Autoinhibition of a clamp-loader ATPase revealed by deep mutagenesis and cryo-EM

Abstract

Clamp loaders are AAA+ ATPases that facilitate high-speed DNA replication. In eukaryotic and bacteriophage clamp loaders, ATP hydrolysis requires interactions between aspartate residues in one protomer, present in conserved “DEAD-box” motifs, and arginine residues in adjacent protomers. We show that functional defects resulting from a DEAD-box mutation in the T4 bacteriophage clamp loader can be compensated by widely distributed single mutations in the ATPase domain. Using cryo-EM, we discovered an unsuspected inactive conformation of the clamp loader in which DNA binding is blocked and the catalytic sites are disassembled. Mutations that restore function map to regions of conformational change upon activation, suggesting that these mutations may increase DNA affinity by altering the energetic balance between inactive and active states. Our results show that there are extensive opportunities for evolution to improve catalytic efficiency when an inactive intermediate is involved.

Clamp-loader complexes are oligomeric AAA+ ATPases that enable high-speed DNA replication by loading sliding clamps onto primer-template junctions for attachment to DNA polymerases ^1–4. The subunits of AAA+ ATPase complexes consist of AAA+ modules that bind ATP at active sites located between subunits in the oligomer (Figure 1A). Nature has elaborated on this conserved functional unit to generate a remarkable variety of molecular motors in which interfacial ATP binding and hydrolysis is coupled to a diversity of functional outputs. Members of different subfamilies of AAA+ ATPases are responsible for many processes in the cell, including DNA replication, the unwinding of DNA and RNA, the remodeling of protein complexes, and protein degradation ^5–11. A fascinating question is how such diversity evolves while preserving the function of the core catalytic engine that powers these motors.

Figure 1. Clamp loader structure and the T4 bacteriophage assay.

A, The T4 clamp loader is shown bound to DNA and sliding clamp (PDB ID: 3U60). The five subunits of the clamp-loader complex are denoted A, B, C, D, and E. In the T4 complex, subunits B through E are the ATPase subunits (gp44) and are identical to each other. Subunit A (gp62) is degenerated and does not have an ATPase domain. We refer to subunit A as the “clasp” subunit because it bridges the two open ends of the sliding clamp, using two extensions referred to as the A and A’ legs. The collar domains of each subunit form a pentameric assembly (transparent). Three catalytically competent ATP-binding sites are formed at the interfaces of subunits B-C, C-D, and D-E. Within each active site is a DEAD-box motif. Asp 107 and Glu 108 in the DEAD-box motif interact with the Mg²⁺ ion and nucleotide, while Asp 110 coordinates Arg 122 of a neighboring subunit. B, The platform uses a modified version of T4 bacteriophage (T4^del) in which the genes encoding the replication proteins of interest are deleted ²⁰. Replication of T4^del is enabled by provision of the missing replication proteins on helper plasmids through incorporation of a Cas12-CRISPR system the DNA encoding the missing replication proteins are recombined into the T4^del phage ^20,51. The plasmids that result in successful phage propagation can be counted by sequencing after cell lysis. C, Rescue mutations (fitness value >0.3) are tallied (y-axis) per residue position (x-axis). The maximum fitness value per position that contains recovery mutations is noted by color on the corresponding mutation count bar. A maximum fitness value of 0.3 is light yellow and positions containing a recovery mutation with a fitness value of 1.5 are dark red. Heatmap data shown in Figure 2 are extracted and paired with nine rescue hotspots.

AAA+ complexes form circular and spiral assemblies that are stabilized by interactions made by the ATP molecules bound at the interfaces between subunits. Nucleic acid or protein substrates are bound in a central channel formed by the spiral AAA+ assembly - in the case of clamp loaders, this substrate is primer-template DNA, the binding of which triggers ATP hydrolysis by the clamp loader and the release of the closed clamp around DNA ^1,12–15. Some of the interfacial interactions in the active state have obvious importance for catalysis – for example, a highly conserved “arginine finger” residue presented by one subunit coordinates the terminal phosphate of ATP bound to the other subunit ^16–18. Other interfacial interactions are conserved within families, but not across all AAA+ ATPases, which raises the question of how the interacting residues change during evolution without loss of function.

An important set of family-specific interfacial interactions are formed by “DEAD-box” motifs found in a subset of AAA+ ATPases. In these AAA+ ATPases, including eukaryotic and bacteriophage clamp loaders, there is a consensus amino-acid sequence motif, D-E-A-D (the “DEAD-box” ¹⁹), at the heart of the network of inter-subunit interactions. The first two residues of the DEAD-box motif (D,E) coordinate the magnesium ion in the ATPase active site and are important for catalysis ²⁰. The fourth residue in the DEAD box is typically an aspartic acid, and in the clamp loaders it forms an ion-pairing interaction with an arginine sidechain presented by the adjacent AAA+ subunit. This interfacial ion-pair, anchored by the DEAD-box motif, is positioned at a key juncture of the interactions that link DNA recognition to the formation of catalytically competent active sites (Figure 1A).

We were struck by the fact that the DEAD-box motif, which is so centrally important to the integrity of the active states of the eukaryotic and bacteriophage clamp loaders ^15,21,22, is incomplete or absent in other AAA+ ATPases, including the bacterial clamp loaders ¹². To better understand the role of the DEAD-box motif in clamp loader function we studied the T4 bacteriophage clamp loader, a well-established model system ^23,24 for which we have developed a high-throughput mutagenesis platform (see Figure 1A for a schematic diagram of the clamp loader and the notation used to identify subunits; the T4 clamp loader consists of one gp62 subunit, denoted A, and four gp44 ATPase subunits, denoted B,C,D, and E). We mapped the mutational responses of a T4 clamp-loader variant with a mild defect in DNA-dependent ATP hydrolysis. This defect was created by replacing the interfacial aspartate residue in the DEAD-box motif of the ATPase subunit (Asp 110 in the gp44 ATPase subunit of the T4 clamp loader) by cysteine (D110C), which attenuates the DNA-dependent ATPase rate of the clamp loader. The relatively mild phenotype of this mutation ²⁰ facilitates the identification of second-site mutations that rescue the fitness of the D110C mutant clamp loader. Such rescue mutations can potentially provide insight into the role of the DEAD-box motif in clamp loader mechanism.

We carried out saturation mutagenesis of the D110C variant of the T4 clamp loader and found that substitutions at many positions in the AAA+ module rescue fitness, as measured by phage propagation. Mutations at these sites do not affect the fitness of the wild-type clamp loader ²⁰. Interestingly, in the D110C background, the rescue function could be fulfilled by different amino acid substitutions at the same site. We interpret the presence of such rescue hotspots as evidence for an as-yet unidentified inactive state of the T4 clamp loader – in this interpretation, the mutations act by destabilizing this inactive state, and the precise identity of the substitution is less important. It is important to note that the rescue mutations could restore activity in other ways, such as by altering clamp affinity or ATPase activity, although it is more difficult to explain the diversity of activating substitutions in this way. We used cryo-EM to identify and structurally characterize such an inactive state, and find that it is a stable conformation of the clamp-loader complex in the absence of DNA. Our results suggest that the D110C rescue mutations increase the affinity of the clamp loader for DNA by altering the energetic balance between the active DNA-bound state and the inactive state to which DNA cannot bind.

Results

Identification of DEAD-box rescue mutations

We have recently developed a platform for high-throughput mutagenesis of T4 bacteriophage replication proteins in their proper functional context ²⁰ (Figure 1B; see methods). A library of helper plasmids is created, each bearing a different variant of the relevant replication proteins. The functional fitness is quantified by counting variants in the input helper plasmid library and also in the phage that are released after lysis of the E. coli cells, by deep sequencing. The fitness of a variant is defined as:

$$fitness={\log }_{10}{\left[\frac{librarycount}{phagecount}\right]}_{mutant}-{\log }_{10}{\left[\frac{librarycount}{phagecount}\right]}_{reference}$$

Here, “mutant” refers to the particular mutation for which the fitness is being measured, and “reference” is the allele (e.g. wild-type or D110C) that serves as the reference. In this logarithmic fitness scale, a value of +1 corresponds to a 10-fold faster rate of phage propagation than the reference variant, while a fitness score of −1 corresponds to a 10-fold slower rate of propagation.

