This study (RB number L21-156 and L13-140) were approved by the Texas Tech University Health Sciences Center Lubbock Institutional Review Board on 12/07/2021 and 07/23/2014, respectively and remained open until all study activities were completed. Informed written consent was obtained from healthy volunteers (HVs) and burn patients (BPs) at University Medical Center, Lubbock, TX, in compliance with ethical practices. HVs had no acute or chronic medical conditions and all subjects were aged 18–89. Blood samples were collected by CRI staff through venipuncture as per the IRB-approved protocol. Blood was collected within 72 h of admission. A total of 25 mL of blood was collected from each person into three BD Vacutainer tubes (Becton Dickinson, NJ, USA), containing 0.35% sodium polyanethole sulfonate as an anticoagulant in 1.7 mL of 0.85% sodium chloride (SPS). Per the IRB protocol, blood samples were de-identified and were given unique numbers by the CRI staff before the samples were sent to the research laboratory. All methods performed on the samples were in compliance with the relevant guidelines and regulations of the IRB-approved protocol.

Acinetobacter baumannii strain A118 was received from ATCC and used in all experiments. This strain was isolated from a blood culture of an intensive care unit patient in 1995 in Buenos Aires City, Argentina [18]. Bacteria were routinely cultured overnight in Luria-Bertani (LB) broth at 37°C with shaking prior to subculturing them into whole blood samples, LB, or LB supplemented with 10% of pooled human serum (LB + 10% serum) (MP Biomedical, CA, USA) in 1:100 ratio. In growth curve assays, bacterial growth of 3 independent replicates was monitored by measuring colony-forming units (CFU) every 2 h over an 8-h period. For bacterial RNA extraction from LB and LB + 10% serum cultures, samples were taken from three individual replicates per condition at 4 and 8 h post inoculation (HPI). The samples were then pelleted, resuspended in LB, treated with RNAprotect (Qiagen, Germany), and subsequently processed for bacterial RNA extraction.

Whole blood from three BD Vacutainer tubes of each subject (HV or BP) was pooled to minimize RBC lysis. Aliquots of 7.5 mL were distributed into three flasks as technical replicates, inoculated with overnight A. baumannii A118 LB culture as previously mentioned, and incubated at 37°C. After 4 h, when bacterial growth reached ~10⁸ CFU/mL, samples were diluted 1:1 with 1X PBS and layered over lymphocyte separation medium (Lonza, Switzerland) and were centrifuged according to the manufacturer’s protocol. Then, the layers containing lymphocytes and granulocytes were discarded, while the RBC and bacterial pellet layer was saved and treated repeatedly with erythrocyte lysis buffer (Qiagen, Germany) and was centrifuged according to the manufacturer’s protocol until all RBCs were lysed and only bacterial cells remained. The pellet was resuspended in LB, treated with RNAprotect (Qiagen, Germany), and then processed for bacterial RNA extraction.

Bacterial RNA was extracted using the RNeasy Mini Kit (Qiagen, Germany) according to the manufacturer’s recommendations followed by an additional on-column DNase treatment to eliminate any remaining traces of genomic DNA. Total RNA concentration and integrity were determined by Nanodrop spectrophotometer (Nanodrop Technologies, DE, USA) and TapeStation 2,200 (Agilent, CA, USA), respectively. rRNA depletion was achieved using NEB Next rRNA Depletion Kit (Human/Mouse/Rat) (New England Biolabs, MA, USA). RNA fragmentation, double stranded cDNA synthesis and adaptor ligation were carried out using NEBNext Ultra II Directional RNA Library Prep according to the manufacture protocol (New England Biolabs, MA, USA). PCR enriched libraries were quantified using picogreen (Thermofisher scientific, MA, USA) and equimolar indexed libraries were pooled. Pooled libraries were assessed using the Agilent Tapestation 2,400 (Agilent, CA, USA). The libraries were then diluted to 250 pM, spiked with 1% phiX libraries (Illumina control), and sequenced using Illumina NovaSeq 6,000 (Illumina, CA, USA).

