SPARC accelerates biliary tract cancer progression through CTGF-mediated tumor–stroma interactions: SPARC as a prognostic marker of survival after neoadjuvant therapy

In biliary tract cancer (BTC), malignancy is strongest at the invasion front. To improve the BTC prognosis, the invasion front should be controlled. We evaluated tumor–stroma crosstalk at the tumor center and at the invasion front of BTC lesions. We investigated the expression of SPARC, a marker of cancer-associated fibroblasts, and determined its ability to predict BTC prognosis after neoadjuvant chemoradiotherapy (NAC-RT). We performed immunohistochemistry to evaluate SPARC expression in resected specimens from patients that underwent BTC surgery. We established highly invasive (HI) clones in two BTC cell lines (NOZ, CCLP1), and performed mRNA microarrays to compare gene expression in parental and HI cells. Among 92 specimens, stromal SPARC expression was higher at the invasion front than at the lesion center (p = 0.014). Among 50 specimens from patients treated with surgery alone, high stromal SPARC expression at the invasion front was associated with a poor prognosis (recurrence-free survival: p = 0.033; overall survival: p = 0.017). Coculturing fibroblasts with NOZ-HI cells upregulated fibroblast SPARC expression. mRNA microarrays showed that connective tissue growth factor (CTGF) was upregulated in NOZ-HI and CCLP1-HI cells. A CTGF knockdown suppressed cell invasion in NOZ-HI cells. Exogeneous CTGF upregulated SPARC expression in fibroblasts. SPARC expression at the invasion front was significantly lower after NAC-RT, compared to surgery alone (p = 0.003). CTGF was associated with tumor–stroma crosstalk in BTC. CTGF activated stromal SPARC expression, which promoted tumor progression, particularly at the invasion front. SPARC expression at the invasion front after NAC-RT may serve as a prognosis predictor.


Introduction
Biliary tract cancers (BTCs) are malignant tumors that originate from bile duct epithelium. BTCs are generally classified into four subgroups, based on anatomic location: gallbladder cancer, intrahepatic cholangiocarcinoma, perihilar cholangiocarcinoma, and distal cholangiocarcinoma (Nakeeb et al. 1996). The global incidence and mortality associated with BTC are increasing (Ouyang et al. 2021). The highest national age-standardized incidence rates were reported in Chile (10.38 per 100,000) and Japan (8.88 per 100,000). Curative resection is the most effective treatment for BTC, but most patients with BTC (> 65%) are diagnosed at a nonresectable stage (Valle 2010). Furthermore, the relapse rate is high among patients that undergo potentially curative surgery (Wang et al. 2011(Wang et al. , 2013. The first-line therapies for unresectable BTC are gemcitabine plus cisplatin ) and gemcitabine plus S-1 (Morizane et al. 2019). Moreover, SWOG 1815 study demonstrated that gemcitabine plus cisplatin and nab-paclitaxel combination therapy could be an effective regimen for locally advanced biliary tract cancer (Rachna et al. 1815). Although these treatments have demonstrated significant antitumor activity, in cases of first-line treatment failure, a second-line chemotherapy has not been established. Novel effective therapeutic approaches are needed to improve the clinical outcome of patients with BTCs.
Cancer-associated fibroblasts (CAFs) comprise one of the components of the tumor stroma. Via paracrine signaling, CAFs contribute to tumor proliferation and invasion, angiogenesis, and immune suppression. A growing body of evidence has suggested that cross-talk between cancer cells and CAFs affects the malignant phenotype and is a key factor in poor clinical outcomes. New treatments that target CAFs are under investigation in several ongoing clinical trials. BTCs are associated with abundant stroma and these tumor cells are strongly associated with CAFs (Aoki et al. 2022). However, CAFs display various phenotypes, because they differentiate from a variety of cells, including pericytes, mesenchymal stem cells, bone marrow cells, fibroblasts, and adipocytes. A study on intrahepatic cholangiocarcinoma revealed diverse CAF subpopulations (Affo et al. 2021).
Secreted protein acidic, rich in cysteine (SPARC) is a matricellular protein that contributes to cell-extracellular-matrix interactions and cellular functions (Alford and Hankenson 2006). SPARC plays a critical role in the desmoplastic response during tissue remodeling, and it is associated with cancer progression. Furthermore, studies in different cancer types showed that high SPARC expression in tumor stroma cells was a marker of poor prognosis (Infante et al. 2007;Bloomston et al. 2006). Moreover, SPARC promoted tumor-cell growth and invasiveness in vitro (Guweidhi et al. 2005;Shi et al. 2007). However, in mice carrying xenograft and orthotopic tumors, the tumors grew faster in SPARC-null mice than in wildtype mice. Thus, the effect of SPARC on cancer has been controversial.
We previously showed that BTC cells at the tumor invasion front possessed high malignant potential (i.e., they underwent the epithelial mesenchymal transition via SMAD4 signaling; Yamada et al. 2013) or acquired immune tolerance via inflammatory cytokine signaling (Kinoshita et al. 2020). In the current study, we hypothesized that, in BTC, stromal SPARC expression would be higher at the tumor invasion front than at the lesion center and that high stromal SPARC expression might independently predict a poor prognosis.

