Background & Summary

The liver has a unique regenerative capacity. After the surgical removal of 70% of the liver mass, the remaining tissue enlarges as hepatocytes proliferate to compensate for the lost tissue and function1,2. Hepatocyte proliferation is critical for liver regeneration because hepatocytes, which account for 80% of the liver mass, regain their weight by hypertrophy and proliferation after 70% partial hepatectomy (PH)3,4. In previous studies, intravenous glucose supplementation5, complementation of altered metabolic capacity6, and suppression of lipid accumulation7 inhibit PH-induced hepatocyte proliferation as a control mechanism for hepatocyte proliferation. Thus, these changes have been considered as necessary triggers for hepatocyte proliferation. However, these studies have investigated the specific phenomena and genes that contribute to the proliferation of hepatocytes during liver regeneration after 70% PH. There have been few comprehensive studies linking each mechanism; thus, the regulatory mechanism of hepatocyte proliferation remains poorly understood.

In recent years, large-scale gene expression analysis has become important in biology and tissue regeneration research. In particular, microarray analysis has long been used to examine gene expression with known nucleotide sequences. In addition, the platform match would allow direct comparative analysis with considerable microarray data accumulated in previous studies. In the 70% PH model, microarray analysis was used to compare normal livers with livers in the proliferative phase to extract genes that control hepatocyte proliferation8, as well as studies that examined gene expression over time in the same model9. However, the microarray data used in previous studies on the mechanism underlying liver regeneration lacked biological replicates, making it impossible to eliminate the effect of individual differences in regenerative capacity. In addition, these data were only associated with time information and were not linked to important liver weights in the 70% PH model.

In this study, a longitudinal microarray dataset of the liver was obtained during liver regeneration after 70% PH in mice. High-purity and high-quality RNA was collected from the liver after obtaining weight and liver weight information from three mice at each time point after 70% PH. High-quality microarray datasets were generated from the collected RNA, and the gene expression profiles were evaluated. The results indicated that the expression level of cell cycle–related genes significantly increased with compensatory hypertrophy of the liver. No genes with significantly altered expression were extracted after 14 days of resection, indicating that the data accurately reflected the state of liver regeneration.

Collectively, the microarray dataset containing sufficient biological replicates obtained in this study will be utilized to elucidate the mechanism underlying liver regeneration after 70% PH based on the body–liver ratio linked to each dataset to accurately determine the degree of regeneration.

Methods

70% PH

All animal experiments were performed in accordance with the Nihon University Animal Care and Use Committee and Nihon University Rules concerning Animal Care and Use. C57BL/6 N mice were purchased from the Charles River Laboratories. 70% PH was performed as previously described10. In brief, the mice (10 weeks old) were anesthetized with isoflurane (Wako Pure Chemical). After restraining the mouse, a 3-cm incision was created in the midline of the abdominal skin and muscle to expose the xiphoid process. The median and left lateral lobes were tied around the top of the lobes using a 4-0 silk surgical thread. Then, the lobes were removed surgically, thereby removing about 70% of the total liver weight. The abdomen of the mice was sutured closed. Afterward, the mice were weighed and sacrificed at 0 day before 70% PH and at 1, 2, 3, 4, 7, and 14 days after PH. The livers were harvested from the mice, and liver weights were measured. Three biological replicates were prepared for each sample category using different mice.

RNA collection, isolation, and quality check

The collected livers were homogenized in TRIzol solution and mixed with an equal volume (300 μL) of the supernatant treated with chloroform. RNA was extracted using RNeasy, and genomic DNA was removed. A NanoDrop ND-1000 spectrophotometer was used to quantify the isolated RNA (NanoDrop Technologies, Wilmington, DE). The intactness of RNA was determined by using the Agilent RNA 6000 Nano Kit on the Agilent 2100 Bioanalyzer system and evaluated by using RNA integrity number (RIN) software algorithm.

Microarray hybridization and data acquisition

Fragmented and labeled cRNA samples were prepared from 250 ng of total RNA using the GeneChip® 3′ IVT PLUS Reagent Kit. cRNA was hybridized to the Mouse Genome 430 2.0 Array at 45 °C for 16 h (60 rpm). After washing, the array was scanned using a GeneChip® Scanner 3000 7 G.

QC stats

Affymetrix raw CEL files were preprocessed using the robust multichip average algorithm11, which includes background correction, quantile normalization, median polish summarization, and log base 2 transformations. The quality of the array data was evaluated by using the Bioconductor R packages “affy” and “simpleaffy,” and QC stats was performed.

Hierarchical clustering analysis

In evaluating the similarity of the expression profiles among the samples, hierarchical clustering analysis12 was performed using R (version 4.0.2) with the expression matrix data as input. In measuring the similarity between each sample, Manhattan was used, and the ward D2 method was used to measure clustering.

Extraction of differentially expressed genes

Differentially Expressed Genes (DEGs) were extracted by focusing on genes with a twofold or more increase or decrease and a P-value of less than 0.05. Student’s t-test was used to test for significant differences. Volcano plots were visualized using TIBCO Spotfire Desktop version 7.9.0 (TIBCO Software).

Statistical analysis

The body–liver ratio data were presented as mean ± SEM. Comparisons between several groups were analyzed using Tukey’s test, and a P-value of 0.01 was considered statistically significant.

Data Records

The dataset of liver microarray data from the 70% PH model mouse was deposited in the Gene Expression Omnibus (GEO) database with the accession number GSE27861813 and is available for public access.

The data can be accessed through https://identifiers.org/geo/GSE278618.

The GEO accession numbers for individual samples are listed in Table 1.

Table 1 Gene Expression Omnibus (GEO) accession numbers and corresponding mouse body weights, liver weights, and body–liver ratios for individual samples of the 70% PH model mouse liver microarray dataset.
Full size table

Technical Validation

Liver regeneration after PH

The body weight and liver weight of the mice were measured to evaluate whether the liver regains weight after 70% PH because of compensatory hypertrophy of the remaining liver (Table 1). The ratio of the liver weight to the body weight (body–liver ratio) significantly increased from 2 to 4 days after resection, indicating that the liver had recovered to a level with no significant difference compared with that before resection by 4 days (Fig. 1).

Fig. 1
figure 1

Liver-to-body weight ratios during liver regeneration. Results are presented as means ± SEM; n = 3. a–b, P < 0.01, Tukey’s test.

Full size image

Assessment of RNA integrity

High-quality RNA for microarray analysis must be obtained to ensure the accuracy of gene expression studies. The purity of the total RNA of all samples was evaluated by measuring A260/280 using a NanoDrop ND-1000 spectrophotometer.

The A260/280 values of all samples ranged from 1.90 to 2.10, indicating high purity (Table 2). The quality of the total RNA of all samples was measured on an Agilent 2100 Bioanalyzer system using the Agilent RNA 6000 Nano Kit and evaluated using the RIN software algorithm. The images (Fig. 2a) and electropherograms (Fig. 2b) showed peak signals of 18 S and 28 S ribosomal bands in each sample, indicating that the RNA from all samples was of high quality and suitable for microarray analysis.

Table 2 RNA purity as determined by Nanodrop 2000.
Full size table
Fig. 2
figure 2

RNA integrity measurement. (a) Gel-like images and (b) electropherograms generated by the Agilent 2100 Bioanalyzer system to determine the quality of the 21 RNA samples. The 18 S and 28 S ribosomal bands are marked on the electropherograms. Fluorescence unit: FU; Nucleotide: nt.

Full size image

Evaluation of microarray data

The quality of the microarray data obtained from all the extracted RNA samples is important for microarray analysis. The QC stats evaluation showed no significant differences in mean background, scale factor, % present value, and 3′–5′ ratio among the microarray data (Fig. 3).

Fig. 3
figure 3

Quality check of the microarray data. Affymetrix quality control parameters for all microarrays: The numbers in the first column indicate present call (41.17%–46.86%), and the numbers below indicate background (30.62–38.82). The horizontal line indicates the scale factor, with all samples inside the 3 × dotted line. Circles and triangles indicate the 3′ 5′ ratio of GAPDH to actin.

Full size image

Expression profiles of the microarray data

Hierarchical clustering was performed to validate the microarray data (Fig. 4a). Hierarchical clustering showed that the livers at 1 to 3 days postresection were classified into different clusters compared with the livers at 0 day. In addition, the three biological replicates in all samples except for 2d-2 and 7d-1 formed one cluster or small clusters in close proximity. This phenomenon is due to individual differences related to the rate of regeneration, as the samples were classified into clusters of consecutive samples. In evaluating whether the gene expression profile returned to normal after PH, the number of differentially expressed genes was extracted by comparing the gene expression profile of each time point with that of the normal liver (Fig. 4b). The gene expression profile returned to normal 14 days after resection, as the genes that were significantly (P < 0.05) altered by more than twofold increase or decrease compared with the normal liver were not extracted after 14 days of resection. The expression level of cell cycle–related genes, such as Mcm2, Mcm3, Mcm4, Mcm6, Mcm7, Cdk1, Ccnb1, Ccnb2, and Mki67, increased significantly from 1 to 4 days after resection, which is consistent with the time of liver hypertrophy, indicating that the microarray data in this study were reliable. This result indicates the reliability of the microarray data in this study.

Fig. 4
figure 4

Gene expression profiles of the microarray data. (a) Dendrogram of signals normalized by robust multichip average for all 21 samples created using hierarchical clustering. The numbers above the dendrogram indicate the approximately unbiased (au) P-values and bootstrap probability (bp). All samples in the dendrogram are colored the same for each day. (b) The volcano plot of genes between 0 day and each cluster. DEGs were extracted for genes with more than twofold change in expression values and P-values less than 0.05 compared with 0 day. Fold change is shown in the x-axis. The P-values are shown in the y-axis. Dots indicate each gene, and numbers indicate the DEG number (red, upregulation; blue, downregulation). Labeling is a gene that promotes the cell cycle.

Full size image