Transport of endocannabinoids across the plasma membrane and within the cell

Abstract

Endocannabinoids are readily accumulated from the extracellular space by cells. Although their uptake properties have the appearance of a process of facilitated diffusion, it is by no means clear as to whether there is a plasma membrane transporter dedicated to this task. Intracellular carrier proteins that shuttle the endocannabinoid anandamide from the plasma membrane to its intracellular targets such as the metabolic enzyme, fatty acid amide hydrolase, have been identified. These include proteins with other primary functions, such as fatty-acid-binding proteins and heat shock protein 70, and possibly a fatty acid amide hydrolase-like anandamide transporter protein. Thus, anandamide uptake can be adequately described as a diffusion process across the plasma membrane followed by intracellular carrier-mediated transport to effector molecules, catabolic enzymes and sequestration sites, although it is recognized that different cells are likely to utilize different mechanisms of endocannabinoid transport depending upon the utility of the endocannabinoid for the cell in question.

Abbreviations

2-AG

2-arachidonoylglycerol

AEA

anandamide

AM404

N-(4-hydroxyphenyl)-arachidonamide

AMT

anandamide membrane transporter

CB

cannabinoid

EMT

endocannabinoid membrane transporter

FAAH

fatty acid amide hydrolase

FABP

Fatty-acid-binding protein

FLAT

FAAH-like anandamide transporter protein

TRPV1

transient receptor potential vanilloid

Introduction

The endogenous arachidonic acid derivatives anandamide (AEA) and 2-arachidonoylglycerol (2-AG) have been the subject of great attention because of their ability to activate cell-surface cannabinoid (CB) receptors [1]. However, these endocannabinoids have additional targets, including intracellular CB receptors [2, 3] and transient receptor potential vanilloid (TRPV1) receptors [4]. In the brain, an important role of endocannabinoids is that of retrograde signalling, whereby they are synthesized and released from postsynaptic locations to stimulate presynaptic CB₁ receptors [5]. To achieve these functions, there must be mechanism(s) by which endocannabinoids can be released from cells, and conversely, extracellular endocannabinoids can be taken up into cells. Intracellular endocannabinoids must also transported to different locations in the cells for both signalling purposes and their catabolism. In this minireview, current knowledge of the ways in which endocannabinoids are transported both across cell plasma membranes and within cells, and how this process can be regulated, is presented. It is the author's intention to be provocative, so hold on to your hats!

Cellular accumulation of endocannabinoids – early studies

The first report [6] of the cellular uptake of AEA came but 1 year after its discovery, where a time-dependent decrease in the medium was noted when C6 glioma or N18TG2 neuroblastoma cells were incubated with trace amounts of this endocannabinoid. In a more detailed study the following year [7], the ‘association’ (i.e. incorporation into the plasma membrane and subsequent intracellular accumulation) of AEA with cultured cortical neurons was reported to be saturable, time- and temperature dependent and to be lost in the presence of BSA. (Throughout this minireview, the word ‘association’ will be used, since the methodology used in many uptake studies does not distinguish between the ligand remaining in the plasma membrane and that actually internalized after the incubation period used.) AEA binds very avidly to albumin [8], and so this loss of activity essentially reflects a reduction in the free, i.e. available for uptake, concentration of AEA, rather than an inhibition of an uptake process per se. AEA association was neither ATP- nor sodium dependent, suggesting a process of facilitated diffusion [9]. In addition, the process appeared to work in both directions, because a time- and temperature-dependent efflux of AEA following preloading of cerebellar granule cells with this ligand was found [9]. A time-dependent clearance of 2-AG by astrocytoma cells was reported in 1999 [10], and although there was no evidence of a loss of the AEA homologue palmitoylethanolamide from the medium when incubated with cultured neurons [7], an apparently saturable and time-dependent association of this ligand with RBL2H3 basophilic leukaemia cells and Neuro-2a neuroblastoma cells was subsequently reported [11].

