34. B. R. Cullen, il1UI.. "6, 973 (1986).
`35. P. E. Pcllctt, K. G. Kousoulas, L. P. Pereira, B.
`Roiz.man,J. Virol. 53, 243 (1985).
`36. J. D. Dignam, R. M. Lcbovitz, R. G. Roeder,
`Nwki&.A.cills RR. 11, 1475 (1983).
`37. Supported by the National Cancer Institute (grants
`CAOM94 and CA19264), the National Institute for
`(grant Ail.24009),
`Al1crgy and Infectious DUcascs
`
`the United Scares Public Hcahh Service, and the
`American Canccr Society (MV2W). N.M., T.M.K.,
`and D.S. were prcdoaoral trainees ot'USPHS train­
`ing grants AI07099, CA09241, and GM7281, re­
`
`spectively
`
`.
`
`23 October 1987; acceptcd 31Dcccmbcr
`
`1987
`
`Reshaping Human Antibodies: Grafting an
`Antilysozyme Activity
`
`quenced as described (7). To reshape the
`NEW heavy chain, we started from a syn­
`thetic gene in an Ml3 vector (Fig. lb)
`containing the framework regions of human
`NEW and the CDR.s from mouse antibody
`Bl-8 (5). Long oligonuclcotides with multi­
`ple mismatches with the template (8) were
`used to replace each of the hypcrvariablc
`loops in tum by site-directed mutagcnesis:
`the central mismatched portion of the ·prim­
`er encoded each CDR of the Dl.3 heavy
`chain, and the 5' and 3' ends of the primer
`were complementary to the Banking frame­
`work regions. Thus after three rounds of
`mutagcncsis, the reshaped gene (Hu V H­
`LYS) encoded the framework regions of
`NEW with the hypcrvariablc regions of
`Dl.3 (Figs. le and 2). This was assembled
`with the heavy-chain constan
`t region of
`human immunoglobulin G2 (HulgG2) (9)
`to give the plasmid pSVgpt-HuVHLYS­
`HulgG2. The plasmid was transfcctod by
`clcctroporation ( 10) into the mycloma line
`J558L (11), which secretes
`a mouse
`�light
`chain. Transfcctants resistant to mycophen­
`olic acid were screen ed for secreti on of
`immunoglobulin by �cl electrophoresis of
`supematants from [3 S]mcthioninc-labcled
`cells. The scacted product (HuVHLYS­
`HulgG2, X) was purified on protein A­
`Scpharosc ; the � light chain was exch�
`for the Dl.3 K light chain in vitro; and the
`HuVHLYS-HulgG2, K antiboqy was puri­
`fied (12). In parallel experiments as contro�
`the mouse Dl.3 variable region (MoVff.
`LYS) was attached to the heavy-chain con­
`stant region of mouse IgGl (MolgGl) in
`a pSVgpt vector
`(pSVgpt-MoVHLYS­
`MolgGl), and antibody was expressed and
`rcassociatod as above (Fig. ld).
`ce quench was used to measure
`Fluorescen
`
`MARTINE VBRHOEYBN, CESAR MILSTEIN, GREG WINTER*
`The production of therapeutic human monodooal antibodies by hybridoma ttcb.nolo­
`gy bas proved difticult, and this bas prompted the "humanizing" of mouse monodooal
`antibodies by recombinant DNA techniques. It waa shown pRViously that the binding
`site for a small bapten could be grafted from the heavy-chain variable domain of a
`mouse antibody to that of a human mydoma protein by transplanting the bypcrvaria­
`blc loops. It is now shown that a large binding site for a protein antigen (lyacnyme) can
`also be transplantcd from mouse
`heavy chain. The success
`to human
`of such
`constructions may be facilitated by an induced-fit mechanism.
