Dense GM-CSFRα-expressing immune infiltration is allied with longer survival of intrahepatic cholangiocarcinoma patients

Human iCCA tissues

Ninety-six formalin-fixed paraffin-embedded tissues were obtained with informed consent from the specimen bank of Cholangiocarcinoma Research Institute, Khon Kaen University, Khon Kaen, Thailand. Tissues were collected from 1998–2012 and were selected under the following criteria: (1) cancers were iCCA, (2) samples were from hepatic resection, (3) clinical data were available for analysis, and (4) patients had no known history of other cancers. Perioperative deaths or patients with an overall survival less than 30 days were excluded. The study protocol was approved by the Ethics Committee for Human Research of Khon Kaen University based on the Declaration of Helsinki (HE571283 and HE611034).

Cell line and cell culture

Four CCA cell lines, KKU-055, KKU-100, KKU-213A, and KKU-213B were obtained from the Japanese Collection of Research Bioresources Cell Bank (Osaka, Japan) (Sripa et al., 2005; Sripa et al., 2020). Two metastatic cell lines, KKU-213L5 and KKU-214L5, were established as previously described (Saentaweesuk et al., 2018; Uthaisar et al., 2016). Cells were maintained in DMEM containing 10% FBS with 1% antibiotic-antimycotic in humidified 5% CO₂ at 37 °C. All cell culture-related reagents were obtained from Gibco (NY, USA).

Antibodies and reagents

The sources of antibodies were as follows: rabbit anti-GM-CSF antibody was from Novus (NB600-632; Novus, St. Louis, MO, USA), mouse anti-GM-CSFRα was from Santa Cruz (4H1; TX, USA), anti-mouse and anti-rabbit EnVision-horseradish peroxidase (HRP)-conjugated antibodies were from DAKO (K4001 and K4003; DAKO, Glostrup, Denmark), phycoerythrin (PE)-conjugated anti-GM-CSFRα was from BioLegend (4H1; San Diego, CA, USA), PE-conjugated mouse IgG was from eBioscience (P3.6.2.8.1; eBioscience, San Diego, CA, USA). ELISA MAX™ Deluxe Set Human GM-CSF and recombinant human GM-CSF (rhGM-CSF) were from BioLegend. All chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA).

Analysis of GM-CSF and GM-CSFRα mRNA expressions in CCA tissues

The GM-CSF and GM-CSFRα mRNA expressions in CCA tissues compared with their normal counterparts were investigated by Gene Expression Profiling Interactive Analysis (GEPIA, http://gepia.cancer-pku.cn/index.html), an online server for cancer and normal gene expression profiling analysis based on the Cancer Genome Atlas (TCGA) database (Tang et al., 2017). The RNA-Seq data were obtained from 36 CCA tissues and 9 adjacent normal tissues. GM-CSF and GM-CSFRα in CCA tissues were categorized into low and high expression by dichotomizing at the median.

Immunohistochemistry

Immunohistochemical staining of GM-CSF and GM-CSFRα was performed as previously described (Vaeteewoottacharn et al., 2019) using rabbit anti-GM-CSF and mouse anti-GM-CSFRα antibodies. Immunoreactivity was developed using 3, 3′ diaminobenzidine. The signals were amplified using the corresponding EnVision-HRP system. The immunohistochemical evaluations were performed by two independent evaluators. GM-CSF expression levels of CCA cells were assessed by H-score (Fitzgibbons et al., 2014), while GM-CSFRα expression was categorized by the density of GM-CSFRα-expressing ICI in the cancer area into light and moderate to dense (Wu et al., 2020).

GM-CSF ELISA

To determine GM-CSF concentration in CCA cultured media (CM), CM was collected at 24 h and determined GM-CSF concentration by GM-CSF ELISA kit following the manufacturer’s instruction. The absorbance was measured at 450 nm by an iMark microplate reader (Bio-Rad, Hercules, CA, USA).

