Abstract

   Background: The clinical value of the biomarkers of bladder cancer (BC)
   is limited due to their low sensitivity or specificity. As a biomarker,
   DLG associated protein 5 (DLGAP5) is a potential cell cycle regulator
   in cancer cell carcinogenesis. However, its functional part in BC
   remains unclear. Therefore, this study aims to identify DLGAP5
   expression in BC and its potential diagnostic and prognostic values.
   Eventually, it predicts the possible RNA regulatory pathways of BC.

   Methods: Data on DLGAP5 expression levels in BC and normal bladder
   tissues were obtained from The Cancer Genome Atlas (TCGA) and the Gene
   Expression Omnibus (GEO) databases. The receiver operating
   characteristic (ROC), Kaplan–Meier survival curves, and the univariate
   and multivariate Cox regression analysis determined the diagnostic and
   prognostic values of DLGAP5 in BC patients. Finally, the StarBase
   predicted the target RNAs and constructed networks using Cytoscape.

   Results: DLGAP5 expression was significantly upregulated in BC tissue,
   verified by the TCGA (p < 0.001), [37]GSE3167, [38]GSE7476, and
   [39]GSE65635 datasets (p < 0.01). BC patients with increased DLGAP5 had
   poor overall survival (OS) (p = 0.01), disease specific survival (DSS)
   (p = 0.006) and progress free interval (DFI) (p = 0.007). The area
   under the ROC curve (AUC) was 0.913. The multivariate Cox analysis
   identified that lymphovascular invasion (p = 0.007) and DLGAP5 (p =
   0.002) were independent prognostic factors.

   Conclusion: Increased DLGAP5 expression was closely associated with a
   poor prognosis in BC patients. In this case, DLGAP5 might be a
   diagnostic and prognostic biomarker for BC. DLGAP5 expression might be
   regulated by NEAT1/MALAT1/XIST/PKD--Hsa-mir-101-3p pathways.

   Keywords: bladder cancer, DLGAP5, immune cell infiltration, RNA
   regulatory pathways, ceRNA network

Introduction

   Bladder cancer (BC) is one of the most common urinary tumors globally
   and the 10th most common cancer. Around 549,393 new cases were
   diagnosed and 199,922 deaths were caused by BC worldwide in 2018
   ([40]Bray et al., 2018). Indeed, BC has significantly impacted
   patients’ health and brought a substantial economic burden ([41]Antoni
   et al., 2017). Although there are various diagnostic methods for BC, it
   is prone to recurrence ([42]Patel et al., 2020), and postoperative
   diagnosis typically requires invasive examination, known as cystoscopy
   ([43]DeGeorge et al., 2017). Therefore, it is significant to look for
   new and non-invasive diagnostic markers. Suppose we can look for
   diagnostic and prognostic markers of BC, it is beneficial to the early
   diagnosis and better treatment to improve the prognosis of patients and
   reduce economic burden. Existing articles have reported that
   ubiquitin-conjugating enzyme E2C (UBE2C) ([44]Morikawa et al., 2013),
   ubiquitin-like with PHD and ring-finger domains 1 (UHRF1), nuclear
   matrix protein22 (NMP-22), human complement factor H-related protein
   ([45]Unoki et al., 2009), leucine-rich and immunoglobulin-like domains
   (LRIGs), and asymptomatic microhematuria (AMH) ([46]Soria et al., 2019)
   are potential prognostic or diagnostic markers of BC. However, the
   molecular mechanisms of BC remain unclear. Furthermore, the clinical
   value of these markers is limited due to their low sensitivity or
   specificity ([47]Ng et al., 2021; [48]Tran et al., 2021). Hence, it is
   urgent to discover new markers of BC for early diagnosis and systematic
   treatment.

   DLG associated protein 5 (DLGAP5), also called Human hepatoma
   up-regulated protein (HURP), has an important biological function as a
   signaling molecule for having a guanylate kinase-associated protein
   homology domain ([49]Tong et al., 2014). Besides, DLGAP5 is a potential
   cell cycle regulator in cancer cell carcinogenesis ([50]Horning et al.,
   2018). A previous study report that DLGAP5 is a mitotic regulator of
   tumorigenesis primarily in cell cycle progression, capture of
   kinetochores, and generation of proper tension across sister
   kinetochores. DLGAP5 can promote microtubule polymerization, bipolar
   spindle formation and the stability of microtubules ([51]Wong and Fang,
   2006). Another study find that DLGAP5 plays an important role in
   multicomponent complex which affect the growth or stability of spindle
   ([52]Koffa et al., 2006). [53]Eissa et al. (2014). found that the
   expression of DLGAP5 (HURP) RNA in urine sampler could be detected,
   which might be a biomarker of bilharzial and nonbilharzial bladder
   cancer. It is reported that DLGAP5 may be a potential prognosis factor
   in lung cancer, breast cancer, pancreatic cancer, endometrial cancer,
   and colorectal cancer ([54]Shi et al., 2017; [55]Branchi et al., 2019;
   [56]Ke et al., 2020; [57]Xu et al., 2020; [58]Zhang et al., 2020).
   Patients with increased DLGAP5 expression have shown worse prognoses in
   above cancers.

