|Home | About | Journals | Submit | Contact Us | Français|
NF-κB activation in the TNF, IL-1 and Toll-like receptor pathways requires Lys63-linked non-degradative polyubiquitination. A20 is a specific feedback inhibitor of NF-κB activation in these pathways and possesses dual ubiquitin editing functions. While the N-terminal domain of A20 is a de-ubiquitinating enzyme (DUB) for Lys63-linked polyubiquitinated signaling mediators such as TRAF6 and RIP, its C-terminal domain is a ubiquitin ligase (E3) for Lys48-linked degradative polyubiquitination of the same substrates. To elucidate the molecular basis for the DUB activity of A20, we determined its crystal structure and performed a series of biochemical and cell biological studies. The structure reveals the potential catalytic mechanism of A20, which may be significantly different from papain-like cysteine proteases. Ubiquitin can be docked onto a conserved A20 surface; this interaction exhibits charge complementarity and no steric clash. Surprisingly, A20 does not have specificity for Lys63-linked polyubiquitin chains. Instead, it effectively removes Lys63-linked polyubiquitin chains from TRAF6 without dissembling the chains themselves. Our studies suggest that A20 does not act as a general DUB but has the specificity for particular polyubiquitinated substrates to assure its fidelity in regulating NF-κB activation in the TNF, IL-1 and Toll-like receptor pathways.
A20 plays an important role in cellular functions by inhibiting cell activation, cytokine signaling and Toll-like receptor pathways 1; 2; 3; 4; 5; 6. It was originally characterized as an early response gene to TNF stimulation in human endothelial cells 7; 8 and its induction is mediated by NF-κB 9. As an NF-κB target gene, A20 is also induced in many other cell types by a wide range of stimuli such as viral infection, LPS treatment, CD40 engagement, CD30 activation and RANKL stimulation 10; 11; 12; 13; 14; 15.
The physiological function of A20 was first revealed by the failure to regulate TNF-induced NF-κB response in A20-deficient mice 16. A20 is also required for termination of Toll-like receptor and intracellular sensing responses to pathogens 17; 18; 19; 20. It is a potential therapeutic tool for the treatment of diseases where inflammatory responses are part of the pathogenic process. Mechanistically, A20 inhibits NF-κB activation at an upstream level leading to impaired activation of the inhibitor of κb (IκB) kinase (IKK). IKK is the kinase that phosphorylates NF-κB bound IκB, leading to IκB degradation and release of NF-κB for nuclear translocation and gene transcription 21. A20 likely directly modulates the function of upstream IKK activating signaling proteins such as TRAF6, RIP and possibly NEMO 22; 23; 24; 25.
NF-κB activation in the TNF, IL-1 and Toll-like receptor pathways requires Lys63-linked non-degradative polyubiquitination 26. Using biochemical purification and in vitro reconstitution, it was shown that TRAF6 is a specific ubiquitin ligase (E3) in Lys63-linked polyubiquitination of itself and other proteins 27; 28; 29 and that TAK1 is the immediate upstream kinase in TRAF6-mediated IKK activation in the IL-1 and Toll-like receptor pathways 30; 31; 32; 33; 34; 35. The E2 required in this ubiquitination is a dimeric complex of Ubc13 and Uev1A. TRAF6-mediated activation of TAK1 requires the TAK1 binding adapter protein TAB2, or its homologue TAB3 36; 37; 38; 39; 40; 41. In a simplified model, TRAF6-catalyzed ubiquitination of itself leads to recruitment and subsequent ubiquitination and activation of the TAB2/TAK1 complex. TAK1 in turn phosphorylates and activates IKK 29; 42. In the related TNF pathway, RIP, rather than TRAF6, becomes polyubiquitinated via the Lys63-linkage and recruits the TAB2/TAK1 complex for IKK activation 42.
Down-regulation of NF-κB activation by A20 depends on its dual ubiquitin-editing functions 24; 43. A20 contains a de-ubiquitinating enzyme (DUB) domain at its N-terminal portion 44 and multiple zinc fingers at its C-terminal portion that show redundancy in inhibition of NF-κB activation and other functions 8; 45. While the DUB domain of A20 de-ubiquitinates Lys63-linked polyubiquitinated TRAF6 and RIP, the A20 C-terminal zinc fingers act as ubiquitin ligase to mediate Lys48-linked polyubiquitination. In the TNF pathway, the first function down modulates Lys63-linked polyubiquitination of RIP1 required for NF-κB activation and the second function leads to RIP1 degradation. Both functions cooperate to lead to inhibition of NF-κB activation to terminate receptor signaling 17; 24; 43.
