LETTERdoi:10.1038/nature10014

Preserving the membrane barrier for smallmolecules during bacterial protein translocationEunyong Park1 & Tom A. Rapoport1

Many proteins are translocated through the SecY channel inbacteria and archaea and through the related Sec61 channel ineukaryotes1. The channel has an hourglass shape with a narrowconstriction approximately halfway across the membrane, formedby a pore ring of amino acids2. While the cytoplasmic cavity of thechannel is empty, the extracellular cavity is filled with a short helixcalled the plug2, which moves out of the way during protein trans-location3,4. The mechanism by which the channel transports largepolypeptides and yet prevents the passage of small molecules, suchas ions or metabolites, has been controversial2,5–8. Here, we haveaddressed this issue in intact Escherichia coli cells by testing thepermeation of small molecules through wild-type and mutant SecYchannels, which are either in the resting state or contain a definedtranslocating polypeptide chain. We show that in the resting state,the channel is sealed by both the pore ring and the plug domain.During translocation, the pore ring forms a ‘gasket-like’ sealaround the polypeptide chain, preventing the permeation of smallmolecules. The structural conservation of the channel in all organ-isms indicates that this may be a universal mechanism by which themembrane barrier is maintained during protein translocation.

Bacteria offer a unique opportunity to test the permeation of smallmolecules through the protein translocation channel because thechannel is located in the plasma membrane and is therefore accessiblein intact cells. To test the permeability of the resting SecY channel, wecompared E. coli wild-type SecY, which is expected to be sealed, with aplug-deletion mutant of SecY (SecY(DP)), which should be constitu-tively open (Supplementary Fig. 1). Although a new plug may formfrom neighbouring polypeptide segments in this mutant9, it probablyblocks the channel only transiently8. Wild-type and SecY(DP) channelswere expressed under an inducible promoter at about the same level asthe endogenous protein and expression of the SecY(DP) mutant causedonly a moderate growth defect (Supplementary Fig. 2).

First, we studied the permeation of a relatively large (525 Da),uncharged cysteine-modifying reagent, biotin-PEG2-maleimide (BM;Fig. 1a), which can cross the outer membrane through porins butcannot cross the inner membrane10 (Supplementary Fig. 1). WhenBM was added to wild-type E. coli cells, few proteins were biotinylated(Fig. 1b, lane 5). By contrast, in the SecY(DP) mutant, several proteinswere strongly modified, particularly a protein of about 30 kDa (Fig. 1b,lane 8). Most of the modified proteins were located in the cytosol, asdemonstrated by cell fractionation (Fig. 1c). The extent of modificationwas about the same after treatment with the transcription inhibitorrifampicin (Fig. 1b, lane 9), which clears all SecY channels of translo-cating polypeptides (see Fig. 2c). Thus, permeation of BM occursprimarily through resting SecY(DP) channels. We also found thatmany signal sequence suppressor (prl) SecY mutants allow the per-meation of BM, although to a lesser extent than the SecY(DP) mutant(Supplementary Fig. 3). Channel opening in the absence of a trans-location substrate explains why these mutants translocate proteinswith defective or missing signal sequences11–13.

Next, we used osmotic swelling and bursting of cells (Supplemen-tary Fig. 1) to test the permeability of the resting SecY channel to

xylitol, an uncharged sugar of 152 Da. Escherichia coli cells were con-verted to spheroplasts and diluted into an iso-osmotic solution ofxylitol. Spheroplasts containing wild-type SecY channels did not takeup xylitol and therefore the turbidity of the sample changed little overtime (Fig. 1d). In contrast, the SecY(DP) mutant allowed xylitol topermeate, particularly when the channel was cleared of translocatingchains by rifampicin (Fig. 1d). Finally, we used osmotic swelling andbursting to test the permeation of charged and small (35 Da) Cl2 ions.Spheroplasts were diluted into an iso-osmotic solution of KCl in thepresence of valinomycin, an ionophore that allows the K1 counter-ions to move directly through the lipid bilayer. The data showed thatwild-type SecY did not conduct Cl2 ions, in contrast to the SecY(DP)mutant (Fig. 1e). We conclude that the resting wild-type channel isimpermeable to the small molecules tested and that the plug domain of

1Howard Hughes Medical Institute and Department of Cell Biology, Harvard Medical School, 240 Longwood Avenue, Boston, Massachusetts 02115, USA.

Xylitol KCl + valinomycin

Time (s)

Rela

tive t

urb

idity

Rela

tive t

urb

idity

0

0.2

0.4

0.6

0.8

1

0 100 200 3000.2

0.4

0.6

0.8

1

0 25 50 75 100

Empty vector

WT

WT (Rif-treated)

SecY(ΔP)

SecY(ΔP) (Rif-treated)

◄ p30

a

b c

d e

T P M C T P M C

Anti-SecY(Membranes)

Anti-MBP(Periplasm)

Anti-TetR(Cytoplasm)

130

9572

55

43

34

26

17

◄ p30

WT SecY(ΔP)SecY

(kDa)

Str

ep

tavid

in–H

RP

Rif

BM

WT SecY(ΔP)SecY

Anti-SecY

Anti-TF

17

9572

554334

26

10

130

(kDa)

Str

ep

tavid

in–H

RP

– + + – + + – + +

– – + – – + – – +

Emptyvector

Time (s)

1 2 3 4 5 6 7 8 9

Biotin-PEG2-maleimide (BM)

Mr = 525.62

Figure 1 | Testing the permeability of the resting SecY channel. a, Structureof the modification reagent BM. b, Wild-type (WT) SecY or the SecY(DP)mutant were expressed under the inducible Tet promoter. Cells were incubatedwith BM and the proteins were separated by SDS–polyacrylamide gelelectrophoresis (SDS–PAGE), then blotted with streptavidin–HRP conjugate,SecY antibodies or trigger factor (TF) antibodies (loading control). Whereindicated, rifampicin (Rif) was added before BM. Endogenous SecY was taggedat its C terminus, abolishing recognition by SecY antibodies (SupplementaryFig. 2). p30, a prominent biotinylated protein. c, Experiment conducted as inb, but cells (T) were fractionated into periplasm (P), membranes (M) andcytosol (C) after incubation with BM. Fractionation was controlled byimmunoblotting for the indicated marker proteins. MBP, maltose-bindingprotein; TetR, tetracycline repressor. d, e, Spheroplasts were diluted into iso-osmotic solutions of xylitol (d) or KCl containing valinomycin (e) and thechange in turbidity was followed over time.

