Review

Introduction

Upon translation, a eukaryotic protein is often destined for further processing and modification. The former includes proteolytic cleavage and peptide splicing. As for protein modification, over 200 forms have been identified, and equipped with genome and proteome sequences, high-sensitivity mass spectrometry is revealing novel ones (Banks et al., 2000; Nalivaeva and Turner, 2001; Mann and Jensen, 2003). Most eukaryotic proteins are modified in one form or the other. Stable modifications, such as disulfide formation, glycosylation, lipidation, biotinylation and lipoylation, are essential for maturation of newly synthesized proteins into proper structural and functional states. Many other covalent modifications are transient and play important roles in regulating protein function. Instead of single-site modification, various proteins are modified at multiple sites, a phenomenon referred to as multisite modification. The multiplicity of modification sites on a protein often correlates with its biological importance and the complexity of the corresponding organism. Here, I utilize selected proteins, including histones, RNA polymerase II, p53, Cdc25C and PDGF receptor-, to illustrate the complexity of multisite modification and discuss how a diverse array of protein modules recognizes specific covalent modifications to regulate protein function in an orchestrated fashion.

Multisite protein modification

It is well known that multisite phosphorylation serves as a common mechanism for regulating protein function in eukaryotes (Cohen, 2000; Holmberg et al., 2002). One extreme example is RPB1, the largest subunit of RNA polymerase II (Figure 1a). Dependent on the organism, the C-terminal domain (CTD) of RPB1 consists of 25–52 heptapeptide repeats with the consensus sequence YSPTSPS. Each repeat contains five phosphorylatable residues, with Ser2 and Ser5 known as predominant phosphorylation sites (Kobor and Greenblatt, 2002; Buratowski, 2003; Hampsey and Reinberg, 2003). Ser2 phosphorylation is seen in coding regions and coupled to 3'-RNA processing, whereas Ser5 phosphorylation is detected primarily at promoter regions and linked to RNA capping (Komarnitsky et al., 2000; Fabrega et al., 2003; Meinhart and Cramer, 2004). Interestingly, both serines are phosphorylated during the M phase of the cell cycle to inhibit RNA splicing and promote gene silencing (Xu et al., 2003).

Figure 1.

Multisite modification of representative proteins. (a) The CTD of RPB1 comprises heptapeptide repeats with the indicated consensus sequence. From yeast to humans, the number (n) of repeats increases from 26 to 52. In addition to phosphorylation, the CTD is subject to other modifications like proline isomerization and glycosylation. (b) p53 is composed of modular domains, including the N-terminal transcriptional activation domains (TA1 and TA2), DNA-binding domain (DBD), nuclear localization signal (NLS), tetramerization motif (TET) and C-terminal regulatory domain (REG). The number at the right refers to total residues of human p53. Modified residues are indicated by single letters along with their positions. p53 is ubiquitinated at its C-terminus (Brooks and Gu, 2003), but the sites are not so well defined. (c) Histone H3 comprises of a flexible N-terminal tail and a C-terminal histone-fold domain. Post-translational modifications are labeled with color letters: P in oval, phosphorylation; A in hexagon, acetylation; U in pentagon, ubiquitination; S in pentagon, sumoylation; M in square, methylation

Full figure and legend (103K)

As shown in Figure 1b, the tumor suppressor protein p53 is also modified at many residues, with about 10 phosphorylation sites clustered in its N-terminal part and several sites located at the C-terminal region. In addition to phosphorylation, p53 is modified by acetylation, sumoylation and ubiquitination. The modification status of p53 is dynamic and fluctuates in response to cellular signaling triggered by DNA damage, proliferation and senescence (Appella and Anderson, 2001; Brooks and Gu, 2003). Similarly, histones are subject to regulation by different modifications, which include acetylation, methylation, phosphorylation, ubiquitination, sumoylation, ADP-ribosylation and citrullination. As illustrated for histone H3 in Figure 1c, modifications of core histones are clustered at the N-terminal tails (Berger, 2002; Fischle et al., 2003a; Zhang, 2003; Khorasanizadeh, 2004). Furthermore, multisite modification occurs in many other eukaryotic proteins, including DNA-binding transcription factors (e.g. Pho4, NF-AT, c-Jun and Elk-1), enzymatic transcriptional coactivators (e.g. p300 and CBP), DNA replication regulators (e.g. PCNA), cell cycle controllers (e.g. cyclins and Cdc25 phosphatases), apoptosis regulators (e.g. BAD), cytoskeletal proteins (e.g. tubulins and neurofilament proteins), and signaling molecules (e.g. Raf-1, MEK1 and tyrosine kinases). Therefore, multisite modification is a common but complex mechanism for regulating protein function in eukaryotic cells.

Competitive same-site modifications

To add another level of complexity to multisite modification, one residue can be modified in many ways. Besides phosphorylation, serine and threonine residues are targets of glycosylation by O-GlcNAC (Figure 2a), raising a potential reciprocal relationship between these two modifications (Slawson and Hart, 2003). Similarly, the hydroxyl group of tyrosine residues can be phosphorylated or sulfated (Figure 2b) (Moore, 2003; Hunter and Eckhart, 2004). Moreover, nitration of the aromatic ring acidifies the hydroxyl group, rendering it unfavorable to phosphorylation (Reinehr et al., 2004). As illustrated in Figure 2c, methylation and ADP-ribosylation may compete to modify the side chains of glutamate, aspartate and arginine residues (Rouleau et al., 2004). For arginine methylation, up to two methyl groups can be added to the guanidino group, with dimethylation being symmetric or asymmetric (Zhang and Reinberg, 2001; Bannister et al., 2002). In addition, arginine residues are subject to citrullination (Vossenaar et al., 2003).

