In common usage, the word "hallucinogen" has become a catchall term for a wide variety of psychoactive substances, including such pharmacologically different chemicals as cannabinoids; NMDA receptor antagonists such as ketamine; 3,4-methylenedioxymethamphetamine; kappa-opioid agonists such as salvinorin A, and lysergic acid diethylamide (LSD). The term hallucinogen is often used in medical and research journals, but is rarely given a precise definition. When used in its most general sense, it is vague, adds confusion to complex neuropharmacological issues, and almost becomes meaningless.

In this article, the use of the term "hallucinogen" will be in reference to two groups: chemicals that are chemically and pharmacologically similar to mescaline [the phenethylamine hallucinogens (Shulgin & Shulgin, 1991)] and the chemicals that are chemically and pharmacologically similar to psilocin and LSD [the indole or tryptamine hallucinogens (Shulgin & Shulgin, 1997)]. These two groups of chemicals, which at their extremes seem to share few chemical similarities (Fig. 1A cf. Fig. 1F), have shared pharmacological targets and produce similar behavioural effects (discussed below). Hence it is sensible to group them together.

Hallucinogens produce effects in the mind that can be so intense, that they have led people to suggest other names for this group, such as 'entheogens' (Ruck et al., 1979) and 'psychotomimetics' (Clark, 1956)1. Entheogen is derived from the Greek, en-theos, which means "god/spirit within" and is in reference to the spiritual and deep effect hallucinogens can have on the mind, while psychotomimetic means "psychosis mimicking", and refers to the similarities between hallucinogen-induced states and schizophrenia and other mental illnesses. These two divergent terms illustrate why the study of hallucinogens is important: an understanding of how hallucinogens affect thought could help explain not only how mental illness affects the mind, but also the nature of consciousness itself.

Once serotonin's (5-HT) presence was demonstrated in the brain in 1953 (Twarog & Page, 1953) it was not long until the chemical similarity between LSD and 5-HT was noted. In the same year Gaddum (1953) demonstrated that LSD antagonized the action of 5-HT in peripheral tissues. Soon afterwards two groups independently proposed the hypothesis that the hallucinogenic action of LSD was due to its ability to block central 5-HT receptors (Gaddum & Hameed, 1954; Woolley & Shaw, 1954). This idea held sway, even after it was shown that the brominated LSD analogue BOL, which was a potent 5-HT antagonist, was devoid of hallucinogenic action (Cerketti & Rothlin, 1955; Woolley & Shaw, 1954), that BOL could block the hallucinogenic effects of LSD, and that several other LSD analogues, which were weaker 5-HT antagonists, were more potent in regards to their hallucinogenic action (Gogerty & Dille, 1957; Votava et al., 1958). Anden et al. (1968) were the first to suggest that the primary action of hallucinogens was as a 5-HT receptor agonist, and with the body of evidence already mentioned, this idea was rapidly accepted by most. The agonist activity has been confirmed with evidence from cortical brain slices showing that this class of hallucinogens are 5-HT receptor partial agonists (Marek & Aghajanian, 1996), although there is still some debate as to whether all hallucinogens are partial agonists (Villalobos et al., 2004).

It was discovered early in research, that hallucinogens of various types reduced serotonin turnover in the brain and this was mirrored by the fact that systemically applied (injected into the body or given orally, not injected into the brain) hallucinogens inhibited the firing of cells in the dorsal raphe nucleus (Aghajanian et al., 1970; Aghajanian et al., 1968). Furthermore, it was shown that indole hallucinogens, microiontophoretically applied to the dorsal raphe, inhibited the firing of cells there (Aghajanian et al., 1972; Aghajanian & Hailgler, 1975; de Montigny & Aghajanian, 1977). Indeed, the ability for these chemicals to inhibit raphe cell firing seemed a likely hypothesis for their mechanism of action, as these cells are the source of cortical serotonin and hence inhibiting them could cause changes throughout the brain consistent with the broad effects of hallucinogens.

There were problems with this theory, however, because systemically applied phenethylamine hallucinogens had a mixed effect on raphe neurons, inhibiting about half, having no effect on some neurons and even exciting others (Aghajanian et al., 1972; Aghajanian & Hailgler, 1975). Phenethylamine hallucinogens had no effect on raphe neurons when applied directly onto the cells (Aghajanian et al., 1972; Aghajanian & Hailgler, 1975). Furthermore, the effects LSD induced in the raphe were found not to be proportional to the behavioural effect, the effects of the raphe outlasted the behavioural effects, and tolerance to LSD-induced behavioural effects was not proportional to the tolerance induced in the raphe (Trulson et al., 1981). Finally it was shown that the effect indole hallucinogens have on the raphe was mediated by activating 5-HT_1A receptors, an effect that is shared with many non-hallucinogenic drugs and is almost certainly unnecessary for hallucinogenic activity. Although this makes the idea that hallucinogens cause their unique action via inhibiting raphe cell firing extremely unlikely, it is not impossible that lowered 5-HT release in part mediates hallucinogen action, as systemically applied phenethylamines do inhibit some raphe neurons (Aghajanian et al., 1972), and decrease 5-HT turnover (Anden et al., 1974).

