Multi-metal restriction by calprotectin impacts de novo flavin biosynthesis in Acinetobacter baumannii

Summary

Calprotectin (CP) inhibits bacterial viability through extracellular chelation of transition metals. However, how CP impacts general metabolism remains largely unexplored. We show here that CP restricts bioavailable Zn and Fe to the pathogen Acinetobacter baumannii, inducing an extensive multi-metal perturbation of cellular physiology. Proteomics reveals severe metal starvation, and a strain lacking the candidate Zn^II metallochaperone ZigA possesses altered cellular abundance of multiple essential Zn-dependent enzymes and enzymes in de novo flavin biosynthesis. The ΔzigA strain exhibits decreased cellular flavin levels during metal starvation. Flavin mononucleotide provides regulation of this biosynthesis pathway, via a 3,4-dihydroxy-2-butanone 4-phosphate synthase (DHBPS; RibB) fusion protein, RibBX, and authentic RibB. We propose that RibBX ensures flavin sufficiency under CP-induced Fe-limitation allowing flavodoxins to substitute for Fe-ferredoxins as cell reductants. These studies elucidate adaptation to nutritional immunity and define an intersection between metallostasis and cellular metabolism in A. baumannii.

In Brief

The host protein calprotectin (CP) inhibits pathogen growth via sequestration of essential transition metals. This work shows that CP impacts the homeostasis of multiple metals as well as flavin biosynthesis in Acinetobacter baumannii and identifies potential targets for the development of antibacterial agents in the infected host.

Graphical Abstract

Introduction

Transition metals are important to microbial physiology and execute key catalytic, structural, and regulatory functions that broadly impact all phases of cellular physiology. The mammalian host takes advantage of this essentiality by restricting growth of invading bacteria through a process termed nutritional immunity, in which transition metals are withheld from microbial invaders by mammalian metal-binding proteins (Damo, et al., 2013; Hood and Skaar, 2012; Weinberg, 1975). Host neutrophils and other immune cells secrete a number of metal-chelating innate immune proteins at sites of infection, including S100-family proteins. One such protein is calprotectin (CP, S100A8/A9 oligomer), which is a versatile extracellular metal-sequestering protein that forms high thermodynamic and/or kinetic stability coordination complexes with Zn^II, Mn^II, Fe^II and Ni^II (Damo, et al., 2013; Nakashige, et al., 2015; Nakashige, et al., 2017). CP-dependent metal-binding restricts the availability of these metals and inhibits the growth of pathogens (Corbin, et al., 2008; Zygiel and Nolan, 2018). While the physiological impact of CP is expected to be niche- and pathogen-dependent, little is known about how CP induced stress impacts metabolism in different pathogens (Besold, et al., 2017; Wakeman, et al., 2016).

Acinetobacter baumannii is a Gram-negative bacterial pathogen commonly associated with hospital-acquired infections and is a leading cause of ventilator-associated pneumonia (Doyle, et al., 2011). Increasing reports of community-acquired and multidrug resistant Acinetobacter infections has attracted considerable public attention (Harding, et al., 2018). Calprotectin (CP) inhibits A. baumannii growth and protects against A. baumannii infection in a mouse model of pneumonia (Hood, et al., 2012). A transposon mutagenesis screen identified a strain lacking the Zn uptake transporter znuB, which is regulated by the Zn uptake repressor Zur, as being defective for growth in the presence of CP and is consistent with the finding that CP induces Zn but not Mn starvation (Hood, et al., 2012). Subsequent transcriptomic experiments with the Δzur strain identified the Zur regulon as well as other Zn-responsive genes, which included the putative Zn^II chaperone, ZigA (Zur-induced GTPase A) (Mortensen, et al., 2014); follow-up work revealed that the ΔzigA strain is sensitive to CP stress (Mortensen, et al., 2014; Nairn, et al., 2016). While the importance of Zur and the induction of known Zur-regulated genes in response to CP has been defined (Mortensen, et al., 2014; Nairn, et al., 2016), it is not yet known how other metal homeostasis and metabolism systems are perturbed by CP stress.

Elucidating a global understanding of CP-induced perturbations in bacterial transition metal distribution network(s) and downstream physiological changes in cellular metabolism is important but remains largely understudied (Zygiel and Nolan, 2018). Cells must alter their metabolism when they are metal-starved. A number of essential metabolic pathways are initiated with a Zn^II metalloenzyme, including the biosynthesis of folate, queuosine-tRNA, riboflavin, and in some organisms, tetrahydrobiopterin (Phillips, et al., 2012). Given that enzymes that require Zn^II or another transition metal for activity are positioned at critical hubs in cellular physiology, it is unclear how bacterial cells balance cellular metabolism under CP-induced metal-deficient conditions. In addition, while studies of how cells regulate metal transport are widespread, these studies have tended to underestimate how intermediary metabolism adapts to host-induced metal starvation. In A. baumannii, CP moderately induces the genes required for L-histidine catabolism (hut), which are proposed to contribute to the release of Zn^II from Zn-His₂ complexes, thus increasing access by Zn^II-dependent processes under extreme Zn limitation (Nairn, et al., 2016). Additionally, iron starvation has been linked to induction of flavin biosynthesis in bacteria and eukaryotes where substitution of flavodoxins for ferredoxins as cellular reductants was proposed (LaRoche, et al., 1996; Tognetti, et al., 2007). A global, systems-level view of how CP impacts metabolic processes is clearly needed.

The studies presented here describe the metabolic impact of nutritional immunity in A. baumannii. Using A. baumannii ΔzigA as a tool to exacerbate multi-metal (Zn and Fe) starvation induced by CP, we observe an incrementally increased abundance of two related enzymes of the flavin biosynthesis pathway, RibBX and RibB, consistent with cellular efforts to ensure sufficient flavin supply under conditions of extreme metal limitation or mis-allocation, predicted of the basis of global changes in Zn^II proteome. These observations are consistent with our discovery that ΔzigA is unable to maintain appropriate flavin abundance during metal starvation and that ribBX contributes to the ability of A. baumannii to synthesize flavins and overcome nutrient metal limitation. Our findings support a dual regulatory model explaining how A. baumannii maintains cellular flavin sufficiency required for central metabolic processes in an effort to overcome CP-induced metal starvation. Our studies also validate ongoing efforts to develop pharmacological strategies against A. baumannii by identifying flavin biosynthetic enzymes as required to survive nutrient limitation. Notably, these enzymes are established antimicrobial targets in other bacterial pathogens, underscoring the general significance of these findings (Islam, et al., 2015).

Results

Calprotectin induces multi-metal restriction in A. baumannii

Although calprotectin is known to induce a Zn^II starvation in A. baumannii (Hood, et al., 2012), the extent to which other metal homeostasis systems are impacted by CP is not known. To address the question, a global transcriptomic analysis (RNAseq) (Figure 1) and a proteomic analysis (Figure 2) were performed on a wild-type (WT) strain grown in the presence or absence of physiologically relevant concentrations of CP (Clohessy and Golden, 1995). We first measured the metal concentration in the culture media (60% LB + CP buffer) with or without 200 μg/mL CP (Figure 1A; Figure S1). CP notably depletes the culture medium of major metals Zn and Fe, with no effect on Cu. Mn and Co are also significantly decreased by CP.

Figure 1. Transcriptomic analysis of WT and WT+CP A. baumannii.

(A) Metal content in the LB-Tris buffer growth media treated with 0 (grey) or 200 μg/mL CP (red). (B) RNAseq analysis of untreated WT vs. 200 μg/mL calprotectin treated WT (WT+CP) A. baumannii cells from 4 biological replicates. The fold-change in expression for each locus tag is indicated (see Table S1 for a complete list of genes). Gene names are indicated according to the NCBI annotation; otherwise, the locus tag (A1S_xxxx where ‘xxxx’ represents the locus tag number) is indicated. Genes regulated by the Zn uptake regulator (Zur) (yellow symbols), involved in Fe homeostasis (green symbols), involved in Cu uptake (red symbols) and associated with the histidine utilization (hut) operon (purple circles) (Nairn, et al., 2016) are highlighted. (C) qRT-PCR validation of selected genes identified by RNA-seq. A1S_0092, putative ferric siderophore uptake protein; A1S_0170, OprC outer membrane copper receptor; A1S_0242, FeoA; A1S_0652, FeoA; A1S_1214, benzoate 1,2-dioxygenase β subunit; A1S_1647, siderophore biosynthesis protein; A1S_1844, CatC3, muconolactone δ-isomerase; A1S_1868, porin for benzoate transport; A1S_2382, BasD; A1S_2580, siderophore biosynthesis protein. . *p < 0.05 as determined by Student’s t test with hypothetical value of 1. Data are the mean combined from 3 independent experiments ± S.D. See also Figure S1; Table S1.

