Proc. Natl. Acad. Sci. USA Vol. 88, pp. 999-1003, February 1991 Biochemistry
Zinc fingers, zinc clusters, and zinc twists in DNA-binding protein domains (transcription factors/steroid receptors/hormone receptors/GAL4/zinc binding site)
BERT L. VALLEE*t, JOSEPH E.
COLEMANt, AND DAVID S. AULD*§
*Center for Biochemical and Biophysical Sciences and Medicine and §Department of Pathology, Harvard Medical School, Brigham and Women's Hospital, 250 Longwood Avenue, Boston, MA 02115; and tDepartment of Molecular Biophysics and Biochemistry, Yale University, New Haven, CT 06510
Contributed by Bert L. Vallee, October 25, 1990
We now recognize three distinct motifs of ABSTRACT DNA-binding zinc proteins: (i) zinc ringers, (ii) zinc clusters, and (iii) zinc twists. Until very recently, x-ray crystallographic or NMR three-dimensional structure analyses of DNA-binding zinc proteins have not been available to serve as standards of reference for the zinc binding sites of these families of proteins. Those of the DNA-binding domains of the fungal transcription factor GAL4 and the rat glucocorticoid receptor are the frst to have been determined. Both proteins contain two zinc binding sites, and in both, cysteine residues are the sole zinc ligands. In GAL4, two zinc atoms are bound to six cysteine residues which form a "zinc cluster" akin to that of metaflothionein; the distance between the two zinc atoms of GAL4 is "'3.5 A. In the glucocorticoid receptor, each zinc atom is bound to four cysteine residues; the interatomic zinc-zinc distance is - 13 A, and in this instance, a "zinc twist" is represented by a helical DNA recognition site located between the two zinc atoms. Zinc clusters and zinc twists are here recognized as two distinctive motifs in DNA-binding proteins containing multiple zinc atoms. For native "zinc ringers," structural data do not exist as yet; consequently, the interatomic distances between zinc atoms are not known. As further structural data become available, the structural and functional significance of these different motifs in their binding to DNA and other proteins participating in the transmiission of the genetic message will become apparent.
The three-dimensional structures of 12 zinc enzymes and their zinc binding sites have been established by x-ray crystallographic analysis. This has focused attention on the zinc chemistry underlying their catalytic functions (1). Their zinc binding sites are characteristic of those of other members of their respective families (1, 2) and show that primary structures alone are unsuitable for such comparisons. The results of such studies constitute compelling evidence regarding the minimum information required to recognize and identify zinc binding sites of other zinc proteins. A functionally heterogeneous group of DNA-binding proteins that includes transcription factors, retroviral low molecular weight nucleic acid-binding proteins, adenovirus EJA gene products, aminoacyl tRNA synthetases, large tumor (T) antigen, bacteriophage proteins, and hormone receptors has been inferred to be zinc proteins based solely on their primary structures (3). Combinations of conserved cysteine and histidine residues separated by variable numbers of amino acids have been formulated into a popular model known as a "zinc finger" (4), in which the zinc chelate forms its "knuckle," as it were. Patterns such as these are thought further to signal the presence ofzinc. The desirability ofzinc analysis together with tertiary structure determinations to confirm such predictions by documenting the zinc binding sites and the
function of intervening amino acid sequences has been emphasized (2). Relevant chemistry of zinc appears to limit the protein sites of its interactions to the N, 0, and S donor groups of histidine, glutamic and aspartic acids, and cysteine (1, 2). However, the presence, absence, or combination of these ligands is not predictable on the basis of sequence analysis alone (3). X-ray crystallographic and/or NMR structure determinations are indispensable for this purpose (2). The NMR-based structure of the fungal transcription factor GAL41 of Saccharomyces cerevisiae (6) has already underscored the pertinence of these comments. Subsequently, both 1H and "3Cd NMR analyses of the DNA-binding domains of GAL4 (5-7) and the glucocorticoid receptor, GR, have been reported (8, 9). These, in turn, now represent standards of reference for the structures of the functional domains of other members of these protein families. Their zinc binding sites are distinctly different from those of zinc fingers and are referred to here, linguistically, as "zinc twists" in the GR type and "zinc clusters" in the GAL4 type.
MATERIALS AND METHODS The three-dimensional structures of GAL4 (5-7, 10) and of GR (8, 9) have been determined by 'H and 1'"Cd NMR. Literature searches for additional GAL4- and GR-like DNAbinding domain sequences utilized the National Biomedical Research Foundation Protein Identification Resource and translated GenBank data base files of the Molecular Biology Computer Research Resource at Harvard Medical School. Protein sequences were aligned and families were formed by the use of a pattern-induced multialignment algorithm for amino acid sequences (11).
RESULTS GR enhances or represses the transcription of numerous steroid-responsive genes by binding to specific DNA sequences named glucocorticoid-responsive elements upstream of the promoters for these genes. GRs, in fact, consist of three functional domains: an N-terminal one that modulates, a central zinc-containing one that binds the glucocorticoid responsive element, and a C-terminal one that binds hormones (12, 13). A GR receptor has been isolated from rat Abbreviations: GR, glucocorticoid receptor; TFIIIA, transcription factor IIIA. tTo whom reprint requests should be addressed. IThree-dimensional solution structures have now been obtained for both the zinc(II) and cadmium(II) derivatives of the DNA-binding domain of GAL4 (the 62 N-terminal residues) by two-dimensional NMR methods. Both zinc(II) and cadmium(II) form a binuclear cluster with ligands consisting of Cys-11, Cys-14, Cys-21, and Cys-31 to one metal ion and Cys-21, Cys-28, Cys-31, and Cys-38 to the second metal ion, with Cys-21 and Cys-31 acting as bridging ligands (5).
The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.
999
1000
Proc. Natl. Acad. Sci. USA 88 (1991)
Biochemistry: Vallee et al.
(14), mouse (15), and human tissues in a and 3 forms (16). The primary structure of the central DNA-binding domain in each of these proteins is the same. GR is a member of a superfamily of receptors that bind steroid and thyroid hormones, retinoic acid, and vitamin D3. Their DNA-binding domains contain a segment of 60 amino acids with nine conserved cysteines and one conserved histidine; in addition, in the steroid receptor subfamily, there is a tenth conserved cysteine. A 150-amino acid segment of human GR forms a domain that contains two zinc atoms. It extends approximately 40 amino acids beyond both ends of the minimal DNA-binding domain (17). 1H NMR (8) and 113Cd NMR (9) spectra of a 70-amino acid segment of this central domain reveal that eight cysteines and two zinc atoms form two tetrahedral zinc binding sites with an interatomic zinc-zinc distance of :13 A (8). The two other cysteine and the histidine residues are not ligands to these zinc binding sites. The 15-amino acid spacer that separates the two zinc atoms forms the DNA recognition site (counting from the last cysteine ligand of the first zinc to the first cysteine ligand of the second zinc). The human (18) and rat (19) mineralocorticoid receptors activate gene transcription in response to aldosterone. The primary structures of their DNA-binding domain are nearly identical to that of GR and the corresponding residues in the progesterone and androgen receptors isolated from human (20, 21), rat (22), chicken (23), and rabbit (24) (Fig. 1). Eight cysteines of the GR family are involved in the two zinc binding sites, which are surrounded by residues that are highly conserved (Fig. 1). Fifty-five of the residues (71%) of the 10 proteins are invariant, and the remainder are very similar. In all of these proteins, the region from 457 to 467 (numbering system of rat GR) is invariant; in the mammalian GR reference protein, this is the site of an a-helix (Fig. 2). A zinc finger, based on one zinc atom as described for transcription factor IIIA (TFIIIA) (4), is not present. The GAL4 transcription factor from S. cerevisiae regulates the expression of the genes encoding the galactosemetabolizing enzymes (25). It consists of 881 amino acid residues, but limited proteolysis localizes its DNA-binding domain to 62 residues of the N-terminal region (6, 7, 10). The N-terminal 38 amino acids of this domain include six cysteine residues; among these, four were predicted to form a Cys2Cys2 zinc finger (26). Two segments of GAL4, residues 1-147 and 1-62, bind two zinc or two cadmium atoms (6, 7, 10). Peptides from transcription factor LAC9 (see below) that 440
II]
443
,L
LV
GR
L
GR3 M R1
L
MR 3
457
\/
C)C;~~
4o 92
482?