We determined, using the phage propagation assay, that the D110C mutation leads to a ~6-fold reduction in phage propagation (Extended Data Figure 1A). Other mutations, such as D110S, reduce phage propagation by a factor of ~100 or more. Biochemical measurements with the purified wild-type T4 clamp loader showed that it has an affinity for primed DNA in the nanomolar range (K_D ~ 40 nM) in the presence of the sliding clamp (Extended Data Figure 2A). Under identical conditions, the D110C clamp loader mutant showed no detectable DNA binding. The D110C clamp loader does, however, retain a weak affinity for DNA, since the addition of DNA stimulates the rate of ATP hydrolysis (Extended Data Figure 2B). We determined the crystal structure of the D110C clamp-loader variant bound to primed DNA and the sliding clamp, relying on high protein and DNA concentration to drive complex formation (Extended Data Figures 1B, 3 and Table 1). The overall structure of the D110C complex is very similar to that of the wild-type complex ²², and the structure suggests that Cys 110 is deprotonated and negatively charged at those sites, thereby mimicking the aspartate residue and explaining the mild phenotype of the D110C mutant. There are two complexes in the asymmetric unit of the crystal. At four of the six interfacial ATPase sites in the two complexes the termini of the cysteine sidechains are within ~3 Å of the termini of the sidechains of Arg 122, from adjacent subunits, consistent with ion pair formation.

Table 1.

X-Ray data collection, refinement and validation statistics

	D110C T4 Bacteriophage Clamp Loader with Sliding Clamp and DNA (8UK9)	WT T4 Bacteriophage Clamp Loader with Sliding Clamp and DNA (8UH7)
Data collection Space group	P 21 21 21	P 1 2 1
Cell dimensions
a, b, c (Å)	95.243 231.991 264.511	93.35 118.32 132.89
a, b, g (°)	90, 90, 90	90.00 102.05 90.00
Resolution (Å)	46.87 – 3.1 (3.211 – 3.1)	49.65 – 2.63 (2.92 – 2.63)
R merge	0.05745 (1.571)	0.325 (2.096)
I / sI	5.80 (0.46)	6.9 (1.1)
Completeness (%)	(ellipsoidal) 95.1 (64.0)	(ellipsoidal) 92.9 (68.3)
Redundancy	2.0 (2.0)	6.3 (6.9)
Refinement
Resolution (Å)	46.87 – 3.1 (3.211 – 3.1)	49.65 – 2.64 (2.73 – 2.64)
No. reflections	76439 (470)	55477 (250)
Rwork / Rfree	0.2525/0.2643	0.1948/0.2466
No. atoms
Protein	35422	17800
Ligand/ion	248	124
Water	0	34
B-factors
Protein	131.1	46.49
Ligand/ion	94.67	26.04
Water	0	26.00
R.m.s. deviations
Bond lengths (Å)	0.003	0.003
Bond angles (°)	0.66	0.54

^*Highest resolution shell shown in parentheses

^*One crystal was used for each structure

For saturation mutagenesis of the D110C clamp loader we used a library for the AAA+ module of the T4 clamp loader that had been developed previously ²⁰, and we fixed the D110C point mutation into this library (see Methods). The resulting double-mutant library (D110C plus additional single-site mutations) was used for phage propagation. The D110C phage DNA sequence without a second mutation is used as the reference allele when calculating fitness (see Figure 1B). In this logarithmic fitness scale, the D110C clamp loader itself has a fitness value of zero. Extended Data Figure 4 shows the reproducibility of these fitness scores.

The D110C variant of the clamp loader is markedly more sensitive to mutations than is the wild-type complex. Strikingly, the saturation-mutagenesis data identify mutations that restore the fitness of the D110C variant (regions shaded red in Figures 1C and 2). For this analysis, we consider mutations that yield a fitness value of >0.3, corresponding to a 2-fold increase of phage propagation over the variant with the D110C clamp loader, as gain of fitness variants (“rescue mutations”). Using this criterion, we find that ~80 positions in the AAA+ module have at least one rescue mutation (see red boxes and bar plot in Figure 1C). Of these positions, 10 have at least one mutation with a fitness score above 1.0, corresponding to wild-type levels of activity, or essentially complete repair of the replication defect. Thus, the clamp loader has a distributed capacity to compensate for the defect in the DEAD-box motif through point mutations.

Figure 2. Deep mutagenesis of the D110C mutant clamp loader.

The top panel in each block shows the sequence conservation, represented as a sequence logo ⁵², in the AAA+ modules of >1000 clamp-loader sequences from phage genomic data ²⁰. The first heat map in each block shows the mutational sensitivity of the wild-type T4 clamp loader, as determined previously ²⁰. The second heatmap in each block shows the mutational sensitivity of the D110C variant, using D110C as the reference allele. In these heatmaps, blue pixels correspond to variants that are less fit than the reference (white), with the minimum fitness (darkest blue) being 1000-fold worse than the reference. Red pixels are variants that are more fit than the reference, with the maximum fitness of 1000-fold better than the reference. Nine rescue hotspots are boxed across the two panels, these positions are discussed in the main text and in Figure 1C.

Purification and biochemical measurements of clamp-loader complexes with 5 rescue mutations show that these mutations increase DNA affinity or the ATP hydrolysis rate, or both, relative to the D110C clamp loader (Extended Data Figure 5). For example, the P50K variant of the D110C clamp loader has DNA-binding affinity and DNA-stimulated ATPase rates that are both comparable to that of the wild-type clamp loader. The G143R variant of the D110C clamp loader also has DNA affinity comparable to wild-type clamp loader, but in this case the ATPase rate is not completely recovered. Other mutations have weaker effects. The D98W mutation has a small increase in DNA affinity, but no improvement in ATP hydrolysis.

The locations of the rescue mutations do not readily explain why these mutations improve fitness. The rescue mutations do not occur at sites of obvious functional importance, since they do not overlap with sites of sequence conservation (Figure 2 and Extended Data Figure 6). A clue to the mechanism of rescue comes from observing an unexpected pattern in the deep mutagenesis data. At many sites where rescue mutations occur, more than one kind of mutation results in improved fitness. We refer to such sites as “rescue hotspots”. Nine of the most prominent rescue hotspots are highlighted in Figure 1C and Figure 2. Some of these rescue hotspots are sites where 10 or more different substitutions increase the fitness of the D110C variant. The average value of the maximum fitness score for substitutions at these sites is ~1.1, which corresponds to wild-type levels of fitness. The ability of very different substitutions at a particular site to rescue fitness suggests to us that these mutations may increase the activity of the D110C clamp loader by non-specific effects, which we demonstrate to be the destabilization a less active autoinhibited state, but could have been any number of things, involving changes in clamp binding, DNA binding or ATPase activity.

An inactive conformation of the clamp loader without DNA

We used cryo-EM to determine the structure of the wild-type T4 clamp-loader complex in the presence and absence of primed DNA and the sliding clamp (Table 2 and Extended Data Figure 7). In order to minimize perturbation of the equilibrium between different conformational states of the clamp loader complex we did not use cross-linking reagents in any of our cryo-EM samples. We first determined the cryo-EM structure of the clamp-loader:clamp complex in the presence of primed DNA and ADP•BeF₃. The resulting structure resembles closely the crystal structure of the same complex determined previously ²² (Figure 3A). If the two structures are aligned using the AAA+ module of subunit C, the root-mean-square deviation in Cα positions calculated over the other AAA+ modules (residues 1–230 of chains B, D, and E) is ~ 2.4 Å. The ATP analog, ADP•BeF₃, is visualized at the ATPase sites of the B, C, and D subunits, with the BeF₃ moiety at each site coordinated by the sidechain of the arginine-finger residue (Arg 151) from neighboring subunits. Cryo-EM density corresponding to an ADP molecule is seen at the ATPase site of subunit E, consistent with the crystal structure ²². We refer to this structure as the “active complex”.

