MC

YS*UR

R

nflcsstlwebttictotccehhfsttA

d

R1

Protein Expression and Purification 17, 249–259 (1999)Article ID prep.1999.1125, available online at http://www.idealibrary.com on

1CA

atrix-Assisted Refolding of Single-Chain Fv–ellulose Binding Domain Fusion Proteins

evgeny Berdichevsky,* Raphael Lamed,* Dan Frenkel,* Uri Gophna,* Edward A. Bayer,†ima Yaron,‡ Yuval Shoham,‡ and Itai Benhar*,1

Department of Molecular Microbiology and Biotechnology, The George S. Wise Faculty of Life Sciences, Tel-Avivniversity, Ramat Aviv 69978, Israel; †Department of Biological Chemistry, The Weizmann Institute of Sciences,ehovot, Israel; and ‡Department of Food Engineering and Biotechnology, Technion IIT, Haifa, Israel

eceived April 26, 1999, and in revised form July 9, 1999

ihctgct

cccfshtntrrtttd

pmntiysmdecg

We describe a method for the isolation of recombi-ant single-chain antibodies in a biologically active

orm. The single-chain antibodies are fused to a cellu-ose binding domain as a single-chain protein that ac-umulates as insoluble inclusion bodies upon expres-ion in Escherichia coli. The inclusion bodies are thenolubilized and denatured by an appropriate chao-ropic solvent, then reversibly immobilized onto a cel-ulose matrix via specific interaction of the matrixith the cellulose binding domain (CBD) moiety. The

fficient immobilization that minimizes the contactetween folding protein molecules, thus preventingheir aggregation, is facilitated by the robustness ofhe Clostridium thermocellum CBD we use. This CBDs unique in retaining its specific cellulose bindingapability when solubilized in up to 6 M urea, whilehe proteins fused to it are fully denatured. Refoldingf the fusion proteins is induced by reducing with timehe concentration of the denaturing solvent while inontact with the cellulose matrix. The refolded single-hain antibodies in their native state are then recov-red by releasing them from the cellulose matrix inigh yield of 60% or better, which is threefold origher than the yield obtained by using published re-

olding protocols to recover the same scFvs. The de-cribed method should have general applicability forhe production of many protein–CBD fusions in whichhe fusion partner is insoluble upon expression. © 1999

cademic Press

Many recombinant proteins, particularly proteins ofiagnostic and therapeutic potential such as antibod-

1 To whom correspondence should be addressed at Green Bldg.,oom 202, Tel-Aviv University, Ramat Aviv 69978, Israel. Fax:

s973-3-6409407. E-mail: benhar@post.tau.ac.il.

046-5928/99 $30.00opyright © 1999 by Academic Pressll rights of reproduction in any form reserved.

es, lymphokines, receptors, enzymes, and enzyme-in-ibitors, have been produced from transformed hostells containing recombinant DNA. The host cells areransformed with an expression vector containingenes encoding the proteins of interest and then areultured under conditions that favor the production ofhe desired protein.

Often, the heterologous protein produced by the hostells precipitates within the cell to form refractile (in-lusion) bodies. For its isolation in the native (biologi-ally active) state, the protein should be separatedrom cell debris, solubilized with a chaotropic agentuch as high concentrations of urea or guanidiniumydrochloride, and then refolded by gradual removal ofhe denaturant (reviewed in 1). In cases in which theative state of the protein is dependent on the forma-ion of intramolecular disulfide bonds, the protein iseduced while in the denatured state by addition ofeducing agents (2). The formation of disulfide bonds inhe protein during the refolding process is initiated byhe addition of oxidizing agents in the refolding bufferhat is used during removal of denaturant from theenatured protein or subsequent to it.Denaturation followed by refolding of recombinant

roteins from inclusion bodies is frequently the onlyeans by which the protein may be recovered in the

ative form. However, refolding is an empirical processhat has to be optimized for each particular protein ofnterest (1). Under conventional folding conditions, theield of renaturation is often exceedingly low due toide reactions such as aggregation or formation of ther-odynamically stable, but nonnative, folding interme-

iates. Furthermore, refolding should be performed atxtremely low protein concentrations due to the kineticompetition between correct folding and incorrect ag-regation (3). Aggregation during refolding may be

omewhat inhibited by the addition of cosolvents such

249

airt

ipmtiif(tcrostbhoptaadopctbtn

itfankmmdlpanipwpit

t

a(mdFbtncsobcCfamhzptf

brbCsdtsiCffdifmtpasltmd(

abaCvl

250 BERDICHEVSKY ET AL.

s amino acids (arginine) or other salts or by refoldingn the presence of molecular chaperones (4,5), but suchemedies are inadequate, particularly for large indus-rial scale.

Aggregation during refolding may be prevented bymmobilization of the denatured protein onto a solidhase prior to initiation of refolding. Some proteinsay be charged under refolding conditions, so that

hey may be immobilized onto ion-exchange matricesn a low-ionic strength denaturant. Refolding is thennduced by gradual removal of denaturant and releaserom the matrix by increasing the salt concentration6). This approach, while appropriate for charged pro-eins, is not adequate for proteins for which the netharge under pH conditions that are compatible withefolding is too small to allow efficient immobilizationnto an ion-exchange matrix (as is the case with manyingle-chain antibodies, our unpublished observa-ions). Furthermore, the protein may be immobilizedy multiple interactions with the matrix, which mayinder its ability to fold properly. When immobilizationf the denatured protein by itself is not an option, it isossible to engineer an affinity tag linked to the N-erminus or to the C-terminus of the protein. Smallffinity tags, such as a hexahistidine tag, or a hexa-rginine tag, were applied for such purposes. Histi-ine-tagged proteins may be immobilized and refoldedn a metal-carrying support (7), while arginine-taggedroteins may be immobilized and refolded on a supportarrying polyanionic groups (8). One must assume thathe folding of the immobilized protein is not perturbedy its proximity to the matrix and that placing a shortagging peptide at its terminus does not divert it into aonproductive folding pathway.Refolding of an immobilized protein with fewer phys-

cal constraints may be achieved by its immobilizationhrough a polypeptide fusion partner. In folding ofusion proteins, when the individual fusion partnersre joined by a flexible peptide linker, the fusion part-ers fold independent of each other (9). However, mostnown protein tags (such as glutathione S-transferase,altose-binding protein, staphylococcal protein A, andost bacterial and fungal cellulose and chitin-binding

omains) will not be useful for the purpose of immobi-ization of a denatured fusion protein onto the appro-riate affinity support. This is because the proteinffinity tags themselves are not functional in the de-atured state. A possible solution is to use as an affin-

ty tag a protein whose compact and stable structurerevents it from unfolding under the conditions underhich its fusion partner is completely denatured. Suchroteins are rare, and one example, as discussed below,s the cellulose-binding domain from the Clostridiumhermocellum cellulosome.