Paired-end reads of each sample were mapped to the A. baumannii A118 reference transcriptome (NCBI RefSeq assembly: ASM1467273v1) by the pseudo-alignment-based tool, Kallisto (Bray et al., 2016) with 100 bootstraps per sample for quantification of transcript sequences. Then, Kallisto output files were imported to R software (version 4.2.1) (R Core Team, R, 2013) and DESeq2 (version 1.36.0) (Love et al., 2014) was used to normalize transcript counts and test for differential gene expression by the Wald test after integrating the dispersion estimate. The p-values were corrected for multiple testing using the Benjamini–Hochberg procedure (Benjamini and Hochberg, 1995). Differentially expressed genes (DEGs) were identified based on their corresponding adjusted p-value (padj ≤ 0.05 and |log2foldchange| ≥ 1).

GO analysis of DEGs was performed by STRING¹ (Szklarczyk et al., 2019). False Discovery Rate (FDR) describes the significance of the enrichment, with p-values adjusted for multiple testing within each category using the Benjamini–Hochberg procedure. Strength values were calculated by observed/expected ratio. Signal values were calculated as a weighted harmonic mean between the observed/expected ratio and -log(FDR). The FDR emphasizes larger terms, as they are more likely to yield lower p-values, whereas the strength value gives prominence to smaller terms, which often have a high foreground-to-background ratio but cannot achieve low FDR values due to their limited size. However, signal parameter aims to balance these two metrics, creating a more intuitive ranking of enriched terms (von Mering et al., 2005). Therefore, we selected it as the primary metric for displaying GO terms in our graphs. The other key enrichment display setting parameter defined as term similarity by STRING, filters terms using the Jaccard index to measure similarity between gene sets within a category (Szklarczyk et al., 2025). Terms were prioritized by p-value, and those exceeding a user-defined similarity threshold were excluded. The enrichment display settings for our analysis were adjusted as follows: Term similarity ≥0.70, FDR ≤ 0.05, Signal ≥ 0.50, Strength ≥ 0.25, and Minimum count in the network was set to 10. For network construction, we used the Full STRING network as the network type, selecting all available options for active interaction sources. The minimum required interaction score was set to 0.4, and the maximum number of interactions to show for both the first and second shell was set to none.

Biofilm formation of bacteria was assessed as previously described (Hammond et al., 2011; Alqahtani et al., 2021) with some modifications. Briefly, standardized bacterial cultures of A. baumannii A118 and P. aeruginosa PAO1 were prepared based on OD₆₀₀ measurements. Then, 10 μL of each culture was added to each well of a 24-well plate containing 1 mL of fresh LB or LB containing different percentages of serum, ranging from 5 to 20% depending on the experiment. One sterile 6 mm cellulose disk (Becton Dickinson and Company, MD, USA) was placed in each well, and plates were incubated at 37°C shaking incubator for 24 h. After incubation, to determine the number of CFUs present in the biofilm, the disks were removed with sterile forceps, rinsed twice with 1x PBS to remove planktonic cells and were placed individually in microtubes containing 1 mL of 1x PBS. The microtubes were vortexed vigorously for 1 min to dislodge biofilms from the disks and CFU counts were then measured. Each experiment was performed in triplicates.

Antimicrobial susceptibility testing was conducted as previously described (Huang et al., 2020; Le et al., 2021) with some modifications following the procedures recommended by the CLSI (Wikler, 2006). After OD₆₀₀ adjustment, 100 μL of bacterial LB cultures were inoculated on Mueller-Hinton agar plates (MH) (Sigma-Aldrich, MA, USA) with and without 10% pooled human serum. Antimicrobial commercial E-strips (Liofilchem S.r.l., Italy) for Chloramphenicol (C), Imipenem (IMI), Amikacin (AK), Rifampicin (RD) and Tobramycin (TOB) were aseptically placed onto individual plates. Then, plates were incubated at 37°C for 18 h. Each antibiotic was tested in three replicates. Minimum inhibitory concentration (MIC) readings were performed according to the manufacturer’s protocol.

