Sage Journals: Discover world-class research

Abstract

Cervical cancer is the fourth most common malignancy among women worldwide, and continued research to discover biomarkers or therapeutic targets will aid early diagnosis and treatment of this cancer. Here, we investigated novel cervical cancer biomarkers using integrated analysis of high-throughput sequencing data from the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) databases. We have identified nine genes of interest that appear to be involved in cervical cancer development: SPARCL1, SYCP2, KIF4A, PRC1, TOP2A, LAMP3, KIF20A, MCM2, and APOBEC3B. Furthermore, gene ontology (GO) and co-expression analysis of these differentially expressed genes indicated that SPARCL1 may play a core role in cervical cancer development. Further, we analyzed the expression of these nine genes during the progression of cervical cancer, and found that SPARCL1 is also related to precancerous lesions and migration processes during cervical cancer pathogenesis. Finally, we validated these observations by investigating SPARCL1 expression in cervical cancer tissue and serum samples. The diagnostic specificity of serum SPARCL1 in cervical cancer occurrence was also compared with other high incidence diseases. All of these data indicate that SPARCL1 may be a novel cancer predictive marker and a potential therapeutic target for tumor development and progression in cervical cancer.

Keywords

Cervical cancer SPARCL1 marker

1. Introduction

Cervical cancer is the second largest type of malignant tumor in women, second only to breast cancer, with more than 1.5 million new cases diagnosed each year and over 300,000 deaths [1, 2]. It has also been associated with persistent infections with oncogenic human papilloma viruses (HPVs). However, while HPV infection is an indispensable factor, it alone is not sufficient to cause cancer. In fact, the majority of acute HPV infections induce low grade precursor lesions that are cleared spontaneously after several months in more than 90% of cases, with less than 10% of cases eventually progressing to high grade lesions or invasive cancer [3]. Cervical intraepithelial neoplasia is the typical type of precancerous lesion leading to cervical cancer, and in the evolution from precancerous lesion to malignant tumor, the expression of a variety of oncogenes and invasion-related genes is significantly altered [4]. At present, the key molecules causing cervical intraepithelial neoplasia to develop into cervical cancer have not been elucidated [5]. Thus, the discovery of novel biomarkers that will allow these molecular events to be monitored in histological or cytological specimens is essential and will likely help improve the detection of high risk lesions in both primary screening and triage settings [6].

The Gene Expression Omnibus (GEO) repository at the National Center for Biotechnology Information (NCBI) archives and freely disseminates microarray and other forms of high-throughput data generated by the scientific community. The database has a minimum information about a microarray experiment (MIAME)-compliant infrastructure that includes fully annotated raw and processed data, which includes expression data for a variety of tumors [7, 8]. The Cancer Genome Atlas (TCGA) project has analyzed mRNA expression, microRNA expression, promoter methylation, and DNA copy number in 306 cervical squamous cell carcinomas and 3 normal cervical tissue samples [9, 10]. Thus, combining the GEO and TCGA datasets may provide important insight into novel biomarkers of these tumors. Indeed, screening of tumor biomarkers based on the GEO and TCGA data has recently revealed a series of highly specific and sensitive markers [11, 12]. Compared with conventional screening methods, analyzing high-throughput data based on bioinformatics methods allows researchers to obtain stable and reliable biomarkers in a larger number of clinical samples.

Meta-analysis is a statistical analysis method that combines the results of multiple scientific studies using a statistical approach in an effort to increase power, improve estimates of the size of the effect, and/or to resolve uncertainty when reports disagree [13, 14]. With regard to screening biomarkers, combining multiple GEO datasets can help to validate the expression of multiple genes of interest between normal and tumor tissues, analyze the specificity and sensitivity of a biomarkers, and even correlate the expression of genes of interest with patients prognosis [15, 16].

Here, we analyzed three independent datasets from the GEO database and TCGA-CSCC, which contains normal cervical and cervical cancer samples. In doing so, we identified nine-gene signatures that were associated with cervical cancer occurrence: SPARCL1, SYCP2, KIF4A, PRC1, TOP2A, LAMP3, KIF20A, MCM2, and APOBEC3B. Differences in gene expression were then evaluated using meta-analysis, which further proved that these nine genes are indeed related to cervical cancer pathogenesis. Diagnostic meta-analysis also indicated that the nine gene signatures have prominent specificity and sensitivity in cervical cancer diagnosis. GO analysis and co-expression analysis showed that SPARCL1 may be a core factor in the development of cervical cancer. Subsequently, we analyzed the expression of those nine-gene signatures in the progression of cervical cancer and found that SPARCL1 is also related to precancerous lesions and the migration process of cervical cancer. Finally, we validated the correlation of SPARCL1 in cervical cancer development and progression in cervical cancer tissue and serum samples. These data indicate that SPARCL1 may be a novel predictive marker and therapy target for tumor occurrence and progression in cervical cancer.

2. Materials and methods

2.1 Materials

Cervical cancer tissue microarrays (CR602a and CR806) were purchased from Biomax US (MD, USA). The anti-SPARCL1 polyclonal antibodies were obtained from ProteinTech Group, Inc. (Wuhan, China). The human SPARCL1 ELISA Kit (EME-H8276) was purchased from KITZOME (Shenzhen, China). All other kits or reagents were purchased from the Beyotime Institute of Biotechnology (Shanghai, China).

