TUMOUR SAMPLES

The study comprised a selected series of 102 sporadic colorectal adenocarcinomas obtained from 101 patients undergoing surgery at the Royal Brisbane Hospital between 1989 and 1999. Of these, 93 cases were derived from a cohort of 303 cancers previously characterised for clinicopathological and molecular features, including MSI,2 and nine MSI-H cancers were added to this series to increase the statistical power of comparison with this group. Altogether, 32 were MSI-H cancers (with band shifts in at least three of six markers), 30 were MSI-L (with band shifts in one or two microsatellite markers), and 40 were MSI-null or MSS. The study was approved by the Royal Brisbane Hospital Ethics Committee.

IMMUNOHISTOCHEMISTRY

Routinely processed paraffin embedded blocks were obtained from the Department of Pathology, Royal Brisbane Hospital. Tumour sections (4 μm) were affixed to adhesive slides (Menzelgläser, Braunschweig, Germany) and air dried overnight at 37°C. Sections were dewaxed in xylene and rehydrated through descending grades of alcohol to water, then transferred to Tris buffered saline (TBS), pH 7.4. Heat based antigen retrieval was performed using the technique of Shi and colleagues.15 Sections were boiled twice for five minutes each in 0.01 M citric acid buffer, pH 6.0, in a domestic 600 kW microwave oven, then cooled to room temperature, prior to washing in three changes of TBS for five minutes each. Endogenous peroxidase activity was quenched in 1% H₂O₂, 0.1% sodium azide in TBS. After washing three times for five minutes in TBS, non-specific antibody binding was blocked by incubating the sections in 4% commercial skim milk powder for 15 minutes. After a brief rinse in TBS the sections were placed in a humidified chamber and incubated for 20 minutes in non-immune goat serum (Zymed Corporation, San Francisco, California, USA). Primary antibody was applied at the appropriate titre and sections were incubated overnight at room temperature. The following antibodies were diluted in TBS: anti-CD3 (rabbit polyclonal, 1:300), anti-CD8 (murine monoclonal, clone C8/144B, 1:30), and anti-CD20 (clone L26, 1:50) from Dako (Carpinteria, California, USA); and CytoDEATH (clone M30, 1:200) from Boehringer Mannheim (Mannheim, Germany). Sections were then thoroughly washed in three changes of TBS for five minutes each (with inclusion of 0.5% v/v Triton X-100 in the first M30 wash). The CD3 sections were incubated with prediluted goat antirabbit secondary antibody (Zymed), while the CD8, CD20, and M30 CytoDEATH sections were incubated with prediluted goat antimouse secondary antibody (Zymed) for 30 minutes. Sections were washed in three changes of TBS and prediluted streptavidin-horseradish peroxidase (Zymed) was applied for 15 minutes. After three washes in TBS, antigenic sites were revealed using 0.05% 3,3′-diaminobenzidine (DAB) in TBS with H₂O₂ as substrate. Sections were lightly counterstained in Mayer's haematoxylin, dehydrated, cleared, and mounted using DePeX (BDH Gurr, Poole, UK). Appropriate negative and positive controls (normal human colon for M30 and lymph node for lymphocyte markers) were included in all staining runs.

For M30 CytoDEATH/CD3 double staining, M30 staining was performed to the DAB step, as described above. Sections were then reblocked with 4% skim milk powder and goat serum and incubated with anti-CD3 (1:200) for two hours at room temperature. Following standard goat antirabbit and streptavidin incubations, purple VECTOR VIP (Vector Laboratories Inc., Burlingame, California, USA) was applied for 3–4 minutes and the slides were washed well in water before counterstaining with methyl green (Vector Laboratories) for three minutes at 60°C.

TIL and apoptosis scoring was performed by two independent observers (JMM-R and AEB-H). For each case, the number of intraepithelial CD3⁺ and CD8⁺ lymphocytes in the 10 most infiltrated fields of 200 tumour nuclei were counted and the five highest values were averaged to obtain the final lymphocyte score per 1000 neoplastic cells. To define TIL positive tumours, the cut off of 5% (that is, ⩾50 CD3⁺ lymphocytes per 1000 tumour nuclei) was chosen for statistical and pathological reasons. There was a natural separation in the bimodal distribution of CD3⁺lymphocytes at a score of 50/1000 tumour nuclei and a pathologist (JRJ) could not easily detect a CD3 count of less than 5% by routine haematoxylin and eosin staining. In addition, the ratio of intraepithelial lymphocytes (IEL) to epithelial cells in the normal human colon is approximately 5:100.16 The distribution of apoptotic epithelial cells was graded in several areas of each tumour. Firstly, 10 randomly selected fields of 200 nuclei from within the middle of each tumour were counted and averaged to obtain a final score of random per cent apoptosis (random %). Secondly, apoptosis at the luminal surface and growing (advancing) edge of the tumour was scored as 0 (<1%), + (1–5%), ++ (5–10%), +++ (10–15%), or ++++ (>15%). Finally, epithelial cell apoptosis occurring within the glandular lumen was graded semiquantitatively as 0 (little or none), + (some), or ++ (widespread) according to the degree of M30 positive staining. M30 negative lumen are likely to contain apoptotic TILs, inflammatory cells, and necrotic debris. When the observers' scores differed by more than 1% for TIL counts or 2% for apoptosis scores, cases were discussed over a double headed microscope and the final grade was reached by consensus.

