Transmembrane and Juxtamembrane Structure of αL Integrin in Bicelles

The accepted model for the interaction of α and β integrins in the transmembrane (TM) domain is based on the pair αIIbβ3. This involves the so-called outer and inner membrane association clasps (OMC and IMC, respectively). In the α chain, the OMC involves a GxxxG-like motif, whereas in the IMC a conserved juxtamembrane GFFKR motif experiences a backbone reversal that partially fills the void generated by TM separation towards the cytoplasmic half. However, the GFFKR motif of several α integrin cytoplasmic tails in non-bicelle environments has been shown to adopt an α-helical structure that is not membrane-embedded and which was shown to bind a variety of cytoplasmic proteins. Thus it is not known if a membrane-embedded backbone reversal is a conserved structural feature in α integrins. We have studied the system αLβ2 because of its importance in leukocytes, where integrin deactivation is particularly important. Herein we show that the backbone reversal feature is not only present in αIIb but also in αL-TM when reconstituted in bicelles. Additionally, titration with β2 TM showed eight residues clustering along one side of αL-TM, forming a plausible interacting face with β2. The latter orientation is consistent with a previously predicted reported polar interaction between αL Ser-1071 and β2 Thr-686.


Introduction
Integrins are heterodimeric type I transmembrane (TM) proteins formed by non-covalent association of an a and a b subunit. In humans, 18 a-chains can interact with 8 different bchains to form 24 different a/b heterodimers [1]. Each subunit contains a large extracellular domain, a single membranespanning a-helix and a short cytoplasmic tail [2]. By spanning the membrane, integrins serve as a dynamic linkage between cytoplasm and extracellular space, transducing signals across the membrane to mediate various functions [3]. Information transduction in integrins is bidirectional, inside-out or outside-in. Inside-out signaling involves separation of integrin cytoplasmic tails and TM domain, and subsequent ectodomain conformational changes, which alter the affinity of integrins for extracellular ligands [4][5][6]. Deletions or mutations that are expected to destabilize a/b association of TM or cytoplasmic domains in the a and b subunits, have been shown to activate integrins [7][8][9]. Conversely, inactivation is achieved by enhancing heterodimerization, e.g., using intermolecular disulfide bonds; in fact disulfide scanning of the aIIbb3 TM region revealed the TM heterodimer interface [10].
Two models of heterodimeric a/b TM interaction were proposed based on electron microscopy and a computational approach showing different crossing angles for 'resting' and 'activated' states [5,11]. For the first model, interaction was mediated via a GxxxG-like motif located in both a and b TM chains. This proposed mode of interaction is reminiscent of that found previously for the homodimer glycophorin A (GpA) [12]. For the second proposed mode of interaction [11,13], the b helix is rotated by approximately 90u from its orientation in the GpA-like model. This latter model has been validated for aIIbb3 using activating mutations at the transmembrane domain [14], cysteine cross-linking of the full integrin [10,15], and NMR studies of the aIIbb3 transmembrane heterodimer structure in bicelles [16]. Thus, this is currently accepted to represent the interaction of the two TM domains of aIIb and b3 in the resting state.
This accepted model has been further refined in the case of aIIbb3 [15][16][17]. Specifically, the two TMs are in close proximity at the extracellular half of the membrane, forming an outer membrane association clasp (OMC) that involves the GxxxG-like motif in aIIb, Gly-972xxxGly-976, and Gly-708 in b3. The TMs separate towards the cytoplasmic half, and the void is partially filled with an inner membrane association clasp (IMC), formed by a juxtamembrane backbone reversal in aIIb that participated by two phenylalanines, Phe-992 and Phe-993. The latter are part of a conserved a chain GFFKR motif. Lastly, a salt bridge between Arg-995 of aIIb and Asp-723 of b3 has also been identified [16]. At present, no other detailed structural model for other integrin TMs is available.
