Very late relapse in diffuse large B-cell lymphoma represents clonally related disease and is marked by germinal center cell features

With currently available treatment protocols, a complete first remission can be obtained in approximately 70% of diffuse large B-cell lymphoma (DLBCL) patients. Long-term disease-free survival can be achieved in about 40% of the patients as a whole group.¹  Most relapses occur within the first 2 to 3 years after diagnosis.²  Relapses after 4 years or more do occur but are rare. Relapse after 24 months after diagnosis is reported at a rate of about 2.2% per year.³  One of the most important questions regarding late relapses is whether they represent true, clonally related disease or de novo unrelated second, possibly therapy-induced, malignancy. Except for a few case reports and isolated cases in larger series,^3-6  no systematic data are available on this issue.

In terms of clinical follow-up strategies after treatment for DLBCL, it would be relevant to identify a subgroup of patients who are at risk for late relapse and who could be considered to have relapsed disease with a long first remission. Development of a second de novo non-Hodgkin lymphoma might be an indication of a (genetic) predisposition for B-cell lymphoma and may guide treatment choice in these patients.

From the clinical files of The Netherlands Cancer Institute, 13 cases of patients with DLBCL who presented with a relapse after a disease-free interval of more than 4 years after treatment since 1983 and of whom suitable histological material for immunohistochemical and molecular studies from both episodes was available were retrieved. The slides of all study samples and other relevant biopsy material were reassessed and reclassified according to the World Health Organization classification,⁷  and clinical data were collected (Table 1).

Immunohistochemistry on paraffin-embedded biopsy samples was performed using standard antigen retrieval methods and included MIB-1, CD20, CD79a, CD3, CD21, bcl-2, bcl-6 (all from DAKO, Glostrup, Denmark), and CD5 and CD10 (from Novacastra, Newcastle upon Tyne, United Kingdom).

After DNAisolation from representative biopsy samples of the initial episode and the relapse, immunoglobulin heavy (IgH) chain gene CDR3 and CDR2 regions were amplified and analyzed on a 3700XL capillary ABI Prism Sequencer (Applied Biosystems, Foster City, CA).⁸  Monoclonal rearranged products were isolated from agarose gels and sequenced using the ABI Prism BigDyePrimer Cycle Sequencing Ready Reaction Kit (Applied Biosystems). The most closely related immunoglobulin heavy chain variable (VH), diversity (DH), and joining (JH) genes were identified as compared to germ line sequences (VBASE) to record the patterns of shared and ongoing mutations.

Bcl-2/IgH polymerase chain reaction (PCR) assays for translocations in the bcl-2 major breakpoint region (MBR) and minor cluster region (MCR) regions were performed.^8,9 

DNA from lymphoma tissue and noninvolved normal tissue isolated as above was used for loss of heterozygosity (LOH) analysis on cases 6, 7, and 10 using a panel of 12 randomly distributed polymorphic microsatellite markers (D1S228, D1S158, D3S158, D4S1572, D4S430, D5S644, D5S346, D9S288, D9S63, D13S158, D14S65, and D18S58) with fluorescently labeled primers, allowing semiquantitative analysis using a 3700XL ABI sequencer. After normalization, the ratio between the 2 alleles was assessed, resulting in an LOH index. An LOH index below 0.5 was interpreted as LOH, between 0.75 and 0.5 as allelic imbalance, and above or equal to 0.76 as retention of heterozygozity.¹⁰ 

Patients with DLBCL with a relapse after a disease-free interval of less than 2.5 years were selected to match the cases on age, stage, and primary (extra)nodal localization. On the basis of sufficient histological material for immunohistochemistry and Ig/bcl-2 PCR, 38 controls were included (3 controls for most of the cases,⁶  4 for 2 cases, 5 for one case, and 2 for 4 cases). The conditional logistic regression model was used to model the matched binary data.

Patients with signs of transformed follicular lymphoma on the basis of an indolent component in any localization (bone marrow) at presentation or at relapse were excluded as in the cases.

Relapse of disease after a disease-free interval of 3 years and, even more so, very late relapses after 6 years or more mark a small but distinct group of DLBCL with a clearly different behavior with an estimated frequency of relapse of 2.2% per year after 24 months (including relapses with discordant histology of indolent lymphoma).³  A major, yet not fully resolved question remains whether these relapses are true clonally related DLBCL or actually represent the development of a second, unrelated DLBCL.

In this study, we could show a clonal relationship between the first presentation and the relapse after an interval of 4 to 17 years (median, 81 months; range, 50-207 months) in 10 of 13 cases. In 9 patients, IgH genes could be successfully amplified and sequenced for both episodes, showing identical variable-diversity-joining (VDJ)complexes. Identical bcl-2/IgH translocations in the MBR and MCR, respectively, were demonstrated in one of these cases and in one additional case, both indicative of a clonal relation (Table 2). The clonal relation of the 3 unresolved cases was further studied for LOH using a panel of hypervariable markers randomly distributed over the genome. Case 5 showed LOH of the same allele in 3 of 11 markers with gain of LOH on 3 loci but loss of LOH in 1 marker over time. This is suggestive of a clonally related process. On the basis of dissimilar patterns in 4 of 12 markers in case 9 and 9 of 12 markers in case 6, the patterns are suggestive of an independent origin of the tumors. Due to a relatively low frequency of LOH, the results are not fully conclusive, and since no strong clonal lymphoid marker was available in these cases, the results should be interpreted with some caution.

