the entry of hepatocytes into the cell cycle in the absence of previous liver cell damage. 13 Among the broad spectrum of chemical mitogens, it is remarkable that many of them are ligands of nuclear receptors of the steroid/thyroid superfamily, including 1,4-bis [2-(3,5 -dichloropyridyloxy)]benzene TCPOBOP (abbreviated thereafter as TCP). Studies with knockout mice have shown that the initial signalling elicited by liver regeneration and direct mitogens is different. [1][2][3]14,15 Moreover, no change in the activation of transcriptional factors implicated in rat liver regeneration (such as, NF-kB, AP-1, STAT3) has been observed in nuclear receptor-mediated hepatocyte proliferation. 13 In this context, it should be also mentioned that the entry of hepatocytes into the S phase of the cell cycle in mice is robustly anticipated in mitogen-treated animals (18 h instead of the 30-36 h required after 2/3 PH, with mitotic figures being evident at 24 vs. 48 h). 16 Although existing studies analysed early responses of the mouse liver transcriptome at early times after treatment with TCP-an agonist of the constitutive androstane receptor (CAR)- 17,18 or soon after PH, 12 the involvement of miRNAs as critical regulators in the priming phase of hepatocytes in these two proliferative conditions, has not been studied so far.
In an attempt to investigate whether deregulation of miRNA expression could play an essential role in the priming phase of hepatocytes and whether differences might exist between the two proliferative conditions (compensatory regeneration and direct hyperplasia), we performed a transcriptomic and miRNomic analysis on the liver of mice sacrificed 1, 3 and 6 h after 2/3 PH (liver regeneration) or after a single dose of TCP (direct hyperplasia).

| Animals and treatments
Guidelines for the Care and Use of Laboratory Animals were followed during the investigation. All animal procedures were approved by the Ethical Commission of the University of Cagliari and the Italian Ministry of Health. Three-month-old CD-1 female mice (30 g) were fed a laboratory chow diet provided by Ditta Mucedola (Settimo Milanese, Italy) with free access to food and water. All experiments were performed in a temperature-controlled room with alternating 12-h dark-light cycles. TCP (Sigma-Aldrich), was dissolved in dimethyl sulphoxide/corn oil. A single dose of 3 mg/kg body weight was administered by gavage. PH was performed by removal of 70% of the liver mass as originally described by Higgins and Anderson. 19 In the first set of experiments, mice were sacrificed 24, 36 and 48 h after PH or TCP. Bromodeoxyuridine (BrdU) (100 mg/kg) was injected intraperitoneally 2 h before sacrifice. To investigate the role of miRs in the priming of hepatocytes, mice were sacrificed 1, 3 and 6 h after each treatment. Three mice were used per group at each time point. Liver segments were fixed in formalin for histology or snap-frozen in liquid nitrogen and kept at -80°C until use.

| RNA and miR isolation
Total RNA was isolated with the RNeasy Plus Mini isolation kit (Qiagen) from 3 livers of untreated and treated mice, subjected to PH or TCP treatment. RNA was quantified by Nanodrop spectrophotometer (Thermo Scientific) and its integrity was evaluated by Agilent Bioanalyzer 2100. Only RNA samples with a RIN (RNA Integrity Number) ≥7 were included in the study. using Bowtie. 22 The R/Bioconductor package "DESeq2" was used to identify differentially expressed genes and miRs. Data were filtered according to read count value (threshold ≥6 reads). Only miRs having an adjusted P-value of ≤0.05) and fold change value of 1.3 were considered for further analysis.

| Microarray
For time-course expression profiling, total liver RNA was extracted and purified from the liver of three animals before (t=0) or 1, 3 or Only genes whose expression differed by at least 1.5-fold from the median in at least 20% of the arrays and characterized by the 50th percentile of intensities >300 were retained. The false discovery rate-adjusted P-values were calculated using the Benjamini-Hochberg procedure. To identify the differentially expressed genes, F-test and Multivariate Permutation Test were applied. Further, the genes were filtered based on their fold change values (±1.5).

| miR Target gene Prediction
The R package multiMiR 23 was used to predict and validate miR-mRNA target interactions. List of DE genes and miRs passing the cut-off value was used as input. Among the databases in the multiMiR package, the validated databases (miRecords, miRTarBase and TarBase) and top 10% results of the predicted database (DIANA-microT ElMMo, Microcosm, miRanda, miRDB, Pictar, PITA, TargetScan) were used for analysis.

