The bioactive sphingolipids ceramide, sphingosine and sphingosine-1-phosphate (S1P) are important signalling molecules that regulate a diverse array of cellular processes. Most notably, the balance of the levels of these three sphingolipids in cells, termed the ‘sphingolipid rheostat’, can dictate cell fate, where ceramide and sphingosine enhance apoptosis and S1P promotes cell survival and proliferation. The sphingosine kinases (SKs) catalyse the production of S1P from sphingosine and are therefore central regulators of the sphingolipid rheostat and attractive targets for cancer therapy. Two SKs exist in humans: SK1 and SK2. SK1 has been extensively studied and there is a large body of evidence to demonstrate its role in promoting cell survival, proliferation and neoplastic transformation. SK1 is also elevated in many human cancers which appears to contribute to carcinogenesis, chemotherapeutic resistance and poor patient outcome. SK2, however, has not been as well characterized, and there are contradictions in the key physiological functions that have been proposed for this isoform. Despite this, many studies are now emerging that implicate SK2 in key roles in a variety of diseases, including the development of a range of solid tumours. Here, we review the literature examining SK2, its physiological and pathophysiological functions, the current knowledge of its regulation, and recent developments in targeting this complex enzyme.
Sphingolipids have emerged as important signalling molecules that can regulate a vast range of cellular processes, including cell survival, proliferation, migration, differentiation and inflammation [1-3]. Although many enzymes are involved in regulating the relative levels of the sphingolipids, the sphingosine kinases (SKs) SK1 and SK2 are of particular interest as their activity can be dynamically regulated by external stimuli . As a result, altering the levels or activity of the SKs can play a key role in controlling cell fate, making them attractive targets in the development of cancer therapeutics [4-9]. Indeed, given its unequivocal role in promoting cell survival and proliferation, targeting SK1 has proved effective in attenuating tumour growth and progression in a range of cancer models (reviewed in ). However, investigating the therapeutic potential of targeting SK2 in cancer has only very recently gained momentum, and although the findings are promising, the mechanisms behind SK2 function and regulation remain poorly understood. Indeed, the current literature focuses largely on SK1, and although the exact cellular roles of SK2 are yet to be deciphered, what is currently known about this isoform suggests that its functions and mechanisms of regulation are extremely complex. Here, we review what is currently known about SK2 function and regulation in both physiological and pathophysiological settings, and the recent developments made in targeting this SK isoform.
Sphingosine-1-phosphate (S1P) and sphingolipid signalling
S1P is a biologically active phospholipid that is derived from a network of sphingolipid synthesis and degradation pathways . Ceramide resides at the centre of this network and can be deacylated by ceramidase to form sphingosine, which can in turn be phosphorylated by the SKs to produce S1P (Fig. 1). The aforementioned reactions are reversible and the only exit point from the cycle is the irreversible degradation of S1P to hexadecenal and phosphoethanolamine. S1P acts as both an extracellular and intracellular signalling molecule through a range of different pathways . It can be transported out of the cell where it exists in high nanomolar concentrations in the blood [11, 12]. This export is mediated by members of the ABC transporter family [13, 14] and, more recently noted, by spinster homolog 2 . Upon exit from the cell, S1P can bind to and activate a family of five G-protein coupled receptors (GPCRs), referred to as S1P1–5 , to elicit autocrine or paracrine signalling. Interestingly, recent evidence suggests that S1P may engage its GPCRs by lateral diffusion within the outer leaflet of the plasma membrane rather than directly from the extracellular space . The S1P receptors are coupled to various heterotrimeric G proteins, which subsequently activate a number of downstream signalling pathways via the modulation of extracellular-signal-regulated kinases 1/2 (ERK1/2), small GTPases (Rac and Rho), the phosphoinositide 3-kinase (PI3K)/AKT pathway and phospholipase C (PLC), ultimately modulating cell proliferation, survival and migration [6, 18]. These responses can vary in a cell- or tissue-specific manner as a result of the differential expression of the S1P receptors and the various G proteins they couple to .
In addition to its extracellular signalling roles, S1P can also act directly on a number of intracellular targets. Specifically, S1P has been shown to directly inhibit histone deacetylase (HDAC) 1/2 activity , activate the E3 ubiquitin ligase activity of tumour necrosis factor (TNF) receptor-associated factor 2 (TRAF2) , interact with prohibitin 2 to mediate mitochondrial respiration  and possibly modulate the activity of p21-activated kinase 1 . A recent study also suggested that S1P may contribute to cytochrome c release from mitochondria via modulation of BAK, although the direct interaction of S1P with BAK was not described . Through the activation of these varied signalling pathways, S1P can act to promote cell survival, proliferation and migration, as well as regulate differentiation, angiogenesis and inflammation [6, 25, 26]. Interestingly, an intracellular role for the S1P5 GPCR has also been proposed, where the colocalization of the receptor and SKs to centrosomes may indicate a novel function in regulating cell division .
