Intracellular Acinetobacter baumannii acts as a transient reservoir in lung infection via a “persist and resist” strategy

Intracellular A. baumannii Can Seed De Novo Infection in Mouse Lungs.

A. baumannii was long considered an extracellular pathogen; however, recent studies have demonstrated that it can replicate within ACVs in mouse and human macrophages in vitro and has also been detected within polymorphonuclear neutrophils (PMNs) and AMs, the predominant cell types in BALF in vivo (6, 8, 20). AMs play a key role in antibacterial defense, eliminating bacteria through phagocytosis and coordinating the subsequent immune response (6, 21). Given the recent discovery of an intracellular A. baumannii population in the lung during murine infection, the infectivity and contribution of phagocytosed bacteria to the infection outcome remain unexplored.

We hypothesized that intracellular A. baumannii could serve as a transient reservoir during lung infection. To test this hypothesis, we performed BALF transfer experiments from infected to uninfected mice and assessed de novo lung infection in the latter, enabling us to assess the fate of bacteria following internalization (Fig. 1A). We intranasally infected TLR4-deficient C3H/HeJ mice (22) with ~10⁹ colony-forming units (CFU) of either the modern clinical isolate A. baumannii strain Ab398 or Ab19606, a lab-adapted strain. We used C3H/HeJ mice, which are suitable for studying A. baumannii infection with a low inoculum, as demonstrated in a recent chronic lung infection model, and to ensure compatibility with mice used in the subsequent BALF transfer step (11). At 24 hpi, red blood cell–depleted BALFs were pooled and treated with colistin to eliminate extracellular bacteria. The predominant cell types present in BALF at this time point are macrophages and neutrophils (6). In colistin-treated BALF, these immune cells harbored an intracellular population of A. baumannii (Fig. 1B). Complete elimination of extracellular bacteria was confirmed by BALF antibiotic protection assays (SI Appendix, Fig. S1A). Bacterial loads in the lungs, kidneys, and spleen were equivalent between the two strains (SI Appendix, Fig. S1B).

Colistin-treated BALFs containing approximately 10⁴ to 10⁵ intracellular bacteria were intranasally transferred to a second group of cyclophosphamide-treated C3H/HeJ mice. Although C57BL/6 mice intranasally infected with 10⁸ CFU of Ab19606 had a lower BALF intracellular bacterial burden 24 hpi than those infected with Ab398 (8), in our experimental settings, the use of C3H/HeJ mice and an intranasal inoculum of 10⁹ CFU led to similar amounts of intracellular bacteria in the BALF transferred for both strains. Cyclophosphamide is a potent immunosuppressive agent that induces severe neutropenia, mimicking the condition of highly immunocompromised patients affected by A. baumannii infection. It is widely accepted in the field for use in animal models (23). Neutropenia also slows the kinetics of bacterial clearance in the lungs. At 24 hpi, bacterial burdens in lungs, spleens, and kidneys of the recipient mice were assessed by CFU enumeration. Mice infected with BALF carrying intracellular Ab398 resulted in a significant increase in bacterial burden in the lung 24 hpi (~10⁵ to 10⁶ CFU), indicating that not only are intracellular Ab398 viable, but are also capable of initiating de novo lung infection in naive mice (Fig. 1C). Consistent with the low-dose intranasal inoculation in C3H/HeJ mice, which leads to localized lung infection rather than systemic spread, CFU counts in the spleen and kidneys were undetectable (11). Mice that received BALF from Ab19606-infected mice exhibited a significantly lower bacterial burden in their lungs at 24 hpi (~10²). This difference in infection outcome could be attributed to impaired egress from BALF cells. However, we ruled out this hypothesis, as both Ab19606 and Ab398 strains were able to exit cells present in the BALF at comparable levels, with Ab19606 even showing slightly higher egress between 1 and 2 h post–BALF collection (SI Appendix, Fig. S1C). Together, our result demonstrates that intracellular bacteria, at least for some virulent strains such as Ab398, can act as a potentially infectious transient reservoir. Upon egress from the host cells, this intracellular bacterial population can seed de novo lung infections. Within a single infection, infected neutrophils and macrophages may serve as a transient reservoir for A. baumannii during infection, underscoring the importance of studying its intracellular lifestyle.

