Structural framework of the GABARAP–calreticulin interface – implications for substrate binding to endoplasmic reticulum chaperones Yvonne Thielmann 1 , Oliver H. Weiergra ¨ ber 1 , Jeannine Mohrlu ¨ der 1,2 and Dieter Willbold 1,2 1 Institut fu ¨ r Neurowissenschaften und Biophysik, Molekulare Biophysik, Forschungszentrum Ju ¨ lich, Germany 2 Institut fu ¨ r Physikalische Biologie und BMFZ, Heinrich-Heine-Universita ¨ tDu ¨ sseldorf, Germany The neurotransmitter 4-aminobutyrate (GABA) medi- ates synaptic inhibition in the brain and the spinal cord [1]. GABA receptors can be categorized into type A (GABA A ) receptors, which are ligand-gated chloride channels, and type B (GABA B ) receptors, which are G-protein-coupled and modulate the activity of potas- sium and calcium channels [2]. GABA A receptors are relevant drug targets for benzodiazepines, barbiturates Keywords 4-aminobutyrate type A receptor-associated protein (GABARAP); calreticulin; protein–protein interaction; structure model; X-ray crystallography Correspondence O. H. Weiergra ¨ ber, Institut fu ¨ r Neurowissenschaften und Biophysik, Molekulare Biophysik, Forschungszentrum Ju ¨ lich, 52425 Ju ¨ lich, Germany Fax: +49 2461 612020 Tel: +49 2461 612028 E-mail: o.h.weiergraeber@fz-juelich.de D. Willbold, Institut fu ¨ r Physikalische Biologie und BMFZ, Heinrich-Heine- Universita ¨ t, 40225 Du ¨ sseldorf, Germany Fax: +49 2461 612023 Tel: +49 2461 612100 E-mail: d.willbold@fz-juelich.de Database The atomic coordinates and structure factor amplitudes (code 3DOW) have been deposited in the Protein Data Bank (http://www.pdb.org) (Received 14 October 2008, revised 2 December 2008, accepted 12 December 2008) doi:10.1111/j.1742-4658.2008.06857.x The 4-aminobutyrate type A receptor-associated protein (GABARAP) is a versatile adaptor protein that plays an important role in intracellular vesi- cle trafficking, particularly in neuronal cells. We have investigated the structural determinants underlying the interaction of GABARAP with cal- reticulin using spectroscopic and crystallographic techniques. Specifically, we present the crystal structure of GABARAP in complex with its major binding epitope on the chaperone. Molecular modeling of a complex con- taining full-length calreticulin suggests a novel mode of substrate interac- tion, which may have functional implications for the calreticulin ⁄ calnexin family in general. Abbreviations CRT(178–188), CH 3 CO-SLEDDWDFLPP-NH 2 ; ER, endoplasmic reticulum; GABA, 4-aminobutyrate; GABARAP, 4-aminobutyrate type A receptor-associated protein; GATE-16, Golgi-associated ATPase enhancer of 16 kDa; HSQC, heteronuclear single quantum coherence; P-domain, proline-rich domain; SPR, surface plasmon resonance; Ubl, ubiquitin-like protein. 1140 FEBS Journal 276 (2009) 1140–1152 ª 2009 The Authors Journal compilation ª 2009 FEBS and general anesthetics [3]. GABA A receptor-associ- ated protein (GABARAP) was initially found in a two-hybrid screen to interact with the cytoplasmic loop connecting transmembrane helices 3 and 4 of the GABA A receptor c2-subunit. This interaction was con- firmed by colocalization experiments in cultured corti- cal neurons and by coimmunoprecipitation of GABARAP with GABA A receptor subunits from brain extracts [4]. GABARAP belongs to a protein family that is evo- lutionarily highly conserved, from yeast to mammals. Atg8 from Saccharomyces cerevisiae has been identified as an essential regulator of the autophagic machinery, which serves to nonselectively sequester cytoplasmic material for vacuolar degradation [5]. Mammalian orthologs of this family include glandular epithelial cell protein 1, Golgi-associated ATPase enhancer of 16 kDa (GATE-16), light chain 3 of microtubule-asso- ciated protein 1, and GABARAP [3]. All these proteins belong to the superfamily of ubiquitin-like proteins (Ubls). They share the charac- teristic b-grasp fold, as first demonstrated by the crystal structure of GATE-16 [6], and are subject to a modification process that is similar to the ubiquitin- type conjugation machinery. After proteolytic cleavage, leading to exposure of a C-terminal glycine residue, these Ubls are coupled to an E1 enzyme via a thioester bond, further transferred from the E1 enzyme to an E2 enzyme, and finally conjugated to phosphatidylserine or phosphatidylethanolamine. Con- sequently, at the end of the conjugation process, GABARAP and related proteins are attached to cellular membranes instead of proteins, as in the case of ubiquitin [7,8]. Available crystal structures [9–11] as well as NMR structures of GABARAP [12] show the expected simi- larity to other Ubls. In GABARAP, the Ubl core domain comprising the b-grasp fold is extended by an N-terminal segment containing two additional a-heli- ces. We have recently determined the first three-dimen- sional structure of GABARAP complexed with a ligand [13]. This structure highlights the interactions of apolar residues of a synthetic peptide with GABA- RAP’s hydrophobic pockets. These pockets were probed previously with indole derivatives [14] and have also been described for GATE-16 [6]. Despite this structural knowledge for GABARAP, data for complexes with its native interaction partners as well as conjugating enzymes are still needed to understand its biological function on a molecular level. We have previously identified calreticulin and the heavy chain of clathrin as potential binding part- ners [15,16]. In the case of calreticulin, immunofluo- rescence staining of neuronal cells