Expression and characterization of soluble forms of the extracellular domains of the β, γ and ε subunits of the human muscle acetylcholine receptor

Authors


S. J. Tzartos, Department of Biochemistry, Hellenic Pasteur Institute, GR11521 Athens, Greece
Fax: +30 210 6478842
Tel: +30 210 6478844 or +30 2610 969955
E-mail: tzartos@mail.pasteur.gr, tzartos@upatras.gr

Abstract

The nicotinic acetylcholine receptor (AChR) is a ligand-gated ion channel found in muscles and neurons. Muscle AChR, formed by five homologous subunits (α2βγδ or α2βγε), is the major antigen in the autoimmune disease, myasthenia gravis (MG), in which pathogenic autoantibodies bind to, and inactivate, the AChR. The extracellular domain (ECD) of the human muscle α subunit has been heterologously expressed and extensively studied. Our aim was to obtain satisfactory amounts of the ECDs of the non-α subunits of human muscle AChR for use as starting material for the determination of the 3D structure of the receptor ECDs and for the characterization of the specificities of antibodies in sera from patients with MG. We expressed the N-terminal ECDs of the β (amino acids 1–221; β1–221), γ (amino acids 1–218; γ1–218), and ε (amino acids 1–219; ε1–219) subunits of human muscle AChR in the yeast, Pichia pastoris. β1–221 was expressed at ≈ 2 mg·L−1 culture, whereas γ1–218 and ε1–219 were expressed at 0.3–0.8 mg·L−1 culture. All three recombinant polypeptides were glycosylated and soluble; β1–221 was mainly in an apparently dimeric form, whereas γ1–218 and ε1–219 formed soluble oligomers. CD studies of β1–221 suggested that it has considerable β-sheet secondary structure with a proportion of α-helix. Conformation-dependent mAbs against the ECDs of the β or γ subunits specifically recognized β1–221 or γ1–218, respectively, and polyclonal rabbit antiserum raised against purified β1–221 bound to 125I-labeled α-bungarotoxin-labeled human AChR. Moreover, immobilization of each ECD on Sepharose beads and incubation of the ECD–Sepharose matrices with MG sera caused a significant reduction in the concentrations of autoantibodies in the sera, showing specific binding to the recombinant ECDs. These results suggest that the expressed proteins present some near-native conformational features and are thus suitable for our purposes.

Abbreviations
AChR

nicotinic acetylcholine receptor

ECD

extracellular domain

MG

myasthenia gravis

β1–221

amino acids 1–221 of the human AChR β subunit

γ1–218

amino acids 1–218 of the human AChR γ subunit

ε1–219

amino acids 1–219 of the human AChR ε subunit

The nicotinic acetylcholine receptor (AChR) is a member of the superfamily of ligand-gated ion channels, which also includes the glycine, γ-aminobutyric acid A, and 5-HT3 receptors [1]. Its physiological role is to mediate the fast chemical transmission of electrical signals in response to acetylcholine released from the nerve terminal to the end-plate.

The muscle AChR is a transmembrane glycoprotein (≈ 290 kDa) located on the postsynaptic membrane of the neuromuscular junction and is composed of five homologous subunits in the stoichiometry α2βγδ (embryonic muscle) or α2βεδ (adult muscle), with the subunits arranged around a central ion pore [2,3]. Each mature subunit (after cleavage of the signal peptide) consists of three domains, an extracellular domain (ECD) (210–220 residues), a membrane-spanning domain, and an intracellular domain [3]. The N-terminal ECD of each of the two α subunits contains the major part of the binding site for the cholinergic ligands. The two sites are nonequivalent, one being formed at the interface between one α subunit and the γ/ε subunits and the other between the second α subunit and the δ subunit [4]. The γ/ε and δ subunits play a major role in shaping the ligand-binding sites and also in maintaining cooperative interactions between the α subunits [5–7]. The β subunit is an important determinant in receptor localization, as shown by studies on the properties of hybrid muscle AChRs, in which the muscle β subunit was replaced by its neuronal counterpart [8].

In addition to its physiological function, the muscle AChR is involved in the pathology of the autoimmune disease, myasthenia gravis (MG), being the main antigen against which MG autoantibodies are produced. These autoantibodies bind to AChR molecules at the neuromuscular junction, leading to their loss and the weakness and fatigability of the voluntary muscles, the main symptoms of MG [9]. A proportion of patients lacking autoantibodies against the AChR harbors antibodies against the muscle-specific kinase, MuSK [10].

The pathophysiological importance of the AChR necessitates the solution of its 3D structure. Current knowledge of its structure is mainly based on data from electron images of the AChR found in large amounts in the electric organ of the marine ray, Torpedo californica[3]. The acquisition of the crystallographic structure of the mollusc acetylcholine-binding protein [11] has provided an insight into the ligand-binding domain of nicotinic receptors. However, the fact that this protein is most closely related to the α7 subunit of the neuronal AChR (24% identity of amino acids) than each of the muscle AChR subunits (22% on average) necessitates the solution of the structure of the mammalian AChR molecule. A prerequisite for this is the availability of large amounts of native, soluble AChR molecules, a target that can be partially achieved by expression of the ECDs of the AChR subunits in heterologous expression systems. Several studies have been carried out on the expression of the muscle-type α subunit ECD in bacterial systems, in which the protein is expressed in large amounts, but is unglycosylated and forms inclusion bodies, requiring refolding to allow partial renaturation [12–14]. Other studies involved the expression of different subunits (whole subunits or ECDs) in mammalian systems, in which the protein has the correct structure, but is only produced in limited amounts because of the inherent difficulty in scaling up expression in cell culture or oocytes [15,16].

