Yaf9 and the YEATS domain protein family.
Our interest in Yaf9 was based on a previous identification of its paralog, Taf14, as a member of the SWI/SNF complex and also on its sequence similarity to proteins such as AF9 and ENL, which are involved in human leukemias. Searches using the Basic Local Alignment Search Tool (BLAST) algorithm showed that Yaf9 was significantly more similar to AF9 (BLAST E-value of 1e −16) (and also to ENL) than was Taf14 (BLAST E-value of 1e −9); therefore, we named the protein Yaf9 (yeast AF9). However, more recent searches have revealed a much larger family, including proteins in Schizosaccharomyces pombe and Caenorhabditis elegans and additional human proteins, such as Gas41 (BLAST E-value of 1e −29) (Fig. ). All members bear a defining N-terminal YEATS domain and therefore have been termed YEATS family members (Fig. ). However, close inspection of these protein sequences and analysis by BLAST suggest the presence of three additional sequence elements, which we termed the A, B, and C boxes, that we believe are useful for classifying the members (Fig. ). The A box is present in many members but is absent in Taf14 and Sas5. The B box defines the C terminus of Yaf9, Gas41, and related proteins and is predicted to form a coiled coil (Fig. and data not shown). Although the B box is absent in AF9 and ENL, these two proteins share homology at their C termini, and this region is also predicted to form a coiled coil. The C box is present only at the C termini of Taf14 and Sas5. Taken together, these results indicate that the family appears to be comprised of three sequence classes: a Yaf9/Gas41 class (A box, with a B box at the C terminus), an AF9/ENL class (A box, with a coiled coil at the C terminus), and a Sas5/Taf14 class (C box at the C terminus).
FIG. 1. Yaf9 localization and YEATS domain alignments. (A) Diagram of Yaf9 domain structure. Boxes represent the YEATS domain, the A box, and the B box. Asterisks indicate the positions of site-directed mutations for yaf9SDM; the arrow at position 186 indicates (more ...) The YEATS family is essential in S. cerevisiae.
To determine the importance of the YEATS family in yeast cells, we isolated strains bearing combinations of null mutations in the three family members (Yaf9, Sas5, and Taf14). Although no individual member was essential, strains lacking two of the three members showed reduced growth ability, and a strain lacking all three was nonviable (Table ). These results establish the YEATS family as being essential in yeast cells.
Yaf9 is a stable member of the NuA4 complex.
Initially, we determined that Yaf9 localizes to the nucleus (Fig. ), consistent with a role in transcription. To identify associated proteins, Yaf9 was tagged at the C terminus with one of two different tandem affinity tags. One consisted of a tandem HA3-His7 tag, and the other consisted of a tandem protein A-calmodulin-binding protein (TAP) tag (see Materials and Methods). DNA encoding each tag was targeted to the 3′ end of chromosomal YAF9 by homologous recombination. The tags did not affect function, as strains bearing these tags lacked the phenotypes displayed by the null allele (described below). For the strain with the HA3-His7 tag, the final anti-HA affinity column afforded nine proteins, which were identified by mass spectrometry to be Eaf1, Eaf2/God1, Eaf3, Eaf5, Eaf6, Epl1, Esa1, Act1, and Nbn1/Yng2, all members of the NuA4 complex.
To assess the stability of the association, we performed coimmunoprecipitation analyses with anti-HA beads and the eluate from the Resource Q column (Fig. ). NuA4 members could be depleted from the fraction with anti-HA beads (Fig. , lane 4) and recovered in the eluate (lane 6). Yaf9-NuA4 complexes resisted extensive washing with 400 mM NaCl, showing that Yaf9 is stably associated with NuA4.
FIG. 2. Yaf9 is a stable component of the NuA4 HAT complex. (A) Purification of HA-tagged Yaf9 yields the NuA4 complex. Extracts were chromatographed on SP Sepharose, DEAE, nitrilotriacetic acid-nickel, Resource Q, and anti-HA-Sepharose columns sequentially (see (more ...)
The alternative TAP tag procedure enabled the rapid preparation of a pure Yaf9-NuA4 complex (Fig. ), which was confirmed by mass spectrometric sequencing and immunoblot analysis (data not shown). TAP-Yaf9 itself stains poorly with silver and migrates at approximately the same mass as Yng2 (36 kDa). However, mass spectrometric analysis of NuA4 prepared by conventional chromatographic methods (18
) clearly identified peptides corresponding to Yaf9 (data not shown). Taken together, these data establish Yaf9 as a stable component of NuA4.