Figure 2. LC-MS/MS proteomic analysis for WT and WT+CP A. baumannii.

Protein profiles of untreated WT vs. 200 μg/mL calprotectin treated WT (WT+CP) A. baumannii cells from 4 biological replicates. (A) Venn diagram for proteins detected at least 3 times in 4 replicates. (B) Histogram plot of the distribution of normalized abundance for all proteins detected in the untreated WT strain, as representative of all 4 growth conditions (see Figure S2A-D). The top-most 50% abundant proteins are indicated with red dash line with the bars shaded red. (C) Volcano plot for proteins detected at least 3 times for each condition among 4 biological replicates. The significance threshold was set at p≤0.001 and fold-change in protein abundance at >1.3 as shown in red dash box for proteins are up (right top) and down (left top) in WT+CP. Filled circles are shaded yellow for Zn-binding proteins, green for proteins involved in iron homeostasis, purple for Cu-binding proteins and pink for proteins involved in transcription or translation. (D) Proteins that are strongly changed in WT+CP conditions. Normalized abundance for proteins that was detected only in WT or WT+CP in all 4 replicates with the significance threshold set at the top-most 50% abundance (see Figure 2B). See also Figure S2; Table S2–3.

We next performed a transcriptomic analysis (Figure 1B; Table S1) and an LC-MS/MS, label-free proteomic analysis, in which we detected up to 1094 proteins in these cytoplasmic lysates, which corresponds to 28.9% of the A.baumannii proteome (Figure 2; Table S2; Figure S2A-D). Of the 67 genes that are significantly increased in CP-treated relative to untreated WT cells, 14 also displayed increased transcription in a strain lacking the Zn^II-responsive transcriptional regulator zur (Δzur) (Mortensen et al., 2014) (Figure 1B). Furthermore, among the most up-regulated genes are Zur-regulated targets including genes involved in cellular Zn uptake and zigA (Mortensen, et al., 2014); increased abundance of ZigA is further confirmed in the proteomic analysis (Figure 2D).

Both experiments reveal that CP induces both Zn^II and Fe starvation. Relative to Fe starvation, the RNAseq (Figure 1B) reveals a dramatically decreased transcription of genes encoding bacterioferritins, but increased transcription of genes involved in the global Fe limitation response, including feoAB, which brings Fe^II into the cell, and Fe^III-siderophore receptors, as confirmed by quantitative RT-PCR (Figure 1C). Consistently, putative achromobactin and acinetobactin siderophore biosynthesis/utilization proteins become more cell-abundant with CP stress, some at a level approaching that of ZigA (Figure 2D). The upregulation of siderophore uptake systems was further confirmed by robust siderophore production following CP exposure (Figure S2E-F). In addition, changes were observed in cellular abundances of Fe-proteins, including those involved in Fe-S cluster biogenesis (IscS and IscR), and the fumarase (FumA) (Figure 2C-D) characterized as part of the Fe-sparing response (Lee and Helmann, 2007). These results are consistent with a significant CP-mediated impact on Fe homeostasis and suggest a strong perturbation of Fe metabolism in A. baumannii.

These findings also revealed an unexpected intersection of CP-induced Zn and Fe starvation with Cu homeostasis. We observed a transcriptomic down-regulation of a putative Cu importer oprC (outer membrane porin, A1S_0170) (Yoneyama and Nakae, 1996) (Figure 1) coupled with detection of a previously unannotated protein V5VEM0, a putative copper storage protein (Csp) (Straw, et al., 2018) only in CP-treated cells (Figure 2D; Table S3). These findings suggest that Zn- and Fe-starved A. baumannii may decrease bioavailability of Cu under these CP-treatment conditions.

Cell-abundant metalloproteins induced by CP stress are established therapeutic targets

We also found 26 proteins significantly increased in normalized cellular abundance in CP-treated lysates including two known Zn^II metalloenzymes (Figure 2C), β-carbonic anhydrase, MtcA2, which may be important for survival, invasion, and pathogenicity (Aspatwar, et al., 2018) and succinyl-diaminopimelate desuccinylase, DapE, which is involved synthesis of meso-diaminopimelate (m-Dap), a component of the peptidoglycan of all Gram-negative bacteria. Fifteen proteins were detected only in CP-treated cells at high cellular levels (Figure 2D), including Zn metalloenzyme QueD, which is involved in queuosine-tRNA biosynthesis (Vinayak and Pathak, 2010), and ZigA, which is required for bacterial growth under Zn-limiting conditions and disseminated infections in mice (Nairn, et al., 2016).

A. baumannii ΔzigA is more susceptible to CP-induced metal starvation

As the candidate metallochaperone ZigA (Nairn, et al., 2016) is among the most abundant metalloproteins that is detected only in CP-treated WT cells (vide infra), investigation of how the ΔzigA strain adapts to CP stress relative to the WT strain may identify crippled ZigA-dependent processes. The addition of CP gave rise to a significant decrease in growth yield of exponentially growing cells (t=4 h), but only in the ΔzigA strain (Figure 3A; Figure S3A-B). CP induces a measurable decrease in total cell-associated Zn and Fe, but only in the WT strain (Figure 3B), a finding consistent with transcriptomic and proteomic analysis (Figures 1–2). In addition, the ΔzigA strain exhibits increased expression of the Zn uptake genes (Nairn, et al., 2016), which may account for these findings. Total cell-associated Cu is also lower only in the WT strain (Figure 3C), consistent with the down regulation of oprC (Figure 1). The ΔzigA strain, in contrast, takes up slightly more Mn and Ni when treated with CP (Figure 3C). These CP-induced trends in growth yield and in cellular metal content largely persist at 6 and 8 h of growth (Figure S3C-F). These results reveal that CP restricts Zn and Fe availability and the loss of ZigA has little further impact on total cell-associated metal, but with a perturbation in metal allocation to cellular targets. This is representative of “adaptive” physiology, since metallation can increase the stability of metalloproteins (Figure 3D).

Figure 3. Calprotectin (CP) induces metal starvation in A. baumannii.

Growth yield at 4 h (OD 600 nm) (A) of cell cultures treated with 0 or 200 μg/mL CP when cell samples were collected for determination of total cell-associated metal by ICP-MS (panels B, C). Total cell-associated metal of WT and ΔzigA cells are shown in (B) for zinc (Zn) and iron (Fe) and (C) for other transition metals. The results shown reflect the mean of 3 independent replicates ± S.D. *p≤0.05 as determined by Student’s t test. (D) Rationale and design of this study using ΔzigA as a tool to study the effects of extreme metal limitation. See also Figure S3.

As an effort to monitor a perturbation in metal allocation, we found the proteomic profiles of unstressed WT and ΔzigA cells are extremely similar but show marked differences in the presence of CP (Figure 4A-C; Table S4; Figure S4A-B). In order to understand these differences, a bioinformatics approach was employed to predict the entire Zn proteome of A. baumannii (Table S5). We predicted 213 Zn-proteins corresponding to ≈5% of the proteome, similar to predictions for E. coli (Andreini, et al., 2006), and these proteins are anticipated to impact a considerable range of cellular metabolism (Figure S4C), with 84 high-probability Zn-proteins were detected in at least three replicates in the proteomic analysis (Figure S4D). Overall, zinc-proteins are modestly more abundant (by protein count) in the WT relative to the ΔzigA strain in the presence of CP, including MtcA2 (Figure 4B) and the global Fe uptake regulator (Fur) (Figure S3E), with the striking exception of tRNA-guanosine transglycosylase (Tgt), involved in queuosine-tRNA biosynthesis (Vinayak and Pathak, 2010), which is detected only in CP-treated ΔzigA cells at high abundance (Figure 4C; Table S4)

Figure 4. LC-MS/MS proteomic analysis for WT and ΔzigA A. baumannii.

Protein profiles of WT (WT+CP) vs. ΔzigA (ΔzigA+CP) A. baumannii cells treated with 200 μg/mL calprotectin from 4 biological replicates. (A) Venn diagram and (B) Volcano plot for proteins detected at least 3 times under each condition in 4 replicates. The significance threshold was set at p≤0.001 and fold-change in protein abundance at >1.3 as shown in red dash box. Yellow, Zn-binding proteins. (C) Proteins that are strongly changed in WT+CP and ΔzigA+CP cells. Normalized protein abundance for proteins that were only detected in WT+CP or ΔzigA+CP for in all 4 replicates with the significance threshold was set at the top-most 50% abundance (see Figure S2A-D). (D) Normalized protein abundances of enzymes of the flavin biosynthetic (rib) pathway (compare to Figure S4F) in the WT (grey) or WT+CP (red) cells. * p≤0.05 as determined by Student’s t test. Except for the RibF, which was only detected twice out of 4 replicates in WT, all proteins shown were detected in at least 3 of the 4 biological replicates. The mean of independent replicates ± S.D. is shown. (E) Riboflavin biosynthesis pathway in A. baumannii. Metabolites are indicated in bold and Zn^II metalloenzymes are highlighted in blue. See also Figure S4; Table S4–5.

