Research ResourcePhosphoproteomics

Mitotic Substrates of the Kinase Aurora with Roles in Chromatin Regulation Identified Through Quantitative Phosphoproteomics of Fission Yeast

See allHide authors and affiliations

Science Signaling  28 Jun 2011:
Vol. 4, Issue 179, pp. rs6
DOI: 10.1126/scisignal.2001588

Abstract

Kinases of the Aurora family are essential for the proper execution of mitosis in eukaryotes, and Aurora inhibitors are in clinical trials as anticancer drugs. We applied site-specific quantitative phosphoproteomics in conjunction with chemical inhibition of Aurora to identify mitotic Aurora substrates in fission yeast on a proteome-wide scale. We detected 8000 phosphorylation events, of which we assigned almost 6000 to a specific residue; 220 were reduced in cells exposed to the Aurora inhibitor. After controlling for unspecific effects of the inhibitor, we classified 70 sites (on 42 proteins) as probable targets of Aurora, which enabled refinement of the consensus sequence for phosphorylation by Aurora. Several of the substrate candidates were known targets of Aurora, validating the approach, but most represented newly detected Aurora substrates. The involvement of these Aurora substrates in diverse aspects of chromatin dynamics suggests that in addition to its established role in controlling chromosome compaction and attachment to the mitotic spindle, Aurora influences other aspects of chromatin architecture and function during mitosis.

Introduction

The serine-threonine protein kinases of the Aurora family are essential for the proper execution of mitosis in eukaryotes (1). Metazoan Aurora A is important for efficient entry into mitosis and bipolar spindle formation (2). Metazoan Aurora B and the single Aurora in yeast (known as Ipl1 in budding yeast and Ark1 in fission yeast) ensure that chromosomes become properly segregated by controlling chromosome compaction and chromosome attachment to the mitotic spindle, as well as cytokinesis and abscission (3). Aurora family kinases have been implicated in tumorigenesis, and both Aurora A and Aurora B are currently explored as targets for anticancer therapy (4, 5).

Crucial for understanding the molecular functions of Aurora is the identification of the cellular substrates of the kinase, but few in vivo substrates are known. In particular, kinetochore-associated proteins are phosphorylated by Aurora B or yeast Aurora, and this regulation is important for proper chromosome attachment (616). Members of the centralspindlin complex are phosphorylated by Aurora B to promote cytokinesis and control abscission timing (1721). Furthermore, Aurora B and yeast Aurora phosphorylate histone H3 on serine-10 (Ser10) during mitosis (22), leading to dissociation of heterochromatin proteins (23, 24), which may support propagation of heterochromatin through cell division (25, 26). To identify additional substrates of Aurora, we combined quantitative phosphoproteomics with chemical inhibition of Aurora in fission yeast and screened for proteome-wide changes in mitotic phosphorylation that depend on Aurora.

Results

Phosphoproteome comparison in the presence and absence of kinase activity identifies Aurora substrates

We sought to identify substrates of yeast Aurora by comparing the phosphoproteome in mitotic cells in the presence and absence of the kinase activity of Aurora. We created a fission yeast (Schizosaccharomyces pombe) strain expressing an analog-sensitive (AS) version of Aurora from the endogenous locus (ark1-as3) (27) and carrying a conditional cdc25 mutation (28). When shifted to the nonpermissive temperature, these cells arrest just before mitosis. We then released the cells into mitosis in the presence or absence of 1NM-PP1, an inhibitor specific for AS kinases (29). We harvested the cultures when the cells had reached a mitotic index of 80 to 90%, meaning that most of the cells were in mitosis (Fig. 1A). We compared the protein phosphorylation events occurring in the absence or presence of Aurora activity by site-specific quantitative phosphoproteomics (30), using stable isotope labeling by amino acids in cell culture (SILAC) (31) (Fig. 1A). Cells in the inhibitor-treated culture were labeled with heavy lysine (H); cells in the control culture, to which the solvent dimethyl sulfoxide (DMSO) was added, were labeled with light lysine (L). Thus, phosphorylation events that were reduced by the inhibitor treatment are characterized by a low H/L ratio. Because Aurora is expected to phosphorylate some of its substrates only when chromosomes are not properly attached to the mitotic spindle (32), we repeated the experiment with the microtubule-depolymerizing drug methyl 2-benzimidazole carbamate (MBC) to perturb chromosome attachment under otherwise identical conditions (Fig. 1A, experiment 2). MBC efficiently depolymerizes fission yeast microtubules (33), and most of the cells in the treated cultures failed to separate their spindle pole bodies in mitosis (Fig. 1A, experiment 2), indicating a failure to assemble a mitotic spindle. To ensure that changes in phosphorylation were specific to inhibition of Aurora and not resulting from inhibition of other kinases that could be targeted by 1NM-PP1, we subjected a strain expressing wild-type ark1+ to the same inhibitor and control treatment (Fig. 1A, experiment 3). Phosphorylation events that depend on the kinase activity of Aurora were expected to have a reduced H/L ratio in experiment 1 or 2 or both, but not in experiment 3. We performed two technical replicates for each experiment and after phosphopeptide enrichment observed 8254 phosphorylation events on 1544 proteins (about one-third of the fission yeast proteome); 530 of the observed phosphoproteins had not been identified in either of two previous fission yeast phosphoproteome analyses (fig. S1) (34, 35). More than 6000 of the phosphorylation events were detected in all three experiments (Fig. 1B), and we could assign more than 5800 to a specific residue with a median localization probability >98% (table S1). Most (>94%) of the phosphorylation events quantified in each experiment did not significantly change upon inhibitor treatment (Fig. 1A and table S1). Parallel quantitative analysis of the proteome indicated that changes in protein abundance upon inhibitor treatment were negligible (fig. S2). Strong reduction of phosphorylation events upon inhibitor treatment was more pronounced in cells expressing the inhibitor-sensitive ark1-as3 than in cells expressing wild-type ark1+ (Fig. 1A), suggesting that we had identified specific Aurora targets. We grouped the phosphorylation events that were reduced in experiment 1 or 2 into five classes of confidence on the basis of their repeated detection and mode of regulation in experiments 1 to 3 (Fig. 1C and table S1). We considered the 70 phosphorylation sites on 42 proteins in the highest confidence classes 1 and 2 as Aurora-dependent. Most of these phosphorylation sites mapped into protein regions outside well-defined domains (table S2), in agreement with the global analysis of substrates of cyclin-dependent kinase 1 (Cdk1), another major cell cycle kinase (36). Among the 150 phosphorylation events in classes 3 to 5 (Fig. 1C and table S1), there are likely additional Aurora-dependent sites, but the current data are too sparse to distinguish them from nonspecific inhibitor-dependent or stochastically occurring changes.

