Characterization of the Attachment of Three New Coliphages onto the Ferrichrome Transporter FhuA

ABSTRACT Receptor-binding proteins (RBPs) allow phages to dock onto their host and initiate infection through the recognition of proteinaceous or saccharidic receptors located on the cell surface. FhuA is the ferrichrome hydroxamate transporter in Escherichia coli and serves as a receptor for the well-characterized phages T1, T5, and phi80. To further characterize how other FhuA-dependent phages attach to FhuA, we isolated and published the genomes of three new FhuA-dependent coliphages: JLBYU37, JLBYU41, and JLBYU60. We identified the egions of FhuA involved in phage attachment by testing the effect of mutant fhuA alleles containing single-loop deletions of extracellular loops (L3, L4, L5, L8, L10, and L11) on phage infectivity. Deletion of loop 8 resulted in complete resistance to SO1-like phages JLBYU37 and JLBYU60 and the previously isolated vB_EcoD_Teewinot phage, but no single-loop deletions significantly altered the infection of T1-like JLBYU41. Additionally, lipopolysaccharide (LPS) truncation coupled with the L5 mutant significantly impaired the infectivity of JLBYU37 and JLBYU60. Moreover, significant reductions in the infectivity of JLBYU41 were observed upon LPS truncation in the L8 mutant strain. Analysis of the evolutionary relationships among FhuA-dependent phage RBPs highlights the conservation of L8 dependence in JLBYU37, JLBYU60, Teewinot, T5, and phi80, but also showcases how positive selective pressure and/or homologous recombination also selected for L4 dependence in T1 and even the lack of complete loop dependence in JLBYU41. IMPORTANCE Phage attachment is the first step of phage infection and plays a role in governing host specificity. Characterizing the interactions taking place between phage tail fibers and bacterial receptors that better equip bacteria to survive within the human body may provide insights to aid the development of phage therapeutics.

T he possibility of phage therapy was explored in the early 1900s to treat bacterial infections, but research in the United States ceased with the discovery of antibiotics (1,2). However, the rise in antibiotic-resistant infections is becoming a major issue, causing a resurgence in phage therapy. In 2019 alone, 1.27 and 4.95 million deaths were attributed to or associated with antibiotic-resistant infections, respectively, and infections caused by Escherichia coli alone were responsible for the most deaths in both cases (3). The number of antibiotic-resistant infections is growing at an alarming rate and mandates the discovery of alternative treatments. Utilization of phage therapeutics is a potential candidate and multiple cases of successful phage therapies have been reported (4)(5)(6)(7). However, further understanding of the processes behind a successful phage infection could aid in the development of phage therapy.
Unlike antibiotics, phages can be very selective regarding which hosts they infect. This selection partially occurs at the initial stages of infection during host attachment; however, host defenses such as restriction modification systems (8), CRISPR-Cas (9), and bacterial suicide (10) can also prevent a phage from completing its replication cycle (11). Phage attachment occurs in two binding steps, classified as reversible and irreversible binding (12)(13)(14)(15). Specialized receptor-binding proteins (RBPs) located at the tips of phage tail fiber and spike proteins allow them to bind to receptors located on the surface of the bacterial cell. Reversible binding initially occurs on cell structures that are easily accessible, which helps to loosely tether the phage to the cell, but the phage can still detach. Irreversible binding triggers ejection of the genetic material into the cell, and phage replication follows (14,15). Phages can bind to a single receptor or may require a multiple-receptor mechanism in which the RBPs and receptors used during reversible and irreversible binding are not the same (13). For example, lipopolysaccharide (LPS) has been shown to be highly involved in phage attachment and can be bound either individually or coupled with other receptors (16,17). Reversible binding to the O-antigen, outer and inner core, or lipid A of LPS can also aid in phage stability and infection efficiency by leading it to the cell surface where irreversible attachment occurs (14,15). Mutations that eliminate portions of LPS, such as the O-antigen in the MG1655 strain of E. coli or removal of the heptose core in a DwaaC mutant, can reduce or even block phage infection (18,19). The host range of a phage is highly dependent upon the receptors it targets, and analyzing how tail fibers interact with their receptors is crucial to unlocking the mysteries behind phage host specificity.
Studies have shown that some phages attach to multiple receptors (20)(21)(22) or target receptors that could give their host an environmental advantage. For example, iron uptake in Gram-negative cells is mediated through outer membrane porins (23,24), which play a key role in determining the virulence of human pathogens (25,26). Phages T1, T5, and phi80 have been shown to adhere to the ferric hydroxamate transporter FhuA (15,21,(27)(28)(29)(30) and may be candidates for treating certain bacterial infections. While FhuA is not unique to pathogenic strains, loss of additional iron transporters does result in a fitness cost when the host is in an iron-limited environment (31).
The structure of FhuA consists of a 22-stranded antiparallel b-barrel (b1 to b22) that is blocked by an N-terminal cork domain. It is also composed of 11 loops (L1 to L11) and 10 turns (T1 to T10) which form at the extracellular and periplasmic faces, respectively. Active transport of ferrichrome (Fc) through FhuA has been shown to bind to L3, L11, b7, and b9 (32,33) and is dependent on the inner membrane TonBcomplex (21). FhuA has also been shown to transport the antibiotics albomycin (34) and rifamycin CGP 4832 (35,36) and the bacteriocins colicin M and microcin J25 (37). Attachment of phages T1 and phi80 is both FhuA-and TonB-dependent (15), but T5 infection occurs independently of TonB (27). Additional dependence upon TonB could be used to ensure DNA ejection is triggered within live cells, and the conformational changes it confers in FhuA may provide the signal for irreversible phage attachment.
