<?xml version="1.0" ?>
<tei>
	<teiHeader>
		<fileDesc xml:id="0"/>
	</teiHeader>
	<text xml:lang="en">
		<p>The UPS is a fundamental component of normal cell growth and<lb/> proliferation. The UPS
			is also important for cancer cell growth, as<lb/> highlighted by the recent approval of
			the proteasome inhibitor,<lb/> Velcade, for the treatment of relapsed multiple myeloma
				<ref type="biblio">8</ref> . Recent<lb/> studies have investigated the mechanism of
			Velcade action by<lb/> examining the transcriptional response to the drug in human
			and<lb/> yeast cells <ref type="biblio">9,10</ref> . These studies indicate that
			proteasome inhibition does<lb/> not substantially alter bulk transcription of the genome
			in myeloma<lb/> cells or in Saccharomyces cerevisiae. However, a group of genes are<lb/>
			repressed by Velcade, including human growth and survival genes<lb/> and yeast genes
			involved in the biosynthesis of amino acids <ref type="biblio">9,10</ref> . These<lb/>
			yeast genes are regulated by the b-ZIP transcriptional activator Gcn4,<lb/> which
			promotes the expression of more than 500 genes <ref type="biblio">11</ref> . Gcn4 is
			a<lb/> target for UPS-mediated degradation <ref type="biblio">12</ref> through the
			E3-ubiquitin<lb/> ligase SCF Cdc4 . Ligases comprise the final, substrate recognition
			step<lb/> of the ubiquitination cascade. SCF Cdc4 ubiquitinates and targets for<lb/>
			proteolysis Gcn4 molecules that have been phosphorylated by the<lb/> cyclin-dependent
			kinases (CDKs) Srb10 and Pho85 <ref type="biblio">(refs 5, 7)</ref>. We<lb/> have
			attempted to explain the role of the UPS in the function of Gcn4<lb/> and other
			activators.<lb/></p>

		<p>To assess the impact of proteolysis on Gcn4 function, we treated a<lb/> yeast strain
			(pdr5D) that is sensitive to proteasome inhibitors <ref type="biblio">10</ref> with<lb/>
			the Velcade analogue MG132 <ref type="biblio">(ref. 4)</ref>. Reverse
			transcriptase-mediated<lb/> polymerase chain reaction (RT–PCR) analyses confirmed that,
			as<lb/> for Velcade <ref type="biblio">10</ref> , treatment with MG132 substantially
			reduced the<lb/> transcription of several Gcn4 targets, including HIS4 and CPA2, in<lb/>
			minimal medium, in comparison with dimethylsulphoxide (DMSO)<lb/> alone <ref
				type="figure">(Fig. 1a and Supplementary Fig. S1a)</ref>. Similar results were<lb/>
			obtained when Gcn4 was highly induced by amino acid starvation<lb/> (<ref type="figure"
				>Fig. 1a</ref>, 2Leu; see Supplementary Information for discussion of<lb/> media).
			These effects were largely dependent on Gcn4, because we<lb/> observed similar decreases
			in the transcription of a reporter driven<lb/> exclusively by six Gcn4-binding sites 13
				(<ref type="figure">Fig. 1a</ref>, GCRE6–LacZ).<lb/></p>

		<figure>Figure 1 | UPS-dependent proteolysis positively regulates inducible<lb/>
			transcriptional activators. a, Indicated strains were grown in minimal or<lb/>
			starvation (2Leu) medium, treated with MG132 (50 mM) or DMSO, and<lb/> processed for
			RT–PCR of the indicated transcripts. pdr5D enables the uptake<lb/> of MG132 into yeast.
			b, A pdr5D strain was induced with galactose, treated<lb/> with MG132 and prepared for
			RT–PCR of GAL1 and ACT1. c, wild-type<lb/> (WT) and pre1-1, pre4-1 strains were grown in
			minimal medium at 27 8C<lb/> and processed for RT–PCR of HIS4 and ACT1. d, Conditionally
			expressed<lb/> ubiquitin was depleted from strains while expression from a
			complementing<lb/> plasmid encoding either WT ubiquitin, no ubiquitin (D) or
			K48R-ubiquitin<lb/> was induced (see Supplementary Information). Samples were prepared
			for<lb/> RT–PCR as above. e, WT (white bars), cdc34-2 (grey bars) and gcn4D (black<lb/>
			bars) strains expressing LacZ from the HIS3 or GCRE6 promoter were<lb/> grown in minimal
			or starvation medium (3-AT) at 30 8C and then processed<lb/> to test b-galactosidase
			(b-gal) activity. Standard deviations are from three<lb/> replicates. n.d., not
			detectable. f, WT, cdc34-2 and cdc4-1 strains were grown<lb/> at 27 8C or shifted to 30
			8C for 1 h. RT–PCR was performed for HIS4, ARG1<lb/> and ACT1.<lb/></figure>

