Control of cell division by an Acinetobacter baumannii protein with a novel nucleotidyl-cyclase-like fold

Abstract

The multidrug-resistant sepsis pathogen Acinetobacter baumannii has diverged from model γ-proteobacteria in fundamental ways, hindering efforts to develop new lines of attack against the microbe. A major area of divergence is cell division. The pathogen lacks several widely conserved division enzymes, such as FtsEX, and instead possesses a suite of atypical gene products showing no sequence similarity to any well-characterized proteins. Key among these unusual proteins is AdvA. In previous Tn-seq studies, we identified AdvA as essential for A. baumannii division and fluoroquinolone resistance. The protein comprises an N-terminal transmembrane/periplasmic region connected to a C-terminal unannotated cytoplasmic domain. Interestingly, most advA transposon insertions were lethal unless they occurred within the linker between these two regions. The roles of AdvA in cell division and the basis for positional transposon effects were unclear. Here, we combine mutagenesis with fluorescence localization, two-hybrid, and structural analyses to investigate how AdvA domains function in assembling and activating the A. baumannii divisome. We show that AdvA depletion profoundly disrupts divisome construction at Z-rings. This dependence is based on numerous interactions by AdvA with divisome proteins, with its N-terminal region binding multiple components and the cytoplasmic domain binding one (the early protein ZipA). In addition, we identified substitutions in FtsB and FtsW that suppress AdvA essentiality, consistent with a role in divisome protein activation as well as recruitment. Finally, we determined the structure of the AdvA cytoplasmic domain, which revealed a novel 3D-fold resembling adenylyl/guanylyl cyclases with striking deviations. Most notably, AdvA lacks canonical catalytic and dimerization sites and contains unusual features, including a positively charged tip promoting fluoroquinolone resistance and an extra C-terminal helix essential to divisome interactions and cell division. The critical nature of the most C-terminal structure in AdvA helps explain the positional effects of transposon insertions and facilitated identification of a distant structural homolog in the pathogen Pseudomonas aeruginosa. These results illuminate a new type of control protein governing bacterial division that could be exploited for improved antimicrobial strategies targeting nosocomial infections.

Introduction

Acinetobacter baumannii is a Gram-negative γ-proteobacterium that causes a wide range of challenging multidrug-resistant (MDR) diseases in healthcare settings, commonly manifesting as bloodstream infections, pneumonia, wound infections, and sepsis [1]. These diseases affect the most critically ill patients and result in increased hospital stays, medical care costs, and risk of mortality [1]. A. baumannii resistance to formerly “last-line” antibiotics such as carbapenems is now widespread, leaving few effective treatment options [2]. This problem only worsened during the COVID-19 pandemic, which coincided with increased rates of hospital-acquired carbapenem-resistant infections with the bacterium [3, 4]. Due to its broad resistance and ability to spread, A. baumannii has been designated an urgent threat to public health and a top priority for research and development of new therapies by the Centers for Disease Control and the World Health Organization [5, 6]. Efforts to inhibit the pathogen will be boosted by a better understanding of its essential cell biology, including the fundamental process of bacterial division, which has the potential to provide multiple essential antimicrobial targets [7].

Our understanding of Gram-negative cell division has derived from research in Escherichia coli and other model organisms. The canonical pathway follows three main stages: (1) midcell placement of tubulin-like FtsZ filaments to form the Z-ring; (2) assembly at the Z-ring of the divisome, the dynamic multiprotein complex that carries out division; and (3) divisome-mediated constriction and septation [8, 9]. Transitions between these stages depend on the ABC transporter-like complex FtsEX that is conserved across a wide range of bacteria [10]. In stage 1, midcell-positioned Z-rings are stabilized and membrane-anchored by two proteins, ZipA and the actin homolog FtsA, and recruit FtsEX [11, 12]. Binding of FtsX to FtsA modulates FtsA oligomerization, likely creating the signal to begin stage 2, the ordered recruitment of additional divisome components [10, 13]. These include the chromosome translocase FtsK; the FtsQLB subcomplex which regulates septum synthesis; FtsW and FtsI peptidoglycan synthases; and FtsN, the activator of stage 3 septal wall synthesis. Commencing stage 3 also requires FtsEX, which activates peptidoglycan hydrolysis by amidases and their regulatory partners [10, 14]. Peptidoglycan processing likely triggers feedback loops involving FtsN that further activate septal wall synthesis by FtsWI, ultimately driving septation and cell separation [10].

While several aspects of the above division paradigm are likely to apply to A. baumannii, the pathogen shows notable differences, despite belonging to the same γ-proteobacterial class as E. coli. First, in A. baumannii, some core divisome proteins are either altogether lacking or show substantial evolutionary divergence from their E. coli counterparts. Most notably, A. baumannii has no orthologs of FtsEX [15, 16]. The pathogen also lacks homologs of several Z-ring regulators (ZapBCD, SlmA, SulA) as well as amidase AmiC [16]. Although a likely ortholog of FtsN exists in A. baumannii (ACX60_RS17385/GO593_07290), it shows extensive divergence from the E. coli prototype (16% identity [16]) and an alternative annotation for the A. baumannii gene (spor) has been proposed [17]. Moreover, A. baumannii versions of ZipA, FtsQ, and FtsL have greatly diverged (26-27% identity with E. coli orthologs) [16]. Second, A. baumannii encodes a number of atypical proteins that have no close similarity to any known protein in other organisms yet are linked to cell division [15, 18, 19]. These observations indicate that unique mechanisms have evolved to drive cell division in A. baumannii and highlight the need for research to understand and exploit this essential process for controlling the pathogen.

The most notable of the A. baumannii-specific proteins is AdvA. We discovered AdvA through transposon-sequencing (Tn-seq) gene-antibiotic profiling, which revealed a phenotypic signature defined by hypersusceptibility to fluoroquinolone antibiotics tightly connecting AdvA to division pathways [15]. We showed that AdvA depletion resulted in dramatic terminal filamentation and blocked growth, revealing the essential nature of the protein [15, 20]. Interestingly, analysis of the insertion sites in viable mutants within the dense transposon mutant banks used in Tn-seq revealed a highly biased distribution in advA, with transposons tolerated only at a specific region roughly halfway into the gene (Fig 1A, purple and green bars). These insertions would create advA hypomorphs producing a partial protein ending at residues 171 or 208-244 (of 435 total), either before or after a central predicted coiled coil (CC). The region N-terminal to these sites is predicted to contain two transmembrane (TM) helices flanking a periplasmic domain [15]. The absence of viable insertions within this region is consistent with an absolutely critical function for the N-terminal domains of AdvA. The paucity of viable transposon insertions downstream of residue 244 was surprising, however, because all such insertions would produce larger proteins with intact N-terminal domains. This result suggests that parts of the C-terminal region of AdvA negatively affect the protein. Nothing was known about the C-terminal domain, other than predicted localization in the cytoplasm.

Fig 1. AdvA domain requirements for A. baumannii growth and division.

(A) Diagram shows the AdvA proteins tested in gene-swap experiments. Bar plot above the diagrams indicates the corresponding positions of transposon insertions in advA. Bar height corresponds to Tn-seq read abundance [20]. Absence of a bar suggests the position is intolerant of transposon insertion. Labels for each construct are given in the adjacent panel; numbers refer to AdvA residues. (B) Diagram shows the gene-swap strategy. Swap alleles are fused to msGFP2 and expressed via Ara induction [P(Ara)-advA-msGFP2], while endogenous advA is under IPTG control at its native chromosomal site [P(IPTG)-advA]. (C) Colony growth of A. baumannii AdvA swap strains on solid IPTG 0, Ara 0.5% medium. - indicates strain contained an insertless vector (no swap allele). White arrow indicates efficient but smaller colony growth with 1-230 mutant. (D) Analysis of cell length in the A. baumannii AdvA swap strains from B. Strains were imaged by phase-contrast microscopy after 3 h incubation in IPTG 0, Ara 0.5% medium, and cell length determined via MicrobeJ (n ≥ 232 cells per strain). Each symbol represents one cell; blue line indicates median. P<0.0001 when comparing each swap allele to WT control by Kruskal-Wallis test. (E) Colony growth of WT and AdvA^1-230 swap strains on solid IPTG 0 Ara 0.5% medium with or without CIP at 0.0625μg/ml.

Based on its essentiality and localization to sites of active division [15], we hypothesized that AdvA makes contact with conserved core divisome proteins in ways critical to their function. Consistent with this idea, a recent study detected two-hybrid interactions between AdvA (referred to as Aeg1) and ZipA and FtsN, and identified amino acid changes in FtsA (E202K) and FtsB (E65A) that bypassed AdvA essentiality [21], suggesting that AdvA may function through these core proteins. More work was needed to define the potential binding and controlling functions of AdvA in cell division, clarify the basis for positional effects of transposons in the gene, and understand its unique structure that has eluded annotation.

In this paper, we address the above problems and reveal the integral role of AdvA and its domains in assembly of the A. baumannii divisome. We confirm the N-terminal TM/periplasmic domains of AdvA are essential and reveal their function in binding numerous core divisome proteins. We show that the cytoplasmic C-terminal domain influences these interactions and binds to the early protein ZipA. AdvA is required to assemble the WT divisome, but we identified bypass suppressors in divisome components that function in septum formation, suggesting a role for AdvA as a checkpoint for activating septal synthesis. Finally, we show by X-ray crystallography that the AdvA cytoplasmic domain adopts a unique three-dimensional fold resembling adenylyl and guanylyl cyclase (AC/GC) enzymes but with major differences not before seen in this enzyme class, including a critical terminal structure that helps explain positional transposon effects in advA. Collectively, these studies uncover mechanisms of control central to FtsEX-independent cell division in A. baumannii that could inform the development of antimicrobial strategies to combat the pathogen.

Results

Truncation analysis confirms the positional effects of advA insertions and reveals essential protein domains.

To examine the basis for the unusual distribution of sites in advA tolerating insertional inactivation (Fig 1A), we engineered a series of advA truncation variants and used a dual-promoter gene-swapping strategy to study their effects in A. baumannii. We replaced endogenous advA with an isopropyl β-D-1-thiogalactopyranoside (IPTG)-dependent allele at its native chromosomal locus [P(IPTG)-advA], and used a plasmid to introduce the engineered advA mutant alleles downstream of the arabinose (Ara) promoter and translationally fused to GFP [P(Ara)-advA-msGFP2] (Fig 1B). With this design, endogenous AdvA can be replaced by the AdvA-GFP constructs by culturing in IPTG- Ara+ medium. The swap strains show the expected phenotypes in control conditions: efficient growth when WT AdvA is made from either locus (IPTG+ Ara−, or IPTG- Ara+ with advA(WT) swap gene), and no growth when expression is off at both loci (IPTG- Ara−, or IPTG- Ara+ without a swap gene) (S1A Fig). We used this system to examine a variety of AdvA mutants: C-terminal truncations ending after TM2 or the CC (aa 1166 or 1-230, corresponding to positions of most of the viable insertional mutants), as well as three additional downstream locations (1-287, 1-339, or 1-391); cytoplasmic fragments with or without the CC (165-435 and 220-435); and a variant lacking the predicted periplasmic domain by substitution with a short linker (Δ^37-134) (Fig 1A). All gene swap strains showed the predicted IPTG-dependent growth pattern on control Ara- media (S1B Fig). In addition, western blot analysis confirmed that Ara induction of all the AdvA-GFP fusions yielded intact proteins, with all but one (165-435) at levels reasonably close to that of the WT control construct (S2A Fig).