Figure 2.

Competitive same-site modifications. (a) Serine (S) or threonine (T) residues are subject to phosphorylation or modification by O-GlcNAC. (b) The hydroxyl group of tyrosine (Y) residues is the target of phosphorylation and sulfation, whereas the aromatic ring can be modified by nitration. (c) The carboxyl group of aspartate (D) and glutamate (E) residues is a target for ADP-ribosylation and methylation. (d) The lysine (K) side chain can be modified by hydroxylation, methylation, acetylation, biotinylation and addition of ubiquitin and ubiquitin-like protein modifiers. (e) A ubiquitin is linked to a target protein through an isopeptide bond, and the seven lysine residues of ubiquitin are potential targets for addition of polyubiquitin chains

Full figure and legend (20K)

Lysine modifications are even more complicated. Besides hydroxylation, lysine residues are subject to acetylation, methylation, ubiquitination, neddylation and sumoylation (Figure 2d). Like lipoic acid, biotin is the prosthetic group of several enzymes, but recent studies suggest that biotinylation also occurs in other proteins for dynamic regulation of protein function (Narang et al., 2004). Lysine methylation leads to addition of up to three methyl groups to the -amino group, and importantly, different methylation states have distinct functional consequences (Sims et al., 2003). For ubiquitination, mono- or poly-ubiquitin can be linked to lysine residues. These two different modifications exert distinct effects. In addition, competitive mono- and poly-ubiquitination have been documented (Li et al., 2003a). All seven lysine residues of ubiquitin are targets for polyubiquitin chain formation (Figure 2e) (Peng et al., 2003). Functional consequences of polyubiquitination are dependent on which lysine is used. For example, Lys48-linked polyubiquitin often labels proteins for degradation, whereas Lys29- or Lys63-linked ubiquitin chains are involved in other cellular functions like DNA repair and endocytosis (Pickart, 2001).

There are three SUMO proteins in mammals, SUMO1, SUMO2, SUMO3 and SUMO4 (Bohren et al., 2004; Guo et al., 2004). The latter two are highly homologous to each other, so SUMO1 modification may have effects distinct from those by addition of SUMO2, SUMO3 or SUMO4 (Muller et al., 2001). Unlike SUMO1, the other three SUMO proteins can form poly-SUMO chains. Interestingly, one SUMO1 molecule may be added to poly-SUMO chain formed by the other three SUMO proteins (Tatham et al., 2001). There is also evidence indicating that mono- and poly-sumoylation have distinct functional consequences (Li et al., 2003b).

Reminiscent of protein phosphorylation and glycosylation, different modifications create gridlock at lysine residues, with one modification excluding others (Freiman and Tjian, 2003). For example, sumoylation of IB blocks its ubiquitination and subsequent degradation (Desterro et al., 1998), while acetylation inhibits the ubiquitination of Smad7, SREBP and p53 (Gronroos et al., 2002; Li et al., 2002; Giandomenico et al., 2003), the sumoylation of Sp3 (Braun et al., 2001; Ross et al., 2002; Sapetschnig et al., 2002; Ammanamanchi et al., 2003) and the methylation of histone H3 at Lys9 (Jenuwein and Allis, 2001; Kouzarides, 2002; Fischle et al., 2003a). Competitive same-site modifications not only increase the diversity of multisite modification, but may also reduce leaky levels of undesired modifications to increase signaling sensitivity.

Histone code and others

From above, it is evident that multisite modification constitutes a complex layer of molecular information beyond the amino-acid sequence of a given protein, so an intriguing question is how such molecular information is conveyed to regulate its function. The 'histone code' hypothesis (Strahl and Allis, 2000; Turner, 2000,2002; Jenuwein and Allis, 2001; Fischle et al., 2003a), 'chromatin signaling' model (Schreiber and Bernstein, 2002) and 'histone interaction surface' model (Kurdistani et al., 2004) have been proposed to explain how different histone modifications regulate chromatin structure and function. In addition, hypotheses similar to that of 'histone code' have been put forward for modifications of p53 (Appella and Anderson, 2000,2001), RPB1 (Buratowski, 2003), p300/CBP (Gamble and Freedman, 2002; Legube and Trouche, 2003) and tubulins (Westermann and Weber, 2003). For cellular signaling, it is well known that multisite phosphorylation serves as an essential way for transducing molecular signals (Cohen, 2000; Holmberg et al., 2002). No matter which model or hypothesis is used, it is clear that modifications of a protein at multiple sites form a complex regulatory program for the qualitative and quantitative control of its function. Emerging evidence suggests that this program displays characteristics of a 'dynamic molecular fingerprint or barcode' (also see below).

Combinatorial characteristics of multisite modification

The functional consequence of a specific modification event could be context-dependent. For example, acetylation of histone H4 at Lys5 and Lys12 is required for chromatin assembly and promotes chromatin compaction and gene silencing (Kelly et al., 2000). However, when Lys8 and Lys16 are also acetylated, acetylation of Lys5 and Lys12 is linked to gene activation. Phosphorylation of histone H3 at Ser10 is linked to gene activation, whereas phosphorylation of both Ser10 and Ser28 marks condensed chromatin (Cheung et al., 2000). Combinatorial effects have also been shown for multisite modification of nonhistone proteins. For example, separate or combined phosphorylation at Ser2 and Ser5 of the RPB1 CTD (Figure 1a) directly couples transcription to different stages of RNA processing (Komarnitsky et al., 2000; Fabrega et al., 2003; Meinhart and Cramer, 2004).

Switch- and gauge-like effects of multisite modification

Consistent with the combinatorial characteristics, multiple modification events on a protein frequently interplay with each other and their functional consequences are often multifaceted. Modifications at different sites could be independent of each other, with each being sufficient to achieve the maximal output. In this case, each modification event serves as a simple 'on/off' switch. Alternatively, modifications at two or more sites synergize with each other to impart an exponential effect, thereby generating a combined switch. This is reminiscent of the countdown mechanism that has been proposed for multisite phosphorylation of the CDK inhibitor Sic1 (Orlicky et al., 2003). Modification events at different sites could also have additive effects, thereby producing a linear output and regulating protein function in a quantitative manner. Such a 'gauge-like' mode of action has been documented for histone acetylation and PDGF receptor phosphorylation (Schreiber and Bernstein, 2002). Therefore, multisite modification is important for coordinating the qualitative and quantitative control of protein function in vivo.

Modification-specific protein modules

Theoretically, a specific modification event could modulate protein function through 'loss-of-function' and/or 'gain-of-function' mechanisms (just like a genetic mutation) (Yang, 2004). As one of the latter, modifications create docking sites for protein modules, a prototype of which is the phosphotyrosine-specific SH2 domain (Pawson and Nash, 2003). As listed in Table 1, protein modules recognizing phosphoserine, phosphothreonine, acetyllysine, methyllysine, methylarginine, ubiquitin and SUMO have also been identified. Moreover, there are uncharacterized modules with potential ability to specifically bind modified residues (Maurer-Stroh et al., 2003). Both SH2 and PTB domains recognize phosphotyrosine-containing motifs and there are over 10 types of phosphoserine/threonine-specific modules (Table 1), suggesting that for each modification, there could be more than one type of modules. Modification-specific modules are selective and able to differentiate subtle differences in the modification status. For example, the chromodomain of HP1 preferentially binds to trimethylated Lys9 of histone H3 (Fischle et al., 2003b; Min et al., 2003; Khorasanizadeh, 2004), the Tudor domain selects symmetric dimethyl-arginine over the other two methylated forms (Sprangers et al., 2003), and different ubiquitin-binding domains can recognize distinct states of ubiquitination (Table 1).

Table 1 - Modification-specific recognition by protein modules.

Full table

Hypothetically, two different modification-specific modules could display identical binding specificity towards the same modified motif. Alternatively, the same modification-specific module may exist in different multiprotein complexes. Related to the latter scenario, Gcn5 acetyltransferase is a subunit of distinct multisubunit complexes (Carrozza et al., 2003). Since its bromodomain recognizes histone H4 acetylated at Lys16 (Owen et al., 2000), different complexes may impart distinct effects, suggesting that a specific modification event may be a subject of different interpretations.

Selectivity of modification-specific protein modules

The existence of diverse modification-specific modules raises the interesting question of how they decode the complex molecular information conveyed by multisite modification. As illustrated in Figure 3a, PDGFR- possesses multiple tyrosine residues in the cytoplasmic domain (Heldin and Westermark, 1999). Upon binding to PDGF, this receptor dimerizes and trans-phosphorylates these tyrosine residues, thereby recruiting specific SH2-containing signaling proteins and triggering multiple signaling pathways (Heldin and Westermark, 1999). In addition to the phosphotyrosine moiety, flanking sequences contribute to the binding specificity of the SH2 domains. Such a signaling theme holds true for many tyrosine kinases and signaling adaptors (Schlessinger, 2000; Pawson and Nash, 2003).

Figure 3.

Modification-specific recognition by protein modules. (a) PDGFR- is composed of an extracellular domain, a transmembrane motif (TM) and a cytoplasmic region. Phosphorylatable tyrosine (Y) residues are indicated along with their positions. Upon phosphorylation, these residues recruit SH2-containing proteins, including Src, PI-3 kinase (PI3K), GAP, SHP2 and PLC-1. The number at the right refers to the total residues of human PDGFR-. (b) Cdc25C has a bipartite domain organization containing the N-terminal regulatory region and C-terminal catalytic domain. Within the regulatory region, there is an NES. Phosphorylatable serine (S) and threonine (T) residues are illustrated along with their positions. Kinases, such as Cdc2, Chk1 and Plk1, phosphorylate Cdc25C and generate binding sites for phospho-specific binding modules of Pin1, Plk1 and 14-3-3

Full figure and legend (103K)

Cdc25C phosphatase is a major cell cycle regulator in mammals and has a bipartite domain organization (Figure 3b). Its C-terminal phosphatase domain dephosphorylates Cdc2 to promote G2–M transition (Hutchins and Clarke, 2004), whereas the N-terminal regulatory domain contains multiple phosphorylation sites. During mitosis, Cdc2 directly phosphorylates human Cdc25C at Thr48, Thr67, Ser122, Thr130 and Ser214 (Izumi and Maller, 1993; Strausfeld et al., 1994; Bulavin et al., 2003), thereby contributing to efficient retention of Cdc25C within the nucleus for Cdc2 activation (Bulavin et al., 2003; Margolis et al., 2003). While phosphorylation of Thr48, Thr67 or Ser214 recruits Pin1 through its WW domain (Hutchins and Clarke, 2004), Thr130 phosphorylation generates a docking site for the Polo boxes of Plk1 (Elia et al., 2003). Other kinases, such as Chk1, phosphorylate Ser216 of Cdc25C and promote specific binding by 14-3-3 proteins, thereby leading to the cytoplasmic retention of Cdc25C and cell cycle arrest at the G2-M checkpoint. Therefore, just like tyrosine phosphorylation of PDGFR-, multisite phosphorylation of Cdc25C recruits specific protein modules.

