TRAF6 interacts directly with P379VQLSY384 of LMP1
We examined all TRAF proteins involved in LMP1 signaling regarding their potential to directly bind to the LMP1 signaling domain (Figures 1A and 1B, Supplementary Figure 1A). The purified recombinant TRAF domains of TRAF1, 2, 3, and 5 interacted with P204QQAT of CTAR1 in pull-down assays with glutathione S-transferase (GST)-coupled LMP1181−386 (Figure 1B). Mutation of P204xQxT into A204xAxA abolished LMP1 binding of TRAF1, 2, and 5. Residual amounts of TRAF3 were recruited by the A204xAxA mutant, which can be explained by contacts of TRAF3 with LMP1 residues adjacent to the P204xQxT core motif 55.
Investigating the interaction between LMP1 and TRAF6 we made the surprising observation that also recombinant His-TRAF6310−522, which includes the TRAF domain of TRAF6, was efficiently recruited by GST-LMP1 (Figure 1B). In contrast to all other TRAF proteins tested, TRAF6 recruitment to LMP1 was not affected by mutation of CTAR1, but was eliminated by the exchange of tyrosine 384 into glycine. In accordance with this finding, Flag-TRAF6 wildtype only co-immunoprecipitated with HA-LMP1 from HEK293 cells if CTAR2 was intact (Supplementary Figure 1B). These experiments provided the first evidence for a direct protein-protein interaction as the molecular basis of TRAF6 recruitment to LMP1.
To further substantiate this result and to narrow down the LMP1 sequences involved in LMP1-TRAF6 interaction, we tested the ability of His-TRAF6310−522 to interact with immobilized LMP1-derived peptides, which incorporate CTAR1 or CTAR2 sequences (Figure 1C and Supplementary Figure 1C). Recombinant His-TRAF2311−501 was used as control. Specificity of TRAF interaction was confirmed by including peptides, which harbored alanine exchanges within the TRAF2-binding motifs of CD40 (P250VQET to A250VAEA, peptides 1 and 2, respectively), LMP1 (P204QQAT to A204QAAA, peptides 6 and 7), and the TRAF6 binding motif of CD40 (Q231EPQEINF to A231EAQAINF, peptides 1 and 3). CD40-derived amino acids 244-273 lacked the TRAF6 binding site (peptide 4). Additional mutation of the TRAF2-binding motif within peptide 4 resulted in peptide 5. We have shown previously that amino acids 371-386 of LMP1 are sufficient to induce TRAF6-dependent CTAR2 signaling 30. To determine whether these sixteen amino acids contain the complete TRAF6 binding site of LMP1, they were included as peptide 8. Within peptide 8, Y384 and Y385, or the cryptic TRAF interaction motif P379xQxS were mutated (peptides 9 and 10, respectively). Further, CTAR2 amino acids 357-386 were spotted (peptide 11), in which P379xQxS was mutated (peptide 12).
Both TRAF2 and TRAF6 specifically interacted with their designated binding sites within CD40, confirming accuracy of the peptide array (Figure 1C). Further, TRAF2 bound to P204QQAT of CTAR1 (peptides 6 and 7), but not to CTAR2 (peptides 8 to 12), which excludes the possibility of direct TRAF2 interaction with the cryptic TRAF interaction motif of CTAR2. Notably however, TRAF6 was efficiently captured by the CTAR2 peptides 8 (16mer) and 11 (30mer), whereas it did not bind to CTAR1 (peptide 6). Mutation of Y384 and Y385 to AA (peptide 9) and P379xQxS to AxAxA (peptides 10 and 12) abolished direct TRAF6 binding to CTAR2 (Figure 1C).
We performed an alanine exchange mutagenesis scan from G378 to Y385 of LMP1 to precisely map the residues that are involved in TRAF6 binding. We developed a highly reliable mix-and-measure screening assay for the LMP1-TRAF6 interaction based upon the Perkin-Elmer AlphaScreen technology, by which the effects of mutations on this protein-protein interaction (PPI) can be detected and quantified directly (Figure 1D). Light emission at 520-620 nm is directly proportional to the affinity of the two protein components of the assay. Each of the LMP1 amino acids P379, V380, Q381 and Y384 was essential for direct TRAF6 recruitment to the LMP1 signaling domain (Figure 1E). Mutation of Y385 had only minor impact on TRAF6 binding at the lowest TRAF6 concentration tested in the assay (100 nM), whereas the side chains of G378, L382 and S383 were dispensable for interaction. Of note, the resulting TRAF6 binding sequence P379VQxxY exactly matches the NF-κB- and JNK-inducing region of CTAR2 37, 39. This finding strongly suggested that the direct binding of TRAF6 to this sequence is in fact the molecular basis for CTAR2 signaling.
