3. P. T. Todorov et al., Cancer Res. 58, 2353 (1998); K. Hirai, H. J. Hussey, M. D. Barber, S. A. Price, M. J. Tisdale, ibid., p. 2359. 4. Protein sequences: Human ZAG [T. Araki et al., Proc. Natl. Acad. Sci. U.S.A. 85, 679 (1988)]; mouse ZAG [H. Ueyama, H. Naitoh, I. Ohkubo, J. Biochem. 116, 677 (1994)]; rat ZAG [A. Fueyo, J. A. Urı ´a, J. P. Freije, C. Lo ´pez-Otı ´n, Gene 145, 245 (1994)]. Human class I MHC sequences [P. J. Bjorkman and P. Parham, Annu. Rev. Biochem. 59, 253 (1990)]; MHC homolog se- quences, SWISS-PROT [A. Bairoch and R. Apweiler, Nucleic Acids Res. 26, 38 (1998)]. Sequence identi- ties: ZAG and HLA-A2, 36%; ZAG and FcRn, 27%; ZAG and mouse CD1d, 23%; ZAG and HFE, 36%; ZAG and MIC-A, 29%. 5. D. R. Madden, Annu. Rev. Immunol. 13, 587 (1995). 6. E. M. Beckman et al., Nature 372, 691 (1994); A. R. Castan ˜o et al., Science 269, 223 (1995). 7. R. P. Junghans, Immunol. Res. 16, 29 (1997); V. Ghetie and E. S. Ward, Immunol. Today 18, 592 (1997); N. E. Simister, E. J. Israel, J. C. Ahouse, C. M. Story, Biochem. Soc. Trans. 25, 481 (1997). 8. J. N. Feder et al., Nature Genet. 13, 399 (1996); J. N. Feder et al., Proc. Natl. Acad. Sci. U.S.A. 95, 1472 (1998). 9. L. M. Sa ´nchez, C. Lo ´pez-Otı ´n, P. J. Bjorkman, Proc. Natl. Acad. Sci. U.S.A. 94, 4626 (1997). In thermal denaturation studies, ZAG denatures with a transi- tion midpoint of 65°C, compared to 57°C for the peptide-filled class I molecule H-2K d and 45°C for empty K d [M. L. Fahnestock, I. Tamir, L. Narhi, P. J. Bjorkman, Science 258, 1658 (1992)]. 10. V. Groh et al., Proc. Natl. Acad. Sci. U.S.A. 93, 12445 (1996). The crystal structure of MIC-A reveals a major rearrangement in domain organization compared to the structures of class I molecules and ZAG. The MIC-A 1-2 platform is displaced from its 3 domain by 113.5° compared to class I molecules. As a result, the MIC-A 1-2 platform makes no contact with its 3 domain [P. Li, S. T. Willie, S. Bauer, D. L. Morris, T. Spies, R. K. Strong, personal communication]. 11. Protein structures: HLA-A2 [Protein Data Bank (PDB) code 2CLR] [E. J. Collins, D. N. Garboczi, D. C. Wiley, Nature 371, 626 (1994)]; Mouse CD1 (PDB code 1CD1) [Z.-H. Zeng et al., Science 277, 339 (1997)]; Rat FcRn (PDB code 3FRU) [W. P. Burmeister, L. N. Gastinel, N. E. Simister, M. L. Blum, P. J. Bjorkman, Nature 372, 336 (1994)]; Human HFE (PDB code 1A6Z) [J. A. Lebro ´n et al., Cell 93, 111 (1998)]. Molecular surface areas buried by interaction were calculated with X-PLOR [A. T. Bru ¨nger, X-PLOR. Ver- sion 3.1: A System for X-ray and NMR (Yale Univ. Press, New Haven, CT, 1992)] with a 1.4 Å radius. Identification of pocket residues and calculation of groove surface areas were done based upon earlier analyses of human and mouse class I structures [M. A. Saper, P. J. Bjorkman, D. C. Wiley, J. Mol. Biol. 219, 277 (1991); M. Matsumura, D. H. Fremont, P. A. Peterson, I. A. Wilson, Science 257, 927 (1992)] as described in the CD1 ( Z.-H. Zeng et al.) and HFE ( J. A. Lebro ´n et al.) structure papers. Cut away molecular surfaces of grooves (Fig. 3B) were generated as de- scribed in the CD1 structure paper. 12. S. Shibata and K. Miura, Nephron 31, 170 (1982). 13. Structure determination and refinement: HKL [Z. Otwi- nowski and W. Minor, Methods Enzymol. 276, 307 (1997)]. SHARP [E. De La Fortelle and G. Bricogne, ibid., p. 472]. Solomon [ J. P. Abrahams and A. G. W. Leslie, Acta Crystallogr. D 52, 30 (1996)]. CCP4 programs [CCP4: Collaborative Computational Project No. 4, Daresbury, UK, Acta Crystallogr. D 50, 760 (1994)]. O [T. A. Jones and M. Kjeldgaard, Methods Enzymol. 