Restriction endonucleases that cleave RNA/DNA heteroduplexes bind dsDNA in A-like conformation

Abstract

Restriction endonucleases naturally target DNA duplexes. Systematic screening has identified a small minority of these enzymes that can also cleave RNA/DNA heteroduplexes and that may therefore be useful as tools for RNA biochemistry. We have chosen AvaII (G↓GWCC, where W stands for A or T) as a representative of this group of restriction endonucleases for detailed characterization. Here, we report crystal structures of AvaII alone, in specific complex with partially cleaved dsDNA, and in scanning complex with an RNA/DNA hybrid. The specific complex reveals a novel form of semi-specific dsDNA readout by a hexa-coordinated metal cation, most likely Ca2+ or Mg2+. Substitutions of residues anchoring this non-catalytic metal ion severely impair DNA binding and cleavage. The dsDNA in the AvaII complex is in the A-like form. This creates space for 2'-OH groups to be accommodated without intra-nucleic acid steric conflicts. PD-(D/E)XK restriction endonucleases of known structure that bind their dsDNA targets in the A-like form cluster into structurally similar groups. Most such enzymes, including some not previously studied in this respect, cleave RNA/DNA heteroduplexes. We conclude that A-form dsDNA binding is a good predictor for RNA/DNA cleavage activity.

Figure 1.

AvaII activity on dsDNA and RNA/DNA oligonucleotide heteroduplexes. A total of 10 pmoles of dsDNA or RNA/DNA were incubated for 30 min at 37°C with amounts ranging from 10⁻⁴ to 100 pmoles of the AvaII dimer. S: substrate; P: product.

Figure 2.

AvaII in open and closed conformations. Open form is observed for the enzyme in nucleic acid free state and in the scanning complex with RNA–DNA heteroduplex. Closed form is captured in the complex with partially cleaved dsDNA. The protein homodimer is colored yellow/green and shown in ribbon representation and the nucleic acids are depicted as white sticks.

Figure 3.

AvaII overall structure. AvaII dimer with specifically bound dsDNA seen looking onto the vertically oriented dimer axis (left), and looking down the dimer axis (right). One of the AvaII protomers is colored in rainbow representation (blue to red from N- to C-terminus), the other one is shown in gray. DNA is shown in black.

Figure 4.

AvaII active site. Selected active site residues around the two catalytic metal ions are shown in all atom representation. DNA is shown in the region around the scissile phosphodiester bond. The figure is based on the coordinates for the complex with partially cleaved DNA, but for clarity, only the cleaved DNA conformation is shown. The color coding is as in Figure 3. Metal ions are shown in green.

Figure 5.

AvaII target sequence recognition. Due to the pseudo-palindromic symmetry, only a half-site of the recognition sequence is independently recognized, interactions with the other half-site are symmetric. The figure shows in the interactions of (A) the outer G:C pair, (B) the inner G:C pair and (C) the central A:T pair (only one of the possible A:T pair binding modes is shown for clarity). The color coding is as in Figure 4. The composite omit map was contoured at 1 rmsd. (D) EMSA assays for protein DNA binding with 1 μM dsDNA (29 bp), containing either a single miscognate (GGSCC) or cognate (GGWCC) site. Protein dimer amount in each series is 0, 10, 20, 50 and 100 pmol.

Figure 6.

RNA/DNA binding propensities deduced from available restriction endonuclease dsDNA complex structures. (A) Maximal clash score values for structurally characterized Type II restriction endonucleases. For each grafted 2′-OH group, only the maximum clash was taken into account. (B) Cumulative distribution of maximal clashes of 2′-OH groups in Type II restriction endonucleases. The Z-scores were calculated by subtracting the average and dividing by the standard deviation. The Z-scores for EcoO109I were calculated taking into account (brown) or ignoring (orange) the outermost nucleotide pairs of the recognition sequence. The probability for a score to be this low (one sided test) or as far from the average (two sided test) is below 5% in both cases.

Figure 7.

Structure based search for endonucleases that bind dsDNA in the A-like form and verification of their RNA/DNA cleaving activity. (A) DALI server (29) clustering of PD-(D/E)XK restriction endonucleases based on their structure. Recognition sequences are indicated, with bases bound in A-form according to the 3DNA program (31) in red. The percent cleavage of RNA and DNA strands is literature based (7). ^#For EcoRV and HinP1I the assignment was inconsistent between the structures. (B) Cleavage of dsDNA with a single target site for the indicated restriction endonucleases. The top DNA strand was labeled. The digestion pattern is as expected based on the location of cleavage sites. (C) Cleavage of RNA/DNA, the labeled top strand is DNA. (D) Cleavage of RNA/DNA, the labeled top strand is RNA. The sequence of substrates was in all cases the same except for T/U differences (Supplementary Figure S15).

Figure 8.

A-DNA form of AvaII bound dsDNA. Comparison of the double stranded helix conformation in (A) AvaII complex, (B) EcoO109I complex (12), (C) ideal A-DNA and (D) ideal B-DNA. Only five central base pairs of the duplexes were used for superposition. The ideal oligonucleotides were generated with 3DNA (31). The magenta line indicates the positions and distance between the phosphates of the scissile bonds in the complexes and the corresponding atoms in the duplexes.

Figure 9.

Steric conflicts of the 2′-OH groups in silico introduced into the AvaII and EcoO109I (13) dsDNA complexes. 2′-OH groups were grafted onto the 2′-deoxyriboses, without alteration of their puckers, with the help of the CNS program (30). Maximal and cumulative clashes were then scored as in a recent analysis of methyl group steric conflicts (32). The 1.1 Å threshold for functionally relevant steric clashes refers to methyl groups on DNA bases, but we suspect that a similar threshold applies to the clashes of 2′-OH groups (red horizontal line). For each nucleotide maximal and cumulative clash is indicated. Additionally the clashes with nucleic acid and protein were independently summed. Values were averaged for both DNA strands. The AvaII and EcoO109I target sequences are shaded in sepia.

Figure 10.

Similarity of PD-(D/E)XK restriction endonucleases that cleave RNA/DNA heteroduplexes (and of EcoO109I). (A) Cumulative plot of pairwise distances between the core regions of PD-(D/E)XK restriction endonucleases. Core regions were taken from a published alignment (10) supplemented with newly available data for selected RNA/DNA cleaving enzymes. (B) Core region alignment of two groups of heteroduplex cleaving enzymes. EcoO109I does not cleave RNA/DNA heteroduplexes, but is included in both panels for comparison.