Protein Interactions within the Set1 Complex and Their Roles in the Regulation of Histone 3 Lysine 4 Methylation*

Set1 is the catalytic subunit and the central component of the evolutionarily conserved Set1 complex (Set1C) that methylates histone 3 lysine 4 (H3K4). Here we have determined protein/protein interactions within the complex and related the substructure to function. The loss of individual Set1C subunits differentially affects Set1 stability, complex integrity, global H3K4 methylation, and distribution of H3K4 methylation along active genes. The complex requires Set1, Swd1, and Swd3 for integrity, and Set1 amount is greatly reduced in the absence of the Swd1-Swd3 heterodimer. Bre2 and Sdc1 also form a heteromeric subunit, which requires the SET domain for interaction with the complex, and Sdc1 strongly interacts with itself. Inactivation of either Bre2 or Sdc1 has very similar effects. Neither is required for complex integrity, and their removal results in an increase of H3K4 mono- and dimethylation and a severe decrease of trimethylation at the 5′ end of active coding regions but a decrease of H3K4 dimethylation at the 3′ end of coding regions. Cells lacking Spp1 have a reduced amount of Set1 and retain a fraction of trimethylated H3K4, whereas cells lacking Shg1 show slightly elevated levels of both di- and trimethylation. Set1C associates with both serine 5- and serine 2-phosphorylated forms of polymerase II, indicating that the association persists to the 3′ end of transcribed genes. Taken together, our results suggest that Set1C subunits stimulate Set1 catalytic activity all along active genes.

Set1 is the catalytic subunit and the central component of the evolutionarily conserved Set1 complex (Set1C) that methylates histone 3 lysine 4 (H3K4). Here we have determined protein/protein interactions within the complex and related the substructure to function. The loss of individual Set1C subunits differentially affects Set1 stability, complex integrity, global H3K4 methylation, and distribution of H3K4 methylation along active genes. The complex requires Set1, Swd1, and Swd3 for integrity, and Set1 amount is greatly reduced in the absence of the Swd1-Swd3 heterodimer. Bre2 and Sdc1 also form a heteromeric subunit, which requires the SET domain for interaction with the complex, and Sdc1 strongly interacts with itself. Inactivation of either Bre2 or Sdc1 has very similar effects. Neither is required for complex integrity, and their removal results in an increase of H3K4 mono-and dimethylation and a severe decrease of trimethylation at the 5 end of active coding regions but a decrease of H3K4 dimethylation at the 3 end of coding regions. Cells lacking Spp1 have a reduced amount of Set1 and retain a fraction of trimethylated H3K4, whereas cells lacking Shg1 show slightly elevated levels of both di-and trimethylation. Set1C associates with both serine 5-and serine 2-phosphorylated forms of polymerase II, indicating that the association persists to the 3 end of transcribed genes. Taken together, our results suggest that Set1C subunits stimulate Set1 catalytic activity all along active genes.
Paf1 and Rtf1, components of the Paf1 complex, have also been shown to be required for the Rad6/Bre1-mediated monoubiquitination of H2B on lysine 123 (29,30). Through a so-called "trans-tail" process, di-and trimethylation of H3K4 depends on H2B ubiquitination (31,32), and components of the Paf1 complex (Paf1, Rtf1, and Ctr9) are required for di-and trimethylation of H3K4 (24). However, although Rtf1 also appears important for monomethylation of H3K4, neither Paf1 nor ubiquitination of histone H2B are essential for H3K4 monomethylation (27,33). Interestingly, a screen meant to identify components that selectively regulate H3K4 trimethylation identified the Bur cyclin-dependent kinase whose inactivation impairs the recruitment of the Paf elongation complex to genes and H2B ubiquitination (34,35), suggesting that low levels of ubiquitination are sufficient to promote mono-and dimethylation of H3K4. In this study, we have investigated protein/protein interactions within Set1C and determined their role in the integrity of the complex and the regulation of H3K4 methylation.

