The SKOV3 and ID8 cell lines were obtained from ATCC and Sigma-Aldrich, respectively. We tested for mycoplasma contamination and verified cells by short tandem repeat (STR) test; morphology was confirmed by pathologist before the experiments. Cells were cultured in DMEM medium mixed with 10% fetal bovine serum (Biosharp, Hefei, Anhui, China), 100 U/mL penicillin, and 100 U/mL streptomycin (Phygene, Fuzhou, Fujian, China) at 37 °C with 5% CO${}_2$.

The SPARC over-expressing plasmid was synthesized and the coding area of SPARC, based on pcDNA3.1+ vector, was added. The SPARC-over-expression (OE) vector was transfected into SKOV3 cells and ID8 cells by Lipofectamine 8000 (Beyotime, Shanghai, China). The empty plasmid was also transfected as the internal control.

SKOV3-Vector/ID8-Vector cells and SKOV3-SPARC/ID8-SPARC cells (4 $\times$ 10${}^5$) were grown in 6-well plates to a density greater than 90%. A single scratch in the center of the plate was made by a p1250 pipette tip. After 24 h, photomicrographs were taken though a light microscope (EVOS, ThermoFisher, Waltham, MA, USA). The cell merging front was calculated by ImageJ (version 1.54f, LOCI, University of Wisconsin, Madison, WI, USA).

The cells (1 $\times$ 10${}^4$ cells) were cultured in a 100-mm petri dish and incubated for 10 days with 10 mL medium, and adding 1 mL medium every other day. The colonies were then fixed with methanol for 10 min, then stained with 0.5% crystal violet for 10 min. The cells were then washed twice with phosphate buffered saline (PBS) at room temperature. Photomicrographs were then taken after dish dried.

SKOV3-Vector/ID8-Vector cells and SKOV3-SPARC/ID8-SPARC cells (SKOV3: 8 $\times$ 10${}^5$ cells per well; ID8: 3 $\times$ 10${}^5$ cells per well) were added to the upper chamber in one 24-well Transwell plate (Corning, Corning, NY, USA) with serum-free medium. Fetal bovine serum (FBS) (10%) was added to the lower chamber. After 24 h, the cells that migrated into the lower chamber were fixed with methanol and stained with 0.5% crystal violet. Photomicrographs were taken through a light microscope (EVOS, ThermoFisher, Waltham, MA, USA), and ImageJ software was used to count cells.

SKOV3-Vector/ID8-Vector cells and SKOV3-SPARC/ID8-SPARC cells were seeded into 96-well plates at 5 $\times$ 10${}^3$ cells. CCK-8 (10%) (Beyotime, Shanghai, China) as added to each well on Days 1, 2, and 3. The cells were incubated for 30 min at 37 °C and read at 450 nm by a plate reader.

A 24-well plate was precoated with 100 µL fibronectin (20 µg/mL), 100 µL collagen I (100 µg/mL), and 100 µL poly-l-lysine (100 µg/mL) in PBS, and placed in a cell incubator for 1 h. The PBS was replaced and 300 µL blocking buffer (0.5% bovine serum albumin (BSA) in medium) was added to each well for an additional 60 min. Then, the blocking buffer was replaced with 1 $\times$ 10${}^5$ cells per well (SKOV3-Vector/ID8-Vector cells and SKOV3-SPARC/ID8-SPARC cells) in serum-free medium. After incubating in the cell incubator for 30 min, the medium was wiped out, and the adherent cells were fixed for 5 min with 100 µL methanol. The cells were then stained with 100 µL of 0.5% crystal violet for 5 min. Finally, the plate was flushed with PBS and air-dried. Photomicrographs were taken and the cells were counted by ImageJ software.

Agarose solution (1%) was prepared and autoclaved, then spread on the bottom of the 100-mm dish when the temperature dropped to about 40 °C. The cells (1 $\times$ 10${}^4$ cells) were seeded and incubated in a 100-mm dish with complete medium for 5 days until the cells formed spheres. Then, these cells were poured into another 100-mm dish without agarose coating, and incubated in the same condition for 10 days.

UCSC Xena and Genotype-Tissue Expression (GTEx) databases were explored by UCSCXenaShiny, an R software package (version 4.23, Lucent Technologies, Murray Hill, NJ, USA) [16]. The mRNA expression and DNA methylation of SPARC were screened with this application (Plugin Version: v0.1.0) in both tumor samples and normal samples, if any. A p $\le$ 0.05 was considered statistically significant. The correlation between SPARC mRNA with immune signatures was analyzed with a Spearman test using signatures of Tumor Immune Estimation Resource (TIMER) database, including T cell CD4${}^{+}$, T cell CD8${}^{+}$, myeloid dendritic cell, neutrophil, macrophage, and B cell, were calculated. A related heatmap was plotted with p values. Similarly, SPARC mRNA level correlations with Stemness, tumor mutation burden (TMB), and microsatellite instability (MSI) were calculated. A circle map of correlation values was plotted. Moreover, the mRNA expression level of SPARC in TCGA and GTEx databases was extracted and highlighted in a violin plot for ovarian cancer.

