Correcting direct effects of ethanol on translation andtranscription machinery confers ethanol tolerancein bacteriaRembrandt J. F. Hafta, David H. Keatinga, Tyler Schwaeglera, Michael S. Schwalbacha, Jeffrey Vinokura,Mary Tremainea, Jason M. Petersb,c,1, Matthew V. Kotlajichb, Edward L. Pohlmanna, Irene M. Onga, Jeffrey A. Grassa,Patricia J. Kileya,d, and Robert Landicka,b,e,2

aGreat Lakes Bioenergy Research Center and Departments of bBiochemistry, cGenetics, dBiomolecular Chemistry, and eBacteriology, Universityof Wisconsin-Madison, Madison, WI 53706

Edited by Tina M. Henkin, Ohio State University, Columbus, OH, and approved May 13, 2014 (received for review January 29, 2014)

The molecular mechanisms of ethanol toxicity and tolerance inbacteria, although important for biotechnology and bioenergyapplications, remain incompletely understood. Genetic studieshave identified potential cellular targets for ethanol and haverevealed multiple mechanisms of tolerance, but it remains difficultto separate the direct and indirect effects of ethanol. We usedadaptive evolution to generate spontaneous ethanol-tolerantstrains of Escherichia coli, and then characterized mechanisms oftoxicity and resistance using genome-scale DNAseq, RNAseq, andribosome profiling coupled with specific assays of ribosome andRNA polymerase function. Evolved alleles of metJ, rho, and rpsQrecapitulated most of the observed ethanol tolerance, implicatingtranslation and transcription as key processes affected by ethanol.Ethanol induced miscoding errors during protein synthesis, fromwhich the evolved rpsQ allele protected cells by increasing ribo-some accuracy. Ribosome profiling and RNAseq analyses establishedthat ethanol negatively affects transcriptional and translationalprocessivity. Ethanol-stressed cells exhibited ribosomal stalling atinternal AUG codons, which may be ameliorated by the adaptiveinactivation of the MetJ repressor of methionine biosynthesisgenes. Ethanol also caused aberrant intragenic transcription termi-nation for mRNAs with low ribosome density, which was reducedin a strain with the adaptive rho mutation. Furthermore, ethanolinhibited transcript elongation by RNA polymerase in vitro. Wepropose that ethanol-induced inhibition and uncoupling of mRNAand protein synthesis through direct effects on ribosomes andRNA polymerase conformations are major contributors to ethanoltoxicity in E. coli, and that adaptive mutations in metJ, rho, andrpsQ help protect these central dogma processes in the presenceof ethanol.

Aliphatic alcohols such as ethanol are important microbialbioproducts whose toxic effects are known to limit their

production in both bacteria and yeast (1–4). Thus, elucidatingthe mechanisms by which alcohols exert toxic effects and un-derstanding modes of microbial tolerance are important both tounderstand basic microbial physiology and to engineer microbeswith more efficient fermentative capacities (5, 6). Escherichia coliis a model of choice for these studies. Its well-defined physiologyand powerful genetic tools have allowed the identification ofmultiple effects of ethanol on biomolecules and cellular processes.One well-established toxic effect of ethanol on E. coli is an

increase in cell-envelope permeability (7–9). E. coli exposed toethanol exhibit reduced peptidoglycan cross-linking, which isdetrimental to viability, and altered membrane-lipid composi-tion, which may represent an attempt to cope with ethanol stress(10, 11). Ethanol induces broad transcriptional changes in E. colithat extend beyond membrane-stress responses (12, 13), how-ever, suggesting that membrane effects explain only a part ofthe toxicity of ethanol. Consistent with this idea, widely variedapproaches have successfully been used to achieve modest eth-anol tolerance in E. coli, including random transposon insertion

(14, 15), overexpression of native gene libraries (16–18), over-expression of protein chaperones (19), engineered oxidationof ethanol (20), transcriptional rewiring through the cataboliteactivator protein (21) or the transcription factor σ70 (22), andmodulation of cellular fatty-acid composition (23). The mecha-nisms by which these genetic changes confer tolerance remainunclear. Additional mechanistic studies are needed to determinethe most physiologically relevant impacts of ethanol and to dis-tinguish primary effects from downstream consequences.An intriguing possibility raised by previous results is that the

fundamental processes of transcription and translation may beaffected by ethanol. For instance, biochemical studies haveshown that purified E. coli ribosomes are prone to misreadingerrors when treated with ethanol (24–26) although in vivo effectsof ethanol on translational error have not been reported. Also,ethanol-treated E. coli produce the alarmone (p)ppGpp (27),a signal of translation inhibition (28). Detrimental effects of

Significance

Microbially produced aliphatic alcohols are important bio-commodities but exert toxic effects on cells. Understanding themechanisms by which these alcohols inhibit microbial growthand generate resistant microbes will provide insight into mi-crobial physiology and improve prospects for microbial bio-technology and biofuel production. We find that Escherichiacoli ribosomes and RNA polymerase are mechanistically af-fected by ethanol, identifying the ribosome decoding center asa likely target of ethanol-mediated conformational disruptionand showing that ethanol inhibits transcript elongationvia direct effects on RNA polymerase. Our findings provideconceptual frameworks for the study of ethanol toxicity inmicrobes and for the engineering of ethanol tolerance thatmay be extensible to other microbes and to other short-chain alcohols.

Author contributions: R.J.F.H., D.H.K., M.S.S., M.T., J.M.P., M.V.K., P.J.K., and R.L. designedresearch; R.J.F.H., D.H.K., T.S., M.S.S., J.V., M.T., J.M.P., M.V.K., E.L.P., and J.A.G. performedresearch; R.J.F.H., D.H.K., T.S., M.S.S., M.T., J.M.P., M.V.K., E.L.P., I.M.O., and J.A.G. ana-lyzed data; and R.J.F.H., D.H.K., J.M.P., M.V.K., and R.L. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

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. GSE56408).Lists of noncoding RNA regions, pseudogenes, and gene sets used in ribosome profilinganalysis are available from GEO under accession no. GSE56372.1Present address: Department of Microbiology and Immunology, University of California,San Francisco, CA 94143.

2To whom correspondence should be addressed. E-mail:landick@biochem.wisc.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1401853111/-/DCSupplemental.

E2576–E2585 | PNAS | Published online June 9, 2014 www.pnas.org/cgi/doi/10.1073/pnas.1401853111

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

020

ethanol on transcription are suggested by two independentreports linking decreased Rho activity to ethanol tolerance inE. coli by a point mutation in rho (14, 15) or by inhibition of Rhoactivity by overexpression of YaeO/Rof (18). Rho is a hexamericRNA helicase that can terminate transcription when it is notcoupled with translation (29). Reduced Rho activity leads tonumerous changes in gene expression that might confer ethanoltolerance (14, 30). The reported effects of ethanol on ribosomesand (p)ppGpp induction suggest that uncoupling of translationfrom transcription could additionally cause maladaptively highRho termination within genes in ethanol-stressed cells.We sought to identify the primary effects of ethanol on E. coli

and to understand mechanisms of tolerance using a two-stagestrategy. First, we selected for tolerance-conferring mutations bydirected evolution, allowing cultures to accumulate spontaneousmutations while minimizing bias against mutations in essentialgenes that can accompany transposon mutagenesis or over-expression. Second, we tested for physiological effects of ethanolsuggested by the evolved alleles and investigated the mechanismsby which evolved alleles counteracted ethanol toxicity. Ourresults demonstrate that ethanol has detrimental effects oncentral dogma processes in vivo, including increases in trans-lational error, ribosome stalling, and intragenic Rho-dependenttranscription termination, as well as on transcript elongation invitro. Mutant alleles isolated in this study help to ameliorate invivo effects of ethanol, underscoring the physiological relevanceof transcription and translation in ethanol toxicity and resistance.

ResultsMutations in metJ, rho, and rpsQ Confer Ethanol Tolerance andImprove Fermentative Yields. To reveal physiologically importanttargets of ethanol, we performed serial-passage evolution experi-ments to select for spontaneous mutations that conferred a growthadvantage to E. coli in minimal glucose medium containing in-creasing concentrations of ethanol up to 65 g/L (Fig. 1A). Weisolated clonal strains from the evolved cultures by growing culturealiquots on nonselective plates and selecting single colonies forfurther study. We tested ethanol tolerance of clonal strains byfollowing growth in media containing 40–65 g of EtOH/L and se-lected the three highly tolerant isolates MTA156, MTA157, andMTA160 for genomic sequencing (Table 1). MTA156 andMTA160 were independently isolated from a single evolvedculture whereas MTA157 was isolated from a separate cultureevolved in parallel. Mutations present in these strains (Table S1)affected genes involved in various pathways, but six of the elevenmutant alleles encoded variant proteins with clear ties to tran-scription and translation, including two variants of transcriptiontermination factor Rho, two variants of the master repressor ofmethionine synthesis (MetJ), a variant ribosomal protein S17(RpsQ), and a variant of the conserved transcription elongation/termination factor NusA.

