Proc. Natl. Acad. Sci. USAVol. 85, pp. 2320-2323, April 1988Medical Sciences

Isolation and sequencing of a cDNA coding for the human DF3breast carcinoma-associated antigenJ. SIDDIQUI, M. ABE, D. HAYES, E. SHANI, E. YUNIS, AND D. KUFE*Laboratory of Clinical Pharmacology and Division of Immunogenetics, Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA 02115

Communicated by Emil R. Unanue, December 9, 1987

ABSTRACT The murine monoclonal antibody (mAb) DF3reacts with a high molecular weight glycoprotein detectable inhuman breast carcinomas. DF3 antigen expression correlateswith human breast tumor differentiation, and the detection ofa cross-reactive species in human milk has suggested that thisantigen might be useful as a marker of differentiated mam-mary epithelium. To further characterize DF3 antigen expres-sion, we have isolated a cDNA clone from a Agtll library byscreening with mAb DF3. The results demonstrate that this309-base-pair cDNA, designated pDF9.3, codes for the DF3epitope. Southern blot analyses of EcoRI-digested DNAs fromsix human tumor cell lines with 32P-labeled pDF9.3 haverevealed a restriction fragment length polymorphism. Varia-tions in size of the alleles detected by pDF9.3 were alsoidentified in Pst I, but not in HindIII, DNA digests. Further-more, hybridization of 32P-labeled pDF9.3 with total cellularRNA from each of these cell lines demonstrated either one ortwo transcripts that varied from 4.1 to 7.1 kilobases in size.The presence of differently sized transcripts detected bypDF9.3 was also found to correspond with the polymorphicexpression of DF3 glycoproteins. Nucleotide sequence analysisof pDF9.3 has revealed a highly conserved (G+C)-rich 60-base-pair tandem repeat. These rmdings suggest that thevariation in size of alleles coding for the polymorphic DF3glycoprotein may represent different numbers of repeats.

A human breast carcinoma-associated antigen has beenidentified by using a murine monoclonal antibody (mAb),designated DF3. mAb DF3 was prepared against a mem-brane-enriched fraction of a human breast carcinoma meta-static to liver (1). DF3 antigen is expressed on the apicalborders of secretory mammary epithelial cells and in thecytosol of less differentiated malignant cells (1). DF3 antigenexpression also correlates with the degree of breast tumordifferentiation and estrogen receptor status (2). These find-ings and the detection of DF3 antigen in human milk (3) havesuggested that mAb DF3 reacts with a differentiation antigenexpressed in breast carcinoma cells.The DF3 antigen has been characterized as a high molec-

ular weight mucin-like glycoprotein (4, 5). DF3 antigen inhuman MCF-7 breast carcinoma cells consists of two distinctglycoproteins with molecular weights of 330,000 and 450,000(4, 5). Moreover, previous studies have demonstrated thatDF3 antigen circulates at elevated levels in the plasma ofpatients with breast cancer (6). The circulating mAb DF3-reactive antigens also have molecular weights ranging fromapproximately 300,000 to 450,000. However, the electropho-retic mobility patterns for circulating DF3 antigen differamong individuals (6). Subsequent studies in family mem-bers have demonstrated that the electrophoretic heterogene-ity of plasma DF3 antigen is determined by codominantexpression of multiple alleles at a single locus (7).

The present studies were performed to further examinegenetic mechanisms responsible for the heterogeneity ofDF3 antigen expression. We describe the isolation of apartial cDNA clone,t designated pDF9.3, coding for the DF3antigen. This cDNA clone encompasses 309 nucleotides ofDF3 mRNA and has tandemly repeated sequences. By usingthis clone, we also demonstrate that the heterogeneity ofDF3 antigen is related to size variations in DF3 alleles andDF3 transcripts.

MATERIALS AND METHODSLibrary Screening. An oligo(dT)-primed cDNA library

prepared from human MCF-7 breast carcinoma cells in Agtllwas kindly provided by P. Chambon (Institut de ChimieBiologique, Strasbourg, France) (8). Immunologic screen-ing of the Agtll library was performed as described (9) byusing affinity-purified mAb DF3 (0.25 ,ug/ml) and anti-mouse IgG conjugated with alkaline phosphatase (PromegaBiotec, Madison, WI). Positive plaques were isolated andthe phage was further purified to homogeneity by repeatedantibody screening. DNA was isolated from mAb DF3-positive recombinant phage, digested with EcoRI, and elec-trophoresed in 1.2% agarose gels containing ethidium bro-mide to determine the size of the insert.