1 2 M A Y 2 0 1 1 | V O L 4 7 3 | N A T U R E | 2 3 9

Macmillan Publishers Limited. All rights reserved©2011

SecY is required for the seal. It should be noted that the SecY(DP)mutant did not allow passage of K1 (Supplementary Fig. 4), Na1 orSO4

22 ions (not shown). Thus, in agreement with previous results14,the open channel still provides a barrier to some molecules (seeSupplementary Discussion), which may explain the relatively minorgrowth defect of the SecY(DP) mutant (Supplementary Fig. 2).

To study the permeability of the active SecY channel, we developed amethod to occupy the channels in vivo with a defined co-translationaltranslocation intermediate. The model substrate (NC100) contains100 amino acids (Fig. 2a) including the signal sequence of DsbA,which targets it to the signal recognition particle (SRP)-dependentco-translational translocation pathway15. NC100 also contains asequence from an unrelated protein, a Myc tag for detection and theSecM-stalling sequence16–18. After synthesis of the SecM sequence, theribosome stalls on the mRNA with the nascent chain associated aspeptidyl-tRNA (Fig. 2a). This construct was synthesized from aninducible promoter in cells expressing the SecY channel from a con-stitutive promoter. The insertion of the nascent chain into the channelwas verified by in vivo disulphide crosslinking: addition of an oxidantto the cell culture led to efficient crosslinking of a single cysteine atposition 19 in NC100 to a single cysteine at position 68 of the plugdomain of SecY (SecY68C; Supplementary Fig. 5). Channel insertionwas dependent on the hydrophobicity of the signal sequence and wasstrongly reduced in the absence of SRP (Supplementary Fig. 6). Whenthe expression of SecY68C was diminished by changing the start codon,about 70% of SecY68C was crosslinked to NC100, as judged by the

reduction of the non-crosslinked SecY band upon addition of theoxidant (Fig. 2b, lane 2 versus lane 1; quantification was confirmedby loading different amounts, see lanes 7–9). When NC100 expressionwas not induced, a lower percentage of SecY68C was crosslinked toendogenous proteins (Fig. 2b, lane 4 versus lane 2; white arrowheads);these crosslinks disappeared over time in the presence of rifampicin(Fig. 2b, lanes 6 and 11). These results indicate that most of the SecYmolecules can be occupied by NC100. Given the almost 1:1 molar ratioof nascent chain to SecY, a single SecY copy may be sufficient forco-translational translocation of a nascent chain, as was previouslyproposed19,20.

We used the new method to ask whether the open SecY channel,represented by the SecY(DP) mutant (see Fig. 1), can be blocked for smallmolecules by NC100. The SecY(DP) mutant expressed from a constitu-tive promoter was leaky for the modification reagent BM but inductionof NC100 abolished permeation (Fig. 2c, lane 8 versus lane 7). Inhibitionof transcription by rifampicin released NC100 from ribosomes andrestored leakiness for BM (lane 9). NC100 with a defective signalsequence containing two arginines in the hydrophobic core (RR) didnot block BM permeation (lane 10). With the wild-type SecY channel,no leakage was observed regardless of whether or not NC100 wasexpressed (lanes 1–5). Thus, the open pore of the SecY(DP) mutant issealed upon binding of the ribosome–nascent-chain complex to theSecY channel.

To test whether the seal is provided by channel insertion of thenascent polypeptide, we expressed chains of different lengths. Nascent

NC100

CuPh3

Rif

72

55

43

34

95

SecY

% loading

9572

55

43

34

26

17

130

(kDa)

Signal seq. WT WTWT WT RR WTWTWT WTRR

– + + + + – ++ + +– – – + – – –– + –

– – + + + – +– + +NC100

RifBM

Anti-Myc

Anti-SecY

Anti-TF

WT SecY68C SecY(ΔP)

Str

ep

tavid

in–H

RP

1 2 3 4 5 6 7 8 9 10

◄ p30

Anti-Myc

Anti-SecY

+ CuPh3 +CuPh3 β-MENC length

(amino acids)

9572

5543

3426

17

(kDa)

9572

55

43

34SecY

NC

(kDa)

a

b d

c

Nascent chain length (amino acids)

Rela

tive p

30

bio

tin

yla

tio

n(1

00

% =

no

NC

in

du

ctio

n)

SecY(ΔP)/+Rif(1.06 ± 0.04)

WT SecY/+Rif(0.05 ± 0.006)

Nascent chain length (amino acids)

Fra

ctio

nal ra

te o

ftu

rbid

ity d

ecre

ase (m

in–1)

SecY(ΔP)/+Rif(0.23 ± 0.02)

WT SecY/+Rif(0.002 ± 0.001)

Fra

ctio

nal ra

te o

ftu

rbid

ity d

ecre

ase (m

in–1)

