Divalent cations regulate the folding and activation status of integrins during their intracellular trafficking

Abstract

Integrins are divalent cation-dependent, αβ heterodimeric adhesion receptors that control many fundamental aspects of cell behaviour by bi-directional signalling between the extracellular matrix and intracellular cytoskeleton. The activation state of cell surface integrins is tightly regulated by divalent cation occupancy of the ligand-binding pocket and by interaction with cytoplasmic adaptor proteins, such as talin. These agents elicit gross conformational changes across the entire molecule, which specify the activation state. Much less is known about the activation state of newly synthesised integrins or the role of cations during the early folding and trafficking of integrins. Here we use a number of well-characterised, conformation-specific antibodies to demonstrate that β1-integrins adopt the bent, inactive conformation after assembly with α-integrins in the endoplasmic reticulum. Folding and assembly are totally dependent on the binding of Ca²⁺ ions. In addition, Ca²⁺ binding prevents integrin activation before its arrival at the cell surface. Activation at the cell surface occurs only following displacement of Ca²⁺ with Mg²⁺ or Mn²⁺. These results demonstrate the essential roles played by divalent cations to facilitate folding of the β-integrin subunit, to prevent inappropriate intracellular integrin signalling, and to activate ligand binding and signalling at the cell surface.

Introduction

The expression of functional integrin molecules is a highly orchestrated process that begins with the assembly of integrin heterodimers in the endoplasmic reticulum (ER), and ends with priming and ligand-induced activation of the molecule at the cell surface (Hynes, 2002). Despite the structural heterogeneity exhibited by the 24 different mammalian integrin dimers, the processes of priming and activation are thought to be conserved. Throughout these processes, changes in the conformation of the molecule have a key role in both facilitating intracellular trafficking and in converting a passive receptor at the cell surface into a high-affinity, highly specific adhesion molecule. These changes in conformation involve transition from an inactive form with low affinity for ligand, to a primed form with high affinity for ligand, to a fully activated ligand-bound cell adhesion receptor (Luo and Springer, 2006). Each stage in the activation process is characterised by gross conformational changes in the integrin structure. The inactive form is thought to exist as a hairpin or bent structure with the globular ligand-binding domains facing the membrane. Upon priming, there is a dramatic straightening of the molecule, which forms a more extended structure, with the binding domains now protruding from the membrane. Evidence to support such a switchblade movement of the integrin ectodomain comes from the crystal structure of the αVβ3 (Xiong et al., 2001) and α_IIbβ₃ (Zhu et al., 2008) integrins, which have a bent conformation that is considered to be the inactive form. Furthermore, electron microscopy, hydrodynamic volume and antibody epitope-mapping studies have revealed the transition between a bent and extended form upon priming and ligand binding (Beglova et al., 2002; Takagi et al., 2002). Although our understanding of the events that underlie affinity regulation has significantly advanced in the last few years, there are still crucial aspects of the process that remain unclear. During inside-out signalling, integrin activation is regulated by binding of intracellular proteins such as talin to the β-integrin cytoplasmic tail (Tadokoro et al., 2003), which leads to the separation of the α- and β-integrin legs (Vinogradova et al., 2002; Luo et al., 2004; Anthis et al., 2009) and increased affinity for ligand (Luo and Springer, 2006). Because intracellular signalling molecules could bind to integrins and activate them inside the cells, it is still not known how cells avoid unwanted intracellular signalling. Above all, we do not know the conformational state of newly synthesised integrin molecules and whether or not they become primed or activated inside the cell.

Integrins contain several cation-binding sites that regulate the ligand-binding affinity of the receptor. Different cations have markedly different effects on ligand affinity: in general, Mn²⁺ supports ligand binding, Mg²⁺ does so to a lesser extent, and Ca²⁺ does not support ligand binding at all (Gailit and Ruoslahti, 1988). It has also been proposed that divalent cations can themselves cause pronounced conformational changes that result in a shift in the equilibrium between the active and inactive forms (Mould et al., 1995). It is clear that ligand binding and cation binding are intimately linked because all of the regions implicated in ligand recognition lie at, or close to, cation-binding sites (Xiong et al., 2001; Xiong et al., 2002). Ca²⁺ and Mg²⁺ are the predominant physiological cations present in the cells in mM concentrations (Montero et al., 1995; Laurant and Touyz, 2000), with Mn²⁺ present at much lower levels (1–14 μM) (Schramm and Brandt, 1986; Smith et al., 1994). Although the binding of cations to integrin at the cell surface is well established, very little is known about the timing of cation association during biosynthesis and trafficking through the secretory pathway.

In humans, the biosynthesis of the integrin heterodimers begins with the selective assembly of one of the 18 α-integrin subunits with one of the 8 β-integrin subunits in the ER. Both subunits require binding to their partner to ensure correct folding (Huang and Springer, 1997; Lu et al., 1998). To ensure a level of quality control of cell surface receptors, integrin heterodimers cannot be transported from the ER to the plasma membrane unless they have attained their native structure (Ho and Springer, 1983; Kishimoto et al., 1987). However, constitutively active integrins can be expressed at the cell surface (Larson et al., 1990), demonstrating that the quality control does not discriminate between inactive and primed conformations. The ability of cells to allow only heterodimers with the ability to become primed to exit the ER is a crucial aspect of integrin biosynthesis.

