Theoretical Development of DnaG Primase as a Novel Narrow-Spectrum Antibiotic Target.

The widespread use of antibiotics to treat infections is one of the reasons that global mortality rates have fallen over the past 80 years. However, antibiotic use is also responsible for the concomitant rise in antibiotic resistance because it results in dysbiosis in which commensal and pathogenic bacteria are both greatly reduced. Therefore, narrow-range antibiotics are a promising direction for reducing antibiotic resistance because they are more discriminate. As a step toward addressing this problem, the goal of this study was to identify sites on DnaG primase that are conserved within Gram-positive bacteria and different from the equivalent sites in Gram-negative bacteria. Based on sequence and structural analysis, the primase C-terminal helicase-binding domain (CTD) was identified as most promising. Although the primase CTD sequences are very poorly conserved, they have highly conserved protein folds, and Gram-positive bacterial primases fold into a compact state that creates a small molecule binding site adjacent to a groove. The small molecule would stabilize the protein in its compact state, which would interfere with the helicase binding. This is important because primase CTD must be in its open conformation to bind to its cognate helicase at the replication fork.

Introduction

The widespread use of antibiotics to treat infections is one of the reasons that global mortality rates have fallen over the past 80 years.[1] However, the concomitant rise in antibiotic resistance is projected to counter those successes with 10 million deaths by infections per year worldwide by 2050.[2] Many of the deadliest pathogens are Gram-negative and Gram-positive bacteria (Table ). When broad-range antibiotics are taken orally to treat gastrointestinal infections (Benthesicymus cereus, C. burnetii, Cosmarium difficile, E. coli, E. faecium, F. tularensis, V. cholerae), it results in long-term adverse changes to hundreds of gut microbes that play positive roles in human health.[3] The result is a decrease in the immune response, reduced vitamin production, and an increase in antibiotic resistance. Narrow-range antibiotics overcome these disadvantages and have already been developed to combat tuberculosis, hospital-acquired diarrhea, and even antibiotic resistance.[4−6]

Table 1

Pathogenic Gram-Negative and Gram-Positive Bacteriaa

organism	genera	relevance
Gram-negative Bacteria
Burkholderia mallei	Burkholderia	glanders[7]
Burkholderia pseudomallei	Burkholderia	melioidosis[8]
Coxiella burnetii	Coxiella	Q fever[9]
Escherichia coli	Enterobacteriaceae	diarrhea[10]
Francisella tularensis	Coccobacillus	tularemia[11]
Neisseria gonorrhoeae	Neisseria	gonorrhea[12]
Pseudomonas aeruginosa	Enterobacteriaceae	pneumonia, sepsis[13]
Ralstonia solanacearum	Ralstonia	pathogenic to plants[14]
Vibrio cholerae	Vibrio	cholera[15]
Yersinia pestis	Yersiniaciae	bubonic plague[16]
Gram-positive Bacteria
Bacillus anthracis	Bacilli	anthrax[17]
Bacillus cereus	Bacilli	emetic and diarrheal syndrome[18]
Clostridioides difficile	Clostridia	gastrointestinal infection (C diff)[19]
Enterococcus faecalis	Bacilli	endocarditis[20]
Enterococcus faecium	Bacilli	gastrointestinal infection[21]
Staphylococcus aureus	Bacilli	MRSA, staph infection[22]
Staphylococcus epidermidis	Bacilli	hospital-acquired skin infection[23]
Staphylococcus haemolyticus	Bacilli	blood infection[24]
Streptococcus agalactiae	Bacilli	neonatal infection[25]
Streptococcus pneumoniae	Bacilli	pneumonia, ear & sinus infection[26]

These bacteria have medical and military importance, and several are on the HHS and USDA Select Agents and Toxins list.[27]

A good example relating to the effects of broad- and narrow-range antibiotics on the gut microbiome is Clostridioides difficile infection (C diff).[19] The resulting diarrhea is often acquired at nursing homes or hospitals, where it persists because the organism forms spores that are difficult to disinfect. Broad-range antibiotics lead to a rise of antibiotic resistance when microbes reinhabit the gut as the antibiotic concentration is decreasing but still modestly high. There are two narrow-range antibiotics marketed for C diff. Their use prevents dysbiosis, or wide-range loss and imbalance, of gut microbiota, which is a major factor in the recurrence of infection. Fidaxomicin targets RNA polymerase[28] from Gram-positive anaerobes, including staphylococci and enterococci.[28] Its use has been shown to preserve most gut microbiota. Ridinilazole targets the process of cell division from an even narrower range of Gram-positive anaerobes than fidaxomicin.[29] Neither of these antibiotics passes through the intestinal wall, which leads to high concentrations in the gut and favorably high percent of antibiotic inhibition of their targets.

