Time-resolved Fourier Transform Infrared Spectroscopy of the Nucleotide-binding Domain from the ATP-binding Cassette Transporter MsbA

Background: The dynamics of coupling of ATP hydrolysis with transport in ATP-binding cassette transporters is not well understood. Results: Characterization of ATP hydrolysis in the MsbA nucleotide-binding domain by time-resolved FTIR spectroscopy revealed two rate constants for ATP binding with dimerization and hydrolysis. Conclusion: ATP hydrolysis is rate-limiting. Significance: The identification of the IR fingerprints of the motor domain will facilitate real-time analysis of the full-length MsbA transport cycle. MsbA is an essential Escherichia coli ATP-binding cassette (ABC) transporter involved in the flipping of lipid A across the cytoplasmic membrane. It is a close homologue of human P-glycoprotein involved in multidrug resistance, and it similarly accepts a variety of small hydrophobic xenobiotics as transport substrates. X-ray structures of three full-length ABC multidrug exporters (including MsbA) have been published recently and reveal large conformational changes during the transport cycle. However, how ATP hydrolysis couples to these conformational changes and finally the transport is still an open question. We employed time-resolved FTIR spectroscopy, a powerful method to elucidate molecular reaction mechanisms of soluble and membrane proteins, to address this question with high spatiotemporal resolution. Here, we monitored the hydrolysis reaction in the nucleotide-binding domain of MsbA at the atomic level. The isolated MsbA nucleotide-binding domain hydrolyzed ATP with Vmax = 45 nmol mg−1 min−1, similar to the full-length transporter. A Hill coefficient of 1.49 demonstrates positive cooperativity between the two catalytic sites formed upon dimerization. Global fit analysis of time-resolved FTIR data revealed two apparent rate constants of ∼1 and 0.01 s−1, which were assigned to formation of the catalytic site and hydrolysis, respectively. Using isotopically labeled ATP, we identified specific marker bands for protein-bound ATP (1245 cm−1), ADP (1101 and 1205 cm−1), and free phosphate (1078 cm−1). Cleavage of the β-phosphate–γ-phosphate bond was found to be the rate-limiting step; no protein-bound phosphate intermediate was resolved.

MsbA is an essential Escherichia coli ATP-binding cassette (ABC) transporter involved in the flipping of lipid A across the cytoplasmic membrane. It is a close homologue of human P-glycoprotein involved in multidrug resistance, and it similarly accepts a variety of small hydrophobic xenobiotics as transport substrates. X-ray structures of three full-length ABC multidrug exporters (including MsbA) have been published recently and reveal large conformational changes during the transport cycle. However, how ATP hydrolysis couples to these conformational changes and finally the transport is still an open question. We employed time-resolved FTIR spectroscopy, a powerful method to elucidate molecular reaction mechanisms of soluble and membrane proteins, to address this question with high spatiotemporal resolution. Here, we monitored the hydrolysis reaction in the nucleotide-binding domain of MsbA at the atomic level. The isolated MsbA nucleotide-binding domain hydrolyzed ATP with V max ‫؍‬ 45 nmol mg ؊1 min ؊1 , similar to the full-length transporter. A Hill coefficient of 1.49 demonstrates positive cooperativity between the two catalytic sites formed upon dimerization. Global fit analysis of time-resolved FTIR data revealed two apparent rate constants of ϳ1 and 0.01 s ؊1 , which were assigned to formation of the catalytic site and hydrolysis, respectively. Using isotopically labeled ATP, we identified specific marker bands for protein-bound ATP (1245 cm ؊1 ), ADP (1101 and 1205 cm ؊1 ), and free phosphate (1078 cm ؊1 ). Cleavage of the ␤-phosphate-␥-phosphate bond was found to be the rate-limiting step; no protein-bound phosphate intermediate was resolved.
