Molecular targets underlying SUMO-mediated neuroprotection in brain ischemia



SUMOylation (small ubiquitin-like modifier conjugation) is an important post-translational modification which is becoming increasingly implicated in the altered protein dynamics associated with brain ischemia. The function of SUMOylation in cells undergoing ischemic stress and the identity of small ubiquitin-like modifier (SUMO) targets remain in most cases unknown. However, the emerging consensus is that SUMOylation of certain proteins might be part of an endogenous neuroprotective response. This review brings together the current understanding of the underlying mechanisms and downstream effects of SUMOylation in brain ischemia, including processes such as autophagy, mitophagy and oxidative stress. We focus on recent advances and controversies regarding key central nervous system proteins, including those associated with the nucleus, cytoplasm and plasma membrane, such as glucose transporters (GLUT1, GLUT4), excitatory amino acid transporter 2 glutamate transporters, K+ channels (K2P1, Kv1.5, Kv2.1), GluK2 kainate receptors, mGluR8 glutamate receptors and CB1 cannabinoid receptors, which are reported to be SUMO-modified. A discussion of the roles of these molecular targets for SUMOylation could play following an ischemic event, particularly with respect to their potential neuroprotective impact in brain ischemia, is proposed.

Abbreviations used

α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor


CCAAT/enhancer-binding protein beta


dynamin related protein 1


excitatory amino acid transporter 2


estrogen receptor β


glucose transporter


glycogen synthase kinase-3β


hypoxia inducible factor-1


homeodomain-interacting protein kinase 2


heat shock factor 1


interferon-stimulated gene 15


inhibitor of κB


c-Jun N-terminal kinase


kainate receptor


long-term potentiation


metabotropic glutamate receptor


mixed lineage kinase 3


nuclear factor kappa-light-chain-enhancer of activated B cells


oxygen/glucose deprivation


protein inhibitor of activated STAT1


Ran binding protein 2


reactive oxygen species


sentrin-specific proteases


small ubiquitin-like modifier


ubiquitin-like modification

Post-translational modifications are critical events in signaling cascades that enable cells to efficiently, rapidly and reversibly respond to extracellular stimuli. This is particularly important in the CNS, where extremely complex and finely tuned processes, such as synaptic communication, take place. Apart from playing a key physiological role, post-translational modifications, such as the well-characterized phosphorylation and ubiquitynation processes, mediate the synaptic dysfunction and neuronal death occurring in neurodegenerative diseases. SUMOylation, which involves the covalent attachment of the 97-amino acid small ubiquitin-like modifier (SUMO) protein, has been shown to be crucial for cell viability and to be activated in the post-ischemic brain (reviewed in Yang et al. 2008a). We update this topic by focusing on key molecular targets for SUMOylation that could play a potential neuroprotective role following an ischemic event.

Although the exact function of SUMOylation and the identity of disease-modified SUMO targets remain largely unknown, evidence supports the involvement of SUMOylation in an endogenous neuroprotective response (Lee et al. 2007, 2009b, 2011; Datwyler et al. 2011; Cimarosti et al. 2012). Given the neuroprotective potential of SUMOylation in brain ischemia, it is tempting to speculate that a detailed investigation into the specific disease-modified SUMO targets, as well as the underlying mechanisms and pathways, may identify new molecular targets for future drug discovery. Here, we briefly review the current studies investigating the roles of SUMOylation in brain ischemia, focusing on the identification of molecular targets underlying SUMO-mediated neuroprotection. The hypothesis that SUMOylation of specific proteins, such as transcription factors, signaling kinases, transporters, ion channels and receptors, can counteract the cascade of damaging events following ischemia will be explored.

Post-translational modification by SUMO

SUMOylation involves the attachment of the SUMO protein to lysine residues, usually within a consensus motif, on target proteins, modulating their activity, stability and subcellular localization. Three SUMO isoforms are present in the brain: SUMO-1 shares ~ 50% sequence identity with SUMO-2 and SUMO-3, which differ by only three amino acids. An important aspect of SUMOylation is that, despite being a covalent modification, it is a highly labile process, allowing cells to respond rapidly to varying cellular demands (reviewed in Cimarosti and Henley 2008). Critical in the SUMOylation pathway is Ubc9, the only known SUMOylating E2-conjugating enzyme, and sentrin-specific proteases (SENPs), a family of seven SUMO-specific de-SUMOylating proteases; nevertheless, E1-activating and E3-ligating enzymes also play a role (reviewed in Gareau and Lima 2010).

The SUMO targets identified so far are largely transcription factors and other nuclear proteins involved in gene expression or DNA integrity, but can also be extranuclear proteins involved in diverse cellular processes, e.g. cell signaling, plasma membrane depolarization, and signal transduction (reviewed in Hay 2005; Geiss-Friedlander and Melchior 2007; Gareau and Lima 2010). In this regard, the elements required for SUMOylation have been shown to be present at the plasma membrane (Plant et al. 2010).

