www.elsevier.com/locate/yprep

Protein Expression and Purification 55 (2007) 319–324

Two-color selection for amplified co-production of proteinsin mammalian cells

Zahra Assur a, Ira Schieren b, Wayne A. Hendrickson a,b,*, Filippo Mancia a,*

a Department of Biochemistry and Molecular Biophysics, Columbia University, New York, NY 10032, USAb Howard Hughes Medical Institute, Columbia University, New York, NY 10032, USA

Received 17 March 2007, and in revised form 26 April 2007Available online 16 May 2007

Abstract

Structural and functional analyses for many mammalian systems depend on having abundant supplies of recombinant multi-proteincomplexes that can be produced best, or only, in mammalian cells. We present an efficient fluorescence marking procedure for establish-ing stable cell lines that overexpress two proteins in co-ordination, and we validate the method in the production of monoclonal antibodyFab fragments. The procedure has worked without fail on all seven of seven trials on Fabs, which are being used in the crystallization ofG-protein coupled receptors. This manner of efficient selection may readily be adapted for the co-production of other complexes of twoor more proteins.� 2007 Elsevier Inc. All rights reserved.

Keywords: Antibodies; Protein overexpression; Mammalian cells; Protein complexes; Fluorescence

Recent proteomic studies have strengthened the conceptthat functional units within a cell typically consist of macro-molecular assemblages as opposed to individual proteins[1]. Accordingly, there is a justified and increasing interestin structure determination and biophysical characterizationof macromolecular complexes [2]. These interests requirethe production of stable complexes, which for the most partdepends on recombinant expression in appropriate hosts.Since reconstitution from separately expressed proteins isoften frustrated by instability of the isolated components,there are obvious advantages for simultaneous overexpres-sion of multiple proteins in the same call. Bacteria, giventheir cost-effectiveness and robustness, have been by farthe most successful hosts for protein production to supportbiophysical studies, and strategies for the co-expression ofgenes in Escherichia coli are available [2,3]. Bacteria, and

1046-5928/$ - see front matter � 2007 Elsevier Inc. All rights reserved.

doi:10.1016/j.pep.2007.04.025

* Corresponding authors. Address: Department of Biochemistry andMolecular Biophysics, Columbia University, 630 West 168th Street, NewYork, NY 10032, USA. Fax: +1 212 305 7952.

E-mail addresses: wayne@convex.hhmi.columbia.edu (W.A. Hendrick-son), fm123@columbia.edu (F. Mancia).

yeast or insect cells for that matter, might be poorly suitedfor the production of mammalian proteins [4,5]. These pro-teins may be toxic to the host, may require a better matchedfolding machinery, and may need post-translational modifi-cations unavailable in the heterologous host.

Expression of recombinant proteins in mammalian cellscan be achieved through transient transfection, viral infec-tion or stable integration of expression constructs into thehost genome. Transient co-transfection of separate plas-mids carrying the individual genes represents possibly themost frequently employed technique for functional studies.Unfortunately, this has yet to translate into a generalapproach for producing abundant amounts of material asrequired for applications such as structure determination.Stable cell lines have obvious advantages for amplifiedrecombinant protein production. However, the integrationof the expression constructs into the genome of the host cellinevitably leads to a wide spectrum in protein synthesis lev-els within the same batch of transfected cells, dependingprimarily on the number of integrants and their sites ofintegration. The method therefore relies on being able toselect for cells capable of producing the most protein.

320 Z. Assur et al. / Protein Expression and Purification 55 (2007) 319–324

Through coupling of the expression of each polypeptidechain with that of different antibiotic-resistant markers,targeting a specific genomic locus for integration, or incor-porating all the components in a single plasmid one canincrease the frequency of stable transfectants able to syn-thesize all the desired components. These provide onlymarginal relief, however, in the task of identifying the bestexpressing cells, which has typically involved time-consum-ing rounds of screening to yield lines with the requiredcharacteristics [6].