I-
A G R N D
I N
PA
R
LV
s -r GS
K
L
A G R N D
N
PA
R
LV
S
T
GS
K
L
A G R N D
N
PA
R
LV
G
T
GS
K
L
A G R N D
NI
PA
R
L. V
G
T
GS
K
L
A G R N D
NI
PA
R
L
G
T
GS
K
L
A G R N D
I NI
PA
R
OS
K
L
A G R N D
NI
PA
R
A G R N D
I N
PA
R
4*
LI
G; T
PRS *
LI
G
T
GS
K
c I
495
K,
PR
-
(7)
conserve the six cysteines and the ligand spacing characteristics of GAL4 also contain 2 mols of zinc per mol of peptide (27). 'H and "3Cd NMR show that six cysteines of the GAL4 (1-62) segment interact with two zinc atoms to form a binuclear zinc thiolate cluster in which the interatomic zinczinc distance is -3.5 A. Cys-21 and Cys-31 are bridging cysteines, while the pairs Cys-11/-14 and Cys-28/-38 bind as monodentate ligands to the first and second zinc atoms, respectively (Fig. 3). The thiolate-metal cluster of metallothionein was the first and, thus far, only zinc/cadmium cluster ever to have been recognized in a natural product (28); that of GAL4 shown here is based on the same principles of coordination (Fig. 3). Again, the structure does not resemble the zinc finger structure proposed for native TFIIIA (4, 29). The six cysteines of GAL4 and their characteristic spacing are conserved in several other fungal transcription factors (Fig. 4). These include those that regulate lactose and galactose metabolism in Kluyveromyces lactis, LAC9 (30); pyrimidine metabolism in S. cerevisiae, PPR1 (31); arginine metabolism in S. cerevisiae, ARGRII (32); quinic acid metabolism in Aspergillus nidulans, QUTA (33); the fouraminobutyric acid (GABA) pathway in S. cerevisiae, UGA3 (34); and quinate/shikimate metabolism in Neurospora crassa, qa-iF (35). In the GAL4 family, the residues flanking the invariant cysteines are not well conserved (Fig. 4), though some similarity exists in the region between Cys-14 and Cys-28. In 4/6
PR
\
)
FIG. 2. Schematic of the DNA recognition site for the GR. The amino acids from 457 to 467 form an a-helix (8).
460
S] T
/
/
/
GS
I
3
1
/
AH't
T
LI
G
T
GS
K
L
A S R N D
F
N
PS
R
AR -
T
L
G
T
GS_
K
L
A S R N D
r N
PS
R
FIG. 1. Zinc ligands of the DNA-binding domain of steroid hormone receptors. The shaded area denotes the NMR standard of reference for each family. An asterisk denotes those for which zinc was not measured. Black vertical columns indicate the proposed metal-binding ligands based on the standard of reference. The sources of these proteins are indicated by arabic supercripts: 1) rat; 2) mouse; 3) human; 4) chicken; and 5) rabbit. MR, PR, and AR refer to mineralocorticoid, progesterone, and androgen receptors, respectively.
Proc. Natl. Acad. Sci. USA 88 (1991)
Biochemistry: Vallee et al.
60
34 37 33 36
5 7
57 59
41 44 48 50
1315 19 21 24 26
METALLOTHIONEIN
GAL4 FIG. 3. Zinc thiolate clusters of metallothionein (28) and GAL4. The positions of the bridging cysteinyl residues are identified by filled circles and those of singly bound cysteines by open circles. The zinc atoms are designated by the Roman numerals. The cysteine-zinc connectivity of GAL4 is taken from the complete sequential assignment data and calculated three-dimensional structure (5).
all eight proteins, residues 15, 17, 18, and 20 are nearly always lysine or arginine; only one noncysteine residue, Pro-26, is strictly conserved. In the case of GAL4, mutation of this proline residue results in a transcription factor that binds zinc much less tightly (26). The functional consequence of this marked diversity is unknown but may signal a markedly different structure for this zinc cluster site.
DISCUSSION Evidence that zinc is crucial to DNA and RNA synthesis and cell division emerged in the seventies (36-39). The finding of Wu and colleagues (40) that Xenopus TFIIIA, which activates the transcription of the 5S RNA gene, is a zinc protein that contains from 2 to 3 mol of zinc per mol of protein focused interest on the role of zinc in the process of transcription. Subsequently, the sequence of TFIIIA was shown to contain nine repeat sequences of about 30 amino acids in each of which two cysteine and two histidine residues are conserved (29). Each 7S particle contains 7-11 zinc atoms (29), although this stoichiometry has been debated (41, 42). The two cysteine and two histidine residues per 30-amino acid unit were proposed to form a tetrahedral coordination complex with each of the nine zinc atoms to generate peptide domains-zinc fingers (4)-that interact with DNA. A series
of observations that includes limited propteolytic degradation, extended x-ray absorption fine structure measurements (43), synthesis of peptide domains and NMR of their zinc complexes (44-47), absorption spectroscopy of the corresponding cobalt complexes (48), and computer modeling (49) has been thought to be consistent with the proposed structure and its functional role. The implications have generated extraordinary experimental effort in the field of DNA-binding proteins. Based on the TFIIIA model, numerous sequences have been reported, many of which have been thought to reflect both structural and functional similarities. Ensuing generalizations have led to a revision of past ideas as well as predictions of new principles and mechanisms describing DNA-protein interactions. Zinc finger has become a widely used description of this model to encapsulate all of these implications and inferences, thereby constituting a convenient linguistic vehicle that conveys the underlying chemical, biological, and mechanistic details. As a consequence, the term has been used to describe virtually any relatively short sequence that contains four or more cysteine and/or histidine residues and is believed to function as a nucleic acid-binding domain (3). The strictest classification would confine the term zinc finger to those proteins that (i) have one or multiple repeats of about 30 amino acids each, (ii) conserve both two cysteine and two histidine residues and their spacing, (iii) feature the two aromatic residues and the leucine highlighted in the TFIIIA repeats (50), and (iv) presumably have a threedimensional structure that resembles a finger. Further classifications have been proposed, such as "classical" and "nonclassical" zinc fingers as well as a subclassification based on sequence homology (51). However, zinc finger has rapidly become a part of the language and a concept conveying even more general ideas. While the structures of zinc fingers have been inferred and thought to be confirmed by organic syntheses, direct structure determinations by x-ray or NMR of proteins containing multiple TFIIIA-like structures have not as yet been reported, though it is but a matter of time until they will be. All of our past efforts to identify and define the zinc binding sites of enzymes have been based on three-dimensional structures of a given protein as determined by x-ray or NMR methods that then served as the template for comparison with other members of its family. In the absence of such data for DNA-binding zinc proteins, similar comparisons could not be undertaken. In the present discussion, we adopt the "zinc finger model" as originally proposed for the multiple zinc domains of TFIIIA (4, 29) and extended since then to other proteins with similarly spaced Cys2-His2 sequences. How-
14
11
28 I21 E 31 I,.-
38
LAC9
HQAJ
DA
I
RKKKWK
S KTVPT
TN
LKYN LO
VYS
LAC9A
HQA
DA
R KKKWK
S KRVP T
TN
LKYN LO
VYS
PPR1
RTA
KR
RLKK I K
DQEFPS
KR
AKLEVP
VS L
PPR1A
RTA
KR
RLKK I K
DQEFPS
KR
AKLEVP
VS L
A QUTA *SR *
D GAQP I
ST
DLRHPH
QR
ASLS RP EKSN LP
TYR
ARGRII*
DS RSKKDK FTG WT RGRKVK
DIT K I RKKR
SEDKPV
RD
RRLS FP
DGI OPA
FP
VSQG RS TYQ
I
UAG3
*
qa-1 F
*SRA DO
KHG
RAAREK
1001
GGY I YI
FIG. 4. Zinc ligands of the DNA-binding domains of fungal transcription factors. For the key to the figure, see the legend to Fig. 1.
