Impact of Cell-of-Origin and MYC/BCL2 Status on the Risk of Central Nervous System Relapse in Primary Breast Diffuse Large B-Cell Lymphoma
Article information
, Ga-Young Song3, Ho-Young Yhim,1,2, Dok Hyun Yoon4, Kyu Yun Jang5, Sang Eun Yoon6, Jin Seok Kim7, Jeong-Ok Lee8, Hyeon-Seok Eom9, Hyewon Lee9, Kyoung Ha Kim10, Ka-Won Kang11, Young Rok Do12, Soon Il Lee13, Han Sang Lee13, Hyo Jung Kim14, Ae Ri Ahn5, Deok-Hwan Yang3, Won Seog Kim6, Jae-Yong Kwak1,2
Abstract
Purpose
Primary breast diffuse large B-cell lymphoma (DLBCL) is a rare entity with a distinct relapse pattern involving the central nervous system (CNS). However, data regarding predictors of CNS relapse in this population remain limited.
Materials and Methods
CNS relapse was retrospectively analyzed in two multicenter cohorts comprising 53 patients with newly diagnosed primary breast DLBCL, including a prospective trial and real-world cohort, all treated with rituximab-based immunochemotherapy. The impact of baseline clinical parameters, cell-of-origin, and MYC/BCL2 dual expression (DE) status on CNS relapse was assessed using a multivariate Cox regression model, separately conducted for the overall study set (n=53) and the immunohistochemical study set (n=36).
Results
By the CNS-International Prognostic Index (CNS-IPI), most patients were classified as low or intermediate risk; no patients were classified as high risk. With a median follow-up of 58.8 months, the 4-year risk of CNS relapse was 15.6% in the overall study set and 14.2% in the immunohistochemical set. MYC/BCL2 DE was identified in 14 patients (38.9%) and was significantly associated with increased risk of CNS relapse (4-year risk, 30.7% vs. 0%, p=0.001). Patients with non-germinal center B-cell–like subtype had a numerically higher risk of CNS relapse. However, in multivariate analysis, only MYC/BCL2 DE status was associated with CNS relapse. Synchronous bilateral involvement was also an independent predictor of CNS relapse in both study sets. CNS-IPI was not discriminatory for CNS relapse.
Conclusion
MYC/BCL2 DE and synchronous bilateral breast involvement may help identify patients at higher risk for CNS relapse. Further studies are warranted.
Introduction
Primary breast diffuse large B-cell lymphoma (DLBCL) is a rare extranodal subtype of non-Hodgkin lymphoma, accounting for less than 3% of all extranodal lymphomas [1,2]. It predominantly affects middle-aged to elderly women and presents most commonly as a unilateral breast mass [3-9]. Although it usually manifests as localized stage disease, primary breast DLBCL has been associated with less favorable clinical outcomes compared to nodal DLBCL [4,6,7]. In addition, this entity exhibits a distinct treatment failure pattern, with disease relapse frequently occurring at extranodal sites such as the ipsilateral or contralateral breast or the central nervous system (CNS), highlighting its characteristic tendency for extranodal dissemination [3-10].
CNS relapse remains a rare but devastating complication in DLBCL [11,12]. Numerous studies have explored the incidence and associated risk factors for CNS relapse, including clinical features such as extranodal involvement > 1 site [13], elevated lactate dehydrogenase (LDH) levels [14], Eastern Cooperative Oncology Group performance status ≥ 2 [14], and specific extranodal involvements [15,16]. These clinical parameters have been collectively incorporated into the CNS–International Prognostic Index (CNS-IPI), a scoring system proposed to predict the risk of CNS relapse [11]. Moreover, non-germinal center B-cell–like (non-GCB)/activated B-cell–like (ABC) cell-of-origin (COO) subtypes [17,18] and biological aberrations such as MYC and BCL2 protein dual expression (DE) [17] or rearrangements [19] have been associated with a heightened risk of CNS relapse in nodal DLBCL. However, the CNS-IPI has limited utility in primary breast DLBCL [3], and reliable biologic predictors for CNS relapse are lacking. To date, no biologic correlative study has specifically addressed the risk of CNS relapse in primary breast DLBCL, a distinct clinicopathologic entity with a proclivity toward CNS involvement.
