Genetic Factors in Familial Manifestation of Primary Mediastinal Large B-Cell Lymphoma over Two Generations

Primary mediastinal large B-cell lymphoma (PMBL) is a rare lymphoid malignancy, making up 6–12% of all large B-cell lymphomas. It occurs more often in women (46–76% of cases) with a median age of presentation of 32 years (range: 27–42) [1].

It is well established that a positive familial history of lymphoid malignancy is associated with an increased risk of lymphoma [2]. Although it is difficult to completely exclude nongenetic risk factors (e.g., dietary influence, shared environmental exposure), common genetic make-up is regarded as the most important factor contributing to such an increased risk. A handful of rare genetic syndromes and variety of single nucleotide polymorphisms have been implicated to the increased genetic susceptibility for lymphomas [2‒4]. The majority of the latter are common genome variants (>5% population frequency) with low to modest effect, identified by genome-wide association studies (GWAS) [2]. Rare single variants with moderate to strong predisposing effects, except for rare syndromic diseases, have not yet been identified. Moreover, as such studies require a large number of cases, it is often difficult to identify risk factors associated with a specific disease subtype. Owing to the rare occurrence of PMBL, familial risk factors for this disease have not been established. In one study, a KMT2A c.5533C>A, p.(His1845Asn) variant has been associated with familial clustering of PMBL [5]. Earlier, we contributed to the identification of a new TIRAP sequence variant c.241C>T p.(Arg81Cys) implicated in increased familial risk of PMBL [6].

Here, we report on a familial clustering of PMBL over two generations with a father and a son affected. We performed HLA typing, whole-exome sequencing (WES), and genome-wide aCGH analysis on the patients’ tumors and peripheral blood cells obtained upon disease remission and the peripheral blood of their non-affected relatives, searching for a possible genetic predisposition factor.

Mediastinal masses were detected in the father (patient I-1) and son (patient II-1) in the course of the workup of a symptomatic cough. The father was diagnosed in January 2006 at the age of 67 years, and the son in August 2013 at the age of 48. Patient’s I-1 tumor, diagnosed on a transthoracal core needle biopsy, displayed a large cell lymphoma composed of slightly polymorphic centroblastoid cells, surrounded by compartmentalizing argyrophilic fibers. The tumor cells were positive for CD20, CD30 (partially), CD79a, p63, and BOB1, while being negative for CD15, CD23, and LMP1 of EBV. Patient’s II-1 tumor was diagnosed on a thoracoscopic excisional biopsy and showed a partially necrotic, large-cell lymphoma composed of polymorphic centroblastoid cells with clear cytoplasm, surrounded by compartmentalizing argyrophilic fibers. The tumor cells were positive for CD20, CD23, CD45, CD79a, p63, and BCL2, while remaining negative for CD15, CD30, C-MYC, and EBER. Diagnoses of PMBL were established. Both patients suffered from stage IB bulky disease confined to the mediastinum, without further organ or bone marrow involvement, as established by CT (I-1) and PET-CT (II-1) scans. While there were no tumor-associated complications in the father, the son’s lymphoma caused thrombosis of the left subclavian vein and hoarseness due to tumor-associated palsy of the left nervus laryngeus recurrens. A weight loss >10% and elevated LDH accompanied both clinical presentations.

Treatment consisted of 6 courses of R-CHOP without radiotherapy in patient I-1 and 6 cycles of DA-EPOCH without radiotherapy in patient II-1. Both patients achieved a partial response at interim staging and complete remission at the end of therapy. The son’s secondary symptoms were completely relieved after therapy. After a follow-up of 17 years (father) and 10 years (son), both remain in complete remission without any therapy- or tumor-related secondary problems.

Materials and methods to obtaining the results described below are defined in online supplementary materials (for all online suppl. material, see https://doi.org/10.1159/000532053). aCGH revealed a complex aberrant tumor genome of patient II-1 with 54 different copy number segments (deletions or duplications) larger than 400 kbp affecting 15 autosomes and the X chromosome (Fig. 1). Tumor copy number profile of patient I-1 showed hyperdiploid genome with trisomies 9, 12, and 20 as well as gain in the short arm of chromosome 6. Both tumors of patients I-1 and of II-1 displayed a classical spectrum of PMBL-associated genetic aberrations: amplification of 9p24 (affecting the JAK2, PDL1/2 loci), duplication of the HLA locus at 6p, and partial deletion and amplification of chromosome 16 affecting the SOCS1 and CIITA locus, respectively. Other aberrations (e.g., gain of the REL locus at 2p16.1, deletion of the HLA at 6p21.3, and TNFAIP3 at 6q23.3) that are frequently reported in PMBL were absent.

At the nucleotide level, WES of tumor II-1 identified a total of 182 nonsynonymous somatic variants (online suppl. Table 1). Except for frameshift deletion of HIST1H3B, the majority of these variants have unclear functional consequences for tumor development as they have not been previously described as frequently occurring in PMBL [7, 8]. The tumor of patient I-1 was not sequenced due to insufficient material available and poor quality of extracted DNA.

