Revealing bacterial targets of growth inhibitorsencoded by bacteriophage T7Shahar Molshanski-Mora, Ido Yosefa, Ruth Kiroa, Rotem Edgara, Miriam Manora, Michael Gershovitsb, Mia Lasersonb,Tal Pupkob, and Udi Qimrona,1

aDepartment of Clinical Microbiology and Immunology, Sackler School of Medicine, and bDepartment of Cell Research and Immunology, George S. WiseFaculty of Life Sciences, Tel Aviv University, Tel Aviv 69978, Israel

Edited* by Sankar Adhya, National Institutes of Health, National Cancer Institute, Bethesda, MD, and approved November 24, 2014 (received for review July13, 2014)

Today’s arsenal of antibiotics is ineffective against some emergingstrains of antibiotic-resistant pathogens. Novel inhibitors of bacte-rial growth therefore need to be found. The target of such bacterial-growth inhibitors must be identified, and one way to achieve thisis by locating mutations that suppress their inhibitory effect. Here,we identified five growth inhibitors encoded by T7 bacteriophage.High-throughput sequencing of genomic DNA of resistant bacterialmutants evolving against three of these inhibitors revealed uniquemutations in three specific genes. We found that a nonessential hostgene, ppiB, is required for growth inhibition by one bacteriophageinhibitor and another nonessential gene, pcnB, is required forgrowth inhibition by a different inhibitor. Notably, we found apreviously unidentified growth inhibitor, gene product (Gp) 0.6,that interacts with the essential cytoskeleton protein MreB andinhibits its function. We further identified mutations in two distinctregions in the mreB gene that overcome this inhibition. Bacterialtwo-hybrid assay and accumulation of Gp0.6 only in MreB-express-ing bacteria confirmed interaction of MreB and Gp0.6. Expression ofGp0.6 resulted in lemon-shaped bacteria followed by cell lysis, aspreviously reported for MreB inhibitors. The described approachmay be extended for the identification of new growth inhibitorsand their targets across bacterial species and in higher organisms.

target identification | host takeover | bacterial cytoskeleton |bacteriophage biology | high-throughput DNA sequencing

Bacteria have evolved to overcome a wide range of antibiotics;in some bacteria, the resistance mechanisms against most

conventional antibiotics have been identified (1, 2). This increasingthreat is spurring the identification of novel antimicrobials againstnovel molecular targets in the pathogens (e.g., refs. 3–6). There arecurrently only a few host molecules targeted by antibiotics. Thesetargets (and examples of the antibiotics against them) are hostRNA polymerase (rifampicin), topoisomerase (quinolones), cellwall (penicillin), membranes (polymyxin), ribosome (tetracyclines,aminoglycosides, macrolids), and synthesis of nucleic-acid pre-cursors (sulfonamides, trimethoprim). Increasing the arsenal ofbacterial targets and antimicrobial drugs against them is valu-able, and novel strategies to increase this repertoire are there-fore of great importance.One strategy for the identification of novel antibacterial tar-

gets is to determine how bacteriophages shut down their host’s bio-synthetic pathways and enslave its machinery during infection.Phages have coevolved with bacteria for over 3 billion years andhave thus developed molecules to specifically and optimally in-hibit or divert key metabolic functions. Examples of bacterialtargets inhibited by phage-derived products include the δ subunitof the DNA polymerase III clamp loader, inhibited by geneproduct (Gp) 8 of the coliphage N4 (7); the Staphylococcus au-reus putative helicase loader, DnaI, inhibited by ORF104 ofbacteriophage 77 (5); a key enzyme of folate metabolism, FolD,inhibited by Gp55.1 of the coliphage T4 (8); and the essentialcell-division protein, filamenting temperature-sensitive mutant Z(FtsZ), inhibited by Gp0.4 of the coliphage T7 (9). These examples

suggest that there are other phage products that may inhibitother bacterial targets.A model for the systematic study of host–virus interactions

and for elucidating phage antibacterial strategies is provided bybacteriophage T7 and its host, Escherichia coli. The laboratorystrain E. coli K-12 shares many essential genes with pathogenicspecies, such as E. coli O157:H7 and O104:H4, and therefore,growth inhibitors against it should prove effective against thesepathogens as well. E. coli has been studied extensively, and theputative functions or tentative physiological roles of over half ofits 4,453 genes have been identified. T7 is a virulent phage thatupon infection of its host, E. coli, produces over 100 progenyphage per host in less than 25 min. It is an obligatory lytic phage,and therefore, its successful growth cycle always results in lysisof the host. Despite extensive knowledge of the T7 phage, themechanism by which it manipulates host functions remains ob-scure. Specific functions have been attributed to over half of the56 T7 Gps (9–12); all of the phage structural Gps are wellcharacterized, as are those Gps that take part in phage DNAreplication. However, most of the remaining Gps that take overthe host machinery have not yet been characterized, and the hostproteins with which they interact have not been identified. Wehypothesized that some of these Gps would inhibit E. coli growthby targeting specific essential proteins.Here we propose an approach to searching for antibacterial

targets using whole-genome DNA sequencing. The basic un-derlying principle is that many resistance mutations againstgrowth inhibitors arise in target genes. Therefore, by expressing

Significance

Antibiotic resistance of pathogens is a growing threat to hu-man health, requiring immediate action. Identifying new geneproducts of bacterial viruses and their bacterial targets mayprovide potent tools for fighting antibiotic-resistant strains.We show that a significant number of phage proteins are in-hibitory to their bacterial host. DNA sequencing was used tomap the targets of these proteins. One particular target wasa key cytoskeleton protein whose function is impaired fol-lowing the phage protein’s expression, resulting in bacterialdeath. Strikingly, in over 70 y of extensive research into thetested bacteriophage, this inhibition had never been charac-terized. We believe that the presented approach may bebroadened to identify novel, clinically relevant bacteriophagegrowth inhibitors and to characterize their targets.

Author contributions: S.M.-M., I.Y., R.K., R.E., M.M., and U.Q. designed research; S.M.-M.,I.Y., R.K., R.E., and M.M. performed research; S.M.-M., I.Y., R.K., R.E., M.M., M.G., M.L.,T.P., and U.Q. analyzed data; and U.Q. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.1To whom correspondence should be addressed. Email: ehudq@post.tau.ac.il.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1413271112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1413271112 PNAS | December 30, 2014 | vol. 111 | no. 52 | 18715–18720

MICRO

BIOLO

GY

a growth inhibitor and identifying resistance mutations usingwhole-genome sequencing, one may be able to identify its target.High-throughput sequencing has been recently used to identifygenetic interactions (e.g., refs. 13–15) but not host–virus inter-actions. Advances in DNA sequencing technology, as well as itsaccessibility and affordability, are enabling its application to theidentification of bacterial targets at high throughput and minimalcost. We used this approach to search for bacterial targets of T7bacteriophage proteins. We cloned most of the uncharacterizedgenes of T7 bacteriophage and tested their inhibition of bacterialgrowth. We then isolated mutants that are resistant to thesegrowth inhibitors and identified the arising mutations by high-throughput sequencing. The mutations arose multiple times, inunique genes for each growth inhibitor, indicating a uniquemechanism for overcoming the inhibition. Of particular interestwas inhibition of the essential cytoskeleton protein, MreB, whichwas further validated by genetic and biochemical methods.

ResultsHigh-Throughput Sequencing Identifies a Characterized Inhibitor–Target Interaction. We hypothesized that expression of growthinhibitors in E. coli would result in resistance mutations in thetarget genes and that these mutations could be identified usinghigh-throughput DNA sequencing of the genomes of the se-lected resistant mutants (Fig. 1). To test the feasibility of thisapproach, we expressed the growth inhibitor Gp0.4 of T7 bac-teriophage, which inhibits the division protein FtsZ (9). Twelvemutants resistant to this growth inhibitor were isolated, and theirgenome was deep-sequenced. We expected to identify a specificmutation that renders FtsZ refractory to Gp0.4 inhibition andhence enables survival of the bacteria encoding this mutation, aswe previously reported (9). Indeed, following analysis of thesequencing results (described in the next section), we identifieda unique mutation encoded only by resistant mutants expressingGp0.4. Other genes were also mutated in these resistant mutants,but none were unique, as they were also mutated in mutantsresistant to other growth inhibitors. The specific mutation was aninsertion mutation of 6 nt in ftsZ, shown to confer resistanceto Gp0.4 expression (9). The frequency of this mutation was61.54%, suggesting it appears in the majority of the 12 resistantmutants (Table 1). As expected, this mutation in ftsZ was onlyobserved in mutants resistant to the Gp0.4 growth inhibitor, andnot in other resistant mutants. This finding validated the feasi-bility of the approach for specifically identifying novel targets ofphage-derived growth inhibitors.

