Germinal center‐derived lymphomas: The darkest side of humoral immunity

Summary One of the unusual features of germinal center (GC) B cells is that they manifest many hallmarks of cancer cells. Accordingly, most B‐cell neoplasms originate from the GC reaction, and characteristically display abundant point mutations, structural genomic lesions, and clonal diversity from the genetic and epigenetic standpoints. The dominant biological theme of GC‐derived lymphomas is mutation of genes involved in epigenetic regulation and immune receptor signaling, which come into play at critical transitional stages of the GC reaction. Hence, mechanistic studies of these mutations reveal fundamental insight into the biology of the normal and malignant GC B cell. The BCL6 transcription factor plays a central role in establishing the GC phenotype in B cells, and most lymphomas are dependent on BCL6 to maintain survival, proliferation, and perhaps immune evasion. Many lymphoma mutations have the commonality of enhancing the oncogenic functions of BCL6, or overcoming some of its tumor suppressive effects. Herein, we discuss how unique features of the GC reaction create vulnerabilities that select for particular lymphoma mutations. We examine the interplay between epigenetic programming, metabolism, signaling, and immune regulatory mechanisms in lymphoma, and discuss how these are leading to novel precision therapy strategies to treat lymphoma patients.

of high-affinity light zone GC B cells recycle back to the dark zone for additional rounds of somatic hypermutation and proliferation, whereas the majority of low-affinity GC B cells undergo apoptosis.
The GC B cell is at a particularly high risk for undergoing malignant transformation, due to attenuation of certain DNA damage and cell proliferation checkpoints, which is essential for immunoglobulin affinity maturation. Although the GC reaction is tightly regulated, somatic hypermutation can disrupt this delicate equilibrium by generating off-target mutations that enable B cells to gain selective advantages. Along these lines, a majority of healthy individuals are believed to harbor premalignant clonal populations of mutant B cells, although at the current time there is no way to identify who is at risk for transforming into overt disease.
Germinal center-derived lymphomas are markedly heterogeneous, as befitting a tissue of origin that naturally undergoes rapid clonal diversification. They are among tumors with the highest burden of somatic mutations, and physically manifest in very diverse manners.
Because of this, the classification schemes for these tumors have been in constant evolution, as improving technologies allow deeper insight into their genetic and epigenetic features. This, combined with major advances in understanding GC biology, including at the epigenetic, metabolic, signaling, and immune synapse levels, have revealed how intricate GC regulatory processes can be hijacked in a multitude of ways to facilitate lymphomagenesis. Importantly, these discoveries point toward novel therapeutic paradigms for controlling and even curing lymphomas that were formerly refractory to existing treatments.

| G C B CELL S INHERENTLY FE ATURE HALLMARK S OF MALIG NANT TR ANS FORMATION
B cells transiting the GC reaction manifest phenotypic features that mimic many of the canonical biological hallmarks of cancer. 13 Some of these include: • Massive proliferation and clonal expansion: Licensing of cell proliferation is initiated by transient induction of MYC early in the GC reaction, as well as upon receiving strong T FH cell help in the light zone for dark zone reentry and clonal expansion. 14,15 In the GC dark zone, proliferation is maintained through the checkpoint suppressing actions of the transcriptional repressors EZH2 and BCL6, and through cyclin D3 activation downstream of GSK3. [16][17][18][19][20] • Inactivation of tumor suppressors: To allow cell cycle progression despite high levels of stress, GC B cells downregulate tumor suppressors such as TP53 as well as cell cycle checkpoint genes (eg, CDKN1A, CDKN1B). 18,21 • Genome instability: Somatic hypermutation and class-switch recombination are mediated by activation-induced cytidine deaminase (AICDA). 22 The error-prone DNA polymerase eta (Polη) additionally introduces DNA point mutations when repairing AICDA-induced lesions. 23 • Resistance to DNA damage: To facilitate somatic hypermutation, GC B cells downregulate major DNA damage sensor and response proteins including ATR, CHEK1, and TP53. 24,25 Dampening of DNA damage checkpoints combined with AICDA activity puts cells at high risk of accumulating off-target genetic alterations.