Very few of the over 200 essential bacterial protein-encoding genes[30] have been exploited as drug targets. More than 10 of those genes code for replication enzymes, some of which are under development as targets.[31] The two enzymes in this study are helicase (DnaB) and primase (DnaG). DnaB helicase is the homohexameric enzyme that couples ATP hydrolysis with unwinding upstream DNA at the replication fork (note that the name DnaB helicase is used here for the protein that is named DnaC helicase in Staphylococcus aureus and Bacillus subtilis because, in all other bacteria, DnaB is the name of the replicative helicase, and DnaC is the name of the helicase loading enzyme). DnaG primase is the specialized DNA-dependent RNA polymerase that synthesizes short oligoribonucleotide polymers called primers. It is required because DNA polymerases lack the ability to initiate chain synthesis but are very efficient at elongating from primers. Both are good antibiotic targets because they are encoded by single genes that produce a few dozen copies of an enzyme that plays critical roles during the initiation and elongation phases of DNA replication.[32−36]

The goal of this study was to identify the DnaG primase residues that are predicted to have the strongest differences between Gram-positive and Gram-negative bacteria. Since bacterial primase has a distinct fold compared to eukaryotic primase, there should be fewer antibiotic side effects. First, the highly conserved primase N-terminal zinc-binding domain (ZBD) was examined as a case study. Then, the primase C-terminal helicase-binding domain (CTD) residues were studied. Although the primase CTD sequences are very poorly conserved, they have highly conserved protein folds. The primase ZBD and CTD were found to have functional differences between Gram-negative and Gram-positive bacteria.

Results

The goal of this project was to identify residues in DnaG primase that play an important role and that are conserved within a Gram phylum but different between them. The plan was to work out the methods for discovering these residues by reanalyzing how this was accomplished for the highly conserved primase ZBD and then apply those methods to the more challenging primase CTD. The CTD lacks sequence conservation even within the phyla. However, the available solved structures suggest a higher level of structural conservation that showed distinct differences between Gram-positive and Gram-negative organisms.

Multiple Sequence Alignments

Given the coevolution of the DnaB–DnaG interaction in G. stearothermophilus, S. aureus, and E. coli, the sequences of primase and helicase from select pathogenic bacteria were aligned (Table ) as the first step to learn which features were conserved and which might be useful for drug target sites. The sequence analysis shows a distinctive separation of primase and helicase in Gram-positive and Gram-negative organisms (Figure ). Furthermore, the organismal clustering is also mostly conserved: the two Burkholderia species have the same sequence; E. coli, V. cholerae, and Y. pestis are closely related; S. aureus and S. epidermis are closely related; E. faecium and E. faecalis are closely related; and S. pneumoniae and S. agalactiae are closely related. Also conserved is that F. tularensis is the most distantly related for both proteins among the Gram-negative organisms and C. difficile among the Gram-positive organisms. As the farthest outliers, it may be difficult to develop reliable homology models for the enzyme from these organisms. One difference between the trees is that B. cereus and B. anthracis have identical primase sequences, whereas their cognate DnaB helicases are not identical but closely related.

Figure 1

Phylogenetic full-length protein sequence trees for (A) DnaG primase and (B) DnaB helicase.

