Crystal Structure of the Homing Endonuclease I-CvuI Provides a New Template for Genome Modification*

Background: Homing endonucleases are one of the templates used in genome engineering. Results: Structural and biochemical analysis shows that I-CvuI can be a new scaffold to engineer novel DNA specificities. Conclusion: The increase in catalytic ion concentration may strengthen I-CvuI selectivity. Significance: This finding expands the homing endonuclease repertoire for redesigning new protein-DNA interactions. Homing endonucleases recognize and generate a DNA double-strand break, which has been used to promote gene targeting. These enzymes recognize long DNA stretches; they are highly sequence-specific enzymes and display a very low frequency of cleavage even in complete genomes. Although a large number of homing endonucleases have been identified, the landscape of possible target sequences is still very limited to cover the complexity of the whole eukaryotic genome. Therefore, the finding and molecular analysis of homing endonucleases identified but not yet characterized may widen the landscape of possible target sequences. The previous characterization of protein-DNA interaction before the engineering of new homing endonucleases is essential for further enzyme modification. Here we report the crystal structure of I-CvuI in complex with its target DNA and with the target DNA of I-CreI, a homologue enzyme widely used in genome engineering. To characterize the enzyme cleavage mechanism, we have solved the I-CvuI DNA structures in the presence of non-catalytic (Ca2+) and catalytic ions (Mg2+). We have also analyzed the metal dependence of DNA cleavage using Mg2+ ions at different concentrations ranging from non-cleavable to cleavable concentrations obtained from in vitro cleavage experiments. The structure of I-CvuI homing endonuclease expands the current repertoire for engineering custom specificities, both by itself as a new scaffold alone and in hybrid constructs with other related homing endonucleases or other DNA-binding protein templates.

Homing endonucleases (HEs), 3 also known as meganucleases, generate an accurate double-strand break that can be used to promote gene targeting through homologous recombination. These enzymes recognize long DNA sequences; their recognition sites vary between 12-45 bp in length. HEs are highly sequence-specific enzymes and display a very low frequency of off-target cleavage, even in whole genomes. Due to this rare-cutting property, they have been used for the manipulation of the genomes of mammalian and plant cells (1)(2)(3). However, the use of meganuclease-induced recombination has long been limited by the repertoire of natural meganucleases amenable of customization.
HEs are grouped in several families; among them the LAGLI-DADG family is the most abundant. The family name derives from the conserved amino acid sequence motif containing the catalytic aspartic residue. The nucleases of this family cleave their DNA target along the minor groove to generate cohesive 4-bp-long 3Ј-OH overhangs. Two classes of LAGLIDADG nucleases can be found. One of them contains a single copy of the motif acting as homodimers, which bind palindromic or near-palindromic DNA target sequences. The other class contains two copies of the LAGLIDADG motif acting as monomers, which recognize and cleave non-palindromic DNA sequences (4). Catalysis requires the conserved acidic residue at the C termini of the two LAGLIDADG sequences in the active site and the coordination of divalent metal cations for phosphodiester hydrolysis (5). Members of the LAGLIDADG family have been engineered to specifically target new DNA sequences emerging as powerful tools for gene targeting. The most used template is I-CreI, whose scaffold has been redesigned using a combinatorial approach to target new specific DNA sequences (6).
I-CvuI is a homologue of I-CreI, and therefore a homodimeric LAGLIDADG homing endonuclease. This enzyme is * This work was supported by the European Community's Seventh Framework Programme (FP7/2007-2013) under BioStruct-X (Grant 283570). This work was also supported by the Ministerio de Economía y Competitividad (BFU2011-23815/BMC (to G. M.) and CTQ2014-56966-R (to F. J. B.)), the Fundación Ramón Areces (to G. M.), the Comunidad Autónoma de Madrid (CAM-S2010/BMD-2305 (to G. M.)). The authors declare that they have no conflicts of interest with the contents of this article. The atomic coordinates and structure factors (codes 5A72, 5A74, 5A77, and 5A78) have been deposited in the Protein Data Bank (http://wwpdb.org/). 1 To whom correspondence may be addressed. E-mail: jprieto@cnio.es. 2 To whom correspondence may be addressed. E-mail: guillermo. montoya@cpr.ku.dk.
coded by the genome of the chloroplast large subunit ribosomal RNA gene of the green algae Chlorella vulgaris. I-CvuI recognizes and cleaves a 24-bp-long DNA palindromic sequence (see Fig. 1). Here we report the crystal structure of I-CvuI in complex with its target DNA, hereafter termed Sro1.3, and with the target DNA of I-CreI, hereafter termed C1221 (7). We have solved the structures in the presence of Ca 2ϩ , Mg 2ϩ , and Mn 2ϩ to characterize I-CvuI cleavage and its DNA recognition pattern, and we have performed a comparison with I-CreI, the best characterized meganuclease template. The structure of this novel HE expands the current repertoire for engineering custom specificities used as a new scaffold or in hybrid constructs with other related HEs (8). Furthermore, after evaluating the metal dependence of I-CvuI catalysis, we analyzed in vitro the metal concentration dependence of the phosphodiester hydrolysis reaction using catalytic ions ranging from non-cleavable to cleavable concentrations.