SUMOylation and ischemic neuroprotection

Mammalian hibernation serves as a natural model of ischemia because of the similarities in reduced cerebral blood flow and, consequently, low levels of oxygen and glucose. However, no brain damage occurs in hibernation. Thus, the observation that SUMOylation is dramatically increased in the brains of hibernating ground squirrels suggests that SUMOylation may provide ischemic neuroprotection (Lee et al. 2007) and raises the possibility that SUMOylation of specific proteins may contribute to this process. This study also showed that over-expression of the SUMOylating Ubc9 in SHSY5Y cells increased tolerance to oxygen/glucose deprivation (OGD), an in vitro model of ischemia, which was reversed by over-expression of dominant negative Ubc9 (Lee et al. 2007).

In vivo studies from adult mice subjected to hypoxia showed a remarkable increase in both SUMO-1 protein and mRNA levels in the brain and heart (Shao et al. 2004). Enhanced global SUMOylation was subsequently observed in several models of brain ischemia. Both transient and permanent global or focal ischemia, in mice and rats, induced a rapid, substantial, and long-lasting increase in SUMOylation (Cimarosti et al. 2008; Yang et al. 2008b,c). These increases, mostly in SUMO-2/3, but also in SUMO-1 conjugation, were particularly prominent in the cortex, striatum and hippocampus of the infarcted and/or non-ischemic brain. Hypothermia induced SUMO-2/3 conjugation, translocation to the nucleus and modified gene expression (Yang et al. 2009), further suggesting a neuroprotective role for SUMOylation. More recently, it has been shown that transient ischemia in spinal cord neurons induced a marked increase in SUMO-2/3 conjugation, while conjugation by SUMO-1 remained unaltered (Wang et al. 2012).

To expand these primary observations, subsequent studies used techniques to manipulate the global levels of SUMOylation. Over-expression of SUMO-1 and SUMO-2 increased the resistance to OGD, whereas knockdown of endogenous SUMO-1, but not SUMO-2, increased the susceptibility of SHSY5Y cells and cortical neurons (Lee et al. 2009b). Consistently, SUMO-2/3 silencing in cortical neurons decreased viability following OGD (Datwyler et al. 2011), whilst over-expression of SENP-1 in hippocampal neurons decreased SUMO-1 and SUMO-2/3 conjugation and increased OGD-induced cell death (Cimarosti et al. 2012). In Ubc9 transgenic mice subjected to permanent focal ischemia, Ubc9 levels, and ultimately global SUMO-1 and SUMO-2/3 conjugation, were inversely correlated with the infarct size (Lee et al. 2011). Such studies support a neuroprotective role for SUMOylation in ischemia, although molecular targets and mechanisms were not identified.

Lee et al. (2012) have recently extended their initial study in hibernating ground squirrels by showing that not only SUMOylation but also conjugation by other ubiquitin-like modifiers (ULMs), such as interferon-stimulated gene 15 (ISG15), neural precursor cell expressed, developmentally down-regulated 8 gene, ubiquitin-fold modifier 1, and fau ubiquitin-like protein 1, are increased in the brain of these animals during torpor (Lee et al. 2012). These increases were accompanied by depression of two families of miRNAs (miR-200 and miR182), which are known to have a role in the expression and regulation of ULM modification of target proteins. Inhibition of these miRNA families increased the global conjugation of ULMs to protein targets and improved SHSY5Y cells resistance to cell death observed after OGD, suggesting that SUMOylation is neuroprotective in this system. In fact, conjugation by ISG15 or ISGylation has been previously shown to be increased following focal ischemia (Nakka et al. 2011). In this study, knockout mice lacking either ISG15 or the ligating enzyme ubiquitin-activating enzyme E1-like protein, the first enzyme of the ISGylation cycle, showed increased mortality, exacerbated infarct size and worsened neurologic recovery compared with wild-type animals.

Although previous studies have used over-expression and knockdown of SUMO, Ubc9 and/or SENP in vitro (Lee et al. 2007, 2009b; Datwyler et al. 2011; Cimarosti et al. 2012), as well as Ubc9 transgenic animals (Lee et al. 2011), these data might suggest an artificial regulation of SUMOylation levels. To elucidate the in vivo role of protein SUMOylation in brain ischemia, an interesting advance would be to apply transgenic technology to the enzyme involved in the SUMO regulatory pathway, i.e. SENP/Ubc9 knockouts; however, conditional knockout animals are still pending. Current understanding of Ubc9 function suggests an essential role in normal embryonic development in mammals, as Ubc9-null mouse die during the first developmental stages (Nacerddine et al. 2005). In addition, the demonstration that SUMO-2/3 can compensate for function in SUMO-1 knockout mice indicates that SUMO-1 is dispensable in normal development as well as in later stages of life (Zhang et al. 2008).

