DksA is a conserved master regulator of stress response in Acinetobacter baumannii

Abstract Coordination of bacterial stress response mechanisms is critical for long-term survival in harsh environments for successful host infection. The general and specific stress responses of well-studied Gram-negative pathogens like Escherichia coli are controlled by alternative sigma factors, archetypically RpoS. The deadly hospital pathogen Acinetobacter baumannii is notoriously resistant to environmental stresses, yet it lacks RpoS, and the molecular mechanisms driving this incredible stress tolerance remain poorly defined. Here, using functional genomics, we identified the transcriptional regulator DksA as a master regulator for broad stress protection and virulence in A. baumannii. Transcriptomics, phenomics and in vivo animal studies revealed that DksA controls ribosomal protein expression, metabolism, mutation rates, desiccation, antibiotic resistance, and host colonization in a niche-specific manner. Phylogenetically, DksA was highly conserved and well-distributed across Gammaproteobacteria, with 96.6% containing DksA, spanning 88 families. This study lays the groundwork for understanding DksA as a major regulator of general stress response and virulence in this important pathogen.


INTRODUCTION
Acinetobacter baumannii , a ubiquitous Gram-negati v e aerobe, has emerged as one of the most notorious human pathogens for healthcare institutions globally ( 1 ). Recently, it has been recognized by the World Health Organization as one of three top pathogens in critical need of new antibiotic therapies ( 2 ) due to its extremely high levels of antimicrobial resistance ( 3 , 4 ). During its lifetime, A. baumannii must adapt and thri v e in frequently changing stresses, particularly for successful host infection and persistence in the environment ( 5 ). This pathogen displays a remar kab le ability to withstand a wide range of stresses for prolonged periods, including living on desiccated surfaces for up to 250 days ( 6 ), survival in commonly used hospital disinfectants and biocides ( 7 ), as well as tolerance to stresses encountered during host infection like metal toxicity and oxidati v e agents ( 8 , 9 ). A. baumannii 's resilience in harsh environments can be largely attributed to its superior permeability barrier and ability to pump out toxic chemicals via efflux mechanisms ( 4 , 10-13 ), but how stress responses are coordinated remains poorl y understood. Thus, ma pping the molecular mechanisms underpinning various stress tolerance strategies in A. baumannii is crucial to ultimately tackle deadly clinical infections caused by this pathogen.
Bacterial stress response systems are energetically costly, and global defense mechanisms can involve a significant proportion of cell components ( 14 ). Regulation of stress response at a cellular le v el is largely controlled by master regula tors tha t r edistribute the limited stor es of RNA polymerase to transcribe genes involved in survival and / or adaptation ( 14 ). Two major interconnected pathways coordinate bacterial str ess r esponse: the general str ess r esponse, and the stringent response. The general stress response system is regulated by an alternati v e sigma factor of RNA polymerase, RpoS (also called 38 and S ), and is well characterized in Esc heric hia coli and other common Gramnegati v e microbes. RpoS plays a pleiotropic role in the cell, acti vating genes involv ed in metabolism, pr otein pr ocessing, tr ansport, and tr anscriptional regulation during starvation and other environmental challenges (14)(15)(16). Conversely, the stringent response is controlled by guanosine penta-or tetra-phosphate (p)ppGpp, a product of relA / spoT activation (17)(18)(19). Howe v er, A. baumannii and related species in the Moraxella family do not harbor a gene encoding RpoS ( 20 , 21 ) and thus it is likely that additional major stress regulators are acti v e that have not yet been characterized.
To understand how str ess r esponses ar e coordinated and regulated in A. baumannii , we investigated two biolo gicall y important metal stresses: excess copper and zinc. These metal ions are essential in all forms of life including in bacterial pathogenesis ( 8 ), yet become toxic at high concentrations ( 22 ). Host immune responses cle v erly e xploit both the essentiality and toxicity of copper and zinc ions to clear invading bacteria and pre v ent infection ( 8 ). In this study, we use a functional genomics technique, transposon insertion sequencing ( 23 ), to identify genes influencing the fitness of A. baumannii under copper and zinc stress, uncovering roles for efflux, membrane and envelope biogenesis. We pinpoint a previously o verlook ed global regulator DksA as the major coordinator of stress response in A. baumannii . Using transcriptomic and phenotypic profiling, we re v eal how DksA acts as a switch between the two metal stressors by regula ting transla tional machinery and metabolism. We outline a vital role of DksA in protection against other infection r elevant str essors, in vivo host infection, maintaining muta tion ra tes and desicca tion, and r etaining antibiotic r esistance. Finall y, phylo genetic studies confirm its wide conservation across Gammaproteobacteria and molecular studies point to key differences between DksA in A. baumannii and E. coli . Together, our results demonstra te tha t DksA is a crucial, conserved component of stress protection regulation in A. baumannii .

Bacteria strains, media, and growth conditions
The wild-type A. baumannii strains used were ATCC 17978 (NCBI accession number: CP012004.1) and AB5075 UW (NCBI accession number: CP008706.1). The Tn 26 insertion mutant deri vati v es of AB5075 UW were purchased from the Manoil Laboratory ( 24 ) and used for individual growth assays to validate TraDIS results. A total of 28 mutants were used for the individual growth assays. The dksA ::kan mutant deri vati v e of ATCC 17978 was constructed for this study using the previously published protocol ( 25 , 26 ). To confirm that both the ATCC 17978 and AB5075 UW dksA:: Tn 26 mutants contained no secondary mutations, we performed whole genome sequencing on each single gene mutant ( > 20 × coverage on an Illumina MiSeq platform) and employed the Snippy pipeline version 4.3.6 ( https:// github.com/tseemann/snippy ) to ensure no additional mutations were present. All primers used in this study are listed in Supplementary Table S4. All chemicals used in this study were obtained from Sigma-Aldrich (Australia) unless otherwise stated.
For routine overnight culturing of A. baumannii strains, a single colony from cation (calcium and magnesium ions) adjusted Mueller Hinton II (MH) agar plate (Becton Dickinson, USA), containing beef extract (3.0 g), acid hydrolysate of casein (17.5 g), starch (1.5 g) and agar (15 g) per litre of deionised water was used to inoculate 5 ml of MH broth medium and grown for 16 h at 37 • C with shaking.

Construction of transposon mutant library
The ATCC 17978 A. baumannii dense transposon library used in this study was constructed using the protocol as previousl y described ( 27 ). Briefly, transposomes wer e pr epar ed by using an EZ-Tn 5 transposase (Epicentre Biotechnology) and a custom Tn 5 transposon carrying a kanamycin resistance cassette amplified from the plasmid pUT Km ( 28 ) using the primer set as described previously ( 29 ). The transposomes (0.25 l) were electroporated into 60 l of fr eshly pr epar ed electrocompetent cells using a Bio-Rad GenePulser II set to 1.8 kV, 25 F and 200 in a 1-mm electrode gap (Bio-Rad). For preparation of electrocompetent cells, 125 ml cultures were grown in 500 ml baffled flasks at 37 • C in an Infor HT shaking incubator (Switzerland) at 200 rpm until they reached mid-log phase i.e. optical density at 600 nm (OD 600 ) = 0.5. The cultures were then placed on ice for 15 min with occasional swirling before centrifugation for 10 min at 4 • C, washed twice with ice-cold 10% glycerol in MilliQ (MQ) w ater. The w ashed electrocompetent cells were then resuspended with 150 l of ice-cold 10% glycerol. The cells were resuspended in 1 ml of SOC medium and incubated at 37 • C with shaking at 200 rpm for 2 h then spread on MH-agar supplemented with 7 g / ml kanamycin (Sigma-Aldrich, Australia). Usually, 12-16 transformations were performed for each batch. Number of transformants in each batch ranged from 10000 to 50000. A pproximatel y 250 000 m utants were collected from a total of 10 batches and stored as glycerol stocks a t −80 • C .

