Detection, Role and Prognostic Value of Cancer-associated Fibroblasts in Colorectal Cancer

Abstract

Cancer-associated fibroblasts (CAFs) — mesenchymal cells in the tumor stroma, play one of the leading roles in tumor progression in many different tumors, including colorectal cancer. Scientists have described many markers for CAFs, but none of them is specific. We performed immunohistochemistry tests using five antibodies (αSMA, POD, FAP, PDGFRα, PDGFRβ) to investigate CAFs in three zones of 49 colorectal adenocarcinomas: apical, central, and invasive edge. We revealed the reliable correlation between high PDGFRβ and PDGFRα value in the apical zone and deeper invasion (T3–T4) (p = 0.0281 and p = 0.0137). High αSMA level in apical zone (p = 0.0001), αSMA level in central zone (p = 0.019), POD level in apical zone (p = 0.0222), POD level in central zone (p = 0.0206) and PDGFRβ level in apical zone (p = 0.014) correlated reliably with the presence of metastasis in lymphatic nodules. For the first time, focused on the inner layer of CAF adjacent to tumor complexes. We observed that cases with inner αSMA expression were significantly more often (p = 0.023) characterized by the presence of regional lymph node metastasis compared with cases with mix of CAF markers (p = 0.007) and with cases with inner POD expression (p = 0.024). The found relationships between the level of markers and the presence of metastases indicate their clinical significance.

Introduction

Tumor stroma or tumor microenvironment (TME) includes the immune cells, capillaries, basement membrane, activated fibroblasts, and extracellular matrix (ECM) surrounding the cancer cells [1]. Today experts see that cancer cells’ autonomous genetic defects and TME control tumor progression and metastasis [2]. One of the critical components of the TME is cancer-associated fibroblasts (CAFs) — non-epithelial non-immune cells with mesenchymal characteristics located in the tumor stroma or along its edge [3]. In contrast to customarily activated fibroblasts, CAFs are characterized by increased synthesis of growth factors and chemokines. Also, CAFs have more evident proliferative activity and the ability to migrate. There are some data [1] on the existence of cancer-associated mesenchymal stem cells (MSCs) in the CAF population. CAFs play one of the leading roles in tumor progression and provide a wide range of fibrotic stromal changes of most known tumors. For example, CAFs produce ECM-degrading proteins — matrix metalloproteinases (MMPs), providing the migration of cancer cells. CAFs are described in many different tumors, including colorectal cancer [3].

A generally accepted marker of activated fibroblasts, CAFs in particular, is α-smooth muscle actin (αSMA), which is also expressed in smooth muscle cells and fibroblasts surrounding non-tumoral crypts [4]. Other markers are considered to be fibroblast activation protein (FAP), platelet-derived growth factor receptors α and β (PDGFRα and PDGFRβ), podoplanin (POD) [5]. All these markers are not specific for CAFs. Thus, the inflammatory cells also express FAP, PDGFRs are found on pericytes [1, 6] and POD in lymphatic vessels. Several markers used to identify CAFs are typical for customarily activated (non-resting) fibroblasts surrounding non-tumoral crypts.

Tumors are sometimes described as “unhealing wounds” since a long-term accumulation of cancer cells in any tissue initiates a chronic reparation reaction. The role of myofibroblasts in wound healing is well researched and understood, but their functional role in tumor progression and metastasis is complex and controversial. The results of several studies demonstrate that CAFs can not only potent the tumor progression but also inhibit cancer cell proliferation and growth [7–9]. The term “tumor budding” is generally understood to mean the presence of isolated single cancer cells or clusters of up to four cancer cells located at the invasive tumor front colorectal carcinoma (CRC). Tumor budding was first announced as an independent prognostic factor for early carcinomas in 2016 and was included in the obligatory characteristics of CRC by WHO guidelines in 2019 [10]. The presence of tumor buds in the invasive tumor front is associated with the loss of cancer cell junctions and increased motility. These features represent the epithelial-mesenchymal transition, which plays an essential role in tumor invasion and metastasis [11].

Our study aims to identify CAFs in CRC by various markers, evaluate the colocalization of markers, their relative position with tumor buds, and search for relationships with clinical and pathological characteristics CRC.