0

20

40

60

80

100

60 70 80 90 100

WT-NC

RR-NC

0

0.05

0.1

0.15

0.2

0.25

60 70 80 90 100

WT-NC

RR-NC

0

0.2

0.4

0.6

0.8

1

1.2

60 70 80 90 100

WT-NC

RR-NC

e

f

g

Lane 1 2 3 4 5 6 7 8 9 10 11

100 62 100 82 100 50 26 100 97% SecY 100 32

NC100DsbA signal sequence

hydrophobic core Myc tagSecM stalling

sequence

4 15 73 85 101Cys 19

tRNA

mRNA

Ribosome

SecY

Cytoplasm

Periplasm

60

39 Cys 68

PlugCys 19

NH2

70 80 90 100

110

120

70 80 90 100

110

120

– + – + – + – – – – +

– – – – 15 m15 m 1 h 1 h– – –

+ + – – – – + + + – –

SecY::NC-RNA

Figure 2 | Testing the permeability of a translocating SecY channel.a, Schematics of the model nascent chain NC100 (upper panel) and its insertioninto the SecY channel (lower panel). The cysteines indicated in NC100 andSecY form a disulphide bridge. Residues 39–60 are inside the channel. b, NC100with Cys 19 was expressed under an arabinose-inducible promoter togetherwith SecY68C under the endogenous promoter. The start codon of SecY68C waschanged from AUG to GUG. Where indicated, cells were treated with theoxidant copper(II)(1, 10-phenanthroline)3 (CuPh3). Rifampicin was addedbefore oxidation for 15 min or 1 h, as indicated. The samples were analysed bySDS–PAGE, followed by blotting with antibodies against SecY::NC-RNA, SecYcrosslinked to tRNA-associated NC100. White arrowheads indicate SecY68C

crosslinked to endogenous proteins. % SecY, percentage of non-crosslinkedSecY. c, NC100 with either a wild-type (WT) or defective (RR) signal sequencewas expressed together with SecY68C or the SecY(DP) mutant. Where indicated,cells were pre-treated with rifampicin before addition of BM. The samples wereanalysed by SDS–PAGE, followed by blotting with streptavidin–HRP conjugate

or with antibodies against Myc (detects NC100), SecY or TF. d, Nascent chains(NC) of different lengths, all with Cys 19, were expressed together with SecY68C

and crosslinked with CuPh3. The samples were analysed by SDS–PAGE with orwithout prior reduction with b-mercaptoethanol (b-ME), followed byimmunoblotting. Red arrowhead, crosslinked SecY and NC-tRNA. e, Nascentchains of different lengths with either wild-type (WT-NC) or defective (RR-NC) signal sequences were expressed together with the DP channel. BM wasadded to the cells and biotinylated proteins were detected by SDS–PAGEfollowed by blotting. The modification of p30 was quantified and normalizedwith respect to the value obtained without nascent chain expression (error bars,s.d.; n 5 3). f, As in e, but spheroplasts were diluted into an iso-osmotic solutionof xylitol and the initial, linear rate of turbidity decrease was determined (errorbars, s.d.; n 5 3). Spheroplasts were also analysed after treatment withrifampicin. The SecY(DP)/1Rif sample showed an initial lag phase, which wasignored. g, As in f, but with dilution into iso-osmotic KCl containingvalinomycin.

RESEARCH LETTER

2 4 0 | N A T U R E | V O L 4 7 3 | 1 2 M A Y 2 0 1 1

Macmillan Publishers Limited. All rights reserved©2011

chains of 90 amino acids or longer were inserted into the SecY channel,as demonstrated by disulphide crosslinking (Fig. 2d). These chainsalmost completely blocked the permeation of BM through theSecY(DP) channel (Fig. 2e and Supplementary Fig. 7). Some reductionof BM permeability was also observed with chains of 60 to 80 residues,which were not inserted into the SecY channel. These chains are prob-ably still targeted to the channel, as signal sequence mutants (RR) didnot prevent BM permeation. Thus, the formation of a ribosome–channel junction may provide a barrier for BM, but complete blockageof permeation requires channel insertion of the nascent chain. Similarresults were obtained when the permeation of xylitol was analysed byosmotic swelling and bursting of spheroplasts (Fig. 2f). The permea-tion of Cl2 ions through the SecY(DP) channel was inhibited onlymoderately by short nascent chains (Fig. 2g), but was reduced by about90% by channel-inserted nascent chains, although not quite to the levelseen with the wild-type channel. Chains with a defective signalsequence did not block Cl2 permeation significantly. Consistent withour observation that short chains do not block the permeation of thesmallest molecules through the SecY(DP) channel efficiently, thesecells grew significantly more slowly than those expressing channel-inserted chains (Supplementary Fig. 8). Wild-type SecY did not causechain-length-dependent growth behaviour (Supplementary Fig. 8) anddid not allow significant Cl2 uptake into spheroplasts regardless ofnascent chain length, even when potentially counteracting pump activitywas abolished by energy depletion (Supplementary Fig. 9). Collectively,these results show that the nascent chain itself provides an effective seal.Consistent with this notion, expression of the post-translational, SecA-dependent substrate proOmpA in the SecY(DP) mutant significantlyreduced BM permeation, whereas a signal sequence mutant did not(Supplementary Fig. 10). The level of channel blockage was lower thanthat seen with a stalled ribosome–nascent-chain complex, probablybecause proOmpA occupies the SecY channel only transiently. It shouldbe noted that some Cl2 permeation was observed in vitro with the wild-type SecY channel engaged in SecA-mediated translocation14,21 (seeSupplementary Discussion).

