CD44 epithelial isoform inversely associates with invasive characteristics of colorectal cancer

Colorectal cancer (CRC) is one of the leading causes of cancer-related deaths throughout the world. In spite of the advent of multimodality approaches to the management of CRC, the global burden of CRC is expected to increase up to 60% by 2030 [1].

The insight that cancer originates from a subpopulation of tumor cells, named cancer stem cells (CSC), has generated considerable enthusiasm among researchers. Compelling evidence supports the pivotal role of CSC in the initiation and maintenance of the tumor, as well as their capability to dictate invasion, metastasis and therapeutic resistance [2]. Accordingly, several clinical applications including therapeutic, diagnostic and prognostic implications have been introduced for CSC [3–5].

A variety of CSC markers have been proposed for CRC [6,7]. Among them, CD44 and CD24 are two of the most investigated colorectal CSC (CCSC) markers and CD44⁺/CD24⁺ cells are introduced as the subpopulation with a greater clonogenic ability and tumor initiation potential [8].

CD44 is a transmembrane glycoprotein with several isoforms which is involved in several cellular functions including angiogenesis, cell–cell interactions, cellular adhesion and cell migration [9]. Its overexpression has been noticed as an early event occurring prior to the transformation of colorectal adenoma to carcinoma [10]. CD44 knockdown prevents tumor formation [11]. Moreover, injection of just a single CD44⁺ cell is able to form tumor sphere and develop into the tumor if xenografted into nude mice [11].

CD24 is also a transmembrane glycoprotein which functions in cell–cell and cell–matrix interactions. Despite its expression in many different cancer subtypes, there is much ambiguity regarding its role as the CSC marker, mainly due to its versatile physiological functions [2].

The paradigm of CD44 and CD24 as the CCSC markers has stimulated numerous studies to translate this information into the clinical implications such as prognosis [12–14]. Yet, there is no consensus regarding the prognostic significance of these markers in CRC [15,16]. Thus, further studies are needed to reach a consensus in this regard.

Here, we aimed to evaluate the local expression of CD44 and CD24 in association with the clinicopathologic characteristics of disease in a large cohort of CRC patients to explore the role of these markers as CSC more comprehensively.

Patients & methods

In a retrospective study, formalin-fixed paraffin-embedded (FFPE) tissues from 494 patients with biopsy-proven colorectal adenocarcinoma who were operated at our center from 2008 to 2017 were included. Patient demographics, clinical and pathologic characteristics were extracted from the available database which included the gender, age, location of the tumor, tumor size, pathological Tumor, Node, Metastasis (pTNM) stage, lymphovascular invasion, neural invasion, metastasis status and tumor differentiation.

Construction of tissue microarrays

To construct tissue microarray (TMA) blocks, hematoxylin and eosin-stained slides were first reviewed by a pathologist and three different regions that were the most representative of the tumor were marked on the slides. The marked regions were aligned with the corresponding FFPE block and using a tissue arrayer (Minicore; ALPHELYS, Plaisir, France) cores with 0.6-mm diameter were harvested from donor blocks and transferred into a recipient block. Accordingly, three copies of TMA blocks were constructed, each containing a sample from a different region of the same FFPE block. These TMA blocks were used for immunohistochemical staining. The average score of the three samples was considered the final score for each tumor.

Immunohistochemical staining

Immunohistochemistry (IHC) was performed on 4-μm thick sections of colorectal cancer TMA slides (Superfrost plus, Thermo Scientific, Braunschweig, Germany). Briefly, the sections were deparaffinized in xylene and rehydrated by passing through the ethanol series. Subsequently, the endogenous peroxidase activity was blocked by incubating the slides in 3% hydrogen peroxide for 20 min at room temperature, followed by antigen retrieval was done by autoclaving slides in citrate buffer (pH = 6.0) for 10 min. After three washes with Tris-buffered saline, the sections were incubated overnight at 4°C with specific primary antibodies against human CD44 (rabbit monoclonal anti-CD44 epithelial isoform, ab51037, Abcam, Cambridge, UK) and CD24 (mouse monoclonal anti-CD24, ab31622) with the optimal dilution of 1/100 and 1/800, respectively. After washing, the TMA slides were incubated with anti-rabbit/anti-mouse Envision (Dako, Denmark) as the secondary antibody for 10 min. The IHC slides were visualized with 3, 3′-diaminobenzidine (DAB, Dako, Denmark) substrate as chromogen for 15 min at room temperature (RT). Then, the sections were lightly counterstained with hematoxylin, dehydrated through a graded ethanol series followed by xylene, and finally mounted. As negative controls, the slides were incubated with Tris-buffered saline instead of the primary antibody.

