tain other mutations that are responsible for the loop repair defect. We examined loop repair in 13 additional cell extracts (18), including several made from tumor cell lines that exhibit microsatellite instability. All extracts efficiently repaired a heterodu- plex containing a five-base loop. One ex- tract was from an endometrial tumor cell line (HEC59) known to be defective in mismatch repair and to contain a mutation in one hMSH2 allele (4). Thus, whether hMSH2 participates in loop repair remains to be resolved. Candidate gene products that may be re- quired for repair of DNA with loops include putative mismatch repair proteins already identified, such as other MSH or MLH ho- mologs, or proteins yet to be discovered. The latter could include a human homolog of a yeast protein that specifically binds to DNA substrates containing loops of three to nine bases, a protein found even in yeast msh2 and msh3 mutants (19). The possible existence of mutant cell lines defective in some but not all forms of heteroduplex repair is suggested by reports indicating both qualitative and quantitative differences in the stability of various microsatellite alleles in tumor cells and tumor cell lines (5, 6). Identification of extracts defective in repair of loops but not mismatches would reinforce the suggestion that mismatch and loop repair activities have one or more distinct requirements. Note added in proof: The recent demon- stration that purified human MSH2 protein binds to DNA containing loops of up to 14 nucleotides (20) is consistent with a possi- ble role for MSH2 protein in loop repair activity. REFERENCES AND NOTES 1 R. Fishel et al., Cell 75, 1075 (1993); F. S. Leach et al., ibid., p. 1215; C. E. Bronner et al., Nature 368, 258 (1994). 2. N. Papadopoulos et al., Science 263, 1625 (1994). 3. R. Parsons et al., Cell 75, 1227 (1993). 4. A. Umar et al., J. Biol. Chem. 269,14367 (1994). 5. D. Shibata, M. A. Peinado, Y. lonov, S. Malkhosya, M. Perucho, Nature Genet. 6, 278 (1994). 6. R. Wooster et al., ibid., p. 152. 7. Y. lonov, M. A. Peinado, S. Malkhosyan, D. Shibata, M. Perucho, Nature 363, 558 (1993); L. A. Aaltonen et al., Science 260, 812 (1993); S. N. Thibodeau, G. Bren, D. Schaid, ibid., p. 816; A. Undblom, P. Tanner- gard, B. Werelius, M. Nordenskj6ld, Nature Genet. 5, 279 (1993); R. A. Lothe et al., Cancer Res. 53, 5849 (1993); H.-J. Han, A. Yanagisawa, Y. Kato, J.-G. Park, Y. Nakamura, ibid., p. 5087; J. Risinger et al., ibid., p. 5100; P. PeltomAki eta/., ibid., p. 5853; N. M. Mironov et al., Cancer Res. 54, 41 (1994); R. Honchel, K. C. Hailing, D. J. Schaid, M. Pittelkow, S. N. Thibodeau, ibid., p. 1 159; C. J. Yee, N. Roodi, C. S. Vemfier, F. F. Pard, ibid., p. 1641; L. A. Aaltonen et a!., ibid., p. 1645; V. Shridhar, J. Siegfreid, J. Hunt, M. M. Alonso, D. I. Smith, ibid., p. 2084; A. Merlo et al., ibid., p. 2098; M.-G. Rhyu, W.-S. Park, S. J. Mettzer, Oncogene 9, 29 (1994); C. Wu et a!., ibid., p. 991. 8. M. Strand, T. A. Prolla, R. M. Liskay, T. D. Petes, Nature 365, 273(1993). 9. B. 0. Parker and M. G. Marinus, Proc. Nati. Acad. Sci. U.S.A. 89,1730 (1992); M. Carraway and M. G. Marinus, J. Bacteriol. 175, 3972 (1993). 10. J. S. Beckman and J. L. Weber, Genomics 12, 627 (1992). 11. T. A. Kunkel, Biochemistry 29, 8003 (1990). 12. J. Holmes, S. Clark, P. Modrich, Proc. Natl. Acad. Sci. U.S.A. 87, 5837 (1990). 13. D. C. Thomas, J. D. Roberts, T. A. Kunkel, J. Biol. Chem. 266, 3744 (1991). 14. Even with unrepaired DNA heteroduplexes, plaques that are exclusively blue or colorless are obtained upon electroporation of an E. coll strain that is defec- tive in mismatch repair. This is believed to reflect selective strand loss in vivo. 15. W.-h. Fang and P. Modrich, J. Biol. Chem. 268, 11838 (1993). 16. Two explanations for repair of looped heterodu- plexes can be considered other than repair of DNA containing several consecutive unpaired bases. One is that the sequences of the extra bases allow forma- tion of multiple, adjacent shorter loops that are rec- ognized by the same system that recognizes mis- matches. However, except for the substrate contain- ing four extra bases (Fig. 1), the sequences of the looped heteroduplexes used here make this possi- bility unlikely. A second explanation is that these sub- strates inadvertently contain an unknown mismatch elsewhere that signals concomitant repair of the loop. This possibility is also unlikely because these same heteroduplexes are efficiently repaired in an extract that is defective in mismatch repair. 