J. Michels MD, PhD, Department of Medical Oncology, British Columbia Cancer Agency, 600 West 10th Avenue, Vancouver, BC V5Z E5F, Canada. E-mail: firstname.lastname@example.org
In the present study, we investigated the expression of Mcl-1 and Bcl-2 by immunohistochemistry in 85 patients of grades 1–3 and transformed follicular lymphoma (FL). In lymphoma tissue, centroblasts uniformly expressed high levels of Mcl-1 (Mcl-1hi) whereas centrocytes demonstrated low Mcl-1 expression (Mcl-1lo). Bcl-2 expression in centroblasts/centrocytes was reciprocal to Mcl-1 staining in most cases. A high number of Mcl-1hi centroblasts in tissue sections (≥200/high-power field) correlated with poor overall survival (P < 0·001), independent of the International Prognostic Index and FL grade. This suggests that the number of centroblasts with strong Mcl-1 staining is associated with clinical outcome in FL patients.
Follicular lymphoma (FL) comprises a spectrum of disease, with different histological grades and the tendency to undergo transformation into high-grade lymphoma (Bastion et al, 1997). Resistance to apoptosis contributes to both the pathogenesis of FL and failure of antilymphoma therapies (McDonnell & Korsmeyer, 1991). Mcl-1 and Bcl-2 are members of the Bcl-2 family proteins with antiapoptotic function. Beside the role of Bcl-2 overexpression in FL development, deregulated Mcl-1 expression may also have an important function in lymphomagenesis. Enforced expression of Mcl-1 promotes the development of lymphoma in mice with high probability (80%; Zhou et al, 2001), and Mcl-1 knockdown experiments have directly demonstrated an essential survival function of Mcl-1 in non-Hodgkin lymphoma (Michels et al, 2004). Mcl-1 expression in FL of different grades and a potential association with clinical outcome are not well described.
We examined the immunohistochemical expression of Mcl-1 and Bcl-2 and their potential correlation with clinical outcome in FL.
Materials and methods
Paraffin-embedded diagnostic biopsy material was obtained with Local Research Ethics Committee approval. Grading of FL was performed according to the World Health Organization (WHO) recommended method of Mann and Berard (Nathwani et al, 2001). Tissue material was stained using the primary antibodies S19 rabbit polyclonal Mcl-1 antibody (Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA) at 1:500 or Bcl-2, clone 124, mouse monoclonal antibody (Dako Ltd, Ely, UK) at 1:50. The overall intensity of Mcl-1 and Bcl-2 immunostaining was scored blinded to clinical outcome data. Mcl-1 staining was scored absent (score 0, <50% of cells with weak Mcl-1 staining) or strong positive (score 2+, ≥50% of cells with strong Mcl-1 staining). Weak positive staining (score 1+) was used for intermediate staining conditions. Bcl-2 staining were absent (score 0, <50% of cells positive for Bcl-2) or positive (score 1+, >50% of cells positive for Bcl-2).
Clinical outcome data were compared with the log-rank test. Hazard ratios were estimated using Cox proportional hazards regression. Correlation of continuous variables was tested for using the Spearman's rank correlation coefficient. The Statistical Package for the Social Sciences (spss) software, version 11.5 for Windows (SPSS UK Ltd, Surrey, UK), was used for analysis. The P-values of ≤0·05 (two-sided) were considered as statistically significant
Expression of Mcl-1 and Bcl-2
A predominant weak diffuse cytoplasmic staining of Mcl-1 was detected in malignant centrocytes (Mcl-1lo) but strong cytoplasmic staining with pronounced perinuclear accentuation (Mcl-1hi) was uniformly seen in malignant centroblasts (Fig 1). Overall intensity of Mcl-1 staining in tissue sections was related to the number of Mcl-1hi centroblasts (Fig 1A–D). Tissue sections from all 29 patients with grade 1 or 2 FL contained few malignant centroblasts and therefore stained weak for Mcl-1 (score 1+). Cases of grade 3 FL demonstrated increasing overall staining intensity for Mcl-1 proportional to an increasing number of Mcl-1hi stained centroblasts. Mcl-1 staining in transformed FL was strong (score 2+) due to high numbers of centroblasts. Bcl-2 was highly expressed in malignant centrocytes of most cases (score 1+ in 93% and score 0 in 7% of cases). Bcl-2 expression in malignant centroblasts was often reciprocal (Fig 1E,F): centroblasts in all tumour samples from patients with grade 1 FL were Bcl-2-negative. Centroblasts in some cases of grade 2/3 FL (11% and 13%, respectively) and centroblasts from most transforming FL (83%) stained strongly for Bcl-2.