GEPIA is a customizable pan-cancer analysis online tool for sequencing data from TCGA with clinical parameters. The GEPIA database was accessed to obtain SPARC transcript expressions among the major stages of ovarian cancer sorted by transcripts per million [17]. An F value was calculated.

The Kaplan-Meier plotter for ovarian cancer was accessed to evaluate SPARC expression value with clinical outcome [18]. The probe SPARC (212667_at) was used for estimating overall survival or progression-free survival by the Kaplan-Meier method. A positive independent biomarker was considered as log-rank p $\le$ 0.05 as a threshold. In addition, specific drug-based overall survivals (Platin, taxane, docetaxel, paclitaxel, gemcitabine) were also calculated.

Frequent genetic mutations in ovarian-cancer are considered one of the main reasons for extremely poor clinical outcome. The ROC plotter database (website database with therapy effect associated with sequencing data of cancers) was accessed to explore the general drug response for chemical application in treating ovarian cancer. SPARC (200665_s_at) was validated in both drug-responder and non-responder cohorts, and the Area Under Curve (AUC) with p-value was calculated [19]. In addition, specific drugs (Platin, taxane, docetaxel, paclitaxel, gemcitabine) were applied by ROC plotter in both drug-response and non-response cohort. The relapse-free survival value at 6 mo (n = 1347) was set as the analysis cohort.

The Timer2.0 database was accessed to provide an estimation of immune infiltration with SPARC level. The Immune association model was applied for evaluation of the correlation of SPARC expression (log2 TPM (transcripts per kilobase million)) with infiltration of 9 major immune cell types (B cell; T cell CD4${}^{+}$; T cell CD8${}^{+}$; macrophage; cancer-associated fibroblast; natural killer (NK) cell; mast cell; myeloid dendritic cell; neutrophil) in ovarian cancer. The Rho with p value was calculated.

The Timer2.0 database was applied using cancer exploration model [20]. The correlation between SPARC RNA-seq expression (log2 TPM) and CD274, CDH1, CDH2, CTNNB1, FN1, MMP2, MMP9, VIM. ZEB1, and MUC1 were calculated by Pearson correlation with a p value. The immune checkpoint genes CTLA4 and PDCD1 were also counted with SPARC expression.

The LinkedOmics online database was used for the identification of DEGs. The methods were previously described [21]. Briefly, the Pearson correlation of SPARC with all RNA-seq transcripts was calculated. All positive-Pearson value related DEGs (p $\le$ 0.01) were selected for GO and KEGG enrichment analysis.

The Drug-Gene Interaction Database (https://www.dgidb.org/) is an online webtool that contains genetic information on drug-gene interactions, based on the biomarkers of target therapy for ovarian cancer [22]. Using the high or low expression of SPARC with the median value of the database cohort, the drug response for ovarian cancer patients was calculated.

Two groups were compared with Prism software (GraphPad, San Diego, CA, USA) using a two-tailed unpaired Student’s t-test. All data were represented as mean $\pm$ standard error of the mean (SEM). Differences were considered statistically significant if p $\le$ 0.05.

According to the pan-cancer expression analysis comparing tumor and normal tissue, SPARC is differentially expressed in ACC, BRCA, CESC, CHOL, COAD, DLBC, ESCA, GBM, HNSCC, KIRC, KIRP, LAML, LGG, LIHC, OV, PAAD, PRAD, READ, SKCM, STAD, THCA, THYM, and UCEC. With the exception of CESC, PRAD, and UCEC, the cancer subtypes showed an increased expression of SPARC (Fig. 1A). It is interesting that the DNA methylation of SPARC did not affect mRNA higher expression in tumor. The beta-value was higher in BLCA, BRCA, CESC, CHOL, ESCA, HNSCC, KIRC, KIRP, LUAD, LUSC, PAAD, PCPG, PRAD, and UCEC (Fig. 1B). This inconsistent status between mRNA expression and DNA methylation indicated post-transcriptional modification of SPARC. Importantly, SPARC is generally positively correlated with immune signatures (TIMER). In ovarian cancer, SPARC expression is positively related to T cell CD8${}^{+}$, T cell CD4${}^{+}$, neutrophil, myeloid dendritic cell, and macrophage, and negatively related with B cell (Fig. 1C). SPARC shows no association with TMB and MSI status (Fig. 1E,F). However, SPARC is generally negative with tumor stemness indicating potential differential functions (Fig. 1D).