Experimental Procedures
Protein Expression, Purification, and Crystallization-I-CvuI expression, purification, protein⅐DNA complex formation in the presence of Sro1.3 (5Ј-TCAGAACGTCGTACGACGTTC-TGA-3Ј) and crystallization have been described previously (9). Here we have incubated the complex in the presence of both non-catalytic ions (Ca 2ϩ at 2 mM) and catalytic ions (Mg 2ϩ and Mn 2ϩ at 2-20 and 2 mM, respectively). For comparison reasons, we also obtained the I-CvuI⅐DNA complex using the analogue DNA target from I-CreI, C1221 (5Ј-TCAAAACGTCGTACG-ACGTTTTGA-3Ј), in the presence of 2 mM Mg 2ϩ . The C1221, a palindromic 24-bp DNA target, derived from the pseudo-palindromic I-CreI wild-type target (10), was used here as a template for the in vitro cleavage experiments because this palindromic scaffold is cleaved in vivo by I-CreI with the same efficiency as the wild-type pseudo-palindromic target (11).
I-CvuI DNA Binding Studied by Isothermal Titration Calorimetry-Isothermal titration calorimetry experiments were performed in a MicroCal Auto-iTC200 (Malvern). Both protein and DNA samples were extensively dialyzed in 25 mM Hepes, 150 mM NaCl, 10 mM CaCl 2 , pH 7.4. I-CvuI (6 M) and Sro1.3 or C1221 (both at 65 M) were loaded into the calorimeter cell and titration syringe, respectively. Titrations were carried out using 25 injections of 1.5 l each injected at 180 -240-s intervals. Data analysis was performed on the concentrationnormalized heats by nonlinear regression using the MicroCal PEAQ-ITC analysis software (Malvern).
I-CvuI DNA Binding Studied by Microscale Thermophoresis-Microscale thermophoresis experiments were performed on a NanoTemper Monolith NT.115 instrument using 40% LED and 20 -40% IR-laser power. Laser on and off times were set at 30 and 5 s, respectively. Cy5-labeled oligonucleotides were purchased from TAG Copenhagen A/S (Copenhagen, Denmark). The concentration of the labeled oligonucleotides was kept constant at 25 nM, and the corresponding endonuclease binding patterns were titrated in 1:1 dilutions with the highest concentrations at 500 nM or 8 M. Samples were prepared in buffer containing 25 mM Hepes, 50 mM NaCl, 0.2% Pluronic F-127, 0.5 mg/ml BSA, pH 7.4, supplemented with either 10 mM CaCl 2 or 10 mM MgCl 2 , and loaded into standard treated capillaries (NanoTemper Technologies) for measurements. The K D was determined by fitting the change in thermophoretic depletion to the quadratic solution of the mass acton law. (The fluorescence during thermodiffusion was averaged for 1 s after the 10 s of IR-laser heating due to the higher signal to noise in this region. Overall, the IR-laser was on for 30 s.) Error bars are the standard deviation between three independent repeats (n ϭ 3).
I-CvuI in Vitro DNA Cleavage-To analyze the effect of cations, the I-CvuI enzymatic activity was assayed against Sro1.3 and C1221 included either in a BamHI-linearized yeast plasmid or in an XmnI-linearized pGEM-T plasmid, respectively, as described previously (12). The cleavage of linearized plasmid was performed at variable cation concentrations (between 1 and 10 M for MgCl 2 or MnCl 2 ) against 100 ng of substrate and with 150 ng of protein. In the assays, the wild-type I-CvuI was incubated with the DNA at 37°C in 10 mM Tris pH 8, 50 mM NaCl and the corresponding cation using a final volume of 25 l. The reactions were stopped by the addition of 6ϫ Buffer Stop (45% glycerol, 95 mM EDTA (pH 8), 1.5% (w/v) SDS, 1.5 mg/ml proteinase K, and 0.048% (w/v) bromophenol blue) and incubated at 37°C for 15 min. Samples were then electrophoresed in 0.6% agarose gels, and the intensity of the bands was observed under UV light.
Structure Determination, Model Building, and Refinement-All data were collected at 100 K, using a PILATUS 6M detector both at the PX beamline at the Swiss Light Source (SLS, Villigen) and at XALOC (ALBA, Barcelona, Spain). Data processing and scaling were accomplished using XDS (13) and SCALA from the CCP4 package (14) (see Tables 1 and 2). The structures were solved by molecular replacement as implemented in the program PHASER (15). The search model was based on the Protein Data Bank (PDB) entries 1G9Z (I-CreI-DNA-Mg 2ϩ ) or 1G9Y (I-CreI-DNA-Ca 2ϩ ) according to the a priori knowledge of catalytic or non-catalytic conditions. The structures were then subjected to iterative cycles of model building and refinement with Coot (16) and PHENIX (17). The identification and analysis of the protein-DNA hydrogen bonds and van der Waals contacts were achieved with the Protein Interfaces, Surfaces and Assemblies service (PISA) at the European Bioinformatics Institute. DNA structures were analyzed using 3DNA (18).

