Insights Into Mechanisms of Biofilm Formation in Acinetobacter baumannii and Implications for Uropathogenesis

Multidrug resistant Acinetobacter baumannii is a serious healthcare threat. In fact, the Center for Disease Control recently reported that carbapenem-resistant A. baumannii is responsible for more than 8,500 infections, 700 deaths, and $281 million in healthcare costs annually in the United States with few, if any, treatment options available, leading to its designation as a pathogen of urgent concern and a priority for novel antimicrobial development. It is hypothesized that biofilms are, at least in part, responsible for the high prevalence of A. baumannii nosocomial and recurrent infections because they frequently contaminate hospital surfaces and patient indwelling devices; therefore, there has been a recent push for mechanistic understanding of biofilm formation, maturation and dispersal. However, most research has focused on A. baumannii pneumonia and bloodstream infections, despite a recent retrospective study showing that 17.1% of A. baumannii isolates compiled from clinical studies over the last two decades were obtained from urinary samples. This highlights that A. baumannii is an underappreciated uropathogen. The following minireview will examine our current understanding of A. baumannii biofilm formation and how this influences urinary tract colonization and pathogenesis.


INTRODUCTION
Acinetobacter baumannii is a public health menace recently rising to prominence due to the rapid increase in antibiotic resistance and infection rates. Infections caused by A. baumannii account for ∼2% of all healthcare-associated infections in the United States and Europe (Sievert et al., 2013;Magill et al., 2014;Lob et al., 2016) and this rate is nearly doubled in Asia and the Middle East (Lob et al., 2016). Globally, it is estimated that nearly 45% of all A. baumannii isolates are multidrugresistant (MDR; resistant to ≥3 antibiotics) with rates as eclipsing 70% in Latin America and the Middle East (Giammanco et al., 2017). We have reached a critical tipping point where antibiotic discovery cannot keep up with the rapidly evolving antibiotic resistance of A. baumannii without some type of intervention. Hence, the World Health Organization (WHO) and Centers for Disease Control (CDC) have signified A. baumannii as a pathogen of critical importance for the discovery of novel antimicrobials (WHO, 2017;CDC, 2019).
Acinetobacter baumannii primarily causes infections of the lung or bloodstream (Peleg et al., 2008). However, it was recently reported that up to one-fifth of all A. baumannii isolates are obtained from urinary sources, implying that this organism is an underappreciated uropathogen (Di Venanzio et al., 2019). Catheter-acquired urinary tract infections (CAUTI) are one of the most common hospital-acquired infections accounting for an estimated 100,000 infections annually in the United States (Zarb et al., 2012;Magill et al., 2014). It is hypothesized that bacterial biofilm formation along the catheter surface is the most important factor in the establishment of bacteriuria (Stickler, 2008). Acinetobacter baumannii's increasing prevalence in CAUTIs is due to its adept ability to form biofilms, with an estimated >75% of all isolates capable of forming a biofilm (Abdi-Ali et al., 2014;Azizi et al., 2016;Thummeepak et al., 2016). Therefore, understanding the mechanisms responsible for A. baumannii biofilm biogenesis and maturation are critical for elucidating the basis for uropathogenesis and may help with the development of future CAUTI anti-biofilm therapies. The following minireview examines existing data focused on the genetic regulation of A. baumannii biofilm lifestyle and its contribution to uropathogenesis as well as identifies current knowledge gaps to be addressed moving forward.