We then investigated if these late-relapsing tumors showed characteristic biologic features that may explain a divergent clinical behavior. On the basis of immunohistochemistry a germinal center (GC) immunophenotype as defined by combined expression of CD10, bcl-6, and bcl-2 protein^11,12  could be assigned to 9 of 13 cases. Shared mutations for the initial DLBCL and the relapse in comparison to the most closely related germ-line VH gene were found in 5 of 9 cases, and signs of ongoing mutation over time also in 5 of 9 cases with a range of 2 to 20 mutations. Intraclonal variation within one episode was not analyzed for this study. In case 3, only a 5′ stretch of CDR3 sequence could be obtained, consisting of JH4. No V or DH genes were included. Ongoing VH hypermutation is a feature that is considered to be characteristic for GC B-cells and is reflected in follicular lymphoma (FL).¹³  Post-GC B-cells and their malignant counterparts do show the imprint of a passage through a GC, but lack subclonal variation by ongoing hypermutation. Somatic hypermutation is also reflected in a subset of DLBCL.¹⁴  Although in our cases it cannot be excluded that these mutations are induced by chemotherapy, this is not very likely on the basis of the pattern of mutations in framework and complementary determining regions. Taken together, biologic features of GC B-cells were seen in 10 of 13 cases. In a matched control series of 38 patients with DLBCL who relapsed within 26 months after reaching complete remission (median, 10 months), a GC immunophenotype was found in only 21% of the cases. Compared to typical DLBCL, the late-relapsing cases show a relative high frequency of limited stage disease (6 patients presented initially with stage I or II extranodal disease and 4 with limited nodal disease). Since the control series was matched to the cases for significant risk factors for relapse, including age, stage, and extranodal localization (data on lactate dehydrogenase [LDH] and performance status were insufficiently available for matching or evaluation) and treatment was comparable in both groups, these findings support the notion that GC features are significantly correlated to the risk for late relapse in DLBCL, especially in limited stage disease (estimated odds ratio (OR) 7.04, hypothesis of OR = 1 can be rejected at P = .005 (2-tailed test) based on immunohistochemical data only).

How do these findings relate to the recently described gene-expression patterns in DLBCL^15,16 ? In a study based on the first series of DLBCL by Alizadeh et al¹⁵ , Lossos and coworkers showed that Ig gene hypermutation was shown to be found almost exclusively in DLBCL with a GC-like expression signature on microarray analysis in contrast to cases with an activated B-cell signature.¹⁷  This series is too small and follow-up too short, however, to assess a specific risk for late relapse. Interestingly, t(14;18) translocation also was shown to be restricted to cases with a GC-expression profile.¹⁸ 

Taken together, evidence from several immunohistochemical, molecular, and expression-profiling studies is converging that DLBCL with GC features encompass a distinct group of DLBCL with a better overall survival but possibly a characteristic risk for late relapse. In about 20% of the cases, the GC features may be related or even induced by the presence of t(14;18). Without necessarily being related to follicular lymphoma as a disease, the genetic makeup may result in a similar pattern of clinical behavior with late relapses that are so characteristic of FL. Provocatively, it may be suggested that late-relapsing DLBCL actually represents clinically overt transformed episodes of cryptic follicular lymphoma. The frequency of cushionlike paratrabecular tumor loci in the bone marrow of DLBCL patients that are morphologically highly reminiscent of FL and clinical studies showing late relapse of DLBCL with divergent histology of “low-grade disease” and, vice versa, the presence of a divergent follicular tumor component as a risk factor for late relapse in DLBCL, may support this hypothesis.^4,19  Although bcl-2 protein expression is a strong prognostic marker for overall and disease-free survival, it should be noted that the presence of t(14;18) in DLBCL has no such predictive value.^20-23 

The clinical implication of the finding of clonally related late-relapsing disease could be that these patients may be treated with second-line chemotherapy and then, if sensitive, would go to receive high-dose chemotherapy and an autologous stem cell transplant. A second de novo DLBCL might be treated again with first-line therapy.

In conclusion, we have found evidence that late-relapsing DLBCL actually represents truly clonally related disease episodes in most cases and that the characteristic clinical behavior may be related to the biologic features of germinal center cells.

We thank Dr P. Groenen and Dr J. H. J. M. van Krieken for amplification and sequencing of case 8. The pathologists and the Departments of Pathology of the Medisch Centrum Alkmaar, Alkmaar, Academic Medical Center, Onze Lieve Vrouwe Gasthuis, Sint Lucas/Andreas Ziekenhuis Slotervaart Ziekenhuis and Free University, Amsterdam; Ziekenhuis Gooi-Noord, Blaricum, Kennemer Gasthuis, Haarlem; Spaarne Ziekenhuis, Heemstede; Leiden University Medical Centre, Leiden; Ysselmeerziekenhuizen, Lelystad; and Ziekenhuis “de Heel” Zaandam are kindly acknowledged for providing histological material for this study.