In this report, we present the expression and characterization of the ECDs of the β, γ and ε subunits of the human muscle AChR. We describe their expression in a soluble, glycosylated form and in satisfactory amounts using the yeast Pichia pastoris expression system, which combines the speed of bacterial systems with the advantages of eukaryotic expression systems (e.g. post-translational modification) and which had been successfully used in the past by our group to express the ECDs of human muscle α1 [17] and human neuronal α7 [18]. CD analysis of amino acids 1–221 of the human AChR β subunit (β1–221) showed that the protein has a β-structure with a contribution from α-helices. Two conformation-dependent mAbs (one anti-β and one anti-γ) specifically bound to their cognate ECDs, whereas autoantibodies in MG sera, the binding of which is highly conformation-dependent [19,20], bound to all three ECDs.

As all three ECDs were expressed in satisfactory amounts and were recognized by human MG autoantibodies, they may be suitable as starting material for preliminary biophysical and structural studies and for the study of MG.

Results

Rationale for the construction and testing of AChR ECD variants

N-Terminal addition of the FLAG peptide (DYKDDDDK) or addition of the first transmembrane amino acid of the mouse muscle α subunit, a proline, which is conserved in human AChR subunits, results in higher expression of the mouse muscle α ECD [21]. To test the effect of these additional epitopes/tags on the yield of the present proteins, we constructed a set of eight human γ ECD variants (γ, amino acids 1–218) with or without a proline at position 219 and/or the FLAG epitope and/or a 6-His tag (6-HIS) (Fig. 1A). We then performed small-scale cultures for each protein and quantified the amounts of expressed protein in the culture supernatant using dot-blots and a series of supernatant dilutions. Expression varied depending on the presence of the different modifications (Fig. 1B). The yield of amino acids 1–218 of the human AChR γ subunit (γ1–218) without additional tags was taken as the 100% reference (≈ 0.3 mg·L−1, see below). Addition of the proline residue had no significant effect on expression (less than 10%). The presence of the FLAG tag increased expression of γ1–218 by almost 20%, but did not improve expression of γ1–219. The presence of the HIS tag alone reduced the expression of both constructs by 30–40% (Fig. 1B, bars 2 and 6), and further addition of FLAG to γ1–218HIS gave 100% expression (construct FLAG/γ1–218HIS, bar 4). Strangely, when both epitopes were present on γ1–219, no expression was observed (Fig. 1B, lane 8). We purified two γ ECD variants, FLAG/γ1–218HIS and γ1–219HIS from 1-L cultures, obtaining ≈ 0.3 mg·L−1 and 0.2 mg·L−1 protein, respectively. We then constructed β1–221HIS and FLAG/β1–221HIS and expressed, purified and quantified them using 2-L cultures. The results showed that expression was increased threefold when the protein carried the FLAG tag (2 mg·L−1 protein instead of 0.7 mg·L−1).

Figure 1.

 Expression of the γ ECD variants. (A) Schematic representation of the various γ ECD constructs. The drawings depict the polypeptides with their tags/epitopes; the additional amino acid, proline, is shown as a black bar at the C-terminus of some γ ECD(s). (B) Relative yields of the different γ ECD constructs. All yields were expressed as a percentage of the yield of the nontagged γ1–218 construct, measured as the pixels for the positive dot-blots of the culture expressing γ1–218. The results shown are the mean from five experiments.

Expression and purification of the β, γ and ε ECDs

As (a) the presence of the HIS tail on the constructs greatly facilitates purification, (b) its negative effect on the yield of γ1–128 was considerably counteracted by the addition of the FLAG epitope, and (c) the presence of the proline residue did not improve expression, we proceeded to large-scale expression of the β, γ and ε ECDs using constructs carrying both the FLAG and 6-HIS tags and no additional proline (i.e. FLAG/β1–221HIS, FLAG/γ1–218HIS and FLAG/ε1–219HIS) (Fig. 2A). The yields ranged from 2 mg·L−1 culture for FLAG/β1–221HIS to 0.3–0.8 mg·L−1 for both FLAG/γ1–218HIS and FLAG/ε1–219HIS. The ECDs were purified using Ni2+/nitrilotriacetate affinity chromatography under native conditions. Typically, the proteins were eluted with 150 mm imidazole, although some protein was eluted at 100 mm (less than 10% of the total). Each protein migrated on SDS/PAGE with an apparent molecular mass of ≈ 35 kDa compared with the estimated molecular mass of ≈ 29 kDa, which was apparently due to the glycosylation of the product in the yeast cell (see below). The proteins were ≈ 90% pure, based on quantification of the protein bands on Coomassie Brilliant Blue-stained SDS/polyacrylamide gel (Fig. 2B).

Figure 2.

 Purification and deglycosylation of the AChR ECDs. (A) Schematic representation of the constructs used for expression of the β1–221, γ1–218 and ε1–219 ECDs of the human AChR in yeast P. pastoris. The arrowhead indicates the cleavage site of the α-factor peptide after secretion, and the circles indicate putative glycosylation sites (Asn-X-Ser motif). (B) SDS/PAGE of the proteins purified by Ni2+/nitrilotriacetate metal affinity chromatography stained with Coomassie Brilliant Blue; the left lane in each panel contains molecular mass markers, and the right lane the test protein. (C) Deglycosylation of the β, γ, and ε ECDs using N-glycosidase F. Purified proteins (1 µg) were incubated for 3 h at 37 °C in the absence (lane 1) or presence (lane 2) of N-glycosidase F, then the mixture was analyzed by SDS/PAGE (12% gel) and western blotting using anti-FLAG mAb M2. The arrows indicate the bands corresponding to the glycosylated (upper) and deglycosylated (lower) forms of each protein.

