Proc. Nati. Acad. Sci.

USA
Vol. 88, pp. 9578-9582, November 1991
Biochemistry

The two-hybrid system: A method to identify and clone genes for

proteins that interact with a protein of interest
(yeast/GAL4/transcriptional activation/transcriptional siendng/dimerization)
CHENG-TING CHIEN*, PAUL L. BARTELt, ROLF STERNGLANZ*, AND STANLEY FIELDStI
Departments of *Biochemistry and Cell Biology and tMicrobiology, State University of New York, Stony Brook, NY 11794
Communicated by Keith R. Yamamoto, August 2, 1991 (received for review May 14, 1991)

ABSTRACT We describe a method that detects proteins leads to the transcriptional activation of a reporter gene
capable of interacting with a known protein and that results in containing a binding site for GAL4.
the immediate availability of the cloned genes for these inter- In this paper, we demonstrate the utility of this in vivo
acting proteins. Plasmids are constructed to encode two hybrid approach (designated the two-hybrid system) to screen a
proteins. One hybrid consists of the DNA-binding domain of random library for an interacting protein. Our test case is the
the yeast transcriptional activator protein GAL4 fused to the yeast SIR4 gene product, a protein involved in the transcrip-
known protein; the other hybrid consists of the GAL4 activa- tional repression of the silent copies of mating type informa-
tion domain fused to protein sequences encoded by a library of tion (6). We first showed by construction of GAL4 hybrids
yeast genomic DNA fragments. Interaction between the known with SIR4 domains that SIR4 is capable of dimer formation
protein and a protein encoded by one of the library plasmids and that this dimerization can be mediated by a small
leads to transcriptional activation of a reporter gene containing C-terminal fragment of SIR4. We then transformed yeast with
a binding site for GAL4. We used this method with the yeast both a plasmid encoding the GAL4 DNA-binding domain
SIR4 protein, which is involved in the trinscriptional repres- fused to part of SIR4 and a library of plasmids containing
sion of yeast mating type information. (a) We used the two- yeast genomic fragments fused to the GAL4 activation do-
hybrid system to demonstrate that SIR4 can form homodimers. main. By screening for transcription of a GAL4-dependent
(ii) A small domain consisting of the C terminus of SIR4 was reporter gene, we identified a plasmid from the library
shown to be sufficient to mediate this interaction. (iii) We carrying a fragment of the SIR4 gene. This method of
screened a library to detect hybrid proteins that could interact identifying interacting proteins has obvious applications in
with the SIR4 C-terminal domain and identified SIR4 from this other systems.
library. This approach could be readily extended to mamma-
lian proteins by the construction of appropriate cDNA libraries MATERIALS AND METHODS
in the activation domain plasmid.
Yeast Strains and Methods. Yeast strains used were
GGY1::171 (7) and W303 (8). Yeast were grown in YEPD or
Specific interactions between proteins form the basis of many selective minimal medium (9). Transformation was by the
essential biological processes. Additionally, transforming pro- high-efficiency method of Schiestl and Gietz (10). 3-Galac-
teins of tumor viruses in many cases exert their effect through tosidase activity was assayed either on plates or in liquid;
their interactions with cellular proteins; for example, the yeast transformants were replica-plated to SSX plates or
simian virus 40 (SV40) large tumor (T) antigen binds to the liquid cultures were assayed for 4-nitrophenyl f3-D-
cellular proteins p53 and Rb (1, 2). Consequently, considerable galactoside cleavage as described (4). SSX plates contained
effort has been made to identify those proteins that bind to 6.7 g of yeast nitrogen base, 14 g of agar, 40 mg of 5-bromo-
proteins of interest. Typically, these interactions have been 4-chloro-3-indolyl f-D-galactoside, 100 ml of 1 M potassium
detected by using coimmunoprecipitation experiments in phosphate (pH 7.0), 100 ml of 20% (wt/vol) sucrose, 100 ml
which antibody to a known protein is used to precipitate of 1Ox amino acids minus leucine and histidine (9), and H20
associated proteins as well. Such biochemical methods, how- to 1 liter. Escherichia coli MH4 (11) contained the leuB
ever, result only in the identification of the apparent molecular mutation, which can be complemented by the yeast LEU2
mass of the associated proteins; obtaining cloned genes for gene.
