This is the eighth in a series of Forum interviews with PRF science advisors.
Marshall Devor, PhD, has been a leader in the study of neuropathic pain mechanisms for over three decades. His pioneering experiments in rats introduced the idea that pain susceptibility is heritable, and led to the discovery of genes that regulate pain sensitivity and the risk for chronic pain. Since 1975 he has worked at the Hebrew University of Jerusalem in Israel, where he is now the Alpert Professor of Pain Research in the Institute of Life Sciences. He has chaired several large international congresses on pain and serves on the editorial boards of numerous journals. Devor spoke with Megan Talkington about the discoveries that first attracted him to neuroscience, the newest lines of research in his own lab, and the fundamental inconsistencies in current pain theory that keep him up at night. The following is an edited transcript of their conversation.
What is going on in your laboratory that you are most excited about?
I have a bunch of things going on, mostly having to do with neuropathic pain and its mechanisms. But the thing that I’m most excited about is something a little different. It has to do with the most powerful form of pain control that we know—general anesthesia.
With anesthetics, we’ve got a series of molecules that, at appropriate doses, completely block the most intense pain imaginable. They have other interesting effects as well, like making a person unconscious. Yet if you look in a textbook, it says nearly nothing about how they work. The most you will find is a discussion of the molecular receptors. People generally presume that anesthetic agents cause neurons everywhere in the nervous system to be less responsive until the brain gets to a state in which it simply can’t support whatever consciousness is.
But there is another option that has a precedent in pain research. We had morphine, which is a very good pain reliever and, like general anesthetics, has receptors all over the brain. John Liebeskind and others showed years ago that morphine, when given systemically at limited doses, acts on a small cluster of neurons in the ventrolateral part of the midbrain PAG [periaqueductal gray], and the remaining pain control is achieved by descending pathways, that is, by a neural network. I wondered: Is it conceivable that general anesthetics work that way, too?
We did the same sort of study that Liebeskind et al. did, which was to microinject tiny amounts of general anesthetic (pentobarbital was used at first and then we tried other things as well) throughout the brain [see Devor and Zalkind, 2001]. And we found one little spot—not too far from the PAG, but not in the PAG—where if you inject a tiny amount of anesthetic, as little as a thousandth of the systemic dose, the animal loses consciousness, loses righting reflex, and loses response to noxious stimuli.
This suggests that anesthetic molecules, at least the barbiturates, may be working on a small cluster of neurons, and that the four things they do—turning off pain, movement, memory, and consciousness—are done by neural networks. We are now trying to work out what the pathways are—the particular cells involved, where they send their axons, and what neurons elsewhere in the brain synapse on these cells. Does an individual cell in this area send its axon in a branched pattern to serve all four functions? Or is it possible that only certain cells are turning off pain, and different ones in the same area are turning off consciousness? In the latter case, maybe one could get the incredible quality of pain relief that general anesthetics deliver without the loss of consciousness and motor control.
This is something really new, which basically nobody else in the pain field has talked about. There are a handful of labs interested in anesthesia that also are thinking in terms of networks, and, of course, people in the sleep arena have been thinking of networks for a long time. These things are actually part of a single problem, but we’ll have to see how they converge as the work goes along.
If you think about it, understanding the morphine pathway also turned out to be really important for understanding other endogenous pain control effects. For example, certain kinds of placebo responses tap into the same circuitry. Here I am talking about a different pain control circuitry—a parallel one, perhaps even more powerful, that does not depend on opiate receptors and presumably uses a different pathway and different pharmacology. Mother Nature gave us opiates a couple of thousand years ago; who knows, maybe now we can come up ourselves with something that works by way of another endogenous control system.
Of course, I like my regular work on mechanisms of neuropathic pain as well. But you asked what I’m particularly excited about, and that is it.
Could you tell me a bit about your work on neuropathic pain?
