Integrin α1 Has a Long Helix, Extending from the Transmembrane Region to the Cytoplasmic Tail in Detergent Micelles

Integrin proteins are very important adhesion receptors that mediate cell-cell and cell-extracellular matrix interactions. They play essential roles in cell signaling and the regulation of cellular shape, motility, and the cell cycle. Here, the transmembrane and cytoplasmic (TMC) domains of integrin α1 and β1 were over-expressed and purified in detergent micelles. The structure and backbone relaxations of α1-TMC in LDAO micelles were determined and analyzed using solution NMR. A long helix, extending from the transmembrane region to the cytoplasmic tail, was observed in α1-TMC. Structural comparisons of α1-TMC with reported αIIb-TMC domains indicated different conformations in the transmembrane regions and cytoplasmic tails. An NMR titration experiment indicated weak interactions between α1-TMC and β1-TMC through several α1-TMC residues located at its N-terminal juxta-transmembrane region and C-terminal extended helix region.


Introduction
Integrins are cell adhesion receptors that mediate cell-cell and cell-extracellular matrix interactions, regulating cell growth and function. These receptors transmit bidirectional signals across the plasma membrane and contribute to the regulation of development, immune responses, inflammation, hemostasis, and to the development of many human diseases, including infection, autoimmunity, and cancers [1,2,3,4]. They are hetero-dimeric, type I transmembrane proteins consisting of a and b subunits. Each subunit contains a relatively large extracellular domain, a single transmembrane domain (TM), and a short cytoplasmic tail (CT) [4]. In mammals, 18 a subunits and 8 b subunits can form 24 different hetero-dimers that are expressed in particular tissues and bind to particular ligands. There are two forms of integrin signaling, which are known as outside-in and inside-out signaling.
Structural characterizations of the extracellular domains of integrins have long been studied [13,14,15,16]. However, few studies on the transmembrane and cytoplasmic (TMC) domains of integrins have been reported. In recent years, the TM and TMC domains of integrin aIIb and b3 were studied, alone or in complex, in organic solvents, detergent micelles, bicelles, or lipids using NMR (nuclear magnetic resonance) methods [3,17,18,19,20,21,22]. Also, interaction interfaces between aIIb/ b3 TM helices were studied using cysteine scanning and disulfide bond formation methods [23]. Multiple hydrophobic and electrostatic contacts within the membrane proximal helices of aIIb and b3 were revealed [3]. However, very few reports about the structures of integrin a1-TMC and b1-TMC are available.
Here, integrin a1-TMC (G1135-K1179) and b1-TMC (V717-K798) were over-expressed using a bacterial system and purified in LDAO (lauryl-dimenthylamine-n-oxide) detergent micelles. The solution structure of a1-TMC in detergent micelles was determined using NMR. The structure determined showed a long helix, extending from the transmembrane region to the cytoplasmic tail of integrin a1-TMC, which differed from the previously reported structure of integrin aIIb-TMC. Backbone 15 N relaxation data for a1-TMC in LDAO micelles also confirmed the extended helix. A chemical shift perturbation study of a1-TMC with the addition of integrin b1-TMC illustrated intensity attenuation in aqueous/ membrane interfacial residues of integrin a1-TMC, indicating weak interactions between integrin a1-TMC and b1-TMC at these residues.

Materials and Methods
Cloning and Over-expression of Human Integrin a1/b1 TMC Synthetic oligonucleotides encoding integrin a1-TMC (G1135-K1179) and b1-TMC (V717-K798) were amplified and subcloned into expression vector pET21b (Novagen) with a C-terminal 66His-tag. The recombinant protein was expressed using BL21(DE3) Gold in M9 medium at 25uC for 15 h. To achieve over-expression of isotope-labeled integrin a1-TMC, 1 g/L 15 N-NH 4 Cl and 3 g/L 13 C-D-glucose (Cambridge Isotope Laboratory) were used as the sole nitrogen and carbon sources, respectively.

