Different primary thyroid B-cell lymphomas show overlapping mutation profiles, suggesting involvement of a common pathogenic process

Abstract

Primary thyroid lymphomas commonly originate from a background of Hashimoto’s thyroiditis and comprise largely extranodal marginal zone lymphoma of mucosa-associated lymphoid tissue (EMZL), diffuse large B-cell lymphoma (DLBCL), and follicular lymphoma (FL). Thyroid EMZL harbors a distinct mutation profile, but whether this discriminates them from thyroid FL and DLBCL is unknown. To investigate this, we have examined 42 EMZL (11 BCL6-translocation [tr]+ve), 21 FL (5 BCL2-tr+ve, 10 BCL6-tr+ve, 1 both BCL2/BCL6-tr+ve) and 34 DLBCL of the thyroid. Targeted next generation sequencing revealed a remarkable overlap in the mutation profile among thyroid EMZL, BCL2-tr-ve FL and DLBCL, all showing frequent mutations in TET2, IGLL5, TNFRSF14, CD274, GNA13, FAS, KLF2 and TNFAIP3. In contrast, BCL2-tr+ve FL of the thyroid showed frequent BCL2, KMT2D, CREBBP, EZH2, but not TET2 and CD274 mutations. Genomic analysis of BCL6 translocation by targeted locus capture next generation sequencing showed different genomic configurations between thyroid FL and EMZL. In thyroid FL, the majority of BCL6 translocations placed its coding exons under the transcriptional control of the IGH switch region super-enhancer or its partner genes, potentially resulting in BCL6 constitutive expression. In contrast, the majority of BCL6 translocations in thyroid EMZL juxtaposed the BCL6 gene to the IGHJ/D region without encompassing the Eμ enhancer or its partner genes in an opposite orientation, thus less likely to lead to constitutive BCL6 transactivation. The above genetic changes likely dysregulate B-cell maturation and peripheral tolerance, thus offer significant molecular insights into the pathogenesis of thyroid lymphomas, particularly underpinning autoimmunity in the lymphomagenesis and potentially explaining the overlap in histopathology between EMZL and FL.

Introduction

Primary thyroid lymphomas commonly originate from a background of chronic lymphocytic thyroiditis (Hashimoto’s thyroiditis, HT) and comprise largely extranodal marginal zone lymphoma of mucosa-associated lymphoid tissue (EMZL), diffuse large B-cell lymphoma (DLBCL), and followed by follicular lymphoma (FL).¹ The histological diagnosis of these thyroid lymphomas can be readily made by integrated histological and immunophenotypic assessment in most cases. However, a high proportion of thyroid FL are negative for BCL2 translocation and lack expression of CD10. These cases often pose significant problems in differential diagnosis from thyroid EMZL that frequently show prominent follicular colonization, hence considerable overlap in their morphological and immunophenotypic presentations.²

We previously showed that thyroid EMZL had a distinct mutation profile characterized by concurrent mutations in TET2, CD274 (PD-L1) and TNFRSF14, distinguishing them from those of other B-cell lymphomas including EMZL at other anatomic sites.^3-5 However, the genetic profile of thyroid FL and DLBCL remains to be unravelled, particularly whether there is any difference in the genetic profile between thyroid FL and EMZL, and if so, whether such genetic differences can help their differential diagnosis. In addition, BCL6 translocation is occasionally seen in EMZL, but seemingly more frequent in those of the thyroid.^6-8 This is rather paradoxical given that BCL6 functions as a master regulator of the germinal center (GC) reaction, preventing GC B-cell exit by repressing the plasma cell differentiation program.^9,10 To address these questions, and to further understand the pathogenesis of thyroid lymphomas, we performed targeted next generation sequencing (NGS) of 172 lymphoma genes and integrated analyses in a large cohort of primary thyroid lymphoma including 42 EMZL, 21 FL and 34 DLBCL. We also investigated the genomic configuration of BCL6 translocation in thyroid FL and EMZL by targeted locus capture NGS in order to understand the paradoxical finding of BCL6 translocation in thyroid EMZL.

Methods

Patients and clinical data

A total of 97 cases of primary thyroid B-cell lymphoma (94 from Ito Hospital, Tokyo) were successfully investigated, including 24 cases of EMZL that had been previously studied.³ The histological diagnosis was reviewed with the help of additional immunohistochemistry (Online Supplementary Table S1) and interphase fluorescence in situ hybridization for BCL2 and BCL6 translocations by expert hematopathologists (AW and ADA). The final diagnosis compromised 42 EMZL (11 with BCL6 translocation), 21 FL (5 with BCL2 translocation, 10 with BCL6 translocation, 1 case with both BCL2 and BCL6 translocations, 5 lacking both BCL2 and BCL6 translocations) and 34 DLBCL (Online Supplementary Table S2).