Here we present and validate procedures for the rapidselection of mammalian cells that co-express multiple pro-teins. We do this by coupling the expression of each proteinchain of a complex to a separate fluorescent marker, andwe test the system in applications to the high-level expres-sion of antibody Fab fragments. Proper functionality of theexpressed complexes was demonstrated by assessing correctassembly of the antibody fragments and their ability to rec-ognize the antigen, a 5HT2c serotonin receptor.

Materials and methods

Cloning of Fv chains

Cloning of Fv regions, inclusive of signal peptidesequences, was performed with appropriate kits and degen-erate primers (Novagen) following standard proceduresand guidelines. The amplified fragments were cloned intopGEM-T vector (Promega). A second PCR was used tointroduce appropriate restriction sites for cloning intoexpression vectors.

Construction of expression vectors

pFM1.2 [7] was used to generate two vectors for the sep-arate expression of the two Fab chains. For light chainexpression, GFP was substituted for RFP, taken frompIRES2-DsRed-Express (Clontech). The CH-His6 andCL regions of D1.3 anti-lysozyme antibody were removedfrom pASK84 [8], and cloned into the MCS regions of therespective vectors. The heavy and light chains of the Fvregions were then cloned into the matching vectors to gen-erate the final expression constructs. The vectors forexpression of heavy and light chains were named pFM-FabH and pFMFabL, respectively.

Cell culture and generation of stable lines

HEK293 cells were maintained at 37 �C, in a humidifiedenvironment enriched with 5% CO2. HEK293 cells weregrown in DMEM (Chemicon) supplemented with 10%FBS (Hyclone), Penicillin, Streptomycin, L-glutamine(Pen/Strep/L-glu; Sigma).1 293 GntI� cells [9] carrying sta-

1 Abbreviations used: Pen/Strep/L-glu, Penicillin, Streptomycin, L-gluta-mine; aDDM, n-dodecyl-a-D-maltopyranoside; CHS, cholesterylhemisuccinate.

ble integration of an inducible expression cassette for5HT2c, were grown in DMEM/F12 (Gibco) supplementedwith 10% FBS, Pen/Strep/L-glu, 4 lg/mL blasticidin (Invit-rogen) and 500 lg/mL G418. A plasmid carrying resistanceto puromycin was mixed with the two expression vectors ina 1:5:5 ratio prior to transfection with lipofectamine (Invit-rogen). Stable integrants were selected by addition of 5 lg/mL puromycin to the growth medium. Production cell linesfrom single double fluorescent colonies were selected eitherby FACS sorting in Autoclone mode and subsequent visualinspection of the resulting clones, to identify single, highlyfluorescent colonies, or by manual picking of the mostintense double fluorescent colonies after antibioticselection.

FACS sorting

Data were collected using a Beckman Coulter Altra flowcytometer equipped with Autoclone. Untransfected (con-trol) and stably integrated cells were passed through thecell sorter. The viable cell population was determined usingthe forward and side scatter characteristics of the cells. Thefluorescent cells were excited at the 488 nM line of a kryp-ton–argon laser. RFP emission was detected using a 590/20 nm band pass filter. GFP emission was detected with a525/30 nm band pass filter. Viable cells (100,000) were col-lected for each pool, according to their fluorescence profile.Typically, the top fluorescent cells of the double positivepopulation (corresponding to 0.1% of the viable cell popu-lation) were chosen for cloning to single cell purity. Thiswas achieved with the flow cytometer in Autoclone mode,in 96-well plates.

Fab purification

Na–Hepes, pH 7.5, NaCl and imidazole were addedto the harvested medium to 20, 400 and 10 mM finalconcentrations, respectively. The medium was added toequilibrated Ni–NTA (Qiagen) on a column. The col-umn was washed with 10 column volumes of 20 mMNa–Hepes, pH 7.5, 400 mM NaCl, 25 mM imidazole.Purified Fab was eluted with 200 mM imidazole in thesame buffer.

