This article is mentioned in:

Introduction

Cancer has contributed to the rising rate in disease-associated mortality in the last few decades. The demand for early-stage diagnosis, prognosis and therapeutic targets is increasing. The serum deprivation response (SDPR) gene has been characterized to serve a critical role in breast cancer as a tumor suppressor (1,2), and recently showing a characteristic gene signature in specific types of cancer, including oral cancer, thyroid cancer and liposarcoma (3–5).

SDPR localizes to chr2q 32–33, also known as caveolae associated protein 2, and has been known for its role in caveolae formation (6–8). Its encoding protein SDPR, which is overexpressed in serum starved cells, was firstly identified as a substrate for protein kinase C (PKC) phosphorylation, an interaction that targets PKC in caveolae formation (9). Caveolae are plasma membrane microdomains involved in multiple biological processes, including lipid metabolism, endocytosis, cellular signal transduction, cell proliferation and migration (7,10).

Previous studies have gradually implicated the differential expression and tumor suppressor function of SDPR in cancer progression and metastasis; reduced SDPR expression has been observed in breast (1,2,11–13), kidney (11,14) and prostate (11,15) cancer. In breast cancer, SDPR was identified as a tumor suppressor (1,2). SDPR serves an anti-metastatic function by promoting apoptosis (1), and depletion of SDPR induced epithelial-mesenchymal transition (EMT) through transforming growth factor-β (TGF-β) signaling activation (2). The reduction of SDPR expression has been reported to be associated with significantly reduced survival in patients with breast cancer, who underwent therapy (1).

In addition to breast cancer, it has been suggested that SDPR may serve as a tumor suppressor gene, with a broader clinical relevance, in other types of cancer. In oral cancer, it was identified that SDPR-negative patients had high tumor progression (5), whereas in sarcoma (SARC), a lower SDPR expression was observed in more aggressive or dedifferentiated tumor forms (4). In addition, it was suggested that the differently expressed SDPR gene can be used as a possible diagnostic marker to discriminate malignant tumors from benign formations, not only in serum from patients with kidney tumors (14), but also in follicular thyroid carcinomas (3). Nevertheless, gene expression alterations of SDPR in various types of cancer and their relevance to clinical outcomes remain unclear.

In the present study, mRNA levels of SDPR were compared in various unique tumor tissue datasets compared with normal tissue datasets, indicating that SDPR was downregulated in various types of cancer. In addition, downregulated SDPR was found to be significantly associated with the survival of the patients, not only in previously reported breast cancer cases, but also in brain, lung and soft tissue tumors. However, SDPR failed to emerge as a frequent target for gene mutational inactivation in a previous next-generation sequencing study (16). As SDPR has been reported to be hypermethylated and silenced in breast cancer cell lines, it was suggested that it is likely to be epigenetically inactivated in cancer (1). In order to examine the mechanism underlying SDPR downregulation in cancer, the present study analyzed and found SDPR gene methylation alteration between cancer and normal tissues. Furthermore, the present study investigated the methylation sites relevant to the survival of patients with lung adenocarcinoma (LUAD) and SARC. In addition, the potential transcription factor binding sites of the SDPR promoter were analyzed. The potential transcription factor GATA binding protein 2 (GATA2) was identified from the analysis of genes that are co-expressed with SDPR. The results of the present study provide additional insight in understanding the underlying molecular mechanisms of SDPR in cancer, in addition to revealing a novel approach in the clinical prognostic evaluation and treatment of LUAD and SARC.

Materials and methods

Gene expression analysis using oncomine platform

The profile of SDPR gene expression level in various types of cancer was identified in Oncomine™ Platform and the detailed datasets are available online (https://www.oncomine.org/) (17). By comparing with normal tissues, the mRNA expression-fold of SDPR in cancer tissues was obtained using the parameters of P-value <10−4 or 0.01, fold-change >2, and gene ranking in the top 10%. To adjust the false discovery rate, the P-values were corrected by using the Benjamini-Hochberg procedure (B-H method) (18) in R, version 3.5.0 (https://www.r-project.org/).

