Protein Science (1995), 4:421-432. Cambridge University Press. Printed in the USA. Copyright 0 1995 The Protein Society
Recombinant immunoglobulin variable domains generated from synthetic genes provide a system for in vitro characterizationof light-chain amyloid proteins
PRISCILLA WILKINS STEVENS,'.296 ROSEMARIERAFFEN,'p6 DEBORAH K. HANSON,' YA-LI DENG,' MARIA BERRIOS-HAMMOND,' FLORENCE A. WESTHOLM,' CHARLES MURPHY,3 MANFRED EULITZ,4 RONALD WETZEL,' ALAN SOLOMON,3 MARIANNE SCHIFFER,' AND FRED J. STEVENS'
' Center for Mechanistic Biology and Biotechnology, Argonne National Laboratory, Argonne,
Illinois 60439 Department of Biomedical Engineering, Northwestern University, Evanston, Illinois 60208 Human Immunology and Cancer Program, Department of Medicine, University of Tennessee Medical Center/Graduate School of Medicine, Knoxville, Tennessee 37920 GSF Institute of Clinical Molecular Biology, Munich 70, Marchioninistrasse 25, Federal Republic of Germany Macromolecular Sciences Department, SmithKline Beecham Pharmaceuticals, King of Prussia, Pennsylvania 19406
(RECEIVED October 20, 1994; ACCEPTED December 23, 1994)
Abstract
The primary structural features that render human monoclonal light chains amyloidogenic are presently unknown. To gain further insight into thephysical and biochemical factors that result in the pathologic depositionof these proteins as amyloid fibrils, we have selected for detailed study three closely homologous protein products of the light-chain variable-region single-gene family VKIV.Two of these proteins, REC and SMA, formed amyloid fibrils in vivo. The third protein, LEN, was excreted by the patient at levels of 50 g/day with no indication of amyloid deposits. Sequences of amyloidogenicproteins REC and SMA differed from the sequence of the nonpathogenic protein LEN at 14 and 8 amino acid positions, respectively, and these amino acid differences have been analyzed in terms of the three-dimensional structure of the LEN dimer. To provide a replenishable source of these human proteins, we constructed synthetic genes coding for the REC, SMA, and LEN variable domains and expressed these genes in Escherichia coli. Immunochemical and biophysical comparisons demonstrated that the recombinant VKIVproducts have tertiary structural features comparable to those of the patient-derived proteins. This welldefined set of three clinically characterized human KIVlight chains, together with the capability to produce these KIVproteins recombinantly, provide a system for biophysical and structuralcomparisons of two different amyloidogenic light-chain proteins and a nonamyloidogenic protein of the same subgroup. This work lays the foundation for future investigations of the structural basis of light-chain amyloidogenicity. Keywords: amyloid proteins; immunoglobulin light chains; kappa IV domains; recombinant human VL; synthetic genes; variable-domain dimerization
Over the past 30 years, detailed analyses of the structure and biophysical properties of immunoglobulin molecules have probed many aspects of Ig interactions and effector functions(Padlan, 1994). More recently, Ig genes have been cloned and altered by Reprint requests to: Fred J. Stevens, Center for Mechanistic Biology and Biotechnology, Argonne National Laboratory, Argonne, Illinois 60439; e-mail:
[email protected]. Abbreviations: ~ g immunoglobulin; , AL amyloidosis, lightchain amyloidosis; V, variable; C, constant; CDR, complementarity determining WiOn; FR, framework KgiOn; J, joining; r, recombinant; Gdn-HCI, guanidine hydrochloride; Fop,,, apparent fraction of unfolded protein; IPTG, isopropyl P-D-thiogalactopyranoside;TES, Tris-EDTA-sucrose buffer (200 mM Tris, 0.5 mM EDTA, 0.5 M sucrose, pH 8.0). The first two authors contributed equally to this project.
42 1
mutagenesis to investigate effects of the changes on biological activities; synthetic Ig genes have been generated for the production of unique antibody reagents for medical and diagnostic purposes (Borrebaeck, 1992). We are interested in examining another important aspect of Ig biology: namely the structural features of Ig light chains that lead to their pathologic deposition in tissue as renal tubular casts, basement membrane precipitates, and, in particular, amyloid fibrils. Amyloidosis is a pathologic condition in which protein isdeposited extracellularly throughout the body in the form of insoluble fibers. These tissue deposits impair organ function and lead to death' Defining features Of amy1oid deposits include the assembly of a primarily p-structure protein into nonbranching, insoluble fibrils of diameter 7-10 nm and a char-
P. W . Stevens et al.
422 acteristic green birefringence of bound Congo red dye when viewed under crossed polars- features indicating that proteins within the fibers arehighly ordered (Stone, 1990). Many different types of proteins are known to form amyloid. In the case of light-chain-associated amyloidosis, the deposits contain a monoclonal Ig light chain (Glenner et al., 1971) or, more often, a light-chain fragment that consists of the variable domain (VL) and varying amounts of the constant domain (CL), or a mixture of fragment and full-length light chain. However, thepresence of monoclonal light chains is not invariably associated with amyloid deposition; these proteins can be depositedin vivo in other pathologic forms,or, in some cases, they do not formdeposits at all (Solomon, 1986; Buxbaum, 1992; Eulitz, 1992). Considerable effort hasbeen devoted to determining the sequences of amyloid-associated Ig light chains in order to identify specific V-region residues responsible for fibril formation. Despite the substantial number of protein sequences derived fromamyloid-associatedlight-chainprecursors(i.e., Bence Jones proteins) and proteins extracted from amyloid deposits, as well as deduced from cDNAscloned from plasma cells of pa20 tients with AL amyloidosis, no common primary structuralfeatures have been identified that distinguish an amyloid from a non-amyloid light chain (e.g., see section I1 of Natvig et al., 1991; Aucouturier et al., 1992; Buxbaum, 1992, and references cited therein). However, the predisposition of certainlight chains to form amyloid has been evidenced by the predominance of Xvs. K-type proteins and thepreferential association oflight chains of the VXVI subgroup with this disease process (Solomon et al., 1982). Several types of in vivo and in vitro experimentalsystems have been used to probe themolecular differences between nonpathogenic and amyloidogeniclight chains and to identify particular proteins for detailed biochemical and biophysical studies (Solomon et al., 1991, 1992; Myatt et al., 1994). Although the usefulness of these experimental systems has been demonstrated, the availabilityof human light chains for study is limited, and the material extracted from tissue is often denatured and poorly soluble. 80 will make possible In this report,we developed a system that future studies foridentifying physical and biochemical features involved in light-chain amyloidogenicity. Two human Ig K IV proteins known to beamyloid-associated were isolated, and their sequences compared toa third KIVprotein that did not form amyloid deposits. Synthetic Ig gene constructs encoding the three proteins were prepared andused to produce human KIV Ig lightchain variable domains by recombinant methods in E. coli. In this manner,we synthesized in bacteria human VKIV proteins, the native forms of which are known to produce, or to fail to produce, pathologic depositsclinically. The ability to generate and modify human KIV light chains by recombinant meanswill provide a unique system for identifying structural and chemical features of Ig light chains that lead to amyloid formation.
tein proved non-nephrotoxic and non-amyloidogenic when tested in an in vivo experimental modelsystem (Solomon et al., 1991, 1992). The KIV Bence Jones protein REC was isolated from the urineof a patient with histologically proven AL amyloidosis. The light-chain SMA was extracted (Pras et al., 1968) from lymph node-derived amyloid fibrils of another AL amyloidosis patient.
Sequencing of LEN, REC, and SMA proteins and comparison of sequences After purifying proteins LEN, REC, and SMA fromurine samples or tissue extracts, we determined the V-region amino acid sequences of REC and SMA and resequenced the aminoterminus of LEN (Fig. 1). The protein sequence determined for LEN was identical to that previously published (Schneider &
I-------------------FR1"----------------------1-10 LEN D I V M T Q S P D S L A V S L G E R A T I N C K S SU4 ......................... REC .............. P . . . . . . . . . . germ . . . . . . . . . . . . . . . . . . . . . . . . .
