Biochemical and mechanistic analysis of the cleavage of branched DNA by human ANKLE1

Abstract ANKLE1 is a nuclease that provides a final opportunity to process unresolved junctions in DNA that would otherwise create chromosomal linkages blocking cell division. It is a GIY-YIG nuclease. We have expressed an active domain of human ANKLE1 containing the GIY-YIG nuclease domain in bacteria, that is monomeric in solution and when bound to a DNA Y-junction, and unilaterally cleaves a cruciform junction. Using an AlphaFold model of the enzyme we identify the key active residues, and show that mutation of each leads to impairment of activity. There are two components in the catalytic mechanism. Cleavage rate is pH dependent, corresponding to a pKa of 6.9, suggesting an involvement of the conserved histidine in proton transfer. The reaction rate depends on the nature of the divalent cation, likely bound by glutamate and asparagine side chains, and is log-linear with the metal ion pKa. We propose that the reaction is subject to general acid-base catalysis, using a combination of tyrosine and histidine acting as general base and water directly coordinated to the metal ion as general acid. The reaction is temperature dependent; activation energy Ea = 37 kcal mol−1, suggesting that cleavage is coupled to opening of DNA in the transition state.


INTRODUCTION
Recent evidence suggests that LEM-3 / ANKLE1 plays an important role in providing a final opportunity to process persistent DNA junctions that have escaped enzymatic processing, and that could lead to DNA bridges linking chromosomes thus blocking cell di vision. Ev ery physical connection between sister chromatids must be removed before cells can divide. Connections between segregating chromatids exist as chromatin bridges (1)(2)(3), or ultrafine bridges ( 4 , 5 ) that are associated with BLM and PICH helicases. LEM-3 was originally identified in a genetic screen for embryonic lethality in Caenorhabditis elegans resulting from exposure to ionizing irradiation ( 6 ). ANKLE1 is the corresponding ortholog in humans and other metazoans ( 7 ).
LEM-3 / ANKLE1 protein sequences include predicted GIY-YIG nuclease domains at the C-terminus, and LEM3 was shown to exhibit nucleolytic activity on supercoiled DN A ( 6 ). We have recentl y shown that human ANKLE1 carries out structure-selecti v e nucleolytic cleavage directed towards a variety of branched DNA species ( 8 ). These branchpoints would be expected to result from the failure to process recombination intermediates fully, or perhaps replica tion intermedia tes, and such structur es could r esult in the formation of inter-chromosomal linkages. A number of lines of evidence suggest that LEM-3 / ANKLE1 might be responsible for processing DNA bridges during anaphase: (1) LEM-3 is known to interact genetically with the Holliday junction-processing enzymes MUS81, SLX1 and SLX 4, with embryonic lethality resulting in lem-3 double mutants ( 3 , 9 ) (2) Cytological experiments have revealed that LEM-3 accumula tes a t the cellular mid-body where the two daughter cells abscise from each other ( 10 ) (3) In lem-3 mutants treated with DNA damaging agents, and when DNA replication or chromosome decondensation are compromised then chromosomal bridges persist ( 3 ).
Thus ANKLE1 has very much the substrate specificity expected for a nuclease whose role is to catch unprocessed or partially processed junctions late in cell division ( 8 ), and its ortholog LEM3 has been shown to be located just at the point in the dividing cells consistent with that role ( 10 ). AN-KLE1 has thus been described as the 'enzyme of last resort'. Despite clear indications of the probable importance of human ANKLE1, very little is known about how the enzyme operates as a nuclease. Like the SLX1 component of the multi-component SLX1-SLX4-MUS81-EME1 resolution complex it appears to be a GIY-YIG nuclease; these ar e r elati v ely rar e in eukaryotes. Ther e have been a number of crystallo gra phic studies of different GIY-YIG nucleases (11)(12)(13)(14)(15)(16)(17)(18), but very little mechanistic analysis. Given its likely significance in cell division it is a potential therapeutic target, and ther efor e it is of both intellectual and practical importance to learn more concerning the mechanism of ANKLE1 nuclease. Moreover aspects of catalytic mechanism learned for ANKLE1 can lik ely pro vide insight into the mechanism of other GIY-YIG nucleases, including SLX1.