277, 173 (1997)]. R free [A. T. Bru ¨nger, Nature 355, 472 (1992)]. The Native II data set (Table 1) was used for refinement. After rigid-body refinement of eight do- mains in the asymmetric unit (1-2 and 3 for each of four ZAG molecules) using CNS [A. T. Bru ¨nger et al., Acta Crystallogr. D 54, 905 (1998)], the four molecules were subjected to restrained NCS torsion-angle refine- ment using the maximum likelihood target function. Tight NCS restraints (300 kcal/mol . Å 2 ) were applied to all regions except for flexible loops and residues in- volved in lattice contacts. Intermediate rounds of model building and refinement included the calculation of SIGMAA-weighted [R. J. Read, Acta Crystallogr. A 42, 140 (1986)] simulated annealing omit maps [A. Hodel, S.-H. Kim, A. T. Bru ¨nger, Acta Crystallogr. A 48, 851 (1992)]. Final rounds of rebuilding and refinement in- cluded tightly restrained individual atomic temperature factor refinement (temperature factor rms deviation for bonded main chain and side chain atoms is 5.7 and 8.8 Å 2 , respectively). The model consists of residues 5 through 277 (average B: 48 Å 2 ) with nine carbohydrate residues (average B: 61 Å 2 ) for molecule 1, residues 5 through 278 (average B: 56 Å 2 ) with 11 carbohydrate residues (average B: 80 Å 2 ) for molecule 2, residues 6 through 278 (average B: 57 Å 2 ) with four carbohydrate residues (average B: 107 Å 2 ) for molecule 3, and resi- dues 6 through 249 and 258 through 276 (average B: 62 Å 2 ) with five carbohydrate residues (average B: 90 Å 2 ) for molecule 4 (Wilson B = 64 Å 2 ). Excluding regions that deviate from the NCS, the domains in the NCS- related ZAG monomers are very similar (0.04 Å rms deviation for C atoms). Ramachandran plot statistics (Table 1) are as defined by G. J. Kleywegt and T. A. Jones [Structure 4, 1395 (1996)]. 14. Extensive carbohydrate density is found at Asn 239 (nine ordered carbohydrate residues in molecule 2) and to a much lesser extent at Asn 89 and Asn 108 in all four ZAG molecules (Fig. 1) (13). Crystal structures of glycoproteins rarely show more than three ordered carbohydrate residues at each glycosylation site [D. E. Vaughn and P. J. Bjorkman, Structure 6, 63 (1998)]. The Asn in the fourth potential N-linked glycosyla- tion site (Asn 92 ) does not show density correspond- ing to carbohydrate. The bond between Asn 92 and Gly 93 can be cleaved by hydroxylamine, confirming that Asn 92 is not glycosylated (18). 15. M. Takagaki et al., Biochem. Biophys. Res. Commun. 201, 1339 (1994); O. Ogikubo et al., ibid. 252, 257 (1998); M. Pfaff, in Integrin-Ligand Interaction, J. A. Eble and K. Ku ¨hn, Eds. (Chapman & Hall, New York, 1997), pp. 101–121. 16. V. A. Tysoe-Calnon, J. E. Grundy, S. J. Perkins, Bio- chem. J. 277, 359 (1991); D. Lancet, P. Parham, J. L. Strominger, Proc. Natl. Acad. Sci. U.S.A. 76, 3844 (1979); A. Bauer et al., Eur. J. Immunol. 27, 1366 (1997). 17. Structural features that prevent ZAG from binding  2 M include the following residues, which clash with  2 M when it is positioned on the ZAG structure either by interacting with 3 or with 1-2: Ile 13 , Thr 15 , Leu 30 , Arg 40 , Gln 98 , Tyr 118 , Lys 122 , Val 234 , His 236 , Trp 245 . 18. L. M. Sa ´nchez and P. J. Bjorkman, unpublished results. 19. G. F. Gao et al., Nature 387, 630 (1997). 