Herein, we investigated the incidence and risk factors for CNS relapse in patients with newly diagnosed primary breast DLBCL using data from prospective trial and real-world practice. We aimed to identify clinical and immunohistochemical (IHC) predictors of CNS relapse in our dataset, with particular focus on COO classification and MYC/BCL2 DE.
Materials and Methods
1. Study design and patients
Two independent cohorts were screened for eligibility to construct two distinct datasets for analysis. Cohort 1 included patients with newly diagnosed primary breast DLBCL who were enrolled in a multicenter, single-arm, phase 2 clinical trial (NCT01448096) conducted by the Consortium for Improving Survival of Lymphoma (CISL) between January 2012 and July 2017. This trial design was described in full elsewhere [3]. Briefly, patients received six cycles of R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone) combined with four doses of prophylactic intrathecal methotrexate. Eligibility criteria for this trial included histologically confirmed primary breast DLBCL, defined as lymphoma involving the breast as the major extranodal site with or without nodal involvement [20], and age between 20 and 75 years. Patients with baseline CNS involvement or evidence of disseminated extranodal disease were excluded. Cohort 2 included patients with histologically confirmed primary breast DLBCL, diagnosed between January 2012 and December 2020, who were treated with R-CHOP or similar immunochemotherapy at institutions participating in the CISL network outside of clinical trials. Except for the absence of an age restriction in cohort 2, the inclusion and exclusion criteria for this retrospective cohort were consistent with those of the phase 2 trial. Based on these two cohorts, we defined the overall study set as patients with available clinical data and the IHC study set as patients for whom formalin-fixed paraffin-embedded (FFPE) tumor specimens, obtained at initial diagnosis, were available for IHC assessment of COO and MYC/BCL2 protein expression status.
2. Treatment and assessment
Patients in cohort 1 received six 21-day cycles of standard R-CHOP, with methotrexate (12 mg) administered intrathecally on day 1 or 2 of the first four cycles as CNS prophylaxis. Radiotherapy was not routinely administered and was reserved for patients with bulky disease or residual lesions on end-of-treatment imaging. Treatment protocols in cohort 2 followed institutional standards and were based on R-CHOP or similar regimens, without predefined protocols for CNS prophylaxis or radiotherapy. Data regarding the administration of prophylactic methotrexate, use of radiotherapy, treatments, and clinical outcomes were retrospectively collected using a study-specific case report form for cohort 2.
Pretreatment evaluations included complete blood count, serum biochemistry, LDH levels, computed tomography (CT) of the chest, abdomen, and pelvis, fluorodeoxyglucose–positron emission tomography (PET), and bone marrow aspiration and biopsy. CNS involvement at baseline was excluded through clinical evaluation, cerebrospinal fluid cytology, and, where indicated, neuroimaging. The IPI and CNS-IPI [11] were calculated to determine prognosis and the risk of CNS relapse. Tumor response was assessed according to the Lugano 2014 criteria [21]. In cohort 1, PET-CT was performed at baseline, after three cycles, and at the end of treatment, whereas cohort 2 assessments followed local practice standards. CNS relapse was defined as imaging and/or cytological confirmation of CNS involvement in patients with clinical suspicion.