In addition to the biochemical studies summarized above, pharmacological studies indicated that the uptake process could be inhibited by substrate analogues, such as N-(4-hydroxyphenyl)-arachidonamide (AM404) [12]. This compound, which could itself be cleared from the medium when incubated with astrocytoma cells [10], was more potent than its N-3-hydroxyphenyl positional isomer as an inhibitor of the cellular association of AEA [12], suggesting some form of structural requirement for inhibition. Further studies have identified several compounds capable of inhibiting the cellular association of endocannabinoids, of which the five most commonly used are the arachidonoyl derivatives AM404 [12], VDM11 [13] and UCM707 [14], and the oleoyl derivative OMDM-1 and its enantiomer OMDM-2 [15]. These compounds are active in vivo (an example is the ability of AM404 to potentiate the antinociceptive effect of AEA [12]) and can block release as well as uptake [16, 17] (an effect consistent with in vivo data showing CB₁ antagonist-like actions of VDM11 [18]). The compounds do not produce obvious cannabis-like signs in drug discrimination tests [19], raising the possibility that endocannabinoid uptake is a drug-able target for disease states where a potentiation of extracellularly mediated endocannabinoid signalling is beneficial [20]. AEA uptake is also inhibited by other arachidonoyl compounds such as 2-AG, arachidonic acid, noladin ether and virodhamine, but not by other N-acylethylanolamines such as palmitoylethanolamide and stearoylethanolamide [11, 17, 21]; whether this reflects substrate competition or inhibition is not yet clear.

‘Antigonish’ or the case of the elusive anandamide membrane transporter

The observations summarized above are entirely consistent with the suggestion that endocannabinoids are accumulated via a transporter. Although early work [7] did not specify where in the cell such a carrier is expressed, it has evolved into a distinct entity, the ‘anandamide membrane transporter’ (AMT); or more generally the ‘endocannabinoid membrane transporter’ (EMT) with a localization upon the plasma membrane (see e.g. Refs [1, 22] for an example in a review article and a pharmacology textbook, respectively). An AMT/EMT protein has not, to my knowledge, been cloned, and so it is important to consider whether current data are also consistent with a simpler model without the need to invoke the existence of such a transporter.

What are the requirements of a simpler model? In essence, the simpler model requires the ability of the endocannabinoid to associate with and translocate the plasma membrane without binding to a carrier, and thereafter to relocate within the cell to either its intracellular receptors or its catabolic enzymes. A key factor here is that if the initial translocation is passive in nature, there must be processes within the cell that reduce the free intracellular concentration of the endocannabinoid, otherwise a equilibrium will be reached. In addition, the whole process, i.e. passive diffusion + subsequent intracellular events, should display the properties reported in the early studies, such as saturability and sensitivity to inhibition by the uptake inhibitors. These criteria are discussed below.

Saturability of uptake and sensitivity to inhibition by uptake inhibitors is not an early event

The transport of AEA through resealed human red blood cell membranes is massively fast (90% of equilibrium between compartments is found within 5 s using an extracellular BSA concentration of 30 μm [23]). In RBL2H3 cells incubated with 100 nm [¹⁴C]AEA, considerable metabolism of the ligand had taken place within 25 s [24]. These data suggest that at very short incubation times, enough AEA has entered the cell so that the properties of a plasma membrane transporter should be apparent. However, and in contrast to what is seen at longer incubation times (generally, 4–10-min incubation times are used, although saturability and sensitivity to inhibition has been reported by 90 s for RBL2H3 cells [25]), neither sensitivity of the AEA association to inhibition by uptake inhibitors nor saturability with respect to the added AEA concentration is seen [24, 26, 27]. An example of the latter in P19 embryonic carcinoma cells is shown in Fig. 1. The temperature dependency of the association of AEA with cells is more difficult to assess, because the free concentration of AEA in the medium, i.e. that available for uptake, is itself temperature dependent. However, when this is compensated for, no temperature dependency of the AEA association with ND7/23 mouse neuroblastoma × rat dorsal root ganglion neurone hybrid cells using an incubation time of 5 min, or for RBL2H3 cells using an incubation time of ~ 1 min, was seen [28]. The association of [³H]AEA with rabbit platelets following an incubation time of 1 min shows no saturability, and a slightly lower association at 37 °C than at 0–4 °C [29]. The temperature sensitivity of the association of AEA with P19 cells is much greater using a 15 min incubation time than a 4 min incubation time [28]. A recent study with U937 cells, demonstrated similar sensitivities of AEA association to inhibition by the uptake inhibitors UCM707 and OMDM-2 over a wide range of incubation times, the lowest tested being 30 s [17]. However, even at this short time, considerable hydrolysis of the AEA was seen, and so these findings do not unequivocally locate the action of the compounds to the cell surface. Thus, the bulk of available data would argue that at very short incubation times, but where a sizable proportion of AEA can cross into the cells and be metabolized [24], AEA association does not behave like a facilitated transporter-mediated process. That these properties are clearly visible at longer incubation times suggests that they reflect intracellular events (such as ligand redistribution and/or metabolism, about which more anon) rather than processes occurring at the plasma membrane.