T r ansposon mutant libr ary metal str ess challenge and tr ansposon-dir ected insertion site sequencing (T r aDIS) of mutant library
A pproximatel y 10 9 viable mutant cells were inoculated into 10 ml MH broth and grown at 37 • C for 8 h with shaking (200 rpm). The culture (500 l) containing a pproximatel y 10 9 cells was sub-cultured into 10 ml fresh MH broth with or without 6 mM CuSO 4 or 3 mM ZnSO 4 in duplicate and grown for 16 h at 37 • C with shaking (200 rpm). Genomic DNA was then extracted from approximately 10 10 cells using the DNeasy UltraClean Microbial Kit (Qiagen) according to the manufacturer's protocol. Sequencing and analysis of the transposon mutant library were performed as described previously ( 30 , 31 ). The primer sets used for PCR amplification of Tr aDIS fr agments and sequencing were described previously ( 29 ). Samples were sequenced on a HiSeq2500 Illumina sequencing platform at the Wellcome Sanger Institute, generating a pproximatel y 2 million 50 bp single-end reads per sample as previously described. TraDIS sequence r eads wer e deposited in the European Nucleotide Archi v e under accession number ERP118051 and analysed using the BioTraDIS pipeline with default parameters as described previously ( 31 ). The final ATCC 17978 Tn 5 library density was > 110 000 unique mutants.

Time kill assay for the selection of copper and zinc concentr ations f or mutant libr ary challenge
To identify subinhibitory concentrations of CuSO 4 and ZnSO 4 for treatment of the Tn 5 transposon library we performed time kill assays. A pproximatel y 10 9 cells from an overnight culture of ATCC 17978 was sub-cultured into 10 ml MH broth spiked with varying concentrations of CuSO 4 (0, 3, 6, 8, 16 and 24 mM final concentration) or ZnSO 4 (0, 3, 4, 8, 16 and 24 mM final concentration) and incubated at 37 • C with shaking. At 0, 1, 2, 4, 5 and 24 h time points, 100 l samples were taken and 10-fold serially diluted in sterile PBS and 10 l of each dilution was then spotted on MH-agar pla tes. Pla tes were incuba ted a t 37 • C overnight and colonies were enumerated to determine the surviving cells.

T r anscriptomic analyses
Three independent cultures of A. baumannii strain ATCC 17978 and its dksA mutant were grown overnight in 5 ml MH broth with shaking at 200 rpm at 37 • C. The overnight cultur es wer e diluted 200-fold in fr esh MH br oth and gr own to mid-log phase (OD 600 of 0.55). Each culture was divided into three flasks, with two cultures treated with either 6 mM CuSO 4 or 3 mM ZnSO 4 and one left untreated as a control and grown for 40 mins. RNA extraction was carried out using the miRNeasy mini kit (Qiagen) and DNA was eliminated using the TURBO DNA-free kit (Ambion Inc., USA) as per manufacturer's instructions. Libraries were constructed using a Uni v ersal Prokaryotic RNA-Seq and Prokary otic An yDeplete ® Libr ary prepar ation kit (Tecan, USA) as per to the manufacturer's instructions. The samples were sequenced on Novaseq Illumina platform, producing ∼3 million 150 bp paired-end reads per sample ∼ 25 Gbp in total. The raw sequencing data was deposited under GEO accession number GSE169081. Reads were quality controlled using FastQC and trimmed using bbduk (v38.79) with the included adapters.fa file and parameters ktrim = r k = 23 mink = 11 hdist = 1 qtrim2 = t trimq = 10 tpe tbo. Reads were then mapped using bbmap (v38.79) with parameters k = 13 and ambig = toss against the A. baumannii genome (accession CP000522) and plasmids (accessions CP000523, CP012004, CP012005), sorted using samtools (v1.6), and quantified using HTSeq (v0.12.4) with default parameters. Read counts were aggregated using a custom perl script and used as the basis for differential expr ession analysis. Differ ential expr ession analysis was performed in R language, using the edgeR package (v3.30.3) using the quasi-likelihood fit and test functions (glmQL-Fit, glmQLFTest). Genes differ entially expr essed, as defined by > 3-fold change and P adj < 0.05, are listed in Supplementary Table S2. The function of genes in A. baumannii were allocated using eggNOG-mapper, and the resulting gene ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways were used for gene set enrichment analysis (GSEA) using Fry, a fast approximation of the ROAST gene set test included in the edgeR package.
For visualising metabolic pathways in our RNA sequencing data, we used the Omics Dashboard Tool from both EcoCyc ( 32 ) and MetaCyc ( 33 ). Enrichment or depletion of metabolic pathways was then analysed using the Fisher's exact test hypothesis and significant values of < 0.05. Enrichment or depletion scores ( −log 10 P values) for each pathway in the dashboard were downloaded, and figur es wer e cr eated using the GraphPad Prism software (Graph-Pad Software Inc). We also downloaded tables showing list of genes from the dashboard and calculated the percentage of transcripts that increased or decreased, as shown in Figure 5 D. We could not map 1338 genes out of 2470 significant up or downregulated in at least one condition due to a lack of functional annotation.

Animal infection experiments
The Galleria mellonella infection experiments were performed as previously described ( 34 ). Briefly, triplicate assays of 5 larvae (200-230mg) were injected with 1 × 10 7 cells of A. baumannii strains AB5075 UW or ATCC 17978 and their dksA mutants (total n = 15 per strain). Survival and health of larvae were assessed and scored e v ery 24 h postchallenge for 7 days according to the G. mellonella Health Index Scoring System ( 35 ).
The in vivo mice model was used for enumeration of A. baumannii AB5075 UW and the dksA mutant in different host niches: blood, nasopharyngeal tissue, bronchioalveolar lavage, BAL, lung tissue, pleural cavity, PL, spleen tissue and li v er. Female BALB / c mice were intranasally challenged with 2 × 10 8 colon y f orming unit (CFU) and colonization was examined 24 h post-challenge using bacterial plate counting, as previously described ( 36 ). All procedures performed in this study were conducted with a view to minimize the discomfort of the animals. All experiments are approved by the University of Adelaide Animal Ethics Committee (Animal Welfare Assurance number A5491-01; pr oject appr oval number S-2019-080) and were performed in strict adherence to guidelines dictated by the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes.

Mutant phenotypic assays
For all growth phenotypic assays, a single colony from Lysogeny broth (LB) agar plates was used to inoculate 5 ml of LB broth medium. Overnight cultures were diluted to an OD 600 of 0.01 in 105 l LB broth with or without stress treatments in 96-well plates. We supplemented ZnSO 4 (1.5 mM), CuSO 4 (3 mM or 5 mM) and H 2 O 2 (0.5 mM) in LB medium for zinc, copper, and oxidati v e stresses respecti v ely. For all growth assays, cultures were incubated at 37 • C for 16 h with shaking at 200 rpm in a PHERAstar FS Spectrophotometer (BMG Labtech). Cell growth was monitored at 0.1 h intervals by measuring OD 600 . Growth curves were used to calculate area under curve (AUC) using the GraphPad Prism software. The difference in AUC between WT and mutants was then used as a proxy for fitness under different stress conditions.

