State of the Art ReviewsPain

Translational pain research: Lessons from genetics and genomics

See allHide authors and affiliations

Science Translational Medicine  13 Aug 2014:
Vol. 6, Issue 249, pp. 249sr4
DOI: 10.1126/scitranslmed.3007017

Abstract

Pharmacological, surgical, psychological, and alternative medicine approaches for the treatment of chronic pain, including neuropathic pain, provide only partial relief for most patients, with the efficacy of existing medications often blunted by dose-limiting side effects arising from drug actions on cells outside the pain-signaling axis. The development of more effective treatments for pain—particularly chronic pain states such as neuropathic pain—has been hampered by lack of predictive animal models and biomarkers, variation in pain characteristics between patients or on a day-to-day basis for single patients, patient stratification on the basis of symptoms rather than mechanism, and a high rate of placebo responses. We discuss genetic and genomic approaches to translational pain research. We review examples of the identification and validation of human pain targets through rodent genome-wide association studies (GWAS) and global mRNA expression studies, functional screening in flies and mice, human GWAS and whole-exome sequencing studies, and the targeted candidate gene approach. These and other emerging genetic and genomic strategies are likely to facilitate the development of new, more effective pain therapeutics.

INTRODUCTION

Nociceptive pain, triggered by an external noxious stimulus such as a pinprick or a hot object, is an adaptive response and facilitates survival. In contrast, neuropathic pain (somatosensory pain that persists beyond the healing of the initial insult) is maladaptive, exacting substantial economic and quality of life costs. Globally, about one-third of adults suffer from chronic pain (1). A recent report from the U.S. Institute of Medicine concluded that about 100 million U.S. adults suffer from chronic pain, including neuropathic pain, at an economic cost of hundreds of billion dollars annually (2). We need a better understanding of the pathophysiology of, and more effective treatments for, persistent pain.

Neuropathic pain is associated with abnormal changes in cells of the normal pain-signaling axis in the peripheral (PNS) and central nervous systems (CNS) (35). Preclinical studies have identified a plethora of molecular substrates in the PNS and CNS that appear to contribute to neuropathic pain, suggesting many potential targets for therapeutic intervention (35). Nevertheless, as described in the accompanying Review (6), it has been difficult to translate these findings into the clinic. Preclinical animal models do not accurately predict human therapeutic responses. Further, only a minority of people with acute pain goes on to develop neuropathic pain, and its duration and severity are quite variable, complicating human studies. Variability at the molecular, cellular, and neural networks and systems levels may contribute to patient-to-patient differences in the manifestations of the pain experience, introducing complexity into clinical studies. Stratification of patients on the basis of symptoms has not proven helpful in understanding this variability, and interpretation of human studies is further confounded by large placebo responses. Thus, there has been increasing interest in genetic and genomic approaches that may facilitate identification of the most opportune target molecules for pain therapy and more effective patient stratification for clinical studies and personalized therapy.

Several comprehensive reviews have discussed genomic approaches to the study of complex traits (79) and animal models for the study of pain (10). Here, we focus on examples of recent studies in which genetic approaches coupled with functional analyses are contributing to translational chronic pain research. We discuss advantages and limitations of each approach and prospects for the future.

FROM ANIMALS TO MAN

Pain has both peripheral and central components that are affected by the large network of gene products that comprises pain-signaling pathways. Preclinical studies of several animal models that mimic at least some aspects of human pain syndromes have identified hundreds of genes that are thought to contribute to the pain phenome (11). An important result of these animal studies has been to link variants of the human cognates of these genes to pain in human patients, as discussed below.

Genome-wide association studies in mice point to human pain genes

The Mouse Genome Database contains multiple examples of gene alleles that cause different disease phenotype from that caused by the analogous human gene, raising questions about translation of animal data to humans in the absence of studies on the human cognate gene (12). Nevertheless, inbred mouse strains have been used to identify new pain genes through a quantitative trait locus (QTL) analysis, initially independently and later in combination with covariance analysis of global transcription patterns. Some of the single-nucleotide polymorphisms (SNPs) identified in studies on mice have also been found in the human cognate genes and reported to be risk factors for pain susceptibility in humans.

For example, the spontaneous pain model autotomy (a result of the scratching and biting of a rodent hindpaw after nerve injury) is highly variable among animal strains, a fact that led to the identification of the QTL pain1 on mouse chromosome 15, which predisposes rodents to this type of neuropathic pain behavior (13). Further mapping including SNP and RNA expression analysis showed pain1 to correspond to the human γ2 subunit of the voltage-gated calcium channel, encoded by the CACNG2 gene. This gene has lower expression in the high autotomy strains than in the low autotomy strains (Fig. 1A) (14). A naturally occurring hypomorph mutation of the Cacng2 gene in mouse (Cacng2stg) produces the phenotype “stargazer” (15). In these mice, peripheral nerve injury is followed by a high incidence of autotomy, recapitulating the finding that high autotomy is linked to reduced abundance of the γ2 subunit (Fig. 1B) (14). Profiling of SNPs in human CACNG2 in a cohort of women who underwent mastectomy, some of whom had experienced postmastectomy pain, identified a three-SNP haplotype associated with increased risk of pain (Fig. 1C) (14).

Fig. 1. Identification of the pain-related gene CACNG2.