Results
Overall Structure of the I-CvuI⅐Sro1. 3 Complex-I-CvuI in complex with a 24-bp DNA duplex, corresponding to the palindromic sequence of its wild-type target, was crystallized as an enzyme⅐substrate complex with Ca 2ϩ and as an enzyme⅐ product complex with Mn 2ϩ . The overall fold of I-CvuI in complex with its DNA target is similar to I-CreI (Fig. 1). The structure shows a homodimer (with monomer A and monomer AЈ) with a clear two-fold symmetry axis between the LAGLIDADG helix of each monomer (␣1 and ␣Ј1). Each monomer contains the typical ␣␤␤␣␤␤␣ topology of the LAGLIDADG family having three extra helices at the C terminus. Two antiparallel ␤-sheets, ␤1-␤4 and ␤Ј1-␤Ј4, form a concave surface with an inner cylindrical shape where the DNA molecule is accommodated (Fig. 1A). The non-cleaved DNA and the cleaved I-CvuI DNA structures differ in the status of the DNA molecule ( Fig.  2), but the protein moieties are similar with a C ␣6 -162 rmsd of 0.27 Å.
I-CvuI versus I-CreI DNA Binding-I-CreI is the member of the LAGLIDADG family that has been extensively used in protein-DNA interaction redesign (7). Therefore, due to our final aim to provide a new scaffold for meganuclease design, we have compared the I-CvuI structure with this member of the family. I-CvuI shows respectively 37% identity and 55% sequence similarity with I-CreI. The crystal structures of I-CvuI and I-CreI (19) share a common overall folding; nevertheless some secondary structure differences are observed (Fig. 1B). As a consequence of their primary sequence differences, the main topology differences are located at ␣1, ␣3-␣4, and ␣6. Particularly interesting are the higher length of ␣6 in I-CvuI than in I-CreI and the presence of 2 at the end of ␣4 in I-CvuI, whereas in I-CreI, 2 is located at the end of ␣3.
The DNA interaction patterns in I-CvuI⅐Sro1.3 and I-CreI⅐C1221 complexes are very different (Fig. 3A), although the DNA sequences of the Sro1.3 and the C1221 are almost identical, differing only at positions Ϯ9 (Fig. 3A). However, the targets conserve only the interactions between the residues Gln-29 (Gln-26 in I-CreI) with Ϫ6G strandB and Arg-73 (Arg-70 in I-CreI) with Ϫ3G strandB and Ϫ4T strandA (Fig. 3A). The superimposition of the complexes I-CvuI⅐Sro1.3 and I-CreI⅐C1221 show that although both protein structures maintain the overall folding (C␣ rmsd 1.32 Å), the DNA structures display clear differences. The origin of the change in the folding of both targets seemed to be located at the positions where the sequences are different (Ϯ9) (Fig. 3, B and C), and to analyze the details, we +3 +4 +5 +6 +7 +8 +9 +10 +11

