HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

J. Biol. Chem., Vol. 281, Issue 29, 20107-20119, July 21, 2006

This Article Abstract Full Text (PDF) Supplemental Data All Versions of this Article: 281/29/20107    most recent M512588200v1 Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Request Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by King, D. A. Articles by Ellenberger, T. Search for Related Content PubMed PubMed Citation Articles by King, D. A. Articles by Ellenberger, T. Social Bookmarking                      What's this?

Domain Structure and Protein Interactions of the Silent Information Regulator Sir3 Revealed by Screening a Nested Deletion Library of Protein Fragments^*

From the Department of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, Boston, Massachusetts 02115

Received for publication, November 28, 2005 , and in revised form, May 19, 2006.

	ABSTRACT

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

 
Transcriptional silencing in yeast is mediated by the interactions of silent information regulator (Sir) proteins with chromatin and with one another. The stable association of Sir3 with Sir4 is mediated by a C-terminal region of Sir3 that has additional functions including the dimerization of Sir3. We have developed a simple, robust expression screening methodology that allows for the unbiased identification of functional protein domains expressed from nested-deletion libraries of full-length genes. Using these methodologies, Sir3 dimerization was shown to be mediated by two separate domains. One of these domains also binds cooperatively to the C-terminal coiled-coil motif of Sir4 and dimerization further increases the affinity of Sir3 for Sir4. The resulting Sir3-Sir4 complexes form progressively higher order assemblies with increasing protein concentration, with implications for the mechanism of gene silencing.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

 
The Saccharomyces cerevisiae genome is divided into actively transcribed regions known as euchromatin and transcriptionally silent regions resembling heterochromatin of higher eukaryotes. In yeast, the silent information regulator (Sir)² proteins are the principle factors establishing silent DNA, and serve as a model system for understanding transcriptionally silent heterochromatin in higher eukaryotes (1). Some of the molecular aspects of DNA silencing have been elucidated at the mating type loci and telomeres of the S. cerevisiae genome (2). Gene silencing requires Sir2, a broadly conserved NAD-dependent histone deacetylase (3–5). Yeast Sir3 and Sir4 proteins have no obvious homologues in higher organisms, but they are essential for gene silencing in S. cerevisiae (6, 7). The Sir2, Sir3, and Sir4 proteins physically interact with one another, and with transcription factors and histones to establish silent DNA (8, 9). Interactions with transcription factors bound to silencer elements initially recruit the Sir2, Sir3, and Sir4 proteins to the silencer (10–15). Following this nucleation step, the histone deacetylase activity of Sir2 enables high affinity interactions of Sir3 with the deacetylated N-terminal tails of histones H3 and H4, allowing the Sir protein complexes to spread along chromatin by "sequential deacetylation" (1, 2, 5, 10–12, 16, 17, 19–22).

The interaction of Sir3 with Sir4 is an essential feature of the gene silencing mechanism. This protein-protein interaction stabilizes Sir protein complexes on the silencer element, a prerequisite for the spreading of Sir proteins to establish silent DNA. The C-terminal regions of both Sir3 and Sir4 participate in their physical and functional interactions (Fig. 1A) (11, 12, 22–25). A C-terminal coiled-coil domain of Sir4 (residues 1198–1358) interacts with Sir3 (11). We and others have reported crystal structures of the Sir4 coiled-coil (Sir4-CC; residues 1271–1346) (26, 27). Guided by the structure, we showed that a small patch of surface residues located in the middle of the Sir4-CC mediates the interaction with Sir3 (27).

The C-terminal region of Sir3 is required for interactions with Sir4 in vitro (summarized in Fig. 1A) and for silencing activity in vivo (11, 12). N-terminal deletions of Sir3 beyond amino acid 522 prevent interaction with immobilized Sir4-CC, whereas slightly longer Sir3 constructs are pulled down efficiently (27). However, the Sir3 C-terminal region also participates in interactions with histones H3 and H4 and mediates Sir3 dimerization (12, 17, 28). We sought to identify the residues that are responsible for these different functions of Sir3, to assess the role of each interaction in DNA silencing.

The histone deacetylase Sir2 forms specific complexes with both Sir3 and Sir4. These physical interactions serve to co-localize the Sir proteins to regions of silent DNA and to regulate histone modifications affecting chromatin structure and accessibility. Crystal structures of several archaeal Sir2 homologs and the human SIRT2 protein have been reported, alone and in complexes with the NAD⁺ cofactor or an acetyl-lysine substrate peptide (29–32). The structures show a conserved Rossman-type nucleotide-binding fold and are suggestive of a mechanism for enzymatic deacetylation.

In contrast, there has been little progress on the structural biology of higher order Sir protein complexes. The stoichiometry and molecular interactions of Sir3 and Sir4 within silencing complexes are unknown. There is evidence that the subunit makeup of Sir3-Sir4 protein complexes can be altered by the products of Sir2 deacetylation, O-acetyl-ADP-ribose and deacetylated histones (33). Consistent with this variability, the overexpression of Sir3 in vivo results in silent DNA with abundant Sir3 but reduced amounts of Sir2 and Sir4 (34). Current models for the assembly of Sir proteins on chromatin do not account for this apparent plasticity in subunit composition.

Biochemical and structural studies of Sir3 (978 residues) and Sir4 (1358 residues) have been hampered by the large size and limited solubility of these proteins as well as the lack of homologous proteins that might otherwise reveal the important conserved features. As such, it has been difficult to divide the Sir proteins into functional domains that can be studied in isolation. After failing to obtain well diffracting crystals of Sir3 alone or in complex with the Sir4-CC, we developed an unbiased methodology to identify Sir3 fragments that stably interact with Sir4, as a prelude to thermodynamic and structural studies of these protein-protein interactions. Here we describe the application of these methods to Sir3, and the identification of two protein interaction domains mediating Sir3 dimerization and its interaction with the Sir4-CC.