EXPERIMENTAL PROCEDURES
The Two-hybrid System in Escherichia coli-Protein interactions have been studied by a bacterial two-hybrid approach based on the functional reconstitution of adenylate cyclase activity of Bordetella pertussis in the E. coli cya strain (36). Pairs of proteins to be tested were fused to the two catalytic domains T18 and T25 of adenylate cyclase using plasmids pKT25link and pUT18link and their derivatives (36,37). After co-transformation of the BTH101 strain by the two plasmids expressing the fusions, 3 ml of LB medium supplemented with ampicillin and kanamycin and 0.5 mM isopropyl 1-thio-␤-D-galactopyranoside were inoculated with ten randomly chosen transformants and grown at 30°C to an A 600nm of 1. Thereafter, 2-l drops were spotted on McConkey plates supplemented with 1% maltose and on X-gal plates. The plates were incubated 3 days at 30°C. An interaction is considered positive if detection of ␤-galactosidase activity is obtained in Ͼ3 independent transformations.
Epitope Tagging of Set1, Mutant Strains, and TAP Purification-Set1 was tagged at its N terminus with a 9-Myc epitope as follows. A NotI restriction site was introduced just after the ATG codon of SET1 into a genomic fragment extending from Ϫ497 ϩ1179 bp in respect to the ATG of SET1. This DNA fragment was cloned into the integrative plasmids Yiplac128 (LEU2) and Yiplac204 (TRP1). A 9-Myc epitope was introduced in-frame into NotI. The resulting plasmids (p9MycSET1-LEU2 or p9MycSET1-TRP1) were digested with SnaBI prior to transformation of yeast cells. The tagged 9-Myc-Set1 was fully functional for all of the set1 phenotypes that were tested. Full deletions of set1, swd1, swd3, bre2, sdc1, spp1, and shg1 were in the BY4741 background (Euroscarf). All experiments were done in BY4741, except TAP tag introduction and TAP purification, which were performed in strain MGD-353-13D according to Roguev et al. (2).
Preparation of Histones, Microsequencing Analysis, and Western Blot Analysis-Histones were isolated by a procedure described in Ref. 38. Histone preparations were subjected to 16% SDS-PAGE and either Coomassie Blue-stained or transferred onto a 0.2-m Protran nitrocellulose membrane. For Western blot analysis, the antibodies used to detect di-and trimethylated H3K4 were anti-H3K4me2 and -H3K4me3 (Abcam). N-terminal sequence determination was performed by stepwise Edman degradation using an automatic sequencer model Procise 494 from Applied Biosystems. Histone H3 bands from blotted protein on 0.2-m polyvinylidene difluoride, stained with Coomassie blue, were excised and put on a Problott cartridge (Applied Biosystems). For all samples, a pulse liquid polyvinylidene difluoride protein program was performed, and phenylthiohydantoin amino acids were identified by high pressure liquid chromatography using a Brownlee C18 column (5 m, 220 ϫ 2.1 mm) with an 18-min gradient of B (acetonitrile/ isopropyl alcohol, v/v) in A (3.5% H 4 folate/water) from 16 -46% B at 55°C and with the wavelength set at 269 nm. A solution of 20 phenylthiohydantoin amino acids standard (10 pmol) was used for calibrating every run.
Simultaneous Set1 and Rap1 detection in yeast whole-cell extracts was carried as follows. Whole-cell extracts were prepared by trichloroacetic acid precipitation. Identical amounts of whole-cell extracts were resolved by SDS-PAGE, transferred to nitrocellulose membrane, and blocked in 5% milk/phosphate-buffered saline. Anti-Set1 (provided by Dr. Peter Nagy, Molecular Oncology Testing, Medical Research Center, Iowa City, IA) and anti-Rap1 (provided by Dr. David Shore, University of Genève, Geneva, Switzerland) antibodies (1:1000 dilution) were incubated for 3 h in 5% milk/phosphate-buffered saline. Infrared-labeled secondary antibodies IRDye 800 antirabbit goat IgG (Rockland) and Alexa Fluor 680 anti-mouse goat IgG (Molecular Probes) were used at a dilution of 1:10000. Emitted fluorescence of the secondary antibodies was detected by scanning on an Odyssey Imager (LI-COR Bioscience). Antibodies BWG16, H14, and H3 used to detect the different forms of pol II CTD in Western blot were purchased from Covance.