T C A G A A C G T C G T A C G A C G T T C T G A A G T C T T G C A G C A T G C T G C A A G A C T
.PSVLMEAALLADKVGLLTDGKKRTILAENVRECLKKLGHVV

I-CvuI Target Recognition Pattern and Site Recognition
Flexibility-Previous work reveals that I-CvuI is a highly specific enzyme. It displays no or low activity when it binds the DNA targets of homologous enzymes such as for example the one of I-CreI (8). We demonstrate here that the binding affinity of I-CvuI for the C1221 palindromic target of I-CreI is lower as compared with the affinity for its own target in the presence of both non-catalytic and catalytic cations (Figs. 4 and 5). I-CvuI shows a 3-fold higher affinity (53.3 versus 131.9 nM) for its own target versus the C1221 as measured by isothermal titration calorimetry (Figs. 4B and 5B). However, I-CreI cleaves the Sro1.3 in a similar degree to C1221 (8). Thus, to address specificity issues and how they could affect the site recognition flexibility for redesign purposes, we also solved the crystal structure of the complex formed by I-CvuI and C1221 (Fig. 6A, bottom), thus deciphering the protein-DNA contacts with this sequence. The Sro1.3 and C1221 are palindromic. They differ only in the two positions Ϯ9 of the whole 22-bp target (Fig. 6A). Exchanging the G/C base pairs at Ϯ9 in the Sro1.3 by A/T base pairs in the C1221 promoted rearrangements in the protein-DNA interaction interface. Seven out of eight amino acids are involved in direct interactions to bases with Sro1.3 (Fig. 6A,  top). These interactions were conserved in the I-CvuI⅐C1221 complex structure. A new extra residue, Gln-41, appeared involved in the I-CvuI⅐C1221 interface (Fig. 6A, bottom). This new interaction does not alter the conformation of the DNA target (Fig. 6B). However, although Ϫ9G strandA in Sro1.3 interacts specifically with Arg-33, the corresponding base in the palindromic C1221 (Ϫ9A strandA ) interacts with Gln-41 and Arg-43 through water-mediated contacts (Fig. 6C). Furthermore, Ϫ9C strandB in Sro1.3 does not display protein contacts, whereas the base in the same position in C1221 (Ϫ9T strandB ) interacts specifically with Arg-33. An additional difference is that Ϫ8T strandB in Sro1.3 interacts with Arg-43 through a watermediated contact, but its counterpart in C1221 shows a direct interaction with Arg-33. Active Site and Cleavage Mechanism-Divalent metal ions play an essential role in the phosphodiester catalysis in meganucleases (20). The conserved acidic residues at the active site coordinate the divalent metal ions. The general mechanism of cleavage of the phosphodiester bonds of DNA requires a nucleophile to attack the electron-deficient phosphorus atom, a general base to activate the nucleophile, and a general acid to protonate the leaving group. Positively charged groups are also needed to stabilize the phosphoanion transition state. The role of divalent metal ions during cleavage in LAGLIDADG meganucleases has been thoroughly studied (21,20). Three metal sites have been observed in the active center called A, B, and C. Sites A and B are located in opposite locations, whereas site C occupies the center of the active site ( Fig. 2A). The metal at site C is shared between the two catalytic sites and participates in a two-metal ion cleavage mechanism to hydrolyze the phosphodiesters of both strands (20).
The positions of the metal ions in the crystal structures of the I-CvuI were analyzed by collecting anomalous diffraction data, using crystals grown in the presence of CaCl 2 and MnCl 2 (Fig. 2,  B and C). The structure of the enzyme⅐substrate complex included the non-catalytic Ca 2ϩ ions, which were visualized by examining the anomalous difference Fourier maps. In contrast with I-CreI, which shows three Ca 2ϩ atoms in the active site (5), only two anomalous peaks could be detected in sites A and B of the I-CvuI active center (Fig. 2B). The Ca 2ϩ atoms are hexacoordinated with phosphates from both DNA strands (ϩ3G strandA , Ϫ1A strandB , and ϩ1A strandA , Ϫ3G strandB , respec-

