Critical role of spatial interaction between CD8⁺ and Foxp3⁺ cells in human gastric cancer: the distance matters

Abstract

Purpose

In various cancer types, an abundance of FoxP3⁺ regulatory T cells (Treg) has been associated with an unfavorable outcome. Yet, the role of Treg on cancer immunity has been shown to be complex. In single cell marker technique, other tumor-infiltrating lymphocytes (TILs) such as cytotoxic CD8⁺ T cells (CTL) also influenced prognosis. This study for the first time investigates the concurrent spatial distribution pattern of CD8⁺ and FoxP3⁺ TILs and their prognostic impact in human gastric cancer.

Materials and methods

Tumor tissue microarrays of 50 patients with surgically treated adenocarcinoma of the cardia were studied. An immunohistochemical double staining of CD8⁺ and FoxP3⁺ TILs was performed. Cell counts and cell-to-cell distances in tumor epithelium and stroma were evaluated with image-processing software. Metastasis-free survival, no-evidence-of-disease survival, and overall survival were investigated (mean follow-up time 6.9 years).

Results

High intraepithelial infiltration of CD8⁺ and FoxP3⁺ TIL was associated with the improved 10-year metastasis-free survival (83 vs. 54 %, p = 0.04 and 85 vs. 59 %, p = 0.09, respectively). Considering cell-to-cell distance and comparing patients with functional (30–110 μm) versus nonfunctional distances of CD8⁺ and FoxP3⁺ TILs, 10-year survival rates differed between 89 and 55 % (p = 0.009), respectively.

Conclusion

Prognostic influence of tumor-infiltrating immune cells in gastric cancer critically depends on their cell-to-cell distance. FoxP3⁺ TILs must be located within a distance between 30 and 110 μm of CD8⁺ T cells to positively impact on prognosis.

Electronic supplementary material

The online version of this article (doi:10.1007/s00262-013-1491-x) contains supplementary material, which is available to authorized users.

Introduction

FoxP3 is a transcription factor characteristic for immune-suppressive T regulatory cells (Treg). Treg inhibit effector T cells both by direct cell–cell contact and cytokine secretion [1]. An accumulation of FoxP3⁺ Treg was detected in various cancer types. Therefore, the prognostic effect of this subtype of tumor-infiltrating lymphocytes (TIL) has come into focus. In several tumor settings, high numbers of Treg were associated with adverse clinical outcome [2–5]. However, recently conflicting data indicated that in tumor types, high intratumoral infiltration of FoxP3⁺ Treg was associated with a better clinical outcome [6]. Currently, contradictory mechanisms regarding the effect of Treg on tumor cells are discussed considering both the pro-inflammatory and the regulatory function on surrounding lymphocytes [7]. Therefore, there is not only a need to define the prognostic significance of single TIL subsets but also to clarify their interrelations between TILs in a tumor model.

Therefore, we focused on immunosurveillance mechanisms of Treg in gastric adenocarcinoma of the cardia. Although overall incidence of gastric adenocarcinoma is decreasing, cancer of the gastric cardia is on the rise [8]. There are two major different growth patterns in gastric adenocarcinoma: intestinal type with glandular growth pattern and diffuse type with dispersed tumor cells. Loss in cell–cell-adhesion molecules as cadherin is a hallmark of diffuse-type gastric cancers and poorly differentiated intestinal-type cancer [9]. This along with the amount of stromal collagen has been seen as a determinant of TIL infiltration density [10]. Ohno et al. showed that increased stromal collagen worked as a barrier for cytotoxic T lymphocytes (CTL) infiltration. CTL are considered to play a protective role against human neoplasms as antitumor effects in vivo are dominantly achieved by cell-mediated immunity [11]. Tumor stage at the time of diagnosis remains the most powerful prognostic indicator for gastric cancer. In curative treatment of gastric cancer, a systemic therapy along with the surgical resection comes more and more into focus [12]. In this context, further markers for clinical decision-making are investigated. Our goal was to characterize the interrelationship between CTL and Treg among TIL in human gastric cancer of the cardia. In the present study, double stainings for CD8⁺ and FoxP3⁺ TIL were performed in human gastric cancer so that in addition to the distribution density of TIL, the topographical relationship could be calculated to assess a potential relevance of cell–cell interactions.