Analysis of Lysogens for Fusion Protein. Lysogenization ofEscherichia coli Y1089 with phage and induction of fusionprotein with isopropyl ,B-D-thiogalactoside (IPTG) were per-formed as described (9, 10). The lysate of IPTG-inducedlysogen was subjected to electrophoresis in NaDodSO4/7.5% polyacrylamide gels (11) and transferred to nitrocellu-lose filters for immunoscreening (12).

Southern and RNA Transfer Blot Analyses. The humanbreast carcinoma cell lines (BT-20, T47D, MCF-7, ZR-75-1),an ovarian carcinoma cell line (OV-D), and the HL-60promyelocytic leukemia cell line were maintained in expo-nential phase (13-15). High molecular weight DNA and totalcellular RNA were isolated by the guanidine isothiocya-nate/cesium chloride method (16). The DNA was digestedwith EcoRI, Pst I, or HindIII. The DNA fragments wereseparated by electrophoresis in 0.6% agarose gels and thentransferred to nylon membranes. The prehybridization andhybridization conditions were as described in the Zeta-Probemanual (Bio-Rad). The purified RNA (20 ,ug) was analyzedby electrophoresis in 1.5% agarose/formaldehyde gels; thiswas followed by transfer to nitrocellulose paper. The hydri-dization conditions were as described (15). The pDF9.3cDNA probe was labeled with [32P]dCTP (Amersham) by therandom primer method (17) to a specific activity of -101ocpm/,ug of DNA.

Abbreviation: mAb, monoclonal antibody(ies).*To whom reprint requests should be addressed.tThis sequence is being deposited in the EMBL/GenBank data base(Bolt, Beranek, and Newman Laboratories, Cambridge, MA, andEur. Mol. Biol. Lab., Heidelberg) (accession no. J03651).

2320

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 27

, 202

2

Proc. Natl. Acad. Sci. USA 85 (1988) 2321

Immunoblot Analysis. Cells were suspended in phosphate-buffered saline [0.15 M NaCl/0.01 M sodium phosphate, pH7.4 (PBS)], 0.2 mM phenylmethylsulfonyl fluoride, andaprotinin (0.015 trypsin inhibitor unit/ml). The suspensionswere sonicated and protein concentration was determined bythe Bio-Rad protein assay. The protein samples (100 gg)were analyzed by electrophoresis in NaDodSO4/3-15% gra-dient polyacrylamide gels and transferred to nitrocellulosepaper (18). The nitrocellulose filters were washed with 5%bovine serum albumin in PBS for 1 hr at room temperatureand incubated with mAb DF3 (0.25 ug/ml) for 2 hr, rabbitanti-mouse immunoglobulin for 1 hr, and then 1251I-labeledprotein A for 2 hr. The filters were washed five times, dried,and exposed to x-ray film.

Nucleotide Sequence Analysis. The 309-base-pair (bp)pDF9.3 cDNA insert was subcloned into the EcoRI site of E.coli phage M13mp8 and M13mp9. The DNA sequence wasdetermined by sequencing both strands by means of thedideoxy chain-termination method (19) using Klenow frag-ment DNA polymerase I (New England Biolabs) and [a-35S]dCTP (Amersham).

RESULTS

Isolation and Characterization of cDNA Clones Coding forDF3 Antigen. mAb DF3 was used to screen the Agtll libraryprepared from MCF-7 cells. Screening of 800,000 plaquesyielded three positive clones that were further purified byrepeated antibody screenings. Physical mapping showed thateach of these recombinant clones contained inserts of similarsize and that they had similar restriction maps (data notshown). One clone, designated pDF9.3, was characterizedfurther. A B-galactosidase fusion protein was prepared byinfecting E. coli Y1089 with pDF9.3 and then analyzed byimmunoblotting. The Agtll lysogen produced a protein cor-responding in molecular weight and antigenicity to P3-galactosidase (Fig. 1, lane 1). mAb DF3 was unreactive with,f-galactosidase and other antigens present in the bacteriallysate (data not shown). In contrast, the recombinantpDF9.3 lysogen produced a fusion protein with an estimatedmolecular weight of 126,000 that reacted with mAb DF3(Fig. 1, lane 2) and the anti-,B-galactosidase antibody (datanot shown).

