Progress in Pain Genetics: Human Studies Move Ahead
This is Part 2 of a four-part report on selected talks from the 10th IASP Research Symposium, The Genetics of Pain: Science, Medicine and Drug Development, held 7-9 February 2012 in Miami, Florida, US. See also Part 1, Part 3, and Part 4, or download a PDF of the entire report.
Mice and rats led off in Miami, but before long, talk turned to human genetic studies. In contrast to the large body of mouse work, human studies are still in their infancy, most agreed. Numerous challenges—recruiting patient cohorts, defining phenotypes, accounting for clinical and genetic heterogeneity, finding funding for large genetic studies for pain—all affect the pace of discovery. Getting to see the big picture may be some time away, but bits and pieces are coming into focus.
Inna Belfer, University of Pittsburgh, US, reviewed work on the best established and most published pain gene, catechol-O-methyltransferase (COMT). First studied in 2003, COMT is now the subject of more than 100 publications relating to pain (reviewed in Belfer and Segall, 2011). Belfer reviewed what is known about the gene, and presented some new data suggesting that COMT variants play different roles in nociceptive versus neuropathic pain.
The COMT protein metabolizes catecholamine neurotransmitters including dopamine, epinephrine, and norepinephrine. As a terminator of neural signaling, COMT has been implicated in mood, cognition, stress response, and pain. COMT genetic variation involves over 28 common single nucleotide polymorphisms (SNPs) that are inherited in complex patterns. For one highly studied missense variant, Val158Met (or SNP identifier rs4680), the Met allele is associated with decreased thermal stability and enzymatic activity of COMT, and thus higher levels of dopamine and other neurotransmitters. The Met allele has been studied extensively for its effect on dopamine-related cognitive phenotypes, including mental processing speed, mood, and schizophrenia.
Belfer’s previous work established that four SNPs in COMT define three haplotypes that account for low, medium, and high sensitivity to experimental pain, differences in endogenous pain control via the μ-opioid receptor, and susceptibility to common musculoskeletal pain conditions. High pain risk is linked to extremely low enzyme activity, and average and low pain risks are associated with higher activity. Interestingly, both high and low pain haplotypes contain the Val158 variant, but differ at other SNPs that influence mRNA structure and protein expression levels. This shows that evaluating haplotypes, rather than individual SNPs, is necessary to understand functional variation in genes.
Dopamine, epinephrine, and norepinephrine have both pro- and anti-nociceptive actions, and so the exact pathways by which variations in COMT enzyme activity affect pain are complex and mostly unknown. Belfer pointed to two negative association studies (Armero et al., 2005; Mylius et al., 2010) that got her thinking about a possible dichotomy in catecholamine signaling in nociceptive and neuropathic pain. High-activity COMT variants protect against nociceptive pain, but their potential to lower epinephrine and norepinephrine in the spinal cord could lead to less anti-nociceptive signaling through β-adrenergic receptors and more pain. Could high-activity COMT alleles be a risk factor for spinal cord or neuropathic pain?
To investigate this idea, Belfer and colleagues did a post-hoc analysis of COMT variants in 11 common strains of mice that had been phenotyped previously (Mogil et al., 1999). Strains of mice vary in their COMT activity, and this has been shown to be associated with pain sensitivity (Segall et al., 2010). Belfer presented unpublished data suggesting that, across the strains, COMT haplotypes were differentially associated with nociceptive and neuropathic pain modalities. A low COMT haplotype was a risk factor for nociceptive pain, and high COMT was protective, as observed previously in human studies. However, in the mice, high COMT alleles conveyed a risk for neuropathic pain. The mechanism likely involves signaling through the adrenergic receptor, but dopamine signaling or effects on the immune system may also play a part.
Intermediate phenotypes: breaking it down
William Maixner, University of North Carolina, Chapel Hill, US, presented results from the genetics portion of OPPERA, the largest prospective human study of chronic pain and its risk factors to date (see PRF related news story). OPPERA (Orofacial Pain: Prospective Evaluation and Risk Assessment) is following 3,276 initially pain-free adults for seven years, watching for onset of temporomandibular (jaw) pain. Data were recently published from a case-control study within OPPERA that analyzed 3,295 SNPs in 358 candidate genes in 348 cases and 1,612 controls (Smith et al., 2011). In the study, no SNP exceeded the statistical threshold for significance, although previously reported associations of the COMT and HTR2A genes were supported. Suggestive associations were seen for five genes (NR3C1, CAMK4, CHRM2, IFRD1, and GRK5) not implicated before in risk of pain or nociceptive pathways.
The data, Maixner said, “may not be as rosy as we would like them to be, and the interpretation is problematic.” In pain, as in other fields, genomewide association results have not been robust or reliable. Many findings point to SNPs or genes without known function, and with low odds ratios that explain only a very small portion of the trait variation. One problem, Maixner said, is that SNPs are often treated as a univariate phenomenon, and researchers need to start looking at gene-gene interactions and using pathway analysis and other computational methods.
Another complication of association studies in pain and other conditions is that phenotypes such as case status (whether subjects have a disease or not) are complex. When people talk about “pain genes,” Maixner said, they use the term “pain” to mean a favorite phenotype or disease of interest. But each gene determines only a small part of the overall complex phenotype. That is where endophenotypes, or intermediate phenotypes, come in.
