Determination of the Border between the Transmembrane and Cytoplasmic Domains of Human Integrin Subunits*

In this study we have determined the position of the C-terminal end of the transmembrane domains of human integrin subunits (α2, α5, β1, β2) in microsomal membranes using the glycosylation mapping technique. In contrast to the common view, the transmembrane helices were found to extend roughly to Phe1129 in α2, to Phe1026 in α5, to Ile757 in β1, and to His728 in β2. The α-carbon of the conserved lysine present near the C-terminal end of the transmembrane helix (Lys1125 in α2, Lys1022 in α5, Lys752 in β1, and Lys724 in β2) is buried in the plasma membrane, and the charged amino group most likely reaches into the polar head-group region of the lipid bilayer. A possible role for the conserved lysine in integrin function is discussed.

Integrins are cell adhesive receptors composed of non-covalently linked ␣ and ␤ subunits. Each subunit consists of a large extracellular domain, a transmembrane helix (TMH), 1 and a short cytoplasmic tail of usually less than 60 amino acids. Cell attachment to extracellular matrix via integrins is necessary for normal cell growth and differentiation. Integrins are also involved in cellular processes that require migration of cells, e.g. angiogenesis and extravasation of lymphocytes. Upon ligand binding, clustering of integrins leads to formation of focal contacts containing signaling complexes (1,2).
The short cytoplasmic domains of integrin ␤ subunits have multiple functions: to establish contact with the actin cytoskeleton, to start signaling cascades, and to regulate the conformation of the extracellular domain of the receptor and thereby the ability to bind extracellular ligands (3)(4)(5)(6). All these functions depend on interactions with cytoplasmic proteins, some of which mediate outside-in signaling and some that regulate extracellular ligand binding affinity (so called inside-out signaling) (7). ␣ subunits of integrins appear to have a regulatory role over the ␤ subunits, possibly by hindering ␤ subunits from binding to cytosolic proteins in the absence of bound extracellular ligand (8,9). In addition, specific signals are also generated by cytoplasmic tails of ␣ subunits (10 -12). Thus, the cytoplasmic domains are indispensable for proper functioning of integrins as demonstrated by many studies.
However, for both integrin subunits, the border between the cytoplasmic domain and the C-terminal end of the transmembrane domain is unclear. Usually the cytoplasmic domain of integrins is assumed by most authors to start at the first charged residue after the continuous stretch of 23 hydrophobic amino acids. In a few cases the cytoplasmic domain has instead been suggested to be 4 and 5 amino acids shorter for the ␣ and ␤ subunits, respectively (see Fig. 1) (13)(14)(15). Interestingly, all presently known ␣ and ␤ subunits, except for ␤4, exhibit a similar pattern in this region: a single conserved, positively charged amino acid (Lys/Arg), a short stretch of hydrophobic amino acids, and highly polar sequence unlikely to be buried in the plasma membrane. The regions between and C-terminally adjacent to the predicted ends of the TMHs of ␣ and ␤ subunits ( Fig. 1) are involved in affinity regulation and in dimerization of integrins (4, 16 -20).
In this study, we have used a glycosylation mapping technique to determine the position of the C-terminal end of the TMHs of human integrin ␣ subunits (␣2, ␣5) and ␤ subunits (␤1, ␤2) in microsomal membranes. Our results establish that the ␣-carbon of the conserved lysine and the following short hydrophobic stretch of all tested subunits are buried in the membrane. Possible functional implications of this finding are discussed.
Expression in Vitro-Synthesis of mRNA from pGEM1 by SP6 RNA polymerase and translation in reticulocyte lysate in the presence and absence of dog pancreas microsomes was performed as described (25). Proteins were analyzed by SDS-polyacrylamide gel electrophoresis, and gels were quantitated on a Fuji BAS1000 phosphoimager using the MacBAS 2.31 software. The extent of glycosylation of a given construct was calculated as the quotient between the intensity of the glycosylated band divided by the summed intensities of the glycosylated and nonglycosylated bands. In general, the glycosylation efficiency varies by no more than Ϯ 5% between different experiments, and the precision in the minimal glycosylation distance (MGD) determinations is Ϯ 0.2 residues.