We selected for detailed study a single clonal strain, MTA156,which displayed improved growth in the presence of ethanol (Fig.1B and Fig. S1) comparable with that reported for highly ethanol-tolerant E. coli strains described by various groups (1, 17, 18, 20,22). Genomic resequencing of MTA156 identified mutations af-fecting ispB, lptF, metJ, rho, rpsQ, and topA (Table 1 and TableS1). Replacement of these loci with wild-type alleles reducedethanol tolerance in each case except that of topA (Fig. S1A).Tolerance conferred by mutations in genes involved in LPS

transport (lptF) and quinone biosynthesis (ispB) likely act byameliorating effects of ethanol on processes occurring in theE. coli cell envelope. The other three tolerance-conferring alleleshad clear ties to transcription and translation: metJ[ΔE91], rho[L270M], and rpsQ[H31P] (Table 2). Combining these evolvedalleles in an otherwise wild-type background improved aerobic

A

B

-1 0 1 2 3

0.125

0.250

0.500

Wild-typemetJ rho rpsQ

0.080

0.750

MTA156

Time (hr relative to ethanol addition)A

600

Response of growing cultures to ethanol stress

T0T1 T2

MTA156

g EtOH/L 50 55 60 65

Passage

(MG1655)

0 10 18

Directed evolution for ethanol tolerance

4 24

Wild-type

40 g EtOH/L

Growth rates (hr-1)

Wild-type 0.71 ± 0.03 0.11 ± 0.01MTA156 0.48 ± 0.04 0.16 ± 0.01

metJ rho rpsQ 0.49 ± 0.03 0.17 ± 0.02

T0 T2

Fig. 1. Selection of mutations conferring ethanol tolerance. (A) Summary ofdirected evolution experiment, with number of serial passages at each stepof increasing ethanol concentration indicated. (B) Representative growthcurves showing response of MG1655, MTA156, and ethanol-tolerant triplemutant (EP61) to ethanol addition. Mean growth rates for pre- and post-stress cultures are shown ± SEM for four biological replicates. Time pointsnoted were used for sampling in ribosome-profiling experiments. Wild-typeMG1655 and the ethanol-tolerant triple mutant EP61 were sampled beforeethanol addition (T0), during acute stress (T1), and during chronic stress (T2).

Table 1. Strains used in the current study

Strain Relevant genotype Source

MG1655 E. coli K-12 F– λ– ilvG– rfb-50 rph-1 (83)MTA156 MG1655 ispB[I32L] lptF[ΔE265] metJ[ΔE91] rho[L270M] rpsQ[H31P] topA[H122L] This studyMTA157 MG1655 metJ[V33E] nusA[R258G] setB[ΔL33-V393] tqsA[N234Y] This studyMTA160 MG1655 ispB[I32L] lptF[ΔE265] metJ[ΔE91] rho[L207Q] rpsQ[H31P] topA[H122L] This studyEP23 MG1655 metJ::kan This studyEP49 MG1655 rpsQ[H31P] This studyEP61 MG1655 metJ[ΔE91] rho[L270M] rpsQ[H31P] This studyMTA376 MG1655 ΔldhA::FRT pJGG2 (Z. mobilis pdc adhB) This studyMTA722 MTA376 metJ[ΔE91] rho[L270M] rpsQ[H31P] This studyRL2325 MG1655 rho[L270M] This studyRL2739 MG1655 rpsL150 This study

Haft et al. PNAS | Published online June 9, 2014 | E2577

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

020

growth in the presence of ethanol to a level near that of MTA156(Fig. 1B and Fig. S1B). The triple mutant (strain EP61) alsoshowed increased ethanol tolerance under anaerobic conditions(Fig. S1 C and D). An ethanologenic strain bearing the evolvedmetJ, rho, and rpsQ alleles produced 30% more ethanol than thebase strain on a per-cell basis in a synthetic mimic of lignocel-lulose hydrolysate, indicating that the ethanol-tolerant pheno-type can promote biofuel production under industrially relevantconditions (Fig. S1E).The tolerance associated with mutations in metJ, rho, and rpsQ

implicated translation and transcription as targets of ethanol.MetJ represses expression of genes involved in methionine bio-synthesis (31). Rho terminates transcription uncoupled fromtranslation, halting mRNA synthesis when translation terminatesprematurely (29). RpsQ is ribosomal protein S17, an essentialcomponent of the 30S subunit that plays a role in maintainingtranslational accuracy (32). We focused on phenotypes relatedto these three targets to study the effects of ethanol on transcriptionand translation within the cell.

Ethanol Induces Toxic Translational Misreading. The evolved rpsQ[H31P] allele is identical to the nea301 rpsQ allele previouslyselected as a neamine-resistance mutation (33). The nea301 al-lele encodes a variant of protein S17 that increases translationalaccuracy in vitro in the presence of ethanol and other misreadingagents (25). The in vivo effects of ethanol on translational ac-curacy have not been reported, but isolation of a hyperaccurateribosomal mutation led us to hypothesize that ethanol wouldstimulate translational misreading, which would in turn be re-duced by the evolved rpsQ allele.We tested this hypothesis by measuring misreading in the

presence of varying concentrations of ethanol using a sensitiveassay developed by Kramer and Farabaugh (34). In this assay, anactive-site codon of firefly luciferase (F-luc) is mutated (K529N)such that it produces an inactive enzyme if correctly translated;if ribosomal error leads to insertion of a lysine at this position,an active enzyme is produced instead. The firefly luciferase istranslationally fused to a wild-type jellyfish luciferase active un-der different chemical conditions, which was used to control forchanges in transcription/translation rate, protein stability, andother factors. All measurements of mutant F-luc activity werenormalized to an isogenic strain expressing wild-type F-luc fusedto jellyfish luciferase to control for effects of ethanol on the twoenzymes. Ethanol caused a dose-dependent increase in normal-ized F-luc activity expressed by wild-type E. coli, up to ninefoldhigher than untreated controls at 40 g of EtOH/L (Fig. 2A),representing an increase in the miscoding error rate similar tothat previously reported for inhibitory concentrations of strep-tomycin (34). Strikingly, the ethanol-tolerant triple mutant andan otherwise wild-type strain bearing the rpsQ[H31P] alleleshowed dramatically reduced levels of F-luc activity at all con-centrations of ethanol, such that F-luc activity in the mutantstrains grown at 40 g of EtOH/L was similar to wild-type cultureswithout ethanol. These results demonstrate that ethanol increasestranslational misreading in vivo and that the evolved rpsQ allelestrongly reduces translational misreading in the presence or absenceof ethanol.We assessed the physiological importance of ethanol-induced

errors in protein synthesis by measuring synergistic toxicity between

ethanol and antibiotics that affect different steps in translation.Unrelated stressors at sublethal concentrations generally havemultiplicative effects on culture growth when combined whereasstressors that are functionally related vary from this pattern (35,36). When combined with streptomycin, an antibiotic thatinduces translational misreading, ethanol caused approximatelyfourfold greater toxicity than predicted by the multiplicativemodel for unrelated stressors (Fig. 2B). In contrast, ethanol didnot show cooperative synergy with chloramphenicol, whichinhibits peptidyl transfer in the ribosome but does not decreasetranslational accuracy (37), suggesting that ethanol does not

Table 2. Selected mutations conferring ethanol tolerance

Gene Nucleotide change Amino acid change Measured effect of mutation

metJ AAAGAGATC→AAGATC ΔE91 Up-regulation of methionine biosynthesis gene expression (Table S2)rho CTG→ATG L270M Reduced transcription termination by Rho (Fig. S3)rpsQ CAC→CCC H31P Increased accuracy during protein synthesis (Fig. 2A)

µ1

untreated

treatedµ2

µ1

µ2

GlyAsnLeu

GlyLysLeu

Correcttranslation

Ethanol concentration (g/L)

Nor

mal

ized

luci

fera

se a

ctiv

ity

0.00

0.01

0.02

0.03

Wild-type

1.2x

5.1x

1.6x

1.8x

3.1x

8.8x

rpsQ

1.4x 2.2x6.1x

metJ rho rpsQ

0 10 20 40

Rel

ativ

e fit

ness

0.0

0.2

0.4

0.6

0.8

1.0

EtOH Str Str +EtOH

predict

Str +EtOHactual

Cm Cm +EtOH

predict

Cm +EtOHactual

Translational misreading

Synergy between ethanol and translation inhibitors

A

B

Time

log

A59

5

Fitness =

Log-phase growth rates

ggaAAUcuc

Mistranslation

Mutant lucmRNA

Inactiveluciferase

Activeluciferase

Fig. 2. Ethanol induces toxic translational error. All panels show meanvalues and SEM for at least three biological replicates. (A) Normalized K529Nfirefly luciferase activity for wild-type (MG1655), ethanol-tolerant triplemutant (EP61), and rpsQ[H31P] (EP49) in the presence of indicated ethanolconcentrations. Activity for the K529N mutant requires misincorporation oflysine at an AAU codon. Numbers above bars show fold increase in errorrelative to no-ethanol condition for the same strain. (B) Effects of treatmentof wild-type MG1655 with ethanol (40 g/L) and translational inhibitors (1.6μg Str/mL, 2 μg Cm/mL) singly and in combination. For combinatorial stresses,predicted fitness values based on a no-synergy model are shown.

