Even as NaV1.7 blockers were (and are) being developed, we were also taking another approach: using the human genome to predict whether an existing medication would help a particular person. This approach, variously called “precision medicine,” “personalized medicine,” or “individualized medicine,” uses the DNA of each specific patient to give the clinician a molecular compass that points to the most effective medication. Currently we do not have that molecular compass in pain medicine. The clinician usually begins by selecting a particular medication, based on the patient’s description of the pain, the cause of the pain or its pattern, and other aspects of the patient’s history. The most effective medication, the dosage, and the dosing schedule can vary from patient to patient. The choice of medication, the dose, and the dosing schedule are usually adjusted based on the patient’s reports of pain relief and side effects, in a process that requires many visits to, or telephone consultations with, the clinic. There are many forms and dosages of existing medications, and many potential combinations, so in any single patient only a subset of the treatment regimens can be tried. Even when based on the best knowledge available, the process includes a degree of trial-and-error and may provide less-than-optimal relief.
Many factors—including differences in the type of injury or disease process responsible for the pain, differences in other medications that the patient may be taking, differences in diet, and environmental factors—can contribute to variability in the responses of different people to medication. In the past this added up to a sense that a given patient’s response to a particular pain medication was unpredictable.
In most people with inherited erythromelalgia, pain medications other than opiates provide no relief or only minimal relief, and even opiates provide only partial relief. Yet there are exceptions to every rule. In 2009 we encountered one unusual family with inherited erythromelalgia, in whom the disorder was responsive to treatment with a medication called carbamazepine. This family was a reminder that molecular genetic analysis of very rare families can be informative.
Developed in the 1960s, carbamazepine is a nonspecific sodium channel blocker. Rather than blocking the activity of a single subtype of sodium channel, it inhibits all sodium channel subtypes. It exerts its blocking effect in an activity-dependent manner, inhibiting the channels when they are active. This medication was initially used as a treatment for epilepsy and is still used in that way. It has also proven effective in the treatment of a particular form of neuropathic pain, trigeminal neuralgia, in which patients experience attacks of severe, lancinating pain in the face. But carbamazepine is not usually effective in other forms of neuropathic pain. And, in most patients with inherited erythromelalgia, carbamazepine is not helpful. The family we encountered in 2009 was different. In this unusual family carbamazepine provided significant pain relief. The initial patient identified in this family (called the “proband” by geneticists) suffered from erythromelalgia with symptoms that began in infancy. Carbamazepine provided significant pain relief. Pain in the proband’s two children was also relieved by carbamazepine. One child had approximately 56 pain attacks per week before starting carbamazepine, but only two attacks per week while being treated with this medication. The other child had a similar response. These children could not tolerate wearing socks or shoes and could not participate in athletics prior to treatment. But they were able to wear socks and shoes, to run, and to play soccer when treated.
What was this unusual family trying to tell us? Most people with inherited erythromelalgia are not helped by carbamazepine. Yet, in this unique family, carbamazepine was remarkably helpful. Their genes, we reasoned, were playing an important role in shaping their response to carbamazepine. In 2011, we made a bet that we could predict, on the basis of analysis of DNA, that a particular patient would be more, or less, responsive to a particular medication. The paper that follows (Yang et al. 2012) represents an early step toward a genomically guided “first time around” approach that will select, for each person, the most effective medication for pain.
In our initial study on the DNA from this family (Fischer et al. 2009) we found a mutation of NaV1.7, V400M, which replaced a single amino acid, valine at position 400 (V400), with methionine. Voltage-clamp analysis showed us that the mutation made it easier to activate the channel, much like other inherited erythromelalgia mutations. The mutation also had an unusual pro-excitatory effect, impairing inactivation of the channel. Each of these changes would be expected to increase the excitability of DRG neurons, thereby producing pain.
To determine why carbamazepine relieved pain in this unusual family, we next examined the effect of carbamazepine on the V400M mutant channels. We observed that exposure to carbamazepine restored activation of the mutant channels to close-to-normal levels, while not affecting activation of normal, wild-type channels. These results showed us that the V400M mutation had multiple effects on the channel: First, the mutation produced gain-of-function changes in the NaV1.7 channel, thereby leading to overactivity in pain-signaling neurons and to pain. Second, and unexpectedly, the mutation sensitized the channel to carbamazepine.
Each gene within our genome has many variants, even in people with no history of disease. Our studies on the V400M family showed us that there are variants within the SCN9A gene—small changes that substitute one out of the nearly 1,800 amino acids that make up the NaV1.7 sodium channel—that alter sensitivity of the channel to a medication. Hundreds of variants—amino acid substitutions at specific sites within the channel—had been reported within various databases for the SCN9A gene; most had not yet been functionally studied so that their clinical implications were not yet understood. As we talked about the V400M mutation, we wanted to ask this question: Might other amino acid substitutions decrease or increase sensitivity of the channel to particular medications, as the V400M mutation did for carbamazepine? Might genomic variability, of the type that occurs in each of us, be used to guide the treatment of pain?