Once the receptor has been bound, the phage ejects its genetic material into the host cell. How this process occurs, however, is completely dependent on the structure of the phage. Tailed phages fall within the Caudoviricetes class and consist of icosahedral heads and tails that can vary in length and complexity. The well-known phages lambda (38), T1 (15), and T5 (27,28) fall within this class but all contain noncontractile tails. T4 is also a tailed phage, but it has a more complex tail structure that contracts upon host recognition (39,40). Both contractile and noncontractile phages eject their genetic material into the host cell following irreversible attachment to their receptor(s). Previously, contractile and noncontractile tailed phages were classified as Myoviridae and Siphoviridae, respectively, but recent updates to phage phylogeny have allowed for phages to be further distinguished (41) and these terms are now used only to refer to the overall structure of the phage. For example, T1-like phages contain noncontractile tails, so it can be said that they have Siphoviridae structure. However, further classification categorizes them within the family Drexlerviridae, which is further divided into different subfamilies (e.g., Tunavirinae and Tempervirinae) and genera (Tunavirus, Hanrivervirus, and Warwickvirus) (42).
While some phages have been shown to utilize FhuA as their receptor, relatively few FhuA-dependent phages have been characterized to identify (i) which regions of FhuA are involved in phage attachment or DNA ejection, (ii) the role of LPS in FhuA-dependent phage infection, and (iii) the evolutionary relationships of the RBPs of FhuA-dependent phages. To address this, we constructed six fhuA loop deletion strains that contained alterations to L3, L4, L5, L8, L10, and L11. The effect of each mutant allele on phage infection was then explored with and without the presence of the LPS heptose core. We also compared the evolutionary relationships of the RBPs of FhuA-dependent phages and their close relatives to analyze whether they coincided with FhuA loop-dependency results.
This work shows which extracellular loops of FhuA play a role in FhuA-dependent phage attachment as well as whether LPS aids in FhuA discovery in three novel coliphages. We also compare these findings against the phylogenetic relationships of their FhuA-dependent RBPs to highlight changes in selective pressure that contributed to the diversity seen in the binding profiles in FhuA-dependent phages. Further understanding of the components of phage receptors that encode host recognition and how their RBP counterparts evolve over time is key to unlocking the secrets that drive host specificity, and could aid in the exploration of phages as potential therapeutics.

RESULTS
FhuA-dependent phage isolation and characterization. Phages were isolated and purified from raw sewage samples taken along the Utah Wasatch front and tested on E. coli strains containing single-gene knockouts of potential phage receptors. Three FhuA-dependent phages were discovered due to their inability, and then restored ability, to infect a MG1655 DfhuA::kan and a DfhuA::kan pJL002 complement strain, respectively. The fhuA complement plasmid (pJL002) was generated by cloning fhuA from MG1655 and ligating it into pKG116 via the NdeI and KpnI sites using standard techniques. To aid in the analysis of these phages, we extracted and sequenced their genomes using Illumina technology. De novo genome assembly was performed, and gene annotations were assigned by comparing each predicted gene with its codon potential and protein BLASTp results. The published phages are named JLBYU37 (accession no. OK272488), JLBYU41 (OK272479), and JLBYU60 (OK272474) and their genomes are publicly available in the NCBI database.
Genome analysis revealed that T1-like phage JLBYU41 belongs to the Hanrivervirus genus, whereas JLBYU37 and JLBYU60 fall into the Dhillonvirus genus as SO1-like phages (42). This was also supported by scanning transmission electron microscope images of the phage which displayed their Siphoviridae structure (Fig. 1). JLBYU37 and JLBYU41 formed 4-and 7mm bullseye plaques, respectively, but JLBYU60 had plaques that averaged ;3 mm in diameter in LB top agar containing 0.7% agar and consistently contained incomplete halos. All phage plaques were hazy in appearance, suggesting the possibility of a lysogenic lifestyle, but lysogen generation could not be confirmed by standard techniques.
Genomic and structural protein comparisons. Phage genomes are very mosaic, but structural homology of phage proteins can sometimes be found even when little gene identity is observed. Because of this, relationships within a group of phages may be uncovered using structural data even if there is jumbled identity at the nucleotide level (43). To analyze the conservation of both gene and structural homology among our FhuA-dependent phages, dot-plot comparisons were performed using genomic (nucleotide) sequences (Fig. 2) as well as the protein (amino acid) sequences for the phage major capsid proteins (MCPs) (Fig. 3) and RBPs (Fig. 4). MCPs are more conserved than RBPs and are considered the gold standard when comparing phage relationships. The phages included in this analysis are listed in Table 1 and were pulled from the top MCP BLASTp hits for phages JLBYU37, JLBYU41, and JLBYU60. Previously identified FhuA-dependent phages T1, T5, and phi80 and their top hits were also included, except for the hits from phi80 because they all stemmed from bacterial genomic sequences that most likely harbored lysogenic phage. Since JLBYU37 and JLBYU60 are SO1-like phages and phi80 is a Lambdavirus, SO1 and lambda were also included in our analysis. T4 was used as the outgroup because it also falls within the Caudoviricetes class but contains a contractile tail.
