scFv Single Chain Antibody Variable Fragment as Inverse Agonist of the (cid:1) 2 -Adrenergic Receptor*

Antibodies directed against the second extracellular loop of G protein-coupled receptors were shown to possess functional activities. Using a functional monoclonal antibody against the human (cid:1) 2 -adrenergic receptor, a scFv fragment with high affinity for the target epitope was constructed and produced. The fragment recog-nized the (cid:1) 2 -adrenergic receptors on A431 cells, blocked cAMP accumulation induced by the (cid:1) 2 -agonist salbutamol, and decreased basal cAMP accumulation in the same cells. Their in vitro activity was tested on neonatal rat cardiomyocytes. The antibody fragments blocked the chronotropic activity induced by the (cid:1) 2 -agonist clen- buterol. They also decreased the in vivo heart beating frequency of mice pretreated with bisoprolol (a (cid:1) 1 -adre- nergic receptor antagonist) for 4 min after injection. The immunological approach presented here may serve as a strategy for the synthesis of a new class of allosteric modulators for G protein-coupled receptors.

The G protein-coupled receptor family is one of the main targets of currently used drugs (1). Most of the structural insights into this family of receptors were obtained from studies of the ␤ 2 -adrenergic receptor, the first of this family of neurotransmitter receptors to be cloned and sequenced. This receptor is an integral membrane protein consisting of seven membrane spanning ␣-helices, which form a pharmacophore pocket, linked together by extra-and intracellular loops (2). One of the pharmacological challenges posed by this family of receptors is the presence of multiple subtypes, all recognizing the same endogenous ligands. This suggests a high conservation of the pharmacophore for a particular family of receptors, thus explaining the difficulty to synthesize drugs (agonists or antagonists), specific for one of these subtypes. Autoantibodies directed against cardiovascular G protein coupled receptors, functionally interfering with the target, have been described in several cardiovascular diseases. Most of these autoantibodies are directed against the second extracellular loop and are exquisitely specific for one of the receptor subtypes in view of the highly variable structure of this domain (3). Lebesgue et al. (4) reported the selection of monoclonal antibodies directed against a synthetic peptide whose sequence was derived from the second extracellular loop of G protein-coupled receptors. The selected monoclonal antibody against the ␤ 2 -adrenergic receptor had partial agonist activity as a dimer and antagonist activity as a monovalent Fab fragment (4,5). This antibody was used to construct a scFv fragment (single chain variable fragment), which was cloned, sequenced, and expressed in Escherichia coli. In this study, we describe the sequence, the immunochemical, pharmacological, and physiological properties of this scFv fragment. These results open the way for the development of new strategies to synthesize molecules, which are highly specific for one of the subtypes of a particular G proteincoupled receptor family and can allosterically modulate their activity.

Construction of the Single Chain Antibody Gene
The single chain antibody gene fragment encoding the heavy and light variable chain of the monoclonal antibody 6H8 (4) was prepared as described in Ref. 7. scFv 6H8 was created by joining the 6H8 V H and V L genes together by PCR splicing with overlap extensions using oligonucleotides that encoded a 15-amino acid linker (G 4 -S) 3 between the C-terminal of the V H and the N-terminal of the V L gene. The ends of the variable gene were modified by PCR using as primers, VHRev (5Ј-GGT GCA GCT GCA GCA GTC AGG GTC TGA GC-3Ј), which encodes the N-terminal wild type sequence of the V H containing a PstI site, VHFor (ACC ACC GGA TCC GCC TCC GCC TGA GGA GAC TGT GAG CGT-3Ј), which encodes the C terminus of the VH and a part of the linker. VLRev (5Ј-GGA GGC GGA TCC GGT GGT GGC GGA TCT GGA GGT GGC-3Ј) and VLFor, containing a XhoI site which encodes 6 His residues (5Ј-GCA ATT CCT CGA GTT AGT GAT GGT GAT GGT GAT GTT TGA-3Ј), were used to amplify and modify the V L domain. The scFv gene was inserted in frame with sequence PelB of the expression vector pSW1 (8) between the PstI and XhoI sites. The constructed vector pSW1-6H8 His 6 was cloned in HB 2151 E. coli strain.

Bacterial Expression of scFv-6H8
The bacterial expression of the recombinant scFv protein and extraction of soluble periplasmic protein are described in Ref. 7. The periplas-* This work was supported in part by grants from the Centre National de la Recherche Scientifique. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AJ574851.

Purification of the scFv Construction
The periplasmic extract was incubated for 1 h at 4°C with 500 l of nickel-nitrilotriacetic acid-agarose beads (Qiagen) and washed with PBSI buffer, and the recombinant protein was eluted with 1 ml of PBS 1 supplemented with 500 mM imidazole and immediately dialyzed against PBS.

