The crystal structure of the second Z-DNA binding domain of human DAI (ZBP1) in complex with Z-DNA reveals an unusual binding mode to Z-DNA

Abstract

Mammalian DAI (DNA-dependent activator of IFN-regulatory factors), an activator of the innate immune response, senses cytosolic DNA by using 2 N-terminal Z-DNA binding domains (ZBDs) and a third putative DNA binding domain located next to the second ZBD. Compared with other previously known ZBDs, the second ZBD of human DAI (hZβ_DAI) shows significant variation in the sequence of the residues that are essential for DNA binding. In this article, the crystal structure of the hZβ_DAI/Z-DNA complex reveals that hZβ_DAI has a similar fold to that of other ZBDs, but adopts an unusual binding mode for recognition of Z-DNA. A residue in the first β-strand rather than residues in the β-loop contributes to DNA binding, and part of the (α3) recognition helix adopts a 3₁₀ helix conformation. The role of each residue that makes contact with DNA was confirmed by mutational analysis. The 2 ZBDs of DAI can together bind to DNA and both are necessary for full B-to-Z conversion. It is possible that binding 2 DAIs to 1 dsDNA brings about dimerization of DAI that might facilitate DNA-mediated innate immune activation.

The innate immune response is essential for protection from foreign invasion, acting as an immediate cellular defense mechanism. Nucleic acids are known as one of the triggers for activation of the innate immune response (1–3). Recent reports have indicated the presence of a new cytosolic DNA sensor that can initiate an innate immune response independent of the endosomic Toll-like receptor 9 (4, 5). Z-DNA binding protein 1 (ZBP1), also known as DLM-1, was identified as the first innate immune activator that senses cytosolic DNAs (4). In response to foreign DNA, this protein activates type I IFN and other immune responses and was therefore named DAI (DNA-dependent activator of IFN-regulatory factors; ref 4). DAI contains 2 tandem Z-DNA binding domains (ZBDs or Zα and Zβ) at its N terminus and a third DNA binding region (D3) located next to the second ZBD. D3 is a novel domain and is reported to bind right-handed B-DNA (4). Upon activation, the C terminus of DAI binds to Tank binding kinase 1 (TBK1), a serine/threonine kinase, and to IFN regulatory factor 3 (IRF3), a transcription factor (4). The N-terminal region, including D3, is thought to be essential for sensing DNA, as shown by its ability to bind to Z-DNA and synthetic B-DNA (4). For the full activation of an in vivo DNA-dependent immune response, all 3 DNA-binding regions are required (5). At the molecular level, it has been suggested that dimerization of DAI results in activation of the innate immune response (5). At the cellular level, it is known that the localization of DAI and its association with stress granules is regulated by ZBDs (6, 7).

ZBDs that are found in DAI are also found in the editing enzyme dsRNA adenosine deaminase (ADAR1) in vertebrates and in fish PKZ protein kinase containing Z-DNA-binding domains and in the E3L of pox viruses (Fig. 1). They all bind tightly and specifically to Z-DNA and usually are able to convert B-DNA to Z-DNA (8–14). Structural studies of ZBDs have shown that they all share an α/β architecture consisting of 3 β-strands and 3 α-helices (9, 15–17). Although they have limited similarity at the amino acid sequence level, they share a common binding mode to the Z conformation of DNA with many conserved residues involved in DNA binding. ZBDs recognize Z-DNA with conformational specificity by using the recognition helix α3 and the wing, which is made up mostly of an antiparallel β-sheet (β2 and β3) and a β-loop (9, 15–17). In their interaction with Z-DNA, positively charged or polar residues play major roles in recognizing the zig-zag conformation of the Z-DNA backbone. There are many direct ionic interactions or water-mediated hydrogen-bonding interactions with the DNA backbone phosphate groups (15). In addition, the CH–π interaction between Tyr and the C8 of guanine in the syn conformation is known to be a key interaction for Z-DNA specific binding (15). The β-sheet wing also contributes to the interaction through water-mediated hydrogen bonds and hydrophobic interaction involving Trp, Thr, and Pro residues (15).

Fig. 1.