Materials and methods

Patient selection

For the present study, a total of 135 patients were enrolled with gastric adenocarcinoma of the cardia treated at this university hospital. The clinical data on the study group were previously published by Haas et al. [13, 14]. In brief, inclusion criteria were as follows: (1) an invasive gastric adenocarcinoma of the cardia, intestinal subtype according to the Laurén classification (patients with esophageal Barrett’s adenocarcinoma were excluded); (2) treatment by complete surgical resection (R0); (3) no (neo-) adjuvant radiotherapy or chemotherapy was performed; (4) staging performed in accordance with the system of the International Union Against Cancer (UICC 2002). Thus, a study group of 50 patients was further investigated. Tumor samples were processed into tissue microarrays (TMA), which were constructed as previously described [14]. Data concerning histopathological features, therapy procedures, actual metastases state, recurrence, and disease-specific survival were acquired from the medical records of this university hospital and regularly scheduled follow-up visits from the patients’ primary care physicians. The median follow-up time was 6.9 years.

Immunohistochemistry

Double stainings were performed with CD8 (DakoCytomation, Hamburg, Germany) and FoxP3 (Abcam, Cambridge, UK)-specific antibodies. In brief, sections were incubated overnight with a FoxP3-specific antibody and afterward with biotin-labeled secondary antibody (biotin rabbit-anti-goat). Biotin was visualized using streptavidin-biotinylated alkaline phosphatase complex (DakoCytomation, Hamburg, Germany). Fast Red was used as chromogen. A double staining enhancer (Zymed, San Francisco, California, USA) was used followed by an avidin and biotin block (Avidin/Biotin Blocking Kit, Vector Laboratories, Inc. Burlingame, CA). Slides were incubated with primary antibodies specific for CD8 after applying a post-block-solution; secondary antibodies connected covalently with an AP-Polymer against mouse (ZytoChem-Plus, Berlin, Germany) were used with Fast Blue (Sigma-Aldrich, Taufkirchen, Germany) as chromogen.

Quantification of TILs

Stained slides were scanned with a high throughput scanner (Zeiss, Mirax MIDI Scan, Göttingen, Germany) at a magnification of 1:200. Evaluation of the T cell density was performed with semiautomatic image-processing software (Biomas, Erlangen, Germany). At least two TMA core sections were evaluated per patient. In the double stains, both the epithelial and stromal compartments were selected on tissue images so that both compartments could be evaluated separately. The epithelial compartment was defined as the atypical glandular tumor cell nests, whereas the stromal compartment was defined as the connective tissue in between atypical glands (Fig. 1a). The areas (in mm²) of the stromal and epithelial compartment were automatically calculated, and positive cells were labeled manually and counted with the software. The numbers of lymphocytes per mm² were counted, and the value was referred to as distribution density. Mean cell-to-cell distances between CD8⁺ and CD8⁺ lymphocytes, FoxP3⁺ and FoxP3⁺ lymphocytes, CD8⁺ and FoxP3⁺ lymphocytes as well as FoxP3⁺ and CD8⁺ lymphocytes were automatically calculated in both compartments.

Fig. 1.