Analysis of the C-terminally truncated mutants on IPTG- Ara+ swap medium was remarkably consistent with the previous Tn-seq results. First, the 1-230 variant corresponding to the portion of the protein produced in most of the viable transposon mutants enabled nearly 100% colony formation efficiency on the swap medium, albeit with smaller colony size compared to the WT control (Fig 1C, white arrow). Second, the other C-terminal truncations did not enable efficient colony formation, consistent with the dearth of viable transposon insertion sites outside of residue positions 208-244 [15] (Fig 1A,C). Of note, the 1-166 and 1-339 truncations did support partial growth at high seeding densities but not at 10⁻⁵ and 10⁻⁶ dilution, consistent with presence of a small number of viable transposon insertions at corresponding positions in advA in the Tn-seq data (Fig 1A,C). To confirm that the completely ineffective C-terminally truncated variants (1-287 and 1-391) were not negatively affected by fusion to GFP, we constructed and tested untagged versions. These untagged constructs were just as deficient at supporting growth as the tagged constructs (S3A Fig). Third, we tested the 1-230 truncation for hypersusceptibility to the fluoroquinolone ciprofloxacin (CIP), a phenotype characteristic of advA hypomorphs including the viable transposon mutants examined in Tn-seq screens [15]. The 1-230 truncation was indeed unable to support colony formation during low-level CIP exposure (i.e., at a dose below the MIC for WT A. baumannii; Fig 1E). Together, these results confirm the phenotypes previously identified by Tn-seq and indicate that C-terminal, cytoplasmic segments of AdvA beyond the CC domain have the potential to negatively modulate the growth-promoting function of the N-terminal region of the protein.

To further examine the importance of the N-terminal domains for AdvA function, we tested additional variants lacking the TM helices and/or periplasmic domain. The two N-terminal truncations lacking all of these domains failed to support growth during gene swap, whether or not the CC region was included (165-435 and 220-435; Fig 1A,C). Similarly, Δ^37-134 lacking only the periplasmic domain also failed to rescue colony growth (Fig 1A,C). These results further support the idea that the N-terminal TM and periplasmic domains of AdvA provide functions essential for A. baumannii growth.

Lastly, we considered the possibility that the truncations that lacked complementing function might show dominant negative effects. None of the non-complementing variants showed any growth inhibiting activity, however, when overexpressed in WT A. baumannii harboring native advA (S3B Fig).

TM domains mediate AdvA mid-cell localization.

To connect the variant growth phenotypes to effects on cell division as well as AdvA localization, we examined the gene swap strains by microscopy. Using the same strategy from Fig 1, strains were precultured in permissive medium, washed, and resuspended at low density in liquid IPTG- Ara+ medium. Western blot analysis revealed that culturing in these conditions led to progressive depletion of endogenous WT AdvA, with ~40% and 10% remaining at 1 and 3 h (T1 and T3), respectively (S4 Fig). We used these time points to examine the cell morphology effects of AdvA variants, focusing on development of increased cell length or filamentation, which is indicative of a block in cell division. At T1, all test strains including the control lacking a swap allele displayed cell length roughly similar to the cells expressing WT AdvA-GFP, indicating that ~40% residual AdvA levels were sufficient to prevent obvious division block (S1C Fig). A different situation was seen with the more extensive depletion of native AdvA at T3. Compared to the control strain expressing a WT swap allele, the insertless vector control showed significantly increased median cell length (2 vs >5 μm), as did all AdvA truncations (Fig 1D). The extent of these cell division defects tended to match the growth defects of each variant in Fig 1C. For instance, 1-166, 1-230, and 1-339 (growth-complementing) caused relatively modest increases in cell length compared to the 1-287, 1-391 and cytoplasmic fragments (growth-non-complementing), although some cell chaining was also evident in the former set (see Fig 2B, which shows representative images from Fig 1D analysis). A notable exception was the Δ^37-134 variant, which caused only a slight cell length defect, despite being completely ineffective at permitting colony growth on solid medium (Fig 1D, Fig 2B).

Fig 2. AdvA domain requirements for mid-cell localization.

(A-B) A. baumannii harboring the indicated AdvA-msGFP2 construct from vector pYDE334 were analyzed by phase-contrast and fluorescence microscopy after 1 (A) and 3 (B) h incubation in swap medium (Ara 0.5%, no IPTG). (C) Assessment of AdvA localization in E. coli. The AdvA(WT)-msGFP2 construct was expressed in E. coli DH5α via the same plasmid as in A-B. Cells were cultured with 0.5% Ara for 2 h. Given are representative phase-contrast (left) and fluorescence (center) micrographs and demograph showing AdvA-msGFP2 fluorescence intensity along medial axes of bacteria (right). In the demographs, the y-axis indicates cell position from −4 to +4 μm. Scale bar, 4 μm.

AdvA localizes to midcell sites of division in A. baumannii [15]. We reasoned that a feature of the essentiality of the AdvA N-terminal TM and periplasmic domains was their ability to promote proper localization to these midcell sites. We examined AdvA-GFP localization in the above samples by fluorescence microscopy. WT AdvA-GFP showed the expected localization at midcell and sites of constriction at both T1 and T3 (Fig 2A, B). In general, the localization of AdvA truncation variants again tended to match their growth effects. 1-166, 1-230, and 1-339 each showed localization to midcell/constriction sites at T1, and this pattern continued at T3 albeit much less efficiently than WT; 1-287 and 1-391 showed mainly peripheral signals with poor midcell localization; and the C-terminal fragments (165-435 and 220-435) yielded nonlocalized cytoplasmic fluorescence, although some diffuse signals at constriction sites were seen sporadically with 220-435 in filamentous cells at T3 (Fig 2A, B). Interestingly, Δ^37-134 lacking the periplasmic domain showed relatively efficient midcell localization at T1 and T3, again despite being ineffective at promoting colony formation (Fig. 2A, B). Δ^37-134 differs from the 165-435 construct effectively only by addition of the 2 TM domains. Collectively these results suggest that the TM domains of AdvA are required for localization to division sites and either promotion of cell division or, at minimum, inhibition of cell elongation. Moreover, they show that none of the truncated fragments tested was sufficient to completely phenocopy the growth, cell morphology, and localization characteristics mediated by WT AdvA.

Of note, we found that midcell localization of AdvA observed in A. baumannii does not extend to the distant bacterial host, E. coli. Expression of full-length AdvA-GFP in E. coli resulted in localization to the cell periphery without detectable enrichment at midcell or constriction sites (Fig 2C). Localization to sites of division thus likely depends on specific molecular determinants of the division apparatus that have evolved in A. baumannii.

AdvA domains bind distinct sets of A. baumannii divisome proteins.

Given the essentiality of AdvA for cell division and its midcell localization, we tested the hypothesis that the protein functions in A. baumannii divisome assembly. First, we examined binding by AdvA to the conserved core divisome proteins of the pathogen. To this end we used the bacterial adenylate cyclase (AC)-based two-hybrid (BACTH) system [22, 23]. In this assay, proteins are fused to one of two fragments from the Bordetella pertussis CyaA class II AC (T18 and T25) and expressed in E. coli, producing cyclic AMP (cAMP) and cAMP-receptor protein (CRP)-dependent reporter signal if the proteins interact. We focused on WT AdvA and three truncations: 1-166 (TM and periplasmic domains), 1-230 (TM, periplasmic, and CC domains), and 165-435 (CC and remaining cytoplasmic domains) (Fig 3A). We fused the C-terminus of each fragment to T18 and T25. Western analysis using a T18 antibody confirmed that each AdvA fusion was produced intact and that protein levels were similar between the WT and truncation constructs (S5A Fig). As a control, we showed that no individual AdvA construct yielded any detectable BACTH signal when expressed alone (i.e., with the insertless complementary vector), indicating that AdvA itself does not raise cAMP levels (Fig 3A). We tested the ability of AdvA to self-interact by co-expressing both the T18 and T25 fusions. Reporter activity consistent with homodimerization was detected for the WT, 1-166, and 1-230 fragments, while no evidence of self-interaction was detected for the 165-435 cytoplasmic fragment (Fig 3A). The N-terminal membrane-tethered portion of AdvA thus appears to be sufficient for dimerization, even in the absence of the cytoplasmic domain or CC.

Fig 3. AdvA dimerizes and binds multiple core divisome proteins.

Two-hybrid analysis of AdvA interactions in E. coli. Reporter color was evaluated after growth on solid indicator medium. (A) Analysis of AdvA homodimerization with the indicated WT or truncated construct fused to T25 or T18 fragment of B. pertussis CyaA. In first and second columns, strains contained the insertless vector with the reciprocal CyaA fragment. (B) Analysis of AdvA-T18 interactions with the indicated A. baumannii division protein fused to T25.

We next examined AdvA interactions with core divisome proteins. We constructed a panel of T25 fusions to 9 core A. baumannii divisome components (ZipA and FtsA, -K, -Q, -L, -B, -W, -I, and -N); FtsZ was not analyzed because we were unable to construct a CyaA fusion that showed any two-hybrid signals with its predicted partners, A. baumannii FtsA and ZipA [24]. When the candidate panel was coexpressed with T18 fusions to AdvA, we detected interactions between AdvA^WT and FtsA, ZipA, FtsL, and FtsN, as well as a possible weak interaction with FtsB (Fig 3B), consistent with previous findings [21]. Compared with full-length AdvA, the 1-166 and 1-230 fragments showed positive two-hybrid signals with an expanded spectrum of candidates: in addition to the above strong hits, the previously weak signal with FtsB intensified, and new interactions were observed with FtsW, FtsI, and FtsQ particularly with AdvA^1-166 (Fig 3B). FtsK was the only candidate from our panel that did not bind any of the AdvA fusions; nonetheless we confirmed that the FtsK construct showed evidence of functionality, in the form of interaction with complementary fusions to FtsQ, its predicted divisome recruit by analogy with E. coli [25] (S5B Fig). While the difference in the interaction spectrum of AdvA^WT and C-terminally truncated constructs could be consistent with negative modulation of N-terminal function by the cytoplasmic domain as proposed above, an alternative possibility is that the truncated constructs place the T18 fragment in a more favorable position to interact with the complementary CyaA fragment of certain partner proteins. By contrast with the AdvA N-terminal domains, the 165-435 cytoplasmic region interacted only with ZipA (Fig 3B). Together, these results show that the N-terminal TM/periplasmic and C-terminal cytoplasmic portions of AdvA make distinct sets of interactions with the A. baumannii divisome, with the former capable of broad binding to at least 8 core divisome proteins, and the latter making more limited contacts involving ZipA.