The first evidence that the hallucinogens act via the 5-HT₂ receptor was the paper by Glennon et al. (1983). This paper, using a drug discrimination task, showed that in rats trained to discriminate the phenethylamine hallucinogen DOM (Fig. 1C) from saline (imagine rats who learn to press one lever when they are given the training drug and another when they are given anything else), the DOM stimulus generalized to both indole and other phenethylamine hallucinogens, but not other 5-HT receptor agonists. Trained rats are said to "generalize" in their responses when they press the lever corresponding to the training drug when they are given another kind of drug; essentially the rats are saying "this drug feels like the training drug in some way". It was also shown that this stimulus generalization was blocked by the 5-HT₂ receptor antagonist, ketanserin. Glennon et al. (1984a) then reported that the binding-affinity (how strongly a chemical 'fits' into a neural receptor) of various hallucinogens for 5-HT₂ subtypes, but not for other serotonin receptors, was tightly correlated with the drug's dose in human usage and ED50 (the dose of drug needed to get 50% generalization) in rat drug discrimination experiments (Fig. 2).

This correlation has been shown several times (Glennon et al., 1986; Glennon et al., 1984b; Sanders-Bush et al., 1988), and along with a large number of agonist/antagonist experiments (reviewed by Nichols, 2004) it is generally agreed that 5-HT₂ receptor subtypes are the primary site of action of hallucinogens.

As the documented number of 5-HT receptor subtypes proliferated, the main question became which subtype mediated the hallucinogenic action of hallucinogens. LSD binds potently to the 5-HT_{1A/1B/1D/1E/1F} (Hoyer, 1988; Lovenberg et al., 1993b), to the 5-HT_2A/2B/2C (Porter et al., 1999) and to the 5-HT_5A/5B/6/7 receptors (Matthes et al., 1993; Monsma et al., 1993; Ruat et al., 1993) but does not bind to the 5-HT3/4 receptors (Gerald et al., 1995; Peroutka & Hamik, 1988). All of the phenethylamine hallucinogens studied only bind to the 5-HT_2A/2B/2C receptors (Adham et al., 1993; Erlander et al., 1993; Lovenberg et al., 1993a; Nelson et al., 1999; Pierce & Peroutka, 1989; Titeler et al., 1988; Zgombick et al., 1992). Given that these 5-HT₂ receptors are the only shared targets between LSD and the phenethylamine hallucinogens and that the 5-HT_2B receptor is only expressed very weakly in the brain (Pompeiano et al., 1994; Schmuck et al., 1994), it seems that the 5-HT_2A/2C receptors are the only remaining candidates as the site of hallucinogenic action for this class of drugs.

There are a number of reasons why the 5-HT₂ receptor has been identified as the primary site of action for hallucinogens. First, there is a preponderance of 5-HT₂ to 5-HT_2C receptors in the neocortex (Pompeiano et al., 1994; Wright et al., 1995). Also, when 5-HT₂ receptor antagonists are used to block the stimulus cue of the phenethylamine hallucinogen DOI (Fig. 1B) and LSD in drug discrimination studies, there is a tighter correlation between the antagonists' affinity for the 5-HT_2A receptor and their ability to block the interoceptive cue than there is for the 5-HT_2C receptor (Fiorella et al., 1995). This means that it appears that blocking the 5-HT_2A receptor more directly blocks the effects in animals associated with hallucinogenic effects than blocking the 5-HT_2C receptor. Schreiber et al. (1994) reported that they were able to block drug discrimination in rats using agonists selective for 5-HT_2A but not 5-HT_2C.

Further evidence for 5-HT_2A's central role in hallucinogenic activity comes from research with DMT. The indole hallucinogen N,N-dimethyltryptamine (DMT) does not produce tolerance to its hallucinogenic effects in humans (Strassman, 1996) and in line with this, in cultured fibroblasts expressing 5-HT_2A/2C receptors, 5-HT_2A receptor does not show desensitization to DMT, but the 5-HT_2C receptor does (Smith et al., 1998). Finally, in human trials, the indole hallucinogen psilocybin produced "altered states of consciousness" which were blocked by ketanserin, which has an affinity for the 5-HT_2A 10-30 times that of the 5-HT_2C receptor (Vollenweider et al., 1998).