Deglycosylation of β1–221, γ1–218 and ε1–219

Each recombinant protein carries at least one Asn-X-Ser motif (glycosylation pattern for eukaryotes), β1–221 at Asn141, γ1–218 at Asn30 and Asn141, and amino acids 1–219 of the human AChR ε subunit (ε1–219) at Asn66 and Asn141. To verify that the recombinant proteins were glycosylated in the yeast cell (as suggested by the observed difference in the molecular mass of the purified proteins on SDS/PAGE), each protein was deglycosylated with peptide–N-glycosidase F. For each of the three proteins, this resulted in the appearance of a band migrating at the expected mass of ≈ 29 kDa (Fig. 2C), confirming that the proteins were glycosylated.

Gel-filtration analysis of polypeptides

To examine the solubility and oligomerization state of the recombinant polypeptides, we performed FPLC analysis in detergent-free solution (50 mm phosphate buffer, 300 mm NaCl, pH 8.0) in the presence of trace amounts of 125I-labeled soluble 66-kDa and 29-kDa protein markers. To verify that the observed peaks on the FPLC coincided with the presence of our proteins, dot-blots were performed using anti-β (mAb 73) or anti-γ (mAb 67) [22](Fig. 3). As the expected molecular mass of a monomer of each of the three ECD proteins was ≈ 30–32 kDa, the results showed that β1–221 was probably eluted as a dimer with an apparent molecular mass of 60–65 kDa (Fig. 3A), whereas γ1–218 was mainly present as an oligomer (possibly trimers-pentamers) (Fig. 3B). ε1–219 displayed a similar pattern to γ1–218 (data not shown), indicative of an oligomeric state. Although these elution patterns are typical of the proteins produced, occasional preparations showed a considerable percentage (10–20) of higher aggregates.

Figure 3.

 Gel filtration analysis of the polypeptides. (A) 2.0 mg β1–221 or (B) 2.0 mg γ1–218 protein was run on a Superose-12 column (Amersham-Pharmacia) at a flow rate of 0.5 mL·min−1, together with 125I-labeled protein markers of known molecular mass (66 and 29 kDa). The fractions were screened for ECD protein by dot-blots using anti-β (mAb 73) or anti-γ (mAb 67). The position of the 158-kDa (aldolase) marker is also shown (nonradioactive, obtained from a separate run).

CD spectra

When the β subunit ECD was subjected to far-UV CD analysis to examine its secondary structure, the CD spectrum in 50 mm phosphate buffer containing 0.15 m NaCl, pH 8.0, was characterized by a positive Cotton effect in the 190–200 nm region (peak ≈ 196 nm) and a negative effect in the 200–240 nm region (Fig. 4), suggesting a major contribution from a β-sheet structure. However, the quite high negative dichroism intensity over a relatively wide region ≈ 215 nm is indicative of the presence of bands at 208 and 222 nm, characteristic of a contribution of α-helical regions [23].

Figure 4.

 Far-UV CD spectrum of β1–221.

Binding of mAbs to the ECDs using ELISA

ELISAs were performed using the conformation-dependent mAbs 73 (binds to an epitope on the extracellular side of the β subunit) [22] and 67 (binds to an epitope on the extracellular side of the γ subunit) [22] and the nonconformation-dependent mAb M2 (anti-FLAG). As a negative control, mAb 25 [24] was used, which recognizes an epitope on Electrophorus electricus AChR, but not on mammalian AChR. Figure 5 shows that mAbs 73 and 67 specifically recognized their cognate proteins, whereas mAb 25 did not bind to any of the three polypeptides, as expected. The strong and specific binding of the mAbs to the appropriate ECD suggested the correct folding of at least β1–221 and γ1–218. Owing to the unavailability of a conformation-dependent ε subunit mAb, only binding of anti-FLAG mAb was tested.

Figure 5.

 mAb binding to β, γ, and ε ECDs using ELISA. ELISA plates were coated with one of the three ECDs or BSA as a control, and the binding of mAbs tested by ELISA as described in Experimental procedures (duplicate samples). mAb 73, checker-board bars; mAb 67, dark gray bars; FLAG mAb, light gray bars. mAb 25 (black bars) was used as the negative control.

Binding of the rabbit anti-β serum to recombinant β1–221 and human AChR

Purified β1–221 was used to raise a rabbit anti-β ECD serum. After three immunizations, the antiserum was tested for its ability to bind to the antigen (β1–221) using ELISA. The results (Table 1) showed strong and specific binding to β1–221 (≈ 1.8 absorbance units), with relatively weak cross-reactivity with either α1–210 or yeast proteins (≈ 0.4 absorbance units). The antiserum was then tested for its ability to bind to native human TE671 AChR [25] in RIA experiments. The high titer of the anti-(β ECD) serum for native AChR (870 nm, Fig. 6) further suggests that recombinant β1–221 retains some native-like conformational features.

Table 1.   Binding of the rabbit anti-(β ECD) serum to purified β1–221 in ELISA tests. Results shown are the mean from two experiments. Recombinant human α ECD (α1–210) was used to test for nonspecific binding of the rabbit anti-(β ECD) serum to a protein related to β1–221, rather than to a totally unrelated protein, such as BSA. The purified β1–221 used for immunization was purified from a P. pastoris yeast culture and possibly contained traces of yeast culture components (e.g. peptides originating from yeast protein degradation and other metabolic by-products). To eliminate the possibility that rabbit antibodies raised against such components could lead to spurious ELISA results, a control yeast supernatant sample was prepared as described in Experimental procedures. BSA was used as a negative control.
 Rabbit anti-(β ECD)
serum (A450)
Normal rabbit
serum (A450)
Purified β1–2211.800.10
Purified α1–2100.400.08
Yeast supernatant0.390.10
BSA0.050.04
Figure 6.