DUBs hydrolyze the isopeptide bonds at ubiquitin C-termini that have been linked to other ubiquitin molecules, specific protein substrates or small molecules. The functions of these enzymes may be quite diverse and include liberation of monomeric ubiquitin molecules from polyubiquitin and or polyubiquitinated protein substrates to recycle ubiquitin, reversal of regulatory ubiquitination, and regeneration of ubiquitin from side reactions with small molecule nucleophiles 46; 47. Most DUBs fall into three structurally distinct families of cysteine proteases, the ubiquitin C-terminal hydrolases (UCHs), the ubiquitin-processing proteases (UBPs) and the ovarian tumor (OTU) domain-containing enzymes 46; 47. The OTU superfamily comprises a group of putative cysteine proteases that are homologous to the OTU gene product of Drosophila 48. A20 has been predicted to be a member of the OTU superfamily 48. The ~100 identified OTU family members include proteins from eukaryotes, viruses and pathogenic bacteria.
In vivo, the A20 N-terminal DUB domain specifically de-ubiquitinates Lys63-linked polyubiquitinated TRAF6 and RIP 17; 24; 43. It is presumed that this is achieved via specific disassembling of Lys63-linked polyubiquitin chains. It has also been proposed that like other DUBs, A20 may have an active site similar to papain-like cysteine proteases 44; 48. To elucidate the molecular basis for the DUB activity of A20, we determined its crystal structure and performed a series of biochemical and cell biological studies. The structure reveals the potential catalytic mechanism of A20, which may be significantly different from papain-like cysteine proteases. Ubiquitin can be docked onto a conserved A20 surface without steric clash; this interaction exhibits charge complementarity. Surprisingly, A20 does not have specificity for Lys63-linked polyubiquitin chains. Instead, A20 effectively removes Lys63-linked polyubiquitin chains from TRAF6. Our studies argue that the specificity for particular substrates assures A20 the fidelity in regulating NF-κB activation in the TNF, IL-1 and Toll-like receptor pathways rather than acting as a general DUB.
The A20 DUB domain (residues 1–370) structure was determined from a four-wavelength Se-MAD data set on a twinned crystal (Table 1, Figure 1(a)). Although twinning was suspected from the beginning, the presence of non-crystallographic translational symmetry masked detection of twinning by intensity distributions 49. A very similar twinning case was previously observed in the crystal of the superoxide dismutase from the hyperthermophilic crenarchaeon Pyrobaculum aerophilum 50. Fortunately, the pseudo-crystallographic twofold operator is close to the twinning operator, resulting in an interpretable electron density map. The twinning operator was applied only in the late stage of the structural refinement. The final model contains six molecules of A20 in the crystallographic asymmetric unit and is well defined in the 2Fo-Fc map (Figure 1(b)).
The structure folds into the shape of a teakettle with approximate dimensions of 80 Å; × 40 Å; × 56 Å; (Figure 1(c), 1(d)). It is composed of ten β-strands and ten α-helices, arranged into an anterior α-helical domain (the mouth), a central β-sandwich domain (the lid) and a posterior α-helical domain (the body). The central β-sandwich lid domain has a five-stranded β-sheet (β1–β5) opposed by a three-stranded β-sheet (β6, β9, and β10). The strands β9 and β10 further extends upward to form a β-hairpin that packs against the β-hairpin formed by strands β7 and β8. The posterior body domain is formed by helices α4–α9 while the anterior mouth domain is formed by helices α1–α3 and α10. The helices are mostly from the N-terminal half whereas the β-strands are mostly from the C-terminal half (Figure 2). The overall fold of the structure is different from any known structures in the PDB database.
The active site of A20 is formed near the center of the A20 surface in the orientation shown (Figure 1(c), 1(d)), at the junction between the lid domain and the body domain. Helix α4 from the body domain harbors the catalytic Cys103 and the β4– β5 loop of the lid domain harbors the catalytic His256, both of which are important for A20 catalysis 44. A catalytic cleft runs horizontally at the domain interface. To the right lies the somewhat concave predicted ubiquitin binding surface (see below). Most of the A20 surface is fairly featureless except that the active site cleft is narrowed to the right by an elevation formed by the β1-α4 loop preceding the catalytic Cys103 and the α7–α8 loop. To the left of the active site may lay the potential binding surface for substrates such as TRAF6 and RIP1. Therefore, the surface of A20 creates an expansive platform for catalysis, ubiquitin binding and interaction with substrates (Figure 1(d)).
The mouth domain of the teakettle extends out to connect to the C-terminal domain containing multiple zinc fingers. Both the N- and C-termini of the DUB domain reside at the mouth domain, allowing it to be inserted into other sequences. In fact, in the two A20-like proteins Cezanne and TRABID, the DUB domains are located at different regions of the sequence, one in the middle of the protein close to the N-terminus, one near the C-terminus 51.