Prognoscan database analysis

The association between SDPR expression and survival in different types of cancer was analyzed using the PrognoScan database (http://www.abren.net/PrognoScan/) (19), and presented as a Kaplan-Meier plot, in which survival curves for high (red) and low (blue) expression groups dichotomized at the optimal cut-point were plotted. The P-values were adjusted for multiple correlation testing using the Miller and Siegmund formula (20), according to Prognoscan database and shown as a corrected P-value (Pcor). The threshold was adjusted to corrected P-values at <0.05.

MethHC database analysis

The MethHC database was used in the analysis of SDPR DNA methylation alternation in cancer. MethHC (http://MethHC.mbc.nctu.edu.tw/) is a systematic database integrating DNA methylation data from The Cancer Genome Atlas (TCGA; http://cancergenome.nih.gov/abouttcga/policies/informedconsent), which includes >6,000 DNA methylation data generated by Illumina HumanMethylation450K BeadChip in 18 types of cancer. The methylation status of DNA was represented as β-values (0–1) (21), and the average β-value of SDPR was presented as a boxplot by comparing the transcript expression in tumor samples and matched normal samples in all 18 types of cancer. To adjust the false discovery rate, the P-values were corrected using the B-H method in R software.

Methsurv database analysis

Methsurv database (https://biit.cs.ut.ee/methsurv) was utilized for survival analysis in different types of cancer based on SDPR methylation patterns. In Methsurv, the gene methylation data was from TCGA Genome Data Analysis Center Firehose (http://gdac.broadinstitute.org/) (22), using the HM450K array, which covers 486,428 CpGs. The methylation status of DNA was represented as β-values (0–1) (23). The methylation pattern was annotated by probes indicating subregions of the query gene, according to the annotation file (Human genome build 27) provided by Illumina [TSS to-200 nucleotides upstream of TSS (‘TS200’); covering-200 to-1500 nucleotides upstream of TSS (‘TSS1500’; first exon (‘1st exon’); ‘5′UTR’, ‘body’ and ‘3′UTR’]. Clustering analysis was plotted and visualized using a heatmap by integrating Methsurv settings with ClustVis (https://biit.cs.ut.ee/clustvis/) (24). The survival analysis of each type of cancer between the low-methylated and the high-methylated groups in specific methylation sites was visualized using Kaplan-Meier plots. Multivariable survival analysis was performed using a Cox proportional-hazards model. Age and sex were used as covariates in the multivariable prediction models. The hazard ratio (HR) with 95% CI was derived from Cox fitting. The goodness-of-fit of the Cox proportional hazard model was assessed using a likelihood-ratio (LR) test, and presented as an LR P-value. The methylation status of SDPR in different LUAD clinical stage samples was shown as violin plots after grouping samples according to stage.

Identification of potential transcription factor for SDPR

The transcription start site (TSS) of SDPR was indicated by the University of California, Santa Cruz (UCSC) Genome Browser (http://genome.ucsc.edu), and the DNA sequence from 2,000 bp nucleotides upstream and 500 bp nucleotides downstream of the TSS was used as a potential promoter sequence for SDPR. PROMO at the ALGGEN server was subsequently used to identify the putative transcription factor binding sites in this promoter sequence (25). The co-expression profiles of the SDPR gene in LUAD were identified and presented as the pattern of a heat map using the Oncomine database (17), in which the node correlation value is computed as the average of all pair-wise correlations among genes. The node correlation value >0.5 was used to define significant co-expressing genes. Finally, the intersection of the above two profiles was investigated in order to identify the potential transcription factor regulating SDPR expression.

Results

Downregulation of SDPR in various types of cancer

The expression of SDPR is nearly ubiquitous in normal tissues and increased expression levels have been reported in the heart and lungs, while lower expression levels have been indicated in the kidney, the brain, the pancreas, skeletal muscle and the liver (8). To explore the gene expression alteration of SDPR in tumor tissues, the present study compared the mRNA levels of SDPR in various unique tumor tissue datasets compared with normal tissue datasets from the Oncomine™ Platform (17). Consistent with previous studies, the analysis showed that the expression of SDPR was downregulated in breast (1,11–13), kidney (11,14) and prostate cancer (11,15), and SARC (4,26). In addition, the present results identified that SDPR is downregulated in bladder, lung, cervical, colorectal, gastric, ovarian and pancreatic cancer (Fig. 1).