1002
Biochemistry: Vallee et al.
Proc. Natl. Acad. Sci. USA 88 (1991)
ever, the application of the single-finger model to the multifinger transcription factors should await conclusive structural data on functional DNA-binding domain(s) of the protein. NMR structures of GAL4 and GR have now been determined and allow a comparison of consensus sequences of their respective families in a manner completely analogous to that previously presented for zinc enzymes (1, 2, 52, 53). In both GR and GAL4, cysteine residues have been known to be the only zinc ligands (Figs. 1 and 4). Many of the inferences regarding the mode of zinc binding sites drawn from the consensus sequences are at variance with those that the NMR structures now demonstrate. This calls for a reevaluation of earlier views and comparisons with the NMR data now available. The chemical details and the stoichiometry of the zinc binding sites in all of these DNA-binding proteins differ specifically and characteristically from those of the zinc enzymes previously communicated (1, 2, 52). Importantly, virtually all DNA-binding zinc proteins contain multiple zinc atoms. Hence, the interatomic zinc-zinc distances can be measured and may constitute a parameter of potential significance. In GAL4, two zinc atoms form a zinc cluster with six cysteine residues binding each of two zinc atoms. The zinc-zinc distance is -3.5 A (Fig. 5). GR also contains two zinc atoms, but these form two separate and distinct sites. In each site, four different cysteine ligands bind to one zinc atom, and the zinc-zinc interatomic distance is ==13 A (Fig. 5). In TFIIIA, these inter-zinc distances are not known, since a three-dimensional structure has not been reported. It may well be that these zinc-zinc distances are also characteristic as, perhaps, are those of other binding modes yet to be recognized. It is particularly important to emphasize the difference in the mode of binding of the GR proteins from that of the zinc fingers. The DNA-binding site of GR is located between and anchored by the two zinc atom complexes (Fig. 2). In zinc fingers, the proposed DNA recognition site is between two successive pairs of ligands to one zinc atom. It follows that the former binding mode can only be expected in proteins containing at least two zinc atoms, while the latter can occur in proteins containing one or more zinc atoms.
DOMAIN ZINC FINGER
TFIIIA
C®
I
ZINC TWIST
C
0
i
\
/ci
C
/
Glu Rec
13
GAL4
3.5
X C
ZINC CLUSTER
PROTEIN dZn-Zn
STRUCTURE
C
FIG. 5. The characteristics of the zinc binding sites of DNAbinding domains of transcriptional factors. The zinc-zinc interatomic distance, dz,,zn, is given in A.
The DNA-binding motifofGR appears to be helical (8), and we have therefore described it as a "twist," although it need not necessarily be an a-helix in other proteins, of course. While the cysteine spacings-2, 6, 6, and 2-in the first zinc domain of GR and that of GAL4 are identical, the structure of their zinc complexes differ significantly (Fig. 5). Beyond that, the cysteine spacings 5, 9, 2, and 4 for the second zinc domain of GR differ from all of the above. 1 The fact that some cysteine residues (e.g., in GRs) do not serve as ligands and the capacity of some cysteine residues to form bridging ligands between two zinc atoms (Fig. 4) virtually precludes structural predictions based solely on primary structures. At present, the number of structure determinations of these DNA-binding zinc proteins would make it premature to extrapolate tertiary structure and its function broadly. Extensions of the results on GAL4 and GR to other members of their families are clearly warranted, as shown in Figs. 1 and 4. However, other proteins containing multiple zinc atoms with ligands that may include variable numbers of ligands such as glutamic or aspartic acid and histidine require additional structural templates. In proteins containing multiple zinc centers, the zinc-zinc distances may well prove to be important indices of both overall structure and function of any given zinc protein. We now recognize three distinct motifs of DNA-binding zinc proteins: (i) zinc fingers, (ii) zinc clusters, and (iii) zinc twists. As additional structural data become available, their implication to the function and modes of interaction of these proteins both with DNA as well as other proteins will become clear. The detection of yet additional motifs revealing other permutations of zinc chemistry may be expected. 'The structural zinc atom of alcohol dehydrogenase is arranged in a tetrahedral site formed by four cysteine residues; the spacing between its cysteine ligands is 2, 2, and 7 amino acid residues (2). Apart from the fact that this arrangement concerns only one zinc atom per subunit, it also differs significantly from that found in GAL4 and GR.