In this study, we characterised various conformational states of β1-integrins during biosynthesis and trafficking to the cell surface using a variety of conformation-specific antibodies. We found that, after assembly of the α- and β-integrins in the ER, integrin heterodimers adopt the bent, inactive conformation, which is dependent on the binding of Ca²⁺. We show that Ca²⁺ binding in the ER is essential for correct folding and assembly, and to maintain intracellular integrin in a bent conformation until it reaches the cell surface. Taken together, these results demonstrate that Ca²⁺ has a crucial role in integrin folding, assembly and trafficking, and maintains the receptors in an inactive form until they reach the cell surface.

Results

Probing the folding and assembly of integrins using monoclonal antibodies

Many monoclonal antibodies have been raised that are reactive towards a variety of conformations of β1-integrins (Mould, 1996; Humphries, 2000; Byron et al., 2009). Our initial aim was to use a selection of these antibodies to determine the folding, assembly and trafficking of integrins. Antibodies were first characterised in terms of their conformation specificity. Radiolabelled β1-subunits were generated by expressing the protein in an in vitro translation system supplemented with semi-permeabilised (SP) HT1080 cells (Wilson et al., 1995). The translation was carried out under reducing conditions to generate material that was translocated into the ER of SP-cells but which failed to fold correctly because of a lack of disulphide bond formation (Jessop et al., 2007). In addition, translation was carried out in the absence of a reductant to allow disulphide formation and correct folding. Finally, correctly folded β1-subunit was denatured with SDS in the presence or absence of reductant to generate denatured β1-subunits with intact or reduced disulphide residues. The various forms of β1-subunit were then immunoisolated with four different monoclonal antibodies, 8E3, 9EG7, TS2/16 and JB1A, which were previously reported to exhibit a wide range of functional activities.

A translation product synthesised in the absence of reducing agent with an approximate molecular weight of 100 kDa was immunoisolated by all the antibodies tested (Fig. 1A, lanes 1,5,9,13). This protein has been shown previously to be fully translated, glycosylated β1-subunit (Jessop et al., 2007). In addition, JB1A, but none of the other antibodies, recognised a translation product with an approximate molecular weight of 80 kDa that corresponds to unglycosylated, non-translocated β1-subunit (Fig. 1A, lanes 13–16). When translation was performed in the presence of reductant, only JB1A recognised the synthesised translation product (Fig. 1A, lane 14). Denatured β1-subunit with and without the reduced disulphides was immunoisolated by 8E3 and JB1A (Fig. 1A, lanes 3,4,15,16), whereas 9EG7 only immunoisolated the denatured protein if the disulphides were still intact (Fig. 1A, lanes 7 and 8). TS2/16 did not recognise any translation products that had been reduced or denatured. These results demonstrate that JB1A recognises all forms of the protein tested and that TS2/16 only recognises non-denatured, folded β1-subunit. The epitope recognised by 8E3 forms during correct folding of the protein and cannot be destroyed by denaturation in SDS, either with or without reducing agent. However, if the protein is prevented from folding following synthesis, the epitope for 8E3 does not form.

Fig. 1.

Reactivity of monoclonal antibodies towards β1-integrin. (A) Human β1-integrin was translated in a reticulocyte lysate containing radiolabelled methionine and SP cells for 2 hours. Translations were carried out in the absence (lanes 1,3,4,5,7,8,9,11,12,13,15,16) or presence (lanes 2,6,10,14) of reducing agent. Translation products were treated with the indicated antibodies under native (lanes 1,2,5,6,9,10,15,16) or denaturing conditions in the absence (lanes 3,7,11,15) or presence (lanes 4,8,12,16) of added DTT. Isolated translation products were separated by reducing SDS-PAGE and visualised by exposure to film. Radiolabelled ER form of β1-integrin (β1′) is indicated. (B) HT1080 cells were radiolabelled for 30 minutes and chased for a further 2 hours. Cell lysates were treated with the indicated antibodies under native (lanes 1,4,7,10) or denaturing conditions in the absence (lanes 2,5,8,11) or presence (lanes 3,6,9,12) of added DTT. Isolated proteins were separated by reducing SDS-PAGE and visualised by exposure to film. The identity of radiolabelled proteins is as indicated: α, α-integrin subunits; β1′, ER form of β1-integrin subunit; β1, Golgi form of β1-integrin subunit. Molecular size markers are shown in kDa.