High-Throughput Sequencing Reveals Targets of Growth Inhibitors.To search for targets of novel growth inhibitors, we cloned 14uncharacterized genes from T7 phage on a plasmid: genes 0.5,0.6, 1.1, 1.4, 1.5, 1.6, 1.8, 2.8, 3.8, 4.1, 4.2, 4.3, 4.7, and 5.3.These genes were cloned downstream of an L-arabinose–inducible promoter, plated in the presence of D-glucose (tran-scription repressor) or L-arabinose (transcription activator), andbacterial viability was monitored in both cases. All bacteriaencoding these 14 genes grew well on plates supplemented with0.2% (wt/vol) D-glucose. However, among these 14 genes, fivewere inhibitory to the host upon L-arabinose induction—genes0.6, 1.6, 3.8, 4.3, and 5.3—whereas the other nine were not in-hibitory (Fig. 2). To identify the Gps targeted by these fivegrowth inhibitors, we expressed them in E. coli and isolated 15–83 independent resistant mutants from each of the five trans-formants encoding the growth inhibitors (Table S1). In all cases,resistant mutants emerged at a frequency of ∼2 × 10−7. To dis-criminate between genomic mutations and mutations occurringin the growth inhibitor itself, we excluded mutants in which theplasmid lost its inhibitory effect due to disruptive mutations inthe growth inhibitor. To this end, we extracted plasmids fromthose colonies that have become resistant to the phage inhibitors,

retransformed them into another strain, and checked theirinhibition. We found that all of the resistant mutants formed inresponse to expression of Gp1.6 or 5.3 evolved from mutations inthe plasmid (Table S1), and these mutants were eliminated fromfurther studies. Genomic DNA was extracted from mutantsresistant to genes 0.4 (positive control), 0.6, 3.8, and 4.3, whoseplasmids retained growth inhibition, and sequenced using anIlluminaHiSEq. 2500. The sequencing results were analyzed byapplying several filters. Mutations occurring in less than 5% ofthe reads were discarded. Mutations passing this filter werefurther processed only if more than 90% of the mutations ina single gene were uniquely mapped to a specific group ofresistant mutants (Table 1). As described above, this analysisidentified FtsZ as the top target of Gp0.4, as expected. It alsoidentified the accumulation of missense mutations in mreB,encoding the cytoskeleton protein MreB, in mutants resistant toGp0.6. Moreover, it determined that disruptive mutations ingenes ppiB and pcnB accumulate in mutants resistant to Gp 3.8and 4.3, respectively (Table 1). Note that the above filtersyielded a unique mutated gene for every group of resistantmutants, and each was found to confer resistance, as described inthe following sections.

Nonessential Targets PpiB and PcnB Validated as Contributing toInhibitor Functionality. Both ppiB and pcnB encode nonessentialproteins, and therefore, resistance could be validated using E. colilacking these genes. We transformed the plasmids encoding genes

T7 phageE. coli

Growth inhibitor?YesNo

Resistant mutants

YesNo

High throughput sequencing

Chromosomalmutation?

PBAD

Growth inhibition

Non-inhibitory interaction

No interaction

T7 phage proteinT7 phage mutant proteinE. coli target proteinE. coli target mutant protein

Fig. 1. Schematic representation of the approach used to identify noveltargets of bacteriophage growth inhibitors. E. coli bacteria are transformedwith plasmids cloned with genes derived from T7 bacteriophage down-stream of an inducible promoter (PBAD). Expression of noninhibitory genesresults in viable bacteria, and these genes are excluded from further analysis.Expression of inhibitory genes results in resistant mutants. These resistantmutants are isolated and then tested for plasmid growth inhibition. Plasmidsare extracted from these clones, and their growth inhibition is validatedby retransformation into E. coli bacteria, as described in SI Materials andMethods. Mutants whose plasmids have lost growth inhibition are excludedfrom further analysis. Mutants whose plasmids are inhibitory are suspectedof having mutations in the genome that confer resistance. Genomes of thesemutants are extracted, sequenced, and analyzed.

18716 | www.pnas.org/cgi/doi/10.1073/pnas.1413271112 Molshanski-Mor et al.

3.8 and 4.3 into strains deleted in ppiB and pcnB, respectively.These transformants were then plated on medium having 0.2%D-glucose or 0.2% (wt/vol) L-arabinose to control expression of thegrowth inhibitor. Indeed, as expected, lack of PpiB conferred re-sistance to Gp3.8 expression (Fig. 3A), and lack of PcnB conferredresistance to Gp4.3 expression (Fig. 3B). A positive control,demonstrating that both Gp3.8 and Gp4.3 inhibit growth in thesesettings, was carried out in an isogenic strain, whose growth wasinhibited in the presence of L-arabinose, as expected. These resultsindicated that Gp3.8 and Gp4.3, respectively, require the ppiB andpcnB genes or their products for growth inhibition. One possibilityfor the growth inhibition is that Gp3.8, a putative homing endo-nuclease (10), or Gp4.3 inhibits growth by cleaving unique re-striction sites in the ppiB or pcnB genes, respectively. In thisscenario, the growth inhibitors cleave the chromosome at thesegenes, but deletion of these genes eliminates the restriction sitesfrom the chromosome, thus rescuing the bacteria. If this is thecase, then complementing the deletion mutants with plasmidsencoding ppiB and pcnB genes should not restore growth in-hibition by Gp3.8 and Gp4.3. To test this, we transformed E. colimutants lacking ppiB with plasmids expressing PpiB and Gp3.8or expressing PpiB and a control plasmid. Similarly, we trans-formed E. coli mutants lacking pcnB with plasmids expressingPcnB and Gp4.3 or expressing PcnB and a control plasmid. Wethen monitored growth in the presence of transcription repressor(D-glucose) or inducer (L-arabinose) of the growth inhibitors.Notably, at this stage, we did not add antibiotics to which theppiB or pcnB encoding plasmids confer resistance, and therefore,

plasmid cleavage and consequent loss should not result in bac-terial death. We observed that plasmids encoding the ppiB andpcnB genes did restore growth inhibition in the respective de-letion mutants (Fig. 3 C and D). The control strains encodingonly the complementing genes did not inhibit growth, indicatingthat toxicity is not due to their expression but rather due to thegrowth inhibitors. This result shows that growth inhibition in thepresence of the respective growth inhibitors results from the inter-actions with the PpiB and PcnB proteins and rules out the possi-bility that growth inhibition is caused by DNA cleavage of uniquerestriction sites located in the genes encoding them. Taken together,these results validated our approach and expanded its original scopeto the identification of nonessential genes that mediate growth in-hibition of some growth inhibitors. A putative mechanism bywhich these nonessential genes mediate the growth inhibition isproposed in Discussion.

