Mutation of TP53, translocation analysis and immunohistochemical expression of MYC, BCL-2 and BCL-6 in patients with DLBCL treated with R-CHOP

Diffuse large B-cell lymphoma (DLBCL) is an aggressive lymphoma with diverse outcomes. Concurrent translocation of MYC and BCL-2 and/or BCL-6, and concurrent immunohistochemical (IHC) high expression of MYC and BCL-2, have been linked to unfavorable treatment responses. TP53-mutated DLBCL has also been linked to worse outcome. Our aim was to evaluate the aforementioned issues in a cohort of 155 patients uniformly treated with R-CHOP-like therapies. We performed direct sequencing of TP53 exons 5, 6, 7 and 8 as well as fluorescence in-situ hybridization (FISH) of MYC, BCL-2 and BCL-6, and IHC of MYC, BCL-2 and BCL-6. In multivariate analysis, TP53 mutations in L3 and loop-sheet helix (LSH) associated with a risk ratio (RR) of disease-specific survival (DSS) of 8.779 (p = 0.022) and a RR of disease-free survival (DFS) of 10.498 (p = 0.011). In IHC analysis BCL-2 overexpression was associated with inferior DFS (p = 0.002) and DSS (p = 0.002). DLBCL with BCL-2 and MYC overexpression conferred inferior survival in all patients (DSS, p = 0.038 and DFS, p = 0.011) and in patients with non-GC phenotype (DSS (p = 0.013) and DFS (p = 0.010). Our results imply that in DLBCL, the location of TP53 mutations and IHC analysis of BCL-2 and MYC might have a role in the assessment of prognosis.

the not-otherwise-specified (NOS) category, but were implied to have negative prognostic significance 3 . A diversity in biology and clinical outcome exists within the individual categories. DH lymphomas are usually clinically very aggressive with poor responses to first-line treatments and with short remissions.
The TP53 tumor suppressor gene located at chromosome region 17p13.1 encodes the p53 protein, which is involved in the regulation of cell cycle, DNA repair, apoptosis, and senescence after various stress signals, such as DNA damage and inflammation 4 . Loss of p53 function allows proliferation of cells with DNA damage and promotes neoplasia in transgenic p53-null mice 5 . Wild-type p53 functions as a cell-cycle checkpoint and a sensor of DNA damage in the cell 6 , and new functions keep emerging such as a role as a suppressor of inflammation 7 , and regulation of glucose metabolism 8 .
The TP53 gene is mutated in about 20% of cases of DLBCL 9 , and most of the published mutations affect p53-DNA interactions, resulting in a partial or complete loss of transactivation functions 10 . TP53 differs from other tumor suppressor genes in its mode of inactivation. While most tumor suppressor genes are inactivated by mutations leading to absence of protein synthesis or production of a truncated protein, more than 80% of TP53 alterations are missense mutations that lead to the synthesis of a stable full-length protein 11 . The location of the resulting amino-acid substitution is usually within the central DNA-binding domain (DBD) of p53, resulting in a loss of DNA-binding activity with consequent failure to transcriptionally activate target genes. The most commonly mutated areas in the DBD are loop-sheet-binding helices (LSHs) L2 and L3. Various mutations have different consequences for the function of the p53 protein 12,13 . Some mutations are associated with a loss of function and others with a gain of function.
Only a small number of studies combining TP53 mutation analysis, translocation data and double-expressor status in DLBCL have been published 14 . The results of previous studies imply that patients with combined mutation of TP53 and double-hit translocation fare poorly 15 . Our series is one of the largest investigated to date. In the present study, we pursue the clinical importance of TP53 mutation types combined with translocation and IHC data in patients with DLBCL.

