†These authors contributed equally.
Background and objective: Liver regeneration (LR) is a complex process
influenced by various genes and pathways, the majority of the of research on LR
focus on the initiation and proliferation phase while studies on termination
phase is lacking. We aimed to identify potential genes and reveal the underlying
the molecular mechanisms involved in the precise regulation of liver size during
the termination phase of LR.
Materials and methods: We obtained the rat liver tissue gene datasets
(GSE63742) collected following partial hepatectomy (PH) from the Gene Expression
Omnibus (GEO) of the National Center for Biotechnology Information (NCBI), from
which, this study screened the late stage LR samples (7 days post-PH) using the
R/Bioconductor packages for the identification of differentially expressed genes
(DEGs). Afterwards, we performed enrichment analysis using the database for
annotation visualization and integrated discovery (DAVID) online tool. Moreover,
the Search Tool for the Retrieval of Interacting proteins (STRING) database was
employed to construct protein-protein interaction (PPI) networks based on those
identified DEGs; the PPI network was then used by Cytoscape software to predict
hub genes and nodes. Animal experimentation (Rat PH model) was performed to
acquire liver tissues which were then used for western blot analysis to verify
our results.
Results: The present study identified together 74 significant DEGs,
among which, 51 showed up-regulation while 23 presented as down-regulated. As
revealed by KEGG pathway enrichment analysis, DEGs were mostly related to
pathways such as retinol metabolism, steroid hormone synthesis, transforming
growth factor-
When injury occurs, a series of signaling pathways in the liver are activated to
initiate liver regeneration (LR), and the body’s metabolic needs are satisfied by
the rapid expansion of the remaining liver tissue. In 1931, Higgins et
al. [1] first demonstrated its unique characteristics by performing animal
experiments. After approximately 70% partial hepatectomy (PH), the remaining
liver in rats could basically recover to the weight, volume and function before
operation within 5–7 days, and then the cells returned to static state. It is
considered that the whole process of liver regeneration includes three phases
including initiation, proliferation and termination [2]. At the initiation phase,
a series of cytokines including TNF-
The gene expression data set (GSE63742) of rat liver tissue after PH operation was selected and downloaded by gene expression omnibus (GEO), a high-throughput gene expression database of the National Center for Biotechnology Information (NCBI), with “liver regeneration” as the key word. The data set contained altogether 57 samples at 9 time points after PH operation. In the present study, 6 samples (3 in sham-operation group and 3 in PH late phase group) were selected as the original data of gene expression at the termination phase of liver regeneration in rats.
The “affyPLM” package of R language (https://www.r-project.org/) was employed to verify the quality of regression calculation of original data. Meanwhile, the results were evaluated and displayed using Relative log expression (RLE) box plot, normalized unscaled standard errors (NUSE) box plot and RNA degradation plot [5].
The “affy” and “limma” [6, 7] packages of R language were used to process
the original chip data as follows: ① Background correction and quantile
homogenization; ② Screening differently expressed gene (DEGs) under the
following conditions: Absolute value of Log2 (fold change, FC)
On-line tool DAVID database (https://david.ncifcrf.gov/home.jsp) [8] was adopted to analyze the enrichment of DEGs by Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene ontology (GO) [9, 10]. GO enrichment analysis included Biological process (BP), Cellular component (CC) and Molecular function (MF). In addition, the analysis results were visualized by “ggplot2” package of R language.
The DEGs obtained in the above steps was input into the STRING database (https://string-db.org/). At the same time, the protein-protein interaction network (PPI) was constructed on the condition of medium confidence of 0.400. PPI network information was downloaded and imported into the Cytoscape software. The key nodes (hub nodes) and key genes (hub genes) in the network were analyzed and predicted by using the cytohubba plug-in [11, 12].
A total of 6 Rats (male, 8-week-old, 260–280 g, SD) acquired from Experimental
Animal Center of Chongqing Medical university (Chongqing, China) and were
maintained in specific pathogen free (SPF) barrier (sterile polycarbonate cages,
sustained at 24
PH Group underwent the procedure described by Higgins et al. [1]. Firstly, rats were given intraperitoneal injection of sodium pentobarbital (50 mg/kg) for anesthesia, then the left and middle lobes of the liver were ligated and resected (approx. 70% of rat liver weight). The same protocol was performed on the sham groups but without undergoing PH. At 7 days postoperatively (late phase of liver regeneration), rats were anesthetized and sacrificed, then the liver tissues were collected for Western blotting analysis.
The acquired liver tissues were lysed (RIPA lysis buffer and PMSF (Beyotime,
Shanghai)), which were quantified using BCA protein quantitative kit (Beyotime,
Shanghai). Protein samples was separated by electrophoresis of SDS-PAGE and then
transferred to polyvinylidene difluoride (PVDF) membranes. The membranes were
first blocked in 5% skim milk (37
As shown in Fig. 1, the RLE values in each group were relatively uniform and close to 0, while the NUSE values in each group were relatively uniform and close to 1. Typically, a greater deviation from the corresponding center implies a larger variability of supposed sample, thus meaning that the sample was invalid for further analysis. RNA degradation plot indicated that the degradation of 5’ end was more significant than that of 3’ end in each group, and the trend among each group was relatively parallel. All the above results demonstrated that the chip had excellent quality and stable detection results, which provided reliable data for subsequent analysis.

