Identification and Characterization of Differentially-Regulated Type IVb Pilin Genes Necessary for Predation in Obligate Bacterial Predators

Bdellovibrio bacteriovorus is an obligate predator of bacteria that grows and divides within the periplasm of its prey. Functions involved in the early steps of predation have been identified and characterized, but mediators of prey invasion are still poorly detailed. By combining omics data available for Bdellovibrio and like organisms (BALO’s), we identified 43 genes expressed in B. bacteriovorus during the early interaction with prey. These included genes in a tight adherence (TAD) operon encoding for two type IVb fimbriae-like pilin proteins (flp1 and flp2), and their processing and export machinery. Two additional flp genes (flp3 and flp4) were computationally identified at other locations along the chromosome, defining the largest and most diverse type IVb complement known in bacteria to date. Only flp1, flp2 and flp4 were expressed; their respective gene knock-outs resulted in a complete loss of the predatory ability without losing the ability to adhere to prey cells. Additionally, we further demonstrate differential regulation of the flp genes as the TAD operon of BALOs with different predatory strategies is controlled by a flagellar sigma factor FliA, while flp4 is not. Finally, we show that FliA, a known flagellar transcriptional regulator in other bacteria, is an essential Bdellovibrio gene.

retraction of type IVa pilus, an appendage shown to be essential for predation in both B. bacteriovorus and in the epibiotic predator, Bdellovibrio exovorus 17,18 . Type IVa pili usually govern twitching motility in other bacteria 19 , but in B. bacteriovorus they are assumed to facilitate invasion into the prey 20 . Other predation-essential genes encoding for proteases, regulatory proteins, signal transduction related proteins and flagellar motility, and genes important for cellular structure and organization were identified by transposon mutagenesis performed in HI genetic backgrounds, combined with a screen for alterations in predatory phenotype, or by systematic homology searches [12][13][14]18 .
In this study, we applied computational integrated genome and transcriptome analyses to identify genes essential for predation, based on the rationale that the AP and the earlier steps of prey invasion are unique to obligate predators. A core of 43 AP genes was identified, including genes in an operon coding for type IVb fimbriae-like protein (flp) genes 21 that was further characterized. Further protein domain sequence analyses uncovered additional flp B. bacteriovorus genes. Genetic and phenotypic analyses showed that three of these four flp IVb genes are essential for predation and although they are expressed during AP, they are under different transcriptional regulation: flp1 and flp2 are co-regulated with flagellins by the flagella sigma factor FliA, while flp4 is regulated independently by an unknown regulator.

Results
Identification of putative attack phase core genes. Successful initial recognition and attachment to the prey by Bdellovibrio depends on factors expressed during the AP 22 . In order to identify genes that are involved in the initial step of predation, we searched for known B. bacteriovorus AP genes that are present in other diverse BALOs. Four hundred twenty one B. bacteriovorus HD100 genes that are expressed in AP 23 were compared to three fully sequenced Bdellovibrio genomes (Bdellovibrio exovorus JSS 24 , Bdellovibrio sp. Arhs 25 and Bdellovibrio bacteriovorus W (https://www.ncbi.nlm.nih.gov/assembly/GCA_000525675.1), three fully sequenced Bacteriovorax strains (Bacteriovorax SEQ25_V, Bacteriovorax BAL6_X, Bacteriovorax BSW_11IV 4 and the fully sequenced Halobacteriovorax marinus SJ 26 . B. bacteriovorus HD100 and B. bacteriovorus W are periplasmic fresh water BALOs, Bdellovibrio sp. ArHS is a thermophilic BALO, and B. exovorus JSS is a fresh water epibiotic predator within the Bdellovibrionaceae. H. marinus SJ and the three Bacteriovorax strains are marine periplasmic BALOs belonging to the Bacteriovoracaceae 1,4,25,26 . Forty-three genes that are certain orthologs between the examined species that had high (0.2) BLAST score ratios (BSR) were identified in this computational screen. To further refine our screen, looking for additional evidence for evolutionary conserved genes 27 , we searched the 43 "AP core" gene set for members that are syntenous in all examined BALOs. Twelve of the 43 genes were found to be syntenous and present in four putative operons ( Fig. 1 and Table S1).