Competition assays were also performed to further con-firm that the epitope expressed by pDF9.3 shares homologywith that identified by mAb DF3 on the DF3 glycoprotein.Thus, mAb DF3 was preincubated with purified DF3 antigen(13) before immunoblot analysis of the pDF9.3 fusion pro-tein. Preincubation of mAb DF3 with increasing amounts ofpurified DF3 antigen progressively inhibited reactivity of theantibody with the fusion protein (Fig. 1, lanes 3-5). Thisfinding indicates that the epitope on the fusion proteinoriginates from the same reading frame that codes for theDF3 epitope.

Southern Blot Analysis of Genomic DNA. Identification ofthe cDNA was further studied by Southern blot hybridiza-tions using 32P-labeled pDF9.3 prepared by subcloning the309-bp insert into the EcoRI site of pUC8. Southern blotanalyses of genomic DNAs from the human tumor cell linesdigested with EcoRI, Pst I, and HindIII are shown in Fig. 2.Hybridization of the 309-bp cDNA with the EcoRI and Pst IDNA digests revealed restriction fragment length poly-morphisms. The EcoRI digest yielded two fragments rangingfrom 7 to 12 kilobases (kb) in size for DNAs from each of thecell lines except BT-20. Similar findings were obtained withthe Pst I fragments, which ranged in size from 3.5 to 6 kb.The single EcoRI and Pst I restriction fragments obtainedwith BT-20 DNA indicate the presence of two alleles ofidentical size or only a single allele. In contrast to theseresults, digestion of each of the DNA preparations withHindIII revealed only a single fragment of 23 kb. This findingcorresponds to the absence of a HindIl restriction site in thealleles identified by pDF9.3.RNA Transfer and Immunoblot Analyses of DF3 Expres-

sion. Total cellular RNA was prepared from each of thehuman tumor cell lines and monitored by RNA transfer blotanalysis for transcripts that hybridized to the pDF9.3 probe.A single 4.7-kb mRNA was detectable in BT-20 cells (Fig.3A). In contrast, cell lines derived from the other breast andovarian carcinomas expressed two transcripts that ranged insize from approximately 4.1 to 7.1 kb (Fig. 3A). Further-more, no hybridization was detectable with RNA fromHL-60 cells (Fig. 3A).These findings by RNA transfer blot analysis were com-

pared to those obtained by immunoblotting with mAb DF3and extracts prepared from each of the cell lines. The resultsindicate concordance in patterns of expression at the RNA

1 2 3 4 5A

Q~O 0rN C

U- a

m 2 NF--C

kb

9.45-

6 .5 -

4.3

B

ON LL0

co

F C- ccm 2 N 0O

.1m, .sR .,

kb

23.1 -

9.46.5 -

4.3 -

2.3

C

o tl-L~n 0N acoLL ,

F- C) cc 't > -1m2EN -OI

Kb .s

23 1- -.....,

9.4 -

6.5 -

4.3

FIG. 1. Immunological identification of pDF9.3-encoded recom-binant antigen. Agtll and pDF9.3 recombinant phage were used tolysogenize E. coli Y1089. Protein extracts were prepared from thelysogens and 10 ,ul of the bacterial lysates was electrophoresed inNaDodSO4/7.5% polyacrylamide gels, electroblotted onto nitrocel-lulose, and incubated with anti-p-galactosidase antibody (lane 1) ormAb DF3 (lane 2). mAb DF3 was also preincubated with 10 ng (lane3), 100 ng (lane 4), and 500 ng (lane 5) of purified DF3 antigen priorto probing the filters. Antibody binding was detected by an enzyme-linked immunoabsorbant assay by using nitroblue tetrazolium assubstrate. Several apparent proteolytic breakdown products werenoted when using the anti-f3-galactosidase antibody.

2.3 2.3-

FIG. 2. Southern blot analysis of genomic DNA with the pDF9.3probe. DNAs (20 ,ug) from human tumor cell lines were digested tocompletion with EcoRI (A), Pst 1 (B), and HindIII (C) and electro-phoresed in 0.6% agarose gels. The gels were denatured and theDNA fragments were transferred to nylon filters. The filters were

hybridized with the 32P-labeled pDF9.3 cDNA insert. The filterswere then washed and exposed to x-ray film.

Mr X 10 3

116-93-

Medical Sciences: Siddiqui et al.