MUTATION ANALYSIS

P53 mutations were detected using single strand conformational polymorphism analysis as described previously.6 Detection of mutations in codons 12 and 13 of the K-ras proto-oncogene using non-radioactive polymerase chain reaction-restriction fragment length polymorphism were described previously.3 Genetic alterations in the coding 8-guanine repeat in the BAX gene were analysed as previously described.7

STATISTICAL ANALYSIS

A non-parametric approach to analysis was used as the primary variables of interest, that is, CD3, CD8, and random apoptosis counts were not normally distributed. This type of analysis is also more conservative than a traditional parametric approach. Therefore, mean and median age, CD3, CD8, and random apoptosis counts were compared across MSI levels using a Kruskal-Wallis non-parametric analysis of variance. Also, K-ras,p53, and BAX status were compared with CD3 or random apoptosis counts using a Kruskal-Wallis analysis of variance. Differences in the frequency distributions of categorical factors, such as sex, stage, and degree of differentiation, across MSI levels were examined using Pearson's χ² statistic. Spearman's non-parametric correlation coefficient (rho) was used to measure the degree of association between CD3 or CD8 count and random apoptosis per cent. In addition, the degree of association between CD3 count or random apoptosis and Bcl-2 or p53 protein expression was measured using Spearman's correlation coefficient. A p value less than 0.05 was considered statistically significant.

CLINICOPATHOLOGICAL FEATURES OF MSI-H, MSI-L, AND MSS COLORECTAL CANCERS

Clinicopathological data from the 102 sporadic colorectal carcinomas surgically resected from 101 patients are shown in table 1. The majority of MSI-H cancers (28/32, 87.5%) occurred in the right colon (proximal to the splenic flexure). In contrast, MSI-L and MSS cancers were found predominantly in the left colon (distal to the splenic flexure) and rectum (table 1). The MSI-H cancers were also characterised by a significantly higher proportion of mucinous carcinomas (composed of >50% mucinous lakes). In addition, more than half (18/32, 56.3%) of the MSI-H cancers were poorly differentiated compared with 4/30 (13.3%) MSI-L and 9/40 (22.5%) MSS tumours. There was a tendency towards more advanced disease in the MSI-L and MSS groups as 12/30 (40.0%) MSI-L and 16/40 (40.0%) MSS cancers had metastasised to the lymph nodes or distal sites (Duke's stages C and D) whereas only 7/31 (22.6%) of MSI-H cancers had progressed to stage C or stage D (table 1). However, this trend did not reach statistical significance (p=0.071).

TILs IN SPORADIC COLORECTAL CANCER ARE ASSOCIATED WITH BOTH THE MSI-H AND MSI-L PHENOTYPES

Immunohistochemical analysis revealed an abundance of TILs in MSI-H colorectal cancers with a mean count of 68.4 (SD 46.9) positively stained CD3 lymphocytes per 1000 tumour nuclei (table 2). The MSI-L subgroup had an intermediate mean CD3 count of 36.3 (26.5) and the mean CD3 count for the MSS group was the lowest at 24.6 (24.1). The distribution of CD3⁺ count is summarised according to MSI in fig 1. Altogether, 37 of 102 tumours were scored as TIL positive (containing >50 CD3⁺ lymphocytes per 1000 tumour nuclei). Of these, 23/32 (71.9%) were MSI-H, 9/30 (30%) were MSI-L, and only 5/40 (12.5%) were MSS (table 2). The difference in both median CD3⁺ count and percentage of assigned TIL score between these groups was highly significant (p<0.001) (table 2), confirming that these TILs were predominantly associated with the MSI and not the MSS phenotype. Further immunohistochemistry and quantitation with anti-CD8 found CD8⁺ counts similar to CD3⁺counts in all 102 cancers, with CD8⁺ cells comprising 75–80% of the CD3⁺ population (fig 2, table 2). Therefore, TILs appeared to be predominantly CD3⁺/CD8⁺ T cells. This result was confirmed by the absence of positive intraepithelial anti-CD20 (B cell) staining in serial sections of the same tumours (fig 2D and data not shown). CD4 staining was also attempted but was unsuccessful on this series of tumours (data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Distribution of CD3⁺ tumour infiltrating lymphocytes (TILs) in colorectal cancer. All 102 tumour CD3⁺ counts are represented by a triangle. The score of 50 CD3⁺ lymphocytes was chosen as the cut off for TIL positive cancers (broken line) and the horizontal bars represent the mean count for each group. MSS, microsatellite stable; MSI-L, microsatellite instability-low; MSI-H, microsatellite instability-high.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Example of a typical undifferentiated microsatellite instability-high (MSI-H) cancer heavily infiltrated with lymphocytes. (A) Standard haematoxylin and eosin stained section. (B) Serial section of the same tumour stained with anti-CD3, revealing T lymphocytes. (C) Serial section stained with anti-CD8. Note that most CD3⁺ T cells also show positive staining with anti-CD8. (D) Serial section of the same tumour stained with anti-CD20, a B cell marker. Note that no intraepithelial cells have stained positively in this section (original magnification ×60).

