The role of sphingolipids in meibomian gland dysfunction and ocular surface inflammation

Abstract

Inflammation occurs in response to tissue injury and invasion of microorganisms and is carried out by the innate and adaptive immune systems, which are regulated by numerous chemokines, cytokines, and lipid mediators. There are four major families of bioactive lipid mediators that play an integral role in inflammation – eicosanoids, sphingolipids (SPL), specialized pro-resolving mediators (SPM), and endocannabinoids. SPL have been historically recognized as important structural components of cellular membranes; their roles as bioactive lipids and inflammatory mediators are recent additions. Major SPL metabolites, including sphingomyelin, ceramide, ceramide 1-phosphate (C1P), sphingosine, sphingosine 1-phosphate (S1P), and their respective enzymes have been studied extensively, primarily in cell-culture and animal models, for their roles in cellular signaling and regulating inflammation and apoptosis. Less focus has been given to the involvement of SPL in eye diseases. As such, the aim of this review was to examine relationships between the SPL family and ocular surface diseases, focusing on their role in disease pathophysiology and discussing the potential of therapeutics that disrupt SPL pathways.

Introduction

Inflammation is a pathophysiological process that occurs in response to various insults, including tissue injury and invasion of microorganisms. The innate immune system is activated in response to foreign molecules displaying pathogen-associated molecular patterns (PAMPs) and stressed or injured cells producing damage-associated molecular patterns (DAMPs), which prompts recruitment of granulocytes to clear pathogens, production of pro-inflammatory cytokines including tumor necrosis factor-alpha (TNF-α), interleukin (IL) 1β, and IL-6, and lipid mediators.¹ While there are many different classes of molecules responsible for driving the inflammatory response, bioactive lipids are a class of lipid mediators that are involved not only in the initiation and potentiation of inflammation but also in the ultimate cessation or resolution of the inflammatory response.² Bioactive lipids can be divided into four major families – eicosanoids, sphingolipids (SPL), specialized pro-resolving mediators (SPM), and endocannabinoids - each of which plays a unique role in inflammation.^2-5

As describing the clinical implications of each family of bioactive lipids is beyond the scope of a single article, this review will focus on SPL, lipid mediators implicated in a number of inflammatory conditions involving the vascular, neurologic, and gastrointestinal systems.^6-18 Less emphasis has been given to the role of SPL with respect to eye disorders, and as such, in this review, we focus on the role of SPL as it relates to ocular surface diseases.

Sphingolipid Biology

SPL are a group of lipids that have baffled scientists for decades. Lipids in the brain were first described by a neurochemist in the 1880s, and were named “sphingosine” owing to their sphinx-like (enigmatic) chemical nature.¹⁹ It wasn’t until the early 2000s that SPL biology was better understood. First, as many of the regulatory enzymes in SPL metabolism were cloned, this allowed the development of knockout mice for the creation of animal models for further studies on SPL biology.^{16, 20} Second, advances in mass spectrometry technology allowed for simultaneous analysis and quantification of the numerous SPL metabolites.^{16, 20} Finally, specific antagonists and agonists of receptors involved in SPL metabolism were developed, further paving the way for additional cell culture, animal, and eventually human studies.^{16, 20} SPL are ubiquitous membrane lipids that play important structural as well as cell signaling roles.²¹ As a family, they mediate a wide range of processes and are implicated in many chronic inflammatory conditions.² SPL with the most extensively-studied bioactivity are ceramide (Cer), sphingosine 1-phosphate (S1P), ceramide 1-phosphate (C1P), and their precursor molecule sphingomyelin (SM). As a group, they play a part in regulating apoptosis, cell growth, inflammation, angiogenesis, intracellular trafficking, and other processes.^{2, 22-25} In general, bioactive SPL have been found to have pro-inflammatory effects, with both pro- and anti-apoptotic roles reported. However, the complex and often contradictory effects of SPL signaling pathways preclude them from being described as having strictly one effect. In fact, the interconvertability of these molecules and observations that their biological effects often oppose each other has given rise to the concept of a “sphingolipid rheostat”, which attempts to describe how the balance of these molecules’ activity influences the prevailing tone of inflammatory, apoptotic, and survival signals which ultimately determines a cell’s fate (Figure 1).^{26, 27}

Figure 1:

A simplified representation of the “sphingolipid rheostat.” Ceramide, generally a pro-apoptotic metabolite, and sphingosine 1-phosphosphate (S1P), a generally anti-apoptotic metabolite, are readily interconverted through the actions of ceramidase (CDase), sphingosine kinase (SphK), S1P phosphatase (S1PP), and ceramide synthetase (CerS).

Sphingomyelin

SM is the most abundant SPL in mammalian cells.^{28, 29} De novo synthesis of SM occurs in the endoplasmic reticulum through various reactions to combine serine and palmitoyl-CoA.³⁰ SM is also made from Cer in the Golgi apparatus (sphingomyelin synthase 1) and the plasma membrane (sphingomyelin synthase 2).³⁰ Ultimately, the vast majority of SM resides in the plasma membrane as a key component of membrane lipid rafts, with nearly 70% of all cellular SM residing in these membrane lipid rafts.³¹ Lipid rafts play roles in membrane trafficking, protein sorting, and contain key signal transduction molecules, including G-protein coupled receptors, protein kinase C receptors, and adenylyl cyclase.^{32, 33} While SM itself is not thought to be bioactive, through its role as a key component of membrane lipid rafts and the precursor molecule for Cer, it certainly is indirectly involved in maintaining cell signaling functions.³⁰

Ceramide

Cer is the keystone of SPL metabolism and is implicated in inflammation both directly and indirectly (via its downstream metabolites C1P and S1P). Cer may be generated de novo in the endoplasmic reticulum and in the Golgi apparatus from serine and palmitoyl-CoA, formed via endolysosomal hydrolysis and salvage of complex SPL, or synthesized in membranes via acidic and neutral sphingomyelinases (aSMase and nSMase, respectively) (Figure 2).^{34, 35} Events that occur early during immune system activation can lead to Cer generation. For example, cytokines (e.g., TNFα, IL-1β) and activated tumor necrosis factor receptors (TNFR) can induce SMases to convert membrane-bound SM to Cer, which has been shown in cell culture and mouse models to have various pro-inflammatory and pro-apoptotic effects.^36-38 As the keystone intermediate of SPL metabolism, Cer may serve as a precursor for S1P or C1P, either of which may augment or oppose the effects of Cer, as described in more detail below.