Resected specimens and patient characteristics
We collected clinical data and tumor specimens from 136 patients with BTC treated between 2005 and 2020 at Osaka University Hospital or Osaka Cancer Institute. The use of resected samples was approved by the Human Ethics Review Committee of the Graduate School of Medicine, Osaka University (No. 20493). Written informed consent was obtained from all patients included in the study.

Immunohistochemistry
Paraffin blocks of resected specimens were cut into 3.5-µm sections, deparaffinized with xylene and ethanol, and bathed in citrate buffer at 95 °C for 40 min for antigen retrieval. Endogenous peroxidase activity was inhibited by immersing sections in a 3.0% hydrogen peroxidase solution in methanol for 20 min. Non-specific binding sites were blocked in 1 mol/L PBS with 10% normal rabbit serum from the Avidin/Biotin Blocking Kit (Vector Laboratories Inc., Burlingame, California, USA). Then, sections were incubated at 4 °C overnight with 10 µg/mL polyclonal goat anti-SPARC antibody (AF941, R&D systems, Minneapolis, Minnesota, USA). After washing with PBS, sections were loaded with secondary antibody from the Avidin/Biotin Blocking Kit (Vector Laboratories Inc) for 1 h. Sections were stained with avidin-biotin complex reagents (Vector Laboratories Inc) and 3,3′-diaminobenzidine (DAB) and counter-stained with hematoxylin. Finally, sections were dehydrated in graded concentrations of ethanol and xylene and mounted on slides.

SPARC immunohistochemistry
SPARC immunostaining was evaluated separately in tumor cells and stroma cells. Staining intensities were scored as absent = 0; weak = 1; moderate = 2; or strong = 3. The area stained was quantified as the ratio of the stained area to the entire section area (e.g., a section with 75% stained cells would be scored as 0.75). The overall staining score was calculated as the intensity score × quantity score. Figure 1a-h demonstrates representative samples of SPARC immunostaining in resected specimens. Data for tumor cells represent the sum of scores in four fields (× 200) and data for stroma cells represent the sum of scores in two fields (× 100). Both tumor cells and stroma cells were evaluated at the lesion center and at the invasion front ( Fig. 1i-k). e negative, f weak, g moderate, and h strong (all at original magnification × 100; scale bars = 100 µm). i-m Comparison of SPARC immunostaining in sections from the lesion center and at the invasion front. i A tumoral section immunostained for SPARC; insets indicate j the lesion center and k the invasion front (× 20 magnification); j, k insets in i are shown at × 200 magnification. l, m Line graphs show comparisons of SPARC scores at the lesion center and the invasion front in l tumoral cells and m stromal cells in sections from 92 patients with BTC treated with upfront surgery. Solid lines indicate that the score at the invasion front was higher than at the lesion center. Dotted lines indicate that score at lesion center was higher than at the invasion front