Most of the genes in the four putative operons encode for poorly characterized or unknown proteins. Genes bd0228 to bd0231 form a putative operon that encodes a serine protein kinase, an uncharacterized protein and a sporulation protein R, SpoVR. Similar operons were found in many non-sporulating and non-predatory bacteria Overlap of attack phase genes between strains of Bdellovibrio and like organisms, identifying putative predation-essential genes. 124 Bdellovibrionaceae and 55 Bacteriovoracaceae genes exhibit high similarity (BSR ≥ 0.2) to 421 B. bacteriovorus HD100 AP-induced genes. The overlap of the three sets depicts 43 AP core genes, 11 of which cluster in four B. bacteriovorus HD100 genomic regions. Each of these regions is a putative operon as indicated by RNAseq expression data (BdelloViewer, http://www.weizmann.ac.il/molgen/Sorek/ bdello_browser/viewer/index.php?meta=bdello_combined&jump=bdello,bdello_sRNAs#1 23 . The Venn diagram was created using Biovenn 69 .
from the beta-, gamma-and delta-proteobacteria. SpoVR is a non-essential gene that is involved in spore cortex formation in Bacillus subtillis and Myxococcus xanthus 28,29 . Genes bd1103 and bd1104 are in a putative operon and encode two conserved proteins of unknown function, also found in many bacteria. The putative operon bd3753-bd3754 encodes for ClpX and ClpP proteins respectively, forming the ClpXP complex that degrades misfolded proteins and thus assists in maintaining cell proteostasis 30 . Interestingly, the E. coli ClpXP complex is activated under carbon starvation, the physiological state of the attack phase 31,32 .
The syntenous region bd0109-bd0113 is part of a putative operon including genes bd0109-bd0119. It was previously annotated as a tight adherence locus (TAD) 33 . This operon encodes for proteins that process and export fimbriae like proteins (Flp) to the outer membrane. Flp proteins were found to be crucial for bacteria-host interaction in human, animal, and plant pathogens 34, 35 . B. bacteriovorus HD100 genome encodes four different flp pilin genes. The TAD locus operon encodes components essential for the maturation and export of Flp proteins 34 . The region includes genes coding for the ATPase that drives Flp export, TadA (bd0111); an inner membrane export component, TadB (bd0110); the outer membrane export proteins RcpA (bd0112) and RcpC (bd0113) 33,36 . However, TadV, a prepilin peptidase crucial for pilin maturation, is missing from the operon and was not found in the entire B. bacteriovorus genome. Further analysis indicates that TadV is missing from all known Bdellovibrio genomes, but is present in Halobacteriovorax and all of the Bacteriovorax genomes (Fig. 2).
Based on their genomic position within the TAD locus, their small size, and the hydrophobic character of the encoded proteins, two B. bacteriovorus flp pilin protein-encoding genes were previously identified (bd0119, flp1, and bd0118, flp2) 33 . We searched BALOs genomes for additional flp pilin genes, encoded outside the TAD locus. Flp pilins are short and not well conserved 37 . We thus searched not with specific flp pilin sequences (i.e., by BLAST searches) but with conserved amino acid motifs. We first constructed a position weight matrix (PWM) of conserved motifs from different Flp pilin protein sequences (Experimental procedures, Fig. 3). The PWM corresponded to a known conserved Flp proteins sequence domain, a GXXXXEY followed by a stretch of hydrophobic amino acids 37 . Searching the B. bacteriovorus genome with the PWM identified three putative flp pilin encoding genes: bd0119 (flp1), bd3297 (flp3) and an un-annotated open reading frame within the bd2719-bd2720 intragenic region that we designated as bd4000.