| GC entry
Upon antigen encounter, naive B cells move to the T-B border of the follicle to interact with CD4+ T cells. There, the duration of their interaction depends on their specificity and affinity for the encountered antigen. 36 Resulting co-stimulatory signals induce B-cell proliferation at the outer B cell follicle and then migration to the center of the follicle to form a nascent GC. 37 Along these lines, genetic lesions that occur prior to the GC reaction, such as those affecting TET2 (arising in hematopoietic stem cells) or BCL2 (in pre-B cells), may confer preferential initial expansion and survival of mutant cells, resulting in an expanded population of GC B cells at risk for acquiring a "second hit." [38][39][40]

| Dark zone to light zone transition
Germinal center B cells move to the light zone after undergoing a defined number of cell divisions ranging from 1 to 6, depending on several factors including BCR affinity for antigen. 41,42 Aberrant retention of B cells in the dark zone proliferative stage of development would be expected to foster malignant transformation and an aggressive phenotype. This situation is best represented by BL, which can manifest a gene expression profile similar to dark zone GC B cells. 11 It is likely that the characteristic MYC translocation occurring in these tumors enables sustained proliferation due at least in part to enhanced metabolic sufficiency.

| Cyclic reentry to the dark zone
About 10%-30% of B cells possess sufficient antigen affinity to evade apoptosis but yet do not differentiate into plasma cells or memory B cells and instead reenter the dark zone for further rounds of somatic hypermutation and proliferation ( Figure 1). 14,15,[42][43][44][45] Cooperation of CD40 and BCR signaling activated in an affinitydependent manner leads to the induction of MYC 46 and mTORC1. 47 Transient MYC-and mTORC1-activation enables light zone cells to become anabolic and generate the pools of metabolic precursors that will be required for proliferation in the dark zone. Normally, this process is strictly compartmentalized so that GC light zone cells do not proliferate while they undergo MYC/mTORC1-dependent anabolic charging ( Figure 1). In contrast, in the GC dark zone MYC is silenced and mTORC1 activity reduced, which may limit the ability of rapidly dividing cells to undergo continuous replication. In part, this separation of anabolism and proliferation is due to the actions of the transcriptional repressor BCL6, which simultaneously silences MYC and multiple cell cycle checkpoint genes. 48 Any loss of this MYC-BCL6 mutual exclusivity is inherently dangerous and could unleash unlimited proliferation, as occurs in DLBCL and BL.

| GC exit to plasma cell differentiation
Commitment to the plasma cell fate is associated with highly stable B-T FH contacts and requires robust CD40 signaling and transcriptional reprogramming. 49 Disruption of this process can favor the accumulation of proliferating plasmablastic cells. This effect appears to underlie the tumorigenic effect of mutations that affect the transcription factor PRDM1 (BLIMP1), which is the master regulator of plasma cell differentiation. 50 Translocations that induce constitutive expression of BCL6 may also lead to aberrant repression of PRDM1. 35,51 PRDM1 loss occurs almost exclusively in patients with ABC-DLBCLs, many of which manifest a plasmablastic transcriptional profile ( Figure 1).

