Colorectal cancer (CRC) is one of the most common gastrointestinal cancers worldwide, with high morbidity and mortality rates [1]. As with other malignancies, evidence for the tumour-supporting capacities of the tumour stroma has rapidly been accrued in CRC [2–4]. The tumour stroma is composed of heterogeneous cell types and components including cancer-associated fibroblasts (CAFs), small vessels, immune cells, and extracellular matrix proteins such as collagen. Because of their broad range of tumour-supporting properties, components of cancer stroma are believed to be promising targets of cancer therapy.
CAFs are a major component of the tumour stroma and are a highly heterogeneous population of cells with different phenotypes and functions. This heterogeneity is considered to be due partly to their different origins; resident fibroblasts [5,6], bone-marrow-derived mesenchymal stromal cells [7], mature adipocytes [8], and other cells exist within the tumour microenvironment [4]. Another source of CAFs could potentially be tumour cells after the process of epithelial-to-mesenchymal transition [4]. Because of the heterogeneity of CAFs, specific and common markers for CAF have not yet been identified; however, many attempts have been made to identify markers classifying CAFs. In CRC, single-cell RNA sequencing enabled the classification of CAFs into two types, CAF-A and CAF-B, according to their gene expression status [9]. CAF-A is defined by its expression of decorin (DCN), fibroblast activation protein (FAP), and COL1A2, indicating its extracellular matrix-remodelling capabilities. In contrast, CAF-B has been reported to express cytoskeletal genes and other known markers of activated myofibroblasts such as ACTA2 (commonly referred to as alpha-smooth muscle actin or α-SMA) [9].
DCN is a small cellular or pericellular matrix proteoglycan belonging to the small leucine-rich proteoglycan family, and is composed of a protein core containing leucine repeats with a glycosaminoglycan chain consisting of either chondroitin sulphate or dermatan sulphate [10]. DCN is a component of connective tissue, binds to type I collagen fibrils, and plays a role in matrix assembly [10]. The Dcn-deficient knockout mouse shows reduced inflammatory reactions during contact dermatitis due to a defect in leukocyte recruitment and altered interferon gamma function [11,12]. Dcn was shown to have anti-tumourigenic properties in an experimental murine tumour model and is capable of suppressing the proliferation of various tumour cell lines [13,14].
PDPN, FAP, and ACTA2 have been reported as CAF markers [9,15]. PDPN encodes a type-I integral membrane glycoprotein with diverse distribution in human tissues [16]. The physiological function of this protein has not been determined but it is described as a differentiation antigen and influenza-virus receptor [16]. FAP is a non-classical serine protease belonging to the S9B prolyl oligopeptidase subfamily that cleaves several bioactive peptides and structural proteins such as human fibroblast growth factor 21 and denatured collagen I and III [17]. FAP is expressed during embryonic development; however, its expression under physiological conditions is very low in most adult tissues [18]. In contrast, FAP expression is observed on the activated stromal fibroblasts of more than 90% of all human carcinomas [17]; therefore, FAP-targeting therapeutics such as radionuclide therapy may become novel treatments for various types of cancer [19,20]. Actins are a family of globular multifunctional proteins that form microfilaments. Mutations in this gene cause a variety of vascular diseases such as moyamoya disease, and multisystemic smooth muscle dysfunction syndrome [21]. ACTA2, one of six different actin isoforms, is often used as a marker of myofibroblast formation [22].
Collagen is the most abundant protein and is the main structural protein in the extracellular matrix of various connective tissues as well as the tumour stroma [23,24]. Collagen consists of amino acids bound together to form a triple helix of elongated fibrils known as a collagen helix [24]. To date, 28 types of human collagen have been identified, over 90% of which in the human body is type I collagen [24]. Fibroblasts, including CAFs in the tumour microenvironment, are the most common cell type that creates collagen [23]. Collagens are extremely important in the tumour microenvironment for modulating crucial steps in tumourigenesis such as proliferation, apoptosis, angiogenesis, invasion, and metastasis [23].
The present study examined the expression status of DCN, PDPN, FAP, ACTA2, and collagen in CRC and analysed their association with clinicopathological features and clinical outcome. For further characterisation and classification of CRCs, hierarchical clustering analyses were performed.
Materials and methods Tissue samplesThe institutional ethical review boards of Aichi Medical University Hospital (2020-H122, 1 March 2021) and Nagoya City University Graduate School of Medical Sciences (60-23-0014, 14 July 2023) approved this project without collecting patient consent by giving them the opportunity to opt-out. The project was performed in accordance with the Declaration of Helsinki. Two hundred and sixty-nine formalin-fixed paraffin-embedded (FFPE) samples of primary colorectal tumours resected at Aichi Medical University Hospital from 2009 to 2012 were collected depending on the availability of tissue samples and clinical information. After surgery, patients were followed up for up to 90 months. All tumours were diagnosed as invasive [25] and were naïve to chemotherapy or radiotherapy. Tumours with glandular formation (>50%) or mucin production (>50% of the area) were defined as having a differentiated or mucin-producing histology. Tumour budding was counted under a magnification of ×200 at a hot-spot and scored as follows: BD1, 0–4; BD2, 5–9; and BD3, ≥10 [26]. A single 4.5-mm core tumour tissue sample derived from an FFPE specimen was assembled into multiturn blocks containing up to 30 samples. All cores were obtained from invasive areas, and approximately 20% of cores contained an invasive front. Non-neoplastic colonic mucosae adjacent to the tumour were also immunohistochemically analysed.
