Function and regulation of protein neddylation

Gwénaël Rabut, Matthias Peter

Author Affiliations

  1. Gwénaël Rabut*,1 and
  2. Matthias Peter*,1
  1. 1 Swiss Federal Institute of Technology Zürich (ETH), Institute of Biochemistry, 8093, Zürich, Switzerland
  1. *Tel: +41 44 633 4587; Fax: +41 44 632 1269; E‐mail: gwenael.rabut{at} or Tel: +41 44 633 6586; Fax: +41 44 632 1269; E‐mail: matthias.peter{at}


Neddylation is the post‐translational protein modification that is most closely related to ubiquitination. However, ubiquitination is known to regulate a myriad of processes in eukaryotic cells, whereas only a limited number of neddylation substrates have been described to date. Here, we review the principles of protein neddylation and highlight the mechanisms that ensure the specificity of neddylation over ubiquitination. As numerous neddylation substrates probably remain to be discovered, we propose some criteria that could be used as guidelines for the characterization of neddylated proteins.

See Glossary for abbreviations used in this article.


Ubiquitination is a prominent post‐translational protein modification that regulates most cellular functions in eukaryotes (reviewed in Glickman & Ciechanover, 2002; Haglund & Dikic, 2005; Weissman, 2001). It consists of the covalent attachment of the small protein ubiquitin to target proteins, thereby modifying their biochemical properties and protein partners (Hicke et al, 2005). Besides ubiquitination, several related protein‐modification systems function in eukaryotes (Kerscher et al, 2006; Kirkin & Dikic, 2007). In particular, the NEDD8 protein—also known as Rub1 in Saccharomyces cerevisiae—is the closest relative to ubiquitin and can similarly be conjugated to substrate proteins in a process known as neddylation. Here, we review the characteristic features of protein neddylation and deneddylation. Moreover, we summarize the currently proposed NEDD8 substrates and suggest some criteria that might help to identify new neddylated proteins (Sidebar A).

Sidebar A | Criteria for the characterization of neddylation substrates

Minimal criteria for the identification of genuine NEDD8 substrates

  1. NEDD8 is covalently attached to the substrate in vivo

  2. Neddylation occurs under endogenous conditions in vivo

  3. Neddylation depends on specific components of the neddylation machinery in vivo

Criteria for further characterization of NEDD8 substrates

  1. Determination of the neddylated lysine residues

  2. Characterization of the phenotype of non‐neddylatable mutants

  3. Identification of a specific NEDD8 ligase in vivo

  4. Reconstitution of the neddylation reaction in vitro

  5. Identification of a NEDD8 isopeptidase in vivo

Conservation and functional importance of neddylation

NEDD8 was initially identified as a gene that is highly expressed in the embryonic mouse brain (Kumar et al, 1992). It was, however, soon realized that NEDD8 is highly conserved in most eukaryotes (plants, slime molds, fungi and animals; Burroughs et al, 2007; Kumar et al, 1993; Rao‐Naik et al, 1998) where it is expressed in most, if not all, tissues (Carrabino et al, 2004; Hori et al, 1999; Kumar et al, 1993; Rao‐Naik et al, 1998), suggesting an important function of NEDD8 in eukaryotic cells. Indeed, neddylation is essential for the viability of most model organisms, including Schizosaccharomyces pombe, Caenorhabditis elegans, Drosophila, Arabidopsis and mouse (Dharmasiri et al, 2003; Jones & Candido, 2000; Kurz et al, 2002; Osaka et al, 2000; Ou et al, 2002; Tateishi et al, 2001), with the notable exception of S. cerevisiae (Lammer et al, 1998; Liakopoulos et al, 1998). In addition, deregulated neddylation might be involved in the aetiology of some human diseases such as neurodegenerative disorders (Dil Kuazi et al, 2003; Mori et al, 2005) and cancers (Chairatvit & Ngamkitidechakul, 2007; Salon et al, 2007). Indeed, MLN4924, which is a general inhibitor of neddylation (Langston et al, 2007), shows substantial activity in a broad range of preclinical tumour models, raising the possibility that components of the neddylation pathway might be promising therapeutic targets. It is therefore important to characterize which cellular proteins are neddylated, how this modification affects their function, and how neddylation is catalysed and regulated (Sidebar B).

Sidebar B | In need of answers

  1. Which proteins are neddylated in vivo and how does this modification regulate their activity?

  2. What are the components, in particular NEDD8 ligases, which function with Ubc12 to neddylate specific targets in vivo?

