Abstract
Infections caused by Acinetobacter baumannii, a Gram-negative opportunistic pathogen, are difficult to eradicate due to the bacterium’s propensity to quickly gain antibiotic resistances and form protective bacterial multicellular communities known as biofilms. The A. baumannii DNA damage response (DDR) mediates antibiotic resistance acquisition and regulates RecA in an atypical fashion; both RecALow and RecAHigh cell types are formed in response to DNA damage. In this study, we show that RecA levels modulate biofilm development, formation and dispersal, through bfmR, the global biofilm regulator. RecA loss results in surface attachment and prominent biofilms while elevated RecA leads to diminished attachment and dispersal. Recalcitrance to treatment may be explained by DDR induction, common during infection, and the balance between biofilm maintenance in low RecA cells, and increased mutagenesis and dispersal to reach new niches in high RecA cells. These data highlight the importance of understanding fundamental biology to better treat bacterial infections.
Impact The mechanism of biofilm formation and dispersal in A. baumannii, shown here to depend on RecA levels, contributes to the understanding of recalcitrant infections caused by this important pathogen.
Introduction
Acinetobacter baumannii is an emerging Gram-negative opportunistic pathogen and one of the ESKAPE pathogens, a group of bacteria responsible for most hospital-acquired infections (Rice, 2008). A. baumannii outbreaks in hospitals are difficult to eradicate, due to increased multi-drug resistance (MDR) (Peleg et al., 2008) and its ability to form biofilms (Eze et al., 2018). A. baumannii infections are very dangerous to immunocompromised individuals, causing various illnesses, including pneumonia, septicemia, and wound infections (Smith et al., 2007).
When gene products involved in antibiotic binding or processing on the chromosome are mutated, resistance is acquired (Blair et al., 2014). One response pathway underlying multidrug resistance (MDR) is the DNA damage response (DDR). Mutagenesis results from induction of error-prone DNA polymerase genes which are part of the DDR regulon (Cirz et al., 2007; Fuchs and Fujii, 2013). We have shown that in A. baumannii, RecA-dependent induction of multiple error-prone polymerases in response to DNA damage leads to clinically relevant antibiotic resistance (Norton et al., 2013). In Escherichia coli and many other bacteria, the cells’ main recombinase, RecA, and the global transcriptional repressor, LexA, manage the DDR, also known as the SOS response (Little and Mount, 1982). In contrast, there is no known LexA homologue in A. baumannii, indicating the A. baumannii DDR circuitry is regulated differently (Ching et al., 2017; Hare et al., 2014; MacGuire et al., 2014; Norton et al., 2013). We also showed that in response to DNA damage there are two RecA cell types, low recA expression (RecALow) and high recA expression (RecAHigh), regulated by a 5’untranslated region in the recA transcript(Ching et al., 2017; MacGuire et al., 2014).
Bacterial cells within biofilms are often less sensitive to chemical and physical challenges, including antibiotics (Anderl and Franklin, 2000; Singh et al., 2016). Like many biological processes, the biofilm cycle in A. baumannii is not fully understood, though key biofilm genes have been described, including the biofilm master regulator gene bfmR that controls adhesive csu pili and genes important for virulence and desiccation tolerance (Farrow et al., 2018; Tomaras et al., 2008, 2003). It is also known that genes encoding Bap protein and efflux pumps are important for biofilms (Brossard and Campagnari, 2012; He et al., 2015; Loehfelm et al., 2008; Tomaras et al., 2008, 2003). The BfmR response regulator is part of a two-component system (BfmRS) (Farrow et al., 2018; Geisinger and Isberg, 2015; Russo et al., 2016; Tomaras et al., 2008; Wang et al., 2014) that results in BfmR phosphorylation. It is unknown if bfmR expression is induced by another factor or whether de-repression by phosphorylated BfmR, with lower affinity for its operator site (Draughn et al., 2018), is enough for induction. The A. baumannii biofilm extracellular matrix has been reported to contain poly-ß-(1-6)-N-acetyl-glucosamine, mannose and extracellular DNA (Bales et al., 2013; Choi et al., 2009; Hardouin et al., 2014; Sahu et al., 2012).
In summary, the DDR and the biofilm cycle are two developmental pathways in bacteria to assure genomic stability and survival. In some bacteria, antibiotics and DNA damaging agents have been observed to induce biofilm formation, suggesting a link between these two pathways (Linares et al., 2006; Takajashi et al., 1995). However, the mechanisms linking the DDR and biofilms varies between bacterial species. For example, in Streptococcus mutans, Escherichia coli (Costa et al., 2014; Inagaki et al., 2009), and Clostridium difficile (Walter et al., 2015) there is a direct correlation between DDR and biofilms in which both pathways are induced in response to environmental signals. However, we showed that during Bacillus subtilis biofilm development, induction of the DDR shuts off biofilm matrix genes in a subpopulation of cells (Gozzi et al., 2017), indicating an inverse correlation between DDR and biofilm formation. In this case, DDR signaling complements the biofilm cycle, and cells that are DDR-induced could potentially leave the biofilm to search for new niches. Altogether, the varied relationships between DDR and the biofilm cycle are intriguing and show that cells have evolved different ways to balance these strategies, which could have downstream effects in treatment and clinical practices.