Solution NMR Spectroscopy of Human Integrin a1-TMC
A set of multi-dimensional NMR experiments of 15 N-or 13 C/ 15 N-labeled integrin a1-TMC were conducted at 30uC, using a 600 MHz Bruker spectrometer equipped with a TXI cryoprobe. NMR spectra, including HSQC (hetero-nuclear single quantum correlation spectroscopy), HNCO, HNCA, HNCACB, CBCA(CO)NH, CC(CO)NH, HBHA(CO)NH, and HCC(CO)NH, were collected to obtain chemical shift assignments of backbone and side chain atoms. 15 N-edited NOESY-HSQC spectra (mixing time 100 ms) were collected to confirm the chemical shift assignments and to generate distance restraints for structure calculations. All NMR spectra were processed using NMRPipe [24] and analyzed using NMRView [25].

Residual Dipolar Coupling (RDC) Experiment
For backbone amide RDC measurements, a 6.5% polyacrylamide gel was prepared. Liquid gels (300 mL) will polymerize overnight at room temperature in 6-mm Teflon casting tubes. The polymerized gels were incubated for 2 h in 5 mL RDC buffer (50 mM NaH 2 PO 4 -Na 2 HPO 4 , pH 6.5) and then in 5 mL RDC buffer, supplemented with 0.5% LDAO and 10% D 2 O. Then, the gel was incubated with 2 mL 15 N-labeled integrin a1-TMC sample in the same detergent/buffer solution at room temperature for 2 days. The protein-soaked gel was then stretched into a 5-mm NMR tube using a device similar to that developed by the Bax group [8].
One-bond 1 H-15 N RDCs [26] were measured by acquiring a pair of spectra to yield semi-TROSY (TROSY in the 1 H dimension and anti-TROSY in the 15 N dimension) and semi-TROSY (TROSY in the 15 N dimension and anti-TROSY in the 1 H dimension) resonances using 15 N-labeled integrin a1-TMC in stretched gels. The couplings were obtained from the 15 N resonance frequency differences of the two semi-TROSY contour peak components.

Structure Calculations
Backbone dihedral angle restraints were obtained from the backbone chemical shifts using TALOS+ [27]. Extensive side chain NMR resonance assignments were not possible, such that the 1 H-1 H NOEs (nuclear Overhauser effect) used to derive distance restraints for structural calculation were limited primarily to short-range backbone HN-HN distances. Backbone dihedral angle, NOE, and one-bond 1 H-15 N RDC restraints were used to calculate the structure of integrin a1-TMC with Xplor-NIH [28]. The final ten structures with the lowest energy were verified using PROCHECK-NMR [29]. Chemical shifts have been deposited in BioMagResBank (accession 17424). The structural coordinates have been deposited in PDB (accession 2L8S).  15 N NOE values were determined from peak ratios observed between two spectra collected with or without a 3 s power presaturation in the proton channel.

Primary Sequence Analysis of the TMC Domains of Different Integrins
The transmembrane regions of the 18 a integrins are wellconserved, as shown in Figure S1A. Integrin a1 has the shortest Cterminus among all a subunits and has a specific PLKKKMEK polybasic sequence [30]. A conserved GFFKR motif in a subunits is considered to be an interaction site between integrin a1 and b1, similar to the combination of hydrophobic and electrostatic

Ramachandran plot statistics (%)
Most favored regions 81.9 Additional allowed regions 11.4 Generously allowed regions 5.7 Disallowed regions 1. interactions between aIIb and b3 [3]. This conserved GFFKR motif is known to play an important role in the regulation of integrin function, while deletion of the specific PLKKKMEK sequence has been reported to affect a1/b1-dependent signal transduction [30]. A tentative topology map of integrin a1-TMC, including the transmembrane helix, is shown in Figure S1B, with the conserved GFFKR motif highlighted.