The OMPs extraction procedure was followed as previously described (Beasley et al., 2020) with some minor modifications. LB and LB + 10% serum were inoculated with A. baumannii A118 as previously mentioned and grown for 4 h. Then, cultures were centrifuged and cells were washed with sterile water and lysed by sonication (10 min pulses at 50% power 3 times) using a Kinematica Polytron P10-35 PCU-11 homogenizer (Kinematica AG, Switzerland). Lysed cells were centrifuged for 20 min at 18,450 × g to obtain a pellet containing both inner and outer membrane proteins. This pellet was resuspended in sterile water and sarcosyl (MilliporeSigma, Germany) was added to a final concentration of 1%. The mixture was then incubated for 1 h at room temperature with continuous rotation to remove the inner membrane proteins. The samples were subsequently ultracentrifuged for 1 h at 100,000 × g. Finally, the pellets were collected and washed three times with sterile water. The resulting pellet, rich in OMPs, was resuspended in sterile water. The total protein concentration of each sample was then measured using the Bradford assay (Bio-Rad, Hercules, CA, USA). Equal amounts of the prepared OMPs (30 μg) were separated by SDS-PAGE and gels were stained by coomassie blue to visualize the OMP profiles. We selected an estimated <75 kDa band that appeared to be strongly upregulated upon the growth of A. baumannii A118 in LB + 10% serum. Selected protein band was excised from the SDS gel, protein was eluted and identified by LC–MS/MS (Texas Tech University Center for Biotechnology and Genomics, Lubbock, TX). For data processing, Proteome Discoverer (Ver. 2.4) (Thermo Scientific, MA, USA) was used to generate peptide mass spectra, mass list, coverage percentage and Score Sequest HT.

The samples were prepared for ICP-MS analysis using a previously established protocol with minor modifications (Wakeman et al., 2014). Bacterial serum-treated cultures were collected, pelleted, and washed in 500 μL of Chelex-treated 1X phosphate-buffered saline (PBS) at two time points (4 and 8 HPI). Three biological replicates were collected for each time point. The pellet was resuspended in 1 mL of Chelex-treated, metal-free water and sonicated at 100% amplitude for 30 s per sample. To prevent cross-contamination, the sonication probe was thoroughly cleaned between samples by wiping and rinsing with Chelex-treated water. To monitor for potential metal contamination during the process, a water control was sonicated after every two samples. Protein concentrations were normalized across samples, and 100 μL of each sample was digested by adding 1 mL of 50% Optima-grade nitric acid (Fisher Scientific). Digestion was performed overnight at 50°C in metal-free 15-mL conical tubes. Following digestion, samples were filtered (0.22 μm, nylon, VWR), and gravimetrically diluted as needed so that the concentrations of target elements fell within the calibration range (0 to 500 ppb) of the inductively coupled plasma mass spectrometer (ICP-MS) (Agilent 7,900). Concentrations of magnesium (Mg), zinc (Zn), copper (Cu) and potassium (K) were measured in helium mode while iron (Fe) was measured in hydrogen mode. Quality analysis and quality control included the use of rhodium (Rh-103) and indium (In-115) as internal standards, as well as blanks and continuing calibration checks which were run with every batch of 10 samples. Calibration standards were prepared in 15 mL polypropylene tubes (VWR) with 1% HNO3. Detection limits for the quantified elements were estimated using the lowest calibration standard with 80 to 120% recovery: Fe = 0.5 ppb, Mg = 1.0 ppb, Zn = 0.1 ppb, Cu = 0.1 ppb, K = 1.0 ppb. Internal standard recoveries were 87 to 92% and continuing calibration check recoveries were in the range 92 to 116%. Each metal ion measurement from the biological replicates was normalized by subtracting the corresponding blank value for that specific ion.

Galleria mellonella infection assay was performed as previously described (Gaddy et al., 2012) with some modifications. Larvae weighing 200 to 400 mg that were kept on wood chips in the dark at 4°C, were used for this assay. A. baumannii A118 was cultured in either LB or LB + 10% serum for 8 h. Bacterial cultures were diluted in 1X PBS and adjusted to a concentration of 1 × 10⁵ CFU/mL based on the OD600. The bacterial inocula were verified by plating serial dilutions on LB agar and counting CFU after overnight incubation at 37°C. Each experiment was conducted six times with 10 larvae per group, totaling 60 G. mellonella larvae for each condition. Each larva was injected with 10 μL of either sterile 1X PBS or a standardized bacterial culture from LB or LB + 10% serum at the last left proleg and incubated at 37°C in sterile Petri dishes, with viability monitored every 24 h for a total of 120 h. Larvae were evaluated based on their color and by gently prodding them with a glass rod; those that showed no response were classified as dead. Survival curves were generated using the Kaplan–Meier method (Kaplan and Meier, 1958) using GraphPad Prism (GraphPad software, San Diego, CA, USA).

