Genetic Biomarkers of Panic Disorder: A Systematic Review

(1) Background: Although panic disorder (PD) is one of the most common anxiety disorders severely impacting quality of life, no effective genetic testing exists; known data on possible genetic biomarkers is often scattered and unsystematic which complicates further studies. (2) Methods: We used PathwayStudio 12.3 (Elsevier, The Netherlands) to acquire literature data for further manual review and analysis. 229 articles were extracted, 55 articles reporting associations, and 32 articles reporting no associations were finally selected. (3) Results: We provide exhaustive information on genetic biomarkers associated with PD known in the scientific literature. Data is presented in two tables. Genes COMT and SLC6A4 may be considered the most promising for PD diagnostic to date. (4) Conclusions: This review illustrates current progress in association studies of PD and may indicate possible molecular mechanisms of its pathogenesis. This is a possible basis for data analysis, novel experimental studies, or developing test systems and personalized treatment approaches.


Introduction
Anxiety disorders are a very wide group of mental disorders which includes, according to ICD-11 (World Health Organization, 2018), a variety of conditions with panic disorder (PD) being one of the most common and often chronic [1,2]. It is predominantly characterized by panic attacks and anticipatory anxiety. Many researchers report that PD can noticeably reduce the quality of a patient's life [3]. Moreover, Sherbourne et al. (1996) report that PD may alter a patient's quality of life is even more dramatically than other severe chronic diseases including diabetes, cardiovascular and lung diseases [4].
The prevalence of PD is estimated at 1-3%, with females suffering from it twice as frequently as males [5]. Family and twin studies estimated PD heritability at 0.48 [6], with a relative risk of PD in proband's first-degree relatives being 6-17 times higher than average risk across the whole population [7,8].
Concordance in monozygotic twins is approximately 20.7-73.0% [6]. Both genetic and environmental factors are considered to be involved in the development of PD. It has been shown that being in stressful conditions can provoke panic attacks in people predisposed to anxiety [9].
Molecular genetic markers, along with psychological examination and evaluation of environmental factors causing stress, are promising in regard to creating effective PD-predisposition tests. This is especially important for people who are frequently exposed to stressful conditions due to their profession and/or lifestyle. Development of such predisposition tests may help to take PD-risks into 1.
Searching for genes (proteins) linked to PD through reported polymorphism associations. The initial list of genes was acquired using PathwayStudio 12.3 software. Search was performed by adding "Panic Disorder" object and extracting all proteins and complexes linked to "Panic Disorder". Linkage type: GeneticChange was used. This type allows to search for reported polymorphism-condition associations. In accordance to ResNet and Pathway Studio features, the acquired list included not only protein-encoding genes but microRNA-encoding genes as well, if they are reportedly linked to PD.

2.
Acquiring a reference table via "View relation table". This function provides access to reference IDs and data-mined "Sentences" containing association reports from original articles. On this step additional search was performed using PubMed (http://www.ncbi.nlm.nih.gov/pubmed/), TargetInsights (https://demo.elseviertextmining.com/) and GoogleScholar (https://scholar.google. ru/). A total of 229 references supporting each gene-PD link were identified and taken into further screening.

3.
Screening was performed by manually analyzing "Sentences" extracted by Pathway Studio from full-text articles. All Sentences were carefully revised by at least three authors, full texts were manually studied for additional details, if necessary. All irrelevant studies identified in this step were excluded from further analysis. A total of 127 irrelevant studies were identified, including 50 literature reviews with no original experiments; 50 articles on irrelevant topics misinterpreted by text-mining algorithms; 27 studies involving model organisms, and lacking human patient studies.

4.
Full texts of 102 remaining articles were manually revised and assessed for eligibility by authors.
Original research papers with statistically significant results were taken into the further analysis: positive association of certain gene modification (or one of its variants) with the disease; interaction with another gene; association of genotypes or alleles of different genes; the presence of the mutation in families with high occurrence of the disease; cosegregation of genetic alteration with manifestations of the disease.
a. All articles reporting lack of association were transferred into a separate "No Association" table. This table was analyzed separately (see Step 6). This also included studies reporting an association with the particular symptoms but no association with the disorder in general. A total of 32 articles were excluded due to a lack of significant association. b.
All remaining articles reporting results irrelevant to our study (i.e., treatment response studies) were excluded. A total of 15 articles were excluded due to these reasons.