NuA4 bearing Yaf9 (purified by the TAP procedure) displayed robust acetyltransferase activity that was highly specific for histone H4 (Fig. ). However, activity was essentially equal to that displayed by NuA4 lacking Yaf9 (isolated from a yaf9Δ strain bearing Epl1-TAP) (data not shown), suggesting that Yaf9 is not required for HAT activity or for H4 specificity. In addition, SDS-PAGE and Western analysis revealed the presence of all other NuA4 subunits in purified preparations from yaf9Δ strains (data not shown), indicating that Yaf9 is not required for complex assembly. Finally, strains lacking YAF9 showed little or no reduction in their bulk levels of H4 acetylation (Fig. ), further suggesting that Yaf9 does not have a great impact on NuA4 H4 HAT activity or that its impact is limited to certain loci (such as telomeres; see below).
Clear evidence for the presence of Yaf9 in the SWR1 complex was recently demonstrated by others (27
). Although SWR1 complex members were not clearly detectable by SDS-PAGE analysis in our preparations, components of SWR1 could be detected by immunoblot analysis of preparations from Yaf9 immunoprecipitations performed at a low stringency, and the Ruv helicases could be detected by mass spectrometric analysis at low levels (data not shown). Consistent with stringency affecting the Yaf9-Swr1 association, preparations of the SWR1 complex purified at a low stringency by others contained only substoichiometric amounts of Yaf9, although its presence was definitive. In addition, the placement of a tag on Yaf9 could, in principle, have slightly compromised its association with the SWR1 complex selectively.
Yaf9 is required for proper DNA repair, DNA metabolism, and growth at low temperatures.
A strain lacking YAF9
was viable but was unable to grow at 15°C (Cs−
), on medium containing 0.03% methyl methanesulfonate (MMS; a DNA-damaging agent) or caffeine (5 to 15 mM; elicits a stress response), or after exposure to UV radiation (80 J/m2
) and grew very slowly on 50 mM hydroxyurea (HU; an inhibitor of deoxynucleoside triphosphate synthesis) (Fig. ). nua4
complex mutants share most of these phenotypes (MMS−
, caffeine negative, and HU−
), making it impossible to attribute these phenotypes in the yaf9Δ
strain to a function in one particular complex (Table ). However, we found that swr1Δ
strains were also Cs−
, while NuA4 mutants (esa1
, or arp4
) were not (data not shown), suggesting that Yaf9 promotes growth in the cold through its function exclusively through SWR1/Htz1. Furthermore, we found that yaf9Δ esa1-L327S
double mutants were nonviable (Table ), with the combination of these two mutations possibly reducing NuA4 function below a critical threshold. Taken together, these results are consistent with functions for Yaf9 in both the SWR1 complex and the NuA4 complex.
FIG. 3. Structure-function analysis of Yaf9. (A) Phenotypes conferred by yaf9Δ and complementation tests with plasmids bearing Yaf9 derivatives. Plasmids used were empty vector pM25, pFlag-YAF9, pFlag-yaf91-186, and pFlag-yaf9SDM. Strains used were WT (more ...) Functional requirements for the YEATS and C-terminal domains.
The Yaf9 C terminus contains a B box that is predicted to form a coiled coil. The function of this domain is not known, although leukemogenic MLL fusions to YEATS family members all contain their extreme C terminus, suggesting a crucial role for this domain. Deletion of the B box (Yaf91-186) conferred a null phenotype (Fig. ), although the derivative was produced well (Fig. ). Interestingly, full-length Yaf9 coprecipitated Esa1, whereas Yaf91-186 did not, showing that the Yaf9 C terminus is critical for assembly into NuA4 (Fig. ).