One contrarian line I have been on for a long time concerns the classical notion that pain is solely the business of A- and C-fiber nociceptors. This idea is entrenched in all of our thinking, and an awful lot of current research deals with the physiology of nociceptors. When someone steps on your toe, the pain felt is certainly a result of nociceptor activation. And if you’ve got an ongoing, deep, burning pain, that’s probably also nociceptors. But what really bothers people with neuropathic pain is tactile allodynia, when light touch hurts. Even forgetting about neuropathic pain, if you have a sunburn, or if you were banging a nail and hit your finger, touching the bruised or slightly burned skin hurts. That’s where almost all everyday pain comes from.
If this pain had to do with C-fiber nociceptors, then it should take an entire second for the signal to get from your finger to your spinal cord, and there it still has to get to perception. But you know perfectly well that if you have a burned hand and you touch something, you feel pain instantly. There are many reasons to believe that rapidly propagating Aβ touch fibers, and not the slower C-fiber nociceptors, are carrying that pain signal, so I study A-fibers in relation to pain.
Here’s an experiment we published recently. It started with experiments that I did in the 1970s, when I wondered whether part of the variability in pain among different people might be genetically determined. Back then, I took outbred rats and found that some of them were prone to developing pain behavior after nerve injury, while some showed little pain after the same injury. We took the high-pain animals and bred them, and we took the low-pain animals and bred them, and then continued the selection process. After several generations, all of the offspring of the pain-prone animals were neuropathic pain-prone, and all of the animals of the pain-protected animals were pain-free.
We’ve gone on to do a lot of things with these animals, including looking for genes that are responsible for the differences in pain behavior. But our new work looks at differences in the response of sensory neurons to nerve injury. One of the dramatic things we’ve found is that the Aβ afferents start to make CGRP [calcitonin gene-related peptide]—which is thought to be perhaps the pain neurotransmitter—but that this only happens in pain-prone animals. In the pain-protected animals, the A-fibers don’t begin to make CGRP [Nitzan-Luques et al., 2011].
Normally, Aβ fibers sense touch; they don’t cause pain. But, if in the presence of neuropathy, injured Aβ afferents start making CGRP and releasing it centrally, the central sensitization that they generate could cause light touch to be painful. Moreover, since injured Aβ afferents tend to fire spontaneously from ectopic locations, this so-called phenotypic switching effect in Aβ afferents could also contribute to spontaneous pain.
Are there other places where you think the field needs a shakeup?
One thing that’s been bugging me is the concept of the pain matrix. We all love the pain matrix. The pain matrix is the group of regions in the brain that become activated in functional MRI or PET when a painful stimulus goes in. That’s all very satisfying—the cortex is presumably where consciousness resides, and now we have all these areas that process pain, so a pain matrix makes perfect sense.
However, as Penfield and others showed beginning in the 1950s, if you electrically stimulate the visual cortex, directly, you get a visual percept—you see flashes of light. If you stimulate the olfactory cortex, you experience smells. If you stimulate the auditory cortex, you hear sounds. Every now and again you actually get organized sounds, music, which is absolutely incredible. But Penfield and his followers showed that if you stimulate the somatosensory cortex, or the anterior cingulate cortex, or any of the other accessible areas in the pain matrix, you don’t get pain sensation. You get tinglings and paresthesias, but you almost never get pain. It began to dawn on me that there may be a real crisis in our understanding. The pain matrix doesn’t meet the fundamental requirement that when its cells are activated, the individual experiences a pain percept.
Sometimes people say, “Yes, but pain is much more complicated than vision and Mozart.” That sounds very suspicious. And then they say that, really, you need to get the entire matrix going—it’s not enough just to stimulate at one point. But consider epileptic seizures, where you have very large areas of cortex firing. Before an actual seizure, there is often a pre-sensation, called an aura. This is due to the onset of cortical firing. Smell is a very common aura, as are vision and tingling and all sorts of other sensory experiences. But it is exceedingly rare for anybody to report pain as a prelude to a seizure, even though all the areas in the pain matrix fire like crazy. In fact, the pain matrix areas are particularly prone to epileptic seizures.