(A) RNA expression. The relative expression of Cacng2 (stargazer, Cacng2stg) is elevated in strains of mice that show lower autotomy, both before and after nerve injury. Relative expression was calculated as expression for each mouse divided by the mean value for expression across all mice. Each square indicates data from a single mouse. Horizontal line, the Cacng2 expression average for each group; line at 1, mean Cacng2 expression in all mice. C3H, AKR, C57, C58, and CBA, strains of mice tested. (B) Autotomy is elevated in mouse strains with a Cacng2 hypomorphic genotype (Cacng2stg/stg) (n = 11). Heterozygote (Cacng2stg/+) (n = 13), wild-type (WT) (Cacng2+/+) (n = 23). Autotomy is expressed as percent of animals with an autotomy score ≥9 on a scale of 0 to 11 (14). Mean autotomy score is indicated within each bar. P = 0.0002, hypomorphs versus heterozygotes; P = 0.000073, hypomorphs versus WT (Fisher’s exact test). (C) SNP analysis. Map locations and haplotypes for the 12 chromosome 22 SNPs used for genotyping human CACNG2. Darker red squares indicate haplotypes, specifically linkage disequilibrium (LD) between the SNPs that intersect at that square. Vertical lines, statistical significance of the allele association with the pain phenotype (in −log P units). Dashed horizontal line, P = 0.05 (−log P = 1.3); red squares, LD value (D′) for the two SNPs that intersect at that square. The higher the D′, the darker the color of the square. Adapted with permission from Nissenbaum et al. (14).

In another example, an unbiased genome-wide screen of a discovery cohort of 15 inbred mouse strains (Fig. 2, A to F) led to the identification of a nonsynonymous SNP (rs48804829) in the gene P2rx7. This gene encodes the adenosine triphosphate–gated P2X7 purinergic receptor, which has been associated with reduced allodynia (pain resulting from a normally nonnoxious stimulus such as a light touch of the skin) in the spared nerve injury mouse model but not with mechanical nociceptive activity in uninjured animals (16). P2X7 is polymorphic in humans and has been associated with bone disorders, infectious diseases, inflammatory and cardiovascular disorders, cancer, and psychiatric disorders (17). Biallelic rs48804829 SNP, which causes a P451L substitution in the cytoplasmic domain of the P2X7 receptor, is responsible for the association with spared nerve injury–induced allodynia in a validation cohort of another 15 mouse strains, with the biallelic L451 variant associated with hypoalgesia. The P451L substitution does not prevent activation of the P2X7 receptor but rather blocks its coupling to pannexin 1 channels to form the nonspecific cationic pore. Thus, the P2X7 receptor may play a role in chronic pain states (18).

Fig. 2. Haplotype-based genetic mapping of allodynia to P2X7 in mice and human.

(A) Allodynia in 18 inbred mouse strains (% allodynia, the percentage of maximum possible allodynia summed over all determinations of withdrawal thresholds for each mouse strain over the total period of testing). (B) Allodynia in 18 mouse strains arranged by strain from least (left) to most (right) allodynia, as standardized by z score. Black bars, strains with the P2X7 Pro451 allele; white bars, strains with the Leu451 allele (the NZB/BIn strain has the Leu451 allele). (C to F) Effect of P2X7 genotype on pain-related behavior in mice subjected to nerve injury (Pro451 allele versus Leu451 allele). (C) Baseline withdrawal thresholds to mechanical stimulation. (D) Ipsilateral allodynia to mechanical stimulation. (E) Contralateral allodynia to mechanical stimulation (negative values represent hypoalgesia manifested as an increase in the threshold stimulus to induce paw withdrawal). (F) Ipsilateral allodynia (the average percent change from baseline on postoperative days 5 and 7). Baseline was measured in noninjured mice, and allodynia was measured after spared nerve injury in each strain. n = 8 to 26 mice per strain. *P < 0.05, **P < 0.01, ***P < 0.001 compared to the opposite genotype, Student’s t test. (G) Genetic association of the human P2RX7 gene with chronic pain. Significance of the association (−log P) between 23 SNPs (in genomic order) in and near the human P2RX7 gene and the level of chronic pain in patients with postmastectomy pain (PMP) (red) or osteoarthritis (OA) (blue). A third set of P values (purple) is shown for the combination of the first two P values, assessed by the optimally weighted z test. Horizontal line, P < 0.05 corrected for the number of effectively independent multiple comparisons after accounting for LD among the SNPs. (H) LD between each pair of SNPs in terms of D′. Red, strong LD (D′ = 1); pink, moderate LD. Blue (D′ = 1) and white (D′ < 1) indicate low confidence (by log10 odds score) in the value of D′. No haplotype blocks (regions of widespread high LD) were observed among the SNPs. (I) Positions of the 23 genotyped SNPs in a 90.3-kb region containing the P2RX7 gene. (Top) Intron and exon structure of the 52.7-kb gene. Red, SNPs with significant association with pain. Adapted with permission from Sorge et al. (16).

Extending these findings from mouse to humans, Sorge et al. (16) found an association of a P2X7 three-SNP haplotype [hyperfunctional SNP rs208294 (H155Y), hypofunctional SNP rs7958311 (R270H), and intronic SNP rs208296 of unknown function] with postmastectomy pain in women, explaining 4.5% of the trait variance seen in these patients (Fig. 2, G and H). They also confirmed an association of pain susceptibility with different P2X7 haplotypes in patients with osteoarthritis, explaining 1% of genetic variability. A meta-analysis of the combined data for 23 SNPs in postmastectomy and osteoarthritis patients showed a significant association of reduced chronic pain with the hypofunctional SNP (rs7958311). These results suggest that a genetic characterization might be useful in predicting an individual’s risk for developing chronic pain.