Fig. 1

Comparison of the mitotic phosphoproteome in the presence and absence of kinase activity of Aurora. (A) Schematic outline of the three experiments (Exp) (top), representative pictures of the cultures at the time of freezing in liquid N2, the mitotic index (MI) of each culture, and diagrams showing for each quantified phosphorylation event the change in abundance between heavy (H; inhibitor-treated) and light (L; control-treated) culture (H/L; x axis) as a function of the signal intensity (y axis) recorded in the mass spectrometer. The dashed lines mark the border for a decrease or increase in phosphorylation by more than a factor of 4 [log2 (H/L) <−2 or >2]. Most phosphorylation events (small gray circles) did not change in abundance between the two cultures. Significantly regulated phosphorylation events (P < 0.05) are depicted as red circles, and sites additionally matching the R-X-[ST] motif as dark red triangles. In the pictures of the cultured cells, DNA is stained with DAPI, and mitotic spindle pole bodies are labeled by the kinase Plo1, tagged with green fluorescent protein (Plo1-GFP). (B) Overlap among the three experiments for all phosphorylation events (Total, n = 8254), all quantified phosphorylation events (Quantified, n = 7082), and phosphorylation events exhibiting a significantly decreased H/L ratio upon inhibitor treatment (P < 0.05) (Decreased H/L ratio, n = 413). (C) Classification of phosphorylation events (P events) significantly reduced in experiment 1 or experiment 2. A down arrow (↓) indicates significant reduction; an up arrow (↑) significant increase; a right arrow (→) no significant change upon inhibitor treatment. n.d., not detected. Phosphorylation events in class 5 could not be localized (Local.) to a specific amino acid within the phosphopeptide, whereas all phosphorylation events in the other classes could be localized and are referred to as phosphorylation “sites” in the text. Phosphorylation sites in class 1 and 2 were considered dependent on Aurora activity. Five of the proteins with class 1 phosphorylation sites also have a class 2 phosphorylation site, so that the total number of proteins with class 1 or 2 sites is 42.

The presence of proteins in classes 1 and 2 that have previously been identified as substrates of yeast Aurora, or whose orthologs in metazoan organisms have been identified as substrates of Aurora B, validated the results. Class 1 contains Spc7 (homologous to Caenorhabditis elegans KNL-1 and human KNL1, which is also known as Blinkin) (table S1). In vertebrate cells, phosphorylation of this kinetochore protein by Aurora B is required for proper chromosome attachment to the mitotic spindle (12, 37). Furthermore, Bir1 (also known as Survivin), Pic1 (also known as INCENP), and Sgo2, all of which are interaction partners and reported substrates of yeast Aurora or whose orthologs are substrates of metazoan Aurora B (3, 3840), were present in class 1 or 2, as was the condensin subunit Cnd2 (homologous to Drosophila Barren and human CAP-H). Condensin localization to mitotic chromosomes requires Aurora in fission yeast and Aurora B in metazoans (4144). Human Aurora B phosphorylates CAP-H (45, 46), and fission yeast Aurora–dependent phosphorylation of Cnd2 is required for condensin function (46, 47). Fission yeast Aurora is also required for the localization of the protein phosphatase Clp1 to the contractile ring during anaphase (48), and a direct interaction partner of Clp1, called Nsk1, was among the proteins in class 2. Across all eukaryotes, the best-established Aurora B or yeast Aurora substrate is histone H3, which is phosphorylated on Ser10 (22). We observed the doubly modified peptide with Ser10 phosphorylation and lysine-14 (Lys14) acetylation, and the abundance of this peptide was substantially reduced upon Aurora inhibition in both experiments 1 and 2, but did not change in experiment 3 (fig. S3), corroborating that our strategy identified direct Aurora substrates.

Sequence analysis of the Aurora-dependent phosphorylation sites reveals a common motif