`F Oil PASSIVE IMMUNITY Oil ANTIBODY
`of
`contrast, the three-dimensional structure
`the complex of lysozymc and the mouse
`, monoclonal anti­
`therapy in humans
`antibody Dl.3 has been solved (6), and
`bodies designed to eliminate toxins,
`viral and bacterial pathogens, or other cells
`about 690 A2 of the solvent-accessible sur­
`would ideally be of human origin (1). Un­
`face of the antibody is buried on complex
`formation. Both V H and light-chain variable
`fortunately it has proved difficult to make
`(V L) domains make c:xtcnsivc contacts to
`human monoclonal antibodies by hybrido­
`ma technology (2). Chimeric antibodies
`lysozymc, but most
`of the hydrogcn-bond­
`arc made to the CDR.s of the
`with mouse
`variable and human constant
`ing contacts
`domains have been constructed by linking
`heavy chain. Thus, of 12 hydrogen bond
`inn:tactiom proposed (6), 9 arc made to the
`together the genes encoding each domain
`binant anti­
`(3, 4), and expressing the recom
`heavy chain. We have� the hypcrvaria­
`blc loops of the human NEW heavy chain (5)
`bodies in myeloma cells. However, the
`mouse variable region may itself be seen as
`with those from the Dl.3 heavy chain.
`foreign (1). We have therefore attempted to
`The variable domains of the mouse anti­
`insert the antigen-binding sil'C of a mouse
`body to lysozymc were cloned and sc-
`antibody, rather than the whole variable
`Ha
`region, directly into a human antibody. In
`Fig. 1. Vectors for heavy-chain c:xprcssion. (a)
`L.,._ ..... v•tlfl'-... ---J8
`Mouse vNP gene (4) and designated hen: as
`previous work, the three heavy-chain hyper- L:..
`,.;;;...
`'°""'
`variable regions [or complementarity-deter-
`MoVaNP, (b) reshaped HuVNP gene (5) desig-
`Hb L
`natcd here as HuVaNP, (c) HuVaLYS gene
`mining rcgions (CDR.s)] from a mouse anti-
`HuVtlfl'
`8
`•
`' i •
`with the hcaty-chain constant region gene of
`'°""' eofla1
`2
`3
`body to a haptcn were transplanted onto the
`human
`IgG2, (d) MoVaLYS gene with the
`t region gene of mouse IgGl,
`framework regions of the heavy-chain vari-
`able (VH) domain of a human mycloma � i _:. ,,:"�:." •. u
`c
`heavy-chain constan
`� �
`immunoglobulin heavy-chain enhancer (Igh cnh)
`and(•) the backbone of the pSVgpt vector
`with
`protein. In combination with the mouse '°""' coRa1
`2
`,
`1tu111G2
`light chain, the
`(4). The Va domains arc denoted in black, stip-
`tightly to hapten (5). Although it seems �'°�.:. "'"�·�:.:,,.,.,.,._, �1: • C: �! �orana�:::;,=>�=
`reshaped heavy chain bound
`d
`likely that both heavy and light chains make
`1ysozymc (Dl.3), and human mycJoma protein
`e
`(NEW), respectively. The constant domaim arc
`contacts to the haptcn, the relative contribu-
`98 denoted in black or open boxes to signify sc-
`E EH
`tion of each chain is unknown. Nor is it clear
`qucoccs encoding mouse or human constant do-
`w
`y
`p o tc en
`th small h dro h b ·
`c
`hapt
`hcth
`Ecogpt
`mains. lg pro, heavy-chain promoter; L, leader
`NP (4-hydroxy-3-nitrophcnacctylaminocap-
`cxon;
`H, Hind ill B B HI S Sac I E, "-
`;
`, am ;
`,
`; ....,...
