Protein Expression and Purification 23, 398–410 (2001)doi:10.1006/prep.2001.1525, available online at http://www.idealibrary.com on

Expression and Purification of the Extracellular Domain ofHuman Myelin Protein Zero

Jeremy P. Bond, Raul A. Saavedra,*,1 and Daniel A. KirschnerDepartment of Biology, Boston College, Chestnut Hill, Massachusetts 02467; and*Department of Neurobiology & Anatomy, MCP-Hahnemann University, Philadelphia, Pennsylvania 19129

ous protocols, including bacterial growth to ,15 ODusing orbital shakers and the use of diafiltration,

Received April 5, 2001, and in revised form July 27, 2001

DEDICATED TO THE MEMORY OF DR. RANDALL SWARTZ

Myelin protein zero (P0), an adhesion protein of theimmunoglobulin superfamily, is the major protein ofperipheral nervous system myelin in higher verte-brates. Protein zero is required for the formation andmaintenance of myelin structure in the internode,likely through homophilic interactions at both theextracellular and the intracellular domains. Mutationsand deletions in the P0 gene correlate with hereditaryperipheral neuropathies of varying severity. Compari-sons between the human and rat isoforms, whose three-dimensional structure has been determined by X-raycrystallography, suggest that these disease-associatedgenetic alterations lead to structural changes in theprotein that alter P0–P0 interactions and hence affectmyelin functionality. Knowing the crystal structures

of native and altered human P0 isoforms could help toelucidate the structural changes in myelin membranepacking that underlie the altered functionality. Alter-ations of P0 extracellular domain (P0-ED) are of addi-tional interest as previous X-ray diffraction studies on

1 Current address: National Institute of Neurological Disorders &Stroke, National Institutes of Health, Neuroscience Center, Suite3208, 6001 Executive Blvd., Rockville, MD.

2 Abbreviations used: AMP, ampicillin; AR, amylose resin; BSA,bovine serum albumin; DEAE, diethylaminoethyl; DTT, dithiothrei-tol; ETB, “Enhanced” Terrific Broth; hP0, human protein zero; hP0-ED, extracellular domain of human protein zero; HPLC, high-perfor-mance liquid chromatography; IPTG, isopropyl b -D-thiogalactoside;K-Phos, potassium phosphate; LB, Luria broth; LB/AMP, Luria brothcontaining 100 mg/ml ampicillin; MBP, maltose binding protein; OD,optical density at 600 nm; OSS, osmotic shock solution; PCR, polymer-ase chain reaction; P0, protein zero; PNS, peripheral nervous system;PVDF, polyvinylidene fluoride; rP0, rat protein zero; SDS–PAGE,sodium dodecyl sulfate–polyacrylamide gel electrophoresis; X-gal,5-bromo-4-chloro-3-indolyl b -D-galactoside.

398

myelin membrane packing suggest that P0-ED mole-cules can assume distinct adhesive arrangements.Here, we describe an improved method to express andpurify human P0-ED (hP0-ED) suitable for crystallo-graphic analysis. A fusion protein consisting of maltosebinding protein fused to hP0-ED was secreted to theperiplasm of Escherichia coli to allow an appropriatefolding pathway. The fusion protein was extracted viaosmotic shock and purified by affinity chromatogra-phy. Factor Xa was used to cleave the fusion protein,and a combination of affinity and ion-exchange chro-matography was used to further purify hP0-ED. Wedocument several significant improvements to previ-

which result in yields of ,150 mg highly pure proteinper liter of medium. q 2001 Academic Press

Protein zero (P0-glycoprotein, or P0) is the major inte-gral protein in peripheral nervous system (PNS) myelinof higher vertebrates (1, 2), where it plays a crucialrole in myelin formation and maintenance of membraneapposition and packing in the internodal region of themyelin sheath (3, 4). P0 knockout mice show severehypomyelination and myelin degeneration (5), and sev-eral peripheral neuropathies of varying severity corre-late with mutations and deletions in the human P0(hP0) gene (6).

It has been proposed that membrane packing at the

extracellular apposition in PNS myelin is mediatedthrough homotypic interactions of apposed P0 mole-cules (1, 7, 8). A detailed understanding of the adhesiveprotein–protein interactions between P0 extracellular

1046-5928/01 $35.00Copyright q 2001 by Academic Press

All rights of reproduction in any form reserved.

O

EXPRESSION AND PURIFICATI

domains (P0-ED) was elucidated from the crystal struc-ture of rat P0-ED (9), which reveals a tetrameric assem-bly. Full sequence P0 shows similar assembly in amembrane mimetic environment that more closely ap-proximates the lipid environment in myelin membranes(10). The crystal structure reveals three different molec-ular interfaces between individual P0-ED molecules.The first interaction occurs between molecules havingthe same orientation and results in the tetrameric ar-rangement of “head-to-tail” P0-ED molecules formingtogether a doughnut-like assembly. The second interac-tion (twofold) is a lateral association between antiparal-lel molecules, oriented as if they were extending fromapposing, close-packed membranes. The third interac-tion is a head-to-head association between antiparallelmolecules, oriented as if anchored in apposed mem-branes that are swollen apart from one another.

The human P0 gene maps to chromosome 1q22–q23(11), and allelic point mutations are associated withCharcot–Marie–Tooth type 1B (12) and Dejerine–Sottas syndrome (13). The human P0 sequence is 97%identical to that of rat P0, with substitutions Tyr10 →His, Gln16 → Arg, and Ser77 → Arg in the extracellulardomain (11). The crystal structure shows Tyr10 locatednear the carboxyl terminus, close to the base on theinner side of the tetrameric, doughnut-like assembly.The hydroxyl group of Tyr10 is hydrogen-bonded to oneoxygen of the Glu8 side chain, while the other oxygen ofthe Glu8 side chain is hydrogen-bonded to the hydroxylgroup of Tyr116. Further, the aromatic ring of Tyr10is in close proximity to the aromatic rings of Tyr116and Phe118. The Tyr10 → His substitution would beexpected to prevent the formation of the hydrophobicpocket and to alter local hydrogen bonding. The Gln16residue is located in the middle of the inside face of thetetrameric assembly, and the side chain is hydrogen-bonded to the His86 side chain near the head-to-tailinterfaces of the tetramer. The Gln16 → Arg substitu-tion would introduce an additional positive charge andlikely alter local side chain interactions, which mayaffect tetrameric packing. The Ser77 residue is locatedat the top of the P0-ED molecule in the region of thehead-to-head interface. The Ser77 side chain is hydro-gen-bonded to the Asp75 side chain. The substitutionSer77 → Arg would introduce a new positive chargeand likely disrupt this interaction, affecting the head-to-head association of P0-ED molecules. The introduc-tion of additional positive charges by these three substi-tutions could result in significant tertiary and quater-nary structural differences between human P0-ED andrat P0-ED.