Dot-plot comparisons of the genomes of T5-like (here, cluster A) and SO1-like phages (cluster B) each separate into 1 cluster, but T1-like phages divide into clusters C and D (Fig. 2). Phage JLBYU41 falls within the Hanrivervirus genus (cluster C) of T1-like phages; however, a visible distinction is observed compared to the Tunavirus genus (cluster D). No phages showed nucleotide sequence homology with the well-studied phages phi80, lambda, or T4 except for slight homology between lambdaviruses phi80 and lambda (cluster E).
To further understand structural and functional similarities between these phages, dot-plot comparisons of the amino acid sequences of their predicted MCPs (Fig. 3) and RBPs (Fig. 4) were also generated using the same set of phages. MCP analysis revealed high protein sequence conservation within each genus, but various levels of homology occurred between T1-like Hanriverviruses, Warwickviruses and Tunaviruses. The clusters generated in the MCP comparison mirrored those found in the genomic comparisons. Comparisons of their predicted RBPs separated into 4 clusters but T5, IrisVonRoten, IME178 and PGN590 no longer showed sequence similarity with any other phages. Homology between clusters C and D was still noted but similarities vanished when the C-terminal regions of Tunavirus RBPs were compared.
The diversity of our FhuA-dependent phages was further characterized by using BLAST and BLASTp to compare the sequence identity of their genomes, MCPs, and RBPs. The genomes of SO1-like phages JLBYU37 and JLBYU60 are 93.45% identical, with 98.36% and 99.30% identity shared between their MCPs and RBPs, respectively. This analysis led to the detection of another FhuA-dependent SO1-like phage previously isolated from our institution called Teewinot (44). The predicted RBP for Teewinot is 99.21% identical to that of JLBYU37 (gp22) and 98.68% identical to that of JLBYU60 (gp24). Genomic comparisons of is a T1-like phage that falls within the genus Hanrivervirus, but SO1-like phages JLBYU37 and JLBYU60 belong to the genus Dhillonvirus. All phages generate bullseye plaques that vary in size, but JLBYU60 generates incomplete halos surrounding the center clearing.
Classification of FhuA loop dependency. FhuA complement strains containing single-loop deletions for loops 4, 5, 8, 10, and 11 were designed to replace the same residues as outlined by Endriss et al. (33) and are summarized in Table S1 in the supplemental material. Each loop was replaced with a short peptide chain to reduce the effect that deletions would have on the structure of the b-barrel. A loop 3 deletion , and Lambdavirus (brown circles) was observed. The names of FhuA-dependent phages identified in this study are highlighted in red, and sequences listed on the vertical axis are compared against themselves on the horizontal axis in the same order. Regions of gene similarity are represented by dots, and diagonal lines suggest high sequence conservation between the compared sequences. Genomes running on the reverse strand were rearranged to provide more accurate alignments, and 5 unique clusters were observed. Tunaviruses resemble hanriverviruses and warwickviruses (cluster C) but were found to be different enough to separate them into their own cluster (D). mutant was also made, but modifications to the Endriss design were made to shift the site for deletion to begin with L247 to conserve Fc binding sites Y244 and W246 (33). To confirm the availability and functionality of the FhuA loop deletions in the outer membrane, each mutant was tested with a Fc growth promotion assay.
Complementation of fhuA in strain MG1655 DfhuA::kan with pJL002 resulted in growth promotion, but normal growth also occurred in the DfhuA::kan pKG116 (empty vector control) when Fc was added as the only iron source. Reduction and restoration of growth could only be observed in DfhuA::FRT DfepA::kan pKG116 and DfhuA::FRT DfepA::kan pJL002, respectively, due to the removal of an additional iron transporter FepA. When tested in the DfhuA::FRT DfepA::kan background, fhuA mutant alleles for L4, L5, L8, and L10 (pJL005 to pJL008) displayed growth enhancement. Although L3 and L11 (pJL004 and pJL009) did not show growth enhancement, these mutant Individual L3 and L5 mutants showed a moderate role in the infection of JLBYU37 and JLBYU60, but Teewinot and JLBYU41 were only affected by the L5 mutant. To discover the possible role of LPS in infection efficiency, phage sensitivity was also analyzed in a DfhuA::FRT DwaaC::kan strain expressing each of the fhuA mutant alleles (pJL003 to pJL009). As a result, the removal of the inner and outer core of LPS significantly affected the adsorption of phages JLBYU37 and JLBYU60 when L5 was removed (noted by one and two stars in Fig. 5, respectively) but the infection of JLBYU41 was significantly impacted by the L8 mutant (noted by two stars in Fig. 5).