Figure 1. SDPR mRNA expression in different cancer types. (A) With a P-value <0.01 threshold level, the comparison illustrated the number of datasets with SDPR mRNA overexpression (left column, red) and downregulation (right column, blue) in cancer compared with normal tissue. (B) With a P-value <10−4 threshold level, the comparison of the number of datasets with SDPR mRNA overexpression (left column, red) and downregulation (right column, blue) in cancer compared with normal tissue. The cell color was determined by the best gene rank percentile for the analyses within the cell. Common thresholds: Fold change of 2, gene rank at the top 10%. SDPR, serum deprivation response; CNS, central nervous system.

Furthermore, gene expression levels of SDPR in cancer subtypes were estimated as a fold-change and gene rank (Table I). The fold change was from-44.083 (invasive ductal breast carcinoma) to-2.102 (cervical squamous cell carcinoma). The gene rank was between 6 and 1% (in the top %). For instance, the entire subtype dataset of breast cancer showed 1% downregulated gene ranks. Downregulated gene expression of SDPR was also observed in lung cancer (in top 3–1%), pancreatic cancer (in top 1%) and SARC (in top 4–1%).

Table I. SDPR expression in cancer.

Table I.

SDPR expression in cancer.

A, Bladder cancer
Author, year	Cancer subtype	P-value	Fold change	Gene rank (%)	Samples (n)	(Refs.)
Sanchez-Carbayo et al, 2006	Superficial bladder cancer	<0.001	−6.264	1	28	(34)
Kim et al, 2010	Superficial bladder cancer	<0.001	−3.076	3	126	(35)
Sanchez-Carbayo et al, 2006	Infiltrating bladder urothelial carcinoma	<0.001	−3.517	2	81	(34)
Kim et al, 2010	Infiltrating bladder urothelial carcinoma	<0.001	−2.315	6	62	(35)
B, Breast cancer
Author, year	Cancer subtype	P-value	Fold change	Gene rank (%)	Samples (n)	(Refs.)
Curtis et al, 2012	Invasive lobular breast carcinoma	<0.001	−4.857	1	148	(12)
TCGA Breast, Current study	Invasive lobular breast carcinoma	<0.001	−10.124	1	36	–
Curtis et al, 2012	Tubular breast carcinoma	<0.001	−5.474	1	67	(12)
Curtis et al, 2012	Medullary breast carcinoma	<0.001	−8.162	1	32	(12)
Curtis et al, 2012	Invasive ductal breast carcinoma	<0.001	−44.083	1	1,556	(12)
TCGA Breast, Current study	Invasive ductal breast carcinoma	<0.001	−25.347	1	389	–
Curtis et al, 2012	Mucinous breast carcinoma	<0.001	−5.711	1	46	(12)
C, Cervical cancer
Author, year	Cancer subtype	P-value	Fold change	Gene rank (%)	Samples (n)	(Refs.)
Biewenga et al, 2008	Cervical squamous cell carcinoma	<0.001	−2.102	5	40	(36)
Scotto et al, 2008	Cervical squamous cell carcinoma	<0.001	−2.123	3	32	(37)
D, Colorectal cancer
Author, year	Cancer subtype	P-value	Fold change	Gene rank (%)	Samples (n)	(Refs.)
Skrzypczak et al, 2010	Colorectal adenocarcinoma	<0.001	−3.48	4	45	(38)
Ki et al, 2007	Colon adenocarcinoma	<0.001	−2.413	1	50	(39)
E, Gastric cancer
Author, year	Cancer subtype	P-value	Fold change	Gene rank (%)	Samples (n)	(Refs.)
D'Errico et al, 2009	Gastric intestinal type adenocarcinoma	<0.001	−2.126	3	26	(40)
Cho et al, 2011	Diffuse gastric adenocarcinoma	<0.001	−4.626	4	31	(41)
