Abstract

   Thrombospondin 1 and thrombospondin 2 (THBS1 and THBS2) share similar
   multifunctional domains, and are known to be antiangiogenic. However,
   the expression pattern of THBS1 and THBS2 is different, and the
   specific role of THBS2 in different subtypes of lung cancer remains
   largely unclear. To evaluate the significance of THBS1 and THBS2 in the
   development of lung cancer, the present study performed a
   microarray-based systematic-analysis to determine the transcript levels
   of thrombospondins and their relation to the prognosis in lung cancer.
   THBS1 was in general underexpressed in lung cancer; in contrast, mRNA
   levels of THBS2 were markedly overexpressed in a number of datasets of
   non-small cell lung carcinoma (NSCLC), including lung adenocarcinoma
   (AC) and squamous cell carcinoma. Similar expression pattern of THBS1
   and THBS2 was verified in pulmonary AC cell lines with real-time PCR
   analysis. The survival of lung AC patients with high THBS2 mRNA
   expression levels was poorer than patients with low levels of
   expression of THBS2. In a microarray-based analysis, genes coexpressed
   with THBS1 or THBS2 were determined. Pulmonary AC patients with a high
   expression level of sevenTSHB1-coexpressed genes (CCL5, CDH11, FYB,
   GZMK, LA-DQA1, PDE4DIP, and SELL) had better survival rates than those
   with a low expression level. Patients with a high expression of seven
   TSHB2-coexpressed genes (CHI3L1, COL5A2, COL11A1, FAP, MXRA5, THY1, and
   VCAN) had poor survival rates. Downregulation of VCAN and THBS2 with
   shRNA inhibited the cell proliferation in the A549 cell line. In
   summary, THBS1 functions as a tumor suppressor in lung adenocarcinoma.
   However, THBS2 may play a double-edged role in the progression of lung
   AC, i.e. anti-angiogenic and oncogenic function. Further study on the
   mechanism underlying the activity of THBS2 is warranted to have further
   implications for cancer diagnosis and treatment of pulmonary AC.

Introduction

   Lung cancer is the leading cause of cancer mortality in the world in
   recent decades, accounting for about 20% of all cancer deaths in both
   men and women [[34]1]. Histologically, there are two major types of
   lung cancer, non-small cell lung cancer (NSCLC) and small cell lung
   cancer, with 85% of cases due to NSCLC. NSCLC can be divided into three
   main subtypes: adenocarcinoma (AC, 40% of lung cancers), squamous cell
   carcinoma (SCC, 25–30% of lung cancers), and large cell carcinoma (10%
   of lung cancers). Overall, the 5-year survival rate for patients with
   NSCLC is less than 18%, and it is only about 7% for patients with small
   cell lung cancer [[35]1]. Metastatic spread was reported in more than
   70% of NSCLC patients with advanced-stage disease, with the metastases
   mainly affecting the brain, liver and bone sites. In all cases, the
   patients died within 18 months or soon after. Investigating changes in
   the tumor-associated microenvironment during cancer progression is
   important for targeted therapy and improvement of clinical outcomes in
   lung cancer [[36]2]

   Thrombospondins (THBSs or TSP) are secreted glycoproteins, with various
   functional domains involved in embryonic development, wound healing
   [[37]3], angiogenesis [[38]4], and inflammatory response [[39]5,
   [40]6]. THBSs are subdivided into two subgroups: subgroup A and
   subgroup B. Subgroup A includes THBS1 and THBS2, which can form
   trimers. Subgroup B, which includes THBS3, THBS4, and THBS5 (also
   referred to as cartilage oligomeric matrix protein [COMP]), can form
   pentamers. A distinct feature of subgroup A is the presence of three
   thrombospondin type 1 (TSR) repeats, which interact with CD36 and
   beta-1 integrins. The interaction of the TSR domain and membrane CD36
   in endothelial cells suppresses cell migration and induces apoptosis,
   which results in the inhibition of angiogenesis. Only THBS1 contains an
   RFK motif located between the first- and second-repeat of the TSR
   domain and responds to the activation of transforming growth
   factor-beta [[41]7]. Except for THBS5/COMP, all THBS proteins contain
   an N-terminal domain as a signature motif and are involved in cell
   adhesion through binding with several receptors or ligands, such as
   calreticulin and integrins [[42]8]. Five THBSs have a carboxy-terminal
   domain, which interacts mainly with CD47, in addition to at least three
   copies of epidermal growth factor-like domain (type 2 repeats) and
   several copies of calcium-binding domains (type 3 repeats) [[43]9].