Evaluation of immunohistochemistry slides & scoring

Stained slides were scored on a multiheaded microscope by two observers Asieh Sadeghi and Maryam Abolhasani (AS and MA). The observers were blinded to the clinical and pathologic characteristics of the samples. In difficult cases, the score was given based on a consensus between the two observers. The slides were initially scanned at 10× magnification to obtain a general impression of the overall distribution of the stain. The positive cores were then assessed for localization and scored semiquantitatively at higher magnifications. For semiquantitative scoring, the intensity of staining was scored as 0 (absent), 1 (weak), 2 (moderate) or 3 (strong) (Figure 1). Also, the percentage of positive tumor cells were scored as 1 (positive tumor cells <25%), 2 (positive tumor cells between 25–50%), 3 (positive tumor cells more than 50% and ≤75%) and 4 (positive tumor cells >75%). The evaluation of histochemical score (H-score), a range from 0 to 300, was achieved by multiplying the intensity of staining and the percentage of positive tumor cells [17]. The samples were categorized as low (H-score <150) or high (H-score ≥150) expression as well. The mean of H-score as a cut-off value, has been used in the several previous studies to categorize the tumors with various expression levels of markers [18–22]; therefore, in the present study, the mean of H-score of (cut-off = 150) has been used to categorize samples as with high (H-score >150) or low (H-score ≤150) CD44 or CD24 expression. Although we agree that ROC curve will help to determine the cut-off of H-score, but the follow-up data could not be applicable for our study in this step, as the data are not available (the study design was a retrospective design).

Statistical analysis

Statistical analyses were performed using SPSS for Windows, version 16. Descriptive data were presented as the mean ± standard deviation or number and percentage. The association of the CD44 and CD24 expression with the clinicopathologic characteristics of the patients was evaluated once using the H-score of the markers (t-test or one-way ANOVA for parametric and Mann–Whitney U test and Kruskal–Wallis H test for nonparametric variables) and once based on the expression pattern (high or low) of the markers (χ² test). Pearson's or Spearman's correlation coefficient tests were used for the evaluation of potential correlations. Median split approach was used for the categorization of quantitative variable such as the tumor size. A p-value of ≤0.05 was considered statistically significant.

Results

Study population

The expression of CD44 and CD24 using IHC was evaluated in 494 biopsy-proven CRC samples that had undergone colectomy. The patients’ population included 259 males and 235 females with the mean age of 59.4 ± 14.9 years. The mean tumor size was 5 ± 2.7 cm. The pathologic stages II and IV were, respectively, the least and most frequent pTNM stages of this study. Metastatic CRC was present in 28 cases.

CD44 expression

The mean CD44 H-score was 99 ± 62.3. Based on this H-score, CD44 expression status was found to be low in 319 (64.6%) cases and high in 74 (15%) cases. CD44 expression was not measurable in the remaining 101 (20.4%) cases due to the considerable section detachment from the slides. No significant association was found between the immunoreactivity pattern of CD44 and clinicopathologic characteristics of the patients.

The CD44 H-score was significantly higher in lower stages (p = 0.038). In this regard, the mean CD44 H-score was 116.2 ± 72.7 in stage I, 100.9 ± 61.8 in stage II, 90.5 ± 57 in stage III and 83.7 ± 53 in stage IV of CRC (Table 1). A significant inverse correlation was also found between the CD44 H-score and stage of CRC (r = -0.148; p = 0.004). No other significant correlation was observed between the CD44 H-scores and clinicopathologic characteristics of the tumors. Moreover, the CD44 H-score was considerably lower in patients with lymphatic invasion in comparison with those without lymphatic invasion (91 ± 54.8 vs 104 ± 65.9). This difference was marginally significant (p = 0.05). No other significant association was found between the CD44 H-score and clinicopathologic characteristics of the tumors.

CD24 expression

The mean CD24 H-score was 147 ± 90.5. Accordingly, CD24 expression status was found to be low in 217 (43.9%) cases and high in 193 (39.1%) cases. CD24 expression was not measurable in the remaining 84 (17%) cases due to the considerable section detachment from the slides. No significant association was found between the CD24 expression pattern and clinicopathologic characteristics of the tumors (Table 2). No significant association was found between the CD24 H-score and clinicopathologic characteristics of the patients. No significant correlation was observed between the CD24 H-scores and clinicopathologic characteristics of the tumors as well. Moreover, no significant correlation was observed between CD24 and CD44 H-scores (r = 0.015; p = 0.78).

