The ubiquitin-related SUMO system controls many cellular signaling networks. In mammalian cells, three SUMO forms (SUMO1, SUMO2 and SUMO3) act as covalent modifiers of up to thousands of cellular proteins. SUMO conjugation affects cell function mainly by regulating the plasticity of protein networks. Importantly, the modification is reversible and highly dynamic. Cysteine proteases of the sentrin-specific protease (SENP) family reverse SUMO conjugation in mammalian cells. In this Cell Science at a Glance article and the accompanying poster, we will summarize how the six members of the mammalian SENP family orchestrate multifaceted deconjugation events to coordinate cell processes, such as gene expression, the DNA damage response and inflammation.
Small ubiquitin-related modifier (SUMO) proteins are members of the ubiquitin protein family (Cappadocia and Lima, 2018; Flotho and Melchior, 2013; Gareau and Lima, 2010). Whereas yeast, insects and nematodes have one single SUMO gene known as smt3, human cells encode four SUMO forms (SUMO1, SUMO2, SUMO3 and SUMO4). SUMO1–SUMO3 are ubiquitously expressed and function as post-translational modifiers through their covalent attachment to lysine residues of target proteins. Prior to conjugation, SUMO1–SUMO3 are C-terminally processed from their respective precursor forms (pre-SUMO) by cysteine proteases of the ubiquitin-like-specific protease 1 (Ulp) or sentrin-specific protease (SENP) family (Hickey et al., 2012; Nayak and Müller, 2014). SENPs cannot process pre-SUMO4, which consequently most likely does not form SUMO conjugates (Owerbach et al., 2005). At the amino acid level, SUMO1 shares ∼50% sequence identity with SUMO2 and SUMO3, which differ in only three amino acids after processing and in the following will therefore be mostly denoted SUMO2/3 (Müller et al., 2001).
The conjugation of SUMO1–SUMO3 to proteins, termed SUMOylation, is mediated by a machinery comprising the dimeric SUMO-activating enzyme subunit 1 and 2 (SAE1–SAE2; also termed UBA2–AOS1), the E2 conjugating enzyme UBC9 (also known as UBE2I) and E3 SUMO ligases (see poster) (Cappadocia and Lima, 2018; Flotho and Melchior, 2013; Gareau and Lima, 2010). SUMOs can be attached to target proteins as single moieties to either one (mono-SUMOylation) or multiple acceptor lysine residues (multiple mono-SUMOylation). Notably, SUMOylation, in particular by SUMO2/3, can also give rise to polymeric chains, in which SUMO–SUMO linkages occur through lysine residues within SUMO moieties (poly-SUMOylation) (see poster) (Ulrich, 2008; Vertegaal, 2010).
SUMOylation generally modulates cellular protein networks by regulating the dynamics of protein–protein interactions (Gareau and Lima, 2010; Raman et al., 2013). For instance, SUMO conjugation can prevent protein–protein interactions by masking interaction sites. More frequently, however, the conjugation initiates the recruitment of binding partners that harbor distinct SUMO-interaction modules, termed SIMs (see poster) (Gareau and Lima, 2010; Raman et al., 2013). Some SIMs exhibit specificity for SUMO1 or SUMO2/3. Moreover, specialized multivalent SUMO-binding modules preferentially bind to SUMO targets that are multi- or poly-SUMOylated. The best-studied multi- or poly-SUMO-mediated process is the SUMO-targeted E3 ubiquitin ligase (StUbL) pathway (see poster) (Geoffroy and Hay, 2009; Sriramachandran and Dohmen, 2014). In this stress-induced pathway, a multi- or poly-SUMOylated protein is recognized by specific E3 ubiquitin ligases, such as mammalian RNF4 or RNF111, which trigger subsequent proteolytic or non-proteolytic ubiquitylation events. Notably, RNF4 not only ubiquitylates the target protein, but also SUMO itself, thereby generating hybrid ubiquitin–SUMO chains (Guzzo et al., 2012; Lallemand-Breitenbach et al., 2008; Tatham et al., 2008). Hybrid chains have been implicated in the DNA damage response (DDR), where they signal through binding to the double-strand break repair factor RAP80 (also known as UIMC1) (Guzzo et al., 2012). The deubiquitylases USP7 and USP11 can deubiquitylate hybrid SUMO-ubiquitin chains to counter the RNF4 pathway (Hendriks et al., 2015; Lecona et al., 2016).
