High-resolution view of bacteriophage lambda geneexpression by ribosome profilingXiaoqiu Liua,1, Huifeng Jiangb,1,2, Zhenglong Gub, and Jeffrey W. Robertsa,3

aDepartment of Molecular Biology and Genetics and bDivision of Nutritional Sciences, Cornell University, Ithaca, NY 14853

Contributed by Jeffrey W. Roberts, May 23, 2013 (sent for review April 19, 2013)

Bacteriophage lambda is one of the most extensively studiedorganisms and has been a primary model for understanding basicmodes of genetic regulation. Here, we examine the progress oflambda gene expression during phage development by ribosomeprofiling and, thereby, provide a very-high-resolution view oflambda gene expression. The known genes are expressed in a pre-dictable fashion, authenticating the analysis. However, many pre-viously unappreciated potential open reading frames becomeapparent in the expression analysis, revealing an unexpectedcomplexity in the pattern of lambda gene function.

Bacteriophage lambda is an original and exemplar organismthat has guided the discovery of basic genetic regulatory pro-

cesses, including transcription repression, activation, and anti-termination (1, 2). Lambda has provided an important model tounderstand the interaction of a virus with its host. Programs oflambda gene expression that establish and maintain the prophagestate, and that mediate the lytic pathway of growth upon infectionor induction, are well understood in terms of the basic biochemicalmechanisms and the timing of regulatory protein function. Boththe classic, focused analysis of gene expression and the modernglobal approaches such as array hybridization (3, 4) have beenapplied to elucidate the complex pattern of gene function duringlambda growth. A recent method, ribosome profiling, providesa further detailed and precise view of gene expression by capturingthe instantaneous translation sites of all of the ribosomes in a cell(5–8). We have applied ribosome profiling to the process of lyticgrowth of bacteriophage lambda to map in detail the expressionof lambda proteins and to infer unique loci of translation. We, ofcourse, confirm the known patterns of gene expression, but we alsoexpand understanding in several ways, including a complete cat-alog of gene expression, the discovery of unique functionalopen reading frames (ORFs), and the discovery of bacterialgenes expressed during phage development.

Results and DiscussionOverview of Method and Approach. We chose temperature in-duction of the classic cI857 repressor mutant of lambda in a ly-sogen of Escherichia coli MG1655 to synchronize the lytic process,sampling the lysogen and control nonlysogen both before and 2, 5,10, and 20 min after shifting the temperature from 32 °C to 42 °C.The last sample time was chosen to be before any significant celllysis, but during the later stages of lytic gene expression. Se-quencing produced ∼106 ribosome prints per sample, which weremapped onto both phage and bacterial genomes and visualized inthe Gbrowse genome browser at EcoWiki (9). Total protectednucleotides within ORFs were summed to determine the densityof translation of each reading frame (5). We take this number toindicate the overall rate of translation, although obviously we areassuming that pauses in translation do not excessively affect theoverall rate. Because the expression level is not normalized for thecopy number of the replicating phage DNA, it thus encompassesboth the effect of DNA template availability on mRNA synthesisand the efficiency of utilization of messengers by ribosomes.

General Features of the Translation Pattern. As shown in moredetail below, the pattern of ribosome occupancy over individualprotein coding sequences follows that previously reported, dis-playing highly asymmetric distributions of protected sites due to

strong apparent translation pauses within coding segments (10).There frequently are clusters over translation initiation andtermination sites, reflecting slow steps at these stages of trans-lation. As reported by Li et al. (10) and discussed further below,pauses are associated with upstream Shine–Dalgarno sequences.The general expression pattern matches in good detail expect-

ations from decades of detailed study of lambda gene regulation(Fig. 1). All known lambda genes and previously annotated ORFsare expressed during lambda development (except orf206b andprobably NP_59778.1), as are the newly identified translatedORFs that we report below.

The Uninduced Lysogen. In the uninduced lambda lysogen, cI, rexA,rexB, lom, and bor are the major bacteriophage genes beingtranslated, in agreement with expectation (Dataset S1). Othersappear significantly over background, including the early genesea8.5 and ea59 (both of unknown function), and the immediateearly genes N and cro; the latter presumably become transcribedwhen repression occasionally fails, although there clearly is notenough expression of gene N to allow expression of most of thedelayed early set of genes that depend on the geneN transcriptionantiterminator. A few other genes appear detectably over back-ground (Dataset S1). Two array experiments (3, 4) measuringRNA agreed in identifying in uninduced lysogens most of the setof five that we find, but each report found other genes significantlyexpressed that we do not find and, in fact, there was littleagreement between the two array studies about these others.Differences with our measurements presumably reflect the factthat RNAs are not uniformly translated, in addition to uncer-tainties of array measurements.

Gene Expression During Lytic Growth. After repression is relieved,lambda gene expression occurs in two waves (11, 12). De-repression enables promoters pL and pR to function, providingexpression of genes N and cro at the earliest time; these genes arethe only two lytic genes highly expressed at 2 min after de-repression (Fig. 2 A and B). N is a transcription antiterminatorthat potentiates transcription of the early genes to the right of Nand cro. Early genes are expressed increasingly from 5 to 10 min,and then less at 20 min (Fig. 2 A and B). The decrease in earlygene expression in the last interval is attributed to the activity ofthe lytic repressor Cro, which represses both pL and pR as itsconcentration builds up in the cell (13). The last of the early geneson the right is Q, which encodes an antiterminator that providesexpression of all of the late genes (14). Late gene expression onlyappears significantly at 10 min, and then increases greatly by20 min (Fig. 2C), reflecting its dominance in the last period of

Author contributions: X.L. and J.W.R. designed research; X.L. performed research; X.L., H.J.,and Z.G. analyzed data; and X.L. and J.W.R. wrote the paper.

