Ketamine has been around for many years, firstly as a dissociative anaesthetic and then as a psychedelic drug. But it might become best known for it’s powerful antidepressant properties (Berman et al 2000; Zarate et al 2006). Compared to existing antidepressants, which take around 2 weeks to work, ketamine exerts a large antidepressant effect on the first day of treatment.
The robust antidepressant effect of ketamine also occurs in patients who have not found relief with existing drugs or with ECT. In the latest study to be reported, 24 patients with treatment-resistant depression underwent up to 6 sessions of intravenous ketamine (0.5mg/Kg in 40 mins) over ~2 weeks. Over 70% of patients responded to ketamine, and the overall reduction in depression was large and rapid (Murrough et al 2013) (Figure 1).
To date a major issue has been the lack of persistence of the antidepressant effect. In previous studies, involving a single ketamine treatment, depression returned within one week of the session or less. In the study by Murrough et al, this was extended to an average of 18 days. This is an improvement, but further work will be needed to solve the problem of the relatively short-lived antidepressant effect of ketamine.
An understanding of the mechanism by which ketamine alleviates depression may be necessary if we are to extend the duration of it’s beneficial effects. Pre-clinical work suggests that ketamine boosts the health and integrity of synapses and neuronal networks. Much of the action is believed to take place within dendritic spines, and involves local protein synthesis (Duman et al 2012) (Figure2).
Two molecules of relevance are mTOR and GSK-3. Ketamine enhances local protein synthesis by activating mTOR and by inhibiting GSK-3. [GSK-3 inhibits mTOR]. A drug, such as lithium, which inhibits GSK-3 might enhance the antidepressant effect of ketamine. This has now been demonstrated in pre-clinical studies (Liu et al 2013). The clinical question, which will now be addressed in trials is whether lithium treatment extends and enhances the antidepressant effects of ketamine. Lithium has been used for treatment-resistant depression for many years, and has a good evidence base (Bauer et al 2010) so that the combination of ketamine and lithium presents as an interesting and relatively straightforward strategy for stubborn depression.
However it is somewhat odd that the proposed mechanism for ketamine involves new protein synthesis and synaptogenesis (which take time, and are sustained) whereas the clinical effects of ketamine are very rapid (and transient). Other mechanisms may have more explanatory power. For instance a recent fMRI study showed that ketamine decreased the connectivity of limbic and prefrontal regions which are known to be overactive in depression (Scheidegger et al 2012). More provocatively, it appears that the antidepressant effect of ketamine depends upon the extent of the acute psychological reaction produced by the drug. Although the dissociative/psychedelic properties of ketamine are sometimes regarded as unwanted “side-effects”, a recent paper showed that the acute psychedelic and subsequent antidepressant effects are related (Sos et al 2013).
Relationship of ketamine’s antidepressant and psychotomimetic effects in unipolar depression.
Ketamine and other NMDA (N-methyl-D-aspartate) antagonists produce fast-acting antidepressant-like effects, although the underlying mechanism is unclear. Furthermore, high affinity NMDA antagonists such as ketamine are associated with psychotomimetic effects. To date the link between the antidepressant and psychotomimetic effects of ketamine has not been explored. We examined the relationship between the antidepressant and psychotomimetic effects of a single ketamine infusion in subjects diagnosed with major depressive disorder.
In a double-blind, cross-over, placebo-controlled, two weeks clinical trial we studied the effects of ketamine (0.54 mg/kg within 30 min) in a group of 27 hospitalized depressive patients.
Higher intensity of psychotomimetic symptoms, measured using BPRS, during ketamine administration correlated with alleviation in mood ratings during the following week with maximum on day seven. Ketamine was superior to placebo in all visits (day 1, 4, and 7) assessed by MADRS with effect size (Cohen´s d) of 0.62, 0.57, and 0.44 respectively. There was no significant correlation between ketamine and nor-ketamine plasma levels and MADRS score change at any study time point.
The substantial relationship between ketamine’s antidepressant and psychotomimetic effects was found. This relationship could be mediated by the initial steps of ketamine’s action, trough NMDA receptors, shared by both ketamine’s clinical effects.
