Morphometric analysis of the balance between CXCR3+ T cells and FOXP3+ regulatory T cells in lymphocyte-rich and conventional gastric cancers
- First Online:
- Cite this article as:
- Ohtani, H. & Yoshie, O. Virchows Arch (2010) 456: 615. doi:10.1007/s00428-010-0921-9
- 130 Views
Lymphocyte-rich gastric carcinomas (Ly-rich GCs) are characterized by the formation of lymphoid stroma. Our previous study has shown abundant infiltration of CXCR3+ T cells and their frequent clustering with CXCL9 (monokine induced by interferon-γ)-expressing stromal cells in the lymphoid stroma of Ly-rich GCs. Foxp3+ regulatory T cells (Tregs) suppress immune responses in the peripheral tissues and play a role in the immunosuppression of cancer tissues. The present study was therefore undertaken to evaluate the significance of the balance between CXCR3+ T cells and Tregs in 44 Ly-rich and 37 conventional GCs by morphometrical analyses of immunohistochemistry. Compared with the pronounced infiltration of CXCR3+ T cells in the lymphoid stroma, the numbers of Foxp3+ cells were relatively low in Ly-rich GCs. Therefore, the ratios of CXCR3+/Foxp3+ cells were much higher in Ly-rich GCs than in conventional GCs. This suggests the occurrence of T-helper type 1 (Th1)-shifted immune responses in Ly-rich GCs. On the other hand, conventional GCs were characterized by a paucity of CXCR3+ T cells and a relative abundance of Tregs. Furthermore, the stroma inside the cancer was characterized by even less CXCR3+ cells, suggesting a strongly immunosuppressive microenvironment. Since Tregs are known to express CCR4, we also examined the tissue distribution of cells expressing its ligand CCL22. CCL22 was not detected in conventional GCs and only sparsely detected in dendritic cells but not in cancer cells in Ly-rich GCs. To conclude, Tregs may play a more important role in conventional GCs than in Ly-rich GCs from the viewpoint of immunosuppression.
KeywordsGastric cancerLymphoid stromaCXCR3Regulatory T cellsTumor immunity
Human cancer tissue is associated with varying degrees of tumor-infiltrating lymphocytes (TILs). Although the significance of TILs as an indicator of host immune responses against cancer cells may not be so straightforward , TILs or a subset of TILs are generally associated with a better prognosis of patients in various cancers, suggesting their protective role in antitumor immunity [2–6]. As an extreme variation, some cancers are characterized by a prominent lymphocytic reaction in the stroma. Among gastric cancers (GCs), such cases are designated as “gastric cancer with lymphoid stroma” or “lymphoepithelioma-like carcinoma (LELC).” They generally show a better prognosis than conventional GCs [7, 8]. Cancer cells in typical LELC are frequently associated with Epstein-Barr (EB) virus infection [9, 10]. In the present paper, we use the term “lymphocyte-rich GCs (Ly-rich GCs)” to include typical LELC and other GCs associated with a rich distribution of lymphocytes in the stroma (for details, see the first section of “Materials and methods”).
Type 1 immune responses are exerted by type 1 effector T cells (T-helper type 1 [Th1] cells, cytotoxic T cells [CTLs]) and natural killer cells (NK cells) and are important in host defense against intracellular pathogens and tumor cells. CXCR3 is one of the representative chemokine receptors for type 1 immune cells, especially Th1 cells [11, 12]. Previously, we have shown that the majority of TILs in lymphocyte-rich GCs express CXCR3, while CXCL9 (also called monokine induced by interferon-γ [MIG]), one of the cognate chemokine ligands of CXCR3, is abundantly expressed by dendritic-shaped stromal cells . As a result, frequent “CXCR3+ T cell–CXCL9+ stromal cell clustering” was observed, which featured the lymphoid stroma of Ly-rich GCs . Furthermore, CXCR3+ T cells clustering with CXCL9+ stromal cells are frequently positive for Ki-67, suggesting some ongoing immune responses .
Recent advances in tumor immunology have revealed that cancers develop by escaping the host immune responses through stepwise processes [14, 15]. Consequently, established cancer tissue is generally considered to be a strongly immunosuppressive microenvironment. In this context, our above-mentioned results from Ly-rich GCs may seem rather inconsistent to the current concept. Therefore, to understand the immunosuppressive aspect of GCs, the present study was undertaken to analyze the potential role of forkhead box protein-3 (Foxp3)-positive regulatory T cells (Tregs) as the mediator of immunosuppression [16, 17]. We compared the distribution density of CXCR3+ cells and that of Foxp3+ cells by morphometric analyses. We reveal here that Ly-rich GCs are characterized not only by the abundance of CXCR3+ cells but also a relative paucity of Foxp3+ cells, resulting in a high ratio of CXCR3+ cells to Foxp3+ cells (termed here as “CXCR3/Foxp3 ratio”). On the other hand, conventional GCs (i.e., GCs other than Ly-rich GCs) are characterized by a relative abundance of Foxp3+ cells and a low CXCR3/Foxp3 ratio, especially within the cancer stroma. Thus, our findings may provide a new insight into the possible differences in immunosuppressive mechanisms between Ly-rich GCs and conventional GCs.