SUMO and brain injury mechanisms

Autophagy is a normal cellular process involved in cell death, which is different from ATP-dependent apoptosis, and is characterized by the formation of double-membrane autophagosomes, which are targeted to lysosomes where their contents are digested by hydrolases (Choi et al. 2013). In the ischemic penumbra, cellular death is slower than at the core of the infarct (Branston et al. 1974) and, although the mechanisms of cell death in this particular region remain largely unknown, in vivo studies suggest that autophagic and apoptotic pathways may be involved in both in vivo and OGD paradigms of ischemia (Rami and Kogel 2008; Wen et al. 2008). Studies in neurons using an ‘ischemic solution’ to mimic the ionic, pH, oxygen, and glucose changes seen in disease provided further proof of a simultaneous activation of autophagy and apoptosis during disease states, this process being more damaging to astrocytes than neurons (Pamenter et al. 2012). In this regard, proteins involved in autophagy and apoptosis have been reported to be modified by SUMOylation (Cajee et al. 2012); moreover, a range of DNA-repair proteins are modified by SUMO (Bergink and Jentsch 2009).

Mitochondria can undergo a process of fission, a mechanism responsible for increasing mitochondrial number and capacity as well as enabling ‘mitophagy’, the cellular mechanism responsible for the recycling of damaged mitochondria through autophagy by lysosomes. Recent literature has suggested that mitophagy is involved in ischemic injury and contributes to stroke damage (Liu et al. 2012; Pamenter et al. 2012). Harder et al. (2004) provided the first evidence of the involvement of SUMOylation in mitochondrial fission, identifying dynamin related protein 1 (DRP1), a controller of the final step of mitochondrial fission, as a SUMO-1 target (Harder et al. 2004). SUMO-1 was also proposed to protect DRP1 against degradation (Harder et al. 2004). Following this study, SENP-5 was also shown to contribute to the modulation of mitochondrial fission (Zunino et al. 2007); ablating SENP-5 led to an increase in the production of reactive oxygen species (ROS) as well as enhancing mitochondrial fission. These studies suggest that SUMOylation might play an important role in the modulation of ROS production and, thus, regulation of the oxidative stress response, through the maintenance of normal mitochondrial functioning and recycling. Such SUMO-mediated regulation of mitochondrial function in cardiac cells, including studies in ischemic/reperfusion injury models, has been proposed to be cardioprotective (Zungu et al. 2011).

At present, studies demonstrating the contribution of SUMOylation to cell death in CNS disease are limited. Odagiri et al. (2012) suggested autophagy and SUMOylation were involved in the formation of Marinesco bodies. These nuclear inclusions, present in several conditions such as chronic hypoxic encephalopathy, were shown to associate with SUMO-1 and SUMO-2 as well as other proteins involved in autophagy (Odagiri et al. 2012). The non-covalent interaction of SUMO-1 with parkin, a gene encoding E3 ubiquitin ligase and associated with cell death in Parkinson's disease, was reported to increase parkin translocation to the nucleus and auto-ubiquitynation activity (Um and Chung 2006). In addition, a SUMO-mediated increase in ubiquitynation of the parkin binding partner Ran binding protein 2 (RanBP2) has been shown to lead to the degradation of RanBP2 (Um et al. 2006). Recent in vitro studies showed degradation of SENP-3, a SUMO-2/3-specific protease, during OGD via a cathepsin B-dependent mechanism (Guo et al. 2013); decreased levels of SENP-3 allowed for prolonged DRP1 SUMOylation, which, in turn, suppressed DRP1-mediated cytochrome c release as well as caspase-mediated cell death. These studies suggest that SUMOylation can be neuroprotective through increased resistance to ischemia-induced apoptosis.

Ischemic events can activate microglia and astrocytes to cause inflammation, moreover, SUMOylation may also be a protective mechanism for glia as well as neurons; indeed, we have recently reported that SUMO-1 over-expression prevents beta amyloid induction of a reactive astrocyte phenotype (Hoppe et al. 2013). In astrocytes, the CCAAT/enhancer-binding protein beta transcription factor, which regulates expression of proinflammatory nitric oxide synthase type 2, is a target for SUMO (Akar and Feinstein 2009) and SUMOylation of nuclear receptors LXRalpha and LXRbeta can protect against inflammatory responses (Lee et al. 2009a). These studies suggest that astrocytes represent an important cellular target for SUMO within the tripartite synapses and that SUMOylation may be protective against brain inflammation because of astrocyte reactivity.

Potential SUMO targets in ischemia

The identification of SUMO targets under physiological conditions is surrounded by confounding issues, such as the fact that SUMOylation is highly labile and that only a small percentage of the total protein is modified at any one time. Despite these challenges, more than hundred SUMO targets have been characterized and many more potential targets have been identified (Da Silva-Ferrada et al. 2012; Becker et al. 2013). A recent study using quantitative proteomics from SUMO-3 over-expressing B35 cells exposed to OGD identified several SUMO-3 targets, mostly nuclear proteins involved in gene expression, which, despite not yet being validated in neurons, provide a valuable resource for future investigations (Yang et al. 2012).