| qRT-PCR analysis
The same cDNA used for gene sequencing was used also for qRT-PCR analysis. Total RNA was retro-transcribed using the High Capacity

| Histology and Immunohistochemistry
Liver sections were fixed in 10% of buffered formalin and processed for immunohistochemistry (IHC). For BrdU detection, field (at x40 magnification). Ten to 50 fields per liver were scored. A segment of the duodenum, an organ with a high rate of cell proliferation, was used as a positive control for BrdU incorporation.

| Statistical analysis
All data were expressed as the mean ± SD. Differences between groups were compared using one-way analysis of variance ANOVA with the use of GraphPad Software Inc., San Diego, CA. A value of p < 0.05 was considered as a significant difference between groups. Seventy-two hours after transfection, total RNA was extracted with Maxwell ® RSC miRNA Tissue Kit (Promega, Madison, WI, USA) according to manufacturer protocol. Total RNA was retro-transcribed starting from 0.25μg RNA/sample using the High Capacity Kit (Thermo Fisher). Gene expression analysis was performed using the specific TaqMan probes (Thermo Fisher) hCCND1 (hs00765553_ m1), and hACTIN (hs99999903_m1) as endogenous control. MiRNA expression was evaluated using the specific Taqman miRNA assay kits (Thermo): hsa-miR-106b-5p #000442 and RNU48 #01006 (as endogenous control). PCR runs were performed with ABI Prism 7900HT (Applied Biosystems).

| Hepatocyte proliferation following PH or TCP treatment.
According to previous reports, 16 the measurement of labelling index of hepatocytes from mice subjected to TCP and PH showed that while a high number of hepatocytes was in an active S phase as early as 24 h after TCP treatment, almost no BrdU-positive cells could be observed at the same time point after PH ( Figure 1A, B). However, the peak of DNA synthesis was observed in both groups 36 h after treatment ( Figure 1A, B) with a trend towards a return to quiescence at 48 h. Since these data suggest that different molecular events are responsible for the accelerated entry of hepatocytes into S phase observed after TCP treatment, we investigated the possible involvement of miRs in the priming phase of liver cells. To this aim, we analysed the expression profiles of mouse hepatic mRNA and miR at 1, 3 and 6 h after PH or TCP. Identification of miR and mRNA expression abundance was evaluated by NGS (miR) and Illumina microarray (mRNA) in the same samples.

| Global gene expression profiles in regenerating livers after PH and TCP.
Global transcriptome changes at 1, 3 and 6 h revealed a total of 810 and 527 genes differentially expressed (DE) in the PH and TCP groups, respectively (Table S1 and S2). Hierarchical clustering analysis of the PH differentially expressed genes stratified them into two major clusters: (1) control liver and PH 1 h and (2) PH 3 and 6 h ( Figure 1C). Similarly, it also stratified differentially expressed genes after TCP into two major clusters: (1) control liver and TCP 1 h and (2) TCP 3 and 6 h ( Figure 1D). Figure 2A  We also performed DAVID functional analysis using Gene

As shown in
Ontology annotation. As shown in Figure 3A, among the genes upregulated after PH, many were related to apoptosis, cell cycle regulation and cell cycle arrest. Interestingly, genes classified as negative regulators of cell proliferation (Sox9, Cdkn1a, Jun, Trp53inp1, Agt, Gja1, Tgif1) or of ERK1/ERK2 cascade (Ats3, Dusp1, Dusp6, Timp3, Ptpn1) were upregulated following surgery ( Figure S2). On the other hand, genes related to cellular metabolism were among the most deregulated after TCP ( Figure 3B). Interestingly, only genes positively related to the cell cycle (Gadd45b, Sgk1, and Ccnd1) were observed following xenobiotic treatment ( Figure S2). No evidence of increased expression of negative regulators of cell proliferation, cell cycle arrest or of ERK1/ERK2 cascade was observed after TCP ( Figure 3B).
The most downregulated pathways in both experimental conditions are listed in Figure S3. While most of them involved metabolic pathways, none was directly related to cell cycle/cell proliferation.