In its native form, P0 is glycosylated at Asn93 withan N-linked nonasaccharide (14). Molecular modelingsuggests that the carbohydrate helps maintain the ori-entation of P0-ED at the membrane surface (8), andthe oligosaccharide plays a major role in intercellular

N OF MYELIN PROTEIN ZERO 399

contacts between cells expressing P0 (15, 16). The citedcrystallographic study (9) did not address the role ofthe Asn93-linked carbohydrate and thus limits compar-isons between the molecular contacts evident from thecrystals and in vivo P0–P0 interactions (17).

Previous studies have analyzed alterations of the hP0gene that cause hereditary demyelinating peripheralneuropathies and found that the severity of differentphenotypes correlates to the alterations in primary se-quence (6, 18). It is hypothesized that these alterationswould affect tertiary and quaternary protein structure,resulting in modified homophilic interactions and modi-fied membrane packing in internodal myelin. Loss ofprotein adhesive functionality might explain the de-creased conduction velocities and hypomyelination ob-served in patients with peripheral neuropathies. Thecrystal structure of native and abnormal hP0-ED iso-forms could provide further insight into the molecularmechanisms of these peripheral neuropathies and indi-cate possible therapeutic approaches.

In the research reported here, the hP0-ED gene wasinserted into the pMal-p2 vector (New England Biolabs(NEB), Beverly, MA), resulting in a fusion protein withmaltose binding protein (MBP) and hP0-ED separatedby a Factor Xa cleavage site. The product of pMal-p2expression is secreted to the Escherichia coli periplasm,which is advantageous for proteins containing disulfidebonds or extracellular domains of transmembrane pro-teins, as it provides a folding pathway that more closelymimics the native environment (19, 20). Moreover, MBPdoes not contain any cysteine residues that could inter-fere with the usual disulfide bond formation of P0-ED(21). The E. coli cytoplasm is a reducing environmentthat is unlikely to allow correct disulfide bond forma-tion. The periplasm, by contrast, contains proteins thatdo catalyze disulfide bond formation (22, 23); however,this compartment also harbors higher protease activity.Thus, secretion usually results in an 8- to 16-fold de-crease in protein expression compared with cytoplasmicexpression (pMal Instruction Manual, NEB).

Owing to the organization of the pMal-p2 polylinker,the resulting hP0-ED product contained a Val2 → Sersubstitution. This substitution was unavoidable be-cause the recognition sequence of a restriction enzymeoverlaps the Factor Xa cleavage site. Although the pre-cise effect of this substitution has not yet been deter-mined, it is unlikely to dramatically alter protein struc-ture. The rP0-ED sequence contains Ile1 → Sersubstitution, and the crystal structure shows Val2 lo-cated near the head-to-tail interface within the tetra-

mer. The Val2 side chain is exposed to this protein sur-face and proximal to the side chains of Tyr48 and Asn878from an adjacent two-fold related P0-ED molecule, butat distances greater than their van der Waals radii.The Val3 residue, however, is probably more important

,

400 BOND, SAAVEDRA

for stabilizing local structure because it forms an inter-nal hydrophobic pocket with Val100 and Phe23. Whilea Val2 → Ser substitution introduces a substantiallyless hydrophobic sidechain, both amino acids show sim-ilar b -strand propensity (24). The crystal structure ofrP0-ED shows a short b -strand composed of residues2–4. Whether the folding of P0-ED is more dependenton maintaining this secondary structure or the hydro-phobicity of the residue at the two position is not known.

Screening of conditions for forming crystals of hP0-

ED that are suitable for X-ray crystallographic analysis requires large quantities (,10 mg) of highly purifiedprotein. Here we describe a protocol that allows large-scale production of hP0-ED and a purification proce-dure that results in .99% purity.

MATERIALS AND METHODS

Materials

DH5a competent bacteria were obtained from Gibco-BRL (Gaithersburg, MD). The plasmid, restriction en-zymes, and ligase were from New England Biolabs. TaqDNA polymerase was obtained from Perkin–ElmerCetus (Foster City, CA). Reagents were from Sigmaunless otherwise specified. A clone of the human P0cDNA was generously provided by Dr. KiyoshiHayasaka (Akita University School of Medicine, AkitaCity, Japan). SDS–PAGE was conducted with ReadyGel precast 10–20% linear gradient gels (Bio-Rad)and stained with GelCode Blue Stain Reagent orSilverSNAP Stain (Pierce).

PCR of Human P0 cDNA

Primers were designed based on the published hu-man P0 cDNA sequence (11). The 58-end primer (58-TAGA AGG ATT TCA GTT TAC ACC GAC AGG GAGGTC-38) was designed with an upstream XmnI restric-tion site (underlined), whereas the 38-end primer (58-TA GAA TTC CTA CCT AGT TGG CAC TTT TTC AAAGAC-38) was designed with a downstream EcoRI re-striction site (double underlined). The TA sequence wasadded to the 58-end of the primers to facilitate DNArestriction and subcloning of the fragment. PCR wascarried out as directed by Perkin–Elmer Cetus, Inc.,using a DNA thermal cycler of their making. The pa-rameters for the 30 cycles were denaturation, 1 min at948C; hybridization, 1 min at 558C; and extension, 1min at 728C. The PCR product was electrophoresed ona 1.2% agarose gel, resulting in a single band of 409bp, as expected for hP0-ED. The band was excised from

the gel and placed in dialysis membrane, and the PCRproduct was electroeluted from the gel fragment. Next,the PCR product was digested with EcoRI in One-Phor-All buffer (Parmacia Biotech, Piscataway, NJ). The re-action was incubated at 378C for 2 h, and the enzyme

AND KIRSCHNER

was heat-inactivated at 758C for 10 min. The PCR prod-uct was ethanol-precipitated, resuspended in dH2O,and digested with XmnI in NEBuffer 2 (NEB) in thepresence of BSA. Incubation and inactivation condi-tions were the same as in the previous digestion.

Subcloning Human P0-ED Gene into pMal-p2 Vector

The pMal-p2 vector was also digested with EcoRI andXmnI and recombined with the digested PCR productin the presence of ATP and DTT. T4 DNA ligase andappropriate buffer were added to the mixture. Ligationwas conducted overnight at 168C and inactivated byincubating at 658C for 10 min.

The ligation reaction was then mixed with competentDH5a cells and incubated for 30 min on ice, then 428Cfor 1 min, and finally returned to ice for 30 min. Thecells were added to 450 ml Luria broth (LB) and placedin an orbital incubator at 378C for 1 h. Next, the cellswere spread on LB agar plates containing ampicillin(AMP) and grown overnight at 378C. Replica plates con-taining IPTG and X-gal were incubated at 378C for 12 hand then placed at 48C for 24 h to allow the chromogenicreaction to fully develop.