	EXPERIMENTAL PROCEDURES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

 
Identification of Functional Protein Domains from a Nested Deletion Expression Library: Methodology and Biases—We wished to create a large and unbiased library expressing random Sir3 protein fragments. We examined several methods to create nested gene deletions and compared a series of reporter vectors to select in-frame fragments from the library. A variety of endonucleases and exonucleases have been used to generate nested deletion libraries (35). However, the extent of these reactions can be difficult to control, leading to incomplete representation of coding sequences in a library. To more completely sample a large number of overlapping gene fragments with randomly distributed 5' and 3' termini, we used mung bean nuclease to create nested gene deletions (Fig. 1B). Mung bean nuclease is commonly used to remove protruding 3' termini of double-stranded DNA. Although it is highly specific for single-stranded DNA (selectivity of 30,000:1 for single-stranded DNA: double-stranded DNA), this enzyme endonucleolytically cleaves double-stranded DNA at high enzyme concentrations and extended incubation periods (36). By exploiting the slow activity of mung bean nuclease on double-stranded DNA, nested gene deletions can be generated that finely sample a large number of 5' and 3' end points (Fig. 1).

pT7CF Expression Vector and Screening Modifications—From a large array of random DNA fragments, we sought a method for selecting coding sequences in the correct reading frame that could be expressed in Escherichia coli and subsequently screened for various functions. We devised an expression vector for random gene fragments fused to E. coli type I chloramphenicol acetyltransferase (CAT), which serves as a convenient reporter group for expression screening of nested deletion libraries (Fig. 1B) (37). In our pT7CF expression vector, random coding sequences appended with an appropriate adaptor (see below) are cloned into an engineered MluI restriction site flanked on the 5' side by a His₆ affinity tag sequence and on the 3' side by the coding sequence for full-length CAT (Fig. 1B). The CAT coding sequence (residues 1–219) from pPROEX HT-CAT (Invitrogen) was amplified by PCR to contain an MluI restriction site (ACGCGT) immediately 5' of the start codon and a stop codon inserted after residue 219. The amplified CAT gene was cloned into the pCRT7/NT-TOPO vector (Invitrogen) to create the pT7CF (T7-promoter CAT Fusion) vector that appends an N-terminal His₆ affinity tag to the expressed proteins. Following affinity purification, the His₆ tag can be removed by cleavage with enterokinase. pT7CF was digested with MluI and gel purified before ligating a phosphorylated 9-mer linker (5'-PO₄-CGCGGTTTG-3') to the ends of the cut vector. The linker was present in a 1,000-fold molar excess over pT7CF in the ligation reaction with T4 DNA ligase done overnight at 16 °C. After addition of linkers, the pT7CF vector was purified by gel electrophoresis in 1% agarose to remove unligated oligos and any recircularized vector.

Preparation of a SIR3-nested Deletion Library and Selection for In-frame SIR3 Coding Sequences—The full-length SIR3 gene was amplified by PCR and 3 µg of the amplified DNA was digested with mung bean nuclease (10–50 units; New England Biolabs) for 3–16 h at 30 °C. The nuclease was inactivated with 0.1% SDS and the DNA was exchanged into 30 µl of 10 mM Tris-Cl (pH 8.5) using a QIAquick spin column (Qiagen). Five µl of the nuclease reaction mixture was analyzed by agarose gel electrophoresis to confirm the extent of the digestion (supplemental Fig. S1A), and the remainder of the library was incubated with T4 DNA polymerase (1 unit of T4 polymerase/µgof DNA, 50 µg/ml bovine serum albumin, and 100 µM each of 4 dNTPs) at 12 °C for 15 min to blunt end the DNA. T4 DNA polymerase was inactivated by adding 1 µl of 0.5 M EDTA and heating the sample at 75 °C for 20 min. The mixture of insert DNAs was then exchanged into 10 mM Tris-Cl (pH 8.5) using a spin column (Qiagen).

A double-stranded adaptor sequence with a 5-bp long 3' overhang was ligated to the blunt-ended DNA inserts prior to cloning into pT7CF (supplemental Fig. S1B). The adaptors consist of the following 5'-phosphorylated oligonucleotides: 13-mer, 5'-PO₄-TCCGTCACCAAAC-3', and 8-mer, 5'-PO₄-GTGACGGA-3'. The DNA strands were annealed by heating at 50 °C for 10 min then slowly cooling to room temperature and incubating at 16 °C overnight before aliquoting and storing frozen (–80 °C) until use. Adaptors were appended to the blunt-ended inserts by incubating with T4 DNA ligase (New England Biolabs) at 16 °C overnight in the presence of a 40-fold molar excess of adaptor DNA. T4 DNA polymerase was heat-inactivated and the ligation mixture was purified by agarose gel electrophoresis to remove excess unligated adaptors and trivial SIR3 fragments. Following electrophoresis, a block of agarose containing a smear of DNA fragments ranging in length from 500 to 2900 bp (the full-length SIR3 gene) was cut out and then embedded in another agarose gel and refocused into a discrete band by reversing the polarity of electrophoresis. The resulting DNA band was cut out and extracted into 30 µl in 10 mM Tris-Cl (pH 8.5) (Qiagen). This selection of inserts with a defined size proved crucial for the elimination of trivial gene fragments, enriching the library for the expression of polypeptides capable of folding into stable, functional domains (typically 100–300 amino acids in length).

The library was constructed by ligating the adapted DNA inserts into 25 ng of the linkered pT7CF vector overnight at 16 °C using T4 DNA ligase (New England Biolabs). T4 DNA ligase was inactivated at 65 °C (10 min) and the ligation mixture was precipitated with ethanol in the presence of glycogen carrier. The DNA pellet was subsequently taken up in 5 µl of sterile distilled water, transformed into 50 µl of electrocompetent E. coli BL21(DE3) cells, and the cells were recovered in 250 µl of SOC for 1 h. The entire ligation mixture was then plated on 245-mm square (Corning) LB agar plates containing 20 µg/ml chloramphenicol (Cam), and incubated at 37 °C overnight. The use of complementary linker and adaptor sequences prevented self-ligation of the pT7CF vector, enriching the library for recombinant clones, and greatly enhancing the effectiveness of downstream functional assays.

Library Expression and CAT Toxicity—Gene expression in pT7CF is under the control of a T7 promoter that is compatible with the commonly used E. coli expression host BL21(DE3) (36). Transformed cells can be grown either on Cam to select for the expression of functional CAT fusion proteins or on ampicillin to recover all transformants containing pT7CF. We initially intended to use the fused CAT reporter group to select for clones expressing highly soluble proteins that confer resistance to chloramphenicol. However, cells transformed with the parental pT7CF vector had an extremely low plating efficiency on media containing the inducer of protein expression, isopropyl -D-thiogalactoside (IPTG), suggesting that overexpression of CAT is toxic to the cells. The transformed cells survived best under basal expression conditions without IPTG (supplemental Fig. S2A). Similar evidence of toxicity associated with high level CAT expression was obtained during growth in liquid media containing IPTG and 30 µg/ml chloramphenicol (Fig. S2A, inset). Further tests confirmed that high levels of soluble CAT are expressed under inducing conditions at 37 °C. The amount of basal or "leaky" expression is also considerable, and only 3-fold less than the induced levels of CAT expression. It is unknown why this modest increase in intracellular CAT levels is strongly toxic to E. coli, and this finding was not reported in earlier studies (37, 38). Importantly, the resulting transformation efficiency of the pT7CF vector was opposite to our expectations: basal expression of CAT conferred resistance to Cam, whereas higher levels of CAT expression caused a precipitous decrease in the number of transformants.