SUMOylation is a highly dynamic and reversible process, so that typically only a small fraction of a SUMO target is modified at a given time (Hay, 2005). This is due to the rapid deconjugation by Ulp/SENP proteases. These enzymes therefore determine both the level of processed SUMO and the extent of substrate modification by counterbalancing SUMO conjugation.
SENPs as maturation enzymes and deconjugases of SUMO
The founding members of the Ulp/SENP family are the Saccharomyces cerevisiae Ulp1 and Ulp2 proteins, initially described and characterized by Li and Hochstrasser (Li and Hochstrasser, 1999, 2000). The human SENP family consists of seven members, SENP1, SENP2, SENP3, SENP5, SENP6, SENP7 and SENP8 (Hickey et al., 2012; Mukhopadhyay and Dasso, 2007; Yeh et al., 2000) (see poster). SENP8, however, exhibits specificity for the ubiquitin-like Nedd8 protein and does not act on SUMO. Ulp/SENP proteins belong to the clan CA of cysteine proteases, which share a conserved catalytic domain characterized by a papain-like fold. Sequence comparison and phylogenetic analysis of the catalytic domain illustrates the pairwise evolutionary relationship between SENP1 and SENP2, SENP3 and SENP5, and SENP6 and SENP7. The analysis further depicts that the SENP1 and SENP2, and SENP3 and SENP5 pairs emanate from the Ulp1 branch, while the more divergent SENP6 and SENP7 forms belong to the Ulp2 branch (see poster) (Hickey et al., 2012). In addition to their conserved catalytic domains, SENPs possess specific N-terminal domains, which are involved in their regulation and substrate selection (Box 1) (Hay, 2007).
The spatial distribution of SENPs is mostly determined by their variable N-terminal region. Distinct protein interaction domains therein recruit SENP activity to specific cellular subcompartments. SENP3 and SENP5 are targeted to the nucleolus through the binding of their N-terminal region to the nucleolar scaffold protein nucleophosmin (NPM1) (Castle et al., 2012; Haindl et al., 2008; Raman et al., 2014; Yun et al., 2008). Chromatin-targeting of SENP3 is in turn regulated by its association with protein flightless-1 homolog (FLII) (Nayak et al., 2017). Alternative splicing or post-translational modifications (PTMs) in the variable N-terminal region enlarges the regulatory repertoire of SENP proteins. Splice variants with N-terminal regions of different length have been described for SENP2 and SENP7 (Bawa-Khalfe et al., 2012; Jiang et al., 2011). The longest murine SENP2 isoform localizes to the nucleus and partially concentrates in PML nuclear bodies, whereas a short form (SENP2S) with a truncation of the first 87 amino acids is found in the cytosol (Jiang et al., 2011). A third isoform (SENP2M) with an alternative N-terminus is concentrated in cytoplasmic vesicles that colocalize with Golgi markers (Jiang et al., 2011). The N-terminus of the longest SENP2 isoform contains multiple elements that cooperate in targeting it to the nuclear pore complex. Such targeting involves binding to karyopherins and components of the nuclear pore complex (Chow et al., 2012; Goeres et al., 2011; Hang and Dasso, 2002; Zhang et al., 2002). Human SENP7 also exists in at least two forms (Bawa-Khalfe et al., 2012). The long isoform SENP7L is nuclear, whereas the shorter isoform SENP7S localizes exclusively to the cytosol (Bawa-Khalfe et al., 2012). SENP7L contains a conserved heterochromatin protein 1 homolog (HP1)-box (PxVxL), which determines the mutual recruitment of SENP7 and HP1α to heterochromatin (Maison et al., 2012; Romeo et al., 2015). SENP7S lacks the HP1-binding domain, explaining its cytosolic distribution (Bawa-Khalfe et al., 2012).
Regulation of SENPs by PTMs is another important way to control their subcellular distribution and abundance. In response to disturbed flow, SENP2 undergoes phosphorylation in epithelial cells by the RPS6 kinase, thereby promoting its cytoplasmic translocation (Heo et al., 2015). The kinase mTOR phosphorylates SENP3 within its N-terminal region and so fosters interaction with NPM1 and nucleolar targeting (Raman et al., 2014). In mitotic prometaphase, SENP3 becomes hyperphosphorylated and is dephosphorylated upon mitotic exit (Klein et al., 2009). SENP3 is also phosphorylated in an in vitro model of ischemia through a pathway that involves the unfolded protein response (UPR) kinase PERK (also known as EIF2AK3) (Guo et al., 2013). This induces lysosomal degradation of SENP3 and concomitant upregulation of SUMO2 and SUMO3 targets, which may contribute to the cytoprotective effect of enhanced SUMOylation in ischemic cells (see Box 2). SENP7 is targeted for degradation by the UPS through the E3 ubiquitin ligase adaptor SPOP (Zhu et al., 2015). Degradation of SENP7 promotes cellular senescence, possibly through enhanced SUMOylation of HP1α (Zhu et al., 2015).
SENPs have a dual function as processing enzymes for pre-SUMO and deconjugases of SUMO conjugates (see poster). In the deconjugation reaction, SENPs cleave an isopeptide bond that links SUMO moieties to the ε-amino group of lysine residues, while processing hydrolyzes a peptide bond close to the C-terminus of SUMO precursors (Hickey et al., 2012). This clips off the very C-terminal amino acids from SUMO1, SUMO2 and SUMO3 to expose two glycine residues (a diGly motif), which are a prerequisite for activation and conjugation. In SUMO1–SUMO3, the diGly motif is preceded by a glutamine (Q) and threonine (T), while SUMO4 exhibits a PTGG motif, in which the proline residue confers resistance to SENP-mediated cleavage (see poster) (Owerbach et al., 2005). The crystal structures of the catalytic domains of SENP1 or SENP2 in complex with SUMO precursors or the SUMO-conjugated model substrate RanGAP1 illustrate the proteolytic mechanism of SENP-mediated catalysis and partially explain their preference for SUMO paralogs (Reverter and Lima, 2004, 2006; Shen et al., 2006a,b; Xu et al., 2006). The substrate contacts the catalytic cysteine residue through a shallow tunnel, in which the diGly motif and the scissile bond are positioned over the active site. In the SENP–preSUMO and SENP–SUMO1–RanGAP1 complexes, the scissile peptide bond is kinked at a right angle to the C-terminal tail of SUMO and adopts an atypical cis configuration (see poster), which likely destabilizes this bond to promote cleavage (Huang and Schulman, 2006). In vitro protease assays with the isolated catalytic domains demonstrate the processing activity of SENP1 and SENP2 on all three SUMO precursors. However, they exert differential activities towards distinct precursors (see poster). SENP2 is most active on SUMO2, followed by SUMO1 and SUMO3, whereas SENP1 prefers SUMO1 over SUMO2 and SUMO3 (Kolli et al., 2010; Reverter and Lima, 2004, 2006; Shen et al., 2006a). These differences have been attributed to the amino acid sequences of the C-terminal tail (Reverter and Lima, 2004, 2006; Shen et al., 2006a,b; Xu et al., 2006). In vitro data for SENP3 processing activity are still missing, but SENP5 has been found to have a pronounced preference for SUMO2 cleavage (Kolli et al., 2010; Mikolajczyk et al., 2007). Notably, SENP6 and SENP7 are almost inactive during SUMO maturation (Kolli et al., 2010; Lima and Reverter, 2008). These data indicate that differential precursor processing may control the availability of conjugatable SUMO in vivo. However, the in vivo contribution of a given SENP to processing of a distinct pre-SUMO protein and the physiological relevance of this process has remained largely elusive. It is also unclear how the N-terminal regions of SENPs influence the cleavage reaction in vivo.
Deconjugation by SENP family members has been most thoroughly studied in vitro on SUMO-modified RanGAP1 as a model substrate (Kolli et al., 2010; Mikolajczyk et al., 2007; Reverter and Lima, 2006; Shen et al., 2006a). SENP1 and SENP2 release all SUMO forms from RanGAP1 with a comparable efficiency. By contrast, SENP3 and SENP5 only have a low isopeptidase activity against RanGAP1–SUMO1 and preferentially deconjugate SUMO2/3 from RanGAP1 (Kolli et al., 2010; Mikolajczyk et al., 2007) (S.M., unpublished observations). SENP6 and SENP7 also act preferentially on RanGAP1–SUMO2, but most efficiently cleave di- and poly-SUMO2/3 chains that are linked through internal lysine residues (Kolli et al., 2010; Lima and Reverter, 2008; Mukhopadhyay et al., 2006; Shen et al., 2009). Notably, the catalytic domains of SENP6 and SENP7 have four conserved loop insertions that are missing in other SENPs. Structural and biochemical work on SENP6 and SENP7 has revealed the importance of loop 1 in selectively binding to SUMO2/3-conjugated substrates (Alegre and Reverter, 2011; Lima and Reverter, 2008). Moreover, loop 1 contributes to the preference of SENP6 and SENP7 for cleaving di- and poly-SUMO2 (Alegre and Reverter, 2011; Lima and Reverter, 2008).
Determinants of in vivo substrate specificity
One conundrum of the SUMO system is how the relatively small set of conjugating and deconjugating enzymes accurately controls the SUMOylation status of thousands of target proteins. There are some examples of distinct deconjugation events that regulate specific biological processes, but in many cases, a given SENP may act on larger groups of SUMOylated proteins. The conception of protein group SUMO modification, established by Jentsch and Psakhye, signifies that each conjugating or deconjugating enzyme controls the modification status of larger groups of proteins (Jentsch and Psakhye, 2013; Psakhye and Jentsch, 2012). These groups are often part of functional networks and, in many cases, are physically associated in macromolecular complexes. In mammalian cells, this is reflected by the compartmentalization of SUMO conjugates in microscopically discernible subcellular domains, such as the nuclear pore complex (NPC), promyelocytic leukemia (PML) nuclear bodies, DNA repair foci or the nucleolus (Raman et al., 2013). Notably, distinct SENP family members are also dynamically associated with these sites (see poster). Alternative splicing and post-translational modifications of SENPs are important determinants of their localization and activity (see Boxes 1 and 2), implying that spatial control is a key regulatory principle in SENP function. SENP1 and SENP2 (the long isoform) are enriched at the NPC and in PML nuclear bodies in interphase cells (Chow et al., 2012; Goeres et al., 2011; Hang and Dasso, 2002; Zhang et al., 2002). After nuclear envelope breakdown in mitosis, GFP–SENP1 and GFP–SENP2 can be detected at the kinetochore (Cubenas-Potts et al., 2013). SENP3 and SENP5 exhibit a prominent nucleolar concentration with subfractions found at chromatin or mitochondria (Gong and Yeh, 2006; Haindl et al., 2008; Nayak et al., 2017, 2014; Nishida et al., 2000; Yun et al., 2008; Zunino et al., 2009, 2007). SENP6 and SENP7 (the long isoform) are mainly found in the nucleoplasm, where they at least partially reside at chromatin (Lima and Reverter, 2008; Maison et al., 2012; Mukhopadhyay et al., 2006; Shen et al., 2009).