AdvA is required for assembly of the divisome.

Having established that AdvA localizes to midcell and binds multiple divisome proteins, we further tested the hypothesis that AdvA functions in divisome assembly by determining the degree to which divisome protein recruitment to the Z-ring requires AdvA. To this end, we used a dual-label fluorescence strategy enabling concurrent tracking of Z-rings and divisome proteins in individual exponential-phase cells with or without AdvA. Our strategy had 3 key elements: (1) we located Z-rings by using a zapA-mCherry fusion gene replacing native zapA [26]; (2) we tracked a panel of divisome proteins (ZipA, FtsK, -B, -W, and -N) as GFP fusions transiently expressed via Ara induction from a plasmid; (3) we assessed the effects of stringent AdvA depletion by using the conditional AdvA strain (P(IPTG)-advA) cultured without IPTG, compared with a WT AdvA⁺ control (Fig 4A). To engender stringent AdvA depletion while cells still grew exponentially, we took advantage of a permissive condition (5% sucrose) that allowed continual growth of P(IPTG)-advA bacteria despite absence of IPTG (S6A,C Fig). This permissive condition presumably depends on leaky advA expression, since an additional copy of lacI blocked the permissive effect of sucrose [P(IPTG)-advA-GFP lacI_2x], as did advA CRISPRi, which efficiently represses the gene to ~0.5% levels compared to control [20] (S6A, B Fig). We passaged cells for 4 h in IPTG- sucrose medium (an initial passage for 3 h, followed by a second for 1 h in the presence of Ara to transiently induce GFP fusions; S6C Fig) before analyzing subcellular protein localization by fluorescence microscopy. In WT control strains harboring divisome protein-GFP constructs, midcell localization of ZapA-mCherry and all GFP fusions was observed, supporting the utility of this strategy for locating Z-rings and divisome proteins (Fig. 4B).

Fig. 4. AdvA depletion disrupts recruitment of core divisome proteins to Z-rings.

(A) Diagram shows the strategy for tracking localization of division proteins during AdvA depletion in A. baumannii. AdvA-depletion test strains contained the IPTG-dependent advA allele replacing endogenous advA as in Figs 1–2, while control strains contained the native allele. Strains contained a zapA-mCherry fusion at the zapA locus to label Z-rings and a plasmid-borne, Ara-inducible divisome gene fusion to GFP [P(Ara)-X-msGFP2] to track divisome protein localization. (B-C) Shown are representative phase-contrast and fluorescence micrographs (left) and fluorescence profile demographs (right) for WT control (B) and AdvA depleted test strains (C) as in Fig. 2. Strains were passaged in IPTG- sucrose 5% medium for 4 h total culture time to allow stringent depletion of AdvA. Division protein-GFP fusions were induced with Ara in the final 1 h of culture before microscopy. Scale bar, 4 μm.

By contrast to the results with WT, AdvA depletion caused profound defects in the ability of both early and late divisome proteins to properly recruit to division sites. In this condition, cell filamentation and chaining occurred, and while a portion of cells showed no localized ZapA-mCherry signal, many cells retained clear ring patterning (Fig. 4C). This signal designating Z-rings can be seen in the demographs in the form of a ZapA-mCherry midline signal in shorter cells and V-patterns in progressively longer/chaining cells. In such cells that had detectable Z-rings, none of the tested divisome proteins efficiently co-localized, indicating defects in the earliest stages of ordered divisome protein recruitment in the absence of AdvA (Fig. 4C). These results support the conclusion that AdvA is a key member of the A. baumannii divisome that is required to appropriately assemble the structure and allow cell division.

The AdvA requirement is bypassed by amino acid changes in FtsB and FtsW.

Despite their essentiality under WT conditions, certain E. coli core divisome proteins become dispensable when activating mutations alter other components, which has led to insights into the control mechanisms of the divisome. These include mutations in ftsA bypassing essentiality of ZipA [27], ftsL and ftsB bypassing FtsN [28, 29], and ftsW bypassing FtsEX [13]. In A. baumannii, ftsA(E202K) and ftsB(E65A) mutants were shown to bypass the growth defect of AdvA depletion [21]. We sought to determine whether additional mutations within division or other loci could bypass the essentiality of AdvA, potentially informing its roles in divisome function.

To obtain such mutations, we took advantage of the P(IPTG)-advA-GFP lacI_2x strain and isolated spontaneous IPTG-independent mutants bypassing AdvA. The extra lacI served to limit off-target mutations affecting advA repression, while the GFP tag enabled secondary screens to further exclude such mutants. Using this strategy, we obtained two distinct suppressors, both of which mapped in late divisome genes: ftsB(D68G) and ftsW(A213V) (Fig 5A). The affected residues are located in periplasmic (FtsB) and TM (FtsW) sites of the divisome (S7 Fig). To validate their suppressor functions, we re-engineered each mutation in a completely different background containing inducible dCas9 for efficient AdvA depletion by CRISPRi [20]. When advA CRISPRi is induced with anhydrotetracycline (aTc) in an otherwise WT strain, cells fail to from colonies (Fig 5B). By contrast, the same advA knockdown was no longer lethal in the ftsB(D68G) or ftsW(A213V) mutants and yielded growth efficiencies similar to the controls (no CRISPRi induction or induction using a non-targeting guide RNA, Fig 5B). The mutations also largely reversed the extreme filamentation characteristic of AdvA depletion, reducing median cell length to levels approaching that of the AdvA-replete controls (Fig 5C, D). These results validate the mutations in ftsB and ftsW as bona-fide advA suppressors. The suppressors also rescued depletion of ZipA, but not FtsK, in analogous CRISPRi colony growth experiments (S7B Fig). Given the roles of FtsB and FtsW in mediating septal cell wall synthesis [13, 28, 30, 31] and their interactions with AdvA (Fig 3B), these findings raise the possibility that AdvA influences the activation of septal wall synthesis enzymes in addition to their midcell recruitment. Since both AdvA and ZipA are bypassed by the same suppressors and interact physically, such activation may also involve coordination with ZipA.

Fig 5. Suppressor mutations in ftsB and ftsW bypass advA essentiality.

(A) AdvA⁻ suppressors isolated in AFA29 [P(IPTG)-advA-GFP lacI_2x) using non-permissive IPTG-medium. Mutations were identified by breseq analysis. (B) Validation of AdvA⁻ suppressors. ftsB and ftsW mutations were re-constructed in a CRISPRi parental strain. Strains harbored non-targeting control (−) or advA-targeting (+) sgRNA. Serial dilutions were spotted on solid medium with or without CRISPRi inducer (aTc, 50ng/ml). (C and D) Analysis of suppression of the AdvA⁻ division defect. CRISPRi strains from B were cultured for 3 h with CRISPRi inducer and imaged by phase-contrast microscopy. (C) Representative phase-contrast micrographs. Scale bar, 10 μm. (D) Analysis of cell lengths as in Fig. 1D (n ≥ 98 cells). One data point for WT/CRISPRi-advA is not shown (length > 50 μm). ** P<0.01, **** P<0.0001, ns P>0.05 when comparing each strain to WT/CRISPRi-control by Kruskal-Wallis with Dunn’s multiple comparisons test.

The AdvA cytoplasmic domain contains an AC/GC-like fold with novel features.

To understand the molecular basis of control of the A. baumannii divisome, we analyzed the structure of AdvA, focusing on its unannotated cytoplasmic region. To this end we purified recombinant cytoplasmic AdvA (residues 158-435, Fig 6A) and performed X-ray crystallography. The crystal structure covered AdvA residues 226-420, potentially a result of partial hydrolysis of AdvA during crystallization. The structure determined at 1.65 Å revealed a 3D fold comprised of six α-helices and two short η-helices surrounding a central seven-stranded β-sheet (Fig 6B). Structural similarity searches of the Protein Data Bank (PDB) using the DALI server [32] revealed that this fold shows striking resemblance to the cyclase homology domain (CHD) of type III nucleotidyl cyclases (NCs), including the widespread class III ACs and cyclic GMP (cGMP)-producing guanylyl cyclases (GCs) [33]. Almost all of the top structural matches were AC and GC enzymes. For instance, the AdvA^226-420 structure aligns closely with the AC domain of Pseudomonas aeruginosa CyaB [34](Dali Z-score 16.2, RMSD of 2.7 Å over 160 aligned residues, Fig 6C). Despite the structural similarity, the aligned residues showed very low primary sequence identity (10%). Indeed, while AdvA^226-420 maintains the characteristic core architecture of ACs such as CyaB, topology analysis revealed that their secondary structure elements have different connectivity (Fig 6D). These observations are consistent with large evolutionary divergence between AdvA and the canonical AC enzyme.

Fig 6. AdvA C-domain crystal structure reveals novel AC/GC-like fold with atypical topology.

(A) Linear diagram highlighting AdvA features. Blue shading denotes region for which structure was solved by X-ray crystallography (residues 226-420, PDB: 7SUA). Bar plot above the diagrams indicates positions of Tn-seq reads in advA (purple, Mariner; green, Tn10-ATS) as in Fig. 1A. (B) X-ray crystal structure of AdvA^226-420 (PDB: 7SUA). Helices are numbered and colored light blue; strands are medium blue; loops are gray. (C) Structural alignment of AdvA^226-420 with the AC domain of P. aeruginosa CyaB (PDB: 3R5G). Key differences between these structures are noted by arrows: AdvA-C lacks the arm domain present in AC/GC enzymes and contains two α-helices not present in most AC/GC enzymes. (D) Protein topology maps of AdvA^226-420 (top) and CyaB (bottom) generated by PDBsum. Cylinders denote α-helices, large arrows denote α-sheets, and small arrows indicate directionality of the protein chain. Boxed is the arm domain of CyaB (bottom) that is absent from AdvA.

The AdvA^226-420 structure has a number of notable features dissimilar from CyaB and other CHD structures. First, AdvA lacks the “arm” formed by 2 antiparallel β-strand extensions in almost all AC and GC enzymes (Fig 6C). The arm mediates enzyme dimerization, which is critical to formation of the active site [33]. Second, AdvA^226-420 contains two additional α-helices (α5 and α6) not present in canonical AC/GC enzymes. α5 is situated in a region that is usually a loop in AC/GC enzymes, while α6 (residues 403-419) extends beyond the typical C-terminus of AC/GCs along the surface roughly opposite that of the canonical dimerization arm (Fig 6C). This additional terminal helix is absent from all known NC enzymes.