Nonetheless, a therapeutic benefit of P2X7 receptor engagement for patients with pain has yet to be demonstrated. A clinical trial of an allosteric modulator of P2X7 (GSK1482160) did not support its further development for treatment of chronic inflammatory pain (19). A second clinical trial with an antagonist (CE-224,535) failed to show efficacy better than placebo in patients with rheumatoid arthritis (20).

The unbiased approach described above, which capitalizes on a rich array of rodent models and resources, can identify molecular targets that appear to be relevant to human pain. Nevertheless, the resulting targets may have only a small effect on pain or none at all.

Functional screening in flies points to human pain genes

Reasoning that avoidance behavior to a noxious stimulus might be conserved between flies and humans, Neely et al. (21) capitalized on the well-studied model Drosophila to unmask the role of a voltage-gated calcium channel subunit in pain behavior. Drosophila’s power as a model organism is based on the conservation of key molecules and molecular signaling pathways between flies and humans and the availability of well-characterized investigative tools.

Using a Drosophila RNAi library, Dietzl et al. (22) knocked down individual genes and screening for noxious heat avoidance in a behavioral assay (Fig. 3, A to D), identifying the straightjacket gene, which encodes the homolog of the human α2δ3 subunit of the voltage-gated calcium channel, the product of the gene CACNAD2D3. Although α2δ3 expression is brain-specific in mouse, in humans, it is also expressed in heart and skeletal muscle, and enhances trafficking of the voltage-gated calcium channel to the plasma membrane (23). Knockout of α2δ3 in mice substantially impairs the acute response to noxious heat and delays the onset of adjuvant-induced thermal hyperalgesia (Fig. 3, E and F), without affecting mechanical pain or basic motor skills and coordination, general exploratory activities, and anxiety (21).

Fig. 3. Regulation of thermal nociception by voltage-gated calcium channel subunit α2δ3 in adult Drosophila, mouse, and human.

(A) Diagram of voltage-gated calcium channels showing the relative position of members of the α2δ3 subunit family. (B) Effect of RNA interference (RNAi) knockdown of α2δ3 (Straightjacket, stj, in Drosophila) on noxious thermal avoidance in adult Drosophila (% avoidance of noxious temperature). stj-IR1, inverted repeat 1; stj-IR2, inverted repeat 2, both crossed to elav-Gal4;UAS-DCR2. (C) Efficiency of stj mRNA knockdown in elav-Gal4>UAS-stj-IR1/2 adult fly brains, measured by quantitative polymerase chain reaction. stj values in flies treated with stj-specific shRNAs are expressed relative to those in flies treated with a control shRNA. (D) Temperature-induced paralysis for control and elav-Gal4>UAS-stj flies. Data in (B) to (D) are presented as mean values ± SEM. **P < 0.01, Student’s t test. (E) Acute thermal nociception response in mutant α2δ3−/− (n = 16) and control α2δ3+/+ mice (n = 12) in a hot plate assay. Littermates were used as controls. Values represent the latency of response to the indicated temperatures. (F) Thermal pain in α2δ3+/+ (n = 10) and α2δ3−/− (n = 21) mouse littermates in a hot plate assay. Complete Freund’s adjuvant (20 μl) was injected into the hindpaws, and the mice were tested at 54°C. Data for (E) and (F) are presented as mean values ± SEM. *P < 0.05, **P < 0.01, and ***P < 0.001, comparing mutant to control mice; #P < 0.05, comparing sensitization to baseline (day −2) of the same genotype (Student’s t test). (G) Effect of polymorphisms in CACNA2D3 (α2δ3) on acute thermal pain in a cohort of 189 healthy volunteers, as measured by heat wind-up–induced sensitivity (temporal summation of brief, repetitive heat pulses). The minor allele of the SNP rs6777055 (C/C, n = 5) was significantly associated with reduced thermal pain sensitization compared to the other genotypes (CA + AA, n = 184). PCA, principal components analysis of raw pain scores used to derive a thermal wind-up score. (H) Influence of polymorphisms in CACNA2D3 (α2δ3) on pain sensitivity in 169 patients with lumbar chronic root pain 1 year after discectomy. The minor alleles of two SNPs rs6777055 (C/C, n = 8) and rs1851048 (A/A, n = 17) were significantly associated with reduced lumbar pain sensitization compared to the other genotypes (rs6777055: CA + AA, n = 153; rs1851048: A/G + G/G, n = 137). Genotyping was not always successful, hence the slightly different total numbers in the chronic pain group. Data in (G) and (H) are presented as mean values ± SEM. *P < 0.05, Student’s t test. All panels are adapted with permission from Neely et al. (21).

In human CACNAD2D3, a minor allele of the intronic SNP rs6777055 (present in 4% of a control population in the homozygous state) was associated with reduced thermal pain sensitivity in a cohort of 189 healthy volunteers (Fig. 3G) and, together with the minor allele of another intronic SNP, rs1851048, was associated with reduced pain within the first year after surgery in a cohort of 169 Caucasian adults assessed in a prospective observational study of discectomy (surgical removal of herniated disc material that impinges on a nerve root) (Fig. 3H) (21). As discussed below, the functional significance of intronic SNPs remains to be determined.

Successful application of this approach to the development of new pain medications requires that findings in Drosophila can be extrapolated to humans. Although there are no current clinical trials testing an α2δ3-based treatment, the relatedness of α2δ3 to α2δ1, the target of the first-line neuropathic pain therapeutic pregabalin (23), suggests that α2δ3 might also be a druggable target for pain.