Solution NMR Backbone Resonance Assignment of Integrin a1-TMC
A high-quality HSQC spectrum for a1-TMC was obtained in LDAO micelles, which was the basis for further resonance assignments and structural determination of the protein in LDAO. With collection of a full set of triple resonance and threedimensional solution NMR spectra, sequential resonance assignments were achieved for backbone nuclei ( 13 C a , 13 C b , 13 CO, amide 15 N/ 1 H) of integrin a1-TMC in LDAO micelles. The HSQC spectrum with each resonance assigned to residues of integrin a1-TMC is shown in Figure 1A. In total, 43 sets of backbone carbon resonances (including 13 CO, 13 C a and 13 C b ) and 38 backbone amide ( 1 H, 15 N) resonances were assigned. There were still four residues (G1135, L1142, M1177, E1178) that could not be assigned, probably due to peak overlap of the narrowly dispersed HSQC spectrum of the sample in detergent micelles, or microsecond-millisecond motion of these residues.
The secondary structure of integrin a1-TMC in LDAO micelles was analyzed using TALOS+ [27] from assigned chemical shift values of 13 CO, 13 C a , 13 C b , amide 1 H, and 15 N (Fig. S2). Sitespecific secondary structure prediction indicated that, in total, 24 residues (L1142-K1165) were shown in an a-helix secondary structure, corresponding to the transmembrane helix of integrin a1-TMC.

Structural Calculation and Description
The solution structure of integrin a1-TMC was determined using Xplor-NIH [28], based on 212 NOE, 60 dihedral angle, and 32 backbone 1 H-15 N RDC restraints. Ten lowest energy structures were selected out of 100 calculated structures. Structural computation statistics regarding the quality and precision of integrin a1-TMC are summarized in Table 1. The backbone superimposition of the final ten conformers is presented in Figure 2A. In this structure, a kink was observed in the transmembrane helix at the position of G1152 (Fig. 2B). A stretch of helix with 28 residues was observed to extend from the transmembrane helix to the conserved GFF motif. This conformation of the integrin a1-TMC domain in LDAO micelles was also consistent with the backbone 15 N relaxation data (Fig. 1B-D). The longitudinal relaxation T 2 values of residues W1143-F1168 were similar (about 30 ms; Fig. 1C) and their steady-state NOE values were above 0.5 (Fig. 1D), indicating that these residues (W1143-F1168) form a stable secondary structural region, flanked by two flexible terminals.

Structural Comparison of Integrin a1-TMC with other Reported Integrin TMC Domains
Using solution NMR or computation modeling methods, several structures of integrin transmembrane helix (TM) and/or C-terminal tails (CT) have been determined over the past 10 years. Most of the reported integrin TM or CT structures are integrin aIIb/b3, which play important roles in primary platelet adhesion [3,19,20,21,22,31,32]. Previously, several TM or CT domain structures of integrin aIIb/b3 have been studied in different amphipathic environments (such as detergent micelles, phospholipid bicelles) or organic solvents. Some minor structural differences were observed for the same integrin segments in the different environments.
Notably, in the cytoplasmic region of integrin a1 and aIIb, the GFF (1167-GFF-1169 in a1 and 991-GFF-993 in aIIb) has been reported to form a helix in detergent and organic solvent (Fig. 3B,  3E) [19], or a GFF reverse turn with its two Phe residues immersed back into the hydrophobic region of bicelles (Fig. 3C, 3D) [20,22]. Moreover, the GFF helical conformation [33] and a GFF reverse turn [31,34] can be readily obtained through different computation modeling. In particular, a CS-Rosetta prediction by Yang et al. showed that the GFF reverse turn was the majority conformation, while the GFF helical conformation was seen in a small proportion [19]. In addition to the two different conformations of aIIb/b3 in different conditions [19,20], integrin a/b heterodimer formation efficiencies were also affected by different membrane-mimicking environments [35]. In light of these results and the complex physiological function of integrins, it is possible that integrin a1-TMC could be in multiple conformations (helix or reverse turn). Transitions between different conformations could be induced by environmental changes and/or specific physiological processes (e.g., activation/inactivation or monomer/dimer formation).
On the other hand, it was previously reported that the conserved GFFKR motifs in different integrins have different correlations with their functions. For example, the aIIb F992A or F993A mutation can activate aIIb/b3 [36] while the FF/AA mutation in this motif had little effect in the activation of aV/b3 [37]. Probably, the conformations of the GFFKR motif in these two integrins are different. Previously, it was reported that the aIIb/b3 association was sensitive to the integrity of the aIIb(R995)-b3(D723) salt bridge [20,36], the KR residues in GFFKR motif having undefined structure might provide some flexibility for the salt bridge formation between integrin a1 and integrin b1.
Here, an extending helical conformation of integrin a1-TMC was determined using solution NMR in detergent micelles, indicating a majority helical conformation of integrin a1-TMC in its monomeric form in micelles. Whether a conformation with a GFF reverse turn can be observed in the a1/b1 complex or in lipid bilayers need to be examined in future studies.