   Recently, bioinformatics analysis have been widely used in cancers to
   identify new biomarkers for improving diagnosis and treatment
   ([59]Kristensen et al., 2014). Moreover, the ceRNA hypothesis was
   proposed by [60]Salmena et al. (2011). In this paper, it described
   clearly the theory of ceRNA, which revealed a new interaction mechanism
   between RNA having a great biological significance. The ceRNA
   hypothesis may explain disease processes and provide opportunities for
   new therapies. For instance, it is known that ceRNA with the binding
   site of miRNA can combine with miRNA and keep it silent. ([61]Thomson
   et al., 2016). As far as we can retrieve, this study is the first to
   analyze the differential expression and determine the diagnostic and
   prognostic values of DLGAP5 in BC. Eventually, it has explored the
   potential DLGAP5-related RNA regulatory pathways of BC.

Material and Methods

RNA-Sequencing Data and Bioinformatics Analysis

   We obtained 10,534 samples from TCGA to conduct a pan-cancer analysis,
   including 33 types of tumors. Nineteen pan-carcinoma tissues and 414
   tumor tissues of the bladder from TCGA were identified to analyze
   differential DLGAP5 expression. Additionally, the [62]GSE3167,
   [63]GSE7476, and [64]GSE65635 datasets were obtained to validate the
   DLGAP5 expression level between normal and tumor tissues of the bladder
   from the GEO database.

Diagnostic and Prognostic Analyses of DLGAP5

   The Receiver Operating Characteristic (ROC) survival curve was
   constructed to analyze the diagnostic values of DLGAP5 in BC patients.
   Firstly, The Kaplan–Meier (KM) method constructed the overall survival
   (OS) curve, the disease specific survival (DSS) curve, and the
   progress-free interval (PFI) curve, illustrating the prognostic
   association between DLGAP5 expression level and OS, DSS, and PFI.
   Secondly, the univariate and multivariate Cox regression analysis
   analyzed the prognostic values of DLGAP5 in BC patients; when p < 0.05,
   the Cox regression analysis result is statistically significant.

Construction of the Clinical Characteristics Table

   The characteristic clinical information of 414 BC patients, including
   age, TNM stage, pathologic stage, histologic grade, lymphovascular
   invasion, whether a smoker or not, OS event, DSS event, and PFI event,
   were obtained from the TCGA database. Based on the DLGAP5 expression
   level, the DLGAP5-positive BC patients were divided into high and low
   expression groups. Then, the correlation between DLGAP5 expression
   level and clinical characteristics was analyzed.

Construction of the Protein-Protein Interaction Network

   Related genes of DLGAP5 were acquired to construct the PPI network
   using the STRING online tool ([65]https://string-db.org/). The
   screening criteria were as follows: 1) minimum required interaction
   score: highest confidence (0.900); 2) maximum number of interactors to
   show no more than ten interactors.

Enrichment Analysis

   The R software (version 4.1.1) clusterProfiler package analyzed the GO
   enrichment of DLGAP5 and its related genes. A bubble plot was
   constructed for visualizing these enrichment results. The R software
   also performed the KEGG pathway enrichment analysis of DLGAP5 and its
   associated genes. When the adjusted p-value < 0.05, the enriched
   functions and pathways were identified as significant results.

Association Analysis Between Immune Cell Infiltration and DLGAP5

   The GSVA package with single-sample Gene Set Enrichment Analysis
   (ssGSEA) of the R software analyzed and visualized the association
   between DLGAP5 expression and 24 kinds of immune cells, including aDC,
   B cells, CD8 T cells, Cytotoxic cells, DC, Eosinophils, iDC,
   Macrophages, Mast cells, Neutrophils, NK CD56bright cells, NK CD56dim
   cells, NK cells, pDC, T cells, T helper cells, Tcm, Tem, TFH, Tgd, Th1
   cells, Th17 cells, Th2 cells, and Treg. The applied statistical method
   was Spearman’s analysis. A lollipop graph was constructed to depict it.
   Then, the correlation between DLGAP5 expression and the enrichment of
   statistical immune cells was determined in more detail. For further
   exploration, the association between statistical immune cells and
   high/low-DLGAP5 was analyzed.

Prediction of Target miRNAs of DLGAP5

   It is reported that miRNAs can repress the function of mRNAs having a
   negative relation ([66]McGeary et al., 2019). In our study, we want to
   explore the miRNAs of DLGAP5. The StarBase (version 2.0)
   ([67]http://starbase.sysu.edu.cn/index.php) is a free and
   multifunctional database, which can provide the most comprehensive
   CLIP-seq data for exploring the interaction of all kinds of RNAs
   ([68]Li et al., 2014). The StarBase can predict target miRNAs of DLGAP5
   for us, including seven online miRNA databases, namely: PITA, RNA22,
   miRmap, microT, miRanda, PicTar, and TargetScan. The screening criteria
   were as follows: mammal, human, hg19, miRNA: all, CLIP-Data: high
   stringency (≥3), Degradome Data: with or without data, pan-cancer: with
   or without data. The target miRNAs were those found in at least three
   databases. Then, a correlation network of DLGAP5 and seven target
   miRNAs was constructed using the Cytoscape software.