5.
The remaining 55 articles formed the basis of our review. Each paper was manually studied for association availability (significance was assessed by revising p-values). Official gene names, polymorphism IDs, sample sizes, p-values for each association, and other relevant data were manually extracted and revised. All articles reporting significant associations were considered eligible. Small sample size was not considered an exclusion criterion. However, we are aware that the sample size may result in significant bias. To allow estimating this bias we provide sample sizes in the Sample column. 6.
After analyzing the articles reporting associations, we performed a manual analysis of the "No Association" table. The analysis was performed by additional full-text revision and extraction of relevant data.
The algorithm scheme, along with the number of articles added or excluded on each step is presented in Figure 1.
Genes 2020, 11, x FOR PEER REVIEW  3 of 27 polymorphism IDs, sample sizes, p-values for each association, and other relevant data were manually extracted and revised. All articles reporting significant associations were considered eligible. Small sample size was not considered an exclusion criterion. However, we are aware that the sample size may result in significant bias. To allow estimating this bias we provide sample sizes in the Sample column. 6. After analyzing the articles reporting associations, we performed a manual analysis of the "No Association" table. The analysis was performed by additional full-text revision and extraction of relevant data.
The algorithm scheme, along with the number of articles added or excluded on each step is presented in Figure 1.

Results
A total of 229 studies were analyzed, with 127 being excluded during screening and 15 being excluded during other steps. The remaining articles were taken into further analysis and separated into two tables. Table 1 provides the results of the analysis of 55 articles reporting association. Studies reporting lack of association were analyzed separately, results of analyzing 32 studies are provided in Table 2.

Results
A total of 229 studies were analyzed, with 127 being excluded during screening and 15 being excluded during other steps. The remaining articles were taken into further analysis and separated into two tables. Table 1 provides the results of the analysis of 55 articles reporting association. Studies reporting lack of association were analyzed separately, results of analyzing 32 studies are provided in Table 2. Allele Ins is more frequent in patients (p = 0.002). Genotype ID is associated with PD (p = 0.003). Patients with allele D had higher risk of respiratory type PD (p = 0.034).
Genotype-allele haplotype rs28632197:TT (AVPR1B) + rs878886:G (CRHR1) shows strongest association with PD (p = 0.00057). 15 more associations below the significance threshold found. [15] BDNF (brain-derived neurotrophic factor) BDNF is one of the most abundant neurotrophins found in brain tissues. It regulates neuronal plasticity: the formation and maintenance of synaptic contacts. May be involved in the pathogenesis of mental diseases including major depression and PD. Haplotype GC (rs6265:G, rs16917204:C) is more frequent in patients (p = 0.0009). No association with single SNPs. A total of 3 SNPs studied. [16]  Cholecystokinins are neuropeptides that play role in satiety and anxiety. CCK-4 tetrapeptide s ability to induce symptoms of panic attacks is known.
[17] SNP is associated with PD (p = 0.004). An analysis using all alleles was also significant (p = 0.038). A total of 16 alleles of CCKBR and several CCK and CCKAR polymorphisms were studied. [21] CNR1 (cannabinoid receptor 1) Endocannabinoid system may play role in the regulation of stress, anxiety, depression, and addictive disorders.
Allele rs12720071:G is associated with PD (p = 0.012) and more frequent in females. Total of 12 SNPs studied.
See also CNR2.