Cellulosomes are multienzyme complexes devoted to

he efficient degradation of cellulose and hemicellulose by d

highly specialized class of cellulolytic microorganisms10). The cellulosome in C. thermocellum comprises nu-erous subunits. Some, but not all, cellulases also have a

istinct noncatalytic cellulose-binding domain (CBD).ungal and bacterial CBDs appear to be different, theacterial form being much larger (11). The cellulase sys-em of C. thermocellum contains both cellulosomal andoncellulosomal cellulases and xylanases. CBDs have re-ently been described for noncatalytic proteins, calledcaffoldins (10), which are responsible for the integrationf enzyme subunits into the cellulosome complex. CBDsind to, and can be eluted from, cellulose under mildonditions and specific reagents are not required. TheBDs retain their cellulose-binding properties when

used to heterologous proteins. They were proposed asffinity tags for protein purification and for enzyme im-obilization (10,11). Other types of CBD fusion proteinsave also been described for protein purification or en-yme immobilization purposes (11–14). Thus, CBDs ap-ear as promising candidates for application as affinityags for protein recovery through binding and elutionrom cellulose.

C. thermocellum CBD (a type IIIa clostridial CBD) haseen cloned (15), expressed, and purified from Esche-ichia coli (16) and its three-dimensional structure haseen solved at high resolution (17). A unique property oflostridium thermocellum CBD is that it maintains itspecific cellulose binding properties under conditions un-er which most proteins are denatured and nonfunc-ional. We found that this CBD, alone or when fused toingle-chain antibodies, will bind reversibly to cellulosen up to 6 M urea. This property makes C. thermocellumBD an ideal candidate for an affinity tag to be applied

or matrix-assisted (in this case, cellulose-assisted) re-olding of fusion proteins. When a protein–CBD fusion isenatured, allowed to bind cellulose, and then refolded, its immobilized onto the cellulose matrix while the proteinused to CBD is still in the unfolded state. Gradual re-oval of the denaturant then facilitates the refolding of

he protein into the native state. Finally, the refoldedrotein can be separated from the cellulose matrix bypplying appropriate elution conditions. When neces-ary, the protein can be separated from CBD by proteo-ytic cleavage and chromatographic separation. In addi-ion, after being refolded while bound to the celluloseatrix, the protein–CBD fusion may by recovered by

igestion of the cellulose matrix by a cellulolytic enzyme18).

Recently we incorporated C. thermocellum CBD intophage display system for efficient isolation of ligand-

inding proteins, particularly single-chain antibodies,nd for engineering the matrix-binding properties ofBD itself (19). The modular design of our expressionectors makes it possible to proceed directly from iso-ation of the ligand-binding protein to its efficient pro-

uction as described herein. Our examples are of the

psr

M

C

sItG(bAtaca

MtsFtGpmmaPrcpo

F

Fna

251MATRIX-ASSISTED REFOLDING OF CBD FUSION PROTEIN

roduction of single-chain antibodies (scFvs)–CBD fu-ion proteins, many of which are notoriously difficult toefold by standard refolding protocols (20).

ATERIALS AND METHODS

onstruction of scFv–CBD Bacterial ExpressionVectors

Initially we constructed vectors for phage display ofcFv–CBD fusion proteins as described previously (19).n those vectors, the C-terminus of the scFv was fusedhrough a short tether composed of Ala-Ala-Ala-Gly-ly to the N-terminus of the C. thermocellum CBD

GenBank Accession No. X68233). Next, we introducedy PCR the sequence encoding Asp-Tyr-Lys-Asp-Asp-sp-Asp-Lys-Leu between the short tether to the N-

erminus of the CBD cassette. This sequence comprisesFLAG epitope (21), followed by an enterokinase

leavage site. A complete expression cassette was then

IG. 1. Expression vectors pFEKCA3d-scFv (A) and pH6T-FEKCucleotide and amino acid sequences of the scFv-Flag-enterokinase-t the top.

ssembled by PCR and cloned into pET3d (Novagen, S

adison, WI) for intracellular expression. A scheme ofhe constructed expression vector pFEKCA3d-scFv ishown in Fig. 1A. A second expression vector, pH6T-EKCA3d-scFv (Fig. 1B), was constructed by linking ofhe sequence encoding for Met-(His)6-Leu-Val-Pro-Arg-ly-Ser to the 59 end of the scFv cassette ofFEKCA3d-scFv. When expressed from the latter plas-id the scFv–CBD fusion protein is preceded by aetal-binding His tag followed by the thrombin cleav-

ge site. (Details regarding PCR primer sequences andCR conditions are available upon request.) For bacte-ial expression, the plasmids were introduced into E.oli strain BL21(DE3) (Novagen), carrying the T7 RNAolymerase gene on its chromosome under the controlf an IPTG-inducible lacUV5 promoter.

ermentation

Typically, cultures were grown to an OD600 of 2–3 in

d-scFv (B) used for expression of scFv–CBD fusion proteins. Theregion in A and of the His tag-thrombin site-scFv in B are shown

A3CBD

uperbroth medium (35 g/liter tryptone, 20 g/liter

yge3sdpci

R

i(at2aefupitor6ST5dpwu2r

pbtfuNum(tf

S

toa

bi(wci32

cfcllMbmc1pm52

m(sPb

C

lmtllsitv

rjwtTyWtw

252 BERDICHEVSKY ET AL.

east extract, 5 g/liter NaCl, supplemented with 1%lucose and 100 mg/ml ampicillin) at 37°C. Proteinxpression was induced by addition of 1 mM IPTG forh. Cells were harvested by centrifugation and the

cFv–CBD fusion protein was isolated and purified asescribed below. Induction of pFEKCA3d-scFv- orH6T-FEKCA3d-scFv-carrying cells resulted in the ac-umulation of the recombinant proteins as insolublenclusion bodies.

