Homodimerization and Heterodimerization Requirements of Acinetobacter baumannii SOS Response Coregulators UmuDAb and DdrR revealed by Two-Hybrid Analyses.

The multi-drug resistant pathogen Acinetobacter baumannii displays unusual control of its SOS mutagenesis genes, as it does not encode a LexA repressor, but instead employs the UmuDAb repressor and a small DdrR protein that is uniquely found in Acinetobacter species. We used bacterial adenylate cyclase two-hybrid analyses to determine if UmuDAb and DdrR coregulation might involve physical interactions. Neither quantitative nor qualitative assays showed UmuDAb interaction with DdrR. DdrR hybrid proteins, however, demonstrated modest head-to-tail interactions in a qualitative assay. The similarity of UmuDAb to the homodimer-forming polymerase manager UmuD and LexA repressor proteins suggested that it may form dimers, which we observed. UmuDAb homodimerization required a free C-terminus, and either small truncations or addition of a histidine tag at the C-terminus abolished this homodimerization. Amino acid N100, crucial for UmuD dimer formation, was dispensable if both C-termini were free to interact. However, mutation of G124, necessary for LexA dimerization, yielded significantly less UmuDAb dimerization, even if both C-termini were free. This suggests that UmuDAb forms dimers like LexA, but may not co-regulate gene expression involving a physical association with DdrR. The homodimerization of these coregulators provides insight into a LexA-independent, coregulatory process of controlling a conserved bacterial action such as the mutagenic DNA damage response.


Introduction
The bacterial SOS response is carried out by a network of genes that enables bacteria to survive DNA damage but is also associated with mutagenesis (Sutton et al. 2000). In the repressor (Fornelos et al. 2015;Caveney et al. 2019). This example suggests the hypothesis that UmuDAb and DdrR coregulate gene expression through physical interaction, which we sought to test in this study. Furthermore, although UmuDAb is known to bind to an Acinetobacter-specific SOS box (Aranda et al. 2013), it is not known if it is capable of forming homodimers, like the LexA and UmuD proteins it is similar to (Hare et al. 2012a). LexA and UmuD self-associate via different parts of these proteins (Giese et al. 2008;Ollivierre et al. 2011), so testing whether UmuDAb behaves more like LexA would help indicate whether its repressive actions might include a role for DdrR, similar to the LexA-gp7 interaction. We therefore used a bacterial two-hybrid system to test for UmuDAb-DdrR heterodimer and homodimer associations. This assay also facilitated the mutation of these proteins and their requirements for interaction. Our findings did not detect direct interactions between UmuDAb and DdrR hybrid proteins but found that UmuDAb forms homodimers with requirements like those of LexA. That this LexA-like UmuDAb repressor only functions in the presence of a chromosomally encoded coregulator, DdrR, suggests additional LexA-independent methods of controlling the mutagenic SOS response beyond those previously identified for bacteria.

Bacterial strains, plasmids, and growth conditions
Acinetobacter baumannii ATCC 17978 was grown in Miller LB broth (Fisher Scientific) or on LB agar plates at 37 °C. Escherichia coli DH5α and XL1-Blue strains were grown overnight at 37 °C in LB with either 100 μg/mL ampicillin (for pUT18 and pUT18C plasmids) or 50 μg/mL kanamycin (for pKT25 and pKNT25 plasmids). The E. coli BTH101 reporter cells containing both bait and prey vector plasmids were grown at 30 °C overnight in LB broth containing both antibiotics and 0.5 mM IPTG to induce protein production from all pUT18(C) and pK(N)T plasmids, or on BD Difco MacConkey agar containing both antibiotics, 0.5 mM IPTG, and 1% maltose.

Recombinant plasmid construction
Primers to amplify ddrR and umuDAb from A. baumannii ATCC 17978 genomic DNA were designed using Primer3 (Rozen and Skaletsky 1999) and are described in Table 1. We performed PCR using PHUSION II polymerase according to its protocol (Thermofisher Scientific). After digestion with the restriction enzymes PstI HF and BamHI HF (New England Biolabs), the PCR products were cloned into the pUT18c, pUT18, pKT25, and pKNT25 vectors provided by the BACTH system (Euromedex, France). This produced recombinant plasmids encoding hybrid proteins consisting of UmuDAb or DdrR fused to the 18 or 25 kD adenylate cyclase (AC) domains (AC 18 or AC 25 denotes the 18 kD or 25 kD domains of AC, respectively). Gene fidelity and frame was confirmed by DNA sequencing. The resulting gene/enzyme subunit plasmid orientations are listed in Table 2 and Fig. 1. To test for hybrid protein production from E. coli AB1157, BTH101, and DH5α cells SDS-PAGE gel, then Western-blotted with antibodies targeting the AC18 subunit (Millipore-Novagen) and the C-terminal end of UmuDAb (Hare et al. 2012a).