ImmunohistochemistryThe antibodies used in the present study are summarised in supplementary material, Table S1. Immunohistochemistry was performed using a Leica Bond-Max (Leica Biosystems, Bannockburn, IL, USA) or Ventana BenchMark XT (Roche Diagnostics, Basel, Switzerland) automated immunostainer. Signals were visualised using 3,3′-diaminobenzidine. Collagens were stained with Picro-Sirius Red Stain (ScyTek Laboratories, Inc., Logan, UT, USA). DCN, FAP, PDPN, ACTA2, and collagen-positive areas were evaluated using ImageJ software (NIH, Bethesda, MD, USA) according to our previous studies [27,28]. DCN and collagens within non-neoplastic submucosal areas and smooth muscles such as vascular wall and muscularis propria were eliminated from the evaluation. To assess DCN expression in CRC cells (cDCN) at the cut-off value of 5%, sections were observed by microscopy. p53 immunoreactivity was classified as follows: wild type, overexpression, cytoplasmic, and complete loss [29,30]. The number of phospho-histone H3 (PHH3)-positive cells was counted under a microscope (×400). Cyclin A (CCNA), geminin (GMNN), and Ki-67 labelling indices were determined by counting over 500 tumour cells per case under a high-power field (×400) [31]. FOXP3+ immune cells within the tumour microenvironment [27] and CD8+ immune cells infiltrating into CRC cells were counted at hot-spots under a high-power field (×400)
Fluorescent immunohistochemistryFluorescent immunohistochemistry was performed with a Leica Bond-Max automated immunostainer. Signals were visualised using secondary antibodies labelled with fluorescein or tetramethylrhodamine applied at a dilution of 1:500 (Molecular Probes, Thermo Fisher Scientific K.K., Tokyo, Japan). Autofluorescence was attenuated using a Vector TrueVIEW Autofluorescence Quenching Kit (Vector Laboratories, Inc., Burlingame, CA, USA).
Gene mutation analysesKRAS mutation status was ascertained from patients' medical records. BRAF V600E mutation analyses were performed by polymerase chain reaction (PCR) direct sequencing using the following primers: BRAF Forward: tgc ttg ctc tga tag gaa aat g and BRAF Reverse: cag ggc caa aaa ttt aat cag t.
Statistical analysesStatistical analyses were performed using EZR software version 1.41 [32] or R software version 4.2.2. The chi-squared test, Fisher's exact test, Mann–Whitney U test, or Kruskal–Wallis test was performed to analyse the statistical correlations between categorical data. Simple Bonferroni correction for multiple hypothesis testing was applied for adjustment at a two-sided alpha level. Hierarchical clustering analyses with heat maps were performed using a Euclidean distance measure and Ward linkage with the R software and complexheatmap-2 package. This analysis was done using logarithmic transformed expression data. The analyses were ‘unsupervised’ with no assumptions made with regard to patients’ clinical or pathological characteristics.
For survival analyses, Kaplan–Meier survival estimates were calculated with the log-rank test. Cox proportional hazards regression analysis was performed to determine the associations of survival with the classified groups adjusting for confounders. The model included the following variables: sex (male versus female), age, tumour size, primary tumour location (right-sided colon versus left-sided colon versus rectum), pT stage (pT2 versus pT3 versus pT4), tumour histopathology (moderately to well-differentiated versus poorly differentiated), mucin production (positive versus negative), lymph node metastasis (positive versus negative), peritoneal metastasis (positive versus negative), distant organ metastasis (positive versus negative), surgical status (complete versus incomplete resection), mismatch repair system status (preserved versus deficient), and the results of hierarchical clustering analyses (group 3 versus group 2 versus group 1).
Transcriptomic analyses of colon and rectal adenocarcinoma using the cancer genome atlas (For transcriptomic analyses of the TCGA data, the UCSC Xena program (
Representative images of immunohistochemistry and collagen staining are presented in Figure 1A and supplementary material, Figure S1. In non-neoplastic colonic mucosae, DCN was expressed in the stromal cells of the mucosal layer. DCN was also detected on the basement membrane and collagen bundles within the mucosal and submucosal layers, indicating the binding capacity of DCN to collagen fibrils [10]. Similar to the DCN expression in TCGA data, DCN expression was higher in non-neoplastic tissue (supplementary material, Figure S2A). DCN expression was undetectable in non-neoplastic colonic epithelial cells. Stromal cells within the mucosal layer variably expressed ACTA2 and PDPN; however, FAP expression was almost negative in non-neoplastic colonic mucosae. In CRC tissues, CAFs variably expressed DCN. Limited cases of CRC (6%, 16/269) showed DCN expression in cancer cells (cDCN). Collagen detected within cancer stroma was considered as type I collagen owing to the dominant expression of COL1A1 and COL1A2 (supplementary material, Figure S2B). CAFs variably expressed ACTA2, PDPN, and FAP. In contrast, as well as non-neoplastic colonic epithelial cells, CRC cells expressed ACTA2, PDPN, and FAP at undetectable levels. The cut-off values for each parameter were defined using receiver operating characteristic (ROC) curves for patient survival at 5 years (supplementary material, Figure S3).
Enlarge this image.Figure 1. Survival significance of stromal factors in CRC. (A) Representative images of stromal marker expression in non-neoplastic colonic mucosae and CRC. (B–G) Kaplan–Meier curves for patients classified by (B) total DCN (tDCN), (C) DCN in CRC cells (cDCN), (D) collagen, (E) ACTA2, (F) PDPN, and (G) FAP expression. Note that DCN expression in both CRC and stromal cells (B) uniquely showed significantly worse clinical outcomes (p = 0.044).
The clinical, pathological, and immunohistochemical features of the analysed tumours are summarised in Table 1 according to stromal marker expression. For DCN (p = 0.0088) and collagen (p = 0.00034), their expression was associated with advanced pT stage. In contrast, PDPN was most frequently expressed in pT2 tumours (p = 0.00062). In the TCGA cohort, COL1A1 (p = 0.0088), FAP (p = 0.0022), and ACTA2 (p = 0.0010) were significantly associated with advanced pT stage (supplementary material, Figure S4A). PDPN (p = 0.0075) and ACTA2 (p < 0.0001) were highly expressed in younger patients. Collagen positivity was associated with peritoneal metastasis (p = 0.0069). ACTA2 showed an association with rectal location (p = 0.0050), tubule-forming histology (p = 0.00024), and mucin production (p = 0.0044). FAP expression was associated with mucin production (p = 0.0030).
