T cell subpopulations in lymph nodes may not be predictive of patient outcome in colorectal cancer

Colorectal cancer is estimated to cause 639,000 deaths world wide per year [1]. The prognosis following surgery depends on disease stage, and this also determines the need for additional treatment. However clinico-pathological stage characteristics alone provide imperfect prognostic information. For example, approximately 25% of patients with disease localised to the primary site (UICC Stage I and II) relapse after surgery and may have benefited from adjuvant therapy [2], whereas 25% of patients with regional lymph node metastases (UICC Stage III) are cured by surgery alone [3]. Various ways to improve the prognostic accuracy of staging include increasing the number of lymph nodes analysed [4, 5], increasing the sensitivity of the tests used to detect lymph node metastases [6] and using microarray technology to analyse gene expression [7, 8]. However these methods do not take onto account potentially important host-related factors such as the immune response.

The immune response has long been associated with eradication of tumours [9]. More recently, it has become clear that T cells in the tumour are positively associated with good patient prognosis [10, 11] in colorectal cancer. CD4 or CD8+ T cells expressing IFNγ, or the IFNγ inducing transcription factor Tbet, are the cells most likely involved at the tumour site [12, 13].

In immune responses to infection, the effector CD4 and CD8 T cell populations are held in check by a third population of cells - regulatory T cells (Tregs). While there are numerous subtypes of T cells with regulatory function, the majority of suppressive function is mediated by Foxp3+ CD4+ Tregs. As expected, low numbers of these Foxp3+ Tregs have been associated with improved patient outcome in breast and colorectal cancers [14‐16]. However, some authors report an association between high numbers of Tregs and positive patient outcome [17, 18], although Salama et al found a negative association between patient outcome and high frequency of Tregs in the non-tumour associated tissue [18]. More recently, Chaput et al identified a population of CD8+Foxp3+ T cells in a cohort of colorectal cancer patients that had suppressive activity and were proposed to mediate tumour escape [19].

The immune response is initiated in the lymph nodes, and although analyses of T cell subsets in the lymph nodes of breast cancer patients have been performed [20], the effect of these T cell subsets on colorectal cancer patient outcome had not been explored. We hypothesised that the priming environment of an anti-tumour immune response would be a useful predictor of patient outcome. In this study we examined the lymph nodes of Stage II colorectal cancer patients to identify CD4+, CD8+ and Foxp3+ cell populations and correlated these with patient outcome, alone, and in combination with other clinico-pathological variables.

Thirty three patients with stage II colon cancer were included; 13 with and 18 without recurrence after 5 years of follow up. Of the 13 patients with recurrent disease, four recurred locally and nine had systemic disease (seven liver, one lung, and one lung and brain). Patient characteristics are summarised in Table 1. For each patient, between 1 and 33 lymph nodes were available for analysis (median = 10). Within each lymph node, between one and 15 sections were examined for CD4, CD8 and FoxP3 percentage (median = 10). For those nodes for which multiple sections were available, the "within-node" standard deviation was calculated to assess the consistency of immunological signal being obtained. Similarly, for those patients from whom multiple lymph nodes were sampled, the "within-patient" (i.e., "between-node" for the same patient) standard deviation was calculated. Finally the average immunological "signal " was calculated for each patient (for each of FoxP3, CD8 and CD4) and used to assess inter-patient variability by determining the "between patient" standard deviation.

Figure 1 shows immunohistochemical staining for CD4, CD8 and Foxp3 respectively. For all three measures of immunological activity (CD4, CD8 and FoxP3), the within-node variability was around half the level of the within-patient (between-node) variability (CD4: 5.81% vs 10.40%, CD8: 2.25% vs 4.24%, FoxP3: 0.24% vs 0.63%), indicating that replicate measurements obtained from the same node were relatively consistent in all cases. The same was not true, however, of nodes taken from the same patient, with the between-node standard deviation approximately the same as the between-patient standard deviation for all three measures of immunological activity (CD4: 10.40% vs 9.12%, CD8: 4.24% vs 4.15%, FoxP3: 0.63% vs 0.68%). That is, the variation in CD4, CD8 and FoxP3 percentages between nodes from the same patient was as great as the variation observed from one patient to another.

Given the large amount of within-patient variability that was observed across multiple lymph nodes from the same patient, the task of identifying differences in immunological activity between different groups of patients could be expected to be very challenging, as is reflected in the results presented below.

In this paper, we have described the analysis of T cell populations in the lymph nodes of Stage II colorectal cancer patients. We were unable to find any association between CD4, CD8 or Foxp3+ (presumed Tregs) and cancer recurrence or with other clinico-pathological variables.