ATP-binding cassette (ABC) 3 transporters form a ubiquitous family of integral transmembrane proteins responsible for the active transport of a wide variety of compounds (ranging from single ions to whole proteins) across biological membranes (1)(2)(3)(4). They share a common overall architecture consisting of two transmembrane domains (TMDs), found to be responsible for transport substrate (allocrite) specificity, and two nucleotide-binding domains (NBDs), which energize transport by ATP (substrate) hydrolysis. All conserved sequence motifs used for classification of the family are located in these NBDs. During the last decade, considerable progress has been made in the structure elucidation of full-length ABC transporters (5). Structures trapped in different states of the transport cycle support the alternating access model, in which the allocrite-binding site in the TMD is exposed sequentially to the intra-and extracellular sides. Direction of transport is determined by differences in binding affinities in the two conformations.
One subfamily of ABC transporters exports a broad range of small hydrophobic drugs across cell membranes in bacteria and eukaryotes. They can confer multidrug resistance and are therefore of large medical interest (6). The first identified and probably best studied representative of this class is human P-glycoprotein (ABCB1) (7). Its mouse homologue is the first eukaryotic ABC transporter, for which a full-length x-ray structure has been reported (8). P-glycoprotein has been trapped in the inward-facing open conformation in these crystals. Threedimensional structures for two bacterial homologues were solved previously: Sav1866 from Staphylococcus aureus has been crystallized in the outward-facing conformation (9, 10), whereas three structures of MsbA have been reported as inward-facing open (from Escherichia coli), inward-facing halfclosed (form Vibrio cholera), and outward facing (from Salmonella typhimurium) (11).
MsbA is a homodimeric transporter located in the cytoplasmic membrane of Gram-negative bacteria. It is essential for the flipping of lipid A from the inner to the outer membrane leaflet (12,13) but, in addition, exports similar xenobiotics as P-glycoprotein or another bacterial multidrug transporter, LmrA from Lactococcus lactis (14). The 65-kDa monomer is composed of an N-terminal TMD with six membrane-spanning helices and a C-terminal NBD (Fig. 1). MsbA shows significant sequence similarities to other multidrug exporters. The TMD sequence of MsbA is 30% identical and 46% similar to the TMD of human P-glycoprotein, whereas the NBD region is 51% identical and 66% similar to P-glycoprotein (15). MsbA can therefore be used as a prokaryotic model for eukaryotic ABC multidrug exporters.
Despite the large impact of the x-ray structures on our understanding of ABC transporters, they can only provide static snapshots of the transport cycle. Therefore, time-resolved techniques have to be employed to gain insight into the dynamics of ABC transporters. The conformational motion of MsbA (and the alternating access of the putative allocrite-binding chamber) during transport was investigated by electron paramagnetic resonance (16 -20) and cysteine cross-linking (21). Fluorescence quenching was used to identify allocrites (reviewed in Ref. 6) and their putative binding sites (22,23). Infrared spectroscopy with detection of secondary structure, hydrogen/deuterium exchange rates, and linear dichroism have been used to determine orientation and accessibility profiles of helices in lipid bilayers, e.g. for LmrA (24) and HorA (25).
It would be very beneficial to investigate allocrite translocation and ATP hydrolysis simultaneously with high spatiotemporal resolution. Time-resolved FTIR spectroscopy is ideally suited to elucidate molecular reaction mechanisms, as the infrared vibrations monitor the conformational and chemical state of the protein, allocrites (i.e. lipid A) or substrate (i.e. ATP), without the need of invasive labeling (26). The use of caged compounds allows the synchronization of reactions with a short laser pulse (27). This has been successfully employed for nucleotide hydrolysis in small GTPases (e.g. Ras (28), Rap (29), and Ran (30)) and ATPases (e.g. CopB (31), Na,K-ATPase (32), Ca-ATPase (33,34), or Eg5 kinesin (35)).
In this study, we employed a cysteine-free MsbA-NBD construct to obtain the first time-resolved FTIR spectra of the ATP hydrolysis reaction of an ABC transporter. Employing isotopically labeled ATP analogues, we identified the IR fingerprints of ␣-, ␤-, and ␥-phosphate groups. The kinetic data allow the direct identification of the rate-limiting steps for the catalytic cycle.