In contrast to the above model for the juxtamembrane backbone reversal, the GFFKR motif in several a integrins e.g., in aIIb, a1, aM, aL, aX or a4 [18][19][20][21][22][23][24], formed an a-helical structure that was not membrane-embedded. However, most of these reports included the cytoplasmic tail without the TM domain, although in some cases they were myristoylated and attached to detergent micelles. Other reports included the TM domain, but the structure was determined in the presence of organic solvents or in presence detergent micelles. Thus, it is possible that the failure to observe such reverse turn in the above examples was due either to the absence of the TM domain, or to the fact that detergent micelles or organic solvents do not provide the appropriate environment for such structural feature.
Although the membrane-embedded reverse backbone has been observed in the presence of the TM domain and lipids, whether in bicelle systems or membranes [15,16], it has only been reported for aIIb integrin. Therefore, observing a membrane-embedded reverse turn in an a integrin-TM different from aIIb would support that this structural feature is conserved, and that it represents one of the possible conformations of the GFFKR motif in all integrins.
We have chosen to use the system aLb2 because of its importance in leukocytes, where it is involved in adhesion, migration, cytotoxicity, proliferation, and antigen presentation [34]. The importance of the GFFKR motif in aL has been demonstrated by looking at the effect of deletion of this motif in a transgenic mouse, which led to a constitutive aLb2-mediated cell adhesion [35]. Additionally, in these fast-migrating cells, integrin deactivation is particularly important, and we proposed previously that the latter could be facilitated by hydrogen bond interaction between TMs [36][37][38]. Thus, to elucidate if the reverse turn at the GFFKR motif is also found in aL when the peptide is reconstituted in bicelles, and to delineate the aL interfacial residues involved in aL/b2 interaction, we have expressed, purified and determined the structure of aL TM in bicelles by solution NMR. Further, titration experiments with non-labeled b2 TM show which residues in aL are most affected by this interaction.

Experimental Procedures
Integrin aL TM and b2 TM Expression and Purification The DNA sequence encoding the TM region of human aL integrin, residues 1057-1103 (Fig. 1A), was cloned into a pET-28b expression vector, whereas the sequence for human TM b2 integrin, residues 675-710, was cloned into pNIC28-Bsa4. G685F and G689F mutations at in the GxxxG-like region of b2 TM were subsequently introduced by site-directed mutagenesis with the relevant pair of primers. All constructs were expressed with an Nterminal 6-His tag, cleavable by tobacco etch virus (TEV) protease.
The plasmids were transformed into E. coli strain BL21 codonplus for protein expression. For expression of non-labeled peptides, the culture was grown in terrific broth (TB) media at 37uC until an OD 600 of 2. Protein expression was induced overnight at 25uC by adding 1 mM IPTG. The cells were harvested by centrifugation at 75006g and stored at 280uC.
For expression of stable isotope-labeled peptides, the culture was initially grown in LB media at 37uC until an OD 600 of 0.7. The cells were collected and resuspended into M9 minimal media at 25% of the initial volume, to achieve a high-density culture as described previously [39,40]. The media was supplemented with 15 N-NH 4 Cl, 13 C-glucose, 13 C/ 2 H-glucose, or D 2 O (Cambridge Isotope Laboratories), according to the desired labeling scheme: 15 N-labeled, 15 N/ 13 C-labeled, and 15 N/ 13 C-labeled with partial (50%) and complete (99%) deuteration. Cultures were further grown for 1 h before inducing protein expression with 1 mM IPTG at 25uC overnight. Cells were harvested as mentioned above and stored at 280uC.