A dual regulatory model of flavin biosynthesis in A. baumannii in response to CP

We found RibB is detected only in CP-stressed ΔzigA but is undetectable in the WT strain in both conditions (Figure 4C-D). Like many bacteria, A. baumannii synthesizes riboflavin, flavin mononucleotide (FMN), and flavin adenine dinucleotide (FAD) from guanosine 5’-triphosphate (GTP) via the combined actions of the Zn^II-dependent GTP cyclohydrolase II (GCHII), encoded by ribA, and from ribulose-5’-phosphate (Ru5P), a product of the pentose phosphate pathway, via the activity of 3,4-dihydroxy-2-butanone 4-phosphate synthase (DHBPS), encoded by ribB (Figure 4E). In addition to an impact of CP of Zn-protein abundance (vide infra), while RibA abundance is unchanged by CP treatment, an uncharacterized RibB-domain containing fusion protein that we designate RibBX is more abundant in the WT strain in the presence of CP (p≤0.05) (Figure 4D and Figure S4F), indicating the two RibB- enzymes associated with riboflavin biosynthesis (rib), also play a role in the response to CP.

The increased cell abundance of RibBX induced by CP stress implies a functional role that is compensatory to RibB under CP stress (Figure 4E), although the function of the RibBX is unknown. This is particularly true for the C-terminal RibX domain that is homologous to RibA, but appears to lack Zn^II coordinating residues of authentic RibA. We measured the DHBPS and GCHII activity of RibBX (Figure 5) and find that RibBX possesses robust DHBPS activity that is comparable to that of authentic DHBPS (RibB) (Figure 5A), but is completely devoid of GCHII activity, consistent with the loss of all three cysteine ligands to the Zn (Singh, et al., 2013) (Figure 5B). We then noted that purified RibBX solution is yellow, which suggests the presence of a bound flavin (Figure S5A); indeed, LC-MS/MS analysis reveals sub-stoichiometric FMN (Figure S5B). We then used FMN-stripped RibBX to show that FMN binds reversibly to RibBX, with a K_a of 1.4±0.5 μM⁻¹ and a subunit binding stoichiometry near 1:1 by isothermal titration calorimetry (Figure 5C). We then reasoned that FMN might regulate the DHBPS activity of RibBX because in some organisms FMN regulates flavin biosynthesis through a canonical FMN-sensing riboswitch (Baird, et al., 2010), which is predicted to fold upstream of ribB in A. baumannii, but not upstream of ribBX. To test this hypothesis, we measured the DHBPS activity of both RibBX and RibB in the presence of increasing FMN (Figure 5D; Figure S5C), and only RibBX is inhibited by FMN but with a K_i value of 7.8 μM (comparable to K_a). We note that this K_i is 3–4 orders of magnitude higher than observed for FMN riboswitch-mediated regulation of ribB in other bacteria.

Figure 5. Biochemical characterization of RibBX.

(A) Michaelis-Menten plot for DHBPS activity with various substrate Ru5P concentrations. Error bars represent S.D. from 3 replicates. RibBX is in red. RibB (blank) is the authentic DHBPS in A. baumannii. (B) Michaelis-Menten plot for GCHII activity with various substrate GTP concentrations. Error bars represent S.D. from 3 replicates. RibBX is in red. RibA (blank) is the authentic GCHII in A. baumannii. (C) ITC titration of FMN to RibBX. The panel is shown as a representative fitting of 3 replicates. (D) Inhibition of DHBPS activity with 200 μM FMN. Reaction rate is normalized to the value with 0 μM FMN. * p≤0.05 using Student’s t-test. The mean of 3 independent replicates ± S.D. is shown. See also Figure S5.

To gain additional insight into the FMN-mediated inhibition of RibBX activity, we determined its structure by x-ray crystallography using molecular replacement with Mtb RibBA (Singh, et al., 2013) (see Table S6 for structure statistics). A. baumannii RibBX crystallizes as a pseudo-D₂ symmetric homotetramer (Figure S5F). In solution, however, RibBX is a dimer and the oligomerization state is not significantly affected by the presence of bound FMN (Figure S5D). As anticipated, the overall architecture of RibBX is similar to that of Mtb RibBA, with a canonical RibB-domain (Islam, et al., 2015) (Figure 6A). The RibX-domain adopts a topology highly similar to the GCHII (RibA) domain of Mtb RibBA, but with no Zn^II site present (Figure 6B). We hypothesized that FMN binds in RibX-domain and allosterically regulates RibB-domain activity (see Figure S5G). These structural and functional findings suggest a dual regulatory model of flavin biosynthesis that is employed by cells experiencing transition metal perturbation (Figure 6C).

Figure 6. Crystal structure of A. baumannii RibBX.

(A) Superposition of structure of RibBX (bright orange and pale yellow) and Vibrio cholerae RibB (Islam et al., 2015) (VcRibB) in the D-ribulose 5-phosphate-bound form (blue). Substrate binding residues are indicated for RibBX (bold). (B) Superposition of structure of RibBX (pale cyan and pale yellow) and Mycobacterium tuberculosis RibBA (Singh, et al., 2013) (MtRibBA; Rv1415) in the apo form (magenta). Zinc (Zn) binding residues in MtRibBA (stick representation). (C) Proposed dual regulatory model of the riboflavin biosynthesis pathway that becomes operative under extreme metal limitation mediated by CP. See also Figure S5; Table S6.

Flavin supplementation complements the ΔzigA growth phenotype

We hypothesized in the dual regulatory model that the cellular flavin supply is tightly regulated by balancing the activities and cell abundance of RibA, RibB and RibBX. To elucidate the importance of ribBX, we found that a ΔribBX strain has a significant growth defect during Zn starvation (Figure 7A). Given the differential abundance of RibBX in the WT and ΔzigA strains, the contribution of RibBX to the A. baumannii response to Zn limitation was interrogated in a system uncoupled from any transcriptional regulation that may exist for ribBX. ribBX was expressed under a constitutive promoter in WT and ΔzigA A. baumannii, and these strains were grown in Zn limitation induced by TPEN. Constitutive expression of ribBX did not alter growth of WT or ΔzigA in Zn-replete conditions. However, in Zn-starved conditions, the ΔzigA strain experienced growth inhibition upon expression of ribBX whereas ribBX expression improved growth in WT A. baumannii during Zn limitation (Figure 7B; Figure S5E). These data revealed that RibBX helps overcome nutrient Zn limitation in A. baumannii.

Figure 7. Riboflavin can rescue the growth of ΔzigA.

(A) Growth of ΔribBX and ΔzigA is impaired with 40 μM TPENThe mean of at least 3 independent replicates ± S.D. is shown. (B) Constitutive expression of ribBX in the presence of 20 μM TPEN improves growth of WT A. baumannii but impairs growth of ΔzigA. (C) Cellular FAD levels are notably decreased only in the ΔzigA strain in the presence of CP. (D) Riboflavin partially rescues the growth phenotype of ΔzigA with fumarate as sole carbon source in Zn-deplete conditions. See text for additional details. * p ≤ 0.05, * p ≤ 0.01 and **** p < 0.0001 as determined by one-way ANOVA with Tukey multiple comparisons test. The mean of at least 3 independent replicates ± S.D. is shown. See also Figure S5–7.

Cellular flavin levels are markedly lower in ΔzigA under CP stress

The proteomics data suggest an impact on cellular flavin levels mediated by CP stress and/or ZigA (Figure 4D) as summarized in our dual regulatory model (Figure 6C). To test this, we quantified FAD levels as a proxy of flux through the riboflavin biosynthesis pathway. The ΔzigA strain has notably reduced total cellular FAD levels, but only under CP stress (Figure 7C), a finding consistent with the detection of RibB only in these cells, whose levels are predicted to be strongly repressed by a canonical FMN-sensing riboswitch under unstressed conditions (Serganov, et al., 2009) (Figure 4C). Both CP-stressed (Figure 7C) and TPEN-treated (Figure S6A-B) ΔzigA cells show reduced flavin supply and induce a growth yield phenotype (Figure 3A; Figure S5E); indeed, the ΔribBX strain has lower levels of FAD even in Zn-replete conditions (Figure S6A).