Two-hybrid analysis
To provide a prokaryotic cell system that lacked UmuDAb and DdrR proteins in which to test for UmuDAb and DdrR hybrid protein associations, we used the Euromedex BACTH (Bacterial Adenylate Cyclase Two-Hybrid) System Kit. This kit uses E. coli BTH101 cells, which were transformed with five ng of each of two plasmids in each plasmid pair described in Table 3 (for wild-type UmuDAb or DdrR proteins) and Table 4 (for mutant UmuDAb combinations), plated to MacConkey-maltose agar plates, and grown at 30°C overnight. For all results presented here, the next day, four to six biological independent cultures were started with isolated colonies from the transformation and grown with shaking at 30°C overnight. As per Battesti and Bouveret (2012), each culture was inoculated with several random colonies from a transformation plate. A beta-galactosidase activity assay was performed on the IPTG-induced overnight cultures to measure the Cya+ phenotype that results from hybrid protein interaction, as described in the BACTH system manual Annex II and reviewed in Battesti and Bouveret (2012). For a qualitative colorimetric assay, two μL of each culture was placed onto a MacConkey-maltose reporter plate and incubated at 30°C overnight. The degree of redness in the spot growth indicated the degree of hybrid protein interaction. All plate images had brightness increased 20% in PowerPoint for clarity.

UmuDAb mutant construction
We made two C-terminal truncations of the umuDAb gene A1S_1389 to encode UmuDAb W192X G193X and UmuDAb R201X, and two amino acid substitutions in the hypothesized dimerization interface of UmuDAb to encode UmuDAb N100D and UmuDAb G124D. Mutations were made using the Quick Change II kit (Agilent Technologies) with the recombinant UmuDAb BACTH plasmids as templates. Primers (Table 1) were designed with the online Quick-Change Primer Design tool (Agilent). The primers for UmuDAb W192X G193X simultaneously introduced two base-pair changes, modifying the codons for W192 and G193 to stop codons. For simplicity, this mutant will be henceforth referred to as UmuDAb W192X.

I-TASSER modeling
I-TASSER (Iterative Threading ASSEmbly Refinement) is a hierarchical method for protein structure and function prediction that identifies structural templates from the PDB by a multiple threading approach (Zhang 2008;Roy et al. 2010;Yang and Zhang 2015). It iteratively constructs full-length atomic models through simulations. The modeling was performed by submitting the A. baumannii A1S_1389 (UmuDAb) and A1S_1388 (DdrR) protein sequences to the online server accessible at http://zhanglab.ccmb.med.umich.edu/I-TASSER/. Images in this manuscript were rendered using EzMol Webserver to visualize pdb files attained from iTASSER modeling (Reynolds et al. 2018).

Statistical analyses
Statistics were calculated with GraphPad Prism software. Miller units were normalized using a natural log transformation before a one-way ANOVA analysis. Dunnett's multiple comparisons post-test was performed to compare values from samples expressing hybrid proteins to the negative control; Tukey's multiple comparison post-test was used to compare values from samples containing either wild type or mutant hybrid proteins to each other, with significance for both at the P < 0.05 level chosen. All statistical analyses included the positive and negative controls performed in the same experiments as the data presented in each figure.

Predicted structures of UmuDAb and DdrR
The amino acid sequence of UmuDAb (A1S_1389) was submitted to I-TASSER (Yang and Zhang 2015) to predict its structure. It was predicted to be most similar, in a TM-align alignment (Zhang and Skolnick 2005) to the structure of LexA from Thermotoga maritima (Zhang 2009;Witkowski et al. 2016) (Fig. 2a). Both the confidence (C) score of 0.05 and the structural similarity score (TM) of 0.72 for UmuDAb, whose structure has not been experimentally determined, was similar to that of the well-studied and crystallized LexA from E. coli (1.46 and 0.92, respectively). The T. maritima LexA, like other LexA proteins, forms dimers via interactions of its C terminus. The 3D models' similarity in the C-terminal regions of UmuDAb and LexA led us to hypothesize that if UmuDAb formed homodimers, the C-terminus of UmuDAb would be required for this interaction. Additionally, as UmuDAb shares a similar function to LexA (Hare et al. 2006;Aranda et al. 2013), we proposed that the amino acid G124 would also be required for UmuDAb homodimerization. Alternately, the higher percent identity of UmuDAb to UmuD than LexA (Hare et al. 2006) suggested that the UmuDAb N100 (N41 in UmuD) could be required for self-association.