Molecular Biology and Protein Expression-Based on the fulllength MsbA protein sequence from E. coli K-12 (UniProtKB accession number P60752), a cysteine-free (C88S and C315S) codon-optimized gene was designed and synthesized (Mr. Gene GmbH, Regensburg, Germany), including a C-terminal His 10 tag and a stop codon. The gene was cloned into the pET28b expression vector (Merck) using the NcoI and BamHI restriction sites, yielding vector pFS1. Based on this full-length vector, a NBD expression construct was created using the fol- lowing primers: 5Ј-ATTCACCATGGATGGCAGAAGGCA-AACGTGTGATCGAACGT-3Ј and 5Ј-ATCCGGGATCCT-TAGTGGTGATGGTGATGATGGTGGTGGTG-3Ј (with the underlined bases indicating the NcoI and BamHI sites, the underlined boldface bases representing an inserted alanine codon, and the boldface bases indicate the annealing sequence). The PCR product was ligated into the pET28b vector, creating the MsbA-NBD expression vector pFS2. This construct includes amino acids 330 -582 of full-length MsbA with two additional residues at the N-terminal end (Met-Ala) and His 10 at the C-terminal end. The integrity of the plasmid construct was confirmed by DNA sequencing. The plasmid was then transformed into E. coli NEB5␣ cells for plasmid amplification and into E. coli ER2566 cells for protein expression. Both strains were from New England Biolabs.
Production and Purification of MsbA-NBD-For production of MsbA-NBD, 1 liter of LB medium containing 100 g/ml kanamycin was inoculated from an overnight culture to A ϭ 0.1. The main culture was incubated at 37°C at 125 rpm, and protein expression was induced by the addition of 1 mM isopropyl ␤-D-thiogalactopyranoside (Biosynth AG, Staad, Switzerland) at A 600 ϭ 0.6 -0.8. After induction, cells were cultured for 3 h, harvested by centrifugation, and stored at Ϫ20°C. Frozen cells were thawed on ice and resuspended in 30 ml of lysis buffer (50 mM Hepes (pH 7.5), 150 mM NaCl, 10 mM MgCl 2 , and 20 mM imidazole). Cells were disrupted using a fluidizer (Microfluidics) at 12,000 p.s.i. To spin down cell fragments and not disrupted cells, 35 ml of lysis buffer was added, and the lysate was centrifuged for 30 min at 10,000 ϫ g. After ultracentrifugation of the supernatant for 45 min at 100,000 ϫ g, the soluble fraction was filtered through a 0.2-m filter and applied to a 1-ml Ni-NTA (nickel-nitrilotriacetic acid) Superflow column (Qiagen). After a washing step, the protein was eluted from the column with lysis buffer containing 300 mM imidazole by gravity flow. For anion exchange chromatography, the eluate was transferred into low salt buffer (50 mM Hepes (pH 7.5), 2 mM NaCl, and 2 mM MgCl 2 ) using a PD-10 desalting column (GE Healthcare). The protein solution was loaded onto a DEAE FF 5-ml column (GE Healthcare) using an Ä KTApurifier 100 system (GE Healthcare). Bound protein was eluted via a salt gradient mixing low salt buffer and high salt buffer (50 mM Hepes (pH 7.5), 1 M NaCl, and 20 mM MgCl 2 ). Peak fractions were collected; pooled; concentrated (Vivaspin 20 concentrator with a molecular mass cutoff of 10,000, Sartorius Stedim Biotech GmbH, Göttingen, Germany) up to 10 mg/ml; desalted in 5 mM Hepes (pH 7.5), 5 mM NaCl, and 2 mM MgCl 2 ; and used for ATPase activity assays and FTIR spectroscopy.
ATPase Activity Assays-The MsbA-NBD ATPase activity was quantified by monitoring the amount of liberated inorganic phosphate using a modified procedure described previously (41). Briefly, samples of 50 g of MsbA-NBD were incubated in a volume of 100 l containing 50 mM Hepes (pH 7.5), 10 mM MgCl 2 , and 3 mM ATP at 37°C for 30 min at 750 rpm in a Thermomixer (Eppendorf, Hamburg, Germany). The reaction was started by the addition of ATP and stopped by the addition of 400 l of molybdate reagent containing 0.5% (w/v) (NH 4 ) 2 Mo 7 O 24 , 2% (w/v) H 2 SO 4 , and 0.5% (w/v) SDS. Complex formation was induced by the addition of 20 l of 10% (w/v) ascorbic acid. Non-enzymatic phosphate release was monitored with samples containing the inactive His 6 -tagged GTPase mutant Rab11a(Q70L)D43 instead of MsbA-NBD. After a 2-min incubation at 37°C, the absorbance was read at 750 nm using a NanoDrop 2000c spectrophotometer (PEQLAB Biotechnologie GmbH, Erlangen, Germany). All values are triplicate determinations. In addition, we used an HPLC-based ATPase activity assay in which the ATPase reaction was stopped by shock-freezing the samples in liquid nitrogen. For Michaelis-Menten kinetics, substrate concentrations from 0.2-4.0 mM ATP were used. Data were analyzed using either Equation 1, assuming no cooperative effects during hydrolysis, or Equation 2, which assumes cooperativity. V and V max are velocity and maximal velocity of the reaction, respectively.