MreB Validated as the Target of Phage Inhibitor Gp0.6. The thirdidentified putative target was mreB. Because MreB is essential tothe host (16), validation of the interactions could not be carried outby its deletion. We first validated the high-throughput sequenc-ing results by Sanger sequencing of the 10 independent resistantclones. This procedure identified 10 independent mutations in themreB gene (Table S2). These mutations were in agreement with thedeep-sequencing analysis. Notably, the mutations were clustered

Table 1. Resistant mutants revealed by high-throughputsequencing

Mutatedhost gene

Expressed toxic gene

Mutation type Mutationa Gp0.4 Gp0.6 Gp3.8 Gp4.3

ftsZ Insertion 105351 61.54b <5 <5 <5ftsZ Mismatch 105363 7.32 <5 <5 <5ftsZ Mismatch 106107 5.80 <5 <5 <5mreB Mismatch 3398235 <5 12.00 <5 <5mreB Mismatch 3398242 <5 8.22 <5 <5mreB Mismatch 3398244 <5 5.48 <5 <5mreB Mismatch 3398246 <5 9.46 <5 <5mreB Mismatch 3398247 <5 5.41 <5 <5mreB Mismatch 3398635 <5 19.05 <5 <5mreB Mismatch 3398649 <5 20.73 <5 <5mreB Mismatch 3398649 <5 9.76 <5 <5mreB Mismatch 3398649 <5 7.32 <5 <5ppiB Deletion 553223 <5 <5 12.12 <5ppiB Deletion 553226 <5 <5 8.91 <5ppiB Deletion 553302 <5 <5 13.92 <5ppiB Deletion 553342 <5 <5 7.25 <5ppiB Deletion 553377 <5 <5 6.41 <5ppiB Deletion 553378 <5 <5 9.88 <5ppiB Mismatch 553427 <5 <5 7.69 <5ppiB Deletion 553452 <5 <5 6.67 <5ppiB Insertion 553461 <5 <5 6.33 <5pcnB Deletion 158087 <5 <5 <5 7.69pcnB Mismatch 158090 <5 <5 <5 8.16pcnB Mismatch 158091 <5 <5 <5 9.28pcnB Insertion 158432 <5 <5 <5 7.62pcnB Mismatch 158504 <5 <5 <5 11.76pcnB Mismatch 158631 <5 <5 <5 5.61pcnB Mismatch 158645 <5 <5 <5 9.00pcnB Mismatch 158769 <5 <5 <5 7.23pcnB Mismatch 158888 <5 <5 <5 5.15

aNumbering refers to E. coli strain K-12, accession no. NC_000913.bPercent mutations.

Gp0.5

Gp0.6

Gp1.1

Gp1.4

Gp1.5

Gp1.6

None

Gp0.4

Gp3.8

Gp4.1

Gp4.2

Gp4.3

Gp4.7

Gp5.3

Gp1.8

Gp2.8

repressed induced

Fig. 2. Identification of T7 bacteriophage Gps inhibiting bacterial growth.E. coli NEB5α bacteria transformed with a plasmid encoding the indicated Gpwere serially diluted in 10-fold increments. These dilutions (highest dilutionon the Right) were then inoculated on LB agar supplemented with 0.2%D-glucose (repressed) or 0.2% L-arabinose (induced). Images of LB agar platesrepresent one out of two experiments showing similar results.

Molshanski-Mor et al. PNAS | December 30, 2014 | vol. 111 | no. 52 | 18717

MICRO

BIOLO

GY

into two main segments, 154–159 and 288–292, indicating that thepossible binding pocket(s) of Gp0.6 to MreB is near the residuesencoded by these mutations. These results confirmed that re-sistance mutations to Gp0.6 expression accumulate in mreB.To validate experimentally that the mutations in mreB alone

confer resistance to Gp0.6 expression, we constructed a strain car-rying a mutation conferring resistance to Gp0.6 expression,MreBE288G. The mutation was generated by isolation of theGp0.6-resistant mutant, followed by transduction of the mutatedmreB gene into fresh E. coli culture (see SI Materials andMethods for details). Colonies transduced with this mutation orcontrol transductants were tested for resistance to Gp0.6 bytransforming them with the inducible plasmid encoding Gp0.6and plating on Luria–Bertani (LB) medium supplemented withL-arabinose. Three clones carrying the mreB mutation were re-fractory to Gp0.6, whereas three clones lacking the mutationremained sensitive to Gp0.6 (Fig. 4A). These results confirmedthat the mutation E288G in mreB is sufficient to render the cellresistant to Gp0.6 growth inhibition.

Gp0.6 and MreB Interact in the Bacterial Cell. To genetically testwhether Gp0.6 and MreB interact, we used the bacterial two-hybrid system, a bacterial version of the yeast two-hybrid assay(17). This system enables the identification of protein–proteininteractions in vivo. mreB was fused, in frame, to one domain,T18, of the cAMP cyclase gene. Gene 0.6 was fused, in frame, toanother domain, T25, of the cAMP cyclase gene. Under theseconditions, the cAMP cyclase enzyme is activated only if the twodomains are brought into close proximity by the interactions ofthe two tested proteins. Its activity results in LacZ induction,which is easily detected as blue-colored colonies on LB plates

supplemented with 5-bromo-4-chloro-3-indolyl-β-D-galactopy-ranoside (X-gal). As a positive control, we used the two domains,each fused to the leucine zipper domain (ZIP). As a negativecontrol, we used two proteins that do not interact—MreB fusedto the T18 domain and ZIP fused to the T25 domain. E. colicotransformed with the indicated plasmids were plated on LBplates supplemented with the appropriate antibiotics and X-gal.Cotransformation of plasmids encoding MreB fused to T18 andGp0.6 fused to T25 resulted in LacZ activity similar to thepositive control (Fig. 4B). As expected, the negative controlyielded no LacZ activity. These results indicated that MreB andGp0.6 interact in vivo.To further validate these interactions, we attempted to carry

out a pull-down assay. Crude extracts were prepared from bac-teria cotransformed with the Gp0.6-encoding plasmid along withplasmids encoding either MreB or a control protein, CheZ.Notably, Gp0.6 was detected by Western blot only in bacteria

degQ

ppiB

Gp3.8A

repressed inducedB

degQ

pcnB

Gp4.3

repressed inducedC

repressed induced

ppiB

ppiB + Gp3.8

ppiB + control

D

repressed induced

pcnBpcnB + controlpcnB + Gp4.3

Fig. 3. Requirement of ppiB and pcnB for inhibition by Gp3.8 and Gp4.3,respectively. E. coli bacteria lacking the indicated gene and transformedwith a plasmid encoding either Gp3.8 (A) or Gp4.3 (B) were serially diluted in10-fold increments. These dilutions (highest dilution on the Right) were theninoculated on LB agar supplemented with 0.2% D-glucose (repressed) or0.2% L-arabinose (induced). Deletion mutants of ppiB (C) or pcnB (D) weretransformed with plasmids encoding the indicated genes. These trans-formants were serially diluted in 10-fold increments, and dilutions (highestdilution on the Right) were then inoculated on LB agar supplemented with0.2% D-glucose (repressed) or 0.2% L-arabinose (induced). Images of LB agarplates represent one out of two experiments showing similar results.

MreBE288G

MreBwtA Gp0.6

repressed inducedB

ZIP-T18+ T25-ZIP

MreB-T18+ T25-ZIP

MreB-T18+ T25-Gp0.6

C

Anti-CBP

Coomassie

kDaD

Anti-CBP

Coomassie

3525

100705540

2535

Gp0.6

Fig. 4. Validation of Gp0.6 interaction with MreB. (A) E. coli bacteriaencoding either wild-type MreB (MreBwt) or MreBE288G transformed witha plasmid encoding Gp0.6 were serially diluted in 10-fold increments. Thesedilutions (highest dilution on the Right) were then inoculated on LB agarsupplemented with 0.2% D-glucose (repressed) or 0.2% L-arabinose (in-duced). (B) E. coli bacteria resistant to Gp0.6 inhibition (encoding MreBE288G)were cotransformed with the indicated plasmids and serially diluted in 10-fold increments. These dilutions (highest dilution on the Right) were theninoculated on LB agar supplemented with X-gal and incubated for 24 h. (C)E. coli bacteria resistant to Gp0.6 inhibition (encoding MreBE288G) harboringa plasmid encoding a calmodulin-binding peptide (CBP)-tagged Gp0.6 werecotransformed with plasmids encoding the indicated his6-tagged proteins.Protein expression was induced by addition of 0.2% L-arabinose and 1 mMIPTG. Whole-cell lysates were then prepared, and a sample was electro-phoresed on a 12% (wt/vol) polyacrylamide gel. Coomassie staining to vali-date equal loading of the purified proteins (Top) and Western blot usinganti-CBP to detect Gp0.6 (Bottom) were carried out. (D) E. coli bacteria re-sistant to Gp0.6 inhibition (encoding MreBE288G) or encoding the MreBwt

were transformed with a plasmid encoding CBP-tagged Gp0.6. Gp0.6 ex-pression was induced by addition of 0.2% L-arabinose. Coomassie stainingto validate equal loading (Top) and Western blot using anti-CBP to detectGp0.6 (Bottom) were carried out. Gel and plate images represent one out ofat least two experiments showing similar results.