Confocal microscopy on fluorescent cells

Cells were plated on 8-well CC2 treated glass slides(Nalge Nunc Int.) at a density of 1 · 105 per well. Thenext day, the cells were fixed with 4% paraformaldehydein 0.1 M phosphate buffer for 10 min at room tempera-ture. Cells were then washed 3 times with PBS and incu-bated for 30 min with the nuclear stain TOTO-3 iodide(Invitrogen) diluted at 1:2000 in PBS. Stained cells wereviewed and photographed on a Zeiss confocal micro-scope at 40· magnification in three separate emissionchannels, at 520 nM (GFP), 570 nM (RFP) and633 nM (TOTO-3).

Z. Assur et al. / Protein Expression and Purification 55 (2007) 319–324 321

Immunocytochemistry

293 GntI� cells carrying stable integration of an induc-ible expression cassette for 5HT2c were induced with tetra-cycline (2 lg/mL) and sodium butyrate (5 mM). Two dayslater, induced and uninduced (control) cells were plated on8-well CC2 treated glass slides (Nalge Nunc Int.) at a den-sity of 1 · 105 per well. After 24 h, cells were fixed with 4%paraformaldehyde in 0.1 M phosphate buffer for 10 min atroom temperature. The fixed cells were blocked with 5%goat serum in PBS and incubated for 1 h with hybridomasupernatants (for full-length Ab), and medium from Fab-expressing cells (for Fab). The secondary antibody, usedat a dilution of 1:1000, was Cy3-conjugated anti-mouseIgG (Jackson Immunoresearch). TOTO-3 iodide (Invitro-gen) was used diluted at 1:2000 as a nuclear stain. Labeledcells were visualized and photographed by confocal micros-copy as described in the previous paragraph.

Fab-antigen complex reconstitution and analysis

Purified Fab was concentrated to �2 mg/mL. 5HT2cexpressed in E. coli as a fusion to periplasmic maltose-bind-ing protein (MBP) was purified on an immobilized strepta-vidin column by means of an SBP epitope tag [10] fused tothe C-terminus of the receptor, by prior solubilization ofthe isolated membranes at 5 mg/mL membrane protein

Fig. 1. Co-expression of two polypeptide chains with fluorescent markers.cytomegalovirus (CMV) promoter. The gene for the heavy chain of the Fab isfused to its C-terminus. The light chain, in a separate vector is shown in yellowcoupled to that of green (GFP) and red (RFP) protein, respectively, by an intFluorescence data from 100,000 viable HEK293 cells expressing 2A11 after antRFP (vertical axis) and GFP (horizontal axis). The orange lines are representatfluorescence (horizontal lines from bottom to top, in increasing order) and fCorrelation of Fab 2A11 expression levels with fluorescence. Cells were sorteconfocal microscopy in the emission channels corresponding (from top to boshown under the respective columns, run out, after purification, on a Coomassieline. A clonal cell line expressing 2A11 Fab was analyzed by FACS as describ

concentration with 1% (w/v) n-dodecyl-a-D-maltopyrano-side (aDDM; Anatrace) and 0.2% cholesteryl hemisucci-nate, (CHS; Anatrace). MBP was removed by cleavage ofthe fusion with TEV protease [11] at an engineered site inthe linker region. Purified 5HT2c was concentrated to�1 mg/mL. For reconstitution, detergent was added tothe concentrated Fab to 0.1% final concentration, mixedwith 5HT2c in approximately a 2:1 molar ratio and incu-bated on ice for 2 h. Samples were run on a 7.5 mm · 60 cmTSKgel G3000SW column (TOSOH Bioscience) at a flowrate of 0.3 mL/min in 20 mM Na–Hepes, pH 7.0,200 mM NaCl, 1 mM EDTA, 0.025% aDDM.