Construction of ceRNA Networks

   The StarBase was also used to predict lncRNAs and circRNAs of seven
   selected target miRNAs. The screening criteria of lncRNA were as
   follows: mammal, human, hg19, CLIP Data: stringency> = 5, Degradome
   data: with and without data. The screening criteria of circRNA were as
   follows: mammal, human, hg19, CLIP Data: stringency > = 5, Degradome
   data: stringency > = 3. A ceRNA network was constructed encompassing
   all selected target miRNAs, lncRNAs, and circRNAs and DLGAP5 using the
   Cytoscape software.

Prognostic Analysis of Hsa-Mir-124-3p and Hsa-Mir-101-3p

   The pan-cancer analysis module of StarBase finished the survival
   analysis of hsa-mir-124-3p and hsa-mir-101-3p and the analysis of
   miRNA-DLGAP5 co-expression.

Statistics Analysis

   The R software was used to perform statistical analysis and
   visualization. The Chi-square test, the Fisher’s exact test, and the
   Wilcoxon rank-sum test were used in clinical information analysis
   ([69]Wang et al., 2021). The Spearman’s analysis was used in immune
   cell infiltration. The cox regression analysis was used in the
   prognostic analysis of Kaplan–Meier. The Wilcoxon rank-sum test and
   matched t-test were used in the differential expression analysis.
   Log-rank test was used in the prognostic analysis of Hsa-mir-101-3p.
   Statistical significance was set at p-value <0.05 ([70]Ge et al.,
   2021).

Results

Clinical Characteristics of BC Patients

   The results determined that histologic grade (p < 0.001),
   lymphovascular invasion (p = 0.027), OS event (p = 0.046), DSS event (p
   = 0.023), and PFI event (p = 0.036) exhibited a statistical difference.
   No statistical difference in age (p = 0.618) and TNM stage (p = 0.721,
   0.147, 0.882) was observed ([71]Table 1).

TABLE 1.

   The clinical characteristics in BC with low and high expression of
   DLGAP5
   Characteristic Low expression of DLGAP5 High expression of DLGAP5 P
   n 204 204
   Age, n (%) 0.618
   <=70 112 (27.5%) 118 (28.9%)
   >70 92 (22.5%) 86 (21.1%)
   T stage, n (%) 0.721
   T1 2 (0.5%) 1 (0.3%)
   T2 60 (16%) 59 (15.8%)
   T3 96 (25.7%) 98 (26.2%)
   T4 33 (8.8%) 25 (6.7%)
   N stage, n (%) 0.147
   N0 116 (31.7%) 121 (33.1%)
   N1 18 (4.9%) 28 (7.7%)
   N2 45 (12.3%) 30 (8.2%)
   N3 4 (1.1%) 4 (1.1%)
   M stage, n (%) 0.882
   M0 103 (49.8%) 93 (44.9%)
   M1 5 (2.4%) 6 (2.9%)
   Pathologic Stage, n (%) 0.637
   Stage I 2 (0.5%) 0 (0%)
   Stage II 66 (16.3%) 64 (15.8%)
   Stage III 67 (16.5%) 73 (18%)
   Stage IV 68 (16.7%) 66 (16.3%)
   Histologic Grade, n (%) <0.001[72] ^a
   High Grade 184 (45.4%) 200 (49.4%)
   Low Grade 19 (4.7%) 2 (0.5%)
   Lymphovascular invasion, n (%) 0.027[73] ^a
   No 55 (19.6%) 75 (26.7%)
   Yes 85 (30.2%) 66 (23.5%)
   Smoker, n (%) 0.385
   No 59 (14.9%) 50 (12.7%)
   Yes 139 (35.2%) 147 (37.2%)
   OS event, n (%) 0.046[74] ^a
   Alive 125 (30.6%) 104 (25.5%)
   Dead 79 (19.4%) 100 (24.5%)
   DSS event, n (%) 0.023[75] ^a
   Alive 149 (37.8%) 123 (31.2%)
   Dead 51 (12.9%) 71 (18%)
   PFI event, n (%) 0.036[76] ^a
   Alive 128 (31.4%) 106 (26%)
   Dead 76 (18.6%) 98 (24%)
   [77]Open in a new tab
   ^a

   p <0.05.