Then, people say that it’s not enough that lots of different areas in the matrix go off. You’ve got to have precisely timed activity, and you have to get the pattern of activity in all the different pain matrix areas right. I can’t deny this is a logical possibility. But if that were true, then if you messed up any of these areas with a lesion, you’d lose the detailed coordination and you’d expect the patient to be blind to pain. But of course, that never happens. If anything, cortical lesions are a cause of chronic central pain. So you see, we’ve got a problem here!
What’s the solution?
This is one of the things that keeps me up at night. We say we’re making progress toward understanding pain, and yet we don’t even have P1. We don’t have a primary pain cortex. The only real candidate for P1 is the insular cortex. But so far, the data indicate that only a small minority of insular stimulation trials evoke pain, that insular seizures only rarely have a painful aura, and that insular lesions do not reliably cause analgesia. Also, most of the data, such as they are, come from one lab—François Mauguière’s group in Lyon, France. We’ll have to wait and see how this works out.
But what if pain perception really is primitive? What if it occurs subcortically? You can certainly stimulate places subcortically and evoke pain. Maybe the cortex isn’t necessary for pain experience at all, and this whole pain matrix thing is a misconception! I’m just playing devil’s advocate, of course.
Now, I don't doubt for a second that information on pain reaches the cortical pain matrix areas. For example, if you are planning for the future, something the prefrontal cortex presumably does, it’s good to know that something hurt. That way you can try to avoid it the next time. But the cells whose activity is pain, it seems to me, are not in the cortex. This is a problem. Because if pain isn’t in the cortex, and pain requires consciousness, that means that consciousness is also not in the cortex. This is an even more heretical statement to make. But it follows from the fundamental observations.
You mentioned some of the groundbreaking work you have done on pain genetics. Where did you get the original idea to look for genetic determinants of pain susceptibility?
It came from a professor here at the Hebrew University named Israel Lieblich, who had worked as a postdoc at Caltech with Jim Olds. In the 1950s, Olds and Milner discovered pleasure centers in the brain. They showed that you can stick an electrode into a rat’s brain, and if you get it in the right place, the animal likes the stimulation. This is so counterintuitive. The reasonable thing to expect from an electrode passing 60 Hertz into a limbic area would be an epileptic seizure. But, in fact, they got pleasure. The animals would learn to press a bar to get the electrical stimulation. And if you asked them to go into a wheel and run five times and then climb a ladder and turn left to get the electric stimulation—they would learn that. They would learn anything you wanted if you offered electrical stimulation in the brain as a reward—it was much better than if you starved them and then offered food as a reward. The phenomenon is incredible. It’s one of those phenomena that brought me into neuroscience in the first place.
In any event, in the 1960s and 1970s, there was a lot of interesting research trying to get some handle on the mechanism of this thing, and then people kind of ran out of ideas. But Lieblich came up with a good idea. He knew that some of the rats were better self-stimulators than others. He took the high self-stimulators and mated them, and the low self-stimulators and mated them, and then kept doing selective breeding for high or low self-stimulation.
When I first came to the Hebrew University in 1975, Israel had already done seven or eight generations of selection and had a very convincing separation of high and low self-stimulator rats. His idea was to look to see what in the brain was co-selected with self-stimulation. Was there some neurotransmitter system, or some structure that was big in one group and small in the other? What correlated with this selection pressure for pleasure? I thought this was so clever. I had worked on the self-stimulation phenomenon myself as an undergraduate—I knew the literature and knew this was an entirely new idea. Then Israel died suddenly of a heart attack at the age of 48, and the project ended.
Before Israel died, I suggested we look at pain in the animals that were selected for high and low pleasure and see if there was a difference between the strains. And by God, there was! The high-pleasure strains were also high-pain ones [Inbal et al., 1980]. That was really the first indication in animals that there is a genetics of pain. It also indicates some sort of deep connection between self-stimulation—the pleasure thing—and pain, in terms of circuitry. I then went on to do my own selection for pain behavior, which is really based on the one Israel did for self-stimulation.
Moving back to the present, what do you see as the way forward in pain genetics?