F, Leukemia
Author, year	Cancer subtype	P-value	Fold change	Gene rank (%)	Samples (n)	(Refs.)
Haferlach et al, 2010	Pro-B acute lymphoblastic	<0.001	−4.65	3	70	(42)
Haferlach et al, 2010	Acute myeloid leukemia	<0.001	−2.442	4	542	(42)
Haferlach et al, 2010	B-cell acute lymphoblastic Leukemia	<0.001	−4.513	4	147	(42)
Haferlach et al, 2010	T-Cell acute lymphoblastic leukemia	<0.001	−4.366	4	174	(42)
Haferlach et al, 2010	B-cell childhood acute lymphoblastic Leukemia	<0.001	−4.187	5	359	(42)
G, Lung cancer
Author, year	Cancer subtype	P-value	Fold change	Gene rank (%)	Samples (n)	(Refs.)
Garber et al, 2001	Lung aenocarcinoma	<0.001	−9.404	2	40	(43)
Okayama et al, 2012	Lung aenocarcinoma	<0.001	−5.976	1	226	(44)
Selamat et al, 2012	Lung aenocarcinoma	<0.001	−6.617	1	58	(45)
Su et al, 2007	Lung aenocarcinoma	<0.001	−3.803	2	27	(46)
Landi et al, 2008	Lung aenocarcinoma	<0.001	−4.783	2	58	(47)
Hou et al, 2010	Lung aenocarcinoma	<0.001	−7.204	3	45	(48)
Garber et al, 2001	Squamous cell lung carcinoma	<0.001	−11.994	1	13	(43)
Wachi et al, 2005	Squamous cell lung carcinoma	<0.001	−5.833	1	5	(49)
Hou et al, 2010	Squamous cell lung carcinoma	<0.001	−8.845	2	27	(48)
Garber et al, 2001	Large cell lung carcinoma	<0.001	−9.981	1	4	(43)
Crabtree et al, 2009	Large cell lung carcinoma	<0.001	−10.929	3	19	(50)
H, Other cancer
Author, year	Cancer subtype	P-value	Fold change	Gene rank (%)	Samples (n)	(Refs.)
Yoshihara et al, 2009	Uterine Corpus Leiomyoma	<0.001	−2.418	1	50	(51)
I, Ovarian cancer
Author, year	Cancer subtype	P-value	Fold change	Gene rank (%)	Samples (n)	(Refs.)
TCGA Ovarian, Current study	Ovarian serous adenocarcinoma	<0.001	−38.254	1	37	–
Pei et al, 2009	Ovarian serous cystadenocarcinoma	<0.001	−3.519	4	586	(52)
J, Pancreatic cancer
Author, year	Cancer subtype	P-value	Fold change	Gene rank (%)	Samples (n)	(Refs.)
Barretina et al, 2010	Pancreatic carcinoma	<0.001	−2.616	1	36	(26)
K, Sarcoma
Author, year	Cancer subtype	P-value	Fold change	Gene rank (%)	Samples (n)	(Refs.)
Barretina et al, 2010	Pleomorphic myxofibrosarcoma	<0.001	−8.447	1	3	(26)
Barretina et al, 2010	Pleomorphic liposarcoma	<0.001	−4.486	1	23	(26)
Barretina et al, 2010	Dedifferentiated liposarcoma	<0.001	−4.804	1	46	(26)
Barretina et al, 2010	Myxofibrosarcoma	<0.001	−6.96	1	31	(26)
Barretina et al, 2010	Leiomyosarcoma	<0.001	−3.417	2	26	(26)
Barretina et al, 2010	Myxoid/round cell liposarcoma	<0.001	−2.673	4	20	(26)

SDPR gene expression alteration and survival of patients

The PrognoScan platform, which integrates published cancer microarray datasets with clinical annotation (19), was used in the systematic meta-analysis and to determine the prognostic value of SDPR in multiple datasets. Survival analysis consists of two steps, patient grouping and comparing the determined risks of these groups. Since gene expression is continuous data, the PrognoScan platform employed a minimum P-value approach for grouping patients in the survival analysis, which determines the optimal cut-off point in the continuous gene-expression measurement.