A major contribution of neuroscience to the humanities is the knowledge that the structure of the brain is moulded by the experiences the mind goes through – the phenomenon known as plasticity. It means that the circuits of the brain are sculpted by habitat, schooling, language, relationships, and culture, as well as by the unfolding genetic programme. The action occurs below the micrometre scale – at synapses (the points of connection between neurons) – and involves the exquisite choreography of a number of molecular machines. These molecular processes are so fundamental for cognition that their failure (whether driven by gene mutation or by harsh environments) results in neuropsychological disability. A major locus of plasticity (and hence, cognitive disability) is the dendritic spine.
Principal neurons in the brain, such as cortical pyramidal neurons, express tens of thousands of small protruberances on their dendritic trees. These structures (dendritic spines) receive excitatory information from other neurons, and are highly dynamic. They can adjust their responsiveness to glutamate (the major excitatory neurotransmitter), becoming stronger (potentiation) or weaker (depression), as local circumstances dictate. This strengthening (LTP) or weakening (LTD) can be transient, or persist over long periods and as such, serves as an ideal substrate for learning and memory at synapses and in circuits. Potentiated spines increase in size, and express more AMPA glutamate receptors, whilst the opposite pattern occurs in synaptic depression to the extent that spines can be ‘absorbed’ back into the dendritic tree.
Over the course of childhood, dendritic spines (excitatory synapses) increase in number, but their numbers are ‘pruned’ back during adolescence to reach a plateau. Enriched environments have been shown to increase spine density, impoverished environments the opposite. In common psychiatric disorders, spine density is altered. For example, the most robust histological finding in schizophrenia is a reduction of spine density in the frontal cortex, auditory cortex and the hippocampus. In major depression, spines (and dendrites) are lost in the hippocampus. In autism, spine density actually increases. Finally, in Alzheimer’s and other dementias there is a catastrophic, and progressive loss of cortical and sub-cortical spines.
REGULATION OF THE SPINE:
The molecular biology of dendritic spines involves hundreds of proteins, but the outlines are now reasonably well understood. Scaffolding proteins [such as PSD95, shank(s), AKAP, stargazin and homer(s)] provide structural support and provide orientatation for membrane bound receptors, ion-channels and their downstream signalling pathways. The scaffold (post-synaptic density), facilitates effective signalling by ensuring that the correct protein partners are in close apposition. The scaffold is also tethered to proteins which bridge the synaptic cleft (cell adhesion molecules) and to bundles of actin filaments which provide the structure and force for spine enlargement (and retraction).
There is a constant remodelling of the actin cytoskeleton within the spine in response to synaptic and network signalling. Remodelling is via small, cytoplasmic G-proteins from the RHOfamily. Some family members promote the growth and stabilisation of actin filaments, whereas others promote actin disassembly. Mutations in the proteins which regulate actin dynamics are a cause of learning disability. Finally local protein synthesis (and degradation) occurs within dendritic spines, is tightly controlled and is essential for plasticity. Abnormalities in local protein synthesis within the spine underlie learning disability syndromes such as fragile X, neurofibromatosis and tuberous sclerosis.
Recent years have seen glutamate synapses move to centre stage in neuropsychiatry. This is not surprising given the role of pyramidal neurons (glutamate containing neurons) in information processing, and the role of glutamate transmission in learning and memory [see link]. But it is remarkable that so many psychological and cognitive disorders appear to ‘coalesce’ at dendritic spines.
The enclosed vector-graphic image [click here] highlights a selection of some of the proteins which are now known to be involved in autism, learning disability and schizoprenia.
Research will continue to decipher the complexity (and beauty) of the dendritic spine, but potential treatments are starting to emerge for disorders like fragile X, (which until recently were thought to be not amenable for drug treatment, as was the case for schizophrenia until the 1950s). Molecular neuroscientists will, almost certainly, continue to uncover more treatment targets. The task for psychiatry, as ever, is to keep abreast of neuroscience in all it’s complexity (and beauty).
Rapid Dissemination of Information
Glutamate and GABA are the archetypal ‘fast’ transmitters. If a neuron in the brain ‘wishes’ to communicate rapidly with another cell, the chances are that it will utilise glutamate or GABA. Of course, glutamate neurons exert an excitatory influence on the cells they contact, whereas GABA, at least on first glance, is inhibitory.