Although the overall structure of the A20 DUB domain is different from other members of the cysteine protease superfamily such as papain, the ubiquitin C-terminal hydrolases (UCH) Yuh1 and UCH-L3 52; 53 and the ubiquitin-specific processing proteases (UBP) HAUSP, USP14 and USP2 54; 55; 56, the arrangement of catalytic Cys and His residues is similar and conserved (Figure 3(a), 3(b)). The distance between the thiol of Cys103 and the Nδ1 atom of His256 is 3.2Å;, suggesting that the observed A20 active site is at least close to a catalytically productive conformation. This is unlike the situation in HAUSP, in which the active site exists in an unproductive conformation before ubiquitin binding 54.
The A20 active site is most similar to that of otubain 2 57 (Figure 3(c)). A structural homology search by DALI shows that 81 out of the 352 ordered residues of the A20 DUB domain can be superimposed with an RMSD of 1.5 Å; onto the otubain 2 structure, which contains 221 residues (Figure 2, 3(d)). The superimposed regions include three β-strands (β2, β4, β5) of the central β-sheet and four α-helices (α4, α5, α6, α8), which contain the active site. They reside at approximately residues 100–250 of A20 and are within the predicted OTU domain. OTU is a drosophila protein involved in oocyte morphogenesis 58. Sequence signatures similar to OTU have been shown to define a new class of cysteine proteases including A20 and otubain 2 24; 48. However, on a sequence level, A20 does not share significant homology with otubain 2. Only 19 residues are identical between A20 and otubain 2 (Figure 2).
Because we were unable to obtain a complex of A20 with a ubiquitin-based suicide inhibitor such as ubiquitin aldehyde (Ubal) or Ub-Br2 for structural studies (see below), we inspected the A20 surface to see if we could predict its ubiquitin binding site. Strikingly, mapping of residues conserved among different species of A20 showed a highly conserved surface patch to the right of the active site (Figure 4(a)). In contrast, the reverse side of A20 does not have conspicuous conserved patches (Figure 4(b)). When we superimposed A20 active site with the corresponding regions of Yuh1-Ubal (PDB ID: 1XD3) 52 and HAUSP-Ubal complexes (PDB ID: 1NBF) 54, the two complexes representative of the UCH and the UBP families of DUBs, we found that the bound ubiquitins were both placed to the right of the active site cleft near the conserved patch (Figure 4(c)).
One important feature for the bound ubiquitins is that the locations of the C-terminal residues are very similar, although the remaining parts deviate from each other. Because the ubiquitin locations from both Yuh1 and HAUSP have some degrees of steric clashes with the A20 surface, we adjusted the ubiquitin location without affecting much of the interactions at the C-terminus. Thus, an A20-ubiquitin complex with little steric clash between A20 and the bound ubiquitin can be modeled (Figure 4(c), 4(d)). This may also suggest that A20 likely does not undergo dramatic conformational rearrangement upon ubiquitin binding.
This modeled complex indicates that ubiquitin binding of A20 may require the surface formed by α4, α7, α8, β2, β3, β4, β5, and the loops between them (Figure 2). Like in other DUB-ubiquitin complexes, most of the specificity determinants appear to lie within the last five residues of ubiquitin, which assume an extended β-strand-like conformation and meander at the junction of the lid domain and the body domain. These residues are recognized extensively by A20. The two hydrophobic residues, Leu71 and Leu73 of ubiquitin, contact hydrophobic residues Pro226 and Leu227 at the disordered β2–β3 junction and Tyr252, Ser254 and His255 at β4 and the following loop (Figure 4(a), ,2).2). The disordered β2–β3 loop, as well as the disordered α6–α7 loop, may become ordered upon ubiquitin binding. The two positively charged residues, Arg72 and Arg74 of ubiquitin, make extensive polar and non-polar interactions including ion pairs with Glu192 and Glu193. Gly75 and Gly76 pass through the narrowest part of the active site cleft, which is buttressed by the unusual β1-α4 loop and the elevated α7–α8 loop (Figure 4(a), 1(c), ,22).
The modeled interaction between A20 and ubiquitin has both hydrophobic and hydrophilic components (Figure 4(a), 4(e), 4(f), 4(g)). In addition, a majority of the ubiquitin binding surface of A20 is negatively charged (Figure 4(e)), which complements the positively charged surface of ubiquitin (Figure 4(g)). Both Lys48 and Lys63 are exposed on the modeled bound ubiquitin, suggesting that the binding can accommodate longer ubiquitin chains (Figure 4(f)). On the opposite side of the active site, the contacting region for target proteins such as TRAF6 or the upstream ubiquitin in a polyubiquitin chain may be found. This second binding site may consist of α1, α2, α3, α4, α10 and the neighboring loops.
