A high proportion of bone marrow T cells with regulatory phenotype (CD4+CD25hiFoxP3+) in Ewing sarcoma patients is associated with metastatic disease

Abstract

Immunosuppressive CD4+CD25^hiFoxP3+ T cells (T_reg cells) have been found at increased densities within the tumor microenvironment in many malignancies and interfere with protective antitumor immune responses. Osseous Ewing sarcomas (ESs) are thought to derive from a bone marrow (BM) mesenchymal cell of origin, and microscopic marrow involvement defines a subpopulation of patients at a high risk of relapse. We hypothesized that BM-resident T cells may contribute to a permissive milieu for immune escape of ESs. Using 6-color-flow cytometry, we investigated the pattern of immune cell subset distribution including NK cells, γδ T cells, central and effector memory CD8+ and CD4+ T cells as well as T cells with regulatory phenotype (T_reg cells) in BM obtained at diagnosis from 45 primary or relapsed ES patients treated within standardized protocols. Although patients at relapse had an inverted CD4:CD8 T-cell ratio, neither CD8+ effector/memory T-cell subsets nor T_reg cells significantly differed from patients at diagnosis. No significant associations of innate and effector/memory T-cell subpopulations with known risk factors were found, including age, gender, tumor site, primary metastases and histological tumor response. By contrast, T_reg cells were found at significantly higher frequencies in patients with primary metastatic disease compared with localized ESs (5.0 vs. 3.3%, p = 0.01). Thus, increased BM T_reg cells in patients with metastasized ES may reflect an immune escape mechanism that contributes to the development of metastatic disease. Immunotherapeutic strategies will have to adequately consider the regulatory milieu within areas of Ewing tumor-immune interactions. © 2009 UICC

Even though the prognosis of Ewing sarcoma (ES) has been improved by multimodal treatment concepts including polychemotherapy, radiotherapy and surgery, patients with primary metastatic disease still have an unfavorable prognosis.1, 2 ES has a number of characteristics that render them amenable to novel immune-based therapeutic strategies. Potential tumor-associated antigens are provided by the unique molecular translocation and the consequent aberrant expression pattern of downstream proteins.3 Indirect support for a role of autologous T cells in maintaining remission in high-risk ES was the observation that early lymphocyte recovery after the first cycle of chemotherapy is an independent favorable prognostic factor.4 Controversial reports exist regarding the expression of adequate HLA class I molecules on ES cells for efficient interaction with tumor-reactive T cells.5, 6 Unfortunately, attempts at exploiting the autologous immune system for the treatment of ES were hampered by insufficient precursor cell frequencies and inadequate functionality of peripheral blood-derived tumor-reactive T cells.3, 7

Several lines of evidence suggest that the bone marrow (BM) compartment is critically involved in ES formation and progression: Although the cell of origin in this disease has long remained obscure, malignant transformation is now thought to occur in a multipotent mesenchymal precursor cell that, at least in osseous ES, likely resides in BM.8, 9 Importantly, the presence of microscopic marrow involvement defines a minor population of patients at a high risk of relapse1, 2 and submicroscopic BM disease was identified by molecular screening in many patients.10 These disseminated tumor cells may be the source of systemic relapse even in patients with apparently localized tumors.11, 12 From an immunological point of view, colocalization of tumor cells with BM-resident T cells may contribute to the initiation of tumor-specific immune responses. Indeed, besides its established role in hematopoesis, BM has recently emerged as a critical organ for the induction and maintenance of antitumor immune responses. Important T-cell events occurring in the BM include both the efficient priming of naïve T cells via resident dendritic cells,13, 14 and the recruitment and reactivation of antigen-experienced memory T cells.15–17 In a murine tumor model, BM dormant tumor cells provided persistent antigenic stimulation for maintaining vaccine-induced immune control.18 In patients with either leukemia or solid cancer, functional antitumor T cells were found to be enriched in the marrow compartment,19–22 and these BM T cells correlated with the maintenance of remission.23 Potential players of effective antitumor immune responses include both effector memory T cells (T_EM), which provide immediate protection, and central memory T cells (T_CM), that recirculate in secondary lymphoid organs and maintain a high potential for re-expansion. Besides T cells with classical αβ T-cell receptors, a surveillance function against malignant cells has been attributed to cells with innate cytolytic function and MHC-independent target recognition, including NK cells, γδ T cells and NKT cells.24–26