Here, we demonstrate that the DDR and the biofilm cycle intersect in A. baumannii. We show that loss of RecA in A. baumannii results in increased cell-to-surface adherence, in part due to elevated expression of bfmR with a concomitant increase in Csu pili. We further identify novel regulation of bfmR through the RecA-dependent UmuDAb gene product. We observed that DNA damaged biofilm pellicles shut off csuAB gene expression, turn on recA, and cells disperse. The observed inverse relationship between DDR and biofilm formation may lead to a heterogeneous population combining physical (biofilm) and genetic (elevated mutagenesis) protection to environmental challenges.
Results
Loss of RecA promotes biofilms in A. baumannii
A. baumannii ATCC 17978 ΔrecA cells (recA::km, Table S1 (Aranda et al., 2011)) grown in standard LB medium and shaking conditions consistently displayed observable surface attachment on the sides of glass tubes (Fig. 1A, red arrow). In comparison, the parental wild-type strain had little to no surface attachment (Fig. 1A). This observation suggested that A. baumannii ΔrecA cells form more biofilm than WT cells. Since lack of RecA can presumably affect bacterial growth (Cox et al., 2008), we performed growth measurements of free-living planktonic WT and ΔrecA cells. We also measured growth in ΔrecA cells complemented with a low-copy plasmid, pNLAC1(Luke et al., 2010), containing recA under its own promoter (Ching et al., 2017) (ΔrecA (pNLAC1-recA), Table S1), which we will refer to as the complemented ΔrecA mutant. We observed a slight growth disadvantage in ΔrecA cells whenever the starting inoculum size was <107 cells that was rescued by the extrachromosomal copy of recA (Fig S1A & B). Our result is largely consistent with a previous report that the ΔrecA strain showed no noticeable growth defects compared to the parental WT strain (Aranda et al., 2011).
To further assay biofilm formation, we prepared pellicle biofilms (formed at the air-liquid interface) with ΔrecA and WT cells as indicated in materials and methods. WT cells did not form a visible pellicle (Fig. 1B, left) at 24 hrs while ΔrecA cells (Fig. 1B, middle) formed a visible and opaque pellicle as early as 24 hrs, consistent with previous observations (Fig. 1A). At 48 and 72 hrs the WT cells formed a pellicle that was smooth and with striations while the ΔrecA pellicle was thicker and granular on the surface (Fig. S2). Cell adherence to the sides/bottom of the wells were measured by Crystal Violet staining. We observed a significant increase in surface-attached cells in ΔrecA biofilms at 24 hrs compared to WT cells (Fig. 1C). These data suggest that ΔrecA makes prominent biofilms compared to the WT strain. To confirm that the observed ΔrecA biofilms are indeed due to loss of RecA, we tested the ΔrecA complemented strain for biofilm formation. At 24 hrs, there was less observable pellicle formation in the complemented strain (Fig 1B, right), and surface architecture was similar to WT (Fig. S2).
Biofilm formation is also studied at the colony level (growth at the air-surface interface) (Vlamakis et al., 2013). Typically, robust colony biofilms have more wrinkling surface architecture due to differential cell death and mechanical forces as a result of increased extracellular matrix (Asally et al., 2012). Thus, we grew colony biofilms on solid medium plates containing Congo red and Coomassie brilliant blue, dyes that detect changes in biofilm matrix components (Ghodke et al., 2019; Surgalla and Beesley, 1969). At 48 hrs, the ΔrecA colony biofilm had differential coloring to the WT and complemented strain indicative of increased biofilm matrix production (Fig 1D). We noticed that the ΔrecA colony biofilms also had decreased diameter size throughout the length of the experiment. Notably, the complemented ΔrecA colony is similar in size to WT but with an intermediate coloration (Fig. 1D). We have previously observed partial complementation of colony biofilms phenotypes with plasmid borne ectopic gene copies, possibly due to the sensitivity of the process to gene dosage (Ching et al., 2018). The data obtained from these experiments suggest that the ΔrecA strain forms robust biofilms.
Inverse relationship between RecA and biofilm is not exclusive to A. baumannii ATCC 17978
To assess whether this finding was exclusive to A. baumannii ATCC 17978, we used a transposon derivative of A. baumannii AB5075 in which the transposon is inserted approximately in the middle of the RecA ORF (Gallagher et al., 2015) (Table S1). A. baumannii AB5075 is a highly virulent clinical isolate (Jacobs et al., 2014). For pellicle biofilms at 48 hrs, the AB5075 insertional recA::Tn mutant had significantly increased cell adherence compared to the parental strain (Fig. S3A). Furthermore, the AB5075 recA::Tn mutant colony biofilm is also smaller than the parental strain and displays differences in colony morphology, including changes in opacity and striations on the edge (Fig. S3B). These data suggest that the observation that RecA negatively impacts biofilm development is not exclusive to the ATCC 17978 strain.
RecA modulates the biofilm matrix in A. baumannii
Since ΔrecA forms prominent biofilms, we were curious to observe closely the differences between ΔrecA and WT biofilms, especially regarding the extracellular matrix. Thus, we analyzed ΔrecA and WT pellicle biofilms by scanning electron microscopy (SEM). To observe pellicle biofilms from the WT and ΔrecA strains we used a fixing procedure that takes advantage of cationic dyes binding to negatively charged polysaccharides and preserve these for imaging (Erlandsen et al., 2004). We found that there were noticeable differences between the matrix and biofilm structures of the WT and ΔrecA strains (Fig. 1E). At 24 hrs WT cells have some matrix and connection between cells, while ΔrecA cells were embedded in an observable thick matrix (Fig. 1E). This finding suggests that ΔrecA cells produce extracellular matrix earlier in their biofilm development.
