Clinical significance of cytogenetic aberrations in bone marrow of patients with diffuse large B-cell lymphoma: prognostic significance and relevance to histologic involvement

A total of 1585 DLBCL cases were referred from six tertiary hospitals in Korea: Seoul National University Hospital (n = 646; 1996 to 2011); Asan Medical Center (n = 484; 2001 to 2009); National Cancer Center of Korea (n = 236; 2004 to 2009); Yonsei University Hospital (n = 118; 2004 to 2009); Hallym University Hospital (n = 57; 2004 to 2009); and Korea Cancer Center Hospital (n = 44; 2005 to 2009). The cases were selected on the basis of diagnoses established according to the 2008 World Health Organization (WHO) classification criteria for primary tissue biopsy specimens [9]. BM biopsies were conducted for staging purposes at the time of the initial diagnosis. The treatment protocols were heterogeneous but generally conformed to international standards, including combination chemotherapy using CHOP-like regimens for front-line therapy, as well as salvage chemotherapy followed by stem cell transplantation for refractory cases. The baseline patient characteristics are summarized in Table 1. All of the patients were Korean, and the median age was 57 years (range, 2–91 years). A total of 1128 patients (71.2%) received R-CHOP as the initial therapy, 380 patients (24.0%) received a therapy other than R-CHOP (157 CHOP and 223 other regimens), and 77 patients (4.9%) received an unknown treatment or no treatment. The median follow-up time was 25.7 months (range, 0.1-211.1 months).This study was reviewed and approved by the institutional review board of each hospital.

A primary DLBCL diagnosis was established by examining hematoxylin and eosin (H&E)-stained sections of diagnostic biopsies from various tissues with IHC stains, including CD3, CD20, CD5, CD10, BCL2, BCL6, and IRF/MUM1, according to the diagnostic protocol of each institute. The type of DLBCL based on the cell of origin, either germinal center B-cell-like (GCB) or non-GCB, was defined in 877 cases using the algorithm of Hans et al. [10]. In general, the BM biopsies were performed bilaterally. Wright-stained BM smears and H&E-stained sections of BM biopsies were reviewed by hematopathologists at each institute. IHC staining was performed at the discretion of hematopathologists at each institute. The BM reports were reviewed centrally, and additional IHC staining targeting CD3, CD20, and CD79a was performed for cases with discrepant results between histologic examination and cytogenetic tests, to confirm the initial BM histologic diagnosis. The presence of benign lymphoid aggregates was distinguished from lymphoma involvement using previously described criteria [11]. Flow cytometric analysis was performed for some cases with diffuse infiltration of lymphoma cells (n = 46). According to the BM results, the cases were dichotomized into those with histological BM involvement (BMI_histo⁺) and those without (BMI_histo^-).

Conventional cytogenetic tests using the G-banding technique were performed on BM aspirates from all the patients. The cytogenetic tests were performed locally, and the reports were reviewed centrally by two of the investigators (SYK and DSL). Cytogenetic studies using standard techniques were performed as a part of the diagnostic work-up at the time of initial diagnosis. Conventional G-banding cytogenetic analysis was performed using the short-term unstimulated culture (24–48 h) of BM cells. At least 20 metaphases were analyzed, whenever possible. Clonal abnormalities were defined as at least two cells with the same aberration if the aberration is a chromosome gain or a structural rearrangement, or 3 or more cells with the same chromosome missing. The karyotypes were recorded according to the International System for Human Cytogenetic Nomenclature (ISCN) [12]. We classified the karyotype results according to the complexity of the chromosomal abnormalities observed. A complex karyotype was defined as ≥ 3 chromosomal abnormalities, in accordance with previous studies [13, 14]. A monosomal karyotype was defined as either a single autosomal monosomy in the presence of one or more structural aberration or two or more distinct autosomal chromosome monosomies [15, 16]. In the interpretation of specific abnormalities, numerical aberrations included gains, losses of chromosomes (aneuploidy) and changes in ploidy. Structural aberrations included abnormalities such as deletions, translocations, isochromosomes, and duplications. Translocations included balanced reciprocal translocations and translocations with unknown partners (additions).