Correlation of Protein Solubility with Cam Resistance—To determine whether similar growth behavior was observed for CAT fusion proteins, we cloned genes for 2 proteins of known solubility into pT7CF upstream of CAT. The human homologue of Rad23, hHR23B (409 amino acids), is wholly soluble when overexpressed in BL21(DE3). Conversely, -integrase (356 amino acids) is completely insoluble when expressed in E. coli (not shown). Again, no recombinants survived growth in the presence of IPTG. Under basal expression conditions, only the cells expressing the soluble hHR23B-CAT fusion survived, albeit at low levels of Cam (25 µg/ml, supplemental Fig. S2B). The insoluble -integrase-CAT fusion failed to confer resistance to Cam. Further tests showed that both proteins fused to CAT were expressed at significantly lower basal levels than the CAT protein alone. However, fusion to CAT did not alter the solubility of either hHR23B (fully soluble) or -integrase (completely insoluble). Unlike the CAT fusion vector created by Davidson and co-workers (pCFN1) (37), which saw survival of soluble fusions up to 400 µg/ml Cam, the stringency of our selection was higher, with cells expressing soluble hHR23B-CAT fusion only showing limited survival up to 50 µg/ml Cam. In contrast, cells expressing the insoluble -integrase-CAT do not survive with or without IPTG, even at limiting concentrations of Cam. Addition of IPTG caused a 3–5-fold increase in expression of these CAT fusion proteins and was toxic to the cells (supplemental Fig. S2B). However, under basal expression levels we could exploit this stringent Cam selection to eliminate insoluble/out of frame fusion proteins from the library (see "Results").

Screening for Sir3 Domains That Interact with Sir4-CC—The GST-Sir4-CC bait protein and a point mutant (GST-Sir4-CC_(M1307N)) that fails to bind Sir3 were overexpressed in E. coli and purified separately (27). The purified proteins were stored at –80 °C in 20 mM Hepes (pH 8.0), 350 mM NaCl, 15 mM glutathione, 5 mM DTT and glycerol. For the pulldown assays, the proteins were exchanged into 50 mM Hepes (pH 8.0), 500 mM KCl, and 2.5 mM DTT. GST-Sir4-CC (100 µl of 10 µg/ml) was bound to the wells of a 96-well glutathione-derivatized assay plate (BD Biosciences) for 1 h at 4 °C with agitation on a rotating platform. The protein solution was then decanted and the plates were used immediately for the pulldown assay.

Sir3-CAT library transformants surviving Cam selection were chosen at random and grown at 37 °C with agitation in 96-well deep well blocks containing 1.5 ml of LB with 20 µg/ml Cam. The CAT fusion proteins were expressed for 2 h with 1 mM IPTG. The cells were pelleted at 3,000 x g and the cell pellets were frozen at –80 °C. The plates were thawed at room temperature for 5 min and the cell lysates were resuspended in 200 µl of cold (4 °C) Lysis Buffer (50 mM Hepes (pH 8.0), 500 mM KCl, 1 mM DTT, 1 mM EDTA, and 0.2 mg/ml PMSF). Twenty µl of lysozyme (200 µg/ml) and 0.1% Triton X-100 were added to lyse the cells by shaking (300 rpm on a rotating platform) at 4 °C for 30 min. The lysate was treated with 20 µl of 0.5 µg/ml DNase I in 900 mM MgSO₄ and 100 mM MnCl₂ and shaken for 30 min more. Insoluble material was pelleted at 7,300 x g for 30 min and the supernatant was transferred to a new 96-well plate. Prior to the pulldown assay, the total protein concentrations of cell lysates expressing Sir3-CAT fusion proteins were measured by a Bradford assay. The working concentration of each lysate was normalized by diluting 4-fold into buffer containing 50 mM Hepes (pH 8.0), 500 mM KCl, 5% (w/v) milk, and 0.05% Tween 20, to achieve concentrations within a 1.5-fold range of the mean value. Alternatively, the amounts of CAT activity in the lysates could be measured with a colorimetric assay based on 5,5'-dithiobis(2-nitrobenzoic acid) reduction (39).

One hundred µl each of the diluted cell lysates containing Sir3-CAT fusion proteins (100 µg/ml of total protein concentration) were incubated with immobilized GST-Sir4-CC in 96-well plates (see above) that were rotated (200 rpm) at 4 °C for 1 h. After the binding reaction, the plates were washed five times (250 µl/well) with 50 mM Hepes (pH 8.0), 500 mM KCl, and 0.05% Tween 20, and the bound Sir3-CAT fusion proteins were eluted by washing the plates with 200 µl of 50 mM Hepes (pH 8.0), 500 mM KCl, 15 mM glutathione, and 1 mM DTT for 30 min. The eluates were blotted onto nitrocellulose using a vacuum manifold (slot-blot apparatus; Bio-Rad). The membrane was then blocked with 5% milk in Tris-buffered saline at room temperature for 30 min. The bound proteins were treated with anti-CAT antibody derivatized with digoxigenin (Roche) for 2 h at 25 °C in Tris-buffered saline containing 5% milk and 0.05% (v/v) Tween 20. The plates were washed twice with Tris-buffered saline, 0.05% Tween 20, incubated for 45 min with anti-digoxigenin antibody derivatized with peroxidase (Roche), and washed two more times. A final wash without detergent was followed by addition of diaminobenzidine to visualize the antibody-bound proteins.

Directed Subcloning, Expression, and Purification of N-terminal-truncated Sir3 Proteins—Sir3-(464–978) and Sir3-(464–728) were expressed with C-terminal His₆ tags using pET24a (Novagen). BL21(DE3) cells were grown to mid-log phase at 37 °C, chilled on ice, then protein expression was induced with 1 mM IPTG by incubation overnight at 16 °C. The cells in growth medium were pelleted and resuspended in Lysis Buffer containing 50 mM Hepes (pH 7.6), 500 mM KCl, 5% glycerol, 5 mM -mercaptoethanol, and 1 mM PMSF then lysed with a pneumatic cell disrupter (Avestin Inc.). Insoluble material was pelleted at 30,000 x g for 30 min and the lysate was bound to nickel-NTA-Sepharose (Amersham Biosciences). The column was washed once with cell Lysis Buffer and the bound protein was eluted in Lysis Buffer plus 250 mM imidazole. The eluted protein was subsequently diluted 5 times in Dilution Buffer (50 mM Hepes (pH 7.6), 1 mM EDTA, 1 mM PMSF, and 1 mM DTT), and any precipitate was pelleted at 30,000 x g and the resulting supernatant was filtered. The clarified sample in low salt buffer was loaded onto a SP-Sepharose ion exchange column (Amersham Biosciences) and eluted with a gradient of 100–500 mM KCl in buffer containing 50 mM Hepes (pH 7.6), 1 mM EDTA, 1 mM PMSF, and 1 mM DTT. Pooled fractions containing Sir3 were concentrated (3 ml) and loaded onto a Superdex 200 gel filtration column. The eluted protein was concentrated to 10 mg/ml in S200 Buffer (50 mM Hepes (pH 7.6), 500 mM KCl, 1 mM DTT, 1 mM PMSF, and 1 mM EDTA). Sir3 was divided into aliquots, flash frozen in liquid nitrogen, and stored at –80 °C prior to use.