Multiple sequence alignments of the first ZBDs (Zα) and the second ZBDs (Zβ). Zαs include human ADAR1 (hZα_ADAR1), mouse ADAR1 (mZα_ADAR1), human DAI (hZα_DAI), mouse DAI (mZα_DAI), goldfish PKZ (caZα_PKZ), zebrafish PKZ (drZα_PKZ), and Yaba-like disease poxvirus E3L (yabZα_E3L). Zβs include human DAI (hZβ_DAI), mouse DAI (mZβ_DAI), goldfish PKZ (caZβ_PKZ), and zebrafish PKZ (drZβ_PKZ). The secondary structures of hZβ_DAI and hZα_ADAR1 are shown by a rectangle indicating a helix or an arrow for a β-strand. The 3₁₀ helix is shown in green. Residues of hZβ_DAI and hZα_ADAR1 involved in DNA interactions are marked by blue and red circles, respectively. Highly conserved residues are highlighted in yellow. The residue numbers of both N and C termini are shown.

In DNA-mediated activation of the innate immune response, the N-terminal ZBDs are believed to be essential for the full activity of DAI (4, 5). The first DNA binding domain of DAI (Zα_DAI, also known as Zα_DLM-1 or Zα_ZBP1) was well characterized as a canonical ZBD in structural and biochemical studies (9). The second ZBD (Zβ_DAI) was also shown to bind Z-DNA based on its binding specificity for Z-DNA and its ability to convert B-DNA to Z-DNA, as measured in the CD spectrum (6, 14). However, some residues that are known to mediate ZBD/DNA interactions in other systems are not present in hZβ_DAI, suggesting that the binding mode of Zβ_DAI with DNA might be different from that of other canonical ZBDs (Fig. 1). In particular, the conserved residues in the β-loop of the wing region are missing in hZβ_DAI (Fig. 1). Therefore, an intriguing question emerges as to how this second ZBD recognizes Z-DNA despite the limited sequence similarity to canonical ZBDs. In addition, it would also be interesting to explore how hZαβ_DAI, which consists of hZα_DAI, hZβ_DAI, and a linker, recognizes DNA and to determine the role of DNA binding by hZαβ_DAI in the innate immune response.

To address these questions about the binding mode of hZβ_DAI to Z-DNA, we have solved the crystal structure of hZβ_DAI complexed with dsDNA. In addition, we have carried out Z-DNA converting assays of hZβ_DAI and mutants that lack some of the residues involved in Z-DNA binding found in other structures. Although hZβ_DAI has the common overall fold of ZBDs, its binding mode to DNA is clearly distinguishable from that of other ZBDs. In particular, residues in the β-loop of the wing do not engage Z-DNA at all and are not directly involved in its recognition. Furthermore, biochemical studies of hZαβ_DAI reveal that both ZBDs of DAI are involved in DNA binding, and such binding could play a role in the proposed oligomerization and activation of DAI (5).

Results

Overall Structure of the hZβ_DAI/Z-DNA Complex.

The structure of hZβ_DAI (residues 103–166) bound to dsDNA in the Z conformation, d(TCGCGCG)₂, was solved at 1.45-Å resolution by using multiwavelength anomalous dispersion (MAD) data with an R factor of 15.3% and an R_free of 19.3% (Table S1). The asymmetric unit contains 2 hZβ_DAI molecules called chain A and chain B, and 1 duplex DNA in the Z conformation (Fig. 2A). The DNA strands facing protein chains A and B are designated as C and D, respectively (Fig. 2A). The 2 hZβ_DAI/ssDNA complexes are related by pseudo noncrystallographic 2-fold symmetry in the asymmetric unit and are nearly identical with 0.68-Å rmsd between 63 Cα atoms of the protein and all atoms of the 6 nucleotides (pCpGpCpGpCpG). The 5′-deoxythymine of chain C and R166 residues in chains A and B were not modeled because of structural flexibility.

Fig. 2.