The authors declare no conflict of interest.

Data deposition: The data reported in this paper have been deposited in the Gene Ex-pression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. GSE47509).1X.L. and H.J. contributed equally to this work.2Present address: Tianjin Institute of Industrial Biotechnology, Chinese Academy of Sci-ences, 32 XiQiDao,Tianjin Airport Economic Park, Tianjin 300308, China.

3To whom correspondence should be addressed. E-mail: jwr7@cornell.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1309739110/-/DCSupplemental.

11928–11933 | PNAS | July 16, 2013 | vol. 110 | no. 29 www.pnas.org/cgi/doi/10.1073/pnas.1309739110

lambda gene expression as phage structural proteins accumulate.Fig. S1 and Dataset S1 document in more detail the patterns ofearly and late gene expression, including translation of newlyrecognized ORFs in both periods.Previous work has shown that although transcription across the

lambda late gene region is approximately uniform, reflecting thesingle mRNA synthesized for the late genes under influence ofthe gene Q antiterminator, the yields of various proteins from thelate gene transcript are very different (15, 16). This differentialtranslation of late genes is apparent in the ribosome profile (Fig. 2C and D). It is of interest to compare the instantaneous rate ofexpression of each transcript to a measure of the ultimate re-quirement for each structural protein in the phage particle (2, 17).Fig. 2E shows some correlation, in particular for the two mostabundant proteins D and E, and for some but not others of theleast abundant proteins. It is noteworthy that rexB, a geneexpressed in the prophage, shows increased expression at latetimes, consistent with previous genetic evidence (18). rexB istranscribed from the independent promoter pLIT, and it is likelythat the enhanced expression of rexB at late times is due to in-creased gene dosage as phage DNA is replicated.

The Pattern of Ribosome Progression. As reported (10), the distri-bution of ribosomes across reading frames is highly heteroge-neous, displaying strong pause sites that are correlated bothwith sites of initiation and termination, and with internal Shine–Dalgarno sequences that presumably stall ribosomes by annealingwith the end of 16S RNA, as in initiation. To illustrate the generalpattern of translation, Fig. 2D shows a display of the distributionof ribosomes over a segment of approximately 11 kB of the lategenes. Note that the scales are drastically different for displays ofthe different time samples in Fig. 2D; thus, there is a difference of∼100 in scale between 5 and 20 min, but the overall patterns arequite similar. In addition to internal pause sites, there frequentlyare collections of ribosomes over initiation and terminationcodons of the reading frame, as noted (10). Fig. S2 uses theanalysis of Li et al. (10) to confirm that ribosomes in the codingsequence tend to pause where Shine–Dalgarno sequences arepositioned upstream to interact with the ribosome.

Frameshifting. A frameshift occurs between lambda genes G andT, resulting in a fusion protein between these reading frameswhen approximately 3.5% of the ribosomes slip back a nucleo-tide at the “slippery” sequence 5′-GGGAAAG-3′ near the endof the G reading frame (19). There is a distinct concentration ofribosomes over this sequence, and at the normal terminationcodon of the G reading frame, which is evident upon examiningthe profiles (Fig. 2F). A typical protected RNA segment coveringthe slippery sequence (in bold) is TCTGCGGGAAAGTGTTC-GACGGT, and a typical segment covering the termination co-don of G is TCGAGGGTGAGCTGAGTTTTGCCCT.

Translation of a Regulatory RNA.Regulation of late gene expressionin lambda and related phages occurs through transcription anti-termination, by the product of gene Q, of a constitutive transcriptfrom the late gene promoter (14); in lambda, this RNA (lambda 6SRNA, not to be confused with the cellular 6S RNA) is 200 nt long.An ORF, orf-64, begins within this RNA and extends through thetranscription terminator. orf-64 shows ribosome-protected frag-ments at a low but significant level, ∼5% the level of the adjacentQ gene and ∼10% the level of gene S, the first late gene regulatedby Q (Fig. 3A). Although it is clear from biochemical analysis withpurified proteins that the Q protein and transcription elongationfactor NusA suffice to cause antitermination of transcription at theterminator of 6S RNA (14), the occurrence of concurrent trans-lation in the cell could modulate or enhance the antiterminationprocess. It may be significant that a cluster of ribosomes appearsover the upstream half of the intrinsic transcription terminatorstem, a configuration that would inhibit termination. Such struc-ture is reminiscent of attenuation control in bacterial operons (20)and could suggest a distant evolutionary relationship betweenthese regulatory mechanisms.

Unique Translated ORFs.A notable feature of the ribosome profilesis the abundance of translation in regions of the genome withoutcharacterized genes but in potential ORFs of unknown function,many of which have not been annotated (Dataset S2 and Fig. S3).We discuss below selected instances of such translation that seemto be noteworthy. However, we also attempted to catalog all such

Fig. 1. An overview of gene expression measuredby ribosome profiling across the lambda phage ge-nome during lambda prophage induction. Bar plotscentered on dashed line circles from inside to out-side show expression levels by ribosome profiling atdifferent times after shifting the culture to 42 °C:blue, 0 min; cyan, 2 min; green, 5 min; orange, 10min; and red, 20 min. Annotated genes of lambdaare shown on the outside. At each time point, thereads per kilobase per million (RPKM) for lambdagenes was normalized by scaling between 0 and 1,with 531 being the maximum at 0 min and 9,091 at 2min. Because the RPKM values for some genes at 5,10, and 20 min are too high, the maximum valueswere set at 10,000, where RPKM is reads per kilo-base of coding sequence per million mapped reads,as originally denoted (31).