A Yaf9 derivative lacking the YEATS domain was unstable and almost undetectable in extracts by immunoblot analysis. Therefore, we prepared a Yaf9 derivative bearing site-directed mutations (SDM) that replaced three consecutive conserved residues in the YEATS domain (PPF, residues 80 to 82) with alanines (Fig. ). This derivative (Yaf9SDM) was stably produced (Fig. ) and behaved as a moderate hypomorph; it fully complemented yaf9Δ phenotypes related to caffeine and HU sensitivity, partially complemented phenotypes related to cold and UV, but only weakly complemented MMS sensitivity (Fig. ). To determine whether the YEATS domain is required for the functions of other YEATS proteins in yeast cells, we created an identical site-directed mutation in Taf14 (termed taf14SDM) and tested it for complementation of taf14Δ phenotypes. We found that the taf14Δ mutation conferred sensitivity to temperature (Ts−), caffeine (5 mM), HU (50 mM), UV (80 J/m2), and MMS (0.03%). Although expressed at WT levels (data not shown), the taf14SDM mutation only partially complemented UV and caffeine sensitivity and failed to complement Ts−, MMS−, and HU− phenotypes (Table ). Taken together, these results show that the YEATS domain is important for many of the functions of Yaf9 and Taf14, providing the first evidence for a function for the YEATS domain in vivo.
Strains lacking Yaf9 display reduced transcription of genes near certain telomeres and HMR.
To better understand the impact of Yaf9 on transcription, we compared the transcription profiles for yaf9Δ and WT strains. Overall, 224 genes are downregulated more than twofold, whereas 48 genes are upregulated more than twofold. These changes could largely be reversed by transforming the strain with a plasmid bearing WT YAF9 (data not shown). The majority of the changes observed did not fall into clear gene classes or pathways (the results obtained with the entire microarray are available on request). However, an examination of the changes relative to the physical chromosomal map revealed a striking position effect; a significant fraction of the genes (65 of 224) that were downregulated more than twofold (defined as “affected ” genes) in the yaf9Δ strain were positioned within 20 kb of the telomere (Fig. ).
FIG. 4. Genes proximal to telomeres are downregulated in yaf9Δ and esa1-L254P strains. (A) Numbers of genes downregulated in yaf9Δ (red) and esa1-L254P (green) strains or upregulated in yaf9Δ (blue) and esa1-L254P (black) strains (compared (more ...)
To depict this relationship, we first parsed the 40 kb of DNA proximal to telomeres into eight successive intervals of 5 kb and considered the remainder of the genome (distances of >40 kb) as the final interval (Fig. , abscissa). Affected genes were represented on the ordinate in terms of either total number (Fig. ) or the fraction of genes affected in the particular interval (Fig. ). Although the majority of the genes affected were not within 20 kb of the telomere (Fig. ), the density of the affected genes was exceptionally high within 20 kb of the telomere (~25%) compared to the bulk of the genome (~2%). Stated differently, although the telomere-proximal 20-kb region contains only ~5% of all genes, it contains ~29% of the affected genes. This overlap is highly significant; the χ2 value generated by 65 genes affected within 20 kb of a telomere is 184.9, corresponding to a P value of <0.001. Affected genes were not lost from the telomere due to shortening; genomic DNA isolated from yaf9Δ strains was labeled and hybridized to the array and revealed no loss of telomere-proximal genes (data not shown).
Extensive overlap of affected telomeres and genes in htz1Δ and yaf9Δ strains.
Interestingly, only a subset of telomeres is affected in yaf9Δ strains. We define affected telomeres as those bearing within 20 kb three or more genes that are downregulated more than twofold. Certain telomeres are highly homologous to telomeres on other chromosomes. For this subset, it is difficult to attribute the downregulation of a transcript observed to gene silencing at a particular gene or telomere, as the transcript (labeled cDNA) derived from one telomere will hybridize to the spot on the microarray representing both homologs. Therefore, we apply the additional criterion that the telomere must contain unique affected genes, which may lead to an underestimation of the number of telomeres affected. According to these stringent criteria, affected telomeres in yaf9Δ strains are 2R, 3R, 4L, 4R, 6L, 11L 13R, 14R, and 15L (see Table S1 in the supplemental data), with several other telomeres being moderately affected. Affected telomeres do not appear to share an obvious common element (such as Y′) or attribute that would distinguish them clearly from other telomeres.
Our results show a striking overlap with telomeres affected in strains lacking Htz1 (see below and Table S1 in the supplemental data) (34
). In keeping with the results and the focus of studies by others on Htz1 and Swr1, we chose to focus our detailed analysis on telomere 14R and the region between silent mating type locus HMR
and telomere 3R (Fig. ). At telomere 14R, the loss of Yaf9 causes a significant downregulation of proximal genes (Fig. ). Likewise, several genes in the region between telomere 3R and HMR
(YCR099C/100C/101C/104W) are affected (Fig. and data not shown). Consistent with a role for NuA4 in this process, a mutation in the catalytic subunit of NuA4 that reduces H4 acetylation, esa1-L254P
), confers significant downregulation of genes at 14R (Fig. and ), although it has only a slight impact at 3R (data not shown). The impact of esa1-L254P
on telomeres is not as extensive as that observed with the yaf9Δ
strain (see Table S1 in the supplemental data). However, we observed a significant overlap in the genes affected in esa1-L254P
strains, and this overlap was quite pronounced when we considered the subset of affected genes located near telomeres (Fig. ).