| Transcription Factors-Dependent Pathways
Next, we analysed transcription factor (TF)-dependent pathways differentially activated in the livers of PH and TCP mice. By examining the top 20 TFs in each group, we found striking differences between the two proliferative stimuli. Indeed, while RB1 was the most significantly downregulated TF in the PH livers at all the analysed time points it was not listed among the first 20 TFs after TCP ( Figure 4A, B). In addition, while STAT3 was among the most significantly upregulated TF after PH ( Figure 4A), it was profoundly downregulated 1 h after TCP treatment ( Figure 4B). Furthermore, C/EBPβ, a TF that initiates a cascade of gene expression responsible for proliferation, 24 was strongly upregulated by TCP at all the examined time points, whereas it was not modified within the first 6 h after surgery. As shown in Figure 4C showing that no change in the activation of transcription factors implicated in liver regeneration such as, NF-kB, AP-1 and STAT3 has been observed in nuclear receptor-mediated hepatocyte proliferation. 13 As to FOXO1-strongly upregulated after PH and unchanged or only slightly upregulated post-TCP-it is interesting to note that forced expression or conditional activation of FOXO factors led to reduced Cyclin D1 expression. 25,26

| miR expression in PH or TCP-treated mice
To investigate miRs differentially expressed in the liver following the two proliferative stimuli, we applied time-course analysis using the R/Bioconductor package "DESeq2". Hierarchical clustering analysis in the PH group stratified the three time points into 2 major clusters:

1) Controls (CO), PH 1 and 3 h and 2) PH 6 h. Hierarchical clustering
analysis on TCP samples did not display a clear separation of the different time points (Figure 5A, B).
A striking difference in the number of differentially expressed miRs was found between the two proliferative conditions ( Figure S4 and S5). Indeed, while the expression of 102 miRs was significantly modified after PH, only 10 miRs were found differentially expressed in the TCP group. As to PH, the Venn diagram showed that the highest number of miRs was deregulated at 1 h (75 miRs) and 6 h (68 miR-NAs) post-surgery, whereas 23 miRs resulted commonly altered at all the time points ( Figure 5C, D). Notably, the three most upregulated miRs 1 h after PH, miR-124-3p, miR-9-3p and miR-9-5p ( Figure S4) act as inhibitors of proliferation in several cell types. 27,28 Similar to what was observed with mRNA expression, a much lower number of deregulated miRs was found after TCP treatment at all the analysed time points, with only 9 miRs being commonly altered at all the time points. Differently from PH, TCP caused downregulation of all the miRs at each time point, with the only exception of miR-382-5p that was upregulated 6 h after administration of the drug ( Figure 5E, Figure S5). Interestingly, out of the 9 miRs deregulated 1 h after TCP, only 3 were altered in PH livers ( Figure 5E).

| miR-mRNA interactions
To identify a possible link between differentially expressed miRs and genes, among the identified dysregulated genes, we selected multiMiR validated/predicted miR targets. In both experimental  (Table S3).
After PH, we also found upregulated both genes considered to be positively correlated to induction of proliferation (ie c-fos and Ccnl1), as well as negative regulators of cell cycle (Gadd45a and Cdkn1a). In particular, Gadd45a was upregulated at all the time points and its upregulation was paralleled by downregulation of miR-301b-3p, miR-484 and miR-19b-3p (validated) and miR-130a-3p, miR-130b-3p and miR-301-3p (predicted) to target it, as early as 1-h post-surgery. Moreover, downregulation of miR-301 was associated with the upregulation of its target gene Cdkn1a ( Figure 6A and Table S3).
Nfkbiz-another negative regulator of the cell cycle 12,29 was upregulated 3 and 6 h after PH. Interestingly, in rat liver Nfkbiz is a target of miR-376b. 30 In our study, however, the role of miR-376b is unclear as it was upregulated at 1 and 3 h after PH and downregulated at 6 h (Table S3). Whether mouse Nfkbiz is a target of miRs other than miR-306b will require further studies.
In the TCP experimental group, no cell cycle negative regulatorother than Gadd45a-was significantly modified compared to control liver. Remarkably, Ccnd1, the gene encoding for cyclin D1, responsible for the G0-G1 transition, was upregulated 6 h after TCP treatment. Such upregulation was paralleled by downregulation of cyclin D1 targeting miRs (miR-20a-5p, miR-20b-5p and miR-17-5p) Colour is determined by Z-score; the Z-score >2 and <−2 is considered significant. Blue colour indicates suppressed disease /biological function or canonical pathways; orange indicates activated disease/biological function or canonical pathways; (D) Heatmap for top 20 canonical pathways and diseases and biological functions at 1, 3 and 6 h post-TCP. Colour is determined by Z-score; the Z-score >2 and <−2 are considered significant. Blue colour indicates suppressed disease /biological function or canonical pathways; orange indicates activated disease/biological function or canonical pathways as early as 1 h after TCP ( Figure 6B). Notably, after PH, downregulation of the same miRs was observed only at 6 h ( Figure S4). Only 2 miRs resulted exclusively deregulated in the liver of TCP-treated mice: miR-106b-5p, predicted to target Ccnd1, which was downregulated at all the time points, and miR-32-5p targeting other positive regulators of the cell cycle, namely Sgk1 and Pik3cb 31,32 ( Figure 6B; Table S3). To validate the NGS results, we performed qRT-PCR analysis on a selected set of genes and miRNAs. As shown in Figure   S6A, the expression of all the investigated genes (Ccnd1, Cdkn1a, Cype2b10, Socs3, Gadd45b and Gadd45a) was deregulated similar to what observed by NGS.
QRT-PCR analysis was also performed to validate the changes in the levels of miR-106b-5p predicted to target Ccnd1 and whose expression was downregulated after TCP ( Figure 5E) and of miR-301b and miR-455-predicted to target Cdkn1a and Socs3, respectively and found upregulated after PH ( Figure S4). As shown in Figure   S6B, the expression of these miRs was deregulated in accord to NGS results.
Next, to further validate the effect of miR-106b-5p on Ccnd1 we transduced two human liver carcinoma cell lines (Mahlavu and HepG2) with this miRNA and measured the mRNA levels of its target gene. As shown in Figure S7, expression of miR-106b-5p led to a significant downregulation of Ccnd1.