Quality verification of Microarray. (A) In the RLE box plot, the boxes centered at 0 low quality arrays had
great variability with large deviation from 0. (B) In the NEUS box plot,
median standard errors from the PLM scaled to 1, the boxes centered at 1, median
standard error
Principal component analysis (PCA) was performed on the chip samples, and the results showed a significant difference between the two groups, with different characteristics (Fig. 2A). Cluster analysis of DEGs showed that the differences within the group were small and the repeatability was good. The gene expression pattern in PH group was significantly different from that in SHAM group. After normalizing the chip data, 51 genes were up-regulated and 23 genes were down-regulated, with a total of 74 DEGs. The results were displayed by volcano map (Fig. 2B). Gene expression across the samples can be visualized through the heatmap (Fig. 2C), green denotes up-regulation and red signifies down-regulation. Full details of the 74 expressed genes along with their fold change and ID given can be observed in the Supplementary Table 1 along with p-values, average expression and gene title (Supplementary Table 1).

PCA and DEGs screening at the termination phase of liver regeneration and in sham operation control group. (A) PCA on sham-operation group and PH group. (B) Volcano plot. Black represents non-DEGs, red stands for up-regulated DEGs, and green represents down-regulated DEGs. (C) Hierarchical clustering heatmap of the 74 up- and down-regulated DEGs screened in the GSE63742 dataset. Each row represents one gene, the right column represents for PH samples, the left column represents SHAM samples.
The results of KEGG visualization analysis of DEGs (Fig. 3) showed that a large
number of DEGs in LR termination phase were closely related to seven signaling
and metabolic pathways, including steroid hormone synthesis, retinol metabolism,
mitogen activated protein kinase signaling pathway and transforming growth
factor-

KEGG and GO enrichment analyses on DEGs. (A) Distribution of DEGs associated with the termination phase of LR in different KEGG pathways. (B) Distribution of DEGs associated with the termination phase of LR in biological processes (BP) of GO category. (C) Distribution of DEGs associated with the termination phase of LR in the cell component (CC) of GO category. (D) Distribution of DEGs associated with the termination phase of LR in molecular function (MF) of GO category.
PPI network constructed by STRING database and analysis by Cytoscape software could further clarify the protein interactions involved in DEGs and the key genes involved in regulation. The results showed that (Fig. 4) the PPI network contained 46 relationships of 67 node proteins, among which most proteins had interaction relationships. The cytohubba plug-in was used to screen key nodes and genes in the PPI network. The top 10 key genes included Cyp2c13, Ste2, Mup5, LOC259244, Rup2, Hsd3b5, UST4r, Btg2, Cyp2c11 and Myc.

Construction of protein-protein interaction (PPI) network and hub gene analysis on DEGs associated with the termination phase of liver regeneration. (A) Construction of PPI network using the STRING tool (medium confidence of 0.400. PPI). (B) The top 10 hub genes predicted using the Cytoscape software by adopting the cytohubba plug-in.
A total of 74 genes were identified as differentially expressed through
microarray analysis. Two genes, cytochrome P450, subfamily 2, polypeptide 11
(Cyp2c11) , alpha-1-B glycoprotein (A1bg) were chosen, due to the fact that they
were the most up and down-regulated genes respectively in our analysis. In the
western blot analysis, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used
as loading control. Western blotting showed that Cyp2c11 was upregulated and A1bg
was downregulated in the PH group compared with the SHAM groups, (*p