Opposed to these effector cell populations, BM has been shown to provide a significant reservoir for CD4+CD25^hiFoxP3+ T cells with suppressive function (T_reg cells)27 that have a key function in the control of peripheral tolerance to self-antigens and alloantigens28 and have emerged as a major impediment to functional immune responses to malignancies. CD4+CD25^hiFoxP3+ T_reg cells have been found at increased densities within the tumor microenvironment of patients with various malignancies29–32 and are thought to contribute to tumor immune escape. Not only T_reg cells with a memory phenotype, but also naïve CD45RA+CCR7+ T_reg cells were detected at increased frequencies in cancer patients.33 Most of these studies have focused on the relative frequencies of CD4+CD25+FoxP3+ T_reg cells among tumor-infiltrating T cells. Recently, the intratumoral balance between cytotoxic and regulatory T cells was found to be even more predictive of survival than the single T-cell subtype in some cancers.34, 35

The significance of T_reg cells and effector T cells localized within the BM microenvironment of Ewing tumor patients has not been explored yet. In this study, we quantified the proportions of naïve and memory CD4+CD25^hiFoxP3+ T_reg cells as well as various cytotoxic immune cell subsets, including γδ T cells, NK cells, NKT cells and CD8+ effector/memory T-cell subpopulations in the BM of patients with primary and relapsed ES. On the basis of the hypothesis that BM-resident immune cells may contribute to a permissive microenvironment for the growth of ES, we investigated specific correlations of these individual T-cell subsets to known risk factors of the disease.

Material and methods

Patients and treatment

On the basis of the availability of pretherapeutic BM aspirates, 40 patients with osseous ES, diagnosed between 1998 and 2009, were recruited from University Children's Hospital Muenster in a retrospective manner. All patients were included into the multicenter EICESS 92 or E.U.R.O Ewing 99 trials, which have been approved by the Institutional Review Board, with central documentation of initial and follow-up data in the study headquarter in Muenster, Germany. Five additional patients were recruited at relapse of their disease. In 1 patient, BM samples were available both at primary diagnosis and relapse. All BM aspirates were taken from tumor-free areas of the iliac crest before administration of primary or relapse anticancer treatment, respectively. In 5 patients, peripheral blood samples were obtained simultaneously. Informed consent was obtained from all patients in accordance with the Declaration of Helsinki. For comparative studies, BM samples were obtained from 3 additional pediatric patients with benign, noninflammatory disorders as part of a routine diagnostic procedure. These included a 5-year-old boy with idiopathic thrombocytopenic purpura, a 17 months old girl with autoimmune granulocytopenia, and a 3-year-old girl with mediastinal hemangioma.

Assessment and definition of patient and treatment-related variables

The following variables were evaluated for their distribution in the patient cohort and for possible correlations with outcome: age, gender, tumor site (extremity vs. axis), primary metastases and tumor response to chemotherapy according to Salzer-Kuntschik et al.36 A good response to chemotherapy was defined as <10% viable tumor cells (response grades 1–3).

Cell preparation

Mononuclear cells (MCs) were isolated from freshly aspirated BM by density gradient centrifugation, resuspended in RPMI culture medium and frozen in liquid nitrogen.