This work was supported by a grant from the Endowment for Research in Human Biology, Inc. (Boston) (to B.L.V. and D.S.A.) and Grants DK09070 and GM21919 from the National Institutes of Health (to J.E.C.). 1. Vallee, B. L. & Auld, D. S. (1990) Proc. Natl. Acad. Sci. USA 87, 220-224. 2. Vallee, B. L. & Auld, D. S. (1990) Biochemistry, 29, 56475659. 3. Berg, J. M. (1986) Science 232, 485-487. 4. Klug, A. & Rhodes, D. (1987) Trends Biochem. Sci. 12, 464-469. 5. Pan, T. & Coleman, J. E. (1991) Biochemistry 30, in press. 6. Pan, T. & Coleman, J. E. (1990) Proc. Natl. Acad. Sci. USA 87, 2077-2081. 7. Pan, T. & Coleman, J. E. (1990) Biochemistry 29, 3023-3029. 8. Hard, T., Kellenbach, E., Boelens, R., Maler, B. A., Dahlman, K., Freedman, L. P., Carlstedt-Duke, J., Yamamoto, K., Gustafsson, J.-A. & Kaptein, R. (1990) Science 249, 157-160. 9. Pan, T., Freedman, L. P. & Coleman, J. E. (1990) Biochemistry 29, 9218-9225. 10. Pan, T. & Coleman, J. E. (1989) Proc. Natl. Acad. Sci. USA 86, 3145-3149. 11. Smith, R. F. & Smith, T. F. (1990) Proc. Natl. Acad. Sci. USA 87, 118-122. 12. Evans, R. M. (1988) Science 240, 889-895. 13. Beato, M. (1989) Cell 56, 335-344. 14. Chang, C., Kokontis, J., Chang, C.-T. & Liao, S. (1987) Nucleic Acids Res. 15, 9603. 15. Danielsen, M., Northrop, J. P. & Ringold, G. M. (1986) EMBO J. 5, 2513-2522. 16. Hollenberg, S. M., Weinberger, C., Ong, E. S., Cerelli, G., Oro, A., Lebo, R., Thompson, R. B., Rosenfeld, M. G. & Evans, R. M. (1985) Nature (London) 318, 670-672.
Biochemistry: Vallee et al. 17. Freedman, L. P., Luisi, B. F., Korszun, Z. R., Basavappa, R., Sigler, P. B. & Yamamoto, K. R. (1988) Nature (London) 334, 543-546. 18. Arriza, J. L., Weinberger, C., Cerelli, G., Glaser, T. M., Handelin, B. L., Housman, D. E. & Evans, R. M. (1987) Science 237, 268-275. 19. Patel, P. D., Sherman, T. G., Goldman, D. J. & Watson, S. J. (1988) Mol. Endocrinol. 3, 1877-1885. 20. Misrahi, M., Atger, M., D'Auriol, L., Loosfelt, H., Meriel, C., Fridlansky, F., Guiochon-Mantel, A., Galibert, F. & Milgrom, E. (1987) Biochem. Biophys. Res. Commun. 143, 740-748. 21. Tilley, W. D., Marcelli, M., Wilson, J. D. & McPhaul, M. J.
(1989) Proc. Natl. Acad. Sci. USA 86, 327-331.
22. Yarbrough, W. G., Quarmby, V. E., Simental, J. A., Joseph, D. R., Sar, M., Lubahn, D. B., Olsen, K. L., French, F. S. & Wilson, E. M. (1990) J. Biol. Chem. 265, 8893-8900. 23. Gronemeyer, H., Turcotte, B., Quirin-Stricker, C., Bocquel, M. T., Meyer, M. E., Krozowski, Z., Jeltsch, J. M., Lerouge, T., Gamier, J. M. & Chambon, P. (1987) EMBO J. 6, 39853994. 24. Loosfelt, H., Atger, M., Misrahi, M., Guiochon-Mantel, A., Meriel, C., Logeat, F., Benarous, R. & Milgrom, E. (1986) Proc. Natl. Acad. Sci. USA 83, 9045-9049. 25. Laughon, A. & Gesteland, R. F. (1984) Mol. Cell. Biol. 4, 260-267. 26. Johnston, M. (1987) Nature (London) 328, 353-355. 27. Halvorsen, Y.-D. C., Nandabalan, K. & Dickson, R. C. (1990) J. Biol. Chem. 265, 13283-13289. 28. Schultze, P., Worgotter, E., Braun, W., Wagner, G., Vasdk, M., Kagi, J. H. R. & Wuthrich, K. (1988) J. Mol. Biol. 203, 251-268. 29. Miller, J., McLachlan, A. D. & Klug, A. (1985) EMBO J. 4, 1609-1614. 30. Salmeron, J. M., Jr., & Johnston, S. A. (1986) Nucleic Acids Res. 14, 7767-7781. 31. Kammerer, B., Guyonvarch, A. & Hubert, J. C. (1984) J. Mol. Biol. 180, 239-250. 32. Messenguy, F., Dubois, E. & Descamps, F. (1986) Eur. J. Biochem. 157, 77-81. 33. Beri, R. K., Whittington, H., Roberts, C. F. & Hawkins, A. R. (1987) Nucleic Acids Res. 15, 7991-8001.
Proc. Natl. Acad. Sci. USA 88 (1991)
1003
34. Andre, B. (1990) Mol. Gen. Genet. 220, 269-276. 35. Geever, R. E., Huiet, L., Baum, J. A., Tyler, B. M., Patel, V. B., Rutledge, B. J., Case, M. E. & Giles, N. H. (1989) J. Mol. Biol. 207, 15-34. 36. Vallee, B. L. (1977) Experientia 33, 600. 37. Vallee, B. L. (1977) in Biological Aspects of Inorganic Chemistry, ed. Dolphin, D. (Wiley, New York), pp. 37-70. 38. Auld, D. S. (1979) Adv. Chem. Ser. 172, 112-133. 39. Vallee, B. L. & Falchuk, K. F. (1981) Philos. Trans. R. Soc. London Ser. B 294, 185-197. 40. Hanas, J. S., Hazuda, D. J., Bogenhagen, D. F., Wu, F. Y.-H. & Wu, C.-W. (1983) J. Biol. Chem. 258, 14120-14125. 41. Shang, Z., Liao, Y.-D., Wu, F. Y.-H. & Wu, C.-W. (1989) Biochemistry 28, 9790-9795. 42. Han, M. K., Cyran, F. P., Fischer, M. T., Kim, S. H. & Ginsburg, A. (1990) J. Biol. Chem. 265, 13792-13799. 43. Diakun, G. P., Fairall, L. & Klug, A. (1986) Nature (London) 324, 698-699. 44. Lee, M. S., Gippert, G. P., Soman, K. V., Case, D. A. & Wright, P. E. (1989) Science 245, 635-637. 45. Carr, M. D., Pastore, A., Gausepohl, H., Frank, R. & Roesch, P. (1990) Eur. J. Biochem. 188, 455-461. 46. Neuhaus, D., Nakaseko, Y., Nagai, K. & Klug, A. (1990) FEBS Lett. 262, 179-184. 47. Pdrraga, G., Horvath, S., Hood, L., Young, E. T. & Klevit, R. E. (1990) Proc. Natl. Acad. Sci. USA 87, 137-141. 48. Frankel, A. D., Berg, J. M. & Pabo, C. 0. (1987) Proc. Natl. Acad. Sci. USA 84, 4841-4845. 49. Gibson, T. J., Postma, J. P. M., Brown, R. S. & Argos, P. (1988) Protein Eng. 2, 209-218. 50. Frankel, A. D., Bredt, D. S. & Pabo, C. 0. (1988) Science 240, 70-73. 51. South, T. L., Blake, P. R., Sowder, R. C., III, Arthur, L. O., Henderson, L. E. & Summers, M. F. (1990) Biochemistry 29, 7786-7789. 52. Vallee, B. L. & Auld, D. S. (1989) FEBS Lett. 257, 138-140. 53. Vallee, B. L. & Auld, D. S. (1991) in Matrix Metalloproteinases and Inhibitors, eds. Birkedal-Hansen, H., Werb, Z., Welgus, H. & Van Wart, H. (Fischer, Stuttgart, F.R.G.), in press.