Results

We first adapted a previously described fluorescence-selection vector for the high-level production of proteinsin mammalian cells [7] for use with a second color. Thisvector carries a strong, constitutive promoter sequence,derived from cytomegalovirus (CMV [12]). Target geneexpression is coupled via an IRES element [13] to that ofgreen fluorescent protein (GFP [14]), and was modified toreplace the coding sequence for GFP with that for red fluo-rescent protein (RFP [15]). The target genes for our currentapplication are those for antibody Fab fragments corre-sponding to monoclonal antibodies (MAbs) against a ratG-protein coupled receptor (GPCR) for serotonin

(a) Schematic of the expression constructs. Expression is driven by thein blue colors (VH and CH1 D1.3), followed by a hexa-histidine tag (H6)

(VL) and orange (CL D1.3). Expression of the heavy and light chains isernal ribosome entry site (IRES). (b) Fluorescence profile of sorted cells.ibiotic selection are shown. Fluorescence is plotted in logarithmic scale forive of the 4 · 4 grid used to select negative, +, ++ and +++ cells, for RFPor GFP fluorescence (vertical lines left to right, in increasing order). (c)d according to their fluorescence profile, and the pools photographed byttom) to RFP, GFP and the nuclear stain TOTO-3. The resulting Fab is

blue-stained SDS–PAGE gel. (d) Fluorescence profile of a production celled for panel (b).

Fig. 2. Fluorescence levels, yields and folding of expressed Fabs fromproduction cell lines. (a) Cells secreting a specific Fab fragment, 1C4,2A11, 5D8 or 1A12, as indicated on the top of each correspondingcolumn, were visualized by confocal microscopy for fluorescence fromRFP (top row), GFP (middle row) and TOTO-3 (bottom row). (b) Thedifferent Fabs were run, after purification, on an SDS–PAGE denaturinggel after treatment with loading dye with (+) and without (�)b-mercaptoethanol. The identity of each Fab is shown above the corre-sponding lane.

322 Z. Assur et al. / Protein Expression and Purification 55 (2007) 319–324

(5HT2c [16]), which were generated as previously described[17]. In this instance, we further modified the GFP- andRFP-containing vectors to include coding sequences forthe conserved heavy-chain CH1 and light-chain CLdomains from the extensively characterized, anti-lysozymeantibody D1.3 [18] ahead of the respective IRES elements.CH1 and CL domains do not confer antigen specificity tothe antibody, and are interchangeable, even between spe-cies [6,19]. These provided the scaffold to express clonedVH and VL domains as Fab fragments. Six histidine resi-dues were genetically fused to the C-terminus of CH1 forpurification purposes [8]. Finally, coding sequences forVH and VL domains, inclusive of signal peptide sequences,were cloned from our hybridoma cells [17] using estab-lished protocols and inserted in-frame into vectors withmatching C regions. A schematic of the two resultingexpression constructs, one for each chain of the Fab frag-ment, is shown in Fig. 1a.

After co-transfection of the two plasmids into HEK293cells, together with that for a selectable marker for antibi-otic resistance, stable integrants were selected. As expected,the resulting antibiotic-resistant colonies displayed a widespectrum of fluorescence levels, due to the presence of bothRFP and GFP. First, then, we wanted to determine if thecells with both red and green fluorescence also actuallysecreted antibody fragments. Second, we were interestedin verifying the extent to which fluorescence intensity cor-related with Fab expression. In other words, we wantedto test the predictive value of our fluorescent markers forthe production of desired protein products.

Using a fluorescence-activated cell sorter (FACS), weclassified cells according to their two-color fluorescenceprofiles into a 4 · 4 grid of �, +, ++ and +++ for eachcolor. The vast majority of cells after antibiotic selectionwere non-fluorescent, or only green, and populations felloff progressively with intensity for each color (Fig. 1b).Pools of cells with pertinent characteristics were selectedand analyzed. Cells from these selected pools were visual-ized by confocal microscopy for fluorescence from RFP,GFP and a nuclear stain, and Fabs secreted from these dif-ferent pools were purified by metal affinity chromatogra-phy and observed by SDS–PAGE. Results from suchexperiments are shown in Fig. 1b for 2A11, a Fab fragmentfrom an anti-5HT2c MAb [17]. We find that while negativecells (RFP�/GFP�) showed no detectable antibody frag-ments, Fab expression was always associated with fluores-cence. Moreover, robust Fab expression could only bedetected in double-positive cells, in amounts that were cor-related to the levels of GFP and RFP fluorescence.