Structural comparison of hZβ_DAI and hZα_ADAR1 and their interactions with Z-DNA. (A) Overall structure of the hZβ_DAI/Z-DNA complex. The protein and DNA are drawn as a ribbon diagram and stick model, respectively. The N and C termini, the secondary structure elements of hZβ_DAI, and 5′ and 3′ of DNA are labeled. (B) The structural overlap of hZβ_DAI and hZα_ADAR1 near the Z-DNA binding site. The ribbon diagrams of hZβ_DAI (blue) and hZα_ADAR1 (green) are overlapped, and the stick model of a double-strand Z-DNA is drawn in brown and gray. Structural deviation near α3 and the β-wing are labeled and marked by red circles. (C) Stereoviews of the protein–DNA interfaces of hZβ_DAI (chain A)/Z-DNA (Left) and hZβ_DAI(Chain B)/Z-DNA (Right). The Cα chains of the ZBDs are colored sky blue, and the residues involved in DNA contact are depicted in sky blue stick models. The Z-DNA backbone is drawn as a red stick model, and bases are drawn as gray stick models. Water molecules are shown in green. Hydrogen bonds are drawn as dashed lines. (D) Schematic diagrams of the protein–Z-DNA interactions in hZβ_DAI (Left) and hZα_ADAR1 (Right). Hydrogen bonds and van der Waals contacts are represented by dashed black lines and pink lines, respectively. The CH–π interaction between the conserved Tyr and the C8 of a syn-guanine is indicated by black circles, and waters are shown by green circles. The protein–DNA interactions in hZα_ADAR1 are identical on both sides of the Z-DNA, thus only 1 side is shown.

Similar to other structures of ZBDs bound to Z-DNA (9, 15, 16), hZβ_DAI also has α/β topology with 3 helices (α1, α2, and α3) packed against 3 β-stands (β1, β2, and β3) (Fig. 2A and Fig. S1). However, hZβ_DAI differs from other ZBDs by the presence of a 3₁₀ helix at the N terminus of α3 (Fig. 2 and Fig. S1). Instead of the long continuous α-helix (α3) present in other Zαs, hZβ_DAI has a kinked helix with a mixed conformation of 3₁₀- and α-helices. Furthermore, the most unique conformational difference in hZβ_DAI is found in the β-sheet wing, where hZβ_DAI has disengaged the β-loop of the wing from Z-DNA so it no longer makes contact (Fig. 2B and Fig. S1). Thus, the β-loop of hZβ_DAI has limited interaction with Z-DNA. In the complex, dsDNA adopts a typical Z-DNA conformation, with guanine nucleotides in the syn conformation and C3′-endo sugar puckering, except for G6, which is C2′ -endo. Whereas previous Z-DNA structures have the Z_I conformation in their phosphodiester backbones (9, 15, 16), both strands of Z-DNA bound to hZβ_DAI have a Z_II conformation for G4pC5, which further illustrates the different binding mode of hZβ_DAI compared with other ZBDs (Fig. S2).

Protein–DNA Interactions in the Recognition Helix.

Interactions mediated by the most conserved core residues of ZBDs (Y145, N141, and W162) are maintained in the hZβ_DAI/Z-DNA complex, although the position of the water molecules involved in hydrogen bonding varies somewhat. In hZβ_DAI, N141 forms 1 direct and 1 water-mediated interaction with the phosphate backbone of Z-DNA, whereas N173 of hZα_ADAR1 contacts DNA through 2 water-mediated hydrogen bonds (Fig. 2C and Fig. S3) (15). However, except for those core residues, the other interactions with Z-DNA seem to be different for hZβ_DAI. It is likely that R142 of hZβ_DAI plays a role in Z-DNA recognition similar to R174 of hZα_ADAR1, but hZβ_DAI does not form a bond with the ribose ring of G6, as found in hZα_ADAR1 (Fig. 2 C and D and Fig. S3). Also, K138 of hZβ_DAI is located in a region that is between K169 and K170 of hZα_ADAR1. However, K138 of hZβ_DAI plays a significantly different role than the 2 positively charged residues in other ZBDs (Fig. 2 C and D and Fig. S3). K138 forms a bond to the phosphate atom of the DNA strand and has only limited additional interactions, whereas the 2 positively charged residues of hZα_ADAR1 contact the 4 phosphate groups of 1 DNA strand by using hydrophobic and coulombic interactions. It is interesting that K138 spans across the Z-DNA molecule and interacts with the C5 phosphate group on the opposite DNA strand. Cross-interactions of this type are not found in Zα_ADAR1 (Fig. 2D). For K138 to point to and contact DNA in that way, a conformational change would be necessary in the recognition helix (α3) and the following loop. Indeed, the N-terminal end of the recognition helix is changed to a 3₁₀ conformation and the following β-loop is moved away from the DNA and toward α2 (Fig. 2C and Fig. S3).