Comparative Dual-Proteomic Analysis.

A. baumannii lacks the canonical virulence factors (T3SS and T4SS) employed by most intracellular Gram-negative bacteria to survive and replicate within macrophages. To identify factors that promote bacterial survival and replication within cells and contribute to interfere with phagosome maturation, we used a comparative dual-proteomic approach (Fig. 2A). In this study, we infected mouse J774.A1 macrophages with two replicative A. baumannii clinical isolates, Ab398 and Ab803, at a multiplicity of infection (MOI) of 10 (8). These strains belong to different multilocus sequence typing (MLST) categories and possess different capsule types and lipooligosaccharide (LOS) outer cores (SI Appendix, Table S1). Noninfected macrophages and bacteria grown in cell culture media under static conditions served as controls. At 1 hpi, cells were washed and colistin was added to the media for J774.A1-containing samples (infected and uninfected), to kill any extracellular bacteria. At 4 hpi, cells were washed and lysed, and protein content was precipitated. Subsequently, we performed proteomic analyses using a data-independent acquisition (DIA) mass spectrometry approach. Overall, we detected 6,271 and 6,354 macrophage proteins in samples infected by Ab398 and Ab803, respectively (Fig. 2B). Additionally, we detected 663 and 513 bacterial proteins in samples infected by Ab398 and Ab803, respectively, 452 of which were conserved and detected between the two strains.

Proteome Profiling of A. baumannii–Infected J774.A1 Macrophages.

We plotted the Log₂[fold change (FC)) values of host proteomic data following infection by Ab398 against those following infection by Ab803 (Fig. 3A). The linear distribution of the data along the line of identity indicates a highly similar macrophage response to both A. baumannii strains. A Venn diagram shown in SI Appendix, Fig. S2 A–C compares data obtained upon infection by the two strains. To explore host proteome remodeling upon intracellular infection, we conducted gene set enrichment analysis (GSEA) and calculated normalized enrichment scores (NES) for hallmark mouse gene sets (24). We identified 11 to 12 enriched gene sets [false discovery rate (FDR) < 0.05], respectively, with the most significant, considering both enrichment score and gene counts being TNFα signaling via NF-κB, Inflammatory response, and IFNγ response (Fig. 3B and SI Appendix, Fig. S3). These three pathways share overlapping gene/protein content. Label-free quantitation (LFQ) values for enriched proteins of which levels were significantly altered in at least one strains are presented in Fig. 3 C–E, along with the Log₂(FC) values for Ab398 and Ab803 infections. Macrophage initially detect LOS, the major outer membrane component in A. baumannii, via binding to TLR4, triggering NLRP3 inflammasome activation and IL-1β production (25). Although NLRP3 levels remained unchanged in our proteomic data, IL-1β levels significantly increased upon infection (Ab398: Log₂(FC) = 6.5, q-val = 0; Ab803: Log₂(FC) = 4.2, q-val = 4.7 × 10⁻⁴). Activated macrophages produce numerous proinflammatory cytokine and chemokines including IL-1β, TNFα, Cxcl2, Ccl4, Ccl2, and Cxcl10, all of which were markedly upregulated in infected macrophages in our data [Fig. 3 C–E and Dataset S1, (6)]. Additionally, A. baumannii–infected macrophages showed increased expression of nitric oxide synthase (NOS2, Ab398: Log₂(FC) = 5.2, q-val = 2.4 × 10⁻⁴; Ab803: Log₂(FC) = 3.4, q-val = 1.3 × 10⁻³) due to LOS and TNFα signaling (26, 27). Similarly, levels of NOD2, an intracellular pathogen recognition receptor sensing A. baumannii peptidoglycan, were elevated in Ab398 [Ab398: Log₂(FC) = 1.5, q-val = 4.2 × 10⁻²; (28)]. Our proteomic findings suggest macrophages undergo proinflammatory M1 polarization upon A. baumannii infection, however, additional experiments are required to confirm this polarization (29). These results are consistent with the findings of Kho et al., in THP-1 human macrophages infected by A. baumannii with or without polymyxin B, suggesting that the proinflammatory response observed in our experiments is triggered during initial contact between A. baumannii and the macrophages, rather than during its intracellular replication (30).