2.2 Differential gene screening

To identify relevant studies, we obtained sequencing data from TCGA database. We also retrieved all relevant studies from the GEO database. The dataset search was limited to those in the English language and that were completed in or before October 2016. To increase the sensitivity of the search, both mesh-terms and free words were used. The search terms included: “cervical cancer”; “normal”; and “tissue” or “tumor” or “cancer” or “carcinoma” or “neoplasm”.

Three GEO datasets (GSE6791, GSE7803, GSE390 01) which contain normal cervical tissue as well as cervical cancer tissue were downloaded. The “Series Matrix Files” and their corresponding platform files were integrated by Perl language to create a new generated file containing gene symbols and their corresponding expression in each sample. We then analyzed differential gene expression in the normal cervical group and cervical cancer group by R language based on the limma library. These data were used to generate the corresponding heatmap, volcano figure, and table of differentially expressed genes.

Files from TCGA-CSCC project were downloaded from TCGA database. The download terms included “Transcriptome Profiling” and “Gene Expression Qua- ntification” for the Data Category and “RNA-Seq” and “HTSeq-Counts” for the Experimental Strategy and Workflow Type, respectively. The analysis included 306 cervical squamous cell carcinoma and 3 normal cervical tissue samples. Similar to our analysis above, the files were integrated with Perl language to create a new generated file containing the gene symbols and their corresponding expression in each sample. We then analyzed the differences in gene expression in the normal and cervical cancer groups by R language based on the edgeR library. These data were then used to generate the corresponding heatmap, volcano figure, and differential gene expression table. The genes identified in both our GEO and TCGA database analyses are presented with a Venn-diagram.

2.3 Data extraction and meta-analysis

Data extraction was repeated independently by three investigators, and a consensus was reached for each dataset. For the seven datasets available in the GEO database that included the expression and related patient data and sample characteristics (e.g., normal or cervical cancer tissue, TNM stage) were extracted for each of the nine genes of interest. We then calculated the expression of the nine genes in normal or cervical cancer tissue using GraphPad Prism 5. We analyzed the obtained data for the Mean, SD, and Total using Review Manager 5.3. An I ${}^{2}$ greater than 50% was defined as significant heterogeneity. If significant heterogeneity existed between studies, we used a random effects model or used the fixed effects model. All graphical representations were generated as forest plots.

We also undertook a bivariate diagnostic meta-analysis to obtain pooled estimates of sensitivity and specificity, positive and negative likelihood ratios, and a summary diagnostic odds ratio. Summary Receiver Operator Characteristic (sROC) curves were then constructed using the bivariate model to produce a 95% confidence ellipse within the ROC space. Each data point in the summary ROC space represents a separate study.

2.4 GO enrichment analysis

The nine differentially expressed genes were also analyzed for their GO term enrichment using the Gene Ontology Enrichment Analysis Software Toolkit (GOEAST). The results of this GO analysis are grouped by Biological Process, Cellular Component, and Molecular Function.

2.5 Co-expression network using TCGA data

The nine genes of interests (SPARCL1, SYCP2, KIF4A, PRC1, TOP2A, LAMP3, KIF20A, MCM2, and APOBEC3B) discovered in our TCGA screening were analyzed for their co-expression and network connections. Briefly, all of the gene expression values were merged into one file. The weight of their relative coefficients were calculated and extracted by R language based on the WGCNA library using a threshold value of 0.80. The co-expression network was visualized by Cytoscape (version 3.4.0).

2.6 Immunohistochemical staining and assessment of SPARCL1 expression

Tissues microarrays were dewaxed and antigens were retrieved using high pressure for 3 min. Then, the activity of endogenous peroxidases was blocked using 3% hydrogen peroxide for 10 min at room temperature. After immersion in normal goat serum for 30 min at 37 ${{}^{\circ}}$ C, tissues were subjected to primary antibody incubation at 4 ${{}^{\circ}}$ C overnight. Subsequently, sections were washed with PBS for 15 min at room temperature and incubated with secondary antibody that had been labeled with biotin for 30 min at 37 ${{}^{\circ}}$ C. After washing with PBS for 15 min at room temperature, the sections were incubated with horseradish peroxidase (HRP) complex for 30 min at 37 ${{}^{\circ}}$ C and visualized using diaminobenzidine (DAB). All immunohistochemical images were obtained using an Olympus BX51 microscope equipped with either a 20 $\times$ , 40 $\times$ , or 100 $\times$ objective lens (Olympus Corporation, Japan) and a DP 50 camera (Olympus Corporation, Japan). Images were processed using DPC controller software (Olympus Corporation, Japan).

The immunohistochemical staining was evaluated using a semi-quantitative scoring method. The scores were determined independently by two pathologists. SPARCL1 staining was scored according to staining intensity using the following grading system: no staining $=$ 0, light positive staining $=$ 1, medium positive staining $=$ 2, and strong positive staining $=$ 3. The positively stained area was scored as follows: less than 5% $=$ 0, 5–25% $=$ 1, 26–50% $=$ 2, 51–75% $=$ 3, and greater than 75% $=$ 4. A final score was calculated by multiplying the intensity score by the expression score and the rating (from 0 to 12) was determined for each case as previously described [17].

Figure 1.

Integrated analysis reveals nine genes of interest with altered expression during cervical cancer development. A. Venn-diagram presenting 72 distinct genes with a fold change greater than 2 between the four datasets investigated. B. Venn-diagram presenting 72 distinct genes with a fold change greater than 10 between the four datasets investigated. C–F. The expression of the nine differentially expressed genes displayed in clustering heatmaps. G–I. Top ten GO terms of the nine genes of interest based on their biological process (G), cellular component (H), and molecular function (I).