ecovery of Inclusion Bodies

The bacterial paste (typically, 5 g from 500 ml ofnduced culture) was suspended in 70 ml of 50 mM TrisHCl), pH 8.0/20 mM EDTA (TE 50:20) and lysed byddition of 20 mg lysozyme (Sigma) for 1 h at roomemperature. Five milliliters of 5 M NaCl and 5 ml of5% Triton X-100 (Sigma, Rehovot, Israel) were addednd the cells were disrupted with a mechanical homog-nizer (Tissumizer, Heidolph, Germany). The insolubleraction of the disrupted cells was collected by centrif-gation at 20,000g for 20 min at 4°C and was resus-ended in TE 50:20 containing 1% Triton X-100. Thensoluble fraction was collected again by centrifuga-ion, resuspended in 40 ml of TE 50:20, and sedimentednce more. The collected insoluble material was en-iched in its scFv–CBD fraction, which accounted for0–80% of the inclusion body protein as judged byDS–polyacrylamide gel electrophoresis (SDS–PAGE).he inclusion bodies were solubilized in 8 M urea/TE0:20 and reduced by addition of 50 mM 2,3,-dihy-roxybutane-1,4-dithiol (DTE; Sigma, Israel). For com-lete solubilization and reduction, the inclusion bodiesere left for 2 h at room temperature. The urea-insol-ble material was removed by centrifugation at2,000g for 10 min, and the urea-soluble protein wasefolded as described below.

When the scFv–CBD fusion protein was made fromH6T-FEKCA3d-scFv, the urea-solubilized inclusionodies were subjected to metal-chelate affinity chroma-ography on a Ni-NTA resin (Qiagen, Germany) asollows. EDTA was omitted from the washes, and therea-solubilized inclusion bodies were loaded onto ai-NTA column. The column was washed with 10 col-mn volumes of 8 M urea/50 mM Tris (HCl), pH 8.0/20M imidazole and eluted with 8 M urea/50 mM Tris

HCl), pH 8.0/250 mM EDTA. The eluted protein washen reduced by addition of 50 mM DTE prior to re-olding.

tandard Refolding and Purification of scFv–CBDFusion Proteins

The standard refolding protocol is a modification ofhe protocol described by Buchner et al. (2), which isptimized for the refolding of single-chain antibodies

nd fusion proteins thereof. Refolding was initiated s

y rapid dilution of the solubilized and reducednclusion bodies into 100 volumes of 20 mM TrisHCl), pH 7.5/1 mM EDTA/250 mM NaCl (TBS) thatas adjusted to pH 9.5 by addition of NaOH. Typi-

ally, 3.3 mg of inclusion bodies protein was dilutednto 100 ml of TBS to a final protein concentration of3 mg/ml. The refolding solution was left at 4°C for4 –72 h without stirring.Following refolding, precipitates were removed by

entrifugation at 22,000g for 20 min. The soluble re-olded protein was adsorbed with 1 g of crystallineellulose (Sigmacell 20; Sigma) for 1 h, and the cellu-ose was recovered by centrifugation. The cellulose pel-et was washed twice with 40 ml of TBS adjusted to 1

NaCl and once with 40 ml of distilled water. Theound scFv–CBD was eluted twice with 10 ml of 100M NaOH/100 mM NaCl, pH 13.0. The eluates were

ombined and neutralized to pH 7.5 by dialysis against00 volumes of TBS at 4°C for 16 h. When required, theurified protein was concentrated using a Centriconicroconcentrator (Amicon, Beverley, MA) with a

0,000-kDa cutoff. The purified protein was stored at20°C.Protein concentration of all fractions was deter-ined by the Coomassie Plus Protein Assay Reagent

Pierce, Rockford, IL) using bovine serum albumin as atandard. Proteins were analyzed throughout by SDS–AGE. Gels were stained with Coomassie brilliantlue.

ellulose-Assisted Refolding and Purificationof scFv–CBD Fusion Proteins

Cellulose-assisted refolding was carried out as fol-ows: the solubilized reduced inclusion bodies were

ixed with an equal volume of either 4 g/10 ml crys-alline cellulose in TBS or 0.2 g/10 ml amorphous cel-ulose prepared by phosphoric-acid digestion of crystal-ine cellulose as described (18). The suspension wastirred for 1 h at room temperature, to allow for bind-ng to cellulose, and then transferred into dialysisubes and dialyzed for 24–72 h at 4°C against 100olumes of TBS adjusted to pH 9.5 with NaOH.Following refolding, the cellulose composite was

ecovered and washed twice with 40 ml of TBS ad-usted to 1 M NaCl and once with 40 ml of distilledater. The cellulose-bound scFv–CBD was eluted

wice with 10 ml of 100 mM NaOH/100 mM NaCl.he eluates were combined and neutralized by dial-sis against 100 volumes of TBS at 4°C for 16 h.hen required, the purified protein was concen-

rated using a Centricon microconcentrator (Amicon)ith a 50,000-kDa cutoff. The purified protein was

tored at 220°C.