Figure 1. Example of the cellular association of [³H]AEA at different incubation times. The main graph shows the cell-associated [³H]AEA at different added concentrations of [³H]AEA. Values for the tritium retained by wells alone are also shown. The data for the wells at all three time points and for the cells at a 45 s incubation were better fitted by a line passing through the origin than by a Michaelis–Menten curve, whereas the converse was seen for cells at 4 and 15 min incubation. The insert shows a replot of the slopes and intercept (at t = 0) from linear regression analysis at each AEA concentration. The figure is redrawn from the original data of Sandberg & Fowler [27]

AEA can translocate artificial membranes; role of cholesterol

In cell membranes fractionated from cells expressing caveolin-1 (F11 dorsal root ganglion × neuroblastoma cells) and from those lacking this protein (BV-2 microglial cells), 2-AG is associated primarily with the lipid raft fractions, whereas AEA is found in both lipid raft and non-lipid raft fractions [30, 31]. N-Acylethanolamine compounds can form complexes with cholesterol [32], and the treatment of C6 glioma cells with cholesterol increases the observed rate of AEA association [33]. Conversely, depletion of cholesterol reduces the AEA association with cells [34, 35] (but see Ref. [36]). These data would suggest that membrane cholesterol is an important initial determinant of AEA uptake. Consistent with this, studies in artificial lipid monolayers and bilayers have demonstrated that the interaction of AEA with the monolayers is highly cholesterol dependent. Thus, in the absence of cholesterol, there is a very rapid interaction of AEA with the monolayers (measured as the increase in the surface pressure of the membranes), but this peaks after a few seconds and then declines. Further, very little translocation of AEA across artificial lipid bilayers is seen in the absence of cholesterol [37]. In the presence of cholesterol, however, the interaction of AEA with the monolayers is slower but more robust, and translocation across the bilayers is seen [37]. Although the translocation is relatively slow (time points used were in hours, rather than minutes), these data do demonstrate that this lipophilic endocannabinoid can cross membranes lacking proteins. More recently, Kaczocha et al. [38] investigated the hydrolysis of AEA by synthetic large unilamellar vesicles containing FAAH trapped on the inside. The authors found considerable hydrolysis of externally applied AEA within 5 min, indicating a more rapid rate of transport across the synthetic vesicles than was seen across the artificial bilayers in Di Pasquale et al. [37]. Addition of a high concentration (40%) cholesterol to the large unilamellar vesicles increased the rate of hydrolysis to about double, as did the addition of coprostanol, which does not form lipid rafts. By contrast, addition of the charged sterol cholesterol sulphate did not increase the observed rate of hydrolysis [38].

