Our approach to charting proteomic environments relies upon the sequential use of TAP and mass spectrometry to identify stable protein assemblies. In a typical TAP pull-down experiment, LC-MS/MS analysis identified over 500 proteins containing stoichiometric and transient bona fide protein interactors, along with a large number of background proteins of diverse origin and abundance. To dissect the composition of complexes, we employed a layered data mining approach. First, we sorted out common background proteins and then distinguished proteins specifically enriched in the TAP isolation using semi-quantitative estimates of their abundance (Figure 1).

From 1,301 unique open reading frames (ORFs) in the master table, only 63 proteins (less than 5% of all identified proteins) matched the above selection criteria, comprising 9 stable protein complexes connected by 12 proteomic hyperlinks. Three out of these nine (ASTRA (for ASsembly of Tel, Rvb and Atm-like kinase), Snt2C and Sc_Rpd-LE (for Rpd3L expanded with Set3C core); Figure 2) are reported here for the first time, whereas the other six (complexes I-VI) have been characterized previously (note that the prefixes Sc_ and Sp_ refer to proteins from S. cerevisiae and S. pombe, respectively; the suffix 'C' always refers to the protein complex).

Chromatin Central comprised four distinct protein assemblies, including: the histone deacetylase Rpd3p (Sc_Rpd3S, Sc_Rpd3L 2425, Sc_Rpd-LE and Sc_Snt2C); at least two histone acetyltransferase complexes, Sc_NuA4 26 and SAGA/SLIK 27; and two ATP-dependent chromatin remodeling complexes, Sc_Swr1C and Sc_Ino80C 2829. The compositions of the individual protein complexes (Tables 1, 2, 3, 4, 5) were compared with previous reports. Surprisingly, we found some discrepancies with data from the best proteome maps even though they were also obtained by TAP tagging 1112. In contrast, our results agree with several publications describing the biochemical and functional characterization of the individual complexes. In particular, complexes I, V and VI are identical to the previously reported Sc_Rpd3S, Sc_Swr1C and Sc_INO80C, respectively 24252829.

In addition to the 12 known members of Sc_Rpd3L (complex II) 2425, we identified 2 novel subunits, including the 72 kDa protein Sc_Dot6p (ORF name YER088C) and its 59 kDa homolog Sc_Tod6p (Twin of the Dot6; ORF name YBL054W). Their sequences share 31% identity; 46% similarity and both possess the chromatin specific SANT domain 30. Furthermore, the involvement of Sc_Dot6 in the regulation of telomere silencing has been indicated 31.

In addition to the 14 known members of Sc_NuA4 (complex IV) 2632, another new protein, the 72 kDa Sc_Yap1p (ORF name YML007W), which is a member of a family of fungal specific transcriptional activators, was identified as a subunit. Within Sc_Set3C (complex III) 10 we also identified a new member, the 55 kDa protein Sc_Tos4p (ORF name YLR183C). It is a putative transcription factor of the forkhead family. Tagging Sc_Tos4p pulled down the entire Sc_Set3C, except for the hyperlink Sc_Hst1p 5 (also, see Figure S2 in Additional data file 2 and Table S2 in Additional data file 1).

We identified 12 proteomic hyperlinks in Chromatin Central (Figure 2). One of these proteins, the 422 kDa Sc_Tra1p (ORF name YHR099W) is a core member of Sc_NuA4 and SAGA/SLIK 27, effectively also hyperlinking these two acetyltransferase complexes into Chromatin Central. Our attempts to TAP-tag Sc_Tra1p failed. However, Sc_Tra1p was co-purified when other Sc_NuA4 and also ASTRA subunits were sequentially tagged (Figure 2; also see Figure S2 in Additional data file 2 and Table S2 in Additional data file 1).

Notably, core-subunits of the histone deacetylase complex Sc_Set3C 10 were co-purified in sub-stoichiometric amounts with subunits of the Sc_Rpd3L complex (Table S2 in Additional data file 1). Sc_Set3C and Sc_Rpd3L complexes regulate overlapping target genes 333435 and synthetic lethal screens have revealed genetic links between components of these complexes 36.