Fast transmitters bind to receptors on membrane-spanning ion channels. An ion-channel is in constant flux between various conformations: e.g. open, closed, desensitised. Binding of fast transmitter ‘causes’ the ion channel to snap open for brief periods, and ions rush down their concentration gradients causing an abrupt, short-lived, change in the local membrane potential of the post-synaptic cell (Figure 1). From start to finish the whole process is over within tens of milliseconds, and constitutes a discrete electrical signal (termed an excitatory or inhibitory post-synaptic potential; EPSP, IPSP).
Neurotransmission v neuromodulation
Fast transmission, as a concept, pre-supposes slow transmission. The classical slow transmitters are the monoamines, e.g. noradrenaline and dopamine. These substances are used as transmitters by neurons within specific brainstem nuclei, whose axons project to numerous subcortical structures and large areas of cortex. There are relatively few monoamine neurons (tens of thousands), but their projections show massive arborisation within the ‘higher centres’ and the limbic system. Anatomically, glutamate and GABA signalling is characterised by point-to-point communication between narrowly separated (and tethered) pre-synaptic and post-synaptic elements, whereas for monoamine systems, the release sites (boutons) and post-synaptic receptors are not necessarily in close proximity. In contrast to glutamate and GABA, which convey a fast, discrete, short-lived electrical signal, monoamines evoke slower-onset, diffuse, longer-duration biochemical changes in their target neurons. Monoamine systems are not optimised for the rapid dissemination of specific information, but instead for modulating those neurons that are.
Ensemble formation and Gestalts
Pyramidal neurons (the principal output neuron of the hippocampus and cortex) use glutamate as a transmitter to communicate rapidly with neurons in ‘lower centres’ such as the striatum, thalamus, pontine nuclei and the cord although most communication is with other pyramidal neurons. Pyramidal neurons organise themselves into ensembles. This process, in which pyramidal neurons fire in synchrony for brief periods of time is thought to be essential for object perception and for movement, speech and thinking.
Consider a pyramidal neuron ‘sitting’ at resting-membrane-potential (-70mV). It receives tens of thousands of excitatory (glutamate) inputs on its dendritic spines, (dynamic structures that are moulded by experience over a lifetime). A single excitatory input (by itself) has little overall impact on the pyramidal neuron. But when numerous EPSP’s from a multitude of inputs arrive ‘synchronously’, the depolarisation may be sufficient for the pyramidal neuron to fire an action potential (AP). In short, the pyramidal neuron is recruited (by the ensemble) into joining the ensemble.
It can be grasped that for AP firing to occur in a pyramidal neuron, there has to be a convergence of excitatory information from numerous sources. Excitatory inputs come from various thalamic nuclei and from stellate cells (in primary sensory cortices), although the overwhelming majority come from other pyramidal neurons. Regardless of the source, timing is key. In order to generate enough depolarisation to trigger an AP, inputs must arrive (and summate) within the same narrow time window (of the order of milliseconds).
Precise Timing and cortical dynamics
The output of a pyramidal neuron (AP spiking) is finely controlled. Precise timing is so fundamental for cortical processing that various auxiliary neurons appear to be tasked with a pacemaker role. These neurons utilise GABA as a transmitter. Classical neuroscience conceptualised GABA containing neurons as nothing more than inhibitory interneurons – this is no longer tenable. There are various populations of GABA containing neuron, which have been classified according to their morphology, their location in the cortex, which proteins they use to sequester calcium, and their electrophysiological properties. Some are even excitatory. For simplicity, we shall restrict ourselves to a simple classification based upon where the GABA neuron contacts the pyramidal neuron (Figure 2).
Contacts formed with the dendrites of pyramidal neurons function as inhibitory interneurons in the classical sense (i.e. they oppose excitatory drive), whereas GABA neurons targeting the soma or the proximal axon (of the pyramidal neuron) function as pacemakers. We can consider how these GABA pacemaker neurons are optimised for their task. Firstly they have very fast dynamics, swifter for example than the pyramidal neurons that they make contact with. Secondly, they provide a very strong and reliable signal to the pyramidal neuron by engulfing the soma or the proximal axon with numerous terminals. A strong, brief, recurrent signal to the soma and proximal axon creates a series of time windows, which determine precisely when the pyramidal neuron fires. Thirdly, individual pacemaker neurons make contact with numerous local pyramidal neurons. And finally, groups of pacemaker neurons are connected by electrical synapses (gap junctions) so that they can function as an interconnected single entity, a syncytium. For completion, pyramidal neurons make strong, reliable synapses (excitatory) with pacemaker neurons.