The use of archival tissues for research was approved by the ethics committees of the institutions involved in the study.

DNA extraction and quality assessment

In each case, tumor rich areas (>30%) were microdissected on formalin-fixed paraffin-embedded (FFPE) tissue slides. DNA was extracted and assessed for quality by polymerase chain reaction (PCR) of variably sized genomic fragments as previously described.¹¹

Mutation analysis by targeted next generation sequencing

The target panel included 172 genes recurrently mutated in MZL, FL and DLBCL (Online Supplementary Table S3). FFPE tissue DNA (100ng) were fragmented using the Covaris E220 Focused Ultrasonicator (Covaris, Brighton, UK). For each DNA sample, an indexed library was prepared using the xGen™ UDI-UMI indexes (IDT, Coralville, IA, USA) and pooled for target enrichment using TWIST probes (TWIST Biosciences, San Francisco, CA, USA). The enriched DNA targets were amplified by PCR and then pooled together for sequencing using the Illumina NextSeq 2000 platform (2x100bp paired-end sequencing protocol). The sequence data analysis, variant calling, and filtering were performed as described in our previous studies.^12,13

For DNA sample with suboptimal quality (PCR amplification of genomic fragment ≤300bp), targeted sequencing was performed in duplicate and only variants detected by both replicates were considered as a true change (Online Supplementary Figure S1).

Mutation signature analysis

This analysis was based on all somatic single base substitutions including synonymous, non-synonymous, splice site and UTR changes using Sigprofiler assignment.¹⁴

Genomic analysis of chromosome translocation by next generation sequencing

The genomic configuration of BCL6 translocation was investigated by targeted locus capture next generation sequencing (TLC-NGS) as described previously and a customized NGS panel (Online Supplementary Methods and Online Supplementary Table S4).¹⁵ The IGH genomic segments including super enhancer positions were according to those described previously using the hg19 assembly.¹⁶

Re-analysis of single-cell RNA-sequencing data from mice germinal center B cells

TET2 inactivation by mutation may cause genome-wide hypermethylation, hence potentially impact on gene expression in B cells and their phenotype. To explore this, we revisited the single cell sequencing data from a previous mouse study.¹⁷

Germinal center B cells sorted from the spleens of tumor-bearing Tet2-deficient (MxTR) and wild-type (MxWT) mice were used for library preparation for RNA sequencing (RNA-seq: MxRT, N=5; MxWT, N=4) and whole-genome bisulfite sequencing (WGBS: MxTR, N=2; MxWT, N=2). Sequencing data were reanalyzed as previously described.¹⁷ Differential expression data from RNA-seq were analyzed using DESeq2.¹⁸ The methylation rate of each base by WGBS was visualized in the UCSC Genome Browser.

Statistical analysis

The comparison of mutation load and gene expression among different groups was assessed using Wilcoxon Rank-Sum test with two-sided P values. Association between mutation frequencies and clinical parameters among lymphoma groups and their subsets was examined using the Fisher exact test. Mutation association was analyzed using the somaticInteractions function in the Maftools R Bioconductor package.¹⁹

Results

Histopathological and immunophenotypic features of thyroid lymphoma

In comparison with EMZL of other sites, thyroid cases showed prominent follicular colonizations, albeit to a variable extent in some cases. Among the 40 cases of thyroid EMZL with complete CD21, CD10 and BCL6 immunohistochemistry data, 38 showed extensive follicular colonization, predominantly type 1 (poorly defined follicle/mantle zone) or type 2 (follicle and mantle zone variably preserved) pattern in 24 and 14 cases, respectively.²⁰ Irrespective of different follicular colonization patterns, 30 cases showed virtually absent CD10 expression in all or most colonized follicles, albeit variable BCL6 staining (Figure 1). The remaining 8 cases showed a mixed pattern with some of the colonized follicles displaying partial CD10 positivity consistent with the presence of residual GC B cells.