Next, we tested whether the specific sequence of the nascent chaininside the channel affects the permeability for small molecules. Wedetermined that residues ,39 to 60 of NC100 are inside the centralpore (Fig. 2a), on the basis that approximately the last 40 residues areinside the ribosome and residues 19–34 are close to the plug domain onthe periplasmic side of SecY (Supplementary Fig. 11). We then variedthe sequence of the nascent chain inside the SecY channel, making itmore hydrophilic or hydrophobic than the original NC100 sequenceor replacing parts with stretches of glycines (Supplementary Fig. 12).All variants completely blocked BM and xylitol permeation throughthe SecY(DP) channel as effectively as the original NC100 chain (Sup-plementary Fig. 12 and data not shown) and the more hydrophobicchains were more potent in blocking Cl2 permeation (SupplementaryFig. 12). With the wild-type channel, little or no permeation wasobserved, regardless of which nascent chain was expressed. Thus,many different sequences of a translocating polypeptide can blockthe permeation of small molecules through the pore.

Next, we tested the role of the pore ring of SecY in sealing the restingchannel. Replacement of just one of the six isoleucines in the pore by aglycine caused only a little BM permeation but the channel becameprogressively more leaky with increasing numbers of glycine substitu-tions (Fig. 3a). Alanine substitutions had less severe effects. BM per-meation occurs through the resting channel because addition ofrifampicin had no effect (Supplementary Fig. 13a). The pore-mutantchannels were also permeable to xylitol (see below) and Cl2 ions (Sup-plementary Fig. 13b). In addition, the most severe pore mutants causeda strong growth defect (Supplementary Fig. 13c): the cells died imme-diately after induction of these SecY mutants (Supplementary Fig. 2).Cell death is probably caused by dissipation of the membrane potentialbecause flow cytometry using the voltage-sensitive dye DiBAC4(3)showed that the membrane potential was decreased to about the same

extent as that seen at a 20mM concentration of the ionophore carbonylcyanide m-chlorophenyl hydrazone (CCCP) (Supplementary Fig. 14).Taken together, these results show that the pore ring has an importantrole in maintaining the seal of the resting SecY channel and that a leakychannel is inconsistent with cell viability.

Finally, we investigated the role of the pore ring in sealing the activechannel. Expression of NC100 blocked the permeation of BM througheven the most severe SecY pore mutants (Fig. 3b, lanes 4 versus 3 and10 versus 9) but expressing a signal sequence mutant (RR) of NC100did not block permeation (Fig. 3b, lanes 6 and 12). By contrast, thepermeation of the smaller molecules xylitol (Fig. 3c), Cl2 ions (Fig. 3d)or K1 ions (Supplementary Fig. 15) was not prevented by expression ofNC100 in a pore mutant containing three glycines. With a mutantcontaining only one glycine in the pore ring, NC100 expressionblocked xylitol (Fig. 3e) but not Cl2 ions (Fig. 3f). The high permeabilityfor Cl2 ions was maintained regardless of the sequence inside the SecYchannel (Supplementary Fig. 16). Thus, a translocating channel requiresisoleucines all around the pore ring to prevent small Cl2 ions frompassing, whereas pore defects are tolerated for larger molecules (Sup-plementary Fig. 1).

Our results show that the wild-type SecY channel is sealed for eventhe smallest molecules, both in its resting state and when translocatinga polypeptide. A simple model explains how this is achieved (Fig. 4). In

SecY(WT)/ + Rif

SecY(IIIG)/ + WT-NC100

SecY(IIIG)/ + Rif

SecY(IIIG)/ + RR-NC100

0.2

0.4

0.6

0.8

1

0 1 2 3 4 5 6 0 1 2 3

Rela

tive t

urb

idity

Time (min) Time (min)

0

0.2

0.4

0.6

0.8

1

0 50 100 150

–NC / –SecY(GGGI)

–NC / + SecY(GGGI)

+ WT-NC100 / + SecY(GGGI)

+ RR-NC100 / + SecY(GGGI)

Time (s) Time (s)

Rela

tive t

urb

idity

0.4

0.6

0.8

1

0 25 50 75 100

◄ p30

95725543

34

26

17

130

10(kDa)

Str

ep

tavid

in–H

RP

◄ p30

– + + + + + + + + + + + +BM

Anti-SecY

Anti-TF

9572

55

43

34

26

130

Str

ep

tavid

in–H

RP

NC100

SecY

Rif

–

–

–

+

–

–

–

+

–

+

+

–

+

+

+

+

+

–

Signal seq. RR

–

–

–

+

–

–

–

+

–

+

+

–

+

+

+

+

+

–

SecY(GGGI) SecY(GGGG)

RRWTWTWTWTWT WTWTWTWTWT

Anti-Myc

Anti-SecY

Anti-TF

1 2 3 4 5 6 7 8 9 10 11 12

17(kDa)

a

b

e f

c d

Vect

or o

nly

Vect

or o

nly

WT

(IIII)

GIII

IIIG

GIIG

IGG

IIG

GG

GG

GI

GG

GG

AAAI

AAAA

SecY(Δ

P)

Figure 3 | Permeability in pore ring mutants. a, Wild-type SecY or poremutants in which Ile 86, Ile 191, Ile 278 or Ile 408 (I in all positions) werereplaced by Gly or Ala (G or A in the relevant positions) were expressed underthe Tet promoter. Cells containing the SecY(DP) mutant or an empty vectorwere also analysed. After treatment of cells with BM, the samples were analysedby SDS–PAGE and blotting with streptavidin–HRP conjugate, or withantibodies against SecY or TF. b, NC100 with WT or RR signal sequence wasexpressed from the inducible arabinose promoter together with SecY poremutants under the Tet promoter. Where indicated, rifampicin was addedbefore BM. The samples were analysed by SDS–PAGE, followed by blottingwith streptavidin–HRP conjugate or with antibodies against Myc, SecY or TF.Addition of rifampicin does not clear the mutant channels of nascent chains(lanes 5 and 11), probably because peptidyltransferase activity is compromisedin these dying cells. c, d, As in b, but spheroplasts containing the GGGI poremutant were diluted into iso-osmotic xylitol (c) or iso-osmotic KCl containingvalinomycin (d), then the turbidity change was followed over time. e, f, As inc, d, but with the IIIG pore mutant under the constitutive promoter with a GUGstart codon. The cells were treated with rifampicin or induced for NC100expression and then diluted into iso-osmotic xylitol (e) or KCl containingvalinomycin (f).