In contrast to S1P, the sphingolipids ceramide and sphingosine, which both lie directly upstream of S1P in the sphingolipid biosynthetic pathway (Fig. 1), are promoters of apoptosis. Ceramide has been found to directly activate a number of protein targets that appear to mediate its pro-apoptotic functions, including protein phosphatases 1 , 2A  and 2C , protein kinase C (PKC) ζ [30, 31], the kinase suppressor of ras  and cathepsin D . Furthermore, sphingosine can bind to and initiate the phosphorylation and inactivation of the pro-survival adaptor protein 14-3-3 . Given these opposing functions, the current dogma is that a delicate balance must be maintained between the relative levels of pro-apoptotic ceramide and sphingosine and pro-survival S1P, with this equilibrium, referred to as the ‘sphingolipid rheostat’, playing an important role in dictating cell fate [6, 35]. SK is a central regulator of this equilibrium, residing at the critical junction between pro-proliferative, pro-survival S1P and pro-apoptotic sphingosine and ceramide (Fig. 1). Therefore, understanding how SKs are regulated is important to understand their role not only in controlling cell fate but also in aberrant survival and proliferative signalling in disease states like cancer. Indeed, there are a multitude of studies that implicate S1P and the SKs in cancer development, survival and metastasis, and this field has been extensively reviewed [4-9, 36, 37].
Two SKs exist in mammalian cells, namely SK1 and SK2. In humans they are generated from two distinct genes, SPHK1 and SPHK2, which are located on chromosome 17 (17q25.2) and 19 (19q13.2), respectively. Although human SK1 and SK2 vary considerably in size (384 and 618 amino acids for SK1a and SK2a, respectively) , they share 80% similarity and 45% overall sequence identity, with almost all of the polypeptide sequence of SK1 aligning with regions of the larger SK2. All known eukaryotic SKs share five highly conserved regions within their sequence (termed C1–C5; see Fig. 2), which appear to encompass the regions necessary for ATP binding and catalysis . The additional residues present in SK2 result from both its extended N-terminus and an additional central proline-rich region not found in SK1 or in any other protein (Fig. 2). While there are currently no crystal structures available to allow for comparisons between the two SKs, the central proline-rich region of SK2 appears to coincide with the sphingosine binding region of these enzymes . Consistent with this, biochemical information regarding substrate specificity and inhibition of the two SKs suggests that significant differences exist in their substrate binding pockets. As discussed below, SK2 is more promiscuous than SK1 in the substrates it can phosphorylate, and several isoform-selective inhibitors that target the sphingosine binding site of these enzymes have been developed [47, 48]. Surprisingly, due to the high degree of sequence conservation in the putative ATP binding regions, a recent study has also reported an ATP competitive inhibitor that shows high selectivity for SK1 , suggesting that at least some divergence also occurs between SK1 and SK2 in this pocket. Experimental evidence has demonstrated that the single point mutant G212E completely abolishes SK2 catalytic activity  just like the comparable G82D mutation in SK1 , via interfering with ATP binding .
While the activity of both SK isoforms can be enhanced by various cytokines and growth factors (see below), both enzymes also possess intrinsic catalytic activity independent of eukaryotic post-translational modifications that results in cellular SK activity even in the absence of agonist stimulation [52-54]. This basal SK activity has been proposed to facilitate a housekeeping role in maintaining cellular sphingosine and ceramide levels .
Although both SK isoforms are ubiquitously expressed in all human tissues, some differential expression is apparent, with SK1 most highly expressed in lung, spleen and leukocyte  whereas SK2 is highest in the kidney and liver . Furthermore, while both SKs are expressed throughout embryonic development in the mouse, SK1 expression peaks earlier (E7–E11), with SK2 being more strongly expressed in the later stages (E15–E17) .
