Genome-wide screens reveal Escherichia coli genes required for growth of T1-like phage LL5 and V5-like phage LL12.

The host factor requirements of phages and mechanisms of mutational phage insensitivity must be characterized for rational design of phage cocktails. To characterize host dependencies of two novel Escherichia coli phages, the T1-like siphophage LL5 and the V5-like myophage LL12, forward genetic screens were conducted against the Keio collection, a library of single non-essential gene deletions in E. coli str. BW25113. These screens and subsequent experiments identified genes required by phages LL5 and LL12. E. coli mutants deficient in heptose II and the phosphoryl substituent of heptose I of the inner core lipopolysaccharide (LPS) were unable to propagate phage LL5, as were mutants deficient in the outer membrane protein TolC. Mutants lacking glucose I of the LPS outer core failed to propagate LL12. Two additional genes encoding cytoplasmic chaperones, PpiB and SecB, were found to be required for efficient propagation of phage LL5, but not LL12. This screening approach may be useful for identifying host factors dependencies of phages, which would provide valuable information for their potential use as therapeutics and for phage engineering.


Determination of genes required for phage propagation
The Keio collection consists of a total of 3,985 individual gene knockout mutants in the E. coli K-12 strain BW25113. Each gene knockout is represented twice in the collection (the results of two independent experiments) [1], thus the total collection contains 7,970 mutants, with each independent gene knockout mutants represented with even and odd numbers. Phages LL5 and LL12 were screened against the entire odd-numbered series of 3,985 Keio single-gene knockouts as described in Materials and Methods. E. coli mutants that were unable to support phage growth, as indicated by their growth to an OD550 of at least 0.2 or 0.11 at 8 h in the presence of phage LL5 or LL12, respectively, were considered positive hits in this initial screen. Using this selection criteria, 37 knockout mutants (21 mutants for each phage) were selected for further investigation (Tables S1, S2). For each of these initial hits, the screening experiment was repeated using the same odd-numbered mutant and its even-numbered counterpart from the collection. From this second experiment, 11/21 mutants identified against LL5 and 9/21 mutants identified against LL12 were found to produce the same phenotype in at least one of the paired knockouts, and these were retained for further study.
The efficiency of plating (EOP) of phage LL5 on the retained mutant strains was determined by spot titer.
The observed plating efficiency of phage LL5 was reduced by at least ~20-fold in eight mutants (Table S1).
This plating defect was confirmed by titration of LL5 in full plate assays, in which only four mutants showed an EOP reduction of ~20-fold or greater. In order to confirm the plating phenotype in a clean genetic background, the kanamycin resistance cassettes from these Keio mutants were transduced by P1 into the parental E. coli strain BW25113. Markers could be transduced from all four Keio mutants into the parental strain, and three showed a similar plating defect as the corresponding Keio mutant, indicating the phenotype was linked to the disrupted locus (Table S1). One mutant, waaQ, showed a ~25-fold reduction in EOP in the Keio mutant but its P1 transductant exhibited only a very mild EOP defect of 0.3, despite having the waaQ deletion confirmed in the transductant by PCR and sequencing. This suggests an abnormality or additional defect in the original waaQ Keio mutant; this mutant was not examined further.
The same approach was applied to confirm the phenotypes of the Keio mutants identified from the screens against phage LL12 (Table S2). Spot titer assays showed that the EOP of phage LL12 was reduced in only three of the nine initially identified Keio mutants. This plating defect could be replicated via full plate plaque assay in all three mutants. Only one of these mutants could be P1 transduced into the parental strain BW25113 and same plating phenotype was observed in the P1 transductant as in the Keio mutant (Table   S2). The other two Keio mutants were resistant to P1 infection and could not be transduced.
Based on the genes identified in these initial screens, additional mutants from the odd-and even-numbered Keio sets were subjected to targeted re-screening by directly determining the phage EOP by the spot method. For both phages, genes involved in LPS biosynthesis (waaP for LL5, and waaP, waaG and gmhA for LL12, Tables S1, S2) were identified and confirmed, but these genes represented only parts of the known biosynthetic pathway. Additional Keio mutants in gmhA, waaE, waaC, waaF, waaY, waaI, and waaB were obtained and confirmed by PCR and sequencing of the mutant locus. A tolC mutant was obtained and tested based on the similarity between LL5 and phage TLS, which is known to use TolC as a receptor. Strong EOP defects (< 10 -7 ) were identified in the gmhA, waaE, waaC, waaF, and tolC mutants against LL5, and in waaE, waaC and waaF in LL12 (an EOP defect in gmhA against LL12 was already identified in the initial screen) ( Table 2).

Discovery rates from gene knockout libraries
From the initial 37 "hits" identified in the untargeted screen, it was established that three E. coli genes were needed each for of phage LL5 and LL12 propagation (Tables S1, S2), which gives a false positive gene discovery rate of ~86% for both phages in the initial screen. It was noted that many genes predicted to play roles in phage propagation were not "hit" in the initial screen, so targeted screens were performed against additional mutants. Ultimately, eight genes affecting LL5 propagation and six genes affecting LL12 propagation were confirmed (Table 2), which translates to a false-negative gene discovery rate of at least 62% for LL5 and 50% for LL12 in the initial untargeted screen. In this process, it was discovered that in several instances the mutant genes were still intact in either the odd-or even-numbered sets, and in a few cases genes were still intact in both sets. These findings highlight the potential utility of using multiple independently-generated mutant libraries when conducting large forward genetic screens as described here. Gene discovery rates are likely to be higher for biological pathways comprised of multiple genes, as this increases the probability that at least some genes in the pathway will be detected in initial screens. This highlights that the results from high-throughput forward genetic screens should be interpreted only after rigorous confirmation of the mutant genotypes and their phenotypes. Table S1. Continued. Targeted re-screening. Based on the results of the initial screening, individual mutants were confirmed for gene deletions and obtained from other sources as necessary. The genes waaC, waaF, and tolC were found to be intact in the copy of the Keio collection used for initial screening.