these proteins is often a difficult process. In one approach, this Construction of GAL4 Activation Domain Plasmids. Plas-
problem has been circumvented by the use of purified proteins mid pGX13 was derived from 401-N (12) and pKLC15 (13)
as probes against bacterial expression libraries, where a pos- and contains as a HindIII fragment sequences encoding the
itive signal for an interacting protein is accompanied by the first five codons from SV40 T antigen, the nuclear localiza-
availability of the corresponding gene (3). tion signal from T antigen, amino acids 11-34 from T7 RNA
We have described a method by which a protein-protein polymerase, and amino acids 768-881, containing an activa-
interaction is identified in vivo through reconstitution of the tion domain, from GAL4. Site-directed mutagenesis was
activity of a transcriptional activator (4). The method is based used to destroy the termination codon of GALA and create a
on the properties of the yeast GAL4 protein, which consists Bgl II site at this location. The HindIII fragment was inserted
of separable domains responsible for DNA-binding and tran- into the unique HindIII site of pAE7, which is pAAH5 (14)
scriptional activation (5). Plasmids encoding two hybrid that has had its two BamHI sites removed by sequentially
proteins, one consisting of the GAL4 DNA-binding domain linearizing with BamHI and filling-in the overhang with DNA
fused to protein X and the other consisting of the GAL4 polymerase I (Klenow fragment). Finally, oligonucleotides
activation domain fused to protein Y, are constructed and containing the BamHI sites in all three frames were inserted
introduced into yeast. Interaction between proteins X and Y into the Bgl II site to generate the three pGAD vectors. Our
initial activation domain plasmids (pGAD1, pGAD2, and
The publication costs of this article were defrayed in part by page charge
payment. This article must therefore be hereby marked "advertisement" Abbreviations: SV40, simian virus 40; T, tumor.
in accordance with 18 U.S.C. §1734 solely to indicate this fact. fTo whom reprint requests should be addressed.
9578
 Biochemistry: Chien et al. Proc. Natl. Acad. Sci. USA 88 (1991) 9579
pGAD3) have the HindIII fragment from pGX13 inserted into plates to SD-His and SD-Leu identified colonies that had lost
pAE7 in the orientation such that transcription of the fusion one or both plasmids, and replica plating to SSX plates
protein is driven by a promoter lying within sequences for (supplemented with histidine and leucine) indicated whether
termination of ADHI transcription. Plasmids have also been ,B-galactosidase was expressed. Any colonies that remained
constructed in the opposite orientation (pGAD1F, pGAD2F, blue after loss of the DNA-binding domain plasmid were
and pGAD3F) in which transcription of the fusion protein eliminated. In a second approach, the blue colonies were
gene is driven by the ADHI promoter. Whereas the ADHI grown in SD-Leu liquid medium and plasmid DNA was
promoter results in additional expression of the fusion pro- isolated from the culture (19) and transformed into E. coli by
tein, it leads to only minor changes in the level of GAL.J-lacZ electroporation. The activation domain (pGAD) plasmids
transcription derived by interaction of the two hybrids. from the library were identified by their ability to comple-
Construction of GAL4 Protein Fusions. Plasmid pES15 ment an E. coli leuB mutation due to the plasmid-borne LEU2
(from Elisa Stone, State University of New York at Stony gene. These plasmids were then individually reintroduced
Brook) containing the entire SIR] sequence in pUC18 was into yeast GGY1::171 either alone or with the DNA-binding
cut at the Nru I site 52 base pairs 5' to the initiator ATG and domain plasmid. Yeast transformants that expressed /3-ga-
an oligonucleotide containing an Xho I site was inserted. The lactosidase when they contained the library plasmid alone
resultant plasmid was cut with Xho I and the SIR] fragment were eliminated. Alternatively, the activation domain plas-
was ligated to Sal I-digested pMA424 (15) to create pKL5. An mid was cut with Xho I and Sal I to assay for a 1.0-kilobase
Xho I fragment containing the entire SR2 sequence was fragment diagnostic of the GALA gene (20). Plasmids encod-
prepared by the polymerase chain reaction from pJM100 (16) ing interacting proteins were sequenced by the dideoxynu-
and was ligated to Sal I-digested pMA424 to produce pCTC7. cleotide method (21) using a primer corresponding to codons
pJR104 (17), which carries the SIR3 gene, was digested with 758-763 of the GALA sequence.