Material and Methods

The research material consisted of 49 colon adenocarcinomas. Only surgical material was used in the work, not biopsies. The study did not include material from patients who received neoadjuvant therapy before surgery. The postoperative colon was delivered in a 10% buffered formalin solution, buffered, at a pH of 7.4–7.6. The surgical material was grossed in such a way that a tumor along the entire thickness and the adjacent intact mucous membrane of the colon were presented in one sample (paraffin-embedded tissue).

We performed immunohistochemistry tests to investigate CAFs using monoclonal antibodies directed against the antigens POD (Abcam, clone EPR22182), FAP (Abcam, clone SP325), PDGFRα (Abcam, clone EPR22059-270), and PDGFRβ (Abcam, clone Y92). We used Pan-Cytokeratin (PCK, Dako, rabbit, polyclonal) to contrast and identify tumor buds. We used the duplex mark technology. Two reactions were carried out on each slice in parallel: with one of the studied CAFs marker and with anti-αSMA antibodies. The duplex mark was detected using Double Stain IHC Kit: M&R on human tissue (HRP/Green&AP/Red, Abcam ab210061) according to the recommended manufacturer’s method. Preprocessing (deparaffinization, rehydration, and antigen retrieval) was carried out in PT-Module (Thermo Scientific, England) for 20 min (95–98 °C) at pH 8,0. Endogenous enzyme blocking was carried out using VectorLab Bloxall Solution (20 min). The immunohistochemical reactions were carried out in a semiautomatic mode using Autostainer 480S (Thermo Scientific, UK).

We evaluated the obtained specimens using Leica DM4000B microscope, Leica DFC495 camera, and a specialized advanced morphometric analysis system Leica Application Suite X (LAS X). The markers were isolated and counted individually by color separation. In each case, we assessed with a magnification of the objective × 20 in three zones of the tumor: apical (closest to the lumen of the organ), central, and invasive edge. The area of interest is areas with a pronounced stroma and the most pronounced reactions of both markers in fusiform cells. We excluded the αSMA reaction in the vessel walls, the muscularis mucosae, and intestinal muscularis propria from the calculation.

The Pearson’s linear correlation coefficient was used as a measure of the markers’ colocalization.

The significance of differences in markers expression levels was established using parametric and, in some cases, nonparametric methods. One- and multi-factor analysis of variance (ANOVA) and the F-test were used as parametric methods. The distribution pattern was checked using the Shapiro–Wilk and Kolmogorov–Smirnov tests. The median and Mann–Whitney U-test were used for pairwise comparison as nonparametric methods; for multiple comparisons, the Kruskal–Wallis test. Linear discriminant analysis was used to establish threshold values.

All procedures performed in the current study were approved by the Ethics Committee of Lomonosov Moscow University (#3/17 by 17.04.2017) in accordance with the 1964 Helsinki Declaration and its later amendments.

Results

Baseline Characteristics

The study includes mainly patients aged 61–80 with TNM stage T3. Morphologically all of them have nonspecific low-grade adenocarcinoma, in some cases with a mucinous component (Table 1).

Table 1.

Clinical and pathological characteristics of the cohort

	Men	Women	Total
Age range
41–50	0	6	6
51–60	2	3	5
60–70	7	7	14
71–80	5	14	19
Above 80	2	3	5
Total	16	33	49
T (tumor)
T1	2	0	2
T2	2	4	6
T3	6	24	30
T4	6	5	11
Total	16	33	49
N (nodes)
N0	4	15	19
N1a	2	4	6
N1b	2	7	9
N1c	2	4	6
N2a	3	2	5
N2b	3	1	4
Total	16	33	49
M (metastasis)
M0	14	32	46
M1	2	1	3
Grade
Low
High
Total	16	33	49
Presence of mucinous component
Yes	1	5	6
No	15	27	43
Total	16	33	49
Localisation
Cecum	2	4	6
Ascending colon	3	2	5
Transverse colon	3	3	6
Sigmoid colon	2	13	15
Rectosigmoid junction	2	4	6
Rectum	4	7	11
Total	16	33	49

Comparison of the Expression Level of CAFs Markers Between Each Other in Different Zones

The studied markers POD, FAP, PDGFRα, PDGFRβ were not expressed in the adjacent intact mucous membrane of the colon in all cases. αSMA reaction was observed in muscularis mucosae and crypt fibroblasts, which formed the edging of each crypt.