Cell culture
The human gallbladder carcinoma cell line, NOZ (RRID: CVCL_3079), was purchased from the JCRB cell bank (Tokyo, Japan). The NOZ cell line was established from ascites of human gallbladder carcinoma by Homma et al. (1988). A human intrahepatic cholangiocarcinoma cell line, CCLP1 (RRID: CVCL_0205) was kindly provided by Dr. Gregory J. Gores (Mayo Clinic, Rochester, Minnesota, USA). A normal human dermal fibroblast cell line (NHDF) was purchased from Ronza, (Tokyo, Japan). All cell lines were grown in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 100 units/mL penicillin and streptomycin at 37 °C in a humidified incubator with 5% CO 2 . Cell lines were sub-cultured every 3-4 days and used within 10 passages. All experiments were performed with mycoplasmafree cells as tested with MycoAlert (Ronza, Tokyo, Japan).

Highly invasive BTC clones established with the in vitro selection method
In vitro selection was carried out to establish HI clones of NOZ and CCLP1 BTC cell lines, as described previously (Takagi et al. 2020). Briefly, highly invasive cells were selected in Corning BioCoat™ Matrigel Invasion chambers (Corning, NY, USA). Each chamber included a companion plate and a cell culture insert that comprised an 8-μm-pore polyethylene terephthalate membrane coated with a thin layer of Matrigel basement-membrane matrix. We seeded 1.0 × 10 5 cells on each insert, and chambers were incubated at 37 °C, 5% CO 2 for 24 h. Thereafter, cells that had migrated through the insert pores into the opposite side of the chamber were collected and cultured in a 10-mm dish to 80% confluency. This procedure was repeated six times to establish NOZ-HI and CCLP1-HI clones. mRNA microarray analysis mRNA microarray assays were performed with a 3D-Gene Human Oligo chip, 25k (Toray Industries, Tokyo, Japan). We compared gene expression between the NOZ-parent and NOZ-HI cells. We determined that RRM1 and dCK mRNA levels were generally upregulated in the NOZ-HI clone. Normalized data were analyzed to determine whether gene expression was upregulated or downregulated.

Invasion and migration assays
Cell invasion was assayed in Corning BioCoat™ Matrigel Invasion chambers (24-well, 8-μm pores, Corning) according to the manufacturer's protocol. Briefly, 2.5 × 10 4 cells were seeded in triplicate on the cell culture inserts. After 24 h, cells that had passed to the opposite side of the membrane were fixed with 100% methanol and 1% toluidine blue. Eight microscopic fields were randomly selected for cell counting.
Cell migration was assayed with TC-inserts (SARSTEDT, Numbrecht, Germany) according to the manufacturer's protocol. Briefly, 1.5 × 10 4 cells were seeded in triplicate on the cell culture inserts. After 24 h, cells that had passed to the opposite side of the membrane were fixed with 100% methanol and 1% toluidine blue. Eight microscopic fields were randomly selected for cell counting.

Cell proliferation assay
Cell proliferation was assayed with the Cell Counting kit-8 (Dojindo laboratories, Kumamoto, Japan), according to the manufacturer's protocol. Briefly, NOZ-parent and NOZ-HI cells were plated at a density of 1.0 × 10 3 cells/well in a 96-well plate and treated with DMEM containing 0.5% FBS overnight. Cells were then treated with FBS-free DMEM

Fibroblast cell exposure to recombinant human CTGF protein
NHDF cells were seeded at a density of 3.0 × 10 5 cells in a six-well plate overnight. The next day, the medium was exchanged for FBS-free medium without or with 3 µg/mL rhCTGF. After 24 h or 48 h, supernatants were collected for ELISA, and cells were harvested for qRT-PCR and Western blotting.

BTC cell line exposure to gemcitabine and radiation
NOZ-HI and CCLP1-HI cells were irradiated with 2 Gy, from a 137-Cs source, at room temperature. Cells were fied by SPARC expression intensity (low or high) in stromal cells at the lesion center. c Recurrence-free survival and d overall survival, stratified by SPARC expression in stromal cells at the invasion front also exposed to complete medium containing Gemcitabine (1.5 ng/mL for NOZ-HI, 10 ng/mL for CCLP1-HI) for 48 h. Then, these treated cells were co-cultured with NHDF cells for 24 h. Afterward, NHDF cells were harvested and SPARC , FAP, COL1A1 and ACTA2 mRNA expression level were evaluated with qRT-PCR. . β-Actin served as the endogenous control. A melting curve analysis was performed to distinguish specific products from non-specific products and primer-dimers. Relative expression was calculated as the ratio of specific mRNA to endogenous β-actin mRNA in each sample.