Analysis of gastric cancer tissues. a Epithelial (EC) and stromal (SC) compartments were marked manually, and the different surface areas were calculated automatically. b The count of positively stained cells was set into relation to the surface area, and mean distance values between the different cell types per each compartment were calculated by the software. Distance to the nearest cell indicated by red arrows FoxP3⁺-to-FoxP3⁺, blue arrows CD8⁺-to-CD8⁺, yellow arrows FoxP3⁺-to-CD8⁺, and green arrows CD8⁺-to-FoxP3⁺. c–f Scatter plots showing the correlation of FoxP3 and CD8⁺ distribution densities in the stromal and intraepithelial compartment

When measuring the cell-to-cell distance, some points have to be considered: There is a relationship between the distribution density of cells and the distances between the cells. For uniformly distributed cells inside an area, the distances between cells would be expected to be the inverse square root of the lymphocytes per area:

$$\text{Cell-to-cell distance}=\frac{p0}{\sqrt{n}}$$ 1

where n is the number of cells per area and p0 is the matching coefficient (Supplementary figure 1A). However, this equation is only true for theoretical cells without any extension and with a uniform spatial distribution. Therefore, to fit the data points, we used a similar equation and introduced a constant for the Y offset due to the diameter of the cells and the nonuniform distribution of the cells to fit the data points:

$$\text{Cell-to-cell distance}=\frac{p0}{\sqrt{n}}+p2$$ 2

where p2 is an offset in the Y direction (Supplementary figure 1B). Additionally, this function is not defined for zero. Toward zero the function`s value goes to infinity. As there are no infinite distances between cells, another parameter is introduced which moves the function to the left:

$$\text{Cell-to-cell distance}=\frac{p0}{\sqrt{n+p1}}+p2$$ 3

where p1 is an offset in the X direction (Supplementary figure 1C). The constants p0, p1, and p2 are calculated by the computer algebra system Techplot (Techplot7, SFTek, Braunschweig, Germany).

Cut-off values

Median values were used as cut-off in distribution densities. To estimate cut-off values for the cell-to-cell distances, a Gaussian distribution was used. The Kolmogorov–Smirnov test was performed to determine whether each variable had a near normal distribution. A Gaussian distribution was fitted, and the cut-off points were defined by a value of one standard deviation from the mean (68.3 % CI) (Supplementary figure 2).

Statistical methods

For all immunohistochemical markers, a cut-off value for the definition of subgroups was the median. Overall survival, no-evidence-of-disease (NED) survival, and metastasis-free survival rates were calculated according to Kaplan and Meier and compared by the log-rank test (SPSS, Chicago, IL, USA). Correlation analysis for possible association between the different variables was performed using Pearson’s correlation coefficient. Overall survival was defined as the period from surgery until death or end of the follow-up. NED survival was defined as the period from surgery until evidence of recurrence (local, nodal, or distant recurrence). Metastasis-free survival was the period until appearance of distant metastases.

Results

Study group

Clinical and histological characteristics of the study group were recently published [13, 14]. Overall survival, NED survival, and metastasis-free survival of the study group were 75.0, 71.2, and 73.1 % at 10 years, respectively. The median follow-up was 6.9 years. Median NED survival, metastasis-free survival, and overall survival time for all patients were 13.2 years (range 11.0–15.5 years), 13.6 years (range 11.4–15.8 years), and 13.9 years (range 11.7–16.1 years), respectively.

TIL counts

FoxP3⁺ and CD8⁺ TILs were analyzed simultaneously by an immunohistochemistry double staining technique. Representative microscopic images are shown in Fig. 1a and b. Cells in the intraepithelial and stromal compartment were counted separately as shown in Fig. 1a. Numbers of intraepithelial FoxP3⁺ and CD8⁺ cells correlated positively with numbers of stromal CD8⁺ cells [Fig. 1c, Pearson’s correlation coefficient (r) = 0.63, p < 0.001] and FoxP3⁺ cells (Fig. 1d, r = 0.62, p < 0.001). Distribution densities of Treg and CTL correlated positively in both the intraepithelial (Fig. 1e, r = 0.68, p < 0.001) and stromal compartment (Fig. 1f, r = 0.86, p < 0.001, Supplementary table 1). As published before (14), an increasing stromal FoxP3 + TIL infiltration was inversely correlated with UICC stage (Pearson’s correlation coefficient, r = −0.40; p = 0.001), number of lymph node metastases (Pearson’s correlation coefficient, r = −0.36; p = 0.009), and N category in general (Pearson’s correlation coefficient, r = −0.36; p = 0.023). For CD8 + cells, no such correlation could be found.