For example, underlying a persistent pain condition such as temporomandibular joint disease and contributing to its complex phenotype are signs and symptoms that can be measured individually. These include mood, anxiety, depression, tissue injury, state of pain regulatory systems, proinflammatory state, and others. These more defined endophenotypes or intermediate phenotypes are each affected by genes that may have a dramatic effect on the endophenotype, despite having a small effect on the overall complex phenotype. So, by defining endophenotypes that are more proximal to genes, it should be possible to better identify genetic associations with complex pain conditions.
Not all patients are created equal
One way to “clean up” phenotypes, and discover endophenotypes, is by dividing heterogeneous populations into homogeneous subgroups. To do that, Maixner is using statistical clustering techniques to group patients based on detailed characterizations including quantitative sensory testing, and psychological and physiological measures. Using a cohort of 199 chronic temporomandibular disorder (TMD) patients and 200 controls, the analysis identified three clusters: One was mostly controls, the second had an approximately sevenfold increased chance of having TMD, and the third had a 60-fold increased chance of TMD. The clusters had different phenotypic profiles: The second cluster had more pressure pain sensitivity than the first, and the third had more psychological distress, autonomic differences, and pain sensitivity relative to the other two. Patients in cluster three had more severe pain clinically, more comorbid pain conditions, and more headaches, indicating that the clusters defined a gradation of severity.
When the analysis was applied to the OPPERA cohort, Maixner and colleagues attained similar results. If a patient was in cluster three at baseline, he or she was 2.5 times more likely to develop TMD, and 85 percent of the TMD patients fell into cluster three. The researchers are now looking at genetic profiles within the clusters.
Even when individual polymorphisms do not achieve statistically significant scores in association studies, they can suggest hypotheses for further testing. When Maixner and colleagues compared 129 people with chronic TMD and 231 very healthy “supercontrols,” they found no hits that attained significance on their own. However, two of the top three genes were for proteins involved in the epidermal growth factor (EGF) signaling pathway, namely, the EGF receptor (EGFR) and one of its ligands, epiregulin (EREG). Maixner showed unpublished data from a collaboration with Jeffrey Mogil indicating that EGF and the EGFR are involved in producing acute and inflammatory pain, but not neuropathic pain, after nerve injury in mice.
Continuing the theme of pathways, Maixner’s colleague at the University of North Carolina, Luda Diatchenko, presented new data on pathway and phenotype analysis of the OPPERA case-control cohort genetic data. To better detect genetic associations, Diatchenko divided patients into two phenotypic groups: One group had only localized facial pain (believed to result from direct activation of muscle nociceptors), and the other had widespread pain (presumably due to central sensitization). In a discovery cohort, none of the top SNPs approached statistical significance after correction for multiple testing. Interestingly, the top SNPs in the two groups were different.
Diatchenko then proceeded to pathway analysis, where genes are grouped into pathways based on their functions and relationships, and genetic data are queried for associations to pathways, rather than individual SNPs or genes. The approach posits that biological pathways, rather than specific polymorphisms, are important for clinical effects, and in a given population there may be associations with different genes, but they will cluster into a limited number of biologically important pathways. Starting with commercially available software (Ariadne Genomics Pathway Studio) that was made for analyzing RNA expression, Diatchenko and colleagues modified it to analyze SNPs. They first combined p values for all SNPs in one gene to give a gene level p value, and from there calculated a pathway p value from all genes in any of 248 pathways.
In this analysis, patients with just facial pain showed a significant association with the serotonin receptor pathway, including two serotonin receptor genes and two MAP kinase genes. In a discovery cohort, four SNPs in the serotonin pathway combined to cause a 2.3-fold increased risk of TMD with local pain, and the same four SNPs replicated in the OPPERA cohort, giving a significant 1.6-fold increased risk. In patients with widespread pain, a T cell receptor signaling pathway was significantly associated in the discovery cohort, although that did not hold up in the replication cohort.
Beyond genetics
Jörn Lötsch, Goethe Universität, Frankfurt am Main, Germany, has studied the effects of genetic variation on the response to opioids. But in his talk, Lötsch presented a different kind of data—on the response of genes to opioids. His recent work suggests that opioid treatment leads to epigenetic changes in pain patients, which could potentially contribute to pain.
Epigenetic changes are stable, environmentally induced modifications of the genome, such as methylation, that affect gene expression without changing DNA base sequence (reviewed in Doehring et al., 2011). There is a small but growing literature on epigenetics and pain (see PRF related news story). It had been shown that methadone-maintained opioid addicts have increased methylation near OPRM1, the gene for the μ-opioid receptor (Nielsen et al., 2009), but it was unclear whether the change was a cause or an effect of drug use. That inspired Lötsch and Alexandra Doehring to look at pain patients, and Lötsch reported unpublished results showing the same increase in methylation in chronic opioid-treated subjects.
The results suggest that opioids have epigenetic effects, and open up the possibility that the observed changes in DNA methylation could contribute to pain. “As DNA methylation has been shown in animal experiments to produce pain, it would be reasonable to ask whether drug-induced methylation affects pain in patients treated with these drugs,” Lötsch said. Those experiments are ongoing.
Article Contributors:
