Mapping Putative Contact Sites Between Subunits in a Bacterial ATP-binding Cassette (ABC) Transporter by Synthetic Peptide Libraries

J. Mol. Biol. (2007) 369, 386–399

doi:10.1016/j.jmb.2007.03.043

Mapping Putative Contact Sites Between Subunits in a Bacterial ATP-binding Cassette (ABC) Transporter by Synthetic Peptide Libraries Bettina Blüschke 1 , Viola Eckey 1 , Britta Kunert 1 , Susanne Berendt 1 Heidi Landmesser 1 , Michael Portwich 2 , Rudolf Volkmer 2 and Erwin Schneider 1 ⁎ 1

Humboldt Universität zu Berlin, Institut für Biologie/Bakterienphysiologie, Chausseestr. 117, D-10115 Berlin, Germany 2

Institut für Medizinische Immunologie, CharitéUniversitätsmedizin Berlin, Schumannstr. 20-21, D-10098 Berlin, Germany

The maltose ATP-binding cassette transporter of Salmonella typhimurium is composed of the soluble periplasmic receptor, MalE, and a membraneassociated complex comprising one copy each of the pore-forming hydrophobic subunits, MalF and MalG, and of a homodimer of the ATPhydrolyzing subunit, MalK. During the transport process the subunits are thought to undergo conformational changes that might transiently alter molecular contacts between MalFG and MalK2. In order to map sites of subunit–subunit interactions we have used a comprehensive peptide mapping approach comprising large-scale microsynthesis of labelled probes and array techniques. In particular, we screened the binding of (i) MalFGderived soluble biotinylated peptides to immobilized MalK, and (ii) radiolabelled MalK to MalFG-derived cellulose membrane-bound peptides. The first approach identified seven peptides (10mers) each of MalF and MalG that specifically bound to MalK. The peptides were localized to TMDs 3 and 6, periplasmic loop P4 and cytoplasmic loops C2 and C3 of MalF, while MalG-derived peptides localized to the N terminus, TMDs 4–6, periplasmic loop P1 and cytoplasmic loop C2. Peptides from C3 and C2, respectively, of MalF and MalG partially encompass the conserved EAAmotif, known to be crucial for interaction with MalK. These results were basically confirmed by screening MalFG-derived peptide arrays consisting of 16mers or 31mers with radiolabelled MalK. This approach also allowed us to perform complete substitutional analyses of peptides in question. The results led to the construction of MalFG variants that were subsequently analyzed for functional consequences in vivo. Growth experiments revealed that most of the mutations had no phenotype, suggesting that the mutated residues themselves are not critical but part of a discontinuous binding site. However, two novel mutations affecting residues from the EAA motifs of MalF (Ile417Glu) and MalG (Phe203Gln/Asn), respectively, displayed severe growth defects, indicating their functional importance. Together, these experimental outcomes identify specific molecular contacts made between MalK and MalFG that extend beyond the well-characterized EAA motif. © 2007 Elsevier Ltd. All rights reserved.

*Corresponding author

Keywords: ATP-binding cassette; maltose transport; Salmonella typhimurium; protein–protein interactions; synthetic peptide arrays

Abbreviations used: ABC, ATP-binding cassette; CAPE, cellulose-amino-hydroxypropyl ether; DIPEA, diisopropylethylamine; Fmoc-, 9-fluorenyl methoxy carbonyl-; RP-HPLC, reverse phase high pressure liquid chromatography; MALDI-TOF, matrix-assisted laser desorption ionisation-time of flight; TBTU, O-(benzotriazol-1-yl)-N, N,N′,N′-tetramethyluronium-tetrafluoroborate. E-mail address of the corresponding author: [email protected] 0022-2836/$ - see front matter © 2007 Elsevier Ltd. All rights reserved.

Contact Sites in a Bacterial ABC Transporter

Introduction ATP-binding cassette (ABC) transporters exist in all living organisms and form one of the largest superfamilies. They are integral to almost every biological process and physiological system. ABC transport systems mediate the uptake or export of an enormous variety of substances across cell membranes, from small ions to large polypeptides, at the expense of ATP.1 They share a common architectural organization comprising two hydrophobic transmembrane domains (TMDs) that form the translocation pathway and two hydrophilic nucleotide binding (ABC) domains (NBDs) that hydrolyze ATP. In prokaryotes, these domains are mostly expressed as separate protein subunits, whereas in eukaryotes, especially in mammalian cells, they are usually fused into a single polypeptide chain. The maltose ABC transporter of Escherichia coli/ Salmonella typhimurium is one of the best characterized systems that can serve as a model for ABC importers.2,3 The transporter is composed of the extracellular (periplasmic) receptor, the maltose binding protein (MBP or MalE), and the membrane-bound complex comprising the hydrophobic subunits, MalF and MalG, and two copies of the ATPase (ABC) subunit, MalK. MalF has a molecular mass of 55 kDa and is predicted to span the membrane eight times. It is further characterized by a large periplasmic loop (20 kDa) of, thus far, unknown function, connecting putative TMDs 3 and 4. MalG is predicted to consist of six TMDs and has a molecular mass of 33 kDa (see also Figure 3). MalK and closely related ABC subunits contain a unique C-terminal extension which, in the crystal structure of the MalK dimer, contributes substantially to monomer–monomer contacts.4 Transport of maltose is assumed to be initiated by interaction of substrate-loaded MalE with periplasmic loops of MalFG, thereby triggering conformational changes that result in ATP hydrolysis at the MalK subunits and eventually in substrate translocation.5 Binding of liganded MalE and ATP occurs rather simultaneously and a transient stable complex of MalFGK2 with MalE is formed.6 ATP-dependent closure of the MalK dimer interface coincides with opening of MalE and maltose release to the transporter.7 ATP hydrolysis is required to reset the transporter to the resting state by opening the contacts between the MalK subunits.8 ATP-induced conformational changes of the isolated MalK dimer4,9 as well as of the subunits in the assembled complex10–12 have been well documented. However, little is known on the binding sites by which the subunits contact each other. It is well established that association of the MalK subunits with MalF and MalG requires at least the so-called EAA sequence motifs (consensus EAAx3Gx9Ix1LP; x, any amino acid) which are conserved in the last cytoplasmic loop regions of TMDs of ABC import systems10,13 (see also Figure 3). The role of the EAA motif in subunit–subunit interactions was confirmed by the crystal structure of the E. coli vitamin

387 B12 transporter (BtuCD), thus far the only ABC import system for which structural information is available.14 Nonetheless, it is unclear whether these interactions are sufficient to allow formation of a stable MalFGK2 complex that requires treatment with 6.6 M urea to release MalK.15 Moreover, contact sites might be altered transiently as a result of conformational changes during the transport cycle. Thus, here, we have employed biotinylated soluble and cellulose membrane-bound synthetic peptides derived from MalFG to map their interactions with the ATP-hydrolyzing MalK subunits. This approach has been widely used to map antibody epitopes and to identify protein interaction domains, to name just a few applications. 16 Our study, complemented by in vivo data of selected mutants reveals contact sites that extend beyond the EAA motif, including peptide regions localized to TMDs and periplasmic loops of MalFG.