Zinc fingers, zinc clusters, and zinc twists in DNA-binding protein domains (transcription factors/steroid receptors/hormone receptors/GAL4/zinc binding site)
BERT L. VALLEE*t, JOSEPH E.
COLEMANt, AND DAVID S. AULD*§
*Center for Biochemical and Biophysical Sciences and Medicine and §Department of Pathology, Harvard Medical School, Brigham and Women's Hospital, 250 Longwood Avenue, Boston, MA 02115; and tDepartment of Molecular Biophysics and Biochemistry, Yale University, New Haven, CT 06510
Contributed by Bert L. Vallee, October 25, 1990
We now recognize three distinct motifs of ABSTRACT DNA-binding zinc proteins: (i) zinc ringers, (ii) zinc clusters, and (iii) zinc twists. Until very recently, x-ray crystallographic or NMR three-dimensional structure analyses of DNA-binding zinc proteins have not been available to serve as standards of reference for the zinc binding sites of these families of proteins. Those of the DNA-binding domains of the fungal transcription factor GAL4 and the rat glucocorticoid receptor are the frst to have been determined. Both proteins contain two zinc binding sites, and in both, cysteine residues are the sole zinc ligands. In GAL4, two zinc atoms are bound to six cysteine residues which form a "zinc cluster" akin to that of metaflothionein; the distance between the two zinc atoms of GAL4 is "'3.5 A. In the glucocorticoid receptor, each zinc atom is bound to four cysteine residues; the interatomic zinc-zinc distance is - 13 A, and in this instance, a "zinc twist" is represented by a helical DNA recognition site located between the two zinc atoms. Zinc clusters and zinc twists are here recognized as two distinctive motifs in DNA-binding proteins containing multiple zinc atoms. For native "zinc ringers," structural data do not exist as yet; consequently, the interatomic distances between zinc atoms are not known. As further structural data become available, the structural and functional significance of these different motifs in their binding to DNA and other proteins participating in the transmiission of the genetic message will become apparent.
The three-dimensional structures of 12 zinc enzymes and their zinc binding sites have been established by x-ray crystallographic analysis. This has focused attention on the zinc chemistry underlying their catalytic functions (1). Their zinc binding sites are characteristic of those of other members of their respective families (1, 2) and show that primary structures alone are unsuitable for such comparisons. The results of such studies constitute compelling evidence regarding the minimum information required to recognize and identify zinc binding sites of other zinc proteins. A functionally heterogeneous group of DNA-binding proteins that includes transcription factors, retroviral low molecular weight nucleic acid-binding proteins, adenovirus EJA gene products, aminoacyl tRNA synthetases, large tumor (T) antigen, bacteriophage proteins, and hormone receptors has been inferred to be zinc proteins based solely on their primary structures (3). Combinations of conserved cysteine and histidine residues separated by variable numbers of amino acids have been formulated into a popular model known as a "zinc finger" (4), in which the zinc chelate forms its "knuckle," as it were. Patterns such as these are thought further to signal the presence ofzinc. The desirability ofzinc analysis together with tertiary structure determinations to confirm such predictions by documenting the zinc binding sites and the
function of intervening amino acid sequences has been emphasized (2). Relevant chemistry of zinc appears to limit the protein sites of its interactions to the N, 0, and S donor groups of histidine, glutamic and aspartic acids, and cysteine (1, 2). However, the presence, absence, or combination of these ligands is not predictable on the basis of sequence analysis alone (3). X-ray crystallographic and/or NMR structure determinations are indispensable for this purpose (2). The NMR-based structure of the fungal transcription factor GAL41 of Saccharomyces cerevisiae (6) has already underscored the pertinence of these comments. Subsequently, both 1H and "3Cd NMR analyses of the DNA-binding domains of GAL4 (5-7) and the glucocorticoid receptor, GR, have been reported (8, 9). These, in turn, now represent standards of reference for the structures of the functional domains of other members of these protein families. Their zinc binding sites are distinctly different from those of zinc fingers and are referred to here, linguistically, as "zinc twists" in the GR type and "zinc clusters" in the GAL4 type.
MATERIALS AND METHODS The three-dimensional structures of GAL4 (5-7, 10) and of GR (8, 9) have been determined by 'H and 1'"Cd NMR. Literature searches for additional GAL4- and GR-like DNAbinding domain sequences utilized the National Biomedical Research Foundation Protein Identification Resource and translated GenBank data base files of the Molecular Biology Computer Research Resource at Harvard Medical School. Protein sequences were aligned and families were formed by the use of a pattern-induced multialignment algorithm for amino acid sequences (11).
RESULTS GR enhances or represses the transcription of numerous steroid-responsive genes by binding to specific DNA sequences named glucocorticoid-responsive elements upstream of the promoters for these genes. GRs, in fact, consist of three functional domains: an N-terminal one that modulates, a central zinc-containing one that binds the glucocorticoid responsive element, and a C-terminal one that binds hormones (12, 13). A GR receptor has been isolated from rat Abbreviations: GR, glucocorticoid receptor; TFIIIA, transcription factor IIIA. tTo whom reprint requests should be addressed. IThree-dimensional solution structures have now been obtained for both the zinc(II) and cadmium(II) derivatives of the DNA-binding domain of GAL4 (the 62 N-terminal residues) by two-dimensional NMR methods. Both zinc(II) and cadmium(II) form a binuclear cluster with ligands consisting of Cys-11, Cys-14, Cys-21, and Cys-31 to one metal ion and Cys-21, Cys-28, Cys-31, and Cys-38 to the second metal ion, with Cys-21 and Cys-31 acting as bridging ligands (5).
The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.
999
1000
Proc. Natl. Acad. Sci. USA 88 (1991)
Biochemistry: Vallee et al.
(14), mouse (15), and human tissues in a and 3 forms (16). The primary structure of the central DNA-binding domain in each of these proteins is the same. GR is a member of a superfamily of receptors that bind steroid and thyroid hormones, retinoic acid, and vitamin D3. Their DNA-binding domains contain a segment of 60 amino acids with nine conserved cysteines and one conserved histidine; in addition, in the steroid receptor subfamily, there is a tenth conserved cysteine. A 150-amino acid segment of human GR forms a domain that contains two zinc atoms. It extends approximately 40 amino acids beyond both ends of the minimal DNA-binding domain (17). 1H NMR (8) and 113Cd NMR (9) spectra of a 70-amino acid segment of this central domain reveal that eight cysteines and two zinc atoms form two tetrahedral zinc binding sites with an interatomic zinc-zinc distance of :13 A (8). The two other cysteine and the histidine residues are not ligands to these zinc binding sites. The 15-amino acid spacer that separates the two zinc atoms forms the DNA recognition site (counting from the last cysteine ligand of the first zinc to the first cysteine ligand of the second zinc). The human (18) and rat (19) mineralocorticoid receptors activate gene transcription in response to aldosterone. The primary structures of their DNA-binding domain are nearly identical to that of GR and the corresponding residues in the progesterone and androgen receptors isolated from human (20, 21), rat (22), chicken (23), and rabbit (24) (Fig. 1). Eight cysteines of the GR family are involved in the two zinc binding sites, which are surrounded by residues that are highly conserved (Fig. 1). Fifty-five of the residues (71%) of the 10 proteins are invariant, and the remainder are very similar. In all of these proteins, the region from 457 to 467 (numbering system of rat GR) is invariant; in the mammalian GR reference protein, this is the site of an a-helix (Fig. 2). A zinc finger, based on one zinc atom as described for transcription factor IIIA (TFIIIA) (4), is not present. The GAL4 transcription factor from S. cerevisiae regulates the expression of the genes encoding the galactosemetabolizing enzymes (25). It consists of 881 amino acid residues, but limited proteolysis localizes its DNA-binding domain to 62 residues of the N-terminal region (6, 7, 10). The N-terminal 38 amino acids of this domain include six cysteine residues; among these, four were predicted to form a Cys2Cys2 zinc finger (26). Two segments of GAL4, residues 1-147 and 1-62, bind two zinc or two cadmium atoms (6, 7, 10). Peptides from transcription factor LAC9 (see below) that 440
II]
443
,L
LV
GR
L
GR3 M R1
L
MR 3
457
\/
C)C;~~
4o 92
482?