Does this principle apply to other proteins? This appears to be the case at least for RPB1, p53 and histone H3. As discussed above (Figure 1a), the CTD of RPB1 is subject to dynamic multisite phosphorylation. Ser2 phosphorylation creates a specific binding site for the CTD-interacting domains (CIDs; Table 1) present in several 3'-RNA processing factors (Meinhart and Cramer, 2004) and perhaps also for the WW domain of Set2, a lysine methyltransferase that modifies histone H3 at Lys36 (Hampsey and Reinberg, 2003; Krogan et al., 2003). Phosphorylation of the CTD at Ser5 (Figure 1a) recruits mRNA capping factors such as RNA guanylyltransferase (Fabrega et al., 2003), whereas phosphorylation at both Ser2 and Ser5 can be recognized by the WW domain of Pin1, a peptidyl–prolyl cis–trans isomerase (Verdecia et al., 2000; Xu et al., 2003). Histone H3 is heavily modified at multiple sites (Figure 1c). The chromodomains of HP1 and Polycomb recognize trimethylated Lys9 and Lys27, respectively (Fischle et al., 2003b; Min et al., 2003), whereas Lys36 methylation may recruit the ATP-dependent chromatin remodeler Chd1 (chromodomain protein 1) through its chromodomain (Krogan et al., 2003). In addition, acetylation of histone H3 at Lys9 recruits bromodomain-containing proteins (Marmorstein, 2001; Zeng and Zhou, 2002). As for p53 (Figure 1b), phosphorylation at Ser33, Thr81 and Ser315 stimulates the interaction with the WW domain of Pin1 (Zacchi et al., 2002; Zheng et al., 2002), whereas acetylation at Lys382 generates a high-affinity site for the bromodomain of CBP (Mujtaba et al., 2004). In addition, phosphorylation at Ser315 creates a consensus binding site for the Polo boxes of Plks (Elia et al., 2003). Moreover, acetylation of Lys320 and sumoylation of Lys386 may generate specific docking sites for bromodomains and SIM proteins, respectively (Table 1). It is also possible that multisite monoubiquitination of p53 recruits proteins with ubiquitin-binding modules (Table 1) (Li et al., 2003a). Consistent with this possibility, multisite monoubiquitination of tyrosine kinase receptors has been proposed to create specific binding sites for ubiquitin-binding domains (Haglund et al., 2003). Therefore, although further investigation is needed to substantiate these speculations, recognition by protein modules has emerged as a common theme to decode the molecular information conveyed by multisite modification.

Avidity in modification-specific recognition by protein modules

A modified motif may recruit a protein module, but this recruitment may not be solely responsible for interaction between two proteins. First, a second site is frequently involved (Figure 4a). For example, CBP recognizes acetylated Lys382 and interacts with PxxP motifs of p53 (Dornan et al., 2003; Mujtaba et al., 2004). Second, modification of the same protein at multiple sites may recruit a protein with multiple modules (Figure 4b), as evidenced by TAF1 and Brd2, each of which possesses double bromodomains and binds to partners with dual acetyllysines at high affinity (Jacobson et al., 2000; Kanno et al., 2004). In addition, many signaling adaptors contain two and more SH2 domains (Pawson and Nash, 2003; Schlessinger and Lemmon, 2003), Polybromo possesses six bromodomains (Xue et al., 2000), and many BRCT proteins have multiple BRCT repeats (Manke et al., 2003; Yu et al., 2003). Third, besides covalent linkage, different modules are also brought together by complex formation (Figure 4c). For example, each 14-3-3 molecule has one site for phosphoserine/threonine recognition, and dimerization generates dual binding sites (Yaffe et al., 1997). Reminiscent of this, yeast and mammalian chromatin-remodeling complexes possess different subunits with bromodomains (Carrozza et al., 2003), so these subunits may cooperate to recognize targets with multiple acetyllysines. Fourth, one module can contain more than one site for modification recognition (Figure 4d). For example, more than one phosphate are needed for optimal recognition by the MH2 domain of Smad2 (Wu et al., 2001) or the WD40 domain of -TrCP (Wu et al., 2003). Finally, modifications at different sites on a protein may cooperate to interact with one binding pocket in an allovalent manner (Figure 4e), as demonstrated for the recognition of six phosphorylation sites on the CDK inhibitor Sic1 by the SCF^Cdc4 ubiquitin ligase (Orlicky et al., 2003).

Figure 4.

Cooperativity in modification-specific recognition. (a, b) Module A recognizes a modification (labeled with the letter X in an oval) at site 1 and this specificity is further increased by docking of module B at site 2. The two modules are covalently linked. If site 2 is modified (labeled with the letter Y in a hexagon, as in (b)), the two modifications, X and Y, can be of the same type or different ones. This model also applies to situations where one target contains more than two modification sites. (c) Two modules associate with each other through dimerization and bind to double modification sites on a target. (d) One module contains two binding pockets for double modification sites on a target protein. (e) The module is monovalent, but the target contains more than one modification site. These sites may independently recruit multiple molecules with the same module. Alternatively, these sites may act in an allovalent manner (Orlicky et al., 2003)

Full figure and legend (83K)

