Media, growth conditions, and yeast strains:

The strains used in this study are listed in Table 1. They are all derivatives of W303 (Thomas and Rothstein 1989), with the exception of JRY2069 (J. Rine) and its derivatives, LPY2557 and LPY10974. Yeast extract/peptone/dextrose (YPD)-rich medium, supplemented synthetic medium (SC) lacking the appropriate nutrient for plasmid selection, minimal medium, and liquid sporulation medium were prepared as described (Sherman 1991). 5-Fluoroorotic acid (5-FOA) plates were prepared by adding 5-FOA to a final concentration of 0.1% to supplemented synthetic medium (Boeke et al. 1987). Yeast strains were grown at 30° unless otherwise noted.

Yeast genetic methods:

Standard yeast genetic methods were used for mating and sporulation. For recovery of double mutants, at least 18 and up to 54 tetrads/diploid were dissected. All markers segregated normally and mutant combinations were recovered at the expected frequencies. Yeast transformations were performed using the lithium acetate method (Ito et al. 1983). Gene disruptions were confirmed by molecular amplification and phenotypic characterization.

Plasmids:

The plasmids used in this study are listed in Table 2. pLP1727 was constructed by inserting the HindIII–BamHI HSL7 fragment of pLP1699 into pRS426 (2μ URA3 vector) opened with HindIII and BamHI. To make a plasmid containing a tagged version of HSL7 under the control of its own promoter, a PCR fragment containing HSL7-3HA∷KanMX was amplified from the genomic DNA of DLY4569 (strain from Daniel J. Lew) and cloned into the pCR-Blunt II-TOPO vector (Invitrogen, San Diego). Then this plasmid was digested with KpnI and NotI and the resulting 5.4-kb fragment was cloned into pRS314 (CEN TRP1 vector) digested with KpnI and NotI, yielding pLP1947. pLP1968 was obtained by direct mutagenesis on pLP1947. Expression of the fusion proteins Hsl7p-3HA and Hsl7p-G386A-R387A-3HA was confirmed by protein immunoblotting using the monoclonal antibody HA.11 from Covance. pLP1969 was created by inserting the KpnI–NotI HSL7-3HA∷KanMX fragment from pLP1947 into pRS412 (CEN ADE2 vector) opened with KpnI and NotI. pLP1970 was obtained by direct mutagenesis on pLP1968. A fragment containing wild-type GCN5 was excised from pLP1641 (2μ URA3 GCN5) by XhoI–EagI double digestion and subcloned into pRS414 digested with XhoI and EagI, yielding pLP2058. Wild-type RPD3 under the control of its own promoter was amplified by PCR from the genomic DNA of LPY5 (W303-1a) and this PCR fragment was cloned into the pCR-Blunt II-TOPO vector from Invitrogen. Then a 1.6-kb fragment containing RPD3 was excised from this vector by digestion with XhoI and PacI and subcloned into pFA6a-KanMX-13myc (Longtine et al. 1998) digested with SalI and PacI. Finally, a 3.8-kb band was excised by PvuII–SacI double digestion and ligated into pRS314 (CEN TRP1 vector) digested with SmaI and SacI. pLP1944 was obtained by direct mutagenesis on pLP1943. Expression of the tagged proteins Rpd3p-13myc and Rpd3p-H150A-H151A-13myc was confirmed by Western blotting analysis using an anti-myc antibody from the 9E10 hybridoma (ATCC CRL 1729). Mutations were generated using site-directed mutagenesis and confirmed by automated sequencing (UCSD Cancer Center). Sequences of the primers used in this study are available on request.

Dilution assays:

A single colony was inoculated in 3 ml of YPD or in SC lacking the appropriate amino acid for plasmid selection. Cells were grown for 24 hr at 30°. Cultures were normalized to an initial concentration of 1.0 A₆₀₀/ml in sterile water and serially diluted fivefold. Approximately 3 μl of each serial dilution were spotted onto the appropriate plates. Cells were incubated for 48 hr prior to photography.

Qualitative and quantitative mating assays:

For qualitative mating assays, cells were patched onto YPD plates for 24 hr and then replica plated to YPD to assay for growth and onto a lawn of mating-type testers, MATa (LPY142) or MATα (LPY78), on minimal medium to assay for successful diploid formation. For quantitative mating assays, cells were grown to midlogarithmic phase in YPD. They were diluted appropriately to obtain 400 colonies/plate. Then they were either plated directly onto YPD to quantify the total number of cells or mixed with MATa (LPY142) or MATα (LPY78) mating-type testers and plated onto minimal medium for diploid selection to evaluate the number of mating-competent cells. The mating efficiency is defined as the number of cells that mated per the total number of cells. Four experiments were performed for each strain.

Histone extraction and Western blot analysis:

Histones were purified essentially as described (Edmondson et al. 1996). Cells were grown to midlogarithmic phase and then 100 OD A₆₀₀ units of cells were harvested and washed in sterile water. Cells were resuspended in 5 ml of buffer A (50 mm Tris, pH 7.5, 30 mm DTT) and incubated at 30° for 15 min with gentle shaking. The pellet was washed with 10 ml of buffer S (1.2 m sorbitol, 20 mm HEPES, pH 7.4) and then resuspended in 5 ml of buffer S with 0.2 ml of 10 mg/ml zymolyase 100T and incubated at 30° with gentle shaking until spheroplasts had formed (∼1 hr). The following procedures were done on ice with prechilled solutions. A total of 10 ml of ice-cold buffer B (1.2 m sorbitol, 20 mm PIPEs, pH 6.8, 1 mm MgCl₂) was added to the spheroplasts and they were harvested at 4500 × g for 5 min at 4°. The pellet was resuspended in 5 ml of ice-cold NIB buffer [0.25 m sucrose, 60 mm KCl, 14 mm NaCl, 5 mm MgCl₂, 1 mm CaCl₂, 15 mm 2-(N-morpholino) ethanesulfonic acid, pH 6.6, 0.8% Triton X-100, 1 mm PMSF, 1 mm NaF], held on ice-water for 20 min, and then centrifuged at 4500 × g for 5 min at 4°. This step was repeated twice. Then the pellet was resuspended in 5 ml of wash buffer A (10 mm, Tris pH 8, 0.5% NP-40, 75 mm NaCl, 30 mm Na-butyrate, 1 mm NaF and 1 mm PMSF) and held in ice water for 15 min three times. Next, the pellet was washed twice in 5 ml of wash buffer B (10 mm Tris, pH 8, 0.4 m NaCl, 30 mm Na-butyrate, 1 mm NaF, and 1 mm PMSF) with a holding time in ice water of 10 min for each wash. After centrifugation at 4500 × g, pellet was resuspended with 1 ml of cold 0.4 n H₂SO₄ and incubated for 1 hr in ice water with occasional vortexing. After a 10-min centrifugation at 9500 × g, precipitation was initiated by adding 100% trichloroacetic acid to a final concentration of 20% and incubated in ice water for 1 hr. Precipitate was collected by a 30-min centrifugation at 13,000 × g, washed once in 500 μl of chilled acidified acetone (acetone with 0.1% HCl) and then in 500 μl of chilled 100% acetone, and air dried. Resulting histones were resuspended in 250 μl of 1× Laemmli loading buffer. A total of 10–25 μl of each sample was loaded onto a 18% SDS–acrylamide gel and transferred to nitrocellulose. The blot was probed with a 1:1000 dilution of an antibody directed against dimethylated arginine 17 of H3 (Upstate Biotechnology, Lake Placid, NY) followed by incubation with a 1:10,000 dilution of a goat anti-rabbit immunoglobulin G conjugated to horseradish peroxidase (W401B; Promega, Madison, WI). Detection was performed using the Western Lightning Chemiluminescence Reagent Plus from PerkinElmer Life Sciences.