Protein–DNA Interactions in the Wing.

Most ZBDs have 1 or 2 Pro residues that contribute to DNA binding through hydrophobic interactions plus polar residues, such as Thr or Asn, that bind the DNA through water-mediated hydrogen bonds (Fig. 2 and Fig. S3) (9, 15, 16). Moreover, Pro residues in this position are known to be essential for the pathogenicity of the vaccinia E3L protein, which has a ZBD in its N-terminal region (18). hZβ_DAI does not have residues that play a role equivalent to Pro or Thr in the wing region (Fig. 1), but it still has nearly the same Z-DNA-converting activity as hZα_ADAR1 (6, 14). Therefore, it was anticipated that there would be another residue in the wing region that must bind to Z-DNA. Recent mutagenesis data suggest that K160 might play this role, because mutation of K160 into Glu reduced Z-DNA binding activity (19). Although the β-loop in the wing of hZβ_DAI is disengaged from the DNA compared with other ZBDs (Fig. 2), K160, located in the tip of the loop, might reach the backbone phosphate group of Z-DNA. But, in the current crystal structure, K160 is unlikely to be involved in DNA binding because the side chain of K160 in each subunit is disordered (Fig. 3). Thus, neither K160 nor any other residue in the β-loop of hZβ_DAI, systematically participates in DNA recognition.

Fig. 3.

The Z-DNA binding mode of hZβ_DAI. (A) Z-DNA recognition by R124 of hZβ_DAI. A magnified view near the wing region is drawn for the interfaces between hZβ_DAI and Z-DNA [chains A and C (Left) and chains B and D (Right)]. A σ-weighted 2 F_o − F_c omit map contoured at 1.2 σ was generated by omitting R124, R160, and the waters involved in the R124-mediated interaction. Possible hydrogen bonds among R124, water and the phosphate group of G2 are shown by dashed lines, and their distances are indicated in Å. (B) Surface charge distributions of the chain A (Left) and chain B (Right) viewed along the DNA binding cleft. The red and blue areas represent negatively and positively charged surfaces, respectively. Z-DNA chains are shown as stick models, and R124, K160 and each nucleotide are labeled.

However, an unusual Z-DNA binding mode is found in the hZβ_DAI wing. ZBDs usually recognize the zig-zag backbone of Z-DNA, but hZβ_DAI, lacking binding interactions through the β-loop of the wing, uses a different method. R124, which is located in the first β-stand of hZβ_DAI, plays that role (Figs. 2 and 3). The Nε of R124 from chain A forms a water-mediated hydrogen bond with O6 of G2; NH1 of R124 forms direct hydrogen bonds with O1P and O2P of the G2 phosphate group. In chain B, NH1 of R124 is connected to O1P of the G2 phosphate group through water (Fig. 3).

It is interesting that all ZBDs lacking Pro residues in the wing region have a conserved positively-charged residue corresponding to R124 of hZβ_DAI (Fig. 1). For example, the second ZBDs of zebrafish PKZ (drZβ_PKZ) and goldfish PKZ (caZβ_PKZ) lack the Pro residues in the wing, but have a Lys residue at the same site as R124 of hZβ_DYI (Fig. 1). In this regard, ZBDs can be divided into 2 groups on the basis of the presence of the Pro residues in the wing region, and accordingly, the binding mode to Z-DNA. Group I ZBDs, which have the Pro residue in the wing, can recognize Z-DNA by using a recognition helix and the β-loop of the wing, whereas group II ZBDs have a positively charged residue in the β1 strand, instead of the Pro residue in the wing. In Z-DNA binding, it is expected that the positively charged residue in the first β-stand (β1) of other group II ZBDs will play the same role in Z-DNA binding as the residues in the β-loop of the wing of group I ZBDs.