To explore the possibility that JLBYU41 utilizes multiple extracellular loops during attachment, we constructed double-loop deletion mutants L3L5, L4L5, and L5L8 (pJL010 to pJL012) and individually tested for phage infectivity in both the DfhuA::FRT DfepA::kan and DfhuA::FRT DwaaC::kan backgrounds. Fc growth enhancement was observed for each of the double-mutant strains. As predicted, the L5L8 double-loop deletion mutants were completely resistant to JLBYU37, JLBYU60, and Teewinot, but the mutation also conferred full resistance to JLBYU41. Interestingly, the L4L5 mutant was resistant against all phages except for Teewinot. Removal of the LPS heptose core, however, completely blocked Teewinot infection in the L4L5 mutant. Only JLBYU41 could infect the L3L5 deletion strain, but LPS truncation resulted in significantly decreased infectivity (noted by 1 star in Fig. 5). Evolution of FhuA-dependent phages. Phages are constantly in an evolutionary arms race to maintain their ability to infect their hosts, and one way a phage can evolve is by acquiring changes in its RBPs. Generally, phage RBPs have higher levels of conservation in their N termini but show greater diversity at the C-terminal end (45) suggesting that the C terminus may be under positive selective pressure or undergo homologous recombination (46). Because of this, it can be difficult to understand phage relatedness based upon nucleotide sequences alone. Some phages, however, maintain structural similarities even when there is little homology at the nucleotide level (47)(48)(49). Consequently, our phylogenetic analysis was performed using the amino acid sequences for both the RBPs and MCPs in each phage, the latter of which are highly conserved. Although MCPs and RBPs are both structural proteins, we would expect to see altered evolutionary paths due to the role RBPs can play in phage evolution. Comparisons between these trees may reveal unique evolutionary relationships that may appear due to changes in selective pressure or RBP recombination, and may give insight into the role tail fiber evolution may have played in diversifying FhuA-dependent phage attachment. To do this, we performed phylogenetic analysis using the same set of predicted phage MCPs and RBPs that were analyzed in our dot-plot comparisons (See Table 1).
The predicted evolutionary relationships for JLBYU37, JLBYU60, Teewinot, and JLBYU41 are represented by a tree generated from their MCP sequences in Fig. 6A. As expected, each phage clusters with members of its respective genus with T1-like Warwickviruses, Hanriverviruses and Tunaviruses diverging early from the ancestors of Tequintaviruses and Dhillonviruses. The ancestor giving rise to Lambdaviruses phi80 and lambda branched from the lineage that gave rise to all T1-like, T5-like and SO1-like phages.
Analysis of the evolution of their putative RBPs, however, suggests a different evolutionary history (Fig. 6B). One of the most noticeable differences is the distribution of Tequintavirus phages throughout, which are monophyletic in the MCP tree. Tequintavirus IME178 branched off very early from the ancestor that gave rise to Dhillonvirus, Lambdavirus, T1-like phages, and the remaining tequintaviruses in this analysis. The RBPs of Dhillonvirus form a clade with PGN590 and branch off from the ancestor giving rise to T5 1 IrisVonRoten tequintaviruses. This varies from the relationships observed for their MCPs, which show Dhillonvirus and Tequintavirus as sister taxa. GEC-vB_N5, OSYSP, and phiLLS tequintaviruses branch off from the clade giving rise to lambdaviruses and T1-like phages. Lambda and phi80 are more distantly related to each other here than in the MCP tree, which depicts them as sister taxa, but are more closely related to tequintaviruses and T1-like phages. This coincides with their corresponding receptors since lambda binds to the LamB receptor and phi80 is FhuA-dependent, like some other T1-like phages. Warwickviruses and tunaviruses are monophyletic, which is consistent with the findings from the MCP tree, but Hanrivervirus RBPs are now paraphyletic. Importantly, although the clade containing FhuA-dependent phage JLBYU41 shares a node with warwickviruses, none of the warwickviruses isolated in our lab (JLBYU01, JLBYU07, JLBYU10, JLBYU16, JLBYU19, or JLBYU26) use FhuA as their sole receptor (personal observation). The differences between the MCP and RBP trees suggest that known evolutionary processes (e.g., recombination or positive selective pressure) could have altered receptor specificity.
To determine the role the C-terminal RBP sequences may have had on these observed differences, we also constructed a phylogenetic tree for the C-terminal region of the RBPs (Fig. S1). The C-terminal start site was chosen by first comparing the RBPs from FhuA-dependent T5-like, T1-like, and SO1-like phages with their homologs in a multiple sequence alignment (not shown). RBP length varied from 585 to 1,192 amino acids and the sequence identities between all RBP sequences were very different, so sequence identity could only be compared between phages within the same genus. The cutoff for the C-terminal start site for each comparison was chosen to be upstream of the region where a reduction in percent identity was observed, and is outlined in Table S3. The C-terminal RBP tree was then constructed as previously described. We then performed a log-likelihood test on the C-terminal RBP tree to determine whether there were any significant changes in the predicted phylogeny compared to the full RBP tree. This was done using IQ-TREE web server topology evaluation with 10,000 replicates. Significant differences were found using the one-sided Kishino-Hasegawa (P = 0.0004), Shimodaira-Hasegawa (P = 0.0004), and approximately unbiased tests (P = 3.03 Â 10 26 ). As a control, we also ran this evaluation on the full-length RBP tree, and all tests resulted in no significant differences (P = 1). This suggests that natural processes of phage evolution (such as positive selective pressure or homologous recombination) involving the C-terminal regions of the phage RBPs have occurred throughout their evolutionary history and may be responsible for the observed differences in FhuA-dependent phage attachment.

DISCUSSION
Effect of FepA on Fc uptake. In this paper, we analyzed the binding of three novel phages with the FhuA protein from E. coli by characterizing which regions of FhuA are required for a successful phage infection and whether the presence of the LPS heptose core contributes to infection efficiency. At the heart of our analysis was a comparison of the ability of these phages to infect different versions of FhuA with mutated extracellular loops and also to probe the role of LPS in infection. We also analyzed the potential lineage of FhuA-dependent RBPs and its correspondence with FhuA-dependent phage binding profiles.