Altogether, the composition of individual complexes in Chromatin Central accords well with the published biochemical evidence. Furthermore, the sequential tagging approach revealed four novel subunits in three previously characterized complexes as well as three novel protein assemblies.

We next asked if the Chromatin Central environment is conserved between the distantly related fungi S. cerevisiae and S. pombe. In contrast to S. cerevisiae, no systematic biochemical isolation of protein complexes has yet been performed in S. pombe; however, complexes can be isolated with essentially the same TAP methodology with a similar success rate 737. We exploited the architecture of Chromatin Central in S. cerevisiae to choose strategic baits for the work in S. pombe. The closest homologues of six S. cerevisiae hyperlinks (products of CLR6, ALP13, YAF9, SWC4, RVB1, TRA1 and TRA2 genes) were subjected to TAP tagging and immunoaffinity isolation, followed by mass spectrometric identification of corresponding interactors (Figure 3). For accuracy, we also isolated complexes associated with three more conserved subunits, encoded by PNG2, SWC2 and IES6. Thus, the characterization of each complex relied upon at least two independent TAP purifications targeting different baits.

As in the S. cerevisiae experiments, the identified proteins, together with their A-indices, were combined into a master table (Tables S2 and S4 in Additional data file 1). We also compiled a list of 250 common background proteins for S. pombe in the same way as we did for S. cerevisiae (Table S1 in Additional data file 1). Interestingly, the average A-index of common background proteins was almost identical in both yeasts (1.3 and 1.4 in the fission and budding yeasts, respectively), and, therefore, we used the same conservative threshold of 1.75 to define stoichiometric interactors.

Chromatin Central shows a very similar architecture in both yeasts (Figures 2 and 4). To assess the similarities more closely, we focused on orthologues, recognized by overall sequence similarity (best hits in forward and reciprocal BLAST searches) and similar composition of structural domains (Tables 1, 2, 3, 4, 5). Altogether, in both Chromatin Central environments we identified 47 pairs of confident orthologues and six pairs with marginal confidence (Figure S3 in Additional data file 2) out of a total of 139 proteins. For other S. cerevisiae and S. pombe proteins, BLAST searches identified no clear orthologous pairs (Tables 1, 2, 3, 4), although some of them might be functional orthologues (such as Sc_Ume1p and Sp_Prw1p).

More than half the subunits of Sc_Rpd3S and Sc_Rpd3L (complexes I and II) are orthologous to the members of corresponding S. pombe complexes Sp_Clr6S and Sp_Clr6L; however, we reveal (Figure 4 and Table 2) further similarities than previously documented 38. In addition to the previously reported subunits, we identified Sp_Cti6p, Sp_Rxt2p, Sp_Rxt3p, Sp_Dep1p and Sp_Pst3p. Our study also revealed that Sp_Clr6L, like Rpd3L in the budding yeast, is hyperlinked to the NuA4 histone acetyltransferase complex via an MRG-family protein, Sp_Alp13p.

Complex IV (Sp_NuA4) comprises orthologues of the 12 core members of the Sc_NuA4 complex, including its catalytic subunit Sp_Mst1p (ORF name SPAC637.12c) 394041 (Table 1). Complexes V and VI include the closest homologues of the S. cerevisiae ATP-dependent helicases Sc_Swr1p and Sc_Ino80p (ORF names SPAC11E3.01c and SPAC29B12.01, respectively), together with 20 subunits orthologous to members of Sc_Swr1C and Sc_Ino80C (Table 3). The corresponding chromatin remodeling complexes in S. cerevisiae catalyze replacement of histone H2A with its variant Htz1p 294243. Complexes V and VI in the fission yeast both associate with Sp_Pht1p, which is the S. pombe orthologue of Sc_Htz1p (Table S2 in Additional data file 1). Therefore, it is likely that these S. pombe complexes (now called Sp_Swr1C and Sp_Ino80C) are also H2A.z chaperones.

Interestingly, while characterizing the composition of Sp_Ino80C, we identified a 17 kDa core subunit, whose gene has not been annotated as an ORF in the S. pombe genome (Figure S4 in Additional data file 2). The protein has no homologues within the Saccharomyces genus, yet possesses some remote similarity to a non-annotated genomic region in Schizosaccharomyces japonicus. We call this newly discovered S. pombe gene, IEC5 short for (Ino Eighty Complex subunit 5 [GenBank:FJ493251]).