Cell pellets were resuspended in lysis buffer containing 20 mM Tris pH 8.0, 300 mM NaCl, 5 mM imidazole, 2 mM b-mercaptoethanol and 10% glycerol. Complete lysis was achieved by adding 1.5% Triton X-100, followed by sonication and microfluidization. The crude cell lysate was clarified by centrifugation at 40,0006g and applied onto pre-equilibrated Ni-NTA resin (Bio-Rad Profinity IMAC Ni 2+ -charged). The resin was washed with the 30 column volume (CV) of washing buffer A and 3 CV of washing buffer B. Washing buffer A consists of lysis buffer supplemented with 25 mM imidazole while washing buffer B consists of washing buffer A supplemented with 1.8 mM Ndodecyl-b-D-maltopyranoside (DDM). The bound peptide was cleaved on-column by adding TEV protease with 3 hours incubation period at room temperature. Cleaved peptide was collected, TCA-precipitated and lyophilized. The peptides were extracted by adding methanol to the lyophilized powder and removing the precipitates by centrifugation.
Peptide identity and purity was assessed by SDS-PAGE and MALDI-TOF MS ( Fig. S1 in File S1). When deemed necessary, the peptides were further purified by using reversed-phase HPLC. Purification was achieved by applying isopropanol-acetonitrile (80/20 (v/v) with 0.1% TFA) linear gradient on a Phenomenex Jupiter C18 semipreparative column (250610 mm, 300 Å pore size, 5 mm particle size). The peptides were lyophilized and stored in 220uC.
NMR Spectroscopy -NMR experiments were performed at 30uC using Bruker Avance-III 800 and Avance-II 700 NMR spectrometers with cryogenic probes. Sodium 2,2-dimethyl-2-silapentane-5sulfonate (DSS) was used as the internal reference for 1 H nuclei. The chemical shifts of 13 C and 15 N nuclei were calculated from the 1 H chemical shifts. The NMR data were processed using TopSpin 3.1 (www.bruker-biospin.com) and analyzed using CARA (www. nmr.ch). Sequence-specific assignment of backbone 1 H N , 15 N, 13

Structure Calculation
NOE distance restraints to calculate the structures of aL TM domain were obtained from 15 N-NOESY-HSQC and 13 C-NOESY-HSQC spectra, respectively. Backbone dihedral angle restraints (Q and y) were derived from 13 C', 13 C a , 13 C b , 1 H a and 1 H b chemical shift values using TALOS [44]. The short-range and medium range NOE connectivities were used to establish the sequence-specific 1 H NMR assignment and to identify elements of the regular secondary structure. Structure calculations were performed using CYANA 3.0 [45,46] and visualized using MOLMOL [47] and PyMOL (Delano Scientific). CNS 1.3 [48,49] was used to refine the structure using the standard simulated annealing protocol. A total of 20 structures have been calculated and the structure statistics are summarized in Table S1 in File S1. The structure of integrin aL TM domain has been deposited on the Protein Data Bank (PDB) with ID 2m3e, and the assigned chemical shifts have been deposited on the Biological Magnetic Resonance Bank (BMRB) with ID 18958.

Structure of the aL TM Domain in Phospholipid Bicelles
The structure of bicelle-reconstituted integrin aL TM domain (aL) was determined by using a set of 2D and 3D NMR experiments on the 15 N/ 13 C-labeled peptide with varying deuteration levels. Upon reconstitution into DHPC-DMPC bicelles, aL forms a linear a-helix which extends approximately from Leu-1065 to Val-1089 (Fig. 1A-B and Fig. S2 in File SS). This a-helix comprises ,50% (25 residues) of the whole peptide. This is significantly lower than ,80-90% a-helix estimated by FTIR when this peptide was reconstituted in lipid bilayers (Fig.  S3A in File S1), but the value is comparable to ,60% (30 residues) estimated from CD when the peptide was reconstituted in bicelles ( Fig. S3B in File S1).