Simultaneous deletion of HSL7 and specific genes encoding chromatin-modifying enzymes results in synthetic sickness and lethality:

To probe potential connections between the Hsl7p arginine methyltransferase and chromatin, we began by constructing a series of double mutants with genes encoding known histone-modifying enzymes. Deletion of GCN5, which encodes the histone acetyltransferase (HAT) targeting mainly Lys14 of histone H3, was synthetically lethal with deletion of HSL7. After dissection of diploids heterozygous for HSL7 and GCN5 deletions, spores corresponding to hsl7Δ gcn5Δ double mutants were never recovered (Figure 1A). In contrast, double-mutant spores covered by a plasmid-borne copy of GCN5 or HSL7 could be recovered (Figure 1B and data not shown), demonstrating that lethality is due to a failure of mitotic growth.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.—

The SAGA complex is required for viability in the absence of Hsl7p. (A) Synthetic lethality is specific between HSL7 and GCN5 deletions. Dissection plates are of the following diploid strains: hsl7Δ (LPY8660) crossed to gcn5Δ (LPY8279); hsl7Δ (LPY8660) crossed to hpa2Δ (LPY4692); hsl7Δ (LPY8660) crossed to hpa3Δ (LPY4690); and hsl7Δ (LPY9548) crossed to sas3Δ (LPY1590). Inferred (hsl7Δ gcn5Δ) or deduced (hsl7Δ hpa2Δ, hsl7Δ hpa3Δ, or hsl7Δ sas3Δ) double mutants are circled. (B) Enzymatic activity of Gcn5p is required in an hsl7Δ mutant. The plasmid-shuffling experiment is on a 5-FOA plate. The double mutant hsl7Δ gcn5Δ covered by a 2μ URA3-HSL7 plasmid (LPY8378) was transformed with wild-type GCN5 (pLP1518), gcn5-KQL (pLP1521), gcn5-LKN (pLP1520), or empty vector (pRS414). Transformants were streaked onto a Ura− Trp− growth control plate and onto a 5-FOA plasmid-shuffling plate. (C) Integrity of SAGA is required in an hsl7Δ mutant. Plasmid-shuffling experiment is on a 5-FOA plate. The double mutant hsl7Δ spt7Δ200 covered by a 2μ URA3 HSL7 plasmid (pLP1727) was transformed either with a plasmid carrying a wild-type HSL7 gene (pLP1699) or with an empty vector (pLRS314) yielding LPY10446 and LPY10445, respectively. Similarly, the hsl7Δ spt7Δ300 double mutant covered by a 2μ URA3 HSL7 plasmid (pLP1727) was transformed with either pLP1699 or pRS314, yielding LPY10448 and LPY10447. These strains were streaked onto Ura− Trp− growth control and 5-FOA plasmid-shuffling plates.

To investigate other possible connections between acetylation by the GCN5-related N-acetyltransferase superfamily and Hsl7p function, we determined whether other members of this family, like Hpa2p and Hpa3p, were essential in the absence of HSL7. Hpa2p acetylates histones H3 and H4 in vitro with a preference for Lys14 of histone H3 shared by Gcn5p (Angus-Hill et al. 1999), and Hpa3p shares 49% sequence identity and 81% sequence similarity with Hpa2p. We observed that deletion of these genes was not lethal in an hsl7Δ mutant strain (Figure 1A). Sas3p, a member of a different HAT acetyltransferase family, the MYST family, has the same substrate specificity as Gcn5p. It acetylates Lys14 of histone H3 and was shown to have an overlapping role with Gcn5p in cell cycle progression (Howe et al. 2001). To determine whether SAS3, like GCN5, plays an essential role in hsl7Δ mutant strains, we constructed a double-mutant hsl7Δ sas3Δ and assessed its viability. Deletion of SAS3 was not lethal when combined with HSL7 deletion (Figure 1A), demonstrating that the role of GCN5 in an hsl7Δ strain is distinct from the role of SAS3.

Gcn5p and other transcriptional coactivators that possess histone acetyltransferase activity, such as TAF_II250, p300, and the related CREB-binding protein (CBP), p300/CBP-associated factor (P/CAF), have been widely investigated. Critical domains and residues essential for their HAT activity have been characterized (Ogryzko et al. 1996; Neuwald and Landsman 1997; Wang et al. 1998). Interestingly, some of these transcriptional coactivators, notably p300/CBP, have critical functions that are independent of their HAT activity (Korzus et al. 1998; Sierra et al. 2003). Thus, we investigated whether Gcn5p HAT activity was required in an hsl7Δ mutant through plasmid-shuffling experiments (Boeke et al. 1987) with two well-characterized gcn5 point mutants. The gcn5-LKN mutant displays acetyltransferase activity similar to wild-type Gcn5p whereas gcn5-KQL encodes a dead catalytic mutant of Gcn5p (Wang et al. 1998). A gcn5Δ hsl7Δ strain covered by a URA3-HSL7 plasmid was transformed either with a TRP1 plasmid encoding gcn5-LKNp or with a TRP1 plasmid encoding gcn5-KQLp. After plating on 5-FOA to identify cells that had lost the wild-type HSL7 plasmid, the strain containing the gcn5-LKN mutation in combination with the HSL7 deletion grew, yet the strain carrying the gcn5-KQL dead catalytic mutant was unable to survive the plasmid shuffling (Figure 1B). Thus, the acetyltransferase activity of Gcn5p was specifically required for viability of an hsl7Δ strain.