Searching other BALO genomes with the PWM revealed that all include similarly located flp pilin open reading frames, i.e. flp sequences in the TAD locus and others that are not TAD associated (Fig. 3, Fig. S1). While only one TAD-associated flp gene was found with the PWM in Bdellovibrio genomes, Bacteriovorax and Halobacteriovorax genomes include three TAD associated flp pilin genes. Interestingly, bd0118, annotated as flp2 in B. bacteriovorus 33 was not identified as a Flp pilin-encoding gene by our PWM. Sequence analysis demonstrated that Flp2 includes a trimmed Flp domain. The Flp2 sequence starts with MXXEY that is followed by a stretch of hydrophobic amino acids (Fig. 3). This domain differs from the consensus Flp domain as it misses four amino acids: the glycine residue that is recognized by the prepilin peptidase and the three residues C-terminal to it that typically are at the N-terminal end of the processed, mature pilin 38 . Consequently, while the Flp motif is usually 5-20 amino acids from the protein initiator methionine, in Bd0118 it directly follows it (Fig. 3). Genes encoding proteins with such a trimmed Flp motif were also observed within the TAD loci of other Bdellovibrio strains, but not in Bacteriovorax (Figs 2 and 3).
In order to exclude the possibility of mis-annotation of the coding region start position, we added its upstream 21 amino acids to the Bd0118 sequence and compared it to the PWM. No complete flp motif was found in Bd0118 or its orthologs in other BALO genomes. flp3 and flp4 were annotated as Flp proteins in additional Bdellovibrio strains based on the synteny of their surrounding genes with those of the B. bacteriovorus HD 100 flp3, and flp4 genes (Table S2).
A similar screen in four Bacteriovoracaceae genomes identified three TAD-associated and three non-TADassociated flp pilin genes in H. marinus SJ, Bacteriovorax Seq25_V and Bacteriovorax BAL6_X. Bacteriovorax BSW11_IV was the only exception as it included two flp genes from each kind.  (Fig. 4). Gene bd0199, which is only expressed during GP, served as a control. A protein analysis by mass spectrometry (MS) of wild type B. bacteriovorus cultures in which Bd0110, Bd0111, Bd0112, Bd0113 and the Bd0119 (Flp1) were detected, confirmed that TAD operon-encoded proteins were indeed produced in AP (Table S3). could not be deleted, as a screen of about 400 plaques of each for putative double recombinants did not yield any. For comparison, the flp3 deletion mutants were identified at a frequency of one per 20 plaques screened with a confirmation rate of 60%. This strongly suggests that both the typical flp1 and flp4 type IVb pilin genes, and the flp2 truncated Type IVb pilin gene, are essential for predation in a wild type background. To examine this hypothesis, these flp genes were deleted separately and also together in the background of HI M11.1, a B. bacteriovorus HD100 spontaneous host-independent type II mutant 16 . This mutant can grow axenically, in the absence of prey but retains predatory capabilities, albeit with reduced efficiency. This yielded the double mutant HIΔflp1-2, the triple mutant HIΔ flp1-2-4 and the three single mutant strains HIΔflp1, HIΔflp2, and HIΔflp4. These five mutant strains grew axenically, i.e. in the absence of prey on a rich PYE medium as expected from HI strains. However, none was able to prey on living E. coli cells, demonstrating their indispensability in the early stages of the predatory interaction (Fig. 5, Fig. S2). Finally, to confirm that the predation inability of these deletion strains was caused by the introduced deletions, the strains were complemented with the deleted genes under their native promoter by conjugating plasmids pPROBE-NT-flp1-2, pPROBE-NT-flp1, pPROBE-NT-flp2 and  Both TAD-associated flp1, flp2, and non-associated flp4 gene knockouts (Δflp1, Δflp2, and Δflp4, respectively) are unable to prey, while Δflp3 had no effect on predation. All strains regained the predatory phenotype of the parental strain (M11.1) upon complementation (comp.Δflp1, comp.Δflp2 and comp. Δflp4). In order to prevent polar effects, all knockouts were constructed as in-frame deletions, with only small non-functional peptides encoded in the remaining sequence (9-12 amino acids long). Complementation for each mutation was carried out using pPROBE-NT. Control strains carried an empty vector. pPROBE-NT-flp4 each to its corresponding deletion strain. Complementation resulted in complete restoration of the predatory HI phenotype (Fig. 5, Fig. S2), causing a four-order of magnitude decrease of the prey population. Deletion of flp3 in the HI M11.1 background did not yield any noticeable phenotypic change, confirming that, as in the wild type background and under laboratory conditions, flp3 is not required for predation. To address the possibility of functional complementation between flp genes, plasmid pPROBE::pflp1-flp3, bearing flp3 under the flp1 promoter was introduced in the HIΔflp1and in the HIΔflp4 backgrounds. No complementation was observed, as predation was not restored. flp1, flp2, and flp4 do not mediate binding to biotic or abiotic surfaces. We examined if the non-predatory phenotype of the flp deletions is due to a reduced capacity for surface adhesion, as is the case in other bacteria where flp genes mediate surface attachment [39][40][41][42][43][44][45] . For that purpose, biofilm formation by the parental HI strain, M11.1 and the type IVb pilins triple knockout strain ∆flp1-2-4, was examined. No significant reduction in biofilm formation was observed (Fig. 6).