[S] denotes substrate concentration, K m and K 0.5 are concentrations where V is at 50% of V max , and H is the Hill coefficient. All measurements were done in triplicates. FTIR Measurements-FTIR measurements were performed in 125 mM Hepes, 125 mM NaCl, and 50 mM MgCl 2 at pH 7.5 at a protein concentration of 8.6 mM. The band assignment experiments were performed with npeATP as the caged compound in the presence of 20 mM DTT. Kinetic experiments were performed with the faster photolyzing phpATP. Caged ATP and protein were used at a 1:1 molar ratio to guarantee single-turnover conditions. All preparations were done under red light due to the photo instability of the caged compounds.
The samples were prepared between two CaF 2 windows with a 20-mm diameter and 2-mm thickness. One window possessed a concentric groove with a depth of 0.2 mm and a width of 2 mm, surrounding a deepened area 10 mm in diameter. The resulting sample thickness was 7 m. A thin 2-3-mm wide silicon grease film was applied on both window rims. Protein solution and the caged ATP were applied onto the inner area of the grooved window. The sample was concentrated in a nitrogen stream to a volume of ϳ0.8 -1 l and subsequently covered with the second window. The assembly was fixed in a blackened metal cuvette, which was then mounted in the spectrometer. All spectra were recorded at 293 K. At higher temperatures, the intrinsic hydrolysis rate is accelerated, which prevents base-line distortions, but the protein becomes less stable. The spectra were taken between 1900 and 900 cm Ϫ1 .
Photolysis was performed with an LPX 240 XeCl excimer laser (Lambda Physik AG, Göttingen, Germany) at 308 nm. The laser energy was 130 -140 mJ per flash with a pulse duration of 20 ns. Twelve or 40 flashes at 500 Hz were used to achieve conversion of caged ATP (phpATP or npeATP, respectively).
After sample equilibration for 2 h at 293 K, FTIR spectra between 1900 and 900 cm Ϫ1 were recorded on an IFS80v or IFS66v/s spectrometer (Bruker) equipped with a liquid nitro-

Time-resolved FTIR Spectroscopy of MsbA-NBD
gen-cooled mercury cadmium telluride detector. The spectral resolution was 4 cm Ϫ1 using an aperture of 4 -5 mm, a doublesided forward-backward data acquisition mode, and an instrumental scanner speed of 320 kHz. Before photolysis, a reference spectrum was taken by averaging 400 scans. After the laser flashes, which started the reaction, data were collected for 15 min. For the averaged interferograms, a zero filling factor of 2 and the Blackman-Harris three-term apodization function were used. For Fourier transformation, the Mertz phase correction was applied. Absorbance spectra and absorbance time courses (A(t)) were calculated from the single-channel spectra (I(t)) according to A(t) ϭ Ϫlog (I(t)/I 0 ), where I 0 is the singlechannel spectrum before photolysis at t ϭ 0. The kinetics of spectral evolution between 1800 and 950 cm Ϫ1 were approximated to multiple exponentials in the period of 1 ms to 700 s by global fit analysis (42), in which the absorbance change (⌬A(,t)) is fitted with a sum of n r exponentials with apparent rate constants (k l ) and amplitudes (a l ) according to Equation 3.
In the amplitude spectrum a l , the disappearing bands face downwards, and the appearing bands face upwards. The amplitude spectrum a 0 compares the state before the triggering of the reaction by the irradiation with the state after the flashes, extrapolated to time t ϭ 0.