Given the lack of the dimerization arm in AdvA, we used AlphaFold predictions [35] to model how the domain may dimerize in comparison to the well-characterized CHD dimers. In AC/GC enzymes including CyaB, the CHDs form an arm-mediated, wreath-like dimer with the shared central cleft providing intermolecular catalytic sites (S8A Fig) [33, 34]. All of our AdvA structural models placed the CHD-like domains in a different arrangement. Modelling the complex formed by two AdvA^226-420 proteins resulted in a low confidence prediction of interactions along the face containing the α6 helix (S8B Fig). Modelling full-length AdvA predicts the CHD-like domain to be positioned adjacent to the CC, with an unstructured linker separating the two domains, and a PAS-like fold is predicted for the periplasmic domain (S8C Fig 6). Modelling the complex of two full AdvA proteins predicts, albeit with low confidence, an extensive dimerization surface along the CC, TM, and periplasmic domains, with the CHD-like domains interacting with the CC helices but not with each other (S8D Fig). In combination with the absence of self-interaction observed in two-hybrid assays for the cytoplasmic fragment of AdvA (Fig. 3B, AdvA^165-435), these observations argue against the CHD-like domains forming prototypical, arm-mediated catalytic dimers.

Analysis of the residues of AdvA^226-420 that share positions with the active sites of AC and GC enzymes revealed further differences between AdvA and canonical cyclases. We used DALI to compare the structures of AdvA^226-420, CyaB, and another structural homolog, the GC Cya2 from cyanobacterium Synechocystis PCC6803 [36] (PDB: 2w01; RMSD 2.5 Å over 160 residues aligned with AdvA). CyaB and Cya2 show a number of matching conserved residues, several of which serve as the AC/GC active sites [37] (S9A Fig, positions indicated by #). Strikingly, of the 8 AC/GC active site residues, only 2 (R334 and R354) are identical in AdvA, with AdvA either containing distinct residues at the other positions or lacking the position altogether (e.g., T351 on the arm of CyaB) (S9A,B Fig). In contrast to the residues conserved in the AC/GC protein family, the AdvA family has a unique spectrum of conserved sites. A region with a prominent number of such sites comprises positions 235-241 and 320-329 (S9A Fig), which together map to an exposed tip in the AdvA 3-D fold involving the C-terminal end of helix α2 and nearby loops (S9B Fig). Residues at these positions in AdvA are completely dissimilar to those in CyaB and Cya2, and 320-329 in particular corresponds to a region of low conservation in the cyclase proteins.

The absence of most AC/GC active site residues and the critical dimerization arm in AdvA suggests that the protein is unlikely to have canonical AC or GC activity. Our findings from multiple in vivo and in vitro tests agree with this prediction. First, recall that no AdvA construct showed an AC reporter signal when expressed individually in BACTH assays, which use CRP to detect cAMP production (see Fig. 3A). Second, we used another E. coli-based reporter assay that detects both AC and GC activities by virtue of dual CRP detectors (WT CRP and a cGMP-sensitive mutant, CRP_G) [38]. Neither full-length AdvA nor two different AdvA cytoplasmic fragments containing the CHD-like domain showed a clear reporter signal in this system, in contrast to two control enzymes, A. baumannii CavA (AC) [39] and Azospillrillum GuaA (GC) [38] (S10A Fig). Third, an A. baumannii mutant lacking the AdvA CHD-like domain showed no discernible defect in cAMP or cGMP production by ELISAs (S10B Fig). Fourth, purified AdvA^158-435 lacked detectable NC activity in vitro, as measured by pyrophosphate release, with four individual nucleotides (ATP, GTP, CTP, and UTP; S10C Fig). We also found that the same incubations released negligible free phosphate, consistent with absence of NTPase activity as well (S10D Fig). Finally, we found that mutating the AdvA residues at positions aligning with critical canonical AC/GC catalytic sites (e.g., R334 and R354, which were conserved with AC/GCs, or D283 or E350, where there was divergence from AC/GCs; S9 Fig) had no adverse effect on colony formation or CIP susceptibility in gene swap experiments (S11 Fig), consistent with the sites related to canonical AC/GC activity having little importance for AdvA division-promoting functions.

Together, these results support the idea that AdvA is not a traditional AC/GC enzyme, showing essentially no canonical NC activity at least under the conditions tested. Instead, the protein contains a novel cyclase-like structure with numerous features that distinguish it from known enzymes and that may provide unique mechanisms to control cell division.

The unique terminal helix and positively-charged tip of the CHD-like domain have critical functions.

The extra α6 helix is an unusual addition to the AC/GC fold in AdvA. Its conservation across the AdvA family suggests the helix may have a role in AdvA function, an idea supported by the previously tested 1-391 variant, which had a truncation site near this helix abolishing its complementing activity. To further test this prediction, we created an additional truncated advA allele deleting the α6 helix (AdvA^1-403-GFP) (Fig 7A). We confirmed that the resulting protein was produced intact and at levels equivalent to WT (S2B Fig). We then examined its effects using the gene-swap approach. AdvA^1-403-GFP was dramatically defective in supporting A. baumannii colony formation (Fig 7B) and in blocking cell elongation (Fig 7C,E), in contrast to the parallel control allele which supported high-efficiency colony formation and short-rod shape. The growth defect of AdvA^1-403 was observed even when testing an untagged construct (S3A Fig). In addition, AdvA^1-403-GFP was defective at midcell localization at 1 and 3 h (Fig 7D, E) mimicking the results with the 1-287 and 1-391 truncations (see Fig 2). Lastly, we analyzed the protein-protein interactions of AdvA^1-403 by two-hybrid assay, fusing the truncated gene to CyaA T18 and T25. The AdvA^1-403-T18 hybrid was produced intact in E. coli and at levels similar to the WT construct (S5A Fig). We found that AdvA^1-403 does not form homodimers (Fig. 7F); however, AdvA^1-403 was able to interact with full-length AdvA^WT, as evidenced by the strongly positive BACTH signal in the AdvA^1-403-T25, AdvA^WT-T18 arrangement (Fig 7F). The heterodimer signal was weaker in the reciprocal -T18/-T25 pairing, possibly due to differences in plasmid copy number of the vectors used to express each hybrid [23]. To account for this variability, we tested both T18 and T25 fusions to AdvA^WT and AdvA^1-403 for interactions with the A. baumannii divisome proteins. AdvA^WT-T25 showed interactions with the same divisome proteins that were hits when AdvA^WT-T18 was used as the BACTH bait (Fig 7G). By contrast, AdvA^1-403 was unable to interact with any of the divisome proteins in either fusion arrangement except for a possible weak interaction seen between AdvA^1-403-T25 and ZipA-T18 in the form of a very faint signal (Fig 7G, white arrow). The two-hybrid results reveal a critical role for the unusual extended C-terminus of the CHD-like domain in enabling AdvA dimerization and all divisome binding functions of the N-terminal domains. More broadly, the unique α6 helix appears essential to AdvA midcell recruitment as well as the protein’s functions in promoting division, blocking filamentation, and enabling A. baumannii proliferation.

Fig. 7. The atypical terminal helix of the cyclase-like domain is critical for AdvA function.

(A) Protein map of AdvA^WT and AdvA^1-403 shown as in Fig 1A. (B) AdvA^1-403 fails to support A. baumannii colony growth. Colony growth was examined after gene swap with the indicated AdvA construct fused C-terminally to msGFP2 as in Fig 1B. (C-E) Microscopy analysis of AdvA^1-403 reveals mislocalization and dramatic cell elongation. The indicated advA-msGFP2 gene swap strains were visualized after culture in IPTG- Ara 0.5% medium for 1 and 3 h by phase-contrast and fluorescence microscopy. (C) Cell lengths from phase contrast images at 3 h were analyzed as in Fig 1D (n ≥ 206 cells per strain, P<0.0001 when comparing vector or 1-403 to WT control by Kruskal-Wallis test). Representative micrographs and demographs of AdvA-msGFP2 subcellular fluorescence are given for 1 h (D) and 3 h (E) time points. (F and G) BACTH analysis in E. coli of AdvA^1-403 dimerization (F) and interactions with A. baumannii core divisome proteins (G). - indicates two insertless T18 and T25 vectors. White arrow indicates partial reporter signal with the AdvA^1-403-T25, ZipA-T18 pairing.

We also examined the importance of the conserved tip region of the CHD-like domain. This region includes multiple basic residues, including K239 and R324, which contribute to a prominent positively charged surface surrounding a short cleft when viewing the electrostatic surface of the structure (Fig 8A). We mutated K239 and R324 to alanine and examined mutant complementing and localization phenotypes in the gene swap system. While both mutant alleles supported A. baumannii colony growth on IPTG- Ara+ medium lacking antibiotics, the R324A mutant showed hypersusceptibility to CIP, mimicking the 1-230 truncation which entirely removes the AC/GC-like domain (Fig 8B). Consistent with its colony growth effects, in antibiotic-free gene-swap medium R324A caused a subtle but significant increase in cell length (Fig 8C, D). R324A also slightly reduced the efficiency of AdvA midline localization, showing increased peripheral signal compared to the WT control (Fig 8D). These results identify a single key residue in the positively charged tip of AdvA that is important for ensuring efficient bacterial division and intrinsic resistance to a division-targeting antibiotic.

Fig. 8. AdvA residue R324 is important for intrinsic fluoroquinolone resistance and efficient cell division.

(A) Electrostatic surface potential of AdvA^226-420 showing the positively-charged tip region (head-on view) and positions of residues K239 and R324. (B) AdvA K239 and R324 were each mutated to alanine, C-terminally fused to msGFP2, and tested for effect on A. baumannii colony formation after allele swap, alongside reference alleles (advA WT and 1-230). Shown are protein maps, with asterisks representing point mutation, followed by swap allele label and images of colony growth after serial dilution on solid Ara 0.5% medium with or without CIP (0.1 μg/ml). (C-D) Microscopy analysis of AdvA^R324A effect on division and subcellular localization in A. baumannii. Gene swap strains producing the indicated AdvA-msGFP2 construct were cultured with Ara 0.5% (no CIP) for 3 h and visualized by phase-contrast and fluorescence microscopy. (C) Analysis of cell lengths as in Fig. 1D (n ≥ 260 cells per strain). One data point for WT allele is not shown (length > 15 μm). **** P<0.0001, Mann-Whitney test. (D) Representative micrographs and demographs of subcellular fluorescence. Scale bar, 4 μm.

Sequence and structural comparisons reveal close and distant AdvA homologs.

BLASTp analysis of the full-length protein sequence revealed that closely related homologs of A. baumannii AdvA are restricted to the Moraxellaceae family of γ-proteobacteria (Fig 9A). As expected, orthologs with highest similarity were in Acinetobacter (sequence identity ranging from 51-97%), while other genera, including Psychrobacter and Moraxella, showed more divergence (22-41% identity with A. baumannii AdvA). Regarding the Acinetobacter AdvA orthologs, a stretch of low sequence conservation and/or gaps was seen in the linker connecting the CC and CHD-like domains (between residues 210-230), which corresponds well to the advA sites enriched in viable transposon insertions (Fig 9A). In accord with the previous analysis of the CHD-like domain, virtually all of the close orthologs (126/128) had arginine at AdvA residue position 324, and multiple α6 helix residues were conserved (Fig 9A). The 13 most C-terminal residues of AdvA predicted to form an unstructured tail after α6 were either poorly conserved or absent from the vast majority of orthologs.