Increased Expression of DLGAP5 in BC

   A pan-cancer analysis of differential DLGAP5 expression was performed.
   It revealed that DLGAP5 expression level increased in BC, breast
   cancer, cervical squamous carcinoma, cholangiocarcinoma, colon
   adenocarcinoma, esophageal carcinoma, glioblastoma multiforme, head,
   and neck squamous cell carcinoma, kidney cancer, hepatocellular liver
   carcinoma, lung adenocarcinoma, lung squamous cell carcinoma,
   pancreatic adenocarcinoma, rectum adenocarcinoma, prostate
   adenocarcinoma, stomach adenocarcinoma, thyroid carcinoma, and uterine
   corpus endometrial carcinoma ([78]Figure 1A). In addition, DLGAP5
   expression level significantly increased in BC tissues than in normal
   tissues (p < 0.05), as validated from the TCGA database (p < 0.001)
   ([79]Figures 1B,C) and the [80]GSE3167, [81]GSE7476, and [82]GSE65635
   datasets from the GEO database (p < 0.01) ([83]Figures 1D–F).

FIGURE 1.

   [84]FIGURE 1
   [85]Open in a new tab

   Different expression level of DLGAP5. (A) A graph was constructed to
   show the expression levels of DLGAP5 between normal tissue with bladder
   tumor tissue for 33 types of cancers. High expression of DLGAP5 in
   bladder tumor (***). (B) The different expression of DLGAP5 in bladder
   tumor tissues compared to adjacent tissues from TCGA database (p <
   0.001). (C) The different expression of DLGAP5 in bladder tumor samples
   and matched adjacent samples from TCGA database (***). (D) [86]GSE3167,
   (E) [87]GSE7476 and (F) [88]GSE65635 from GEO database were used to
   verify high expression of DLGAP5 in bladder tumor (p < 0.01). *: p <
   0.05, **: p < 0.01, ***: p < 0.001.

Diagnosis and Prognosis Analysis

   The KM survival curves of OS (p = 0.01), DSS (p = 0.006), and PFI (p =
   0.007) depicted that the increased DLGAP5 expression was associated
   with poor prognosis for BC patients ([89]Figures 2A–C). The ROC
   survival curve was constructed to analyze the diagnostic values of
   DLGAP5 in BC patients. The result identified that the ROC curve (AUC)
   area was 0.913 ([90]Figure 2D). The univariate Cox analysis revealed
   that age (p = 0.006), TNM stage (p < 0.001), pathologic stage (p <
   0.001), lymphovascular invasion (p < 0.001), and DLGAP5 (p = 0.041)
   were associated with OS events. The multivariate Cox analysis
   determined that lymphovascular invasion (p = 0.007) and DLGAP5 (p =
   0.002) were associated with OS events ([91]Table 2).

FIGURE 2.

   [92]FIGURE 2
   [93]Open in a new tab

   The diagnostic and prognostic analysis Of DLGAP5. (A) High DLGAP5 had
   worse overall survival (OS) for bladder cancer patients (p = 0.01, HR =
   1.48). (B) The bladder cancer patients with high expression of DLGAP5
   had worse disease specific survival (DSS) (p = 0.006, HR = 1.66). (C)
   The bladder cancer patients with high expression of DLGAP5 had worse
   progress free interval (DFI) (p = 0.007, HR-1.52). (D) The diagnostic
   analysis of DLGAP5 in bladder cancer patients. the area under the ROC
   curve (AUC) was 0.913, CI: 0.850–0.976. When the p < 0.05, the result
   was regarded as statistical significance.

TABLE 2.

   The univariate and multivariate cox analysis
   Characteristics Total (N) Univariate analysis Multivariate analysis
   HR (95% CI) P value HR (95% CI) P value
   Age (< = 70 vs. > 70) 407 1.514 (1.128−2.031) 0.006[94] ^a 1.165
   (0.630−2.152) 0.627
   Gender (Male vs. Female) 407 0.874 (0.631−1.210) 0.417
   T stage 373
   T1 and T2 122 Reference
   T stage (T1 + T2 vs. T3) 193 1.917 (1.302−2.824) < 0.001[95] ^a 3.155
   (0.702−14.182) 0.134
   T stage (T1 + T2 vs. T4) 58 2.867 (1.779−4.620) < 0.001[96] ^a 2.974
   (0.616−14.354) 0.175
   N stage (N0 + N1 vs. N2 + N3) 365 2.240 (1.609−3.120) < 0.001[97] ^a
   1.459 (0.684−3.112) 0.329
   M stage (M0 vs. M1) 207 3.310 (1.582−6.926) 0.001[98] ^a 1.712
   (0.456−6.425) 0.426
   Pathologic stage (Stage I + Stage II vs. Stage III + Stage IV) 405
   2.203 (1.521−3.190) < 0.001[99] ^a 0.416 (0.079−2.196) 0.301
   Histologic grade (LG vs. HG) 404 2.899 (0.717−11.718) 0.135
   Lymphovascular invasion (No vs. Yes) 280 2.249 (1.547−3.268) <
   0.001[100] ^a 2.696 (1.316−5.522) 0.007[101] ^a
   DLGAP5 (Low vs. High) 407 1.362 (1.013−1.831) 0.041[102] ^a 2.925
   (1.501−5.699) 0.002[103] ^a
   [104]Open in a new tab

   HR: Hazard ratio, LG: low grade, HG: High grade.
   ^a

   P value < 0.05.