Survival analysis

Patients with higher intraepithelial FoxP3⁺ and CD8⁺ TIL counts had a significantly better clinical outcome concerning the metastasis-free survival (Fig. 2a, c). Ten-year metastasis-free survival was 82 versus 59 %, p = 0.09 and 83 versus 54 %, p = 0.04, respectively. Considering the stromal compartment only, no influence of TIL infiltration was seen on the metastasis-free survival (p = 0.77 and 0.66; Fig. 2b, d). No significant influence of intraepithelial TIL counts was shown on NED survival or overall survival. CD8⁺/FoxP3⁺ ratio, both intraepithelial and stromal, was not associated with differences in metastasis-free survival, NED survival, and overall survival.

Fig. 2.

Kaplan–Meier analysis for patients with high and low FoxP3⁺ cells and CD8⁺ cells in intraepithelial and stromal compartment of tumor tissue. The median was used as cut-off value (the respective values can be seen in the curve caption). a Impact of intraepithelial, b stromal FoxP3⁺ cells, c impact of intraepithelial CD8⁺ cells, and d stromal CD8⁺ cells. Patients with counts above and below median were compared by the log-rank test. e, f Correlation of FoxP3⁺-to-FoxP3⁺ and CD8⁺-to-CD8⁺ distances in the stromal and intraepithelial compartment

TIL cell-to-cell distances

Distances between the nearest adjacent TILs were assessed using an image-processing software as shown in Fig. 1b. In both the epithelial and the stromal compartment, the shortest distance between CD8⁺ cells and between FoxP3⁺ cells was calculated as well as the shortest distance between those two cell types (Fig. 1b). Mean distances CD8⁺-CD8⁺, FoxP3⁺-FoxP3⁺, CD8⁺-FoxP3⁺, and FoxP3⁺-CD8⁺ in the intraepithelial compartment were longer than in the stromal compartment (Table 1, Supplementary table 2). In contrast to the cell distribution density, there was no significant correlation between the distances CD8⁺ to CD8⁺ and FoxP3⁺ to FoxP3⁺ in stromal compartments (r = 0.252, p = 0.071). Cell distances CD8⁺ to CD8⁺ and FoxP3⁺ to FoxP3⁺ in the intraepithelial (r = 0.600, p < 0.001) and stromal compartment (r = 0.653, p < 0.001) shared a positive correlation (Fig. 2e, f). The data fit by Eq. 3 was used to identify lymphocytes uniformly distributed in the tissue. If cells are uniformly distributed, a high correlation between distribution density and cell-to-cell distance would appear. In result, there would be only a limited deviation of the values from the curve progression. The correlation of the distribution density of intraepithelial CD8⁺ cells per area with the calculated distances between the CD8⁺ cells was high (r ² = 0.82). This indicates a uniform distribution of the cells (Fig. 3a). The correlation of the distribution density with the distance of stromal and intraepithelial CD8⁺ and FoxP3⁺ cells was much weaker (r ² < 0.4). This deviation from the curve progression and therefore from the expected values indicates a nonuniform pattern (Fig. 3b–d). A deviation from the expected values to shorter distances might indicate a presence of functionally active cells and mirror their cell-to-cell interactions.

Table 1.

Distribution density of FoxP3⁺ and CD8⁺ cells and shortest average distance between adjacent cells. Mean values were compared by the students’ t test

	Cells (n)	Intraepithelial	Stromal	P
Distribution density (cells/mm2)	FoxP3 (50)	78	130	0.05
	CD8 (50)	159	164	0.88
Distance cell-to-cell (μm)	FoxP3-FoxP3 (49)	79	43	0.002
	CD8-CD8 (49)	57	40	0.01
	FoxP3-CD8 (50)	92	68	0.05
	CD8-FoxP3 (50)	125	70	0.001

Fig. 3.