To extend these findings to endogenous newly synthesised β1-subunits, HT1080 cells were pulse-labelled for 30 minutes followed by a chase for 2 hours. These conditions are known to allow the assembly of αβ-integrin dimers and their transport to the Golgi complex (Huang et al., 1997). Two distinct forms of β1-subunit can be recognised under these conditions: an ER form (β1′) that migrates in the same position as the fully translocated translation product and a Golgi form (β1) that has a slower mobility as a result of modification of the oligosaccharide side chains (Huang et al., 1997). Following labelling, β1-subunits were immunoisolated from cell lysates with antibodies as indicated (Fig. 1B). JB1A recognised both the β1 and β1′ forms of the protein under all conditions tested, confirming that this antibody is not conformation specific. In addition, under native conditions, some slower-migrating radiolabelled proteins were co-immunoisolated with mobility that was similar to that of the α-subunits (Fig. 1B, lane 10). HT1080 cells make several α-chains, any of which might assemble with β1-subunits. TS2/16 recognised the β1′-subunit and β-subunit forms and co-immunoisolated assembled α-subunits, but β1-subunit recognition was lost after denaturation (Fig. 1B, lanes 7–9). By contrast, 9EG7 recognised only the β1′-subunit under native conditions, but could recognise both the β1′ and β1 forms following denaturation in the absence, but not presence of reductant (Fig. 1B, lanes 4–6). Finally, 8E3 recognised the β1′-subunit only under native conditions, and weakly recognised both the β1′-and β-subunits under denaturing conditions, irrespective of whether the disulphides had been reduced (Fig. 1B, lanes 1–3). These results demonstrate that the panel of antibodies recognised different conformations of the β1-subunit. Most interestingly, under native conditions, 8E3 and 9EG7 recognised the monomeric ER form; however, the epitopes only become exposed in the Golgi form after denaturation. This observation suggests that the epitopes are present in monomeric β1-integrins, but become lost following assembly with α-integrin subunits. The results for 9EG7 extend those published previously which demonstrate that this antibody can recognise an αβ-integrin heterodimer, but only following binding to Mn²⁺ ions (Bazzoni et al., 1995). In addition, these two antibodies have been shown to report unbending of β1-integrin molecules during their activation (Mould et al., 2005; Askari et al., 2010). It is, therefore, highly likely that 9EG7 recognises an epitope in the heterodimer that only becomes exposed following chain unbending or denaturation.

Early folding events during the biosynthesis of β1-integrin

After characterisation of the various forms of β1-integrins recognised by the panel of antibodies, the folding, assembly and trafficking of endogenous protein was then studied in HT1080 cells. HT1080 cells were pulsed with ³⁵S-labelled amino acids and chased for varying lengths of time (Fig. 2). In most of the time courses, two labelled proteins at approximately 66 kDa and 220 kDa were observed, which bound non-specifically to Protein-G–Sepharose (Fig. 2A). All the antibodies recognised and isolated the β1′-integrin ER form, which diminished in intensity following 60 minutes of chase (Fig. 2B,D–F). The β1-integrin Golgi form was only recognised by TS2/16 and JB1A and appeared after 45 minutes. In addition, α-integrins were co-immunoisolated by TS2/16 and JB1A, presumably following assembly with the β1-integrin. A precursor–product relationship was seen between the β1′ and the β1-integrins, as demonstrated by quantification of the TS2/16 autoradiograph (Fig. 2C). As expected, no Golgi form of β1-integrin was recognised by 9EG7 or 8E3 (Fig. 2E,F).

Fig. 2.

The folding, assembly and trafficking of β1-integrin. HT1080 cells were radiolabelled for 30 minutes and chased for varying lengths of time as indicated. (A) Control cell lysate treated with Protein-G–Sepharose alone to indicate any radiolabelled proteins that bound non-specifically to the resin. Cell lysates were treated with monoclonal antibodies (B) TS2/16, (D) JB1A, (E) 9EG7 or (F) 8E3. Immunoisolated material was separated by SDS-PAGE under reducing conditions and radiolabelled proteins were visualised by autoradiography. The proteins that bind non-specifically to Protein-G–Sepharose are indicated by arrows in all the time courses. The identity of radiolabelled proteins specifically isolated with the antibodies is indicated as in Fig. 1. (C) The amount of TS2/16-immunoisolated β1′-integrin, β1-integrin and α-integrin subunits were quantified and plotted versus time.

Together, these results indicate that newly synthesised β1-integrins acquire their native conformation early, following synthesis in the ER. Exit from the ER and transport to the Golgi results in loss of the 9EG7 and 8E3 epitopes. The co-isolation of α-integrins with the ER form of β1-integrin (Fig. 2B, 15 and 30 minute time points) indicates that assembly occurs within the ER and is a prerequisite for exit from the ER. No Golgi-modified chains were recognised and no α-integrins were co-immunoisolated by 9EG7 or 8E3 indicating that assembly with the α-integrin prevents their reactivity towards these antibodies. Hence newly synthesised, Golgi-localised β1-integrin molecules exhibit an inactive, bent conformation.