Colonies from the master plates corresponding towhite colonies on the replica plates were selected andgrown in LB/AMP at 378C. The cells were induced at308C with IPTG (100 mg/ml), and grown for an addi-tional 2 h. Induced cells were electrophoresed on precast10–20% SDS–PAGE gels (Bio-Rad). Positive coloniesthat showed an extra band at ,57 kDa, the expectedmolecular weight of the MBP/hP0-ED fusion protein(43 kDa, MBP; 14 kDa, hP0-ED), were stored in 85%glycerol and maintained at 2708C.

Microsequencing of Fusion Protein

Microsequencing was performed to confirm theproper expression of the fusion protein MBP/hP0-ED.Positive cells were grown in LB/AMP at 378C to ,0.5OD. The cells were induced with IPTG and incubatedfor a further 2 h. The fusion protein was extracted fromcell periplasm by osmotic shock and then purified asdescribed in the pMAL Instruction Manual (CatalogNo. 800, NEB). Affinity chromatography was used topurify the MBP/hP0-ED fusion protein from the peri-plasmic fraction, and the resulting elution was concen-trated using Microcon vials (Millipore, Bedford, MA) to,1 mg/ml. Factor Xa (Catalog No. P8010L, NEB) wasused to cleave the fusion protein. The resulting digestwas electrophoresed on a 10–20% SDS–PAGE gel, andtransferred to membrane (Immobilon PVDF, Millipore,

Bedford, MA) with Bjerrum and Schafer-Nielsen trans-fer buffer (48 mM Tris–HCl at pH 9.2, 39 mM glycine,20% methanol) using a semi-dry electrophoretic trans-fer cell (Bio-Rad, Hercules, CA). The PVDF membranewas lightly stained with Coomassie blue, and the band

O

EXPRESSION AND PURIFICATI

at ,14 kDa was excised from the membrane. The blot-ted band was microsequenced using an Applied Biosys-tems Model 475A/900 protein sequenator and con-firmed to be hP0-ED, including the substituted serinefor valine at the second position (Dr. D. B. Teplow,Biopolymer Laboratory, Brigham and Women’s Hospi-tal, Harvard Medical School). This clone was used forall subsequent studies.

Large-Scale Expression of MBP/hP0-ED FusionProtein

Bacteria were grown at 378C, unless otherwise noted.Owing to the unstable nature of the vector, and thesubsequent differing levels of protein expression, colo-nies were first screened for high fusion protein expres-sion. Bacterial stabs were taken from frozen stocks,streaked onto LB/AMP plates, and incubated overnight.Individual colonies were selected from the plates andstreaked onto discrete regions of secondary LB/AMPplates. Following streaking, individual colonies wereplaced in LB/AMP solution and grown for 12 h whilethe secondary plates were incubated overnight. Thecells in solution were induced with IPTG for 6 h at 308C.The cells were harvested by centrifugation at 14,000gfor 5 min. The pellets were electrophoresed on SDS–PAGE gels to determine the relative levels of fusionprotein expression of individual colonies, which oftenshowed significant variation (see Results). Coloniesthat showed the highest level of fusion protein expres-sion were selected from the corresponding region on thesecondary plate and placed in LB/AMP solution. Thesecultures were grown to ,0.5 OD, and then 200 ml wasadded to 200 ml LB/AMP containing Antifoam A (Cat.No. 6457, Sigma) in a 2-liter baffled culture flask (Nal-gene, Rochester, NY). The culture was grown until ,0.5OD, and then 100 ml was added to 900 ml “Enhanced”Terrific Broth (ETB; 20 g tryptone, 40 g yeast extract,50 g glycerol, 4 g glucose, 1 ml antifoam A to 1000 mlwith dH2O) prewarmed to 378C. The resulting 1000-mlsolution was split into 250-ml aliquots and placed infour 2-liter baffled flasks. These cultures were grownuntil ,0.5 OD, and the temperature was reduced to308C as significant proteolysis of the fusion protein re-sulted at 378C induction. One hour later IPTG wasadded to a final concentration of 0.42 mM (100 mg/ml),and during induction pH .7.0 was maintained withconstant monitoring. After ,8 h of induction the cells

reached ,15 OD and had a wet weight of ,70 g. Cellswere harvested by centrifugation and the fusion proteinextracted via osmotic shock as described in the pMALInstruction Manual (Catalog No. 800, NEB).

N OF MYELIN PROTEIN ZERO 401

Purification of hP0-ED

Step 1: Affinity chromatography. Affinity chroma-tography was conducted at 48C. To remove any re-maining bacterial contamination the osmotic shock so-lution (OSS) was sterile filtered by passing the solutionthrough a 90-mm stainless-steel filter holder (Millipore)with glass fiber prefilter (Cat. No. AP1509000, Milli-pore) and 0.45-mm hydrophilic Durapore membrane(Cat. No. HVLP09050, Millipore). The sterile filteredOSS was then concentrated using a miniplate bio-concentrator (Cat. No. 1571304; Millipore), and in 6 hthe volume was reduced 12-fold to ,500 ml. Amyloseresin (100 ml) was poured into an XK 26/70 column(Pharmacia Biotech) and washed with 10 column vol-umes of amylose resin (AR) buffer (20 mM Tris–HCl(pH 7.4), 200 mM NaCl, 1 mM EDTA, 1 mM NaN3, 1mM DTT) using low pressure liquid chromatography(Econo System, Bio-Rad). Affinity chromatography wascarried out as instructed by the supplier (pMal Instruc-tion Manual, NEB), except column flow rates were dou-bled without reduction in column performance. The fu-sion protein eluted as a single peak as determined byUV monitor at 280 nm. Fractions containing the fusionprotein were pooled and concentrated with Ultrafree-15 centrifugal filtration devices (Cat. No. UFV2 BTK10, Millipore) to ,10 mg/ml.

Step 2: Fusion protein cleavage. The concentratedeluate from the AR column was digested for 24 h withFactor Xa at room temperature. Optimal cleavage wasobtained using 10 mM Tris–HCl (pH 8.0), 500 mMNaCl, 2 mM CaCl2, 1 mM NaN3, 0.02% SDS. Dansyl-glu-gly-arg-chloromethyl ketone (2 mM final concentra-tion) was added to the digest to inhibit irreversiblyFactor Xa. The digest suspension was concentratedwith Ultrafree-15 devices and diafiltrated against ARbuffer. By removing the digestion buffer and the MBP-bound maltose, this step replaced one dialysis step.

Step 3: MBP removal via affinity chromatogra-phy. After Factor Xa treatment, the digestion suspen-sion contained predominantly MBP. This was removedby using the AR column as previously described, andthe flow-through was collected and concentrated withUltrafree-15 devices. SDS–PAGE confirmed the re-moval of MBP and any undigested fusion protein.

Step 4: Ion-exchange chromatography. Diethylami-noethyl (DEAE) Sepharose (Cat. No. DFF-100, Sigma)was poured into a 0.7 3 20 cm column and washed with10 vol of DEAE buffer (20 mM Tris–HCl (pH 8.0), 25mM NaCl). This step was performed at room tempera-ture and negligible hP0-ED degradation was observed.