Antibodies and flow cytometry analysis

For immunophenotyping of T cells, cells were stained with fluorescein-conjugated monoclonal antibodies (all from BD Pharmingen, Heidelberg, Germany). Antibodies conjugated to fluorescein isothiocyanate (FITC), phycoerythrin (PE), phycoerythrin-cyanin 7 (PE-Cy7), peridinin chlorophyll protein (PerCP), allophycocyan (APC) or allophycocyan-Cy7 (APC-Cy7) were used in the following combinations: Panel 1, CD3-PerCP/CD56-PE-Cy7/TCRαβ-FITC/TCRγδ-PE/CD86-APC, HLA-DR-APC-Cy7; Panel 2, CD4-PerCP/cyFoxP3(clone PCH101)-AlexaAPC/CD25-APC-Cy7/CD 45RA-FITC/CD45RO-PE, CCR7-PE-Cy7; Panel 3, CD3-PerCP/CD4-FITC/CD8-APC-Cy7/CD45RA-PE, CCR7-APC, CD 45RO-PE-Cy7. For intracellular FoxP3 staining, cells were fixed with fixation/permeabilization buffer (eBioscience, San Diego, CA) for 30 min, washed twice in 1× permeabilization buffer (eBioscience) and resuspended in 100 μl 1× permeabilization buffer containing 2% normal rat serum for blocking. After 15 min of incubation at 4°C, fluorochrome-conjugated anti-human FoxP3 antibody (eBioscience) was added for additional 30 min of incubation, followed by 2 washing steps in permeabilization buffer before analysis. For each sample, lymphocytes were acquired by forward and side scatter plotting on a BD FACS Canto flow cytometer (Becton Dickinson), and a gate was set on CD3+ cells within the lymphocyte gate. The percentage of FoxP3-positive cells was determined after gating on CD4+ cells. Samples containing <15,000 CD3+ cells and samples containing <5,000 CD4+ cells were excluded from analysis. FACS DIVA software (Becton Dickinson) was used for analysis. Frequencies of CD4+CD25^hi T cells in BM are shown as a fraction of CD4+ T cells.

Statistical analyses

Comparison between groups was performed using the Student t-test whereas the equality of variance was assessed using Levene's test. A p-value <0.05 was defined as statistically significant. All p-values reported are two-sided. To assess the degree of linear relationship between 2 variables, the Pearson Product Moment Correlation was used. The distributions of the time-to-event variables were estimated using the Kaplan–Meier method, and comparisons were based on the log-rank test with a significance level of 0.05. All statistical analyses were performed using the SPSS statistical software package (SPSS 16.0 for Windows; SPSS, Chicago, IL).

Results

Patient characteristics

The characteristics of the 40 patients with ES at primary diagnosis are shown in Table I. Our cohort of patients is well representative of patients with this disease, as analyzed within large collaborative study groups.1, 2 In 1 of these patients and in 5 additional patients, BM samples were obtained at relapse before reinitiation of chemotherapy. Relapsed patients were 8–25 years of age (median of 14 years). Four patients had systemic relapses, 1 had a local relapse. Relapses had occurred 15–22 months (median of 19 months) after primary diagnosis.

Table I. Characteristics of Ewing Sarcoma Patients

	Primary diagnosis (n = 40)	Relapse (n = 5)
Age, median age (range) (years)	14 (3–20)	14 (8–25)
Gender
Male	24	4
Female	16	1
Male-to-female ratio	1.5	4
Tumor site
Axial	19	2
Extremity	21	3
Primary metastases
Absent	29
Pulmonary only	6
Bone and/or BM	2
Pulmonary and bone and/or BM	3
Tumor response of primary tumor
Good (Grades 1–3)	29	4
Poor (Grades 4–6)	4	1
Median time of relapse after primary diagnosis		19 (15–22 months)
Type of relapse
Localized		1
Systemic		4

Relative proportions of cytolytic effector cell subpopulations in BM with regard to patient related variables

In patients at primary diagnosis, the MCs within the lymphocyte gate contained 7.5 ± 5.9% (1.6–23.5%) CD3−CD56+ NK cells. 56.6 ± 13.7% (32.1–78.9%) were CD3+ T cells (CD3+CD56−), with a proportion of 6.2 ± 5.2% (1.7–26.7%) γδTCR+ T cells and 1.5 ± 1.0% (0.0–3.9%) CD3+CD56+ NKT cells (Fig. 1a). Among αβTCR+ T cells, the CD4+/CD8+ T-cell ratio was 1.14 ± 0.39, consistent with published data on BM T-cell subsets in healthy adult donors.17, 19, 37 Although the patients at relapse of the disease had comparable proportions of NK cells, NKT cells and γδ T cells to patients at primary diagnosis, an inverted ratio of CD4+ to CD8+ T cells was found (Fig. 1b). This observation was confirmed in direct comparison of paired BM samples obtained at both diagnosis and disease relapse in 1 individual patient (CD4/CD8 ratio of 1.01 vs. 0.28). Costaining with CD45RA and CCR7 allowed quantitative assessment of the relative proportions of CD8+ T-cell subpopulations with distinct antigen experience and memory function (Fig. 2a). In the cohort of patients with primary ES, the predominant subset had a naïve T-cell phenotype (T_N; CD45RA+CCR7+), representing 56.4 ± 15.3% of CD8+ T cells, whereas 16.8 ± 9.1% were T_EM (CD45RA−CCR7−), 8.3 ± 7.5% were T_CM (CD45RA−CCR7+) and 16.7 ± 9.5% had a differentiated effector cell phenotype (T_EMRA; CD45RA+CCR7−) (Fig. 2b). The relative proportions of naive and effector/memory T-cell subpopulations in the 5 relapsed patients were not significantly different from those at primary diagnosis of the disease, and except for an increased number of T_EMRA cells in the patient with mediastinal hemangioma, effector subpopulations in the 3 children with nonmalignant disorders were within the range of those found in Ewing tumor patients.

Figure 1. (a) Relative proportions of NK cells (CD3−CD56+), γδ T cells (CD3+γδTCR+), NKT cells (CD3+CD56+) and αβ T cells (CD3+ αβTCR+) within the lymphocyte gate (NK cells) or as a fraction of CD3+ T cells (all others) in 40 patients with primary ES (filled circles), 3 patients with relapsed ES (open circles) and 3 children with nonmalignant disorders (filled triangles). The lines indicate the mean proportions of the respective cell populations. In 3 of the 6 patients with relapsed ES and 5 of the 40 patients with primary ES, sufficient material was not available for these analyses. (b) Quantification of CD4+ and CD8+ T cells as fractions of CD3+ T cells in 40 patients with a primary diagnosis of ES (grey boxes) and 6 patients at relapse of disease (white boxes). Boxes indicate median, 25th and 75th percentiles. The bars indicate 10th and 90th percentiles, whereas outliers are shown as filled circles.

Figure 2. (a) Classification of CD8+ T cells as naïve (T_N), central memory (T_CM), effector memory (T_EM) and CD45RA+ effector memory (T_EMRA) cells by staining with CD45RA− and CCR7-specific antibody; representative example of many. (b) Relative proportions of T_N, T_CM, T_EM and T_EMRA cells as a fraction of CD8+ T cells in 40 patients with primary ES (filled circles), 5 patients with relapsed ES (open circles) and 3 children with nonmalignant disorders (filled triangles). The lines indicate the mean proportions of the respective cell populations. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

To identify potential associations of BM effector cell subpopulations with known risk factors of the disease, we categorized our patient cohort according to previously defined prognostic variables,1, 2, 38 including age (< or >14 years), gender, tumor site (extremityversus axis), the presence or absence of primary metastases, and histological tumor response to neoadjuvant chemotherapy according to Salzer-Kuntschik et al.36 The initial tumor volume was not included in these analyses, because alternative documentation within the 2 applied treatment protocols did not allow any valuable comparisons. A comprehensive analysis of the proportions of each of the individual effector cell subpopulations (CD4+ cells, CD8+ cells, CD8+ T_N cells, CD8+ T_EM cells, CD8+ T_CM cells, CD8+ T_EMRA cells, NK cells, NKT cells and γδ T cells) and these patient related variables failed to reveal any significant associations. With regard to treatment outcome, none of the parameters assessed at primary diagnosis of the disease correlated with the risk of relapse within either 3 years (n = 4) or 5 years of follow-up (n = 9). Thus, the relative proportions of individual cytolytic immune effector cell subpopulations in the BM of newly diagnosed ES patients neither correlate with prognostic variables characterizing the disease nor with outcome.