`roatc) simply binds to a hydrophobic pocket RI restriction sites. The reshaped NEW heavy chain was expressed from a vector derived from the
`HI fragment, oontains heavy-chain promoter
`at the base of the hypcrvariablc loops. By pSVgpt-HuVaNPvccror(b). ThustheHuVaNPgcnc,cloncdinitiallyinMl3mp9asaHindill-Bam
`, the CDRs of the Bl-8 antibody, and the framework
`•regions of human NEW (5). The CDRs of Bl-8 wen: then replaced with those of Dl.3 using Jong
`mutagcnicoligonuclcocidcs (sec Fig. 2). ThcHind ill-Barn HI fragment, now carrying the HuV51LYS
`scne (c) was excised from the Ml3 Vector and cloned into the pSV Vector (e) along with a Barn HI
`fragmen t encoding the heavy-chain constan
`pSVgpt-MoVaLYS-MolgGl vector to exp� recombinant Dl.3 is swnmari7.Cd in (d).
`t region of human IgG2 (9). The construction of the
`SCIBNCB, VOL. 239
`
`Medical Rcscardl <:.ounci1 Laborarory ofMolccular Biol­
`ogy, Hilb Road, Cambridge CB2 2QH, England.
`*To whom rorrcspondc:ncc should be addrcacd.
`153.+
`
`• • •
`
`er
`
`1 of 3
`
`BI Exhibit 1068
`
`

`

`CDRs could be supported by the NEW
`Fig. 2. Sequence of the reshaped ��::.
`V H domain. The reshaped domain .....,.
`.._. _ •tart•
`HuV HLYS is based on the <:M:lW1C-11GMJtJ•toACart:Al•rGAGl'<TCllCAG1==11CAG:rrM::TGMJCAC11CAGG11a:TC
`framework regions with the same (or simi­
`lar) contacts as arc used with the Dl.3
`HuVHNP gene (4), with the frame.
`nating with the hypavariablc re­ (M G ! S C
`framework regions. However, we also iden­
`work regions of human NEW alter­
`tified some potential problems. For exam­
`gions of mouse 01.3 (7). Three ACM>T�MATGGGtGACMTGACATCCM:T1'TGCCT1'
`ple, in CDR..l of Dl.3, Tyr-32 makes a van
`oligonuclcotides I, II, and ill were
`der Waals contact with the Phc·27 in FRI;
`10
`Si!p!!l
`1
`5
`(Q V 8 S)Q V Q L Q £ S G P G L V R
`used to replace each of the Bl-8
`in NEW this phenylalanine is replaced by
`CDR.s with those from 01.3. Each
`15
`25
`30 ---2!!!!_
`serine, and this contact is therefore lost
`oligonµclcotidc has 12 nucleotides
`20
`V S G S T r S Oi.....l-lLJ =
`P S Q T L S L T C T
`at the 5' end and 12 nucleotides at
`when the ·Dl.3 CDRs arc mounted on
`the 3' end which ·arc complcmcn­
`tary to the NEW framework re- crlJ 11 v a 0 4� P G R G 4� e 11 I G 15a � c I
`NEW framework regions. Furthcnnore, in
`-UQTT'nlllGGT
`gions, whereas the central portion GTMAC
`antibody D 1.3 the loop includ.ir).g the end of
`encodes COR.s 1, 2, or 3 of the I P s� � P 6� "pr;-: IL t; s I a v T 7� L v o
`FRl bulges out from the surface, whereas in
`NEW (and other solved structures), it is
`01.3 antibody. The primers arc -�-----'----.---�--- P��r II
`L�
`pinned bade to the surface. Despite these
`c.omplcmcn� to the marked por•
`T 7� K N Q r 8� L a8� 8s cs v T 8! A D T A v
`tions of the coding sttand of the
`problcnu, the reshaped antibody is able to
`reshaped domain. Three rounds of
`io5
`mutagcncsis were used, transfcct- go
`•.5 Cll!J 100.
`bind lysozymc.
