Disruption of KLHL6 Fuels Oncogenic Antigen Receptor Signaling in B-Cell Lymphoma

Diffuse large B-cell lymphoma (DLBCL) is the most common hematologic malignancy in the western world. Although anthracycline-based chemoimmunotherapy is curative in the majority of patients, about one-third of the patients experience relapse with dismal outcome (1). Despite the discovery of multiple compounds that are lethal to a subset of B-cell lymphomas, successful precision medicine strategies in DLBCL have been limited to targeting surface antigens of mature B cells with antibodies (2), antibody–drug conjugates (3), or chimeric antigen receptor–equipped T cells (4). Disappointments in the efficacy of small-molecule inhibitors targeting oncogenic B-cell receptor signaling arise from the extraordinary clinical, biological, and genetic heterogeneity of DLBCL, resulting in short-lived responses and inadequate patient stratification (5, 6). A better molecular understanding of the oncogenic mechanisms in DLBCL is warranted for rational personalized medicine approaches to emerge. Hence, functional characterization of the genomic drivers in DLBCL is critically needed.

B-cell antigen receptor (BCR) conveys signals that determine the expansion, differentiation, and survival of maturing B cells in the germinal center (GC) reaction (7, 8). In B-cell lymphomas, malignant cells have hijacked pathways that originate from the BCR through somatic mutations and maintain a proliferating phenotype through transcriptional dysregulation and differentiation block (9). In activated B-cell (ABC)–like DLBCL, such genetic alterations cause chronic BCR signaling and activation of nuclear factor kappa B (NF-κB; refs. 10, 11). In addition, stereotypic usage of VDJ-recombined immunoglobulin gene segments gives BCRs capable of binding self-antigens (12). Moreover, antigen-independent tonic mode of BCR signaling is required for the survival of B cells and repurposed for oncogenic growth in germinal center B-cell (GCB) DLBCL (13, 14). As such, many lymphomagenic mutations mechanistically potentiate these modes of BCR signaling by increasing BCR levels on the cell surface (10), activating converging and downstream pathways (15–18), or disrupting negative feedback loops of the signaling (19, 20).

The germinal center regulator of B-cell maturation Kelch-like family member 6 (KLHL6) is recurrently mutated in DLBCL, but a potential link to oncogenic BCR signaling has remained elusive. KLHL6 belongs to the bric-a-brac tramtrack broad (BTB)-Kelch family of proteins, which function as exchangeable substrate adapters for Cullin-3-RING-ligase (CRL) E3 ubiquitination machinery targeting proteins for degradation. KLHL6 was first cloned from ovine Peyer’s patch tissues, and its expression is restricted to human GC B cells and most GC-originating B-cell lymphomas (21, 22). The importance of KLHL6 expression for B-cell function was demonstrated in vivo and in vitro, as Klhl6 knockout mice had severely reduced mature B-cell numbers, impaired GC formation following ovalbumin immunization and had blunted responses to BCR stimulation in vitro (23). However, the indispensable role of KLHL6 in the GC reaction is controversial, and a potential link with BCR signaling lacks mechanistic evidence (24).

Somatic mutations in the N-terminal BTB domain of KLHL6 have been reported in 1% to 10% of DLBCLs (25–27) and 1% to 3% of chronic lymphocytic leukemia (CLL; refs. 28, 29). In DLBCL, KLHL6 mutations have neither been linked to any molecular subtypes nor are they involved in emerging genomic taxonomies (30, 31). More recently, BTB domain mutations have been found in autoreactive B-cell clones responsible for autoantibody production in cold agglutinin disease (32). First efforts to understand the consequences of these mutations have linked the loss of KLHL6 with NF-κB activation through stabilization of RNA decay factor Roquin-2, which leads to downregulation of the negative regulator A20 (TNFAIP3; ref. 33). Some BTB domain mutations were suggested to result in impaired Cullin-3 binding proposing a loss-of-function. However, these first studies have relied on protein interaction discovery in cell lines that do not recapitulate the molecular context of GC-derived B-cell lymphomas. Thus, how KLHL6 mutations promote lymphomagenesis remains incompletely characterized, impeding rational therapeutic targeting of these alterations.

To determine the phenotypic consequences of various KLHL6 gene mutations found in DLBCL, we examined the molecular portrait of KLHL6 in DLBCL and applied high-throughput proteomic profiling and functional characterization to discover how these mutations alter BCR expression and function. We found that KLHL6 interacts with the BCR and regulates its surface levels. Notably, mutations in the BTB domain led to increased BCR levels and signaling, thereby adding to the multitude of targetable mechanisms of oncogenic BCR signaling in B-cell malignancies.

To get an overview of KLHL6 protein expression in DLBCL tissues, we first performed immunostaining of tissue sections from reactive lymphoid specimen and diagnostic DLBCL samples. In lymphoid tissues, KLHL6 protein expression was mainly constrained to GC B cells that were not spatially restricted to either light zone (LZ) or dark zone (DZ) of the GC (Fig. 1A and B; Supplementary Fig. S1A and S1B). At the single-cell level, KLHL6 expression varied across GC B-cell subpopulations with the highest expression in transitional B cells whereas cells undergoing DZ and memory B-cell differentiation had downregulated KLHL6 (Supplementary Fig. S1C). Colocalization studies demonstrated that KLHL6 localized on perinuclear vesicles concentrated in the immediate vicinity of the Golgi apparatus (Fig. 1C and D; Supplementary Fig. S1D and S1E). This physiologic GC-like KLHL6 expression pattern (KLHL6^GC+) was evident in most malignant cells in the examined DLBCL tissues and cell lines (Fig. 1C–F). In a subset of DLBCL cases and cell lines, however, malignant cells did not express KLHL6 or they exhibited a varying degree of cytosolic protein expression pattern (KLHL6^GC–; Fig. 1E and F).

In the discovery cohort of 134 primary DLBCLs, the overall immunoreactivity for KLHL6 antibody was high, with the KLHL6^GC+ phenotype present in 67% of the samples (Fig. 1G; Supplementary Fig. S1F; Supplementary Table S1A). KLHL6^GC+ was strongly enriched in the GCB subtype, whereas 44% of the ABC DLBCLs were KLHL6^GC– (Fig. 1H). At the mRNA level in an expanded data set of 624 DLBCLs (34), KLHL6 gene expression was significantly higher in GCB DLBCLs compared with ABC DLBCLs (Fig. 1I and J). In detail, KLHL6 expression was the highest in the lymphomas dominated by B cells of S1 state, corresponding to GC B cells (Fig. 1I–K; ref. 35), 90% of which were deemed KLHL6^GC+ on the protein level. KLHL6 gene expression correlated with the expression of other GCB markers including BCL6 and MEF2B and 90% (70 of 78) of the KLHL6^GC+ were also BCL6+ by IHC (Fig. 1L; Supplementary Table S1B and S1C). Of note, neither elevated KLHL6 (3q27.1) gene expression nor KLHL6^GC+ phenotype was associated with BCL6 (3q27.3) translocations (Fig. 1L; Supplementary Table S1C).

To further characterize the subset of lymphomas that were KLHL6^GC−, we expanded our analysis to the genomic level. We found that somatic mutations in MYD88 and other ABC drivers linked to the MCD/C5 genotype were associated with KLHL6^GC− (Fig. 1M; Supplementary Fig. S1G). Besides aberrations characteristic of MCD/C5 subtype, we observed that KLHL6 mutations itself more common with KLHL6^GC− phenotype (Fig. 1M; Supplementary Fig. S1G). Clinically, KLHL6^GC+ did not associate with underlying patient characteristics besides cell of origin (Supplementary Table S1D). In contrast, although there was a trend toward poor survival for KLHL6^GC− in the discovery cohort (Fig. 1N), KLHL6^GC− phenotype translated to poor survival in the ABC DLBCL subtype independently of the International Prognostic Index (IPI; Fig. 1O). Collectively, these findings show that KLHL6^GC+ is characteristic of GCB DLBCLs, whereas this phenotype is absent in a subset of ABC DLBCLs with MCD/C5-like features and poor outcome.