With the exception of one case (grade 3B), all other thyroid FL were grade 3A (Figure 2C), and all showed predominantly a nodular growth pattern, and also prominent lymphoepithelial lesions including “MALT balls” (Figure 2). Like classic FL, BCL2-translocation [tr]+ve thyroid FL expressed both CD10 and BCL6, albeit a single case with both BCL2 and BCL6 translocations was CD10 negative (Figure 2). Of the 15 BCL2-tr-ve FL, only 6 showed CD10 expression, with the remaining cases negative for CD10 including 7 and 2 with and without BCL6 translocation, respectively (Figure 2B, C). The FL diagnosis in these cases, including those negative for both BCL2 and BCL6 translocations, was supported by observation of expression of other GC markers (BCL6, HGAL, LMO2, GCET1, STMN1) in neoplastic B cells within and outside B-cell follicles (Figure 2B, Online Supplementary Figure S2). In contrast, none of these GC markers was positive in the neoplastic B cells outside B-cell follicles in 21 cases of thyroid EMZL examined.

Figure 1.Histopathology of a representative primary thyroid extranodal marginal zone lymphoma of mucosa-associated lymphoid tissue. (A) A thyroid extranodal marginal zone lymphoma of mucosa-associated lymphoid tissue (EMZL) without BCL6 translocation (case EMZL-37) shows diffuse infiltration of neoplastic B cells with prominent lymphoepithelial lesions including “MALT ball”, and also extensive follicular colonization. CD10 immunohistochemistry highlights residual germinal center (GC) B cells, while neoplastic B cells in the colonized follicle are CD10 negative. BCL6 positivity likely identifies both residual GC B cells and the colonized neoplastic B cells. (B) A thyroid EMZL with BCL6 translocation (case EMZL-04) displays diffuse infiltration of neoplastic B cells with prominent lymphoepithelial lesions. CD21 immunohistochemistry highlights the colonized follicles that are virtually negative for CD10 but show variable BCL6 expression, indicating total overrun by neoplastic B cells. tr+ve: translocation positive, tr-ve: translocation negative; LEL: lymphoepithelial lesion. Magnification 40x for the 2 images that show LEL; magnification 4x for all other low power images.

Figure 2.Histopathology of representative primary thyroid follicular lymphoma. (A) A BCL2 translocation positive follicular lymphoma (FL) (case FL-05) shows follicular growth pattern and extensive lymphoepithelial lesions (LEL) including “MALT ball” in the interfollicular region (IF). The neoplastic cells including those involved in the LEL are strongly positive for CD10 and BCL2. Magnification: top, left = 2x; top, middle and right = 40x. Bottom = 4x (insert, LEL = 40x). (B) A BCL2 and BCL6 translocation negative FL (case FL-19) displays vaguely follicular growth pattern with little involvement of the IF. The neoplastic B cells in both the follicle center (FC) and IF are positive for CD10, BCL6 and HGAL. Magnification: top, left = 2x; top, middle = 40x; all other panels = 4x (insert, LEL = 40x). (C) Summary of immunophenotype of primary thyroid FL. CD10 is more often negative in BCL2 translocation negative FL including those with BCL6 translocation. *FL-04 showed no apparent neoplastic involvement in the IF. IHC: immunohistochemistry; NA: not available; STMN1: Stathmin 1; tr+ve: translocation positive; tr-ve: translocation negative.

Thyroid DLBCL were unremarkable; all were negative for BCL2 translocation and 4 had BCL6 translocation. Of the 34 cases investigated, 27 and 17 showed BCL6 and CD10 expression by immunohistochemistry, respectively.

Mutation signature among different thyroid B-cell lymphomas

All somatic variants including synonymous, non-synonymous, indels and UTR changes were included in the mutation load analysis (Online Supplementary Table S5). Overall, the mutation load was significantly higher in DLBCL than FL (P=0.006) and EMZL (P=0.005), but there was no significant difference between the two latter groups (Figure 3A). As many of the genes (N=37) investigated are targets of the somatic hypermutation (SHM) process, a separate analysis of mutations in these genes was performed, and this showed a similar trend, with DLBCL cases having a significantly higher mutation load in SHM targets than FL (P=0.021) and EMZL (P=0.014) (Online Supplementary Figure S3).

We next compared the mutation signature among the three lymphoma groups by combining all the somatic single base substitutions of each group together as none of the individual cases had a sufficient number of variants for such analysis. As expected, the cytidine deaminase (AID) associated mutation signature SBS84 was similarly found among the three lymphoma groups (Online Supplementary Figure S3).²¹ Similarly, the clock-like/age-related mutation signature SBS5 was also high among the three lymphoma groups.²¹

Mutation profile among different thyroid B-cell lymphomas

The pathogenic mutations were compared among the three lymphoma groups with subset analysis according to translocation status.