Different injuries, common pain pathway

Genes that are critical for pain signaling should be common to multiple species and to multiple injury types. Consequently, global gene expression studies of injury in rodents have been used to identify pain targets. These studies have the advantage of using an unbiased approach, but may be limited by the poor predictive power of rodent models for the human therapeutic response.

For example, guanosine triphosphate cyclohydrolase (GCH1), the rate-limiting enzyme for 6(R)-l-erythro-5,6,7,8-tetrahydrobiopterin (BH4) synthesis (24), is a modulator of peripheral neuropathic and inflammatory pain in rodents. BH4, an essential cofactor in the production of several neurotransmitters (25), is significantly up-regulated in dorsal root ganglion (DRG) neurons after axonal injury or soft tissue inflammation within the projection fields of DRG neurons, a result of GCH1 up-regulation (24). Short hairpin RNA (shRNA) knockdown of GCH1 (26) or small-molecule blockers of the BH4 biosynthetic pathway (24, 27) ameliorate neuropathic and cancer pain in rodents. Building on these rodent studies, genomic variants of human cognates of these targets were shown to influence pain thresholds. A common GCH1 gene haplotype (rs8007267, rs3783641, rs8007201, rs4411417, and rs752688) is present in 15.4% of individuals in a Caucasian control population (24). The haplotype was linked to lower injury-induced GCH1 and BH4 levels, although the molecular mechanism for this effect is unknown. The association of this GCH1 haplotype with reduced pain after sensitization, but not basal thresholds of pain in humans, was statistically significant. Pain-protective GCH1 haplotypes in humans have been associated with reduced experimental pain (28), shorter time on treatment for a variety of persistent pain conditions (29), favorable outcome after surgical correction of lumbar degenerative disc disease (30), and a delay in onset of cancer pain (31). Sulfasalazine, a U.S. Food and Drug Administration–approved anti-inflammatory drug, blocks sepiapterin reductase, an enzyme in the BH4 synthetic pathway, produces analgesia in a rat model of diabetic neuropathy (29), and is being evaluated in clinical trials for multiple pain indications including diabetic neuropathy (clinical trial identifier: NCT01667029). In contrasting independent studies, GCH1 haplotypes were not associated in patients with dental pain (32), chronic pancreatitis (33), provoked vestibulodynia (34), or HIV-associated pain in South African adults of black African ancestry (35).

HUMAN GENOME-WIDE ASSOCIATION STUDIES

Genome-wide association studies (GWAS) provide an unbiased alternative approach that requires fewer a priori assumptions and, because they are on humans, can—if adequately powered—identify loci with small size effect on pain phenotype. An extensively studied example is catecholamine-O-methyltransferase (COMT), which regulates the enzymatic activity that inactivates the neurotransmitters dopamine, epinephrine, and norepinephrine (36). In GWAS studies, a three-SNP haplotype [two synonymous SNPs (rs4633 and rs4818) and the nonsynonymous SNP (rs4680, which produces V158M)] is associated with a 30-fold difference in enzymatic activity and a significant change in pain sensitivity (37, 38). It is not known, however, which neurotransmitters or synapses are affected by the COMT pain-regulating haplotype, the mechanism by which this effect is produced, or whether the effect will manifest in different pain types (39). Three recent GWAS studies with large cohorts (>1000) of patients with migraine (4042), with and without quantitative expression assays in lymphoblastoid cell lines, reported association with additional loci, but the loci were different in the three studies, raising questions about their validity. A GWAS study in a large discovery and replication cohort in patients with joint-specific chronic widespread pain, followed by quantitative expression studies in mice, identified two novel loci on chromosome 5, although results for SNPs in COMT, GCH1, and the μ opioid receptor (OPRM1) did not reach statistical significance (43). Larger, well-defined cohorts may be needed for discovery of meaningful pain associations, especially when the pathophysiology is incompletely understood.

Even when associations are discovered, GWAS may not elucidate underlying disease mechanisms. Limitations of the GWAS approach include the need for large patient cohorts, failed replication in some studies, and sensitivity to gender and ethnicity. Ideally, these studies should be carried out using standardized phenotyping methods, which can present a challenge for multicenter studies. Even when statistically significant, the effect of SNP may explain only a small percent of trait variance in patients. Stratification of patients by underlying pain mechanisms, rather than by symptomatology or clinical diagnosis, may improve the yield of this approach.

WHOLE-EXOME SEQUENCING

Whole-exome sequencing, which makes no a priori functional assumptions, has been used to identify rare alleles that contribute to the heritability of complex diseases and health-related traits (44). Whole-exome sequencing of pain-sensitive and pain-insensitive individuals selected from 2500 volunteers from the Twins UK cohort has identified rare (minor allele frequency <5%) variants associated with pain (45). No single rare variant with a large size effect on pain was found, but network analysis identified rare variants in up to 30 genes within the angiotensin pathway, which is linked to pain in animal models (46). The angiotensin II receptor is expressed on most small- and medium-sized DRG neurons, and the specific blocker EMA401 attenuates capsaicin-induced calcium influx into cultured human and rat DRG neurons (47). In an initial phase 2 clinical trial (clinical trial identifier: ACTRN12611000822987), EMA401 significantly reduced mean pain intensity after 4 weeks of treatment compared to placebo in patients with postherpetic neuralgia. As noted below, next-generation sequencing technology is likely to add to this powerful discovery approach.