Results Enzyme-linked absorbance studies: binding of MalFG-derived biotinylated soluble peptides to immobilized MalK As an initial approach to map the putative contact sites between MalK and MalFG we analyzed MalFG-derived soluble peptides for their capability to bind to purified MalK. To this end, 15mers representing the complete primary structures of MalF and MalG (total of 168 and 96 peptides, respectively) that overlapped by nine amino acid residues were synthesized on a cellulose membrane, biotinylated and subsequently cleaved off the solid phase as described.17,18 Mixtures of two consecutive peptides were then incubated with purified MalK previously immobilized to the wells of a microtiter plate. After washing off unbound peptides, the amounts of retained peptides were quantified enzymatically after incubation with peroxidaseconjugated avidin. In order to account for unspecific binding the obtained read-outs were corrected by subtracting absorbance values of the respective peptide mixes with immobilized bovine serum albumin (BSA). This unspecific reaction was in the range from zero to 40% of the absorbance measured with MalK. A total of 16 and 12 putative binding regions were identified in MalF and MalG, respectively (not shown). Subsequently, in order to narrow the number of possibly interacting peptides, a new set of peptides representing the above identified binding regions was synthesized. This time the peptides were 10mers overlapping by nine amino acid residues and each was incubated separately with immobilized MalK. Again, immobilized BSA was used as a control. As a representative example of the data obtained, Figure 1 shows the absorbance values of peptides representing the region Cys304 to Val327 of MalF. Peptide 9 (K312-GKAIYRVL-L321) displayed

388

Contact Sites in a Bacterial ABC Transporter

of the peptides within the topological models of MalF and MalG are shown in Figure 3(a) and (b) (residues in blue). As expected from the localization of the MalK dimer, peptides within cytoplasmically exposed loop regions of both proteins (MalF: C2 and C3; MalG: N terminus and C2) are involved in binding. However and most surprisingly, peptides from transmembrane and periplasmic domains

Figure 1. Binding of soluble biotinylated peptide 9 of MalF to immobilized MalK. Peptide binding assays were performed in microtiter plates in the presence (black bars) or absence (grey bars) of ATP/Mg2+ as described in Materials and Methods. Absorbance readings were corrected for unspecific binding by substracting absorbance values obtained with BSA in parallel assays. A representative data set out of three independent assays is shown. While absolute absorbance values varied between experiments the ratio of read-outs of each peptide to the strongest binder (K312-L321) did not change significantly (<30%).

the strongest signal while the readings for both neighboring peptides were substantially lower. These data indicate that in particular residues K312 and L321 are crucial for binding to MalK in the context of peptide 9 (see also below). The presence of ATP/Mg2+ in the incubation buffer seemed to slightly weaken the binding. In Figure 2 a comparative binding assay is shown with those peptides from each of the analyzed regions of MalF and MalG that exhibited strongest binding to MalK. Each column represents the average of three experiments that were corrected for unspecific binding to BSA as described above. By using a cut-off of 0.2, ten out of 17 MalF-derived peptides were considered to exhibit significant binding activities, with peptides 4, 5, 8, 9 and 11 of MalF giving the most intense signals. Among those, peptides 3 and 5 also displayed substantial binding to BSA and thus were not studied further. Only binding of peptides 12 and 14 was strongly affected by ATP/Mg2+. Interestingly, peptide 13 encompassing the N-terminal part of the conserved EAA motif exhibited almost no binding activity for MalK. In case of MalG, eight out of 14 peptides gave absorbance values of 0.2 or higher. The most intense signals were obtained with peptides 1 and 12–14. With the exception of peptide 12, all displayed somewhat weaker binding in the presence of ATP/ Mg2+. Peptide 11 that comprises the N-terminal part of the EAA motif was only weakly reacting, while peptide 12 representing the C-terminal region of the motif displayed the strongest signal of all peptides analyzed. Peptide 5 exhibited some binding to BSA and was thus not further considered. Table 1 compiles those MalFG-derived peptides that exhibited specific binding to MalK. The positions of each

Figure 2. Comparative binding of soluble biotinylated MalFG-derived peptides to MalK that displayed strongest binding intensities in the individual assays. Assay conditions were as described in Materials and Methods and in the legend to Figure 1. (a) and (b) MalF-derived peptides. (c) MalG-derived peptides. Peptide MalF-6 was analyzed in the absence of ATP/Mg2+ only. The absorbance values of each peptide represent the average of three ((a) and (b)) and two (c) experiments, respectively. Deviations from the mean were <39% ((a) and (b)) and <30% (c).

Contact Sites in a Bacterial ABC Transporter Table 1. MalFG-derived peptides with strong and specific binding affinity for immobilized MalK Effect of MgATP

Topological localizationb

Sequence

MalF 4 8 9 11 12 14 17

− − − − + + −

TMD 3 (TMD 4), C2 C2, (TMD 5) TMD 6 TMD 6, C3 C3 P4

V79-LFPLVCTI-A88 M300-VLACLVQW-E309 K312-GKAIYRVL-L321 L379-GYPYMMIL-C388 C388-MGLLKAIP-D397 Q412-NFFKITLP-L421 L467-VSYTYRIA-F476

MalG 1 6 9 10 12 13 14

(+) − (+) (+) − (+) (+)

N terminus P1 TMD 4 TMD 4, C2 C2, (TMD 5) (TMD 5), P3 TMD 6

K7-SQKLRLLI-T16 F79-PVLLWLWN-S88 Y166-LGGIALHV-W175 H173-VWTIKGYF-E182 W200-QAFRLVLL-P209 I220-LSFIAAIT-E229 D262-FAAAAVLS-A271

Peptidea

a b

Numbering is according to Figure 2. The topological localization is according to Figure 3.

were also identified. These peptides localize to TMDs 3–5 and periplasmic loop P4 of MalF, and to TMDs 4–6 and loop P1 of MalG.

389 Array studies: probing MalF and MalG-derived peptide arrays for MalK binding In order to obtain further support for these findings, we wished to perform detailed substitutional analyses of the identified peptides. However, using again soluble peptides would have been too laborious and time-consuming even with the nanosynthesis approach. Thus, we took advantage of the SPOT synthesis technology that permits parallel synthesis and screening of a large number of cellulose membrane-bound peptides. First, cellulose-bound peptide arrays representing the complete MalF and MalG sequences were screened for binding of MalK. The peptide arrays consisted of 16mers or 31mers, that overlapped with adjacent peptides by 15 and 30 amino acid residues, respectively. The cellulose membranes were incubated with 35S-labelled MalK, washed to remove excess label, and retained radioactivity was visualized by phosphoimaging. The results obtained with peptide arrays containing 16mers are shown in Figure 4(a) (MalF) and (b) (MalG). For both proteins identical signal patterns were obtained with 31mers (not shown). Thirteen and ten signal rows, respectively, were found for MalF and MalG, albeit with different

Figure 3. Topological models of MalF (a) and MalG (b) according to Froshauer et al.37 and Boyd et al.,38 respectively. Residues from peptide regions identified by enzyme-linked binding assays and arrays are shown in blue. Residues suggested from substitutional analyses to be crucial for interaction with MalK are shown in red. Residues discussed in the text are indicated by their codon number. Residues that are part of the EAA consensus sequence are underlined. P, Periplasmic loops; C, cytoplasmic loops.

