Published online before print June 17, 2008
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
,
,
,,,,1
* Centre de Recherche des Cordeliers, Université Paris Descartes, Paris, France;
Université Pierre et Marie Curie-Paris6, Paris, France;
INSERM, Paris, France; and
Immunology Unit, Hopital Européen Georges Pompidou, Paris, France
1 Correspondence: Centre de Recherche des Cordeliers, UMRS 872, rue de l’Ecole de Médecine, Paris, F-75006 France. E-mail: herve.fridman{at}crc.jussieu.fr
ABSTRACT
Colorectal cancer is the second most common cause of cancer-related death. Novel prognostic factors should be identified and validated to refine the present tumor-node-metastasis system. The presence of immune cells infiltrating colorectal cancers is a common phenomenon. However, the current belief is that clinically detectable human tumors escaping immune surveillance are no longer kept in check by the immune cells of the tumor microenvironment. Despite studies showing the influence of immune cell infiltrates on the behavior of colorectal carcinomas, this parameter is not currently recognized as a reliable prognostic factor. We showed that the nature, functional orientation, density, and location of immune cells within distinct tumor regions could provide a prognostic factor superior to and independent of criteria related to the anatomic extent of the tumor. The strength of the immune reaction identified in our studies could advance our understanding of cancer evolution and have important consequences for clinical practice.
Key Words: tumor immunology • memory T cells • prognostic factors
INTRODUCTION
To grow, invade, and metastasize, a tumor interacts with its microenvironment, composed of diverse cells of various origins that form the tumor matrix, its vascularization, its lymphatic network, and its neurologic accompaniment. These structures have dual roles: They contain the tumor and provide it with oxygen and nutrients but also favor its dissemination [1 ]. The microenvironment also contains cells of the immune system, including inflammatory infiltrates of innate immunity [macrophages, neutrophils, mast cells, NK cells, and immature dendritic cells (DC)] and infiltrates of the adaptive immune response [mature DC, B lymphocytes, and T lymphocytes comprising Th1 or Th2 cells, cytotoxic, and regulatory T cells (Tregs)]. Tumor-infiltrating T cells live in a milieu of chemokines and cytokines produced by the tumor cells, stromal cells, as well as the infiltrating immune cells. The interactions between the malignant cells and the local immune infiltration are complex and result in a balance between tumor-promoting and tumor-controlling effects. The role of chronic inflammation in favoring tumor development is well documented [2 3 4 ]. In contrast, acute inflammation may be anti-tumoral, and this property, unveiled by William Coley a century ago, has led to the use of therapeutic agents that stimulate inflammation and innate immunity, such as bacillus Calmette-Guerin for superficial bladder cancer [5 ] or more recently, CpG [6 ]. The role of adaptive immune reactions is still a matter of debate. The immunosurveillance theory [7 ] has dominated the field for a long time but has been difficult to demonstrate. The strongest support comes from clinical observations of increased cancer incidence in immuno-depressed patients [8 ] and the development of tumors originating from a transplanted kidney when the recipient was subjected to immunosuppressive drugs [9 ]. The latter observations established that in humans, tumor cells may remain in a dormant state for years. A clear demonstration of a protective role for the immune system against tumors came from the analysis of immunocompromised mice. Animals with specific disruptions of genes involved in adaptive or innate immunity exhibit higher sensitivities to chemically induced and more importantly, spontaneous tumor appearance [10 ]. These observations led R. D. Schreiber and co-workers [11 ] to propose the cancer immunoediting theory, in which a tumor interacts with the immune system in three sequential phases: the elimination phase, when the immune system clears most tumor cells; the equilibrium phase, when the tumor is present but contained [12 ]; and finally, the escape phase, when the tumor grows, invades, and metastasizes. The last phase is a consequence of immune exhaustion or inhibition or results from the emergence of tumor-cell variants. In man, it is difficult to establish the reality of this theory. Although the presence of high numbers of tumor-infiltrating lymphocytes, first observed in melanoma [13 , 14 ], is often of good prognosis, there are many reports of local lymphocyte inactivation, immune deviation, or the presence of Tregs, which are associated with the occurrence of metastasis and poor clinical outcome [15 ].
