BDNF and CGRP interaction: Implications in migraine susceptibility

Abstract

Objectives: Migraine pathophysiology involves several pathways. Our aims were to explore a possible role of the brain-derived neurotrophic factor gene (BDNF) in migraine susceptibility; to study, for the first time, the calcitonin gene-related peptide gene (CGRP); and a possible interaction between the two.

Methods: Using a case-control approach, four tagging single nucleotide polymorphisms (SNPs) (rs7124442, rs6265, rs11030107, and rs2049046) of BDNF and one tagging SNP—rs1553005—of CGRP were analyzed in 188 cases and 287 controls. A multivariable logistic regression was performed, adjusting for gender. Allelic and haplotypic frequencies were estimated. Interaction was assessed by a stepwise multivariable-logistic regression and confirmed by a multifactor dimensionality reduction analysis.

Results: No significant main effects were found; however, a significant interaction was found between BDNF and CGRP, showing an increased risk for the AT-genotype of rs2049046 and the GC-genotype of rs1553005 (odds ratio = 1.88, 95% confidence interval: 1.20–2.93) for migraineurs.

Conclusion: Our data support the hypothesis of an interaction between BDNF and CGRP in migraine susceptibility that should be further explored.

Keywords

Migraine association BDNF CGRP pathophysiology

Introduction

Migraine is a common multifactorial disease (1,2). Several different processes are involved in its pathophysiology, such as alteration of pain and sensory input, increased sensitivity of the cortex leading to aura, central pain facilitation, neurogenic inflammation and brain stem nociceptor sensitization (3,4).

BDNF (brain-derived neurotrophic factor) is important for neuronal growth, development and survival in the central nervous system, and is the most abundant neurotrophin in the brain (5,6). BDNF is expressed in the nociceptive sensory neurons, acting as a central pain modulator at both the spinal and supra-spinal level, contributing to central sensitization and modulating the activation of glutamatergic, N-methyl-d-aspartate (NMDA) receptors (7–9). A significant decrease in the levels of BDNF in platelets was found in migraineurs, showing that it may also have a role in migraine pathophysiology (8). One of the mechanisms involved in migraine is central sensitization, characterized by activity-dependent plastic changes in second-order trigeminal neurons (10). BDNF has been pointed out as a mediator of the trigeminal nociceptive plasticity (9,11). A previous study evaluated several genes that can be expressed after cortical spreading depression (CSD), the mechanism hypothesized to underlie migraine pathophysiology, especially migraine with aura (MA) (10,12). BDNF was one of the genes found to be differentially expressed after CSD (12).

Several studies showed that neurogenic inflammation is associated with pathogenesis of migraine (13). When neuronal excitability occurs, leading to primary brain dysfunction, the perivascular trigeminal sensory nerve fibers are activated, resulting in the release of vasoactive peptides, such as substance P and calcitonin-gene related peptide (CGRP), from trigeminal fibers (13). CGRP is the most abundant neuropeptide in perivascular sensory trigeminal nerve fibers and one of the mediators of neurogenic inflammation; it is a potent vasodilator (14,15), enhancing the activation of second-order neurons, thus contributing to pain transmission (13,16). CGRP has an active role in the activation and sensitization of nociceptors at the peripheral and central level (17), modulating nociceptive transmission in the trigeminovascular system (15).

The involvement of CGRP in migraine pathophysiology has been supported by the finding that the serum levels of CGRP are increased during migraine attacks and there is a decrease after treatment with triptans, although this was not observed in a more recent study (18–21).

CGRP is co-expressed with BDNF in trigeminal ganglion neurons in rats. In addition, CGRP enhances BDNF release from trigeminal neurons in vitro; this is independent of extracellular calcium and is abolished by a CGRP-receptor antagonist (11).

A case-control study found no evidence of a BDNF single nucleotide polymorphism (SNP) (rs6265) being associated with migraine susceptibility (22). The aims of this study were (i) to further explore a possible role of BDNF in migraine susceptibility and (ii) to study, for the first time, CGRP as a candidate gene. Furthermore, since both genes are co-expressed and involved in migraine pathophysiology (3), we also assessed their possible interaction in migraine’s susceptibility.