Gcn5p acts as the catalytic subunit of several macromolecular complexes including SAGA (Spt7-Ada-Gcn5 acetyltransferase coactivator complex) and SLIK/SALSA, an alternate form of SAGA (Grant et al. 1997; Pray-Grant et al. 2002; Sterner et al. 2002). To determine whether the integrity of one or both of these Gcn5p-containing complexes is essential in hsl7Δ strains, we used the previously characterized spt7 mutant alleles spt7Δ200 and spt7Δ300. In spt7Δ200 mutants, the SLIK/SALSA complex is lost but SAGA is intact, whereas in spt7Δ300 mutants SLIK/SALSA is preserved but SAGA is lost (Wu and Winston 2002). Loss of the SLIK/SALSA complex in hsl7Δ resulted in synthetic sickness, whereas loss of SAGA caused death (Figure 1C). Thus, both SLIK/SALSA and SAGA integrity are important in the absence of HSL7, although SAGA function appears most critical.

In addition to GCN5, a gene encoding another key acetyltransferase, ESA1, has a significant role when HSL7 is deleted. ESA1 is an essential HAT in yeast and it preferentially acetylates Lys5 and Lys8 of nucleosomal H4, nucleosomal H2A, and Lys14 of H3 on free histones (Smith et al. 1998; Allard et al. 1999; Clarke et al. 1999). A strain carrying a thermosensitive allele of ESA1, esa1-L254P, lacking HAT activity at restrictive temperatures (Allard et al. 1999; Clarke et al. 1999) was crossed to an hsl7Δ strain and double mutants were assayed for viability. The esa1-L254P and hsl7Δ single mutants grew at 34°, but the double mutants esa1-L254P hsl7Δ grew more poorly at this temperature, indicating that the need for HAT activity of Esa1p is greater when HSL7 is mutated (Figure 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.—

Synthetic sickness of the esa1-L254P hsl7Δ double mutant. Cell growth of wild-type (LPY5), esa1-L254P (LPY5056), hsl7Δ (LPY8660), and esa1-L254P hsl7Δ (LPY10587) strains was assayed by plating fivefold dilutions of cells on YPD plates; cells were grown at 30° and 34°.

During the search for HATs required in the absence of HSL7, we unexpectedly observed that the deacetylase Rpd3p was also essential. As shown by Rundlett et al. (1996), rpd3 mutant strains have increased acetylation at Lys5 and Lys12 of histone H4 and RPD3-mediated deacetylation is involved in regulating both heterochromatic silencing and gene expression. After sporulation and dissection of two different diploids heterozygous for deletions of HSL7 and RPD3, none of the spores (0 of 45) corresponding to the double mutant hsl7Δ rpd3Δ germinated (Figure 3A and data not shown). Thus, simultaneous disruption of HSL7 and RPD3 is synthetically lethal. In parallel with the Gcn5p activity studies, we asked whether inactivation of Rpd3p deacetylase activity is lethal in the absence of Hsl7p. A double mutant hsl7Δ rpd3Δ covered by a URA3-HSL7 plasmid was transformed with a HIS3 plasmid encoding the previously characterized catalytic dead mutant rpd3-H150A-H151A (Kadosh and Struhl 1998). This strain was unable to survive the 5-FOA plasmid shuffling, indicating that Rpd3p deacetylase activity is required in hsl7Δ mutant strains (Figure 3B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.—

The Rpd3p/Sin3p complex is essential in the absence of Hsl7p. (A) Synthetic lethality between HSL7 and RPD3 deletion. The dissection plate of a diploid resulting from a cross between hsl7Δ (LPY8660) and rpd3Δ (LPY4889) is shown. Inferred double-mutant spores are circled. (B) Enzymatic activity of Rpd3p is required in hsl7Δ mutants. The double mutant hsl7Δ rpd3Δ covered by a 2μ URA3 HSL7 plasmid (LPY9295) was transformed with wild-type RPD3 (pLP1943), rpd3-H150A-H151A (pLP1944), or empty vector (pRS314). Transformants were streaked onto Ura− Trp− control growth and 5-FOA plasmid-shuffling plates. (C) Synthetic lethality between HSL7 and SIN3 deletion. The dissection plate of diploid resulting from a cross between hsl7Δ (LPY8660) and sin3Δ (LPY10429) is shown. Inferred double-mutant spores are circled.

Rpd3p exists in a large protein complex that contains the Sin3p corepressor (Kasten et al. 1997) and sin3 mutants have phenotypes very similar to rpd3 mutants. To address the requirement for the Rpd3p/Sin3p complex in the absence of Hsl7p, a diploid strain heterozygous for HSL7 and SIN3 deletions was sporulated and dissected. No spores corresponding to hsl7Δ sin3Δ double mutants were recovered from the 15 presumed double mutants obtained (Figure 3C and data not shown), indicating that SIN3 is an essential gene when HSL7 is deleted. Two other well-characterized histone deacetylases, Sir2p and Hda1p, were not required for viability of an hsl7Δ strain (Table 3 and data not shown). Thus, the RPD3/SIN3 deacetylase complex is specifically required for viability in the absence of Hsl7p.