Endogenous β1-integrin is mainly in the inactive conformation

To extend studies with radiolabelled β1-integrin subunits that report the conformation of newly synthesised protein, the activation status of the total cellular integrin complement was determined. β1-integrin receptors were immunoisolated from cell lysates using 9EG7 and isolated β1-integrins detected by western blotting with JB1A (Fig. 3A). As the 9EG7 epitope has been shown to be cation sensitive (Bazzoni et al., 1995; Mould et al., 1998), immunoisolation was also carried out following treatment of the lysate with Ca²⁺, Mn²⁺ and Mn²⁺ in combination with a 50 kDa ligand-binding fragment of fibronectin. No β1-integrins were isolated either when no cations were added, or following the addition of Ca²⁺ (Fig. 3A, lanes 1 and 2). However, in the presence of added Mn²⁺ ions alone, or in combination with the 50 kDa fragment of fibronectin, β1-integrins were isolated (Fig. 3A, lanes 3 and 4). Both the ER and Golgi forms were identified, which further demonstrates that assembly with the α-integrin occurs within the ER, and that even the ER-localised integrin heterodimer adopts a bent conformation. In addition, the majority of β1-integrins in HT1080 cells appeared to be in an inactive conformation because they are not recognised by 9EG7 in the absence of added Mn²⁺.

Fig. 3.

Effect of Ca²⁺, Mn²⁺ and fibronectin on total and cell-surface β1-integrins. (A) β1-integrin was immunoisolated with 9EG7 from HT1080 cell lysate either in the absence of added cations (lane 1), with added Ca²⁺ (lane 2), Mn²⁺ (lane 3) or Mn²⁺ with fibronectin (lane 4) and western blotted with JB1A antibody. Black arrowheads indicate immunoisolated precursor β1-integrin of ~100 kDa (β1′) and ~125 kDa mature β1-integrin (β1). The arrow shows a 50 kDa IgG heavy chain band (HC) present in all lanes, which is due to reactivity with the anti-mouse secondary antibody used for blotting. (B) HT1080 cells were biotinylated and cell lysis was performed either in the absence of added cations (lane 1) or with added Ca²⁺ (lane 2) or Mn²⁺ (lane 3). β1-integrin was immunoisolated from cell lysate with 9EG7 and subjected to SDS-PAGE and western blotting with avidin conjugated to peroxidase. Black arrowhead indicates the mature β1-integrin band. 9EG7 binding to cell surface β1 heterodimers in HT1080 cells was analysed by flow cytometry; in presence of (C) Ca²⁺, (D) Ca²⁺ and 50K fragment of fibronectin, (E) Mn²⁺ and (F) Mn²⁺ with 50K fragment of fibronectin. Solid line, 9EG7; dotted line, rat IgG control. (G) HT1080 cells were radiolabelled for 30 minutes and chased for varying lengths of time as indicated. Cell lysis was performed in lysis buffer with 2 mM EGTA to remove Ca²⁺ ions and lysates were immunoisolated with 9EG7. Immunoisolated material was separated by SDS-PAGE under reducing conditions and radiolabelled proteins were visualised by autoradiography. Black arrowheads indicate precursor β1-integrin of ~100 kDa (β1′) and ~125 kDa mature β1-integrin (β1). (H) HT1080 cells were radiolabelled for 30 minutes and chased for 2 hours. Cell lysis was performed in lysis buffer containing added EGTA (lane 1), EGTA and Mn²⁺ (lane 2) and EGTA and Ca²⁺ (lane 3) and lysates were immunoisolated with 9EG7. Immunoisolated material was separated by SDS-PAGE under reducing conditions and radiolabelled proteins were visualised by autoradiography. The identity of radiolabelled proteins is as follows: α, α-integrin chains immunoisolated with β1-integrin antibody as a result of assembly; β1′, precursor β1-integrin subunit; β1, mature β1-integrin subunit.

To differentiate between intracellular and cell surface integrins, cell surface proteins were biotinylated and β1-integrins immunoisolated. Isolated, biotinylated proteins were detected by blotting with peroxidase-conjugated avidin (Fig. 3B). Cell lysis was performed either in the absence or presence of Ca²⁺ or Mn²⁺. As was the case with total cellular integrins, no β1-integrin subunits were isolated in the absence of added cations or following the addition of Ca²⁺ (Fig. 3B, lanes 1 and 2), but in the presence of Mn²⁺, a 125 kDa mature β1-integrin was observed (Fig. 3B, lane 3). In addition, a slightly slower-migrating protein was also seen, which is likely to represent biotinylated α-integrins co-isolated with the β1-integrins. This result indicates that most cell surface β1-integrins are in an inactive conformation in HT1080 cells. To confirm these biotinylation results, we also performed flow cytometry on HT1080 cells to analyse 9EG7 binding to cell surface integrins (Fig. 3C–F). Binding of 9EG7 was only observed in the presence of Mn²⁺, with or without protein ligand (Fig. 3E,F). No binding was seen in the presence of Ca²⁺ even after the addition of protein ligand (Fig. 3C,D).