The present manuscript will review our attempts to apply large-scale analysis of many immunological parameters (lymphocyte densities and localization, activation and memory states, lymphocyte-associated receptors, inflammatory cells) to large cohorts of human colorectal tumors from patients with long-term follow-up to dissect the local immune reactions and investigate a role for natural immunity in the control of cancer [16 , 17 ]. Our observations lead to the striking conclusion that the in situ immune reaction might be the strongest parameter influencing clinical outcome, regardless of the local tumor extension and its spread to lymph nodes.
Th1 ORIENTATION WITH MEMORY T CELLS IN THE TUMOR CORRELATES WITH THE ABSENCE OF EARLY SIGNS OF METASTATIC INVASION
Several reports have documented that deposits of single tumor cells or very small clusters of neoplastic cells outside the primary site of a tumor are common features of colorectal cancers and have even been observed in a subset of clinically localized tumors [18 ]. Whether circulating neoplastic cells represent metastatic dissemination or are merely cancer cells without metastatic potential is still controversial. This finding, however, changes the general concept of an organ disease toward a general disease. Tumor cells face the immune system. Immune defenses challenge tumor cells at the primary site but also in lymphovascular and perineural structures around the tumor, in lymph nodes, in the blood stream, in bone marrow, and later, in distant organs. To the best of our knowledge, no studies have investigated whether the nature of the immune infiltrate at the primary site influences the appearance of early signs of metastatic invasion, i.e., vascular emboli (VE), lymphatic invasion (LI), and perineural infiltration (PI; collectively referred to as "VELIPI").
We first investigated a cohort of 959 colorectal tumors from patients presenting with diverse tumor-node-metastasis (TNM) grading and from all colorectal locations (from right colon to rectum) with up to 15 years of clinical follow-up (mean duration of follow-up: 44.5 months; min:max values until progression/death or last follow-up: 0–214 months) for the presence or absence of VELIPI. From this large cohort, we confirmed the debated, pejorative prognostic value of the presence of VELIPI in tumors. Patients with no sign of invasion in their primary tumor (VELIPI-negative) had longer disease-free survival (DFS) and overall survival (OS) than patients with early signs of metastatic invasion (VELIPI-positive) based on Log-rank tests comparing patient groups (all P<0.001).
We then analyzed tumor-infiltrating immune cells by gene-expression profiling, flow cytometry, and in situ immunohistochemistry and correlated the data with the VELIPI status of the tumor. After analyzing immune-associated gene expression in 75 randomly selected tumors, we established that increased levels of genes associated with Th1 orientation (T-bet, IFN-
) and effector function [CD8
, granzyme B (GZMB), and granulysin] were associated with the absence of signs of invasion (VELIPI). In contrast, we found no correlation between the VELIPI status of the tumors and markers of inflammation [IL-8, vascular endothelial growth factor, carcinoembryonic antigen (CEA)-related cell adhesion molecule 1, matrix metalloproteinase-7, cyclooxygenase-2, and thrombospondin-1] or Th2 orientation (GATA-3). Next, we performed detailed analyses of immune infiltrates in 39 tumors using flow cytometry with a large array of mAb. Hierarchical clustering showed that markers of T cell migration (CD62 ligand, CCR7, CD103, CD49d, and CXCR3), activation (HLA-DR, CD98, CD80, CD86, and CD134), and differentiation (CD45RO, CD45RA, CD27, CD28, CCR7, and CD127) were increased in VELIPI-negative tumors. Colorectal cancers without tumor embolies presented with significant increases in T cell subpopulations from early memory (CD45RO+CCR7–CD28+CD27+) to effector memory CD8 T cells (CD45RO+CCR7–CD28–CD27–).
According to recent literature, Tregs may play an important role in host anti-tumor immune response [19 , 20 ]. We determined by flow cytometry that VELIPI-positive tumors had increased frequencies of CD4+CD25hi T cells among CD3+ cells compared with VELIPI-negative tumors, but no difference between groups was found when CD4+CD25hi cells were assessed as a percentage of total cells. Moreover, expression levels of forkhead box P3, CTLA-4, glucocorticoid-induced TNFR, IL-10, and TGF-β mRNA (often associated with Treg) were not increased in either group of patients [21 ]. The data did not reveal any obvious associations between Treg infiltration and early signs of metastatic invasion.