We also submitted the amino acid sequence of DdrR (A1S_1388) to I-TASSER for structure prediction (Yang and Zhang 2015). The C-terminal 36 amino acids of the DdrR model contained two alpha-helices of 15 and 14 amino acids separated by a beta strand of five amino acids (Fig. 2b). This sized structure is similar to the N-terminal 36 amino acids of gp7 (7 kD), which contains two 16-and 12-amino acid alpha helices separated by a three amino acid beta strand (shown in the crystal structure 6N7O in the RCSB Protein Data Bank (Caveney et al. 2019)). Although these I-TASSER models both showed a similar structure, none of the PDB hits of threading templates were shared for these two proteins, i.e., DdrR was not directly modeled on gp7. Neither of these proteins' (or the Acinetobacter baylyi strain ADP1 DdrR) C scores were very high (gp7: −1.47; DdrR: −2.62 and −3.42 from 17978 and ADP1, respectively). The TM scores, however, for all three proteins ranged from 0.34 -0.53 (gp7), suggesting the likelihood of a similar structure. (A TM score < 0.17 suggests random similarity, and > 0.5 suggests correct topology.) The prediction of a similar structure, nevertheless, may reflect a conserved function. Since two gp7 homodimers interact with one LexA dimer to enhance DNA binding (Fornelos et al. 2015), we speculated that DdrR associates with itself, and subsequently UmuDAb, in a similar fashion.

Hybrid UmuDAb-AC does not directly interact with Hybrid DdrR-AC protein
We used an E. coli bacterial-based adenylate cyclase two-hybrid system (BACTH), available as a standardized kit from Euromedex, to probe possible A. baumannii protein interactions.
As neither the UmuDAb and DdrR regulators, nor the UmuDAb SOS binding site (nor gp7) are encoded or used by E. coli strains, this decreased the possibility that overexpression of these regulators would lead to abnormal or toxic changes in E. coli gene expression.
Furthermore, testing in a heterologous system provides the advantage of a more direct test of the interactions between the two test partners, which rules out requirements for other A. baumannii proteins (Karimova et al. 1998(Karimova et al. , 2000Battesti and Bouveret 2012). We used this system to investigate whether UmuDAb interacts with DdrR and whether UmuDAb or DdrR interact with themselves. The BACTH vectors pUT18 and pUT18C encode the 18 kD AC domain, and pKT25 and pKNT25 encode the 25 kD AC domain. Cloning the A. baumannii umuDAb or ddrR genes into each of these four vectors yielded eight plasmids. Each plasmid encoded a hybrid protein with either the N-or C-terminus of UmuDAb (or DdrR) fused inframe to an 18 or 25 kD AC domain (See Fig. 1 and Table 2). We transformed pairs of these hybrid plasmids into E. coli BTH101 cells (Table 3). Interactions between the two hybrid proteins generate a Cya+ phenotype that we quantified (in Miller units) and observed qualitatively as color on MacConkey-maltose plates. The advantage of this method is in allowing analyses by both the colorimetric assay, which is very sensitive at detecting interactions in this assay, as well as the quantifiable beta-galactosidase assays that can show statistically meaningful differences between hybrid protein configurations as well as wildtype and mutant hybrid proteins. The observations of the greater sensitivity seen in the colorimetric assay is consistent with previous work by Battesti and Bouveret, 2012. UmuDAb represses SOS genes like LexA (Norton et al. 2013;Aranda et al. 2013;Hare et al. 2014). Based on interactions between the small protein gp7and LexA that facilitate LexA-DNA binding (Caveney et al., 2019), we hypothesized that the 9 kD DdrR interacts with UmuDAb in a similar manner. We tested all eight possible fusion protein configuration pairings for interaction between hybrid UmuDAb-AC and DdrR-AC proteins ( Table 2). None of these eight hybrid protein pairs showed a significant increase from a negative control pair of plasmids consisting of empty pKT25 and pUT18 vectors (Fig. 3a), and grew to similar culture densities, suggesting that expression did not cause toxicity that compromised cell growth. The same cultures measured with the beta-galactosidase assay were spotted to MacConkey-maltose reporter plates, and also showed no visible indication of interaction between the UmuDAb and DdrR AC-fusion proteins (Fig. 3b). Furthermore, we conducted multiple pull-down assays in which His-tagged UmuDAb or His-tagged DdrR was the bait and the other possible binding partner was the prey. All of these failed to show binding of the prey protein to the bait in eluates.