It is readily apparent that the interconnectivity of pyramidal neurons and GABA interneurons favours the emergence of oscillations, with successive, precisely timed periods of integration followed by periods of AP discharge. Experiments have shown that the population of neurons in an active ensemble generate the rhythm, whilst the rhythm puts precise constraints upon when an individual neuron can fire.
Systems and levels
For slow, diffuse modulators such as noradrenaline, it makes sense to talk of a system. To recap, noradrenaline [NA] is synthesized by no more than tens of thousands of neurons, confined to discrete nuclei within the brainstem, and is ‘sprayed’ from en-passant boutons over large territories of CNS tissue, in a hormone-like manner. Crucially, the release patterns of noradrenaline [and other neuromodulators] can be clearly mapped onto distinct behavioural states, the most marked differences arising in the sleep-state [noradrenaline – ‘off’] versus the waking-state [noradrenaline – ‘on’]. Since the extracellular concentrations of noradrenaline [and other neuromodulators] can inform directly about higher brain/mind levels, the idea of a noradrenergic system has utility.
Glutamate and GABA are too ubiquitous as fast point-to-point transmitters for the term ‘system’ to be applicable in the same way. Particular patterns of behaviour cannot be mapped onto the release of GABA or glutamate at a specific locus. All we can say is that neurons in an ensemble use glutamate and GABA to communicate with each other. Whereas transient fluctuations in the extracellular concentrations of GABA/glutamate do not reveal anything about behaviour, the dynamics of neuronal ensembles correspond with distinct behavioural states. Again the sleep wake-cycle is illustrative. Oscillatory activity generated by the ensemble can be mapped unambiguously onto the sleep-state and the waking-state.
Learning & Memory
In the 1970s it became clear that excitatory connections onto pyramidal neurons could be made stronger, if they were subjected to particular patterns of input. This was the first experimental support for an idea that can be traced back to Ramon y Cajal – the idea that synapses are modifiable (plastic) and that such plasticity might serve as the physical basis of memory.
There are various forms of plasticity, but the most widely studied is NMDA-dependent long-term potentiation (LTP). In the early 1980’s, researchers based in Bristol showed that NMDA receptor antagonists could block the initiation of LTP [and subsequent behavioural experiments, (most famously, by Richard Morris in Edinburgh) showed that such drugs could inhibit new learning].
NMDA receptor channels are found at the heads of dendritic spines, adjacent to the glutamate terminal. AMPA receptor channels are found in the same locale. When activated, both receptor channels produce an excitatory-post-synaptic-potential (EPSP). In the case of the AMPA receptor, the EPSP is mediated by sodium ions flowing into the spine. For NMDA receptors, the EPSP is mediated by a combination of sodium and calcium ions. [It is the calcium signal that initiates LTP (Figure 3). Early-phase LTP is mediated by phosphorylation of AMPA receptors (increasing their conductance) and by insertion of new AMPA receptors into the post-synaptic membrane].
AMPA and NMDA receptor channels differ in one other key property. The NMDA channel is voltage-dependent. At membrane potentials less than -50mV, the NMDA channel remains closed, even if glutamate is bound to the receptor. For the NMDA channel to snap open, the membrane potential must be already depolarised to at least -30mV. So two conditions are necessary for NMDA conductance; binding of glutamate and membrane depolarisation. For this reason, the NMDA receptor is said to be a coincidence detector (or in engineering terms, an AND gate).
Sufficient post-synaptic depolarisation can occur from backward-propagating action potentials (APs) or from temporally or spatially summated excitatory input to a dendritic branch. Research in the last decade has revealed that the timing of pre-synaptic activity (glutamate release) and of post-synaptic activity (post-synaptic-depolarisation) is critical in determining whether synaptic strength will be altered. Pre and post synaptic ‘events’ must occur within approximately 20 milliseconds, otherwise synaptic strength remains unchanged. This form of plasticity, known as Spike-Timing-Dependent-Plasticity (SDTP), is likely to become increasingly relevant as we begin to conceptualise ‘micro-circuit’ abnormalities in major neurodevelopmental disorders. Two final points about SDTP will be made here. Plasticity is bidirectional (potentiation or depression) depending on the order of pre and post-synaptic events. And conventional modulators such as dopamine can impact upon the timing rules and alter the direction of the plasticity, (LTP or LTD).