An immunosuppressive microenvironment that supports tumor growth, dissemination, and immune escape could also be the consequence of myeloid-derived suppressor cells [22 ], which when found in the spleens of mice with colon cancer, block T cell function through NO and arginase production [23 ]. Additionally, a recent report showed that microvesicles released by melanoma and colorectal carcinoma cell lines could promote the differentiation of monocytes into myeloid-derived suppressor cells producing TGF-β [24 ]. Whether such cells are detected in human colorectal carcinomas and influence the nature and the functionality of T cell subpopulations, such as memory T cells, remains to be determined.
By immunohistochemistry-based tissue MicroArray analysis of 415 tumors, we determined the density of memory T cell (CD45RO+) infiltration in tumors. Immunostaining was quantified using a dedicated image analysis workstation, providing a robust and reliable method [17 ] to determine the density of immunostained cells. We confirmed that a high density of memory T cells (CD45RO+ cells) in the primary tumor correlated with the absence of early signs of metastatic invasion (VELIPI). Importantly, a high density of memory T cells also correlated with the absence of lymph node invasion or distant metastasis.
Together, our results showed a positive correlation between a specific immune orientation—Th1 differentiation with cytotoxic and memory components—in tumors and the absence of microscopic evidence of early signs of invasiveness around the tumor and metastasis in distant organs. An appealing interpretation of these data is that even when a tumor has already reached a clinical stage, tumor-associated (expression of antigenic determinants, danger signals, etc.) and inflammatory elements might convene to recruit an efficient, adaptive immune reaction that keeps tumor emboli in check. The natural history of a colorectal cancer could associate an immune escape at the tumor site with a degree of recognition of tumor emboli leading to an equilibrium phase following surgical removal of the tumor.
AN ADAPTIVE IMMUNE REACTION IN TUMORS CORRELATES WITH THE ABSENCE OF TUMOR RECURRENCE
We applied bioinformatics approaches to investigate functional patterns and the coordination of immune cell populations within primary tumors. Reanalysis of genomic data showed 39 highly significant combinations of comodulated genes (e.g., CD3/CD8, CD8/T-bet, CD8/IFN-, IFN-/granulysin (GLNY), IL-10/TGF-β; all P<0.0001). We performed pair-wise comparisons by measuring the similarity between profiles using Pearson correlation coefficients. The relationships implied by these correlations were visualized using hierarchical clustering of a correlation matrix, which revealed three major clusters of comodulated genes related to inflammation, immunosuppression, and Th1/cytotoxic orientation. Strikingly, the cluster of genes associated with Th1 and cytotoxic phenotypes (T-bet, IFN-regulatory factor-1, IFN-, CD3
, CD8, granulysin, and GZMB) was the sole cluster associated with relapse. The relapse rate inversely correlated with the level of expression of Th1 and cytotoxic-associated genes. Only 20% of patients with high expression of these genes underwent relapse, in contrast to 80% of patients with low expression of these genes in their primary tumors. The Log-rank tests comparing the DFS times reached significance (P<0.05). We found no statistical correlation between relapse rates and the expression of genes associated with immunosuppression and inflammation.
Microarray technology has permitted the development of multiorgan cancer classifiers, identification of tumor subclasses, discovery of progression markers, and prediction of disease outcome in many types of cancer, including colorectal cancer [25 26 27 ]. The use of low-density arrays focusing on immune genes combined with novel bioinformatic tools allowed us to identify comodulated genes with known physiological functions strongly associated with clinical outcome. These data reinforce the hypothesis that an efficient Th1 response, required for the establishment and maintenance of T cell memory, might control tumor recurrence by mechanisms such as preventing tumor dissemination but probably also by rendering migrated tumor cells unable to develop clinically detectable metastases.