To do this, we first analyzed the effect of mutant fhuA constructs containing singleloop deletions on phage infection. Because loop deletions could potentially disrupt protein stability or membrane insertion, we developed a Fc growth promotion assay to check each mutant for membrane integration, FhuA functionality, and phage accessibility. As a negative control for these experiments, we observed a reduction of growth in a strain harboring an empty vector that did not encode fhuA. This loss of growth enhancement was only observed when fepA (ferric enterobactin transporter) was also removed. Enterobactin is another siderophore that has been shown to bind ferric iron with high affinity and utilizes the outer membrane transporter FepA. Although FepA is predicted to be ;11Â less abundant than FhuA (50), these results suggest that enterobactin may be able to steal enough ferric iron from dipyridyl (an iron chelator that was added to the M9 plates to bind any remaining free iron) or Fc to support normal growth in the DfhuA::kan strain. Possible explanations for this observed bypass of FhuA dependence are (i) the presence of another siderophore which has a higher affinity for iron and can transport stolen iron from Fc or dipyridyl, or that (ii) enough iron remained in the plates to support regular growth unless both fhuA and fepA were both removed.
Growth promotion in a DfhuA::FRT DfepA::kan pJL002 (FhuA complement) strain, however, suggests that the presence of enterobactin does not convert all Fc to its apoform. Growth enhancement was also observed in a DfhuA::kan pKG116 strain (empty vector control) when only HCl was added, but the subsequent deletion of fepA resulted in no growth enhancement. Importantly, complementation of the double-knockout strain, DfhuA::FRT DfepA::kan, with pJL002 led to only minimal growth with the HCl negative control, but full growth enhancement with the addition of both HCl and Fc.
Role of L8 in FhuA-dependent phage infection. Removal of L8 resulted in complete loss of infection by SO1-like phages JLBYU37, JLBYU60, and Teewinot. This pattern coincides with previous studies that show that removal of L8 results in complete or significant reduction in phi80 or T5 phage infection, respectively (33). To further understand how mutations in the L8 mutant could confer phage resistance, we compared the AlphaFold2 predicted structure of our L8 mutant against the solved structures of FhuA with (8A8C) and without (2FCP) the T5 phage attachment protein pb5, which has recently been solved and published (51) (Fig. 7).
The plausibility of the AlphaFold2 prediction of the L8 mutant structure was assessed through superimposition onto FhuA (2FCP) and by calculating the root mean square deviation (RMSD, Fig. 7A) for the placement of the backbone C-alpha atoms. Only the regions that encode the 6ÂHis tag of FhuA (2FCP) and the mutagenized section of L8 (Fig. 7B) resulted in a high maximum RMSD or low similarities between the two structures (indicated by the red stars in Fig. 7A). The top-down views of FhuA before (Fig. 7C) and after irreversible phage binding (Fig. 7D) are shown as surface model structures to show the residues exposed to the extracellular matrix. The structure of pb5 was removed from FhuA (8A8C) to aid in the visualization of L8 positioning and lumen exposure.
These comparisons highlight two residues, F557 (L8, yellow) and Y325 (which is in L4, cyan), which extend out into the entrance of the lumen and are shifted upon phage attachment. Replacement of residues DPEGSFF (D552-F558) with NSEG in the L8 mutant (Fig. 7E) results in a 3-amino acid reduction in loop length and a 56% reduction in solvent accessible surface area. This reduction results in a miniature L8 (shown in red) that is not predicted to stick out into the opening of the lumen. Superimposed structures of FhuA with and without phage attachment (Fig. 7F) clearly show a shift of Y325 and F557 placement (shown as stick structures) upon T5 attachment, but Y325 placement in the L8 mutant is expected to resemble FhuA structure prior to phage interaction (Fig. 7G) rather than post-T5 attachment (Fig. 7H).
Y325 and F557 predominance in the entrance of the lumen and shift upon T5 attachment suggests that these residues may also be involved in JLBYU37, JLBYU60, and Teewinot attachment. One observation is that these tyrosine and phenylalanine residues are hydrophobic amino acids that are fully exposed to the extracellular environment. Hydrophobic residues have been shown to play a supportive role in phage receptor binding (52), but whether they play a role in supporting initial RBP binding or serve as targets for irreversible phage attachment is still yet to be determined.
Removal of additional residues from L8 may also have a role in phage susceptibility. Analysis of the residues comprising L8 revealed the presence of two charged amino acids (D552 and E554) and two hydrophobic amino acids (F557 and F558). Both types of amino acids have been shown to play a role in ligand binding and may play an important role in tethering the phage to the cell. D552 and E554 have been predicted to play an important role in maintaining the rigidity of L8 (53), and collapse of L8 into the lumen of FhuA has been shown to confer T5 resistance (51). Removal of D552 may have adverse effects on the stability of L8 and alter its orientation on the cell surface. Consequently, this transition could confer complete phage resistance by blocking RBP access to the lumen of FhuA or preventing access to residues on L8 that are important for attachment. While the removal of D552 may cause L8 instability, we do not believe that the truncated version would be able to obstruct the lumen enough to prevent RBP access. Residues Y325 and F557 are both very susceptible to phage interaction, but removal of Y325 in the L4 mutant did not appear to impede JLBYU37, JLBYU60, and Teewinot infection.