E2578 | www.pnas.org/cgi/doi/10.1073/pnas.1401853111 Haft et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

020

impede the peptidyl transfer reaction. These results strongly in-dicate that sublethal ethanol stress induces physiologically dam-aging levels of translational misreading, perhaps compromisingthe conformation of the ribosomal decoding site known to betargeted by streptomycin (38, 39).The observed synergy between ethanol and streptomycin sug-

gested that ribosomal proteins other than RpsQ classically impli-cated in translational accuracy, like RpsL (protein S12), might alsoplay a role in ethanol tolerance. Consistent with this hypothesis,we observed that the rpsL150 allele (40), which confers strepto-mycin resistance, increased fitness during growth at a modestconcentration (20 g/L) of ethanol (Fig. S2). Ethanol has also beenreported to rescue streptomycin-dependent mutants with lesionsmapping at or near rpsL (41), which generally display a hyper-accurate translation phenotype (42), implying a potent effect ofethanol on ribosome decoding activity.

Ribosome Profiling Reveals Ethanol Induction of Ribosomal Termination.Translational misreading has been shown to cause ribosome stall-ing and termination (43), and (p)ppGpp production in ethanol-treated E. coli may be a consequence of stalled ribosomes (27). Toinvestigate whether ethanol stress might cause aberrant termina-tion of polypeptide synthesis, we assessed the genome-wide effectsof ethanol on mRNA levels and ribosomal distribution acrosscellular mRNAs using ribosome profiling coupled with RNAseq(44). These techniques measure mRNA abundance and ribosomaloccupancy with high resolution, revealing the density of ribosomes

on various mRNAs and ribosome abundance along the lengths ofmRNA transcripts.We designed the ribosome-profiling experiment to compare

three physiological conditions apparent in ethanol-stress experi-ments (Fig. 1B): logarithmic growth before ethanol stress (T0), theperiod of growth cessation due to acute stress (T1), and the phasein which cells had resumed growth under chronic ethanol stress(T2). Practical and cost limitations on the number of samplesprocessed precluded examining the single and double rho, metJ,and rpsQmutants individually; thus, we limited the experiment tocomparing the triple mutant to wild type at the three indicatedtime points.To assess ribosome processivity, we examined the change in

relative ribosome occupancy from the 5′ end to the 3′ end ofORFs, dividing each ORF into eight even segments and averagingsignals across sets of genes. Decreases in relative occupancy at the3′ end of messages would indicate aberrant translational termi-nation between the 5′ and the 3′ end of the ORF. Because ribo-some occupancy is the ratio of site-specific ribosome abundanceover site-specific mRNA abundance, we first quantified mRNAsignals from RNAseq across ORFs (Fig. 3 A–C) and then used theratio of raw ribosome footprint signal to mRNA signal to generateplots of ribosome occupancy (Fig. 3 D–F). To facilitate genome-wide comparisons, we identified a set of 3,048 protein-codinggenes with unambiguously mapped reads in all RNAseq andribosome footprinting datasets. Initially, we looked for evidenceof translation termination by averaging relative occupancy values

All genes: ribosomesAll genes: mRNA

1 2 3 4 5 6 7 80.8

0.9

1.0

0.8

0.9

1.0

WT T0WT T1 *WT T2

Mut T0Mut T1 *†Mut T2

Rel

ativ

em

RN

A s

igna

l

ORFsegments

1 2 3 4 5 6 7 80.7

0.8

0.9

1.0

Rel

ativ

erib

osom

es/m

RN

A

0.7

0.8

0.9

1.0

Mut T0Mut T1 *Mut T2

WT T0WT T1 *WT T2

1 2 3 4 5 6 7 8

ORFsegments

1 2 3 4 5 6 7 8

Low ribosome density: mRNA

0.7

0.8

0.9

1.0

0.7

0.8

0.9

1.0

T0 †T1 *†T2 *†

Rel

ativ

em

RN

A s

igna

l

T0T1 *T2 *

1 2 3 4 5 6 7 81 2 3 4 5 6 7 8 0.6

0.7

0.8

0.9

1.0

0.6

0.7

0.8

0.9

1.0

T0T1T2

T0T1T2

Low ribosome density: ribosomes

Rel

ativ

erib

osom

es/m

RN

A

1 2 3 4 5 6 7 81 2 3 4 5 6 7 8

High ribosome density: mRNA

0.7

0.8

0.9

1.0

0.7

0.8

0.9

1.0

Rel

ativ

em

RN

A s

igna

l

T0T1T2

T0T1T2

1 2 3 4 5 6 7 81 2 3 4 5 6 7 8

High ribosome density: ribosomes

0.6

0.7

0.8

0.9

1.0

0.6

0.7

0.8

0.9

1.0

T0T1 *T2

T0T1 *T2

Rel

ativ

erib

osom

es/m

RN

A

1 2 3 4 5 6 7 81 2 3 4 5 6 7 8

* = Significantly different from same strain at T0 (P < 0.01) † = Significantly different than wild-type at same time point (P < 0.01)

mRNA abundance

RNAseq

ORFORF

mRNA densityRibosome profiling

nuclease digestion footprintsequencing

ORFORF

D

E

F

A

B

C

mRNA

# re

ads

ORF

ribosome

# re

ads

ORF

ribosome density

Fig. 3. Ribosome profiling reveals ethanol-induced ribosome halting and Rho activity. (Top) A summary of ribosome profiling sample treatment for side-by-side RNAseq and ribosome footprinting, adapted from Ingolia et al. (44). (A–C) Relative mRNA signal levels from 5′ to 3′ for groups of ORFs: “all genes”(the 3,048 genes represented in mRNA and ribosome footprint datasets), high ribosome density quintile (610 genes), and low ribosome density quintile (610genes). (D–F) Relative ribosome occupancy signal levels (ribosome footprint signal divided by mRNA signal) from 5′ to 3′ for groups of ORFs as in A–C. Datafrom wild-type (MG1655) cultures are shown in gray squares, and data from the ethanol-tolerant triple mutant (EP61) cultures are shown in colored triangles.Error bars show SEM. T0, T1, and T2 represent prestress, acute stress, and chronic stress conditions (Fig. 1B).

Haft et al. PNAS | Published online June 9, 2014 | E2579

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

020

across this set of “all genes” (Fig. 3D). Ethanol-stimulatedtranslational termination was evident and followed similar pat-terns in both wild-type and mutant cultures. Before stress,ribosomes were widely distributed across mRNAs. During acutestress (T1), relative ribosomal occupancy near the 3′ ends ofgenes dropped from ∼0.95 to ∼0.75. After cells resumed growthunder chronic stress (T2), genome-wide ribosome occupancyshifted back to near prestress patterns. These observations in-dicate that the acute stress phase, which is coupled to growthcessation, was characterized by widespread aberrant terminationof translation within mRNAs. By the time growth resumed, bothstrains had largely corrected this defect.In analyzing the ribosome profiling data, we observed that

mRNAs with higher prestress ribosome density exhibited greaterdecreases in 3′ ribosome occupancy. To assess this pattern fur-ther, we separated the “all gene” set into quintiles based on ri-bosome density in unstressed wild-type cells and examined theupper and lower quintiles. The upper ribosome density quintileexhibited a decrease in occupancy at the 3′ end of genes that was,on average, greater in magnitude than the analogous decrease inthe set of all genes (Fig. 3E). This result indicates that a highdensity of ribosomes on the message could not prevent ethanol-induced ribosomal halting. The low ribosome density quintile, incontrast, did not exhibit such a decrease in ribosome occupancy(Fig. 3F). Although this flatter ribosome occupancy curve couldindicate a smaller effect of ethanol on translation of low ribosomedensity mRNAs, a likely alternative explanation is that translationtermination on these messages might be coupled to termination oftranscription. Because ribosome occupancy is ribosome abun-dance corrected for mRNA abundance, parallel decreases in bothwould appear as relatively flat ribosome occupancy across a mes-sage, as observed for the low-density quintile.Coupling of ethanol-induced translation termination to tran-

scription termination could explain why we and others found thatmutations reducing Rho activity confer ethanol tolerance. Rhoterminates transcription of untranslated mRNA molecules; thus,poorly translated mRNAs might be particularly susceptible toethanol-induced Rho-dependent termination because an etha-nol-induced increase in ribosomal termination could not becompensated by other ribosomes on the mRNA. We thereforeused the RNAseq data to assess changes in intragenic tran-scription termination after ethanol treatment.