Bacterial cloning and mutagenesis of human ANKLE1
N-terminal deletions of human ANKLE1 were generated by PCR of the coding region of isoform 1 of human An-kle1 (NCBI Reference Sequence: NP 689576), cloned into pGEX6P , using the Q5 Site-Directed mutagenesis kit (New England Biolab E0554S). DNA primers were chosen to delete sequences between the N-terminus and chosen positions (Supplementary Table S1). PCR products were transformed into E. coli DH5 ␣ and single colonies were grown in LB broth containing 100 mg / l carbenicillin and 50 M IPTG. After 16 h, induced protein production was analysed by SDS PAGE from an aliquot of the culture medium, while the r emaining cultur e was used to purify plasmid DNA by mini-prep (Qiagen). A deleted form beginning at methionine 331 (ANKLE1 331-615 ) was found to be fully acti v e, and used in all the experiments described in this work. This was found to contain a mutation R492H that facilitated growth in E. coli . Site directed mutation of acti v e site residues were performed by PCR using the same Q5 site directed mutagenesis kit and the appropriate primers (Supplemen-tary Table S1). All DNA constructs were verified by DNA sequencing.

Expression and purification of ANKLE1 331-615 R492H in E. coli
Ar ctic Expr ess E. coli were transformed with pGEX6P-ANKLE1 331-615 R492H and its mutants and grown in LB supplemented with 100 mg / l carbenicillin and 20 mg / l gentamicin at 37 • C until the OD 600 of the culture reached 0.6. The temperature of the culture was lowered to 12 • C. When the absorbance of the culture at 600 nm was between 1.0 and 1.5, ANKLE1 331-615 R492H production was induced by the addition of IPTG to a final concentration of 200 M and incubated for 24 h. Bacterial cells were harvested by centrifuga tion a t 7500 g for 15 min a t 4 • C . The cell pellet was resuspended in 40 ml / l of cell culture of 25 mM HEPES (pH 7.5), 1M NaCl, 5% glycerol, 0.5 mg / ml lysozyme and 1 tablet of complete protease inhibitors (Roche) and incuba ted a t room tempera ture for 15 min. Triton X100 was added to a final concentration of 0.2% (w / v) and the extract was then sonicated on ice at 30 W for 1.5 min using 10 s pulses separated by 30 s pauses using a FB705 sonicator (Fischer Scientific). The cellular extract was clarified by centrifuga tion a t 20 000 g for 30 min at 4 • C (Beckman rotor JA25.50). The supernatant was mixed with glutathione sepharose 4B beads (GE healthcar e), which wer e preequilibrated with 25 mM HEPES (pH 7.5), 1 M NaCl, 5% glycerol. The beads were then washed with the same buffer. GST-ANKLE1 331-615 R492H was eluted as a fusion protein by incubating the beads with 25 mM HEPES (pH 7.5), 50 mM NaCl, 5% glycerol and 20 mM gluta thione. Alterna ti v ely, GST-ANKLE1 331-615 R492H was cleaved with human 3C protease while bound to the glutathione column overnight in 25 mM HEPES (pH 7.5), 150 mM NaCl, 5% glycerol. GST fusion or non-fusion ANKLE1 331-615 R492H were further purified by chromato gra phy on Fracto gel SO 3 − (Merck) in 25 mM HEPES (pH 7.5), 5% glycerol with a linear NaCl gradient from 50 mM to 1 M, followed by purification on Mono S 5 / 50 GL (GE Healthcare) in 25 mM HEPES (pH 7.5), 10% glycerol, 0.5 mM EDTA with a linear NaCl gradient from 50 mM to 1 M.

Denaturing gel electrophoresis of proteins in SDS-containing buffer
ANKLE1 331-615 R492H purity was analysed by polyacrylamide gel electrophoresis in denaturing condition. Protein samples were subjected to electrophoresis in 10% polyacrylamide Bis-Tris NuPAGE (Invitrogen) in 50 mM MOPS, 50 mM Tris (pH 7.7), 3.47 mM SDS, 1 mM EDTA. Protein bands were detected using Instant Coomasie Protein Stain (Abcam) and electrophoretic migration was compared to standard proteins (PageRuler Plus Prestained Protein Ladder, Thermo Scientific).