Adiposomes sequester AEA

In cerebellar granule cells, the steady-state levels of extra- and cell-associated [³H]AEA and [¹⁴C]urea were compared following incubation with these markers. For the freely diffusible urea molecule, equilibrium was reached when ~ 0.5% of the added urea was cell-associated. By contrast, the corresponding number for added AEA was ~ 30% [39]. A separate study using RBL2H3 cells estimated the ratio of cell-associated to extracellular AEA concentrations of > 20 : 1 [40]. Hillard & Jarrahian [39] concluded that their calculations ‘suggest that AEA within the cell is either metabolized which maintains the concentration gradient or it is removed to a compartment that is not free to equilibrate with the extracellular AEA pool.' [39]. Subsequent experiments using biotin-labelled AEA and cells engineered to express the AEA hydrolytic enzymes FAAH-1 and -2 in different subcellular localizations demonstrated that lipid droplets (adiposomes) are capable of sequestering and metabolizing AEA [41, 42].

AEA-binding proteins and FAAH regulate AEA uptake

There has been some controversy as to the importance of FAAH in the regulation of AEA uptake. In theory, as pointed out in Hillard & Jarrahian [39], FAAH, by reducing the intracellular AEA concentration, will act to preserve the inward gradient and hence uptake of AEA. Inhibitors of FAAH should thus reduce the observed rate of cell association of AEA. Although this is true for some cells, such as RBL2H3 basophilic leukaemia cells [24, 40], it is not the case for others (such as cortical astrocytes [12]) (see Fig. 2 for examples of cell lines with different sensitivities to FAAH inhibitors). AEA association has also been reported in uptake experiments using brain synaptosomes and cortical neuronal cultures from FAAH^−/− mice [25, 45-47], and so it is clear that although FAAH can act as a regulator of AEA uptake, it is by no means the only regulator found in cells.

Figure 2. The influence of inhibitors of endocannabinoid hydrolysis upon the association of AEA and 2-AG with different cells. (A) Cells (RBL2H3 rat basophilic leukaemia cells, PC3 human prostate cancer and R3327 AT-1 rat prostate cancer cells) were preincubated for 10 min at 37 °C with either vehicle or the selective FAAH inhibitor URB597 prior to addition of [³H]AEA (added concentration 100 nm) followed by incubation for a further 10 min. (B) RBL2H3 cells were preincubated for 10 min with the nonselective endocannabinoid hydrolysis inhibitor methyl arachidonoyl fluorophosphonate (MAFP) prior to addition of either [³H]2-AG or [³H]AEA (added concentration 100 nm) and incubation for a further 10 min. *P < 0.05, **P < 0.01, otherwise not significant from vehicle control [paired t-test (A), Dunnett's test following significant one-way analysis of variance (ANOVA) for repeated measures for (B)]. The figures are drawn from the original data of Thors et al. [43] and Fowler & Ghafouri [44].

An additional way of reducing the free intracellular AEA concentration is for the endocannabinoid to bind to intracellular proteins. Fatty-acid-binding proteins (FABP)-5 and -7, albumin, heat shock protein 70 and the FAAH-like anandamide transporter protein (FLAT) have been identified as AEA-binding proteins and shown to influence the cellular association of AEA [47-50], and a novel inhibitor of FABP5 reduces carrageenan-induced thermal hyperalgesia in mice in a manner blocked by a combination of CB₁ and CB₂ receptor antagonist/inverse agonists [51]. The compound also is active in the formalin model of persistent pain, but in this case it only produces significant effects on the first phase of the nociceptive response to formalin at the dose tested [51], a result also seen with AM404 [52], whereas UCM707 blocks both phases [52].

Just as the importance of FAAH in regulating the uptake of AEA is dependent upon the cell type used, it is likely that the relative contributions of the binding proteins to the uptake of AEA will vary from cell to cell. One nice example of this is in U937 cells, where differentiation into macrophages increases expression of FABPs: in the differentiated cells, AEA association is an order of magniture more sensitive to inhibition by the FABP inhibitor BMS309403 than in the undifferentiated cells [17]. Otherwise, little is known about this aspect as yet, although preliminary data using the FLAT inhibitor ARN272 [47] to detect FLAT-dependent cellular AEA association has indicated that whilst inhibition of this protein produces a large reduction in the association of AEA with cultured rat cortical neurons [47], a more modest reduction is seen in R3327 AT-1 prostate cancer and C6 glioma cells, and no reduction was found for RBL2H3 cells [53]. Further work in this area is needed, not least in view of recent data questioning the role of FLAT as an intracellular carrier of AEA [54].