DdrR self-interactions depend on head-to-tail orientation
We tested whether a hybrid DdrR-AC interacted with another DdrR-AC protein, possibly in a manner reminiscent of how RecA forms filaments (Stasiak and Egelman 1986) or how the bacteriophage GIL01 gp7 protein forms dimeric four-helix bundles to mediate LexA-DNA interactions in Bacillus thuringiensis (Caveney et al. 2019). The four possible pairs of DdrR hybrid proteins were tested for DdrR self-interaction. In one pair, the DdrR N-termini were unbound, as the C-termini of both DdrR proteins were bound to an AC domain (termed NN; see Fig. 1). Another pair had their C-termini unbound and both N-termini of DdrR bound to an AC domain (termed CC). The remaining two pairings (NC and CN) had one N-terminus bound on one DdrR-AC, and one C-terminus bound on the other DdrR-AC (with two different configurations possible, depending on whether the N or C end of DdrR was fused to the 18 vs. the 25 kD AC domain).
Quantitative beta-galactosidase assay data from these four pairings showed no interaction (no statistical difference from the negative controls; Fig. 4a). However, there was a modest but consistent appearance of red-colored bacterial growth for two of the hybrid protein pairs on MacConkey-maltose, in two independent transformations (Fig. 4b). These positive interactions between DdrR hybrid proteins were only observed when opposite termini were available for each protein, i.e., in the NC and CN orientation pairs. There was no color change for the pairings with the same DdrR terminus free (NN or CC). These consistent results for the four different configurations of pairings suggest that DdrR proteins associate in a mirror-image arrangement of DdrR monomers.

UmuDAb self interactions are stronger when its C-terminus is unbound to adenylate cyclase domain
UmuDAb may have a LexA-like mechanism of action. However, previous observations have not determined whether this repressor, which has similarities to both UmuD and LexA, forms dimers, such as are important in LexA-like repressor function. To test whether UmuDAb binds to itself to form dimers, we tested four different hybrid protein pairs in the two-hybrid assay, similar to those used to test DdrR-DdrR interactions. All UmuDAb-UmuDAb hybrid protein configurations (CC, NC, and CN) except one (NN) showed significant interactions when compared to the negative control (Fig. 5a), suggesting that UmuDAb binds to itself. These three configurations all have in common at least one UmuDAb hybrid protein that contains the C-terminus of UmuDAb unbound to an AC domain. The highest beta-galactosidase activity was seen in the CC and NC configurations, with no significant difference in their activity (Fig. 5a). However, the CN configuration (see Fig. 1) yielded significantly less activity than the NC (and CC) configurations, even though both NC and CN contained one UmuDAb with its C-terminus unbound to an AC domain. These pairs differed only in which AC domain (the 18 kD or the 25 kD) was bound to the UmuDAb with its C-terminus free. Analysis of these data suggests that the association of UmuDAb with itself is improved when the 25 kD AC domain, not the 18 kD AC domain, is fused to the N-terminus of UmuDAb. The relative differences in the CC, NC, CN, and NN pairs' interactions were also clearly seen in the qualitative results seen on the MacConkeymaltose agar plates (Fig. 5b).

UmuDAb mutations that reduce self interactions are located in the C-terminal domain and required for LexA dimerization
To investigate which domains or specific amino acids within UmuDAb were necessary to form dimers, we constructed hybrid proteins containing UmuDAb mutations known to affect the dimerization of UmuD (N41D (Peat et al. 1996;Ollivierre et al. 2011), which aligns with N100 of UmuDAb (Hare et al. 2006)), and LexA (G124D; (Giese et al. 2008)). These were tested in both the CC and NC hybrid protein pairs that yielded the highest and equal levels of beta-galactosidase activity. Mutations were made in either the "N", the "C", or both hybrid proteins in an NC or CC pair. Surprisingly, the effects of the N100D and G124D mutations tested were different in these two hybrid protein pairs, depending on whether the 18 kD AC domain was fused to the N-or C-terminus of UmuDAb (i.e., CC or NC, respectively) (Fig. 5c, 5d).
These assays testing the effect of the N100D and G124D mutations indicated that the N100 amino acid involved in UmuD dimerization was not required for UmuDAb-UmuDAb association if both C-termini of UmuDAb were available in the CC configuration (Figure 5a, 5c). There was no significant decrease in beta-galactosidase activity in a hybrid protein pair containing the N100D mutation in the C N C, CC N , or C N C N pairs relative to the wild type CC pair (Fig. 5a). However, in the NC configuration, where only one C-terminus of UmuDAb was available for interaction, the N100 amino acid was required for full UmuDAb-UmuDAb interactions. There was significantly less beta-galactosidase activity in all three of the hybrid protein pairs (N N C, NC N , and N N C N ) than in the wild type NC protein pair, and the N N C mutant pair did not show significant interaction at all (the N N C and N N C N mutant pairs did indicate interaction).