Roles and regulation of SK1
Of the two SKs, SK1 is by far the most well studied. SK1 has been widely described as a pro-survival, pro-proliferative enzyme, demonstrated initially by studies showing that overexpression of SK1 increases cell survival and proliferation  and induces neoplastic transformation . Numerous studies have implicated SK1 in cancer development and progression, where expression levels of SK1 are found to be upregulated in a number of human solid tumours [59-65] and higher SK1 expression is correlated with poor patient prognosis [62, 63, 66-72]. Notably, targeting SK1 by chemical inhibition or genetic ablation has successfully reduced tumour growth in mice [59, 71, 73-80]. The oncogenic signalling mediated by SK1 is dependent upon its activation and translocation to the plasma membrane . SK1 is a cytoplasmic protein that, upon agonist stimulation, can be phosphorylated by ERK1/2 at Ser225, which results in a 14-fold increase in catalytic activity and also facilitates the transport of SK1 to the plasma membrane via binding to calcium- and integrin-binding protein 1 [82, 83]. Interestingly, other phosphorylation-independent mechanisms of SK1 translocation to the plasma membrane have also been documented [84, 85]. At this location, SK1 catalyses the formation of S1P from plasma-membrane-associated sphingosine, which appears to facilitate both the efficient release of S1P to act extracellularly on cell surface S1P receptors and the interaction with intracellular targets to mediate downstream signalling and promote cell survival, proliferation, migration, differentiation, angiogenesis and inflammation [2, 3, 6].
Roles of SK2
SK2-induced apoptosis and growth arrest
Contrary to the roles of SK1 in pro-proliferative, pro-survival signalling, many of the early studies examining SK2 function found that its overexpression induced cell cycle arrest and apoptosis [49, 86, 87]. A putative BH3 domain was identified within SK2, which has been proposed to mediate its pro-apoptotic functions, at least in part, through interaction with Bcl-xL, a pro-survival Bcl-2 family member . Since this SK2–Bcl-xL interaction was only demonstrated following SK2 overexpression, however, the physiological significance of the proposed association remains unclear. More recently, mitochondrial localization of SK2, and specifically S1P generation at this site, was shown to contribute to BID-mediated activation of BAK, and subsequent mitochondrial membrane permeabilization and cytochrome c release . Another recent study also demonstrated that nuclear S1P and dihydro-S1P produced specifically by SK2 inhibited HDAC1/2 activity, leading to increased histone acetylation at distinct promoters and resulting in enhanced transcription of cyclin-dependent kinase inhibitor p21 and transcriptional regulator c-fos . While this is likely to contribute to the growth arrest previously associated with SK2, the notion that SK2 can act as an epigenetic and transcriptional regulator suggests that there may be more downstream effects of its nuclear activity yet to be elucidated.
While a number of the initial studies describing pro-apoptotic effects of SK2 are based on data from forced overexpression of this protein, several studies also suggest this role for endogenous SK2. For example, small interfering RNA (siRNA) mediated knockdown of endogenous SK2 in HEK293 or mouse embryonic fibroblasts prevented the induction of apoptosis by TNF-α [24, 86], while mesangial cells taken from SK2−/− mice displayed greater resistance to staurosporine-induced apoptosis than wild-type or SK1−/− cells . Thus, these studies support the notion that SK2 can have an opposite role to SK1 in the control of cell survival. Indeed, the tissue distribution and developmental expression of the SKs would also suggest that they may have differing roles.
Interestingly, although there is compelling evidence to suggest that SK2 can have a physiological role in mediating apoptosis, there are now many studies also supporting a role for SK2 in promoting survival and proliferation, much like SK1. siRNA-mediated knockdown of SK2 has been shown to enhance apoptosis and decrease chemotherapeutic resistance in a number of cancer cell types [63, 90-92]. Interestingly, a recent study also proposed that nuclear S1P, produced by SK2, can act as an antagonist to the retinoic acid receptor β, attenuating ligand-stimulated tumour suppressor effects of this nuclear receptor in human colon carcinoma cells . This is an intriguing concept, as nuclear-localized SK2 and S1P have been shown in a number of other studies to have exclusively anti-proliferative roles [20, 86, 87].
Notably, there appears to be at least some functional redundancy between the two enzymes as deletion of either Sphk1 or Sphk2 in mice produces no gross phenotype [94-96], and yet a double knockout of the two genes is embryonic lethal from defects in neurological and vascular development .
SK2 in cancer
Despite the apparent roles, in some situations, of SK2 in inducing apoptosis and growth arrest described above, a number of studies have emerged that demonstrate a role for SK2 in cancer promotion, similar to SK1. Notably, targeting SK2 in a range of cancer cell lines can have more of an anti-cancer effect than targeting SK1 [63, 97]. A number of in vivo studies have reported that tumour growth can be significantly attenuated following the genetic ablation of SK2 in MCF-7 breast tumour xenografts  or the pharmacological inhibition of SK2 in a range of mouse tumour models, including breast [48, 99, 100], kidney , pancreatic , liver  and colon cancer . Furthermore, SK2 was shown to play a role in transforming growth factor (TGF) β induced migration of oesophageal cancer cells  and epidermal growth factor (EGF) induced migration of breast cancer cells , suggesting a potential role for SK2 in metastasis.