CD4+CD25^hiFoxP3+ T_reg cells and patient related variables

T cells with natural regulatory function have a characteristic phenotype of CD4 and CD25^hi coexpression with the transcription factor FoxP3. By intracellular costaining of CD4+CD25^hi T cells with a FoxP3-specific antibody, a population of T cells with T_reg phenotype (referred to as “T_reg cells”) was identified in patient BM (Fig. 3a). Among the CD4+ T cells of patients with primary tumors, 3.8 ± 1.5% cells were FoxP3+ T_reg cells (Fig. 3b). 38.0 ± 17.3% of these cells coexpressed CD45RA, representing the naive subtype of T_reg cells.33 The relative proportions of BM T_reg cells and of naïve T_reg cells did not correlate with age, gender, tumor localization or tumor response to neoadjuvant chemotherapy. Furthermore, no differences were found in relapsed patients to those at primary diagnosis of the disease. However, a statistically significant discrepancy was found in patients with and without metastatic disease at diagnosis: Compared with patients with localized tumors (n = 29), patients with metastatic disease (n = 11) had significantly increased frequencies of T_reg cells (5.0 ± 1.6% vs. 3.3 ± 1.3%, p = 0.01; Fig. 3c). Among the T_reg cells in these patient subgroups, the relative proportions of T_reg cells with naïve or memory phenotype, respectively, did not significantly vary. Although there was a trend toward higher T_reg cell numbers in patients with bone and/or BM dissemination (5.3 ± 1.3%, n = 5) compared with those with isolated lung metastases (4.6 ± 2.0%, n = 6), the difference between these small groups did not reach statistical significance (p = 0.48).

Figure 3. (a) Gating strategy used to quantitate CD4+CD25^hi FoxP3+ T_reg cells as well as CD45RA+ and RA-negative T_reg cell subpopulations (one representative example). (b) Relative proportions of T_reg cells as a fraction of CD4+ T cells, and relative proportions of CD45RA+ and RA-negative T_reg cells in 20 (right panel) or 40 (left panel) patients with primary ES (filled circles), 6patients with relapsed ES (open circles) and 3 children with nonmalignant disorders (filled triangles). The lines indicate the mean proportions of the respective cell population. (c) Proportions of CD4+CD25^hiFox P3+ T_reg cells as a fraction of CD4+ T cells in Ewing sarcoma patients without metastases (n = 29, grey box) and with primary metastatic disease (n = 11, white box). Boxes indicate median, 25th and 75th percentiles. The bars indicate 10th and 90th percentiles, whereas outliers are shown as filled circles. (d) Kaplan–Meier analysis of event-free survival (EFS) in patients with T_reg cell proportions of <3.65% and >3.65% (n = 16 in each group). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Compared with the relative frequencies of T_reg cells alone, the relative ratios of T_reg cells and T cells with cytolytic effector or memory properties may provide a more accurate parameter for the immunosuppressive milieu within a specific compartment.34, 35 Therefore, we calculated the relative ratios of total CD8+ T cells to CD4+CD25^hiFoxP3+ T_reg cells (CD8/T_reg), and of CD8+CD45RA−CCR7− T_EM cells to T_reg cells (T_EM/T_reg), respectively, for each BM sample, and analyzed these ratios with regard to patient related variables. Patients with localized disease had significantly higher T_CM/T_reg ratios (3.4 ± 3.9) than patients with metastatic disease (1.1 ± 0.7, p = 0.005). Furthermore, there was a trend toward an association between an increased CD8/T_reg cell ratio (39.1 ± 32.0 vs. 22.5 ± 13.4, p = 0.106) and the absence of metastases.

In 5 patients in which both BM and peripheral blood samples were available at primary diagnosis, we performed a comparative analysis of FoxP3+ T-cell numbers in the 2 compartments. The T_reg cell content in peripheral blood did not correlate with that in BM in these 5 patients (Pearson coefficient 0.22, p = 0.72) (Table II). Thus, it is very unlikely that our finding would have been reproduced by using peripheral blood.