revealed significant colocalization with GABARAP in punctuate structures, probably corresponding to a vesicular compartment [15]. Calreticulin is a multifunctional lectin-like 46 kDa Ca 2+ -binding chaperone predominantly located in the endoplasmic reticulum (ER). It is found in a wide range of species and is involved in intracellular Ca 2+ homeostasis as well as ER Ca 2+ storage capacity [17]. Within secretory pathways, it functions as an impor- tant chaperone involved in quality control [18]. Studies on calreticulin knockout mice indicate that the protein is essential for early cardiac development [19]. Recently, cell surface calreticulin has attracted particu- lar attention because of its role as a phagocytic signal on apoptotic cells, implicating the protein in processes such as autoimmunity and cancer [20]. Moreover, it was found to be retrotranslocated from the ER lumen into the cytosol [21], and has been ascribed specific functions in protein transport and gene expression (see Discussion for details). The N-terminal and C-terminal segments of calreticulin are predicted to fold into a composite globular domain, whereas the intervening sequence forms an arm-like structure often referred to as the proline-rich domain (P-domain) [17]. In this study, we investigated the interaction of GABARAP with different calreticulin fragments, including the complete P-domain as well as an undecamer peptide {CH 3 CO-SLEDDWDFLPP-NH 2 [CRT(178–188)]} comprising the principal GABARAP- binding motif [15]. In particular, we determined the three-dimensional structure of the latter peptide associ- ated with the GABARAP molecule. The binding mode of this native ligand turned out to differ significantly from the artificial peptide investigated previously. Moreover, our data provide evidence for additional contacts mediated by the calreticulin P-domain. On the basis of these observations, we present a detailed molecular model of the native complex. Beyond the specifics of this particular interaction, our model offers conceptual insights into the function of the calnexin ⁄ calreticulin family in general. Results Binding constants determined by surface plasmon resonance (SPR) spectroscopy Using SPR, the binding of an analyte in solution to an immobilized partner can be measured directly [22]. Therefore, we investigated the interaction of GABA- RAP with the calreticulin P-domain (amino acids 177– 288) and related peptides with this technique (Fig. 1). Y. Thielmann et al. Complex structure of GABARAP and calreticulin FEBS Journal 276 (2009) 1140–1152 ª 2009 The Authors Journal compilation ª 2009 FEBS 1141 Evaluation of steady-state binding signals yielded dissociation constants of 930 ± 120 nm for the GABARAP–P-domain interaction (Fig. 1A,B) and 11.5 ± 1.1 lm for G ABARAP binding to CR T(178–188 ) (Fig. 1C,D). Comparison of the dissociation constants of both complexes suggests that binding of additional residues not included in the undecamer peptide may account for the higher affinity of the P-domain. In fact, full-length calreticulin binds with an even lower dissociation constant of 64 nm and an estimated mean lifetime of 20 min [15]. Therefore, the globular domain is likely to contribute to the association with GABA- RAP as well. We also investigated a variant of CRT(178–188) in which the tryptophan residue was replaced by alanine [W183A-CRT(178–188)]. Binding of this peptide to GABARAP was not saturable up to a ligand concentration of 1 mm (data not shown). Obviously, the mutation shifted the dissociation constant from 11.5 lm into the millimolar range. We conclude that the tryptophan side chain plays a key role in the affinity of the CRT(178–188)–GABARAP complex. Characterization of complexes by NMR spectroscopy High-resolution liquid-state NMR spectroscopy is a powerful technique for in vitro studies of the structure and dynamics of soluble biological macromolecules. NMR also allows the identification and characteriza- tion of molecular interactions of soluble complexes [23]. 1 H 15 N heteronuclear single quantum coherence (HSQC) experiments performed with GABARAP and the W183A-CRT(178–188) peptide showed only small changes of chemical shifts for distinct amino acids (Fig. 2B). In contrast, incubation with the native CRT(178–188) ligand induced large chemical shift changes throughout the GABARAP spectrum and the disappearance of certain peaks (Fig. 2A). Again, the mutation of Trp183 to alanine in CRT(178–188) had a tremendous effect on the binding properties of the molecule. Similar to the results with CRT(178–188), HSQC titration experiments with GABARAP and the entire P-domain (Fig. 2C) showed large chemical shift changes. In addition, we observed disappearance of AC BD Fig. 1. SPR measurements of calreticulin fragments binding to immobilized GABARAP. (A) Calreticulin P-domain and (C) CRT(178–188) were injected into the flow cell at a range of concentrations (10 n M to 5 lM, and 100 nM to 100 lM, respectively). Sensorgrams are shown in dark gray, with black bars indicating the average response at equilibrium for every concentration. In (B) and (D), the respective average responses (d) are fitted to a 1 : 1 binding model (black curves). Complex structure of GABARAP and calreticulin Y. Thielmann et al. 1142 FEBS Journal 276 (2009) 1140–1152 ª 2009 The Authors Journal compilation ª 2009 FEBS resonances caused by broadening of the line width of the chemical shift. Line broadening was reduced by heating the sample from 25 to 35 °C (data not shown), which