RARE MONOGENIC PAIN DISORDERS VALIDATE PAIN-SIGNALING TARGETS

In general, not all rare alleles are necessarily pathogenic, and their absence from control population cohorts may not establish disease relevance (48). Nonetheless, rare monogenic (Mendelian) and highly penetrant human disorders can provide compelling evidence for target validation when they yield a strong genotype-phenotype correlation. These disorders can come to light from studies on families, especially large families with a pain phenotype and multiple affected and unaffected family members. Alternatively, this approach may be based on extrapolation from a well-studied rare genetic disorder to a more common disease, or may be based on biological plausibility, in which functional analysis can establish a causal relationship between the gene and the phenotype.

In the absence of strong evidence from linkage analysis, however, the targeted candidate gene approach carries high risk because of the possibility of not finding pathogenic variants in the candidate gene, and the significant initial investment needed to identify the culprit gene and understand its contribution to disease. Moreover, functional analysis of selected gene variants may require specialized methods for assessment of the gene product under study, ideally in the native cell type where the gene is normally expressed, because some variants have functional effects that are different in heterologous expression systems. Despite these challenges, the targeted gene approach, focused by linkage analysis, has established a major role of voltage-gated sodium channel Nav1.7, which drives impulse firing in DRG neurons including nociceptors, in human pain (49). This approach has also implicated TRPA1, which transduces chemical irritants, in pain in humans (50).

Nine distinct sodium channels share a common overall structural motif, but with different amino acid sequences and with different biophysical and pharmacological properties (51). Nonselective sodium channel blockers have been used to treat clinical pain (52), but their efficacy is blunted by dose-limiting side effects (ataxia, confusion, sedation, and cardiac arrhythmias) caused by block of sodium channels in the CNS and cardiovascular and skeletal muscles (53). The identification of sodium channels that are preferentially or selectively expressed within nociceptors would allow selective block of nociceptors activity, alleviating pain without off-target side effects.

Nav1.7 is preferentially expressed within peripheral (DRG and sympathetic ganglion) neurons including nociceptors (54, 55), amplifies small depolarizations (56, 57), and thus sets the gain on pain signaling (58). It is found along the axons of DRG neurons, including their peripheral nerve terminals within the epidermis, where it contributes to the initiation of pain signaling, and within their terminal arborizations within the spinal cord dorsal horn, where it may contribute to impulse invasion of the terminal axon or synaptic transmission (59, 60). Genomic variants of Nav1.7 have been linked to four human pain disorders, three of these monogenic, and polymorphisms of Nav1.7 likely influence pain sensitivity in the general population (49).

Inherited erythromelalgia (IEM), a rare disorder characterized by excruciating pain in the limbs typically triggered by mild warmth, is inherited in a classical Mendelian autosomal dominant manner (61) (Fig. 4A). Linkage analysis and candidate gene sequencing identified missense mutations in SCN9A, the gene encoding Nav1.7, in the first kindred studied and other families with IEM (62, 63). Functional studies showed that these mutated channels might cause the disease because the mutations alter the gating properties of Nav1.7 channels and expression of the mutant channels increases the excitability of DRG neurons (63, 64). Structural modeling of the F1449V mutation shows that the amino acid substitution increases the diameter of the pore from 0.6 to 1.1 Å, destabilizing the hydrophobic ring that acts as the activation gate (Fig. 4, B and C) (65, 66), making the channel easier to open so that activation occurs with weaker stimuli (Fig. 4, D and E). DRG nociceptor neurons expressing the mutant channel fire at higher than normal frequencies, and in response to small stimuli that do not normally provoke firing (Fig. 4, F and G) (63).

Fig. 4. A gain-of-function mutation in sodium channel Nav1.7 in a family with IEM.

(A) Multigeneration segregation of erythromelalgia with the c.4393T>G mutation in SCN9A, which causes the substitution of phenylalanine at codon 1447 with valine (F1449V) in Nav1.7. Circles, females; squares, males; arrow, proband; black symbols, subjects affected with erythromelalgia; white symbols, unaffected subjects; red symbols, affected subjects who were tested and found to be heterozygous for the T4393G mutation; cyan symbols, unaffected subjects tested and found to be homozygous for the WT T4393 allele. (B and C) Atomic-level modeling of Nav1.7 channels based on crystal structures of sodium and potassium channels. A hydrophobic ring is formed by the C-terminal ends of S6 helices, which contributes to an activation gate that stabilizes the preopen state of the channel. Cytoplasmic views of the C-terminal ends of the S6 helices of WT (B) and F1449V (C) mutant channels. (D to G) Effects of the F1449V mutation on the biophysical properties of Nav1.7. (D) Both the WT (black) and F1449V (orange) mutant Nav1.7 channels produce rapidly activating and inactivating sodium currents of similar amplitude. (E) Voltage dependence of activation and steady-state fast inactivation of F1449V mutant channels, determined by fitting data to Boltzmann functions. Activation is shifted in a hyperpolarizing direction, lowering the threshold stimulus for channel opening, whereas fast inactivation is shifted in a depolarized direction, making more channels available to open at depolarized potentials. (F and G) Current-clamp recordings in DRG neurons showing that the F1449V mutant channels cause an increased frequency of action potential firing in response to small stimuli (G) compared to WT channels (F). Panels (A) and (D) to (G) were adapted with permission from Dib-Hajj et al. (63); panels (B) and (C) were adapted with permission from Lampert et al. (65).