In the present study, PrognoScan indicated a significant association between microarray SDPR expression and cancer prognosis in several tests: Brain (1/8), breast (10/40), lung (6/19) and soft tissue (1/1). In all these 18 tests, low SDPR expression was associated with poor survival (data not shown), suggesting its protective function in cancer malignancy. Decreased SDPR expression was significantly associated with decreased overall survival (OS) and relapse-free survival (RFS) of patients with breast cancer (Fig. 2A and B), consistent with a previous study (1). In addition, SDPR downregulation was significantly associated with OS and RFS in patients with lung cancer adenocarcinoma (Fig. 2C and D), OS in patients with glioma (Fig. 2E) and distant-recurrence free survival (DRFS) in liposarcoma (Fig. 2F).

Figure 2. Survival curves comparing patients with high and low expressions of serum deprivation response were plotted from the PrognoScan database. (A) OS and (B) RFS survival curves comparing patients with high (red) and low (blue) expression in breast cancer were plotted from the PrognoScan database using the threshold of corrected P-value <0.05. (C) OS and (D) RFS survival curves comparing patients with high (red) and low (blue) expression in lung adenocarcinoma were plotted from the PrognoScan database using the threshold of corrected P-value <0.05. (E) OS survival curve comparing patients with high (red) and low (blue) expression in (E) glioma was plotted from the PrognoScan database as the threshold of corrected P-value <0.05. (F) DFRS survival curve comparing patients with high (red) and low (blue) expression in liposarcoma was plotted from the PrognoScan database as the threshold of corrected P-value <0.05. 95% confidence intervals for each group are indicated by dotted lines. OS, overall survival; RFS, relapse-free survival; DRFS, distant-recurrence free survival; HR, hazard ratio; Pcor, corrected P-value.

SDPR is hypermethylated in specific types of cancer

In a previous next-generation sequencing study, SDPR was not identified as a frequent mutational gene target (16) and it was also observed that SDPR was epigenetically silenced in breast cancer cell lines (1). Therefore, the present study hypothesized that DNA methylation alteration, which is an important epigenetic regulator for transcription, may be one of the mechanisms for SDPR gene expression differences. In order to reveal the underlying molecular mechanisms responsible for SDPR downregulation, MethHC was used to identify differential DNA methylation data between tumor and non-tumor tissues, which included 18 human types of cancer in >6,000 samples. MethHC is a database that systematically integrates DNA methylation and mRNA expression data from TCGA. The methylation profile of SDPR across 18 tumors is presented in Fig. 3. In this profile, significantly SDPR gene methylation differences were observed in most types of cancer (17/18). Consistent with previously published experimental data, SDPR was significantly hypermethylated in breast invasive carcinoma compared with normal breast tissues (1). In addition, SDPR was observed to be significantly hypermethylated in bladder urothelial carcinoma, cervical squamous cell carcinoma, head and neck squamous cell carcinoma, LUAD, lung squamous cell carcinoma (LUSC), pancreatic adenocarcinoma, prostate adenocarcinoma, SARC, skin cutaneous melanoma, stomach adenocarcinoma and uterine corpus endometrial carcinoma (P<0.005). Statistical differences were also found in colon adenocarcinoma, rectum adenocarcinoma and thyroid carcinoma (P<0.05). These results suggested that DNA methylation may be responsible for SDPR downregulation in these types of cancer. As shown in Fig. 1, SDPR was downregulated in kidney cancer. However, significant SDPR hypomethylation was found in both kidney renal clear cell carcinoma (KIRC) and kidney renal papillary cell carcinoma (KIRP) (P<0.005), suggesting the existence of other regulatory pathways.