Sal I and the SIR3-containing fragment was ligated into
pUC18. The resulting plasmid (pKL3) was digested first with RESULTS
Bcl I and HindIII and then with Xho II and HindIII and the
two appropriate fragments were ligated into BamHI-digested Construction of Activation Domain Vectors. The principle of
pGAD3 to yield pKL13. The SIR4 gene was subcloned as a using two hybrid proteins to detect protein-protein interac-
HindIII fragment from pJR643 (from Jasper Rine, Univ. of tions is shown in Fig. 1. A gene fusion is first generated that
California, Berkeley) into pUC18 to produce pCTC15. A encodes the protein under study (protein X) as a hybrid with
BamHI fragment encoding amino acids 839-1358 was in- the DNA-binding domain of GAL4. This domain enables the
serted into the BamHI site of pMA424 to generate pCTC17. hybrid to localize (22) to the nucleus and bind in a site-specific
This same BamHI fragment was used to generate the acti- fashion to DNA (5). For the method to work, the protein X
vation domain hybrid pCTC18. The SIR4 fragment corre- domain of this hybrid must fail to activate transcription of the
sponding to amino acids 1262-1358 was generated as a reporter gene. Second, a gene fusion is introduced that
BamHI fragment by the polymerase chain reaction using as encodes a protein Y sequence as a hybrid with the GAL4
template pJR643. This fragment was inserted into the BamHI activation domain; this hybrid also enters the nucleus. Pro-
site of pMA424 to generate pCTC23 and into pGAD2 to tein Y may be encoded by a known gene or represent a library
generate pCTC24. The SNF4 gene was derived from plasmid
pSL321 (18). The Cla I site at codon 20 and the polylinker DNA-binding domain hybrid
Kpn I site after codon 321 (18) were converted to BamHI sites
by insertion of oligonucleotide adaptors, and the resultant
BamHI fragment was inserted into pGAD3 to generate pCTC14.
Construction of Yeast Genomic Libraries. Yeast genomic
DNA was isolated from strain W303 by spheroplasting the
cells with Zymolyase and adding 2 vol of lysis buffer [3%
(wt/vol) Sarkosyl/0.5 M Tris Cl/0.2 M EDTA, pH 9] to
liberate the cell contents. Cell debris was pelleted after
addition of 0.5 vol of 5 M potassium acetate, followed by Activation domain hybrids encoded by a library
ethanol precipitation of the DNA. The DNA was partially
digested with Sau3A and was fractionated on 5-30% (wt/vol) I ) GAL4 (768-881)
sucrose gradients to isolate fragments from 2 to 6 kilobases
in size. This DNA was ligated to pGAD1, pGAD2, or pGAD3
DNA that had been digested with BamHI and dephosphory-
lated with calf intestinal alkaline phosphatase. Ligation mixes GAL I-IacZ
were used to transform E. coli DH5 by electroporation. After UASG
growth on ampicillin-containing plates overnight, colonies
were pooled and frozen at -70°C. These frozen stocks were
used to grow the ampicillin-resistant transformants for prep- Interaction between DNA-binding domain
aration of plasmid DNA. hybrid and a hybrid from the library
Screening Activation Domain Libraries. Yeast strain
GGY1::171, which is His- and Leu-, was transformed si- < I Y ) GAL4 (768-881)
multaneously with both a DNA-binding domain plasmid and
one of the libraries of genomic DNA fragments. After 3-4
days growth on SD-His-Leu plates (9), His', Leu+ trans- \\J AL4 (1-147)
formants were replica-plated to SSX plates and were incu- UASG GAL 1-lacZ
bated until blue colonies appeared. False positives due to
cloning of the GAL4 gene into the pGAD vectors were FIG. 1. Strategy to detect interacting proteins using the two-
eliminated by one of several strategies. In one approach, blue hybrid system. UASG is the upstream activation sequence for the
colonies were grown to saturation nonselectively in YEPD yeast GAL genes, which binds the GAL4 protein. The libraries of
liquid and replated on YEPD plates. Replica plating of these activation domain hybrids are constructed in the pGAD vectors.