Pooled analysis of the reported KLHL6 mutations in DLBCLs (7.6%, 303/4,011 patients; Supplementary Table S2A and S2B) revealed that the mutations were mostly of missense type (89%) and four recurrent amino acid residues accounting for 33% of the reported mutations (Fig. 2A–C). As established, the N-terminal BTB domain was targeted by recurrent point mutations most commonly affecting leucine 65 and leucine 90 with proline (KLHL6^L65P) or phenylalanine (KLHL6^L90F) substitutions, respectively (Fig. 2B). Additionally, we identified two recurrently mutated residues in the C-terminus of KLHL6 in the Kelch domain (Fig. 2A). These Kelch mutations switched evolutionary conserved negatively charged glutamic acid residues (E547 and E568) to mostly positively charged lysines (KLHL6^E547K and KLHL6^E568K; Fig. 2B and C). In silico structural modeling (36) revealed that these residues were situated side-by-side on the outer surface of the Kelch domain, lining the putative substrate–binding surface of the domain (Fig. 2D and E). Interestingly, these C-terminal mutations have not been reported in CLL (37) with 2.8% KLHL6 mutation rate, solely relating to the N-terminus (Supplementary Fig. S2A). In 928 DLBCLs (38), mutations in SGK1 were enriched in patients with either N- or C-terminal mutations, however, select NF-κB pathway genes and hypermutable targets were enriched in cases with exon 1 mutations encoding the BTB domain mutation hotspots of KLHL6, whereas mutations in TNFRSF14, TET2, and EBF1 were enriched in cases with Kelch domain mutations (Fig. 2F; Supplementary Fig. S2B).

Next, we characterized KLHL6 mutants for subcellular protein expression pattern in the available DLBCL samples. In the discovery set, we found that the association between the loss of KLHL6^GC+ localization pattern and KLHL6 mutations was limited to the BTB domain compromising N-terminal mutations, whereas Kelch domain affecting mutations retained physiologic attenuation (Fig. 2G and H; Supplementary Fig. S2C). Moreover, KLHL6^L65P localized in cytosolic aberrant vesicles in the DLBCL cells, whereas in a patient with KLHL6^L90F in the original dataset KLHL6 was expressed throughout the cytosol (Fig. 2H). Notably, a similar cytosolic pattern was seen in one DLBCL sample from our training set which upon genomic interrogation of KLHL6 exon one carried a double mutation of KLHL6 (KLHL6^A93D-S94C; Fig. 2H).

Introduction of wild-type (WT) and different KLHL6 mutants to GCB and ABC DLBCL cell lines with endogenous KLHL6 expression recapitulated the findings from DLBCL samples and confirmed that BTB domain mutants were localized in nonphysiologic cytosolic patterns (Fig. 2I and J; Supplementary Fig. S2D). By contrast, WT and Kelch domain mutant KLHL6 displayed the GC-like pattern (KLHL6^GC+, Fig. 2I and J; Supplementary Fig. S2E). Furthermore, ectopic BTB domain mutants abolished the physiologic expression pattern of endogenous KLHL6 to a varying degree, suggesting that these mutations could possess dominant capabilities over the endogenous protein (Fig. 2I and J). Together, these findings reveal a dichotomous pattern of mutations targeting the BTB and Kelch domains of KLHL6 in DLBCLs that disrupt subcellular localization in malignant B cells.

To uncover B-cell pathways affected by KLHL6 mutations, we introduced C-terminally tagged KLHL6 constructs to SuDHL5 GCB cells with high KLHL6 gene expression and KLHL6^GC+ phenotype, affinity-purified the lysates for the baits with their bound complexes and identified the interacting proteins with high-throughput mass spectrometry (AP-MS, Fig. 3A; Supplementary Table S3A–G).

The protein–protein interactome of WT KLHL6 reflected its role as a CRL substrate adapter for ubiquitination and subsequent proteasomal degradation of target proteins (Fig. 3B). These interactors included Cullin-3, Kelch-chaperone NuDCD3 (39), the CRL deneddylating Cop9-signalosome complex, and immunoproteasome components. Additionally, we identified proteins involved in BCR signaling of which subunit CD79A and the adapter protein BANK1 were detected with the highest confidence (Fig. 3B). Furthermore, we recorded interactions with SUGT1, involved in membrane trafficking of MHC I and II complexes in B cells (40), and a previously reported interactor Roquin-2 (33). Heterotypic interactions with other BTB-Kelch family proteins KLHL7 and KLHL26 proposed that KLHL6 forms heterodimeric configurations characteristic to BTB-Kelch family members. Notably, KLHL7 and KLHL26 have been shown not to dimerize with each other (41), and we found that these proteins were expressed in different subcellular compartments of DLBCL cells and colocalized with endogenous KLHL6 (Fig. 3C). While KLHL7 and KLHL6 proteins colocalized in perinuclear vesicles, the colocalization with KLHL26 was more diffusely present in perinuclear membranous structures.

To delineate mutation-dependent changes in KLHL6 protein interactions, we compared the proteomic data between WT and mutant KLHL6 baits. All baits were ubiquitinated in multiple residues and most of the protein–protein interactions were retained by the mutant baits, suggesting that they remained involved in CRL activities (Supplementary Table S3A–G). Strikingly, the BTB domain mutations L65P and L90F ablated heterodimerization of KLHL6 with KLHL7 and KLHL26 (Fig. 3D–F). In contrast, the Kelch-domain variant KLHL6^E568K bait was more tightly engaged by these BTB-Kelch family proteins (Fig. 3F). Conflicting with a previous report, we found that Cullin-3 co-purified with all the baits in every replicate, albeit to a lesser degree with KLHL6^L65P (Fig. 3F). These patterns prompted us to ask whether BTB mutant KLHL6-baits still formed CRL complexes through homodimerization with endogenous KLHL6. We recovered WT and mutant KLHL6 peptide spectra from AP-MS data of both BTB-mutant baits, strongly arguing in favor of homodimerization with endogenous KLHL6 (Fig. 3G and H). Additionally, affinity purifications (APs) with several KLHL6 WT and patient-derived mutant baits co-purified endogenous KLHL6, thus confirming the formation of a homodimer CRL that was not disrupted by the examined mutations (Fig. 3I).

Given that KLHL6 mutants could form homodimer CRLs, we asked how mutant KLHL6 baits would affect the KLHL6^WT interactome (Fig. 3J). A stronger association between KLHL6^E568K and CRL components, the Cop9-signalosome, and CRL-activating posttranslational modification Nedd8 suggested that this Kelch variant could be involved in constitutively active forms of CRL complexes. However, KLHL6^E568K did not differ in the interaction with CD79A and BANK1, in contrast to BTB mutant KLHL6^L90F (Fig. 3J).

Finally, to determine which KLHL6 interactions were shared by diverse molecular contexts of B-NHL, we examined the KLHL6 interactome in a broader panel of DLBCL cell lines. Most of the CRL components and many other interactors were confirmed throughout the cell line panel and, notably, all the pull-downs confirmed interactions with the BCR, including both IgM and IgG isotypes (Fig. 3K; Supplementary Fig. S3A and S3B; Supplementary Table S3H and S3I). Altogether, these experiments indicate that KLHL6 forms homodimers and heterodimers and interacts with the CRL machinery and BCR components. Specifically, recurrent BTB mutations compromise the heterodimeric constellations.

Apart from the protein components that form the KLHL6-CRL machineries, their interactions with the substrates they process are likely more transient and, thus, less likely to be detected by purification of protein complexes. Therefore, to obtain a molecular map of potential substrates recognized by KLHL6, we complemented our AP-MS interactome screen with KLHL6 baits fused with a Kelch domain adjacent proximity–dependent biotin ligase (BioID2; Fig. 4A; ref. 42).

We uncovered additional KLHL6 interactors with high confidence and validated many of the AP-MS interactors with BioID2 (Fig. 4B–D; Supplementary Table S4A–C). Novel interactors included ubiquitin turnover–related proteins such as deubiquitinating enzyme 25 (USP25) and E3 ubiquitin ligase UBR4 (Fig. 4B). We found further support for the localization of KLHL6 in endoplasmic reticulum (ER) and protein cargo–related compartments with high-confidence interactors, such as the Serca-type Ca²⁺ pumps, ER exit site marker SEC16A, and BANK1-associated Ca²⁺ channel inositol 1,4,5-trisphosphate (IP3) receptor type 2 (ITPR2; Fig. 4B; Supplementary Fig. S4A; ref. 43).

By contrast, some KLHL6 AP-MS interactors were poorly captured by BioID2. Among these, the N-terminal BTB-mediated heterodimeric interactions with KLHL7 and KLHL26 were poorly evidenced as labeling distance and protein stoichiometry limit the detection of interactors with C-terminal proximity labeling. By combining the data from the AP-MS and BioID2 interaction screens, we revealed that besides CRL components and electrolyte pumps, BCR signaling molecules CD79A and BANK1, together with LuZP1, a centrosomal actin modulator involved in ciliogenesis (44, 45), were captured by both affinity and proximity interactome screening methods, strongly suggesting that these proteins represented substrates of the KLHL6-CRL (Fig. 4C and D).