probably relates to an increased tumbling rate at the higher temperature. According to the known assignment of native GABARAP resonances, the major binding site for all calreticulin fragments is located in the hydrophobic pockets hp1 (Ile21, Tyr25, Ile32, Lys48, and Leu50) and hp2 (Lys46, Tyr49, Phe60, and Leu63). Structure of the GABARAP–CRT(178–188) complex The three-dimensional structure of the GABARAP– CRT(178–188) complex was investigated by X-ray crystallography. Using poly(ethylene glycol) MME 550 as precipitating agent, we obtained crystals belonging to space group I23, containing one copy of the com- plex in the asymmetric unit. Initial phases were deter- mined by molecular replacement with the crystal structure of GABARAP [9] as a search model, and the structure was refined to 2.3 A ˚ . Several segments in the GABARAP structure display elevated temperature fac- tors and weaker electron density, which indicates enhanced conformational freedom. This applies to the N-terminus as well as the a3–b3 and b3–a4 loops of GABARAP. The N-terminal four residues of the pep- tide ligand (Ser178 to Asp181) could not be built, because the electron density was very sparse in the respective region. A remarkable lattice contact is estab- lished by a Zn 2+ tethering three symmetry-equivalent copies of GABARAP; these molecules contribute resi- dues His69 (no. 1), His99 and Glu101 (no. 2) and Glu112 (no. 3) to ion coordination. Figure 3 shows a sketch of the overall structure of the complex (for a close-up view including GABARAP side chains, see Fig. S1). The GABARAP molecule (shown as a ribbon model) displays a b-grasp fold (light blue), which is A B C Fig. 2. 1 H 15 N-HSQC spectra of GABARAP and calreticulin con- structs. (A) Superimposed HSQC spectra of [ 15 N]GABARAP alone (red contour lines) and in the presence of a stoichiometric equiva- lent of CRT(178–188) (black). Large chemical shift changes appear throughout the spectrum. (B) Superimposed HSQC spectra of [ 15 N]GABARAP alone (red contour lines) and in the presence of a four-fold stoichiometric excess of W183A-CRT(178–188) (blue). Minor chemical shifts of distinct amino acids occur. (C) Superim- posed HSQC spectra of [ 15 N]GABARAP alone (red contour lines) and in the presence of 0.5 (blue) and 1 (green) stoichiometric equiv- alents of the calreticulin P-domain. During titration, large chemical shift differences appear throughout the spectrum. In addition, line broadening is observed. Y. Thielmann et al. Complex structure of GABARAP and calreticulin FEBS Journal 276 (2009) 1140–1152 ª 2009 The Authors Journal compilation ª 2009 FEBS 1143 characteristic for the superfamily of Ubls. This com- pact domain consists of a four-stranded mixed b-sheet (strands labeled b1 through b4) and two a-helices (a3 and a4) packed against its concave surface. A specific feature of the GABARAP family is an extension by two N-terminal helices (a1 and a2) on the convex face of the b-sheet. CRT(178–188) assumes an extended conformation and makes close contact with the GABARAP molecule, burying 490 A ˚ 2 of solvent-acces- sible surface. The central part of the ligand (Asp184 to Leu186; gray in Fig. 3) forms main chain hydrogen bonds with strand b2 of GABARAP (Lys48 and Leu50), and can thus be thought of as an intermolecu- lar extension of the central b-sheet. In contrast, the terminal peptide segments (dark blue) are engaged in side chain hydrogen bonds to Lys48, Glu17 and Arg28 of GABARAP. Overall, the interaction between CRT(178–188) and GABARAP appears to be domi- nated by hydrophobic contacts established by Trp183, Phe185 and Leu186 of the peptide. The individual side chains involved are listed in Table 1. The calreticulin peptide is anchored by the indole moiety of Trp183, which contacts residues from helix a2, strands b1 and b2 and the a4–b4 loop (hp1, see below). The side chain of Phe185 reaches out across strand b2, interacting with apolar side groups from the a2–b1 loop. Finally, the C-terminal part of CRT(178–188) is held in posi- tion by hydrophobic contacts of Leu186 with strand b2, helix a3 and the b2–a3 loop (hp2). As expected, the GABARAP residues involved in ligand binding agree well with those displaying medium to slow exchange rates in our NMR experiments (included in Table 1). Notably, the hydrophobic pock- ets engaged in complex formation of GABARAP and CRT(178–188) are also crucial for the GABARAP–K1 peptide complex (see below for details). Conformational changes upon complex formation The substantial structural knowledge available for GABARAP [9–13] enables us to delineate the require- ments and consequences of complex formation. Figure 4 shows an alignment of nonliganded GABA- RAP [12] (light gray) with the GABARAP–K1 peptide complex [13] (shades of red) and the complex investi- gated in this study (shades of blue). (For this compari- son, the solution structure of nonliganded GABARAP (1KOT) was preferred over available X-ray structures (1GNU, 1KJT), because the latter contain a lattice contact that partially mimics the effect of ligand bind- ing. In contrast, crystal packing interactions in the K1 and CRT(178–188) complexes do not involve the hydrophobic surface of GABARAP, suggesting that this part of the structure should be relatively unaf- fected.) The overview at the top gives an impression of the overall variation among the three structures. The most significant backbone displacements occur in Fig. 3. Overview of the GABARAP–CRT(178–188) complex. GABA- RAP is depicted as a ribbon