--------CDRI-----------------I------------FR2--a b c d e f 30 40 LEN SM REC germ
S Q S V L Y S S N S K N Y L A W Y Q Q K P G Q P P . . . . . . . . .NR . . . . . . . . . L . . . . . .NL.DA.FDT.T . . . . . . . . . . . . ......... N...............
""""-I""c~~""I-""""-""""-""--
50 LEN SU4 REC germ
60
K L L I Y W A S T R E S G V P D R F S G S G S G T
......................... ........S ................ .........................
-FR3---------------------------------I----C~R390
70 LEN SM4 REC germ
D F T L T I S S L Q A E D V A V Y Y C Q Q Y Y S T . . . . . . . . . . . . . . . . . . . H . . . . H
......................... .........................
------I---------FR~-----I 100 LEN SU4 REC germ
108
P Y S F G Q G T K L E I K R . Q T . . q g t k . . L . . .PT ..G . . .V . . . .
.
Fig. 1. Amino acid sequences of KIV variable regions. Amino acidsequences of the variable regions of the LEN, SMA, and REC K I V proteins are shown relative to the nonamyloidogenic LEN protein sequence. LEN ocA dot indicates that an amino acid residue identical toof that curs at a particular position. The six-residue insert foundin CDRI of KIVproteins is identified as positions 27a-27f. Small letters at residues Results 100-103 of the SMA protein indicate positions where the amino acid sequence of SMA could not be unambiguously determined; we have asIsolation of LEN, REC, and SMA proteins sumed that the SMAsequence is identical to LEN atthese four positions. 1-3 (CDRI, The KIVBence Jones protein LENwas isolated from the urine Sequence numbering, complementarity determining regions CDR2, and CDR3), and framework regions1-4 (FRI, FR2, FR3, and of a patient with multiple myeloma (Solomon, 1985) who, deFor reference, the predicted FR4) are marked as in Kabat et (1991). al. spite excretion of upto 50 g of this protein daily, had no renal amino acid sequenceof the germline humanKIVV-gene is included o n the bottom line (Klobeck et al., 1985). dysfunction or evidence of amyloidosis. Additionally,this pro-
Recombinant VKIVdomains of amyloid proteins Hilschmann, 1974) with the exception that position 9 contained an Asp rather thanAsn residue. The sequence of VKIVLEN is identical to thatencoded by the single KIVgermline V-segment exon (Klobeck et al., 1985) except for an Asn to Ser substitution at position 29. The VL sequences ofREC and SMA differed from that of LEN by 14 and 8 amino acid residues, respectively. In REC, 8 of the 14 residues that differed from LEN occurred in the first complementarity determiningregion (CDRl). They included three charge changes: neutral to negative at positions 27d and 29 and positive to neutral at position 30. REC also contained two more Pro residues than LEN, at positions 15 (in the first framework region [FRl]) and 96 (in CDR3). The only other V-exon difference between the REC and LEN sequences was a Ser-Thr substitution at position 53 (in CDRZ). The sequence of REC residues 96-108 was identical to that encoded by the germline K joining-segment exon 54 (Hieter et al., 1982), but differed from the LEN sequence at three sites. Four of the eight residue differences between proteins SMA and LEN were located in CDR3. Positions 89 and 94 in SMA contained His residues, whereas in LEN, the positions had Gln and Thr, respectively. The CDRl of SMA differed from LEN by two amino acids (positions 29 and 30), but was of similar charge. The other two amino acid differences occurred in position 40 (Pro in LEN and Leu in SMA) and at aninternal position, 106, which contained the hydrophobic amino acids Ile in LEN and Leu in SMA. Notably, except for a changefrom Ser at position 97 in LEN to Thr in both REC and SMA (Thr occurs at this position in most K light chains), all amino acid variations between REC and LEN are different from amino acid variations between SMA and LEN. Four amino acid positions of SMA in the highly conserved FR4, however, could not be established.
423
S29N Y96Q
OR
Analysis of amino acid differences in terms of three-dimensional structure We crystallized the VKIV LEN protein and determined its tertiary structure by X-ray diffraction (M. Schiffer, C.-H. Chang, C. Ainsworth; Brookhaven Protein Data Bank 1LVD) employing the molecular replacement method and using as the search molecule the VKIlight-chain dimer REI (Epp et al., 1975). As is characteristic of the KIVsubgroup, CDRl of LEN contained a six-residue insertion relative to theCDRl of REI. In the partially refined structure ( R = 2470, 1.8-A resolution), all sidechain atomicpositions of residues of LEN that differ from REI have been located, as well as the backbone atoms of the extra six CDRl residues. LEN is the first human KIVprotein to be analyzed crystallographically; nevertheless its basic domain structure demonstrated the expected homology to the well-conserved VL conformation. Because we expect little change in thebackbone coordinates of the LEN VKdimer during further refinement, we generated computer displays of REC and SMA structures by substituting appropriate amino acids at the positions that distinguish these two proteins from LEN. These structural models of REC and SMA VKIV domains showed that the majority of differences from LEN are spatially localized on theend of the light-chain dimer where the six CDRs form apocket, whereas onlytwo amino acid substitutions occurred in each protein on the opposite end of the dimer (Fig. 2). Internal P-sheet framework residues were extremely well conserved among the three proteins.
Sma Fig. 2. Computer graphics representations of rREC and rSMA structures using LEN backbone coordinates. The VKIVdomain dimer backbone structure of the nonpathologic protein LEN (Brookhaven Protein Data Bank ILVD) is used to illustrate the positions of amino acid substitutions found in the amyloidogenic proteins REC (above) and SMA (below). Each dimer representation shows monomers that are essentially symmetric around a vertical central axis. On both monomers of each dimer, complete side chains indicate amino acid positions that differ from LEN. Labels for these positions, however, occur in only one of the two monomers; labels list the LEN amino acid residue prior to the position number and the REC or SMA amino acid substitution after the position number. REC and SMA dimer representations retain the LEN backbone coordinates and have not been energy minimized.
Construction of synthetic genes encoding the VL domains of proteins LEN, REC, and SMA The amino acid sequence data for the KIVproteins were used to construct synthetic genes encoding the VL portions of proteins LEN, REC, and SMA to provide a replenishable source of all three proteins for physical and chemical analyses. A clone
424
P. W. Stevens et al.
containing the human germline KIVV-segment exon (Klobeck et al., 1985),synthetic oligonucleotides for the human K J-segment exon used in LEN (52 but with Ser, not Thr, in the second position; Hieter et al., 1982), and the expression vector pASK40 (Skerra et al., 1991) were used to construct a plasmid that encoded a germline-typeVKIVdomain. From this template, we derived plasmidsencoding the LEN VL and SMA VL using invitro mutagenesis (Kunkel et al., 1987), recombinant PCR (Higuchi et ai., 1988),and insertion of synthetic oligonucleotides. A fourth plasmid containing a coding region for the VKIV domain of REC was generated de novo from eight overlapping synthetic oligonucleotides using recursive PCR (Prodromou & Pearl, 1992). The sequences of the coding regions of these plasmids are provided in Figure 3.
0rpA Signal Sequence M K K T A I A germ I en sma rec
I
A
V
A
L
Protein expression and purification The vector pASK40 positions a coding region for the ompA leader sequence prior to thepolylinker so that therecombinant proteins produced are directed to the periplasmic space. The oxidizing environment of this compartment facilitates folding and disulfide-bond formation within the recombinant VKIVdomains (Skerra et al., 1991; Skerra, 1993). In addition, by isolating the recombinant VKIV domainsfrom periplasmic extracts (Pluckthun & Knowles, 19871, a number of extraneous E. coli proteins were eliminated in this step (lanes 1-3, Fig. 4). The recombinant VKIVdomains were purified to apparent homogeneity through a combination of ion-exchange and gel-filtration chromatography (lanes 4-6, Fig. 4). During electrophoresis, purified rREC
A
G
F
A
T
-4 V
1 atg aaa aag aca gct atc gcg att gca gtg gca ctg gct ggt ttc gct acc gtt
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . 4FRl A
germ 55 1 en sma rec
Q
A
D
I
V
M
T
Q
S
P
D
S
L
A
V
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . gcg . gct . . gta . . . tcc . . . . gca gta agc CCA G
sma rec
15 L
gct cag gca gac atc gtg atg acc cag tct cca gac tcc ctg gct gtg tct ctg
cm1 germ l a 1 en
S
gta
E
R
A
T
I
N
C
K
S
S
Q
S
V
L
Y
27f S S
ggc gag agg gcc acc atc aac tgc aag tcc agc cog agt gtt tta tac agc tcc
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ggt gaa cgc
. . . .
tgt aaa
.
tct
Q
Q
.