Construction of DNA spla y ed Y-junctions
Oligonucleotides were purchased from Sigma-Aldrich.
Splayed Y-junction was pr epar ed by the hybridization of r-and x-strands (Supplementary Table S1). Wher e r equir ed, strands wer e radioacti v ely-[5 -32 P]-labelled by polynucleotide kinase (Thermo Scientific). 2 M of oligonucleotide was incubated with 0.17 M [ ␥ -32 P]-ATP at 37 • C in 50 mM Tris (pH 7.6), 10mM MgCl 2 , 5 mM DTT, 0.1mM spermidine for 30 min. 50 mM EDTA was added to terminate the kinase reaction before the addition of the other oligonucleotide in order to generate the splayed Y-junction substrate. Oligonucleotides were hybridized to gener ate splay ed Y-junctions by slow cooling from 80 • C to 4 • C . The substra te was purified by electrophoresis in 8% (29:1) polyacrylamide gels in 90 mM Tris-boric acid (pH 8.3), 2 mM EDTA.

Measurement of binding affinity by fluorescence anisotropy
The r-strand was synthesized with fluorescein attached at the 3 terminus, using a 6-fluorescein CPG column (Glen Research 20-2961). The splayed Y-junction was generated by hybridization with the h-strand of J3. Fluor escent anisotrop y measur ements wer e performed with a SLM Aminco 8100 fluorescence spectrometer. The excitation w avelength w as selected using an interference filter of 482.5 ± 31 (fwhm) nm, and the emission was selected using a monochrometer set at 530 nm with slits set at 8 (fwhm) nm. Fluorescence was measured at 20 • C using 100 nM splayed Y-junction in 20 mM HEPES-NaOH (pH 7.0 or 8.0), 50 mM KCl, 2 mM MnCl 2 , 0.1 mg / ml BSA. ANKLE1 331-615 R492H Y453F was added to the indicated concentrations. Fluor escence anisotrop y ( r ) was measur ed as : where I VV and I VH are fluorescence intensity with the polarizers parallel and crossed respecti v ely, and G is the ratio of detector sensitivity between vertical and horizontal polarization, and was measured for each determination in these studies. The data were fitted to the following binding model : where r 0 is the anisotropy in the absence of protein, r is the change in anisotropy over the full range of protein concentra tion, K d is the dissocia tion constant for ANKLE1 binding to the splayed Y-junction, [ Ank ] and [ Y ] are the ANKLE1 331-615 R492H and splayed Y-junction DNA concentrations respecti v ely.

Structur al pr ediction by AlphaFold
The structure of Ankle1 was predicted by the software Al-phaFold (model Q8NAG6). The overall confidence of the model for the nuclease domain (residue 448-566) of Ankle1 was mostly very high (64%) and the positioning of the backbone of the residues of the mutants in this study were either very high (Y453, Y486, K519 and E551) or confident (H498 and N565). For the measurement of cleavage rate as a function of temperatur e, conditions wer e identical apart from the incuba tion tempera ture.

Kinetic analysis of ANKLE1 cleavage reactions
For the study of the effect of flap lengths, a lower enzyme concentration was used (200 nM) and the products were resolved on a 15% (19:1) polyacrylamide sequencing gel, that was dried before exposure. All other conditions were as above.
For measurement of cleavage rate as a function of pH the following buffers were used for the indicated pH ranges : MES (pH 5.51, 6.01 and 6.39), MOPS (pH 6.5 and 7.2) and HEPES (pH 6.86, 7.58 and 7.8). pH was measured for the reaction buffer (20 mM buffer, 50 mM KCl and 10 mM MnCl 2 ) using an Accumet AB150 pH meter (Fischer Scientific) at 31.6 • C.