Fig 9. Analysis of AdvA homologs.

(A) Heatmap shows amino acid sequence similarity of close AdvA homologs. A. baumannii AdvA is the first row, followed by sequence homologs from Refseq prokaryote representative genomes, using BLASTp with filters E<0.01, coverage 75-100. Taxon is indicated by label at left. Within heatmap, blue shading represents BLOSUM similarity score (darker blue, higher score/conservation), and white indicates gap. Diagram above the heatmap shows AdvA residue positions and protein features, and corresponding advA Tn-seq mutant abundance as in Fig. 1A. (B) Structural alignment of AdvA^226-420 (PDB: 7SUA) with Alphafold model of the CHD-like domain of P. aeruginosa PA4961 (residues 324-512; pTM 0.91). Arrowhead denotes structurally conserved C-terminal helix. (C) Genomic regions of A. baumannii advA and P. aeruginosa PA4961. PA4961 predicted protein features are shown below the gene map.

To identify possible remote homologs of AdvA beyond the Moraxellaceae not evident through sequence comparisons, we performed an expanded search for proteins whose structures align with the AdvA^226-420 crystal structure. We required that candidate structural homologs contained a C-terminal helix matching that of AdvA and lacked the dimerization arm of AC/GCs. As no protein in the PDB passed these criteria, we used Foldseek to search the AlphaFold Protein Structure Database [40]. This led to the identification of PA4961 from P. aeruginosa PAO1. PA4961 is an uncharacterized protein annotated as a histidine kinase (HK). The predicted structure of the C-terminal segment of the protein (residues 324-512) significantly matches AdvA^226-420 (Fig 9B, RMSD of 1.115 Å for 78 pruned atom pairs and 4.995 Å across all 172 pairs) and resembles a CHD, not a HK catalytic domain. Like AdvA, PA4961 has the unusual property of lacking the canonical AC/GC catalytic residues and dimerization arm while containing an extra C-terminal α-helix. Overall, PA4961 and AdvA have low amino acid sequence identity (12%) but have the same predicted TM topology with analogous periplasmic domains, and both genes share a genomic context oriented divergently from serB (Fig 9C). Interestingly, although few studies have examined PA4961, genome-wide screens have linked mutations in the locus (or its PA14 equivalent PA14_65570) to defective fitness in bacteriological medium [41], increased antibiotic susceptibility [42, 43], and cell filamentation [44]. AdvA and PA4961 are thus likely to be distant homologs and examples of a new type of bacterial control protein employing a CHD-like fold.

Discussion

In these studies, we have defined key functional and structural features of AdvA, a distinctive essential protein that bears no sequence resemblance to known proteins and is critical to A. baumannii cell division. Our Z-ring recruitment and bypass suppressor data support the conclusion that AdvA essentiality is based on at least two key functions, assembling and activating the multicomponent divisome in the microorganism. We revealed by truncation analysis that AdvA contains N- and C-terminal regions with different contributions to A. baumannii division protein interactions and viability. The N-terminal periplasmic and TM domains are required for dimerization and myriad interactions with early and late divisome proteins, rendering them essential for viability. The cytoplasmic C-terminal domain, by contrast, interacts with one early divisome component, ZipA, and can be removed without sacrificing viability but at the cost of cell division and fluoroquinolone susceptibility defects. Removing a small segment from the end of the C-domain, however, had the striking effect of blocking AdvA divisome interactions and A. baumannii division. We can infer from these results that the cytoplasmic domain has the ability to control the activities of the protein, potentially through conformations that recapitulate effects of truncations. To illuminate these control mechanisms, we solved the crystal structure of the cytoplasmic domain, which revealed a novel 3D fold resembling AC/GC enzymes with crucial differences—absence of canonical active sites and presence of novel structural elements essential to division and CIP susceptibility. This work has thus shed light on functions of AdvA fundamental to cell division and their relationship to unique structural features of the protein.

Our integrated mutant, interaction, and localization data support a model for the critical role of AdvA in assembling and activating the A. baumannii divisome to promote cell division (Fig 10A). AdvA localizes to midcell early in divisome assembly likely via recruitment by a Z-ring binding partner (e.g., ZipA and/or FtsA, although interaction with FtsZ has not been ruled out) (Fig 10A, 1). AdvA is likely a pre-formed dimer/oligomer upon arrival, given the two-hybrid results. Our results also indicate that AdvA makes both inner membrane and cytoplasmic contacts with ZipA through its different domains. Interestingly, modeling the structure of the cytoplasmic ZipA-AdvA complex (Fig 10B; AlphaFold3 confidence scores ipTM = 0.89, pTM = 0.92) reveals that the predicted binding interface involves the two key features of the AdvA CHD-like domain—R324 and other residues of the positively charged tip, which form salt bridges with acidic residues in ZipA; and the essential α6 helix, predicted to interact with a portion of the ZipA disordered linker. Detection of cytoplasmic ZipA and/or other cytoplasmic molecules by the CHD-like domain (Fig 10A, 2) may serve as signals that promote an active state in the AdvA N-terminal domains, enabling interaction with a plethora of divisome proteins including FtsA, FtsQLB, FtsWI, and FtsN (Fig. 10A, 3). By analogy with E. coli [10, 45], an interaction that modulates FtsA may drive recruitment of the subsequent proteins, while activating interactions between AdvA and the latter complexes would promote synthesis of the septal wall (Fig 10A, 4). Suppressor mutations in ftsB and ftsW bypass AdvA essentiality and to large degree rescue the AdvA⁻ cell division defect. These mutations could be bypassing an AdvA-mediated checkpoint for activating septal peptidoglycan synthesis proteins FtsWI, as analogous suppressors in FtsB and FtsW in E. coli resulted in constitutive forms that hyperactivate division [28, 30, 31]. By bridging early and late divisome proteins and enabling activation of the latter, AdvA would provide unique mechanisms that replace critical functions of FtsEX, the prototypical controller of division widespread in other bacteria that is absent from Acinetobacter. In support of this idea, all of the taxa harboring close AdvA orthologs (Fig 9A) also lack FtsEX. Whether AdvA also functions, like FtsEX, in activating septal cell wall hydrolysis via regulatory partners of peptidoglycan amidases [14, 46] remains to be determined.

Fig. 10. Model for AdvA-dependent assembly and activation of the A. baumannii divisome.

(A) Model depicts roles of AdvA in divisome construction and control to drive cell division independent of FtsEX, which is absent from Acinetobacter. (1) AdvA (blue) is recruited early to midcell division sites, potentially through binding ZipA (green) and/or FtsA and forms a dimer. Our results show that both the AdvA N-terminal TM/periplasmic domains and cytoplasmic CHD-like domain bind ZipA. (2) AdvA detects cytoplasmic signals, such as the ZipA C-terminal domain (green circle) or possibly nucleotide-related molecules (triangles), using its CHD-like structure. Activating signals may propagate to the AdvA N-terminal TM and periplasmic domains (dotted arrow). (3) The active N-terminal domains modulate FtsA and drive recruitment of downstream divisome proteins (Fts Q, L, B, W, I, N; arrows). (4) AdvA binding to direct partners FtsN and/or QLB leads to activation of WI and cell septum synthesis. The terminal α-helix of AdvA (dark blue) is critical to all the above functions, likely by facilitating interaction with early divisome components (ZipA) to bring AdvA to the Z-ring, as well as by stabilizing an AdvA state competent for dimerization and late divisome protein interactions. (B) Structural model of the cytoplasmic AdvA-ZipA complex predicted by AlphaFold3 (ipTM = 0.89, pTM = 0.92). Ribbon diagrams for A. baumannii AdvA^226-420 and A. baumannii ZipA^171-339 are given in light blue and green, respectively. Side chains are shown for basic residues in the conserved tip of AdvA and acidic residues in the interacting surface of ZipA, and R324 and α6 of AdvA are labelled. Surface overlays show electrostatic potential (blue, positive charge; red, negative charge).

Control of AdvA cell division functions via its cytoplasmic domain appears to hinge on the critical α6 helix, based on the dramatic phenotypes of the AdvA^1-403 mutant (blocked homodimerization, divisome interactions, mid-cell localization and ultimately division). It is intriguing that the α6 region has such global effects, when two-hybrid results indicate the entire cytoplasmic domain binds just one core divisome protein (ZipA) and the domain is predicted to provide minimal self-interaction surface. We propose a two-fold explanation for the striking defects of AdvA^1-403. First, the CHD-like domain may contain an inhibitory function that is prevented by α6; for instance, α6 could act as a “latch” between AdvA subunits, enabling the remainder of the protein to stably adopt a dimeric state competent for divisome interactions and control. The inability of AdvA^1-403 to form dimers may be the basis for most of the binding defects of the protein, if binding requires dimer/oligomer-dependent AdvA surfaces or conformations. Second, independent of dimer formation, α6 may be critical for interaction with the early Z-ring protein ZipA, and thus for AdvA recruitment to midcell division sites. This idea is supported by the two-hybrid data showing that ZipA interacts with AdvA^165-435 but not AdvA^1-403, despite neither having the ability to homodimerize (Fig 3A,B), and by the structural model (Fig 10B). Notably, the critical requirement in A. baumannii for the C-terminus of AdvA containing α6 helps to explain the unusual positional effects of transposon insertions in advA on Tn-seq fitness [15]. Insertions after the possibly inhibitory sites in the CHD-like domain producing truncations lacking α6 would likely yield globally inactive forms of the protein mimicking AdvA^1-403 and blocking division; whereas insertions in the linker upstream of the domain would avoid such inhibitory latchless cytoplasmic fragments, yielding dimeric N-terminal domains that provide hypomorphic binding functions.

We have uncovered a novel structural homolog of the type III NC fold, most closely related to the CHD of AC/GC enzymes but without the typical catalytic residues, dimerization arm, or topology. Combined with the lack of AC/GC activity in multiple assays, our data strongly suggest that the AdvA domain is not an AC or GC. Diguanylate cyclases (DGCs) are also members of the type III NC family, containing the same core fold as AC/GCs while lacking a dimerization arm [33, 47], and we considered a possible relationship with these enzymes. AdvA bears lower resemblance to DGCs than AC/GCs, however, with fewer aligning residues (120 vs 160) and lower DALI Z-score (10.9 vs 16.3) when comparing the closest DGC and AC/GC matches to AdvA. The absence of a GGDEF motif or PPi release from nucleotides in vitro further suggest that AdvA is not a DGC. AdvA^165-435 also lacks appreciable NTPase activity in vitro, at least under the conditions tested, together raising the possibility the protein uses mechanisms independent of NTP hydrolysis. By contrast, our mutational data indicate the domain is critical to division efficiency and can globally modulate AdvA self and divisome interactions. Why would Acinetobacter evolve a variant NC protein for such a role? The domain may have a yet unidentified enzymatic activity, in the context of the complete protein, that influences divisome interactions. Alternatively, as suggested above, AdvA may use the CHD-like fold to sense nucleotide-related ligands important for control of the A. baumannii cell cycle, or to exploit interactions with proteins that are traditional binding partners of ACs, such as G-proteins [33]. Experiments examining these alternative mechanisms are in progress.