Interactions between Integrin a1-TMC and b1-TMC in LDAO Micelles
Interactions between the TMC domain of integrin a1 and b1 are known to be important for cell adhesion, probably due to integrin clustering. The C-terminal tail of integrin a1 plays an essential role in both physiological and pathological angiogenesis [38]. Deletion of the entire cytoplasmic tail of integrin a1 or mutations in several amino acids distal to the highly conserved GFFKR motif have been reported to have a similar phenotype to parental a1-null cells, resulting in malfunctions in angiogenesis and endothelial cell proliferation [38]. The highly conserved GFFKR motif in the a1 tail has been proposed to form a salt bridge with conserved residues in the b1 tail. However, the detailed interaction between TMC domains of a1 and b1 is not yet understood. Due to the complex enthalpy solvent effects of detergent mixing or exchange between two membrane protein samples, isothermal titration calorimetry assay is not suitable for analyzing interactions between a1-TMC/b1-TMC in detergent micelles. Thus, NMR titration experiment was employed to study the interaction between a1-TMC and b1-TMC. First of all, we acquired the HSQC spectrum of b1-TMC to make sure it's a well folded sample (Fig. S3). Then, a series of 1 H-15 N HSQC spectra of 15 N-labeled integrin a1-TMC with different concentrations of non-labeled b1-TMC were acquired and the processed spectra are shown in Figure 4. Surprisingly, no obvious chemical shift perturbation was observed anywhere in the spectrum, indicating no pronounced conformational change of integrin a1-TMC upon addition of b1-TMC, maintaining the major helical conformation in the GFF motif. However, intensity attenuations were observed in several resonances with increasing integrin b1-TMC concentration (Fig. 4). These resonances with attenuated intensities were mapped to two regions: the N-terminal juxta-transmembrane region (V1140, W1143, V1144, I1145, S1148, A1151, and G1152) and the C-terminal juxta-transmembrane region (L1161, A1162, L1163, W1164, K1165, I1166, G1167, F1168, F1169, K1170, and R1171). These residues are marked with arrows at the top of Figure 4.
According to solution NMR relaxation theories, peak intensity attenuation or missing peaks are attributed to intermediate-time scale (microsecond to millisecond) conformational exchanges [39,40,41]. Thus, intensity attenuation or the 'disappearance' of residues are hypothesized to indicate interaction and these residues are located in the direct interface between integrin a1/b1. This hypothesized a1/b1 interaction interface is possibly similar to interactions in integrin aIIb/b3 and consistent with previous mutagenesis and deletion studies of integrin a1 [42,43,44]. In the complex structure of aIIb/b3 TMC, aIIb residues W968, G972, G976, L979, L980, and R995 are involved in the dimer interface [19,20]. Their corresponding residues in a1 are V1144, S1148, G1152, L1155, L1156, and R1171. Here, the HSQC peaks of four of them (V1144, S1148, G1152, R1171) were obviously attenuated in the NMR titration experiment with integrin b1-TMC, while perturbations of the other two residues (L1155, L1156) were not apparent because they were crowded by L1158 and I1160. Also, the aIIb/b3 residues involved in the dimer interface are largely conserved in a1/b1. Those observations implied that the dimer interface of a1/b1 TMC is similar to that of aIIb/b3 TMC. Further structural studies of a1/b1 TMC complex will illustrate detail interaction surfaces. Residue mutations and deletions in the two regions have been reported to interfere with associations between a and b subunits and between integrin and cytoplasmic binding partners, thus interfering with downstream signal transduction, leading to inhibition of cell spreading and stress fiber formation [42,43,44]. Thus, titration results of integrin a1 with the addition of integrin b1-TMC provide preliminary insights about the interaction interfaces between the two proteins, and provide a basis for further detailed studies of signal transduction in fibrosis, angiogenesis, or cancer cells.