18718 | www.pnas.org/cgi/doi/10.1073/pnas.1413271112 Molshanski-Mor et al.

expressing the MreB but not CheZ (Fig. 4C). This surprisingresult prevented a controlled pull-down assay as the level ofGp0.6 was significantly different between the samples. However,it showed that the increase in the steady-state level of Gp0.6was specific to MreB expression. To strengthen this finding, weshowed that Gp0.6 did not accumulate in E. coli cotransformedwith plasmids encoding four other randomly chosen proteins(Fig. 4C). Furthermore, the steady-state level of Gp0.6 wasdetectible only in E. coli encoding the mreBwt gene in its chro-mosome and not in an isogenic strain encoding a Gp0.6-resistantmreBE288G allele (Fig. 4D). Thus, the specific accumulation ofGp0.6 only in strains encoding a Gp0.6-sensitive MreB proteinfurther confirmed the specific MreB’s interaction with Gp0.6.

MreB Inhibition by Gp0.6 Changes Bacterial Morphology. Inhibitionof MreB is morphologically manifested as “lemon-shaped” bac-teria (18). We therefore expected that if the binding of Gp0.6 toMreB was inhibitory, this morphology would be seen. We alsoexpected that bacteria refractory to Gp0.6 growth inhibitionwould show normal morphology following Gp0.6 expression.Indeed, light microscopy showed a typical lemon shape for E. colibacteria harboring the plasmid encoding Gp0.6 after 2 h of 0.2%L-arabinose induction (Fig. 5A). In contrast, the morphology ofE. coli encoding a resistant mreB mutant was not affected byGp0.6 expression, as expected. Moreover, as reported for MreBinhibition (18, 19), many cells expressing the wild-type MreBunderwent lysis during that time (Fig. 5B). Altogether, theseresults indicated that Gp0.6 inhibits MreB function in vivo.

DiscussionWe demonstrated that high-throughput DNA sequencing of re-sistant bacterial mutants can identify the targets of uncharac-terized growth inhibitors or reveal proteins that mediate theirinhibition. Unique mutations corresponding to single genes werefound in mutants resistant to each of the four tested growth

inhibitors—one that had been previously reported and threeunreported phage-derived growth inhibitors.Deletion of the nonessential gene ppiB eliminated Gp3.8

growth inhibition. PpiB is a cytoplasmic peptidyl-prolyl cis-transisomerase involved in protein folding (20). We speculate that thisprotein is involved in folding of the Gp3.8 growth inhibitor, andthus in its absence, Gp3.8 is not folded into its active state,resulting in a noninhibitory protein.Another nonessential gene, pcnB, eliminated the inhibitory

effect of Gp4.3. PcnB has been shown to maintain high copynumbers of plasmids, and therefore, lack of this protein probablyresults in decreased copy number of the plasmid encoding thegrowth inhibitor. This is likely the reason for the reduced growthinhibition by Gp4.3 (21). Indeed, disruption of pcnB reducedgrowth inhibition also by Gp3.8 (Fig. S1), suggesting that Gp3.8also requires high plasmid copy number to exert its inhibition.Nevertheless, in the initial high-throughput sequencing analysis,we did not find pcnB disrupted in response to the expression ofGp3.8. A possible explanation for this result is that alternativemutations in ppiB, which overcome growth inhibition by Gp3.8,are more frequent than pcnB mutations. Growth inhibition byGp0.4, 0.6, 1.6, and 5.3 was not significantly reduced in the ab-sence of pcnB (Fig. S1) probably because they inhibit growtheven at medium and low doses, and therefore, reduced plasmidcopy number does not alleviate their inhibition.An essential Gp that may serve as a novel target for antibiotics

is MreB, which was inhibited by Gp0.6. We showed that at least11 different mutations can render MreB resistant to Gp0.6 in-hibition. The fact that we obtained only one mutation twicewhereas the other mutations were represented once suggests thatthere are other mutations that alleviate MreB inhibition byGp0.6. The obtained mutations were clustered in two regions,suggesting that Gp0.6 binds to MreB in the pockets formed bythese encoded regions. We further demonstrated that Gp0.6binds to the wild-type MreB and that expression of Gp0.6 inE. coli results in a phenotype that is characteristic of MreB in-hibition. As an inhibitor, Gp0.6 can be used to study bacterialcytoskeleton arrangement, and it can also potentially serve asa new tool in the fight against antibiotic-resistant bacteria. Froma therapeutic viewpoint, MreB is a possible target for antibiotics,as it is an essential bacterial protein that is conserved across mostrod-shaped bacteria and absent in eukaryotes (16, 22). Indeed,some chemical compounds known to depolymerize MreB filamentshave been suggested as antibacterial agents (e.g., refs. 23–25). Fur-ther studies on Gp0.6 to facilitate its use as an antimicrobial com-pound should determine the minimum effective peptide length forinhibition, its MreB-inhibition capability across pathogenic bacterialspecies, its stability inside and outside mammalian tissues, and itspenetration efficiency into the bacteria.What does the phage gain from inhibiting MreB by Gp0.6?

Under standard laboratory conditions (LB medium, 37 °C, andaerated culture), we could not detect any significant advantagefor T7 phages that encode Gp0.6 versus those that lack it.Nevertheless, under other as-yet unidentified conditions, thisgene’s inhibition might be useful for phage growth. It is in-triguing that the newly identified inhibitor of MreB, Gp0.6, isencoded near Gp0.4 (203 bp away), a recently identified in-hibitor of FtsZ (9). MreB and FtsZ are key cytoskeletal proteinsin E. coli. They are responsible for proper chromosome segre-gation and movement, as well as for cell-wall integrity duringelongation and division (22, 26–29). Interestingly, two proteins ofBacillus Φ29 phage interact with FtsZ and MreB, but in contrastto the T7 proteins, they do not inhibit them but rather exploitthem for phage replication (30). The fact that FtsZ has beenshown to confer a competitive advantage to the phage merely viainhibition of FtsZ suggests that this is also the case for MreB.We speculate that inhibiting MreB, a protein that is importantfor cell-wall stability and maintenance (22, 26, 27), results in

MreBP154LMreBwt

4

A

B

4

uninduced

induced

Fig. 5. Morphology of E. coli expressing Gp0.6. (A) E. coli encoding the wild-typeMreB protein (MreBwt) or the Gp0.6-resistant protein MreBP154L harboringthe pBAD33-Gp0.6 plasmid were induced with 0.2% L-arabinose (induced) todrive expression from the plasmid promoter or left uninduced (uninduced), asdescribed in SI Materials and Methods. Images were taken after 2 h of growth.(Scale bar, 4 μm.) Images represent one out of 10 experiments showing similarresults. (B) A single MreBwt bacterium expressing Gp0.6 for ∼2 h is shown inimages taken at five 1-s intervals. Arrow indicates order of events. This series ofimages represents one out of dozens observed.

Molshanski-Mor et al. PNAS | December 30, 2014 | vol. 111 | no. 52 | 18719

MICRO

BIOLO

GY

loosening of the cell wall, which in turn allows smoother releaseof new virions. This speculation is supported by our observationthat bacterial lysis is induced by MreB inhibition by Gp0.6, butdirect evidence is required to establish this. Also worth noting isthe fact that an E. coli growth inhibitor, YeeV, manifests in-hibition of both FtsZ and MreB in the same polypeptide (18).This single-polypeptide structure, which simultaneously inhibitsboth MreB and FtsZ, suggests that Gp0.4 and Gp0.6 performsimilar functions and are thus encoded from a single operon inthe bacteriophage genome.Taken together, the results presented in this article show that

using the described approach, it is possible to search for novelbacterial growth inhibitors and their targets. The finding thata significant proportion (five out of 14) of the uncharacterizedphage products inhibited growth of their host highlights thisapproach’s potential for finding more such products in otherbacterium–phage systems. The robustness of the approach isemphasized by the fact that it revealed an inhibitor that hadremained unknown during almost 70 y of extensive research intothe T7 phage. In addition to targeted essential genes, the ap-proach also identified nonessential genes that mediate thegrowth inhibition.Despite the robustness of the approach, it has some limitations.