`x
`G s L
`Y c A R I ii a jCx a i p x I 11 G o
`mg. irito the EuheridtitJ coli .....,;"
`�.,....,
`It � conceivable that the several contacts
`TATTA���
`priller m
`BMH71-18mutL, plating on a
`made by the FRl/CDRl loop to lysozymc
`110
`v T v s s -1
`lawn of BMH71-18, and probing
`infected colonies with the mutagen- �. · · · · .3'
`make only a small contribution to the overall
`cncrgctics of binding. Problems in this re­
`ic primers as in (8). In washing the Mutagenic priMre are I 5' C'tG TCT CAC CCA GTT 'tAC NX ATA GCC
`filters, it was necessary to gradually � � � � � i:x:5�MTTGTCArCAC'tCCTCTM�CCATTTCTC�'�I�'� �
`gion would then have only a slight dfcct on
`increase the wash tcmpctaturc from NX CCA GTA GTC MG CC't ATA A'tC 'tC't ere TCT TGC ACA ArA 3'
`the affinity. It is also possible that the shape
`65° to 75°C to distinguish the mutant from wild-type doocs. The yield of mutants at each step was
`of the binding site could adjust to the
`about 5%.
`binding of antigen (18) and therefore that
`small errors in assembling the CD Rs on a
`new framework might be "self-correcting"
`when the antigen binds. The energetic price
`for such putative self-correcting induced fit
`would be low, given the small los.s in bind­
`ing affinity suggested by the competition
`data of Fig. 3. It is not possible to deduce
`the exact change in affinity from the compe­
`tition data, as the competition may only
`rcfk:ct the relative on-rates. However, even
`if the difference in affinities were tcnfoJd,
`this would correspond to no more than the
`lo.ss of a single hydrogen bond interaction
`(19). The shape of the antigenic cpitope
`could also adjust to the binding of antibody
`(20), although no structural change in lyso­
`zymc was detected in its complex with the
`mouse D 1.3 antibody (is). Furthermore, the
`whole surface of interaction could reorien­
`tate slightly, perhaps by � on side
`chains (21), to make a set of new contacts.
`This is an attractive model fur large surfaces
`making multiple interactions. Thus sclf-cor­
`rcction could occur at' the level of the anti­
`body, the antigen, or the complex.
`In conclusion, the structure of antibodies
`may incorporate fC1llU.rCS that ass� the
`grafting of hypcrvariablc regions. The criti­
`cal features arc that framework contacts arc
`largely COQSCrved between Vu and V L qo­
`mains, between the sheets within a domain,
`and to the CDR 'loops. In addition, the
`shape of the antigen binding s.itc or the
`antigenic determinant (or both) might be
`able to adjust to each other in an induced-fit
`mechanism.
`JlBFBJlBNCBS AND NOTBS
`
`Fig. 3. Competition of reshaped and cnginccrcd
`mouse antibody for lysozymc. The wells of a
`microritcr plate were ooattd with lysozymc at a
`concentration of 100 µ.g/ml in pbosphatc-buiF­
`crcd saline (PBS), and remaining binding sites on
`the plastic were blcxktd with 1 % bovine scnim
`albumin (in PBS). The binding of 12'J-labclcd
`mouse �tibody 01.3 (specific activity, >0.6 "'Ci
`per micrograms ofOl.3) to lysozymc was com­
`peted with in=ing coocmttari� of(•) cngi­
`nccrcd antibody (MoVHLYS-MolgGl, K) or (0)
`reshaped antibody (HuVHLYS-HulgG2, K), and
`the data presented arc typical of two independent
`cxpcrimcnts. Antibody 01.3 and reassembled
`antibody D 1.3 competed with the Jabclcd 01.3 as
`effectively as did the cnginccrcd antibody 01.3.
`Radioiodination was carried out by the chlora·
`mine T method (22).
`
`the binding oflysozymc to thes<: antibodies.
`The binding to both antibodies was tight,
`with 2 mol oflysozymc molecules binding l
`mol of antibody. Our results indicate that
`the affinity is better than 5 nM, although
`previous work had indicated a val� of 20 to
`40 nM (6, 13). The antibodies were then
`compared in a competition assay fur lyso­
`zymc; 125I-labeled mouse antibody Dl.3
`competing with mouse antibody or with the
`reshaped antibody for binding to lysozymc.