The second human pain disorder to be linked to Nav1.7 is paroxysmal extreme pain disorder (PEPD; also known as familial rectal pain), in which patients experience severe lower body pain in response to rectal stimulation and, later in life, develop periocular and perimandibular pain (67). This disorder, like IEM, is autosomal dominant, and linkage analysis followed by functional analysis of the candidate gene showed that PEPD is produced by missense mutations in Nav1.7. The mutant channels that cause PEPD show a gain of function and render the DRG neurons hyperexcitable (68), but, in contrast to IEM mutations, the inactivation of the Nav1.7 channel is impaired (67). Patients with PEPD usually respond to treatment with carbamazepine, which restores inactivation of PEPD mutant channels (67).

In contrast to these gain-of-function disorders, channelopathy-associated insensitivity to pain (CIP) is an autosomal recessive disorder, caused by loss-of-function coding or splicing mutations in SCN9A (69). Patients with CIP are insensitive to pain and report painless burns, fractures, dental extractions, and childbirth. These patients have apparently normal autonomic function (69), an observation that is clinically important but unexpected because Nav1.7 is present in sympathetic ganglion neurons as well as in nociceptors (54, 70).

Although the number of patients reported is smaller, a similar approach has recently identified gain-of-function substitutions in kindreds with rare pain phenotypes (71, 72) in another peripheral sodium channel, Nav1.9, which regulates nociceptor excitability (73, 74). Two gain-of-function Nav1.9 mutations in multigeneration families segregate with episodic lower extremity pain accompanied by severe sweating. A different de novo gain-of-function mutation in Nav1.9 in two unrelated individuals was associated with congenital insensitivity to pain; the paradoxical association between a gain of function of Nav1.9 and loss of sensitivity to pain has not yet been fully explained.

This approach has also pointed to the involvement of another channel in pain. A gain-of-function point mutation in TRPA1, a membrane-associated sensor of environmental irritants, has been found in a family with upper body pain, in a disorder called familial episodic pain syndrome (50). The mutation increases by fivefold the inward current upon channel activation. Antagonists of TRPA1 inhibit the mutant channel in in vitro assays (50), suggesting that TRPA1 blockers might be useful clinically.

FROM RARE GENETIC DISORDERS TO COMMON DISEASES

These successes with highly penetrant monogenic disorders with strong phenotypes have led to understanding of more common disorders. Gain-of-function mutations arise from substitution of various amino acids within Nav1.7 in about 30% of patients with idiopathic small fiber neuropathy, a relatively common disorder in which patients experience profound pain, usually with onset in adulthood (75, 76). These Nav1.7 variants, by causing a variety of changes in channel function, produce hyperexcitability and spontaneous firing in nociceptor neurons, which likely contribute to evoked and spontaneous pain, respectively (7579).

Functional SNPs in SCN9A may also bias sensitivity to pain or susceptibility to chronic pain in broad populations. A nonsynonymous SNP in SCN9A (rs6746030; Nav1.7/R1150W), present in about 30% of a control Caucasian population, produces a moderate increase in firing frequency in nociceptors (80) (Fig. 5). The minor allele at this site (W1150) alters pain threshold and is associated with higher pain scores in human subjects with osteoarthritis, sciatica, and traumatic limb amputations, consistent with the idea that Nav1.7 contributes to human pain sensitivity and to clinical pain disorders (81). An intronic SNP (rs6754031), of unknown functional significance, has also been reported to associate with fibromyalgia in a small cohort of Mexican women (82).

Fig. 5. Enhanced firing of neurons in humans carrying a chronic pain–associated SNP within SCN9A.

(A) Traces showing action potentials elicited from DRG neurons transfected with hNav1.71150R. The three traces illustrate the response to current injection of 1×, 2×, and 3× threshold (250, 500, and 750 pA). Stimulus duration, 1 s. (B) Traces of action potentials elicited from neurons transfected with the hNav1.71150W construct. The three traces illustrate the response to current injection at 1×, 2×, and 3× threshold (150, 300, and 450 pA). (C) Number of action potentials (defined as spikes overshooting 0 mV) elicited by stimulations with 1-s current injection pulses of various intensities. The mean response of neurons expressing hNav1.71150W (orange) (n = 30) was significantly higher than that of neurons expressing hNav1.71150R (black) (n = 48) for stimuli above 100 pA [analysis of variance (ANOVA) with repeated measures, P < 0.05]. Error bars are means ± SEM. Adapted with permission from Estacion et al. (80).

Nav1.7 may also be a major regulator of noninherited pain. Nav1.7 accumulates abnormally in painful human neuromas, tangles of blindly ending, abortively regenerating axons that form at the distal ends of injured nerves after, for example, traumatic limb amputations and similar injuries (83). Injured axons within neuromas generate abnormal ectopic impulses (84, 85), and the abnormal accumulations of Nav1.7 within these injured axons suggest that, even in the absence of a family history of pain or a variant in the SCN9A gene, increased expression of Nav1.7 can contribute to chronic neuropathic pain after traumatic injury. Within the axonal endings in the neuroma, Nav1.7 coaccumulates with activated mitogen-activated protein kinases (MAPKs) (83), which can phosphorylate the channel and reduce the strength of the stimulus needed for Nav1.7 activation, thereby contributing to enhanced DRG neuron excitability (86). Nav1.7 expression also increases in DRG neurons in animal models of inflammation (87) and diabetic neuropathy (88, 89). Nav1.7 may therefore contribute to nociceptor hyperexcitability in a broad range of nongenetic pain disorders.

