Mismatch binding, ADP-ATP exchange and intramolecular signaling during mismatch repair

Abstract

The focus of this article is on the DNA binding and ATPase activities of the mismatch repair (MMR) protein, MutS—our current understanding of how this protein uses ATP to fuel its actions on DNA and initiate repair via interactions with MutL, the next protein in the pathway. Structure-function and kinetic studies have yielded detailed views of the MutS mechanism of action in MMR. How MutS and MutL work together after mismatch recognition to enable strand-specific nicking, which leads to strand excision and synthesis, is less clear and remains an active area of investigation.

1. MutS actions in MMR

Mismatch repair is conducted by a group of proteins whose task is to recognize errors in DNA, resect the error-containing strand and coordinate its resynthesis; reviewed in [1-5]. The θ-shaped MutS dimer searches duplex DNA for mispaired or looped out bases and binds these sites with high affinity, selecting them as targets for repair (Fig. 1); reviewed in [6,7]. Subsequent interaction of the MutL dimer with MutS and DNA leads to nicking of the strand flanking the error either by MutL-stimulated MutH endonuclease activity (in methyl-directed repair systems; e.g., E. coli and like γ-proteobacteria) or by MutL endonuclease activity (in methyl-independent repair systems; e.g., other bacteria and eukaryotes); reviewed in [8-11]. While MutH specifically recognizes and nicks the unmethylated nascent strand of a newly replicated, hemimethylated duplex, MutL nicking can be directed to the nascent strand via asymmetric interactions with the sliding clamp (β/PCNA) that confers processivity on replicative DNA polymerases. Once nicked, this strand is excised past the mismatch and the DNA replication machinery is recruited to complete resynthesis and ligation. The core MMR proteins, MutS and MutL, are evolutionarily conserved and manifest in eukaryotes as heterodimers—MutSα (Msh2-Msh6), which is specialized for mispaired bases and 1-2 base loops, MutSβ (Msh2-Msh3), which is specialized for larger loops, as well as MutLα (human MLH1-PMS2/yeast Mlh1-Pms1) and MutLγ (Mlh1-Mlh3) that work with MutSα and MutSβ, respectively (some eukaryotic MutS and MutL homologs also function in DNA recombination); reviewed in [12,13]. Both MutS and MutL proteins possess ATPase activity, which drives an ordered series of conformational changes and interactions between the proteins and DNA leading from mismatch recognition to strand nicking and subsequent repair. Below is a summary of our current understanding of ATPase-coupled MutS actions at the initiation of MMR (and a brief discussion of MutL as well), based on structure-function and kinetic studies of E. coli, T. aquaticus (Taq), S. cerevisiae, and human proteins (kinetic data are provided for Taq MutS as an example). Many aspects of the MutS mechanism are broadly consistent across these model systems and these are discussed in the main, although there are a few differences in detail that may be attributable to variations in the proteins themselves or in the experimental approaches and/or the depth of study.

Fig. 1. T. aquaticus MutS bound to a single T-loop DNA.

Crystal structures of Taq MutS dimer, in apo form (PDB code: 1EWR) and bound to DNA containing a single T-loop (PDB codes: 1EWQ). The mismatch binding domains I are shown in pink, T-loop in black, stacking Phe in red, ADP in stick representation, and IAANS fluorophore location marked with an asterisk (see data in Fig. 2, 3).