Combined CD44 & CD24 expression

In 352 out of 494 cases, both CD44 and CD24 expression was measured. The groups were categorized as CD44^lowCD24^low (150 cases, 30.4%), CD44^highCD24^low (35 cases, 7.1%), CD44^lowCD24^high (132 cases, 26.7%) and CD44^highCD24^high (35 cases, 7.1%). No significant association or correlation was found between CD44/CD24 status and clinicopathologic characteristics of the patients.

Discussion

Prognostic markers assist the clinician in planning the most efficacious and least invasive therapeutic strategy for each patient, whereas many prognostic markers have been proposed for CRC, several attempts are being made to find more reliable prognostic markers [23,24]. Recently, considerable interest has been devoted to the application of CSC markers as prognostic markers in different tumors, including CRC [15].

Here, we studied the prognostic significance of CD44 and CD24, as the two potential CCSC markers, in a relatively large cohort of CRC patients. Although many earlier investigations have reported an association between CD24 expression and invasive characteristics of the CRC, CD24 individually or in combination with CD44 showed no prognostic significance in CRC cases of this study.

Choi et al. studied the local expression of CD24 on 307 CRC patients using IHC. According to their results, CD24 expression was associated with several unfavorable clinical parameters of disease including tumor size, pTNM stage, and tumor invasion [25]. Weichert et al. also evaluated the immunoreactivity of CD24 in 147 colorectal carcinomas and two colon carcinoma cell lines using IHC. Based on their results, strong cytoplasmic CD24 expression correlated significantly to the shortened survival in cases without distant metastases [26]. By contrast, the study of Ahmed et al. who assessed the CD24 expression with IHC in ten whole tissue sections of adenoma and 345 CRCs, showed that in spite of the upregulation of CD24 as an early and common event during the development of CRC, it was not a prognostic marker in CRC [27].

Our results were in accordance with the study of Ahmed et al. Based on our results, CD44 was inversely associated with the stage of CRC and lymphatic invasion. In this respect, the higher expression of local CD44 was seen in lower stages of CRC. Moreover, in patients with positive lymphatic invasion, the local expression of CD44 was lower. As CSCs are established as a subset of tumor cells responsible for tumor initiation, progression, invasion, migration and metastasis [28], rationally a higher percentage of CSCs and subsequently an overexpression of the CSC markers are expected in more advanced cancers. Although this rationale has been supported in the majority of earlier investigations, the inverse association of CSC markers with invasive characteristics of the tumor has also been reported. According to the study of Jing et al., the 5-year disease-free survival rate of the patients with high local CD44 mRNA expression was significantly lower than that of the patients with low expression [29]. Many other investigations have demonstrated a prognostic link between tumor invasive characteristics of CRC and different CD44 isoforms [30,31]. By contrast, the study of Hoda et al., who evaluated the local expression of CD44 by IHC in 60 CRC patients, revealed an inverse correlation of CD44 with tumor invasiveness, lymphatic node status and TNM stage, although not statistically significant [32]. Our results were in accordance with the study of Hoda et al. The observed disagreement regarding the direct or inverse association of CD44 with tumor invasive characteristics could be attributed to the isotype switch as described by Biddle et al. The study of Biddle et al. revealed that cancer stem-like cells with an epithelial phenotype express variant isoforms of CD44, whereas in cancer stem-like cells that have undergone an epithelial-to-mesenchymal transition these variant isoforms are downregulated and the standard CD44 isoform is upregulated [33]. Based on this report, tumor bulk is composed of heterogeneous CSC populations with different CD44 isoforms, and the results of different investigations are only comparable if they evaluate the same CD44 isoform. In this study, we evaluated the epithelial isoform of CD44 (isoform 10) which is reported to be predominantly expressed in colon cancer [34]. However, we did not find any other similar study evaluating this isoform. Thus, further isoform-specific studies are needed to extensively clarify the prognostic role of CD44 in CRC.

It is well known that stage and lymph node metastasis are the two main important parameters for determination of the colorectal cancer patient's survival [35–37]. Although we agree that follow-up data will help to clear the role of these markers in colorectal cancer tumorigenesis and prognostic significance, but the collection of data could not be applicable for our study in this step, as the data are not available (the study design was a retrospective design).

Conclusion

Although CD24 individually or in combination with CD44 is not associated with the clinicopathologic characteristics of the tumor, CD44 epithelial isoform is inversely associated with the invasive characteristics of the tumor including pTNM stage and lymphatic invasion. Thus, expression of this isoform could be regarded as a marker of invasiveness in CRC patients.