Analysis of the large set of potential yeast Aurora phosphorylation sites in the highest confidence classes 1 and 2 should reveal a common amino acid motif surrounding the phosphorylated site. Comparison of these sites to all phosphorylated sites in the data set revealed a strong preference for arginine (R, Arg) at position −2 (Fig. 2A). This partly matches the previously established motif for yeast Aurora-dependent phosphorylation, [RK]-X-[ST]-[ILV] (6). However, hydrophobic amino acids at position +1 were not prominent in class 1 and 2 sites (Fig. 2A and fig. S4). We examined the previously reported yeast Aurora and metazoan Aurora B phosphorylation sites and found that more than 65% of them did not have a hydrophobic amino acid at position +1 (fig. S5 and table S3). One example is the sequence surrounding Ser10 of histone H3, which has threonine (T) at position +1. Furthermore, among the published yeast Aurora and metazoan Aurora B sites, Arg at position −2 was more prevalent than Lys (K) (45 versus 4 occurrences) (table S3). On the basis of these published phosphorylation sites, as well as those that we identified here, we propose that R-X-[ST] is the common consensus motif for phosphorylation by Aurora B–type kinases. Aurora A and Aurora B have different functions, different localization, and different binding partners during mitosis (1, 49). However, the amino acid sequence similarity between Aurora A and Aurora B is high (75% identity within the kinase domain of the human proteins), and a single amino acid change converts human Aurora A into a protein that binds the Aurora B–interacting proteins INCENP and Survivin and complements for Aurora B (50, 51). Furthermore, some studies found that Aurora A and Aurora B phosphorylate the same site on target proteins (15, 5255), and [RKN]-R-X-[ST]-[ILVM] was identified as consensus for Aurora A on the basis of phosphorylation of synthetic peptides in vitro (56). We therefore hypothesize that R-X-[ST] is recognized by Aurora A as well as Aurora B. The R-X-[ST] consensus is furthermore shared with protein kinase A (57), and other determinants, like localization, must provide specificity.

Fig. 2

Consensus motif analysis of phosphorylation sites that depend on the kinase activity of fission yeast Aurora. (A) Class 1 and 2 sites were examined for the relative enrichment (red) or depletion (blue) of amino acids flanking the phosphorylation site (left side). The heat map depicts the ±log10 P values (derived by application of Fisher’s exact test). The occurrence of arginine (R) at position −2 was significantly more frequent (P = 2.03 × 10−12) in the 70 class 1 and 2 sites compared to all 5812 phosphorylation sites in the data set. Consensus motif analysis according to biochemical properties of amino acids (right side) showed an enrichment of positively charged amino acids at position −2 (P = 7.48 × 10−7). No significant (P < 1 × 10−4) enrichment of hydrophobic amino acids was observed at position +1. (B) Total number of considered sites (# sites) and percentage and number of sites matching the indicated R-X-[ST] consensus among all sites in a certain category. In contrast to sites from classes 1 and 2, phosphorylation sites that exhibited a decrease in phosphorylation due to nonspecific effects of the inhibitor (NS.↓) did not show enrichment for the R-X-[ST] motif. See table S1 for complete data.

Class 1 and 2 sites that are directly targeted by Aurora should conform to the R-X-[ST] consensus. About 30% of all class 1 and 2 sites matched this consensus, and the percentage was higher (~60%) when we considered only those sites that exhibited the greatest reduction in phosphorylation in Aurora-inhibited cells (H/L <0.25) (Fig. 2B). Sites that exhibited reduced phosphorylation due to nonspecific effects of the inhibitor (table S1) were not enriched for the R-X-[ST] motif (Fig. 2B).

There are several possible reasons for the occurrence of class 1 and 2 sites that did not match the consensus. Four sites matched the amino acid sequence K-X-[ST] (Id 556, 3455, 4377, and 5021; table S1). Because both Lys and Arg are positively charged and some of the yeast Aurora or metazoan Aurora B sites reported in the literature have Lys at position −2 (table S3), it is likely that these sites are targeted by Aurora, just like the R-X-[ST] sites. Other Aurora-dependent sites that did not match the R-X-[ST] consensus may have been falsely reported for technical reasons: Some of the sites were in the vicinity of another site matching the consensus. When these sites appeared together on one doubly phosphorylated peptide and when there were no other quantified phosphopeptides containing only one of the sites, we could not distinguish whether either or both sites were exhibiting the reduction in phosphorylation. In such cases, we reported both as reduced in order not to miss any potential sites. For several such doubly phosphorylated peptides (Id 2826, 3455, 5021, 6068, and 9099; table S1), there was independent evidence from another peptide (Id 2825, 3456, 5020, and 9346; table S1) that phosphorylation of the R-X-[ST] site was reduced upon Aurora inhibition, but it was impossible to determine whether phosphorylation of the neighboring non–R-X-[ST] site was also reduced. In two cases of adjacent sites that we examined closely (Elg1-Thr34 Ser36; Spc7-Ser4 Ser9), the R-X-[ST] site was phosphorylated by fission yeast Aurora in vitro, but the adjacent non–R-X-[ST] site was not (table S5). Thus, our motif analysis may contain sites that are falsely reported as reduced and Aurora-dependent. This technical issue does not jeopardize the substrate analysis: Even though we cannot say with certainty whether phosphorylation of both sites was reduced upon Aurora inhibition, phosphorylation of the peptide did depend on Aurora activity.

We repeated the analysis of the frequency of motif occurrence by including the K-X-[ST] sites and excluding non-[RK]-X-[ST] sites from doubly phosphorylated peptides (fig. S6). In this analysis, only two of the sites exhibiting strong reduction (H/L <0.25) did not fit the [RK]-X-[ST] consensus. We confirmed that one of these sites (Orc1-Ser9, Id 5035) was phosphorylated by fission yeast Aurora in vitro (table S5) despite the absence of the consensus motif.

The remaining class 1 and 2 sites that did not match the consensus may represent secondary sites that depend on kinases downstream of Aurora or sites that have passed the significance threshold by chance. We cannot distinguish these latter sites from secondary sites by formal parameters. Consistent with the possible identification of secondary sites, enrichment for the R-X-[ST] motif was greater among the sites that exhibited a greater reduction in phosphorylation after Aurora inhibition and, thus, are more likely to represent direct Aurora substrates (Fig. 2B and fig. S6). Phosphorylation of secondary sites still depends on the kinase activity of Aurora and, hence, gives an indication of pathways that are regulated by Aurora.