model with the b-grasp domain and the N-terminal extension colored in light blue and light gray, respec- tively. The ligand backbone is shown in dark blue (terminal seg- ments) and gray (b-strand). The apolar side groups docking to GABARAP are drawn in stick mode (gold). Table 1. Overview of hydrophobic interactions between GABARAP and CRT(178–188), as revealed by the crystal structure, and extent of chemical shift changes of GABARAP resonances in the corre- sponding 1 H 15 N-HSQC experiment. +, minor effect; ++, large chemical shift change; +++, absence of peak from spectrum; NA, not applicable, since prolines do not appear in 1 H 15 N-HSQC spectra; NE, not evaluated due to spectral overlap. CRT(178–188) GABARAP Contacts in crystal structure Effects in HSQC experiment Trp183 Ile21 ++ Pro30 NP Ile32 ++ Lys48 +++ Leu50 +++ Phe104 +++ Phe185 Arg28 + Tyr25 +++ Leu186 Tyr49 +++ Val51 NE Phe60 ++ Leu63 ++ Ile64 NE Arg67 ++ Complex structure of GABARAP and calreticulin Y. Thielmann et al. 1144 FEBS Journal 276 (2009) 1140–1152 ª 2009 The Authors Journal compilation ª 2009 FEBS helix a3 and the adjacent a3–b3 loop. A more detailed view of the alignment is given in the bottom panels of Fig. 4 for the two hydrophobic pockets. Binding of Trp183 (dark blue) in hp1 induces a slight shift of helix a2 (blue), similar to the effect of Trp11 in the K1 peptide (dark red). Notably, the two complexes differ in both the position and conformation of these trypto- phan side chains. Specifically, Trp183 extends deeper into the pocket, and this is reflected by the side chain configuration of Lys48 and Phe104, which are altered most notably as compared to the GABARAP–K1 complex. Binding of Phe185 does not have obvious consequences for the conformation of either hp1 or hp2. In contrast, Leu186 (dark blue) leads to a large displacement of helix a3 (blue). Binding of this leucine alone exhibits almost the same effect as was reported for Trp6 and Leu9 in the GABARAP–K1 complex (shades of red), resulting in hp2 assuming an open conformation. This spatial rearrangement appears to be chiefly mediated by the displacement of Leu63. On the other hand, the side chain conformation of Arg67 remains similar to that of the nonliganded protein, thus not exposing additional hydrophobic surface, which is needed in the GABARAP–K1 complex for insertion of Trp6 (dark red) into hp2. Model of the GABARAP–calreticulin interaction Unfortunately, attempts to cocrystallize GABARAP with the entire calreticulin molecule or the P-domain have been unsuccessful. In order to gain more insight into the three-dimensional arrangement of the native complex, we have built a homology model incorporat- ing available data on the soluble portion of calnexin [24] and the calreticulin P-domain [25], in addition to the GABARAP–CRT(178–188) complex investigated in this study (Fig. 5). The GABARAP-binding epitope on calreticulin is located at the N-terminal junction between the globular domain and the arm domain. Intriguingly, the corresponding residues could not be resolved in the X-ray structure of calnexin serving as the major template, suggesting significant conforma- tional freedom in this region. In agreement with this notion, our model indicates that, at least when com- plexed with GABARAP (blue), this portion of calreti- culin forms a protrusion emerging from the base of the hp1 hp2 Fig. 4. Comparison of GABARAP structures. Top: overview of nonliganded and liganded GABARAP structures in ribbon and coil rep- resentation: light and dark blue, GABARAP and CRT(178–188) (this study); light and dark red, GABARAP and K1 peptide [13]; light gray, nonliganded GABARAP [12]. Bottom: detailed view of hydrophobic pockets hp1 and hp2. The structures are depicted as above, with selected side chains taking part in the interaction appearing as stick models. For visual clarity, additional GABARAP side chains involved in the interaction (Glu17, Tyr25, Arg28, Pro30, Ile32, Tyr49, Leu50, Val51, Phe60 and Ile64) have been omitted. Y. Thielmann et al. Complex structure of GABARAP and calreticulin FEBS Journal 276 (2009) 1140–1152 ª 2009 The Authors Journal compilation ª 2009 FEBS 1145 arm domain, largely devoid of tertiary interactions with neighboring segments (Fig. 5A,C). We therefore propose that the residues forming the N-terminal junc- tion between the two domains of calreticulin (as well as calnexin) constitute a versatile interaction site that may be adapted to accommodate a variety of ligands. The potential implications for the structure and func- tion of these chaperones are discussed below. Figure 5B,D includes surface representations of GABARAP (light gray), with colored patches denoting residues with major shifts in our HSQC titrations with CRT(178–188) (blue) and the calreticulin P-domain (red). Localization of these amino acids is consistent with the spatial arrangement of GABARAP and calreticulin in our model. Discussion Members of the GABARAP family of Ubls have been implicated in several aspects of membrane vesicle traf- ficking in eukaryotic cells. An important step towards understanding these functions at a molecular level was the discovery of a peculiar conjugation mechanism resulting in covalent linkage of these proteins to mem- brane lipids [7,8]. On the other hand, knowledge of the protein–protein interactions implicated in the various biological roles of GABARAP and its relatives is only beginning to emerge. In a search for novel cellular targets of GABARAP, we have recently identified