[22]
COMT (catechol-O-methyltransferase) COMT enzyme is involved in the inactivation of the catecholamine neurotransmitters. The link between COMT activity and depressive and anxiety disorders is known, as well as its involvement in the associated biochemical processes.
[ Genotype rs4680:AA is associated with PD and increases anxiety (p = 0.01). [26]  Rare allele rs8192506:G is found to be protective and almost three times more frequent in controls (p = 0.032). [33] DRD1 is known to be linked with depression and anxiety.   Germany.
Epigenetic study. Lower hypermethylation of the FOXP3 promoter region is found in female PD patient subsample compared to control (p = 0.005). The male subsample lacks this association. [34] GABRA6 (γ-aminobutyric acid receptor, α 6 subunit) GABRA6 subunit is well known for its link to anxiety in mammals. It is the main inhibitor of neurotransmitters in the CNS which regulates several physiological and psychological processes such as anxiety and depression. Allele rs3219151:T and genotype rs3219151:TT are associated with PD (p = 2.00 × 10 −7 , p = 2.18 × 10 −6 , respectively). Patients carrying genotype rs11060369:AA (TMEM132D), and allele rs3219151:T (GABRA6) had significantly stronger response to frightening image demonstration (both p < 0.01). A total of 5 SNPs were studied. See also COMT and TMEM132D. [24] GABRA5 (γ-aminobutyric acid type A receptor alpha5 subunit) The GABA-ergic system is inhibitory. GABA receptors are targets for benzodiazepines that are used for PD treatment. SNP rs35399885 is associated with PD (p = 0.05).
A total of 10 SNPs in GABRA3 and GABRA5 were studied. See also GABRA3.  Germany.
Negative life events correlated with a decreased mean methylation in patients-the correlation was observed mainly in the female subsample (p = 0.01). A total of 38 methylation sites in promotor/2 intron of GAD1 and 10 promoter sites in GAD2 were studied. Germany.
Only rs3749034:A × gender remained significant in a combined sample (p = 0.045), but not in the replication sample. A total of 13 SNPs studied. [37] GHRL (ghrelin and obestatin prepropeptide) Ghrelin is well-known for its anxiogenic and anxiolytic effects in rodents. Obestatin in contrast, decreases anxiety-like behavior in rodents. Hypocretin receptor 2 is expressed exclusively in the brain. Its agonist orexin-A may have an anxiogenic effect in rodents. It is also may be linked to PD via its role in respiration which seems to be altered in PD. Allele rs2653349:A is associated with PD in total sample and female subsample (p = 0.015, p = 0.0015 respectively).
Other studied SNP rs2271933 (HCRTR1) showed no association. [40] HTR1A is known to be associated with anxiety disorder and anxiogenic stimuli response. Allele rs6295:G is associated with PD with agoraphobia (p = 0.03). [27] HTR2A (5-hydroxytryptamine receptor 2A) HTR2A over-and under-expression is known in PD patients and thus may lead to PD.
In Spanish sample: SNPs rs11763020 (p < 0.00008) is associated with PD. Associations in other samples lack significance. A total of 712 SNPs in MIR genes were studied. See also MIR22. See also MIR22. [50] NPS (neuropeptide S) Neuropeptide S can produce behavioral arousal and anxiolytic-like effects in rodents. Sweden.
Allele rs10895068:A is more frequent in patients (p = 0.01). Association is found in female subsample (p = 0.0009), male subsample lacks association. PROGINS insertion was also studied, no association found. [57] RGS2 (regulator of G protein signaling 2) RGS2 protein regulates G protein signaling activity and modulates receptor signaling neurotransmitters involved in the pathogenesis of anxiety diseases. Allele rs10801153:G is associated with PD in a combined and 2nd sample (p = 0.017 for the combined sample). 5 SNPs (rs16834831, rs16829458, rs1342809, rs1890397, rs4606) in RGS2 were also studied, no association found. [59] SLC6A4 (solute carrier family 6 member 1, SERT) The norepinephrine transporter is responsible for the reuptake of norepinephrine and dopamine into the presynaptic nerve endings. Transmembrane protein 132D is associated with the severity of anxiety symptoms in psychiatric patients. A gene may be involved in anxiety (anxiety-related behavior).
Five patient samples were studied. In the combined sample: Allele rs7309727:C is found to be risk (p = 1.1 × 10 −8 ).
No significance in the Japanese sample. A total of 4 SNPs were studied. [64] 1 Official gene name according to NCBI database. 2 Possible reasons for the association, including possible molecular/cellular level mechanisms. 3 Polymorphic locus identifiers (if present), if associated with a disease in conjunction with another marker, this marker is also included. 4 Sample parameters, including country, gender ratio, and other possibly important information (i.e., concomitant diseases) for patients and controls if this information is given in the original paper. 5 A summary of the authors' findings regarding the association of polymorphisms with PD. 6 Link to the original paper.    No significant associations found. [88] (1) nP = 88. Germany.
The authors tested the hypothesis that the heterogeneity of CO2 reactivity in patients with PD and its possible relation to 5-HTT promoter polymorphism. No association found. [90]  No significant associations found. [88] 218 A/C nP = 107, nC = 161. South Korea.
No significant associations found. [82] 1 Official gene name according to NCBI database. 2 Polymorphic locus identifiers (if present). 3 Sample parameters, including country, gender ratio, and other possibly important information (i.e., concomitant diseases) for patients and controls if this information is given in the original paper. 4 A summary of the authors' findings regarding the association of polymorphisms with PD. 5 Link to the original paper.