RESULTS
Cloning, Expression, and Purification-On the basis of the structure and sequence of MsbA (Protein Data Bank code 3B60 (11)), we designed several constructs for the isolated NBD. A soluble MsbA-NBD construct starting at amino acid 330 of fulllength MsbA with a His 10 tag at the C terminus led to good expression and purification (Fig. 2). A clear band for MsbA-NBD could already be assigned in the homogenate fraction and was enriched in the supernatant of an ultracentrifugation step.
Pure MsbA-NBD could be eluted from the Ni-NTA affinity column. Remaining contaminants were eliminated with an additional anion exchange step. We can now routinely produce highly purified MsbA-NBD with a yield of 3 mg of protein from 1 liter of culture.
Catalytic Activity-MsbA-NBD catalytic activity was assayed with both a colorimetric and a HPLC-based assay for different substrate concentrations. A typical Michaelis-Menten plot is shown in Fig. 3, resulting in a K m of 1.31 mM and an ATP catalytic activity V max of 56 nmol mg Ϫ1 min Ϫ1 . NBDs of ABC transporters have been shown to dimerize upon ATP binding (43,44) and have to interact for ATP hydrolysis (45). Therefore, we also checked for cooperativity by fitting the data with the Hill equation (Equation 2 and Fig. 3). This resulted in V max ϭ 45 nmol mg Ϫ1 min Ϫ1 , K 0.5 ϭ 0.85 mM, and a Hill coefficient of 1.49, thus demonstrating positive cooperativity of MsbA-NBD in ATP binding and hydrolysis.
FTIR Spectroscopy-Because of the relatively slow turnover of MsbA-NBD, we employed fast scan FTIR in transmission mode, which provides millisecond time resolution (46). Concentrated samples were mounted between CaF 2 windows. High quality difference spectra were recorded between 1900 and 900 cm Ϫ1 at 293 K. To initiate the hydrolysis reaction, caged phpATP (supplemental Fig. 1) was used, which liberates ATP after flash photolysis with a rate in the sub-microsecond time range below the resolution of the fast scan technique. The global fit analysis of the absorbance changes between 1800 and 950 cm Ϫ1 according to Equation 3 with two rate constants (n r ϭ 2) resulted in a good agreement with the data, and we obtained the apparent rate constants k 1 ϭ 1.4 Ϯ 1.2 s Ϫ1 and k 2 ϭ (10.8 Ϯ 2.0) ϫ 10 Ϫ3 s Ϫ1 and the amplitude spectra a 0 , a 1 , and a 2 (Fig. 4). In these amplitude spectra, positive bands are indicative of appearing absorbance, whereas negative bands represent loss of absorbance. In the initial amplitude spectrum a 0 , the dominating negative band at 1264 cm Ϫ1 represents the removal of the caged group after the laser flashes. It represents mainly the changes induced by photolysis of the caged compound. The absorbance changes in a 1 in the phosphate region between 1300 and 950 cm Ϫ1 reflect a rearrangement of ATP and the binding niche after photolysis. The amplitude spectrum a 2  shows clear marker bands of ATP hydrolysis and phosphate release. How these bands were assigned will be discussed next.