BIOFILM FORMATION Bacterial Cell Adherence
The initial step involved in the shift from planktonic to biofilm formation is surface contact and irreversible attachment (reviewed in Petrova and Sauer, 2012;Armbruster and Parsek, 2018). Acinetobacter baumannii has the ability to form biofilms on a wide range of surfaces including abiotic surfaces, like stainless steel and polypropylene, as well as host epithelial cells (Greene et al., 2016). Many virulence factors have been implicated in bacterial cell adherence, however the plasticity observed in A. baumannii genomes leads to significant strainspecific variations in biofilm formation. Investigation into the presence of known biofilm-associated genes in A. baumannii clinical isolates across several publications (Loehfelm et al., 2008;Badmasti et al., 2015;Zeighami et al., 2019) has shown that the most highly conserved genes were CsuE, the proposed tip subunit of the chaperone-usher pili (Csu), and OmpA (reported 81-100% detection). For the biofilm-associated protein (Bap) and class A extended β-lactamase blaPER-1 enzyme, detection was variable ranging from 30-66% to 2-64% of isolates, respectively. The Csu assembly system is composed of pilin subunits CsuA/B, CsuA, CsuB, and CsuE and transport proteins CsuC and CsuD, is highly conserved in biofilm-forming isolates and critical for adherence to abiotic surfaces, but not host surfaces (Tomaras et al., 2003;de Breij et al., 2009).Outer membrane protein A (OmpA) is a prominent porin that contributes to drug resistance, adhesion to epithelial cells and biofilm formation on plastic surfaces (C.H. Choi et al., 2008;. Anti-OmpA serum and antibodies blocked A. baumannii's adherence and subsequent invasion of host cells (Schweppe et al., 2015). Biofilmassociated protein (Bap) is a surface-exposed, highly divergent protein that is required for adherence to bronchial cells and structural integrity and water channel formation within the biofilm (Loehfelm et al., 2008;Brossard and Campagnari, 2012;De Gregorio et al., 2015). One study found that disruption of the Bap gene led to significant reductions in biofilm thickness and volume, interbacterial cell adhesion and ability to form higher order structures on medically relevant abiotic surfaces (Loehfelm et al., 2008). Another recent study found that the variation in the bap coding sequence across A. baumannii lineages results in differential functions during biofilm development with some versions displaying better adherence properties and others forming more complex biofilms (Skerniskyte et al., 2019). β-lactamase blaPER-1-expressing strains displayed significantly increased cell adhesiveness and biofilm formation compared to strains lacking the β-lactamase (H.W. Lee et al., 2008). However, additional publications report no or limited correlation between blaPER-1 expression and biofilm formation (Sechi et al., 2004;Rao et al., 2008); thus, more research is required to elucidate its role.
Other virulence factors that have been implicated in adherence and biofilm formation include Pap, Prp, Cup, and Type IV pili systems as well as Acinetobacter trimeric autotransporter (Ata) (reviewed in Eijkelkamp et al., 2014;Longo et al., 2014;Harding et al., 2018). The pap operon encodes proteins homologous to P pili in E. coli, which has been found to be important for migration of bacteria from the bladder to the kidney (Wullt et al., 2000). The prpABCD operon encodes a photoregulated pilus associated with light-regulated motility and biofilm formation in ATCC 17978 (Wood et al., 2018). In addition, this operon is conserved in several other A. baumannii strains, including the hyper-biofilm forming MAR002, which displayed a 25-fold increase in the prpD homolog in sessile cells (Alvarez-Fraga et al., 2016). CUP2 pili were recently discovered as a prp operon homolog in UTI pathogen UPAB1, which when deleted resulted in reduced adhesion to both the catheter surface and bladder lumen in a CAUTI murine model (Di Venanzio et al., 2019). Type IV pili, encoded by the pil operon, have been shown to play a role in adhesion to cells and stainless steel (Ronish et al., 2019). Ata is a surface-exposed protein that has been shown to play an important role in biofilm formation as well as adherence to host cells and various host extracellular components (Bentancor et al., 2012;Weidensdorfer et al., 2019).

Biofilm Formation Cues and Detection
Following adhesion to a surface, the bacterial cells are now primed to continue the shift to the biofilm state. The next step in biofilm formation involves environmental signal sensing and signal transduction, which will lead to downstream cellular responses. Many signals and signaling components that have been implicated in the control of biofilm formation and virulence factor production in A. baumannii are described below.