Contact Sites in a Bacterial ABC Transporter

390

Figure 3 (legend on previous page)

intensities. The identified peptide regions were largely identical to those found in the first round of enzyme-linked binding assays with immobilized MalK (not shown). Thus, all but one peptide (L379D397 of MalF; see Figure 4(a)) suggested above to be involved in MalK binding (Table 1) were among those cellulose-bound peptides that had strongly retained [35S]MalK. Substitutional analyses Then, in order to identify amino acid residues within these MalFG peptides that are crucial for binding of MalK, substitutional analyses of selected peptides (16mers) were performed. In these experiments every position was substituted one-at-a-time by all other genetically encoded amino acids (except cysteine). Thus, all possible single site substitution analogs were synthesized on cellulose membranes and screened for binding of radiolabelled MalK as described above. The peptides considered encompass all MalG-derived peptides and five out of seven MalF-derived peptides as listed in Table 2. MalFderived peptides 11 and 12 (L379-D397) were excluded as the respective cellulose-bound peptides exhibited weak or no binding to MalK in the initial screening (Figure 4(a), see also above). Instead, a strongly reacting peptide (R112-T127) from signal row 3 (Figure 4(a)) was included in the study. Those arrays in which amino acid substitutions affected binding of MalK are shown in Figures 5 and 6. In case of MalF-derived peptide R112-T127 (Figure 5(a)), which is localized to the large periplasmic P2

loop (see Figure 3(a)), bulky polar residues at positions 118 and 121 seemed to be somewhat favorable for binding MalK, since glycine or hydrophobic residues weakened the binding. In peptide V306L321 (Figure 5(b)), representing the second cytoplasmic loop (C2), residues Y317, R318 and V319 appear to be sensitive for substitutions. However, while Y317 tolerated more or less only hydrophobic replacements the substitution profiles of R318 and V319 are less clear. In case of R318 signals were obtained with all replacing residues, albeit with weaker intensities. V319 also showed weaker signals with most replacements but not with tryptophan and tyrosine, while, interestingly, proline resulted in a more intense signal. The result for peptide A404-L419, representing the third cytoplasmic loop and encompassing most of the EAA motif (see Figure 3(a)) is shown in Figure 5(c). It is obvious that residues N413, F414, F415, and I417 cannot be replaced by either aspartate or glutamate without complete loss of binding. Moreover, glycine at positions 415–417 and proline at position 416 substantially affected binding. Residue I417 could only be replaced by other aliphatic hydrophobic residues and by phenylalanine. Analysis of MalG-derived peptide P6-L21 (Figure 6(a)) revealed only a minor effect of negatively charged or small residues replacing L13. The signals obtained with peptide E190-L205 encompassing the EAA motif are generally weak and thus somewhat hard to detect on a printout (Figure 6(b)). However, close inspection on the computer screen revealed that residues R204 and L205 are hardly replaceable

Contact Sites in a Bacterial ABC Transporter

391

Figure 4. MalK binding to cellulose-bound peptide arrays. Peptide arrays (16mers) derived from the sequences of MalF (a) and MalG (b) were probed for MalK binding. Peptide-bound cellulose membranes were prepared and incubated with 35S-labelled MalK as described in Materials and Methods. Last spots of rows (right) and N-terminal residues of peptides of the first spots of rows (left) are indicated.

without loss of binding activity. R204 tolerated only substitution by lysine and arginine as does L205. The same holds true for F203. Interestingly enough, lysine or arginine at position 201 substantially increased binding of MalK. Together, these findings might indicate that a positively charged residue is crucial in this region of the peptide chain. Peptide V223-V247 (Figure 6(c)), representing the third periplasmic loop of MalG, showed clear substitutional effects on binding of MalK at position 243. Residue Y243 was found to be irreplaceable by negatively charged residues and asparagine but also by small hydrophobic residues. Effect of selected point mutations in malFmalG on maltose transport in vivo Do these results reflect contact sites of MalK and MalFG in the context of the assembled protein complex in vivo? To address this question, we

introduced selected single, and double mutations in malF and malG and analyzed their consequences on transporter activity. To this end, E. coli strain ES66 (pFS1) was transformed with plasmids harboring the respective malF and malG alleles and the resulting transformants were assayed for growth on maltose as sole source of carbon and energy. Strain ES66 carries a deletion of the entire malB region encoding the maltose transport proteins while plasmid pFS1 provides the malK and malE genes in trans. Residues that were mutated according to the results of the substitutional analyses included K118(→G), Y317 (→Q), K416(→E), and I417(→E) of MalF, and L13 (→A), F203(→N,Q) and Y234(→A,N) of MalG (see Figure 3(a) and (b)). Residues from TMDs were not considered in order to exclude possible effects on correct membrane insertion and folding. The results expressed in doubling times of cells grown in liquid medium are summarized in Table 3. Most of the mutations allowed growth of the transformant very

Contact Sites in a Bacterial ABC Transporter

392 Table 2. Peptides subjected to complete substitutional analysis

Peptide MalF F78-VLFPLVCTIAIAFT-N93 R112-SYQAGKTYNFGLYP-T127 V306-QWEALKGKAIYRVL-L321 Q412-NFFKITLPLLIKPL-T427 G462-YTDLLVSYTYRIAF-E477 MalG P6-KSQKLRLLITHLGL-L21 T75-PPPFPVLLWLWNSV-K90 G168-GIALHVWTIKGYFE-T183 E190-AAALDGATPWQAFR-L205 I220-LSFIAAITEVPVAS-L235 D262-FAAAAVLSAIPITL-V277

Number of corresponding peptides Topological from Table 1 localization 4 − 8+9 14 17

TMD 3 P2 C2 C3 P4

1

N terminus, TMD 1 P1 TMD 4, C2 C2 TMD 5, P3 TMD 6

6 9 + 10 12 13 14

The peptides identified by binding to immobilized MalK (see Table 1) are in bold.

similar to control cells carrying the malFG wild-type alleles. A slightly slower growth was consistently observed with MalG-Y243A or MalG-L13A, however combination of both mutations did not result in an additive effect. Interestingly, while MalF-K416G did not cause a growth defect, although the substitution was shown to reduce binding of MalK (Figure 5(c)), the I417E mutant exhibited much slower growth than wild-type. Likewise, replacing MalG-F203 by asparagine or glutamine substantially affected growth of the transformants. In both cases, this was not caused by dislocalization of MalK as assayed by cell fractionation and subsequent immunoblotting. About the same amounts of MalK were found associated with the cytoplasmic membrane for wild-type and both mutants. Furthermore, localization of MalF to the membrane indicated that complex assembly had most likely occurred (Figure 7; data for I417E not shown). This conclusion is based on the earlier observation that assembly defects cause proteolytic cleavage of MalF.19 Possible effects on

the interaction with MalE can of course not be ruled out. Two mutants were subsequently selected for further characterization. MalF(I417E)GK 2 and MalFG(Y243A)K2 were purified from overproducing strains and incorporated into liposomes according to standard protocols.15 The resulting proteoliposomes were assayed for MalE-maltose-stimulated and vanadate-sensitive ATPase activity.15 Both criteria are widely used as a measure for transport activity in vitro.5,6,11,12,15 In accordance with the growth experiments, MalF(I417E)GK2 displayed no measurable ATPase activity, while MalFG(Y243A)K2 exhibited 86% of wild-type activity (100% = 2.1 μmol Pi/min per mg). Moreover, the latter activity was completely inhibited by vanadate (0.1 mM) (data not shown). Together, the observation that most of the mutants exhibited no phenotype is consistent with the notion that discontinuous binding sites exist between MalFG and MalK. If only of structural importance replacing one or two residues from these sites by unfavorable residues as predicted from the peptide arrays should not cause a functional defect. In contrast, the severe growth defects of mutants producing the variant MalG-F203N/Q and MalF-I417E, respectively, are likely suggesting that these residues are functionally critical for the transport process. The observation that complex assembly is unchanged compared to wild-type provides further evidence for these residues being also part of a discontinuous contact site.