A striking observation is the dramatic increase in SUMO conjugation upon exposure of cells to stress. Upon hypoxic, osmotic, thermal or proteotoxic stress, cells exhibit an enhanced pattern of SUMO conjugation (Saitoh and Hinchey, 2000). Several lines of evidence indicate that this is at least partially due to changes in the activity or abundance of SENPs. Reduced activity of SENPs has been described under thermal, oxidative or hypoxic stress (Kunz et al., 2016; Pinto et al., 2012; Xu et al., 2008). An emerging aspect of SENP regulation is their redox-dependent regulation, suggesting SENPs function as cellular redox sensors. For example, severe oxidative stress induced by H2O2 inactivates SENP1 by reversible oxidation of its catalytic cysteine residue (Xu et al., 2008). Redox-sensitivity of SENP1 might be physiologically relevant in SENP1-mediated insulin secretion (Ferdaoussi et al., 2015; Gooding et al., 2015) (see poster). Gooding et al. have shown that adenylosuccinate (S-AMP), an intermediate of glucose-induced purine metabolism, requires SENP1 to amplify insulin secretion under high-glucose stimulation (Gooding et al., 2015). Importantly, Ferdaoussi and coworkers determined the importance of the SENP1 redox state for proper insulin secretion (Ferdaoussi et al., 2015). They demonstrated that NADPH and reduced glutathione (GSH) are crucial for SENP1-triggered insulin secretion and, accordingly, oxidation prevents the SENP1-dependent amplification of exocytosis. An important redox regulation has also been proposed for SENP3. Interestingly, severe oxidative stress inactivates SENP3, whereas mild stress stabilizes SENP3 by protecting it from ubiquitylation (Huang et al., 2009). Under normal growth conditions, SENP3 is targeted for degradation by the ubiquitin ligase C-terminus of Hsc70-interacting protein (CHIP; also known as STUB1), but oxidation of two cysteine residues within SENP3 under mild oxidative stress prevents this process (Yan et al., 2010).
Ischemia–reperfusion injuries are also characterized by severe oxidative stress and, interestingly, a dramatic upregulation of SUMO conjugation is also observed in murine brain following cerebral ischemia–reperfusion (Yang et al., 2008). As outlined in Box 1, this most likely involves reduced stability of SENP3 (Guo et al., 2013). Although reduction of SENP3 leads to enhanced SUMO2 and SUMO3 modification of a whole set of proteins, one key target for SENP3-mediated deSUMOylation appears to be the dynamin-related GTPase Drp1 (also known as DNM1L), which functions a master regulator of mitochondrial fission and fusion (Guo et al., 2013). SUMOylation of Drp1 has been proposed to regulate Drp1 recruitment to the outer mitochondrial membrane in order to control mitochondrial fission and apoptosis (Guo et al., 2013). Inhibition of SENP3 was shown to prolong Drp1 SUMO2 conjugation, which in turn, prevents Drp1 association with mitochondria and inhibits apoptotic cell death (Guo et al., 2013). These data are consistent with the idea that reduction of SENP3 activity is a cytoprotective response to ischemia–reperfusion injury. Notably, however, this regulatory pathway appears to be specific for SENP3 and SUMO2/3, as neural-specific depletion of SENP2 in mice leads to enhanced apoptosis and mitochondria-mediated neurodegenerative disorders, which has been linked to prolonged SUMO1 conjugation to Drp1 (Fu et al., 2014). However, again one needs to emphasize that lack of SENP2 affects multiple pathways. Indeed, Yeh and coworkers found that mice with partial deficiency of SENP2 develop spontaneous seizures and sudden death, which they explained by hyper-SUMOylation of multiple K+ channels (Qi et al., 2014).
Selected cellular pathways controlled by SENPs
In the following section, we summarize some key findings that exemplify the role of SUMO deconjugation in selected cellular pathways. We focus on SENP-controlled processes that are experimentally validated by phenotypes observed upon genetic or siRNA-mediated inactivation of a given SENP in cell culture or in animal models. In some cases, these phenotypes have been linked to impaired deSUMOylation of distinct targets. However, it needs to be emphasized that the subset of SUMOylated proteins that are targeted for deconjugation by a given mammalian SENP family member is not yet defined. At this stage, this makes it difficult to pinpoint the biological effects of inactivation to the lack of a single SUMO-deconjugation event.