Table 1 Expression of stromal markers in CRC based on clinicopathological features
| Number (%) | Stromal markers | |||||||||||
| Collagen | DCN | FAP | PDPN | ACTA2 | ||||||||
| 269 (100%) | Area | p value | Area | p value | Area | p value | Area | p value | Area | p value | ||
| Age | † | |||||||||||
| Young | 134 (50%) | 912 (321.5, 2406.25) | p = 0.86 | 2013 (507, 5968) | p = 0.53 | 1293.5 (544, 2995.75) | p = 0.090 | 430.5 (128.5, 1456.5) | p = 0.0075 | 4699.5 (2812.75, 6775.5) | p < 0.0001* | |
| Old | 135 (50%) | 1031 (383.5, 2109) | 2468 (631, 7150) | 967 (471.5, 2276) | 261 (41, 981.5) | 3193 (1834, 5385) | ||||||
| Sex | † | |||||||||||
| Male | 143 (53%) | 1147 (327, 2601) | p = 0.50 | 3114 (743.5, 7472.5) | p = 0.11 | 1174 (506.5, 2673) | p = 0.45 | 416 (75.5, 1467) | p = 0.20 | 3986 (2180.5, 6102.5) | p = 0.55 | |
| Female | 126 (47%) | 889.5 (359.5, 2021.75) | 1661.5 (558.5, 4340) | 1014 (397.25, 2544.5) | 229.5 (64.5, 1010.25) | 3635.5 (2173.75, 6155.5) | ||||||
| Size | † | |||||||||||
| <5 cm | 145 (54%) | 862 (288, 1880) | p = 0.11 | 1601 (452, 4669) | p = 0.86 | 1096 (444, 2469) | p = 0.86 | 379 (128, 1172) | p = 0.019 | 4116 (2501, 6233) | p = 0.16 | |
| ≥5 cm | 123 (46%) | 1160 (388.5, 2625.5) | 3294 (897, 8846.5) | 1164 (475, 2884.5) | 191 (39, 1143) | 3614 (1936, 5892) | ||||||
| Primary site | ‡ | |||||||||||
| Right-sided colon | 124 (46%) | 1053 (447.75, 2115.5) | p = 0.44 | 2837 (754.5, 7059.5) | p = 0.037 | 1320 (534.75, 2637.25) | p = 0.13 | 443 (65.5, 1285.75) | p = 0.31 | 3720 (2036.75, 5683) | p = 0.0050 | |
| Left-sided colon | 86 (32%) | 768 (293.5, 2557.2) | 1235 (339.8, 4116.8) | 884.5 (280, 2193.75) | 202.5 (61.5, 967.75) | 3425.5 (1966.25, 5648) | ||||||
| Rectum | 59 (22%) | 1147 (282, 2218.5) | 2245 (733.5, 7873.5) | 1174 (580.5, 3037) | 341 (84, 1226) | 4919 (3208, 7799) | ||||||
| pT stage | ‡ | |||||||||||
| T2 | 36 (13%) | 534 (160.5, 1403) | p = 0.00034* | 914.5 (340.25, 2872.25) | p = 0.0088 | 1336.5 (725.25, 2497.75) | p = 0.55 | 1119 (302.5, 2014.25) | p = 0.00062* | 3633 (3153.25, 4807) | p = 0.60 | |
| T3 | 189 (70%) | 917 (320, 2091) | 2245 (609, 6738) | 1143 (442, 2635) | 234 (60, 1023) | 3829 (2242, 6231) | ||||||
| T4 | 44 (16%) | 2001.5 (756.5, 3342.25) | 3535.5 (1115.75, 9533) | 925.5 (359, 2859.25) | 522.5 (47.5, 1034.75) | 3338.5 (1510.25, 6898.25) | ||||||
| Histological differentiation | † | |||||||||||
| Well to moderately | 242 (90%) | 939.5 (352.5, 2285) | p = 0.90 | 2045 (595, 6902.25) | p = 0.94 | 1106 (473, 2583) | p = 0.89 | 338.5 (83.5, 1194.5) | p = 0.069 | 4051 (2443.25, 6232.5) | p = 0.00024* | |
| Poorly | 27 (10%) | 994 (299, 2151.5) | 2473 (882, 4913.5) | 1336 (371, 2992) | 125 (50.5, 449) | 1905 (973, 3144.5) | ||||||
| Mucin production | † | |||||||||||
| Positive | 14 (5%) | 1197.5 (720.75, 2896) | p = 0.21 | 3272 (1140.5, 6788.25) | p = 0.39 | 382 (109.25, 848.5) | p = 0.0030 | 128.5 (64.5, 1175) | p = 0.44 | 1824 (1448, 3306.5) | p = 0.0044 | |
| Negative | 255 (95%) | 932 (334.5, 2211.5) | 2058 (593.5, 6730) | 1181 (506.5, 2714.5) | 339 (73.5, 1171.5) | 3858 (2337, 6232) | ||||||
| Tumour budding | ‡ | |||||||||||
| BD1 | 183 (68%) | 874 (334.5, 2608) | p = 0.90 | 1670 (485.5, 6093) | p = 0.14 | 834 (340.5, 2177.5) | p = 0.00015* | 339.1 (73.5, 1212.5) | p = 0.57 | 3703 (2227.5, 5924.5) | p = 0.63 | |
| BD2 | 61 (23%) | 1073 (421, 1983) | 3379 (1086, 7051) | 1850 (857, 3191) | 305 (128, 1104) | 4182 (2418, 6831) | ||||||
| BD3 | 25 (9%) | 1159 (341.5, 1775) | 2482 (617, 7085) | 2303 (924, 3389) | 261 (16, 978) | 3316 (1948, 5837) | ||||||
| Lymph node metastasis | † | |||||||||||
| Positive | 124 (49%) | 1063 (388, 2577.25) | p = 0.13 | 2438.5 (594.25, 7321.25) | p = 0.34 | 1289 (507.5, 2879.5) | p = 0.18 | 273 (74.75, 955) | p = 0.29 | 3912 (2207.25, 6363) | p = 0.46 | |
| Negative | 129 (51%) | 836 (338, 2042) | 1956 (595, 4727) | 1011 (432, 2448) | 339 (66, 1465) | 3657 (2242, 5701) | ||||||
| Distant organ metastasis | † | |||||||||||
