. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . REPORTS
`Structure of the FKBP12-Rapamycin
`Complex Interacting with the Binding
`Domain of Human FRAP
`Jungwon Choi,* Jie Chen, Stuart L. Schreiber, Jon Clardyt
`
`Rapamycin, a potent immunosuppressive agent, binds two proteins: the FK506-binding
`protein (FKBP12) and the FKBP-rapamycin-associated protein (FRAP). A crystal struc(cid:173)
`ture of the ternary complex of human FKBP12, rapamycin, and the FKBP12-rapamycin(cid:173)
`binding (FRB) domain of human FRAP at a resolution of 2. 7 angstroms revealed the two
`proteins bound together as a result of the ability of rapamycin to occupy two different
`hydrophobic binding pockets simultaneously. The structure shows extensive interac(cid:173)
`tions between rapamycin and both proteins, but fewer interactions between the proteins.
`The structure of the FRB domain of FRAP clarifies both rapamycin-independent and
`-dependent effects observed for mutants of FRAP and its homologs in the family of
`proteins related to the ataxia-telangiectasia mutant gene product, and it illustrates how
`a small cell-permeable molecule can mediate protein dimerization.
`
`A dividing cell must pass various check(cid:173)
`points as it proceeds through the cell cycle,
`and error-free division requires the ability to
`rectify DNA lesions (1, 2). The ataxia-telan(cid:173)
`giectasia mutant (A TM) gene product, the
`catalytic subunit of the DNA-dependent pro(cid:173)
`tein kinase, and the products of yeast genes
`such as TORI, TOR2, and MECJ (ESRJ) are
`members of a family of large molecular size
`proteins that participate in cell cycle progres(cid:173)
`sion and checkpoints as well as in DNA repair
`and recombination ( 1-4 ). All members of the
`A TM family contain a COOH-terminal ki(cid:173)
`nase domain (1-4), but no structural informa(cid:173)
`tion is available for any domain of any family
`member. Human FRAP (rat RAFT), one
`member of this family, is a 289-kD protein
`that binds FKBPl 2-rapamycin, regulates p 70
`ribosomal protein S6 kinase, and is required
`for G 1 cell cycle progression in several cell
`types (5-7).
`The potent immunosuppressive agents
`FK506 (8) and rapamycin (9-11) (Fig. lA)
`share the same cellular target: the 12-kD
`FK506-binding protein, FKBPl 2 ( 12, 13). Al(cid:173)
`though both rapamycin and FK506 bind to
`FKBP12 with high affinity [dissociation con(cid:173)
`stant (KJ), 0.2 to 0.4 nM], they effect im(cid:173)
`munosuppression through different mecha(cid:173)
`nisms. Whereas FK506 interrupts the signal
`from the T cell receptor, rapamycin inter(cid:173)
`rupts the signal from the interleukin-2 re(cid:173)
`ceptor and the receptors for other cytokines
`and growth factors (14). Binding of similar
`molecules, such as FK506 and rapamycin,
`
`J. Choi and J. Clardy, Department of Chemistry, Baker
`Laboratory, Cornell University, Ithaca, NY 14853-1301,
`USA.
`J. Chen and S. L. Schreiber, Howard Hughes Medical
`Institute and Department of Chemistry and Chemical Bi(cid:173)
`ology, Harvard University, Cambridge, MA 02138, USA.
`
`*Present address: Department of Chemistry, Suwon Uni(cid:173)
`versity, Kyunggi 445-773, South Korea.
`tTo whom correspondence should be addressed.
`
`to the same protein can disrupt different sig(cid:173)
`nals because the protein-ligand complexes,
`(not the individual components) are the
`active entities {14}. FKBP12-FK506 inhib(cid:173)
`its calcineurin, a serine-threonine phos(cid:173)
`phatase (15), whereas FKBP12-rapamycin
`binds to FRAP (5). Thus, small cell-perme(cid:173)
`able molecules can induce proteins to asso(cid:173)
`ciate in a manner similar to that by which
`growth factors (such as human growth hor(cid:173)
`mone), cytokines (such as interleukin-2),
`and intracellular signaling proteins (such as
`GRB2) induce dimerization of signaling
`proteins. These naturally occurring induc(cid:173)
`ers of dimerization have inspired the design
`of synthetic molecules that induce dimer(cid:173)
`ization of intracellular proteins and result
`in the activation of cytoplasmic signaling
`pathways (16).