Ca²⁺ prevents activation of β1-integrins

The lack of activation of β1-integrins suggests that a mechanism exists to maintain the molecule in a bent conformation during trafficking through the cell. The bent inactive conformation could be maintained by binding of an adaptor protein or cation. Integrins contain several cation-binding sites, and binding of distinct divalent cations can regulate the ligand-binding affinity of the receptor at the cell surface. To determine whether cation binding maintains newly synthesised integrins in an inactive state, the time course of integrin biosynthesis was repeated in the presence of EGTA in the lysis buffer to chelate cations such as Ca²⁺. In the presence of EGTA, 9EG7 immunoisolated the ER form of β1-integrin (Fig. 3G). In contrast to the previous time course, 9EG7 also immunoisolated the Golgi form (from 60 minute chase). This result implies that in the absence of cations, the 9EG7 epitope becomes accessible on mature β1-integrin, indicating a role for cations in the prevention of unbending and activation. To determine which cation stabilises the inactive, bent conformation, excess Mn²⁺ or Ca²⁺ was added to the lysis buffer before immunoisolation with 9EG7 (Fig. 3H). Both Golgi and ER forms of β1-integrins were immunoisolated in the absence and presence of Mn²⁺, together with co-assembled α-integrins (Fig. 3H, lanes 1 and 2). The gel was exposed for longer than the time course to reveal the α-integrins, resulting in an increased background. In the presence of Ca²⁺, only the ER-unassembled form of β1-integrin was immunoisolated (Fig. 3H, lane 3). These results demonstrate that it is the binding of Ca²⁺ to the integrin chains that maintains any resulting integrin molecules in a bent conformation before and during trafficking of the receptor to the cell surface.

Ca²⁺ is required for folding and intracellular trafficking of

β1-integrin Given the role of Ca²⁺ in the maintenance of inactive receptors, it was important to establish when Ca²⁺ became bound to β1-integrins and whether it was required for correct folding. To address the role of intracellular Ca²⁺, free Ca²⁺ was depleted from the ER by treating cells with low concentrations of thapsigargin, which causes leakage of Ca²⁺ from the ER by inhibiting the ER sarcoplasmic/endoplasmic reticulum Ca²⁺-ATPase (SERCA) pump (Thastrup et al., 1990). It has been shown previously that treating cells with such a low concentration of thapsigargin depletes free Ca²⁺, preventing the folding of the Ca²⁺-binding LDL receptor, but does not cause a general defect in protein folding (Pena et al., 2010).

HT1080 cells were either untreated or depleted of free Ca²⁺ from the ER by treating with thapsigargin during the 30 minute starvation, 30 minute pulse and for various periods of chase (Fig. 4). As previously described, in the absence of Ca²⁺ depletion, TS2/16 immunoisolated correctly-folded ER-localised β1-integrins as well as αβ-integrin heterodimers that had been transported to the Golgi (Fig. 4A). However, following Ca²⁺ depletion, only a small fraction of β1-integrins were isolated by this antibody (Fig. 4B). However, β1-integrins were still recognised by 9EG7 following Ca²⁺ depletion (Fig. 4D). In contrast to the situation in the absence of Ca²⁺ depletion, the ER form was immunoisolated throughout the time course by 9EG7, indicating that no ER to Golgi transport had occurred. In support of this conclusion, JB1A immunoisolated only the ER form of β1-integrins throughout the time course, with no Golgi form observed or co-isolation of α-integrins (Fig. 4F). These results demonstrate a dramatic effect of Ca²⁺ depletion on β1-integrin folding, assembly and trafficking. Folding of the monomer was compromised, because very little reactivity was seen with TS2/16. The 9EG7 epitope was formed, but no transport was seen from the ER and no assembly with α-integrin chains. These results indicate that Ca²⁺ is required early in the folding pathway of the integrin receptor, is loaded onto the molecule in the ER and is required to ensure assembly with the α-integrin subunit.

Fig. 4.

Ca²⁺ is required to ensure β1-integrin folding and trafficking to the Golgi. HT1080 cells were starved and pulse-labelled for 30 minutes, and chased for up to 2 hours in the absence (A,C,E) or presence of thapsigargin (B,D,F). Cell lysates were collected at varying time points as indicated, β1-integrin immunoisolated with TS2/16 (A,B), 9EG7 (C,D) or JB1A (E,F) and then subjected to SDS-PAGE and autoradiography. The identity of radiolabelled proteins is as indicated in previous figures. Arrows indicate non-specific protein bands that were also pulled down by Protein-G–Sepharose alone (see Fig. 2).

To further explore the assembly of the β1-integrin with a specific α-integrin subunit, immunoisolation was carried out with the α5-integrin-specific monoclonal antibody mAb11 (Clark et al., 2005) (Fig. 5A). Assembly of α5β1-integrin was seen in HT1080 cells as evidenced by the co-isolation of β1-integrins with mAb11 (Fig. 5A, lanes 3,4). In Ca²⁺-depleted cells, only the α5-integrin band was observed (Fig. 5A, lanes 5–8), showing lack of α5β1-integrin assembly in absence of Ca²⁺. These results demonstrate that correct folding and association of α5-integrin with β1-integrin subunits requires Ca²⁺ in the ER and that depletion of Ca²⁺ inhibits trafficking of α5β1-integrin molecules to the Golgi.

Fig. 5.

