PHLPP2: A Prognostic Biomarker in Adenocarcinoma of the Rectum

Background/Aims: Adenocarcinoma of the rectum (READ) is typically diagnosed at advanced stages due to a lack of early-onset specific features. Materials and Methods: The study used bioinformatics analysis of READ ribonucleic acid sequencing data from The Cancer Genome Atlas database to identify differentially expressed genes (DEGs). Overlapping genes between DEGs and autophagy-associated genes were screened for prognosis-associated DEGs, which were then validated in the OncoLnc database. Results: A total of 129 autophagy-associated DEGs were identified, with 17 genes found to be associated with READ prognosis. Multivariate Cox regression analysis revealed that only the PHLPP2 gene was significantly associated with READ prognosis (hazard ratio = 0.442, P = .026), and its low expression correlated with low survival in patients with brain lower-grade glioma (P = .00623) and pancreatic adenocarcinoma (P = .00109). Conclusions: PHLPP2 expression may serve as a READ-specific prognostic biomarker and is involved in the PI3K-Akt signaling pathway.


INTRODUCTION
Adenocarcinoma of the rectum (READ) is a type of colorectal cancer (CRC) that originates in the rectum and rectal tube.It is one of the malignant tumors with poor prognosis in patients with advanced tumors.The 5-year survival ratio of patients diagnosed with early READ was 70%-90%, and that of patients with advanced READ was less than 60%, 1 even less than 30%. 2,3However, most READs were diagnosed at advanced stages due to the lack of specific features of early READ.
Over the past 2 decades, a number of high-quality studies have shown that clinical factors are associated with the early-onset, progression, and prognosis of CRC, [4][5][6] while little data has been reported on READ. 2,7,8Also, the analysis of genetic research is also unfair.For instance, various genetic factors including messenger ribonucleic acids (mRNAs), microRNAs (miRNAs), long noncoding RNAs, and mutations have shown an association with the prognosis of CRC. 9-146][17][18] Therefore, identification of diagnostic or prognostic biomarkers for READ is still necessary.
Advances in computational bioinformatics have attracted a great amount of interest in the field of cancer research.Computational bioinformatics analysis is of great value in identifying potent prognostic biomarkers, and some are likely to be used clinically.For instance, Pan et al 19 showed that integrating 5 CRC-related miRNAs (including miR-15b, miR-17, miR-21, miR-26b, and miR-145) and serum carcinoembryonic antigen provided good diagnostic performance in CRC prognosis.Hansen et al 20 showed that patients with a loss mutation of caudal-related homoeobox transcription factor 2 (CDX2) had a poor prognosis in 2 Denmark clinical cohorts.The correlation of CDX2 loss mutation with colon cancer prognosis had been previously reported by Dalerba et al 21 using bioinformatics analysis.Since computational bioinformatics facilitates the identification of potential biomarkers, its popularization will provide a valuable reference to the features of cancers with unknown or unclear pathogenesis, including READ.
This study aimed to identify potential prognostic biomarkers in READ based on computational bioinformatics analysis, which would provide a novel genetic reference to the pathogenesis and development of READ.

MATERIALS AND METHODS The Cancer Genome Atlas Data Collection
The RNA-seq data (Illumina HiSeq 2000 RNA Sequencing)  were extracted from The Cancer Genome Atlas (TCGA) database.A total of 177 samples, including 158 samples with clinical information and 9 nontumor adjacent tissues, were analyzed.The data files were downloaded and used for further analyses.

Identification of Differentially Expressed Genes
The differentially expressed genes (DEGs) in the READ tumor samples were identified using the R Limma package (https ://bi ocond uctor .org/packa ges/r eleas e/bio c/ htm l/lim ma.ht ml; version 3.6.1).Differentially expressed genes were screened out according to the following criteria: log 2 (fold change) > 1, P < .05,and false discovery rate (FDR) < .05.

Extraction of Autophagy-Associated Genes
To understand the molecular changes mediated by autophagy, the autophagy-associated genes were extracted from the Comparative Toxicogenomics Database (CTD; http: //ctd base.org/a bout/ ; 2020 update) using the search keyword "autophagy" and the National Center for Biotechnology Information (NCBI) gene database with the following search phrase: (autophagy) AND "Homo sapiens."Also, the items included in the Human Autophagy Database (HADb; http: //www .autophagy .lu/)were downloaded.Then, the genes overlapped between DEGs and at least 1 of the 3 databases (CTD, HADb, and NCBI) were retained and used for further analyses.