If the mutation rate in the chromosome is significantly less thanthat of the growth inhibitor encoded on the plasmid, it becomesdifficult to obtain chromosomal resistance mutants that maintainplasmid growth inhibition. This situation could occur in caseswhere multiple mutations are required to overcome growth in-hibition—for example, if the target is not a single gene, but rathera complex, membrane, cell wall, DNA, or RNA. Depending onthe frequency of the mutation in the genome compared with thaton the plasmid, it may occasionally be possible to select resistant

mutants in such cases as well, after thorough screening. Alterna-tively, two different plasmids encoding the same growth inhibitorcan be cotransformed into the bacteria, thus reducing the proba-bility of growth inhibitor loss that would have to occur simulta-neously on both plasmids. Another downside of our approach isthat the optimal therapeutic agents are those to which resistanceforms at low frequency—that is, those that are best masked fromthis approach. Nevertheless, despite this issue, we believe thatidentifying several such antibacterial substances and combiningthem in a mixture should overcome the formation of simultaneousresistance against all of them. Moreover, the main advantage ofthis approach is that it can recognize targets of uncharacterizedantibacterial substances and may elucidate novel pathways andmechanisms operating in the cell, as demonstrated in this study.These results, in turn, may pave the way for the structural designof more potent inhibitors of those targets, leading to effectiveantibacterial substances.

Materials and MethodsThe reagents, bacterial strains, phages, plasmids, and oligonucleotides usedin this study are listed in SI Materials and Methods and Table S3. Plasmidconstruction, growth inhibition assays, other genetic and biochemical assays,as well as the isolation of resistant mutants, analysis of the high-throughputdata, and microscopy are all described in SI Materials and Methods.

ACKNOWLEDGMENTS. We thank Ken Gerdes and Manuela Castro-Camargofor plasmids and Eyal Molchansky for graphics. The research was funded byEuropean Research Council Starting Grant 336079 (to U.Q.), Israeli Ministryof Health Grant 9988-3 (to U.Q.), Israel Science Foundation Grants 268/14and 1092/13 (to U.Q. and T.P.), and International Reintegration Grant GA-2010-266717 (to R.E.). A fellowship of the Edmond J. Safra Center forBioinformatics at Tel-Aviv University was granted to M.G.

1. Levy SB, Marshall B (2004) Antibacterial resistance worldwide: Causes, challenges andresponses. Nat Med 10(12, Suppl):S122–S129.

2. Spellberg B, Bartlett JG, Gilbert DN (2013) The future of antibiotics and resistance.N Engl J Med 368(4):299–302.

3. Edgar R, Friedman N, Molshanski-Mor S, Qimron U (2012) Reversing bacterialresistance to antibiotics by phage-mediated delivery of dominant sensitive genes.Appl Environ Microbiol 78(3):744–751.

4. Yosef I, Kiro R, Molshanski-Mor S, Edgar R, Qimron U (2014) Different approaches forusing bacteriophages against antibiotic-resistant bacteria. Bacteriophage 4(1):e28491.

5. Liu J, et al. (2004) Antimicrobial drug discovery through bacteriophage genomics. NatBiotechnol 22(2):185–191.

6. Fischbach MA, Walsh CT (2009) Antibiotics for emerging pathogens. Science 325(5944):1089–1093.

7. Yano ST, Rothman-Denes LB (2011) A phage-encoded inhibitor of Escherichia coliDNA replication targets the DNA polymerase clamp loader. Mol Microbiol 79(5):1325–1338.

8. Mattenberger Y, Mattson S, Métrailler J, Silva F, Belin D (2011) 55.1, a gene of un-known function of phage T4, impacts on Escherichia coli folate metabolism andblocks DNA repair by the NER. Mol Microbiol 82(6):1406–1421.

9. Kiro R, et al. (2013) Gene product 0.4 increases bacteriophage T7 competitiveness byinhibiting host cell division. Proc Natl Acad Sci USA 110(48):19549–19554.

10. Molineux IJ (2005) The Bacteriophages, eds Abedon ST, Calendar RL (Oxford UnivPress, Oxford), pp 275–299.

11. Tran NQ, Rezende LF, Qimron U, Richardson CC, Tabor S (2008) Gene 1.7 of bacte-riophage T7 confers sensitivity of phage growth to dideoxythymidine. Proc Natl AcadSci USA 105(27):9373–9378.

12. Sberro H, et al. (2013) Discovery of functional toxin/antitoxin systems in bacteria byshotgun cloning. Mol Cell 50(1):136–148.

13. Davis BM, Waldor MK (2009) High-throughput sequencing reveals suppressors ofVibrio cholerae rpoE mutations: One fewer porin is enough. Nucleic Acids Res 37(17):5757–5767.

14. Stanley SA, et al. (2012) Identification of novel inhibitors of M. tuberculosis growthusing whole cell based high-throughput screening. ACS Chem Biol 7(8):1377–1384.

15. Ioerger TR, et al. (2013) Identification of new drug targets and resistance mechanismsin Mycobacterium tuberculosis. PLoS ONE 8(9):e75245.

16. Kruse T, Bork-Jensen J, Gerdes K (2005) The morphogenetic MreBCD proteins ofEscherichia coli form an essential membrane-bound complex. Mol Microbiol55(1):78–89.

17. Karimova G, Pidoux J, Ullmann A, Ladant D (1998) A bacterial two-hybrid systembased on a reconstituted signal transduction pathway. Proc Natl Acad Sci USA 95(10):5752–5756.

18. Tan Q, Awano N, Inouye M (2011) YeeV is an Escherichia coli toxin that inhibits celldivision by targeting the cytoskeleton proteins, FtsZ and MreB. Mol Microbiol 79(1):109–118.

19. Bendezú FO, de Boer PA (2008) Conditional lethality, division defects, membraneinvolution, and endocytosis in mre and mrd shape mutants of Escherichia coli.J Bacteriol 190(5):1792–1811.

20. Compton LA, Davis JM, Macdonald JR, Bächinger HP (1992) Structural and functionalcharacterization of Escherichia coli peptidyl-prolyl cis-trans isomerases. Eur J Biochem206(3):927–934.

21. Lopilato J, Bortner S, Beckwith J (1986) Mutations in a new chromosomal gene ofEscherichia coli K-12, pcnB, reduce plasmid copy number of pBR322 and its de-rivatives. Mol Gen Genet 205(2):285–290.

22. White CL, Gober JW (2012) MreB: Pilot or passenger of cell wall synthesis? TrendsMicrobiol 20(2):74–79.

23. Iwai N, Nagai K, Wachi M (2002) Novel S-benzylisothiourea compound that inducesspherical cells in Escherichia coli probably by acting on a rod-shape-determiningprotein(s) other than penicillin-binding protein 2. Biosci Biotechnol Biochem 66(12):2658–2662.

24. Gitai Z, Dye NA, Reisenauer A, Wachi M, Shapiro L (2005) MreB actin-mediated seg-regation of a specific region of a bacterial chromosome. Cell 120(3):329–341.

25. Robertson GT, et al. (2007) A novel indole compound that inhibits Pseudomonasaeruginosa growth by targeting MreB is a substrate for MexAB-OprM. J Bacteriol189(19):6870–6881.

26. Carballido-López R (2006) Orchestrating bacterial cell morphogenesis. Mol Microbiol60(4):815–819.

27. Shih YL, Rothfield L (2006) The bacterial cytoskeleton. Microbiol Mol Biol Rev 70(3):729–754.

28. Kruse T, Møller-Jensen J, Løbner-Olesen A, Gerdes K (2003) Dysfunctional MreB in-hibits chromosome segregation in Escherichia coli. EMBO J 22(19):5283–5292.

29. Varma A, Young KD (2009) In Escherichia coli, MreB and FtsZ direct the synthesis oflateral cell wall via independent pathways that require PBP 2. J Bacteriol 191(11):3526–3533.

30. Muñoz-Espín D, et al. (2009) The actin-like MreB cytoskeleton organizes viral DNAreplication in bacteria. Proc Natl Acad Sci USA 106(32):13347–13352.

18720 | www.pnas.org/cgi/doi/10.1073/pnas.1413271112 Molshanski-Mor et al.

Supporting InformationMolshanski-Mor et al. 10.1073/pnas.1413271112SI Materials and MethodsReagents, Strains, and Plasmids. LB medium (10 g/L tryptone, 5 g/Lyeast extract, and 5 g/L NaCl) and agar were from Acumedia.Antibiotics, lysozyme, and L-arabinose were from Calbiochem.D-Glucose, sodium chloride, and the phosphate salts were fromMerck. Isopropyl-β-D-thiogalactopyranoside (IPTG) and X-galwere from Bio-Lab. Restriction enzymes, ligation enzymes, andPhusion High-Fidelity DNA Polymerase were from New Eng-land Biolabs. Benzonase was from Novagen. EDTA-free minicomplete was from Roche. Imidazole was from Sigma. Thebacterial strains, plasmids, and oligonucleotides used in thisstudy are listed in Table S3.