A separate issue is whether the binding proteins act as true carriers, i.e. that they shuttle the AEA from the inside of the plasma membrane to either intracellularly located binding sites for signalling proteins or metabolic enzymes. This question cannot be answered by measuring rates of AEA cell association after transfection and/or inhibition of the proteins involved (in theory, the proteins could bind the AEA and thus affect its free extra : intracellular concentration; but then pass on the endocannabinoid to an unidentified carrier molecule), but some data are emerging. FLAT has structural properties suited to this purpose in that it is associated with membranes but can detach to move to the cytosol [47] (but see Ref. [54]). The ability of the FLAT inhibitor ARN272 to increase plasma AEA concentrations and to produce CB₁ receptor-dependent antinociceptive effects in mice in vivo [47] are consistent with (but not definitive proof of) the contention that this is a true carrier. The AEA homologue oleoylethanolamide is a powerful activator of the nuclear receptor peroxisome proliferator-activated receptor α [55], and in HeLa cells transfected with the ligand binding domain of proliferator-activated receptor α fused to a reporter group, the ability of oleoylethanolamide to elicit a signal is decreased following knockdown of FABP5 [50]. Additionally, overexpression of a cytoplasmically restricted FABP5, which would be expected to compete with the native FABP5 for the binding of oleoylethanolamide, but which would prevent its delivery to the nucleus if the molecule had significant shuttling properties, reduced the activation of peroxisome proliferator-activated receptor α by this N-acylethanolamine [50].

In contrast to AEA, little is known about the intracellular components involved in 2-AG uptake. In J744 and U937 macrophage cell lines, 2-AG is accumulated in a time-dependent manner and incorporated into phospholipids, although the degree of incorporation into phospholipids differs between the cells [17, 56]. The association of 2-AG with a variety of cells is not reduced by inhibition of the 2-AG hydrolytic enzymes [44, 57] (see Fig. 2B; indeed, an increased cellular association has been seen in U937 cells [17]), or to overexpression of FLAT in HEK293 cells [47]. The association of 2-AG with human CCF-STTG1 astrocytoma cells is, however, saturable (with an incubation time of 4 min at 37 °C) and can be inhibited both by AM404 and the acyl-CoA synthetase inhibitor triacsin C [57]. In U937 cells, the association is blocked by UCM707 and OMDM-2, and by AEA, noladin ether and virodhamine, but not by N-acylethanolamines other than AEA [17]. Although AM404 reduces levels of 2-AG, arachidonic acid and phospholipids (but not di- and triacylglycerols) associated with the CCF-STTG1 cells following the incubation with 30 nm [³H]2-AG, only the phospholipid levels are affected by triacsin C [57], suggesting that AM404 acts proximally to triacsin C in these cells. It would be useful to know whether similar results are found following knockdown of FABP5.