Cleavage of cruciform structures in negatively supercoiled plasmid DNA
The plasmid pHRX3 ( 19 ) was used in these experiments as it contains the sequence of x strand of J3. Negati v ely supercoiled pHRX3 was purified from E. coli DH5 ␣ with two rounds of CsCl density gradient centrifugation. Cleavage was carried out similarly to cleavage of splayed Y-junctions; 5 nM pHRX3 was preincubated with 2 M ANKLE1 331-615 R492H in 20 mM MES (pH 5.5), 50 mM KCl, 0.1 mg / ml BSA a t 31.6 • C . After 3 min incubation MnCl 2 was added to a final concentration of 10 mM to initiate the cleavage reaction. Aliquots were removed at chosen times and the reaction terminated by addition of 50 mM EDTA final concentration. 20 g Proteinase K (Thermo Scientific) was added and incubated a further 1 h at 37 • C. DNA was separated by electrophoresis in a 1% agarose gel in 90 mM Tris-boric acid (pH 8.3), 2 mM EDTA. DNA was stained with SYBR Safe and quantified by fluorescence on a Typhoon FLA 9500 fluorimager (GE Healthcare). Reaction progress of cleavage of supercoiled DNA was fitted to two exponential functions : where f cl is the fraction of supercoiled DNA cleaved at time t , and k 1 and k 2 are the two rate constants.
Permanganate probing of DNA structure 100 nM splayed Y-junction DNA radioacti v ely-[5 -32 P]labelled on the indicated strand was pre-incuba ted a t room temperature for 5 min in the presence of 0, 20 nM, 200 nM or 2 M ANKLE1 331-615 R492H in 10 mM cacodylate (pH 6.5), 50 mM KCl, 50 g / ml calf thymus DNA and either 1 mM EDTA (for the acti v e enzyme) or 10 mM MgCl 2 (for the Y453F mutant). The 20 l reactions were initiated by adding 2 l of a 12.5 mM KMnO 4 solution and terminated by addition of 1.5 l of ␤-mercaptoethanol after 1 min. Reaction products wer e pr ecipitated by addition of 1 / 10 volume of 3 M sodium acetate and 3 volumes of ethanol, resuspended in 100 l of 1 M piperidine and incuba ted a t 95 • C for 30 min. Samples were dehydra ted by addition of n -butanol and removal of the solvent phase, followed b y v acuum desiccation. Dried samples wer e r esuspended in formamide and analysed on denaturing 15% polyacrylamide sequencing gels. These were dried, exposed to storage phosphor screens and visualised with a Typhoon FLA 9500 fluorimager (GE Healthcare). Purine sequence markers were generated by subjecting the same substrates to reaction with formic acid ( 20 ) followed by piperidine cleavage as described above.

Expression of human ANKLE1 in bacteria
While we have previously expressed human ANKLE1 in insect cells ( 8 ) it would clearly facilitate analysis to express acti v e enzyme in bacteria, where site-specific mutants could be easily generated. N-terminal deletions of human ANKLE1 were generated by PCR of the coding region of isoform 1 of human ANKLE1 (Figure 1 A). We found that the Cterminal fragment beginning at methionine 331 (see the protein sequence in Supplementary Figure S1) was fully acti v e, so this fragment (termed ANKLE1 331-615 ) was used in all subsequent analysis. Sequencing re v ealed that this contains a R492H mutation that facilitated growth in E. coli without pre v enting acti vity. This mutation is present in all constructs used in this work. hANKLE1 331-615 R492H was expressed as an N-terminal fusion with GST, allowing purification on glutathione sepharose 4B beads. The fusion protein was cleaved while bound to the glutathione column with human 3C protease to release the isolated hANKLE1 331-615 R492H fragment. This was further purified by chromato gra phy on two sequential ion-exchange columns. The purity of the final hANKLE1 331-615 R492H product was demonstrated by polyacrylamide gel electrophoresis under denaturing conditions (Figure 1 B).

Human ANKLE1 331-615 R492H expressed in bacteria is active
The pure human hANKLE1 331-615 R492H expressed in bacteria is highly acti v e. The pattern of cleavage on a splayed Y-junction DNA (Supplementary Figure S2) shows cleavage from 3 nt into the duplex from the branchpoint, progressi v ely moving down the duplex region. This is exactly what was observed using the form of human ANKLE1 expressed in insect cells ( 8 ). hANKLE1 331-615 R492H cleaves the splayed Y-junction DNA with a rate of 52 ± 1.7 × 10 −3 s −1 under our standard conditions in Mn 2+ ions a t 37 • C . While the standard substrate has two single-stranded regions of 25 nt in length, these could be substantially shortened (either the 3 or the 5 strand, or both symmetrically) with retention of activity (Supplementary Figure S2).

ANKLE1 331-615 R492H is monomeric in free solution
Holliday junction-resolving enzymes in general bind to f our-wa y DNA junctions in dimeric form, and introduce symmetrically-paired cleavages within the lifetime of the DNA-protein complex (21)(22)(23)(24). While ANKLE1 will cleave f our-wa y junctions, it does so more slowly than splayed Y-junction DNA ( 8 ), its best substrate. It ther efor e seemed probable that ANKLE1 acts in monomeric form. We analyzed the state of multimerization of hANKLE1 331-615 R492H using gel filtration. Our purified protein was applied to a Super de x 75 (10 / 300) column and its elution profile measured by light absorption (Figure 2 A). In parallel a sample of carbonic anhydrase (molecular mass = 29000 Da) was applied to the same column. The elution profiles of the two proteins are closely similar. Since the calculated molecular mass for a monomer of ANKLE1 331-615 is 33000 Da we conclude that hANKLE1 331-615 R492H is monomeric in solution.