Table II. Comparative Analysis of FoxP3+ T-Cell Numbers From Bone Marrow (BM) and Peripheral Blood (pB) in 5 Patients With Ewing Tumors

Patient	% FoxP3+ T cells
	BM	pB
A	4.23	3.67
B	3.00	6.21
C	4.35	5.79
D	3.10	2.96
E	3.09	2.92

Correlation of effector cell subpopulations and CD4+CD25^hiFoxP3+ T_reg cells with overall and relapse free survival

Within a follow-up of 5 years in 32 evaluable patients with primary diagnosis of ES, events had occurred in 9 patients, defined as either progressive disease, new metastases or a local or systemic relapse. Six of the patients had died of the disease. Seven of the relapses occurred in patients without initial metastases (n = 23), whereas 2 of the 6 patients with bone and/or BM metastases relapsed. None of the 3 evaluable patients with lung metastases had relapsed within the 5 year follow-up. A Kaplan–Meier analysis failed to reveal any associations of the relative proportions of innate immune cells (NK cells, γδ T cells or NKT cells) or effector/memory T cells (T_CM, T_EM) with the event-free or overall survival of the patients. The difference between the relative proportions of BM T_reg cells in patients with and without primary metastases was not reflected in the event-free and overall survival of the patients (p = 0.314) (Fig. 3d). Furthermore, no significant associations were found between the ratios of total CD8+ T cells to CD4+CD25^hiFoxP3+ T_reg cells, CD8+CD45RA−CCR7− T_CM cells and CD8+CD45RA−CCR7− T_EM cells to T_reg cells, respectively, and event-free survival.

Discussion

As a first step toward understanding the role of BM in the immune control and/or escape of ES, we performed an analysis of the immune effector subpopulations in pretherapeutic marrow of a cohort of pediatric and young adult ES patients. While a comparison with normal, age-matched controls was limited by the lack of available BM samples from healthy children and adolescents, the investigated patient cohort allowed an intrinsic statistical correlation of individual T-cell subpopulations with known risk factors of the disease. Even though lymphocyte subpopulations have been shown to vary with age,39 no significant associations were found between any of the lymphocyte or T-cell subsets and patient age below or above 14 years, arguing against a major contribution of age-dependent variables to the observed differences.

The major finding of our study was a significantly increased prevalence of BM T_reg cells in patients with metastasized ES compared with localized tumors. Natural T_reg cells, characterized by coexpression of CD4, high levels of CD25, and the transcription factor FoxP3, can inhibit tumor-specific CD8+ and CD4+ T-cell effector functions through incompletely understood mechanisms.28 First evidence for a role of T_reg cells in tumor immunopathology was based on the demonstration that T_reg cell depletion in mouse models of cancer enhanced antigen-specific immunity and tumor rejection.40 In cancer patients, T_reg cells were found over-represented in the peripheral blood and/or tumor environment of various tumor types,29–32 and were capable of suppressing protective antitumor T-cell responses.41–43 Thus, promoting the expansion of CD4+CD25+ T_reg cells appears to be a mechanism by which tumor cells restrain the function of T cells and escape autologous immune responses. The association of higher proportions of BM T_reg cells in our cohort of ES patients with primary metastases compared with ES patients with localized tumors may suggest a role of T_reg cells in tumor dissemination. In an unrelated malignancy, acute lymphoblastic leukemia, we found similar proportions of T_reg cells among the residual BM lymphocytes of 25 pediatric patients to patients with localized ES (data not shown), further supporting a specific role of T_reg cells in metastatic ES.