There is no structure for the entire primase–helicase complex. The structures of several individual domains have been determined (Figure A). Primase consists of three domains: Zinc-Binding Domain (ZBD), RNA Polymerase Domain (RPD), and C-Terminal Domain (CTD). Helicase consists of two domains: N-Terminal Domain (NTD) and the Helicase Motor/Body. The CTD structure has been solved from three different organisms—two Gram-positive and one Gram-negative (Figure A). The structures of three multidomain proteins, one of which is a complex, have been solved (Figure B). The ZBD–RPD structure from A. aeolicus, a hyperthermophilic organism, suggests that the linker between the two domains is not flexible. The DnaB helicase homohexamer structure from G. stearothermophilus shows as a trimer of dimers. The DnaB–DnaG CTD complex from G. stearothermophilus established the ratio of one CTD for every two DnaB-NTDs.

Figure 2

Domain organization of the DnaG–DnaB complex. The structures of (A) three individual domains and (B) three multidomain proteins and complexes have been determined. The organisms are abbreviated as follows: Gstea, G. stearothermophilus; Ecoli, E. coli; Saure, S. aureus; Mtube, M. tuberculosis and Aaeol, A. aeolicus. The structures from the PDB are 1D0Q;[37]13TW;[38]2LZN;[39]1Z8S;[40]2R5U;[41]2AU3;[42]2R6A;[43] and 2R6C.[43]

Zinc-Binding Domain

The primase ZBD domain residues are highly conserved (Figure for 20 organisms; Supporting Information for 40 organisms covering a wider range of classes, Figures S1–S7). The Gram-positive primases are 37% identical over this region, and the Gram-negatives are 47% identical. This 52-residue region encompasses the four zinc-binding residues C, H, C, and C (Figure A gray highlighted residues). It is a potential site for the binding of a narrow-range antibiotic because it is the domain responsible for the initiation, which shows distinct phyletic differences. Enzymatically, S. aureus and G. stearothermophilus primase initiate at the triplet sequence 5′-d(CTA),[39,44] whereas the E. coli primase initiates from 5′-d(CTG).[45] A mutagenesis study of the residues that differ between Gram-positives and Gram-negatives but that are conserved in one or the other of those phyla showed that the S. aureus residues Ile56 and Cys57 (Figure A purple highlighted residues) were responsible for initiating from 5′-d(CTA) rather than 5′-d(CTG).[46]

Figure 3

Primase zinc-binding domain. (A) The ZBD sequence and secondary structure alignment showing the zinc-binding residues in gray and the residues responsible for initiation specificity in purple (for Gram-positive organisms) and pink (for Gram-negative organisms). The numbers along the top relate to the residues in the S. aureus sequence. The five beta strands in the 1D0Q.pdb secondary structure from G. stearothermophilus are shown next. The identical (*), highly conserved (:), and conserved (.) residues are shown separately for the Gram-positive and Gram-negative organisms. (B) Primase ZBD from G. stearothermophilus showing the specificity residues, Ile58 and Phe59, in purple.

The primase ZBD structure has been determined from G. stearothermophilus(37) and Aquifex aeolicus.[42] They are members of a unique protein family called “zinc ribbons,” in which the zinc is coordinated by three cysteines and one histidine that hold together five antiparallel beta strands (Figure B). The two or three amino acids between the zinc-chelating residues are “knuckles” with highly conserved sequences. Before and after the zinc ribbon structure are α helices that form a bundle. The two residues responsible for recognizing the third nucleotide of the initiation sequence are located at the end of one beta strand located in the center of the five-strand antiparallel sheet (Figure B). The exposed Ile58 may stack on the key nucleotide, and the buried Phe59 may ensure the beta-strand maintains its conformation.

Since the single-stranded DNA template is bound much more strongly to the RNA polymerase domain,[42,47] the ZBD must have enough flexibility to fold onto the DNA in a sequence-specific manner. With regard to drug binding, beta sheets are challenging drug target sites because their flat surfaces do not form binding pockets.[34] A more promising strategy might be to reduce the ability of the ZBD to interact with the single-stranded DNA template that is bound to the adjacent RNA polymerase domain.[42,47] This could be achieved by developing molecules that bind to the α helical bundle or the linker between the ZBD and the RNA polymerase domains. However, the structure of the linker and its flexibility has not been established yet.