Mutation rate assay
Acquisition of resistance to rifampicin (Rif R ) from rifampicin-sensiti v e (Rif S ) A. baumannii strain AB5075 UW and its dksA ::Tn 26 mutant (Rif S → Rif R assay) was used to determine muta tion ra tes. A single colony of each strain was inoculated in 5 ml MH broth and allowed to propagate overnight at 37 • C with shaking at 200 rpm. The overnight cultures were diluted in 5 ml of fresh LB medium and allowed to grow to an OD 600 of 0.6. The exponentially growing cultures were further diluted 10 000-fold and 150 l distributed in 20 wells in 96-well pla tes and incuba ted a t 37 • C with shaking a t 200 rpm. Aliquots (100 l) were then plated on rifampicin MH agar pla tes. The concentra tion of rifampicin in plates was 25 g / ml rifampicin (Sigma-Aldrich). The plates were then incubated for 24 h at 37 • C to detect Rif R mutant colonies. For CFU counts, aliquots of a ppropriatel y diluted cultures were plated on MH-agar plates. The mutation rates were then estimated from the number of resistant colonies per culture and the total CFU counts by using the Luria-Delbr ück fluctuation test ( 37 ) and Ma-Sandri-Sarkar maxim um likelihood anal ysis ( 38 ). The fluctuation anal ysis calculator (FALCOR) w e b tool was used for the analysis ( 39 ).

Desiccation assay
Stationary phase (overnight grown) cells were harvested from 1 ml samples of MH cultures by centrifugation, and then were washed twice with 1 ml of sterile PBS and resuspended with MQ w ater. MQ w ater w as used to pre v ent additional osmotic stress during drying of the cell suspensions. Cell suspensions in MQ water were adjusted to an OD 600 of 5.0, and then 10 l droplets of each adjusted suspension were deposited onto a plastic (polystyrene) surface in 24-well sterile plates. The samples w ere allow ed to dry for a pproximatel y 1 h in a biosafety cabinet at ambient temperature.
To estimate the survival time of A. baumannii cells, dried samples were incubated in a desiccator at ambient temperature in dark. The initial number of viable cells was determined by plating 100 l of serially diluted cultures on LB pla tes in triplica te. To determine viability after drying, 200 l of PBS was added onto each dried sample. The samples wer e r ehy drated b y incuba ting a t room tempera ture for 10 min, mixed thoroughly by pipetting the suspensions up and down. The suspended cultur es wer e serially diluted in PBS and 10 l of each dilution was then spotted onto LB agar pla tes. The pla tes were incuba ted a t 37 • C overnight. The viable cells on the dried surface were then inferred from the number of CFU r ecover ed from each dried sample. To determine the length of survival time of desiccated samples, six dried samples were sampled e v ery 1 or 2 days for 58 days.

Biolog phenotypic micr oarra y
The phenomes of A. baumannii ATCC 17978 and its dksA mutant were assayed with the Biolog Phenotype MicroAr-rayTM (PM) system ( 40 ) to identify compounds that could serve as sole carbon sources (PM1-2; 190 compounds). Additionally, sensitivities to stress conditions (PM9-10; 192 conditions) were also tested. All phenotypic tests were performed as per manufacturer's instructions. Following inocula tion, all PM pla tes were incuba ted in an OmniLog reader (Biolo g) aerobicall y at 37 • C for 48 h. Reduction of the tetr azolium-based dy e (colour less) to formazan (violet) was monitored and recorded at 15 min intervals by an integrated charge-coupled device camera. The resultant data were analysed with the supplied manufacturer's software, resulting in a time-course curve for colorimetric change equating to respiration rate.

Respiration activity assay
For respiration assays, wild-type ATCC 17978 and dksA mutant cells in 5 ml MH broth were grown to mid-log phase (OD 600 = 0.5) at 37 • C with shaking at 200 rpm and treated with 6 mM CuSO 4 or 3 mM ZnSO 4 for 40 min, 1 ml cultur es wer e centrifuged for 1.5 min and resuspended with fresh MH medium containing 0.1% tetrazolium dye and chlor amphenicol (200 g / ml). Chlor amphenicol was used to pre v ent further protein synthesis allowing us to capture respira tion sta tus during 40 mins of copper or zinc treatment. 150 l of cells were then transferred into 96-well culture plates. The plates were incubated in an OmniLog reader (Biolog) aerobically at 37 • C for 6 h. Reduction of the tetr azolium-based dy e (colour less) to formazan (violet) was monitored and recorded at 15 min intervals by an integrated charge-coupled device camera. The resultant data were analysed with the supplied manufacturer's software as in the Biolog phenotypic microarray assay.

Serum growth inhibition assay
For the serum growth inhibition assay 10 5 CFU in 10 l fr om exponentially gr owing cells in MH wer e transferr ed into 100 l of 50% serum in MH plus 0.1% tetrazolium dye in 96-well micropla tes. The pla tes were then incubated in an OmniLog reader (Biolog) aerobically at 37 • C for 48 h.
Nucleic Acids Research, 2023, Vol. 51, No. 12 6105 Reduction of the tetr azolium-based dy e (colour less) to formazan (violet) was monitored and recorded at 15 min interv als b y an integrated charge-coupled device camera. The resultant data were analysed with the supplied manufacturer's software.

Biofilm formation and capsule synthesis
For biofilm f ormation assa y we used the previously published method ( 41 ). Briefly, overnight cultur es wer e diluted 100-fold in 100 l LB broth in 96-well dish. Cells were then incubated for 24 h at 37 • C without shaking. Bacterial cells wer e r emoved by pipetting and washed thr ee times with PBS to remove unattached cells. We then added 125 l of a 0.1% crystal violet (CV) aqueous solution and incubated for 15 mins at room temperature. After rinsing 3 times with water and drying for 2 h, 125 l of 30% acetic acid in water was added to each well, incubated for 15 mins to allow complete solubilisation of CV and 125 l of solubilised CV was transferred a new flat bottom microtiter plate. Biofilm formation was then estimated by measuring absorbance in a plate at 550 nm using 30% acetic acid solution as a blank.
For qualitati v e estimation of capsule le v els, we used a density gradient centrifugation method as previously described ( 42 ), which is based on the effect of cell-associated capsule on bacterial density. Briefly, 1 ml of overnight grown cultur es wer e centrifuged, washed with PBS and r esuspended in 1 ml PBS. The OD 600 of the cell suspensions was then adjusted to 1, translating to a pproximatel y 8 × 10 8 cells / ml, and 400 l of the cell suspensions were loaded gently on the top of a solution of 37.5% (AB5075 UW) or 47.5% (AT CC17978) Per coll in PBS. A second layer of 60% P ercoll w as included to aid visualisation of the cells following centrifugation. The tubes containing biphasic Percoll solution and cell suspension were centrifuged for 5 mins at 3000 x g.

Minimal inhibitory concentration (MIC) assay
The three wild-type strains ( A. baumannii AB5075 UW, ATCC 17978, and E. coli K-12) and their dksA single gene knockouts were streaked from frozen stocks on an MH pla te overnight a t 37 • C . A single colony was inoculated in 10 ml of MH in a 50 ml falcon tube and shaken at 200 rpm in 37 • C until an OD 600 of 0.5 was reached. Antibiotic twofold dilutions were prepared in triplicate in 96-well plates to a volume of 140 l using a multichannel pipette. A 1 / 400 dilution was made in PBS for each of the cultures once they had reached OD 600 of 0.5. 15 l of the culture dilutions was dispensed into each well, bringing the final volume to 155 l. Each plate was covered with an AeraSeal ™ film (Sigma Aldrich, cat. A9224-50EA) and incubated at 37 • C for overnight with shaking (200 rpm). Plates wer e r ead at OD 600, and MICs were reported at the lowest concentration where the majority of wells had 80% growth inhibition compared to the positi v e control.