Complex VI, ASTRA, is the same as the orthologous complex in S. cerevisiae except that the S. pombe genome encodes for two Tra1 homologues and only one, Tra1, is present in ASTRA (Table 4). The other, Tra2, is a subunit of Sp_NuA4 and presumably the S. pombe SAGA/SLIK complexes. In S. cerevisiae, the single Tra1 was found in all three complexes.

As we observed in S. cerevisiae for Sc_Rpd3L, some Sp_Set3C subunits co-purified in sub-stoichiometric amounts with Sp_Clr6L and vice versa, when Sp_Set3p was used as bait (Table S2 in Additional data file 1). Notably, the three subunits (Sp_Snt1p, Sp_Hif2p, and Sp_Set3p) isolated together with Clr6L are the orthologues of the three (Sc_Snt2, Sc_Sif2, and Sc_Set3) isolated with Rpd3L. However, in contrast to the Sc_Set3C complex, which consists of eight subunits, the Sp_Set3C complex contains only four proteins (Table 2).

In a few instances we identified proteins with domains that are not present in the corresponding orthologous complexes in the other yeast, including Sp_Msc1p (ORF name SPAC343.11c), which is a member of the Sp_Swr1C complex. The function of this protein is not known, although Ahmed et al. 44 suggested that Msc1 is involved in chromatin regulation and DNA damage response. Msc1 contains a remarkable composition of domains, including three PHD fingers 45, PLU-1 46, zf-C5HC2, JmjC and JmjN 47. It was recently shown that the Msc1 PHD fingers act as an E3 ubiquitin ligase 48, while in other proteins the JmjC domain mediates histone demethylation 49. None of the Sc_Swr1C subunits possess these domains or appears to be remotely similar to Sp_Msc1 (Table S5 in Additional data file 2).

We identified nine hyperlinks within Chromatin Central in S. pombe, all of which are orthologues to corresponding proteins in the budding yeast. As our attempts to purify TRA2 failed (as they did in S. cerevisiae), it remains unclear if, similar to the budding yeast, this protein is also shared between Sp_NuA4 and an assembly orthologous to SAGA/SLIK 50.

We independently validated some of the novel proteomics observations, namely the insights regarding Set3C and Swr1C, using quantitative genetic interaction data from S. cerevisiae 51 and S. pombe 5253.

Our proteomic data suggest that Set3C contains a conserved core complex of four proteins (Set3, Hos2, Snt1, and Sif2) and physically interacts with Rpd3L in both S. cerevisiae and S. pombe. To validate these findings, we compared the correlation coefficients of genetic profiles of the two Set3C core components in both yeasts (Sc_Set3 and Sc_Hos2; Sp_Set3 and Sp_Hda1) against genetic profiles of a set of 239 direct homologs in the two species. As expected, the correlation between the two core members (Set3 and Hos2) and Sif2, is well beyond the 90th percentile (Figure 5a), suggesting they act in concert. In contrast Hst1, also a subunit of Sc_Set3C, shows a much lower correlation, which is consistent with its absence from the core complex (as was previously shown for Sc_Hst1 10) or not being a part of the complex at all (Sp_Hst1). Furthermore, correlation of the Set3C core with the Sc_Rpd3L subunit Sp_Pho23 (Sp_Png2) is also high in both yeasts and higher than the correlation with one of the Rpd3S subunits (Rco1, Sp_Cph1). The functional division within Sc_Set3C becomes even more obvious when examining individual interactions of Set3C core and extension subunits. Members of the Set3C core display stronger positive genetic interactions with each other, compared to the Set3C extension subunits, and their genetic interaction patterns differ from patterns of Swr1C, SAGA and Prefoldin members (Figure 5b). Taken together, these data provide genetic evidence that Sc_Set3C encompasses two functional modules, one of which (Set3C core) interacts closely with Rpd3L. This functional evidence corroborates our proteomic mapping data, suggesting that the Set3C complex in S. pombe is only represented by core subunits, while the orthologous complex in S. cerevisiae has an extension of four extra subunits. In both yeasts, the core Set3C interacts with Rpd3L to form a distinct module referred to as Rpd3LE.