The single flp1, flp2, and flp4 mutants as well as the triple flp1-2-4 knockout were then examined for direct binding to E. coli prey cells. No significant difference to the parental strain M11.1 was evident (Fig. 6). Direct observation under phase contrast microscopy showed that the Δflp1-2-4 mutant as well as each single flp mutant can bind to prey cells for significant lengths of time (>10 min).
The B. bacteriovorus HD100 and B. exovorus JSS TAD operons are regulated by the flagellar sigma factor FliA. A transcriptomic analysis of B. bacteriovorus AP and GP 23 suggested that the flagellar sigma factor, fliA (bd3318) acts as the main regulator of AP-expressed genes, including those encoded in the TAD operon. To demonstrate FliA control over TAD, we attempted to delete the B. bacteriovorus sigma factor FliA, encoded by gene bd3318. No mutants were obtained after screening 300 plaques of wild type HD100 and 300 colonies of a HI derivative. We then focused our efforts on the HI strain for technical convenience. As FliA is a central regulator of a cell cycle phase, we conjectured that a fliA knockout would be lethal or grow slowly. Accordingly, the knockout procedure was modified (see experimental procedures). No mutants were uncovered under any of the incubation conditions employed: colonies that appeared after 2 to 3 weeks of incubation were all found to be wild type. When low, 2% and 3% sucrose solutions were used for counter-selection, 30 to 50% of the colonies were merodiploid. No full recombinant was obtained. Therefore, the inability to knockout fliA led us to the conclusion that FliA is essential in B. bacteriovorus, unlike what was found in other bacteria 46 .
The flp1 promoter (i.e., promoter of the TAD operon) (Fig. 7, Fig. S3) was examined in a heterologous E. coli system: fliA coding regions (bd3318 or a11q_0617 for B. bacteriovorus and B. exovorus, respectively) were cloned into pBAD/gIIIA under an arabinose inducible promoter. Since the examined promoters were cloned upstream of a promoter-less β-galactosidase (see experimental procedures), recognition of these promoters by a sigma factor, such as FliA, should result in a measurable β-galactosidase biochemical activity. In other gram-negative bacteria, flagellin-encoding genes are regulated by FliA 46 , as was also predicted for B. bacteriovorus 11,23 . Therefore, B. bacteriovorus HD100 and B. exovorus JSS flagellin (bd0606 and a11q_1936 respectively) promoters were chosen as positive controls. β-galactosidase under the TAD operon (that includes flp1 and flp2) or under the regulation of flagellin promoters was strongly induced by FliA, as compared to an empty pBBR1-MCS2-LacZ vector. This validated the hypothesis that FliA regulates expression of flp1 and of all the other AP genes present with it in the TAD operon (Fig. 7). In contrast, the flp3 and flp4 promoters from both B. bacteriovorus HD100 and B. exovorus JSS remained silent under these conditions (Fig. 7).