| THE G ENE TI C BA S IS OF G C-DERIVED B-CELL LYMPHOMA S
Germinal center-derived B-cell lymphomas have been defined (DLBCL, FL, and BL) based on histopathology and gene expression profiles. 7-10 DLBCL has been further classified into the GCB-and ABC-DLBCL subtypes based on gene expression. 9 An additional primary mediastinal subtype of DLBCL tends to occur in younger individuals with a bias toward females, and features strong NF-κB signaling signatures. 56 However, in the genome-sequencing era it has become evident that these diseases are far more intricate.
Indeed, GC-derived lymphomas are among the most genetically complex and heterogeneous of all tumor types. It is estimated that there are ~150 highly recurrently mutated genes (defined as occurring in >5% of patients) in DLBCL. 57 There is also a heavy burden of copy number and structural variations in these tumors. 58 All of this is reasonable to expect, given the inherent mutability of their cellof-origin. Strikingly, more than 50% of genes mutated in DLBCL are transcription factors or chromatin modifiers. 57,59 There are frequent mutations of immune signaling pathway genes including antigen presentation, BCR signaling, NF-κB signaling, PI3K, Toll-like receptor (TLR), and NOTCH signaling. Table 1 provides an integrated and curated list of genes that are recurrently mutated in BL, FL, and DLBCL subtypes; Table 2 gives an overview of lymphoma mutated genes, for which mechanistic studies have yielded insight into their role in the GC reaction and lymphomagenesis.  TA B L E 1 (Continued) Germinal center (GC)-derived Diseases and their subtypes are distinguished by colors: BL, Burkitt lymphoma in dark green; FL, follicular lymphoma in light orange; followed by DLBCL, diffuse large B-cell lymphoma subtypes: GCB, germinal center B-like in dark blue; ABC, activated B cell-like in red; the mostly GCB-related C3/EZB and C4 clusters in medium and light blue respectively; the mostly-ABC-related C1/BN2, C5/MCD and N1 clusters in pink, purple and orange respectively; and the instability cluster C2 in green. a Gain-of-function lesions (GOF, red) can result from copy number gains, translocations or mutations. Loss-of-function lesions (LOF, blue) can result from copy number losses, mutations or DNA methylation.
Recent efforts have attempted to improve the molecular classification of DLBCL through integrative analysis of their genomic profiles. 58,81 These studies included recurring mutations, copy number variations, and chromosomal rearrangements ( is overexpressed. 83 The following sections illustrate some of the biological mechanisms and therapeutic vulnerabilities induced by lymphomaassociated somatic mutations.

| BCLAND CONTROL OF THE G C PHENOT YPE
Any discussion of GC mutations and lymphomagenesis requires special consideration of the BCL6 transcriptional repressor, a critical master regulator of the GC phenotype. Expression of BCL6 is required for GC formation, and its actions are dependent on its direct repression of >1000 target genes (reviewed in 84 ). Although first cloned as being frequently translocated in B-cell lymphomas and linked to the GC reaction, BCL6 is in fact widely expressed in many cell types. Evolutionarily, the first recognizable BCL6 gene appeared over 500 million years ago in early vertebrates. 85 Functional studies suggest that the ancestral role of BCL6 was to enable vertebrate cells to adapt to stress conditions, downstream of heat shock factor 1 (HSF1). 85 Presumably, the existence of BCL6 as an inducible stress tolerance protein facilitated evolution of the humoral immune response, where B cells must be able to tolerate the major stresses associated with massive proliferation and mutagenesis. 85 Lymphoma cells inherit dependency on BCL6 from their GC cell-of-origin and are almost universally dependent on BCL6 for their survival and proliferation. 84,86 To drive the GC phenotype, BCL6 binds and represses genes The C1/BN2, C5/MCD and N1 subtypes colored in pink, purple and orange respectively, are mostly ABC-related. The C3/EZB and C4 subtypes colored in medium and light blue respectively, are mostly GCB-related. The instability C2 subtype with no specific association to ABC or GCB is colored in green.
a Not used for classification. b The percentage of cases in the Schmitz et al study represents the predicted prevalence of the indicated DLBCL subtypes in a population-based cohort, as the cohort was deliberately enriched for ABC and COO-unclassified cases.
the radar" of immune regulatory networks until encountering the is lethal to lymphoma cells, but can also result in induction of target genes like BCL2 resulting in a process described as "oncogene switching." Along these lines combination of BCL6 and BCL2 inhibitors is highly synergistic in killing lymphoma cells. 100