Materials and methods
We use the term “lymphocyte-rich gastric cancer (Ly-rich GC)” to include both LELC and other GCs associated with abundant TILs in the whole stroma (see  for details). The present study enrolled 44 cases of Ly-rich GCs (EBV+: 35 cases; EBV−: 9 cases) and 37 cases of conventional GCs (EBV+: 2 cases; EBV−: 35 cases). The 35 EBV− cases of conventional GCs were consecutively sampled by their histopathological numbering. Intramucosal carcinoma (pTis and pT1[M]) were not included in this study, because typical lymphocyte-rich stroma is observed only when carcinoma cells have started invading the submucosa. All cases were staged as described by tumor, node, and metastasis (TNM) Classification, 7th edition . The present study was approved by the ethical committee of Mito Medical Center.
Enzyme-linked immunohistochemistry (immunoperoxidase method) was performed using formalin-fixed, paraffin-embedded tissue sections as described previously . The EBV status was checked by in situ hybridization for EBER in all cases . The primary antibodies used in this study were mouse monoclonal antibodies to human CXCR3 (CD183; clone 1C6, IgG1; BD, Franklin Lakes, NJ; used at 1:400 = 1.25 μg/mL) and Foxp3 (clone 236A/E7, IgG1; Abcam, Cambridge, MA; used at 1:100), and affinity-purified rabbit polyclonal antibody to human CCL22 (MDC) (PeproTech, Rocky Hill, NJ; used at 1:250 = 4 μg/mL). Tissue sections were incubated with primary antibodies overnight. The staining methods of other lymphocyte markers were described previously . Heat antigen retrieval methods applied before incubation with primary antibodies were as follows: for anti-CXCR3 and anti-CCL22, 10 mmol/L Tris HCl buffer + 1 mmol/L EDTA, pH 9.0, at 95°C for 60 min; for anti-Foxp3, 1:5 dilution of DAKO (Glostrup, Denmark) high-pH buffer (S3308, DAKO), at 95°C for 60 min. Anti-mouse or anti-rabbit Envision (DAKO) was used as the secondary antibody. Diaminobenzidine (DAB; DAKO) was used as the chromogen. Endogenous peroxidase activity was inactivated by immersing tissue sections in 3% H2O2 for 5 min after incubation with primary antibodies.
Enzyme-linked double immunohistochemistry for CD4 or CD8 and Foxp3
The immunoperoxidase method for CD4 or CD8 detection was also performed as described above. Tissue sections were then treated with a 1:5 dilution of DAKO high-pH retrieval solution at 95°C for 20 min to inactivate antibodies and enzymes used in the first step. Then immunohistochemistry for Foxp3 was performed using a 1:40 dilution of anti-Foxp3 antibody. TrueBlue (KPL, Gaithersburg, MA) was used as the chromogen.
The percentage of Foxp3+ cells among CD4+ or CD8+ cells (i.e., the labeling index of Foxp3) was determined using tissue sections double-stained for CD4 or CD8 and Foxp3. The number of total CD4+ (or CD8+) cells and that of CD4+ (or CD8+) Foxp3+ was manually counted in the same areas using an ocular grid (10 × 10-mm lattice) with a 400× microscopic field (40× objective lens and 10× ocular lens using a BX51 microscope; Olympus, Tokyo, Japan). The area of one lattice was 0.0625 mm2. At least three fields were counted for each case, and the numbers were averaged. The labeling index of Foxp3+ cells among CD4+ (or CD8+) cells was calculated by the following formula: the number of CD4+ (or CD8+) Foxp3+ cells divided by the total number of CD4+ (or CD8+) cells in the same field. This analysis was performed in five representative cases of Ly-rich GCs. We also measured the distribution density of CXCR3+ cells and that of Foxp3+ cells in all 81 cases of GC by the same method described above. The results were expressed by the average number of three areas. In this analysis, the most densely distributed areas were selected for cell counting. The size of cancer tissues we examined were 2 to 4 cm in length.