Gel Electrophoresis and Western Blot Analysis
SDS-PAGE analysis was performed as a standard procedure using 12.5% acrylamide gels followed by staining with Coomassie Brilliant Blue (Serva) or immunoblotting. For Western blot analysis, the proteins were transferred from the gels onto a PROTRAN nitrocellulose transfer membrane (Schleicher & Schuell) using a mini trans-blot system (Bio-Rad) in transfer buffer (25 mM Tris-HCl, 190 mM glycin, 20% methanol, pH 8.3). The membranes were soaked 1 h in PBS-T (20 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 150 mM NaCl, 2.7 mM KCl, pH 7.4) supplemented with 5% nonfat milk powder and 0.1% Tween 20. This was followed by a 1-h incubation with anti-His tag antibody conjugated to horseradish peroxidase 1/2000 (Sigma). The antibody was diluted in the blocking solution PBS-T milk. The proteins on the membranes were revealed by the classical procedure of the ECL reagents (Amersham Biosciences, Saclay, France).

Immunochemical Characterization of the scFv Fragment
The instrument BIACORE 3000 and all the reagents for analysis were obtained from BIACORE (Uppsala, Sweden). The low carboxylated dextran matrix (B1) was activated with 50 l at 5 l/min of a mixture 0.2 M N-ethyl-NЈ-dimethylaminopropyl carbodiimide and 0.05 M N-hydroxysuccinimide. Streptavidin was immobilized with the standard BIACORE protocol at a density of 0.05 pmol/mm 2 . The biotinylated peptide (0.1 mg/ml in PBS, pH 6.0) was then immobilized on the streptavidin at a flow rate of 5 l/min for 7 min. Kinetic studies were performed as described in Ref. 5.

Protein Concentration Determination
The total protein concentration of the purified scFv (active ϩ nonactive proteins) was determined using the BCA kit (Pierce) and by measuring absorbance at 280 nm. The extinction coefficient was determined using the Expasy protparam tool available on the web (www.expasy. org/tools/protparam.html). Immunocytochemistry A431 cells were fixed 5 min with 2% paraformaldehyde and permeabilized with PBS Triton X-100 0.1% for 1 min. Slides were saturated with PBS supplemented with nonfat dry milk 5%. ScFv 6H8 and scFv 9C2 (9) (control) were incubated overnight at 4°C. After three washes with PBS, anti-His tag antibody was added and incubated for 1 h at room temperature. Rabbit anti mouse IgG (H ϩ L) antibody Alexa conjugated (1/200, Molecular Probes, Junction City, OR) was allowed to react with the fixed primary antibody for 1 h at room temperature. 4Ј,6-Diamidino-2-phenylindole (1 g/ml, Sigma) was used for nuclear staining.

Pharmacological Characterization of the scFv Fragment
cAMP Response by A431 Cells Treated in Vitro with scFv-The biochemical effects of scFv 6H8 on the ␤ 2 -adrenergic receptor were assessed by measuring the intracellular cAMP concentration of A431 cells (10). Cells were seeded in 6-well culture plates 24 h before stimulation and then washed and incubated with 1 ml of Hanks' balanced medium buffered with 10 mM HEPES, containing 100 M isobutylmethylxanthine to block cAMP hydrolysis. After 15 min, scFv was added at different concentrations, and 30 min after the first treatment salbutamol was added to a final concentration of 10 nM. Stimulation of the cells was performed during 15 min, then the supernatant was aspirated and the reaction was stopped by adding 1 ml of boiling water. The cAMP content was determined using a competitive immunoenzymatic assay (BIOTRAK cAMP, Amersham Biosciences). The protein concentrations of the samples were determined using BCA kit (Pierce). The concentra-tion of cAMP was reported on the protein concentration, related to the number of cells/well; the results were expressed as pmol of cAMP/mg of protein. The results were normalized using cAMP content of untreated cells as 100%. Results are from duplicates of three independent experiments.
Beating Frequency of Neonatal Rat Cardiomyocytes in Culture-Rat neonatal cardiomyocytes were prepared from ventricles of 1-2-day-old Wistar rats using a modified method (11). The cells were cultured as monolayers for 4 days at 37°C in SM 20-1 medium supplemented with 10% heat-inactivated calf serum and 2 mM fluorodeoxyuridine and exhibited a spontaneous basal pulsation rate of about 160 beats/min. The cardiomyocyte cultures were washed with fresh medium containing serum and incubated for 30 min at 37°C with the same medium under stationary conditions for 2 h. The flasks were transferred to the heatable stage of an inverted microscope, and the increase of the beating rate was determined for at least 10 observations for each experimental point in three independent experiments.
Physiological Characterization of the scFv Fragment on Mice in Vivo-Ten female BALB/c mice, 20 g each, were habituated to tail intravenous injections at the same time as being kept in individual mouse electrocardiogram (ECG) boxes as described in Ref. 12, for 4 continuous weeks.
Six-lead ECGs (DI, DII, DIII, AVL, AVR, AVF) were recorded from 10 conscious mice for 60 min. Mice were not anesthesized to prevent ECG modifications (13). The mice were pretreated ip with an injection of 200 g of the ␤ 1 -adrenergic antagonist bisoprolol. After 10 min 100 l of NaCl 0.8% or scFv fragments were injected intravenously at an active concentration of 110 nM. Heart rates were measured by counting QRS peaks in 10 s windows. The cardiac rhythm of scFv-treated mice and NaCl (negative control mice) are compared in a paired match Student's t test.