Ca²⁺ is required for association of α5-integrin and β1-integrin subunits, but does not prevent folding and trafficking of MHC Class I. (A) HT1080 cells were starved and pulse-labelled for 30 minutes, and chased for up to 2 hours in the absence (lanes 1–4) or presence (lanes 5–8) of thapsigargin. Cell lysates were collected at varying time points as indicated, and α5-integrin subunits were immunoisolated with mAb11 and then subjected to SDS-PAGE and autoradiography. The identity of radiolabelled proteins is as described previously. Arrow indicates non-specific protein bands that were pulled down by Protein-G–Sepharose alone. (B,C) HT1080 cells were starved and pulse-labelled for 30 minutes, and chased for up to 2 hours in the absence or presence of thapsigargin as indicated. (B) Cell lysates were collected and immunoisolated with an antibody against β2-microglobulin and then subjected to SDS-PAGE and autoradiography. Black arrowhead indicates MHC I co-precipitated with β2-microglobulin. (C) Cell lysates were prepared and treated with Endo-H as indicated and then immunoisolated with anti-MHC I antibody W6/32. Black arrowhead indicates undigested MHC I and the line below indicates MHC I digested with EndoH enzyme.

It has been shown previously that the folding of influenza virus haemagglutinin is not affected by the conditions of Ca²⁺ depletion used in our experiments (Pena et al., 2010). To determine the effect of Ca²⁺ depletion on the folding machinery in the ER, the folding, assembly and Golgi transport of endogenous MHC Class I molecules was studied. During the biosynthesis of MHC Class I molecules, the heavy chain is first translocated across the ER membrane before assembly with β2-microglobulin and loading with peptide in the ER (Morrice and Powis, 1998). Both the initial assembly with β2-microglobulin, peptide loading and the subsequent transport to the Golgi is dependent upon the action of Ca²⁺-dependent ER chaperones, such as calnexin (Tector and Salter, 1995) and calreticulin (Sadasivan et al., 1996). Following pulse labelling of HT1080 cells, β2-microglobulin was immunoisolated to determine whether it was associated with the assembled heavy chain. In the absence of Ca²⁺ depletion, MHC Class I heavy chain was co-isolated at both 0 and 120 minutes of chase (Fig. 5B). After Ca²⁺ depletion, heavy chain was still isolated, demonstrating that thapsigargin treatment did not grossly affect the ER folding machinery. In addition, a conformation-specific antibody (W6/32) that only recognises correctly assembled and peptide-loaded MHC Class I molecules (Barnstable et al., 1978), was able to immunoisolate heavy chains even after Ca²⁺ depletion (Fig. 5C). To follow transport of newly synthesised MHC Class I molecules to the Golgi, the immunoisolated heavy chains were treated with endoglycosidase H (Endo-H). In the absence of Ca²⁺ depletion, heavy chains were Endo-H sensitive after 0 minutes of chase (Fig. 5C, lane 2). Heavy chains become Endo-H resistant after 120 minutes of chase (Fig. 5C, lane 4), indicating modification of their oligosaccharide side chains in the Golgi. Similarly, in thapsigargin-treated cells, MHC class I molecules were Endo-H sensitive after 0 minutes of chase, but became resistant to digestion after 120 minutes, indicating their successful transport to the Golgi. As the Ca²⁺ depletion had little effect on the folding of MHC Class I, we conclude that the dramatic effect on β1-integrin folding and assembly was due to the requirement for Ca²⁺ during folding of this protein in the ER, rather than a general defect in the protein folding machinery.

Discussion

In this study, we used conformation-sensitive monoclonal antibodies to track the folding and assembly of β1-integrin, and investigate the conformational state of newly synthesised integrin molecules. Our major findings are: (1) β1-integrins adopt a bent, inactive conformation after assembly with an α-integrin subunit in the ER, before transport to the Golgi; (2) correct folding and assembly are dependent on the binding of Ca²⁺ ions, which have to be present from the start of the process; and (c) the integrin molecule remains in an inactive form throughout the secretory pathway. Although the role of cations during the activation of integrins at the cell surface is well established, our results show that the binding of cations is also important to prevent any activation during intracellular trafficking.

Monoclonal antibodies have become essential tools to study the structure and function of integrins because they recognise distinct conformational states of the receptors (Mould, 1996). Many of these antibodies have been used to study changes in the ligand-binding affinity of integrins on the cell surface (Humphries, 2000); however, in this study, we used their specificities to report defined conformational states during β1-integrin biosynthesis. In particular, we noted that two conformation-specific antibodies, 8E3 and 9EG7, which have been shown to recognise the unbent form of β1-integrins, also react with monomeric β1-integrin subunits. 9EG7 was of particular interest because it seems to recognise an epitope that requires the formation of a disulphide and once this disulphide is formed, the epitope is not lost even after denaturation. The epitope has been mapped previously to within a cysteine-rich stretch (residues 495–602) at the back of the β1-integrin knee region (Bazzoni et al., 1995). Because this epitope also appears following Ca²⁺ depletion in the ER, we can conclude that this region of the integrin can still fold under these conditions. The epitope for TS2/16 has been mapped to the β1-integrin A-domain (Hemler et al., 1984; Tsuchida et al., 1997), therefore the dramatic effect on the assembly of β1-integrin and α-integrin in the absence of free Ca²⁺ would indicate that a lack of folding of the A-domain results in a lack of assembly and subsequent transport from the ER. Given that we did not see a dramatic effect on the folding of other ER proteins we can conclude that lack of assembly was the direct result of a lack of Ca²⁺ binding.