LETTER RESEARCH

1 2 M A Y 2 0 1 1 | V O L 4 7 3 | N A T U R E | 2 4 1

Macmillan Publishers Limited. All rights reserved©2011

the resting state, the plug is located in the centre of SecY, interactingwith the pore ring residues and sealing the channel (Fig. 4a). Duringtranslocation, the plug is displaced and the pore ring forms a gasket-like seal around the translocating polypeptide chain to prevent the freeflow of ions (Fig. 4b). The translocating chain itself serves as the majorobstacle for small molecules; without it, the open pore allows manysmall molecules to pass. Whenever the polypeptide leaves the channel,either towards the extracellular side after termination of translocation(Fig. 4c), or sideways into lipid after the arrival of a hydrophobictransmembrane sequence (Fig. 4d), the plug returns and re-seals thechannel. This mechanism applies to both co-translational and post-translational translocation. However, in co-translational translocation,the junction between the ribosome and the channel seems to providean additional barrier for larger molecules, preventing metabolites andother cytosolic molecules from reaching the channel. Given thesequence conservation of the SecY and Sec61 channels, these principlesmay be universal. However, in prokaryotes, a tight seal is essential forcell viability, whereas in eukaryotes, the intracellular endoplasmicreticulum membrane may tolerate some leakiness22–24, which couldexplain why Sec61 pore mutants in Saccharomyces cerevisiae have onlyminor growth defects25.

METHODS SUMMARYAll strains and plasmids used in this study are listed and described inSupplementary Tables 1 and 2. The subunits of the heterotrimeric SecY complexwere expressed from either the Tet promoter in the pTet vector (Figs 1 and 3a–d),or the endogenous rplN promoter on either the pACYC-SecYEG vector (Figs 2c,e–g and 3e, f) or the pRSY vector (Fig. 2d). The latter vector also contained the SRPcomponents Ffh, 4.5S RNA and FtsY under their own promoters. NC100 and itsvariants were expressed from the arabinose promoter on the separate pBAD-NC100 plasmid, except in Fig. 2b, where both NC100 (from the arabinose pro-moter) and SecY (from the rplN promoter) were expressed from the same plasmidand the SRP components were expressed from another. Except in Fig. 2d, an E. colistrain was used in which chromosomal SecY is tagged at the carboxy terminus witha calmodulin-binding peptide (CBP). To measure permeability to BM, cell cultureswere incubated with 0.4 mM BM, then quenched with 20 mM b-mercaptoethanoland lysed in SDS sample buffer. For cell fractionation, cells were converted intospheroplasts by treatment with EDTA and lysozyme in the presence of 18% sucrose.The spheroplasts were lysed by sonication and the membranes were separatedfrom the cytosol by ultra-centrifugation. To measure permeability for other smallmolecules, spheroplasts were diluted 20-fold with iso-osmotic solutions and theabsorbance was followed at 500 nm in a spectrophotometer. To disulphide-crosslink nascent chains with SecY, 0.25 mM copper(II)(1,10-phenanthroline)3

(CuPh3) was added to the culture directly, followed by quenching with 20 mMN-ethylmaleimide and lysis in SDS sample buffer.

Full Methods and any associated references are available in the online version ofthe paper at www.nature.com/nature.

Received 24 January; accepted 22 March 2011.

1. Rapoport, T. A. Protein translocation across the eukaryotic endoplasmic reticulumand bacterial plasma membranes. Nature 450, 663–669 (2007).

2. van den Berg, B. et al. X-ray structure of a protein-conducting channel. Nature 427,36–44 (2004).

3. Harris, C. R. & Silhavy, T. J. Mapping an interface of SecY (PrlA) and SecE (PrlG) byusing synthetic phenotypes and in vivo cross-linking. J. Bacteriol. 181, 3438–3444(1999).

4. Tam, P. C., Maillard, A. P., Chan, K. K. & Duong, F. Investigating the SecY plugmovement at the SecYEG translocation channel. EMBO J. 24, 3380–3388 (2005).

5. Hamman, B. D., Hendershot, L. M. & Johnson, A. E. BiP maintains the permeabilitybarrier of the ER membrane by sealing the lumenal end of the translocon porebefore and early in translocation. Cell 92, 747–758 (1998).

6. Liao, S., Lin, J., Do, H. & Johnson, A. E. Both lumenal and cytosolic gating of theaqueous ER translocon pore are regulated from inside the ribosome duringmembrane protein integration. Cell 90, 31–41 (1997).

7. Simon, S. M. & Blobel, G. A protein-conducting channel in the endoplasmicreticulum. Cell 65, 371–380 (1991).

8. Saparov,S.M.et al.Determining theconductanceof theSecYprotein translocationchannel for small molecules. Mol. Cell 26, 501–509 (2007).

9. Li, W. et al. The plug domain of the SecY protein stabilizes the closed state of thetranslocation channel and maintains a membrane seal. Mol. Cell 26, 511–521(2007).