Notably, we observed that the itaconate-synthesizing enzyme Acod1 (also known as IRG1) was significantly induced upon A. baumannii infection in J774.A1 macrophages (Ab398: Log₂(FC) = 3.0, q-val = 3.7 × 10⁻⁴; Ab803: Log₂(FC) = 2.9, q-val = 4.8 × 10⁻⁴). Itaconate inhibits bacteria species such as Mycobacterium tuberculosis and Salmonella, while others like Pseudomonas aeruginosa and Staphylococcus aureus can exploit it as a source of nutrients (31). To our knowledge, Acod1 induction has not been previously reported for infection by A. baumannii. We confirmed this finding by immunoblot (SI Appendix, Fig. S4A). In our analysis, we included the nonreplicative strain Ab19606, which also induced Acod1 (IRG1), suggesting that induction of Acod1 is not a response specific to intracellularly replicating A. baumannii, consistently with its induction via the TLR4/NF-кB signaling pathway (32). To test whether intracellular Ab19606 could be eliminated due to a higher sensitivity to itaconate, we added itaconic acid (IA) to the culture media. Although neither Ab19606, Ab398, nor Ab803 strains were able to use IA as a sole carbon source (SI Appendix, Fig. S5A), none of the strains’ growth was affected by the addition of IA to the media (SI Appendix, Fig. S5 B–D). Similarly to Acod1, the induction of two additional A. baumannii–induced host proteins identified in our proteomics data was confirmed; IL-1β and Serpine1/PAI1. Both proteins were induced in a replication-independent manner, given that the nonreplicative strain Ab19606 also led to their induction (SI Appendix, Fig. S4 B and C).

Taken together, intracellular A. baumannii infection triggers a macrophage inflammatory response, characterized by the production of proinflammatory cytokines and chemokines, and the induction of Acod1, probably due to initial contact with A. baumannii rather than its intracellular replication itself.

Proteomic Profiling of Intracellular A. baumannii.

Two intracellular A. baumannii strains, Ab398 and Ab803, were used in our dual comparative proteomic analysis of infected J774.A1 macrophages. Both strains survive and replicate within ACVs in J774.A1 macrophages (8). Proteomic analysis was conducted at 4 hpi as the peak of replication typically occurs between 4 and 8 hpi in J774.A1 cells. Bacterial cultures grown under static condition for 4 h in DMEM + 10% heat-inactivated fetal bovine serum (hi-FBS)—the same media used for infection—served as baseline. Input samples grown in LB were also analyzed but excluded from the final analysis, as the bacterial growth state following a 16 h preculture differs significantly from the tested conditions. To allow comparison between sample groups, A. baumannii proteins observed during infection were normalized to the median of the LB groups as shown in SI Appendix, Fig. S6. A. baumannii exhibits extensive genetic diversity, with a large pan-genome exceeding 20,000 genes. As a result, not all genes are conserved between the two strains. Fig. 4A shows the Log₂(FC) of all 452 conserved proteins detected in both strains. Intracellular infection of J774.A1 macrophages induced highly similar changes in A. baumannii protein levels across these conserved proteins, as most data aligned closely with the line of identity. Among these 452 bacterial proteins, 73 exhibited significant alterations in abundance during intracellular infection compared to their respective control (static growth in culture media, SI Appendix, Fig. S2 D–F).

Significantly altered proteins belonged to diverse pathways, as demonstrated by the analysis of clusters of orthologous groups (COG) categories (Fig. 4B). In both strains, a higher proportion of proteins involved in “lipid transport and metabolism” and “energy production and conversion” increased during macrophage infection. Conversely, proteins that decreased during infection were primarily associated with “translation, ribosomal structure and biogenesis,” “nucleotide transport and metabolism,” or even “cell cycle control, cell division, chromosome partitioning.” Fig. 4C shows the most significantly altered proteins on both strains.