1-1

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 27

, 202

2

2322 Medical Sciences: Siddiqui et al.

Ar-L() 0

N IL IaF

N:t >

01m 2 N F- o I

.028S-

18S-

B

Mr X 10-3

200-

1 16-93-

o N U)

L N

m 2 N 1

C]> -I

66-

FIG. 3. RNA transfer blot analysis with pDF9.3 and immunoblot-ting with mAb DF3. (A) Total cellular RNA (20 ,g) from humantumor cell lines was electrophoresed in a 1.5% agarose/formaldehydegel, transferred to nitrocellulose, and hybridized with the 32P-labeledpDF9.3 cDNA insert. (B) Extracts of the human tumor cells wereanalyzed by NaDodSO4/3-15% polyacrylamide gel electrophoresis,immunoblotted with mAb DF3, and then allowed to react with rabbitanti-mouse immunoglobulin and '25I-labeled protein A.

and protein levels (Fig. 3B). Thus, BT-20 cells expressed asingle transcript and a single DF3 glycoprotein, whereas theother epithelial cell lines expressed two transcripts and twoDF3 antigens. Moreover, HL-60 cells had no detectableRNA and no detectable mAb DF3-reactive species. Thesefindings further suggested that the transcripts detected byRNA transfer blot analysis code for the DF3 core proteinand that the size of these transcripts determines the size ofthe mAb DF3-reactive glycoproteins.

Nucleotide Sequence of pDF9.3. The reactivity of the fusionprotein with mAb DF3 indicated that the cDNA insertcontained an open reading frame that codes for the DF3epitope. The nucleotide sequence of pDF9.3 was found to behighly rich (85%) in G-C base pairs (Fig. 4). Moreover, thesequence was found to consist entirely of 60-bp tandemrepeats. These repeats were nearly identical with the excep-tion of some transversions (Fig. 4). Furthermore, compari-son of the pDF9.3 sequences with that of all genes withknown sequences failed to reveal any significant homology.

DISCUSSIONWe have previously demonstrated that DF3 antigen inhuman breast tumors and milk is comprised of mucin-like

glycoproteins with molecular weights ranging from 300,000to >450,000 (4, 5). DF3 antigenicity was found to be sensi-tive to neuraminidase and proteases (4, 5). These resultssuggested that sialyl oligosaccharides on a peptide backboneare required for DF3 antigenicity. In the present study, mAbDF3-positive plaques were isolated by using a Agtll cDNAlibrary prepared from human MCF-7 breast carcinoma cells.The MCF-7 cells have been shown to express DF3 antigen(13). One of the positive A clones (pDF9.3) was furtherpurified and found to produce a B3-galactosidase fusionprotein that specifically reacted with mAb DF3. The reac-tivity of mAb DF3 with plaques from this expression libraryand the fusion protein suggests that this antibody reacts withthe core protein of DF3 antigen. However, DF3 antigenicityhas also been shown to be sensitive to neuraminidase (4, 5).Thus, mAb DF3 binding to the protein may be enhanced bythe presence of glycosidic linkages.Although patients with breast cancer and certain other

carcinomas have higher levels of circulating DF3 antigen,the electrophoretic mobilities of the mAb DF3-reactivespecies are similar to those in normal subjects (6, 20).Indeed, more recent results have indicated that the variationin electrophoretic mobility of circulating DF3 antigen amongfamily members is related to a genetically determined poly-morphism (7). The present findings support this geneticpolymorphism. Thus, considerable fragment size variationwas observed after hybridization of the pDF9.3 probe toEcoRI and Pst I restriction digests of DNA from differentcell lines. The EcoRI restriction fragments varied from 7 to12 kb in size, and the different cells had only one or twobands. Furthermore, the Pst I fragments varied from 3.5 to 6kb and each DNA preparation similarly yielded one or twobands. In contrast, hybridization of pDF9.3 probe to HindIIIDNA digests revealed only one 23-kb band and indicatedthat this restriction enzyme has digestion sites outside theregion identified by this probe.The variation in allele size identified with pDF9.3 corre-