9580 Biochemistry: Chien et al. Proc. Natl. Acad. Sci. USA 88 (1991)

of sequences (Y1, Y2, Y3, . . ., Yn). Any activation domain Table 1. Reconstitution of GALA activity by SNF1 and
hybrid carrying a protein Y sequence that can bind to protein SNF4 hybrids
X might be capable of reconstituting proximity of the two Transformant GAL1-
GAL4 domains in a manner that activates transcription of a LacZ
reporter gene.
DNA-binding-domain Activation-domain
In the previous report from this laboratory (4) using the hybrid hybrid activity
two-hybrid system to detect the interaction of the yeast SNF1 1. GAL4-(1-147)-SNF1 1
and SNF4 proteins, the activation domain hybrid consisted of 2. SNF4-(1-322)-GAL4-(768-881) 1
the yeast SNF4 protein at its N terminus and the GAL4 3. GAL4-(768-881)-SNF4-(22-320) 1
activation domain at its C terminus. Although the C-terminal 4. GAL4-(1-147)-SNF1 SNF4-(1-322)-GAL4-(768-881) 171
location for the GAL4 activation domain corresponds to its 5. GAL4-(1-147)-SNF1 GAL4-(768-881)-SNF4-(22-320) 144
normal position within the GAL4 protein, this location is Values for ,B-galactosidase activity (28) are the mean of assays on
inconvenient for the construction of a library of activation at least two transformants, each assayed at least twice. The standard
domain hybrids. We therefore determined whether the GAL4 errors were typically 30%o of the mean for values >2 units.
activation domain (amino acids 768-881) could function at
the N terminus of a protein fusion. This placement allows the DNA-binding domain plasmid, the SNF4 hybrid in plasmid
use of the same vector-encoded yeast promoter and initiation pGAD3 led to a substantial level of GAL1-LacZ activity.
codon for each gene fusion and requires a single fusion joint This activity was comparable to that observed with the
between the activation domain and a heterologous protein. In original SNF4 hybrid, which carried the GAL4 activation
addition, we included within the activation domain fusion a domain at its C terminus (4). This reconstruction experiment
sequence for nuclear targeting. We constructed a set of three
indicated that the pGAD vectors were suitable for testing
plasmids, pGAD1, pGAD2 and pGAD3, as shown in Fig. 2, other defined protein combinations and as recipients in the
that differ only in the reading frame of the unique BamHI site. construction of libraries.
These vectors include an initiation codon, the nuclear local- SIR Protein Interactions. The mating type of Saccharomy-
ization signal from SV40 T antigen (23), which has been ces cerevisiae is determined by the genetic information
shown to function in yeast (12, 24), and the GAL4 activation present at the MAT locus. Two other copies of mating type
domain (residues 768-881) followed by the BamHI cloning information, HML and HMR, are kept transcriptionally silent
site. and serve as donors for transposition of sequences to MAT.
We used a pGAD vector to construct a hybrid between the This transcriptional repression is due to the specific action of
GAL4 activation domain and the yeast SNF4 protein to test four SIR gene products (SIR1 through SIR4) and the involve-
for the interaction between the SNF1 and SNF4 proteins. As ment of the proteins ABF1, RAP1, and histone H4, which
shown in Table 1, in cells carrying the SNF1 protein in the have additional roles in the yeast cell (for review, see ref. 25).
The mechanism of silencing and the biochemical interactions
among these proteins are unknown.