Figure 2:

Sphingolipid synthesis and influence on inflammatory pathways. Activation of TNFR1 induces SMase and SphK, which form Cer and S1P, respectively. Cer activates NF-;B and cPLA₂, but may also be converted to C1P, which both activates cPLA₂ and inhibits TACE, thus potentially exerting both pro- and anti-inflammatory effects. TNFR1 activation also induces SphK, which converts Sph to S1P. S1P may pass outside of cellular membranes via Spns2 channels and acts on various S1P receptors (S1PR1-5). Downstream effects of S1PR signaling include modulation of cellular activity via ERK1/2, AKT, PLC, Rho, and JNK, as well as activation of NF-κB, cPLA₂, and COX. Activation of these pathways by Cer, C1P, and S1P may help induce or maintain a pro-inflammatory state by facilitating the production of prostaglandins, thromboxanes, and leukotrienes. Ceramide (Cer); Ceramide 1-phosphate (C1P); Ceramide synthase (CerS); Sphingomyelinase (SMase); Sphingomyelin (SM); Sphingosine (Sph); Sphingosine kinase (SphK); Sphingosine 1-phosphate (S1P); Sphingosine 1-phosphate receptor (S1PR); Tumor necrosis factor receptor 1 (TNFR1); Tumor necrosis factor-alpha converting enzyme (TACE); Cytosolic phospholipase A2 (cPLA₂); Cyclooxygenase (COX); Lipoxygenase (LOX); S1P transporter spinster homolog 2 (Spns2).

Sphingosine 1-phosphate

S1P is generated when sphingosine, derived from Cer, is phosphorylated by the two isoforms of sphingosine kinases, SphK1 and SphK2 (Figure 2). This process occurs intracellularly in the cytosol (SphK1) and in the nucleus or occasionally mitochondria (SphK2).^{39, 40} In most cases, S1P is exported from cells via specific transporters and acts extracellularly through five specific G protein-coupled receptors (S1PR1 – S1PR5). However, additional studies have also implicated an intracellular role for S1P as a second messenger in cell signaling pathways.^{41, 42} Finally, intracellular levels of S1P are closely regulated through the degradation pathways of complex SPL and Cer. S1P can be dephosphorylated back to sphingosine, which can then be used for Cer generation in the salvage pathway, or S1P lyase can irreversibly cleave S1P to generate a hexadecenal and an ethanolamine phosphate in the final step of SPL metabolism that releases these compounds from SPL metabolic cycle.^{42, 43}

S1P has been noted to have multiple roles, including regulation of inflammation through lymphocyte trafficking, both pro- and anti-apoptotic functions, and regulation of vascular integrity and angiogenesis.⁴⁴

S1P is usually found in high concentrations in the blood and lymph, primarily in erythrocytes and endothelial cells.⁴⁴ Erythrocytes lack S1P degrading enzymes, leading to high S1P concentrations, and erythrocytes and endothelial cells both express specific S1P transporters, also contributing to the high circulating S1P concentration.^{45, 46} Conversely, S1P concentrations are low in intracellular and interstitial compartments creating an S1P gradient that is crucial for immune cell trafficking.⁴⁴ Briefly, S1PR1 is highly sensitive to S1P, and the receptor is internalized rapidly in circulating lymphocytes in the high-S1P environment in plasma. The internalized S1PR1 is crucial for circulating lymphocytes to enter secondary lymphoid organs against the S1P gradient. In the low-S1P environment of secondary lymphoid organs, the receptor is gradually re-expressed on the cell surface, allowing the lymphocytes to egress secondary lymphoid organs with the S1P gradient.⁴⁴ In inflammatory responses, naïve lymphocytes are presented antigens in secondary lymphoid organs, and then these activated lymphocytes must exit the lymphoid organs to carry out their immune function.⁴⁴

Multiple studies have investigated the role of S1P in trafficking lymphocytes in and out of lymphoid organs.⁴⁴ In one study, knockout mice with hematopoietic cells deficient in S1PR1 had a decreased concentration of circulating lymphocytes, but high concentrations of lymphocytes within the thymus.⁴⁷ Furthermore, when these S1PR1-deficient lymphocytes were transferred into wildtype mice, the lymphocytes were able to enter but not exit secondary lymphoid organs.⁴⁷ Together, this suggests that S1P-S1PR1 signaling is required for egress of lymphocytes from lymphoid organs but is not a requirement for entry into the lymphoid organs.

Regarding the roles of S1P in apoptosis, in a cell culture model of human embryonic kidney cells (HEK 293) and human fibroblasts, overexpression of the enzyme sphingosine kinase (SphK), which phosphorylates sphingosine into S1P, was shown to promote cell survival.⁴⁸ Further studies in various cells lines, including HEK 293 and human fibroblasts comparing the two isoforms of SphK found that upregulation of intranuclear SphK2 inhibited DNA synthesis and promoted apoptosis, whereas upregulation of cytosolic SphK1 promoted cell survival and growth, suggesting two S1P phenotypes, one pro-apoptotic and one anti-apoptotic.^{49, 50}

Finally, S1P is an important regulator of vascular integrity and normal angiogenesis. Vascular permeability is facilitated by the loosening of tight junctions between vascular endothelial cells in response to inflammatory mediators like leukotrienes and histamine, allowing plasma proteins, complement, and leukocytes to leak from blood vessels into the target tissues.⁴⁴ Mice selectively deficient for plasma S1P (through deletion of both SphK1 and SphK2) had increased vascular leakage compared to control mice in a model of anaphylaxis generated by histamine.⁵¹ In another study, knockout mice deficient in ApoM, a key plasma S1P transport protein, also had increased vascular leakage compared to control mice.⁵² Additionally, when knockout mice were given ApoM/S1P or an S1PR1 agonist (to simulate the effects of S1P on its receptor), levels of vascular leakage returned to levels similar to control mice.⁵² These studies suggest that decreased plasma S1P and S1PR1 activation increase vascular permeability. Furthermore, S1P signaling may serve as a checkpoint for vascular integrity to limit an exaggerated response to pro-inflammatory molecules.

