The GEPIA2 (Tang et al., 2019), UALCAN (Chandrashekar et al., 2017), and GEO databases were used to analyze the difference of SPARC expression levels between LIHC tumor and non-tumor tissues. By using the “Expression DIY” module of GEPIA2 and the “TCGA Analysis” module of UALCAN, differential expression of SPARC mRNA was obtained. UALCAN was also used to analyze the promoter methylation level of SPARC. GSE14323 (Levy et al., 2016), GSE14520 (Roessler et al., 2010; Wang et al., 2021), and GSE121248 (Wang et al., 2007) datasets contain the gene expression profiling of LIHC tumor and their matched normal adjacent tissues, which were used to validate the analysis results based on GEPIA2 and UALCAN. GSE58208 and GSE49515 (Shi et al., 2014) datasets contain the gene expression profiling of PBMC from LIHC patients and healthy subjects, which were used to evaluate the difference of SPARC expression in PBMC.

Formalin-fixed tissue microarray which contained 48 LIHC and 48 matched adjacent controls (D097Lv01, http://www.bioaitech.com/chip-design/13c0de19d7e94d4a99952af585089847.html) was purchased from Bioaitech (Xi’an, China). A 1:200 dilution of anti-SPARC mAb (D10F10, Cell Signaling Technology) was used as the primary antibody. PANNORAMIC and CaseViewer 2.4 software (3DHISTECH, Hungary) were used for image acquisition and analysis. IHC analysis was conducted by two independent pathologists. The intensity was scored as 0 (negative), 1 (weak), 2 (moderate), and 3 (strong). The extent of positive cells less than 5% was scored as 0, 6%–25% was scored as 1, 26%–50% was 2, 51%–75% was 3, and >75% was 4. The final score was determined by multiplying the intensity and extent positivity scores. The detail information of tissue microarrays and IHC scores is available in Supplementary Table S1.

The KM plotter bioinformatic tools (Menyhárt et al., 2018) were used to evaluate the potential prognostic value of SPARC expression in LIHC patients. We chose the “auto select best cutoff” model, which means that all possible cutoff values were computed and the best performing threshold was used as a cutoff.

Ethical approval was obtained from the Ethics Committee of Tangdu Hospital, Fourth Military Medical University (TDLL-202106-04). Informed consent was exempted for this study. Serum from 39 LIHC patients and 32 healthy controls was collected. PBMC from 10 LIHC patients and 10 healthy individuals was collected. PBMC was obtained by using Ficoll separation according to the operation manual (Dakewe Biotech, China).

Serum SPARC levels were measured by the ELISA assay (ELH-SPARC-1, RayBiotech, United States). In brief, 100 μl diluted serum were added into test wells and incubated for 2.5 h at room temperature. Then, the solution was discarded and washed 4 times, and 100 μl biotinylated antibody was added and incubated for 1 h. After washing 4 times, streptavidin was added and incubated for 45 min. The TMB reagent was added into wells, incubated for 30 min, followed by adding stop solution and absorbance determination at 450 nm.

HepG2 cells were cultured in the DMEM (Gibco, United States) plus 10% heat-inactivated fetal bovine serum (FBS, Sijiqing Biotec, China) at 37°C with 5% CO₂ in a humidified incubator. The SPARC coding gene was cloned into the pcDNA3.1 + expression vector. Then the SPARC/pcDNA3.1+ and control plasmids were transfected into HepG2 cells using Lipofectamine 2000 (Invitrogen, United States) according to the manufacturer’s instruction. Briefly, HepG2 cells were seeded in 6-well plate and cultured until 70–80% confluence. Premixed lipofection and plasmid DNA (10 μl: 4 μg) were added into the wells and incubated for 24 h. The stable cell models were screened by G418. The expression levels of SPARC in cell models were investigated by the quantitative real-time PCR (qRT-PCR). The primers used for qRT-PCR are listed in Table 1.

Cell proliferation was investigated by the CCK8 assay. In short, cells were seeded into 96-cell plate at a density of 2 × 10³ cells/well, with six replicate wells per group. After 24 and 48h, a 10 µl CCK8 reagent (KeyGEN BioTECH, Nanjing, China) was added, followed by incubation for 2 h. The absorbance value was measured at 450 nm by using the microplate reader (BIO-TEK Epoch, United States).