		<p>We also tested the proteasome dependence of two regulons (GAL<lb/> and INO) that might
			have eluded the Velcade microarray analysis<lb/> because they were not expressed under
			the growth conditions that<lb/> were employed. MG132 substantially decreased induction
			of the Gal4<lb/> target, GAL1, on the addition of galactose <ref type="figure">(Fig.
				1b)</ref>. Similarly, on the<lb/> removal of inositol, transcription of INO1, a
			target of the activators<lb/> Ino2 and Ino4, was largely abrogated by MG132 <ref
				type="figure">(Supplementary<lb/> Fig. S1b)</ref>. In all cases treatment with MG132
			did not affect the<lb/> constitutive transcription of ACT1.<lb/></p>

		<p>These findings indicate that the 26S proteasome might promote<lb/> the function of some
			inducible transcriptional activators. Focusing<lb/> on Gcn4, we tested additional
			components of the UPS for their<lb/> impact on transcription. A strain (pre1-1 and
			pre4-1) mutated for the<lb/> peptidase activity of the proteasome showed reduced
			transcription of<lb/> HIS4 in comparison with a wild-type (WT) strain <ref type="figure"
				>(Fig. 1c)</ref>
			<ref type="biblio">14</ref> . We also<lb/> tested transcription of HIS4 in strains that
			conditionally express<lb/> alternative versions of ubiquitin <ref type="biblio">6</ref>
			. When endogenous ubiquitin was<lb/> depleted in the presence of a vector plasmid (UbD)
			or depleted in a<lb/> cell expressing K48R ubiquitin, HIS4 messenger RNA levels
			were<lb/> sharply decreased in comparison with a depleted strain that<lb/> expressed WT
			ubiquitin <ref type="figure">(Fig. 1d)</ref>. Again, the ACT1 mRNA remained<lb/>
			constant. These findings confirm that ubiquitination and proteolysis<lb/> are important
			for Gcn4 function. Importantly, because the<lb/> K48R mutant cannot form chains that
			target substrates to the<lb/> proteasome, these findings also indicate that, in contrast
			to<lb/> the regulation proposed for Gal4–VP16 <ref type="biblio">(ref. 15)</ref> and
			c-Myc <ref type="biblio">16,17</ref> ,<lb/> mono-ubiquitination might not be able to
			sustain Gcn4 activity.<lb/></p>

		<p>We next examined the impact of Cdc34–SCF Cdc4 , the specific<lb/> E2–E3 ubiquitin ligase
			for Gcn4 <ref type="biblio">(refs 5, 7)</ref>, on activator function.<lb/>
			Gcn4-dependent expression of b-galactosidase was evaluated in WT,<lb/>
			temperature-sensitive cdc34-2, and gcn4D strains. After growth at the<lb/>
			semi-permissive temperature of 30 8C, the cdc34-2 strain exhibited a<lb/> fourfold
			decrease (relative to WT) in reporter expressed from the<lb/> HIS3 promoter (HIS3P) in
			minimal medium and a roughly 15-fold<lb/> reduction on starvation by 3-aminotriazole
			(3-AT, <ref type="biblio">ref. 11</ref>) <ref type="figure">(Fig. 1e)</ref>.<lb/>
			Expression from GCRE6–LacZ was also compromised about 10-fold<lb/> in cdc34-2 <ref
				type="figure">(Fig. 1e)</ref>. In addition, expression from both promoters was<lb/>
			extinguished in gcn4D cells under all conditions <ref type="figure">(Fig. 1e)</ref>.
			RT–PCR<lb/> analysis was then used to assess effects on endogenous targets.<lb/>
			Gcn4-dependent transcription of HIS4 and ARG1 was defective in<lb/> both cdc34-2 and
			cdc4-1 (a thermosensitive allele of the SCF F-box<lb/> protein Cdc4) strains at 30 8C
				<ref type="figure">(Fig. 1f)</ref>. In contrast, ACT1 transcript<lb/> levels in all
			strains were similar. These results point to a stimulatory<lb/> role for Cdc34–SCF Cdc4
			in Gcn4-mediated transcription.<lb/></p>