RESULTS
Glycosylation Mapping-The glycosylation mapping technique has previously been described in detail (21). Briefly, the lumenally oriented active site of the endoplasmic reticulum enzyme oligosaccharyltransferase was used as a fixed point of reference against which the position of a TMH in the endoplasmic reticulum membrane could be measured ( Fig. 2A). Based on the variation in glycosylation efficiency for a set of constructs differing only in the position of a glycosylation acceptor site, it was possible to define a MGD, i.e. the number of residues in the nascent chain needed to bridge the distance between the end of the TMH and the oligosaccharyltransferase active site. By calibration of the MGD scale against TMHs whose position in the lipid bilayer had been determined by direct techniques such as x-ray crystallography, NMR, or fluorescence quenching measurements, the point where a TMH exited from the lipid environment could be determined to within less than Ϯ 1 residue.
Previously, we used this technique to study how the position of model TMHs in the endoplasmic reticulum membrane changed in response to single mutations such as the introduction of a proline or a charged residue (21,22). We had also calibrated our measurements against two different TMHs of known position in the membrane, the TMH of the H-subunit from the photosynthetic reaction center and the TMH from the phage M13 major coat protein, as well as against model poly-Leu TMHs of varying lengths (21,22). These studies showed that the MGD measured from the first residue after the hydrophobic region of a typical TMH is 9.5-10.5 residues.
Determination of the Membrane-embedded Parts of Integrin Subunits-For the studies of the TMH of ␣ subunits, a segment encoding residues 989 -1028 was amplified from the ␣5 and 1096 -1131 from ␣2 cDNAs and cloned into a series of previously constructed vectors based on the well characterized protein leader peptidase (Lep) (Fig. 2A). The vectors differ only in the position of a single Asn-Ser-Thr glycosylation acceptor site downstream of the TMH and, thus, allow facile determination of the C-terminal MGD for any TMH.
The results of in vitro transcription/translation of three Lep-␣2 and Lep-␣5 constructs in the absence and presence of dog pancreas microsomes are shown in Fig. 2B, and the MGDdetermination is shown in Fig. 2C. Essentially identical results were obtained for both tested ␣ subunits. The glycosylation efficiency drops from 54% for the Asn 88 construct to 6% for the Asn 87 construct for ␣2 subunit and from 85% for the Asn 88 construct to 6% for the Asn 87 construct for ␣5 subunit. Since the expected MGD value for a TMH longer than ϳ23 residues is ϳ10 residues (21), this allows the C terminus of the TMHs of ␣2 and ␣5 to be positioned relative to the reference TMHs (Fig.  2D). Even with allowance for a rather wide margin of error, this clearly places the ␣-carbon of Lys 1125 in ␣2 and Lys 1022 in ␣5 more than one helical turn below the membrane-water interface, similar to the position of Lys 40 in M13 coat protein (26,27) and to single lysine mutations in a poly-Leu TMH (22).
Similarly, for the studies of ␤ subunits, a segment encoding residues 723-761 from the ␤1 and 695-933 from the ␤2 cDNAs was amplified and cloned into the Lep vectors. Very similar results were obtained (Fig. 2, B and C), placing the ␣-carbon of Lys 752 in ␤1 and Lys 724 in ␤2 well below the membrane-water interface (Fig. 2D).
We conclude that the membrane-embedded parts of the TMHs of the ␣ subunits extend roughly to Phe 1129 in ␣2 and to Phe 1026 in ␣5. For the ␤ subunits, the TMH of ␤1 extends to Ile 757 , and the TMH of ␤2, to His 723 . As proposed in the socalled snorkel model (28 -30), the long, aliphatic part of the side chain of the membrane-embedded Lys in integrin ␣ and ␤ subunits most likely extends toward the membrane surface, placing the positively charged terminal amino group in the polar head-group region of the lipid bilayer.