Next, we asked whether manipulation of KLHL6 expression impacted the protein levels of these interactors. In the examined KLHL6^GC+ GCB cell lines, overexpression of KLHL6 WT downregulated the protein levels of BCR components CD79A and CD79B, whereas CRISPR-Cas9 or RNAi-mediated silencing increased their levels (Fig. 4E). By contrast, the impact on immunoglobulin heavy chain (IgH) levels was not evident (Fig. 4E). In the ABC cell lines U2932 and HBL1, however, the impact of KLHL6 manipulation was only modest on the expression levels of BCR components (Fig. 4E). To clarify these observations, we examined the KLHL6-BioID2 interactome in U2932 and validated BANK1 and LuZP1 interactions, whereas support for BCR components was minimal (Supplementary Fig. S4B; Supplementary Table S4D). BANK1 and LuZP1 levels were also downregulated in response to KLHL6 overexpression (Fig. 4F; Supplementary Fig. S4C). Together, these experiments suggest that KLHL6 targets BCR components, BANK1, and LuZP1 proteins for degradation in the GCB context.

To examine the functional connection between KLHL6 and BCR signaling, we tested if KLHL6 interactions were altered upon BCR activation by performing AP-MS for KLHL6 bait at 60 seconds after BCR cross-linking in SuDHL5 cells (Fig. 5A; Supplementary Fig. S5A and S5B). We observed that although the protein complex of KLHL6 remained mostly unchanged, the heterodimerization with KLHL7 was strengthened and the interactions with UBR4 and immunoproteasome were induced (Fig. 5B–D; Supplementary Fig. S5C and S5D). However, in contrast to the manipulation of KLHL6, overexpression or CRISPR-Cas9 silencing of KLHL7 did not similarly alter the levels of BCR components in cell lysates suggesting that KLHL6-KLHL7 heterodimerization was dispensable for BCR downregulation (Fig. 5E and F; Supplementary Fig. S5E–G).

Given that the KLHL6 mutants retained homodimerization, however, we proceeded to characterize the consequences of different KLHL6 mutants for BCR expression and signaling in DLBCL cells. Analysis of total and surface protein levels of CD79A, CD79B, and IgM/IgG in isogenic DLBCL cell lines with endogenous KLHL6 (parental), or overexpression of KLHL6^WT or recurrent mutants revealed differential impacts by the BTB and Kelch mutations on the BCR phenotype. Strikingly, the BTB domain mutants KLHL6^L65P and KLHL6^L90F failed to downregulate and even increased the levels of CD79A and CD79B across all tested cell lines (Fig. 5G and H; Supplementary Fig. S5H and S5I). In contrast, Kelch domain mutants KLHL6^E547K and KLHL6^E568K retained their ability to promote downregulation of these molecules also in the ABC-DLBCL cell line U2932, in which the overexpression of KLHL6^WT had little impact (Fig. 5H; Supplementary Fig. S5H and S5I). The changes in the total levels of CD79A and CD79B were recapitulated by similar changes in the surface levels, whereas the surface level of IgM/IgG (IgH) followed the changes in CD79A and CD79B (Fig. 5I and J; Supplementary Fig. S6), which is in agreement with the requirement of CD79A-CD79B heterodimer for IgH transportation to the plasma membrane (14). In detail, the cell surface levels of all BCR components CD79A, CD79B, and IgM were significantly decreased across the GCB cell lines following ectopic WT KLHL6 expression, whereas no such impact was seen in the ABC cell line U2932 (Fig. 5I and J). However, the overexpression of the Kelch domain mutants decreased the surface levels of BCR across all cell lines (Fig. 5I and J). In contrast, BTB mutants did not downregulate BCR surface levels, and KLHL6^L65P even increased the surface levels of all BCR components similarly to KLHL6 silencing by short hairpin RNA (shRNA) in both ABC and GCB cell lines (Fig. 5I and J). These phenotypic differences in the surface levels of BCR components led us to hypothesize that BCR responses could differ between DLBCL cells expressing different KLHL6 mutants.

We engaged surface BCRs of isogenic DLBCL cells with endogenous levels or overexpressing WT or mutant KLHL6 proteins to profile changes in protein phosphorylation of intracellular proximal signaling kinases. In line with the lowest level of surface BCRs, overexpression of the Kelch domain mutant KLHL6^E568K attenuated the phosphorylation of proximal BCR-activated kinases such as Src family kinases (SFK, e.g., Lyn), SYK, BTK, and phospholipase C γ (PLC-γ) compared with the parental cells in GCB cell lines (Fig. 5K and L; Supplementary Fig. S7A). However, no significant change was observed in the ABC cell line U2932 with only a minor effect of KLHL6^E568K on the BCR surface levels (Fig. 5I–L). Overexpression of KLHL6 WT had a similar but less significant effect. The effects of Kelch domain mutants contrasted the overexpression of the BTB domain mutant KLHL6^L65P, which led to significantly potentiated proximal BCR signaling responses in the GCB cell line SuDHL5 and the ABC cell line U2932 (Fig. 5L). In concordance with the surface expression levels, the shRNA knockdown of KLHL6 (sh_#2118) mimicked that of KLHL6^L65P and led to significantly potentiated BCR signaling as compared with shRNA control cells (Fig. 5K and L; Supplementary Fig. S7A).

BCR engagement in the absence of a co-stimulatory signal induces apoptosis of B cells in vitro and the magnitude depends on intracellular Src kinase activity (46). Therefore, we hypothesized that manipulation of KLHL6 expression should alter these responses. When unchallenged, U2932 and SuDHL5 cells with various KLHL6 manipulations grew at a similar rate, however, silencing of KLHL6-sensitized and WT overexpression desensitized lymphoma cells for activation-induced cell death (AICD; Fig. 5M; Supplementary Fig. S7B and S7C). Notably, in SuDHL5 cells, the BTB domain mutants did not compromise cell viability upon stimulation whereas these mutants sensitized U2932 cells to AICD similarly to KLHL6 knockdown (Fig. 5M). Despite the lower BCR expression levels in the Kelch domain mutant expressing cells, the viability was comparable with transduction controls and parental cell line (Fig. 5M).

Having established a functional link between KLHL6 and the BCR in cell lines, we re-examined the discovery cohort by immunofluorescent (IF) microscopy for KLHL6 and BCR marker (CD79B) at the protein level (Fig. 6A; Supplementary Fig. S8A–D). Quantification of the CD79B staining revealed that KLHL6^GC+ tumors had lower CD79B levels than the tumors with KLHL6^GC- phenotype, whereas no such association was seen on CD79B gene expression (Fig. 6B and C; Supplementary Fig. S8E). The difference in CD79B protein was confirmed among the major subtypes, and KLHL6 and CD79B mutations showed a mutually exclusive trend (Fig. 6D–F; Supplementary Fig. S8F). Notably, the tumors with KLHL6 BTB mutations had a trend toward higher CD79B intensity, whereas CD79B mutations did not (Fig. 6G and H). In the tumors, KLHL6^L65P and KLHL6^L71V mutations were associated with membrane-like CD79B staining, whereas other BTB mutants had variable CD79B patterns (Fig. 6I; Supplementary Fig. S8G). Yet, the staining for CD79B was low in the tumor carrying KLHL6^E547K mutation (Fig. 6I; Supplementary Fig. S8H). Taken together, these experiments and characterization suggest that KLHL6 targets the BCR and disruptions in the gene result in an aberrant BCR phenotype.

Finally, to address these strikingly different phenotypes that the different KLHL6 mutants induced to lymphoma cells, we sought to investigate their molecular characteristics beyond the BCR components.

In DLBCL samples, BANK1 gene expression was linked to opposing molecular features to KLHL6 expression, including higher expression associated with ABC subtype, mutations in BANK1 interactor MYD88 (47) and association with S3 and S4 B-cell states (Supplementary Fig. S9A–E). We introduced KLHL6 constructs to SuDHL4 cells, with KLHL6^GC+ phenotype and high BANK1 expression, and observed the strongest colocalization between BANK1 and KLHL6^L90F and KLHL6^A93D-S94C mutants in the cytosol (Fig. 7A and B). By contrast, colocalization was compromised between BANK1 and KLHL6^L65P and Kelch domain mutants as supported by AP-MS interactome data with the strongest association between BANK1 and KLHL6^L90F (SaintScore > 0.95; Fig. 7B and C; Supplementary Table S5A). KLHL6^L90F and KLHL6^A93D-S94C promoted BANK1 degradation even more effectively than WT KLHL6 overexpression across all the tested cell lines, whereas KLHL6^L65P and the Kelch domain mutations failed to promote BANK1 downregulation (Fig. 7D). These findings together with our AP-MS data supporting homodimeric CRL-formation highlight that KLHL6^L90F remains functional for the downregulation of BANK1, whereas the other recurrent BTB domain mutant KLHL6^L65P exhibits loss of function.