Liu et al. PNAS | July 16, 2013 | vol. 110 | no. 29 | 11929

GEN

ETICS

potential ORFs in a systematic way: Dataset S2 lists 55 potentialORFs distinct from known genes and previously annotated ORFs,all of which display significant translation over background. Thislist was obtained by (i) computing all potential ORFs greater than5 aa in length with either ATG or GTG initiation and a terminationcodon (∼3,000); (ii) removing all ORFs with apparent translationlevels below a chosen background; and (iii) removing from this listpreviously recognized genes or ORFs, and completely overlappingORFs, leaving 55. Some of these potential ORFs display significantupstream matches to the Shine–Dalgarno initiation sequence, butmany do not (Dataset S2); however, this feature is not a re-quirement for significant translation.There is no evidence that any of these ORFs represent func-

tional genes. Furthermore, we cannot rule out the possibility thatat least some of the signal, particularly the weakest, reflects anentirely irrelevant background association of RNA with ribo-somes. Many potential ORFs are from regions of the genomedesignated “inessential,” such as the “b2” region and the segmentbetween bor and the right end (2) (Fig. S3). It is clear, however,that inessential in the laboratory context does not mean useless innature. Furthermore, the level of expression of many of the po-tential ORFs in Dataset S2 is comparable to that of known genes(as we describe below), consistent with the notion that this list

could include candidates for regions of important translation ac-tivity, including potential active polypeptide products.We first consider potential translated ORFs that are coded in

the predominant direction of known transcription. In severalcases, these ORFs occur between genes of the late region whenthere is space—a rare occurrence, because throughout the lategenes, a termination codon generally meshes closely with or evenoverlaps the next initiation codon (2). In one case, 150 bp sep-arates the termination of lambda tail gene L from the initiationof tail gene K; an ORF of 76 amino acids (ORF 322) starts 6 bpafter the termination of L and proceeds through the first 11amino acids of K (Fig. 3B). Juhala et al. (21) noted that this ORFis homologous to the beginning of the counterpart of gene K inthe lambda-related phage HK022, and they suggested thata frameshift mutation might have severed a gene ancestral tolambda K. Our data shows that ribosome occupancy of ORF 322,largely clustered just after the initiation codon, is comparable tothat of adjacent genes. Because it is expected that transcriptionof the late gene region is uniform, this result means that thetranslation activity of ORF 322 is significant. Possibly ORF 322and lambda K together constitute the active protein that corre-sponds to the related protein of HK022.In a second case, found at the beginning of the late gene

transcript, there is an interval of 375 bp between the end of the

Fig. 2. Expression patterns of the lambda genome from ribosome profiling. (A–C) Gene expression pattern in the early left operon (A), early right operon (B), andlate gene operon (C) at different times after induction of a lambda prophage. Reads were summed over each reading frame. (D) Higher resolution view of theribosome occupancy profile for some late genes during lambda prophage induction. (E) Correlation of gene expression levels for phage structural genes with theprotein copy numbers in the purified lambda particle. (F) Ribosome occupancy profile at the translational frameshift region of lambda genes G and T. The red dashedrectangle indicates the frameshift site. Ribosomes are stalled at the slippery sequence GGGAAAG in the G reading frame; a few of these signals (approximately 3.5%)shift back one base on the mRNA into the T reading frame and continue to the end of T (19). A cluster of ribosomes over the G termination codon also is apparent.

11930 | www.pnas.org/cgi/doi/10.1073/pnas.1309739110 Liu et al.

previously recognized orf-64 and the first nucleotide of gene S,the first recognized late gene (2). Two additional ORFs occur inthis interval: ORF 915 (of 15 aa) and ORF 916 (of 12 aa). Thedensity of translation over ORF 916 is approximately one-halfthat of gene S (Fig. 3A and Dataset S2).The inessential lambda late gene stf is interrupted by a muta-

tion in laboratory strains, resulting in an N-terminal fragmentencoded by orf-401, and a potential C-terminal fragment encodedby orf-314 (22). orf-401 is translated at a level comparable tosome late genes, and orf-314 is translated much more weakly.One of two potential ORFs between the fragments (ORF 438)appears to be significantly translated. Thus, there is the potential

for expression of the distal parts of the interrupted stf gene, butno evidence that it has any significance.Between the leftward genes exo and Ea22, there is an interval

of 953 bp with several previously recognized leftward-directedORFs (orf60a, orf63, orf61), as well as orf73 (also ORF 2310 inDataset S2) identified with a function named bin (23). Weidentify another, ORF 2313 (31 aa), which is shown in Fig. 3C,which is expressed comparably to the known genes like exoand Ea22.In several cases, there are translated ORFs in regions previously

known to be transcribed from both strands. Thus, in the region ofpredominant leftward transcription to the left of gene N, the genesieB is transcribed rightward (Fig. 4A). Mostly overlapping sieB is

Fig. 3. Examples of expressed unique ORFs in the prevailing direction of transcription. (A) Ribosome occupancy profiles downstream of gene Q, with 20 minof induction time. Some ribosome occupancy was found on orf-64 (red), which extends beyond the 194-nt pR’ transcript that is terminated (in the absence ofthe gene Q antiterminator) at nucleotide 44780. The sequence of the terminator is shown, including the hairpin (red) and poly U (underlined) sequence at theend. Two other unique ORFs (ORF 915 and ORF 916) in this region also showed significant ribosome occupancy. (B) Ribosome occupancy profiles betweengenes L and K, including ORF 322, with 20 min of induction time. (C) Ribosome occupancy profiles between ea22 and orf61, with 20 min of induction time.