Given that the 5-HT_2A receptor is the site of action for hallucinogens, an attempt to localize the receptor may give insights into how hallucinogens affect consciousness. Early studies attempting to localize the receptor involved in hallucinogen action used halogenated LSD analogues, which presumably had the same non-specific binding pattern as LSD and hence did not report 5-HT_2A receptor specifically (Engel et al., 1984; McKenna & Saavedra, 1987; Wong et al., 1987). Most recent studies have used very specific antibodies to localize 5-HT_2A receptors (a technique called immunocytochemistry), and generally, their results have been in complete agreement. Jakab and Goldman-Rakic (1998) examined macaque brains and reported dense 5-HT_2A receptor immunoreactivity (places where the antibodies bound specifically) throughout all cortical regions, specifically in the frontal, prefrontal, temporal and occipital cortex. Throughout the cortical sheet there were two intensely stained bands, consisting of layer II and III, and layer V and VI.

The cerebral cortex is a convoluted 'sheet' of brain tissue about 2-5mm thick, whose folds form the outer inch of the human brain and give the brain its signature uneven, bumpy appearance. The cerebral cortext is itself made up of layers, a feature that was described as early as 1776. The cerebral cortex (which is also referred to by its major part, the neocortex) is still usually described as being divided into six distinct layers, labeled I through VI. Although the function of the different cortical layers are not yet well identified, each layer is generally stereotyped as having inputs from certain other layers and locations in the brains, as well as outputs that reach into other layers or other areas of the brain. The possible significance of the effect on layer V pyramidal cells is commented on in the "Putting it all together" section below.

Most, if not all, of the pyramid-shaped cells (called 'pyramidal cells') in this experiment reacted to the 5-HT_2A immunoassay and were 'labeled' by the antibodies as having these serotonin receptors. Many large- and medium-sized interneurons were also labeled, while most small-sized interneurons were unlabeled.

There were also a small percentage of weakly labeled presynaptic sites, often containing dense core vesicles. Miner et al. (2003) quantified the different locations of 5-HT_2A receptors in the rat prefrontal cortex using various immunocytometric techniques. Out of 325 identifiable structures, it was reported that 73% of immunoreactive sites were postsynaptic, and the majority of these, extrasynaptic. Twenty-four percent of the labeled profiles were presynaptic, belonged to thin, unmyelinated axons and were often seen to contain dense core vesicles, that is to say, terminals which looked like monoaminergic neurons.

The 5-HT_2A receptor is a protein, and like all other proteins is produced from messenger RNA (mRNA). By localizing mRNA, one can localize the cell bodies which produce the protein in question. This technique is called "in situ hybridization". In contrast to immunocytochemistry, in situ hydrbization doesn't tell you exactly where on the cell the protein is located, but what it tells you is which cells produced the protein in the first place. This is very useful when it comes to neurons which project long distances. For instance, neurons of the thalamus largely project to the cortex. Immunocytochemistry would just tell you that there are 5-HT_2A receptors in the cortex, in situ hybridization on the other hand can tell if the receptors are on cells which originated in the thalamus. The thalamus is another area where hallucinogens are speculated to work (Lambe & Aghajanian, 2001; Marek et al., 2001) and two studies using in situ hybridization have shows that the reticular and lateral geniculate nuclei as well as the zona incerta of the thalamus all contain 5-HT_2A receptor mRNA (Cyr et al., 2000; Pompeiano et al., 1994).

A description of the research into the mechanism of hallucinogen activity in the brain is necessarily highly technical. Those unfamiliar with neurophysiology will find this section very dense and may want to skip down to the Putting it all together conclusion where this is summarized.

In slices of the cerebral cortex, the most marked result of 5-HT_2A receptor activation is an increase in postsynaptic potentials, the electrical hallmarks of synaptic activity. Postsynaptic potentials essentially come in two flavours: excitatory postsynaptic potentials (EPSPs) and inhibitory postsynaptic potentials (IPSPs). EPSPs are "excitatory" because they bring the cell closer to firing. EPSPs are almost exclusively caused by the synaptic release of glutamate. IPSPs on the other hand are "inhibitory" because they reduce the likelihood that the cell will fire. IPSPs are caused by the synaptic release of GABA. In rat piriform cortex pyramidal cells, DOI and LSD induce IPSPs by directly exciting GABAergic interneurons (Marek & Aghajanian, 1996). However, 5-HT_2A receptor activation causes different effects in other parts of the brain. In medial prefrontal cortex (mPFC) pyramidal cells, 5-HT_2A receptor activation causes some IPSPs but primarily causes an increase in the amplitude and especially frequency of spontaneous EPSPs with a small effect on electrically evoked EPSPs, especially in the later phases of the EPSP (Aghajanian & Marek, 1997; Aghajanian & Marek, 1999; Marek & Aghajanian, 1999; Marek & Aghajanian, 1998).