10. Zhang, W., Bogdanov, M., Pi, J., Pittard, A. J. & Dowhan, W. Reversible topologicalorganization within a polytopic membrane protein is governed by a change inmembrane phospholipid composition. J. Biol. Chem. 278, 50128–50135 (2003).

11. Bieker, K. L., Phillips, G. J. & Silhavy, T. J. The sec and prl genes of Escherichia coli. J.Bioenerg. Biomembr. 22, 291–310 (1990).

12. Derman, A. I., Puziss, J. W., Bassford, P. J. & Beckwith, J. A signal sequence is notrequired for protein export in prlA mutants of Escherichia coli. EMBO J. 12,879–888 (1993).

13. Smith, M. A., Clemons, W. M. Jr, DeMars, C. J. & Flower, A. M. Modeling the effects ofprl mutations on the Escherichia coli SecY complex. J. Bacteriol. 187, 6454–6465(2005).

14. Dalal, K. & Duong, F. The SecY complex forms a channel capable of ionicdiscrimination. EMBO Rep. 10, 762–768 (2009).

15. Schierle, C. F. et al. The DsbA signal sequence directs efficient, cotranslationalexport of passenger proteins to the Escherichia coli periplasm via the signalrecognition particle pathway. J. Bacteriol. 185, 5706–5713 (2003).

16. Nakatogawa, H. & Ito, K. The ribosomal exit tunnel functions as a discriminatinggate. Cell 108, 629–636 (2002).

17. Nakatogawa, H.& Ito, K.Secretionmonitor, SecM,undergoes self-translation arrestin the cytosol. Mol. Cell 7, 185–192 (2001).

18. Woolhead, C. A., Johnson, A. E. & Bernstein, H. D. Translation arrest requires two-way communication between a nascent polypeptide and the ribosome. Mol. Cell22, 587–598 (2006).

19. Menetret, J. F. et al. Single copies of Sec61 and TRAP associate with anontranslating mammalian ribosome. Structure 16, 1126–1137 (2008).

20. Becker, T. et al. Structure of monomeric yeast and mammalian Sec61 complexesinteracting with the translating ribosome. Science 326, 1369–1373 (2009).

21. Schiebel, E. & Wickner, W. Preprotein translocation creates a halide anionpermeability in the Escherichia coli plasma membrane. J. Biol. Chem. 267,7505–7510 (1992).

22. Le Gall, S., Neuhof, A. & Rapoport, T. A. The endoplasmic reticulum membrane ispermeable to small molecules. Mol. Biol. Cell 15, 447–455 (2004).

23. Heritage, D. & Wonderlin, W. F. Translocon pores in the endoplasmic reticulum arepermeable to a neutral, polar molecule. J. Biol. Chem. 276, 22655–22662 (2001).

24. Roy, A. & Wonderlin, W. F. The permeability of the endoplasmic reticulum isdynamically coupled to protein synthesis. J. Biol. Chem. 278, 4397–4403 (2003).

25. Junne, T., Kocik, L. & Spiess, M. The hydrophobic core of the Sec61 translocondefines the hydrophobicity threshold for membrane integration. Mol. Biol. Cell 21,1662–1670 (2010).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We thank P. Walter, H. Bernstein and G. Phillips for materials,D. Boyd for advice, C. Akey for discussions and C. Akey, A. Osborne and A. Salic forcritical reading of the manuscript. The work was supported by a grant from the NIH(GM052586). T.A.R. is a Howard Hughes Medical Institute investigator.

Author Contributions E.P. performed the experiments and E.P. and T.A.R. wrote themanuscript.

Author Information Reprints and permissions information is available atwww.nature.com/reprints. The authors declare no competing financial interests.Readers are welcome to comment on the online version of this article atwww.nature.com/nature. Correspondence and requests for materials should beaddressed to T.A.R. (tom_rapoport@hms.harvard.edu).

SecY

Translocating

polypeptide

Plug

Pore ring residues

a b c

d

Figure 4 | Model for the maintenance of the membrane barrier by the SecYchannel. a, SecY in the resting state: the plug is located in the centre, interactingwith pore ring residues and sealing the channel. b, The plug is displaced duringtranslocation; the pore ring forms a gasket-like seal around the translocatingpolypeptide chain to prevent the free flow of ions. c, d, When the polypeptideleaves towards the extracellular side (c) or sideways into lipid after the arrival ofa hydrophobic transmembrane sequence (d), the plug returns to re-seal thechannel.

RESEARCH LETTER

2 4 2 | N A T U R E | V O L 4 7 3 | 1 2 M A Y 2 0 1 1

Macmillan Publishers Limited. All rights reserved©2011

METHODSBacterial strains and plasmids. All strains and plasmids used in this study arelisted and described in Supplementary Tables 1 and 2. PCR reactions were per-formed with Phusion polymerase (New England Biolabs) or KOD polymerase(Novagen). Site-directed mutagenesis was performed using either the Quikchange(Stratagene) or Phusion (New England Biolabs) mutagenesis kits. All constructswere verified by sequencing. E. coli DH5a strain was used for all cloning procedures.