Flow cytometry (FCM) was used to analyze the cell cycle and apoptosis. The cell cycle was analyzed by ethanol-fixed cells stained with propidium iodide (PI) in buffer containing RNase A. The DNA content was assessed by using the FACS calibur flow cytometer (BD Biosciences). Cell apoptosis was detected using the annexin V–FITC apoptosis detection kit (KeyGEN BioTECH, Nanjing, China). Briefly, cells were collected and washed with PBS twice, and stained for 15 min at room temperature with annexin V-FITC and (PI). Following that, cell apoptosis was examined using FACS calibur. The percentage of apoptotic cells was calculated with CellQuest 6.0 (BD Biosciences).

All procedures of animal experiments were processed according to animal welfare guidelines approved by Air Force Medical University’s Institutional Animal Care and Use Committee. Male BALB/c nude mice were used. The effect of high SPARC expression on tumor growth in vivo was assessed by subcutaneous injection of 2×10⁶ HepG2 ^SPARC or HepG2 ^NC cells into a BALB/c nude mouse. Nude mice were purchased from Animal Experiment Center, Fourth Military Medical University (age: 6–8 weeks; body weight: 20–25 g). Tumor size was measured every 2 days with a vernier caliper. The volume of tumor was calculated according to the formula: a×b ² × 0.5 (a, largest diameter and b, perpendicular diameter).

The gene set which was positively or negatively correlated to SPARC in LIHC was screened by the LinkedOmics analysis tool (http://www.linkedomics.org) based on TCGA data (Vasaikar et al., 2018). The gene ontology (GO) and KEGG pathway analyses were performed by DAVID Bioinformatics Resources (https://david.ncifcrf.gov/tools.jsp) (Huang et al., 2009). R Bubble diagrams were used for visualization of the enriched GO term and KEGG pathway.

Student’s t-test was used to analyze the difference of IHC scores between tumor tissues and controls. Student’s t-test was used to analyze the difference of SPARC expression between the LIHC and control, both in tissues and in PBMC. Serum SPARC levels were expressed as the median (inter quartile range, IQR). Wilcoxon’s test was used to analyze the difference of serum SPARC levels between the LIHC and healthy controls. Student’s t-test was used to analyze the difference of the cell proliferation rate between different groups. p < 0.05 was considered to be statistically significant.

We first analyzed the SPARC expression levels in LIHC patients based on the TCGA and GEO databases. GEPIA2 and UALCAN analyses showed an increased mRNA expression level of SPARC (Figures 1A,B). And moreover, the results were validated in GSE121248, GSE14520, and GSE14323 datasets (Figures 1C–E). These data showed that SPARC expression levels were higher in LIHC tumor tissues than those in control normal tissues. Notably, there was no significance of the SPARC expression between LIHC patients with different stage or grade (Figures 1F,G). And moreover, the UALCAN analysis showed a lower promoter methylation level of SPARC in LIHC than that in normal controls (Figure 1H), which indicated demethylation might contribute to an increased expression of SPARC in LIHC.

Furthermore, the IHC analysis of tissue microarrays was used to investigate the protein expression levels of SPARC in LIHC tissue. The IHC analysis revealed an increased level of the SPARC protein in LIHC tissues compared to adjacent non-tumor tissues (Figure 1I). Taken together, these results demonstrated that SPARC was increased in LIHC tissues.

Furthermore, the relationship between the SPARC expression and prognosis of LIHC patients was analyzed by using the KM plotter. The results showed that there was no significant correlation between the SPARC mRNA expression and LIHC patients’ overall survival (OS) (Supplementary Figure S1A), whether or not the patients were with hepatitis virus (Supplementary Figures S1B,C) or alcohol consumption (Supplementary Figures S1D,E). However, in LIHC patients treated with sorafenib, although the number of patients was small (n = 29), high SPARC expression showed a favorable prognosis (HR = 0.19, logrank p = 0.0053; Supplementary Figure.S1F).

SPARC was expressed in a variety of cells, including PBMC. The human protein atlas (HPA) database showed that SPARC was expressed in total PBMC, monocytes, neutrophil, basophil, eosinophil, etc (Figure 2A). Thus, in addition to analyzing the SPARC expression in tumor tissues, we also analyzed the difference of SPARC mRNA expression levels in PBMC from LIHC patients and normal controls based on GSE58208 and GSE49515 datasets. These datasets showed a higher SPARC expression in LIHC’s PBMC than in control’s PBMC (Figures 2B,C).