Surface-attached cells withstand antibiotic treatment
Biofilms have been noted to protect bacteria from antibiotic exposure (Anderl and Franklin, 2000; Singh et al., 2016). To determine if there were any differences in antibiotic susceptibility of the WT and ΔrecA biofilms compared to free-living cells, surface-attached WT and ΔrecA biofilms were treated with the bactericidal aminoglycoside gentamicin. First, we determined the MIC of gentamicin for exponentially growing free living planktonic cells and found it to be the same for both WT and ΔrecA cells (1.88 μg/mL, error within 2-fold). Aranda et al. determined the MIC in free living planktonic cells to Amikacin and Tobramycin, both aminoglycosides, and reported that these were also the same for the WT and ΔrecA strains (1.5 μg/mL and 0.38 μg/mL, respectively) (Aranda et al., 2011).
To assess whether biofilms withstand gentamicin treatment differently to planktonic cells, the spent growth medium from 72 hr pellicle biofilms was removed, the wells were washed to eliminate non-attached cells and gentamicin was added in fresh growth medium at different concentrations, and wells were incubated for 24 hrs. After exposure to gentamicin, the growth medium and non-attached cells were removed, and Crystal Violet staining was performed as before. Strikingly, there were more surface-attached ΔrecA cells suggesting that they withstood treatment better than WT cells (Fig. 2A). For the ΔrecA strain, the percentage of surface-attached cells remaining after antibiotic treatment plateaus above 50% relative to no treatment after exposure to 0.47–15 μg/mL of gentamicin (Fig. 2B), representing over 3-fold excess of ΔrecA surface-attached cells compared to WT (at 15 μg/mL the percentage of WT surface-attached cells decreases to ~16% relative to no treatment).
To determine if the remaining surface-attached cells were viable, we directly measured the viability of surface-attached cells before and after gentamicin treatment, by LIVE/DEAD staining and quantified the number of cells within 5 independent fields of view. We find that without gentamicin and compared to the WT, the ΔrecA strain had ~ 2-fold more viable surface-attached cells (Table 1). The gentamicin treated biofilms (Table 1) shows that ΔrecA cells in a biofilm survive approximately 4-fold better than WT cells upon antibiotic exposure (Table 1). Taken together, the evidence indicates that biofilms made by ΔrecA protect cells from antibiotic treatment.
Increasing RecA diminishes biofilm formation
Together, our data suggest that there is an inverse relationship between RecA and biofilms in A. baumannii. To further test this relationship, we artificially increased RecA levels in the WT cells and expected that a RecA overproducing strain would form poor biofilms with fewer surface-attached cells than WT or ΔrecA biofilms. Thus, we introduced the plasmid borne copy of recA under its own promoter (pNLAC1-recA, Table S1) into the WT strain, which we refer to as recA++ (Table S1). We first measured the intracellular RecA concentration of recA++ relative to the WT strain. To do this, we purified A. baumannii RecA and Escherichia coli RecA to test whether an antibody raised against E. coli RecA recognizes A. baumannii RecA equally well. There is no difference between the detection of the E. coli RecA or A. baumannii RecA for the antibody (Fig. S4A&B). Using semi-quantitative western blot analysis, we found that the relative level of RecA in recA++ cells were between 3- to 5-fold higher than WT in both basal and DNA-damage inducing conditions than the parental strain (Fig. S4C). We calculated that after DNA damage, RecA relative levels in the WT strain increased by ~3-fold relative to basal. Thus, recA++ at basal conditions has similar levels of RecA as the WT strain treated with 10X MIC of ciprofloxacin (Fig. S4C).
We observed that recA++ biofilms had significantly less surface-attached cells compared to the WT and ΔrecA biofilms, but still above the limit of detection (Fig 3A, dashed line shows limit of detection).
The recA++ strain grown in planktonic conditions had a slight growth defect which was like ΔrecA cells in shaking conditions. These strains doubling time was 0.15X slower than WT based on these growth curves (Fig. S1 C&D).
RecA levels influence surface attachment
Since there were minor growth defects in planktonic shaking conditions for ΔrecA and recA++ cells (Fig. S1) in the conditions tested, we decided to concurrently measure biofilm attachment and bacterial growth in the first 13 hours of static growth conditions. We inferred that difference in biofilms due to growth would be apparent at the early stages of biofilm formation where there may still be active cell division. To do this, we set up biofilms as indicated in materials and methods and measured surface attachment while also determining colony forming units (CFUs) from 0-13 hrs at regular intervals. Strikingly, at 6 hrs we detected an increase in surface-attached cells for ΔrecA compared to WT and recA++, both of which showed little attachment (Fig. 3B). The colony forming unit counts were comparable for all the strains (Fig. 3C). These data show that ΔrecA cells attach to surfaces earlier than the WT or the recA++ strains, as our previous assays suggested (Fig. 1B). This also demonstrates that our findings for biofilm formation are directly comparable and not, in part, due to growth differences. In this experiment, as well as in the one shown previously (Fig. 1B), strains contained the empty plasmid (pNLAC1; Table S1) for direct comparison to the complemented strain. We observed that regardless of pNLAC1 (Fig. 3B, C) ΔrecA forms robust biofilms, while WT and recA++ do not. Overall our data indicates an inverse relationship between RecA and biofilm development.