To determine whether SIR proteins can bind to each other
as indicated by a transcriptional signal in the two-hybrid
system, we constructed a number of hybrid genes encoding
SIR proteins fused to one of the GAL4 domains and intro-
duced these genes singly or in pairwise combinations into a
yeast reporter strain (Table 2). When present as fusions with
the GAL4 DNA-binding domain, SIR1 (line 1), SIR2 (line 4),
and SIR4 (lines 7 and 8) failed to activate transcription of
GALJ-acZ, indicating that these regions of the SIR proteins
do not function as activation domains. We did not detect a
signal for interaction of SIR1 with SIR3 or SIR4 (lines 2 and
3), SIR2 with SIR3 or SIR4 (lines 5 and 6), or SIR4 with SIR3
(line 9). Only with SIR4 present in both the DNA-binding
domain and activation domain plasmids (line 10) was GALI-
Pstl
lacZ transcription significantly above background. We note
that we constructed only five of the eight possible GAL4
Pvu
fusions with SIR proteins and assayed only 6 of 16 possible
combinations. Lack of a signal is not strong evidence against
a specific interaction; the hybrid proteins might not be stable,
they might not include the residues necessary for interaction
or the GAL4 domains might occlude a site of interaction. For
SIR1 and SIR2, however, the hybrids with the GAL4 DNA-
binding domain are capable of complementing sirl and sir2
mutations, respectively, indicating that the protein fusions
Vector Reading frame of BamHI site are sufficiently stable and active to cooperate with the other
pGAD1 GGA TCC proteins required for transcriptional silencing. For SIR4, the
pGAD2 XGG ATC CXX
hybrids contain no more than 520 residues of this 1358-
residue protein and do not complement a sir4 mutation.
pGAD3 XXG GAT CCX However, the SIR4 hybrids exhibit an activity designated
FIG. 2. Restriction map of the activation domain plasmid. The
anti-SIR (26), in which overexpression of the C-terminal end
GAL4 activation hybrid is transcribed from a promoter labeled P and of the SIR4 protein causes a dominant derepression pheno-
contains in addition a nuclear localization signal from SV40 T type.
antigen, which is not indicated. The three vectors differ in the reading The SIR4 constructions that resulted in a transcriptional
frame of the BamHI cloning site, as indicated. amp, Ampicillin- signal from the reporter gene both contained the C-terminal
resistance gene; ori, origin of replication from pBR322. 520 amino acids of SIR4. It has been noted that the most
 Biochemistry: Chien et al. Proc. Natl. Acad. Sci. USA 88 (1991) 9581

Table 2. Reconstitution of GAL4 activity by SIR hybrids

Transformant GAL1-LacZ
DNA-binding-domain hybrid Activation-domain hybrid activity
1. GAL4-(1-147)-SIR1-(1-678) 1
2. GAL4-(1-147)-SIR1-(1-678) GAL4-(768-881)-SIR3-(17-978) 1
3. GAL4-(1-147)-SIR1-(1-678) GAL4-(768-881)-SIR4-(839-1358) 2
4. GAL4-(1-147)-SIR2-(1-562) 1
5. GAL4-(1-147)-SIR2-(1-562) GAL4-(768-881)-SIR3-(17-978) 2
6. GAL4-(1-147)-SIR2-(1-562) GAL4-(768-881)-SIR4-(839-1358) 2
7. GAL4-(1-147)-SIR4-(839-1358) 1
8. GAL4-(1-147)-SIR4-(1262-1358) 1
9. GAL4-(1-147)-SIR4-(839-1358) GAL4-(768-881)-SIR3-(17-978) 1
10. GAL4-(1-147)-SIR4(839-1358) GAL4-(768-881)-SIR4-(839-1358) 141
11. GAL4-(1-147)-SIR4-(1262-1358) GAL4-(768-881)-SIR4-(1262-1358) 201
12. GAL4-(1-147)-SIR4-(839-1358) GAL4-(768-881)-SIR4-(1205-1358) 167
13. GAL4-(768-881)-SFI1 1
14. GAL4-(1-147)-SIR4-(839-1358) GAL4-(768-881)-SFI1 96
/3-Galactosidase activity was determined as in Table 1.