Exogenous flavin partially rescues the growth deficiency of ΔzigA under CP stress

Our dual regulatory model predicts that supplementation of the ΔzigA strain with riboflavin might rescue this growth phenotype observed during CP- or TPEN-mediated metal limitation. To test this prediction, we employed a minimal growth medium with a carbon source(s) that would increase the cellular demand for both flavins and Fe, and supplement with riboflavin, as riboflavin is the precursor to FMN and FAD (Figure 4E) (Nwugo, et al., 2011). We used succinate and fumarate, two TCA cycle intermediates, for this purpose. Succinate dehydrogenase is a 4Fe-4S, flavin and heme-requiring enzyme that interconverts succinate and fumarate, while fumarase (FumA) is a 4Fe-4S cluster-containing enzyme, that converts fumarate to malate and whose cellular levels are decreased in CP-stressed cells (Figure 2C). Both WT and the ΔzigA strain have similar growth kinetics in Zn-replete minimal media with either succinate or fumarate as the sole carbon source but not riboflavin in Zn-replete conditions (Figure S6C-E). Strikingly, in Zn-deplete conditions, the ΔzigA strain has a dramatic growth defect, which can be partially rescued by supplementation with low levels of riboflavin (Figure 7D; Figure S6F-G), but not by folate, the biosynthesis of which is also initiated from a Zn enzyme, GTP cyclohydrolase I (FolE) (Figure S7A-B). Neither WT or ΔzigA A. baumannii is able to utilize riboflavin as a carbon source, revealing that this growth rescue is not from additional carbon source supplementation (Figure S6C). Growth of ΔzigA and the effect of riboflavin supplementation under Zn-deplete conditions is not observed in a rich growth medium, revealing that increasing the cellular Fe requirement via the use of fumarate or succinate as the sole carbon source also exacerbates the demand for flavins (Figure S7C-D). Consistent with this prediction, both the ΔzigA and ΔribBX strains have significant growth defects compared to the WT strain when succinate is provided as the sole carbon source and Fe is restricted by the Fe chelator EDDHA (Figure S7E-F).

Discussion

The findings presented here reveal a CP-induced multi-metal (Fe and Zn) perturbation of transition metal bioavailability that is exacerbated by the loss of ZigA in a way that impacts central metabolic processes in an important human pathogen. These processes include de novo flavin biosynthesis, the product of which functions as an essential cofactor involved in many aspects of cellular redox and energy metabolism (Bacher, et al., 2000), and potentially other Zn-requiring enzymes including the β-carbonic anhydrase (MtcA2) (Supuran and Capasso, 2017) and enzymes (QueD and Tgt) of the queuosine-tRNA biosynthesis pathway (McCarty and Bandarian, 2012). We describe a dual regulation model of flavin biosynthesis that is orchestrated by the DHBPS RibB and a unique, structurally characterized RibBX which functions in an effort to maintain cellular flavin sufficiency under conditions of CP-induced metal starvation and the loss of ZigA. The fact that RibBX homologs are widespread in Proteobacteria (Brutinel, et al., 2013) underscores the importance of this dual regulation model of flavin biosynthesis in other bacteria.

It is now well established that CP is a broad-spectrum chelator capable of inducing microenvironmental niche- and pathogen-dependent transition metal starvation (Zygiel, et al., 2019; Zygiel and Nolan, 2018). The niche-dependence is likely due to the spectrum of total metals defined by the site of infection (Cassat, et al., 2018). However, it is often not clear how much of this total metal is bioavailable at infection sites, since the metal speciation in complex environments is not typically known (Juttukonda, et al., 2017). As proxy for this, and from the perspective of comparing data obtained under different culture conditions, we have surveyed a sampling of typical growth media and found that the total metal composition and the ability of CP to deplete metals from these media varies dramatically (Figure S1). We conclude that although CP is clearly capable of depleting divalent metals, what metals CP depletes from the growth medium appears strongly influenced by the relative concentrations of total metal and total CP-binding sites. The coupling between the niche-dependent metal depletion by CP and intrinsic metal requirements in different bacteria (Lisher and Giedroc, 2013) potentially unifies previously conflicting responses to CP in various organisms.

Here, using a standard growth medium, we confirmed an adaptive response to Zn starvation regulated by Zur (Mortensen, et al., 2014), but also revealed a strong Fe limitation response, in that siderophore biosynthesis and utilization genes are found to be notably upregulated at both the transcriptome and proteome levels in A. baumannii. These findings are fully consistent with our metal analysis (Figure S1), and the previous observation that CP reduces Fe uptake in multiple bacteria (Zygiel, et al., 2019). On the one hand, Zn^II starvation may affect the functional integrity of the global Fe regulator Fur, since Fur contains a structural Zn^II site. On the other hand, CP-induced a Zur-regulatd putative stand-alone siderophore receptor protein (A1S_0092), which may have a unique role in transition metal acquisition and/or represent regulatory overlap between Zur and Fe limitation. Our findings also provide further evidence that transition metal homeostasis systems are interconnected (Hassan, et al., 2017). Genes detected in RNAseq involved in Fe metabolism are not differentially expressed in the Δzur strain, demonstrating that their upregulation is mediated through a Zur-independent mechanism (Mortensen, et al., 2014). The detection of a candidate cell-abundant Csp3 suggests that cytoplasmic Csp3 traps Cu^I (Dennison, et al., 2018) and thus potentially allows A. baumannii to cope with the restricted bioavailability of Zn and Fe (Figure 3D); further, A. baumannii have an as-yet unknown nutritional need for Cu.

ZigA is a key player in response to Zn starvation and infection in mice, but how ZigA functions at the molecular level is unclear. ZigA exhibits weak GTPase activity and binds Zn^II with high affinity, and is therefore proposed to be a Zn metallochaperone from the COG0523 family (Nairn, et al., 2016). As a result, the loss of ZigA may exacerbate cellular metal deficiency due to a failure to mobilize or allocate cellular Zn from metal complexes to high priority protein targets that emerge under conditions of extreme metal limitation, with little or no impact on total metal. Regardless of the mechanism, we employed the ΔzigA strain as a tool to assess changes in the proteome that could be reporting on perturbations in metabolism that result from the loss of ZigA under these Zn/Fe-starved conditions. We identified two cellular processes, de novo flavin biosynthesis and queuosine-tRNA biosynthesis, that appear strongly impacted by extreme CP-induced metal depletion.

It is known that in bacteria and plants, flavodoxins are required as non-metal substitutes for ferredoxins, thus prioritizing (ribo)flavin biosynthesis under these conditions (Pi and Helmann, 2017; Sepulveda Cisternas, et al., 2018; Tognetti, et al., 2007). In H. pylori, riboflavin is reported to be secreted from cells and is thought to participate in Fe reduction to Fe^II and acquisition potentially through FeoAB (Figure 1B), with the flavin biosynthesis pathway found to be upregulated under Fe deficiency (Worst, et al., 1998); this is true in other bacteria as well (da Silva Neto, et al., 2013; Vasileva, et al., 2012). Recent work reveals that riboflavin and Fe levels are reciprocally regulated in Vibrio cholerae (Sepulveda-Cisternas, et al., 2018), and we observe elements of this in A. baumannii, since the ΔribBX strain, like the ΔzigA strain, is sensitive to Fe starvation (Figure S7E-F). Although in B. subtilis, ribBA expression is reported to be Fur-regulated (Worst, et al., 1998), it is unknown how the transcription of the ribBX operon is regulated in A. baumannii. We showed here that while RibBX has WT-like DHBPS activity under conditions where authentic RibB is not detected in cells, RibBX is subjected to product inhibition by cellular FMN, but only at very high concentrations, and thus can be used in CP-treated cells to bypass the regulatory control of the FMN-sensing riboswitch upstream of ribB under these conditions (Figure 6C) (Serganov, et al., 2009). We also found that two enzymes, QueD and Tgt, also become more cell-abundant in CP-stressed cells thus suggesting that queuosine-tRNA biosynthesis, a second pathway downstream of FolE, is prioritized under these conditions. The cellular logic for this is unknown but strongly implicates this tRNA modification as involved in response to CP-induced metal starvation in A. baumannii.

In summary, we show here that CP treatment of A. baumannii under these growth conditions strongly restricts the availability of Zn and Fe. The bacteria respond by inducing an acute cellular response to Zn and Fe deficiency while limiting the bioavailability of Cu. CP stress impacts flavin biosynthesis, and it is plausible that Fe homeostasis and flavin biosynthesis are reciprocally regulated in A. baumannii. Taken together, these results outline significant crosstalk between metal homeostasis systems and a key intersection with flavin, and perhaps queuosine, biosynthetic pathways. These results also identify cell-abundant antimicrobial targets, including MtcA2 (Aspatwar, et al., 2018; Supuran and Capasso, 2017), DapE (Gillner, et al., 2013; Starus, et al., 2015), ZigA and RibBX, for the development of antibacterial strategies for the treatment of A. baumannii infection while this multi-drug resistant organism is actively subjected to nutritional immunity in the infected host.