THE INTRATUMORAL LOCATION OF INFILTRATING T CELLS IS AN IMPORTANT FEATURE FOR PREDICTION OF CLINICAL OUTCOME
In colorectal tumors, T cells are distributed in different tumoral areas: in the center of the tumor (CT) within the stroma and tumor glands, in the invasive margin (IM), and in lymphoid structures resembling germinal centers at the tumor periphery. Previous studies have suggested a protective role for immune infiltrates in colorectal tumors [28 29 30 31 32 ], although this conclusion has been challenged [33 ]. Variations in the descriptions of tumor types and subjective grading of host lymphoid responses have favored nonreproducibility of previous findings. The use of underpowered analyses has also been problematic. We aimed to determine in a large cohort of patients if the determination of functional orientation, density, and location of immune cell populations within distinct tumor territories could improve the accuracy of predicting patient outcome.
The CT and the IM of 415, 119, and 75 colorectal cancers (three independent series) were investigated for the presence of total T lymphocytes (CD3), CD8 T cell effectors, and their associated cytotoxic molecule (GZMB), as well as memory T cells (CD45RO). Densities of immune cells in 7384 immunostainings were quantified with the use of a dedicated image analysis workstation. High densities of CD3, CD8, CD45RO, and GZMB-positive cells in each tumor region (CT and IM) correlated with a favorable prognosis for DFS (log-rank P values ranging from 1.0x10–2 to 4.8x10–6) and OS (log-rank P values ranging from 5.5x10–3 to 7.9x10–8; Fig. 1 ). The combined analysis of CT plus IM regions strongly increased the accuracy of predicted DFS and OS times for the different groups of patients compared with single-region analysis (CT/IM-high vs. CT-high or IM-high; CT/IM-low vs. CT-low or IM-low; Fig. 1 ). For example, CD45RO CT/IM density was associated with highly significant P values for DFS and OS analyses (P values of 6.6x10–6 and 8.6x10–6, respectively). The respective 5-year DFS and OS rates were 76.9% and 70.5% for patients with CD45RO CT/IM-Hi tumors and 24.2% and 22.5% for patients with CD45RO CT/IM-Lo tumors (Fig. 1) . For each marker studied, a low density of cells in a single tumor region markedly impaired the beneficial effect of the immune cell infiltrate at the tumor site. For example, the respective 5-year DFS and OS rates were 50.5% and 44.1% for patients with CD45RO CT/IM-heterogeneous tumors (LoHi or HiLo). Data were reanalyzed using twofold cross-validation [34 ] with 100 repetitions using the median cell density as a cut-off (50% of patients with high and 50% of patients with low density). The immune markers remained highly significant parameters associated with DFS and OS survival. Thus, immune cell organization and densities within tumors strongly influenced prognosis of colorectal cancer patients. This may indicate that the coordination of the immune reaction among tumor regions is required to mount an effective immune response but also that the immune compartments in each tumor region may control different tumor events (e.g., tumor emboli, lymph node metastasis, distant metastases).
 View larger version (39K): [in a new window] |
Figure 1. (A) The H&E section of colon cancer (original magnification, x40); CT (original magnification, x100); IM (original magnification, x100). (B) Kaplan Meier curves for the duration of DFS according to memory T cell (CD45RO) density in the IM of the tumor (CD45ROIM), in the CT of the tumor (CD45ROCT), in the CT and the IM evaluated as the mean of CT + IM densities for CD45RO+ cells (CD45ROmeanCT+IM), and evaluated in combination (CT and IM) for CD45RO+ cell densities (CD45ROCT/IM). A combined analysis of the CT and the IM (CT/IM) provides the most accurate method to discriminate patient outcome. HR, Hazard ratio.
|
A note of caution is needed when interpreting the correlations we found between the nature and density of the in situ immune reaction and the signs of tumor dissemination and clinical outcome. Our observations provide strong but indirect evidence of immune-mediated control of colorectal cancer progression. We cannot exclude the possibilities that intratumoral lymphocytes represent an epiphenomenon or that they influence tumor cells and its stroma in such a way that they attenuate the metastatic capacity of tumor cells.