SK2 in inflammation and immune cell regulation
The role of SK in inflammation and immune cell function has been widely investigated, with most studies focusing on SK1 and its role in promoting inflammation. SK1 can be both post-translationally activated and transcriptionally upregulated by a number of inflammatory signalling molecules such as TNF-α, interleukin (IL)-1β, interferon (IFN)-γ, IgE and C5a [106-108], and is shown to regulate monocyte, macrophage and neutrophil function during the inflammatory response . However, despite SK1 having a clear role in promoting/enhancing inflammation, the role of SK2 in the inflammatory response is controversial, with many of the studies suggesting that SK2 may in fact be anti-inflammatory. It was demonstrated, using a breast cancer xenograft model, that SK2-deficient MCF-7 cells had increased levels of pro-inflammatory cytokines and decreased levels of anti-inflammatory IL-10, which coincided with a decrease in tumour growth . Furthermore, unlike SK1, siRNA-mediated knockdown of SK2 in a murine collagen-induced arthritis model led to more aggressive disease and production of pro-inflammatory cytokines . Notably, Samy et al. employed an adoptive transfer model of inflammatory bowel disease (IBD) where T-cell-deficient C.B-17 scid mice were injected with SK2−/− T cells, and found that mice receiving these cells had increased levels of pro-inflammatory cytokines and worsened intestinal inflammation than those receiving wild-type T cells, seemingly due to enhanced IL-2 responsiveness and increased expression of activated pSTAT5 . Therefore, SK2 may play a role in negatively regulating IL-2 signalling by attenuating STAT5 activation.
In contrast to the studies described above, the pharmacological inhibition of SK2 demonstrated anti-inflammatory effects in murine IBD models of ulcerative colitis  and Crohn's disease  and in rodent models of inflammatory arthritis , suggesting that SK2 can promote inflammation. However, it is intriguing that opposite effects are observed only with the pharmacological inhibition of SK2 compared with genetic ablation and RNA interference, suggesting either that the loss of SK2 protein elicits a different effect in comparison with its inhibition or that the SK2-specific inhibitor may be affecting different pathways.
It also appears that SK2 may have opposite roles to SK1 in mast cell function. Upon IgE-mediated crosslinking of the FcεRI receptor, the Src family kinases Lyn and Fyn facilitate the activation and translocation of the SKs to the plasma membrane . In murine mast cells, SK2 has emerged as the major producer of intracellular S1P and, unlike SK1, appears to mediate calcium influx and initiate downstream activation of PKCα, PKCβ and nuclear factor κB (NF-κB), leading to degranulation and the production of eicosanoids and cytokines . However, in human mast cells SK1 appears to be more important, initiating degranulation, migration and cytokine production, with the roles of SK2 seemingly limited to the production of TNF-α and IL-6 . This variation in SK function between species is intriguing, in particular as S1P strongly induces degranulation of human mast cells but only weakly in murine mast cells [115, 116]. It has therefore been suggested that SK2 may function as an intrinsic regulator of mast cell responses, independent of the S1P receptors, whereas SK1, which is largely responsible for the production of circulating S1P, may regulate extrinsic mast cell responsiveness [115, 116]. Indeed, sphingosine is reported to inhibit calcium influx [117, 118] and so SK2 may function, at least in murine mast cells, by decreasing the intracellular levels of sphingosine, thus allowing for calcium entry and the subsequent activation of downstream signalling pathways.
S1P plays an important role in immune cell function by regulating lymphocyte egress from lymphoid tissues, and it appears that SK2 may mediate this response by regulating the transport and circulation of S1P between tissues. A study by Sensken et al. demonstrated that the uptake of blood-borne S1P into peripheral tissues is SK2-dependent, and suggested that intracellular SK2 may play a role in importing S1P into cells and directing it to S1P lyase for degradation and maintenance of an S1P gradient . In agreement, SK2−/− mice resist lymphopenia induced in wild-type mice by the inhibition of S1P lyase in lymphoid tissues . This is an intriguing concept, particularly as a number of groups have reported that SK2−/− mice, surprisingly, have significantly increased levels of plasma S1P compared with wild-type mice [12, 115, 119-121]. Another obvious explanation for this phenomenon is that SK1 may be upregulated as a compensatory mechanism for SK2 ablation, and indeed this has been examined with somewhat conflicting results. Some studies have found no differences in SK1 mRNA levels or activity in SK2−/− mice compared with wild-type mice [12, 120]. Conversely, a recent study found increased SK1 mRNA and protein levels in SK2−/− mice , which was proposed to arise from a compensatory mechanism involving a reduction of HDAC1/2 inhibition by SK2, leading to a decrease in histone acetylation and an indirect transcriptional upregulation of SK1 [20, 121]. Notably, however, transgenic mice with ubiquitous overexpression of SK1 (approximately 20-fold over endogenous) did not show a significant increase in blood S1P levels .