Short-range intramolecular signaling

To achieve avidity or exert antagonistic effects, different modification sites may cluster within a small motif of a given protein. Such physical proximity facilitates short-range interactions among different modification events (Figure 5). An interplay between two modifications sites on the same protein represents an 'intramolecular signaling' event. Such events have been well illustrated for histone modifications (Figure 5a). The chromodomain of HP1 interacts with methylated Lys9 of histone H3, and phosphorylation of Ser10 blocks this interaction (Berger, 2002; Fischle et al., 2003a). Tandem bromodomains of TAF1 bind to histone H3 diacetylated at Lys9 and Lys14 (Jacobson et al., 2000), and Ser10 phosphorylation stimulates Lys14 acetylation and TFIID association (Clements et al., 2003). As illustrated in Figure 5b–e, short-range intramolecular signaling also regulates the interaction between Cdc25C and 14-3-3 proteins (Bulavin et al., 2003), BAD and 14-3-3 proteins (Konishi et al., 2002), CREB and CBP (Lu et al., 2003), and STAT1 and its SH2 domain (Rogers et al., 2003).

Figure 5.

Agonistic and antagonistic effects of neighboring modifications. (a) The amino-acid sequence of residues 9–14 of histone H3 (Figure 1c) is shown along with modifications. Lys9 methylation recruits HP1 via its chromodomain (chromo) and phosphorylation of Ser10 interferes with this recruitment (illustrated with an arrow marked by a red minus sign in an oval). Tandem bromodomains (bromo) of TAF1 bind to histone H3 acetylated at Lys9 and Lys14, whereas Ser10 phosphorylation stimulates Lys14 acetylation (illustrated with an arrow marked by a green plus minus sign in an oval). (b) Ser216 phosphorylation of Cdc25C recruits 14-3-3 proteins, whereas Ser214 phosphorylation prevents Ser216 phosphorylation and/or interferes with 14-3-3 binding to Ser216 (illustrated with an arrow marked by a red minus sign in an oval). (c) Ser136 phosphorylation of the proapoptotic protein BAD creates a binding site for 14-3-3 proteins, whereas Ser128 phosphorylation inhibits 14-3-3 recruitment. (d) Ser133 phosphorylation of CREB generates a binding site for the KIX domain of CBP or p300, which acetylates Lys136 and strengthens KIX binding. (e) Tyr701 phosphorylation of STAT1 promotes dimerization through its SH2 domain. Lys703 resides within a consensus sumoylation site and its sumoylation may affect Tyr701 phosphorylation and/or SH2 binding. (f) Phosphorylation of histone H2A.X at Ser139 creates a docking site for BRCT domains of PTIP. It is unclear if ubiquitination of Lys119 affects PTIP association. (g) p53 acetylation at Lys382 generates a binding site for the bromodomain of CBP. It remains to be determined whether Ser392 phosphorylation stimulates Lys386 sumoylation and if the sumoylation facilitates Lys382 acetylation and subsequent association with CBP

Full figure and legend (89K)

This generalization leads to several testable predictions. In response to DNA repair, the histone variant H2A.X is specifically phosphorylated at Ser139 (Rogakou et al., 1998). Recent studies indicate that BRCT domains of the Pax transactivation-domain interacting protein PTIP recognize this phosphorylation (Manke et al., 2003; Yu et al., 2003). A nearby residue, Lys119, is ubiquitinated (Jason et al., 2002), raising the interesting possibility that the ubiquitination affects the BRCT association (Figure 5f). The bromodomain of CBP recognizes Lys382-acetylated p53 (Figure 5g) (Mujtaba et al., 2004). Adjacent residues, Lys386 and Ser392, are known sumoylation and phosphorylation sites, respectively. By analogy to the interplay between interdependent phosphorylation and sumoylation of HSF1 (Hietakangas et al., 2003), Ser392 phosphorylation may stimulate Lys386 sumoylation. Since PML association is required for Lys382 acetylation (Pearson et al., 2000), the sumolyation may in turn affect the acetylation and subsequent CBP association.

Long-range intramolecular signaling

In addition to neighboring effects, various modification sites may spread in different parts of a protein and thus promote long-range interactions. As illustrated in Figure 6a, Cdc2-mediated Thr67 phosphorylation of Cdc25C stimulates Ser216 dephosphorylation and subsequent dissociation of 14-3-3 proteins, thereby serving as a positive feedback for the nuclear retention of Cdc2 (Margolis et al., 2003). Phosphorylation of Cdc25C at Thr130 by Cdc2 generate a docking site for Plk1 and recruit it for Ser198 phosphorylation (Figure 6a) (Toyoshima-Morimoto et al., 2002). Ser198 is located within the NES of Cdc25C, so the phosphorylation blocks CRM1 association and nuclear export of Cdc25C.

Figure 6.

Regulation by long-range intramolecular signaling. (a) In response to mitogenic signals, Cdc25C dephosphorylates and activates Cdc2 to promote G2–M transition. Thr67 phosphorylation promotes the dephosphorylation at Ser216, whereas Thr130 phosphorylation recruits Plk1 through its Polo-box domain. Once recruited, Plk1 phosphorylates Ser198 to inhibit the function of the NES. (b) DNA damage activates Cdc2 to phosphorylate Ser315 of p53 (Figure 1b), which generates a potential docking site for the Polo-box domain of Plks and facilitates Ser20 phosphorylation of p53. This modification then stimulates CBP association for Lys382 acetylation. Similarly, Ser46 phosphorylation stimulates the same acetylation

Full figure and legend (91K)