The AR column flow-through fractions were concen-trated and the buffer was exchanged for DEAE bufferwith Ultrafree-15 devices. The sample was loaded ontothe DEAE column at 0.25 ml/min and washed with 10column volumes of DEAE buffer. A 25–500 mM NaCl

,

Gaussian fitting with the program PeakFit (Jandel Sci-

402 BOND, SAAVEDRA

gradient eluted the proteins in discrete peaks, and thefractions containing the hP0-ED were determined bySDS–PAGE. (We found that purification of P0 by gelfiltration was problematic, owing to the propensity ofthe protein to self-associate.)

Step 5: Final purification hP0-ED with ion exchangeHPLC. The hP0-ED fractions from the DEAE columnwere pooled and concentrated with Ultrafree-15 devicesand then dialyzed against 20 mM Tris–HCl (pH 8.0)using Slide-A-Lyzer dialysis products (Pierce) at 48Cover 24 h. An anion exchange medium (POROS 10 HQ)was packed into a PEEK 4.6 mmD/100 mmL columnwith a POROS Self Pack packing device using a BioCADSPRINT system (all components from PerSeptive Bio-systems, Framingham, MA). HPLC protocol was ac-cording to instructions (PerSeptive Biosystems), withflow rate 5 ml/min. Proteins were eluted with a 100–350mM NaCl gradient (containing 20 mM Tris–HCl, pH8.0) over 20 min. The hP0-ED peptide eluted at ,175mM NaCl, and the fraction resulted in a single ,14-kDa band when analyzed by silver staining of SDS–PAGE gels.

Protein Quantitation

Protein levels after each purification step (Table 1)were quantitated from densitometer tracings of theSDS–PAGE gels, which were digitized using a Molecu-lar Dynamics Personal SI densitometer (Sunnyvale,CA) at 50 mm resolution. The machine readout was inoptical density units, calibrated internally, and con-firmed to be linear from 0 to 3 OD using the knownoptical density of a calibration step wedge (Kodak

Photographic Step Tablet No. 2, Eastman Kodak Com-

Factor Xa digestion 2398 331 57 13.8Second amylose affinity 433 291 50 67.4DEAE column 235 224 39 95.7POROS HQ column ,153 153 27 .99.9

a Mass of hP0-ED/MBP fusion protein is shown in parentheses.

AND KIRSCHNER

pany, Rochester, NY). Protein profiles of individuallanes were obtained after subtraction of the polynomial-fit background curve. The integral areas of proteinbands were obtained from the protein profiles after

TABLE 1

Purification of hP0-ED Expressed in the Periplasm ofE. coli DH5a Cells Grown in 1000 ml “Enhanced”

Terrific Broth

Total hP0-ED hP0-EDprotein massa Recovery purity

Purification step mass (mg) (mg) (%) (%)

Osmotic shock ,7900 575 (2300) 100 7.3Amylose affinity 2608 403 (1612) 70 15.4

entific Software). Protein mass was obtained by densi-tometric comparison with prestained molecular mark-ers and known quantities of ovalbumin and lysozyme(Sigma).

RESULTS

Fusion Protein Expression

The PCR protocol resulted in a single band of ,409bp, as expected for the hP0-ED fragment (data notshown). Subsequent ligation into the pMal-p2 vectorand transformation into DH5a cells resulted in ,10%white colonies. Colonies were selected from the masterplate, grown in LB/AMP, and induced with IPTG asdescribed. Protein expression was assayed using SDS–PAGE (Fig. 1), which indicated the variable level offusion protein expression among different colonies. Thecolony with the highest level of fusion protein expres-sion was determined by visual inspection and used forlarge-scale bacterial growth. Microsequencing con-firmed that the N-terminal 12 residues were correct forhP0-ED, including the Val2 → Ser substitution.

Factors Affecting Bacterial Growth

Proteolysis of the fusion protein was minimized bymaintaining the pH .7.0 of the media (Fig. 1, lanes6–8). Various buffering conditions were tested, but evenstrong buffering required pH adjustment during growthto maintain pH .7.0 (Fig. 2). Of the various bufferstested K-Phos at pH 7.4 proved optimal, allowing highgrowth rates and cell concentrations (Fig. 2A), whileproviding the most effective long-term buffering capac-ity (Fig. 2B).

To assess the influence of ionic strength on bacterialgrowth, various media containing a range of buffer con-centrations were tested. Buffers up to 400 mM wereexamined, but at these ionic strengths growth ratesbecame erratic and prohibitively slow; sometimes nogrowth occurred. Optimal growth was obtained usingmedia containing no additional buffer, as increasingbuffer concentration decreased growth rates (Fig. 3A).

Controlling pH variation of bufferless media proveddifficult, but dramatic pH variation could be avoidedwith routine monitoring and precise NaOH addition(Fig. 3B). Comparison of OSS obtained from cells grownat different buffer conditions showed similar proteinprofiles, and the fluctuating pH of the bufferless growthmedia did not significantly affect fusion protein expres-sion or degradation (Fig. 4).

Large-scale bacterial growth was conducted usingETB, with routine pH monitoring and adjustment withNaOH. Growth rates varied, but typically cells weregrown for several hours until ,1 OD was reached andthen were induced at 308C and grown until ,15 OD.

expression levels of MBP/hP0-ED (arrow). Lane 2 shows high expression of fusion protein, while lane 3 shows slightly reduced fusion proteinexpression. Fusion protein expression is negligible in lane 4 and minor in lane 5. Lane 6 indicates an osmotic shock solution (OSS) from

h

rol

and silver staining confirmed .99.9% purity by densito-metric analysis (Fig. 5, lane 9; Table 1). The increased

the colony showing the high fusion protein expression of lane 2, witfusion protein (open arrowhead) and MBP (filled arrowhead) are thelane 6, but without maintaining pH .7, and low molecular weight p7 with further induction. Significant fusion protein degradation resu

Overall, a total of 12–14 h of growth resulted in OD13–15. At OD .10, the speed of the orbital shaker wasincreased to 400 rpm to improve media aeration. Thecells were harvested, and the periplasmic fraction wasisolated via osmotic shock.

Fusion Protein Purification

The OSS containing the fusion protein was sterilefiltered and concentrated, mixed with AR buffer (20%v/v), and passed over an amylose column. The columnwas washed with 10 column volumes of AR buffer, andmaltose was used to elute the fusion protein. The amy-lose resin bound the fusion protein and MBP with ahigh degree of specificity, resulting in a relatively cleancolumn elution (Fig. 5). Cleavage of the fusion proteinwas accomplished with Factor Xa, and the digestionsolution became opaque as the reaction proceeded. Fur-ther digestion of the fusion protein (i.e., .48 h) lead toa wispy granular precipitate, which settled upon cessa-tion of mixing. Some precipitate adhered to the wall

of the digestion vessel, forming fine long fibers. Theprecipitate could be mostly resolubilized by dialysisagainst AR buffer, addition of SDS, or raising the tem-perature to 378C. Increasing salt concentration repre-cipitated the aggregates. Analysis of the precipitate by

pH .7 maintained throughout growth and induction. MBP/hP0-EDprominent proteins. Lane 7 shows the OSS from the same colony asducts (star) become obvious. Lane 8 shows the same sample as lane

ts when pH .7 is not maintained.