Fluorescence in situ hybridization (FISH) was performed in some cases (n = 235). The following target regions were investigated using the accompanying probes: 14q32/IGH using an IGH dual-color, break-apart rearrangement probe (n = 201); cMYC/IGH using a dual-color, dual-fusion probe (n = 11); BCL2/IGH using a dual-color, dual-fusion probe (n = 29); 9p21/p16 using a p16/CEP 9 dual-color probe (n = 178); 3q27/BCL6 using a BCL6 dual-color, break-apart rearrangement probe (n = 95); 8q24/cMYC using a cMYC break-apart probe (n = 39) or a cMYC/IGH using dual-color, dual-fusion probe (n = 11); 1q25 using a 1p32/1q25 probe (n = 49); 17p13/TP53 using a P53 probe (n = 34); and 18q23/BCL2 using a BCL2/IGH dual-color, dual-fusion probe (n = 29; all probes from Abbott/Vysis, Downers Grove, IL, USA). We analyzed interphase cells according to the manufacturer’s instruction and the ISCN criteria. At least 200 nuclei per sample were scored for normal or abnormal FISH signals. The normal cut-off values for translocation, deletion, or amplification were based on the mean (± 3SD) and the binomial distribution function [17] analyzed of 40 negative controls. The cut-off values for IGH break-apart probe was 2%, and cMYC/IGH probe was < 0.5%.

The data were compared using the Mann–Whitney and Kruskal-Wallis tests for continuous variables and χ² test for categorical variables. Each numerical abnormality and the specific locus of each structural abnormality were dichotomized as present or absent, and hierarchical clustering was performed using Pearson correlation distance metrics and Wald linkage tests. The probabilities of overall survival (OS) and progression-free survival (PFS) [18] were plotted according to the Kaplan-Meier method, and the log-rank test with Bonferroni correction for multiple testing was used to compare the survival curves. A multivariate analysis was performed using the Cox regression method. The following parameters were analyzed for multivariate analysis: advanced age, gender, IPI risk groups, history of R-CHOP treatment, BMI_histo⁺ vs BMI_histo^-, ≥ 2 abnormalities vs normal karyotype or 1 abnormality, and presence of several specific cytogenetic abnormalities abnormalities, which were associated with poor prognosis in univariate analysis and found in a minimum of 5 patients. Variables in the final model were selected using stepwise selection procedure with a threshold of P = 0.05. The statistical analyses were performed using SPSS version 15 (SPSS, Chicago, IL, USA) and the R statistical package (R Development Team 2012). A probability level of 0.05 was considered significant in the univariate analysis. When multiple hypothesis testing was performed, the P value was adjusted by Bonferroni correction.

A total of 259 patients (16.3%) had BM involvement, as determined through histologic examinations. Among the 259 BMI_histo⁺ patients, 181 (69.9%) exhibited lymphoma cells on BM aspirate smears, and all 259 patients demonstrated lymphoma involvement in a BM biopsy. The median percentage of lymphoma cells in the BM aspirate smear was 6.2% (range, 0-98%). Compared with the BMI_histo^– group, the BMI_histo⁺ group had a poorer performance status, higher lactate dehydrogenase (LDH) levels, more advanced stage tumors, and more prevalent extranodal involvement; consequently, these patients had higher IPI scores (Table 1).

Chromosomal abnormalities were detected in 192/1585 patients (12.1%), of whom 124/192 (64.6%) were BMI_histo⁺ (Table 2). Among the 192 patients with chromosomal aberrations, 42 (21.9%) exhibited the following single numerical aberrations: 33 patients had Y chromosome loss; 7 patients had a loss or gain of other single chromosomes; and 2 patients had hyperdiploid clones (Additional file 1: Table S1). Among the 42 patients with the above single numerical aberrations, 4/42 (9.5%) were BMI_histo⁺, which was not significantly different from the proportion of BMI_histo⁺ patients with normal karyotypes (135/1393, 9.7%; P = 0.971). There were 12 patients with single structural abnormalities, of whom 4 patients exhibited single deletions [del(13q) in 2 patients and del(20)(q11.2) in 2 patients], and 1 patient had a duplication of Y chromosome; none were BMI_histo⁺. Seven patients exhibited translocations, including t(2;11)(p21;q23), t(4;10)(q28;p13), t(6;18)(p23;p11), t(10;11)(q22;q23), add(12)(q24), t(3;14)(q27;q32), and t(14;18)(q32;q21); the patients with the latter 3 types of abnormalities were BMI_histo⁺. The remaining 138/192 patients (71.9%) exhibited ≥ 2 aberrations. Ten of these patients exhibited 2 chromosomal abnormalities: 6 patients had 1 structural aberration and 1 numerical aberration, and 4 patients had 2 structural aberrations. Among the patients with 2 chromosomal abnormalities, 6/10 (60%) were BMI_histo⁺. A total of 128 patients exhibited complex karyotypes. Many of these patients had highly complex abnormalities, with a median of 9 total chromosomal abnormalities (range, 3–25) including many structural abnormalities (median, 7; range, 0–20). Among the patients with complex karyotypes, 111/128 were BMI_histo⁺ (86.7%). Consequently, 117/138 patients (84.8%) with ≥ 2 chromosomal aberrations were BMI_histo⁺. Among the patients with complex karyotypes, the percentages of abnormal metaphases were higher in the BMI_histo⁺ group compared with the BMI_histo^- group (P = 0.024). The percentages of metaphases with Y chromosome losses were higher in the BMI_histo⁺ group (P = 0.021), however, there was no significant differences in the number of metaphases for patients with loss of single chromosomes other than Y, polyploidy or single structural abnormalities between the BMI_histo⁺ and BMI_histo^- groups. Among the patients with complex karyotypes, only a weak correlation was found between the percentage of metaphases with aberrant karyotypes and the percentage of lymphoma cells in BM aspirates (r = 0.365). Our results indicate that a finding of BM cytogenetic aberrations involving single numerical abnormalities alone cannot be regarded as sufficient evidence of BM involvement of lymphoma cells, considering the low concordance between such findings and the histologic results. If multiple chromosomal abnormalities were considered as definite evidence of the presence of lymphoma cells, the 21/1585 (1.3%) patients with isolated cytogenetic aberrations without histologic involvement would be reassessed as having BM involvement.