AAT C l T
.
GAC GCC tct
FRZ N
N
K
germ 153 aac aat aag I en TCT sma AGG rec l T C GAC ACG
N
Y
L
A
W
Y
K
P
G
Q
P
P
45 K
aac tac tta gct tgg tac cag cag aaa cca gga cag cct cct aag
. . . . . . . . . . . . . . . . . . . . . . . . . . . . CTC ggg . . . . . ACT . . . . . . . . ggt . . . aaa
FR3
CDRZ L germ 298 I en sma rec
sma rec
I
Y
W
63
A
S
T
R
E
S
G
V
P
D
R
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . TCT . . . . gta . . cgg . . S
G
S
G
T
D
F
T
L
T
I
S
S
L
Q
rec
A
81 E
ggc agc ggg tct ggg aca gatttc act ctc act atc agc agc ctg cog gct gaa
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ggt . ggt . . . . . . . . . . . . caa gca . FR4 99
CM3
sma
S
ttg
D germ 316 1 en
F
ctg ctc att tac tgg gca tct acc cgg gaa tcc ggg gtc cct gac cga ttc agt
G germ 262 I en
L
V
A
V
Y
Y
C
Q
Q
Y
Y
S
T
P
Y
S
F
G
gat gtg gca gtt tat tac tgc cog cog tac tac tcc act ccg tac tcc ttc ggt
. . . . . . . . . . . . . . . . . . . . . . . . . CAT . . . . CAC . CAG ACC . . gac gta gct gta tac . tgt . caa . . . act cct CCT ACT ttt ggc 108 Q
germ 378 cag I en sma rec CGT
G
T
K
L
E
I
K
*
*
ggt act aaa ctg gaa atc aaa tga taa
. . . . . . . . . . . . . . CTC . ggc act aag GlT gag . .
. .
CGC CGG tga CGC
Fig. 3. Nucleotide sequences of KIVcoding regions. Nucleotide sequences of the light-chain coding regions of pkIVexOOl (germ), pkIVlen004 (lea), pkIVsma007 ( m a ) , and pkIVrec006 (rec) plasmids, with the germline construct as the reference sequence. A dot indicates that a codon identical tothat of the germline construct occurs at a particular position. Alternative codons are shown in lowercase letters if the nucleotide change did not alter the amino acid sequence; capital letters are used for codons that code for amino acid residuesdifferent from those encoded in thegermline construct. For convenience, the amino acid sequence encoded by the germline construct is included, together withannotations of complementarity determining regions 1-3 (CDR1, CDR2, and CDR3), framework regions 1-4 (FR1, FR2, FR3, and FR4), and light-chain amino acid sequence numbering as inKabat et al. (1991), with the ompA leader sequence assigned numbers -21 through -1. An arrow indicates the position at which the ompA leader sequence is cleaved when the recombinant protein is transported to the periplasm of the E. coli host.
Recombinant VKIV proteins domains of amyloid
kDa
M
1
2
425
3
4
5
6
94.0 67.0
Fig. 4. Purification of recombinant VKIVdomains analyzed by SDS-PAGE. Various samples obtained during purification of recombinant V K I V domains were dissolved in SDS-PAGE loading buffer with 1.25% 0-mercaptoethanol and analyzed on a 13% acrylamide gel stained with Coomassie blue. Samples were prepared from I-L shaker-flaskcultures of rLEN, rREC, and rSMA. Lane 1: rLEN, total proteins prepared by solubilizing pelleted whole cells in loading buffer. Lane 2: rLEN, cytoplasmic proteins prepared by solubilizing spheroplasts in loading buffer. Lane 3: rLEN, 10 pg of periplasrnic proteins from the supernatant comprising the peri4, 5 , 6: rLEN, rREC, plasmicfraction.Lanes rSMA, respectively; 5 pg of purified recombinant V K I V proteins.
43.0 30.0 20.1
14.4
consistently showed a decreased mobility relative to the other domains, the intradomain disulfide bond adjacent to this Trp two VKIV domains, presumably because its lower pl (3.5 for dramatically quenched its fluorescence (Hurle et al., 1994). The rREC versus 8.5 for rLEN and rSMA) resulted in lower SDS fluorescence of the LEN VKIV domain in its native conformabinding (cf. Dunn, 1993). Amino-terminal sequenceanalysis of tion was higher than formost VL domains because it contained purified rLEN, rSMA, and rREC VL verified that the ompA a second, unquenched Trp residue at position 50. Nevertheless, signal sequence had indeed been properly cleaved. total fluorescence of the LEN VKIV domainin native buffer was Typically a I-L overnight shaker-flask culture yielded more less than half (28.7%) that of the samemolecule in denaturant, than 20 mg rLEN, whereas yields of the amyloid-associated suggesting that in native buffer one of the LEN VKIV resiTrp rREC and rSMAwere consistently less, approximately 4 mg/L dues was highly quenched and that the disulfidewas therefore and 7.5 mg/L, respectively. In general, synthesis of rLENwas in its normal position. Underidentical conditions, rLEN exhibrelatively insensitive to theE. coli host strain used. On the other ited a very similar fluorescence yield when compared to denahand, production of rREC was extremely sensitive to host strain tured material (29.0%). The close agreement of these ratios and required the use of strain BL26 (which lacks both theompT suggested that rLEN was folded into a native conformation and outer membrane protease [Grodberg& Dunn, 19881 and theIon also that both VKIV preparations were very pure. cytoplasmic protease [Chung & Goldberg, 19811) with a low Unfolding of LEN and rLEN VKIV domains in increasing growth temperature (30°C) and slow agitation rate (100-1 15 concentrations of the denaturant Gdn-HCI, plotted as the Fop,, rpm). All threeVKIVdomains were, therefore, prepared in (Finn et al., 1992) in Figure 5 , demonstrated that, within experstrain BL26 with the growth conditions determined tobe optiimental variance, the LEN and rLEN domains unfoldedidenmal for rREC production. tically in response to Gdn-HCI. The original fluorescence data Quantities of REC and SMA proteins isolated from patient samples were very limited; only patient-derived LEN protein was sufficient for extensive comparisons with the recombinant VKIV domains. By comparative immunodiffusion analysis using a rab121 . ' t bit polyclonal anti-LEN Bence Jones protein-specific antiserum, I the rLEN and LEN VKIV fragments formed precipitin reactions of complete identity, as did the rREC and rSMA proteins (data 8 0 not shown). Immynological similarity of patient-derivedand re0 combinant human V K I V proteins provided a preliminary indication that the recombinantlight-chain domains had probably folded properly in E. coli. .I,._
,
0
Stabilities of LEN and rLEN V~IVproteins
I
1
Gdn-HCI (M)