Multi-factor analysis of variance highlighted that αSMA expression levels in each of three tumor zones differed significantly (significance level p = 0.0017). In particular, αSMA expression level was significantly higher in the invasive edge and central zone compared to the apical zone. POD level also differed depending on the zone (significance level p = 0.00006). However, the relation was different: POD level was significantly higher in the apical zone than in the central zone and invasive edge. We found no significant difference between expression levels of other markers in different tumor zones.

We noted higher rates of PDGFRβ, POD, αSMA and low rates of PDGFRα and FAP when comparing the values of markers between each other.

Correlation of Marker Levels and T

The reliable correlation between high PDGFRβ value in the apical zone and deeper invasion (T3–T4) was revealed (p = 0.0281) using the median test. A similar correlation was observed between high PDGFRα value in invasive edge and deeper invasion (p = 0.0137).

Correlation of Marker Levels and N

The Kruskal–Wallis test showed multiple correlations between studied markers’ expression levels and the presence of metastasis in regional lymph nodes (Fig. 1). High αSMA level in apical zone (p = 0.0001), high αSMA level in central zone (p = 0.019), high POD level in apical zone (p = 0.0222), high POD level in central zone (p = 0.0206), and high PDGFRβ level in apical zone (p = 0.014) correlated reliably with the presence of metastasis.

Fig. 1.

Identified relationships between CAF markers expression and the presence and number of metastases in regional lymph nodes (N). a Apical αSMA expression grouped by lymph node metastasis (p < 0.01); b Central αSMA expression grouped by lymph node metastasis (p < 0.01); c Apical POD expression grouped by lymph node metastasis (p < 0.01); d Central POD expression grouped by lymph node metastasis (p < 0.01); e Apical PDGFRb expression grouped by lymph node metastasis (p < 0.01 only between N1 and N2)

Correlation of Marker Levels and M

The Kruskal–Wallis test showed the correlation between high POD level in invasive edge and the presence of distant metastasis (M1, p = 0.0264).

Correlation of POD Level and Tumor Buds

Tumor buds — isolated single cancer cells or clusters of up to four cancer cells located at the invasive tumor front; were found in 53.5% of cases. We used POD labeling to identify CAFs around the tumor buds.

A chi-squared test was used to compare POD expression levels around the tumor buds and in invasive edge with no tumor buds. The analysis did not show any significant differences between groups (p > 0.01). This underlines the uselessness of evaluation of POD expression exactly around the tumor buds and that there is no tropism of POD to tumor buds.

There was a reliable correlation (p = 0.023) between the presence of tumor buds and deeper invasion (TNM stages T3–T4). We showed the trend towards the correlation (p = 0.068) between the presence of tumor buds and the presence of regional lymph nodes metastasis, but not their amount.

Also, we showed the trend towards the correlation between expression of POD around the tumor buds and the tumor depth of invasion (p = 0.088), but not the presence of metastasis. Interestingly, no significant correlation was observed between the expression of POD in invasive edge with no tumor buds and clinicopathological characteristics.

Correlation of Marker Levels and the Presence of Mucinous Component

A chi-squared test showed no significant differences between POD expression levels in tumors with and without mucinous component (p = 0.678), neither in the invasive edge nor around isolated tumor buds. However, we noticed a marked expression of POD directly around extracellular mucin in adenocarcinomas with mucinous features. Due to the small size of the cohort, we could not confirm this pattern statistically (p = 0.181). However, the obtained value may indicate a possible statistical trend, and the pattern can be established on a larger cohort.

We have not found studies about the examination of POD in the stroma of mucinous adenocarcinomas. However, Oe S. and co-workers (2010) [12] showed a higher frequency of POD expression in clear cell carcinoma of the ovary compared to other histological types. This fact confirms the potential tropism of POD to tumors containing polysaccharides.

Colocalization of Markers

We used correlation analysis to find the correlation between markers expression in each zone of the tumor (colocalization). These tests showed strong inverse relationship between αSMA and POD levels in apical zone (r = 0.394, p = 0.006). Also, we found a little less strong relationship between PDGFRβ/PDGFRα in invasive edge (r = 0.468, p = 0.043).