Interestingly, murine SK2 was found to interact with the cytoplasmic region of the murine IL-12 receptor β1, mediating downstream IL-12 signalling and production of IFN-γ . It still remains to be elucidated whether this interaction and function of SK2 also occurs in humans.
SK2 in other diseases
SK2 also appears to be influential in a number of other disease states. A recent study demonstrated that the inhibition or downregulation of SK2 resulted in decreased proteolytic activity of BACE1, the rate-limiting enzyme for amyloid-β peptide production . Targeting SK1 also had the same effect, although the role of SK2 in this process appeared more prominent. Interestingly, SK2 activity was also found to be upregulated in the brain of Alzheimer's disease patients, suggesting a role for SK2 in this disease .
There are conflicting reports on the role of SK2 in ischaemia-reperfusion (IR) injury, although generally it appears to play a protective role. The genetic deletion of SK2, but not SK1, in mice significantly increased kidney damage following renal IR , and SK2 was found to mediate the protective effects of ischaemic preconditioning in cerebral [126, 127] and myocardial [128, 129] IR injury. Interestingly, cerebral ischaemia was found to increase SK2 mRNA levels in the brain  and hypoxic preconditioning rapidly and transiently upregulated SK2 activity and protein levels in cerebral microvessels . Conversely, however, SK2 inhibition greatly reduced liver injury and improved survival following hepatic IR, coinciding with a reduction in liver S1P levels and mitochondrial permeabilization , suggesting that the role of SK2 in IR injury may be tissue-specific.
It has also been reported that the pharmacological inhibition of SK2 resulted in reduced disease severity in rodent models of osteoarthritis  and diabetic retinopathy . It therefore appears that the roles of SK2 are complex and, as such, the dysregulation of SK2 can facilitate the development of a number of highly varied disease states. In this regard, SK2 is emerging as a promising therapeutic target in many of these diseases.
Regulation of SK2
Like SK1, the catalytic activity of SK2 can be rapidly increased upon stimulation by a number of agonists, including EGF , TNF-α , IL-1β , crosslinking of the IgE receptor FcεRI , and phorbol esters . Hypoxia was also found to rapidly activate SK2 in vivo  and in cultured lung cancer cells . Like SK1, SK2 activation can occur via phosphorylation by ERK1/2 . In cells, agonist-induced activation of SK2 increases its catalytic activity by 2- to 6-fold , which is comparable with that observed for SK1 . Interestingly, however, in vitro phosphorylation of SK2 by ERK1 resulted in a modest doubling of catalytic activity of this enzyme , which is less than the 14-fold increase observed for the same analysis of SK1 . Enhanced SK2 activity following interaction with eukaryotic elongation factor 1A has also been described, suggesting that alternative mechanisms of SK2 activation may occur in addition to that involving phosphorylation .
The subcellular localization of the SKs, and hence the compartmentalization of generated S1P, is widely accepted to play an important role in dictating their function [136, 137]. It is well established that in order for SK1 to mediate pro-survival, pro-proliferative and oncogenic signalling it must relocalize from the cytoplasm to the plasma membrane . By comparison, the subcellular localization of SK2 appears to be much more complex than for SK1, in line with its more complex functions. SK2 possesses nuclear localization (NLS) and export (NES) signals [45, 87], with the NLS positioned within the N-terminus and the NES located within the central proline-rich region that is not conserved in SK1 (Fig. 2). Interestingly, when localized to the nucleus, SK2 has been shown to inhibit DNA synthesis , as well as regulate epigenetic modifications via interaction with and modulation of HDAC1/2 . SK2 localization also appears to vary according to both cell type and cell density. For example, SK2 predominantly localizes to the nucleus in HeLa cells whereas in HEK293 cells it is mainly cytoplasmic . Moreover, it has been described that, as COS-7 fibroblasts became more confluent in culture, the proportion of SK2 localized to the nucleus increased , suggesting that SK2 may be involved in a contact inhibition response to decrease cell proliferation.
It has also been demonstrated that under stress conditions SK2 localizes to the endoplasmic reticulum (ER) . S1P production at the ER can fuel the sphingosine salvage pathway driven by ER-localized S1P phosphatases and ceramide synthases, to ultimately generate pro-apoptotic ceramide . Interestingly, artificially targeting SK1 to the ER or nucleus can allow the otherwise pro-survival enzyme to promote apoptosis [49, 87]. Moreover, mitochondrial localization of SK2 has been shown to promote apoptosis via S1P and BAK-dependent membrane permeabilization and cytochrome c release [22, 24]. Therefore the additional pro-apoptotic functions of SK2 not shared with SK1 seem to be mediated by its localization to intracellular membranes like the ER, mitochondria and nucleus. It should be noted that a lipid binding domain has been identified within the N-terminal region (residues 1–175) of SK2, not conserved in SK1 . This domain was shown to be a requirement for SK2 to interact with phosphoinositides and may therefore facilitate its differential localization at internal membranes .