Finally, we uncovered a genetic interaction between Hsl7p and the methyltransferase Set1p, which targets Lys4 of histone H3 (Nislow et al. 1997; Briggs et al. 2001). Simultaneous deletion of both SET1 and HSL7 often resulted in cell death as shown by tetrad dissection of a diploid strain heterozygous for SET1 and HSL7 deletions (Figure 4 and data not shown): of 45 presumed double mutants, 40 double mutants were inviable and 5 double mutants grew extremely poorly after dissection. However, deletion of DOT1, which encodes a methyltransferase that targets Lys79 of H3, or deletion of HMT1, which encodes an arginine methyltransferase that modifies Arg3 of H4, did not affect hsl7Δ mutant strains (Table 3 and data not shown). Thus, the deletion of HSL7 is detrimental for cell growth only in the absence of Set1p, but not in the absence of Hmt1p or Dot1p, two other histone methyltransferases identified in S. cerevisiae.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.—

Deletion of SET1 causes cell death of hsl7Δ mutants. Tetrad dissection of diploid hsl7Δ/+ set1Δ/+ resulting from a cross between LPY8660 and LPY11036. Double-mutant spores that grew poorly are underlined. Inferred double-mutant spores are circled.

Collectively, our data (summarized in Table 3) define critical and specific genetic interactions between HSL7 and genes encoding the known chromatin modifiers Gcn5p, Esa1p, Rpd3p, and Set1p. In addition, we show that GCN5/HSL7 and RPD3/HSL7 genetic interactions involve the enzymatic activity of Gcn5p and Rpd3p, respectively, and the complexes through which they act.

Deletion of HSL7 increases silencing at HMR and decreases silencing at the rDNA array:

In addition to their roles in transcriptional activation, chromatin-modifying enzymes also influence silencing, a form of gene regulation in which regions of the genome are epigenetically inactivated by changes in chromatin structure. In S. cerevisiae, several regions of the genome are subject to transcriptional silencing, including the telomere proximal regions, the silent mating-type loci HMR and HML and the rDNA array (Rusche et al. 2003). Mutation of SET1 affects silencing (Nislow et al. 1997); therefore, as a potential histone methyltransferase, we assayed the effects of deleting HSL7 on silencing.

First, we assayed silencing at telomeres using an ADE2 marker integrated or close to the telomeric region of the chromosome V right arm (Singer and Gottschling 1994). We found that silencing at telomeres is not affected by deletion of HSL7 (data not shown).

Next, to study silencing at HMR, we used strains containing hmrΔE∷TRP1. The hmrΔE∷TRP1 construct contains a deletion of the Rap1p-binding site at HMR-E, but retains the ARS consensus sequence and the Abf1p-binding site, resulting in a partial loss of silencing. In addition, the TRP1 gene has been inserted in place of most the a1 and all of the a2 genes at HMR. (Sussel and Shore 1991). In this assay, a sir2Δ strain completely disrupted for silencing at HMR was used as a control. When compared to a wild-type or a sir2Δ strain, an hsl7Δ hmrΔE∷TRP1 grew poorly on medium lacking tryptophane. As a control, we showed that deletion of HSL7 does not impair the ability of a strain with the wild-type TRP1 gene in its original genomic location to grow on medium lacking tryptophan (Figure 5A). These data suggest that Hsl7p may partially antagonize silencing.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.—

Silencing is affected in hsl7Δ mutants. (A) Deletion of HSL7 restores silencing in a strain slightly defective for silencing at HMR (hmrΔE∷TRP1). Fivefold dilutions of wild type (LPY4912), hsl7Δ (LPY10243), gcn5Δ (LPY8286), rpd3Δ (LPY10242), sir2Δ (LPY4980), set1Δ (LPY11037), and esa1-L254P (LPY5058) were plated on SC as a control medium and on a Trp− plate to assay silencing. Decreased growth on plates lacking tryptophan indicates improved silencing. (B) HSL7 deletion does not rescue silencing in an HMRa-e** mutant. Growth control and mating assay plates used the following strains: HMRa-e** (JRY2069), hsl7Δ HMRa-e** (LPY10974), and sas3Δ HMRa-e** (LPY2557). (C) Mating efficiency of the hsl7Δ strain is decreased compared to a wild-type strain. The quantitative mating assay was done with wild-type strains MATa (LPY5) and MATα (LPY79) and with hsl7Δ strains MATa (LPY8660) and MATα (LPY9548). (D) Deletion of HSL7 decreases silencing at the rDNA array. Cell growth of strains carrying the cassette ADE2-CAN1 inserted into the rDNA array of wild-type (LPY4909), esa1-L254P (LPY5055), and two different hsl7Δ mutants (LPY11013, LPY11014) was assayed by fivefold dilutions on Ade− Arg− plates as a control medium and on Ade− Arg− plates with 16 μg/ml of canavanine to assay silencing. Plates were grown at both 30° and 34°. Decreased growth on canavanine indicates impaired silencing.

In comparison to the deletion of the Rap1p-binding site, deletion of both the Rap1p- and the Abf1p-binding site results in complete loss of silencing at HMR, and mutations in some genes, such as SAS2 and SAS3, can suppress this defect (Reifsnyder et al. 1996). However, deletion of HSL7 did not suppress the silencing defect of this fully derepressed mutant silencer (Figure 5B), demonstrating that its role is distinct from the MYST HATs at HMR.

Impaired silencing at the HM loci results in mating deficiency. Thus, we asked whether deletion of HSL7, which improves silencing at HMR, could affect mating efficiency. Surprisingly, we found that MATa hsl7Δ and MATα hsl7Δ mutant strains had slight mating defects compared to a wild-type strain (Figure 5C). The gcn5Δ and rpd3Δ strains share this phenotype with hsl7Δ mutants. Indeed, gcn5Δ and rpd3Δ mutants had increased silencing at the hmrΔE silencer (Figure 5A) and yet are also slightly mating defective (Vidal and Gaber 1991 and data not shown).

Finally, to address whether Hsl7p regulates gene expression at the rDNA array, a strain that contains the marker ADE2-CAN1 inserted into the array was assayed for silencing (Fritze et al. 1997). CAN1 expression confers toxic dominant sensitivity to the arginine analog canavanine. As a consequence, strains defective for silencing at the rDNA have decreased growth on plates containing canavanine. As shown in Figure 5D, hsl7Δ strains grew more poorly than the wild-type strain on canavanine plates, indicating that they were partially defective for silencing at the rDNA array. This phenotype was even more pronounced when strains were grown at 34°, indicating that growth at elevated temperature increases hsl7Δ silencing defects at the rDNA array.

In conclusion, in hsl7Δ strains, silencing at the telomeres was not affected. However, silencing was perturbed at both the silent mating-type loci and the rDNA array. Indeed, hsl7Δ mutants showed improved silencing at HMR and decreased silencing at the rDNA array.