Gentamicin uptake assay
The gentamicin accumulation assay was performed using the method described previousl y ( 43 ). Briefly, A. baumannii AB5075 UW wild-type and its dksA ::Tn 26 mutant were grown to OD = 0.6 in MH broth. Culture aliquots (500 l) wer e transferr ed to 2 ml sterile Eppendorf tubes, and gentamicin-Texas Red conjugate was added to each sample at a final gentamicin concentration of 500 g / ml. Reactions were protected from light and incubated for 30 min at 37 • C with shaking at 200 rpm. Cells were then pelleted by centrifuga tion a t 8000 g for 1 min, washed with 400 l PBS, and the pellet resuspended in 1 ml DMSO and stored at −20 • C prior to measurement.
Photophysical measurements were performed with a FLS980 photoluminescence spectrometer (Edinburgh Instruments) equipped with a Xe1 Xenon Arc Lamp (450 W ozone free, excitation range 230-1000 nm) for steady-state measurements. Excitation (lex) was performed at 550 nm, and emission spectra were recorded in DMSO at 28 • C with 1 nm step-size, 0.1s integration time, and slit-width of lex = lem = 1.5 nm for both strains.

Complementation of dksA gene homologs
Three dksA gene constructs were designed for complementation in A. baumannii cells. These include A. baumannii AB5075 UW full-length dksA (FL-Ab; amino acids 1-178), A. baumannii AB5075 UW truncated dksA (Tr-Ab; amino acids 46-178) and E. coli MG1655 full-length dksA (FL-Ec; amino acids 1-157). The gene fragments wer e pur chased from Integrated DN A Technolo gies (Supplementary Table  S5). Each gene fragment was cloned in the pVRL2Z (using EcoRI and NotI restriction enzymes) plasmid. Insertion of the gene fragment in the respecti v e plasmids was confirmed via PCR followed by Sanger sequencing using sequencing primers specific for each plasmid (Supplementary Table S4).
The pVRL2 plasmid containing the cloned dksA sequences (FL-Ab, FL-Tr and FL-Ec) under an arabinose inducible promoter were transferred into A. baumannii AB5075 UW lacking dksA by electroporation as described previously in ( 44 ). The complementation of dksA (FL-Ab, FL-Tr and FL-Ec) was investigated by performing growth phenotypic assays in LB with or without the addition of added stresses. For oxidati v e and zinc stresses, we used 0.5 mM H 2 O 2 and 1.5 mM ZnSO 4 respecti v el y, w hile for antibiotic stress we used rifampicin (0.4 g / ml final concentration). For all complementation experiments we also added 0.5% arabinose in the growth medium.

Expr ession, purification, and differ ential scanning fluorimetry for thermal melting assay of DksA
A. baumannii full length dksA was cloned into the pOPINE-3C-eGFP plasmid using NcoI and PmeI restriction enzymes. Correct insertion of the gene fragment was confirmed via PCR (primers in Supplementary Table S4) followed by Sanger sequencing. The DksA protein was expressed in E. coli BL21 cells and purified as previously described ( 45 ) using IMAC and SEC procedures in the presence of reducing agent tris(2-carboxyethyl)phosphine (TCEP, 1 mM). For ta g cleava ge (GFP-His 6 ), the pooled protein fractions post-SEC were incubated with HRV-3C protease overnight at 4 • C. The cleaved DksA protein was subsequently r ecover ed using r e v erse IMAC. The purity of the protein was verified using sodium dodecyl sulfate polyacrylamide gel electrophoresis, showing a single band at ∼21 kDa when visualised with Coomassie blue dye.

Scr eening f or dksA and rpoS in Gammaproteobacteria
From the 104 665 Gammaproteobacteria genomes in the Genome Tax onom y Database (GTDB) release 202 ( 46 ), 1686 genomes were selected as representati v e genomes -one for each genus within the class, and one for each species in the Acinetobacter and Moraxella genera. Each genome was screened for DksA (accession AKA33312.1 from A. baumannii ) and RpoS (accession NP 417221.1 from E. coli ) using BLASTp ( 47 ). We consider ed DksA pr esent if the hit had ≥ 40% amino acid identity and ≥50% coverage; RpoS was consider ed pr esent if the hit had ≥ 50% amino acid identity and ≥50% coverage. These cut-offs were selected after visualising the distribution of identity and coverage values for each gene (Supplementary Figure S4A, B). Due to high sequence similarity of DksA with TraR (Supplementary Figure S4C) and RpoS with other sigma factors, sequence similarity alone was not enough to distinguish DksA from TraR and RpoS from other sigma factors. Since DksA is relati v ely larger than TraR and RpoS is well known for its size of 38 kDa, we also use size information when distinguishing these proteins from other proteins. For RpoS amino acid length between 288-400 was chosen whereas for DksA ≥ 118 was chosen for cut-offs. To visualise the distribution of DksA and RpoS across the phylogeny of Gammaproteobacteria, we subset the GTDB v202 bacterial phylogeny using ape v5.6 ( 48 ) to select only genomes we screened. The resulting phylogeny was visualised in R using g gtr ee v3.0.4 ( 49 ) and g gtr eeExtr a v1.2.3 ( 50 ).

Genes r equir ed f or copper and zinc str ess toler ance in A. baumannii
To identify the genetic networks important in A. baumannii for survival of two infection-relevant stresses, copper and zinc, we employed the fitness-based functional genomics technique, transposon directed insertion-site sequencing (TraDIS) ( 23 , 30 ). A high-density, r andom tr ansposon library was generated in A. baumannii wild-type (WT) strain ATCC 17978 containing > 110 000 unique Tn 5 mutants and challenged with subinhibitory le v els of copper (6 mM) or zinc (3 mM) for 16 h (Supplementary Figure S1A, B). TraDIS sequencing was performed as previously described ( 27 ) and analyzed using the TraDIS Toolkit ( 31 ) to determine relati v e insertion frequencies. Non-essential genes w hose m utants had decr eased in abundance r elati v e to untreated controls were considered necessary for metal-stress tolerance and those w hose m utants increased in abundance as metal stress sensiti v e (using standar d cut-offs of log 2 FC and P adj < 0.05).
The TraDIS analysis under copper stress identified 45 tolerance genes with decreased mutant fitness and 32 sensitivity genes with increased mutant fitness (Figure 1 A, Supplementary Table S1). Under zinc stress, 92 tolerance genes and 31 sensitivity genes were identified (Figure 1 B, Supplementary Table S1). As a sanity-check of our TraDIS genotype-phenotype screens, known metal tolerance genes were identified among the mutants with decreased abundance, such as the copper exporter copAB in copper treated samples (Figure 1 C) and czcABCD transport genes in the zinc treated samples (Figure 1 D) ( 51 , 52 ). The phenotypic growth of these r epr esentati v e control genes was validated using defined copA and czcD Tn 26 insertion mutants in A. baumannii strain AB5075 UW ( 24 ), with and without copper and zinc treatment in LB. As expected, altered growth was observed only in the presence of their respecti v e metals (Figure 1 E, F) and no drastic growth defect was observed compared to WT in untrea ted LB , confirming their role as specific metal resistance genes.
Next, we validated the copper and zinc growth phenotypes of a di v erse collection of genes identified from the TraDIS analysis ( n = 26) that were not previously associated with metal resistance, using single mutants in the AB5075 UW background ( 24 ) (see Supplementary Figure  S2  Besides the known copper and zinc efflux genes, the TraDIS analysis also identified se v eral genes involved in other cellular functional groups ( 53 ) including cell wall / envelope / membrane biogenesis, and global regulators involved in translation and ribosome synthesis for both copper and zinc stresses (Supplementary Figure S3). These data indica ted tha t no single pa thwa y can fully account f or the A. baumannii metal tolerance profile, and that multiple layers of gene regulation are required for adaptation to metal stresses. While the copper and zinc stress responses involved distinct gene networks, we also detected a subset of overlapping metal tolerance genes. For example, Tn 5 insertions in genes associated with membrane integrity and capsule synthesis ( wzb , galU, pgi and lptE ) were depleted in both copper and zinc str ess (Figur e 2 A, Supplementary Table S1). Similarities of metal sensitivity genes were also observed, for instance, disruption of dcaP, an outer membrane pore f orming protein f or nutrient uptake ( 54 ), increased tolerance to both copper and zinc (Figure 1 G, H; Supplementary Figure S2). Interestingly, disruption of ompA resulted in increased tolerance only to zinc stress ( Figure 1H; Supplementary Figure S2).