Each production cell line was cloned from a singleRFP+++/GFP+++ progenitor cell and chosen from afew such cultures to assure that maximal fluorescence wasmaintained (Fig. 2a). For these experiments, theRFP+++/GFP+++ grid was typically set to include only0.1% of the cells. FACS analysis of a production cell linefor Fab 2A11, after having been maintained for severalmonths in continuous culture is shown in Fig. 1d. Cells

were grown in adherent monolayers, and for Fab 2A11the yield was quantified to be above 30 mg of purified pro-tein per liter of medium from the endpoint of a time courseexperiment (Fig. 3) and from subsequent scale-ups (datanot shown). Comparable yields have also been obtainedfrom each of six other Fab productions attempted with thissystem, three reactive to 5HT2c, and three to a differentGPCR. Relative yields were judged from the relative inten-sities of Coomassie blue-stained bands on gels. Three ofthese, all from antibodies against 5HT2c, are shown inFig. 2. Ultimate production levels (Fig. 2b) seem to corre-late with maximal fluorescence across these proteins(Fig. 2a) much as Fab production correlates with fluores-cence for a given protein (Fig. 1c).

In contrast to our 100% success rate so far from two-color selection in HEK293 cells, alternative methods haveproved far less reliable for us. We attempted Fab produc-tion by digestion of the intact IgG with papain, but onlyone of four different antibody-secreting hybridoma celllines tried could be adapted to grow in serum-free mediumand scaled-up for Fab production. We also attempted to

Fig. 3. Time course of Fab expression. 18 · 106 cells expressing Fab 2A11were equally divided into 6, 10 cm-diameter dishes, in 10 mL of medium.At each time point, the medium from a plate was removed, cells counted,and the Fab purified. 1/10th of each purified sample was run on an SDS–PAGE gel, stained with Coomassie blue. The length of each incubationperiod is indicated above the matching lane, in days, and the correspond-ing number of cells, in millions, below. The yield of Fab resulting from a12 day incubation period was quantified by absorbance reading at280 nM.

Z. Assur et al. / Protein Expression and Purification 55 (2007) 319–324 323

express recombinant Fabs in E. coli following establishedprocedures (reviewed in [20]), but achieved yields in themicrograms per liter range for the two antibodies tested.Reported expression levels of Fabs in E. coli [20,21] seemhighly variable, while in the yeast Pichia pastoris, properFab assembly seems to be the major rate limiting factor[22]. In both these hosts, high-density cell growth in appro-priate fermenters appears to have been essential for veryhigh yields [20,22]. Indeed, if required, yields from our celllines could most likely be improved dramatically by opti-mizing cell growth conditions, as in current practice forindustrial production of antibodies in mammalian cells [6].

Fig. 4. Functionality of the expressed Fab. (a) Immunocytochemistry using recafter induction of 5HT2c expression. (b) Overlay of elution profiles from size-ex5HT2c (blue trace), purified Fab (green trace) and reconstituted 5HT2c-Fab c

The light and heavy chains of the purified Fab appearedto be disulfide-bond linked, in agreement with what is typ-ical for a correctly folded antibody (Fig. 2b). Furthermore,functionality of the expressed Fab was assessed by immu-nohistochemistry on antigen-expressing cells, showing astaining profile essentially identical to that observed withthe intact IgG antibody derived from hybridomas(Fig. 4a). Finally, in preparation for crystallization trials,the Fab-5HT2c complex was successfully reconstitutedwith purified components as shown by size-exclusion chro-matography (Fig. 4b). All these data are consistent withproper assembly and function of the expressed Fab.