Rho-Dependent Transcription Termination Is Increased During EthanolStress. Intragenic transcription termination results in decreasedmRNA levels from the 5′ end to the 3′ end of ORFs. We testedfor this pattern by dividing the RNAseq reads into the same genesets used to analyze ribosome occupancy (Fig. 3 A–C). In the “allgene” gene set, we observed statistically significant decreases in 3′abundance relative to 5′ abundance, indicative of transcriptionaltermination, during acute stress for both wild type and the mutant(T1) (Fig. 3A). This change was markedly greater in magnitude forwild-type cells, suggesting that the variant Rho expressed by themutant may reduce intragenic transcription termination. However,mRNAs with high ribosome density did not exhibit obvious eth-anol-induced decreases in 3′-proximal signal (Fig. 3B), suggestingthat high ribosome density may protect transcripts from termination.The low ribosome-density gene set exhibited a similar pattern

to the set of all genes, but with stronger and more persistenteffects (Fig. 3C). For these genes, mRNA levels were maximal atthe 5′ end and dropped as transcription proceeded toward the 3′end in both strains. The decrease in mRNA level after the 5′ endof the gene was greatest during acute ethanol stress and only par-tially recovered during chronic stress, indicating that ethanolinhibited the ability of cells to produce full-length transcripts of thesegenes even after cells had resumed active growth. Although bothstrains exhibited this general pattern of transcription termination,the tolerant mutant had a clear advantage in producing full-

length transcripts, maintaining significantly higher mRNA levelsacross the length of genes than wild type at all time points.These results suggested that Rho[L270M] reduced transcription

termination. To test this possibility, we performed an RNAseqanalysis of a strain containing only the rho(L270M) mutation. Theexperiment confirmed the reduced rho activity phenotype, re-vealing increased readthrough of a set of previously defined Rho-dependent terminators (45) in the rho[L270M] strain comparedwith wild type (Fig. S3). Of 114 Rho-dependent terminators forwhich fold change could be reliably calculated using our RNAseqdataset (Dataset S1), 84% exhibited increased terminator read-through in the mutant, and nearly half (45%) exhibited a twofoldor greater increase (Mann–Whitney U test; P < 0.001). Thus, therho[L270M] mutation confers a global defect in Rho-dependenttranscription termination.We concluded that, during ethanol stress, Rho aberrantly

terminates transcription of hundreds of genes. RNA polymerases(RNAPs) synthesizing mRNAs with fewer ribosomes than aver-age were more prone to Rho-dependent termination. Especiallyfor these poorly translated mRNAs, the reduced activity of Rho[L270M] favored production of full-length transcripts. Thus, ourresults are consistent with the hypothesis that ethanol-inducedtranslational termination uncouples translation from transcrip-tion, stimulating Rho to terminate mRNA elongation.

Ethanol Inhibits mRNA Synthesis by RNA Polymerase in Vitro. An-other possible way ethanol could increase Rho-dependent tran-scription termination is to slow transcript elongation by RNApolymerase, which would increase the time available for Rhoaction. To investigate this possibility, we tested whether modestconcentrations of ethanol could affect transcript elongation byRNAP in vitro in the absence of translation. In the presence ofethanol at concentrations of 30 g/L or 60 g/L, the average rate oftranscript elongation by E. coli RNAP was reduced by 10–30%,and transcriptional pausing at a subset of sites was exacerbated(Fig. 4). Thus, ethanol not only may increase chances for Rholoading by causing translational misreading and termination, butalso may increase Rho action by slowing RNAP and thus increasingthe kinetic window within which Rho can effect termination.

Ethanol Inhibits Translation of Nonstart AUG Codons. The resultsdescribed above (see Ribosome Profiling Reveals Ethanol Inductionof Ribosomal Termination) strongly implicated ribosome stalling inthe toxic effect of ethanol on E. coli. Such stalling could be ran-dom or could be biased toward specific sites in the genome. Wehypothesized that ribosomes in ethanol-treated cells might beprone to stop at AUG codons due to methionine limitation becausea mutation in the master repressor of methionine biosynthesis,metJ, contributed to ethanol tolerance. We therefore used the ri-bosome profiling data to assess ethanol-induced changes in ribo-some occupancy at start and nonstart AUG codons compared withcodons for other amino acids and stop codons (Fig. 5). Increasedcodon occupancy in the ribosome profiles reflects increased dwelltime of ribosomes at those codons relative to others, thus revealingcodons whose translation was inhibited by ethanol.Ethanol had small effects on ribosome occupancy at most

codons, but strongly affected occupancy at nonstart AUGcodons. Nonstart AUG occupancy dramatically increased duringacute toxicity (T1) and only partially recovered during chronictoxicity (T2) in both strains. This pattern correlates with theoverall pattern of translation termination observed by ribosomeprofiling (Fig. 3D), which was strongly induced at T1 but largelyrecovered by T2. The magnitude of the ethanol effect on nonstartAUG occupancy was less for the mutant than for wild type at bothtime points (Wilcoxon matched-pairs rank test; P < 0.001). Weinfer that the metJ[ΔE91] allele protects against ethanol stress atleast in part because it leads to increased expression of methioninebiosynthesis genes (Table S2), potentially increasing the methio-

E2580 | www.pnas.org/cgi/doi/10.1073/pnas.1401853111 Haft et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

020

nine pool available to the cell. Consistent with this hypothesis,deletion of the metJ repressor or addition of excess methionineimproved ethanol tolerance of wild-type E. coli (Fig. S4).In contrast to nonstart AUG codons, ribosome occupancy at

AUG start codons increased in the mutant after ethanol additionbut decreased in the wild-type strain (Fig. 5). In principle, rela-tive start codon occupancy reflects the rate of translation initi-ation relative to elongation (i.e., how much time that ribosomesspend at start codons relative to other codons). Because rates of

translation initiation and growth are positively correlated inE. coli (46), these different responses likely reflect the lessereffect of ethanol on mutant versus wild-type growth rates (<3×versus >6×, respectively) (Fig. 1B).

DiscussionOur study of the effects of ethanol on central dogma processeswas motivated by the discovery that significant ethanol tolerancein E. coli could be recapitulated by mutations in genes encoding

2 Minutes

4 Minutes

8 Minutes

0.3 0.6 0.9 1.2 1.5Transcript length (kb)

B

Rel

ativ

e In

tens

ity

A NoEtOH

60 g EtOH/L

Tran

scrip

tion

Time

30 g EtOH/L In vitro transcription -/+ EtOH

201

307

404

527622

RO

03060

2340270240

4410340300

8500450430

Mean transcript length (nt)

g EtOH/LTime (min)

mea

n ±

15%

sig

nal

No EtOH30 g EtOH/L60 g EtOH/L

Fig. 4. Ethanol slows transcription in vitro. (A) Representative gel lanes showing in vitro transcript elongation in the presence of 0 g, 30 g, or 60 g ethanol/L.Elongation complexes (10 nM) were halted at the end of a 26-nt C-less cassette and then exposed to ethanol. NTPs (30 μM each) were added subsequently,and products from 2-min, 4-min, and 8-min extensions were resolved by gel electrophoresis. RO indicates template run-off products. Dots to the left of the gelpanels indicate the mean transcript length of RNA products for the experiment shown, and vertical lines show regions containing 15% of the signal upstreamand downstream from the mean. (B) Densitometric traces of gel lanes in A. Height of traces corresponds to relative densitometric signal at a given nucleotideposition. Mean average transcript lengths from two independent experiments are shown.

Ala

Arg

Asn A

sp Cys

Gln

Glu

Gly His

Ile Leu

Lys

Met

(sta

rt)

Met

(non

-sta

rt)

Phe Pr

o

Ser

Thr

Trp

Tyr

Val

Stop

-0.4

0.0

0.4

0.8

1.2

WT: T1/T0WT: T2/T0 Mut: T1/T0Mut: T2/T0

T0: no EtOH T1: 10’ EtOH T2: 70’ EtOH

Log 2

T2/T0

T1/T0

Log 2

Δ codonoccupancy

codon

Log 2

cha

nge

in m

ean

codo

n oc

cupa

ncy

Ethanol-induced changes in ribosome occupancy by residue

WT T0 Mut T0 WT T2 Mut T21.4

1.6

1.8

2.0Non-start AUG codons

Mea

n co

don

occu

panc

y

Fig. 5. Ethanol-induced changes in ribosome occupancy by encoded amino acid. Log2 values of mean ethanol-induced ribosome occupancy changes forcodons encoding the indicated amino acid are shown. Changes were measured as the ratio of poststress occupancy to prestress occupancy. Data from wild-type (MG1655) cultures are shown in blue, and data from the ethanol-tolerant triple mutant (EP61) cultures are shown in red. The Inset at upper right showsgenome-wide mean ribosome occupancy values at nonstart AUG codons + SEM for T0 and T2.

