Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2017 Jul 13.
Published in final edited form as: Psychiatr Genet. 2012 Dec;22(6):271–278. doi: 10.1097/YPG.0b013e3283586248

A survey of putative anxiety-associated genes in panic disorder patients with and without bladder symptoms

Ryan L Subaran a,b, Ardesheer Talati a,d, Steven P Hamilton f, Phillip Adams b,d, Myrna M Weissman a,d,c, Abby J Fyer a,e, Susan E Hodge a,b,d
PMCID: PMC5508589  NIHMSID: NIHMS872040  PMID: 23018769

Abstract

Background

We have previously described a subtype of panic disorder (PD) that we termed ‘bladder syndrome’, characterized by urological and bladder symptoms (and possibly interstitial cystitis) in the patients and/or their family members and confirmed the validity of this subset in family linkage and association analysis. In this study, we determine (a) whether 20 single-nucleotide polymorphisms (SNPs) reported in the literature can be replicated in a new PD dataset and (b) whether dividing the sample into those with and without the ‘bladder syndrome’ can help to resolve the genetic heterogeneity within this new sample.

Methods

We selected 20 putative associated SNPs from the literature, taken from studies published since 2004. We tested these SNPs for association in a sample of 351 PD patients and 552 controls, and then divided them into subgroups of 92 patients from bladder families and 259 from nonbladder families.

Results

(a) When analyzed in all PD patients, none of the 20 SNPs appeared to be replicated (except for SLC6A4 from our previous study, but in a sample that overlaps substantially with that in our previous report). (b) However, some intriguing findings emerged when we separated bladder from nonbladder families: SLC6A4, reported by us previously, yielded stronger evidence than before (P = 0.0018) when examined only in nonbladder families, and in contrast, is not statistically significant in bladder families. Two other markers yielded nominally significant results in bladder families – rs5751876 in ADORA2A (P = 0.046) and rs12579350 in TMEM16B (P = 0.035) – but were not significant in nonbladder families. (c) Two markers had noticeably lower P-values when we differentiated the women and analyzed them separately – rs12579350 in TMEM16B (P-value decreased from 0.035, as above, to 0.00055) and a different SNP in ADORA2A, rs4822492 (P-value decreases from 0.07 to 0.028).

Significance

Our results indicate that most of the 20 reported associations do not hold up when PD is analyzed as one group. However, our findings provide further evidence that PD with bladder symptoms may be genetically different from PD without bladder. We suggest that it is worth pursuing SLC6A4 in nonbladder PD, and ADORA2A and TMEM16B in bladder PD. Also, the possibility of a male–female difference in PD is worth pursuing. We also briefly discuss issues of replication and multiple tests.

Keywords: ADORA2A, ANO2, genetics, interstitial cystitis, panic disorder, SLC6A4

Introduction

Panic disorder

Panic disorder (PD) is a debilitating yet common psychiatric condition, with a lifetime prevalence of 3.4–4.7% in the USA (Kessler et al., 2006; Schumacher et al., 2011). Although it is clear that PD shows at least moderate heritability (Hettema et al., 2001; Schumacher et al., 2011), the genes responsible for this effect have remained elusive (Hamilton, 2009; Maron et al., 2010). Linkage studies have failed to identify specific PD genes (with one exception; see below), and there have been only a handful of reports of genetic association, with even fewer being replicated (Maron et al., 2010).

Failure to replicate can result from spurious initial associations that arise from small sample sizes. However, other explanations are also possible, including undetected population stratification, the so-called ‘replication fallacy’ (Gorroochurn et al., 2007 and see the Discussion section), sample heterogeneity between studies, etc.

One method of minimizing potential heterogeneity in PD studies has been to focus on populations that are presumably more genetically homogenous (Thorgeirsson et al., 2003). Another important safeguard against heterogeneity comes from paying very careful attention to phenotype (e.g. Greenberg, 1992; Ioannidis et al., 2009), and stratifying affected individuals by subphenotypes (Hamilton et al., 2003).

The bladder syndrome

In a previous work, we identified a syndrome characterized by specific bladder, thyroid, and cardiovascular problems, as well as severe migraines. Several of these symptoms had already been reported in PD (Gorman et al., 1988; Lydiard et al., 1994; Stewart et al., 1994; Zaubler and Katon, 1996; Placidi et al., 1998). In our genetic studies of 128 multiplex families with PD (Fyer and Weissman, 1999; Weissman et al., 2000; Hamilton et al., 2003), we observed that these symptoms aggregated nonrandomly in some families, but not others.

Whereas some of the symptoms, such as migraines, are relatively common in the general population, the bladder manifestations are relatively rare, making it all the more intriguing that the bladder phenotype led to the strongest findings in our genetic studies (discussed below). Thus, we refer to the collection of bladder, thyroid, and cardiovascular symptoms as the bladder syndrome.

In subsequent studies, several lines of evidence converged to support the hypothesis that this syndrome represents a separate genetic entity within PD. Linkage analysis found significant evidence for linkage to chromosome 13q (LOD = 3.52) when we analyzed the syndrome families separately (Weissman et al., 2000). As mentioned above, the most striking symptoms were bladder problems, and examination of case records by an urologist indicated that these symptoms represented interstitial cystitis (IC), a chronic and debilitating urological disorder (Hanno, 1994) that is relatively rare [lifetime prevalence around 0.5% (Jones and Nyberg, 1997)]. In an independent sample of patients with IC, we observed a four-fold higher lifetime prevalence of PD than in urological controls without IC; also, the IC patients were significantly more likely to have a first-degree relative with PD than were probands without PD (Weissman et al., 2004). Further evidence for familiality emerged from another study where we included an independent PD case–control sample (Talati et al., 2008). We found an increase in IC, as defined by the consensus criteria developed by NIDDK, in probands with PD, as opposed to controls; also, first-degree relatives of PD probands were at an increased risk for IC, whether or not the proband had IC.

Whether this connection between PD and IC is because of pleiotropy or autonomic dysregulation in both conditions or other factors altogether is not yet known. For further discussion, see Weissman et al. (2004), Talati et al. (2008).