4
426
P. W. Stevens et al.
(not shown) was fit atotwo-state transition model (Santoro& eluting at low protein concentrations with the expected M, Bolen, 1988); results confirmed that the LEN and rLEN VKIV (1 1,500) of a monomer domain and athigh protein concentrations with the expectedM , of the dimer (23,000) and at interdomains had essentially identical folding thermodynamics. By this analysis, LEN had a standard free energy of unfolding mediate concentrations displaying a sharp leading edge and (AGO,,,,) of -9.0 f 1.0 kcal/mol, a cooperativity value for the extended trailing boundary, asexpected for systems with rapid association and dissociation of noncovalent dimers (Winzor & Gdn-HC1 unfolding transition of -5.2 f 0.6 (kcal/mol)/M, and a midpoint of the unfolding transition of1.74 f 0.02 M GdnScheraga, 1963; Stevens, 1989). Computer simulation quantitating VL dimerization constants from chromatographic proHCl (Finn et al., 1992). Corresponding values for rLEN were files (Kolmar et al., 1994) yielded values of approximately 4 x -9.7 k 1 . 1 kcal/mol, -5.5 f 0.6 (kcal/mol)/M, and 1.76 f 0.02 M Gdn-HC1. lo5 M" for both LEN and rLEN. The amyloid-associated domains rREC and rSMA also demonstrated concentration-dependent dimerizationby size-exclusion HPLC analysis of dilution series of light-chain domains chromatography. The rSMA protein behaved chromatographLEN and rLEN, when analyzed by size-exclusion chromatogically in a manner quite similar to rLEN, whereas,in contrast, raphy (Stevens et al., 1980), yielded virtually identical chromatoa significant proportion of rREC eluted as a dimer, even at low graphic profiles, providing further evidence for the proper concentrations (data not shown). Similar analyses of Bence folding of the recombinant VKIVdomain (Fig. 6). Both domains Jones protein REC showed that the intact light chain also exdemonstrated the concentration-dependent dimerization char- hibited an unusually high degree of aggregation, even at low conacteristic of native humanIg light chains (Stevens et al., 1980), centrations (data notshown). Dimerization constants calculated
A
B
0.020
0.50.4
0.6
0.7
0.8 1.0 0.9
0.020
1.1 1.3 1.2
rLEN (VeNt) n
I
LEN (VeNt)
D
1.6
' 7
4
rLEN (rng/rnl) Fig. 6 . Chromatographic analysis of dilution series of LEN and rLEN VKIVproteins. Top: Concentration dependence of V K I V proteins prepared in phosphate buffer with 100 mM NaCl. Each sample was analyzed by small-zone size-exclusion HPLC as described in the Materials and methods. Column eluates were monitored at 214 nm, and elution profiles normalized as described. On each graph, the earliest eluting peak represents the sample with the highest concentration. A: rLEN VKlV protein analyzed at concentrations ranging from 18 to 0.05 mg/mL. Reference concentrations of 18, 5 , 0.5, and 0.05 mg/mL are indicated by solid lines. B: LEN VKIVprotein analyzed at concentrations ranging from 14 to 0.05 mg/mL. Reference concentrations of 14, 5, 0.5, and 0.05 mg/mL are indicated by solid lines. Bottom: Quantitative analysis of self-association affinities of (C) rLEN and (D) LEN VKIVproteins. Best fits were obtained assuming dimer and monomer velocities of 2.1 and 1.25 cells/cycle, respectively. For both rLEN and LEN,instantaneous equilibration was assumed, and best fits were obtained for self-association constants of 4 x IO5 M".
Recombinant VKIVdomains of amyloid proteins
case of AL amyloidosis, no one amino acid position or substitution can be implicated because many different Ig V-region light-chain sequences have been found among AL amyloidassociated proteins. Even among the light chains LEN, REC, and SMAof the relatively uncommon single-gene family, VKIV, Analysis of an rLEN-REC hybrid variant several different positions within theVL domain are involved in the propensity of these proteins to form amyloid, and these In an initial experiment to identify specific structural features of these VKIV domains associated with particular biophysical positions contain particular residues that affectbiophysical characteristics such as dimer formation, exposure of hydrophobic properties relevant for amyloidogenesis,we constructed a plasresidues, solubility, or stability. mid encoding a hybrid rLEN-REC variant that contained the eight REC CDRlresidues that differed from LEN, but not the other six REC substitutions. When this hybrid recombinant Light-chain dimer formation VKIV domain was produced and analyzed by size-exclusion chromatography, dimerization of the hybrid protein was simi-The increased dimerization noted in the amyloid-associated REC and rREC proteins may have contributed to the pathologic delar to that of rLEN and yielded a calculated dimerization conposition of this light chain. The calculated dimerization constant stant of approximately 3 X lo5 "' (Fig. 7). These data suggest for rREC was unusually high, IO' M-', approximately two orthat the apparent enhanced dimerization of rREC either results ders of magnitude higher than that of the nonpathologic LEN from one or more of the limited number of remaining substiand rLEN proteins. The dimerization constants of rLEN, IO5 tutions, or represents a concerted effect of CDRl and nonM-', and rSMA, IO6 M-I, were in the range of the selfCDRl positions. association constantsobserved for other humanIg light chains, lo3 M" to IO6 M" (Maeda et al., 1978; Stevens et al., 1980; Discussion Kolmar et al., 1994). It has been suggested that unusual amino acidswithin the inBiophysical characteristics of individual light ner @-sheets,which form the contactregions at the dimer interchains influence amyloidogenicity face, may be responsible for increasing the dimer stability of The potentialof human light chains to form pathologic depos- amyloidogenic light chains and thereby promoting fibril formation (Dwulet et al., 1985; Liepnicks et al., 1991; Aucouturier its in vivo was previously correlated with the propensity ofthese et al., 1992). The sequence of the amyloid-associated protein components to form pathologic depositsin both in vivo and in REC, however, differed from that of LEN primarily at CDR vitro experimental models (Solomon etal., 1991, 1992; Myatt residues and not atresidues comprising the@-sheetframework et al., 1994). In some types of hereditary amyloidoses, single amino acid changesin normal human proteins may be respon- (see Figs. 1,2). That CDRlof rREC is not the sole determinant of its exceptionally high dimerization potential was demonsible for amyloid fibril formation (e.g., see section VI1 of Natstrated by analysis of the rLEN-REC hybrid, which exhibited vig et al. [1991] and references cited therein). However, in the a dimerization constant on the order of that of rLEN. It is likely that Pro at position 96in CDR3 is also required for rREC's increased dimerization, because the rLEN-REC hybrid, which r contains the LEN-type residue Tyr at position 96, exhibiteda 1.w dimerization constant about two orders of magnitude lower than that of rREC. The nature of the residue at position 96is known to affect light-chain self-association (Stevens et al., 1980). Although considerable sequence variation has been found at this position due to its location at theV-J junction (Hieter et al., 1982), only REC and three other myeloma-associated K light chains that containPro at position 96 have been identified. Two 0.25 of these light chains formed amyloidfibrils in vivo: NIE (Liepnicks et al., 1991; M.D. Benson, pers. comm. ofsequence) and MAL (Rodilla Sala et al., 1991); clinical data regarding the third 0 protein, NIM (Eulitz & Kley, 1977), were not available. I , 0 2 4 6 8 10 12 14 16 I8 20 Although K and A light chains are encoded by different VLConcentration (mglrnl) gene families, at least half of the 10 A myeloma-associated huFig. 7. Chromatographic analysis of dilution series of rLEN, rSMA, man light chains that contain Pro at position 96 have been rLEN-REC, and rRECVKIVproteins. Each sample of VKIVproteins derived from patients with amyloidosis: NIG77 and NIG51 was prepared in phosphate buffer with 100 mM NaCl and analyzed by (Tonoike et al., 1985), TYL (Eulitzet al., 1991), and EMM and small-zone size-exclusion HPLC as described inthe Materials and methFIEG (A. Solomon, unpubl. sequences). Another h protein that ods. Self-associationaffinities were calculated from the chromatographic contains Pro at position 96is AI Bence Jones protein RHE, for profiles for rSMA, rLEN-REC, and rREC VKIVproteins, and values for rLEN (from Fig. 6C)are included for comparison. To allow for comwhich a three-dimensional structure of the Vh dimer hasbeen parisons of thefour sets of data despite differences in velocity and disdetermined crystallographically. Whether or not the RHEdimer persion parameters, the Kp values in this figure were normalized, using formed amyloid invivo is unknown. Its structure, however, dis1 for the monomer peak and 0 for the dimer peak for each VKIVproplayed a unique interdomain interactionwith monomers shifted tein. Experimental data points and lines representing self-association constants with the best fits for each VKIVprotein are plotted. significantly relative to other crystallographically characterized by computer simulation (Kolmar et al., 1994) were about 7 X IO' M-] for rSMA and approximately 2 X lo7 M-l for rREC (Fig. 7).