RecA levels influence density of attachment pili
Since ΔrecA cells attaches to surfaces faster than WT or recA++ (Fig. 3A, B), we wanted to investigate whether ΔrecA had more surface pili to permit attachment. Csu pili are involved in cell adherence in A. baumannii biofilms (Gaddy and Actis, 2009; Pakharukova et al., 2018; Rumbo-Feal et al., 2013; Tomaras et al., 2003). To investigate this, we imaged the ΔrecA, WT and recA++ strains from saturated cultures using transmission electron microscopy (TEM). Remarkably, ΔrecA cells displayed a higher density of surface pili than WT or recA++ cells (Fig. 3D). To determine whether we were observing, in part, Csu pili, we performed a Western blot to detect CsuAB protein from both purified pili (Fig. 3E) and total cell extracts (Fig. 3F). We detected a strong signal for CsuAB pili in ΔrecA cells.
We next constructed a plasmid containing the promoter of the csuAB operon fused to gfp (PcsuAB-gfp) which was introduced into ΔrecA, WT, and recA++ strains. Fluorescence measurements of statically grown cells in YT medium showed a fluorescence signal in ΔrecA cells at ~6 hours, and the signal further accumulated to orders of magnitude more than WT or recA++ (Fig. 4A). Indeed, GFP fluorescence of the WT strain with the same csu reporter plasmid was detected after 36 hours and gradually increased with time. Expression in the recA++ strain was even lower, while growth is comparable (Fig. 4A). Single cell imaging of these cells at 20 hrs showed that fluorescence was easily detectable in ΔrecA, but largely absent in WT or recA++ (Fig. 4B). Quantification of fluorescence of single cells (Fig. 4C) shows significantly elevated average fluorescence in ΔrecA cells compared to recA++ or WT. Notably, recA++ fluorescence is significantly lower than the WT (Fig. 4C). We next measured expression levels of csu pili genes in biofilm pellicle cells using qPCR. csuAB expression is ~80-fold higher in A. baumannii ΔrecA and lower in recA++ compared to WT cells (Fig. 4D). In summary, our data suggested that ΔrecA cells have deregulated CsuAB pili, which leads to faster surface attachment compared to recA++ or WT.
RecA levels and csuAB expression are inversely correlated in DNA-damaged biofilm pellicles
Our findings suggest that increased RecA, which occurs upon DDR induction, will result in decreased biofilms. Due to the absence of a LexA homologue in A. baumannii, it is possible that RecA may serve different regulatory roles as has been previously suggested (Hare et al., 2014). Our results suggest that elevated levels of RecA leads to decreased CsuAB pili (Figs. 3–4). Thus, we hypothesized csuAB expression would decrease in biofilm cells upon treatment with DNA damage, which leads to an increase in RecA. To test this and monitor both RecA and csuAB expression we constructed a plasmid borne transcriptional reporter of Csu genes as the one we used previously (Fig. 4), but used instead mCherry, PcsuAB-mCherry, and we introduced it into the A. baumannii strain already containing a chromosomal PrecA-gfp reporter (Ching et al., 2018). 48 hr pellicle biofilms of the double reporter strain were treated with 0.5X MIC of ciprofloxacin (Cip), which is known to induce the DDR but does not kill cells (MacGuire et al., 2014), for an additional 48 hrs. The biofilm pellicles were examined by confocal microscopy to observe differences in gene expression, and in biofilm thickness.
In untreated biofilms (Fig. 5A), most cells had moderate to high expression of the PcsuAB-mCherry pili reporter with few expressing the PrecA-gfp (Fig. 5A). In comparison, cells isolated from Cip-treated biofilms showed dramatic changes (Fig. 5B). Most of the DNA damaged biofilm cells were fluorescing GFP, an indication of RecA induction, with few cells expressing mCherry. There appeared to be little to no overlap in gene expression between the two reporters, suggestive of mutually exclusive cell types. Most treated cells were also elongated, as cell division inhibition is a common feature of DDR induction (Kreuzer, 2013). Notably, the untreated pellicles showed high cell density and had a greater thickness (28 μm) than the treated pellicles (22 μm), suggesting that biofilm dispersal occurred in response to DNA damage treatment. It is unlikely that thinning is due to cell death by the Cip treatment since we used sub-MIC levels of Cip. Our data suggest that RecA negatively influences expression of Csu pili genes, necessary for expression of the Csu pili machinery and biofilm formation.
RecA levels modulate bfmR expression
Csu pili regulation is dependent on BfmR, a key regulator of biofilm formation in A. baumannii (Tomaras et al., 2008). Increased Csu pili in ΔrecA suggests either higher bfmR expression or perhaps a more active BfmR. To test whether bfmR was differentially expressed we measured expression levels of bfmR in biofilm pellicles using qPCR. We found that bfmR expression is significantly higher in ΔrecA, but not in recA++, compared to WT (~4 folds, Fig. 6A). These results suggest that RecA levels manage bfmR expression that in turn controls Csu pili.
UmuDAb, a RecA-dependent transcription factor, regulates BfmR
While A. baumannii lacks LexA, UmuDAb, is a known repressor of some, but not all, DDR genes (Hare et al., 2014). Notably, UmuDAb has a typical helix-turn-helix DNA binding motif at the amino-terminal end that is cleaved by activated RecA nucleoprotein filament, RecA*, in a way believed to be like that of LexA and other proteins within the same family (Draughn et al., 2018; Hare et al., 2013). Cleavage of UmuDAb by RecA* provides a plausible mechanistic explanation for the observations that high RecA levels (i.e. low UmuDAb) lead to biofilm dispersal. Thus, we hypothesized that UmuDAb may serve as a RecA-dependent regulator of biofilms.
