The diagnostic and prognostic role of RhoA in hepatocellular carcinoma

mRNA and protein expression profile of RhoA in the HPA

By examining the RhoA expression profile in the HPA, we found that the RNA expression of RhoA in the normal liver tissue samples was the lowest compared with that in the other 30 human tissue samples (Figure 1A), and the protein level of RhoA was nearly undetectable in hepatic tissues (Figure 1B). As Figure 1C shows, the level of RhoA in the HCC tissue samples was relatively low among the 17 carcinoma types; however, the expression of RhoA was found to be significantly upregulated in HCC tissues compared with that in normal tissues (Figure 1D). Interestingly, regarding RhoA protein expression, liver cancer exhibited a high positive rate, which ranked second among the positive expression rates of 20 common cancer types (Figure 1E). Figure 1F displays representative IHC images of the distinct RhoA expression levels, including undetected, low, and medium. In addition, Hep G2, which is a frequently used hepatic carcinoma cell line, had a similar RhoA expression pattern as that of the liver cancer tissue shown in Figure 1C, as identified by the Cancer Cell Line Encyclopedia (CCLE) web searching tool. Hence, the RhoA protein level, rather than the gene expression level, may be a more sensitive biomarker to diagnose HCC.

Figure 1. Gene and protein expression profiles of RhoA in tissue samples and cancer cell lines.RhoA mRNA expression (A) data from the GTEx dataset in the Human Protein Atlas) and protein expression (B) data from the Human Protein Atlas) in normal human tissues. RhoA gene expression in common human tumor tissues (C) data from the TCGA dataset in the Human Protein Atlas). Comparison of RhoA mRNA expression between hepatocellular carcinoma tissues and normal liver tissues, including normal TCGA and GTEx data (D) data from GEPIA). RhoA protein expression overview in human tumor tissues (E) data from the TCGA CAB005052 dataset in the Human Protein Atlas) and representative immunohistochemistry (IHC) images (F) pictures from the Human Protein Atlas) with RhoA antibody (1:25, Cat#1600-1, Abcam, Cambridge, UK). RhoA mRNA expression in human cancer cell lines (G, data from RNA cell line category in the Human Protein Atlas).

Prognostic role of RhoA gene expression in liver cancer patients

To explore the association between RhoA mRNA expression and OS in liver cancer patients, we performed Kaplan-Meier survival analysis based on data from TCGA. The results indicated that the HCC patients with high RhoA expression had poorer prognoses (Figure 2A). Notably, the optimal cutoff point of RhoA mRNA expression was calculated based on the X-tile method, which can produce the optimal cutoff value to predict survival time (Figure 2B). Moreover, the tumor-promoting role of RhoA was also demonstrated by two genomic spatial event (GSE) datasets (GSE10186 and GSE 54236), in which the P values were 0.039 and 0.024, respectively (Figure 2C and Figure 2D).

Figure 2. RhoA expression level as an independent prognostic factor in HCC. The high expression level of RhoA suggests poor prognosis based on the TCGA training set (A), and the optimal cutoff point was calculated via the X-tile method (B). High RhoA expression levels were also unfavorable in two GEO validation sets (C, GSE10186; D, GSE54236). Univariate and multivariate Cox regression analyses of clinical indicators and RhoA levels related to prognosis: red bars represent prognostic factors, and blue bars represent nonprognostic factors (E).

To further investigate the independent prognostic value of RhoA, univariate and multivariate Cox regression analyses were conducted. As shown in Figure 2E, vascular invasion (HR: 1.274; 95% CI: 1.092–4.243, P = 0.027; Figure 2E) and high RhoA levels (HR: 2.578; 95% CI: 1.311–5.071, P = 0.006; Figure 2E) were associated with poor OS in the univariate analysis. In addition, multivariate analysis also demonstrated that the aforementioned factors were independent risk factors of OS (P < 0.05; Figure 2E).

RhoA expression is associated with gender

Next, we divided HCC patients from TCGA into a high-risk group and a low-risk group according to the optimal cutoff value of RhoA expression mentioned before to explore the relationships of RhoA expression with different clinicopathological parameters, and we only found that male patients had significantly higher RhoA gene expression levels than female patients (Table 1, Supplementary Table 1, and Figure 3A). However, the other clinical characteristics, such as different stage levels or grade levels of HCC, seemed to have no influence on RhoA expression levels (Table 1, Supplementary Table 1, Figure 3B, 3C).

Figure 3. RhoA mRNA expression in different groups of patients and differential expression verification. Relative expression of RhoA in males and females (A), stage I-II and stage III-IV (B), and grade 1–2 and grade 3–4 (C) from TCGA data. The relative RhoA expression in ten pairs of normal tissues and HCC tissues from Peking Union Medical College Hospital identified through real-time quantitative polymerase chain reaction (D). The representative immunohistochemistry images of HCC tissue (lower right) and adjacent normal tissue (upper right): The red squares shown in the figure represents the enlarged area. (E).

mRNA and protein expression validation of RhoA

We further extracted RNA from 10 frozen HCC and paired adjacent liver tissues to quantify RhoA gene expression. The HCC tissues did have a significantly higher RhoA expression, which was consistent with the pattern identified in the HPA. (Figure 3D). Moreover, the IHC test of RhoA showed that the hepatic cells in the normal tissue were mainly stained negatively except for portions of the interstitial substance (Figure 3E, upper parts), while tumor cells in the HCC tissue from the same section were stained positively (Figure 3E, lower parts). In summary, although the mRNA and protein levels of RhoA were significantly higher in the HCC tissues than in the normal adjacent tissues, the RhoA protein level, rather than the mRNA level, seems to have a better distinction ability, which indicates its potential as a diagnostic biomarker.