SDS–PAGE showed aggregated material with a veryhigh molecular weight that slightly entered the gel (Fig.6). The precipitate protein profile was similar to thatof the supernatant, but with elevated levels of fusionprotein and hP0-ED.

After digestion of the fusion protein and removal ofmaltose from MBP, a second pass of the protein solutionover the amylose resin effectively removed MBP andany undigested fusion protein (Fig. 5, lane 7). A DEAESepharose column was used to purify hP0-ED from Fac-tor Xa and the remaining contaminants. The hP0-EDeluted at ,175 mM NaCl during a linear 25–500 mMNaCl gradient (Fig. 7). Subsequent ion-exchange HPLC(100–350 mM NaCl gradient) eluted hP0-ED (Fig. 8);

EXPRESSION AND PURIFICATION OF MYELIN PROTEIN ZERO 403

FIG. 1. Analysis of MBP/hP0-ED fusion protein expression with 10–20% SDS–PAGE. “White” colonies were selected from a master plateand grown in LB/AMP for 8 h at 378C. Cells were induced for 2 h with IPTG (100 mg/ml) and then harvested by centrifugation and boiledin SDS–PAGE loading buffer for 4 min. Lane 1, molecular weight markers. Lanes 2–5, different induced “white” colonies showing variable

purity at the cost of diminished amounts of proteinrecovered was determined by the need to have molecu-larly homogeneous molecules for crystallization.

DISCUSSION

Myelin protein zero is an adhesive transmembraneprotein that is thought to maintain peripheral myelin

,

graphic trials.

at 308C after 5.5 h of growth. (A) Growth of E. coli in ETB withvarious buffers. The Tris–HCl (8.0) buffer slowed bacterial growthand the K-Phos (6.0) buffer reduced the maximum attainable OD

value. (B) pH of media during bacterial growth. All buffers failed toprovide adequate buffering at high OD values, although the K-Phosbuffers were more effective than the Tris–HCl buffers.

membrane juxtaposition through homophilic interac-tions between extracellular domains (17). The crystallo-graphic structure of the rat P0-ED isoform (9) refinedan earlier structure that used homology modeling (8).The human P0-ED isoform is 97% identical to the ratsequence, and several studies have examined possiblelinks between the disease-causing mutations of P0-EDand alterations in P0-ED structure and function (6,18, 25). Knowing the crystal structures of native and

disease-causing hP0-ED isoforms might provide de-tailed information on the molecular mechanisms under-lying myelin packing abnormalities in demyelinatingneuropathies and may point toward possible therapies.

AND KIRSCHNER

The protocol described here allows large-scale produc-tion of highly pure hP0-ED for use in crystallo-

404 BOND, SAAVEDRA

FIG. 2. Effect of pH on the growth of E. coli, monitored by measuringthe OD at different times during growth (data points). A colony ex-pressing high fusion protein was selected and grown overnight inLB/AMP. A total of 50 ml was added to 5 ml ETB with 200 mMTris–HCl, pH 8.0 (triangles); 200 mM Tris–HCl, pH 7.4 (circles); 200mM potassium phosphate (K-Phos), pH 7.4 (crosses); or 200 mMpotassium phosphate, pH 6.0 (squares). Cells were grown in 50-mlFalcon tubes at 378C, 300 rpm, and induced with IPTG (100 mg/ml)

The DNA fragment that codes for hP0-ED was in-serted into the pMal-p2 vector, chosen for its abilityto secrete the fusion protein to the periplasm, whichprovides a suitable folding pathway for transmembrane

FIG. 3. Effect of buffer on the growth of E. coli, monitored by mea-suring the OD at different times during growth (data points). A colonyexpressing high fusion protein was selected and grown overnight inLB/AMP. A total of 200 ml of overnight growth was added to 200 mlfresh LB/AMP and grown for 6 h at 378C, at which time the OD '0.5and pH '7.0. A total of 20 ml of this growth was added to 200 mlprewarmed media; “Enhanced” Terrific Broth (ETB; 20 g tryptone,40 g yeast extract, 50 g glycerol, 4 g glucose in 1 L dH2O; circles),ETB with 200 mM potassium phosphate buffer (K-Phos) pH 7.4 (ETB/K-Phos; triangles), or ETB with 200 mM K-Phos pH 7.4, 170 mMNaCl, 0.8 mM MgSO4 (ETB/K-Phos/NaCl/MgSO4; crosses). Cells weregrown in 2-L baffled flasks at 378C for 4 h and then induced withIPTG (100 mg/ml) at 308C. The pH was adjusted by the addition ofNaOH during induction, and 50-ml aliquots were removed at 8, 11.5,17.5, and 23.5 h. (A) Growth of E. coli in ETB at various bufferconditions. Increasing buffer concentration reduced cell growth rates.

The erratic growth rates of cells are likely due to removing aliquotsof medium, which improves aeration and hence cell growth. (B) pHof media during bacterial growth. As expected, the buffered mediashowed less dramatic pH fluctuations, and pH .7 proved easier tomaintain. However, all media showed progressively decreasing pH,and maintaining pH .7 became more difficult at high OD.

and electrophoresed on 10–20% precast gels. Lane 1, molecular weight markers; lane 2, ETB OSS at 7.3 OD (8 h growth); lane 3, ETB/K-Phos OSS at 5.5 OD (11.5 h growth); lane 4, ETB/K-Phos/NaCl/MgSO4 OSS at 8.5 OD (17.5 h growth); lane 5, ETB OSS at 12.5 OD (11.5

ero(*p

h growth); lane 6, ETB/K-Phos OSS at 12.6 OD (17.5 h growth); lanbuffer content of the growth medium did not affect the levels of fusion pof the medium did not affect levels of low-molecular-weight productsat high OD. The increased levels of low molecular weight products a

proteins and the proper formation of disulfide bonds(19, 26). However, secretion to the periplasm lowers thelevel of fusion protein expression, and performing ablue–white selection of cells containing the pMal-p2vector is problematic owing to the strength of the ptac

promoter. Transformants induced with IPTG can con-tain mutant plasmids that have either lost part or allof the fusion gene or no longer express it at high levels(pMAL Instruction Manual, Catalog No. 800, NEB).Moreover, cells transformed with an insert in the poly-linker often maintain some lacZa fragment activity,resulting in colonies with subtle shades of blue. Inducedcolonies exhibited a range of colors from blue to white,and transformation with pMal-p2 alone resulted in,50% white colonies. Thus, to prevent false positives,transformants containing the hP0-ED insert were se-lected from master plates using a blue–white chromo-genic reaction on replica plates.