Subjects and methods

Subjects

A sample of 188 unrelated migraine patients from the neurology clinic, at (Hospital de Santo António) HSA, Porto, was sequentially enrolled in this study. Patients with familial hemiplegic migraine were excluded. Control subjects (287), with no personal history of migraine, were ascertained among healthy blood donors and from the Obstetrics and Gynecology Department of HSA. Controls were from the same ethnic and geographical origin (north of Portugal) as cases, and were age-matched to these. A diagnostic interview was performed both in cases and controls, based on the operational criteria of the International Headache Society (IHS), using the same structured questionnaire. The first edition of these criteria (ICHD-I) (23) was used before 2004; revising the diagnosis using the second edition (ICHD-II) (24) showed no differences in patients’ diagnosis (data not shown). Participants gave their written informed consent and the project was approved by the Ethics Committee of HSA.

Genotyping

Genomic DNA was extracted from peripheral blood leucocytes, using a standard salting-out method (25); or from saliva using ORAGENE kits and DNA extraction according to the manufacturer’s instructions (DNA Genotek, Kanata, Ontario, Canada).

SNPs were selected based on a data dump from the International HapMap Project (Release 24, November 2008, on NCBI B36 assembly, dbSNP build 126) (www.hapmap.org) and tagging SNPs were selected using Haploview 4.1 (Broad Institute, Cambridge, MA, USA) (26), capturing all the variation for both genes studied at an r² threshold of 0.80 and with a minor allele frequency (MAF) of 0.10, by an aggressive tagging approach. For BDNF (NM_170732), these included rs712442 (located on chromosome 11: 27,634,117; on 3′ UTR), rs6265 (27,636,992; a functional SNP: p.V66M), rs11030107 (27,650,911; on intron 1) and rs2049046 (27,679,851, according to the Ensembl database; on 3′ gene region). Two of the SNPs were combined as a two-marker haplotype, allowing coverage of a broader region of SNPs tagged. For CGRP (NM_001033953), rs1553005 (located in chromosome 11: 14,950,566; on 3′ gene region) was selected.

Allelic discrimination was performed using Real-Time PCR (iQ5 Real-Time PCR Detection System, Bio-Rad Laboratories). Primers and molecular beacons were designed using Beacon Designer 6.0.

For BDNF, Real-Time PCR reactions were performed using 20 ng/µL DNA, 10 µL of iQ^TM Supermix or iQ^TM Multiplex Powermix, 1.2-8.0 mM MgCl₂ and 0.5 µM of each primer, in 20 µL final reaction volume. Concentration of the probes used was 0.06 µM of probe 1 (labeled with FAM) and 0.25 µM of probe 2 (labeled with Cy3), for rs6265; 0.04 µM of probe 1 (labeled with FAM) and 0.16 µM of probe 2 (labeled with Cy3), for rs2049046; 0.03 µM of probe 1 (labeled with FAM) and 0.15 µM of probe 2 (labeled with Cy3), for rs7124442; 0.03 µM of probe 1 (labeled with FAM) and 0.30 µM of probe 2 (labeled with Cy3), for rs11030107. For the rs1553005 SNP of CGRP, Real-Time PCR reactions were performed using 20 ng/µL DNA, 10 µL of iQ^TM Supermix, 8 mM MgCl₂, 0.5 µM of each primer, 0.03 µM of probe 1 (labeled with FAM) and 0.15 µM of probe 2 (labeled with Texas Red), in 20 µL final volume.

Primer and probe sequences are available upon request. Cycling conditions used were: an initial denaturing at 95°C, for 3 minutes and 30 seconds, followed by 35 cycles of 30 seconds, at 95°C, 60 seconds at 60°C (for annealing, and at this point fluorescence was measured), and 30 seconds at 72°C. Genotypes were determined automatically by the signal processing algorithms of the iQ5 Optical System Software, version 2.0 (Bio-Rad Laboratories); results were compared with the genotypes visually scored, based on the fluorescent emission data depicted (relative fluorescent units [RFUs]). To validate the Real-Time PCR reaction, we included three individuals previously genotyped by sequencing (a heterozygote, a homozygote for variant 1 and a homozygote for variant 2). If genotypes were uncertain, they were additionally sequenced, using the Big Dye Terminator Cycle Sequencing v1.1 Ready Reaction (Applied Biosystems), according to the manufacturer’s instructions, and an ABI-PRISM 3130 XL genetic analyzer (Applied Biosystems).