`FK506 and rapamycin bind FKBP12 in a
`hydrophobic pocket, and an important feature
`of both FKBP12-ligand structures is the large
`fraction, -50%, of ligand exposed on the
`exterior of the complex (17-19). A composite
`surface-a surface with contributions from
`both ligand (FK506 or rapamycin) and the
`protein (FKBP12)-mediates the interaction
`of the FKBP12-ligand complex with its target
`(14 ). Two independent x-ray structures of the
`ternary FKBP12-FK506-calcineurin complex
`show a composite binding surface with exten(cid:173)
`sive contacts between FK506 and its protein
`partners as well as between the protein part(cid:173)
`ners themselves (20, 21). We now present a
`2. 7 A x-ray structure that shows how rapamy(cid:173)
`cin mediates
`the heterodimerization of
`FKBP12 with FRAP.
`Binding of FRAP to FKBP12-rapamycin
`(KJ = 2 nM) is mediated by a small domain,
`the FKBP12-rapamycin-binding (FRB) do(cid:173)
`main, that can be expressed as a 12-kD soluble
`protein (22). Crystals (23) of the ternary
`FKBP12-rapamycin-FRB complex were pre-
`
`pared and the structure was solved (24) by a
`combination of molecular replacement (MR)
`with isomorphous replacement for a single
`anomalously scattering derivative (SIRAS)
`(Fig. 1, Band C). The resulting MR-SIRAS
`map was readily interpretable, and the struc(cid:173)
`ture was refined to a final R factor of 0.193
`(8.0 to 2.7 A data). The ternary complex has
`a roughly rectan$1Jlar sh~pe with 0overall di(cid:173)
`mensions of 60 A by 45 A by 35 A. The two
`protein components are virtually the same
`size, and rapamycin is almost completely bur(cid:173)
`ied between them. The ternary complex fea(cid:173)
`tures an extensive array of rapamycin-protein
`interactions in two binding pockets lined with
`aromatic residues. FKBP12 contains a large ~
`sheet composed of five antiparallel ~ strands
`(17-19). A short amphipathic a helix is
`pressed against this sheet, and rapamycin
`binds in a hydrophobic pocket formed be(cid:173)
`tween the a helix and ~ sheet. Three loops-(cid:173)
`the 40s loop, which is a bulge in ~5, the 50s
`loop connecting ~5 to a, and the 80s loop
`connecting ~2 to ~3--surround and contrib(cid:173)
`ute to the binding pocket. In general, the
`structure of the FKBP12-rapamycin portion of
`the ternary complex does not differ from that
`of the binary FKBPl 2-rapamycin complex
`(18, 19), and the overall root-mean-square
`deviation between the binary and ternary
`complex is 1.14 A for all atoms and 0.49 A for
`main chain atoms.
`The FRB domain of FRAP forms a four(cid:173)
`helix bundle, a common structural motif in
`globular proteins (25). Its overall dimensions
`are 30 A by 45 A by 30 A, and all four helices
`(al to a4) have short underhand connec(cid:173)
`tions similar to the cytochrome b562 fold
`(26); The NH2- and COOH-termini of the
`FRB domain are close to each other, suggest(cid:173)
`ing that an FRB-type domain could be in(cid:173)
`serted into other protein chains. The longest
`helix, a3, has a 60° bend at Tyr2°74, roughly
`one-third of the distance from its NHz-ter(cid:173)
`minus, and the a2 helix has a small region
`(Gly2049 to Leu2051 ) that deviates from a
`standard a helix. Ignoring the first 10 resi(cid:173)
`dues of a3, all four helices are -26 A in
`length and comprise 16 to 19 residues. The
`al and a2 helices are almost parallel (inter(cid:173)
`helical angle of 22°), as are a3 and a4 (20°),
`whereas the crossing angles between the al(cid:173)
`a2 pair and the a3-a4 pair range from 30° to
`60°. The parallel helices are also closer. He(cid:173)
`lices al-a2 and a3-a4 show the shortest
`interhelical distances of 6.3 and 7 .2 A, re(cid:173)
`spectively, whereas other closest interhelical
`contacts range from 10 to 14 A. Most of the
`hydrophobic and aromatic residues are locat(cid:173)
`ed in the interhelical regions, and the hydro(cid:173)
`philic residues are exposed to solvent. The
`first and last helices of the bundle, al and
`a4, form a deep cleft near their crossing
`point, and this cleft, which is lined by six
`aromatic side chains, forms the hydrophobic
`
`SCIENCE • VOL. 273
`
`• 12 JULY 1996
`
`239
`
`West-Ward Exhibit 1057
`Choi 1996
`Page 001
`
`

`

`pocket in which rapamycin binds.