Fig. 1.

Pan-cancer analysis of SPARC. (A) Bar plot of SPARC (secreted protein, acidic and rich in cysteine) expression in tumors from The Cancer Genome Atlas (TCGA) in tumor sample and normal tissue. TPM, Transcripts Per Kilobase Million. (B) The DNA methylation of SPARC (TCGA database) for tumor and normal tissue using a bar plot with beta value. (C) The heatmap of the correlation between SPARC mRNA with immune signature in Pan-cancer. (D) The circle map of the correlation between SPARC mRNA level and cancer stemness through a Pan-cancer analysis. (E) The correlation circle map of SPARC mRNA level and tumor mutational burden (TMB) in Pan-cancer. (F) The correlation circle map of SPARC mRNA level and microsatellite instability (MSI) through Pan-cancer analysis. Inner circle indicates correlation = –1, middle circle indicates correlation = 0, outer circle indicates correlation = 1. Cancer type TCGA study abbreviation list: https://gdc.cancer.gov/resources-tcga-users/tcga-code-tables/tcga-study-abbreviations.

SPARC was significantly increased in ovarian cancer tumor samples comparing normal ovarian tissue from the TCGA/GTEx database (Fig. 2A). According to the major-stage analysis, SPARC increased from early stages to late stages, F = 1.19 (Fig. 2B). Although the dataset lacks Stage I data, the other major stages from Stage IIa to Stage IV have already shown the increased expression pattern with advancing stages. The Kaplan–Meier (KM)-plot results showed that SPARC expression could be used as an ovarian cancer prognostic marker for both overall survival and progression-free survival predictions, p = 0.006 and 0.0013, respectively (Fig. 2C,D). It is important that the higher expression group demonstrated a worse clinical outcome in overall survival and progression-free survival prediction [hazard ratio (HR) = 1.19 and 1.24, respectively]. For general chemical drug response based on SPARC expression level, the bar plot shows that the non-response group had a higher expression than did the response group (Fig. 2E). Receiver operating characteristic (ROC) analysis showed an AUC = 0.585, with p = 5 $\times$ 10${}^{-4}$ (Fig. 2F).

Fig. 2.

SPARC as a biomarker of ovarian cancer. (A) Bar plot of SPARC expression comparing ovarian carcinoma (OV) tumor and normal tissue. (B) A violin plot of SPARC expression in ovarian carcinoma major stages. Pr ($>$ F), the association between the significance probability value and the F Value. (C) The Kaplan–Meier (KM)-plot result of OS using SPARC (212667_at) in ovarian cancer. (D) The KM-plot result of progression-free survival (PFS) using SPARC (212667_at) in ovarian cancer. (E) Box plot of SPARC gene expression level in both non-responder and responder group of chemical treatments. (F) The receiver operating characteristic (ROC) plot of SPARC in predicting general chemical drug response. TCGA/GTEx, The Cancer Genome Atlas/Genotype-Tissue Expression; OS, overall survival; PFS, progressions-free survival; HR, hazard ratio; AUC, Area Under Curve; TPR, true positive rate; TNR, true negative rate.

For specific drug-response-based survival analysis, including Platin, taxane, cocetaxel, paclitaxel, and gemcitabine, SPARC gene expression showed a consistent increase in the non-responder group and a worse clinical outcome. Platin-treated patients (Fig. 3A): AUC = 0.594 with p = 1.7 $\times$ 10${}^{-4}$, and overall survival with a logrank p = 2.2 $\times$ 10${}^{-6}$. Taxane-treated patients (Fig. 3B): AUC = 0.594 with p = 1.2 $\times$ 10${}^{-3}$, and overall survival with a logrank p = 8.3 $\times$ 10${}^{-7}$. Docetaxel-treated patients (Fig. 3C): AUC = 0.726 with p = 0.1, and overall survival with a logrank p = 7.3 $\times$ 10${}^{-2}$. Paclitaxel-treated patients (Fig. 3D): AUC = 0.53 with p = 0.31, and overall survival with a logrank p = 3.4 $\times$ 10${}^{-2}$. Gemcitabine-treated patients (Fig. 3E): AUC = 0.531 with p= 0.39, and overall survival with a logrank p = 6.9 $\times$ 10${}^{-1}$.

Fig. 3.

SPARC as drug response indicator of ovarian cancer. (A) A box plot of SPARC gene expression in both the non-responder and responder group of Platin treatment. ROC plot and drug based overall survival KM-plot (Best p-value). (B) Taxane. (C) Docetaxel. (D) Paclitaxel. (E) Gemcitabine.