Remarkably, there was a considerable overlap in the overall mutation profile among thyroid EMZL, BCL2-tr-ve FL and DLBCL, with frequent mutations in TET2, IGLL5, TNFRSF14, CD274, GNA13, KLF2, TNFAIP3 and FAS (Figures 3A and 4A, Online Supplementary Figures S4 and S5). Such an overlap in mutation profile among the three groups was also seen in the genes not targeted by the SHM process (Figure 4A, Online Supplementary Figure S4). Apart from a few genes, there was no significant difference in the mutation frequencies of the vast majority of genes between EMZL and BCL2-tr-ve FL, nor between BCL2-tr-ve FL and DLBCL (Figures 3B and 4A). Similarly, there was no significant difference in the mutation profile between BCL2-tr-ve FL with and without BCL6 translocation, nor between EMZL with and without BCL6 translocation (Figure 3B, Online Supplementary Figure S6). Nonetheless, thyroid DLBCL showed a significantly higher incidence of mutation than thyroid EMZL in several genes (SGK1, KLHL6, B2M, EBF1, SOCS1, FOXO1, P2RY8, DUSP2, KRAS, CD58) (Figures 3B and 4A).

In contrast, BCL2-tr+ve FL of the thyroid showed frequent BCL2, KMT2D, CREBBP, EZH2, TNFRSF14, but not TET2, CD274 and IGLL5 mutations (Figures 3B and 4A), similar to the mutation profile of nodal FL.²² Nonetheless, all these thyroid cases were associated with HT as evidenced by clinical, laboratory and histological investigations, with 4/5 cases at early stage (IE=2, IIE=2), supporting their thyroid origin (Online Supplementary Table S2).

Analysis of co-occurrence of the mutations revealed a number of significant associations and mutual exclusions (Figure 4B). For example, TET2 mutation was significantly associated with changes in CD274, TNFAIP3, TNFRSF14 and IGLL5, and CD274 mutation was further significantly associated with changes in TNFRSF14, PABPC1, TNFAIP3, FAS and KLF2, but mutually exclusive from changes in KLHL6 and B2M (Figure 4B).

Distinct genomic configuration of BCL6 translocation between thyroid follicular lymphoma and extranodal marginal zone lymphoma of mucosa-associated lymphoid tissue

The genomic configuration of BCL6 translocation in 8 FL and 10 EMZL were investigated by TLC-NGS, and a further 2 FL were studied by a customized NGS panel (Figures 5 and 6).

Among the 10 FL with BCL6 translocation, 3 showed genomic fusion between the BCL6 intron 1 and the centromeric IGHA1 region including 3’regulatory region 1 (3’RR1) super-enhancer in the same orientation, and this genomic configuration may cause BCL6 transactivation due to super-enhancer hijack²³ (Figure 5). A further 5 cases displayed genomic fusion between BCL6 intron 1 or its upstream 5’UTR region and either the 5’ regions of partner genes (HSP90AA1, HMGA1, ZBTB38) or the 3’UTR of MIR29A in the same orientation, and these genomic configurations represented a promoter substitution thus placing BCL6 under the transcriptional control of its translocation partner. In the remaining 2 cases, one showed fusion between the BCL6 intron 1 and the centromeric IGHJ6 region in the same orientation, and the juxtaposed IGHJ region did not contain the Eµ super enhancer, thus lacking the apparent mechanism to transactivate BCL6 expression. In the other case, the translocation placed an upstream region of the BCL6 gene between the IGH and IGK, and the translocation did not involve any BCL6 coding exons, thus unlikely affecting BCL6 expression (Figure 5).

Among the 10 BCL6-tr+ve EMZL, 5 cases showed genomic fusion between BCL6 intron-1 or its upstream 5’UTR region, and the centromeric region of IGHJ6 or D2-8 in the same orientation (Figure 6). As the translocated IGHJ and D2-8 region did not contain the Eµ super enhancer, these translocations lacked any apparent mechanism to drive constitutive BCL6 expression. A further 2 cases displayed genomic fusion between the upstream of the BCL6 5’UTR region and its partner gene (upstream of GRHPR or upstream of IGLL5) in an opposite orientation, with no obvious evidence of super-enhancer hijack that may drive constitutive BCL6 transactivation (Figure 6). In the remaining 3 cases, the translocation fused the BCL6 5’UTR or intron 1 to different partner genes in the same orientation (intron 2 of SGK1 and GAPDH and intron 1 of SMC4), with the latter one potentially representing a promoter substitution.