SENPs as regulators of macromolecular assemblies in the nucleus – PML nuclear bodies and pre-ribosomes
PML protein functions as a scaffold for a nuclear multiprotein complex known as PML nuclear bodies (Uversky, 2017). The transient recruitment of proteins to nuclear bodies affects diverse cellular signaling pathways, ranging from the control of gene expression to protein quality control and immune signaling. Nuclear bodies are a cellular hub for SUMO conjugation–deconjugation and have become the prototypical example of a protein complex whose dynamics are controlled by multiple SUMO–SIM interactions (see poster) (Raman et al., 2013). The number and composition of nuclear bodies are influenced by the degree of PML SUMOylation, which can promote self-assembly through its own SIM or recruit additional SIM-containing proteins. Recent biophysical studies underline the importance of SUMO in providing a platform for multivalent interactions that foster nuclear body assembly (Banani et al., 2016). In line with this, it has been demonstrated that the chain-specific deconjugases SENP6 and SENP7 control the dynamics of PML nuclear bodies (Hattersley et al., 2011; Mukhopadhyay et al., 2006; Shen et al., 2009). Poly-SUMO-modified PML itself is a direct SENP6 substrate, and downregulation of SENP6 therefore induces the formation of a SUMO chain on PML (Hattersley et al., 2011). This in turn triggers an increase in the number and size of nuclear bodies, most likely through their de novo formation and the enhanced recruitment of SUMO-binding proteins to these structures.
Another macromolecular assembly that is exquisitely controlled by Ulp/SENPs are pre-ribosomal particles (Finkbeiner et al., 2011b; Panse et al., 2006) (see poster). Subsequent to their formation in the nucleolus, pre-60S and pre-40S ribosomes undergo extensive remodeling processes before they are exported to the cytosol. Our group has shown that SENP3 plays a pivotal regulatory role in 28S rRNA maturation and 60S ribosome formation (Haindl et al., 2008). SENP3 controls the SUMOylation status of the PELP1 complex, which is implicated in the late steps of nucleolar maturation of pre-60S particles (Finkbeiner et al., 2011a). Transient SUMO conjugation to PELP1 facilitates the recruitment of MDN1, a 60S-remodeling ATPase (Raman et al., 2016). Subsequent deSUMOylation of PELP1 by SENP3 contributes to the release of the MDN1–PELP1 complex from pre-60S particles, allowing further progression of the biogenesis pathway.
SENPs in chromatin remodeling and the control of gene expression
Gene expression is another cellular key process controlled by alterations in SUMO conjugation and deconjugation. Although SUMOylation has occasionally been linked to gene activation (Niskanen and Palvimo, 2017), the modification of transcription factors or transcriptional co-regulators restrains gene activation in most cases (Wotton et al., 2017). A recurrent theme is that chromatin-associated SUMOylation facilitates the establishment of a repressive environment, whereas deSUMOylation by SENPs counters this process, thereby contributing to gene activation (Niskanen and Palvimo, 2017; Wotton et al., 2017). At the center of these activities are chromatin-modifying enzyme complexes that act as epigenetic regulators (see poster). We will illustrate this concept with three selected examples.
In mammalian cells, SENP functions have been connected to pathways that are controlled by polycomb (PcG) and trithorax (trxG) group proteins. PcG and trxG proteins regulate the expression patterns of developmental genes, such as HOX genes, by maintaining their silenced or activated state. Indeed, SUMOylation of CBX4, a subunit of the polycomb repressive complex 1 (PRC1), mediates its chromatin recruitment, whereas deSUMOylation by SENP2 reverses this process (Kang et al., 2010). Accordingly, in the absence of SENP2, the occupancy of PRC1 is increased on promoters of several PcG target genes, including a subset of HOX genes, which leads to severe developmental defects (Kang et al., 2010). Importantly, deSUMOylation can also facilitate the assembly of mammalian trxG complexes that function as ‘writers’ for activating histone marks on HOX gene promoters. This is illustrated by our finding that SENP3 is found at promotors of a subset of HOX genes, where it facilitates the assembly of the MLL/SET methyltransferase complex by deconjugating SUMO2/3 from the subunit RbBP5 (Nayak et al., 2017, 2014). SUMOylation of RbBP5 blocks recruitment of the Ash2L and menin subunits, thus restraining H3K4 methylation; SENP3 relieves this constraint promoting transcriptional activation of a subset of HOX genes.