We considered the BM microenvironment as a site for ongoing immune escape in ES for the following reasons: Even in patients with apparently localized disease, ESs can not be cured by local therapy alone, and disseminated relapses inevitably occur after inadequate systemic chemotherapy.44 Thus, small numbers of circulating tumor cells must be present at diagnosis even in patients with early and localized stages of the disease. Anti-Ewing tumor immune reactions are thus likely to occur beyond the tumor microenvironment in secondary lymphoid organs, including lymph nodes and BM, in all patients, regardless of the microscopic presence of BM metastases. Therefore, it is plausible that colocalization of T cells with microscopic or micrometastatic ES cells in BM may allow for the priming and subsequent expansion of tumor antigen-specific memory T cells. Because of the metastatic properties of BM-resident tumor cell subsets and the preferential location of long-term persisting tumor cells in this compartment,45 these T cells may have special relevance for the control of disseminated minimal residual disease. Indeed, tumor-reactive memory T cells have been found to be enriched in the BM of patients with various cancers, including both hematologic malignancies23, 46 and solid tumors not generally homing to BM,19–21, 47 and human marrow T cells were capable of inducing regressions of autologous tumor xenografts in mice.48 In pediatric patients with acute myeloid leukemia, the presence of antileukemic precursor T cells in autologous BM correlated with long-term remission.23 Furthermore, BM-derived effector T cells specific for preneoplastic plasma cells were detected exclusively in patients with preclinical stages of myeloma,49 suggesting a protective role against progression to cancer. However, the development of clinical metastases in ES patients at diagnosis or at systemic relapse argues against sustained immune control of the disease. Therefore, our finding of an increased proportion of T_reg cells in the BM of patients with primary metastatic disease supports a concept in which local accumulation of T_reg cells in BM impedes protective cellular immune events. Survival of even small numbers of tumor cells in this compartment may contribute to the manifestation of clinically detectable metastases. A unique role of the BM compartment is further supported by our observation that BM T_reg cell proportions did not correlate with those circulating in the peripheral blood.

The origin of the increased proportion of T_reg cells in the tumor microenvironment of malignancies is still a matter of investigation. They may either be specifically recruited from the circulating T_reg population,41 or their phenotype may be induced from resident conventional T cells.50 Furthermore, our data do not allow to answer the question whether the development of overt metastases occurs as a consequence of a permissive immune environment, or whether tumor cell dissemination induces these alterations in the BM T-cell compartment. One means of approaching this issue would be to correlate levels of submicroscopic BM disease quantified by sensitive molecular techniques with the frequency of BM T_reg cells.

Cytotoxic effector T cells coexisting with T_reg cells in the BM micromilieu may shift the balance toward antitumor immune control. Indeed, ESs have been found to be susceptible targets for granzyme/perforin-mediated lysis51 and thus provide adequate targets for cytolytic T cells. Recent attention in tumor immunology has been drawn toward the CD8+ memory T-cell compartment, which appears to be critically important in preventing tumor progression and metastasis.18 Specifically, among adoptively transferred T-cell populations, cells with a central memory phenotype (T_CM) have been found to mediate long-term immune control of cancer in vivo.52 Studies in mice have shown a predominance of activated CD8+ T_CM cells in BM, suggesting a selective tropism for this compartment.53 In humans, by contrast, T_EM cells were found to account for the majority of memory CD8+ BM T cells in several independent investigations.16, 17, 37 Compared with published data obtained in a total of 29 donors with nonmalignant diseases in 2 independent studies,19, 37 the relative proportions of BM T_EM and T_CM subsets of CD8+ T cells in our ES patient cohort were not altered. Furthermore, no correlations of either CD8+ T-cell proportions or T_EM and T_CM subpopulations with any of the known risk factors of ES or with survival were found in our patient cohort, arguing against a tumor-induced increase of either of these individual effector cell subpopulations. Encouraged by recent publications that have emphasized the important role of an unfavorable balance between effector CD8+ T cells and T_reg cells in tumor growth, rather than total cell numbers alone,34, 35 we have further analyzed CD8/T_reg, T_EM/T_reg and T_CM/T_reg ratios, and indeed have found trends for correlations of low ratios with metastatic disease.