As SPARC is a secreted protein, we further investigated the SPARC levels in serum from LIHC patients. As shown in Figure 2D, the serum SPARC levels were significantly decreased in LIHC patients compared to healthy controls [86.19 (IQR: 49.16-140.46) vs. 293.53 (IQR: 245.97-356.66) ng/ml, p < 0.001)]. Furthermore, by using qRT-PCR, we detected the SPARC mRNA expression in PBMC from LIHC patients and healthy individuals. The results showed a lower expression level of SPARC in LIHC’s PBMC than that in healthy individuals (Figure 2E), which is consistent with our results of serum SPARC detection. However, these data were inconformity with the published data in GSE58208 and GSE49515 datasets.

We investigated the endogenous SPARC overexpression on HepG2 cell proliferation in vitro. First, the SPARC overexpressed cell model was constructed and named as HepG2 ^SPARC . The higher expression level of SPARC in HepG2 ^SPARC than that in HepG2 ^NC control cells was identified by qRT-PCR (Figure 3A). Then, by using the CCK8 assay, HepG2 ^SPARC cells showed an enhanced proliferation rate compared to HepG2 ^NC (Figure 3B). And moreover, the FCM analysis showed that there was a significant decrease of HepG2 ^SPARC cells in the G1 phase, while there was an increase in G2 and S phases (Figure 3C). There was no significant difference of cell apoptosis between HepG2 ^SPARC and HepG2 ^NC cells (Figure 3D).

In addition to the cell experiments in vitro, we further investigated the effect of SPARC overexpression on tumor growth in vivo. By using nude xenografted models, we found that SPARC overexpression promoted liver cancer growth (Figure 3E). Taken together, these results demonstrated a promoted effect of high SPARC expression on liver cancer growth.

Furthermore, we evaluated the effects of exogenous SPARC treatment on the HepG2 cell biological behavior. The gradient concentration of recombination SPARC protein was added into the cell cultured medium. Then, CCK8 results showed that exogenous SPARC treatment has no significant effect on HepG2 cell proliferation (Supplementary Figure S2A). The FCM analysis results showed that there was no significant effect of exogenous SPARC on cell cycle and apoptosis (Supplementary Figures S2B,C).

We used the LinkedOmics bioinformatic tool to screen the genes which positively or negatively correlated to SPARC in LIHC (Supplementary Table S2). As shown in the heatmap (Figure 4A), SPARC strongly and positively correlated to the collagen gene family (such as COL4A2, COL4A1, COL3A1, COL6A3, and COL1A2; r > 0.8, p < 0.0001), peroxidasin-PXDN (r = 0.893, p < 0.0001), and junctional adhesion molecule 3-JAM3 (r = 0.871, p < 0.0001). Furthermore, the KEGG pathway and GO enrichment analyses of SPARC association genes were performed (Figures 4B–E; Supplementary Table S2). The results showed that SPARC-correlated gene enrichment in GO_BP (biological process) was mainly related to the extracellular matrix (ECM) organization (GO:0030198), collagen catabolic process (GO:0030574), collagen fibril organization (GO:0030199), and cellular response to amino acid stimulus (GO:0071230) (Figure 4C). GO_CC (cell component) enrichment showed that these genes were mainly expressed in the extracellular matrix (GO:0031012), endoplasmic reticulum lumen (GO:0005788), proteinaceous extracellular matrix (GO:0005578), and extracellular region (GO:0005576) (Figure 4D). The GO_MF (molecular function) analysis showed an enrichment in the extracellular matrix structural constituent (GO:0005201), calcium ion binding (GO:0005509), and platelet-derived growth factor binding (GO:0048407) (Figure 4E). And moreover, the KEGG pathway analysis showed that SPARC-correlated genes were involved in protein digestion and absorption (hsa04974), ECM-receptor interaction (hsa04512), and PI3K-Akt signaling pathway (hsa04151). Taken together, these results indicated a critical role of interaction with the extracellular matrix in SPARC-promoting cancer cell proliferation.