Acinetobacter baumannii and its close relative Acinetobacter nosocomialis have one quorum sensing (QS) system, which plays an integral role in regulating virulence factors, biofilm formation and surface motility (Niu et al., 2008;Clemmer et al., 2011;Bhargava et al., 2015;Subhadra et al., 2019). AbaI is the autoinducer synthase that generates the QS molecule N-(3-hydroxydodecanoyl)-L-HSL (AHL), which at high enough density interacts with the cognate receptor AbaR leading to downstream cellular responses. Several publications have found that AbaI and AbaR gene disruption leads to reduced biofilm formation (Niu et al., 2008;Anbazhagan et al., 2012;Guo and Xiang, 2017). Furthermore, cells cultured in the presence of AHL exhibited increased expression of Csu pili and stimulation of biofilm formation (Luo et al., 2015). Additionally, the activity of AbaI and biofilm production are regulated by iron in a dosedependent manner (Modarresi et al., 2015), suggesting that iron is a possible environmental signal for nutrient limitation and the shift to survival mechanisms.
Several two component systems (TCS) have been shown to play a critical role in biofilm formation. BfmRS is predicted to contribute to the enhanced biofilm formation on abiotic surfaces since a knockout mutant of bfmS displayed drastic reduction in biofilm formation, adherence to eukaryotic cells and serum killing resistance compared to the wildtype strain (Liou et al., 2014). Furthermore, the csu operon is regulated by BfmRS, suggesting that the TCS plays an integral role in the initial adhesion step of biofilm formation (Tomaras et al., 2008;Shin et al., 2009). AdeRS is another TCS implicated in biofilm formation because an adeS deletion mutant resulted in decreased biofilm formation (Richmond et al., 2016). GacSA TCS was initially discovered for its role in citrate metabolism (Dorsey et al., 2002). However, further characterization of a gacS deletion mutant revealed its involvement in the control of pili synthesis, motility, biofilm formation, resistance against human serum, and metabolism of aromatic compounds by the paa operon (Cerqueira et al., 2014). Finally, A1S_2811 is a recently characterized hybrid sensor kinase expressed in an operon with pilGHIJ genes, suggesting a potential link to adhesion. Further, the A1S_2811 deletion mutant displayed a significant reduction in surface motility, pellicle formation and abaI protein (Chen et al., 2017), suggesting a second putative control mechanism associated with QS.
Many other signals and sensing systems have been recently implicated in biofilm formation by A. baumannii. One study showed that cyclic di-GMP may play a role in A. baumannii biofilm formation since small molecule inhibitors of diguanylate cyclase enzymes (DGC) significantly reduced biofilm density (Sambanthamoorthy et al., 2014). Furthermore, another publication identified 2 DGCs that control biofilm and pellicle formation (Ahmad et al., 2020). When these enzymes are overexpressed, it drives early poly-N-acetyl-β-(1-6)-glucosamine (PNAG) production, which is an important biofilm extracellular matrix component. Temperature influences biofilm robustness since 26 • C biofilms displayed significantly increased biofilm mass compared to 30 and 37 • C (Eze and El Zowalaty, 2019). Mussi et al. showed that A. baumannii senses and responds to blue light as motility and biofilm formation were only observed in cultures grown in darkness, with the responsiveness level influenced by temperature (Mussi et al., 2010). The predicted photoreceptor protein is conserved in other A. baumannii strains, suggesting that light sensing is a potential widespread cue in Acinetobacter species. Deletion of A1S_0114 displayed an increase in csuAB expression as well as a decrease in other pilin proteins and ompA . Further, this mutant was unable to form complex 3D biofilm structures on abiotic surfaces and reduced airway epithelial adhesion. Recently, a Zur-regulated lipoprotein ZrlA was described to be involved in biofilm formation and motility through BfmRS signaling and subsequent control of csu expression (E.K. Lee et al., 2020).