Germline DNA investigations by WES and aCGH in affected patients as well as segregation analysis in their unaffected family members (I-2 and II-2) were performed, attempting to find an explanation for the familial clustering of this rare disorder. Closely considered variant candidates fulfilled all the following criteria: (1) nonsynonymous, deleterious (nonsense, truncating, damaging missense mutations as predicted by in silico analysis); (2) located in genes that are involved in immune cell development and function, cell cycle, apoptosis, DNA repair, carcinogen metabolism or are implicated in lymphoma predisposition by currently available studies; 3) present in both affected patients. The presence/absence of a genetic variant in the unaffected brother was not an inclusion/exclusion criterion due to the possibility of reduced-/age-related penetrance of the predisposing factor. Additionally, we looked into genes that were altered in the patients’ tumors, aiming to detect putative tumor suppressors inactivated by a somatic “second-hit.” Variants fulfilling the abovementioned criteria are listed in Table 1. In addition, genetic typing of HLA alleles was performed in patients I-1 and II-2 (Table 2), aiming to compare identified haplotypes with those already implicated in lymphoma predisposition.

These analyses did not reveal a single genetic factor explaining the familial clustering of PMBL. The previously described genetic variants in KMT2A[5] and TIRAP [6] genes were absent. Both patients I-1 and II-2 displayed a heterozygous nonsense variant in the HLA-DRB5 gene c.97C>T, p.(Arg33Ter), leading to truncation of the protein shortly after a signal peptide, resulting in loss of function. The HLA-DRB5 gene that encodes HLA class II DRβ-chain has been implicated in lymphoma predisposition by GWAS [9]. However, this gene has multiple paralogues (DRB1, DBR3, DRB4), suggesting that its function is redundant. Moreover, this particular variant represents a frequent variation in Europeans (MAF=21%). Thus, a strong predisposing effect of this variant is very unlikely.

A heterozygous CYP1B1 missense variant c.1103G>A, p.(Arg368His) was also detected in both patients. The affected gene encodes the CYP superfamily member cytochrome P450 1B1, functioning as heme-thiolate-monooxygenase. This protein is involved in the metabolism of various xenobiotics and is expressed in a large variety of tissues. The currently detected variant is reliably predicted to be damaging by multiple in silico algorithms (MetaLR rankscore 0.85), and in vitro functional studies have shown that this variant leads to decreased metabolic rates of various tested substrates including estradiol, arachidonic acid, and retinoid [10]. Moreover, this variant is heavily debated regarding its role in the development of autosomal recessive primary congenital glaucoma [11]. However, there is evidence that an active CYP1B1 function, rather than abrogation of it, may be linked to increased carcinogenesis [12], which is explained by the fact that CYP1B1 activates procarcinogens, bringing up their carcinogenic effect. It has been demonstrated that Cyp1b1-null mice treated with 7,12-dimethylbenz[a]-anthracene (DMBA) develop aggressive lymphomas less frequently than wild-type mice [13]. Thus, according to current knowledge, inactivating mutations in CYP1B1 as detected in this family should decrease the probability of lymphoma rather than predispose to it.

Other detected variants frequently affect interleukin and interleukin receptor genes (IL34, IL4R), suggesting a possible involvement of interleukin signaling as potential predisposition factor, especially given its important role in immune cell functioning. Additional mutated immune regulators included SEMA7A and BTNL2, which play an important role in integrin-mediated regulation of immune responses and suppressing T-cell proliferation, respectively.

Owing to the rarity of PMBL, no associations are known between HLA alleles and disease risk. However, such data exist for classical Hodgkin lymphoma (cHL), which is a similar disease based on gene expression and somatic mutational signatures [14]. Interestingly, our patients do not have HLA alleles implicated in increased risk for cHL but carry HLA*02 and DRB1*04:04 haplotypes, which are associated with a reduced risk for cHL [15]. Taking into account the limitations of such an analysis, we cautiously conclude that physiological variations of HLA alleles were not among the genetic predisposing factors in this particular case of familial PMBL clustering.

We report a family with occurrence of the same rare lymphoma subtype in two first-degree relatives across two generations. Thorough genetic characterization of tumor and germline genomes as well as segregation analysis in the two unaffected family members could not clearly determine whether this noticeable clustering has occurred due to an underlying genetic predisposition or purely by chance. The fact that we find KMT2A and TIRAP, genes that have been implicated in familial PBML previously [5, 6], unaffected in our present study implicates that genetic alterations in PBML, even in familial-associated cases, are more complex than hitherto appreciated. Our findings are in line with the current concept of lymphoma predisposition being most likely the net effect of an accumulation of factors with low effect size rather than one strong predisposing alteration.

The study was approved by the Ethikkommission Nordwest-und Zentralschweiz (EKNZ) reference No.: 2014-127, issued on 22 September 2014. Written informed consent was obtained from the patients for analysis of their biological samples and publication of their medical cases.

The authors declare that they have no conflict of interests.

The study was funded by the Krebsliga beider Basel, Grant No.: 09-2014.

D.J. designed the study, performed genetic analysis, analyzed the data, and wrote the manuscript; P.L. performed exome sequencing and reviewed the manuscript; A.T. performed histopathological characterization of cases, analyzed the data, and reviewed the manuscript; S.D. designed the study, performed histopathological characterization of cases, and reviewed the manuscript; and F.S. designed the study, performed clinical characterization of cases, and reviewed the manuscript.

Data are not publicly available due to legal reasons. Further inquiries can be directed to the corresponding author.