Because of this, we believe that F557 may play an important role in infection by (i) initially tethering the phage to the cell surface, (ii) serving as an essential residue in conferring irreversible phage attachment, or (iii) triggering DNA ejection. We speculate that F557 may be involved in irreversible attachment or DNA ejection because the L8 mutation conferred complete phage resistance. However, this hypothesis was not tested in this study, and the mechanism behind L8-dependent phage resistance remains unknown. Utilization of the LPS heptose core in infection efficiency. SO1-like phages JLBYU37, JLBYU60, and Teewinot were affected by the L5 mutation, and significant reductions were observed upon the removal of the inner and outer core sugars of LPS. When observing the position of LPS relative to the placement of FhuA extracellular loops, L3 and L5 serve as a crown that partially wraps around the outside of the b-barrel (Fig. 8). LPS truncations resulting from the DwaaC mutation are parallel to the base of L5 (Fig. 8A). Because of this, we propose that these phages utilize the sugars of LPS to bring them close to the surface of the cell and increase their probability of L5 discovery. Next, reversible binding of L5, which may serve as a ramp, leads directly to the lumen where irreversible binding can take place (Fig. 8B). Although a significant reduction was observed in the L3 mutant, this reduction remained the same upon LPS truncation. Consequently, we propose that L3 plays a role in L5 positioning but does not directly interact with phage tail fibers. L3 and L5 have been shown to be essential for Fc binding, whereas L3, L5, and L8 are important for Fc uptake (33). Interestingly, all 3 loops played various roles in phage attachment, so we propose that FhuA-dependent phages which target L5 have a selective advantage since this loop plays an important role in both Fc binding and transport.
The T1-like phage JLBYU41 was also affected by the L5 fhuA mutation, but the L8 deletion had no significant effect on infection until the LPS heptose core was removed. Interestingly, the truncation of LPS in the L5 mutant did not cause a greater reduction in JLBYU41 infection efficiency. Although LPS may play an important role in aiding in L8 discovery, complete resistance to JLBYU41 was only observed when multiple loops were deleted. Because of this, we propose that JLBYU41 infection requires a more complex interaction involving multiple extracellular loops and LPS.
Evolution of FhuA loop preferences. Acquired dependence upon different regions of FhuA implies that various instances of positive selective pressure or homologous recombination may have occurred during the evolution of FhuA-dependent phages. , and the predicted L8 mutant structure (E) are shown as surface models and their b-barrels are colored green, salmon, and gray, respectively. The lumen (blue), wild-type L8 (yellow), mutant L8 (red), Y235 (cyan), and 6ÂHis tag (raspberry) are also indicated. The structure of T5 pb5 is removed from 8A8C for clarity and allows observation of the shifts conferred by T5 pb5 attachment in FhuA L8 (F557) and L4 (Y325). (F to H) Shifts in L4 Y325 and L8 F557 positioning (shown as sticks) differ upon T5 attachment (F), but Y325 in the L8 mutant is predicted to coincide with Y325 placement in unbound FhuA (G) than in a phage-attached structure (H). To determine whether altered and conserved loop dependency across various phage families could be explained phylogenetically, we compared the evolutionary pathways of their predicted RBPs with their reported loop dependency (Fig. 9). The RBP evolutionary tree (Fig. 9A) represents phages from 7 different genera: Tequatrovirus (T4), Tequintavirus (orange), Tunavirus (purple), Hanrivervirus (green), Warwickvirus (red), Dhillonvirus (blue), and Lambdavirus (brown). The phages within these genera have Siphoviridae structure and contain at least one FhuA-dependent phage, except for T4 which has Myoviridae structure and is dependent upon OmpC and LPS (54). Although the targeted phage receptors are not known for every phage in the tree, those that have been experimentally determined target FhuA (13, 15, 21, 27-30, 39, 40, 55, 56), BtuB (cobalamin outer membrane transporter) (57), LptD (LPS biosynthesis) (32,58), and LamB (maltose outer membrane transporter) (38,56,59), indicated by dark green, orange, light green, and gray triangles, respectively (Fig. 9A). Additional T1-like phages that were isolated in our lab (JLBYU07, JLBYU26, JLBYU10, JLBYU19, and JLBYU01) fall within the Warwickvirus genus and are included in this analysis due to their high similarity with JLBYU41. Although similar, these phages were not found to be FhuA-dependent and the receptors they target remain unknown (purple triangles). Evolutionary analysis of predicted RBPs reveals an early split into two major clades, as discussed above. Briefly, the top clade contains the ancestor that gave rise to T5 and IrisVonRoten tequintaviruses, SO1-like Dhillonvirus, and T1-like Hanrivervirus. The bottom clade contains GEC_vB_N5, OSYSP, and phiLLS tequintaviruses, Lambdavirus, and T1-like Tunavirus, Hanrivervirus, and Warwickvirus. One possible reason for the divergence of these two clades is their dependence or lack of dependence on TonB in phage infection. T5 is known to be TonB-independent (21) (top major clade), whereas phi80 and T1 (bottom major clade) can only infect when both FhuA and TonB are present (15). In addition, the ancestor of T5 diverged from the ancestor of IrisVonRoten, which is a BtuB-dependent phage. Interestingly, BtuB and FhuA are both TonB-dependent transporters (60), and the only other identified BtuB-dependent phage is vB_EcoS_PTX06 (Dhillonvirus clade). Whether IrisVonRoten and vB_EcoS_PTX06 can infect independent of TonB expression, however, is unknown, and not enough information is available to conclude whether the generation of these two major clades occurred due to differences in TonB dependence in phage infection.
Dhillonviruses JLBYU37, JLBYU60, and Teewinot all required the presence of L8 (Fig. 9B) and fall within the same major clade as T5, which is also L8-dependent. Dhillonviruses bob and Envy were previously shown to be FhuA-dependent (56), but the loops they utilize in FhuA attachment were not assessed in this study. However, due to the predicted evolution of their RBPs, we propose that they may also be dependent upon L8.