Successive generations of bacteria maintained at 48Cshowed progressive reduction in fusion protein expres-sion, and colonies could not be kept for more than 1

week. SDS–PAGE analysis suggested that either theplasmid was lost completely or the insert removed re-sulted in MBP expression only. Consequently, we usedfresh stabs from frozen stocks for each growth, and theresulting colonies showed significant variation in fusion

7, ETB/K-Phos/NaCl/MgSO4 OSS at 13.0 OD (23.5 h growth). Thetein (arrow) or MBP (arrowhead) expression. The buffer concentration) or the ratio of fusion protein to total protein that appeared reducedpeared to correlate with decreased levels of fusion protein.

protein expression levels. Therefore, screening the colo-nies for high fusion protein expression was necessaryto ensure efficient protein production from large-scalebacterial growth.

The large-scale growth media described here con-tained enough nutrients to grow cells ,15 OD, but atthis cell density oxygen became the limiting nutrient(27). Baffled shaker flasks provide enhanced aerationthrough increased liquid mixing and cavitation, but thevessel opening must allow sufficient oxygen to the gas–liquid interface, and the liquid–gas ratio of the vesselshould not exceed ,10%. Shaker speed was increased to400 rpm at high cell concentration to provide sufficientoxygenation. To reduce foaming at these speeds, whichis known to reduce oxygen transfer across the gas–liquid interface, we added a suitable antifoam agent(e.g., Antifoam A) without any detrimental effects tocells or fusion protein expression.

Proteolysis of the fusion protein occurred when thelarge-scale growth media became acidic. Media con-taining various buffers were investigated; however,

EXPRESSION AND PURIFICATION OF MYELIN PROTEIN ZERO 405

FIG. 4. Analysis of the osmotic shock solutions from bacteria grown in different buffer conditions using SDS–PAGE. Bacteria were grownin ETB, ETB with K-Phos (ETB/K-Phos), and ETB with K-Phos and NaCl/MgSO4 (ETB/K-Phos/NaCl/MgSO4) and induced as described inFig. 3. Aliquots (50 ml) from each medium were removed after 8, 11.5, 17.5, and 23.5 h of growth, and the periplasmic fraction wasprepared using the osmotic shock protocol as described in the text. The OSS was concentrated 10-fold using Ultrafree 15 (5000 MWCO)

buffer reduced bacterial growth rates and pH adjust-ment with NaOH was required at high cell concentra-tions. Media without buffer, but with regular pH moni-toring, maintained a suitable pH range withoutincreased fusion protein degradation. Longer induction

the osmotic shock protocol (lane 2) contained MBP/hP0-ED fusion protein (arrow), MBP (arrowhead), and low-molecular-weight productsor MBP (lane 3), although prolonged column washing did elute some

,

hP0-ED (square), while levels of MBP appeared unaltered (arrow-head). The precipitate also contained additional high molecularweight material that remained at the top of the gel (#).

d predominantly fusion protein and MBP (lane 5). The fusion proteinlane 6). A second pass over the affinity column removed most MBP

n-exchange chromatography (lane 8) and finally purified with HPLC,

times and higher induction temperatures gave higheryields, but unacceptably high levels of fusion proteindegradation. Problems with fusion protein degradationcorrelate with IPTG concentration, and lowering con-centration decreases fusion protein degradation (pMalInstruction Manual, NEB). High levels of fusion proteinexpression can overload the membrane translocationsystem, clog transport channels, and lead to fusion pro-tein degradation. Various IPTG concentrations weretested in this study, but reduced fusion protein expres-sion or increased degradation was not observed withhigh IPTG concentrations.

Previous studies have shown varied MBP levels inthe periplasmic fraction purified by the osmotic shockprotocol (21, 28). These studies cited endogenous MBPexpression and proteolytic activity at the Factor Xa siteas likely sources of periplasmic MBP. Another studyproposed premature termination of transcription ortranslation as the cause of periplasmic MBP (29). Inour experience, limiting the amount of MBP in the per-iplasm proved difficult. Increased glycerol and glucose

406 BOND, SAAVEDRA, AND KIRSCHNER

FIG. 5. Analysis of MBP/hP0-ED fusion protein purification using SDS–PAGE. The protein profiles in lanes 1–8 were detected by Coomassieblue staining while that in lane 9 was detected by silver stain. Lane 1, molecular weight markers; lane 2, OSS; lane 3, affinity column flow-through; lane 4, affinity column wash; lane 5, affinity column elution; lane 6, post Factor Xa digestion; lane 7, second affinity column flowthrough; lane 8, DEAE fraction containing hP0-ED; lane 9, silver stained POROS 10 HQ fraction. The periplasmic fraction resulting from

(star). Affinity column flow-through showed little if any fusion proteinfusion protein and MBP (lane 4). The affinity column elution containewas cleaved by Factor Xa, creating the hP0-ED polypeptide (squareand fusion protein (lane 7). The hP0-ED was further purified with ioresulting in a single product as detected by silver staining (lane 9).

FIG. 6. Analysis of precipitant and supernatate of Factor Xa digestusing SDS–PAGE. Lane 1, molecular weight markers; lane 2, blank;lane 3, supernatant from digest; lane 4, precipitate from digest. Bothsupernatant and precipitate contained similar levels of proteins; how-ever, the precipitate showed increased fusion protein (arrow) and

appeared to be correlated to increased MBP levels(J. P. Bond, personal observation), possibly due to mini-mal maltose formation during medium autoclaving.MBP expression and secretion to the periplasm seems

a(a

FIG. 7. Elution profile of ion-exchange chromatography. The secondcontaining DEAE Sepharose. A linear 25–500 mM NaCl gradient (d(solid line); and SDS–PAGE confirmed the peak containing hP0-ED

to be an unavoidable consequence of utilizing thepMal-p2 vector.

Bacterial contamination is the main cause of reducedlevels of protein binding to the amylose resin, as bacte-ria contain amylase that readily releases amylose fromthe supporting resin (pMal Instruction Manual, NEB).The prefilter and 0.45-mm membrane provided the mostcost-effective sterilization for large volumes, and failureto remove bacteria from the OSS resulted in the lossof considerable column binding capacity. Amylose resinprovided an efficient purification step; however, anyperiplasmic MBP then copurified with the fusion pro-tein. This could explain the presence of an MBP bandin gels that were not treated with Factor Xa (seeFigs. 1 and 5).

Digestion of the fusion protein was conducted in thepresence of high salt and enhanced by the addition of0.02% SDS, which is assumed to partially denature thefusion protein allowing Factor Xa access to the cleavagesite (30). As the digestion proceeded, a fine white precip-itate formed. Dialysis of the sample into low salt buffer(20 mM Tris–HCl, pH 8.0) or the addition of 0.02%SDS caused the precipitate to solubilize. At low ionicstrength, the electrostatic repulsion between two P0-ED molecules is much stronger than their van derWaals attraction (10). Increasing ionic strength de-

creases the electrostatic repulsion, and molecular inter-action occurs as the van der Waals attractive force be-comes dominant. However, the resolubilized precipitatedid not purify in the same manner as supernatant of thedigestion reaction, probably due to nonnative refolding.

amylose column flow-through was loaded onto a 0.7 3 20 cm columnshed line) at 0.2 ml/min resulted in the protein profile shown heresterisk).