Haft et al. PNAS | Published online June 9, 2014 | E2581

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

020

a ribosomal protein involved in decoding (RpsQ), the masterfeedback repressor of methionine biosynthesis (MetJ), and a tran-scription factor that terminates transcription when it becomesuncoupled from translation (Rho). These three mutations ex-hibited specific effects on transcription and translation consis-tent with the hypothesis that ethanol alters the conformations ofribosomes and RNAP in ways that increase misreading, ribosomestalling, and RNAP pausing, leading to increased Rho terminationwhen transcription and translation become uncoupled. RpsQ[H31P]increased translational fidelity and suppressed ethanol-inducedmisreading (Fig. 2). MetJ[ΔE91] up-regulated expression of me-thionine biosynthesis genes and may have partially countered anethanol-induced stalling of ribosomes at nonstart AUG codons(Table S2, Fig. 5, and Fig. S4). Rho[L270M] reduced transcripttermination at Rho-dependent terminators and contributed toamelioration of ethanol-induced premature termination on poorlytranslated genes (Fig. S3 and Fig. 3). Finally, we found that ethanoldirectly slows transcript elongation by RNA polymerase, whichcould increase opportunities for intragenic Rho-dependent ter-mination (Fig. 4).Taken together, these data indicate that multiple effects on

the transcription and translation machinery are important com-ponents of cellular ethanol stress (Fig. 6). In the presence ofethanol, increased ribosome stalling (particularly at nonstartAUG codons) and increased ribosome misreading inhibit poly-peptide synthesis and increase the levels of misfolded and ab-errant proteins in the cell. Ethanol-induced misreading errorsmay exacerbate stalling and chain termination due to ribosomalproofreading after peptide bond formation (43). Slower trans-lation and increased translational termination can decoupletranslation from transcription, promoting Rho-dependent ter-mination of transcription within genes. Ethanol-induced slowingof RNAP further sensitizes transcription to termination by Rho.The ethanol tolerance conferred bymetJmutations may reflect

the inhibitory effects of ethanol on translation that are affectedby methionyl-tRNAMet levels. Ethanol increases ribosome dwell

time on internal AUG codons (Fig. 5), suggesting that methionyl-tRNAMet becomes limiting for protein synthesis; thus, elevatedmethionine levels might partially compensate for this effect.Consistent with this hypothesis, methionine supplementation ormetJ deletion increased ethanol tolerance of wild-type E. coligrown in minimal medium (Fig. S4). Deletions and point muta-tions in metJ are known to increase methionine levels in E. coli(47, 48); the metJ[ΔE91] mutation likely causes a similar effectby elevating transcription of the metABFIKNR genes (Table S2).Interestingly, elevating methionine biosynthesis or supplementingwith methionine also compensates for other stresses in E. coli,including acetate, organic acids, nitrosating agents, and heat shock(49–51). MetA, which catalyzes the first step in methionine bio-synthesis, is known to aggregate during heat stress (51), suggestingthat stress-induced MetA aggregation may reduce functionalMetA and resultant methionine levels. Ethanol is a potent inducerof the heat shock response (27) and thus could similarly causeMetA aggregation, resulting in decreased methionine andmethionyl-tRNAMet levels that could be partially corrected bythe metJ[ΔE91] mutation. Although other compensatory effectson cell growth of elevating methionine biosynthesis remainpossible, effects on translation are consistent with our detectionof ethanol-induced stalling on internal Met codons.In addition to causing toxic effects related to methionine, our

data indicate that ethanol interferes with the processes by whichribosomes and RNAP produce proteins and RNA. Ethanol mayalter the activities of the ribosome and RNAP through inter-actions at specific sites or through less-specific solute effects onmacromolecular conformation. The latter possibility is attractivebecause both the ribosome and RNAP are multisubunit com-plexes whose activities require significant conformationalchanges during repeated cycles of chain extension and becausethe ethanol levels at which toxic effects occur (1-2 M) are suf-ficient to alter the activities of water and solutes at proteinsurfaces (52). Indeed, solute effects are known to alter RNAPpausing and elongation (53) in patterns and magnitude similar tothose we observed for ethanol (Fig. 4). Ethanol appears to de-stabilize proteins by promoting exposure of hydrophobic regions(54, 55) through direct interactions (56). Ethanol destabilizationof protein structure via global effects as a solute may contributeto the well-known induction of the unfolded protein response(heat-shock response) by ethanol in E. coli (57) although etha-nol-induced amino acid misincorporation to generate nonnativeproteins also could contribute. However, ethanol is less wellstudied than other protein-perturbing solutes, and both mRNAdecoding by the ribosome and pausing and elongation by RNAPare complex phenomena. Thus, further study will be needed tounderstand which parts of these complex molecular machinesmay be altered to cause the effects we observed.Nonetheless, our finding that ethanol causes decoding errors

similarly to inhibitors such as streptomycin, coupled with pre-vious findings, suggests that ethanol may alter the conformationof the decoding center of the small ribosomal subunit. Ribo-somes select the correct aminoacyl-tRNA by coupling correctcodon-anticoding pairing in the decoding center to GTP hydro-lysis by elongation factor Tu (EF-Tu), leading to tRNA accom-modation into the peptidyl transferase center (58). Ethanolincreases ribosomal misreading in vitro (24–26) and causeschanges to the ribosome footprint in a manner similar to anti-biotics such as streptomycin and neomycin (59). A mutation inrpsL, encoding a ribosomal protein known to be important forstreptomycin binding and accurate translation (S12), was shownto increase ethanol tolerance in an E. coli rho mutant back-ground (14). Additionally, ethanol tolerance is conferred on E.coli by overexpression (17) of RlmH, which methylates a 23SrRNA pseudouridine near the ribosomal decoding center (60),and of TruB, which catalyzes formation of pseudouridine-55 ontRNAs and is important for efficient translation of certain

mRNA

Correctly translatedpolypeptide Ribosome

5’

EthanolIncreased stalling

Increased pausing

Rho terminationIncreased misreading

RNA polymerase

Translationerror

Coupled transcription and translation with no ethanol

Toxic effects of ethanol on transcription and translation

Fig. 6. Summary of ethanol effects on transcription and translation. In un-stressed cells, transcription and translation are coupled by RNAP pausing re-lieved by ribosome translocation (Upper). Ethanol exerts direct effects onribosomes and RNAP that increase RNAP pausing required for Rho terminationand uncouple transcription and translation via effects on the ribosomedecoding center that allow Rho access to RNAP and decrease translationalaccuracy (Lower).

E2582 | www.pnas.org/cgi/doi/10.1073/pnas.1401853111 Haft et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

020

codons (61). Finally, ethanol has been reported to promote thegrowth of streptomycin-dependent mutants of E. coli in the ab-sence of streptomycin (41, 62), suggesting that ethanol, like strep-tomycin, may induce compensatory conformational changes in thedecoding center of the ribosome. We propose that ethanol disruptsthe natural conformation of the ribosomal decoding center andthus, like streptomycin, allows EF-Tu GTP hydrolysis and accom-modation upon binding of noncognate aminoacyl-tRNAs. An eth-anol-induced rearrangement of the decoding center might alsoaffect the propensity of the ribosome to stall and possibly its abilityto recognize methionyl-tRNA although other, distal effects of eth-anol on the ribosome also could cause translational halting.Our proposal that ethanol exerts its toxic effects on E. coli in

part through direct effects on the ribosome and RNAP contrastswith previous suggestions that ethanol tolerance mutations mod-ifying proteins involved in transcription and translation functionindirectly by “rewiring” gene-expression networks. For example,increased alcohol tolerance of strains with reduced Rho activity(14, 20) or mutations in RNA polymerase subunits or the TATA-binding protein (4, 22, 63) have been attributed to altered ex-pression of specific genes controlled by termination or initiation.Our results do not preclude the possibility that increased expres-sion of some genes contributes to ethanol tolerance; both effectson the transcription/translation machinery and effects on geneexpression may occur, and both may be important. Rather, wesuggest that the potentially important and simpler explanation thatcompensatory mutations may help shield the ribosome and RNAPfrom direct effects of ethanol should not be overlooked. Indeed,some indirect effects of ethanol may also be consequences ofprimary effects on the ribosome or RNAP. For example, wespeculate that the ethanol-induced changes in E. coli membranelipid composition (10) might be caused in part by ethanol’s effectson the ribosome. Ethanol stress results in (p)ppGpp production(27), likely by the RelA synthase bound to stalled ribosomes. (p)ppGpp binds numerous cellular enzymes (64) and is known toinhibit the activity of the phospholipid synthesis enzyme PlsB invivo and in cell-free extracts (65, 66). Mutation of plsB alters fattyacid composition similarly to effects of ethanol (increased un-saturated C18:1 fatty acids) (67), providing a potential link betweenthe effects of ethanol on the ribosome and on membrane com-position (10, 11) mediated by (p)ppGpp. Increased (p)ppGpplevels affect transcription from numerous promoters in the cell(28, 68), which could lead to other downstream effects as well.Thus, our finding that ethanol induces ribosome stalling also po-tentially provides a simple mechanistic explanation for ethanolstimulation of (p)ppGpp production and consequent effects onmultiple cellular functions as a signal of translational stress.A key question raised by our work is whether modes of ethanol