Levels of Arg17 dimethylation of histone H3 are not affected by HSL7 deletion:

In yeast, Hmt1p has been validated as a histone arginine methyltransferase and appears exclusively responsible for methylation of Arg3 of histone H4 (Lacoste et al. 2002). In humans, CARM1 and PRMT5 have been reported as histone methyltransferases, modifying respectively, Arg2, Arg17, and Arg26 of histone H3 (Schurter et al. 2001) and Arg3 of H4 and Arg8 of H3 (Pal et al. 2004). It is not established whether Arg2, Arg8, Arg17, and Arg26 of histone H3 residues are methylated in yeast cells.

The genetic interactions between HSL7 deletion and the histone H3 or H4 N-terminal tail deletions (Ma et al. 1996), as well as the new chromatin-related phenotypes of Hsl7p reported above, prompted us to ask if histones were in vivo targets of Hsl7p. We hypothesized that the synthetic lethal phenotype of gcn5Δ hsl7Δ double mutants could be due to the concomitant loss of Gcn5p acetyltransferase activity and the loss of Hsl7p methyltransferase activity toward an arginine residue of the histones. We constructed histone mutants in which arginine residues were mutated to alanine and we asked whether any of these arginines were essential in gcn5Δ strains. For this experiment, we constructed a strain deleted for GCN5 and for the two loci encoding histones H3 and H4 (HHT1-HHF1, HHT2-HHF2). The viablity of this mutant was maintained by a CEN URA3 plasmid containing the wild-type HHT2 and HHF2 genes. To assess the effect of the arginine mutations in HHT2 or HHF2 when GCN5 is deleted, the strain described above was transformed with CEN TRP1 plasmids carrying a variety of mutations in HHT2 or HHF2, and plasmid-shuffling experiments were carried out on 5-FOA plates. Among the mutations tested, we found that substitution of Arg17 of histone H3 to an alanine was specifically lethal in the absence of Gcn5p (Figure 6A and see discussion). This result suggested that Arg17 of H3 was the target of Hsl7p. To test this hypothesis, methylation levels of Arg17 of H3 were evaluated in hsl7Δ strains. Protein immunoblotting experiments were carried out on purified yeast histones using an antiserum directed against H3 dimethylated at Arg17 (Figure 6B). We found that Arg17 of histone H3 appeared, indeed, to be dimethylated in S. cerevisiae (lane 1). However, this methylation was intact on histones extracted from both hsl7Δ (lane 2) and hmt1Δ strains (lane 3), showing that neither Hsl7p nor Hmt1p are solely responsible for Arg17 dimethylation. Thus, although mutation of Arg17 of H3 to a residue that cannot be methylated mimics the hsl7Δ synthetic lethality with gcn5Δ, it appears not to be due to a change in its dimethylation status.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.—

Levels of dimethyl Arg17 of H3 are not affected by hsl7Δ. (A) Synthetic lethality between gcn5Δ and alteration of Arg17 of H3 to an alanine. Plasmid-shuffling experiments were the following: a gcn5Δ hta2-htb2Δ hht1-hhf1Δ hht2-hhf2Δ strain covered by a CEN URA3 plasmid encoding HHT2 and HHF2 (pJH33) was transformed either with a CEN TRP1 plasmid carrying HHT2 and HHF2 (pLP1775) or with the same plasmid carrying hht2-R17A and HHF2 (pLP1837) (yielding LPY9374 and LPY9377, respectively). Transformants were streaked onto a growth control plate of Ura− Trp− and onto a 5-FOA plasmid-shuffling plate. (B) Dimethylation of Arg17 of H3 is not affected in an hsl7Δ mutant. A Western blot experiment was carried out on purified histones extracted from wild-type (LPY5), hsl7Δ (LPY8660), hmt1Δ (LPY8669), and hht2-R17A (LPY9367) using an antibody that recognizes histone H3 dimethylated at Arg17.

Hsl7p conserved catalytic residues are dispensable for viability of gcn5Δ, rpd3Δ, and set1Δ mutants and for silencing:

Members of the arginine methyltransferase family, such as Hmt1p and PRMT5, have a role in RNA transcription and processing that is dependent on their methyltransferase activity. For example, Hmt1p, the major type I arginine methyltransferase in yeast, is recruited to genes during transcription and functions to modulate protein–protein interactions within a messenger ribonucleoparticle in an RNA-dependent manner (Yu et al. 2004). The type II arginine methyltransferase PRMT5, the human homolog of Hsl7p, is involved in splicing, regulating the assembly of the Sm core by methylation of the Sm proteins (Meister et al. 2001). However, it is not known whether Hsl7p has the same function in yeast. Indeed, Hsl7p has been mostly studied for its role in the control of the morphogenesis checkpoint, a role reported to be independent of its methyltransferase activity (Theesfeld et al. 2003; Yamada et al. 2004).

Crystal structures of several S-adenosylmethionine-dependent methyltransferases define four conserved domains, designated as I, post-I, II, and III. Domain I is known to be the site at which S-adenosylmethionine (donor of the methyl group) is bound to the protein. In studies of human PRMT5, mutation of an arginine codon (R368A) in the GxGRG motif of domain I abolished the methyltransferase activity of the enzyme in vitro (Pollack et al. 1999). Similarly in yeast, in vitro methyltransferase activity of recombinant Hsl7p is inactivated by alteration of the GxGRG motif: G386A, R387A, or G386A-R387A (Cid et al. 2001; Theesfeld et al. 2003).