The resolubilized fusion protein and MBP did not bindthe AR column, an interaction dependent on tertiarystructure (31). The resolubilized hP0-ED showed anom-alous elution from ion-exchange chromatography aswell, and it was not used for any further purification.

During the digestion of shark P0-ED (,5 mg/ml, 57%homologous to rP0-ED), the solution viscosity increasesas the digestion proceeds, resulting in the formation ofa clear gel. This gelation phenomenon is reversible inlow salt buffer, and a similar behavior is observed withpurified rat P0-ED (9). At high concentration (,150mg/ml), rP0-ED shows a reversible transition betweenopalescence at low temperature and translucence athigh temperature. Equilibrium ultracentrifugationshows that tetramer formation has a strong tempera-ture dependence (9), implying that gelation is a likelyconsequence of quaternary P0-ED assembly. Similargelation behavior occurs with pure, synthetic peptidesthat form amyloid (32). Given the 80-A width of theP0-ED tetrameric assembly (10), which is similar tothat of certain Ab peptides in amyloid fibers (33), it wasproposed that P0-ED can form amyloid-like assemblieswith periodic arrangement of subunits and defined byregular b -sheet packing (10). This notion is supportedby our observation that digestion of the MBP/hP0-EDfusion protein resulted in a fine, fiber-like precipitate.

EXPRESSION AND PURIFICATION OF MYELIN PROTEIN ZERO 407

Upon analysis with SDS–PAGE, the precipitate showedelevated levels of fusion protein and hP0-ED relativeto the digest solution, indicating that these proteinsmay be integral components of the fiber-like precipitate.

MBP and any remaining undigested fusion protein

ak

have been able to produce large quantities of a highly

FIG. 8. Elution profile of HPLC ion-exchange chromatography. Thecolumn containing POROS 10 HQ medium. A linear 100–350 mM Nshown here (solid line), and SDS–PAGE confirmed the dominant pea

were effectively removed by passing the digested mix-ture over the AR column a second time. Maltose mustbe unbound from the MBP prior to this step; however,maltose is not efficiently removed from MBP via dial-ysis as, once it is free of one binding site, maltose willoften bind to another site before encountering the dial-ysis membrane (34). Another recommended method uti-lizes a hydroxyapatite column to bind MBP, with malt-ose being eluted with multiple column washes (pMalInstruction Manual, NEB). However, this method istime consuming and further dilutes the target proteinconcentration. Diafiltration, dialysis by repeated sam-ple concentration with buffer exchange, allowed rapidconcentration of the sample while substituting ARbuffer for digestion buffer. All three methods were in-vestigated, and diafiltration proved the most cost-effec-tive and time-efficient for removing maltose from MBP.

Diafiltration was used to concentrate and buffer-exchange the flow-through from the second AR column,

which was then loaded onto the DEAE Sepharosecolumn. Any MBP contamination complicated the ion-exchange chromatography as the isoelectric points ofhP0-ED and MBP are similar ( pI (hP0-ED) 5 4.8, pI

peak containing hP0-ED from the DEAE column was loaded onto aCl gradient (dashed line) at 5 ml/min resulted in the protein profile

contained .99% hP0-ED.

(MBP) 5 4.9 (35); pMal Instruction Manual, NEB), butlow volume flow rates and a gradual salt gradient al-lowed adequate protein separation. Final hP0-ED puri-fication was with POROS 10 HQ Perfusion Chromatog-raphy medium, which allows high protein bindingcapacity and fast flow rates (36).

In summary, this paper describes a cost-effectivetechnique for the production of hP0-ED suitable forcrystallographic trials. Without the use of specializedequipment, and using a commonly available vector, we

408 BOND, SAAVEDRA, AND KIRSCHNER

pure protein (see Table 1). This protocol will be of inter-est to investigators conducting crystallographic studiesof proteins that require secretion to form functional con-formations.

ACKNOWLEDGMENTS

We thank Dr. Hideyo Inouye for helpful discussions, Drs. Randall

Swartz, Karl Lawton, Ross Edwards, and Larry Shapiro for technicaladvice, and Veronica Eckstein for assistance with preparing the fig-ures. We acknowledge Drs. Clare O’Connor and Laura Hake for gener-ous use of their equipment. This research was supported by awardsto D.A.K. from Boston College Institutional Research Support Funds,

EXPRESSION AND PURIFICATIO

the Guillain-Barre Syndrome Foundation International, and NIH-NINDS NS39650.

REFERENCES

1. Lemke, G., and Axel, R. (1985) Isolation and sequence of a cDNAencoding the major structural protein of peripheral myelin. Cell40(3), 501–508.

2. Lemke, G., Lamar, E., and Patterson, J. (1988) Isolation andanalysis of the gene encoding peripheral myelin protein zero.Neuron 1(1), 73–83.

3. Kirschner, D. A., and Ganser, A. L. (1980) Compact myelin existsin the absence of basic protein in the shiverer mutant mouse.Nature 283(5743), 207–210.

4. Giese, K. P., Martini, R., Lemke, G., Soriano, P., and Schachner,M. (1992) Mouse P0 gene disruption leads to hypomyelination,abnormal expression of recognition molecules, and degenerationof myelin and axons. Cell 71(4), 565–576.

5. Martini, R., Zielasek, J., Toyka, K. V., Giese, K. P., and Schachner,M. (1995) Protein zero (P0)-deficient mice show myelin degenera-tion in peripheral nerves characteristic of inherited human neu-ropathies. Nat. Genet. 11(3), 281–286.

6. Warner, L. E., Hilz, M. J., Appel, S. H., Killian, J. M., Kolodny,E. H., Karpati, G., Carpenter, S., Watters, G. V., Wheeler, C.,Witt, D., Bodell, A., Nelis, E., Van Broeckhoven, C., and Lupski,J. R. (1996) Clinical phenotypes of different MPZ (P0) mutationsmay include Charcot–Marie–Tooth type 1B, Dejerine–Sottas,and congenital hypomyelination. Neuron 17(3), 451–460.

7. Saavedra, R. A., Fors, L., Aebersold, R. H., Arden, B., Horvath,S., Sanders, J., and Hood, L. (1989) The myelin proteins of theshark brain are similar to the myelin proteins of the mammalianperipheral nervous system. J. Mol. Evol. 29(2), 149–156.

8. Wells, C. A., Saavedra, R. A., Inouye, H., and Kirschner, D. A.(1993) Myelin P0-glycoprotein: Predicted structure and interac-tions of extracellular domain. J. Neurochem. 61(6), 1987–1995.