  3. What are the functions of NEDD8 isopeptidases?

  4. What are the signals that regulate protein neddylation and deneddylation?

  5. Can cells discriminate between neddylated and mono‐ubiquitinated targets?

  6. Can NEDD8 form chains in vivo?

NEDD8 processing and activation

Similar to ubiquitin, NEDD8 is attached to its substrates by an isopeptide linkage between its carboxy‐terminal glycine (Gly) 76 and a lysine of the target protein. However, NEDD8 genes from all organisms encode non‐conjugatable precursors that contain one or more additional residues beyond Gly 76 that need to be cleaved by C‐terminal hydrolases (Fig 1). This reaction is catalysed by UCH‐L3 (Yuh1 in S. cerevisiae; Linghu et al, 2002; Wada et al, 1998), which can also process ubiquitin precursors (Frickel et al, 2007; Johnston et al, 1999; Wada et al, 1998). However UCH‐L3 knockout mice are viable (Kurihara et al, 2000), indicating that there must be other NEDD8‐processing enzymes in mammals. Indeed, NEDP1 (also known as DEN1 or SENP8)—a protein with similarity to SUMO proteases that is conserved in S. pombe, plants and animals, but not in S. cerevisiae—has been shown to catalyse the processing of NEDD8 precursors (Mendoza et al, 2003; Wu et al, 2003) with remarkable specificity (Gan‐Erdene et al, 2003; Shen et al, 2005).

Figure 1.

Neddylation pathway. Schematic representation of the main steps of the neddylation pathway, including NEDD8 precursor processing, activation by the E1 (UBA3–APPBP1 heterodimer), loading onto the E2 (Ubc12), conjugation to a substrate by an E3 and recycling of NEDD8 by an isopeptidase. APPBP1, APP binding protein 1; NEDD8, neural precursor cell expressed developmentally downregulated protein 8; UBA3, ubiquitin‐like modifier activating enzyme 3; Ubc, ubiquitin‐conjugating.

After its processing, NEDD8 is activated through an ATP‐dependent mechanism catalysed by an activating enzyme, E1, which creates a high‐energy intermediate (Huang et al, 2004a). It is then transferred to a conjugating enzyme, E2 (Huang et al, 2007), that shuttles activated NEDD8 to a ligase, E3, which then ensures specific conjugation of NEDD8 to its substrates (Fig 1). Both NEDD8 E1 and E2 are conserved from yeast to humans (for a review, see Parry & Estelle, 2004). The NEDD8 E1 activity is fulfilled by a heterodimer of APPBP1 and UBA3, which are homologous to the amino‐terminal and C‐terminal domains of the ubiquitin‐activating enzyme, respectively (Liakopoulos et al, 1998; Osaka et al, 1998; Walden et al, 2003a). Contrary to ubiquitin, which can be transferred by multiple E2s, available evidence indicates that Ubc12 functions as the unique E2 of the NEDD8 pathway (Liakopoulos et al, 1998). Indeed, despite the high sequence similarity (76%) and structural similarity between NEDD8 and ubiquitin (Whitby et al, 1998), Ubc12 is exclusively loaded with NEDD8. This insulation of the NEDD8 pathway is achieved by multiple mechanisms. First, the interaction between Ubc12 and the NEDD8 E1 involves a unique N‐terminal extension in Ubc12 (Huang et al, 2004a, 2007), which prevents mischarging by the ubiquitin E1 (Huang et al, 2008). Second, a conserved basic residue in UBA3 acts as a selectivity gate to block misactivation of ubiquitin by the NEDD8 E1: it collides with Arg 72 in ubiquitin but not with the corresponding alanine residue in NEDD8 (Souphron et al, 2008; Walden et al, 2003b). Conversely, Arg 72 is the main determinant responsible for preferential activation of ubiquitin over NEDD8 by the ubiquitin E1 (Lee & Schindelin, 2008; Whitby et al, 1998). It is noteworthy that ubiquitin activation is not as specific as NEDD8 activation, as NEDD8 can be activated by the ubiquitin E1, transferred to ubiquitin E2s and incorporated into polyubiquitin chains in vitro (Whitby et al, 1998). However, this reaction is inefficient and its consequences in vivo have not been investigated.