Results
TP53 mutation. Patient characteristics are summarized in Table 1 with comparison between wild-type (WT) and mutated TP53. Out of 155 patient samples all exon sequencings were successful in 80 samples (51.6%). In 26 (16.8%) samples all but one exon were successful. Sequencing was either unsuccessful in more than one exon or totally unsuccessful in 49 samples (31.6%). Nine missense mutations with eight non-functional and one partially functional mutation in TP53 were detected in our patient material. One silent mutation with a synonymous protein product was also detected. The mutations are presented in Table 2. The total mutation frequency was detected 9.6%. In patient material with successful sequencing, 3 Fig. 1. Despite very low number of patients and events, in multivariate analysis mutation of LSH and L3 remained an independent prognostic variable as the relative risk of death from lymphoma was 8.779 (95% CI, 1.377 to 55.972, p = 0.022). LSH and L3 mutations were also independent prognostic factors for DFS (RR 10.498; 95% CI, 1.710 to 64.449, p = 0.011). Results should be considered suggestive and with caution due to low numbers in the subgroups. Immunohistochemistry. IHC GC and non-GC phenotyping was successful in 141 (91.0%) out of 155 samples. IHC evaluation of MYC and BCL-2 was possible in 128 (82.6%) samples and evaluation of BCL-6 in 129 (83.2%) (Fig. 3). A non-GC phenotype was associated with a trend towards worse DSS (3-year DSS 74.6% versus 83.5%, p = 0.123). High MYC expression correlated with an intermediate IPI score, compared with low-and high-risk scores (p = 0.007). High BCL-6 expression was associated with GC phenotype (p = 0.011).
Immunohistochemical p53 expression was associated with TP53 mutation (p = 0.00017) ( Table 3). The sensitivity of high p53 expression to find TP53 mutated cases was 55.6% and specifity 90.8%, respectively. Corresponding positive and negative predictive values were 31.3% and 96.4%, respectively. p53 immunohistochemical expression did not associate with traditional prognostic factors of DLBCL (performance status, IPI, stage, extranodal involvement) nor survival.
Associations between the studied parameters. Only two patients had concurrent TP53 mutation and BCL-2 translocation. TP53 mutations were located in LSH and L3 motifs (p = 0.021). These patients with concurrent BCL-2 translocation and TP53 mutation had very aggressive primary refractory DLBCL with a dismal outcome (mean DSS only 3 months, p = 0.000000002). There were no other associations between TP53 mutation and
Only a few studies have reported simultaneous analysis of genetic alterations and immunohistochemical protein expression of these genes [21][22][23] . Here we report a study describing p53 mutations, MYC, BCL-2 and BCL-6, translocations and immunohistochemical expression in a cohort of 155 newly diagnosed DLBCL cases. In the present study mutations of TP53 in LSH and L3 motifs were the only mutation types that had a strong association with poor survival. Overall, TP53 mutations were not associated with survival. High BCL-2 expression and DE status also had negative prognostic impact. DH translocation did not predict survival, nor did individual translocations.
Patients were however few and results should be considered suggestive, interpreted with care and need further validation. DH status has been established as a major survival predictor in DLBCL, and in the recently revised WHO classification DH DLBCL is categorized as an entity of its own among high-grade lymphomas (HGL) 18     studies have shown that patients with BCL-2 and MYC DH translocation show poor responses to treatment. However, some recent reports have not been able to confirm the very poor prognosis of this group. These discrepancies have been recently presented in a comprehensive review of the literature published by Rosenthal & Yunes 24 . Moreover, to describe further the biological impact of these translocations it been suggested that patients with MYC translocations should be substratified according to translocation partner 23,[25][26][27][28] . The issue is further complicated by the fact that MYC/BCL-2 and MYC/BCL-6 double hit lymphomas seems to be biologically distinct and probably should be addressed separately 24 . One study has shown that IHC analysis of BCL-2 and MYC expression might have more prognostic impact than FISH alone 18 . This implies that mechanisms other than translocation affect protein expression, which is also supported by the fact more patients have high MYC protein expression than translocation. In the present material, DH status and translocations did not predict survival. This might be due to the rarity of these cases. In the present series limited number of patients with DH lymphoma did not allow for substratification according to MYC translocation partner gene. In contrast to this, IHC predicted worse survival in the BCL-2 expression group and in the DE group using cut-off values established in the previous studies of DE 29 .
In this study we used cut-off value of 40% for MYC positivity, which is used in most previous studies. Work by Ambrosio et al. including a large series of 753 patients with aggressive B-cell lymphoma suggested that cut-off value of 70% might be able to better define the true poor prognosis group of DE lymphomas 30 . We repeated our analyses with this higher cut-off value also. This change increased the statistical power of established correlations but still we did not find correlations with survival. This discrepancy with the results of Ambrosio et al. may be explained by our smaller cohort.