Analysis of Intramacrophage A. baumannii Proteome.

The proteomic results for the bacterial proteins are presented in Dataset S2. The abundance of several proteins involved in defense against host nutritional immunity was increased in intracellular bacteria. Host-imposed nutritional immunity limits bacterial access to trace metals, such as zinc (Zn²⁺), magnesium (Mg²⁺), and iron (Fe²⁺), preventing their proliferation during infection (33). Bacteria counteract by inducing the corresponding transporters. Our proteomic data showed increased levels of both ZnuD and MgtA following macrophage infection (Table 1). In A. baumannii, Zn²⁺ is imported by the Zn²⁺ uptake (ZnuABCD) system, which contributes to bacterial survival in in vivo infections (33–35). Similarly, MgtA type Mg²⁺ transporters, extensively studied in Salmonella, play a crucial role in virulence in mice and replication in macrophages (36).

Proteins encoded by the phenylacetic acid (PAA) catabolism operon were also increased upon intracellular infection, including PaaZ, PaaG, and PaaJ (Table 1). In A. baumannii, this pathway is involved in stress response and virulence. PAA is an intermediate in the catabolism of phenylalanine. In A. baumannii, the virulence of a mutant unable to catabolize PAA was significantly impaired in both murine and zebrafish models of infection (37, 38). Our analysis also revealed increased levels of exopolyphosphatase (Ppx), responsible for the catabolism of polyphosphate (polyP), while levels of Ppk1, the polyphosphate kinase synthetizing polyP, were decreased upon infection. PolyPs are linear chains of phosphate residues linked by high-energy bonds, which can be stored in polyP granules and play an important role in the virulence and stress response in many bacterial species, including P. aeruginosa and A. baumannii (39, 40). Further highlighting the requirements for phosphate in A. baumannii intracellular survival/replication, protein levels of PstS, a phosphate transporter, were also increased in both strains (41). Other stress responses were activated such as response to the oxidative stress through increase of the superoxide dismutase (SodB) and the catalase (KatG) in the Ab398 strain.

In addition, protein levels of LpsB, a glycosyl transferase required for the addition of outer core to the nascent LOS, were increased (42). This is consistent with previously reported reduced virulence of lpsB mutants in serum and in vivo rat and mouse models (43, 44). Also involved in bacterial cell surface modification during the intracellular lifestyle, our proteomic analysis showed increased levels of bacterial lipocalin Blc in both strains, an inner membrane protein expressed in response to starvation or high osmolarity, allowing maintenance of membrane and envelope integrity (45).

Multiple proteins involved in metabolism and energy source utilization were also increased upon macrophage infection, including FahA and Ald1. FahA is a fumarylacetoacetase (also known as HmgC), belonging to the homogentisate (Hmg) degradation pathway (46). Together with the PAA catabolism pathway, the Hmg pathway participates in the degradation of the amino acids phenylalanine and tyrosine. Ald1 is annotated as a long-chain aldehyde dehydrogenase, which could play a role in the catabolism of n-alkanes or ethanolamine utilization (47, 48). In line with these findings, in both Listeria monocytogenes and Salmonella enterica serovar Typhimurium, intracellular replication of mutants in the ethanolamine utilization (eut) pathway was impaired (49, 50). A. baumannii also appears to rely at least partly on intracellular ethanolamine as a source of carbon and nitrogen. Other proteins induced were CatA, a protein involved in aromatic compound metabolism (51), the NADH-quinone oxidoreductase subunit E (NuoE), and the hydroxyisobutyrate dehydrogenase MmsB in Ab398.

Overall, the analysis of the proteins with known function increased in intracellular bacteria was mainly involved in four main functional categories: counteracting nutritional immunity, stress responses, surface modification, and energy source utilization (Table 1). Our analysis additionally identified many proteins of unknown function that are up- or down-regulated upon macrophage infection (Dataset S2). The role of these proteins will be the scope of future work.