DksA is a pleotropic global r egulator f or coor dinating metal str ess r esponse
The ability of the cell to defend itself from metal stresses r equir es not only the activation of stress-specific genes, but also global regulators that coordinate str ess r esponse. In addition to stress-specific mediators, such as metal-specific efflux pumps, we identified three global regula tors tha t potentially play an important role in adaptation to metal stresses in A. baumannii . They are DksA, HvrA (homologue to H-NS) and the GacS / A two component system (Figure 2 A). Interestingly these regulators showed opposite roles in copper and zinc stresses (providing tolerance to one and sensitivity in the other), potentially acting as switches. Since host immune cells exploit both the essentiality and toxicity of zinc and copper during infection ( 8 , 51 ), it is crucial to understand how these global regulators respond to differ ent metal exposur es and coordinate str ess protection and virulence in A. baumannii .
While GacS / A and H-NS have been studied extensively in A. baumannii and are known to be dynamic coordinators of stress tolerance , virulence , motility, and antibiotic resistance ( 10 , 55 , 56 ), the molecular role of DksA is largely uncharacterized in A. baumannii and may play a major role in str ess r esponse. Intriguingly , our T raDIS data suggests that DksA may act in a pleotropic manner, having opposite effects in the two distinct metal conditions ( Figure 2B; increased insertions in copper yet decreased insertions in zinc). Phenotypic fitness assays of a targeted dksA ::Tn 26 mutant of A. baumannii strain AB5075 UW confirmed that DksA disruption is deleterious to the bacteria under zinc str ess (Figur e 2 D), wher eas it is beneficial under copper str ess (Figur e 2 E). We also noticed that the dksA mutant had a comparable growth rate to WT but reached stationary phase much earlier than WT with a significantly lower growth yield (Figure 2 C). Further, relA , which is responsible for the biosynthesis of ppGpp and mediates the stringent response, was detected as being important in zinc stress, but not copper (Supplementary Table S1). Both the TraDIS and fitness assays of a targeted relA ::Tn 26 mutant showed that deletion of relA is deleterious under zinc stress (Supplementary Figure S2C), consistent with the effects of disruption of dksA and suggesting that ppGpp and DksA are important during metal stress. Taken together, we reasoned that DksA could play a key role in survival under multiple stresses that had not yet been fully defined in A. baumannii . Ther efor e we investigated the molecular mechanism by which DksA regulates metal stress using a suite of di v erse phenotypic and genomics analyses.

The role of DksA in virulence and colonization in animal models
The role that DksA plays in A. baumannii virulence was initially tested by employing the Galleria mellonella wax-moth insect model, which has been shown to be an effecti v e in vivo platf orm f or molecular pathogenicity studies ( 57 ). Infection assays of G. mellonella using two different strains of A. baumannii, ATCC 17978 and AB5075 UW and their respecti v e dksA mutants, were performed in triplica te ba tches of larvae, as previously described ( 34 ). The dksA mutants of both A. baumannii strains killed significantly fewer larvae than their WT strains, indicating that an intact DksA is r equir ed for virulence (Figure 3 A). These results spurred us to investigate the role of DksA in infection of a mammalian host. For this, we intranasally challenged BALB / c mice with A. baumannii strain AB5075 UW or its dksA ::Tn 26 mutant deri vati v e. After 24 h the mice were sacrificed, organs were removed, and bacterial load counted. Strikingl y, dksA m utant cells could not be r ecover ed from the blood of any mice ( < 10 2 cells / ml), compared to 2.5 × 10 6 cells / ml for the WT (Figure 3 B). For all tissues (Figure 3 C-G) the dksA mutant could still colonize but showed a significant reduction in bacterial load compared to the WT, except for the liver (Figure 3 H). Recovery of the dksA mutant from the respir atory tr act (nose, bronchoalveolar and lung tissue), was at least 2 orders of magnitude lower than that seen for WT cells (Figure 3 C-E). Our results also confirm the previously published INSeq transposon insertion sequencing based study in which abundance of dksA insertion was decreased following mouse lung infection ( 58 ). A recent study using a mouse model also showed that DksA is r equir ed for A. baumannii infection in the lungs but, in contrast to our animal model and serum sensitivity results, no difference in infection rates in the blood survival was observed ( 59 ).
To further understand the differences in the observed lack of ability of the dksA mutant to survi v e in the blood compared to other tissues, we performed in vitro virulence assays on both A. baumannii strains (ATCC 17978 and AB5075 UW) and their dksA mutants. First, we tested the mutants' ability to propagate in human serum, which we found was greatly reduced for both dksA mutant strains (Figure 3 I), with ATCC17978 showing an inability to grow in serum. Next, we tested the mutants' ability to form biofilm and produce capsule and found that both were incr eased compar ed to WT (Figur e 3 J, K). AB5075 UW is known to produce a thick protecti v e capsule ( 60 ) and this may be one of the reasons that its dksA mutant still retains partial survival in serum. Taken together, these data show that DksA is r equir ed for serum resistance and ultimately to infect the bloodstream, but it seems to r epr ess other virulence determinants, such as biofilm. We specula te tha t the increase in biofilm density resulting from dksA disruption is what allows this mutant to retain the ability to colonize tissue, albeit not as well as WT despite being undetectable in the blood. This may be consistent with a planktonic lifestyle predominating in the blood, where enhanced biofilm formation of the dksA mutant may not aid colonization.