In S. pombe another complex, Swr1C, contains a newly identified subunit, Msc1. Its closest homolog in Sc_Ecm5 is not a part of Swr1C in budding yeast. To independently validate this finding, we examined and compared individual genetic interactions of seven of the Swr1C subunits in both yeasts with the genetic patterns of Sp_Msc1 (Sc_Ecm5). Consistent with our proteomic data, Sp_Msc1, unlike Sc_Ecm5, shows strong positive genetic interactions and a very similar pattern to the other members of the complex (Figure 5c). Hence, pairs of genetic profiles containing Sc_Ecm5/Sp_Msc1 and other members of Swr1C show weak correlation in S. cerevisiae, but strong correlation in S. pombe (Figure 5d). Taken together, these genetic data confirm our proteomic mapping observations.

The overall architecture of Chromatin Central is the same in the two yeasts; however, there is a surprising amount of variation in their subunit composition. For both yeasts, Chromatin Central is based on the same eight complexes, encompassing 53 orthologous protein pairs plus a further 33 proteins that appear to be species specific (23 in S. cerevisiae, 10 in S. pombe). Of these 33, three pairs have a very similar composition of functional domains (Table S5 in Additional data file 2). Hence, a very similar architecture is sustained by a scaffold based on about two-thirds of all the proteins involved, whereas the remaining one-third appears to be less constrained.

Analyses of domain composition (Table S5 in Additional data file 2) revealed that many subunits (19 in S. cerevisiae and 21 in S. pombe) possess bromo-, chromo-, SANT or PHD finger domains, which can bind either methylated or acetylated histones or other chromatin determinants 3054555657, thus potentially targeting their complexes to specific regions of chromatin. Along with the seven enzymes in Chromatin Central, these putative targeting subunits are the most highly conserved subunits between the two yeasts. For example, Sc_NuA4 and Sp_NuA4 complexes retain all four targeting factors: the PHD fingers of Sc_Yng2/Sp_Png2; the bromodomains of Sc_Bdf1/Sp_Bdc1 and the chromodomains of Sc_Eaf3/Sp_Alp13 and Sc_Esa1/Sp_Mst1. Similarly, the different PHD finger subunits of Rpd3S and Rpd3L (Sc_Rco1 and Sp_Cph2 in Rpd3S/Clr6S; Sc_Cti6 and Sp_Cti6 as well as Sc_Pho23 and Sp_Png2 in Rpd3L/Clr6L), which appear to direct these complexes to differentially methylated nucleosomes 58, are conserved. Hence, almost half of the conserved scaffold of Chromatin Central is based on proteins that convey the functions of the environment, that is, the reading and writing of the histone code.

In contrast to the conserved scaffold of Chromatin Central, the proteomic hyperlinks appear to be less constrained. We define proteomic hyperlinks, which are notable features of proteomic environments, as proteins found as stoichiometric subunits of more than one scaffold complex. Hyperlinks do not physically connect complexes; rather, they could exist for one of the following three reasons. First, hyperlinks might reflect a common ancestry for two complexes. In Chromatin Central, it is possible that Ino80C and Swr1C are examples of complexes that share a common evolutionary origin because they not only share four subunits, but also share a similar function related to the histone variant H2A.z 4359. Second, hyperlinks may play a functional role to co-ordinate two complexes. If the hyperlink receives a signal via a post-translational modification, then two complexes should receive the signal at the same time and, hence, be co-regulated. Conversely, if a hyperlink recognizes an epitope or target, then both complexes will be coordinately recruited. In Chromatin Central, the Rpd3S/NuA4 hyperlink Eaf3 plays a role in coordinating these complexes 60. Third, the hyperlink may be a common tool recruited to the complex. In Chromatin Central, the Rvb1/Rvb2 heterodimeric helicase is a good example to consider. It is present in three complexes, Swr1C, Ino80C and ASTRA, presumably because each requires a helicase for action in chromatin.