Discussion
Genes essential for predation by B. bacteriovorus have been previously identified by direct random transposon mutagenesis, screening for mutants defective in predation 12,13 , or by targeting B. bacteriovorus genes whose orthologs in other bacteria encode functions that could potentially affect predation 14 . In this work, we aimed to identify genes that are essential for predation by examining early predation stages of the B. bacteriovorus life cycle. Genomic data have so far been used to differentiate between predatory and non-predatory bacteria, between predators exhibiting an epibiotic versus a periplasmic strategy, and to delineate the core genes of periplasmic predators 4, 24, 26, 47 . By integrating BALOs genomic and transcriptomic data we identified a set of 43 genes designated as "attack phase core genes".
We focused on the TAD operon (bd0109-0119), that encodes flp (type IVb) pilin-encoding genes and their processing and export machinery. Our analyses showed that type IVb pilin genes, while found in various gram negative bacteria, are more numerous in BALOs than in other known bacteria and that they are very diverse in sequence, in their association with the TAD loci, and in their regulation (Figs 2, 3 and 7). The only other known example of non-TAD associated flp genes is one of the two Agrobacterium tumefaciens flp pili 45 .
The B. bacteriovorus TAD-encoding flp pilins (flp1, flp2) and flp4 are expressed during AP in wild type cultures as well as in planktonic and in biofilm HI cultures (possibly in the AP cells of these unsynchronized cultures). Flp1 (Bd0119), along with other TAD locus proteins Bd0110, Bd0111, Bd0112, Bd0113, was also detected by proteome analyses, but Flp2 and Flp4 were not detected, possibly because their hydrophobicity and small mass make detection of Flp pilins very difficult 33 . flp3 was silent under all various conditions we tested and as such is the first documented case of a flp pilin gene in gram-negative bacteria that was not found to be expressed. flp3 does not appear to be a pseudogene and it may be expressed under other environmental conditions. While a type IVa pilus gene had been shown to be essential for B. bacteriovorus predation 14,17 , no functional analysis of type IVb pilin genes have been performed in B. bacteriovorus 33 . Since type IVb pilins are crucial for physical interactions with the eukaryotic host cell in pathogenic bacteria [48][49][50] we posited they might also play an important role in BALO predation.
Attempts at deleting flp1, flp2 and flp4 in a wild type background were unsuccessful. In contrast, a flp3 wild type knockout was viable, in accordance with its lack of expression in both AP and GP. In an HI background, triple flp1-flp2-flp4, double flp1-flp2, as well as all individual flp deletions yielded viable strains that could be grown axenically on a rich medium but were unable to prey. The flp genes were not interchangeable as their mutants were not complemented by each other or by the flp3 gene. Thus, each of flp1, flp2 and flp4 is required for predation, and each may have a specific role but none is crucial to the cell cycle, as mutants in an HI background were viable.
B. bacteriovorus, and all other BALO, include both type IVa and type IVb pilin genes. These pilin-encoding genes differ in their sequence characteristics and in the systems that assemble, process and export them 35 ,  FliA (bd3318 and a11q_617, respectively) activity was examined in an E. coli heterologous system. Flagellin promoters (Bd0606, A11q_1936) and an empty vector were used as positive and negative controls, respectively. The promoters of flp1 genes (bd0119 and A11q_2366) responded strongly to FliA induction in comparison to the controls. In contrast, promoters of the flp3 (bd3297 and a11q_630) and flp4 (bd4000 and a11q_1811) genes did not respond to FliA induction. and different functions have been attributed to each. While in most bacteria PilA pilins are responsible for surface-dependent motility and attachment, the Flp pilins were found to be involved only in surface attachment [38][39][40][41][42][43][44] . In some pathogenic bacteria (such as Pseudomonas aeruginosa, Aggregatibacter actinomycetemcomitans, Haemophilus ducreyi and Agrobacterium tumefaciens) disturbance of flp and TAD components attenuated virulence as well as decreased non-specific surface adhesion that is required for the development of biofilms 38,45,51,52 . In other pathogens, flp deletion reduced virulence or microcolony formation without affecting biofilm formation 53 , or bacteria-cell host interaction 49 . In B. bacteriovorus, deletion of the three active flp genes had no effect on attachment/adhesion and HI biofilm formation. It is remarkable that both Bdellovibrio type IVa and the type IVb pilin -encoding genes, while essential for predation, are unnecessary for adhering or binding to prey cells, suggesting that other adhesion proteins expressed in the AP may be responsible for that step (58).