ANKLE1 331-615 R492H binds to a junction in monomeric form, with nanomolar affinity
We have analysed the binding of catal yticall y-inacti v e hANKLE1 331-615 R492H Y453F to splayed Y-junction DNA by following the fluorescence anisotropy of fluorescein attached to the 3 terminus of the single-stranded section. 100 nM of fluorescent splayed Y-junction was titrated with increasing concentrations of ANKLE1 331-615 R492H and the values of anisotropy plotted in Figure 2 B. The anisotropy of the fluorophore attached to free DNA is 0.06, indica ting tha t the fluorescein is very mobile. On addition of hANKLE1 331-615 R492H the anisotrop y incr eases showing that mobility of the fluorophore becomes restricted on protein binding. There is a linear increase in anisotropy up to a plateau; this is characteristic of stoichiometric binding, where the concentration of DN A is significantl y greater than the K d . The plateau value of the anisotropy of 0.15 is  reached at 100 nM protein, i.e. the same as the concentration of the splayed Y-junction DNA, showing that the stoichiometry is 1:1 protein:DNA. The data can be fitted to a simple binding isotherm. Although the binding affinity cannot be measured with high precision, simulation of the binding isotherms (Supplementary Figure S3A) indicates that the dissociation constant K d is close to 1 nM. These data indica te tha t a single molecule of hANKLE1 331-615 R492H binds to the splayed Y-junction with high affinity.
The titration was repeated at the higher pH value of 8. Under these conditions only a very small change in anisotrop y was measur ed (Supplementary Figur e S3B), indica ting tha t very little binding occurs a t this pH.

ANKLE1 331-615 R492H makes a single incision into a crucif orm f our -w ay junction
The gel filtration data indicate that hANKLE1 331-615 R492H is monomeric in free solution, and the binding experiments show that it binds to splayed Y-junction in monomeric form. We would ther efor e expect that it cleaves only one strand in a DNA branchpoint. The best way in which to distinguish unilateral and bilateral cleavage of a f our-wa y junction is to study the cleavage of a cruciform in a negati v ely supercoiled plasmid (21)(22)(23); the cruciform comprises a four-way junction connecting two stem-loops and the two arms of the plasmid. The principle is shown in Figure 3 A. If an enzyme introduces a single cleavage this releases the supercoiling and the cruciform is reabsorbed so that there is no substrate remaining to be cleaved a second time. In that situation the DNA is cleaved on one strand, leaving an open circular DNA product. Howe v er, two cleavages can be made if they occur sim ultaneousl y, or within the lifetime of the enzyme-DNA complex. In that case the product will be linear plasmid. Since supercoiled, open circular and linear forms of the plasmid DNA are readily separated by agarose gel electrophoresis this provides a straightforward test distinguishing between unilateral and bilateral cleavage.
A time course of cleavage of a supercoiled cruciformcarrying plasmid by hANKLE1 331-615 R492H is shown in Figure 3 B, and the results quantified and plotted in Figure 3 C. Over a 2 min. incubation the nuclease ANKLE1 331-615 R492H converts supercoiled plasmid into open circular, with no linear product visible. This demonstra tes tha t hANKLE1 331-615 R492H introduces a unilateral cleavage into the cruciform junction, consistent with the enzyme acting as a monomer that cleaves one strand only. This is in marked contrast with GEN1 for example, where the open circular form is generated as a transient intermediate and linear DNA is the e v entual product, showing that like other junction-resolving enzymes it acts as a dimer making bilateral cleavages ( 24 ). Reaction progress was best fitted using two exponential functions, indicating a brief lag phase prior to cleavage of the DNA.

