Lipid microdomains (“rafts”) are dynamic, nanoscale regions of the plasma membrane enriched in cholesterol and glycosphingolipids, that possess distinctive physicochemical properties including higher order than the surrounding membrane. Lipid microdomain integrity is thought to affect neurotransmitter signaling by regulating membrane-bound protein signaling. Among the proteins potentially affected are monoaminergic receptors and transporters. As dysfunction of monoaminergic neurotransmission is implicated in major depressive disorder and other neuropsychiatric conditions, interactions with lipid microdomains may be of clinical importance. This systematic review evaluates what is known about the molecular relationships of monoamine transporter and receptor regulation to lipid microdomains. The PubMed/MeSH database was searched for original studies published in English through August 2017 concerning relationships between lipid microdomains and serotonin, dopamine and norepinephrine transporters and receptors. Fifty-seven publications were identified and assessed. Strong evidence implicates lipid microdomains in the regulation of serotonin and norepinephrine transporters; serotonin 1A, 2A, 3A, and 7A receptors; and dopamine D1 and β2 adrenergic receptors. Results were conflicting or more complex regarding lipid microdomain associations with the dopamine transporter, D2, D3, and D5 receptors; and negative with respect to β1 adrenergic receptors. Indirect evidence suggests that antidepressants, lipid-lowering drugs, and polyunsaturated fatty acids may exert effects on depression and suicide by altering the lipid milieu, thereby affecting monoaminergic transporter and receptor signaling. The lipid composition of membrane subdomains is involved in localization and trafficking of specific monoaminergic receptors and transporters. Elucidating precise mechanisms whereby lipid microdomains modulate monoamine neurotransmission in clinical contexts can have critical implications for pharmacotherapeutic targeting.
Plasma membrane structure–function relationships
A growing area of research over the last two decades concerns structural and functional characteristics conferred by the lipid composition of plasma membranes. The distinct physicochemical properties of different lipid species are accommodated in ways that impose organization on the membrane, with consequences for the distribution and regulation of proteins within the membrane bilayer .
Studies in model membrane systems show that lipid domains rich in sphingolipids and cholesterol are highly ordered, existing in a tightly-packed, gel-like state with an aversion to the unsaturated acyl chains of glycerophospholipids, which occur in a loosely-packed, liquid disordered state (reviewed in refs. [2, 3]). These observations support the paradigm that membrane lipid microdomains (lipid rafts) exist as physically and temporally dynamic, nanoscale membrane assemblies enriched in cholesterol, sphingolipids, and glycosylphosphatidylinositol (GPI)-anchored proteins .
Lipid microdomains also have been studied in cell membranes [5,6,7,8,9], leading to a proposed categorization into caveolar and non-caveolar, or planar, rafts . Both raft types exhibit high cholesterol and sphingomyelin content, enrichment in GPI-anchored proteins, cytoskeletal association, and resistance to detergent extraction. Most rafts are reported to be of the caveolar type. Caveolae are small, flask-shaped invaginations of the membrane that function in cell signaling, lipid trafficking, and endocytosis, an important mechanism for downregulating cell-surface levels of proteins such as transporters. Key relevant characteristics are the absence of clathrin in caveolar endocytosis and the presence of caveolin and cavin proteins [11, 12], which are present in brain . Another type of raft-related, non-clathrin-mediated endocytotic mechanism involves flotillin proteins (also known as reggies) that are also present in the central nervous system [13, 14] and analogous functionally, but not homologous structurally, to caveolin . Caveolins and flotillins serve a scaffolding/organizing function, anchoring signaling components to lipid microdomains . In addition to stimulating movement into the cell by endocytosis across the membrane, neurotransmitter signaling also can cause movement of signaling proteins into or out of lipid microdomains by translocation or lateral movement within the membrane bilayer .
However, countervailing structural models also have been proposed. One alternative is a continuous bilayer with an outer membrane leaflet rich in cholesterol and sphingolipids, with a large ordered fraction of the plasma membrane [17, 18]. In another suggested model, sphingolipids form distinct domains not dependent on cholesterol , and a strong role for the actin cytoskeleton is implicated [19,20,21,22,23]. Some investigators propose that lipid domains are larger than nanoscale assemblies and are not organized on the basis of phase-differentiation . The most extreme view suggests that microdomains do not exist but are an artifact of the methods used to study them [21, 24]. Finally, others have attempted to reconcile disparate observations into an integrative model .