`Rapamycin interacts extensively with both
`FKBP12 and FRB.
`Its
`interactions with
`FKBP12 resemble those described for the bi(cid:173)
`nary complex and feature extensive contacts
`with conserved aromatic residues and five hy(cid:173)
`drogen bonds (18 , 19). Trp59 forms the base
`of the binding pocket and contacts the pipe(cid:173)
`colinyl ring (C2 to N7)-the most deeply
`buried portion of rapamycin. A total of 460
`A2 (44%) of the solvent-accessible surface
`area of rapamycin is buried in FKBP12. Rapa(cid:173)
`mycin also interacts with the FRB domain of
`FRAP through close contacts with aromatic
`residues, and a series of interactions along the
`triene arm of rapamycin (C16 to C23) involv(cid:173)
`ing Phe2039, Trp2101, Tyr21os. and Phe21os
`appear especially important (Fig. 2 ). Ser2035,
`Leu2031, Thf209B Asp2102, and Tyr203B also
`make contact with rapamycin. There are no
`hydrogen bonds in the FRB-rapamycin inter(cid:173)
`action, and 340 A2 (33%) of the solvent(cid:173)
`accessible surface area of rapamycin partici(cid:173)
`pate in the interaction. A small, but impor(cid:173)
`tant, conformational change in rapamycin is
`evident between the binary (FKBP12-rapamy(cid:173)
`cin) and ternary (FKBP12-rapamycin-FRB)
`complexes. In crystal structures of uncom(cid:173)
`plexed rapamycin (27) and the FKBP12-rapa(cid:173)
`mycin complex (18), the triene arm of rapa(cid:173)
`mycin is planar, with the three double bonds
`fully conjugated. In the ternary complex, ro(cid:173)
`tations of -15° about C18-C19 and 37°
`about C20-C21 slightly disrupt the conjuga(cid:173)
`tion and move the most deeply buried portion
`of rapamycin, the methyl group attached to
`C23, by 1.6 A (Fig. 1). This conformational
`alteration of rapamycin avoids a close contact
`with Phe2108 and places the deeply buried
`methyl group in a small crease between
`Phe2108 and Leu2031 (Fig. 2). The macrocyclic
`loop of rapamycin is eight carbon atoms larger
`than the corresponding loop in FK506, and
`the triene portion of rapamycin reduces the
`conformational flexibility of the loop and
`holds it away from the surface of FKBP12-it
`organizes the loop for binding into a deep
`pocket. Our analysis suggests that the limited
`tlexibiliry of the triene arm might also be an
`important feature of the structure of rapamy(cid:173)
`cin. If FK506 and rapamycin are overlaid, the
`much shorter loop of FK506 does not ap(cid:173)
`proach the binding pocket of the FRB domain
`of FRAP.