Complete dependence upon L8 in the bottom major clade is only reported for Lambdavirus phi80 (33), which diverged early from the ancestor, giving rise to lambda and T1-like phages. However, as we move down the evolutionary path toward T1-like phages, T1 (Tunavirus) is found to be L4-dependent. This change could be explained by examining the roles the other loops play in FhuA attachment (Fig. 9C). Using this analysis, we find that phi80 is partially dependent upon L4 (33). While it is unknown whether L8 dependence was exchanged for L4 dependence in T1, these changes suggest that alterations in the host or environment may have made it advantageous for the ancestor that gave rise to T1 to become solely dependent upon L4. FhuA is an iron transporter which is expressed under iron-limited conditions (61). Because of this, changes in the amount of iron available in the environment during this time may have been a driving force for FhuA-dependent phage evolution. JLBYU41, however, is not dependent upon L8 or any of the single loops analyzed in this study. This suggests that the selective pressure during its evolutionary history was either not as strong for single-loop dependence or that there was stronger selection to utilize more than one loop. These findings are consistent with the RBP dot-plot comparisons, which show high homology among all genera of T1-like phages except when the C-terminal regions of tunaviruses are involved. This suggests that RBP evolution Journal of Virology among T1-like phages involved changes to the C terminus, which is also supported by the C-terminal RBP phylogenetic analysis.
Conclusions. Phage attachment onto a host is the primary step in the infection process and plays a major role in determining phage host range. Analysis of which FhuA regions contribute to FhuA-dependent phage infection revealed that the SO1-like phages JLBYU37, JLBYU60, and Teewinot are completely dependent upon FhuA L8. We hypothesize that mutations in L8 confer phage resistance by removing important residues, like F557, which may play an important role in initial binding or be directly involved in irreversible binding and DNA ejection. Complete resistance to T1-like JLBYU41 infection, however, was only conferred if multiple FhuA loops were altered. Removal of the LPS heptose core also significantly reduced phage infectivity, suggesting that these phages utilize LPS to guide them to very specific FhuA loops which serve as targets for reversible binding and guide the RBP to its irreversible binding site. Analysis of these FhuA-dependent phage binding profiles against the predicted evolutionary history of known FhuA-dependent phages and their close relatives revealed the conservation of L8 dependence among members of the Tequintavirus, Dhillonvirus, and Lambdavirus genera. However, L4 dependence in T1 and a lack of single-loop dependence in JLBYU41 infection was also observed within the ancestry of T1-like phages. Significant differences between the topology of the full-length and C-terminal RBP phylogenetic trees highlight the effect that positive selective pressure or homologous recombination may have had on FhuA-dependent RBP evolution. While the receptors targeted by each phage in the tree are not known, this comparison highlights possible instances of RBP evolution among FhuA-dependent phages. This study underscores the regions of FhuA that are targeted during FhuA-dependent phage attachment and the contribution of LPS to FhuA discovery, and begins to explore the evolutionary history which led to the diversity of FhuA-dependent RBPs.

MATERIALS AND METHODS
Phage isolation and receptor identification. Coliphages were purified from raw sewage samples collected along the Wasatch front in Utah and spotted onto MG1655 E. coli mutants containing single-gene knockouts of known phage receptors. Single-deletion mutants were generated using Lambda Red recombination as previously described (62). Three FhuA-dependent phages, JLBYU37 (OK272488), JLBYU41 (OK272479), and JLBYU60 (OK272474), were identified using traditional plaque assays with 0.7% top agar (63), and complementation further revealed FhuA as the only phage receptor utilized by these phages. Phage scanning transmission electron microscope images were taken at the Brigham Young University (BYU) Microscopy Laboratory with a Helios NanoLab 600 dual-beam microscope, and phage samples were prepared for imaging by staining with either 2% phosphotungstate and bacitracin or uranyl acetate.
Phage sequencing, genome assembly, and annotation. Phage DNA was isolated using a Norgen Phage DNA Isolation kit (cat. no. 46800) and isolated phage DNA was sequenced using a NEBNext Ultra II FS DNA Library Prep kit (New England Biolabs, cat no. E7805S) and Illumina iSeq 100. Genomes were assembled with Geneious (64) version 8.1 using de novo assembly and medium-low sensitivity as previously described by Sharma et al. (65). All phages circularized upon de novo assembly and the coverage depths for each alignment were 40 to 172 (93 average) for JLBYU37, 13 to 147 (69.6 average) for JLBYU41, and 14 to 369 (175.7 average) for JLBYU60. DNA Master and GeneMark were used to determine open reading frames and each putative gene was annotated using BLASTp (66, 67) from NCBI.
Genomic and structural comparisons. Similar sequences for each phage were collected using BLASTp (66,67) results that had an E value of $-07. Gepard (68) was used to perform whole-genome nucleotide comparisons and individual protein comparisons for putative phage MCP and RBPs.