I-
A G R N D
I N
PA
R
LV
s -r GS
K
L
A G R N D
N
PA
R
LV
S
T
GS
K
L
A G R N D
N
PA
R
LV
G
T
GS
K
L
A G R N D
NI
PA
R
L. V
G
T
GS
K
L
A G R N D
NI
PA
R
L
G
T
GS
K
L
A G R N D
I NI
PA
R
OS
K
L
A G R N D
NI
PA
R
A G R N D
I N
PA
R
4*
LI
G; T
PRS *
LI
G
T
GS
K
c I
495
K,
PR
-
(7)
conserve the six cysteines and the ligand spacing characteristics of GAL4 also contain 2 mols of zinc per mol of peptide (27). 'H and "3Cd NMR show that six cysteines of the GAL4 (1-62) segment interact with two zinc atoms to form a binuclear zinc thiolate cluster in which the interatomic zinczinc distance is -3.5 A. Cys-21 and Cys-31 are bridging cysteines, while the pairs Cys-11/-14 and Cys-28/-38 bind as monodentate ligands to the first and second zinc atoms, respectively (Fig. 3). The thiolate-metal cluster of metallothionein was the first and, thus far, only zinc/cadmium cluster ever to have been recognized in a natural product (28); that of GAL4 shown here is based on the same principles of coordination (Fig. 3). Again, the structure does not resemble the zinc finger structure proposed for native TFIIIA (4, 29). The six cysteines of GAL4 and their characteristic spacing are conserved in several other fungal transcription factors (Fig. 4). These include those that regulate lactose and galactose metabolism in Kluyveromyces lactis, LAC9 (30); pyrimidine metabolism in S. cerevisiae, PPR1 (31); arginine metabolism in S. cerevisiae, ARGRII (32); quinic acid metabolism in Aspergillus nidulans, QUTA (33); the fouraminobutyric acid (GABA) pathway in S. cerevisiae, UGA3 (34); and quinate/shikimate metabolism in Neurospora crassa, qa-iF (35). In the GAL4 family, the residues flanking the invariant cysteines are not well conserved (Fig. 4), though some similarity exists in the region between Cys-14 and Cys-28. In 4/6
PR
\
)
FIG. 2. Schematic of the DNA recognition site for the GR. The amino acids from 457 to 467 form an a-helix (8).
460
S] T
/
/
/
GS
I
3
1
/
AH't
T
LI
G
T
GS
K
L
A S R N D
F
N
PS
R
AR -
T
L
G
T
GS_
K
L
A S R N D
r N
PS
R
FIG. 1. Zinc ligands of the DNA-binding domain of steroid hormone receptors. The shaded area denotes the NMR standard of reference for each family. An asterisk denotes those for which zinc was not measured. Black vertical columns indicate the proposed metal-binding ligands based on the standard of reference. The sources of these proteins are indicated by arabic supercripts: 1) rat; 2) mouse; 3) human; 4) chicken; and 5) rabbit. MR, PR, and AR refer to mineralocorticoid, progesterone, and androgen receptors, respectively.
Proc. Natl. Acad. Sci. USA 88 (1991)
Biochemistry: Vallee et al.
60
34 37 33 36
5 7
57 59
41 44 48 50
1315 19 21 24 26
METALLOTHIONEIN
GAL4 FIG. 3. Zinc thiolate clusters of metallothionein (28) and GAL4. The positions of the bridging cysteinyl residues are identified by filled circles and those of singly bound cysteines by open circles. The zinc atoms are designated by the Roman numerals. The cysteine-zinc connectivity of GAL4 is taken from the complete sequential assignment data and calculated three-dimensional structure (5).
all eight proteins, residues 15, 17, 18, and 20 are nearly always lysine or arginine; only one noncysteine residue, Pro-26, is strictly conserved. In the case of GAL4, mutation of this proline residue results in a transcription factor that binds zinc much less tightly (26). The functional consequence of this marked diversity is unknown but may signal a markedly different structure for this zinc cluster site.
DISCUSSION Evidence that zinc is crucial to DNA and RNA synthesis and cell division emerged in the seventies (36-39). The finding of Wu and colleagues (40) that Xenopus TFIIIA, which activates the transcription of the 5S RNA gene, is a zinc protein that contains from 2 to 3 mol of zinc per mol of protein focused interest on the role of zinc in the process of transcription. Subsequently, the sequence of TFIIIA was shown to contain nine repeat sequences of about 30 amino acids in each of which two cysteine and two histidine residues are conserved (29). Each 7S particle contains 7-11 zinc atoms (29), although this stoichiometry has been debated (41, 42). The two cysteine and two histidine residues per 30-amino acid unit were proposed to form a tetrahedral coordination complex with each of the nine zinc atoms to generate peptide domains-zinc fingers (4)-that interact with DNA. A series
of observations that includes limited propteolytic degradation, extended x-ray absorption fine structure measurements (43), synthesis of peptide domains and NMR of their zinc complexes (44-47), absorption spectroscopy of the corresponding cobalt complexes (48), and computer modeling (49) has been thought to be consistent with the proposed structure and its functional role. The implications have generated extraordinary experimental effort in the field of DNA-binding proteins. Based on the TFIIIA model, numerous sequences have been reported, many of which have been thought to reflect both structural and functional similarities. Ensuing generalizations have led to a revision of past ideas as well as predictions of new principles and mechanisms describing DNA-protein interactions. Zinc finger has become a widely used description of this model to encapsulate all of these implications and inferences, thereby constituting a convenient linguistic vehicle that conveys the underlying chemical, biological, and mechanistic details. As a consequence, the term has been used to describe virtually any relatively short sequence that contains four or more cysteine and/or histidine residues and is believed to function as a nucleic acid-binding domain (3). The strictest classification would confine the term zinc finger to those proteins that (i) have one or multiple repeats of about 30 amino acids each, (ii) conserve both two cysteine and two histidine residues and their spacing, (iii) feature the two aromatic residues and the leucine highlighted in the TFIIIA repeats (50), and (iv) presumably have a threedimensional structure that resembles a finger. Further classifications have been proposed, such as "classical" and "nonclassical" zinc fingers as well as a subclassification based on sequence homology (51). However, zinc finger has rapidly become a part of the language and a concept conveying even more general ideas. While the structures of zinc fingers have been inferred and thought to be confirmed by organic syntheses, direct structure determinations by x-ray or NMR of proteins containing multiple TFIIIA-like structures have not as yet been reported, though it is but a matter of time until they will be. All of our past efforts to identify and define the zinc binding sites of enzymes have been based on three-dimensional structures of a given protein as determined by x-ray or NMR methods that then served as the template for comparison with other members of its family. In the absence of such data for DNA-binding zinc proteins, similar comparisons could not be undertaken. In the present discussion, we adopt the "zinc finger model" as originally proposed for the multiple zinc domains of TFIIIA (4, 29) and extended since then to other proteins with similarly spaced Cys2-His2 sequences. How-
14
11
28 I21 E 31 I,.-
38
LAC9
HQAJ
DA
I
RKKKWK
S KTVPT
TN
LKYN LO
VYS
LAC9A
HQA
DA
R KKKWK
S KRVP T
TN
LKYN LO
VYS
PPR1
RTA
KR
RLKK I K
DQEFPS
KR
AKLEVP
VS L
PPR1A
RTA
KR
RLKK I K
DQEFPS
KR
AKLEVP
VS L
A QUTA *SR *
D GAQP I
ST
DLRHPH
QR
ASLS RP EKSN LP
TYR
ARGRII*
DS RSKKDK FTG WT RGRKVK
DIT K I RKKR
SEDKPV
RD
RRLS FP
DGI OPA
FP
VSQG RS TYQ
I
UAG3
*
qa-1 F
*SRA DO
KHG
RAAREK
1001
GGY I YI
FIG. 4. Zinc ligands of the DNA-binding domains of fungal transcription factors. For the key to the figure, see the legend to Fig. 1.
