In vivo Electrophysiology
(and why I bother)
As a scientist my life has been occupied by trying to quantify things. How do you approach something that is subjective? When it comes to pain, the only measure that should count is what someone tells you. I find one of the biggest challenges of being a pre-clinical pain researcher is what do you measure? Pain is defined as being both a sensory and emotional experience; in this respect all current measures in non-verbal beings feel somehow inadequate. Ultimately, we are restricted to measuring one aspect of the pain experience at a time, often by inference. Typically these encompass reflexive withdrawals, facial grimaces, open-field behaviour, conditioned place preference, calcium imaging and electrophysiology. Before I continue and you consider sending an angry rebuttal, I don't dislike behavioural testing! As with all endpoints it's the interpretation that matters and there is a limitation to what any one assay can tell you. In some ways it's like trying to understand how a bike works by looking at a collection of it's parts. My major issue with reflexive endpoints is that it tells you about one single point on the stimulus-response curve; i.e. the first point at which a motor response is elicited. Is it much different to snapping your fingers, measuring the preyer reflex and claiming to have measured auditory function? It's a bit of a blunt tool. Measuring a change in the withdrawal latency or threshold is one aspect of pain processing, too many papers extrapolate that to mean 'X is critical for pain'. As a field, the study of pain needs to evolve towards a multi-faceted approach encompassing both sensory and emotional dimensions (and I include myself in that). We need better ways of determining spontaneous pain as this is what troubles patients the most.
I am a self-professed ephys nerd. For 10 years I have been more than content to sit at my rig as a one-man data factory churning out dataset after dataset. I have to admit, there is still something strangely satisfying watching a neurone fire in real time to natural stimulation. Am I alone in finding the squeal of a neurone and subsequent afterfiring slightly exhilarating? It might be difficult to gauge from any one of the Dickenson lab papers what we have been trying to achieve, so I will take a moment to try and put everything together (we recently wrote this review which explains it a bit).
When talking to non-scientists, the question of 'what do you do?' is typically followed by 'how does that tell you about pain in humans?' If you struggle to answer the second question, it might be a good time to re-evaluate what you do and why! Let’s consider in vivo electrophysiology for now. I've really enjoyed recording within physiological systems. From a technical point of view there are some things that can only be done in vitro or in slice and these continue to be invaluable to deciphering mechanisms. It’s this latter point that is central to what we were trying to achieve (more on this later). Our lab has primarily focused on recording from polymodal wide dynamic range neurones within the spinothalamic pathway (spinal and thalamic sites). Ron Dubner and colleagues (1, 2), in a series of elegant studies, demonstrated how the fine-tuned coding of these neurones corresponded to the ability to discriminate small changes in stimulus intensity (I'm sure these studies wouldn't be sanctioned now). In the Dickenson lab, a parallel human and rat study corroborated these findings by delivering infrared laser stimuli, looking at how wide dynamic range and nociceptive neurones encoded features such as intensity coding and spatial summation, and how these corresponded to pain ratings in human subjects (3) (the late Donald Price writes an eloquent opinion piece here (4)). These neurones also exhibit wind-up and post-discharges (see video below) that are relevant to understanding temporal summation and second pain/aftersensations in chronic pain states. A second study from the Dickenson lab, also a parallel rat and human study, delivered UVB irradiation to rats and pain-free volunteers. Wide dynamic range neurones exhibited features of central sensitisation in a manner that corresponded to human pain thresholds (5). We are also fortunate that single unit recordings have been obtained from the sensory thalamus of patients undergoing stereotaxic procedures for other conditions (6). From these we can see that polymodal neurones exist that encode sensory modalities and stimulus intensity, but microstimulation at these recording sites also elicit natural sensations. The characteristics of these neurones are remarkably similar to other species such as cat, rodent and primate.
An example of wind-up in the mouse deep dorsal horn. A train of 16 electrical stimuli are delivered at 3x the threshold for activating C-fibres at a frequency of 0.5 Hz.
Note the progressive increase in excitability following repetitive stimulation and the prolonged post-discharges towards the end of the train. Wind-up has been interpreted as the capacity to amplify nociceptive signals and shares many features with central sensitisation.
It's old school electrophysiology but still useful. Hopefully I have convinced you of the merit of studying wide dynamic range neuronal activity to help understand the sensory component of pain in normal and chronic pain states. The characteristics of these neurones are conserved between species, they form part of a critical sensory relay, they encode stimulus intensity, and they exhibit features of sensitisation and excitability that mirrors human pain measures. How can we utilise these features to understand pain mechanisms and drug actions? Something that has led to soul searching across the field of neuroscience is translation and the pain field is no exception. We are not completely devoid of success stories of bench-to-bedside translation (e.g. anti-NGF therapies for osteoarthritis, conotoxin for refractory neuropathic pain, anti-CGRP therapies for migraine) but these are definitely in the minority. Some will question whether pain research should be performed in animals at all. I would counter that by arguing that many features of sensory processing are conserved across species and that drugs we know to be analgesic in humans work in animals as well. It has been proposed that translation could be improved by identifying what works in patients and then studying the underlying mechanisms in rodents (e.g. as has been the case for neuromodulation and botulinum toxin). One of the biggest challenges is the homogeneity of phenotypes and mechanisms in rodent models compared with the quite obvious heterogeneity of symptoms we see in patients. This has led to calls by many to move towards personalised medicine and mechanism-based treatment selection. The basic premise is that patterns of sensory symptoms reflect surrogate markers of pain mechanisms and thus relate to specific drug sensitivity (7). Sounds beautifully simple! But ultimately it is very difficult to identify a mechanism successfully in any given patient as different mechanisms could lead to identical symptoms. Stratification along these lines improves treatment selection when you apply it to large patient cohorts but it’s not perfect and treatment algorithms will likely need to be refined and multifactorial.