C-terminal region of SIR4 contains 12 heptad repeats and a Fifteen positive transformants (of a total of 220,000 trans-
similarity to human nuclear lamins (27). We therefore tested formants) were categorized as to whether GALl-lacZ tran-
whether fragments of SIR4 consisting of essentially only scription required the presence of both hybrids. In one
these repeats (the C-terminal 97 amino acids of SIR4) were protocol, cells were grown nonselectively and screened for a
capable of a transcriptional signal indicating interaction (Ta- return to histidine or leucine auxotrophy, indicating loss of
ble 2, line il). The similar level of GALI-lacZ transcription the marker on the DNA-binding or activation domain plas-
we detect with these small hybrids suggests that one SIR4 mid, respectively. A comparison of the histidine and leucine
protein contacts another through a coiled-coiled interaction requirements with ,B-galactosidase expression indicated
mediated by the heptad repeats. whether both plasmids were required for GAL4 function. In
Screening of an Activation Domain Library. Based on the a second approach, each of the library plasmids was isolated
demonstration of the SIR4-SIR4 interaction in the two- and reintroduced into yeast with or without the DNA-binding
hybrid system, we sought to determine whether this or other domain plasmid, and these transformants were tested for
interactions could be detected by screening a library of total ,f-galactosidase activity. Two of the 15 positives required
sequences present in the activation domain plasmid. With both plasmids to reconstitute GAL4 function. DNA sequence
SIR4 as a fusion with the GAL4 DNA-binding domain, any analysis of the insert from one of these activation domain
protein encoded by an activation domain fusion that can hybrids indicated that it encoded a portion of the SIR4 gene,
interact with SIR4 might reconstitute GAL4 activity. We note beginning at a Sau3A site corresponding to amino acid
that the library must be constructed in the activation domain residue 1205, and restriction digestion analysis indicated that
plasmid to avoid detecting random (and abundant) sequences the insert contained the rest of the SIR4 gene. This hybrid
that can activate transcription when fused to a DNA-binding contains 154 amino acids of the SIR4 protein, slightly more
domain (15). than the small C-terminal SIR4 fragment that we had tested.
Each pGAD vector was ligated to a size-fractionated /8-Galactosidase activity of this hybrid from the library (Table
partial Sau3A digest of yeast genomic DNA, generating 2 x
2, line 12) showed a level comparable to that observed in the
106 individual transformants in E. coli. Colonies containing
reconstruction experiments. The other plasmid that required
library constructions in each pGAD vector were pooled
the fusion of the GAL4 DNA-binding domain and. SIR4 for
separately, and plasmid DNA was prepared. Yeast were
GAL4 function carries the gene designated SF1 (for SIR4-
cotransformed with a mixture of DNAs containing equal interacting protein), whose partial sequence does not corre-
amounts of GAL4-(1-147)-SIR4-(839-1358) and one of the
spond to any yeast gene sequence in available data bases
pGAD Sau3A libraries, selecting for both histidine and (February 1991). The SFI1-containing plasmid alone was
leucine prototrophy. This protocol was preferred over se-
inactive for GAL1-lacZ transcription (Table 2, line 13) but
with GAL4-(1-147)-S1R4-(839-1358) produced 96 units of
quential transformation of first the DNA-binding domain activity (Table 2, line 14).
plasmid and then the library because we have observed some The yeast genome contains =50,000 Sau3A fragments
instability of certain plasmids encoding DNA-binding domain which can be ligated in either orientation to yield 100,000
hybrids. The cotransformation protocol minimizes the possible fusion joints in each of the pGAD libraries. If a gene
amount of time transformants are incubated before they are encoding an interacting protein contains a single appropriate
assayed for 8-galactosidase activity. Transformants were Sau3A site, the probability of detecting the correct GAL4
replica-plated to medium containing 2% sucrose, amino acid hybrid in a given library is 90% by screening 230,000 trans-
supplements lacking histidine and leucine, and 5-bromo-4- formants and 99% by screening 460,000 transformants. Thus
chloro-3-indolyl ,8-D-galactoside (40 ,ug/ml). Approximately it is possible that there are additional proteins capable of
1 transformant per 14,000 turned blue by 5 days, indicating interacting with SIR4 that would be found with additional
transcription of GAL1-lacZ. In addition, we note that the library screening.