For SKOV3 cells, the CCK-8 test demonstrated that overexpression of SPARC at Day 3 significantly increased cell proliferation (Fig. 4A). For the ID8 cell line, overexpressed SPARC produced a significantly increased proliferation at Days 2 and 3 (Fig. 4B). The cell migration ability was evaluated by Transwell, and the wound-healing assay showed that SPARC overexpression promoted cell movement in both SKOV-3 and ID8 ovarian cancer cells (Fig. 4C,E,F). For cell-colony formation and ovarian-cancer-spheroid formation, increased SPARC level enhanced the cell colony and spheroid-formation numbers in both SKOV-3 and ID8 cells (Fig. 5A,B). For cell-to-matrix adhesion, increased SPARC raised extracellular matrix collagen I, fibronectin, and Poly-L-lysine in ovarian cancer cell lines SKOV-3 and ID8 (Fig. 5D,E). Those in vitro experiments validated the GO biological process enrichment result including terms as “cell adhesion, positive regulate cell migration, positive regulation of ERK1 and ERK2 cascade, cell migration, and cell-matrix adhesion” (Fig. 4D). For KEGG enrichment, the terms “focal adhesion, ECM-receptor interaction, cell adhesion molecules, MAPK signaling pathway, and regulation of actin cytoskeleton” were indicated by those cell phenotypes (Fig. 5C).

Fig. 4.

SPARC promoting ovarian cancer cell proliferation and migration. (A) Cell Counting Kit-8 (CCK-8) dot plot of cell proliferation in SKOV3 cells. *indicates p 0.05. (B) CCK-8 dot plot of cell proliferation in ID8 cells. ***indicates p 0.001. (C) Transwell assay showing increasing SPARC (over-expression (OE) group) enhanced cell migrations rate under starving status in both SKOV-3 and ID8 cells. (D) Bubble plot of Gene Ontology (GO) biological process analysis. (E,F) The wound healing assay showed results consistence with Transwell assay results in evaluation of SKOV-3 and ID8 cell migration by increasing SPARC expression level at 24 h. Yellow line indicates starting point. *indicates p 0.05, **indicates p 0.01, ***indicates p 0.001, ****indicates p 0.0001.

Fig. 5.

SPARC regulating ovarian cancer spheroids formation and adhesion to matrix. (A) Colony-formation assay results: SPARC increase promoted colony numbers in SKOV-3 and ID8 cells. (B) Spheroid-formation assay supported SPARC level was associated with cell survival during metastasis process forming mediated spheroids. (C) Bubble plot of Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment demonstrated major that a signaling pathway was involved. (D,E) Cell-to-matrix adherence test showed that higher SPARC level regulated SKOV-3 and ID8 cells adherence capability to extracellular matrix. *indicates p 0.05, **indicates p 0.01, ***indicates p 0.001, ns = not significant.

SPARC level was significantly negatively related to B cell infiltration, with a Rho = –0.306 and p = 8.8 $\times$ 10${}^{-7}$. Similarly, SPARC expression was also negatively associated with infiltration of mast cells and CD4${}^{+}$ T cells, with Rho = –0.128 and Rho = –0.128, and p = 4.44 $\times$ 10${}^{-2}$ and p = 4.30 $\times$ 10${}^{-2}$, respectively. In contrast, higher SPARC level was positively related to CD8${}^{+}$ T cell, macrophage, cancer associated fibroblast, NK cell, myeloid dendritic cell, and neutrophil infiltration levels (Fig. 6). For EMT signatures, SPARC level was strongly associated with CD274, CTNNB1, FN1, MMP2, MMP9, VIM, and ZEB1, indicating a metastasis possibility. However, there were no significant effects for CDH1, CDH2, and MUC1 (Fig. 7). In addition, SPARC was related to immune checkpoint gene CTLA4 (Rho = 0.25; p = 1.09 $\times$ 10${}^{-5}$) and PDCD1 (Rho = 0.191; p = 8.44 $\times$ 10${}^{-4}$).

Fig. 6.

SPARC expression related correlation with specific immune cell type infiltration. Rho = Spearman’s correlation.

Fig. 7.

The correlation dot plot between SPARC level and epithelial-mesenchymal transition (EMT)/immune checkpoint signature markers. Rho = Spearman’s correlation.

Aside from the ABL- (Abelson Murine Leukemia Viral Oncogene Homolog) targeted drug AZD0530 (Fig. 8), all other drug effect predictions showed negative results, including paclitaxel (n = 26). This may have been due to the small number of patients.

Fig. 8.

Target drugs response prediction result using SPARC expression level. Red group: Higher expression than median. Green group: Lower expression than median. *indicated p 0.05, ns = not significant.