In E. coli, DnaA binding to oriC is controlled both before and immediately after initiation by mechanisms that prevent completion of orisome assembly stages 2 and 3 (Leonard and Grimwade, 2005). Before initiation, the DNA bending protein Fis helps maintain the origin in a conformation that reduces DnaA’s ability to bind low affinity sites, until levels of DnaA increase enough to displace Fis from its recognition site (Ryan et al., 2004; Kaur et al., 2014). Since E. coli oriC contains multiple low affinity DnaA binding sites that preferentially bind DnaA-ATP (McGarry et al., 2004; Kawakami et al., 2005), orisome assembly cannot be completed until DnaA-ATP levels rise to a critical level. (Regulation of DnaA-ATP levels is discussed below.) After initiation, the SeqA protein binds hemimethylated GATC motifs in oriC, several of which are inside or overlap low affinity DnaA recognition sites (Lu et al., 1994; Skarstad et al., 2001). SeqA blocks DnaA-ATP from re-occupying low affinity sites and the DUE region for approximately one third of the cell cycle (Nievera et al., 2006).

It isn’t known how many bacterial origins contain low affinity recognition sites with preference for DnaA-ATP, and not all bacteria use Fis or SeqA to regulate orisome assembly (Brézellec et al., 2006; Madiraju et al., 2006). Regardless, the basic paradigm of controlling DnaA’s access to oriC as a way of regulating orisome assembly can be found in many bacterial types. For example, response regulators CtrA, MtrA, and HP1021 inhibit DnaA occupation of oriC in Caulobacter crescentus, Mycobacteria tuberculosis, and Helicobacter pylori, respectively, and by doing so, help prevent untimely initiations (Taylor et al., 2011; Donczew et al., 2015; Purushotham et al., 2015). H. pylori also uses DNA topology to regulate DnaA/oriC interactions (Donczew et al., 2014). In Bacillus subtilis, several proteins have been identified that negatively regulate initiation by inhibiting cooperative binding of DnaA at oriC; these include YabA (Merrikh and Grossman, 2011; Scholefield and Murray, 2013), DnaD (Bonilla and Grossman, 2012; Scholefield and Murray, 2013), and Soj (Scholefield et al., 2012). In several systems, orisome assembly is also controlled by positive regulators that increase DnaA binding to low affinity sites. In E. coli and Caulobacter crescentus, low affinity DnaA binding is stimulated by the DNA bending protein IHF (Grimwade et al., 2000; Siam et al., 2003). Additionally, the E. coli DiaA protein (Ishida et al., 2004), and its homolog in H. pylori, HobA (Natrajan et al., 2007; Zawilak-Pawlik et al., 2007, 2011), bind to DnaA’s domain I and increase weak site occupation.

The studies described above suggest that several different regions of DnaA could be targeted to inhibit DnaA binding. Obviously, blocking the DNA binding domain (domain IV) should inhibit all stages of orisome formation. Although protein–DNA interactions have not traditionally been considered to be “druggable” targets, recent studies have reported success in identifying inhibitors of DNA binding (Huang et al., 2016; Grimley et al., 2017). Further, inhibition of the self-oligomerization regions of DnaA in domains I and III should prevent cooperative binding and thus assembly stages 2 and 3 (Kawakami et al., 2005; Miller et al., 2009; Duderstadt et al., 2010; Scholefield and Murray, 2013). Like protein–DNA interactions, protein–protein interactions have not traditionally been favored as drug targets, but recent reports raise optimism that targeting DnaA oligomerization could be successful (Marceau et al., 2013; Voter et al., 2017).

Several other must be resolved before inhibition of DnaA’s access to oriC can be determined to be a practical antimicrobial strategy. First, it is not yet clear how much binding must be prevented to inhibit replication. All origins contain multiple DnaA binding sites (Leonard and Mechali, 2013), and studies that removed or inactivated DnaA recognition sites in E. coli chromosomal oriC revealed a tremendous plasticity in orisome assembly. Remarkably, deletion of the entire right region of oriC is tolerated in slow growing cells (Stepankiw et al., 2009). Additionally, directed mutations that knocked out binding to individual chromosomal oriC sites had little effect on viability (Weigel et al., 2001; Riber et al., 2009; Kaur et al., 2014). However, eliminating binding to more than one high affinity site did cause loss of viability (Kaur et al., 2014). Similar plasticity was noted in SeqA regulation of the number of occupied DnaA sites; even though loss of SeqA binding would be expected to allow DnaA re-binding at some oriC sites after initiation, mutating individual GATCs had little effect on initiation synchrony (Jha and Chattoraj, 2016). In Bacillus, some individual chromosomal oriC DnaA binding sites were shown to be essential, but others were not (Richardson et al., 2016). These studies, although by no means comprehensive, suggest that any pharmacological strategy should aim to inhibit DnaA binding at a majority of oriC sites, at least until future orisome studies reveal which sites are needed to assemble sub-complexes that carry out the essential mechanical functions of origin activation. Additionally, several studies suggest that assays used to screen for inhibitors of DnaA binding should be based on inhibiting chromosomal oriC rather than cloned origins, since inactivating individual sites is much more detrimental to plasmid oriC function (Weigel et al., 2001). Also, given the diversity in bacterial origin configurations (Leonard and Mechali, 2013), screens using a single bacterial type might not be sufficient to identify agents that act against a broad spectrum of bacteria. It might be necessary to utilize multiple types of bacteria, unless methodology is developed that allows the function of any chromosomal origin to be examined in an easily cultured strain. One strategy, involving heterologous origin transplantation, was described in a recent review (Leonard and Grimwade, 2015).