PPI Network Analysis

   The PPI network included DLGAP5, CCNB1, TOP2A, CDC20, AURKA, TPX2,
   KIF11, ASPM, NUSAP1, CDK1, and TTK ([105]Figure 3B).

FIGURE 3.

   [106]FIGURE 3
   [107]Open in a new tab

   The GO/KEGG enrichment analysis and PPI network. (A) A bubble plot was
   used to show the results of the GO/KEGG enrichment. The color of bubble
   was on behalf of the adjusted value (p.adjust) and the size of bubble
   represent the degree of enrichment. (B) The PPI network was used to
   show DLGAP5 and its correlated genes including CCNB 1, TOP2A, CDC20,
   AURKA, TPX2, KIF11, ASPM, NUSAP1, CDK1, and TTK.

Enrichment Analysis

   The GO enrichment included three parts: Molecular Function (MF),
   Biological Process (BP), and Cellular Component (CC). According to the
   adjusted p-value < 0.05 and q-value < 0.2, four KEGG pathways and five
   terms were selected per GO enrichment parts to perform a bubble plot
   ([108]Figure 3A). The result indicated that KEGG pathways mainly
   focused on the cell cycle, p53 signaling pathway, cellular senescence,
   and viral carcinogenesis. The GO enrichment terms of BP were also
   mainly enriched in the nuclear division, the positive regulation of the
   cell cycle, the signal transduction involved in DNA damage checkpoint,
   the signal transduction involved in cell cycle checkpoint, and the
   regulation of nuclear division. The terms of CC included the spindle,
   the spindle pole, the serine/threonine protein kinase complex, the
   protein kinase complex, and the cyclin-dependent protein kinase
   holoenzyme complex. The terms of MF had protein
   serine/threonine/tyrosine kinase activity, cyclin-dependent protein
   serine/threonine kinase activity, ATP-dependent microtubule motor
   activity plus-end-directed, cyclin binding, RNA polymerase II and CTD
   heptapeptide repeat kinase activity. All selected pathways and terms
   are in [109]Table 3.

TABLE 3.

   The overview of KEGG pathways and GO enrichment terms selected.
   Ontology ID Description p.adjust
   BP GO:0000280 Nuclear division 1.20042E-13
   BP GO:0090068 Positive regulation of cell cycle process 5.47663E-06
   BP GO:0072422 Signal transduction involved in DNA damage checkpoint
   5.84843E-05
   BP GO:0072395 Signal transduction involved in cell cycle checkpoint
   6.01927E-05
   BP GO:0051783 Regulation of nuclear division 2.18971E-10
   CC GO:0005819 Spindle 1.30276E-15
   CC GO:0000922 Spindle pole 1.86509E-11
   CC GO:1902554 Serine/threonine protein kinase complex 0.003376654
   CC GO:1902911 Protein kinase complex 0.00455199
   CC GO:0000307 Cyclin-dependent protein kinase holoenzyme complex
   0.000825205
   MF GO:0004712 Protein serine/threonine/tyrosine kinase activity
   0.006568804
   MF GO:0004693 Cyclin-dependent protein serine/threonine kinase activity
   0.048549788
   MF GO:0008574 ATP-dependent microtubule motor activity,
   plus-end-directed 0.048549788
   MF GO:0030332 cyclin binding 0.048549788
   MF GO:0008353 RNA polymerase II CTD heptapeptide repeat kinase activity
   0.048549788
   KEGG hsa04110 Cell cycle 5.45668E-06
   KEGG hsa04115 p53 signaling pathway 0.003542622
   KEGG hsa04218 Cellular senescence 0.012683523
   KEGG hsa05203 Viral carcinogenesis 0.016449973
   [110]Open in a new tab

   KEGG: kyoto encyclopedia of genes and genomes, GO: gene ontology.