Distribution of Markers in Relation to Tumor Complexes

During image analysis, in addition to the labeling area, we paid attention to the location of each of the pairs of markers relative to the tumor complexes (Fig. 2). These tests highlighted that the inner (nearest to the tumor complexes) layer was different in different cases. Thus, in 7.14% of cases in the POD/αSMA pair, POD turned out to be the nearest; in 40.48% — αSMA (Fig. 2a), and 52.38% of cases, a mixture of markers was observed (in some foci, POD turned out to be the nearest, in others — αSMA, in some fields of view, it was not possible to determine the nearest marker). A similar assessment of the nearest marker in the PDGFRβ/αSMA pair showed that PDGFRβ was the closest at 7.14% (Fig. 2b), αSMA — at 38.09%, and the mixed type at 54.77% (Fig. 2d). We noted that in 93.75% of cases where the nearest layer was αSMA in the POD/αSMA pair, in the PDGFRβ/αSMA pair, αSMA was also the closest. The match for the nearest marker POD-PDGFRβ was slightly lower, 66.67%, the match for the mixture of markers was 65.22% (Table 2).

Fig. 2.

IHC. Differences in the location of the CAFs marker reaction relative to tumor complexes. A Reaction with POD (red) and αSMA (brown), × 200. The nearest CAFs marker is αSMA. B Reaction with PDGFRβ (red) and αSMA (green), × 200. The nearest CAF marker is PDGFRβ. C Reaction with FAP (red) and αSMA (green), × 200. The nearest CAF marker is mixed. D Reaction with PDGFRβ (red) and αSMA (brown), × 200. The nearest CAF marker is mixed

Table 2.

Distribution of markers in relation to tumor complexes

Pair of markers	Nearest to epithelial complex marker (%)		Matching	Nearest to epithelial complex marker (%)	Matching	Nearest to epithelial complex marker (%)	Matching
	POD	PDGFRβ	66.67%	αSMA	93.75%	mixed	65.22%
POD/αSMA	7.14	-		40.48		52.38
PDGFRβ/αSMA	-	7.14		38.09		54.76

Discussion

The established differences between the expression of each marker in different zones indicate significant heterogeneity of CAFs in CRC. The lack of a typical pattern in the distribution of all markers described as CAFs markers means potentially different detectable subpopulations of cells. On the other hand, it shows that it is impossible to identify all CAFs using only one marker. Indeed, the question of whether all these markers belong to CAFs remains unclear. The present paper is based on comparing the studied markers with αSMA, as it is the most conventional marker of CAFs. However, the presence of αSMA also in blood vessels, muscle tissue, and, possibly, in several mesenchymal elements makes it difficult to identify specifically CAFs. Even though we selected fields of view with minimal “extra” elements, and we did not include the inflammatory infiltrate, muscular elements of the intestinal wall, and vessels in the estimated area, it is impossible to exclude all non-CAF elements completely.

We believe that the increased POD level in the apical part may be associated with an increased number of young vessels and fibroblasts. They are the basis of granulation tissue in the part of the tumor surrounding the ulcerative defect, which was observed in 90.5% of cases in the study and accompanied most cases of colorectal cancer. Interestingly, a high POD level in the apical part was associated with many metastases in regional lymph nodes (N2). In contrast, a high POD level in the central part was associated with metastasis in both N1 and N2 groups of tumors. Using discriminant analysis, we established threshold values for POD in the apical part — 2.5368 (p = 0.15) and in the central zone — 1.1548 (p = 0.207). However, these values in the studied cohort are not statistically significant. A larger cohort is likely needed to establish true threshold values for the level of POD expression. The high POD in the invasive edge was also significant since high values in this area were associated with distant metastasis. In the literature, there are conflicting data regarding the role of POD in tumor progression: Kubouchi Y. et al. [13] showed that these CAFs were associated with a poor prognosis in adenocarcinoma, including CRC. In contrast, Yamanashi et al. [14] и Song-Yi C. [15] have shown that PDPN + CAFs correlate with a favorable prognosis in CRC. Our results indicate that the main prognostic value is the presence of tumor buds, and not the POD reaction around them.

The opposite picture was observed with αSMA. Maximum expression of αSMA was detected in the invasive edge and central zone. We considered invasive edge as a site of direct invasion into the adipose tissue (where, according to the latest recommendations, the number of tumor buds should be counted [10]). In this area, there is no granulation tissue and only few vessels that could cause false high values, and the rarely observed inflammatory infiltrate does not express αSMA. In this regard, the obtained data appear to be correct. Using discriminant analysis, we established statistically significant threshold values for the αSMA expression level: in the apical zone — above 1.9464 (p = 0.0285) and in the central zone — above 3.6791 (p = 0.0331).