Altogether, we distinguished 21 proteomics hyperlinks in both Chromatin Central environments, 12 in S. cerevisiae and 9 in S. pombe (Figures 2 and 4; Table S6 in Additional data file 1). These proteins display a variety of physicochemical properties and domains (Table S5 in Additional data file 2 and Table S6 in Additional data file 1). For instance, in S. cerevisiae four hyperlink proteins are enzymes: Sc_Rvb1p and Sc_Rvb2p are DNA helicases, Sc_ Rpd3p is a histone deacetylase, and Sc_Bdf1p is a protein kinase. Sc_Act1p and Sc_Arp4p are cytoskeleton and structural proteins. Sc_Tra1p belongs to a protein kinase family (although its catalytic activity has been questioned 61), whereas no enzymatic activity has been reported for the other three proteins, Sc_Eaf3p, Sc_Yaf9p, Sc_Swc4p. Thus, the hyperlinks display diverse functional roles. However, they are all members of highly conserved protein families with clear homologues even in vertebrates. Also, half of the S. cerevisiae hyperlinks (6 out of 12) are essential, whereas only 3 essential genes were additionally found among the other 73 members of the environment.

Of the twelve hyperlinks in S. cerevisiae Chromatin Central, three are not conserved between the two yeasts. In two out of these three cases, the lost hyperlink is due to gene duplication or deletion. For Rpd3S and Rpd3L, Sin3 is a hyperlink in S. cerevisiae, but in S. pombe, both Clr6S and Clr6L have a dedicated Sin3 homologue. In fact, S. pombe has three Sin3 paralogues, with Sp_Pst2p being exclusively found in Clr6S whereas the other two homologues, Sp_Pst1p and Sp_Pst3p, are exclusively found in Clr6L. For NuA4, SAGA/SLIK and ASTRA, Tra1 is a hyperlink in S. cerevisiae, but in S. pombe, the tra1 gene is duplicated with one paralogue present in ASTRA and another in NuA4 and, presumably, SAGA/SLIK. We have previously noted the same phenomenon regarding a lost hyperlink. In S. cerevisiae, Swd2 hyperlinks Set1C and CPF; however, in S. pombe these two complexes each have a dedicated Swd2 paralogue, Swd2.1 and Swd 2.2. Notably, deletion of the gene encoding Swd2.1 did not provoke Swd2.2 to occupy the missing position in Sp_Set1C 7. In the third case, Bdf1 is a hyperlink between NuA4 and Swr1C in S. cerevisiae, but in S. pombe, it is replaced by two non-orthologous proteins with a similar composition of domains (Tables 1 and 3; Table S5 in Additional data file 2).

Although we have documented only a few examples, this early concordance between hyperlinks and gene duplications/deletions is notable and indicates that a gene duplication in evolution may be especially advantageous when it is a proteomic hyperlink. Gene duplication permits the diversification of the encoded protein. However, unless all genes encoding interacting proteins, particularly members of the corresponding protein complex, are also duplicated, diversification of the duplicated protein will be constrained by the existing spectrum of protein-protein interactions 62. If the duplicated gene encodes a hyperlink, then diversification of the duplicated protein will be less constrained because the duplications can associate and specialize with different complexes (Figure 6). Therefore, the gene duplication of a hyperlink may be more successful than other gene duplications. From a technical point of view, we suggest that the tagging of hyperlinks as an entry point for mapping proteomic environments will be a rewarding focus for mapping and understanding proteomes.

Despite the fact that this work was based on one of the most intensely studied proteomic environments to date, we documented four new subunits in the six known complexes (described above in Results). We also found three new complexes, including two more containing the histone deacetylase Rpd3, to add to the recently described Rpd3S and Rpd3L complexes 2425.

We found the three-member Sc_Snt2 complex in an Rpd3-TAP pull down. Subsequently, we validated the complex by tagging Snt2 (Figure 7a). Sc_Snt2p and the other subunit, Sc_Ecm5p, contain several domains involved in chromatin regulation (Figure 7b). Sc_Snt2p (YGL131C) is one of 18 SANT domain-containing proteins in budding yeast and the sixth identified in Chromatin Central (Table S5 in Additional data file 2). Due to its JmjC domain, Sc_Ecm5p is a putative histone demethylase. Its PHD fingers have been reported to bind methylated H3K36 58.