In analogy with papain-like cysteine proteases including DUBs with known structures, catalysis by A20 may proceed through the function of a catalytic triad. The catalytic Cys residue is thought to become deprotonated with assistance from the His residue and attack the scissile isopeptide bond at the C-terminal Gly76 of ubiquitin. A third catalytic residue, usually an Asp or Asn, stabilizes the catalytic His. While the catalytic Cys and His residues are highly conserved between A20, otubain 2 and other cysteine proteases, the third catalytic residue of A20 is unclear. In otubain 2, the third catalytic residue is Asn226 52; 54; 57. However, the corresponding residue for Asn226 of otubain 2 is Val258 of A20, making it impossible to serve this purpose. In HAUSP and Yuh1, the third catalytic residues are Asp481 and Asp181, respectively. When the active sites of A20, otubain 2, HAUSP and Yuh1 are superimposed, Asp70 of A20 approach the catalytic His256 from the similar direction as these known third catalytic residues, making it a possible candidate (Figure 3(e)). However, the distance between the side chain of Asp70 and the Nε2 atom of His256 is 4.0Å;, which is beyond hydrogen bonding distances.
The A20 structure shows that another candidate for a third catalytic residue to orient the catalytic His or stabilize the imidazolium ion, is the backbone O atom of Thr97 at a distance of 2.9Å; (Figure 3(a), 3(e)). This residue is conserved both structurally and at the level of sequence as Thr45 in otubain 2. Although it is unusual to have a main chain carbonyl oxygen as the third catalytic residue, there is no reason to exclude this possibility given the optimal distance of this hydrogen bonding interaction. In addition, the side chain of Thr97 is also fairly close to His256.
Further structural analysis of A20 revealed that the side chains of residues Asp100, Asp253 and Asn72 are also within 5–10Å; to the Nε2 atom of the catalytic His256 (Figure 3(e)). It has been proposed previously that Asp100 within the Cys sequence conservation region may be the third catalytic residue 48 (Figure 2). However the side chain of Asp100 forms hydrogen bonds with His106, which helps to stabilize the loop for oxyanion hole construction (see below) (Figure 3(e)). Asp253 is at the tight β-hairpin between β4 and β5 that harbors the catalytic His256. It is at the center of an extensive hydrogen bonding network with the main chain amides of His255 and His256 and the Oγ atom of Ser254 and is therefore important for the conformation of the active site. These residues, together with Asp70 and Thr97, may be tested by mutagenesis for their roles in catalysis (see below).
Comparison of the crystal structure of A20 with known DUB-ubiquitin complex structures suggests that main chain amide atoms of Cys103, Gly101 and Asp100 may be exclusively responsible for the oxyanion hole formation in A20 (Figure 4(h)). This is in contrast to the case in most cysteine proteases, in which a Gln or Asn several residues before the active site Cys in sequence and the backbone amide of the catalytic Cys together accommodate the negative potential formed on the carbonyl oxygen atom of the scissile bond 52; 54; 59. Although there is the Asn98 residue 5-residues preceding the catalytic Cys103, its side chain points away from the putative ubiquitin scissile bond (Figure 4(h)). It is possible that this is due, at least partly, to the unusual conformation of the β1-α4 loop that immediately precedes the active site Cys103 (Figure 3(b)). The loop is shifted towards Cys103 and His256, generating a much more spatially restricted active site cleft. It likely makes it impossible for a Gln or Asn side chain to approach the active site for oxyanion hole construction.
The first indication that A20 may be different from most DUBs came from our effort in obtaining a covalent ubiquitin adduct of A20. Most DUBs are sensitive to labeling by ubiquitin-based suicide compounds such as Ubal or Ub-Br2 60; 61; 62; 63, including DUBs that are specific for polyubiquitinated proteins 54; 55; 56. However, we were only able to obtain a minute amount of ubiquitin labeling of A20 by Ubal and a higher but still small amount of ubiquitin labeling by Ub-Br2 (Supplemental Figure (s1)). In contract, Ubal labeled the control DUB Yuh1 efficiently (data not shown). This may be due to an extremely low activity of A20 towards single ubiquitin based substrates. In a cleavage assay using the fluorogenic substrate Ub-AMC, A20 began to show measurable activity at around 10µM concentration while Yuh1 at the same concentration completely cleaved all Ub-AMC substrate within a few minutes (Figure 5(a)).
Although A20 has low activity towards single ubiquitin based substrates, it does cleave polyubiquitin chains 24; 44. In fact, given that the function of A20 is in down-regulating cytokine-induced NF-κB activation through de-ubiquitination of Lys63-linked polyubiquitinated TRAF6, RIP and other signaling component proteins, it is intriguing that A20 cleaves both Lys48- and Lys63-linked polyubiquitin chains. We confirmed this observation using our A20 DUB construct and Lys48- and Lys63-linked tetraubiquitin substrates (Figure 5(b)). While A20 cleaved both substrates, a qualitative difference was observed. A20 disassembled Lys48-linked chains to mono-ubiquitin with no accumulation of diubiquitin, suggesting that Lys48-linked diubiquitin is a good substrate for A20. In contrast, both Lys63-linked diubiquitin and triubiquitin are poor substrates of A20.