It is important to note that, although the overall effect of SUMOylation in conditions of severe cell stress, such as brain ischemia, is regarded as neuroprotection, the exact effect of SUMOylation on specific SUMO targets depends on the target. Similarly to phosphorylation, the functional consequences of SUMOylation on particular proteins are expected to vary greatly: some of these when SUMOylated could protect neurons from cellular death, while others could aggravate the pathological process. In this context, promoting a general change in the global levels of SUMOylation is unlikely the most effective way of managing brain ischemia, with the specific targeting of SUMO substrates providing a more attractive alternative.

SUMOylation was first shown to modify nuclear proteins and is better known for its direct involvement in transcriptional processes, with several SUMO targets being transcription factors (reviwed in Hay 2005; Geiss-Friedlander and Melchior 2007). SUMO exerts its effects on these targets in various ways, e.g. SUMOylation is involved in converting a transcription factor from an activator to a repressor or in regulating the localization and/or stability of numerous transcription factors. While the regulatory roles for SUMOylation in the nucleus have been previously reviewed (Seeler and Dejean 2003; Heun 2007), some examples of key transcription factors and signaling kinases targeted by SUMO, crucial in the pathophysiological mechanisms underlying brain ischemia, are discussed below.

Important extranuclear roles for SUMOylation have began to be unraveled, with several cytosolic and plasma membrane proteins reported to be SUMOylated (reviewed in Wilkinson et al. 2010) and the necessary machinery shown to be present at the plasma membrane (Plant et al. 2010). These include key CNS transporters, ion channels and receptors, strongly implicating SUMOylation in the control of activity dependent neuronal metabolism, glucose homeostasis and coping strategies for excitotoxicity, which could mediate the neuroprotective effects attributed to SUMOylation in brain ischemia (Fig. 1, Table 1).

Table 1. Potential functional consequences of SUMOylation of key CNS transporters, channels and receptors in brain ischemia
Membrane ProteinSUMO isoformPutative modulation siteProposed SUMOylation effectHypothetic role in brain ischemiaExample Referencesa
  1. a

    Because of space restrictions, only references that were also included in the main text are listed.

Glucose TransportersGLUT1SUMO-1UnknownUbc9 decreased GLUT1 protein levelsSUMO-mediated up-regulation of GLUTs could help maintaining the energy supply to the brainGiorgino et al. 2000;Liu et al. 2007;
GLUT4SUMO-1UnknownUbc9 increased GLUT4 protein levels and retarded GLUT4 turnover
Glutamate TransportersEAAT2SUMO-1K571Targeting of SUMOylated caspase 3 cleaved C-terminal fragment of EAAT2SUMO up-regulation of EAAT2 could increase glutamate clearance and decrease excitotoxicity

Chao et al. 2010;

Gibb et al. 2007;

K+ ChannelsK2P1SUMO-1K274SUMOylation led to silencing of the channelRole as yet unclear: inhibition of K2P1 could be modulated by nuclear events

Es-Salah-Lamoureux et al. 2010;

Feliciangeli et al. 2007;

Plant et al. 2010;

Plant et al. 2012;

Rajan et al. 2005;





K221 or K536De-SUMOylation led to hyperpolarizing shift in the voltage-dependence of steady-state inactivationKnocking down Kv1.5 channels elicited tolerance without pre-conditioning; SUMO-mediated inactivation could be neuroprotective

Benson et al. 2007;

Stapels et al. 2010;





K470Inhibition of current: differential effects on channel gating proposedAs Kv2.1 channels are pro-apoptotic, SUMO-mediated inhibition could be neuroprotective

Dai et al. 2009;

Plant et al. 2011;

Pal et al. 2003;

Kainate ReceptorsGluK2SUMO-1K886Internalization of SUMOylated GluK2SUMOylation-induced down-regulation of GluK2 could decrease excitotoxicity (less GluK2 at the post-synaptic membrane)

Feligioni et al. 2009;

Konopacki et al. 2011;

Martin et al. 2007;

Pei et al. 2006;

or increase neurotoxicity (activation of the MLK3-JNK3 pathway)Zhu et al. 2012;
Metabotropic Glutamate ReceptorsmGluR8SUMO-1K882UnknownmGluRs modulation of excitatory synaptic transmission could be neuroprotective

Caraci et al. 2012;

Dutting et al. 2011;

Cannabinoid ReceptorsCB1SUMO-1UnknownSUMOylation of CB1 involved in regulation of p53Inactivation of p53, an agent shown to potentiate ischemic neuronal deathGowran et al.2009;Yonekura et al. 2006
Figure 1.

Key CNS transporters, ion channels and receptors reported to be small ubiquitin-like modifier (SUMO) targets and the potential functional consequences of SUMOylation. Although the function of SUMOylation and the identity of disease-modified SUMO substrates remain in most cases unknown, potential membrane protein targets are shown. Glucose transporters (GLUT1 and GLUT4), different K+ channels subtypes (KV1.5, K2P1 and KV2.1) and GluK2-containing kainate receptors (GluK2), which have been reported to both increase and decrease transmitter release at different synapses, are modulated by SUMOylation. In addition, metabotropic glutamate receptors (mGluR8) and cannabinoid receptors (CB1), which can inhibit pre-synaptic release of transmitters such as glutamate, together with glial glutamate transporters (EAAT2), have been identified as SUMO targets. Noticeably, there are still many unanswered questions regarding not only the physiological role of SUMOylation but also its role in relation to altered protein dynamics associated with brain ischemia. Example references to all the processes illustrated can be found in the main text.