TRAF6 showed a weaker affinity for LMP1 as compared to CD40. The KD of His-TRAF6310−522 interaction with GST-CD40 was 17.8 ± 4 nM in contrast to 77.1 ± 21.7 nM with GST-LMP1, determined by the AlphaScreen PPI assay (Figure 1F). Confirming our previous data, mutation of LMP1 P379xQxxY into A379xAxxA abolished TRAF6 binding. Analogous mutation of the TRAF6 binding motif within CD40, which was included as control, resulted in a loss of TRAF6 interaction as well (Figure 1F).
Alignment of the consensus TRAF6 interaction motif PxExxF/Y/D/E of cellular receptors 45, 51 with the newly identified TRAF6 binding sequence of LMP1 revealed high similarity, with the exception of one striking difference at the central position P0 (Figure 1F). Cellular TRAF6-recruiting sequences carry a glutamic acid at P0 45, 56–58, whereas this position is occupied by glutamine in LMP1. Remarkably, glutamic acid at P0 of the TRAF6 binding motif of CD40 cannot be replaced by any other amino acid, including glutamine, without losing affinity to TRAF6 56. We tested the effect of converting the TRAF6 binding motif of LMP1 into the cellular consensus motif by Q381E mutation. The resulting LMP1 Q381E mutant is capable of binding TRAF6 with significantly enhanced affinity as compared to wildtype LMP1 (Figure 1G, compare to GST-LMP1 wildtype of Figure 1F). Q381E mutation reduced the KD from 77.1 ± 21.7 nM to 8.1 ± 0.9 nM, which is even lower as the KD of TRAF6 interaction with CD40. This may suggest that additional interactions of TRAF6 with LMP1 beyond P0, which are absent in the CD40-TRAF6 complex, stabilize LMP1 interaction with TRAF6 and allow P0 being occupied by glutamine.
Position P3 of LMP1's TRAF6-binding motif is filled by Y384, which has a critical role in LMP1 signaling and viral cell transformation 8, 19, 37, 39. In cellular TRAF6-interacting receptors this position can be occupied by an aromatic or acidic amino acid 45. Accordingly, a permutation scan at P3 of CD40 showed that TRAF6 binding to CD40 still occurs if F238 is mutated into tyrosine or tryptophan 51. To test variability at the P3 position of LMP1, we introduced a Y384F mutation resembling P3 of the TRAF6 binding motif of CD40. The Y384F exchange was not only tolerated by LMP1, but even improved the affinity of LMP1 to TRAF6 (Figure 1G). As expected, Q381A and Y384A exchanges abolished TRAF6 binding (Figure 1G). Taken together, TRAF6 is directly recruited by the JNK- and NF-κB-inducing sequence P379VQLSY within CTAR2 and is, thus, the first identified cellular factor whose binding site exactly matches the signaling-active site of CTAR2. In contrast to cellular receptors, this motif contains glutamine at the central P0 position, likely facilitated by unique structural characteristics of the viral LMP1-TRAF6 complex.
Arginine 392 of TRAF6 discriminates between LMP1 and CD40
To examine whether LMP1 binds to the same region at the surface of TRAF6 as cellular receptors, we mutated amino acids within TRAF6 that are involved in interaction with P−2, P0, or P3 of CD40 and receptor activator of NF-κB (RANK, also known as TRANCE receptor) 45. The capability of the TRAF6 mutants R392A, K469A, F471A, or Y473A to bind to GST-LMP1 or GST-CD40 was analysed in AlphaScreen PPI experiments (Figure 2A). F471 and Y473 of TRAF6 build the binding pocket for amino acid P−2 of cellular receptors 45, 58. F471A or Y473A mutation caused a complete loss of TRAF6 binding to LMP1, as to CD40 (Figure 2A). Hence, this pocket forms an essential interaction with LMP1, most probably with LMP1 residue P379, which occupies P−2 of the P379VQxxY motif. Mutation of K469 into alanine had no effect on TRAF6 interaction with LMP1 or CD40. The side chain of K469 likely forms non-essential charge-charge interactions with the main chain carboxylate of P0 of CD40 45.
The TRAF6 mutant R392A revealed a striking difference regarding LMP1 and CD40 binding. R392 forms an amino-aromatic interaction with F238 at P3 of CD40 45. However, mutation of R392 into alanine had no impact on TRAF6 binding to CD40, whereas interaction with LMP1 was fully eliminated by this mutation (Figure 2A). R392 thus discriminates between LMP1 and CD40. This result suggested a different molecular architecture of the LMP1-TRAF6 complex as compared to CD40-TRAF6.