`The reshaped antibody oompctcs, albeit less
`effectively (about tenfold less) than the
`mouse antibody (Fig. 3). Nevertheless, the
`grafting of hypcrvariablc regions from
`mouse to human framework regions is suffi­
`cient to tranSfcr the lysozymc-binding site,
`an extensive surface of interaction, despite
`the 31 of87 residues that differ between the
`hcavy-chaµt framework regions.
`The results confirm that the hypcrvariablc
`2S MAii.CH 1988
`
`0.1
`
`10 20
`1
`Competing etlbody (µgtml)
`
`regions (mainly loops) fashion the antigcn­
`binding site and that the more conserved
`regions (mainly �-sheet) form a structural
`framework (14). However, the result is re­
`mark.able given the several asswnptions un­
`d«lying the building of an antigen-binding
`site on a new framework regi�ly.
`( i) the V ti and V L domains pack together in
`the same way (15), (ii) the two �-sheets
`within each variable domain pack together
`in the same way (16), (iii) die antigen
`usually binds to the hypcrvariablc regions,
`and (iv) the hypcrvariablc regions arc sup­
`ported by tfic framcwQrk regions in a similar
`way (17). In particular, the hypcrvariablc
`loops arc not stand-alone struc:turcs, but
`make cxtmSivc contacts to the framework
`and to other hypcrvariablc regions. Since
`the aystallographic structures of both par­
`ent antibodies arc known, we could con1irm
`that in the reshaped antibody, the Dl.3
`
`l. lL A. Miller, A. IL Oscroff, P. T. Stnae, IL Levy,
`BU!otl 62, 988 (1983).
`
`llBPOllTS IS35
`
`2 of 3
`
`BI Exhibit 1068
`
`

`

`to 2 mg of antibody (in I ml) was dialy7.cd over­
`2. 0. A. Carson and B. 0. Freimark, Aih. !MMUnol.
`16. A. M. Lcsk and C. J. Chothia, ihi4. 160, 325
`night at 4"C against O.SM tris, pH 8.0, (ii) intcr­
`38, 275 (1986).
`chain disul6dcs were reduced with O. lM 2-mcrcap­
`(1982).
`3. G. L. Bouliannc, N. Hozum.i, M. J. Shulman,
`17. C. Chothia and A. M. Lcsk., ihi4. 196, 901 (1987).
`S. L. Morrison,
`fi:ec sulfhydryls were alkylated with O. lSM iodoacc­
`PeptiAt:s llS Anli!Jms, R. Porter and J. Whelan, Eds.
`
`tocthanol for l hour at room temperature, (iii) the
`18. A. B. Edmundson and K. R. Ely, in Synthetic
`N11t11n (Lon;lqn) 312, 643 (1984);
`M. J. Johnson, L. A. Herunbcrg, V. T. Oi, Prw;.
`tam.ide for 15 minutes at room temperature, (iv) the
`Nllll. Auut. Sei. U.S.A. 81, 6851 (1984).
`p. 107; E. D. Gett.off a
`(Wiley, New Yort, 1986),
`heavy and light chains were fractionated on a Du­
`td., Seiena 235, 1191 (1987).
`4. M. S. Neuberger a td., N11t11n (Lon;lon) 314, 268
`pont Zorbax G250 colwnn in SM guan.idinc hydro­
`(1985).
`19. A. R. Fcnht a td., NlltUre (Lon;lon) 314, 235
`and 20 mM sodium phosphate, pH 8.0.
`5. P. T. Jones, P. H. Dcac, J. Foote, M. S. Neuberger,
`The D 1.3 1< light chain was rcfractionated on high­
`chloride
`(1985).
`G. Winter, ibid. 321, 522 (1986).
`20. P. M. Colman a td., ibid. 326, 358 (1987).
`liquid chromatography (HPLC) and
`6. A.G. Amit, R. A. Mariuxza, S. E. V. Phillips, R. J.
`21. C. Chothia, A. M. Lcsk., G. G. Dodson, D. C.