Chromatin Immunoprecipitation (ChIP) Experiments-ChIPs were done using 50-ml cultures fixed with 1% formaldehyde for 15 min (27). Yeast cells were broken using glass beads, and fixed chromatin was broken by sonication using a Bioruptor (Diagenode). Average DNA fragment lengths were 150 bp. 360-l aliquots of chromatin were incubated with the indicated antibodies in 1.5 ml of siliconized eppendorfs at 4°C for 15-20 h and immunoprecipitated with protein G-Sepharose beads (Dynabeads) for 90 min at 4°C. After washing, the chromatin was eluted from the beads at 65°C for 30 min. Cross-links were reversed by incubation at 65°C for 20 h. DNA was purified using Invitrogen PCR minicolumns (Pure-Link) and eluted in 200 l of water. Samples were subject to real-time PCR using a light cycler (Roche Applied Science) and Sybr Green mix. ChIPs were expressed as a ratio immunoprecipitation:input (IP:input). ChIPs against the methylated states of H3K4 are normalized to the IP H3 :input H3 value. Immunoprecipitation of DNA cross-linked to the Myc-Set1 was performed with 9E10 (anti-Myc)-agarose-conjugated monoclonal antibodies (Santa Cruz Biotechnology). ChIP experiments used to test the presence of methylated H3K4 were performed with anti-H3K4me (Upstate Biotechnology), anti-H3K4me2 and -H3K4me3 (Abcam).

RESULTS
Set1C Composition and Roles of the Subunits-To analyze direct interactions within Set1C and interactions with Set1, we used a bacterial two-hybrid system based on the interactionmediated reconstruction of adenylate cyclase (36). The use of a bacterial two-hybrid system avoids the detection of indirect interactions between yeast proteins mediated by endogenous proteins acting as bridging molecules. In this assay, the proteins of interest were fused to two fragments (T25 and T18) of the catalytic domain of B. pertussis and co-expressed in an E. coli strain deficient for adenylate cyclase. Interaction of the two proteins resulted in a functional complementation between T25 and T18, leading to cAMP synthesis and in turn to transcriptional activation of catabolic operons, such as the lactose operon. The N terminus of Set1 (Set1-(1-283)), the double RNA recognition motif (RRM1-RRM2, Set1-(234 -580)) (39), and a large C-terminal domain (Set1-(580 -1080)), as well as Bre2, Sdc1, Spp1, and Swd3 were cloned either into the T25 or T18 expression plasmids (or in both) to generate recombinant plasmids expressing hybrid proteins in which the polypeptides were fused at the C terminus of either the T25 or T18 fragment of the B. pertussis adenylate cyclase. The efficiencies of functional complementation were detected on McConkey and X-gal plates. In this assay, Bre2 and Sdc1 interact with each other, and Sdc1 also has the ability to interact with itself ( Fig.  1A). Furthermore, the large C-terminal Set1-(580 -1080) fragment interacts with Bre2 and Spp1.
We examined the effect of each Set1C member on Set1 stability. Set1 protein levels were compared with Rap1 (as an internal control) by Western blots in strains lacking each of the nonessential Set1C subunits. We used anti-Set1 mouse monoclonal antibodies (14) and anti-Rap1 rabbit polyclonal antibodies and two different secondary antibodies allowing the simultaneous direct infrared fluorescence detection of the Rap1 and Set1 antibodies (Fig. 1B). The amount of Set1 was similar to wild type in sdc1, bre2, and shg1 mutants. In swd1 and spp1 mutants, the amount of Set1 was reduced by 6-and 2-fold, respectively. Similar results were obtained in an experiment where TAP-tagged Set1 levels were evaluated by the detection of the TAP tag. In this experiment, deletion of either Swd1 or Swd3 reduced TAP-Set1 levels by Ͼ10-fold (data not shown). The expression of Set1 was also tested in several swd2 ts mutants (19). In all of the tested swd2 mutants, Set1 levels were reduced by Ͼ5-fold (not shown). We conclude that the Swd1-Swd3 heterodimer and Swd2 are required for Set1 stability and consequently for Set1C integrity.
We next analyzed the protein complexes associated with TAP-tagged Set1 by mass spectrometry. Because the absence of Swd1 or Swd3 reduces Set1 stability, these experiments were only performed with strains lacking Bre2, Sdc1, Spp1 and Shg1 (Fig. 1C). In accordance with the existence of a Bre2-Sdc1 heteromer, the deletion of either abolishes the association of the other. However, the rest of the complex remained. Therefore, no other subunit requires Bre2-Sdc1 for association with Set1C. Furthermore, the complex does not require Bre2-Sdc1 for integrity. The absence of either of the two other deletable sub-units, Spp1 or Shg1, did not destroy the complex, and all of the other subunits were still associated with Set1 (Fig. 1C). These results are in agreement with previous results showing that, in the absence of Set1, two heteromeric remnants of the complex, Swd1-Swd3 and Bre2-Sdc1, were found (2).