`Although rapamycin interacts extensively
`with both protein partners, the extent to
`which the proteins interact with each other is
`relatively limited (Fig. 2). Two regions of the
`complex show interactions between the pro(cid:173)
`teins: the 40s loop ofFKBP12 with cx4 ofFRB,
`and the 80s loop of FKBP12 with the cxl-cx2
`region of FRB. In the 40s-cx4 interaction, the
`OH group ofTyr2105 and the 0 atom ofLys47
`make a short contact and there is also a
`water-mediated salt bridge. In the other inter-
`
`•
`
`• •
`•
`•
`
`• •
`•
`
`c
`
`2065
`
`N
`
`2060
`
`•
`
`•
`
`•
`
`I •
`•
`
`,C)
`
`I_
`
`'r
`
`c
`
`2023
`
`2074
`
`2110
`
`n4
`
`n1
`
`a3
`
`a2
`
`114
`
`113
`
`n1
`
`FRB
`
`2044
`
`2091
`
`115
`
`2042
`
`2095
`
`80s loop
`
`112
`
`111
`
`FKBP12
`
`Fig. 1. (A} Chemical structure of rapamycin. (B} The 2F0 - Fe electron density of the FRB domain of the
`FKBP12-rapamycin-FRB complex (24) (F0 , observed structure factor; Fe• calculated structure factor}. A
`model of the final structure is embedded in this inttial electron density. (CJ Overall structure of the ternary
`complex between FKBP12 (blue ribbon}, rapamycin (ball and stick), and the FRB domain of FRAP (red
`ribbon}. Secondary structural elements are labeled with the conventional numbering scheme for FKBP12.
`N and C, NHr and COOH-termini, respectively. The drawing in (C) was prepared wtth RIBBONS (31}.
`
`action region, the NH2 group of Arg2042
`makes short contacts with Oy 1 of Thr85 and
`the 0 atom of Gly86, and there are two water(cid:173)
`mediated interactions. Although the number
`of interprotein polar interactions is moderate,
`400 A 2 of solvent-accessible surface area,
`roughly equivalent to the surface area buried
`by FKBP12-rapamycin or FRB-rapamycin,
`participate in the interaction between them.
`The 80s loop of FKBP12 appears to be one
`region where the structure of FKBP12 differs
`between the binary and ternary complexes,
`with the major change being around Ile90,
`where both side chain and main chain devi(cid:173)
`ations are apparent. These deviations move
`FKBP12 away from the FRB domain, suggest(cid:173)
`ing repulsion between the proteins in this
`region.
`The residues that form the rapamycin
`binding pocket of FRAP (Fig. 2) are con(cid:173)
`served
`in yeast Torlp, Tor2p, and rat
`RAFfl-other ATM family members that
`
`bind FKBP12-rapamycin-and thus all four
`proteins are likely to contain a hydrophobic
`pocket with similar architecture and related
`function. Overexpression of the Ser1972 (cor(cid:173)
`responding to Ser2°35 of FRAP) -
`Ile mutant
`of Torlp results in marked inhibition of cell
`growth that is dependent on the kinase activ(cid:173)
`ity of the mutant (28). These results indicate
`that the site bearing the Ser mutation may
`regulate the neighboring kinase domain of
`Torlp (28). The conserved binding pocket
`associated with a regulatory function suggests
`that binding of an as yet unidentified ligand
`may regulate kinase activity. The shape of
`the binding pocket and the cxl-cx4 crossing
`angle are mutually dependent; thus, ligand
`binding could result in a change in the cross(cid:173)
`ing angle, and the domain could function as
`a ligand-dependent conformational switch.
`FKBPl 2 might adjust the crossing angle
`(possibly by having its 80s loop repel the
`cxl-cx2 helical pair of the FRB domain of
`
`240
`
`SCIENCE
`
`• VOL. 273
`
`• 12 JULY 1996
`
`West-Ward Exhibit 1057
`Choi 1996
`Page 002
`
`

`

`. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . 11111111111111111111 . . lllBllllBl . . llllllllBllllBllllBIREPORTS
`
`~6
`
`3,. lle90
`
`~99 -~ ._J Asp37
`ps--\ b
`
`Phe~;?>/
`v --270
`
`(
`
`-::
`
`lie·~
`
`01
`
`3. Thr2098
`-"1m
`~r21os ~2101
`
`tsp 2102
`
`~he21oa
`
`fuu20J1
`
`Gln53
`
`o--e Ligand bond
`o------e Nonligand bond
`0-3 o.e Hydrogen bond and its length
`
`His53 Nonligand residues involved in
`~ hydrophobic contacts
`
`... Corresponding atoms involved
`in hydrophobic contacts
`
`Fig. 2. Interactions of rapamycin with FKBP12 and FRB. (A) A LIGPLOT
`(32) rendering of the interactions of rapamycin with FKBP1 2 and FRB.