Analysis of cell-to-cell distances in gastric cancer tissues. a–d Scatter plots showing the correlation of FoxP3⁺ and CD8⁺ TIL distribution densities with FoxP3⁺-to-FoxP3⁺ and CD8⁺-to-CD8⁺ TIL distances in the stromal and intraepithelial compartment. Data points were fitted using the inverse square-root function (Eq. 1). Data points lying within the parallel horizontal lines were used for Kaplan–Meier analysis and were compared to the values outside the horizontal lines. As stated in the curve caption, cut-off values were 30 and 110 μm. Kaplan–Meier analysis of e metastasis-free survival and f overall survival and the impact of CD8⁺-to-FoxP3⁺ intraepithelial distances of TIL

Cell-to-cell distances and survival

It was hypothesized that a short distance between CD8⁺ and FoxP3⁺ cells would be a prognostically adverse event due to the suppression of a tumor-specific CD8⁺ immune response by Treg in the proximity of the CTL. Using the median average cell-to-cell distance to compare short with long distances, there was no difference in metastasis-free, NED survival, and overall survival of the patients (p > 0.2). Figure 3b–d indicates the distribution pattern of cell-to-cell distances which deviate distinctly from the expected uniform distribution according to lymphocytes distribution density.

The nonuniform distribution of the cells prompted us to select cells with short distances, which deviate from the graph and were outside the expected values. The distances were classified into suitable divisions, and a Gaussian fit was performed. The cut-off values were the values within 1 standard deviation from the mean of the normal distribution in the analyzed group (Supplementary figure 2). In result, patients with cell-to-cell distances ranging from 20 to 50 μm for both CD8⁺ to CD8⁺-TIL and FoxP3⁺ to FoxP3⁺-TIL distances were compared to patients having a lower or higher cell-to-cell distance. For CD8⁺ to FoxP3⁺-TIL distances, the range was defined from 30 to 110 μm.

Patients with CD8⁺ to FoxP3⁺ distances between 30 and 110 μm showed a tendency to an advantage in the metastasis-free survival (Fig. 3e). The 10-year metastasis-free survival was 79.6 % in cases with TIL distances within the target range versus 57.2 % for patients with distances outside the target range (p = 0.072). Moreover, overall survival was significantly improved in patients with distance values from CD8⁺ to FoxP3⁺ TILs and between 30 and 110 μm. Ten-year overall survival was 88.7 % for patients with distance values in the target range versus 55.4 % when distance values were outside the target range (p = 0.009, Fig. 3f, Supplementary table 3).

Discussion

Due to the long follow-up period and the double marker technique, this study could identify new prognostic markers in gastric adenocarcinoma of the cardia. Our previous studies identified high numbers of stromal FoxP3⁺ Tregs to be related to longer disease-free survival and to be an independent prognostic factor for a favorable outcome in gastric cancer [14].

Here, we show that high numbers of intraepithelial FoxP3 + Tregs and CD8 + CTL are related to a longer metastasis-free survival. Against intuitive expectation, high numbers of FoxP3 + Tregs are related to beneficial outcome. In contrary to these results, a better outcome was associated with lower numbers of Treg infiltrations in gastric cancer [15, 16]. One study reports high numbers of FoxP3 + Tregs in sentinel lymph node of gastric cancer to be related to metastatic disease in non-sentinel lymph node metastasis [17]. Although there are contradictory results, we want to stress that FoxP3 + Treg infiltration seems to be able to solely function as a strong immunological marker. A question that remains is the mechanism by which infiltration of CD8⁺ and FoxP3⁺ TIL in carcinomas affects prognosis.