All the statistical analyses were performed using PASW Statistics 18 (SPSS Inc., Chicago, IL), including the Fisher's exact test for the test of contingency table, the Kruskal-Wallis test for the comparison between unpaired two groups, the Pearson correlation coefficient between two variables, and the discriminant analysis to check how variables discriminate two groups. All the tests were judged to be significant when p < 0.05.
Double-labeling immunofluorescence method for CCL22 and CD83
Tissue sections were first immersed in a 1:5 dilution of high-pH antigen retrieval solution at 95°C for 60 min. Nonspecific binding was blocked by Protein Block (X0909, DAKO). The sections were then incubated with a mixture of rabbit antihuman CCL22 (1:100) and mouse monoclonal antihuman CD83 (1:30; Clone 1H4b; Novocastra, Benton Lane, UK) overnight. Alexa fluor 488-labeled donkey antirabbit IgG (1:100 = 20 μg/mL; Molecular Probe, Carlsbad, CA) and Alexa fluor 555-labeled donkey antimouse IgG (1:100 = 20 μg/mL) were applied in a mixture for 30 min. After DAPI (Molecular Probes) nuclear staining, specimens were mounted with ProLong Gold (Molecular Probes). Immunofluorescent observation was performed with a confocal laser scanning microscope (TCS SP5; Leica Microsystems, Wetzlar, Germany).
General histopathological aspects
Stage distribution of cases
A. Ly-rich vs. conventional GCs
Ly-rich GCs (%)
Conventional GCs (%)
B. LELC (a subgroup of Ly-rich GC) vs. others
Nearly all Foxp3+ cells are CD4+ cells in Ly-rich GCs
In Ly-rich GCs, Foxp3+ cells were distributed relatively evenly in the whole area of the stroma. They constituted only a small part of the TILs as compared with abundantly distributed CXCR3+ cells (Fig. 1c, d). Previously, we have shown that the average frequencies of Foxp3+ cells among CXCR3+ cells and CCR4+ cells were 5.1% and 47%, respectively, in Ly-rich GCs . To further characterize Foxp3+ cells, we counted the Foxp3-labeling index among CD4+ and CD8+ cells by double staining in five representative Ly-rich GC cases (Fig. 1e, f). The Foxp3-labeling index among CD4+ cells and CD8+ cells were 98% and 2%, respectively. Therefore, we judged that Foxp3+ cells in Ly-rich GC were mostly CD4+ regulatory T cells (Tregs). This conclusion is consistent with a recent report .
Ly-rich GCs are characterized by a relative paucity in Tregs
We next analyzed the relationship between various parameters. When all GCs were included, the number of CXCR3+ cells positively correlated with the CXCR3/Foxp3 ratio (γ = 0.55, p < 0.0005 [Pearson]; Fig. 2d) and also correlated with the number of Foxp3+ cells (γ = 0.53, p < 0.0005 [Pearson], not shown). Discriminant analysis revealed that the three variables (CXCR3+ cells, Foxp3+ cells, and CXCR3/Foxp3 ratio) contributed to 92.6% discrimination between Ly-rich GCs and conventional GCs. The analyses here also indicated that although the numbers of CXCR3+ and Foxp3+ cells as a whole were in good correlation, the more CXCR3+ T cells infiltrated the cancer tissue, the more CXCR3+ T cells predominated over Tregs
Uneven distribution of TILs in conventional GCs
Stroma inside the cancer in conventional GCs is characterized by a paucity of TILs and a decrease of CXCR3/Treg ratio as compared with the invasive margin
Sporadic dendritic cells (DCs) express CCL22
The present study has shown that the lymphoid stroma of Ly-rich GCs is characterized not only by a drastic increase in CXCR3+ T cells but also by a relative paucity of Tregs, suggesting a microenvironment highly enriched with Th1 cells. In sharp contrast, conventional GCs are characterized by a low number of CXCR3+ T cells and a relative increase in Tregs, suggesting a highly immunosuppressive milieu. These results may raise an interesting issue in terms of tumor immunity and immune evasion in these two groups of GC.