1002
Biochemistry: Vallee et al.
Proc. Natl. Acad. Sci. USA 88 (1991)
ever, the application of the single-finger model to the multifinger transcription factors should await conclusive structural data on functional DNA-binding domain(s) of the protein. NMR structures of GAL4 and GR have now been determined and allow a comparison of consensus sequences of their respective families in a manner completely analogous to that previously presented for zinc enzymes (1, 2, 52, 53). In both GR and GAL4, cysteine residues have been known to be the only zinc ligands (Figs. 1 and 4). Many of the inferences regarding the mode of zinc binding sites drawn from the consensus sequences are at variance with those that the NMR structures now demonstrate. This calls for a reevaluation of earlier views and comparisons with the NMR data now available. The chemical details and the stoichiometry of the zinc binding sites in all of these DNA-binding proteins differ specifically and characteristically from those of the zinc enzymes previously communicated (1, 2, 52). Importantly, virtually all DNA-binding zinc proteins contain multiple zinc atoms. Hence, the interatomic zinc-zinc distances can be measured and may constitute a parameter of potential significance. In GAL4, two zinc atoms form a zinc cluster with six cysteine residues binding each of two zinc atoms. The zinc-zinc distance is -3.5 A (Fig. 5). GR also contains two zinc atoms, but these form two separate and distinct sites. In each site, four different cysteine ligands bind to one zinc atom, and the zinc-zinc interatomic distance is ==13 A (Fig. 5). In TFIIIA, these inter-zinc distances are not known, since a three-dimensional structure has not been reported. It may well be that these zinc-zinc distances are also characteristic as, perhaps, are those of other binding modes yet to be recognized. It is particularly important to emphasize the difference in the mode of binding of the GR proteins from that of the zinc fingers. The DNA-binding site of GR is located between and anchored by the two zinc atom complexes (Fig. 2). In zinc fingers, the proposed DNA recognition site is between two successive pairs of ligands to one zinc atom. It follows that the former binding mode can only be expected in proteins containing at least two zinc atoms, while the latter can occur in proteins containing one or more zinc atoms.
DOMAIN ZINC FINGER
TFIIIA
C®
I
ZINC TWIST
C
0
i
\
/ci
C
/
Glu Rec
13
GAL4
3.5
X C
ZINC CLUSTER
PROTEIN dZn-Zn
STRUCTURE
C
FIG. 5. The characteristics of the zinc binding sites of DNAbinding domains of transcriptional factors. The zinc-zinc interatomic distance, dz,,zn, is given in A.
The DNA-binding motifofGR appears to be helical (8), and we have therefore described it as a "twist," although it need not necessarily be an a-helix in other proteins, of course. While the cysteine spacings-2, 6, 6, and 2-in the first zinc domain of GR and that of GAL4 are identical, the structure of their zinc complexes differ significantly (Fig. 5). Beyond that, the cysteine spacings 5, 9, 2, and 4 for the second zinc domain of GR differ from all of the above. 1 The fact that some cysteine residues (e.g., in GRs) do not serve as ligands and the capacity of some cysteine residues to form bridging ligands between two zinc atoms (Fig. 4) virtually precludes structural predictions based solely on primary structures. At present, the number of structure determinations of these DNA-binding zinc proteins would make it premature to extrapolate tertiary structure and its function broadly. Extensions of the results on GAL4 and GR to other members of their families are clearly warranted, as shown in Figs. 1 and 4. However, other proteins containing multiple zinc atoms with ligands that may include variable numbers of ligands such as glutamic or aspartic acid and histidine require additional structural templates. In proteins containing multiple zinc centers, the zinc-zinc distances may well prove to be important indices of both overall structure and function of any given zinc protein. We now recognize three distinct motifs of DNA-binding zinc proteins: (i) zinc fingers, (ii) zinc clusters, and (iii) zinc twists. As additional structural data become available, their implication to the function and modes of interaction of these proteins both with DNA as well as other proteins will become clear. The detection of yet additional motifs revealing other permutations of zinc chemistry may be expected. 'The structural zinc atom of alcohol dehydrogenase is arranged in a tetrahedral site formed by four cysteine residues; the spacing between its cysteine ligands is 2, 2, and 7 amino acid residues (2). Apart from the fact that this arrangement concerns only one zinc atom per subunit, it also differs significantly from that found in GAL4 and GR.