FIG. 5. Transcriptional profiles of sir3Δ yaf9Δ, sir3Δ yaf9Δ, and esa1-L254P at telomeres 14R and 3R. (A) Physical maps of loci near telomeres 14R (top) and 3R (bottom) with the chromosome (Chr) end at right (black arrowheads). (more ...)
In keeping with a role for Yaf9 in the SWR1 complex, about one-third of the genes downregulated in htz1Δ strains are identical to those downregulated in yaf9Δ strains (Fig. ). As with esa1-L254P, the overlap was even more extensive when we considered the subset of affected genes within 20 kb of the telomere; in that situation, the overlap approached 50%. Taken together, these results suggest that Yaf9 assists both SWR1/Htz1 and NuA4 complexes in telomere-proximal gene expression.
sas5Δ strains show extensive repression of telomere-proximal genes.
Strains lacking Sas2 show extensive downregulation of genes near certain telomeres, and their downregulation is correlated with a reduction in histone H4 acetylation (26
). However, the impact of the associated YEATS protein Sas5 on this process has not been determined. We found that a loss of Sas5 affects telomeres 1L, 2R, 3R, 4R, 5L, 6L, 7L, 9R,11L, 12R, 13R, 14R, 15L, and 16R (using the criteria established above), although many additional telomeres are moderately affected (see Table S1 in the supplemental data). The work of others on sas2Δ
focused on selected telomeres, primarily 6R, which is one of the two telomeres for which gene(s) are lacking on our array and which therefore could not be analyzed. These results extend the importance of the YEATS family in the process of gene expression or antisilencing at telomeres in vivo. These results also suggest that certain telomeres, such as 14R, rely on multiple complexes for antisilencing, whereas others depend primarily on the SAS complex.
Loss of SIR3 partially reverses yaf9Δ gene expression profiles and phenotypes.
Downregulation of genes at telomere-proximal loci might be a consequence of the spreading of Sir3 into these regions, a reduction in histone H4 acetylation in these regions, a reduction in Htz1 replacement in these regions, or a combination of these factors. To help discern among these possibilities, we determined whether downregulated telomere-proximal genes acquire Sir3 and the extent to which their downregulation depends on Sir3. Interestingly, the removal of Sir3 from strains lacking Yaf9 restores the transcription (either partially or fully) of certain genes (PAU6 and YNR074 for 14R; GIT1 and YCR099C/100C for 3R), whereas the transcription of other loci in these regions is not restored. We suggest that genes whose transcription is restored by the removal of Sir3 are those silenced by the Sir complex (in yaf9Δ strains), whereas those that remain silent in the yaf9Δ sir3Δ double mutant are deficient in histone acetylation and/or Htz1 replacement due to the absence of Yaf9 (see below).
As the loss of SIR3 partially restores the profiles of expression of telomere-proximal genes in yaf9Δ, the loss of SIR3 might suppress certain yaf9Δ phenotypes. Compared to yaf9Δ mutants, sir3Δ yaf9Δ double mutants show slightly improved growth at low temperatures and WT growth on medium containing caffeine (Fig. ), suggesting that these defects in yaf9Δ strains are related to telomere-proximal gene repression caused by the spreading of Sir proteins. However, no suppression of phenotypes related to DNA repair and metabolism was observed, suggesting that these defects are not restricted to functions at telomeres and are not related to Sir protein spreading. These results further suggest that Yaf9, as a member of either the SWR1 complex or the NuA4 complex, plays roles both at telomeres and at other loci.
Silencing of telomere-proximal genes in yaf9Δ strains correlates with occupancy by Sir3.