We provide the first experimental evidence that expression of the TAD-associated flp1 and flp2 gene is regulated by FliA, a flagellar sigma factor, both in the periplasmic predator, B. bacteriovorus HD100, and in the epibiotic predator, B. exovorus JSS. Further analyses of promoter sequences suggested that TAD associated flp genes are also regulated by FliA in other BALOs. Notably, flp4, while co-expressed with the TAD-encoded flp1 and flp2 is not similarly regulated by FliA. In many bacteria FliA regulates flagellar and chemotaxis-related genes 46,54 . In several pathogens, FliA has expanded to directly control virulence related genes, such as a cytotoxic protein from Campylobacter jejuni, or indirectly through cylic-di-GMP modulation as with the E. coli fimbriae or Vibrio cholerae cytotoxin [55][56][57] . These FliA-regulated traits are unlinked to the cell cycle. In contrast, we previously found that in B. bacteriovorus FliA expanded to become the "AP master regulator" controlling about 77% of the AP genes 23 . Accordingly, in B. bacteriovorus FliA has turned into an essential gene: B. bacteriovorus FliA knockout was lethal both in wild type and HI mutants, indicating that FliA has expanded to regulate vital functions, a situation not known in any other bacterium. FliA regulation thus appears to be an indispensable component of obligate predators and a central regulator of the Bdellovibrio cell cycle.
Although functional analyses show that type IVa and type IVb pilins mutants are phenotypically undistinguishable from each other, genetic and biochemical analyses suggest that each may have a different role in signaling prey cells to the predator. We propose to integrate our data into previously proposed models 8,18,20 . BALOs reversely bind the prey outer membrane (with a yet unknown component), followed by irreversible anchoring to a prey envelope-associated structure 8 . Type IV pili with their varied pilins, bring about the sensing, transduction, and an initial output of the interaction in the form of tightening of the predator-prey complex. According to a recent study, transition from AP to GP defines a third phase in B. bacteriovorus' life cycle -the transition phase (TP) 8 . The TP expresses a specific transcription profile in response to signals triggered by the prey. Characteristically, response to the prey envelope results in bd0108 shutdown, and in pilA and fliA down-regulation as the first step resulting from attachment to the prey. We suggest that type IVa and type IVb pili generate differentiated signaling outputs: type IVa pilus sensing leads to its retraction through Bd0108-Bd0109 18 (AP genes not regulated by FliA) initiating a signal sensed at the invasive pole of the predatory cell 20 . This signal could be further processed through cyclic-di-GMP regulation 20 , leading to bd0108 and other non-fliA regulated genes (pilA) downregulation; the signal generated by interacting type IVb pilins during attachment and further, during penetration and establishment in the periplasm, brings about alterations in Bdellovibrio FliA-regulated genes, including flp locus genes and pilin genes. This two-pronged signaling scheme, which still has to be validated, would ensure a rapid response based on enzymatic activation (c-di-GMP signaling) followed by a somewhat slower change at the transcription level.
Predation assay of the mutated and complemented strains. The tested strains were grown axenically to OD 600 0.7, centrifuged, washed twice in DNB, and suspended in 5 ml of OD 600 1 of E. coli in DNB supplemented with the appropriate antibiotic. E. coli were counted by serial dilution on LB agar every 24 hours for 120 h. Plates were incubated overnight at 28 °C.
Orthologs identification and synteny analysis. Genuine orthologs were identified using the BLAST Score Ratio (BSR) approach 59 using a ratio cutoff of 0.2. BSR similarity measure is the ratio between the blast score of one protein to another with the blast score of the protein to itself. flp pilins identification. A position weight matrix (PWM) was generated with the MEME program 60 using flp pilins sequences from the Pfam database (entry PF04964) 61 , Fig. 3. The PWM was searched against selected BALO genomes using the MAST program 61 . Only high scoring hits (expected values < 0.1) within short (<100 amino acids) proteins or ORFs were further considered.