Similar experiments testing the ability of UmuDAb G124D mutant proteins to form dimers showed that G124, which is influential in forming LexA dimers (Giese et al. 2008), was required to form UmuDAb-UmuDAb dimers in both the CC and NC configurations. The C G C and CC G mutant hybrid protein activity was significantly lower than the wild type CC activity, although both pairs showed interaction. Furthermore, when these two mutant proteins were combined in the C G C G pair, there was no significant interaction between these proteins. The G124D mutation had even more detrimental effects when present in the NC configuration, as a single mutation in the NC G pair was sufficient to reduce betagalactosidase activity levels to those of the negative control. Similar qualitative results were seen when cultures harboring either N100D or G124D mutant UmuDAb proteins were spotted to MacConkey-maltose reporter plates (Fig. 5d).
The C-terminal amino acids of LexA are also important for its dimerization, as shown by the crystal structure of its dimers (Zhang 2009). Therefore, we truncated the C-terminus of UmuDAb by eleven or three amino acids by constructing two different mutations to convert amino acids W192 or R201 of UmuDAb to stop codons, respectively. These C-terminal truncations negatively affected interactions in all three tested configurations (CC, NC, and CN (data not shown)). UmuDAb W192X and UmuDAb R201X abolished interactions in CC pair, seen in both the beta-galactosidase assay (Fig. 5a) and MacConkey-maltose plates (Fig.  5b). The same results were observed when the C terminus was disrupted, not by a truncation, but by a 6X His tag incorporated at the end of UmuDAb.
proteins. Because both UmuDAb and DdrR are required for repression of SOS genes in A. baumannii, but how they work together to achieve this repression is not known, we tested whether these proteins associated with each other. Additionally, the ability of UmuDAb to form dimers has not been demonstrated. However, the C-terminus of UmuDAb is similar to both the error-prone polymerase accessory UmuD as well as the LexA repressor (Hare et al. 2012a), each of which forms dimers via conserved C-terminal domains as part of their participation in the SOS response. Specifically, UmuDAb shares the same required active site and cleavage site residues with LexA and UmuD, and also contains the specific conserved amino acids required for both LexA and UmuD dimerization (Hare et al. 2012a). As an additional way to learn how the unusual UmuDAb regulator functions in this genus that lacks LexA, we also investigated whether UmuDAb-AC hybrid proteins could selfassociate.
In this study, using a two-hybrid system in a heterologous E. coli background, we observed significant evidence that the A. baumannii UmuDAb can form dimers in vivo. By constructing hybrid proteins where only one end (i.e., the amino-or carboxy-terminus) of UmuDAb was unbound by an AC subunit, we identified the C terminus of UmuDAb as crucial for UmuDAb dimer formation. We were thenable to easily modify the BACTH system plasmids to introduce mutations and truncations in the hybrid proteins. We constructed several mutant UmuDAb hybrid proteins containing changes in amino acids required by either the error-prone polymerase accessory UmuD or the LexA repressor. This helped determine which UmuDAb residues were required for, or involved in, this dimerization, and therefore, whether UmuDAb dimerization resembled either of these differently-acting proteins.
Comparing the results from mutating UmuDAb residues and using multiple hybrid protein pairs where both or only one UmuDAb possessed a free N or C terminus allowed us to dissect the effect of C-terminal interactions from specific amino acids previously known to affect UmuD or LexA dimerization. This resulted in a simple hierarchy of importance of these regions for UmuDAb dimerization, where C termini were more valuable in producing UmuDAb interactions than G124, which was more important than N100. It appeared that if both UmuDAb proteins contain an unbound C-terminus, any contribution of N100 in holding together the dimer, e.g., through hydrogen bonds such as function in UmuD homodimers (Ollivierre et al. 2011), was dispensable. However, in the absence of Cterminus/C-terminus interactions of UmuDAb, mutation of N100 to aspartic acid, which simultaneously disrupts hydrogen bonding and produces destabilizing electrostatic repulsion in UmuD (Ollivierre et al. 2011), disrupted the more tenuous UmuDAb-UmuDAb interactions. A more important requirement was observed for the G124 residue of UmuDAb, which in LexA helps form dimers (Giese et al. 2008). This residue, too, was less crucial to UmuDAb dimerization than having two UmuDAb C-termini free, because a mutation in either one of the UmuDAb hybrid proteins reduced, but did not abolish, UmuDAb interactions when both UmuDAb C-termini were free. However, if only one UmuDAb possessed a free C-terminus (the NC configuration), a single G124D mutation was sufficient to abolish UmuDAb-UmuDAb interactions. This shows a complex set of contacts within the UmuDAb dimer that could be investigated further with purified UmuDAb, outside of the context of the two-hybrid system, such as with native gel electrophoresis.