| CORE EPI G ENE TIC MECHANIS MS DRIVING G C-DERIVED LYMPHOMAG ENE S IS
During the humoral immune response, GC B cells undergo dramatic and rapid-sequence phenotypic changes. 101   KMT2D mutation or deficiency causes a focal loss of H3K4me1 activating chromatin mark predominantly at enhancers (Figure 2). This leads to repression or inability to activate genes involved in CD40, BCR, TLR, and other immune pathways. 103 Importantly, KMT2D mutation renders DLBCL cells resistant to CD40 signaling due to suppression of CD40-responsive enhancers. 103 Since KMT2D mutations F I G U R E 2 Proposed epigenetic driver mechanisms in GC B-cell lymphomas. In the GC reaction, there is transient suppression of enhancers and promoters of genes that regulate immune signaling pathways, antigen presentation, and checkpoints, which revert back to the active state when GC B cells are signaled to exit the GC reaction. Lymphomas arise from failure of GC exit signals to restore expression of these genes through several proposed epigenetic mechanisms: (A) EZH2 is induced in GC B cells and converts H3K4me3 active promoters (green) to H3K4me3/H3K27me3 bivalent promoters (yellow) for transient repression of target genes, which is reversed upon GC exit. EZH2 is an H3K27 methyltransferase and component of the PRC2 polycomb complex that is upregulated in the GC. 16,112,113 Conditional deletion of EZH2 in GC B cells results in failure to form GCs. 16,113 In GC B cells, EZH2 converts gene promoters from an active state marked by H3K4me3 to a bivalent H3K4me3/H3K27me3 "poised" state ( Figure 2). 16 Repression of these genes requires the presence of BCL6, which together with the H3K27me3 mark recruits the BCOR complex through combinatorial tethering. 21 Genes regulated in this manner include cell cycle checkpoint genes such as CDKN1A, antigen presentation genes, and GC exit genes such as IRF4. 16,17,21,114 It is not yet known how the EZH2 H3K27 methylation program is erased in the light zone, but it is reasonable to postulate that this is essential for GC exit. EZH2 is affected by somatic gainof-function mutations in 30% of GCB-DLBCL and FL patients. 115,116 These mutations are always heterozygous and most commonly affect the Y641 residue within the catalytic pocket of the EZH2 enzymatic SET domain. Y641 mutant EZH2 yields more efficient H3K27 trimethylation but loss of H3K27 monomethylation activity. This explains why the mutation is always heterozygous, since the mutant enzyme needs the wildtype protein to monomethylate H3K27. 21,117 Conditional expression of Y641 mutant EZH2 in GC B cells leads to GC hyperplasia and lymphomagenesis. 16 EZH2-specific inhibitors or shRNA cause proliferation arrest and a partial plasma cell phenotype in DLBCL cells, and eventually some degree of apoptosis. 16 These effects occur more rapidly in cell lines harboring EZH2 mutations, and justified the clinical translation of EZH2 inhibitors for patients with FL and DLBCL. 16 Unique among recurrent mutations in DLBCL, TET2 lesions occur in hematopoietic stem cells and hence are by definition "founder" mutations in DLBCL. 38 These are generally missense or truncating mutations that result in protein loss. 39,57 TET2 is an alpha-ketoglurate (aKG)-dependent dioxygenase that converts 5'methylcytosine (5mC) into 5'hydroxymethylcytosine (5hmC). 118 Gene promoter 5mC is linked to transcriptional repression, whereas 5hmC is a transcriptional activation mark at gene enhancers ( Figure 2). 119 Loss of Tet2 results in GC hyperplasia in mice and failure to undergo class-switch recombination and plasma cell differentiation, ultimately leading to the development of B-cell lymphomas. 39 This phenotype is linked to loss of gene enhancer activating 5hmC and H3K27 acetylation marks in GC B cells, at a similar set of genes that are controlled by CREBBP. TET2 and CREBBP mutations are mutually exclusive in DLBCL, suggesting that they control the same pathways. 39 Loss of TET2 renders DLBCL cells dependent on HDAC3, suggesting a potential therapeutic approach for TET2mutant DLBCL patients. 39