Distinct Roles of Set1C Subunits in H3K4 Methylation-Lysines can be mono-, di-, or trimethylated. Previously, we developed a method to quantify the unmethylated and monomethylated forms of H3K4 on histones purified from S. cerevisiae by N-terminal sequencing of H3 (27). This method determines the ratio between the unmethylated and monomethylated H3K4 lysine by measuring the area under the K and Kme peaks at cycle 4 ( Fig. 2A). Consequently, the amounts of di-and trimethylated H3K4 can be deduced (Table 1). Histones were purified from mutant strains disrupted for each one of the seven nonessential members of the Set1 complex. No peaks corresponding to monomethylation of H3K4 in the swd1 and swd3 mutants were detected. The bre2 and sdc1 mutants exhibited ϳ50% of the wild-type monomethylation level, and the levels of di-and trimethylated H3K4 were reduced by ϳ5-fold. The spp1 mutant did not affect monomethylation, and the di-and trimethylated H3K4 amount was reduced by ϳ3-fold. Unexpectedly, the shg1 mutation provoked a small but reproducible increase of the di-and trimethylation H3K4 level and a slight reduction of the amount of unmethylated H3K4. Because N-terminal sequencing could not distinguish between di-and trimethylation, we analyzed these methylations by Western blot with H3K4me2 and H3K4me3 antibodies (Fig. 2B). For precision, these analyses were performed with purified histones (Fig.  2B). They agreed with and extended previous studies with whole-cell extracts (4,14,15,16). Trimethylation was severely affected (but not abolished) in the bre2 and sdc1 mutants, whereas dimethylation was reduced by ϳ4-fold. In the spp1 mutant, dimethylation was not affected; however, trimethylation was reduced by ϳ3-fold.  Table 1. wt, wild type. B, purified histones from the indicated strains were transferred onto nitrocellulose and stained with Ponceau red (top). The histones were analyzed on Western blots with the indicated antibodies for the presence of the dimethylated H3K4 (H3K4me2) and trimethylated H3K4 (bottom, H3K4me3). Therefore, Set1C subunits can be grouped into four classes. 1) Swd1 and Swd3 form a heterodimer that is essential, together with Swd2, for Set1 stability and integrity of the complex and therefore for any H3K4 methylation. 2) Sdc1 and Bre2 form a heteromer that is required for trimethylation of H3K4 and efficient dimethylation but not for Set1C integrity. 3) Spp1 stabilizes Set1 and promotes efficient H3K4 trimethylation. 4) Shg1 seems to slightly inhibit H3K4 di-and trimethylation.  strains (sdc1, swd1, spp1, and shg1). The expression of Myc-Set1 was checked as in Fig. 2. B, ChIP of Myc-Set1 (9E10, Santa Cruz Biotechnology) at positions on CBF5 corresponding to either early 5Ј-coding regions (blue) and 3Ј regions of the coding sequence (green). ChIPs were performed from cells grown in YPD medium (yeast extract, peptone, dextrose). The specificity of the Myc-Set1 ChIP was controlled by performing the ChIP experiment in the untagged Set1 strain (no tag). Quantitative PCR analyses were performed by real-time PCR using a Light Cycler (Roche Applied Science). The values (IP:input) obtained for the binding of the tagged Myc-Set1 were normalized to the values obtained when the ChIP was carried out with the same antibodies in the untagged Set1 strain. This value thus reflects the Myc-Set1-specific binding. C, ChIP experiments were performed with anti-H3K4me1 (Upstate Biotechnology), anti-H3K4me2, -H3K4me3, and -H3Ct (Abcam) antibodies at the same positions in CBF5 as in Fig. 4B. The amounts of methylated H3K4 were normalized to the level of H3. In these experiments, the ChIPs against H3 and the methylated states of H3K4 were performed on the same input; therefore, the values represented correspond to IP:IP H3 . Errors bars were calculated from three independent ChIP experiments. wt, wild type.