Thioredoxin fused to Sir3-(464–572) with a C-terminal His₆ tag was expressed from pET32a (Novagen) using BL21(DE3) cells grown to log phase at 37 °C before chilling on ice and inducing protein expression overnight at 16 °C with 1 mM IPTG. The thioredoxin-Sir3 fusion protein was purified using the same scheme described above, except that Q-Sepharose (Amersham Biosciences) ion exchange chromatography was substituted for the SP-Sepharose step.

To create heterodimers consisting of two different Sir3 fragments, SIR3-(464–978) (Fig. 1A) was cloned into the first cassette of pDuetRSF (Novagen) in conjunction with different N-terminal truncated fragments of SIR3 cloned into the second cassette. The coding sequence for SIR3-(464–978) was amplified by PCR using a C-terminal primer engineered to encode an 8-amino acid Strep-II tag (NH₂-WSHPQFEK-COO). This PCR product was subcloned into the NcoI/NotI sites of cassette 1 of pDuetRSF, which eliminates the N-terminal His₆ tag encoded by the parental vector. Various C-terminal domains of SIR3 were cloned into the second cassette with a C-terminal His₆ tag appended.

Pulldowns to Identify the Minimal Dimerization Domain of Sir3—Pairs of Sir3 fragments were co-expressed in different combinations using the pDuetRSF vectors described above. BL21(DE3) cells harboring these vectors were grown to mid-log phase at 37 °C, chilled on ice, and protein expression was induced with 1 mM IPTG followed by incubation at 16 °C overnight. The induced cells were pelleted and resuspended in Lysis Buffer (50 mM Hepes (pH 7.6), 500 mM KCl, 5% glycerol, 20 mM imidazole, 5 mM -mercaptoethanol, and 1 mM PMSF), and then sonicated. Insoluble cell debris was pelleted at 30,000 x g and the lysate was loaded onto a nickel-Sepharose column (Amersham Biosciences). The column was washed and the bound protein was eluted in Lysis Buffer plus 250 mM imidazole and then analyzed as described in the text. Interaction of GST-Sir3-(464–728) with Sir3-(832–978)-His₆ was tested by separately expressing these two fragments with 1 mM IPTG at 16 °C overnight and co-lysing the cells. Pulldowns on nickel-Sepharose were performed identically as above.

Purification of a Heterodimeric Sir3 Complex—A soluble cell lysate containing the Sir3-(464–978)/Sir3-(522–978) was prepared and bound to nickel-NTA-Sepharose using the conditions described above. The eluant from the nickel column was loaded directly to StrepTactin Superflow (iBA), washed with Strep Buffer (50 mM Hepes (pH 7.6), 500 mM KCl, 5 mM -mercaptoethanol, 1 mM PMSF, and 1 mM EDTA) and subsequently eluted in Strep Buffer plus 2.5 mM desthiobiotin. Following elution from the StrepTactin matrix, the Sir3 heterodimer was diluted 5 times in Dilution Buffer to reduce the salt concentration to 100 mM KCl. Sir3 was loaded to a SP-Sepharose ion exchange column and eluted over a 100–500 mM KCl gradient in a buffer containing 50 mM Hepes (pH 7.6), 1 mM DTT, 1 mM PMSF, and 1 mM EDTA. Pooled fractions containing Sir3 were concentrated to 3 ml and loaded onto a Superdex 200 gel filtration column. Fractions containing the Sir3 heterodimer were concentrated to 10 mg/ml in S200 Buffer (50 mM Hepes (pH 7.6), 500 mM KCl, 1 mM DTT, 1 mM PMSF, and 1 mM EDTA). Sir3 was divided into aliquots, flash frozen in liquid nitrogen, and stored at –80 °C prior to use.

Purification of Sir4-CC—GST-Sir4-CC and GST-Sir4-CC_(M1307N), a point mutant defective in binding to Sir3, were expressed and purified as described (27). For measuring binding by fluorescence anisotropy, a Sir4 construct spanning residues 1198–1358 was mutated to replace cysteine 1326 with serine to enable the unique labeling of cysteine 1262 with the thiol-reactive probe, tetramethylrhodamine-5-maleimide (TMR). The reaction of Sir4-CC with DTNB (40) revealed that both cysteines of native Sir4-CC are solvent accessible. The mutation of Cys¹³²⁶ to serine does not affect Sir3 binding, and the Sir4_(C1326S) mutant protein was purified as described (27). Prior to labeling, Sir4-CC_(C1326S) was reduced with 10 mM DTT in degassed G-25 Buffer (50 mM Hepes (pH 7.2), 100 mM KCl, and 1 mM EDTA). The DTT was removed by passing Sir4-CC over a NAP-25 (Sephadex G-25, Amersham Biosciences) desalting column equilibrated with G-25 Buffer. Sir4-CC was labeled with a 10-fold molar excess of TMR label at 25 °C for 2 h and then quenched with -mercaptoethanol. Excess label was removed by passing Sir4-CC over a NAP-25 column and Sir4-CC concentration was determined by a Bradford protein assay. Based on the absorption at 553 nm (the absorption maximum of TMR) we estimate that 70% of the protein molecules are conjugated to the fluorophore.

Purification of Sir4-CC Complexed with the Sir3 Interacting Domain—GST-Sir4-CC and an interacting fragment of Sir3 (residues 464–728; Fig. 1A) were expressed separately as described here and previously (27). The induced cells from both cultures were pelleted and lysed together in Lysis Buffer before pelleting the insoluble debris. The lysate was bound by batch to glutathione-agarose 4B for 2 h at 4 °C, packed in a column, and washed extensively with Wash Buffer (Lysis Buffer without glycerol or PMSF). The Sir3-(464–728)-Sir4-CC complex was eluted by incubating the column overnight with PreScission protease to release Sir4-CC from GST. The eluted complex was diluted 5-fold in Dilution Buffer and then filtered and loaded to a Q-Sepharose ion exchange column (Amersham Biosciences) that had been pre-equilibrated in Dilution Buffer + 100 mM KCl. The complex was eluted over a 100–500 mM KCl gradient in a buffer containing 50 mM Hepes (pH 7.6), 1 mM DTT, 1 mM PMSF, and 1 mM EDTA. The pooled fractions were concentrated to 3 ml and loaded onto a Superdex 200 gel filtration column. Following purification, the protein complex was concentrated to 10 mg/ml in S200 Buffer (50 mM Hepes (pH 7.6), 500 mM KCl, 1 mM DTT, 1 mM PMSF, and 1 mM EDTA) and then divided into aliquots, flash frozen in liquid nitrogen, and stored at –80 °C until use. The Sir4-CC-Sir3-(464–978) complex was purified in an analogous manner and stored at 4 °C for several days before use for dynamic light scattering (DLS) experiments.