Table 2.

Cryo-EM data collection, refinement and validation statistics

	#1 DNA-bound clamp-loader:clamp (EMDB-42399) (PDB 8UNF)	#2 DNA-free clamploader:clamp (EMDB-42402) (PDB 8UNH)
Data collection and processing
Magnification	81,000	81,000
Voltage (kV)	300	300
Electron exposure (e−/Å2)	50	50
Defocus range (μm)	−0.8 to −2.2	−0.8 to −2.2
Pixel size (Å)	0.524	0.524
Symmetry imposed	No	No
Initial particle images (no.)	4,206,415	17,937,844
Final particle images (no.)	204,448	455,041
Map resolution (Å)	3.15	3.21
FSC threshold	0.143	0.143
Refinement
Initial model used (PDB code)	3U60	3U60, 3U61
Map sharpening B factor (Å2)	-133	-157
Model composition
Non-hydrogen atoms	17810	13271
Protein residues	2147	1683
Ligands	8	4
B factors (Å2)
Protein	64.78	50.59
Ligand	47.45	58.12
R.m.s. deviations
Bond lengths (Å)	0.007	0.004
Bond angles (°)	0.662	0.480
Validation
MolProbity score	1.92	2.52
Clashscore	11.50	46.50
Rotamer outliers(%)	0.44	0.49
Cb outliers(%)	0.05	0.00
Ramachandran plot
Favored (%)	95.12	94.48
Allowed (%)	4.69	5.28
Disallowed (%)	0.19	0.24

Figure 3. Cryo-EM structures of DNA-bound and DNA-free state of the clamp-loader complex.

A, The structure of the T4 clamp loader is shown, including subunits A (gp62, green), B (gp44, cyan), C (gp44, magenta), D (gp44, yellow), and E (gp44, salmon). The open T4 sliding clamp (gp45) is colored grey and primer-template DNA is colored orange. The clasp subunit (gp62, green) bridges the gap in the open sliding clamp with its A leg on the lower part of the clamp and its A′ leg at the top. The primer-template DNA (orange) is bound in the central chamber of the clamp-loader:clamp complex, with the single-stranded template overhang threaded through the gap between the A and A′ legs. B, Surface representation of the cryo-EM structure of the clamp-loader:clamp complex in the absence of DNA (same color scheme as in panel A). The T4 sliding clamp (gp45, grey) is closed. The AAA+ modules of subunits C and D are disordered in this structure (right, magenta and yellow). The distance between the collar domains of subunits C and D and the sliding clamp reduces from ~50 Å in the DNA-bound complex (panel A, right) to ~25 Å in the DNA-free complex (panel B, right), as indicated by double-headed black arrows. C, Schematic diagrams (colored as in panel A) of the T4 clamp-loader:clamp complex in the DNA-free (left) and DNA-bound (right) states. For clarity, the collar domains are not colored. ATPγS molecule (red) are bound to subunits B and E of the clamp-loader in the DNA-free state. In the DNA-bound state of the clamp loader, ADP•BeF₃ is bound to subunits B, C, and D, while ADP (purple) is bound to subunit E.

We do not observe the formation of this DNA-bound active complex of the wild-type clamp loader if the D110C variant of the clamp loaders is used during sample preparation. We also do not observe the formation of the active complex if the non-hydrolysable ATP analog ATPγS is used instead of ADP•BeF₃. Using the fluorescence anisotropy assay described above, we confirmed that DNA binding to the wild-type T4 clamp-loader complex can be readily detected in the presence of ADP•BeF₃ (K_D ~ 40 nM) or ADP•AlF₃ (K_D ~ 100 nM), but not with ATPγS, ADP alone, or in the absence of nucleotides (Figure 4A). The inability of the T4 clamp loader to bind DNA in the presence of ATPγS ^25,26, along with the results of saturation mutagenesis on the D110C variant that showed numerous gain of function sites, motivated us to pursue structural analysis of a DNA-free form of the clamp loader.

Figure 4. The clamp-loader adopts a stable inactive conformation.

A, The fluorescence anisotropy assay was used to measure DNA-binding to the clamp-loader:clamp complex in the presence of 1 mM ADP, ATPγS, ADP•BeF₃, ADP•AlF₃, and no nucleotide, respectively. In this assay, DNA-binding to the clamp-loader:clamp complex is detectable only in the presence of ADP•BeF₃ or ADP•AlF₃. Data are presented as mean values +/− standard deviations (SD), n=3 biologically independent experiments. B, Cryo-EM maps of the wild-type clamp-loader:clamp:DNA complex in the presence of ADP•BeF₃ (left) and the wild-type clamp-loader:clamp complex in the presence of ATPγS (right) are colored as in Figure 4. C, Cryo-EM maps of the clamp-loader(D110C):clamp complex in the presence of ADP•BeF₃ (left) and the clamp-loader(wild-type):clamp complex in the presence of ATPγS (middle) show that both of these two clamp-loader complexes adopt the DNA-free conformation, even when primed DNA is used during sample preparation and is present on the EM grids. Right, the clamp loader alone in the presence of ATPγS adopts the same DNA-free conformation. D, Cryo-EM maps of the wild-type clamp-loader:clamp complex in the presence of ADP (left), ADP•BeF₃ (middle), and ADP•AlF₃ (right) show that these complexes all adopt the distorted DNA-free conformation described in panel A, independent of the nature of the nucleotide bound to the ATPase sites.

We used cryo-EM to analyze the D110C variant of the clamp-loader:clamp complex in the presence of ADP•BeF₃, and the wild-type clamp-loader:clamp complex in the presence of ATPγS. In both of these conditions the clamp loader adopts a novel DNA-free conformation, even when primed DNA is used during sample preparation and is present on the grids (Figure 4C). The cryo-EM structure of the wild-type T4 clamp loader bound to ATPγS and the sliding clamp but without DNA bound was determined at a resolution of 3.2 Å (Figure 3B). The collar domains form a pentameric assembly, as in other clamp-loader structures. The collar assembly packs more closely against the AAA+ modules of subunits B and E than is the case for the DNA-bound complex. Only the AAA+ modules of subunits B and E can be resolved in this cryo-EM structure. Additional diffuse density can be identified in the vicinity of the presumed locations of the AAA+ modules of subunits C and D, at lower contour levels of the cryo-EM map. We used different data processing strategies, including partial signal subtraction, masked 3D classification, and masked 3D refinement, but we were unable to resolve interpretable density for the AAA+ modules of subunits C and D, which implies they are mobile and are no longer held in a fixed position by their neighboring subunits. We refer to this complex as the “inactive complex”, since the interfacial catalytic sites are not assembled.

We also used cryo-EM to analyze the T4 clamp-loader:clamp complex without DNA, in the presence of ADP, ADP•BeF3, and ADP•AlFx, respectively. The clamp-loader complex adopts the same distorted conformation seen in the ATPγS-bound complex under all of these conditions (Figure 4D), suggesting that this DNA-free conformation of the T4 clamp loader is stable and is insensitive to the nature of the nucleotide bound to the ATPase domain. We also analyzed the clamp loader in the absence of sliding clamp and DNA by cryo-EM. The clamp loader alone adopts the same DNA-free inactive conformation (Figure 4C).