Phosphate Band Assignment-The asymmetric stretching vibrations of the phosphate groups of ATP and ADP are usually found between 1100 and 1300 cm Ϫ1 and possess a large absorption coefficient. For the assignment of these vibrations, ␣-, ␤-, or ␥-18 O-labeled caged ATP was used (Fig. 5). The increased mass of the substituted isotope shifts the corresponding IR band of the marked residue to lower wave numbers. Bands were normalized to the bands between 1320 and 950 cm Ϫ1 . For assignment of the phosphate vibrations of NBD-ATP, NBD-ADP, and the released P i , the a 2 hydrolysis spectra of unlabeled ATP and isotope-labeled ATP were compared (Fig. 5A), and double-difference spectra were calculated (Fig. 5B). The bands of ATP are negative due to its decrease in the hydrolysis reaction, whereas the bands of hydrolysis products like ADP and P i are positive in a 2 . Isotopic labeling of [␣-18 O 2 ]ATP results in a downshift of the ATP educt band (red arrow) from 1245 to 1219 cm Ϫ1 , whereas the emerging ADP product band (blue arrow) is shifted from 1205 to 1181 cm Ϫ1 . Thus, we can assign the absorption at 1245 cm Ϫ1 to as (␣-PO 2 ) of ATP bound to MsbA-NBD and the absorption at 1205 cm Ϫ1 to as (␣-PO 2 ) of ADP (spectrum I in Fig. 5, A and B). The hydrolysis of [␣,␤-18 O/␤-18 O 2 ]ATP induces a shift of an educt band from 1217 to 1188 cm Ϫ1 , which is marked in blue and assigned to as (␤-PO 2 ). The original band of unlabeled ATP is masked by a shift of the 1245 cm Ϫ1 band (red) to this position as indicated by the dotted lines in the double-difference spectrum. This band was already assigned to as (␣-PO 2 ) above and indicates some coupling between the asymmetric PO 2 stretching vibration of ␣and ␤-phosphates (spectrum II in Fig. 5, A and B). An ADP product band (orange) is shifted from 1101 to 1083 cm Ϫ1 and assigned to one of the two as (␤-PO 3 ) vibrations. The second as (␤-PO 3 ) vibration (green) is found at 1142 cm Ϫ1 . Here, the shifted band  Amplitude spectra a 0 , a 1 , and a 2 . Spectra are the averages of four single measurements. Infrared absorption of the protein at higher wave numbers is separated from the lower phosphate absorption region by the dotted line.

Time-resolved FTIR Spectroscopy of MsbA-NBD
is again masked as indicated by the dotted lines in the doubledifference spectrum. Using [␤,␥-18 O/␥-18 O 3 ]ATP as a substrate, the ATP educt bands shift from 1160 cm Ϫ1 (red) and 1132 cm Ϫ1 (blue) to 1120 and 1099 cm Ϫ1 , respectively, and are assigned to as (␥-PO 3 ) (spectrum III in Fig. 5, A and B). A band at ϳ1078 cm Ϫ1 has previously been assigned to free P i (31,47,48). This assignment is confirmed by a shift from 1078 to 1047 cm Ϫ1 (green). The less obvious band shift from 994 to 962 cm Ϫ1 (orange) can also be assigned to the release of P i .
A similar analysis for the a 1 double-difference spectra is hampered by a partial mixing of a 0 and a 1 due to the relatively slow photolysis kinetics of the 1-(ortho-nitrophenyl)ethyl group used for the isotopically labeled ATP. Therefore, analysis was performed on the combined a 0 ϩ a 1 spectra (supplemental Figs. 2 and 3). In these spectra, contributions of caged ATP are included as negative peaks, whereas NBD-ATP bands will appear as positive bands. Most of the key features of NBD-ATP can also be assigned in this spectrum. At 1216 cm Ϫ1 , a band for the ATP as (␤-PO 2 ) vibration can be identified, which is masked in the hydrolysis spectrum.
FTIR Kinetics-The assignment of specific phosphate marker bands allows characterization of the two rate constants in relation to hydrolysis. The well characterized free phosphate band at 1078 cm Ϫ1 clearly is related only to k 2 , which therefore represents the P i release (Fig. 6). Because the marker bands for ATP at 1245 cm Ϫ1 (␣ and ␤) and 1132 cm Ϫ1 (␥) also vanish with k 2 , bond breakage between ␤and ␥-phosphates occurs with the same rate, indicating this to be the rate-limiting step for hydrolysis.
In the protein region (1720 to 1350 cm Ϫ1 ) of the amplitude spectra associated with k 1 and k 2 , several strong absorption bands are visible. The near-mirror symmetry is indicative of a reversible conformational change in the protein during the catalytic cycle (Fig. 7A). All major peaks were evaluated with respect to their dominating contributions to either k 1 or k 2 . Notably, most of the changes indeed correlate with these two rates. As a typical example, the kinetic analysis of the band at 1520 cm Ϫ1 is shown in Fig. 7B. Absorption increases with a post-photolysis/pre-hydrolysis rate of k 1 ϭ 1.4 s Ϫ1 and disappears with a rate of k 2 ϭ 10.8 ϫ 10 Ϫ3 s Ϫ1 . Nevertheless, slight deviations from the two-rate fit are visible, pointing to additional rates involved in conformational changes of the protein.