Our results show that A. baumannii relies on a novel NC-related architecture to enable a distinct strategy for controlling cell division. The CHD-like fold may be used by other pathogens including P. aeruginosa, which possesses a distant AdvA ortholog also linked to antibiotic resistance and cell division. Interestingly, P. aeruginosa employs the widespread FtsEX system in its division cycle [48]. How the two diverse pathogens have evolved to use AdvA-family proteins in distinct ways is a key question for future studies. Future work should also focus on resolving the unique molecular contacts that govern divisome assembly and activation in A. baumannii, and illuminating the conformational dynamics of AdvA during this process that are influenced by the cyclase-like domain. Improving our understanding of this new class of control protein has potential to guide development of strategies for blocking division and enhancing antibiotic killing of MDR nosocomial pathogens.

Materials and methods

Bacterial strains, growth conditions, and antibiotics

Bacterial strains used in this work are described in Table S1. A. baumannii strains were derivatives of ATCC 17978. Escherichia coli strains were DH5α, DH5α(λpir), or XL1-Blue for cloning; BTH101 for BACTH; and BL21[DE3] for protein purification and AC/GC reporter assay. All cultures were in lysogeny broth (LB) (10g/L tryptone, 5g/L yeast extract, 10g/L NaCl) at 37°C unless otherwise indicated. LB agar (LBA) medium was supplemented with antibiotics as needed for isolation of strains with plasmids: tetracycline (Tc, 5-10μg/ml), kanamycin (Km, 10μg/ml), carbenicillin (Cb, 50μg/ml) for A. baumannii; Tc (10μg/ml), Km (30-50μg/ml), Cb (50-100μg/ml), ampicillin (Amp, 50-200 μg/ml), and/or chloramphenicol (Cm, 25μg/ml) for E. coli. Liquid cultures were aerated by shaking in flasks or by rotation on a tube roller drum. Medium was supplemented with 100μM IPTG for maintaining P(IPTG)-advA strains and with 5-10% sucrose for maintaining the advA(1-230) strain.

Molecular cloning and strain construction

Plasmids and synthetic DNA oligos/gene fragments used in this study are listed in Tables S2 and S3, respectively. Oligos and gBlocks were purchased from IDT. Blunt-ended PCR products (Q5 High-Fidelity DNA polymerase, NEB) and gBlocks were cloned into the HincII site of pUC18 and sequenced (Genewiz, Quintara Biosciences) before subcloning. Plasmids were introduced into A. baumannii by electroporation.

Allelic exchange:

For chromosomal P(IPTG)-advA, a ~3kb fragment including lacIq-P_T5lac-advA was amplified from pAFE43 with 5’ SphI and 3’ SalI sites, and ~1kb sequence upstream of the advA ORF was amplified from gDNA with 5’ NotI and 3’ SphI sites; the fragments were digested and fused via 3-way ligation into NotI and SalI sites of pSR47S, resulting in pAFE104. For chromosomal zapA-mCherry, a ~800bp fragment including A. baumannii zapA and upstream sequences was amplified from gDNA and subcloned in BamHI and XbaI sites of pAFE25, creating zapA-mCherry flanked by BamHI and PstI sites; a ~1kb fragment downstream of the zapA ORF was amplified from gDNA and cloned using 5’ PstI and 3’ SalI sites; the BamHI-PstI and PstI-SalI fragments from the resulting plasmids were 3-way ligated with BamHI-SalI cut pSR47S, creating pAFE324. pAFE25 was constructed by cloning a synthetic gene fragment harboring 5’ XbaI and 3’ PstI sites in the HincII site of pUC18 in the direction of lacZα. For chromosomal advA-GFP, advA-GFPmut3 was amplified from pAFE42 with 5’ BamHI and 3’ PstI sites; ~1kb of sequence downstream of advA was amplified from gDNA with 5’ PstI and 3’ SalI sites; the fragments were digested and fused via 3-way ligation into BamHI and SalI sites of pSR47S resulting in pAFE65. To re-introduce ftsW or ftsB alleles, ~2kb of sequence centered on each mutation was amplified from mutant gDNA with flanking BamHI and SalI sites and subcloned in pSR47S resulting in pAFE185 and pAFE186, respectively. For chromosomal advA(1-230), ~2kb of partial advA and 5’ flanking sequence was amplified from gDNA with 5′ NotI and 3′ PstI sites and a stop codon following codon 230; ~1kb of sequence 3’ of the advA stop codon was amplified with 5′ PstI and 3’ SalI sites; and fragments were 3-way ligated with pSR47S via NotI and SalI sites, creating pAFE272. Allelic exchange was performed using the sucrose counterselection method and verified by colony PCR [49].

GFP fusions:

A Tc^r vector for Ara gene induction (pYDE344) was constructed by cloning the NsiI-EcoRI fragment from pVRL2 into pYDE153, such that araC-P_BAD replaces the lacI-P_T5lac. (i) To construct advA-msGFP2 under Ara control, advA fragments were generated with 5’ BamHI and in-frame 3’ XbaI sites and subcloned in pAFE225. (ii) Deletions were constructed by PCR using the WT plasmid (pAFE229) as template. To delete parts of the AdvA C- or N-terminus, we replaced the full-length fragment with truncated PCR products using distinct 3’ or 5’ primers, respectively, the latter containing the Shine-Dalgarno and ATG sequence from pFPV25. For internal deletion, we used inverse PCR with primers containing a glycine codon and SpeI sites, followed by self-ligation via SpeI, such that Gly-Thr-Ser replaced the periplasmic domain. (iii) Point mutants were cloned with gBlocks. (iv) For untagged constructs, fragments were re-amplified using distinct primers with EcoRI (5’) and stop codon and PstI sites (3’). (v) For divisome proteins C-terminally fused to msGFP2, gene fragments were amplified from gDNA and subcloned into pAFE225. (vi) To make divisome proteins N-terminally fused to msGFP2, fragments were amplified from gDNA with 5′ SpeI and 3′ PstI sites and subcloned to pAFE309 (pAFE309 constructed by PCR-amplifying msGFP2 using pAFE225 template and primers with 5′ SacI and 3′ SpeI sites and cloning the blunt-ended product in the HincII site of pUC18 in the direction of lacZα). From the resulting plasmids, the EcoRI-PstI fragments (i-v) or SacI-PstI fragments (vi) were subcloned in pYDE334.

Reporter assay plasmids:

For BACTH assay, genes were amplified from gDNA with 5’ SphI and 3’ XbaI sites (C-terminal fusions) or 5’ BamHI and 3’ EcoRI sites (N-terminal fusions), and constructs were subcloned into either pKNT25 and pUT18 (C-terminal fusions) or pKT25 and pUT18C (N-terminal fusions). BACTH plasmids were maintained in E. coli XL1-Blue at 30°C before cotransforming E. coli cya strain BTH101. For AC/GC assay, genes were amplified from gDNA with 5’ NdeI and 3’ SalI or HindIII sites. Constructs were subcloned in pET23(a)+ before transforming E. coli cya CRP/CRP_G reporter strains.

Western blot analysis

Cells were grown to optical density (OD) ~1, and 1 OD unit was harvested by centrifugation. Pellets were resuspended in 50μl Laemmli buffer and heated at 95°C for 5 min. Proteins were separated by 10% PAGE and transferred to low-fluorescence PVDF membranes. SYPRO Ruby blot stain (Invitrogen) was used for visualizing total protein following the manufacturer’s protocol. Blots were blocked (5% milk in TBST) and incubated with primary rabbit anti-GFP antibody (Invitrogen A11122, 1:2000) or rabbit AdvA^165-435 antiserum (1:10,000) followed by secondary goat anti-rabbit IgG, HRP (Invitrogen 65-6120, 1:5000); or primary mouse CyaA antibody 3D1 (Santa Cruz Biotechnology SC-13582, 1:200) followed by secondary rabbit anti-mouse IgG, HRP (A27025, 1:5000). Blots were developed with Western Lightning Plus (Perkin Elmer) and imaged with a ChemiDoc MP (BioRad). Rabbit polyclonal AdvA antiserum was produced at the Pocono Rabbit Farm and Laboratory by immunization with purified recombinant AdvA^165-435 using standard protocols.

Colony formation tests

Bacteria were precultured for 3 h in standard or permissive LB medium (containing 100μM IPTG for gene swaps, or 5% sucrose for advA(1-230)) and washed 2x with PBS. Serial 10-fold dilutions in PBS were spotted on the noted solid medium. Colonies were imaged after overnight growth with transilluminated light on a ChemiDoc MP (BioRad).

Microscopy

Gene swap and divisome tracking strains were precultured in LB with 100μM IPTG and washed as above, seeded at OD 0.02 in LB with 0.5% Ara, and imaged at 1 and 3 h (gene swap). For divisome tracking, cultures were seeded first in LB with 5% sucrose, incubated for 3 h, re-seeded at OD 0.02 in LB containing 5% sucrose and 0.5% Ara, and cells imaged after 1 h. CRISPRi strains were precultured in LB for 3 h, seeded at OD 0.02 in LB with 50ng/ml aTc, and imaged after culture for 3 h. To image cells, a sample was mixed 1:1 with Slow-Fade Diamond (Thermo Fisher) and immobilized on agarose pads (1% agarose in PBS). Micrographs were acquired with 100x/1.4 phase-contrast objective on Zeiss Axio Observer 7 microscope. Fluorescence micrographs used a Colibiri 7 LED light source with 475nm LED and filter set 92 HE for GFP, and 587nm LED and filter set 45 for mCherry. Fiji [50] was used to determine cell boundaries from stacked GFP/mCherry/phase images. Cell length was quantified from phase images using MicrobeJ [51]. To analyze fluorescence profiles, images were background-subtracted and quantified by using MicrobeJ. Results were exported into .mat format and demographs were generated using a custom MATLAB (R2022b) script. Cells were excluded from analysis if their mean fluorescence was outside of the range defined by the lower quartile minus 1.5 times interquartile range (IQR) and the upper quartile plus 1.5 times IQR. Medial axis fluorescence profiles were then arranged by cell length and centered on Y=0.

Bacterial two-hybrid assay

BACTH assays followed described protocols [23]. Overnight cultures of BTH101 cotransformants were spotted on solid BACTH assay medium (LBA with Km 30μg/ml, Cb 50μg/ml, IPTG 500μM, and X-gal 40μg/ml). Plates were incubated overnight at 30°C followed by 24 h at room temperature and reporter color was imaged using dark-field illumination.