,
427
P. W. Stevens et al.
428 VL dimers (Furey et al., 1983) and may be relevant to understanding the dimerization properties of rREC.
Exposure of hydrophobic residues The occurrence of hydrophobic amino acid residues on exposed surfaces may be another factor that contributes to amyloid assembly due to decreased solubility or increased aggregation. For SMA, substitution of the hydrophobic residue Leu at position 40 for the Proin LEN may have promoted pathological deposition of this protein. Hydrophobic substitutions occur at position 40 in amyloid-forming K chains BRE (Liepnicks et al., 1991; M.D. Benson, pers. comm. of amino acid sequence), ARN (Aucouturier et al., 1992), and WR (Westermark et al., 1982), and Ala occupies position 40 in MCG, an amyloid-associated X chain (Schiffer et al., 1973; Fett & Deutsch, 1974) as well as in CUM (Kabat et al., 1991), a K protein of uncertain pathologic nature. Virtually all human light-chain sequences compiled to date contain Pro atposition 40 (686of 722; G.Johnson & T.T. Wu, private comm.; see also Kabat et al., 1991); to date,no example has been found of a human light chain known to be nonamyloidogenic but containing a hydrophobic amino acid residue at position 40. Like SMA, REC also contained a hydrophobic amino acid on what is presumed to be an exposed portion of the VL domain; Phe replaced the Asn residue found at position 28 in both LEN and SMA. Based on the position of Asn 28 in the LEN crystal structure, it is unlikely that Phe 28 of REC is involved in this protein’s enhanced potential for dimer formation. Instead this substitution, like Leu40 in SMA, may contribute to other aspects of amyloid assembly such as decreased light-chain solubility, stacking of domains within amyloid fibrils, or other aggregation phenomena.
Protein solubility and stability Lower isoelectric points have been noted for several amyloidogenic proteins and may contribute to thedecreased solubility of these molecules (Bellotti et al., 1990). Three charge changes in CDRl of REC resulted in its isoelectricpoint being significantly more acidic than that of LEN (3.5 versus 8 . 5 ) . However, a low PI value must not be a necessary condition for amyloid potential. The second amyloidogenic protein in our study, rSMA, had a PIvalue essentially equivalent to thatof rLEN ( 8 . 5 ) , and isoelectric points greater than 8.0 have also been reported for other amyloid-associatedlight chains (MOL, SPR, CAN; Bellotti etal., 1990). Altered protein stability may also contribute to a light chain’s propensity for amyloid fibril formation (Hurle et al., 1994). Consistent with previous data on mutational effects in a number of other protein aggregation systems (Wetzel, 1994),unfolding intermediates of light-chain proteins may be key structures in the assembly of VL domains into amyloid fibrils. According to this line of reasoning, particular amino acid residues at critical positions in the VL domain may facilitate amyloidosis by destabilizing domain structure, decreasing the energy barrier to the formation of a partially folded state under physiological conditions. Whether this state is actually sensitive to aggregation and whether that aggregate takes theform of amyloid or amorphous deposits, may depend on other key sequence elementsthat become exposed to solvent in the partially folded intermediate.
Synthetic genes provide system for analysis of amyloid-implicated residues The unique featureof our set of three VKIV proteins is that the two amyloid-associated light chains REC and SMA can be compared to a closely related light-chain protein, LEN, for which there is extensive clinicaldocumentation verifying its nonpathologic character. It will be possible, therefore, to interpret biophysical characterizations of the recombinant amyloid-associated proteins in the context of a bona fide nonamyloid control protein. Because syntheticgenes have beenconstructed for producing rLEN, rREC, andrSMA, we are now in a position to generate variants of these KIVdomains and to determine the contributions of particular amino acid residues to intermolecular processes that may be important in amyloid fibril assembly in vivo. Preliminary results with a hybrid rLEN-REC VKIV domain demonstrates that such analysis is not only possible, but also informative. A detailed understanding of molecular processes such as dimerization, domain interface interactions, solubility, stability, and higher-order aggregation of light chains will be required for the development of effective therapies that can be applied to AL amyloidosis, presently an incurable disease.
Materials and methods Nomenclature In these studies, LEN refers to the pepsin-derived VKIV proteolytic fragment of the LEN Bence Jones protein isolated from patient urine (Solomon & McLaughlin, 1969); this LEN fragment contains the entire V-region plus 10 residues of the constant region. REC is the REC urinary KIVBence Jones protein, and SMA refers to KIV light-chain protein extracted from lymph node amyloid deposits. The designations rLEN, rREC, and rSMA are used for the recombinant human VKIVdomain proteins. All recombinant VKIVdomains contain 114 amino acid residues, beginning with Asp 1 of FRl and concluding with Arg 108 of FR4 (see Fig. 1). Position 108 is included in the recombinant VKIVdomains because this residue is clearly visible in the crystal structure of purified WAT K light-chain dimer (Huang et al., 1994) and because it affects the PI of the rLEN protein. With Arg 108 as the final residue of the recombinant protein, the isoelectric point of the rLEN VKIV domainis similar to that of the LEN pepsin-derived VKIV domain,which contains an additional nine residues of the constant region (Thr-Val-Ala-AlaAla-Pro-Ser-Val-Phe, starting at Thr 109). The amino acid numbering system employed is that of Kabat et al. (1991). Plasmids containing the VdV-domaincoding regions were designated with the common prefix pkIV followed by a unique alphanumeric identifier.
Isolation of LEN, REC, and SMA proteins The Bence Jones proteins LEN and REC were isolated from urine specimens, LEN from a patient with multiple myeloma and REC froma macroglobulinemia patient with amyloid deposition in the mucosa of the oral cavity (see Stevens etal., 1991, for a discussion of Bence Jones proteins). Protein SMA was extracted from lymph node-derived amyloid fibrils of a patient with lymphoma and microglobulinemia. Purification of lightchain proteins from urine samples was accomplished as de-
429
Recombinant VKZVdomains of amyloid proteins scribed in Solomon (1985); the LEN Bence Jones protein was cleaved with pepsin and its VKIV fragment isolated as described (Solomon & McLaughlin, 1969). The amyloid fibril protein SMA was extracted and purified as described by Pras et al. (1968). Proteins were identified as KIVlight chains using polyclonal rabbit anti-human KIVantisera (Solomon, 1985). Complete V-regions of the REC andSMA proteins were sequenced by methods described previously (Eulitz et al., 1991). LEN had been previously sequenced (Schneider & Hilschmann, 1974), and resequencing of the amino-terminal 30 residues confirmed the original sequence, with the exception of Asp (rather than Asn) at position 9. REC was sequenced from tryptic peptides 1-18, 19-24,25-45, 46-54, and 104-108. In REC, trypsin did notcleave after the arginine at position 61; instead a tryptic peptide extending from positions 55 to 104 was isolated and residues 55-81 were sequenced from theamino-terminal portion of this peptide. Peptide 55-104 showed an extraordinarily high tendency to form aggregates, was insoluble in solvents normally applied to dissolve tryptic peptides, and was resistant to most proteolytic enzymes tried for generating smaller products. Endopeptidase Glu-C was capable of cleaving this peptide, however, and the remaining amino acids of REC were determined from two fragments generated by cleaving peptide 55-104 with trypsin and Glu-C: 62-81 and 82-104. Amino-terminal sequencing of SMA yielded the amino acid sequence for residues 1-27a. SMA tryptic peptides 1-24,25-30, 31-39, 40-61, 62-103, and 104-108 were used to generate the remaining sequence, but because the clinical sample yielded little protein, it was not possible to determine unambiguously the residues at positions 100-103. We have assumed that the SMA amino acids at these four J-segment positions (FR4) are identical to those of LEN. Computer analysis of protein structures Computer graphics structural representations of rREC and rSMA dimers were developed on a Silicon Graphics workstation using Insight I1 software (BioSym). C a coordinates from the LEN dimer structure (Brookhaven Protein Data Bank code 1LVD) were usedto generate the backbone for each dimer structure, with the side chains of amino acid positions that differed from LEN displayed in full. Structural representations of REC and SMA were not minimized. Plasmids and host strains An M13 clone of the germline V-region exon of the human KIV light chain was obtained from ATCC (#61121; Klobeck et al., 1985), and the plasmid pBS+/- was purchased from Stratagene. The vector pASK40, which contains an ompA signal sequence prior to the polylinker (Skerra et al., 1991), wasthe kind gift of A. Skerra. Recombinant proteins utilizing this vector are synthesized as fusions with the ompA leader sequence, which is cleaved during transport of the mature protein into theperiplasmic space of the host E. coli. Three strains of E. coli were used as plasmid hosts: DH5a [F-~8OdlacZAMlSA/acZYA-argF)U169endAl recAl hsdR17 (r;mG) deoR thi-1 supE44 X- gyrA96 relAl] from BRL was used for cloning and plasmid preparation; CJ236 [dutl ungl thil relAl pCJ105(Cmr)] from Bio-Rad was used to prepare