2.7 ELISA

The serum samples collected and experiments conducted in this study were approved by the Chengdu Medical College ethics committee (No: CYYFYEC20 16002) in accordance with approved guidelines. Blood was collected prior to the first surgery and before patients receiving any kind of treatment. The collected serum samples were aliquoted into smaller volumes and stored at $-$ 80 ${{}^{\circ}}$ C until further use. Prior to the assay, the serum samples were thawed on ice.

SPARCL1 levels in the serum were measured and quantitated using an ELISA kit (KITZOME, Shenzhen, China) according to the manufacturer’s instructions. Each serum sample was analyzed three times.

2.8 Statistical analysis

Experiments were performed independently at least three times. One-way analysis of variance and two-tailed t tests were used to compare the effects of various treatments with GraphPad Prism 5 (San Diego, CA). Data are shown as mean $\pm$ SEM. A $p$ -value less than 0.05 was considered statistically significant.

Table 1
Study information on eight GEO datasets which consist of normal cervical and cervical cancer samples

Datasets No.	Contributor	Citation	Country	Platform	Analysis	No. of	No. of
		PMID No.			method	normal	cancer
GSE6791	Pyeon et al. [44]	17510386	USA	GPL570	Array	8	20
GSE78035	Zhai et al. [45]	17974957	USA	GPL96	Array	10	24
GSE39001	Espinosa et al. [46]	23405241	Mexico	GPL201	Array	12	43
GSE46857	Thomas et al. [47]	24403257	India	GPL7025	Array	4	25
GSE9570	Scotto et al. [48]	18506748	USA	GPL96	Array	24	33
GSE63514	Den Boon et al. [49]	26056290	USA	GPL570	Array	24	28
GSE26342	Mine et al. [50]	23651994	USA	GPL1052	Array	20	34
Total						102	207

Figure 2.

Meta-analysis of SPARCL1 expression in normal cervical and cervical cancer tissue. A. Meta-analysis of SPARCL1 expression in normal cervical and cervical cancer samples. B. Sensitivity and specificity of SPARCL1 in cervical cancer diagnosis. C. ROC curve of SPARCL1 as a diagnostic marker of cervical cancer development. D. The expression of SPARCL1 increases ( $P<$ 0.0001) with the degree of precancerous lesions. E. The expression of SPARCL1 changes ( $P<$ 0.0001) depending on the level of cervical cancer metastasis.

Table 2

TOP 10 GO terms in each category (Biological process, Cellular component, and Molecular function) associated with the nine genes of interest analyzed by GOEAST

GOID	Term	Level	Number in our	Number in the	Log	$p$ -value
			list (q)	genome (m)	odd-ratio
Biological process
GO: 0009987	Cellular process	1	35	22558	0.811	5E-05
GO: 0044699	Single-organism process	1	30	18620	0.866	7E-04
GO: 0044763	Single-organism cellular process	2	29	16172	1.02	1E-04
GO: 0071840	Cellular component organization or biogenesis	1	19	6264	1.778	3E-05
GO: 0016043	Cellular component organization	2	19	6118	1.812	2E-05
GO: 0006996	Organelle organization	2	19	3352	2.681	3E-09
GO: 0007049	Cell cycle	2	18	1636	3.637	3E-13
GO: 1902589	Single-organism organelle organization	3	11	1519	3.034	7E-06
GO: 0006259	DNA metabolic process	8	9	1339	2.926	1E-04
GO: 0022402	Cell cycle process	3	18	1278	3.994	8E-15
Cellular component
GO: 0043227	Membrane-bounded organelle	1	25	16287	0.796	0.016
GO: 0043231	Intracellular membrane-bounded organelle	4	25	16131	0.81	0.014
GO: 0044422	Organelle part	2	21	10455	1.184	0.002
GO: 0044446	Intracellular organelle part	4	21	10214	1.217	0.001
GO: 0043228	Non-membrane-bounded organelle	1	17	6898	1.479	0.001
GO: 0043232	Intracellular non-membrane-bounded organelle	4	17	6898	1.479	0.001
GO: 0031974	Membrane-enclosed lumen	1	13	4793	1.617	0.005
GO: 0043233	Organelle lumen	3	13	4734	1.635	0.005
GO: 0044428	Nuclear part	5	13	4734	1.635	0.005
GO: 0070013	Intracellular organelle lumen	4	13	4680	1.652	0.004
Molecular function
GO: 0005488	Binding	1	35	25220	0.65	8E-04
GO: 0036094	Small molecule binding	1	14	6330	1.323	0.019
GO: 1901265	Nucleoside phosphate binding	2	14	6041	1.39	0.013
GO: 0000166	Nucleotide binding	2	14	6040	1.39	0.013
GO: 0016787	Hydrolase activity	1	16	5848	1.63	8E-04
GO: 0097367	Carbohydrate derivative binding	1	14	5172	1.614	0.003
GO: 0032553	Ribonucleotide binding	3	14	4757	1.735	0.001
GO: 0017076	Purine nucleotide binding	2	14	4730	1.743	0.001
GO: 0032555	Purine ribonucleotide binding	3	14	4713	1.748	0.001
GO: 0001882	Nucleoside binding	4	14	4704	1.751	0.001