Fig. 4. Unique translated ORFs in regions of overlapping or predominant opposite strand transcription. (A) Ribosome occupancy profiles downstream ofgene N, with 20 min of induction time. Unique ORFs in this region are shown in green; most are antisense to gene sieB, but in the direction of prevailingtranscription from pL. (B) Ribosome occupancy profiles of eight unique ORFs at the end of lambda genome, with 20 min of induction time; these ORFs areantisense to bor and lambda p78 (NP_597781.1) but in the direction of the Q-dependent late gene transcription. (C) Ribosome occupancy profiles of plusstrand ORFs in the b2 region, with 20 min of induction time. The 14 ORFs shown are antisense to ea47, ea31, and ea59, but in the same direction asQ-dependent late gene transcription, which may thus converge with the N-dependent transcription from pL.

Liu et al. PNAS | July 16, 2013 | vol. 110 | no. 29 | 11931

GEN

ETICS

a set of substantially translated ORFs, all oriented leftward, andpresumably derived from the N-dependent leftward transcription:ORF2423 (24 aa), ORF 3070 (21 aa), ORF 2426 (20 aa), ORF2429(37 aa), ORF2432 (30 aa), ORF2433 (26 aa), andORF2434 (15 aa).Between genes RZ and Nu1 of the rightward-directed late

transcript, there is an interval of 2,278 bp that contains the left-ward-directed gene bor and extends to the cos DNA packagingsite (Fig. 4B). This interval contains the rightward-directed ORF948 (45 aa), ORF 950 (35 aa), ORF 958 (31 aa), ORF 966 (27 aa),ORF 1830 (24 aa), ORF 975 (9 aa), ORF 976 (28 aa), and ORF1832 (33 aa). Most of these display substantial translation.There is a concentration of translated ORFs directed counter

to the prevailing direction of transcription in the “b2” region. Thissubstantial region of approximately 5 kb, which can be deletedwithout impairing growth (2) (although prophage integrationis compromised), contains three well-recognized “early” genes,ea59, ea31, and ea47; all are directed leftward and presumablyexpressed primarily via gene N antiterminator-influenced left-ward transcription from the early promoter pL. (ea59 is also sig-nificantly expressed in the lysogen, so that another promoter alsomay be present.) Several translated ORFs, ORF 467 (53 aa),ORF 472 (14 aa), ORF 486 (14 aa), ORF 488 (17 aa), ORF 494(28 aa), and ORF496 (36 aa) are all within ea47 and are directedrightward (Fig. 4C). Several rightward-directed ORFs, ORF 499(of 21 aa), ORF 1509 (of 5 aa), ORF 503 (of 17 aa), ORF 1510 (of39 aa), and ORF 507 (of 7 aa), are in the 400-bp intergenic regionbetween genes ea47 and ea31 and are expressed comparably tothe three leftward genes (Fig. 4C). Within ea31 are two rightward-directed ORFs (Fig. 4C). Remarkably, ORF 511 is 73 aa long, andORF 1522 is 30 aa long. Both are translated comparably to theadjacent leftward genes ea31 and ea59.There also are rightward-directed ORFs within the early re-

gion to the right of the attachment site (not illustrated). Thus,within ea8.5, the overlapping ORFs 585 (of 130 aa) and 591 (of41 aa) show substantial translation.The source of the transcription for these newly recognized

rightward ORFs is unknown. It may of course be the gene Qantiterminator-influenced late gene transcription, which is ex-pected to proceed efficiently at least through the tail fiber genesstf and tfa, and could easily extend well into the early region, al-though its efficiency likely would be reduced through collisionwith the predominant leftward-directed N-protein-influencedtranscription from promoter pL.A more detailed display of expression data of some potential

ORFs in shown in Fig. S4.

Changes in Host Gene Expression After Lambda Induction. Expres-sion in the host was measured before and after lambda induction

through ribosome profiling of annotated E. coli genes, in thesame experiments described above. We find that although ap-proximately 1,000 genes are down-regulated during the 20 min ofinduction, a substantial number, 120, are up-regulated (DatasetS3). Most of these changes occur primarily late in induction.Approximately one-half of the up-regulated gene expression

has a relatively trivial origin: Genes that surround the attach-ment site attB are replicated along with the prophage (“escapereplication”) and are overexpressed primarily because of theincreased gene copy number (Fig. 5A). This enhanced expressionoccurs in a region of 300–400 kB surrounding the prophage in-tegration site (4). No down-regulated genes were found in thisregion. One well-characterized set of up-regulated genes is thenearby gal operon, overexpression of which depends on bothoverreplication and gene N-dependent antiterminated tran-scription from the phage pL promoter (4). Transcription of galoperon genes is enhanced approximately 13-fold in inducedlysogens relative to nonlysogens (4); we found that ribosomeoccupancy of the gal operon genes galE, galT, and galK increasedapproximately 13–16-fold, in good agreement (Dataset S3).A prominent group of up-regulated genes located far from attB

and, thus, stimulated by some metabolic change in the cells, is theset of SOS DNA damage genes (examples in Fig. 5B) (24). Becauseour induction protocol used temperature inactivation of phagerepressor and not explicit DNA damage (e.g., UV irradiation), theinduction of SOS genes must be due to some aberration in DNAmetabolism that arises during phage growth, likely resulting fromthe replication of phage DNA; this anomaly might be the accu-mulation of single-stranded regions and incompletely replicatedphage chromosomes, for example. It is noteworthy that increasedtranslation of the SOS repressor LexA occurs along with that of theother SOS genes, although the concentration of LexA must bediminished. This result is in fact as expected, because LexA is self-regulated but also rapidly degraded while the SOS-inducing signalis active, a process that overcomes the increased expression due toderepression. It is also consistent with these proposals that SOSgene induction occurs late rather than early in induction.Several other sets of up-regulated genes presumably respond to

the altered cellular conditions of phage growth. Induction ofsubunits of ribonucleotide reductase genes may reflect increasedneed for dNTPs. The set of genes involved in phosphate metab-olism and transport responds to cellular phosphate limitation,which may occur in the induced cell; the involvement of pho genesin lambda growth has been noted (25). The genes cpxP, htpX, anddegP are activated by the two component system encoded by cpxR/A, which senses envelope stress (26), a plausible response to suchphage proteins as the lysis set that are made at late times ofinduction.