Simply put, this research shows that activating the 2A subtype of serotonin receptor (5-HT_2A) causes an increase in synaptic traffic and its role is mainly excitatory.

Applying 5-HT to cortical slices and recording activity from layer V pyramidal cells produced the most robust EPSPs in a distribution pattern closely matching the known distribution of 5-HT_2A receptors. This type of test showed the highest activation in Layer I and Va (Aghajanian and Marek, 1997). If this large increase in EPSPs in the mPFC occurred in a whole animal situation (aka in vivo), one would expect to be able to record an increase in glutamate in the extracellular fluid surrounding these excited cells (as EPSPs are caused by the release of glutamate). Scrugges et al. (2003) showed this with in vivo microdialysis; showing DOI induces an increase in extracellular glutamate in the medial prefrontal cortex (mPFC) of awake animals.

Also, in two different human trials, psilocybin induced a large increase in frontal cortex metabolism with more modest increases in other cortical and subcortical areas (Gouzoulis-Mayfrank et al., 1999; Vollenweider, 2001). As firing/active neurons require more energy than inactive ones, this increase in metabolism is believed to reflect an increase in neuronal activity.

The excitatory postsynaptic potentials (EPSPs) induced by 5-HT activating the 5-HT_2A receptor in mPFC pyramidal cells are Ca2+ and tetrodotoxin sensitive. Seemingly indicating that they are the product of neurons firing (as neuronal firing is blocked by tetrodotoxin) and the release of neurotransmitter-containing vesicles (a Ca2+ dependent process. Also, the EPSPs are inhibited by classical presynaptic inhibitors like the group II/III metabotropic glutamate receptor (mGluR2/3) agonists (Marek et al., 2000) and mu-opioid receptor agonists (Marek & Aghajanian, 1998). These facts together indicate that the EPSPs are not a product of postsynaptic 5-HT_2A receptor activation.

On the other hand, the increase in EPSPs does not seem to be due to an increase in neighboring neurons firing and activating the cells being recorded from, as neighbouring cells have not been recorded as having action potentials when the 5-HT was directly applied (Aghajanian & Marek, 1997). As mentioned above, if the solution that the slice is perfused with to keep it alive is Ca2+ free, then all 5-HT-induced increase in EPSPs evoked by electrical stimulation (which excites neighbouring cells) and spontaneous EPSPs were blocked, but if Ca2+ was replaced with Sr2+, the 5-HT induced increase in spontaneous EPSPs returned.

Previous experiments had only briefly looked at the effect that 5-HT_2A receptor activation has on electrically evoked EPSPs, so Aghajanian and Marek (1999) examined it a bit more closely. It was noted that the increase in evoked currents caused by 5-HT_2A receptor activation induced by DOI were found largely after the classical electrically evoked excitatory postsynaptic potentials (direct stimulation using micro-electrodes). This is shown as the slow sag in current after the quick dip (EPSC) in current in Fig. 3E. (Aghajanian & Marek, 1998; Aghajanian & Marek, 1999). Both the spontaneous and evoked EPSPs had all the hallmarks of "asynchronous" transmitter release; a form of release that happens after the action potential has caused classical synchronous or phasic transmitter release, that can be evoked in solutions where Ca2+ is replaced with Sr2+ (Kirischuk & Grantyn, 2003; Rumpel & Behrends, 1999). Thus, it was hypothesized that hallucinogens had their effect by increasing asynchronous transmission via presynaptic 5-HT_2A receptors on glutamatergic terminals.

Another effect of 5-HT_2A receptor activation in layer V medial pre-frontal cortex (mPFC) cells is an enhancement of a voltage-sensitive persistent sodium current (INaP) (Aghajanian and Marek; 1997). This means that whenever excitatory currents bring a cell closer to its threshold for firing, a supplementary current will be induced, effectively amplifying the synaptically induced current. This may in part explain the increase in EPSP amplitude seen during application of 5-HT_2A agonists.

One way to try to locate which neurons in the brain are responsible for the activity is to destroy specific areas in the brains of rats and mice and then test them to see what changes have occured.