Western blotting
Total proteins were extracted from cell lines with RIPA buffer (Thermo Fisher Scientific) containing a proteaseinhibitor cocktail (Thermo Fischer Scientific) and a phosphatase-inhibitor cocktail (Thermo Fischer Scientific). The protein concentration was measured with the Bradford method. Aliquots of total protein (10 µg) were electrophoresed on sodium dodecyl sulphate polyacrylamide gels containing 10% Tris-HCl (Bio-Rad Laboratories, Hercules, CA, USA). The separated proteins were transferred to polyvinylidene difluoride membranes (Bio-Rad Laboratories). The membranes were blocked with Blocking One (Nacalai Tesque, Kyoto, Japan), then incubated overnight at 4 °C with the following primary antibodies: anti-SPARC (Invitrogen), anti-CTGF (Cell Signaling Technology, Danvers, MA, USA), anti-αSMA (Abcam, Cambridge, UK), and antiβactin (Sigma Aldrich, Saint Louis, MO, USA). Next, membranes were washed with TBS, then incubated with HRPlinked anti-rabbit IgG and anti-mouse IgG (GE Healthcare Biosciences, Piscataway, NJ, USA) at room temperature for 1 h. Antigen-antibody complexes were detected with ECL Prime Western Blotting Detection Reagent (Cytiva, Tokyo, Japan).

ELISA
The concentration of SPARC protein secreted into NHDF cell culture supernatants was measured with a Human SPARC Quantikine ELISA Kit (R&D systems), according to the manufacturer's recommended protocol.

Gene silencing with small interfering RNA
We silenced CTGF expression by transfecting cells with CTGF-specific small interfering RNA (siRNA, Invitrogen). We also transfected unrelated siRNAs (Invitrogen) as a negative control. siRNA transfections were performed with lipofectamine RNA iMAX (Invitrogen), according to the manufacturer's protocols. Transfection efficiencies were confirmed with quantitative real-time PCR (qRT-PCR) and Western blotting.

Paracrine signaling in trans-well co-cultures
We employed trans-well co-cultures to test paracrine signaling between invading tumor cells and stromal fibroblasts. Briefly, 2.0 × 10 5 NHDF cells were seeded into 6-well plates that contained complete medium. Then, in separate plates, 1.2 × 10 5 BTC cells (parental NOZ cells and NOZ-HI cells) were seeded onto trans-well inserts with 1-µm pores (BD Biosciences, San Jose, CA, USA). For a control, we prepared trans-well monocultures, with 1.2 × 10 5 NHDF cells seeded onto trans-well inserts. The next day, the complete medium was removed, and the NHDF cells in plates and the BTC-HI cells on the inserts were washed three times with PBS. Then, the inserts were placed into the NHDF plates, and FBS-free medium was added to the plates (2 mL) and inserts (1 mL).
After 24 h of co-culturing, the NHDF cells were harvested and SPARC mRNA was analyzed with qRT-PCR.

Statistical analysis
All data are expressed as the mean ± standard deviation (SD) of at least three independent experiments.

Stromal SPARC expression is higher at invasion front than at lesion center
The clinico-pathological factors of 120 patients with BTC are summarized in Table 1. Tumoral SPARC immunostaining scores (Fig. 1l, m) were not significantly different between the central and invasion front areas (2.87 ± 3.12 vs. 3.09 ± 3.26. p = 0.475). In contrast, the stromal SPARC score at the invasion front was significantly higher than that at the central region (2.28 ± 1.70 vs. 2.69 ± 1.81. p = 0.014).