Table 3

Clinicopathological association of SPARCL1 with cervical cancer

Factors	Number	Average score	$P$ -value
Age			0.1139
$<$ 60	92	5.641 $\pm$ 0.2894
$\geqslant$ 60	10	4.200 $\pm$ 0.6289
Primary tumor (T)			$<$ 0.0001
T1	64	6.034 $\pm$ 0.1679
T2–T4	38	4.584 $\pm$ 0.4001
Regional lymph nodes (N)			0.0129
N0–N1	80	5.888 $\pm$ 0.1933
N2–N3	22	4.242 $\pm$ 0.4796
Distant metastasis (M)			0.0443
M0	97	5.578 $\pm$ 0.1829
M1	5	3.600 $\pm$ 0.7483
Clinical stage			0.0332
I–II	72	6.056 $\pm$ 0.1993
III–IV	30	4.600 $\pm$ 0.6905
Grade			0.0379
1	21	6.162 $\pm$ 0.3645
2	50	5.440 $\pm$ 0.1899
3	25	4.520 $\pm$ 0.4800

Clinical and TNM staging were referenced to American Joint Committee on Cancer: AJCC Cancer Staging Manual, 6th edition.

3. Results

3.1 Integrated analysis reveals nine differentially expressed gene signatures that are associated with cervical cancer development

By integrating the four unique data sets (three GEO data sets (GSE6791, GSE7803, GSE39001) and TCGA-CSCC data), which contain normal cervical tissue and cervical cancer tissue, we created a Venn-diagram presenting 72 of the same distinct genes that changed greater than 2-fold. This analysis highlighted nine genes whose expression was altered more than 10-fold (Fig. 1A and B). In Fig. 1C–F, the expression of these nine genes in the four datasets is displayed in clustering heatmaps. Furthermore, our GO enrichment analysis of the nine gene signatures indicates a range of functions as highlighted by the list of top ten GO terms for their biological process, cellular component, and molecular function (Fig. 1G and I and Table 1). The full list of GO terms for these nine genes of interest can be found in Supplemental Figs 1–3.

3.2 Meta-analysis of expression level and diagnostic value of the nine genes of interest during cervical cancer development

While searching the GEO database, we excluded invalid datasets by reading their summary and sample characteristics (Supplemental Fig. 4). We included seven GEO datasets in this study (Table 2). Meta-analysis of their mean expression changes indicates that of the nine gene differentially expressed in cervical cancer samples compared to normal tissue (SPARCL1, SYCP2, KIF4A, PRC1, TOP2A, LAMP3, KIF20A, MCM2, and APOBEC3B), SPARCL1 appears to be more highly expressed in normal cervical tissue than in cervical cancer tissue (Fig. 2A and Supplemental Fig. 5). Additional diagnostic meta-analysis revealed that these nine gene signatures all have excellent diagnostic potential, are highly sensitive, and are specific for cervical cancer diagnosis (Fig. 2B and C and Supplemental Figs 6 and 7).

3.3 Progression-related expression and co-expression networks reveal SPARCL1 as an important factor in both cervical cancer development and progression

We analyzed the expression of the nine genes of interest in both precancerous lesions and later stage migration processes. In doing so, we found that the expression of SYCP2, KIF4A, PRC1, TOP2A, KIF20A, MCM2, and APOBEC3B was significantly increased ( $P<$ 0.0001 for all) and depended on the degree of precancerous lesions (Fig. 2D and Supplemental Fig. 8). Moreover, expression of SPARCL1, PRC1, TOP2A, and KIF20A changed significantly ( $P<$ 0.0001 for all) depending on the level of cervical cancer metastasis (Fig. 2E and Supplemental Fig. 9).

Figure 3.

Clinical significance of the expression of SPARCL1 in cervical cancer tissue samples and patient serum. A. Expression of SPARCL1 is related with clinical stage ( $P=$ 0.0332). B. Expression of SPARCL1 is related with tumor grade ( $P=$ 0.0379). C. SPARCL1 expression significantly decreased as cervical cancer progressed. D. Serum SPARCL1 expression was downregulated in cervical cancer compared to normal controls.

Figure 4.

Diagnostic specificity of serum SPARCL1 expression in cervical cancer development. A. SPARCL1 is downregulated in cervical cancer, lung cancer, colon cancer, liver cancer, ovarian cancer, and breast cancer, while no changes were observed in prostate cancer, stomach cancer, or esophageal cancer. B. Expression of SPARCL1 is not associated with senile disease. C. Expression of SPARCL1 is not correlated with other chronic diseases.

A co-expression network of nine gene signatures also highlighted SPARCL1 as an important factor as it interacts with multiple other differentially expressed genes(Supplemental Fig. 10).

3.4 Clinical significance of the expression of SPARCL1 in cervical cancer tissue samples and patient serum

To determine the clinical significance of SPARCL1 expression in patients with cervical cancer, we performed immunohistochemical (IHC) labeling of specimens from a cohort of 102 patients with cervical squamous cell carcinoma, 10 samples of corresponding adjacent cervix tissue, 10 samples of chronic cervicitis, 10 samples of cervical intraepithelial neoplasia, and 8 samples of lymph node metastatic squamous cell carcinoma from the uterine cervix. SPARCL1 expression levels were measured with a semi-quantitative scoring method. Overall, SPARCL1 expression was observed in all patients and was significantly associated with clinical T classification ( $P<$ 0.0001), N classification ( $P=$ 0.0129), M classification ( $P=$ 0.0443), clinical stage ( $P=$ 0.0332), and tumor grade( $P=$ 0.0379) (Fig. 3A and B). No correlation was observed between SPARCL1 expression and age ( $P>$ 0.05) (Table 3).