The subcellular localization of SK2 also influences substrate availability, and recent findings suggest that this may be another major contributor to the pro-apoptotic function of SK2. Both SK1 and SK2 can utilize dihydrosphingosine (dhSph) as a substrate . Since dhSph is generated at internal membranes such as the ER, SK1 targeted to the plasma membrane cannot utilize this substrate pool . Unlike S1P, the specific role of dihydro-S1P (dhS1P) in oncogenic signalling is not well characterized. Interestingly, however, Hait et al. demonstrated that dhS1P and S1P generated by SK2 in the nucleus can both inhibit HDAC1/2 and regulate epigenetic modifications to inhibit proliferation . Furthermore, a very recent study employing photodynamic therapy and nanoparticle technology to treat solid tumours has reported that the anti-tumour efficacy of this novel technology is directly attributed to the production of dhS1P specifically by SK2, leading to a reduction in immune suppressive myeloid cells expanded in response to tumour-associated inflammation . Notably, this study also demonstrated that administering dhS1P in tumour-bearing mice prevented tumour growth and increased survival, as opposed to S1P which increased tumour growth . Therefore, dhS1P and S1P appear to have opposing roles in oncogenic signalling, dependent on their differential cellular compartmentalization. Again, this highlights the importance of SK2 subcellular localization and consequent access to substrates in relation to its function.
While most studies have described SK2 localization to internal organelles to either induce apoptosis or elicit growth arrest, one study has shown SK2 to mediate pro-proliferative S1P receptor signalling when localized to the plasma membrane . This process, however, appears linked to the activation of caspase-1 in apoptotic cells, which cleaves the N-terminus of SK2 and allows it to be exported from the cell, potentially mediated by the flipping of phosphatidylserine following interaction of the enzyme with the plasma membrane .
Although the activating phosphorylation site in SK1 (Ser225) occurs within a region that is divergent in SK2, studies have shown that SK2 catalytic activity also increases following phosphorylation by ERK1/2 . This activation of SK2 has been suggested to involve the phosphorylation of Ser351 and/or Thr578 on SK2a (Ser387 and Thr614 on SK2b) , although notably Ser351 is not conserved in mouse or rat SK2. Phosphorylation of SK2 also appears to regulate its nuclear–cytoplasmic shuttling, with protein kinase D mediated phosphorylation of either Ser419 or Ser421 within the NES of SK2 promoting its nuclear export .
Recent global phosphoproteome analyses have identified 13 other Ser/Thr phosphorylation events in endogenous human SK2, most of which occur within unique regions of SK2 not conserved in SK1 (Fig. 2) [40-44]. Nine of these residues are conserved in mouse and rat, while another three phosphorylations have been detected in mouse and rat SK2 (see Fig. 2) but have yet to be confirmed in human . No functional studies have been performed to define the regulatory significance of these novel phosphorylation sites. However, it is possible that these modifications play a role in the isoform-specific regulation of SK2 and may therefore provide an elegant yet undoubtedly complex mechanism to allow for the emerging functional complexity of SK2.
Two isoforms of SK2 have been described and characterized [38, 86]. Unless otherwise specified, in the literature ‘SK2’ generally refers to the shorter isoform (SK2a or SK2-S), which is consequently the best characterized. The larger isoform (SK2b or SK2-L) has an additional 36 amino acids at the N-terminus and appears to arise from the use of an alternative start codon (Fig. 3). While not expressed in mice, SK2b appears the predominant form of SK2 in several human cell lines and tissues . This may suggest that in fact SK2b is physiologically the more important human SK2 isoform, although very few studies have specifically examined its functions. It has been demonstrated that serum deprivation leads to an increase in SK2b expression and promotes the translocation of SK2b to the nucleus where it can inhibit DNA synthesis . Furthermore, SK2b appears to phosphorylate some of its substrates at a higher rate than SK2a, including FTY720 (4-fold increase) and sphingosine (1.3-fold increase) . This trend would suggest that the N-terminal extension of SK2b may introduce a conformational change that promotes catalytic activity .