High stromal SPARC expression at the invasion front is a marker of poor prognosis
Next, we investigated associations between the intensities of tumoral and stromal SPARC expression levels and the clinico-pathological factors of 92 patients with BTC treated with upfront surgery (Supplementary table 2). First, we divided the patients into groups of low and high tumoral SPARC intensities. When analyzing cells at the central lesion and invasion front separately, the tumoral SPARC intensities were not significantly associated with any clinico-pathological factors. Next, we divided the patients into groups of low and high stromal SPARC intensities (Table 2). We found that low stromal SPARC intensity at the lesion center was significantly associated with higher invasion into the portal vein (p = 0.022), and also high stromal SPARC intensity at the invasion front was significantly associated with advanced pathological T factor (p = 0.041). These results suggested that high stromal SPARC expression contributed to the proliferation of BTC and to tumoral invasion into surrounding tissues. In Supplementary tables 3-5, we also performed similar analyzes separately for each type of biliary tract cancer (There were only 2 patients with gallbladder cancer, so the analysis was omitted). We could find only that low stromal SPARC intensity at the lesion center was significantly associated with higher invasion into the portal vein (p = 0.032). However, as each sample size was too small, we consider that these analyses may not have statistically significance. Next, we analyzed long-term survival in 50 patients with BTC treated with surgery alone, to exclude the effects of perioperative treatment on the prognosis. Figure 2 shows survival of patients stratified by low and high stromal SPARC intensities. In analyzing stromal SPARC intensities at the lesion center, we found no significant difference in survival between the low and high stromal SPARC groups. In contrast, in analyzing stromal SPARC intensities at the invasion front, the high stromal SPARC group had worse RFS and OS rates (RFS; p = 0.033, OS; p = 0.017) than the low stromal SPARC group. We performed univariate and multivariate analyses to test the association between high stromal intensity and RFS and OS (Supplementary tables 6 and 7). In the analysis for OS, high stromal SPARC intensity at the invasion front was an independent predictor of a poor prognosis. In the analysis for RFS, although there was not significant difference, high stromal SPARC intensity at the invasion front was a predictor of a poor prognosis.
To summarize, these immunohistochemical studies showed that SPARC expression was upregulated in stromal cells at the invasion front. Additionally, high stromal SPARC expression at the invasion front predicted a poor prognosis in patients with BTC.

Highly invasive BTC cell lines
Based on our immunohistochemical findings that stromal SPARC expression was higher at the invasion front than at the lesion center, we hypothesized that paracrine signaling might be different between the tumor center and the tumor invasion front. To investigate differences in signaling, we performed in vitro selection to establish highly invasive (HI) BTC cell lines, which imitated BTC cells at the tumor invasion front. We established both NOZ-HI and CCLP1-HI cell lines by repeatedly cloning the most mobile NOZ and CCLP1 cells from the parent BTC cell lines. To confirm the invasive abilities of NOZ-HI and CCLP1-HI cells, we performed invasion assays for the parent cells, the HI cells after three passages, and HI cells after six passages. As shown in Fig. 3a-d, after six passages, the invasive abilities of both NOZ-HI and CCLP1-HI cells were higher than those of their respective parent cells.