Fig. 5. Effect of phage induction on E. coli protein syn-thesis. (A) Increase in E. coli gene translation around theattB site due to escape synthesis. Time at 42 °C: blue, 0 min;cyan, 2 min; green, 5 min; orange, 10 min; and red, 20 min.(B) E. coli genes with functions likely related to lambdaphage growth and significantly up-regulated during pro-phage induction. (C) E. coli genes mostly with knownfunctions not obviously related to lambda phage growthbut significantly up-regulated during prophage induction.(D) Phage induction leads to a decrease in ribosome occu-pancy on E. coli genes, and approximately 30% of totalreads map to the phage genome at 20 min. (E) Decrease ofE. coli gene translation during lambda induction.

11932 | www.pnas.org/cgi/doi/10.1073/pnas.1309739110 Liu et al.

Finally, there is a set of up-regulated genes that mostly haveknown function, but no obvious relation to phage development;examples are shown in Fig. 5C.Phage growth substantially engages the protein production ca-

pacity of the cell. At 20 min after induction, about 30% of the totalribosome-protected RNA reads map to the lambda genome. Of the1,000 down-regulated genes (Datasets S4 and S5), 600 are down-regulated more than twofold. There appears to be a constancy oftotal protein synthetic capacity when the number of ribosome readsfrom phage and host are summed throughout the 20-min period ofphage development; the 30% increase in phage reads is matched bya 30% loss in host reads (Fig. 5 D and E), and the decrease in eachindividual amino acid incorporation into bacterial protein is equalto its incorporation into phage proteins (Fig. S5).

Summary and Further DiscussionRibosome profiling provides a detailed view of gene expressionof both phage and host during the development of bacteriophagelambda. This analysis confirms the expected general patterns ofgene expression but also shows an unexpected complexity of thetranslation landscape. The major finding is that much ribosomeoccupancy occurs in ORFs of unknown function. Although someof these potential ORFs were known from analysis of the lambdagenome sequence, and all could of course be inferred by simpleanalysis, the translation data focuses attention on a substantialnumber that were not previously known to be of interest.Numerous translated ORFs are large enough plausibly to

encode functional proteins, in the range of 20–130 aa. The un-derappreciated importance of small proteins has been recog-nized, and they have a variety of roles that might be relevant tolysogen and phage growth, acting as intercell signaling factors,toxins, and membrane components (27). Few of the newly dis-covered ORFs appear to be expressed in the lysogen at levelssimilar to well known prophage-specific genes (cI, rexA, rexB,lom, bor), although the previously recognized ORFs ea59 andea8.5 are significantly expressed. However, very low expressionmight well be important.Of course, the act of ribosome engagement or ribosome syn-

thetic activity could be important rather than the translation

product itself. Thus, translation of ORFs upstream of expressedgenes has a well-known regulatory function in yeast (28), and thevery act of ribosome binding can change the structure and avail-ability of mRNA for translation of other parts of the message, orfor other functions of RNA such as transcription termination. Aspecific example of an alternate consequence of ribosome functionis the expression of the lambda bar “genes,” which act to sequesterribosomes that must be freed by a peptide hydrolase (29). Anotherpossibility is that engagement of RNA with ribosomes serves toprevent deleterious activity (e.g., forming R loops) of free RNA(30), which might be prevalent where antiterminators prevent Rhofunction, as in most lambda transcription; thus, functional ORFsmight be retained in regions of the genome where transcriptionoccurs, even if no functional polypeptide is produced.Finally, it is of course possible that some or all of these ORFs

have no function at all, representing just background activity ofthe transcription and translation systems. We cannot eliminatethe possibility that there is adventitious and nonfunctional as-sociation of RNA with ribosomes that gives rise to illusorytranslation of at least some of these potential ORFs. Furtherinformatic and directed analysis, and comparative phage geno-mic analysis, would be required to query the function of thisunexpected translation activity.

Materials and MethodsThe strain used in these experiments was E. coli K12 MG1655 (obtained fromF. Blattner, University of Wisconsin, Madison, WI), made lysogenic for λcI857(from M. Gottesman, Columbia University Medical School, New York). Ri-bosome protected RNAs were prepared as described (6), with minor mod-ifications. Detailed protocols and methods of data analysis are provided inSI Materials and Methods.

ACKNOWLEDGMENTS. We thank Ryland Young and Jim Hu (Texas A&MUniversity) for careful and thoughtful reading of the manuscript, SherwoodCasjens (University of Utah) for helpful comments, and Nick Ingolia for pro-viding a protocol before publication. Help with the display and analysis ofthe data was provided by the PortEco project through National Institute ofGeneral Medical Sciences U24GM088849, and this work was supported byNational Institutes of Health Grant GM 21941.

1. Hendrix RW, Roberts JW, Stahl FW, Weisberg RA (1983) Lambda II (Cold Spring HarborLab Press, Cold Spring Harbor, NY).

2. Hendrix RW, Casjens S (2006) Bacteriophage lambda and its genetic neighborhood.The Bacteriophages, ed Calendar RL (Oxford Univ Press, Oxford).