Discussion Evaluation of the methods applied to study membrane protein complexes We have studied the contact sites of the membrane-integral subunits MalFG with the MalK ATPase homodimer of the maltose ABC transporter by using MalFG-derived synthetic peptides: firstly as soluble peptides in the nanomolar range and also

Figure 5. Substitutional analyses of MalF-derived peptides recognized by MalK. Each amino acid of the peptides (16mers) corresponding to spots 112 (a), 306 (b) and 404 (c) in Figure 4(a) (right-hand side of membrane) is substituted by all other 20 L-amino acids (except cysteine) in alphabetical order (shown on top of each membrane) and tested for binding of 35S-labelled MalK as described in Materials and Methods. All spots in the left column comprise the wild-type sequence (wt) of the peptides. Amino acids important for binding are underlined.

Contact Sites in a Bacterial ABC Transporter

393

Figure 6. Substitutional analyses of MalG-derived peptides recognized by MalK. The peptides analyzed correspond to spots 6 (a), 190 (b) and 232 (c) in Figure 4(b). See the legend to Figure 5 for details.

as cellulose membrane-bound peptide arrays both prepared by the SPOT synthesis approach. Preparation of soluble peptides by the SPOT technique resulted approximately in 50–100 nmol peptide per spot (spot area 0.25 cm2).17,18,20–23 Over the past years it was found that SPOT-synthesized soluble peptides (up to 18mers) can in most cases be obtained with a purity above 60%.20,23 More recently, it was demonstrated that even the 34meric FBP28D15N WW domain could be SPOT synthesized with an overall purity of 65%.24 The highly parallel character of the SPOT synthesis approach, the good synthesis quality of so prepared soluble peptides and the preparation of huge numbers of peptides at low costs make them ideal candidates for biological or biochemical solution-based screening studies. SPOT-synthesized soluble peptides have been successfully applied to several biological assay systems like T-cell epitope mapping,22 or membrane-integral receptor studies like the PACAP-VIP interaction.18 However, it should also be mentioned that the method, like all other solid phase peptide synthesis approaches, has its limitations when

sequences that cannot be synthesized by a standard synthesis protocol are concerned. Nevertheless, our control analysis by HPLC and MALDI-TOF mass spectrometry of some selected MalFG-derived sequences revealed peptide identity and purities in the range of 60–90%. However, analytical control of SPOT synthesized TMD-deduced peptides failed due to the fact that these peptides show extreme poor solubility in aqueous buffer systems. Therefore, standard HPLC and MALDI-TOF mass spectrometry analyses are unsuitable control instruments in this case. Nonetheless, it was demonstrated that the coupling efficiency of aliphatic and hydrophobic amino acids is satisfactory.18,25 Furthermore and as demonstrated recently by the successful SPOT synthesis of several 32meric leucine zipper analogs, amphipathic helical peptides can be synthesized with good yields

Table 3. Effects of mutations on cell growth on maltose MalFG variants synthesized

Topological localization

Wild type MalF K118G Y317Q K416E I417E MalG L13A F203N F203Q Y243A Y243N L13A, Y243A L13A, Y243N

Doubling time (h) 5.9

P2 C2 C3 C3

5.9 6.0 5.3 9.2

N terminus C2 C2 P3 P3 N terminus, P3 N terminus, P3

6.3 9.0 >>10 6.3 5.5 6.3 5.5

Cells of strain ES66(pFS1) harboring the respective pES62-97 derivatives were grown in minimal salt medium supplemented with 2% maltose and appropriate antibiotics at 37 °C. For details see Materials and Methods. Data represent average of three independent experiments.

Figure 7. Subcellular localization of MalK (a) and MalF (b) in mutant strains. Cells of host strain ES66(pFS1) (see Materials and Methods) harboring derivatives of plasmid pES67-92 that encode wild-type MalFMalG, MalFMalG(F403N), and MalFMalG(F403Q), respectively, were grown in LB medium (10 ml) supplemented with ampicillin (0.1 mg/ml) and chloramphenicol (0.02 mg/ml) at 37 °C to late exponential phase. Subsequently, cells were desintegrated by passage through a French press and fractions corresponding to low speed pellet (LSP), cytosol (Cy) and cytoplasmic membrane vesicles (CM) were obtained as described.15 Proteins from each fraction (20 μg) were separated by SDS-PAGE, electro-transferred to nitrocellulose membranes and probed with specific antisera raised against MalK and MalF, respectively. St, molecular weight markers.

394 and quality.26 Taking into account that TMDs are amphipathic helical structures composed of aliphatic amino acids it is likely that the SPOT synthesis of the TMD-deduced peptides of this study was successful on the cellulose membrane. Cellulose membrane-bound peptide arrays prepared by the SPOT synthesis technology have emerged over the past decade as a powerful proteomics tool to study molecular recognition events.16 Well-known applications include the identification of linear peptide epitopes that bind to antibodies27 or to protein interaction domains.28,29 Recently, a comprehensive and statistically significant study demonstrated the reliability of array-based measurement of peptide binding affinity.30 The intramembrane comparability was found to be reasonably good and peptide affinities correlated well with the measured spot signal intensities. In contrast, the inter-membrane comparability is quite low and a drawback of the method. In a previous study, we have successfully used peptide arrays to map the binding surfaces of the cytosolic regulatory protein enzyme IIAGlc that interact with MalK.31 In contrast, peptide arrays have not yet been employed to identify contact sites between subunits in a membrane protein complex. Such interactions are likely to include linear as well as conformational (discontinuous) binding sites, which make their identification more difficult. By pepscan experiments sequentially continuous protein binding sites can be readily mapped, which is one the strengths of SPOT technology.16 Discontinuous protein–protein contact sites are formed by sequentially distant fragments of the molecule that are brought into spatial proximity by protein folding. As a result, affinities of these separated fragments are generally low. Therefore, it is a great challenge to map discontinuous contact sites by pepscan experiments. Nevertheless, the first mapping of discontinuous protein–protein contact sites applying the SPOT technology was reported by Reineke et al., 32,33 mapping the discontinuous binding site of interleukin-10. However, because of the low affinity the peptides must be adjusted for optimal length and their specificity has to be tested by additional substitutional analyses and by quantitative affinity measurement by ELISA and/or surface plasmon resonance experiments. Nonetheless, by combining data obtained with soluble and cellulose-bound peptide arrays with the results from mutational analyses, we have identified putative contact sites between MalFG and MalK. While binding assays employing biotinylated soluble peptides and immobilized MalK enabled us to account for unspecific binding events, cellulose bound peptide arrays allowed complete substitutional analyses of peptides in question. Together with the in vivo analysis of mutants, both approaches identified possible interaction sites localized to periplasmic and cytoplasmic loop regions as well as to TMDs of MalF and MalG. So far, only residues from the conserved EAA motif have been demonstrated to be crucial for the