About 15–20% of sporadic colorectal cancers exhibit microsatellite instability (MSI), resulting from defective mismatch repair, commonly a result of hypermethylation of the homosapiens mutL homolog 1 gene [35
]. Patients bearing MSI+ colorectal cancers have favorable clinical outcomes [36
]. The mechanism by which MSI renders cancers less aggressive remains unresolved. One study suggested that the MSI+ phenotype is associated with a favorable response to adjuvant chemotherapy [37
]. The MSI-H subset is characterized by a dense infiltrate of intraepithelial lymphocytes. Substratification of patients according to MSI status may determine if the immune reaction is the major determinant of a more favorable course in patients with MSI-H tumors. The median proportion of tumors with MSI-H is only 17%, according to the National Cancer Institute guidelines [35
]. In our studies, we observed that 
50% of colorectal tumors presented with high immune T cell infiltrates in the CT or IM. Thirty-four percent of patients presented with high immune T cell infiltrates in both regions. The higher proportion of tumors with strong lymphocytic infiltration than that observed for MSI-H tumors, particularly in rectal cancers (where it is below 5%), indicates that the beneficial effect of high immune infiltration in tumors is not an epiphenomenon related to the concomitant MSI-H phenotype. One attractive hypothesis to explain this phenomenon is that the mutator phenotype may increase the production of abnormal peptides able to elicit cytotoxic immune responses against tumor cells. The ability of the immune system to mount anti-tumor immune responses against tumors is illustrated by studies showing the presence of HLA-restricted T cells (CD4+ and CD8+) specific for tumor cells in tumor-infiltrating lymphocytes and PBMCs from patients with colorectal cancer [38
]. Also, large amounts of antibodies against tumor cell determinants (p53, CEA, Ras, topoisomerase II-, histone deacetylase 3 and 5, ubiquitin C-terminal hydrolase L3, tropomyosin, and cyclin B1) [39
] are detectable in the serum of patients with colorectal cancer. This indicates that some of the patients’ T and B cells could target colorectal cancer cells.
We found a positive correlation between the presence of markers for Th1 polarization, cytotoxic T cells, and memory T cells and a low incidence of tumor recurrence. This argues for immune-mediated rejection of persistent tumor cells after surgery. However, functional studies showing ongoing T cell responses against tumors are lacking. We are currently investigating the T cell repertoire of the immune infiltrate for evidence that one or more colorectal cancer antigens are targets of tumor-infiltrating lymphocytes. We hypothesize that memory T cells play a central role in the control of tumor recurrence. Primary CD8+ T cell responses to antigen result in the clonal expansion and contraction of antigen-specific CD8+ T cells and the generation of memory T cells within a few weeks. This memory cell population can be maintained for many years and is capable of rapid responses upon antigen re-exposure [40 ]. Anti-tumor immune responses may persist following surgical removal of a tumor, as occult tumor cells can be detected in blood and bone marrow in a significant percentage of patients with colorectal cancers [18 ]. In mice, protective immunity against colon cancer is mediated in part by long-lived memory T cells [41 ].
THE IN SITU ADAPTIVE IMMUNE REACTION COULD BE THE STRONGEST PROGNOSTIC FACTOR FOR COLORECTAL CANCER
Prognosis of colorectal cancer is classically determined by the TNM staging system [42 ] that relies on the depth of tumor invasion and the absence or presence of nodal and distant metastases (Fig. 2 ). This system is used for estimating prognosis, planning treatment, and measuring outcome end results. The primary treatment modality for colon cancer is surgery. Patients presenting with Dukes’ stage A or TNM I stage are subjected to curative surgery alone. Once the disease has spread to lymph nodes or to distant organ sites, chemotherapy is incorporated into treatment plans. Adjuvant chemotherapy is recommended for patients with nodal disease (Dukes’ stage C). Finally, Dukes’ stage D or TNM IV stage patients (distant metastasis) are treated by chemotherapy, with or without surgery. The TNM staging system provides a risk classification for recurrence of colorectal carcinoma and survival; metastatic patients have the worst prognosis, whereas patients with localized disease have the best. The TNM system, however, is far from being a fully reliable predictor, as 20–30% of patients with clinically localized cancer (Dukes’ A and B or TNM I and II) will experience relapse, and up to 40% with Dukes’ C (TNM III) will relapse regardless of treatment. Additional prognostic markers are therefore needed to identify patients who may be at risk for recurrence. Genetic and cellular markers of tumors, such as p53 expression, K-ras mutation, CEA production, and MSI status of the tumor, have been proposed to predict relapse and patient survival, but none has reached a degree of confidence to be recommended for use in the clinic [43 ].