3. Chen Y, Golding I, Sawai S, Guo L, Cox EC (2005) Population fitness and the regulationof Escherichia coli genes by bacterial viruses. PLoS Biol 3(7):e229.

4. Osterhout RE, Figueroa IA, Keasling JD, Arkin AP (2007) Global analysis of host re-sponse to induction of a latent bacteriophage. BMC Microbiol 7:82.

5. Ingolia NT (2010) Genome-wide translational profiling by ribosome footprinting.Methods Enzymol 470:119–142.

6. Ingolia NT, Brar GA, Rouskin S, McGeachy AM, Weissman JS (2012) The ribosomeprofiling strategy for monitoring translation in vivo by deep sequencing of ribosome-protected mRNA fragments. Nat Protoc 7(8):1534–1550.

7. Stern-Ginossar N, et al. (2012) Decoding human cytomegalovirus. Science 338(6110):1088–1093.

8. Oh E, et al. (2011) Selective ribosome profiling reveals the cotranslational chaperoneaction of trigger factor in vivo. Cell 147(6):1295–1308.

9. Stein LD, et al. (2002) The generic genome browser: A building block for a modelorganism system database. Genome Res 12(10):1599–1610.

10. Li GW, Oh E, Weissman JS (2012) The anti-Shine-Dalgarno sequence drives trans-lational pausing and codon choice in bacteria. Nature 484(7395):538–541.

11. Oppenheim AB, Kobiler O, Stavans J, Court DL, Adhya S (2005) Switches in bacte-riophage lambda development. Annu Rev Genet 39:409–429.

12. Friedman DI, Gottesman M. (1983) Lytic mode of lambda development. Lambda II, edHendrix RW, Roberts JW, Stahl FW, Weisberg RA (Cold Spring Harbor Lab Press, ColdSpring Harbor, NY).

13. Gussin G, Johnson A, Pabo C, Sauer R (1983) Repressor and cro protein: Structure,function, and role in lysogenization. Lambda II, eds Hendrix RW, Roberts JW,Stahl FW, Weisberg RA (Cold Spring Harbor Lab Press, Cold Spring Harbor, NY).

14. Roberts JW, et al. (1998) Antitermination by bacteriophage lambda Q protein. ColdSpring Harb Symp Quant Biol 63:319–325.

15. Ray PN, Pearson ML (1974) Evidence for post-transcriptional control of the morpho-genetic genes of bacteriophage lambda. J Mol Biol 85(1):163–175.

16. Sampson LL, Hendrix RW, HuangWM, Casjens SR (1988) Translation initiation controlsthe relative rates of expression of the bacteriophage lambda late genes. Proc NatlAcad Sci USA 85(15):5439–5443.

17. Dai XX (2009) Expression, purification and characterication of bacteriophage lambdatail tip proteins. PhD thesis (Univ of Pittsburgh, Pittsburgh).

18. Parma DH, et al. (1992) The Rex system of bacteriophage lambda: Tolerance and al-truistic cell death. Genes Dev 6(3):497–510.

19. Xu J, Hendrix RW, Duda RL (2004) Conserved translational frameshift in dsDNA bac-teriophage tail assembly genes. Mol Cell 16(1):11–21.

20. Henkin TM, Yanofsky C (2002) Regulation by transcription attenuation in bacteria:How RNA provides instructions for transcription termination/antitermination deci-sions. Bioessays 24(8):700–707.

21. Juhala RJ, et al. (2000) Genomic sequences of bacteriophages HK97 and HK022:Pervasive genetic mosaicism in the lambdoid bacteriophages. J Mol Biol 299(1):27–51.

22. Hendrix RW, Duda RL (1992) Bacteriophage lambda PaPa: Not the mother of alllambda phages. Science 258(5085):1145–1148.

23. Sergueev K, Court D, Reaves L, Austin S (2002) E.coli cell-cycle regulation by bacte-riophage lambda. J Mol Biol 324(2):297–307.

24. Little JW, Mount DW (1982) The SOS regulatory system of Escherichia coli. Cell 29(1):11–22.

25. Maynard ND, et al. (2010) A forward-genetic screen and dynamic analysis of lambdaphage host-dependencies reveals an extensive interaction network and a new anti-viral strategy. PLoS Genet 6(7):e1001017.

26. Vogt SL, Raivio TL (2012) Just scratching the surface: An expanding view of the Cpxenvelope stress response. FEMS Microbiol Lett 326(1):2–11.

27. Hobbs EC, Fontaine F, Yin X, Storz G (2011) An expanding universe of small proteins.Curr Opin Microbiol 14(2):167–173.

28. Hinnebusch AG (2005) Translational regulation of GCN4 and the general amino acidcontrol of yeast. Annu Rev Microbiol 59:407–450.

29. Ontiveros C, Valadez JG, Hernández J, Guarneros G (1997) Inhibition of Escherichiacoli protein synthesis by abortive translation of phage lambda minigenes. J Mol Biol269(2):167–175.

30. Leela JK, Syeda AH, Anupama K, Gowrishankar J (2013) Rho-dependent transcriptiontermination is essential to prevent excessive genome-wide R-loops in Escherichia coli.Proc Natl Acad Sci USA 110(1):258–263.

31. Ingolia NT, Ghaemmaghami S, Newman JR, Weissman JS (2009) Genome-wide anal-ysis in vivo of translation with nucleotide resolution using ribosome profiling. Science324(5924):218–223.

Liu et al. PNAS | July 16, 2013 | vol. 110 | no. 29 | 11933

GEN

ETICS

Supporting InformationLiu et al. 10.1073/pnas.1309739110SI Materials and MethodsStrains. The strain used in these experiments was Escherichia coliK12 MG1655 (obtained from F. Blattner), made lysogenic forλcI857 (from M. Gottesman).