Plasmid Construction. T7 genes were cloned downstream of thearabinose promoter in plasmid pBAD33 as follows. Each T7 genewas PCR-amplified from wild-type T7 phage using primersSM24F 1–15 and the corresponding SM24R 1–15 (Table S3), allencoding restriction sites SalI and PstI, respectively. The PCRfragment was digested with SalI and PstI and then ligated tocompatibly digested plasmid pBAD33. pUT18-Gp0.6 (Table S3)was constructed by PCR amplification of T7 gene 0.6 usingprimers SM28Fb1 and SM28Rb, encoding restriction sitesBamHI and PstI, respectively. The PCR fragment was digestedwith BamHI and PstI and then ligated to compatibly digestedplasmid pUT18. pBAD18-Gp0.6-CBP was constructed by PCR-amplification of T7 gene 0.6 using primers SM29F and SM29R,encoding restriction sites SacI and NotI, respectively. The PCRfragment was digested with SacI and NotI and then ligated tocompatibly digested plasmid pBAD18-CBP.pBAD33-Gp3.8-A was constructed by blunt ligation of a bla

resistance gene using primers IY289F and MG14R (Table S3)from plasmid pUT18 (Table S3) to a pBAD33-Gp3.8 vectorcleaved with BsaAI. pBAD33-Gp4.3-amp was constructed by li-gating gene 4.3 amplified from T7 phage genome using primersSM24F12 and SM24R12 (Table S3) to pBAD33-Gp3.8-ampcleaved with SalI and PstI.

Growth Inhibition Assay. Indicated E. coli strains (Table S3) har-boring the indicated pBAD33-based plasmids were aeratedovernight in LB medium supplemented with 35 μg/mL chloram-phenicol and 0.2% D-glucose at 37 °C. Overnight cultures werediluted 1:100 in fresh LB medium supplemented with 35 μg/mLchloramphenicol and 0.2% D-glucose at 37 °C and aerated to anOD600 of 0.1. Cultures were serially diluted 10-fold and replica-plated on LB agar supplemented with 35 μg/mL chloramphenicoland either 0.2% D-glucose or 0.2% L-arabinose. Plates were in-cubated overnight at 37 °C, and growth inhibition of the plasmid-encoded gene was determined by lack of growth on the L-arabinose–supplemented versus D-glucose–supplemented medium.

Determining Inhibition of Gp3.8 and Gp4.3 in E. coli pcnB and ppiBComplemented with Plasmids Encoding These Genes. E. coliBW25113ΔppiB was cotransformed with pBAD33-Gp3.8-ampand with pCA24N-ppiB or a control plasmid, pACYC177 (Table S3).E. coli BW25113ΔpcnB was cotransformed with pBAD33Gp-4.3-amp and with pCA24N-pcnB or a control plasmid, pA-CYC177. The transformants were aerated overnight in LBmedium supplemented with 25 μg/mL kanamycin, 35 μg/mLchloramphenicol, 100 μg/mL ampicillin, and 0.2% D-glucoseat 37 °C. Overnight cultures were diluted 1:100 in fresh LBmedium supplemented as above at 37 °C and aerated to anOD600 of 0.1. Cultures were serially diluted 10-fold and rep-

lica-plated on LB agar lacking chloramphenicol, but supple-mented with 100 μg/mL ampicillin, 25 μg/mL kanamycin, and either0.2% D-glucose or 0.2% L-arabinose. Plates were incubated over-night at 37 °C, and growth inhibition of the plasmid-encoded genewas determined by lack of growth on the L-arabinose–supplementedversus D-glucose–supplemented medium.

Isolating Resistant Mutants. Colonies of E. coli NEB5α (Table S3)harboring the indicated pBAD33-based plasmids were individuallypicked and independently aerated overnight in separate tubes con-taining LB medium supplemented with 35 μg/mL chloram-phenicol and 0.2% D-glucose at 37 °C. Overnight cultures (3 mL)were centrifuged and the pellet was resuspended in the same vol-ume of LB medium supplemented with 35 μg/mL chloramphenicol.Each culture (100 μL), originating from an independent overnightculture, was spread on a separate LB agar plate supplemented with35 μg/mL chloramphenicol and 0.2% L-arabinose, and incubatedovernight at 37 °C. To avoid sibling mutants, a single resistantcolony was picked from each plate for further analysis, includingplasmid extraction and retransformation, as described in SI Mate-rials and Methods.

Plasmid Extraction and Retransformation. Each independent re-sistant colony, isolated in an independent procedure to avoidsibling mutants, was transferred to 1.5 mL LB medium supple-mented with 35 μg/mL chloramphenicol and 0.2% L-arabinose,and aerated at 37 °C overnight. Cultures were then centrifugedat 12,000 × g for 30 s, at room temperature (RT). Supernatantswere removed and pellets were resuspended in 100 μL ice-cold25 mM Tris·HCl pH 8.0 supplemented with 10 mM EDTA and50 mM D-glucose. After vortexing, 200 μL of 0.2 N NaOH and1% (wt/vol) SDS were added to the suspension and mixed byinverting the tube several times. The mixture was incubated on icefor 5 min, and 150 μL of 3 M potassium acetate and 6% (vol/vol)glacial acetic acid were added. The mixture was incubated again onice for 5 min and centrifuged at 12,000 × g for 5 min at RT ina bench-top centrifuge. The supernatants were transferred to a newtube, and 0.7 vol of isopropanol was added. The tube was mixed andincubated at RT for 2 min. The mixture was centrifuged at 12,000 × gfor 5 min. The supernatant was removed, and 1 mL of 70% (vol/vol)ethanol was added. The mixture was centrifuged at 12,000 × g for5 min. The supernatant was removed, and the pellet was driedfor 5 min. Finally, the pellet was resuspended in 30 μL Tris-EDTA. A 1-μL aliquot of this suspension was added to freshelectrocompetent cells and plated on LB medium supplementedwith 35 μg/mL chloramphenicol and 0.2% D-glucose. The re-sultant colonies were picked and subjected to toxicity assay.

High-Throughput Sequence Generation and Analysis. Cultures from10 to 12mutants resistant to either Gp0.4, Gp0.6, Gp3.8, or Gp4.3(Table S1) were combined into four different tubes followedby genomic DNA extraction using the NucleoSpin Tissue kit(Macherey–Nagel). Genomic DNA from each of the four in-hibitor-resistant groups of mutants was processed using Illumina’skit (cat. no. 5025064) according to the manufacturer’s in-structions. Reads for each group of mutants were identifiedbased on a barcode sequence. The following barcodes were usedfor the four different groups: GATCAG (Gp0.4), TAGCTT(Gp0.6), GGCTAC (Gp3.8), and GTGGCC (Gp4.3). Sequenc-ing was carried out using Illumina’s HiSEq. 2500, in rapid mode,with single-read runs of 100 bp. The average DNA coverage perresistant strain was 95 reads per bp. The sequencing results were

Molshanski-Mor et al. www.pnas.org/cgi/content/short/1413271112 1 of 5

used to identify putative E. coli genes interacting with eachgrowth inhibitor. The 100-bp reads were aligned to the referencesequence of the E. coli strain K-12 genome (NC000913) usingBowtie (version 2.0) (1). The alignment was visualized usingSAMTools (2). Next, for each inhibitor-resistant group, thefrequency of each mutation (either a specific base pair sub-stitution or an insertion–deletion mutation) was determined asthe percentage of reads that contained the mutation out of thetotal number of reads that covered that locus. Low-frequencymutations—that is, those with frequency lower than 5%—werefiltered out to decrease noise (e.g., sequencing errors). Genesthat harbor such mutations are putative target genes. The list ofgenes was found to include some genes that are shared acrossmultiple strains. Such genes are likely not to be specific fora given growth inhibitor, and we hence aimed to filter out thesegenes. Specifically, for each putative gene, we compiled a list ofmutations. For a gene to be further considered, we requestedthat at least 90% of these mutations appear only in a single strain(i.e., unique mutations). Genes that did not pass this 90%threshold were discarded. The 90% threshold was manuallydetermined to balance specificity and sensitivity. The final out-put of this Python algorithm is a list of putative target genes foreach combined group of resistant mutants to a specific inhibitor.Validation of the sequencing results by Sanger sequencing, mreBtransduction, bacterial two-hybrid assay, and pull-down assaysare all described in SI Materials and Methods.