Inactive conformation of the clamp loader blocks DNA binding

The structures of the DNA-bound and DNA-free states of the clamp-loader:clamp complex provide an explanation for why DNA binding is required to form the catalytically-active complex. In a DNA-bound clamp-loader, the single-stranded portion of the primed DNA binds to the collar domain of the clasp subunit (subunit A), between the two linkers connecting the collar domain to the A and the A’ legs. In the absence of DNA, these linkers switch conformation from extended loops in the active conformation to α helices in the inactive conformation (Figure 5A). This conformational switching leads to tighter interaction between the collar domain of the clasp subunit and the AAA+ module of the B subunit in the inactive conformation. This reduces the distance between the collar domains of subunits C and D and the sliding clamp, from ~50 Å in the DNA-bound complex to ~25 Å in the DNA-free state (Figure 3A and 4B). The compressed space between the collar domain and the sliding clamp can no longer accommodate DNA or the AAA+ modules of the C and D subunits, which are displaced from the main body of the assembly. As a result, DNA cannot bind, and none of the catalytic ATPase sites between the AAA+ modules are assembled.

Figure 5. Conformational differences between the active and inactive states.

A, Surface representations of the cryo-EM structures of the DNA-bound (left) and the DNA-free (right) states of the clamp-loader complex are superimposed with cartoon representation of the clasp subunit (gp62, green). The linker regions (red) connecting the collar domain to the A and the A’ legs in the clasp subunit switch conformation from extended loop (left, DNA-bound state) to α helices (right, DNA-free state). This conformational switch allows the collar domains to pack more closely to the AAA+ module of subunit B, which reduces the distance between the collar domains and the sliding clamp (see also Figure 4A and 4B). B, AAA+ modules of subunits B (purple) and E (salmon) of the inactive DNA-free state of the clamp-loader complex are superimposed with subunit D (yellow) of the active DNA-bound state of the clamp-loader complex. These structures are aligned using domain 1 (residue 1–120) of the AAA+ module. Black arrows highlight the inwards rotation (~ 10°) of domain 2 (residue 160–230) towards domain 1, observed in subunits B and E of the DNA-free complex when compared to subunit D of the DNA-bound complex. C, Schematic diagrams of the active (left, yellow) and inactive (right, pink) conformations of the AAA+ modules. The active conformation of the AAA+ module is adopted by subunits B, C, and D of the DNA-bound state of the clamp-loader complex. The inactive confirmation of the AAA+ module is adopted by subunits B and E of the DNA-free state and subunit E of the DNA-bound state of the clamp-loader complexes. In the active conformation, key residues in the ATPase sites, Asn 139 (sensor 1) and Arg 205 (sensor 2), are positioned near ADP•BeF₃ in a manner consistent with the formation of a catalytically competent active site. In the inactive conformation, these two residues are pulled away from the position of terminal phosphate of ATPγS or ADP.

Rescue hotspots located near sites of conformational change

Comparison of the cryo-EM structures shows that rescue hotspots for D110C are in regions of the AAA+ module that undergo structural change as the T4 clamp loader transitions between the inactive DNA-free and active DNA-bound states (Figures 6A and 6B). For analysis of these structural changes, the E subunit of the DNA-free state and the D subunit of the DNA-bound state of the cryoEM structures are aligned using the central coupler segment (residues 80–95, 114–127 and 146–152).

Figure 6. D110C interrupts the energetic balance between DNA-bound and DNA-unbound states.

A, The positions of the nine rescue hotspots (red spheres) are mapped onto the active clamp loader cryo-EM structure bound to DNA and clamp. The clasp subunit is colored green, while subunits B-E are colored blue, purple, yellow and pink. Two categories of rescue sites are noted. The first category contains sites that sit at the interfaces between subunits, deemed ‘interfacial’ sites (grey areas). The other sites not highlighted in grey are sites internal to the AAA+ module and generally located near the active sites of the ATPase subunits. B, Subunit C of the clamp loader complex in panel A (purple) is aligned with the E subunit of the DNA-unbound clamp loader structure (wheat) along the central coupler regions (residues 80–95, 114–127, and 146–152). The alignment is rotated to expose the full interfacial region of the ATPase module (highlighted in grey). The other three sites mapped onto the AAA+ module (Pro 50, Gly 143 and Thr 156) are in regions populated with loops experiencing internal conformational change between DNA-unbound and DNA-bound structures. C, Gly 143 (red sphere) within subunit B (cyan) is located near the clasp subunit (green). The inactive, DNA-free structure is shown in the top panel and features close contacts between Gly 143 and clasp subunit residue Ala 57. Aligned on the central coupler region, the DNA-bound cryoEM structure (bottom panel) features displacement of subunit A such that no contacts with Gly 143 are made with the clasp subunit. D, A focused view of Gly 143 is taken from panel B. Gly 143 is local to DNA-binding residue Arg 111. When DNA is bound, Arg 111 is pulled away from Gly 143. E, The conceptual free energy landscape of the clamp-loading trajectory is shown for wild-type, D110C mutant, and rescued D110C clamp loader. The D110C mutation destabilizes the active conformation of the wild-type clamp loader when bound to DNA (orange line). The addition of rescue mutations destabilizes the inactive, DNA-free state of the clamp loader when in context of the D110C mutation (dashed line).

There are two key differences between the DNA-bound and DNA-free states of the T4 clamp loader. The first set of conformational differences concerns the overall organization of the active and inactive states, described above. The second set involves changes in the internal structure of the AAA+ modules, depending on whether or not the module is part of a properly assembled interfacial catalytic center. The internal conformations of the AAA+ modules of subunits B, C, and D of the DNA-bound complex resemble each other closely, and we refer to this as the active conformation of the individual AAA+ modules (Figure 5B and 5C). In this active conformation, Asn 139 (Sensor 1)²⁷, a residue that positions a water molecule for attack on the terminal phosphate of ATP, is located ~3 Å away from the BeF₃ moiety. Arg 205 (Sensor 2)²⁷ interacts with the β phosphate of ADP and with the BeF₃ moiety.

Subunit E of the DNA-bound complex, as well as subunits B and E of the DNA-free complex are not at properly configured interfaces, and are in a different conformation, referred to as the inactive conformation of the AAA+ module (Figure 5B and 5C). Based on the local resolution of the cryoEM density map of the DNA-free state, subunit E is better resolved than subunit B, likely due to reduced flexibility within the complex. Because of its higher quality in the cryoEM density map, we chose to use subunit E (DNA-free, inactive) for our detailed structural comparison to subunit D (DNA-bound, active). The difference between the active and inactive conformations of the AAA+ module is mainly due to a change in the angle between Domain 1 (residues 1–120) and Domain 2 (residues 160–230) of the AAA+ module. In switching from the active to the inactive conformation, Domain 2 rotates by ~10° towards Domain 1. This leads to a more closed conformation for the AAA+ module in the inactive conformation. The two ATP sensor residues, Asn 139 and Arg 205, are pulled away from their positions in the active conformation.

Gly 143 is a rescue hotspot in the D110C clamp loader, and was also identified earlier as a site where substitutions can compensate for the loss of integrity of the central coupler due to mutation of Gln 118, a structural lynchpin ²⁰. The cryo-EM structure of the DNA-free inactive state shows that the Cα atom of Gly 143 in subunit B is in van der Waals contact with the sidechain of Ala 56 and the sidechain of Gln 57 in the A subunit (clasp), with interatomic distances of ~4 Å in both cases (Figure 6C). Mutation of Gly 143 is likely to result in steric clashes that increase the energy of the inactive state. When DNA is bound to the clamp loader, the clasp subunit is pulled away from this region, and Gly 143 in the AAA+ modules point towards DNA, but with sufficient additional space to accommodate substitutions.

Substitution of Gly 143 by arginine is particularly activating, consistent with the location of this residue next to DNA in the active complex. Gly 143 is located close to Asp 110 (the interfacial aspartate residue in the DEAD-box motif) and to Arg 111, a residue that is critical for DNA binding. The Cα atom of Gly 143 makes van der Waals contact with the backbone of Asp 110 and Arg 111 in both the DNA-bound and DNA-free states. In the DNA-free state there is change in the conformation of the backbone such that Arg 111 no longer points towards DNA (Figure 6D). This is coupled to a change in the orientation of the Asp 110 sidechain and a correlated change in the catalytically important Sensor 1 residue, Asn 139. Given its close apposition to these residues and to DNA, mutation of Gly 143 could alter the balance between the active and inactive states.