3
Next, we performed immunocytochemistry and qRT-PCR to investigate whether co-culturing fibroblasts (NHDFs) with HI BTC cells affected NHDF SPARC expression. We found that SPARC mRNA expression in NHDFs cells co-cultured with NOZ-HI was significantly higher compared to those co-cultured with NOZ-parental cells or those cultured in monocultures with other NHDF cells (Fig. 3e). Furthermore, an immunocytochemical analysis confirmed that SPARC protein expression was highest in NHDFs after co-culture with NOZ-HI, compared to the other two groups (Fig. 3f, g). Connective tissue growth factor (CTGF) is upregulated and associated with invasiveness in highly invasive BTC cells (BTC-HI), and recombinant human CTGF (rhCTGF) protein promotes SPARC expression in fibroblasts. a Heat map of upregulated genes in NOZ-HI cells shows the 28 genes associated with cytokines or growth factors. b qRT-PCR results show CTGF mRNA expression in NOZ-parental (NOZ-p), NOZ-HI, CCLP1-parental (CCLP1-p), and CCLP1-HI cells. c Western blots show elevated CTGF protein expression in NOZ-HI and CCLP1-HI cells compared to NOZ-p and CCLP1-p cells. d Invasion assay results show the suppression of invasion, when NOZ-HI cells were transfected with small interfering CTGF (siCTGF). e qRT-PCR results show elevated SPARC and ACTA2 mRNA expression in normal human dermal fibroblasts (NHDFs) after exposure to rhCTGF protein. f Western blot shows elevated SPARC and αSMA protein expression in NHDFs after exposure to rhCTGF protein. g ELISA results show elevated SPARC concentrations in supernatants of NHDFs after exposure to rhCTGF for 24 h ◂ Exogenous rhSPARC protein promotes cancer cell growth SPARC was reported to be expressed mainly in the stromal components of malignant tumors. Thus, we compared SPARC expression in fibroblast cells to SPARC expression in BTC tumor cells. Consistent with previous studies, we found that SPARC expression was much lower in BTC cell lines than in fibroblast cells ( Supplementary Fig. 1). Next, we used recombinant human SPARC protein (rhSPARC) to explore how SPARC protein secreted from stromal cells affected the function of BTC cells. We exposed NOZ cells to rhSPARC and performed proliferation, invasion, and migration assays. As shown in Fig. 3h, i, the NOZ-parental and NOZ-HI cells treated with rhSPARC displayed significantly increased proliferation compared to the corresponding untreated control groups. Furthermore, proliferation increased with increasing rhSPARC concentrations. In contrast, the invasion and migration assays did not show any significant differences between the treated and control groups (Fig. 3j, k).

CTGF expression is upregulated in NOZ-HI cells
Based on the results from the co-culture assays, we hypothesized that cytokines or growth factors secreted from BTC cells might upregulate SPARC expression in stromal cells. To identify factors that could upregulate SPARC expression in fibroblast cells, we performed mRNA microarray analyses of NOZ-parent and NOZ-HI cells, and we evaluated associations between SPARC expression and the expression of cytokine-or growth factor-associated genes. Figure 4a shows a heat map of 28 genes identified as cytokines or growth factors that were upregulated by > 1.5-fold in NOZ-HI cells. These genes included interferon-stimulated exonuclease gene 20 (ISG20), interferon-related developmental regulator 1 (ISG20L2), interferon regulatory factor 1 (IRF1), interferon-related developmental regulator 1 (IFRD1), transforming growth factor beta 1-induced transcript 1 (TGFβ1I1), vascular endothelial growth factor A (VEFGA), CTGF, and fibroblast growth factor binding protein 1 (FGFBP1). Among these genes, we focused on CTGF, because CTGF was previously reported to play a pivotal role in tumor-stroma paracrine signaling and accelerated cancer progression (Makino et al. 2018;Lobe et al. 2021). We confirmed CTGF expression in BTC cells. In addition, CTGF mRNA was expressed at higher levels in NOZ-HI and CCLP1-HI cells than in their corresponding parent cells (Fig. 4b). A Western blot analysis revealed that CTGF protein expression was also elevated in NOZ-HI and CCLP1-HI cells (Fig. 4c). Next, we silenced CTGF expression by transfecting cells with CTGF-specific siRNAs to determine whether CTGF was associated with BTC cell invasiveness. We found that the number of invasive cells was reduced in cells transfected with CTGF siRNA (Fig. 4d).

Exogenous rhCTGF protein promotes SPARC expression in fibroblasts
To determine whether CTGF might contribute to upregulated SPARC expression in fibroblasts, we exposed NHDF cells to rhCTGF protein and examined SPARC expression with qRT-PCR, western blotting, and ELISAs. At the same time, we investigated changes in αSMA expression, because αSMA is one of the most common markers of CAFs. The qRT-PCR results revealed that SPARC and ACTA2 expression were elevated in the rhCTGF treatment group. Western blotting showed that SPARC and αSMA protein expression were elevated in the rhCTGF treatment group. The ELISA results showed SPARC protein concentrations in the supernatants after exposure to rhCTGF for 24 h were higher than control group. These results showed that SPARC mRNA expression was elevated after exposure to rhCTGF ( Fig. 4e-g).

