Lymphocytic Reaction to Colorectal Cancer is Associated with Longer Survival, Independent of Lymph Node Count, MSI and CpG Island Methylator Phenotype

Study Population

We utilized the databases of two independent prospective cohort studies; the Nurses' Health Study (N=121,701 women followed since 1976) and the Health Professionals Follow-up Study (N=51,529 men followed since 1986) (27). Every 2 years, participants were sent follow-up questionnaires to identify newly diagnosed cancer in themselves and their first-degree relatives. When a participant reported colorectal cancer, study physicians reviewed medical records, and recorded TNM stage, tumor location, and the number of positive and negative lymph nodes. We collected paraffin-embedded tissue blocks from hospitals where patients underwent tumor resections (27). We excluded cases preoperatively treated with radiation and/or chemotherapy. Based on availability of tissue specimens for pathologic analyses, we included a total of 843 stage I-IV colorectal cancer cases diagnosed up to 2003 (Table 1). Patients were observed until death or June 30, 2006, whichever came first. Ascertainment of deaths included reporting by the family or postal authorities. In addition, the names of persistent nonresponders were searched in the National Death Index. We identified >98% of deaths in the cohorts by these methods. The cause of death was assigned by physicians unaware of tumoral pathologic or molecular data. Written informed consent was obtained from all study subjects. This study was approved by the Human Subjects Committees at Brigham and Women's Hospital and the Harvard School of Public Health.

Histopathologic Evaluations

Tissue sections from all colorectal cancer cases were examined by a pathologist (S.O.) unaware of other data. Of the 834 tumors, ≥3 tumor tissue blocks were available in 284 cases, 2 in 297 cases, and 1 in the remaining 262 cases. Tumor grade was categorized as high or low (≤50% vs. >50% glandular area). Four components of lymphocytic reactions [Crohn's-like lymphoid reaction, peritumoral lymphocytic reaction, intratumoral periglandular reaction, and tumor infiltrating lymphocytes (TIL)] were examined (Figure 1). Crohn's-like reaction was defined as transmural lymphoid reaction. Peritumoral lymphocytic reaction was defined as discrete lymphoid reactions surrounding tumor. Intratumoral periglandular reaction was defined as lymphocytic reaction in tumor stroma within tumor mass. TIL was defined as lymphocytes on top of cancer cells. For any given tumor, each of the 4 lymphocytic reaction components was scored as 0 (absent), 1+ (mild), 2+ (moderate) or 3+ (marked). The overall lymphocytic reaction score (0 to 12) was the sum of scores for the above 4 reaction components. Due to a skewed distribution of the lymphocytic scores (Supplemental Table 1), the majority of cases were in the group of scores 0-2. Most of them had score of 2. Thus, the reference group (i.e., low score) was set as those with scores 0-2, which provided the robust reference group. We divided non-low score cases into two groups because of a wide range of scores from 3 to 12. We set the cutoff between middle and high groups as ≤6 vs. ≥7, because it was in the middle of the total score of 12. A random selection of 398 cases were re-examined by a second pathologist (J.N.G.) unaware of other data, and the correlation on the lymphocytic scores between the two pathologists was good (Spearman correlation r = 0.65, p<0.0001; κ = 0.55 for score <3 vs. ≥3, p<0.0001). Kappa coefficients between the lymphocytic scores by the two pathologists for additional cutoffs were as follows: 0.49 for <2 vs. ≥2; 0.53 for <4 vs. ≥4; 0.46 for <5 vs. ≥5; 0.50 for <6 vs. ≥6; 0.49 for <7 vs. ≥7; 0.55 for <8 vs. ≥8; 0.57 for <9 vs. ≥9; 0.47 for <10 vs. ≥10.

Pyrosequencing of KRAS and BRAF, and Microsatellite Instability (MSI) Analysis

DNA from paraffin-embedded tissue was extracted, and PCR and Pyrosequencing targeted for KRAS codons 12 and 13 (28), and BRAF codon 600 (29) were performed. MSI status was determined using microsatellite markers, D2S123, D5S346, D17S250, BAT25, BAT26, BAT40, D18S55, D18S56, D18S67 and D18S487 (i.e., 10-marker panel) (30). MSI-high was defined as the presence of instability in ≥30% of the markers, MSI-low/microsatellite stability (MSS) as no or <30% unstable markers.

