Study Characteristics
Our study explored the associations between PFDN1/2/3/4/5/6 and liver hepatocellular carcinoma in many aspects including transcriptional levels, protein levels, genes mutation, clinical significance,protein-protein interactions, GO/ KEGG enrichment analysis, and immune cells infiltration. The flow chart of the entire study design was presented in Fig. 1.
mRNA levels of PFDNs in LIHC patients
We first analyzed the transcriptional levels of PFDNs in LIHC and normal liver tissues by using the Oncomine database (http://www.oncomine.org). The results were shown in Fig. 2 and Table 1.the transcriptional levels of PFDN2/3/4/5/6 were significantly increased in LIHC.In Roessler Liver 2 dataset[26], PFDN2 was increased in hepatocellular carcinoma (fold change = 1.793, p = 1.29E-51) versus normal liver tissues.Roessler Liver statistics showed that PFDN2 was up-regulated in hepatocellular carcinoma (fold change = 1.584, p = 5.65E-6) compared to normal liver tissues[26]. Likewise, Wurmbach Liver statistics showed that the expression of PFDN2 was higher in hepatocellular carcinoma than normal liver tissues (fold change = 1.613,p = 6.64E-5)[27]. Roessler Liver 2 statistics showed that PFDN3 was increased in hepatocellular carcinoma (fold change = 2.364,p = 4.76E-66) versus normal liver tissues[26]. Similarly, Chen et al.[28]reported that PFDN3 was up-regulated in hepatocellular carcinoma (fold change = 1.635, p = 3.54E-15) compared to normal liver tissues.
In Roessler Liver dataset[26], PFDN4 was increased in hepatocellular carcinoma (fold change = 3.253, p = 6.11E-11) versus normal liver tissues. Likewise, Roessler Liver 2 statistics showed that the expression of PFDN4 was higher in hepatocellular carcinoma than normal liver tissues (fold change = 2.758,p = 6.50E-7 [26]. Mas et al.[29] showed that PFDN5 was up-regulated in cirrhosis (fold change = 1.696, p = 8.38E-12) compared to normal liver tissues.Additionally, in Chen Liver dataset[28],PFDN6 was increased in hepatocellular carcinoma (fold change = 1.651, p = 4.82E-14) versus normal liver tissues. Similarly,Roessler Liver statistics showed that PFDN6 was higher in hepatocellular carcinoma than normal liver tissues (fold change = 1.540,p = 1.14E-5)[26].
Table 1
Significant changes of the expression of PFDNs in mRNA level between LIHC and normal liver tissues(ONCOMINE)
|
Types of LIHC VS normal liver tissue
|
Fold Change
|
P value
|
T-test
|
Ref
|
PFDN1
|
NA
|
NA
|
NA
|
NA
|
NA
|
PFDN2
|
Hepatocellular Carcinoma
Hepatocellular Carcinoma
Hepatocellular Carcinoma
|
1.793
1.584
1.613
|
1.29E−51
5.65E−6
6.64E−5
|
17.327
5.015
4.569
|
Roessler Liver[26]
Roessler Liver 2[26]
Wurmbach Liver [27]
|
PFDN3
|
Hepatocellular Carcinoma
Hepatocellular Carcinoma
|
2.364
1.635
|
4.76E−66
3.54E−4
|
20.410
8.514
|
Roessler Liver 2[26]
Chen Liver[28]
|
PFDN4
|
Hepatocellular Carcinoma
Hepatocellular Carcinoma
|
3.253
2.758
|
6.11E−11
6.50E−71
|
8.970
21.896
|
Roessler Liver[26]
Roessler Liver 2[26]
|
PFDN5
|
Cirrhosis
|
1.696
|
8.38E−12
|
9.762
|
Mas Liver [29]
|
PFDN6
|
Hepatocellular Carcinoma
Hepatocellular Carcinoma
|
1.651
1.540
|
4.82E−14
1.14E−5
|
8.088
4.972
|
Chen Liver[28]
Roessler Liver[26]
|
NA: Not Available
|
Linkages of mRNA levels of PFDNs with clinicopathological parameters of
LIHC patients
We explored the mRNA expressions of PFDNs in LIHC and normal liver samples via the GEPIA2 database (http://gepia2.cancer-pku.cn/#index). The results showed that the expression levels of PFDN1/2/3/4/5/6 were higher in LIHC tissues than normal liver tissues(Fig. 3A, B).In addition, it was also confirmed that the expressions of PFDNs were higher in LIHC tissues than paracancerous ones based on a matched comparison of 50 samples(Fig. 3C). Subsequently, we evaluated the associations between the expressions of PFDNs family proteins and tumor grades and cancer staging by using the GEPIA2 database and the UALCAN database (http://ualcan.path.uab.edu).From Fig. 4A, we could see that the mRNA expressions of 6 PFDNs family proteins were significantly correlated with tumor grades. The highest mRNA levels of PFDN1/3/4/5/6 were found in grade 4, while the highest mRNA level of PFDN2 was discovered in grade 3. Likewise, form Fig. 4B, we could see that the mRNA levels of PFDN1/3/4/5 significantly changed(p<0.05) in different stages, whereas PFDN2 and PFDN6 did not(p>0.05).