This controversy is fueled by the uncertain validity of extrapolating information from artificial model membranes, combined with technical difficulties inherent in the study of submicroscopic, temporally dynamic nanostructures . Common methods used to isolate lipid microdomain-containing cellular fractions have drawbacks; notably, the use of nonionic detergent-resistant membranes (DRM) [25, 26] produce results that are inconsistent  and are not be adequately representative of in vivo lipid microdomains . Evidence that the detergent itself induces artifacts was presented over a decade ago . Possibly less perturbative methods for isolating raft-like domains include sucrose gradient ultracentrifugation , which ostensibly avoids the pitfalls of DRM, but since this technique includes treating the membranes with alkaline buffer, it may cause saponification, essentially retaining a detergent-like effect. Cholesterol depletion is commonly used under the assumption that subsequent functional loss is evidence for lipid microdomain disruption. Disruptive techniques include: (1) sequestration of cholesterol using filipin or nystatin ; (2) depletion and removal of cholesterol using methyl-β-cyclodextrin ; (3) inhibition of cholesterol synthesis using HMG-CoA reductase inhibitors (statins) ; and (4) genetic approaches using caveolin-1 knockout mice  or caveolin RNA interference (siRNA) . However, because the use of cholesterol-disrupting, and thus lipid raft-disrupting, drugs can produce pleiotropic effects, including effects on non-raft domains  and on the actin cytoskeleton , data from cholesterol depletion experiments ideally require validation. The restoration of cholesterol or replacement with its analog, desmosterol (a non-raft promoting sterol) is sometimes used as an experimental control. Depletion experiments can likewise clarify the role of sphingolipids, using Fumonisin B1 [35, 36], and actin, using the actin filament disrupter cytochalasin D .
The resolution of these controversies will require continuing development of technologies to more directly study the lipid–protein milieu, such as super-resolution microscopy coupled with fluorescence and bioluminescence techniques  harnessed to protein-conjugated quantum dots, which enables nanoscale and microscale quantitation and visualization of receptor clustering in living cells . Sophisticated techniques also include stimulated emission depletion (STED) super-resolution nanoscopy , Förster resonance energy transfer (FRET) [40, 41], fluorescence correlation spectroscopy (FCS) , k-space image correlation spectroscopy (kICS) , and magic-angle spinning nuclear magnetic resonance (HR-MAS NMR) [43, 44], each of which can provide valuable information on the nanometer scale.
Structural characteristics of lipid microdomains are understood to influence plasma membrane proteins in several ways, including (1) providing a scaffolding for spatially co-localizing the membrane proteins with other elements important in cell-cell signaling [45, 46]; (2) affecting lateral movement of proteins ; (3) altering molecular trafficking, e.g., endocytosis of proteins from the cell membrane surface to the cell interior [48,49,50] to achieve receptor recycling, sequestration, and/or downregulation; and (4) directly regulating activity, e.g., modulation of the transport rate or affinity of transporter proteins for substrates by the relative amounts of cholesterol [51, 52].
Lipid microdomains as regulators of monoaminergic signaling
Among membrane proteins affected by lipid microdomains are a variety of monoamine transporters and receptors, important for emotion regulation. Monoamine neurotransmitters comprise a key family of neural signaling proteins, including serotonin (5-hydroxytryptamine, 5-HT), dopamine (DA), and norepinephrine (NE). Serotonin modulates mood, aggression, motivation, appetite, sleep, cognition, and sexual activity; and altered 5-HT signaling has been implicated in neuropsychiatric diseases [53,54,55,56,57]. DA systems control motor function, mood, reward, and cognition [58, 59], dysregulation of which is linked to attention deficit hyperactivity disorder (ADHD), schizophrenia, substance use disorders, and Parkinson’s disease . NE regulates arousal, mood, attention, and stress-responsiveness [61,62,63,64].
Monoaminergic transporters and receptors
Plasma membrane monoamine transporters govern the effects of their respective neurotransmitters within the synaptic cleft by recycling them back into the synaptic boutons, with direct effects on intra-synaptic neurotransmitter concentrations and downstream indirect effects on monoaminergic receptors [65, 66]. Thus, altering the quantity of functional transporters at the neuronal or glial plasma membrane represents one mechanism for regulating synaptic levels of neurotransmitters . These transporters are integral membrane phosphoproteins affected by kinase and phosphatase activities (reviewed in ref. ) and are also pharmacological targets of many psychoactive drugs [68,69,70]. Studies in knockout mice have demonstrated that the mechanisms regulating monoamine neurotransmitter biosynthesis and storage, receptor sensitivity, and transporter expression are interdependent [71,72,73,74,75,76,77,78]. A deeper understanding of lipid microdomain regulation of monoamine transporter function is likely to foster new treatment targets for neuropsychiatric disease.
Monoaminergic receptors exist in multiple subtypes and may be pre-synaptic or post-synaptic. With the exception of 5-HT3, monoaminergic receptors belong to the G-protein-coupled receptor superfamily, with distinct receptor subtype specificity for G-proteins. Depending on receptor subtype and the concomitant G-protein, agonist binding to monoaminergic receptors activates or inhibits second messengers adenylyl cyclase (AC) and cyclic AMP (cAMP), or inositol triphosphate and diacylglycerol.