1.1. A dynamic process of mismatch recognition

MutS was long known to bind mismatches with higher affinity than matched (homoduplex) DNA [14-16], but it was not until single-molecule (sm) tracking of fluorophore-labeled yeast Msh2-Msh6 (MutSα) on λ DNA substrates that it became clear the protein moves on the duplex by 1D diffusion, scanning it for MMR target sites [17]. Subsequent sm-tracking of yeast MutSα as well as sm-FRET and sm-polarization measurements of Taq MutS revealed that the protein rotates along the helical contour of DNA in this search mode, likely making non-specific contacts that interrogate the bendability of the duplex, and then stalls when it encounters a mismatch [18-21]. Initial contact with a mismatch is fast and weak, as evidenced by the 3D collision rate constant of 10⁷ M⁻¹ s⁻¹ (at 40 °C) between Taq MutS and a single T-loop-containing DNA that yields a complex with K_D ~5 μM [22,23]—similar to the low affinity MutS displays for matched DNA [16,24]. This complex then undergoes rate-limiting transformation, involving changes in both MutS and DNA structure, into another complex that has about 10³-fold higher affinity (K_D ~5 nM). Stopped-flow experiments monitoring the conformations of Taq MutS (by fluorescence intensity of IAANS dye-labeled mismatch binding domains I) and DNA (by FRET between fluorophores flanking the T-loop) reveal that the DNA bends in concert with domain I rearrangement during this slow step following initial contact (30 s⁻¹ at 40 °C; Fig. 2) [23].

Fig. 2. Mismatch binding by MutS.

(a) In stopped flow experiments, quenching of IAANS fluorescence reports Taq MutS domain I movement upon DNA binding (0.1 μM MutS_IAANS mixed with 3 μM DNA(+T); blue trace), and increasing FRET between AF488 and AF594 dyes flanking the T-loop reports DNA bending upon MutS binding (0.03 μM DNA(+T)_AF488-AF594 mixed with 3 μM MutS; red trace). (b) The rates of these conformational changes increase hyperbolically with titrant concentration to a maximum of 25 – 30 s⁻¹ with an apparent K_1/2 of 3 – 4 μM. The data fit well to a 2-step binding model in which initial rapid collision between MutS and DNA forms a weak complex (K_D1 = 5 μM) followed by intramolecular conformational changes to form a tight complex (k₂ = 30 s⁻¹, k₋₂ = 0.05 s⁻¹; K_D2 = 2 nM).

The nature of the protein movements in this second step is revealed by sm-FRET between donor and acceptor dyes on Taq MutS domains I that form the bridge of the θ-shaped dimer. They flip between closed (high FRET) and open (low FRET) states in apo MutS, moving from ~30 – 70 Å apart, a dynamic that explains their lack of resolution in the Taq MutS crystal structure [25] and high solvent accessibility in deuterium exchange mass spectrometry of E. coli and yeast MutS [26], in the absence of DNA (Fig. 1). When MutS binds and moves on homoduplex DNA, domains I adopt a relatively open conformation that likely enables scanning for mismatches. Note that the DNA binding clamp domains IV, which close one end of the θ, adopt a dynamic mostly open conformation in apo MutS as well (allowing DNA to access the interior of the dimer), but remain mostly closed when MutS is on DNA (keeping the protein topologically linked to the duplex for mismatch search and recognition; Fig. 1) [25,27]. Upon encountering a mismatch, domains I move toward each other and adopt a closed conformation [19]. The outcome of this transition is likely the stable mismatch recognition complex observed in all crystal structures of Taq, E. coli and human mismatch-bound MutS proteins solved thus far, in which the DNA is bent 45 – 60° at the mismatch site, and domain I of only one MutS subunit (Msh6 in eukaryotic MutSα) has inserted a Phe-X-Glu motif into the minor groove to make specific stacking and hydrogen bonding contacts with the mismatched base (Fig. 1) [25,28-30]. Recent AFM and sm-FRET studies of Taq and E. coli MutS have revealed formation of a slightly bent/unbent DNA complex, which is also considered an important intermediate during the process of recognition [15,24,31-33].

The uniformity of the mismatch-bound MutS structures solved thus far does not explain the variation in repair efficiency observed with different mismatches. Mismatch type, sequence context, even the orientation of MutS binding to an asymmetric mismatch (e.g., Phe stacking against T versus G in G:T) can affect repair efficiency [14,33-37]. These variables likely affect the conformational dynamics of the DNA and protein and hence the lifetime of the final mismatch recognition complex, which appears to be a critical factor for subsequent signaling of repair (e.g., yeast MutSα dissociates about 100-fold faster from the poorly repaired 2-Aminopurine:T mismatch than from the best repaired G:T mismatch) [38]. Thus, transition from a weak collision complex to a stable mismatch recognition complex is a key kinetic proofreading step whereby MutS first selects bona fide mismatches for repair. It sets the stage for subsequent ATP-coupled MutS (and MutL) actions, described in section 2.2, which further determine how effectively a particular mismatch is repaired.