Figure 3.Comparison of mutation frequencies among different primary thyroid lymphomas. (A) Comparison of mutation frequencies among thyroid extranodal marginal zone lymphoma of mucosa-associated lymphoid tissue (EMZL), follicular lymphoma (FL) and diffuse large B-cell lymphoma (DLBCL). Only genes showing a frequency of potentially pathogenic mutation above 10% in any of these lymphoma groups are included, with the exception of BCL2, of which all variants are included. (B) Comparison of mutation frequency between subsets of thyroid EMZL, FL and DLBCL as indicated. In each comparison, all genes showing a mutation frequency above 10% in any of the two subsets included, with the exception of BCL2, of which all variants are included to highlight different mutation profiles among various comparisons. *Significant changes between subsets (P≤0.05). tr+ve: translocation positive; tr-ve: translocation negative.

Figure 4.Mutation profiles among different primary thyroid lymphoma. (A) Heatmap illustration of mutation profile among thyroid extranodal marginal zone lymphoma of mucosa-associated lymphoid tissue (EMZL), follicular lymphoma (FL) and diffuse large B-cell lymphoma (DLBCL). (See Online Supplementary Figure S4 for complete data illustration.) Only genes showing a frequency of potentially pathogenic mutation above 10% in any of these lymphoma groups are included in the data presentation, with the exception of BCL2, of which all variants are included to allow informed comparison among subgroups. *Known and **predicted somatic hypermutation (SHM) targets, respectively. (B) Correlation analysis of mutation seen in top 20 most frequently mutated genes in thyroid EMZL, FL and DLBCL. •Denotes differences which were statistically significant (P<0.05). NA: not available; SNV: single nucleotide variation; tr+ve: translocation positive; tr-ve: translocation negative.

Figure 5.Genomic configuration of BCL6 translocation in primary thyroid follicular lymphoma. Genomic breakpoint sequencing analysis was performed by targeted locus capture-based next generation sequencing (TLC-NGS) or a customized NGS panel with sequence annotations based on human genome (hg19) together with IGH super enhancers according to Mikulasova et al.¹⁶ cen: centromeric; telomeric.

In addition, TLC-NGS also identified additional chromosome translocations in 4 cases of the above BCL6-tr+ve thyroid lymphomas, including 3 FL and one EMZL (Online Supplementary Figure S7). One of these FL showed IGH::BCL2 and the other 2 revealed an IGH involved translocation, one in association with BTG2 but not involving its coding exons, and the remaining one in association with IGK. The EMZL case showed IGH::SOX5, with genomic configuration potentially driving S OX 5 expression (Online Supplementary Figure S7).

Potential impact of genetic changes on follicular lymphoma phenotype

As many of the genes frequently altered in primary thyroid lymphoma are critical for the GC B-cell development, these genetic changes may potentially affect the maturation process of GC B cells, hence determining their differentiation stages that undergo malignant transformation. To investigate this, we first examined the expression pattern of TET2, CD274, TNFRSF14, BCL6 and MME (CD10) in various normal GC B-cell subsets using the single cell transcriptomic data from human tonsils.²⁴ Interestingly, these genes showed a similar trend of expression across different GC B-cell subsets, being abundantly expressed in proliferating dark zone (DZ) GC B cells, reaching the highest expression level in DZ non-proliferating GC B cells and DZ to LZ (light zone) transition cells, then decreasing to the lowest level in LZ proliferating and LZ-to-DZ transition GC B-cell subsets (Figure 7).

The expression of the above genes must be tightly regulated, and their coordinated alterations are deemed critical in GC B-cell maturation and peripheral tolerance. As TET2 mutation in primary thyroid lymphoma is an early event,^3,5 its inactivation by mutations may cause genome-wide hypermethylation, thus potentially affect the above gene expression. To explore this, we examined this gene expression in Tet 2 deficient mouse GC B cells from a previous study.¹⁷ In comparison with the wild-type mouse GC B cells, the Tet2 deficient mouse GC B cells showed a higher level of methylation in Irf4 (promoter), Cd274 (gene body, introns) and Mme (CD10) (CpG site), and a reduced expression of these genes (Figure 8). However, there was no significant alteration in the methylation profile and expression of Tnfrsf14 in the Tet 2 deficient mouse GC B cells.

Correlation analysis of clinical, pathological and genetic data

A history of HT was present at a similar rate among thyroid EMZL (38/40; 95%), FL (17/21; 81%) and DLBCL (28/34; 82%) (Online Supplementary Table S2). The average duration of HT was significantly higher in BCL6-tr-ve EMZL than the BCL6-tr+ve EMZL (P=0.008). Apart from that the DLBCL group showed a significant higher level of serum LDH than EMZL (P=0.005), and elevated sIL-2R than both FL (P=0.005) and EMZL (P=0.0000022), there were no significant differences among the standard clinical and laboratory parameters including TgAb and TPOAb among the three lymphoma groups.