Construction of the Protein-Protein Interaction Network
The protein-protein interaction (PPI) network was constructed for the autophagy-associated DEGs to show the potential interactions between the genes.The predictive interaction pairs were extracted from the STRING source (https ://st ring-db.org/cgi /inpu t.pl; version 11.0).Interaction pairs with a score of higher than 0.4 were downloaded.Then, the PPI network was constructed using the Cytoscape (http: //app s.cyt oscap e.org /apps / all; version 3.8.0).The significant modules with a module score of higher than 5.0 in the PPI network were identified using the MCODE plugin in the Cytoscape (http: //app s.cyt oscap e.org /sear ch?q= MCODE ).

Identification of Adenocarcinoma of the Rectum Prognosis-Associated Genes
The prognosis-associated genes in READ were identified using the Cox regression analysis based on the TCGA cohort.Briefly, the expression profiles of the autophagyassociated DEGs including in the PPI network, clinical overall survival time, and death data were extracted from the TCGA cohort.Then, the univariate and multivariate Cox regression analyses were used to screen the prognosis-associated DEGs.Significant items were identified when P < .05.The associations of the above selected prognosis-associated DEGs with the overall survival and prognosis in other cancers were validated in the OncoLnc database (http: //www .oncolnc.o rg/).

Searching of the Pathways Related to Prognosis-Associated Differentially Expressed Genes
At last, we constructed the molecular regulatory network involving the prognosis-associated DEGs based on the searching result in the KEGG PATHWAY Database (https ://ww w.keg g.jp/ kegg/ pathw ay.ht ml).The pathways associated with the prognosis-associated DEGs and the nodes that interacted with them in the PPI were extracted from the KEGG database.Then, the mRNA-pathway regulatory network was constructed using the Cytoscape (version 3.8.0).

Statistical Analysis
The difference in the expression level of DEGs between tumors with different clinical stages and metastatic statuses was analyzed using the nonparameter Mann-Whitney U-test or the Kruskal-Wallis H test. Also, the Cox regression analysis was performed using the Statistical Package for the Social Sciences 22.0 software (IBM corp., Armonk, NY, USA).Hazard ratio and 95% CI values were analyzed.For all analyses, a significant difference was defined at P < .05.

Screening of the Differentially Expressed Genes in Adenocarcinoma of the Rectum Tumor
Using the aforementioned criteria, a total of 1790 DEGs in tumor samples were screened out from the TCGA cohort.The volcano figure of the DEGs is shown in Figure 1.

Identification of Autophagy-Associated Differentially Expressed Genes in Adenocarcinoma of the Rectum Tumor
Based on the searching in CTD and NCBI databases, 175 and 1619 autophagy genes were identified.Also, 232 autophagy-associated genes were downloaded from the HADb (Figure 2).The Venn diagram indicated that 129 DEGs were included in at least 1 of the 3 databases.The list of the 129 autophagy-associated DEGs is shown in Supplementary Table 1.

Protein-Protein Interaction Construction and Module Identification
A total of 402 interaction pairs between the 129 autophagy-associated DEGs were obtained from the STRING database.Then, the PPI network derived from the interaction pairs was constructed (Figure 3), which consisted of 111 nodes (gene products) and 402 edges (interaction pairs).The 24 nodes with top degrees (≥10) in the PPI network is shown in Table 1, including KIT proto-oncogene, receptor tyrosine kinase (KIT, degree = 28), cyclindependent kinase inhibitor 2A (CDKN2A, degree = 20), and interleukin 17A (IL17A, degree = 20).

Identification of the Prognosis-Associated Genes
Then, all 111 DEGs including in the PPI network were used to identify the prognostic genes.Univariate Cox regression analysis identified 17 genes were associated with the prognosis of READ (Supplementary Table 2 and Table 2), while multivariate Cox regression analysis showed the PHLPP2 gene (downregulated) was the only correlated with the survival outcome of READ (HR = 0.442, 95% CI 0.215-0.906,P = .026;Supplementary Table 2 and  Table 2).Cox regression also indicated that patients with a high expression level of PHLPP2 had a higher survival ratio (HR = .546,95% CI 0.347-0.858,P = .009;Figure 4).