To gain additional insights into the pathogenic roles of Kelch domain mutants, we performed PL-MS with mutant constructs in SuDHL5 cells. According to the principal component analysis (PCA), the proximity interactors of the Kelch domain mutants KLHL6^E547K and KLHL6^E568K overlapped and differed from those of the WT bait (Fig. 7E). Explaining this difference, the KLHL6 interactor LuZP1 was the most prominently lost interaction partner by both Kelch domain variants in the high-throughput data (Fig. 7F and G; Supplementary Table S5B and S5C). By contrast, the pull-downs with KLHL6^L65P bait contained significantly more LuZP1 peptides, indicating substrate trapping properties of this poorly Cullin-3–associated bait (Fig. 7H). Co-staining of LuZP1 and KLHL6 in DLBCL cells revealed that one cytosolic KLHL6 punctum colocalized with centrosomal LuZP1 protein by IF microscopy (Fig. 7I and J). Remarkably, only WT KLHL6 overexpression promoted LuZP1 downregulation, whereas recurrent BTB and Kelch domain KLHL6 mutants did not (Fig. 7K).

Finally, we explored the BioID2 interactions of KLHL6^L65P bait (Supplementary Table S5D). Proximity labeling confirmed the poor association between KLHL6^L65P and Cullin-3 and other Cullin family scaffolds, USP25 and UBR4 (Fig. 7L). However, KLHL6^L65P gained unique interactions and was more strongly associated with KLHL6 interactor SEC16A that colocalized with aberrant KLHL6^L65P vesicles in SuDHL5 cells (Fig. 7L and M). Although mostly descriptive, the above data reveal both shared and unique substrates between recurrent KLHL6 variants that extend beyond dysregulation of BCR expression and signaling.

The landscape of recurrently mutated protein-coding genes in DLBCL is well established (25, 26, 31, 48), but their functional implications have remained largely uncharacterized. Through translational investigation including comprehensive proteomic analyses of KLHL6 and its recurrent mutants in the malignant B-cell context, we discovered novel functionalities for this protein. Our study revealed that BTB domain mutants of KLHL6 fail to heterodimerize and promote degradation of their substrates CD79A, CD79B, and IgM, which constitute the BCR complex. Expression of the BTB-mutant KLHL6^L65P increased BCR surface levels on the plasma membrane and induced a hyper-responsive phenotype to BCR signaling. Moreover, additional functional anomalies exhibited by recurrent BTB and Kelch domain mutants induced phenotypic changes that distinguish them from total genomic ablation of KLHL6. Together, our findings reshape the pathogenic role of KLHL6 mutations in lymphoma, which have previously been postulated as passenger mutations arising from aberrant somatic hypermutation or resulting in complete loss of function (24, 33). We anticipate our findings to extend to other B-cell diseases with KLHL6 mutations, such as CLL and B-cell autoimmunity besides B-NHL.

Discovery of new pathogenic BCR signaling mechanisms could aid rational stratification of patients for different therapeutic interventions (49). The absence of the KLHL6^GC+ phenotype in almost half of ABC DLBCLs could be permissive to the oncogenic chronic active mode of BCR signaling in which sustained proximal BCR signaling and evasion of anergy through increased plasma membrane density of self-reactive BCRs represents a hallmark. Previous studies have demonstrated that lymphoma cells can unlock this oncogenic strategy through CD79B/A mutations, which result in increased BCR surface levels through hindered BCR internalization and loss of negative regulation by phosphorylation by LYN (10). We showed that KLHL6 loss and recurrent BTB mutations add to this hallmark by increasing protein and membrane levels of BCR signaling subunits in cell lines and patients. These phenotypic characteristics have been shown to impact responses to small-molecular inhibitors (50) and could render these lymphomas sensitive to the anti-CD79B antibody–drug conjugate polatuzumab vedotin with promising efficacy especially in the ABC DLBCLs (51, 52). Furthermore, association and potential cooperation between loss of KLHL6^GC+ and MYD88 mutations could promote synergistic BCR and Toll-like receptor 9 (TLR9) signaling via oncogenic My-T-BCR complex, which is targetable with small-molecule inhibitor ibrutinib (18, 53, 54). Examining KLHL6 in the tumor tissues of patients treated with these regimens could reveal a biomarker role for KLHL6 mutations and KLHL6^GC+ phenotype to predict responses besides being a prognostic marker in the ABC DLBCL.

Recurrent BTB domain mutations and the related loss of KLHL6^GC+ phenotype could add to chronic active mode of BCR signaling by disrupting surface BCR homeostasis. Most established CRL complexes, such as KEAP1-CRL, rely on BTB domain mediated homodimer formation and their accurate quality control and spatiotemporal regulation is the basis of their proper function (41, 55). Intriguingly, we observed specific disruption of heterodimers whereas homodimerization remained unaffected. These patterns could be related to KLHL6 regulation and could give rise to dominant negative features of especially the KLHL6^L65P mutant with poor binding to Cullin-3. However, other phenotypic changes and the opposing impact of Kelch domain mutations warrant further explanations. Some aberrant characteristics of KLHL6 mutants might not relate to perturbed BCR surface levels and could arise from disruptions relating to other KLHL6 interactors, such as BANK1 and LuZP1, or they could be linked to previously described mRNA dysregulation and activation of NF-κB signaling through Roquin-2 (33). Our findings provide foundations for future biochemical investigations relating to the regulation of KLHL6-CRLs E3 ligase function that could expose completely new therapeutic strategies to KLHL6 dysregulated diseases.

Multiple mutations in different genes contribute to lymphoma growth by disrupting surface BCR homeostasis. Here, we linked mutational dysregulation of KLHL6 as a mechanism to oncogenic BCR signaling. Distinct impacts of different KLHL6 variants on BCR plasma membrane levels and signaling raise questions for future studies to determine what are the BCRs coupled with these different mutations and what antigens, whether self or foreign, they can recognize. We expect our findings will unlock the exploration of therapeutic BCR targeting following KLHL6 dysregulation in B-cell neoplasia and autoimmunity.

The study was approved by the Ethics Committee of Helsinki University Hospital (HUH), Finland; the National Authority for Medicolegal Affairs, Finland; and Institutional Review Board. The patients treated in Nordic phase II trial protocols signed an informed written consent before study enrollment and sample collection was performed in accordance with the Declaration of Helsinki.

We utilized previously prepared tissue microarrays (TMA) constructed from residual formalin-fixed, paraffin-embedded (FFPE) tumor tissue from diagnostic DLBCL biopsies treated in the Department of Oncology, HUH (56). One to four cores were punched from representative regions of diagnostic tumor tissues. The patients with evaluable TMA cores that had available molecular data produced previously (34) comprised the discovery cohort of the study. Additionally, whole-tissue FFPE sections from available tissues of patients with DLBCL treated according to Nordic dose-densified chemoimmunotherapy protocols were utilized in the training set of tumors to optimize staining procedure, characterize KLHL6 staining prior to discovery set analysis, and as an expansion series to characterize KLHL6 mutants on the protein level by IF and IHC microscopy. These samples were residual tissues from diagnostic biopsies of young (<65 years of age) clinically high-risk (age adjusted IPI ≥ 2) patients with primary DLBCL. Molecular subtype by Hans’ algorithm is annotated in the representative figures.

Mutational data for the discovery set was available from the original publication and the RNA sequencing gene expression data of log₂-transformed fragments per kilobase of transcript per million mapped reads (FPKM) per gene was obtained from the corresponding authors [European Genome-Phenome Archive: EGAD00001003600; ref. 34]. Variant calls from a recent re-analysis of the discovery cohort exomes were available and prioritized in the study (bioRxiv 2023.11.21.567983) and complemented with one in-house validated KLHL6 L65P mutation (Supplementary Fig. S8G). LymphGen classifications (30) were determined using the online tool with revised mutations, without considering copy number alterations and using BCL2 and BCL6 translocations determined previously for available cases (56). Additional DLBCLs with BTB domain mutations were screened with capillary sequencing of KLHL6 exon one from fresh-frozen tissue-derived DNA available from expansion set of tumors and one Kelch domain mutation was detected from targeted deep sequencing data (Supplementary Fig. S8H). Meta-analysis of KLHL6 mutations was performed by pooling together KLHL6 mutations as reported or obtained from corresponding authors in the publications listed in Supplementary Table S2A (31, 34, 38, 48, 57–62).