R

ffspwlscw7lagPgaKeHsfiag

A

pficdwt

A

pgpbfma(prgagTd

R

E

pt5pra35fdmsmmifrtC

ofP

FcpcIwlww5also

253MATRIX-ASSISTED REFOLDING OF CBD FUSION PROTEIN

ecovery of Purified scFv by Proteolytic DigestionWhile Immobilized

As an alternative to basic pH elution of scFv–CBDusion proteins from cellulose, we liberated the scFvrom the fusion protein by utilizing the enterokinaseite we had engineered at the 39 end of the scFvs in theFEKCA expression system (Fig. 1). Here, the proteinas subjected to proteolytic cleavage while immobi-

ized onto the cellulose matrix. Typically, 20 mg ofcFv–CBD fusion protein was immobilized onto 2 mg ofrystalline cellulose. The cellulose-bound protein wasashed and equilibrated in 20 mM Tris (HCl), pH.4/50 mM NaCl/2 mM CaCl2. One unit of enterokinaseight chain (New England Biolabs, Beverley, MA) wasdded and incubation was for 8 h at 23°C. The undi-ested and digested samples were analyzed by SDS–AGE followed by immunoblotting, in which the undi-ested scFv–CBD and the free scFv were detected withn anti-FLAG monoclonal antibody, M2 (Eastmanodak Co., New Haven, CT), recognizing the FLAGpitope at the 39 end of the scFv (Fig. 1), followed withRP-conjugated rabbit anti-mouse antibodies (Jack-

on Laboratories, West Grove, PA). The nitrocelluloselter was developed with the ECL Western blottingnalysis system (Amersham, Buckinghamshire, En-land) as described by the vendor.

nalytical Gel-Filtration Chromatography

To assess the purity and aggregation status of theurified scFvs–CBD, aliquots were subjected to gel-ltration chromatography on a Superose 12 HR 10/30olumn (Pharmacia, Sweden). TBS buffer was used toevelop the column at 0.5 ml/min flow rate. The proteinas monitored at 280 nm. Bio-Rad gel-filtration pro-

ein standards were used to calibrate the column.

nalysis of Antigen Binding by ELISA

To compare the immunoreactivity of the scFv–CBDroteins refolded under various conditions, their anti-en binding properties were compared in an ELISArocedure. Antigen at 10–20 mg/ml in 50 mM NaHCO3

uffer, pH 9.6, was coated onto 96-well ELISA platesor 16 h at 4°C. The plates were blocked with 2% skimilk powder in phosphate buffered saline (PBS), and

ll washes were with PBS containing 0.05% Tween 20Sigma). Serial dilutions of the analyzed scFv–CBDrotein were added to the plates for a 1 h incubation atoom temperature. The scFv–CBD bound to the anti-en on the plate was detected with a rabbit polyclonalnti-CBD antiserum followed with HRP-conjugatedoat anti-rabbit antibodies (Jackson Laboratories).he peroxidase substrate ABTS (Sigma) was used for

evelopment and the color was recorded at 405 nm. i

ESULTS

xpression and Purification of the Anti-b-amyloidPeptide scFv 508(Fv)–CBD Fusion

We isolated the scFv 508(Fv) from an antibody–hage display library using Alzheimer’s b-amyloid pep-ide as a capturing antigen (in preparation). The08(Fv) was cloned into pFEKCA3d (Fig. 1A) for ex-ression. Upon IPTG induction of BL21(DE3) cells car-ying pFEKCA3d-508(Fv), the overexpressed proteinccumulated as insoluble inclusion bodies (Fig. 2, lane). Five hundred milliliters of induced culture yielded.6 g of wet cell paste. Inclusion bodies were recoveredrom the bacterial paste, solubilized, and reduced asescribed under Materials and Methods. A total of 170g of wet inclusion bodies was recovered. The urea-

oluble reduced inclusion bodies were 12 ml at 3.2g/ml (protein concentration), which was approxi-ately 80% pure. One milliliter of solubilized reduced

nclusion bodies (3 mg) was refolded by standard re-olding. The rest (35 mg) was used in cellulose-assistedefolding on crystalline cellulose. Table 1 summarizeshe protein quantities and production yield of 508(Fv)-BD.As shown in the Table 1, for 508(Fv)-CBD the yield

btained by cellulose-assisted refolding was almost 20-old that obtained by standard refolding. The SDS–AGE analysis of the purification procedure is shown

IG. 2. Expression and purification of 508(Fv)-CBD. Lane 1, totalell extract from uninduced BL21(DE3) cells carrying plasmidFEKCA3d-508(Fv). Lane 2, total cell extract from BL21(DE3) cellsarrying plasmid pFEKCA3d-508(Fv) induced for 3 h with 1 mMPTG. Lane 3, washed, solubilized, and reduced inclusion bodies thatere used in refolding. Lane 4, protein that did not bind to crystal-

ine cellulose during cellulose-assisted refolding. Lane 5, proteinashed away from crystalline cellulose with TBS. Lane 6, proteinashed away from crystalline cellulose with distilled water. Lane 7,08(Fv)-CBD recovered from crystalline cellulose by high-pH elutionnd neutralization by dialysis. 5–10 mg protein was loaded in eachane of a 14% SDS–polyacrylamide gel. Proteins were visualized bytaining with Coomassie brilliant blue. The arrow marks the positionf the scFv–CBD fusion protein.

n Fig. 2. Purification to .90% was achieved by cellu-

lpqi

pptEaoafbGlsaraii

E

bcip

Cly

rawiws

II

I

p

wo

c

mti

F5aGs

FfGpLunNiuals

254 BERDICHEVSKY ET AL.

ose-assisted refolding of 508(Fv)-CBD (lane 7). Therotein obtained by standard refolding was of pooruality (mostly aggregated) and failed to bind antigenn an ELISA assay (not shown).

To assess the immunoreactivity of 508(Fv)-CBD pre-ared by cellulose-assisted refolding, binding of theurified protein was tested in ELISA. Similar quanti-ies, with similar serial dilutions, were applied onto theLISA plate that was coated with b-amyloid peptidend developed as described under Materials and Meth-ds. The results are shown in Fig. 3. From the ELISAnalysis it is evident that 508(Fv)-CBD that was re-olded on crystalline cellulose is active, as it binds-amyloid peptide specifically compared to theal6(Fv)-CBD (an anti-b-galactosidase scFv, see be-

ow) control. We produced a panel of three additionalcFvs related to 508(Fv) that differ by point mutationst a single VL CDR3 position, by cellulose-assistedefolding. All were produced with comparable yieldsnd had similar b-amyloid peptide binding propertiesn ELISA assays and eluted as monomer upon analyt-cal gel-filtration analysis (not shown).

xpression and Purification of the b-Galactosidase-Specific Gal6(Fv)–CBD Fusion

The scFv Gal6 specifically binds the E. coli enzyme-galactosidase with high affinity (19). Gal6(Fv) wasloned into pH6T-FEKCA3d (Fig. 1B), and the result-ng plasmid, pH6T-FEKCA3d-Gal6(Fv), was used for