MUTATIONAL ANALYSIS OF BIOLOGICALLY PLAUSIBLE TARGETS

Nav1.8 is also preferentially expressed in peripheral sensory neurons (90). Nav1.8 sodium channels carry most of the sodium current underlying the rising phase of action potentials in DRG neurons (91, 92) and support repetitive action potential firing in response to sustained depolarization (92). They are modulated in a pronociceptive manner by cytokines and their downstream signaling molecule p38 MAPK kinase (9395). In animal studies, knockout of Nav1.8 or ablation of Nav1.8-positive neurons impairs thermal hyperalgesia after inflammation (96, 97) and cold-induced pain (98), indicating that Nav1.8 is important for pain signaling. Nav1.8 is also crucial for full manifestation of the gain-of-function effects of Nav1.7 mutations (70).

Nav1.8 underlies human pain conditions (99). Reasoning that Nav1.8 is preferentially expressed within peripheral sensory neurons and their axons, regulates DRG neuron firing, and has been implicated in pain signaling in rodent models, Faber et al. (99) screened for Nav1.8 mutations in 104 patients with painful predominantly small-fiber neuropathy who did not carry mutations in Nav1.7, and identified seven Nav1.8 substitutions in 9 subjects. Three of these Nav1.8 mutations met criteria for potential pathogenicity by predictive algorithms, and two of the three enhanced Nav1.8’s response to membrane depolarization, producing hyperexcitability in DRG neurons. Thus, candidate gene screening based on the biological plausibility of pathogenicity led to a demonstration of a role for this channel in a relatively common pain condition.

THE TARGETED GENE APPROACH: TRAJECTORY FROM TARGET TO TREATMENT

Nav1.7 is a good target for isoform-specific blockers or modulators of pain because genetic studies have verified its central role in human pain signaling, and there are no serious cognitive, cardiac, or adverse motor effects in individuals with a total loss of Nav1.7. At least some pathogenic variants of Nav1.7 can be bound and affected by therapeutics. There have been occasional reports of patients with IEM who responded to treatment with the local anesthetic derivative mexiletine (100) or the anti-epileptic sodium channel blocking drug carbamazepine (101). Patients with PEPD often respond to carbamazepine (67). In the responsive patients with IEM, the mutation endows the channel with an enhanced response to the pharmacological agent, which causes an enhanced use-dependent block (100), consistent with the state- and use-dependent inhibitory effect of local anesthetics (102). Carbamazepine normalizes the shift of activation in the Nav1.7 V400M IEM mutant channel, a previously unrecognized effect of this class of state-dependent sodium channel blockers (101). In PEPD, carbamazepine restores normal inactivation in the mutant channels (67).

To determine whether in vitro assays might be used to predict the most effective pain therapies for specific patients, Yang et al. (103) used Nav1.7 carrying the V400M mutation as a carbamazepine-responsive “seed” for atomic-level modeling and thermodynamic/pharmacological analyses to predict functional alterations in different variants of Nav1.7. They showed that a mutation in the DI/S4-5 linker, S241T—separated within the linear channel sequence by 159 amino acids from the V400 residue (Fig. 6A)—is in fact located less than 3 Å from the V400M site in the folded structure (Fig. 6B). Electrophysiological recordings then confirmed that the S241T and V400M mutations are thermodynamically coupled. These findings predicted the carbamazepine responsiveness of S241T, which was confirmed at the channel (Fig. 6C) and cellular levels (Fig. 6, D and E). This approach provides a paradigm for determining responsiveness of other common channel variants to drugs, potentially bringing us closer to individualized, genomically guided pain pharmacotherapy.

Fig. 6. Predicted responses of genetic variants of the human Nav1.7 channel.

(A) Schematic of the human Nav1.7 channel linear topology showing the mutations S241T, V400M, and F1449V. (B) Cytosolic view of the structural model of Nav1.7 channel transmembrane domains. (Below) Enlarged image of boxed area containing S241, V400, and F1449 residues. V400, S241, and F1449 are shown as sticks and colored red, black, and yellow, respectively. Note the proximity of S241 and V400. (C) The averaged voltage dependence of activation of S241T mutant channel treated with dimethyl sulfoxide (DMSO) or carbamazepine (CBZ) (30 μM, a clinically relevant concentration), fitted with Boltzmann equation. Carbamazepine causes a 7.0-mV depolarizing shift of activation of the S241T mutant channel. (D and E) Action potential frequency of (D) DMSO (0.1%)–treated or (E) carbamazepine (30 μM)–treated DRG neurons expressing the S241T mutant channel in response to 1-s, 200-pA depolarization current steps. Adapted with permission from Yang et al. (103).

Validation of Nav1.7 as a therapeutic target has begun to propel human studies of Nav1.7 blockers for pain. Although limited by a small number of subjects and a short period of drug administration, a recent study of a small-molecule, nonspecific sodium channel blocker, XEN402, in subjects with gain-of-function variants in Nav1.7 and chronic pain was interpreted by the authors as suggesting target engagement (104). An orally bioavailable small-molecule sodium channel blocker, PF-05089771, which belongs to a new chemical class, has robust selectivity for human Nav1.7 channels (105). This compound is 10- to 900-fold selective for Nav1.7 over other voltage-gated sodium channel isoforms and is selective over potassium and calcium channels by 1000-fold. This blocker is being evaluated in a phase 2 clinical trial in etiologically homogeneous patients with Nav1.7-related IEM (clinical trial identifier: NCT01769274).

Nav1.8 may also be a viable drug target because it plays a role in pain signaling in animals and humans as discussed above. Indeed, a Nav1.8 blocker can attenuate pain in animal studies (106), although this compound is not orally bioavailable and has not been tested in humans. These preclinical animal studies, the role of Nav1.8 in nociceptor excitability, and the identification of Nav1.8 gain-of-function mutations in patients with painful peripheral neuropathy suggest that targeting of Nav1.8 alone, or in combination with Nav1.7, may prove to be an effective pain treatment.