 Binding of the rabbit anti-(β ECD) serum to 125I-α-bungarotoxin-labeled native human AChR. Various amounts of the rabbit anti-(β ECD) serum were incubated with 14 fmol intact 125I-α-bungarotoxin-labeled human AChR, then bound receptor was precipitated with sheep anti-rabbit IgG, and radioactivity was measured. Samples were processed in duplicate, and the results shown are the mean of those of three experiments. The titer of β antibodies in the serum was calculated to be 870 nm.

Binding of human MG antibodies to recombinant ECDs

To further examine the structure of the ECDs produced and their potential as tools for MG studies, we tested their capacity to bind the highly conformation-dependent AChR antibodies present in MG sera. We had previously identified MG patient sera in which the antibodies are mainly directed against the α subunit (anti-α sera) and others with a very small proportion of antibodies against α (nonanti-α sera) [26]. We incubated five nonanti-α and one anti-α (82% antibodies against α) sera with a single ECD (β, γ or ε)–Sepharose or BSA–Sepharose resin, then measured the nonbound AChR antibodies in the initial and final samples. If the AChR antibodies in the MG serum recognized and bound to the recombinant, immobilized proteins, there would be a reduction in the amount of antibodies in the sample incubated with the ECD–Sepharose, and this reduction should be proportional to the percentage of subunit-specific antibodies in each serum. The anti-α serum should display little or no reduction when incubated with any of the test ECDs. Table 2 shows that incubation of each of the five nonanti-α sera (samples 1–5) with different ECD–Sepharose resins resulted in different percentage reductions in AChR antibody titers. In contrast, the anti-α serum (sample 6) did not show any significant reduction in titer when incubated with any of the non-α ECDS, as expected, but 82% loss of antibodies when incubated with α ECD–Sepharose. Similar results were obtained even when much higher serum quantities were incubated with the ECD–Sepharose resins (data not shown), which suggests that the immunoadsorbents in this experiment adsorbed all corresponding subunit antibodies. As the binding to the AChR of AChR antibodies in MG patient sera is highly conformation-dependent, our findings support the presence of native-like conformational features on all three recombinant ECDs.

Table 2.   Adsorption of AChR antibodies from human MG sera by immobilized ECDs. AChR antibody titer given in parentheses in nm. The reduction in total AChR antibodies present in MG sera observed after incubation of sera with β1–221, γ1–218, or ε1–219 immobilized on CNBr–Sepharose beads was measured by RIA using 125I-α-bungarotoxin-labeled native AChR.
SerumReduction (%) in AChR antibodies in MG serum
after incubation with immobilized ECDs
α1–210aβ1–221γ1–218ε1–219
  1. a  Data from [26].

MG 1 (50)3 ± 353 ± 631 ± 129 ± 8
MG 2 (163)8 ± 44 ± 28 ± 227 ± 4
MG 3 (99)1 ± 189 ± 115 ± 222 ± 5
MG 4 (11)3 ± 119 ± 429 ± 512 ± 1
MG 5 (6)6 ± 21 ± 139 ± 718 ± 1
MG 6 (5)82 ± 42 ± 113 ± 46 ± 2
anti-α serum

Discussion

In this paper, we describe the expression of soluble forms of the ECDs of non-α subunits of the human muscle AChR, using the yeast P. pastoris system. We have successfully used this system for the human muscle α ECD (α1–210) [17] and human neuronal type α7 subunit (α7 1–208) [18]. Based on this experience, we embarked on the expression of three of the four non-α ECDs, namely β, γ and ε. The expression of the δ subunit ECD, which is currently under progress, presents major difficulties, which require further investigation.

Aiming to improve expression yields, we designed, constructed and tested different variants of the γ ECD with and without a 6-HIS tail and/or the FLAG epitope and/or the first transmembrane amino acid of the AChR γ subunit (proline), which is found at this position in all human muscle subunits and has been shown to positively affect expression of the mouse AChR α subunit [21]. Our results showed that the addition of a Pro residue and the presence of common epitopes/tags used for purification influenced the expression yield. When the widely used 6-HIS tail was added to the C-terminus of γ1–218, it reduced expression almost twofold, whereas an N-terminal FLAG, a peptide sequence rich in charged residues, improved expression of γ1–218 by 20% and that of γ1–218HIS by 40% (Fig. 1B). The effect of the hydrophilic FLAG epitope was most dramatically seen with the β1–221 ECD, its addition leading to an almost threefold increase in expression. Although Yao et al. [21] showed that addition of a Pro to the mouse α ECD increased expression fourfold, no such effect was seen with the non-α human ECDs in our system. Aiming simultaneously at easy purification (achievable using the 6-HIS tail) and high yields, we used the N-FLAG/ECD/HIS-C constructs for the large-scale expression of the proteins and found that β1–221 was consistently expressed at a concentration of 2 mg·L−1 of culture and the γ and ε ECDs at a concentration of 0.3–0.8 mg·L−1. These yields are an improvement over the previous 0.1–0.2 mg·L−1 expression of the α1–210 protein [17], which, however, was in the monomeric form, in contrast with the three recombinant proteins described here. Gel-filtration analysis showed that β1–221 existed mainly as a dimer, whereas both γ1–218 and ε1–219 were mainly present as oligomers, possibly trimers–pentamers (Fig. 3). The state of the proteins was confirmed by dynamic light scattering experiments (data not shown); the proteins appeared polydisperse with an estimated diameter of 7.5–9.8 nm (β ECD) and 12.0–13.8 nm (γ ECD), suggesting, respectively, a dimeric or an oligomeric structure and confirming the FPLC data, considering that the ‘height’ of the ECD of the AChR is ≈ 6 nm [3]. This difference in solubility between the γ–ε and the β ECDs might be attributed to the primary structure of the protein: in addition to the ‘standard’ cysteine pair (residues 128 and 142) [27], present in all AChR subunit ECDs, both γ and ε carry extra cysteine residues at residues 61, 105, and 115 (γ) and 190 (ε), which could be involved in the formation of intramolecular or intermolecular bonds, leading to oligomer formation. However, if ‘free’ cysteines were the only factors responsible for multimer formation, then the β ECD should exist as a monomer; as this was not the case, exposed hydrophobic regions, which are presumably present in the β ECD, may also contribute to intermolecular association of monomers.