We summarize the literature concerning lipid raft involvement in monoaminergic transporter and receptor regulation and depression (Fig. 1), including (1) co-localization of transporters/receptors with lipid microdomains, ganglioside GM1, and raft-associated proteins in detergent-resistant membranes or low-buoyancy sucrose fractions; (2) lateral transporter movement; (3) association of transporters/receptors with caveolar proteins; (4) lipid influences on receptor transport rate; (5) effects on agonist and antagonist receptor binding; (6) downstream effects on second messengers; and (7) sensitivity to effects of cholesterol depletion (Fig. 1).
Literature search strategy
Animal and human experimental studies on monoamine neurotransmitters, receptors and transporter proteins, and lipid microdomains published in English through August 2017 were identified by searching with MEDLINE Medical Subject Heading (MeSH) terms. In addition, general PubMed searches were conducted. Duplicate papers were removed, and articles were excluded if not clearly germane. The reference lists of prior reviews and research articles were also manually searched to find other potentially relevant studies (see detailed strategy, Table 1.)
Combined PubMed/MeSH searches identified 250 potentially eligible reports, of which 27 review articles were eliminated; 183 were excluded as not directly concerning monoaminergic systems, were focused narrowly on non-brain systems, or were primarily concerned with other elements of the signaling pathways, yielding 37 articles. Twenty additional studies were identified through manual bibliography searching, yielding 57 reports considered in the Results section of this systematic review, which we view as appropriately summarizing the current understanding of lipid microdomain associations with monoaminergic transporters and receptors. Of these investigations into the interactions with lipid microdomains, 21 studies reported on serotonergic systems [16, 33, 34, 36, 51, 52, 90, 93, 117,118,119,120,121,122,123,124, 126,127,128,129]; 25 reported on dopaminergic systems [35, 79–81, 99,100,101,102, 105–108, 110, 134,135,136, 138,139,140,141,142,143,144,145,146]; and 12 reported on noradrenergic systems [46, 113, 147,148,149,150,151,152,153,154,155,156].
Monoamine transporters and lipid microdomains
Transporter relationship to plasma membrane is modulated by phosphatidylinositol 4,5-bisphosphate (PIP2)
The monoaminergic transporters for serotonin (SERT), dopamine (DAT), and norepinephrine (NET) are all examples of neurotransmitter:sodium symporters with a 12-transmembrane domain structure that form oligomers in intact cells [82, 83]. One important mechanism whereby these transporters are anchored in lipid microdomains is through binding to PIP2 . Found in low concentration in lipid microdomains [85, 86], in the cytoplasmic leaflet of the cell membrane , PIP2 nevertheless has an important role in monoaminergic signaling by stabilizing transporter oligomerization [84, 88, 89] and regulating neurotransmitter efflux [88, 89].
Co-localization of SERT with caveolin and the ganglioside known as GM1 has been demonstrated in rat brain synaptosomes . SERT proteins exist in two states with respect to their degree of lateral mobility: more mobile transporters that can diffuse relatively freely, and less mobile transporters localized to cholesterol and GM1 ganglioside-enriched microdomains . In RN46A serotonergic neuronal cells labeled with quantum dots conjugated to the SERT ligand IDT318, confocal imaging reveals SERT clusters in cholesterol-rich membrane microdomains. Studied before and after cholesterol depletion with methyl-β-cyclodextrin, SERT molecules evince greater lateral movement after reduction of cholesterol, consistent with lipid raft disaggregation. In addition, actin filaments have a role in anchoring SERT molecules via their C terminus, since the actin filament disrupter cytochalasin D also results in increased SERT diffusion rates .
Increased SERT mobility after cholesterol depletion or actin disruption is accompanied by decreased transport activity . Functional changes induced by cholesterol depletion result in a mean decrease of 50% in the transport rate (Vmax) of SERT [51, 52], and a concurrent reduction in SERT affinity (Km) for 5-HT, suggesting that lipid microdomains may regulate SERT by promoting a high-affinity state [51, 52]. Consistent with this hypothesis, signaling pathways that upregulate SERT activity, particularly p38 MAPK, act on SERT localized within membrane microdomains . Conversely, in synaptosomes, activation of PKC downregulates SERT, causing redistribution from lipid raft to non-raft fractions .
Trafficking of SERT appears to be homeostatically regulated, as platelet studies have found that application of 5-HT variously induces SERT upregulation [91, 92] or downregulation [93, 94], which may depend on complex factors including 5-HT concentration and acute vs. chronic exposure . SERT downregulation can be mediated by 5-HT binding to 5-HT2A autoreceptors via activation of Gαq proteins and downstream PKC activation [95, 96]. PKC exerts its effects on SERT by phosphorylating the transporter ; under some conditions, the SERT internalization can be blocked by 5-HT . An analogous process has been reported for DAT (see below).
