Purpose: Tumor microenvironment has a strong effect on the survival of follicular lymphoma (FL) patients. The aim of this study was to determine what are the signaling pathways that mediate the cross-talk between lymphoma cells and tumor-infiltrating inflammatory cells and contribute to the clinical outcome of FL patients.

Experimental Design: Gene expression profiling and pathway impact analyses were done from pretreatment lymphoma tissue of 24 patients. The findings were validated immunohistochemically in an independent cohort of 81 patients. All patients were treated with the combination of rituximab and cyclophoshamide-doxorubicin-vincristine-prednisone chemotherapy. In addition, microarray was used to screen the genes differentially expressed between control and rituximab-stimulated B-cell lymphoma cells in culture.

Results: Among the transcripts differentially expressed in the FL tissues between the patients with favorable or adverse outcomes, an overrepresentation of genes associated with the signal transducers and activators of transcription (STAT)5a pathway was observed. In a validation set, a better progression-free survival was observed among the patients with high STAT5a protein expression. In the FL tissue, STAT5a positivity was barely detectable in the neoplastic B cells, but a subpopulation of follicular dendritic cells and T lymphocytes showed prominent STAT5a expression. Rituximab was found to induce the expression of STAT5a-associated interleukin-15 in B-lymphoma cells in culture, thereby providing a possible explanation for the cross-talk between rituximab-stimulated FL cells and their microenvironment.

Conclusion: The findings suggest that STAT5a activity in immunologically active nonmalignant cells acts as molecular predictor for rituximab and cyclophoshamide-doxorubicin-vincristine-prednisone–treated FL patients. Clin Cancer Res; 16(9); 2615–23. ©2010 AACR.

This article is featured in Highlights of This Issue, p. 2483

Translational Relevance

Follicular lymphoma is a rarely curable disease that exhibits clinical heterogeneity. In the prerituximab era, the heterogeneous clinical course has been particularly associated with molecular signatures reflecting putative interactions between tumor cells and infiltrating immune cells. In this study, we combine gene expression profiling and pathway impact analyses with immunohistochemistry and cellular studies to identify signaling pathways, which mediate cross-talk between lymphoma cells and tumor-infiltrating inflammatory cells and contribute to the outcome of FL patients treated with immunochemotherapy. The results provide evidence that the signals in the immunologically active neighboring nonmalignant cells act as molecular predictors also in the postrituximab era of lymphoma therapies.

Follicular lymphoma (FL) represents a challenging malignant disease. It is indolent and chemosensitive but considered rarely curable with conventional chemotherapy due to its propensity to relapse. FL maintains sensitivity to different chemotherapeutic agents for several years but ultimately becomes resistant or transforms into a high-grade lymphoma. The median survival has been in the range of 8 to 10 years (1). Recently, however, a significant improvement of the outcome of FL patients has been obtained by combining a monoclonal anti-CD20 antibody, rituximab, with the induction chemotherapy (24), or by prolonging the remission with rituximab maintenance therapy (57). Despite these advances, responses to treatments are heterogeneous and the treatment outcome is often unpredictable. These aspects raise the need to identify more accurately the patients who will benefit from immunochemotherapy.

Recent gene profiling studies have provided valuable biological information to explain the clinical behavior of FLs and have also led to the discovery of novel molecular predictors for survival (810). The heterogeneous clinical course has been particularly associated with molecular signatures reflecting putative interactions between tumor cells and infiltrating immune cells (8). Importantly, however, the prognostic significance of the tumor microenvironment seems to be highly dependent on a given therapy (11). For example, fludarabine and cyclophosphamide have direct suppressive effects on T-regulatory cells (12, 13) and can therefore affect FLs differently from the other chemotherapeutic agents. Conversely, the therapeutic benefit of CD20 antibody requires the presence of macrophages and their Fcγ-dependent interactions (14). This together with the data showing that the addition of rituximab to chemotherapy can reverse the prognostic effect of tumor-associated macrophage and mast cell contents in FL (15, 16) indicate that signaling between FL cells and tumor-infiltrating effector cells regulates the efficacy of rituximab. Thus, it would be important to understand how the cross-talk between tumor cells and their nonmalignant counterparts is mediated.

We have previously observed that gene expression by nonmalignant tumor cells has a prognostic effect also in rituximab and cyclophoshamide-doxorubicin-vincristine-prednisone (R-CHOP)–treated FL patients (10). The success of using gene expression profiling to predict patient outcome in response to a given therapy suggests that the same data can be used for a systems biology approach to identify signaling pathways or groups of genes having clinical importance in FL. In the current study, we used whole genome expression analysis, gene ontology associations, and immunohistologic validation experiments and identified signal transducers and activators of transcription (STAT)5a activity in the tumor microenvironment as a predictor of outcome in R-CHOP–treated FL patients.

Patients

This is a population-based retrospective analysis for FL patients treated with a combination of R-CHOP regimen. Initially, 24 FL patients were selected for the microarray group. The patients were eligible if they had received R-CHOP, if fresh frozen tissue was available for RNA isolations and microarrays, and if the sample had been taken before R-CHOP treatment. All patients received R-CHOP for the first time. Of these, 17 had primary disease and 7 had relapsed disease.

The verification group consisted of 81 FL patients, who all received R-CHOP regimen as a first-line treatment. The same inclusion criteria were used for the validation group except that instead of fresh frozen tissue, paraffin-embedded lymphoma tissue had to be available for immunohistochemical stainings, All patients had a clinical indication for treatment and were sequentially treated at the Helsinki University Central Hospital during 1999 to 2005. The baseline characteristics of the patients are listed in Table 1. The protocol and sampling were approved by the Ethical Committee at the hospital, the Institutional Review Board, and the Finnish National Authority for Medicolegal Affairs.