Several of the hotspot residues are in regions where structural changes are caused by the rotation of Domain 2 of the AAA+ module with respect to Domain 1. This rotation is correlated with changes in the backbone within several segments of the protein (the N-terminal segment, residues 1–27, which leads into helix α1, residues 28–39; the phosphate-binding P loop, residues 48–56; and the loop connecting domain 1 to domain 2, residues 157–163). For example, Pro 50 is a rescue hotspot that is part of the P loop. In the DNA-bound active structure, the carbonyl group of the peptide linkage between Ser 49 and Pro 50 forms a hydrogen bond with the Sensor 1 sidechain (Asn 139), and this is likely to be important for holding Asn 139 in a catalytically competent conformation (Extended Data Figure 8A). This hydrogen bond is part of a network of interactions made by Asp 110 in the DEAD-box motif. In the inactive state, Asn 139 is no longer anchored by this backbone interaction. The location of Pro 50 at the center of these interactions is consistent with mutations at this site having an effect on the energetic balance between the inactive and active states.

Helix α1 contains two rescue hotspots (Thr 32 and Ser 35). In the inactive DNA-free structure, helix α1 does not form interfacial interactions (Figure 6A). However, when Domain 2 rotates upon DNA binding, one face of helix α1 is positioned close to helices α9 and α10 (residues 212:227) at the B-C and C-D interfaces of the DNA-bound complex, and Thr 32 and Ser 35 now face the interfacial region. The sidechain of Phe 28 in helix α1 packs close to the sidechain of Leu 227 in α10 in the DNA-bound state (Extended Data Figure 8B). There is a change in the orientation of helix α1 in the structure of the DNA-free form, due to the rotation of Domain 2. If this conformation of helix α1 were to be maintained when the complex switches to the DNA-bound form, Phe 28 would be pulled away from interaction with Leu 227. Both residues are highly sensitive to mutation in the D110C context, suggesting that the energetics of this interface is important in DNA binding (Figure 6E).

Discussion

In the present work, we studied the mutational sensitivity of a T4 clamp-loader variant with a mild defect in DNA-dependent ATP hydrolysis. This defect was created by replacing the interfacial aspartate residue in the DEAD-box motif of the gp44 ATPase subunit (Asp 110) by cysteine (D110C). We carried out saturation mutagenesis of the D110C variant of the T4 clamp loader and identified rescue hotspots in the AAA+ module, and observed that many different substitutions lead to the recovery of fitness. The structure of the active DNA-bound form of the clamp-loader:clamp complex failed to provide explanations for why these mutations lead to increased fitness - most of these are in obscure locations that do not suggest a direct functional relationship.

We hypothesized that that there is an as-yet unidentified inactive state of the T4 clamp loader, and that the rescue mutations alter the energetic balance between inactive and active states so as to compensate for the weakening of the interfacial interaction made by Asp 110 in the DEAD-box motif. This explanation – involving an inactive state – is more compatible with the observed wide spectrum of mutations at many rescue positions. In contrast, rescue mutations that work by directly increasing activity would be expected to satisfy more rigid structural demands so as to maintain an active catalytic site, and such mutations would presumably be restricted to only one or two residues at only a few, perhaps conserved, positions.

Cryo-EM analysis of the D110C clamp loader confirmed this proposal by revealing a novel DNA-free and autoinhibited state of the T4 clamp loader, in which DNA binding is blocked. This novel autoinhibited state is also seen in cryo-EM reconstructions of the wild-type clamp loader alone, the wild-type clamp-loader:clamp complex without DNA, or in the presence of ATPγS, an ATP analog. In contrast, the cryo-EM reconstruction of the wild-type T4 clamp-loader:clamp:DNA complex in the presence of the ATP mimic ADP•BeF₃ closely resembles the fully active state, as defined previously by X-ray crystallography ²². The cryo-EM reconstructions of DNA-bound and DNA-free states of the clamp-loader complex reveal that there are two key conformational changes between the two states. The first concerns the proper assembly of the overall complex. The second concerns internal structural rearrangements within individual AAA+ modules. If there is no neighboring subunit to complete the assembly of a catalytically-competent ATPase site, as is the case in subunit E of the DNA-bound state and the subunits of the DNA-free state, the internal structure of the AAA+ module adopts a more closed inactive conformation. D110C rescue mutations occur in regions that undergo conformational changes in transitioning from the inactive DNA-free state to the active DNA-bound state.

Comparison with the mechanism of DEAD-box RNA helicases suggests that clamp loaders have preserved a feature of the RNA helicases, which is to use the DEAD-box motif to couple RNA or DNA binding to ATP hydrolysis. DEAD-box RNA helicases are monomeric proteins that are analogous to dimer units of oligomeric ATPases – each subunit of a monomeric RNA helicase consists of two RecA-type folds, connected by a flexible linker ^28–31. One of these domains maintains the ability to bind ATP, while the other does not. The ATP-binding domain contains the DEAD-box motif. In the presence of RNA, the second aspartic acid of the DEAD-box motif, corresponding to Asp 110 in the T4 clamp loader, coordinates an arginine sidechain presented by the other domain, and this interaction stabilizes the formation of a competent catalytic site ²⁸. When ATP and RNA are absent, the ion pair between the aspartic acid and the arginine is broken, and the catalytic center is dismantled ^28,32,33.

The arginine sidechain that makes the interfacial contact with the final aspartate in the DEAD box of RNA helicases is structurally analogous to the arginine residue (Arg 122) that interacts with the corresponding interfacial aspartate (Asp 110) in the T4 clamp loader. Our results suggest that DNA-dependent interfacial interactions made by the DEAD-box motif in the clamp loader provide part of the energy to transition to the fully assembled and catalytically competent state. ATP is integral to the stability of this active state, providing a mechanism for triggering the release of the sliding clamp on DNA once ATP hydrolysis occurs. Release of DNA then causes disassembly of the active sites, which minimizes cycles of futile ATP hydrolysis.

The limited functional constraints on the inactive states of clamp loaders means that the structures of the inactive states of these machines can diverge dramatically over evolutionary time. This principle is well appreciated in the study of signaling proteins ^34,35. Over time this divergence can lead to changes in the sequences that are part of the catalytic center itself, as the nature of the couplings that are required to transition out of the inactive states are altered – the bacterial clamp loaders have lost the DEAD-box motif altogether, although analysis of structures show that they have maintained the same kind of interfacial interactions in this region, but with very different residues. Alterations in the inactive states allows natural selection to tune the responses of these molecular machines, and contributes to the remarkable diversity of ATP-driven machines that are found in nature.

Methods

Protein expression and purification

The clamp loader complex (gp44 and gp62 with a C-terminal Hisx6 tag) was prepared as described previously ²⁰, with one modification. The cells expressing the D110C clamp-loader variant and double mutants containing D110C were lysed with a French press (Avestin EmulsiFlex C50 cell homogenizer), using 1500 psi pressure at 4°C for 1.5 to 2 minutes, instead of sonication. Subsequent purification via affinity chromatography proceeded as described ²⁰. The sliding clamp (gp45) was purified as described ²⁰.

Preparation of primer-template DNA

The DNA oligonucleotides were obtained from Integrated DNA Technologies (IDT) and had the following sequences:

Template strand (30bp): 5’-TTTTTTTTTTTTATGTACTCGTAGTGTCTG-3’

Primer strand (20bp): 5’-GCAGACACTACGAGTACATA-3’.

DNA oligonucleotides were ordered as lyophilized samples and diluted with 20 mM Tris (pH 7.5) and 1 mM MgCl₂ to a final concentration of 200 μM. For fluorescence anisotropy measurements, the template strand had a TAMRA fluorophore attached at the 5’ end. Primer and template DNA were mixed in an equimolar ratio, brought to 95 °C, and cooled to room temperature over a period of 2 hours.