Our results demonstrate that the majority of integrin molecules expressed by HT1080 cells are in an inactivate state, as judged by a lack of 9EG7 epitope recognition. This regulation of activation is very important for their biological function, as is most evident from considering integrins present on circulating blood cells. The major platelet integrin αIIbβ3 is present at high density on circulating platelets, where it is inactive. If it were not, platelets would bind their ligand fibrinogen from the plasma and aggregate. Integrins carry signals from the outside to the inside of the cell and vice versa. In inside-out signalling, integrin activity is regulated by binding of regulatory proteins to the short cytoplasmic domains of integrins (Legate and Fassler, 2009). So far, two major protein families, talins and kindlins, have been reported to regulate the activity of integrins from within the cells by binding the tails of the β-integrin subunit (Moser et al., 2009). Because these intracellular proteins can bind to integrins and activate them inside the cells, how cells avoid unwanted intracellular signalling is still an unresolved question. Our results suggest tight control of cell surface molecules by cation binding. A similar study has looked at the expression of 9EG7 on various β1-integrin heterodimers at the cell surface, and demonstrated that removal of Ca²⁺ with EDTA or EGTA induced expression of the 9EG7 epitope (Bazzoni et al., 1998). This is consistent with other inhibitory effects of Ca²⁺ on β1-integrin heterodimers at the cell surface (Sonnenberg et al., 1988; Staatz et al., 1989; Mould et al., 1995). Our results extend these observations and provide evidence that Ca²⁺ helps to stabilise an inactive, bent conformation of β1-integrin heterodimers inside the cell. Ca²⁺ is a predominant ion present in mM concentrations in the ER (Montero et al., 1995) and many ER-resident proteins involved in protein folding bind Ca²⁺ through high-affinity binding sites (Macer and Koch, 1988). This previously unreported intracellular function of Ca²⁺ might allow β1-integrin heterodimers to maintain an inactive state in the presence of high free Ca²⁺ concentrations in the secretory pathway.

Our pulse-chase experiments using TS2/16 and JB1A antibodies show that the β1-integrin associates with the α-integrin before transport from the ER to the Golgi. Such a requirement for assembly before transport is in agreement with previous studies on leukocyte integrin biosynthesis; these showed that the transport of α-integrin and β-integrin subunits from the ER to Golgi is dependent upon the formation of αβ-integrin heterodimers (Ho and Springer, 1983; Kishimoto et al., 1987). In addition, it has been reported that the β-integrin propeller domain of α-integrins and the A-domain of β-integrins associate with one another and both are mutually dependent on the formation of αβ-integrin for folding (Huang et al., 1997; Huang and Springer, 1997). The Ca²⁺-binding motifs in integrin α-subunits are predicted to be close to one another on the lower surface of the β-integrin propeller domain, which, although it is not involved in interaction with the β-integrin subunit, appears to stabilise the tertiary structure of this fold (Tuckwell et al., 1992). It is also very interesting that Ca²⁺ ions are shown to stabilise αβ-integrin association: removal of Ca²⁺ resulted in dissociation of α-integrin and β-integrin in detergent-solubilised αIIbβ3 integrin (Jennings and Phillips, 1982) and made α- and β-integrin subunits susceptible to dissociation by high pH in αLβ2 integrin (Dustin et al., 1992). Our results extend these studies by showing that after the depletion of Ca²⁺ from the ER, the β1-integrin subunit cannot attain its native fold and does not associate with the α-integrin subunit. Furthermore, the transportation of both the α- and β-integrins from the ER to the Golgi was blocked in absence of Ca²⁺ in the ER, emphasising the importance of these ions for the proper folding and trafficking of β1-integrins. Taken together, our data shed light on integrin folding during biosynthesis and provide strong evidence for a key mechanism by which inappropriate intracellular signalling is prevented.

Materials and Methods

Antibodies and reagents

Antibodies binding to human β1-integrin were: TS2/16, which were a gift from F. Sanchez (Hospital de la Princesa, Madrid, Spain), 8E3 (Mould et al., 2005), 9EG7 from Dietmar Vestweber (University of Munster, Munster, Germany) and JB1A from J. Wilkins (University of Manitoba, Winnipeg, Canada). Antibody recognising human α5-integrin subunit, mAB11, was a gift from Kenneth Yamada (NIH, Bethesda, MD). Antibody binding to MHC class I was W6/32 (Abcam) and β2-microglobulin was B2M-01 (Abcam). FITC-conjugated rabbit anti-rat-IgG for flow cytometry was purchased from Serotec and peroxidase-conjugated anti-mouse IgG and extravidin-HRP were purchased from Jackson ImmunoResearch. The recombinant 50K fragment of FN comprising type III repeats 6–10 was made as described previously (Mould et al., 1997).