		<p>To explore the molecular basis of the stimulation of Gcn4 function<lb/> by the UPS, we
			examined the effect that proteasome inhibition has on<lb/> the abundance and
			ubiquitination of Gcn4. Western blotting showed<lb/> that MG132 increased the abundance
			of chromosomally encoded<lb/> Gcn4–Myc9 and led to the appearance of a
			high-molecular-mass<lb/> ladder <ref type="figure">(Fig. 2a, lanes 1 and 2)</ref>.
			Immunoprecipitation and western<lb/> analysis with anti-ubiquitin antibodies confirmed
			that this ladder<lb/> was ubiquitinated Gcn4 <ref type="figure">(Fig. 2a, lanes 5, 6, 9
				and 10)</ref>. Because<lb/> transcription mediated by Gcn4 was strongly repressed by MG132<lb/>
			<ref type="figure">(Fig. 1)</ref>, ubiquitination was presumably insufficient to sustain
			Gcn4<lb/> activity. To confirm the specificity of Gcn4 ubiquitination, we<lb/> repeated
			the analysis with a gcn4–3T2S strain. Gcn4–3T2S lacks<lb/> five phosphorylation sites,
			is no longer phosphorylated by Srb10 or<lb/> Pho85 nor ubiquitinated by SCF Cdc4 in
			vitro, and is stabilized <ref type="biblio">7</ref> . As<lb/> predicted, the 3T2S strain
			had much lower levels of ubiquitinated<lb/> Gcn4 <ref type="figure">(Fig. 2a, lanes 3,
				4, 7, 8, 11 and 12)</ref>.<lb/></p>

		<p>We next investigated promoter occupancy by Gcn4 and Gal4 and<lb/> recruitment of RNA
			polymerase II (polII) to target genes. Chromatin<lb/> immunoprecipitation (ChIP) assays
				<ref type="biblio">18</ref> revealed a substantial increase<lb/> in the association
			of Gcn4–Myc9 with the HIS4 promoter in cells<lb/> treated with MG132 <ref type="figure"
				>(Fig. 2b, 9E10)</ref>. Similar experiments showed<lb/> little change in levels of
			TAP-tagged Gal4 at the GAL1 promoter on<lb/> treatment with MG132 <ref type="figure"
				>(Fig. 2c, IgG)</ref>. The galactose-dependent<lb/> decrease in promoter-bound
			Gal80–TAP was also unaffected by<lb/> MG132 <ref type="figure">(Supplementary Fig.
				S3)</ref>. When the ChIP was performed<lb/> with polII, MG132 reduced the signal for
			the promoter, open reading<lb/> frame (ORF) and terminator regions of the Gcn4 target
			HIS4,<lb/> whereas the signal for the ACT1 ORF was unaffected <ref type="figure">(Fig.
				2b,<lb/> polII)</ref>. MG132 also decreased the polII ChIP signal for the GAL1<lb/>
			promoter and ORF <ref type="figure">(Fig. 2c, polII)</ref> and the INO1 promoter (data
			not<lb/> shown).<lb/></p>

		<p>The ChIP analysis was extended to cdc34-2 and cdc4-1 strains. At<lb/> 30 8C, more
			Gcn4–Myc9 <ref type="figure">(Fig. 2d, 9E10)</ref> but less polII was associated<lb/>
			with the HIS4 promoter in the SCF mutants than in the WT.<lb/> Recruitment of polII to
			ACT1 was unaffected in the SCF mutants.<lb/> Results from all the ChIP analyses closely
			parallel the transcription<lb/> results and imply that the UPS is important for
			sustaining the<lb/> interaction of RNA polymerase II with the targets of some
			activators<lb/> despite the increased accumulation of activator (for Gcn4) at the<lb/>
			promoter.<lb/></p>