MICROSATELLITE INSTABILITY IS ASSOCIATED WITH ENHANCED APOPTOSIS OF MALIGNANT EPITHELIAL CELLS

To determine if the presence of TILs correlates with an increased level of tumour cell apoptosis, the frequency and distribution pattern of apoptotic cells in the same selected series of 102 cancers were detected by performing immunohistochemistry with the M30 CytoDEATH antibody. The M30 antibody binds to a neoepitope on cytokeratin-18 exposed only after caspase cleavage thus allowing highly specific and early detection of apoptotic epithelial cells.17 M30 CytoDEATH staining is markedly more sensitive than the morphological identification of apoptosis by haematoxylin and eosin staining (fig 3). Pockets of M30⁺ malignant epithelial cells were found randomly distributed throughout the 102 examined lesions. Significantly, the MSI-H cancers showed on average a twofold higher rate of apoptosis (3.55 (1.99)%) than the MSS cancers (1.53 (1.47)%) while the MSI-L subgroup had an intermediate level of apoptosis (2.52 (1.92)%) in 10 randomly selected fields from each tumour (table 3). This difference in the level of tumour cell apoptosis was highly significant (p<0.0001). On average, apoptotic counts at the surface and growing edge were also significantly higher in the MSI-H subgroup, intermediate in the MSI-L, and lowest in the MSS cancers (p<0.0001 and p=0.002, respectively) (table 3). In addition, epithelial cell apoptosis in the glandular lumen was most prevalent in MSI-H cancers (table 3) although this difference did not reach statistical significance (p=0.059). M30 negative lumen may contain apoptotic TILs, necrotic debris, neutrophils. and phagocytic cells. Importantly, when serial sections were compared there was a paucity of M30⁺staining in areas heavily infiltrated with high CD8⁺ T cells (fig 4) and similarly, M30 CytoDEATH/CD3 double staining showed little colocalisation between CD3⁺ TILs and M30⁺ apoptotic tumour cells (fig 5). Overall, there was a significant but weak-moderate linear correlation between the number of CD3⁺ TILs and M30⁺ random apoptosis scores (r=0.347, p<0.001). The linear correlation between CD8⁺ lymphocytes and M30⁺ apoptosis was of a similar magnitude (r=0.299, p<0.002). However, these linear correlations were lost when the tumours were subdivided into their three subgroups according to the level of MSI.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Serial sections of a microsatellite instability-low (MSI-L) cancer (A) stained routinely with haematoxylin and eosin and (B) staining with the M30 cytoDEATH antibody specific for epithelial cell apoptosis. The brown cytoplasmic staining indicates the many apoptotic tumour cells present in this field (original magnification ×120).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Serial sections of a well differentiated colorectal cancer stained with (A) anti-CD3 and (B) M30 CytoDEATH antibody. Note that only one apoptotic cell is present in this area densely infiltrated with tumour infiltrating lymphocytes (original magnification ×120).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Example of M30 CytoDEATH and CD3 double staining, counterstained with methyl green. The brown DAB staining indicates M30⁺ apoptotic tumour cells (arrows) and purple Vector VIP staining reveals the CD3⁺ T cells present in this tumour infiltrating lymphocyte (TIL) positive microsatellite instability-high (MSI-H) cancer. Note that M30⁺ neoplastic cells and CD3⁺ TILs do not coincide (original magnification ×120).

We also explored the relationship between apoptotic level and mutations in the p53, K-ras, and BAX genes, as well as p53 and Bcl-2 protein expression. In the 79 tumours previously analysed for K-ras status in this laboratory,3 there was a trend towards decreased apoptosis (p=0.135), with the 19 tumours harbouring mutant K-ras demonstrating a mean apoptosis score of 1.56% compared with 2.51% in the 60 tumours with wild-type K-ras. In the 81 tumours from this series previously analysed for p53mutation,6 no significant correlation was found betweenp53 status and apoptosis, with 42 wild-type and 39 mutant p53 cancers having similar mean apoptosis scores (2.39% v 2.20%, respectively). However, over expression of mutant p53 protein, as determined previously by immunohistochemistry,18 was correlated with a low level of apoptosis in 85 tumours from this series (p=0.041). Bcl-2 protein expression was also assessed previously in 65 tumours from the current series.18 No correlation was found when random apoptosis counts were compared across Bcl-2 expression (p=0.618). Finally, BAX status was assessed in the 32 MSI-H cancers of this series, as described previously.6 The apoptotic index in the 16 tumours with insertion or deletion in the polyG tract in the BAX gene (3.78%) was similar to that of the 16 tumours without BAX mutation (3.32%) (p=0.361).