In this study, we do two things. First, we attempt to replicate the findings of 20 positive association reports from the literature. We do this in our PD sample, using the standard approach of including all cases in the association analysis. Second, on the basis of the above evidence for genetic differences between families with and without the bladder syndrome, we analyze the two types of cases separately. We believe that this second approach may help resolve the genetic heterogeneity within PD and potentially provide additional evidence for two distinct forms of PD. In addition, given the complex nature of PD associations, at the end of the paper, we also briefly discuss our negative and positive results, with an emphasis on replication and multiple tests.

Methods and materials

We draw data from two of our previous studies, referred to as the panic linkage study (Hamilton et al., 2003) and the anxiety study (Gyawali et al., 2010; Strug et al., 2010), respectively.

Sample description

Cases

All cases for this study had a definite or a probable Diagnostic and Statistical Manual of Mental Disorders, 4th ed., diagnosis of PD with onset by the age of 31 years and were of non-Hispanic White ancestry. The PD cases were drawn from two previously described populations, one a case–control study of anxiety disorders including panic (Gyawali et al., 2010; Strug et al., 2010) and the other a linkage study of panic (Hamilton et al., 2003). Both datasets had data on bladder symptoms that allowed us to generate PD +bladder and PD −bladder groups. Of the 259 nonbladder cases, 248 were drawn from the anxiety disorder study and 11 from the panic linkage study. In addition, we had 92 ‘PD with bladder’ (PDB) cases, 59 from the anxiety study and 33 from the linkage sample. To qualify as PDB, an individual had to have PD and also had to have at least one first-degree relative with PD, and the individual and/or the first-degree relative also had to have bladder symptoms consistent with an NIDDK definition of IC. Sex distribution was heavily skewed toward women: 207 women and 52 men among the nonbladder cases, and 80 women and 12 men among the PDB cases. The 307 cases from the anxiety study and the 44 from the linkage study represented all cases that fulfilled our criteria and for whom we had DNA.

Controls

The first set of controls (n = 80), all ethnically White, were recruited in the same manner as the independently recruited cases above (Talati et al., 2008). They were interviewed on the Schedule for Affective Disorders and Schizophrenia Lifetime Version modified for Anxiety Disorders (SADS-LA-IV) and were required to have (a) no anxiety disorder, (b) no family history of anxiety disorder, and (b) to be at least 30 years of age, so that they had passed through the age of risk requirement for the onset of PD in the cases. To increase the number of controls, we screened the 4035 individuals who make up the National Institute of Mental Health Human Genetics Initiative psychiatric control sample recruited from across the USA. There were 534 individuals from the National Institute of Mental Health Human Genetics Initiative psychiatric controls older than 30 years of age who screened negative for all psychiatric categories assessed on the Composite International Diagnostic Interview Short Form (CIDI-SF). Of these, we were able to genotype 472 Whites, bringing our total anxiety-free control group to a total of 552 (278 women and 274 men). Table 1 summarizes the characteristics of the cases and controls.

Table 1.

Description of the sample

Panic disorder cases Female Male Total
No bladder syndrome 207 52 259 (248a + 11b)
Bladder syndrome 80 12 92 (59a + 33b)
Total cases 287 64 351 (307a + 44b)
Controls 278 274 552
Open in a new tab
a

From the anxiety study (Gyawali et al., 2010; Strug et al., 2010).

b

From the panic linkage study (Hamilton et al., 2003).

Single-nucleotide polymorphism selection

At the start of this study, we carried out a PubMed search on ‘panic disorder genetic association’ and identified 25 single-nucleotide polymorphisms (SNPs) with statistical significance for association with PD reported since 2004. [For one SNP, rs4822492 in ADORA2A (#5 in our list; see Table 2), the statistical significance was for anxiety disorder, not panic, but we included this SNP as it is also in tight linkage disequilibrium with the other ADORA2A SNP, rs5751876, #4 in our list.] In all cases, we used the same SNPs reported in each study, as our goal was to replicate these associations. (That is, we did not use SNPs in reported linkage disequilibrium with the reported SNPs as proxies or rely on a gene-based approach with other markers within the gene of the reported SNP.) Of these 25 SNPs, five were excluded because of multiplex assay incompatibility as scored by the Sequenom ‘Assay design algorithm’ (Sequenom Inc., San Diego, California, USA) or failure to pass quality control (see below), leaving 20 SNPs. Table 2 summarizes information about these 20 SNPs.

Table 2.

Descriptions and sources of the 20 single-nucleotide polymorphisms analyzed in this study

# SNP Gene Chromosome Study Cases OR or RR P-value Replicated?
1 rs2296972 HTR2A 13 Unschuld et al. (2007) 154 NR 0.0006 –
2 rs6313 HTR2A 13 Unschuld et al. (2007) 154 NR 0.009* –
3 rs6311 HTR2A 13 Unschuld et al. (2007) 154 NR 0.009* –
4 rs5751876 ADORA2A 22 Deckert et al. (1998) 457 1.51 0.01 Hohoff et al. (2010)
5 rs4822492 ADORA2A 22 Childs et al. (2008) 62 NR 0.007 –
6 rs4680 COMT 22 Domschke et al. (2008) 115 NR 0.01 Rothe et al. (2006))
7 rs8016905 BDKRB2 14 Gratacos et al. (2009) 173 3.06/2.79 10−7/10−6 –
8 rs10414815 ZNF574 19 Gratacos et al. (2009) 173 2.03 0.001* –
9 rs6136667 PDYN 20 Gratacos et al. (2009) 173 NR 0.018* –
10 rs945032 BDKRB2 14 Gratacos et al. (2009) 173 3.06 1.0 ×10−7 –
11 rs3813034 SLC6A4 17 Gyawali et al. (2010) 307 1.56–1.81 0.007 –
12 rs5326 DRD1 5 Maron et al. (2005) 127 1.25 0.02 –
13 rs1386494 TPH2 12 Maron et al. (2007) 213 1.88 0.01 Maron et al. (2008)
14 rs860554 PKP1 1 Otowa et al. (2009) 200 NR 4.6 ×10−8 –
15 rs9372078 PLEKHG1 6 Otowa et al. (2009) 200 NR 4.9 ×10−7 –
16 rs17466684 CLU 8 Otowa et al. (2009) 200 NR 6.9 ×10−7 –
17 rs12579350 TMEM16B 12 Otowa et al. (2009) 200 NR 3.7 ×10−9 –
18 rs9302001 BC045767 13 Otowa et al. (2009) 200 NR 3.0 ×10−7 –
19 rs3816995 SDK2 17 Otowa et al. (2009) 200 NR 2.1 ×10−7 –
20 rs8192506 DBI 2 Thoeringer et al. (2007) 126 0.43 0.0298 –
Open in a new tab

NR, not reported; OR, odds ratio; RR, relative risk; SNP, single-nucleotide polymorphism.

a

Denotes failure to pass multiple test correction for panic disorder.