sponded with the presence of differently sized transcripts.Thus, cells with two restriction fragments in the EcoRI orPst I DNA digests had two differently sized mRNAs. Incontrast, BT-20 cells had only one detectable restrictionfragment in these DNA digests and expressed only onetranscript. This relationship also extended to the variation inelectrophoretic mobilities of DF3 antigen. BT-20 cells ex-pressed a single mAb DF3-reactive species, whereas theother epithelial tumor cells expressed two DF3 antigens.Moreover, HL-60 cells had no detectable transcripts and nodetectable DF3 antigen. Taken together, these findings sup-port our previous findings that the heterogeneity of DF3antigen production is controlled by multiple alleles at a singlelocus expressed in an autosomal codominant fashion (7).The nucleotide sequence analysis of pDF9.3 provides a

possible explanation for the variability in restriction frag-ment size and the polymorphic patterns of DF3 expression.In this regard, we have identified a 309-bp cDNA clone thatconsists of multiple tandem repeats. These repeats are(G +C)-rich and encompass 60 bp. Variation in the size ofthe DF3 alleles could thus be due to differences in the

CGCACGGCTG GGGGGGCGGT GGAGCCCGGG GCCGGCCTGC TCTCCGGGGC CGAGGTGACA 60

.CGTG .....C. C................................... .......... 120

......... G ......... 18

.............G ................................... .......... 240

.... ... . ..... ... .....300

FIG. 4. Nucleotide sequence of the pDF9.3 cDNA insert.

Proc. Natl. Acad Sci. USA 85 (1988)

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 27

, 202

2

Proc. Natl. Acad. Sci. USA 85 (1988) 2323

number of these repeats and occur as a result of unequalcrossing-over events. The presence of closely related re-peats may also explain the finding that mAb DF3 binds totwo or more epitopes on the same DF3 molecule (6). Thetotal number of these repeats in the full-length cDNA,however, requires further investigation.

Similar variable tandem repeats have been reported forother genes, including those coding for carcinoembryonicantigen (21), insulin (22), a- and ,3-globulin (23, 24), Epstein-Barr virus (25), c-Ha-ras (26), and a hypervariable minisatel-lite family (27). Furthermore, the human complement recep-tor (CR1) gene consists of homologous repeats =1.6 kb insize (28). Allelic variants of CR1 differ by 1.6 kb and alsocorrelate with variations in size of the CR1 transcripts andproducts (28). The lengths of most internal repeats, how-ever, range between 120 and 300 bp (29). Moreover, homol-ogy of the internal repeats for many vertebrate proteinsranges between only 20% and 50% (29). In contrast, theinternal repeats identified in the present study exhibit aparticularly high degree of homology. This finding couldsuggest that the DF3 gene evolved more recently by dupli-cation of a primordial gene or by exon shuffling.We have recently demonstrated that the DF3 antigen is a

member of a family of related, but not identical, highmolecular weight tumor-associated antigens (13). The DF3glycoprotein is also detected by another mAb, designatedCal, that reacts with a wide range of human tumors (30, 31).The binding site for mAb Cal on the DF3 glycoprotein,however, is distinct from that defined by mAb DF3 (13).Recent studies have described a genetic polymorphism ofthe Cal antigen in human urine (32). Moreover, the locuscoding for these urinary mucins (PUMs, from peanut lectinbinding urinary mucins) is a hypervariable minisatellite ofhuman DNA (33). After completion of the present work, areport appeared that describes the isolation of a partialcDNA clone (pMUC10) that codes in part for PUM (33). TheEcoRI restriction fragments of genomic DNA identified bythe pMUC10 probe are similar to those found in the presentstudy (34). However, the nucleotide sequence of thepMUC10 cDNA has not been published and therefore it isnot possible to directly compare that clone with the pDF9.3cDNA isolated in the present study.

This investigation was supported by Public Health Service GrantsCA-38869 and CA-20531 awarded by the National Institutes ofHealth, Department of Health and Human Services, and by aBurroughs Wellcome Clinical Pharmacology Scholar Award (D.K.).

1. Kufe, D., Inghirami, G., Abe, M., Hayes, D., Justi-Wheeler, H.& Schlom, J. (1984) Hybridoma 3, 223-232.

2. Lundy, J., Thor, A., Maenza, R., Schlom, J., Forouhar, F.,Testa, M. & Kufe, D. (1985) Breast Cancer Res. Treat. 5,269-276.