Protein Binding by Fluorescence Anisotropy—Fluorescent, TMR-labeled Sir4-CC (50–65 nM) was incubated with increasing concentrations of purified Sir3 fragments, as indicated under "Results." Fluorescence anisotropy data were collected at 553 nm (excitation) and 575 nm (emission) wavelengths using a Photon Technology International fluorimeter. For fluorescence polarization measurements, the intensities of emitted light parallel (I_||) and perpendicular (I) to the vertical plane of excitation were measured, to calculate the anisotropic fluorescence emission (A).

(Eq. 1)

In this equation, G is equal to the ratio of intensities perpendicular and parallel to the horizontal plane of excitation (I/I_||). Anisotropy data were fit to a Hill model (41) using SigmaPlot (SYSTAT).

(Eq. 2)

A_max and A₀ are the maximum and minimum anisotropy values, respectively. The variable [S3] is the Sir3 monomer concentration, [S3]₅₀ is the concentration of Sir3 at one-half A_max, and n is the Hill coefficient.

Pulldowns of Various Sir3 Fragments with GST-Sir4-CC—Pulldown assays using immobilized GST-Sir4-CC and various Sir3 fragments as well as the Sir3-(464–978)/Sir3-(522–978) heterodimer were performed as described (27).

Dynamic Light Scattering—Sir4-CC was purified and concentrated in 50 mM Hepes (pH 7.6), 100 mM KCl, 1 mM EDTA, and 2 mM DTT. DLS measurements were carried out using a DynaPro molecular sizing instrument (ProteinSolutions Inc.). A 40-µl sample was passed through a 0.02-µm filter and a 12-µl aliquot of the filtrate was analyzed. Light scattering data were analyzed using the Dynamics (version 5) software (42).

To confirm that Sir4 did not form disulfide bond-linked aggregates, two cysteines in the Sir4-CC were alkylated with iodoacetamide, after first reducing the proteins with 10 mM DTT in a degassed buffer containing 50 mM Hepes (pH 7.6), 100 mM KCl, and 1 mM EDTA and then removing DTT by passing the sample over a NAP25 column (Amersham Biosciences) before alkylation for 30 min at room temperature with 10 mM iodoacetamide. Excess iodoacetamide was subsequently removed by exchange of the buffer with a NAP-25 column. The alkylated protein failed to react with dithionitrobenzoate, confirming the chemical blockage of both cysteines.

	RESULTS

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

 
Isolation of Sir3 Domains Interacting with Sir4—The yeast silent information regulators Sir3 and Sir4 are large proteins that are prone to aggregate, perhaps reflecting their biological activity of spreading along chromatin to form transcriptionally silent DNA (43). To perform crystallographic studies, we sought to identify soluble, interacting protein domains derived from Sir3 and Sir4. In the absence of identified homologues of the yeast Sir proteins, we could not take advantage of amino acid sequence conservation to guide the directed subcloning of individual domains. Instead, we developed a functional screen to identify soluble domains of Sir3 that interact with Sir4 derived from an expression library of random Sir3 protein fragments.

To identify functional domains of yeast Sir3, we constructed a nested deletion library in the pT7CF CAT reporter vector as described under "Experimental Procedures" (Fig. 1B). We first examined the distribution of gene fragments represented in the library by plating cells on ampicillin (100 µg/ml) to select for all transformants. Ninety-six colonies were picked at random and the sizes of the cloned inserts in pT7CF were determined by PCR amplification using primers flanking the MluI cloning site (Fig. 1C). The results show that the DNA size selection imposed during library construction successfully enriched for inserts capable of encoding protein fragments longer than 150 amino acids, which are more likely to adopt a stably folded conformation (11, 12, 27). The distribution of insert sizes is skewed toward a lower limit of 500 bp imposed during construction of the library (see "Experimental Procedures"). There are few inserts larger than 950 bp in length, and none of the 96 transformants examined from the SIR3 library are longer than 1.7 kb. Importantly, the library contains only a minimal number of self-ligated pT7CF plasmids or clones with trivially small inserts.

Sixteen clones selected on ampicillin plates were sequenced, to map the distribution of SIR3 fragments in the library (Fig. 1D). The inserts are distributed across the entire length of the SIR3 gene, vary in size, and have unique 5' and 3' termini. This sampling (from 5,000 to 20,000 total transformants typically obtained) is consistent with a comprehensive library of overlapping SIR3 gene fragments. The randomly distributed 5' and 3' ends of the inserts is further evidence that DNA cleavage with mung bean nuclease does not exhibit appreciable sequence bias, and it is an efficient means for creating nested deletion libraries.

View larger version (27K): [in this window] [in a new window]   FIGURE 1. Nested deletion expression library methodology. A, Sir protein domains used in this article. Asymmetric Sir3 dimerization activity was localized to two domains: residues 464–728 and 832–978. At a minimum, interactions with Sir4 require the Sir3 region spanning residues 464–728. It was previously reported that a Sir3 fragment spanning residues 623–910 binds to histones H3 and H4 (12). The Sir4 protein dimerizes through a C-terminal coiled-coil domain (Sir4-CC; residues 1198–1358) that specifically interacts with Sir3 (26, 27). B, the pT7CF (T7 CAT Fusion) CAT reporter vector and construction of a SIR3-nested deletion library for expression screening. SIR3 fragments cloned into the unique MluI restriction site of pT7CF in the correct reading frame are expressed as CAT fusion proteins with an N-terminal His6 affinity tag. Random SIR3 fragments generated by digestion with mung bean nuclease were subcloned into pT7CF after the addition of short adaptors, which are complementary to linkers added to the pT7CF cloning site. Adaptors are represented as wedges and the linker as a white box. C, size distribution of SIR3 gene fragments in the unselected library (transformants grown on 100 µg/ml ampicillin) shows that most inserts are longer than 500 bp, the lower cut-off used during construction of the library. D, the DNA sequences of 16 randomly selected clones grown in the absence of Cam shows that the unselected library contains fragments from across the entire SIR3 coding region.

View larger version (26K): [in this window] [in a new window]   FIGURE 2. A nested deletion expression library identifies functional protein domains of Sir3. A, a protein-protein interaction assay identifies Sir3 fragments that interact with GST-Sir4-CC. Cell lysates containing Sir3-CAT fusion proteins were incubated in multiwell plates conjugated with GST-Sir4-CC. The plates were washed to remove unbound proteins, and the bound Sir3-CAT proteins were visualized by Western blotting using anti-CAT antibody. The initial positives from the library screen (left panel) were confirmed by directed pulldown experiments using GST-Sir4-CC immobilized on glutathione-conjugated plates (right panel). As controls for binding specificity, a Sir4 point mutant (Sir4(Met1307Asn); labeled G4*) and underivatized plate failed to bind any of the Sir3-CAT proteins. B, Sir3 fragments capable of interacting with Sir4 include residues previously shown to be important for this interaction (residues 464–522 (27)), but these functional Sir3 fragments are considerably shorter than the previously characterized interacting region of Sir3 (residues 464–978). C, Coomassie-stained SDS-PAGE gel showing the purified complex of Sir3-(464–728) and Sir4-CC-(1198–1358). These compact domains form a stable protein complex.