`an aliquot was chcded on analytical HPLC to
`pcrformancc
`Seiena 233, 747 (1986).
`Hodgkin, ibill. 302, 500 (1983).
`Poljak,
`22. L. Hudson and F. C. Hay, Praawu Immunology
`en.sure no contamination with the D 1.3 heavy chain,
`7. M. E. Vcrhocycn, C. Berck, G. Winter, in prqwa­
`(v) appropriate heavy chains were mixed with equal
`Oxford, 1980), p. 240.
`amounts of D l . 3 K tight chain and dialy7.cd against
`(Blackwell,
`tion.
`23. We thank R. Poljak, A. Amit, S. Phillips, and R. A.
`8. P. Carter, H. Bcdoucllc, G. Winter, Nlldeie Aeids
`Ra. 13, 4431
`O.lM trisQ, pH 7.4, at 4°C for 2 days,
`Marriuzza for the crystallographic coordinates of
`rcasscrn bled antibody was purified on a protein A­
`(1985).
`and (vi)
`9. J. Ellison and L. Hood, J>rw;. Nllll.Aaut. &i. U.S.A.
`Dl.3 lysozymc
`
`complex; R. A. Marriuzza, J. Foote,
`79, 1984 (1982).
`C. Chothia, and A. Lcsk for discussions and com­
`eluted at pH 6, and the reshaped
`Scpha.rosc colwnn. The rcasscrn bled mouse anti­
`10. H. Potter, L. Weir, P. Leder, ibid. 81, 7161 (1984).
`ments on the paper; J. FOO!C for use of the Buorcs­
`at pH 4. From 1 mg of anti­
`cc:ncc titration method and associated computer
`body (MolgGl)
`11. V. T. Oi, S. L. Morrison, L.A. Hcncnbcrg, P.
`body, the yield of rcasscm bled antibody was Jess
`antibody (HulgG2)
`Berg, ibid. 80, 825 (1983).
`in advance
`12. Clones secreting antibody were grown in 1-liter
`programs
`
`of publication; and J. Jarvis, P.
`T. Jones, and R. Pannell
`than 20%.
`rollcr bottles, and antibody was purified on protein
`for technical assistance.
`M.V. is a senior rcscarch
`
`13. M. Harper, F. Lema, G. Boulot, R. J. Poljak, Mal.
`assistant of the Belgian
`1-no1. 24, 97 (1987).
`was exchanged
`National Fund for Scientific Research and a fcllow
`The mouse ).. light chain of antibodies
`A�a.rosc-
`14. E. A. Kabat, Aih. p,,,fein Chmr. 32, 1 (1978).
`secreted from the J558L myclonu
`for the D 1.3 K light chain in vitro. For the D l. 3
`of the Commission
`of the European Communities.
`J. Novotny, R. Bruccoleri, M. Karplus,
`15. C. Chothia,
`antibody and the two recombinant antibodies (i) 1
`J. Mol. Biol. 186, 651 (1985).
`
`29 September 1987; accepted 8 February 1988
`
`Peroxisomal Membrane Ghosts in Zellweger
`Syndrome-Aberrant Organelle Assembly
`M. J. SANTos, T. IMANAKA, H. SHIO, G. M. SMALL, P. B. l...AzAR.ow
`Pcroxisomcs are apparently missing in Zellwegcr syndrome; nevertheless, some of the
`integral membrane proteins of the organelle are prcscot. Their distribution was
`studied by immunoftuoresccncc microscopy. In control fibroblasts., pcroxisomcs
`appeared as small dots. In Zellwegcr fibroblasts., the pcroxisomal membrane proteins
`were located in unusual emp ty membrane structures of larger size. These results
`suggest that the primary defect in this di.scasc may be in the mcchan.ism for import of
`matrix proteins.
`ZELLWEGER SYNDROME IS A DISEASE
`in which an entire organelle, the
`peroxisome, appears to be missing,
`as first recognized by Goldfischer et ti/.. (1).