3. COO, immunohistochemistry and fluorescence in situ hybridization analyses
Diagnostic histology for all cases included in this study was reviewed and confirmed by two experienced hematopathologists (K.Y.J. and A.R.A.). All IHC analyses were centrally performed on diagnostic FFPE tissue sections using standard protocols [22]. The IHC slides were reviewed independently by two hematopathologists who were blinded to patients’ clinical outcomes. Each reviewer semiquantitatively scored the percentage of positively stained tumor cells in 5% increments. Any discrepancies between the reviewers were resolved through joint review and consensus on a multiheaded microscope. The COO subtype was determined according to the Hans algorithm [23], based on CD10 (clone 56C6, Novocastra), BCL6 (LN22, Novocastra), and MUM1 (MUM1p, Dako) expression. DE was defined by concurrent MYC (Y69, Ventana Medical Systems) and BCL2 (124, Ventana Medical Systems) protein overexpression, using ≥ 40% positivity for MYC and ≥ 50% for BCL2 as cutoff values [17]. Interphase fluorescence in situ hybridization (FISH) analysis for MYC gene rearrangement was centrally performed using the Vysis LSI MYC Dual Color Break Apart Rearrangement Probe (Abbott Molecular) in cases where sufficient FFPE tissues were available. At least 100 nuclei were evaluated, and MYC rearrangement was defined as the break-apart signals in ≥ 10%. In cases with MYC rearrangement, additional FISH testing to detect BCL2 and BCL6 rearrangement was performed using Vysis LSI BCL2 and BCL6 Dual Color Break Apart Rearrangement Probes (Abbott Molecular). Double-hit or triple-hit lymphoma was defined as concurrent rearrangement of MYC and BCL2 and/or BCL6.
4. Statistical analyses
The primary endpoint of this study was the cumulative rate of CNS relapse, defined as the time from diagnosis to the documentation of CNS involvement. Secondary endpoints included progression-free survival (PFS) and overall survival (OS). Time-to-event outcomes, including cumulative rate of CNS relapse, PFS, and OS, were estimated using the Kaplan-Meier method, and comparisons between groups were performed using the log-rank test. To identify potential risk factors for CNS relapse, univariate and multivariate Cox proportional hazards regression analyses were conducted separately in the overall and IHC study sets. Variables with p < 0.10 in the univariate analysis were included in the multivariate model. Descriptive statistics were presented as percentages for categorical variables and as medians with interquartile ranges (IQRs) for continuous variables. All statistical analyses were performed using R ver. 4.2.0. (R Foundation for Statistical Computing; http://www.r-project.org).
Results
1. Patient characteristics
A total of 53 patients with primary breast DLBCL were identified across two cohorts and included in the overall study set (Fig. 1). Of these, 33 patients were enrolled in a prospective phase 2 trial (cohort 1), and 20 were retrospectively identified from seven institutions (cohort 2). Among them, diagnostic FFPE tissue sufficient for histopathologic analysis was available for 36 patients, comprising the IHC study set. The primary reasons for exclusion from the IHC study set included lack of tumor material (n=8), institutional administrative restrictions (n=7), and patients’ refusal to provide archival tissue for additional testing (n=2).
Study flow. COO, cell-of-origin; DLBCL, diffuse large B-cell lymphoma.
Baseline clinical characteristics of the overall and IHC study sets are summarized in Table 1. The baseline characteristics and outcomes of patients included versus excluded from the IHC study set were generally comparable, suggesting that the IHC study set was reasonably representative of the overall population (S1 Table). The median age was 55 years in the overall study set and 57 years in the IHC study set, with all patients being female. Most patients presented with localized stage disease, unilateral breast involvement, and good performance status. Elevated serum LDH level was observed in approximately one-quarter of patients, and the majority had low or intermediate risk by CNS-IPI. Synchronous bilateral breast involvement was found in three patients (5.7%) of overall set and in two patients (5.6%) in IHC study set. Nearly all patients received CNS prophylaxis, predominantly with intrathecal methotrexate (12-15 mg, 4-6 doses), while a small subset received intravenous high-dose methotrexate (2.5-3 g/m²) twice—after cycle 2 and at the end of R-CHOP. The median follow-up duration was 58.8 months (IQR, 30.3 to 72.9) for the overall study set and 54.4 months (IQR, 20.5 to 71.7) for the IHC study set. The 4-year PFS and OS rates in the overall set were 78.2% (95% confidence interval [CI], 63.0 to 87.7) and 85.5% (95% CI, 70.1 to 93.3), respectively. In the IHC study set, the 4-year PFS and OS rates were 83.6% (95% CI, 64.7 to 92.9) and 89.8% (95% CI, 71.4 to 96.6), respectively.