The Alternative Binding Mode Is Supported by Z-DNA Converting Activity.

The role of each residue in Z-DNA recognition was investigated by a B-DNA-to-Z-DNA CD converting activity assay of various mutants. Residues directly involved in Z-DNA binding are known to stabilize the Z-DNA conformation and enhance Z-DNA conversion in the CD (18, 20, 21). The above results suggest that the kinked recognition helix and R124 of β1 rather than the β-loop of the wing are important in Z-DNA binding of hZβ_DAI. It has been demonstrated that K160 does not have a critical role in Z-DNA recognition because mutation of K160 to Ala (19) has only a slight effect on Z-DNA converting activity (Fig. 4A). The contribution of R124 was confirmed by this method because Z-DNA converting activity was diminished to 70% when it was replaced with Ala (Fig. 4A). It is clear that K160 is not systematically involved in Z-DNA binding in the current complex structure, but from its proximity to Z-DNA, it can be assumed that K160 may participate in this process in some way (Fig. 3). The reduced Z-DNA converting activity of the K160E (19) mutant suggests that K160 is in close proximity to DNA, at least in terms of charge (Fig. 3B), although a direct interaction was not observed. From these results it can be postulated that R124 has a dominant role and K160 is supplementary in Z-DNA binding. It is tempting to hypothesize that K160 performs a role in Z-DNA binding in the absence of R124, and the 70% remaining activity of the R124A mutant comes from the contribution of K160. To verify this hypothesis, both were mutated to Ala, and the activity of this double mutant was tested. As shown in Fig. 4A, Z-DNA converting activity was dramatically reduced, lower than that of either R124A or K160A mutants. CD spectroscopy (Fig. S4) and dynamic light scattering experiments (data not shown) revealed that WT and mutant hZβ_DAI R124AK160A have similar secondary structure contents and homogeneous solution conformations, suggesting that the reduced activity is not caused by the structural instability or denaturation of the mutant. These results suggest that the presence of a positively charged residue near G2 of DNA is critical for Z-DNA binding. The fact that a K138A mutant retained ≈80% activity suggests that K138 may significantly participate in this process. The K138A counterparts in hZα_ADAR1, K169, and K170, have also proven to be important for Z-DNA binding affinity by mutagenesis experiments (20). However, K138 does not seem to be as critical or indispensable when compared with other core residues in the recognition helix, such as Y145, for which an Ala mutant showed no Z-DNA converting activity (Fig. 4A).

Fig. 4.

B-to-Z-DNA conversion activity of hZβ_DAI and hZαβ_DAI. (A) The Z-converting activities were estimated by monitoring the CD signal at 255 nm for 40 min by using 60 μg/ml of double-stranded (dCdG)₆. The molar ratio of hZβ_DAI or its alanine mutants to ds(dCdG)₆ was 4:1. D139A was randomly chosen as a negative control. (B) WT hZαβ_DAI and 2 tyrosine mutants, hZαβ_DAI Y50A and hZαβ_DAI Y145A, were mixed with 15 μM double-stranded (dCdG)_6, at 2:1 and 4:1 molar ratios, and their CD signals were monitored at 255 nm for 60 min. As a control, the Z-DNA converting activity of a single ZBD (hZα_ADAR1) at a 4:1 molar ratio to DNA was also measured under the same condition.

Z-DNA Binding and Oligomerization of hZαβ_DAI.

From structural and biochemical studies on the first ZBD of DAI (6, 9, 14, 19), it was expected that hZα_DAI shared a canonical ZBD fold and a similar binding mode with other members of what we call group I ZBDs. This study reveals that hZβ_DAI belongs to what we call group II ZBDs and has a distinct Z-DNA binding mode. The next questions are how the 2 tandem ZBDs contribute to Z-DNA binding and how Z-DNA binds to hZαβ_DAI. Structural studies will be required to get a firm answer. We would like to know how it may contribute to a possible oligomerization of DAI and ultimately, how this binding contributes to the immune response. We may, however, get clues from solution studies.