Figure 3. Methylation profiles of SDPR across tumors. The differential methylation statuses of each transcript of SDPR in different tumor types are presented as boxplots. Tumor samples are in dark grey and are compared with normal samples in light grey. The shapes of notched boxplots indicated the characteristics of the data in each sub-dataset. P-values were adjusted using the Benjamini-Hochberg procedure. *P<0.05, **P<0.0005 vs. normal sample. SDPR, serum deprivation response; BLCA, bladder urothelial carcinoma; BRCA, breast invasive carcinoma; CESC, cervical squamous cell carcinoma; COAD, colon adenocarcinoma; HNSC, head and neck squamous cell carcinoma; KIRC, kidney renal clear cell carcinoma; KIRP, kidney renal papillary cell carcinoma; LIHC, liver hepatocellular carcinoma; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; PAAD, pancreatic adenocarcinoma; PRAD, prostate adenocarcinoma; READ, rectum adenocarcinoma; SARC, sarcoma; SKCM, skin cutaneous adenocarcinoma; STAD, stomach adenocarcinoma; THCA, thyroid carcinoma; UCEC, uterine corpus endometrial carcinoma.

Gene methylation of SDPR is associated with patient survival in LUAD and SARC

As downregulation of SDPR gene expression was observed in lung cancer and SARC, SDPR was indicated to be hypermethylated compared with normal tissues, and downregulation of SDPR expression was significantly associated with poor patient survival in LUAD and SARC, MethSurv was used to identify whether hypermethylation of SDPR was associated with patient survival in LUAD or SARC. In addition, since the differential methylation levels at the CpG island are tissue-specific, the DNA methylation patterns were analyzed, taking into consideration the differential methylation levels in different gene subregions (Fig. 4A). MethSurv was the first database that indicated an association with overall survival and DNA methylation patterns, in which the methylation levels are from TCGA methylation profile using the HM450k array. Based on the UCSC database, the CpG sites were grouped into gene subregions: ‘TSS200’, ‘TSS1500’, ‘1st exon’, ‘5′UTR’, ‘body’ and ‘3′UTR’ (23). As shown in Fig. 4B and C, clustering visualization in the form of heat maps evaluated the association of methylation levels with the available patient characteristics and gene subregions in LUAD and SARC. According to the heatmaps, methylation sites cg18843739, cg10082589 and cg17809945 in the ‘TSS1500’ subregion showed differential methylation levels in patients with LUAD (461 samples). Similar to patients with SARC, the differently methylated sites were cg07488576 in the ‘1st Exon’ subregion and cg18843739 in the ‘TSS1500’ subregion (261 samples).

Figure 4. Gene methylation of SDPR is associated with patient survival in LUAD and SARC. (A) Diagram showing the relative location from the TSS for each CpG site of SDPR in Methsurv. Heat map depicting clustering analysis of the CpG methylation levels within the SDPR gene in (B) LUAD and (C) SARC. Methylation levels (0 represents fully unmethylated and 1 represents fully methylated) are shown as a continuous variable from blue to red color. Kaplan-Meier plot showed survival curve in higher (red) and lower (blue) methylation groups in methylation sites (D) cg17809945 and (E) cg10082859 in LUAD, and (F) cg07488576 and (G) cg18843739 in SARC. Violin plots showing the methylated levels of (H) cg17809945 and (I) cg10082859 among stage I, II, III and IV LUAD samples. The interquartile range and median methylation levels are shown in each violin plot as boxplots. SDPR, serum deprivation response; LUAD, lung adenocarcinoma; SARC, sarcoma; TSS, transcription start site; TSS1500, covering-200 to-1500 nucleotides upstream of TSS; TSS200, TSS to-200 nucleotides upstream of TSS; UTR, untranslated region; HR, hazard ratio; LR, log-likelihood ratio.