Analysis of Immune Cell Infiltration With DLGAP5

   The R software was used to analyze and visualize the relationship
   between DLGAP5 and 24 kinds of immune cells and construct a lollipop
   graph ([111]Figure 4A). The result identified that DLGAP5 expression
   level had a positive association with Th2 cells (p < 0.001, r = 0.700),
   Th1 cells (p < 0.001, r = 0.280), T helper cells (p < 0.001, r =
   0.271), Tgd (p < 0.001, r = 0.263), aDC (p < 0.001, r = 0.240),
   Macrophages (p < 0.001, r = 0.218), NK CD56dim cells (p < 0.001, r =
   0.177), Tcm (p < 0.001, r = 0.185), Treg (p = 0.018, r = 0.116), and
   Neutrophils (p = 0.035, r = 0.104), while NK CD56bright cells (p <
   0.001, r = −0.413), pDC (p < 0.001, r = −0.332), Mast cells (p < 0.001,
   r = −0.270) and DC (p = 0.004, r = −0.142) had a negative association
   ([112]Figure 4B–O). The enrichment of 15 kinds of statistical immune
   cells in high and low DLGAP5 were further analyzed; Th2 cells (p <
   0.001), Th1 cells (p < 0.001), T helper cells (p < 0.001), Tgd (p <
   0.001), aDC (p < 0.001), Macrophages (p < 0.001), NK CD56dim cells (p <
   0.001), Tcm (p < 0.001), TReg (p < 0.001), and Neutrophils (p = 0.001)
   had higher enrichment in high DLGAP5 than in low DLGAP5 ([113]Figures
   5A–J), while NK CD56bright cells, pDC and Mast cells had higher
   enrichment in low DLGAP5 than in high DLGAP5 ([114]Figures 5K–M). DC (p
   = 0.123) and iDC (p = 0.245) had no statistical significance between
   the two groups ([115]Figures 5N,O).

FIGURE 4.

   [116]FIGURE 4
   [117]Open in a new tab

   Analysis of immune cell infiltration with DLGAP5. (A) a lollipop graph
   was constructed to show the correlation between DLGAP5 with 24 types of
   immune cells. The different degree color of bubble was on behalf of the
   p value and the size of bubble represented the degree of correlation.
   The left terms were negative correlation and the right terms were
   positive correlation when p value < 0.05. 0: p < 0.05, a*: p < 0.01.
   (B) Th2 cells (p < 0.001,1-0.700) (C) T helper cells (p < 0.001,
   t-0.271) (D) Thl cells (p < 0.001,1-0.280) (E) Tgd (p < 0.001, r-0.263)
   (F) aDC (r < 0.001, r = 0.240) (G) NK CD56dim cells (p < 0.001, r =
   0.177) (H) TReg (p = 1.018, r = 0.116) (K) Macrophages (p < 0.001, 1 =
   0.218) (L) Tem (p < 0.01:11,-i = 0.185), and (M) Neutrophils (p-1.035,
   1 = 0.104) had positive correlation with DLGAP5. (I) NK CD56bright
   cells (p < 0.001 (N) pDC (p < 0.001, r = −0.332) (J) Mast cells (p <
   0.001, t = −0.270) and (O) DC (p = 0.004, r--0.142) had negative
   association with DLGAP5.

FIGURE 5.

   [118]FIGURE 5
   [119]Open in a new tab

   The enrichment of 15 kinds of statistical immune cells in high and low
   DLGAP5. Th2 cells (p < 0.001), Thl cells (p < 0.001), T helper cells (p
   < 0.001), Tgd (p < 0.001), aDC (p < 0.001), Macrophages (p < 0.001), NK
   CD56dim cells (p < 0.001), Tam (p < 0.001), TReg (p < 0.001) and
   Neutrophils (p = 0.001) had more higher enrichment level in high level
   than low level of DLGAP5 (A–J), while NK CD56bright cells, pDC and Mast
   cells had more higher enrichment level in low level than high level of
   DLGAP5 (K–M). DC (p,1).123 and iDC (p = 0.245) have no statistical
   significance between two groups (N,O).

Prediction of Target miRNAs

   The target miRNAs were selected in at least three databases. Seven
   target miRNAs were predicted by StarBase, including hsa-mir-101-3p,
   hsa-mir-124-3p, hsa-mir-409-5p, hsa-mir-515-5p, hsa-mir-506-3p,
   hsa-mir-653-5p, and hsa-mir-1252-5p. Cytoscape was used to construct a
   correlation network ([120]Figure 6A).

FIGURE 6.

   [121]FIGURE 6
   [122]Open in a new tab

   DLGAP5-miRNAs correlation network and ceRNA network. (A) A correlation
   network between DLGAP5 with seven target miRNAs. Seven target miRNAs of
   DLGAP5, including hsa-mir-101-3p, hsa-mir-124-3p, hsa-mir-409-5p,
   hsa-mir-515-5p, hsa-mir-506-3p, hsa-mir-653-5p, and hsa-mir-l252-5p,
   were predicted and selected by StarBase database. (B) DLGAP5 and all
   selected target miRNAs, lneRNAs and circRNAs were constructed a ceRNA
   network by using Cytoscape software. (C) The ceRNA network of the
   potential RNA regulatory pathways. The red hexagon represents DLGAP5,
   blue diamond represents lncRNA, yellow round rectangle represents
   miRNA, and green triangle represents circRNA.