A sodium channel is a complicated molecule: nearly 1,800 amino acids, strung together like a string of beads folded into a highly specific three-dimensional configuration. Figuring out the configuration of the protein in three dimensions, or solving the question “what is the precise pattern of folding?” can be formidable. Imagine folding up a long string of beads by crunching it up your hand. There are a multitude of folding configurations, sometimes with red bead “A” coming into close proximity to blue bead “B” located thirty beads down the line, but sometimes not.
Even our bacterial ancestors contain sodium channels. All four domains of the bacterial sodium channel share an identical structure, in contrast to mammalian sodium channels where the four domains have different structures. So, the bacterial channel provides an experimental model that is amenable to crystallographic analysis, which can provide information about the relative locations of the amino acids that make up a protein, even a protein as complex as an ion channel. Crystallography can even indicate the location of atoms within a large molecule. In 2011, William Catterall and his colleagues at the University of Washington, including crystallographer Jian Payandeh, reported the crystal structure of the bacterial sodium channel (Payandeh et al. 2011). This was an important advance; it showed how the sodium channel protein bends and folds in three dimensions, thus revealing how individual amino acids within the bacterial channel come into close contact with each other, and even signaled the locations of some of the atoms within these amino acids, inside the folded channel protein.
A few days after this report appeared, Yang Yang, a new research fellow in our group, knocked on my office door and—with his usual “we can do anything if we try” demeanor—suggested that we construct a structural model, with atomic-level resolution, of the human NaV1.7 channel. His plan was to base this on the crystal structure of the bacterial channel. Tasked with doing this, he used powerful computerized algorithms to individually model each of the four transmembrane domains of the human channel, and then aligned the four domains to a structural template provided by the recently solved bacterial channel. Because the sodium channel protein is a large molecule, and weaves in and out of the cell membrane twenty-four times, there are many thousands of potential solutions to the problem, each depicting a slightly different alignment of the channel’s amino acids within three-dimensional space. Yang, again using computer simulations, identified the best solution—the model with the lowest free energy. This gave us a picture of the three-dimensional structure of the folded string of amino acids within the human NaV1.7 channel in a stable, biologically plausible state. We now had a model, included in Yang et al. (2012), which showed the locations of specific amino acids and some of the atoms within them in the human NaV1.7 channel.
A next step was to use our NaV1.7 model to predict the actions of drugs on specific variants on NaV1.7. The carbamazepine-sensitive V400M mutation provided a starting point. Building upon Yang’s success with structural modeling, and reasoning that the location of the carbamazepine-sensitizing V400M substitution within the folded NaV1.7 channel might be critical for the responsiveness of this channel variant to various medications, we reviewed our database of NaV1.7 mutations. We then substituted various amino acids within the modeled channel to model these mutations, using the V400M substitution as a “seed” to search for other naturally occurring amino acid substitutions that might heighten the channel’s response to carbamazepine.
Our analysis pointed to another inherited erythromelalgia mutation, S241T, as a candidate that might show increased responsiveness to carbamazepine. The site of the substituted amino acid in the S241T mutant channel is located 159 amino acids away from V400M in the linear, unfolded sequence of the channel, which is a long distance. But from our structural modeling we knew that S241T was located only 2.4 Ångstroms (Å)—less than the diameter of a water molecule from the V400M residue—within the folded three-dimensional structure. Beads 241 and 400 in the “necklace” of amino acids of NaV1.7 were located right next to each other in the scrunched-up, folded configuration of the channel. To see whether this might permit us to predict the effect of carbamazepine on the S241T mutant, we next asked whether the relatively close locations of the V400M and S241T mutations within the folded channel structure reflected their working together in a coupled way during activation of the channel. Our thermodynamic analysis showed that, indeed, the V400M and S241T mutations were energetically coupled, working in tandem to effect channel activation through the same or a similar mechanism.
Based on these results, we hypothesized that the atomic proximity of these two mutations, and their mechanistic coupling, might be paralleled by pharmacological coupling. If this were the case, the two variant channels would be expected to share pharmacoresponsiveness to carbamazepine. To test this prediction, we assessed the effect of carbamazepine on the excitability of DRG neurons which we transfected with the S241T mutant channels in tissue culture, and compared the results with those obtained for cells expressing normal, wild-type NaV1.7 channels. As shown in figure 5a–c of Yang et al. (Yang et al. 2012), our analysis showed that exposure to carbamazepine resulted in a doubling of current threshold for DRG neurons expressing the S241T mutant channel, making it harder to fire these cells. And carbamazepine dramatically reduced the number of action potentials produced by DRG neurons expressing the S241T mutant channel. In the aggregate, these observations demonstrated that carbamazepine can attenuate the hyperexcitability of pain-signaling neurons expressing the S241T mutant channel, as predicted by the structural modeling.
These results showed, in DRG neurons carrying the human NaV1.7 channel in tissue culture, that structural modeling and thermodynamic analysis could predict the response of variant sodium channels to a pharmacotherapeutic agent. We published these results in 2012 (Yang et al. 2012). We were not yet there, but we hoped that this study would move us closer to personalized, genomically guided treatment for chronic pain.
A next step would be to study this in humans, to see whether carbamazepine would actually relieve pain in human subjects with the S241T mutation. That study (Geha et al. 2016) is described in chapter 13.
It took four years to do.