Association of PHLPP2 with the Prognosis of Other Human Cancers
Based on the OncoLnc database, we found the high expression of PHLPP2 was correlated with higher survival percent of patients with brain lower-grade glioma (LGG; logrank P = .00623;Figure 5) and pancreatic adenocarcinoma (PAAD; logrank P = .00109;Figure 6).These results showed that patients with low expression of PHLPP2 were at higher risk of poor outcomes for patients with LGG and PAAD.We did not observe its association with the prognosis of other cancers, including the colon adenocarcinoma (COAD, logrank P = .119;Figure 7).

Illustration of the Signaling Pathways Associated with PHLPP2
To illustrate the potential molecular mechanism mediated by PHLPP2, we identified the gene-pathway regulatory network involving PHLPP2 and the DEGs that interacted with it in the PPI network.PHLPP2 was directly enriched in the "hsa04151: PI3K-Akt signaling pathway" (Figure 8).The 3 DEGs including downregulated serum/   glucocorticoid regulated kinase 1 (SGK1), protein kinase AMP-activated catalytic subunit alpha 2 (PRKAA2), and AKT serine/threonine kinase 3 (AKT3) were associated with 3, 19, and 93 pathways, respectively.However, the gene-pathway network was constructed using the PPI network and the pathways that had been reported to be associated with cancers.Accordingly, 23 and 11 pathways related to AKT3 and PRKAA2, respectively, were   retained and used for the construction of the regulatory network.Accordingly, we speculated that the association of PHLPP2 with the development and prognosis of READ may be associated with various signaling pathways, including "hsa04151: PI3K-Akt signaling pathway," "hsa04068: FoxO signaling pathway," "hsa04370: VEGF signaling pathway," "hsa04066: HIF-1 signaling pathway," "hsa04630: JAK-STAT signaling pathway," "hsa04668: TNF signaling pathway," "hsa04010: MAPK signaling pathway," and "hsa04210: Apoptosis."

DISCUSSION
In this study, the DEGs in the tumor samples were identified and used for the screening of prognosis-associated gene.The results showed that only PHLPP2 was associated with the prognosis of READ among the known autophagy-associated genes.PHLPP2 was downregulated in the READ tumor samples as compared with the nontumor adjacent tissues.Also, we found the high expression of PHLPP2 was associated with a higher survival ratio in patients with READ, LGG, and PAAD but not in COAD.These results might show that PHLPP2 was a READ-specific prognostic biomarker.
Autophagy is essential for cell survival and differentiation, as well as for homeostasis and disease development.It is a lysosomal degradation pathway that plays a key role in diverse pathologies, including tumorigenesis, neurodegeneration, inflammation, and aging. 22- 24Accordingly, there has been a tremendous increase in autophagy research in the past 10 years, which has increased the number of autophagy-related genes and proteins reported.6][27] These research studies indicated that autophagy inhibition may be an effective therapeutic strategy in advanced cancers. 25,26,28ence, our present study focused on the association of autophagy-associated genes with the prognosis of READ.
Fortunately, we identified that the expression of the PHLPP2 gene was associated with a high survival probability in patients with READ.
The PHLPP protein directly binds to and inactivates Akt and protein kinase C (PKC). 29,30The knockdown of PHLPP2 is shown to increase the activities of its downstream targets, including GSK3, Akt1, Akt3, and FoxO.
It inactivates Akt via the dephosphorylation of serine 473. 30,31Similarly, PHLPP inactivates PKC via the dephosphorylation of serine 657 (in PCKα). 32Accordingly, PHLPP expression suppresses cell survival and promotes cell apoptosis in cancer cells. 29,30[35][36][37] The PHLPP2 gene is expressed in all organs, but at its highest level in the small intestine, followed by the colon, duodenum, testis, and brain. 38However, the expression of PHLPP2 was reported to be lost or greatly decreased in tumor samples. 29,39Liu et al 39 indicated that the expression of PHLPP1 or PHLPP2 isoform was lost or decreased in more than 70% of colon tumor specimens as compared with the adjacent normal mucosa.We identified that the PHLPP2 gene was downregulated in the READ tumor samples compared with control, and the downregulation of PHLPP2 was associated with a poor prognosis.Also, it was only enriched with the PI3K/AKT signaling.These data suggested its crucial role in regulating READ tumor progression.However, the downregulation of the AKT3 gene in READ tumor indicated that the molecular mechanism underlying PHLPP2-associated READ progression might not be as simple as they appear, and should be examined carefully.
Given the connection to autophagy, PHLPP2 did not directly control or regulate autophagy.Peng et al 35 reported that PHLPP2 inhibited bladder cancer invasion by promoting the degradation of matrix metalloproteinase 2 (MMP2) via p62-mediated autophagy.Jin et al 40 identified that PHLPP2 showed a distinct function in bladder cancer.They found that the inhibition of PHLPP2 in bladder cancer cells promoted BECN1/Beclin1 degradation, attenuated autophagy, and promoted bladder cancer growth.They also identified that PHLPP2 mediated the stabilization of BECN1/Beclin1 indirectly by cullin 4A (CUL4A) and promoted autophagy. 40In other words, the connection between PHLPP2 and autophagy is not direct, and accordingly, the association of PHLPP2 with autophagy in cancers needs to be explored.