Statistical analysis

The Genetic Power Calculator (http://pngu.mgh.harvard.edu/∼purcell/gpc/) was used, assuming a high-risk allele frequency of 0.1, a relative risk for a homozygous genotype of 2.25 and 1.5 in heterozygosity. Hardy-Weinberg equilibrium was tested, using HWE software (http://linkage.rockefeller.edu/ott/linkutil.htm). Demographic data of patients and controls were compared using a chi-square test for qualitative variables. To compare allele frequencies between cases and controls, a chi-square test was used and odds ratios (OR) were estimated, with 95% confidence intervals (CIs). Our initial significance level was set at α = 0.01.

A multivariable-logistic regression was performed (with the most frequent homozygote as the reference), to evaluate association between SNPs and migraine, including the four SNPs of BDNF and the CGRP SNP in the model, and adjusting for gender. We assessed interaction between BDNF and CGRP by a stepwise multivariable-logistic regression. For the logistic regression analyses, significance was set to α = 0.005, using a Bonferroni correction for multiple comparisons. These analyses were performed using SPSS (version 16.0 for Windows).

Gene-gene interactions were also evaluated using multifactor dimensionality reduction (MDR) software (version 2.0) (27,28). MDR is a non-parametric and genetic model-free approach that can identify combination of SNPs involved in disease susceptibility (27). We performed a MDR analysis of the rs2049046*rs1553005 interaction using a forced approach. By using this forced analysis, we evaluated a specific set of attributes and we obtained an unbiased estimate of the testing accuracy. We used a tenfold cross-validation to avoid false-positives (29). The significant results obtained were corrected for multiple testing using the permutation test implemented on the MDR Permutation Tool (version 1.0) (30).

Haplotype frequencies were compared between cases and controls, for the BDNF gene, using Haploview 4.1 (26) with all parameters set at the default values. Frequencies of haplotypes analysed were above 1%, according to the Haploview threshold; to correct for multiple comparisons, when estimating allelic and haplotype frequencies, 10,000 permutations were used.

Results

Demographic and clinical data of 188 unrelated migraineurs and 287 age-matched controls can be found in Table 1. The cases-control ratio was 1:1.5 and no significant differences were found between cases and controls regarding gender.

Table 1.

Demographic and clinical data of migraine patients and controls

Migraine patients (N)	188
Migraine with aura	77
Migraine without aura	111
Females	153
Males	35
Mean age at observation (± SD)	36.14 years ( ± 12.84)
Mean age at onset (± SD)	17.67 years ( ± 8.15)
Family history of migraine	87%
Migraine-free controls (N)	287
Females	217
Males	70
Mean age at observation (± SD)	36.42 years ( ± 12.35)

SD = standard deviation.

A migraine prevalence of 16% has been estimated in a Portuguese population (31,32). We found that we had a power of 64% to detect an association with our sample (for a nominal significance level of 0.05).

Both case and control groups were in Hardy-Weinberg (H-W) equilibrium for the four SNPs of the BDNF gene (rs712442, rs6265, rs11030107 and rs2049046) and for rs1553005 of CGRP. The correlation between the BDNF SNPs was small, denoting the weak linkage disequilibrium (LD) between them; in our sample, these SNPs were also in weak LD (data not shown). To confirm Real-Time PCR efficiency genotyping we have sequenced a random sample. For BDNF SNPs rs6265, rs2049046 and rs712442 all genotypes were concordant and thus 100% efficiency was considered. However, for rs11030107, some genotypes were incorrect; therefore, we sequenced all individuals for this SNP. Regarding CGRP, we also sequenced all individuals and we found 100% of identity with the previous results from Real-Time PCR.