Genotyping and quality control

All SNPs were genotyped using Sequenom iPLEX MassARRAY through the Roswell Park Cancer Institute Genomic Shared Resources (Buffalo, New York, USA). All DNA was prepared as recommended by Roswell Park Cancer Institute Genomic Shared Resources. For quality control, 20 individuals were genotyped in duplicate for all SNPs and four SNPs were genotyped in duplicate for all individuals. SNPs and individuals who yielded nonconcordant duplicates were excluded. In addition, we excluded individuals who yielded genotypes for fewer than 90% of all SNPs and SNPs that yielded calls for fewer than 90% of individuals. When tested for Hardy–Weinberg equilibrium, the 20 SNPs had P-values ranging from 0.01 to 1.00, with a median of 0.59. The two SNPs with P-values below 0.05 (rs17466684 of CLU, #16 in our tables; and rs9302001of BC045767, #18 in the tables) did not yield any results of interest. We had rs3813034 (#11 in Tables 2 and 3), genotypes for 307 individuals from a previous study (Gyawali et al., 2010) and regenotyped this SNP for the same individuals in this study. On comparing the overlap between studies, we observed complete concordance, indicating genotyping fidelity across platforms.

Table 3.

P-values and odds ratios for 20 single-nucleotide polymorphisms, under different analyses

All PD
Cases 351
Controls 552		 No bladder
Cases 259
Controls 552		 No bladder (fm)
Cases 207
Controls 278		 Bladder
Cases 92
Controls 552		 Bladder (fm)
Cases 80
Controls 278
# SNPs P-value OR P-value OR P-value OR P-value OR P-value OR
1 rs2296972 0.77 1.03 0.52 1.08 0.43 1.12 0.58 0.91 0.82 0.96
2 rs6313 0.51 1.06 0.43 1.09 0.80 1.03 0.98 1.01 0.77 0.95
3 rs6311 0.35 1.10 0.27 1.13 0.69 1.05 0.92 1.02 0.77 0.95
4 rs5751876 0.29 1.11 0.81 1.03 0.31 1.15 0.046 1.38 0.021 1.52
5 rs4822492 0.24 1.12 0.62 1.06 0.10 1.25 0.07 1.33 0.028 1.49
6 rs4680 0.69 0.96 0.92 1.01 0.95 0.99 0.27 0.84 0.58 0.90
7 rs8016905 0.17 1.16 0.27 1.14 0.78 1.04 0.26 1.21 0.53 1.13
8 rs10414815 0.63 0.93 0.26 0.83 0.85 0.96 0.33 1.24 0.16 1.42
9 rs6136667 0.40 0.89 0.22 0.83 0.83 0.96 0.74 1.08 0.56 1.15
10 rs945032 0.20 0.85 0.46 0.90 0.27 1.20 0.11 0.70 0.64 0.89
11 rs3813034 0.0125 1.27 0.0018 1.40 0.0041 1.46 0.93 0.99 0.93 1.02
12 rs5326 0.78 1.04 0.70 1.06 0.92 1.02 0.94 0.98 0.66 0.89
13 rs1386494 0.65 1.06 0.49 1.11 0.85 1.04 0.81 0.94 0.44 0.81
14 rs860554 0.31 0.89 0.30 0.88 0.38 0.88 0.70 0.93 0.66 0.91
15 rs9372078 0.89 1.01 0.72 1.04 0.42 1.12 0.72 0.94 0.94 0.99
16 rs17466684 0.91 0.99 0.96 0.99 0.93 1.02 0.87 0.96 0.92 1.02
17 rs12579350 0.32 1.19 0.93 1.02 0.11 1.49 0.035 1.68 0.00055 2.62
18 rs9302001 0.79 1.03 0.82 1.03 0.43 1.11 0.86 1.03 0.38 1.17
19 rs3816995 0.059 1.20 0.15 1.17 0.35 1.13 0.09 1.31 0.12 1.32
20 rs8192506 0.97 1.01 0.79 0.93 0.87 0.95 0.57 1.24 0.89 1.06
Open in a new tab

Shaded fields indicate nominally significant results (P ≤ 0.05).

Numbers (#) correspond to those in Table 2.

PD, panic disorder; SNP, single-nucleotide polymorphism.

Statistical analysis

Allelic association tests were carried out with PLINK (Purcell et al., 2007) to generate χ2 P-values, as well as odds ratios and 95% confidence intervals. Strict Bonferroni correction would require a P-value below 0.000513 to achieve an overall significance level of 0.05 (see the Discussion section). However, given the high priors accompanying the SNPs examined, along with the nonindependent nature of some of these SNPs, we report all uncorrected P-values below 0.05 as being nominally significant (also see the remarks on issues of multiple testing in the Discussion section). With a significance of 0.05, power calculations (http://pngu.mgh.harvard.edu/~purcell/gpc/cc2.html) suggest between 72% power for our smallest stratified sample and 99% power for the total case–control sample to detect true associations, assuming a relative risk of 2.0 and complete LD between the disease allele and the marker allele.