DAT are equally distributed across raft and non-raft domains , and their function is regulated by cholesterol . Cholesterol depletion with methyl-β-cyclodextrin decreases Vmax and Km values for DA uptake  and efflux [100, 101] via DAT. However, in contrast to the SERT findings, cholesterol effects on DAT may not be mediated by caveolar endocytosis, as DAT downregulation induced by methyl-β-cyclodextrin does not occur by reduction of surface DAT expression [99, 101]. Moreover, although methyl-β-cyclodextrin depletes cholesterol across both raft and non-raft membrane domains , cholesterol chelation with nystatin or filipin primarily disrupts raft-localized cholesterol  and has no effect on DA uptake and efflux rates [100, 102]. Additionally, supranormal repletion with desmosterol—a sterol that does not form highly ordered domains and thus does not affect lipid rafts —was as effective as cholesterol in restoring DAT transport rates . Finally, DAT is internalized in a clathrin/dynamin-dependent process . As an alternative explanation to lipid raft- mediated effects, studies with DAT mutations suggest that cholesterol may impact DA uptake by promoting an outward-facing conformation [100, 105].
Contradictory evidence exists concerning a lipid raft role in DAT endocytosis. Lipid raft-mediated DAT endocytosis is linked to the GTPase known as Rin or Rit2 , and to flotillin-1 , both reportedly localized in raft domains. Different regulatory pathways have been suggested for the “raft” and “non-raft” DAT populations , since basal DAT internalization occurs similarly from raft and non-raft domains, while PKC-stimulated DAT internalization occurs only from lipid raft domains . The latter finding does not agree with reports that (1) the clathrin-mediated (i.e., non-raft) endocytosis inhibitor concanvalin A blocks PKC-stimulated loss of surface DAT ; (2) coexpression with a dominant-negative mutant of dynamin I inhibits internalization of DAT ; and (3) interference with clathrin and dynamin significantly reduces PKC-dependent DAT endocytosis, but flotillin depletion does not . However, depletion of flotillins does decrease lateral mobility of DAT , suggesting a limited role for lipid microdomains.
Similar to effects of 5-HT on SERT, DAT substrates, DA and amphetamines, induce internalization and thus downregulation of DA transport [109, 111]. This DA-stimulated trafficking of DAT is accomplished by DA binding to D2 autoreceptors, triggering the downstream activation of Gαi and protein kinase C .
Like SERT, NET is found in lipid microdomains, and NET downregulation is triggered by PKC via a cholesterol-dependent mechanism blocked by filipin and nystatin and accomplished through a lipid raft-mediated internalization that is surprisingly independent not only of dynamin and clathrin but also of caveolae . However, this NET internalization depends on complexing with the neurokinin-1 receptor (NK1R), which occurs in non-raft domains; after agonist binding to NK1R, the NET-NK1R complex translocates to lipid raft membrane domains .
Mechanisms of NET inhibition through internalization appear to differ from those described above for SERT and DAT, in that NE does not demonstrate homeostatic regulation of NET . Treatment of PC12 cells with even high concentrations of NE does not alter NET binding properties . Additionally, in 293-hNET cells without the ability to produce NE, NET is downregulated by desipramine or nisoxetine equally as well as are NE-synthesizing PC12 cells .
Monoamine receptors and lipid microdomains
Serotonin (5-HT) receptors
The 5-HT1A receptor, an ion channel ligated autoreceptor, is coupled to Gαi, which mediates inhibitory effects of 5-HT binding on AC and cAMP . Evidence for involvement of lipid microdomains in 5-HT1A regulation includes findings in cell cultures that 5-HT1A receptors distribute to detergent-resistant lipid microdomain fractions together with caveolin-1 [93, 117]. Cholesterol depletion with methyl-β-cyclodextrin inhibits both agonist  and antagonist  binding and G-protein coupling  to 5-HT1A receptors, effects that are reversible with exogenous cholesterol . More direct evidence is the identification of a cholesterol recognition amino acid consensus motif in the 5-HT1A receptor . In a model membrane system, increasing membrane order by manipulating cholesterol, ergosterol, epicholesterol, or sphingomyelin results in increased 5-HT1A receptor activity .
The 5-HT2 receptor couples to Gαq, increasing inositol triphosphate and diacylglycerol levels . 5-HT2A receptors are abundantly expressed in brain and targeted by virtually all atypical antipsychotic agents. Co-localization of 5-HT2A receptors with caveolins occurs in sucrose gradient fractions in a rat tail artery preparation ; as co-immunoprecipitated in C6 glioma cells, human embryonic kidney (HEK-293) cells, or rat brain synaptic membrane preparations ; and as co-immunoprecipitated in cardiac myocytes . A single study in bovine tracheal muscle, however, found that 5-HT2A co-immunoprecipitated in the non-caveolar fraction , although the immunoprecipitation was not studied after 5-HT stimulation. This may be relevant since in cardiac myocytes, 5-HT2A receptors only co-immunoprecipitate with caveolin-3 (a muscle-specific caveolin) upon serotonin stimulation, and are then redistributed into caveolae microdomains, trafficking that is abolished by caveolin-3 knockdown . However, in C6 glioma cells, 5-HT2A and caveolin-1 co-localize, on the cell surface and in intracellular vesicles, exclusively in the unstimulated condition, as 5-HT agonists administration does not induce 5-HT2A receptor internalization . These dissimilarities may be tissue-specific or may reflect methodological differences.