Expression of CTGF and SPARC in resected specimens
We evaluated relation between CTGF and SPARC immunostaining in resected specimens of BTC. In five SPARC positive specimens, positive reactivity for CTGF was observed in three (60%), whereas, in five SPARC negative specimens, positive reactivity for CTGF was observed in one (20%). We showed representative samples in Supplementary Fig. 2.

Stromal SPARC expression decreased after NAC-RT
Finally, we investigated whether SPARC expression was regulated in resected BTC specimens from patients that c SPARC scores in stromal cells at the invasion front in resections from patients treated with upfront surgery or different NAC regimens. d, e qRT-PCR results show CAF marker expression induced in normal human dermal fibroblasts (NHDFs) cocultured with highly invasive (-HI) cell lines, d NOZ-HI or e CCLP1-HI, and/or exposed to Gemcitabine and radiation, as indicate underwent neoadjuvant therapy. We compared stromal SPARC immunohistochemical scores in tissues from the upfront surgery group (n = 92), the NAC-RT group (n = 21), and the NAC group (n = 23). The average stromal SPARC scores measured at the lesion center were 2.28 ± 1.70 for upfront surgery, 1.95 ± 1.53 for NAC-RT, and 2.28 ± 2.36 for NAC. There was no significant difference among the three groups (Fig. 5a). The average stromal SPARC scores measured at the invasion front were 2.72 ± 1.79 for upfront surgery, 1.29 ± 1.31 for NAC-RT, and 2.13 ± 1.84 for NAC. The SPARC score in NAC-RT group was significantly lower than that in upfront surgery group (p = 0.002; Fig. 5b). Furthermore, we divided the NAC group into three subgroups, based on the neoadjuvant regimen: the GC group received Gemcitabine + Cisplatin, the GCnP group received Gemcitabine + Cisplatin + Nab-paclitaxel, and the GCS group received Gemcitabine + Cisplatin + S1. When we compared the stromal SPARC scores measured at the invasion front among these groups, we found no significant difference between the upfront surgery group and any of the three NAC subgroups (Fig. 5c).
To verify the immunohistochemical result that showed reduced SPARC expression after NAC-RT, we performed an in vitro study. We exposed BTC-HI cells to Gemcitabine and radiation, then co-cultured them with NHDF cells. Next, we evaluated mRNA expression of CAF markers in the NHDF cells. Figure 5d, e shows that in NHDF cells cocultured with NAC-RT-treated NOZ-HI and CCLP1-HI, cells displayed significantly reduced the SPARC, COL1A1, and FAP mRNA expression levels compared to NHDF cells co-cultured with untreated BTC-HI cells. However, ACTA2 mRNA expression was significantly higher in NHDF cells exposed to the NAC-RT-treated BTC-HI cells, compared to untreated BTC-HI cells.