TABLE 1

Purification of 508(Fv)-CBD

Fraction Total protein (mg) % purity Yield %a

nclusion bodies 38.0b 80 100. Standard refolding

Inclusion bodies 3.0 80 100Refoldedc 0.7 80 23Purified 0.2 50 3

I. Cellulose-assisted—crystalline cellulose

Inclusion bodies 35.0 80 100Refoldedd 28.0Purified 17.0 90 55

a The yield is calculated by the quantity of purified scFv–CBD as aercentage of the scFv–CBD amount in the inclusion bodies.

b Amount of urea-soluble inclusion bodies recovered from 5.6 g ofet bacterial paste from which 170 mg of wet inclusion bodies wasbtained.

c Refolded—before purification of the standard refolded protein onellulose.

d Refolded—the amount of protein immobilized on the celluloseatrix, calculated by subtracting the unbound protein and the pro-

ein in the washes from the quantity of inclusion body protein usedn the cellulose-assisted refolding.

rotein expression. Upon IPTG induction, Gal6(Fv)- o

BD accumulated as insoluble inclusion bodies (Fig. 4,ane 3). Five hundred milliliters of induced cultureielded 5 g of wet cell paste.The inclusion bodies were recovered from the bacte-

ial paste and solubilized as described under Materialsnd Methods. A total of 1.3 g of wet inclusion bodiesas recovered. The urea-soluble reduced inclusion bod-

es were 10 ml at 14 mg/ml (protein concentration),hich was approximately 80% pure (Fig. 4, lane 3. The

olubilized inclusion bodies were loaded onto a Ni-NTA

IG. 3. Analysis of antigen (b-amyloid peptide) binding by purified08(Fv)-CBD in an ELISA assay. Black bars, crystalline cellulose-ssisted refolded protein. Open bars, cellulose-assisted refoldedal6(Fv)-CBD used as a negative control. The error bars mark the

tandard deviation of the data.

IG. 4. Purification of Gal6(Fv)-CBD. Lane 1, total cell extractrom uninduced BL21(DE3) cells carrying plasmid pH6T-FEKCA3d-al6(Fv). Lane 2, total cell extract from BL21(DE3) cells carryinglasmid pH6T-FEKCA3d-Gal6(Fv) induced for 3 h with 1 mM IPTG.ane 3, washed, solubilized, and reduced inclusion bodies that weresed in Ni-NTA purification. Lane 4, Inclusion body protein that didot bind to the Ni-NTA resin. Lane 5, proteins released from thei-NTA column by 8 M urea/50 mM Tris (HCl), pH 8.0/20 mM

midazole. Lane 6, purified inclusion bodies recovered from the col-mn by EDTA. Lane 7, purified and neutralized crystalline cellulose-ssisted refolded Gal6(Fv)-CBD. 5–10 mg protein was loaded in eachane of a 14% SDS–polyacrylamide gel. Proteins were visualized bytaining with Coomassie brilliant blue. The arrow marks the position

f the scFv–CBD fusion protein.

cebt

(loictPirccbffm

moCtFmAGsflgst

aTat

pdEtvTaCaftseadftassuppelfi

GC

II

I

I

i

cw

ce

255MATRIX-ASSISTED REFOLDING OF CBD FUSION PROTEIN

olumn, washed, and eluted, which resulted in thenrichment of the scFv–CBD to .90% of the inclusionody protein (Fig. 4, lane 6). The inclusion bodies werehen reduced and refolded.

One milliliter of solubilized reduced inclusion bodies3.3 mg) was refolded by standard refolding. Two mil-iliters (6.6 mg) was used in cellulose-assisted refoldingn either crystalline or amorphous cellulose. As shownn Table 2, for Gal6(Fv)-CBD the yield obtained byellulose-assisted refolding was about threefold morehan that obtained by standard refolding. The SDS–AGE analysis of the purification procedure is shown

n Fig. 4. Purification to .95% was obtained by allefolding methods (Fig. 4, lane 7; only the crystallineellulose-assisted refolded protein is shown). In thisase, the scFv–CBD fusion protein was almost pureefore it was refolded, so in contrast to 508(Fv)-CBDor which cellulose-assisted refolding resulted also inurther purification of the protein, here it servedostly to increase the refolding yield.To test whether cellulose-assisted refolding yieldsonomeric protein, rather then aggregates, an aliquot

f the crystalline-cellulose-assisted refolded Gal6(Fv)-BD was subjected to analytical gel-filtration chroma-

ography on a 25 ml Superose 12 column. As shown inig. 5A, the protein eluted from the column as a singleajor peak as expected for the scFv–CBD monomer.nalysis of the amorphous-cellulose-assisted refoldedal6(Fv)-CBD by analytical gel filtration (Fig. 5B)

howed that while most (;70%) of the protein elutedrom the column as a monomer, about 30% eluted asarger molecular forms, indicating the presence of ag-regated protein in the preparation. Analysis of thetandard refolded Gal6(Fv)-CBD by analytical gel fil-

TAB

Purification o

Fraction Total protein

nclusion bodiesb 16.5. Standard refolding

Inclusion bodies 3.3Refoldedc 1.0Purified 0.52

I. Cellulose-assisted—crystalline celluloseInclusion bodies 6.6Purified 4.5

I. Cellulose-assisted—amorphous celluloseInclusion bodies 6.6Purified 4.4

a The yield is calculated by the quantity of purified scFv–CBD as anclusion bodies were purified on Ni-NTA agarose before refolding.

b 15 mg of Ni-NTA-purified urea-soluble inclusion bodies recovereolumn. In this experiment 5 g of wet bacterial paste yielded 1.3 g of was recovered. Of these, a 20-mg aliquot was used in the subsequenc Refolded—before purification of the standard refolded protein on

ration (Fig. 5C) showed that most of the protein was i

ggregated, while less than 40% eluted as a monomer.hus, the actual quantitative advantage of cellulose-ssisted refolding of Gal6(Fv)-CBD is greater than thehreefold difference shown in Table 2.