Mutations Near the C-terminal End of the Transmembrane Segments Cause a Shift in Membrane Location-To confirm our interpretation of the glycosylation mapping data, we constructed two derivatives of the ␤1 TMH: one where the entire hydrophobic segment between Lys 752 and His 758 was replaced by Asn residues (␤1⌬5N), and one where Leu 753 was replaced by Lys (␤1⌬L-K) (Fig. 3A). Since both Asn 82 constructs were fully glycosylated (data not shown), the four residues HDRR near the lipid-water interface were deleted to facilitate the MGD determination. For both ␤1⌬5N and ␤1⌬L-K, the MGD measurement indicated a substantial shift in the position of the TMH in the membrane (Fig. 3B). Although we have only determined the MGD to within Ϯ 1 residue in this case, it is nevertheless clear that Lys 752 is positioned much closer to the membrane-water interface in these constructs (Fig. 2D). This demonstrates that the hydrophobic segment between Lys 752 and His 758 in wild type ␤1 is embedded in the membrane and that it is pushed out of the membrane when its hydrophobicity is reduced. DISCUSSION In this study the glycosylation mapping technique has been used to determine the position of the TMHs of two human integrin ␣ and ␤ subunits in the microsomal membrane. The TMHs of the ␣ subunits were found roughly to Phe 1129 in ␣2 and to Phe 1026 in ␣5, and the TMHs of the ␤ subunits were found to extend to Ile 757 in ␤1 and to His 728 in ␤2. Interestingly, the ␣-carbon of the single positively charged amino acid present near the C-terminal end of the TMHs (Lys 1125 in ␣2, Lys 1022 in ␣5, Lys 752 in ␤1, and Lys 724 in ␤2) was buried in the membrane in all cases. Thus, the same result was obtained, irrespective of whether tryptophan or tyrosine is N-terminally adjacent to the conserved sequence KXGFFKR of the ␣ subunits. Reductions in the hydrophobicity of the short hydrophobic stretch downstream of Lys 752 in the ␤1 TMH caused a shift in the position of Lys 752 relative to the membrane, confirming that the hydrophobic segment between Lys 752 and His 758 is indeed buried in the membrane in the wild type protein. These results were obtained using microsomal membranes; however, the integrin TMHs can be accommodated in the same way also in the plasma membrane, which actually is thicker than microsomal membranes. Previous measurements (21,22) have shown that MGD ϭ 10.1 for RC-H (counting from the indicated Glu residue), MGD ϭ 10.7 for M13 coat (counting from the indicated Phe residue), and MGD ϭ 9.7 for 23L (counting from the indicated Gln residue). For the integrin subunit ␣2, ␣5, ␤1, and ␤2 TMH, the location relative to the membrane-water interface has been estimated based on the assumption that MGD ϭ 10 residues (arrows). Since the MGD has only been determined to within Ϯ 1 residues for the ␤1⌬5N and ␤1⌬L-K constructs (see Fig. 3), a wider margin of error is indicated in these two cases (hatched). The position of Glu 82 , the first Lep-derived residue, and Gln 85 are indicated. Lys 1125 in ␣2, Lys 1022 in ␣5, Lys 752 in ␤1, Lys 724 in ␤2, and Lys 40 in the M13 coat protein are highlighted in bold. Note that residues HDRR near the membrane-water interface in ␤1 have been deleted in the ␤1⌬5N and ␤1⌬L-K constructs. Residues in the TMH are shown in uppercase.
The highly conserved motif KXGFFKR at the C-terminal end of the ␣ subunit TMH is known to be critically important for integrin function, although the phenotypic effects of modification in this region vary between different integrins. Deletion of the motif from ␣IIb results in a constitutive high affinity state of the ␣IIb␤3 integrin. Similarly, deletion of LLITIHD in ␤3, the opposing region of KXGFFKR motif in ␣IIb, also leads to a high affinity state of the receptor. It has been suggested that the amino acids Arg and Asp in these motifs form a salt bridge and that breaking the bridge locks the integrin in a conformation corresponding to a high affinity state (16,17). This idea is further supported by molecular modeling on integrin subunits ␣IIb and ␤3 showing that the LLITIHD sequence in ␤3 and the KXGFFKR motif in ␣IIb are likely to be associated (31). Similarly, deletion of VGFFK in ␣L increases the affinity of the ␣L␤2 integrin, but in this case the mutation also interferes with post-ligand binding events dependent on the cytoskeleton (19,32). In addition, the KXGFFKR motif has been shown to promote the assembly and/or stabilization of the ␣L␤2 heterodimer (18,19) as well as of several ␤1 integrins (4,20).
These regions have been demonstrated to interact with a variety of different proteins. The intracellular calcium-binding protein calreticulin was reported to bind the sequence KXG-FFKR in ␣ subunits, and this interaction may be required for integrin activation (33,34). A recent report by Coppolino and Dedhar shows that interaction between ␣ subunits and calreticulin is ligand-specific and transient, occurring shortly after ligand binding (35). The synthetic peptide KLLMIIHDRREFA derived from the ␤1 sequence was found to interact with focal adhesion kinase in vitro (36). Focal adhesion kinase is known to be involved in integrin-mediated signaling (37) but appears not to be required for integrin activation (38). Recently, the proteins Rack1 and skelemin were shown to bind to the membrane-proximal region of ␤-subunits in yeast two-hybrid assay (39,40). Rack1 binding to ␤1 and ␤2 integrins requires activation of protein kinase, which is one of the early events after ligand binding of integrins (41,42), and Rack1 may recruit protein kinase C to adhesion complexes by its ability to bind integrins. However, the functional role of Rack1 in integrin signaling remains to be elucidated. Muscle skelemin was found to bind ␤1 and ␤3 subunits but not to ␤2. In Chinese hamster ovary cells the skelemin-like protein colocalized with integrins at early stages of cell spreading (40).