It has already been established that each of the 2 ZBDs of DAI individually possesses Z-DNA converting activity and can bind to Z-DNA independently (6, 14, 19). To investigate their joint binding modes, Z-DNA converting activities were measured for hZαβ_DAI, hZαβ_DAI Y50A, and Zαβ_DAI Y145A (Fig. 4B). It is known that the Tyr residue located in the α3 recognition helix of these domains plays a pivotal role in the conversion to Z-DNA, and substitution with Ala sharply lowers the activity of hZα_ADAR1 (18) and abolishes the activity of hZβ_DAI (Fig. 4B; ref. 19).

In Fig. 4B, 100% of the converting activity (B-DNA to Z-DNA) of 15 μM ds(dCdG)₆ is found with 30 μM hZαβ_DAI, which is equivalent in activity to 60 μM hZα_ADAR1. This indicates that both Zα and Zβ domains of DAI are active in binding Z-DNA. Mutating Y145 of 30 μM hZαβ_DAI to alanine lowers the converting activity to 35%. The analogous mutant in the Zα domain (Y50A) lowers the converting activity to nearly 23%. These results suggest that the binding of Zα_DAI is equal to or slightly more important for converting activity than Zβ_DAI. When the concentration of the mutant (Zαβ_DAI Y150A or Zαβ_DAI Y145A) is doubled to 60 μM to compensate for the loss of Z binding by the mutations, the converting activity approximately doubles but is still below the activity of unmutated 30 μM Zαβ_DAI. There may be many possible explanations for this phenomenon, including various types of steric hindrance. However, it suggests the 2 ZBDs act largely independently of each other. It is known that Zα domains bind to 6 bp of Z-DNA, with a domain largely bound to 1 DNA strand. To get some insight into how Zαβ_DAI domains bind, sizing gel experiments were carried out by using 12-bp fragments that can form Z-DNA (Fig. S5). These support the idea that the dsDNA can recruit 2 Zαβ_DAI domains.

Discussion

Since the first identification of a ZBD (8), there has been great interest in the role and the specificity at their N terminus for the Z conformation of DNA or RNA (22). Recent studies of poxvirus E3L demonstrated that deletion of the ZBD or mutations that eliminate Z-DNA binding resulted in loss of viral pathogenicity. The ZBD is thus essential for virulence (18). It has been proposed that E3L may compete with cellular Z-DNA binding proteins (23). There is also evidence that E3L can act through gene transactivation and conversion of the host cell to an antiapoptotic state (24). Recently, ZBP1 (DLM-1) was reported to be a cytosolic B-DNA binding sensor that regulates the DNA-mediated innate immune response. It was thus renamed DAI (4). It has also been suggested that a DNA-mediated multimerization of DAI may be involved in activation of the innate immune response (5). It is possible that poxvirus E3L competes with DAI for the binding to Z-DNA, thereby preventing activation of the innate immune response. An understanding of the binding mode and spatial arrangements of the ZBDs of DAI is essential for understanding the mechanism of DAI as an IFN activator in the innate immune response.

The second ZBD of DAI shows distinctive features in sequence and structure compared with other ZBDs of defined structure (Fig. 1). The crystal structure analysis supported by mutagenesis and a B- to Z-DNA converting assay showed that hZβ_DAI recognizes Z-DNA by an unusual binding mode. It has a kinked recognition helix containing a 3₁₀-helix segment and a positively charged residue in the β1 strand that play key roles (Figs. 2 and 3). In contrast with other ZBDs, the β-loop in the wing of hZβ_DAI is short and positioned away from the Z-DNA. We conclude that ZBDs can be categorized into 2 groups based on the residues involved in DNA binding. A study of splicing patterns of DAI (ZBP1) reveals considerable complexity (23). The first ZBD (Zα) is spliced out in ≈50% of DAI mRNA transcripts, whereas the second ZBD (Zβ) is present in nearly all splicing variants. Recent work (25) reports that human DAI activity can also be found when Zα is spliced out, which suggests a major role of Zβ_DAI in Z-DNA recognition and immune activation. Substantial sequence variation in the second ZBDs (Zβ) of most Z-DNA binding proteins could also result in the loss of Z-DNA binding activity. For instance, the Z-DNA binding protein ADAR1, which has RNA editing activity, also has 2 tandem ZBDs at its N terminus, but the Zβ domain does not bind to Z-DNA as it is missing the crucial Tyr residue (11, 18). However, hZβ_DAI retains Z-DNA binding activity, despite the considerable sequence variation that leads to alterations in the DNA binding mode.