Transcriptomic and Proteomic Changes
Several studies have compared the transcriptomic and proteomic profiles of A. baumannii grown in various growth conditions, including exponential, late stationary, pellicle and biofilm states, to elucidate the functional and metabolic differences between various bacterial lifestyles (Shin et al., 2009;Cabral et al., 2011;Marti et al., 2011;Chopra et al., 2013;Rumbo-Feal et al., 2013;Han et al., 2014;Kentache et al., 2017;Li et al., 2017;Penesyan et al., 2019). To gain more insight into the differential cellular response associated with biofilms, we compiled transcriptional and proteomic data reported from 9 publications, focusing specifically on up-regulated genes in biofilm/pellicle states compared to exponential growth ( Table 1, Supplementary Tables 1, 2). Our efforts evaluated a total of 854 reported up-regulated genes (473 up-regulated transcripts and 381 up-regulated proteins) across 7 different A. baumannii strains (ATCC 17978, AB5075_UW, A077, A061, A132, 1656-2, BJAB0868), 3 of which were isolated from urinary sources (A077, A061, A132). Overall, we found 132 up-regulated genes to be corroborated between independent strains and/or separate publications ( Table 1). Seventy-six genes were confirmed by both transcriptional and proteomic data with 43 of those genes validated across at least 2 different A. baumannii strains. Further, 35 and 21 genes were verified by at least two independent collections of transcriptional data and proteomic data, respectively. We further broke down these 132 up-regulated biofilm genes into basic biological function categories: Outer membrane proteins, Attachment/Motility, Metabolism, Transcription, Translation, and Hypothetical proteins ( Table 1 and Figure 1). The largest represented categories were metabolism (49 genes), translation (30 genes), and outer membrane proteins (29 genes). This suggests that the transition and maintenance of the biofilm state involves significant changes to metabolic processes and outer membrane composition supported by translational machinery required to produce nascent proteins.
The most highly upregulated genes observed in biofilm associated cells were the csu operon (csuABCDE) exhibiting overexpression levels ranging from 11-to 205-fold increase over exponential phase cells. The other attachment/motility genes identified to be upregulated include pili genes filF, fimA, and papCE. In the metabolic category, we observed the significant upregulation of the phenylacetate degradation operon (paaZABCEFJ), which has been linked to neutrophil evasion and regulation by the GacS/GacA TCS (Cerqueira et al., 2014). Of the 30 translational genes up-regulated, 24 are components composing the small and large subunits of the ribosome, suggesting an overall increase in translational capacity within biofilm cells. In the outer membrane protein category, we observe significant increases in RND efflux pump proteins and iron       ATCC 17978 gene name, gene IDs, and gene description are listed first by combination of confirmed data with the first section presenting genes confirmed by both transcriptional and proteomic data sets, the second section confirmed only transcriptionally and the third section confirmed only proteomically. The up-regulated genes are then broken down into functional subcategories: outer membrane proteins, adhesion/motility, metabolism, transcription, translation, and hypothetical proteins (visual representation of Table 1 is presented in Figure 1). In the transcriptional data column, we compiled (A) fold change of the transcript in biofilm cells compared to exponential phase control cells, (B) the A. baumannii strain tested and (C) the publication from with the data was obtained. In the proteomic data column, we compiled (D) fold change or protein spectra match (PSM) of the protein in biofilm/pellicle cells compared to exponential phase control cells, (E) the A. baumannii strain tested and (F) the publication from with the data was obtained. If there is more than one publication reporting up-regulation of the same gene, the fold-changes, strains, and publications are separated by commas (,) unless the same strain was used across multiple publications. The first listed fold-change number reported corresponds to the first listed strain and publication and so on. All compiled data, including more information on the original gene ID call and biofilm/pellicle growth conditions, can be found in Supplementary Tables 1, 2. **We were not able to integrate all the transcriptional changes observed in Li et al. (2017) since the reported gene code did not align to any sequenced A. baumannii genome. We selected genes that had clear gene annotations relevant to the compiled data set.
FIGURE 1 | Functional categorization of corroborated genes up-regulated in biofilms. Each stacked bar represents the number of corroborated genes confirmed by transcriptional, proteomic, transcriptional and proteomic data sets (as listed in Table 1). The last stacked bar represents the number of genes confirmed in at least two different strains of A. baumannii regardless of data set. The total number of genes compared in each data set are listed after the bar title (i.e., 473 up-regulated transcripts compiled and compared, 381 up-regulated proteins compiled and compared, and so on). Each stacked bar is further broken down into functional subcategories: outer membrane proteins (blue), adhesion and motility (orange), metabolism (gray), transcription (yellow), translation (purple), and hypothetical proteins (green).
acquisition systems, which are important for intrinsic antibiotic resistance and nutrient procurement.