Fluorescence anisotropy DNA binding assay

50 nM primer template DNA labeled at the 5’ end of the template with a TAMRA fluorophore (purchased from IDT) was incubated in buffer A (100 mM HEPES pH 7.5, 150 mM NaCl, 10% (v/v) glycerol), with 1 mM ADP•BeF_x (1mM ADP dissolved in 20 mM HEPES pH 7.5, 1mM BeSO₄, 5mM NaF), 5 mM MgCl₂ and 7 μM sliding clamp (gp45). When preparing ADP•AlF_x, 1 mM BeSO₄ was replaced with 1 mM AlCl₃. ATPγS was purchased from Jena Biosciences. Clamp loader was serially diluted in buffer A and titrated in the above reaction mixture ranging from 0 – 5 μM for a total reaction volume of 100 μL. After a three-hour incubation at room temperature (Extended Data Figure 9) ^36,37, fluorescence polarization was measured by a Biotek Synergy H1 plate reader. Apparent dissociation constants (K_d) were estimated by nonlinear curve fitting to the Hill equation using GraphPad Prism 9.

ATP hydrolysis assay

The ATP hydrolysis rates of the various gp44/62 clamp loader complexes were measured using a coupled enzyme assay in which the decrease in absorbance at 340 nm is monitored as NADH is converted to NAD+ during the regeneration of the ADP product of the ATP hydrolysis reaction ³⁸. Each well’s regeneration system contained 3 mM phosphoenolpyruvate, 0.5 mM NADH, and 1079 units per ml of pyruvate kinase and lactate dehydrogenase (Sigma Aldrich). The reaction buffer contained 25 mM HEPES (pH 7.5) and 9 mM Mg(OAc)₂) ³⁹. The reaction also included 0.2 μM gp44/62, 1.5 μM gp45, and 1.5 μM primed DNA per well. To initiate the reaction, ATP was added as a 1:1 ATP:Mg(OAc)₂ solution for a final concentration of 100 μM ATP in each 50 μl sample, and the assay temperature was maintained at 23 °C. Absorbances were read every 30 s for 40 mins in a Biotek Synergy H1 plate reader. The molar absorption coefficient for NADH used was K_path = 6220 M⁻¹cm^-1.

Crystal structure of the D110C T4 clamp loader mutant bound to DNA and the sliding clamp

For crystallization, D110C mutant clamp loader (6 mg/mL), clamp, and DNA were mixed in a buffered solution (20 mM Tris pH 7.5 and 100 mM NaCl) to a molar ratio of 1:2.7:2.7 immediately prior to setting crystal trays. ADP•AlF_x was formed within this solution by adding ADP (4 mM), NaF (13mM), AlCl₃ (4 mM), and MgCl₂ (20 mM). Crystals were obtained via hanging drop diffusion (1 uL protein complex solution mixed with 1 μL reservoir solution) with 0.1M MES pH 6.5, 9% PEG MME 5000 and 6% 1-propanol reservoir solution (1 mL). We attempted crystallization of D110C mutant clamp loader:DNA:clamp complex with ADP•BeF_x, ADP•AlF_x and ATPγS. Only the use of ADP•AlF_x produced crystals.

For data collection, crystals were soaked with well solution with 10–20% (v/v) glycerol as cryoprotectant. Cryogenic diffraction data were collected at the Advanced Light Source (Laurance Berkeley National Lab, Beamline 5.0.1. with 0.9998 Å wavelength X-ray). Diffraction data were processed using with XDS ⁴⁰ and AIMLESS ⁴¹ and star-aniso ⁴² to a high resolution cutoff of 3.10 Å. Molecular replacement was performed with Phaser ⁴³ as implemented in Phenix suite ⁴⁴ using a combination of PDB IDs 3U60 and 3U5Z as a search model. Refinement was performed using Phenix and model building was done in Coot ⁴⁵. Non-crystallographic symmetry restraints between ATPase subunits were not imposed (see below). The D110C mutant clamp loader structure and data is deposited with the PDB ID 8UK9.

X-ray crystallography: Re-refinement of wild-type clamp loader structure with DNA and clamp

Since we observed structural changes to the ATPase subunits when not bound to DNA, we sought to locate any internal structural changes of the ATPase domain when DNA is bound to the clamp loader. Unfortunately, the imposed non-crystallographic symmetry restraints masked any unique conformational changes between subunits in our previously published T4 clamp loader structure Thus, we reprocessed data from the original T4 bacteriophage clamp loader:clamp:DNA structure deposition (PDB ID: 3U60) without non-crystallographic symmetry restraints with XDS ⁴⁰ and AIMLESS ⁴¹. The original diffraction data also contained high anisotropy, so we also used the star-aniso web server ⁴² to improve the diffraction data processing. This resulted in data with a higher resolution cut off (2.63 Å). From the re-refinement, the free R-value improved from 0.273 to 0.244, and the stereochemical parameters for the model improved. The newly processed data and refined structure is deposited with the PDB ID 8UH7.

Phage-propagation assay

The phage-propagation assay was carried out as described previously ²⁰. Briefly, T4 bacteriophage engineered to lack clamp-loader genes (T4^del) was used to infect E. coli cells (BL21) harboring the CRISPR Cas12a system and plasmid from the constructed libraries. Prior to infection, a sample of the cells containing the libraries was saved for DNA sequencing as the ‘input’ clamp loader gene population. After a phage replication period of ~14–16 hours, the phage present in the culture supernatant were subject to PCR amplification of the recombined clamp-loader locus. The clamp-loader genes present in the input cell sample were also amplified using PCR. Both sample types were then sequenced using Illumina 500 cycle kits with an Illumina Miseq.

Construction of the saturation mutation library of Asp 110 within the ATPase subunit (gp44)

The wild-type clamp-loader helper plasmid used in previous work was used as the base for the introduction of a single degenerate site (codon sequence NNS) at position 110 of T4 bacteriophage gp44 using Golden Gate cloning methods as described ²⁰. Due to the small library size (23 codons introduced by NNS degenerate primers), library coverage at every stage of library generation was well over 1000-fold. This plasmid was used to transform E. coli (BL21 cells) and was subjected to the phage-propagation assay.

Saturation mutagenesis of the D110C T4 clamp-loader variant

To create the library for saturation mutagenesis of the AAA+ module of the ATPase subunit (gp44), the wild-type saturation mutagenesis library generated previously²⁰ was used, in which every residue in the AAA+ module was replaced by each of the 20 amino acids, plus the stop codon, one at a time.

The library was created in two pools. In one pool (pool A), residues 2–118 were varied, one at mutant per plasmid. Residues 116–230 were varied in the second pool (pool B). The D110C mutation was introduced on top of this existing set of mutants using primer-guided mutagenesis. Primers with an annealing area of 32 nucleotides upstream from position 110 and 23 nucleotides downstream that sensed the wild-type sequence were used to introduce a mutagenetic codon for cysteine at position 110. The 5’ ends of these primers contained a recognition motif for the BsaI restriction enzyme for Golden Gate cloning ⁴⁶. As result, residues 2–118 of pool A were varied, except for residues 101–115, which retain wild-type identity as they were sites of primer annealing for the introduction of the D110C mutation. Every residue of pool B was varied. Library diversity during the cloning stages was maintained at >20x and assessed by colony counts of transformants. These two library pools were used to transform E. coli (BL21) and subjected to the phage-propagation assay.

Cryo-EM sample preparation and data acquisition

To assemble the ternary complex of clamp-loader:clamp:DNA, clamp loader, clamp, and primed DNA were mixed at a molar ratio of 1:1.2:1.2 to a final protein concentration of 5–10 mg/ml. ADP, NaF, BeCl₂, MgCl₂ and β-OG were added to a final concentration of 1 mM, 10 mM, 2 mM, 10 mM and 0.015% (w/v), respectively. To assemble the binary complex of clamp-loader:clamp, clamp loader and clamp were mixed at a molar ration of 1:1.2 to a final protein concentration of 5–10 mg/ml. ATPγS, MgCl₂ and β-OG were added to a final concentration of 1 mM, 10 mM and 0.015% (w/v), respectively. The initial cryo-EM trials for these complexes were unsuccessful, due to severely restricted orientations of the particles on the cryo-EM grids. Addition of the detergent n-octyl-β-D-glucopyranoside (β-OG) at a concentration of 0.015% (w/v) during grid preparation eliminated the preferred orientation, enabling single-particle reconstruction of the T4 clamp-loader:clamp: DNA complex at ~3.2 Å resolution.