| Positive | 44 (16%) | 1326.5 (654.5, 3113.25) | p = 0.011 | 2702 (975.75, 9785.75) | p = 0.10 | 1248.5 (531.75, 2856.5) | p = 0.83 | 220 (43.25, 1015.75) | p = 0.33 | 2642 (1430.25, 5547.25) | p = 0.034 | |
| Negative | 225 (84%) | 874 (284, 2135) | 2024 (491, 5949) | 1096 (473, 2635) | 338 (85, 1223) | 3815 (2371, 6231) | ||||||
| Peritoneal metastasis | † | |||||||||||
| Positive | 50 (19%) | 1326.5 (635.5, 3213.75) | p = 0.0069 | 2702 (1037.7, 8527.75) | p = 0.082 | 1248.5 (571, 2860.75) | p = 0.65 | 220 (46.5, 946) | p = 0.31 | 3394.5 (1565.75, 6254.5) | p = 0.19 | |
| Negative | 219 (81%) | 874 (283, 2131) | 2021 (490, 5907.5) | 1096 (471.5, 2609.5) | 338 (85.5, 1279) | 3808 (2362, 6148) | ||||||
| Operation status | † | |||||||||||
| Complete resection | 237 (88%) | 905 (341, 2189) | p = 0.32 | 2051 (595, 5949) | p = 0.15 | 1030 (444, 2448) | p = 0.013 | 292 (66, 1171) | p = 0.50 | 3737 (2268, 5968) | p = 0.56 | |
| Incomplete resection | 32 (12%) | 1368 (366.75, 3026.5) | 2875.5 (996, 11639) | 2530.5 (700, 3926.5) | 669.5 (77, 1201.25) | 4115 (2016, 7907.25) | ||||||
| MMR system status | † | |||||||||||
| Deficient | 31 (12%) | 596.9 (333.5, 2162.5) | p = 0.35 | 2039 (550, 5291) | p = 0.65 | 1366 (364.5, 2757) | p = 0.91 | 319 (85.5, 1287.5) | p = 0.84 | 2268 (1404, 5233.5) | p = 0.018 | |
| Preserved | 238 (88%) | 1016.5 (358, 2342.5) | 2111 (626, 7020) | 1096 (473.2, 2629) | 314 (73.2, 1164) | 3966 (2373.8, 6216.2) | ||||||
Data are shown as median (25th, 75th percentiles).
MMR, mismatch repair.
*The median difference is significant at 0.00077 (0.05/65) level (Bonferroni correction).
†Mann–Whitney U test was used to calculate p values.
‡Kruskal–Wallis test was used to calculate p values.
In the association with cellular proliferation markers, DCN- and collagen-high tumours showed significantly lower proliferative activity in tumour cells. In contrast, PDPN-high tumours showed significantly higher expression of the proliferation markers (Table 2).
Table 2 Cellular proliferation of CRC classified by stromal marker expression
| Total No. | Collagen | DCN | FAP | PDPN | ACTA2 | ||||||||||||
| High | Low | p value | High | Low | p value | High | Low | p value | High | Low | p value | High | Low | p value | |||
| 269 (100%) | 131 (49%) | 138 (51%) | 71 (26%) | 198 (74%) | 133 (49%) | 136 (51%) | 117 (43%) | 152 (57%) | 144 (54%) | 125 (46%) | |||||||
| PHH3 | 6 (3–9.5) | 8 (3–13.75) | 0.011 | 5 (2–9) | 7 (3–13) | 0.009 | 7 (3–10) | 7 (3–13) | 0.41 | 8 (5–13) | 5 (2–10) | 0.00018* | 7 (3–13) | 6 (3–11) | 0.090 | † | |
| CCNA | 34.4 (27.2–43.3) | 37.4 (31.8–45.7) | 0.038 | 34.2 (24.3–42.7) | 36.9 (29.5–45.5) | 0.023 | 36.5 (28.1–45.6) | 35.3 (28.5–43.2) | 0.55 | 40.4 (34.5–47.4) | 33.6 (25.2–40.2) | <0.0001* | 37.7 (29.5–45.1) | 34.9 (26.8–43.6) | 0.12 | † | |
| GMNN | 34.1 (25.2–41.9) | 38.3 (31–45.5) | 0.002* | 31.3 (23.4–41.4) | 37.8 (30.4–45.5) | 0.002* | 37.5 (28.1–44.0) | 36.2 (29.4–43.3) | 0.78 | 38.5 (31.7–46.3) | 34.6 (25.7–41.1) | 0.00017* | 37.6 (28.6–43.8) | 36.0 (25.9–43.5) | 0.30 | † | |
| Ki-67 | 46.0 (31.2–58.8) | 50.7 (35.5–65.4) | 0.053 | 45.7 (31.0–58.7) | 50.0 (34.7–62.9) | 0.33 | 49.8 (36.7–62.4) | 46.4 (31.2–61.7) | 0.30 | 54.4 (40.3–66.4) | 44.4 (31.1–57.3) | 0.00014* | 50.3 (35.9–61.8) | 45.3 (31.7–63.1) | 0.38 | † | |
Data are shown as median (25th, 75th percentiles).
*The median difference is significant at 0.0025 (0.05/20) level (Bonferroni correction).
†Mann–Whitney U test was used to calculate p values.
On the basis of a previous report claiming that DCN suppresses CRC proliferation and migration through the interaction with and stabilisation of E-cadherin in CRC cells [14], we assessed the correlation between cDCN and membranous E-cadherin; however, significant associations were not detected in our cohort (supplementary material, Figure S5A). In contrast, cDCN expression was inversely associated with PHH3 (supplementary material, Figure S5B).