Real-Time PCR for CpG Island Methylation, and Pyrosequencing to Measure LINE-1 Methylation

Sodium bisulfite treatment on tumor DNA and subsequent real-time PCR (MethyLight) assays were validated and performed as previously described (31). We quantified promoter methylation in 8 CIMP-specific genes (CACNA1G, CDKN2A, CRABP1, IGF2, MLH1, NEUROG1, RUNX3 and SOCS1) (32-34). CIMP-high was defined as ≥6/8 methylated promoters using the 8-marker CIMP panel, CIMP-low/0 as 0 to 5 methylated promoters, according to the previously established criteria (33). In order to accurately quantify relatively high LINE-1 methylation levels, we utilized Pyrosequencing as previously described (35, 36).

Immunohistochemistry for p53 and COX-2

Tissue microarrays (TMAs) were constructed (37). p53 (38) and COX-2 (cyclooxygenase-2) immunohistochemistry was performed as previously described (27, 30). Appropriate positive and negative controls were included in each run of immunohistochemistry. All immunohistochemically-stained slides were interpreted by a pathologist (S.O.) unaware of other data. A random sample of 118 and 108 tumors were re-examined for p53 and COX-2, respectively, by a second observer (p53 by K.N., and COX-2 by R. Dehari, Kanagawa Cancer Center) unaware of other data, and the concordance between the two observers was substantial (κ=0.75, p<0.0001, for p53; and κ=0.62, p<0.0001, for COX-2).

Statistical Analysis

All analyses used SAS version 9.1 (SAS Institute, Cary, NC) and all p values were two-sided. The chi square test was used to assess an association between categorical variables. The analysis of variance (ANOVA) was performed to compare mean age and mean LINE-1 methylation level across the 3 lymphocytic categories. The κ coefficients were calculated to assess agreements between the two observers for lymphocytic reaction scores and immunohistochemical data.

In analysis of colorectal cancer-specific survival, death as a result of colorectal cancer was the primary end point and deaths as a result of other causes were censored. The Kaplan-Meier method was used to describe the distribution of survival time, and the log-rank test was performed. We used stage-matched (stratified) Cox proportional hazard models to calculate hazard ratio (HR) of death according to lymphocytic reaction, adjusted for age at diagnosis (continuous), sex, year of diagnosis (continuous), BMI (≥30 vs. <30 kg/m²), family history of colorectal cancer in any first degree relative (present vs. absent), tumor location (proximal vs. distal colon vs. rectum), tumor grade (high vs. low), negative lymph node count (ordinal categories; 0-3, 4-6, 7-12 and ≥13), MSI (high vs. low/MSS), CIMP (high vs. low/0), LINE-1 methylation (continuous), KRAS, BRAF, p53 and COX-2. Tumor stage (I, IIA, IIB, IIIA, IIIB, IIIC, IV, unknown) was used as a matching (stratifying) variable (using the “strata” option in the SAS “proc phreg” command) to minimize residual confounding; this fine tumor staging also included information on the number of positive lymph nodes. When we included all of the 4 lymphocytic reaction components (Crohn's-like reaction, peritumoral reaction, periglandular reaction and TIL) in multivariate Cox models, these 4 variables were collinear with each other. Spearman correlation coefficients among these 4 variables are shown in Supplemental Table 2. Thus, we analyzed each of the 4 components separately in Cox models. In the summary analysis, we used overall lymphocytic reaction score (0 to 12; the sum of scores for the 4 components). We classified tumors into 3 categories [overall reaction score 0-2 (low) vs. 3-6 (middle) vs. 7-12 (high)]. We examined the possibility of a non-linear relation between the lymphocytic reaction score and survival, non-parametrically with restricted cubic splines. In addition, we added squared term of lymphocytic score and/or cubic term of lymphocytic score in a multivariate model, and did not observe any significant improvement in model fitting (all p>0.5 by likelihood ratio tests). Nonetheless, our data presentations were mainly based on categorized groups according to the lymphocytic score, because categorized data were easily understandable. In the Cox regression analyses, the proportionality of hazards assumption was satisfied by evaluating time-dependent variables, which were the cross-product of the ordinal lymphocytic score variable and survival time (p=0.084 for colorectal cancer-specific survival; p=0.18 for overall survival). For cases with missing information on the negative lymph node count, we assign a separate (“missing”) indicator variable. For cases with missing information in other categorical covariates [including BMI (5.8% missing), tumor grade (0.7%), MSI (2.0%), CIMP (3.7%), KRAS (1.3%), BRAF (3.8%), p53 (1.7%) and COX-2 (0.7%)], we included those cases in a majority category, to minimize the number of variables in multivariate Cox models. We confirmed that excluding cases with missing information in any of the covariates did not substantially alter results (data not shown). An interaction was assessed by including the cross product of the ordinal lymphocytic score variable and another variable of interest in a multivariate Cox model, and p values for interaction were conservatively interpreted, considering multiple hypothesis testing.