We asked whether a plasmid-borne hsl7-G386A-R387A allele could complement the defects of an hsl7Δ strain. As shown in Figure 7A, the double point mutant restored a normal bud phenotype to hsl7Δ cells, indicating complementation of the morphogenetic phenotype. Then we performed plasmid-shuffling experiments to determine whether this mutant could suppress the synthetic lethal phenotypes uncovered between HSL7 deletions and inactivation of genes encoding chromatin-modifying enzymes or deletion of the N-terminal tail of histone H3. We found that the hsl7-G386A-R387A mutant supported viability of hsl7Δ gcn5Δ, hsl7Δ rpd3Δ, and hsl7Δ set1Δ double mutants (Figure 7B) and viability of an hsl7Δ strain in which the N-terminal tail of histone H3 was deleted (data not shown). Finally, the double point mutant hsl7-G386A-R387A significantly complemented the silencing phenotype of an hsl7Δ hmrΔE∷TRP1 strain (Figure 7C). Thus, it appears that conserved catalytic residues of the methyltransferase Hsl7p are dispensable for the role of Hsl7p in the morphogenesis checkpoint for viablity of the gcn5Δ, rpd3Δ, and set1Δ mutants or viability of the strains deleted for the N-terminal tail of histone H3, and for silencing.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.—

Hsl7p-conserved catalytic residues are dispensable in rescuing hsl7Δ chromatin-related phenotypes. (A) Expression of mutated HSL7 suppresses the elongated bud phenotype of hsl7Δ. Differential interference contrast microscopy of hsl7Δ (LPY8660) transformed with either HSL7-HA (pLP1947), hsl7-G386A-R387A-HA (pLP1968), or empty vector (pRS314). (B) Methyltransferase activity of Hsl7p appears dispensable in gcn5Δ, rpd3Δ and set1Δ strains. Plasmid-shuffling experiments on 5-FOA are shown. Double mutants hsl7Δ gcn5Δ, hsl7Δ rpd3Δ, and hsl7Δ set1Δ covered by a 2μ URA3 HSL7 plasmid (LPY8378, LPY9295, and LPY8380, respectively) were transformed with HSL7-HA (pLP1947), hsl7-G386A-R387A-HA (pLP1968), or empty vector (pRS314). Transformants were streaked onto a growth control plate of Ura− Trp− and onto a 5-FOA plasmid-shuffling plate. (C) Expression of the dead catalytic hsl7-G386A-R387A mutant rescues defects in silencing at the silent mating-type locus hmrΔE. Dilution assay with strains carrying the TRP1 reporter gene were inserted at hmrΔE; fivefold dilutions of wild type (LPY4912) + empty vector (pRS412), wild type (LPY4912) + HSL7 (pLP1969), hsl7Δ (LPY10243) + empty vector (pRS412), hsl7Δ (LPY10243) + HSL7 (pLP1969), hsl7Δ (LPY10243) + hsl7-G386A-R387A (pLP1970), and sir2Δ (LPY4980) + empty vector (pRS412) were plated onto a Ade− plate as a control medium and onto a Ade− Trp− plate to assay silencing.

Constitutive activation of the morphogenesis checkpoint is responsible for the genetic interactions between HSL7 and chromatin-modifying enzymes:

If loss of Hsl7p catalytic activity was not responsible for the new hsl7Δ chromatin-related phenotypes, it is possible that the phenotypes were due to an independent role of Hsl7p in the morphogenesis checkpoint. Other proteins, like Hsl1p, play an important role in this checkpoint. Indeed, both Hsl7p and Hsl1p are required for Swe1p-targeted degradation at the G₂/M transition. HSL1 encodes a member of the Nim1p protein kinase family (McMillan et al. 1999). Hsl1p is autophosphorylated and promotes the periodic phosphorylation of Hsl7p (Ma et al. 1996; McMillan et al. 1999; Cid et al. 2001). HSL1, like HSL7, was identified by Ma et al. (1996) as an essential gene when the N-terminal tail of H3 is deleted. As shown in Figure 8A, deletion of HSL1 was synthetically lethal with deletion of GCN5 or RPD3. Simultaneous disruption of HSL1 and SET1 led to lethality for 60% (28/46) of the presumed double mutants, whereas the other 40% were able to germinate but were extremely sick (Figure 8A, right, and data not shown). Moreover, we found that deleting HSL1 in an esa1-L254P mutant resulted in synthetic sickness (data not shown). These results suggest that constitutive activation of the morphogenesis checkpoint is responsible for the synthetic phenotypes uncovered between HSL7/HSL1 and GCN5, RPD3, SET1, and ESA1 inactivation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.—

Constitutive activation of the morphogenesis checkpoint is responsible for the genetic interactions between HSL7 and chromatin-modifying enzymes. (A) Synthetic lethality/sickness between HSL1 and GCN5, RPD3, or SET1 deletions. Dissection plates of hsl1Δ (LPY8936) crossed to gcn5Δ (LPY8279), rpd3Δ (LPY4889), or set1Δ (LPY11036). Double-mutant spores that grew poorly are underlined. Inferred double-mutant spores are circled. (B) SWE1 deletion fully rescues the synthetic lethalities. The following strains were streaked onto growth control Ura− and 5-FOA plasmid-shuffling plates for plasmid-shuffling experiments: hsl7Δ gcn5Δ + HSL7-URA3 (LPY8378), hsl7Δ gcn5Δ swe1Δ + HSL7-URA3 (LPY9571), hsl7Δ rpd3Δ + HSL7-URA3 (LPY9295), hsl7Δ rpd3Δ swe1Δ + HSL7-URA3 (LPY10431), hsl1Δ gcn5Δ + GCN5-URA3 (LPY9093), hsl1Δ gcn5Δ swe1Δ + GCN5-URA3 (LPY9095), hsl1Δ rpd3Δ + RPD3-URA3 (LPY10793), and hsl1Δ rpd3Δ + RPD3-URA3 (LPY10769). (C) SWE1 deletion rescues the synthetic sickness of the esa1-L254P hsl7Δ double mutant. Cell growth of wild type (LPY5), esa1-L254P (LPY5056), hsl7Δ (LPY8660), esa1-L254P hsl7Δ (LPY10587), swe1Δ (LPY9400), esa1-L254P swe1Δ (LPY10586), hsl7Δ swe1Δ (LPY8834), and esa1-L254P hsl7Δ swe1Δ (LPY10588) strains was assayed by plating fivefold dilution of cells on YPD at 30° and 34°.

It has been observed that deletion of HSL7 or HSL1 leads to an accumulation of Swe1p, a kinase that inhibits Cdc28p activity by phosphorylation of Cdc28-Tyr19 at the G₂/M transition (Ma et al. 1996; Shulewitz et al. 1999; McMillan et al. 2002). Deletion of SWE1 or mutation of Cdc28p into a nonphosphorylable form fully rescues the G₂/M delay and the morphogenetic defects of hsl7Δ and hsl1Δ strains (Ma et al. 1996). To test the hypothesis that the morphogenesis checkpoint was critical, we asked whether its inactivation rescued the phenotypes associated with HSL7 deletion. First, we found that deletion of SWE1 restored silencing at HMRΔE∷TRP1 and at the rDNA array in an hsl7Δ strain. Indeed, a mutant hsl7Δ swe1Δ HMRΔE∷TRP1 grew like a wild-type strain on medium lacking tryptophan and the double mutant hsl7Δ swe1Δ, containing the marker ADE2-CAN1 inserted into the array, was not sensitive to canavanine (Figure 8B). Thus, inactivation of the morphogenesis checkpoint completely restored silencing to wild-type levels at both the HMR and the rDNA array in an hsl7Δ strain.