Statistics
Statistical analysis was performed using Minitab software. To compare percent of variation of the cardiac rhythm with the rhythm at t ϭ 0, we used a Student's t test with fixed mean of 1. A Student's paired match t test was used to compare NaCl injection with scFv 6H8 or control scFv.

Molecular Modeling
ScFv 6H8 model was built in silico using the Biopolymer, Homology and Discover software of Accelrys (San Diego, CA). The VH and VL chains of the scFv 6H8 were built from the 1cic (Protein Data Bank) and 1jrh (Protein Data Bank), respectively. The sequences of the templates (Fig. 6) were changed to those of the scFv 6H8, and the modified chain was submitted to a steepest descent energy minimization of 2000 steps until a RMS derivative of 0.001 kcal/mol Å. The assembly was again submitted to a conjugate gradient minimization procedure of 2000 steps with the backbone fixed.

Cloning, Sequencing, Expression, and Immunochemical
Characterization of the scFv 6H8 Fragment-The N-terminal part of the VH and VL of the 6H8 antibody was sequenced by Edman degradation to design the primers used for the amplification of the two domains. Unfortunately, the VH domain was blocked by a pyroglutamin residue. The sequence of the VL, DIQMTQ, helped us to design the VLRev primer. The scFvencoding gene derived from the variable regions (V H and V L linked together via a short linker (G 4 -S) 3 ) of the IgG1 6H8 monoclonal antibody (4), with addition of a C-terminal His 6 tag encoding sequence, was inserted in frame with the PelB sequence into the pSW1 expression vector. The sequence of the single chain construction is represented in Fig. 1. We confirmed that the cloned VL gene did not correspond to the aberrant kappa transcript of the sp20 hybridomas (14). The plasmid pSW1-scFv 6H8 was cloned into the HB2151 Escherichia coli strain, and the recombinant protein was expressed and exported to the bacterial periplasm by its leader sequence PelB (15). The scFv 6H8 was easily purified and concentrated by immobilized metal affinity chromatography (Fig. 2).
BIACORE technology allows the analysis of scFv/antigenic peptide interaction in real time. It also allows the determination of the active concentration of analytes in solution (16). The active concentration showed that only 5% of the purified recombinant protein was able to interact with the antigenic peptide H19C; this is mainly due to the bacterial expression sys-tem that has limited capacity to correctly fold the polypeptidic chain and to the natural instability of this chimeric construct (see Ref. 17 for review). The calculated concentration allowed scFv Single Chain Antibody Variable Fragment 36742 us to determine the kinetic parameters of the scFv 6H8 using a BIACORE system and the BIAeval 3.1 software. A Langmuir binding model gave us the following kinetic parameters: the k on ϭ 1.0 ϫ 10 5 M Ϫ1 s Ϫ1 , the k off ϭ 4.8 ϫ 10 Ϫ3 s Ϫ1 , and the equilibrium constant K A ϭ 2.1 ϫ 10 8 M Ϫ1 . These parameters are similar to those calculated under the same conditions for the 6H8 proteolytic Fab fragment (k on ϭ 0.89 ϫ 10 6 M Ϫ1 s Ϫ1 , k off ϭ 6.93 ϫ 10 Ϫ3 s Ϫ1 ) (5).
To assess the ability of the scFv to interact with the receptor, immunocytochemical experiments were performed on A431 cells that express at their surface the ␤ 2 -adrenergic receptor. Fig. 2b shows the presence of a specific vesicular and membrane labeling when using the scFv 6H8 fragment.
Pharmacological Characterization of the scFv Fragment-We next determined whether the scFv construction had pharmacological properties on the ␤ 2 -adrenergic receptors. A431 cells stimulated with salbutamol (10 nM) were treated with different scFv concentrations. Fig. 3 shows that scFv 6H8 was able to inhibit the activation of the ␤ 2 -adrenergic receptor in a dose-dependent manner. Interestingly, it was able to significantly decrease the basal accumulation of cAMP. This is the main characteristic of inverse agonists (see Ref. 18 for review). These results were confirmed in vitro by measuring the beating rate of neonatal rat cardiomyocytes stimulated by clenbuterol (a specific ␤ 2 -agonist) (Fig. 4). The basal beating rate of the cardiomyocytes as well as the clenbuterol dose-response curve were decreased. In view of the fact that the clenbuterol doseresponse curve was not shifted to the right but only showed a decrease in the maximal response, we conclude that the scFv acts in a noncompetitive manner. The left panel shows a nuclear 4Ј,6-diamidino-2-phenylindole labeling and the right panel an immune labeling of the same cells. A specific membranar and vesicular fluorescence at the membrane of the cells incubated with the scFv 6H8 can be seen (I). A nonspecific nuclear labeling can be seen with the scFv 6H8 (I), scFv control (II), or anti-histidine tag alone (III). The nuclear background noise is mainly due to a nonspecific binding of the anti histidine tag antibody.