Table 1.

Characteristics of the microarray and validation groups

Microarray (n = 24)Validation (n = 81)
Median age (y) 53 (38-77) 57 (28-82) 
Male/female (%) 38/62 49/51 
FLIPI 0-2 (%) 54 61 
FLIPI 3-5 (%) 46 36 
Primary disease (%) 71 100 
    Response rates (%) 
        Complete 65 63 
        Partial 23 35 
        Stable 12 
Relapsed disease (%) 29  
    Response rates (%) 
        Complete 86  
        Partial 14  
        Stable  
Microarray (n = 24)Validation (n = 81)
Median age (y) 53 (38-77) 57 (28-82) 
Male/female (%) 38/62 49/51 
FLIPI 0-2 (%) 54 61 
FLIPI 3-5 (%) 46 36 
Primary disease (%) 71 100 
    Response rates (%) 
        Complete 65 63 
        Partial 23 35 
        Stable 12 
Relapsed disease (%) 29  
    Response rates (%) 
        Complete 86  
        Partial 14  
        Stable  

Gene expression profiling and data analyses

24 FL patients were classified into groups of favorable (continuous remission, n = 11) or adverse (relapsed disease, n = 13) outcomes and gene expression profiles of lymphoma tissue were measured using Agilent Human IA oligonucleotide microarrays (10). Raw expression microarray data are available at the ArrayExpress archive (http://www.ebi.ac.uk/microarray-as/ae/; ID: E-MEXP-2305). Samples were divided to groups X and Y, and then a t test was used to identify differentially expressed genes. Genes with the P value of <0.01 (404 genes) were considered as significant. As a control, a t test was done to find differentially expressed genes between the nontreated and relapsed cases. Gene Ontology analysis was done using the GOstat analysis tool (17). Gene Ontology term overrepresentation was calculated with the Fisher's exact test. The signaling pathways that are significantly influenced in these experimental conditions were identified with the OntoTools pathway impact analysis (18).

To identify rituximab-induced genes, four established human B-cell lymphoma cell lines, HF-1 (19), Granta-519 (Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (DSMZ)), OCILy-3, and SuDHL-4 (a gift from Jose A. Martinez-Climent, University of Navarra, Pamplona, Spain) were used. Of these, HF-1 and SuDHL-4 represent germinal center/follicular derived lymphomas, whereas Granta-519 and OciLy-3 show characteristics of mantle cell and activated B-cell type lymphomas, respectively. All cell lines were incubated in a humidified 5% CO2 atmosphere at 37°C. HF-1 and SuDHL-4 cells were cultured in RPMI and Granta-519 in DMEM supplemented with 10% FCS, 2 mmol/L glutamine, 100 U/mL penicillin, and 100 μg/mL of streptomycin. OCILy-3 were grown in Iscove's modified Dulbecco's medium supplemented with 20% human serum, 2 mmol/L glutamine, 55 μmol/L β-mercaptoethanol, 100 U/mL penicillin, and 100 μg/mL of streptomycin. The cells were treated with rituximab (10 μg/mL; Mabthera, Roche) for 3 h, pelleted by centrifugation, and the mRNA was extracted with the Nucleospin RNA II kit (Macherey-Nagel GmbH & Co.) according to the manufacturer's instructions. The gene expression profiles of control and rituximab-stimulated cells were analyzed using the two-channel Agilent Human 44K oligonucleotide microarray (Agilent Technologies). Signal values were calculated with Agilent Scanner. Gene expression analyses were executed with an open-source framework for data analysis Anduril (http://csbi.ltdk.helsinki.fi/anduril/index.html). A probe was removed from the analysis if the signal for either channel was saturated or if the signal values for channels belonged to the lowest 5% in its respective channel. Identical probes were combined with their median. Microarray data are available at ArrayExpress archive (http://www.ebi.ac.uk/microarray-as/ae/; ID: E-MEXP-2317).

Immunohistochemistry and scoring of STAT5a

Immunohistochemical stainings were done on formalin-fixed, paraffin-embedded tissue sections, all on individual slides. Small samples and biopsies were excluded due to overstaining. After deparaffination, heat-induced epitope retrieval (121°C, 3 min), and blocking of endogenous peroxidase, the slides were incubated with anti-STAT5a (1:500, Zymed Laboratories, Inc.), anti-CD3 (1:100, Novocastra Laboratories Ltd.), and anti-FoxP3 (1:150, Serotec Ltd.) antibodies at 4°C overnight. Immunohistochemistry was completed using the Vectastain ABC kit reagents (Vector Laboratories) according to the manufacturer's instructions. 3-Amino-9-ethylcarbazole was used as chromogen and counterstaining was done with hematoxylin. For CD4 staining, the antibody (1:150, clone 4B12, Novocastra Laboratories Ltd.) was incubated for 30 min and the detection was completed with Envision Advanced (Dako Cytomation).

Initially, overall immunohistochemical STAT5a positivity of the lymph node was evaluated by semiquantitative grading into low (occasional sporadic positive cells), intermediate (frequent but single positive cells), or high (abundant, partially confluent areas of positive cells) categories. The absolute number of STAT5a-positive cells was counted per five (two from follicular and three from interfollicular fields) high-power fields (hpf; ×630 magnification) with the Leica DM LB bright-field microscope (Leica Microsystems GmbH) and a camera attached to it (Olympus DP50, Studio Lite 1.0 Software). The representative areas with the most intense staining pattern were first selected with low magnification and these “hotspots” were then digitized, resulting in microscopic images with area sizes of 0.02 mm2. To more precisely determine the prognostic staining pattern and localization of STAT5a-positive cells, we later counted independently three more follicular areas per sample and analyzed them separately. Evaluation of FoxP3 was done similarly, except that images from six hpfs were counted: three from interfollicular and three from follicular areas. CD3+ cells were scored semiquantitatively as previously described (16). All scorings were done blinded.