The genes encoding the soluble Sir3 fusion proteins were sequenced, along with some that formed insoluble inclusion bodies. In all cases, the clones surviving Cam selection have SIR3 and CAT coding sequences in the correct reading frame. Although survival on Cam plates does not guarantee the expression of soluble proteins, the initial Cam selection does eliminate the 5 of 6 of clones predicted to have SIR3 gene fragments inserted out-of-frame with respect to the N-terminal His₆ affinity tag. Additionally, only of these in-frame constructs will be fused to the CAT gene in the correct reading frame. Thus, the initial Cam selection has enriched the library for 6% of the SIR3 random gene fragments that are in the correct reading frame with respect to both the His₆ tag and CAT.

Because the initial Cam selection does not eliminate those Sir3-CAT proteins that are poorly soluble, a functional screen was devised to isolate highly expressed, soluble domains of Sir3 that interact with Sir4. For this purpose, we created a protein interaction screen in a 96-well format. The bait protein, GST-Sir4-CC (Fig. 1A), was expressed and purified as described previously (27) then immobilized on glutathione-conjugated multiwell plates. Three-hundred twenty clones were picked at random from a Sir3 fragment library and grown in deep well blocks. Following induction, the soluble supernatants were adjusted to 100 µg/ml total protein and incubated with the immobilized Sir4-CC. The Sir3 fragments that bound to Sir4-CC were identified by an enzyme-linked immunosorbent assay using a polyclonal antibody against CAT followed by detection with a secondary antibody conjugated to horseradish peroxidase (Fig. 2A). The immobilization of CAT fusion proteins on GST-coated plates substantially decreased CAT activity, as previously reported (38), and this precluded a colorimetric activity assay using dithionitrobenzoate to detect CAT activity in the bound fraction. Immune detection of the eluted proteins proved much more sensitive and reproducible.

Using these methods, we isolated 8 overlapping fragments of the Sir3 protein that specifically interact with Sir4-CC (Fig. 2B). These protein fragments bind specifically to Sir4-CC and fail to bind underivatized plates or a Sir4-CC_(M1307N) point mutant that does not interact with full-length Sir3 (Fig. 2A) (27). Furthermore, we confirmed that CAT alone did not bind to the GST-Sir4-CC protein or to underivatized plates. The Sir3 fragments that bind strongly to Sir4-CC cluster in a region between Sir3 residues 370 and 781 (Fig. 2B), in agreement with previous data implicating residues N-terminal to amino acid 522 as important for interaction with Sir4 (27). One of 8 identified Sir3 fragments (Fig. 2B, white bar) bound Sir4-CC with significantly lower efficiency (not shown), and its sequence only partially overlaps with those of other fragments that bind.

View larger version (32K): [in this window] [in a new window]   FIGURE 3. Sir3-(464–728) is the minimal fragment sufficient for binding the coiled-coil of Sir4. A, a series of Sir3 fragments beginning at residue 464 and ending at different C termini were tested for their ability to stably interact with GST-Sir4-CC. Thioredoxin (Trx) fusions Trx-Sir3-(464–522), Trx-Sir3-(464–572), and C-terminal His6-tagged Sir3-(464–728)-His6 and Sir3-(464–978)-His6 were separately expressed and co-lysed with GST-Sir4-CC. Each Sir3 fragment was incubated with GST-Sir4-CC, bound to glutathione-Sepharose, rigorously washed, and the eluted as described in the text. The input and eluted fractions were visualized by SDS-PAGE and Coomassie Blue staining. B, a summary of experiments testing the interactions of thioredoxin-tagged Sir3 fragments with Sir4-CC reveals that although Sir3 residues 464–522 are necessary for a stable interaction, only Sir3 fragments extending to residue 728 are sufficient for binding Sir4. These results were confirmed by titrating labeled Sir4-CC against the same Sir3 fragments, and measuring binding by fluorescence anisotropy (see Fig. 5A). The Sir3-(464–522) and Sir3-(464–572) proteins also failed to bind Sir4-CC after proteolytic removal of the thioredoxin tag (data not shown).

We next asked if the screen identified the smallest fragment of Sir3 that is capable of interacting with the Sir4-CC. We had previously shown that Sir3 residues 464–522 are essential for this interaction (44), and yet the partially overlapping clone that bound weakly to Sir4-CC (Fig. 2B, white bar) suggested that a smaller polypeptide could be sufficient. We expressed from E. coli two thioredoxin (Trx) fusions of Sir3 encompassing these residues, Trx-Sir3-(464–522) and Trx-Sir3-(464–572), and tested their abilities to bind to GST-Sir4-CC immobilized on glutathione-agarose (Fig. 3). In contrast to monomeric Sir3-(464–728) or dimeric Sir3-(464–978), neither of these smaller Sir3 fragments stably interact with Sir4-CC. Sir3 residues 464–522 are essential, but apparently not sufficient, for the interaction with Sir4. We conclude that Sir3-(464–728) is the minimal domain of Sir3 that stably interacts with Sir4.

Remarkably, all of the Sir4-binding fragments of Sir3 lack the C-terminal region that reportedly functions in Sir3 dimerization (12, 28). This may reflect the size constraint imposed during construction of the expression library, resulting in cloned inserts encoding on average 150–300 amino acids. Furthermore, these results imply that a second region within the C terminus of Sir3 mediates dimerization separate from residues responsible for its interaction with Sir4. This possibility spurred us to further investigate the molecular architectures of these Sir protein complexes.

Identification of Two Separate Domains Mediating Sir3 Dimerization—A C-terminal fragment of Sir3 spanning residues 307–978 dimerizes (12, 28), whereas Sir3-(464–728) failed to dimerize (Fig. 1A), indicating that residues in the extreme C-terminal region of Sir3 promote dimerization. To identify the minimal dimerization region, different pairs of Sir3 fragments were co-expressed from the dual expression cassette of the pRSF vector. A C-terminal fragment, Sir3-(464–978), was inserted into one cassette, in-frame with a 3' Strep-II affinity tag. Several different C-terminal fragments of Sir3 were cloned into the second cassette of pRSF to generate C-terminal His₆-tagged proteins (Fig. 4). These combinations of Sir3 fragments were co-expressed in E. coli BL21(DE3) and cell lysates containing the resulting Sir3 heterodimers were incubated with Ni²⁺-NTA resin in a pulldown experiment. All Sir3 proteins with N termini beginning before residue 832 efficiently pulled-down Sir3-(464–978) (Fig. 4). Sir3 fragments with N-terminal truncations beyond residue 832 were insoluble. Therefore, we operationally define the minimal dimerization region as Sir3 residues 832–978 (Fig. 1A).