Contact Sites in a Bacterial ABC Transporter

interaction between MalFG and MalK.10,13 Peptides derived from this motif, although representing the less well-studied carboxy-terminal end only, were among those that displayed strongest binding to MalK. Thus, this finding is indicative for the suitability of the approach used here to map contact sites between subunits of a protein complex. Below we discuss in detail the identified putative contact sites with respect to other reports. Putative MalK binding sites in TMDs No significant binding of MalK to peptides derived from the N-terminal TMDs 1and 2 of MalF were detected, which is consistent with findings suggesting that both helices are dispensable for function.34,35 Specific binding was however found for peptides V79-A88 (TMD 3), M300-E309 (TMD 4/C2), and L379-D397 (TMD 6/C3). Mutational analysis by Steinke et al.36 identified TMD 3 as crucial for function. Substitutions mainly affected MalF assembly, which does not exclude that contacts to other subunits are disturbed. Furthermore, mutations of T295, C304 and Q307, respectively, in TMD 4 of MalF were found to result in reduced transport activity.36 Also, a deletion mutant affecting TMD 6 caused complete dislocation of MalK.35 In case of MalG our results suggest that TMDs 4–6 might participate in contacting MalK. Linker insertions at positions A171 (TMD 4) and I220 (TMD 5) (see Figure 3(b)), respectively, caused severe defects in protein stability.19 Although speculative, the phenotype could be due to insufficient complex assembly and subsequent degradation of MalG. No other evidence that TMD 6 is involved in MalK binding has thus far been reported. A note of caution however seems appropriate concerning the assignments of TMDs and connecting loop regions. Although the topological models of MalF and MalG (Figure 3(a) and (b)) are based on experimental data using the phoA fusion technique37–39 the transition of the polypeptide chain from TMD to loop and vice versa cannot be precisely determined. Thus, these models do not exclude that peptides derived from one end of a transmembrane helix according to Figure 3(a) and (b) are actually part of the connecting loop. In fact, alternative topological models based on several criteria, including prediction algorithms and mutational data,40 predict TMDs 4 and 6 of MalF to have their carboxyterminal ends at C300 and L387, respectively. Furthermore, according to this model TMD 4 of MalG would encompass residues L167 to I184, thereby placing peptides 9 and 10 (see Table 1) entirely into the membrane. It should also be noted that in two reports using alkaline phosphatase fusions TMD 4 of MalG was differently defined.38,39 Furthermore, our study did not include length analyses of the identified peptides. Thus, it is not unlikely that the minimum length of a peptide displaying MalK binding affinity is smaller than ten. Consequently, peptides extending from loop regions into TMDs (see Figure 3(a) and (b)) might

Contact Sites in a Bacterial ABC Transporter

actually require only loop-derived sequences for binding. Moreover, only those residues might actually be involved in binding that lie on the same face of the helix. Clearly, we will have to await the crystal structure of the complete MalFGK2 complex to clarify this matter. In case of the BtuCD complex for which structural data are available, predicted lengths and sequences of helical and loop regions in fact differ somewhat from those found in the crystal, thereby supporting the above notion.14 Taking these considerations into account, the contact sites identified by our study that localize to TMDs might thus be restricted to regions in TMDs 3 and 6 of MalF and TMDs 4 and 6 of MalG. In any case, it is intriguing that transmembrane regions at all display binding activity for MalK. This might indicate that MalK embeds itself into the membrane during its interaction with MalFG. Protease studies with spheroplasts containing the MalFGK2 complex suggested such a scenario.41 Furthermore, an earlier report on the protease susceptibility of HisP, the ABC protein of the histidine transporter of S. typhimurium, arrived at the same conclusion,42 although it should be noted that the physiological significance of these results was questioned.43 In addition, transmembrane helices that make contacts to MalK might be tilted rather than being oriented perpendicularly to the membrane. The way by which the ten transmembrane helices of BtuC are packed together is at least not in contradiction to this notion,14 also variations in helix packing are likely to exist in TMDs from different ABC transporters. In this respect it is also worth mentioning that heterobifunctional crosslinking of the ABC protein DrrA to its cognate membrane protein DrrB also identified residues localized to TMDs of DrrB as being in close contact to DrrA.44 Putative MalK binding sites in periplasmic loop regions Our study suggested two regions within periplasmic loops P2 (R112-T127) and P4 (L467-F476) of MalF to interact with MalK. Since the P2-peptide also displayed significant binding to BSA, a rather unspecific interaction cannot be ruled out. However, mutants carrying deletions of residues 112 to 121 and 114 to 121 in P2, respectively, were found to partially accumulate MalK in the cytoplasm.35 Based on these and other mutations, these authors speculated that the large P2 loop might be in contact to MalK. Our data are consistent with this notion. A deletion mutant comprising residues 453 to 470 in P4 was defective in maltose transport and also affected MalK localization, leading the authors to suggest that P4 might be critical for both interaction with the substrate binding protein MalE and MalK.35 Peptides identified within periplasmic loops of MalG are part of P1 and P3, respectively. Again, support for the relevance of the region in P1 was

395 obtained by linker insertion mutagenesis. In their study, Nelson and Traxler19 found insertions at positions W84, K90 and A92 (Figure 3(b)) to cause protein instability. The identified peptide region within the third periplasmic loop P3 was also characterized by insertion mutagenesis (at T244 and A246, see Figure 3(b)) and found to be partially assembly-deficient.19 Together, these results suggest that the MalK dimer might also contact extracytoplasmic domains of MalFG. This could be explained by the assumption that the loops are not fully solvent-exposed but partially embedded in the MalFG core. Interestingly, and at least not in contradiction to this notion, MalFP2 and MalG-P1 are relatively resistant to trypsin in the resting state but expose a cleavage site in the presence of ATP. Thus, both loops are likely changing their conformation during the transport cycle.12 Putative MalK binding sites in cytoplasmic loops Our study identified peptides from the cytoplasmically oriented loops C2 and C3 of MalF as well as from the N terminus and from loop C2 of MalG, respectively, as being crucial for MalK binding. MalF-C2 was suggested by Tapia et al.35 to participate in the attachment of MalK to the membrane based on the phenotype of a mutant carrying a deletion of residues 312–318. Our data are perfectly consistent with this notion although a single point mutation (Y317Q) exhibited no effect on cell growth (Figure 5(b)). The conserved EAA motif, which is localized to the C3 and C2 loops, respectively, of MalF and MalG was demonstrated by crosslinking experiments to reside in close proximity to the helical subdomain of MalK.10 These data provide independent evidence for our finding that peptides from both loops are involved in MalK binding. Moreover, the data presented here suggest that residues preceeding the core motif or at its C-terminal end are critical (Figure 3(a) and (b)). In MalF, especially residues in region N413–L419 turned out to be sensitive to replacements, as was the case for the corresponding peptide W200-P209 in MalG. Significant sequence similarity of both peptides likely accounts for this observation. Interestingly, substitution of glutamate for MalF-K416 or MalF-I417 caused different phenotypes. While the strain carrying K416E grew like wild-type, the I417E mutant displayed a severe growth defect (Table 3) and the purified variant lacked ATPase activity. Isoleucine is predominantly present at this position of the EAA motif in ABC membrane proteins closely related to the maltose transporter.45 Only conservative replacements by valine and leucine are found, which might explain the observed phenotype. To our knowledge, the residue has not yet been included in a mutational analysis. With MalG-F203 we have identified another residue from the EAA motif that is crucial for function as mutations to either Gln or Asn resulted