Proliferation assays

In proliferation assays, cells were plated at a density of 2 × 105 cells/mL in 96-well plates and exposed to rituximab and rHu interleukins (IL) 2, 4, 6, and 7 (10 ng/mL; Promokine, PromoCell GmbH) for indicated periods of time. Cell proliferation was measured with the WST-1 reagent (Roche Diagnostics GmbH) according to the manufacturer's instructions.

Statistical analyses

The χ2 test and Mann-Whitney U test were used to assess the significance of the differences in the frequency of prognostic factors. The strength of associations of different factors with continuous variables was tested with Spearman rank correlation. Progression-free survival (PFS) was measured from the start of induction therapy with R-CHOP until the time of disease progression or the end of the observation period in patients without progressive disease. Overall survival was measured from the start of induction therapy until the last follow-up or death from any cause. Both univariate and multivariate analysis were done according to the Cox proportional hazards regression model. Survival rates were estimated using the Kaplan-Meier method and the differences between the subgroups were compared with the log-rank test. The statistical analyses were done with SPSS 14.0 for Windows (SPSS, Inc.).

STAT signature score

The STAT signature score for each patient was calculated by taking the sum of the mRNA expression values of the genes with positive correlation with PFS (STAT5a and Pim1) and dividing it with the sum of the mRNA expression values of the genes with inverse association with outcome (Socs1, ILsRα, IL4R, and IL7).

Patient characteristics

The baseline characteristics of the 24 patients, whose samples were subjected to microarray analysis, are shown in Table 1. All patients received R-CHOP for the first time. Seven patients had received previous therapy [chlorambucil (n = 5), alternating triple therapy (n = 1), and CHOP followed by radiotherapy (n = 1)]. The overall response rate (complete + partial responses) in this study population was 88% (21 of 24). For the subgroups of primary and previously treated patients, the overall response rates were 88% and 100%, respectively. After a median follow-up of 73 months (8-100 months), 11 patients were in remission (median follow-up, 72 months) and 13 had relapsed (median follow-up, 77 months). The median PFS for the whole cohort was 49 months. Based on the clinical follow-up data, the patients were divided into groups with favorable (continuous remission, n = 11) or adverse (relapsed disease, n = 13) outcomes. All previously treated patients had an adverse outcome.

In addition to the microarray group, 81 samples from primary FL patients treated with R-CHOP were available for validation with immunohistochemistry. The clinical characteristics of these patients are shown in Table 1. Overall response rate was 98%. After a median follow-up of 61 months (8-100 months), 68 of the FL patients were alive (84%) and 43 were in remission (53%). Median PFS for the whole cohort was 45 months. No significant differences in patient characteristics were observed between the microarray and immunohistochemical groups (data not shown).

Comparison of gene expression profiles between the groups with favorable and adverse outcomes identifies biologically meaningful signaling pathways

As a first step in exploring the signaling pathways and biological processes involved in the outcome of FL patients in response to R-CHOP, we used our previously established gene expression database from 24 pretreatment lymph node samples (10) with updated follow-up and identified 404 transcripts differentially expressed (P < 0.01) between the groups with favorable and adverse outcomes. In comparison, there were no differentially expressed genes between primary patients and the ones who had received previous treatments. A Gene Ontology analysis of the differentially expressed mRNAs showed a statistically significant overrepresentation of genes involved in the biological processes of lymphocytes, such as activators of lymphocyte differentiation (GO:0045621, P = 0.0126) and regulators of I-κB kinase/NF-κB cascade (GO:0043123, P = 0.028) and B-cell activation (GO:0050871 P = 0.037). According to the Onto Tools pathway impact analysis (18), phosphatidylinositol-3 kinase (PI3K) signaling system and Janus-activated kinase (JAK)-STAT pathways were among the top ones considered relevant for lymphoma biology. When other than primary FL patients were excluded from the analyses, PI3K and JAK-STAT pathways remained as major ones to differentiate the groups with favorable and adverse outcomes. Furthermore, according to t test, STAT5a was the best discriminator of all genes differentially expressed between the groups.

Association of STAT signaling with long-term survival

The association of JAK-STAT signaling with the R-CHOP response together with the literature reports on the importance of cytokine and STAT signaling in lymphoma biology (2024) motivated us to further analyze the prognostic effect of STATs in FL patients treated with immunochemotherapy. First, we tested in detail whether the levels of STAT pathway–related transcripts were different in patients who remained in remission compared with ones who relapsed. When the expression of STATs was compared between the subgroups, the only difference we found was a significantly lower STAT5a mRNA expression in the relapsed patients (Supplementary Fig. S1; Table 2). STAT5a expression correlated negatively with several regulators of STAT activity. These included STAT4 (rs = −0.524, P = 0.009), Socs1 (rs = −0.542, P = 0.006), JAK2 (rs = −0.649, P = 0.001), IL2Rα (rs = −0.751, P = 0.001), IL4R (rs = −0.578, P = 0.003), and IL7 (rs = −0.426, P = 0.038), whereas positive correlation with STAT3 (rs = 0.508, P = 0.011), Pim1 (rs = 0.401, P = 0.052), and IL4 (rs = 0.421, P = 0.041) was found. Of the regulators, IL7, IL4, IL2Rα, Socs1, and Pim1 were also differentially expressed between the patients in remission compared with the ones who had relapsed (Supplementary Fig. S1; Table 2). Expression of other STATs, JAK2, or Socs3 showed no correlation with outcome and there were no probes for STAT5b in the Agilent Human IA array.