scFv Single Chain Antibody Variable Fragment
Physiological Characterization of the scFv Fragment-To investigate the possible in vivo effects of scFv on the heart, we injected the scFv construction in mice. Before scFv injection, mice were pretreated for 10 min with bisoprolol hemifumarate (a specific ␤ 1 -antagonist) to limit the regulatory activity of the ␤ 1 -adrenergic receptor on the heart. Intravenous injection of scFv Single Chain Antibody Variable Fragment 36744 100 l of 110 nM active scFv gave a statistically significant decrease of the beating rate (about 14%), 2 min after the injection (Fig. 5). NaCl or control scFv did not significantly alter the beating rate. The scFv control was injected at the same protein concentration as the 6H8 scFv. The intravenous injection of the inverse agonist ICI 118,551 at 10 mg/kg decreased the beating rate to a larger extent (26%). DISCUSSION G protein-coupled receptors are one of the main targets of cardiovascular and neurological drugs, since most of the neurotransmitter receptors belong to this family. The pharmacological existence of subfamilies of these receptors was first assessed by showing that the same neurotransmitters had different effects on different tissues. The molecular basis of this diversity was confirmed by cloning, sequencing, and expressing molecules coded by different genes but recognizing the same neurotransmitters. The putative structure of these proteins suggests that the neurotransmitters are localized in an intramembrane pocket, which only allows a limited amount of variability to embed different agonist and antagonists. The difficulty to synthesize molecules with an exquisite specificity for one of the receptor subtypes could be explained by the constancy of the pharmacophore pocket belonging to members of the same family. In contrast, the extracellular domains of receptors of the same family can vary more widely, which suggests the possibility to raise antibodies possessing the exquisite specificity sought for by the pharmacologists.
Drug discovery based on immunological reagents was already forwarded more than 20 years ago, but only recently, due to the expansion of biotechnological tools, could it be considered as a realistic goal. Indeed, recent results have shown that such approach could lead to functional peptides derived from antibodies directed against cell receptors (19,20). The finding that autoantibodies against G protein-coupled receptors were present in and responsible for different cardiovascular diseases prompted us to use a similar approach for the development of antibody fragments of low molecular weight interfering with this family of receptors.
Cloning and sequencing of the variable regions of the 6H8 monoclonal antibody in comparison with the corresponding germ line showed 19.4 and 52.6% variability respectively for the heavy chain and for the light chain. The sequence of the FIG. 5. Physiological effects of scFv 6H8 on mouse heart, in vivo cardiac physiology measurements. a, scheme of experimental timing: the ECG of all the treated mice were recorded during 30 min, 10 min before intravenous treatment, and 20 min after injection. Mice were pretreated by intraperitoneal injection with 5.10 -7 mol of bisoprolol hemifumarate (␤ 1 -specific antagonist) 10 min before intravenous injection of scFv 6H8, scFv 9C2 (negative control), or NaCl. ECG windows (1 s) represent the cardiac rhythm of one of the ten mice treated by 11 pmol scFv 6H8 (upper) and one of the six treated mice with scFv control (lower) at t ϭ 0 (injection) and t ϭ 2 (2 min after injection). b, mean (ϮS.E.) of the variation of the cardiac rhythm at different time after intravenous injection reported on the cardiac rhythm at t ϭ 0 in percent. Asterisks show that only at t ϭ 1 and 2 min the percent values are significantly different from 100% at p Ͻ 0.05 in a Student's t test. c, percent of mice, reacting with a decreased heart beating frequency, treated with scFv 6H8 or scFv control 9C2 in the confidence level in a paired match Student's t test at p Ͻ 0.05 at different times after intravenous injection. The times when mice treated are statistically different from the control NaCl treated mice are shown with asterisks.  scFv Single Chain Antibody Variable Fragment 36746 of the known three-dimensional structure allowed us to construct a realistic model of the scFv fragment in silico. Ten amino acids of the CDR of the scFv 6H8 are different from the sequence of the Protein Data Bank templates (Fig. 6). These amino acids seem to form a small groove at the VH-VL interface in this flat paratope where the essential epitopic Trp-173 of the H19C peptide or the second extracellular loop of the ␤ 2 -adrenergic receptor could fit in.