		<figure>Figure 2 | Defects in the UPS lead to the accumulation of ubiquitinated<lb/> Gcn4
			and impair the association of RNA polymerase II with Gcn4 and Gal4<lb/> targets. a,
			pdr5D strains expressing Gcn4–Myc9 (WT) or Gcn4–3T2S–<lb/> Myc9 (3T2S) were grown in
			minimal medium and treated with MG132 or<lb/> DMSO. Immunoprecipitations (IPs) were
			performed with 9E10 antibodies<lb/> recognizing Myc9, and subsequent western blots of
			the input (lanes 1–4) or<lb/> immunoprecipitation (IP; lanes 5–12) samples were probed
			with 9E10 (lanes<lb/> 1–8) or anti-Ub antibodies (lanes 9–12). b, A WT strain, as above,
			was<lb/> processed for ChIP analysis with 9E10 and anti-polII antibodies. PCR was<lb/>
			performed to amplify the promoter, ORF and terminator regions of HIS4<lb/> and the ORF
			of ACT1. c, ChIP analysis of Gal4–TAP and polII was<lb/> performed with
			galactose-induced strains treated with MG132 or DMSO.<lb/> The promoter and ORF regions
			of GAL1 were amplified. IgG was used to<lb/> retrieve Gal4–TAP. d, ChIP analysis of
			Gcn4–Myc9 and polII was performed,<lb/> as in b, with WT, cdc34-2 and cdc4-1 strains
			that were grown at 27 8C, then<lb/> shifted to 30 8C for 1 h. kb, kilobases; n.d., not
			done.<lb/></figure>

		<p>To test whether turnover of the activator itself can promote<lb/> function, we evaluated
			the stabilized Gcn4–3T2S <ref type="biblio">(ref. 7)</ref>. In minimal<lb/> medium the
			levels of total protein <ref type="figure">(Fig. 3a)</ref> and promoter-associated<lb/>
			Gcn4–Myc9 <ref type="figure">(Fig. 3b)</ref> were about twofold to threefold higher in
			the<lb/> gcn4–3T2S strain. Despite such increases, gcn4–3T2S did not alter the<lb/>
			expression of Gcn4 targets in minimal <ref type="figure">(Fig. 3c, SM þ ; Fig. 3e,<lb/>
				CDC34)</ref> or starvation medium <ref type="figure">(Fig. 3c, 2Leu)</ref>,
			indicating a<lb/> possible decrease in specific activity. Most importantly,
			gcn4–3T2S<lb/> partly alleviated the deleterious effects of an impaired UPS on<lb/>
			Gcn4-dependent transcription. For example, MG132 inhibited<lb/> HIS4 transcription in
			gcn4–3T2S much less than in GCN4 <ref type="figure">(Fig. 3d)</ref>.<lb/> In addition,
			cdc34-2 diminished HIS3P–LacZ expression only about<lb/> 1.2-fold in gcn4–3T2S, in
			comparison with more than 2.5-fold in<lb/> GCN4 <ref type="figure">(Fig. 3e)</ref>. The
			cdc34-2 mutation had no impact on gcn4D<lb/>
			<ref type="figure">(Fig. 3e)</ref>. A similar epistatic relationship was seen on
			deletion of<lb/> SRB10 and PHO85, which stabilizes Gcn4 to a similar extent to the<lb/>
			gcn4–3T2S mutation <ref type="biblio">7</ref> . Deletion of both CDKs (DD) slightly
			increased<lb/> LacZ expression (1.2-fold) and, as with gcn4–3T2S, cdc34-2 only<lb/>
			mildly affected expression (1.2-fold) in the DD strain <ref type="figure">(Fig.
				3f)</ref>. The<lb/> suppression of cdc34-2 required the deletion of both kinases,
			because<lb/> srb10D and pho85D single mutants remained relatively sensitive to<lb/>
			cdc34-2 <ref type="figure">(Fig. 3f)</ref>. We note that gcn4–3T2S is not completely
			refrac-<lb/>tory to UPS inhibition, indicating that, in addition to Gcn4, the UPS<lb/>
			might also promote transcription through other factors. Neverthe-<lb/>less, these
			findings indicate that proteolysis of CDK-phosphorylated<lb/> Gcn4 by means of the UPS
			might be important in sustaining<lb/> maximal expression of Gcn4 targets and that, in
			the absence of<lb/> phosphorylation, Gcn4 activity is less dependent on its
			turnover.<lb/></p>

