A total of 37 patients underwent surgery for malignant pleural mesothelioma at the Nagoya University Hospital between January 2007 and December 2016. All patients were reviewed for age, sex, histological subtype, pathological invasion (pT), lymph node metastasis (pN), and neoadjuvant therapy. Tumor classification was performed based on the TNM Classification of Malignant Tumors (UICC) 7th edition (30). Patients who had another cancer, who had undergone several surgeries, or who had undergone only cytoreductive surgery were excluded. Based on histological and immunohistochemical analyses, patients with the biphasic or sarcomatoid subtype were also excluded because of the difficulty in distinguishing the mesothelioma cells from CAFs. In total, 22 samples were ultimately analyzed. Human mesothelioma tissues were obtained with informed patient consent at the time of surgery at Nagoya University Hospital (Nagoya, Japan). This study was carried out in accordance with the principles of the Helsinki Declaration for human research and approved by the Ethics Committee of Nagoya University Graduate School of Medicine (protocol no. 2017-0127).

Four-micron- thick serial sections were cut from formalin-fixed and paraffin-embedded tissue and were stained with hematoxylin and eosin (H&E) or Elastica-Masson or for immunohistochemistry (IHC). The following antibodies were used for IHC: Anti-CTGF (goat polyclonal, dilution 1:50; Santa Cruz Biotechnology, Inc.; cat. no. 14939), anti-AE1/AE3 (mouse monoclonal, dilution 1:100; Biocare Medical; cat. no. ACR011A, B, C), anti-Ki-67, clone SP6 (rabbit monoclonal, dilution 1:100; Abcam; cat. no. 16667), and anti-αSMA (mouse monoclonal, dilution 1:50; Dako; cat. no. M0851). High-temperature antigen retrieval for CTGF and Ki-67 was performed using Immunosaver (Nisshin EM, Tokyo, Japan) and that for αSMA and AE1/AE3 was performed using 10 mM Tris (hydroxymethyl) aminomethane/1 mM ethylenediaminetetraacetic acid (TE) buffer, pH 9.0. Following antigen retrieval, the sections were dipped for 30 min in methanol containing H₂O₂ (0.3% vol/vol) to quench endogenous peroxidase activity and subsequently blocked with Protein Block Serum-Free Ready-to-use (Dako). For CTGF staining, the avidin-biotin complex method using peroxidase was employed, as previously described (31). For AE1/AE3, Ki-67, and αSMA staining, Histofine Simple Stain MAX-PO (Multi; Nichirei, Tokyo, Japan) was used as the secondary antibody. Color development was performed with 3,3′-diaminobenzidine (DAB, Dako) or HistoGreen (EUROBIO/ABCYS, Courtaboeuf, France). All images were obtained using an Olympus BX53 microscope (Olympus, Japan; objective lens PlanApo N 2X and U PlanApo 4X and 10X) and a DP22/U-TV0.5XC camera.

We performed double staining for αSMA and AE1/AE3. DAB was used to stain αSMA in the first step, and HistoGreen was used to stain AE1/AE3 in the second step. Because both antibodies are mouse monoclonal, high-temperature TE buffer was used to inactivate the anti-αSMA antibody after DAB staining.

The entire tumor mass of each specimen was digitized using the microscope and camera described above. Based on previous studies (15,32), we analyzed the images using ImageJ 1.50i (http://rsb.info.nih.gov/ij/) and the color deconvolution plugin (http://imagej.net/Colour_Deconvolution) for ImageJ and Fiji to implement staining separation via the method of Ruifrok and Johnston (33). Fibrosis was detected as a light green color in Elastica-Masson staining. The light-green positive area was extracted with the color deconvolution plugin (vector: Feulgen Light Green), and the area was converted to black (threshold: Upper cutoff, 214; lower cutoff, 0). After this processing, most of the remaining pixels in the image were those originally stained in light green, and we calculated the total area occupied by these pixels. This area was divided by the entire tumor area, and the extent of fibrosis (fibrotic area index) was determined as previously described (15). To obtain the αSMA-positive area (αSMA area index) and AE1/AE3-positive area in the entire tumor, we used the same method used to determine the fibrotic area index (vector: H DAB; threshold: Upper cutoff, 200; lower cutoff, 0). Most of the cases had the reactive pleural fibrosis with mesothelioma cell invasion. For the fibrotic and αSMA area indices, we evaluated the entire tumor mass including pleural fibrosis since invasion to parietal and visceral pleura is a common pattern of mesothelioma invasion (2) and since the fibrosis can contribute to mesothelioma progression. To minimize the influence of tumor heterogeneity, all the fields in the entire tumor were evaluated. The average of each entire tumor mass area was 141.7 mm². We used a 2× objective lens and the average fields of view for each tumor mass was 7.6.

We used the entire tumor mass of each specimen. The CTGF immunostaining intensity in mesothelioma cells and CAFs was assessed as follows: 0, negative; 1, weak; 2, moderate; and 3, strong. In addition, the H-score for CTGF (CTGF score) was calculated using the following formula: [1× (% of cells with an intensity of 1)+2× (% of cells with an intensity of 2)+3× (% of cells with an intensity of 3)] (34,35). All the fields were evaluated by a registered pathologist (YO). The average of each entire tumor mass area was 141.7 mm². We used a 4× objective lens and the average fields of view for each tumor mass was 24.4. For the Kaplan-Meier survival curve, the CTGF score for mesothelioma cells and CAFs was modified by the tumor area (AE1/AE3-positive area) and stromal area (αSMA-positive area) as follows: (CTGF score × AE1/AE3 or αSMA area index). Using these modified CTGF scores, the patients were divided into two groups (low or high).

