Tumour infiltrating lymphocytes correlate with improved survival in patients with oesophageal adenocarcinoma

Virchow first described the association of lymphocyte infiltration with solid tumours in 1863 [1]. Immune responses against malignant cells, whether systemic or within the tumour microenvironment, are increasingly implicated as a determining factor in tumour progression [2]. The most comprehensive studies, to establish this hallmark of cancer [3], have been in colorectal tumours [4]. Specific tumour infiltrating lymphocyte (TIL) subsets communicate and function to affect tumour growth, and the balance of these effects leads to either tumour regression or tumour promotion. An anti-tumour effect is mediated by the combination of cancer cell lysis and the production of cytotoxic cytokines, supported by cluster of differentiation (CD)4+ T helper cells. In contrast, T regulatory cells, Forkhead box P3+ (FOXp3+) T cells, act to suppress the immune response of other cells and so affect the inflammatory process indirectly [2]. Therefore, understanding the impact of specific subsets of immune cells that infiltrate tumours is important for making rational decisions in the development of targeted therapies.

Oesophageal adenocarcinoma (OAC) is becoming increasingly common in the Western world, and despite a variety of strategies to improve outcome [5], survival remains poor at 10–15 % at 5 years [6]. Radical treatment with curative intent includes neoadjuvant therapy and oesophagogastrectomy. Neoadjuvant chemotherapy (NAC) delivers improved survival in a small percentage of patients, in whose tumours there is a significant pathological response to treatment [7]. However, approximately two-third of patients present with advanced, incurable disease at diagnosis. Biological therapies such as growth factor blockers, poly-ADP ribose polymerase (PARP) inhibitors, vaccines and monoclonal antibodies are being evaluated and are yet to enter routine clinical practice [8]. The rapid rise in incidence of OAC in the west means that novel therapies are urgently required, either as single agents or for use in combination with conventional treatments.

In particular, the ability of CD8+ lymphocytes to recognise tumour antigens has been well documented [9], and tumour regression is observed when tumour-reactive T lymphocytes invade cancers [10]. Before considering immune therapies as a potential treatment in OAC, it is important to understand whether immune cells are present in OAC, in which subsets of cells are observed and whether or not they have prognostic significance.

The effects of TILs have been studied in a range of solid tumours, and high TIL density correlates with better survival (reviewed in [11]). A few initial studies have examined the association of TILs in OAC with prognosis [12, 13]. These studies have provided conflicting results, and no study has assessed TIL density in resected tumours after NAC and whether or not there is a link to outcome. Schumacher et al. observed intratumoural CD8+ T cell infiltration to correlate with improved survival in a small heterogeneous cohort of OAC (n = 37) and oesophageal squamous cell carcinoma (OSCC) (n = 33) patients [14]. Additional studies have shown that the overall grade of TIL density at the invasive margin correlates with improved survival in patients with gastroesophageal cancer [12]. In contrast, Zingg et al. did not find any independent associations between differing TIL subsets and survival in OAC patients who received multimodal therapy with either neoadjuvant chemoradiotherapy or surgery alone. However, they did demonstrate in univariate analyses that particular subtypes of TIL conferred a better survival when dichotomised at median counts (CD3+, CD8+, FOXp3+, CD8+:CD4+) [15]. Therefore, the prognostic value of TILs in OAC remains to be established and the composition of TIL density following NAC in OAC is yet to be assessed.

In this study we have analysed the level of TILs in a large homogeneous cohort of oesophagogastric cancer patients following radical treatment with curative intent. We obtained three major findings: firstly, we have shown associations between pathological stage of disease and TIL density and confirmed the independent association of particular TIL subsets and survival; secondly, we have shown significant correlation between TIL subtypes in OAC, and finally, we found patients with a significantly increased pathological response to NAC had higher levels of TILs in their resected tumour, most notably with CD4+ and CD8+ TILs for local tumour regression and lymph node response, respectively.