Marek et al., (2001) lesioned (damaged/destroyed) the medial thalamus and amygdala of rats in an attempted to define the location and source of the 5-HT_2A receptors which are activated by their agonists and cause increases spontaneous EPSPs in the layer V of the medial prefrontal cortex (mPFC). The neurotoxic chemical N-methyl-D-aspartate (NMDA) was infused by cannula into the brains of living animals in order to lesion the midline and intralaminar nucleus of the thalamus, and using radiofrequencies, the basolateral nucleus of the amygdala was lesioned. These nuclei were chosen because they are classical sources of excitatory input into the mPFC and they are one of the few nuclei whose projections share the laminar pattern 5-HT_2A receptors distribution in the mPFC (Bacon et al., 1996; Berendse & Groenewegen, 1991).

Roughly 18 days after the damage, the animals were sacrificed and cortical slices were taken to ensure the accuracy of the lesion, electrophysiological work was done and 5-HT_2A, mu-opioid and mGluR2/3 receptors were imaged via autoradiography. It was reported that the thalamic lesions were largely specific to the basolateral, midline and intralaminar nuclei, and where they weren't, the damage spilled onto thalamic nuclei with no input to layer V. In cortical slices where the midline and intralaminar nuclei were lesioned, there was no change in spontaneous or chemically evoked EPSPs recorded in mPFC pyramidal neurons, indicating that there was no change in the mPFCs sensitivity to excitatory input. However, there was 60% decrease in the frequency of 5-HT induced EPSPs. Lesions of the amygdala had no effect on 5-HT induced EPSPs. Lesions to the thalamus lead to a decrease in mGluR2/3 (~20%) and mu-opioid receptor binding (20-50%) throughout the layers of the cortex, but an increase in 5-HT_2A receptor binding.

These results, although not conclusive, give weight to several ideas. For one, the dramatic decrease in 5-HT_2A activation-mediated EPSPs in mPFC pyramidal cells caused by midline and intralaminar lesions indicate that these nuclei, but not the basolateral nucleus of the amygdala, are a major source of the cortical afferents responsible for the EPSPs. It had previously been shown that mGluR2/3, and mu-opioid receptor activation inhibits 5-HT_2A mediated EPSPs, the decrease in their density after thalamic lesioning indicated that some of these receptors were located on the thalamocortical neurons. Finally, the fact that thalamic lesions caused a significant decrease in EPSP frequency yet caused an increase in 5-HT_2A receptor density, indicates that there is a very limited amount, if any 5-HT_2A receptors on the projection of midline and intralaminar neurons in the cortex. Therefore it is very unlikely that 5-HT_2A receptors on the presynaptic terminals of thalamocortical neurons are responsible for the increase in EPSPs in mPFC pyramidal cells caused by 5-HT_2A receptor agonists.

These results with the other research looking at the 5-HT hallucinogens are somewhat paradoxical: a presynaptic action of 5-HT_2A receptors on presynaptic terminals devoid of 5-HT_2A receptors. One way to avoid the paradox would be a retrograde messenger. Conceivably, 5-HT_2A receptor activation on the postsynaptic terminal could lead to the generation of a retrograde messenger, which could activate the presynaptic terminal. This possibility gained more ground after Lambe and Aghajanian (2001) showed the similarity between the EPSPs induced by 5-HT_2A receptor activation and antagonists of Kv1.2- and possibly Kv3-containing potassium channels, channels which are blocked by the retrograde messenger arachidonic acid (Poling et al., 1995; Poling et al., 1996).

It was reported that there was a correlation between the affinity of various potassium channel antagonists for Kv1.2-containing channels and their ability to induce EPSPs in layer V mPFC pyramidal cells. There were many similarities between the EPSPs induced by the Kv1.2 antagonist mu-dendrotoxin (DTX) and 5-HT. The EPSP traces recorded looked similar in both amplitude and frequency; they were induced most strongly in the same layers of the cortex and they were inhibited by mu-opioid agonists and thalamic lesions. Importantly, 5-HT induced EPSPs were only slightly additive with DTX-induced EPSPs, and when DTX was added with tetraethylammonium (TEA), a general potassium channel antagonist, 5-HT could not induced any more EPSPs at all.

To exclude the possibility of a ceiling effect, it was shown that nicotine induced EPSPs were completely additive with DTX induced EPSPs. It is believed that the other potassium channel that were blocked by TEA are Kv3.2-containing channels, as it is highly expressed in thalamic neurons (Weiser et al., 1994) and its expression is highly decreased by thalamic lesions (Moreno et al., 1995). Interestingly, the Kv1.2 and Kv3.2 are the only two potassium channels which have been shown to be inhibited by arachidonic acid, a product of 5-HT_2A receptor activation (discussed below) (Poling et al., 1995; Poling et al., 1996).