Table 2.

Correlation of STAT pathway gene expression with prognosis as measured by the patient outcome predictor scores

GeneAverage expression on microarray*Significance
In remissionrelapsedt test (P)Cox regression (P)
STAT5a 1.4049 1.0978 0.001 0.001 
STAT2 1.5899 1.5348 0.689 0.479 
STAT3 2.3953 2.0036 0.179 0.133 
STAT4 1.7102 2.0893 0.190 0.075 
STAT6 4.6210 5.2775 0.333 0.227 
IL2 1.2758 1.7533 0.193 0.271 
IL2Rα 8.1235 24.3535 0.035 0.004 
IL4 2.0715 1.6903 0.047 0.083 
IL4R 5.3156 12.9004 0.095 0.012 
IL7 2.5326 4.4112 <0.001 <0.001 
IL7R 2.3177 2.3985 0.836 0.757 
Socs1 1.4386 1.9663 0.034 0.002 
Socs3 4.2836 4.5815 0.810 0.772 
Jak2 1.0714 1.2074 0.159 0.105 
Pim1 1.0952 0.7744 0.049 0.025 
STAT Score 0.4758 0.3544 0.001 0.002 
GeneAverage expression on microarray*Significance
In remissionrelapsedt test (P)Cox regression (P)
STAT5a 1.4049 1.0978 0.001 0.001 
STAT2 1.5899 1.5348 0.689 0.479 
STAT3 2.3953 2.0036 0.179 0.133 
STAT4 1.7102 2.0893 0.190 0.075 
STAT6 4.6210 5.2775 0.333 0.227 
IL2 1.2758 1.7533 0.193 0.271 
IL2Rα 8.1235 24.3535 0.035 0.004 
IL4 2.0715 1.6903 0.047 0.083 
IL4R 5.3156 12.9004 0.095 0.012 
IL7 2.5326 4.4112 <0.001 <0.001 
IL7R 2.3177 2.3985 0.836 0.757 
Socs1 1.4386 1.9663 0.034 0.002 
Socs3 4.2836 4.5815 0.810 0.772 
Jak2 1.0714 1.2074 0.159 0.105 
Pim1 1.0952 0.7744 0.049 0.025 
STAT Score 0.4758 0.3544 0.001 0.002 

*Values are on a log10 scale.

In univariate analyses, STAT5a, Socs1, Pim1, IL7, IL2Rα, and IL4R also had a prognostic effect on PFS (Table 2). Figure 1A shows that PFS for patients with high STAT5a levels was significantly better than for the ones with low STAT5a levels (P = 0.002). The outcome according to the Follicular Lymphoma International Prognostic Index (FLIPI) could also separate the high-risk patients from low- and intermediate-risk groups (P = 0.131, Fig. 1B).

Fig. 1.

The outcome of R-CHOP–treated FL patients according to STAT5 activity and FLIPI. A, PFS according to STAT5a low (<median) and high (>median) mRNA levels. B, patients according to FLIPI 0-2 (n = 14) and FLIPI 3-5 (n = 10) distinction. C, PFS according to STAT score levels (low versus high).

Fig. 1.

The outcome of R-CHOP–treated FL patients according to STAT5 activity and FLIPI. A, PFS according to STAT5a low (<median) and high (>median) mRNA levels. B, patients according to FLIPI 0-2 (n = 14) and FLIPI 3-5 (n = 10) distinction. C, PFS according to STAT score levels (low versus high).

Close modal

The data suggest that increases in some STAT pathway components affecting the pathway activity are correlated with favorable prognosis; however, decreases in others may be equally as important. To test whether the STAT activity affects the outcome in FL, we developed a STAT5 activity score that included changes in six components affecting both STAT activity and outcome. For this signature, we used the expression of STAT5a, Socs1, Pim1, IL2Ra, IL4R, and IL7. We constructed a signature score to report a high score when the STAT5 activity is high. This six-gene signature score had a significant association with the outcome (Table 2), with a higher score in patients in remission. The score also significantly correlated with PFS (Fig. 1C). When the analyses with the single STAT pathway components or STAT score were restricted to primary FL patients, no major changes in the results were observed (Supplementary Fig. S2).

STAT5a protein levels and their association with outcome

Because the clinical outcome of FL patients has been particularly associated with the molecular signatures reflecting interactions between lymphoma cells and infiltrating immune cells (8, 10), we next analyzed the correlation of STAT5a activity with the transcripts highly expressed in macrophages, mast cells, T-lymphocytes, and endothelial cells. Interestingly, a significant association of STAT5a activity score with the expression of CD4 (rs = 0.537, P = 0.007) was found. In addition, a negative association was seen between STAT5a and CD68 (rs = −0.443, P = 0.030), and CD31 (rs = −0.800, P < 0.001) transcripts. The data suggest that the high STAT5a activity reflects the biological characteristics of the nonmalignant immune cells and especially of CD4-positive T-lymphocytes, macrophages, and endothelial cells within the lymphoma.

To directly identify the cells with STAT5a activity and validate the gene expression data, we performed immunohistochemical stainings for STAT5a on paraffin-embedded lymphoma tissue from 81 FL patients. Overall, differences in immunohistochemically defined STAT5a protein expression were less apparent than the differences in mRNA levels. In the FL tissue, STAT5a expression localized both to perifollicular and follicular areas, but in the neoplastic follicles, STAT5a expression was commonly low (Fig. 2A). Generally, STAT5a positivity colocalized with CD4-positive T lymphocytes (Fig. 2B). Considering this together with the positive association of STAT5a activity with CD4 mRNA expression, it seems likely that the majority of STAT5a expression is derived from CD4-positive T lymphocytes. However, no correlation of STAT5a expression with CD3 or FoxP3 immunoreactivity was found. Instead, CD3 correlated positively with FoxP3 (rs = 0.468, P = 0.001). Additional analyses identified a prominent STAT5a positivity in a subpopulation of follicular dendritic cells (Fig. 2C). Taken together, the data suggest that STAT5a activity is primarily derived from nonmalignant inflammatory cells, including CD4-positive T lymphocytes and follicular dendritic cells.