Star Methods

Contact for Reagent and Resource Sharing

Further information and requests for materials should be directed to and fulfilled by the Lead Contacts, David Giedroc (giedroc@indiana.edu) and Eric Skaar (eric.skaar@vanderbilt.edu).

Experimental Model and Subject Details

Bacterial strains and reagents

Experiments were performed on A. baumannii ATCC 17978 and mutant derivatives of this strain, including the ΔzigA mutant generated previously (Nairn, et al., 2016) and the ΔribBX mutant generated as described below. Cloning was performed in E. coli DH5α and protein expression in E. coli BL21 (DE3). Strains were cultured in Luria Broth (LB) at 37 °C with aeration unless otherwise noted, and OD₆₀₀ was used to measure bacterial growth as previously described (Nairn, et al., 2016). Kanamycin (Sigma) was used at 40 μg/mL. Ampicillin (Sigma) was used at 100 μg/mL for E. coli and 500 μg/mL for A. baumannii. Recombinant human CP were expressed and purified as previously described (Corbin, et al., 2008). S100A8 and S100A9 cloned into pET1120 were transformed separately into E. coli C41 (DE3) cells for expression. Cells were induced at 37 °C at OD₆₀₀ between 0.6 and 0.8 with the addition of 1 mM isopropyl thio-β-D-galactoside (IPTG) and allowed to grow 4 hours post-induction. Cells were harvested by centrifugation (6500 rpm, 20 min, 4 °C), and resuspended in Lysis Buffer (50 mM Tris pH 8.0, 100 mM NaCl, 1 mM EDTA, 1 mM phenylmethane sulfonyl fluoride (PMSF), 0.5 % Triton X-100). Once homogenized, sample was sonicated (10 min, 50 watts, 5 s on/10 s off), and centrifuged (20,000 rpm, 20 min). The supernatant was filtered and loaded onto a SourceQ column (General Electric) with flow rate at 1 mL/min. After loading, the column was washed with three column volume (CV) Buffer A (20 mM Tris, pH 8.0) and eluted with a gradient from 0 to 100% Buffer B (20 mM Tris, pH 8.0, 1 M NaCl) over 10 CV. S100A8/A9 containing fractions were pooled according to SDS-PAGE, concentrated, and loaded onto S75 column (General Electric). The column was washed with 1 CV S75 Buffer (20 mM Tris, pH 8.0, 100 mM NaCl), S100A8/A9 containing fractions pooled according to SDS-PAGE, flash frozen, and stored at –80°C.

Methods Details

A. baumannii mutant generation

To generate in-frame the ΔribBX strain, approximately 1000 base pairs of DNA in both the 5’ and 3’ flanking regions surrounding the open reading frame, leaving the N-terminal ~200 nucleotides of ribBX intact to limit polarity, were amplified from A. baumannii genomic DNA using primers listed in Supplemental Table S7, and the kanamycin resistance gene aph was amplified by PCR from the pUCK1 plasmid using primers listed in Supplemental Table S7 (Menard, et al., 1993). The 3 PCR products and pFLP2 vector were assembled together using Gibson assembly (New England Biolabs) and sequence verified (Hoang, et al., 1998). pFLP2 was then electroporated into A. baumannii, plated onto LB Km 40 agar, and grown overnight at 37 °C. Transformants were patched to LB Km40 or LB agar with 10 % sucrose to isolate Km^R and sucrose^S merodiploids. Merodiploid strains were grown in LB supplemented with 10 μg/mL riboflavin overnight at 37 °C to resolve the plasmid. Cultures were serially diluted, plated on LB agar with 10 % sucrose, and incubated at 37 °C overnight. The resulting Km^R sucrose^R strains were screened for loss of ribBX and replacement with aph by multiple PCR reactions.

A. baumannii expression construct generation

The ribBX gene was amplified from WT A. baumannii genomic DNA using primers listed in Supplemental Table S7 and incorporated a C-terminal cMyc tag. The amplified gene product was ligated into digested pWH1266 containing the constitutive r01 promoter, which was generated as previously described (Mortensen, et al., 2014): the 16s gene r01 was amplified from A. baumannii genomic DNA and ligated into the pWH1266 expression vector. The integrity of the r01 promoter and ribBX was confirmed by sequencing. The empty vector and the ribBX expression vector were transformed into WT and ΔzigA by electroporation.

Growth in calprotectin or metal chelators

Overnight A. baumannii cultures started from independent colonies as replicates were subcultured 1:50 in LB for 1 h. Back-diluted cultures were then inoculated 1:100 into LB containing various concentrations of tetrakis-(2-pyridylmethyl)ethylenediamine (TPEN) as indicated, and growth was monitored over time. For growth in medium containing recombinant human WT CP, back-diluted cultures were inoculated 1:100 into LB containing 40% CP buffer (20 mM Tris-HCl pH 7.5, 100 mM NaCl, 5 mM β-mercaptoethanol, 3 mM CaCl₂) supplemented with CP, and growth was monitored over time. To assess growth during Fe limitation, back-diluted cultures were inoculated 1:100 into M9 minimal media supplemented with 0.5% sodium succinate dibasic hexahydrate and 10 μM ethylenediamine-di(o-hydroxyphenylacetic acid) (EDDHA, LGC Standards), and growth was monitored over time.

ICP-MS quantification of total metals

The quantification of total metal content of the growth media used here (LB + 40 % CP buffer) and total cell-associated metal was carried out largely as described before on a Perkin-Elmer Elan DRCII ICP-MS (Nairn, et al., 2016). Briefly, the growth media was incubated with 200 μg/mL calprotectin (CP) at 37 °C for 2 h. CP was then removed via filtration with a pre-washed and pre-equilibrated 3 or 10 kDa filter with the flowthrough eluent analyzed for total metal content and compared with the same media prior to filtration, essentially as described (Nakashige, et al., 2015). The CP used for these studies contained 0.11 protomer mol•equiv Zn, 0.08 protomer mol•equiv Cu, 0.01 protomer mol•equiv Mn, 0.03 protomer mol•equiv Ni and 0.04 protomer mol•equiv Fe and is thus 85% metal-free (2 metal sites per protomer); quantitation of the total metal chelated by CP in replicate experiments was found to be 95 ±7% of expected value based on the concentration of CP-binding sites (site 1 + site 2, per protomer). For bacterial lysates, strains were grown in the presence of the 200 μg/mL CP as described above. Cells were normalized to the same OD₆₀₀ and pelleted at 4 °C, then washed 3 times with ice-cold, chelexed PBS buffer. Cell pellets were stored at –80 °C before measurements. Samples were incubated with 30% nitric acid at 95° for 10 min and 65 °C for 1 h. The volumes were then adjusted to 3 mL with 2.5% nitric acid and subjected to elemental quantification on a Perkin-Elmer Elan DRCII ICP-MS equipped with AS-93 autosampler run alongside elemental standards. Cellular metal content was normalized to protein amount which was measured using a standard Bradford assay, as described previously (Jacobsen, et al., 2011).

RNA Sequencing

WT A. baumannii was grown overnight in 3 mL LB. Cultures were reseeded 1:50 in LB and grown for 1 h at 37 °C with shaking. These cultures were reseeded 1:100 in 40 % LB, 60 % CP buffer plus or minus 250 μg/ml CP and grown for 7 h at 37 °C with shaking. Cultures were pelleted at 4 °C with 2,500 × g for 8 min and then air dried on ice. Each pellet was suspended in 1 mL TRIzol and transferred to tubes containing lysing matrix B (MP Biomedicals). Bacteria were lysed using a FastPrep-24 (MP Biomedicals) bead beater for 45 s at 6 m/s. Two hundred μL chloroform then was added to each tube. After brief vortexing, the tubes were centrifuged for 15 min at 4 °C and the upper layer was transferred to another new tube. RNA was purified using the RNeasy preparation kit (Qiagen) according to the RNeasy lipid tissue directions. DNA contamination was removed by adding 8 μL RQ1 (Promega), 12 μL 10x RQ1 buffer, and 2 μL RNase inhibitor (Promega) to each sample and incubating at 37 °C for 2 h. Samples were then cleaned up using the RNeasy miniprep RNA cleanup protocol (Qiagen). To ensure purity of the RNA from DNA contamination, an aliquot of RNA was removed for reverse transcription (RT), including no-RT controls, and assessed by PCR. Prior to sequencing, RNA was quantified on the Synergy 2 with Gen5 2.00 software (BioTek).