No significant differences were found between cases and controls regarding allele frequencies (Table 2). Genotypic frequencies can be found in Table 3. Results from the multivariable logistic models showed that in the global sample no significant independent effects of BDNF or CGRP were found (Table 3). A significant interaction was found, however, between BDNF and CGRP, showing an increased risk for the AT-genotype of rs2049046 and the GC-genotype of rs1553005 (OR = 1.88, 95% CI: 1.20-2.93, p = .005) as shown in Table 4. Other possible two-way interactions were non-significant (data not shown).

Table 2.

Allele frequencies of SNPs studied in migraine patients and controls

	Alleles (%)
BDNF SNP	All patients	Controls	X²	OR (95% CI)	p
rs7124442
C	112 (29.8)	167 (29.1)	0.053	1.03 (0.78–1.38)	.82
T	264 (70.2)	407 (70.9)
rs6265
A	76 (20.2)	113 (19.7)	0.04	1.03 (0.75–1.43)	.84
G	300 (79.8)	461 (80.3)
rs11030107
G	85 (22.6)	124 (21.6)	0.13	1.06 (0.78–1.45)	.72
A	291 (77.4)	450 (78.4)
rs2049046
T	175 (46.5)	273 (47.6)	0.095	0.96 (0.74–1.25)	.76
A	201 (53.5)	301 (52.4)
rs7124442/ rs2049046
TA	92 (24.5)	141.4 (24.6)	0.003	0.99 (0.73–1.34)	.95
Others	284 (75.5)	432.6 (75.4)
CGRP SNP	Alleles (%)
	All patients	Controls	X²	OR (95% CI)	p
rs1553005
C	124 (33.0)	174 (30.3)	0.75	1.13 (0.86–1.50)	.39
G	252 (67.0)	400 (69.7)

SNP = single nucleotide polymorphism. BDNF = brain-derived neurotrophic factor (gene). CGRP = calcitonin gene-related peptide (gene). OR: odds ratio; CI: confidence interval.

Table 3.

Results from the multivariable logistic regression, gender-adjusted

Genotype	Migraine	Controls	OR	CI 95%	p
BDNF
rs7124442					.91
TT (ref)	96 (51.1)	149 (51.9)	1.00	—	—
TC	72 (38.3)	109 (38.0)	1.13	0.58–2.22	.72
CC	20 (10.6)	29 (10.1)	1.03	0.33–3.22	.96
rs6265					.94
GG (ref)	118 (62.8)	183 (63.8)	1.00	—	—
GA	64 (34.0)	95 (33.1)	1.09	0.60–2.00	.51
AA	6 (3.2)	9 (3.1)	1.26	0.30–5.35	.55
rs11030107					.55
AA (ref)	114 (60.6)	174 (60.6)	1.00	—	—
AG	63 (33.5)	102 (35.5)	0.87	0.49–1.54	.63
GG	11 (5.9)	11 (3.8)	1.55	0.46–5.24	.48
rs2049046					.69
AA (ref)	51 (27.1)	79 (27.5)	1.00	—	—
AT	99 (52.7)	143 (49.8)	1.20	0.63–2.28	.58
TT	38 (20.2)	65 (22.6)	1.01	0.35–2.89	.99
CGRP
rs1553005					.16
GG (ref)	78 (41.5)	140 (48.8)	1.00	—	—
GC	96 (51.1)	120 (41.8)	1.40	0.94–2.08	.09
CC	14 (7.4)	27 (9.4)	0.86	0.42–1.77	.69

(ref) = reference category. OR = odds ratio. CI = confidence interval.

Table 4.

Interaction results found between CGRP*BDNF, assessed by a stepwise logistic regression analysis

CGRPBDNF* interaction	OR	Migraine CI 95%	p
rs1553005*rs2049046			.03
GG*AA	1.00	—	—
GC*AT	1.88	1.20–2.93	.005
GC*TT	0.88	0.46–1.69	.71
CC*AT	0.54	0.27–1.97	.73
CC*TT	0.34	0.04–2.97	.33

CGRP = calcitonin gene-related peptide (gene). BDNF = brain-derived neurotrophic factor. OR = odds ratio. CI = confidence interval.