We did not find a complex similar to Sc_Snt2C in S. pombe. Previously, we showed that the fission yeast orthologue of Sc_Snt2p is also found in a complex with the JmjC domain proteins Sp_Lid2p and Sp_Jmj3p 717. However, these two complexes, Sc_Snt2C and Sp_Lid2C, are very different. From our IPs and Coomassie gels, we note that the Sc_Snt2C is much less abundant than either Rpd3S or Rpd3L.

The second new Rpd3-containing complex, Rpd3-LE, also contains a minor fraction of cellular Rpd3. Rpd3-LE is an extension of the Rpd3L complex, which includes the core of the Set3 complex. We have four reasons to be confident of this assignment. First, we reciprocally observed the relationship between Rpd3L and core Set3C in both yeasts. Second, in S. cerevisiae we observed only the core proteins of Set3C, but not the full eight-membered Set3C, associated with Rpd3L. This accords with our previous dissection of Sc_Set3C, which showed that the complex has a core of these four proteins 10. Third, Sp_Set3C consists of only the four core proteins found in Sc_Set3C and does not include homologues of Sc_Hst1p or Sc_Hos4p. It also does not include a homologue of the cyclophilin Sc_Cph1p, which is required to fine-tune the sporulation-specific activity of Sc_Set3C 1063. Fourth, genetic interaction data clearly support the division of Sc_Set3C into core and extension components, as well as the interaction of the core with Rpd3L.

Taken together, these data indicate that Set3C has two versions, one that is small, common to both yeasts and interacts with Rpd3L, and the another that is larger and specific to S. cerevisiae. Sc_Set3C functions during vegetative growth to regulate several inducible genes and to co-operate with Rpd3 333435. It has also been shown to be a negative regulator during meiosis 10. Partly because transcriptional control of meiosis appears to be poorly conserved 64, we suggest that the vegetative functions in S. cerevisiae entirely relate to the smaller complex.

We also discovered the ASTRA complex, which is composed of orthologous subunits in both yeasts and contains a Tra1 member of the ATM-like kinase family (Figures 2, 4 and 7C,D; Table 4). ASTRA includes the ATP-dependent DNA helicase Rvb1/2 heterodimer, the telomere binding protein Tel2p together with two uncharacterized Tel2-interacting proteins (Tti1p and Tti2p), and a WD-repeat containing protein (Asa1p) of unknown function in both yeasts.

Protein assemblies similar to ASTRA have not been identified in other organisms. However, five subunits have clear human homologues so the complex may be widespread. Furthermore, Sc_Tel2p, as well as the C. elegans (CLK-2 65) and human (TELO 66) orthologues are required for telomere length regulation. Sc_Tel2p also influences telomere position effect and interacts directly with double-stranded telomeric DNA 6768. Recently, Hayashi et al. 69 suggested that a Tel2-Tti1 heterodimer recognizes a common domain of phosphatidyl inositol 3-kinase related kinases (PIKK) in the fission yeast and is a component of multiple PIKK assemblies, which corroborates our findings in S. cerevisiae (Table S2 in Additional data file 1). Hence, we suggest that ASTRA is the Tra-specific Tel2/CLK-2/TELO complex involved in telomeric chromatin regulation and will be found in many eukaryotes.

The stringency of our protein selection criteria may have excluded some lowly abundant interactors, as well as proteins only interacting with a particular subunit of the complex, rather than with the entire complex cores. This might include several interesting possible further connections to Chromatin Central (Tables S2, S3 and S4 in Additional data file 1). For example, we observed several subunits of the SWI-SNF global transcription activator complex 707172, RSC chromatin remodeling complex 7273, Regulator of Nucleolar silencing and Telophase exit (RENT) 7475, Anaphase Promoting Complex (APC) 37, cytoplasm to nucleus signaling complexes TORC1 and TORC2 6976, the proteasome, DNA replication licensing complex 77, the CCR4/NOT transcription complex 78 and various transcription factors.