This work was supported by a grant from the Endowment for Research in Human Biology, Inc. (Boston) (to B.L.V. and D.S.A.) and Grants DK09070 and GM21919 from the National Institutes of Health (to J.E.C.). 1. Vallee, B. L. & Auld, D. S. (1990) Proc. Natl. Acad. Sci. USA 87, 220-224. 2. Vallee, B. L. & Auld, D. S. (1990) Biochemistry, 29, 56475659. 3. Berg, J. M. (1986) Science 232, 485-487. 4. Klug, A. & Rhodes, D. (1987) Trends Biochem. Sci. 12, 464-469. 5. Pan, T. & Coleman, J. E. (1991) Biochemistry 30, in press. 6. Pan, T. & Coleman, J. E. (1990) Proc. Natl. Acad. Sci. USA 87, 2077-2081. 7. Pan, T. & Coleman, J. E. (1990) Biochemistry 29, 3023-3029. 8. Hard, T., Kellenbach, E., Boelens, R., Maler, B. A., Dahlman, K., Freedman, L. P., Carlstedt-Duke, J., Yamamoto, K., Gustafsson, J.-A. & Kaptein, R. (1990) Science 249, 157-160. 9. Pan, T., Freedman, L. P. & Coleman, J. E. (1990) Biochemistry 29, 9218-9225. 10. Pan, T. & Coleman, J. E. (1989) Proc. Natl. Acad. Sci. USA 86, 3145-3149. 11. Smith, R. F. & Smith, T. F. (1990) Proc. Natl. Acad. Sci. USA 87, 118-122. 12. Evans, R. M. (1988) Science 240, 889-895. 13. Beato, M. (1989) Cell 56, 335-344. 14. Chang, C., Kokontis, J., Chang, C.-T. & Liao, S. (1987) Nucleic Acids Res. 15, 9603. 15. Danielsen, M., Northrop, J. P. & Ringold, G. M. (1986) EMBO J. 5, 2513-2522. 16. Hollenberg, S. M., Weinberger, C., Ong, E. S., Cerelli, G., Oro, A., Lebo, R., Thompson, R. B., Rosenfeld, M. G. & Evans, R. M. (1985) Nature (London) 318, 670-672.
Biochemistry: Vallee et al. 17. Freedman, L. P., Luisi, B. F., Korszun, Z. R., Basavappa, R., Sigler, P. B. & Yamamoto, K. R. (1988) Nature (London) 334, 543-546. 18. Arriza, J. L., Weinberger, C., Cerelli, G., Glaser, T. M., Handelin, B. L., Housman, D. E. & Evans, R. M. (1987) Science 237, 268-275. 19. Patel, P. D., Sherman, T. G., Goldman, D. J. & Watson, S. J. (1988) Mol. Endocrinol. 3, 1877-1885. 20. Misrahi, M., Atger, M., D'Auriol, L., Loosfelt, H., Meriel, C., Fridlansky, F., Guiochon-Mantel, A., Galibert, F. & Milgrom, E. (1987) Biochem. Biophys. Res. Commun. 143, 740-748. 21. Tilley, W. D., Marcelli, M., Wilson, J. D. & McPhaul, M. J.
(1989) Proc. Natl. Acad. Sci. USA 86, 327-331.
22. Yarbrough, W. G., Quarmby, V. E., Simental, J. A., Joseph, D. R., Sar, M., Lubahn, D. B., Olsen, K. L., French, F. S. & Wilson, E. M. (1990) J. Biol. Chem. 265, 8893-8900. 23. Gronemeyer, H., Turcotte, B., Quirin-Stricker, C., Bocquel, M. T., Meyer, M. E., Krozowski, Z., Jeltsch, J. M., Lerouge, T., Gamier, J. M. & Chambon, P. (1987) EMBO J. 6, 39853994. 24. Loosfelt, H., Atger, M., Misrahi, M., Guiochon-Mantel, A., Meriel, C., Logeat, F., Benarous, R. & Milgrom, E. (1986) Proc. Natl. Acad. Sci. USA 83, 9045-9049. 25. Laughon, A. & Gesteland, R. F. (1984) Mol. Cell. Biol. 4, 260-267. 26. Johnston, M. (1987) Nature (London) 328, 353-355. 27. Halvorsen, Y.-D. C., Nandabalan, K. & Dickson, R. C. (1990) J. Biol. Chem. 265, 13283-13289. 28. Schultze, P., Worgotter, E., Braun, W., Wagner, G., Vasdk, M., Kagi, J. H. R. & Wuthrich, K. (1988) J. Mol. Biol. 203, 251-268. 29. Miller, J., McLachlan, A. D. & Klug, A. (1985) EMBO J. 4, 1609-1614. 30. Salmeron, J. M., Jr., & Johnston, S. A. (1986) Nucleic Acids Res. 14, 7767-7781. 31. Kammerer, B., Guyonvarch, A. & Hubert, J. C. (1984) J. Mol. Biol. 180, 239-250. 32. Messenguy, F., Dubois, E. & Descamps, F. (1986) Eur. J. Biochem. 157, 77-81. 33. Beri, R. K., Whittington, H., Roberts, C. F. & Hawkins, A. R. (1987) Nucleic Acids Res. 15, 7991-8001.
Proc. Natl. Acad. Sci. USA 88 (1991)
1003
34. Andre, B. (1990) Mol. Gen. Genet. 220, 269-276. 35. Geever, R. E., Huiet, L., Baum, J. A., Tyler, B. M., Patel, V. B., Rutledge, B. J., Case, M. E. & Giles, N. H. (1989) J. Mol. Biol. 207, 15-34. 36. Vallee, B. L. (1977) Experientia 33, 600. 37. Vallee, B. L. (1977) in Biological Aspects of Inorganic Chemistry, ed. Dolphin, D. (Wiley, New York), pp. 37-70. 38. Auld, D. S. (1979) Adv. Chem. Ser. 172, 112-133. 39. Vallee, B. L. & Falchuk, K. F. (1981) Philos. Trans. R. Soc. London Ser. B 294, 185-197. 40. Hanas, J. S., Hazuda, D. J., Bogenhagen, D. F., Wu, F. Y.-H. & Wu, C.-W. (1983) J. Biol. Chem. 258, 14120-14125. 41. Shang, Z., Liao, Y.-D., Wu, F. Y.-H. & Wu, C.-W. (1989) Biochemistry 28, 9790-9795. 42. Han, M. K., Cyran, F. P., Fischer, M. T., Kim, S. H. & Ginsburg, A. (1990) J. Biol. Chem. 265, 13792-13799. 43. Diakun, G. P., Fairall, L. & Klug, A. (1986) Nature (London) 324, 698-699. 44. Lee, M. S., Gippert, G. P., Soman, K. V., Case, D. A. & Wright, P. E. (1989) Science 245, 635-637. 45. Carr, M. D., Pastore, A., Gausepohl, H., Frank, R. & Roesch, P. (1990) Eur. J. Biochem. 188, 455-461. 46. Neuhaus, D., Nakaseko, Y., Nagai, K. & Klug, A. (1990) FEBS Lett. 262, 179-184. 47. Pdrraga, G., Horvath, S., Hood, L., Young, E. T. & Klevit, R. E. (1990) Proc. Natl. Acad. Sci. USA 87, 137-141. 48. Frankel, A. D., Berg, J. M. & Pabo, C. 0. (1987) Proc. Natl. Acad. Sci. USA 84, 4841-4845. 49. Gibson, T. J., Postma, J. P. M., Brown, R. S. & Argos, P. (1988) Protein Eng. 2, 209-218. 50. Frankel, A. D., Bredt, D. S. & Pabo, C. 0. (1988) Science 240, 70-73. 51. South, T. L., Blake, P. R., Sowder, R. C., III, Arthur, L. O., Henderson, L. E. & Summers, M. F. (1990) Biochemistry 29, 7786-7789. 52. Vallee, B. L. & Auld, D. S. (1989) FEBS Lett. 257, 138-140. 53. Vallee, B. L. & Auld, D. S. (1991) in Matrix Metalloproteinases and Inhibitors, eds. Birkedal-Hansen, H., Werb, Z., Welgus, H. & Van Wart, H. (Fischer, Stuttgart, F.R.G.), in press.