In agreement with earlier reports, we confirmed that the density of TILs played an important factor for the clinical outcome [6]. In some studies, high numbers of tumor-infiltrating Treg were shown to be a predictor for adverse survival in advanced gastric cancer [3, 18, 19]. Interestingly, recent data challenged these results by indicating that Treg infiltration in cancer types such as colorectal adenocarcinoma was associated with improved prognosis [20, 21]. These observations are supported by the results of a recent study (20): There accumulating Tregs in human cancer had heterogeneous phenotypes. On the one hand, inducible Treg promoting tumor growth by down-regulating antitumor immune responses were detected. On the other hand, natural Treg could be seen, which influence tumor growth by controlling cancer-associated inflammation [22]. What makes this issue even more complex is the fact that FoxP3 can also be expressed in the tumor cell cytoplasm and in peritumoral epithelial cells. There the accumulation of FoxP3⁺ in tumor cells was associated with decreased lymph node metastases [23]. For the analysis of FoxP3 stainings in our study, only a nuclear positivity of small cells with lymphoid morphology and scant cytoplasm was interpreted as being a positive staining. In our study, tumor cells did not stain positive for FoxP3.

Most studies analyzing FoxP3⁺ TILs and their effect on tumor progression and survival focused on the analysis of TIL in single stains [2, 4, 6, 7, 16–22]. Consequently, the complexity of the functional capacity of FoxP3⁺ TILs on tumor immunity could not be fully appreciated. Here, we examined FoxP3⁺ and CD8⁺ TILs in parallel by using the double marker technique. The present study shows a crucial impact of the topography of TIL on their prognostic impact. When using the median value as a cut-off, distances between TILs did not show an influence on the development of metastasis nor an advantage in survival. Yet, when reclassified in relation to their distribution pattern (Fig. 3a–d), a favorable effect on metastases-free and overall survival became evident (Fig. 3e, f). This suggests that there could be a number of inactive TILs, which are distributed randomly and cannot perform an appropriate immune response. The tumor itself might escape host immune response when actively disarranging TIL by inhibiting their co-localization. CTL priming demands two signals from tumor cells: on the one hand foreign peptides presented by MHC class I molecules and on the other hand co-stimulatory molecules [24]. This fact suggests that topography and spatial arrangement of tumor cells and TIL, respectively, might play a key role to receive these co-stimulatory signals. The biologic effect of Treg might therefore depend on tumor microenvironment and localization in the tumor site. It has been shown that the number of regulatory T cells among the circulating peripheral blood mononuclear cells increases with tumor stage [16]. Mizukami et al. [7] already showed that a peritumoral localization of FoxP3 + Treg in advanced gastric cancer is prognostically beneficial as opposed to a diffuse infiltration pattern in the tumor. In addition, lymph nodes with metastases were found to have a more intense FoxP3 + Treg infiltration than unaffected lymph nodes [25]. This may indicate that topographical distribution of lymphocytes both in the primary tumor and in the draining lymphatic system plays an important role in immune cell functionality. With respect to our results, high counts of TILs only improved survival, when they were allocated in proximity of the neoplastic cells, and thus the epithelial compartment of the tumor. Only in these positions, they appeared to be able to mediate antitumor effects. Perhaps, this result arises from the biologic peculiarity of different cancer types. This theory might explain the dual role in which high numbers of tumor-infiltrating T cells seem to play in carcinogenesis.

In conclusion, this study is the first to report the prognostic significance of cell-to-cell distance between TILs in gastric adenocarcinoma. The intratumoral spatial arrangement of CD8⁺ and FoxP3⁺ lymphocytes might mediate their functionality and their antitumoral effects. Our results suggest that the prognostic significance of TILs depends on their localization and distribution in the tumor tissue. When considering risk of metastases, assessing TIL distances could play a role for prognostic stratification. This is of importance, as to date only the degree of invasion and presence of nodal metastases at the time of diagnosis remain the most powerful prognostic indicators for gastric cancer.

Electronic supplementary material

Below is the link to the electronic supplementary material.