With the MDR analysis, we found that the best model for the interaction between rs2049046 and rs1553005 showed a testing balanced accuracy of 0.53 and a cross-validation consistency of 10/10 (Table 5). After permutation testing, this model was still significant (p = .002), which confirmed the interaction result found in the logistic regression analysis. Figure 1 shows the combinations associated with high and low risk. The varied patterns of high and low risk cells, with no linearity across each multifactor dimension, provide evidence of gene-gene interaction. The dendogram in Figure 2 shows a strong interaction between the two SNPs, which is indicative of a non-additive effect.

Figure 1.

Interaction graphic for rs2049046 and rs1553005. High-risk combinations are depicted as dark-shaded cells, and low-risk combinations as light-shaded cells. For each cell the left bar indicates the cases and the right bar indicate the controls.

Figure 2.

Interaction dendogram for the SNPs genotyped. The dendogram clearly shows a strong interaction effect between rs2049046 and rs1553005 (darker lines suggest that there is a synergistic relationship: the shorter the lines, the stronger the interaction). SNPs = single nucleotide polymorphisms.

Table 5.

Interaction results found between CGRP*BDNF, assessed by the MDR method

Best model	Cross-validation consistency	Testing-balanced accuracy	OR (95% CI)	p (after permutation)
rs1553005*rs2049046	10/10	0.53	1.84 (1.25–2.69)	.002

CGRP = calcitonin gene-related peptide (gene). BDNF = brain-derived neurotrophic factor. OR = odds ratio. CI = confidence interval. MDR = multifactor dimensionality reduction.

For BDNF, we also performed a haplotype-based analysis. No differences were found between cases and controls for any of the haplotypes evaluated (data not shown).

Discussion

The aim of this study was to assess the possible involvement of BDNF and CGRP in migraine susceptibility and their possible interaction. A previous study found no evidences of BDNF as a risk factor for migraine, regarding the functional SNP rs6265 (22). We now used a haplotype-block tagging approach, selecting SNPs that cover whole gene common variation. No significant differences were found between cases and controls regarding the SNPs analysed.

For CGRP, only one tagging SNP (rs1553005) was selected with the software used, covering also the whole variation within this gene, with the parameters specified for the European population. Several polymorphisms of CGRP are described as non-informative in dbSNP (http://www.ncbi.nlm.nih.gov/SNP) in European populations, although they may be informative for other populations. This may be a reason why these SNPs are not captured when searching for tagging SNPs.

For CGRP, no differences were also found between cases and controls. No differences were found regarding gender for BDNF and CGRP (data not shown).

We have also assessed a possible interaction between BDNF and CGRP. The Hierarchically Well-Formulated Rule (HWFR) is commonly used when a logistic regression is performed and assumes that lower order terms of the highest order interaction term are included in the model (33). With limited number of SNPs (as in our study), a stepwise logistic regression without HWFR may be a better approach to test for SNP-SNP interactions (33). Therefore, we performed a stepwise logistic regression, allowing that the final model included only the interaction terms between the BDNF and the CGRP SNPs, in an exploratory analysis of a putative interaction between the two genes. Further studies are now needed to confirm these results.

A significant interaction between rs2049046 and rs1553005 was found, showing an increased risk for the double heterozygotes AT and GC (OR = 1.88, 95% CI: 1.20–2.93) when compared with the most frequent double homozygotes (AA of rs2049046 and GG of rs1553005).

However, logistic regression parameters may not be completely reliable due to the limited number of subjects in each group (34).Therefore, we employed a MDR analysis. This method combines genotypes into a single dimension with two groups (high or low risk). MDR is also able to detect a high-order interaction in the absence of any statistically significant main effects. The combination of cross-validation and permutation testing minimizes the occurrence of false positive results (27). Furthermore, MDR is able to detect gene-gene interactions even in small sample sizes (35), which gives us confidence in our results. With the MDR analysis we also found a highly significant two-way interaction between those SNPs associated with an increased risk for migraine, confirming the results of the logistic regression analysis. Therefore, an interaction between these variants, or between variants in LD with them, may be involved in migraine susceptibility.