The results from a range of experiments suggested that the recombinant polypeptides are, at least to some extent, properly folded. Firstly, they were glycosylated, like native AChR [28] (Fig. 2C). Even though we lack direct evidence about the site and structure of the glycosylation sites on the ECDs, indirect evidence of correct glycosylation of our ECDs comes from our previous studies on the α ECD [17]: deglycosylation abolished α-bungarotoxin activity, strongly suggesting that glycosylation was at the right site and possibly of correct structure. Secondly, the CD spectrum of the β ECD indicated a folded protein consisting mainly of β-sheet (Fig. 4). The solved crystallographic structures of the molluscan Lymnaea stagnalis[11] and Bullinus truncatus[29] acetylcholine-binding proteins, which provide the prototypes for the AChR ligand-binding domain, show a predominance of β-sheet, and the CD spectra for these proteins largely resemble our spectra [29] and are also similar to those for mouse α1 expressed in mammalian cells [16] or yeast [21] and the Torpedoα ECD expressed in Escherichia coli[13]. These results suggest that the acetylcholine-binding proteins and the β ECD have similar structures and that the secondary structures of a non-α ECD (β) and the α ECD resemble one another, being largely composed of β-structure. Thirdly, conformation-dependent anti-β and anti-γ (mAbs 73 and 67, respectively) bound specifically to their cognate ECD (Fig. 5), and a polyclonal serum raised against the β1–221 polypeptide specifically bound to native AChR in RIA experiments (Fig. 6). Finally, AChR antibodies in different MG sera were specifically adsorbed by matrix-immobilized ECDs, with variable concentrations of AChR antibodies being retained on each ECD matrix (up to 89% of β antibodies for MG serum 3; Table 2). The presence of antibodies against several AChR subunits in a single serum (e.g. MG serum 1; Table 2) is interesting, although it was not unexpected because of previous indirect information (e.g. from competition experiments between mAbs against different subunits and MG sera) [22]. The actual autoantigen in anti-AChR-mediated MG is still uncertain. It may be intact AChR, AChR subunit(s) or fragments, or an AChR cross-reactive molecule. The polyspecificity of the sera may either mean that the autoantigen is an intact AChR or that epitope spreading occurred after initial induction by a single AChR subunit or a cross-reactive molecule. The adsorption results also indicated the presence of a considerable percentage (29–39%) of γ antibodies in three of the five tested nonanti-α MG sera (e.g. MG sera 1, 4 and 5; Table 2). The γ subunit, present in the fetal isoform of the AChR, is replaced by the ε subunit in adult muscle; however, this fetal isoform is expressed in myoid cells in the thymus [30,31], and γ expression persists into adulthood in mouse and bovine ocular fibers [32,33], justifying the presence of γ antibodies in adult MG sera. The majority of AChR antibodies in MG sera are directed against various nonlinear, conformation-dependent epitopes on the extracellular part of the AChR molecule, a fact that has prevented their characterization using synthetic peptides or denatured recombinant polypeptides obtained using prokaryotic expression systems [20,34]. In addition, the immune responses against the non-α AChR subunits have not been examined as carefully as those against the α subunit, even though the differential expression of the different subunits may be highly significant in the pathogenesis of MG [35].

All four ECDs of the Torpedo AChRs have previously been expressed as soluble proteins using baculovirus-infected insect cells [36], and the proteins showed proper folding, but the amounts produced were insufficient for crystallization trials. Moreover, the α ECDs of Torpedo and human AChR, which have been produced as inclusion bodies in bacteria in quantities sufficient for structural studies [13,14], require denaturing conditions for solubilization of the protein and refolding and also do not undergo post-translational modifications. In the present study, we obtained stable Pichia clones expressing satisfactory amounts of three non-α ECDs (β1–221, γ1–218 and ε1–219) which were in a soluble, secreted form and probably correctly folded, a fact that may permit preliminary crystallization trials. Crystallization trials require protein samples of concentration ≈ 10 mg·mL−1, purity of at least 95%, and monodispersity. Based on the yields of our yeast cultures (0.5–2 mg·L−1), a medium-scale expression would suffice to provide material that, after purification and gel filtration, should be sufficiently concentrated. The risk in this case would be the putative formation of aggregates that would render the sample unusable for downstream processing, especially for the recombinant γ1–218 and ε1–219, which were already in the form of oligomers; this approach, however, could possibly be applicable to β1–221, which is dimeric, stable on concentration (data not shown), and exhibits the highest expression yield. For the γ and ε ECDS, improvement in their solubility is required before attempts at structural trials. We are working towards this by constructing mutant forms of the proteins. Nevertheless, these polypeptides, together with the already produced α1–210 [17], were all specifically recognized by human AChR antibodies in MG sera, allowing their immediate use for the detailed study of the specificities of the antibodies in MG sera and the development of antigen-specific therapeutic approaches.