To verify the relevance of our findings on LMP1-TRAF6 interaction in vivo, we expressed Flag-tagged TRAF6 wildtype or the mutants R392A, K469A, F471A and Y473A together with HA-tagged LMP1 in HEK293 cells and performed co-immunoprecipitations of both proteins (Figure 2B and Supplementary Figure 2A). Confirming our previous results, each of the mutations R392A, F471A or Y473A, abolished TRAF6 interaction with LMP1, whereas K469A mutation had no negative effect on the interaction between both proteins in HEK293 cells. Confocal immunofluorescence studies in HeLa cells further verified these results (Figure 2C). Flag-TRAF6 wildtype and the K469A mutant co-located to a high extent with HA-LMP1 clusters, demonstrating their interaction with LMP1 in situ. In contrast, the TRAF6 mutants R392A, F471A and Y473A showed a strongly decreased co-localization with HA-LMP1, which was comparable to the LMP1Δ371-386 mutant lacking the TRAF6 interaction site (Figure 2C and Supplementary Figure 2B). In the absence of LMP1, all TRAF6 mutants showed a similar cytoplasmic distribution as TRAF6 wildtype (Supplementary Figure 2C). In summary, these results demonstrated that binding of TRAF6 to LMP1 involves the same TRAF6 residues in the cellular context as in our interaction studies with recombinant proteins, which strongly argues for the same and direct mechanism of LMP1-TRAF6 complex formation in vivo as in vitro.
Direct binding of TRAF6 to LMP1 is required for CTAR2 signaling
CTAR2 signaling is defective in TRAF6-deficient mouse embryonic fibroblasts (MEFs) and can be rescued by exogenous TRAF6 expression 19, 27, 28, 30. To demonstrate that direct interaction of LMP1 and TRAF6 is indeed the molecular basis for CTAR2 signaling, we tested the TRAF6 mutants that are defective in direct LMP1 binding for their potential to rescue CTAR2 signaling in NF-κB reporter assays in TRAF6-/- MEFs (Figure 3A). TRAF6-/- cells were transfected with the CTAR1 mutant A204xAxA, which signals towards NF-κB only through CTAR2, or the inactive double mutant A204xAxA/∆371-386, together with wildtype TRAF6 or the TRAF6 mutants R392A, F471A and Y473A. Comparable protein expression levels were confirmed by immunoblot analysis (Supplementary Figure 3A). In the absence of TRAF6, CTAR2 was unable to induce NF-κB reporter activity (Figure 3A, see w/o). As expected, expression of wildtype TRAF6 or the TRAF6 mutants alone (co-transfection with inactive A204xAxA/∆371-386) induced NF-κB to similar levels (grey bars), demonstrating that all mutants fully retained their downstream signaling capacity. However, only TRAF6 wildtype, but none of the binding-defective mutants, was able to rescue CTAR2 signaling to NF-κB (green bars). This result showed that the direct interaction of TRAF6 with LMP1 is critical for activation of CTAR2-mediated NF-κB signaling.
To further confirm this result, we retrovirally transduced TRAF6-/- MEFs, which stably express NGFR-LMP1, with TRAF6 wildtype or the TRAF6 mutants R392A, F471A and Y473A. NGFR-LMP1 is a fusion construct of the extracellular and transmembrane domains of the p75 nerve growth factor (NGF) receptor (NGFR) with the intracellular signaling domain of LMP1 13, 34. Instant NGFR-LMP1 activity can be triggered at the cell surface by incubation of the cells with an α-NGFR primary antibody and subsequent crosslinking by a secondary antibody (Figure 3B). Antibody crosslinking of NGFR-LMP1 caused a rapid degradation of IκBα, which is indicative for activation of the canonical NF-κB pathway in wildtype MEFs (Supplementary Figure 3B and ref. 34 34). In TRAF6-/- cells this pathway was defective (Supplementary Figure 3B). Exogenous expression of TRAF6 wildtype in TRAF6-/- cells restored activation of the canonical NF-κB pathway upon NGFR-LMP1 crosslinking (Figure 3C). In contrast, the TRAF6 mutants R392A, F471A and Y473A, which are unable to directly bind to LMP1, were also ineffective in rescuing canonical NF-κB activation by CTAR2 (Figure 3C). Taken together, our data demonstrated that CTAR2 only induces NF-κB if TRAF6 is directly recruited to CTAR2.