20. D. N. Garboczi et al., ibid. 384, 134 (1996); K. C. Garcia et al., Science 274, 209 (1996); Y. H. Ding et al., Immunity 8, 403 (1998); K. C. Garcia et al., Science 279, 1166 (1998). 21. Superpositions based on C atoms in the platform  strands reveal that the ZAG platform is more similar to classical class I MHC molecules than to any of the class I homologs [rms deviations for superpositions of platforms: ZAG and HLA-A2, 1.3 Å (147 C atoms); ZAG and CD1, 1.1 Å (86 C atoms); ZAG and FcRn 1.0 Å (88 C atoms); ZAG and HFE 1.0 Å (115 C atoms)]. 22. L. M. Sa ´nchez, A. J. Chirino, P. J. Bjorkman, G. Hatha- way, P. G. Green, K. Faull, unpublished results. 23. Figures 1, 2A (right), 2B, 3A, and 3C were made using MOLSCRIPT [P. J. Kraulis, J. Appl. Crystallogr. 24, 946 (1991)] and RASTER-3D [E. A. Merritt and M. E. P. Murphy, Acta Crystallogr. D. 50, 869 (1994)]. Elec- trostatic calculations were done and Figs. 2A (left) and 3B were made using GRASP [A. Nicholls, R. Bharadwaj, B. Honig, Biophys. J. 64, A166 (1993)]. 24. ZAG, CD1, HFE, and FcRn contain prolines within their 2 domain helices at a position corresponding to Val 165 in classical class I MHC molecules (4). The FcRn and CD1 helices are kinked at a position near their proline residues, whereas the ZAG and HFE helices are similar to the 2 domain helices of class I molecules (11). Substitution of Val 165 for proline in the mouse class I molecule H-2D d did not interfere with binding and presentation of peptides to T cells, suggesting that no major structural rearrangements occurred [D. Plaksin, K. Polakova, M. G. Mage, D. H. Margulies, J. Immunol. 159, 4408 (1997)]. 25. We thank G. Hathaway, P. G. Green, and K. Faull for mass spectrometric analyses. ZAG coordinates have been deposited in the PDB (code 1zag). L.M.S. was supported by a grant from the U.S. Department of Defense Breast Cancer Research Program. 21 December 1998; accepted 18 February 1999 Acoel Flatworms: Earliest Extant Bilaterian Metazoans, Not Members of Platyhelminthes In ˜ aki Ruiz-Trillo, 1 Marta Riutort, 1 D. Timothy J. Littlewood, 2 Elisabeth A. Herniou, 2 Jaume Bagun ˜a ` 1 * Because of their simple organization the Acoela have been considered to be either primitive bilaterians or descendants of coelomates through secondary loss of derived features. Sequence data of 18S ribosomal DNA genes from non–fast evolving species of acoels and other metazoans reveal that this group does not belong to the Platyhelminthes but represents the extant members of the earliest divergent Bilateria, an interpretation that is supported by recent studies on the embryonic cleavage pattern and nervous system of acoels. This study has implications for understanding the evolution of major body plans, and for perceptions of the Cambrian evolutionary explosion. “Since the first Metazoa were almost certainly radial animals, the Bilateria must have sprung from a radial ancestor, and there must have been an alteration from radial to bilateral sym- metry. This change constitutes a most difficult gap for phylogeneticists to bridge, and various highly speculative conjectures have been made” (1, p. 5). So began Libbie Hyman’s discussion on the origin of bilaterian Metazoa, and despite a century of morphological studies and a decade of intensive molecular work, the nature of the simplest bilaterian animal remains elusive (1, 2). Paleontological and molecular data indicate that most bilaterian phyla ap- peared and diversified during the Cambrian explosion (3, 4). Three main clades emerged— R EPORTS www.sciencemag.org SCIENCE VOL 283 19 MARCH 1999 1919