Finally, S1P promotes vascular integrity and maturation of vessels. In a mouse model, S1PR1 deficient mice had increased embryonic hemorrhage compared to normal mice.⁵³ Notably, these mice had normal vasculogenesis but had incomplete maturation of these vessels.⁵³ When S1PR1 deficiency was limited to endothelial cells, these mice still had incomplete vascular maturation.⁵⁴ Finally, mice unable to produce S1P (induced by SphK1 and SphK2 deletion), also demonstrated incomplete maturation of vessels.⁵⁵ Together, these studies confirm an important role in the S1P-S1PR1 axis in the regulation of vascular maturation in an endothelial-cell-dependent manner.

In summary, S1P has multiple roles including regulation of inflammation through immune cell trafficking, both pro- and anti-apoptotic roles in a context and tissue-dependent fashion, and regulation of vascular integrity and regulation of normal angiogenesis, underscoring the complexity of S1P metabolism and signaling.

Ceramide 1-phosphate

C1P is generated when Cer is phosphorylated by membrane-bound ceramide kinase (CerK). This process occurs primarily in membrane compartments within cells, such as the endoplasmic reticulum or Golgi apparatus. C1P is transported to the plasma membrane and likely binds to one or more not yet identified G protein-coupled receptors to carry out its actions.^{56, 57} C1P synthesis has been found to be regulated by inflammatory stimuli (IL-1β and calcium ionophore A23187), which increase CerK activity and thus C1P conversion from Cer (Figure 2).^58-61

As with all SPL, C1P has complex actions, with both pro- and anti-inflammatory effects, anti-apoptotic effects, and roles in angiogenesis.^{25, 57, 62-66} Supporting its pro-inflammatory role, in an in vitro study of A549 human lung adenocarcinoma cells stimulated with IL-1β and calcium ionophore A23187 had increased activity of CerK and increased production of arachidonic acid (AA) and thromboxane and prostaglandin, which are both downstream pro-inflammatory eicosanoids, thus supporting the idea that C1P is an upstream activator of pro-inflammatory molecules.⁶⁰ Other studies have demonstrated anti-inflammatory effects. In a model of murine peritoneal macrophages, treatment with lipopolysaccharide (LPS), a potent pro-inflammatory activator of macrophages, generated high levels of the inflammatory cytokine TNFα.⁶⁷ Blockade of CerK, thus inhibiting C1P generation in this study, significantly increased LPS-induced TNFα production, thus suggesting an anti-inflammatory role for C1P in this context.⁶⁷

C1P has also been shown to have anti-apoptotic effects. In a cell culture model of bone-marrow-derived macrophages, deprivation of macrophage colony-stimulating factor (M-CSF) induces apoptosis. Deprivation of M-CSF resulted in increased SMase, increased Cer, and decreased intracellular C1P. In this experiment, addition of exogenous C1P promoted cell survival through direct inhibition of SMase which resulted in reduced Cer generation.⁶³ In a follow-up study using a model of alveolar macrophages that undergo apoptosis when incubated in the absence of serum, the group found that the alveolar macrophages under apoptotic conditions had higher levels of Cer, but only slight activation of SMases, indicating that in this model, Cer was primarily generated through the de novo pathway rather than through hydrolysis of SM. Exogenous C1P again blocked apoptosis; however, in this model, it was through blockade of serine palmitoyl transferase, a key enzyme in de novo Cer synthesis.⁵⁷ Together these two studies suggest that C1P has an anti-apoptotic effect, and can carry out this effect through blockade of Cer generation by both the SMase and de novo pathways.

Finally, there is some evidence to support a role of C1P in angiogenesis as well. In a murine model of hindlimb ischemia induced by ligation of the femoral artery, administration of intravenous exogenous C1P five hours after occulusion resulted in significantly increased hindlimb perfusion on day 14 compared to mice injected with saline vehicle only (p<0.05; flow ratios measured by Doppler ultrasound).⁶⁸ These data suggest a role for C1P in reperfusion of ischemic tissues, but further studies are needed to fully understand the mechanism of the observed effects. In another study, dermal microendothelial cells (DMECs) from wildtype or CerK deficient mice were studied in vitro.⁶² CerK-deficient DMECs and wildtype DMECs treated with a CerK-inhibitor had significantly impaired capillary-like tube formation compared to untreated wildtype DMECs. Addition of vascular endothelial growth factor, fibroblast growth factor, and tumor necrosis factor was unable to rescue angiogenesis in the CerK-deficient or wild type cells treated with the CerK-inhibitor.⁶² Finally, in both CerK-deficient cells and treated wild type cells, levels of Cer were increased with no change in S1P levels compared to untreated wildtype cells. Together, these data suggest that C1P has a role in regulating angiogenesis through a novel pathway that is independent of S1P signaling.

Overall, much like S1P, C1P’s roles in regulating inflammation, apoptosis, and angiogenesis are complex, context-dependent, and not yet fully understood.

Interactions between Sphingolipids and Eicosanoids

While the roles of eicosanoids, a class of bioactive lipids involved in inflammation, will not be covered in detail in this review, there is growing evidence of substantial interplay between SPL and critical components of the eicosanoid-driven inflammatory response, which necessitates mention here. Evidence supports an expanded view of the classic eicosanoid synthesis pathways that includes the involvement of Cer, C1P, and S1P (Figure 3). In cell cultures, the combined actions of Cer, S1P, and C1P suggested a synergistic mechanism of eicosanoid formation via upregulation of the rate-limiting enzyme in eicosanoid formation, cytosolic phospholipase A₂(cPLA₂).²⁵

Figure 3:

Inflammatory signaling roles of sphingolipids (SPL), arachidonic acid (AA), docosahexaenoic acid (DHA), and eicosapentaenoic acid (EPA). Synthesis of Ceramide occurs via three main pathways: de novo synthesis, hydrolysis of sphingomyelin (SM), and salvage from complex SPL and gangliosides (GG). Ceramide generated via lysosomal SM hydrolysis can activate NF-κB and CAAT/enhancer binding proteins (c/EBP), both of which are critical inflammatory gene transcription. Ceramide may also activate cytosolic phospholipase A2 (cPLA₂), though some studies have shown that this effect on cPLA2 is primarily due to downstream ceramide-1-phosphate (C1P) and sphingosine-1-phosphate (S1P). Ceramide may be phosphorylated to C1P, which activates cPLA₂ and increases AA formation. C1P has also been shown to inhibit TNF activation, suggesting both pro- and anti-inflammatory roles. S1P increases activity of NF-κB, cPLA₂, and cyclooxygenase-2 (COX-2), thus activating early inflammatory mediators as well as increasing AA production and synthesis of inflammatory products by COX. COX-1 and COX-2 are enzymes which synthesize prostaglandins (PGs), leukotrienes (LTs), thromboxanes, and prostacyclins from AA, which are critical for initiating and sustaining inflammatory responses. Meanwhile, DHA and EPA are converted to anti-inflammatory signaling products by lipoxygenases (LOX). DHA forms D-series resolvins and maresins, which inhibit pro-inflammatory CD4+ Th1 and Th17 cell activity and increase anti-inflammatory regulatory T-cell (Treg) activity, while EPA forms E-series resolvins that inhibit CD4+ Th2 cell activity. Both D and E series resolvins and maresins also have other anti-inflammatory effects such as inducing neutrophil apoptosis and inhibition of cytokine production etc., all of which are important for resolving inflammation. Sphingomyelin (SM); Ceramide 1-phosphate (C1P); Arachidonic acid (AA); Docosahexaenoic acid (DHA); Eicosapentaenoic acid (EPA); Cyclooxygenase (COX); Lipoxygenase (LOX); Prostaglandin (PG); Leukotriene (LT).