Suppressor isolation and mapping by whole-genome sequencing

Colonies of AFA29 on LBA with 1mM IPTG and 5 μg/ml Tc were directly streaked on LBA plates, or were inoculated into parallel tubes of liquid LB, precultured 3h, and spread on multiple LBA plates. After overnight incubation, colonies were purified in parallel on LBA with or without 5 μg/ml Tc and incubated overnight. For isolates maintaining Tc^r (marker of pEGE305 providing extra copy of lacI) colonies from the LBA plate were inoculated into LB liquid. The parent strain was also inoculated in LB with or without 100μM IPTG to provide reference controls. When cultures reached OD of 0.3, samples were analyzed by fluorescence microscopy. For suppressor candidates that maintained a low GFP signal similar to the uninduced parent, cultures were continued until they reached saturation, stocked, and gDNA extracted (DNeasy kit Qiagen). gDNA was quantified by SYBR green assay. Illumina libraries were prepared using a modified small-volume Nextera tagmentation method and sequenced at TUCF-Genomics (single-end 100 bp, HiSeq 2500) [52, 53]. Reads were mapped to NZ_CP012004 and mutations identified using breseq [54].

Expression, purification and crystallization of A. baumannii AdvA

The gene encoding ACX60_00475 from A. baumannii ATCC 17978 (residues 158-435) was cloned using gene-specific primers (Forward: 5’TACTTCCAATCCAATGCCCGTGCCATTGCTCGCCCAACA3’; reverse 5’TTATCCACTTCCAATGTTAAGATGCTGCACTACTCGGGCT 3’) and ligation independent cloning as described previously [55, 56] into pMCSG53 vector, containing an N-terminal 6×His tag followed by a TEV protease cleavage site, encoding Amp resistance, and genes for rare codons. E. coli BL21(DE3)(Magic) cells [55] were transformed with the plasmid and the overnight starter culture was grown in LB supplemented with 130 μg/ml Amp and 50 μg/ml Km at 37°C and 220 rpm. The next day, 3 liters of M9 medium (High Yield M9 SeMet media, Medicilon Inc.) supplemented with 200 μg/ml Amp and 50 μg/ml Km were inoculated at 1:100 dilution with the overnight culture and incubated at 37°C and 220 rpm. Protein expression was induced at OD=1.8-2 by an addition of 0.5 mM IPTG and the culture was further incubated at 25°C and 200 rpm for 14 hours. The cells were harvested by centrifugation at 6,000 x g for 10 min, re-suspended (5 ml of buffer/1g of cells) in lysis buffer (50 mM Tris pH 8.3, 500 mM NaCl, 10% glycerol, 0.1% IGEPAL CA-630) and frozen at −30°C until purification.

Frozen pellets were thawed and sonicated at 50% amplitude, in 5 s x 10 s cycle for 20 min at 4°C. The lysate was cleared by centrifugation at 18,000 x g for 40 min at 4°C, the supernatant was collected. The protein was purified in one step by IMAC followed by size exclusion chromatography using ÅKTAxpress system (GE Healthcare) as previously described with some modifications [57]. The cell extract was loaded into a His-Trap FF (Ni-NTA) column in loading buffer (10 mM Tris-HCl pH 8.3, 500 mM NaCl, 1 mM Tris (2-carboxyethyl) phosphine (TCEP), and 5% glycerol) and the column was washed with 10 column volumes (cv) of loading buffer and 10 (cv) of washing buffer (10 mM Tris-HCl pH 8.3, 1M NaCl, 25 mM imidazole, 5% glycerol). Protein was eluted with elution buffer (10 mM Tris pH 8.3, 500 mM NaCl, 1 M imidazole), loaded onto a Superdex 200 26/600 column, separated in loading buffer, collected and analyzed by SDS-PAGE. 6xHis-tag was cleaved by recombinant TEV protease in ratio 1:20 (protease:protein) overnight at room temperature. The cleaved protein was separated from recombinant TEV protease, un-cleaved protein and 6xHistag peptide by Ni-NTA-affinity chromatography using loading buffer followed by loading buffer with 25mM imidazole. The cleaved protein was collected in loading buffer with 25mM imidazole. Both fractions were analyzed by PAGE for purity and 6xHis tag cleavage.

The protein was concentrated to 8.8 and 16.6 mg/ml in loading buffer without NaCl and with 500 mM NaCl, respectively. Both protein samples were used to set up 2-μl crystallization drops (1μl protein:1μl reservoir solution) in 96-well plates (Corning) using commercial Classics II, PACT, and JSCG+ (QIAGEN) crystallization screens. Diffraction quality crystal grown from the condition with 0.1M Bis-Tris (pH 6.5) and 20% (w/v) PEG 5000 MME (Classics II, condition 46) were flash cooled in liquid nitrogen for data collection.

Structure Determination, Refinement and Validation

Diffraction quality crystals were screened and several data sets were collected. The data sets were collected at beam line 21ID-F of the Life Sciences-Collaborative Access Team (LS-CAT) at the Advanced Photon Source (APS), Argonne National Laboratory. Images were indexed, integrated, and scaled using HKL-3000 [58]. The data set with the 1.65 Å resolution was used for structure solution and refinement. The structure was determined by the Single Anomalous Dispersion (SAD) method using anomalous signal from the Se-Met protein derivative. Crystal belongs to the orthorhombic space group P2₁2₁2₁ with one protein chain in the asymmetric unit. The initial model (residues 221-413) was built using the HKL3000 structure solution package. Data quality, structure refinement and the final model statistics are shown in Table 1. The model was refined using REFMAC [59] and additional residues were manually added during visual inspections in Coot [60]. The water molecules were generated automatically using ARP/wARP [61], ligands (PEG and EDO) were fit into electron density maps manually and the structure was further refined in REFMAC. The final stages of refinement were carried out with the Translation–Libration–Screw (TLS) group corrections, which were generated by the TLS Motion Determination (TLSMD) server [62, 63]. The quality of the model during refinement and the final validation of the structure was done using MolProbity [64] (http://molprobity.biochem.duke.edu/). The structure was deposited to the PDB (https://www.rcsb.org/) with the assigned PDB code 7SUA. Structural alignments and molecular graphics procedures were performed with ChimeraX [65].

Table.1. Data quality and refinement statistics.

Values in parentheses are for the outer shell.

PDB Accession Code	7sua
Data Collection
Space group	P212121
Unit cell parameters (Å; °)	a = 50.76; b = 59.78; c = 75.63
	α = β = γ = 90.00
Resolution range (Å)	30.00 - 1.65 (1.68 - 1.65)
No. of reflections	27,958 (1,354)
Rmerge (%)	6.9 (68.7)
Completeness (%)	99.5 (99.0)
⟨I/σ(I)⟩	24.3 (2.4)
Multiplicity	4.3 (4.1)
Wilson B factor	22.7
Refinement
Resolution range (Å)	27.06 - 1.65 (1.70 - 1.65)
Completeness (%)	99.2 (94.8)
No. of reflections	26,532 (1,920)
Rwork/Rfree, (%)	17.6/20.8 (26.5/24.6)
Protein chains/atoms	1/1,664
Ligand/Solvent atoms	11/251
Mean temperature factor (Å2)	25.8
Coordinate Deviations
R.m.s.d. bonds (Å)	0.006
R.m.s.d. angles (°)	1.365
Ramachandran plot
Favored (%)	100.0
Allowed (%)	0.0
Outside allowed (%)	0.0

Enzyme assays

AdvA^158-435 was also purified without cleavage of tag and without SeMet for enzyme assays. For this purpose, E. coli BL21 (DE3) pLysS cells were transformed with the pMCSG53-advA(168-435) plasmid; the overnight starter culture was in LB with Cm and Cb; and 500ml of Terrific Broth (TB) with 50μg/ml Cb and 25μg/ml Cm was used as protein production medium. After protein induction as described above, cultures were centrifuged and cell pellets frozen at −80°C. Pellets were thawed, resuspended in 5ml lysis buffer/1g of cells, and 15 μL benzonase (Sigma) was added per 50 mL to reduce viscosity. Cells were lysed by two passes through a French press, and lysates were clarified by centrifugation at 4°C for 1 h at 16,000 x g. Using gravity flow, 2.5 mL bed volume of Ni-NTA column was washed with 3 mL and then 10 mL of lysis buffer. Lysate was passed through column, followed by wash with 3× 10 mL of washing buffer. Protein was eluted with 5 mL of elution buffer with 250mM imidazole, then 5 mL with 500mM imidazole. The two elution fractions were merged, dialyzed against three changes of 1x loading buffer without TCEP at 4°C. After dialysis changes, TCEP was added to 1 mM final concentration. Protein aliquots were stored at 5.7 mg/ml concentration at −80°C.

NTP hydrolysis reactions were performed in 100 μL volumes containing NTP hydrolysis reaction buffer (25 mM HEPES, 150 mM KCl, 2 mM MgCl2, and 2 mM TCEP at pH 7.4.) and 0.5 mM NTP. NTP stock solutions were prepared at 100 mM by dissolving NTP powder (ATP, GTP, CTP, or UTP; Jena Bioscience) in 25 mM HEPES, pH 7.4. For measuring pyrophosphate release, 5 μM of 6xHis-tagged AdvA^158-435 protein or equivalent volume of AdvA loading buffer was added to reactions. Reactions were mixed by pipetting, then incubated at 37°C for 1 h. Pyrophosphate was measured using the MAK168 fluorogenic PPi sensor kit (Sigma-Aldrich) according to manufacturer instructions, with PPi standards prepared in NTP hydrolysis reaction buffer. Fluorescence was measured using a BioTek Synergy H1 microplate reader. For measuring phosphate release, reactions were prepared as above, except a set of control reactions were set up with addition of 5 μM of apyrase (NEB). Reactions were diluted two-fold by addition of 100 μL reaction buffer and phosphate concentration measured using the Sigma-Aldrich Malachite Green Phosphate Assay Kit (MAK307) according to manufacturer instructions, with phosphate standard solutions prepared in NTP hydrolysis reaction buffer and absorbance at 620 nm detected on a BioTek Synergy H1 microplate reader.

To detect AC/GC activities of AdvA in vivo, the E. coli CRP/CRP_G reporter system was used as described [38]. E. coli BL21[DE3] cya crp strains harbored a CRP (the cAMP sensor CRP, or dual cAMP/cGMP sensor CRP_G) and a heterologous NC, both IPTG-inducible from plasmids. NC controls included insertless vector (negative control), A. baumannii CavA (AC), and Azospillrillum sp. B510 GuaA (GC). Colonies forming overnight on LBA containing Tc 10μg/ml and Cb 50 μg/ml were patched on reporter plates (LBA containing Tc 10μg/ml, Cb 50 μg/ml, X-gal 40 μg/ml, and IPTG at 25 or 50 μg/ml). Plates were incubated overnight at 37°C, followed by room temperature for one day, and imaged using white-light transillumination.