To ask whether deletion of SWE1 rescued the viability of hsl7Δ gcn5Δ or hsl7Δ rpd3Δ strains, hsl7Δ swe1Δ double mutants were crossed either with gcn5Δ carrying a GCN5-URA3 plasmid or with rpd3Δ carrying a RPD3-URA3 plasmid. After sporulation and dissection, double mutants and triple mutants with covering plasmids were streaked onto 5-FOA. Although hsl7Δ gcn5Δ and hsl7Δ rpd3Δ double-mutant segregants were not viable, the triple mutants hsl7Δ gcn5Δ swe1Δ and hsl7Δ rpd3Δ swe1Δ grew with no covering plasmid. Thus, the SWE1 deletion completely rescued the synthetic lethalities observed between hsl7Δ and gcn5Δ or rpd3Δ strains. Similarly, the synthetic lethalities hsl1Δ and gcn5Δ or rpd3Δ were rescued by swe1Δ (Figure 8C). The double mutant hsl7Δ swe1Δ or hsl1Δ swe1Δ strains were also crossed with a set1Δ strain. Dissection of the resulting diploids revealed that hsl7Δ set1Δ swe1Δ and hsl1Δ set1Δ swe1Δ triple mutants grew as well as set1Δ mutants, demonstrating that deletion of SWE1 suppressed the synthetic lethality/sickness between hsl7Δ/hsl1Δ and set1Δ mutations (data not shown). We also found that deletion of SWE1 rescued the synthetic sickness of the double mutant hsl7Δ esa1-L254P (Figure 8D).

To determine whether the effects of the SWE1 deletion were specific to its role in the morphogenesis checkpoint, we utilized a cdc28 mutant strain in which Tyr19, the target of Swe1p, is mutated to a nonphosphorylable residue (cdc28-Y19F), resulting in inactivation of the morphogenesis checkpoint. Like swe1Δ, the cdc28-Y19F allele fully rescued the synthetic lethal or synthetic sickness phenotypes of hsl7Δ gcn5Δ, hsl7Δ rpd3Δ, hsl7Δ set1Δ, hsl1Δ gcn5Δ, hsl1Δ rpd3Δ, and hsl1Δ set1Δ double mutants (data not shown). Thus, inactivation of the morphogenesis checkpoint—either by deletion of the SWE1-encoded kinase responsible for Cdc28p phosphorylation or by mutation of Cdc28p into a nonphosphorylable form—suppresses the synthetic interactions between HSL7/HSL1 deletions and deletion of the chromatin-modifying enzymes. Together, these results show that the abrogation of the morphogenesis checkpoint suppresses both chromatin-related phenotypes of an hsl7Δ strain: the silencing defects and the genetic interactions with chromatin-modifying enzymes.

Deletion of SWE1 rescues the thermosensitivity of hsl7Δ, hsl1Δ, gcn5Δ, and rpd3Δ mutant strains:

As shown in Figure 9A, we found that constitutive activation of the morphogenesis checkpoint by hsl7Δ results in impaired cell growth at elevated temperatures (37°). However, abrogation of this checkpoint by swe1Δ rescues the thermosensibility phenotype of an hsl7Δ strain. The temperature-sensitive phenotype is shared by hsl1Δ strains and comparable rescue is observed in combination with deletion of SWE1 (supplemental Figure S1 at http://www.genetics.org/supplemental/).

• Download figure
• Open in new tab
• Download powerpoint

Figure 9.—

Suppression of hsl7Δ, gcn5Δ, and rpd3Δ temperature sensitivity by SWE1 deletion. (A) Cell growth of wild-type (LPY5), hsl7Δ (LPY8660), swe1Δ (LPY9400), and hsl7Δ swe1Δ (LPY8834) strains was assayed by plating fivefold dilutions on YPD at 30° and 37°. (B) Cell growth of wild-type (LPY5), gcn5Δ (LPY8155), gcn5Δ swe1Δ (LPY8653 and LPY8654), rpd3Δ (LPY4889), and rpd3Δ swe1Δ (LPY9667 and LPY9668) strains was assayed by plating fivefold dilutions on YPD at 30° and 37°.

The gcn5Δ and rpd3Δ strains also showed slow-growth phenotypes at 37° that were relieved in the swe1Δ background (Figure 9B). Thus, in addition to rescuing hsl7Δ phenotypes, inactivation of the morphogenesis checkpoint also alleviates gcn5Δ and rpd3Δ phenotypes, implying that there is a crosstalk between the chromatin-modifying enzymes Gcn5p and Rpd3p and the morphogenesis checkpoint even in the presence of Hsl7p.

Synthetic lethality between HSL7 deletion and deletion of the N-terminal tails of histone H3 or H4 may be due to the absence of post-translational modifications on the tails:

HSL7 was identified earlier as a gene essential in the absence of the N-terminal tail of histones H3 or H4 (Ma et al. 1996). Because the histone N-terminal tails may be multiply post-translationally modified by chromatin-modifying enzymes, we speculated that these synthetic lethalities result from the absence of post-translational modifications of these tails. Rather than initially searching for modifications that are essential in the absence of Hsl7p, we searched for the enzymes that mediate these modifications. We identified three: the acetyltransferase Gcn5p, the deacetylase Rpd3p, and the lysine-methyltransferase Set1p are all essential in an hsl7Δ strain. Likewise, deletion of HSL7 in a strain encoding a thermosensitive allele of the essential histone acetyltransferase Esa1p conferred synthetic sickness at semipermissive temperature. Importantly, these interactions are specific since deletion of genes encoding other acetyltransferases (Hpa2p, Hpa3p, Sas3p), deacetylases (Sir2p, Hda1p), and methyltransferases (Hmt1p, Dot1p) did not result in synthetic sickness or lethality in hsl7Δ mutants.