During cervical cancer progression, SPARCL1 expr- ession was observed to significantly decrease (Fig. 3C). We further analyzed SPARCL1 expression in normal and cervical cancer patient serum. We found that serum SPARCL1 was downregulated in cervical cancer compared to normal tissue (Fig. 3D). Taken together, these data corroborate our conclusions based on our meta-analysis.

3.5 Diagnostic specificity of serum SPARCL1 in cervical cancer development

We tested serum SPARCL1 in healthy, high risk patients previously diagnosed with tumors (cervical cancer, lung cancer, colon cancer, liver cancer, prostate cancer, stomach cancer, ovarian cancer, esophageal cancer, and/or breast cancer), senile disease (diabetes, hypertension, atherosclerosis), and other chronic diseases (chronic hepatitis B, chronic obstructive pulmonary, chronic bronchitis, chronic renal failure). We found that SPARCL1 is downregulated in patients with cervical cancer, lung cancer, colon cancer, liver cancer, ovarian cancer, and breast cancer, while showing little to no change in prostate cancer, stomach cancer, esophageal cancer (Fig. 4A), senile disease (Fig. 4B), or chronic diseases (Fig. 4C).

4. Discussion

In recent years, with the implementation of the human genome project, widespread using of sequencing and gene chips has increased. This increase has also been further supplemented by the accessibility of multiple open databases, such as TCGA, GEO and Oncomine, which store vast amounts of gene expression data for both normal and diseased tissues [18, 19]. Integrating multiple studies from these databases allows researchers to investigate novel tumor biomarkers that have important research value [20, 21]. Based on the high morbidity and mortality rates associated with cervical cancer in gynecological tumors, we used an integrated analysis to evaluate novel biomarkers in cervical cancer in multiple TCGA and GEO data sets. Gene screening demonstrated that nine genes of interest were differentially expressed (SPARCL1, SYCP2, KIF4A, PRC1, TOP2A, LAMP3, KIF20A, MCM2, and APOBEC3B) and are likely related to cervical cancer development and progression.

Meta-analysis as a systemic method for evaluating multiple studies [14]. In the present study, the nine differentially expressed genes were further analyzed in terms of their specificity and sensitivity as cervical cancer biomarkers using both differential and diagnostic meta-analysis. In doing so, we found that the nine gene signatures were significantly different in normal and cervical cancer tissues. Thus, the expression of the nine genes can indeed be used as diagnostic biomarkers of cervical cancer.

Of the nine genes of interest identified in this study, SYCP2 has been previously related to HPV-associated cancer, screening, and development of cervical cancer [22, 23]. The expression and prognostic significance of KIF4A [24, 25], TOP2A [26, 27], LAMP3 [28, 29], KIF20A [30, 31], and MCM2 [32, 33] have also been reported in a wide variety of tumor types, including a Phase I/II clinical trial of a peptide vaccine derived from KIF20A in patients with advanced pancreatic cancer [31]. Moreover, APOBEC3B appears to be related with HPV, hepatitis virus B (HBV), and associated cancers [34, 35]. Thus, it seems likely that these genes are related to cervical cancer and that our biomarker screening methods are valid, resulting in a list of valuable differentially expressed genes that have application potential.

Progression-related expression and co-expression networks also revealed that SPARCL1 may play an important role in both cervical cancer occurrence and progression. Our analysis revealed that this particular gene may be involved in the expression changes observed in the other eight genes, making it a particularly important and interesting biomarker. Recent studies have shown that SPARCL1 is down-regulated in a number of cancers, including colorectal cancer [36], gastric cancer [37], breast cancer [38], prostate cancer [39], hepatocellular carcinoma, and non-small cell lung cancer [40]. However, no studies have been conducted concerning its clinical significance and diagnostic value in cervical cancer. Thus, we analyzed the expression of SPARCL1 during cervical cancer progression by tissue array and found that SPARCL1 expression is also down-regulated.

Currently, the biggest limitation for the clinical application of tumor markers is poor specificity [41]. Most tumor biomarker research has focused on differences in normal and tumor tissue or serum in patients, while ignoring research on the interference expression of these candidate biomarkers in other common diseases [42, 43], including cardiovascular disease and chronic diseases. Thus, evaluating the expression of potential cancer biomarkers in multiple diseases can provide a new method for investigating their specificities. In this study, we found that SPARCL1 is specifically regulated in a wide variety of tumors, while have no obvious changes in other senile or chronic diseases compared with normal samples. Taken together, our data indicates that SPARCL1 may be an excellent tumor biomarker and its clinical application warrants further investigation.

Footnotes

Acknowledgments

This work was funded by the Foundation of Sichuan Province Science and Technology Agency (2014JY00 39), Foundation of Sichuan Province Education Office (15ZB0245), Foundation of Sichuan Medicine Institute (16PJ112), and Foundation of Chengdu Medical College (CYTD15-03).

Conflict of interest

The authors have no competing financial interests to disclose.

Supplementary data

The supplementary files are available to download from http://dx.doi.org/10.3233/CBM-170501.

References

Ferlay

Soerjomataram

Dikshit

Eser

Mathers

Rebelo

Parkin

D.M.