The spectrum of newly detected Aurora substrates suggests a role for the kinase in chromatin restructuring during mitosis

The presence of known Aurora substrates in class 1 and 2 and the motif analysis indicated that our strategy successfully identified direct Aurora substrates. Most of the proteins in class 1 or 2 had previously not been reported as Aurora targets. Most of these proteins are nuclear and many play a role in chromatin dynamics (Fig. 3, fig. S7, and table S4). Among these new candidate substrates are heterochromatin proteins and nucleolar proteins, as well as proteins implicated in the posttranslational modification of histones, in DNA replication, in DNA damage signaling, in RNA processing, and in nucleocytoplasmic transport (Fig. 3).

Fig. 3

Aurora targets identified in this study. Proteins with phosphorylation sites in classes 1 and 2 were grouped according to their GO Slim classification or GO classification for biological process with additional manual curation. The question mark group represents proteins without any well-defined function. Colored boxes indicate from left to right: (box 1) whether the protein is in class 1 (red) or class 2 (yellow); (box 2) whether at least one of the phosphorylation sites in the protein exhibited a strong reduction (H/L <0.25) by inhibitor treatment in experiment 1 or 2 (red, yes; yellow, no); (box 3) whether at least one of the phosphorylation sites in this protein that was reduced by Aurora inhibition matched the R-X-[ST] consensus (red, yes; yellow, no); (box 4) whether the protein localizes to the nucleus (red indicates nuclear localization of this protein or of a homologous protein from another organism; gray indicates cytoplasmic or otherwise nonnuclear localization; white indicates that no information was available); (box 5) whether any phosphorylation site in this protein exhibited reduced phosphorylation by inhibitor treatment in experiment 3 (gray, yes; red, no). The circles indicate whether the protein was phosphorylated by fission yeast Aurora in vitro (table S5) (red, yes, at least one of the sites that depended on Aurora in vivo was phosphorylated in vitro; gray, phosphorylation by Aurora could not be detected; hatched red and gray, the protein was phosphorylated by Aurora but on sites different from the ones suggested by the in vivo analysis). The phosphorylation sites represented by box 5 were in all cases different from the sites that put the protein into class 1 or 2, which were sites that did not exhibit a reduction in phosphorylation in experiment 3. Protein names for S. pombe are given. Alternate names, homolog names, or names of the protein complex that the protein is part of are indicated in parentheses. H.s., Homo sapiens; S.c., Saccharomyces cerevisiae.

To confirm some of the newly identified putative targets as direct Aurora substrates, we expressed candidate targets from different functional classes (Fig. 3) as recombinant proteins and tested their phosphorylation by fission yeast Aurora in vitro. Ten of 11 proteins tested were phosphorylated by Aurora, and 8 of those 10 were phosphorylated on at least one of the sites found by our in vivo analysis (Fig. 3 and table S5), confirming that these proteins are direct Aurora targets. Sixteen of the 19 phosphorylation sites found on these proteins after incubation with Aurora matched the [RK]-X-[ST] consensus; of the remaining 3, 2 had several Lys and Arg residues in position −3 to −6, and only for one site (Mdb1-Ser156), we could not detect a resemblance to the consensus. This confirms that Aurora has a strong preference for [RK]-X-[ST] sites. Of the nine phosphorylation sites that had been found to be Aurora-dependent in vivo, but that were not phosphorylated in vitro, five did not fit the [RK]-X-[ST] motif, and four of these (Id 2826, 6068, 6256, and 9099; table S1) may have been falsely reported because of their presence on a doubly phosphorylated peptide together with an [RK]-X-[ST] site. Another four sites were not phosphorylated in vitro despite matching the consensus. Because two of these sites belonged to class 1 and therefore had been found to be Aurora-dependent in vivo in two independent experiments, we suspect that the failure to phosphorylate these sites may have resulted from the conditions of the in vitro kinase assay, due to missing binding partners or improperly folded proteins. Hence, technical reasons may contribute to the failure to confirm all substrates in this assay.

We next addressed whether phosphorylation at identified sites is of functional consequence. Others have shown that phosphorylation of the condensin subunit Cnd2 at the site identified by us and at adjacent Aurora consensus sites is required for condensin binding to mitotic chromosomes and mitotic chromosome compaction (46, 47). Furthermore, we found that mutation of Aurora-dependent phosphorylation sites within the kinetochore protein Spc7 caused sensitivity to the microtubule-depolymerizing drug benomyl (Fig. 4), which is an indication of a mitotic defect. Combination of the Spc7 phosphorylation site mutants with mutations in Nuf2 (58), another outer kinetochore protein, caused a synthetic growth defect (Fig. 4C), suggesting that phosphorylation of Spc7 by Aurora is important for proper kinetochore-microtubule attachment. Mutation of Aurora-dependent phosphorylation sites in the origin recognition complex subunit Orc1 caused a synthetic growth defect and phenotypic enhancement when combined with deletion of the DNA damage checkpoint kinase gene chk1+ (Fig. 4, E and F). Similar observations have been made for orc1 temperature-sensitive mutants (59). Hence, the functionality of Orc1 is impaired by mutation of the phosphorylation sites. Together, these data suggest that both expected and unexpected Aurora-dependent phosphorylations identified in this study may be of functional relevance.