Experimental procedures

Bacterial and yeast strains, growth conditions, plasmids and DNA manipulations

The E. coli K-12 strain TOP10F′ (Invitrogen, San Diego, CA, USA) was used for replication of plasmid DNA. Cloning of the ORFs encoding the β1–221, γ1–218 and ε1–219 ECDs was performed by standard techniques [37]. Luria–Bertani broth and agar were used for amplification of transformed bacteria. Ampicillin (100 µg·mL−1) was used in liquid or solid media.

The vector pPIC9 (Invitrogen) was used to clone the ORFs in-frame with a leader sequence allowing secretion of the produced protein after cleavage of the secretion signal. An oligonucleotide, 5′-GTAGATTACAAGGATGACGATGACAAAG-3′ encoding the FLAG sequence, DYKDDDDK, was introduced into the vector between the unique SnaBI and EcoRI sites. This allowed the subsequent in-frame cloning of our PCR products with a 5′-EcoRI site in such a way that the cloned ORF was expressed as a polypeptide carrying the FLAG peptide at its N-terminus. The resulting plasmid was named pPIC9/FLAG.

Cloning using PCR

We used PCR to amplify the extracellular region of each of the β, γ and ε subunits, using the plasmid templates, pcDNA3.1/Beta, pcDNA3.1/Gamma and pcDNA3.1/Epsilon (cDNA clones of the human β, γ and ε AChR subunits in pcDNA3.1 respectively; all kindly provided by D. Beeson, University of Oxford, UK) [38]. PCR was performed on the appropriate template for each subunit on a Perkin-Elmer (Boston, MA, USA) thermal cycler; 5 min denaturation at 94 °C was followed by 25 cycles of 94 °C for 20 s, 58 °C for 30 s, and 72 °C for 90 s, and a final 5-min extension step at 72 °C. The reaction mix consisted of 10 ng template, 50 mm each dNTP, 20 pmol each primer, and 1 U Taq DNA polymerase in a volume of 50 µL 10-fold diluted reaction buffer (Promega, Madison, WI, USA). For β1–221, the forward primer was 5′-GCGGAATTCTCGGAGGCGGAGGGTCGAC-3′ and the reverse primer 5′-ATAGTTTAGCGGCCGCTCAATGGTGATGGTGATGGTGCTTGCGGCGGATGATGAG-3′. For the γ1–218 variants (some with an additional C-terminal Pro giving γ1–219), the forward primer 5′-GGTGTAGAATTCCGGAACCAGGAGGAGCGC-3′ was used in all cases, together with the reverse primer (a) 5′-ATAGTTTAGCGGCCGCTTACTTGCGCTGGATGATGAGCAGG-3′ for γ1–218, (b) 5′-ATAGTTTAGCGGCCGCTTAGTGATGGTGATGGTGATGCTTGCGCTGGATGAGCAGG-3′ for γ1–218HIS (γ1–218 with a 6-HIS tag at its 3′ end to facilitate purification), (c) 5′-ATAGTTTAGCGGCCGCTTAGGGCTTGCGCTGGATGAGCAGG-3′ for γ1–219, or (d) 5′-ATAGTTTAGCGGCCGCTTAGTGATGGTGATGGTGATGGGGCTTGCGCTGGATGAGCAGG-3′ for γ1–219HIS. For the ε1–219 variants (ε1–220 with additional Pro), the forward primer 5′-GGTGTAGAATTCAAGAACGAGGAACTGCG-3′ was combined with (a) 5′-ATAGTTTAGCGGCCGCTTACTTCCGGCGGATGATGAGCGAG-3′ for ε1–219, (b) 5′-ATAGTTTAGCGGCCGCTTAGTGATGGTGATGGTGATGCTTCCGGCGGATGATGAGCGAG-3′ for ε1–219HIS, (c) 5′-ATAGTTTAGCGGCCGCTTACGGCTTCCGGCGGATGATGAGCGAG-3′ for ε1–220, or (d) 5′-ATAGTTTAGCGGCCGCTTAGTGATGGTGATGGTGATGCGGCTTCCGGCG-GATGATGAGCGAG-3′ for ε1–220HIS (underlined EcoRI and NotI). The PCR products were purified (Qiagen PCR clean-up kit; Qiagen, Hilden, Germany), EcoRI–NotI digested, repurified, and cloned into the EcoRI–NotI-digested pPIC9 or pPIC9/FLAG plasmid. Each PCR product was cloned into both plasmids. Sequencing was used to verify the identity of the inserts.