To understand how A20 cleaves its substrates, we used a physiological substrate, Lys63-linked polyubiquitinated TRAF6. TRAF6 is a RING-type E3, which in conjunction with E1 and the E2 dimeric enzyme Ubc13/Uev1A catalyzes its auto-ubiquitination 27; 29. We ubiquitinated bacterially expressed GST-TRAF6 with Lys63-linked chains using the in vitro ubiquitination reaction. We then added A20 or its catalytic mutant C103A to determine whether A20 can de-ubiquitinate TRAF6. The wildtype A20 effectively de-ubiquitinated TRAF6 in a dose-dependent manner (Figure 5(c)). This de-ubiquitination manifested itself as disappearance of polyubiquitinated TRAF6. The higher order polyubiquitinated species were more susceptible to the de-ubiquitination. However, we did not observe a concomitant accumulation of lower order polyubiquitinated TRAF6, suggesting that the de-ubiquitination reaction did not simply transform higher order to lower order polyubiquitinated TRAF6, but rather removed the entire polyubiquitin chains by cleaving at the TRAF6-ubiquitin junction. As a control, the C103A mutant of A20 failed to de-ubiquitinate TRAF6.
Because we used Lys63-linked polyubiquitinated GST-TRAF6 and performed the de-ubiquitination assay in the presence of glutathione beads, we could observe the appearance of free ubiquitin chains in the supernatant of the reaction as it progressed. Surprisingly, the pattern of the polyubiquitin ladder remained the same throughout the time course (Figure 5(d)), suggesting that there was negligible cleavage of the Lys63-linked polyubiquitin chains themselves. This suggests that instead of disassembling Lys63-linked polyubiquitin chains, in the presence of a specific substrate, A20 preferentially cleaves at the junction between the specific substrate and the polyubiquitin chain to remove its Lys63-linked polyubiquitination, leading to down-regulation of NF-κB activation. This specificity may be due to specific recognition of polyubiquitinated substrates as shown by interaction of A20 with TRAF6 in transfected cells (Figure 5(e)). In addition, A20 transfection significantly reduced TRAF6 auto-polyubiquitination (Figure 5(e)).
To determine whether Asp70, Asn72, Thr97, Asp100 or Asp253 may be the third catalytic residue, we mutated each to Ala (Figure 6(a), 6(b), 6(c)). The D70A, N72A and T97A mutations did not have drastic effects on the ability of A20 to de-ubiquitinate Lys48-linked polyubiquitin or diubiquitin. These results are most surprising for D70A, as this residue situates at a location similar to the third catalytic residues in HAUSP, Yuh1 and otubain 2 and may move sufficiently close to His256 upon substrate interaction. The D100A mutation compromised A20-mediated catalysis, but this may be due to its hydrogen bonding with His106 and therefore its role in maintaining the conformation of the catalytic loop. The D253A mutant did not produce soluble protein, indicating that the mutation may have led to compromised A20 stability. As a control, mutation of the catalytic Cys (C103A) completely abolished A20 activity. The mutagenesis data prompted us to conclude that unlike papain-like cysteine proteases and other DUBs, A20 most likely uses main chain carbonyl oxygen of Thr97 to orient the catalytic His256.
To validate our model of A20-ubiquitin interaction, we selected conserved surface residues at the putative ubiquitin-binding site. The selective residues were either mutated to Ala such L157A, Y159A, F224A and L227A or to charged residues such as S190K and E192K to maximize the disruption effect at the interface. The E193Q mutant has the only conservative substitution. We tested these mutants on their abilities to de-ubiquitinate Lys48-linked polyubiquitin chains or Lys48-linked diubiquitin (Figure 6(a), 6(b)). Both experiments showed that all mutations are defective except the conservative mutation of E193Q.
If these residues affect ubiquitin binding, they should also exert deleterious effects in de-ubiquitination of a specific substrate, the Lys63-linked polyubiquitinated TRAF6. As predicted, with the exception of E193Q, these mutants also abolished or reduced TRAF6 de-ubiquitination, in both the appearance of free polyubiquitin chains (Figure 6(c), left panel) or in the ubiquitination status of GST-TRAF6 bound to glutathione beads (Figure 6(c), right panel). These data strongly support the involvement of these residues in ubiquitin binding and validate our model of the A20-ubiquitin complex.