In addition, we observed absorption changes in the protein region already in a 0 , which represent initial nucleotide binding after ATP release. However, our current data do not allow a reproducible fit of these early events, which might be resolved with higher time resolution by utilizing the experimentally more challenging step scan technique (49). In summary, the protein region of the infrared spectra contains detailed information about residues within the NBD protein that undergo significant reversible conformational changes during the catalytic cycle.

DISCUSSION
Catalytic Activity-ABC transporters couple ATP hydrolysis in the motor domain with allocrite translocation in the TMD. Although the ultimate goal is to study the complete transport process in the holoprotein, a rational approach is to first characterize the ATP hydrolysis reaction in the isolated NBD. As the NBDs in ABC exporters are usually fused to the TMD, expression constructs of isolated NBDs frequently suffer from poor protein yield and solubility. Correct definition of domain  boundaries has been shown to be crucial for a successful study (50). Although our first attempts to overexpress the isolated NBD of MsbA were based on erroneous crystallographic models, boundaries defined on the corrected model (11) resulted in a construct with a good expression profile and satisfactory protein solubility up to 15 mg/ml in buffer. Purified substrate-free MsbA-NBD elutes as a monomeric enzyme during size exclusion chromatography (data not shown). The ATPase activity falls within the lower range of reported activities of the fulllength transporter (22,(51)(52)(53)(54). MsbA is found to be 2-4-fold stimulated by the binding of the allocrite lipid A to the transmembrane region, which is not included in our MsbA-NBD construct (19,51,53). A Hill coefficient of 1.5 indicates slight cooperativity in ATP binding and is consistent with values for the isolated NBD of the hemolysin transporter HlyB (43) and the mitochondrial transporter Mdl1p (44). These data therefore clearly demonstrate that MsbA-NBD represents a functional ABC motor domain.
Nucleotide Interaction with MsbA-NBD-We used the MsbA-NBD protein to study the time course of ATP hydrolysis by FTIR spectroscopy. Utilizing caged ATP to trigger the reaction, the kinetic data can be fitted with two major rate constants after the fast (and not experimentally resolved) photolysis of the cage group. During the first experimentally resolved post-photolysis/pre-hydrolysis rate, a hydrolysis-competent conformational state is formed. This step includes formation of the canonical head-to-tail NBD dimer (Fig. 1). The second, much slower rate represents decrease in ATP, increase in proteinbound ADP, and release of free phosphate and thereby monitors the hydrolysis. On the basis of the assignment of the phosphate bands, we can draw conclusions on the nature of the bound nucleotide states: the amplitude spectra (Figs. 4 and 6) and kinetic data clearly show the formation of an ATP-enzyme reaction intermediate in the pre-hydrolysis reaction with rate k 1 . A similar E⅐S reaction intermediate has been observed in the case of P-glycoprotein (55). In the post-hydrolysis phosphate state (bands facing upwards in Fig. 4), the dominant band at 1101 cm Ϫ1 represents the asymmetric ␤-PO 3 2Ϫ vibration, which is shifted 11 or 20 cm Ϫ1 with respect to free ADP or MgADP, respectively (56). A very similar shift has been found in the case of protein-bound GDP: for free GDP in solution, the asymmetric ␤-PO 3 2Ϫ vibration has been measured at 1115 cm Ϫ1 , which also is shifted by 11 cm Ϫ1 for Ras-bound MgGDP to 1104 cm Ϫ1 (47). Therefore, the majority of the hydrolysis product ADP remains bound to MsbA-NBD under the singleturnover conditions employed in our experiment. This is consistent with the observation that MsbA binds both ATP and ADP with similar dissociation constants of 1-2 mM (53). MsbA-NBD saturated with ADP is again monomeric in solution, as indicated by gel filtration experiments (data not shown).