| DISCUSS ION
In the present study, we unveiled the miR-mRNA networks involved Previous studies demonstrated that, in mice, the entry of hepatocytes into the S phase of the cell cycle is robustly anticipated in TCP-treated animals compared to PH. 16 Indeed, an active DNA synthesis takes place 24 h after TCP, a time when virtually no dividing hepatocytes can be seen in livers undergone PH. In this context, it is interesting to note that Cyclin D1 induction occurs early after treatment with TCP, but not after PH. 16 For this reason, the identification of Ccnd1 targeting miRs already downregulated 2 h after TCP may explain the increase of cyclin D1 and the consequent faster entry into the cell cycle. Importantly, miR-106b-5p, Cdkn1a. Furthermore, we observed a strong upregulation of the oncosuppressor miR-34 family. Of note, the increased expression of miR-34 observed during the termination phase of liver regeneration has been reported as a potential 'stop' signal. 39 This finding together with the report of Sun 40 showing that miR-34a targets the 3' untranslated mRNA region of Ccnd1, supports the concept that an active control on mitogenic signals operates after PH, thus leading to a delay in the entry into the cell cycle. Conversely, no evidence of miR-34 deregulation was observed at any time point in TCP-treated livers ( Figure S4).
Another interesting observation that could justify the different kinetics of S phase entry of the two proliferative stimuli is the finding that while after PH pathways related to inflammatory response were among the most deregulated, pathways involving upregulation of metabolic changes, especially lipid metabolism-required for sustaining cell proliferation-were the most modified after TCP.
In conclusion, the analysis of the microRNAs-mRNA networks performed in the present study unveils on the one hand that different miRs are implicated in the early phase of hepatocyte proliferation induced by mitogenic stimuli of different nature, and on the other hand that miRs, such as miR-106b-5p are critical in regulating the levels of the main cyclin implicated in the G1-S transition of the cell cycle.
A limitation of this study relies on the impossibility to functionally validate some of the present findings since primary hepatocytes in vitro do not express CAR and do not proliferate after TCP. 41 Nevertheless, the present work, performed in a strictly controlled experimental condition, contains a number of novel findings that can be helpful for a better comprehension of the molecular mechanisms responsible for deciphering proliferative signals in totally diverse conditions, such as those where hepatocyte replication is needed F I G U R E 6 mRNA-miR interaction networks of cell cycle-related genes in mice subjected to PH or treated with TCP. (A) mRNA-miR networks of Socs3, Gadd45a and Cdkn1a in mice subjected to PH showing that downregulation of several miRNAs is associated with the upregulation of their target genes; (B) mRNA-miR networks of Ccnd1-miR-106b-5p and Sgk1-Pik3cb-miR-32-5p in TCP-treated mice. The panel shows miRs predicted/validated to target Ccnd1 and downregulated after TCP (left side), and all the predicted target genes of miR-106b-5p and miR-32-5p -the only two miRs differentially expressed only in TCP mouse liver (middle and right side) to replace cell loss (ie PH or chemically induced necrosis) or those where proliferation occurs in the intact liver (TCP as well as other hepatomitogens, such as T3 and PPAR ligands). Columbano, Andrea Perra and Silvia Giordano contributed to conceptual framework, supervised the study and revised the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request. Raw data can be Andrea Perra https://orcid.org/0000-0002-8098-899X