Validation of High-Throughput Analysis by Sanger Sequencing. Eachindependent mutant resistant to Gp0.6 was streaked on an LBagar plate supplemented with 35 μg/mL chloramphenicol and0.2% L-arabinose. A single colony from this plate was picked,and its DNA served as the template for amplification of themreB gene using primers SM27F and SM27R. The PCR productwas purified and Sanger-sequenced using primers SM27F andSM27R (Table S3).

Transduction ofmreBMutation.The gene encoding MreBE288G wasgenerated de novo in a ΔdegQ::kan strain (Table S3) by isolationof spontaneous resistant mutants as described above. degQis located ∼20 kb away from mreB. The probability of co-transduction of a marker located 20 kb from another gene is∼50% (3), and we selected ΔdegQ::kan as the donor strain in theP1 transduction to transduce MreBwt and MreBE288G at equalfrequencies. Overnight culture of this strain was diluted 1:100 inLB medium supplemented with 25 μg/mL kanamycin, 5 mMCaCl2, and 0.2% D-glucose. After 1 h of aeration at 37 °C, 0, 10,or 50 μL of phage P1 was added. Cultures were aerated for 1–3 huntil lysis occurred. The resulting P1 lysate (0, 10, or 50 μL) wasmixed with 100 μL overnight culture of the recipient strain K-12(Table S3) and 1.25 μL of 1 M CaCl2. After incubation at 30 °Cfor 30 min, 100 μL of 1 M Na-citrate (Merck) and 500 μL LBmedium were added to each tube. Cultures were incubated at37 °C for 1 h, and then 3 mL of warm LB medium supplementedwith 0.7% (wt/vol) agar was added to each tube. The suspension

was poured onto a plate containing 25 μg/mL kanamycin andincubated overnight. Emerging transductant colonies werestreaked several times on LB plates containing 25 μg/mL kana-mycin, and their mreB gene was then sequenced to identifycolonies encoding MreBwt and MreBE288G. Three independentcolonies from each type were used in the toxicity assay.

Bacterial Two-Hybrid Assay. E. coli strain BTH101, transducedwith the gene encoding MerBE288G, was cotransformed with theindicated plasmids. Cotransformants were then plated on LBplates supplemented with 100 μg/mL ampicillin, 50 μg/mLkanamycin, 0.5 mM IPTG, and 40 μg/mL X-gal. Plates werewrapped in aluminum foil, incubated at 30 °C for 24 h, and thenscanned on an Epson perfection V700 scanner.

Assessment of Gp0.6 Steady-State Levels upon Coexpression of His6-Tagged Proteins. E. coli K-12 ΔdegQ mreBE288G was transformedwith pCA24N encoding either mreB or cheZ, ormdoG or bioB orlyx or norV, all with a His6 tag. Bacteria was cotransformed withpBAD18-Gp0.6 with a CBP tag. Overnight cultures of thesecotransformants, grown at 37 °C, were diluted 1:50 in 10 mL LBmedium and aerated at 37 °C to an OD600 of 0.5. IPTG (1 mM)and 0.2% L-arabinose were added, and growth continued for anadditional 2 h. The cultures were then centrifuged at 4 °C for 20min at 4,000 × g. The pellet was resuspended in 0.5 mL lysisbuffer (20 mM sodium phosphate buffer pH 7.4, 300 mM NaCl,10 mM imidazole, 200 μg/mL lysozyme, 2.5 U/mL benzonase,and minicomplete EDTA-free tablet) and stored at –80 °C for atleast 16 h. The samples were thawed in a RT water bath andincubated for 60 min on ice. They were then frozen for 2 min inliquid nitrogen and thawed for 3 min in a RT water bath. Thisfreeze-thaw cycle was repeated twice more. Samples weretransferred 5 times in a 1 mL syringe with a 23 G needle. Celldebris was removed by centrifugation at 4 °C for 10 min at19,150 × g. Protein concentration was measured with a Nanodrop2000 spectrophotometer (Thermo), and equal amounts of totalprotein were loaded on a 12% (wt/vol) polyacrylamide gel.Coomassie staining was performed to validate equal amount ofsample loading. Western blotting was performed to detect theCBP-tagged Gp0.6 using anti-CBP epitope tag (Merck), ac-cording to the manufacturer’s instructions.

Microscopy. The specified E. coli strains harboring pBAD33-Gp0.6 or pBAD33 plasmids were aerated overnight in LB me-dium supplemented with 35 μg/mL chloramphenicol and 0.2%D-glucose at 37 °C. These overnight cultures were diluted 1:50 infresh LB medium supplemented with 35 μg/mL chloramphenicol at37 °C, and aerated to an OD600 of 0.1. Cultures were then eitherinduced with 0.2% L-arabinose or uninduced, for 2 h at 37 °C withshaking. Each sample was centrifuged at 12,000 × g for 2 min, atRT. The bacteria were resuspended to an OD600 of 10. Bacteria(5 μL per slide) were viewed under an Olympus Provis AX-70microscope. Images were taken by an Olympus DP72 camera.

1. Stanley SA, et al. (2012) Identification of novel inhibitors of M. tuberculosis growthusing whole cell based high-throughput screening. ACS Chem Biol 7(8):1377–1384.

2. Ioerger TR, et al. (2013) Identification of new drug targets and resistance mechanismsin Mycobacterium tuberculosis. PLoS ONE 8(9):e75245.

3. Wu TT (1966) A model for three-point analysis of random general transduction. Ge-netics 54(2):405–410.

Molshanski-Mor et al. www.pnas.org/cgi/content/short/1413271112 2 of 5

Gp0.6Gp1.6

NoneGp0.4

Gp3.8Gp4.3

Gp5.3

repressed induced

Gp0.6Gp1.6

None

Gp0.4

Gp3.8Gp4.3

Gp5.3

ydhQ

pcnB

Fig. S1. Inhibition of bacterial growth in the absence of pcnB. E. coliΔpcnB or control E. coliΔydhQ transformed with plasmids encoding the indicated growthinhibitors were serially diluted in 10-fold increments. These dilutions (highest dilution on the Right) were then inoculated on LB agar supplemented with 0.2%D-glucose (repressed) or 0.2% L-arabinose (induced). Images of LB agar plates represent one out of two experiments showing similar results.

Table S1. Isolation of independent resistant mutants for eachtoxin

ToxinResistant colonies

isolatedColonies encodingfunctional toxin

Colonies lackingfunctional toxin Sequenced

0.4 29 28 1 120.6 17 10 7 101.6 15 0 15 03.8 30 13 17 104.3 83 17 66 105.3 25 0 25 0

Table S2. Sanger sequencing of the individual Gp0.6-resistantmreB mutants

Resistant mutant Mutationa Amino acid substitution

1 G3,398,649A P154L2 G3,398,649A P154L3 G3,398,649T P154Q4 G3,398,649C P154R5 T3,398,635C T159A6 T3,398,247A E288V7 C3,398,246A E288D8 C3,398,244A R289L9 C3,398,242A G290C10 A3,398,235T V292E

aNumbering refers to E. coli strain K-12, accession no. NC_000913.