Some Psychiatry: The K-Hole and beyond
Ketamine, a drug that has attracted the attention of psychiatrists in the past few decades, ‘blocks’ the NMDA channel. It has been used as a model psychosis, and latterly has been demonstrated to have acute anti-depressant properties. (It certainly impairs new learning, as would be expected).
Downstream of NMDA blockade, there is no clear consensus as to how ketamine produces a psychosis. Counter-intuitively (for a glutamate antagonist), ketamine increases the excitability (spiking) of pyramidal neurons. Ketamine also increases the power of gamma band (~40 Hz oscillations) and some have proposed that ‘kernels’ of ‘abnormal’ gamma underlie the psychotic-like effect.
But the behavioural pharmacology of ketamine is far from straightforward. Rating-scales used in schizophrenia research, are probably not ideal for capturing the nuances of the drug. Those who have taken a more phenomenological approach [in the sense of ‘bracketing-out’ existing assumptions, whilst focussing on clear descriptions] have identified a much richer and more complex behavioural psychopharmacology, which includes euphoria, near-death experiences, the cessation of time, the dissolution of the ego, and the experience of being immersed in fractal geometries or boundless oneness (Jansen K, Ketamine: Dreams & Realities 2000).
Close observation reveals the dose-dependent emergence of an oneroid (dream-like) state, and other catatonic features (ambitendency, posturing) but not a classic paranoid psychosis. Researchers have also tended to assume that ketamine can ‘cause’ negative symptoms, but reports of euphoria, terror and awe are inconsistent with this categorisation. Motor output (which includes speech of course) is certainly restricted following ketamine, but because the concurrent inner world is a kaleidoscope of strange, mystical and fantastic experiences with extremes of emotion, the overall picture is far removed from the negative syndrome.
Nevertheless, ketamine is frequently championed as the most convincing drug-model of schizophrenia because it can induce negative symptoms, on a rating scale. The irony perhaps is that the ketamine experience might actually be more schizophrenia-like than many of its proponents have suggested. Ketamine elicits phenomena, which are now very rarely encountered in psychiatric clinics, given the modern-day domination of the softer, paranoid form of the illness.
Paul Janssen’s genius was in predicting that a drug which blocked the effects of amphetamine in animals, would be an effective treatment for those cases of schizophrenia that resembled an amphetamine psychosis (characterised by agitation, hallucinations and delusions)[link]. That drug was haloperidol, and that class of drug (D2 dopamine receptor antagonists) changed the landscape of psychiatry.
Janssen’s logic would also suggest that a drug which inhibited the effects of ketamine in animals, would be an effective treatment for those cases of schizophrenia which resemble ketamine-elicited psychopathology (characterised by bizarre, inaccessible dream-like states, and psychotic motor phenomena. i.e. cases where ECT becomes a sensible option). A pharmacological antagonist of ketamine (in animals) proved to be ineffective against human paranoid schizophrenia. Perhaps this could have been predicted, by closer attention to the phenomenology of ketamine. The question now is whether ‘The Lilly compound‘ has efficacy against non-paranoid schizophrenia?
New Brain Cells
In the last decade it has become clear that new cells can form in the adult brain. This happens in a region known as the hippocampal complex. The hippocampal complex is found deep inside either temple and is crucial for memory and emotion. The hippocampal complex inhibits the human stress response, but can itself be damaged by persistent stress, leading to a vicious cycle in which the stress response is amplified further and depression ensues.
Experimental work suggests that neurogenesis (the birth of new neurons) in the hippocampal complex is vital for the action of conventional antidepressant drugs. Exercise and enriched environments also promote neurogenesis, whilst stress has the opposite effect.The current picture is that hippocampal health (including the birth of new neurons) is essential for protecting the organism against the effects of stress, so that if hippocampal functioning is compromised, anxiety and depression can emerge.
There has been recent interest in the antidepressant properties of a natural molecule called curcumin. This substance is found in the herb turmeric. As well as a foodstuff, turmeric has been used for centuries in traditional Indian medicine (Ayurveda). In pre-clinical studies, curcumin exhibited clear antidepressant effects.
Research has focused on the mechanism of action of curcumin. Remarkably it appears that curcumin can also increase the birth of new neurons in the hippocampal complex. This is an intriguing finding which hints at the possibility of a new class of antidepressant drug.
A new paper from researchers at King’s College London provides an excellent summary of work in this area. The full paper can be read here.