Diagnostic model based on RhoA protein levels in liver cancer patients

To test our hypothesis, we performed additional IHC staining with 30 normal tissue samples and 134 HCC tissue samples from an HCC patient cohort from PUMCH to detect the protein expression profile of RhoA. Consistent with the previous results, the HCC samples had significantly higher RhoA protein levels (Figure 4A). Next, we divided patients into two groups, one with a high RhoA protein level (3+) and the other with a low level (0−2+). Similar to the outcome of the analysis of the prognostic function of RhoA gene expression, the HCC patients with longer OS times had lower RhoA protein levels (Figure 4B). Considering that the protein level of RhoA has better tissue specificity than gene expression between liver tumor tissues and adjacent normal tissues, the RhoA protein expression based on the results of the IHC analysis of normal tissues was mainly undetected (0) or low (1+), while that of the cancerous tissues was usually medium (2+) or high (3+). We proposed a diagnostic model based on this phenomenon (Figure 4C). After validation, the sensitivity and specificity of the model were 92.5% and 90.0%, respectively (Figure 4D). The area under the curve (AUC) of the receiver operating characteristic (ROC) curve was 0.913 (Figure 4E), which indicates the good performance of this model. To exclude false positive caused by antibody non-specificity, we performed immunohistochemical tests on some samples with another RhoA antibody (1:50, AF6352, Affinity biosciences, USA) and the results are in line with previous one (data not shown). Overall, the diagnostic model based on the RhoA IHC results expanded our current detection approaches for HCC.

Figure 4. Diagnostic model of RhoA protein expression in liver cancer patients. The staining intensities of RhoA via immunohistochemistry chips from PUMCH patient samples (A). Kaplan-Meier curves of overall survival (B) of liver cancer patients with high RhoA protein expression levels (3+) and low RhoA protein expression levels (0–2+). Diagram (C), sensitivity and specificity validation (D) and receiver operating characteristic curve (E) of the diagnostic model according to RhoA immunohistochemistry level.

RhoA expression is modulated by genetic amplification and related KEGG pathways

Next, we explored the mechanisms of RhoA dysregulation using deep sequencing data from the University of California, Santa Cruz. After visualizing the somatic mutation, copy number variation, DNA methylation, and expression of RhoA, we found only four HCC samples with single nucleotide polymorphisms (SNPs), which demonstrated that point mutation did not play an essential role in RhoA expression (Figure 5A). Notably, the copy number variation and gene expression of RhoA shared a coexpression pattern, and the significant association was further validated via regression analysis (r = 0.563, P < 0.05, Figure 5B). However, the DNA methylation profile of the HCC tissues showed little variation (Figure 5A). To depict the exact relationships between DNA methylation and gene expression levels, we analyzed 405 samples with both methylation data and gene expression levels. As shown in Figure 5C, the increase in RhoA DNA methylation only slightly attenuated RhoA gene expression (r = −0.108, P < 0.05, Figure 5C). Furthermore, to detect the primary KEGG pathways in the two groups (RhoA high and low), we performed GSEA and found that samples with high levels of RhoA were enriched with genes mainly involved in the lysosome, pathogenic Escherichia coli infection, purine metabolism and pyrimidine metabolism pathways (Figure 5D, left part), while those with low levels of RhoA had predominant genes involved in the hedgehog signaling pathway, linoleic acid metabolism, olfactory transduction and taste transduction (Figure 5D, right part).

Figure 5. RhoA dysregulation and related KEGG analysis. Multiomic data of RhoA in liver cancer tissues are displayed in a heatmap (A). The correlation between RhoA gene expression level and copy number variation level (B) or DNA methylation level (C) were determined by regression analysis. The top 4 Kyoto Encyclopedia of Genes and Genomes pasthways identified via gene set enrichment analysis of tissues with high and low Rho expression levels (D).

Therefore, it seems that the copy number amplification level of RhoA, rather than the decrease in the DNA methylation level, plays a major role in the overexpression of RhoA in HCC tissues.