Transcription factors

SUMOylation directly modifies a number of the hypoxia signaling cascade mediators required to allow cells to adapt to low oxygen levels, including the key transcriptional factor hypoxia inducible factor-1 (HIF-1), which regulates genes involved in angiogenesis, cell survival and glucose metabolism (Nunez-O'Mara and Berra 2013). HIF-1 consists of an oxygen-regulated α subunit and a constitutively expressed β subunit and HIF-1α, HIF-2α and HIF-1β have been shown to be post-translationally modified by SUMO-1 and/or SUMO-2/3 (Tojo et al. 2002; Shao et al. 2004; Berta et al. 2007; van Hagen et al. 2010). Most of the research so far has focused on the consequences of SUMOylation on the stability/activity of HIF-1α with controversial findings being reported (Bae et al. 2004; Berta et al. 2007; Carbia-Nagashima et al. 2007; Cheng et al. 2007; Xu et al. 2010). The direction in which SUMOylation modifies HIF-1α stability/activity remains to be fully clarified. One explanation for these contrasting observations could be because of the use of over-expression systems, which do not account for an indirect effect of SUMOylation on other proteins other than HIF-1α. In fact, it is important to note that not only HIF but also other constituents of the hypoxia signaling cascade are SUMO targets, e.g. CREB-binding protein/p300 coactivator and the ubiquitin E3 ligase Von Hippel–Lindau tumor suppressor (Girdwood et al. 2003; Kuo et al. 2005; Huang et al. 2009; Cai et al. 2010).

In addition to enzymatic regulation of SUMO conjugation to target proteins described above, ROS, generated by oxidative stress, have been shown to regulate the SUMO pathway and, more specifically, to have effects on transcription factors (Bossis and Melchior 2006). Exposure to low concentrations of ROS induced de-SUMOylation of several transcription factors, including c-Jun, a key component in the oxidative stress response. In cells undergoing different types of stress, including heat, the heat shock factor 1 (HSF1), which occurs physiologically in a non-DNA-binding form, is converted into a DNA-binding form, promoting the transcription of heat shock genes (Hilgarth et al. 2003). Stress-induced HSF1 modification by SUMO-1 acting at lysine residue 298 (K298) has been shown to increase its DNA binding capacity (Hong et al. 2001). Phosphorylation has been shown to be essential for HSF1 stress-induced SUMOylation (Hietakangas et al. 2003), suggesting an interplay between these two post-translational modifications in HSF1 modulation, which was subsequently confirmed by the discovery of a phosphorylation-dependent SUMOylation motif in HSF1 (Hietakangas et al. 2006). Contradicting previous findings by Hong et al. (2001), this study showed that SUMOylation does not interfere with the capacity of HSF1 to bind to DNA, as the HSF1 SUMOylation-deficient K298R mutant was able to bind to heat shock genes (Hietakangas et al. 2003). HSF1 can also be modified by SUMO-2/3 and can also be SUMO-2/3-ylated; stressed cells show nuclear accumulation of heat-shock protein 27 bound to HSF1 (Brunet Simioni et al. 2009). SUMO-2/3 modification blocks HSF1 transactivation capacity, but maintains HSF1 DNA-binding ability. A protective role against ischemic insults has been proposed for HSF1. In transgenic mice over-expressing HSF1, apoptosis and infarct size in cardiac tissue were reduced following ischemia/reperfusion (Zou et al. 2003). Silencing HSF1 in vivo abrogated the cardioprotection induced by whole-body hyperthermia pre-conditioning (Yin et al. 2005). The underlying mechanisms of HSF1-mediated protection remain incomplete and, despite the controversy surrounding current findings, SUMOylation-induced increase in HSF1 DNA-binding capacity is a prominent premise to be further explored.

SUMOylation has been shown to indirectly modulate nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB), through the interaction with inhibitor of κB (IκB). Under normoxic conditions, NF-κB is kept inactive in the cytoplasm through association with IκB, while under hypoxic conditions IκB undergoes phosphorylation and ubiquitynation-mediated degradation, allowing NF-κB activation and translocation to the nucleus (Schneider et al. 1999). IκBα is a SUMO-1 target and SUMO-1 conjugation prevented the ubiquitynation-induced degradation of IκBα, thus limiting NF-κB activation (Desterro et al. 1998). Recently, IκBα has been shown to be also a SUMO-2/3 target and, by contrast to SUMO-1 conjugation, SUMO-2/3 conjugation facilitated NF-κB dissociation from IκBα, leading to NF-κB activation in response to an ischemic event (Culver et al. 2010). The role of NF-κB in ischemia is still controversial. NF-κB has five known subunits, p50, p52, RelA, RelB and cRel, with p50 being the best characterized (Schneider et al. 1999). In transient and permanent focal ischemia models, reduced infarct size was observed in knockout mice for NF-κB p50 subunit, suggesting a pro-apoptotic role for NF-κB (Schneider et al. 1999; Nurmi et al. 2004). Contrary to these studies, in a permanent focal ischemia model, increased infarct size as well as apoptotic cell death were observed in knockout mice for NF-κB p50 subunit, suggesting a neuroprotective role for NF-κB (Li et al. 2008).