Fig. 4

Mutation of Aurora-dependent phosphorylation sites in the kinetochore protein Spc7 or the origin recognition complex protein Orc1 impairs their functionality. (A) Schematic diagram of the outer kinetochore protein Spc7 with phosphorylation sites identified in vivo (table S1) indicated above. The fragment tested for phosphorylation by fission yeast Aurora in vitro and the result from this assay (table S5) is indicated below. aa, amino acids. (B) Serial dilution growth test of various Spc7 phosphorylation site mutants compared to the wild-type (WT) strain. One of the plates contained the microtubule-depolymerizing drug benomyl, as indicated. Cells expressing spc7, in which three possible phosphorylation sites have been mutated (3A or 3D), showed an increased sensitivity to benomyl. (C) Serial dilution growth test of the indicated strains. Nuf2 is one subunit of the outer kinetochore Ndc80 complex. Cells expressing both the temperature-sensitive nuf2-2 mutant (58) and spc7-3A or -3D exhibited a synthetic growth defect. (D) Schematic diagram of the origin recognition complex subunit Orc1 with phosphorylation sites identified in vivo (table S1) indicated above. Full-length Orc1 was tested for phosphorylation by Ark1 in vitro and the result from this assay (table S5) is indicated below. (E) Tetrads resulting from crossing a chk1Δ (“Δ”) strain with an orc1+ (“1”), orc1-S6A S9A (“A”), or orc1-S6D S9D (“D”) strain. The chk1Δ orc1-S6A S9A double mutants showed reduced colony size compared to the single mutants, indicating a synthetic growth defect. Tetrads framed by a black box were chosen for the serial dilution growth test, shown in (F). (F) Serial dilution growth test for the tetrads indicated by a black box in (E). The hydroxyurea (HU) sensitivity of chk1Δ is increased by the presence of the orc1 phosphorylation site mutants.

Yeast Aurora and metazoan Aurora B have been best characterized for their role in regulating mitotic chromosome segregation through phosphorylation of kinetochore proteins. In contrast, few substrates related to chromatin have previously been reported. In addition to histone H3, tousled-like kinase (TLK-1), which plays a role in chromatin assembly and the regulation of transcription; MYB-binding protein 1A (MYBBP1A), a nucleolar protein, which has been reported to bind transcription factors; and NSUN2, which is named for NOP2/sun domain and is a nucleolar RNA methyltransferase, have been identified as targets of Aurora B (6062). Our data indicate that Aurora has a more widespread role in regulating chromatin.

To further corroborate a functional link between fission yeast Aurora and the newly identified substrates, we analyzed whether candidate substrate and Aurora mutants would show similar genetic interactions. When we searched for direct genetic interactors of the identified substrates (6365), we found that several candidate substrate genes had common genetic interactors (Fig. 5A and fig. S8). As an example, six of the candidate substrates showed a negative genetic interaction with cbp1 (CENP-B) (Fig. 5A and fig. S8). CENP-B is a conserved DNA binding protein required for genome integrity in fission yeast (66). An ark1 temperature-sensitive mutant was synthetically lethal with deletion of cbp1 (fig. S8), indicating that Aurora and the candidate substrates could be functionally related. Similar observations were made for genetic interactions with arp4 (also called arp42, encoding a subunit of the SWI-SNF and the RSC chromatin-remodeling complexes), chk1 (encoding a DNA damage checkpoint kinase), and swi3 (encoding a protein homologous to vertebrate tipin, a component of the replication fork protection complex) (Fig. 5A). In all these cases, the ark1 temperature-sensitive mutant exhibited genetic interactions similar to those reported for many of the substrates (Fig. 5A and fig. S8). Several candidate substrate genes or genes with which candidate substrates exhibited genetic interactions are required for resistance toward DNA-damaging agents (Fig. 5 and fig. S8). The temperature-sensitive ark1 mutants also exhibited an increased sensitivity to DNA-damaging agents (Fig. 5B). These data confirm a functional relationship between Aurora and some of the newly identified substrates and suggest that Aurora is a major regulator of mitotic chromatin and has a role in protecting from DNA damage (Fig. 5C).

Fig. 5

Aurora shows similar genetic interactions as identified substrates and is required for resistance to DNA damage. (A) Left side: Genetic interactions reported for some of the identified Aurora substrates (84). Genes required for resistance to DNA damage are shown in gray. Right side: Serial dilution growth test at 28°C of the indicated yeast strains compared to the WT strain. One of the plates contained camptothecin (CPT), as indicated, to assay for sensitivity to DNA damage. (B) Serial dilution growth test of the indicated yeast strains. The degree of sensitivity of the ark1 mutant strains to the DNA-damaging agents CPT and HU correlated with the severity of the temperature-sensitive allele (ark1-T8 is the weakest allele). Ctf18, an alternative replication factor C subunit, is one of the newly identified Aurora substrates. Deletion of the ctf18+ gene causes CPT and mild HU sensitivity (89). (C) Model showing well-characterized functions of Aurora on the left side and potential functions suggested by the range of identified substrates on the right side. The preferentially centromeric localization of Aurora during early mitosis is depicted by the red oval, and phosphorylation of kinetochore proteins, chromatin-associated proteins, and other nuclear proteins is symbolized by the small circles labeled with “P.”

Discussion

The identification of all cellular substrates of a kinase is crucial for understanding its molecular mechanism of action, but comprehensive kinase substrate identification is still a challenging task (67). The approach we have taken here is powerful because substrates of the kinase are identified in vivo in the native cellular context. Similar strategies have been used to identify substrates of the kinase Cdk1 (36) and to assess global perturbations of kinases and phosphatases in yeast (68).

At present, one limitation of this approach is the inability to capture the entire phosphoproteome in one experiment, and therefore, despite excellent coverage, some substrates or phosphorylation sites escape detection. We alleviated this effect by including a technical replicate for each experiment. It is difficult to say whether we missed any substrates, because other than histone H3, Cnd2, and Bir1 (Survivin) (40, 47, 69), each of which we identified, no fission yeast Aurora targets have been described. However, on the basis of work in other organisms, several conserved outer kinetochore proteins are likely phosphorylated by Aurora (6, 10, 12), and of these we identified only Spc7 (human KNL1, also known as Blinkin) (Fig. 3), suggesting that there are additional Aurora substrates not covered by our analysis.