Clinical characteristics
2. Incidence and patterns of CNS relapse
CNS relapse occurred in seven patients in the overall study set and in four patients in the IHC study set. The 4-year cumulative incidence of CNS relapse was 15.6% (95% CI, 6.7 to 27.9) in the overall set and 14.2% (95% CI, 0.4 to 26.3) in the IHC study set (Fig. 2A and B). The median time to CNS relapse was 19.2 months (range, 6.6 to 39.3 months) in the overall set and 18.2 months (range, 6.6 to 39.3 months) in the IHC study set. All but one CNS relapse occurred within the first two years of initial diagnosis. Approximately half of the CNS relapses occurred as isolated CNS events, while the remaining CNS relapses were accompanied by concurrent systemic progression. The distribution of CNS involvement sites is described in Table 2. These findings suggest a continued risk of CNS relapse in patients with primary breast DLBCL, even with the use of traditional intrathecal or systemic methotrexate prophylaxis.
Cumulative risk of central nervous system relapse and its association with immunohistochemical features. Cumulative risk of central nervous system (CNS) relapse in the (A) overall study set and the (B) immunohistochemical study set. (C) Distribution of CNS relapse events according to MYC/BCL2 dual expression status and cell-of-origin subtype among patients in the immunohistochemical study set. (D) Comparison of the CNS relapse risk by MYC/BCL2 dual expression status. CI, confidence interval; DE, dual expression; GCB, germinal center B-cell–like.
Distribution of CNS relapse
3. COO, MYC/BCL2 DE status, and CNS relapse risk
Among the 36 patients in the IHC study set, the majority (n=30, 83.3%) were classified as the non-GCB subtype according to the Hans algorithm, while six patients (16.7%) were categorized as GCB. MYC/BCL2 DE was identified in 14 patients (38.9%), of whom 13 were non-GCB. All four CNS relapse events occurred exclusively in patients with MYC/BCL2 DE and the non-GCB subtype (Fig. 2C). However, there was no statistically significant difference in the cumulative incidence of CNS relapse according to COO classification (non-GCB vs. GCB; 17.4% [95% CI, 0 to 31.9] vs. 0%; hazard ratio [HR], 28.11 [95% CI, not applicable], p=0.105) (Table 3). In contrast, a significant difference in the 4-year cumulative incidence of CNS relapse was observed based on DE status (DE vs. non-DE, 30.7% [95% CI, 1.17 to 52.0] vs. 0%; HR 90.60 [95% CI, not applicable], p=0.001) (Table 3, Fig. 2D).
Univariate analysis for CNS relapse in the overall set and immunohistochemical study set
FISH analysis was performed in 20 patients with sufficient FFPE material. Only one patient harbored a MYC rearrangement, without concurrent BCL2 or BCL6 rearrangements, and thus none of the patients was classified as double-hit or triple-hit lymphoma. Notably, the patients with MYC rearrangement did not experience CNS relapse during follow-up.
Univariate analysis of CNS relapse was conducted separately for the overall and IHC study sets, including key baseline variables (Table 3). In the overall set, synchronous bilateral breast involvement was significantly associated with increased risk of CNS relapse, while the presence of B symptoms showed a trend toward heightened risk. In the IHC study set, synchronous bilateral breast involvement, bulky (> 10 cm) mass, intermediate-risk CNS-IPI, and MYC/BCL2 DE were significant variables associated with CNS relapse. To identify independent prognostic factors for predicting CNS relapse, multivariate Cox regression analysis were performed for each dataset, including variables with p < 0.10 in the univariate analysis. In the overall study set, synchronous bilateral breast involvement (HR, 3.37; 95% CI, 1.39 to 8.15) was an independent factor for predicting CNS relapse (Table 4). In the IHC study set, bilateral breast involvement (HR, 3.73; 95% CI, 1.28 to 10.86) and MYC/BCL2 DE (HR, 2.42×104; 95% CI, not applicable) were identified as independent predictors of CNS relapse (Table 4).