Structure and DNA Binding Properties of Endonuclease I-CvuI
NOVEMBER 27, 2015 • VOLUME 290 • NUMBER 48

JOURNAL OF BIOLOGICAL CHEMISTRY 28731
tively), three water molecules, and the side chains of Asp-23-AspЈ-23 in each LAGLIDADG motif. We followed a similar strategy for the enzyme⅐product complex including the presence of Mn 2ϩ ions (Tables 1 and 2). Three anomalous peaks in sites A, B, and C were detected due to the presence of this metal in the I-CvuI active center (Fig.  2C). The central Mn 2ϩ ion is shared between the two acidic catalytic residues. The two external Mn 2ϩ ions are located in positions equivalent to the Ca 2ϩ cations. They also display similar interactions with neighboring atoms. The Mn 2ϩ ion in site A is coordinated with the side chain of Asp-23, the carbonyl of AlaЈ-22, the 5Ј-phosphate of ϩ3G strandA , the phosphate of Ϫ1A strandB , and a water molecule outside the active site. The Mn 2ϩ in site B interacts with the 5Ј-phosphate of ϩ1A strandA , the phosphate of Ϫ3G strandB , the main chain carbonyl of Ala-22, the side chain of AspЈ-23 in the LAGLIDADG motif of the other monomer, and a water molecule outside the active site. The Mn 2ϩ in site C interacts with both catalytic residues Asp-23-AspЈ-23, the carbonyls of the 3Ј-phosphates ϩ3G strandA , Ϫ3G strandB , and the 5Ј-hydroxyls of ϩ2C strandA , Ϫ2C strandB . Thus, our data show that the cleaved complex contains the expected three metal sites and follows a mechanism similar to other members of the LAGLIDADG family (20,21). The comparison of the I-CvuI Ca 2ϩ and Mn 2ϩ anomalous maps (Fig. 2) suggests that I-CvuI may follow a mechanism similar to the monomeric I-DmoI (22). However, the structural organization of the homodimeric I-CvuI active site is symmetric, suggesting that there is no cleavage preference for one strand over the other during catalysis, such as in the case of I-DmoI (20).
Besides the residues involved in the metal binding, other critical residues in HE catalysis are responsible for the generation of the basic pocket needed for favoring the transition state (23). In this sense, Lys-102 is located in the enzyme active site and is a candidate to act as a Lewis acid (stabilizing the pentacoordinate transition state) or to activate a proton donor in the cleavage reaction. A sequence alignment with other homodimeric HEs suggests that the other residue involved in the basic pocket in I-CvuI should be Arg-54. Interestingly, in I-CvuI, a glutamine is found in this position instead of the conserved arginine/lysine (8). By analogy with I-CreI, additional residues such as Arg-73 in I-CvuI (Arg-70 in I-CreI) were located at the active site in the same positions. The mutation of these amino acids abolishes I-CreI endonuclease activity (24  Catalysis Is Metal Type-and Concentration-dependent-As described previously (8), I-CvuI catalysis is dependent on the type of metal. To fully understand the cleavage characteristics of I-CvuI, we analyzed the activity of the enzyme in the presence of Ca 2ϩ and Mg 2ϩ , combining structural, binding, and in vitro cleavage information using its target Sro1.3. As has been observed in other meganucleases, the presence of 2 mM Ca 2ϩ did not promote cleavage of the DNA target, in agreement with our enzyme-substrate structure (Fig. 2B). On the other hand, in the presence of the same concentration of Mn 2ϩ , the crystal structure revealed that both DNA strands were cleaved and that three metal ions were located at the active site (positions A, B, and C, Fig. 2C).
To investigate the metal ion type and concentration dependence for the catalysis, we performed in vitro cleavage experiments (Fig. 4A), showing a metal concentration dependence of phosphodiester hydrolysis when using Mg 2ϩ . Cleavage at 2 mM Mg 2ϩ is not complete (Fig. 4A), whereas at 10 mM, the enzyme cleaves the target completely (Fig. 4A).
In addition, by in vitro cleavage experiments, we checked the I-CvuI activity of the C1221 target in the presence of different concentrations of Mg 2ϩ . In contrast with the I-CvuI⅐Sro1.3 complex, high concentrations of Mg 2ϩ inhibited the cleavage of the C1221 target, whereas low concentrations allowed partial cleavage (Fig. 5A). Moreover, the crystal structure of I-CvuI⅐C1221 in presence of 2 mM Mg 2ϩ showed a mixture between the non-cut and cut target, in agreement with the in vitro cleavage experiments. We could not obtain crystals of this protein⅐DNA complex at higher Mg 2ϩ concentrations, hampering the collection of structural information. Therefore, to address whether this behavior is related to a catalytic cation-dependent affinity, we performed microscale thermophoresis experiments for the I-CvuI⅐C1221 complex at different magnesium concentrations (Fig. 5D). Interestingly, I-CvuI binding decreases for its nonspecific target DNA at higher cation concentrations (Fig. 5D). In fact, the presence of 10 mM Mg 2ϩ completely abolishes the binding, in agreement with the in vitro cleavage experiments that showed no cleavage at 10 mM Mg 2ϩ (Fig. 5, A and D).