Loss of Individual Set1C Subunits Differentially Affects the Distribution of H3K4 Methylation at the 5Ј and 3Ј Coding
Regions of Active Genes-We looked for the effects caused by the loss of individual Set1C subunits on the distribution of Set1 protein as well as H3K4 methylation along the coding region of two active genes, RPL17A and CBF5. Results for both were similar; therefore, only CBF5 is presented (Fig. 3). To locate Set1, it was Myc-tagged at its N terminus. The tag did not interfere with the ability of Set1 to methylate H3K4 (27). We first tested the expression of the Myc-tagged Set1 in swd1, sdc1, and spp1 cells. As expected, the amount of the Myc-tagged Set1 in swd1 was compromised (Fig. 3A), and occupancy on CBF5 was diminished (Fig. 3B). The presence of Myc-Set1 at the 5Ј end of the coding region was not affected in sdc1 or spp1 mutant cells but was slightly increased at the 3Ј end of CBF5. These results indicate that Spp1 and Sdc1 were not required for the association of Set1 with the coding region of CBF5 (and RPL17A).
Next, we looked at mono-, di-, and trimethylation of H3K4 at the 5Ј and 3Ј ends of CBF5 in sdc1, bre2, spp1, and shg1 cells (Fig. 3C). A set1 mutant was used to control the specificity of the ChIP. The amounts of methylated H3K4 are normalized to the level of H3. As can be seen, loss of Bre2 or Sdc1 had the same effect, namely an increase of monomethylation and a loss of trimethylation at the 5Ј end and a loss of dimethylation at the 3Ј end. Loss of trimethylation at the 5Ј end was the largest effect by far. Loss of Spp1 produced a similar profile; however, both the losses of 5Ј or 3Ј end dimethylation were less pronounced. Loss of Shg1 produced a different profile. Only moderate changes in mono-and trimethylation at either the 5Ј or 3Ј ends were observed. However, dimethylation was increased at the 5Ј end and decreased at the 3Ј end.
These results are in reasonable agreement with the global analyses (Fig. 2). The Bre2-Sdc1 heteromer appears to be required for 5Ј and 3Ј end dimethylation. In its absence, 5Ј end monomethylation increases, reflecting the possibility that the transition of monomethylation to trimethylation at 5Ј ends depends on Bre2-Sdc1. Potentially, the loss of 5Ј end trimethylation precedes the loss of 3Ј end dimethylation. Some support for this proposition can be found in the similar effects caused by loss of Spp1, where losses of 5Ј end trimethylation and 3Ј end dimethylation are both milder.
Set1C Associates with Both Phosphorylated Forms of pol II-The Paf1 complex recruits Set1C to chromatin (probably because of a physical interaction between the two complexes and pol II phosphorylated at Ser 5 of its CTD) (23,24). The Paf1 complex associates with elongating pol II over the entire mRNA-coding region (40), and several reports show that Set1 and dimethylation can be found on the middle and end of actively transcribed genes (25)(26)(27). However CTD Ser 5 phosphorylation is limited to the initiating pol II, and the elongating pol II found toward the 3Ј end of genes is characterized by CTD Ser 2 phosphorylation (41). We therefore examined which pol II CTD-phosphorylated forms co-purify with Set1C (Fig. 4). Although Set1C was poorly associated with nonphosphorylated pol II (Fig. 4, upper panel; compare the signals in cell extracts with the ones of TAP-purified proteins), we found that Set1C associates with pol II phosphorylated at both Ser 5 and Ser 2 (Fig. 4). Although a significant fraction of Set1C may be released from pol II after the loss of Ser 5 phosphorylation, this result suggests that Set1C can also associate with pol II, which is CTD-phosphorylated on Ser 2 .

DISCUSSION
The subunit composition of the Set1 complex has been comprehensively characterized in both budding and fission yeasts (this study and in Refs. 2, 3, 13-15, 21, and 22), and recently the human counterpart has been shown to be nearly identical (16,17). These data indicate that Set1C is strongly conserved in eukaryotes, as suggested in Ref. 2. Although it has been purified several times from different laboratories and sources, the size of the complex has not been established yet. By sizing column or sucrose gradient centrifugation, the complex has been twice FIGURE 4. Set1 interacts with both forms of phosphorylated pol II. Shg1-TAP, Bre2-TAP, and Rpb3-TAP were affinity-purified with their associated proteins. Whole-cell extracts (left) and affinity-purified proteins (right) of the indicated strains were analyzed by Western blot, with the indicated antibodies (BWG16, H14, and H5) recognizing nonphosphorylated pol II, pol II CTD phosphorylated at Ser 5 , and pol II CTD phosphorylated at Ser 2 , respectively. The Rbp3-TAP lanes are from the same blot; however, the exposure was for 10ϫ less time. NOVEMBER 17, 2006 • VOLUME 281 • NUMBER 46 estimated to be ϳ1 MDa and twice ϳ500 kDa (2,13,14,16). Because the smallest possible size for the 8-subunit yeast complex is 360 kDa (that is, only 1 molecule of each subunit/complex), the size estimates permit three different propositions. Is the complex (i) monomeric with one active site, (ii) dimeric, as suggested (15), with two active sites, or (iii) agglomerative?