Yeast transformation and dot-blot screening of positive clones

Plasmids (10 µg) encoding the β1–221 ECD (with or without the FLAG epitope) and the various γ1–218 and ε1–219 ECDs were linearized using SacI (for β1–221) or SalI or SacI (for γ1–218 and ε1–219) and electroporated into freshly made competent GS115 P. pastoris cells. Selection of positive transformants (cells able to grow in the absence of histidine) was achieved by plating on regeneration dextrose plates (1 m sorbitol, 2% dextrose, 1.34% yeast nitrogen base, 4 × 10−5% biotin, 0.005%l-glutamic acid, l-lysine, l-methionine, l-leucine, and l-isoleucine, 2% agar) without histidine. Small-scale cultures of single colonies were tested after growth overnight in 3 mL BMGY medium (1% yeast extract, 2% peptone, 100 mm potassium phosphate, pH 6.0, 1.34% yeast nitrogen base, 4 × 10−5% biotin, 1% glycerol) and resuspension of the cells in 3 mL BMMY medium to induce expression (BMMY medium is identical with BMGY, but contains 0.5% methanol instead of glycerol) (day 0). Methanol was added to 0.5% every 24 h to maintain induction, and 0.75 mL liquid medium was removed every 24 h after day 0 to test for the expression and secretion of the produced protein. The cleared supernatant was tested on dot-blots using mAb 73, mAb 67, or anti-FLAG mAb M2 (Sigma, St Louis, MO, USA) to test for the expression of β, γ or all ECDs, respectively. After the initial screening, phosphate buffers with a pH of 6.5 or 7.0 were also tested, and the pH 7.0 buffer was finally adopted for large-scale expression. Expression levels of the different γ or ε variants were estimated by quantification of the positive signal on dot-blots of culture supernatant (at serial dilutions) using imagej software (http.//rsb.info.nih.gov/ij/).

Large-scale expression and purification of proteins

The best expressing clone was selected for each protein. A 0.1-mL sample of a small overnight culture of 20 mL BMGY medium was used to inoculate 1 L fresh BMGY medium. After growing to an A600 of 3 (≈ 18–20 h), the cells were spun down, washed, and resuspended in 3 L BMMY medium to induce expression. On day two, the cultures were cleared of cells by centrifugation for 20 min at 2500 g (Jouan 11175372 M4 rotor), and the supernatant concentrated using a Millipore (Bedford, MA, USA) ultrafiltration system (filter cut-off 10 kDa); these steps and all subsequent steps were performed at 4 °C. The concentrate was dialyzed overnight against 50 mm phosphate buffer, 2 m NaCl, pH 8.0, for the γ and ε ECDs or 50 mm phosphate buffer, 0.5 m NaCl, pH 8.0, for the β ECD before binding of the protein to 1.5 mL pre-equilibrated Ni2+/nitrilotriacetate/agarose (Qiagen). The protein was purified under native conditions following the manufacturer's instructions. Eluates were analyzed by SDS/PAGE (12% gel) and Coomassie blue staining or western blotting using mAb 73 (for the β ECD) or anti-(FLAG M2) (Sigma). The purity of the protein was estimated from Coomassie Brilliant Blue-stained gels and quantification of the bands using imagej software, and protein concentration was determined using the Bradford method (Bio-Rad, Hercules, CA, USA).

In vitro deglycosylation

A sample (1 µg) of purified protein was deglycosylated by incubation for 3 h at 37 °C with 1000 U N-glycosidase F (New England Biolabs, Frankfurt, Germany) in a final volume of 50 µL under the conditions recommended by the manufacturer for a nondenatured protein. The protein was then precipitated by the addition of 200 µL methanol/acetone (1 : 1, v/v), incubation at −20 °C for 20 min, centrifugation for 15 min, and resuspension in 15 µL distilled water. The samples were analyzed by SDS/PAGE and western blotting using FLAG mAb M2.

FPLC analysis of polypeptides

To determine the size of β1–221, γ1–218 or ε1–219, FPLC analysis on a Superose-12 column (Amersham-Pharmacia, Munich, Germany) was performed in 50 mm sodium phosphate buffer/300 mm NaCl, pH 8.0, at a flow rate of 0.5 mL·min−1. Samples of each fraction (normally 1 and 10 µL of each 0.5-mL fraction) were tested for the presence of the specific protein by dot-blotting with FLAG mAb M2.

Radioactive labeling of protein markers and α-bungarotoxin

α-Bungarotoxin (24 µg) or 2 µg either bovine erythrocyte carbonic anhydrase (≈ 29 000 Da) or BSA (≈ 66 200 Da) (both from Fluka; Sigma-Aldrich, Athens, Greece) was labeled, respectively, with 2 mCi or 0.1 mCi 125I, using the chloramine T method [39], loaded on to a G50-Fine column (Amersham-Pharmacia), and the labeled protein collected and stored at −20 °C. Approximately 100 000 c.p.m. of each of the 125I-labeled protein markers was loaded on every FPLC run as size markers.

Preparation of rabbit anti-(β subunit) serum

An 8-week-old female New Zealand White rabbit was injected subcutaneously with ≈ 0.5 mg purified β1–221 protein in 50% (v/v) complete Freund's adjuvant, followed by three injections at monthly intervals in 50% incomplete Freund's adjuvant. One week after the last injection, antiserum was collected, aliquoted, and stored at −20 °C in the presence of 0.05% sodium azide. Use of experimental animals abides by law 2015/27-2-1992 of the Greek Republic and Presidential Decree 160/3-5-1991 in accordance with directive 86/609EOK of the Council of Europe for protection of vertebrates/animals used for experimental or other research purposes.

CD spectra

CD spectra were measured at 20 °C using a Jasco model J-715 spectropolarimeter (located at NCSR, Demokritos, Athens, Greece) in semi-automatic slit adjustment mode. The scan speed was set at 50 nm·min−1, the response time at 2 s, and the scan range at 180–260 nm. Optical activity was expressed as the mean residue ellipticity (Θ), in degrees·cm2·dmol−1, based on a mean residue weight of 115 for the β ECD polypeptide. The derived spectrum represents the mean of eight scans and was corrected for light scattering by buffer subtraction. The protein concentration was optimized as 0.2 mg·mL−1, and the quartz cell path length was 1 mm. All samples were optically homogeneous.