DnaG-DnaB Interface

Residues at the DnaG–DnaB interface should be narrow-range antibiotic targets because there are distinct enzymatic differences between genera and even species of bacteria. First, E. coli primase activity is strongly stimulated by its cognate DnaB helicase,[48]S. aureus primase is stimulated only about 2-fold by its DnaB helicase,[49] and C. difficile primase is only active in the presence of its cognate DnaB helicase.[50] The structural basis for these functional differences has not been determined but are likely related to the different affinities of the catalytic subdomain for the ssDNA template and/or the different affinities of primase C-terminal domain for the cognate helicase N-terminal domain. Information about the primase–helicase interface interactions is based on studies of the X-ray structure of thermophilic G. stearothermophilus proteins.[47] At room temperature, the binding affinities for thermophilic proteins are much stronger than those from mesophilic bacteria because thermophilic proteins are less prone to unfolding.[51]

Sequence alignments supported the idea that there are structural differences between Gram-positive and Gram-negative CTDs (Figure ) despite sharing no identical residues and almost no conserved residues. The residues at interface A (purple highlights in Figure ) are in the final helix of the protein and encompass four of the six residues that show any degree of sequence conservation. The residues at interface B of Gram-positive organisms (orange highlights) are Aspartate and Tyrosine in four Gram-positive organisms with possible corollary residues in the Gram-positive sequences but not in the Gram-negative sequences. The residues highlighted in green form the small molecule binding site in the S. aureus DnaG closed conformation. The secondary structures of Gram-negative CTDs have many more helix-favoring residues at the site where Gram-positives have a turn. This suggests that Gram-negative CTDs lack the flexibility to change conformation, which indicates there are significant structural differences with the Gram-positive primase CTDs.

Figure 4

Primase CTD sequence alignment. The primase CTD from the organisms in Table were aligned with numbers corresponding to the S. aureus sequence and with two arrows showing the location of two critical G. stearothermophilus residues. Along the top are the secondary structures from Gram-positive organisms and, along the bottom, from Gram-negative organisms. Among Gram-positives, there are two closed structures and one open. The red box and delta sign show the major structural differences between the open and closed conformations. In the center of the sequences are the few conserved residues (: is highly conserved;. is conserved). Along the bottom are the four open conformations from Gram-negative organisms. The green highlighted residues fill the small molecule binding pocket of the S. aureus closed conformation. The orange highlighted residues are at interface B. The purple highlighted residues are buried at interface A. The gray highlighted residues are conserved helix-stabilizing residues, two of which are not conserved in Gram-positive sequences, where the conformational change occurs.

G. stearothermophilus primase CTD adopts two conformations.[40,43] Its open conformation binds to two helicase NTDs (Figure A). The formation of this complex increases the affinity between DnaB helicase and the DNA polymerase III holoenzyme by 500-fold, which may allow rapid transfer of the short RNA primer to the DNA polymerase.[52,53] The homology model of S. aureus primase CTD sequence in its open conformation (Figure B) was created using the G. stearothermophilus structure as a template (Figure B).[39] The two DnaG-DnaB interfaces are distinct. Interface A involves the extreme C-terminal helical hairpin of DnaG CTD and the five-helix bundle of DnaB NTD 1. In contrast, interface B is between the five-helix bundle of DnaG CTD and the five-helix bundle of DnaB NTD 2.

Figure 5

Bacterial primase CTD conformational changes and interfaces. (A) Diagram of the S. aureus primase CTD in its open and closed conformations. Interfaces A and B are between the two DnaB NTD and the “open” conformation of a single primase CTD. Interface A residues are in purple and interface B in orange. Small molecules, such as acyclovir, bind to the groove formed by the closed conformation to form interface C in green. (B) Key interface residues are shown in the open conformation of S. aureus primase CTD, a homology model using 2R6C (G. stearothermophilus) as a template structure. (C) Key interface residues are shown in the closed conformation of S. aureus primase CTD (2LZN).

The other CTD conformation is compact (Figure C) and was observed in the solution structures of two Gram-positive primases.[39,40] The difference from the open conformation is that there is a bend in the longer helix of the terminal hairpin. In the S. aureus primase CTD compact conformation, the resulting shorter terminal hairpin forms several weak interactions with its five-helix bundle to form a closed conformation. This forms a groove into which small molecules are able to bind with low millimolar concentrations affinity[39] (Figure ) to prevent interface B from forming. The currently identified small molecules can serve as seed molecules in the search for others with higher affinity.