A third isoform of SK2 (SK2c; Genbank™ accession number EF107108) has been reported, which possesses an N-terminal extension and an additional C-terminal sequence  (Fig. 3). SK2c mRNA has been reported to be detected in some human cells, but no other analysis of this putative SK2 isoform has been described . In silico analysis has predicted a fourth SK2 splice variant (Genbank™ accession number AK000599) that differs from SK2a at the C-terminus (from residue 256; see Fig. 3), although there is yet no physical evidence for the existence of this SK2 isoform. Notably, this putative SK2 variant does not possess the C4 and C5 regions proposed to contain the sphingosine binding and catalytic residues, respectively [4, 145], which have previously been shown to be essential for SK activity . Therefore, even if this SK2 variant was expressed it is unlikely to have SK activity.
SK2 substrate specificity
Both SK1 and SK2 can utilize d-erythro-sphingosine and d-erythro-dihydrosphingosine as substrates; however, SK2 appears to be more promiscuous and can phosphorylate a range of other biological and synthetic lipids with much greater efficiency than SK1 . These substrates include phytosphingosine , ω-biotinyl d-erythro-sphingosine  and, surprisingly, the SK1-specific inhibitor d,l-threo-dihydrosphingosine . SK2 is also responsible for phosphorylating the immunosuppressive agent FTY720, converting it to its active form FTY720-P [144, 147, 148]. Although SK1 can also phosphorylate FTY720 in vitro, albeit less efficiently than SK2 [144, 147, 148], levels of FTY720-P in SK2−/− mice administered FTY720 are negligible, indicating that SK1 does not significantly contribute to FTY720 phosphorylation in vivo . The ability of SK2 to phosphorylate a larger pool of substrates suggests that the sphingosine binding pockets of SK1 and SK2 differ slightly in conformation. Although the functional significance of this is currently unknown, it has allowed for these subtle structural differences to be exploited in the generation of isoform-specific SK inhibitors.
A number of SK inhibitors have been generated which show potential for development as therapeutics for cancer and some other diseases . The majority of these specifically inhibit either SK1 or both SK isoforms. In the last few years, however, several SK2-specific inhibitors have emerged that demonstrate promising therapeutic properties (Table 1). Notably, recently developed potent SK1-selective inhibitors have demonstrated surprisingly limited anti-cancer properties compared with less isoform-specific SK inhibitors [47, 150, 151]. This notion, coupled with the promising anti-cancer effects of the current SK2-specific inhibitors (described below), suggests that in some cancers SK2 may be a more important target for cancer therapy.
Table 1. Structures and inhibitory properties of SK2 inhibitors. ND, not determined
There is growing interest in the therapeutic effects of the first described SK2-selective small molecule inhibitor ABC294640 . This orally bioavailable agent specifically targets SK2 in a sphingosine-competitive manner with a Ki of 9.8 μm. No effect was observed with ABC294640 on SK1, or the closely related diacylglycerol kinase, at concentrations up to 100 μm [47, 48]. There have been a number of studies documenting the effects of this inhibitor in a range of disease models. Notably, ABC294640 has been shown to significantly decrease tumour growth in vivo in an array of different tumour models in mice [48, 99-103]. Furthermore, ABC294640 also seems to have therapeutic potential for a number of other diseases, where attenuated disease progression was observed in rodent models of osteoarthritis , rheumatoid arthritis , Crohn's disease , ulcerative colitis  and diabetic retinopathy . Biochemically, ABC294640 has been reported to decrease S1P and increase ceramide levels in cells [47, 48], decrease plasma S1P levels in mice , inhibit TNF-α-induced NF-κB activation , as well as inhibit the activation of STAT3, AKT and ERK2, and decrease the expression of STAT3 and ERK2 . There are conflicting reports on the mode of action whereby ABC294640 induces cell death, with some studies demonstrating that apoptotic pathways are activated [48, 100] and others describing the presence of autophagy markers [47, 152]. Perhaps the mechanism by which SK2 inhibition leads to cell death is tissue- and disease-specific, as different diseases may exploit different roles of SK2, and so inhibition of SK2 in each case may affect different pathways. Nevertheless, targeting SK2 with ABC294640 appears to have significant therapeutic potential and, as such, this compound is currently in phase I clinical trials for the treatment of advanced solid tumours.
Notably, in addition to its SK2 inhibitory properties, ABC294640 can also bind to the oestrogen receptor and act as a partial antagonist like tamoxifen . While this additional action of ABC294640 complicates some studies into the role of SK2 in disease, it does suggest that, with dual targeting of SK2 and the oestrogen receptor, this agent may be particularly beneficial in the treatment of oestrogen-receptor-positive breast cancer.