Cer itself is not thought to have direct actions on the eicosanoid pathway. However, as the precursor of both C1P and S1P, it is crucial in the relationship between SPL and eicosanoids.⁶⁹

C1P directly upregulates eicosanoid formation. As discussed above, human lung epithelial cells stimulated with IL-1β and calcium ionophore A23187 had increased CerK activity (yielding more C1P) with an associated increase in AA and thromboxane and prostaglandin, both downstream pro-inflammatory eicosanoids.⁶⁰ In a follow-up study in the same cell line, inhibition of cPLA₂ extinguished the ability of C1P to stimulate AA release.⁷⁰ Together, these studies demonstrate the role of C1P in upregulating AA production through direct activation of the rate-limiting enzyme cPLA₂.

S1P also has roles in directly activating eicosanoid formation. In human lung epithelial cells, S1P was found to increase AA production in a time and dose-dependent manner. In the same study, inhibition of cPLA₂ led to decreased AA production, suggesting that S1P also acts through direct activation of cPLA₂.⁷¹ In another cell culture study using human fibroblasts, stimulation with TNFα led to an increase in S1P and eicosanoids, prostaglandin E₂ (PGE₂) and cyclooxygenase (COX).⁷² Inhibition of S1P lyase, the enzyme that degrades S1P, also led to high levels of PGE₂ and COX, providing further support to the role of S1P in generation of these eicosanoids.⁷² Conversely, inhibition of SphK1 (preventing S1P formation) led to decreased S1P, PGE₂, and COX.⁷² Finally, exogenous Cer and Sph were unable to generate PGE₂ or COX in the presence of SphK1 inhibition, suggesting that the effects of Cer and Sph on eicosanoid generation are mediated by S1P.⁷² Together, these results further support the role of S1P in upregulating eicosanoid generation.

In summary, SPL have a complex relationship with other bioactive lipid mediators, such as eicosanoids. This relationship is a two-way street in which eicosanoid-driven production of pro-inflammatory cytokines (such as IL-1β) can increase SPL synthesis while SPL sub-types (i.e., Cer, C1P, and S1P) can increase eicosanoid synthesis.^{25, 60, 69-72}

Sphingolipids have been implicated in a number of ocular diseases

SPL have been implicated in the pathogenesis of several ocular diseases.³⁴ As several recent reviews have covered that subject well,^{34, 73-76} a concise description of the inflammatory eye diseases will be discussed here. Animal studies have suggested a role of Cer, SM, and S1P in inflammatory eye conditions, specifically in uveitis. In an endotoxin-induced uveitis model in rats, aqueous humor and retinas were harvested in the acute inflammatory period (24 hours after endotoxin induction), and lipids were extracted. All five classes of phospholipids were detected in both uveitis and control rats; however, in uveitis rats, there was a significant increase in the ratio of Cer: total phospholipids and SM: total phospholipids.⁷⁷ In another study, a mouse model of experimental autoimmune uveitis (EAU), EAU mice were treated with FTY720, a drug known to downregulate the expression of the S1PR1. S1P signaling through S1PR1 is a key step in T cell egress from secondary lymphoid tissue, and blockade of this receptor prevents migration of T cells to inflammatory sites.⁷⁸ In this study, treated EUA mice had a higher percentage of CD4+ T cells in lymph nodes and a lower percentage of CD4+ T cells in the eye when compared to untreated EUA mice, suggesting that FTY720 was successful in preventing egress of CD4+ T cells from lymphoid tissue, and establishing a role for S1P signaling in uveitis.⁷⁸ Taken together, these studies suggest that Cer, SM, and S1P are involved in promoting increased inflammatory cell infiltration in uveitis and that blockade of S1P-S1PR1 signaling may prevent inflammatory cell infiltration into uveal tissue.

SPL have also been implicated in age-related macular degeneration (AMD).^79-80 Inhibition of Cer synthesis using desipramine, an SMase inhibitor,⁸¹ and overexpression of acid ceramidase (N-acylsphingosine amidohydrolase 1, ASAH1), which catalyzes the conversion of Cer to sphingosine,⁸² have both been shown to protect photoreceptors from oxidative stress and apoptosis in vitro, while overproduction of Cer was shown to have the opposite effect on RPE cells.⁸³ In a mouse model of laser-induced choroidal neovascularization, mice injected with sonepcizumab (a monoclonal antibody that selectively binds and blocks S1P) had a significantly reduced area of choroidal neovascularization and decreased leakage on fluorescein angiography as compared to sham-treated animals, linking S1P to choroidal neovascularization.⁸⁴ In the same model, mice injected with sonepcizumab had a decreased influx of macrophages into the ischemic retina, as well as less retinal neovascularization.⁸⁴ Other studies have found similar effects with respect to SPL and retinal neovascularization. In a mouse model of ischemia-induced retinopathy (exposure to air with 75% oxygen for 5 days followed by room air), mice deficient in S1P receptor (S1PR2) had less vitreous chamber neovascularization and more normal revascularization into areas of ischemia compared to control mice.⁸⁵ Together, these two studies link S1P to pathological retinal angiogenesis.

SPL have also been implicated in glaucoma. In a metabolome-wide association study, plasma collected from individuals with primary open-angle glaucoma (POAG) had higher levels of sphingosine and lower S1P compared to controls, highlighting the potential involvement of SPL in glaucoma.⁸⁶ In a study pairs of donor normal human eyes, normal intraocular pressure was maintained with an infusion of buffered saline. In each pair, the anterior chamber of one eye was exchanged with a solution of S1P, while the contralateral eye was maintained with saline, and the outflow facility was monitored for three hours after the anterior chamber exchange. Eyes with saline containing S1P had decreased outflow facility through Schlemn’s canal as compared to the control contralateral eye,⁸⁷ suggesting an effect of S1P on aqueous outflow dynamics.

Sphingolipids and ocular surface diseases

Ocular surface diseases encompass a number of conditions often placed within the umbrella of “dry eye”. Dry eye (DE) is a term that includes both aqueous tear deficiency (ATD) and evaporative dry eye (EDE) subtypes. Both subtypes have been closely associated with meibomian gland dysfunction (MGD), which is defined by clinical signs including meibomian gland (MG) plugging, MG dropout, abnormal meibum quality, and hyperemia and/or telangiectasias along the eyelid margin.⁸⁸

Inflammation is a key component of ocular surface diseases, and while it has been more closely linked to ATD, inflammation also plays a key role in EDE and MGD.⁸⁹ Animal and human studies have shown abnormalities in inflammatory markers in DE. In an animal model, adult rats with dry eye (induced by desiccating stress) had higher levels of IL-1β, IL-6, and TNF-α expression in lacrimal glands compared to control rats.⁹⁰ In humans, various inflammatory markers have been found elevated in the tears of individuals with DE (variably defined), including cytokines (e.g., IL-1β, IL-6, TNF-α), chemokines, and prostaglandins (e.g., PGE₂ and PGD₂).⁹¹ In addition, cellular abnormalities have been described in DE, including elevated numbers of macrophages, monocytes, and T cells on the ocular surface in desiccating stress-induced models in rats and in humans.^90-95 Given that SPL are found in every cell and play important roles in regulating inflammatory responses, there is biological plausibility that SPL abnormalities may contribute to ocular surface diseases.

SPL are found in the meibum and tears of healthy individuals

SPL are a component of meibum, the lipid product produced by the meibomian glands, and have also been reported in tears.^96-103 Specifically, meibum is composed of synthesized lipids, membrane-derived lipids, and bacterial degradation products.¹⁰⁴ Non-polar lipids are the major lipid type in meibum (77%), whereas polar lipids are a minor component (8%), with SPL comprising about 30% of the polar component.¹⁰⁴

SPL are also a component of the tear film lipid layer (TFLL), the layer that slows evaporation of the muco-aqueous layer and stabilizes the tear film.¹⁰⁴ The source of polar lipids, including SPL, in tears is of keen interest. As described above, polar lipids comprise a minor fraction of meibum lipids. However, they are a major component of tears, suggesting that SPL and other polar lipids are derived from other sources as well, with ocular surface epithelial cells and the lacrimal gland being the most likely sources.^105-111

Regardless of the source of SPL in the TFLL, there has been a growing interest in the relationship between SPL and ocular surface diseases. While several studies have examined the impact of non-polar and polar lipids on various facets of DE and MGD^{107-110, 112-117}, in this review, we focus specifically on data that link SPL to DE/MGD clinical phenotypes and sub-clinical metrics. In this regard, SPL may have structural and functional consequences in maintaining ocular surface health and driving various ocular surface diseases.

SPL composition can impact meibum and tear film structure

SPL are important structural components of meibum, and alterations in SPL composition have been variably associated with TFLL instability and MGD in computer simulation (in silico), in vitro, animal, and human studies.^{104, 106, 118, 119}

In one study, Cer and SM were positively correlated with TFLL stability.¹²⁰ This study used molecular dynamic simulations - computer simulations designed to analyze physical movements of single molecules - to model the TFLL and carry out various in silico simulations to study TFLL characteristics.¹²⁰ The physiologic TFLL normally decreases the overall surface tension in tears to allow for spreading of the tears over the ocular surface, and in this study, the authors found that when the TFLL was modeled to be deficient in Cer or SM, the TFLL had increased surface tension compared to the physiologic TFLL.^{106, 120} These models suggest that Cer and SM play a role in stabilizing the TFLL and thus a decrease in Cer and SM would be expected to promote a pathologic TFLL state.

Conversely, other in vitro,¹²¹ and animal^{119, 122} studies have found increased Cer to be associated with TFLL instability and MGD. For example, in an in vitro study, meibum was collected from healthy individuals and used to create monolayer meibomian lipid films. The hysteresis and rigidity of the lipid films were analyzed using a Langmuir trough technique.¹²¹ This study found that pure physiologic meibomian lipid films had minimal hysteresis and rigidity, but sequentially increasing the Cer % in the lipid films increased the rigidity and hysteresis, leading to destabilization of the lipid films.¹²¹

Animal studies have also supported the association of increased Cer and pathologic states. In a model of epinephrine-induced MGD, meibum Cer % was higher in MGD vs. control rabbits (14-19% vs 10-11%).¹¹⁹ These data suggest a correlation between increased polar lipids (Cer) and pathologic changes in meibomian glands. Another animal study supported the correlation between Cer and clinical findings of MGD. In this model, knockout mice deficient in the enzyme stearoyl-CoA-desturase-1 (SCD1) were compared to control mice.¹²² SCD1 is found in meibomian glands and normally inhibits excessive SPL synthesis. Thus, SCD1-deficient mice had unregulated SPL synthesis, specifically, unregulated Cer through overexpression of the enzyme serine palmitoyltransferase, the initial enzyme in de novo Cer synthesis. These mice were found to have increased MG plugging with poor quality (toothpaste-like) meibum, MG atrophy with meibography, and shrinkage and atrophy of MG acini on histopathologic sections compared to wildtype mice, suggesting that the increase in Cer was associated with these morphologic changes.¹²² In the same study when serine palmitoyltransferase (and consequently de novo Cer generation) was inhibited in the knockout SCD1-deficient mice, MG morphology improved and the eyelid margin demonstrated decreased keratinization.¹²² Overall, while in silico models of the TFLL found that Cer contributed to TFLL stability, in vitro and animal studies found that increased Cer destabilized the TFLL and was associated with pathological MG morphology including keratinization, MG atrophy and plugging, and poor quality meibum. It is possible that some Cer is required for normal TFLL functioning as shown in the in silico model, but higher than normal levels of Cer contribute signs of MGD.