1.5% agar LB plates with and without 10% pooled human serum, were prepared fresh for each experiment. Each twitching motility plate was stab inoculated with a colony of bacteria at the agarose/petri plate interface and placed in a 37°C, humidified incubator for 48 h. To visualize the bacteria at the interface, agarose was removed from each plate and the opaque zone of bacterial movement on the plastic surface was stained by 0.1% crystal violet for 10 min. Then plates were washed several times with DI water until the rinse ran clear.

Statistical analyses and graphical plotting of all experiments results except RNA-Seq analysis were conducted using R 4.2.1 and GraphPad Prism 9.5.1 (GraphPad Software, Inc., San Diego, CA). One-way ANOVA followed by Tukey’s multiple-comparison test was used to assess statistical significance of ICP-MS and biofilm formation data.

We analyzed the transcriptomic changes of A. baumannii A118 grown in WBBP compared to WBHV, with each condition including samples from three patients or healthy volunteers. In the PCA plot, PC1 accounts for 33.65% of the total variance, representing the largest source of variation in the dataset and indicating a moderate but meaningful degree of separation among the samples (Supplementary Figure 1). RNA-Seq analysis comparing A. baumannii A118 grown in WBBP versus WBHV identified 118 DEGs (padj ≤ 0.05, |log2FoldChange| ≥ 1), with 59 genes upregulated and 59 downregulated under the WBBP condition (Supplementary Figure 2; Supplementary Table 1).

GO analysis of downregulated genes revealed that, two GO terms within the biological process category were significantly enriched under the WBBP condition, both showing substantial overlap (Figure 1A). According to STRING’s GO analysis, the nonribosomal peptide biosynthetic process term had the higher signal value and STRING assigned a total of 17 genes—within our network and the background—to this term. Eight downregulated genes from our DEGs were associated with this term (Supplementary Figure 3). The second enriched GO term—siderophore biosynthetic process—includes 7 of the same aforementioned 8 genes—excluding only gamma-glutamyltransferase—and 14 genes in total were annotated with this term based on the STRING database.

The presence of multiple acinetobactin biosynthesis and transport genes within both GO terms led us to manually examine the full list of downregulated genes to better understand how WBBP affects the production and transport of this siderophore. Out of the 18 genes in the acinetobactin operon, 11 were found to be downregulated in our dataset (Figure 1B).

Although STRING GO analysis did not reveal a significantly enriched GO term among the upregulated genes, it identified a local network cluster labeled: “Nitrate assimilation and nitronate monooxygenase activity” (FDR = 0.0087, signal = 0.67, count in network = 4 of 6). This network cluster highlights two of the most highly upregulated genes within our DEGs: nirB (nitrite reductase large subunit, log2FoldChange = 4.48) and nirD (nitrite reductase small subunit, log2FoldChange = 4.04). The other two genes in this network cluster—nasA (nitrate reductase, log2FoldChange = 2.42) and a nitrate/nitrite transporter (log2FoldChange = 3.23)—were also among the most highly upregulated DEGs, further supporting the activation of nitrate assimilation pathways in response to WBBP (Supplementary Table 1).

To better understand the effects of specific components of the human blood that would influence A. baumannii A118 gene expression, we also analyzed transcriptomic changes in response to 10% human serum added to LB. While the whole blood experiment provided insight into host-pathogen interactions in physiologically relevant but complex ex vivo conditions, the LB + 10% serum model allowed us to isolate and investigate the effects of serum-derived signals in a more defined environment. By comparing these datasets, we aimed to distinguish between general serum-driven responses and those specific to the altered immune/nutritional status of burn patients.

The initial evaluation of serum’s effect on A. baumannii A118 involved assessing its standard growth curve over an 8-h period. The growth curve analysis was conducted in both LB and LB + 10% serum (Figure 2). The initial bacterial inoculum for both media was similar and around 2–1.8 × 10⁷ CFU/mL. As the incubation continued, significant differences in growth between the two conditions emerged. At the 2 HPI, the CFU/ml in LB decreased to 7.0 × 10⁶ CFU/mL, while the decline in LB + 10% serum, was more moderate, reaching 1.75 × 10⁷ CFU/mL. This period might represent an adaptation phase, typically showing a reduction from the starting inoculum. Therefore, the addition of human serum appears to have attenuated this decline compared to pure LB.