We also analyzed the immunohistochemistry (IHC) of 6 PFDNs family proteins in LIHC tissues and normal liver tissues by using the HPA database (https://www.proteinatlas.org/). The result showed that medium or high expressions of PFDN1/2/3/4/5 were detected in LIHC tissues, whereas medium or low protein expressions of them were found in normal liver tissues(Fig. 5A). PFDN6 protein was not detected in LIHC tissue and normal liver tissue. All in all, the protein expression levels of PFDNs were higher in LIHC tissue than normal liver tissue.
The prognostic value of PFDNs in LIHC patients
We evaluated the prognostic value of PFDNs in the survival of LIHC via the online tool of Kaplan-Meier Plotter (http://kmplot.com/). The curves of overall survival(OS), relapse-free survival (RFS), progression-free survival(PFS) and disease-specific survival (DSS) were showed in Fig. 5B. High mRNA expressions of PFDN1/2/3/4 were relevant to poor OS(p<0.05). Besides, high mRNA expressions of PFDN2/3/4 were related to poor RFS and DSS(p<0.05), and high expression of PFDN6 was associated with poor RFS(p<0.05). However, the expression of PFDN5 was not significantly correlated with the survival of LIHC patients(p>0.05).
PFDN1/2/4 were independently correlated with prognosis in LIHC patients
After we showed the significant correlation between PFDN1/2/3/4/6 expressions with the survival of LIHC patients by K-M survival analysis, and then, we attempted to analyze the independent prognostic value of PFDNs in OS and PFS for hepatocellular carcinoma patients. The clinical data of 374 LIHC patients were acquired from The Cancer Genome Atlas (TGCA, https://portal.gdc.cancer.gov) database, and their basic clinical characteristics were presented in Supplementary Table 1.Subsequently, we used the statistical method of Cox regression analysis to assess these data. The results were shown in Supplementary Table 2 and Supplementary Table 3. Univariate analysis indicated that high T stage (HR = 2.109, p<0.001), high expression of PFDN1(HR = 1.454, p<0.035), PFDN2(HR = 1.460, p<0.032),PFDN3(HR = 1.277, p = 0.016),and PFDN4(HR = 1.789, p = 0.001) were significantly associated with poorer OS of LIHC patients. Besides, high T stage(HR = 2.384, p<0.001), high expression of PFDN1(HR = 1.390, p<0.028) and PFDN4 (HR = 1.494, p = 0.007) were significantly relevant to poorer PFS of LIHC patients.In multivariate analysis, we could see that T stage (HR = 2.178, p<0.001),PFDN1(HR = 1.187, p<0.017),PFDN2(HR = 1.3–71,p<0.021), and PFDN4 (HR = 1.465, p<0.024) were independently correlated with poorer OS of LIHC patients;T stage (HR = 2.349,p<0.001) was independently related to poorer PFS of LIHC patients. Taken together, high PFDN1/2/4 expressions were independent prognostic factors in OS for LIHC patients.
Genetic mutations and co-expression of PFDNs in LIHC patients
Genetic mutations and co-expression of the PFDNs family proteins were analyzed via the cBioPortal database(http://www.cbioportal.org/). As were shown in Fig. 6A and B. In total, 55%(199/360) of patients with LIHC had at least two genetic mutations. PFDN1,PFDN2,PFDN3,PFDN4,PFDN5 and PFDN6 were altered in 12%, 35%, 14%, 14%, 8%, and 13% of the 360 complete samples(Fig. 6A).Similarly, the degree of genetic mutations of PFDNs family proteins was presented in the Expression Heatmap(Fig. 6B), the frequency of PFDN2 mutation was highest among PFDNs in LIHC. Moreover, we analyzed the correlations among the 6 PFDNs family numbers, from Fig. 6C, we could see that there were meaningful and positive associations as follows: PFDN1 with PFDN2/4/5; PFDN2 with PFDN1/4/5/6; PFDN4 with PFDN1/2/5/6; PFDN5 with PFDN1/2/4/6; PFDN6 with PFDN2/4/5.