Our sample size is a limitation of our study and does not give us enough power to detect a variant with an OR < 1.5. Also, rare alleles may also be involved in the susceptibility of migraine (36), although we were interested in studying the role of common variants (MAF > 0.10). Therefore, we could not exclude that these genes may be independent risk factors for migraine, although this was not detected in our sample. However, some gene-gene interactions occur even without the presence of a significant main effect (37) and future association studies should take this into account.

Nonetheless, for common variation, we covered the whole gene region with the SNPs studied for both genes. Importantly, our high case:control ratio increases the power to detect any effects in our sample. In addition, cases and controls were matched for age-at-observation and gender, and were from the same population and geographic region; several studies using lineage markers, sensitive to population stratification, have showed there is no substructure within the Portuguese population (38–40). Also, we corrected for multiple testing, to prevent type-I errors, and used logistic regression analyses to incorporate all SNPs in the model and analyze their effects together. Additionally, we set our initial level of significance at p < .01, to use a stringent statistical criteria as suggested by Bird et al. (41). Furthermore, the MDR analysis strongly confirmed the interaction results found with the logistic regression analysis. The application of this method has already led to the identification of gene-gene interactions for several diseases, such as sporadic breast cancer, essential hypertension, schizophrenia and familial amyloid neuropathy (29).

Several findings support a role of CGRP in initiating and sustaining migraine episodes (42). CGRP has been extensively studied regarding its possible role in migraine treatment. Antagonists for CGRP receptors may act at the central level and modulate nociceptive trigeminovascular transmission in the cat, and are arising as a new generation of migraine drugs (42–44). Furthermore, studies in transgenic mice showed that modulation of one of the CGRP receptors (RAMP1) levels may contribute to migraine susceptibility and, in particular, to photophobia (45). More interestingly, a recent study showed that adenosine triphosphate (ATP)-gated P2X3 receptors may be involved in chronic pain (including migraine) and CGRP seems to enhance its transcription, involving also mediation by BDNF (46). BDNF is also a mediator of neuronal survival and plasticity of dopaminergic, cholinergic and serotonergic neurons, and its involvement in migraine seems clear as a mediator of the trigeminal nociceptive plasticity (9,11,47).

Our data support the hypothesis of an interaction between BDNF and CGRP in migraine susceptibility. The results obtained also reinforce the importance of replication studies with these genes in larger samples and in other populations. Interaction between candidate genes for migraine pathophysiology should be further explored, in order to have a broader knowledge of genes interplay in susceptibility for this common disease.

Footnotes

Acknowledgements

We would like to thank all patients and controls for participating in this study. We thank also Dr Serafim Guimarães and nurses Teresa Gomes and Palmira Gouveia for their assistance with sample collection, Paula Magalhães for technical assistance and Jorge Pinto-Basto and Zoltán Bochdanovits for their helpful suggestions.

This study was supported in part by grants of Fundação para a Ciência e Tecnologia (FCT) (POCTI-034390/99/FCT), Sociedade Portuguesa de Cefaleias (SPC) and the National Headache Foundation (NHF) (USA). CL was the recipient of an FCT fellowship (SFRH/BD/17761/2004). The authors report no conflicts of interest.

References

Kalfakis

Panas

Vassilopoulos

Malliara-Loulakaki

. Migraine with aura: segregation analysis and heritability estimation. Headache 1996; 36: 320–322.

Devoto

Lozito

Staffa

D'Alessandro

Sacquegna

Romeo

. Segregation analysis of migraine in 128 families. Cephalalgia 1986; 6: 101–105.

Menken

Munsat

Toole

. The global burden of disease study: implications for neurology. Arch Neurol 2000; 7: 418–420.

Mazaheri

Hajilooi

Rafiei

. The G-308A promoter variant of the tumor necrosis factor-alpha gene is associated with migraine without aura. J Neurol 2006; 253: 1589–1593.

St Clair

Ott

Feng

. Brain-derived neurotrophic factor gene C-270T and Val66Met functional polymorphisms and risk of schizophrenia: a moderate-scale population-based study and meta-analysis. Schizophr Res 2007; 91: 6–13.