`Hydrogen bonds are shown by dashed lines (with lengths in angstroms),
`and hydrophobic interactions by surface dashes. (B) Close-up of the inter(cid:173)
`actions of rapamycin (ball and stick) with the FRB domain of FRAP (red
`ribbon). Residue numbering is based on that for FRAP. (C) Complementa(cid:173)
`rity plot of rapamycin with FKBP12 and the FRB domain of FRAP (33). High
`complementarity is indicated by purple. FKBP12 is on the right and the
`opening of the deep cavity where the pipecolinyl moiety is buried is visible.
`The FRB domain of FRAP is on the left, and the most deeply buried methyl
`group is shown disappearing into the FRB cavity.
`
`FRAP) so that the FRB pocket is optimized
`for rapamycin binding.
`Mutation of Ser2035 to Thr in FRAP
`renders the protein insensitive to the in(cid:173)
`hibitory effects of the FKBP12-rapamycin
`complex; the extra bulk of Thr2035 may
`prevent binding of rapamycin in the pock(cid:173)
`et (7) (Fig. 2). Mutations of Trp2101 and
`Phe 2108 also confer rapamycin resistance
`these residues also
`interact
`(29). and
`strongly with rapamycin (Fig. 2).
`Our data provide a structural frame(cid:173)
`work for understanding the rapamycin(cid:173)
`based dimerization of FKBP12 and FRAP.
`Comparison of the FKBP12-FK506-cal(cid:173)
`cineurin structure (20, 21) with
`the
`FKBPl 2-rapamycin-FRB structure reveals
`different strategies for the two dimeriza(cid:173)
`tion modes. Whereas FK506-induced
`dimerization features extensive protein-
`
`protein interactions, rapamycin-induced
`dimerization does not. Because rapamycin(cid:173)
`induced protein dimerization can form the
`basis for regulating gene transcription and
`other cellular processes (30), such struc(cid:173)
`ture-based modifications of the interac(cid:173)
`tion might have important practical con(cid:173)
`sequences. The structure also provides in(cid:173)
`sights into structural features and possible
`regulation of the A TM family of proteins.
`
`REFERENCES AND NOTES
`
`1. C. T. Keith and S. L. Schreiber, Science 270, 50
`(1995).
`2. V. A. Zakian, Ce// 82, 685 (1995).
`3. T. Hunter, ibid. 83, 1 (1995).
`4. S. P. Jackson: Curr. Biol. 5, 1210 (1995).
`5. E. J. Brown et al., Nature 369, 756 (1994).
`6. D. Sabatini, H. Erdjument-Bromage, M. Lui, P.
`Tempst, S. Snyder, Cell 78, 35 (1994).
`7. E. J. Brown et al., Nature 377, 441 (1995).
`
`8. H. Tanaka et al., J. Am. Chem. Soc. 109, 5031
`(1987).
`9. D. C. N. Swindells, P. S. White, J. A. Findlay, Can. J.
`Chem. 56, 2491 (1978).
`10. S. N. Sehgal, H. Baker, C. Vezina, J. Antibiot. 28,
`727 (1975).
`11 . C. Vezina, A. Kudelski, S. N. Sehgal, ibid., p. 721.
`12. M. W. Harding, A. Galat, D. E. Uehling, S. L. Schrei(cid:173)
`ber, Nature 341, 758 (1989).
`13. J. J. Siekierka, S. H. Y. Hung, M. Poe, C. S. Lin, N. H.
`Sigal, ibid., p. 755.
`14. S. L. Schreiber, Science 251, 283 (1991 ).
`15. J. Liu et al., Cell 66, 807 (1991).
`16. D. M. Spencer, T. J. Wandless, S. L. Schreiber, G. R.
`Crabtree, Science 262, 1019 (1993); P. J. Belshaw,
`S. N. Ho, G. R. Crabtree, Proc. Natl. Acad. Sci.