T r anscriptomics to define the DksA-dependent str ess r esponse on a molecular level
To identify the molecular mechanism underlying the dynamic role of DksA in stress protection and virulence in A. baumannii , we conducted RNA-sequencing (RNAseq) on the ATCC 17978 dksA mutant and WT, treated with or without copper or zinc str ess. Differ ential expr ession of 13.2% (504) of the total genes in the ATCC 17978 genome was observed for the dksA mutant compared to WT, indica ting tha t loss of DksA af fects a large proportion of genes, e v en without stress induction (using a cut-off of log 2 FC > 1.5 change and P adj < 0.05, Supplementary Table  S2). The e xpression le v els of the two adjacent genes either side of dksA ( nudF, cpdA and gluQ, ftsW ) showed no significant expression change in our transcriptomic analysis, indica ting tha t ther e wer e no polar effects arising from deletion of dksA . Under copper and zinc stress, the number of differ entially expr essed genes incr eased to > 1 / 5 of all genes (22.6% (859) genes and 23.6% (898) genes, respecti v ely) for the dksA mutant compared to the control, suggesting that DksA is a master stress regulator in A. baumannii .
Numerous pathways were found to be heavily impacted under copper and zinc stress including translation, respiration, ATP synthesis, amino acid synthesis, aromatic compound degradation, co-factor synthesis, nucleoside and nucleotide synthesis and oxidati v e stress protection as in both the dksA and WT strains (Figure 4 A), based on gene ontology classification ( 33 ). Transcriptional changes in the copper-treated dksA mutant differed markedly from the copper-treated WT and in some cases had opposite effects. For e xample, e xpression of genes involved in aromatic compound degradation increased under copper stress in the WT strain but decreased in the dksA mutant under both treated and untreated conditions (Figure 4 A, 1st, 2nd and 3rd panels from left). Conversely, expression of genes involved in protein translation decreased under copper stress in the WT strain but increased in the dksA mutant (Figure 4 A, 2nd and 3rd panels from left). The translation pathway includes ribosomal-protein (r-protein) genes; expression of these genes largely decreased in WT cells under copper str ess (Figur e 4 B middle panel). Howe v er, in the dksA cells expression of these genes was increased (log 2 FC 1.5-3.7) with or without copper treatment (Figure  4 B, top and middle panels). Curiously, under zinc stress the transcription of r-protein genes was relati v ely unaffected in both WT and dksA cells (Figure 4 B, lower panel).
It is well known that nutrient limitation, such as iron, induces the stringent response in bacteria ( 61 ), which is primarily characterized by a down-regulation of r-proteins ( 19 ). In E. coli , DksA disrupts the interaction of RN A pol ymerase (RNAP) with DNA by directly binding to RNAP during the stationary growth phase, decreasing the transcription of r-proteins; thus a strain lacking DksA constitu-ti v ely e xpresses r -proteins and r -RNAs throughout different growth phases ( 18 , 19 , 62 ). Ther efor e, observed downr egulation of r-protein genes (induction of stringent response) in the WT cells under copper stress could be due to dysregulation of iron and zinc homeostasis, whereas it remained constituti v ely high in cells lacking DksA.
In fact, genes responsible for biosynthesis, uptake and export of siderophores for iron acquisition such as acinetobactin ( bauA-F ), baumannoferrin ( bfnA-L ) and fimbactin ( fbsA-Q ) gene clusters ( 63 ) were upregulated up to 180-fold under copper stress in both the dksA and WT strains (Figure 4 C). We also noted the upregulation of genes involved in zinc uptake and metabolism such as rpmE2, zigA and  Table  S2). These genes encode the alternati v e ribosome subunit of the 50S protein L31, a zinc metallochaperone and zinc uptake protein respecti v ely and are known to play crucial roles in cellular physiology during zinc limitation (64)(65)(66)(67). In contrast, the expression of both siderophore metabolism clusters and zinc metabolism genes were not affected under zinc stress in WT (Figure 4 C). We therefore hypothe-sized that copper str ess incr eases synthesis of siderophores and zinc metabolism proteins that may be r equir ed to compensate for the metal starvation of the iron-sulfur (Fe-S) and zinc-dependent proteins. Consistent with this hypothesis, when we supplemented the growth medium with subinhibitory le v els of ZnSO 4 and / or FeCl 3 in the presence of copper stress, we observed that fitness of WT A. baumannii improved significantly compared to the copper stress alone (Figure 4 E). These results suggest that copper stress induces both iron and zinc limitation, consistent with previously reported interdependencies of copper and zinc homeostasis in this organism ( 68 ).
The stringent response in bacteria is also highly correlated with the cellular concentration of initiating nucleotide triphosphates, ATP and GTP (69)(70)(71). Most microorganisms use the Fe-S-dependent branched electron transport chain composed of NADH-quinone oxidoreductases and quinol oxidases to efficiently couple electron exchange for ATP production by the F 1 F 0 ATPase during aerobic respiration ( 72 , 73 ). Since aerobic respiration contributes to more than 70% the total ATP production during bacterial growth ( 74 ), we compared respiration activities in A. baumannii under copper and zinc stresses as a proxy for ATP production. Both WT and dksA exhibited similar levels of respiratory activities with or without zinc stress (Figure 4 F, right  panel). In contrast, copper str ess r esulted in a drastic reduction in respiration for WT cells (Figure 4 F, left panel). A reduction of respiratory activity was also noted in the dksA strain under copper stress, but the effect was not as se v ere as in WT cells. This finding was also consistent with the observed significant reduction in expression of nuoA-N , c y dAB and atpA-I genes, which encode enzymes r equir ed for NADH:quinone oxidoreductase electron exchange, and cytochrome d ubiquinol oxidase and ATP synthesis respecti v ely, in WT under copper stress but not in the dksA or in treated cells (Figure 4 D). Collecti v ely, these data suggest that copper stress inhibits respiration in A. baumannii and DksA plays a role in exacerbating this effect under copper stress.
Recently, it has been proposed that oxidation of cysteine residues of the zinc finger is required for the allosteric activation of DksA-dependent stringent response to protect bacteria from H 2 O 2 ( 75 , 76 ). Copper stress is known to generate hydroxyl radicals ( • OH) through the Fenton-like reaction ( 77 ). We ther efor e hypothesized tha t the activa tion of DksA-dependent stringent response under copper stress was due to a redox-switch of DksA by the oxidation of cysteine residue. Under zinc stress, DksA remains as a zincbound reduced form and therefore fails to activate stress response.
To test our hypothesis, we employed nanoscale differential scanning fluorimetry (nanoDSF) analysis on the heterolo gousl y expressed and purified full-length A. baumannii AB5075 UW DksA protein (in presence of the reducing agent, TCEP) in vitro . The fluorometric technique monitors changes in intrinsic tryptophan (and tyrosine to some extent) fluorescence as a result of folding or unfolding of the protein as a function of the temperature ( 78 ). The A. baumannii DksA protein has only one tryptophan residue (W74), predicted to be in one of the ␣-helices forming the coiled-coil region (AlphaFold structure prediction, data not sho wn). As sho wn in Figure 4 G, purified DksA displayed a red-shifted emission with a melting temperature (T M ) of 57 • C, with a mild destabilization observed following the addition of the strong oxidizing agent H 2 O 2. In the presence of excess zinc, the assay showed a blue-shifted emission with significantly lower T M of 46 • C, indicating possible allosteric changes in the chemical environment around the lone tryptophan (and potentially tyrosines) leading to the destabi-lization of DksA. The addition of the strong oxidizing agent H 2 O 2 to the DksA-Zn sample did not lead to a major thermal stability shift as compared to zinc alone, supporting our hypothesis that excess zinc potentially locks and inhibits the allosteric activation of DksA.