A. baumannii Survival and Replication in Macrophages Is Multifactorial.

We engineered Ab398 deletion mutants in genes corresponding to the most altered proteins in each of the identified functional categories and their complemented strains. No significant growth defects were observed when these mutants were grown in DMEM + hi-FBS (SI Appendix, Fig. S7A). We subsequently assessed the impact of these mutations on replication within J774.A1 macrophages over 24 h, respective to that of the parental Ab398 strain (Fig. 5 A–H; all mutants were tested simultaneously, but datasets are graphed separately to allow better visualization). The complete summary of the data is shown in SI Appendix, Fig. S8. To complement the analysis, we conducted confocal microscopy analysis of J774.A1 infected with each mutant at 24 hpi (Fig. 5I). All the mutants displayed impaired replication ability in macrophages to different extents. Both mutant strains involved in nutritional immunity (∆mgtA and ∆znuD) displayed moderate but reproducible replication defects at all times, with the ΔmgtA showing a more dramatic, statistically significant effect at 12 hpi. This aligns with findings in Salmonella, in which mgtA and mgtCB are required at later time during macrophage infection (36). ∆paaG and ∆ppx mutants showed markedly reduced replication while ∆fahA and ∆ald1 mutants displayed little to no replication in J774.A1 macrophages. Interestingly, an lpsB-deficient mutant, in which we could confirm the absence of LOS outer core (SI Appendix, Fig. S9A), was unable to survive in macrophages and was readily eliminated over time. We showed that this defect was not due to increased sensitivity to acidic pH as the ∆lpsB mutants grow similarly to the wildtype strain in LB buffered at pH5 (SI Appendix, Fig. S7B). In line with data obtained by antibiotic protection assays, confocal microscopy showed a significant reduction in intramacrophage bacterial burdens (Fig. 5I, green signal), with the most affected mutants being ∆lpsB and ∆ald1. Interestingly, overexpression of lpsB both in the ∆lpsB and in the parental strain led to decreased susceptibility to vancomycin (SI Appendix, Fig. S9 B and C). Vancomycin is a large compound targeting the peptidoglycan that is typically used to treat Gram-positive bacterial infections. However, previous work showed that if the Gram-negative outer membrane barrier is compromised, vancomycin crosses the outer membrane and kills the bacteria (52). Increased tolerance to vancomycin upon overexpression of lpsB therefore suggests a role of lpsB in bacterial surface modification.

Intracellular replication defects of most mutants were complemented by addition of the deleted genes (SI Appendix, Fig. S10), except for ∆mgtA and ∆ald1 mutants, in which in trans addition of the deleted gene lead to lower intracellular fitness for ΔmgtA::mgtA and no effect for ∆ald1::ald1.

Collectively, these results reveal that A. baumannii employs a multifactorial strategy to survive and replicate within macrophages. This strategy includes counteracting host-imposed nutritional immunity and activating mechanisms of stress response (including polyP and PAA catabolism) and adapting to its intracellular niche through the utilization of ethanolamine and increased catabolism of tyrosine and phenylalanine. Furthermore, our experiments show that an intact LOS is required for intracellular survival within macrophages.

Bacterial Strains and Culture Conditions.

Plasmids and strains used in this work are listed in SI Appendix, Table S3. All A. baumannii overnight cultures were inoculated with 5 to 10 colonies to average potential phase variation effects, although low to no phase variation could be observed for the strains used here (65). Bacterial cultures were grown at 37 °C in Lennox broth/agar (LB) under agitation. Cultures and plates were supplemented with 10 μg/mL chloramphenicol, 50 μg/mL apramycin, or 10% sucrose when appropriate.

Cell Culture Conditions.

The J774A.1 mouse macrophage cell line (ATCC TIB-67) was cultured in Dulbecco’s Modified Eagle Medium (DMEM) High Glucose (Sigma-Aldrich) supplemented with 10% heat-inactivated Fetal Bovine Serum (hi-FBS, Corning) at 37 °C and 5% CO₂.

Murine BALF Transfer.