Discussion

By investigating the genetic changes of primary thyroid EMZL, FL and DLBCL, we have unravelled several novel observations. 1) Apart from BCL2-tr+ve FL, there is a remarkable overlap in the mutation profile among BCL2-tr-ve FL, EMZL, and DLBCL of the thyroid. 2) The core of these commonly mutated genes involves TET2, TNFRSF14 and CD274, distinguishing them from other B-cell lymphomas, including those at other mucosal sites. 3) BCL6 translocation is recurrently seen in both thyroid FL and EMZL, but there is a fundamental difference in its genomic configuration between the two lymphoma entities. As many of these frequent genetic changes target genes that regulate B-cell maturation and tolerance in the peripheral lymphoid tissues, the above novel findings offer significant molecular insights into the pathogenesis of these thyroid lymphomas. Like nodal FL, thyroid BCL2-tr+ve FL showed frequent mutations in BCL2, TNFRSF14, KMT2D, CREBBP and EZH2. Intriguingly, none of these thyroid BCL2-tr+ve FL showed TET2 and CD274 mutations. It is important to note that these BCL2-tr+ve FL were associated with HT, and mainly at early clinical stages, not secondary to a systemic nodal disease. As BCL2 translocation is an early event in FL development, occurring at the pre-B cell stage due to erroneous IGH VDJ recombination, it is pertinent to speculate that the aberrant BCL2 overexpression caused by the translocation may favor the subsequent genetic changes with cooperative oncogenic activities, thus giving rise to a unique mutation pattern seen in all IGH::BCL2 positive neoplasms, including in situ follicular B-cell neoplasm, FL and transformed FL.^12,25-29

The remarkable overlap in the mutation profile (TET2, IGLL5, TNFRSF14, CD274, GNA13, KLF2, TNFAIP3, FAS) (Figure 3A) among BCL2-tr-ve FL, EMZL and DLBCL of the thyroid is intriguing, but biologically plausible. As these lymphomas originate from a common background of HT, their development most likely involves the same pathological process of autoimmunity that underpins HT. HT is characterized by development of autoreactive B cells, and histologically extensive GC reactions, a process critically depending on T-helper cells. The molecular mechanisms that underpin the breach of peripheral tolerance may drive the evolution of autoreactive B cells and their clonal expansion. As B cells undergo reiterative GC reactions and relentless exposure to SHM activities, this may lead to acquisition of genetic abnormalities and eventual malignant transformation. In this context, the occurrence of malignant B-cell clone can be viewed as a part of the overall autoimmune process, but as an extreme presentation. Thus, the overlapping mutation profile among BCL2-tr-ve FL, EMZL and DLBCL of the thyroid is likely driven by their common etiology, i.e., the pivotal pathological process associated with HT.

Figure 6.Genomic configuration of BCL6 translocation in primary thyroid extranodal marginal zone lymphoma of mucosa-associated lymphoid tissue. Genomic breakpoint sequencing analyses was performed by targeted locus capture-based next generation sequencing (TLC-NGS) with sequence annotations based on human genome (hg19) together with IGH super enhancers according to Mikulasova et al.16 cen: centromeric; telomeric.

Figure 7.Re-analyses of single-cell RNA sequencing data of germinal centre (GC) B cells from the tonsil cell atlas project.²⁴ UMAP illustration shows the overall distribution of the indicated genes among various subsets of germinal center B cells (GC B cells). Violin plots with overlaid box plots compare the expression level of the indicated genes (TET2, CD274, TNFRSF14, BCL6, MME [CD10], and CD40) among various subsets of GC B cells. DZ: dark zone; LZ: light zone; MBC: memory B cells; PC: plasma cells.

In support of the above speculation, several genes among those frequently mutated in thyroid lymphoma are involved in the governance of peripheral tolerance. CD274 (PD-L1) and TNFRSF14 encode ligands for co-inhibitory surface receptor PD1 and BTLA on T-helper cells, respectively, and their inactivation by somatic mutation/deletion most likely constrain their negative regulation on T-helper cell function. PD-L1 deficiency or impairing PD1/PD-L1 interaction can cause autoimmune disorders in animal models.^30,31 Similarly, immune checkpoint inhibitors, particularly anti-PD1 antibody, frequently cause cancer patients to develop autoimmune thyroiditis and/or autoantibodies including anti-thyroperoxidase and anti-thyroglobulin.^32,33 Moreover, patients with germline PD1 or CD274 (PD-L1) loss-of-function mutations show early-onset endocrine autoimmunity.^34,35 Cd274 deficiency in B cells appears to promote T-helper cell function, consequently enabling B cells a growth advantage.^36,37 Similarly, Tnfrsf14 deficient B cells gain an enhanced growth advantage due to exaggerated T-cell help through increased CD40/CD40L co-stimulation.^38,39 The concurrent inactivation of both CD274 (PD-L1) and TNFRSF14 in thyroid lymphomas may synergise in their enhancement of T-helper cell function, thus breaching peripheral tolerance and providing growth advantage to autoreactive B cells.