| DYS REG UL ATI ON OF G C ME TABOLIS M
Many B-cell lymphomas originating in the GC present an exceptionally high proliferation index. 120 This implies massive metabolic requirements in order to generate sufficient energy and support anabolism for repeated growth and division cycles.  (Figures 3 and 4). 58,81 In the normal GC context, constitutive mTORC1 activation (through deletion of the mTORC1 inhibitor Tsc1 or constitutive activation of RagA) promotes a temporary clonal expansion in the dark zone. 47 However, this is followed by the progressive extinction of mTORC1-constitutively active GC B cells, which feature a competitive disadvantage due, at least in part, to a failure to undergo affinity maturation. 47 Therefore, while mTORC1 activation must be transient to support the GC reaction, constitutive mTORC1 activation seems to be highly supportive of lymphoma survival.  (Figure 3). 125 High expression levels of the key PPP enzyme G6PD in DLBCL patients associate with poor prognosis. 125 Furthermore, pharmacological inhibition of PP2A or G6PD induces cell death in the DLBCL cell line OCI-Ly10, with a strong synergistic effect of combined inhibition. 125 PP2A inhibition or PP2A plus G6PD inhibition also increases survival of mice engrafted with primary human DLBCL cells. 125 In normal B cells, PPP is normally kept low through the action of the transcription factors PAX5 and IKZF1 that specifically repress G6PD and other PPP enzymes. However, BCR activation can stimulate Glut1-mediated glucose uptake and redirect, as cells progress through G1/S, glucose usage from primarily glycolysis to the PPP. 32 This generates NADPH, which provides antioxidant protection to proliferating cells. Furthermore, glutamine might be used in place of glucose to replenish the TCA cycle in activated B cells, and the imported glucose would instead serve ribonucleotide biosynthesis through the PPP (Figure 3). 30 Diffuse large B-cell lymphomas also develop addiction to the mitochondrial protein deacetylase SIRT3 regardless of DLBCL mutation profile or cell-of-origin. 126 SIRT3 stimulates glutaminolysis by directly activating mitochondrial glutamine dehydrogenase F I G U R E 3 Metabolic dysregulation in GC-derived B-cell lymphoma. In the mitochondrion, the tricarboxylic acid (TCA) cycle produces reduced NADH and FADH2 that are used by complexes I to V of the electron transport chain (ETC) to generate ATP through oxidative phosphorylation (oxphos). The TCA can be fueled by fatty acid-or pyruvate-derived acetyl-CoA. Alternatively, glutamine can be used to generate alpha-ketoglutarate (aKG). DLBCLs have developed dependency on SIRT3 to replenish the TCA cycle (also known as anaplerosis). SIRT3 stimulates glutaminolysis by activating the glutamine dehydrogenase (GDH). Glucose can be converted into pyruvate through glycolysis or used through the pentose phosphate pathway (PPP) to generate ribose 5-phosphate (R5P) and NADPH. Some GC-derived B-cell lymphomas depend on PP2A and G6PD, a key PPP enzyme, to switch glucose carbon usage from glycolysis to the PPP. This provides antioxidant protection and supports ribonucleotide biosynthesis in proliferating cells. Pyruvate can also be converted into lactate as part of the "aerobic glycolysis" that tumor cells use to "bypass" the TCA cycle, to generate some ATP and to create biomass. This is known as the Warburg effect and can be induced by MYC and HIF1-alpha stabilization in DLBCL and FL. Finally, mTORC1 activation in the GC happens downstream of T cell-positive selection signals via the PI3K/AKT/mTOR pathway or downstream of nutrient signaling via activation of RagA/C and the v-ATPase. These components either carry gain-of-function (GOF) mutations, are hyper-activated, or are expressed at high levels in GC-derived B-cell lymphomas, resulting in mTORC1 constitutive activation. mTORC1 is the master regulator of anabolism and while constitutive activation is detrimental to GC B cells, it appears to favor lymphoma growth. G6PD, glucose-6-phosphate dehydrogenase; ROS, reactive oxygen species; v-ATPase, vacuolar ATPase proton pump (GDH) to enhance TCA activity and generate alpha-ketoglutarate (aKG) (Figure 3). Sirt3 depletion yields reduced abundance of TCA metabolites and is rescued by GDH overexpression or addition of an aKG analog (to bypass glutaminolysis). 126 Sirt3 knockout impairs BCL2-driven lymphomagenesis but has no effect on normal GCs. 126