All animal experiments were approved by the Washington University Animal Care and Use Committee, and we have complied with all relevant ethical regulations. For BALF transfer, 3 d before infection, 6- to 8-wk-old female C3H/HeJ (Jackson Laboratory, Bar Harbor, ME) mice were treated intraperitoneally with 100 μL of 300 mg/kg (6 mg/mouse) of cyclophosphamide (TCI), dissolved in PBS and filtered. On the day of the first infection, overnight bacterial cultures were subcultured at a 1:200 dilution in fresh LB and grown with shaking at 37 °C for 3 h to mid-exponential growth phase. Cultures were washed twice with PBS and normalized to OD_600nm = 100. Naive C3H/HeJ mice were anesthetized with 4% isoflurane and were intranasally inoculated with 10⁹ CFU (35 µL of culture). At 24 hpi, mice were killed by isoflurane overdose, and BALF was collected in 1 mL of PBS-EDTA 1 mM. BALFs were pooled followed by red blood cell lysis by incubation for 5 min on ice in 5 mL of Pharm Lyse Buffer diluted in PBS (BD Biosciences 555899), and the reaction was stopped by the addition of 30 mL of PBS. Cells were then incubated for 1 h at 37 °C in DMEM 20% hi-FBS with colistin 50 μg/mL for Ab398 [Minimal inhibitory concentration (MIC); MIC_Ab398= 2 μg/mL] and 100 μg/mL for Ab19606 (MIC_Ab19606= 4 μg/mL) under agitation to eliminate all extracellular bacteria. Cellular content was centrifuged at 600×g for 6 min, washed twice with PBS, and resuspended in PBS (10 BALFs pooled, treated, and resuspended in 300 μL). Subsequently, cyclophosphamide-treated mice were intranasally inoculated with 35 μL of the pooled treated BALF. Intracellular bacterial count was determined by lysis of the host cells by Triton X-100 (0.025%), and serial dilutions of bacterial suspensions were plated for CFU counting. At 24 hpi, mice were killed, and CFU in the lungs, spleen, and kidneys were quantified by serial dilution plating the homogenized organs. Additionally, half of the lung homogenate (500 µL) was plated onto LB chloramphenicol to lower the bacterial detection limit (as compared to the 30 µL usually plated for CFU measurements).

Antibiotic Protection Assay.

Bacterial intracellular replication was performed as previously described (8). Briefly, J774A.1 cells were seeded 16 h before the experiment in 48-well plates (3 × 10⁵ cells/well). A. baumannii overnight cultures were inoculated from fresh LB-chloramphenicol plates. Bacterial cultures were pelleted for 5 min at 6,500 rpm and washed twice with PBS. Bacterial cultures were normalized to OD_600nm = 1, and appropriate volumes were added to 500 μL of DMEM high glucose (Sigma-Aldrich), 10% hi-FBS per well to achieve an MOI = 10. Infected cells were centrifuged for 10 min at 200×g and incubated at 37 °C and 5% CO₂. 1 hpi, macrophages were washed thrice with PBS and treated with DMEM supplemented with 10% FBS and 50 μg/mL of colistin to eliminate extracellular bacteria. At the indicated time points, cells were washed thrice with PBS and lysed with 500 μL of Triton X-100 (0.025%) per well. Serial dilutions of bacterial suspensions obtained were plated. Bacterial replication was assessed by CFU counting. Replication FC for each timepoint was calculated by ratio compared to the average of the two technical duplicates at the 2 h timepoint. Data were collected from at least three independent experiments.

Dual Comparative Proteomic Analysis.

GSEA.

The GSEA was performed using the software GSEA (version 4.3.3). Imputed LFQ data were loaded into the software, and pathway enrichment was conducted using the mouse molecular signature database MH: hallmark gene sets (mh.all.v2024.1). Enrichment scores were visualized using SRplot online platform (71).

Bacterial Homolog Identification and COG Annotation.

Genes in two strains were considered homologs if they had at least 70% identity and 90% coverage. This analysis was performed using reciprocal BLAST hit (RBH) on the usegalaxy.eu platform. Bacterial protein fasta files were submitted to EggNOG to obtain predicted COG annotation of each bacterial protein (72).

Additional Material and Methods.

A full description of additional material and methods is available in SI Appendix.