Forman

and Bray

, Cancer incidence and mortality worldwide: Sources methods and major patterns in GLOBOCAN 2012, Int J Cancer 136 (2015), E359-E386.

Torre

L.A.

Bray

Siegel

R.L.

Ferlay

Lortet-Tieulent

and Jemal

, Global cancer statistics 2012, CA Cancer J Clin 65 (2015), 87–108.

Sikorski

, The risk of cervical cancer in female partners of HPV-infected males, Cent European J Urol 62 (2009), 9–14.

Mejía

Muñoz

Trueba

Tinoco

and Zapata

, Prevalence of human papillomavirus types in cervical cancerous and precancerous lesions of ecuadorian women, J Med Virol 88 (2016), 144–152.

Shi

Jiang

Fang

Liu

X.I.

and Chen

, Overexpression of SPARC correlates with poor prognosis in patients with cervical carcinoma and regulates cancer cell epithelial-mesenchymal transition, Oncol Lett 11 (2016), 3251–3258.

Wentzensen

and von Knebel Doeberitz

, Biomarkers in cervical cancer screening, Dis Markers 23 (2007), 315–330.

Barrett

Troup

D.B.

Wilhite

S.E.

Ledoux

Rudnev

Evangelista

Kim

I.F.

Soboleva

Tomashevsky

and Edgar

, NCBI GEO: Mining tens of millions of expression profiles – database and tools update, Nucleic Acids Res 35 (2007), D760-D765.

Ghosh

Huang

Wilson

K.D.

Dey

and Wu

J.C.

, Dissecting the oncogenic and tumorigenic potential of differentiated human induced pluripotent stem cellsand human embryonic stem cells, Cancer Res 71 (2011), 5030–5039.

Hayes

D.N.

Grandis

J.R.

and El-Naggar

A.K.

, The cancer genome atlas: Integrated analysis of genome alterations in squamous cell carcinoma of the head and neck, J Clin Oncol 31 (2013), 6009.

10.

Cancer Genome Atlas Research Network: Integrated genomic analyses of ovarian carcinoma, Nature 474 (2011), 609–615.

11.

Beuselinck

Job

Becht

Karadimou

Verkarre

Couchy

Giraldo

Rioux-Leclercg

Molinié

Sibony

Elaidi

Teghom

Patard

J.J.

Méjean

Fridman

W.H.

Sautès-Fridman

de Reyniès

Oudard

and Zucman-Rossi

, Molecular subtypes of clear cell renal cell carcinoma are associated with sunitinib response in the metastatic setting, Clin Cancer Res 21 (2015), 1329–1339.

12.

Pineda

Real

F.X.

Kogevinas

Carrato

Chanock

S.J.

Malats

and Van Steen

, Integration analysis of three omics data using penalized regression methods: An application to bladder cancer, PloS Genet 11 (2015), e1005689.

13.

Tan

and Wang

, Prognostic role of platelet-lymphocyte ratio in colorectal cancer: A systematic review and meta-analysis, Medicine (Baltimore) 95 (2016), e3837.

14.

Liu

Zhu

and Jiang

, A meta-analysis: Is low-dose computed tomography a superior method for risky lung cancers screening population? Clin Respir J 10 (2016), 333–341.

15.

Wang

Xue

Pan

Zhang

Kuang

Zhou

Xiong

Zeng

and Yang

, The long noncoding RNA MALAT-1 is anovel biomarker in various cancers: Ameta-analysis based on the GEO database and literature, J Cancer 7 (2016), 991–1001.

16.

Pashaei

Guzel

Ozgurses

M.E.

Demirel

Aydin

and Ozen

, A meta-analysis: Identification of common mir-145 target genes that have similar behavior in different GEO datasets, PloS One 11 (2016), e0161491.

17.

D.M.

Zhang

Liu

R.Y.

Sang

Y.X.

Zhou

G.C.

Yang

J.L.

Tong

A.P.

and Wang

C.T.

, Phosphorylation and changes in the distribution of nucleolin promote tumor metastasis via the PI3K/Akt pathway in colorectal carcinoma, FEBS Lettv 588 (2014), 1921–1929.

18.

Wang

Meng

Dai

Yang

Wang

Chen

Yang

Wang

and Jiang

, Identification of associations between small molecule drugs and miRNAs based on functional similarity, Oncotarget 7 (2016), 38658–38669.

19.

Xie

Qin

Liu

Zhang

Wang

Shi

Zhou

and Liu

, Cep70 overexpression stimulates pancreatic cancer by inducing centrosome abnormality and microtubule disorganization, Sci Rep 6 (2016), 21263.

20.

Iyer

M.K.

Niknafs

Y.S.

Malik

Singhal

Sahu

Hosono

Barrette

T.R.

Prensner

J.R.

Evans

J.R.

Zhao

Poliakov

Cao

Dhanasekaran

S.M.

Y.M.

Robinson

D.R.

Beer

D.G.

Feng

F.Y.

Iyer

H.K.

and Chinnaiyan

A.M.

, The landscape of long noncoding RNAs in the human transcriptome, Nat Genet 47 (2015), 199–208.

21.

Chen

Sun

Wei

Zhang

Rezaeian

A.H.

Teruya-Feldstein

Gupta

Liang

Lin

H.K.

Hung

M.C.

and Ma

, LIFR is a breast cancer metastasis suppressor upstream of the hippo-YAP pathway and a prognostic marker, Nat Med 18 (2012), 1511–1517.