We compared the clinical and biological characteristics of the BMI_histo⁺ patients with normal karyotypes to patients with abnormal karyotypes. Compared with the patients with normal karyotypes, the patients with a single numerical abnormality presented with less aggressive disease, as evidenced by their lower LDH levels, less involvement of extranodal sites, and lower IPI risk scores. In contrast, the patients with ≥ 2 chromosomal abnormalities had a poorer performance status, higher LDH levels, more advanced disease stages, more prevalent extranodal involvement, and thus, higher IPI scores (Table 3).

The histologic examinations, conventional cytogenetics, and FISH results of a limited number of patients were compared (Table 4). Among the 235 patients for whom FISH studies were performed, there were 3 BMI_histo^- patients with normal karyotypes and abnormal 14q32/IGH FISH results, although the frequencies of abnormal FISH signals were low (5%, 7%, and 8%). A single BMI_histo^- patient exhibited cytogenetic abnormalities of the 8q24 locus despite normal FISH results using a specific probe.

We analyzed the frequencies of specific cytogenetic aberrations among the patients with chromosomal aberrations other than single numerical aberrations (n = 150) (Figure 1). The chromosomes most frequently involved were chromosome 1, 3, 6, 14, and 18. The most common numerical aberrations were trisomy 18, trisomy 7, trisomy 3, loss of Y, and loss of 13. The predominant structural aberrations involved the following loci: rearrangements involving 14q32, 19q13, 19p13, 1p32-p36, 3q27, 8q24, 18q21-q23, 1cen-1q12, 9p22-p24, 11q23-q25, 16q22-q24, and; deletions of 6q; and duplications of 1q. The well-known oncogenes and lymphoma-related genes that exhibited frequent breakpoints, including BCL6 (3q27), JAK2 (9p22), IGH (14q32), and BCL2 (18q21), as well as other possible oncogenes with breakpoints, are indicated in Figure 1D. Monosomal karyotypes were observed in 70 patients (46.7%), and hyperdiploidy was noted in 24 cases (16.0%). Reciprocal translocations were observed in 90 patients (60.0%); among them, 19 carried reciprocal translocations involving the 14q32/IGH region with defined partners. The translocation partners were the following: t(8;14)(q24;q32) in 5 cases; t(14;18)(q32;q21) in 4 cases; t(3;14)(q27;q32) in 3 cases; t(1;14)(q21;q32); t(1;14)(q25;q32); t(3;14)(p25;q32); t(6;14)(q25;q32); t(9;14)(p13;q32); t(9;14)(q13;q32); and t(14;19)(q32;q13).

We analyzed the prognostic impacts of chromosomal abnormalities according to the complexity of the chromosomal aberrations involved (Figure 2). There was a significant difference in prognosis between patients with ≥ 2 abnormalities and those with a normal karyotype or a single abnormality both in the BMI_histo⁺ group (5-year OS, 22.0% vs. 52.7%; P < 0.001) and BMI_histo^- group (5-year OS, 31.8% vs. 66.5%; P < 0.001). In BMI_histo^- groups, the patients with a single numerical abnormality and those with single structural abnormality did not exhibit significant differences in OS compared with those with normal karyotypes (P = 0.422 and 0.137, respectively). The number of BMI_histo⁺ patients with single abnormalities was too small to assess prognostic impact (Additional file 2: Figure S1). There was no significant difference in OS and PFS according to the number of abnormalities among the patients with ≥ 2 abnormalities (Additional file 3: Figure S2).