The conserved GFFKR motif follows immediately the a-helix, with Phe-1091 and Phe-1092 reinserting into the bicelle (Fig. 1B-C) and forming a backbone reversal, only observed previously in aIIb in bicelles [16]. The structure is consistent with Phe-1091 making hydrophobic contacts to Leu-1086 and Val-1089, whereas the side-chain of Phe-1092 packs against Phe-1083 and Leu-1086, as indicated by NOE connectivities between these residues (Fig. 1C and Fig. 2). Thus Phe-1083, Leu-1086, Phe-1091 and Phe-1092 constitute a hydrophobic surface which potentially forms an interaction site with b2-TM. The superposition of our aL-TM model and that reported for aIIb-TM shows high structural similarity between the two proteins (Fig. 1D). This strongly suggests that the backbone reversal of the GFF motif is a characteristic structural feature of a integrin subunits when reconstituted in bicelles. Alignment of the rest of the TM domains produced an aC RMSD of 0.83 Å . In comparison, the aC RMSD with a recently publised a1 TM structure in micelles for which no reverse turn was observed [24] (Fig. 3A). This number is considerably lower than the 35 exchange-protected residues estimated from FTIR using the peptide reconstituted in lipid bilayers (see IR spectra before and after exchange in Fig. S3C in File S1).
The dynamic properties of aL were studied using a steady-state [ 1 H, 15 N] heteronuclear NOE experiment. The majority of the residues correspond to a well-folded structure, with the central part, Val-1059 to Asn-1095, being the most stable (Fig. 3B). The backbone RMSD of the central TM region from a set of 20 conformers is 0.23 Å , indicating a well-defined structure. On the other hand, the N and C termini, especially the latter, are highly dynamic regions (see also Fig. 1B).

Interaction between the TMs of aL and b2
To determine what residues in aL mediate interaction with b2, we titrated unlabeled b2 TM into 15 N-labeled aLTM, and monitored the changes in the [ 1 H-15 N]-HSQC spectra. Qualitative examination of the spectra at a 1:2 aL/b2 molar ratio revealed eight residues with pronounced intensity reduction upon addition of b2 peptide, namely Leu-1067, Tyr-1068, Ser-1071, Gly-1072, Gly-1074, Gly-1075, Tyr-1087 and Phe-1091 (Fig. 4A). Met-1064 also showed intensity reduction, albeit less obvious, whereas Asn-1095 experienced a chemical shift change. Intensity of these resonances was reduced with increasing b2 peptide, until  the signal disappeared (Fig. 4B). A summary of these changes for the a/b molar ratio 1:2 is shown in Fig. 4C. The same results were observed when the titration was performed with a b2 TM mutant where the 'G' residues in the GxxxG-like motif were changed to Phe, i.e., G685F and G689F (data not shown). The latter confirms that b2 does not interact with aL via its GxxxG-like motif, consistent with the reported aIIb/b3 model of interaction for resting integrins [10,16,50]. The affected residues cluster on one side of the aL-TM, with residues Ser-1071 and Gly-1075, located at positions 1 and 5 in the OMC region GxxxG-like motif, having the same orientation as the residues involved in the IMC (Fig. 5). At the same time, these locations are separated by a stretch on non interacting residues, consistent with the TM separation previously observed for the aIIb/b3 system (see above). Thus, the putative IMC and OMC regions of aL TM are oriented to form a plausible interacting face with b2 TM.
The presence of ab heterodimerization in our titration is supported by comparison of the TROSY-HSQC spectrum of 0.1 mM aL with that of 0.6 mM concentration obtained in identical conditions shows that increasing aL concentration does not lead to intensity reduction of the peaks (Fig. S4 in File S1). Therefore, the peak intensity reduction observed in our titration experiment is not caused by aL homodimerization. Also, membrane-mimicking screening studies [51] show a predominant presence of ab heterodimers in a 1:1 integrin ab-TM mixture at 0.1 mM, identical concentration to that used here, or even at 0.01 mM. In the latter report, in contrast, is was shown that homomeric species increase in DPC and SDS micelles, and also in organic solvents [51].