Metabolic labelling, pulse chase and immunoisolation

Approximately 3×10⁶ HT1080 cells were starved of essential amino acids for 30 minutes before pulse labelling for 30 minutes using 11 μCi/ml of [³⁵S]methionine protein labelling mix (PerkinElmer). Cells were washed and incubated in complete DMEM3 medium for various chase times. Cells were lysed in buffer [50 mM Tris-HCl buffer, pH 7.4 containing 1% (v/v) Triton X-100, 150 mM NaCl, 2 mM ethylenediaminetetraacetic acid (EDTA) and 0.5 mM phenylmethylsulphonyl fluoride (PMSF)]. Clarified lysates were incubated for 1 hour at 4°C with Protein-G–Sepharose beads (Sigma) to remove proteins that may bind non-specifically, before being incubated with antibodies with protein-G–Sepharose beads overnight at 4°C. Beads were washed three times with RIPA buffer (50 mM Tris-HCl buffer, pH 8.0 containing 1% (w/v) deoxycholic acid, 0.1% (w/v) SDS, 1% (v/v) Triton X-100, 500 mM NaCl and 0.5 mM PMSF) before being resuspended in SDS-PAGE sample buffer containing 50 mM DTT and boiling for 5 minutes. Samples were separated by SDS-PAGE under reducing conditions on a 7% Tris-acetate gel, which was dried and exposed to Kodak BioMax MR film (GRI, Braintree, UK).

In vitro transcription and translation

Human β1-integrin was transcribed from a pSPUTK vector linearised with EcoRV. SP6 polymerase was used for transcription reactions. RNA transcripts were translated in a rabbit reticulocyte lysate (Flexilysate, Promega) in the presence of SP-cells and [³⁵S]methionine (PerkinElmer) as described previously (Jessop et al., 2007). Initiation of protein synthesis was allowed to proceed for 5 minutes at 30°C before inhibitionwith 1 mM ATCA (Sigma), followed by incubation at 30°C for 2 hours to allow elongation and post-translational modification. Translation products were immunoisolated with anti-β1-integrin monoclonal antibodies before electrophoresis. Samples were separated by SDS-PAGE under reducing conditions on a 7% Tris-acetate gel, which was dried and exposed to Kodak BioMax MR film (GRI).

Biotinylation of cell surface proteins

Approximately 10⁷ HT1080 cells were detached and resuspended in 1 ml PBS (without Ca²⁺ and Mg²⁺). NHS-sulfo-Biotin (Sigma) was added to cells at 0.25 mg/ml final concentration and incubated at room temperature for 30 minutes. Cells were washed twice with PBS and then with Tris-buffered saline. Cells were divided into four aliquots and then lysis was performed in lysis buffer (see above) containing different cation compositions [without EDTA and with 1 mM CaCl₂, MnCl₂ or MnCl₂ with 50 K fragment of fibronectin (10 μg/ml)] for 15 minutes at 4°C. Lysates were centrifuged at 14,400 g for 5 minutes. Integrin subunits were immunoisolated with anti-α5-integrin or anti-β1-integrin monoclonal antibodies before electrophoresis. Samples were separated by SDS-PAGE and blotting was performed using extravidin-conjugated HRP.

Flow cytometry

Cells were detached, washed and resuspended to a final concentration of 10⁶–10⁷ cells/ml in HEPES-buffered saline (HBS) containing 4.5 g/l glucose and 1% FBS. Cell aliquots (50 μl) were incubated with primary antibody 9EG7 or rat IgG (control) at a final concentration of 10 μg/ml in the same buffer supplemented with 1 mM MnCl₂ or CaCl₂ with or without 10 μg/ml 50 K fragment of fibronectin for 1 hour at 4°C. Cells were washed in the respective buffers three times and isolated by centrifugation at 1100 g for 30 minutes. FITC-conjugated anti-rat secondary antibody (50 μl) diluted to 5 μg/ml in PBS + 1% FBS, was added followed by incubation for a further 30 minutes at 4°C. Cells were washed and made up to a final volume of 300 μl in HBS before immediate analysis on a Dako Cyan flow cytometer using an excitation wavelength of 488 nm and a 530/40 nm emission filter.

Depletion of Ca²⁺ from the ER

To deplete Ca²⁺ from the ER, 100 nM of the Ca²⁺-ATPase inhibitor thapsigargin was added to starvation, pulse, and chase media. The cells were starved for 30 minutes and pulse-labelled for 30 minutes with 11 μCi/ml of [³⁵S]methionine protein labelling mix (PerkinElmer). After various chase times, the cells were treated with 20 mM N-ethylmaleimide (NEM) to block free sulphydryl groups and prevent any disulfide bond formation and lysis was performed in lysis buffer (see above) with 20 mM NEM. After centrifugation the cell lysates were used for immunoisolation as described above.