22.

Masterson

Sorgeloos

Winder

Lechner

Marker

Malhotra

Sudhoff

Jani

Goon

and Sterling

, Deregulation of SYCP2 predicts early stage human papillomavirus-positive oropharyngeal carcinoma: A prospective whole transcriptome analysis, Cancer Sci 106 (2015), 1568–1575.

23.

Martinez

Wang

Hobson

K.F.

Ferris

R.L.

and Khan

S.A.

, Identification of differentially expressed genes in HPV-positive and HPV-negative oropharyngeal squamous cell carcinomas, Eur J Cancer 43 (2007), 415–432.

24.

Wang

Xie

Chen

Tan

and Jiang

, The role of kif4A in doxorubicin-induced apoptosis in breast cancer cells, Mol Cells 37 (2014), 812–818.

25.

Taniwaki

Takano

Ishikawa

Yasui

Inai

Nishimura

Tsuchiya

Kohno

Nakamura

and Daigo

, Activation of KIF4A as a prognostic biomarker and therapeutic target for lung cancer, Clin Cancer Res 13 (2007), 6624–6631.

26.

Arriola

Marchio

Tan

D.S.

Drury

S.C.

Lambros

M.B.

Natrajan

Rodriguez-Pinilla

S.M.

Mackay

Tamber

Fenwick

Jones

Dowsett

Ashworth

and Reis-Filho

J.S.

, Genomic analysis of the HER2/TOP2A amplicon in breast cancer and breast cancer cell lines, Lab Invest 88 (2008), 491–503.

27.

Nielsen

K.V.

Ejlertsen

Møller

Jørgensen

J.T.

Knoop

Knudsen

and Mouridsen

H.T.

, The value of TOP2A gene copy number variation as a biomarker in breast cancer: Update of DBCG trial 89D, Acta Oncol 47 (2008), 72–734.

28.

Nagelkerke

Mujcic

Bussink

Wouters

B.G.

van Laarhoven

H.W.

Sweep

F.C.

and Span

P.N.

, Hypoxic regulation and prognostic value of LAMP3 expression in breast cancer, Cancer 117 (2011), 3670–3681.

29.

Sun

Wang

Zhu

Mei

Wang

Zhang

and Huang

, Prognostic value of LAMP3 and TP53 overexpression in benign and malignant gastrointestinal tissues, Oncotarget 5 (2014), 12398–12409.

30.

Taniuchi

Nakagawa

Nakamura

Eguchi

Ohigashi

Ishikawa

Katagiri

and Nakamura

, Down-regulation of RAB6KIFL/KIF20A a kinesin involved with membrane trafficking of discs large homologue 5 can attenuate growth of pancreatic cancer cell, Cancer Res 65 (2005), 105–112.

31.

Asahara

Takeda

Yamao

Maguchi

and Yamaue

, Phase I/II clinical trial using HLA-A24-restricted peptide vaccine derived from KIF20A for patients with advanced pancreatic cancer, J Transl Med 11 (2013), 291.

32.

Dudderidge

T.J.

Stoeber

Loddo

Atkinson

Fanshawe

Griffiths

D.F.

and Williams

G.H.

, Mcm2, geminin and KI67 define proliferative state and are prognostic markers in renal cell carcinoma, Clin Cancer Res 11 (2005), 2510–2517.

33.

Torres-Rendon

Roy

Craig

G.T.

and Speight

P.M.

, Expression of mcm2 geminin and ki67 in normal oral mucosa oral epithelial dysplasias and their corresponding squamous-cell carcinomas, Br J Cancer 100 (2009), 1128–1134.

34.

Leonard

McCann

J.L.

Starrett

G.J.

Kosyakovsky

Luengas

E.M.

Molan

A.M.

Burns

M.B.

McDougle

R.M.

Parker

P.J.

Brown

W.L.

and Harris

R.S.

, The PKC/NF-κB signaling pathway induces APOBEC3B expression in multiple human cancers, Cancer Res 75 (2015), 4538–4547.

35.

Suspène

Guétard

Henry

Sommer

Wain-Hobson

and Vartanian

J.P.

, Extensive editing of both hepatitis B virus DNA strands by APOBEC3 cytidine deaminases in vitro and in vivo , Proc Natl Acad Sci USA 102 (2005), 8321–8326.

36.

Naschberger

Liebl

Schellerer

V.S.

Schütz

Britzen-Laurent

Kölbel

Schaal

Haep

Regensburger

Wittmann

Klein-Hitpass

Rau

T.T.

Dietel

Méniel

V.S.

Clarke

A.R.

Merkel

Croner

R.S.

Hohenberger

and Stürzl

, Matricellular protein SPARCL1 regulates tumor microenvironment-dependent endothelial cell heterogeneity in colorectal carcinoma, J Clin Invest 126 (2016), 4187–4204.

37.

Jakharia

Borkakoty

and Singh

, Expression of SPARC like protein 1 (SPARCL1) extracellular matrix-associated protein is down regulated in gastric adenocarcinoma, J Gastrointest Oncol 7 (2016), 278–283.

38.

Oliveras-Ferraros

Vazquez-Martin

Cuyàs

Corominas-Faja

Rodríguez-Gallego

Fernández-Arroyo

Martin-Castillo

Joven

and Menendez

J.A.