The histone deacetylases Rpd3p and Clr6p pulled down the complete TRiC chaperonin complex 79 in S. cerevisiae and S. pombe, respectively (Figures 2 and 4; Tables S3 and S4 in Additional data file 1). Similarly, TRiC was co-purified with the Hos2 deacetylase complex Set3C in S. cerevisiae 5. Hence, we suggest that TriC chaperonin activity is specifically related to type I histone deacetylases.

Transformations for both yeasts were performed as described 79. Genes of interest were tagged by in-frame fusion of the ORFs with a PCR generated targeting cassette encoding the TAP-tag and a selectable marker. Correct cassette integrations were confirmed by PCR and Western blot analysis. Two S. cerevisiae strains with TAP-tagged genes YGR099W and YJR136C were obtained from Euroscarf (Frankfurt am Main, Germany). Breaking and extraction of yeast cells was performed as described 7 with modifications 10. Purified proteins were concentrated according to Wessel and Fluegge 93 and loaded onto one-dimensional gradient (6-18%) polyacrylamide gels.

Protein bands were visualized by staining with Coomassie. Full lanes were cut into approximately 30-40 slices; to enhance the detection dynamic range, visible bands were always sliced separately. Excised gel plugs were cut into approximately 1 mm × 1 mm × 1 mm cubes and in-gel digested with sequencing grade modified porcine trypsin (catalogue number V5111, Promega, Mannheim, Germany) as described in 94. Then, 1 μl aliquots were withdrawn directly from in-gel digests for the protein identification by MALDI peptide mapping. The rest of the peptide material was extracted from the gel pieces with 5% formic acid and acetonitrile and recovered peptides dried down in a vacuum centrifuge.

Where specified, 1 μl aliquots of in-gel digests were analyzed on a REFLEX IV mass spectrometer (Bruker Daltonics, Bremen, Germany) using AnchorChip probes (Bruker Daltonics) as described in 9596. Peaks were manually selected and their m/z searched against MSDB protein database of S. cerevisiae or S. pombe species using MASCOT 2.0 software (Matrix Science Ltd, London, UK), installed on a local two CPU server. Mass tolerance was set to 50 ppm; variable modifications: oxidized methionines; one misscleavage per tryptic peptide sequence was allowed. Spectra were calibrated externally using m/z of known abundant trypsin autolysis products as references. Protein hits whose MOWSE score exceed the value of 51 (the threshold confidence score suggested by MASCOT for p < 0.05 and the corresponding species-specific database) were considered significant, but were only accepted upon further manual inspection, which made sure that the m/z of all major peaks in the spectrum matched the masses of peptides from the corresponding protein sequences or known tryptic autolysis products.

Dried peptide extracts were re-dissolved in 20 μl of 0.05% (v/v) trifluoroacetic acid and 4 μl were injected using a FAMOS autosampler into a nanoLC-MS/MS Ultimate system (Dionex, Amstersdam, The Netherlands) interfaced on-line to a linear ion trap LTQ mass spectrometer (Thermo Fisher Scientific, San Jose, CA, USA). The mobile phase was 95:5 H₂O:acetonitrile (v/v) with 0.1% formic acid (solvent A) and 20:80 H₂O:acetonitrile (v/v) with 0.1% formic acid (solvent B, Lichrosolv grade). Peptides were first loaded onto a trapping microcolumn C18 PepMAP100 (1 mm × 300 mm ID, 5 mm, Dionex) in 0.05% trifluoroacetic acid at a flow rate of 20 ml/minute. After 4 minutes they were back-flush eluted and separated on a nanocolumn C18 PepMAP100 (15 cm × 75 μm ID, 3 μm, Dionex, Sunnyville, CA, USA) at a flow rate of 200 nl/minute in the mobile phase gradient: from 5-20% of solvent B in 20 minutes, 20-50% B in 16 minutes, 50-100% B in 5 minutes, 100% B during 10 minutes, and back to 5% B in 5 minutes; %B refers to the solvent B content (v/v) in A+B mixture. Peptides were infused into the mass spectrometer via a dynamic nanospray probe (Thermo Fisher Scientific) and analyzed in positive mode. Uncoated needles Silicatip, 20 μm ID, 10 μm tip (New Objective, Woburn, MA, USA) were used with a spray voltage of 1.8 kV, and the capillary transfer temperature was set to 200°C. In a typical data-dependent acquisition cycle controlled by Xcalibur 1.4 software (Thermo Fisher Scientific), the four most abundant precursor ions detected in the full MS survey scan (m/z range of 350-1,500) were isolated within a 4.0 amu window and fragmented. MS/MS fragmentation was triggered by a minimum signal intensity threshold of 500 counts and carried out at the normalized collision energy of 35%. Spectra were acquired under automatic gain control in one microscan for survey spectra and in three microscans for MS/MS spectra with a maximal ion injection time of 100 ms per each microscan. M/z of the fragmented precursors were then dynamically excluded for another 60 s. No precompiled exclusion lists were applied.