activation domain library transformed alone caused a back-
ground of approximately the same ratio of blue colonies,
which results from cloning of the GAL4 gene in the pGAD DISCUSSION
vectors. (The high background is due to the fact that the The approach of using two GAL4 hybrids to detect protein-
GAL4 gene can be present anywhere within the insert se- protein interactions has been extended to three additional
quence and need not be in a defined position, orientation, and applications. (i) It can be used with available genes to test
reading frame as does the gene for an interacting protein.) pairwise combinations for interaction. Such testing of SIR
9582 Biochemistry: Chien et al. Proc. Natl. Acad. Sci. USA 88 (1991)

proteins provides strong evidence for a SIR4-SIR4 complex. this method in a strain carrying a GAL4-dependent selectable
(ii) A positive signal for interaction allows a rapid means to gene. Such a strain would allow the assaying of cDNA
identify the specific domains responsible for the protein- libraries of high complexity while requiring the use of rela-
protein contacts. For SIR4, hybrids carrying only 7% of the tively few plates. This approach might thus lead to the
protein were used to demonstrate that the contacts appear to identification of various interacting proteins of mammalian
be mediated by a series of heptad repeats present at the C origin.
terminus. (iii) A library of total sequences fused to the
activation domain can be screened to detect a plasmid We thank Mark Swanson for some of the plasmids used in these
encoding an interacting protein. We used such a yeast library experiments and Joe Lipsick for comments on the manuscript. This
to identify a SIR4 insert and another gene. work was supported by U.S. Public Health Service Research Grants
GM28220 to R.S. and CA54699 to S.F. and by grants from the Procter
By testing a series of pairwise combinations of SIR pro- and Gamble Company and New York State Science and Technology
teins in the two-hybrid system, we detected a signal for Foundation to S.F.
interaction only between SIR4 and itself. This result suggests
that SIR4 may function in silencing as a dimer or higher-order 1. Lane, D. P. & Crawford, L. V. (1979) Nature (London) 278,
multimer. It also provides a possible explanation for the 261-263.
anti-SIR activity of SIR4 C-terminal fragments (26), observed 2. DeCaprio, J. A., Ludlow, J. W., Figge, J., Shew, J.-Y., Huang,
as derepression of the silent mating type loci in a SIR' strain: C.-M., Lee, W.-H., Marsilio, E., Paucha, E. & Livingston,
the overproduced SIR4 fragments may bind to the wild-type D. M. (1988) Cell 54, 275-283.
3. Nelbock, P., Dillon, P. J., Perkins, A. & Rosen, C. A. (1990)
SIR4 protein and prevent it from exerting its activity. Science 248, 1650-1653.
In using the two-hybrid approach for screening a library of 4. Fields, S. & Song, 0. (1989) Nature (London) 340, 245-246.
activation-domain hybrids, a major advantage is the imme- 5. Keegan, L., Gill, G. & Ptashne, M. (1986) Science 231, 699-
diate availability of the cloned gene for the interacting 704.
protein. In addition, only a single plasmid construction is 6. Rine, J. & Herskowitz, I. (1987) Genetics 116, 9-22.
required to use this method; there is no necessity to prepare 7. Gill, G. & Ptashne, M. (1987) Cell 51, 121-126.
either antibody or purified protein for the biochemical de- 8. Thomas, B. J. & Rothstein, R. (1989) Cell 56, 619-630.
tection of interactions. The interactions are detected in vivo, 9. Sherman, F., Fink, G. R. & Hicks, J. B. (1986) Methods in
Yeast Genetics (Cold Spring Harbor Lab., Cold Spring Harbor,
under conditions that may be similar to those that occur NY).