Characterizing the Critical Residues at Interface A through Virtual Mutations

The PRODIGY program was used on the G. stearothermophilus DnaB–DnaG co-crystal to identify the interatomic contacts at interface A within a defined cutoff distance of 5.5 Å (Table ). Six of the residues were the same as those previously identified as making multiple contacts at the interface: E572T, F577V, L578E, A581K, A584L, and I588V. Nine other residues were more solvent-exposed.

Table 2

Predicted ΔΔG for Single Virtual Mutations of G. stearothermophilus DnaG CTD Interface Aa

virtual mutation	ΔΔG (kcal/mol)
Wild-type	0.0 (ΔG = −6.2 kcal/mol)
T569G	–0.3
A580G	0.0
A581G	0.0
L578G	0.0
A584G	0.0
E565G	0.1
I588G	0.1
K575G	0.2
K591G	0.2
K568G	0.2
F577G	0.2
R582G	0.2
K585G	0.4
K592G	0.4
E572G	0.5

PRODIGY identified the residues at the interface within a defined cutoff distance of 5.5Å (SI Figure ).

Figure 8

Colocalization of key residues at interface C. Homology model of the closed conformation of C. difficile primase CTD that was created using 2LZN (S. aureus) as a template structure. The green residues are equivalent to those that form the small molecule binding pocket in S. aureus primase. The red residues are the equivalent residues that are predicted to most perturb the DnaB interaction in G. stearothermophilus as described above.

To establish whether there were any residues that played key roles in the interface interaction, each of the 15 G. stearothermophilus residues was virtually mutated singly to glycine using the DeepView-Swiss-PdbViewer, and the strength of the binding interaction between the resulting mutant and helicase NTD 1 was quantified with PRODIGY. This measurement estimates the relative importance of the residues to the overall interaction and would be different from in vitro binding strengths because both interfaces A and B would have to be disrupted. The free energy was −6.2 kcal/mol for the wild-type primase interface. When each of the 15 residues was singly mutated, three of them weakened ΔG by more than 0.2 kcal/mol: E572, K585, K592 (Table ). The strongest perturbation was caused by mutating E572. This is the only residue that made contact with multiple adjacent residues and was exposed to solvent.

To determine whether there was an interface A sequence code, all six buried interface residues were simultaneously mutated to glycine (Table ). The resulting ΔΔG of 0.6 kcal/mol was significant but indicated that each residue makes a modest individual contribution (and an average of 0.1 kcal/mol) to the overall affinity. To determine whether the contributions of those six residues were conserved among Gram-positive organisms, they were virtually mutated to the equivalent sequence in S. aureus. The small value for ΔΔG indicated that the S. aureus residues were capable of substitution, which was consistent with phyletic conservation.

Table 3

Predicted ΔΔG for Multiple Virtual Mutations of G. stearothermophilus DnaG CTD Interface Aa

virtual mutation	ΔΔG	notes
E572G, F577G, L578G, A581G, A584G, I588G	0.6	all closest contacts to glycine
E572T, F577V, L578E, A581K, A584L, I588V	–0.2	all closest contacts to S. aureus
F577G, L578G, A581G, A584G, I588G	0.2	closest contacts to glycine, except E572
E572G, K585G, K592G	1.1	three most perturbing to glycine
K585G, K592G	0.7	second & third most perturbing to glycine
E572G, K585Q, K592G	0.3	two of three most perturbing to glycine, K585 to S. aureus
E572G, K585W, K592G	0.9	two of three most perturbing to glycine, K585 to E. coli

PRODIGY identified the residues at the interface within a defined cutoff distance of 5.5 Å (Supporting Information Figure S8).