In situ hybridization (ISH) analysis was performed using four-micron-thick formalin-fixed and paraffin-embedded human tissue sections with the RNAscope technology (RNAscope 2.5 HD Detection Kit; Advanced Cell Diagnostics) and a custom-designed probe of human Meflin according to the manufacturer's instructions, as previously described (28,29). Briefly, tissue sections were baked in a dry oven (HybEZ II Hybridization System; Advanced Cell Diagnostics) at 60°C for 1 h, deparaffinized, and incubated with H₂O₂ solution (Pretreat 1 buffer) for 10 min at room temperature. The slides were boiled in target retrieval solution (Pretreat 2 buffer) for 30 min, incubated with protease solution (Pretreat 3 buffer) for 30 min at 40°C, incubated with the probe for 2 h at 40°C, and then successively incubated with Amp1 to 6 reagents. Staining was visualized with DAB, followed by counterstaining with hematoxylin. The RNAscope probe was as follows: Human Meflin (ISLR) (NM_005545.3, region 275–1322; cat. no. 455481). The slides were evaluated, as previously described (29).

ImageJ software was used for obtaining the merged image of αSMA + AE1/AE3 and Meflin. To detect the Meflin-positive area, we used the same method used to determine the αSMA area index (vector: H DAB; threshold: Upper cutoff, 190; lower cutoff, 0). The Meflin-positive area was indicated by red color and merged with αSMA + AE1/AE3 using Image Calculator.

The data were analyzed using GraphPad Prism 5 (GraphPad Software). The correlation between the expression of CAF markers and various clinicopathological features was analyzed by the Fisher's exact test. Correlation analysis was performed using non-parametric method (Spearman's rank correlation coefficient). The overall survival rate was calculated according to the Kaplan-Meier method and compared using the Log-rank test if not indicated otherwise. Gehan-Breslow-Wilcoxon test was used if crossover between the groups was observed at late timepoints. A P-value of <0.05 was considered as indicative of statistical significance.

Through H&E, IHC staining of αSMA + AE1/AE3, and Elastica-Masson staining, a high extent of fibrosis was observed in the reactive pleura present adjacent to mesothelioma (Fig. S1). To confirm the correlation between fibrosis or CAFs and mesothelioma patient features, we first evaluated the density of fibrosis and the presence of CAFs expressing αSMA using paraffin-embedded sections of mesothelioma samples. Although epithelioid mesothelioma is positive for AE1/AE3 (2,36), approximately 20% of sarcomatoid mesothelioma is negative for AE1/AE3 (37,38). In addition, it has been reported that reactive spindle cells can be positive for AE1/AE3 in sarcomatoid mesothelioma (39). Although the H&E and IHC staining distinguished mesothelioma cells from CAFs in epithelioid mesothelioma (Fig. S2), our cases of biphasic or sarcomatoid mesothelioma contained cells for which it was difficult to clarify whether they were mesothelioma cells or CAFs. Thus, we excluded cases of biphasic and sarcomatoid mesothelioma. The area indices of fibrosis and αSMA were quantified based on color deconvolution (Fig. 1A-D), and the clinicopathological findings are summarized (Tables I and II). The indices of fibrosis and αSMA did not correlate with the clinicopathological features. In addition, no significant differences in overall survival were found based on the fibrotic area index (Fig. 1E). However, a significant difference (P=0.0262) was found based on the αSMA area index (Fig. 1F).

To confirm CTGF as a prognostic factor and potential targets, we next evaluated the expression of CTGF and Ki-67 using paraffin-embedded sections. Mesothelial cells, which are nontumorous, did not express CTGF (Fig. S3), whereas obvious CTGF expression was observed in both mesothelioma cells and CAFs (Figs. 2A-F and S2). Biphasic and sarcomatoid mesotheliomas were excluded after the IHC staining results were examined, as described above. Heterogeneity in CTGF expression was observed in both mesothelioma cells and CAFs. We therefore adopted a semiquantitative scoring system (CTGF score; see Materials and methods) to quantify the expression of CTGF in each tumor sample (Fig. 3A and B) and compared these scores with the numbers of Ki-67-positive cells (Ki-67 index). The CTGF score for CAFs but not that of mesothelioma cells was correlated with the Ki-67 index for mesothelioma cells (Fig. 3C and D). The CTGF score for CAFs was also correlated with the αSMA area index, while that for mesothelioma cells was not (Fig. 3E and F). No significant correlations were found between CTGF expression in mesothelioma cells and patient prognosis (Fig. 3G). However, CTGF expression in CAFs correlated with poor prognosis (Fig. 3H). Notably, the clinicopathological features (pathological invasion, lymph node metastasis, and stage) and sensitivities to neoadjuvant chemotherapy of the examined mesothelioma cases did not correlate with CTGF expression in either mesothelioma cells or CAFs (Tables III and IV), suggesting that CTGF in CAFs could be used as a marker that specifically predicts patient prognosis.

We next investigated Meflin expression by RNA ISH using paraffin-embedded sections. Mesothelial cells, which are nontumorous, did not express Meflin (Fig. S2), while CAFs expressed Meflin (Fig. 4A and B). More Meflin⁺ CAFs were observed in the αSMA-negative area than in the αSMA-positive area. Meflin expression did not correlate with the clinicopathological features (Table V). Additionally, no significant differences were found in patient prognosis according to Meflin expression (Fig. 4C).