Flow cytometry

GM-CSFRα expressions on surface CCA cell lines were evaluated by flow cytometry (LSR II flow cytometry; BD Biosciences, San Jose, CA, USA). Cells were stained by PE-conjugated anti-GM-CSFRα. PE-conjugated mouse IgG1 was used as isotype control. Data were analyzed by FlowJo software V.10.7.2 (Tree Star, Woodburn, OR, USA). GM-CSFRα expression levels were calculated as mean fluorescence intensity (MFI) of anti-GM-CSFR α-stained cells/MFI of isotype control-stained cells and were expressed as the relative MFI (Vaeteewoottacharn et al., 2019).

MTT assay

The effects of GM-CSF on KKU-055 and KKU-213B cell numbers were determined by MTT assay. Cells were seeded at 2 × 10³ cells per well onto a 96-well plate and treated with 0, 1, and 10 ng/ml rhGM-CSF for 24, 48, and 72 h. MTT solution was added to yield a final concentration of 0.5 mg/ml, and cells were incubated for an additional 3 h. Then formazan crystals were dissolved in acidified isopropanol (0.04 N HCl in isopropanol). The absorbance at 540 nm was determined by a Sunrise microplate reader (Tecan US, Inc., Morrisville, NC, USA).

Scratch-wound assay

The scratch-wound assay was performed to determine the effect of GM-CSF on CCA migration. KKU-055 and KKU-213B were seeded into 24-well plates as a monolayer until confluent, and then the scratch was made using a sterile 200-µl tip, and cells were allowed to migrate in 0, 1, and 10 ng/ml rhGM-CSF containing media for 24 h. The images of wound closure area at 0, 6, 12, 18, and 24 h were taken by Nikon’s DS-Fi2 camera connected to a Nikon Eclipse inverted microscope (Nikon Instruments Inc., Tokyo, Japan) and then calculated by ImageJ software V.1.48 as previous described (Grada et al., 2017). The percentage of wound closure was calculated as follows; (area at time 0 - area at an indicated time) × 100/area at time 0.

The estimation of immune cell infiltration

The ICI in CCA tissues was analyzed by a web-based tool, the Tumor Immune Estimation Resource (TIMER2.0, http://timer.cistrome.org/), based on computational algorithms of the TCGA database (Li et al., 2020). The association between GM-CSF and GM-CSFRα mRNA expression levels and levels of ICI, including neutrophils, dendritic cells (DCs), all macrophages, M2 macrophages, myeloid-derived suppressor cells (MDSCs), and CD8+ T cells were estimated and plotted as correlation scatter plots. The correlation coefficients between variables were presented as rho values.

Statistical analysis

Statistical analyses were performed using SPSS (V.26.0) (IBM, NY, USA) as follows: the associations between categorical variables were determined by Pearson’s chi-square tests (χ²), Kaplan–Meier survival curves were compared using the log-rank test, and multivariate analyses using Cox proportional hazard models were performed to identify independent prognostic factors for survival of patients, then the backward selection was tested to avoid confounding factors (Grant, Hickey & Head, 2019). The quantitative data were presented as mean ± SD of three independent experiments unless otherwise specified. The multiple comparisons of means were analyzed by ANOVA while comparing two groups was performed by student’s t-test using GraphPad Prism (V.7.0) (GraphPad Software, Inc., San Diego, CA, USA). A p-value less than 0.05 was considered of statistical significance.

Demographic characteristics of iCCA patients

The clinical data of 96 iCCA cases are shown in Table 1. The patient ages ranged from 33–76, with a median age of 57. Fifty-eight were males and 38 were females (a male-to-female ratio is 3:2). Histologically, 35 were papillary subtypes and 61 were non-papillary. Cancers were classified according to the 7^th edition of the American Joint Committee in Cancer (AJCC) staging system (Edge & Compton, 2010). Six-point three percent was stage I, 10.4% was stage II, 24.0% was stage III, and 59.4% was stage IV. Thirty-five of 85 cases (41.2%) had lymph node (LN) metastasis while distant metastasis was observed in 10 of 78 cases (12.8%). The survival times calculated from surgery to death ranged from 41 to 2,509 days, with a median survival time of 302.