Uniform manifold approximation and projection analysis (UMAP) of 624 DLBCL transcriptomes (34) was performed using R package UMAP (version 0.2.10.0). B-cell ecotypes and cell states were obtained from supplementary information of Steen and colleagues (35). Single-cell RNA sequencing–based results of KLHL6 expression in various GC B-cell populations was characterized from supplementary files of Holmes and colleagues (63).

SuDHL4, SuDHL5, SuDHL8, dB, HBL1, Ri-1 (Riva), U2932, NuDUL-1, Ocily3, Ocily7, Ocily8, and Ocily19 were obtained from Professor Karen Dybkaer (Aalborg University Hospital, Denmark). SuDHL10 cell line was obtained from Dr. Caroline Heckman [Institute for Molecular Medicine Finland (FIMM), Helsinki Institute of Life Science (HiLIFE), University of Helsinki, Finland]. The cells were cultured at 37°C in 5% carbon dioxide and maintained in humid conditions in their respective culture mediums as listed in Supplementary Table S6. Cell line authentication was performed for select cell lines at the Institute of Molecular Medicine Finland (FIMM, HiLIFE Unit, University of Helsinki) using the GenePrint25 system (Promega). The authentication of HBL1 with no publicly available reference genotype profile available was performed by confirming that cell line did not match with any of the available cell line profiles. The cells were confirmed negative for mycoplasma using the MycoAlert kit (Lonza).

Human embryonic kidney cells (293FT) were cultured in DMEM 10% (Supplementary Table S6) and plated the day prior transfection to Petri dishes (500,000 or 1,500,000 cells to dishes 5 or 10 cm in diameter, respectively) in antibiotic-free culture medium. For 10-cm plates, the lipid-based transfection reaction was set by first adding 36 μL DharmaFECT Duo (Dharmacon) transfection agent to 1,200 μL of room temperature (RT) Opti-MEM (Gibco) medium, and the mixture was incubated for 5 minutes. Then, the 6 to 12 µg plasmids were added mixing the reaction thoroughly, and micelles were allowed to form for 20 minutes in RT prior adding the mixture dropwise to cells. After 18 to 24 hours of incubation in cell incubator, the culture medium was removed and replenished with antibiotic-containing culture medium.

We used 293FT cells to produce lentiviruses for transduction. 293FT cells were kept from reaching confluence prior to virus production and plated on culture dishes with antibiotic-free medium on the day before transfection. After 24 hours, the cells were co-transfected with 2.25 to 4.5 µg lentiviral packaging plasmids (both gifts from Didier Trono, pCMVR8.74, Addgene #22036, or psPAX2, Addgene #12260), 0.75 to 1.5 µg envelope plasmids [a gift from Bob Weinberg, pCMV-VSV-G, Addgene #8454 or a gift from Didier Trono, pMD2.G (VSV-g), Addgene #12259], and 3 to 6 µg desired constructs, with the protocol described above. After an overnight incubation, the supernatant was removed and replenished with fresh antibiotic-containing culture medium. After 48 hours (72 hours from transfection), the lentivirus-containing supernatant was harvested with a syringe and run through a 0.45 μm filter. If not immediately used for transduction, the viral supernatant was aliquoted and stored in −70°C for later use.

Lymphoma cell lines were transduced with the spinoculation method. The target cells, passaged the day prior to transduction, were collected 50,000 to 250,000 cells per transduction, centrifuged and the supernatant was removed. Then, the cells were suspended to 500 to 1,000 μL of lentiviral supernatant supplemented with 8 μg/mL polybrene. Spinoculation was performed by centrifugating the cells in 24-well plates or conical tubes for 2 hours with 800 g in RT. After spinoculation, the viral supernatant was removed, and the cells were suspended and plated in fresh culture medium. Fourty-eight hours after transduction, the cells transduced with vectors containing antibiotic selection markers were supplemented with the respective antibiotics [puromycin (0.5 µg/mL)] for positive selection of transduced cells. Cells with green fluorescent protein (GFP) as selection marker were selected with fluorescent activated cell sorting (FACS).

The full-length human KLHL6 cDNA was cloned from cDNA synthetized from RNA extracted from SuDHL4 cell line with available sequencing data (Cancer Cell Line Encyclopedia) and in-house RNA sequencing data indicating abundant transcript expression encoding a WT protein (Uniprot ID Q8WZ60, 621 amino acids). Constructs with C-terminal Strep-HA and BioID2 tags were manufactured by restriction-ligation cloning and followed by their transfer into mammalian overexpression vector. Full-length coding sequence of KLHL6 with and without a stop codon was PCR amplified and cloned into a C-terminal Strep-tag-HA cassette in pcDNA3.1 backbone using KpnI and EcoRV restriction enzymes to produce nontagged KLHL6 and KLHL6-strHA cDNAs, respectively. Constructs of KLHL6 fused to C-terminal BioID2 were cloned into MCS-BioID2 backbone with restriction-ligation cloning with BspEI. For gateway cloning into lentiviral mammalian expression vector pLex_307 (a gift from David Root, Addgene #41392), the complete CDS of each construct was PCR amplified with gateway recombination compatible primers for recombination into entry vector (pENTR-223). Purified PCR products were subsequently destination cloned by L.R reaction (Gateway Cloning, Life Technologies, Thermo Fisher Scientific) into pLex_307. Additionally, the C-terminus V5-tagged constructs of KLHL7, KLHL26 and BANK1 were prepared from ORFeome collection of pENTR221 gateway compatible cDNAs by L.R reaction into destination vector pLex_307. The clones without stop-codons for BANK1 (clone id: 100008707), KLHL7 (100068501) and KLHL26 (100073736) were used. The gateway cloning was performed at the Genome Biology Unit supported by HiLIFE and the Faculty of Medicine, University of Helsinki, and Biocenter Finland.

Site-directed mutagenesis of tagged and nontagged KLHL6 constructs was performed for pertinent pENTR-223 entry vectors using QuikChange II Site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer’s protocol using half the reaction volumes. Primer sequences for mutagenesis were designed using the manufacturer’s Primer Design Program online tool (Supplementary Table S6). After the reactions, the resulting plasmids were Sanger sequenced for successful mutagenesis prior to their recombination into the destination vector pLex_307.

We designed single-guide RNA (sgRNA) constructs targeting KLHL6 and used nontargeting sgRNAs derived from the GeCKO V2 library (identifiers HGLibA_14895 and HGLibA_14096) as controls in CRISPR-Cas9 experiments to produce isogenic knockout cell line. Recombinant DNA encoding KLHL6 or non-mammalian sgRNA targets were cloned into lentiCRISPRv2GFP backbone [a gift from David Feldser (64), Addgene #82416] and the lentivirus particles were produced as described above. After transduction, SuDHL5 cells were cultured for multiple passages until the GFP positive cells were batch and single-cell sorted into U-bottom 96-well plates with 100 µL culture medium using Sony SH800 cell sorter. The resulting sorted batch was used to assess editing and knockout efficiency, and the appropriate number of single-cell clones were expanded and screened for isogenic frameshift mutations leading to premature stop codons with capillary sequencing.

Comparison between KLHL6 and KLHL7 silencing was performed in SuDHL4 cells using another set of sgRNAs (Supplementary Table S6). SuDHL4 cells were transduced with lentiCas9-EGFP (a gift from Phil Sharp and Feng Zhang, Addgene plasmid #63592) and an isogenic GFP+ single cell line was cloned as described above using the SH800 sorter. Then, two KLHL7 targeting guides from the Brunello library, two targeting KLHL6 and two control guides targeting intergenic regions of chromosome 2 and MYC were cloned to lentiGuide-Puro (a gift from Feng Zhang; Addgene plasmid #52963) vector, transduced to Cas9-EGFP+ SuDHL4 cell line and positively selected using puromycin. Gene-silencing efficacy of KLHL7 targeting guides was evaluated with capillary sequencing of target regions with Inference of CRISRP Editing (ICE) software (Synthego) indicating knockout scores of 71 and 75 for KLHL7.3 and KLHL7.4 sgRNAs, respectively. Finally, isogenic clones of edited cells were single cell cloned as described above, screened for frameshifts leading to premature stop codons with capillary sequencing and analyzed for their BCR phenotypes (Supplementary Fig. S5G).