Prediction of Target LncRNAs and CircRNAs to Construct the ceRNA Network

   StarBase was used to predict the lncRNAs and circRNAs of seven target
   miRNAs. A ceRNA network was constructed encompassing all selected
   target miRNAs, lncRNAs, and circRNAs, and DLGAP5 using Cytoscape
   software ([123]Figure 6B). Finally, Hsa-mir-101-3p had 6 lncRNAs, and
   12 circRNAs, Hsa-mir-124-3p had 5 lncRNAs, and 14 circRNAs,
   Hsa-mir-409-5p had 3 lncRNAs, and 3 circRNAs, Hsa-mir-515-5p had 2
   lncRNAs, and 23 circRNAs, Hsa-mir-506-3p had 5 lncRNAs and 15 circRNAs,
   Hsa-mir-653-5p had 4 lncRNAs, and 6 circRNAs, and Hsa-mir-1252-5p had 8
   lncRNAs, and 13 circRNAs. The ceRNA hypothesis indicated that miRNA was
   down-regulated when gene expression increased. We have retrieved papers
   in the PubMed database and selected downregulated miRNAs from seven
   target miRNAs; The selected miRNAs were hsa-mir-124-3p and
   hsa-miR-101-3p, which reported and confirmed deregulating expression in
   BC. Then, we selected three upregulated lncRNAs and two upregulated
   circRNAs of the two downregulated miRNAs in BC. The upregulated lncRNAs
   were NEAT1, MALAT1, and XIST, while the two upregulated circRNAs were
   PKD2, and VIM. Finally, a ceRNA network of potential RNA regulatory
   pathways to regulate BC was constructed ([124]Figure 6C).

Prognosis Analysis of Hsa-Mir-124-3p and Hsa-Mir-101-3p

   The increased expression of hsa-mir-101-3p had a positive prognosis for
   BC patients (p = 0.041). The hazard ratio was 0.73 ([125]Figure 7A).
   However, hsa-mir-124-3p had no significance for the prognosis of BC
   patients (p = 0.71) ([126]Figure 7B). Hence, we explored the
   co-expression between hsa-mir-101-3p with DLGAP5 in BC. The result
   determined that the hsa-mir-101-3p expression level was negatively
   associated with DLGAP5 expression level (r = −0.415). The p-value was
   2.21e-18 ([127]Figure 7C), regarded as statistically significant.

FIGURE 7.

   [128]FIGURE 7
   [129]Open in a new tab

   Prognosis analysis of hsa-mir-124-3p and hsa-mir-101-3p. (A) The K-M
   survival curve of hsa-miR-101-3p. High expression of hsa-miR-101-3p had
   positive overall survival for bladder cancer patients (Log-Rank p =
   0.041, Hazard Ratio = 0.73). Green curve represented high expression of
   hsa-mir-101-3p and brown curve represented low expression of
   hsa-mir-101-3p. (B) The K-M survival curve of hsa-mir-124-3p. The
   result was no statistical significance (Log-Rank p = 0.71, Hazard Ratio
   = 1.07). (C) The co-expression between hsa-mir-101-3p with DLGAP5 in
   bladder cancer patients. The result shows that hsa-mir-101-3p
   expression level had negative relation with DLGAP5 expression level (r
   = −0.415).The p value was 2.21e-18.

Discussion

   The DLGAP5 expression profile in BC has not yet been systematically
   investigated. This study demonstrated that DLGAP5 was overexpressed and
   a negative prognostic factor for BC patients by bioinformatics
   analysis. Previous articles determined that DLGAP5 overexpression had
   also been shown in lung cancer, breast cancer, pancreatic cancer,
   endometrial cancer, and colorectal cancer ([130]Shi et al., 2017;
   [131]Branchi et al., 2019; [132]Ke et al., 2020; [133]Xu et al., 2020;
   [134]Zhang et al., 2020). Meanwhile, increased DLGAP5 in the above
   cancers has an adverse prognosis. Coincidentally, the study’s findings
   were consistent with previous studies. Horning AM and colleagues
   demonstrated that DLGAP5 played an essential role in cancer cell
   carcinogenesis ([135]Horning et al., 2018). The characteristic clinical
   analysis found that the expression of DLGAP5 had a significant relation
   with histologic grade. Therefore, this study hypothesizes that DLGAP5
   plays a vital role in the pathogenesis of BC and may be an excellent
   prognosis biomarker for BC patients. The ROC analysis is typically used
   to estimate the diagnostic value, which has an excellent diagnostic
   efficiency ([136]Mandrekar, 2010). As far as we can retrieve, DLGAP5 as
   a diagnostic marker of BC has not been reported. Surprisingly, we
   demonstrated that DLGAP5 had an excellent diagnostic performance in BC
   patients (AUC = 0.913), implying that DLGAP5 is a potential diagnostic
   biomarker of BC.