Preparation of Ribosome Protected RNA. The procedure was asdescribed (1), with minor modifications. One hundred-milli-liter portions of Mops minimum medium (2) were inoculatedwith 2 mL of overnight culture of MG1655 and MG1655(λcI857) and incubated with shaking at 32 °C to an opticaldensity at 600 nm (OD600) of 0.4–0.5; cultures were processed0, 2, 5, 10, and 20 min after transfer to a shaking water bath at42 °C. Chloramphenicol was added to 0.1 mg/mL 2 min beforeharvesting the cells to freeze the translating ribosomes. Thecell culture was mixed with an equal mass of ice in centrifugetubes and centrifuged at 8,000 rpm (10,000 × g) for 10 min at4 °C to harvest the cells. The cell pellet was resuspended in 0.5mL of ice-cold lysis buffer (20 mM Tris·Cl at pH 7.5, 100 mMNH4Cl, 10 mM MgCl2, and 0.1 mg/mL chloramphenicol),supplemented with 50 μL of lysozyme solution (Sigma), mixedby pipetting, and frozen for 5 min in liquid nitrogen. Thesuspension was thawed, refrozen, thawed again, and supple-mented with 15 μL of 10% (wt/vol) deoxycholate to completethe cell lysis. Cell debris was removed by centrifugation at10,000 rpm for 10 min at 4 °C. One hundred OD260 units ofthe ribosome preparation was digested with 20,000 gel units ofmicrococcal nuclease (M0247; New England Biolabs) and 20units of DNase I (M0303S; New England Biolabs) at 25 °C for1 h, followed by gentle layering onto a sucrose gradient con-sisting of 2.5 mL layers of 10, 20, 30, and 40% sucrose sol-utions in polysome buffer (20 mM Tris·Cl at pH 7.5, 100 mMNH4Cl, 10 mM MgCl2, 2 mM 2-mercaptoethanol, and 0.1 mg/mL chloramphenicol) that had equilibrated overnight at 4 °C.The gradient was centrifuged in an SW41 Ti rotor at 35,000rpm (151,000 × g) for 3 h at 4 °C, and monosome fractionswere collected.

Library Construction and Deep Sequencing. Ribosome footprintlibraries were prepared as reported, with minor modifications(1, 3). RNA was extracted from monosomes by the hot acidphenol method, precipitated with NaOAc and ethanol, anddephosphorylated by T4 polynucleotide kinase (M0236; NewEngland Biolabs) for 1 h at 37 °C. RNA was loaded ontoa 15% Mini-PROTEAN TBE-Urea Precast Gel (Bio-Rad)and run until the ∼30-nt region was resolved. The region from28 nt to 42 nt was excised, based on the mobility of definedRNA oligonucleotides and a 10-bp DNA ladder. The gelfraction was eluted overnight in 300 mM NaOAc at pH 5.5,1 mM EDTA, and 0.1 U/μL SUPERas In (Ambion no.AM2694), followed by ethanol precipitation. PolyA tails wereadded to the purified RNA fragments by E. coli polyA poly-merase (New England Biolabs) at 37 °C for 20 min in thebuffer provided. The tailed RNA molecules were reversetranscribed by using barcoded primers and SuperScript III(Invitrogen) to generate the first-strand cDNA. After reversetranscription, RNA was removed from RNA-DNA duplexesby incubation for 15 min at 98° in 0.1 M NaOH, followed bythe addition of an equal concentration of HCl. Reverse-transcription products were loaded onto a 10% poly-acrylamide TBE-urea gel. The band of the first-strand cDNAsynthesis was excised and recovered by using DNA gel elutionbuffer (300 mM NaCl and 1 mM EDTA). Purified first-strand

cDNA was circularized by 50 U CircLigase (Epicentre). Cir-cularized ribosomal footprint cDNA was amplified by PCRusing the Phusion High-Fidelity enzyme (New England Biol-abs) according to the manufacturer’s instructions. The PCRproducts were separated on a nondenaturing 8% poly-acrylamide TBE gel. Mixed DNA samples from differentbarcoded samples typically were used for cluster generation.Deep sequencing was performed with the Illumina Hi-SEQ2000.One reverse-transcription primer sequence (Adap1) is shown

below; the barcode is underlined. The other 11 primer sequencesare the same except that the barcode sequence is replaced byGCAT, GTAC, GATC, GCTA, GTCA, ACGT, TCGA, ATGC,CGAG, TGTG or AGCG./5Phos/GACTGATCGTCGGACTGTAGAACTCT/iSp18/CA-

CTCA/iSp18/CAAGCAGAAGACGGCATACGATTTTTTTTT-TTTTTTTTTTTVN.

Sequence Alignments and Normalization. Reads were partitionedaccording to the 4-nt 5′ barcode, trimmed of barcode from the5′ end, and all contiguous trailing A residues were removedfrom the 3′ ends of reads. SOAP2 (Bowtie v0.12.7), allowingup to two mismatches, was used to perform the alignments.First, reads that aligned to E. coli rRNA and tRNA sequenceswere discarded. All remaining reads were aligned to the viral(J02459) and E. coli (U00096) genomes. Reads with uniquealignments were used to compute the total number of reads ateach position in the viral and E. coli genomes. Footprintdensities were calculated in units of reads per kilobase permillion (RPKM) to normalize for gene length and totalviral and E. coli reads per sequencing run. The pattern ofribosome protection was highly reproducible over twoanalyses (Dataset S5), with a Pearson correlation coefficientof 0.976.