The mutation frequency of TP53 is considered to be about 20% in de novo DLBCL 19,20 . Earlier studies of DLBCL have shown that most mutations occur in hot-spot regions, and mutations in LSH and L3 are associated with worse prognosis, while patients with L2 mutations show survival similar to those in WT groups. In a study by Young et al. in the pre-rituximab era concerning a cohort of 477 patients, 102 of the DLBCL cases were TP53-mutated. Mutations in LSH and L3 were associated with worse survival and TP53 mutations in the DBD region were considered markers of poor prognosis 19 . In a later study by Xu-Monette et al. a rituximab-treated cohort of 506 patients was studied. Of these, 112 patients with mutation of TP53 were detected and mutation was associated with worse prognosis. The study also established IHC-detected p53 as a suitable surrogate marker of mutation. A cut-off value of 50% quantified patients into a probable TP53 mutation group and IHC of p53 was shown to have prognostic potential. Deletion of TP53 was not associated with poor prognosis, only point mutations 20 . These studies, as well as other studies performed at rituximab era, established that mutations at the DBD region of TP53 were prognostic in regards to survival in DLBCL, regardless of treatment 14,21,22 .
In the present study, LSH and L3 mutations of TP53 were associated with poor survival. The other TP53 mutations were located in β-sheets (in non-DNA-binding domains) and did not predict survival. LSH and L3 mutations of TP53 mutations were however only detected in three patients and the statistical analysis should be considered with care. Moreover, while the value of p53 pathway in carcinogenesis is evident, the big picture seems to be much more complicated and cover a broader issue than just p53 gene mutations. p53 pathway is a complexed one with over 50 genes and proteins affected. Several genetic events commonly discovered in DLBCL, like ATM (Ataxia telangiectasia mutated) mutations and deletions, MDM2 (murine double minute 2) deletions and ARF (alternate reading frame of CDKN2A locus) loss may induce p53 dysfunction despite unaltered gene 31 . To add more complexity to the issue, it has been shown in CLL, that patients harboring bi-allelic loss of p53 function have a dismal prognosis 32 . Recently a large comprehensive study revealing the molecular subtypes of DLBCL, verified the same phenomena in DLBCL as well 33 . Together these facts imply that a broader approach discovering the genetic landscape of the disease should be preferred in the future.
Patients with TP53 mutations had a high frequency of primary refractory diseases. An interesting finding in our data was, however that among patients with mutated TP53 (n = 10), in whom primary treatment was successful, no relapses were detected. This might imply that these patients would possibly benefit from more intensive primary treatments or from different treatment strategies such as new targeted therapies, e.g. kinase inhibitors idelalisib or ibrutinib. In chronic lymphocytic leukemia, an effect of these drugs has been shown to be independent of functional TP53 genes 34,35 . In DLBCL ibrutinib has shown promise in treatment of ABC subtypes in a phase-2 trial 36 . These arising therapies warrant new studies to discover their therapeutic potential in high-risk DLBCL.
In our series TP53 mutation frequency was lower than previously reported, i.e. 12.5% versus 20%, and this difference might be explained by selection bias, because small samples were excluded from the study 20 . Although difficulties were expected with sequencing of paraffin-embedded samples, the total success rate of sequencing was not optimal. To improve the success ratio, we excluded small biopsy samples, e.g. core needle samples, and only selected the exons that harbour most of the functional mutations. In addition, we divided exons 5 and 8 into two parts to improve the output.
Because gene sequencing is a challenging method to apply to routine clinical practice, it would be attractive to use IHC as a surrogate marker to find the mutated cases. We found that high p53 protein expression correlated with TP53 gene mutations. However, it did not have statistically significant prognostic value, and half of the cases with strong expression had wild type p53 gene. These findings imply that p53 immunohistochemistry might be used for screening of the mutations but is not able to substitute sequencing 20 .