Regarding KRAS and BRAF mutation status, BRAF mutant tumours showed higher DCN expression (Table 3). Based on the usefulness of p53 immunohistochemistry as a surrogate marker to predict TP53 mutation status, we classified our cohort according to p53 expression status, as follows: wild type, overexpression, cytoplasmic, and complete loss. However, no significant association was detected between p53 immunoreactivity and stromal marker expression (Table 3).
Table 3 Expression of stromal markers in CRC based on gene mutation and p53 expression status
| Number (%) | Characteristics of CAFs | |||||||||||
| Collagen | DCN | FAP | PDPN | α-SMA | ||||||||
| 269 (100%) | Area | p value | Area | p value | Area | p value | Area | p value | Area | p value | ||
| KRAS/BRAF status | † | |||||||||||
| Wild type | 15 (39%) | 1010 (333.5, 2873) | p = 0.60 | 1325 (645.5, 4258) | p = 0.036 | 1440 (710.5, 2681) | p = 0.92 | 225 (156, 910) | p = 0.53 | 3232 (1932.5, 5927.5) | p = 0.54 | |
| KRAS mutant | 19 (50%) | 1691 (744, 3022) | 3114 (1572, 5910) | 1451 (408.5, 2653.5) | 340 (75, 595.5) | 5027 (3198.5, 6596) | ||||||
| BRAF mutant | 4 (11%) | 1269 (1158.5, 1637) | 10200.5 (7944, 12473.75) | 1444.5 (1100.75, 1578.25) | 92 (50.25, 237.75) | 3185.5 (2502.75, 4423.5) | ||||||
| p53 expression | † | |||||||||||
| Wild-type pattern | 59 (23%) | 804 (375, 2476) | p = 0.12 | 2121 (568, 6495) | p = 0.13 | 1218 (354.5, 2607.5) | p = 0.93 | 287 (62.5, 925.5) | p = 0.33 | 3241 (2025.5, 5034.5) | p = 0.37 | |
| Overexpression | 143 (55%) | 917 (343, 2131) | 2024 (602, 6093) | 1143 (549, 2437) | 285 (73.5, 1170) | 3986 (2279.5, 6365) | ||||||
| Cytoplasmic expression | 10 (4%) | 330.5 (175.25, 1665) | 485.5 (205, 1909.25) | 996 (391, 2975) | 769 (280.25, 2347.25) | 3435.5 (2928, 4445.75) | ||||||
| Complete loss | 50 (19%) | 1240.5 (713.75, 3267) | 2633 (730.25, 8649.75) | 1361.5 (493.75, 3085.25) | 664 (82.5, 1304.25) | 4242.5 (2101, 5946.25) | ||||||
Data are shown as median (25th, 75th percentiles).
*The median difference is significant at 0.005 (0.05/10) level (Bonferroni correction).
†Kruskal–Wallis test was used to calculate p values.
Survival analyses of patients withAmong the analysed CAF markers, patients with DCN-high CRC uniquely had a significantly worse 5-year survival rate (57.3% versus 79.0%; p = 0.044; Figure 1B). Note that cases with DCN expression in CRC cells tended to have a worse clinical outcome (54.7% versus 74.6%; p = 0.32; Figure 1C). Collagen (p = 0.14), ACTA2 (p = 0.58), PDPN (p = 0.28), and FAP (p = 0.077) did not show significant associations with patient survival in our cohort. Different from our immunohistochemistry-based survival data, the CRC patients with higher ACTA2 expression uniquely showed worse survival in the TCGA cohort (p = 0.017, supplementary material, Figure S6). These discrepancies may originate from differential mRNA and protein expression and/or intermingled expression analyses for tumour, immune, and stromal cells in the TCGA cohort.
Classification and characterisation ofThe results of hierarchical clustering analyses using five independent stromal markers are presented in Figure 2A. DCN and collagen were classified into an identical tree. In contrast, ACTA2 and FAP were classified in the same tree. In the present study, based on the cohort size, patients were classified into three groups with specific characteristics: group 1, solid and cancer cell-rich tumours with the highest tumour cell content with lower expression of stromal markers; group 2, PDPN-dominant, presenting moderate DCN and higher PDPN, FAP, and ACTA2 expression; group 3, DCN-dominant, showing the highest DCN and the lowest PDPN expression. The DCN-dominant tumours moderately expressed FAP and ACTA2 without a significant difference compared with solid and PDPN-dominant tumours (Figure 2A,B).
Enlarge this image.Figure 2. Hierarchical clustering analyses identify three independent clusters. CRC patients were divided into three groups according to the expression of five stroma-related markers by hierarchical clustering analyses. (A) Heat map for the hierarchical clustering analyses (left panel). Representative core images of DCN and PDPN immunohistochemistry for each group (right panel). (B) Stromal marker expression in the three groups. Group 1, the solid group (cancer cell-rich, DCNLowPDPNLow); group 2, the PDPN-dominant group (DCNMidPDPNHigh); and group 3, the DCN-dominant group (DCNHighPDPNLow).
The clinicopathological characteristics of each group are summarised in Table 4. The DCN-dominant group showed significantly advanced pT stage (p = 0.0038) and a tendency for higher peritoneal (p = 0.026) and distant organ metastasis (p = 0.033). Note that no significant association was detected between tumour budding and the CAF-based subgroups. Survival analyses identified significantly favourable clinical outcomes in patients with solid tumours (Figure 3A). From Cox regression analyses, the solid group [hazard ratio (HR) = 0.60, 95% CI = 0.29–1.26, p = 0.18] and PDPN-dominant group (HR = 0.50, 95% CI = 0.26–0.96, p = 0.037) were considered more favourable than the DCN-dominant group. Age (HR = 1.04, 95% CI = 1.01–1.07, p = 0.004), rectal origin (HR = 2.06, 95% CI = 1.02–4.19, p = 0.045), tumour budding (BD3) (HR = 2.38, 95% CI = 1.05–5.35, p = 0.037), poorly differentiated histopathology (HR = 5.24, 95% CI = 2.50–11.0, p < 0.0001), and peritoneal metastasis (HR = 7.02, 95% CI = 1.68–29.26, p = 0.008) were identified as potential risk factors for CRC patients (Table 5).