To compare the immunoreactivity of Gal6(Fv)-CBDrepared by different refolding schemes, b-galactosi-ase binding by the purified proteins was tested inLISA. Similar quantities, with similar serial dilu-

ions were applied onto the ELISA plate that was de-eloped as described under Materials and Methods.he results are shown in Fig. 6, in which a cellulose-ssisted refolded anti-fluorescein antibody, 4-4-20(Fv)-BD, served as a negative control. From the ELISAnalysis it is evident that Gal6(Fv)-CBD that was re-olded on crystalline cellulose is the most active, withhe protein refolded on amorphous cellulose beingomewhat less active and the standard refolded proteinven less active. The reduced immunoreactivities of themorphous cellulose-assisted refolded and of the stan-ard refolded scFv-CBD are in agreement with theraction of monomeric protein observed in the gel-fil-ration analysis (Fig. 5). This indicates that cellulose-ssisted refolding using a crystalline cellulose matrixhould be preferred. The concept behind matrix-as-isted refolding in general is that by immobilizing thenfolded protein before refolding is initiated, loss ofrotein due to aggregation is minimized because therotein molecules are prevented from interacting withach other. The larger particle size of crystalline cellu-ose should perform better in that capacity than thene fibers comprising amorphous cellulose.To demonstrate the possibility of recovering purifiedal6(Fv) by proteolytic digestion, 20 mg of Gal6(Fv)-BD was immobilized onto crystalline cellulose. The

2

al6(Fv)-CBD

g) % purity Yield %a Aggregation

90 100

90 10090 3095 17 111

90 10095 72 2

90 100 195 70

rcentage of the scFv–CBD amount in the inclusion bodies. Here the

rom 20 mg of urea-soluble inclusion bodies that was loaded on theinclusion bodies, from which 140 mg of urea-soluble inclusion bodiesteps.llulose.

LE

f G

(m

pe

d fett s

mmobilized fusion protein was treated with 1 unit of

eaitiboectsa

D

p

misShfstTtfpmtl

f2splcclglmmoioFtdrNGsitTfIftr

srchpctf

FGtraa

256 BERDICHEVSKY ET AL.

nterokinase for 8 h and analyzed by immunoblottings described under Materials and Methods. As shownn Fig. 6, the undigested fusion protein was detected byhe anti-FLAG monoclonal antibody M2 (lane 1), whilen the digested sample, only the faster-migrating scFvand could be visualized. This indicates that digestionf the Gal6(Fv)-CBD fusion protein at the engineerednterokinase site, while immobilized on cellulose,ould liberate the scFv completely. As shown in lane 3,he scFv can be separated from CBD by digestion inolution. This, however would make an additional sep-ration step to remove CBD necessary.

ISCUSSION

In this study we have shown that recombinant fusion

IG. 5. Analytical gel-filtration chromatography of purifiedal6(Fv)-CBD on a Superose 12 HR 10/30 column. Aliquots of crys-

alline cellulose-assisted refolded (A), amorphous cellulose-assistedefolded (B), and standard refolded (C) were loaded onto the columnnd eluted with TBS. Absorbance was monitored at 280 nm. Therrow marks the elution position of the scFv–CBD monomer.

roteins comprising a single-chain Fv linked to C. ther- f

ocellum CBD can by refolded from E. coli-producednclusion bodies in high yield. The refolding protocol isimple and based on cheap available materials.cale-up should be feasible with little difficulty, and weave prepared batches of .50 mg purified proteins

rom 1-liter shake-flask cultures of several differentcFvs. In all cases, the yields were at least 3-fold betterhan those obtained by standard refolding protocols.he specific binding activities were similar to those ofhe same scFv–CBD fusions that have been isolatedrom bacterial periplasmic fractions of cells carryinghage-display vectors (not shown). In all cases, 10-foldore scFv–CBD could be isolated from a similar cul-

ure volume by using intracellular expression and cel-ulose-assisted refolding compared to secretion.

Several methods for the isolation of refolded scFvs orusion proteins thereof have been described (2,20,22–5). Our refolding method is based on the reagentystem described in (2) with some modifications. Inarticular, and in striking difference from the pub-ished protocols, we can refold the protein at a highoncentration. Most of the published refolding proto-ols are based on diluting the denatured protein to veryow concentration (about 100 mg/ml), to minimize ag-regation. Since aggregation is prevented by immobi-izing the denatured protein prior to denaturant re-

oval, large quantities of protein can be handled usingodest reagent volumes. This property is shared by

ther refolding methods relying on denatured proteinmmobilization (6–8). However, these methods (as ourwn) may not be suitable for any protein of interest.or example, eight different scFvs we had isolated inhe lab do not bind ion-exchange resins, so the methodescribed in (6) is not applicable for their production byefolding. Moreover, when we used the protocol fori-NTA-assisted refolding (7) for the isolation ofal6(Fv)-CBD from pH6T-FEKCA3d, or from a con-

truct in which Gal6(Fv) had a C-terminal His tag, thesolation yield was low. We could recover 20% or less ofhe material we loaded onto the resin in an active form.his may result from the proximity of the tag to the

olding protein which may reduce the folding efficiency.n our case, the N- and C-termini of CBD are distalrom the cellulose-binding planar strip (17), so a pro-ein linked to either terminus may be more free toefold.In initial attempts to recover several scFv–CBD fu-

ions by cellulose-assisted refolding, the proteins weecovered were not of sufficient purity. Apparently, therystalline cellulose we used (Sigmacell 20) is not aighly selective matrix and binds other contaminatingroteins in addition to the CBD fusion. Other availableellulose matrices for the isolation of CBD fusion pro-eins are commercially available (14), some may per-orm better in the removal of contaminating proteins

rom the inclusion bodies. In our case, we could obtain

rtCCvdftOH(trh(wClslfldCetpaSc

Fac ative control. The error bars mark the standard deviation of the data.