The results of this study, combined with previous biochemical analyses, indicate that each ZBD of DAI binds to DNA in a distinctive mode, but both act together in inducing the conversion to Z-DNA. DAI also has a DNA binding region (D3) located at the C terminal to hZβ_DAI. It has been shown that all 3 DNA-binding domains of DAI are required for full activation of the mouse DNA-dependent immune response (5). The linker between hZα_DAI and hZβ_DAI may help position the 2 ZBDs in close enough proximity to allow for their full accessibility to Z-DNA. It has been suggested that the linker may be involved as it enhances the DNA binding activity of ZBDs (6). Interestingly, artificial dimerization of DAI was shown to activate the innate immune response, even in the absence of DNA (5), which suggests the possibility that DNA-mediated multimerization of DAI might also act to evoke the innate immune response. Some data presented here may provide clues for the molecular mechanism underlying such multimerization. Two ZBDs (Zαβ_DAI) are able to bind dsDNA in tandem or by encircling the DNA. If the Z-forming region is ≥12 bp, then 2 DAIs can be attached in close proximity, and such dimerization could result in activation. An analogous example is the dsRNA-induced dimerization of PKR by binding 2 tandemly arranged RNA binding domains, which results in phosphorylation of the C-terminal domain (26). It has also been proposed that PKZ is activated in a DNA-dependent manner by using its 2 tandemly arranged ZBDs (27).

B-DNA activates DAI for IFN production. It is possible that Zαβ_DAI recognizes and binds to a transient Z-conformation in a long B-DNA. It is also possible but less likely that Zαβ_DAI, in concert with the third DNA binding domain, may bind to B-DNA. We have observed that hZα_ADAR1 has a binding affinity for B-DNA with a dissociation constant (K_d) of the order of 10⁻⁵ M (data not shown), and this binding is likely to be nonspecific. In contrast, hZα_ADAR1 has a K_d near 10⁻⁸ M to Z-DNA (8). It will be of interest to make similar measurements for hZαDAI and its subunits. It is likely that possession of 2 ZBD domains and a linker maximizes DNA binding activity and possibly facilitates dimerization of DAI, leading to IFN induction.

This study contributes to the understanding of the structure and function of DAI and provides a foundation for elucidating DAI function in the innate immune response. However, it is clear that more work will be required to uncover the detailed mode of dsDNA binding to DAI that leads to IFN production.

Materials and Methods

Protein Preparation.

hZβ_DAI (residues 103–166) and hZαβ_DAI (residues 8–166) were prepared as described (14, 19). Point mutations were introduced by using QuikChange according to the manufacturer's instructions (Stratagene). All mutations were confirmed by DNA sequencing, and mutant proteins were purified in the same way as WT hZβ_DAI.

CD.

The conversion of double-stranded (dCdG)₆ from B-DNA to Z-DNA was monitored by its CD spectrum, which was measured with a Jasco J-810 CD spectrometer at 25 °C in a 1-mm quartz cell. For kinetic comparisons of hZβ_DAI and its mutants, changes in the CD signal at 255 nm were measured for 40 min by using 60 μg/ml of ds(dCdG)₆ in the CD buffer [10 mM Hepes-NaOH (pH 7.5), 10 mM NaCl, and 0.1 mM EDTA]. Similarly, the CD signal of hZαβ_DAI and its mutants were monitored for 60 min. The volume of the added protein did not exceed 5% of the total reaction volume.

Structure Determination.

The crystallization and data collection of wild and selenomethionine-substituted hZβ_DAI were carried out as described (14). The structure was solved via Se MAD phasing and model building and refinement (SI Text). The final model refined at 1.45 Å with R_work and R_free of 15.5% and 19.3%, respectively, contains 2 domains of hZβ_DAI and a DNA duplex with the Z conformation in the asymmetric unit. The refinement statistics are summarized in Table S1.