| Adaptation to low nutrient and oxygen levels
Germinal center B cells are exposed to reduced oxygen availability in the light zone and depend on the serine/threonine protein kinase Gsk3 to survive under hypoxic conditions in vivo. 31  and is rather inefficient in producing ATP, but helps to create biomass ( Figure 3). 133 Therefore, response to hypoxia-induced metabolic imbalances might facilitate anabolism in GC-derived lymphoma cells.

| DIS RUP TION OF S IG NALING PATHWAYS
Signal transduction is hijacked in lymphomas to promote survival of cells that would otherwise be negatively selected in the GC. As detailed in Table 1, a large number of signal transducers are mutated or translocated in lymphomas of GC origin, including receptors, adapter proteins, kinases, phosphatases, ubiquitin ligases, GTPases or ultimately, transcription factors. 57,58,81,106 Although mutation frequency of most individual genes is low, when these signal transducers are grouped in pathways they can be subtype defining.
Perhaps the paradigm for this would be the concurrent constitutive activation of the BCR and TLR pathways in extranodal ABC-DLBCL (C5/MCD). 57,58,81,106

| BCR and PI3K pathways
Antigen recognition triggers "active" BCR and PI3K signaling. 134 However, B cells can also receive "tonic" BCR signaling, which is antigen-independent and can be rescued by PI3K activation alone.
This tonic BCR signaling is typical of B1 and follicular B cells (reviewed in 135 ) but both active and tonic BCR signaling can be hijacked by GCderived lymphomas to promote survival.

Active BCR signaling. Many lymphomas manifest "chronic-active"
BCR signaling that is reminiscent of antigen-dependent BCR and PI3K signaling. This phenomenon was first demonstrated in ABC-DLBCL using a loss-of-function RNA interference screen to identify dependence on BCR signaling mediators. 136 Following this effort, a number of reports confirmed the presence of somatic mutations in BCR pathway genes such as TNFAIP3, CARD11, and CD79A/B (Figure 4). [137][138][139] For example, FL and ABC-DLBCL (C5 and MCD cases) preferentially carry mutations in genes characteristic of active BCR signaling (eg, CD79A/B or CARD11).

Tonic BCR signaling. BL and GCB-DLBCL (C3 and EZB cases)
instead present alterations of indirect modulators of the PI3K pathway (eg, PTEN loss or MIR17HG amplification) that are more related to "tonic" BCR signaling (Figure 4). In support of this notion, two-thirds of BL cell lines show reduced survival after CD79B or SYK knockdown but remain insensitive to (further downstream) CARD11 knockdown or IKK pharmacological inhibition ( Figure 4). 83 On the other hand, C1 and BN2 cases feature mutations in  Figure 4). 145 This "My-T-BCR" supercomplex co-localizes with mTOR, driving pro-survival NF-κB and mTOR signaling. Notably, presence of this supercomplex in patient samples was predictive of response to BTK inhibitors. 145 MYD88 mutations are also present in 15% of C1 DLBCLs but they are almost exclusively non-L265P in this group. 58  On the other hand, NOTCH1 is upregulated upon B-cell activation and necessary for differentiation into antibody-producing cells, in synergy with BCR signaling and CD40 or BAFF co-stimulation. 148,149 NOTCH1 is affected by gain-of-function mutations in the N1 group of DLBCLs, 81 and in BL. 150   and SOCS1 (12%). Constitutive activation of these genes is meant to enhance the GC response and prevent apoptosis, in favor of lymphoma survival and proliferation.