The identification of a high number of unexpected targets raises the question of whether our analysis indeed specifically identified Aurora substrates. In particular for sites in class 1, we are, however, confident that their phosphorylation depends on Aurora, because (i) we have repeatedly detected these phosphorylation sites in cells with active Aurora, (ii) we have repeatedly found less of the phosphorylated peptide in Aurora-inhibited cells, and (iii) we did not find significant changes in the abundance of the phosphorylation of these sites when testing for nonspecific effects of the inhibitor (Fig. 1). It is still possible that some of these proteins are not direct Aurora substrates, but secondary targets. However, we can easily identify a common motif for most of these sites (Fig. 2), and most of the substrates that we tested (8 of 11) were phosphorylated by Aurora in vitro on the same sites identified in vivo (Fig. 3 and table S5). These data indicate that at least a subset of the identified sites that exhibited reduced phosphorylation upon fission yeast Aurora inhibition are directly targeted by Aurora. If we assume that a protein from class 1 or 2 is a direct substrate if it contains an [RK]-X-[ST] site that is less phosphorylated upon Aurora inhibition, then 12 of 13 proteins (92%) with at least one phosphorylation site strongly reduced (H/L <0.25) by Aurora inhibition and 21 of all 42 proteins in class 1 and 2 (50%) would be direct substrates (table S1).

The spectrum of candidate substrates that we identified suggests that Aurora influences chromatin during mitosis more broadly than previously thought (Fig. 5C). Given the generally high conservation in function between yeast Aurora and metazoan Aurora B, we speculate that such influence on chromatin may also occur in metazoans. Mitosis-specific modification of chromatin could serve a number of tasks. Changes to chromatin may be required to segregate chromosomes. Given the complex topology of chromatin and the many associated proteins and RNA molecules, it is generally surprising how the two DNA strands can be separated without causing damage. It has been proposed that proper chromosome segregation requires removal of many proteins and RNA from chromatin, because these would interfere with segregation (70). Aurora B–type kinases and condensin have been suggested to have a role in this “clearing” process (70). The Smc5-Smc6 complex is required for segregating chromosomes specifically after DNA damage, presumably by aiding in the removal of the cohesin complex from chromatin (71). The spectrum of Aurora substrates that we identified, which includes condensin and Smc6 (Fig. 3), is consistent with a potential role for Aurora in preparing chromosomes for segregation. Mitotic modification of chromatin may also help to prevent chromatin damage as a result of the condensation that occurs during mitosis (72), or it may modulate DNA damage signaling such that unperturbed mitotic progression is ensured (73). A role of Aurora in these processes is consistent with the increased DNA damage sensitivity of ark1 mutants (Fig. 5, B and C). Furthermore, mitotic modification of chromatin may affect subsequent phases of the cell cycle. The mitotic phosphorylation of histone H3 by Aurora has been proposed to facilitate transcription from heterochromatin in the subsequent S phase, which reinforces heterochromatin silencing by processing of these transcripts into small interfering RNA (26). Consistent with such a role, circumstantial evidence implicates mammalian Aurora B in maintaining epigenetic chromatin states in differentiated cells (74, 75). Thus, Aurora may have a role in heterochromatin inheritance (Fig. 5C). Because we found several proteins implicated in DNA replication as Aurora targets, Aurora may also contribute to the organization of the program of replication origin firing in S phase, which may be set up during mitosis (76, 77). Regardless of the precise consequences of the Aurora-dependent phosphorylation events, our data indicate that Aurora has previously unappreciated functions in regulating chromatin during mitosis and is involved in the resistance toward DNA damage. These additional functions should be considered in basic research as well as in the clinical application of Aurora inhibitors.

Materials and Methods

S. pombe growth

All S. pombe yeast strains are listed in the Supplementary Materials. For SILAC experiments, a yeast strain carrying a temperature-sensitive cdc25-22 mutation, the ark1-as3 allele, and plo1+-GFP was grown for 10 cell cycles at 25°C in Edinburgh minimal medium (EMM) supplemented with 12C614N2 l-lysine (Lys0, “light,” L) or 13C615N2 l-lysine (Lys8, “heavy,” H) until it reached a concentration of 8 × 106 cells/ml. Cells were arrested before mitosis by shifting them to 36°C for 4.5 hours. Heavy-labeled cells were treated for 20 min with 5 μM 1NM-PP1 (4-amino-1-tert-butyl-3-(1′-naphthylmethyl)pyrazolo[3,4-d]pyrimidine; Toronto Research Chemicals), and light-labeled cells were treated with the same amount of the solvent DMSO. In the second experiment, MBC (25 μg/ml; Sigma-Aldrich) was additionally added 10 min before the addition of 1NM-PP1 or DMSO. The third experiment was identical to the first except that a strain expressing ark1+ instead of ark1-as3 was used. Cells were released from the cdc25-22 arrest by reducing the temperature back to 25°C and were grown for 20 min at 25°C before harvesting. Cells were harvested by centrifugation (3000g, 5 min, 4°C). The resulting cell pellet was washed three times in ice-cold double-distilled H2O and finally resuspended in the residual amount of liquid left after decanting the supernatant. The suspension was dropped in liquid N2 to freeze cells as droplets. The droplets were stored at −80°C until further processing. Typically, 1.5 × 109 cells were used for each treatment (1NM-PP1 or DMSO) in one experiment.

Serial dilution growth tests were performed by growing cells to logarithmic growth phase in YEA (yeast extract containing adenine sulfate) medium and spotting a 1:5 serial dilution onto YEA plates, when indicated containing 10 μM camptothecin (CPT; Sigma-Aldrich), 4 mM hydroxyurea (HU; Sigma-Aldrich), or benomyl (8 μg/ml; Sigma-Aldrich).