As shown in Figure 6b, DNA damage activates Cdc2 to phosphorylate Ser315 of p53, which creates a consensus binding site for the Polo boxes of Plk3 and may thus recruit it for Ser20 phosphorylation (Xie et al., 2001; Elia et al., 2003). Like Ser46 phosphorylation, Ser20 phosphorylation strengthens the association with p300 and CBP (Appella and Anderson, 2001), leading to p53 acetylation in its C-terminal regulatory domain. Hypoxia, oxidative stress and other cellular insults may induce p53 acetylation in a similar manner (Ito et al., 2001). In addition, oncogenic Ras induces PML expression and relocalizes p53 and CBP to PML bodies, thereby leading to Lys382 acetylation and p53-dependent senescence (Pearson et al., 2000). PML-IV, a PML isoform derived from alternative splicing, promotes p53 phosphorylation at Ser46 by HIPK-2 and acetylation at Lys382 to stabilize p53 for transcriptional activation (Bischof et al., 2002). In addition to Cdc25C and p53, long-range intramolecular signaling has been shown for other proteins, such as STAT1 (Zhu et al., 2002). Thus, like short-range interplay, long-range intramolecular signaling is involved in coordinating different modification events of a given protein.

Concluding remarks

Multisite modification of a protein constitutes an extra layer of molecular information beyond the amino-acid sequence and forms a combined and coordinated regulatory program (Figure 1). This regulatory program is dynamic, which not only refers to the reversibility of many modifications but also emphasizes the kinetics, that is, duration and order of appearance of different modification events. As a 'gain-of-function' mechanism, covalent modifications may recruit a diverse array of modules, including the SH2 domain, 14-3-3 protein, WW domain, Polo box, BRCT repeat, bromodomain, chromodomain, Tudor domain and motifs capable of binding to ubiquitin and related protein modifiers (Table 1). Mechanistic impact of a specific modification event could be multifaceted, as supported by the findings that methylation of histone H3 at Lys9 blocks acetylation at this residue and creates a binding site for the chromodomain of HP1 (Figure 5a) and that acetylation of p53 at Lys382 enhances DNA binding, blocks ubiquitination and specifically interacts with the bromodomain of CBP (Figure 5g). Multisite modification promotes agonistic and antagonistic interplay among different modification sites, as reflected by combinatorial action and intramolecular signaling.

An important issue is how such principles derived from selected examples apply to other proteins. To fully address this issue, it will be necessary: (i) to catalogue multisite modification in various proteins; (ii) to identify the responsible enzymes and establish how they add and remove different modifications; (iii) to determine how modification-specific modules may recognize the modifications (Table 1); and (iv) to elucidate how different modification sites interplay with each other to coordinate the spatiotemporal regulation of protein function in vivo. Since multisite modification plays fundamental roles in regulating protein functions, its comprehension will not only provide important insights into various normal biological processes, but will also shed good light on the pathogenesis of different human diseases such as cancer.