396 in severe growth defects. In related proteins the residue aligns only with other aromatic residues, although it is not part of the consensus EAA sequence. 45 Since MalK was properly associated with the membrane (Figure 7) the residue might also be part of a discontinuous binding epitope for MalK. This is in contrast to mutations affecting the highly conserved A192(→D) and the invariant G196(→P) that caused mislocalization of MalK.13 Here, our data suggest a role in complex assembly rather than in direct MalK binding, since binding affinity of the respective biotinylated peptide to MalK was weak (Figure 2(c), compare peptides 11 and 12). Interestingly, peptide C388-D397 derived from the MalF-C3 loop displayed binding to immobilized MalK only in the absence of ATP/Mg2+, suggesting that the site of interaction is recognized when the MalK dimer resides in the open conformation.4 This is in perfect agreement with results from a previous crosslinking study demonstrating that homobifunctional thiol-specific linkers paired MalFS403C with A85C of MalK in the absence of ATP only.10 Effect of ATP A few biotinylated peptides displayed significantly weaker binding to MalK in the presence of ATP/Mg2+ (Figure 2). Crystal structures of the soluble MalK dimer suggested that ATP but not ADP binding triggers a transition from an open to a closed state.4,8 Thus, our findings might imply that the closed MalK dimer has less contact sites on the TMDs. Furthermore, all but one of these peptides were derived from MalG. Although the effect on a single peptide is difficult to discuss in the absence of a crystal structure of the complete transporter, this observation indicates that the closed MalK dimer might interact differently with MalF and MalG. For the EAA motifs of both subunits this has already been proposed on the basis of mutant analyses13 and crosslinking experiments.10 Moreover, the structure of the multidrug ABC exporter Sav1866 revealed different contacts between the NBDs and two helices that connect the NBDs to the TMDs.46 In conclusion, our data provide detailed insight on subunit–subunit interactions that might occur during the transport cycle of an ABC import system and will set the stage for experiments to further define the contact sites, e.g. by means of crosslinking and surface plasmon resonance.

Materials and Methods Bacterial strains and plasmids E. coli strain JM109 (Stratagene, La Jolla, USA) was routinely used for cloning purposes and as host for plasmid pGS91-1 (malKht).47 E. coli strain ES66 is a derivative of strain ED16913 and carries a malTc mutation that causes constitutive synthesis of the positive regulator

Contact Sites in a Bacterial ABC Transporter of the maltose regulon, MalT (laboratory collection). Plasmid pFS1 carries malE under the control of its chromosomal promoter and malK under control of the ptrc promoter on vector pSE380 (laboratory collection). Derivatives of pES67-92 (malFG under the control of the trc promotor on plasmid pSU1848 carrying single point mutations) were obtained by site-directed mutagenesis using Stratagene's Quikchange kit. Growth experiments Cells of E. coli strain ES66 carrying the indicated plasmids were grown in 5 ml of M9 minimal medium49 supplemented with maltose (2%, w/v), ampicillin (0.1 mg/ml), and chloramphenicol (0.02 mg/ml) at 37 °C under vigorous agitation. Protein purification MalFGK2 (wild-type and variants) and MalE were prepared as described.15 MalK was either purified as an N-terminal His6 fusion or tag-free as described.15,50 [35S]methionine-labelling of MalK E. coli strain JM109(pGS91-1)15 (20 ml) was grown in minimal medium E51 supplemented with 0.2% Casamino acids (omitting methionine), 2 μg/ml thiamine, 0.5% (w/v) glucose, 0.1 mg/ml ampicillin at 37 °C to A650nm = 0.5. Cells were harvested by centrifugation for 5 min at 5000 g, resuspended in pre-heated fresh medium and incubated with 0.5 mM IPTG at 37 °C for 12 min under vigorous shaking. Subsequently, [35S] methionine (250 μCi) was added and incubation continued for 5 min. Cells were then cooled on ice for 20 min, harvested by centrifugation as above, washed once with cell lysis buffer (50 mM Tris-HCl (pH 8), 150 mM NaCl, 20% (v/v) glycerol, 0.1 mM phenylmethylsulfonylfluoride) and subsequently disrupted by ultrasonication (3× for 1 min with 1 min intervals; output 60%; Branson sonifier). The cytosolic fraction was then recovered by centrifugation for 30 min at 12,000g and MalK was purified by Ni-NTA chromatography as described elsewhere.15 The average specific radioactivity was 5 × 105 cpm/μg MalK. Peptide synthesis on cellulose membranes (SPOT synthesis) Cellulose-bound peptide libraries were prepared by semi-automatic SPOT synthesis using a SPOT robot (INTAVIS AG, Köln, Germany). SPOT synthesis was carried out as described in the SPOT synthesis protocol52 and arrays were synthesized on modified Whatman 50 cellulose membranes (Whatman, Maidstone, UK). Sequence files and design of the arrays were generated with the in house software LISA. Peptides derived from S. typhimurium MalF (GenBank accession no. CAA38190) and MalG (GenBank accession no. CAA38191) were used for pepscan analyses. To this end, two peptide arrays consisting of 16meric or 31meric peptides, overlapping by 15, and 30 amino acid residues, respectively, were synthesized. Complete substitutional and length analyses of the interacting 16mer-peptides were generated using the software LISA and subsequently synthesized as described.20,53

Contact Sites in a Bacterial ABC Transporter Screening of cellulose membrane-bound peptide arrays Before screening, the dried membranes were washed for 10 min in ethanol, 3 × 10 min in TBS (50 mM Tris-HCl (pH 8), 137 mM NaCl, 27 mM KCl) and subsequently incubated in TBS, supplemented with 5% blocking buffer (Sigma, München, Germany) and 5% (w/v) sucrose, for 3 h at room temperature. After washing with TBS, peptide arrays were incubated with 35S-labelled MalK (105 cpm) in blocking buffer, supplemented with 20% glycerol for 6 h at room temperature with gentle shaking. Unbound MalK was removed with TBS and peptide-bound MalK was visualized and quantified using a phosphoimager and associated software (Fuji, Japan). Nano-synthesis of biotinylated peptides The synthesis of cleavable biotinylated peptides on a CAPE (cellulose-amino-hydroxyl propylether) membrane was carried out as described.17 After completion of the synthesis, a spacer consisting of three β-alanine residues was attached to the N terminus of each peptide. Subsequent biotinylation was achieved by using 0.6 M Fmoc-lysine(biotinyl-ε-aminocaproyl)-OH, previously activated with TBTU and DIPEA. Briefly, Fmoc-lysine-Biotin, TBTU, and DIPEA were mixed in N-methyl-pyrrolidon, incubated for 5 min, and centrifuged to remove undissolved Fmoc-lysine-Biotin. This step was repeated three times for each peptide. Removal of Fmoc and side-chain deprotection was performed as described.20 Biotinylated peptides were finally released by punching out the spots followed by treatment of each spot with 0.125 M NaOH in 50% acetonitrile for 30 min. Subsequently, the pH was adjusted to 2–4 by adding hydrochloric acid, the peptides were dried in a vacuum centrifuge and then dissolved in 50 mM Tris-HCl (pH 7.5), containing 50% (v/v) DMSO to a final concentration of 1.2 mM. Some peptides were quality-controlled by RP-HPLC and MALDI-TOF mass spectrometry. Peptide binding assay Binding of soluble biotinylated peptides derived from MalF and MalG was assayed by incubation with purified MalK immobilized in the wells of a microtiter plate (Polysorb™, Nunc, Wiesbaden, Germany). For each peptide to be analyzed two wells were incubated over night with purified MalK (1 μg) in 50 mM Tris-HCl (pH 8), 0.1 mM phenylmethylsulfonylfluoride (PMSF), 20% glycerol (buffer 1) or with 0.5 μg MalK in buffer 1 containing 5 mM MgCl2, 10 mM ATP at 4 °C (final volume: 100 μl). Under these conditions, MalK displayed a specific ATPase activity of 5 nmol Pi/min per mg (control value at 37 °C: 377 nmol Pi/min per mg; see also Schmees et al.47). Thus, hydrolysis of ATP during the incubation time was negligible. Subsequently, the wells were washed three times with 50 mM Tris-HCl (pH 7.5), 200 mM NaCl, 10% glycerol (buffer 2). To block unspecific binding to the surface of the Polysorb surface the wells were then incubated with 0.5% (w/v) BSA in buffers 1 and 2, respectively, for 30 min at room temperature. A third well was also incubated with buffer 1 containing BSA and used as reference. After three washing steps with buffer 2 containing 0.05% (v/v) Tween 20, peptides (12 μM each) were added to wells 1 and 3 in 50 mM Tris-HCl (pH 7.5),