Another area of study has been the impact of alterations in fatty acid chain lengths on the structure of the TFLL and meibum.^{123, 127} Ceramide synthase is an enzyme that produces different subsets of Cer containing between 14 and 32 carbons, and typically, SPL are primarily composed of saturated (i.e. C16:0) and monounsaturated (i.e. C18:1) fatty acid chains.^{123, 128} In general, unsaturated and shorter chain fatty acids are more soluble than a saturated and longer fatty acids. Solubility is thought to have an inverse relationship with TFLL stability.^123-128 Animal and human studies have found that long-chain ceramides - those less soluble and more destabilizing to the TFLL – were higher in MGD. For example, in a mouse model of MGD induced by SCD-1 deficiency, which allows for unregulated SPL metabolism, SCD1 deficient mice had a higher concentration (pmol) of C16:1, C22:0, and C24:0 compared to wildtype mice (p<0.05).¹²² A similar relationship was found in human MGD studies, with higher long-chain ceramides more common in MGD. Specifically, meibum was collected from 43 individuals, and meibum quality was graded (0, clear; 1, cloudy; 2, granular; 3, toothpaste; 4, no meibum extracted) and analyzed by mass spectroscopy. Twenty-one individuals were classified as having poor quality meibum (grade 2-4) and 22 were classified with good quality meibum (grade 0 or 1). Individuals with poor vs good quality meibum had a higher percentage of ceramides with longer (less-soluble) fatty acid carbon chains (C24:0 14.96% vs 1.94%, p<0.01; C24:1 7.60% vs 3.91%, p<0.01; C26:0 5.56% vs 0.36%, p=0.01), supporting the notion that less soluble lipids contribute to poor quality meibum and may contribute to TFLL instability.²⁶

Taken together, the above in silico, in vitro, animal, and human studies suggest that there is an optimal range of SPL composition in meibum and the TFLL, and variations outside of the optimal range may alter the structure of meibum and the TFLL, leading to disease.

SPL compositional differences are noted in individuals with a variety of ocular surface diseases

When examining the contribution of SPL composition to ocular surface disease in humans, it is not possible to dissect which facet of SPL biology (structure vs function) drives the noted associations. The bioactive roles of SPL likely contribute to ocular surface diseases, predominately through their regulatory effects on apoptosis and inflammation. However, this link has not been as extensively studied as the role of alterations in SPL composition on the structure of meibum and the TFLL.

Overall, SPL composition has been found to differ in individuals with various DE/MGD phenotypes compared to controls. One study analyzed SPL difference in individuals with (n=27) and without (n=10) DE symptoms (Ocular Surface Disease Index, OSDI > 21 vs OSDI <12.9). Individuals with DE symptoms had a higher Cer % (0.04% ± 0.008% vs 0.031% ± 0.005%, p>0.05) and SM % (0.061% ± 0.014% vs 0.032% ± 0.005%, p<0.05) compared to those with no symptoms, although the difference was only significant with relationship to SM.¹⁰⁹ These data suggest an association between higher meibum SM % and DE symptoms.

However, results have not been uniform across studies. In a study of 21 individuals with blepharitis and aqueous tear deficiency (ATD), 26 with blepharitis but no ATD, and 22 controls, individuals with blepharitis and ATD had a lower meibum SM % (7.2% ± 4.8%, p <0.05) as compared to individuals with blepharitis alone (12.6% ± 5.6%) or controls (14.0±8.1).¹¹⁸ Other SPL classes, including Cer and S1P, were not assessed in this study. These studies suggests that alterations in SM, the key precursor of all other SPL classes, may play a role in DE symptoms and ATD, albeit in different directions.

We assessed the relationship between SPL composition and MGD, with some similarities and some differences compared to prior studies. In our study, meibum was collected from the inferior MGD and graded on a scale of 0-4 (good = 0 or 1, poor = 2-4). Individuals with poor meibum quality had a higher meibum SM % (58.67% ± 20.82% vs 21.84% ± 38.18%, p<0.01) compared individuals good quality. However, dissimilar to prior studies, Cer % was lower in individuals with poor meibum quality (33.6% ± 16.98% vs 49.49% ± 19.25%, p<0.01). S1P % levels followed the same pattern as Cer and were lower in those with poor meibum quality (0.16% ± 0.25% vs 0.31% ± 0.23%, p=0.05).²⁶ Taken together, both increased and decreased meibum SM % have been noted in relationship to various ocular surface diseases in cross-sectional studies.

A follow-up study by our group examined SPL composition in both meibum and tears and again found differences in composition between individuals with poor vs good meibum quality.¹²⁹ Similar to the first study, individuals with poor meibum quality were found to have less Cer (15.1% vs 17.1%, p>0.05), more SM (75.9% vs 65.8%, p<0.005), and a lower Cer/SM ratio (0.20 vs 0.25, p=0.02) in meibum compared to individuals with good quality meibum.¹²⁹ Interestingly, these differences were not replicated in tears, with no significant differences noted between the groups.¹²⁹ These data support the notion that other sources beyond meibum lipids may contribute to tear film lipids, or that eyelid margin keratinization and other eyelid changes may cause sequestration of meibum in the MG, contributing to a different SPL profile in tears vs meibum.

Finally, we also analyzed activity of both the neutral and acidic isoforms of SMase (nSMase and aSMase), the enzyme responsible for cleaving SM to Cer, in tears.¹³⁰ We found that SMase activity, measured in relative fluorescein units, was lower for both forms in individuals with poor vs good meibum quality (SMase, 4205 vs 23,748, p<0.001; aSMase 0 vs 30,893, p<0.001).¹³⁰ SMase activity was inversely correlated with SM % in meibum (rho −0.33, p<0.05), which is an expected finding, as increased activity in SMase would result in less SM.¹³⁰ This relationship was similarly seen for SM % in tears (nSMase, rho −0.30, p<0.05 and aSMase −0.29, p<0.05).¹³⁰ In addition, a positive correlation was noted between SMase and S1P % in tears (nSMase, rho 0.51, p<0.001 and aSMase 0.49, p<0.001).¹³⁰ Decreased SMase activity in individuals with poor meibum quality may explain the higher SM % and lower Cer % noted in the meibum of individuals with poor vs good meibum quality. However, to truly establish this relationship, SMase activity in meibocytes, not tears, would need to be analyzed. Together, while relationships between SPL classes, SMase activity, and MGD symptoms and signs need further characterized in larger, longitudinal studies, current data suggest alterations in SPL metabolic pathways in MGD.