FcmiCCdHsl

257MATRIX-ASSISTED REFOLDING OF CBD FUSION PROTEIN

efolded protein of satisfactory homogeneity only whenhe inclusion bodies were highly enriched in the scFv–BD fusion protein (as was the case with 508(Ser)(Fv)-BD; Fig. 2). Since there is a large batch-to-batchariability in the quality of the inclusion bodies (mostlyependent on the induction level, which varies for dif-erent scFvs), additional purification steps are requiredo recover a highly homogeneous protein preparation.ur solution for that problem was the addition of theis tag linked to the N-terminus of the fusion protein

Fig. 1B). Thus it was possible to purify the protein inhe denatured state on a Ni-NTA resin before it wasefolded. Metal-chelate affinity chromatography isighly efficient as a single-step purification method26,27), and in our case little additional purificationas necessary after it was applied to enrich the scFv–BD fraction in the inclusion bodies. When the scFv is

iberated from CBD by specific proteolytic digestion (ashown in Fig. 7), the quality of the inclusion bodies is ofesser concern, as the contaminants are not liberatedrom the cellulose matrix under these conditions. On aarger scale, the protease will have to be removed afterigestion is completed. The possibility of liberating theBD fusion protein from the matrix is based on ourxperience with recovering the CBD-containing C.hermocellum cellulosome by affinity digestion (18). Aroblem exists with the absence of commercially avail-ble purified cellulases that may serve this purpose.ince we believe that our method is of general appli-

IG. 6. Analysis of antigen (b-galactosidase) binding by purifiedssisted refolded protein. Open bars, amorphous cellulose-assistedrystalline cellulose-assisted refolded 4-4-20(Fv)-CBD used as a neg

Gal6(Fv)-CBD in an ELISA assay. Black bars, crystalline cellulose-refolded protein. Striped bars, standard refolded protein. Dotted bars,

ability for the recovery of proteins other than single- i

IG. 7. Immunoblot analysis of Gal6(Fv) liberated by digestion ofellulose-immobilized Gal6(Fv)-CBD. Lane 1, undigested protein im-obilized on cellulose, boiled in Laemmli sample buffer before load-

ng the SDS–polyacrylamide gel. Lane 2, supernatant of Gal6(Fv)-BD that had been digested while immobilized. Lane 3, Gal6(Fv)-BD digested in solution. The scFv–CBD and free scFv wereetected with an anti-FLAG monoclonal antibody followed withRP-conjugated rabbit anti-mouse antibodies and an ECL HRP

ubstrate. The upper arrow indicates the undigested scFv-CBD. Theower arrow indicates the free Gal6(Fv). The numbers on the left

ndicate the size of molecular weight markers in kDa.

cdp

lsCptmi(pgmawsbbtc

trsettgtsdpytmtatcttCfeawcthcSoct

alcbCscfen

R

1

1

1

1

1

258 BERDICHEVSKY ET AL.

hain antibodies, downstream purification steps can beesigned based on the properties of the specific fusionartner.Although our cellulose-assisted refolding method al-

ows for refolding at high protein concentration, carehould be taken not to overload the matrix with theBD fusion protein. Under such conditions, unfoldedrotein monomers may interact to form multimers onhe matrix, which will result in reduced active mono-er recovery after refolding. An indication of such

ncidence is our experience with amorphous celluloseFig. 5), of which we could use less to immobilize therotein, but the recovered material contained aggre-ated forms. The binding capacity of different celluloseatrices may vary, depending mostly on particle size

nd particle surface area. As a general rule, one shouldork well below the binding capacity of the matrix. For

imilar reasons, the binding of solubilized inclusionodies to the cellulose matrix should be performed inatch, rather than on a column, to avoid “crowding” ofhe monomers that is expected to occur at the top of aolumn.

Finally, the elution conditions we apply to recoverhe refolded proteins from the cellulose matrix areather harsh (100 mM NaOH, pH 13.0). Although thecFvs we have produced by this method tolerated thelution conditions, other fusion partners may be lessolerant of such conditions. Moreover, our refolded pro-eins may have been unfolded by the high pH andradually refolded again during dialysis. Two observa-ions make this unlikely in our case: first, severalcFv–CBD fusions were efficiently recovered by imme-iate neutralization with Tris buffer following alkalineH elution from cellulose rather than by gradual dial-sis (not shown). Second, during dialysis the concen-ration of the eluted protein was quite high (about 0.25g/ml). Dialysis of unfolded scFv–CBD at that concen-

ration results in loss of .80% of the protein due toggregation. However, high pH elution may result inhe unfolding of proteins. This difficulty should be over-ome by reengineering the elution properties of CBD sohat it may be recovered under milder conditions. Forhat purpose, we used rational design based on theBD 3D structure (17) and the phage-display system

or CBD (19) to identify CBD mutants having alteredlution properties. One such mutant, having Asp56nd Trp118 both mutated to alanine, binds to celluloseith the same high capacity. However, its dissociation

onstant is increased about 20-fold. We incorporatedhis mutant CBD into our expression vectors describederein and found that the scFv–CBD fusion proteinan be eluted by 0.01 M NaOH or 3% triethylamine.uch conditions are milder and acceptable for numer-us protocols for the elution of proteins from affinityolumns. In addition, scFv–CBD fusions with this mu-

ant CBD are more efficiently recovered from cellulose, 1

s over 95% of the cellulose-bound protein can be re-eased under these conditions, while the recovery effi-iency with wild-type CBD does not exceed 80% ofound protein. We are now in pursuit of additionalBD mutants with varying elution profiles (pH, ionictrength, etc.), which will not only broaden the appli-ability of our system for the recovery of a multitude ofusion partners, but will also serve as tools for efficientxploitation of cellulose binding domains in other ave-ues of biotechnology (10).

EFERENCES

1. Lilie, H., Schwarz, E., and Rudolph, R. (1998) Advances in re-folding of proteins produced in E. coli. Curr. Opin. Biotechnol. 9,497–501.

2. Buchner, J., Pastan, I., and Brinkmann, U. (1992) A method forincreasing the yield of properly folded recombinant fusion pro-teins: Single-chain immunotoxins from renaturation of bacterialinclusion bodies. Anal. Biochem. 205, 263–270.

3. Kiefhaber, T., Rudolph, R., Kohler, H. H., and Buchner, J. (1991)Protein aggregation in vitro and in vivo: A quantitative model ofthe kinetic competition between folding and aggregation. Bio-technology 9, 825–829.