Blockade of presynaptic potassium channels causes depolarization that leads to an increase in intracellular Ca2+ concentration and hence increases the rate at which neurotransmitters are spontaneously released. Despite conventional thinking, this effect has been shown to be voltage sensitive sodium channel dependent (Lambe & Aghajanian, 2001; Sanchez-Prieto et al., 1996) and may explain why the increase in EPSPs caused by 5-HT_2A receptor activation is TTX-sensitive (a blocker of voltage sensitive sodium channels). Arachidonic acid can be produced via the action of phospholipase A2 (PLA2), an enzyme that 5-HT_2A receptor activation is known to stimulate along with its classical linkage to phospholipase C (PLC) (Kurrasch-Orbaugh et al., 2003). Therefore, it was hypothesized that the EPSPs induced by 5-HT_2A receptor agonists are caused by the activation of postsynaptic 5-HT_2A receptors, the production of arachidonic acid by PLA2, and its action on presynaptic potassium channels.

This hypothesis raises an interesting problem; although hallucinogens' behavioral potencies are tightly correlated with their 5-HT_2A receptor affinity, there seems to be no correlation between behavioral potencies and their ability to stimulate second messenger production (Nichols, 2004). For example, LSD is approximately 20 times more potent than the hallucinogenic amphetamine DOB in drug discrimination experiments (Glennon et al., 1984a). But in mouse fibroblasts, DOB is slightly more potent than LSD at inducing arachidonic acid release, and only around seven times less potent than LSD at stimulating PLC-mediated inositol phosphate production. It is also possible that one second messenger is 'pro-hallucinogenic' while another is 'anti-hallucinogenic'. This theory could explain the lack of correlation between second messenger production and behavioral effect, as it would not be the potency of a drug at stimulating second messenger production, but the ratio of the potency at producing one second messenger to another. On top of that, the 5-HT_2A receptor pathway has one more card up its sleeve, as it has been shown to be strongly linked to phospholipase D (PLD) in a ADP-ribosylation factor-dependent manner (Mitchell et al., 1998; Robertson et al., 2003). Unfortunately, no studies have looked at the ability of hallucinogens to activate PLD.

The situation becomes more difficult still when one tries to include the results of Puig et al., (2003). In that research, DOI (50-300åµg/kg I.V.) was given to anaesthetized animals and the firing rates of mPFC neurons were recorded. It was shown, as expected from the above results, that DOI increased the overall firing rates 238% in a largely 5-HT_2A receptor dependent manner (DOI excited 38% of neurons, on average 481%, 32% were unaffected and 30% were inhibited, on average 11%). However the DOI-induced increase in firing rate was not inhibited by mu-opioid or mGluR2/3 agonists or lesions to the midline and intralaminar thalamic nuclei (i.e. the same nuclei which Marek et al. (2001) lesioned and reported that 5-HT_2A-mediated spontaneous EPSPs were massively reduced).

Interestingly, injection of excitatory drugs (the GABA-A antagonist bicuculline) into the thalamus produced an increase in cellular firing which physiologically and pharmacologically mirrored the EPSPs of Aghajanian and Marek: they were reduced by mu-opioid or mGlurR2/3 agonists and would presumably be blocked by thalamic lesions. Exactly what these findings say about those put forward by Aghajanian and Marek is unclear. They could suggest that the excitatory drive is actually derived from cortico-cortical projections, something which is supported by Aghajanian and Marek's (1998; 1999) work on electrically evoked EPSPs (the evoked EPSPs were caused by stimulating the majority of inputs onto the cell, most of which were excitatory cortico-cortical projections).

While there is an overarching theme that hallucinogens act by increasing excitatory cortical transmission, two papers challenge this classical stance with evidence that 5-HT_2A receptor agonists inhibit N-methyl-D-aspartate (NMDA) receptor-mediated currents (Arvanov et al., 1999a; Arvanov et al., 1999b). It was reported that in layer V pyramidal cells in mPFC slices, DOB at low concentrations (0.01-1åµM) caused an increase in the currents induced by perfusion with 10µM of NMDA. On the other hand at high concentrations (>1µM) it produced a Ca2+/calmodulin kinase II-dependent blockade of NMDA-induced currents.