The active center of ANKLE1
At the present time there is no crystal structure of ANKLE1 or any ortholog including LEM3. Howe v er, the structure of human ANKLE1 has been modelled using AlphaFold ( 25 ), and was downloaded from Uniprot ( 26 ) as model Q8NAG6. The putati v e acti v e center of the enzyme is shown in Figure 4 A. The predicted secondary structure conforms closely to that of other GIY-YIG nucleases for which structures have been determined by X-ray crystallo gra phy, including I-TevI ( 11 ), UvrC ( 12 ), R. Eco 29kl ( 13 ) and SLX1 ( 15 , 18 ). Each comprises three antipar allel ␤-str ands surrounded by ␣-helices. An alignment of putati v e acti v e site sequences in vertebrates is shown in Figure 4 B; all the expected conserved critical amino acids are present in human ANKLE1 and 100% conserved. These are shown on the AlphaFold prediction in stick form. They comprise the tyrosine residues of the GIY-YIG motif (Y453 and Y486), together with a histidine (H498), glutamate (E551), lysine (K519) and asparagine (N565). Each occurs in the position equivalent to that of the nucleases of known structure.
Each putati v e acti v e center residue has been mutated, and the rate of cleavage of the splayed Y-junction substrate measur ed (Figur e 4 C, D and Table 1 ) in the presence of Mn 2+ ions. Muta tions a t each amino acid led to enzymes that wer e impair ed in cleavage activity, from a 60-fold reduction (K519A) to no detectable activity (Y453F and E551A). The  r esults ar e consistent with these conserved amino acid side chains playing important roles in the catalytic mechanism of the nuclease. The proximity of Y453, Y486 and H498 in the modelled acti v e center and their sensitivity to mutation suggests a potential role in general acid-base catalysis, as suggested for UvrC ( 12 ) and R. Eco 29kl ( 13 ). Gi v en the pr obable r ole of a divalent metal ion in catalysis it is also v ery possib le that E551 and N565 coordinate a metal ion, discussed below.

The dependence of ANKLE1 331-615 R492H cleavage on the nature of the metal ion
Our standard conditions for the study of the cleavage of DNA junctions by hANKLE1 331-615 R492H include Mn 2+ ions. We hav e e xplored the rate of cleavage on the splayed Y-junction in different divalent metal ions. The activity is 3.5-fold lower in the presence of Mg 2+ ions, but the initial rate of cleavage is 5.2 fold faster in Co 2+ ions ( Table 1 ). The enzyme is essentially inacti v e in the presence of Ca 2+ ions. Thus the rate follows the order Co 2+ > Mn 2+ > Mg 2+ >> Ca 2+ . While the low activity in Ca 2+ ions is likely a result of differences in size and coordination, cleavage activity is log-linear with the p K a of the other three ions (Figure 5 A). These data are consistent with proton transfer by metal ionbound water molecules in the transition state, or potentially with the metal ion acting as a Lewis acid. As we discuss below, it is likely that a divalent metal ion is bound into the acti v e site to act as a general acid in the hydrolysis of the phosphodiester bond.
The dependence of ANKLE1 331-615 R492H cleavage rate upon pH Gi v en the possible involvement of tyrosine and histidine side chains in proton transfer in catalysis we explored the dependence of the rate of cleavage of the splayed Yjunction by hANKLE1 331-615 R492H upon pH. Cleavage rate was measured in the presence of Mn 2+ ions over the range 5.5-7.8 (Figure 5 B). The rate rises over the pH range 5.5-7.2, well fitted by a single ionization corresponding to a p K a = 6.9 (Figure 5 C). As pH values rise above 7.2 the rate falls faster than can be fitted by a second ionization. The pH dependence of the rate of cleavage by hANKLE1 331-615 R492H H498A was also measured and the data plotted in Figure 5 B and C. These data exhibit a different profile, with a steady increase in cleavage rate across the whole range of pH. The measured value of the p K a coupled with the effect of the H498A mutation on both rate and pH profile indicates that histidine 498 likely participates in proton transfer in the reaction mechanism for phosphodiester hydrolysis.
The dependence of ANKLE1 331-615 R492H cleavage rate upon temper atur e, and possible opening of the junction The progressi v e nature of the cleavage of DNA junctions by ANKLE1 suggests an opening of the dsDNA that progresses into the DNA with successi v e cleavage e v ents. This might be reflected in the energy of activation for cleavage, and we ther efor e investigated the temperature dependence of the cleavage of the splayed Y-junction by hANKLE1 331-615 R492H in the presence of Mn 2+ ions. The data are presented in the form of an Arrhenius plot of the natural logarithm of the observed cleavage rate ( k obs ) as a function of inverse temperature in Figure 6 . The plot is linear, with a gradient corresponding to an activation energy E a = 36.8 kcal mol −1 .
We have also measured the rate of cleavage of a supercoilstabilized cruciform as a function of temperature (Supplementary Figure S4). The Arrhenius plot is linear, with a gradient corresponding to an activation energy E a = 36.9 kcal mol −1 . This is almost the same as the value measured for the splayed Y-junction.
This relati v ely large acti vation energy would be consistent with a r equir ement for structural distortion of the DNA substrate concurrent with bond cleavage. One possibility would be an opening of the DNA helix. We explored whether any unpairing of the helix of the double-stranded portion of the splayed Y-junction DNA prior to cleavage could be detected using probing by potassium permanganate. Unstacked thymine nucleobases can be oxidised at the C5-C6 double bond to form cis -5,6-dihydroxy thymine that renders the backbone susceptible to cleavage by 1 M piperidine at 95 • C. Splayed Y-junction was radioacti v ely  (open circle symbols). The lowering of rate above pH 7 cannot be fitted to a simple ionization process, and measurements of binding indica te tha t af finity becomes lower at these pH values (Supplementary Figure S3). In ( C ), the rate of cleavage is plotted on a linear scale as a function of pH over the range 5.5 to 7.0. Rates have been measured in triplicate, and the error bars are one standard deviation. In most cases these are smaller than the symbols. These data for hANKLE1 331-615 R492H have been fitted to a single ionization indicated by the line. From this a p K a of 6.9 has been calculated.