AEA-binding proteins and FAAH are targets for uptake inhibitors

The ‘elephant in the room’ that has been ignored so far is the fact that the association of AEA with a wide variety of cells is sensitive to compounds like AM404, VDM11, UCM707 and OMDM-1/2. It has been argued above that these compounds act intracellularly, because they do not affect the association of AEA at short incubation times [24, 26] (but see Ref. [17]). In this respect, endocannabinoid release, as assessed in electrophysiological experiments measuring striatal long-term depression, is blocked by the intra- but not the extracellular application of VDM11 [16]. An action of uptake inhibitors upon FAAH- and AEA-binding proteins would, however, provide an explanation for their ability to affect the association of AEA with cells. It is now well established that AEA uptake inhibitors can inhibit FAAH, and in RBL2H3 cells, this is an important component of their action [24]. However, it would be a gross oversimplification to suggest that the compounds are simply FAAH inhibitors, because they are capable of inhibiting the association of AEA with cells that are not sensitive to FAAH inhibitors, or which lack FAAH [12, 45, 46] (see also Ref. [17]). Indeed, there is no significant correlation between the FAAH inhibitory and uptake inhibitory potencies of a series of AEA analogues (Fig. 3), which would have been expected had FAAH been the sole target of the uptake inhibitors. However, the binding of [³H]AEA to FLAT is inhibited by AM404 and OMDM-1 [47], and AM404, VDM11, OMDM-1 and OMDM-2 inhibit the binding of 12-N-methyl-(7-nitrobenz-2-oxa-1,3-diazo)aminostearate to purified FABP5 [50]. With respect to the latter, the cellular association of AEA with FAAH-transfected HeLa cells is reduced by these compounds, whereas they are inactive when the content of FABP5 in these cells is decreased by introduction of shRNA for this protein [50]. By contrast, Chicca et al. [17] found rather similar potencies of the uptake inhibitors UCM707 and OMDM-2 in undifferentiated and differentiated U937 cells, which express different levels of FABPs. The authors argued that this finding indicated that ‘these compounds exert their activity independently from the expression of FABPs’ [17]. However, if the compounds inhibited the binding of AEA to FABPs in the intact cells with the same potency as to its other (and possibly myriad) intracellular targets, the same result would have been found. Thus, there are enough intracellular targets for the uptake inhibitors to explain their ubiquitous effects on the cellular accumulation of AEA.

Figure 3. Comparison between the potencies of 19 head group analogues of AEA towards inhibition of rat forebrain FAAH activity and of the cellular association of AEA with cerebellar granule cells. Spearman's rho for the data (expressed as log values) was 0.29 (P > 0.2). The graph is drawn from the original data given in Tables 1 and2 of Jarrahian et al. [21].

In a different approach, Ligresti et al. [35] used the TRPV1 receptor as a biosensor for AEA internalization, because AEA acts on the cytosolic face of this receptor. In HEK 293 cells transfected with this reporter system, the observed response to addition of AEA to the medium was clearly reduced by bovine serum albumin, by OMDM-1, by depletion of membrane cholesterol, by addition of a high concentration of a competing fatty acid or by an inhibitor of FABP4. The response to the TRPV1 ligand capsaicin, however, was not particularly sensitive to these treatments. The effects of the compounds upon the TRPV1 signal were significantly reduced when the AEA was delivered to the cells in poly(ε-caprolactone) nanoparticles [35]. The authors argued that this was evidence of a plasma membrane transporter, because the inhibition of uptake produced by the compounds tested were lost when the need of AEA to cross the plasma membrane was bypassed by the use of the nanoparticles. It is also the case, however, that the nanoparticles would shield the AEA from binding to intracellular carrier molecules such as FABP5 and FLAT before delivering the ligand to the biosensor, and so their data, elegant though it is, does not provide hard proof of a plasma membrane transporter.

From the above data, compelling evidence for an AMT/EMT is lacking, and this author feels that the use of such a term is both premature and over-used (hence the reference to the poem ‘Antigonish’ by William Hughes Mearns on the second page of this Minireview. The first verse is alluded to here; for those readers unfamiliar with this ditty, http://en.wikipedia.org/wiki/William_Hughes_Mearns will do it for you. The reference may seem a bit brutal in the present context, but there are many articles in the literature describing either ‘anandamide transporter inhibitors’ or changes in ‘anandamide transporter activity’ in the literature, i.e. a defined entity, when what is used/measured are inhibitors of anandamide accumulation or changes in the ability of cells to accumulate this endocannabinoid.) My recommendation would be to use as default the simpler model whereby endocannabinoids can cross plasma membranes as a result of their lipophilicity, and whereby their accumulation is controlled intracellularly by carrier proteins, by intracellular sequestration and/or by metabolism. The ‘simpler’ model is admittedly more complex in as much as it requires a bunch of different intracellular components, but is it simpler in the sense that it uses identified cellular proteins and components. In fairness to aficionados of AMT/EMT, it should be pointed out that solid evidence that a plasma membrane transporter does not exist is also lacking, and its future identification would, of course, have massive implications. However, proving something to be absent may not be possible.