The potentiating effect of DOB was Ca2+ sensitive, and presumed to be of presynaptic origin, in line with the above research. The AMPA receptor antagonist CNQX was added, to block the effect of any spontaneously released glutamate, and it was then seen that even low concentrations of DOB would decrease NMDA-induced currents, with an IC50 of 130nM (IC50 is the "inhibitory concentration 50%", or the concentration necessary to cause 50% inhibition). This was repeated with LSD, and it was shown to have an IC50 of 9nM. The authors stated that the concentration of LSD needed to induce an enhancing effect on NMDA induced currents was >300nM and far above the 10-20nM plasma concentration found in recreational use (Aghajanian & Bing, 1964; Hawks & Chiang, 1986), and hence unphysiological.

However, it should be noted that although potentiation of NMDA receptor-mediated currents seems unlikely to be a cause of the hallucinogenic action of 5-HT_2A receptor agonists, that does not say they can not act by causing glutamate release as described above. Indeed, if hallucinogens worked solely by inhibiting the NMDA receptor in some fashion, one would expect hallucinogens to generalize to NMDA-receptor antagonists in drug discrimination experiments, but this has not been shown consistently (Jones et al., 1998; West et al., 2000). Furthermore, if inhibiting the NMDA glutamate receptor was the complete explanation of hallucinogen action, it seems surprising that mGluR2/3 agonists, which inhibit 5-HT_2A receptor-mediated glutamate release in vitro, also inhibit behavioral aspects of hallucinogens such as head shakes (Klodzinska et al., 2002) and drug-induced stimulus control (Winter et al., 2004). However it seems possible that hallucinogens work by both increasing glutamate release and inhibiting NMDA-receptors. Indeed, if hallucinogens have varying abilities to inhibit NMDA-receptors and induce EPSPs in cortical pyramidal cells, it could help explain how different hallucinogens seem to produce the wide ranging subjective effects reported by recreational users (Shulgin & Shulgin, 1991; Shulgin & Shulgin, 1997).

Another possibly important effect of hallucinogens is their ability to alter the firing of the locus coeruleus (LC), the major source of noradrenaline (NA) in the brain. Systemic administration of either phenethylamine or indole hallucinogens to anaesthetized animals causes a 5-HT_2A receptor dependent decrease in the spontaneous activity of the LC, but an increase in activity evoked by sensory stimulation (Aghajanian, 1980; Rasmussen & Aghajanian, 1986; Rasmussen et al., 1986). This effect was shown to be mediated by 5-HT_2A receptors extrinsic to the LC, as hallucinogens microiontophoretically applied to LC cell bodies did not produce this effect. The ability of 5-HT_2A receptor agonists to inhibit LC firing was blocked by local administration of GABAA antagonists and the stimulatory action of hallucinogens on LC activity was blocked by NMDA receptor antagonists, but not non-NMDA ionotropic glutamate receptor antagonists (Chiang & Aston-Jones, 1993).

The LC has been likened to a "novelty detector" due to its increased activity in response to novel stimuli. The release of NA also seems to gate sensory information to a degree, and to increase the signal to noise ratio, by inhibiting basal neuronal activity but increasing responses to sensory stimulation (reviewed by Sara et al., 1994; Woodward et al., 1991). Interestingly, the laminar distribution of the alpha1-adrenoreceptor closely matches that of the 5-HT_2A receptor (Palacios et al., 1987), and NA induces a alpha1-adrenoreceptor-dependent increase in EPSPs in layer V mPFC cells in a similar fashion to 5-HT (Marek & Aghajanian, 1999). It seems possible then that the increase in sensory-evoked LC firing could produce some of the cognitive effects induced by hallucinogens, such as ordinary objects appearing fascinating.

Unfortunately, the majority of recent research into hallucinogens has focused on isolated brain regions, specifically using in vitro cortical brain slices. This preparation is usually devoid of the spontaneous activity seen in the neocortex of an intact animal. Although one can replicate the slow spontaneous rhythmic activity seen in vivo if the bathing medium closely matches the extracellular ionic composition seen in the brain in situ, one can not replace the cortical and subcortical inputs lost when the section is taken from the brain (Sanchez-Vives & McCormick, 2000). Indeed, these subcortical and especially the thalamic inputs seem critical to this complex activity when it is hypothesized that thalamic projections are the neurons that supply the increased excitatory input in response to hallucinogens. Furthermore, cortical input from the locus coeruleus and the raphe nuclei are lost. Yet one can attempt to build a larger picture out of the available research.