Two of the most reliable measures of pain mechanisms that can be identified in patients are temporal summation (enhanced spinal amplification) and conditioned pain modulation (decreased descending pain control). As it happens these are both features that can be measured in spinal neurones! NMDA antagonists block both neuronal wind-up and temporal summation of pain, and drug potency is increased in neuropathic rats. There is also evidence that patients with enhanced temporal summation can benefit from ketamine treatment (8). It stands to reason that drugs that are capable of inhibiting neuronal wind-up in rodent models may benefit patients with exacerbated temporal summation. I have worked on a few studies for a collaborator in Oxford demonstrating how ion channel gene deletion can affect this measure potentially identifying novel targets for drug discovery.
Conditioned pain modulation (CPM) is a test based on the knowledge that 'pain inhibits pain.' Applying two distant noxious stimuli recruits top-down inhibitory signalling pathways that supress the pain intensity of the applied stimuli. Diffuse noxious inhibitory controls (DNIC) are the analogous process in rodents and is considered the key neural substrate to CPM (see below). However with CPM there are many other considerations as the subject will be influenced by cognitive factors such as stress, anticipation, expectation etc, hence the distinction between CPM and DNIC. The loss of CPM in many neuropathic patients is hypothesised to represent reduced descending inhibitory signalling and this can be restored in these patients by drugs that block noradrenaline reuptake (e.g. duloxetine (9) and tapentadol (10)). These are studies we have successfully back-translated. I have shown how the balance of serotonergic (11) and noradrenergic (12) modulation of nociception shifts across stimulus intensities and sensory modalities in neuropathic rats. We have seen that this altered activity leads to a loss of DNIC in neuropathic rats, mirroring observations in some patients. DNIC can be restored in these animals by tapentadol and reboxetine (NRI), demonstrating that the restoration of DNIC is reliant on the noradrenergic system (13). But we have also shown that reducing facilitatory drive alone can restore DNIC. These findings suggest that targeting mechanisms/pathways that reduce descending facilitation and/or augment descending inhibition can restore DNIC (and presumably CPM). It might be expected that treatments that restore deficient DNIC in rodents could benefit patients with low CPM.
Diffuse noxious inhibitory controls
In rats, DNIC can be recruited by applying a heterotopic noxious stimulus. While recording from wide dynamic range spinal neurones under anaesthesia, a noxious von Frey (60g) is applied to the hind paw and the response quantified. This is then repeated with a clamp applied to the ear (CS – conditioning stimulus). In the sham animals, we can see that DNIC are active and reduces the neuronal response. The effect of this concurrent 10s conditioning is quite short lived and we can see the response returns to normal within a minute (and the test can be repeated). In the neuropathic animals (SNL), the ear clamp no longer has the same effect and DNIC are absent.
Another project I have worked on was also based on quantitative sensory testing (QST)-led observations in patients. Earlier I mentioned that patients stratified according to their sensory profile may reflect different pain mechanisms. It has been hypothesised that patients with small fibre neuropathy and severe mechanical hyperalgesia have pain driven predominantly by central mechanisms, whereas those with intact fibre function and thermal hyperalgesia have pain driven largely by peripheral mechanisms (I should point out here there is a sensory loss group as well and that individual patients can exhibit combinations of these phenotypes in adjacent areas). There is some evidence that the first group may be more responsive to pregabalin (14), and the second group to sodium channel blockers (15, 16). We wanted to use our electrophysiological recordings to perform a rat version of QST. We can apply many of the same stimuli used during QST and quantify responses in a way that some behavioural experiments aren’t amenable to. We’ve also seen earlier how the characteristics of these neurones correspond to QST measures. I found that the rat sensory profile looks a lot like the thermal phenotype (17). Using local lidocaine, I showed in the spinal nerve ligation model of neuropathy that ongoing activity in peripheral neurones drives central neuronal firing in ascending pathways. This activity was also sensitive to inhibition by systemic oxcarbazepine and local licarbazepine. So this neuropathic model seems to share some drug sensitivity and features with the thermal sub-group. It may help explain why oxcarbazepine and similar acting drugs have had no overall effect in clinical trials where patients are grouped by aetiology, but there is improved efficacy once patients are stratified.
I have also shown how pregabalin affects these neuronal responses and it seems to inhibit brush and punctate mechanical responses much more than heat evoked responses (18). Further work in the Dickenson and Porreca labs has shed light on how gabapentin relieves ongoing pain through central mechanisms (19). We see here how the homogeneity of mechanisms in rodent models complicates translation to patients. Both systemic pregabalin and oxcarbazepine have anti-nociceptive effects in this model (and may look equally effective if the only endpoint was a reflexive one). However, we can see clear differences in the neuronal responses they inhibit and sites of action. Hopefully it emphasises how important it is to dissect out drug mechanisms and how studying neuronal activity can help.
My current work with my good friend Kirsty involves combining optogenetic, electrophysiological and behavioural approaches to identify top-down circuits that mediate descending control of pain. I have also been working on a parallel rat and human study investigating the interaction between DNIC/wind-up and CPM/temporal summation (I'll provide a link to the data soon).