Table 4 Clinicopathological features of CRC classified by hierarchical clustering
| Total No. | Characteristics of CAFs | |||||
| Group 1 | Group 2 | Group 3 | ||||
| Solid | PDPN-dominant | DCN-dominant | ||||
| 269 (100%) | 98 (36%) | 123 (46%) | 48 (18%) | p values | ||
| Age, years (mean ± SD) | 68.6 ± 12.6 | 69.3 ± 10.9 | 67.7 ± 13.9 | 69.6 ± 12.6 | p = 0.53 | † |
| Sex | p = 0.13 | ‡ | ||||
| Male | 143 (53%) | 46 (47%) | 66 (54%) | 31 (65%) | ||
| Female | 126 (47%) | 52 (53%) | 57 (46%) | 17 (35%) | ||
| Size, cm (mean ± SD) | 5.0 ± 2.6 | 4.6 ± 2.7 | 5.1 ± 2.5 | 5.6 ± 2.3 | p = 0.074 | † |
| Primary site | p = 0.14 | ‡ | ||||
| Right-sided colon | 124 (46%) | 40 (41%) | 60 (49%) | 24 (50%) | ||
| Left-sided colon | 86 (32%) | 40 (41%) | 31 (25%) | 15 (31%) | ||
| Rectum | 59 (22%) | 18 (18%) | 32 (26%) | 9 (19%) | ||
| pT stage | p = 0.0038* | ‡ | ||||
| T2 | 36 (13%) | 22 (22%) | 12 (10%) | 2 (4%) | ||
| T3 | 189 (70%) | 67 (69%) | 87 (71%) | 35 (73%) | ||
| T4 | 44 (16%) | 9 (9%) | 24 (19%) | 11 (23%) | ||
| Histological differentiation | p = 0.42 | ‡ | ||||
| Well to moderately | 242 (90%) | 88 (90%) | 113 (92%) | 41 (85%) | ||
| Poorly | 27 (10%) | 10 (10%) | 10 (8%) | 7 (15%) | ||
| Mucin production | p = 0.82 | § | ||||
| Negative | 255 (95%) | 94 (96%) | 116 (94%) | 45 (94%) | ||
| Positive | 14 (5%) | 4 (4%) | 7 (6%) | 3 (6%) | ||
| Tumour budding | p = 0.25 | § | ||||
| BD1 | 183 (68%) | 74 (76%) | 79 (64%) | 30 (63%) | ||
| BD2 | 61 (23%) | 15 (15%) | 33 (27%) | 13 (27%) | ||
| BD3 | 25 (9%) | 9 (9%) | 11 (9%) | 5 (10%) | ||
| Lymph node metastasis | p = 0.083 | ‡ | ||||
| Negative | 129 (48%) | 54 (55%) | 58 (47%) | 17 (35%) | ||
| Positive | 124 (46%) | 38 (39%) | 59 (48%) | 27 (56%) | ||
| Distant organ metastasis | p = 0.033 | ‡ | ||||
| Negative | 225 (81%) | 88 (90%) | 102 (83%) | 35 (73%) | ||
| Positive | 44 (19%) | 10 (10%) | 21 (17%) | 13 (27%) | ||
| Peritoneal metastasis | p = 0.026 | ‡ | ||||
| Negative | 219 (81%) | 87 (89%) | 98 (80%) | 34 (71%) | ||
| Positive | 50 (19%) | 11 (11%) | 25 (20%) | 14 (29%) | ||
| Operation status | p = 0.56 | ‡ | ||||
| Complete resection | 237 (88%) | 89 (88%) | 107 (91%) | 41 (79%) | ||
| Incomplete resection | 32 (12%) | 9 (12%) | 16 (9%) | 7 (21%) | ||
| MMR status | p = 0.53 | § | ||||
| Preserved | 237 (88%) | 84 (86%) | 110 (89%) | 44 (92%) | ||
| Deficient | 32 (12%) | 14 (14%) | 13 (11%) | 4 (8%) | ||
MMR, mismatch repair.
*The p values are significant at 0.0038 (0.05/13, Bonferroni correction).
†One-way ANOVA was used to calculate p values.
‡Chi-squared test was used to calculate p values.
§Fisher's exact test was used to calculate p values.
Enlarge this image.Figure 3. Survival and characteristics of subgroups classified by hierarchical clustering analyses. (A) Kaplan–Meier curves for the subgroups. Group 1, the solid group (cancer cell-rich, DCNLowPDPNLow); group 2, the PDPN-dominant group (DCNMidPDPNHigh); and group 3, the DCN-dominant group (DCNHighPDPNLow). (B) Fluorescent immunohistochemistry for DCN and PDPN in a CRC case. Note that DCN and PDPN are expressed in CRC CAFs in a mutually exclusive manner. (C) Numbers of FOXP3+ and CD8+ immune cells in CRCs. T, tumour.
Table 5 Multivariable Cox proportional hazards model of CRC
| Hazard ratio | 95% CI | |||
| Min | Max | p value | ||
| Group 3, DCN-dominant (reference) | ||||
| Group 2, PDPN-dominant | 0.50 | 0.26 | 0.96 | 0.037* |
| Group 1, solid | 0.60 | 0.29 | 1.26 | 0.18 |
| Age | 1.04 | 1.01 | 1.07 | 0.004* |
| Sex (female) | 0.60 | 0.32 | 1.11 | 0.10 |
| Size | 0.96 | 0.84 | 1.10 | 0.54 |
| Primary site (reference: right-sided colon) | ||||
| Left-sided colon | 0.89 | 0.43 | 1.86 | 0.76 |
| Rectum | 2.06 | 1.02 | 4.19 | 0.045* |
| pT stage (T2: reference) | ||||
| T3 | 2.34 | 0.66 | 8.24 | 0.19 |
| T4 | 2.31 | 0.56 | 9.61 | 0.25 |
| Tumour budding (BD1: reference) | ||||
| BD2 | 1.81 | 0.95 | 3.44 | 0.072 |
| BD3 | 2.38 | 1.05 | 5.35 | 0.037* |
| Poorly differentiated histology | 5.24 | 2.50 | 11.00 | <0.0001* |
| Mucin production | 2.44 | 0.88 | 6.77 | 0.086 |
| Lymph node metastasis | 1.66 | 0.88 | 3.16 | 0.12 |
| Distant organ metastasis | 0.74 | 0.18 | 2.96 | 0.67 |
| Peritoneal metastasis | 7.02 | 1.68 | 29.26 | 0.008* |
| Incomplete resection | 2.16 | 0.98 | 4.78 | 0.058 |
| Deficient mismatch repair system | 1.01 | 0.38 | 2.69 | 0.98 |
*Statistically significant at 0.05 level.