MethSurv was used to identify which methylation sites were significantly associated with patient survival. By analyzing all 11 methylation sites in LUAD hypermethylation, it was indicated that two sites in LUAD (cg17809945 and cg10082589, located in the ‘TSS1500’ subregion of SDPR) were significantly associated with poor overall survival (Fig. 4D and E). In SARC, 2 out of 13 methylation sites (cg07488576 in the ‘1st Exon’ subregion and cg18843739 in the ‘TSS1500’ subregion) were identified to be associated with poor survival (Fig. 4F and G). The results were consistent with the heatmap presented in Fig. 4B and C. In this analysis, CpG (cg10082589; HR=1.887; 95% CI: 1.371–2.596; LR test P-value=0.001) was identified as the optimal survival methylation marker for LUAD (Fig. 4E) and CpG (cg07488576; HR=2.406; 95% CI: 1.608–3.599; LR test P-value=0.00001) was identified as the optimal survival methylation marker for SARC (Fig. 4F). In addition, when LUAD samples were grouped according to clinical stage, the methylation levels of cg17809945 (Fig. 4H) and cg10082589 (Fig. 4I) were indicated to be higher in late stages, suggesting that they may be involved in LUAD progression. However, no association between methylation levels of CpG sites and clinical stages was observed in SARC (data not shown).

Prediction of transcription factors regulating SDPR expression in LUAD

Since epigenetic patterns may not be the only reason for gene expression alternations in cancer, transcriptional regulation by deregulated transcription factors was investigated. The UCSC database was used to identify the promoter sequences of SDPR and PROMO was subsequently used at the ALGGEN server showing potential transcriptional factors on the promoter (Fig. 5A). Meanwhile, genes that were co-expressed with SDPR were analyzed using the Oncomine database, which were subsequently grouped into normal lung tissue and LUAD (Fig. 5B). By comparing the potential transcriptional factors with genes significantly co-expressed with SDPR (node correlation value >0.5), GATA binding protein 2 (GATA2; node correlation value=0.807) was identified as a potential transcription factor. GATA2 is a member of the GATA family, serves as a transcriptional activator during development and carcinogenesis, and was indicated to be epigenetically repressed in both human and mouse lung tumors (27). This result suggested that GATA2 may be a potential transcription factor regulating SDPR gene expression in LUAD.

Figure 5. Prediction of TFs regulating SDPR expression. (A) TFs predicted using PROMO in the SDPR promoter sequence. The numbers in the square brackets represent the annotation for each TF in the PROMO database. (B) SDPR is coexpressed with the indicated genes across a panel of 58 lung adenocarcinoma and 49 normal samples. Normalized gene expressing levels are shown from least expressed (blue) to most expressed (red) within each row. The asterisks indicate the GATA2 results. SDPR, serum deprivation response; TF, transcription factor; GATA2, GATA binding protein 2.

Discussion

SDPR has been previously reported to show characteristic gene signatures in specific types of cancer, including breast (1,2,11–13), thyroid (3,28), oral (5) and kidney (11,14) cancer, and SARC (4,26). In breast cancer, SDPR was identified as a novel tumor suppressor, which was significantly associated with patient survival (1,2). However, a systemic profile of SDPR alterations or analysis of its relevance in clinical outcomes in different types of cancer has yet to be performed, to the best of our knowledge. In the present study, SDPR downregulation was observed in bladder, breast, lung, kidney, cervical, colorectal, gastric, ovarian and pancreatic cancer. Consistent with previous studies (1,2), SDPR downregulation was significantly associated with patient survival in breast cancer. Furthermore, the analysis also indicated that decreased expression of SDPR was significantly associated with poor OS and RFS in patients with adenocarcinoma of lung cancer, OS in patients with glioma and DRFS in liposarcoma.

In breast cancer, SDPR, which is partially silenced by DNA methylation, has been elucidated to execute an anti-metastatic function by promoting apoptosis (1), and depletion of SDPR-induced EMT through TGF-β signaling activation, according to previously published experimental studies (1,2). To clarify why SDPR expression was altered in other types of cancer, the gene methylation level of SDPR was subsequently profiled in 18 types of cancer and was compared with normal tissues. It was indicated to be significantly altered in 17 out of 18 types of cancer.

If DNA methylation of SDPR is involved in its downregulation in LUAD or SARC, it might be associated with patient survival. As not all of the methylation sites are responsible for gene expression, the most significant methylation sites require further investigation. Therefore, the present study analyzed the DNA methylation patterns in LUAD and SARC, considering the differential methylation levels in different gene subregions of SDPR. To the best of our knowledge, the present study for the first time, indicated that CpG (cg10082589) serves as the optimal survival methylation marker for LUAD, and CpG (cg07488576) serves as the optimal survival methylation marker for SARC. Furthermore, when LUAD samples were grouped according to clinical stage, the methylation level of CpG (cg10082589) and CpG (cg1780995) were higher in late stages, suggesting that they may be involved in LUAD progression.