The association of 5-HT2A receptors with caveolae microdomains has functional consequences, inhibiting 5-HT2A receptor-mediated expression of the atrial natriuretic factor (ANF) gene and activation of nuclear factor of activated T cells (NFAT) in cardiac myocytes ; and stimulating calcium flux in rat brain synaptic membranes . Physiologic responses to 5-HT stimulation are disrupted by the cholesterol depletion with methyl-β-cyclodextrin and restored by cholesterol repletion in the bovine airway smooth muscle  and rat tail artery  preparations, and are abolished by caveolin-1  or caveolin-3  knockdown.
The only 5-HT receptor that does not bind G-proteins, 5-HT3 is a member of the Cys-loop pentameric receptor family consisting primarily of five 5-HT3A monomers (although some also contain 5-HT3B). Serotonin binding to 5-HT3 triggers fast signal transduction across synapses by opening Ca2+ channels presynaptically and Ca2+, Na+, or K+ channels postsynaptically . Using fluorescence resonance energy transfer technology (FRET), lipid microdomains can be measured on a nanometer scale , and ligand binding to 5-HT3A receptors can be quantitated . Cholesterol depletion with methyl-β-cyclodextrin reduces 5-HT3 function, affecting peak amplitude and kinetics of serotonin-stimulated cation currents . In HEK-293 and neuroblastoma (N1E-115) cells transfected with 5-HT3A receptors, 5-HT3 receptor proteins localize exclusively in the low buoyant, lipid raft-containing fractions of the sucrose gradient, along with a high content of cholesterol, caveolin-2, and flotillin-1, and a number of antidepressant and antipsychotic medications . In addition to co-localization of these drugs with 5-HT3 receptors, drug concentrations within lipid microdomain fractions correlated highly with their inhibition of serotonin-induced cation currents. This non-competitive antagonism at the 5-HT3 receptor was not caused by increased receptor internalization, as demonstrated by immunofluorescence, receptor density in clathrin-coated vesicles, and electrophysiology ; the antidepressant antagonism of 5-HT3 also was shown to be independent of the association of 5-HT3 with lipid microdomains [127, 128].
The 5-HT7 receptor is coupled to Gαs, stimulating cAMP levels . Caveolin-1, cholesterol, sphingomyelin, and gangliosides all regulate 5-HT binding to the 5-HT7 receptor as well as agonist-induced internalization and signaling. Caveolin-1 and 5-HT7 receptors co-localize in lipid microdomains . Reduction of cholesterol, by methyl-β-cyclodextrin treatment or cholesterol synthesis inhibition, reduces agonist and antagonist binding to the 5-HT7 receptor in cell culture, with downstream decreases of 5-HT-induced CREB phosphorylation . Independent of cholesterol-mediated effects, inhibition of sphingolipids and gangliosides attenuates agonist binding at 5-HT7 receptors .
Dopamine (DA) receptors
Dopamine receptors are categorized as D1-class (D1 and D5) or D2-class (D2, D3, D4) based on G-protein receptor coupling and associated respective stimulation or inhibition of AC and production of cAMP: stimulating via Gαs/Gαolf with D1-class, and inhibiting via Gαi/Gαo with D2-class , although D3 may also couple to Gαs [131, 132]. In the central nervous system, D1-class receptors are primarily post-synaptic, while D2-class have both pre- and post-synaptic distributions .
D1-class (D1 and D5) receptors
D1 receptor: Using both detergent solubilization of rat brain  and detergent-free HEK-293 preparations [35, 81], D1 receptors have been predominantly localized to lipid microdomains. Within the membrane, D1 localization depends on receptor conformation . As expected, D1 co-fractionates with Gαs, caveolin-2, flotillins, and other signal transduction proteins . However, D1 receptor endocytosis can occur via both clathrin-dependent (i.e., non-lipid raft) [134, 135] and caveolar (raft) [81, 135] mechanisms. In heterologously D1-expressing human renal proximal tubule (hRPT) and HEK-293 cells, a D1-class receptor agonist fenoldopam increases AC protein in lipid rafts, but not in non-lipid raft domains . Consistent with lipid raft localization, disruption of lipid microdomains with methyl-β-cyclodextrin reduces basal AC activity .
D5 receptor: The D5 subtype is a high-affinity receptor  occurring predominantly in cytoplasmic fractions and not in lipid raft-like subdomains in two studies of rat frontal cortex [80, 138]. Like D1 receptors, D5 receptors co-localize with and enhance the activity of anti-oxidizing enzyme paraoxonase in the plasma membrane of hRPT cells and HEK-293 cells heterologously expressing D5. However, in contrast to D1 receptors, D5 agonist binding in HEK-293 cells increases paraoxonase activity  and AC protein  only in non-rafts although cholesterol depletion with methyl-β-cyclodextrin does reduce AC activity .