naturally. Because the background of GALJ-lacZ transcrip- 10. Schiestl, R. H. & Gietz, R. D. (1989) Curr. Genet. 16, 339-346.
tion is negligible, even interactions that reconstitute only a 11. Hall, M. N., Hereford, L. & Herskowitz, I. (1984) Cell 36,
low level of GAL4 function produce a detectable signal. Thus 1057-1065.
it may be possible to detect transient interactions, such as 12. Benton, B. M., Eng, W.-K., Dunn, J. J., Studier, F. W.,
those occurring during only a limited portion ofthe cell cycle. Sternglanz, R. & Fisher, P. A. (1990) Mol. Cell. Biol. 10,
Finally, the stability of the f3-galactosidase protein allows the 353-360.
accumulation of a weak signal over time. 13. Ma, J. & Ptashne, M. (1987) Cell 48, 847-853.
14. Ammerer, G. (1983) Methods Enzymol. 101, 192-201.
Comparison of the signal generated by the combination of 15. Ma, J. & Ptashne, M. (1987) Cell 51, 113-119.
GAL4-(1-147)-SIR4-(839-1358) or GAL4-(1-147)-SIR4- 16. Mullen, J. R., Kayne, P. S., Moerschell, R. P., Tsunasawa, S.,
(1262-1358) with three other SIR4 hybrids with the activation Gribskov, M., Colavito-Shepanski, M., Grunstein, M., Sher-
domain indicates that all six combinations are similarly man, F. & Sternglanz, R. (1989) EMBO J. 8, 2067-2075.
effective at reconstituting GAL4 function (Table 2 and un- 17. Kimmerly, W. J. & Rine, J. (1987) Mol. Cell. Biol. 7, 4225-
published results). This result further supports the hypothesis 4237.
that the structure of the two interacting hybrids can be highly 18. Celenza, J. L., Eng, F. J. & Carlson, M. (1989) Mol. Cell. Biol.
variable, with the major requirement being only the presence 9, 5045-5054.
of the interacting domains. Such flexibility suggests that 19. Hoffman, C. S. & Winston, F. (1987) Gene 57, 267-272.
20. Laughon, A. & Gesteland, R. F. (1984) Mol. Cell. Biol. 4,
numerous other proteins may be detected in this system. 260-267.
Reconstruction experiments indicate that such interacting 21. Sanger, F., Nicklen, S. & Coulson, A. R. (1977) Proc. Natl.
protein pairs as p53-SV40 large T antigen (unpublished Acad. Sci. USA 74, 5463-5467.
results), retinoblastoma protein-SV40 large T antigen (T. 22. Silver, P. A., Keegan, L. P. & Ptashne, M. (1984) Proc. Natl.
Durfee and W.-H. Lee, personal communication), and the .8 Acad. Sci. USA 81, 5951-5955.
and y subunits of the yeast pheromone-responsive guanine 23. Kalderon, D., Roberts, B. L., Richardson, W. D. & Smith,
nucleotide binding protein (K. Clark and M. Whiteway, A. E. (1984) Cell 39, 499-509.
personal communication) give positive signals in this system. 24. Nelson, M. & Silver, P. (1989) Mol. Cell. Biol. 9, 384-389.
In addition, other screens for interacting proteins using the 25. Alberts, B. M. & Sternglanz, R. (1990) Nature (London) 344,
193-194.
libraries described here have detected candidate yeast pro- 26. Marshall, M., Mahoney, D., Rose, A., Hicks, J. B. & Broach,
teins capable of binding to p53 (unpublished results) and J. R. (1987) Mol. Cell. Biol. 7, 4441-4452.
RAP1 (C. Hardy and D. Shore, personal communication). 27. Diffley, J. F. X. & Stillman, B. (1989) Nature (London) 342, 24.
Although interaction in these experiments has been detected 28. Miller, J. H. (1972) Experiments in Molecular Genetics (Cold
by screening for blue colony color, it should be feasible to use Spring Harbor Lab., Cold Spring Harbor, NY).