3. Abe, M. & Kufe, D. (1984) Cancer Res. 44, 4574-4577.

4. Sekine, H., Ohno, T. & Kufe, D. (1985) J. Immunol. 135,3610-3615.

5. Abe, M. & Kufe, D. (1986) J. Cell. Physiol. 126, 126-132.6. Hayes, D., Sekine, H., Ohno, T., Abe, M., Keefe, K. & Kufe,

D. (1985) J. Clin. Invest. 75, 1671-1678.7. Hayes, D., Sekine, H., Marcus, D., Alper, C. & Kufe, D. (1988)

Blood 71, 436-440.8. Walter, P., Green, S., Greene, G. M., Krust, A., Bornert, J. M.,

Heltsch, J. M., Staub, A., Jensen, E., Scrace, G., Waterfield,M. & Chambon, P. (1985) Proc. Natl. Acad. Sci. USA 82,7889-7893.

9. Young, R. A. & Davis, R. W. (1983) Proc. Natl. Acad. Sci.USA 80, 1194-1198.

10. Young, R. A. & Davis, R. W. (1983) Science 222, 778-782.11. Laemmli, U. K. (1970) Nature (London) 227, 680-685.12. Burnette, W. N. (1981) Anal. Biochem. 112, 195-203.13. Abe, M. & Kufe, D. (1987) J. Immunol. 139, 257-261.14. Friedman, E., Hayes, D. & Kufe, D. (1986) Cancer Res. 46,

5189-5194.15. Sariban, E., Mitchell, T. & Kufe, D. (1985) Nature (London)

316, 64-66.16. Davis, L. G., Dibner, M. D. & Battey, J. F. (1986) Basic

Methods in Molecular Biology (Elsevier, New York), pp. 130-135.

17. Feinberg, A. P. & Vogelstein, B. (1983) Anal. Biochem. 132,6-13.

18. Towbin, H., Staehelin, T. & Gordon, J. (1979) Proc. Natl.Acad. Sci. USA 76, 4350-4354.

19. Sanger, F., Nicklen, S. & Coulson, A. (1977) Proc. Natl. Acad.Sci. USA 74, 5463-5467.

20. Sekine, H., Hayes, D., Ohno, T., Keefe, K.A., Schaetal, E.,Bast, R. C., Knapp, R. & Kufe, D. (1985) J. Clin. Oncol. 3,1355-1363.

21. Zimmermann, W., Ortlieb, B., Friedrich, R. & von Kleist, S.(1987) Proc. Natl. Acad. Sci. USA 84, 2960-2964.

22. Bell, Q. I., Selby, M. J. & Rutte, W. J. (1982) Nature (London)295, 31-35.

23. Higgs, D. R., Goodbourn, S. E. Y., Wainscoat, J. S., Clegg, J.B. & Weatherall, D. J. (1981) Nucleic Acids Res. 9, 4213-4224.

24. Spritz, R. A. (1981) Nucleic Acids Res. 9, 5037-5047.25. Speck, S. H., Pfitzner, A. & Strominger, J. (1986) Proc. Natl.

Acad. Sci. USA 83, 9298-9302.26. Capon, D. J., Chen, E. Y., Levison, A. D., Seeburg, P. H. &

Goeddell, D. V. (1983) Nature (London) 302, 33-37.27. Jeffreys, A. J., Wilson, V. & Thein, S. L. (1985) Nature

(London) 314, 67-73.28. Holers, V. M., Chaplin, D. D., Leykam, J. F., Gruner, B. A.,

Kumar, V. & Atkinson, J. P. (1987) Proc. Natl. Acad. Sci. USA84, 2459-2463.

29. Li, W. L. (1983) in Evolution of Genes and Proteins, eds. Nei,M. & Koehn, R. K. (Sinauer, Sunderland, MA), pp. 14-37.

30. Ashall, F., Bramwell, M. & Harris, H. (1982) Lancet ii, 1-6.31. McGee, J., Woods, J., Ashall, F., Bramwell, M. & Harris, H.

(1982) Lancet ii, 7-10.32. Swallow, D., Griffiths, B., Bramwell, M., Wiseman, G. &

Burchell, J. (1986) Dis. Markers 4, 247-254.33. Swallow, D., Gendler, S., Griffiths, B., Corney, G., Taylor-

Papadimitriou, J. & Bramwell, M. (1987) Nature (London) 328,82-84.

34. Gendler, S., Burchell, J., Duhig, T., Lamport, D., White, R.,Parker, M. & Taylor-Papadimitriou, J. (1987) Proc. Natl. Acad.Sci. USA 84, 6060-6064.

Medical Sciences: Siddiqui et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 27

, 202

2