Though not confirmed in our gene list, the pgaABCD operon encodes the enzymes that produce PNAG, an important structural component for biofilm formation (A.H. Choi et al., 2009). Further, it has been shown that expression of pgaB is positively correlated with biofilm formation capacity in clinical isolates from burn wound infections in Iran (Amin et al., 2019). It is clear that the A. baumannii growth state results in different transcriptional, proteomic, and metabolic profiles, which account for variable cellular responses.

Recent Developments
As mentioned previously, A. baumannii has only just begun to be recognized as an important uropathogen. A recent study discovered that a large conjugative plasmid (pAB5) in the MDR A. baumannii urinary isolate UPAB1 increases virulence in a firstof-its-kind CAUTI murine model of infection (Di Venanzio et al., 2019). Furthermore, UPAB1 grew better than ATCC 19606 in pooled human urine in vitro and co-localized with fibrinogen similar to previous observations in common UTI pathogens such as E. faecalis and MRSA (Walker et al., 2017;Xu et al., 2017). To identify adhesins involved in colonization of the bladder, Di Venanzio et al. identified two loci encoding putative CUP pili (CUP1 and CUP2). Deletion of these operons revealed loss of distinct surface appendages observed in the wildtype control and reduction in bacterial burden both on the catheter implant and within the bladder. Further, loss of pAB5 resulted in significantly reduced bacterial burden on the implant and within the bladder; however, the presence of pAB5 attenuated virulence and dissemination to other organs in an acute pneumonia murine model, which led the researchers to conclude that pAB5 confers niche specificity. To identify potential virulence factors differentially regulated by pAB5, researchers utilized proteomic and transcriptional approaches. Overall, their data indicated that pAB5 repressed type VI secretion system and differential regulation of PNAG biosynthesis and CUP1/2 pili are influenced by growth condition; thus, indicating that plasmid-encoded genes may influence biofilm formation and uropathogenesis by modulating the expression of chromosomal genes.
Another recent publication supports the hypothesis of niche-specific plasmid acquisition. They found distinct genome expansions in strains isolated from the similar sites of infections whereas strains isolated from another site of infection maintained different plasmids (Yakkala et al., 2019).
Given the wide-ranging phenotypic changes observed during the transition from planktonic to biofilm growth, it is likely that there are many levels of regulation involved in coordinating the cellular response. In recent years, the role of small RNAs (sRNA) in transcriptional regulation networks have been increasingly recognized. To this end, Alvarez-Fraga et al. compared the expression of sRNAs in ATCC 17978 biofilm cells and found 60 sRNAs were differentially regulated compared to planktonic cells . Additionally, they were able to show that sRNA 13573 is involved in the biofilm formation and attachment to eukaryotic cells, suggesting that biofilm biogenesis and adhesion properties in ATCC 17978 are coordinately regulated. Interestingly, another group found a distinct set of differentially expressed sRNAs in A. baumannii strain MTCC1425 compared to ATCC 17978, suggesting that the sRNAs involved in transcriptional control display some strain specificity (Sharma et al., 2014). Mangas et al. compared nearly 2000 A. baumannii genomes. They observed that strains carrying CRISPR systems were enriched for biofilm-associated genes (>70 vs. <2% non-CRISPR strains), suggesting a link between CRISPR immunity and biofilm formation (Mangas et al., 2019). Previous research has shown that Cas3 endonuclease is involved in the control of biofilm formation in both gram-positive and gram-negative bacteria (Tang et al., 2019;Cui et al., 2020).

PERSPECTIVES
While investigations into the mechanisms behind A. baumannii biofilm formation and CAUTI-associated pathogenesis have expanded recently, there remains many questions left to be addressed in order to produce a fully developed model.
