High nuclear expression of STAT3 is associated with unfavorable prognosis in diffuse large B-cell lymphoma

Seventy-four consented patients with DLBCL in the Beijing Cancer hospital from 2001-2007 were studied. In 58 patients, 27 cases were treated with R-CHOP and 31 cases with CHOP as first-line regimens. The clinical research protocol was approved by our Institutional Review Board (IRB). Archived formalin-fixed and paraffin-embedded tumor tissues were obtained from our Department of Pathology.

4 μm thick sections were mounted on APES-coated slides. After dewaxing in xylene and rehydrating in a gradient concentration of ethanol, the slides were immersed in methanol containing 0.3% hydrogen peroxide for 15 minutes to block endogenous peroxidase activity. All slides were pretreated with an antigen retrieval method by heating the slides in an autoclave in citrate buffer (10 mM, pH 6.0) for 90 seconds except those stained for P-STAT3. EDTA-Tris buffer (1 mM, pH 9.0) was used for pretreating before P-STAT3 staining. After rinsing in TBS (pH7.6), the specimens were incubated for 2 h at 37°C with anti-STAT3 antibody (sc-7179 rabbit polyclonal antibody, Santa Cruz Biotechnology) for STAT3, anti-P-STAT3 antibody (9145, rabbit monoclonal antibody, Cell Signaling Technology) for P-STAT3^Tyr705, and antibodies for BCL6, CD10, MUM-1 (Santa Cruz Biotechnology). Subsequently, all slides were incubated with Envision HRP antibody working fluid (Dako Company) for 30 minutes at 37°C, and then developed with DAB-H₂O₂ solution (Dako Company). The cell nuclei were stained with Meyer's hemotoxylin. The normal tonsil tissue was used as a negative control and breast cancer tissue stained positive was used as a positive control for STAT3 and P-STAT3 in all experiments. For technical details, see the manufacturer's instructions for each reagent.

IHC staining was evaluated by two independent experienced pathologists, who were blinded to the clinical data. As for the nuclear staining, at least 100 tumor cells per specimen were counted and only specimens showing moderate to strong immunoreactivity were considered positive. Staining was considered strong positive when > 75% of tumor cell nuclei were stained positive for STAT3 and > 30% of tumor cell nuclei for P-STAT3. Specimens stained positive for STAT3 ≤ 75% and ≤ 30% for P-STAT3 were considered weak immunoreactivity.

The Chi-square test was used to analyze the consistence of expressions of STAT3 in nucleus and P-STAT3. Correlation analysis of the STAT3 expression and the P-STAT3 level with clinicopathological variables was performed by two-sided Chi-square test. Kaplan-Meier method was used to estimate difference of OS. OS was defined as the time from diagnosis to death or the last follow-up. The Cox regression model was used to evaluate the prognostic value. The statistical software SPSS16.0 was used for all the statistical analysis.

All patients had complete follow-up information from the Tumor Registry Office in our hospital. The clinicopathological characteristics are summarized in Table 1. Fifty five patients were younger than 60 years old. Male and female patients were 30 and 44, separately. Twenty nine patients were diagnosed with B symptoms, 50 patients had stage III-IV diseases and 50 patients were diagnosed with the non-GCB subtype.

Among the 74 patients, 66 cases (89.19%) had the STAT3 expression, including 19 cases (25.7%) with strong nuclear staining of STAT3, and 24 cases (32.4%) with strong nuclear staining of P-STAT3. Representative staining outcomes were shown in Figure 1. There existed a consistence between the STAT3 expression and the P-STAT3 level (P = 0.001), indicating the reliability and accuracy of our IHC analysis (Table 2).

We observed the associations of the STAT3 nuclear expression with IPI score and molecular subtypes, but no statistical significances were reached (P = 0.099 and P = 0.061, respectively). No association was found between the STAT3 nuclear expression and other factors, including B symptoms, age of onset, clinical stage, and erythrocyte sedimentation rate (ESR), lactate dehydrogenase (LDH), and tumor size (Table 1).

Kaplan-Meier analysis showed that strong STAT3 nuclear expression was correlated with poorer OS (P = 0.005) (Figure 2). Other factors such as serum LDH level, clinical stage, B symptoms, tumor size, and IPI score were also shown to be correlated with OS (data not shown) as reported in other studies, which confirmed our data is reliable. A forward stepwise multivariate Cox model analysis, incorporating the above factors, demonstrated that the nuclear expression of STAT3 (P = 0.001), LDH level (P = 0.002) and tumor size (P = 0.025) were independent prognostic factor for OS.

To analyze the prognostic implication of STAT3 in term of Rituximab therapy, we stratified all patients into two subgroups, the CHOP subgroup and the R-CHOP subgroup. In CHOP subgroup, high nuclear expression of STAT3 predicted poor survival (P = 0.001). In R-CHOP subgroup, 2 of 19 cases died of DLBCL in low STAT3 cohort and 3 of 8 cases died in high STAT3 cohort. No significant association was observed between the expression of STAT3 and prognosis (P = 0.216) in the R-CHOP subgroup. But the survival curve showed that high STAT3 expression indicated poor OS in the first 40 months. Thus, it needs to increase the sample size to confirm this result (Table 3, Figure 3).