1.2. ATP hydrolysis switches MutS to an ADP-bound state capable of mismatch recognition

Each MutS subunit contributes differentially to the DNA binding and ATPase activities of the dimer, and crosstalk between both subunits and both activities is essential for mismatch recognition and subsequent initiation of DNA repair. Resolving these intertwined mechanisms has involved more than a decade of genetic and biochemical studies, including bulk and single molecule kinetic measurements of multiple events in the reaction. A current model of the ATPase reaction mechanism is shown in Fig. 4. For reference, in a MutS homodimer, “A” is the subunit from which Phe-X-Glu makes specific contacts with the mismatched base (Msh6 in MutSα) and “B” is the other subunit (Msh2 in MutSα). Key data on which the model is founded are described here.

Fig. 4. Model mechanism of coupled MutS DNA binding and ATPase activities.

(a) When MutS is not at a mismatch, its ATPase reaction cycle involves rapid ATP binding to the high (A/Msh6) and low (B/Msh2) affinity sites (which closes domains I and IV, rendering MutS unable to load on DNA), followed by fast and slow hydrolysis at these sites, respectively. A/Msh6 loses ADP while B/Msh2_ADP species (in which domains I/IV can open) accumulate in steady state. (b) When MutS arrives at a mismatch, the weak initial encounter is followed by slow conformational changes in both MutS and DNA to form a tight mismatch recognition complex (in which domains I close in, Phe-X-Glu contacts the base and DNA is bent). (c) Mismatch-induced rapid ADP dissociation from B/Msh2 and ATP binding to both subunits is again followed by slow conformational changes in MutS and DNA (domains I and II move away from the mismatch, domains IV cross over and DNA unbends) that enable both interaction with MutL and formation of the MutS sliding clamp. MutL binds MutS before it begins sliding to form a stalled complex at the mismatch; alternately MutS may begin sliding first and a stalled MutS-MutL complex forms distant from the mismatch. The former enables the complex to mark the mismatch site and the latter could enable the search for a strand discrimination signal and nicking site.

First, nucleotide-binding measurements revealed that the ATPase active sites in domains V of the two MutS subunits bind nucleotides with different affinities [39-44]. For example, in nitrocellulose filter-binding assays Taq MutS binds one ATPγS (a non-hydrolysable ATP analog that mimics ATP binding) with high affinity (K_D ~1 μM) and the other with low affinity (K_D ~30 μM), and the same is true for ADP [40]. Experiments with E. coli, Taq, yeast and human MutS proteins showed that the dimer is also capable of binding one ATP (ATPγS or AMPPNP) and one ADP simultaneously [40-42,44,45]. This asymmetric nucleotide binding activity is recapitulated during ATP hydrolysis. From rapid quench experiments measuring pre-steady state ATP binding and hydrolysis (Taq, yeast MutS) and stopped flow experiments measuring phosphate release (Taq, E. coli, yeast, human MutS), we know that free MutS binds ATP rapidly (10⁶ M⁻¹ s⁻¹ at 40 °C for Taq MutS) and catalyzes a burst of hydrolysis—of only one ATP molecule per dimer (10 s⁻¹ at 40 °C for Taq MutS; Fig. 3a) [24,39,40,44,46]. Subsequent phosphate release is also fast, and one or more steps after that limit the steady state turnover rate (k_cat = 0.3 s⁻¹ at 40 °C for Taq MutS). The other MutS subunit also hydrolyzes ATP, but at a much slower rate (~30-fold) that is indistinguishable from k_cat, suggesting that this event may limit catalytic turnover [40]. Finally, ADP release from at least one of the MutS subunits is slow and likely helps limit the turnover rate as well [47,48]; Mg²⁺ influences ADP release, but the effects are not completely resolved since it appears to stabilize ADP in human MutSα and destabilize it in E. coli MutS [44,49].