17. That these extracts are not generally inactive for all DNA transactions is indicated by the fact that they are competent for simian virus 40 (SV40) ongin-de- pendent DNA replication. The relative replication ef- ficiency of the extracts was: HeLa, 100%; TK6, 120%; HCT1 16, 90%; LoVo, 41% (4). 18. Additional cell lines examined include the trans- formed lymphoblast line MT1 [V. S. Goldmacher, R. A. Cuzick, W. G. Thilly, J. Biol. Chem. 261, 12462 (1986)]; the colon tumor cell lines SW48 and SW480 [A. Leibovitz et al., Cancer Res. 36, 4562 (1976)l, DLD1 and HCT15 [D. L. Dexter, J. A. Bar- bosa, P. Calabresi, Cancer Res. 39, 1020 (1979)1, LS 180 and LS1 74T [B. H. Tom et al., In Vitro 12, 180 (1976)1; the endometrial tumor cell lines HEC59 [T. Morisawa, J. Jpn. Soc. Clin. Cytol. 26, 433 (1987)] and AN3CA [C. J. Dawe, W. G. Ban- field, W. D. Morgan, M. S. Slat- ick, H. 0. Curth, J. Nat!. Cancer Inst. 33, 441 (1964)]; the pancreatic cancer cell line HPAF-11 [R. S. Metzgar et al., Can- cer Res. 42, 601 (1982)1; the breast cancer cell lines T47D [H. C. Freake, C. Marcocci, J. Iwasaki, I. Macintyre, Biochem. Biophys. Res. Commun. 101, 1131 (1981)] and ZR-75-1 [L. W. Engel etal., Can- cer Res. 38, 3352 (1978)]; and the ovarian terato- carcinoma cell line PA-1 [J. Zeuthen et al., Int. J. Cancer 25, 19 (1980)1. 19. J. J. Miret, B. 0. Parker, R. S. Lahue, Environ. Mol. Mutagen. 23 (suppl. 23), 45 (1994). 20. R. Fishel, A. Ewel, S. Lee, M. K. Lescoe, J. Griffith, Science, in press. 21. We thank R. M. Schaaper and J. W. Drake for com- ments on the manuscript. 29 July 1994; accepted 1 1 October 1994 Ligands for EPH-Related Receptor Tyrosine Kinases That Require Membrane Attachment or Clustering for Activity Samuel Davis,* Nicholas W. Gale, Thomas H. Aldrich, Peter C. Maisonpierre, Vladimir Lhotak, Tony Pawson, Mitchell Goldfarb, George D. Yancopoulos* The EPH-related transmembrane tyrosine kinases constitute the largest known family of receptor-like tyrosine kinases, with many members displaying specific patterns of ex- pression in the developing and adult nervous system. A family of cell surface-bound ligands exhibiting distinct, but overlapping, specificities for these EPH-related kinases was identified. These ligands were unable to act as conventional soluble factors. However, they did function when presented in membrane-bound form, suggesting that they require direct cell-to-cell contact to activate their receptors. Membrane attachment may serve to facilitate ligand dimerization or aggregation, because antibody-mediated clustering ac- tivated previously inactive soluble forms of these ligands. Intercellular communication is often medi- ated by protein factors produced in one cell and recognized by receptors on the surface of other cells. Many of these factors, such as insulin and nerve growth factor, bind to and activate cell surface receptors with intrinsic protein tyrosine kinase activity (1). Ligand- mediated activation of these receptor ty- rosine kinases regulates cell growth, survival, S. Davis, N. W. Gale, T. H. Aldrich, P. C. Maisonpierre, M. Goldfarb, G. D. Yancopoulos, Regeneron Pharmaceuti- cals, 777 Old Saw Mill River Road, Tarrytown, NY 10591, USA. V. Lhotak and T. Pawson, Division of Molecular and De- velopmental Biology, Samuel Lunenfeld Research Insti- tute, Mount Sinai Hospital, Toronto, Ontario M5G 1X5, Canada. *To whom correspondence should be addressed. and differentiation in various cell types (1). There remain numerous receptor-like ty- rosine kinases whose ligands have yet to be identified, and many of these orphan recep- tors are specifically expressed in the nervous system (2). The EPH-related kinases consti- tute the largest known family of orphan re- ceptor-like tyrosine kinases, with several members of this family displaying specific expression in the developing and adult ner- vous system (2-17). In the adult a number of EPH-related kinases, such as EHK1 (16), EHK2 (16), and ELK (7), are restricted in their expression to discrete neuronal popula- tions, including locus coeruleus neurons and the dopaminergic neurons in the substantia nigra. To identify ligands that might func- SCIENCE * VOL. 266 * 4 NOVEMBER 1994 m m 816 on November 17, 2014 www.sciencemag.org Downloaded from on November 17, 2014 www.sciencemag.org Downloaded from on November 17, 2014 www.sciencemag.org Downloaded from on November 17, 2014 www.sciencemag.org Downloaded from