Vanderbilt Technologies for Advanced Genomics Core Facility (VANTAGE) prepared RNA-seq libraries from 1.5 μg of A. baumannii total RNA using the following protocol. The integrity of the total RNA was evaluated using the Agilent Bioanalyzer Nano RNA chip. The Ribo-Zero rRNA removal kit for Gram-negative bacteria (Epicentre) was used to remove rRNA by following the manufacturer’s protocol. The rRNA-reduced RNA was used as an input to the TruSeq stranded mRNA sample preparation kit (Illumina), skipping the mRNA selection step and going directly into the RNA fragmentation and random hexamer priming step. The RNA was converted into double-stranded cDNA, adapter ligated, and enriched with PCR, replacing the enzyme from the kit with KAPA HiFi DNA polymerase to create the final cDNA sequencing library. The cDNA library underwent quality control (QC) by running on an Agilent Bioanalyzer HS DNA assay to confirm the final library size and on an Agilent Mx3005P quantitative PCR (qPCR) machine using the KAPA Illumina library quantification kit to determine concentration. A 2 nM stock was created, and samples were pooled by molarity for multiplexing. From the pool, a 10.5 pM concentration was loaded into each well for the flow cell on the Illumina cBot for cluster generation. The flow cell was then loaded onto an Illumina HiSeq 2500, utilizing v3 chemistry and HTA 1.8 for a paired-end 50-bp run. The raw sequencing reads in BCL format were processed through CASAVA-1.8.2 for FASTQ conversion and demultiplexing. The RTA chastity filter was used, and only the PF (pass filter) reads were retained for further analysis.

Transcriptomic analysis

The Illumina HiSeq 2500-generated FASTQ reads were processed by using the Bayesian adapter trimmer Scythe (version 0.992; http://github.com/vsbuffalo/scythe) to trim 3= adaptor sequence contaminants from the reads. EDGE-pro v1.3 (Estimated Degree of Gene Expression in Prokaryotes) software (Magoc, et al., 2013) then was used to align the reads with Bowtie2 v2.1.0 (Langmead and Salzberg, 2012) and estimate gene expression directly from the alignment output. The FASTA of the reference genome sequence (.fna), protein translation table with coordinates of protein coding genes (.ptt), and a table containing coordinates of tRNA and rRNA genes were downloaded from NCBI (ftp://ftp.ncbi.nih.gov/GenBank/genomes/Bacteria/Acinetobacter_baumannii_ATCC_17978_uid17477). The .fna, .ptt, and .rnt files were concatenated from the main genome and two native plasmids into single input files, and they were used as inputs into EDGE-pro. EDGE-pro was run with default parameters, except for defining the read length as 50 bp (-l 50) and using 16 threads (-t 16) on a 64-core Linux server. On average, more than 96% of the reads were uniquely aligned. Alignment statistics and other QC metrics were calculated from the aligned BAM file using the RSeQC program suite (Wang, et al., 2012). The output of EDGE-pro was an RPKM value table for the genome and plasmids. A script provided with EDGE-pro called edgeToDeseq.perl was used to concatenate the output of the various samples into a single count table for input into DESeq. The expression level of each gene was determined using DESeq (Anders and Huber, 2010) in the statistical programming package R-3.0.0 (http://www.r-project.org). Differences in expression comparing untreated WT A. baumannii to CP-treated A. baumannii were considered significant based on a P value less than or equal to 0.01.

Quantitative RT-PCR

Overnight cultures of A. baumannii were diluted 1:50 in LB and grown for 1 h at 37°C. Cultures were then diluted 1:100 into LB plus or minus 250 μg/ml CP and grown for 8 h. Cultures were pelleted at 4°C and resuspended in 1:1 acetone : ethanol prior to storage at −80°C until processing. For RNA extraction, cells were pelleted and resuspended in LETS buffer and lysed using Lysis Matrix B tubes (MP Biologicals) and a FastPrep-24 (MP) bead beater. Samples were heated to 55°C for 5 min and pelleted at 15,000 rpm for 10 min. The top phase was combined with TRIzol and incubated at room temperature for 5 min. Chloroform was mixed with each sample, incubated for 3 min, and centrifuged for 15 min at 4°C at 15,000 rpm. Following centrifugation, the upper aqueous phase was transferred to a new tube and incubated with isopropyl alcohol for 10 min at room temperature to precipitate the RNA. Samples were centrifuged at 4°C for 10 min at 15,000 rpm. Supernatant was removed, and the pellet was washed twice with 70% ethanol and dissolved in water. DNA contamination was removed by adding RQ1 and RQ1 buffer (Promega), and RNAse inhibitor (Thermo), and the samples were incubated at 37°C for 2 h. Following DNAse treatment, RNA was purified using RNease mini kit (Qiagen) following manufacturer’s recommendation. RNA was quantified, and 2 μg of RNA was used for cDNA synthesis. cDNA synthesis and qRT-PCR was performed as previously described using the 2^ΔΔCT method (Mortensen, et al., 2014; Nairn, et al., 2016). Primers used for qRT-PCR are listed in Supplemental Table S7.

LC-MS/MS proteomic analysis

Protein samples were denatured in 8 M urea, 100 mM ammonium bicarbonate solution, then incubated for 45 min at 56 °C with 10 mM dithiothreitol (DTT) to reduce cysteine residues. The free cysteine residue side chains were then alkylated with 40 mM iodoacetamide for 1 h in the dark at room temperature. The solution was diluted to 1 M urea and 1:100 (w:w) ratio of trypsin was added and the samples were digested at 37 °C overnight. Peptides were desalted by Zip-tip and injected into an Easy-nLC HPLC system coupled to an Orbitrap Fusion Lumos mass spectrometer (Thermo Scientific, Bremen, Germany). Peptide samples were loaded onto a home-made C18 trap column (75 μm × 20 mm, 3 μm) in 0.1% formic acid. The peptides were separated using an Acclaim PepMapTM RSLC C18 analytical column (75 μm × 150 mm, 2 μm) using an acetonitrile-based gradient (Solvent A: 0% acetonitrile, 0.1% formic acid; Solvent B: 80% acetonitrile, 0.1% formic acid) at a flow rate of 400 nL/min. A 30 min gradient was implemented as follows: 0–0.5 min, 2–9% B; 0.5–24 min, 9–26% B; 24–26 min, 26–100% B; 26–30 min, 100% B followed by re-equilibration to 2% B. The electrospray ionization was carried out with a nanoESI source at a 260 °C capillary temperature and 1.8 kV spray voltage. The mass spectrometer was operated in data-dependent acquisition mode with mass range 400 to 1600 m/z. The precursor ions were selected for tandem mass (MS/MS) analysis in Orbitrap with 3 sec cycle time using HCD at 35% collision energy. Intensity threshold was set at 5e3. The dynamic exclusion was set with a repeat count of 1 and exclusion duration of 30 s. The resulting data were searched against an Acinetobacter baumannii database (Uniprot UP000094982, 3,780 entries, with the database downloaded on 07/06/2017 from Uniprot) in Proteome Discoverer 2.1. Carbamidomethylation of cysteine residues was set as a fixed modification. Protein N-terminal acetylation, oxidation of methionine, protein N-terminal methionine loss, protein N-terminal methionine loss and acetylation, and pyroglutamine formation were set as variable modifications. A total of 3 variable modifications were allowed. Trypsin digestion specificity with two missed cleavage was allowed. The mass tolerance for precursor and fragment ions was set to 10 ppm and 0.6 Da respectively.

Statistical rationale and bioinformatics analysis

Data analysis was performed as previously described (Kentache, et al., 2017; Park, et al., 2014). Proteins detected fewer than 3 times in 4 replicate lysates were excluded from our statistical analysis. The statistical analysis of these data was completed using unpaired, two-tailed student’s t test with Welch’s correction. Functional information for the selected proteins were gathered from the National Center for Biotechnology Information (NCBI, https://www.ncbi.nlm.nih.gov/) and the SEED database (http://pubseed.theseed.org/?page=Minimal) (Overbeek, et al., 2005). Metabolism pathway information for the selected proteins were obtained from the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (https://www.genome.jp/kegg/pathway.html).

Siderophore quantification

Siderophore production was quantified using a Chrome Azurol S (CAS) assay as described previously (Alexander and Zuberer, 1991; Schwyn and Neilands, 1987). Briefly, A. baumannii was grown for 36 h in the presence of 250 μg/mL CP or buffer as described above. Supernatants were collected and filtered through a 0.22 μm filter (Millipore). Supernatants were mixed 1:1 with the chrome azurol S shuttle solution(Alexander and Zuberer, 1991), and OD₆₃₀ was determined after a 30 min incubation at room temperature. Values were normalized to starting cell density as determined by OD₆₀₀.