T cells have long been known to play a role in eradicating tumours. Colorectal cancer has been particularly well studied, with several laboratories showing a positive association between patient survival and effector (IFNγ+) T cell infiltration into the tumour [10, 11]. It was expected that the regulatory T cell infiltration into the tumour would be negatively associated with patient outcome; however, regulatory (FoxP3+) T cells have been shown to have a protective role in colorectal cancer, in contrast to their negative role in many other cancers [17]. The positive effect of FoxP3+ T cells has been proposed to be a result of their effects on other T cells that are promoting tumour growth [25].

T cell immune responses are initiated in the lymph nodes by cells, such as dendritic cells, presenting tumour antigens to responding specific T cells. These activated T cells then migrate to the tumour and specifically destroy it. Munn et al proposed that the tumour draining lymph node is a unique immunological environment where the presence of regulatory T cells could mediate a suppressive effect on anti-tumour immune responses [26]. Indeed, depletion of Tregs enhances effector T cell responses in tumour draining lymph nodes [27]. Recent data also indicated that the presence of Foxp3+ T cells in tumour draining lymph nodes of colorectal cancer patients correlated with disease progression [28]. Given the associations between Treg infiltration in primary colorectal tumours and patient outcome [18], we questioned whether Tregs in the regional lymph nodes could be predictive of patient survival.

Our data is in contrast to Khort et al [20], who described a population of CD4 cells in the axillary lymph node could predict outcome in breast cancer patients. Although our sample was smaller, there were no apparent trends in the data to indicate that a larger sample would be likely to yield significant results. In fact, given the amount of variation in immunological activity that we observed in lymph nodes taken from the same patient, the use of lymph nodes for prognostic purposes would seem to be extremely challenging. Even if a difference in activation existed between patients with "good" and "poor" prognosis, detection of a statistically significant difference would require collection of large numbers of both patients and nodes. For per-patient prognosis, the inter-node variability would make accurate prediction almost impossible, with the good and poor responders likely to be indistinguishable from one another. This is likely due to the background of non tumour-specific T cell overshadowing the presence of tumour specific responses - indeed, the majority of studies looking at T cells as predictors of outcome in this disease have been restricted to the tumour tissue [11, 12, 17, 18, 21, 29].

We did not identify the sentinel nodes, which are believed to be the primary priming site for the anti-tumour immune response, however data exists to indicate that there is often more than one sentinel node and it's spatial relationship to the tumour can vary considerably [30].

Immunotherapy of cancer patients is difficult due to the specific nature of the adaptive immune response and the absence of easily identifiable tumour specific antigens. The current study looked only at total T cell populations in the lymph node, and it may be that tumour specific T cell populations were present in different frequencies in patients with and without recurrence, but not able to be identified as such.

A further complication is the lack of healthy control tissue. Studies comparing immune response in colorectal cancer patients have used blood of healthy patients [14, 15]; however the scope of our study was to investigate the role of lymph nodes for predicting patient outcome, and mesenteric lymph nodes from healthy controls were not obtainable. We compromised by using matching lymph node tissue from IBD patients, as has been previously published [15] but are aware of the difficulties of using immune tissue from patients with an immune mediated inflammatory disease.

However, an interesting finding was the difference between colorectal cancer patients and inflammatory bowel disease patients with respect to CD4 expression. IBD patients had a higher CD4 frequency that is not surprising given the inflammatory nature of IBD and the proven role for CD4 cells in driving this disease [23]. However, no difference was seen between cancer patients and IBD patients in Foxp3+ cells. This indicates that the Treg population was not diminished in IBD patients, a finding in direct contrast to Clarke et al. We are currently investigating this further to examine the role of other T cell subpopulations.

Foxp3 is recognised as the most specific Treg marker; however, there are reports of Foxp3 expression in effector T cells, especially in humans [31]. It is possible that the Foxp3 cells detected in our study were effector rather than regulatory cells. Studies are underway to further characterise these cells, using a panel of regulatory markers. Clarke et al found that Foxp3+ cells recovered from mesenteric lymph nodes of CRC patients exhibited regulatory activity against CD4 T cells [15], so it seems likely that Foxp3+ cells in our study have regulatory function.

We found no correlation between major T cell populations in regional lymph nodes and cancer recurrence in patients with stage II colon cancer. A more detailed analysis of T cell sub-populations will be required to determine whether characterisation of the immune response in regional lymph nodes can inform prognosis in colorectal cancer.