   THBS1 is the most studied gene among the THBS family. It plays a
   functional role in inhibiting tumor growth, cell migration, and
   neovascularization; it also acts as an endogenous tumor suppressor by
   interacting with its receptors, CD36 and CD47, or activating
   transforming growth factor-beta signaling [[44]10]. Decreasing levels
   of THBS1 in NSCLC were reported to be correlated with worse prognoses
   [[45]11]. Several THBS1-based compounds are in development for cancer
   therapy [[46]12]. On the other hand, the expression profile of another
   subgroup A member, THBS2, is variable in different types of cancers.
   The expression of THBS2was up-regulated in some types of cancers
   [[47]13–[48]15], but was down-regulated in other types of cancers
   [[49]16, [50]17]. High levels of THBS2 and fibroblast growth factor-2
   in the serum of NSCLC patients predicted poor survival rates [[51]18].
   In contrast, overexpression of THBS2 suppressed tumor growth in
   squamous cell carcinomas and Lewis lung carcinoma xenograft mouse tumor
   models [[52]19]. A truncated-recombinant THBS2 protein inhibited tumor
   growth and angiogenesis in vivo [[53]20].

   Although THBS subgroup A members share many structural domains or
   functional motifs, there are few studies systematically evaluating
   their expression patterns or importance in human lung cancer using
   clinical microarray databases. The Oncomine cancer microarray database
   integrates gene expression data and clinical data, and contains 20
   major cancers and over 4,700 microarray experiments [[54]21]. The
   Kaplan–Meier plotter database integrates gene expression data and
   clinical data and contains information on 22,277 genes and their
   effects on survival in 2,437 lung cancer patients [[55]22]. In the
   present study, both the Oncomine and Kaplan–Meier plotter databases
   were used to investigate whether the transcript levels of THBS1 and
   THBS2 were altered in lung and correlated with the clinical prognosis.
   In addition, the functional characteristics and molecular mechanism of
   THBS2 and its coexpressed genes were investigated in a
   systematic-analysis. The results may shed light on the role of THBS2 in
   the tumor microenvironment during the progression of lung AC. To the
   best of our knowledge, this is the first study to systematically
   evaluate the correlation between transcript levels of THBSs and
   clinical outcomes in lung cancer patients using a systematic-analysis.

Methods

Oncomine database analysis

   Analysis of THBSs expression change in common or selected cancer
   tissues was performed by using the online cancer microarray database
   Oncomine, ([56]www.oncomine.org, Compendia biosciences, Ann Arbor, MI,
   USA) [[57]21]. mRNA expression of clinical specimens of tumor and
   normal (cancer vs. normal) was compared and extracted between April
   2015 and June 2016. The threshold search criteria used for the [58]Fig
   1 (gene summary view) in the study were a p-value < 1E-4, a fold change
   > 2 and a gene rank in top 5%. P-values and fold-changes presented in
   this study for differential expression analysis of THBS genes were
   calculated with Oncomine using a two-sided Student’s t-test and
   multiple testing corrections. To examine the THBS2 or THBS1 coexpressed
   genes, the Oncomine tool was utilized to conduct the coexpression
   analysis of the microarray datasets. Three datasets (Garber Lung,
   Gordon Lung, and Landi Lung) were selected for the coexpression
   analysis, with each dataset consisting of >50 AC samples. To explore
   the THBS2 coexpressed gene in breast and gastric cancer, three breast
   cancer datasets (Ivshina Breast, Minn Breast 2 and Schmidt Breast) and
   three gastric cancer datasets (Chen Gastric, DErrico Gastric and
   DErrico Gastric) were analyzed. The top 5% of genes within each dataset
   were selected by using the co-expression score. The genes that appeared
   in at least two of the three datasets were defined as THBS2 and THBS1
   coexpressed genes.

Fig 1. THBS2 was overexpressed in lung adenocarcinoma tissue.

   [59]Fig 1
   [60]Open in a new tab

   (A) Expressions of THBS1 and THBS2 in 20 common cancers were compared
   with corresponding normal tissues (Oncomine Database). The threshold of
   search criteria for datasets of cancer vs. normal analysis was a
   p-value <1E-4, a fold change >2, and a gene rank in top 5%. Analysis of
   (B) THBS1 and (C) THBS2 mRNA expression in lung normal tissue and lung
   AC tissue by using Oncomine database. Relative (D) THBS1 and (E) THBS2
   expression of six human lung AC cell lines (A549, H1299, CL1-0, CL1-1,
   CL1-2 and CL1-5) compared with the mean value of a normal lung cell
   line (IMR90). THBS1, Thrombospondin 1. THBS2, Thrombospondin 2. AC,
   adenocarcinoma.* represent a P value < 0.05.