		<p>Components of the UPS have been posited to activate transcrip-<lb/>tion by multiple
			mechanisms <ref type="biblio">2,3</ref> . In numerous examples, including<lb/> the
			activation of Gcn4 in response to ultraviolet radiation <ref type="biblio">19</ref>
			and<lb/> transcription of NF-kB <ref type="biblio">(ref. 2)</ref> and oestrogen receptor
			targets <ref type="biblio">20</ref> , the<lb/> UPS seems to mediate signalling upstream
			of the activator. As<lb/> discussed in <ref type="figure">Supplementary Fig. S3</ref>,
			this mechanism probably does<lb/> not account for our findings. Meanwhile, subunits of
			the 19S cap of<lb/> the proteasome have been suggested to have a positive role in<lb/>
			transcription that is independent of their proteolytic function <ref type="biblio"
				>21,22</ref> .<lb/> In addition, it has been proposed that the ubiquitination
			of<lb/> Gal4–VP16 and c-Myc transiently increases the activity of these<lb/> factors
			before proteolysis <ref type="biblio">15–17</ref> . Our results differ substantively
			from<lb/> these examples, in that neither the 19S cap (whose activity is not<lb/> known
			to be affected by inhibition of the 20S proteases <ref type="biblio">23</ref> ) nor<lb/>
			ubiquitination was sufficient to achieve maximal transcription of<lb/> Gcn4 targets <ref
				type="figure">(Figs 1d and 2a)</ref>. Instead, we found that inhibition of<lb/> the
			proteasome and genetic manipulations of the UPS, the CDKs for<lb/> Gcn4, and Gcn4 itself
			all provided evidence that turnover of Gcn4<lb/> normally enhances its
			function.<lb/></p>

		<p>Proteasome activity also seems to sustain inducible transcription<lb/> mediated by Gal4
			and Ino2/4. Interestingly, activation of promoter-<lb/>associated Gal4 in galactose
			medium requires Srb10-dependent<lb/> phosphorylation <ref type="biblio">24,25</ref> ,
			and this activated form has a short half-life <ref type="biblio">24</ref> .<lb/> This
			indicates that Gal4 might be regulated by degradation in a<lb/> manner similar to that
			of Gcn4. We previously proposed a model<lb/> consistent with these current findings in
			which proteolysis is required<lb/> to remove &apos;spent&apos; activators and to reset
			the promoter <ref type="biblio">3</ref> . The initial<lb/> &apos;pioneer round(s)&apos;
			of transcription would not involve the UPS, but<lb/> subsequent rounds would be
			stimulated by turnover of the spent,<lb/> promoter-bound activator to allow binding of a
			fresh molecule. This<lb/> mechanism places Gcn4, Gal4 and Ino2/4 into a class of
			regulatory<lb/> factors—including securin, p21 and p27—whose activity is required<lb/>
			early in a process but whose subsequent turnover or removal<lb/> promotes completion of
			the process or subsequent reaction cycles.<lb/> We call this phenomenon &apos;activation
			by destruction&apos; and believe that,<lb/> given the diversity of the examples listed
			in Supplementary Table S2,<lb/> it might represent a regulatory mechanism for a large
			class of factors<lb/> and might be an important determinant of infection and
			disease.<lb/></p>

		<head>METHODS<lb/></head>

		<p>Yeast strains, growth conditions and extract preparation. A complete list of<lb/> yeast
			strains used in this study is provided in Supplementary Table S1. All strains<lb/> were
			derived from the S288C background, except RJD 2505 and RJD 3137–3141,<lb/> which were
			derived from the W303 background. Strains were constructed and<lb/> grown in accordance
			with standard protocols <ref type="biblio">26</ref> . A description of the various<lb/>
			media used in the study is given in Supplementary Information. MG132<lb/> (American
			Peptide) was added to cultures of pdr5D strains to a final concen-<lb/>tration of 50 mM.
			All extracts were prepared by lysis with glass beads (Sigma) in a<lb/> Fast Prep (Bio
			101) device. Ubiquitin derivative analysis <ref type="figure">(Fig. 1d)</ref> is
			described in<lb/> the Supplementary Methods.<lb/></p>

		<p>RT–PCR analysis. mRNA was prepared using RNeasy kits (Qiagen). Total<lb/> mRNA (200 ng)
			and 10 pmol of oligo(dT) were used to reverse-transcribe<lb/> complementary DNA
			(Stratagene). One-tenth of the cDNA reaction was then<lb/> used for 20–22 cycles of PCR
			and products were resolved on 2% agarose gels.<lb/> Primer sequences are available from
			the authors on request.<lb/></p>