Molshanski-Mor et al. www.pnas.org/cgi/content/short/1413271112 3 of 5

Table S3. Bacterial strains, plasmids, and oligonucleotides used in this study

Bacteria/plasmids/oligonucleotides Description Source

Bacterial strainsNEB5α F−, ϕ80lacZΔM15Δ(lacZYA-argF), U169, deoR, recA1, endA1, hsdR17

(rk−, mk

+),gal −,phoA, supE44, λ−, thi−1, gyrA96, relA1New England Biolabs

BW25113ΔdegQ F−, Δ(araD-araB)567, ΔlacZ4787(::rrnB-3), λ−, rph-1, ΔdegQ::kan,Δ(rhaD-rhaB)568, hsdR514

(1)

BW25113ΔydhQ F−, Δ(araD-araB)567, ΔlacZ4787(::rrnB-3), λ−, rph-1, ΔydhQ::kan,Δ(rhaD-rhaB)568, hsdR514

(1)

BW25113ΔppiB F−, Δ(araD-araB)567, ΔlacZ4787(::rrnB-3), λ−, rph-1, ΔppiB::kan,Δ(rhaD-rhaB)568, hsdR514

(1)

BW25113ΔpcnB F−, Δ(araD-araB)567, ΔlacZ4787(::rrnB-3), λ−, rph-1, ΔpcnB::kan,Δ(rhaD-rhaB)568, hsdR514

(1)

BW25113ΔdegQ mreBE288G F−, Δ(araD-araB)567, ΔlacZ4787(::rrnB-3), λ−, rph-1, ΔdegQ::kan,Δ(rhaD-rhaB)568, hsdR514, mreBE288G

This study

BTH101 mreBE288G F-, cya-99, araD139, galE15, galK16, rpsL1 (Str r), hsdR2, mcrA1,mcrB1mreBE288G

This study

K12 mreBE288G Wild-type, mreBE288G This studyAG1 endA1 recA1 gyrA96 thi-1 relA1 glnV44 hsdR17(rK- mK+) (2)

PlasmidspBAD33 L-arabinose–inducible expression vector, p15A origin of replication,

araC, camr(3)

pBAD33-Gp0.4 pBAD33 cloned with T7 0.4 gene, camr This studypBAD33-Gp0.5 pBAD33 cloned with T7 0.5 gene, camr This studypBAD33-Gp0.6 pBAD33 cloned with T7 0.6 gene, camr This studypBAD33-Gp1.1 pBAD33 cloned with T7 1.1 gene, camr This studypBAD33-Gp1.4 pBAD33 cloned with T7 1.4 gene, camr This studypBAD33-Gp1.5 pBAD33 cloned with T7 1.5 gene, camr This studypBAD33-Gp1.6 pBAD33 cloned with T7 1.6 gene, camr This studypBAD33-Gp1.8 pBAD33 cloned with T7 1.8 gene, camr This studypBAD33-Gp2.8 pBAD33 cloned with T7 2.8 gene, camr This studypBAD33-Gp3.8 pBAD33 cloned with T7 3.8 gene, camr This studypBAD33-Gp3.8-amp pBAD33 cloned with T7 3.8 gene, ampr This studypBAD33-Gp4.1 pBAD33 cloned with T7 4.1 gene, camr This studypBAD33-Gp4.2 pBAD33 cloned with T7 4.2 gene, camr This studypBAD33-Gp4.3 pBAD33 cloned with T7 4.3 gene, camr This studypBAD33-Gp4.3-amp pBAD33 cloned with T7 4.3 gene, ampr This studypBAD33-Gp4.7 pBAD33 cloned with T7 4.7 gene, camr This studypBAD33-Gp5.3 pBAD33 cloned with T7 5.3 gene, camr This studypACYC177 Negative control for complementation assay (4)pUT18C-zip Positive control for bacterial two-hybrid assay (5)pKT25-zip Positive control for bacterial two-hybrid assay (5)pKT25 Backbone vector for bacterial two-hybrid assay (5)pUT18 Backbone vector for bacterial two-hybrid assay (5)pUT18-mreB pUT18 cloned with mreB, ampr (6)pKT25-Gp0.6 pKT25 cloned with mreB, kanr This studypCA24N-mreB pCA24N cloned with His6-mreB, camr (2)pCA24N-cheZ pCA24N cloned with His6-cheZ, cam

r (2)pCA24N-mdoG pCA24N cloned with His6-mreB, camr (2)pCA24N-bioB pCA24N cloned with His6-cheZ, cam

r (2)pCA24N-lyx pCA24N cloned with His6-mreB, camr (2)pCA24N-norV pCA24N cloned with His6-cheZ, cam

r (2)pCA24N-ppiB pCA24N cloned with His6-ppiB, cam

r (2)pCA24N-pcnB pCA24N cloned with His6-pcnB, cam

r (2)pBAD18-CBP L-arabinose–inducible expression vector, ampr Laboratory collectionpBAD18-Gp0.6-CBP pBAD18 cloned with T7 Gp0.6-CBP This study

Oligonucleotides 5′→3′ Amplified geneSM24F1 atatGTCGACttacttatgagggagtaatga 0.5SM24R1 attaCTGCAGtcatttgcgtagtgccccttSM24F2 atatGTCGACaaaggggcactacgcaaatg 0.6SM24R2 attaCTGCAGtcagtgcgttctgtcatagcSM24F3 atatGTCGACaagagaggactttaagtatg 1.1SM24R3 attaCTGCAGttactgaccctcccagctacSM24F4 atatGTCGACttataaggag acactttatg 1.4

Molshanski-Mor et al. www.pnas.org/cgi/content/short/1413271112 4 of 5

Table S3. Cont.

Bacteria/plasmids/oligonucleotides Description Source

SM24R4 attaCTGCAGttaccagttaactatactccSM24F5 atatGTCGACtaaaggaggtacacaccatg 1.5SM24R5 attaCTGCAGttaagcgtgaccatctggcaSM24F6 atatGTCGACactaaaggagacactatatg 1.6SM24R6 attaCTGCAGtcagaacacctccttgattcSM24F7 atatGTCGACacataaggataaatgttatg 1.8SM24R7 attaCTGCAGtcattgacctctcgatttccSM24F8 atatGTCGACaagacggagacttctaagtg 2.8SM24R8 attaCTGCAGttagcgccgtaacctgccatSM24F9 atatGTCGACgcattggaggtcaaataatg 3.8SM24R9 attaCTGCAGctatcggaatcgtgcgaattgSM24F10 atatGTCGACcaactgtgggagtagtgatg 4.1SM24R10 attaCTGCAGtcattggtttacctcctgagSM24F11 atatGTCGACagacaaaggtaaagcacatg 4.2SM24R11 attaCTGCAGttagaatgggactctccagcSM24F12 atatGTCGACactaaaggagacacaccatg 4.3SM24R12 attaCTGCAGttactcaaagaatttggaaagSM24F13 atatGTCGACctataggagatattaccatg 4.7SM24R13 attaCTGCAGttatcgtgacttaacaatctSM24F14 atatGTCGACgatacaggaggctactcatg 5.3SM24R14 attaCTGCAGctatagttttatgcctttgtSM24F15 atatGTCGACgaggaggatgaagagtaatg 0.4SM24R15 attaCTGCAGtcactcagcagattctaaagSM27F gttatgcgtattctcgtatcag mreBSM27R gtggttggtaaagtaagcggSM28Fb1 atatCTGCAGtatgatgaagcactacgttatgc 0.6 (BACTH)SM28Rb attaGGATCCaagtgcgttctgtcatagccggSM29F atat GAGCTCaaaggggcactacgcaaatg 0.6 (expression assay)SM29R attaGCGGCCGCgtgtgcgttctgtcatagccggIY289F gtgcagatagagttgcccattagctgtctttcgctgctgagggtgacgatcccgcaaatgtgcgcggaaccccta blaMG14R taagtctagattcttttgccgacagaatcgggcgagaagaggtaccaggcgcggtttgatgatccttttaaattaaaaat bla

aRestriction enzyme recognition sites capitalized.

1. Baba T, et al. (2006) Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: The Keio collection. Mol Syst Biol 2:0008.2. Kitagawa M, et al. (2005) Complete set of ORF clones of Escherichia coli ASKA library (a complete set of E. coli K-12 ORF archive): Unique resources for biological research. DNA Res

12(5):291–299.3. Guzman LM, Belin D, Carson MJ, Beckwith J (1995) Tight regulation, modulation, and high-level expression by vectors containing the arabinose PBAD promoter. J Bacteriol 177(14):

4121–4130.4. Chang AC, Cohen SN (1978) Construction and characterization of amplifiable multicopy DNA cloning vehicles derived from the P15A cryptic miniplasmid. J Bacteriol 134(3):1141–1156.5. Karimova G, Pidoux J, Ullmann A, Ladant D (1998) A bacterial two-hybrid system based on a reconstituted signal transduction pathway. Proc Natl Acad Sci USA 95(10):5752–5756.6. Fenton AK, Gerdes K (2013) Direct interaction of FtsZ and MreB is required for septum synthesis and cell division in Escherichia coli. EMBO J 32(13):1953–1965.

Molshanski-Mor et al. www.pnas.org/cgi/content/short/1413271112 5 of 5