Short hairpin RNAs (shRNAs) were used to knockdown KLHL6 expression in lymphoma cell lines. Bacterial clones with plasmids (pLKO.1) encoding KLHL6 targeting shRNA library (TRCN0000159143, TRCN0000159320, TRCN0000161116, TRCN0000161652, TRCN0000162118, TRCN0000164063) by the RNAi Consortium (TRC) were from Sigma-Aldrich MISSION shRNA library, distributed by Genome Biology Unit core facility. Knockdown efficiency was evaluated by qPCR and Western blotting, and shRNAs TRCN0000159143 and TRCN0000162118 referred in the manuscript as “sh_#9143” and “sh_#2118,” respectively, were selected and used in the representative knockdown experiments. Non-mammalian targeting scrambled (SHC002) and GFP targeting (SHC003) control shRNA plasmids (pLKO.1) obtained from the laboratory of Jorma Keski-Oja were used as RNAi controls.

Capillary sequencing for PCR amplified select tumor DNA and cellular DNA or plasmid DNA was performed at the Institute of Molecular Medicine Finland (University of Helsinki supported by HiLIFE and Biocenter Finland) using ABI3730XL DNA Analyzer (Applied Biosystems) and BigDye version 3.1 (Applied Biosystems) and the data were manually examined using the SnapGene Viewer tool (version 4.2.6).

Cell viability and growth was measured using CellTiter-Glo 2.0 Cell Viability Assay (CTG, Promega) according to the manufacturer’s protocol and the luminescence was measured with FLUOstar Omega Microplate Reader (BMG Labtech). For monitoring growth of SuDHL5 and U2932 cells transduced with different KLHL6 or control constructs, 1,000,000 live cells were plated in 10 mL of cell culture medium and baseline cell viability was measured after 24 hours. Then, the viability of the cells was measured every 24 hours using CTG in three technical replicates and the cell culture was topped up to 10 mL with fresh culture medium. The culture was maintained until confluency and the measured luminescence was normalized to the signal of the first measurement.

Lymphoma cells were cultured in their respective culture medium and passaged the day before the experiments. On the day of the stimulation, 25,000 cells per well were plated into white flat-bottom 96-well plates in 50 µL of their respective culture medium. Goat anti-human IgM F(ab′)2 (Southern Biotech) was first diluted in PBS and then mixed with cell culture medium to be added to the cells in 1:1 ratio for desired final concentrations (0.3, 1, 3, and 10 µg/mL). Unspecific purified goat IgG (Supplementary Table S6) at concentration 10 µg/mL was used for control cells. The viability of the cells was measured after 48 hours of stimulation using CTG. The measured luminescence was normalized to the signal of the control treatment.

Fluo-8 Calcium Flux Assay Kit (Abcam, ab112129) was used according to manufacturer’s instructions to assess kinetics of BCR cross-linking induced Ca²⁺ flux in SuDHL5 cells. For the experiments, 250,000 SuDHL5 cells per well were plated on black 96-well plates coated with Poly-L-lysine and incubated in dye-loading solution for 1 hour (30 minutes in cell culture incubator and 30 minutes in RT). Then, the plates were centrifugated once, the supernatant was removed, 100 µL of Hank’s Balanced Salt Solution (HBSS, Gibco) was added and the prepared plates that were placed in FLUOstar Omega microplate reader. In the experiment, the wells were injected with 100 µL goat anti-human IgM F(ab′)2 or unspecific goat IgG (10 µg/mL) using plate reader’s injectors and the fluorescence was monitored at excitation/emission 490/525 nm per second for 5 minutes.

Cells were cultured as described above and passaged the preceding day before harvesting. Proteins were extracted by lysing the cell pellets in RIPA Lysis Buffer with brief sonication. The lysate was clarified by centrifugation and the protein concentrations were determined using DC Protein Assay and measured with FLUOstar Omega. The separately processed cytoplasmic and nuclear protein fractions were extracted from 10 μL cell pellets using NE-PER Nuclear and Cytoplasmic Extraction Reagents kit (Thermo Fisher) according to the manufacturer’s protocol. All lysates were supplemented with protease and phosphatase inhibitor cocktails if phosphoproteins were investigated. Laemmli sample buffer supplemented with ß-mercaptoethanol was added to the lysates, which were then denaturated in a 95°C heat block for 5 minutes.

The samples were run in 4% to 15% precast gradient gels (Bio-Rad). After electrophoresis (100V, 75–80 minutes) in electrophoresis chambers, the proteins were transferred to a 0.2 µm nitrocellulose membrane with a Trans-Blot Turbo Transfer system (3–7 minutes, 2.5 A, up to 25 V). Total protein load was detected with No-Stain Protein Labeling Reagent (Invitrogen) for normalization. The membrane was blocked and the indicated primary and secondary antibodies were diluted in 5% milk in TBS-Tween (0.1% Tween 20). The detection of proteins was performed using Odyssey Fc Imaging System and Image Studio Lite software (LI-COR Biosciences; Supplementary Table S6) or Azure 500 Imaging System (Azure Biosystems; Supplementary Table S6). Quantification of the protein levels and adjustment of images for publication were performed using the standard tools of AzureSpot Pro or Image Studio Lite software with default settings (Supplementary Table S6). All unprocessed Western blot images of the figures are shown in Supplementary file Unprocessed Western blots, and raw image files and additional blots can be inquired from the corresponding authors.

qPCR to determine KLHL6 mRNA transcript levels in cell lines and genetically manipulated lymphoma cells. RNA was isolated with Nucleospin RNA kit (Macherey–Nagel) according to manufacturer’s protocol and the complementary DNA (cDNA) was synthetized from 500 ng of RNA using iScript cDNA synthesis kit (Bio-Rad). The cDNA was prepared according to the manufacturer’s protocol and diluted ½ with ddH₂O. The qPCR reactions were prepared on 96-well plates in triplicate using 5 μL of TaqMan Universal PCR Master Mix and 0.5 μL of TaqMan probes (Applied Biosystems), 2.5 μL of molecular grade H₂O and 2 μL of diluted cDNA and run with ABI 7500 Fast Real-time PCR (Thermo Fisher) equipment. The TaqMan probe assays (Thermo Fisher) used for KLHL6 (Hs01092201_m1) were normalized to reference housekeeping assay for GAPDH (Assay ID Hs02786624_g1). The data was analyzed and the relative quantifications were calculated using 7,500 Fast Real-Time PCR System Software (Applied Biosystems, version 2.3).

IHC staining was performed using 4- to 5-μm-thick tissue and TMA FFPE sections attached on microscopy slides. After deparaffinization with xylene, the sections were permeabilized using Tris-EDTA buffer (10 mmol/L Tris base, 1 mmol/L EDTA solution, 0.05% Tween 20, pH 9.0) and blocked with hydrogen peroxide. Primary antibodies were diluted in Normal antibody Diluent (ImmunoLogic) and incubated overnight in +4°C. Unconcentrated mouse anti-human KLHL6 antibody (clone 92C, CNIO, Supplementary Table S6) was used at 1:3 dilution (65). After primary antibody incubation, the sections were treated with BrightVision anti-mouse HRP (ImmunoLogic), ImmPACT Diaminobenzidine Peroxidase Substrate Kit (Vector), and stained with Mayer’s Hematoxylin Solution (Sigma-Aldrich). The slides were washed two times with PBS for 5 minutes between all the steps, and finally for 10 minutes under running water after hematoxylin treatment.

IHC was analyzed together with an expert hematopathologist using a dual-head microscope. Light microscopy figures for publication were obtained from stained tissue sections scanned with Pannoramic 250 Flash III, with 20× air objectives (3DHISTECH, Budapest, Hungary) and analyzed in Pannoramic viewer environment (version 1.15.4).