Molecular model of the LMP1-TRAF6 complex
Our experiments with the TRAF6 mutant proteins showed that LMP1 binds to the same PPI interface of TRAF6 as CD40 and other cellular receptors. To gain structural insights into the binding of TRAF6 to LMP1, we used Molecular Operating Environment to derive an in silico model of the LMP1-TRAF6 complex (Figure 4A). The sequence alignment between LMP1 and TRAF6-binding receptor peptides (see Figure 1F) showed no indication for significant structural differences in the proximity of position P−2 between LMP1 and cellular receptors, because this position is always occupied by a proline. Accordingly, P397 of LMP1 is located in the hydrophobic indentation formed primarily by TRAF6 residues M450, F471 and Y473 (Figure 4A). In line with this finding, mutation of the TRAF6 residues F471 and Y473 abolished LMP1 binding (see Figure 2).
At P−1 of LMP1, hydrogen bonds are formed between the main chain of V380 and the main chain of TRAF6 residue G472. The loss of TRAF6 binding of the LMP1 mutant V380A might be related to a loss of surface contacts between the side chains of LMP1 V380 and TRAF6 V474.
The most significant difference between LMP1 and the consensus TRAF6-binding sequence PxExxF/Y/D/E is that in cellular receptors P0 is occupied with glutamic acid, while LMP1 carries a glutamine at this position. Even more, an exchange of glutamic acid by glutamine is not tolerated at P0 of CD40 56. For cellular receptors it has been shown that the side chain carboxylate of glutamic acid at P0 forms a strong hydrogen bond network with the backbone amide nitrogen atoms of L457 and A458 45, 58. Also in our LMP1-TRAF6 model, hydrogen bonds are formed between Q381 at P0 of LMP1 and the amide NH atoms of L457 and A458 of TRAF6 (Figure 4B). Yet, due to the different charge of the side chains, the strength of Q381 interaction with TRAF6 is weaker compared to E235 of CD40 with TRAF6. Accordingly, the Q381E exchange significantly increases the affinity between LMP1 and TRAF6 (see Figure 1F).
Mutation of L382 at P1 to alanine does not impair LMP1-TRAF6 binding. This is consistent with the model, which indicates that hydrogen bonds at this position are formed by the peptide backbone with the TRAF6 residues G470 and R392 and are hence invariant to changes of the side chain (Figure 4C). In addition, R392 forms another hydrogen bond with S383 at P2 of the LMP1 main chain. These interactions explain the critical role of R392 for TRAF6 interaction with LMP1. When TRAF6 binds to CD40, R392 adopts a different conformation and forms a hydrophobic pocket for F238 at P3 45. The main chain carbonyl at P2 of CD40 is, thus, not within hydrogen bonding distance to R392. Whereas the side chain of F238 of CD40 shows a kinked orientation towards TRAF6, Y384 of LMP1 adopts a rather stretched conformation along the surface of TRAF6. This orientation enables Y384 at P3 of LMP1 to build non-polar surface contacts with R392 and V374 of TRAF6 over a large area. Taken together, the unique conformation of the LMP1-TRAF6 interface at P1 to P3 enables additional stabilizing contacts between LMP1 and TRAF6, which are not present in CD40-TRAF6, and may well explain why LMP1 tolerates glutamine at P0.
NMR spectroscopy reveals shifting of TRAF6 residues upon LMP1 binding
To confirm the binding position of LMP1 at TRAF6 proposed by our biochemical and modeling data, we recorded NMR spectra of TRAF6 in its free form as well as bound to the LMP1 peptide G378PVQLSYYD (Figure 5A). Addition of the peptide caused significant shifts as well as line broadening in some peaks in the TRAF6 spectra, a clear indication of binding. Based on a previously published partial backbone chemical shift assignment of TRAF6 59, several TRAF6 residues of interest could be assigned to peaks in the recorded spectra. Of those TRAF6 residues previously tested for their functions in LMP1 binding (see Figures 2 and 3), F471 and K469 are highlighted in both spectra (R392 and Y473 have not been assigned by Moriya and colleagues 59). Upon addition of the LMP1 peptide, the peaks corresponding to these residues are broadened beyond detection, indicating that these residues contribute strongly to binding. Because mutation of K469 had no effect on LMP1 interaction, this result supports a role of the K469 backbone in LMP1 binding. Next, the chemical shift pattern caused by the LMP1 peptide was calculated and plotted onto the modeled LMP1-TRAF6 complex (see Figure 4). The overall assigned shift perturbations caused by the addition of LMP1 peptide are clustered around the TRAF6 PPI surface and confirmed LMP1 binding at this position (Figure 5B).