Of the TIL subsets analysed, CD8+ T cells had the most significant independent association for both CSS (p = 0.003) and DFS (p < 0.0001). This has been previously suggested to be important in OAC; however, independent association has not been universally verified [12, 14, 15]. Zingg el al. [15] identified on univariate analysis a number of TILs as significant predictive factors; however, these were not independent factors when taking into account the stage of disease. In contrast, Schumacher et al. [14] identified CD8+ TILs to be a significant predictive factor of survival independent of disease stage. The reasons for this discrepancy of findings between studies may relate to the clinicopathological factors used to build the multivariate statistical model, the location of TILs counted and the cut-off values used to dichotomise the TIL frequencies into high and low groups. In our large and homogeneous cohort, we choose to evaluate TIL number as a continuous variable, counted intratumoural TILs and entered a comprehensive number of known clinicopathological predictors of survival into the univariate and subsequent multivariate statistical models. We believe this to be the most robust analysis of the association of TILs with survival on OAC that has been conducted to date.

In addition to the observed effects TILs have on disease recurrence, this study demonstrates that TILs play a role at different stages of disease. The significant association observed between TILs with T and N stage supports the possibility that intratumoural T cells prevent tumour progression throughout the disease process. These findings are supported by previous smaller studies showing correlation with low TILs and higher stage of disease [14, 15].

A novel observation is that the response to chemotherapy links with TIL counts. Previous studies have highlighted increased infiltrate of TILs in OAC tumours after NAC when compared to surgery alone [13]. We did not find any statistically significant difference between TIL levels in those patients who received NAC and those that proceeded directly to surgery as a whole. It would be preferable to evaluate TILs in preoperative biopsies to assess response prior to neoadjuvant therapy. This was attempted but was unsuccessful due to the paucity of tumour material in the small diagnostic endoscopic biopsies unlike the diagnostic biopsies taken for other tumours that tend to be larger, for example breast and colon cancer. However, different and heterogeneous chemotherapy regimens were used in previous studies making comparison difficult. We did find that patients who had a significant response to NAC (as assessed by TRG and LN response) had a higher level of immune infiltrate. This important finding has been shown in other cancer sites where loss of CD4+ Treg [23] and TIL numbers at the margins of liver metastases [24] predicted for response to chemotherapy [25‐27]. The potential immunological mechanisms by which cytotoxic chemotherapy can provide anti-tumour activity are being increasingly highlighted. These include: subverting immunosuppressive mechanisms; exerting stimulatory effects in immune cells; and modulating dying tumour cells, so they regain visibility to the host immune response [28]. The combination of immunological therapies with conventional chemotherapy has been suggested to provide a synergistic effect if the host immunological response is harnessed appropriately [29]. In addition to the effects of chemotherapy on TILs, previous studies have highlighted spontaneous pathological regression of tumours in 13.7 % (n = 17/124) of OAC cases that have received no NAC when using TRG to assess the resected tumour specimen. This may represent the host immunological response and successful immune attack [30]. Our study would support this hypothesis as significant response to NAC was associated with increased TILs most notably with CD8+ TILs. In other tumour sites, CD8+ infiltration has been shown to correlate with specific immunogenic antigen expression and improved survival [31]. These findings highlight the T cell effector potential and a means to harness the patient’s immune response in cancer.

It is also of interest that increased frequencies of TILs correlate with improved response following the use of trastuzumab [32] in light of the ToGA study [33] that may lead to the adjuvant use of trastuzumab in OAC. This adjuvant therapy may lead to a further increase in the adaptive immune response in the tumour as seen in breast cancer [34]. We have previously shown that patients are more likely to respond to chemotherapy if acute phase proteins are normal (serum albumin) by assessing systemic markers of nutrition and the inflammatory response (neutrophil/lymphocyte ratio, serum albumin) [18]. It is suggested that suboptimal immunological and nutritional status may contribute to tumour development through subversion of tumour immunity [35, 36] and this is particularly pertinent to OAC. We found that patients with lower TIL levels were less likely to respond to chemotherapy and had lower serum albumin levels. From our data, it is not possible to know whether high TIL levels lead to a greater responsiveness to chemotherapy or chemotherapy leads to a higher number of TILs in those that respond, and it may well be a combination of the two. The association of poor nutritional status with lower TIL levels and a lack of response are logical and may well be the most likely explanation.