Regarding CD4+ T cells, our major finding besides variable T_reg proportions was an inverted ratio of CD4/CD8 T cells in patients at relapse of the disease. Based on published data, we attribute this finding to the preceding chemotherapy these patients had undergone. Indeed, various investigators report a more dramatic and sustained effect of intensive chemotherapy on the CD4+ than CD8+ T-cell subpopulations in both children and adults with solid tumors and leukemia, resulting in statistically significant decreases of the CD4/CD8 T-cell ratio and delayed recoveries of over 12 months in many patients.54–57

Besides antigen-specific T cells, innate immune cells can eliminate cancer cells and are thought to be involved to the host response against tumors.58 The precise role of marrow NK cells, NKT cells or γδ T cells in this regard is as yet unknown. In 1 study, γδ T cells were found at increased relative frequencies within the rare nonmalignant lymphocyte population in the BM of 6 children with newly diagnosed acute lymphoblastic leukemia.59 We found a wide variation of innate immune effector cells among the mononuclear BM cells of ES patients, and failed to demonstrate any correlation with prognostic parameters. However, a recent publication revealed reduced NK cell cytotoxic function in ES patients compared with age-matched controls, whereas total numbers of NK cells in peripheral blood were normal.60 On the other hand, ES cell lines were shown to express NK cell receptor ligands and were susceptible to NK cell lysis. Because T_reg cells can profoundly inhibit NK-cell effector functions,61 the increased numbers of BM T_reg cells reported here may thus contribute to impaired NK cell function and escape of ES from NK-cell mediated lysis in vivo.

Unexpectedly, the association of increased T_reg cell frequencies with primary metastases was not reflected in an increased rate of metastatic relapse and corresponding decreased overall or relapse-free survival. Likely explanations are both the small numbers of high-risk patients and the risk-adjusted intensification of treatment according to the EICESS 92 and E.U.R.O Ewing 99 trials. Only a minority of 6 patients in our cohort had high-risk metastatic disease, i.e., combined or bone/BM metastases, and a superior outcome to that reported within large multi-institutional trials was found for both subgroups of patients with isolated lung metastases and systemic disease in our cohort.1, 2, 38 Recent treatment advances have reduced systemic relapses at least in patients with isolated lung metastases, who now approach the survival rates of patients with nonmetastatic disease.2 Thus, though BM T_reg cells appear to be involved in the development of primary metastatic disease, the highly efficient polychemotherapy may have offset a potential impact of increased T_reg cells on the subsequent course of the disease and on survival.

The major limitation of our study is its restriction to immune cell phenotypes in the absence of functional studies regarding the potential reactivity of individual T-cell clones to tumor-associated antigens. The quantification of individual effector cell subsets provides only indirect clues of the nature of the immune cell milieu. For example, in a recent study, T_CM cells with specificity for CMV were found to be enriched in BM compared with peripheral blood, whereas the relative proportions of T_CM cells in both compartments were not different.16 Thus, only functional experiments based on T cells with defined specificities for Ewing tumor-associated antigens can reveal whether relevant antitumor T-cell subpopulations are present within the BM of the patients. Unfortunately, an analysis of tumor-reactive T cells is complicated by the lack of known ES-associated antigens. Even though several antigens have now been identified to be overexpressed in ES, T cells directed against these antigens can not be expected to represent the cellular immune response in individual patients. The use of autologous ES cells as targets of functional experiments may be a means of overcoming this limitation, and we are currently prospectively collecting samples for these functional studies.

In summary, we demonstrate for the first time an association of the T_reg cell content in pretherapeutic marrow samples of patients with ES with the presence of metastatic disease. Whether and to what extent these cells are indeed involved in creating an immunosuppressive environment for the specific escape of ES cells from immune control will be subject of functional studies. The development of improved strategies for the immune targeting of ES may rely on efficient in vivo depletion of T_reg cells, as suggested by recent experience with other types of cancer.62, 63 In this regard, the profound lymphodepletion associated with high dose chemotherapy and subsequent autologous stem cell transplantation performed in high-risk ES patients may create a suitable venue for the implementation of immune-based treatment strategies.

Acknowledgements

This work was supported by a grant from the Innovative Medizinische Forschung (IMF) program of the University of Muenster Faculty of Medicine (to C.R. and C.C.).