`U.S.A. 93, 4604 (1996).
`17. G.D. Van Duyne, R. F. Standaert, P. A. Karplus, S. L.
`Schreiber, J. Clardy, Science 251, 839 (1991).
`18. G. D. Van Duyne, R. F. Standaert, S. L. Schreiber, J.
`Clardy. J. Am. Chem. Soc. 113, 7 433 (1991 ).
`19. G. D. Van Duyne, R. F. Standaert, P. A. Karplus, S. L.
`Schreiber, J. Clardy, J. Mo/. Biol. 229, 105 (1993).
`20. C. R. Kissinger et al., Nature 378, 641 (1995).
`21. J. P. Griffith et al., Cell 82, 507 (1995).
`22. J. Chen, X. F. Zheng, E. J. Brown, S. L. Schreiber,
`
`SCIENCE
`
`• VOL. 273
`
`• 12 JULY 1996
`
`241
`
`West-Ward Exhibit 1057
`Choi 1996
`Page 003
`
`

`

`Proc. Natl. Acad. Sci. U.S.A. 92, 4947 (1995).
`23. The expression and purification of recombinant hu(cid:173)
`man FKBP12 (19) and the FRB domain of human
`FRAP (22) have been described. Crystals of
`FKBP12-rapamycin-FRB were grown in 2 to 3
`weeks at room temperature from hanging drops
`prepared from FKBP1 2 [10 mg/ml, in 10 mM tris(cid:173)
`HCI (pH 8.0)], two equivalents of rapamycin (in
`methanol), and one equivalent of FRB [10 mg/ml, in
`50 mM tris-HCI (pH 8.0)]. The well solution con(cid:173)
`tained 20% (w/v) polyethylene glycol 8000, 10%
`methypentanediol, and 10 mM tris-HCI (pH 8.5).
`The rod-shaped crystals are orthorhombic, space
`group P2 12 12 1 with cell constants a = 44.63,
`b = 52.14, and c = 102.53 A, and contain one
`ternary complex in the asymmetrical unit.
`24. Data to a resolution of 2.7 A (43,447 measure(cid:173)
`ments of 6920 unique reflections, 98.5% complete,
`Rsym = 0.071) were collected from a crystal of
`dimensions 0.3 mm by 0.2 mm by 0.1 mm with the
`use of a San Diego multiwire area detector on a
`Rigaku RU-200 rotating anode x-ray source. Ex(cid:173)
`perimental phases were obtained from MR and
`SIRAS. MR with X-PLOR [AT. Brunger, J. Kuriyan,
`M. Karplus, Science 235, 458 (1987)] and the
`FKBP12-rapamycin model (19) yielded a clear so(cid:173)
`lution, but the resulting electron density map was
`noisy. A mercury derivative was prepared (2 mM
`HgCl2 , overnight), and the two heavy atom sites
`were refined with PHASES [W. Furey and S.
`Swaminathan, ACA Abstr. 18, 73 (1990)]. Anoma(cid:173)
`lous dispersion measurements were included in
`this data set and 16 cycles of solvent flattening
`were applied (PHASES). The resulting electron
`density map clearly showed the four-helix-bundle
`architecture of FRB. The FKBP12-rapamycin por(cid:173)
`tion of the structure was well defined in the initial
`electron density map, and minor changes in the
`backbone of the 30s loop and some side chains
`were sufficient to fit the model. For the FRB portion,
`most of a polyalanine chain could be traced for the
`helical regions of the initial map. After several cy(cid:173)
`cles of positional refinement (X-PLOR), loop re(cid:173)
`gions could also be traced and the side chains
`assigned. CHAIN [J. S. Sack, J. Mo/. Graphics 6,
`244 (1988)] was used for model fitting and building
`the structure. A total of 95 residues in the FRB
`domain of FRAP (three residues in the NH2 -termi(cid:173)
`nal and two residues in the COOH-terminal regions
`showed no electron density and were not includ(cid:173)
`ed), all residues of FKBP12, all atoms of rapamycin,
`and 23 water molecules were included in the final
`model. FRB residues are numbered according to
`FRAP numbering. The current R factor is 0.193
`(R,, •• = 0.299) for data from 8 to 2.7 A. The root(cid:173)
`mean-square deviations of bond lengths and bond
`angles are 0.008 A and 1.48°, respectively. The
`average temperature factors for all atoms and main
`chain atoms are 17.0 and 14.7 A2. respectively.