Fig. 3. ATPase activity of MutS.

(a) Pre-steady state rapid quench experiments with Taq MutS reveal a burst of [³²P]-ADP formation at 10 s⁻¹ (40 °C) followed by a slow steady state rate at k_cat = 0.3 s⁻¹ (green trace). The burst amplitude indicates rapid hydrolysis of one ATP per MutS dimer. Pre-incubation of MutS with T-loop DNA results in complete inhibition of the burst and a residual slow k_cat = 0.3 s⁻¹ remains (red trace). (b) Stopped flow traces of MutS_IAANS mixed with ATP show that domains I of the free protein rearrange concurrently with ATP binding (signal increase at 10⁶ M⁻¹ s⁻¹) and switch right back with ATP hydrolysis (signal decrease at 10 s⁻¹), whereas one ADP-bound MutS_IAANS-DNA(+T) complex shows a lag after ATP binding to A subunit (likely reflecting ADP release from B subunit) followed by slow change in conformation (signal increase at 2 s⁻¹) to a stable ATP-bound complex.

ATP concentration dependence of the hydrolysis kinetics showed that the subunit that binds ATP with high affinity also hydrolyzes it rapidly (a finding confirmed by analysis of mixed wild type and mutant MutSα heterodimers, as described below) [40]. Interestingly, this fast asymmetric hydrolysis is maximal when both the tight and weak sites bind ATP, indicative of active communication between the two subunits during the reaction [39,40]. These functional data are in line with crystal structures of MutS proteins showing that the ATPase sites are located at the dimer interface and have a composite nature, with each subunit contributing residues to the site on the other subunit (Fig. 1) [50,51], and that ATP binding tightens the interface [26,43,52,53]. Bulk kinetic measurements of fluorophore-labeled Taq MutS domains I under the same reaction conditions reveal robust allosteric signaling between the ATPase and DNA binding sites despite their distant locations (~70 Å apart), with the protein flipping rapidly between two distinct conformations, one upon ATP binding to both subunits (10⁶ M⁻¹ s⁻¹ at 40 °C) and the other upon ATP hydrolysis at the tight site (10 s⁻¹ at 40 °C; Fig. 3b) [23]. Complementary sm-FRET data indicate that the DNA binding domains approach each other to form a closed clamp when bound to ATP and isomerize into a dynamic open/close state when bound to ADP [19]. This binary ATP binding- and hydrolysis-driven conformational switching mechanism enables differential MutS interactions with DNA that underlie its function in MMR (Fig. 4).

The kinetic data summarized above reveal the nature, stoichiometry, rates and overall order of events in the reaction; however, they do not identify the role of each subunit in the mechanism and how their activities are linked with each other. That information was obtained primarily from radiolabeled nucleotide cross-linking as well as ATPase kinetics experiments with wild type and mixed MutSα heterodimers containing Walker A/B ATPase site mutants of either Msh2 or Msh6. Cross-linking data with both yeast and human MutSα showed that Msh6, which makes specific contact with the mismatch (A subunit), binds ATP with high affinity and ADP with lower affinity, whereas Msh2 (B subunit) binds ATP with low affinity and ADP with higher affinity [41,44]. Other interesting results from these experiments include: (i) ATP binding to Msh6 weakens ADP at Msh2 (yeast MutSα) [41]; (ii) ATP binds tightly to Msh6 whether Msh2 is nucleotide free or ADP-bound (yeast and human MutSα) [41,44]; and (iii) Msh2 can bind ATP tightly if the Msh6 site is empty (yeast and human MutSα) [44,46]; these findings are incorporated into a model ATPase mechanism described further below (Fig. 4). Corresponding ATPase kinetics showed that a yeast MutSα mixed heterodimer in which Msh6 is wild type and Msh2 cannot bind nucleotides (Walker A mutant) undergoes a fast burst of ATP hydrolysis and then slow catalytic turnover like wild type protein (although the burst phase is less robust, consistent with the observation that ATP binding to both sites on Taq MutS stimulates maximal hydrolysis) [40,46]. A mixed heterodimer in which Msh6 is wild type and Msh2 can bind ATP but not hydrolyze it (Walker B mutant) also undergoes fast ATP hydrolysis, but then cannot go through catalytic turnover (indicating that a subsequent round of ATP hydrolysis by Msh6 is inhibited until Msh2 also hydrolyzes ATP). Mixed heterodimers in which Msh2 is wild type and Msh6 cannot bind or hydrolyze ATP (Walker A or B mutants) exhibit a slow, linear ATPase rate [46]. These data confirm that A/Msh6 is responsible for the burst of ATP hydrolysis by free MutS (and likely by MutS scanning matched DNA), whereas B/Msh2 hydrolyzes ATP at a much slower rate. Moreover, it is apparent that crosstalk between the two subunits directs the order and affinity of ATP binding, hydrolysis and product release at each active site.