Since testing the most buried residues did not identify any critical interface residues, the three solvent-exposed interface residues were examined (Table ). When E572, K585, and K592 were virtually mutated to glycine, the free energy was perturbed by 1.1 kcal/mol, or about 0.4 kcal/mol/residue. The location of these amino acids can be seen in Figure . This large contribution per residue was confirmed by a perturbation of 0.7 kcal/mol when only two of those residues were mutated (K585 and K592). To determine the degree of phyletic conservation, K585 was mutated to glutamine, the equivalent residue in the S. aureus. This mutant had nearly the same free energy as wild-type primase, consistent with conservation of function within the phylum. Next, K585 was mutated to tryptophan, the equivalent residue in the E. coli. This mutant had nearly the same free energy as the glycine mutant, consistent with a difference in function between Gram-negative and Gram-positive organisms.

Figure 6

G. stearothermophilus DnaG CTD key interface A residues—E572, K585, and K592—on the terminal helical hairpin in blue as determined by virtual mutation.

DnaB Interface Residues

The DnaB NTD sequences are much more highly conserved than primase CTD sequences (Figure ) despite being members of the same protein fold family. Among Gram-positive bacteria, DnaB NTD sequences are 24% identical over 51 residues and 31% identical among Gram-negative (Figure ). All five residues at interface A (purple highlights) are either conserved or adjacent to identical or highly conserved residues, suggesting that there is a scaffold for orienting the more variable interface residues in similar stereospecific locations. This is consistent with a species-specific code.[48] Given their high sequence and structural conservation, the available DnaB NTD structures are good templates for creating homology models of the other DnaB helicase sequences (Periago et al., preliminary results).

Figure 7

DnaB helicase NTD sequence alignment. The DnaB NTD from the organisms in Table were aligned with numbers corresponding to the S. aureus sequence. Along the top are the secondary structures from Gram-positive organisms and, along the bottom, from Gram-negative organisms. The secondary structure for M. tuberculosis is shown below the main alignments. The Gram-negative organisms and M. tuberculosis have a linker structure (red box and delta symbol) that differs from Gram-positives. In the center of the sequences are the identical (*), highly conserved (:) and conserved (.) residues. The purple highlighted residues are at interface A. The yellow highlighted residues are at interface B.

Discussion

DnaB helicase and DnaG primase are antibiotic targets because they interact to carry out critical reactions during DNA replication. These enzymes are narrow-range targets because Gram-positive and Gram-negative primases and helicase have distinct specificities, which are the result of three domains—the primase zinc-binding domain, the primase CTD, and the helicase NTD. Although the primase zinc-binding domain is responsible for distinct differences between the phyletic substrate specificity, its protein fold lacks a pocket for small molecule binding. On the other hand, Gram-positive primase CTD creates a groove when the C-terminal hairpin folds into its closed conformation. This groove creates a small molecule binding pocket in S. aureus primase that can serve as a model for the C. difficile primase (Figure ). One of the three critical interface A residues, E572, is near that pocket. Therefore, a promising direction would be to extend the size of the small molecule so that it extends into the adjacent groove to also bind E572. Such binding will favor the closed conformation of primase CTD, which cannot bind to the helicase because the interface A residues are sterically obscured.

Experimental Section

The organism sequences were obtained from UniProt.[54] Clustal Omega[55] was used to create the multiple sequence alignments in a phylip format. SplitsTree4[56] used the phylip format from Clustal Omega to create the phylogenetic trees. Chimera[57] was used to examine the 3D model structures from their PDB IDs. The 3D homology models were created using the modeling program from SWISS-MODEL[58] using template structures from the PDB. PRODIGY[59] was used to predict the amino acids in closest proximity (5.5Å or less) as well as the ΔG for the protein–protein binding interaction. PRODIGY has the option to select the temperature, and the ΔG values were calculated at 37 °C due to its biological relevance. DeepView-Swiss-PdbViewer[60] was used to introduce virtual mutations to the wild-type protein, and PRODIGY was used to predict the ΔG at 37 °C for each mutation. ΔΔG was calculated by subtracting the ΔG wild-type from the ΔG mutation. PRODIGY predicts the binding affinity using the formula reported by Vangone and Bonvin.[61] It counts the number of Interatomic Contacts (ICs) made at the interface of a protein–protein complex within a 5.5 Å distance threshold and classifies them according to the polar/apolar/charged character of the interacting amino acids. This information is then combined with properties on the Non-Interacting Surface (NIS), which was previously shown to influence the binding affinity.[62]