Our previous report was mainly focused on Th1/Th2 responses in Ly-rich GCs and demonstrated a predominant infiltration of CXCR3+ cells over CCR4+ cells . Tregs suppress the type 1 and type 2 immune responses and are regarded to be important for the control of immune responses [16, 17]. Tregs are also considered to participate in the immune evasion mechanism of cancer . Therefore, the effector cell–Treg balance could be important for understanding the immune status of cancer tissues. In fact, a high CD8 to Treg ratio in cancer tissues was associated with a prolonged survival of patients with cancer of the ovary , liver , and kidney . We therefore focused our attention on the balance between CXCR3+ T cells and Tregs in Ly-rich and conventional GCs. By analyzing the CXCR3/Foxp3 ratio, our present study has shown that Ly-rich GCs are characterized by a relative paucity of Tregs in the abundance of CXCR3+ TILs. This, in turn, underscores a relative abundance of Tregs along with a general paucity of TILs in conventional GCs. Therefore, Tregs may play a less major role in Ly-rich GCs than in conventional GCs as the mediator of immune evasion. The immune cell infiltrate in conventional GCs, if present, also tends to be more prominent along the invasive margin than in the stroma inside the cancer. The analysis of the 12 selected cases of conventional GCs clearly demonstrated that the stroma within the cancer is associated with even less CXCR3+ T cells (i.e., less TILs) and more Tregs than the invasive margin. The results from conventional GCs are thus fully consistent with the current concept that tumor tissue is a strongly immunosuppressive microenvironment. We may regard the findings from conventional GCs as the basic feature of cancer tissue.
Concerning the function of Tregs in conventional GCs, Mizukami et al.  have reported that cases with diffuse distribution of Tregs are associated with a poorer prognosis than cases with peritumoral distribution of Tregs. However, Ly-rich GCs, usually associated with a better prognosis, generally show diffuse distribution of Tregs as shown in the present paper. Therefore, the balance between effector T cells and Tregs may also have a strong impact on the significance of TILs.
Tregs are known to frequently express CCR4, and Curiel et al.  have reported that cancer cells and tumor-associated macrophages produce CCL22 to attract CCR4+ Tregs to ovarian cancer tissues. For GCs, Mizukami et al.  have reported that CD14+ cells in the tumor stroma express CCL17 or CCL22. Our present data showed that CCL22 is only sporadically expressed by a part of CD83+-mature DCs but not by cancer cells in Ly-rich GCs. The production of CCL22 by cancer cells may be dependent on the type of cancer. DC expression of CCL22 is consistent with a previous report . As shown in “Results,” the absolute number of Tregs is higher in Ly-rich GCs than in conventional GCs (Fig. 2b), with the number of Tregs being approximately 10% to 20% of CXCR3+ T cells in Ly-rich GCs (Fig. 2d). Thus, the expression of CCL22 by a part of DCs could, at least partly, explain the increased recruitment of Tregs in Ly-rich GCs. We may need further analyses to clarify the mechanism(s) of Treg recruitment in Ly-rich GCs and conventional GCs.
Given that Tregs may not be a dominant immunosuppressive factor in Ly-rich GCs, we may need to analyze possible roles of other immunosuppressive factors such as immunosuppressive cytokines TGF-β and IL-10 in the immune evasion mechanism of Ly-rich GCs. In particular, Ly-rich GCs are frequently associated with EBV, which is known to induce IL-10 . Another important issue to be addressed in the future is the possible involvement of IL-17-producing T helper (Th17) cells in Ly-rich GCs, which are known to play the major role in chronic inflammation and autoimmune pathogenesis, and selectively express CCR6 . These analyses are now ongoing.
In conclusion, our present morphometrical analyses on the balance between CXCR3+ T cells and Foxp3+ Tregs have revealed that Ly-rich GCs are characterized by an absolute and relative abundance of CXCR3+ effector T cells over Tregs in contrast to conventional GCs. On the other hand, conventional GCs, particularly in the stroma inside the cancer, are characterized by an absolute paucity of CXCR3+ effector T cells and a relative abundance of Tregs. The data suggest that Tregs may play a more dominant role in conventional GCs than in Ly-rich GCs from the viewpoint of immunosuppression in cancer tissue. The present findings also highlight the unique immunologic aspect of Ly-rich GCs and the complexity of immune evasion mechanisms used by these types of cancer.
We are grateful to Dr. Masaaki Miyazawa (Kinki University School of Medicine) for critical reading of the manuscript, Dr. Takashi Nakayama (Kinki University School of Medicine) for valuable discussion, Dr. Noriko Kimura (Hakodate National Hospital) and Dr. Hiroshi Naganuma (Sendai City Hospital) for supplying surgical materials, Ms. Fumiko Date (Tohoku University Graduate School of Medicine) for clerical assistance, and Ms. Megumi Maeda and Ms. Shoko Kajiwara (National Institute for Materials Science [NIM.S.]–Leica Bioimaging Laboratory) for technical assistance in confocal laser scanning microscopy. The present study was partly supported by the National Hospital Organization Collaborative Clinical Research Grant as well as the “Nanotechnology Network Project” and “HAITEKU” from the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan.
Conflict of Interest
The authors declare that they have no conflict of interest.