Neddylated proteins and their E3s

The first identified targets of NEDD8 were Cdc53 in S. cerevisiae (Lammer et al, 1998; Liakopoulos et al, 1998) and CUL4A in human cells (Osaka et al, 1998). Both are members of the cullin family of proteins—which has between three and six members in all eukaryotes from yeast to humans—suggesting that neddylation is an important mechanism for regulating cullin function. Indeed, all yeast and mammalian cullins—S. cerevisiae: Cdc53, Cul3 and Rtt101 (Laplaza et al, 2004); S. pombe: Pcu1, Pcu3 and Pcu4 (Osaka et al, 2000; Zhou et al, 2001); and mammalian: CUL1, CUL2, CUL3, CUL4A, CUL4B and CUL5 (Hori et al, 1999; Jones et al, 2008)—are neddylated on a conserved lysine in their C‐terminal domain in vivo (Table 1). This lysine is conserved in the vertebrate specific cullin‐related proteins PARC and CUL7 (Pan et al, 2004), which are probably also neddylated (Jones et al, 2008), although neddylation of Cul7 is controversial (Skaar et al, 2007). By contrast, Apc2—a subunit of the anaphase‐promoting complex that shows similarity to cullins (Zheng et al, 2002a)—is not neddylated (Pan et al, 2004).

View this table:
Table 1. Proteins reported to be neddylated

Cullins function as scaffolds for the assembly of multisubunit ubiquitin E3s. They interact tightly with a RING‐domain protein—Rbx1 or Rbx2 (Ohta et al, 1999)—which recruits charged ubiquitin E2s into the complex and catalyses the ubiquitination of cullin substrates (Seol et al, 1999). By analogy, it has been proposed that Rbx1 could also function as a NEDD8 E3 for cullins (Kamura et al, 1999). Indeed, Rbx1 interacts with both cullins and Ubc12 (Dharmasiri et al, 2003; Morimoto et al, 2003), is required for cullin neddylation in insect cells (Kamura et al, 1999; Megumi et al, 2005) and is sufficient for Cul1 neddylation in vitro (Morimoto et al, 2003). However, Dcn1—a protein conserved from yeast to humans—is also required for efficient neddylation of several cullins in vivo (Kurz et al, 2005). Dcn1 interacts directly with Cdc53, Rbx1 and Ubc12, and stimulates Cdc53 neddylation in vitro when Ubc12 is present in limiting amounts (Kurz et al, 2008; Yang et al, 2007). Moreover, a cullin mutant that efficiently binds to Rbx1 but fails to interact with Dcn1 is not efficiently neddylated (Kurz et al, 2008), suggesting that Rbx1 is not sufficient for cullin neddylation in vivo. Further experiments are required to determine how Dcn1 promotes cullin neddylation and whether it requires functional Rbx1 for its activity.

It has become evident that cullins are not the only class of proteins modified by NEDD8 (Table 1), indicating that neddylation might regulate many cellular processes. Interestingly, several neddylated proteins seem to be either substrates or components of ubiquitin E3s, revealing an intriguing relationship between ubiquitination and neddylation. For example, the tumour suppressor protein p53 and its relative p73 are both neddylated and ubiquitylated on several lysines by the RING‐domain protein Mdm2, which also self‐neddylates (Watson et al, 2006; Xirodimas et al, 2004). Similarly, the RING‐domain protein c‐Cbl can neddylate and ubiquitinate the EGFR upon its stimulation (Oved et al, 2006). pVHL, which is a well characterized component of a Cul2‐based ubiquitin E3, is also neddylated by an uncharacterized E3 (Stickle et al, 2004). Finally, several ribosomal proteins can be modified by NEDD8 (Xirodimas et al, 2008) and ribosomes are also regulated by ubiquitination (Kraft et al, 2008). BCA3 (Gao et al, 2006) and the APP intracellular domain (Lee et al, 2008) are the only neddylated proteins identified so far that have not been implicated in an ubiquitination pathway.

Despite several proteomic approaches (Jones et al, 2008; Li et al, 2006; Norman & Shiekhattar, 2006; Xirodimas et al, 2008), it is still unclear how many other proteins are modified by NEDD8. These studies confirmed that cullins are abundant NEDD8 substrates, but failed to identify other previously characterized neddylated proteins except p53 (Li et al, 2006), indicating that non‐cullin NEDD8 substrates are only weakly expressed and/or modified in steady‐state conditions. Although this does not preclude an important function of their neddylation, it calls for careful characterization of putative NEDD8 targets. We propose a minimum of three criteria that should be shown for genuine NEDD8 substrates (Sidebar A). Covalent attachment of NEDD8 to its target (criterion i) should be detectable under endogenous conditions (criterion ii). Moreover, neddylation should be dependent on bona fide components of the neddylation machinery (criterion iii) such as the NEDD8 E1 or Ubc12—the sole E2 currently known to function specifically with NEDD8. Although this criterion has only been shown for neddylation of cullins (Collier‐Hyams et al, 2005; Liakopoulos et al, 1998), and in part for p53 and pVHL (Table 1; Russell & Ohh, 2008; Xirodimas et al, 2004), it is important to exclude fortuitous misactivation of NEDD8 by the ubiquitin E1 and misloading on an ubiquitin E2 in vivo. Note, however, that this criterion cannot be demonstrated using the Ubc12‐C111S mutant because overexpression of this construct depletes free NEDD8 (Wada et al, 2000), thereby preventing specific and adventitious substrate neddylation. Additional criteria to further characterize NEDD8 substrates include the identification of the neddylated lysine residues (criterion iv), ideally with an associated mutant phenotype if neddylation is prevented (criterion v). Identification of a NEDD8 E3 (criterion vi) and reconstitution of substrate neddylation in vitro (criterion vii) underline the fact that the modification is, indeed, specific and direct. Finally, identification of a deneddylation activity (see below) provides further support for a reversible modification (criterion viii).