DISCUSSION
We hav e e xpressed the C-terminal fragment of human AN-KLE1 comprising amino acids 331-615 in E. coli and found that it is acti v e on splayed Y-junction junctions and supercoil-stabilized crucif orm f our-wa y DNA junctions with r ates compar ab le to full-length ANKLE1 e xpressed in insect cells ( 8 ). A branchpoint is essential for ANKLE1 cleavage of DNA (i.e. a duplex is a very poor substrate), but the length of either or both single-stranded sections is not critical. We show here that hANKLE1 331-615 R492H is predominantly monomeric in free solution, and binds to a splayed Y-junction as a monomer, with high affinity ( K d ∼ 1 nM). This is fully consistent with the demonstration that the enzyme cleaves the f our-wa y junction of a supercoilstabilized cruciform structure unilaterally. Thus although a f our-wa y junction is a substrate f or ANKLE1, the result is a junction containing a single cleaved strand rather than complete resolution. Gi v en that the N-terminal region of hANKLE1 that is not present in the 331-615 fragment has ankyrin repeats this could mediate pr otein-pr otein interactions that might be important in its function. Howe v er, the observation that hANKLE1 331-615 R492H acts in monomeric form is consistent with the primary substrate being the splayed Y-junction or similar structure, and AN-KLE1 is most acti v e on that branchpoint. Many GIY-YIG nucleases act in monomeric form ( 16 , 27 ). It is interesting to note that SLX1 (another GIY-YIG nuclease) is directly involved in four-way junction resolution, but works in partnership with another nuclease MUS81 to achie v e the two cleavages r equir ed for r esolution ( 1 , 28-31 ).
The pattern of cleavages on a splayed Y-junction by fulllength ANKLE1 and hANKLE1 331-615 R492H are essentially identical, and both are consistent with a progressi v e mov ement along the dsDNA from the junction with the single-stranded DNA. We show here that the cleavage reaction is strongly temperature dependent, corresponding to a large activation energy E a = 37 kcal mol −1 ; such a large activation energy is likely to arise from distortion of the DNA structure. The activation energy is virtually identical for cleavage of both the splayed Y-junction and the f our-wa y junction of a cruciform, so is unlikely to arise from distortion of the junction structur e. Mor eov er we observ e no reacti vity of the dsDNA to permanganate on addition of hANKLE1 331-615 R492H in EDTA or hANKLE1 331-615 R492H Y453F in the presence of Mg 2+ ions, so in the absence of DNA cleavage the enzyme does not appear to induce local DNA melting in the ground state. DNA distortion might ther efor e involve unwinding of the DNA that preserves base pairing and does not result in elevated reactivity to permanganate. Or ANKLE1 might open the DN A locall y as it cleaves, so str and separ ation is part of the activation process. Having made the first cleavage the enzyme can then make a second opening and cleavage, and so moves pro gressivel y into the dsDNA section of the junction.
The structure of the acti v e center of human ANKLE1 modelled using AlphaFold shows three antiparallel strands of ␤ sheet, with the putati v e functional tyrosine residues on the N-terminal (Y486) and central (Y453) strands. These are flanked above and below by helices on which are displayed the key histidine H498, glutamate E551 and aspar agine N565. This arr angement of protein structure and the disposition of critical side chains is typical of a number of GIY-YIG nucleases that have been determined structurally (11)(12)(13)(14)(15)(16). The putati v e acti v e residues of ANKLE1 ar e cluster ed with a radius of 4.3 Å , and all exhibit significantly impaired activity on mutation.
Like the majority of junction-selecti v e nucleases, AN-KLE1 catalyses the hydrolysis of a specific phosphodiester linkage in a DNA junction, with transfer of the phosphate to the 5 -terminus that is cr eated (Figur e 7 A). In principle this is likely to be catalysed by general acid-base catalysis in which the water nucleophile is deprotonated and the 3oxyanion leaving group is protonated. Many nucleases employ two-metal ion mechanisms in which metal ion-bound water molecules act as general base and general acid. Howe v er the GIY-YIG nucleases depart from this paradigm.