We constructed a ΔumuDAb strain as we have previously done (Norton et al., 2013) and tested biofilm formation. We found that the ΔumuDAb strain formed significantly weaker biofilms than any of the other strains previously studied in pairwise comparisons (Fig 6B). In fact, ΔumuDAb formed significantly worse biofilms than recA++ (Fig. 6B). Consistent with this observation, TEM showed no visible pili on the surface of ΔumuDAb cells (Fig 6C). Furthermore, we detected almost no bfmR transcript in ΔumuDAb cells compared to WT (Fig 6A). Contrary to the known UmuDAb repressor function (Hare et al., 2014; Witkowski et al., 2016), our findings suggest that UmuDAb has a positive effect on bfmR and consequently on csu. Taken together, our data suggest that UmuDAb is an inducer of bfmR. Our group is currently investigating the mechanism by which UmuDAb performs this function.
Discussion
Bacteria largely exist in the environment as biofilms (Costerton et al., 1995). Generally, cells in biofilms have different characteristics and gene expression profiles from their free-living planktonic counterparts. Here, we find that in A. baumannii ATCC 17978, RecA levels, a key DDR gene product, influences biofilm development in either the ATCC 17978 (Figs. 1–6) or AB5075 (Fig. S3) strains. Our data also show that cells lacking recA produce robust biofilms (Fig. 1). Importantly, ΔrecA biofilms on an abiotic surface are more resilient (Fig. 2) and are viable after antibiotic treatment (Table 1) providing additional evidence to the concept that biofilms protect cells from antibiotic exposure (Anderl and Franklin, 2000; Singh et al., 2016). Thus, recurring A. baumannii infections may be the product of cells remaining adhered to equipment surfaces or implanted devices.
We found that RecA levels influence biofilm formation through Csu pili density. ΔrecA cell adherence is dependent on higher csuAB and increased Csu pili compared to either the WT or the recA++ strains (Figs. 3, 4). Our results further suggest that csuAB expression is due to induction of bfmR (Fig. 6). It has been previously shown that bfmR and csuAB expression is higher in biofilm cells (Rumbo-Feal et al., 2013) consistent with our findings.
Aranda et al. phenotypically characterized the ΔrecA insertional mutant used in this study. Free-living ΔrecA cells had decreased survival during heat shock, desiccation, UV, and certain antibiotic treatment (Aranda et al., 2011). Importantly, the ΔrecA strain had much lower pathogenicity in a mouse model, indicating its importance in this process (Aranda et al., 2011). These results demonstrate the significance of RecA in survival and virulence in free-living planktonic cells. However, our findings demonstrate the importance of investigating both biofilm and free-living states. For example, it has been shown that strong A. baumannii biofilm formers are less frequently antibiotic-resistant, consistent with having lower levels of RecA and subsequently mutagenesis (Wang et al., 2018). Moreover, biofilm cells were more resistant to eradication (Wang et al., 2018), again consistent with the observation that ΔrecA surface-attached cells are more difficult to eradicate with antibiotics (Fig. 2, and Table 1). For example, certain antibiotics have lowered penetration of biofilms (Anderl and Franklin, 2000; Singh et al., 2016) and there is a nutrient gradient within the biofilm which leads to different metabolic states of biofilm cells (Werner et al., 2004). Additionally, planktonic free-living cells have been shown to more readily gain higher level antibiotic resistances compared to biofilm cells during exposure to ciprofloxacin (Santos-Lopez et al., 2019).
Our findings have significant implications on understanding A. baumannii survival to antibiotic treatment and possibly antibiotic resistance. We show that upon DNA damage, the DNA damage response is induced, leading to induction of recA with simultaneous shutoff of biofilm genes, leading to thinner biofilms (Fig. 5). We have observed that this occurs through lowered bfmR and csuAB expression upon elevated RecA levels. Quantification of RecA upon DNA damage, and in recA++ cells, demonstrates how sensitive biofilm gene expression is to a change in RecA levels (Figs. S4, 1-3). This may be due to the UmuDAb sensitivity to cleavage by RecA* (Hare et al., 2013).
Our evidence suggests that UmuDAb is an inducer of bfmR and of biofilm formation (Fig. 6). A summary and model of genes is shown in Fig. 6D. Based on previous findings, we have observed bimodality of recA expression in planktonic cells (Ching et al., 2017; MacGuire et al., 2014), and there may also be heterogeneity in RecA levels between cells in biofilms. Thus, within the biofilm there may be two cell types: RecALow cells maintain biofilms and low mutagenic potential while RecAHigh cells with high mutagenic potential (Norton et al., 2013) can disperse and search for new niches. This observed inverse relationship can allow the population to combine physical (biofilm) and genetic (elevated mutagenesis) protection to environmental challenges, including antibiotic exposure in a host.