397 200 mM NaCl, 1 mM dithiothreitol (DTT), 0.1 mM PMSF, 20% glycerol, 0.05% Tween 20 (buffer 3) and to well 2 in buffer 3 containing 5 mM MgCl2, 10 mM ATP (buffer 4). Incubation took place overnight at 4 °C under gentle shaking. Subsequently, the wells were washed with buffer 3 and 4, respectively, and bound peptides were detected enzymatically after reaction with peroxidaseconjugated avidin (1 μg/ml) for 20 min at room temperature. After five washing steps per well with buffers 3 and 4, respectively, substrate solution (100 mM acetate/citrate buffer (pH 6), 8.5 mM H2O2, 5.5 mM tetramethylbenzidine) was added. The reaction was terminated after 100 s by adding 0.05 M Na2SO4 in 2 M H2SO4 and absorbance readings for each well were recorded with a microplate reader (SCT Labinstruments, Crailsheim, Germany). Analytical procedures Preparation of proteoliposomes, ATPase assay, protein determination, SDS-PAGE and immunoblotting were performed as described elsewhere.15,21

Acknowledgements We thank H. Landmesser for excellent technical assistance. This research was supported by the Deutsche Forschungsgemeinschaft (SFB 449, B14 and Z1).

References 1. Holland, E. B., Cole, S., Kuchler, K. & Higgins, C. F. (eds) (2003). ABC Proteins: From Bacteria to Man, Academic Press, New York. 2. Schneider, E. (2003). Import of solutes by ABC transporters-the maltose and other systems. In ABC Proteins: From Bacteria to Man (Holland, E. B., Cole, S., Kuchler, K. & Higgins, C., eds), pp. 157–185, Academic Press, New York. 3. Davidson, A. L. & Chen, J. (2004). ATP-binding cassette transporters in bacteria. Annu. Rev. Biochem. 73, 241–268. 4. Chen, J., Lu, G., Lin, J., Davidson, A. L. & Quiocho, F. A. (2003). A tweezers-like motion of the ATPbinding cassette dimer in an ABC transport cycle. Mol. Cell, 12, 651–661. 5. Davidson, A. L., Shuman, H. A. & Nikaido, H. (1992). Mechanism of maltose transport in Escherichia coli: transmembrane signaling by periplasmic binding proteins. Proc. Natl Acad. Sci. USA, 89, 2360–2364. 6. Chen, J., Sharma, S., Quiocho, F. A. & Davidson, A. L. (2001). Trapping the transition state of an ATP-binding cassette transporter: evidence for a concerted mechanism of maltose transport. Proc. Natl Acad. Sci. USA, 98, 1525–1530. 7. Austermühle, M. I., Hall, J. A., Klug, C. S. & Davidson, A. L. (2004). Maltose-binding protein is open in the catalytic transition sate for ATP hydrolysis during maltose transport. J. Biol. Chem. 279, 28243–28250. 8. Lu, G., Westbrooks, J. M., Davidson, A. L. & Chen, J. (2005). ATP hydrolysis is required to reset the ATP-

Contact Sites in a Bacterial ABC Transporter

398

9.

10.

11.

12.

13.

14. 15.

16.

17.

18.

19.

20.

21.

22. 23.

binding cassette dimer into the resting-state conformation. Proc. Natl Acad. Sci. USA, 102, 17969–17974. Schneider, E., Wilken, S. & Schmid, R. (1994). Nucleotide-induced conformational changes of MalK, a bacterial ATP binding cassette transporter protein. J. Biol. Chem. 269, 20456–20461. Hunke, S., Mourez, M., Jèhanno, M., Dassa, E. & Schneider, E. (2000). ATP modulates subunit-subunit interactions in an ATP-binding cassette transporter (MalFGK2) determined by site-directed chemical cross-linking. J. Biol. Chem. 275, 15526–15534. Mannering, D. E., Sharma, S. & Davidson, A. L. (2001). Demonstration of conformational changes associated with activation of the maltose transport complex. J. Biol. Chem. 276, 12362–12368. Daus, M. L., Landmesser, H., Schlosser, A., Müller, P., Herrmann, A. & Schneider, E. (2006). ATP induces conformational changes of periplasmic loop regions of the maltose ATP-binding cassette transporter. J. Biol. Chem. 281, 3856–3865. Mourez, M., Hofnung, N. & Dassa, E. (1997). Subunit interactions in ABC transporters: a conserved sequence in hydrophobic membrane proteins of periplasmic permease defines an important site of interactions with the ATPase subunit. EMBO J. 16, 3066–3077. Locher, K. P., Lee, A. T. & Rees, D. C. (2002). The E. coli BtuCD structure: a framework for ABC transporters and mechanism. Science, 296, 1091–1098. Landmesser, H., Stein, A., Blüschke, B., Brinkmann, M., Hunke, S. & Schneider, E. (2002). Large-scale purification, dissociation and functional reassembly of the maltose ATP-binding cassette transporter (MalFGK2) of Salmonella typhimurium. Biochim. Biophys. Acta, 1565, 64–72. Reineke, U., Volkmer-Engert, R. & SchneiderMergener, J. (2001). Applications of peptide arrays prepared by the SPOT-technology. Curr. Opin. Biotechnol. 12, 59–64. Licha, K., Bhargava, S., Rheinländer, C., Becker, A., Schneider-Mergener, J. & Volkmer-Engert, R. (2000). Highly parallel nano-synthesis of cleavable peptidedye conjugates on cellulose membranes. Tetrahedron Letters, 41, 1711–1715. Bhargava, S., Licha, K., Knaute, T., Ebert, B., Becker, A., Grotzinger, C. et al. (2002). A complete substitutional analysis of VIP for better tumor imaging properties. J. Mol. Recogn. 15, 145–153. Nelson, B. D. & Traxler, B. (1998). Exploring the role of integral membrane proteins in ATP-binding cassette transporters: analysis of a collection of MalG insertion mutants. J. Bacteriol. 180, 2507–2514. Wenschuh, H., Volkmer-Engert, R., Schmidt, M., Schulz, M., Schneider-Mergener, J. & Reineke, U. (2000). Coherent membrane supports for parallel microsynthesis and screening of bioactive peptides. Biopolymers (Peptide Sci.), 55, 188–206. Stein, A., Seifert, M., Volkmer-Engert, R., Siepelmeyer, J., Jahreis, K. & Schneider, E. (2002). Functional characterization of the maltose ATP-binding-cassette transporter of Salmonella typhimurium by means of monoclonal antibodies directed against the MalK subunit. Eur. J. Biochem. 269, 4074–4085. Kamradt, T. & Volkmer-Engert, R. (2004). Crossreactivity of T lymphocytes in infection and autoimmunity. Mol. Diversity, 8, 271–280. Ay, B., Streitz, M., Boisguerin, P., Schlosser, A., Kern, F. & Volkmer, R. (2007). Sorting and pooling strategy: A

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35. 36.