D2-class (D2, D3, D4) receptors
D2 receptor: Studies of the D2 receptor subtype have produced conflicting results. Immunoreactivity assays find a diffuse distribution of the D2 receptor subtype spanning cytoplasmic, detergent-soluble and detergent-resistant membrane fractions in rat frontal cortex . D2 receptors are expressed either endogenously in brain or exogenously in HEK-293T cells and localize to lipid microdomains [140, 141], but are unaffected by cholesterol depletion using methyl-β-cyclodextrin . Internalization of D2 receptors through caveolar endocytosis  depends on glycosylation state , but dynamin-dependent internalization is also reported [143, 144].
D3 receptor: Although less abundant than other DA receptors, the D3 receptor has a much higher affinity for agonists than D2 receptors. In human proximal tubule cells, D3 receptors co-localize with G-protein-coupled receptor kinase 4 (GRK4) both in non-raft fractions (90%) and, together with caveolin-1, in lipid raft fractions (10%) isolated using detergent-free sucrose fractionation and co-immunoprecipitation methods . Cholesterol depletion using methyl-β-cyclodextrin causes redistribution of raft-associated D3 to non-raft fractions . In contrast to D2 receptors, glycosylation of the D3 receptor is essential for clathrin-dependent internalization . In brain, D3 receptors are expressed mainly in limbic regions; they have a low susceptibility to downregulation through internalization, consistent with a role as autoreceptors to monitor DA concentrations [144, 146].
Norepinephrine (NE, adrenergic) receptors
Agonist binding to β1 and β2 adrenergic receptors (β-AR), linked to Gαs proteins and thence to AC, catalyzes the formation of cAMP, with downstream effects such as heart muscle contraction and smooth muscle relaxation. In cardiac myocytes, β1-AR is primarily located in non-raft membrane regions, whereas β2-AR is exclusively localized to caveolae  along with signaling components such as AC and Gαs proteins [148, 149]. However, in cardiomyocytes, agonist binding to β2-AR triggers redistribution of the receptor out of caveolae , causing sequestration of β2-AR receptors away from the effectors and reducing downstream signaling events . A dissenting view is that localization and signaling of β2-AR do not depend on caveolae, but rather are controlled by protein signaling complexes comprised of scaffold proteins interacting with the receptor carboxy-terminus .
Crystallographic studies identify a consensus motif in β2-AR  indicating a cholesterol binding site. The effects of β2-AR on cAMP are altered by cholesterol depletion, with disagreement as to whether AC activity and cAMP production are inhibited [148, 149] or enhanced [150, 154, 155]. A disinhibitory mechanism is proposed for enhanced cAMP production, in which β2-AR-induced caveolar endocytosis and concomitant reduction in receptor number is interrupted by the loss of cholesterol .
Studies with caveolin-3 dominant-negative mutants in rat cardiomyocytes find that caveolin-3 directly regulates β2-AR functioning. Caveolin-3 loss leads to β2-AR-cAMP signaling that mimics heart failure, which is rescued by caveolin-3 overexpression .
Summary of findings
The evidence supports an association of lipid microdomains with SERT and NET, but is less clear with respect to DAT. SERT co-localizes with caveolin and GM1 and is visualized in cholesterol-rich membrane microdomains [16, 90], and its functioning is adversely affected by cholesterol depletion [51, 52]. NET demonstrates cholesterol-dependent trafficking [46, 113] and internalization that is mediated neither by clathrin nor caveolin . In contrast, DAT occurs across raft and non-raft domains , and, although DAT are affected by changes in cholesterol , these effects are not clearly lipid raft-mediated.
There is strong evidence for lipid microdomain regulation of serotonin function for 5-HT1A, 5-HT2A, and 5-HT7 receptors, although leading to very different downstream effects: 5-HT1A couples to Gαi, with inhibitory effects on AC and cAMP; 5-HT7 couples to Gαs, stimulating cAMP levels; and 5-HT2 couples to Gαq, increasing inositol triphosphate and diacylglycerol levels . The 5-HT1A receptors co-localize with caveolin-1 in a palmitoylation-dependent manner , and cholesterol depletion affects binding to 5-HT1A receptors and signaling [118, 119], effects that are reversed by replenishing cholesterol . Most studies of 5-HT2A [34, 122, 123] but not all  find co-localization with caveolins. Findings also are conflicting with regard to caveolin association in agonist-stimulated  vs. unstimulated  conditions; however, in all studies responses to agonist binding to 5-HT2A receptors were affected by cholesterol depletion or caveolin knockdown [34, 122,123,124]. Lipid rafts regulate 5-HT7 binding, the ensuing functional sequelae, and receptor internalization, although it should be noted that all studies have been performed by one group [33, 36, 129]. Lipid microdomain associations with 5-HT3 have been studied primarily in the context of psychotropic medications, many of which co-localize with the 5-HT3 receptor in lipid rafts (see section “Lipid microdomains and antidepressant treatment” below).