To explore which region of A20 may be involved in TRAF6 interaction, we identified another conserved patch on the surface of A20, which is a narrow strip to the left of and below the active site (Figure 4(a)). This conserved surface is not used in the modeled A20-ubiquitin complex. We recently showed that Lys124 of TRAF6 is a site of Lys63-linked TRAF6 polyubiquitination 29. The N-terminal region of TRAF6 contains a RING and 5 zinc-fingers and Lys124 is at the junction between the RING and the first zinc-finger. Because tandem RING and zinc-fingers tend to form elongated structures, we speculated that A20 might use a long surface strip to the left side of the active site, including the identified conserved residues, to interact with TRAF6. We selected two residues for mutation in this conserved patch, T118 and D119, and generated potentially highly disruptive mutations, T118Q and D119K (Figure 6(a), 6(b), 6(c)). Not surprisingly, neither mutation altered the ability of A20 to de-ubiquitinate Lys48-linked polyubiquitin chains (Figure 6(a), 6(b)). But surprisingly, they also did not affect the ability of A20 to de-ubiquitinate Lys63-linked polyubiquitinated TRAF6 (Figure 6(c)). Because this patch and the predicted ubiquitin-binding surface are the only conserved surfaces of A20 (Figure 4(a), 4(b)), we speculate that the non-conserved region of the A20 surface may be important for interacting with substrates such as TRAF6 and RIP.
There are several unexpected findings in this structural and functional analysis of A20. First, A20 does not have a preferential specificity for Lys63-linked chains. Instead, it recognizes the specific polyubiquitinated substrates and cleaves at the substrate-polyubiquitin junctions. This is a surprising finding because it has been assumed based on the physiological function of A20 that it is a Lys63-specific DUB 24; 43. The surface of A20 surrounding the active site is broad and open, which makes it ideal for cleaving between substrates and their linked polyubiquitin chains. This is in contrast to the presence of a crossover loop directly over the active site of Yuh1, which presumably only accommodate the pass through of small or unstructured substrates 52. Even in comparison with otubain 2 within the same OTU superfamily, the significantly different fold and surface features around the active site would suggest that they have very different repertoires of substrates in vivo.
Second, A20 appears to be a “stripped down” version of an enzyme as many main chain elements appear to mediate its catalysis, which differs from other DUBs and papain-like cysteine proteases in general. It is possible that the minimal active site may underlie the apparent low catalytic activity of A20. This and the specificity of A20 for particular substrates rather than for any polyubiquitin chains make A20 ideal for its physiological function in reversal of regulatory ubiquitination to inhibit NF-κB signaling. Transfection of A20 does not alter the overall level of ubiquitinated cellular proteins, conjugated through either Lys63 or Lys48 44, which is consistent with the notion that A20 does not have a global de-ubiquitinating activity and does not participate in ubiquitin recycling. These characteristics of A20 appear to be specific for A20 and not general for other OTU family members as in the same study the OTU family member Cezanne reduced the buildup of polyubiquitinated cellular proteins upon transfection 44.
A20 is induced by NF-κB transcription, which in turn inhibits NF-κB activation by cytokines and pathogen ligands in a feedback manner. Because Lys63-linked ubiquitination has been shown to play diverse roles such as kinase activation, control of gene transcription and DNA repair and replication 64, it makes biological sense that A20 only de-ubiquitinates the specific substrates in the pathway. Both the low catalytic activity and the specificity for particular substrates of A20 assure its fidelity in this function.
While this manuscript was under preparation, Komander et al. also reported the crystal structure of the A20 DUB domain 65. Although the structure was solved at a lower resolution of 3.2Å; with more disordered regions, the major structural features are nonetheless very similar. However, there may be one major difference in our conclusions regarding the catalytic mechanism of the A20 DUB domain. Similar to our structural analysis, Komander et al. also proposed that Asp70 might be the third catalytic residue, which they argue, is supported by the significant effect of the D70A mutation on catalysis. Although our D70A mutation also had observable effects in some of our assays (Figure 6(b), 6(c)), the magnitude of the effects, in our assessment, does not qualify the residue as the third catalytic residue. In the classical case of serine proteases, mutations of the third catalytic residues often reduce the catalytic rates by 100-fold. One compromising scenario is that both Asp70 and the main chain carbonyl oxygen of Thr97 participate in orientating His256 during catalysis, but Kamander et al. and we differ in our assessment of the relative contributions.
While the N-terminal DUB of A20 mediates de-ubiquitination of proteins such as TRAF6, RIP and possibly other signaling component proteins 17; 24; 25, the C-terminal Zn-finger domain of A20 mediates Lys48-linked polyubiquitination for proteasomal degradation 24. Interestingly, for RIP and possibly for TRAF6, A20-mediated de-ubiquitination is a pre-requisite for A20-mediated Lys48-linked polyubiquitination 24. It is possible that only polyubiquitinated substrates interact with A20 and therefore the pre-requisite is due to recruitment of A20 to the NF-κB activation complex. Once recruited, A20 can mediate de-ubiquitination and ubiquitination using the two different regions of the protein and possibly on different sites of the substrates. Although A20 has some activity in disassembling Lys48-linked polyubiquitin chains, the general low DUB activity of A20 likely yields to its Lys48-linked polyubiquitination, leading to the degradation of the substrates. Therefore, it appears that the specificity of A20 DUB for TRAF6 and RIP controls both the de-ubiquitination and Lys48-linked ubiquitination to inhibit NF-κB activation.