A general concern across all pathogenic organism studies is that in vitro assays have been important for identifying virulence factors responsible for pathogenesis. However, studies within animal models of these putative virulence factors have often lacked direct correlation with in vivo outcomes, including in A. baumannii studies (Wand et al., 2012;Giannouli et al., 2013;Zimbler et al., 2013;Beceiro et al., 2014;Lazaro-Diez et al., 2016). These results highlight the importance of the confirmation of virulence in vivo, especially in models reflecting human infection. The first CAUTI murine infection model was recently established and requires more investigation for validation (Di Venanzio et al., 2019), but is a good first step in addressing this concern.
Another major complication that is evident across the array of A. baumannii pathogenesis publications is that some of the biological roles associated with identified virulence factors seem to be strain specific. For example, Wood et al. described and characterized a light-regulated pilus system involved in ATCC 17978 biofilm formation; however, this operon displayed no changes in expression in the hyper-biofilm producing strain MAR002 (Alvarez-Fraga et al., 2016;Wood et al., 2018). Further, Eze and El Zowalaty observed significant strain variation in biofilm formation across strains tested under differing temperatures, nutrient levels and agitation conditions (Eze and El Zowalaty, 2019). Future work should investigate conservation and incorporate several different A. baumannii lineages to strengthen the original discovery.
One observation we encountered while compiling upregulated genes involved in biofilm cell growth is the wide variation in methods used to measure biofilm formation (Shin et al., 2009;Cabral et al., 2011;Marti et al., 2011;Rumbo-Feal et al., 2013;Nait Chabane et al., 2014;Alvarez-Fraga et al., 2016;Kentache et al., 2017;Li et al., 2017;Penesyan et al., 2019). Publications reported using different incubation times (24-144 h), incubation temperatures (25-37 • C), abiotic surfaces supplied (glass, polystyrene), and growth conditions (continuous flow, stationary). While we were able to identify a large set of genes up-regulated in biofilm cells despite differential growth conditions, we are concerned that many other genes may have been missed in these studies. For example, previously reported biofilm-associated genes, bap and the pga operon, were not reported to be up-regulated in any publication examined. Moving forward, transcriptional and proteomic profiling over time during biofilm formation and maturation will provide important information into the dynamic, rapidly transitioning cellular responses within sessile cells.
Recently, a novel, phase-variable colony opacity switch has been described in AB5075 and other A. baumannii clinical isolates, in which colonies interconvert at a high-frequency between opaque and translucent variants (Tipton et al., 2015). Further characterization of the two opacity forms showed significant differences in biofilm formation, virulence and transcriptional profiles (Chin et al., 2018). However, none of the publications discussed in this review mentioned focusing a specific phase variant, which likely means their results were generated from a mixed pool of cell types. This implies that transcriptional and proteomic data may be missing important differences since the average of the mixed population may match control even though one subpopulation could have the gene significantly up-regulated and the other subpopulation has the gene significantly down. This leads us to wonder what the individual contributions that each of the phase types have in biofilm formation and maturation. It is important to note that we have not observed colony opacity variation in ATCC 17978 and therefore, this phenotypic variation may not apply to studies using this strain.
Finally, there are nearly 2500 A. baumannii genome sequences publicly available comprising a core genome of ∼2,200 genes and a collapsed pan-genome size of almost 20,000 genes (Chan et al., 2015;Mangas et al., 2019), showing the broad variation across this pathogen. Furthermore, 42% of the pan-genome is of unknown function displaying our superficial knowledge of the roles these genes play in A. baumannii growth, virulence and environmental adaptability. Overall, the accumulation of decades of research has revealed many genes that are involved in the transition from planktonic growth to biofilms in A. baumannii. Only recently has this organism begun to be appreciated as a uropathogen and research into this area has commenced. Many more studies are required to fully understand how biofilmassociated genes may contribute to urinary tract infection. As we gain more insight into the underlying mechanisms of biofilm formation and uropathogenesis, this work will lay the foundation for potential anti-infective targets to combat surmounting obstacle of MDR A. baumannii.

AUTHOR CONTRIBUTIONS
JC conceived and wrote the majority of the manuscript. PR contributed to manuscript revision. Both authors read and approved the submitted version.

FUNDING
This work was supported by funding from the Department of Veterans Affairs I01 BX001725, IK6BX004470 and NIH R21AI142489 to PR.