Our extensive pairing of different AC domains to each possible terminus of UmuDAb and DdrR also showed that the particular AC domain fused to a hybrid protein influenced that protein's interactions with a partner. This was indicated by the UmuDAb-UmuDAb interactions ( Fig. 5a-b), where the NC and CN configurations contained only one unbound UmuDAb C-terminus (Fig. 1), but NC yielded significantly higher amounts of betagalactosidase activity than CN. Comparison to the pairs displaying the highest activity (CC and NC), which both contained the p25U plasmid, suggests that having AC25 rather than AC18 fused to the N terminus of UmuDAb allows for better interaction and higher activity. The physical orientation or configuration of the AC 18 domain may have prevented efficient UmuDAb-UmuDAb interaction. Alternately, the stoichiometry of configurations may also affect interactions, as the BACTH vectors have different copy numbers (p18 plasmids are high copy; p25 plasmids are low copy).
Surprisingly, although DdrR co-regulates SOS genes with UmuDAb (Peterson et al. 2020), we did not observe direct interaction between UmuDAb and DdrR with this assay. We do not think that either protein's expression was absent, as Western blots verified the DdrR-AC and UmuDAb-AC hybrid proteins' production (Supplemental figure S1). Additionally, positive interactions were observed in other plasmid pairings (e.g., the UD-UD pairs), containing these same plasmid preparations, transformed on the same days, into the same cell strains. Nor do we think that overexpression may have led to toxicity, as cells co-transformed with both DdrR and UmuDAb hybrid proteins grew to similar culture densities as the UmuDAb-UmuDAb pairs and negative controls. We also think it unlikely that intrinsic lexA and recA expression in the E. coli BTH101 cells interfered with UmuDAb and DdrR interactions, because under the conditions of the assay (lack of DNA damage), lexA and recA are expressed at basal (uninduced) levels, and RecA does not associate with LexA (Little 1984(Little , 1991. We also think it improbable that LexA would interfere specifically with UmuDAb-DdrR interactions but not UmuDAb-UmuDAb interactions. Our observation of strong UmuDAb-UmuDAb binding interactions show that the uninduced amounts of LexA in the E. coli BACTH test strain did not prevent this interaction from occurring. Several possibilities exist for the lack of interaction we observed. First, the qualitative DdrR-DdrR interactions we observed required the N-and C-terminus of different DdrR fusion proteins to be free, i.e., unbound to AC domains. If pre-existing DdrR dimers are important in forming an association with UmuDAb, our BACTH system conditions, which contained only one form of a DdrR-AC fusion protein (i.e., with either the N-or C-terminus free, but not a combination of both) to interact with UmuDAb, would have prevented observing a UmuDAb-DdrR interaction. Furthermore, this possibility is consistent with the crystal structure of the scaffold protein gp7 forming similar dimers before its association with LexA, where the N terminus of one monomer associates with the C terminus of the other monomer in a dimeric four-helix bundle facilitated by interactions between beta-strands of each gp7 monomer (Caveney et al. 2019). As two similarly sized and organized helices and beta-strands were also predicted with TM-align modeling (Zhang and Skolnick 2005) to exist in a similar configuration in the C-terminus of DdrR (Fig. 2b), this possibility will be tested for DdrR-DdrR associations, using purified proteins lacking fusion to AC domains. Sensitive methods such as Western analyses of DdrR-DdrR dimers in native gels can be used to identify the weak interactions suggested by the qualitative MacConkey indicator plates.
UmuDAb and DdrR may indeed physically interact but be prevented from doing so by the AC domains fused to them in this assay. This is consistent with the results of Supplemental Fig S1, where UmuDAb was not seen in a Western blot when fused to AC25. However, this hybrid protein (25U) was present in the protein pair that yielded the strongest positive signal in the beta-galactosidase assay, which would not have occurred in the absence of hybrid protein expression. Hybridization with AC domains can cause proteins to be unstable or poorly folded; this can make the interaction between these proteins difficult to identify (Battesti and Bouveret 2012). Alternately, UmuDAb and DdrR may have transient interactions that were not stable enough to initiate the signaling feedback cascade for positive interaction results. Finally, these proteins may not interact directly with each other, but through an intermediate present in the A. baumannii proteome but not in the E. coli strain used for the BACTH analyses, or they may coregulate target genes without physical interaction.