Fig. 2.

Immunohistochemical analysis of FL tissue for STAT5a and CD4 expression. A, STAT5a immunoreactivity (red) was most abundant in perifollicular areas but was also involved in neoplastic follicles (high-expression pattern shown). Magnification, ×100. B, STAT5a expression strikingly colocalized with CD4+ T cells, as shown by the CD4 staining of a consecutive serial section of the corresponding neoplastic follicle. Magnification, ×100. In A and B, representative examples of neoplastic follicles are shown. C, STAT5a immunoreactivity was detected in follicular dendritic cells (arrowhead) and lymphocytes (arrow) of neoplastic follicles. Magnification, ×1,000. D, PFS according to STAT5a high (>median, 48 positive cells/hpf) and low(≤median) immunoreactivity.

Fig. 2.

Immunohistochemical analysis of FL tissue for STAT5a and CD4 expression. A, STAT5a immunoreactivity (red) was most abundant in perifollicular areas but was also involved in neoplastic follicles (high-expression pattern shown). Magnification, ×100. B, STAT5a expression strikingly colocalized with CD4+ T cells, as shown by the CD4 staining of a consecutive serial section of the corresponding neoplastic follicle. Magnification, ×100. In A and B, representative examples of neoplastic follicles are shown. C, STAT5a immunoreactivity was detected in follicular dendritic cells (arrowhead) and lymphocytes (arrow) of neoplastic follicles. Magnification, ×1,000. D, PFS according to STAT5a high (>median, 48 positive cells/hpf) and low(≤median) immunoreactivity.

Close modal

Subsequently, we analyzed the prognostic significance of immunohistochemically defined STAT5a expression. Stainings were initially categorized semiquantitatively into the low, intermediate, or high expression of STAT5a. Because this method showed significant association with patient outcome (P = 0.024 according to PFS in the Cox univariate analysis), we were encouraged to count the absolute number of STAT5a-positive cells. Again, high STAT5a expression (>median = 48/hpf; range, 4-108/hpf) was associated with better PFS. Of the 40 patients within the STAT5a-high group, 27 (68%) were in remission (median PFS not reached) compared with 16 (39%) of the remaining 41 patients in the STAT5a-low group (median PFS, 46 months; Fig. 2D). Clinically based FLIPI could also separate the high-risk patients from low- and intermediate-risk groups (data not shown). In a multivariate analysis with FLIPI, STAT5a expression (low versus high) had prognostic value on PFS (relative risk, 2.45; 95% confidence interval, 1.196-5.018; P = 0.014). The evaluation of STAT5a-positive cells separately in follicular areas did not have any predictive value. The independent association of STAT5a expression with PFS in the validation group of FL patients confirms the reproducibility of the microarray data.

Cytokine induction in FL cells in response to rituximab

To understand how the cross-talk between FL cells and their nonmalignant counterparts could be mediated, we performed further experiments in HF-1, SuDHL-4, Granta-519, and OciLy-3 B-cell lymphoma cells. Consistent with previous studies (25, 26), exposure of HF-1 cells to rituximab decreased their growth significantly (Fig. 3A), whereas in Granta-519 cells, the growth-inhibitory effect of rituximab was less apparent (Fig. 3B). Likewise, germinal center–derived SuDHL-4 cells reduced their growth, whereas OCI-Ly3 cells did not respond to rituximab at all (data not shown). The growth-inhibitory effect was cytokine independent, as none of the tested ILs (IL-2, IL-4, IL-6, IL-7) had direct effects on the growth of HF-1 or Granta-519 cells, and IL-2 could not further enhance the suppressive effect of rituximab (Fig. 3).

Fig. 3.

Effect of rituximab on the proliferation of HF-1 and Granta-519 cells. HF-1 (A) and Granta-519 (B) cells were exposed to rituximab or different ILs as indicated and analyzed for growth at indicated time periods. Columns, mean of three parallel samples; bars, SEM.

Fig. 3.

Effect of rituximab on the proliferation of HF-1 and Granta-519 cells. HF-1 (A) and Granta-519 (B) cells were exposed to rituximab or different ILs as indicated and analyzed for growth at indicated time periods. Columns, mean of three parallel samples; bars, SEM.

Close modal

Considering STAT5a activity and its prognostic effect in the tumor microenvironment, we next used our microarray database to identify STAT5a-activating ILs differentially expressed between control and rituximab-stimulated cells. This analysis revealed 4.1- and 2.7-fold inductions of IL15 and IL23A genes in HF-1 cells in response to rituximab (Fig. 4). Of these cytokines, IL-15 has been identified as a pivotal activator of T regulatory and effector cells (27, 28), regulator of natural killer cell proliferation, and enhancer of rituximab-mediated cytotoxicity (29), whereas IL-23A has been shown to have potent CD8+-mediated antitumor and antimetastatic activity in carcinoma and melanoma tumor models (30). In comparison, IL4, which signals through STAT4, was also induced, whereas IL2, which also signals through STAT5a and regulates Tregs, was not induced by rituximab. In addition, IL6 signaling through STAT6 was not affected. When we analyzed expression of ILs in rituximab-sensitive SuDHL-4 cells, IL-15 was the only rituximab-inducible cytokine. In comparison, no induction was observed in Granta-519 and OciLy-3 cells. Together, the data offers a possible explanation for the cross-talk between rituximab-stimulated FL cells and their microenvironment.