In addition, by analyzing both the co-expressed genes with SDPR and the putative transcription factor binding sites on SDPR, GATA2 was identified as a potential transcription factor for SDPR transcription in LUAD. Transcriptional factor GATA2 regulates genes critical for embryonic development, self-renewal maintenance (29), functionality of blood-forming (30) and lymphatic vessel valve development (31). GATA2 has been reported to be frequently epigenetically repressed in both human and mouse lung tumors, and aberrant GATA2 methylation occurred early during lung carcinogenesis (27). GATA2 may serve a role in the downregulation of SDPR in LUAD.

Analysis of the associations between SDPR expression and DNA methylation with patient survival in LUSC was additionally performed. However, neither the OS nor disease-free survival of patients had been observed to be significantly associated with SDPR expression alternations, according to the Prognoscan database. Furthermore, similar to LUAD, the same 13 CpG sites grouped in gene subregions based on the UCSC database were analyzed using MethSurv. However, the results indicated no significant association between DNA methylation data and patient survival (data not shown). Since gene expression patterns differ in the subtypes of lung cancer, this may provide novel insight for the examination of potential mechanisms of SDPR function in LUAD.

SDPR was also downregulated in kidney cancer. Significant hypomethylation was found in both KIRC and KIRP, suggesting other regulatory pathways. By comparing methylation patterns with patient survival in KIRC and KIRP, the present study indicated that hypomethylation in the ‘TSS200’ and ‘TSS1500’ subregions of SDPR was significantly associated with longer survival time, while hypomethylation in the gene ‘body’ and the ‘3′UTR’ subregions was associated with poor OS (data not shown). The function of DNA methylation status seems to vary in context; this may be due to hypermethylation in the promoter region inducing the downregulation of gene expression, while hypermethylation in the gene body may not block and may even stimulate transcription elongation, and the gene body methylation may have an impact on splicing (32). In addition, long non-coding RNA (lncRNA) SDPR-antisense (SDPR-AS) has been verified to be co-expressed with SDPR, and elevated lncRNA SDPR-AS increases the OS in renal cell carcinoma, suggesting the possibility that lncRNAs may serve a regulatory role in the SDPR pathway (33).

The specific methylation CpG sites, which are significantly associated with patient survival, require further study in order to verify the present results, such as pyrosequencing and reverse transcription-quantitative PCR. The potential transcription factor GATA2 should also be experimentally investigated to determine whether it is responsible in SDPR transactivation. Despite taking age and sex into consideration as covariates in survival analysis based on the Methsurv database, due to the lack of available clinical data, the survival analysis based on PrognoScan is univariable. Since several factors, such as co-morbidities, performance status and treatments, may potentially affect the prognosis of patients with cancer, it is important to consider all potential relevant specific features for specific cancer types in future clinical investigations of SDPR.

In summary, the present study suggested that the role of SDPR as a tumor suppressor may have broader clinical relevance beyond breast cancer. The present study on SDPR may help to examine its underlying molecular mechanism in cancer progression, reveal novel perspectives for prognostic evaluation in specific cancer, and provide insight for further research in the field.

Acknowledgements

Not applicable.

Funding

The present study was supported by Foundation of Sichuan Educational Committee, People's Republic of China (grant nos. 17ZA0164 and 18CZ0010) and Foundation of Chengdu University of Traditional Chinese Medicine, China (grant nos. ZRQN1660 and CGZH1709).

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Authors' contributions

YW and YL designed the present study. YW analyzed the gene expressing data, patient survival data and the potential transcriptional factors of SDPR. ZS analyzed the co-expressing genes with SDPR. YL and PL analyzed the gene methylation data. YW wrote the manuscript and YL revised it. All authors read and approved the final manuscript.

Ethics approval and consent to participate

Not applicable.

Patient consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

References