This study suggests that the A. baumannii repressor UmuDAb forms dimers in a manner more similar to LexA than UmuD (Hare et al. 2006(Hare et al. , 2012a, which helps describe the actions of this LexA-like repressor and its corepressor that are unique to this genus. This UmuDAb-DdrR coregulatory system also shares some similarities to the several bacterial systems in which a corepressor (either horizontally acquired or native to the bacterial host) has evolved to use the chromosomal LexA protein to corepress target genes found in mobile genetic elements (plasmids or prophages) (Fornelos et al. 2016). It is striking that similar processes of coregulation of an SOS repressor by a small protein are conserved across such different evolutionary lineages, involving ddrR in the chromosome of the Gram-negative Acinetobacter genus vs. gp7, residing in a prophage of the Gram-positive Bacillus thuringiensis (Verheust et al. 2003;Fornelos et al. 2015). However, the Acinetobacter SOS response system also displays several significant differences from these systems. First, neither umuDAb nor ddrR show evidence of being part of a prophage in any A. baumannii strain (Di Nocera et al. 2011), nor of being horizontally acquired (Hare et al. 2012b). This is in contrast to the corepressors encoded by plasmids in E. coli (Butala et al. 2012;Kamenšek et al. 2015) or bacteriophages in Vibrio (Quinones et al. 2005). Secondly, although the chromosomal LexA of B. thuringiensis interacts with gp7 to repress prophage genes and establish a lysogenic state, UmuDAb and DdrR regulate both non-mobile element genes (umuC, A1S_2008; umuDAb/ddrR, A1S_1389/1388) as well as horizontally-acquired error prone polymerases (A1S_0636/0637 on pAB3, and prophage genes A1S_1174/1174 and A1S_2015) ( (Hare et al. 2014;Peterson et al. 2020). Finally, the UmuDAb repressor does not function as a repressor in the absence of its corepressor, DdrR (Peterson et al. 2020), unlike these systems of coregulated repressors in other species.
This work helps develop our understanding of a LexA-independent method of controlling the mutagenic SOS response beyond those previously identified for bacteria (Davis et al. 2002;Campoy et al. 2003). Learning how UmuDAb regulates the SOS DNA damage response in the absence of LexA can lead to treatments needed to prevent A. baumannii from gaining new resistance to clinically used antibiotics after DNA damage (Norton et al. 2013;Jara et al. 2015). In other species, LexA appears to be a non-traditional, yet promising drug target (Mo et al. 2018). This reinforces the need to consider the SOS repressor UmuDAb, and its corepressor DdrR, as logical targets for therapeutics to reduce or eliminate SOS repair and mutagenesis needed for cell survival. The increasing occurrence of multidrug resistance to such drugs as carbapenem has moved A. baumannii from being a 'serious' to an 'urgent threat' to human health, according to the Centers for Disease Control (Centers for Disease Control and Prevention 2019). This lends importance to developing more effective ways to combat this pathogen and identify parallel SOS response genes to target in other emerging pathogens that do not have LexA as their SOS response regulator.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material. Explanation of name codes used to refer to hybrid protein pairs used in BACTH analyses. AC18 = the 18 kD domain of adenylate cyclase contained in the vectors pUT18 and pUT18C. AC25 = the 25 kD domain of adenylate cyclase contained in the vectors pKT25 and pKNT25. When describing a pair of hybrid proteins, the protein containing the 18 kD AC domain is listed first, followed by the protein containing the 25 kD AC domain. The "C" or "N" part of the hybrid protein pair name refers to which end of UmuDAb or DdrR is free (unbound to an AC domain). The subscripts "U" or "D" refer to UmuDAb or DdrR. Hence the UmuDAb-DdrR hybrid protein pair called N U C D consists of the C-terminus of UmuDAb fused to the 18 kD AC domain (plasmid pUD18), and the N-terminus of DdrR fused to the 25 kD AC domain (plasmid p25DR), both transformed into the same E. coli BTH101 cells for subsequent two-hybrid analysis. I-TASSER models of A. baumannii ATCC 17978 proteins (a) UmuDAb, showing Cterminal domain S119 and K156 serine protease activity-conferring amino acids and A83/G84 cleavage site amino acids, for comparison to UmuD and LexA, which share the same motifs. Boxed amino acid names denote potential dimerization interface amino acids N100 and G124 and the W192 and R201 amino acids where C-terminal truncation mutations were constructed for subsequent two-hybrid system analysis. This model had a C-score of 0.05 (scores typically range from −5 to 2, with higher scores