Zinc Proteome prediction of A. baumannii

Using a previously published approach (Valasatava, et al., 2016), we created two libraries of Hidden Markov Model profiles (Eddy, 1998): a library of Zn-binding Pfam domains (Finn, et al., 2014), and a library of Zn-binding structural motifs. The Pfam domain library was created by merging two lists: 1). a list of Pfam domains with known 3-dimensional structure that contain a Zn-binding site extracted from MetalPDB (Putignano, et al., 2018) in which each of these domains could be associated with the residues responsible for Zn-binding and with their positions within the domain sequence and 2). a list of Pfam domains without a known 3D structure but annotated as Zn-binding, obtained by text mining of the annotations in the Pfam database. The procedure resulted in a set of 573 Pfam domains: 541 with an associated Zn-containing 3D structure, and an additional 32 annotated as Zn-binding domains. The library of Zn-binding structural motifs was created by splitting into fragments the Zn-binding sites stored in MetalPDB as of June 2017, as previously described (Rosato, et al., 2016). Only one representative was kept for Zn-binding sites that, though found in different PDB structures, fall in the same position of the same protein domain. Sites that are not physiologically relevant based on literature inspection, e.g., Zn-substituted structures, spurious sites, were manually removed from the dataset. This procedure resulted in a set of 6450 zinc-binding motifs derived from 2651 Zn-binding sites.

The Zn proteome of Acinetobacter baumannii was obtained by using the hmmscan tool (Eddy, 1998) to search each bacterial sequence for the profiles contained in the two libraries. A bacterial sequence was identified as containing a potential Zn-binding site if at least one of the following conditions was verified: A) The profiles of all the fragments of a given site matched the sequence with an e-value lower than 10⁻³ and the corresponding ligands are conserved in the sequence. B) The profile of a domain with associated ligands matched the sequence with an e-value lower than 10⁻⁵ and the ligands are conserved in the sequence. C) The profile of a domain with no associated ligands matched the sequence with an e-value lower than 10⁻⁷. These predictions were integrated by adding the proteins annotated as Zn-binding in the UniProt database (UniProt Consortium, 2018). In total, 213 proteins of Acinetobacter baumannii were identified as Zn-binding proteins.

FAD quantification

Overnight A. baumannii cultures were back-diluted 1:50 into LB for 1 h. Cultures were then inoculated 1:100 into LB containing 40% CP buffer (20 mM Tris-HCl pH 7.5, 100 mM NaCl, 5 mM β-mercaptoethanol, 3 mM CaCl₂) supplemented with 200 μg/ml CP and grown for 6 h. Optical densities were normalized across strains, and cells were pelleted and washed with PBS. Total cellular FAD was quantified using the FAD assay kit (Abcam). Pellets were resuspended in FAD assay buffer and lysed twice using Lysis Matrix B tubes (MP Biologicals) and a FastPrep-24 (MP) bead beater. Supernatants were collected, and protein concentration was determined by BCA assay (Thermo Scientific). Samples were deproteinized by adding ice-cold 4 M perchloric acid and incubated on ice for 5 min. Samples were centrifuged for 2 min at 4°C at 13,000 × g, and supernatants were transferred to a new Eppendorf tube. Icecold 2 M potassium hydroxide was added until the pH was 6.5–8. Samples were centrifuged for 15 min at 4°C at 13,000 × g, and supernatants were collected. To quantify FAD, 50 μL reaction mixtures were prepared by combining 46 μL of 1:50 dilution of sample with 2 μL OxiRad probe and 2 μL enzyme mixture (Abcam). OD₅₇₀ was monitored over 30 min, and values were plotted against an FAD standard curve and normalized to total protein. Data are averaged from 3 independent experiments.

Purification of A. baumannii proteins

Oligonucleotide primers used for cloning are listed in Supplemental Table S7. The gene sequences of ribA, ribB and ribBX from Acinetobacter baumannii ATCC 17978 (locus tags A1S_3107, A1S_0823 and A1S_3388, respectively) were obtained from the SEED database. The gene ribBX encodes a 373 amino acid protein that is 41% identical and 56% similar to the homolog encoded by Mycobacterium tuberculosis H37Rv (locus tag Rv1415; NP_215931; PDB: 4I14) over their entire length (Figure S4B) (Singh, et al., 2013). The calculated molecular mass and theoretical isoelectric point (pI) value for the A. baumannii RibBX are 40,899 Da and 5.62, respectively (https://web.expasy.org/protparam/). Full length proteins were cloned into a pHIS plasmid at the NcoI site using an isothermal assembly method and was expressed with an N-terminal His-tag (Gibson, et al., 2009).

These resulting plasmids were transformed into E. coli BL21(DE3) for expression. Cells were grown in Luria–Bertani (LB) medium supplemented with 100 μg/mL ampicillin and shaking at 37 °C to OD₆₀₀ of 0.6, at which time ITPG was added to final 0.5 mM. The cells were allowed to induce at 16 °C overnight. For each 0.8 L cell pellet, cells were suspended in 40 mL lysis buffer (25 mM Tris, 500 mM NaCl, 2 mM TCEP, pH 8) and sonicated at 60 % power (3 s on; 9 s off, on ice), for total 15 min/L cells. This solution was centrifuged at 10,000 g (20 min, 4 °C). The supernatant was filtered with 0.22 μm for loading onto a 5 mL HisTrap column (GE Healthcare). The Ni-NTA column was run at 3 mL/min, and protein was eluted with a 4–40% buffer B (25 mM Tris, 500 mM NaCl, 2 mM TCEP, pH 8) gradient over 20 column volumes. The protein-containing fractions were pooled, cleaved with TEV protease and dialyzed again buffer (25 mM Tris, 500 mM NaCl, 2 mM TCEP, pH 8) at 4 °C overnight. Solutions were filtered with 0.22 μm and run through Ni-NTA column. Non-tagged protein factions are pooled and concentrated to ~3 mL for loading on Hi LoadTM 16/600 SuperdexTM 200 or 75 (GE Healthcare). Running buffer was (25 mM Tris, 200 mM NaCl, 2 mM TCEP, pH 8), and protein was eluted at a flow-rate of 1 mL/min. Protein-containing fractions were pooled, concentrated and flash-frozen in liquid nitrogen, stored at –80 °C, and buffer exchanged prior to use. Protein mass was confirmed by LC-ESI-MS or trypsin digested MALDI-MS and was judged to be 90% pure by overloaded SDS-PAGE gels. For RibBX, the residual FMN bound to RibBX during purification was confirmed by LC-MS/MS and UV-Vis spectroscopy.

GCHII activity assay

The GTP cyclohydrolase II (GCHII) activities of RibA and RibBX were measured essentially as previously described (Singh, et al., 2013). Briefly, reaction mixtures were prepared in a buffer containing 50 mM Tris, 100 mM NaCl, 10 mM MgCl₂, 2 mM TCEP, 10 μM ZnSO₄, pH 8.0 in a total volume of 200 μL, with reactions initiated by the addition of enzyme to a final concentration of 2 μM with 0~1 mM GTP. To monitor the formation of 2,5-diamino-6-ribosylamino-4(3H)-pyrimidinone 5’-phosphate (DARP), the absorbance at 310 nm was monitored at 37 °C for 20 min and the concentration of DARP was calculated using an extinction coefficient of Abs₃₁₀ = 7.43 mM⁻¹cm⁻¹. The linear range of these kinetic traces was used to calculate the initial rate and these rates were used to fit to the Michaelis-Menten steady-state model.

DHBPS activity assay

The 3,4-dihydroxy-2-butanone 4-phosphate synthase (DHBPS) activities of RibB and RibBX were measured by a colorimetric method as described (Singh, et al., 2013). Briefly, reaction mixtures contained 50 mM Tris, 150 mM NaCl, 10 mM MgCl₂, pH 7.5 in a total volume of 60 μL. Reactions were initiated by the addition of enzyme to final 1 μM with final 0~1 mM ribulose-5-phosphate (Cayman). After incubation at 37 °C for 15 min, 48 μL of a saturated creatine (Fisher Scientific) solution was added, followed immediately by 24 μL 1-naphthol (Fisher Scientific) (dissolved at 35 mg/mL in 1 M NaOH). The absorbance at 525 nm was measured after 45 min using plate reader (BioTek, Synergy Neo2). The standard curve was constructed with 0~500 μM 2,3-butadione (Sigma-Aldrich) for calculating product (DHBP) formation. Initial rates were calculated for Michaelis-Menten fit. Inhibition of FMN was measured in the presence of 0~1 mM FMN (Cayman).

FMN binding by ITC

All ITC titrations were carried out as described before (Grossoehme and Giedroc, 2009) using a MicroCal VP-ITC calorimeter. 300 μM FMN was placed into the syringe with ~30 μM RibBX protomer in the reaction cell. Injections (2~8 μL) were made at a rate of 2 μL s⁻¹, with 180~240 s allowed for equilibration of the mixture between injections. All reactions were conducted in at least triplicate in 50 mM Tris, 150 mM NaCl, 10 mM MgCl₂, pH 7.5 at 25.0 °C. The Origin 7.0 Software package provided by MicroCal was used and data were fitted as single-site binding model with the n and K_a values optimized during the fit.