		<p>b-Galactosidase assays. The HIS3P–LacZ, HIS4P–LacZ and GCRE–LacZ repor-<lb/>ter
			constructs and the protocol to measure b-galactosidase activity were as<lb/> described
			previously <ref type="biblio">27</ref> . Reported activity was normalized to total
			extract protein<lb/> as measured by bicinchoninic acid assay (Pierce). Relative
			activities are reported<lb/> with average WT activity set to 1.<lb/></p>

		<p>Western blots. Except as noted, extracts were prepared by immediate boiling of<lb/> cell
			pellets in 2 £ Laemmli SDS sample buffer followed by lysis with glass beads.<lb/> Equal
			amounts of total protein (as judged by Coomassie staining) were resolved<lb/> by
			SDS–PAGE. Blots were probed with 9E10 antibodies to recognize Gcn4–Myc9<lb/> or FK1
			antibodies (Affiniti) to recognize ubiquitin. Horseradish peroxidase-<lb/>coupled goat
			anti-mouse secondary antibodies (Bio-Rad) were used for<lb/> detection.<lb/></p>

		<figure>Figure 3 | The UPS has little effect on the activity of stable,<lb/>
			non-phosphorylated versions of Gcn4. a, GCN4 (WT), gcn4–3T2S or<lb/> gcn4D strains were
			grown in minimal medium and processed for western<lb/> blotting with 9E10 antibodies.
			Protein half-life data (t 1/2 ) were reported<lb/> previously 7 . n.a., not applicable.
			b, ChIP analysis of the above strains was<lb/> performed for the HIS4 promoter. c,
			RT–PCR of the indicated transcripts<lb/> were performed with WTand gcn4–3T2S strains
			grown in minimal (Min) or<lb/> starvation (2Leu) medium. d, RT–PCR analysis of HIS4 and
			ACT1 was<lb/> performed with WTand 3T2S strains in the presence and absence of
			MG132.<lb/> e, CDC34 (open bars) and cdc34-2 (filled bars) strains with the
			HIS3P-LacZ<lb/> reporter and expressing either WT, 3T2S or null (gcn4D) versions of<lb/>
			GCN4 were grown in minimal medium and processed for b-galactosidase<lb/> (b-gal)
			activity. Standard deviations were calculated from three replicates.<lb/> f, WTand
			cdc34-2 strains with a HIS4P-LacZ reporter and harbouring WTor<lb/> deleted versions of
			SRB10 and/or PHO85 were treated as in e.<lb/></figure>

		<p>Immunoprecipitations and detection of ubiquitinated Gcn4. DMSO-treated<lb/> or
			MG132-treated cultures were treated with formaldehyde (final concentration<lb/> 1%) for
			20 min to trap ubiquitinated intermediates. Crosslinking was quenched<lb/> and extracts
			were made in ChIP buffer <ref type="biblio">18</ref> . A 10% sample of the extract
			(whole cell<lb/> extract) was removed and boiled in SDS sample buffer. The remainder
			of<lb/> the extract was incubated at 4 8C with 9E10 antibodies coupled to Protein
			A–<lb/> Sepharose (Sigma) beads. Proteins were eluted and crosslinks were reversed
			by<lb/> being boiled in SDS sample buffer. Proteins samples were then processed for<lb/>
			western blotting. For FK1 (anti-ubiquitin) western blots, the nitrocellulose was<lb/>
			boiled before incubation with the antibody.<lb/></p>

		<p>ChIP assays. ChIP assays were performed as described <ref type="biblio">18</ref> ; 9E10
			antibodies were<lb/> used to immunoprecipitate chromatin fragments associated with
			Gcn4–Myc9 and<lb/> antibodies against the carboxy-terminal domain of the largest subunit
			of RNA<lb/> polymerase II (anti-polII; Covance) were used to immunoprecipitate
			fragments<lb/> associated with polII. Protein G–Sepharose beads (Amersham
			Biosciences)<lb/> were used to precipitate antibody–antigen complexes. Rabbit
			immunoglobulin<lb/> (Ig) G–agarose (Sigma) was used to immunoprecipitate
			Gal4-associated<lb/> fragments. Primer sequences are available from the authors on
			request.<lb/></p>

	</text>
</tei>