Limitations of this study include its retrospective nature and the associated biases on selection and collection. Patients were excluded if they had a complete pathological response to NAC due to the lack of tumour in the resection specimen and also if there was insufficient material collected. However, this cohort is representative of current clinical practice. We found increasing nodal burden (p < 0.0001) to be the best independent prognosticator for worse survival. The cohort was homogenous in terms of staging, histology and treatment algorithms. In addition, patient, tumour factors and survival data are in keeping with published western cohorts making our findings applicable to these populations. The excluded patients with complete pathological response (TRG1) potentially will have had the highest immune response with high TIL levels. This hypothesis is supported by analysis of the cytotoxic response of patients with a complete pathological response in breast cancer patients. Granzyme B and TiA1, cytolytic granules, expressing cells were observed at higher frequency in specimens that had undergone a pathological complete response [25].

Additional criticism could be expressed with regard to the use of TMA cores not being representative of the tumour as a whole. However, multiple studies have utilised this approach and excellent correlation between the two has been established [37]. With respect to scoring TILs, the technique used has been described frequently by other studies with good interobserver and intraobserver correlation. Furthermore, these data were analysed with and without dichotomization so as to limit false-positive results [38, 39].

An unexpected finding in our study was the association of higher TILs for some subsets (CD4+, CD8+ and FOXp3+) in female patients. This may represent a type 1 error; however, it is well documented that immune cell numbers vary with age and gender. Possible explanations include the inhibitory effect of oestrogen on T-suppressor cells or its stimulatory effect on T helper cells [40]. However, this speculation is outside the remit of this study and will require further analysis. The reason why higher infiltration of regulatory FOXp3+ cells showed favourable prognosis in this study is complex but has been reported in gastric and other tumour types [41, 42]. This warrants further functional analysis, but others have suggested a synergistic role of regulatory cells with other TIL subsets within the tumour [43].

With regard to our novel finding, it remains to be seen whether the composition of TILs can predict for response to chemotherapy prior to treatment and the functional role the immune response plays in improving the response to chemotherapy. Emerging evidence from other solid organ tumours suggests that this might be possible [44]. However, as yet the required sensitivity and specificity have not been met for this to enter clinical practice. A potential clinical application of the TIL response could be in the selection of patients for specific adjuvant therapies, specifically using immunomodulators. In our series, a significant response to chemotherapy (TRG2 or LN downstaging) was associated with high TIL levels, suggesting these patients may benefit from further adjuvant chemotherapy. It would be intriguing in this group in particular to evaluate the long-term outcome after the use of immunomodulators, that release pre-existing anti-tumour T cells, as has been proposed for aPD1 and aCTLA4 therapies. In contrast, the group of patients who had no or a partial response to chemotherapy (TRG 3–5 or no LN downstaging) with high TIL levels in the pathological specimen additionally may benefit from further adjuvant or alternative chemotherapy in combination with an immunomodulator. In contrast, those patients with low TILs may warrant the use of second-line therapies but it seems less likely that immune attack will be successful, unless second-line therapy can overcome the barriers to immunological visibility of the cancer.

In summary, the results of this study show that local immune responses, in particular the adaptive immune response, are important independent predictors of CSS and DFS in patients with OAC undergoing radical treatment. These findings highlight the role of the adaptive immune response in preventing tumour recurrence in OAC. A generalised immune response was observed with high correlation between TIL subtypes and at all stages of disease. This is pertinent with regard to the design of therapies as it suggests all patients would benefit from treatment that appropriately boosts the immune response. A significant response was associated with higher infiltration with TILs post-therapy. Further work analysing the function of these TIL subsets and the correlation of pretreatment TIL densities with response to therapy may unearth predictors of response to chemotherapy in addition to aid design of novel treatments or as a mechanism to improve response to current therapies.