Microscopy of S. pombe cells

Cells were fixed in −20°C methanol at the moment when freezing the rest of the culture in liquid N2. After washing once with PEM [100 mM Pipes (pH 6.9), 1 mM EGTA, and 1 mM MgSO4]/methanol and once with PEM, DNA was stained with DAPI (4′,6-diamidino-2-phenylindole, 1 μg/ml; Sigma-Aldrich). Images were acquired on a Zeiss Axio Imager microscope coupled to a charged-coupled device camera and were processed with MetaMorph software (Molecular Devices Corporation). Typically, a Z stack of about 3-μm thickness, with single planes spaced by 0.3 μm, was acquired and subsequently projected to a single image. Shown are color-combined pictures of maximum intensity projections of the GFP Z stack (green channel) with single plains of the DAPI Z stack (blue and red channel).

In vitro kinase assays

Full-length proteins or protein fragments (see table S6) were expressed in Escherichia coli strain BL21 and purified on Ni-NTA agarose beads (Qiagen). Ark1 and Pic1 were coexpressed in E. coli strain BL21 and similarly purified. Under these conditions, Pic1 copurified with His-Ark1. Proteins were stored in 50 mM tris-HCl (pH 7.5), 150 mM NaCl, and 20% glycerol at −80°C. For in vitro phosphorylation, 100 μg of substrate protein was incubated with 2 μg of Ark1 and Pic1 in a buffer containing 50 mM tris-HCl (pH 7.5), 100 mM NaCl, 20 mM MgCl2, 1 mM dithiothreitol, and 1 mM adenosine 5′-triphosphate in a total volume of 100 μl for 30 min at 30°C. The reaction was stopped by adding 400 μl of methanol, followed by chloroform-methanol precipitation of proteins. Precipitated proteins were resuspended in denaturation buffer (6 M urea, 2 M thiourea) and digested in solution by endoproteinase Lys-C (Waco) or trypsin (Promega) as described previously (78). About 20% of each digest was desalted on C18 Stage Tips (79) and analyzed directly by liquid chromatography–mass spectrometry (LC-MS), whereas the remainder of the digest was subjected to phosphopeptide enrichment by TiO2 chromatography (78) before LC-MS analysis.

Protein extraction

The frozen cell droplets of the heavy and light cultures were mixed according to their cell count in a 1:1 ratio and protein extracts were prepared with a ball mill cell grinder (Mixer Mill 400; Retsch) under cryogenic conditions. The cell powder was resuspended in denaturation buffer [6 M urea, 2 M thiourea, and 1% n-octyl glucoside (NOG)], and insoluble material was removed by centrifugation (13,000g, 10 min, 4°C). Both supernatant (#I) and pellet were further processed. The pellet was resuspended in denaturation buffer, 500-μm glass beads were added, and the mixture was subjected to 20-s shaking in a FastPrep machine (FP120, Qbiogene). The resulting suspension was centrifuged (13,000g, 10 min, 4°C), the pellet was discarded, and the protein concentration in the supernatant of this step (#II) and the supernatant #I was determined with a BCA (bicinchoninic acid) assay (Novagen). Both supernatants were further processed in parallel for the in-solution proteolysis and phosphopeptide enrichment. The procedure was identical for each of the three SILAC experiments and was always repeated once (technical replicate). In the replicate procedure, only supernatant #I was processed, because supernatant #II typically had a low protein concentration.

In-gel proteolysis for proteome analysis

For proteome analysis, 150 μg of protein extract was separated by 4 to 12% gradient SDS–polyacrylamide gel electrophoresis (SDS-PAGE) (Invitrogen). The resulting lanes were cut into 15 slices, which were subjected to in-gel digestion as previously described (80), but with endoproteinase Lys-C instead of trypsin. The resulting peptide mixtures were desalted with C18 Stage Tips (79) before LC-MS measurements.

In-solution proteolysis and phosphopeptide enrichment for phosphoproteome analysis

Six milligrams of total protein was further processed as described previously (78) with the following changes: In-solution digestion was performed with endoproteinase Lys-C; elution of bound peptides in the strong cation exchange (SCX) chromatography was done with a gradient of 0 to 50% SCX solvent B; the elution of bound phosphopeptides from the TiO2 beads was performed with 3 × 100 μl of a modified elution buffer (40% NH3, 60% ACN).

LC-MS/MS

All LC-MS analyses were performed on an EasyLC nano-HPLC (Proxeon Biosystems) coupled to an LTQ Orbitrap XL mass spectrometer (Thermo Scientific). Peptide separation was performed on a nano-HPLC column (75 μm by 15 cm) packed in-house with 3-μm C18 beads (Dr. Maisch). Peptides were loaded in a solvent containing 0.5% acetic acid and eluted with a segmented gradient of 5 to 60% of solvent containing 80% acetonitrile and 0.5% acetic acid.

The LTQ Orbitrap was operated in the positive ion mode. Precursor ions were recorded in the Orbitrap mass analyzer at a resolution of 60,000. Precursor ion fragmentation and acquisition were performed in the LTQ mass analyzer. For proteome analysis, the 10 most intense precursor ions were sequentially fragmented in each scan cycle, whereas for phosphoproteome analysis, the 5 most intense precursor ions were fragmented by activation of neutral loss ions at −98, −49, and −32.6 relative to the precursor ion (multistage fragmentation). In all measurements, sequenced precursor masses were excluded from further selection for 90 s. The target values for the LTQ were 5000 charges [tandem mass spectrometry (MS/MS)] and for the Orbitrap 106 charges (MS); the maximum allowed fill times were 150 and 500 ms, respectively.

MS data processing and analysis

The MS data were processed and quantified with MaxQuant (v1.0.14.3) (81). Generated peak lists were searched against a target-decoy database (82) consisting of the S. pombe proteome (http://www.sanger.ac.uk) and 262 frequently observed contaminants with the Mascot (Matrix Science) search engine (v.2.2.0). Carbamidomethylation of cysteine was set as fixed modification, whereas protein N-terminal acetylation, oxidation of methionine, and phosphorylation of serine, threonine, and tyrosine were set as variable modifications. Initial precursor mass tolerance was set to 7 parts per million (ppm) at the precursor ion and 0.5 dalton at the fragment ion level.