Discussion
In this study, we reported that SPARC and CTGF were associated with paracrine signaling between BTC tumors and the stroma. Additionally, this was the first study to show that SPARC expression was suppressed in human samples after NAC-RT. Our main findings were (1) in resected specimens, stromal SPARC was highly expressed at the invasion front and high stroma SPARC expression was a marker for a poor prognosis; (2) exogenous SPARC protein promoted BTC cell growth; (3) CTGF expression was upregulated in established BTC-HI cells, and knocking down CTGF in HI cells suppressed their ability to invade; (4) exogenous CTGF upregulated SPARC expression in fibroblast cells; and (5) NAC-RT suppressed SPARC expression in stromal cells at the invasion front.
This study showed that although high stromal SPARC expression at the invasion front was an independent marker of a poor prognosis, high stromal SPARC expression at the lesion center was not associated with the prognosis. Previous studies reported that high SPARC expression in tumor stroma cells indicated a poor prognosis in patients with several types of cancer (Infante et al. 2007;Bloomston et al. 2006). However, to our knowledge, the current study was the first to investigate stromal SPARC levels separately at the tumor invasion front and at the lesion center. Additionally, exogenous SPARC protein promoted the proliferation of both NOZ-parental and NOZ-HI cells.
CTGF is a member of the cell communication network family. Due to its unique structure, CTGF plays a multifunctional regulatory role by binding to both ligands and receptors (Jun and Lau 2011;Kular et al. 2011). In cancer, CTGF contributes to tumor malignancy by promoting cancer cell proliferation (Kim and Son 2019), migration (Chen et al. 2007), invasion, and the epithelial-mesenchymal transition (Liu et al. 2006). Additionally, CTGF contributes to building a fibrotic and inflammatory-tumor microenvironment, which supports cancer progression (Chen et al. 2007;Jiang et al. 2013;Kim et al. 2020). We showed that CTGF mRNA expression was upregulated in two BTC-HI cell lines and that knocking down CTGF suppressed the ability to invade. These results suggested that CTGF increased the malignant potential of BTC cells. Additionally, CTGF has been reported to mediate interactions between tumor and stromal cells, and it accelerated tumor progression in hepatocellular carcinoma (Makino et al. 2018;Mazzocca et al. 2010) and cholangiocarcinoma (Lobe et al. 2021). Those reports showed that CTGF derived from tumor cells activated stromal cells, which resulting in cancer progression. A phase I/II trial was performed to test the efficacy of Pamrevlumab, a CTGF neutralizing antibody, in combination with gemcitabine and nab-paclitaxel, in treating locally advanced pancreatic cancer (Picozzi et al. 2020). This trial showed that neoadjuvant chemotherapy with Pamrevlumab enhanced resection rate in locally advanced pancreatic cancer. The phase III study is also ongoing (NCT 03941093). The present study showed that CTGF mRNA expression was higher in BTC-HI cells than in parental BTC cells, and that adding exogeneous CTGF upregulated SPARC expression in fibroblast cells. In addition, exposure to exogeneous CTGF increased the expression of αSMA mRNA, which is a marker of CAFs. These results suggested that CTGF secreted from BTC tumor cells could activate fibroblasts.
Previous reports showed that the rates of curative resection were 68.2% in gallbladder cancer and 68.1% in bile duct cancer (Miyakawa et al. 2009). These rates are much lower than the curative resection rates for other gastrointestinal cancers. The low curative resection rate may be due to the proximity to major vessels and the strong infiltrative nature of BTC. A few studies reported that patients treated with chemoradiotherapy (CRT) before surgery showed a high CRT response rate, and subsequently, a high R0 resection rate for advanced BTC (McMasters et al. 1997;Sumiyoshi et al. 2018). We also previously reported that patients treated with NAC-RT (3 cycles of full dose gemcitabine and 50-60 Gy radiation) had significantly longer RFS and OS times than patients without NAC-RT treatment (Kobayashi et al. 2017), and the curative resection rate was 90% in patients with initially resectable BTC (Kobayashi et al. 2016). The current study demonstrated that stromal SPARC expression in patients treated with NAC-RT was significantly lower than in patients treated with upfront surgery. Moreover, the RFS and OS times in patients treated with NAC-RT were significantly longer than those in patients treated with upfront surgery. A previous report studied the presence of SPARC in fibroblasts from both primary resected samples and neoadjuvant-treated samples of esophageal carcinomas. They showed that SPARC was associated with malignant factors, such as a higher pT category, lymph node metastasis, lymphatic invasion, and perineural invasion (Galván et al. 2020). However, several other studies have reported that radiation induced changes in the tumor microenvironment that favored tumor growth and invasion by activating CAFs. Moreover, several preclinical studies showed that radiotherapy induced CAF activation, which promoted tumor invasion and spread.
The current report had several limitations. First, our cohort was small for the immunohistochemical study. In particular, both the NAC (n = 23) and the NAC-RT (n = 21) groups were small. Future studies should include more patients to investigate whether SPARC might serve as a predictive marker in patients treated with neoadjuvant therapies. Second, we did not determine whether a CTGF blockade could contribute to suppressing cancer cell malignancy.

Conclusion
In conclusion, we showed that CTGF contributed to BTC cancer cell progression by inducing SPARC expression in stromal cells at the invasion front. Furthermore, our immunohistochemical analysis results suggested that SPARC expression might serve as a predictive marker in patients treated with upfront surgery and NAC-RT.