Now we can construct a model detailing transient events in the ATPase mechanism when MutS is not bound to a mismatch (Fig. 4a). In the absence of DNA (and likely while scanning matched DNA), the A/Msh6 active site binds ATP rapidly and with high affinity and commits it to fast hydrolysis when the weak B/Msh2 site is also occupied with ATP (Msh2_ATP-Msh6_ATP → Msh2_ATP-Msh6_ADP). After ATP hydrolysis, Msh6 switches to a weak nucleotide binding site and can release the bound ADP. If Msh2 remains a weak site at this stage, both subunits could lose nucleotides and the reaction would effectively be aborted and start over. But we know that the reaction goes forward and Msh2 hydrolyzes ATP, so one possibility is that as Msh6 becomes a weak site after ATP hydrolysis, Msh2 switches to a tight nucleotide binding site that can hydrolyze its bound ATP and retain the ADP product (Msh2_ADP-Msh6). When Msh6 binds ATP again (Msh2_ADP-Msh6_ATP), Msh2 switches back to a weak site that can release the bound ADP (Msh2-Msh6_ATP) and continue catalytic turnover. Note that the model maintains asymmetry in the identity and affinity of nucleotides occupying the two subunits throughout the reaction cycle, simply flipping A/Msh6 and B/Msh2 between alternating states. This is a simple model to explain asymmetry of the ATPase cycle even in the early homodimeric form of MutS.

As noted earlier in section 1.1, in the absence of DNA, domains I and IV of nucleotide-free MutS are in a dynamic open/close equilibrium, which allows DNA access to the interior of the protein dimer. This is also true for MutS pre-incubated with ADP or mixed ADP+ATP [19]. Upon interaction with DNA, domains IV adopt a closed state that enables a topological link between MutS and DNA, whereas domains I apparently retain flexibility to interrogate base pairs for mismatches as the protein moves along the duplex (Taq MutS structures show a shift in B subunit domain I position, and more mobile B domain IV and A domain I in the ADP-bound versus nucleotide-free protein bound to DNA [50]). In contrast, if MutS is pre-incubated with non-hydrolyzable ATPγS before mixing with DNA, both domains I and IV are stabilized in a closed state that precludes DNA binding [19,22]. By favoring accumulation of ADP-bound rather than ATP-bound MutS, the ATPase mechanism generates species with the structure and dynamic properties necessary to search, find and respond to mismatches in DNA.