CHALLENGES AND OPPORTUNITIES

Genetic and genomic studies of pain (Table 1) have helped to advance the search for new pain therapeutics. Nonetheless, these approaches face challenges.

Table 1. Representative genetic and genomic approaches for translational pain research in humans.
View this table:

Noncoding variants

Thus far, genetic and genomic approaches have been most informative in understanding the function of exonic or coding gene variants. Yet, the bulk of mammalian genomic sequences are noncoding. For example, the SCN9A gene spans more than 180 kilo–base pair (kbp) on chromosome 2 but contains only about 6 kbp of coding sequence. Functional testing, by necessity, has been limited to exonic variations because intronic mutations are difficult to assess in heterologous expression systems or in vitro. The relevant sequences can be several thousand base pairs long, and intronic sequences diverge substantially in orthologous genes from different species. Although exonic gain- or loss-of-function substitutions in SCN9A have been functionally profiled (49), intronic variants have not. For example, in one study (81), 5 of 27 SNPs in SCN9A showed significant association with pain conditions, but only the exonic SNP rs6746030 was functionally tested. The remaining four were assumed either to be in LD with other causative mutations or to have effects, not yet understood, on RNA splicing or regulatory functions. The intronic SNP rs208296 in P2RX7—part of the haplotype associated with postmastectomy pain (16)—and both SNPs in human CACNAD2D3 (rs6777055 and rs1851048) associated with reduced pain (21) are of unknown functional significance. The same could also be said about numerous synonymous exonic SNPs whose effects may only be unmasked in native neurons.

The need for predictive animal models

Although invaluable for elucidating pathophysiological mechanisms (10), animal pain models have poor predictive power for successful translation into clinical pain treatment. Animal models (for example, sciatic nerve axotomy leading to autotomy) may not precisely mimic pain in humans. A recent study has questioned the reliability of animal models in general to reflect human inflammatory disorders (107). An alternative approach is to study experimental pain in human volunteers. Ethically approved studies of ultraviolet B–induced experimental pain in human subjects followed by validation of the findings in animals have shown concordance between the injury-induced changes in cytokine expression profile in humans and rodents (108). However, the use of pain-inducing protocols in humans is limited.

Limitations of GWAS

The use of GWAS for the discovery of pain genes has faced a major challenge: the need for large cohorts of affected individuals with uniform pain phenotypes and matched controls. Pain that is classified as neuropathic by one clinician may be classified differently by a second clinician. Clinical phenotyping is also complicated by the similarity of pain symptoms associated with diverse underlying mechanisms, day-to-day variability in the pain experienced by a given subject, and the absence of clinically validated objective biomarkers for pain. A further challenge is presented by the paucity of studies of diverse ethnic cohorts. Rigorous patient stratification is essential. Studies of large cohorts of subjects and replication of findings from discovery cohorts in independent validation cohorts can add rigor by eliminating false-positive associations. Meta-analysis of multiple GWAS studies can also identify variants with small size effects and minimize false-positive findings in small cohort studies; however, meta-analysis can also be subject to bias. Finally, GWAS can identify associations between disease phenotypes and gene variants but, in the absence of functional assessment, cannot elucidate mechanisms of disease.

FUTURE DIRECTIONS

New technologies will undoubtedly increase the power and translational reach of genetic and genomic approaches to pain. Induced pluripotent stem cell (iPSC) technology and the ability to differentiate these cells into sensory neurons (109) may allow researchers to functionally profile noncoding and silent SNPs in cell-based systems that are close to a native background. These iPSC-based systems could offer robust cellular platforms for pharmacogenomic studies, especially when tagged with genetic markers that are preferentially expressed within nociceptive neurons.

Next-generation sequencing technology will also provide a powerful discovery tool to identify novel targets. For example, tetrad (one affected proband, two asymptomatic parents, and one unaffected sibling) whole-genome sequencing has been used to analyze a singleton case of another neuronal excitability disorder, epilepsy. This method identified a mutation in the voltage-gated sodium channel Nav1.6 linked to epilepsy, which produces epileptogenic-like discharges when expressed in hippocampal neurons (110). Similar analyses in patients with pain disorders that have not been associated with known genetic variants could be productive. Transcriptome analysis with RNA-seq (111) may also identify additional new targets for pain research by virtue of its exquisite sensitivity to detect differences in transcript levels under pathological conditions.

Although genetic and genomic approaches have not yet, in themselves, yielded new pain therapeutics approved for clinical use, these and other new technologies are likely to pinpoint and validate additional targets and to help to delineate basic mechanisms underlying chronic pain. We believe that genetic and genomic approaches will contribute to new stratification and diagnostic criteria for pain, personalized pharmacotherapy for pain, and development of new, more effective classes of pain therapeutics.

REFERENCES AND NOTES

  1. Acknowledgments: We thank the members of our group for valuable discussions. The Center for Neuroscience and Regeneration Research is a Collaboration of the Paralyzed Veterans of America with Yale University. Funding: Work in the authors’ laboratory is supported in part by grants from the Rehabilitation Research and Development Service and Medical Research Service, Department of Veterans Affairs, and the Erythromelalgia Association. Author contributions: S.G.W. and S.D.D.-H. planned, wrote, and revised multiple drafts of the article. Competing interests: The authors declare that they have no competing interests.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article