Discussion
Our analysis revealed 40 genes linked to PD. Among them, the majority of associations (in 38 genes) were marked by at least one SNP, 3 (5HTR2A, CCKBR, RGS2) had additional repeat/deletion polymorphism markers, 1 (COMT) had additional microsatellite polymorphism marker, and 1 (GAD1) had additional methylation alterations associated with PD. The remaining 2 genes (FOXP3, MAOA) were associated with PD via methylation alterations with no reported associated SNPs.
For the COMT gene, its' link to PD was supported by 9 studies which may be considered the most reliable or most well-studied gene associated with PD. For the majority (25 out of 40) of associations, the link to PD was supported by just 1 study. Associations of polymorphisms in CCKBR, HTR1A, and TPH2 genes were supported by 3 studies. Remaining associations in genes 5HTR2A, ACE, ADORA2A, ASIC1, CRHR1, GAD1, MAOA, MIR22, NPS, PDE4B, and SLC6A4 were supported by 2 studies.
COMT is also among the top genes by the number of associated polymorphisms, however, it only has 5 associated markers, while SLC6A4 has 8 associated markers and ADORA2A, and TMEM132D have 6 each.
The majority of genes (16 out of 40) have 2 associated polymorphisms, 7 out of 40 have 3 or 4 associated polymorphisms.
The rest of the genes can be attributed to the following functional categories: • 13%-involved in the regulation of the work of genes (including miRNA); FOXP3, IKBKE, MIR22, MIR339, MIR491. • 10%-involved in the reception of extracellular signals and system of secondary messengers functioning; NTRK3, PDE4B, RGS2, TMEM132D. • 5%-involved in the general functioning of neurons; ASIC1, BDNF. • 5%-not fitting to any of these categories others; GLO1, MBL2.
Among the 30 non-associated genes, 16 are present in both tables while 14 are not present in Table 1.
This data represents studies available via ResNet, Google Scholar, PubMed, and TargetInsight databases. Despite the large number of articles covered, several limitations are present. First of all, some articles may be absent from the listed databases. This means that our analysis is not completely exhaustive. Second, only articles available in English were included.
In our work, we summarized literature data on PD biomarkers. Our review may provide a strong base for creating effective test-systems and predisposition tests for early PD diagnostics. This is especially important for people working or living in stressful conditions, as PD is known to be triggered by environmental factors and causes a severe reduction in quality of life [4,93]. Moreover, our study as well as the following studies may be used to reveal possible pathogenesis mechanisms and signal pathways underlying PD. This knowledge is not just theoretical as it may allow developing a personalized approach for each PD case. Personalized medicine is considered a promising field in the case of PD due to the lack of serious advances in treatment in recent years [94]. We believe that recently proposed for personalized psychiatry machine learning approaches already applied to some PD cases [95] may be even more effective if proper genetic screening is conducted. In light of this, reviewing and systemizing PD biomarkers is the way to bring technology and medicine even closer together and revolutionize the field of personalized psychiatry.

Conclusions
The information summarized in our work illustrates current progress in association studies of panic disorder and may indicate possible molecular mechanisms of the pathogenesis of panic disorder. This information can serve as a basis for both further data analysis, novel experimental studies, or creating effective test-systems and personalizing medicine in the future.