To directly determine the extent of Sir3 spreading in yaf9Δ strains, we performed ChIP of Sir3 in WT and yaf9Δ strains, each bearing a SIR3 allele encoding a Myc13-tagged Sir3 derivative integrated at the SIR3 locus. The levels of Myc13-tagged Sir3 protein were virtually identical in WT and yaf9Δ strains (Fig. ). We then assessed the enrichment of particular loci by two methods: (i) genome-wide occupancy determinations involving ChIP combined with an immobilized array of the entire yeast genome parsed into open reading frame and intergenic fragments (G-ChIP) and (ii) qPCR. For telomere14R, telomere-proximal PAU6 is highly occupied by Sir3 in both WT and yaf9Δ strains, as indicated by both G-ChIP (Fig. ) and qPCR (Fig. ). Note that Fig. depicts the relative enrichment of Sir3 in the yaf9Δ strain compared to the WT strain, whereas the intensity of Cy5 (red) in the G-ChIP assay (Fig. ) reflects the absolute levels of Sir3. In the yaf9Δ strain, a redistribution of Sir3 is clearly observed at telomere 14R, consistent with spreading from the telomere end toward internal genes (Fig. ). Likewise, genes interposed between HMR and telomere 3R display increased Sir3 occupancy. Also, a reduction in occupancy at HMR itself was revealed by qPCR, suggesting that the spreading of Sir3 to the intervening region may come from (or at the expense of) HMR; however, a very significant amount of Sir3 remains (Fig. and data not shown).
FIG. 6. Spreading of Sir3 in yaf9Δ strains. (A) Sir3-Myc levels are identical in WT and yaf9Δ strains, as shown by immunoblot analysis of whole-cell extracts derived from WT and yaf9Δ strains. (B) Sir3-Myc occupancy at telomere 14R in (more ...) Loss of Yaf9 correlates with loss of H4 hyperacetylation at telomeres 14R and 3R.
To test whether downregulation near telomeres is correlated with reductions in H4 acetylation, we performed ChIP analyses with an antibody raised against the hyperacetylated (at K5, K8, K12, and K16) tail of histone H4. For WT cells, we observed high levels of H4 acetylation at particular loci in both regions. Interestingly, these loci are either near or at the boundary of Sir3 spreading in WT cells (COS10
for 14R and GIT1
for 3R) (Fig. ). However, RDS1
is both occupied by Sir3 and significantly acetylated, suggesting that H4 acetylation may not entirely prevent Sir3 binding or that the region is dynamic with respect to acetylation and deacetylation or Sir binding. For yaf9Δ
cells, we observed a significant reduction in the acetylation of these loci but not of several other control loci. However, as similar reductions at these loci have been observed in strains lacking Htz1 (34
), we cannot directly attribute the entire effect to Yaf9 assisting acetylation by NuA4 (see Discussion).
FIG. 7. Yaf9 is important for histone H4 acetylation and is essential for Htz1 deposition at telomeres 14R and 3R. Relative acetylated H4 (Ac-H4) enrichment (A) and Htz1 enrichment (B) for WT (red) and yaf9Δ (blue) strains are shown. Enrichment was quantified (more ...) Htz1 occupancy at telomeres is reduced significantly in yaf9Δ strains.
Our genetic and genomic results all suggest a close functional connection between Yaf9 and Htz1, raising the possibility that Yaf9 may assist the SWR1 complex in the deposition of Htz1 in vivo. To test this possibility, we used ChIP and qPCR to assess the occupancy of Htz1 in WT and yaf9Δ strains. To this end, we deleted YAF9 from a strain bearing an HTZ1 allele encoding an HA3-tagged version of Htz1 integrated at its genomic locus. Initially, we performed ChIP for both strains and quantified the enrichment of loci at telomeres 14R and 3R by qPCR. As observed with our H4 ChIP analyses, only certain loci are occupied by Htz1. Interestingly, although the sample size is small, there appears to be a significant correlation between loci showing H4 hyperacetylation and loci showing Htz1 occupancy, raising the possibility that these two functions are linked. Remarkably, Htz1 occupancy is significantly reduced at the telomeric loci tested in the yaf9Δ strain (Fig. ). Interestingly, the reduction in Htz1 occupancy in the yaf9Δ strain may not be restricted to telomeres, as Htz1 occupancy is also reduced at GAL7.
To determine whether deletion of YAF9 affected the assembly of the SWR1 complex, we performed gel filtration analyses with WT and yaf9Δ strains bearing a Myc-tagged Swr1 derivative. For both strains, the SWR1 complex eluted at a molecular mass of 700 to 850 kDa (data not shown), suggesting that the complex remains intact in the yaf9Δ strain.