When patients were classified according to IPI risk groups, BMI_histo⁺ patients with ≥ 2 cytogenetic abnormalities presented significantly poorer prognosis compared with normal karyotypes in the high/intermediate-risk group (5-year OS, 35.4% vs. 69.4%, P < 0.001; Figure 3A). BMI_histo^- patients with ≥ 2 abnormalities revealed significantly poorer prognosis in the high/intermediate and high risk groups (5-year OS, 31.8% vs. 58.4%, P = 0.027 for high/intermediate risk group; 0% vs. 40.6%, P < 0.001 for high risk group, respectively; Figure 3B).

When R-CHOP-treated patients were analyzed separately, ≥ 2 chromosomal abnormalities were associated with poorer prognoses compared with normal karyotypes or 1 abnormality among both the R-CHOP-treated BMI_histo⁺ patients (5-year OS, 33.9% vs. 70.9%; P < 0.001) xand the BMI_histo^- patients (5-year OS, 39.5% vs. 72.2%; P = 0.003; Additional file 4: Figure S3).

The presence of a monosomal karyotype had no apparent prognostic value among patients with ≥ 2 abnormalities (5-year OS, 25.4% vs. 20.0%; P = 0.274). There was no significant prognostic value of hyperdiploidy among patients with ≥ 2 abnormalities (5-year OS, 24.7% vs. 23.1%; P = 0.413). Among the 877 patients for whom IHC information was available, the non-GCB group demonstrated a lower OS than the GCB group (5-year OS, 58.6% vs. 71.2%, respectively; P = 0.004). However, among patients with ≥ 2 abnormalities, there was no significant difference in OS between the non-GCB and GCB types (5-year OS, 25.4% vs. 23.4%, respectively; P = 0.467).

We investigated the prognostic impacts of the specific chromosomal abnormalities. Figure 4A presents hazard ratios (HRs) obtained by univariate Cox analysis for OS and PFS according to specific chromosomal abnormalities found in ≥ 8 patients using BMI_histo⁺ patients with normal karyotype as a reference group among 327 patients with either cytogenetic abnormalities or were BMI_histo⁺. The presence of structural abnormalities at 16q22-q24 was significantly associated with a higher risk for both OS (HR, 5.86, P < 0.001) and disease progression (HR, 4.05, P < 0.001). The abnormalities at 6p21 (HR for OS, 4.89, P < 0.001), 12q22-q24 (HR, 4.28, P < 0.001), and -17 (HR, 4.49, P < 0.001) were also significant association with adverse prognosis. Consistent results were observed in the R-CHOP-treated patients (Additional file 5: Figure S4). In addition, frequent loci of cytogenetic abnormalities, including 11q21-q23, 19q13, 18q21, 1q21-q23, 8q24, 19p13, 3q27, 6q, and 14q32 were also associated with adverse outcomes.

To identify the subgroups of BM-involved DLBCL that exhibited distinct cytogenetic aberrations, we performed cluster analyses using the loci associated with poor survival in the univariate analysis and frequent and characteristic breakpoints, including 14q32, 3q27, 8q24, 19p13, and 19p13. Based on the hierarchical cluster analysis, 138 patients with ≥ 2 chromosomal abnormalities were segregated into clusters with characteristic patterns of chromosomal abnormalities (Figure 4B). When we compared prognoses among the clusters, Cluster 7, which was composed of loci associated with poorer prognoses in the univariate analysis, revealed the poorest prognosis (Figure 4C).

A multivariate analysis was performed for 327 patients who had either chromosomal abnormalities or were BMI_histo⁺ (Table 5). When the presence of ≥ 2 cytogenetic abnormalities was analyzed with other covariates, the presence of ≥ 2 cytogenetic abnormalities was also significantly associated with a worse OS (HR, 2.49; 95% CI, 1.75-3.54; P < 0.001). The high IPI score was strongly associated with a poor prognosis, whereas R-CHOP treatment was strongly associated with a better prognosis. When the specific chromosomal abnormalities which were associated with adverse prognosis in the univariate analysis were analyzed, the aberration at 19p13 was selected as an independent adverse prognostic factor (HR, 2.67; 95% CI, 1.50-4.76; P = 0.001), in addition to 7q22, 12q22-q24, 18q21, and 16q22-q24. When PFS was analyzed, aberrations at 19p13 and 8q24 emerged as factors independently associated with disease progression (HR, 3.02 and 2.61, respectively; P < 0.001 and P < 0.001, respectively). When 200 R-CHOP treated patients among 327 patients were analyzed separately, 19q13 (HR, 3.36, P = 0.003), 12q22-q24, 19p13, and 8q24 were independently predicted poor OS, and 19p13 and 8q24 were associated with disease progression (Table 5).