Finally, our group also examined relationships between SPL composition, SMase activity and clinical metrics of DE/MGD. Interestingly, no significant associations were found between meibum SPL composition and DE/MGD symptoms or signs.¹²⁹ However, some significant associations were noted in tears, including a positive correlation between S1P % and DE symptoms (Dry Eye Questionnaire-5, DEQ5, rho 0.31, p<0.01 and OSDI, rho 0.41, p<0.001) and nSMase activity and DE symptoms (DEQ5, rho 0.29, p<0.05 and OSDI, rho 0.30, p<0.05). This suggests that increased S1P, hydrolyzed by SMase from SM, contributes to DE symptoms.^{129, 130} With regards to signs, Schirmer’s score was negatively correlated with Cer % (rho −0.49, p<0.001) and positively with S1P % (0.29, p<0.05) in tears.¹²⁹ In addition, tear aSMase activity was negatively correlated with MG dropout (rho −0.35, p<0.01) and plugging (rho −0.30, p<0.001).¹³⁰ Together, these associations suggest that SM-derived Cer and S1P are related to some DE/MGD signs. However, these observations need more careful review in longitudinal studies.

Together, data demonstrate that SPL composition varied in individuals with ocular surface diseases compared to controls but relationships were not consistent across disease phenotypes (Table 1). This may be due to methodologic issues (sampling, cross-sectional design) or may suggest that different SPL signatures are found in different diseases. Given the cross-sectional nature of most studies, it is difficult to comment on the dynamic equilibrium between lipid species, or distinguish whether alterations primarily impact the structure or function of meibum and the TFLL. Further studies are needed to better characterize mechanisms that underlie these noted associations.

Table 1.

Comparison of changes in sphingolipid composition by patient population

Patient Population	Source	Cer (%)	SM (%)	S1P (%)
DE Symptoms109 vs No DE Symptoms	Meibum	Increased	Increased	Not studied
Blepharitis + ATD vs Blepharitis118	Meibum	Not studied	Decreased	Not studied
Blepharitis + ATD vs Controls118	Meibum	Not studied	Decreased	Not studied
MGD: Individuals with poor quality meibum vs good quality meibum26	Meibum	Decreased	Increased	Decreased
MGD: Individuals with poor quality meibum vs good quality meibum129	Meibum	Decreased*	Increased	Not studied
MGD: Individuals with poor quality meibum vs good quality meibum129	Tears	Increased*	Increased*	Not studied

DE, dry eye; ATD, aqueous tear deficiency; MGD, meibomian gland dysfunction; Cer, ceramide; SM, sphingomyelin; S1P, sphingosine 1-phosphate

^*Difference was not statistically significant

Clinical Implications for Ocular Surface Disease

A better understanding of the relationships between bioactive lipids, inflammation, DE, and MGD is important so as to evaluate potential therapeutic targets within SPL pathways. With our current understanding, a number of targets and therapeutic candidates have emerged.

Given that changes in Cer % in meibum have been associated with increased TFLL rigidity in studies of meibomian lipid films in vitro¹²¹, with morphologic MG changes in animal models of MGD^{119, 122}, and with signs of MGD^{26, 129, 130} and symptoms of DE^{26, 109, 129, 130} in humans, modulation of Cer represents a possible therapeutic target. In the mouse model of MGD induced by SCD1 deficiency described above¹¹², control mice were first treated with an SCD1 inhibitor, which allowed for uninhibted Cer generation through the de novo pathway. Mice demonstrated morphologic signs of MGD including MG atrophy, MG plugging, increased expression of keratin-10, a biomarker of keratinization, and increased CD45 positive inflammatory neutrophils within the meibomian glands.¹¹² When the mice were co-treated with the SCD-1 inhibitor and an inbitor of serine palmitoyltransferase, thus inhibiting de novo Cer synthesis, mice maintained normal MG morphology, showed reduced infiltration of inflammatory neutrophils, and decreased biomarkers of keratinization.¹¹² These data suggest that inhibition of Cer within meibomian glands can maintain normal MG morphology and prevent increased inflammatory cell infiltration.

Another candidate molecule is Fingolimod (FTY720), a medication that is FDA-approved for the treatment of relapsing multiple sclerosis.¹³¹ FTY720 is primarily thought of as an agonist of the S1P receptor after the molecule is phosphorylated by sphingosine kinase. However, FTY720 has also been shown to be an inhibitor of ceramide synthase (CerS), and thus an inhibitor of Cer generation, in human lung endothelial cells.¹³² In a model of light induced retinal degeneration in albino rats, quantitative analysis by mass spectroscopy showed increased Cer levels in the retina, generated through the de novo pathway, in rats exposed to light vs. controls (p<0.01). Systemic FTY720 protected retinal photoreceptors from light-induced damage, with preservation of retinal photoreceptor function on electroretinogram and structure on histopathologic analyses. Concomitantly, the light-induced increase in retinal Cer was blocked by FTY720, potentially underlying the improvement in clinical phenotype.¹³³ These results in retinal disease present a case for studying the role of FTY720 driven Cer inhibition in MGD, especially given the dissimilar results – higher Cer% associated with TFLL instability (in vitro), higher Cer% associated with MG morphologic changes (animal), but lower Cer% associated with individuals with poor meibum quality (human) – in prior cross-sectional studies.

Although Cer inhibition was shown to be protective in animal studies, given the numerous roles of Cer throughout the body, inhibition of systemic de novo Cer synthesis may have other detrimental consequences outside of the eye, so candidate molecules would likely need to preferentially inhibit Cer in the MGs.