Another instructive example for SUMO-mediated gene repression is that of KAP1 (also known as TIF1β or TRIM28). SUMOylation of KAP1 facilitates the chromatin recruitment of the SETDB1 histone methyltransferase and components of the NuRD complex (e.g. CHD3) ultimately triggering repressive heterochromatic marks, such as tri-methylation of H3K9 and H4K20 (Ivanov et al., 2007). SENP1 and SENP7 have been proposed to reverse KAP1 SUMOylation and concomitantly relieve SUMO-mediated repression. Although SENP1 may preferentially act on KAP1–SUMO1 conjugates (Li et al., 2007), SENP7-mediated removal of SUMO2/3 chains from KAP1 appears to be particularly important in the response to DNA damage (Garvin et al., 2013). In this context, deSUMOylation of KAP1 contributes to the establishment of a permissive chromatin environment for DNA repair. DeSUMOylation of KAP1 impairs recruitment of the chromatin remodeler CHD3, thereby facilitating chromatin relaxation and homologous recombination (HR) (Garvin et al., 2013).
SUMO chain editing by SENP6 in the control of genome integrity
Genetic and biochemical data in both lower and higher eukaryotes underscore the key role of the SUMO system in the maintenance of genome integrity (Bergink and Jentsch, 2009). The role of SENP7 in HR has already been addressed above, but a balanced SUMO conjugation–deconjugation and especially the control of SUMO chain formation is also important during replication stress. In the Fanconi anemia repair pathway, SUMO-primed ubiquitylation by RNF4 cooperates with the action of the segregase p97 (also known as VCP) to extract the FANCI–FANCD2 complex from the sites of DNA lesions (Gibbs-Seymour et al., 2015). SENP6 antagonizes the StUbL pathway by limiting SUMO chain formation on FANCI. Considering the involvement of RNF4 in several DDR pathways, SENP6 may have a more general function in balancing chromatin association of protein complexes that are engaged in DNA repair. Along this line, SENP6 was also reported to control HR during replication by controlling the modification status of the replication factor RPA70 (also known as RPA1) (Dou et al., 2010). During unperturbed replication, SENP6 is associated with RPA70, thus limiting its SUMOylation. Double-strand breaks trigger the dissociation of SENP6 from RPA70 and so induce its SUMOylation, which results in the recruitment of Rad51 to the DNA damage foci and initiates DNA repair through HR (Dou et al., 2010).
SENPs in inflammatory signaling and the adaptive immune response
The nuclear factor (NF)κB signaling pathway is a key regulator of inflammation and cell survival. The activation of NFκB is under the exquisite control of PTM networks. The kinase NEMO (also known as IKKγ or IKBKG) is part of the IKK complex that triggers NFκB activation. In response to DNA damage, NEMO is modified by SUMO1, whereas pro-inflammatory stimuli such as lipopolysaccharide (LPS) trigger NEMO modification by SUMO2/3 (Liu et al., 2013; Wuerzberger-Davis et al., 2007) (see poster). SUMOylation of NEMO contributes to NFκB activation and accordingly, deSUMOylation attenuates NFκB signaling (Liu et al., 2013). SENP2 counters NFκB activation upon DNA damage through deconjugation of SUMO1 from NEMO (Lee et al., 2011), whereas SENP6 attenuates the canonical Toll-like receptor (TLR)-triggered inflammation by cleaving SUMO2/3 from NEMO (Liu et al., 2013). Depletion of SENP6 enhances the NFκB-mediated induction of proinflammatory genes, as seen in SENP6-knockdown mice, which are more susceptible to endotoxin-induced sepsis (Liu et al., 2013). SENP6 facilitates removal of polyubiquitin chains on NEMO by fostering binding of the deubiquitylase CYLD, ultimately dampening IKK complex activation (Liu et al., 2013).
Notably, NEMO is not the only SENP target in the NFκB signaling pathway. Transcriptional activation of many LPS- or TLR-response genes requires an initial de-repression step, in which complexes formed by nuclear receptors and co-repressors are actively removed from the promoters of NFκB target genes. This clearance step involves deSUMOylation of nuclear receptors by SENP3 (Huang et al., 2011). The activity of NFκB may thus be orchestrated by a coordinated and possibly also tissue-specific interplay of several SENP family members in both the cytoplasmic and nuclear steps of NFκB signaling. Along this line, it is worth noting that, in mouse adipocytes, SENP1-mediated NEMO deSUMOylation limits inflammatory responses. Adipocyte-specific deletion of SENP1 in mice aggravates diabetes phenotypes by triggering unrestricted NFκB activation and production of proinflammatory cytokines (Shao et al., 2015). Intriguingly, SENP1 and SENP2 themselves are NFκB target genes, implying that there is an intricate negative feedback mechanism in SENP-regulated NFκB signaling (Lee et al., 2011).