Our data suggest two important elements in the catalysis of phosphodiester bond cleavage by ANKLE1. First, a tomic muta tion of Y453 in w hich onl y the phenolic oxygen is removed leads to undetectab le le v els of acti vity. The reaction rate of the unmodified enzyme depends on pH over the range 5.5-7.0 corresponding to a p K a = 6.9. That would be consistent with a histidine undergoing proton transfer in the transition state, and the pH profile of the H498A mutant is substantially altered. In the structure predicted by AlphaFold, N2 of H498 is 2.9 Å from the phenolic oxygen of Y453. A very similar arrangement of tyrosine and histidine was observed in the crystal structures of Eco 29kI ( 13 ) and SLX1 ( 17 ) with histidine N2 2.6 and 2.7 Å from the oxygen of the tyrosine respecti v ely. We ther efor e hypothesize that the proton from Y453 is transferred to H498, and the phenolic oxygen then acts as a general base to deprotonate the nucleophilic water molecule (Figure 7 B). We propose that the second element in the catalysis by AN-KLE1 involves a divalent metal ion. We have observed that the ANKLE1 reaction rate is lo g-linearl y dependent on the p K a of the ion present, implying that an ion-bound innersphere water molecule is involved in proton transfer in the Nucleic Acids Research, 2023, Vol. 51, No. 11 5753 transition state (Figure 7 B). The ion is likely coordinated by E551 and N565; both E551A and N565D are totally inacti v e. A di valent metal ion was observed 2.1 Å from the carboxylate group of the corresponding glutamate in the crystal structure of UvrC ( 12 ) indicating inner-sphere coordina tion. Muta tion of tha t glutama te, and also the corresponding one of SLX1 ( 15 ), led to complete loss of activity of both nucleases.
The general acid-base catalysis model we are proposing for human ANKLE1 is very similar to those proposed on the basis of crystallo gra phic data for other GIY-YIG nucleases ( 12 , 13 ). Howe v er, kinetic ambiguity does not allow us to distinguish a model in which the tyrosine acts as general base and metal ion-bound water as the general acid or vice versa on the basis of our kinetic data. But comparison with crystallo gra phic structures of other GIY-YIG nucleases bound to DNA shows that in general the critical tyrosine is observed to be on the side of the 5 O while the metal ion and the glutamate are found on the side of the 3 O ( 13,14 , 16 ). We ther efor e pr opose that in the hydr olysis of the phosphodiester linkage by ANKLE1 the tyrosine (assisted by the histidine) acts as the general base to deprotonate the water nucleophile, while the hydrated metal ion acts as general acid to protonate the O3 leaving group as depicted in Figure 7 B. This model is consistent with all the available mechanistic data for ANKLE1, together with structural data for other GIY-YIG nucleases. Gi v en the apparent similarities in acti v e centers beween the GIY-YIG nucleases it is likely that they employ closely similar catalytic mechanisms, and that the mechanistic analysis described here can be extrapolated to the other enzymes. This might be of particular interest in the case of SLX1 that participates in f our-wa y junction resolution in the context of the SLX1-SLX4-MUS81-EME1 complex ( 1 , 28 , 32-34 ), where it introduces one of the two requir ed cleavage r eactions and is thus comparable with AN-KLE1 cleaving one strand in Y-shaped DNA junctions.

DA T A A V AILABILITY
The data underlying this article are available in the article and in its online supplementary material.

NOTE ADDED IN PROOF
While this paper was in submission Chan and co-w ork ers published a paper showing that human ANKLE1 is localized at the cell midbody: Jiang, H., Kong, N., Liu, Z., West, S. C. and Chan, Y. W. (2023) Human endonuclease AN-KLE1 localizes at the midbody and processes chromatin bridges to pre v ent DNA damage and cGAS-STING activation. Adv. Sci. 10, e2204388.

SUPPLEMENT ARY DA T A
Supplementary Data are available at NAR Online.