The other SPL that has been extensively studied as a therapeutic target in many diseases is S1P. Given the associations between S1P and signs of MGD and symptoms of DE in human studies, this is also a compelling therapeutic target.^{26, 129, 130} As mentioned, FTY720 is most well-known for its action after it is phosphorylated, as an agonist of the S1PR1 receptor on lymphocytes. This leads to internalization of this receptor and prevention of egress of lymphocytes from secondary lymphoid organs, effectively sequestering inflammatory cells within these organs and reducing the population of peripherally circulating inflammatory cells.²⁰

Several animal studies have investigated the role of FTY720 in DE. In one study, non-obese diabetic (NOD) mice - mice with increased inflammatory cytokines in tears - were used as a DE model.¹³⁴ Mice were divided into six groups (n=24 in each group) and treated with 0.001%, 0.005%, and 0.05% FTY720 (FTY) eye drops, along with 0.05% cyclosporine A (CsA), and saline (controls). Inflammation was assessed by immunohistochemistry in the conjunctival epithelium, and inflammatory cytokines were profiled by PCR in lacrimal glands and the tears after 8 weeks of treatment. Mean TBUT was longer in FTY720 and CsA treated mice compared controls (p<0.01) (0.05% FTY, 3.50±0.19; 0.005% FTY, 4.14±0.26; 0.001% FTY, 2.68±0.31; 0.05% CsA, 3.38±0.32; normal saline, 1.86±0.27), and mice treated with 0.005% FTY had the longest mean TBUT compared to all other treatment groups (p<0.05). Furthermore, FTY720 had an effect on ocular surface inflammation. While control mice had increased levels of IL-1β in tears from weeks 2-8, levels decreased and remained low in all other groups (p<0.01). FTY720 and CsA also impacted lacrimal gland inflammation as mRNA levels of both IL-1β and TNFα within lacrimal glands were significantly lower (p<0.01) in treated mice vs. controls. Of all doses, the 0.005% and 0.05% FTY720 doses balanced efficacy and side effects.^{134, 135}

Similar findings were noted in other studies that treated NOD mice with various FTY720 concentrations.¹³⁶ Compared to wildtype and FTY-treated NOD mice, untreated NOD mice had higher levels of IL-1β, CD30L, CXCL16, IL-6, IL-17F, and IL-21 in the bulbar conjunctiva as measured by a protein assay (p<0.01).

Lower concentrations of FTY have also been found to have a positive effect on inflammation. In a separate experiment, NOD mice received eyedrops 3 times daily with FTY720 (0.001%, 0.005%. and 0.05%) or normal saline or received 0.05% topical FTY270 co-administered with W146, a S1PR1 antagonist, which effectively blocked the actions of FTY720.¹³⁶ Relative to saline-treated mice, FTY720-treated mice had significantly decreased levels of S1PR1 (the receptor necessary for leukocyte migration) and phosphorylated MAPKs (a key pathway in cytokine generation), as well as decreased numbers of CD45-positive leukocytes (representing infiltrating neutrophils) in the conjunctiva as assessed by Western blot (p<0.01), indicating it was efficacious at decreasing the amount of inflammatory cells and cytokines on the ocular surface, with 0.05% FTY720 being the most effective concentration this experiment.¹³⁶ Finally, the NOD mice co-treated with FTY270 and the S1PR1 antagonist W146 had similar findings to the control NOD mice, supporting the importance of S1P receptor signaling in ocular inflammation. Together, these studies suggest a potential role for topical FTY720 in treating ocular surface disease through reducing the inflammatory milieu on the ocular surface. It is important to note however, that NOD mice were found to have abberant patterns ocular inflammation,¹³⁷ so the same effects may not be seen in humans, and so further studies investigating dosing, safety, and efficacy in humans are needed.

Finally, although a complete discussion of the role of androgens and ocular surface disease is out of the scope of this review, the interactions between androgens and S1P signaling and the possible therapeutic pathways are worth mentioning. Androgen deficiency, especially in individuals on anti-androgen therapy, has been associated with clinical symptoms and signs of DE.¹³⁸ Moreover, androgens have roles in regulating SPL metabolism, especially through S1P receptor signaling. Further studies are needed to understand the crosstalk between androgens and SPL that may uncover a role for androgen supplementation in modulating the SPL mediated effects in DE/MGD.^137-140

Conclusion

In conclusion, SPL are found in meibum and tears as a low percentage of total lipids. Compositional changes in SPL have been noted to have an effect on MG anatomy and tear film stability in several pre-clinical studies. In humans, various composition differences have been noted in various ocular surface diseases, with discrepancies between studies. However, the family of SPLs has been found to impact ocular surface inflammation and clinical signs of DE/MGD in animal and human studies, providing a basis for further studies that investigate the role of SPL modulators as therapeutic avenues in ocular surface disease. However, several limitations were noted when reviewing the current studies including small sample sizes, dissimilar study populations thus preventing comparison across studies, and primarily cross-sectional studies, thus limiting inferences about dynamic relationships between SPL metabolites. As such, future studies will need to examine meibum and tear SPL composition in varied populations and analyze their relationships to various aspects of DE/MGD (symptoms, tear stability, MG parameters). Beyond cross-sectional studies, longitudinal natural history and intervention studies are needed to clarify the impact of current DE/MGD therapies on SPL metabolism and composition. More information is also needed regarding the interactions and interplay between SPL and other lipid mediators, such as eicosanoids. Overall, while our review highlights the potential role of SPL in DE and MGD, more information is needed on the contributions of SPL to ocular surface health and disease and to explore the potential therapeutic applications of SPL pathway manipulation.

List of abbreviations:

aSMase

Acidic sphingomyelinase

AMD

Age-related macular degeneration

ATD

Aqueous tear deficiency

AA

Arachidonic acid

c/EBP

CAAT/enhancer binding proteins

CaLB/C2

Calcium-dependent lipid binding domain

Cer

Ceramide

C1P

Ceramide 1-phosphate

CerK

Ceramide kinase

CerS

Ceramide synthase

COX

Cyclooxygenase

cPLA₂

Cytosolic phospholipase A2

DAMPs

Damage-associated molecular patterns

DR

Diabetic retinopathy

DHA

Docosahexaenoic acid

DE

Dry eye

EPA

Eicosapentaenoic acid

EDE

Evaporative dry eye

GlcCer

Glucosylceramide

IL

Interleukin

LacCer

Lactosylceramide

LT

Leukotriene

LOX

Lipoxygenase

LOX

Lipoxygenase

MMP-2

Matrix metalloproteinase-2

MGD

Meibomian gland dysfunction

ASAH1

N-acylsphingosine amidohydrolase 1

nSMase

Neutral sphingomyelinase

PAMPs

Pathogen-associated molecular patterns

POAG

Primary open-angle glaucoma

PG

Prostaglandin

RPE

Retinal pigment epithelium

Spns2

S1P transporter spinster homolog 2

SM

Sphingomyelin

SMase

Sphingomyelinase

Sph

Sphingosine

S1P

Sphingosine 1-phosphate

S1PR

Sphingosine 1-phosphate receptor

SphK

Sphingosine kinase

TFLL

Tear film lipid layer

TNF-a

Tumor necrosis factor alpha

TNFR

Tumor necrosis factor receptor

TACE

Tumor necrosis factor-alpha converting enzyme