 View larger version (57K): [in a new window] |
Figure 2. (Left panel) The TNM system is based on three components: the extent of the primary tumor (T), the absence or presence of the disease in the lymph nodes (N), and the absence or presence of distant metastasis (M). This staging system, providing indications of prognosis, is far from being a fully reliable predictor for DFS and OS. (Right panel) A strong adaptive immune reaction (IR) in the CT and in the IM of the tumor is associated with a favorable prognosis, whatever the extent of the primary tumor and the invasion of regional lymph nodes. Conversely, a weak adaptive immune reaction in the CT and in the IM of the tumor is associated with a poor prognosis, even in patients with minimal tumor invasion [International Union Against Cancer (UICC)-TNM stages I–II].
|
In multivariate analysis, after adjusting for tumor invasion (T stage), tumor differentiation, and lymph node invasion (N stage), CD3-CT/IM density was the only independent parameter associated with OS (HR=1.89; P=1.2x10–5) and was the criteria with the highest HR and the smallest P value in DFS analysis (HR=2.4; P=1.4x10–6). Moreover, in patients with coordinated high or low densities of immune markers in both tumor regions (HiHi or LoLo, representing 66% of the cohort), we observed that the histopathological parameters were no longer associated with DFS and OS.
This strongly supports the hypothesis that following surgical removal of the tumor, time to recurrence and survival are governed in large part by the state of the local adaptive immune reaction at the tumor site. We thus propose that the postsurgical immune reaction, and not tumor status, is the ultimate sensor of local recurrence and metastasis and therefore, of clinical outcome [44 ].
CONCLUSIONS
A large-scale analysis of immune parameters in cohorts of colorectal cancers with long-term clinical follow-up revealed that tumor infiltration by immune cells with a Th1 orientation and cytotoxic and memory cells correlates with criteria for tumor dissemination, tumor relapse, and patient survival. Our data could support the existence of natural immune responses to human cancer, which influences the natural course of the tumor process, sometimes for periods of several years. As our analyses were performed on clinically diagnosed and surgically removed tumors, this immunosurveillance would apply to the equilibrium rather than the elimination phase. The nature of the factors that recruit and activate an efficient immune reaction (tumor antigens, chemokines, cytokines, DC, macrophages, NK cells, etc.) needs to be elucidated and may vary between tumors and patients. Conversely, the factors responsible for the lack of an efficient immune response (HLA loss, immune deviation, regulatory cells, myeloid cells with suppressive activity, immunosuppressive factors, tumor cell resistance to killing, etc.) are likely numerous and are important to define and better understand tumor escape and how to counteract it therapeutically.
In any case, the fact that the local immune contexture (i.e., density, phenotype, activation status, and location of immune cells) appears to be the strongest prognostic factor for DFS and OS in colorectal cancer, provided our findings can be validated at other centers, indicates that immunological criteria could be included in tumor staging to improve the identification of high-risk patients who would benefit most from adjuvant therapy. This approach should be extended to other cancers, particularly those where high numbers of infiltrating T cells have been associated with good prognosis.
Received November 18, 2007; revised April 7, 2008; accepted April 21, 2008.
REFERENCES
and lymphocytes prevent primary tumor development and shape tumor immunogenicity Nature 410,1107-1111[CrossRef][Medline]This article has been cited by other articles:
 |
|
 |
|
  S. Ogino, K. Nosho, N. Irahara, J. A. Meyerhardt, Y. Baba, K. Shima, J. N. Glickman, C. R. Ferrone, M. Mino-Kenudson, N. Tanaka, et al. Lymphocytic Reaction to Colorectal Cancer Is Associated with Longer Survival, Independent of Lymph Node Count, Microsatellite Instability, and CpG Island Methylator Phenotype Clin. Cancer Res., October 15, 2009; 15(20): 6412 - 6420. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