By 4 HPI, bacterial growth in LB reached 4.15 × 10⁸ CFU/mL, while that in LB + 10% serum was significantly increased reaching 1.2 × 10⁹ CFU/mL. The period between this point and that at 6 HPI, may represent a short stationary phase. At the 6 HPI, the bacterial concentration in LB broth dropped slightly to 4.05 × 10⁸ CFU/mL, while the presence of serum resulted in a slightly higher concentration—1.55 × 10⁹ CFU/mL—compared to 4 HPI point of this condition.

At the 8 HPI, the bacterial concentration in LB decreased to 1.65 × 10⁸ CFU/mL, indicating a decline in growth, while in LB + 10% serum, the CFU/ml only slightly decreased to 1.03 × 10⁹. While the exact timing of the growth phases may vary between the two conditions, our data demonstrates that compared to growth in LB alone, the addition of human serum significantly enhances the overall growth of A. baumannii A118.

To gain a deeper understanding of how human serum influences gene expression in A. baumannii A118 and the potential role of the growth phase in this influence, we compared bacterial transcriptomic changes at two distinct time points in response to human serum. We selected 4 HPI, which represents the peak of rapid and logarithmic growth in both conditions, and 8 HPI, at which both cultures are likely in the mid-late stationary phase. For each condition and each time point, we examined three biological replicates. Importantly, we obtained the time-points samples at 4 and 8 HPI from the same flask (Supplementary Figures 4, 5). At 4 HPI, the PCA showed a PC1 value of 71% (Supplementary Figure 4), however by 8 HPI, the separation between the LB and LB + 10% serum conditions became more pronounced, with PC1 increasing to 75% (Supplementary Figure 5). This greater variance suggests stronger transcriptional divergence over time. Typically, a higher PC1 value correlates with a larger number of DEGs, which is consistent with our findings.

At 4 HPI, RNA-Seq analysis revealed 197 DEGs between LB and LB + 10% serum (padj ≤ 0.05, |log2foldchange| ≥ 2), with 63 genes upregulated and 134 downregulated in the presence of serum (Supplementary Table 2). In contrast, at 8 HPI, the number of DEGs nearly doubled to 386, with 144 upregulated and 242 downregulated (Supplementary Table 3). These results indicate that the influence of human serum on gene expression becomes more pronounced as bacterial growth progresses, leading to broader transcriptional reprogramming at later stages.

To identify pathways affected by serum at both time points, we performed STRING GO analysis on DEGs from our RNA-Seq data. At 4 HPI, comparing LB + 10% serum to LB, we detected significant enrichment in the biological process category for both up- and down-regulated genes (Figures 3A,C). Among the upregulated genes, the only enriched GO term was siderophore biosynthetic process (5 of 14 genes in the network according to STRING; Figure 3A), which was consistent with our manual search of all 18 genes present on acinetobactin operon. We found 4 out of these 5 identified genes among 18 genes present on acinetobactin operon (Figure 3B). The fifth gene from this GO term is annotated as siderophore biosynthesis protein (protein ID: WP_168726283.1). Among the downregulated genes, two GO terms were significantly enriched—protein secretion and protein transmembrane transport— with 9 of 42 and 9 of 48 genes in the network, respectively (Figure 3C). Together, these findings suggest that serum exposure does not cause a broad response at 4 HPI but rather triggers a specific reprogramming of pathways important for host adaptation—especially iron acquisition—while deprioritizing others. It reflects a calculated and efficient early adaptation strategy to host-derived signals.

The notable downregulation of GO terms related to protein secretion and transmembrane transport at 4 HPI in response to human serum suggests a potential shift in membrane permeability or transporter activity that might affect antibiotic import/export. Such changes could directly influence the bacterium’s susceptibility to antibiotics. Many antimicrobials, such as Imipenem and Rifampicin, rely on specific outer membrane proteins or active transport mechanisms to enter bacterial cells. Similarly, the activity of efflux systems affects intracellular concentrations of antibiotics like Chloramphenicol, Amikacin, and Tobramycin. Therefore, this transcriptomic change prompted us to examine whether the presence of serum influences antibiotic susceptibility by performing MIC assays in MH agar plates versus MH agar plates supplemented with 10% human serum. Human serum was added directly to MH agar plates to ensure continuous exposure during the 18-h MIC assay. This method better reflects serum’s influence on antibiotic susceptibility compared to pre-growing bacteria in serum, which may lose its effect once transferred to serum-free plates.