DksA-dependent regulation affects metabolic pathways
Recently, it has been shown that activation of the stringent response determines survival success of bacteria under stress by modulating metabolic pathways ( 79 ). To test whether DksA-dependent stringent response is important for coordinating cellular metabolism in A. baumannii , we interrogated expression patterns of key metabolic pathways in dksA mutants. In A. baumannii metabolic pathways such as aromatic compound degradation are known to be essential for successful virulence ( 10 ). Most bacteria metabolize aromatic compounds such as catechol and protocatechuate through phenylacetate and ␤-ketoadipate pathways ( 80 , 81 ). In the phenylaceta te pa thway, aroma tic compounds are broken down into succinyl-CoA, whereas the ␤-ketoadipate pa thway genera tes succin yl-CoA and acetyl-CoA bef ore entering the tricarboxylic acid (TCA)-glyoxylate cycle ( Figure  5 A). Bacterial growth on aromatic compounds , acetate , or fatty acids also r equir e the activation of the glyoxylate shunt in the TCA and gluconeogenesis pathways ( 82 ).
In our dksA transcriptomics, the two most differentially expressed pathways were phenylacetate and catechol pathways, encoded by genes in paa ( paaNABCDEFGHK ) and pca ( pcaIJFBDKCHG ) operons respecti v ely (Figure 5 B, C), but showed condition-specific induction. The expression of genes in the paa operon decreased (between 12-and 330fold) in the dksA cells with and without copper stress (Figure 5 B). In contrast, when WT cells were treated with copper, expression of these genes increased (5-to 14-fold; Figure 5 B). Whilst there was no change in expression pca genes with or without copper stress in the dksA mutant, copper str ess incr eased expr ession of the pca operon in WT strain, mirroring effects of the paa operon (incr eased r elati v e to untreated cells 28-to 180-fold; Figure 5 C). Interestingly, under zinc stress the expression of paa genes were not affected in the dksA m utant w hereas it decreased by 3-to 9fold in the WT. In the dksA cells, two important genes responsible for the glyoxylate shunt, aceA encoding isocitrate lyase and glcB encoding malate synthase were also downregulated (by 18-and 5-fold respecti v ely; Figure 5 A, Supplementary Table S2). Consistent with the copper stress impacting iron homeostasis, genes encoding Fe-S-dependent proteins, such as fumC in the T CA cycle, wer e upr egulated under copper str ess. These r esults indicate that a functional version of DksA is needed to activate not only specific metabolic pathways (phenylacetate and catechol) but also central metabolic pathways (TCA cycle) during metal stress in A. baumannii .
The mechanism(s) by which DksA induces paa and pca operons during copper stress remains unclear. Howe v er, it has been proposed that GacS / GacA two-component system operates as a switch between primary and gluconeogenic metabolites in number of bacteria ( 83 ). Additionally, carboxylic acids such as acetate and propionate have been shown to be an environmental cue for the GacS / A system ( 84 , 85 ). We found that expression of gacA was decreased in the dksA mutant (at an average of 3.6-fold) with or without copper str ess. The expr ession of gacA was not affected under both copper and zinc stresses in the WT strain (Supplementary Table S2). Thus, these data suggest that DksA controls major metabolic pathways under metal str esses by r egulating the GacS / A two component signaling system.
To test whether DksA is functionally important for ca tabolism of substra tes associa ted with the TCA cycle and its glyoxylate shunt, we performed Biolog phenotypic arrays tha t calcula te bacterial respira tion ra tes on 192 carbon sources over time (Biolog MicroArrays PM1 and PM2), as described previously ( 40 , 86 ). In line with expression data, the dksA mutant showed significant growth defects in sub-strates (acetic and ketoglutaric acid), and aromatic carbon sour ces r equiring the paa and pca operons such as phenylalanine and 4-hydroxy benzoic acid (Figure 5 E). Taken together, the expression data of all aromatic compoundassociated genes and the phenotypic growth assays suggest that DksA acts as a transcriptional switch for regulating secondary gluconeogenic pathways under stringent conditions ( Figure 5 D).

Confirmation of the essential role that DksA plays in general str ess r esponse
To understand the role of DksA in stress protection in A. baumannii , we investigated the expression pattern of stress responsi v e genes in the dksA mutant of A. baumannii . In E. coli, functional rpoS is r equir ed for expression of oxidati v e stress genes such as superoxide dismutase ( sodC ) and catalases ( katE and katG ) as well as biosynthesis of stress protectants such as trehalose ( otsA and otsB ) ( 87 ). As shown in Figure 6 A, expression of these RpoS-dependent genes ( katE , sodC , otsA and otsB ) was decreased (up to 36-fold). Expression of RpoS-independent stress genes such as btuE, katG and uspA was also decreased (2-to 4-fold; Figure 6 A). The observed regulation of RpoSdependent stress genes by DksA suggests that there may be downstream overlaps in the regulation of stress-related genes orthologs between RpoS in E. coli and DksA in A. baumannii , which lacks RpoS. Howe v er, gi v en that RpoS and DksA both regulate hundreds of genes in the E. coli genome ( 88 , 89 ), we cannot determine the true extent of the direct functional overlap between RpoS in E. coli and DksA in A. baumannii .
To test whether down regulation of stress genes has impacted bacterial ability to cope with external stresses, we investigated the phenotypic effects of disruption of dksA in A. baumannii under oxidati v e and desiccation stresses. Ability to cope with these stresses are paramount for both survival on dry nosocomial environments and virulence ( 6 ). Consistent with the decreased expression of oxidati v e stress genes, the strain lacking a functional dksA was unable to grow in the presence 0.5 mM of H 2 O 2 , a well-known oxidizing agent, whereas the WT cells exhibited relati v ely uninhibited growth at the same concentration of H 2 O 2 (Figure 6 B). Similarly, we found that the viability of the dksA mutant was markedly reduced within 7 days of desiccation on dry surface at room temperature and this trend continued for next 51 days (Figure 6 C). After 58 days of incubation, only 0.03% of the original population of dksA mutants survi v ed. The rate of dying was much slower for WT, with a significant proportion of the original population (up to 16.7%) was still viable up to 58 days (Figure 6 C), suggesting that DksA is essential for survi val under oxidati v e and desiccation stresses.
In addition to physiological adaptation, genetic adaptation also plays an important role for both the short-term survival and long-term evolution of pathogens. RpoS is known to play a crucial role in mutagenesis in E. coli ( 90 , 91 ). To test whether DksA is also involved in mutagenesis in A. baumannii , we compared mutation rates in the WT A. baumannii strain AB5075 UW and its dksA mutant by measuring the frequency of rifampicin resistance mutation acquisition, which is usually conferred by base pair substitution mutations in rpoB , a RNA polymerase subunit B (RNAP B)-encoding gene ( 92 ). We found that the dksA mutant rpoB muta tion ra te was almost 10-fold lower compared to the WT, with rates of 0.04 × 10 −8 (95% CI, 0.02-0.06 × 10 −8 ) and 0.39 × 10 −8 (95% CI, 0.29-0.5 × 10 −8 ) mutations per generation, respecti v ely, as shown in Figure 6 D. These results suggest that DksA is not only essential for A. baumannii survival under stress conditions but also plays an important role in cell mutagenesis.