Protein Dynamics-The kinetic data in the protein region also are dominated by the two rates k 1 and k 2 . Interestingly, the corresponding amplitude spectra are nearly mirror-symmetric (Fig. 7), indicating a mostly reversible induced fit upon MgATP-protein intermediate formation. At the same time, this implies that post-photolysis and post-hydrolysis protein conformations are largely identical but are different from the reaction intermediate. Based on the analysis of NBD structures in different complexes, motion of the helical domain upon formation of the hydrolysis-competent, MgATP-bound conformation has been postulated (57,58). In contrast, the ADP-bound structure largely resembles the nucleotide-free state. In these investigations, the conformation of MgATP-NBD had to be inferred either from hydrolysis-impaired mutant proteins or from structures with bound ATP analogues. Here, we trace these conformational transitions for the first time in a catalytically intact NBD. Our time-resolved data clearly identify the protein-bound MgATP intermediate appearing with rate k 1 and monitor the resetting of the protein concomitant with the hydrolysis reaction with rate k 2 .
Catalytic Cycle-We can combine these data into the scheme shown in Fig. 8. ATP released during photolysis binds to monomeric NBD. During the experimentally resolved apparent rate k 1 , ATP binding triggers the rotatory movement of the helical domain and subsequent dimer formation, thereby generating the full ATP-binding pocket. These protein changes in k 1 most likely represent the power stroke of transport. In follow-up experiments, site-directed mutagenesis will be used to decode these IR absorbance changes to elucidate the mechanistic details of this power stroke. Hydrolysis, release of free phosphate, and dimer dissociation can be modeled with k 2 . Our data do not allow the discrimination between sequential or concomitant ATP hydrolysis in both binding sites. However, no ATP remains in the sample after hydrolysis as indicated by HPLC (data not shown). The hydrolysis product ADP remains mostly FIGURE 8. Scheme of ATP hydrolysis in MsbA-NBD. After photolysis of caged ATP (cg-ATP), free ATP binds to the monomeric NBD. With rate k 1 , domain movement of the helical domain of the NBD and dimerization lead to formation of the two catalytic binding sites. This NBD dimer is similar to the conformation in the AMP-PNP-bound full-length structure in Fig. 1. With rate k 2 , hydrolysis in both catalytic sites, either consecutive or parallel, results in a NBD-ADP complex, which dissociates to monomeric NBD-ADP. The cycle is halted under our single-turnover conditions. Under physiological conditions, nucleotide exchange would immediately reset the NBD to the ATP-bound monomer.
bound to the protein as indicated by the specific infrared marker band appearing with rate k 2 .
These results are in line with the processive clamp model put forth on the basis of biochemical and structural work on isolated NBDs (44,59). Here, the power stroke is facilitated by the dimerization of the NBDs, similar to the NBD switch models generally accepted for ABC importers. Although ATP hydrolysis in both sites is proposed to be sequential, only very small conformational changes between the two hydrolysis events are predicted (59), thereby resulting in one apparent rate constant. Finally, the ADP-bound dimer dissociates, and nucleotide exchange of two ADPs with ATP resets the transporter.
Hydrolysis Limits MsbA-NBD Turnover-One important factor for understanding a catalytic cycle is the nature of the rate-limiting step. Studies on isolated NBDs of two homodimeric ABC exporters led to diverging results. For the hemolysin exporter HlyB, hydrolysis was found to be rate-limiting (60), whereas for the peptide exporter Mdl1, dimerization was the rate-limiting step (61). In both studies, the authors had to modify the catalytic cycle either by using partially inactive mutants or by substituting metal cofactors to gain information about the intermediate catalytic steps.
In our experiments, we can monitor the sequence of events for the first time online in the intact protein with the natural substrate MgATP. The experimental identification of individual spectral marker bands for the most important reaction intermediates (discussed above) allows us to unambiguously assign the ATP binding and dimerization to the fast first rate, whereas the hydrolysis and phosphate release are clearly assigned to the second rate, which is 2 orders of magnitude slower. We therefore propose that hydrolysis is the rate-limiting step in isolated ABC transporter motor domains.
It will be essential to extend our approach to the full-length transporter. Understanding the exact coupling of NBD dynamics to allocrite binding and translocation in the TMD remains the Holy Grail of ABC transporter research. Clarification will require more time-resolved FTIR studies on intact full-length transporters. To this end, we are currently working on investigation of full-length MsbA in a lipid environment.