Here we report results of p53 gene sequencing, MYC, BCL-2 and BCL-6 FISH as well as MYC, BCL-2, BCL-6 and p53 immunohistochemistry in a moderate group of 155 DLBCL cases. Our data suggest that TP53 mutations in LSH and L3, and IHC high expression of BCL-2 and MYC are each independently associated with poor prognosis in patients with DLBCL. The impact of p53 mutations was limited. Together with other existing data, this implies, that in the future studies also the existence of other wild type gene should be taken into account. Although we had a moderate patient population, considering the excellent prognosis of these patients, few events; relapses and disease related deaths occurred. Combining this fact with the rarity of studied molecular features we could not do detailed subgroup analysis and the results should be addressed with caution. However, most of the published studies face this same problem, which should therefore be addressed in a meta-analysis combining several studies. Despite all these limitations we found our study adds knowledge to this field of prognostic impact of molecular events in DLBCL.  (Table 4) were used in the PCR amplification with AmpliTaq-Gold ® (Applied Biosystems) and in BigDye terminator v.1.1 cycle sequencing reactions (Applied Biosystems). For exons 5 and 8, two sets of primers were used in order to keep the PCR product sizes small (<200 bp) and thus suitable for sequence analysis of fragmented DNA. The PCR and sequencing reaction conditions are available upon request. All the sequencing reactions were carried out in both forward and reverse directions and any unclear results were confirmed by re-sequencing of the sample.
The sequencing reaction cleanup was performed with basic ethanol/EDTA precipitation. All the sequence data was analyzed with CodonCodeAligner v4.1.1 (CodoneCode Corporation) and Sequence Scanner v1.0 (Applied Biosystems) software. IARC database version R18, April 2016 was used to analyze the mutational data 37 .
IHC staining and FISH. Immunostaining and fluorescence in situ hybridization (FISH) analyses were performed as previously described 25,38,39 . For these stainings, tissue microarrays were constructed 40 .
Cut-off values were used to divide patients into high-and low-expression groups. Double-expressor (DE) lymphomas were defined as lymphomas with high expression irrespectively to the existence of gene translocations. The cut-off value used were BCL-6 staining over 50%, BCL-2 staining over 70% of the cells positive. For MYC IHC we performed analyses with both the cut-off value of 70% and 40%. The results are given mainly with the latter one.
The following FISH probes were used in accordance with the instructions of the manufacturer. BCL2 FISH DNA Probe, Split Signal, Code Y5407, Dako, Denmark; BCL6 Breakapart probe, LPH 035, Cytocell, United Kingdom; MYC FISH DNA Probe, Split Signal, Code Y5410, Dako, Denmark.
DH lymphomas were defined as those with concurrent MYC and BCL-2 or BCL-6 translocation. Triple-hit lymphomas were defined as lymphomas with MYC translocation combined with both BCL-2 and BCL-6 translocation.
Statistical analysis. Associations between the different variables and clinical parameters were assessed by using Pearson's 2-sided chi-square test. Kaplan-Meier analyses were used to assess survival rates and log-rank tests were used to determine the statistical significance. Disease-specific survival (DSS) was calculated from the date of diagnosis to the date of lymphoma-related death or the last follow-up date. Overall survival (OS) was calculated from the date of diagnosis to death from any cause or last follow-up. Disease-free survival (DFS) was calculated from the date of diagnosis to the date of relapse or date of death from any cause, or last follow-up date, whichever occurred first. p-values < 0.05 were considered significant. To evaluate the independent prognostic potential, all significant associations with survival in univariate analysis were analyzed by means of Cox regression using the enter method. The model included International Prognosis Index (IPI) divided into three categories according to risk, lactate dehydrogenase, Eastern Cooperative Oncology Group (ECOG) performance status, Ann Arbor stage, age, B-symptoms and extranodal involvement. The three IPI categories were as follows: low 0-1, intermediate, 2-3 and high risk, 4-5. Lactate dehydrogenase categories were normal and high. ECOG performance status categories were 0 or 1 and 2, 3 or 4. Ann Arbor-stage categories were Stages I-II and III-IV. Age categories were under 60 and 60 or more. B-symptom categories were no and yes. Extranodal involvement was divided to no extranodal disease or extranodal involvement. All statistical analyses were performed using the Statistical Package for the Social Sciences, v. 22.0 (IBM SPSS, Chicago, IL, USA).
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.