, Acquired resistance to metformin in breast cancer cells triggers transcriptome reprogramming toward a degradome-related metastatic stem-like profile, Cell Cycle 13 (2014), 1132–1144.

39.

Hurley

P.J.

Marchionni

Simons

B.W.

Ross

A.E.

Peskoe

S.B.

Miller

R.M.

Erho

Vergara

I.A.

Ghadessi

Huang

Gurel

Park

B.H.

Davicioni

Jenkins

R.B.

Platz

E.A.

Berman

D.M.

and Schaeffer

E.M.

, Secreted protein acidic and rich in cysteine-like 1 (SPARCL1) is down regulated in aggressive prostate cancers and is prognostic for poor clinical outcome, Proc Natl Acad Sci USA 109 (2012), 14977–14982.

40.

Isler

S.G.

Schenk

Bendik

Schraml

Novotna

Moch

Sauter

and Ludwig

C.U.

, Genomic organization and chromosomal mapping of SPARC-like 1 a gene down regulated in cancers, Int J Oncol 18 (2001), 521–526.

41.

Schmitz-Dräger

B.J.

Droller

Lokeshwar

V.B.

Lotan

Hudson

M.A.

van Rhijn

B.W.

Marberger

M.J.

Fradet

Hemstreet

G.P.

Malmstrom

P.U.

Ogawa

Karakiewicz

P.I.

and Shariat

S.F.

, Molecular markers for bladder cancer screening early diagnosis and surveillance: The WHO/ICUD consensus, Urol Int 94 (2014), 1–24.

42.

Dong

Zhou

Pang

X.W.

Tian

X.J.

Zhang

and Zhang

, The tumor-suppressive function of UNC5D and its repressed expression in renal cell carcinoma, Clin Cancer Res 19 (2013), 2883–2892.

43.

Sawyers

C.L.

, The cancer biomarker problem, Nature 452 (2008), 548–552.

44.

Pyeon

Newton

M.A.

Lambert

P.F.

den Boon

J.A.

Sengupta

Marsit

C.J.

Woodworth

C.D.

Connor

J.P.

Haugen

T.H.

and Smith

E.M.

, Fundamental differences in cell cycle deregulation in human papillomavirus-positive and human papillomavirus-negative head/neck and cervical cancers, Cancer Research 67 (2007), 4605–4619.

45.

Zhai

Kuick

Nan

Ota

Weiss

S.J.

Trimble

C.L.

Fearon

E.R.

and Cho

K.R.

, Gene expression analysis of preinvasive and invasive cervical squamous cell carcinomas identifies HOXC10 as a key mediator of invasion, Cancer Research 67 (2007), 10163–10172.

46.

Espinosa

A.M.

Alfaro

Romanbasaure

Guardadoestrada

Palma

Serralde

Medina

Juárez

Bermúdez

and Márquez

, Correction: Mitosis is a source of potential markers for screening and survival and therapeutic targets in cervical cancer, Plos One 8 (2013), 142–142.

47.

Thomas

Mahantshetty

Kannan

Deodhar

Shrivastava

S.K.

Kumar-Sinha

and Mulherkar

, Expression profiling of cervical cancers in indian women at different stages to identify gene signatures during progression of the disease, Cancer Medicine 2 (2013), 836–848.

48.

Scotto

Narayan

Nandula

S.V.

Subramaniyam

Schneider

Kaufmann

A.M.

Wright

J.D.

Pothuri

Mansukhani

and Murty

V.V.

, Identification of copy number gain and overexpressed genes on chromosome arm 20q by an integrative genomic approach in cervical cancer: Potential role in progression, Genes Chromosomes & Cancer 47 (2008), 755–765.

49.

den Boon

J.A.

Pyeon

Wang

S.S.

Horswill

Schiffman

Sherman

Zuna

R.E.

Wang

Hewitt

S.M.

and Pearson

, Molecular transitions from papillomavirus infection to cervical precancer and cancer: Role of stromal estrogen receptor signaling, Proceedings of the National Academy of Sciences of the United States of America 112 (2015), 3255–3264.

50.

Mine

K.L.

Shulzhenko

Yambartsev

Rochman

Sanson

G.F.O.

Lando

Varma

Skinner

Volfovsky

and Deng

, Gene network reconstruction reveals cell cycle and antiviral genes as major drivers of cervical cancer, Nature Communications 4 (2013), 54–56.

Integrated analysis reveals down-regulation of SPARCL1 is correlated with cervical cancer development and progression

Abstract

Keywords

1. Introduction

2. Materials and methods

2.1 Materials

2.2 Differential gene screening

2.3 Data extraction and meta-analysis

2.4 GO enrichment analysis

2.5 Co-expression network using TCGA data

2.6 Immunohistochemical staining and assessment of SPARCL1 expression

2.8 Statistical analysis

Table 1 Study information on eight GEO datasets which consist of normal cervical and cervical cancer samples

3.1 Integrated analysis reveals nine differentially expressed gene signatures that are associated with cervical cancer development

3.2 Meta-analysis of expression level and diagnostic value of the nine genes of interest during cervical cancer development

3.3 Progression-related expression and co-expression networks reveal SPARCL1 as an important factor in both cervical cancer development and progression

3.5 Diagnostic specificity of serum SPARCL1 in cervical cancer development

4. Discussion

Footnotes

Acknowledgments

Conflict of interest

Supplementary data

References

Table 1
Study information on eight GEO datasets which consist of normal cervical and cervical cancer samples