The role of lipid microdomains in the regulation of both D1- and D2-class dopaminergic receptors is complex. D1 receptors localize in raft domains [80, 81], and agonist binding to AC protein increases in lipid rafts, but raft localization has only been observed in certain receptor conformations . Both caveolar [81, 135] and clathrin-dependent [134, 135] endocytosis have been reported. The D5 receptor, on the other hand, does not localize in lipid microdomains [80, 138]; its effects on paraoxonase activity are seen only in non-rafts , yet it does mediate effects on AC activity that are attenuated with cholesterol depletion . Also poorly understood are lipid raft effects on D2 receptors, which are diffusely distributed across membrane domains [80, 140, 141], but are unaffected by cholesterol depletion . Endocytosis of D2 receptors can be both caveolar  and dynamin-dependent [143, 144].
The evidence strongly supports a key role of lipid microdomains with regard to β2-AR but not β1-AR adrenergic (norepinephrine) receptors: β2-AR are localized to caveolar lipid microdomains  along with relevant signaling components [148, 149], and are dependent on cholesterol [150, 153] and caveolin-3 , with downstream effects on AC and cAMP [148,149,150, 154, 155].
One caveat in interpreting the possible importance of these findings for depression is that studies of lipid microdomains with respect to 5-HT7 and D3 receptors were carried out only in cells derived from kidney, and likewise β-AR were studied exclusively in cardiomyocytes. Therefore, assuming that lipid microdomain effects on 5-HT7, D3, and β-AR have ramifications for brain function and depression, is premature.
Mechanisms of lipid microdomain effects in monoaminergic receptor signaling
Early studies focused on establishing co-localization of receptor and transporter molecules in lipid microdomains. More recent research has expanded the field of study to include related elements of signaling complexes and to begin exploration of structure–function relationships related to lipid microdomains. For instance, a salient characteristic of most monoaminergic receptors is coupling to G-proteins at the cytoplasmic membrane leaflet, and it turns out that lipid microdomains have regulatory effects on G-proteins, impacting localization, trafficking, and signaling. Among the G-protein isoforms associated with monoaminergic receptors, studies in rat lung homogenate find that Gαs (with which 5-HT7 and β2-AR associate) and Gαi (5-HT1A) tend to localize in lipid microdomains, while Gαq (5-HT2A) more than Gαs or Gαi, are enriched in caveolae . As an example of lipid microdomain-mediated trafficking of G-proteins related to monoaminergic signaling, in C6 glioma and MCF-7 adenoma cells that express only endogenous β2-AR, Gαs undergoes internalization within vesicles in response to agonist binding , a process blocked by caveolin knockdown or cholesterol depletion . Furthermore, lipid microdomains can have effects downstream from agonist-receptor binding, e.g., in C6 cells, caveolin knockout or knockdown interferes directly with Gαs stimulation of adenylyl cyclase .
A growing body of research shows that monoaminergic receptors in the membrane not only interact with G-proteins, but also form complexes with a variety of other molecules in association with lipid microdomains, including pre-synaptic receptors , membrane proteins involved in exocytosis [45, 160, 161], and hundreds of adaptor proteins that modulate maturation, internalization, targeting for degradation, and recycling of monoamine receptors .
Indirect effects on monoaminergic function via lipid microdomains also have been observed. For example, the raft-associated cellular prion protein (PrPC) inhibits GSK3β kinase in serotonergic cells in the raphe region, with the downstream effects of suppressing 5-HT1B receptor activity . As 5-HT1B receptors are inhibitory, the net effect is an increase in serotonergic activity.
Lipid microdomains and antidepressant treatment
Although selective serotonin re-uptake inhibitors (SSRI) act acutely on SERT in pre-synaptic neurons, the therapeutic effects of these antidepressants occur after chronic exposure and are believed to occur via downstream effects on post-synaptic 5-HT receptors . Consistent with this hypothesis, accumulation of the active S- but not the inactive R- enantiomer of citalopram triggers Gαs translocation from raft to non-raft regions in C6 glioma cells that do not express SERT, and is not enhanced in cells that contain transfected SERT .
Studies of postmortem brain tissue from suicide decedents with confirmed unipolar major depression find increased accumulation of Gαs in lipid microdomains . Treatment with several chemically distinct antidepressants facilitates translocation of Gαs, but not other G-proteins, from lipid microdomains to non-raft fractions and into closer association with AC, potentiating its activity [165, 167,168,169,170]. This translocation might contribute to the increased cAMP tone and synaptic changes observed subsequent to chronic antidepressant treatment . These studies are also in agreement with observations that rolipram, a cAMP-phosphodiesterase inhibitor, has antidepressant effects . On the other hand, mood stabilizers lithium and valproate have the opposite effect, causing increased Gαs in lipid microdomains , thus suggesting a biphasic effect at the molecular level for treatment of the two “poles” of bipolar illness. Among monoaminergic receptors, D1 and D3, 5-HT7, and β1-AR and β2-AR are known to couple with Gαs and cause increased cAMP, although only D1, 5-HT7, and β2-AR co-localize predominantly in lipid raft fractions. Possible connections between these elements and antidepressant mechanisms could be a promising area of future investigation.