For ELISAs, strains were precultured in permissive medium (LB or LB supplemented with 5% sucrose) for 3 h, washed 2x with PBS, seeded in LB at OD 0.02, and cultured to OD 0.5. 1 ml of OD 0.5 was pelleted at 3000xg for 3 min. Pellets were resuspended with 1 ml 0.1 M-HCl-0.5% Triton X-100 and incubated for 10 min with end-over-end rotation at room temperature [66]. Samples were centrifuged 3000xg for 3 min and supernatants were collected and stored at −20°C prior to use. cAMP and cGMP ELISA kits (Enzo) were used per manufacturer’s specifications.

Protein structural predictions and homology analysis

Alphafold 3 [35] was used for protein structural predictions. Dali [32] and Foldseek [67] were used for structural similarity searches. PDBsum was used to generate protein topology maps [68]. AdvA orthologs were identified using BLASTp [69] using query sequence WP_002000748.1, database refseq_select, maximum E value of 0.01, and minimum coverage of 75%. FastTree [70] was used to generate a phylogeny from the BLASTp hits. Custom scripts were used to assign BLOSUM62 matrix scores to positions in orthologs relative to WP_002000748.1 in an Interactive Tree of Life (iTOL)-compatible format. The phylogeny and heatmap data were visualized using iTOL [71].

Supplementary Material

Supplement 1

S1 Fig. Evaluation of gene-swap system under control conditions. (A) Colony growth of A. baumannii AdvA allele swap strains harboring insertless P(Ara) vector or vector with WT advA swap gene. Solid medium contained IPTG 0 or 500 μg/ml to control expression of the native advA gene, and Ara 0 or 0.5% to control expression of the plasmid-borne swap allele. (B) Colony growth of allele swap strains from Fig 1 on control medium (with or without IPTG 500 μg/ml). (C) Microscopy analysis of the A. baumannii AdvA swap strains from Fig 1 after 1 h incubation in LB IPTG 0, Ara 0.5%. Cell length was determined from phase-contrast micrographs as in Fig 1D (n ≥ 234 cells per strain). Each symbol represents one cell; blue line, median; -, strain contained an insertless vector (no swap allele).

S2 Fig. Western blot analysis of AdvA variants in A. baumannii. Gene swap strains harboring the indicated AdvA truncation/deletion construct (A, B) or point mutant (C) were cultured in LB Ara 0.5% as in Fig 1. Samples were collected at OD 1 and analyzed by SDS-PAGE and Western blot using anti-GFP primary antibody (top) and SYPRO Ruby to detect total protein (bottom).

S3 Fig. Additional analysis of AdvA truncations/deletions: (A) untagged versions behave identically to GFP-tagged constructs; and (B) non-complementing constructs lack dominant-negative effects. (A) Untagged versions of the non-complementing AdvA variants were tested in the A. baumannii gene-swap system for colony formation. Strains were precultured in IPTG 500 μM, dilutions were spotted on LBA with or without IPTG 500 μM or Ara 0.5%, and colony growth was analyzed as in Fig 1. (B) The AdvA-msGFP2 fusions from Fig 1 (top) and the untagged AdvA constructs from panel A (bottom) were tested in a WT (AdvA⁺) A. baumannii background for colony formation. Strains were precultured in LB, dilutions were spotted on LBA with or without Ara 0.5%, and colony growth was analyzed as in A.

S4 Fig. Western blot analysis of AdvA depletion using the conditional P(IPTG)-advA allele. AdvA protein levels were measured in A. baumannii strains harboring the conditional advA allele [P(IPTG)-advA] compared to the WT control. Strains also harbored an insertless plasmid vector (pYDE334), such that no AdvA-GFP swap construct is present. WT control was precultured in LB and P(IPTG)-advA was precultured in LB 100 μM IPTG for 3 h. Cells were washed twice with PBS and reinoculated into fresh LB. At 1 h and 3 h, 1 OD unit was collected and analyzed by SDS-PAGE and western blot using anti-AdvA antiserum. (A) Quantification of AdvA protein levels at 1 h and 3 h. AdvA levels were normalized by dividing AdvA band intensity by total protein intensity determined from SYPRO staining and are shown relative to control at 1 h. Bars show mean ± s.d. * P< 0.05, *** P < 0.001, unpaired t test. (B) Representative western blot.

S5 Fig. Analysis of controls for two-hybrid assays. (A) Western blot analysis of AdvA-T18 fusions in E. coli. E. coli BACTH strains harboring the indicating AdvA-T18 fusions and an insertless pKNT25 plasmid were cultured in LB medium with Km (30 μg/ml), Cb (50 μg/ml), and IPTG (500 μM) at 30°C. Samples were collected at OD 1 and analyzed by SDS-PAGE and Western blot using anti-CyaA primary antibody (top) and SYPRO Ruby for total protein (bottom). (B) BACTH analysis in E. coli of interactions between A. baumannii FtsK and candidate A. baumannii divisome partners. Two-hybrid reporter analysis was performed as in Fig 3.

S6 Fig. Sucrose provides a condition permissive to growth of bacteria with the conditional P(IPTG)-advA allele. (A) A. baumannii conditional advA strains [P(IPTG)-advA-gfp] and [P(IPTG)-advA-gfp lacI_2x] were precultured in LB with 100 μM IPTG for 3 h, washed, and dilutions spotted on LBA with or without 500 μM IPTG or 10% sucrose. (B) A. baumannii CRISPRi strains harboring control non-targeting sgRNA or advA-targeting sgRNA were diluted and spotted on LBA containing CRISPRi inducer (50 ng/ml aTc), 10% sucrose, or no additive. (C) Time course of WT vs AdvA-depleted bacterial growth in the presence and absence of sucrose. WT and P(IPTG)-advA strains with insertless P(Ara) vector (pYDE334) were precultured in LB (WT) or LB with 100 μM IPTG [P(IPTG)-advA] for 3 h. At T0, cells were washed twice with PBS and diluted to OD 0.02 in LB with 0.5% Ara (left) or 5% sucrose osmoprotectant (right). Medium lacked IPTG to deplete native AdvA. After 3 h culture, bacteria were diluted again to OD 0.02 in LB with 0.5% Ara (-sucrose, left; +sucrose, right), followed by continued culture for 3 h. Arrows indicate time of addition of Ara inducer and time point used for microscopy in Fig. 4. Data points show geometric mean ± s.d.

S7 Fig. Location and additional suppressor functions of amino acid substitutions in A. baumannii FtsB and FtsW that rescue AdvA⁻ lethality. (A) Ribbon diagrams show AlphaFold 3 predicted structures for FtsB (pTM = 0.42) and FtsW (pTM 0.92). Predicted topology relative to the inner membrane is indicated. Positions and amino acid side chains affected in suppressor mutants are shown in yellow. (B) Analysis of additional suppressor interactions using CRISPRi. Strains with the indicated genotype (left) harbored non-targeting control sgRNA (−) or an sgRNA targeting zipA or ftsK (+). Serial dilutions were spotted on solid medium with or without CRISPRi inducer (aTc, 50ng/ml).

S8 Fig. Analysis of predicted AdvA full-length and dimeric structures. (A) The CyaB dimer crystal structure (PDB: 3r5g, chains A,B) features a wreath-like organization dependent on the dimerization arms (arrows) and containing a central cleft that provides the active sites (dotted rectangles). B-D AlphaFold 3 structural models of theoretical AdvA homodimers (B, D) and full-length protein (C,D). In C, regions corresponding to the AdvA domains in Fig 6A are indicated, and predicted topology is shown based on the DeepTMHMM model. AlphaFold 3 ipTM and/or pTM confidence values are given below each structural prediction.

S9 Fig. Unique topographical features of the AdvA CHD-like domain: lack of canonical AC/GC active site residues and presence of conserved tip region. (A) Stacked sequence logos show conserved residues at structurally aligned positions between AdvA^226-420 (PDB: 7sua), CyaB AC domain (PDB: 3r5g), and the GC Cya2 from Synechocystis PCC6803 (PDB: 2w01), generated using DALI. Height of a stack of letters indicates sequence conservation at that position within each protein family, with letter height reflecting amino acid frequency at the position. Above logos is the secondary structure and residue numbers of AdvA^226-420. Blue brackets indicate conserved residues that contribute to an exposed tip at one end of AdvA. Arrowhead notes residue R324 in AdvA. # indicates active site residues in AC/GCs [33, 37]. (B) Structural alignment of AdvA^226-420 with CyaB showing side chains and labels for 6 of the conserved catalytic residues in CyaB (yellow) and corresponding AdvA residues (blue).

S10 Fig. AdvA CHD-like domain shows no detectable cyclase or NTPase activity under conditions tested in vivo and in vitro. (A) E. coli CRP/CRP_G reporter system was used to detect AC/GC activities of AdvA constructs. E. coli BL21[DE3] cya crp strains expressed WT CRP (sensing only cAMP) or a mutant CRP, CRP_G (sensing both cAMP and cGMP) from a plasmid, and expressed a heterologous NC from a second plasmid. Expression of both components is controlled by IPTG. Blue color develops from CRP or CRP_G-dependent lacZ activation after growth on reporter plates with the indicated concentration of IPTG. Constructs encoding CavA from A. baumannii and GuaA from Azospillrillum sp. B510 were included to serve as positive controls for AC and GC activity, respectively. The two adjacent images above or below the dotted line were from the same plate. (B) ELISAs were used to examine effect of loss of AdvA CHD-like domain on A. baumannii cellular cAMP and cGMP levels. A. baumannii WT control and a mutant lacking the domain [native gene replaced by advA(1-230)] were cultured in LB to OD 0.5, and cell lysates were analyzed for cAMP and cGMP levels by ELISA. Values were normalized to total protein level determined from BCA assay. Bars show means ± s.d. (C-D) Evaluation of AdvA CHD-like domain NC and NTPase activity in vitro. Purified recombinant AdvA^158-435 containing N-terminal His-tag and cleavage site, or buffer alone, was used in NC assay (C) or NTPase assay (D) with the indicated nucleotide substrate. Apyrase was also used as a positive control for NTPase activity. Data were interpolated from phosphate standard curve, with < indicating values were below the lowest standard value, and > above the highest standard value.

S11 Fig. AdvA function is preserved despite substitutions to positions aligning with the canonical AC/GC active site. AdvA swap constructs were analyzed for effect on A. baumannii colony growth in the presence of fluoroquinolone antibiotic. Swap genes were expressed as fusions to msGFP2 via Ara induction as in Fig 1. Dilutions were spotted on LBA 0.5% Ara with or without CIP at 0.0625μg/ml. Colony formation was analyzed as in Fig 1D.

Table S1. Strains used in this study

Table S2. Plasmids used in this study

Table S3. Synthetic oligonucleotides and gene fragments used in this study