indicating a higher confidence model). The TM score was 0.72, where a score >0.5 indicates a model with correct topology, and the RMSD was 5.3 ± 3.4 A. (b) DdrR, showing a two-helix structure. This model had a C-score of −3.42 and a TM score of 0.34, where a score >0.5 indicates a model with correct topology, and <0.17 indicates random similarity. The RMSD was 11.0 ± 4.6 Å. Two-hybrid assay analyses do not show evidence of association between UmuDAb and DdrR hybrid proteins. Assays were performed on a minimum of six independent biological replicates, with each culture started from a minimum of three first-day transformation colonies to reduce the heterogeneity of results (Battesti and Bouveret 2012). Hybrid protein pairs tested in the assay are denoted by names that begin with either the N-or C-terminus that is free, of the protein (subscript U for UmuDAb; D for DdrR) fused to the 18 kD AC domain, followed similarly by the protein fused to the 25 kD AC domain; see Table 3. (a) Quantitative beta-galactosidase assay results from a minimum of six independent cultures graphed as a box and Tukey whisker plot. One-way ANOVA was performed on the logtransformed Miller units with Dunnett's post-test for samples showing interaction as those significantly different (p < 0.05) from the negative control ("neg"). Only the positive control ("pos") was significantly greater than the negative control. (b) Qualitative results from two μL of independent cultures of each strain containing a pair of hybrid proteins, spotted to MacConkey-maltose agar plates, and incubated overnight. Cells containing the pair of positive control plasmids pKT25-zip and pUT18C-zip ("+") resulted in a positive, redcolored colony. Cells containing the pair of negative control vectors pKT25 and pUT18C ("− ") resulted in a negative result of a pale white colony. Two-hybrid analyses show a modest association between DdrR hybrid proteins. NN = both hybrid DdrR-AC proteins have the N-terminus of DdrR free (unbound to AC); CC = both hybrid DdrR-AC proteins have the C-terminus of DdrR free; NC = N-terminus of DdrR fused to AC18 is free, C-terminus of DdrR fused to AC25 is free, and CN is the opposite configuration of NC. (a) Beta-galactosidase assay results from 8-9 independent cultures from two different transformation experiments, graphed as box and Tukey whisker plot. One-way ANOVA was performed on the log-transformed Miller units with Dunnett's posttest for samples showing interaction as those significantly different (p < 0.05) from the negative control ("neg"). Only the positive control ("pos") was significantly greater than the negative control. (b) Qualitative results from two μL of independent cultures of each strain spotted to MacConkey-maltose agar plates and incubated overnight. Red color in colonies shows the modest association between DdrR hybrid proteins when opposite DdrR termini are unbound to AC domains (NC and CN configurations). There is no indication of association when fused in the same orientation (NN and CC). Two-hybrid analyses indicate interaction between UmuDAb hybrid proteins that depends on C-terminal interactions. (a) Beta-galactosidase assay results from 9-14 independent cultures, graphed as box and Tukey whisker plot. Hybrid UmuDAb protein pairs (see Table 3 or Figure 4 for naming conventions) displaying interaction are designated by an asterisk. These were significantly different from negative controls using Dunnett's post-test following ANOVA analysis, with P < 0.05 chosen as the significance level. Pairs NC and CN were also significantly different from each other by Tukey's post-tests following ANOVA analysis, with P < 0.05. (b) Qualitative results from two μL of IPTG-induced overnight cultures plated on MacConkey-maltose agar plates, showing interaction between UmuDAb hybrid proteins. (c) Beta-galactosidase results from 6-16 independent cultures, graphed as box and Tukey whisker plot of mutant UmuDAb CC-or NC-paired hybrid proteins containing N100D, W192X, R201X, G124D, and C-terminal 6X His tag UmuDAb mutant forms. Hybrid protein pairs displaying interaction as determined by Dunnett's post-tests following ANOVA analysis are denoted with an asterisk. Table 4 contains a specific description of every mutant hybrid protein pair. Brackets denote mutant hybrid protein pairs that were significantly different from the wild-type CC or NC hybrid protein pairs, using Tukey's post-test following ANOVA. (d) Qualitative results from two μL of IPTG-induced overnight cultures on MacConkey-maltose agar plates. The subscript * in C*C, CC*, and C*C* is a place holder for UmuDAb mutant identification above each MacConkey plate image. Mutant names are color-coded to match their boxes plotted in the graph in (c).