We considered two alternative models of the Sir3 dimer interaction, a "parallel" dimer in which residues interact in trans with the same residues of the opposite subunit, and an "antiparallel" dimer in which C-terminal residues interact with N-terminal residues of the opposite subunit. To test these models, we attempted to pulldown a GST fusion of the monomeric Sir3-(464–728) with Sir3-(832–978)-His₆ on Ni²⁺-NTA resin. Strikingly, these two separate domains interact strongly as shown by Coomassie staining and Western analysis (Fig. 4A). The interaction was specific as GST-Sir3-(464–728) failed to bind nickel beads and GST alone was not pulled down by Sir3-(832–978)-His₆ (data not shown). These results indicate that C-terminal residues 832–978 associate asymmetrically with the Sir4 interaction domain (residues 464–728), consistent with an antiparallel Sir3 dimer. Thus, the C-terminal half of Sir3 comprises two separate domains that associate with one another (Fig. 5C), with the N-terminal domain additionally associating with Sir4 (Fig. 1A).

View larger version (28K): [in this window] [in a new window]   FIGURE 4. Sir3 dimerization is mediated by an association of C-terminal residues 834–978 with the Sir4-binding domain. A, the input and eluted fractions from nickel-NTA resin containing either Sir3-(464–978) fused to a Strep-II affinity tag or GST-fused Sir3-(464–728) were analyzed by Coomassie Blue staining and Western blot analysis. Both Sir3 fragments were expressed alone, or co-expressed with the bait fragment Sir3-(832–978)-His6. The binding of immobilized Sir3-(832–978)-His6 to Sir3-(464–978)-II or GST-Sir3-(464–728) was visualized using a monoclonal antibody directed at the Strep-II or GST tags, respectively. The efficient association of Sir3-(832–978) with both Sir3 fragments shows that the dimerization function of Sir3 is mediated by C-terminal residues 832–978 and Sir4-binding residues 464–728. B, summary of all C-terminal fragments of Sir3-His6 tested for interaction with Sir3-(464–978) containing a Strep-II tag. All C-terminal fragments of Sir3, except the insoluble Sir3-(910–978) protein (IN, insoluble; ND, not determined in legend), interact with the longer Sir3 fragment.

The binding affinities of these Sir3 proteins for Sir4 were determined by fluorescence anisotropy using TMR-labeled Sir4-CC (see "Experimental Procedures"). The binding isotherms indicate that monomeric and dimeric Sir3 bind tightly to the Sir4-CC. Whereas the monomeric Sir3-(464–728) protein binds to Sir4 with lower affinity (EC₅₀ = 118 nM; Fig. 5A and Table 1), the affinity of the dimer can only be given qualitatively because we could not reliably measure the anisotropic fluorescent emission of TMR-labeled Sir4-CC at concentrations below 50 nM. The apparent affinity (EC₅₀) of Sir3-(464–978) for the labeled Sir4-CC systematically decreased as the probe concentration approached this lower limit (supplemental Fig. S3). In agreement with pulldown experiments (Fig. 3), the anisotropic emission was unchanged by addition of the smaller Sir3-(464–572) fragment to labeled Sir4-CC (Fig. 5A).

We previously reported calorimetric data showing that Sir3 and Sir4 interact with a 1:1 molar stoichiometry (27). We attempted to independently determine the binding stoichiometry at concentrations of Sir4-CC well above the K_app for binding to Sir3. However, for a cooperative binding interaction the K_app = [EC₅₀]ⁿ, where n = the Hill coefficient. We could not achieve high enough concentrations of the Sir4-CC to ensure stoichiometric binding conditions (Fig. 5B). The cooperative nature of the Sir3-Sir4 interaction and the antiparallel association of the Sir3 dimers are suggestive of a self-assembly process capable of generating higher order protein complexes, which might underlie the spreading of Sir protein complexes on chromatin (43) (Fig. 5C).

View larger version (21K): [in this window] [in a new window]   FIGURE 5. Sir3-(464–728) is the minimal domain that binds cooperatively to Sir4 and the complex is stabilized by Sir3 dimerization. A, titration of 65 nM Sir4-CC with Sir3-(464–728) (circles), as measured by fluorescence anisotropy. The sigmoidal binding curve is indicative of a cooperative binding interaction between Sir3 and Sir4. A thioredoxin-tagged Sir3 protein encompassing residues 464–572 does not bind to Sir4-CC (diamonds). Although these residues are essential for a stable interaction with Sir4, they are not sufficient (see Fig. 3). B, titration of Sir4-CC (1.5 µM) with increasing concentrations of Sir3-(464–978). The concentrations of the Sir3-(464–978) monomer are shown. Even at Sir4-CC concentrations 10-fold greater than the EC50 concentration of the binding interaction, a cooperative binding interaction with Sir3-(464–978) is observed, preventing an analysis of subunit composition under stoichiometric binding conditions. C, our results show that two separate regions of the Sir3 protein contribute to its interactions with Sir4. Several different subunit stoichiometries within the Sir3-Sir4 complex are consistent with the results of fluorescence anisotropy, isothermal titration calorimetry, and x-ray crystallographic analyses (Ref. 27 and this work). The Sir3-(464–728) fragment is a monomer but binds cooperatively to Sir4. Dimerization via the C-terminal and Sir4-interacting domains of Sir3 further increases the affinity of Sir3 for Sir4-CC. The asymmetric dimerization of Sir3 suggests a mechanism for the formation of higher order arrays of Sir proteins. One potential model (shown here, left) allows for C-terminal dimerization residues 832–978 (moon shaped domain) to interact with the Sir4 interaction domain of a second Sir3 molecule (circular domain, residues 464–728). Using two domains for dimerization frees the second Sir3 molecule to recruit another monomer of Sir3. Alternatively, a symmetric Sir3/Sir4 interaction (shown here, right) may assemble complexes cooperatively through Sir3 residues 464–728 to form higher order arrays.

The smaller Sir3-(522–978) subunit lacks the minimal Sir4-interacting region (Fig. 1A) and it does not stably bind to Sir4-CC, either as a homodimer or in complex with the larger Sir3-(464–978) subunit (Fig. 6A). Only the Sir3-(464–978) homodimer stably interacts with Sir4. The lower affinity of the mixture of Sir3-(464–978) and Sir3-(522–978) subunits for Sir4-CC (EC₅₀ = 345 nM versus EC₅₀ < 100 nM for the Sir3-(464–978) dimer) can be explained by the 1:2:1 ratio of functional Sir3 dimers: non-functional heterodimers: non-functional homodimers in the mixture (Fig. 6B). We conclude that both subunits of the Sir3 dimer must contain the Sir4-interacting domain (residues 464–728) for stable binding to Sir4.