Crystal structure of RibBX

RibBX expressed and purified as described above was subjected to polishing on a Superdex 200 10/300 GL column (GE Healthcare) running in 25 mM Tris, 200 mM NaCl, 2 mM TCEP, pH 8.0 at a flow-rate of 0.5 mL/min. Peak fractions were concentrated to 8~12 mg/mL using an Amicon centrifugal concentrator with a 30 kDa cutoff membrane (Millipore) and used for crystallization by vapor-diffusion in 96-well sitting-drop plates. Initial crystallization screening was carried out using commercial kits from MemGold2™ (Molecular Dimensions) by mixing 0.15 μL reservoir buffer and 0.15 μL protein solution and equilibrating against 30 μL reservoir buffer at 293 K. Small crystals appeared after 1 day when RibBX was equilibrated against a reservoir solution containing 10 mM MES, 100 mM NaCl, 150 mM ammonia sulfate, 19% (w/v) PEG 1000. The crystallization conditions were optimized over 19~24% (w/v) PEG1000 over 10~15 mg/mL RibBX while changing the protein:buffer ratio to 2:1 by hanging drop vapor diffusion. The plates were incubated at 20 °C for 1–2 weeks, with crystals appearing before that time.

Crystals were harvested after 4–7 days, cryo-protected in the reservoir solution supplemented with 25% glycerol and flash frozen in liquid nitrogen. Diffraction data were collected at 100 K at the Beamline station 4.2.2 at the Advanced Light Source (Berkeley National Laboratory, CA) and were indexed, integrated, and scaled using XDS (Kabsch, 2010). The structure was solved by molecular replacement using PHASER the PDB 4I14 as search model. The Autobuild function was used to generate a model that was improved by iterative cycles of manual building in Coot (Emsley, et al., 2010) and refinement using PHENIX (Adams, et al., 2010). Torsion-angle non-crystallographic symmetry restraints and secondary structure restraints were used during refinement. MolProbity software (Chen, et al., 2010) was used to assess the geometric quality of the models. PyMOL (PyMOL Molecular Graphics System, Version 1.2r3pre, Schrödinger, LLC.) was used to generate molecular images. Data collection and refinement statistics are indicated in Table S6.

Analytical gel filtration chromatography

A Superdex 200 10/300 GL column (GE Healthcare) was equilibrated and chromatographed in 50 mM Tris, 150 mM NaCl, 10 mM MgCl₂, 2 mM TCEP, pH 7.5 at a flow-rate of 0.5 mL/min and calibrated with the Gel Filtration Calibration Kit HMW (GE Healthcare). A linear fit a plot of K_av vs. log M_r gave R²= 0.90. 100 μM RibBX was pre-incubated with 1.25 eq. FMN at 4 °C for 1 h. 100 μL samples were injected for each chromatography run.

A. baumannii growth with riboflavin

Overnight cultures of A. baumannii were subcultured 1:50 in LB for 1 h. Cultures were then inoculated 1:100 into M9 minimal media supplemented with Vishniac’s trace metal mix with or without ZnCl₂ as previously described (Nairn, et al., 2016). M9 minimal media supplemented with full Vishniac’s trace metal mix contained 36 μM Fe, 5.2 μM Mn, 3.2 μM Cu, 1.8 μM Mo, 13.4 μM Co and 27.2 μM Zn. Sodium succinate dibasic hexahydrate (Sigma) or sodium fumarate dibasic (Sigma) were used as the sole carbon source at 0.5 % (w/v). Riboflavin (Sigma) was supplemented with a final concentration of 10 μg/mL, and OD₆₀₀ was monitored over time. Data are representative of at least 3 independent experiments.

Quantification and Statistical Analysis

Bioassays were performed with at least 3 biological replicates (n=3) as indicated. The averages and standard deviation of the separate replicates and Student’s t test were then calculated using Graphpad Prism and Microsoft Excel. The structure of RibBX was refined with internal statistical analysis as reported in the deposition in the PDB. Detailed quantification and statistical analysis are fully described in the manuscript Figure legend and Methods sections of the manuscript.

Data and software availability

The coordinates for the RibBX crystal structure have been deposited in the Protein Data Bank PDB: 6MNZ. Software used to analyze RNA sequencing results are fully described in the manuscript Methods section with appropriate citations, and data have been uploaded to the National Institutes of Health Gene Expression Omnibus with accession number GSE125491 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE125491). Briefly, raw sequencing reads were processed using CASAVA-1.8.2 for FASTQ conversion. FASTQ reads were processed using the Bayesian adapter trimmer Scythe version 0.992. EDGE-pro version 1.3 was used to align reads with Bowtie2 version 2.1.0. The expression level of each gene was determined using DESeq in the statistical programming package R-3.0.0. Software used for diffraction data are fully described in the manuscript Methods section with appropriate citations. Briefly, diffraction data were collected at 100 K at the Beamline station 4.2.2 at the Advanced Light Source (Berkeley National Laboratory, CA) and were indexed, integrated, and scaled using XDS (Kabsch, 2010). Software used for LC-MS/MS data is xcaliber v4.0. All software used in this study are reported in Method Details and indicated in the Key Resources Table.

Supplementary Material

2

Table S1. RNAseq analysis of untreated wild-type (WT) vs. wild-type A. baumannii treated with 250 μg/mL CP (WT+CP) in 60% TSB (see Supplementary Table 1.xlsx). Related to Figure 1, main text.

3

Table S2. Complete list of proteins detected in a proteomics analysis of untreated WT A. baumannii sorted according to relative abundance (see Supplementary Table 2.xlsx). Related to Figure 2, main text.

4

Table S3. List of tryptic peptides and accompanying MS/MS data used to positively identify a candidate A. baumannii copper storage protein (Csp) in CP-treated cell lysates (see Supplementary Table 3.xlsx). Related to Figure 2D, main text.

5

Table S4. Proteomics analysis of cell lysates from WT vs ΔzigA A. baumannii cells grown in the presence of 200 μg/mL calprotectin (CP) (see Supplementary Table 5.xlsx). Related to Figure 4, main text. Proteomics analysis of cell lysates from untreated WT vs ΔzigA A. baumannii obtained from cells grown in the absence of calprotectin (CP) (see Supplementary Table 4.xlsx). Related to Figure S4.

6

Table S5. Predicted Zn proteins in the A. baumannii ATCC 171978 genome, and in each of the three extrachromosomal plasmids, pAB1, pAB2, pAB3 (see Supplementary Table 6.xlsx). Related to Figure S4.

7

Table S7. Oligonucleotides (see Supplementary Table 7.xlsx). Related to STAR Methods.

Significance.

Vertebrate host-derived nutritional immunity, defined as transition metal withholding, and transition metal toxicity, operate at the host-microbial pathogen interface, in an effort to limit the impact of bacterial infections. Calprotectin (CP) is one of a large family of Ca^II-activated transition metal-binding proteins, secreted by neutrophils that migrate to sites of infection, to starve those bacteria of these essential metals. Although CP is a versatile transition metal chelator, significant questions remain as to what metals are sequestered by CP at infection sites, and how the loss of specific metals impact cellular physiology. Here, we show that CP induces an acute Zn^II and Fe starvation response in the important opportunistic pathogen Acinetobacter baumannii, and probe the function of the Zn-uptake repressor (Zur)-regulated GTPase ZigA in cellular adaptation to CP. We show that CP and the loss of ZigA severely lowers cellular flavin levels, and that a major metabolic response to CP-mediated Fe restriction is a prioritization of de novo flavin biosynthesis. We uncover a dual regulatory model of (ribo)flavin biosynthesis that is operative under these metal-deplete conditions and present the crystallographic structure of a fusion protein, RibBX, which is feedback-inhibited by flavin mononucleotide (FMN) but only at high intracellular flavin. This work investigates the cellular mechanisms of adaption to Festarvation as key feature of CP-induced cellular stress, and further refine our understanding of the intersection of metal homeostasis systems and the flavin biosynthesis pathway in an important human pathogen. Further, in addition to the proteins ZigA and RibBX, we identify a number of Zn metalloenzymes that are established antimicrobial targets and are cell-abundant under conditions of host-derived nutritional immunity, that might be leveraged for the development of antimicrobial strategies against Acinetobacter baumannii.

Highlights.

• Calprotectin (CP) induces zinc- and iron-deficiency in Acinetobacter baumannii.

• Flavin biosynthesis is impaired when zinc- and iron-deficiency is exacerbated.

• A dual regulatory model of flavin biosynthesis ensures flavin supply in CP-stress.

• A widespread fusion protein RibBX impacts flavin supply in CP stress.