1.3. Mismatch recognition switches MutS to an ATP-bound state that licenses repair

When ADP-bound MutS encounters a mismatch, the ensuing conformational changes in both protein and DNA alter the MutS ATPase mechanism and, reciprocally, the interaction of MutS with DNA. As domains I from both subunits close in on DNA with A/Msh6 making specific contacts with the mismatched base, the DNA is bent to form a stable MutS-mismatch recognition complex [19,23]. Transient kinetic experiments monitoring ADP release, ATP binding, DNA binding/bending, and domain I dynamics reveal how the two MutS activities are coupled (Fig. 4b). After MutS recognizes a mismatch, ATP binds rapidly to the nucleotide-free subunit (A/Msh6) [40]. There is a brief lag during which ADP is released from B/Msh2 and replaced with ATP, followed by domain I rearrangement as seen on free MutS binding to ATP (Fig. 3b) [23]. Notably, ADP release occurs 10 – 30-fold faster from the MutS-mismatch complex than from free MutS [47,48]; thus, mismatched DNA works as a nucleotide exchange factor to trigger ADP release and favor ATP binding [44,49]. This ADP-ATP exchange is at the heart of the molecular switch model for MutS function in MMR [47]. The “switch” refers both to the mismatch-induced ADP-ATP exchange and the fact that MutS changes into a conformation wherein ATP hydrolysis by A/Msh6 is inhibited by ~30-fold compared with free MutS (from 10 s⁻¹ to 0.3 s⁻¹ at 40 °C for Taq MutS; Fig.3a) [39,40]. As a consequence, the MutS-mismatch complex is stabilized in an ATP-bound state (Msh2_ATP-Msh6_ATP) for a prolonged period (Fig. 3b) [41]. Contacts between the Phe-X-Glu motif and DNA are necessary for this transition, by selectively forming stable complexes on mismatched versus matched DNA and triggering intramolecular communication between the DNA binding and ATPase sites to block ATP hydrolysis [24,31,38,54-57].

So, what happens with the ATP-bound MutS-mismatch complex? Free MutS forms a closed clamp immediately on binding ATP and quickly reverts back to an open/close dynamic state on ATP hydrolysis. In contrast, mismatch-bound MutS isomerizes slowly into a closed clamp on binding ATP and remains in that state because hydrolysis is suppressed (Fig. 3). The transition involves multiple steps (coupled with ADP release and ATP binding), particularly disengagement of domains I from the mismatch and unbending of DNA, which in turn allow MutS to move away from the site (2 s⁻¹ and 0.3 s⁻¹, respectively at 40 °C for Taq MutS; Fig. 3, 4c) [23,58]. In addition to domains I moving away from the mismatch [58], new structural data indicate that domains IV cross each other (keeping the clamp closed) and the connector domains II move outward [53]; hence, ATP-bound MutS can slide on DNA [41,59] without rotating along the helical contour, as confirmed recently by sm-tracking experiments with Taq and yeast MutS proteins [18-21,60]. It is surprising that after the effort of finding and marking a rare mismatch on DNA (akin to finding a needle in a haystack), MutS adopts a mobile state that can diffuse freely away from the MMR target site. Recently published studies on MutS-MutL interactions following mismatch recognition provide new insights into the next stage of MMR, as described further in section 2.2 [53,58].

2. Ongoing research on how ATPase-coupled actions of MutS and MutL enable MMR and other processing of DNA

Our understanding of MMR is deepening on multiple fronts as new discoveries from molecular to organism levels resolve current models of the pathway and its impact on genome stability. The following are some active areas of research on the mechanisms of action of MutS and MutL proteins.

2.1. Allosteric communication within MutS

The above narrative highlights a critical aspect of MutS function—that its two activities, DNA binding/mismatch recognition and ATPase, are intimately linked such that the ligand at one active site dictates events at the other site. What makes this two-way communication especially fascinating is that it comprises multiple signals within one reaction cycle, transmitted in sub-millisecond time scales across the protein between active sites located 70 – 100 Å apart, and that the information transfer is asymmetric across the two subunits even in the bacterial MutS homodimer. The molecular mechanism underlying allostery in MutS remains unknown.