Differential expressions of GM-CSF and GM-CSFRα were observed in iCCA tissues

To emphasize the significance of GM-CSF and GM-CSFRα on CCA progression, the mRNA expressions of both genes in CCA tissues were checked using a web-based tool, GEPIA, based on the TCGA database. GM-CSF and GM-CSFRα were highly expressed in cancer (Figs. 1A and 1D). Patients with high GM-CSF and GM-CSFRα in CCA tissues tended to have more prolonged overall survival (OS) and longer disease-free (DF) survival times than those with low expressions (Figs. 1B–1C and 1E–1F). Only higher GM-CSFRα expression in cancer tissues compared to the normal counterpart (Fig. 1D) and longer DF survival of patients with high GM-CSFRα- expressed CCA (Fig. 1F) showed statistical significance.

These data prompted determination of the GM-CSF and GM-CSFRα protein expressions in 96 iCCA tissues. GM-CSF expression was evaluated by H-score, while GM-CSFRα was graded by density. The representative figures are demonstrated in Figs. 2A and 2B. Immunohistochemistry staining revealed that CCA cells notably expressed GM-CSF (Fig. 2A), which was detectable in 95 cases (99%). On the other hand, GM-CSFRα was principally observed in non-CCA infiltrating cells localized in the peri-cancerous areas. These patterns morphologically resembled those previously described as immune cell infiltration (ICI) (Ino et al., 2013). GM-CSFRα-expressed ICI was noticeable in all cases (Fig. 2B), while GM-CSFRα-expressing CCA cells were observed in 3 cases (3.1%) (Fig. S1).

Distinct histological subtypes of iCCA expressed GM-CSF differently (Fig. 2C). Therefore, further analyses were performed as the overall iCCA and subtype-specific. The median H-scores of GM-CSF in the overall iCCA, papillary and non-papillary subtypes were 80, 120, and 70. The median H-scores were used for dividing into low and high GM-CSF expressions. From a total of 96 cases, 50 (52.1%) were high GM-CSF (H-scores ≥ 80), while 46 (47.9%) were low (H-scores <80). Although the papillary subtypes expressed higher GM-CSF than the non-papillary ones, this difference was not significant (Fig. 2C). In the papillary subtype, 17 cases of (48.6%, H-score <120) expressed low GM-CSF while 18 cases (51.4%, H-score ≥120) highly expressed GM-CSF. Among the non-papillary subtype, low and high GM-CSF expressions were observed in 27 (44.3%, H-score<70) and 34 cases (55.7%, H-score ≥70).

Light GM-CSFRα-expressing ICI was observed in 62 cases (64.6%), and moderate to dense infiltration was observed in 34 cases (35.4%; moderate = 27 cases, dense = 7 cases) (Table 1). The distributions of GM-CSFRα-expressing ICI were comparable between papillary and non-papillary subtypes (Fig. 2D). Light GM-CSFRα-expressing ICI was detected in 22 cases (62.9%) of papillary iCCA while light GM-CSFRα was noted in ICI of 40 non-papillary cases (65.6%).

GM-CSFRα-expressing ICI was an independent prognostic factor for iCCA patients

Univariate analysis was performed to determine the correlation between GM-CSF or GM-CSFRα expressions and clinical parameters of the iCCA patients using χ² test. The results demonstrated that the papillary subtype expressed higher GM-CSF (Table 1). Even though the patients with higher GM-CSF or moderate to dense GM-CSFRα seemed to have longer survival times, there were no statistically significant differences (Figs. 3A–3B); however, when the expressions of GM-CSF and GM-CSFRα were combined, the results showed high GM-CSF accompanied with moderate to dense GM-CSFRα-expressing ICI was correlated with longer median survival times (474 days) compared to one of these high expressions (329 days) or those who had decreased expressions of both proteins (209 days, p = 0.047) (Fig. 3C). Multivariate Cox regression analysis indicated that non-papillary subtype, TNM stage III, and light GM-CSFRα-expressing ICI were independent poor prognostic indicators (HR) = 2.130; 95% CI [1.046−4.337], p = 0.037; HR = 4.233, 95% CI [1.325–13.522], p = 0.015; and HR = 1.882, 95% CI [1.077–3.287], p = 0.026 (Table 2). It is worth noting that differential expressions of GM-CSF in distinct iCCA subtypes and GM-CSFRα in papillary iCCA were not correlated with the survival times of patients (Fig. S2).