Biofilm formation assay. B. bacterivorous HI M11.1 was grown from an initial 100 μl OD 600 0.7 in 96 well plates at 28 °C. Biofilm formation was measured every 24 hours using the standard crystal violet staining procedure 13 . Sample collection and RNA extraction. RNA was extracted using a MasterPure kit (epicenter) according to the manufacturer's instructions. This was followed by an additional DNAse treatment using TurboDNASE kit Scientific RepoRts | 7: 1013 | DOI:10.1038/s41598-017-00951-w (ambion). RNA integrity was validated by agarose gel electrophoresis. All the RNA samples were tested for DNA contamination by PCR using 16 S rRNA gene primers (8 f and 1492 R in Table S5). RNA extraction from biofilms was performed in glass test tubes: a turbid culture of the host independent strain M11.1 was diluted to OD 600 0.7 into 1 ml fresh PYE for 4 days. Next, the planktonic phase (AP) cells were harvested for RNA extraction. The biofilm was washed three times with HEPES buffer then immersed in RNALatter (Ambion), extracted from the test tube by scraping the flanks, centrifuged and kept at −80 °C until RNA was extracted. Samples were collected as described previously 8 : 1 ml of each time point was centrifuged and immersed in RNAlatter and then kept in −80 °C until RNA was extracted.
Semi-quantitive RT-PCR (sqRTPCR). 100 to 200 ng of RNA were used for cDNA synthesis using Improm II kit (Promega) according to the manufacturer's protocol. 50 µl of RNAse free water were added to the cDNA mixture to a final volume of 80 µl. sqRTPCR was performed using 1 μl of a diluted cDNA mixture, primers as listed in Table S5 and a Readymix polymerase (Lambda biotech) according to manufacturer's instructions. Gels were recorded using a DNR MiniBIsPro BioImager with the DNR Gel Capture software with no further processing (DNR Bio-Imaging Systems, Israel).
Proteome analyses. Analysis of the HD100 proteome was performed by MS. SDS-PAGE separation and digestion was performed as previously described 62 . The resulting peptide mixtures were fractioned by with strong cation exchange and analyzed by subsequent high-performance liquid chromatography coupled to an Orbitrap XL. MS data (survey and fragmentations scans) was analyzed using Thermo Proteome Discoverer ™ (Thermo Fisher Scientific, Version 1.4.1.12) through SEQUEST HT search engine with a UniProt database (5 th November, 2015) containing the whole HD100 proteome 63 . Proteins were considered, if they were identified with at least two unique peptides and a minimum score of 20.
Plasmid construction. All plasmids were constructed by the "Restriction Free" (RF) method as described by 64 . PCR was performed using Q5 polymerase (New England Biolabs) according to manufacturer instructions. Plasmids backbones and primers are listed in Tables S4 and S5, respectively.
For the purpose of in-frame deletions, 500-1000 bp long DNA fragments were amplified from both sides of the target gene and fused together by the RF method to the suicide plasmid pSSK10 65 . Knockout primers were designed to leave a 16-17 amino acid remaining to avoid upstream polar effects.
Complementation was carried by pPROBE-NT, which carried the examined gene with its promoter (usually 200-400 bp upstream to the translation start site).
Examined fliA orthologs were cloned to replace the gIII secretion signal peptide, downstream to the arabinose-induced promoter in pBAD/gIII (Invitrogen). Examined promoters were cloned upstream to a promoter-less full-length β-galactosidase (lacZ) in pBBR1-MCS2-LacZ 66 . B. bacteriovorus. E. coli donor strains were grown with the appropriate antibiotics (listed in Table S4) to an OD 600 of 0.4, then washed twice with DNB and concentrated 10-fold. A 100 µl volume of each donor was mixed with 100 µl of B. bacteriovorus HD100 or B. bacteriovorus M11.1 from a turbid culture or with an E. coli-B. bacteriovorus HD100 overnight-dual culture. The mixture was then spread over a sterile membrane that was placed upon a DNB agar plate. Following overnight incubation at 28 °C, the membrane was washed with 1 ml of DNB, diluted and plated on E. coli double layer agar or PYE agar plates supplemented with 50 µg.ml −1 streptomycin and 25 µg.ml −1 kanamycin. When E. coli MFDpir was used, diaminopimelic acid (DAP) was added to a final concentration of 100 µg.ml −1 .