Multiple modes of endocannabinoid uptake

The discussion above has been deliberately confrontational, given that it has considered the AMT/EMT and passive diffusion model as mutual exclusives. The passive diffusion model does not rule out concomitant endocannabinoid uptake mediated by other pathways, such as an AMT/EMT pathway (or vice versa, depending upon your point of view). Additional processes may also occur: Barker and colleagues [58, 59] have argued that endocytosis is an important route of endocannabinoid uptake, although some of the compounds used by these authors in support of their hypothesis suffer from a lack of selectivity [60]. As argued by Hillard & Jarrahian [61], cells may utilize different mechanisms depending upon whether they employ endocannabinoids as signalling molecules or as a source of arachidonic acid. Indeed, in cancer cells, N-acylglycerols (including 2-AG) are an important source of free fatty acids [62]. Thus, an all-encompassing model of endocannabinoid uptake is most probably an oversimplification, tempting though it may be.

Regulation of endocannabinoid uptake and release

At first sight, the complexity of endocannabinoid transport and the number of proposed different intracellular carriers would suggest that regulation of this process is likely to be difficult, because increased expression of a given component of the uptake would have little effect upon the total uptake capacity of the cell in question, unless it was the rate-limiting step in a dominant pathway. FAAH has a number of potential regulatory sites in its promotor region [63-65] and its expression in cells is affected by a variety of endogenous signalling molecules such as cytokines, leptin, oestrogen and progesterone [64-67] and by treatment of cells with lipopolysaccharide [68]. In theory, such changes in expression would be expected to affect uptake in cells where FAAH plays a dominant role in controlling this process. Human lymphocytes do not appear to fall into this category, because neither progesterone, leptin, cytokines nor lipopolysaccharide uptake affect the cellular association of AEA [64-66, 68]. Oestrogen treatment of human umbilical vein endothelial cells dramatically reduces the FAAH activity and yet increases the rate of cellular association of AEA, both effects being blocked by the oestrogen receptor antagonist tamoxifen [67]. The oestrogen treatment also increases the activity of nitric oxide synthase [67], which may be the primary regulator here, given that nitric oxide donors affect the way in which cells accumulate and metabolize AEA [68, 69], in our hands by mechanism(s) downstream of FAAH [70]. Thus, although it is early days, current data would suggest that despite its complexity in terms of multiple carriers, the cellular accumulation of AEA can be regulated.

Conclusions

We have come a long way to understanding the process(es) whereby endocannabinoids are accumulated since the first report [6] of the cellular uptake of AEA. Controversies remain, and it is wise to keep an open mind with respect to the different potential ways in which endocannabinoids can translocate plasma membranes and be shuttled to their intracellular targets/catabolic enzymes/sites of sequestration. AEA uptake has dominated the field, and there is a need for more studies investigating both the uptake of 2-AG and the release of both endocannabinoids. Thus, for example, we need to know which intracellular molecules can bind 2-AG and thereby act as potential carriers. We need to determine whether endocannabinoid release can adequately be characterized as uptake in reverse, or whether it involves separate pathways that can be regulated. Evidence so far (such as the ability of uptake blockers to prevent release [16, 17, 25] and the finding that AEA release is increased following FLAT transfection [47]) is consistent with, but not proof of, the notion of uptake and release utilizing the same pathways in opposite directions. Finally, further work is needed to understand how endocannabinoid uptake can be regulated, and whether such regulation has consequences for endocannabinoid signalling in vivo in different physiological or pathological situations. If ‘antigonish’ was a brutal way of describing the focus on AMT/EMT, then the opposite, a need for more focus, on these issues could presumably be called ‘agonish’!

Acknowledgements

This article is dedicated to Dr Dale Deutsch, whose original study [6] on AEA uptake and hydrolysis I discovered fortuitously while leafing through the 1 September 1993 edition of Biochemical Pharmacology and which caused me to enter into the endocannabinoid field. The author would also like to thank the Swedish Science Research council (Grant no. 12158), the Swedish Cancer Society (Grant no. CAN2010/437), Lion's Cancer Research Foundation, Umeå University, and the Research Funds of the Medical Faculty, Umeå University for research support.