Bioinformatics data mining and processing

The mRNA level and protein level of RhoA in both normal human tissues and cancerous tissues, as well as gene expression in cancer cell lines, were examined using data from the Human Protein Atlas (HPA) (http://www.proteinatlas.org/) [30, 31]. Representative immunohistochemistry images of RhoA protein expression profiles in HCC samples with different staining intensities were also downloaded from the HPA. The differential mRNA expression level of RhoA between HCC tissues and normal tissues from the TCGA database and the Genotype-Tissue Expression (GTEx) project were analyzed via the Gene Expression Profiling Interactive Analysis (GEPIA) web tool (http://gepia.cancer-pku.cn/). The mRNA transcriptome data and clinical parameters of liver hepatocellular carcinoma (LIHC) patients were obtained from the TCGA data portal (https://portal.gdc.cancer.gov/). Two other HCC validation datasets (GSE10186 and GSE54236) containing mRNA expression data and corresponding survival times were retrieved and downloaded from the GEO database (https://www.ncbi.nlm.nih.gov/geo/). Histological type, somatic mutation, copy number variation, gene expression, and DNA methylation of RhoA in liver cancer were visualized via a heatmap generated from the Xena browser web tool (https://xenabrowser.net/), and data used to analyze the factors affecting RhoA gene expression were downloaded from the Xena browser web tool. Analyzing the data downloaded from the TCGA and GEO databases did not require the approval of an ethics committee. Gene Set Enrichment Analysis (GSEA) was conducted to compare the distinct Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways between the high RhoA level group and the low RhoA level group based on the TCGA HCC mRNA expression data.

Experimental validation with samples from Peking Union Medical College Hospital

From January 2004 to August 2016, thirty paired HCC and para-cancerous normal tissues and 104 HCC tissues from patients who underwent surgery and pathologically diagnosed as HCC at PUMCH were collected and made into four formalin-fixed paraffin-embedded (FFPE) tissue chips, the remaining tissues were frozen in liquid nitrogen and then stored at −80°C. Among these patient samples, the differential RhoA expression of ten frozen HCC tissues and paired adjacent normal tissues were validated by quantitative polymerase chain reaction (qPCR) as described previously [32]. The primer sequences were as follows: RhoA forward primer: 5′-AGCCTGTGGAAAGACATGCTT-3′; RhoA reverse primer: 5′-TCAAACACTGTGGGCACATAC-3′; β-actin forward primer: 5′-GCCGGGACCTGACTGACTAC-3′; β-actin reverse primer: 5′-CGGATGTCCACGTCACACTT-3′. Immunohistochemistry (IHC) sections of four FFPE tissue chips were stained with the RhoA antibody (EPR18134, RabMAb, UK) at a dilution of 1:100 to detect their protein profiles. In addition, nine IHC sections of the cancer margin with both normal and tumor components were used to elucidate the differential protein expression level of RhoA. The study was approved by the Medical Ethics Committee of PUMCH of the Chinese Academy of Medical Sciences (CAMS) & PUMC, and all patients signed informed consent forms.

Immunohistochemistry

Paraffin slices are cut into 3μm thick and placed on anti-unloading slides, baking 1 hour at 72°C. After 2–3 dewaxing with dimethylbenzene, slides are immersed into gradient alcohol (100%, 100%, 95% and 75%) successively. Rinse with PBS (1‰ Tween 20) for 5 times. Not less than 2 minutes each time. Preheat the repair solution in the pressure cooker to boiling then put the slides into the repair solution. After 2.5 minutes, wash slides to room temperature with cool water. Rinse with PBS for 5 times. Incubate with 3% H₂O₂ 10–15 minutes at room temperature. Wash with distilled water first, then PBS for 5 times. Add primary antibody and overnight at 4 °C. Rinse with PBS for 5 times. Incubate with one-step detection system (PV-8000, ZSGB-BIO, China) for 20 minutes at 37°C. Rinse with PBS for 5 times. Color development with DAB (ZLI-9018, ZSGB-BIO, China) 6–8 minutes. Termination reaction with water. Re-satin with hematoxylin and wash with water. Differentiated by hydrochloric alcohol quickly. Return blue for 3–5 minutes via repair solution and rinse with running water for 3–5 minutes. Seal slides after dehydration with gradient alcohol.

Interpretation of immunohistochemical results was performed by two independent experienced pathologists based on the proportion of positive cells without considering dyeing intensity. Histochemical score from 0 to 3 as follows: 0, 1–5% positive cells; 1, 6–25% positive cells; 2, 26–75% positive cells; and 3, >76% positive cells. Samples with discordant scores from the two pathologists were discussed and re-scored.

Statistical analysis

Statistical analysis was performed using R software v.3.5.2. Continuous variables are reported as the means ± standard deviations (SDs). Differences between groups were compared by unpaired Student’s t-test and visualized by GraphPad Prism v.8.0 (GraphPad Inc.). The associations between RhoA expression and clinicopathological parameters were evaluated using the χ² test. The OS curves of the above three HCC patient cohorts were generated based on the best cutoff value (X-tile algorithm) with the survival package in R. Univariate and multivariate Cox regression analyses were performed to identify independent prognostic variables based on RhoA level and other clinical characteristics, including age, gender, body mass index (BMI), stage, grade, fibrosis, AFP, extent of inflammation, vascular invasion, and family history. Unless noted otherwise, P < 0.05 was considered statistically significant. Linear regression analyses were conducted to assess the correlations between RhoA gene expression and copy number variation and between RhoA gene expression and DNA methylation. |R|>3 and P < 0.05 were considered relevant and statistically significant.