uracil-substituted DNA for site-specific mutagenesis (Kunkel et al., 1987); and BL26 [F- ompT hsdS,(rim,) dcm gal lac] from Novagen was used for production of recombinant proteins because it yielded more recombinant VKIV proteinthan other E. coli strains tested. Nucleic acid methodologies Plasmids were prepared and analyzed according to standard methods (Sambrook et al., 1989). Dideoxy sequencing of plasmids was performed with Sequenase kits (United States Biochemical; Sanger & Coulson, 1975), and Muta-Gene In Vitro Mutagenesis kits (Bio-Rad) were used for site-specific mutagenesis according to the uracil-replacement method (Kunkel et al., 1987). PCR (Innis et al., 1990) wasrun on a Perkin Elmer thermal cycler 9600 using either Taq DNA polymerase (Perkin Elmer) or Vent DNA polymerase (New England BioLabs), with temperatures and cycle times determined empirically. Oligonucleotides used for mutagenesis, PCR reactions, and plasmid construction were designed using OLIGO software (Rychlik & Rhoads, 1989) or the Genetics Computer Group Sequence Analysis Software Package (Devereux et al., 1984) and were synthesized at the University of Chicago using Applied Biosystems reagents and equipment. The sequences of all oligonucleotides used to construct plasmids described in this paper and PCR conditions used for various constructions are available upon request. Construction of plasmids containing LEN, REC, SMA and hybrid LEN-RECsequences The Bam HI-Sph I fragment containing the VKIVexon was subcloned from ATCC clone #61121 into the bifunctional vector pBSi-1- and sequenced, confirming the VKIVcoding region sequence reported in GenBank (Klobeck et al., 1985; Benson etal., 1993). The VKIV exon wasamplified by PCR with a sense-strand primer, which positioned a Hinc I1 site at the first codon of the mature KIVlight chain (Asp 1 in Fig. 1) and an antisense-strand primer, which added to the 3‘ end of the VKIVexon the 12 codons of the LEN J-segment (Tyr 96-Arg 108 inFig. l), tandem stop codons, and a Hind 111 site. The amplified fragment was digested with Hinc I1 and Hind 111 and cloned into thevector pASK40 (Skerra et al., 1991), whichhad been digested with Eco RI, blunted with mung bean nuclease, and digested with Hind 111. This generated a complete “germline” VKIV-domain coding region (including both V- and J-segments) following the ompA signal sequence of the pASK40 vector, with an additional Ala codon encoded at the junction of the blunted Eco RI site of the vector with the HincI1 site of the insert. By site-specific mutagenesis (Kunkelet al., 1987), the codon for Asn 29 was mutated to Ser to generate the rLEN VKIV-domain. By recombinant PCR (Higuchi et al., 1988), the additional Ala codons at the beginning of the recombinant germline and rLEN VKIVdomain constructs were removed, and a codon for a terminal Arg (position 108 in Fig. 1) was added at the 3’ end of the rLEN construct. These final plasmids containing the germline and rLEN VdV-domain sequences are called pkIVexOOl and pkIVlen004, respectively. Because the REC sequence contained 14 differences from LEN, many steps would have been required to derive a gene encoding the RECsequence using the germline- or rLEN-encoding
430 plasmids as template. Instead we synthesized the rREC VKIV coding sequence de novo using recursive PCR (Prodromou & Pearl, 1992) with eight overlapping oligos of length 70-85 nucleotides. The rREC sequence synthesized included the region of the pASK40 vector from the Xba I site preceding the signal sequence to the end of the ompA signal sequence plusa coding region for the entire rRECVKIV-domain followed by stop codons and a Hind I11 site. After the Xba I-Hind 111 fragment containing the rRECcoding sequence was cloned into pASK40, it was altered by site-specific mutagenesis to correct a Thr-Ser codon substitution mistakenly included in the original oligonucleotides. The plasmid containing the rREC VKIV-dOmain sequence is called pkIVrec006. The rSMA VKIV-domain construct was generated from the pkIVlen004 plasmid using recombinant PCR (Higuchi et al., 1988) to convert the three LEN residues, Ser 29, Lys 30, and Pro 40, to Asn, Arg, and Leu, respectively (see Fig. 1). The resulting plasmidwas digested with Pst I and Hind111 to remove codons for the carboxy-terminal 30 residues of the VKIVdomain. A series of four oligos, two sense-strand and two antisense-strand, coding for the carboxy-terminal30residues of the rSMA VdV-domain sequence were annealed and ligated into thePst 1- and Hind111-cut plasmid, giving rise to plasmid pkIVsma007, which contained the entire rSMA VKIV-domain sequence. The hybrid &EN-REC VdV-domain construct was generated from the pkIVlen004 plasmid using recombinant PCR (Higuchi et al., 1988) to convert codons for the LEN CDRl residues to those of the REC CDRl(residues 24-34, Fig. 1). The resulting plasmid (pkIVlen006) was identical to pkIVlen004 except for the replacement of nucleotides 136-167 (Fig. 3) with the sequence aut ctt tta gac gcc tcc ttc gat acg aac ace, which encodes the amino acid sequence NLLDASFDTNT, and a silent mutation in the signal sequence (gcg to gct at nucleotides 19-21). The ompA signal sequence and VKIV-domain coding sequences of all plasmidswere confirmed by dideoxy sequencing (Fig. 3).
Expression and purification of recombinant VKZVdomains
P. W. Stevens et al. g for I5 min at 4 "C. The supernatant constituted the periplasmic fraction. The periplasmic fraction from 2 L of starting culturewas dialyzed against 10 mM Tris, pH 8.0. The volumewas reduced by ultrafiltration on anAmicon-stirred cell (YM3 membrane), and the sample was applied at a flow rate of 0.7 mL/min to two 5-mL Econo-Pac Q cartridges (Bio-Rad) connected in series and equilibrated with the same Tris buffer. At this step, rLEN and rSMA eluted inthe flow-through fraction, whereas rREC and the rLEN-REC hybrid bound to the column and were eluted with a 75-mL, 0-100 mM NaCl gradient. Fractions containing VKIV domains were identified by SDS-PAGE on PhastGels(8-25%, Pharmacia), exchanged into 10 mM NaOAc, pH 5.6,by ultrafiltration, and applied at0.7 mL/min to two 5-mL Econo-Pac S cartridges (Bio-Rad) connectedin seriesand equilibratedwith the same acetate buffer. The rLEN and rSMA proteins eluted from the S-cartridgewith a 135-mL, 0-150 mM NaCl gradient; the rREC and rLEN-REC domains eluted in the flow-through fraction. Fractions containing VKIV proteins were exchanged into 20 mM Tris HCl, 150 mM NaCl, pH 7.2, concentratedby ultrafiltration, and applied at a flow rate of 0.5 mL/min to a HiLoad 16/60 Superdex 75 gel-filtration column (Pharmacia) equilibrated with the same buffer. Purified recombinantV K I V proteins appeared to be more than 95% pure by SDS-PAGE analysis (Laemmli, 1970) on 1.5-mm, 13% acrylamide gels using a Bio-Rad Mini-Protean I1 apparatus with proteins stainedwith Coomassie brilliant blue (Fig. 3, lanes 4-6). Purified proteins were exchanged into HPLC buffer (20 mM potassium phosphate, 100 mM NaCl, pH 7.0), concentrated by ultrafiltration to 30-50 mg/mL, and stored at 4 "C. Protein concentrations (mg/mL)were determined at280 nm using an extinctioncoefficient of 2.1 calculated from the amino acid compositionof the LEN VKIV protein as described in Cantor and Schimmel(l980). Theisoelectric points of recombinant proteins were determined using 1 pg of rLEN, rREC, and rSMA domains on Pharmacia PhastGelisoelectric focusing gels with Pharmacia 3-10 standards. Ouchterlony analysis was performed with rabbit antihuman VKIV antiserum as described previously (Solomon, 1985).