Overall, we have identified an inverse relationship between the DDR and biofilm development in A. baumannii. These results demonstrate the complexity of treating pathogens with a DDR that does not follow the paradigm, such as A. baumannii (Ching et al., 2017; MacGuire et al., 2014). Recent work has highlighted the potential for clinical RecA inhibitors to potentiate the effect of antibiotics while hindering antibiotic resistance acquisition by reducing mutagenic capacity in bacteria (Alam et al., 2016). This indeed may be promising in the treatment of certain bacteria, as demonstrated in E. coli (Alam et al., 2016). However, our results show that for A. baumannii, inhibiting RecA may lead to robust biofilm formation. It is thus important to understand the relationships between survival strategies in bacteria, and that different bacteria may have different responses to treatment, based on their fundamental biology. Overall, our results show the complex ways that A. baumannii robustly survives stress by balancing alternative survival strategies.
Methods
Strains and Growth Conditions
Strains and plasmids used are listed in Table S1. A. baumannii strains harboring the PNLAC1-recA plasmid were constructed as before (Ching et al., 2017). The ΔumuDAb strain was constructed using SOE PCR to insert a kanamycin gene cassette in the umuDAb gene as before (Norton et al., 2013). The oligonucleotides used are in Table S1. All bacterial cultures were routinely grown in LB medium, unless otherwise noted, and incubated at 37°C with shaking at 225 rpm for liquid cultures. YT medium used is composed of 2.5g NaCl, 10g Tryptone and 1g Yeast extract per liter. Solid medium contains 1.5% agar (Fisher Bio-Reagents). Antibiotics were used at the following concentrations: Kanamycin (Kan; 35 μg/mL), Gentamicin (Gm; 10 μg/mL), Tetracycline (Tet; 12 μg/mL) and Carbenicillin (Carb; 100 μg/mL).
Growth curve measurements
Strains were diluted in the growth medium as indicated in the respective figure legends and grown in YT medium in 96-well dishes. All strains were grown at least in triplicate. They were incubated in a plate reader (Biotek Synergy H1) at 37°C with shaking for 24 hrs. with measurements of OD600 every 15 mins.
Pellicle & Colony Biofilm Formation
To form pellicle biofilms, cells from a saturated culture were inoculated at a 1:1000 dilution into YT liquid medium in either 6, 12 or 24-well non-tissue culture treated (lacking coating which change surface properties) polystyrene plates. The same number of cells, adjusted based on optical density, were added to each well. Plates were then incubated statically at 25°C. A standard Crystal violet staining procedure was used to quantify adherence to the polystyrene surface (Chen et al., 2015). Early biofilms are formed after 24 hrs of static growth and between 48-72 hrs. the biofilm has matured. After 72 hrs. the biofilms start to deteriorate. In the time course experiment, to correlate viable cells and biofilm formation, samples were taken for CFU counting from 3 wells per experiment. The well content was thoroughly mixed before taking the samples, diluted with PBS, and plated on LB agar. After CFU sampling, the well content was aspirated, washed 3 times with PBS, and stained with Crystal Violet as before (Ching et al., 2018). For colony biofilms, 3 μL of a saturated bacterial culture were spotted on to YT agar plates containing 200μg/mL Congo red and 100μg/mL Coomassie brilliant blue (Congo red plates) (Ching et al., 2018) and incubated for 48 hrs., early colony biofilms, to 96 hrs. at 25°C. Images of biofilms were taken with a Leica MSV269 dissecting scope and a Leica DMC2900 camera, using the same settings. The radius of the biofilm colonies was measured in triplicate from the dissecting microscope images at the same magnification.
Scanning Electron (SEM) and Transmission Electron (TEM) microscopy
Biofilms for SEM were prepared by collecting pellicle biofilms on a glass coverslip treated with a solution of 0.1 mg/mL of polylysine (Fisher Scientific). The coverslips with the biofilms were deposited for at least 24 hrs at 4°C in wells with fixative 1 composed of 25% Glutaraldehyde, 0.1 M Na-Cacodylate buffer (pH 7.2), 0.15% Alcian Blue and 0.15% Safranin. Additional treatment of the coverslips and observation of the biofilms was performed at the Northeastern University Electron Microscopy Core Facility.
For TEM, cultures were grown to saturation in YT liquid medium and streaked for single colonies on YT plates incubated at 37°C. A colony was picked and gently resuspended in 100 μL of phosphate buffered saline (PBS). 10 μL of the resuspended colony was pipetted onto a copper grid, excess liquid wicked away with filter paper, and the grid was dried at room temperature. To negatively stain the samples and observe pili, 10 μL of a solution of 1.5% phosphotungstic acid (PTA) was pipetted onto the cell-containing grids. Liquid excess was wicked away with filter paper, and grids were left to air dry at room temperature. Microscopy was performed with a JEOL JEM-1010 (JEOL USA, Peabody, MA) microscope equipped with a 2k x 2k AMT CCD camera (Advanced Microscopy Techniques, Woburn, MA) at the Northeastern University Electron Microscopy Core Facility.