We recently had the first conference devoted to pain genetics [the 10th IASP Research Symposium, The Genetics of Pain: Science, Medicine and Drug Development, held in Miami Beach, Florida, US, 7-9 February 2012]. They gave me the opening lecture because they felt that I was the first one to bring genetics into the pain field, which to a certain extent I guess is true. The fact of individual variability in pain, however, was very obvious long before I was born. It was logical to think that at least part of the variability could be genetic. Now the evidence is pretty much overwhelming—so what should we do next?
What people mean by pain genetics nowadays is trying to find polymorphisms—common genetic variants among people that account for differences in pain. Another thing that is very popular is making a pain model in rodents by injuring a nerve, and looking at changes in gene expression in sensory neurons or the spinal cord. The trouble with both of those approaches is that too much happens. There are too many genes with small effects that come together to cause pain-free disposition. And if you injure the nerve, then two, three, or maybe 4,000 genes are significantly up- or downregulated. That’s 20 percent of the entire genome!
What has happened is that each person latches onto a gene whose expression has changed and says, “Ah, this is my career; this is what I’m going to write my papers on.” And since we don’t have thousands of investigators, but we do have thousands of gene candidates, lots of genes get shortchanged—they are the wallflowers. Nobody dances with them. We’ve got this crazy situation where individual labs promote particular genes and pathways, and the genes that are lucky enough to have a strong champion get all the attention.
We need strategies to overcome this bias. To give you an example, those rats that I developed that have high and low predisposition to pain have a common genetic background. If you take the pain-prone animals and cut their nerves, then 3,000 genes are up- or downregulated. And if you take the pain-protected animals and cut their nerves, 3,000 genes are up- or downregulated. But you can look at the regulation in the highs, compare that to the regulation in the lows, and see which genes are differentially regulated. In this way, you filter out all the genes that have to do with regeneration or apoptosis or thermoregulation—all sorts of things that happen in a nerve injury model that have nothing to do with pain.
Another good strategy is to take inbred mice and line them up based on a given pain phenotype, and then find the genes whose level of regulation across strains correlates with the pain phenotypes. That should give you some information about which genes are related to pain and not to something else.
I am not trying to point to particular strategies, but rather to the need to do strategic thinking. How do you sift through the large number of candidate genes and candidate hypotheses? Biological systems in general are terribly complicated. And here we are talking not just about an ordinary biological system, but a biological system that has components from receptors in the skin to cognitive and emotional functions. Our system is modulated right from the bottom all the way to the top. Strategic thinking is going to be necessary to move us forward.
You don’t know a priori which genes are going to be the interesting ones. Say that regulation of some gene has a beautiful correlation with pain across strains, and it happens to be a gene that is involved in connective tissue. You can’t throw it out just because it’s not part of our current pain physiology. We don't know enough to be sure that it’s not related to the pain physiology of the future. To look at a list of 500 genes and say, “Yes, there’s the one; it’s the receptor for a blockbuster drug” is not strategic thinking. Our current pain physiology is primitive. We are learning new things all the time.
Thank you very much for sharing your ideas.
It was really nice talking to you.
PRF Related Content:
News: Progress in Pain Genetics: A Meeting of Their Own (14 Mar 2012)
News: Sex and Drugs and Stress and Genetic Variation: Very Complicated Indeed (28 Oct 2011) (see Devor’s comment at end of story)
View Marshall Devor’s profile on Pain Research Forum (requires member log in)
Nitzan-Luques A, Devor M, Tal M. Genotype-selective phenotypic switch in primary afferent neurons contributes to neuropathic pain. Pain. 2011 Oct;152(10):2413-26.
Devor M, Zalkind V. Reversible analgesia, atonia, and loss of consciousness on bilateral intracerebral microinjection of pentobarbital. Pain. 2001 Oct;94(1):101-12.
Inbal R, Devor M, Tuchendler O, Lieblich I. Autotomy following nerve injury: genetic factors in the development of chronic pain. Pain. 1980 Dec;9(3):327-37.
Other Forum Interviews:
Pain and Its Control: A Conversation with Allan Basbaum (6 June 2012)