Lymphoma cells were passaged the day before their harvest for IF analysis. Briefly, the cells growing in their respective standard culturing conditions were collected and washed once with ice-cold PBS. Then the cells were fixed with 4% paraformaldehyde in PBS gently rocking on ice for 15 minutes. Then the cells were washed two times with PBS and H₂O prior to their attachment on microscopy slides. Round areas (∼1 cm ∅) for fixed lymphoma cells were encircled with hydrophobic marker and the microscopy slides were coated with poly-L-lysine (Sigma) by incubating for 5 minutes. Desired numbers of cells (typically 200,000) suspended in PBS were plated on coated regions and were allowed to sediment, attach, and the glasses were allowed to dry o/n. The attached cells were washed with PBS and then permeabilized with 0.1% Triton X-100 in PBS by incubation of 5 minutes and washed three times with PBS. Then, the cells were blocked with 3% bovine serum albumin (BSA) in PBS, washed once and stained with primary antibodies diluted in 3% BSA-PBS. After 1 hour of incubation in RT, the cells were washed three times with PBS and incubated for 30 minutes protected from light in RT with secondary antibodies in 3% BSA-PBS. After one wash with PBS, the DNA was stained with Hoechst (2 µg/mL) for 5 minutes. Then the glasses were washed twice with PBS and the slides were dipped in ddH₂O and allowed to dry briefly. Finally, the cells were mounted with ProLong Gold (Invitrogen) and coverslips, and the glasses were allowed to dry before IF microscopy analysis.

IF analysis of KLHL6 and select markers in human tissues was performed using two or three fluorochromes with a modified staining protocol described previously (66). Tissue and TMA sections (4–5 μm thick) were prepared, deparaffinized and blocked as described above in the IHC section. After endogenous peroxidase blocking, the slides were blocked with 15 minutes incubation in 10% normal goat serum (NGS, Gibco) TBS-T. Then, the 100 μL of diluted primary antibody (mouse anti-human KLHL6) in 10% NGS TBS-T was added and the slides were incubated in +4°C humid conditions o/n. The slides were then washed three times with TBS-T (5 minutes) and 100 μL of secondary antibody (goat anti–mouse HRP, Brightvision) was added. The slides were incubated 30 minutes and washed three times with TBS-T. Then 100 μL of tyramide conjugated with Alexa Fluor 488 was added according to manufacturer’s instructions (Thermo) and incubated for 15 minutes RT until the reagent was washed away with three TBS-T washes and once with H₂O. After the tyramide signal amplification reaction, the first primary antibody was denatured by repeating the antigen-retrieval step. Then the slides were washed once with H₂O and once with TBS-T and blocked again with 10% NGS TBS-T. Next set of primary antibodies was added in 10% NGS TBS-T and incubated o/n. The next morning, the slides were washed three times with TBS-T and the secondary antibodies (diluted 1:300, AF647 and AF750) and 4′,6-diamidino-2-phenylindole2-phenylindole (DAPI; at 1.7 μg/mL concentration) in TBS-T were incubated on the slides for 45 minutes in RT. After three washes with TBS-T and once with H₂O, the slides were allowed to dry 1.5 hours in +37°C until mounting with ProLong Gold reagent and placing coverslips. Antibodies used for IF analyzes and their concentrations in parenthesis were as follows: mouse anti-human KLHL6 (CNIO, 1:3), rabbit anti-human CD79B (1:250), rabbit anti-human GM130 (1:3,000 for lymphoma cell lines, and 1:500 for FFPE tissues), rat anti-human Ki-67 (1:150), rabbit anti-human CD23 (1:100), rabbit anti-human SEC16A (1:100).

Colocalization analyses were performed with CellProfiler (version 3.1.8). Pixel-based colocalization analysis was conducted for the whole images and six to seven images were analyzed per construct. All the thresholds were determined using Otsu two-class thresholding method, and the images were masked accordingly to exclude extracellular spaces and the nuclei of the cells. Mander’s colocalization coefficients were used as a measure of colocalization.

The slides were imaged using a Zeiss Axio Imager upright epifluorescence microscope with Zen 2.3 lite acquisition software. Optical sectioning microscopy was performed using the same microscope equipped with ApoTome (Zeiss). The objectives used were EC Plan-Neofluar 40×/1.30 Oil, Plan Apochromat 63×/1.40 Oil, and EC Plan-Neofluar 100×/1.30 Oil. The samples were imaged using HXP 120 V fluorescent light source and acquired using Hamamatsu Orca Flash 4.0 LT B&W camera. The imaging was performed at the Biomedicum Imaging Unit, University of Helsinki, with the support of Biocenter Finland.

IF scanning of the TMAs and select tissue sections was performed with Zeiss Axio Scan.Z1 (Zeiss, Germany) equipped with EC Plan-Neofluar 20× objective (NA 0.8), Colibri seven light source and 112 HE LED filter set. Quantification of CD79B signal was performed in Zen 2.6 lite software (Zeiss). Briefly, representative TMA punches were quantified for fluorescence in representative regions of tumor cells for AF647 signal with Zen 3.1 lite software.

Interaction screens with AP with Strep-tag and proximity labeling (PL) with BioID2-tag were based on at least four biological replicates with 1 to 1.5 × 10⁹ cells per pull-down. The cells stably expressing the baits were cultured in their standard conditions into large volume, washed and the resulting pellet was snap-frozen with liquid nitrogen. Cells pellets prepared for biotinylation-based PL with BioID2-tag experiments were cultured with no additional biotin supplementation as RPMI contained biotin (0.2 mg/L) and no significant increase in protein biotinylation was observed with biotin supplementation.

For AP with Strep-tag, the cell pellets were thawed and lysed in 3 mL of ice-cold lysis buffer [0.5% IGEPAL, 50 mmol/L HEPES (pH 8.0), 150 mmol/L NaCl, 50 mmol/L NaF, 1.5 mmol/L NaVO3, 5 mmol/L EDTA, with 0.5 mmol/L PMSF and protease inhibitors (Sigma-Aldrich). For PL protein purification approach with BioID2-fused baits, the cell pellets were thawed in 3 mL of ice-cold lysis buffer [0.5% IGEPAL, 50 mmol/L HEPES (pH 8.0), 150 mmol/L NaCl, 50 mmol/L NaF, 1.5 mmol/L NaVO3, 5 mmol/L EDTA, 0.1% SDS, 0.5 mmol/L PMSF, and protease inhibitors]. Proximity lysates were sonicated before adding Benzonase Nuclease (Santa Cruz Biotechnology, sc-202391). Cleared lysate was obtained by centrifugation (16,000 g for 20 minutes), and the lysate was subjected to a one-step purification using Strep-Tactin Sepharose resin (IBA). The purified protein complexes were reduced, alkylated, and digested to peptides for MS analysis. The peptides samples were desalted with C18 macrospin columns (Nest Group) and resuspended for mass spectrometry analysis. A detailed description of the method used here can be found in a previous protocol (67).

The study was carried out using a Q Exactive Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Fisher) with Xcalibur version 2.0.7 SP1 (Thermo Fisher) linked to an EASY-nLC 1000-system through an electrospray ionization sprayer (Thermo Fisher). Four to eight biological replicates per condition were used for each sample for interaction discovery experiments and an equal amount of peptide samples was loaded in the instrumentation for each analysis. Peptides were eluted and separated with a C-18-packed pre-column and an analytical column at a flow rate of 300 nL/minutes using a 60-minute buffer gradient from 5% to 35% buffer B (buffer B: 0.1% formic acid in 98% acetonitrile and 2% HPLC-grade water; buffer A: 0.1% formic acid in 2% acetonitrile and 98% HPLC-grade water), followed by a 5-minute gradient from 35% to 80% buffer B and a 10-minute gradient from 80% to 100% buffer B. Peptides were analyzed in a data-dependent acquisition mode with a 70,000 resolution FTMS complete scan (200–2,000 m/z) and a higher energy collision dissociation scan of the top 20 most abundant ions. For protein identification, Thermo.RAW files were searched using the Sequest search engine against human entries in the UniProtKB/SwissProt database (http://www.uniprot.org/) with 15 ppm MS1 tolerance and 0.05 Da fragment mass tolerance. Carbamidomethylation of cysteine was regarded as a static alteration, whereas methionine oxidation, biotinylation of lysine, and N-termini biotinylation were defined as variable modifications. The provided findings were based on peptides assigned with high confidence in Proteome Discoverer (Thermo Fisher) and a 1% false discovery rate (FDR) by Percolator. Protein purifications and mass spectrometry were performed in the Proteomics Unit at the Institute of Biotechnology supported by University of Helsinki, HiLIFE, and Biocenter Finland.