   To clarify the biological role of DLGAP5 in BC, we investigated its
   most relative genes to perform GO/KEGG enrichment analysis. A previous
   study concluded that DLGAP5 was a mitotic spindle protein and targets
   cell cycle controllers ([137]Horning et al., 2018). Furthermore,
   Hatfield et al. reported that DLGAP5 was a cell cycle gene associated
   with proliferation in acute myeloid leukemia patients ([138]Hatfield et
   al., 2014). Meanwhile, Kuo and colleagues showed that DLGAP5
   overexpression in HEK293 cells promoted the ubiquitination of p53 and
   its degradation by the proteasome ([139]Kuo et al., 2012). Furthermore,
   after DLGAP5 overexpression, the expression of exogenous p53 in H1299
   and Hep3b cells decreased, while their sensitivity to cisplatin-induced
   apoptosis also decreased. In this study, GO/KEGG enrichment analysis
   results demonstrated that DLGAP5 was a vital gene related to the cell
   cycle and p53 signaling pathways. Furthermore, DLGAP5 might be a novel
   therapeutic target of BC.

   In this study, we demonstrated that DLGAP5 overexpression in BC
   patients was prone to lymphovascular invasion. Vittorio Branchi found
   that the knockdown of DLGAP5 led to a significant reduction of the
   invasion and migration in colorectal cancer ([140]Branchi et al.,
   2019). [141]Liao et al. (2013). reported that silencing the expression
   of DLGAP5 suppressed cell growth significantly and colony formation in
   hepatocellular carcinoma cells in vitro. Zhang et al. also reported
   that knockdown of DLGAP5 suppressed cell proliferation, induced G2/M
   phase arrest and apoptosis in ovarian cancer ([142]Zhang et al., 2021).
   However, the functional experiment of DLGAP5 in BC was not observed.
   Based on the facts mentioned above, we hypothesized that DLGAP5 plays
   an essential role in tumorigenesis and the development of BC.

   As we all know, the tumor microenvironment (TME) is complex and closely
   related to tumorigenesis and development, mainly composed of immune and
   stromal cells ([143]Anderson and Simon, 2020). These cells can perform
   both tumor-promoting and anti-tumor functions. TME supports tumor
   growth and development, local invasion, metastasis, and spread
   ([144]Pitt et al., 2016). Therefore, a viewpoint helped that TME was a
   new target for treating and preventing malignant tumors ([145]Pitt et
   al., 2016; [146]Zhou et al., 2019). In the present study, we found that
   the expression of DLGAP5 has the most positive correlation with Th2
   cells and has the most negative correlation with NK CD56bright cells in
   BC. However, it may be beneficial to the further research of new
   therapeutic targets for BC patients.

   The ceRNA hypothesis reveals a new interaction mechanism between RNA,
   which regulates the occurrence and development of diseases ([147]Tay et
   al., 2014; [148]Thomson et al., 2016; [149]Yang et al., 2016).
   According to the ceRNA hypothesis, we selected two down-regulated
   miRNAs from predicted miRNAs, including hsa-mir-124-3p and
   hsa-mir-101-3p, reported and confirmed deregulating expression in BC
   ([150]Zhang et al., 2014; [151]Yuan et al., 2017; [152]Li et al.,
   2019). Previous articles said that hsa-mir-101-3p was related to the
   proliferation and metastasis of BC cells ([153]Liu et al., 2017;
   [154]Li et al., 2019). Besides, previous studies showed that three
   lncRNAs, including NEAT1, MALAT1, and XIST ([155]Hu et al., 2017;
   [156]Guo et al., 2018; [157]Jiao et al., 2018; [158]Liu et al., 2019;
   [159]Shan et al., 2020), and two circRNAs, including PKD2 and VIM
   ([160]Fang et al., 2013; [161]Li et al., 2018), were upregulated in BC.
   Furthermore, NEAT1, MALAT1, XIST, and PKD had reported or validated
   having a close relation with pro chemotherapy resistance of malignant
   tumors ([162]Gutschner et al., 2013; [163]Hu et al., 2017; [164]Kim et
   al., 2018; [165]Li et al., 2018; [166]Shan et al., 2020). However, we
   demonstrated that high expression of hsa-mir-101-3p has an overall
   adverse survival for BC patients, which indicated that hsa-mir-101-3p
   might be a protective factor having a negative relation with DLGAP5 in
   BC patients. The above result was in accord with the ceRNA hypothesis.
   Eventually, we proposed the lncRNA‐miRNA‐mRNA pathway:
   NEAT1/MALAT1/XIST‐‐hsa‐mir‐101‐3p‐‐DLGAP5 and the circRNA‐miRNA‐mRNA
   pathway: PKD‐‐hsa‐mir‐101‐3p‐‐DLGAP5 are potential RNA regulatory
   pathways of BC, which regulated the occurrence, development,
   metastasis, chemosensitivity resistance of BC. There is no doubt that
   this study has some limitations. First, DLGAP5 requires validation in
   further fundamental studies. Some experiments will be conducted to
   verify the conclusions. Second, the number of normal samples is small.

Conclusion

   This study found that the increased expression of DLGAP5 had an
   unfavorable prognosis for BC patients and DLGAP5 might be a diagnostic
   and prognostic biomarker for BC. Additionally,
   NEAT1/MALAT1/XIST/PKD--hsa-mir-101-3p might regulate DLGAP5 expression,
   providing the basis for the further study of BC.

Acknowledgments