MutS crystal structures show that the DNA binding clamp domain IV is connected to the ATPase domain V by two long, linked α helices that form a “lever” arm [6]. This structural element as well as a “transmitter” region where MutS domains II, III and V intersect, are thought to play an important role in communication between the ATPase and DNA binding sites (Fig. 1). The structural data led to the hypothesis that nucleotide occupancy of the ATPase sites triggers conformational changes in the transmitter that are amplified along the lever to cause corresponding changes in the DNA binding sites [50]. Another possibility is that allosteric communication occurs via nucleotide- and DNA-induced changes in the conformational dynamics of the protein [61,62]. This view is supported by computational studies of E. coli, Taqand human MutS proteins. According to normal mode calculations there are strong motional correlations across the ATPase sites and the lever arms that may be the means of propagating information to and from the DNA binding sites [63]. All-atom molecular dynamics (MD) simulations of Taq MutS reveal clusters of correlated amino acid fluctuations between the two active sites. Moreover, there are distinct differences in the dynamics of nucleotide-free versus ATP-bound MutS-mismatch complexes, with the latter presenting more extensive amino acid clusters with a higher degree of correlated motion [64]. These findings suggest that nucleotide- and DNA-generated signals could be relayed via changes in a network of correlated motions instead of (or complementary to) a daisy chain of stable conformational changes between the two sites.

Key questions about this hypothesis include whether there are optimal signaling pathways in the dynamic framework and whether they are conserved, and how might they be identified and experimentally validated. One approach in the works is to define the amino acid network quantitatively based on non-covalent interaction energies between individual residues and use that data to identify centers of high connectivity, in terms of numbers and/or strength of interactions, across the protein and over time (persistence over nanosecond scale MD ensembles). Such information can yield potential paths of communication and identify residues with significant roles as “nodes” or “hubs” for transmitting information (Bhattacharyya M and Vishveshwara S, unpublished data) [65,66]. These could be experimentally tested by mutational analysis to assess their contribution to the mechanism coupling the DNA binding/mismatch recognition and ATPase activities of MutS. An intriguing possibility is that answers to mechanistic questions about allosterism in MutS might yield a functional role for the many conserved amino acids distributed across the protein (distant from both active sites), whose mutation is associated with Lynch syndrome, yet whose contribution to MutS structure-function remains obscure; reviewed in [67-71].

2.2. MutL ATPase activity and actions on DNA in MMR

Each subunit in the MutL dimer comprises two domains joined by an unstructured linker region; reviewed in [6,9,72]. Both domains contribute to the dimerization interface, the N-terminal domain contains a conserved active site from the GHKL family of ATPases and can bind MutS and DNA, and the C-terminal domain contains the endonuclease activity (when present in MutL homologs). ATP binding promotes dimerization of the N-terminal domain, DNA stimulates its weak ATPase activity, and mismatched DNA and ATP favor its interaction with MutS [73-76]. The C-terminal domain forms a stable dimer [77,78]; indeed that was thought to be its predominant function until the discovery of endonuclease activity in this domain, which is responsible for nicking the error-containing DNA strand in organisms that do not contain MutH [9,79,80]. The nicking activity is stimulated by MutS and by the processivity clamp (β/PCNA) and clamp loader proteins [79]; the latter proteins also induce a strand bias on MutL [81]. The linker domain plays a critical role in coupling the ATPase, DNA binding, MutS binding and endonuclease activities of MutL [77,82]. The linkers are flexible in the absence of ATP and enable MutL to adopt a range of ATP/ADP-induced asymmetric conformational states from fully extended (wherein nucleotide free N-terminal domains of the dimer are apart from each other and the C-terminal domains) to fully compact (wherein ATP-bound N-terminal domains are in contact with each other and the linker/C-terminal domains) [83]. This brief outline of MutL structure-function illustrates that the second protein in the MMR pathway likely has as complex an ATPase-driven mechanism of action as MutS. While there have been significant advances in structural analysis of MutL, mechanistic analysis is currently at a nascent stage. Questions that remain under investigation include: what are rate-determining steps in the MutL-catalyzed ATPase reaction; how are they coupled to conformational changes that enable/modulate its transient interactions with DNA, MutS and β/PCNA; what is the nature of the interactions between the different components; what is the temporal order of these events and how do they result in strand-specific nicking (either by MutL or by activation of MutH).