Kaplan–Meier plotter database analysis

   The survival analysis and hazard ratio estimation of the expression
   values of THBS1, THBS2, and coexpressed genes in cancers were performed
   with a Kaplan–Meier plotter online database ([61]www.kmplot.com), which
   contains information on 22,277 genes and their effects on survival in
   2,437 lung cancer, 4,142 breast, and 1,065 gastric patients (June 2015)
   [[62]22]. The hazard ratios (95% confidence intervals) and log rank
   p-values were also computed using the Kaplan–Meier plotter database.
   The hazard ratios were estimated using a Cox-proportional hazards
   model. The progression-free survival rate of the lung cancer patients
   was analyzed. The patient samples were split into two groups with the
   best cut-off and compared using the Kaplan–Meier plotter. The specific
   probes (JetSet best probes) analyzed for recognizing each gene are
   listed in all related tables.

SurvExpress database analysis

   The SurvExpress database was used to further compare the survival rates
   of individuals segregated according to the THBS2 expression levels
   within each risk group (Okayama Lung dataset, [63]GSE31210) [[64]23].
   The risk groups generated from the SurvExpress database were based on
   the prognostic index (PI) [[65]24], and were split by the ordered PI
   (higher values for higher risk) with equal number of samples in each
   group. The PI was computed using the expression levels and values
   obtained from Cox fitting [[66]23].

Gene ontology and pathway enrichment analysis

   Gene ontology and pathway enrichment analysis were conducted to examine
   THBS2 coexpressed genes ([67]S4 Table) by using the Database for
   Annotation, Visualization and Integrated Discovery (DAVID;
   [68]http://david.abcc.ncifcrf.gov/) [[69]25]. The categories
   GOTERM_BP_3, GOTERM_CC_2 and GOTERM_MF_3, were selected, and all
   options were set as defaults. The data listed in the table were with
   hit count >5, and p-value <0.001.

Construction of the gene interaction network

   The gene interaction network was constructed with network building tool
   in MetaCore (Thompson Reuters, New York, NY) as previously described
   [[70]26]. The direct interaction of the identified THBS2 coexpressed
   gene list ([71]S1 Table)was predicted and obtained from MetaCore
   software. Genes with no interactions were not shown in the network.

Cell culture

   The lung cancer cell lines were kind gifts from Dr. Pan-Chyr Yang
   (National Taiwan University, Taipei, Taiwan) [[72]27]. CL1-0, CL1-1,
   CL1-2 and CL1-5 cells were grown in RPMI 1640 media (LONZA,
   Walkersville, MD, USA); A549 and H1299 cells were grown in Dulbecco's
   Modified Eagle Medium (GIBCO, Carlsbad, CA, USA) media. Both culture
   media were supplemented with 10% fetal bovine serum (GIBCO, Carlsbad,
   CA, USA) and100 U/ml penicillin and 100 mg/ml streptomycin (HYCLONE,
   Logan, UT, USA). Cells were maintained at 37°C in 5% CO[2] incubator.

Real-time PCR

   Lysate of human normal lung cell line, IMR90, was a kind gift from Dr.
   Yi-Ching Wang. Total RNA was extracted from cells by using TRIzol
   (Invitrogen, Carlsbad, CA, USA). cDNA was synthesized using MMLV
   reverse transcriptase (Promega, Madison, WI, USA). The following GAPDH,
   THBS1 and THBS2 sense and antisense primers were used as previously
   described: THBS1 5′-TTG TCT TTG GAA CCA CAC CA-3′ and 5′-CTG GAC AGC
   TCA TCA CAG G-3′ [[73]28]; THBS2 5′-CGT GGA CAA TGA CCT TGT TG-3′ and
   5′-GCC ATC GTT GTC ATC ATC AG-3′ [[74]29]; GAPDH 5′-AGC CAC ATC GCT CAG
   ACA C-3′ and 5′-GCC CAA TAC GAC CAA ATC C-3′; VCAN 5’-TCC TGA TTG GCA
   TTA GTG AAG-3’ and 5’-CTG GTC TCC GCT GTA TCC-3’. Real-time PCR was
   performed on a StepOne ™ real-time PCR instrument (Applied Biosystems,
   Foster City, CA, USA) using Fast SYBR Green Master Mix (Applied
   Biosystems). The cycling conditions were 10min at 95°C and 45 cycles at
   95°C for 15s and 60°C for 60s. The 2^ΔΔCt method was used to calculate
   the relative RNA expression, which was normalized with GAPDH
   expression.