In further analyses, fluorescent immunohistochemistry indicated that DCN and PDPN were expressed in a mutually exclusive manner in CRC CAFs (Figure 3B). Additionally, immune cells within the tumour microenvironment were analysed. In the DCN-dominant group, the number of tumour-associated FOXP3+ and tumour-infiltrating CD8+ immune cells was the lowest among the groups (Figure 3C).
The results for subcohort analyses dividing the patients into two groups, with or without post-operative chemotherapy, are presented in Figure 4. All of the chemotherapy-received groups classified by CAF characteristics showed shorter 5-year survival than those without chemotherapy probably due to their advanced disease. Among them, the shortest survival of the DCN-dominant group with chemotherapy is significant.
Enlarge this image.Figure 4. Survival of CRC patients with or without post-operative chemotherapy. The results for subcohort survival analyses are presented. Kaplan–Meier curves for the groups according to the expression of five stroma-related markers classified by hierarchical clustering in patients (A) without or (B) with post-operative chemotherapy. Note that the shorter survival of the DCN-dominant group with chemotherapy is significant.
In the present study, hierarchical clustering analyses successfully classified CRC patients into three groups with specific clinicopathological and immune cell characteristics using five stromal factors. Several studies have reported the classification of CRC using hierarchical clustering analyses [33,34]; however, in these studies, the stromal and tumour cell characteristics as well as molecular and clinical information were intermingled for the analyses. Recently, pancreatic cancer, one of the most stroma-rich tumours, was classified according to stromal factors such as collagen, ACTA2, and FAP, the latter two of which were expressed in pancreatic cancer CAFs in a mutually exclusive manner [35]. For the hierarchical clustering analyses of immunohistochemical data, proteins that are expressed in a mutually exclusive manner may be suitable.
DCN has been reported to suppress CRC proliferation and migration through the interaction with and stabilisation of E-cadherin in CRC cells [14]. In the present study, immunohistochemical analyses revealed undetectable and limited expression of DCN in non-neoplastic colonic epithelial cells and CRC cells (6% of CRC cases), respectively. Furthermore, patients with CRC-expressing cDCN tended to be associated with an unfavourable clinical outcome. To assess the correlation between E-cadherin and DCN, we immunohistochemically compared the membranous expression of E-cadherin in CRC cells with or without cDCN; however, significant associations were not detected in our cohort. In contrast, cDCN expression was inversely associated with PHH3, a cellular proliferation marker that predicts favourable clinical outcome in CRC [31]. Regarding stromal-expressed DCN, the worst clinical outcomes in the DCN-dominant group with post-operative chemotherapy was prominent. These observations indicate the reduced chemosensitivity of these patients due to the desmoplastic stroma. Taken together, although the mechanism should be identified in the future, the authors consider DCN to be a potentially unfavourable marker, at least for CRC patients.
pT stage-related expression was detected for collagen, DCN, and PDPN. Collagen and DCN expression were higher in advanced tumours. In contrast, PDPN-positive CAFs were the most abundant CAF type in pT2 tumours. These observations may be partly due to the different localisation of the DCN- and PDPN-positive CAFs. Following single-cell RNA sequencing of CRC, DCN-positive CAFs were indicated to be identical to CAF-A [9]. The significant associations between DCN and collagen expression support this notion because CAF-A have been reported to express COL1A2. Although DCN and PDPN were expressed in a mutually exclusive manner in CRC CAFs, it is unclear whether PDPN-positive CAFs are identical to CAF-B, which have been reported to be positive for ACTA2 [9], because the association between PDPN and ACTA2 was considered not so significant in our hierarchical clustering analyses. Furthermore, the different characteristics between PDPN-positive CAF and ACTA2-positive CAF have been indicated [15]. It is unclear where the differences in DCN- and PDPN-positive CAFs are initiated; however, the authors speculate that the heterogeneity of the originating cells [4–8], especially their physiological location (e.g. lamina propria, submucosa, or subserosa), has some impact on the characteristics of the differentiated CAFs. The heterogeneity of the CRC cells can be considered as an effector to modulate CAF characteristics.
The significance of cellular proliferation markers in the survival of CRC patients is controversial. Most studies conducted immunohistochemical analyses of proliferation markers such as Ki-67 to predict patient outcome; however, the results were conflicting, and some studies claimed no significant difference in or positive or inverse associations between the expression of cellular proliferation markers and clinical outcomes [36–40]. Our group analysed the expression of PHH3 and other cellular proliferation markers in CRC cells and identified an inverse correlation between CRC cell proliferation and patient survival [31]. In the present study, significant associations between the expression of CAF markers and cellular proliferation markers were detected: collagen and DCN-rich tumours showed lower proliferative activity, whereas higher PDPN expressors showed higher proliferative activity. On the one hand, based on the close communication between CRC cells and CAFs [41], the characteristics of CAFs may have effects on the proliferative activity of CRCs. On the other hand, it has been reported that the cellular proliferation of CRC cells decreases according to pT stage [28,31]. Based on the pT stage-related expression of DCN, collagen, and PDPN, rather than stromal characteristics, pT stage itself may affect the cellular proliferation of CRC cells.