| Gα migration pathway
Another pathway frequently mutated in GC-derived lymphomas is the GC homing pathway involving S1PR2 and GNA13. Sphingosine-1-phosphate (S1P) acts via its receptor S1PR2 and the G-protein GNA13 to inhibit migration of GC cells and control their growth. 153 GC confinement of GC B cells is lost upon GNA13, S1PR2, or P2RY8 (another S1P receptor expressed on GC B cells) loss-of-function, leading to dissemination of GC-origin lymphomas. 154 Accordingly, GNA13 is mutated in 30% of GCB-DLBCL 106 and 15% of BL. 81,150 By genetic DLBCL subtypes, SP1R2 and GNA13 are disrupted in 38% of EZB cases, 81 GNA13 is frequently altered in C3, and both GNA13 and its downstream mediators RHOA and SGK1 are disrupted in C4. 58 Mutations in this pathway may enable GC B cells, which are normally largely confined to their respective follicles, to spread to other sites resulting in systemic dissemination. MHC class II-dependent manner. 165 Somatic mutation of EZH2 is associated with profound silencing of both MHC I and MHC II genes, and EZH2-mutant lymphomas in both mouse and humans manifest reduced expression of these genes and a reduction in lymphoma infiltrating CD4 and CD8 T cells. 114 This repression is due at least in part to increased levels of the H3K27me3 repressive mark at antigen presentation genes, an effect that can be overcome by EZH2 inhibitors. 114 Finally, MHC class II expression may also be affected by the differentiation status of the cells, as seen in the more plasmablastic ABC-DLBCLs that also feature low MHC II levels. 166

| Inhibiting antitumor immunity
Gaining surface markers that inhibit T cells, NK cells, and macrophages is another way to evade antitumor immune surveillance. Effector T cells express the PD-1 receptor on their surface, which upon binding to its cognate ligands PD-L1 and PD-L2 on the surface of antigen-presenting cells, induces T-cell anergy or "exhaustion." The PD-1-PD-L1/PD-L2 axis has therefore gained great interest for anticancer therapeutic intervention. 167 However, most FLs and DLBCLs express relatively low levels of PD-L1, which may explain the relatively poor performance of checkpoint inhibitors in these tumors as compared, for example, to Hodgkin lymphomas. 168 One exception to this may be the ABC-DLBCL C1 subtype that specifically harbors gains, amplifications, and translocations of the PD-L1/PD-L2 locus associated with increased expression. 58,169 PD-L1 levels may be increased through other mechanisms as well.
A recent study showed that NFATc1 activation downstream of BCR signaling was found to be responsible for IL-10 secretion and subse-  Figure 5). 175 It is proposed that TNFRSF14 and BTLA could also interact in cis on the same lymphoma B cell. 176 In line with this, CAR-T cell delivery of a soluble TNFRSF14 protein to CD19+ B cells yield enhanced antitumor efficacy in a lymphoma xenograft model, as compared to CD19-directed CAR-T cells only. 175 Hence, TNFRSF14-BTLA interaction can induce cell autonomous growth inhibitory effects and although it requires BTLA expression to be maintained, engineered immune cells could be used for targeted therapy.
Lymphoma cells can also recruit and "reeducate" surrounding cells to their advantage, for example, by modulating CD70 expres-