Generation of DfhuA loop deletion strains. Mutant fhuA plasmids containing loop deletions were constructed using the NEB Q5 Site-Directed Mutagenesis kit (cat. no. E0554S). FhuA loops were replaced with a 4-or 5-amino acid sequence from loop 2 of OmpF as previously designed except for L3 (33). Each mutant plasmid was transformed into MG1655 DfhuA::FRT DfepA::kan strain to test for FhuA functionality and incorporation into the outer membrane with a Fc growth promotion test. Briefly, MG1655 DfhuA:: FRT DfepA::kan mutant strains containing plasmids pJL004 to pJL009 (see Table S1) were tested for FhuA FIG 9 Legend (Continued) icon on the corresponding loop. Results for T5 (blue), phi80 (green), and T1 (red) are based on data reported by Endriss et al. (33), and loops resulting in at least a 10 2 -fold reduction compared to infectivity on the wild-type strain were included. Loops involved in Fc binding (32,33) are indicated by black stars. (C) Evolutionary shifts in FhuA loop preferences among various FhuA-dependent phages. Loop dependency is reported as the percent reduction that loop deletions caused in Teewinot, JLBYU37, JLBYU60, and JLBYU41 infection efficiency. Fold reductions for T5, T1, and phi80 were also based on results obtained by Endriss et al. (33); however, to maintain data comparability for this analysis, hazy spots were reported as being positive for phage infection. functionality and incorporation into the outer membrane by Fc growth promotion using M9 plates containing 100 mM dipyridyl (DIP) and 0.5 mM sodium salicylate. Next, 10 mM Fc was loaded with 10 mM FeCl 3 by incubating the tubes at 37°C for 15 min. Next, 100 mL of ice-cold sterile double-distilled water (ddH 2 O) was added and spread over the outlined plates. A negative-control plate was also made by adding 10 mM HCl to 10 mM FeCl 3 and preparing it as described above. Each strain was picked with a toothpick and patched first onto the FeCl 3 1 HCl negative Fc control plate and then onto the Fc 1 FeCl 3 plate immediately afterward. Plates were pulled and read following a 24-h incubation at 37°C. Mutant plasmids were also transformed into a MG1655 DfhuA::FRT DwaaC::kan strain to test the effects of the LPS inner and outer core on phage infection.
Phage susceptibility spot test. Spot tests were performed with 0.7% soft agar overlays using traditional methods. Briefly, overnight cultures of each mutant strain were normalized to the same optical density at 600 nm (OD 600 ) reading and poured over a prewarmed plate in 4 mL 0.7% top agar. Once solidified, 5 mL of phage serial dilutions was spotted onto each mutant and incubated overnight at 37°C. An average of five replicates were performed for each phage and loop dependency was reported for mutants exhibiting significant fold reduction (P , 0.005) compared to the results for the MG1655 DfhuA::FRT DfepA::kan pJL002 or MG1655 DfhuA::FRT DwaaC::kan pJL002 strain. Levels of significance were calculated using a Student's t test with two-tailed distribution and two-sample equal variance. Hazy spots were scored positive for phage infection so that the fold reductions reported are due to the first dilution without a clearing or a hazy spot.
Protein structural modeling. Putative phage tail fiber-encoding genes were identified with BLASTp (66, 67) and the monomeric structure predictions for JLBYU37, JLBYU41, JLBYU60, and the FhuA L8 mutant were constructed using AlphaFold2 (69). Superimposition of unbound FhuA (2FCP), FhuA with T5 pb5 (8A8C) (51), and the predicted L8 mutant structures was performed in PyMOL (70) using the alignment plugin. A many-to-one super method alignment was performed using an outlier rejection with 5 cycles and a cutoff of 2.0. RMSD and solvent accessible surface area calculations and images were generated using the PyMOL scripts 'ColorbyRMSD' and 'Get_area,' respectively (71). Superimposition of structures of FhuA with T5 pb5 (8A8C) with the predicted structure of JLBYU37 gp22 was performed with the all-against-all structure comparison from Dali (72) and analyzed in PyMOL.
Phage phylogenetic analysis. Similar sequences for each phage MCP were obtained using BLASTp (66,67) and the genome (nucleotide) and potential RBP (amino acid) sequences were collected from the same set of phages using the NCBI database. Multiple sequence alignments were generated with MAFFT (73) using the E-INS-i method (74), which compares conserved domains and accounts for the generalized affine gap cost (75). Model fit predictions (76) and maximum-likelihood tree generation (77) using the WAG1G4 model for the MCP tree and the LG1F1G4 model for substitution for the RBP tree were performed using IQ-TREE (78) with a resampling size of 1,000 and ultrafast bootstrap approximation (79). RBP C-terminal ends were analyzed using Jalview to show percent sequence identity among members of the same genus since little homology was observed when all sequences were compared. The start of the C-terminal end was chosen to be upstream of regions containing reduced percent identity in all genera. Tree generation was performed as outlined above using IQ-TREE and resulted in a tree using the VT1F1G4 model. Differences in tree topology between the full and C-terminal RBP trees were tested for significance using IQ-TREE topology evaluation. This showed significant differences in the Kishino-Hasegawa (80), , and approximately unbiased tests (82) with 10,000 replicates.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.2 MB.

ACKNOWLEDGMENTS
We acknowledge the BYU Electron Microscopy Facility for providing access to the equipment, Michael D. Standing for his expertise in phage imaging, and Ian Baeten for contributions to phage sample preparation. We also acknowledge Parker Folsom for his previous isolation of SO1-like phage vB_EcoD_Teewinot (44) and Daniel Thompson from the Julianne H. Grose lab for his assistance with phage sequencing. We thank Christopher "Topher" Haynie for running the FhuA L8 mutant AlphaFold2 prediction and assistance learning PyMOL script. Phage isolation and characterization was performed with the help of members from the William R. McCleary lab, and we thank Byron Adams and Brett Pickett for their careful reading of the manuscript.