For all experiments except cloning, we used E. coli strains lacking both Rmf andOmpT because Rmf is known to induce dimerization and inactivation of ribo-somes at stationary phase26 and the OmpT protease has been reported to proteo-lyse SecY27. The Drmf DompT strain was constructed from individual deletionstrains by P1 transduction28. The Drmf DompT secY-CBP strain, in which thechromosomal secY gene is tagged at its C terminus with calmodulin-bindingpeptide (CBP), was constructed in the following way. We synthesized a ‘CBP-RBS-Zeo’ DNA cassette containing a CBP tag, a stop codon, a ribosome bindingsite and a zeocin resistance gene in that order. This cassette was amplified by PCRand electroporated into Drmf DompT cells expressing Lambda red recombinasefrom the pKD46 plasmid29. The resulting cells were selected on zeocin-containingagar plates (Invivogen). Incorporation of the cassette into the chromosome wasverified by PCR, DNA sequencing and immunoblotting using CBP antibodies(Genscript). The Drmf DompT secY-CBP strain was used throughout, except forexperiments in Fig. 2d, where the Drmf DompT strain was used. For all osmoticswelling and bursting experiments, we used a strain containing an additionaldeletion of the glpF gene because GlpF has been shown to allow permeation ofpolyalcohol sugars, including xylitol30. The Drmf DompT secY-CBP DglpF strainwas generated by P1 transduction (see Supplementary Table 1) and did not showany xylitol permeability in osmotic swelling and bursting experiments.

The pTet-SecYEG plasmid, which expresses SecYEG under a tetracycline-inducible promoter, was made by PCR amplification of the SecE-SecY-SecGencoding sequence from the pBAD22-SecYEG plasmid9 and subsequent insertionof this sequence into the pTet vector. The pACYC-SecYEG plasmid, which con-stitutively expresses SecYEG, was constructed by inserting the same SecYEG cod-ing sequence into pACYC184 after fusing it with a 200-base-pair DNA fragmentcontaining the E. coli rplN promoter at its 59 end. The plasmid pBAD-NC100 forexpression of a SecM-stalled nascent chain was generated as follows: we con-structed the pBAD Myc-SecM vector by inserting a DNA fragment encoding aMyc epitope between the PstI and SacI sites of pBAD His/C (Invitrogen) andthen inserting a fragment encoding the 17-residue SecM stalling sequence(FSTPVWISQAQGIRAGP) between the PstI and EcoRI sites. Then we synthe-sized a DNA segment by PCR encoding the DsbA signal peptide fused to asequence from an unrelated protein and inserted it between the NcoI and PstIsites of the pBAD Myc-SecM vector. Other plasmids for different nascent chainlengths and sequences were made by modifying the pBAD-NC100 plasmid. Moreinformation is available in Supplementary Table 2.Cell culture and protein expression. Unless otherwise indicated, E. coli cells weregrown and induced as follows. Cells harbouring the indicated plasmids werepicked from freshly transformed colonies and inoculated into LB medium supple-mented with appropriate antibiotics (100mg ml21 for ampicillin and 50mg ml21

for chloramphenicol). Cultures were grown at 37 uC to log phase (optical density at600 nm (OD600) of about 0.4–0.6) before induction. To induce SecYEG from thepTet vector (Figs 1 and 3a), 200 ng ml21 anhydrotetracyline (aTet) was added for30 min. To overexpress nascent chains, 0.2% arabinose was added for 1 h except inFigs 2b and 2d, where the induction was for 2 h. When both SecYEG (from pTet)and a nascent chain were co-expressed (Fig. 3b–d), 0.07% arabinose was added toE. coli cultures at log phase, then 200 ng ml21 aTet was added after 15 min andinduction was continued for an additional 40 min. Where indicated, 100mg ml21

rifampicin was added to cultures for 1 h at 37 uC.SecY was expressed from the endogenous rplN promoter on either the pACYC-

SecYEG vector (Figs 2c, e–g and 3e, f) or the pRSY vector (Fig. 2d). The latter alsoencodes rare-codon tRNAs and the SRP pathway components (Ffh, 4.5S RNA andFtsY) from their endogenous promoters. For the experiment in Fig. 2b, a fusedvector (pBAD-NC100/SecYEG) was used, encoding both NC100 and SecY,together with pR2HQ4, a vector overexpressing the SRP components and rare-codon tRNAs. Details about these plasmids and their construction are given inSupplementary Table 2. When a nascent chain was intended to saturate the SecYchannels, vectors contained a GUG start codon for SecY instead of AUG (Figs 2b,c, e–g and 3e, f). Saturation of GUG-SecY with the NC100 chains was not affectedby plasmid combinations or by overexpression of the rare-codon tRNAs and theSRP components.In vivo biotinylation and subcellular fractionation. E. coli cells were grown tolog phase at 37 uC in LB medium and induced as described above. After an aliquotof the culture was taken, 0.4 mM biotin-PEG2-maleimide (BM; Pierce) was addedto the culture directly for 30 min at room temperature (23 uC). When a nascent