References

1. Ammanamanchi S, Freeman JW & Brattain MG. (2003) J. Biol. Chem. 278: 35775−35780. | Article | PubMed | ChemPort |
2. Appella E & Anderson CW. (2000) Pathol. Biol. (Paris) 48: 227−245. | PubMed | ChemPort |
3. Appella E & Anderson CW. (2001) Eur. J. Biochem. 268: 2764−2772. | Article | PubMed | ISI | ChemPort |
4. Banks RE, Dunn MJ, Hochstrasser DF, Sanchez JC, Blackstock W, Pappin DJ & Selby PJ. (2000) Lancet 356: 1749−1756. | Article | PubMed | ISI | ChemPort |
5. Bannister AJ, Schneider R & Kouzarides T. (2002) Cell 109: 801−806. | Article | PubMed | ISI | ChemPort |
6. Berger SL. (2002) Curr. Opin. Genet. Dev. 12: 142−148. | Article | PubMed | ISI | ChemPort |
7. Bischof O, Kirsh O, Pearson M, Itahana K, Pelicci PG & Dejean A. (2002) EMBO J. 21: 3358−3369. | Article | PubMed | ChemPort |
8. Bohren KM, Nadkarni V, Song JH, Gabbay KH & Owerbach D. (2004) J. Biol. Chem. 279: 27233−27238. | Article | PubMed | ChemPort |
9. Braun H, Koop R, Ertmer A, Nacht S & Suske G. (2001) Nucleic Acids Res. 29: 4994−5000. | Article | PubMed | ISI | ChemPort |
10. Brehm A, Tufteland KR, Aasland R & Becker PB. (2004) Bioessays 26: 133−140. | Article | PubMed | ISI | ChemPort |
11. Brooks CL & Gu W. (2003) Curr. Opin. Cell Biol. 15: 164−171. | Article | PubMed | ISI | ChemPort |
12. Bulavin DV, Higashimoto Y, Demidenko ZN, Meek S, Graves P, Phillips C, Zhao H, Moody SA, Appella E, Piwnica-Worms H & Fornace AJ, Jr. (2003) Nat. Cell Biol. 5: 545−551. | Article | PubMed | ISI | ChemPort |
13. Buratowski S. (2003) Nat. Struct. Biol. 10: 679−680. | Article | PubMed | ISI | ChemPort |
14. Carrozza MJ, Utley RT, Workman JL & Cote J. (2003) Trends Genet. 19: 321−329. | Article | PubMed | ISI | ChemPort |
15. Cheung P, Tanner KG, Cheung WL, Sassone-Corsi P, Denu JM & Allis CD. (2000) Mol. Cell 5: 905−915. | Article | PubMed | ISI | ChemPort |
16. Clements A, Poux AN, Lo WS, Pillus L, Berger SL & Marmorstein R. (2003) Mol. Cell 12: 461−473. | Article | PubMed | ISI | ChemPort |
17. Cohen P. (2000) Trends Biochem. Sci. 25: 596−601. | Article | PubMed | ISI | ChemPort |
18. Desterro JM, Rodriguez MS & Hay RT. (1998) Mol. Cell 2: 233−239. | Article | PubMed | ISI | ChemPort |
19. Di Fiore PP, Polo S & Hofmann K. (2003) Nat. Rev. Mol. Cell Biol. 4: 491−497. | Article | PubMed | ISI | ChemPort |
20. Dornan D, Shimizu H, Burch L, Smith AJ & Hupp TR. (2003) Mol. Cell. Biol. 23: 8846−8861. | Article | PubMed | ChemPort |
21. Durocher D & Jackson SP. (2002) FEBS Lett. 513: 58−66. | Article | PubMed | ISI | ChemPort |
22. Elia AE, Cantley LC & Yaffe MB. (2003) Science 299: 1228−1231. | Article | PubMed | ISI | ChemPort |
23. Fabrega C, Shen V, Shuman S & Lima CD. (2003) Mol. Cell 11: 1549−1561. | Article | PubMed | ISI | ChemPort |
24. Fischle W, Wang Y & Allis CD. (2003a) Nature 425: 475−479. | Article | PubMed | ISI | ChemPort |
25. Fischle W, Wang Y, Jacobs SA, Kim Y, Allis CD & Khorasanizadeh S. (2003b) Genes Dev. 17: 1870−1881. | Article | PubMed | ISI | ChemPort |
26. Freiman RN & Tjian R. (2003) Cell 112: 11−17. | Article | PubMed | ISI | ChemPort |
27. Gamble MJ & Freedman LP. (2002) Trends Biochem. Sci. 27: 165−167. | Article | PubMed | ISI | ChemPort |
28. Giandomenico V, Simonsson M, Gronroos E & Ericsson J. (2003) Mol. Cell. Biol. 23: 2587−2599. | Article | PubMed | ChemPort |
29. Gronroos E, Hellman U, Heldin CH & Ericsson J. (2002) Mol. Cell 10: 483−493. | Article | PubMed | ISI | ChemPort |
30. Guo D, Li M, Zhang Y, Yang P, Eckenrode S, Hopkins D, Zheng W, Purohit S, Podolsky RH, Muir A, Wang J, Dong Z, Brusko T, Atkinson M, Pozzilli P, Zeidler A, Raffel LJ, Jacob CO, Park Y, Serrano-Rios M, Larrad MT, Zhang Z, Garchon HJ, Bach JF, Rotter JI, She JX & Wang CY. (2004) Nat. Genet. 36: 837−841. | Article | PubMed | ChemPort |
31. Gurevich VV & Gurevich EV. (2003) Structure (Camb.) 11: 1037−1042. | PubMed | ChemPort |
32. Haglund K, Di Fiore PP & Dikic I. (2003) Trends Biochem. Sci. 28: 598−603. | Article | PubMed | ISI | ChemPort |
33. Hampsey M & Reinberg D. (2003) Cell 113: 429−432. | Article | PubMed | ISI | ChemPort |
34. Heldin CH & Westermark B. (1999) Physiol. Rev. 79: 1283−1316. | PubMed | ISI | ChemPort |
35. Hietakangas V, Ahlskog JK, Jakobsson AM, Hellesuo M, Sahlberg NM, Holmberg CI, Mikhailov A, Palvimo JJ, Pirkkala L & Sistonen L. (2003) Mol. Cell Biol. 23: 2953−2968. | Article | PubMed | ISI | ChemPort |
36. Holmberg CI, Tran SE, Eriksson JE & Sistonen L. (2002) Trends Biochem. Sci. 27: 619−627. | Article | PubMed | ISI | ChemPort |
37. Hunter T & Eckhart W. (2004) Cell 116: S35−S39. | PubMed | ChemPort |
38. Hutchins JR & Clarke P. (2004) Cell Cycle 3: 41−45. | PubMed | ChemPort |
39. Ito A, Lai CH, Zhao X, Saito S, Hamilton MH, Appella E & Yao TP. (2001) EMBO J. 20: 1331−1340. | Article | PubMed | ChemPort |