tolerance related to its direct effects on the ribosome and RNAPare broadly applicable to other solvents and other microbes.Some ethanol-tolerant mutants of Saccharomyces cerevisiae carryribosomal mutations that alter antibiotic sensitivities (69), sug-gesting that the toxic effects of ethanol on the ribosome mayoperate in yeast as well as bacteria. Do naturally ethanol-tolerantspecies such as S. cerevisiae protect their ribosomes from etha-nol-induced translational error and their RNAPs from ethanol-induced transcriptional slowing? Do other small-molecule sol-vents have ethanol-like toxic effects? The answers to these questionscould have important consequences for biosynthetic engineering. Iftranslation and transcription are widely prone to inhibition by sol-vents such as ethanol, then engineering microbes with resilientribosomes and RNAPs may enhance biological production of a va-riety of small molecules. More generally, our results highlight theimportance of considering direct effects on central dogma processeswhen evaluating the effects of solutes on microbes.

Materials and MethodsBacterial Strains, Media, and Growth Conditions. All strains are derivatives ofE. coli K12 strain MG1655 (Table 1). Strains MTA156, MTA157, and MTA160were selected after 24 aerobic serial passages of MG1655 in M9 minimal me-dium (70), supplemented with MgSO4 (1 mM), CaCl2 (0.1 mM), and glucose(10 g/L), and increasing amounts of ethanol up to 65 g/L (Fig. 1). We isolatedMTA156, MTA157, and MTA160 as clonal strains from the evolved culturesby single-colony purification. The genomes of MG1655, MTA156, MTA157,and MTA160 were sequenced at the University of Wisconsin-Madison (UW-Madison) Biotechnology Center using the Illumina HiSeq 2000 platform. Un-marked transfer of alleles between MG1655 and MTA156 was achieved by P1transduction, first transducing in an auxotrophic marker from the Keio col-lection (71) linked to the desired locus and then selecting for prototrophictransductants from the appropriate donor, as previously described for rhomutations (72). EP23 was constructed by P1 transduction of the metJ::kanmutation from the Keio collection into MG1655. RL2739 was constructed by λRed recombination of the rpsL150 allele into MG1655 after amplification fromstrain DH10B (40) by PCR using forward primer 5′-GCCTGGTGATGGCGGGATCGand reverse primer 5′-CGCGACGACGTGGCATGGAA.

Fermentation Experiments. For fermentation data shown in Fig. 1E, strainsMTA376 and MTA722 were constructed by introducing the ldhA::kan allelefrom the Keio collection into MG1655 and EP61, respectively, by transduction,followed by removal of the kan marker by FLP recombinase (71). The etha-nologenic PET cassette was expressed in each strain from plasmid pJGG2 (52).Fermentative cultivations were performed in stirred flasks incubated at 37 °Cin an anaerobic chamber with an atmosphere of 10% CO2 + 10% H2. Themedium for fermentations was synthetic corn stover hydrolysate (73) sup-plemented with glucose (60 g/L) and xylose (30 g/L) to mimic 9% glucanloading. End product detection was done as described previously (73).

Rho-Dependent Terminator Readthrough Analysis. MG1655 and RL2325 cul-tures were grown, RNAwas harvested, and RNAseq libraries were prepared asdescribed (30). Sequencing was performed by the Joint Genome Instituteusing an Illumina Genome Analyzer II. Normalized RNAseq reads were sum-med within regions previously defined by terminator readthrough in Rho-inhibited cells (45). Fold changes in terminator readthrough were calculatedby averaging the read counts for two biological replicates, and then dividingthe mutant read count by the wild-type read count. Fold change could not becalculated for terminators with an average read count of zero for the wild-type samples; such terminators were removed from the final analysis. Sta-tistical significance was tested by comparing fold changes for Rho-dependentterminators versus random sites in the genome of the same length (obtainedby rotating the positions of terminator readthrough by 1 Mb).

In Vitro Transcription Assay. Template DNA was made from plasmid pMK110,which was constructed from pIA267 (74) by inserting a region from the E. colibgl operon into the SpeI site using forward primer 5′-GCGAGCACTAGTTGTT-CAAGAATACGCCAGGA and reverse primer 5′-GCGAGCACTAGTGGCGATGA-GCTGGATAAACT. pMK110 template DNA contains a λPR promoter followedby a 26-nucleotide C-less cassette. The linear pMK110 template for transcrip-tion reactions was generated by PCR amplification using forward primer 5′-CGTTAAATCTATCACCGCAAGGG and reverse primer 5′-CAGTTCCCTACTCTCG-CATG. PCR products were electroeluted from an agarose gel and phenolextracted. Core E. coli RNAP and E. coli σ70 were purified as described pre-viously (75, 76). RNAP holoenzyme (core ββ′α2 plus σ70) was prepared by in-cubating twofold molar excess of σ70 with core for 30 min at 30 °C.

Halted elongation complexes were formed by incubating 10 nM linearpMK110 template and 15 nM RNAP holoenzyme in transcription buffer [40mM Tris·HCl (pH 8.0), 50 mM KCl, 5 mM MgCl2, 0.5 mM DTT, and 5% (vol/vol)glycerol] with 150 μMApU, 10 μM ATP and UTP, 2.5 μMGTP, and 0.37 μM (10μCi) [α-32P]GTP for 10 min at 37 °C to stall complexes 26 nucleotides down-stream of the transcription start site. Resumption of transcription wasallowed by adding 30 μM each ATP, UTP, CTP, and GTP, 100 μg rifampicin/mL, and 0.1 U RNasin/μL (Promega). Ethanol was added where indicated.Samples were taken at indicated time points by mixing with an equal vol-ume of 2× stop dye (8 M urea, 30 mM Na2EDTA, and 0.05% bromophenolblue and xylene cyanol). Samples were heated for 2 min at 90 °C andseparated by electrophoresis in a denaturing 6% polyacrylamide gel (19:1acrylamide:bisacrylamide) in 7 M urea, 1.25 mM Na2EDTA, and 44 mM Trisborate (pH 8.3). Gels were exposed to a PhosphorImager screen, which wasscanned using a Typhoon PhosphorImager and quantified using Image-Quant software (GE Healthcare).

Haft et al. PNAS | Published online June 9, 2014 | E2583

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

020

Densitometry profiles were generated by converting pixels from thePhosphorImager scan to transcript positions by comparison with end-labeledMspI fragments of pBR322 using a 6-factor polynomial function. Meantranscript length was calculated from the summed products of the transcriptlength (in nt) times signal intensity divided by total signal intensity.

Ethanol–Antibiotic Synergy Experiments. Indicated strains (MG1655 andRL2739) were cultured in 96-well plates (BD Falcon) in Neidhardt rich mediumwith 2 g glucose/L (77) supplemented with ethanol, translation inhibitors, orboth. Wells were topped with mineral oil to prevent ethanol evaporation asdescribed (78, 79), and growth was tracked by measuring the absorbance at595 nm in a Tecan Infinite F200 plate reader. Fitness was defined as the ratioof stressed to unstressed logarithmic growth rates.

Measurement of Translational Misreading. Cultures for misreading experi-ments were diluted from unstressed overnight cultures and grown to mid-logarithmic phase in aerobic tubes in Neidhardt rich mediumwith 2 g glucose/L(77) supplemented with 100 μg ampicillin/mL (to maintain luciferase-expressingplasmids), 0–40 g EtOH/L, and 10–100 μM isopropyl β-D-1-thiogalactopyranoside(IPTG). Equivalent numbers of cells were harvested from each culture. Cells werelysed, and firefly luciferase (F-luc) and jellyfish luciferase (R-luc) activities weremeasured as described (34). Luminescence was measured using a TecanInfinite F200 plate reader. Values are expressed as F-luc/R-luc ratios nor-malized to the F-luc/R-luc ratio of an isogenic strain carrying the wild-typefirefly luciferase to control for any differences between cultures and dif-ferential effects of ethanol on the two enzymes.