37.

38.

novel tool to map a virus proteome for CD8 T-cell epitopes. Biopolymers, 88, 64–75. Przezdziak,J.,Tremmel,S.,Kretzschmar,J.,Beyermann, M., Bienert, M. & Volkmer-Engert, R. (2006). Probing the ligand binding specificity and analyzing the folding state of SPOT synthesized FBP28 WW domain variants. ChemBioChem. 7, 780–788. Ay, B., Volkmer, R. & Boisguerin, P. (2007). Synthesis of cleavable peptides with authentic C-termini: an application for fully automatic SPOT synthesis. Tetrahedron Letters, 48, 361–364. Portwich, M., Keller, S., Strauss, H. M., Mahrenholz, C. C., Kretzschmar, I., Kramer, A. & Volkmer, R. (2007). A network of coiled-coil associations derived from synthetic GCN4 leucine-zipper arrays. Angew. Chem. Int. Ed. Engl. 46, 1654–1657. Reineke, U., Kramer, A. & Schneider-Mergener, J. (1999). Antigen sequence- and library-based mapping of linear and discontinous protein-protein-interaction sites by spot synthesis. Curr. Top. Microbiol. Immunol. 243, 23–36. Boisguerin, P., Leben, R., Ay, B., Radziwill, G., Moelling, K., Dong, L. & Volkmer-Engert, R. (2004). An improved method for the synthesis of cellulose membrane-bound peptides with free C termini is useful for PDZ domain binding studies. Chem. Biol. 11, 449–459. Landgraf, C., Panni, S., Montecchi-Palazzi, L., Castagnoli, L., Schneider-Mergener, J., Volkmer-Engert, R. & Cesareni, G. (2004). Protein interaction networks by proteome peptide scanning. PLoS Biol. 2, E14. Weiser, A. A., Or-Guil, M., Tapia, V., Leichsenring, A., Schuchhardt, J., Frommel, C. & Volkmer-Engert, R. (2005). SPOT synthesis: reliability of array-based measurement of peptide binding affinity. Anal. Biochem. 342, 300–311. Blüschke, B., Volkmer-Engert, R. & Schneider, E. (2006). Topography of the surface of the signal transducing protein EIIAGlc that interacts with the MalK subunits of the maltose ATP-binding cassette transporter (MalFGK2) of Salmonella typhimurium. J. Biol. Chem. 281, 12833–12840. Reineke, U., Sabat, R., Volk, H. D. & SchneiderMergener, J. (1998). Mapping of the interleukin-10/ interleukin-10 receptor combining site. Protein Sci. 7, 951–960. Reineke, U., Sabat, R., Misselwitz, R., Welfle, H., Volk, H. D. & Schneider-Mergener, J. (1999). A synthetic mimic of a discontinuous binding site on interleukin10. Nature Biotechnol. 17, 271–275. Ehrmann, M. & Beckwith, J. (1991). Proper insertion of a complex membrane protein in the absence of its amino-terminal export signal. J. Biol. Chem. 266, 16530–16533. Tapia, M. I., Mourez, M., Hofnung, M. & Dassa, E. (1999). Structure-function study of MalF protein by random mutagenesis. J. Bacteriol. 181, 2267–2272. Steinke, A., Grau, S., Davidson, A., Hofmann, E. & Ehrmann, M. (2001). Characterization of transmembrane segments 3, 4, and 5 of MalF by mutanional analysis. J. Bacteriol. 183, 375–381. Froshauer, S., Green, G. N., Boyd, D., McGovern, K. & Beckwith, J. (1988). Genetic analysis of the membrane insertion and topology of MalF, a cytoplasmic membrane protein of Escherichia coli. J. Mol. Biol. 200, 501–511. Boyd, A., Traxler, B. & Beckwith, J. (1993). Analysis of the topology of a membrane protein by using a

Contact Sites in a Bacterial ABC Transporter

39. 40. 41.

42.

43.

44.

45.

46.

minimum number of alkaline phosphatase fusions. J. Bacteriol. 175, 553–556. Dassa, E. & Muir, S. (1993). Membrane topology of MalG, an inner membrane protein from the maltose transport system of E. coli. Mol. Microbiol. 7, 29–38. Ehrmann, M., Ehrle, R., Hofmann, E., Boos, W. & Schlösser, A. (1998). The ABC maltose transporter. Mol. Microbiol. 29, 685–694. Schneider, E., Hunke, S. & Tebbe, S. (1995). The MalK protein of the ATP-binding cassette transporter for maltose of Escherichia coli is accessible to protease digestion from the periplasmic side of the membrane. J. Bacteriol. 177, 5364–5367. Baichwal, P. A., Liu, D. X. & Ames, G. F.-L. (1993). The ATP-binding component of a prokaryotic traffic ATPase is exposed to the periplasmic (external) surface. Proc. Natl Acad. Sci. USA, 90, 620–624. Blott, E. J., Higgins, C. F. & Linton, K. J. (1999). Cysteine-scanning mutagenesis provides no evidence for the extracellular accessibility of the nucleotidebinding domains of the multidrug resistance transporter P-glycoprotein. EMBO-J. 23, 6800–6808. Kaur, P., Rao, D. K. & Gandlur, S. M. (2005). Biochemical characterization of domains in the membrane subunit DrrB that interact with the ABC subunit DrrA: identification of a conserved motif. Biochemistry, 44, 2661–2670. Saurin, W., Köster, W. & Dassa, E. (1994). Bacterial binding protein-dependent permeases: charaterization of distinctive signatures for functionally related integral cytoplasmic membrane proteins. Mol. Microbiol. 12, 993–1004. Dawson, R. J. P. & Locher, K. P. (2006). Structure of a

399

47.

48.

49. 50.

51. 52.

53.

bacterial multidrug ABC transporter. Nature, 443, 180–185. Schmees, G., Stein, A., Hunke, S., Landmesser, H. & Schneider, E. (1999). Functional consequences of mutations in the conserved ‘signature sequence’ of the ATP-binding-cassette protein MalK. Eur. J. Biochem. 266, 420–430. Martinez, E., Bartolome, B. & de la Cruz, F. (1988). pACYC 184-derived cloning vectors containing the multiple cloning site and lacZα reporter gene of pUC8/9 and pUC18/19. Gene, 68, 159–162. Miller, J. H. (1972). Experiments in Molecular Genetics, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Schneider, E., Linde, M. & Tebbe, S. (1995). Functional purification of a bacterial ATP-binding cassette transporter protein (MalK) from cytoplasmic fraction of an overproducing strain. Protein Expres. Purif. 6, 10–14. Roth, J. (1970). Genetic techniques in studies of bacterial metabolism. Methods Enzymol. 17, 3–35. Wenschuh, H., Gausepohl, H., Germeroth, L., Ulbricht, M., Matuschewski, H., Kramer, A. et al. (2000). Positionally addressable parallel synthesis on continuous membranes. In Combinatorial Chemistry: A Practical Approach (Fenniri, H., ed), pp. 95–116, Oxford University Press, Oxford. Volkmer-Engert, R., Ehrhard, B., Höhne, W., Kramer, A., Hellwig, J. & Schneider-Mergener, J. (1994). Preparation, analysis and antibody binding studies of simultaneously synthesized soluble and cellulosebound HIV-1 p24 peptide epitope libraries. Lett. Peptide Sci. 1, 243–253.

Edited by I. B. Holland (Received 15 December 2006; received in revised form 12 March 2007; accepted 13 March 2007) Available online 24 March 2007