Western blotting Verification results. (A) Representative Western blotting analysis on the expression of A1GB,
Cyp2c11 and GAPDH (loading control) in the liver of male rats at 6 days after PH.
(B) Relative staining intensity ratios of A1GB and Cyp2c11 compared
between PH group and SHAM group. Data represent mean
Before extracting data, it is of necessity to verify the chip quality at first and then find and reject the problematic chips to ensure the reliability of subsequent analysis. RLE and NUES can reflect the consistency of parallel experiments. An RLE plot exhibits the difference between a gene’s expression level and the median for that gene, assuming that most genes are not differentially expressed. The boxes should be close to 0. A NUSE plot however is scaled so that the median standard error across arrays is 1 for each gene and thus boxes should be close to 1 [13]. If there is a large deviation in the box plot of a chip, it indicates that there is a problem with the chip [14]. RNA degradation is another important factor which can affect chip quality. Since RNA degradation starts from the 5’ end, the fluorescence intensity at the 5’ end is much lower than that at the 3’ end [15]. To assess the viability of the microarray using the RNA degradation plot, we look at the parallelism and similarities between samples slopes in which an extreme difference in one or more of the samples slopes means that the sample is not viable for use. The detection results of the above three indicators show that the chips included in this analysis are reliable in quality, and the batch difference of detection results is small, thus providing credible original data for performing subsequent analysis.
In the proliferation phase, both the regenerated liver and liver tumors show the
strong proliferation ability of liver cells. However, because tumor cells lack
the negative regulation mechanism for termination, their cells can divide
indefinitely. Liver regeneration is a complex and delicate process regulated by
multiple factors. When the regeneration level can satisfy the body’s metabolic
needs, it will trigger a termination signal to stop cell proliferation. In the
present study, altogether 74 DEGs were selected from the rat liver microarray at
the termination phase of liver regeneration, further validated by Western
blotting assay, and analyzed by KEGG and GO enrichment analyses, so as to reveal
the roles of these genes at the termination phase of liver regeneration. Among
the seven pathways screened by KEGG analysis, MAPK and TGF-
Fortier et al. [18] confirmed through the CCl4-induced acute liver
injury model that specific knockout of p38
In comparison with the control group, the top 10 hub genes related to the
termination phase of liver regeneration screened from PPI network were all
up-regulated. Ste2 gene regulates the expression of estrogen sulfotransferase 2,
maintaining the steady state of estrogen by sulfonating and inactivating
estrogen. It has been reported that this enzyme is the direct transcription
target of Nrf2. The latest research shows that Nrf2 is activated in the early
phase of tumorigenesis in rat model of fatty hepatitis with fibrosis, and
knockout of Nrf2 gene can inhibit the amplification of initial cell clone, which
can thus prevent the occurrence of HCC [27, 28]. This may be the reason for the
increased expression of the gene in the late phase of liver regeneration. A study
conducted by Zhao et al. [29] showed that tumor-like growth in the early
phase of liver development is positively and negatively coordinated by
To conclude, liver regeneration is a complex biological process regulated by multiple factors. The main feature of liver regeneration is that it can stop proliferation and produce functional hepatocytes in accordance with the needs of the body at the end phase. In this study, we successfully screened and analyzed the DEGs in rat liver tissue at the end phase of liver regeneration, and preliminarily understood its related functions and signal pathways. In the meanwhile, we also analyzed the hub genes with negative regulatory effect, such as Ste2, Btg2, and Hsd3b5, thus providing a new idea for the screening of liver tumor research targets and the exploration of pathological mechanism.
DEGs, Differentially expressed genes; GEO, Gene Expression Omnibus; HCC,
Hepatocellular carcinoma; HNF4
JG designed and supervised the study. HS. and MW. contributed to the implementation of the research, analysis of the results and writing of the manuscript. All authors have read and approved the manuscript.
All animals’ care and experimental protocols were complied with the Animal Management Rules of the Ministry of Health of the People’s Republic of China and all experimental designs were approved by the Animal Care and Use Committee of Second affiliated Hospital of Chongqing Medical university.
We thank the Experimental Animal Center of Chongqing Medical university (Chongqing, China) for providing us with animals required and we would also like to thank all the peer reviewers for their opinions and suggestions.
This study was supported by grants from nature science foundation of Chongqing (no: cstc2019jcyj-zdxmX0027).
The authors declare no conflict of interest.