RNA interference and lentivirus production

   THBS2 shRNAs and VCAN shRNAs were obtained from the National RNAi Core
   facility (Academia Sinica, Taipei, Taiwan). The TurboFect transfection
   reagent (Thermo Fisher Scientific, Slangerup, Denmark) was used to
   generate the lentiviral particles according to the protocol provided
   from the National RNAi Core facility. The following target sequences
   were used: THBS2 shRNA-1, GTG CCT TCA GAG GAT AAA TAT; THBS2 shRNA-2,
   GTC TTC AAT GAA CGA GAC AAT; VCAN shRNA-1, GCC ACA GTT ATT CCA GAG ATT;
   and VCAN shRNA-2, GTG AAT TTC TCC GCA TCA AAT.

Cell proliferation assay

   After incubation with lentivirus for 24 hours, infected A549 cells were
   seeded in 96-well plates at a cell density of 5,000 cells per well and
   cultured for 48 hours. Cell proliferation was examined with WST-1 Cell
   Proliferation Reagent (Clontech, Mountain View, CA, USA). 10μL WST-1
   reagent was added into culture wells and incubating for 1hr. Absorbance
   was measured at wavelength of 450 nm by using a scanning multi-well
   spectrophotometer.

Statistical analysis

   p-values and fold-changes for differential expression analysis of THBS
   genes generated form Oncomine database were calculated using a
   two-sided Student’s t-test and multiple testing corrections with
   Oncomine. Statistical analyses of the mRNA expression in real-time PCR
   experiments were performed by using One-Way ANOVA in GraphPad Prism 5
   software. Statistical analyses of the cell proliferation assay were
   performed by using the t-test in GraphPad Prism 5 software. P values
   <0.05 were considered significant.

Results

The expression patterns of THBS1 and THBS2 were diverse in various types of
cancer

   To determine changes in THBS and THBS2 transcripts in clinical
   specimens of lung cancer and other cancers, the mRNA level of THBSs in
   various types of cancer was examined using the Oncomine cancer
   microarray database. Based on the data on gene summary views
   (neoplastic vs. normal tissue), THBS2 was significantly up-regulated in
   11 of 20 common cancers ([75]Fig 1A). It was overexpressed in
   colorectal (11 of 36 studies) ([76]S1 Table), gastric (8 of 24 studies)
   ([77]S2 Table), lung (9 of 37 studies) ([78]Table 1 and [79]S3 Table),
   and pancreatic cancer (5 of 12 studies) ([80]S4 Table). In contrast,
   only a small number of studies reported that THBS1 was significantly
   up- ([81]S2 Table) or down-regulated ([82]Table 1, [83]S1 and [84]S3
   Tables) in different types of cancer ([85]Fig 1A). We next focused on
   investigating the changes in THBS1 and THBS2 mRNA expression in lung
   cancer subtypes. The mRNA level of THBS1 was significantly
   underexpressed in small cell lung cancer (a fold change of -5.923)
   ([86]S3 Table). Analysis of the expression levels of THBS1 in the two
   main subtypes of NSCLC in the same microarray dataset revealed
   decreased mRNA levels of THBS1 in both lung AD (a fold change of
   -2.159) ([87]Fig 1B and [88]Table 1) and lung SCC (a fold change of
   -1.968) ([89]S3 Table), although the data did not satisfy the threshold
   criteria set in this study. On the other hand, mRNA levels of THBS2
   were markedly overexpressed in lung AC (an average fold change of
   3.308, n = 6) ([90]Fig 1C and [91]Table 1) and lung SCC (an average
   fold change of 8.915, n = 3) ([92]S3 Table). To verify this finding,
   the mRNA expression of THBS1 and THBS2 was examined in a normal lung
   cell line (IMR90) and six lung AC cell lines (A549 H1299, CL1-0, CL1-1,
   CL1-2, and CL1-5) using real-time PCR. Lower expression levels of THBS1
   were observed in four lung AC cell lines compared with those in normal
   lung cell lines ([93]Fig 1D). The expression level of THBS2 mRNA was
   increased (more than 2-fold) in the A549, CL1-0, CL1-1, CL1-2, and
   CL1-5 lines ([94]Fig 1E). This evidence suggested that THBS1 and THBS2
   may play opposite roles in NSCLC.

Table 1. mRNA expression levels of THBS1 and THBS2 in lung adenocarcinoma.

   Gene P-Value (Cancer/Normal) Fold Change (Cancer/Normal) Ranking (Top%)
   Dataset #Samples Reference
   THBS1 0.001 -2.159 14 Bhattacharjee 203 1
   THBS2 6.20E-23 4.251 1 Landi 107 2
   2.69E-9 3.356 1 Su 66 3
   2.55E-19 3.307 1 Selamat 116 4
   2.83E-8 3.416 2 Stearman 39 5
   1.11E-13 3.965 2 Hou 110 6
   1.52E-11 4.551 1 Wei 50 7
   [95]Open in a new tab

   All references in this table were listed in the [96]S7 Table.