We observed increased MIC values for all tested antibiotics when bacteria were grown on MH agar plates supplemented with 10% human serum, compared to MH alone, with varying degrees of change (Table 1; Supplementary Figure 6). The most notable change was seen with amikacin, where the MIC increased from 0.75 mg/L to 3 mg/L, indicating a four-fold decrease in susceptibility. Rifampicin also showed a substantial increase in MIC from 1.5 mg/L to 4 mg/L. Imipenem and tobramycin both demonstrated moderate increases from 0.5 to 0.75 mg/L, while chloramphenicol exhibited a smaller change, rising from 0.24 to 0.32 mg/L. These findings suggest that the presence of human serum reduces bacterial susceptibility to a broad spectrum of antibiotics, potentially due to serum-induced changes in membrane permeability and transporter expression.

The downregulation of GO terms related to protein secretion and protein transmembrane transport at 4 HPI and increase MIC values of several antibiotics in response to human serum, led us to hypothesize that the OMP profile of A. baumannii A118 grown in LB + 10% serum is potentially different from that in LB alone. To test this, we extracted OMPs from both conditions at 4 HPI and analyzed them via SDS-PAGE, loading equal protein concentrations (Figure 4). Although the overall banding patterns appeared broadly similar between the two conditions, we recognize that SDS-PAGE combined with Coomassie staining lacks the sensitivity and resolution to detect subtle or global shifts in OMP abundance. However, a band just below the 75 kDa marker was visually more prominent in the OMP of bacteria grown in the presence of serum. Therefore, we isolated this band for further analysis and identification. The best match for this band, based on LC–MS/MS analysis, was an outer membrane protein annotated as FepA family TonB-dependent siderophore receptor (Protein ID: WP_228129848.1; molecular weight: 70.6 kDa; number of matched peptides: 15; coverage: 33%; Sequest HT score: 308.77) (Supplementary Table 4). The increase in the production of this protein in the presence of serum correlates with the observed increase in the expression of acinetobactin related genes at 4 HPI.

A manual search of the DEGs at 4 HPI did not identify this gene; however, its presence and relatively high abundance on the SDS-PAGE gel suggests either a post-transcriptional regulation of the acinetobactin operon or a potential increase in the stability of the synthesized protein. Alternatively, this may reflect a temporal disconnect between transcriptomic and proteomic responses, where this gene may be upregulated at earlier time points prior to upregulated at 4 HPI, but the transcriptional level returns to baseline at 4 HPI. However, the protein remains stable and abundant in serum-treated cultures. Additional experiments are required to validate this hypothesis and determine the underlying mechanisms.

At 8 HPI, several significantly enriched GO terms were identified among the downregulated genes (Figure 5A), whereas none were found among the upregulated genes at this time point. Interestingly, the GO term siderophore biosynthetic process was significantly downregulated (5 out of 14 genes in the network) in contrast to the 4 HPI time point, where this term was upregulated. Upon manually reviewing the list of DEGs, we found that 15 out of 18 genes in the acinetobactin operon, as well as entA, were downregulated at 8 HPI (Figure 5B). entA is not located on the acinetobactin operon, however it is essential for the biosynthesis of its precursor. entA was not differentially expressed at 4 HPI, highlighting a more pronounced regulatory shift at 8 HPI.

Importantly, the aforementioned GO term significantly overlaps with the top-enriched GO term based on signal value (Figure 5A)—secondary metabolite biosynthetic process—which underscores that the primary target of downregulation at 8 HPI is acinetobactin related.

Regulation of gene expression related to acinetobactin biosynthesis and transport emerged as a key theme in response to human serum at both time points analyzed. At 4 HPI, we observed upregulation of several acinetobactin related genes in response to serum, with a maximum log2FoldChange of 1.8 (Figure 3B). However, by 8 HPI, not only were more of these genes downregulated, but the extent of downregulation was also more pronounced, reaching log2FoldChange as low as −3.7 (Figure 5B). This pattern led us to hypothesize that during early and logarithmic phases of growth, serum-treated cultures produce high levels of acinetobactin, facilitating iron uptake. As a result, by 8 HPI, bacteria may have accumulated sufficient intracellular iron and, to conserve energy during the transition toward stationary or death phase, began downregulating acinetobactin-related genes. Therefore, we expected higher intracellular iron levels at 8 HPI compared to 4 HPI.