As reviewed above, the medial prefrontal cortex (mPFC) cells that seem to be excited most by application of 5-HT_2A receptor agonists are layer V pyramidal cells. In a theory reviewed by Jones (2002), layer V pyramidal cells have been implicated in the binding of separate sensory stimuli into a discrete conscious event. Major sensory stimuli often lead to high frequency, 20-50Hz "gamma" oscillatory firing between thalamic relay cells and the area of cortex that they project to. These gamma oscillations are dependent on thalamo-cortico-thalamic circuits, and some are believed to code the conscious recognition of the stimuli (Golshani & Jones, 1999; Steriade & Amzica, 1996). This loop is created by thalamic relay "core" cells in specific thalamic nuclei activating layer VI pyramidal cells, and their projections back to the core cells. This loop system would stay in the cortical columns originally activated by the thalamic relay cell, if it did not also activate layer V pyramidal cells. Layer V cells have diverse intracortical projections and, importantly, project to many non-specific thalamic nuclei. Therefore these would "decide" which distant cortical areas could be activated by the original stimulus.

This way, associated stimuli could be bound to form synchronous activity across the cortex. By activating layer V pyramidal cells' coritco-cortical and cortical-thalamic projections, hallucinogens could cause the spread of high-frequency oscillations to areas that would not normally be activated. In sensory cortexes this could produce effects such as the hallucinations and synesthesia. In the frontal cortex the spreading of high frequency oscillations could result in mood changes and alteration to ego perception. Indeed this theory can be used to explain any facet of the hallucinogenic experience where a stimulus produces an inappropriate perception (e.g. an object being recognized as something it is not, a sound being heard as something it is not, though not simple visual effects like fractals). It is interesting to note that the frequency of excitatory postsynaptic potentials (EPSPs) induced by 5-HT_2A receptor agonists was shown to be 36Hz (Aghajanian & Marek, 1998), right in the middle of the gamma frequency, (explained above).

Other 'big picture' theories of hallucinogen action involve their potential action on the thalamic reticular nucleus (TRN). The TRN is thought to work as a filter for transmission of information from the thalamus to the cortex (reviewed by Guillery & Harting, 2003). TRN neurons are nearly exclusively GABAergic, and can switch thalamic relay cells from tonic firing mode to burst firing mode. In tonic mode, the relay cells can accurately transmit sensory information because their firing rate closely reflects the rate and amplitude of excitatory inputs while cells in burst firing mode reach a maximum firing rate when inputs reach 15Hz at rates above 100Hz, but cells in burst fire mode can not fire faster than 15Hz and hence the integrity of information is lost (Kim & McCormick, 1998; McCormick & Feeser, 1990).

It is generally believed that bursts act as some kind of "wake-up call" to the cortex, as bursts are usually firing in response to a new stimulus (Grubb & Thompson, 2005). However, this could just be because bursts are initiated when a cell is excited out of an unexcited (hyperpolarized) state, the kind of state a neuron would be in between presentations of stimuli in an experimental setting i.e. in an anaesthetized animal. Because there is a large difference in the neurophysiological consequences caused by different patterns of cell firing and certain firing modes predominate during different behavioural states (tonic firing is more common in awake animals while burst firing is more common during sleep) (Steriade & Llinas, 1988), it seems unlikely that there is no functional difference between these two firing modes.

In a model proposed by Vollenweider and Geyer (2001) activation of 5-HT_2A receptors on TRN neurons could activate them in the same way that 5-HT_2A receptors activate GABAergic neurons in the cortex, and decrease the ability of the TRN to gate information flow effectively.

This author notes that the zona incerta of the thalamus also expressed 5-HT_2A receptors (Pompeiano et al., 1994), and projects GABAergic neurons to higher-order thalamic relay cells (Bartho et al., 2002). The zona incerta, unlike most thalamic nuclei, receives its cortical input from almost exclusively layer V pyramidal cells (Bartho et al., 2002; Mitrofanis & Mikuletic, 1999), and this is another way by which hallucinogens could effect cortical function.

Along with any cortico-thalamo-cortical action of hallucinogens, their action in potentiating locus coeruleus (LC) firing in response to sensory stimuli will also excite cortical cells, possibly in a synergistic fashion to 5-HT_2A receptor stimulation. The idea that hallucinogens work at least in part by inhibiting firing of cells in the raphe nuclei was once seen as a major hypothesis, but now seems unlikely, thought it may modulate the subjective effect hallucinogens produce. Unfortunately, in the end, most of these theories are speculative rather than being scientific hypthoses based on empirical observations, as there is only a very limited amount of research on hallucinogens and most of that research is restricted to in vitro electrophysiology on cortical cells. Future research needs to look into the effect that hallucinogens have on thalamic cells and thalamocortical oscillations. Since the intracellular cascade responsible for hallucinogen action is not yet fully understood, it is hard to imagine science being able to explain the much more complicated details of hallucinogens' action on consciousness.