Results

Results when all panic disorder cases are analyzed together

When we analyzed all PD cases as one group, 19 of the 20 SNPs tested failed to achieve even nominal significance, and thus did not replicate the association with PD in our sample of 351 cases and 552 controls. The one exception, rs3813034 in SLC6A4, will be discussed separately below. The only other SNP that even approached nominal statistical significance (P = 0.06) was rs3816995 in the SDK2 gene on chromosome 17 (#19 in the tables), which encodes a protein of unknown function homologous to the Drosophila melanogaster sidekick gene product and that was previously found to be associated with PD by GWAS (genome-wide association studies) (Otowa et al., 2009). The ‘all PD’ columns in Table 3 show the results.

Results when panic disorder patients are differentiated into bladder versus nonbladder

The rs3813034 SNP (#11 in the tables), which is functional in the SLC6A4 gene (see the Discussion section), yielded interesting results. Note that the sample in this paper overlaps substantially with the sample in the paper where we initially reported an association with PD (Gyawali et al., 2010; Strug et al., 2010). Thus, the positive result with this SNP does not represent an independent replication. However, what is new is that the significant result when the SNP is analyzed in all PD patients (P = 0.01) becomes even more striking when we consider only the nonbladder cases (present in 52.1% of alleles in cases vs. in 43.8% in controls, P = 0.0018), but yields no evidence at all of association in the bladder families (P = 0.93). A direct comparison of nonbladder vs. bladder cases for this SNP is significant as well (52.1% in nonbladder cases vs. 43.5% in bladder cases, P = 0.044). Note also that the frequency of the risk allele in the bladder PD cases is essentially the same as the frequency in the controls.

Two other SNPs show differences between bladder and nonbladder families – in the opposite direction from the SLC6A4 SNP. The rs5751876 SNP in the ADORA2A gene (#4 in the tables) is not significant when analyzed in all PD patients (P = 0.29) or only in the nonbladder cases (P = 0.81), but it is nominally significant when analyzed in the bladder cases alone [84 of 182 (46.2%) vs. 363 of 940 (33.1%), P = 0.046]. The rs12579350 SNP in the TMEM16B gene (#17 in the tables) shows a similar pattern, with nominal significance in bladder cases [64 of 692 (9.2%) vs. 87 of 1100 (7.9%), P = 0.035], but no significance when all cases are analyzed together or when the nonbladder cases are analyzed separately. Table 3 shows these results (‘no bladder’ and ‘bladder’ columns).

Results when only female bladder and nonbladder patients are analyzed

Both PD and its associated bladder symptoms occur significantly more frequently in women (Crowe et al., 1983; Weissman et al., 1997). The female-to-male ratio for PD is over 4 to 1 in our sample and it is over 6 to 1 for the bladder cases considered separately, whereas the 552 controls are almost identically matched for sex. Therefore, in both the nonbladder and the bladder groups, we reanalyzed the data using only the women. These analyses reduced the number of controls from 552 to 278, with the nonbladder cases reduced from 259 to 207 and the bladder cases from 92 to 80. For the rs3813034 SNP in SLC6A4 (#11), the results among women remained nominally significant in the nonbladder cases [217 of 412 alleles in cases (52.7%) vs. 241 of 556 alleles in controls (43.3%), P = 0.0041] and with no evidence of association in the bladder group (P = 0.92). For the two SNPs that had shown nominal significance in bladder cases and no significance in nonbladder cases, this pattern remained in the women-only analysis. The P-values improved slightly for #4, rs5751878 [74 of 158 (46.8%) vs. 204 of 556 (36.7%), P = 0.021], and improved considerably for #17, rs12579350 [23 of 158 (14.6%) vs. 34 of 556 (6.1%), P = 5.5 × 10−4). The odds ratio for this SNP was 2.62, with a 95% confidence interval [1.49, 4.59]. This may have been because of the fact that all 12 PD-bladder men were homozygous for the nonrisk allele. These results indicate a possible sex-specific role for the TMEM16B in the PD-bladder syndrome. The ‘no bladder (fm)’ and ‘bladder (fm)’ columns of Table 3 summarize these results.

Discussion

Summary of findings

Using our own PD sample, we examined a panel of 20 SNPs that had been reported in the literature to be associated with PD. In the most pessimistic interpretation of our results, one would look at the P-values in our ‘all PD’ analysis (first column of results in Table 3) and conclude that none appeared to be replicated (but also see the points discussed below, under the Issues of replication section), except for the rs3813034 SNP in SLC6A4 on chromosome 17 (#11 in the tables); moreover, this would not count as a replication, as the sample overlaps considerably with the sample in our original report of that association (Gyawali et al., 2010; Strug et al., 2010). [The current sample contains the 307 cases from the anxiety study (Gyawali et al., 2010; Strug et al., 2010), plus 44 additional cases from (Hamilton et al., 2003); see Table 1].

However, a more nuanced look at the results indicates intriguing differences. Our current findings suggest that the previously identified finding of an association at SLCL6A4 (#11 in the tables) may be driven by the PD cases without bladder symptoms, whereas two other loci (ADORA2A and TMEM16B, #4 and #17) provide potential evidence of an association only in the PD cases with bladder symptoms. However, it should be noted that the latter two loci are on chromosomes 22 and 12, respectively, not on 13, where we had found our peak linkage evidence. Yet, as we already have several lines of evidence pointing toward the bladder symptoms representing a separate genetic syndrome (see the Introduction section), these findings buttress the hypothesis that the ‘bladder syndrome’ represents a distinct genetic subtype of PD.

Differentiation of the female PD cases has, perhaps, less prior justification than differentiation of the bladder and nonbladder cases. However, other genes such as the PD candidate gene BDKRB2 have been shown to have sex-specific effects in a range of phenotypes (Madeddu et al., 1996; Asselbergs et al., 2006; Bourdet et al., 2010). For one of our tested SNPs, rs12579350 (#17), the change when we differentiate by sex is striking. Therefore, for future investigations into PD, it may be worthwhile to analyze female cases separately.