calreticulin and the heavy chain of clathrin as potential binding partners [15,16]. In both cases, binding activity could be narrowed down to short contiguous peptide sequences comprising 11 and 13 amino acids, respec- tively, centered on a hydrophobic motif (WxFL). In the current study, we present the first three- dimensional structure of GABARAP complexed with a fragment of such a proposed physiological ligand. Our data largely confirm previous assignments of hydro- phobic patches on the surface of GABARAP, which we have shown to interact with indole derivatives as well as a high-affinity artificial ligand (K1) [13]. Although a detailed analysis of the GABARAP– CRT(178–188) interface reveals significant differences with respect to the K1 peptide, both complexes are critically dependent on the presence of at least one tryptophan side chain in the ligand. In its central part, the calreticulin peptide assumes an extended conforma- tion, aligning parallel to strand b2 of GABARAP. Protein–protein interactions via formation of inter- molecular b-sheets have been observed for several AB CD Fig. 5. Model of the GABARAP–calreticulin interaction, shown in two orientations. (A, C) Overview of the GABARAP–calreticulin complex; GABARAP is shown in light blue and the CRT(178–188) segment in dark blue, with the apolar side chains drawn as stick models (gold); the globular domain and P-domain of calreticulin are depicted in light and dark red, respectively. The calreticulin P-domain bends around the bound GABARAP molecule. (B, D) Detailed view of the GABARAP surface (light gray) in complex with calreticulin, oriented as in (A) and (C), respectively. Blue surface patches indicate the GABARAP residues that are most strongly affected in HSQC spectra in the presence of CRT(178–188); additional candidates found with the calreticulin P-domain are marked in red. Complex structure of GABARAP and calreticulin Y. Thielmann et al. 1146 FEBS Journal 276 (2009) 1140–1152 ª 2009 The Authors Journal compilation ª 2009 FEBS members of the ubiquitin superfamily [26,27]. Intrigu- ingly, one of the GABARAP crystal structures has revealed self-association by a similar mechanism, with the N-terminal six amino acids binding to strand b2of a neighboring molecule [11]. What is the biological significance of the GABA- RAP–calreticulin complex? Although its high affinity and estimated lifetime are clearly indicative of a relevant interaction, definition of the precise biochem- ical context in which it naturally occurs has remained a challenge. As long as direct experimental evidence for a biological function of this complex is missing, even fortuitous binding cannot be completely excluded. This seems rather unlikely, however, given that the two proteins not only interact with apprecia- ble affinity in vitro, but also colocalize in vivo. Inter- estingly, conventional knowledge indicates that the subcellular locations of these two proteins should be mutually exclusive: GABARAP has largely been found associated with intracellular membranes [28], and the lack of sorting signals together with the C-terminal conjugation mechanism suggests that it is linked to phospholipids on the cytosolic leaflet of such membranes. Calreticulin, on the other hand, is well known as a soluble chaperone of the ER lumen [17]. However, the protein is in fact not restricted to the ER, but does exert important functions in other cellular compartments, such as the cytosol [29], the nucleus [30] and the plasma membrane [20]. Impor- tantly, calreticulin found at these locations appears to be derived from the ER pool; export into the cytosol is accomplished by a retrotranslocation process that is distinct from the pathway taken by misfolded proteins leading to ubiquitination and proteasomal degradation [21]. On the basis of these findings, we shall discuss several cellular processes that may be envisaged as involving the formation of a GABARAP–calreticulin complex. Export of the N-cadherin–b-catenin complex from the ER has been shown to be dependent on PX-RICS (a GTPase-activating protein acting on Cdc42) and its interaction partner GABARAP. In HeLa cells expressing GABARAP and PX-RICS, knockdown of either protein with short hairpin RNA prevented transport of N-cadherin to sites of cell–cell contact. Exogenous expression of the respective components restored the subcellular distribution of N-cadherin and b-catenin [31]. On the other hand, N-cadherin is downregulated in calreticulin-deficient mouse embry- onic hearts. This may contribute to the disorganiza- tion in myocardial architecture that led to death of the embryos mostly between day 12 and day 14 post conception [19]. Conversely, if calreticulin is overex- pressed in fibroblasts, the N-cadherin protein level is doubled as compared to control cells [32]. Taken together, both GABARAP and calreticulin are involved in a process that enriches N-cadherin in the plasma membrane at cell–cell contacts. According to these data, it is attractive to speculate that GABA- RAP may recruit calreticulin to the cytosolic surface of transport vesicles carrying N-cadherin. Similar considerations may hold in the case of inte- grins. Calreticulin has been shown to associate with a 3 b 1 integrin dimers, and the interaction site has been mapped to a conserved motif in the intracellular domain of the integrin a-subunit, thus clearly involving cytosolic calreticulin [33]. At the same time, a 3 b 1 inte- grins colocalize with GABA A receptors, suggesting a possible connection to