NEDD8 isopeptidases

Protein neddylation is reversed by NEDD8 isopeptidases in a process known as deneddylation. The best characterized NEDD8 isopeptidase is CSN5, a subunit of the COP9 signalosome (CSN), which deneddylates cullins (reviewed in Cope & Deshaies, 2003; Schwechheimer, 2004; Wei & Deng, 2003). The CSN is conserved from yeast to humans (Schwechheimer, 2004; Wei & Deng, 2003) and its activity is essential for viability in metazoans (Cope & Deshaies, 2003). Genetic evidence in several organisms has revealed that the CSN promotes cullin activity, indicating that cycles of neddylation and deneddylation are required for correct cullin function in vivo (Bosu & Kipreos, 2008; Cope & Deshaies, 2003; Pintard et al, 2003). As CSN inactivation destabilizes many subunits of cullin‐based ubiquitin ligases (reviewed in Bosu & Kipreos, 2008; Wu et al, 2006), it is thought that, at least in some cases, the deneddylating activity of the CSN, as well as the CSN‐associated deubiquitinating enzyme Ubp12, protect components of cullin‐based ubiquitin ligases from autocatalytic degradation.

The cysteine protease NEDP1—which can process NEDD8 precursors (see above)—also functions as a specific NEDD8 isopeptidase. Ala 72 in NEDD8 (Arg 72 in ubiquitin) is an important—but not the sole—determinant of NEDP1 selectivity for NEDD8 over ubiquitin (Reverter et al, 2005; Shen et al, 2005). Compared with the CSN complex, NEDP1 is inefficient at deneddylating cullins (Wu et al, 2003; Yamoah et al, 2005); however, it can process most neddylated proteins either in vitro or upon overexpression in vivo (Table 1 and references therein; see also Mendoza et al, 2003). Deletion of the NEDP1 S. pombe homologue, nep1, leads to an increase of the total cellular NEDD8 conjugates and to a cell‐cycle defect, which has not been associated with increased neddylation of a specific NEDD8 substrate (Zhou & Watts, 2005). Similarly, NEDP1 knockdown in HeLa cells increases the fraction of neddylated BCA3 (Gao et al, 2006), but no other phenotypes have been reported so far. In the absence of specific NEDP1 deneddylation substrates, it is possible that the NEDP1 isopeptidase functions primarily in a salvage pathway to remove adventitiously formed NEDD8 conjugates.

Finally, other proteases show dual specificity for ubiquitin and NEDD8. These include USP21 (Gong et al, 2000), Ataxin‐3 (Ferro et al, 2007), PfUCH54 (Artavanis‐Tsakonas et al, 2006), UCH‐L3 (see above) and UCH‐L1 (Hemelaar et al, 2004), which is a close homologue of UCH‐L3 that, however, cannot process NEDD8 precursors (Wada et al, 1998). Importantly, the in vivo targets of these proteases remain to be investigated.

Direct effects of protein neddylation

Similar to other post‐translational modifications, neddylation of proteins modifies their three‐dimensional surface and, hence, their biochemical properties. As illustrated in Fig 2, its direct effects can be classified into three categories from which further consequences such as changes in subcellular localization or enzymatic activity might follow.

Figure 2.

Direct effects of neddylation. By remodelling the three‐dimensional surface of its substrates, neddylation can achieve the following: (A) induce conformational changes (for example, neddylation of cullins allows their C‐terminal domain and Rbx1 to adopt catalytically active conformations); (B) preclude association with certain partners or compete with other posttranslational modifications (for example, cullin neddylation is incompatible with CAND1 interaction); and (C) provide a novel binding surface to recruit new partners (for example, neddylated EGFR might recruit endocytic proteins such as Eps15 or Hrs). CAND1, cullin‐associated and neddylation‐dissociated 1; EGFR, epidermal growth factor receptor; Eps15, EGFR pathway substrate 15; Hrs, hepatocyte growth factor–regulated tyrosine kinase substrate; Rbx1, RING box.