Discussion

Both the tight correlation that we observe betweenproduction levels of functional antibody fragments andtwo-color fluorescence of transfected HEK293 cells(Fig. 1c) and the high reliability of the procedure (sevensuccesses in seven trials) provide a compelling validationof this selection method. Our application to monoclonalantibody production itself has broad potential. Mono-clonal antibodies are important tools for structural stud-ies on membrane proteins [23]. Indeed, Fabs describedhere are now being used in our crystallization of the5HT2c receptor. Monoclonal antibodies are also gainingan increasingly relevant role in clinical applications[6,19]. This two-color system should have other applica-tions as well, however, as we have already shown thatour previous GFP system works for other secreted pro-teins and for membrane proteins such as GPCRs [7].Moreover, given the availability of several different fluo-rescent proteins [24], one could anticipate being able to

ombinant Fab 2A11 (rFab) and intact antibody (IgG) on cells, before andclusion chromatography, monitored by absorbance at 280 nM, of purifiedomplex (red trace).

324 Z. Assur et al. / Protein Expression and Purification 55 (2007) 319–324

extend this methodology to studies of multi-componentprotein complexes.

In contrast to time-consuming multi-step conventionalamplification systems for generating stably transfectedmammalian cell lines, this fluorescence-based approach issimple and rapid. Fluorescence surveillance is also conve-nient for monitoring and maintaining the continued healthof an established cell line. We hope that these proceduresmay aid in finally bridging the gap between multi-geneexpression in mammalian cells for functional studies, whichis commonplace and routine, and the use of mammaliancells for biophysical studies, which has had very limitedpopularity even as the structural community is increasinglyinterested in macromolecular assemblages.

Acknowledgments

We are grateful to Richard Axel for discussions. Wethank Susan Brenner-Morton for advice and help withcloning of the antibody fragments and Yonghua Sun forexcellent technical assistance. This work was supported inpart by NIH Grants GM68671 and GM75026.

References

[1] A.C. Gavin, G. Superti-Furga, Protein complexes and proteomeorganization from yeast to man, Curr. Opin. Chem. Biol. 7 (2003) 21–27.

[2] C. Romier, M. Ben Jelloul, S. Albeck, G. Buchwald, D. Busso, P.H.Celie, E. Christodoulou, V. De Marco, S. van Gerwen, P. Knipscheer,J.H. Lebbink, V. Notenboom, A. Poterszman, N. Rochel, S.X.Cohen, T. Unger, J.L. Sussman, D. Moras, T.K. Sixma, A. Perrakis,Co-expression of protein complexes in prokaryotic and eukaryotichosts: experimental procedures, database tracking and case studies,Acta Crystallogr. D. Biol. Crystallogr. 62 (2006) 1232–1242.

[3] S. Tan, A modular polycistronic expression system for overexpressingprotein complexes in Escherichia coli, Protein Expr. Purif. 21 (2001)224–234.

[4] S. Geisse, H. Gram, B. Kleuser, H.P. Kocher, Eukaryotic expressionsystems: a comparison, Protein Expr. Purif. 8 (1996) 271–282.

[5] J. Yin, G. Li, X. Ren, G. Herrler, Select what you need: acomparative evaluation of the advantages and limitations of fre-quently used expression systems for foreign genes, J. Biotechnol. 127(2007) 335–347.

[6] J.R. Birch, A.J. Racher, Antibody production, Adv. Drug. Deliv.Rev. 58 (2006) 671–685.

[7] F. Mancia, S.D. Patel, M.W. Rajala, P.E. Scherer, A. Nemes, I.Schieren, W.A. Hendrickson, L. Shapiro, Optimization of proteinproduction in mammalian cells with a coexpressed fluorescentmarker, Structure 12 (2004) 1355–1360.