In cell culture, the 5-HT3 receptor protein localizes exclusively to fractions containing lipid microdomains, which are also the same fractions robustly enriched in antidepressants (desipramine, fluoxetine, and reboxetine) and antipsychotics (fluphenazine, haloperidol, and clozapine) . Concentrations of these psychotropic medications correlate with the magnitude of non-competitive inhibition of 5-HT3 serotonin-induced cation currents . However, the ability of lipid raft co-localizers desipramine and fluoxetine to counter 5-HT3 effects in N1E-115 cells was not affected by cholesterol depletion . Thus, although lipid rafts have an impact on 5-HT3 function, antidepressant antagonism of 5-HT3 appears to be independent of lipid microdomain status [126,127,128].
In addition to studies in postmortem brain and in vitro cell cultures, ex vivo studies of human tissue such as platelets  and B-cells  are now feasible approaches to the study of lipid microdomains in humans. Outcome measures of these approaches are physicochemical microdomain properties such as membrane order  and microdomain clustering  that could be tested alone or in combination with monoaminergic parameters as potential biomarkers for depression or as predictors of antidepressant response.
Lipid microdomains as a target for polyunsaturated fatty acid (PUFA) effects in depression
Lipid raft molecular organization and raft-associated protein distribution are highly susceptible to modulation by long chain n-3 PUFAs, dietarily essential fatty acids that are generally low in Western diets , and very low in major depressive disorder  and in patients at high risk for suicide [177,178,179]. There is strong evidence that n-3 PUFAs, upon incorporation into the plasma membrane as acyl chains of glycerophospholipids, can either prevent or promote lipid raft formation, depending on several factors: cell type, the relative concentrations of eicosapentaenoic acid (EPA, 20:5,n-3), docosahexaenoic acid (DHA, 22:6,n-3), and docosapentaenoic acid (DPA, 22:5,n-3), and the phosphatidyl moiety to which the n-3 PUFA is bound [180,181,182,183].
It has been hypothesized that n-3 PUFA effects on depression [184,185,186] are mediated by their ability to modify lipid rafts, with downstream effects on G-protein signaling . If correct, monoaminergic transporters and receptors are possible candidates as mechanistic intermediaries. Membrane n-3 PUFAs affect oligomerization kinetics of adenosine A2A and D2 receptors via effects on biophysical membrane properties , and in rodent models in a cardiovascular context, α-linolenic acid (ALA, 18:3,n-3) enhances caveolin-3 expression  and prevents cardiac damage due to β-adrenergic overstimulation, an effect blocked by preadministration with methyl-β-cyclodextrin . Other possible lipid microdomain-associated targets for PUFA effects in depression relate to the anti-inflammatory and pro-resolving properties of n-3 PUFA, as inflammation likely contributes to depression etiology in a subset of patients [191, 192]. For example, in B-cell cultures, DHA inhibits Toll-like receptor (TLR) dimerization and recruitment into lipid microdomains .
Antihyperlipidemic drugs and suicide risk
Lower cholesterol is linked to suicide risk [194,195,196], and in nonhuman primates, low cholesterol is associated with lower CSF levels of the 5-HT metabolite 5-HIAA, blunted prolactin response to fenfluramine-stimulated 5-HT release, and increases in aggressive behaviors .
The aggregate evidence presented herein supports a hypothesis that effects of cholesterol-lowering therapies on suicide risk are mediated by influencing lipid raft regulation of plasma membrane serotonergic proteins, since cholesterol depletion decreases the transport rate (Vmax) and affinity (Km) of 5-HT for SERT [51, 52], agonist and antagonist binding to 5-HT1A and 5-HT7 [33, 118, 119], and physiologic responses of 5-HT2A to 5-HT stimulation [122, 124].
Our literature review suggests that lipid microdomains are a significant modulator of monoaminergic neurotransmission, influencing co-localization of transporters/receptors with other signaling proteins and adaptor molecules; trafficking by endocytosis; agonist and antagonist binding; and downstream second messenger-mediated events. Limitations on knowledge reflect the limitations of current experimental methodologies for nanoscale discrimination of spatial and temporal aspects of neuronal signaling. In addition, there is a paucity of clinical studies, which are needed to advance understanding of lipid microdomain effects on monoaminergic signaling with respect to etiology, diagnosis, and treatment of major depression. Thus, the next wave of research should seek to move toward the study of monoaminergic–lipid microdomain interactions in humans with and without depression. One approach would be to measure physicochemical effects on microdomain structure, using ex vivo techniques, of substances already known to be safe for human use that affect lipid microdomains, such as cholesterol-lowering drugs and n-3 PUFAs.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This study was supported in part by the following grants: K08 MH079033 (PI, MES), NIH R01AT008375 (PI, SRS), NIH 8G12 MD007603 (Area Leader, RES). Dr. JJM receives royalties for commercial use of the Columbia-Suicide Severity Rating Scale (C-SSRS) from the Research Foundation for Mental Hygiene.