It was reported that the different histological subtypes exhibited dissimilar disease progression; the papillary subtype seemed less aggressive and acquired a longer survival time (Zen et al., 2006). In agreement with a previous report, the longer median survival time was observed in patients with papillary iCCA that were recruited to the current study (456 vs. 236 days, p < 0.001) (Fig. 3D). Hence, different subtypes should be analyzed separately. The correlation between GM-CSF or GM-CSFRα expressions and clinical characteristics of the patients with distinct iCCA subtypes was performed. Longer median survival time of non-papillary iCCA was positively correlated with moderate to dense GM-CSFRα-expressing ICI (351 vs. 181 days, p = 0.002) (Fig. 3E). The multivariate Cox proportional hazards model indicated that light GM-CSFRα-expressing ICI increased risk of death to 2.788 times in non-papillary iCCA (95% CI [1.299–5.985], p = 0.009), while cancer stage III and lymph node metastasis increased risk of death 6.017 times, and 5.094 times in papillary iCCA (95%CI [1.394–25.959], p = 0.016 and 1.256–20.649, p = 0.023) (Table 3).

GM-CSF had no direct effects on CCA cell proliferation and migration

To assess whether CCA cells produce GM-CSF, GM-CSF production was determined by detecting GM-CSF in conditioned media of CCA cells. The results showed GM-CSF was undetectable in KKU-055 and KKU-100 (<4 pg/ml), but in KKU-213A and KKU-213B, there were 84 ± 6 and 100 ± 13 pg/ml and were higher in metastatic cells, KKU-213L5 and KKU-214L5 (349 ± 21 and 544 ± 56 pg/ml, Fig. 4A). To evaluate the possible roles of GM-CSF on CCA cells, the surface expression of its cognate receptor, GM-CSFRα, on CCA cells was assessed by flow cytometry. GM-CSFRα expressions on CCA cells varied from undetectable in KKU-213B to slight expression in KKU-213A and KKU-213L5 (1.11 and 1.12 times higher than MFI of isotype control). Moderate expressions were detected in KKU-055 and KKU-214L5 (2.44 and 2.17 times), while the highest expression was demonstrated in KKU-100 (9.93 times, Fig. 4B). Two CCA cell lines, KKU-055 and KKU-213B, were selected as representatives of GM-CSFRα-expressing and GM-CSFRα-non-expressing cells. The effects of GM-CSF on cell proliferation and migration were determined under rhGM-CSF treatment. Treatment with GM-CSF did not affect CCA cell proliferation and migration (Figs. 4C and 4D).

GM-CSF and GM-CSFRα expression levels were correlated with specific ICI

As the moderate to dense ICI was correlated with longer survival time of specific iCCA, the potential immune cells were identified by an online tool, TIMER2.0. The infiltrating immune cells, which were reportedly associated with cancer-derived GM-CSF, including neutrophils, dendritic cells (DCs), all macrophages, M2 macrophages, and myeloid-derived suppressor cells (MDSCs), and CD8+ T cells, were selected (Figs. 5A and 5B). The results showed a positive correlation between specific groups of immune cells and GM-CSF or GM-CSFRα mRNA expressions. The GM-CSF expression level was positively correlated with infiltrations of DCs, all macrophages, and CD8+ T cells (Fig. 5A). Similarly, the higher GM-CSFRα expression was correlated with increased neutrophil, DC, and CD8+ T cell infiltrations. In contrast, the expression of GM-CSFRα was inversely correlated with M2 macrophage and MDSC infiltrations (Fig. 5B).