Multivariate Cox regression analysis for predicting CNS relapse in the overall set and immunohistochemical study set
Discussion
In the present analysis, we investigated CNS relapse risk in patients with newly diagnosed primary breast DLBCL using both prospective trial and real-world cohorts, all treated with rituximab-based immunochemotherapy and predominantly receiving prophylactic intrathecal or systemic methotrexate. Although most patients received conventional CNS prophylaxis using intrathecal or systemic methotrexate, CNS relapse still occurred in approximately 15% of cases, with the majority of events developing within two years of diagnosis, consistent with previous reports [3-9]. Notably, CNS relapse occurred even in patients classified as low risk by CNS-IPI, and no patients in our dataset were classified as high risk, demonstrating the limitations of existing prognostic tools in this unique subset. Importantly, although validation studies are needed, we identified for the first time, to our knowledge, that MYC/BCL2 DE is an independent predictor of CNS relapse, while no cases of double-hit lymphoma were detected. These findings suggest the urgent need for studies on biologically informed risk stratification and the development of appropriate prophylactic strategies to improve CNS outcomes in high-risk patients with primary breast DLBCL.
Previous studies in nodal DLBCL have identified several biological features that may contribute to the risk of CNS relapse, including the COO [18], MYC/BCL2 DE [17], double-hit status [19], and dark zone lymphoma [24]. In the present study, MYC/BCL2 DE was identified as an independent predictor of CNS relapse. This finding is consistent with a report by Savage et al. [17], in which MYC and BCL2 co-expression by IHC was significantly associated with an increased risk of CNS relapse among patients treated with R-CHOP, independent of CNS-IPI and COO classification. These results suggest that MYC/BCL2 DE may serve as a biomarker for identifying patients at higher risk of CNS relapse in primary breast DLBCL.
Thirteen of the 14 DE patients were classified as non-GCB by IHC, consistent with previous reports that DE status is frequently associated with the non-GCB/ABC subtype [24-26]. Considering the frequent association between the non-GCB subtype and MYC/BCL2 co-expression, we explored whether the increased risk of CNS relapse might be related to the IHC-defined non-GCB subtype. Although the number of CNS relapses was numerically higher in patients with non-GCB subtype, we did not find a statistically significant difference in CNS relapse risk on multivariate analysis. This result should be interpreted with caution, as the small number of GCB cases (n=6) and absence of CNS relapse events in this group may have limited the statistical power to detect meaningful differences. Furthermore, the markedly high prevalence of the non-GCB subtype (83.3%) in our dataset, which is higher than that generally observed in other nodal DLBCL studies [17,18,27], may also have contributed to this finding. One possible explanation for this predominance is the genetic similarity of primary breast DLBCL to DLBCL of immune-privileged sites, where mutations such as MYD88 and CD79B are frequently observed [28]. Based on molecular classification of DLBCL, these alterations are enriched in the MCD/C5 genetic subtype, which is mostly classified as ABC or non-GCB [29,30]. Therefore, our dataset may inherently harbor a high proportion of CNS relapse-prone patients who are classified as non-GCB phenotype. Given that Bruton’s tyrosine kinase inhibitors have shown clinical activity in MCD-subtype DLBCL [31] and demonstrate adequate CNS penetration [32], these agents could be explored as potential preventive and therapeutic options for CNS relapse in this biologically defined subset of patients.
Double-hit or triple-hit lymphomas were identified in approximately 5% to 10% of nodal DLBCL cases [33]. However, data regarding their prevalence in primary breast DLBCL remain scarce. In our study, FISH analysis was conducted in 20 patients, and no cases of double-hit or triple-hit lymphoma were identified. Nevertheless, the small number of patients with sufficient tissue available for FISH testing limited the assessment of the incidence of double-hit or triple-hit lymphoma and their impact on the risk of CNS relapse.