4. Guise, A. D., West, S. M., and Chaudhuri, J. B. (1996) Proteinfolding in vivo and renaturation of recombinant proteins frominclusion bodies. Mol. Biotechnol. 6, 53–64.

5. Fenton, W. A., and Horwich, A. L. (1997) GroEL-mediated pro-tein folding. Protein Sci. 6, 743–760.

6. Creighton, T. E. (1985) Folding of proteins adsorbed reversibly toion-exchange resins, in “UCLA Symposia on Molecular and Cel-lular Biology New Series Vol. 39” (Oxender, D. L., Ed.), pp.249–258.

7. Glansbeek, H. L., van Beuningen, H. M., Vitters, E. L., van derKraan, P. M., and van den Berg, W. B. (1998) Expression ofrecombinant human soluble type II transforming growth factor-beta receptor in Pichia pastoris and Escherichia coli: Two pow-erful systems to express a potent inhibitor of transforminggrowth factor-beta. Protein Expression Purif. 12, 201–207.

8. Stempfer, G., Holl-Neugebauer, B., and Rudolph, R. (1996) Im-proved refolding of an immobilized fusion protein. Nat. Biotech-nol. 14, 329–334.

9. Garel, J.-R. (1992) Large multidomain and multisubunit pro-teins, in “ Protein Folding” (Creighton, T. E., Ed.), pp. 405–454,Freeman, New York.

0. Bayer, E. A., Morag, E., and Lamed, R. (1994) The cellulo-some—A treasure-trove for biotechnology. Trends Biotechnol.12, 379–386.

1. Ong, E., Greenwood, J. M., Gilkes, N. R., Kilburn, D. G., Miller,R. C. J., and Warren, R. A. J. (1989) The cellulose-bindingdomains of cellulases: Tools for biotechnology. Trends Biotech-nol. 7, 239–243.

2. Assouline, Z., Shen, H., Kilburn, D. G., and Warren, R. A. J.(1993) Production and properties of a Factor-X-cellulose bindingdomain fusion protein. Protein Eng. 6, 787–792.

3. Ramirez, C., Fung, J., Miller, R. C. J., Warren, R. A. J., andKilburn, D. G. (1993) A bifunctional affinity linker to coupleantibodies to cellulose. Bio/Technology 11, 1570–1573.

4. Shoseyov, O., and Karmely, Y. (1995) Kits and methods of de-tection using cellulose binding domain fusion proteins. U.S.patent application No. 460,458.

5. Poole, D. M., Morag, E., Lamed, R., Bayer, E. A., Hazlewood,

1

1

1

1

2

2

2

2

2

2

2

2

259MATRIX-ASSISTED REFOLDING OF CBD FUSION PROTEIN

G. P., and Gilbert, H. J. (1992) Identification of the cellulose-binding domain of the cellulosome subunit S1 from Clostridiumthermocellum YS. FEMS Microbiol. Lett. 78, 181–186.

6. Morag, E., Lapidot, A., Govorko, D., Lamed, R., Wilchek, M.,Bayer, E. A., and Shoham, Y. (1995) Expression, purification,and characterization of the cellulose-binding domain of the scaf-foldin subunit from the cellulosome of Clostridium thermocel-lum. Appl. Environ. Microbiol. 61, 1980–1986.

7. Tormo, J., Lamed, R., Chirino, A. J., Morag, E., Bayer, E. A.,Shoham, Y., and Steitz, T. A. (1996) Crystal structure of abacterial family-III cellulose-binding domain: A generalmechanism for attachment to cellulose. EMBO J. 15, 5739 –5751.

8. Morag, E., Bayer, E. A., and Lamed, R. (1992) Affinity digestionfor the near-total recovery of purified cellulosome from Clostrid-ium thermocellum. Enzyme Microb. Technol. 14, 289–292.

9. Berdichevsky, Y., Ben-Zeev, E., Lamed, R., and Benhar, I. (1998)Phage display of a cellulose binding domain from Clostridiumthermocellum and its application as a tool for antibody engineer-ing. J. Immunol. Methods, in press.

0. Tsumoto, K., Shinoki, K., Kondo, H., Uchikawa, M., Juji, T., andKumagai, I. (1998) Highly efficient recovery of functional single-chain Fv fragments from inclusion bodies overexpressed in Esch-erichia coli by controlled introduction of oxidizing reagent—Application to a human single-chain Fv fragment. J. Immunol.

Methods 219, 119–129.

1. Brizzard, B. L., Chubet, R. G., and Vizard, D. L. (1994) Immu-noaffinity purification of FLAG epitope-tagged bacterial alkalinephosphatase using a novel monoclonal antibody and peptideelution. Biotechniques 16, 730–735.

2. Huston, J. S., Levinson, D., Mudgett-Hunter, M., Tai, M. S.,Novotny, J., Margolies, M. N., Ridge, R. J., Bruccoleri, R. E.,Haber, E., Crea, R., and Opperman, H. (1988) Protein engineer-ing of antibody binding sites: Recovery of specific activity in ananti-digoxin single-chain Fv analogue produced in Escherichiacoli. Proc. Natl. Acad. Sci. USA 85, 5879–5883.

3. Huston, J. S., Tai, M. S., McCartney, J., Keck, P., and Opper-mann, H. (1993) Antigen recognition and targeted delivery bythe single-chain Fv. Cell Biophys. 22, 189–224.

4. Kurucz, I., Titus, J. A., Jost, C. R., and Segal, D. M. (1995)Correct disulfide pairing and efficient refolding of detergent-solubilized single-chain Fv proteins from bacterial inclusion bod-ies. Mol. Immunol. 32, 1443–1452.

5. Skerra, A. (1993) Bacterial expression of immunoglobulin frag-ments. Curr. Opin. Immunol. 5, 256–262.

6. Hochuli, E. (1990) Purification of recombinant proteins withmetal chelate adsorbent. Genet. Eng. 12, 87–98.

7. Crowe, J., Dobeli, H., Gentz, R., Hochuli, E., Stuber, D., andHenco, K. (1994) 63His–Ni-NTA chromatography as a superiortechnique in recombinant protein expression/purification. Meth-

ods Mol. Biol. 31, 371–387.