LMP1-driven B lymphomas are strictly dependent on TRAF6
CTAR2 provides critical signals for effective growth transformation of primary B cells by EBV 8, 60. Because we showed that direct TRAF6 interaction with CTAR2 is required for CTAR2 signaling we next asked whether TRAF6 is necessary for proliferation and survival of LMP1-driven B cell lymphomas. To address this question, TRAF6 was targeted by an ex vivo CRISPR/Cas9 approach in the two LMP1-dependent B cell lymphomas LMP1-CL 37 and 40 derived from the transgenic CD19-Cre;R26LMP1stopfl:CD3εKO mouse model 9, 61. The effect of three different gRNAs targeting the gene of interest (GOI) TRAF6 on tumor cell survival was examined. gRNAs targeting LMP1 as positive or the intracellular adhesion molecule 1 (ICAM1) as negative controls were included in parallel transfections. Cell survival was monitored seven days post transfection as selection score of the gRNAs targeting the GOI versus a non-targeting (NT) gRNA directed against an irrelevant Rosa26 sequence (Figure 6A and Methods).
The knockout of LMP1 resulted in a drastic reduction of survival of the LMP1-CL 37 and 40 lymphomas (Figure 6B and 6C). This result was expected because both lymphomas had been selected for their dependence on LMP1 61 (see Methods). In contrast, targeting of ICAM1 did not affect lymphoma survival. More interestingly, we found that inactivation of TRAF6 by CRISPR/Cas9 caused a massive negative effect on lymphoma survival, which was comparable to the effect of LMP1 targeting itself (Figure 6B and 6C). Hence, both lymphomas are absolutely dependent on TRAF6, demonstrating a previously unappreciated critical role of TRAF6 in the survival of LMP1-dependent B lymphoma cells. These findings further suggested that the direct interaction between LMP1 and TRAF6 is an important factor for lymphoma development and may serve as a novel therapeutic target for inhibitory molecules.
Disruption of the LMP1-TRAF6 complex interferes with lymphoblastoid cell survival
We showed so far that the direct interaction of TRAF6 with LMP1 is required for CTAR2 signaling, and that TRAF6 has an important function in mediating survival of LMP1-driven B cell lymphomas. To prove that the direct LMP1-TRAF6 complex can be targeted in vivo, we aimed to inhibit TRAF6 recruitment to LMP1 by peptides to test the effect of LMP1-TRAF6 PPI disruption on LCL survival. Previously, cell-penetrating TRAF6 inhibitory peptides derived from the TRAF6 binding site of RANK had been used to inhibit receptor interaction of TRAF6 and RANK signaling 45, 62, 63. The RANK sequence RKIPTEDEY contains the motif PxExxY that binds to the same site of TRAF6 as CD40 45. Because LMP1 also interacts with this region at the TRAF6 surface, we reasoned that the RANK-derived peptide should be able to block TRAF6 interaction with LMP1. An alignment of the TRAF6 inhibitory peptide with CD40 and LMP1 sequences is shown in Figure 7A. We used a cell-penetrating version of this peptide, fused to the Antennapedia leader sequence, to inhibit TRAF6 interaction with LMP1. A peptide containing the leader sequence only served as negative control.
Indeed, the TRAF6 inhibitor peptide blocked interaction of TRAF6 and GST-LMP1 in AlphaScreen PPI assays with an IC50 of 177 nM, while the control peptide was inactive (Figure 7B). TRAF6 binding to GST-LMP1 wildtype and the A379xAxxA null mutant demonstrated the dynamic range of the assay and verified that LMP1-TRAF6 inhibition by the peptide was complete (Figure 7B). As expected, the inhibitor peptide had no effect on the recruitment of TRAF2 to LMP1 (Figure 7C). TRAF6 binding to CD40 was inhibited by the peptide as well, albeit with strongly reduced efficiency as compared to LMP1 (Figure 7D).
Finally, we examined the TRAF6 inhibitor peptide for its effects on LCL viability. Two lymphoblastoid cell lines, LCL721 and HA-LCL3, were incubated for three days in the presence of the TRAF6 inhibitor peptide or the control peptide, respectively. The EBV-negative Burkitt’s lymphoma cell line BL41 was included as negative control (Figure 7E). The TRAF6 inhibitor peptide, but not the control peptide, caused a severe reduction of cell viability in both LMP1-dependent LCLs, whereas no such effect was seen in LMP1-independent BL41 cells. This result corroborated our previous results regarding the relevance of TRAF6 function for the survival of EBV/LMP1-transformed cells. It further showed that the direct interaction of TRAF6 with LMP1 is essential for LMP1's pro-survival function and might therefore constitute a novel therapeutic target for inhibitors, for instance small molecule LMP1-TRAF6 PPI inhibitors.