Importance of Environment for Integrin TM Reconstitution
Phospholipid bicelles have been successfully used for structural determination of membrane proteins by solution NMR [52], especially with bicelles of low q-ratios. The inclusion of a small proportion of long-chain lipid makes this system more membranelike than micelles, yet still reasonably small for solution NMR studies. The better suitability of bicelles, relative to micelles, has been demonstrated in the study of the integrin aIIbb3 TM heterodimer in bicelles [16,51]; in micelles, these two TMs were not found to heterodimerize [51,53]. In addition, the unique backbone reversal for aIIb, was reported in bicelles [17] and full length integrin in biological membranes [15]. In presence of detergent micelles or organic solvents, the region encompassing the GFFKR motif adopted an a-helical conformation in aIIb and several other integrins [18][19][20][21][22][23]54], even when the TM domain was also present (e.g., [24,54]). Thus, not only the reconstitution environment may play a critical role for the observation of this reverse turn, but cytoplasmic peptide length and proper conformation of the TM region may be important. Indeed, the reverse turn formed by the GFFKR motif makes contacts with several side-chains in the a TM domain, as we (this paper) and others [16] have observed. Herein a hydrophobic packing is proposed in the aL-TM for the reinserted Phe-1091 and Phe-1092 against Phe-1083 and Leu-1086. The titration data also suggests that Phe-1091 is strongly involved in the interaction with b2 TM, which indicates that the hydrophobic surface formed by Phe-1091 and Phe-1092, Phe-1083 and Leu-1086 is important in the interaction with b2 TM.
We also note that in lipid bilayers (FTIR experiments) 14 more residues were calculated to be protected from H/D exchange relative to the bicelle system (NMR). This higher number is unlikely to only originate from the inevitably larger error incurred when measuring band areas in IR. Rather, these differences must be of experimental origin; in the NMR experiment, bicelles that have a high curvature are re-solubilized in D 2 O. In contrast, in the FTIR experiment, preformed planar lipid bilayers are exposed to D 2 O-saturated air. Whether or not the reverse turn is present in the FTIR experiment is not known, but even considering it is present and completely protected from H/D exchange, this is not sufficient to account for these results. We propose that the extra protection to H/D exchange in membranes is due to the formation of a larger proportion of a-helix in extramembrane domains. Indeed, a higher helical content was observed in the FTIR studies performed in membranes (see above), compared to NMR and CD techniques, where bicelles were used. The increased a-helical content in extramembrane regions in supported lipid bilayers relative to the solution NMR conditions may arise from a far more dehydrated environment in the former. Tight binding of these a-helical domains to the membrane may explain the increased H/D exchange protection. These regions then become unbound, disordered and exposed to solvent when reconstituted in bicelles. The above results suggest that data obtained in organic solvents, micelles or bicelles probably underestimates the helical content and membrane binding behavior of juxtamembrane domains, when compared to lipid bilayers. Indeed, in the present work even the reverse turn formed by the GFFKR motif is not totally protected against H/D exchange in bicelles: Phe-1092 is protected, but Gly-1090 and Phe-1091 are not (Fig. 3A), although as discussed above this may arise from experimental protocol differences. In aIIb-TM, both of the equivalent Phe residues, showed little broadening after exposure to Mn 2+ EDDA2 and H N -H 2 O chemical exchange [17], consistent with some degree of protection. Differences in methodology might also account for the shorter membraneembedded region for aL, relative to aIIb, determined in the present paper.

Structure of aL in Bicelles
The results of aL structure calculation show that the bicelleembedded aL peptide adopts a well-defined, highly a-helical conformation at the membrane-spanning region, as expected from a single-pass TM domain. The TM boundaries we found here are generally in agreement with previous studies [17,54,55]. The rest of the aL structure is remarkably similar to aIIb, especially with regards to the two proposed interaction points with the b2 chain at the two ends of the TM domain, i.e. the OMC and IMC regions.
In the OMC region, both alternative models proposed for a-b heterodimer interaction predict the small residues in the a TM GxxxG-like motif to face the b TM [11], as observed in the aIIbb3 interface [16]. In aL, a polar residue, Ser-1071, was hypothesized to form a polar interaction with Thr-686 at the binding partner b2 [36]. Both the alignment of Ser-1071 with Gly-1075 and the hydrophobic IMC region, together with the sensitivity of these residues to titration with the b2 peptide, are consistent with this part of aL forming a polar interaction with b2. As polar interactions are stronger in non-polar environments [56,57], this has been proposed to have an impact on integrin deactivation, which is important during leukocyte migration [35]. However, further characterization of the aL/b2 interaction is necessary to confirm that point. The OMC region of aL may extend as far as Tyr-1068 and Met-1064, as suggested by the titration experiments.