Fig. 4.

Expression of IL mRNAs in B-cell lymphoma cells in response to rituximab. Messenger RNA isolated from control and rituximab-treated lymphoma cells at 3 h was analyzed by oligonucleotide microarray for the expression of IL mRNAs. Data are presented as fold change compared with expression level in the control cells.

Fig. 4.

Expression of IL mRNAs in B-cell lymphoma cells in response to rituximab. Messenger RNA isolated from control and rituximab-treated lymphoma cells at 3 h was analyzed by oligonucleotide microarray for the expression of IL mRNAs. Data are presented as fold change compared with expression level in the control cells.

Close modal

STAT proteins comprise of a family of transcription factors that regulate diverse cellular events, such as differentiation, proliferation, and cell survival. A variety of cytokines and growth factors activate STAT factors by binding to cell surface receptors, which triggers the activity of receptor-associated JAK family members, including JAK1, JAK2, JAK3, and TYK2. JAKs phosphorylate STAT proteins, leading to their dimerization and transit to the nucleus (27, 31). The transcriptional targets of STAT proteins play roles in cell cycle progression as well as cell survival. Constitutively active STATs, particularly STAT3 and STAT5, often contribute to the malignant phenotype in diverse tumor types, including lymphomas (21, 22, 3235). However, in certain situations, STAT5 is associated with favorable outcome (22, 36) and can act as a tumor suppressor and inhibitor of metastasis (37, 38).

In the present study, we identified STAT5a as a favorable prognostic factor for FL patients treated with immunochemotherapy. We used gene expression profiling and pathway analysis, and found JAK-STAT pathway among the top ones differentially represented between the FL patients with favorable and adverse outcomes in response to R-CHOP regimen. Of the different STAT genes, STAT5a was the only one showing prognostic effect on survival. Pim1, which is a direct STAT5 target gene and previously identified as a negative regulator of STAT5 activity, was also positively correlated with the outcome. In contrast, several other STAT5 regulators had an inverse association with survival. Of these, IL7 gene is of particular interest, as IL-7 has been identified as a crucial cytokine for the STAT5-mediated development of T and B lymphocytes from common lymphoid progenitors, for the maintenance of mature T-lymphocytes (39, 40), for the development and maintenance of dendritic cells (41, 42), and for lymphoma development (21). The association of IL7 expression with adverse outcome in our patient cohort is in line with previous studies (21). However, the inverse association of IL7 and other STAT regulators, such as IL2Ra and IL4R, with STAT5a-associated survival contrasts with the role of STAT5a as a signaling effector for activating survival and proliferative pathways. Currently, the molecular interactions between these STAT5a pathway components are hypothetical. Nevertheless, their association with survival suggest that STAT5a pathway may have previously unidentified functions in FL.

Immunohistochemical analysis of an independent patient cohort was applied to validate the gene expression data. The findings showing that STAT5a expression predicts outcome both at mRNA and protein levels and, in two independent patient cohorts, encourages us to believe that a novel prognostic factor for immunochemotherapy-treated FL patients has been identified. Furthermore, the data show that gene expression technology, which is not currently available for routine clinical use, can be extended to protein level. Most importantly, we discovered that STAT5a expression was not primarily detected in malignant FL cells but rather in CD4-positive lymphocytes and follicular dendritic cells in the FL microenvironment. Similarly, Meier et al. (22) very recently reported that expression of STAT5 in perifollicular and follicular lymphocytes in the FL tissue is associated with improved prognosis.

Another major interest was to determine how the cross-talk between immunochemotherapy-stimulated FL cells and tumor inflammatory cells could be mediated. The finding that rituximab induced germinal center–derived lymphoma cells to express interleukins, which themselves are well-established STAT5a activators, suggests that rituximab can regulate STAT5a signaling in the FL tissue. The mechanism by which R-CHOP improves especially the outcome of the patients with high STAT5a expression is currently unknown but likely to be related to rituximab-triggered inflammatory response in the lymphoma tissue. Based on the current data, we propose that expression of STAT5a is a limiting factor for the efficacy of R-CHOP in FL. According to our hypothesis, the rituximab-dependent cytotoxicity is augmented by rituximab-induced secretion of cytokines, such as IL-15 from the FL cells, if the effector cells have high STAT5a expression. In contrast, if the effector cells have low STAT5a levels, this effector loop is impaired. Considering that STAT5a is primarily expressed in T cells and that IL-15 has been identified as a pivotal activator of natural killer, T regulatory, and effector cells (2729), one might suggest that IL-15–induced STAT5 activity in these cells contributes to the efficacy of rituximab.

In conclusion, we suggest that STAT5a expression in the tumor microenvironment predicts the outcome of FL patients in response to R-CHOP regimen. Our data not only suggest that nonmalignant tumor cells have a profound prognostic effect in FL but also imply that signals from FL cells to the surrounding microenvironment can have a biological effect. Questions of whether and how rituximab modifies STAT5a action and the potential target cells need further examination.

No potential conflicts of interest were disclosed.

We thank Onerva Levälampi and Minna Pietikäinen for the technical assistance and Dr. Michael Wenger and members of the Leppä research group for comments on the manuscript.