`The peroxisome is nearly ubiquitous in eu­
`karyotic cells and functions in fatty acid 13-
`oxidation, plasmalogen biosynthesis, cellular
`respiration (H20rfonning), gluconeogene­
`sis, bile acid synthesis, and purine catabo­
`lism (2). This human genetic disorder, char­
`acterized by profound neurological impair­
`ment, metabolic disrurbance, and neonatal
`death, has taught us much about peroxi­
`some function (3) and promises to teach us
`more about peroxisome assembly. Some
`peroxisomal proteins are synthesized nor­
`mally in Zellweger syndrome (4), but they
`are not assembled into peroxisomes. Many
`of these proteins are rapidly degraded, with
`the result that important soluble matrix en­
`zymes (catalyzing j3-oxidation) ( 4) and
`membrane-bound enzymes (catalyzing the
`initial steps in plasmalogen biosynthesis) ( 4,
`5) are missing or seriously deficient, thus
`The Rockcfcllcr University, New York, NY 10021.
`
`causing severe metabolic abnormalities (3).
`On the other hand, some peroxisomal en­
`zymes (for example, catalase) are present in
`normal amounts in Zellweger cells but are
`located free in the cytosol (4, 6, 7).
`These defects correlate with the known
`facts of peroxisome biogenesis. All peroxi­
`somal proteins investigated thus far are syn­
`thesized on free polyribosomes and are as­
`sembled posttranslationally into preexisting
`peroxisomes ( 8). If the organelle is missing
`(1), newly made proteins will just diffuse
`through the cytosol, unable to enter a per­
`oxisome. Under these circwnstances, it is
`not surprising that many are degraded.
`The autosomal recessive genetics of Zell­
`weger syndrome indicates that a single mu­
`tation is responsible for the defects. One
`explanation would be that the mutation
`prevents the assembly of the peroxisomal
`membrane. In this case peroxisomal integral
`membrane proteins (PxlMPs) (9) should be
`absent, since those that have been studied
`are made on free polyribosomes (10-12)
`and are unlikely to be stable if not integrated
`into a membrane. Two other possibilities
`
`were suggested by the unexpected finding
`that several PxIMPs are present in normal
`amounts in Zellweger liver (13). The peroxi­
`somal membranes could be assembled in
`Zellweger syndrome, but the import of ma­
`trix proteins is defective. 1bis would result
`in empty (or nearly empty) membrane
`ghosts, which would not be recogniz.able by
`electron microscopy or cytochemistry.
`Alternatively, the PxIMPs, in the absence of
`peroxisomes, might have violated the rules
`of protein sorting and inserted into the
`wrong intracellular membrane(s).
`To differentiate among these possibilities,
`we analyzed fibroblasts with a polyspecific
`antiserum against rat liver PxIMPs (polyspe­
`cific anti-PxIMPs) (12) (Fig. lA, lane 1).
`This serum detected three human PxIMPs
`with masses of approximately 140, 69, and
`53 kD in control cells (Fig. lA, lanes 2 to
`4). These PxIMPs were also present in Zell­
`weger fibroblasts (lane 5), and they coscdi­
`mented in equilibriwn density centrifuga­
`tion of Z.Cllweger fibroblast homogenates.
`However, the PxIMPs were found ac an
`abnormally low density of 1.10 g!cm3 (in­
`stead of the usual fibroblast peroxisome
`density of 1.17 g!cm3) [lanes 6 and 7; further
`details in (14)]- Several other cell mem­
`branes also sedimented in the low-density
`region of the gradient where the PxIMPs
`were found (14). Therefore, the fraction­
`ation data arc consistent with the possibility
`that the PxIM.Ps are present in aberrant
`peroxisomal membrane ghosts, but they do
`not exclude erroneous localization in lyso­
`somes, the endoplasmic reticulum, or some
`other low-density organelle. lmmW10f1uo­
`rescence studies were carried out to resolve
`this uncertainty.
`
`SCIENCE, VOL. 239
`
`3 of 3
`
`BI Exhibit 1068
`
`