The SLC6A4, ADORA2A, and TMEM16B genes

Understanding the network of polymorphisms associated with PD is an important step in modeling risk and outcome for patients. More importantly, it provides an insight into the biology of disease expression, an essential component of developing more effective treatments. The genes we find to be most important in our sample by association analysis are SCL6A4, ADORA2A, and THEM16B. The SCL6A4 gene encodes the mammalian serotonin transporter and has been studied in many psychiatric disorders (Goldberg et al., 2009; Caspi et al., 2010). The pathways and signaling surrounding the neurotransmitter serotonin (5-hydroxytryptamine) have been recognized for their important roles in behavior and mood for decades. However, the role of specific genetic variants in the risk for PD has been unclear. Our studies suggest that the functional SNP rs3813034, which alters polyadenylation and the abundance of the SCL6A4 message (Gyawali et al., 2010), is involved in a subset of PD not necessarily accompanied by bladder symptoms. The ADORA2A gene has also been the focus of studies examining the genetics of mood disorders. We had previously reported linkage of ADORA2A to PD in a study of 153 families (including the families of 44 cases from this study) (Hamilton et al., 2004). Subsequently, ADORA2A was found to be associated with PD phenotypes (Deckert et al., 1998; Childs et al., 2008; Hohoff et al., 2010) and anxiety (Deckert et al., 1998; Childs et al., 2008; Hohoff et al., 2010). Deckert et al. (1998) found an association with rs5751876 in a group of 89 PD patients. This was replicated by Hohoff et al. (2010) when the sample was expanded to 457 cases. In our study, its association seems to depend on the bladder cases. Perhaps the most unexpected association in our study is that found in TMEM16B. The TMEM16B gene, also known as ANO2, encodes the Ca2+ -activated chloride channel family member anoctamin 2 (Hartzell et al., 2009; Stephan et al., 2009; Stohr et al., 2009). The rs12579350 SNP of this gene had yielded a genome-wide P-value of 3.7 × 10−9 in a Japanese GWAS of 200 PD cases and 200 non-PD controls (Otowa et al., 2009), although it was not replicated in a follow-up study that increased the sample size (Otowa et al., 2010). In this study, rs12579350 appears to be associated with PD bladder, particularly in women. The biological connection of TMEM16B to PD is unclear, but provides an impetus for further functional studies. These three PD-associated genes, each associated with a slightly different phenotype, may help us understand the biological networks contributing to different kinds of PD syndromes.

Issues of replication

For all complex diseases, especially psychiatric disorders, finding genetic associations that hold up under further scrutiny has proven challenging. It is widely accepted that if an association is not replicated, then the initial finding was likely a false positive (Greene et al., 2009). However, this is not necessarily the case. Possible alternative explanations include the following.

One alternative explanation is the ‘replication fallacy,’ already known in some areas of applied statistics, and elucidated with particular application to genetic association studies by Gorroochurn et al. (2007). Depending on how low the P-value in an initial study was, a second study may have a probability as low as 50% of replicating the study in a sample of the same size, even when the initial study’s result was in fact a true positive. If the second study has a smaller sample than the first study, the replication probability can be even lower than 50%.

A second possibility is that the initial association may be real for the sample collected, but not for a subsequent independently sampled population (Vieland, 2001). With complex diseases, it is very difficult to ensure that the sample collected for the follow-up study is truly comparable with the sample used in the initial study.

Third, there may be unrecognized etiological heterogeneity within any one sample. This is where very careful attention to phenotype can play a critical role. For example, a study of the COMT V158M (rs4680) polymorphism (Domschke et al., 2008) stratified PD cases by neuronal response to fearful faces and in this way was able to replicate a previous reported association. In the current study, the differentiation of bladder from nonbladder cases yields a quite different picture than analysis of all PD cases together. This separation is not ad hoc but is based on our published studies – both linkage and association – indicating genetic differences between the two types of cases.

Finally, there is the question of what significance level may be required in a follow-up study. As pointed out by Vieland (2001), if a significance level of α1 is required in an initial study and a significance level of α2 in a follow-up study, a significance level of α1 × α2 overall is actually required. Thus, if both α1 and α2 are set at 0.05, a significance level of 0.0025 may actually be required. At first glance, it may seem harmless to set such a strict criterion, but unfortunately if one does so, one pays the price in greatly reduced statistical power.

Thus, although replication studies are certainly important, they do not necessarily provide definitive answers to our questions, even when they are negative.

Issues of multiple testing

When attempting to ‘correct’ for multiple analyses on data, logical difficulties and contradictions may develop; in fact, for this reason, some investigators recommend following a completely different paradigm for statistical testing (Royall, 1997, 2000; Strug and Hodge, 2006a, 2006b). However, in the current study, we followed the standard statistical paradigm, and thus we need to address the issue of multiple testing.

The underlying concern in multiple testing arises from the fact that if a number of statistical tests are carried out at, say, the 5% level, the probability that at least one of these tests will yield a ‘significant’ result, if there is no association with any of the markers, will be greater than 5%. We can designate 5% as the ‘nominal’ or the ‘desired’ type I error level, whereas the true probability, the increased one, is the ‘corrected’ level. Turning the question around, if N tests are carried out and a P-value of, say, 0.001 is obtained, this should be considered the nominal P-value; to determine the corrected P-value, the Bonferroni correction formula Pcorr = 1 – (1 – Pnom)N, or, in the usual approximation, Pcorr ≈ N × Pnom, should be used.

However, applying this Bonferroni correction leads to two questions. (i) It assumes the N analyses are all statistically independent; if they are correlated, then the approach is overconservative and the Bonferroni approach loses power. (ii) It assumes that none of the N analyses has prior support; thus, this approach is appropriate for a GWAS, for example. However, this logic does not allow for the possibility that some or all of the analyses are potentially replicating earlier studies or that there may be biological arguments supporting any of the markers.

Returning to our study – on the face of it, Table 3 shows 100 analyses (five each of 20 markers). Of the eight P-values below 0.05, only one – rs12579350 (#17) – is small enough to ‘almost’ survive the standard Bonferroni correction with N = 100: inserting Pnom = 0.0005 and N = 100 into the Bonferroni formula above yields Pcorr = 1 – (0.99945)100 ≅ 0.0535. However, both the issues raised above have relevance for our study: the five tests carried out on each marker are not independent, and therefore, the 100 analyses are not independent [cf. issue (i)]; all 20 loci had been published previously, with some of them having already been replicated [issue (ii)]; and, as discussed in the Introduction section, we have strong previous evidence that the bladder syndrome is a separate genetic entity within PD [also issue (ii)].