Regulation of Histone H3 Lysine 4 Methylation
Previously, it was shown that the complex is based on a scaffold of Set1 (2). Here, we extended the scaffold to include the Swd1-Swd3 heterodimer and probably Swd2. Unexpectedly, we report that the Swd1-Swd3 heterodimer does not need the SET domain for stable interaction with Set1C. This was unexpected, because the mammalian homologues of these proteins (Rbbp5 and WDR5, respectively) have been found in partially characterized complexes with the mammalian H3K4 methyltransferases Mll and Mll2 (42)(43)(44) as well as with the mammalian homologues of Set1 (16 -17). The only obvious region of homology shared by Mll and Set1 is the SET domain, and even their preSET domains are divergent (2). Therefore, logical simplicity would infer that the highly conserved Swd1-Swd3 (Rbbp5-WDR5) heterodimer and its highly conserved interaction with H3K4 methyltransferases should interact with the common element, the highly conserved SET domain. However, stable interaction between S. cerevisiae Set1C and the Swd1-Swd3 heterodimer is not reliant on the Set1 SET domain. Set1 must therefore have, outside of the SET domain, a conserved structural motif required for the interaction with the heterodimer Swd1-Swd3.
Recently, it was discovered that WDR5 binds to dimethylated H3K4 (45). Hence, all characterized complexes conveying H3K4 methylation appear to include a subunit that binds the methylated site. Furthermore, we show that this subunit in yeast is essential for complex integrity. Because the Swd1-Swd3 heterodimer is essential to complex integrity and Set1 protein expression, the total loss of H3K4 methylation in their absence is explained.
By several criteria, including a bacterial two-hybrid system, we conclude that Sdc1 and Bre2 interact with each other and Sdc1 interacts with itself. In most published studies, both mutants bre2 and sdc1 share similar phenotypes (2,4,14,22). Furthermore, in the two-hybrid system, Bre2 interacts with Set1-(580 -1080), and TAP purification indicates that Bre2 and Sdc1 do not associate with a Set1C that lacks the SET domain. These results suggest that Bre2 interacts directly with Set1 and that this interaction requires the SET domain (Set1-(900-1048)).
We conclude that the Bre2-Sdc1 heteromer associates onto the Set1/Swd1/Swd3 scaffold. This association requires the SET domain. Again, the evidence from partially characterized mammalian H3K4 methyltransferase complexes indicates that the association of the Bre2-Sdc1 heteromer (human names ASH2-DPY30) with H3K4 methyltransferases (17,42) is highly conserved. Furthermore, the heteromer also exists independently of H3K4 methyltransferase complexes. In Schizosaccharomyces pombe, it is also found associated with the Jumonji domain protein Lid2 (15), which is the homologue of the trithorax group (trxG) protein Lid (little imaginal discs), (46). Ash2 (Bre2) is itself in the trxG and is highly conserved. Sdc1 is also highly conserved. In Caenorhabditis elegans, its homologue Dpy30 is involved in dosage compensation (47). We report here that Sdc1 has a very strong ability to interact with itself. This is in concordance with fact that the most conserved part of Sdc1 is very similar to the dimerization interface of the regulatory subunit of cAMP-dependent type II protein kinases (2). This interface consists of two ␣-helices that dimerize into a four-helix bundle (48). Hence, we suggest that Sdc1 mediates a dimerization of the Bre2-Sdc1 heteromer, which may contribute to dimerization of the Set1, Swd1-Swd3 scaffold. If this proposition is correct, then the absence of the Bre2-Sdc1 heteromer may cause Set1C to fall into two halves. If so, the residual H3K4 di-and trimethylation seen in the absence of Bre2 or Sdc1 may reflect methylation by a half-complex with only one active site. Conversely, the presence of two active sites in the complex may confer a degree of coordination, which could play a role in a process such as hemimethylation and epigenetic maintenance.