Quantitative measurement of gene expression
To measure bfmR and csu transcripts in WT, ΔrecA, and recA++, we took 500μL of 48 h statically grown pellicle cells. These were harvested by centrifugation and resuspended in 1 mL of RNA Protect Bacteria Reagent (Qiagen) with 30μL of 20 mg/mL lysozyme, incubated at room temperature for 15 min, spun down at 15,000g for 5 min, and cell pellets were stored at −20 °C. Total RNA extraction was carried out by using the Zymo Direct-zol RNA Kit (Zymo) according to manufacturer’s instructions. A NanoDrop One (ThermoFisher) was used to measure RNA concentration and purity. RNA was converted to cDNA using a high capacity cDNA reverse transcription kit according to the manufacturer’s protocol (Applied Biosystems). Expression of bfmR and csuAB pili gene expression by RT-qPCR was performed with cDNA using the qPCR primer pairs listed in Table S1 and following the protocol for Fast SYBR Green Master Mix (Applied Biosystems) using a StepOnePlus real-time PCR instrument (Applied Biosystems). Relative gene expression was standardized using endogenous 16S ribosomal RNA expression and the comparative threshold cycle (ΔΔCT) was calculated for each sample and compared relative to WT expression. Experiments were repeated twice, and each sample was run in biological triplicate. An unpaired two tail T test was used for statistical analysis relative to the parental strain (*=P < 0.05).
Pilus purification
Pilus shear preparations were performed as previously published (Moon et al., 2017). Briefly, similar number of cells grown on a YT agar plate (OD600~20) were collected in 1.5 mL of 1X PBS, placed on ice for 10 min, and vortexed for 1 min. After centrifugation at 13,000 x g at 4°C, cells supernatants (containing the pili) and the respective pellets (containing pili-less cells) were collected. The pili within supernatants were precipitated for 20 hrs with trichloroacetic acid (TCA) (final concentration, 25%), at −20°C. The precipitated pili were collected by centrifugation at 16,000 x g for 10 min at 4°C and resuspended in 50 μL 1X PBS. Total protein was extracted from pili-free cell pellets and whole cell lysates using Bugbuster reagent (Novagen) following the manufacturer’s instructions and quantified using a Bradford assay.
Purification of His-tagged A. baumannii RecA and Immunoblots
The recA gene from A. baumannii was amplified using primers RecALICF and RecALICR (Table S1) and cloned into the pET-His6-TEV-LIC vector (plasmid 29653; Addgene, Cambridge, MA, USA) using Ligation Independent Cloning Protocol (Gradia et al., 2017) and introduced by transformation into DH5α Escherichia coli cells for plasmid maintenance. The induction of expression and purification methods are as previously described for RecA purification from Escherichia coli (Tashjian et al., 2017). RecA purification was confirmed by SDS-PAGE.
For the semi-quantitative western blots, saturated cultures were diluted 1:100 and grown to exponential phase. Cells were then treated with ciprofloxacin (Cip) at 10X the MIC for 3 hrs (Ching et al., 2017). A parallel culture for each strain was left untreated. Cells were collected by centrifugation and whole-protein lysate was extracted using Bugbuster reagent (Novagen) and Pierce Universal Nuclease according to manufacturer’s instructions. Cell free lysates were quantified with a Bradford assay following manufacturer’s instructions (BioRad). Lysate was diluted to 1X with Laemmli buffer and heated for 10 minutes at 95°C. The samples were separated on a 12% Bis-Tris gel (Invitrogen) in MOPS buffer and transferred to a nitrocellulose membrane (Cafarelli et al., 2013). The blot was developed as before (Cafarelli et al., 2013) using a primary anti-RecA antibody (Abcam) at a 1:10,000 dilution (Norton et al., 2013) and an HRP-labelled secondary goat anti-rabbit antibody (Abcam) diluted at 1:40,000. The density of the RecA bands was quantified by ImageJ (Norton et al., 2013). The RecA signal from all samples were compared to a standard of purified RecA protein from A. baumannii. Similar loading of lanes was measured by Commassie Blue staining of a parallel gel. The procedure was done in 4 biological replicates. RecA determinations were similar every time.
To measure the relative amount of Csu pili in the different strains, we performed immunoblots with an anti-Csu antibody (gracious gift from Luis Actis). The pili content of each strain was standardized to the total protein measured in the cell-free extracts from which the pili were isolated. To detect Csu pili from whole cell lysates, similar number of cells were lysed with Bugbuster and Pierce Universal Nuclease as indicated above. The samples were separated in a 12% polyacrylamide gel in MES buffer at 150V for 1 hour. Immunoblot procedures were followed as above using a 1:1000 dilution of the anti-Csu primary antiserum and a 1:40,000 dilution of the HRP labeled secondary anti-rabbit antibody (Abcam). A ChemiDoc MP imaging system (BioRad) was used for chemiluminescence signal detection from RecA and Csu blots.
Construction of the Pcsu-gfp and Pcsu-mCherry expression reporter plasmids and expression assays
A. baumannii ATCC 17978 containing a csu operon (Pcsu-gfp) transcriptional reporter on the Carbenicillin (Carb) resistant plasmid pTU1-A-AB (Moore et al., 2016) was used to assess the expression of Csu in the WT, recA++ and ΔrecA strains. The Pcsu-gfp reporter plasmid was made utilizing a modified version of the EcoFlex Kit (Moore et al., 2016). The EcoFlex kit was a gift from Paul Freemont (Addgene kit #1000000080). All level 1-3 plasmids (Addgene accession numbers 72935-72946) were digested with PstI and dephosphorylated using Quick CIP (both New England Biolabs, Ipswich, MA.). PstI restriction sites were attached to the A. calcoaceticus’ native pWH1266 backbone by PCR amplification with primers ori_ab_pstI_r and ori_ab_pstI_f (Table S1). The amplicon was further digested with PstI and ligated using T4 DNA ligase (Thermo Fisher Scientific) into the linearized level 1-3 plasmids. 5 μL of the ligation mix was introduced by transformation into chemically competent E. coli DH5α. Positive clones were confirmed by a digest with PstI and BsaI (New England Biolabs, Ipswich, MA.). Afterwards, the promoter region of the csu operon was amplified using primers csu_f and csu_r. The pBP_lacZ (Table S1; Addgene accession number 72948) and the purified PCR fragment were digested using SphI and NdeI and the pBP-lacZ plasmid dephosphorylated. The ligation was carried out with T4 DNA ligase (Thermo Fisher Scientific). Colonies were screened using blue-white color in plates with X-Gal and verified by colony PCR and sequencing using Primers pBP_pTU_f and pBP_r.