Correlation of Mcl-1 and Bcl-2 expression and clinical outcome
The baseline clinical characteristics are shown in Table I. At the time of this analysis, 59 of 83 patients (71%) had progressive disease and 40 of 85 (47%) patients had died. On univariate analysis, International Prognostic Index (IPI; P < 0·001) and transformation (P = 0·043) were significantly correlated with clinical outcome, in contrast to morphological grades 1–3 FL (P = 0·39). Mcl-1 expression was not associated with treatment response or progression-free survival (PFS). In contrast, the number of Mcl-1hi centroblasts/high-power field (hpf) demonstrated a significant inverse correlation with overall survival (OS; r = −0·31, P = 0·001) and a high number of Mcl-1hi centroblasts (≥200/hpf vs. <200/hpf) was related to poor outcome [median OS: 24 months, 95% confidence interval (CI): 7·3, 40·3 vs. 98·6 months, 95% CI: 85·4–112·8, P < 0·001]. On multivariate analysis, IPI (HR: 6·9, 95% CI: 2·3–20·7, P = 0·001) and number of Mcl-1hi centroblasts ≥200/hpf (HR: 6·3, 95% CI: 1·7–22·9, P = 0·001) remained statistically independent prognostic factors for poor OS in contrast to transformation status (HR: 0·5, 95% CI: 0·1–2·1, P = 0·34). Positive Bcl-2 expression in malignant centroblasts demonstrated a trend towards worse outcome for PFS (28·0 months, 95% CI: 0·5, 65·3 vs. 13·5 months, 95% CI: 7·0–20·0, P = 0·093) and OS (106 months, 95% CI 78·9, 106·5 vs. 34·5 months, 95% CI: 0·5–84·0, P = 0·145).
Table I. Clinical characteristics of FL (n = 85) patients.
Frequency, n (%)
Median OS (95% CI; months)
The number of patients (n) with available clinical information for a given parameter is indicated. Patients were stratified into the IPI risk groups only if all or if at least four of the five prognostic indices were recorded and the missing variable would not have altered the assignment to a prognostic group. Two patients were not evaluable for response (lost from follow up). The median OS times and P-values (two-sided, log-rank test) are shown.
This study has shown that Mcl-1 is widely expressed in FL on immunohistochemistry similar to the observations in other series (Agarwal & Naresh, 2002). Our results confirm increasing Mcl-1 staining in higher grade FL, as has recently been reported (Cho-Vega et al, 2004). We demonstrated that the increase of Mcl-1 staining intensity with higher grade of FL was due to the accumulation of Mcl-1hi centroblasts rather than a general increase in Mcl-1 staining of malignant tissue. In mantle cell lymphoma and cutaneous T-cell lymphoma, increasing overall Mcl-1 expression has been associated with higher grade or later stage of disease (Khoury et al, 2003; Zhang et al, 2003), suggesting that deregulated Mcl-1 expression may be a feature of disease progression in lymphoma.
We observed that malignant centroblasts, unlike centrocytes, of lower grade FL frequently did not express Bcl-2. The majority of centroblasts in transforming FL; however, expressed high levels of Bcl-2. Patients with high Bcl-2 expression in centroblasts had a trend towards worse clinical outcome. Similar to the prognostic significance of Bcl-2 expression in centroblasts of diffuse large B-cell lymphoma (Gascoyne et al, 1997), this finding could indicate that Bcl-2 expression in malignant centroblasts of FL patients has potential biological and clinical significance. Further studies are therefore warranted to elucidate the relevance of our observation.
The distinct Mcl-1 staining pattern of centroblasts greatly simplified their identification and high numbers of centroblasts in FL specimens appeared to be related to clinical outcome. The association is probably due to cases of transformation included in this analysis although the correlation was independent of the transformation status. Grading of FL by the WHO recommended method of Mann and Berard (Nathwani et al, 2001) has no overall confirmed role in prognostication of FL patients, as in our series (Chau et al, 2003). This is partly caused by the difficult identification of centroblasts in specimens, relying upon morphology alone and the poor reproducibility of grading results (Dardick & Caldwell, 1986). Mcl-1 staining has the potential to revive the prognostic value of grading. Due to the relative small number of cases in this study, our results should be considered preliminary and hypothesis generating. Confirmatory studies are clearly needed to establish the potential role Mcl-1 as a biomarker in FL.
This study was supported by Cancer Research UK. The authors thank Ron Lee, Department of Pathology, University of Southampton, for performing the immunohistochemistry. Dr J. Michels is supported by an unrestricted grant-in-aid from Abbott Laboratories.