For cryoEM grids preparation, QUANTIFOIL^® R 1.2/1.3, 300 mesh, gold grids were rendered hydrophilic by glow-discharge using a PELCO easiGlow^™ system for 45 s at 25 mA. After application of 3.5 μL sample of clamp-loader complexes, grids were plunge-frozen using a Vitrobot Mark IV (Thermo Fisher Scientific, Waltham, MA) with blotting time of 2.5 seconds, blotting force of 20, chamber temperature of +10 °C, and humidity of 95%. The grids were imaged using a Titan Krios G2 transmission electron microscope (Thermo Fisher Scientific, Waltham, MA) operated at 300 kV. Images were recorded on a K3 camera (Gatan, Inc.) operated in super-resolution counting mode with a physical pixel size of 0.524 Å. The detector was placed at the end of a GIF Quantum energy filter (Gatan, Inc.), operated in zero-energy-loss mode with a slit width of 20 eV. All movies were collected at a nominal magnification of 81,000x and a defocus range between −0.8 and −2.2 μm., using the automated data collection software SerialEM ⁴⁷. The nominal dose per frame was ~1 e⁻/Ų and the total dose per movie is ~50 e⁻/Ų. For the clamp-loader:clamp:DNA complex, a total of 10,508 movies were collected. For the clamp-loader:clamp complex, a total of 18,727 movies were collected.

EM image processing

Cryo-EM images were processed using cryoSPARC v4 ⁴⁸. For the clamp-loader:clamp:DNA complex, 10,508 movies were motion-corrected using Patch-Based Motion Correction. After local CTF determination using Patch-Based CTF Estimation, a total of 4,206,415 particles from 10,508 micrographs were automatically selected using template-free Blob_picker and were extracted from the micrographs with a binning factor of 2. A relatively low threshold value was used during the particle picking to include particles with less contrast. More than half of the initial picks are from the background. After iterative 2D classification, most of the background particles were removed based on 2D class averages. The remaining 2,019,311 particles were further processed by ab-initio reconstruction into 4 classes. The resulting 3D maps showed that 3 of the ab-initio reconstruction classes were mostly “junk” particles without recognizable structural features, and the other class showed clear resemblance to the previously reported crystal structure of the clamp-loader:clamp:DNA complex (PDB code: 3U60). This class of particles (754,126) were selected and re-extracted from the micrographs without binning. Another round of ab-initio reconstruction was performed to sort the remaining particles into two classes, a “junk” class (200,803 particles) without any recognizable structural feature and a “clean” class (553,323 particles) resembling the clamp-loader complex. The remaining particles were subjected to iterative rounds of heterogeneous refinement with 2 classes, with each round resulting in a class of particles that yielded improved map quality relative to the previous round. After the map quality stabilized, the remaining 204,448 particles were subjected to a final round of non-uniform refinement, with the Defocus Refinement option turned on. The resulting map (overall resolution: 3.15 Å) was used for model building and refinement.

For the clamp-loader:clamp complex, 18,727 movies were motion-corrected using Patch-Based Motion Correction. After local CTF determination using Patch-Based CTF Estimation, a total of 17,937,844 particles from 18,727 micrographs were automatically selected using template-free Blob_picker and were extracted from the micrographs with a binning factor of 2. A low threshold value was used during the particle picking to include particles with less contrast. After iterative 2D classification, 2D classes with no recognizable feature were removed. 3D Ab-initio reconstruction was performed on two subsets of particles to generate the initial models of sliding clamp and clamp-loader:clamp complex. The first subset of particles (711,674) contains 2D classes showing sliding clamp only. The second subset of particles (314,348) contains 2D classes showing clamp-loader:clamp complex. The resulting models of sliding clamp, clamp-loader:clamp complex, together with a “junk” model generated from background particles, were used as input models for one round of heterogeneous refinement to sort the remaining 9,692,266 particles into 3 classes. Particles corresponding to the clamp-loader:clamp class (2,525,436) were selected and re-extracted from the micrographs without binning. One round of heterogeneous refinement was performed to sort the remaining particles (2,525,436) into 10 classes. After visual inspection, the class of particles (455,041) corresponding to the map with best quality was selected and subjected to a final round of non-uniform refinement, with the Defocus Refinement option turned on. The resulting map (overall resolution: 3.21 Å) was used for model building and refinement.

Model building and refinement

To generate the initial model for the cryo-EM structure of the clamp-loader:clamp:DNA complex (DNA-bound state), the re-refined crystal structure of the T4 clamp-loader:clamp:DNA complex (PDB code: 8UH7) was rigid-body fitted into the reconstructed cryo-EM map of the clamp-loader:clamp:DNA complex, using UCSF Chimera ⁴⁹. UCSF chimera-fitted model was subjected to one round of real-space refinement ⁵⁰ in Phenix. The model was then manually adjusted in Coot, followed by iterative rounds of real-space refinement in Phenix and manual fitting in Coot. Model quality was assessed using the comprehensive model validation tools from Phenix.

To generate the initial model for the cryo-EM structure of the clamp-loader:clamp complex (DNA-free state), the re-refined crystal structure of the T4 clamp-loader:clamp:DNA complex (PDB code: 3U60) was split into five fragments: the pentameric collar domains (residues 233–319 of subunits B, C, D, and E, and residues 40–106 of subunit A), the A leg (residues 2–39, subunit A), the A’ leg (residues 107–187, subunit A), the AAA+ module of subunit B (residues 1–232, subunit B), and the AAA+ module of subunit E (residues 1–232, subunit E). The structure of a closed sliding clamp was taken from the previously reported crystal structure of the T4 clamp-loader:DNA complex bound to a closed sliding clamp (PDB code: 3U61, chain F, G, and H). These six fragments (5 from PDB: 3U60, and 1 from PDB: 3U61) were rigid-body fitted into the reconstructed cryo-EM map of the clamp-loader:clamp complex, using UCSF Chimera. The Chimera-fitted models were then subjected to one round of real-space refinement in Phenix, individually. The six models of fragments were manually adjusted and merged into one model in Coot, followed by iterative rounds of real-space refinement in Phenix and manual fitting in Coot. Model quality was assessed using the comprehensive model validation tools from Phenix. All maps and models were visualized using UCSF Chimera and PyMOL.

Extended Data

Extended Data Fig 1. Mutational tolerance and location of position 110 within T4 bacteriophage clamp loader AAA+ ATPase domains.

A, We used a focused mutagenesis library in which only the identity of the residue at position 110 was varied and measured the fitness of the variants. The fitness assay was conducted as described in Fig. 1B. Fitness of each substitution of position 110 is plotted. The fitness of synonymous codons was also measured to insure internal reproducibility of fitness measurements. Data are presented as mean values +/− standard deviations (SD), n=5 biologically independent experiments B, The crystal structures of the remodeled wild-type clamp loader (translucent) and D110C mutant clamp loaders are aligned on residues 1–120 (domain 1) of subunit C (purple). At each of the ATP-binding sites in the wild-type complex, the sidechain of Asp 110, in the DEAD-box motif, forms hydrogen bonds with an arginine sidechain (Arg 122) provided by the adjacent subunit, and also with the backbone of the catalytically important sensor 1 residue, Asn 139 ²⁷. In the D110C complex, the cysteine residue at position 110 interacts with only one of these residues at a time. When the cysteine sidechain does interact closely with that of Arg 122, the termini of the sidechains are within 3.0 Å of each other. This suggests that the cysteine is deprotonated and negatively charged at those sites, explaining the mild phenotype of the D110C mutant. The inability of the cysteine sidechain to engage Arg 122 and Asn 139 simultaneously results in a subtle change in the interfacial geometry between subunits that is distributed throughout the structure. This leads to a slight change in the overall engagement with DNA, which might result in the weakened DNA affinity.