Because of these complications, we cannot state exactly what the ‘corrected’ P-value(s) should be for this study, but we believe that it is reasonable to argue that our results – those for rs5751876 and rs4822492 in ADORA2A, rs3813034 in SLC6A4, and rs12579350 in TMEM16B, as well as those supporting the separation of bladder from nonbladder families, and the possible differentiation of women from men with PD – should not be dismissed automatically because of multiple tests. They deserve serious consideration, and we hope that we or other investigators will be able to study them further, toward the goal of either replicating or rejecting them.

Conclusion

Achieving definitive and unassailable results in studying the genetics of complex diseases is difficult. Findings that do not meet the general standards for statistical significance need to be replicated in an independent sample before strong conclusions can be drawn. We do not claim that our findings are definitive. Nevertheless, they suggest the potential value of differentiating families with bladder symptoms in future genetic studies of PD.

Acknowledgments

This work was supported by funds from the Interstitial Cystitis Association (PI M.M.W.), National Institutes of Mental Health Grants MH 28274 (to M.M.W.), PO1MH60970 (M.M.W. PI of clinical core), MH37592 (to A.J.F.), MH65213 (R.L.S., S.E.H. PI), and MH48858 (to S.E.H.). S.P.H. was supported by a UCSF bridge award. Dr Talati was funded by a K Award from the National Institute for Drug Abuse (1 K01 DA029598), a NARSAD Young Investigator Award, and support from the Research Foundation for Mental Hygiene Inc. In the past 2 years, Dr Weissman received funding from the National Institute of Mental Health (NIMH), the National Institute on Drug Abuse (NIDA), the National Alliance for Research on Schizophrenia and Depression (NARSAD), the Sackler Foundation, the Templeton Foundation and the Interstitial Cystitis Association; and receives royalties from the Oxford University Press, Perseus Press, the American Psychiatric Association Press, and MultiHealth Systems.

Genotyping was carried out at the Roswell Park Cancer Institute Genomic Shared Resources (Buffalo, NY).

The authors thank Kenia Gomez of Department of Psychiatry, Columbia University for assistance in sample organization.

The authors thank the families and individuals who participated in this study, without whose help this work would not have been possible.

Footnotes

Conflicts of interest

There are no conflicts of interest.