The physicochemical parameters of the epitope-scFv interaction as measured by the BIACORE system allowed us to confirm that the scFv fragment had similar binding properties as the Fab fragment of the 6H8 antibody. The amount of actively binding scFv fragments was, however, only 5% of the total purified amount of protein. The total amount of purified scFv was determined using a BCA kit and absorbance reading at 280 nm. The measured and calculated extinction coefficients are quite similar (respectively, 1.964 and 2.123), which indicates that the measured protein concentration corresponds to the scFv. The active concentration of scFv was taken into account to study the pharmacological and physiological properties of the scFv fragments. Cytochemistry experiment showed a specific labeling of membrane, membrane invagination, and cytoplasmic vesicles of the ␤ 2 -adrenergic receptor as described in Ref. 23. This distribution corresponded to diffuse membrane receptors, activated receptors concentrated in invaginations, and submembranar vesicles and new synthesized receptors transported by cytoplasmic vesicles toward the membrane.
The Fab fragments of the antibody were shown to block the activation of L-type Ca 2ϩ channels by the ␤ 2 -selective agonist clenbuterol, suggesting that the combining site-receptor complex blocked G s protein coupling, adenyl cyclase activation, and protein kinase A phosphorylation, the normal biochemical cascade induced by ␤-agonists. To directly check the mechanism of blocking, we studied the accumulation of cAMP in A431 cells, which have been shown to possess a large amount of ␤ 2 -adrenergic receptors (10). The scFv fragments indeed blocked the cAMP accumulation induced by the ␤ 2 -agonist salbutamol, but moreover they blocked the basal cAMP accumulation in the cells, ascribed to the existence of spontaneous active receptors (24). This means that the scFv fragments behave as inverse agonists, i.e. molecules specifically recognizing the resting conformation of the receptor and shifting the active 3 resting conformational equilibrium to the right. We functionally confirmed the inverse agonist properties of the scFv fragment both in vitro and in vivo.
The scFv fragment was able to decrease the spontaneous beating rate of neonatal rat cardiomyocytes in culture. It also blocked the effect on the same cells of the ␤ 2 -agonist clenbuterol. This effect did not shift the clenbuterol dose-response curve to higher agonist concentrations but decreased the maximal obtained response. These results suggest that the scFv fragment blocks the accessibility of the pharmacophore pocket in a noncompetitive manner.
A conformational change induced by the scFv on the extracellular loop could thus close the pharmacophore pocket for the agonist. The scFv was also able to decrease the beating frequency in vivo of the heart of conscious mice pretreated with the ␤ 1 -agonist bisoprolol. Although the decrease only stayed for 4 min, probably due to the rapid filtration of the scFv by the kidneys (25)(26)(27), it was similar to that observed by the ␤ 2specific inverse agonist ICI 118,551, physiologically confirming its biochemical inverse agonist activity. Recently it was shown that the inverse agonist ICI 118,551 could induce a conformation with high affinity for the G i protein (28). The decrease in beating frequency could thus not only be explained by closing of the pharmacophore pocket by the scFv but also by induction of the same receptor conformation as that induced by ICI 118,551. It must, however, be noted that the low amount of scFv injected must supersede the endogenous catecholamine concentration to exert its physiological effects, explaining the low and transient response.
To summarize, we have completely characterized a polypeptide with a specific inverse agonist activity on the ␤ 2 -adrenergic receptor. It is the first example of a G protein-coupled receptor inverse agonist, shifting by an allosteric mechanism the receptor to its resting conformation. The structural model that we obtained from the scFv fragment could help us to synthesize shorter peptide fragments, which could share the same properties.