As a statistical strategy for identifying high-confidence interactions from AP-MS and PL-MS data, the online tool (http://proteomics.fi/) integrated with Significance Analysis of INTeractome (SAINT) express version 3.6.0 (68, 69) was utilized. Four to eight EmGFP control runs (four Strep-tagged and four nontagged for AP approach, BioID2-tagged for PL experiments) were used as control counts for each hit. High-confidence interactions (HCI) were defined by an estimated Bayesian FDR of ≤0.01. Additionally, the CRAPome database (70) with a cutoff frequency of ≥20% (≥82), except for an average spectral count fold change ≥3 was used to further flag the possible false positive hits. As a complementary approach to SAINT, we applied DESeq2 (71) that normalizes and accounts for dispersion in count data from individual purifications to determine significant differentially interacting proteins between different conditions. R package DESeq2 (version > 1.24.0) was used in the analysis. All the conclusions were evaluated with both models and manual examination of spectral counts.

SuDHL5 cells stably expressing Strep-HA-tagged KLHL6 and EmGFP constructs were cultured into large volumes to produce ∼750 million cells per experiment. Prior to stimulation, 1.5 × 10⁹ cells were collected and sensitized for stimulation by incubation in low serum (0.5%) culture medium for 30 minutes in cell culture incubator. The sensitized cells were then collected, centrifuged, and the medium was removed. Then, the cells were suspended in 10 mL of PBS (RT), split into two 15 mL conical tubes at 1:1 ratio, centrifuged and the PBS was removed to obtain ∼ 600 μL cell slurry for parallel stimulation and non-stimulation control reactions. Then, 600 µL of goat anti-human IgM F(ab′)2 (0.5 mg/mL) or unspecific purified goat IgG unlabeled (diluted to 0.5 mg/mL with PBS) were added to cells or for stimulation and control experiments, respectively, and the reaction was incubated in RT for 60 seconds before snap-freezing in liquid nitrogen. For stimulation of control cells expressing EmGFP baits, only stimulations with anti-human IgM F(ab′)2 was performed. Western blot samples to control successful stimulation containing 10 μL of cell slurry were collected prior to and immediately following the adding of the reagent, and the control samples were snap-frozen together with the falcon tube at 60 second landmark. The samples were stored in −70°C until lysis and the AP-MS was performed in one batch according to the protocol described above.

The surface expression of the BCR components CD79A, CD79B, and IgG or IgM in SuDHL4, SuDHL5, OciLy7, and U2932 cells was measured by flow cytometry. The cells were stained with the following antibodies (Supplementary Table S6): anti–CD79A-PE, anti–CD79B-PerCPCY5.5, anti–CD19-APC, and anti–IgM-Pacific Blue or anti–IgG-FITC on ice for 30 minutes before analysis on a FACS Canto II (BD).

Activation of BCR signaling was performed as previously described (50, 72). Briefly, cell lines were distributed at 5 million cells/mL and rested for 75 minutes, then stimulated with anti-IgM F(ab′)₂ (Jackson ImmunoResearch) or anti-IgG F(ab′)₂ (Jackson ImmunoResearch) at a final concentration of 10, 5, 2.5 µg/mL or left unstimulated for 5 minutes. Cells were fixed by adding paraformaldehyde to a 1.6% final concentration for 5 minutes at RT, then washed with PBS, and permeabilized with ice-cold methanol (final concentration > 90%) and stored in −80°C.

On the day of flow cytometry analysis, cells were rehydrated by washing twice in PBS. For barcoding, the four different activation conditions for each isogenic cell line were resuspended in PBS, and each well was incubated with a specific concentration of Pacific Blue dye (0.5, 0.125, 0.031, and 0 ng/µL) for 30 minutes in the dark. Antibody staining was performed at RT for 30 minutes bere washing and running the samples on a FACS Canto II (BD). The following antibodies from BD were used (Supplementary Table S6): anti–p-BLNK-Ax647, anti–p-BTK-Ax647, anti–p-LCK-PE, anti–p-PLC-γ-Ax488, anti–p-SYK-PE, anti–p-ERK-Ax488, anti–p-STAT5-Ax488, anti–p-p38-PE, and anti–p-AKT1-Ax647. Cytobank Software was used for data analysis, and relative phosphorylation changes were calculated using arcsinh transformation of median fluorescence intensity (MFI) relative to unstimulated cells.

Statistical analyses were performed in R (version ≥ 3.6.1) and GraphPad Prism 9.5.1. Details about statistical testing is described in the figure legends and were nonparametric and two-sided unless otherwise indicated. P values < 0.05 were considered statistically significant and the significance is denoted in the figures as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001.

Survival analysis of the discovery cohort was performed using R packages. Survival analyses were performed with R package survival (version 2.44-1.1). Kaplan–Meier estimates were prepared and visualized with R package survminer (version 0.4.6) and cox regression models; HRs and 95% confidence intervals were visualized with forestplot R package (version 1.9).

The figures for publication were prepared in CorelDRAW 2023 (version 24.4.0.636, Corel Corporation). In general, plots were produced in R environment using ggplot2 R package (version ≥ 3.2.1). Protein–protein interaction data were imported into Cytoscape (Supplementary Table S6) for interaction network visualization. Modeling of KLHL6 3D structure by Alphafold was examined, and the images were acquired in Chimera (version 1.13.1; Supplementary Table S6). Oncoprints and lollipop plots were produced with ComplexHeatmap R package (version ≥ 2.0.0; ref. 73) and maftools R package (version 2.16.0), respectively, and finalized in CorelDraw.

MS data have been deposited to MassIVE (https://massive.ucsd.edu/) with web access MSV000092728. The protein interactions from this publication have been submitted to the IMEx consortium (http://www.imexconsortium.org) through IntAct. Other data that support this study and script to reproduce the analyses are available from the corresponding author upon reasonable request.

L. Meriranta reports grants from Orion Research Foundation, Paulo Research Foundation, Biomedicum Helsinki Foundation, Ida Montin Foundation, and Emil Aaltonen Foundation during the conduct of the study. I. Spasevska reports grants from Foundation KG Jebsen Centre for B-cell malignancies during the conduct of the study. O. Dufva reports grants from Incyte and Gilead Sciences and personal fees from Sanofi outside the submitted work. J.H. Myklebust reports grants from The Research Council of Norway and The Foundation KG Jebsen during the conduct of the study. S. Leppä reports grants from Finnish Medical Council, Sigrid Juselius Foundation, Finnish Cancer Organizations, and Helsinki University Hospital during the conduct of the study, as well as personal fees from AbbVie, Gilead, Incyte, Sobi, and Orion, grants and personal fees from Genmab, Novartis, and Roche, and grants from HUTCHMED, Bayer, and Bristol Myers Squibb outside the submitted work. No disclosures were reported by the other authors.

L. Meriranta: Conceptualization, formal analysis, investigation, visualization, methodology, writing–original draft. S. Sorri: Conceptualization, formal analysis, investigation, visualization, methodology, writing–review and editing. K. Huse: Formal analysis, investigation, visualization, methodology, writing–review and editing. X. Liu: Formal analysis, writing–review and editing. I. Spasevska: Formal analysis, investigation, writing–review and editing. S. Zafar: Formal analysis, writing–review and editing. I. Chowdhury: Formal analysis, writing–review and editing. O. Dufva: Formal analysis, investigation, writing–review and editing. E. Sahlberg: Formal analysis, investigation, writing–review and editing. L. Tandaric: Formal analysis, investigation, writing–review and editing. M.-L. Karjalainen-Lindsberg: Formal analysis, writing–review and editing. M. Hyytiäinen: Supervision, investigation, writing–review and editing. M. Varjosalo: Formal analysis, supervision, funding acquisition, writing–review and editing. J.H. Myklebust: Formal analysis, supervision, investigation, methodology, writing–review and editing. S. Leppä: Conceptualization, resources, supervision, funding acquisition, project administration, writing–review and editing.

The authors thank Anne Aarnio and Sini Miettinen for expert technical assistance and Suvi-Katri Leivonen for commenting on the manuscript. We thank the FIMM Digital Microscopy and Molecular Pathology Unit supported by Helsinki University and Biocenter Finland. Protein purifications and mass spectrometry were performed in the Proteomics Unit at the Institute of Biotechnology supported by the University of Helsinki, HiLIFE, and Biocenter Finland. This study was funded by grants from the Finnish Research Council (S. Leppä), Finnish Cancer Organizations (S. Leppä), Sigrid Juselius Foundation (S. Leppä), University of Helsinki (S. Leppä), and Helsinki University Hospital (S. Leppä). L. Meriranta was supported by grants from Ida Montin, Biomedicum Helsinki, Orion Research, Emil Aaltonen, and Paulo foundations. M. Varjosalo was supported by the Finnish Cancer Organizations.