Extended Data Fig 2. Biochemical assessment of D110C mutant clamp loader.

A, Purified wild-type T4 clamp-loader complex and the D110C variant are assessed for DNA-binding affinity by using a fluorescence polarization anisotropy assay. In previous work, a fluorescent tag on a DNA duplex with a 5’ overhang (representing primed DNA) was used to measure the change in fluorescence anisotropy of the labeled DNA after adding the clamp loader ^22,53,54. In the present experiments, we used primed DNA (50 nM) consisting of a 20 base-pair DNA duplex with a 10-nucleotide overhang labeled with TAMRA dye. The labeled DNA was added to a reaction mixture containing the T4 sliding clamp (7 μM) and the ATP analog ADP•BeF₃ (1 mM). Purified clamp loader (0–5 μM) was then titrated into the sliding-clamp:primed DNA:nucleotide analog mixture and allowed to equilibrate for 3 hours at room temperature before fluorescence anisotropy measurement (see Methods). The anisotropy per clamp loader concentration is plotted for wild-type D110C mutant clamp loader. Data are presented as mean values +/− standard deviations (SD), n=3 biological replicates. B, We measured the kinetics of ATP hydrolysis by using an assay in which the consumption of ATP by the clamp loader is coupled to the oxidation of NADH ^12,38,54. Kinetic measurements were made for both the wild-type clamp loader and the D110C mutant (0.5 μM) bound to clamp (2.5 μM) and the same primed DNA construct used for the fluorescence anisotropy assay (2.5 μM), but without the fluorophore label. ATP hydrolysis rate of the D110C mutant is ~33% of the rate exhibited by the wild-type clamp loader (Supplemental Figure 1D). The DNA concentration used for this assay is ~5-fold higher than the K_D value for the wild-type clamp loader, which presumably accounts for the partial stimulation of ATP hydrolysis by the D110C mutant. Example traces of these kinetic measurements are shown.

Extended Data Fig 3. Electron density of wild-type and D110C T4 bacteriophage crystal structures.

A, The electron density map (2FoFc) of the re-refined wild-type clamp loader structure is shown centered around the active site of subunit D. The map is rendered at a contour level of 0.6176 e/ų (2.09 RMSD). Two view angles of the active site are shown. The nucleotide is the central axis of the view. B, The electron density map (2FoFc) of the D110C clamp loader mutant structure is shown centered around the active site of subunit D. The map is rendered at a contour level of 0.2253 e/ų (2.07 RMSD). Two view angles of the active site are shown. The nucleotide is the central axis of the view. C, The crystal structure of D110C mutated clamp loader (molecule A) is aligned against the wild-type clamp loader structure (gray). The alignment is performed on subunits C and D of both structures. The A, B, C, and E, subunits of the D110C structure are colored green, blue, magenta, yellow, and pink, respectively. The clamp subunits of the D110C structure are colored light blue, purple, and orange.

Extended Data Fig 4. Statistics of D110C clamp loader deep mutagenesis screen.

A, Agreement between fitness measurements from three replicate experiments per assay pool (residues 2–115 assay pool n=1130, and residues 115–230 assay pool n=2076) is shown in scatter plots. Each point in the scatter plot represents the fitness measurements made from the two trials. B, The histogram shows the spread of relative fitness values for the AAA+ module of the ATPase subunit in the background of the D110C mutant clamp loader. C, Small-scale libraries of recovery mutations in the background of D110C were constructed and subjected to the phage replication assay. The fitness of this small library per substitution (x-axis) is plotted against the fitness scores of same mutations as assayed in the full-scale deep mutagenesis screen (y-axis). Data are presented as mean values +/− standard deviations (SD), n=3 biological replicates.

Extended Data Fig 5. Biochemical characterization of D110C rescue mutations.

A, Using TAMRA-labeled double-stranded DNA, fluorescence anisotropy measurements are taken to assess the DNA-binding capabilities of four recovery mutations in the background of the D110C mutant clamp loader (denoted by a star *). These four representative mutants were chosen to span both strong (P50K*) and weaker recovery (D98W*) activity. Data are presented as mean values +/− standard deviations (SD), n=3 biologically independent experiments B, The ATP hydrolysis recovery of four rescue mutations in the background of D110C mutant clamp loader is measured via use of the coupled kinase assay. We determined the ATP hydrolysis rates of each clamp loader variant alone, in the presence of DNA alone, in the presence of clamp alone, and an activated clamp loader complex containing both clamp and duplexed DNA. Data are presented as mean values +/− standard deviations (SD), n=2 to 6 technical replicates, with at least two biological replicates showing the same hydrolysis trends per clamp loader variant studied.

Extended Data Fig 6. Mapping recovery mutations onto structure with wild-type sensitive residues indicated.

Recovery positions having at least 6 substitutions that recover D110C replication fitness are mapped onto the wild-type clamp loader crystal structure (PDB ID: 3U60) in green. Positions that are identified to be particularly sensitive in the wild-type background (Figure 2) are shown in blue.

Extended Data Fig 7. Representative Cryo-EM micrographs.

A. Representative Cryo-EM micrographs of wild-type T4 clamp-loader:clamp with DNA B. Representative Cryo-EM micrographs of wild-type T4 clamp-loader:clamp without DNA.

Extended Data Fig 8. Mechanistic evaluation of rescue hotspot residues.

A, An expanded view from Figure 7B is taken of residue Pro 50 (red spheres). The DNA-unbound structure (wheat) features greater distance between Pro 50 and Asn 139. The distance between ATPyS and Asn 139 in this structure is also greater (red dashes, 5.6 angstroms). When DNA is bound (purple), Pro 50 moves closer to Asn 139, while Asn 139 moves closer to ADP•BeF₃ (black dashes, 4.7 angstroms) B, The interfacial surface seen in Figure 7A is expanded to show five distinct rescue hotspots (red spheres) around the hydrophobic plug residue, Phe 28. The DNA-bound conformation of chains C (purple) and B (blue) is aligned with the DNA-unbound conformation of chain C (wheat) along the central coupler region. The hydrophobic plug is formed by residues Tyr 214, Ile 224, and Leu 227.

Extended Data Fig 9. Kinetic measurement of fluorescence anisotropy over a range of clamp loader concentrations.

The fluorescence anisotropy of clamp loader to fluorescently labeled primer-template DNA is measured over time and across clamp loader concentration. This complex was also bound to the non-hydrolysable analog ADP*BeF3. Very slow equilibration of the anisotropy readout is seen at low clamp loader concentration (between 0.009 uM and 0.039 uM). Clamp loader at a concentration of around 0.08 uM generates consistent anisotropy values after approximately 10,000 seconds (~2.5 hours).

Data availability

Deep mutagenesis data generated during this study have been deposited in the Sequence Read repository with BioProject accession number PRJNA1025274. Deep mutagenesis data of the wild-type clamp loader can be found in our previous publication ²⁰. Crystal structures of the D110C mutant bacteriophage clamp-loader:clamp:DNA complex and reprocessed wild-type bacteriophage clamp-loader:clamp:DNA complex have been deposited in the Protein Data Bank (PDB) with accession codes 8UK9 and 8UH7. The cryo-EM density maps of T4 bacteriophage clamp-loader:clamp DNA-bound complex and T4 bacteriophage clamp-loader:clamp DNA-free complex been deposited to the Electron Microscopy Data Bank (EMDB) under the accession codes EMD-42399 and EMD-42402. The associated coordinates have been deposited to the PDB under accession codes 8UNF and 8UNH. Biochemical data are supplied as Excel spreadsheets, with relevant figures listed in the filename.