Based on seminal studies of in vitro E. coli DNA replication by the Kornberg lab (Sekimizu et al., 1987), DnaA-ATP is the active initiator form, and it is widely accepted that all bacteria share the requirement for DnaA-ATP in origin activation. In E. coli, DnaA-ATP levels are tightly regulated during the cell cycle to ensure precise initiation timing. Prior to the initiation step, DnaA-ATP levels rise due to new synthesis and a combination of recharging systems that include the DARS loci and acidic phospholipids in the membrane, reviewed in Skarstad and Katayama (2013). After initiation, the synthesis of DnaA-ATP is repressed for 1/3 of the cell cycle by SeqA, which binds to hemi-methylated GATC motifs in the dnaA promoter (Campbell and Kleckner, 1990). To inactivate DnaA-ATP, DnaA’s intrinsic ATPase activity is stimulated by the Hda protein associated with the β-clamp (DnaN) (Su’etsugu et al., 2004; Kim et al., 2017). Excess DnaA-ATP can also bind to a high capacity locus, termed datA (Kitagawa et al., 1998), which also stimulates DnaA-ATP hydrolysis (Kasho and Katayama, 2013).

The critical importance of mechanisms regulating DnaA-ATP levels in E. coli is demonstrated by the lethality observed in mutants, such as dnaA(cos) and hda null, that have lost the ability to inactivate DnaA-ATP by hydrolysis (Nishida et al., 2002; Felczak and Kaguni, 2009). DnaA(cos) carries two amino acid substitutions, one that prevents nucleotide binding (A184V), and another that stabilizes the mutated form (Y271H) (Simmons and Kaguni, 2003). Cells harboring dnaA(cos) are non-viable at 30°C, most likely due to over-initiation that results in co-directional replication fork collisions at stalled forks, leading to catastrophic double-stranded breaks (Simmons et al., 2004). A similar lethal phenotype is seen when Hda is inactivated, unless suppressor mutations arise (Riber et al., 2009; Charbon et al., 2011). Interestingly, although diverse suppressor mutations have been identified (Charbon et al., 2011), they all seem to cause tolerance of over-initiation by decreasing the chance of fork collisions, either by reducing initiation frequencies, or by preventing DNA lesions, such as oxidative DNA damage, that would slow forks (Charbon et al., 2014, 2017).

There are several aspects of DnaA inactivation mutants that are relevant to identifying antibiotic targets. First, lethality is caused by increasing, rather than decreasing the initiation frequency (Simmons et al., 2004). The run-away replication observed in DnaA(cos) mutants correlates with the inability to bind adenine nucleotide (Simmons et al., 2003), although it is not clear why loss of nucleotide binding leads to over-replication rather than orisome inactivation. Second, it is not yet known how many other bacterial types use regulation of DnaA-ATP levels as a regulatory mechanism. While some bacteria, such as Caulobacter and most enterobacteria, appear to have homologs of hda (Wargachuk and Marczynski, 2015), others, such as Bacillus, Staphylococcus, and H. pylori, do not (Katayama et al., 2010). DnaA in B. subtilis and S. aureus exchange bound ADP for ADP much more rapidly than E. coli does (Kurokawa et al., 2009; Bonilla and Grossman, 2012), and negative regulation of orisomes in these bacteria is focused mainly on DnaA-DNA interactions. Thus, screens to identify stimulators of DnaA hydrolysis may be ineffective in identifying broad-spectrum antimicrobials. In contrast, the growth inhibition/lethality caused when DnaA can’t hydrolyze ATP suggests that identification of inhibitors of ATP binding or ATPase activity, causing lethality by over-initiation, may be more fruitful. While targeting of the ATPase of AAA+ proteins is still in its infancy there are reports of successful inhibition of this protein class (Chou et al., 2011; Firestone et al., 2012). Targeting of DnaA’s ATPase, however, could generate suppressor mutations that reduce fork collisions (Charbon et al., 2017) and thus be prone to rapid resistance development. Possibly, this could be resolved by combination with an agent that inhibits DNA repair to counteract the actions of suppressor mutations (Simmons et al., 2004; Sutera and Lovett, 2006).