Signaling kinases

The glycogen synthase kinase-3β (GSK-3β) is a constitutively active serine/threonine kinase abundant in the CNS, where it regulates many intracellular signaling pathways, including those involved in cell death (Phukan et al. 2010). GSK-3β plays a pivotal role in brain ischemia: its inactivation has been proposed as a mechanism to promote neuronal survival and its inhibition has been shown to reduce infarct size and suppress neuroinflammation (Liang and Chuang 2007; Collino et al. 2008; Zhou et al. 2011). SUMOylation on K292 of GSK-3β was suggested to be a regulatory mechanism, promoting its nuclear localization and protein stability, decreasing its kinase activity and resulting in the stimulation of cell apoptosis (Eun Jeoung et al. 2008). A subsequent study identified estrogen receptor β (ERβ) as a SUMO-1 target and showed that phosphorylation of ERβ by GSK-3β maximizes its SUMOylation in response to hormone (Picard et al. 2012). SUMO-1 conjugation to ERβ increases its stability by competing with ubiquitin at the same acceptor site and thus preventing ERβ degradation. ERβ plays important roles in estrogen-mediated neuroprotection and the activation of ERβ has been shown to reduce brain damage and blood–brain barrier breakdown following ischemic injury (Carswell et al. 2004; Shin et al. 2013).

The c-Jun N-terminal kinase (JNK) regulates a range of biological processes implicated in neurodegenerative disorders, essential for mediating the apoptotic response of neurons to ischemia-associated stress (Davies and Tournier 2012). JNK, in its phosphorylated state, has been suggested as a SUMO-1 target (Feligioni et al. 2011). Over-expression of SUMO-1 promoted JNK activation and exacerbated cellular death, whereas over-expression of the de-SUMOylating protease SENP-1 inhibited JNK activation and rescued SHSY5Y cells from H2O2-induced death (Feligioni et al. 2011). By contrast, over-expression of SUMO-1 attenuated heat-shock induced ROS generation in HEK293 and HeLa cells, through a decrease in p38 and JNK activation, protecting cells against heat-shock induced cell death (Kim et al. 2011). This protection was also mediated through SUMOylation of NADPH oxidase 2 at the cell membrane, which decreased the activity of the enzyme involved in intracellular ROS generation. Moreover, knockdown of smt3 gene, the only known SUMO gene in Drosophila, has been shown to increase the activity of JNK and induce cell death in vivo (Huang et al. 2011). The SUMO E3-ligase protein inhibitor of activated STAT1 (PIAS1) has a bidirectional role in regulating JNK-dependent oxidative stress responses: PIAS1 is a target as well as a negative regulator of the JNK pathway (Leitao et al. 2011). Whether SUMOylation-mediated regulation of JNK plays a role in brain ischemia, or not, remains to be investigated.

Responses to oxidative stress (see above) may also involve SUMO modification of signaling kinases. SUMOylation of the proapoptotic kinase homeodomain-interacting protein kinase 2 (HIPK2) was shown to protect against ROS-induced acetylation of HIPK2; conversely, increased ROS acted to block SUMOylation (de la Vega et al. 2012). Such findings suggest that inter-dependent dynamic regulation of SUMO/ROS can act as a molecular switch during oxidative stress.


The essential exchange of glucose between the blood and the cytoplasm of the cells is mediated by glucose transporters (GLUTs), which are highly expressed in most cell types, including neurons. GLUT1 and GLUT4 interaction with Ubc9 in skeletal muscle cells provided the first evidence that membrane proteins could undergo SUMOylation (Giorgino et al. 2000). Over-expression of Ubc9 decreased GLUT1, but increased GLUT4, protein levels. As GLUT1 is constitutively present at the cell surface, whilst GLUT4 is targeted to the membrane in response to insulin, it was suggested that SUMOylation decreases basal glucose transport and potentiates insulin responsiveness. A subsequent study confirmed that Ubc9 can regulate the subcellular targeting and up-regulate GLUT4 in adipocytes (Liu et al. 2007). Although these reports demonstrated regulation of GLUTs by Ubc9, whether SUMOylation per se is involved was not investigated. In fact, the mechanism by which Ubc9 up-regulates GLUT4 may be independent of the conjugating activity of Ubc9, since an inactive mutant showed identical effects to those of wild-type Ubc9 (Liu et al. 2007). Moreover, the mechanisms controlling the opposing effects of Ubc9 on GLUT1 and GLUT4 remain to be elucidated. Given the abundant expression of GLUT1 and GLUT4 in neurons, an interesting possibility is that their regulation by Ubc9 could modulate glucose uptake in the brain. The SUMO-mediated up-regulation of GLUT4, for example, could help maintain the energy supply to the brain by enhancing GLUT4 function, thus ameliorating the infarct outcome.