Gdn-HCI unfolding of VKIVproteins BL26 cells transformed with plasmids pkIVlen004, pkIVrec006, Samples of approximately0.5 pM LEN or rLEN in 10 mM sopklVsma007, or pkIVlen006 were grown in 2XYT medium dium phosphate buffer, pH 7.5, were incubated with various (Sambrook et al., 1989) containing 100 pg/mLcarbenicillin concentrations of Gdn-HC1 15-20h at25 "C. The fluorescence reached (Sigma Chemical Company)at 30 "C. When the culture an A,,, of 0.75-1 .O, expression was inducedby addition of IPTG of each 400-pL sample was read ina microcell in a Perkin Elmer MPF-66, using excitation at295 nm and measuring emis(Sigma Chemical Company) toa final concentration of 1 mM. sions at 350 nm. Data were fit to a two-state transition model, Cell growth was continued for an additional 16h. A low agitation with native and denaturedbaselines included in the fit,using the rate (100-115 rpm) and 30 "C temperature were used throughout the growth period because growth at faster agitation rates method of Santoro and Bolen (1988) in the nonlinear leastsquares software NLIN (SAS Institute, Cary, North Carolina). or highertemperaturesfrequentlyresultedin cell lysis, Using the methoddescribed in Finn et al. (1992), LEN and rLEN which contaminated the periplasmic fractionwith cytoplasmic Fappplots were constructed separately from original fluoresproteins. cence data. Preparation of the periplasmic extracts was based on Pluckthun and Knowles (1987). Cells were harvested by centrifugation at 4,000x g 10 min at 4 "C. The cell pellet from each liter Small-zone size-exclusion HPLC of culture was gently resuspended in 20 mL of ice-cold TES The HPLCsystem consisted of a 0.3-cm x 25-cm glass column buffer. One milliliter of a freshly prepared solution of lysozyme (20 mg/mL in TES) was added to the suspension, followed by (Alltech Associates) packed with Superose 12 (Pharmacia) and a Pharmacia 2248 HPLC pump ata flow rate of 0.06 mL/min. 40 mLof diluted TES(1 :1 inHzO). The cells were incubated on The mobile phase consisted of 20 mM potassium phosphate, ice for 1 h with gentle shaking and then centrifuged at 27,000 X
Recombinant VKIVdomains of amyloid proteins
100 mM NaCl, pH7.0. All experiments were performed at room temperature. Protein samples ranging in concentration from 0.005 to 50.0 mg/mL were injected in a volume of 5 pL with a Rheodyne 7010 injection valve. The column eluent was monitored at 214, 280, or 254 nm by an HP 1040A diode array detector. Typical run times ranged from 30 to 45 min. The data were collected and stored asdescribed (Stevens et al., 1986; Stevens, 1989). Chromatograms were normalized by summation of the absorbances at1 O , OO data points collected during the run and by scaling the data so that the integrated area under each elution profile was equal to1 . This allowed evaluation of differences in protein aggregation uncomplicated by peak-height differences due to varying amounts of protein applied. Partition coefficients (K,) were determined by elution volume (V,) and the relationshipK p = (V, - b)/(V, - Vo)where Vo and were the elution volumes of bluedextran (excluded volume) and acetone (included volume), respectively.
Computer simulation of chromatographic data
Computer simulation of VKIVdomain migration in small-zone size-exclusion HPLC was accomplished by explicit calculation of monomer and dimer concentration changes during small time steps as previously described (Kolmar et al., 1994). The simulated column consisted of708 cells, with each cell representing 2.5 pL. Complete monomer/dimer equilibration during each time step was assumed.
Acknowledgments We thank Professor Arne Skerra at the Max-Planck-lnstitut fur Biochemie in Martinsried, Germany for the giftof the plasmid pASK40, and we gratefully acknowledge Mr. Paul Gardner at the University of Chicago, whosynthesized our numerous oligonucleotides.We are grateful to Dr. Merrill D. Benson of Indiana University School of Medicine for communicating thefull V-region sequences of amyloid light chains BRE and NIE, and to Mr. George Johnson and Professor Tai TeWu of Northwestern University for database analyses of human light-chain proteins (Sequences of Proteins of Immunological Interest database,listing as of July 1994). This research was supported by the US Departmentof Energy, Office of HealthandEnvironmentalResearch,undercontract W-31-109ENG-38; by US Public Health Service grantDK43757; and by National Cancer Institute grant CA10056. A.S. is an American Cancer Society Clinical Research Professor.
References Aucouturier P, Khamlichi AA, Preud’homme JL, Bauwens M, Touchard G, Cogne M. 1992. Complementary DNA sequence of human amyloidogenic immunoglobulin light-chain precursors. Biochem J285:149-152. Bellotti V, Merlini G, Bucciarelli E, Perfitti V, Quaglini S, Ascari E. 1990. Relevance of class, molecular weight and isoelectric point in predicting human amyloidicity. Br J Haematol 74:65-69. 1993. GenBank. NucleicAcidsRes Benson D, LipmanDJ,OstellJ. 21:2963-2965. Borrebaeck CAK, ed. 1992. Aniibody engineering: A praciical guide.New York: W.H. Freeman. Buxbaum J. 1992. Mechanisms of disease: Monoclonal immunoglobulin deposition; amyloidosis, light chain deposition disease, and light and heavy chain deposition disease. Hemaiol/Oncol Clin N A m 6:323-346. Cantor CR, Schimmel PR. 1980. Biophysical chemistry, part II: Techniques for ihe study of biological structure and function. San Francisco: W.H. Freeman. p 380.
43 1 Chung CH, Goldberg AL. 1981. The product of the Ion (capR) gene in Escherichia coli is the ATP-dependent protease, protease La. Proc Nut1 Acad Sci USA 78:4931-4935. Devereux J, Haeberli P, Smithies 0. 1984. A comprehensive set of sequence analysis programs for the VAX. Nucleic Acids Res 12:387-395. Dunn MJ. 1993. Gel electrophoresis: Proteins. Oxford: BIOS ScientificPublishers. pp 57-58. Dwulet FE, Strako K, Benson MD. 1985. Amino acid sequence of a X VI primary (AL) amyloid protein (WLT). Scand J Immunol22:653-660. Epp 0, Lattman EE, Schiffer M, Huber R, Palm W. 1975. The molecular structure of a dimer composed of the Bence-Jones protein REI refined at 2.0-A resolution. Biochemistry 14:4943-4952. Eulitz M. 1992. Amyloid formation from immunoglobulinchains. Biol Chem Hoppe-Seyler 373:629-633. Eulitz M, Kley HP. 1977. The primary structure of a human immunoglobulin L-chain (K-type, subgroup 11): Bence Jones protein NIM. Immunochemistry 14:289-297. Eulitz M, Murphy C, Weiss DT, Solomon A. 1991. Serologic and chemical differentiation of human hIII light chain variable domains. J Immunol 146:3091-3096. Fett JW, Deutsch HF. 1974. Primary structure of the Mcg h chain. Biochemisiry 13:4102-4114. Finn BE, Chen X, Jennings PA, Saalau-Bethell SM, Matthews CR. 1992. Reversible unfolding of proteins: Kinetic and thermodynamicanalysis. In: Rees AR, Sternberg MJE, Wetzel R, eds. Protein engineering. Oxford, UK: IRL Press at Oxford University Press. pp 167-189. Furey W, Wang BC, Yo0 CS, Sax M. 1983. Structure of a novel Bence-Jones protein (Rhe) fragment at 1.6 A resolution. J Mol Biol 167:661-692. Glenner GG, Terry W, Harada M, Isersky C, Page D. 1971. Amyloid fibril proteins: Proof of homology with immunoglobulin light chains by sequence analyses. Science 172:1150-1151. Grodberg J, Dunn JJ. 1988. ompTencodes the Escherichia coli outer membrane protease