To elucidate the molecular basis of A20 function, we generated a series of deletion mutants of A20 containing at least its N-terminal DUB domain. These A20 deletions were cloned into the pET vectors and expressed in E. coli by overnight IPTG induction at 20°C. They were purified to homogeneity using Ni-affinity, ion exchange and gel filtration chromatography. The purified proteins were concentrated to approximately 5–10 mg/ml in 20 mM Tri, pH 7.5, 150 mM NaCl and 10 mM DTT. As evidenced by their ability to hydrolyze polyubiquitin chains, these deletion mutants were all functional as DUBs and the shortest deletion (residues 1–355) may represent the catalytic core domain of A20.
Three A20 constructs containing residues 1–430, residues 1–370 and residues 1–355, respectively, were crystallized using hanging drop vapor diffusion. The initial A20 crystals were from the construct containing residues 1–430 and diffracted poorly with strong anisotropy. Further truncations of A20 to residues 1–370 or residues 1–355 yielded crystals of a different morphology, which diffracted to 2.5Å; resolution. The crystallization condition is 1M ammonium sulfate, 300mM Na thiocyanate, 5mM MgSO4, and 25mM Na Cacodylate buffer at pH 6.5. Mutations of residues Q80-K81-K82 of A20 to triple Ser residues to reduce surface side chain entropy greatly improved the reproducibility of the crystallization. Selenomethionyl substitutions were generated on one A20 construct (residues 1–370) and crystallized similarly.
The diffraction patterns of the A20 crystals (residues 1–370) were initially indexed and scaled in space group P31/212 with cell dimensions of approximately 123Å;, 123Å; and 146Å;, which produced indistinguishable statistics with space group P31/2. The crystal has non-crystallographic translational symmetry as evidenced by strong peaks on a Native Patterson at positions (1/3, 2/3, 0) and (2/3, 1/3, 0) and contains three molecules per crystallographic asymmetric unit. By omitting the weaker reflections, the diffraction data can be transformed into a smaller unit cell in space group P31/221 with 1/3 of the original unit cell volume, cell dimensions of 71.0Å;, 71.0Å; and 145.7Å;, and one molecule per crystallographic asymmetric unit.
The structure determination of A20 was pursued by multi-wavelength anomalous diffraction (MAD) using Se as an anomalous scatterer and by multiple isomorphous replacement using Hg and other heavy atoms. Over ten Se-MAD or Se-SAD data sets and multiple data sets from heavy atom soaked crystals were collected, but only the last four-wavelength Se-MAD data set at 2.9Å; resolution collected with 15° wedges for both inverse beam and different wavelengths gave the correct Se positions, determination of the hand of the space group and a map suitable for model building. Twinning was suspected, but the diffraction data repeatedly passed twinning test 49; 66. The non-crystallographic translational symmetry may have altered the intensity statistics and masked the twinning of the crystal 66. The structure determination was successfully performed with data indexed in the small unit cell to avoid the complication from the translational symmetry.
The atomic model built from the small unit cell of the Se-Met crystal was molecular replaced into the large unit cell of the corresponding Se-Met crystal or a native data set at 2.5Å; resolution in space group P3212. Neither generated satisfactory refinement statistics and the refinement was stalled. Upon lowering the symmetry to P32, the refinement progressed but again stalled at fairly high R factors. Only by applying the correct twin law at the final refinement stage, satisfactory refinement statistics could be obtained (Table 1). The final model contains six molecules of A20 in the crystallographic asymmetric unit and residues 3–362 with two disordered loops 155–158 and 222–225.
Human embryonic kidney 293 (HEK293) cells were obtained from the American Type Culture Collection (ATCC) (Rockville, MD) and were cultured as previously described 29. Retroviral packaging cell line GP2-293 was purchased from Clontech (Palo Alto, CA). Monoclonal antibody to Ub was purchased from Santa Cruz Biotechnology (Santa Cruz, CA); anti-FLAG from Sigma (St. Louis, MO); rabbit polyclonal antibody to beta-actin from Cytoskeleton (Denver, CO); goat anti-rabbit IgG-conjugated horseradish peroxidase from BioRad Laboratories, and goat anti-mouse IgG-conjugated horseradish peroxidase from BD Biosciences Pharmingen. Monoclonal antibody to A20 was a generous gift from Dr. S. Singh (Imgenex; San Diego, CA). Monoclonal antibody to HA was a generous gift from Dr. G. B. Mills (University of Texas M. D. Anderson Cancer Center). Yeast E1, Ub, and poly-Ub chains (K63 and K48) were purchased from Boston Biochem (Cambridge, MA). His-tagged proteins Ubc13 and Uev1A were expressed from pET45 in E. coli and were purified using nickel-affinity chromatography as previously described 29.