Our results show that the KLGFF sequence in ␣2 and ␣5, KLLMII in ␤1, and KALIH in ␤2 are buried in the membrane in the absence of interacting proteins. Why, then, is a transmembrane lysine conserved in all known integrin subunits (17 ␣ and 8 ␤ subunits)? One possibility is presented in Fig. 4A. This model is based on the studies discussed above, demonstrating that the regions between arrows 1 and 2 in Fig. 1 are involved in interactions with intracellular proteins and, thus, are likely to be exposed to the cytoplasm. The charged residue may facilitate a transfer of this region out of the hydrophobic A, activation of integrins is suggested to involve a movement of the "mobile region" (black areas in the TMHs) out of the plasma membrane. In the inactive state this segment of both subunits may associate with each other. In this situation the extracellular domain is in a conformation that is incapable of ligand binding, and the cytoplasmic tail of the ␣ subunit masks the cytoplasmic tail of ␤ subunit. In a fully active integrin the mobile regions are exposed to interacting proteins in the cytoplasm, the transmembrane domains are 4 -5 amino acids shorter, and the extracellular domain is able to bind to a ligand. It should be noted that this represents one of several similar possible scenarios; integrin activation may be obtained by moving the TMH of one or both of the integrin subunits in or out of the membrane. B, the positively charged residues of the integrin TMHs are indicated to participate in interactions with other membrane proteins (striped). Gray areas in the plasma membrane indicate polar head-group regions. environment as a result of binding to intracellular proteins. Such a movement could trigger the conformational changes associated with integrin activation and/or ligand binding.
The possibility that the position of the conserved lysine and the following stretch of hydrophobic amino acids in the plasma membrane may contribute to the transmembrane signaling of integrins was first discussed by Williams et al. (43). Our study provides supporting evidence for this view. We suggest that this region in one or both of the integrin subunits is positioned differently relative to the plasma membrane depending on the affinity state of the receptor. For example, in the inactive conformation, the C ␣ s of the lysines could be buried in the plasma membrane. In this situation the TMHs would probably be tilted in the membrane due to their length, and the ␣ and ␤ subunits could associate close to the cytoplasm, e.g. via the Asp-Arg bridge. Upon integrin activation, the TMHs would be shortened by 4 -5 amino acids, possibly induced by binding of cytosolic proteins to the cytoplasmic tails of integrin subunits. The tension exerted by actin cytoskeleton toward integrin clusters may also have a role in preventing backsliding of the mobile region (Fig. 4A).
Only a few studies have directly addressed the role of the conserved TMH Lys (e.g. Lys 752 in human ␤1 subunit) in integrin function. Mutation of the Lys in ␣1 (Lys to Asp) did not impair the ␣1␤1-mediated adhesion to collagen IV but resulted in localization of ␣1␤1 to focal contacts also on a fibronectin substrate (44). The phenomenon of ligand-independent clustering of integrins in focal contacts has been suggested to reflect a disturbed interaction between the ␣and ␤-cytoplasmic domains, resulting in unmasking of binding sites in the ␤-tail for components in existing focal contacts (8,44,45). Mutated chicken ␤1 subunit (Lys to Leu) was also found to localize to focal contacts when expressed in mouse NIH 3T3 cells (46), but in this case it is unclear whether the mutated receptor localized to focal contacts in a ligand-dependent or -independent manner. It is also unknown if this mutation affects the conformation of the extracellular domain.
Another possible role for the positively charged residues in the TMHs of integrins could be to promote interactions with other membrane proteins (Fig. 4B). This would be analogous to the T-cell receptor complex, which depends on charged amino acids in the TMHs for assembly and surface expression on T-cells (47). The T-cell receptor complex is stabilized in the endoplasmic reticulum by salt bridges between negatively charged residues in TMH of CD3 and positively charged residues in TMHs of T-cell receptor chains. In this context it is interesting that several tetraspanin proteins implicated in integrin function, e.g. CD9 and CD81, contain negatively charged glutamic acids in their third TMH (48 -50). Future experiments will be designed to test the validity of the two, not mutually exclusive models for the role of charged residues in integrin TMHs.