that cleaves T7 RNA polymerase during purification.J Bacteriol 170:1245-1253. Hieter PA, Maize1 JV, Leder P. 1982. Evolution of human K J region genes. J Biol Chem 257:1516-1522. Higuchi R, Krummel B, Saiki RK. 1988. A general method of in vitro preparation and specific mutagenesis of DNA fragments: Study of protein and DNA interactions. Nucleic Acids Res 16:7351-7367. Huang DB, Chang CH, Ainsworth C, Brunger AT, Eulitz M, Solomon A, Stevens FJ, Schiffer M. 1994. Comparison of crystal structures of two homologous proteins: Structural origin of altered domain interactions in immunoglobulin light-chain dimers. Biochemistry 33:14848-14857. Hurle MR, Helms LR, Li L, Chan W, Wetzel R. 1994. A role for destabiking amino acid replacements in light-chain amyloidosis. Proc Nail Acad Sci USA 91:5446-5450. Innis MA, Gelfand DH, Sninsky JJ, White TJ. 1990. PCRprotocols: A guide to methods and applications. San Diego, California: Academic Press. Kabat EA, Wu TT, Perry HM, Gottesman KS, Foeller C. 1991. Sequences of proteins of immunological interest, 5th ed. NIH publication no. 91-3242. Washington, DC: US Department of Health and Human Services, US Government Printing Office. Klobeck HG, Bornkamm GW, Combriato G, Mocikat R, Pohlenz HD, Zachau HG. 1985. Subgroup IV of human immunoglobulin K light chains is encoded by a single germline gene. Nucleic Acids Res 13:6515-6529. Kolmar H, Frisch C, Kleemann G, Gotze K, Stevens FJ, Fritz HJ. 1994. Dimerization of BenceJones proteins: Linking the rate of transcription from an Escherichia coli promoter to the association constant of REIv. Biol Chem Hoppe-Seyler 375:61-70. Kunkel TA, Roberts JD, Zakour RA. 1987. Rapid and efficient site-specific mutagenesis without phenotypic selection. Methods Enzymol154:367382. Laemmli UK. 1970. Cleavage of structural proteins during theassembly of the head of bacteriophage T4. Nature 227:680-685. Liepnicks JJ, Benson MD, Dwulet FE. 1991. Comparison of the amino acid sequences of ten kappa I amyloid proteins for amyloidogenic sequences. In: Natvig JB, Fsrre 0, Husby G, Husebekk A, Skogen B, Sletten K, Westermark P, eds. Amyloid and amyloidosis 1990. Dordrecht, The Netherlands: Kluwer Academic Publishers. pp 153-156. Maeda H, Steffen E,Engel J. 1978. Kinetics of dimerization of the BenceJones protein AU. Biophys Chem 9:57-64. Myatt EA, Westholm FA, Weiss DT, Solomon A, Schiffer M, Stevens FJ. 1994. Pathogenic potential of human monoclonal immunoglobulin light chains: Relationship of in vitro aggregation to in vivo organ deposition. Proc Nail Acad Sci USA 91:3034-3038. Natvig JB, Fsrre 0, Husby G, Husebekk A, Skogen B, Sletten K, Westermark P, eds. 1991. Amyloid and amyloidosis 1990. Dordrecht, The Netherlands: Kluwer Academic Publishers.
432 Padlan EA. 1994. Anatomy of the antibody molecule. Mol Immunol 31:169-217. Pluckthun A, Knowles JR. 1987. The consequences ofstepwise deletions from the signal-processing site of 0-lactamase.JBiol Chem 262:3951-3957. Pras M, Schubert M, Zucker-FranklinD, Rimon A, Franklin EC. 1968. The characterization of soluble amyloid prepared in water. J Clin Invest 47:924-933. Prodromou C, Pearl LH.1992. Recursive PCR: A novel technique for total gene synthesis. Protein Eng 5:827-829. Rodilla Sala E, Kratzin HD, Pick AI, Hilschmann N. 1991. Biclonality in amyloidosis patient MAL: One clone producing an amyloidogenic, the 0, Husby other a non-amyloidogenic kappa L-chain. In: Natvig JB, Forre G, Husebekk A, SkogenB, Sletten K, WestermarkP, eds. Amyloid and amyloidosis 1990. Dordrecht, The Netherlands: Kluwer Academic Publishers. p p 161-164. Rychlik W, Rhoads RE. 1989. A computer program for choosing optimal oligonucleotides for filter hybridization, sequencing and in vitro amplification of DNA. Nucleic Acids Res 1723543-8551. Sambrook J, Fritsch EF, ManiatisT. 1989. Molecular cloning: A laboratory manual, 2nd ed. Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press. Sanger F, Coulson AR.1975. A rapid method for determining sequences in DNA by primed synthesis with DNA polymerase.JMol Biol94:441-448. Santoro MM, Bolen DW. 1988. Unfolding free energy changes determined by the linear extrapolation method. 1. Unfolding of phenylmethanesulfonyl a-chymotrypsin using different denaturants. Biochemistry 27:80638074. Schiffer M, Girling RL, Ely KR, Edmundson AB. 1973. Structure of a htype Bence-Jones protein at 3.5 A resolution. Biochemistry 12:4620-463 1. Schneider M, Hilschmann N.1974. The primary structure of a monoclonal immunoglobulin L-chain of K-type, subgroup I V (Bence-Jones protein Len): A new subgroup of the K-type L-chain. Hoppe-Seyler’s Z Physiol Chem355:1164-1168. Skerra A. 1993. Bacterial expression of immunoglobulin fragments. Curr Opin Immunol5:256-262. Skerra A, Pfitzinger I, Pluckthun A. 1991. The functional expression of antibody Fv fragments in Escherichia coli: improved vectors and a generally applicable purification technique. Bio/Technology 9:273-278. Solomon A. 1985. Light chains of human immunoglobulins. Methods Enzymol 116:101-121.
P. W. Stevens et al. Solomon A. 1986. Clinical implications of monoclonal light chains.Semin Oncol 13:341-349. Solomon A, FrangioneB, Franklin EC. 1982. Bence Jones proteins and light chains of immunoglobulins: Preferential association of the VXVI subgroup of human light chains with amyloidosis AL(h). J Clin Invest 70:453-460. Solomon A, McLaughlin CL.1969. Bence-Jones proteins and light chains of immunoglobulins: Formation and characterization of amino-terminal (variant) and carboxyl-terminal (constant) halves. J Biol Chem 244: 3393-3904. Solomon A, Weiss DT, Kattine A A . 1991. Nephrotoxic potential of Bence Jones proteins. N Engl J Med 324:1845-1851. Solomon A,Weiss DT, Pepys MB. 1992. induction in mice of human lightchain-associated amyloidosis. A m J Pathol 140:629-637. Stevens FJ. 1989. Analysis of protein-protein interaction by simulation of small-zone size exclusion chromatography: Stochastic formulation ofkinetic rate contributionsto observed high-performance liquid chromatography elution characteristics. Biophys J55:1155-1167. Stevens FJ, LeBuis DA, Eider WJ, Ainsworth CF. 1986. Macromolecular interactions: Application of microcomputer-controlled, high speed sizeexclusion chromatography. LCGC 4:340-348. Stevens FJ, Solomon A, SchifferM . 1991. Bence Jones proteins: A powerful tool for the fundamental study of protein chemistry and pathophysiology. Biochemistry 30:6803-6805. Stevens FJ, Westholm FA, Solomon A, Schiffer M . 1980. Self-association of human immunoglobulinK I light chains: Role of the third hypervariable region. Proc Null Acad Sci USA 77:1144-1148. Stone MJ. 1990. Amyloidosis: A final common pathway for protein deposition in tissues. Blood 75531-545. Tonoike H, Kametani F, Hoshi A, Shinoda T. 1985. Primary structure of an amyloidogenic Bence Jones protein NiG77. Biochem Biophys Res Commun 126:1228-1234. Westermark P, Sletten K, Pitkanen P, Natvig JB, Lindholm CE. 1982. LOcalized laryngeal amyloidosis: Partial characterization ofan amyloid fibril protein AL. Mol Immunol 19:447-450. Wetzel R. 1994. Mutations and off-pathway aggregation of proteins. Trends Biotechnol 12:193-198. Winzor DJ, Scheraga HA. 1963. Studies of chemically reacting systems on Sephadex. I. Chromatographic demonstration of the Gilbert theory. Biochemistry 2:1263-1267.