First, NEDD8 attachment can induce conformational changes of its targets. It has long been observed that neddylation of cullins stimulates their ubiquitin E3 activity in vitro (Kawakami et al, 2001; Podust et al, 2000; Read et al, 2000), although the molecular mechanism has remained elusive. In the unneddylated state, the cullin C‐terminal domain forms a groove in which the RING domain of Rbx1 is embedded (Zheng et al, 2002a). This conformation constrains the molecular movements of Rbx1 and positions the ubiquitin E2—which interacts with the RING domain of Rbx1—away from its ubiquitination substrates (Zheng et al, 2002a). Crystallographic data have now revealed that the cullin C‐terminal domain undergoes a substantial conformational change upon neddylation that ‘frees’ the RING domain of Rbx1 and allows it to adopt multiple orientations that stimulate substrate ubiquitination in vitro (Duda et al, 2008). Notably, the catalytic effect of neddylation can be mimicked by deleting or mutating the residues of the unneddylated cullin C‐terminal domain that are in contact with the RING domain of Rbx1 (Yamoah et al, 2008; Duda et al, 2008). Therefore, cullin neddylation induces a conformational switch in its C‐terminal domain that relieves Rbx1 from an autoinhibitory mechanism (Fig 2A). Interestingly, EGFR neddylation has similarly been proposed to alter the conformation of its cytoplasmic domain in a manner that exposes previously buried lysine residues for further modifications (Oved et al, 2006).

As a second possibility, neddylation can be incompatible with the interaction of some protein partners (Fig 2B). For example, the cullin inhibitor CAND1 preferentially binds to unneddylated cullins (Goldenberg et al, 2004; Liu et al, 2002; Zheng et al, 2002b). Indeed, the conformational switch induced by Cul1 neddylation not only activates Rbx1, but also prevents it from binding to CAND1 (Fig 2B; Duda et al, 2008). Similarly, neddylation of pVHL is incompatible with its incorporation in Cul2‐containing complexes and therefore stimulates its Cul2‐independent function (Russell & Ohh, 2008). In a variation of this theme, neddylation can compete with other post‐translational modifications. For example, p53 and EGFR have been shown to be neddylated and ubiquitinated on overlapping lysines (Oved et al, 2006; Xirodimas et al, 2004), and excessive neddylation of EGFR can inhibit its ubiquitination (Oved et al, 2006).

Finally, neddylation can stimulate the recruitment of NEDD8‐interacting proteins. For example, NEDD8 can interact directly with the ubiquitin E2 Ubc4 and this interaction has been proposed to participate in the activation of cullin‐based ubiquitin ligases (Sakata et al, 2007). However, in light of the structural data, such a mechanism is unlikely without additional conformational changes, as the NEDD8 surface interacting with Ubc4 is engaged in binding the C‐terminal domain of neddylated cullins (Duda et al, 2008). In another example, NUB1, which was identified in a yeast two‐hybrid screen as a NEDD8‐interacting protein (Kito et al, 2001), has been shown to trigger proteasomal degradation of neddylated but not ubiquitinated proteins (Kamitani et al, 2001). However, its physiological targets have not been identified and the significance of these observations remains unknown. Finally, neddylated EGFR likely recruits proteins of the endocytic machinery (Fig 2C) in a manner similar to mono‐ubiquitinated EGFR, and both modifications cooperate to induce EGFR downregulation (Oved et al, 2006). This observation indicates that neddylation and mono‐ubiquitination can have partly overlapping or synergistic functions.


Protein neddylation has emerged as an essential post‐translational protein modification in the cell. However, as we still do not know how many NEDD8 E3s are functioning under physiological conditions, the extent of the neddylated proteome and its functions remain unclear. Given the strict specificity of NEDD8‐activation mechanisms, it is surprising that some NEDD8 E3s are also ubiquitin E3s and that the currently known functions of protein neddylation are tightly associated with ubiquitination. Future research will be needed to investigate whether certain E3s are strictly using Ubc12 to neddylate their substrates or whether neddylation pathways always work in concert with ubiquitination.


We are grateful to Z.Q. Pan and B. Schulman for sharing their unpublished results. We thank members of the Peter laboratory for critical reading of the manuscript, in particular R. Dechant, T. Kurz, S. Leidel and N. Meyer. Work in the Peter laboratory is supported by the Swiss National Science Foundation, Oncosuisse and the Swiss Federal Institute of Technology Zürich; G.R. is supported by a Human Frontier Science Programme Organization long‐term fellowship.


Gwénaël Rabut & Matthias Peter