GABARAP [34]. It seems conceivable that calreticulin may travel to the plasma membrane on the cytosolic surface of GABARAP- tagged vesicles loaded with either GABA A receptors or integrins (or both). For the two scenarios discussed above, the func- tional role of calreticulin in complex with GABARAP is still unclear, but may be speculated to involve Ca 2+ - dependent regulation of subsequent protein–protein interaction or membrane fusion events. Recent investigations have established that low amounts of calreticulin are exposed on the plasma membrane of most cell types. Intriguingly, surface expression was found to be significantly enhanced by cellular stress, acting as a potent ‘eat me’ signal stimu- lating clearance of apoptotic cells [35]. Moreover, upregulation of calreticulin exposure in tumor cells by certain antineoplastic drugs has been shown to enhance phagocytosis and tumor antigen cross-presen- tation by dendritic cells [20]. Saturable binding of exogenous calreticulin to the surfaces of viable as well as apoptotic cells [35] indicates the presence of specific receptors. On the basis of our results, we speculate that GABARAP (or another member of its family) may constitute such a receptor, tethering calreticulin to membranes via its phospholipid linkage. However, this concept requires that GABARAP should be present on the lumenal leaflet of the ER membrane, which is topographically equivalent to the outer surface of the plasmalemma. Such a localization cannot be excluded, although GABARAP and its relatives do not contain obvious sorting signals. Irrespective of the precise physiological role of the GABARAP–calreticulin complex, its structure sheds light on a general aspect of calnexin and calreticulin function. It is widely accepted that these chaperones provide at least two distinct sites for interaction with folding intermediates in the ER lumen [17]. A specific Y. Thielmann et al. Complex structure of GABARAP and calreticulin FEBS Journal 276 (2009) 1140–1152 ª 2009 The Authors Journal compilation ª 2009 FEBS 1147 binding pocket for Glc 1 Man 9 GlcNAc 2 oligosaccarides has been identified on the surface of the globular domain. In contrast, a general polypeptide interaction site, which is independent of glycosylation, has been postulated, but its location has not been established so far. Such a site can be expected to expose hydrophobic side chains, conferring the ability to stabilize folding intermediates and to prevent them from aggregating. Indeed, the structure of CRT(178–188) in complex with GABARAP reveals a significant hydrophobic interface. Its position within the overall structure of calreticulin, at the socket of the arm-like domain, makes this segment a particularly favorable candidate for a substrate recognition site, as it would provide access to important chaperoning and refolding activi- ties associated with calreticulin. Specifically, it is located in the vicinity of the carbohydrate-binding pocket on the globular domain and of the protein disulfide isomerase ERp57, which is bound to the distal part of the arm domain [36]. Although the precise orientation of the enzyme in this complex is still unknown, the remarkable flexibility of the arm domain, as demonstrated by NMR spectroscopy [25], is likely to enable it to accommodate substrate mole- cules of different size and shape. By virtue of its over- all concave surface, the chaperone is believed to shield the folding intermediate from its surrounding, thus reducing formation of aggregates. The calnexin seg- ment corresponding to CRT(178–188) differs in sequence, but displays significant hydrophobic charac- ter as well. As pointed out previously, the binding motif of calreticulin considered here is remarkably con- served between organisms as diverse as slime molds (exemplified by Dictyostelium discoideum), insects (Drosophila melanogaster) and vertebrates (Homo sapiens) [14]. As this similarity even extends to higher plants (such as Arabidopsis thaliana), it is likely to reflect a fundamental function of calreticulin acquired during early eukaroytic evolution. Besides interaction with substrate proteins, such a function may also involve the formation of specific complexes with other chaperones. In summary, these considerations provide a possible structural foundation for the well-documented affinity of calnexin and calreticulin for partially unfolded poly- peptides. Numerous proteins expressed in the ER have been shown to preferentially interact with one of these chaperones. Among other reasons, this may be related to the differences in the apolar sequence discussed above. Current evidence indicates that the calreticulin frac- tion retrotranslocated into the cytosol exerts distinct functions that are unrelated to the rather promiscuous activities within the ER lumen. Indeed, the chemical milieu in the two compartments is strikingly different, the most prominent example being the Ca 2+ concen- tration, which is four orders of magnitude higher in the ER. Along these lines, it seems conceivable that a general recognition site for partially folded polypep- tides, which plays a crucial role in the chaperoning function of calreticulin, might have been readapted for specific protein–protein interactions in the cytosolic environment, such as with GABARAP. The structural details of such complexes are now beginning to be unraveled. Experimental procedures Expression and purification of proteins The expression and purification of GABARAP have been described previously [37]. The calreticulin P-domain (amino acids 177–288) coding sequence was cloned into pGEX-6P-2 (GE Healthcare, Munich, Germany) using BamH1 and Xho1 restriction sites, and expressed in the Escherichia coli BL21 plysS strain transformed with the plasmid. After affinity purification using glutathione–Sepha- rose 4B (GE Healthcare), the fusion protein of glutathione S-transferase and the P-domain was cleaved with PreScis- sion protease (GE Healthcare). The final purification step was size exclusion chromatography using a Superdex 75 matrix (GE Healthcare). The correct molecular mass was verified by MS. Peptide synthesis The two peptides CRT(178–188) and W183A-CRT(178– 188) were custom synthesized and purified to > 95% by the BMFZ at the University of Du ¨ sseldorf and Jerini BioTools (Berlin, Germany), respectively. SPR spectroscopy SPR studies were carried out on a BiacoreX optical biosen- sor (GE Healthcare). Following the standard procedure of the manufacturer for amine coupling, 1.5 lm GABARAP protein in 10 mm sodium acetate buffer (pH 5.5) was used to perform coupling to the carboxymethylated dextran matrix of a CM5 sensor chip surface. A reference surface was treated identically, but was not exposed to GABARAP for immobilization. Experiments were performed in 10 mm Hepes (pH 7.4), 150 mm NaCl, 3 mm EDTA and 0.005% surfactant P20, using various concentrations of calreticulin P-domain and peptides at a flow rate of 30 lLÆmin )1 at 21.5 °C. Biosensor data were prepared by double referenc- ing [38]. The biaevaluation software package was used for data analysis. Complex structure of GABARAP and calreticulin Y. Thielmann et al. 1148 FEBS Journal 276 (2009) 1140–1152 ª 2009 The Authors Journal compilation ª 2009 FEBS NMR spectroscopy All NMR spectra were recorded on a Varian (Darmstadt, Germany) Unity INOVA spectrometer at a proton frequency of 600 MHz with a Varian Gen 2 HCN cryogenic probe. The sample for the CRT(178–188) binding experiment con- tained 600 lm [ 15 N]GABARAP and 600 lm CRT(178–188) in 25 mm sodium phosphate (pH 7.0), 100 mm KCl, 100 mm NaCl and 5% (v ⁄ v) deuterium oxide. For the experiment with W183A-CRT(178–188), 170 lm [ 15 N]GABARAP and 690 lm W183A-CRT(178–188) were used. These spectra, together with a third one of 200 lm [ 15 N]GABARAP with- out ligand, were recorded at 10 °C. The buffer conditions for experiments with GABARAP and calreticulin P-domain were 25 mm sodium phosphate (pH 7.0), 100 mm NaCl, 3mm EDTA and 7% (v ⁄ v) deuterium oxide. The initial concentration of [ 15 N]GABARAP was 680 lm, and final concentrations were 365 l m [ 15 N]GABARAP and 400 lm P-domain; spectra were recorded at 25 °C. Data were processed with nmrpipe [39] and analyzed with cara [40]. Crystallization The GABARAP–CRT(178–188) complex was prepared by combining 700 lm protein and 770 lm peptide in 10 mm Tris-HCl (pH 7.0). Cocrystallization was achieved using the hanging-drop vapor diffusion method, with the reservoir containing 0.1 m Mes (pH 6.5), 27% (v ⁄ v) poly(ethylene glycol) MME 550 and 10 mm ZnSO 4 . Data collection The X-ray diffraction dataset was collected at 100 K. Prior to cryocooling, crystals were soaked once in a reservoir solution containing 29% (v ⁄ v) poly(ethylene glycol) MME 550 and 5% (v ⁄ v) glycerol. A single-wavelength native dataset was recorded at beam- line ID14-1 of the ESRF (Grenoble, France) tuned to a wavelength of 0.934 A ˚ on an ADSC-Q4R detector. Data processing was carried out with the ccp4 [41] software suite using mosflm and scala. Structure determination Cocrystals of GABARAP and CRT(178–188) belonged to space group I23. The structure was determined by mole- cular replacement using molrep (ccp4) with a single native dataset. The search model was created from the crystal structure of GABARAP (Protein Data Bank code: 1KJT) [9]. Crystals were found to contain one copy of the complex per asymmetric unit, corresponding to a Matthews coeffi- cient of 2.47 A ˚ 3 ÆDa )1 and a solvent content of 50.2%. Fol- lowing rigid-body refinement using the cns [42] package, the model was improved by iterative cycles of manual rebuilding using the program o [43] and refinement with cns. Later stages of refinement and assignment of water molecules were carried out with the phenix [44] package. In order to avoid overfitting in light of the moderate observa- tions-to-parameters ratio, the relative weight of stereochem- ical restraints was increased, resulting in a comparatively low deviation of geometric parameters from library targets. For statistics on data collection and refinement, see Table 2. The final model contains amino acids 1–117 of native GABARAP with an additional N-terminal glycine–serine extension and amino acids 182–188 of the calreticulin ligand. The N-terminal cloning artefact (glycine–serine) is omitted from residue numbering throughout this article. Note that all amino acids of the crystallized protein have to be considered in the Protein Data Bank entry, resulting in a +2 shift in residue numbers. According to Ramachandran plots generated with mol- probity (http://molprobity.biochem.duke.edu), the model exhibits good geometry with one residue (His69) in the dis- allowed region. This histidine is involved in coordination of aZn 2+ at the interface of three GABARAP molecules. Comparative modeling The molecular model of the human GABARAP–calreticulin complex was built with the modeller package [45], using two distinct templates. The template for the calreticulin moiety was created from the crystal structure of canine calnexin (Protein Data Bank code 1JHN) [24] by manually Table 2. Data collection and refinement statistics. Values in paren- theses are for the highest-resolution shell (2.42–2.30 A ˚ ). Data collection Space group I23 Cell dimensions a (A ˚ )(T = 100 K) 97.04 Resolution range (A ˚ ) 39.6–2.3 Beamline ESRF ID14-1 Detector ADSC-Q4R Wavelength (A ˚ ) 0.934 R sym (%) 5.2 (39.9) I ⁄ r(I) 24.5 (4.9) Completeness (%) 99.7 (100.0) Redundancy 6.8 (6.6) Refinement No. reflections 6892 R work (%) 23.2 R free (%) 27.0 No. atoms Protein 1036 Ion 1 Solvent 29 rmsd Bond lengths (A ˚ ) 0.003 Bond angles (°) 0.6 Y. Thielmann et al. Complex structure of GABARAP and calreticulin FEBS Journal 276 (2009) 1140–1152 ª 2009 The Authors Journal compilation ª 2009 FEBS 1149 [...]