Grant Support: Finnish Academy of Sciences (S. Leppä), Finnish Cancer Societies (S. Leppä and M. Taskinen), Sigrid Juselius Foundation (S. Leppä), University of Helsinki (S. Leppä and M. Taskinen), Biomedicum Helsinki Foundation (M. Taskinen), and Helsinki University Central Hospital (S. Leppä). Rituximab (Mabthera) was provided by Roche.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1
Hiddemann
W
,
Buske
C
,
Dreyling
M
, et al
. 
Treatment strategies in follicular lymphomas: current status and future perspectives
.
J Clin Oncol
2005
;
23
:
6394
9
.
2
Herold
M
,
Haas
A
,
Srock
S
, et al
. 
Rituximab added to first-line mitoxantrone, chlorambucil, and prednisolone chemotherapy followed by interferon maintenance prolongs survival in patients with advanced follicular lymphoma: an East German Study Group Hematology and Oncology Study
.
J Clin Oncol
2007
;
25
:
1986
92
.
3
Hiddemann
W
,
Kneba
M
,
Dreyling
M
, et al
. 
Frontline therapy with rituximab added to the combination of cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP) significantly improves the outcome for patients with advanced-stage follicular lymphoma compared with therapy with CHOP alone: results of a prospective randomized study of the German Low-Grade Lymphoma Study Group
.
Blood
2005
;
106
:
3725
32
.
4
Marcus
R
,
Imrie
K
,
Solal-Celigny
P
, et al
. 
Phase III study of R-CVP compared with cyclophosphamide, vincristine, and prednisone alone in patients with previously untreated advanced follicular lymphoma
.
J Clin Oncol
2008
;
26
:
4579
86
.
5
Forstpointner
R
,
Unterhalt
M
,
Dreyling
M
, et al
. 
Maintenance therapy with rituximab leads to a significant prolongation of response duration after salvage therapy with a combination of rituximab, fludarabine, cyclophosphamide and mitoxantrone (R-FCM) in patients with relapsed and refractory follicular and mantle cell lymphomas—results of a prospective randomized study of the German low grade lymphoma study group (GLSG)
.
Blood
2006
;
108
:
4003
8
.
6
van Oers
MH
,
Klasa
R
,
Marcus
RE
, et al
. 
Rituximab maintenance improves clinical outcome of relapsed/resistant follicular non-Hodgkin's lymphoma, both in patients with and without rituximab during induction: results of a prospective randomized phase III intergroup trial
.
Blood
2006
;
108
:
3295
301
.
7
Hochster
H
,
Weller
E
,
Gascoyne
RD
, et al
. 
Maintenance rituximab after cyclophosphamide, vincristine, and prednisone prolongs progression-free survival in advanced indolent lymphoma: results of the randomized phase III ECOG1496 Study
.
J Clin Oncol
2009
;
27
:
1607
14
.
8
Dave
SS
,
Wright
G
,
Tan
B
, et al
. 
Prediction of survival in follicular lymphoma based on molecular features of tumor-infiltrating immune cells
.
N Engl J Med
2004
;
351
:
2159
69
.
9
Glas
AM
,
Kersten
MJ
,
Delahaye
LJ
, et al
. 
Gene expression profiling in follicular lymphoma to assess clinical aggressiveness and to guide the choice of treatment
.
Blood
2005
;
105
:
301
7
.
10
Harjunpää
A
,
Taskinen
M
,
Nykter
M
, et al
. 
Differential gene expression in non-malignant tumour microenvironment is associated with outcome in follicular lymphoma patients treated with rituximab and CHOP
.
Br J Haematol
2006
;
135
:
33
42
.
11
de Jong
D
,
Koster
A
,
Hagenbeek
A
, et al
. 
Impact of the tumor microenvironment on prognosis in follicular lymphoma is dependent on specific treatment protocols
.
Haematologica
2009
;
94
:
70
7
.
12
Beyer
M
,
Kochanek
M
,
Darabi
K
, et al
. 
Reduced frequencies and suppressive function of CD4+CD25hi regulatory T cells in patients with chronic lymphocytic leukemia after therapy with fludarabine
.
Blood
2005
;
106
:
2018
25
.
13
Lutsiak
ME
,
Semnani
RT
,
De Pascalis
R
, et al
. 
Inhibition of CD4(+)25+ T regulatory cell function implicated in enhanced immune response by low-dose cyclophosphamide
.
Blood
2005
;
105
:
2862
8
.
14
Minard-Colin
V
,
Xiu
Y
,
Poe
JC
, et al
. 
Lymphoma depletion during CD20 immunotherapy in mice is mediated by macrophage FcγRI, FcγRIII, FcγRIV
.
Blood
2008
;
112
:
1205
13
.
15
Taskinen
M
,
Karjalainen-Lindsberg
ML
,
Leppa
S
. 
Prognostic influence of tumor-infiltrating mast cells in patients with follicular lymphoma treated with rituximab and CHOP
.
Blood
2008
;
111
:
4664
7
.
16
Taskinen
M
,
Karjalainen-Lindsberg
ML
,
Nyman
H
,
Eerola
LM
,
Leppa
S
. 
A high tumor-associated macrophage content predicts favorable outcome in follicular lymphoma patients treated with rituximab and cyclophosphamide-Doxorubicin-vincristine-prednisone
.
Clin Cancer Res
2007
;
13
:
5784
9
.
17
Beissbarth
T
,
Speed
TP
. 