Expression vectors for FLAG-TRAF6, GST-TRAF6, HA-Ub, His-Ubc13, and His-UeV1A were previously described 29. The retroviral expression vector pMX-IRES-GFP-puro (pMX-BGD) has been previously described 29. FLAG-tagged human A20 was a generous gift from Dr. H. B. Shu (National Jewish Center, University of Colorado Health Science Center) and subsequently cloned into pMX-BGD without a FLAG-tag. All site-directed mutagenesis was performed using the QuikChange kit (Stratagene) and verified by DNA sequencing.
A20 activity was monitored by turnover of the fluorogenic substrate Ub-7-amino-4-methylcoumarin (Ub-AMC). Ub-AMC was purchased from Boston Biochem with a stock concentration of 2.14 mg/ml (250 µM) in 100% DMSO. Fluorescence measurement upon substrate cleavage was performed on SpectraMax M2 plate reader (Molecular Devices) using 384-well plates (Corning). All measurements were performed in 20 µl reactions in 25 mM Tris pH 8.0, 100 mM NaCl and 5 mM DTT at 37°C with 10µM A20 and various concentrations of Ub-AMC. Release of the AMC moiety was monitored at an emission wavelength of 460 nm with the excitation wavelength of 380 nm. After A20 and Ub-AMC were added to the reaction buffer, plate was shaken for 5 sec before fluorescence reading. Reactions were monitored at 10-second intervals.
100 µM Diubiquitin and 5 µM A20 were mixed in a 10 µl reaction with 25 mM Tris buffer at pH 8.0, 5 mM DTT and 100 mM NaCl. The mixture was incubated at 37°C for one hour when SDS-loading buffer was added to stop the reaction. The cleavage was visualized on SDS-PAGE with Coomassie staining.
Transfection of HEK293 cells was performed essentially as described 29. Production of retroviral supernatants from GP2-293 cells transfected with pMX, pMX-FLAG-TRAF6 and pMXA20 and infection of HEK293 cells to generate stable cells lines with the retroviral supernatants has been previously described 29. Following transfection, cells were harvested and lysed in Buffer A for 30 min on ice, and centrifuged at 13,000 rpm for 15 min. Protein was measured on the clarified lysates and equal protein was then processed for immunoprecipitation using anti-FLAG antibody as previously described 29.
To generate polyubiquitinated TRAF6 for A20 DUB assays, purified GST-TRAF6 bound to glutathione agarose beads was mixed in a volume of 30 µl with the following components: 20 mM Hepes pH 7.2, 10 mM MgCl2, 25 µM MG132, 1 mM DTT, 59 µM Ub, 50 nM E1, 850 nM E2 (Ubc13/Uev1A), 1 mM ATP, 30 µM Creatine phosphate, and 1U Creatine kinase. The mixture was incubated at 37°C for 3–4 h with gentle agitation. The beads were washed several times in Buffer A (20 mM Tris, pH 7.4, 250 mM NaCl, 1 mM DTT, 1 mM sodium orthovanadate, 2 mM EDTA, 1% Triton X-100, 2 mg/ml leupeptin, and 2 mg/ml aprotinin) followed by three washes in low salt buffer (20 mM TRIS, pH 7.4, 25 mM NaCl, and 1 mM DTT). The polyubiquitinated TRAF6 bound to beads was then used for de-ubiquitination assays.
De-ubiquitination of ubiquitinated GST-TRAF6 bound to beads was performed in a final volume of 20 µl with the indicated concentration of A20 (or mutants therein) in DUB buffer (25 mM Hepes, pH7.4, 1 mM DTT, and 5 mM MgCl2). Assays were conducted at 37°C for the indicated times. After the incubation, 10 µl of cold 20 mM Hepes (pH 7.4) was added to the reaction and samples were centrifuged for 20 min at 4°C. A volume (20 µl) of the supernatant was collected, boiled in SDS-sample buffer, and subjected to SDS-PAGE and immunoblotted with a Ub-specific antibody. The remaining GST-TRAF6 bound to the beads was then washed as described above followed by addition of SDS-sample buffer, and subjected to SDS-PAGE and immunoblotted with a Ub-specific antibody. De-ubiquitination of free K48- or K63-polyubiquitin chains (500 ng) was performed essentially as described above and stopped by the addition of SDS-sample buffer, boiled and subjected to SDS-PAGE and immunoblotted with a Ub-specific antibody.
We thank Igor Kourinov at NE-CAT of APS for help with diffraction data collection and Jin Wu for maintaining X-ray and computer equipment. This work was supported by National Institute of Health (RO1 AI045937) to HW and institutional start-up funds to BGD. SCL and YCL are postdoctoral fellows of the Cancer Research Institute and ML is a postdoctoral fellow of the American Heart Association.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.