Identified MS/MS spectra were further processed by MaxQuant for statistical validation and quantitation of peptides, sites, and protein groups (83). False discovery rates (82) were set to 1% at phosphorylation site (P site), peptide, and protein group level. Downstream bioinformatics analysis was performed with R (v. 2.9.0; http://www.r-project.org). The interaction network in Fig. 5A and fig. S8C is based on the BioGRID database (84) and was visualized in Cytoscape 2.6.3 (85) with subsequent editing in Adobe Illustrator software. Information on DNA damage sensitivities and additional protein-protein interactions were taken from the PombePD [http://www.proteome.com (86)].

Determination of phosphorylation sites

All phosphorylation events having a reported localization probability of at least 0.75 were considered to be assigned to a specific residue, and we refer to these as phosphorylation sites. This means that the sum of all remaining potential phosphorylation site probabilities was at most 0.25. We used the intensity-weighted ratio significance values reported by MaxQuant to determine significantly changed phosphorylation sites. These values were further corrected for multiple hypothesis testing by the method proposed by Benjamini and Hochberg (87). For this data set, we report significant phosphorylation event ratios based on a Benjamini-Hochberg corrected P value threshold of 0.05.

Consensus motif analysis

We compared the occurrence of each amino acid at the −/+5 flanking positions of all class 1 and 2 sites exhibiting decreased phosphorylation to the corresponding frequencies among all identified and localized phosphorylation sites. To test whether the frequency of a particular amino acid at a specific flanking position was increased or decreased, we assumed the hypergeometric distribution as null distribution, and we derived P values by the application of Fisher’s exact test.

Gene Ontology term enrichment analysis

Gene Ontology (GO) term enrichment analysis was done with the BiNGO (88) plug-in for Cytoscape (85). GO annotation for S. pombe was downloaded from the GO consortium (http://www.geneontology.org/GO.downloads.annotations.shtml; submission date 04/23/2010). The hypergeometric test was applied, and derived P values were adjusted for multiple hypothesis testing by the method of Benjamini and Hochberg (87). GO terms with adjusted P values <0.01 were considered to be significantly enriched. For grouping the candidate substrates into functional classes (Fig. 3), we preferentially used the fission yeast GO slim annotation (http://amigo.geneontology.org/cgi-bin/amigo/slimmer; submission date 03/15/2010).

Supplementary Materials

www.sciencesignaling.org/cgi/content/full/4/179/rs6/DC1

Fig. S1. Comparison of phosphorylated proteins identified in this and two other large-scale studies of fission yeast.

Fig. S2. Comparison of phosphopeptide ratios before and after normalization for protein abundance.

Fig. S3. Fragmentation mass spectra of the histone H3 peptide carrying the known Aurora target site (Ser10).

Fig. S4. Consensus motif analysis of phosphorylation sites that depend on the kinase activity of fission yeast Aurora.

Fig. S5. Consensus motif analysis of previously identified yeast Aurora or metazoan Aurora B phosphorylation sites.

Fig. S6. Reanalysis of the frequency of motif occurrence by including the K-X-[ST] and excluding the doubly phosphorylated sites.

Fig. S7. GO term enrichment of proteins with class 1 and 2 phosphorylation sites.

Fig. S8. Genetic interaction tests of ark1 temperature-sensitive mutants and previously identified genetic interactions of fission yeast Aurora substrates.

Table S1. Summary of MS data on all identified phosphorylation events.

Table S2. Number of phosphorylation sites within structured or unstructured regions.

Table S3. List of previously known yeast Aurora or metazoan Aurora B phosphorylation sites.

Table S4. List of GO terms and their associated proteins from class 1 and 2.

Table S5. List of proteins with class 1 or 2 sites tested for phosphorylation by fission yeast Aurora in vitro.

Table S6. Recombinant proteins or protein fragments used for the in vitro kinase assays.

Experimental Procedures

References

References and Notes

  1. Acknowledgments: We thank Y. Watanabe, O. Weichenrieder, D. Weigel, and members of our laboratory for valuable comments on the manuscript; M. Mann, J. M. Peters, and A. Köhler for advice; S. A. MacNeill, Y. Hiraoka, P. Nurse, and the Yeast Genetic Resource Center for fission yeast strains; E. Illgen and S. Wahl for excellent technical support; and D. Steiner for help with graphics. Funding: Supported by the Max Planck Society, the Deutsche Forschungsgemeinschaft (SFB446 grant to A.K. and S.H.), and the Landesstiftung Baden-Württemberg (Juniorprofessorenprogramm grant to B.M.). Author contributions: A.K., B.M., and S.H. designed the experiments; A.K. performed all experiments except those in Fig. 5A and fig. S8 (done by S.H.); to prepare samples for MS, A.K. had help from S.P. and B.M.; S.P. and B.M. processed and analyzed the MS results; K.K. performed additional computational analysis; and S.H. and B.M. wrote the manuscript with input by A.K. and K.K. Competing interests: The authors declare that they have no competing interests. Data availability: The MS data associated with this manuscript can be downloaded from the ProteomeCommons.org Tranche network using the following hash: aDFbn22hraYj2G/L63ojoA0PN3p8JS/+7oHi17zlW0UUDFKerD9pXTauJ74RWqfr/BK7V5e0qzBOTOpjWk1WhlVn35EAAAAAAAAaHA==. The hash can be used to prove which files were published as part of this manuscript’s data set, and the hash can be used to check that the data have not changed since publication.
View Abstract

Navigate This Article