Plasmids with the expected inserts were then used in a combined cloning and ligation reaction, by using 20 fmol of each entry vector (pBP_csu, pBP_B0012, pBP_egfb, pBP_pET-RBS and pTU1-A-RFP-AB) combined with 2 μL of Cutsmart, 0.5 μL of BsaI-HFv2, 0.5 μL of T4 DNA Ligase and 1 μL of 10 mM ATP (all reagents from New England Biolabs, Ipswich, MA., except T4 DNA Ligase from Thermo Fisher Scientific). The reaction was cycled 50 times in a thermocycler with a 37 °C digestion step for 2 minutes and a 16 °C ligation step for 5 minutes. Any plasmid with no insert is eliminated from the mix by digestion with BsaI at 37 °C for 1 h. The enzymes were heat-inactivated at 80°C for 10 min. 5 μL of the ligation mix was introduced by transformation into chemically competent E. coli DH5α. Colonies with the expected fragment were screened by red-white color in which white colonies contain the desired insert due to the exchange of the mCherry gene originally in the pTU1-A-AB cloning site and confirmed by colony PCR and sequencing the insert using pBP_pTU_f and pTU_r. Plasmids with the desired constructs were introduced by transformation into A. baumannii utilizing standard electroporation protocols(MacGuire et al., 2014). To construct the mCherry csuAB reporter, (Pcsu-mCherry) same primers and approach were used except that the plasmid pBP_ORF-mCherry was used instead.
Microscopy and GFP reporter measurements
To perform confocal microscopy experiments we used a Zeiss Axio Observer.Z1/7. The excitation/emission used were 280/618 nm for the red channel, and 488/509 nm for the green channel. We formed biofilms in wells, and after 48 hrs, 0.5xMIC Cip was added underneath pellicles and dishes were incubated statically for an additional 48 hrs. Pellicles were collected from mock treated and Cip-treated wells on a glass coverslip treated with 0.1 mg/mL of polylysine to facilitate attachment. The side of the pellicle in direct contact with the liquid medium was in contact with the glass, and the side of the biofilm in contact with air remained as such. Z-stacks were of 1μm.
For the GFP reporter measurements, GFP fluorescence was measured in a plate reader (Biotek synergy H1) for at least 24 hrs. at 25°C without shaking. To start the experiment, cultures were diluted 1:100 in YT medium in a 96 multi-well dish. Green fluorescence (Excitation 479 nm and emission 520 nm) and OD600 were measured every 15 mins. Growth and fluorescence data were gathered at least in triplicate for each strain.
MIC and Biofilm Eradication Assays, and Live/Dead determination
Standard assays were used to test for MIC (Andrews, 2001) in planktonic shaking cultures. Pellicles were set up as described above and incubated statically at 25°C for the number of hours indicated in the respective figure legends, at which time cells that were not stuck to the wells’ surface were removed, and wells washed three times with 1X PBS. YT medium containing Gm, at a range of concentrations shown in the respective figures was added to each washed well. The plate was then incubated statically for 24 hrs. at 25°C, after which the spent medium was removed, and wells were washed three times with PBS. Crystal Violet staining was then carried out as before (Chen et al., 2015). LIVE/DEAD staining was performed on resuspended surface-attached cells, and microscopy performed as before (Ching et al., 2018). The LIVE/DEAD stained surface attached cells of the WT, ΔrecA, and recA++ after Gm treatment (10X MIC) were mechanically dislodged. This methodology did not affect cells’ integrity. The number of total cells obtained from the surface-attached fraction varied from strain to strain due to their difference in ability to attach to surfaces. The total number of cells and their viability was determined by counting 5 independent microscope fields from each of the strains.
Competing Interests
The authors declare no competing interests.
Author Contributions
V.G.G., and C.C. initiated research in discussions with YC. B. I. performed preliminary experiments. C.C., P.M., A.R., M. B., B. N., S. R., M.D., W. F., and V.G.G. performed experiments and analyzed data. C.C. and V.G.G. wrote the manuscript.
Funding
M.D. and S.R. were funded by the Undergraduate Research and Fellowships office at Northeastern University. V.G.G. is funded by a stipend from NUSci, an Inclusive Excellence award form HHMI and the NSF REU site award #1757443, Y.C. is funded by an NSF grant (MCB1651732), and A.R. by Northeastern University Provost Dissertation Completion Fellowship.
Acknowledgments
We would like to acknowledge members of the Godoy, Chai, and Geisinger labs at Northeastern University for helpful comments and discussion. We would like to thank G. Bou from Servicio de Microbiologia, Complejo Hospitalario Universitario, La Coruña, Spain for the ΔrecA strain, Luis Actis from the Department of Microbiology, Miami University, for the anti-Csu antibody, and Stephen Lory from Harvard Medical School for access to the AB7505 transposon collection.