[8] A. Skerra, A general vector, pASK84, for cloning, bacterial produc-tion, and single-step purification of antibody Fab fragments, Gene141 (1994) 79–84.

[9] P.J. Reeves, N. Callewaert, R. Contreras, H.G. Khorana, Structureand function in rhodopsin: high-level expression of rhodopsin withrestricted and homogeneous N-glycosylation by a tetracycline-induc-ible N-acetylglucosaminyltransferase I-negative HEK293S stablemammalian cell line, Proc. Natl. Acad. Sci. USA 99 (2002) 13419–13424.

[10] A.D. Keefe, D.S. Wilson, B. Seelig, J.W. Szostak, One-step purifi-cation of recombinant proteins using a nanomolar-affinity streptavi-din-binding peptide, the SBP-Tag, Protein Expr. Purif. 23 (2001) 440–446.

[11] R.B. Kapust, J. Tozser, J.D. Fox, D.E. Anderson, S. Cherry, T.D.Copeland, D.S. Waugh, Tobacco etch virus protease: mechanism ofautolysis and rational design of stable mutants with wild-typecatalytic proficiency, Protein Eng. 14 (2001) 993–1000.

[12] D.R. Thomsen, R.M. Stenberg, W.F. Goins, M.F. Stinski, Promoter-regulatory region of the major immediate early gene of humancytomegalovirus, Proc. Natl. Acad. Sci. USA 81 (1984) 659–663.

[13] S. Vagner, B. Galy, S. Pyronnet, Irresistible IRES. Attracting thetranslation machinery to internal ribosome entry sites, EMBO Rep. 2(2001) 893–898.

[14] R.Y. Tsien, The green fluorescent protein, Annu. Rev. Biochem. 67(1998) 509–544.

[15] M. Knop, F. Barr, C.G. Riedel, T. Heckel, C. Reichel, Improvedversion of the red fluorescent protein (drFP583/DsRed/RFP),Biotechniques 33 (2002) 592–602.

[16] D. Julius, A.B. MacDermott, R. Axel, T.M. Jessell, Molecularcharacterization of a functional cDNA encoding the serotonin 1creceptor, Science 241 (1988) 558–564.

[17] F. Mancia, S. Brenner-Morton, R. Siegel, Z. Assur, Y. Sun, I.Schieren, M. Mendelsohn, R. Axel, W.A. Hendrickson, Productionand characterization of monoclonal antibodies sensitive to confor-mation in the 5HT2c serotonin receptor, Proc. Natl. Acad. Sci. USA104 (2007) 4303–4308.

[18] A.G. Amit, R.A. Mariuzza, S.E. Phillips, R.J. Poljak, Three-dimensional structure of an antigen–antibody complex at 2.8 Aresolution, Science 233 (1986) 747–753.

[19] J.M. Reichert, C.J. Rosensweig, L.B. Faden, M.C. Dewitz, Mono-clonal antibody successes in the clinic, Nat. Biotechnol. 23 (2005)1073–1078.

[20] A. Skerra, Bacterial expression of immunoglobulin fragments, Curr.Opin. Immunol. 5 (1993) 256–262.

[21] S. Corisdeo, B. Wang, Functional expression and display of anantibody Fab fragment in Escherichia coli: study of vector designs andculture conditions, Protein Expr. Purif. 34 (2004) 270–279.

[22] B. Gasser, M. Maurer, J. Gach, R. Kunert, D. Mattanovich,Engineering of Pichia pastoris for improved production of antibodyfragments, Biotechnol. Bioeng. 94 (2006) 353–361.

[23] C. Hunte, H. Michel, Crystallisation of membrane proteins mediatedby antibody fragments, Curr. Opin. Struct. Biol. 12 (2002) 503–508.

[24] K.A. Lukyanov, D.M. Chudakov, S. Lukyanov, V.V. Verkhusha,Innovation: photoactivatable fluorescent proteins, Nat. Rev. Mol.Cell. Biol. 6 (2005) 885–891.