The Sir4 Coiled-coil Forms Higher Order Complexes Alone and in Complex with Sir3—A key feature of current models for the assembly of silent DNA is the self-association of the Sir proteins and their spreading on chromatin (1). Sir3 forms higher order oligomers in vitro (33), and we investigated if the Sir4-CC also could self-associate into larger oligomers. Although the crystal structure of the Sir4-CC (27) gave no indication of an extended packing arrangement, the Sir4-CC is prone to self-associate during purification.

The native oligomeric state of purified Sir4-CC was measured at different protein concentrations by DLS. The autocorrelation data were analyzed by a regularization algorithm to determine the hydrodynamic radius (R_H) of the primary intensity peak (46) and converted to M_r using a log-log relationship of globular protein standards. The results reveal a clear dependence of the native M_r on protein concentration (Fig. 7A), consistent with the Sir4-CC forming complexes of increasing size as the protein concentration is increased from 2 to 12 mg/ml. These results are not consistent with the formation of random aggregates because the average R_H of the Sir4-CC increases more sharply than polydispersity with increasing protein concentration (Fig. 7B). This self-association behavior of Sir4-CC was also observed in high ionic strength buffer (0.5 M KCl) that would disfavor nonspecific electrostatic interactions. To rule out the formation of disulfide bond-linked oligomers of the Sir4-CC, two exposed cysteines on the surface of the coiled-coil (27) were alkylated by treatment with iodoacetamide. The alkylated Sir4-CC shows a similar increase in M_r with increasing protein concentration, and we conclude that the Sir4-CC has an intrinsic self-association behavior that could contribute to the cooperative assembly of Sir protein complexes.

View larger version (24K): [in this window] [in a new window]   FIGURE 6. Both subunits of the Sir3 dimer are required to stabilize the complex with the Sir4 coiled-coil. A, pulldown assay using GST-Sir4-CC immobilized on glutathione-agarose as bait to capture Sir3 dimers formed by co-expression of Sir3-(464–978) and Sir3-(522–978) (Fig. 1A). Proteins in the input (I), flow through (FT), wash, and elution (E) fractions were visualized by SDS-PAGE and Coomassie Blue staining. Only Sir3 homodimers consisting of the larger Sir3-(464–978) subunits are captured in complex with Sir4-CC. Sir3-(522–978) homodimers and Sir3-(522–978)/Sir3-(464–978) heterodimers fail to stably interact with Sir4-CC. Right, Western blot analysis of input and bound fractions using anti-Strep-II antibody to visualize Sir3-(464–978) and anti-His6 to visualize Sir3-(522–978). B, titration of 65 nM TMR-labeled Sir4-CC with the same mixture of Sir3 homodimers and heterodimer as described in panel A. The binding affinity of this mixture of Sir3 dimers ([EC]50 = 345 nM) is 3–4-fold less than the affinity of fully functional Sir3 dimers (cf. Fig. 5 and Table 1), consistent with Sir4 binding activity being limited to the Sir3-(464–978) dimers constituting one-fourth of the total dimers in this mixture of subunits.

	DISCUSSION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

 
Coupling Reading Frame Selection to Functional Screens of Libraries Expressing Random Protein Fragments—Several publications have correlated the overexpression of soluble proteins in E. coli with the activity of a fused reporter group. In vivo protein solubility assays based on fusing the coding sequence of a target protein to the gene encoding CAT, the -fragment of -galactosidase, or the green fluorescent protein have been reported (37, 47, 48). CAT is a trimer of 25-kDa subunits conferring Cam resistance, and the enzyme remains active when fused to the C terminus of several different proteins (49, 50). Davidson and co-workers (37) showed that expression of insoluble proteins fused to CAT conferred less resistance to Cam than did expression of soluble CAT fusion proteins. In their experiments, clones expressing insoluble CAT fusions typically survived at only one-half the Cam concentration (typically <250 µg/ml Cam) that was tolerated by cells expressing soluble CAT fusions. By selecting for growth at higher Cam concentrations, transformants expressing soluble proteins could be enriched over insoluble counterparts present in excess in the starting library.

View larger version (18K): [in this window] [in a new window]   FIGURE 7. Higher order complexes of Sir4 alone and in complex with Sir3. A, left, the hydrodynamic radius (RH) of Sir4-CC changes as a function of protein concentration in buffers containing 100 mM KCl (triangles), 500 mM KCl (squares), or 100 mM KCl supplemented with iodoacetamide to block disulfide bond formation (circles). The apparent increase in molecular weight suggests that the Sir4-CC assembles into higher order oligomers as the protein concentration is raised. Right, polydispersity values (CP/RH, %) for the same samples shown on the left side of panel A. The decreasing polydispersity of Sir4-CC with increasing protein concentration is inconsistent with the formation of large nonspecific protein aggregates, and instead the data suggest a progressive assembly of Sir4-CC multimers. B, as for Sir4-CC alone, the hydrodynamic radius of an equimolar complex of Sir3-(464–978)/Sir4-CC increases in step with increasing protein concentration. The error bars represent the absolute polydispersity of the sample. In contrast to Sir4-CC alone, the complex with Sir3 shows a modest increase in polydispersity as the concentration is increased.

We developed several different vectors with a T7 promoter for the expression of CAT fusion proteins. However, all of these vectors expressed similarly high levels of CAT that caused cell death upon induction with IPTG (not shown). The pCFN1 vector, kindly provided by Dr. Alan Davidson, expresses CAT fusion proteins using an IPTG-inducible trc promoter (37). Under inducing conditions, pCFN1 produced about 3-fold less CAT protein than the pT7CF vector. E. coli JM101 cells transformed with pCFN1 survived on Cam plates containing IPTG, whereas BL21(DE3) cells transformed with pT7CF did not grow on Cam plates with IPTG (data not shown). Although the source of CAT toxicity is unknown, it could be caused by a metabolic imbalance related to depletion of acetyl-CoA.

After Cam selection for in-frame coding sequences, many of the Sir3-CAT fusion proteins are poorly soluble, and some are proteolyzed in E. coli (not shown), necessitating further screening for well behaved and functional protein domains. False positives lacking a Sir3 polypeptide can arise either by proteolysis of an expressed Sir3-CAT fusion or by the internal initiation of mRNA translation. Recent developments in reading frame selection methods could address these potential problems (52–58). However, most of these limitations are resolved by coupling the CAT-based selection to a functional assay.

A secondary protein interaction screen of the Sir3-CAT fusions surviving Cam selection eliminated the insoluble constructs from the library and successfully identified well expressed Sir3 fragments that interact with Sir4. Th