Considering that both high αSMA and high POD values in the apical zone are poor prognostic factors, the result of correlation analysis seems to be logical, as it demonstrates a direct relationship between these markers in the apical zone.

Our finding of a direct link between PDGFRβ and PDGFRα in invasive edge, which was obtained using correlation analysis, contrasts with previous results reported in the literature. For instance, the authors point out that a high level of PDGFRα is associated with a good prognosis, while a high level of PDGFRβ is associated with a poor prognosis in most solid tumors. For this reason, some authors interpret PDGFRα as a marker of a tumor-suppressive population of CAFs [16]. Our direct correlation between PDGFRβ and PDGFRα relates them to the same population of CAFs. Probably, the obtained variation is associated with the assessment of the expression of markers by zones and not for the entire specimen. Therefore, we feel strongly that our findings more precisely represent the biology of CAFs subpopulations and their distribution in a tumor.

Attention to the proximal relative to tumor complexes (nearest) layer of CAFs is drawn for the first time. We believe that the demonstrated differences seem to be extremely interesting since, probably, it is the nearest layer of fibroblasts that has a more significant influence on the epithelial structures, determining the behavior and progression of the tumor (Fig. 2). Remarkably, the nearest layer of CAFs is similar in different pairs of markers. In particular, αSMA appeared to be the closest in 93.75% of cases in both pairs of markers (Fig. 2a). Using contingency tables and chi-squared test, we have established that cases of colorectal cancer, in which αSMA expression appeared to be the nearest to tumor complexes, were significantly (Fig. 3) more often (p = 0.023) characterized by the presence of regional lymph node metastasis compared with cases in which there was a mix of markers (p = 0.007) and with cases in which POD expression was in the foreground (p = 0.024). It seems likely that a search for correlations between prognostic clinical and pathological characteristics and the inner layer of fibroblasts may be promising.

Fig. 3.

The relationship between the characteristics of the expression of markers and the number of lymph node metastases

Conclusions

We have carried out the identification and assessment of CAFs level using five markers: αSMA, FAP, PDGFRα, PDGFRβ и POD. Considering that the different CAF markers had different level in central/apical tumor zone, this suggests that a single fibroblast marker cannot recapitulate the heterogeneous composition of CAFs in the tumor stroma. Despite the marked heterogeneity of the distribution of markers depending on the tumor zone, we have established a higher POD expression level in the apical area and αSMA in the invasive edge. The found relationships between the level of markers and the presence of metastases indicate the importance of the value of each of the markers and their clinical significance. For the first time, focused on the inner layer of CAF adjacent to tumor (epithelial) complexes. It became possible due to identifying CAFs using duplex labeling technology on one section. The obtained statistical tests indicate the clinical significance of the inner layer: cases with inner αSMA were significantly more often (p = 0.023) characterized by the presence of regional lymph node metastasis compared with cases in which there was a mix of markers (p = 0.007) and with cases in which POD expression was in the foreground (p = 0.024). Taken together, these findings stimulate additional studies of CAFs and their features on a larger cohort better to understand the molecular biological characteristics of colorectal cancer.

Author Contribution

Nina Alexandrovna Oleynikova, MD, PhD — Conceptualization, Data Curation, Investigation, Methodology, Project Administration, Validation, Visualization

Ilya Alexandrovich Mikhailov, graduate student — Formal Analysis, Review & Editing

Timofey Yurevich Zavidnyi, student — Data Curation, Investigation, Writing (Original Draft Preparation)

Nataliya Vladimirovna Danilova, MD, PhD — Resources, Visualization, Review & Editing

Olga Andreevna Kharlova, MD, PhD — Funding Acquisition, Resources

Pavel Georgievich Malkov, MD, PhD, Professor — Project Administration, Software, Supervision

Funding

We thank all the participants for their contribution to the study. This work was sponsored by Russian Foundation for Basic Research (RFBR, grant «Perspektiva» № 19-315-60006) and was carried out using equipment purchased under the development program of the Lomonosov Moscow State University until 2021.