Another transporter that has been suggested to be SUMOylated is the astroglial glutamate transporter EAAT2 (excitatory amino acid transporter 2 a.k.a. GLT-1 in rodents), vital for the clearance of synaptically released glutamate and the prevention of excitotoxicity. In a mouse model of amyotrophic lateral sclerosis, a fragment of EAAT2 cleaved by caspase 3 has been shown to be conjugated by SUMO-1 (Gibb et al. 2007). The SUMOylated EAAT2 fragment is targeted to promyelocytic leukemia nuclear bodies, whose proposed functions include the regulation of gene transcription, leading the authors to suggest a possible novel mechanism by which EAAT2 could contribute to the pathology of amyotrophic lateral sclerosis. More recent evidence suggests that the SUMOylated EAAT2 fragment accumulates within nuclei of spinal cord astrocytes (Foran et al. 2011). However, the functional implications of SUMOylating caspase 3 cleaved-EAAT2 remain unknown. Further investigation is also required to determine whether the entire transporter is SUMOylated and whether SUMOylation of EAAT2 is specific to amyotrophic lateral sclerosis. SUMOylation of full-length EAAT2 could have an alternative effect than that demonstrated for fragmented EAAT2. Considering that up-regulation of EAAT2, and consequential increases in glutamate uptake, has been shown to be neuroprotective in brain ischemia (Chao et al. 2010), it will be important to elucidate the effects of EAAT2 SUMOylation on ischemia-induced excitotoxicity.

Ion channels

Currently, studies of effects of SUMOylation on ion channels are largely limited to potassium (K+) channel isoforms, which are crucial in excitable neurons, where they help shape the action potential and set the resting membrane potential. K2P channels regulate background leak currents, acting to stabilize neuronal excitability. In the plasma membrane of Xenopus oocytes, SUMOylation silences K2P1 channels, whereas de-SUMOylation restores channel activity (Rajan et al. 2005). Mutation of a C-terminal lysine residue in the putative SUMOylation site prevented K2P1 SUMOylation and rendered the channel constitutively active, suggested to explain why K2P1 functional expression had not been reported previously. These findings were challenged by a subsequent study that failed to observe the characteristic band shift by immunoblotting, necessary to corroborate the proposed SUMOylation of K2P1 (Feliciangeli et al. 2007). However, effects were later substantiated, including the demonstration that SUMOylation of one of the two lysines present in wild-type K2P1 channels (formed as subunit dimers) was sufficient to silence the channel (Plant et al. 2010). More recent study by Plant et al. (2012) has shown that K2P1 subunits form heterodimeric functional channels with the K2P3 or K2P9 subunits of the two-P domain, acid-sensitive K+ channel in cerebellar granule neurons, and that SUMOylation of K2P1 suppresses the activity of these channels. Consistent with previous findings (Plant et al. 2010), one SUMO monomer was sufficient to silence channels with mixed subunits containing K2P1 (Plant et al. 2012).

Voltage-dependent KV1.5 channels have also been shown to be SUMOylated by all three SUMO isoforms, both in vivo and in vitro (Benson et al. 2007). Blockage of SUMOylation, either by mutation of the two SUMO-modifiable lysines or over-expression of SENP-2, led to a hyperpolarizing shift in the voltage dependence of steady-state inactivation, but had no effect on current amplitude or activation properties. Another voltage-dependent K+ channel, KV2.1 has been reported to be modified by SUMOylation. In HEK293 cells, direct infusion or over-expression of SUMO-1 resulted in KV2.1 current inhibition, an effect that was enhanced by Ubc9 and reversed by SENP-1 (Dai et al. 2009). Furthermore, over-expression of SUMO-1 in human β-cells or insulinoma cells inhibited native KV currents and caused a decrease in cell excitability in the latter, ascribed to a SUMO-induced increase in KV2.1 inactivation kinetics and a slowing of recovery from inactivation. Further studies have shown KV2.1 to be SUMOylated at a C-terminal lysine residue (K470) and, by contrast to the study by Dai et al. (2009), SUMOylation caused a positive shift in the voltage-dependence of activation, consistent with promotion of closed gating states (Plant et al. 2011). In this study, SUMO-1-3 increased and SENP-1 decreased cellular excitability in hippocampal neurons; such SUMO effects could be reproduced by over-expression of wild-type KV2.1, but not a KV2.1 K470Q mutant channel (Plant et al. 2011).