Few clinical parameters have been associated with CNS relapse in primary breast DLBCL. Clinical parameters such as stage-modified IPI [5] and bulky disease [8] have been reported as potential risk predictors. However, these parameters have not been consistently validated across studies, and their predictive value remains uncertain. In contrast, our study demonstrated that synchronous bilateral breast disease appears to confer a higher risk of CNS relapse, which is in line with previous report [6]. This finding may suggest that bilateral breast disease at presentation indicates a more extensive burden of extranodal involvement, which may facilitate hematogenous dissemination to the CNS. Larger studies are needed to validate this observation and provide further insight.
Our study has several limitations. First, it was a retrospective analysis involving a relatively small number of patients, which may introduce unexpected selection bias and limit generalizability of the findings. Moreover, because of the small sample size, the number of patients who experienced CNS relapse events was also limited, with no events observed in the non-DE subgroup. This small number of CNS events might influence our results, as evidenced by the large HR estimates and wide range of CIs. Therefore, the findings should be interpreted with caution and considered exploratory rather than definitive. To address these limitations, future studies should focus on enrolling a larger patient population in a prospective manner, with systematic procurement of diagnostic tissue suitable for biomarker analysis. However, given the rarity of primary breast DLBCL, such studies may necessitate international collaboration, as conducting them within a single institution or an individual study group would be impractical. Second, our analysis focused specifically on the impact of DE status, COO, and double-hit status on the risk of CNS relapse. As a result, we were unable to conduct additional comprehensive genomic analyses. Recent genomic studies have shown that the MCD/C5 subtype, including rare extranodal DLBCL such as primary breast and testicular DLBCL, harbors a MYD88 mutation [29,30,34,35]. In addition, a recent study from the British Columbia group suggested that the presence of BCL6 and/or PDL rearrangements is associated with an increased risk of CNS relapse in testicular DLBCL, a well-known extranodal DLBCL with a high risk of CNS relapse [34]. Thus, comprehensive genomic analyses to identify genomic biomarkers in primary breast DLBCL are warranted to refine risk stratification. Nevertheless, given the limitations of current clinical risk models for predicting CNS relapse in primary breast DLBCL, our study is, to our knowledge, the first to demonstrate that MYC/BCL2 DE is a significant predictor of CNS relapse in this unique patient population.
In conclusion, CNS relapse occurred in a substantial proportion of patients who had received intrathecal or systemic methotrexate prophylaxis, with most events occurring within two years of diagnosis. While traditional clinical risk models, such as CNS-IPI, were not discriminatory in this unique subtype, our findings identify MYC/BCL2 DE as an independent predictor associated with an increased risk of CNS relapse. Furthermore, bilateral breast involvement at initial presentation was also associated with higher risk of CNS progression. Therefore, future studies with larger cohorts and comprehensive genomic profiling are warranted to refine predictive models and guide CNS-directed preventive strategies in patients with primary breast DLBCL.
Electronic Supplementary Material
Supplementary materials are available at Cancer Research and Treatment website (https://www.e-crt.org).
Notes
Ethical Statement
The study was approved by the institutional review board at each institution and conducted in accordance with the Declaration of Helsinki. All patients provided written informed consent.
Author Contributions
Conceived and designed the analysis: Lee CH, Song GY, Yhim HY.
Collected the data: Yoon DH, Yoon SE, Kim JS, Lee JO, Eom HS, Lee H, Kim KH, Kang KW, Do YR, Lee SI, Lee HS, Kim HJ, Yang DH, Kim WS, Kwak JY.
Contributed data or analysis tools: Jang KY, Ahn AR.
Performed the analysis: Lee CH, Song GY, Yhim HY.
Wrote the paper: Lee CH, Song GY, Yhim HY.
Critically revised manuscript and approved final version: Lee CH, Song GY, Yhim HY, Yoon DH, Jang KY, Yoon SE, Kim JS, Lee JO, Eom HS, Lee H, Kim KH, Kang KW, Do YR, Lee SI, Lee HS, Kim HJ, Ahn AR, Yang DH, Kim WS, Kwak JY.
Conflicts of Interest
Conflict of interest relevant to this article was not reported.
Funding
This paper was supported by Fund of Biomedical Research Institute, Jeonbuk National University Hospital.