References

  1. Asselbergs FW, Williams SM, Hebert PR, Coffey CS, Hillege HL, Navis G, et al. The gender-specific role of polymorphisms from the fibrinolytic, reninangiotensin, and bradykinin systems in determining plasma t-PA and PAI-1 levels. Thromb Haemost. 2006;96:471–477. [PubMed] [Google Scholar]
  2. Bourdet B, Pécher C, Minville V, Jaafar A, Allard J, Blaes N, et al. Distribution and expression of B2-kinin receptor on human leukocyte subsets in young adults and elderly using flow cytometry. Neuropeptides. 2010;44:155–161. doi: 10.1016/j.npep.2009.12.005. [DOI] [PubMed] [Google Scholar]
  3. Caspi A, Hariri AR, Holmes A, Uher R, Moffitt TE. Genetic sensitivity to the environment: the case of the serotonin transporter gene and its implications for studying complex diseases and traits. Am J Psychiatry. 2010;167:509–527. doi: 10.1176/appi.ajp.2010.09101452. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Childs E, Hohoff C, Deckert J, Xu K, Badner J, De Wit H. Association between ADORA2A and DRD2 polymorphisms and caffeine-induced anxiety. Neuropsychopharmacology. 2008;33:2791–2800. doi: 10.1038/npp.2008.17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Crowe RR, Noyes R, Pauls DL, Slymen D. A family study of panic disorder. Arch Gen Psychiatry. 1983;40:1065–1069. doi: 10.1001/archpsyc.1983.01790090027004. [DOI] [PubMed] [Google Scholar]
  6. Deckert J, Nothen MM, Franke P, Delmo C, Fritze J, Knapp M, et al. Systematic mutation screening and association study of the A1 and A2a adenosine receptor genes in panic disorder suggest a contribution of the A2a gene to the development of disease. Mol Psychiatry. 1998;3:81–85. doi: 10.1038/sj.mp.4000345. [DOI] [PubMed] [Google Scholar]
  7. Domschke K, Ohrmann P, Braun M, Suslow T, Bauer J, Hohoff C, et al. Influence of the catechol-O-methyltransferase val158met genotype on amygdala and prefrontal cortex emotional processing in panic disorder. Psychiatry Res. 2008;163:13–20. doi: 10.1016/j.pscychresns.2007.04.016. [DOI] [PubMed] [Google Scholar]
  8. Fyer AJ, Weissman MM. Genetic linkage study of panic: clinical methodology and description of pedigrees. Am J Med Genet. 1999;88:173–181. [PubMed] [Google Scholar]
  9. Goldberg TE, Kotov R, Lee AT, Gregersen PK, Lencz T, Bromet E, Malhotra AK. The serotonin transporter gene and disease modification in psychosis: evidence for systematic differences in allelic directionality at the 5-HTTLPR locus. Schizophr Res. 2009;111:103–108. doi: 10.1016/j.schres.2009.03.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Gorman JM, Goetz RR, Fyer M, King DL, Fyer AJ, Liebowitz MR, Klein DF. The mitral valve prolapse–panic disorder connection. Psychosom Med. 1988;50:114–122. doi: 10.1097/00006842-198803000-00002. [DOI] [PubMed] [Google Scholar]
  11. Gorroochurn P, Hodge SE, Heiman GA, Durner M, Greenberg DA. Non-replication of association studies: ‘pseudo-failures’ to replicate? Genet Med. 2007;9:325–331. doi: 10.1097/gim.0b013e3180676d79. [DOI] [PubMed] [Google Scholar]
  12. Gratacos M, Costas J, De Cid R, Bayes M, Gonzalez JR, Baca-Garcia E, et al. Identification of new putative susceptibility genes for several psychiatric disorders by association analysis of regulatory and non-synonymous SNPs of 306 genes involved in neurotransmission and neurodevelopment. Am J Med Genet B Neuropsychiatr Genet. 2009;150B:808–816. doi: 10.1002/ajmg.b.30902. [DOI] [PubMed] [Google Scholar]
  13. Greenberg D. There is more than one way to collect data for linkage analysis. What a study of epilepsy can tell us about linkage strategy for psychiatric disease. Arch Gen Psychiatry. 1992;49:745–750. doi: 10.1001/archpsyc.1992.01820090073012. [DOI] [PubMed] [Google Scholar]
  14. Greene CS, Penrod NM, Williams SM, Moore JH. Failure to replicate a genetic association may provide important clues about genetic architecture. PLoS One. 2009;4:e5639. doi: 10.1371/journal.pone.0005639. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Gyawali S, Subaran R, Weissman MM, Hershkowitz D, McKenna MC, Talati A, et al. Association of a polyadenylation polymorphism in the serotonin transporter and panic disorder. Biol Psychiatry. 2010;67:331–338. doi: 10.1016/j.biopsych.2009.10.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Hamilton SP. Linkage and association studies of anxiety disorders. Depress Anxiety. 2009;26:976–983. doi: 10.1002/da.20615. [DOI] [PubMed] [Google Scholar]
  17. Hamilton SP, Fyer AJ, Durner M, Heiman GA, Baisre De Leon A, Hodge SE, et al. Further genetic evidence for a panic disorder syndrome mapping to chromosome 13q. Proc Natl Acad Sci USA. 2003;100:2550–2555. doi: 10.1073/pnas.0335669100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Hamilton SP, Slager SL, De Leon AB, Heiman GA, Klein DF, Hodge SE, et al. Evidence for genetic linkage between a polymorphism in the adenosine 2A receptor and panic disorder. Neuropsychopharmacology. 2004;29:558–565. doi: 10.1038/sj.npp.1300311. [DOI] [PubMed] [Google Scholar]
  19. Hanno PM. Diagnosis of interstitial cystitis. Urol Clin North Am. 1994;21:63–66. [PubMed] [Google Scholar]
  20. Hartzell HC, Yu K, Xiao Q, Chien LT, Qu Z. Anoctamin/TMEM16 family members are Ca2 + -activated Cl− channels. J Physiol. 2009;587:2127–2139. doi: 10.1113/jphysiol.2008.163709. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Hettema JM, Neale MC, Kendler KS. A review and meta-analysis of the genetic epidemiology of anxiety disorders. Am J Psychiatry. 2001;158:1568–1578. doi: 10.1176/appi.ajp.158.10.1568. [DOI] [PubMed] [Google Scholar]
  22. Hohoff C, Mullings EL, Heatherley SV, Freitag CM, Neumann LC, Domschke K, et al. Adenosine A(2A) receptor gene: evidence for association of risk variants with panic disorder and anxious personality. J Psychiatr Res. 2010;44:930–937. doi: 10.1016/j.jpsychires.2010.02.006. [DOI] [PubMed] [Google Scholar]
  23. Ioannidis JP, Thomas G, Daly MJ. Validating, augmenting and refining genome-wide association signals. Nat Rev Genet. 2009;10:318–329. doi: 10.1038/nrg2544. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Jones CA, Nyberg L. Epidemiology of interstitial cystitis. Urology. 1997;49:2–9. doi: 10.1016/s0090-4295(99)80327-6. [DOI] [PubMed] [Google Scholar]
  25. Kessler RC, Chiu WT, Jin R, Ruscio AM, Shear K, Walters EE. The epidemiology of panic attacks, panic disorder, and agoraphobia in the National Comorbidity Survey Replication. Arch Gen Psychiatry. 2006;63:415–424. doi: 10.1001/archpsyc.63.4.415. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Lydiard RB, Greenwald S, Weissman MM, Johnson J, Drossman DA, Ballenger JC. Panic disorder and gastrointestinal symptoms: findings from the NIMH Epidemiologic Catchment Area project. Am J Psychiatry. 1994;151:64–70. doi: 10.1176/ajp.151.1.64. [DOI] [PubMed] [Google Scholar]