DksA is highly conserved and widely distributed across gammaproteobacteria
To better understand whether the unique role of DksA in str ess r esponse is limited to A. baumannii , we analyzed the distribution of DksA in 1686 r epr esentati v e bacterial species from 88 different families across Gammaproteobacteria (Supplementary Figure S4A, B) as well as the archetypal str ess-r esponse protein RpoS. Due to a high sequence similarity of DksA with TraR (Supplementary Figure S4C) and RpoS with other sigma factors, such as RpoD ( 17 , 93 ), sequence similarity alone was not enough to distinguish DksA from TraR and RpoS from other sigma factors. Since DksA is relati v e larger than TraR and RpoS is well known for its size of 38 kDa, we also examined the length of protein sequences in addition to the sequence similarity to accurately distinguish these proteins from other similar ones using the GTDB database (see the methods section and Supplementary Figure S4C for detail). As shown in Figure 7 A, DksA showed extremely high conservation and could be detected in 85 of 88 (96.6%) Gammaproteobacteria families.
Howe v er, we une xpectedly observ ed that 34% of families did not harbor RpoS at all (starred in Figure 7 A). Overall, 54.9% of 1686 r epr esentati v e bacterial species r epr esentati v es analyzed did not harbor RpoS, although almost all Enterobacteriaceae have RpoS. These results suggest that DksA is more widely distributed amongst Gammaproteobacteria than RpoS, although its overr epr esentation in Enterobacteriaceae means it is commonly studied in relation to stress response.
To test whether DksA has a direct role in providing antibiotic stress protection in A. baumannii , we investigated the effect of dksA disruption on antibiotic stress in A. baumannii and E. coli by estimating minimum inhibitory concentration (MIC) for 10 antibiotics across different classes (excluding the mutant selection marker for each strain; Figure 7 B). We found that despite having very different antibiotic resistance profiles, both A. baumannii strains exhibited increased sensitivity to the majority of antibiotics upon disruption of dksA and the trend w as alw ays reduced MIC. In the dksA mutants, 5 out of 10 (50%) antibiotics had a decreased MIC (2 to 16-fold) for AB5075 UW and 6 out of 10 (60%) for ATCC 17978 (Figure 7 B, Supplementary Table S3). The dksA mutant strain of E. coli also showed a decreased MIC for six antibiotics (60%), but unlike A. baumannii strains, the MIC for two antibiotics (amikacin and rifampicin) was surprisingly increased by 2-to 4-fold (Figure 7 B, Supplementary Table S3), suggesting DksA has a unique role in controlling antibiotic resistance in A. baumannii .
The mechanisms by which DksA protects A. baumannii from antibiotic stress could be related to specific efflux pump activation and / or defense against endogenously antioxidants generated antibiotics such as gentamicin ( 94 ). Expression of two known resistance-related genes, ACX60 00045 and emrB encoding AdeT and EmrB efflux pumps, respecti v el y, was significantl y decreased (up to 5fold) in the dksA mutant compared to WT (Supplementary Figure S5A). AdeT is known to be involved in aminoglycoside efflux ( 95 ), and phenotypically dksA mutants were more sensiti v e to amikacin, gentamicin, and kanamycin. We hypothesized that removing DksA would affect the accumulation of antibiotics in mutant cells and investigated this using intracellular accumulation of gentamicin in the A. baumannii dksA mutant using a gentamicin uptake assa y. We f ound that the dksA mutant cells accumulated a significantly higher le v el of gentamicin compared to WT cells (Supplementary Figure S5B). We did not observe a significant increase or decrease in expression of the other major efflux pumps that A. baumannii harbors, namely adeIJK and adeABC (Supplementary Figure S5A ) . Taken together this suggests that DksA provides protection from stresses not only b y activ ating stringent response but also by activa ting specific ef flux pumps and reducing accumulation of toxic compounds inside the cells. These findings contrast to a previous stud y tha t observed significant increases in expression of efflux genes adeB, adeIJ, abeM and tetA for the dksA mutant of A. baumannii ( 96 ), which would indica te tha t the presence of DksA is detrimental for antibiotic resistance gene activation and conflicts with their gi v en phenotypic resistance results. This discrepancy could be due to differences in the test conditions, for example, use of overnight culture for RNA extraction in transcription assays.
To further examine DksA's role in stress response, we compared the ability to tolerate oxidati v e stress between dksA mutants of two A. baumannii strains (AB5075 UW and ATCC 17978) and E. coli K-12 strain BW25113 by analyzing the growth phenotypes in the presence of exogenous H 2 O 2 . The dksA mutants of both A. baumannii strains displayed a significant growth defect in the presence of as little as 0.6 mM H 2 O 2 , whereas their respecti v e WT cells grew in up to 5 mM H 2 O 2 (Figure 7 C). Howe v er, in E. coli , dksA disruption had no impact on H 2 O 2 survival compared to WT (Figure 7 C). This data suggests that DksA is r equir ed to protect from a high le v el of oxidati v e stress is in A. baumannii, but is not essential for H 2 O 2 stress protection in E. coli .
It is worthwhile noting that A. baumannii has a larger DksA (178 amino acids) compared to E. coli (151 amino acids) (Supplementary Figure S4C), and we suspected that the larger protein may function differently to the smaller version. To test if this difference in size contributed in observed phenotypic differences between A. baumannii and E. coli , we cloned different DksA versions into pVRL2 ( 44 ) and transformed into the A. baumannii AB5075 UW dksA mutant. These versions were the full length dksA from A. baumannii (dksAFL-Ab), the truncated (dksATr-Ab) minimized version from A. baumannii (consensus of conserved amino acids) and the full length dksA from E. coli K-12 (dksAFL-Ec). We found that the dksA mutant complemented with dksAFL-Ab had the greatest le v el of phenotypic rescue in all three test conditions (i.e. oxidati v e, zinc and rifampicin str esses), wher eas dksAFl-Ec showed the incomplete complementation (Figure 7 D). Together these results suggest that the full DksA protein is most effecti v e in stress protection in A. baumannii and that the shorter E. coli version of DksA cannot r estor e function as effectively in A. baumannii .

CONCLUSIONS
Our systematic genomics-based approach has uncovered DksA acts as a master regulator of stress response and virulence in A. baumannii and here we present the intricate details of how DksA controls stress tolerance. Our genotypic and phenotypic results allowed us to outline the pleotropic activity of DksA, acting on several core biological processes to protect the bacterial cell from stressors. The overall strategy that A. baumannii employs to use DksA to overcome numerous activities that are RpoS-controlled in other wellstudied pathogens, like those in the Enterobacteriaceae family, can be rationalized in terms of its adapti v e advantages. While RpoS positi v ely r egulates many genes r equir ed for stress protection, it also ad versel y ef fects the utiliza tion of secondary carbon sources such as acetate and succinate as primary energy sources ( 97 ). In A. baumannii , DksA appears to play broader roles beyond regulating the direct stress genes, but also positi v ely regulating secondary carbon metabolism and energy r esour ces without ex erting notable trade-of fs associa ted with RpoS.  Table S5 for details about the length ( dksA FL-Ab) and truncated ( dksA Tr-Ab) of A. baumannii AB5075 UW and full the length dksA ( dksA Fl-Ec) from E. coli K-12.
Our work provides insight into the disparate roles of DksA under seemingly similar stresses (such as copper and zinc), indica ting tha t DksA acts as a sophisticated and sensiti v e molecular switch for stress response. Our study reiterates the importance of assessing gene function in less wellstudied bacterial species and warns against blindly transferring function between species based solely on sequence homology, as DksA seems to play a varied role in E. coli and A. baumannii str ess r esponse. We highlighted that DksA is highly conserved in Gammaproteobacteria and is almost always pr esent ( > 95%), compar ed to the ar chetypical str ess r esponse r egulator RpoS, which is pr esent far less commonly ( < 45%). In this study, we present an initial characterization of a conserved, DksA-mediated general stress response that provides a blueprint of a stress adaptation strategy in A. baumannii , which may be also appliable to other bacterial species lacking RpoS, including the key pathogens Neisseria gonorrhoeae , Camp ylobacter jejuni and Bordetella pertussis . Our work raises the question whether other fundamental str ess r esponse systems across di v erse bacterial classes are yet to be characterized.

DA T A A V AILABILITY
The raw sequencing data was deposited under GEO accession number GSE169081. TraDIS sequence reads were deposited in the European Nucleotide Archi v e under accession number ERP118051.