9. Shapiro, L., Doyle, J. P., Hensley, P., Colman, D. R., andHendrickson, W. A. (1996) Crystal structure of the extracellulardomain from P0, the major structural protein of peripheral nervemyelin. Neuron 17(3), 435–449.

10. Inouye, H., Tsuruta, H., Sedzik, J., Uyemura, K., and Kirschner,D. A. (1999) Tetrameric assembly of full-sequence protein zeromyelin glycoprotein by synchrotron x-ray scattering. Biophys. J.76(1 Pt. A), 423–437.

11. Hayasaka, K., Ohnishi, A., Takada, G., Fukushima, Y., andMurai, Y. (1993) Mutation of the myelin P0 gene in Charcot–Marie–Tooth neuropathy type 1. Biochem. Biophys. Res. Com-mun. 194(3), 1317–1322.

12. Bird, T. D., Ott, J., and Giblett, E. R. (1982) Evidence for linkageof Charcot–Marie–Tooth neuropathy to the Duffy locus on chro-mosome 1. Am. J. Hum. Genet. 34(3), 388–394.

13. Patel, P. I., and Lupski, J. R. (1994) Charcot–Marie–Tooth dis-ease: A new paradigm for the mechanism of inherited disease.Trends Genet 10(4), 128–133.

14. Uyemura, K., Kitamura, K., and Miura, M. (1992) Structureand molecular biology of P0 protein, in “Myelin: Biology and

Chemistry” (R. E. Martenson, Ed.), pp. 481–508. CRC Press,Boca Raton, FL.

15. Filbin, M. T., and Tennekoon, G. I. (1991) The role of complexcarbohydrates in adhesion of the myelin protein, P0. Neuron7(5), 845–855.

N OF MYELIN PROTEIN ZERO 409

16. Filbin, M. T., and Tennekoon, G. I. (1993) Homophilic adhesionof the myelin P0 protein requires glycosylation of both moleculesin the homophilic pair. J. Cell. Biol. 122(2), 451–459.

17. Kirschner, D. A., Inouye, H., and Saavedra, R. A. (1996) Mem-brane adhesion in peripheral myelin: Good and bad wraps withprotein P0. Structure 4(11), 1239–1244.

18. Kirschner, D. A., and Saavedra, R. A. (1994) Mutations in demye-linating peripheral neuropathies support molecular model of my-elin P0-glycoprotein extracellular domain. J. Neurosci. Res.39(1), 63–69.

19. Riggs, P. (1995) Expression and purification of maltose-bindingprotein fusions, in “Current Protocols in Molecular Biology”(F. M. Ausubel, Ed.), pp. 16.16.11–16.16.14, Wiley, New York.

20. Georgiou, G., and Valax, P. (1996) Expression of correctly foldedproteins in Escherichia coli. Curr. Opin. Biotechnol. 7(2),190–197.

21. di Guan, C., Li, P., Riggs, P. D., and Inouye, H. (1988) Vectorsthat facilitate the expression and purification of foreign peptidesin Escherichia coli by fusion to maltose-binding protein. Gene67(1), 21–30.

22. Hsiung, H. M., Mayne, N. G., and Becker, G. W. (1986) High-level expression, efficient secretion and folding of human growthhormone in Escherichia coli. BioTechnology 4, 991–995.

23. Bardwell, J. C., McGovern, K., and Beckwith, J. (1991) Identifica-tion of a protein required for disulfide bond formation in vivo.Cell 67(3), 581–589.

24. Minor, D. L., Jr., and Kim, P. S. (1994) Context is a major determi-nant of beta-sheet propensity. Nature 371(6494), 264–267.

25. Kirschner, D. A., Szumowski, K., Gabreels-Festen, A. A.,Hoogendijk, J. E., and Bolhuis, P. A. (1996) Inherited demyelinat-ing peripheral neuropathies: Relating myelin packing abnormali-ties to P0 molecular defects. J. Neurosci. Res. 46(4), 502–508.

26. Georgiou, G., and Valax, P. (1999) Isolating inclusion bodies frombacteria. Methods Enzymol. 309, 48–58.

27. Costilow, R. N. (1981) Biophysical factors in growth, in “Manualsof Methods for General Bacteriology” (P. E. A. Gerhardt, Ed.),pp. 66–78. American Society for Microbiology, Washington, DC.

28. Maina, C. V., Riggs, P. D., Grandea, A. G. d., Slatko, B. E., Moran,L. S., Tagliamonte, J. A., McReynolds, L. A., and Guan, C. D.(1988) An Escherichia coli vector to express and purify foreignproteins by fusion to and separation from maltose-binding pro-tein. Gene 74(2), 365–373.

29. Lauritzen, C., Tuchsen, E., Hansen, P. E., and Skovgaard, O.(1991) BPTI and N-terminal extended analogues generated byfactor Xa cleavage and cathepsin C trimming of a fusion proteinexpressed in Escherichia coli. Protein Expr. Purif. 2(5–6),372–378.

30. Ellinger, S., Mach, M., Korn, K., and Jahn, G. (1991) Cleavageand purification of prokaryotically expressed HIV gag and envfusion proteins for detection of HIV antibodies in the ELISA.Virology 180(2), 811–813.

31. Szmelcman, S., Schwartz, M., Silhavy, T. J., and Boos, W. (1976)Maltose transport in Escherichia coli K12. A comparison of trans-port kinetics in wild-type and lambda-resistant mutants as meas-ured by fluorescence quenching. Eur. J. Biochem. 65(1), 13–19.

32. Kirschner, D. A., Inouye, H., Duffy, L. K., Sinclair, A., Lind, M.,and Selkoe, D. J. (1987) Synthetic peptide homologous to betaprotein from Alzheimer disease forms amyloid-like fibrils in vitro.Proc. Natl. Acad. Sci. USA 84(19), 6953–6957.

33. Inouye, H., Fraser, P. E., and Kirschner, D. A. (1993) Structureof beta-crystallite assemblies formed by Alzheimer beta-amyloidprotein analogues: Analysis by x-ray diffraction. Biophys. J.64(2), 502–519.

34. Silhavy, T. J., Szmelcman, S., Boos, W., and Schwartz, M. (1975)

410 BOND, SAAVEDRA, AND KIRSCHNER

On the significance of the retention of ligand by protein. Proc. 36. Afeyan, N. B., Gordon, N. F., Mazsaroff, I., Varady, L., Fulton,Natl. Acad. Sci. USA 72(6), 2120–2124. S. P., Yang, Y. B., and Regnier, F. E. (1990) Flow-through particles

35. Inouye, H., and Kirschner, D. A. (1991) Folding and function ofthe myelin proteins from primary sequence data. J. Neurosci.Res. 28(1), 1–17.

for the high-performance liquid chromatographic separationof biomolecules: Perfusion chromatography. J. Chromatogr.519(1), 1–29.