As noted earlier, the transition from mismatch recognition to DNA processing in MMR is also under active investigation, including recruitment and activation of MutL after MutS is stabilized at the mismatch in an ATP-bound state and able to form a sliding clamp on DNA (Fig. 4). Unlike MutS, which tracks the DNA helix while searching for mismatches, MutL appears capable of hopping between different sites on DNA [84], which could enable it to locate mismatch-bound MutS efficiently (MutL and MutS interactions with β/PCNA may also help by limiting the search area [85]). A recent sm-tracking study visualized yeast MutSα-MutLα complex diffusing away from the mismatch upon addition of ATP [21]. However, other studies indicate that MutS and MutL form long-lived multi-protein complexes at or near a mismatch [86-90]. Two new studies of E. coli and Taq MutS-MutL complexes argue for the latter model. A crystal structure of the E. coli mismatch-bound MutS homodimer crosslinked with the N-terminal domain of MutL in the presence of ATP reveals substantive rearrangement of MutS domains, the most striking being movement of the connector domain (II) out of the center of the dimer. This rearrangement enables the connector to interact with MutL and forms a new channel for DNA that allows MutS sliding [53]. Sm-FRET and bulk kinetic analysis of Taq proteins shows that MutL binds and traps MutS at the mismatch as it undergoes ATP-induced conformational transitions to form a sliding clamp but before it slides away from the site [58]; conversion of ATP-bound MutS to a clamp can take a few seconds, allowing for MutL interaction [19,23]. SPR and bulk kinetic measurements of the crosslinked E. coli MutS-MutL complex also indicate that MutL traps ATP-bound MutS on DNA as it transitions to the sliding clamp state [53]. Conceptually there are sound functional reasons for both stalled and mobile complexes—localized MutS-MutL complexes mark the mismatch, constrain strand nicking in its vicinity, and may help regulate strand excision, while mobile MutS can clear the site, allowing multiple rounds of MutS loading to amplify the recognition signal, and help MutL locate strand-discrimination signals and/or nicking sites distant from the mismatch. One intriguing hypothesis is that the ATPase reaction and differential nucleotide occupancy of the active sites on MutSMutL might dictate the choice of stationary or mobile behavior as needed. Transient kinetic studies of the coordinated workings of MutS and MutL are ongoing in order to resolve these fundamental mechanistic questions about the process of MMR and how it maintains genome stability.

2.3. MutS and MutL functions on DNA beyond MMR

MutS and MutL homologs are also involved in DNA metabolic processes other than canonical mismatch repair [91]. Some of these promote genome stability, such as suppression of homeologous recombination (by disrupting exchange between heteroduplex DNA substrates) [12,92], while others promote genome instability, such as expansion of triplet nucleotide repeat (TNR) sequences [93], processing of DNA oxidative damage lesions [94], and somatic hypermutation for antibody diversity [95]. For example, Msh2-Msh3 (MutSβ), which binds and initiates repair of insertion/deletion loops (post-replication or during recombination), also binds double strand/single strand junctions (for removal of 3’non-homologous tails in double strand break repair) [96], as well as secondary structures formed by repeat sequences (leading to TNR expansion and related neurological disorders) [97]. Recent crystal structures show that like MutSα, MutSβ also bends DNA containing loops to select them as targets for MMR [98]. There are however, differences in the protein-DNA interactions and in nucleotide binding compared with MutSα, and steady state measurements suggest differences in their ATPase activities and potentially differences within MutSβ depending on the type of DNA substrate as well [99,100]. Resolving variations in the DNA binding and ATPase mechanisms, and the core coupling mechanism between the two, will be important to understand how MMR proteins have adapted to perform multiple functions on DNA

Highlights.

• Kinetically coupled MutS DNA binding and ATPase activities enable mismatch repair.

• In mismatch search mode, the ATPase mechanism favors ADP-bound MutS.

• Upon mismatch recognition, changes in both MutS and DNA stabilize ATP-bound MutS.

• Subsequent ATP-induced changes in MutS and DNA allow MutL recruitment to the site.

• MutS-MutL complexes stalled at/near the mismatch initiate DNA excision and repair.