SLR080811 has been recently described as an SK2-specific inhibitor . This agent is a competitive inhibitor with respect to sphingosine, with a Ki of 1.3 μm for SK2 and a 10-fold weaker affinity for SK1 (Ki = 12 μm). Interestingly, SLR080811 was not found to have significant effects on survival or proliferation of U937 human leukaemia cells, despite reducing S1P generation and total S1P levels in these cells . The reasons for this unexpected finding remain unclear, with further studies required to examine whether this is specific to this cell line or a more general effect.
Notably, use of this inhibitor in vivo resulted in an increase in blood S1P levels in wild-type mice , which is the same unexpected phenomenon observed in SK2−/− mice [12, 115, 119, 120] but different from that observed with ABC294640 . Again, the reasons for these differences require further analysis.
Amidine-based compounds have recently been described as SK inhibitors . To this end, a quaternary ammonium salt, labelled trans-12a, was synthesized and found to selectively inhibit SK2 over SK1, with Ki values of 8 and 60 μm, respectively . In U937 human leukaemia cells, trans-12a was reported to significantly reduce the production of FTY720-P, suggesting inhibition of SK2, although there was no detectable change in S1P levels in these cells .
K145 is another SK2-selective inhibitor that acts in a competitive manner with respect to sphingosine, with a Ki of 6.4 μm . No significant inhibition of SK1 or ceramide kinase was observed by K145 at concentrations up to 10 μm . At 10 μm, K145 also caused a > 40% decrease in the activity of CaMKIIβ, and at 4 μm a significant decrease in phospho-ERK and phospho-AKT was observed, implying that K145 may function as a dual-pathway inhibitor . Cellular studies demonstrated that, in contrast to SLR080811, K145 can inhibit growth and induce apoptosis in U937 human leukaemia cells. In vivo K145 was found to reduce tumour volume in JC mammary adenocarcinoma and U937 leukaemia xenograft models , with comparable anti-tumour activity by both oral administration (50 mg·kg−1) and intraperitoneal injection (15 mg·kg−1).
Lim et al. have recently synthesized an SK2-selective inhibitor, (R)-FTY720 methyl ether, based on the structure of the SK2-specific substrate FTY720 . (R)-FTY720-OMe is a competitive inhibitor with respect to sphingosine, inhibiting SK2 with a Ki of 16.5 μm and showing no significant inhibition of SK1 activity at 50 μm . In LNCaP prostate cancer cells, (R)-FTY720-OMe was found to decrease expression of SK2 and stimulate autophagy, but not apoptosis, in these cells .
SG-12 was synthesized amongst a panel of 15 potential SK2-specific inhibitors and was found to inhibit SK2 with an IC50 of 22 μm while not affecting SK1 activity . SG-12 induced cell death in CHO-K1 cells, consistent with SK inhibition, but since it also inhibits PKC the actual mechanism mediating these effects is not clear . Interestingly, it has now been proposed that SG-12 is phosphorylated by SK2 and this modification is critical for its ability to induce cell death .
SKI-II is one of the most commonly used SK inhibitors . Among numerous findings, SKI-II has been shown to decrease S1P production and induce apoptosis in tumour cells in vitro [59, 76] and inhibit tumour growth in a mammary adenocarcinoma xenograft mouse model . SKI-II has been used widely as a direct SK1 inhibitor [133, 160-162], although more recent studies suggest it acts mainly to target SK1 through enhancing degradation of this enzyme [163, 164]. While commonly stated as an SK1-specific inhibitor, analyses of this compound revealed that SKI-II inhibits SK2 with slightly higher affinity than SK1 (Ki values of 7.9 and 16 μm for SK2 and SK1, respectively) , but, contrary to SK1, does not induce degradation of SK2 .
Concluding remarks and future perspectives
There is now considerable evidence that SK2 can physiologically perform both pro- and anti-survival functions in cells. It is not currently understood how this is regulated, but it is almost certainly influenced by the subcellular localization of SK2, which may vary according to the cell and tissue type, disease state or environmental stimuli. It is notable, however, that the role for SK2 specifically in cancer and disease, as determined by pharmacological or genetic ablation in vivo, is overwhelmingly in support of survival and proliferation. Therefore, despite its ability to promote apoptosis under some conditions, it still appears that targeting SK2 in cancer and other diseases will provide substantial therapeutic benefits. However, as the functional complexity of SK2 becomes more apparent (Fig. 4), it stands to reason that the mechanisms regulating its contrasting roles need to be identified in order to develop therapeutics to specifically target the pro-survival functions of this enzyme.
This work was supported by the Fay Fuller Foundation, an Australian Postgraduate Award and Dawes Scholarship (to HAN), and a Senior Research Fellowship and Project Grant (626936) from the National Health and Medical Research Council of Australia (to SMP).