It will be of interest to determine whether alteration of the target histone residues of Gcn5p, Esa1p, Rpd3p, and Set1p to unmodifiable residues mimics the absence of these enzymes and results in cell death when combined with deletion of HSL7. Clearly, this search will be complex because the majority of the chromatin-modifying enzymes have multiple histone target residues. Consequently, it seems likely that combinations of multiple mutations will ultimately prove responsible for the phenotypes observed.

The identification of chromatin-modifying enzymes that target the N-terminal tail of H3 or H4 and are essential in the absence of Hsl7p strongly implies that particular histone tail modifications, rather than the tails alone, are required to overcome the activation of the morphogenetic checkpoint in an hsl7Δ strain.

A critical role for arginine 17 of histone H3 in gcn5Δ:

In the course of this work, we uncovered a genetic interaction that initially appeared to recapitulate the gcn5Δ hsl7Δ synthetic lethality, in that substitution of Arg17 of H3 to alanine was lethal when combined with deletion of GCN5. To test the hypothesis that Hsl7p methylates Arg17 of H3, we checked the levels of dimethylation of Arg17 of H3 in a hsl7Δ strain and observed that they were unchanged. This result indicates either that there is another arginine methyltransferase able to substitute for Hsl7p or that Arg17 of H3 is not an Hsl7p substrate. Consistent with the latter possibility, we were able to differentiate the two double mutants gcn5Δ hsl7Δ and gcn5Δ hht2-R17A regarding their checkpoint dependency. Indeed, the synthetic lethality between hsl7Δ and gcn5Δ can be rescued by inactivation of the morphogenesis checkpoint by SWE1 deletion. Thus, we assessed the viability of the gcn5Δ hht2-R17A double mutant when the morphogenesis checkpoint was inactivated by SWE1 deletion. If gcn5Δ hht2-R17A lethality was due to the absence of the methylation of Arg17 of H3 by Hsl7p, we would expect that gcn5Δ hht2-R17A swe1Δ triple mutants would be viable. However, deletion of SWE1 did not rescue the gcn5Δ hht2-R17A synthetic lethality (supplemental Figure S2 at http://www.genetics.org/supplemental/). Thus, the role of Arg17 of H3 in a gcn5Δ strain is critical, yet independent of the morphogenesis checkpoint.

The discovery of a crucial function for a single residue in H3 is particularly noteworthy. Earlier studies had defined combinations of multiple H3 and H4 tail mutations that are synthetically sick or lethal with gcn5Δ (see, for example, Zhang et al. 1998). Mutation of arginine 17 is distinct from these both because it is not an acetylatable site and because of its extreme consequences specifically in gcn5Δ mutants, but not in the other hsl7Δ-interacting mutations or in wild-type cells (data not shown). It is not yet known whether the lethal defects are a consequence of transcriptional errors or problems in other genomic processes, but defining the contribution of arginine 17 should prove of great interest.

hsl7Δ chromatin-related phenotypes are independent of Hsl7p methyltransferase activity:

The genetic interactions uncovered between the putative histone methyltransferase Hsl7p and the H3 or H4 amino termini raised the possibility that Hsl7p participates directly in accomplishing combinatorial histone modification patterns (Strahl and Allis 2000) by methylating arginine residues. Indeed, Hsl7p is a member of the PRMTs family and, until recently, was reported as a histone arginine methyltransferase whose targets were histones H2A and H4 in vitro (Pollack et al. 1999; Lee et al. 2000). However, these results could not be confirmed by an independent study (Cid et al. 2001). Moreover, during the course of our work, Miranda et al. (2006) demonstrated that Hsl7p can methylate H2A but not H4 and this methylation has been observed to date only in vitro, not in vivo.

In agreement with this, we found that the chromatin-related phenotypes that we uncovered for hsl7Δ mutants do not appear due to the loss of Hsl7p methyltransferase activity toward histones or other substrates. Indeed, all these phenotypes are complemented by hsl7-G386A-R387A, which encodes a protein reported to be without in vitro catalytic activity (Theesfeld et al. 2003; Yamada et al. 2004). Additionally, when we searched for histone mutants whose phenotype mimicked the loss of the putative histone arginine methyltransferase, none of the mutants constructed (hhf2-R3A, hhf2-R19A, hht2-R2A, hht2-R8A, hht2-R17A, and hht2-R26A) recapitulated hsl7Δ's morphological defects (data not shown). Moreover, we showed that the chromatin-related phenotypes of hsl7Δ are due to the constitutive activation of the morphogenesis checkpoint, an activation unrelated to the methyltransferase activity of Hsl7p (Theesfeld et al. 2003; Yamada et al. 2004).

HSL7/HSL1, GCN5, and RPD3 genetically interact with each other and have common phenotypes:

The hsl7Δ, gcn5Δ, and rpd3Δ mutants share similar phenotypes. Indeed, they have increased silencing at a hmrΔE mutant silencer (Figure 5A) and they are all slightly mating defective compared to a wild-type strain. As with HSL7 or HSL1 deletions, deletion of GCN5 and RPD3 genetically interacts with deletion of histone H3 or histone H4 N-terminal tails. Deletion of GCN5 has been reported to be synthetically lethal with the loss of either the H3 or the H4 amino terminus (Zhang et al. 1998) whereas deletion of RPD3 is lethal only with the loss of the H4 N-terminal tail (Sabet et al. 2004). In addition to the synthetic lethality observed between HSL7 or HSL1 and GCN5 or RPD3 deletions, we observed that double mutant gcn5Δ rpd3Δ is synthetic sick (Sun and Hampsey 1999 and supplemental Figure S3 at http://www.genetics.org/supplemental/), indicating that GCN5 and RPD3 also interact genetically. We also showed that deletion of SWE1, in addition to its effect on an hsl7Δ strain, is also able to rescue some defects of gcn5Δ or rpd3Δ strains. In conclusion, these results show that gcn5Δ, rpd3Δ, and the morphogenesis checkpoint are part of a previously unreported genetic network.

Our data establish that a specific subset of chromatin-modifying enzymes becomes essential when the morphogenesis checkpoint is constitutively activated. Three of these enzymes, Gcn5p, Rpd3p, and Set1p, are previously defined global transcriptional regulators. An attractive possibility is that the transcription networks that they define will ultimately point to regulatory circuits that are essential for bypassing Cdc28p inactivation when the morphogenesis checkpoint is constitutively activated.