To test this, we performed ICP-MS analysis on the same serum-treated cultures used for RNA-Seq analysis. Samples were collected at 4 HPI, returned to the incubator, and re-sampled at 8 HPI. ICP-MS results revealed an increase in intracellular levels of several metal ions, including iron, at 8 HPI compared to 4 HPI in serum-treated cultures (Figure 6). While the differences did not reach statistical significance—likely due to limited sensitivity of the instrument causing high variability of the reads between biological replicates— The observed 2.22-fold increase in iron concentration at 8 HPI supports our hypothesis that iron accumulation contributes to the downregulation of acinetobactin genes at this time point. The only other metal ion that showed a fold increase greater than 2 was zinc, with a 2.38-fold increase at 8 HPI.

We observed that several biofilm-associated genes were upregulated at both 4 and 8 HPI in response to human serum (Supplementary Figures 7, 8), indicating a potential role for serum in enhancing the biofilm-forming ability of A. baumannii A118. Among the upregulated genes at 4 HPI were those of the csu operon, which contribute to early stages of biofilm formation and are also involved in twitching motility. Therefore, we conducted series of experiments to assess the potential effects of serum on both twitching motility and biofilm formation. While our transcriptome analysis was conducted at 4 and 8 HPI, both twitching motility and biofilm formation assays are commonly conducted on cultures grown to 24 or 48 HPI. Thus, it is feasible that a considerable time—24 h or more—is required for proteins encoded by the early regulated genes to produce the observed two phenotypes.

To test whether serum influences biofilm formation, we first compared biofilm production by A. baumannii A118 in LB alone versus LB + 10% serum (Figure 7). While biofilm formation appeared to increase in the presence of 10% serum, the difference was not statistically significant. It possible that at higher concentrations, serum may produce a more prominent effect on biofilm formation. Therefore, we expanded the experiment to test a range of serum concentrations (5–20%). This gradient revealed a consistent, dose-dependent increase in biofilm formation, with statistical significance observed at the 20% serum level. It is possible that this observed effect is not unique to A. baumannii A118. Rather, at 20% concentration, serum may enhance biofilm formation by other Gram-negative pathogens. To explore this possibility, we examined the effect of human serum on biofilm formation by Pseudomonas aeruginosa— a Gram-negative opportunistic pathogen. We selected PAO1, a well-characterized and virulent P. aeruginosa laboratory strain known for its robust and consistent biofilm formation under various laboratory conditions. As expected, PAO1 produced more biofilm than A. baumannii A118 in pure LB, and this difference was statistically significant (Figure 7). However, in contrast to its effect on A. baumannii A118, the addition of 10% human serum to LB significantly reduced PAO1 biofilm formation by 2.8 × 10³-fold. These findings reveal a distinct trait of A. baumannii A118 and will be expanded in details in future studies.

These results align with our twitching motility assays, which also showed increased motility in A. baumannii A118 when exposed to 10% serum (Supplementary Figure 9), further supporting the hypothesis that serum influences both biofilm formation and twitching motility in this strain.

Our RNA-Seq results and supporting phenotypic assays, all pointing toward serum-induced changes in several bacterial virulence factors. To assess whether these effects translated to a detectable increased in vivo virulence, we used the well-established G. mellonella infection model. To capture he full effects of serum, we chose to grow bacteria for 8 h as we observed the biggest DEGs numbers in our 8 HPI RNA-Seq analysis. Larvae were injected with 10 μL of either sterile 1X PBS as the control or standardized bacterial cultures grown in either LB or LB supplemented with 10% human serum. The survival curves over the 5-day period reveal a notable difference (Figure 8). The PBS control group exhibited minimal mortality, maintaining 93.3% survival. Larvae infected with LB-grown bacteria showed a gradual decline to 90% by 48 h and 86.6% by 72 h, remaining stable thereafter. In contrast, larvae infected with serum-exposed bacteria experienced a rapid drop to 61.6% survival within 24 h, decreasing further to 31.6% by 96 h. These results suggest that exposure to serum enhances the virulence potential of A. baumannii A118.