40. Izumi T & Maller JL. (1993) Mol. Biol. Cell. 4: 1337−1350. | PubMed | ISI | ChemPort |
41. Jacobson RH, Ladurner AG, King DS & Tjian R. (2000) Science 288: 1422−1425. | Article | PubMed | ISI | ChemPort |
42. Jason LJ, Moore SC, Lewis JD, Lindsey G & Ausio J. (2002) Bioessays 24: 166−174. | Article | PubMed | ChemPort |
43. Jenuwein T. (2001) Trends Cell Biol. 11: 266−273. | Article | PubMed | ISI | ChemPort |
44. Jenuwein T & Allis CD. (2001) Science 293: 1074−1080. | Article | PubMed | ISI | ChemPort |
45. Kanno T, Kanno Y, Siegel RM, Jang MK, Lenardo MJ & Ozato K. (2004) Mol. Cell 13: 33−43. | Article | PubMed | ISI | ChemPort |
46. Kelly TJ, Qin S, Gottschling DE & Parthun MR. (2000) Mol. Cell. Biol. 20: 7051−7058. | Article | PubMed | ChemPort |
47. Khorasanizadeh S. (2004) Cell 116: 259−272. | Article | PubMed | ISI | ChemPort |
48. Kobor MS & Greenblatt J. (2002) Biochim. Biophys. Acta 1577: 261−275. | Article | PubMed | ISI | ChemPort |
49. Komarnitsky P, Cho EJ & Buratowski S. (2000) Genes Dev. 14: 2452−2460. | Article | PubMed | ISI | ChemPort |
50. Konishi Y, Lehtinen M, Donovan N & Bonni A. (2002) Mol. Cell 9: 1005−1016. | Article | PubMed | ISI | ChemPort |
51. Kouzarides T. (2002) Curr. Opin. Genet. Dev. 12: 198−209. | Article | PubMed | ISI | ChemPort |
52. Krogan NJ, Kim M, Tong A, Golshani A, Cagney G, Canadien V, Richards DP, Beattie BK, Emili A, Boone C, Shilatifard A, Buratowski S & Greenblatt J. (2003) Mol. Cell Biol. 23: 4207−4218. | Article | PubMed | ISI | ChemPort |
53. Kurdistani SK, Tavazoie S & Grunstein M. (2004) Cell 117: 721−733. | Article | PubMed | ISI | ChemPort |
54. Legube G & Trouche D. (2003) EMBO Rep. 4: 944−947. | Article | PubMed | ChemPort |
55. Li M, Brooks CL, Wu-Baer F, Chen D, Baer R & Gu W. (2003a) Science 302: 1972−1975. | Article | PubMed | ISI | ChemPort |
56. Li M, Luo J, Brooks CL & Gu W. (2002) J. Biol. Chem. 277: 50607−50611. | Article | PubMed | ISI | ChemPort |
57. Li Y, Wang H, Wang S, Quon D, Liu YW & Cordell B. (2003b) Proc. Natl. Acad. Sci. USA 100: 259−264. | Article | PubMed | ChemPort |
58. Lu Q, Hutchins AE, Doyle CM, Lundblad JR & Kwok RP. (2003) J. Biol. Chem. 278: 15727−15734. | Article | PubMed | ChemPort |
59. Manke IA, Lowery DM, Nguyen A & Yaffe MB. (2003) Science 302: 636−639. | Article | PubMed | ISI | ChemPort |
60. Mann M & Jensen ON. (2003) Nat. Biotechnol. 21: 255−261. | Article | PubMed | ChemPort |
61. Margolis SS, Walsh S, Weiser DC, Yoshida M, Shenolikar S & Kornbluth S. (2003) EMBO J. 22: 5734−5745. | Article | PubMed | ChemPort |
62. Marmorstein R. (2001) Nat. Rev. Mol. Cell. Biol. 2: 422−432. | Article | PubMed | ISI | ChemPort |
63. Maurer-Stroh S, Dickens NJ, Hughes-Davies L, Kouzarides T, Eisenhaber F & Ponting CP. (2003) Trends Biochem. Sci. 28: 69−74. | Article | PubMed | ISI | ChemPort |
64. Meinhart A & Cramer P. (2004) Nature 430: 223−226. | Article | PubMed | ChemPort |
65. Min J, Zhang Y & Xu RM. (2003) Genes Dev. 17: 1823−1828. | Article | PubMed | ISI | ChemPort |
66. Minty A, Dumont X, Kaghad M & Caput D. (2000) J. Biol. Chem. 275: 36316−36323. | Article | PubMed | ISI | ChemPort |
67. Moore KL. (2003) J. Biol. Chem. 278: 24243−24246. | Article | PubMed | ISI | ChemPort |
68. Mujtaba S, He Y, Zeng L, Yan S, Plotnikova O, Sachchidanand Sanchez R, Zeleznik-Le NJ, Ronai Z & Zhou MM. (2004) Mol. Cell 13: 251−263. | Article | PubMed | ISI | ChemPort |
69. Muller S, Hoege C, Pyrowolakis G & Jentsch S. (2001) Nat. Rev. Mol. Cell Biol. 2: 202−210. | Article | PubMed | ISI | ChemPort |
70. Nalivaeva NN & Turner AJ. (2001) Proteomics 1: 735−747. | Article | PubMed | ChemPort |
71. Narang MA, Dumas R, Ayer LM & Gravel RA. (2004) Hum. Mol. Genet. 13: 15−23. | Article | PubMed | ChemPort |
72. Orlicky S, Tang X, Willems A, Tyers M & Sicheri F. (2003) Cell 112: 243−256. | Article | PubMed | ISI | ChemPort |
73. Owen DJ, Ornaghi P, Yang JC, Lowe N, Evans PR, Ballario P, Neuhaus D, Filetici P & Travers AA. (2000) EMBO J. 19: 6141−6149. | Article | PubMed | ChemPort |
74. Pawson T & Nash P. (2003) Science 300: 445−452. | Article | PubMed | ISI | ChemPort |
75. Pearson M, Carbone R, Sebastiani C, Cioce M, Fagioli M, Saito S, Higashimoto Y, Appella E, Minucci S, Pandolfi PP & Pelicci PG. (2000) Nature 406: 207−210. | Article | PubMed | ISI | ChemPort |
76. Peng J, Schwartz D, Elias JE, Thoreen CC, Cheng D, Marsischky G, Roelofs J, Finley D & Gygi SP. (2003) Nat. Biotechnol. 21: 921−926. | Article | PubMed | ISI | ChemPort |
77. Pickart CM. (2001) Annu. Rev. Biochem. 70: 503−533. | Article | PubMed | ISI | ChemPort |
78. Reinehr R, Gorg B, Hongen A & Haussinger D. (2004) J. Biol. Chem. 279: 10364−10373. | Article | PubMed | ChemPort |
79. Rogakou EP, Pilch DR, Orr AH, Ivanova VS & Bonner WM. (1998) J. Biol. Chem. 273: 5858−5868. | Article | PubMed |