The above lymphomagenic process is likely augmented by the concurrent loss-of-function mutations in TNFAIP3, KLF2 and FAS. TNFAIP3 encodes a global negative regulator of the canonical NF-βB activation pathway and attenuates signaling transduction from several surface receptors, including BCR, TNFR, TLR and IL1βR, by inactivating their downstream signaling molecules through its ubiquitin editing activities.⁴⁰ Similarly, KLF2 encodes a transcription factor/repressor that can negatively regulate NF-βB and NOTCH2 activities triggered by a range of surface receptor signaling.⁴¹ FAS encodes a surface receptor for apoptosis signaling and mutant FAS can act as a dominant-negative, preventing FAS interaction with its downstream adaptor molecule FADD to mediate apoptosis.⁴² Taken together, TNFAIP3, KLF2 and FAS inactivation may promote the apoptosis evasion of autoreactive B cells and their clonal expansion.⁴³

The above mutations may also have a major impact on the histological presentation of lymphoma. Deficiency of Tnfrsf14 or Cd274 alone increases GC B-cell competitiveness in mice.^36,38 A combined inactivation of both CD274 (PD-L1) and TNFRSF14 in B cells may significantly enhance their entry, expansion and retention in B-cell follicles due to enhanced T-helper cell signals.³¹ These biological impacts may underpin the follicular growth pattern of thyroid FL as well as the extensive follicular colonization of thyroid EMZL. In this context, it is pertinent to point out that follicular colonization in thyroid EMZL is the most prominent among EMZL of various sites, often involving most or the majority of B-cell follicles, thus posing significant difficulty in differential diagnosis from BCL2-tr-ve FL.² In a similar context, pediatric-type follicular lymphoma (PTFL) and pediatric nodal marginal zone lymphoma (PNMZL) also exhibit substantial overlap in their clinical, morphological and genetic features.^44,45 The shared genetic changes between the two lymphomas have been used to argue for considering these different lymphomas to be a single biological entity with variations in their histological presentations.^44,45 The somatic genetic alterations and tumor microenvironmental factors that influence GC entry (for example T-cell dependent B-cell activation) and GC exit (via BCL6 downregulation) are likely to have a major impact on the phenotypic presentation of these lymphomas. However, studies so far have largely focused on mutations within coding gene sequences, while other potentially relevant genetic changes, such as somatic mutations in BCL6 transcriptional regulatory (non-coding) regions that could affect GC exit,^46,47 have still to be explored. Thus, the currently available data are not sufficient to support considering these lymphomas as a single biological entity.

The finding of recurrent BCL6 translocation in EMZL is rather paradoxical as BCL6 is a master transcriptional factor that orchestrates the GC reaction while repressing the plasma cell differentiation program,^9,10 hence maintaining the GC phenotype. Our finding of significant differences in the genomic configuration of BCL6 translocation between thyroid FL and EMZL provides a plausible explanation. In thyroid FL, the majority of BCL6 translocations place the BCL6 coding exons under the transcriptional control of the IGH switch region super enhancer (3’RR1) or its partner genes, known as super enhancer hijack and promoter substitution, respectively. Among the 5 cases with BCL6 translocation genomic configuration showing promoter substitution, all the partner genes (HSP90AA1, HMGA1, ZBTB38, MIR29A) appear to be highly expressed in B cells.^48-53 These genomic configurations most likely enable BCL6 constitutive expression. In contrast, the majority of BCL6 translocations in thyroid EMZL juxtapose the BCL6 gene to the IGHJ/IGHD region without encompassing the Eµ enhancer or partner genes in an opposite orientation. In the 3 cases of thyroid EMZL where BCL6 is placed in the same orientation with its partner gene, 2 cases involved coding exons of the partner gene (GAPDH, SGK1), potentially resulting in chimeric transcripts and possible protein products with unknown function; the remaining case with the partner gene SMC4 may represent a true promoter substitution, but SMC4 appears not to be highly expressed in B cells.^49,50 Overall, BCL6 translocation in thyroid EMZL lacks the apparent genomic configuration that enables BCL6 constitutive expression, hence minimal impact on the lymphoma phenotypic presentation.