Lymphocytic reaction to colorectal cancer

We examined lymphocytic reaction under a light microscopy in 843 stage I-IV colorectal cancers identified in the two prospective cohort studies. We scored Crohn's-like reaction, peritumoral reaction, intratumoral periglandular reaction and tumor infiltrating lymphocytes (TIL) as: 0 (absent), 1+ (mild), 2+ (moderate) or 3+ (strong). Then, we calculated overall lymphocytic reaction score (0 to 12) as the sum of scores for the 4 components. As shown in Table 1, tumors with middle (score 3-6) to high (score 7-12) lymphocytic scores were associated with older age, proximal location, the larger negative node count, stage I-III, high tumor grade, MSI-high, CIMP-high, LINE-1 hypermethylation, BRAF mutation, and negative p53 (all p<0.003).

Overall lymphocytic reaction score and colorectal cancer survival

We examined patient survival according to the overall lymphocytic reaction score. There were a total of 373 deaths including 225 colorectal cancer specific deaths. Five-year colorectal cancer-specific survival was 73% among patients with a score of 0-2, 83% among patients with a score of 3-6, and 89% among patients with a score of 7-12 (log-rank p<0.0001), and five-year overall survival was 67 % among patients with a score of 0-2, 77% for a score of 3-6, and 88% for a score of 7-12 (log-rank p=0.002) (Figure 2A).

We examined the possibility of a non-linear relation between the lymphocytic reaction score and survival, non-parametrically with restricted cubic splines (Figure 2B-C). This flexible method allowed us to examine the relation with survival without any categorization of the lymphocytic reaction score. Increasing lymphocytic reaction score was associated with a progressive decrease in mortality.

In multivariate Cox regression analyses, overall lymphocytic reaction score was associated with a significant improvement in colorectal cancer-specific and overall survival (P_trend =0.008 and P_trend=0.002, respectively) (Table 2). The attenuation of effect of lymphocytic reaction score on colorectal cancer-specific survival in the multivariate analysis was principally due to adjustment for tumor stage. A high lymphocytic reaction score was inversely associated with stage IV (Table 1). No other major confounders were identified.

Each lymphocytic reaction component and colorectal cancer survival

We examined patient survival according to each of the 4 lymphocytic reaction components (Figure 3). All components of the lymphocytic reactions were associated with improved colorectal cancer-specific survival (long-rank p<0.014). We examined the prognostic significance of each component in stage-adjusted and multivariate Cox models (Table 3). Crohn's-like reaction, peritumoral reaction, and periglandular reaction appeared to be associated with long cancer-specific survival in multivariate analysis (P_trend<0.053) (Table 3).

Stratified analysis of lymphocytic reaction score and survival

We further examined whether the prognostic influence of the lymphocytic reaction score was modified by any of the other patient and tumoral features (Supplemental Figure). The prognostic effect of the lymphocytic reaction score was not significantly modified by any of the variables examined (all P_interaction >0.30). Notably, the effect of the lymphocytic reaction score did not significantly differ across tumor stages (P_interaction =0.32) or between the two independent cohort studies (P_interaction =0.73).