Discussion
LAGLIDADG homing endonucleases are one of the scaffolds widely used to create highly precise protein "cutters" capable of generating a double-strand break in a desired genome region  NOVEMBER

Structure and DNA Binding Properties of Endonuclease I-CvuI
once their DNA specificity has been redesigned. However, their use is limited to the repertoire of known HEs not covering the whole genome sequence. To overcome this problem, we could redesign already known scaffolds or characterize novel HEs whose DNA pattern recognitions fulfill our DNA target requirements. Usually, the re-engineering of well known scaffolds, such as I-CreI, to recognize a given DNA sequence is expensive and time-consuming. Therefore, it could be cheaper to find other HEs that may offer the same or similar effects. Indeed, the current massive genome sequencing helps us find new scaffolds for these biotechnological purposes that greatly expand the number of DNA sequences that can be targeted. I-CvuI is an example of a novel HE with another DNA target pattern recognition giving us another useful tool for gene therapy. To check whether we can use I-CvuI as a new scaffold for recognizing new DNA sequences, we proceeded to its biochemical characterization (Fig. 4), and analyzed its DNA target recognition pattern using the crystal structure of the I-CvuI⅐Sro1.3 complex (Fig. 3). In addition, we also checked the structural and biochemical properties of I-CvuI binding to C1221 (Figs. 5 and 6). Our work shows how this HE recognition pattern is able to bind both the I-CvuI and the I-CreI palindromic targets. The comparison of the DNA recognition patterns of I-CreI and I-CvuI shows that the changes in the protein template are minimal; however, just the change of the bases at the Ϯ9 positions according to their sequence differences imposes important conformational differences. This effect has been observed previously in I-CreI variants targeting the human XPC (25) and RAG1 (26) genes. Furthermore, we have shown how the metal concentration influences I-CvuI catalysis using its DNA target and its quasi-analogue palindromic I-CreI target. Following some of our previous studies with members of the LAGLI-DADG homing endonuclease family, we suggest that the increase of the cation concentration may induce subtle conformational changes in the target that decrease binding (27). Thus, an increase of Mg 2ϩ ions could make more stringent the differences between C1221 and the Sro1.3, leaving the enzyme in a less favorable scenario and avoiding cleavage of the non-target DNA. Therefore, our data suggest that the DNA pattern recognition of I-CvuI can be altered with few changes in the DNA sequence, suggesting the possibility to use this new scaffold for HE redesign.
Thus, I-CvuI offers an alternative scaffold to well established HEs already used in the development of redesigned enzymes for therapeutic and biotechnological applications. The characterization of the I-CvuI⅐DNA complex should help in the production of intelligent molecular scalpels that recognize and substi-

Structure and DNA Binding Properties of Endonuclease I-CvuI
NOVEMBER 27, 2015 • VOLUME 290 • NUMBER 48 tute certain DNA sequences, may avoid the high cost of engineering some other traditional HE scaffolds, and may be useful for therapeutic and biotechnological purposes.