Lipsky

Marini

. Brain-derived neurotrophic factor in neuronal survival and behavior-related plasticity. Ann NY Acad Sci 2007; 1122: 130–143.

Thompson

Bennett

Kerr

Bradbury

McMahon

. Brain-derived neurotrophic factor is an endogenous modulator of nociceptive responses in the spinal cord. Proc Natl Acad Sci USA 1999; 96: 7714–7718.

Blandini

Rinaldi

Tassorelli

. Peripheral levels of BDNF and NGF in primary headaches. Cephalalgia 2006; 26: 136–142.

Mannion

Costigan

Decosterd

. Neurotrophins: peripherally and centrally acting modulators of tactile stimulus-induced inflammatory pain hypersensitivity. Proc Natl Acad Sci USA 1999; 96: 9385–9390.

10.

Pietrobon

. Migraine: new molecular mechanisms. Neuroscientist 2005; 11: 373–386.

11.

Buldyrev

Tanner

Hsieh

Dodd

Nguyen

Balkowiec

. Calcitonin gene-related peptide enhances release of native brain-derived neurotrophic factor from trigeminal ganglion neurons. J Neurochem 2006; 99: 1338–1350.

12.

Urbach

Bruehl

Witte

. Microarray-based long-term detection of genes differentially expressed after cortical spreading depression. Eur J Neurosci 2006; 24: 841–856.

13.

Durham

. Calcitonin gene-related peptide (CGRP) and migraine. Headache 2006; 46(Suppl 1): 3–8.

14.

Lassen

Haderslev

Jacobsen

Iversen

Sperling

Olesen

. CGRP may play a causative role in migraine. Cephalalgia 2002; 22: 54–61.

15.

Greco

Tassorelli

Sandrini

Di Bella

Buscone

Nappi

. Role of calcitonin gene-related peptide and substance P in different models of pain. Cephalalgia 2008; 28: 114–126.

16.

Striessnig

. Pathophysiology of migraine headache: insight from pharmacology and genetics. Drug Discovery Today: Disease Mechanisms 2005; 2(4): 453–462.

17.

Benemei

Nicoletti

Capone

Geppetti

. Pain pharmacology in migraine: focus on CGRP and CGRP receptors. Neurol Sci 2007; 28(Suppl 2): 89–93.

18.

Goadsby

Edvinsson

. The trigeminovascular system and migraine: studies characterizing cerebrovascular and neuropeptide changes seen in humans and cats. Ann Neurol 1993; 33: 48–56.

19.

Goadsby

Edvinsson

Ekman

. Vasoactive peptide release in the extracerebral circulation of humans during migraine headache. Ann Neurol 1990; 28: 183–187.

20.

Tvedskov

Lipka

Ashina

Iversen

Schifter

Olesen

. No increase of calcitonin gene-related peptide in jugular blood during migraine. Ann Neurol 2005; 58: 561–568.

21.

Edvinsson

. Blockade of CGRP receptors in the intracranial vasculature: a new target in the treatment of headache. Cephalalgia 2004; 24: 611–622.

22.

Marziniak

Herzog

Mossner

Sommer

. Investigation of the functional brain-derived neurotrophic factor gene variant Val66MET in migraine. J Neural Transm 2008; 115(9): 1321–1325.

23.

International Headache Society. Classification and diagnostic criteria for headache disorders, cranial neuralgias and facial pain. Cephalalgia 1988; 8: 1–96.

24.

International Headache Society. International Classification of Headache Disorders. 2nd edn. Cephalalgia 2004; 24(Suppl 1): 1–160.

25.

Miller

Dykes

Polesky

. A simple salting out procedure for extracting DNA from human nucleated cells. Nucleic Acids Res 1988; 16: 1215–1215.

26.

Barrett

Fry

Maller

Daly

. Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 2005; 21: 263–265.

27.

Ritchie

Hahn

Roodi

. Multifactor-dimensionality reduction reveals high-order interactions among estrogen-metabolism genes in sporadic breast cancer. Am J Hum Genet 2001; 69: 138–147.