MS/MS spectra were exported as dta (text format) files using BioWorks 3.1 software (Thermo Fisher Scientific) under the following settings: peptide mass range, 500-3,500 Da; minimal total ion intensity threshold, 1,000; minimal number of fragment ions, 15; precursor mass tolerance, 1.4 amu; group scan, 1; minimum group count, 1.

Dta files were merged into a single MASCOT generic format (mgf) file and searched against a database of S. cerevisiae or S. pombe proteins using MASCOT v.2.2 installed on a local two CPU server. Tolerance for precursor and fragment masses was 2.0 and 0.5 Da, respectively; instrument profile, ESI-Trap; variable modification, oxidation (methionine); allowed number of miscleavages, 1; peptide ion score cut-off, 15.

Hits were considered confident if two or more MS/MS spectra matched the database sequences and their peptide ion scores exceeded the value of 31 (the threshold score suggested by MASCOT for confident matching of a single peptide sequence at p < 0.05). For each identified protein, the number of matched peptides and of MS/MS spectra were exported from MASCOT output to Excel spreadsheets using a script developed in-house, which further created a non-redundant list of protein hits detected in all analyzed bands of the same IP experiment. If the same protein was sequenced in several bands, only the analysis that produced the highest number of matched peptides and spectra was reported.

Where specified, recovered tryptic peptides were sequenced de novo by nanoelectrospray tandem mass spectrometry on a QSTAR Pulsar i quadrupole time-of-flight mass spectrometer (MDS Sciex, Concord, Canada) as described 97. MS/MS spectra were interpreted manually using BioAnalyst QS v.1.1 software and candidate sequences searched against the genomic sequence of S. pombe by the tblastn program at the NCBI BLAST server. The search found, with several matched peptides, a non-annotated segment at chromosome 1. Subsequently, the full length sequence of the gene was produced by 5'-RACE and determined at the DNA sequencing facility in MPI CBG, Dresden.

Protein sequence database searches were carried out using a stand-alone version of NCBI-BLAST and the PSI-BLAST (position-specific iterated BLAST) interface at the NCBI 98. To identify orthologous genes in S. cerevisiae and S. pombe, we performed automated BLAST searches using sequences of all subunits of all identified complexes. Potential orthologues were further evaluated by reciprocal BLAST searches using all hits whose E-values were less than 10-fold higher than the E-value of the best hit. Best hits in reciprocal searches were regarded as orthologues. If no reciprocal best hit pair was identified, PSI-BLAST searches were carried out against all fungal sequences in the non-redundant protein database. E-values and PSI-BLAST iterations for highly divergent orthologues are shown in Figure S3 in Additional data file 2. Multiple sequence alignments also shown in Figure S3 in Additional data file 2 were done manually by using pair-wise alignments produced by BLAST as a template. Residues that were conserved in at least 75% of sequences are highlighted. Identification of protein sequence domains was carried out using a stand-alone version of the InterproScan software 99 against the Superfamily and HMM-Pfam databases using default settings.

Quantitative genetic interaction profiles in S. cerevisiae and S. pombe were generated as described 51100. Pearson correlation coefficients were calculated between all possible pairs of genetic profiles for an overlapping set of genes in both species and the data corresponding to Set3 and Hos2 (Hda1) are presented as scatter plots in Figure 5.