FOXP3+ Tregs harbouring an immunosuppressive function have been reported to suppress the antitumour immune response and accelerate tumour immune evasion within the tumour microenvironment in many cancer types [42,43]. In contrast, the tumour-suppressive effects of Tregs with prolonged survival in Treg-containing tumours have been reported in several cancer types including CRC [44,45]. In CRC, inflammatory antimicrobial responses through Th17 cells have been considered to accelerate tumour promotion via angiogenesis and inflammatory reactions. A potential explanation for the favourable role of Tregs in CRC may be the attenuated pro-inflammatory and tumour-enhancing response of Th17 cells by Tregs [44]. Another possibility is the infiltration and the antitumour immune effects of FOXP3-lower-expressing non-suppressive T cells, functionally distinct subpopulations of tumour-infiltrating FOXP3+ T cells induced by interleukin-12 and transforming growth factor-beta secreted by surrounding tissues in a Fusobacterium nucleatum intestinal bacteria-dependent manner [46]. Similar to a previous larger cohort study [47], patients with DCN-dominant tumours containing significantly fewer FOXP3+ cells showed the worst prognosis in the present study. It is unclear why DCN-high tumours contained fewer FOXP3+ cells; however, the modulatory effects of DCN on the recruitment of leukocytes into the tissue may have some role in this phenomenon [11,12]. As another possibility, the non-proteolytic infiltration of lymphocytes may be suppressed by the tumour-associated dense collagen colocalising with DCN [48].
Studies describing the associations between CAF characteristics and CRC molecular subtypes or CRC gene mutations are sparse [49–51]. Recently, it was reported that consensus molecular subtype 4 (CMS4, mesenchymal type) [52] CRC has a high CAF content expressing DCN in their microenvironment [51]. At the same time, this report identified that a subset of CMS1 (immune type with frequent BRAF mutation) and CMS2 (canonical type with TP53 mutation) CRCs also had a high CAF content and the CAF subtypes were not clearly different between the CRC molecular subtypes [51]. Regarding gene mutation, in the present study, DCN expression was higher in CRCs carrying BRAF or KRAS mutations than in tumours without them (p = 0.036). To best of our knowledge, associations between CRC mutations and DCN expression have not been reported. Cancer cells and surrounding fibroblasts communicate each other [41,53,54], and gene mutation status in cancer cells may affect the activation or differentiation of the surrounding fibroblasts [50,54]; however, the mechanisms underlying this have not been fully elucidated. In pancreatic cancer, it has been reported that mutated p53 in cancer cells, along with the activation of Janus kinase 2-signal transducers and activators of transcription (STAT) 3 signalling, influences the collagen production of pancreatic stellate cells by paracrine stimulation [55]. In the present study, p53 immunohistochemistry, a surrogate marker to predict TP53 mutation status, was performed; however, no correlation with CAF marker expression was detected. Further investigations including co-culture experiments using colonic epithelial cells with induced gene mutations and fibroblasts or CAFs may reveal the precise mechanisms in the near future.
The limitations of this study include the number of CRC patients. Classification into three groups was performed using hierarchical clustering analyses. A larger cohort with gene expression and mutation analyses might identify better and more precise classification of CRC patients as well as validate our present results. In the present study, serial sections of tumour tissue arrays were used in the cohort analyses. Three-colour fluorescent immunohistochemical staining was performed in a representative case; however, the application of multi-colour immunohistochemical staining in whole slide sections might be useful for the further characterisation of tumours, CAFs, and tumour-infiltrating immune cells within the tumour microenvironment.
In the present study, CRCs were classified according to the expression of five stromal markers. Univariate analyses using independent stromal markers identified clinicopathological characteristics and a unique association between DCN expression and patient survival. Furthermore, hierarchical clustering analyses using all the markers successfully identified survival models for CRC patients. For the mechanistic aspects, advanced pT stage, frequent peritoneal and distant metastasis, and decreased CD8+ and FOXP3+ cells in the DCN-dominant group were identified. In future, tumour stroma-targeting therapies may be candidate treatments for patients with CRC.
AcknowledgementsWe thank Naoki Igari (Aichi Medical University), Taeko Yamauchi, and Koji Kato (Nagoya City University) for their assistance with tissue preparation and immunohistochemical staining.
This work was supported by a Grant-in-Aid for Scientific Research (C) (to ShI, 20K07410 and 23K06446) from the Japan Society for the Promotion of Science.
Author contributions statementSuI and ShI prepared the figures and tables and wrote the manuscript. SuI, AK, KeK and ShI performed the histopathological and immunohistochemical analyses. AK, ME, NO and KuK collected and analysed the clinical data. SuI, MK, TO, CW, AU, HT, ShT and ShI performed the statistical analyses. TT provided archival tissue samples from Aichi Medical University Hospital. KeK and SaT provided the research facilities. ShI conceived, designed and supervised the overall study. All authors have read the final manuscript and given their final approval to the submitted version.
Data availability statementThe datasets used and/or analysed during the present study are available from the corresponding author on reasonable request.
© 2024. This work is published under http://creativecommons.org/licenses/by-nc-nd/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
; Kasai, Kenji 6 ; Kasugai, Kunio 2 ; Takiguchi, Shuji 1 ; Takahashi, Satoru 3 ; Inaguma, Shingo 7
1 Department of Gastroenterological Surgery, Nagoya City University Graduate School of Medical Sciences, Nagoya, Japan
2 Division of Gastroenterology, Department of Internal Medicine, Aichi Medical University School of Medicine, Nagakute, Japan
3 Department of Experimental Pathology and Tumor Biology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Japan
4 Department of Public Health, Nagoya City University Graduate School of Medical Sciences, Nagoya, Japan
5 Surgical Pathology, Aichi Medical University School of Medicine, Nagakute, Japan
6 Department of Pathology, Aichi Medical University School of Medicine, Nagakute, Japan
7 Department of Experimental Pathology and Tumor Biology, Nagoya City University Graduate School of Medical Sciences, Nagoya, Japan; Department of Pathology, Aichi Medical University School of Medicine, Nagakute, Japan; Department of Pathology, Nagoya City University East Medical Center, Nagoya, Japan