A chemical shift change was also observed in Asn-1095. This residue was reported to be involved in heterodimerization of the aLb2 cytoplasmic domain [21]. The equivalent residue in aIIb, Asn-996, experienced a large chemical shift change upon heterodimerization with b3 TM domain [16]. It is possible that the changes observed in aIIbb3 Asn-996 are indirectly caused by a nearby salt bridge, Arg-995(aIIb)-Asp723(b3) [16]. This salt bridge should also be present in aLb2, between aL Arg-1094 and b2 Asp-731 [21]. Unfortunately, peak crowding of a large portion of the cytoplasmic tail prevented us from examining the behavior of other residues near this region, e.g. Arg-1094 and Lys-1097.
In summary, we have shown that the conserved GFFKR motif in a integrins forms a structural motif, a backbone reversal, that can be observed when TM domain, cytoplasmic tail and a lipid environment consisting or membranes or bicelles are present, although cytoplasmic tail length may also be an important variable for the stability of this turn [54]. This backbone reversal has been now observed in two different a integrins TMs, aIIb and aL. This is likely to be one of the structural forms of this motif, present in the inactive or intermediate activation state of integrins. We also show that for aL/b2, interaction in the TM domain proper takes place through a remarkably similar structural organization to aIIb/b3, with features reminiscent of OMC and IMC clasps. Further studies will include mutagenesis at the predicted a-b interface to confirm the observed interaction, and crucially, determination of interhelical NOE contacts that can unequivocally describe the precise nature of the polar interaction in aL/b2 pairs. This work is presently underway.

Supporting Information
File S1 Includes Table S1, Figures S1-S3 . Table S1. Structure statistics for the selected 20 structures of integrin aL TM domain. Figure S1. Expression and purification of integrin aL TM and b2 TM peptides. (A) SDS-PAGE of methanol-extracted (meOH) and HPLC-purified (HPLC) aL-TM and b2-TM peptides. (B) MALDI-TOF MS spectra of the HPLC fractions corresponding to aL TM and b2 TM peptides. The letters inside square brackets identify the charged species forming a particular peak: M, H, Na, K corresponds to peptide, proton, sodium ion, and potassium ions, respectively. The charge of each species is indicated outside the square bracket. Figure S2. Secondary structure of the peptide aL used, based on NOE connectivity. Sequential and medium-ranged NOE connectivity between residues are displayed as bands under the respective residues. The d aN (i,i+4) connectivities indicate the ahelical structure, extending from Leu-1065 to Val-1088. Figure  S3. Secondary structure of integrin TM peptides. (A) FT-IR spectra of integrin TM a (upper) and b (lower) peptides reconstituted in DMPC lipid bilayers indicating that both TM peptides are almost completely a-helical. The amide I peak centered at 1657 cm 21 (blue) and the Fourier self-deconvolved spectra (red) are shown; (B) CD spectra of integrin TM peptides in DHPC-DMPC bicelles, showing minima at 209 and 222 nm, indicating a high proportion of a-helical form for both peptides; (C) Amide I and amide II regions of integrin TM peptides when bulk water was removed (red) and after D 2 O-saturated air exposure (blue). The decrease in the amide II area at ,1550 cm 21 indicates H-D exchange. Figure S4. Effect of increasing concentration on the [ 1 H-15 N]-TROSY-HSQC spectrum of aL-TM. aL TM peptide at 0.1 mM (A) or 0.6 mM (B). aL was reconstituted into DHPC-DMPC bicelles (3% w/v, q = 0.3) buffered with 50 mM potassium phosphate at pH 6.5 and spectra were recorded at 305 K. (DOCX)