B. bacteriovorus knockouts.
After conjugation with the appropriate pSSK10 derived construct, plaques or colonies were picked up and checked for merodiploidicity by PCR (Table S4). Selected plaques or colony were transferred into 5 ml of OD 600 1 of an E. coli suspension or to PYE, respectively, supplemented with streptomycin and grown for several days. An aliquot was transferred to OD 600 1 of an E. coli suspension or PYE both supplemented with streptomycin and 5% sucrose and grown for 24 and 72 hours, respectively, then diluted and plated on E. coli double layer agar or PYE agar plate both supplemented with 50 µg.ml −1 streptomycin. Plaques and colonies appeared after a week of incubation. Plaques were transferred to a 96 wells plate, each well containing 150 to 200 μl of a OD 600 1 E. coli suspension. HI colonies were transferred to a grid PYE agar plate (53 colonies per plate), then after appropriate incubation, samples were examined for allelic exchange by PCR. Parental strains (WT HD100 or HI) and the appropriate merodiploid strain were used in each PCR reaction as a control.
Selection for fliA knockout strain was performed with the following modifications: Recombinants were screened as above in counter-selection tubes containing PYE and 5% sucrose. After 2, 4 and 5 days (instead of after 3 days in the standard procedure), suspensions were plated on PYE agar and incubated for a week. Colonies were examined for allelic exchange by PCR. Thereafter, plates were incubated for an additional 2 to 3 weeks under humid conditions, to enable the development of slowly growing colonies, typically 10 to 20 which were examined for allelic exchange by PCR. This experiment was performed three times. E. coli fliA knockout. E. coli fliA was knocked out using a Lambda-red recombination system 67 . Briefly, E.
coli DH5α/pKD46 was induced for lambda red recombinase by adding 0.05% arabinose. A linear DNA fragment containing a chloramphenicol resistance cassette flanking 30 bp with the corresponding upstream and downstream to E. coli fliA was created by PCR using primers: fliAlambdared-F and fliAlambdared-R and pKD3 as a template. Next, the liner fragment was electroporated into the induced lambda red E. coli DH5α/pKD46. This resulted in chloramphenicol resistant E. coli colonies that were isolated and validated for fliA knockout using primers Colifliadel-R and Colifliadel-F. E. coli -Bdellovibrio attachment assay. Attachment assays were performed according to Milner et al. 20 with slight modifications. HI and HI flp mutants were grown in PYE medium to an OD 600 of 0.5, washed twice in DNB, and mixed with E. coli. The experiments were carried out at the final cell concentrations of 10 6 and of 10 7 ml −1 . The Bdellovibrio -E. coli co-culture was incubated for recovery of 30 min in 28 °C, then concentrated by low speed centrifugation (4000 rpm for 10 min), the pellet was re-suspended in fresh DNB and re-incubated in 28 °C. 10 µl were taken at different time points, placed on positively charged polylysine-coated adhesive microscope slide (X-tra slides, Leica) and examined under phase contrast microscopy. As a negative control, E. coli was replaced with Bacillus subtilis, a gram-positive bacterium that Bdellovibrio does not attach to. B. bacteriovorus and B. exovorus flp pilin gene promoters response to FliA in an E. coli heterologous system. Examined promoters were cloned upstream to β-Galactosidase (lacZ) in pBBR1-MCS2-LacZ. The resulting constructs were introduced into E. coli DH5α ∆fliA together with FliA cloned under arabinose-induced promoter in pBAD. The endogenous E. coli fliA was knocked out in order to reduce the background signal. Each E. coli strain carrying the two plasmids was grown overnight, then diluted into fresh LB to OD 600 0.05, supplemented with either 0.5% glucose or 0.5% arabinose. When cultures reached mid-log (OD 600 0.4-0.6), bacteria were harvested and measured for LacZ activity following a standard protocol 68 . Final β-Galactosidase activity was calculated by the following formula: In this manner any background signals stemming from E. coli were eliminated.