... representation of the binding interface Alignment of the calreticulin sequence to the calnexin-based template was performed using the structure-sensitive algorithm implemented in modeller Ten models of the complex were built in a single run, employing the automodel environment Out of these, the structure with the most favorable values in the modeller objective function as well as the discrete optimized protein... GABAA-receptor-associated protein links GABAA receptors and the cytoskeleton Nature 397, 6 9–7 2 5 Ohsumi Y (2001) Molecular dissection of autophagy: two ubiquitin-like systems Nat Rev 2, 21 1–2 16 1150 6 Paz Y, Elazar Z & Fass D (2000) Structure of GATE16, membrane transport modulator and mammalian ortholog of autophagocytosis factor Aut7p J Biol Chem 275, 2544 5–2 5450 7 Ichimura Y, Kirisako T, Takao T, Satomi... K (2001) NMR structure of the ¨ calreticulin P-domain Proc Natl Acad Sci USA 98, 313 3–3 138 26 Stebbins CE, Kaelin WG & Pavletich NP (1999) Structure of the VHL–ElonginC–ElonginB complex: implications for VHL tumor suppressor function Science 284, 45 5–4 61 27 Huang L, Hofer F, Martin GS & Kim SH (1998) Structural basis for the interaction of Ras with RalGDS Nat Struct Biol 5, 42 2–4 26 28 Kittler JT, Rostaing... structure of GABARAP and calreticulin Y Thielmann et al replacing the two distal modules of the arm domain by the terminal portion of the rat calreticulin arm domain solved by NMR spectroscopy (Protein Data Bank code 1HHN) [25] The X-ray structure of human GABARAP complexed with amino acids 18 2–1 88 of human calreticulin (this study) was introduced as a second template for accurate representation of the binding. .. insights from the interfacial and thermodynamic properties of hydrocarbons Proteins 11, 28 1–2 96 Supporting information The following supplementary material is available: Fig S1 Close-up view of the GABARAP–CRT(17 8– 188) complex This supplementary material can be found in the online version of this article Please note: Wiley-Blackwell is not responsible for the content or functionality of any supplementary... ¨ ¨ Willbold D (2007) Identification of clathrin heavy chain as a direct interaction partner for the c-aminobutyric acid type A receptor associated protein Biochemistry 46, 1453 7–1 4543 17 Gelebart P, Opas M & Michalak M (2005) Calreticulin, a Ca2+ -binding chaperone of the endoplasmic reticulum Biochem Cell Biol 37, 26 0–2 66 FEBS Journal 276 (2009) 114 0–1 152 ª 2009 The Authors Journal compilation ª 2009... Calreticulin exposure dictates the immunogenicity of cancer cell death Nat Med 13, 5 4–6 1 21 Afshar N, Black BE & Paschal BM (2005) Retrotranslocation of the chaperone calreticulin from the endoplasmic reticulum lumen to the cytosol Mol Cell Biol 25, 884 4–8 853 22 Karlsson R, Roos H, Fagerstam L & Persson B (1994) ¨ Kinetic and concentration analysis using BIA technology Methods 6, 9 9–1 10 23 Stangler T, Hartmann... Driscoll PC & Keep NH (2002) The X-ray crystal structure and putative ligand-derived peptide binding properties of c-aminobutyric acid receptor type A receptor-associated protein J Biol Chem 277, 555 6–5 561 11 Coyle JE, Qamar S, Rajashankar KR & Nikolov DB (2002) Structure of GABARAP in two conformations: implications for GABAA receptor localization and tubulin binding Neuron 33, 6 3–7 4 12 Stangler T, Mayr... (2008) ¨ An indole -binding site is a major determinant of the ligand specificity of the GABA type A receptor-associated protein GABARAP ChemBioChem 11, 176 7– 1775 15 Mohrluder J, Stangler T, Hoffmann Y, Wiesehan K, ¨ Mataruga A & Willbold D (2007) Identification of calreticulin as a ligand of GABARAP by phage display screening of a peptide library FEBS J 274, 554 3–5 555 16 Mohrluder J, Hoffmann Y, Stangler... Forschungsgemeinschaft (DFG) to D Willbold (Wi1472 ⁄ 5) References 1 Moss SJ & Smart TG (2001) Constructing inhibitory synapses Nat Rev Neurosci 2, 24 0–2 50 2 Enna SJ & Mohler H (2007) The GABA Receptors ¨ Humana Press, Totowa, NJ 3 Chen Z-W & Olsen RW (2007) GABAA receptor associated proteins: a key factor regulating GABAA receptor function J Neurochem 100, 27 9–2 94 4 Wang H, Bedford FK, Brandon NJ, Moss . Structural framework of the GABARAP–calreticulin interface – implications for substrate binding to endoplasmic reticulum chaperones Yvonne Thielmann 1 , Oliver. detailed view of the alignment is given in the bottom panels of Fig. 4 for the two hydrophobic pockets. Binding of Trp183 (dark blue) in hp1 induces a slight shift of helix a2 (blue), similar to the effect. throughout the GABARAP spectrum and the disappearance of certain peaks (Fig. 2A). Again, the mutation of Trp183 to alanine in CRT(17 8–1 88) had a tremendous effect on the binding properties of the molecule.