SD-Like Motif Identification. After mapping onto the genome, themiddle position of each read was used to represent the A site. Ifthe number of reads was fivefold more than the average reads ofthe gene the position was defined as pausing site, and the 20-bpflanking sequences up and down were used to identify motifenrichment. Compared with the same length of random se-quence in genome, the occurrence of motifs was evaluated by theFisher t test.

Unique ORF Prediction. The analysis and figures in this reportused the GenBank accession no. J02459 that contains 73annotated ORFs in the lambda phage genome. To identifyunique ORFs in the genome, we first computed all potentialORFs that are longer than 18 bp and contain a start codon(ATG or GTG) and a stop codon, thereby finding 2,809 po-tential ORFs of more than 5 aa. Second, we defined a back-ground in all regions in E. coli and lambda that are notcontained in ORFs or in rna genes and, thus, are expected tobe ribosome free (the null model). (This background mighthave both experimental and computational origins.) Finally,we summed reads from different times at each position andcalculated RPKM in each putative ORF. The RPKM of eachputative ORF was compared with the distribution of RPKMin the null model. We retained putative ORFs if their RPKMis significantly higher than the null model (P < 0.05, onesample t test). Comparing these ▪▪▪ with the known 73 genes,we removed overlapping ORFs if the RPKM of the non-

Liu et al. www.pnas.org/cgi/content/short/1309739110 1 of 5

overlapping region is not significantly higher than the nullmodel (P < 0.05). We also removed overlapping ORFswithin putative ORFs. The final result was to identify 55 pu-tative ORFs shown in Dataset S2, in two categories of confi-dence of expression level: P < 0.01 (black) or P < 0.05 (red).

Differential Gene Translation Analysis. All experimental sampleswere collected in duplicate. Based on the correlation bet-ween the replicates, we set up a threshold with RPKM of

10 for the genes whose translation could be determined re-producibly.

Genome Browser. Bam database files that were uploaded intoa genome browser (http://heptamer.tamu.edu/gb2/gbrowse/)that displays the lambda phage or E. coli annotated genomealong with representations of the ribosome footprints. On thissite, ribosome occupancy profiles for selected time points (0,2, 5, 10, and 20 min) can be viewed simultaneously.

1. Ingolia NT, Brar GA, Rouskin S, McGeachy AM, Weissman JS (2012) The ribosomeprofiling strategy for monitoring translation in vivo by deep sequencing of ribosome-protected mRNA fragments. Nat Protoc 7(8):1534–1550.

2. Ingolia NT, Ghaemmaghami S, Newman JR, Weissman JS (2009) Genome-wide analysisin vivo of translation with nucleotide resolution using ribosome profiling. Science 324(5924):218–223.

3. Neidhardt FC, Bloch PL, Smith DF (1974) Culture medium for enterobacteria. J Bacteriol119(3):736–747.

Fig. S1. Time course of expression in detail of lambda genes. (A) Early leftward genes. (B) Early rightward genes. (C) Late genes. (D) Unique ORFs in the earlyoperons. (E) Unique ORFs in the late operon. (F) Unique ORFs in the b2 region. (G) Time course of expression of unique ORFs versus previously characterizedmajor genes and ORFs.

Liu et al. www.pnas.org/cgi/content/short/1309739110 2 of 5

Fig. S2. The anti-Shine–Dalgarno sequence drives translational pausing in bacteria (A) and lambda phage (B). B Right shows an alignment of the sequencesenriched at pause sites against the Shine–Dalgarno consensus sequence. Reads on translational pause sites, defined as sites where the number of reads isfivefold more than the average for the gene, were mapped onto the genome and the middle position was used to represent the A site. The 20-bp flankingsequence upstream and downstream was used to identify motif enrichment. Compared with the same length of the randomly selected sequence in genome,the occurrence of motifs was evaluated by the Fisher test: Yellow, P < 0.01; black, P = 0.01; blue, P > 0.01.

Fig. S3. Unique potential ORF distribution in the lambda genome.

Liu et al. www.pnas.org/cgi/content/short/1309739110 3 of 5

Fig. S4. Expression of unique ORFs, displayed on a log scale to visualize a range of expression. (A and B) Time course of expression of ORF821, ORF2423,ORF3070, ORF2426, ORF2429, ORF2432, ORF2433, and ORF2434, showing strand specificity and typical pR and pL expression. The expression of ORF821 from pRis significant compared with the neighboring genes ren and NinB. (C) Time course and strand specificity of expression of unique ORFs showing typical late geneexpression. (D) Time course and strand specificity of expression of unique ORFs in the b2 region; the ORFs arise from the plus strand, and ea47 from the minusstrand. (E) Time course and strand specificity of expression of the unique ORF 2232 upstream of the expressed gene ea59. (F) Time course and strand specificityof expression of unique ORFs 2496 and 2498 antisense to cro and to cI. CII-dependent promoter pRE is upstream of these ORFs and is the likely source of theRNA, consistent with the time course of cII expression.

Liu et al. www.pnas.org/cgi/content/short/1309739110 4 of 5

Fig. S5. Amino acid use in phage and E. coli genes. (A) Use of amino acids, defined as the frequency of an amino acid in a gene multiplied by the RPKM. Theamino acid use in the induced lysogen and nonlysogen control was estimated for each time. (B–F) For each amino acid, the consumption in lambda and thedecrease in consumption in E. coli was calculated. The correlation of use of amino acids between E. coli and lambda is shown at different times of induction.

Other Supporting Information Files

Dataset S1 (XLSX)Dataset S2 (XLSX)Dataset S3 (XLSX)Dataset S4 (XLSX)Dataset S5 (XLSX)

Liu et al. www.pnas.org/cgi/content/short/1309739110 5 of 5