Figure 8.Epigenomic and transcriptomic alterations in germinal center B cells from Tet2-deficient and wild-type mice. (A) Visualizations of base-resolution DNA methylation profiles at representative gene loci (Mme, Cd274, Irf4, and Tnfrsf14) derived from whole-genome bisulfite sequencing (WGBS) of germinal center (GC) B cells sorted from the spleens of Tet 2 -deficient (Tet2^-/-) and wild-type (Tet2^wt) mice. The y-axis represents base-resolution methylation levels derived from WGBS. Differentially methylated regions (q-value ≤ 0.25 and |methylation difference| > 30%) are highlighted with red squares. (B) Normalized expression values of the indicated genes based on DESeq2 analysis of RNA-sequencing data (Tet2^-/-, N=5; Tet2^wt, N=4).^17,18

TET2, CD274 and TNFRSF14 inactivation may also affect the GC B-cell maturation process, potentially influencing their differentiation stage that undergoes malignant transformation, and hence the neoplastic cell immunophenotype. As shown in tonsils, during normal GC reaction, the downregulation of CD274 and TNFRSF14 expression in LZ centrocytes may dampen their inhibitory effect on T-follicular helper cells, allowing them to receive more T-cell help, preparing for GC exit or re-entry to DZ for further SHM. Tet2 deficiency increased the proportion of GC centrocytes and impaired GC exit.⁵⁴ Although the molecular mechanism of how Tet 2 deficiency dysregulates the GC reaction is not fully understood, Tet 2 deficient mouse GC B cells showed an increased level of DNA methylation in Cd274, Mme (CD10) and Irf4, accompanied by their reduced expression in comparison with wild-type controls.¹⁷ This, together with the inactivation of both CD274 and TNFRSF14, may promote the GC B-cell maturation process toward LZ centrocytes and memory B-cell differentiation. On the other hand, BCL6 constitutive expression by translocation and/or other mechanisms may prevent the late GC B cells from GC exit by repressing transcription factors (BLIMP1) that are critical for plasma cell differentiation.^9,10 Taken together, it is conceivable that these genetic changes may cooperate in their oncogenic activities, promoting the GC B-cell differentiation process but impairing their GC exit, thus favoring malignant transformation of the late GC B-cell subsets. This may potentially explain why a high proportion of BCL2-tr-ve FL of the thyroid is CD10 negative.

In conclusion, our findings offer significant molecular insights into the pathogenesis of thyroid lymphomas, bridging autoimmunity, somatic genetic changes and lymphomagenesis together, and potentially explain the overlap of histological presentations between thyroid EMZL and FL.

Footnotes

• Received May 28, 2025
• Accepted March 6, 2026

Correspondence

Ming-Qing Du mqd20@cam.ac.uk

Disclosures

ES and HF are employees of Cergentis BV, the company that developed the Targeted Locus Capture (TLC) technology used in this study. All the other authors have no conflicts of interest to disclose.

Contributions

M-MT is responsible for FISH, targeted NGS, immunohistochemistry data collection and analyses, and prepared and wrote the manuscript; M-QD is responsible for FISH, targeted NGS, immunohistochemistry data collection and analyses, case contribution and pathology, study conception, coordination and research funding, and prepared and wrote the manuscript; NW is responsible for study conception, coordination and research funding, case contribution and pathology; ES, HF, ZC, AG, FW and TM are responsible for FISH, targeted NGS, immunohistochemistry data collection and analyses; MF and MS-Y are responsible for mouse experimental data; KS, KI, LR-B, ADA and AW are responsible for case contribution and pathology. All authors contributed to manuscript preparation and writing, and read and approved the final version of the manuscript for publication.

Funding

The study was supported by research grants from Blood Cancer UK (22011, 24011, 19011, 19010), Cancer Research UK (CRCBPA-Nov23/100001, C8333/A29707) and an International Collaborative Award from the Pathological Society of Great Britain and Ireland, UK (ICA 1019 01). M-MT was supported by a BBSRC DTP PhD studentship (BBSRC BB/M011194/1). MF was supported by Gilead’s Research Scholars Program and The Leukemia & Lymphoma Society (3442-25). The Human Research Tissue Bank of the Cambridge University Hospitals NHS Foundation Trust is supported by the NIHR Cambridge Biomedical Research Centre.

Acknowledgments

The authors would like to thank Dr. Ana Toribio and Jessica Ferreira Gouveia for their assistance with Illumina sequencing.