| Epigenetic therapy
The which also raises concerns about their use in combination with immunotherapy regimens. 165,186 Perhaps a more precise approach for targeting the epigenome is provided by specific EZH2 inhibitors, which are particularly effective in suppressing EZH2 mutant lymphoma cells and EZH2 mutant lymphomas in humans. 16,187 Wildtype DLBCLs and FLs may also respond, given that EZH2 is an essential protein in GC B cells from which lymphomas arise. Important considerations with these agents include that (a) sustained target suppression is probably important, and may be difficult to achieve especially in more aggressive tumors, (b) aside from EZH2 mutation, there is no biomarker to predict which EZH2 wildtype patients might respond, and (c) these drugs are mostly cytostatic, so that combination with other agents is needed to achieve maximal effect. Among these, combination with BH3 mimetics was shown to be highly synergistic and is well suited to the setting of EZH2 mutant FLs and EZB DLBCLs. 16 Given the role of mutant EZH2 in suppressing MHC I/II, it seems likely that these drugs would greatly enhance the efficacy of immunotherapies such as checkpoint inhibitors. 114 Although current regimens involve continuous and prolonged dosing, this can lead to the development of resistance and may not be necessary if the drug is used in more brief cycles in combination. 188 It will be necessary to follow the outcomes of clinical trials with different EZH2 inhibitors of varying potency, mechanism of action and pharmacokinetics to fully understand the optimal manner in which to use this modality in the clinic. are also highly active against DLBCL cells, which may be partly related to PRMT5 acting as a corepressor for BCL6. 190,191 As PRMT5 is also involved in RNA-splicing, the actions of these inhibitors are likely also pleiotropic.

| Targeting metabolic vulnerabilities
Lymphomas must adapt to high energy demands to sustain survival and proliferation under scarce nutrient and oxygen conditions. 192  usage from glycolysis to the PPP also represents a possible therapeutic option. 125 In addition, certain DLBCLs defined by gene expression profiling manifest addiction to glycolysis, while other "OxPhos-DLBCLs" depend on the electron transport chain (ETC) activity. 193,194 OxPhos-DLBCLs were reported to require mitochondrial palmitate oxidation to produce ATP, 193 or the mitochondrial translation machinery to raise ETC protein levels. 194 Inhibition of the mitochondrial fatty acid oxidation 193 or mitochondrial translation machinery 194

| Immunotherapy
The majority of immunotherapies require potent tumor antigen presentation to T cells. However, unlike normal light zone GC B cells, which are the cell-of-origin of FL and DLBCL, lymphoma B cells exhibit low levels of antigen presentation proteins as well as other immune defects. Low MHC class II expression has been associated with poor outcome in DLBCL, likely due to impaired immune surveillance. 155 Therefore, therapeutic intervention to restore the immune system's antitumor activity would likely prove to be powerful in GC-derived Bcell lymphomas. Clinical trials for PD-1/PD-L1 blockade have shown some response in refractory and relapsed FL (40% ORR) and DLBCL (36%) but not as striking as in non-GC-derived classical Hodgkin lymphoma (87%). 168,218 A combinational approach with synergistic drugs to jointly target several immune evasion mechanisms would therefore likely be highly beneficial. More recently, clinical trials with chimeric antigen receptor (CAR)-T cell therapy targeting CD19 cells have shown especially high response rates and durability of remission in refractory and relapsed FL and DLBCL. 219,220 Together, the advances in CAR-T cell biology, in GC-derived B-cell lymphomas identity, and pathogenesis mechanisms will allow developing efficient and precise medicine for each patient, taking into account the patient's tumor unique molecular profile and using the patient's own T cells.

| CON CLUDING REMARK S
Recent years have seen dramatic acceleration of our understanding of the humoral immune response and lymphomagenesis.
Sequencing the lymphoma genome has illuminated many new critical regulators of the GC reaction and mechanistic studies of these mutations shed light both on normal and malignant GC Bcell biology. Along with these advances we are nearing the point where lymphoma genomes will dictate selection of precision therapies geared to reverse the effect of specific mutations or target biologically defined patient subsets. One area of special interest will be using knowledge of how immune regulatory networks are suppressed in GC-derived lymphomas to develop molecular targeted plus immunotherapy regimens, with real potential to eradicate disease even in the most difficult to treat lymphoma patients.

CO N FLI C T O F I NTE R E S T
AM receives research support and has consulted for Janssen Pharmaceutical. He is also a scientific advisor to KDAC pharmaceuticals.