chain was overexpressed, the incubation was shortened to 15 min. After quenchingthe reaction with 20 mM b-mercaptoethanol and a further incubation of 15 minon ice, the cells were collected by brief centrifugation. For SDS–PAGE analysis, thesame numbers of cells were lysed in SDS sample buffer and loaded on a gel(0.1 OD600 per lane). For subcellular fractionation, the cells were re-suspendedin ice-cold spheroplasting buffer (100 mM Tris-HCl, pH 8.0, 18% sucrose).Conversion of cells into spheroplasts was carried out by addition of 2 mMEDTA and 0.1 mg ml21 hen egg lysozyme and incubation for 10 min on ice.The spheroplasts were sedimented by centrifugation (9,000 r.p.m., 10 min) andthe supernatant containing periplasmic proteins was removed. The spheroplastswere re-suspended in 100 mM Tris-HCl, pH 8.0, 150 mM NaCl and lysed bysonication. The lysate was subjected to centrifugation for 1 h at 51,000 r.p.m. ina TLA100.3 rotor (Beckman) to pellet the membranes. The supernatant (cytosolicproteins) and the membranes were analysed by SDS–PAGE.Osmotic swelling and bursting experiments. Cells were placed on ice for afew minutes and harvested by centrifugation at 5,000 r.p.m. for 7 min. Afterre-suspension in one-fifteenth culture volume of 20 mM Tris-HCl, pH 7.2 and18% sucrose (0.619 mol kg21), the cells were converted to spheroplasts by additionof 2 mM EDTA and 0.1 mg ml21 lysozyme at 4 uC. When nascent chains wereexpressed, 4 mM MgSO4 was added 3 min after initiation of spheroplasting toavoid potential adverse effects of EDTA on ribosome–nascent chain complexes.Efficient (.95%) spheroplasting was verified by phase contrast microscopy and byimmediate lysis on dilution into water (confirmed by measuring turbidity). Todetermine permeability for various small molecules, the spheroplast suspensionwas mixed rapidly at room temperature with a 19-fold volume of an iso-osmoticsolution of xylitol (0.616 mol kg21) or KCl (0.342 mol kg21). Absorbance at500 nm was recorded in a spectrophotometer. The time between mixing andmeasurement was about 5 s. To obtain more accurate iso-osmotic solutions,osmotic coefficients were taken into account (1.05 for sucrose, 0.90 for KCl and1.00 for xylitol). For measurement of chloride permeability, 10mM valinomycinwas included in the KCl solution. When the rate of turbidity decrease was deter-mined, initial slopes were calculated by linear regression. These numbers werenormalized with respect to the initial turbidity.In vivo expression and crosslinking of nascent chains. To disulphide-crosslink anascent chain with SecY, 0.25 mM CuPh3 was added directly to the inducedculture. After gentle rocking at room temperature for 10–20 min, the cells werecollected by centrifugation and re-suspended in ice-cold TMP100 buffer (50 mMTris-acetate, pH 7.2, 25 mM Mg(OAc)2, 0.1 M KOAc) containing 20 mMN-ethylmaleimide. The cells were lysed by sonication and the samples analysedby non-reducing SDS–PAGE and immunoblotting. Where indicated, the sampleswere treated with 2% b-mercaptoethanol or 0.2 mg ml21 RNase A before loadingthem onto the gel.Testing for SRP-dependence of nascent chain insertion. WAM121 E. coli cells(Supplementary Table 1) were transformed with pRARE/SecYEG and pTac-NC100 (Supplementary Table 2). Cells from a saturated culture, grown in LBmedium supplemented with 0.1% arabinose, were washed three times with RMmedium (1 3 M9 salt, 1 mM MgSO4, 2% casamino acids and 0.2% glucose) sup-plemented with ampicillin and chloramphenicol and then inoculated into freshRM medium for 3.5 h at 37 uC. Incubation was then continued in either theabsence or presence of 0.1% arabinose. After an additional 1 h, overexpressionof the nascent chain was induced for 2 h by addition of 1 mM isopropyl thioga-lactopyranoside (IPTG). In vivo crosslinking experiments were performed asabove.Mal-PEG modification of nascent chains. pBAD-NC100 plasmids encodingsingle cysteines at various positions were transformed into Drmf DompT cellsharbouring the pRARE plasmid. After induction of nascent chain expression,the cells were harvested and lysed by sonication in TMP100 buffer. Mal-PEG((methyl-PEG12)3-PEG4-maleimide, Pierce) was prepared as a 125 mM solutionin dimethyl sulphoxide (DMSO) and added to the cell homogenate at 2 mM finalconcentration. After incubation for 1 h at room temperature, the reaction wasstopped by addition of 33 mM N-ethylmaleimide. The samples were subjectedto SDS–PAGE and analysed by immunoblotting with antibodies against Myc.Flow cytometry. To monitor the membrane potential of cells, SecY was inducedfor the indicated time period, then cells were diluted to ,5 3 106 cells ml21 inphosphate buffered saline containing 2mM bis-(1,3-dibutylbarbituric acid)trimethine oxonol (DIBAC4(3), Invitrogen). After a few minutes, the cell suspen-sion was injected into a flow cytometer (FACScalibre, BD) for analysis. The flowrate was set at ‘low’. Signals from cells were readily distinguishable from back-ground signals as a separate population in forward vs. side scattering (FSC versusSSC) plots. This population was gated to exclude background events from furtheranalysis. The DIBAC4(3) signal was measured in the green fluorescence channel(FL1-H). A total of 50,000 events were counted in each experiment.

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2011

SDS–PAGE, immunoblotting and densitometry analysis. SDS–PAGE was per-formed using Bis-Tris gels (Invitrogen) with MES-SDS running buffer. Images ofimmunoblots were taken with a CCD-based device (Fujifilm LAS-3000) and astandard ECL reagent. Image J software was used for densitometry analysis.

26. Wada, A., Yamazaki, Y., Fujita, N. & Ishihama, A. Structure and probable geneticlocation of a ‘‘ribosome modulation factor’’ associated with 100S ribosomes instationary-phase Escherichia coli cells. Proc. Natl Acad. Sci. USA 87, 2657–2661(1990).

27. Akiyama, Y. & Ito, K. SecY protein, a membrane-embedded secretion factor of E.coli, is cleavedby theOmpTprotease invitro. Biochem.Biophys.Res. Commun. 167,711–715 (1990).

28. Baba, T. et al. Construction of Escherichia coli K-12 in-frame, single-gene knockoutmutants: the Keio collection. Mol. Syst. Biol. 2, 2006.0008 (2006).

29. Datsenko, K. A. & Wanner, B. L. One-step inactivation of chromosomal genes inEscherichia coliK-12 using PCR products. Proc.Natl Acad. Sci.USA97, 6640–6645(2000).

30. Heller, K. B., Lin, E. C. & Wilson, T. H. Substrate specificity and transport propertiesof the glycerol facilitator of Escherichia coli. J. Bacteriol. 144, 274–278 (1980).

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2011