Although the above studies convincingly demonstrate that different K+ channels can be experimentally modulated by SUMOylation, it is evident that modulation is reported to occur by diverse effects on channel gating and, therefore, that precise effects of SUMOylation on neuronal excitability are likely to be dependent on ion channel complement within specific neurons. Moreover, functional consequences for such effects under pathophysiological conditions warrant further investigation. In particular, SUMO regulation of K+ channels in ischemia is likely to be complex. For example, agents that activate K2P leak channels are generally considered neuroprotective in ischemia (Es-Salah-Lamoureux et al. 2010). However, KV1.5 encoding gene Kcna5 expression has been shown to be decreased in ischemic-tolerant brains (Stapels et al. 2010) and, as KV2.1 channels have been shown to initiate apoptosis (Pal et al. 2003), the data above are consistent with a neuroprotective role for SUMO. The presence of consensus motifs for SUMOylation in multiple ion channel families raises the hypothesis that SUMOylation could be a widespread mechanism of ion channel regulation, also warranting further investigation.


Kainate receptors (KARs) are ionotropic glutamate receptors widely expressed in the CNS, where they regulate pre- and post-synaptic neuronal responses. Agonist-induced activation causes SUMOylation of GluK2, a KAR subunit, in neurons, resulting in receptor internalization (Martin et al. 2007). Internalization led to decreased KAR-mediated excitatory post-synaptic currents in hippocampal slices, which were enhanced by de-SUMOylation. Recently, protein kinase C phosphorylation of GluK2 has been shown to be necessary for SUMOylation-dependent internalization of GluK2-containing KARs (Konopacki et al. 2011). At a functional level, SUMOylation was shown to increase KAR-evoked glutamate release in synaptosomes, while decreasing release evoked by KCl or α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) activation (Feligioni et al. 2009). Such studies point to complex, possibly differential, regulation of pre- and post-synaptic receptors and to the potential involvement of further SUMO targets in transmitter release. As KARs have been implicated in ischemic neuronal death, with knockdown of GluK2 being neuroprotective against OGD (Pei et al. 2006), SUMOylation-induced reduction in GluK2-mediated excitation could contribute to neuroprotection. It has recently been shown that global brain ischemia evokes a sustained elevation of GluK2 SUMOylation and SUMOylation-induced endocytosis of GluK2-containing kainate receptors (Zhu et al. 2012). However, this increased internalization of GluK2 promoted its binding with mixed lineage kinase 3 (MLK3) and activation of the MLK3-JNK3 pathway, which may be responsible for ischemic neurotoxicity. More recently, over-expression of either dominant negative Ubc9 or SENP-1 was shown to prevent chemical long-term potentiation (LTP)-induced increase in AMPAR surface expression (Jaafari et al. 2013). Despite AMPAR subunits not being identified as SUMO targets, this study suggested SUMOylation as an indispensable mechanism for AMPAR trafficking during LTP.

The metabotropic glutamate receptors (mGluRs) are another potential SUMO target widely expressed in the CNS. Group III mGluRs, including mGluR4, mGluR6, mGluR7, and mGluR8, are G-protein coupled receptors, which can modulate pre-synaptic transmitter release. A recent study has shown that Group III mGluRs, more specifically mGluR8b, can be modified by SUMO-1 in vivo (Dutting et al. 2011). As mGluRs modulate excitatory synaptic transmission and are linked to processes of neuroprotection (Caraci et al. 2012), further work is needed to define the functional consequences of SUMOylating these receptors.

Cannabinoid CB1 receptors are one of the most abundant G-protein coupled receptors in the brain and are a SUMO substrate. Interestingly, CB1 is basally SUMOylated, with agonist activation leading to de-SUMOylation (Gowran et al. 2009). In the same study, CB1 agonist activation led to de-SUMOylation of the tumor suppressor protein p53. As p53 has been shown to potentiate ischemic neuronal death (Yonekura et al. 2006), CB1-mediated SUMO-modulation of p53 might contribute to its stability and activity. Although further studies are required, SUMOylation of CB1 may play a role in the pathophysiology of stroke, not least through the regulation of p53, but also through other processes, such as excitotoxicity, modulated by CB1.

Concluding remarks

SUMOylation is an important regulatory mechanism for the function and fate of multiple proteins, including a number of nuclear and extranuclear proteins. SUMOylation at the cell membrane is involved in the reversible control of CNS transporters, ion channels, and receptors, constituting a unique mechanism for adaptive tuning of metabolic, electrical and other requirements of the cell (Fig. 1). A growing body of evidence supports the hypothesis that SUMOylation might be part of an endogenous neuroprotective response in situations of cellular stress, especially in brain ischemia. However, it is important to keep in mind that data from Ubc9 transgenic animals, as well as ectopic SUMO and/or SENP expression, might not account for artificial/unphysiological effects on protein SUMOylation and that in vivo data with conditional loss of function of SUMOs, Ubc9 and SENPs, key proteins in the SUMOylation pathway, are still lacking. As knowledge of the molecular targets and mechanisms involved in SUMOylation remain incomplete, basic research into the functional consequences of SUMOylation of the proteins reviewed here, and further potential targets, under physiological and pathophysiological conditions, may translate into future therapeutics for brain ischemia.


We apologize to all researchers whose work could not be appropriately discussed or cited because of space limitations. This study was supported by a grant from the Royal Society. V.S. is recipient of a University of Reading Research Endowment Trust Fund PhD studentship. The authors have no conflicts of interest to declare.