The absence of Bre2 or Sdc1 reduces the dimethylation of H3K4 and severely affects trimethylation, consistent with previous results (4,22). ChIP experiments showed that mutation of sdc1 did not reduce the presence of Set1 in the 5Ј and 3Јcoding regions of active genes. Consistent with this observation, we found that inactivation of Bre2 or Sdc1 resulted in an increase of H3K4 mono-and dimethylation at the 5Ј coding regions of active genes at the expense of trimethylation. These results suggest that Bre2 and Sdc1 regu- The core complex is made by Set1, the Swd1-Swd3 heterodimer, and probably Swd2. According to results obtained with higher eukaryotes (46), this core complex may mediate binding to dimethylated H3K4 (dashed lines), although it has not been shown that any of the Swd proteins is able to bind dimethylated H3K4. Shg1, Spp1, and the heteromer Bre2-Sdc1 are not required for complex assembly. The interaction of the heteromer Bre2-Sdc1 with Set1C requires the presence of the SET domain (Set1-(900 -1040)), whereas the region of interaction with Spp1 lies in the Set1-(560 -900) region. The catalytic SET domain likely integrates different inputs from Bre2-Sdc1 and Spp1 (12). Sdc1 is shown to contact only one of the two molecules of Bre2, because previous observations indicate that the stoichiometry of TAP-tagged Bre2 and Sdc1 is 2:1, whereas the ratio of TAP-tagged Sdc1 and Bre2 is 1:1 (2). In this model, the Set1 complex is proposed to dimerize through the ability of Sdc1 to make homodimers.
late Set1 histone methyltransferase activity rather than affecting the presence of Set1 in specific chromosomal regions. We further observed that deleting BRE2 or SDC1 abolished H3K4 dimethylation at the 3Ј end of coding regions. Therefore Bre2 and Sdc1 promote the catalytic activity of Set1 in 3Ј gene regions transcribed by the elongating pol II. 6 Consistent with this proposition, we showed that Set1C associates with both serine 5 and serine 2 phosphorylated forms of pol II. Recent studies indicate that trimethylated H3K4 peaks at the beginning of the transcribed portions of genes, dimethylated H3K4 is most enriched in the middle of genes, and monomethylated H3K4 is found predominantly at the end of genes (26). Our results indicate that Set1C can associate with pol II along the complete length of transcribed regions, with different subunits differentially stimulating catalytic activity according to the phase of the transcriptional cycle.
In addition to the highly conserved heteromers that interact with both Set1-and Mll-type H3K4 methyltransferases, Set1C includes three individual proteins that do not appear to interact with any subunit other than Set1. They include the plant homeodomain finger protein Spp1, the WD40 repeat protein Swd2, and the pioneer protein Shg1. The twohybrid analysis also revealed an interaction between Spp1 and Set1-(580 -1080). However, in contrast to Bre2, Spp1 co-purifies with Set1-(1-900), suggesting that it interacts with a region of Set1 lying between residues 580 and 900, which includes the N-SET domain (12). The counterpart for Spp1 in the human Set1C is CFP (CXXC finger protein), also known as CGBP for CpG-binding protein. Both proteins include a very similar plant homeodomain finger. Additionally CFP possesses a domain that binds unmethylated CpG dinucleotides.
Swd2 is probably a homodimer (2) whose function remains unclear; however, it is also a subunit of CPF (cleavage and polyadenylation factor) and therefore could play a similar role for Set1C and CPF at the 3Ј end of genes. The smallest subunit, Shg1, does not appear to be well conserved in evolution. The S. pombe Set1C has a counterpart; however, our data base searches have only found homologues among other fungi (not shown). It appears to play a moderately repressive role on Set1C methylation activity.
We extend pre-existing data on the structure of Set1C (2,4,(13)(14)(15)(16)22), which permits the advancement of a model (Fig. 5). The core complex is made by Set1, the Swd1-Swd3 heterodimer, and probably Swd2. Hence, the association between the H3K4 enzyme and the H3K4 binding domain is fundamental to complex integrity. The C-terminal region of Set1 contacts Bre2. In turn, Bre2 forms an heteromer with Sdc1, which has the ability to interact with itself. Spp1 interacts with a region of Set1 lying between residues 580 and 900. Both the Bre2-Sdc1 heteromer and Spp1 differentially regulate H3K4 di-and trimethylation during the transcription cycle of pol II. Further work is required to resolve the size and composition of the complex as well as the role of Swd2. This information will advance the understanding of the most evolutionarily conserved, and to date best understood, SET domain complex.