28.

Moore

. Computational analysis of gene-gene interactions using multifactor dimensionality reduction. Expert Rev Mol Diagn 2004; 4: 795–803.

29.

Moore

Gilbert

Tsai

. A flexible computational framework for detecting, characterizing, and interpreting statistical patterns of epistasis in genetic studies of human disease susceptibility. J Theor Biol 2006; 241: 252–261.

30.

Greene

Himmelstein

Nelson

. Enabling personal genomics with an explicit test of epistasis. Pac Symp Biocomput 327–336.

31.

Pereira Monteiro

Maio

Calheiros

. Overlap of migraine and tension-type headaches in a population-based study. In: Olesen

(ed.) Headache classification and epidemiology. Vol 4, New York: Raven Press, 1994, 103–106.

32.

Pereira Monteiro

Barros

. Familial incidence of primary headaches in a general population. In: Olesen

Bousser

M-G

(eds.) Genetics of headache disorders. Vol 8, Philadelphia: Lippincott-Williams & Wilkins, 2000, 37–40.

33.

Lin

Desmond

Bridges

Jr Soong

. Variable selection in logistic regression for detecting SNP-SNP interactions: the rheumatoid arthritis example. Eur J Hum Genet 2008; 16: 735–741.

34.

Coffey

Hebert

Ritchie

. An application of conditional logistic regression and multifactor dimensionality reduction for detecting gene-gene interactions on risk of myocardial infarction: the importance of model validation. BMC Bioinformatics 2004; 5: 49–49.

35.

Ritchie

Hahn

Moore

. Power of multifactor dimensionality reduction for detecting gene-gene interactions in the presence of genotyping error, missing data, phenocopy, and genetic heterogeneity. Genet Epidemiol 2003; 24: 150–157.

36.

Hemminki

Forsti

Bermejo

. The ‘common disease-common variant’ hypothesis and familial risks. PLoS ONE 2008; 3: e2504–e2504.

37.

Reijmerink

Bottema

Kerkhof

. TLR-related pathway analysis: novel gene-gene interactions in the development of asthma and atopy. Allergy 2010; 65: 199–207.

38.

Pereira

Prata

Amorim

. Diversity of mtDNA lineages in Portugal: not a genetic edge of European variation. Ann Hum Genet 2000; 64: 491–506.

39.

Pereira

Fondevila

Philips

Amorim

Carracedo

Gusmao

. Genetic characterization of 52 autosomal SNPs in the Portuguese population. Genetics Supplement Series. Forensic Sci Int 2008; 1: 358–360.

40.

Beleza

Gusmao

Lopes

. Micro-phylogeographic and demographic history of Portuguese male lineages. Ann Hum Genet 2006; 70: 181–194.

41.

Bird

Jarvik

Wood

. Genetic association studies: genes in search of diseases. Neurology 2001; 57: 1153–1154.

42.

Recober

Russo

. Calcitonin gene-related peptide: an update on the biology. Curr Opin Neurol 2009; 22: 241–246.

43.

Storer

Akerman

Goadsby

. Calcitonin gene-related peptide (CGRP) modulates nociceptive trigeminovascular transmission in the cat. Br J Pharmacol 2004; 142: 1171–1181.

44.

Mehrotra

Gupta

Chan

. Current and prospective pharmacological targets in relation to antimigraine action. Naunyn-Schmiedebergs' Arch Pharmacol 2008; 378: 371–394.

45.

Recober

Kuburas

Zhang

Wemmie

Anderson

Russo

. Role of calcitonin gene-related peptide in light-aversive behavior: implications for migraine. J Neurosci 2009; 29: 8798–8804.

46.

Simonetti

Giniatullin

Fabbretti

. Mechanisms mediating the enhanced gene transcription of P2X3 receptor by calcitonin gene-related peptide in trigeminal sensory neurons. J Biol Chem 2008; 283: 18743–18752.

47.

Angelucci

Brene

Mathe

. BDNF in schizophrenia, depression and corresponding animal models. Mol Psychiatry 2005; 10: 345–352.