  27. Madeddu P, Pinna Parpaglia P, Anania V, Glorioso N, Chao C, Wang C, Chao J. Sexual dimorphism of cardiovascular responses to early blockade of bradykinin receptors. Hypertension. 1996;27:746–751. doi: 10.1161/01.hyp.27.3.746. [DOI] [PubMed] [Google Scholar]
  28. Maron E, Nikopensius T, Koks S, Altmae S, Heinaste E, Vabrit K, et al. Association study of 90 candidate gene polymorphisms in panic disorder. Psychiatr Genet. 2005;15:17–24. doi: 10.1097/00041444-200503000-00004. [DOI] [PubMed] [Google Scholar]
  29. Maron E, Toru I, Must A, Tasa G, Toover E, Vasar V, et al. Association study of tryptophan hydroxylase 2 gene polymorphisms in panic disorder. Neurosci Lett. 2007;411:180–184. doi: 10.1016/j.neulet.2006.09.060. [DOI] [PubMed] [Google Scholar]
  30. Maron E, Toru I, Tasa G, Must A, Toover E, Lang A, et al. Association testing of panic disorder candidate genes using CCK-4 challenge in healthy volunteers. Neurosci Lett. 2008;446:88–92. doi: 10.1016/j.neulet.2008.09.052. [DOI] [PubMed] [Google Scholar]
  31. Maron E, Hettema JM, Shlik J. Advances in molecular genetics of panic disorder. Mol Psychiatry. 2010;15:681–701. doi: 10.1038/mp.2009.145. [DOI] [PubMed] [Google Scholar]
  32. Otowa T, Yoshida E, Sugaya N, Yasuda S, Nishimura Y, Inoue K, et al. Genome-wide association study of panic disorder in the Japanese population. J Hum Genet. 2009;54:122–126. doi: 10.1038/jhg.2008.17. [DOI] [PubMed] [Google Scholar]
  33. Otowa T, Tanii H, Sugaya N, Yoshida E, Inoue K, Yasuda S, et al. Replication of a genome-wide association study of panic disorder in a Japanese population. J Hum Genet. 2010;55:91–96. doi: 10.1038/jhg.2009.127. [DOI] [PubMed] [Google Scholar]
  34. Placidi GP, Boldrini M, Patronelli A, Fiore E, Chiovato L, Perugi G, Marazziti D. Prevalence of psychiatric disorders in thyroid diseased patients. Neuropsychobiology. 1998;38:222–225. doi: 10.1159/000026545. [DOI] [PubMed] [Google Scholar]
  35. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet. 2007;81:559–575. doi: 10.1086/519795. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Rothe C, Koszycki D, Bradwejn J, King N, Deluca V, Tharmalingam S, et al. Association of the Val158Met catechol O-methyltransferase genetic polymorphism with panic disorder. Neuropsychopharmacology. 2006;31:2237–2242. doi: 10.1038/sj.npp.1301048. [DOI] [PubMed] [Google Scholar]
  37. Royall R. Statistical evidence: a likelihood paradigm. New York: Chapman & Hall; 1997. [Google Scholar]
  38. Royall R. On the probability of misleading evidence. J Am Stat Assoc. 2000;95:760–768. [Google Scholar]
  39. Schumacher J, Kristensen AS, Wendland JR, Nothen MM, Mors O, Mcmahon FJ. The genetics of panic disorder. J Med Genet. 2011;48:361–368. doi: 10.1136/jmg.2010.086876. [DOI] [PubMed] [Google Scholar]
  40. Stephan AB, Shum EY, Hirsh S, Cygnar KD, Reisert J, Zhao H. ANO2 is the cilial calcium-activated chloride channel that may mediate olfactory amplification. Proc Natl Acad Sci USA. 2009;106:11776–11781. doi: 10.1073/pnas.0903304106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Stewart W, Breslau N, Keck PE., Jr Comorbidity of migraine and panic disorder. Neurology. 1994;44:S23–S27. [PubMed] [Google Scholar]
  42. Stohr H, Heisig JB, Benz PM, Schoberl S, Milenkovic VM, Strauss O, et al. TMEM16B, a novel protein with calcium-dependent chloride channel activity, associates with a presynaptic protein complex in photoreceptor terminals. J Neurosci. 2009;29:6809–6818. doi: 10.1523/JNEUROSCI.5546-08.2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Strug LJ, Hodge SE. An alternative foundation for the planning and evaluation of linkage analysis. I. Decoupling ‘error probabilities’ from ‘measures of evidence’. Hum Hered. 2006a;61:166–188. doi: 10.1159/000094709. [DOI] [PubMed] [Google Scholar]
  44. Strug LJ, Hodge SE. An alternative foundation for the planning and evaluation of linkage analysis. II. Implications for multiple test adjustments. Hum Hered. 2006b;61:200–209. doi: 10.1159/000094775. [DOI] [PubMed] [Google Scholar]
  45. Strug LJ, Suresh R, Fyer AJ, Talati A, Adams PB, Li W, et al. Panic disorder is associated with the serotonin transporter gene (SLC6A4) but not the promoter region (5-HTTLPR) Mol Psychiatry. 2010;15:166–176. doi: 10.1038/mp.2008.79. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Talati A, Ponniah K, Strug LJ, Hodge SE, Fyer AJ, Weissman MM. Panic disorder, social anxiety disorder, and a possible medical syndrome previously linked to chromosome 13. Biol Psychiatry. 2008;63:594–601. doi: 10.1016/j.biopsych.2007.07.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Thoeringer CK, Binder EB, Salyakina D, Erhardt A, Ising M, Unschuld PG, et al. Association of a Met88Val diazepam binding inhibitor (DBI) gene polymorphism and anxiety disorders with panic attacks. J Psychiatr Res. 2007;41:579–584. doi: 10.1016/j.jpsychires.2006.06.001. [DOI] [PubMed] [Google Scholar]
  48. Thorgeirsson TE, Oskarsson H, Desnica N, Kostic JP, Stefansson JG, Kolbeinsson H, et al. Anxiety with panic disorder linked to chromosome 9q in Iceland. Am J Hum Genet. 2003;72:1221–1230. doi: 10.1086/375141. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Unschuld PG, Ising M, Erhardt A, Lucae S, Kloiber S, Kohli M, et al. Polymorphisms in the serotonin receptor gene HTR2A are associated with quantitative traits in panic disorder. Am J Med Genet B Neuropsychiatr Genet. 2007;144B:424–429. doi: 10.1002/ajmg.b.30412. [DOI] [PubMed] [Google Scholar]
  50. Vieland VJ. The replication requirement. Nat Genet. 2001;29:244–245. doi: 10.1038/ng1101-244. [DOI] [PubMed] [Google Scholar]
  51. Weissman MM, Bland RC, Canino GJ, Faravelli C, Greenwald S, Hwu HG, et al. The cross-national epidemiology of panic disorder. Arch Gen Psychiatry. 1997;54:305–309. doi: 10.1001/archpsyc.1997.01830160021003. [DOI] [PubMed] [Google Scholar]
  52. Weissman MM, Fyer AJ, Haghighi F, Heiman G, Deng Z, Hen R, et al. Potential panic disorder syndrome: clinical and genetic linkage evidence. Am J Med Genet. 2000;96:24–35. doi: 10.1002/(sici)1096-8628(20000207)96:1<24::aid-ajmg7>3.0.co;2-e. [DOI] [PubMed] [Google Scholar]
  53. Weissman MM, Gross R, Fyer A, Heiman GA, Gameroff MJ, Hodge SE, et al. Interstitial cystitis and panic disorder: a potential genetic syndrome. Arch Gen Psychiatry. 2004;61:273–279. doi: 10.1001/archpsyc.61.3.273. [DOI] [PubMed] [Google Scholar]
  54. Zaubler TS, Katon W. Panic disorder and medical comorbidity: a review of the medical and psychiatric literature. Bull Menninger Clin. 1996;60:A12–A38. [PubMed] [Google Scholar]

RESOURCES