GOstat: find statistically overrepresented Gene Ontologies within a group of genes
.
Bioinformatics
2004
;
20
:
1464
5
.
18
Draghici
S
,
Khatri
P
,
Tarca
AL
, et al
. 
A systems biology approach for pathway level analysis
.
Genome Res
2007
;
17
:
1537
45
.
19
Eray
M
,
Tuomikoski
T
,
Wu
H
, et al
. 
Cross-linking of surface IgG induces apoptosis in a bcl-2 expressing human follicular lymphoma line of mature B cell phenotype
.
Int Immunol
1994
;
6
:
1817
27
.
20
Alvaro
T
,
Lejeune
M
,
Camacho
FI
, et al
. 
The presence of STAT1-positive tumor-associated macrophages and their relation to outcome in patients with follicular lymphoma
.
Haematologica
2006
;
91
:
1605
12
.
21
Abraham
N
,
Ma
MC
,
Snow
JW
, et al
. 
Haploinsufficiency identifies STAT5 as a modifier of IL-7-induced lymphomas
.
Oncogene
2005
;
24
:
5252
7
.
22
Meier
C
,
Hoeller
S
,
Bourgau
C
, et al
. 
Recurrent numerical aberrations of JAK2 and deregulation of the JAK2-STAT cascade in lymphomas
.
Mod Pathol
2009
;
22
:
476
87
.
23
Guiter
C
,
Dusanter-Fourt
I
,
Copie-Bergman
C
, et al
. 
Constitutive STAT6 activation in primary mediastinal large B-cell lymphoma
.
Blood
2004
;
104
:
543
9
.
24
Lam
LT
,
Wright
G
,
Davis
RE
, et al
. 
Cooperative signaling through the signal transducer and activator of transcription 3 and nuclear factor-{κ}B pathways in subtypes of diffuse large B-cell lymphoma
.
Blood
2008
;
111
:
3701
13
.
25
Harjunpää
A
,
Junnikkala
S
,
Meri
S
. 
Rituximab (anti-CD20) therapy of B-cell lymphomas: direct complement killing is superior to cellular effector mechanisms
.
Scand J Immunol
2000
;
51
:
634
41
.
26
Mattila
AM
,
Meri
S
. 
Responses to rituximab vary among follicular lymphoma B cells of different maturation stages
.
Scand J Immunol
2008
;
68
:
159
68
.
27
Hennighausen
L
,
Robinson
GW
. 
Interpretation of cytokine signaling through the transcription factors STAT5A and STAT5B
.
Genes Dev
2008
;
22
:
711
21
.
28
Passerini
L
,
Allan
SE
,
Battaglia
M
, et al
. 
STAT5-signaling cytokines regulate the expression of FOXP3 in CD4+CD25+ regulatory T cells and CD4+CD25- effector T cells
.
Int Immunol
2008
;
20
:
421
31
.
29
Moga
E
,
Alvarez
E
,
Canto
E
, et al
. 
NK cells stimulated with IL-15 or CpG ODN enhance rituximab-dependent cellular cytotoxicity against B-cell lymphoma
.
Exp Hematol
2008
;
36
:
69
77
.
30
Lo
CH
,
Lee
SC
,
Wu
PY
, et al
. 
Antitumor and antimetastatic activity of IL-23
.
J Immunol
2003
;
171
:
600
7
.
31
Imada
K
,
Leonard
WJ
. 
The Jak-STAT pathway
.
Mol Immunol
2000
;
37
:
1
11
.
32
Joliot
V
,
Cormier
F
,
Medyouf
H
,
Alcalde
H
,
Ghysdael
J
. 
Constitutive STAT5 activation specifically cooperates with the loss of p53 function in B-cell lymphomagenesis
.
Oncogene
2006
;
25
:
4573
84
.
33
Qin
JZ
,
Kamarashev
J
,
Zhang
CL
, et al
. 
Constitutive and interleukin-7- and interleukin-15-stimulated DNA binding of STAT and novel factors in cutaneous T cell lymphoma cells
.
J Invest Dermatol
2001
;
117
:
583
9
.
34
Bessette
K
,
Lang
ML
,
Fava
RA
, et al
. 
A Stat5b transgene is capable of inducing CD8+ lymphoblastic lymphoma in the absence of normal TCR/MHC signaling
.
Blood
2008
;
111
:
344
50
.
35
Kelly
JA
,
Spolski
R
,
Kovanen
PE
, et al
. 
Stat5 synergizes with T cell receptor/antigen stimulation in the development of lymphoblastic lymphoma
.
J Exp Med
2003
;
198
:
79
89
.
36
Nevalainen
MT
,
Xie
J
,
Torhorst
J
, et al
. 
Signal transducer and activator of transcription-5 activation and breast cancer prognosis
.
J Clin Oncol
2004
;
22
:
2053
60
.
37
Sultan
AS
,
Xie
J
,
LeBaron
MJ
, et al
. 
Stat5 promotes homotypic adhesion and inhibits invasive characteristics of human breast cancer cells
.
Oncogene
2005
;
24
:
746
60
.
38
Zhang
Q
,
Wang
HY
,
Liu
X
,
Wasik
MA
. 
STAT5A is epigenetically silenced by the tyrosine kinase NPM1-ALK and acts as a tumor suppressor by reciprocally inhibiting NPM1-ALK expression
.
Nat Med
2007
;
13
:
1341
8
.
39
Dai
X
,
Chen
Y
,
Di
L
, et al
. 
Stat5 is essential for early B cell development but not for B cell maturation and function
.
J Immunol
2007
;
179
:
1068
79
.
40
Johnson
SE
,
Shah
N
,
Panoskaltsis-Mortari
A
,
LeBien
TW
. 
Murine and human IL-7 activate STAT5 and induce proliferation of normal human pro-B cells
.
J Immunol
2005
;
175
:
7325
31
.
41
Esashi
E
,
Wang
YH
,
Perng
O
, et al
. 
The signal transducer STAT5 inhibits plasmacytoid dendritic cell development by suppressing transcription factor IRF8
.
Immunity
2008
;
28
:
509
20
.
42
Vogt
TK
,
Link
A
,
Perrin
J
,
Finke
D
,
Luther
SA
. 
Novel function for interleukin-7 in dendritic cell development
.
Blood
2009
;
23
:
3961
68
.

Supplementary data