ELISA

ELISA plates (Maxi-Sorb; Nun Roskilde, Denmark) were coated, as described previously [14], using 0.25 µg purified recombinant protein (β1–221, γ1–218 or ε1–219) per well. Control wells were coated with BSA (0.25 µg). Additional control wells were coated with 0.25 µg α ECD (α1–210) or 100 µL yeast culture supernatant prepared as follows: 100 mL of a culture of P. pastoris GS115 strain was spun, and the supernatant filtered, concentrated 40-fold, and dialyzed against 50 mm phosphate buffer, pH 8.0.

The plates were washed with phosphate-buffered saline, pH 7.5 (NaCl/Pi) and blocked for 30 min at 37 °C with blocking solution (5% nonfat milk in NaCl/Pi), then incubated for 1 h at 25 °C with primary antibody in blocking solution; mAbs were used at a 1 : 100 dilution (the concentration of the undiluted mAb ‘stock solution’ was 0.1–0.5 mg·mL−1), and the rabbit antiserum was used at dilutions of 1 : 100–1 : 10 000. After three washes with blocking solution, the plates were incubated for 1 h at 25 °C with secondary antibody [horseradish peroxidase-conjugated rabbit anti-rat IgG (Dako, Glostrup, Denmark) in the case of the mAbs and sheep anti-rabbit IgG (Dako)] at a 1 : 500 dilution in blocking solution. No secondary antibody was used when the FLAG mAb M2 was used, as the antibody was supplied in its horseradish peroxidase-conjugated form (Sigma). The ELISA plate was developed using 3,3′,5,5′-tetramethylbenzidine ready-to-use substrate (MBI-Fermentas, St Leon-Rot, Germany), stopping the reaction with 0.2 m H2SO4. The plate was read at 450 nm on a microtiter plate reader.

Preparation of ECD–Sepharose beads

ECD (0.25 mg) mixed with BSA (1.25 mg, as carrier) were bound to 0.25 g CNBr-activated Sepharose beads (Pharmacia, Munich, Germany) according to the manufacturer's protocol as described previously [26]. The beads were then diluted in NaCl/Pi/2% BSA/0.05% NaN3 so that 120 µL of the mixture contained 1 µg recombinant protein. Control beads were prepared using 1.5 mg BSA.

Use of the ECD–Sepharose matrix for binding AChR antibodies in MG sera

Depending on the AChR antibody titer, different dilutions of sera were prepared: the MG sera were diluted 1 : 10 (for serum titer 5 nm) to 1 : 500 (for titer 290 nm) supplemented with normal human serum to a final serum dilution of 1 : 10. This guaranteed that the amount present in the untreated sample would immunoprecipitate ≈ 50% of the labeled AChR. A 40-µL portion of the dilution was incubated for 2 h at 4 °C with 120 µL Sepharose–ECD or Sepharose–BSA matrix (final volume 160 µL), and then duplicate 40-µL samples of supernatant (containing 1 µL serum) were tested in the RIA described below.

RIA for MG sera or rabbit anti-(β1–221) serum

We tested the ability of MG sera to precipitate α-bungarotoxin-labeled human AChR prepared from either TE671 cells or a mixture of CN21/TE671 cells (CN21 cells express the ε and γ AChR subunits at a ratio of approximately 2 : 1 [40], whereas TE671 cells express only the γ subunit AChR [25]). The TE671-derived AChR was used when sera had been preincubated with γ1–218–Sepharose beads, and the mixed AChR was used when the sera were preincubated with the other ECD–Sepharose beads. Human AChR (14 fmol) was labeled for 4 h at 4 °C with 50 000 c.p.m.

125I-α-bungarotoxin (≈ 70 fmol) in a volume of 20 µL, then 40 µL of the ‘treated’ MG serum (preincubated with either ECD–Sepharose or BSA–Sepharose) was added. The mixtures were incubated for 16–18 h at 4 °C, then immune complexes were precipitated by incubation for 1.5 h at 4 °C with 10 µL goat anti-(human γ-globulin) serum (Lampire Biological Laboratories, Pipersville, PA, USA), followed by centrifugation. The samples were washed twice by the addition of 1 mL NaCl/Pi (pH 7.4)/0.5% Triton X-100 and centrifugation at 3500 g for 10 min at 4 °C. The precipitated radioactivity was counted on a γ-counter. If the antibodies bound specifically to the test ECD protein, a reduction in the amount of antibodies would be detected in the MG serum sample incubated with ECD–Sepharose (sample 1) compared with the same sample incubated with BSA–Sepharose (sample 2). This reduction should be proportional to the fraction of autoantibodies reactive with the subunit. The percentage immunoadsorption was estimated as 100 × {[Δc.p.m. (Sample 2)]–[Δc.p.m. (Sample 1)]}/[Δc.p.m. (Sample 2)], where Δc.p.m. is the difference between the individual sample c.p.m. value (from precipitated 125I-labeled AChR) and that from the corresponding negative serum control sample.

When the rabbit anti-(β1–221) serum was used supplemented with normal rabbit serum to a total volume of 0.1 µL, the immune complexes were precipitated using 0.3–1 µL goat anti-rabbit IgG (Sigma) and incubation for 1.5 h at 4 °C, all other steps being identical with those described above.

Acknowledgements

This work was supported by grants from the Quality of Life program of the EU (contract QLG3-CT-2001-00225), the Association Française contre les Myopathies (AFM), and the Muscular Dystrophy Association of USA (MDA). We are grateful to Anna Kokla for excellent technical assistance, and Dr D. Beeson (University of Oxford) for the cDNA clones.

Ancillary