Ribosome Profiling. Cultures were grown aerobically in 1-L volumes of M9minimal medium (70), supplemented with MgSO4 (1 mM), CaCl2 (0.1 mM),and glucose (10 g/L) in vigorously shaken (225 rpm) Fernbach flasks. Ethanolwas added to 40 g of EtOH/L once A600 of cultures reached 0.3. This ethanolconcentration was chosen because it allowed comparison of wild-type andmutant phenotypes under conditions in which both could grow and both

exhibited growth responses to ethanol. Ribosome profiling and librarygeneration were performed as described by Oh et al. (80), with cells har-vested by rapid vacuum filtration onto 0.2-μm filters and flash-frozen inliquid N2 to arrest ribosomes. After cryogenic lysis, monosomes were purifiedby sucrose gradient fractionation.

Sequencing was performed at the UW-Madison Biotechnology Centerusing an Illumina HiSeq 2000 set for 50-bp single-end reads. Raw reads weretrimmed by 2 nt from the 5′ end (to remove any nontemplated nucleotidesadded by reverse transcriptase) and were then mapped using the Burrows-Wheeler Aligner (81) to the MG1655 genome (National Center for Bio-technology Information accession no. NC_000913). Reads were not trimmedto individual codons to avoid errors arising from the variation in ribosomefootprint dimensions at different genomic locations (82). Signals mapping tononcoding RNA regions were removed from the dataset, and each datasetwas normalized to reads per million per position before further analysis.

Data manipulation and analyses were performed with custom Perl scripts.Statistical analyses were performed using GraphPad Prism (GraphPad Soft-ware). For metagene analyses (gene-segment analyses and codon-typeanalyses), pseudogenes and genes not represented in one or more datasetswere excluded, leaving 3,048 genes in the “all genes” dataset. High-trans-lation and low-translation quintiles were defined as the gene sets withhighest/lowest ribosome-occupancy-to-RNA signal ratios in MG1655 beforeethanol treatment. For gene-segment analyses, differences between data-sets were assessed using area under the curve (rectangular midpointapproximations) for each gene in the set.

ACKNOWLEDGMENTS. We thank our Great Lakes Bioenergy ResearchCenter collaborators and R.L. laboratory colleagues for critical reading ofthe manuscript. We are grateful to Gene-Wei Li and David Burkhardt foradvice and assistance with ribosome profiling experiments, P. Chu for tech-nical assistance, and Gerwald Jogl for helpful discussions of ribosome decod-ing. This work was funded by the Department of Energy Great Lakes BioenergyResearch Center (DOE BER Office of Science Grant DE-FC02-07ER64494).

1. Yomano LP, York SW, Ingram LO (1998) Isolation and characterization of ethanol-

tolerant mutants of Escherichia coli KO11 for fuel ethanol production. J Ind MicrobiolBiotechnol 20(2):132–138.

2. Dien BS, Cotta MA, Jeffries TW (2003) Bacteria engineered for fuel ethanol pro-duction: Current status. Appl Microbiol Biotechnol 63(3):258–266.

3. Zhao XQ, Bai FW (2009) Mechanisms of yeast stress tolerance and its manipulation forefficient fuel ethanol production. J Biotechnol 144(1):23–30.

4. Alper H, Moxley J, Nevoigt E, Fink GR, Stephanopoulos G (2006) Engineering yeasttranscription machinery for improved ethanol tolerance and production. Science

314(5805):1565–1568.5. Fischer CR, Klein-Marcuschamer D, Stephanopoulos G (2008) Selection and optimi-

zation of microbial hosts for biofuels production. Metab Eng 10(6):295–304.6. Wackett LP (2008) Biomass to fuels via microbial transformations. Curr Opin Chem

Biol 12(2):187–193.7. Ingram LO (1990) Ethanol tolerance in bacteria. Crit Rev Biotechnol 9(4):305–319.8. Sikkema J, de Bont JA, Poolman B (1995) Mechanisms of membrane toxicity of hy-

drocarbons. Microbiol Rev 59(2):201–222.9. Segura A, et al. (2012) Solvent tolerance in Gram-negative bacteria. Curr Opin Bio-

technol 23(3):415–421.10. Dombek KM, Ingram LO (1984) Effects of ethanol on the Escherichia coli plasma

membrane. J Bacteriol 157(1):233–239.11. Ingram LO, Vreeland NS (1980) Differential effects of ethanol and hexanol on the

Escherichia coli cell envelope. J Bacteriol 144(2):481–488.12. Brynildsen MP, Liao JC (2009) An integrated network approach identifies the iso-

butanol response network of Escherichia coli. Mol Syst Biol 5:277.13. Gonzalez R, et al. (2003) Gene array-based identification of changes that contribute

to ethanol tolerance in ethanologenic Escherichia coli: Comparison of KO11 (parent)to LY01 (resistant mutant). Biotechnol Prog 19(2):612–623.

14. Freddolino PL, Goodarzi H, Tavazoie S (2012) Fitness landscape transformationthrough a single amino acid change in the Rho terminator. PLoS Genet 8(5):e1002744.

15. Goodarzi H, Hottes AK, Tavazoie S (2009) Global discovery of adaptive mutations. NatMethods 6(8):581–583.

16. Nicolaou SA, Gaida SM, Papoutsakis ET (2012) Exploring the combinatorial genomicspace in Escherichia coli for ethanol tolerance. Biotechnol J 7(11):1337–1345.

17. Woodruff LB, Boyle NR, Gill RT (2013) Engineering improved ethanol production inEscherichia coli with a genome-wide approach. Metab Eng 17:1–11.

18. Woodruff LB, et al. (2013) Genome-scale identification and characterization of eth-anol tolerance genes in Escherichia coli. Metab Eng 15:124–133.

19. Zingaro KA, Papoutsakis ET (2012) Toward a semisynthetic stress response system toengineer microbial solvent tolerance. MBio 3(5):e00308–12.

20. Goodarzi H, et al. (2010) Regulatory and metabolic rewiring during laboratory evo-lution of ethanol tolerance in E. coli. Mol Syst Biol 6:378.

21. Chong H, et al. (2013) Improving ethanol tolerance of Escherichia coli by rewiring itsglobal regulator cAMP receptor protein (CRP). PLoS ONE 8(2):e57628.

22. Alper H, Stephanopoulos G (2007) Global transcription machinery engineering: A newapproach for improving cellular phenotype. Metab Eng 9(3):258–267.

23. Luo LH, et al. (2009) Improved ethanol tolerance in Escherichia coli by changing thecellular fatty acids composition through genetic manipulation. Biotechnol Lett 31(12):1867–1871.

24. Friedman SM, Berezney R, Weinstein IB (1968) Fidelity in protein synthesis: The role ofthe ribosome. J Biol Chem 243(19):5044–5048.

25. Phoenix P, Melançon P, Brakier-Gingras L (1983) Characterization of mutants of Es-cherichia coli with an increased control of translation fidelity. Mol Gen Genet 189(1):123–128.

26. So AG, Davie EW (1964) The effects of organic solvents on protein biosynthesis andtheir influence on the amino acid code. Biochemistry 3:1165–1169.

27. VanBogelen RA, Kelley PM, Neidhardt FC (1987) Differential induction of heat shock,SOS, and oxidation stress regulons and accumulation of nucleotides in Escherichiacoli. J Bacteriol 169(1):26–32.

28. Potrykus K, Cashel M (2008) (p)ppGpp: still magical? Annu Rev Microbiol 62:35–51.29. Peters JM, Vangeloff AD, Landick R (2011) Bacterial transcription terminators: The

RNA 3′-end chronicles. J Mol Biol 412(5):793–813.30. Peters JM, et al. (2012) Rho and NusG suppress pervasive antisense transcription in

Escherichia coli. Genes Dev 26(23):2621–2633.31. Marincs F, Manfield IW, Stead JA, McDowall KJ, Stockley PG (2006) Transcript analysis

reveals an extended regulon and the importance of protein-protein co-operativity forthe Escherichia coli methionine repressor. Biochem J 396(2):227–234.

32. Bollen A, Cabezón T, de Wilde M, Villarroel R, Herzog A (1975) Alteration of ribo-somal protein S17 by mutation linked to neamine resistance in Escherichia coli. I.General properties of neaA mutants. J Mol Biol 99(4):795–806.

33. Yaguchi M, et al. (1976) Alteration of ribosomal protein S17 by mutation linked toneamine resistance in Escherichia coli. II. Localization of the amino acid replacementin protein S17 from a neaA mutant. J Mol Biol 104(3):617–620.

34. Kramer EB, Farabaugh PJ (2007) The frequency of translational misreading errors inE. coli is largely determined by tRNA competition. RNA 13(1):87–96.

35. Andrusiak K, Piotrowski JS, Boone C (2012) Chemical-genomic profiling: Systematicanalysis of the cellular targets of bioactive molecules. Bioorg Med Chem 20(6):1952–1960.

36. Barker CA, Farha MA, Brown ED (2010) Chemical genomic approaches to study modelmicrobes. Chem Biol 17(6):624–632.

37. Lando D, Cousin MA, Privat de Garilhe M (1973) Misreading, a fundamental aspect ofthe mechanism of action of several aminoglycosides. Biochemistry 12(22):4528–4533.