`25. N. L. Harris, S. R. Presnell, F. E. Cohen, J. Mo/. Biol.
`236, 1356 (1994).
`26. F. Lederer, A. Glatigny, P. H. Bethge, H. D. Bellamy,
`F. S. Mathews, ibid. 148, 427 (1981).
`27. J. A. Findlay and L. Radics, Can. J. Chem. 58, 579
`(1980).
`28. X. F. Zheng, D. Fiorentino, J. Chen, G. R. Crabtree,
`S. L. Schreiber, Ce// 82, 121 (1995).
`29. M. C. Lorenz and J. Heitman, J. Biol. Chem. 270,
`27531 (1995).
`30. V. M. Rivera et al., Nature Med., in press.
`31. M. Carson, J. Mo/. Graphics 5, 103 (1987).
`32. A. C. Wallace, R. A. Laskowski, J.M. Thornton, Prot.
`Engin. 8, 127 (1995).
`33. A. Nicholls, K. Sharp, B. Honig, GRASP Manual (Co(cid:173)
`lumbia Univ. Press, New York, 1992).
`34. We thank S. Ealick, J. Liang, and R. Gillilan for dis(cid:173)
`cussions. The Cornell work was funded in part by
`USPHS grant CA59021 (to J.C.) and the Harvard
`work by an Irvington Institute Fellowship (to J.C.) and
`USPHS grant GM38625 (to S.L.S.). S.L.S. is a
`Howard Hughes Medical
`Institute
`Investigator.
`Atomic coordinates have been deposited in the Pro(cid:173)
`tein Data Bank under the accession number 1 FAP.
`
`15 March 1996; accepted 15 May 1996
`
`Long-Term Lymphohematopoietic
`Reconstitution by a Single CD34-Low/Negative
`Hematopoietic Stem Cell
`Masatake Osawa,* Ken-ichi Hanada, Hirofumi Hamada,
`Hiromitsu Nakauchit
`
`Hematopoietic stem cells (HSCs) supply all blood cells throughout life by making use of
`their self-renewal and multilineage differentiation capabilities. A monoclonal antibody
`raised to the mouse homolog of CD34 (mCD34) was used to purify mouse HSCs to near
`homogeneity. Unlike in humans, primitive adult mouse bone marrow HSCs were de(cid:173)
`tected in the mCD34 low to negative fraction. Injection of a single mCD34101-, c-Kit+,
`Sca-1 +, lineage markers negative (Lin-) cell resulted in long-term reconstitution of the
`lymphohematopoietic system in 21 percent of recipients. Thus, the purified HSC pop(cid:173)
`ulation should enable analysis of the self-renewal and multilineage differentiation of
`individual HSCs.
`
`CD34 is a marker of human HSCs, and all
`colony-forming activity of human bone mar(cid:173)
`row (BM) cells is found in the CD34-positive
`fraction (1 ). Clinical transplantation studies
`that used enriched CD34 + BM cells also in(cid:173)
`dicated the presence of HSCs with long-term
`BM reconstitution ability within this frac(cid:173)
`tion (2). After isolation of the human
`CD34 gene, the mouse homolog (mCD34)
`was isolated by cross-hybridization (3). To
`examine the expression and function of
`mCD34, we raised a monoclonal antibody
`(mAb), 49E8 [rat immunoglobulin G2a
`(IgG2a)), to mCD34 by immunizing rats
`with a glutathione-S-transferase (GST)(cid:173)
`mCD34 fusion protein. This mAb stained
`BaF3 cells transfected with a full-length
`mCD34 cDNA but not mock-transfected
`cells (4). Murine cell lines such as PA6,
`NIH 3T3, Ml, and DAl, shown by reverse
`transcriptase-polymerase chain
`reaction
`(RT-PCR)
`to contain mCD34 mRNA,
`were also stained by this mAb, indicating
`that 49E8, although specific for a GST(cid:173)
`mCD34 fusion protein, could also recognize
`the native form of mCD34 as expressed on
`various cell types ( 4 ).