CONCLUSIONS
This study showed a positive correlation between PHLPP2 expression and READ prognosis.The PHLPP2 gene was downregulated in the tumor samples, and its high expression level was correlated with a higher survival ratio in patients with READ.It may be a READ-specific prognostic biomarker, providing a novel reference for treatment of READ.However, the association of it with PI3K/AKT signaling and the autophagy in READ progression needs to be explored.

Figure 1 .
Figure 1.The volcano figure of the DEGs in the TCGA tumor samples compared with the controls.Upregulated (log 2 [fold change] >1, P < .05,and false discovery rate (FDR) < .05)and downregulated DEGs (log 2 [fold change] <−1, P < .05,and FDR < .05)were indicated as red and blue nodes, respectively.DEGs, differentially expressed genes; TCGA, The Cancer Genome Atlas.Figure 2. The Venn diagram representing the DEGs that overlapped between the autophagy-associated genes in the 3 databases.Overlapping genes indicated by red stars (n = 129) were the identified autophagy-associated DEGs and used for further analyses.CTD, Comparative Toxicogenomics Database; DEGs, differentially expressed genes; HADb, Human Autophagy Database; NCBI, National Center for Biotechnology Information.

Figure 3 .
Figure 3.The protein-protein interaction (PPI) network consisting of the autophagy-associated genes.The overall PPI network consisting of the 111 differentially expressed genes (DEGs) associated with autophagy.The node size corresponds to interaction degree, and node color indicates log2[fold change] level.Orange and blue notes upregulation (log2[fold change] >1) and downregulation (log2[fold change] <−1), respectively.The darker the color, the greater the |log 2 (fold change)| value.

Figure 4 .
Figure 4. Cox regression analysis showing the association of PHLPP2 expression with the prognosis of patients with rectum adenocarcinoma.P = .009.HR, hazard ratio.

Figure 5 .
Figure 5.The association of PHLPP2 expression with the prognosis of brain lower grade glioma.The analysis was performed based on the OncoLnc database, logrank P = .00623.

Figure 7 .
Figure 7.The association of PHLPP2 expression with the prognosis of colon adenocarcinoma.The analysis was performed based on the OncoLnc database.

Figure 8 .
Figure 8.The potential gene-pathway regulatory network involving PHLPP2 in rectum adenocarcinoma.Blue nodes are downregulated genes (log2[fold change] < −1) in rectum adenocarcinoma.The pathways (green nodes) were extracted from the KEGG database.

Figure 6 .
Figure 6.The association of PHLPP2 expression with the prognosis of pancreatic adenocarcinoma.The analysis was performed based on the OncoLnc database, logrank P = .00109.

Table 1 .
Top 20 Nodes with Relatively High Interaction Degree in the Protein-Protein Interaction Network FC, fold change.

Table 1 .
The List of the Autophagy-Associated Genes Overlapped between the Differentially Expressed Genes in Rectum Adenocarcinoma

Table 2 .
Cox Regression Analysis for the Prognosis-Associated Genes in Rectum Adenocarcinoma Patients from the TCGA Database

Table 2 .
Cox Regression Analysis for the Prognosis-Associated Genes in Rectum Adenocarcinoma Patients from the TCGA Database (Continued) HR, hazard ratio.CI, confidential interval.TCGA, The Cancer Genome Atlas.