38. Demirci H, et al. (2013) A structural basis for streptomycin-induced misreading of thegenetic code. Nat Commun 4:1355.

39. Gromadski KB, Rodnina MV (2004) Streptomycin interferes with conformationalcoupling between codon recognition and GTPase activation on the ribosome. NatStruct Mol Biol 11(4):316–322.

40. Durfee T, et al. (2008) The complete genome sequence of Escherichia coli DH10B:Insights into the biology of a laboratory workhorse. J Bacteriol 190(7):2597–2606.

E2584 | www.pnas.org/cgi/doi/10.1073/pnas.1401853111 Haft et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

020

41. Momose H, Gorini L (1971) Genetic analysis of streptomycin dependence in Escher-ichia coli. Genetics 67(1):19–38.

42. Agarwal D, Gregory ST, O’Connor M (2011) Error-prone and error-restrictive muta-tions affecting ribosomal protein S12. J Mol Biol 410(1):1–9.

43. Zaher HS, Green R (2009) Quality control by the ribosome following peptide bondformation. Nature 457(7226):161–166.

44. 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.

45. Peters JM, et al. (2009) Rho directs widespread termination of intragenic and stableRNA transcription. Proc Natl Acad Sci USA 106(36):15406–15411.

46. Bremer H, Dennis PP (1996) Modulation of chemical composition and other param-eters of the cell by growth rate. Escherichia coli and Salmonella: Cellular and Mo-lecular Biology, ed Neidhardt FC (American Society for Microbiology, Washington,DC), pp 1553–1569.

47. Nakamori S, Kobayashi S, Nishimura T, Takagi H (1999) Mechanism of L-methionineoverproduction by Escherichia coli: The replacement of Ser-54 by Asn in the MetJprotein causes the derepression of L-methionine biosynthetic enzymes. Appl Micro-biol Biotechnol 52(2):179–185.

48. Usuda Y, Kurahashi O (2005) Effects of deregulation of methionine biosynthesis onmethionine excretion in Escherichia coli. Appl Environ Microbiol 71(6):3228–3234.

49. Flatley J, et al. (2005) Transcriptional responses of Escherichia coli to S-nitrosoglutathioneunder defined chemostat conditions reveal major changes in methionine biosynthesis.J Biol Chem 280(11):10065–10072.

50. Mordukhova EA, Lee HS, Pan JG (2008) Improved thermostability and acetic acid tol-erance of Escherichia coli via directed evolution of homoserine o-succinyltransferase.Appl Environ Microbiol 74(24):7660–7668.

51. Gur E, Biran D, Gazit E, Ron EZ (2002) In vivo aggregation of a single enzyme limitsgrowth of Escherichia coli at elevated temperatures. Mol Microbiol 46(5):1391–1397.

52. Record MT, Jr., Zhang W, Anderson CF (1998) Analysis of effects of salts and un-charged solutes on protein and nucleic acid equilibria and processes: A practical guideto recognizing and interpreting polyelectrolyte effects, Hofmeister effects, and os-motic effects of salts. Adv Protein Chem 51:281–353.

53. Chan CL, Landick R (1997) Effects of neutral salts on RNA chain elongation andpausing by Escherichia coli RNA polymerase. J Mol Biol 268(1):37–53.

54. Brandts JF, Hunt L (1967) The thermodynamics of protein denaturation. 3. The de-naturation of ribonuclease in water and in aqueous urea and aqueous ethanol mix-tures. J Am Chem Soc 89(19):4826–4838.

55. Bull HB, Breese K (1978) Interaction of alcohols with proteins. Biopolymers 17(9):2121–2131.

56. Lehmann MS, Mason SA, McIntyre GJ (1985) Study of ethanol-lysozyme interactionsusing neutron diffraction. Biochemistry 24(21):5862–5869.

57. Straus DB, Walter WA, Gross CA (1987) The heat shock response of E. coli is regulatedby changes in the concentration of σ 32. Nature 329(6137):348–351.

58. Giudice E, Gillet R (2013) The task force that rescues stalled ribosomes in bacteria.Trends Biochem Sci 38(8):403–411.

59. Jerinic O, Joseph S (2000) Conformational changes in the ribosome induced bytranslational miscoding agents. J Mol Biol 304(5):707–713.

60. Baldridge KC, Contreras LM (2014) Functional implications of ribosomal RNA meth-ylation in response to environmental stress. Crit Rev Biochem Mol Biol 49(1):69–89.

61. Urbonavicius J, Durand JM, Björk GR (2002) Three modifications in the D and T armsof tRNA influence translation in Escherichia coli and expression of virulence genes inShigella flexneri. J Bacteriol 184(19):5348–5357.

62. Gorini L, Rosset R, Zimmermann RA (1967) Phenotype masking and streptomycindependence. Science 157(3794):1314–1317.

63. Klein-Marcuschamer D, Santos CN, Yu H, Stephanopoulos G (2009) Mutagenesis ofthe bacterial RNA polymerase alpha subunit for improvement of complex pheno-types. Appl Environ Microbiol 75(9):2705–2711.

64. Kanjee U, Ogata K, Houry WA (2012) Direct binding targets of the stringent responsealarmone (p)ppGpp. Mol Microbiol 85(6):1029–1043.

65. Heath RJ, Jackowski S, Rock CO (1994) Guanosine tetraphosphate inhibition of fattyacid and phospholipid synthesis in Escherichia coli is relieved by overexpression ofglycerol-3-phosphate acyltransferase (plsB). J Biol Chem 269(42):26584–26590.

66. Merlie JP, Pizer LI (1973) Regulation of phospholipid synthesis in Escherichia coli byguanosine tetraphosphate. J Bacteriol 116(1):355–366.

67. Cronan JE, Jr., Weisberg LJ, Allen RG (1975) Regulation of membrane lipid synthesis inEscherichia coli: Accumulation of free fatty acids of abnormal length during inhibitionof phospholipid synthesis. J Biol Chem 250(15):5835–5840.

68. Dalebroux ZD, Swanson MS (2012) ppGpp: Magic beyond RNA polymerase. Nat RevMicrobiol 10(3):203–212.

69. Bandas EL, Zakharov IA (1980) Induction of rho- mutations in yeast Saccharomycescerevisiae by ethanol. Mutat Res 71(2):193–199.

70. Miller JH (1972) Experiments in Molecular Genetics (Cold Spring Harbor LaboratoryPress, Plainview, NY).

71. Baba T, et al. (2006) Construction of Escherichia coli K-12 in-frame, single-geneknockout mutants: The Keio collection. Mol Syst Biol 2:2006.0008.

72. Das A, Court D, Adhya S (1976) Isolation and characterization of conditional lethalmutants of Escherichia coli defective in transcription termination factor rho. Proc NatlAcad Sci USA 73(6):1959–1963.

73. Schwalbach MS, et al. (2012) Complex physiology and compound stress responsesduring fermentation of alkali-pretreated corn stover hydrolysate by an Escherichiacoli ethanologen. Appl Environ Microbiol 78(9):3442–3457.

74. Artsimovitch I, Landick R (2002) The transcriptional regulator RfaH stimulates RNAchain synthesis after recruitment to elongation complexes by the exposed non-template DNA strand. Cell 109(2):193–203.

75. Gribskov M, Burgess RR (1983) Overexpression and purification of the sigma subunitof Escherichia coli RNA polymerase. Gene 26(2-3):109–118.

76. Toulokhonov I, Zhang J, Palangat M, Landick R (2007) A central role of the RNApolymerase trigger loop in active-site rearrangement during transcriptional pausing.Mol Cell 27(3):406–419.

77. Neidhardt FC, Bloch PL, Smith DF (1974) Culture medium for enterobacteria.J Bacteriol 119(3):736–747.

78. Sasson V, Shachrai I, Bren A, Dekel E, Alon U (2012) Mode of regulation and the in-sulation of bacterial gene expression. Mol Cell 46(4):399–407.

79. Zaslaver A, et al. (2006) A comprehensive library of fluorescent transcriptional re-porters for Escherichia coli. Nat Methods 3(8):623–628.

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

81. Li H, Durbin R (2009) Fast and accurate short read alignment with Burrows-Wheelertransform. Bioinformatics 25(14):1754–1760.

82. O’Connor PB, Li GW, Weissman JS, Atkins JF, Baranov PV (2013) rRNA:mRNA pairingalters the length and the symmetry of mRNA-protected fragments in ribosome pro-filing experiments. Bioinformatics 29(12):1488–1491.

83. Blattner FR, et al. (1997) The complete genome sequence of Escherichia coli K-12.Science 277(5331):1453–1462.

Haft et al. PNAS | Published online June 9, 2014 | E2585

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Janu

ary

28, 2

020