Genetic mutations and expression heatmap of PFDNs in LIHC patients (cBioPortal). (C) The correlations among different expressions of the 6 PFDNs family proteins (cBioPortal). (D)PPI network among the 6 PFDNs family proteins(STRING). (E)PPI network for PFDNs and their 50 frequently neighboring genes(STRING). (F)Nine hub genes includingRPL24, RPL37, RPL37A, RPS15, ZNF317, MYS6B, LSM7, PNAJC7, and ATP5Q2 were presented in PPI network (Cytoscape).
Established protein-protein interaction (PPI) network and analyzed the functions and pathways of PFDNs in LIHC patients
Firstly, we selected 50 most frequently altered neighboring genes of the 6 PFDNs proteins by using the GEPIA2 database. And then, we established the PPI network among the 6 PFDNs family proteins(Fig. 6D) and another PPI network for PFDNs and their 50 frequently neighboring genes (Fig. 6E) by using the tool of STRING(https://string-db.org/). Lastly, we screened out nine hub genes in these frequently neighboring genes by using the software of Cytoscape (https://cytoscape.org/). The result was showed in Fig. 6F, nine hub genes including RPL24, RPL37, RPL37A,PRS15, ZNF317, MYS6B,LSM7,PNAJC7, and ATP5Q2 were tightly related to the alterations of PFDNs.
Subsequently, we predicted the functions of PFDNs and their 50 frequently altered neighboring genes by analysis of GO and KEGG using DAVID6.8(https://david.ncifcrf.gov/).The GO enrichment analysis showed GO:00000398(mRNA splicing via spliceosome),GO:0006364(rRNA processing),GO:000 0184(nudcar-transcribed mRNA catabolic process),GO:0006457(protein folding), GO:0005634(nudeus),GO:0005737(cytoplasm),GO:0005654(nudeoplasm),GO:0005515(protein binding),GO:0003723(RNA binding) and GO:0044822(poly(A)RNA RNA binding) were meaningful modulated by alterations of the 6 PFDNs family proteins (Fig. 7A, B, C). Moreover, KEGG pathway enrichment analysis indicated that hsa03010 (Ribosome), hsa03040(Spliceosome) and hsa03013(RNA transport) were significantly related to alterations of the 6 PFDNs family proteins(Fig. 7D).
Immune cell infiltration of PFDNs in LIHC patients
Lastly, we attempted to analyse the correlation between 6 PFDNs family proteins and infiltrated immune cells by using the TIMER database (https://cistrome.shinyapps.io/timer/). The results were shown in Fig. 8.PFDN1 expression was positively correlated with the infiltration of B cells (Cor = 0.319,p = 1.46E-9), CD8+T cells (Cor = 0.273,p = 1.63E-7),CD4+T cells(Cor=0.357,p = 9.07E-12),macrophage(Cor = 0.404,p = 7.72E-15), neutrophil(Cor = 0.358,p= 7.12E-12) and dendritic cells (Cor = 0.351,p = 2.79E-11)(Fig. 8A).PFDN2 expression was positively associated with the infiltration of B cells (Cor = 0.153,p = 4.42E-3), CD4+T cells (Cor = 0.137,p = 1.12E-2) ,macrophage(Cor=0.126,p = 1.98E-2) and neutrophil(Cor = 0.123,p = 2.19E-2) (Fig. 8B).PFDN3 expression was positively related to the infiltration of B cells(Cor = 0.353, p = 1.61E-11),CD8+T cells(Cor = 0.32,p = 3.35E-09),CD4+T cells(Cor = 0.312,p = 3.35 E-9),macrophage(Cor = 0.421,p = 4.72E-16),neutrophil(Cor = 0.453,p = 6.84E-19) and dendritic cells (Cor = 0.453,p = 1.23E-18) (Fig. 8C). PFDN4 expression was positively relevant to the infiltration of B cells (Cor = 0.316,p = 2.05E-9), CD8+T cells (Cor = 0.329,p = 4.37E-10), CD4+T cells(Cor = 0.204,p = 1.36E-4), macrophage (Cor = 0.419,p = 6.67E-16), neutrophil(Cor = 0.265,p = 5.63E-7) and dendritic cells (Cor = 0.346,p = 5.16E-11) (Fig. 8D). PFDN5 expression was positively correlated with the infiltration of B cells(Cor = 0.244,p = 4.82E-6), CD8+T cells(Cor = 0.338,p = 1.36E-10),macrophage(Cor = 0.244,p = 5.15E-6),neutrophil(Cor = 0.132,p = 1.40E-2) and dendritic cells (Cor = 0.278,p = 1.93E-7) (Fig. 8E). However, PFDN6 was not significantly correlated with the above six types of infiltrated immune cells(all p>0.05)(Fig. 8F).