Recent data also point to an important function of SENPs in the innate immune system by controlling the cGAS (cyclic GMP-AMP synthase)–STING (stimulator of interferon genes) signaling pathway. This pathway senses cytosolic DNA in cells infected by certain viruses or bacteria, and triggers the expression of inflammatory genes through interferon 3 (IRF3) (Maringer and Fernandez-Sesma, 2014). SENP7 has been proposed to potentiate cGAS–STING activation by reversing an inhibitory SUMOylation of cGAS (Cui et al., 2017). Consistent with this idea, mice lacking SENP7 are more susceptible to herpes simplex virus 1 (HSV-1) infection and exhibit an impaired gene expression pattern upon induction of the cGAS–STING axis (Cui et al., 2017).
SENPs in cell cycle progression
Genetic data in the yeast S. cerevisiae first established a pivotal function of timed SUMO conjugation–deconjugation in the control of the cell division cycle. Inactivation of either the E2-conjugating enzyme UBC9 or the SUMO-deconjugating enzyme Ulp1 prevents cell cycle progression at the G2/M transition (Li and Hochstrasser, 1999; Seufert et al., 1995). In mammalian cells, SUMO2/3 localize to centromeres, whereas SUMO1 is detected at the mitotic spindle and spindle midzone. SUMOylation is highly dynamic during cell cycle progression and, accordingly, mammalian SENPs have also been linked to cell cycle progression (Cubenas-Potts et al., 2013; Zhang et al., 2008). Knockdown of SENP1 delays progression through mitosis because of a failure in sister chromatid separation at metaphase (Cubenas-Potts et al., 2013). Interestingly, overexpression of SENP2 induces a prometaphase arrest due to a defective targeting of the microtubule motor protein CENP-E to kinetochores, further underlining the importance of balanced SUMO conjugation–deconjugation for mitotic progression (Zhang et al., 2008). In line with this, lack of SENP5 results in the formation of binucleate cells, pointing to a role for SENP5 in cytokinesis (Di Bacco et al., 2006). Notably, SENP5 relocalizes from nucleoli to the mitochondrial surface at the G2/M transition, which promotes mitochondrial fission during mitosis (Zunino et al., 2009). Depletion of the chain-specific protease SENP6 in cell culture causes chromosome misalignment, prolonged mitotic arrest and chromosome missegregation due to impaired inner kinetochore assembly and centromere organization (Mukhopadhyay and Dasso, 2010). This defect could be secondary to SENP6 function in S phase. Accordingly, SENP6 has been proposed to protect inner kinetochore proteins from degradation during S phase by antagonizing the StUbL pathway (Mukhopadhyay and Dasso, 2010). Taken together, these data point to crucial functions of SENPs for proper mitotic progression. However, the specific mitotic substrates of SENPs remain to be determined.
Conclusions and perspectives
The above-mentioned examples illustrate the importance of SENP proteins in cellular processes by ensuring balanced SUMO conjugation–deconjugation. However, to provide an integrated view of SENP function we still need to define the set of SUMOylated proteins targeted by a given SENP family member. Many published data on SENP substrates are based on overexpression of SENP family members, which does not reflect their physiological functions. Importantly, the field also lacks specific inhibitors to directly target SENP activity. Most studies therefore rely on the use of siRNA or CRISPR-mediated inactivation of SENPs, which makes it difficult to distinguish between direct and indirect effects of SENP inactivation. The availability of chemical inhibitors would also be a prerequisite for targeting SENPs in human disease, including cancer, where a misregulation of SENPs has been observed (Seeler and Dejean, 2017).
We thank Anne Gärtner for help with the images of SENP1 structures and all members of the SUMO signaling group for critical discussion of the manuscript.
This work was funded by the Deutsche Forschungsgemeinschaft (DFG) collaborative research centers (SFB815 and SFB1177), the LOEWE Ub-Net and DFG grant MU-1764/4.