`We next examined adult mouse BM for
`expression of mCD34. Four-color fluores(cid:173)
`cence-activated cell sorter (FACS) analysis
`was done after sequential staining of BM cells
`with a combination of lineage-specific mAbs
`to CD4, CDS, B220, Gr-1, Mac-1, and
`TERl 19, and then a mixture of mAbs to c-Kit
`(ACK-2), Ly6A/E (Sca-1), and mCD34 (5).
`
`M. Osawa and H. Nakauchi, Department of Immunology,
`Institute of Basic Medical Sciences and Center for
`Tsukuba Advanced Research Alliance, University of
`Tsukuba, Tsukuba Science-City, lbaraki 305, Japan.
`K.-i. Hanada and H Hamada, Department of Molecular
`Biotherapy, Cancer Chemotherapy Center, Japanese
`Foundation for Cancer Research, Tokyo 170, Japan.
`
`*Present address: KIRIN Pharmaceutical Research Lab(cid:173)
`oratory, Gunma 371, Japan.
`tTo whom correspondence should be addressed.
`
`Monoclonal antibody 49E8 reacted with
`2.5 ± 0.5% (mean ± SD) of total BM cells,
`with most of the positive cells occurring in the
`Lin- fraction (Fig. lA). More than 90% of
`the c-Kit + Sea 1 + Lin - cells previously
`shown to contain primitive HSCs (6) stained
`brightly with 49E8, whereas the remainder
`were low to negative (Fig. lB). The frequency
`of mCD34 + c-Kit + Sca-1 + Lin - cells and
`mCD34- c-Kit+ Sca-1+ Lin- cells among
`total nucleated BM cells was 0.073 ± 0.028%
`(mean ± SD, n = 5) and 0.004 ± 0.003%
`(mean ± SD, n = 5), respectively.
`To determine whether mouse HSCs ex(cid:173)
`press mCD34, we sorted subpopulations by
`FACS and examined their stem cell activity.
`Within the c-Kit+ Sca-1 + Lin- population,
`the frequency of interleukin-3 (IL-3 )-de(cid:173)
`pendent colony-forming unit culture (CFU(cid:173)
`C) per 200 cells was 20.0 ± 3.9% (mean ±
`SD, n = 8) (7) for mCD34+ cells but only
`0.16 ± 0.4% (mean ± SD, n = 8) in the
`CD34- fraction. Similarly, mCD34+ cells
`contained 14.1 ± 3.4% (mean ± SD, n =
`15) day 12 CFU spleen (CFU-S) per 200
`cells, whereas in the mCD34 - fraction this
`value was 1.6 ± 1.7% (mean± SD, n = 15)
`(8). Thus, colony-forming activity was pos(cid:173)
`itively correlated with mCD34 expression
`among c-Kit+ Sca-1 + Lin- cells. When
`these cells were cultured in the presence of
`both IL-3 and stem cell factor (SCF), how(cid:173)
`ever, 80% of mCD34- c-Kit+ Sca-1 +Lin(cid:173)
`cells formed large multilineage colonies (7).
`For in vivo analyses, c-Kit+ Sca-1 +Lin(cid:173)
`cells were fractionated into mCD3410/- (Fr.
`1), mCD341" (Fr. 2), and c034+ (Fr. 3)
`subpopulations according to their mCD34
`expression by FACS (Fig. 2A). Although
`100 c-Kit+ Sca-1 + Lin- cells were sufficient
`to radioprotect a lethally irradiated mouse,
`injection of 300 cells from either the Fr. 1 or
`Fr. 3 subpopulation {Fig. 2A) alone showed
`poor radioprotective ability (9). When cells
`
`242
`
`SCIENCE
`
`• VOL. 273
`
`• 12 JULY 1996
`
`West-Ward Exhibit 1057
`Choi 1996
`Page 004
`
`