- Research article
- Open Access
- Open Peer Review
Phenotype, origin and estimated prevalence of a common long QT syndrome mutation: a clinical, genealogical and molecular genetics study including Swedish R518X/KCNQ1families
© Winbo et al.; licensee BioMed Central Ltd. 2014
- Received: 12 November 2013
- Accepted: 14 February 2014
- Published: 19 February 2014
The R518X/KCNQ1 mutation is a common cause of autosomal recessive (Jervell and Lange Nielsen Syndrome- JLNS) and autosomal dominant long QT syndrome (LQTS) worldwide. In Sweden p.R518X accounts for the majority of JLNS cases and is the second most common cause of LQTS. Here we investigate the clinical phenotype and origin of Swedish carriers of the p.R518X mutation.
The study included 19 Swedish p.R518X index families, ascertained by molecular genetics methods (101 mutation-carriers, whereof 15 JLNS cases and 86 LQTS cases). In all families analyses included assessment of clinical data (symptoms, medications and manually measured electrocardiograms), genealogy (census records), haplotype (microsatellite markers) as well as assessment of mutation age and associated prevalence (ESTIAGE and DMLE computer software).
Clinical phenotype ranged from expectedly severe in JLNS to surprisingly benign in LQTS (QTc 576 ± 61 ms vs. 462 ± 34 ms, cumulative incidence of (aborted) cardiac arrest 47% vs. 1%, annual non-medicated incidence rate (aborted) cardiac arrest 4% vs. 0.04%).
A common northern origin was found for 1701/1929 ancestors born 1650-1950. Historical geographical clustering in the coastal area of the Pite River valley was shown. A shared haplotype spanning the KCNQ1 gene was seen in 17/19 families. Mutation age was estimated to 28 generations (95% CI 19;41). A high prevalence of Swedish p.R518X heterozygotes was suggested (~1:2000-4000).
R518X/KCNQ1 occurs as a common founder mutation in Sweden and is associated with an unexpectedly benign phenotype in heterozygous carriers.
- Long QT Syndrome
- Genotype-phenotype correlations
- Clinical phenotype
- Founder mutation
- Mutation age
- Prevalence estimate
Loss-of-function mutations in the KCNQ1 gene cause both the autosomal recessive Jervell and Lange-Nielsen syndrome (JLNS) and the autosomal dominant type 1 long QT syndrome (LQTS), also known as the Romano-Ward syndrome . The KCNQ1 gene encodes the α-subunit of a voltage-gated potassium ion channel (Kv7.1) regulating both inner ear endolymph flow and cardiac action potential duration via the delayed rectifier potassium current (IKs). Kv7.1 function loss corresponds to congenital hearing loss in JLNS, as well as variable QT prolongation and propensity for arrhythmia, presenting as syncope or (aborted) cardiac arrest, in both syndromes.
Several hundred different mutations with variable effect on Kv7.1 function have been reported, and while mutation-specific risk-stratification could be of considerable clinical importance in LQTS, few mutations are common enough to allow such characterization .
In Sweden, two KCNQ1 mutations dominate the mutation spectrum regarding LQTS and JLNS [3, 4]. The p.R518X and p.Y111C mutations account for over 25% of Swedish LQTS index cases with identified mutations , and p.R518X alone has been identified as the major cause of JLNS in Sweden . Regarding p.Y111C, strong founder effects during the population development of a northern river valley region have previously been shown to result in the enrichment of this specific founder mutation in the population [5, 6].
All living p.R518X mutation-carriers (n = 97) answered a questionnaire regarding personal LQTS history (occurrence of symptoms and beta-blocker therapy duration and compliance). Symptomatic cases were interviewed by one of the authors regarding symptoms (debut, type, frequency and triggers). Anamnestic data regarding four deceased JLNS siblings was obtained during an interview with their living sibling. The data included information regarding deafness, cardiac symptoms (debut, type and frequency) and age at, and cause of, death.
Symptomatic LQTS was defined as syncope/transient but complete loss of consciousness. Life-threatening cardiac events were defined as aborted cardiac arrest requiring resuscitation or as sudden cardiac death.
QT interval duration was measured manually, preferably in lead II as a mean of three consecutive QT intervals, and corrected for heart rate by Bazett’s formula (QT/√R-R), using the mean of the R-R intervals (lead III) preceding the measured beats, in standard 12- lead electrocardiograms obtained from medical records, when available recorded in absence of beta-blocker therapy. Beta-blocker therapy was not discontinued in any mutation-carrier in order to obtain a recording in absence of therapy.
Statistical analyses regarding clinical phenotype
Clinical parameters were explored within p.R518X genotype groups using Chi-square test (Fisher’s Exact test, for correlations between nominal variables), and Nonparametric test (Mann Whitney U test, for analysis of variance between continuous (scalar) and nominal variables). To assess the interfamilial variation of phenotype, the annual incidence rate of life threatening cardiac events in absence of beta-blocker therapy in p.R518X heterozygotes was compared to that of a large mixed LQT1 population , by Fisher’s Exact test. For clinical parameters a two-tailed value for p < 0.05 was considered statistically significant.
Investigating the origin of the p.R518X mutation
Genealogical investigations were performed using parish records and genealogical databases at the Umeå University research archive and the Swedish archive information homepage (http://www.svar.ra.se). Both parental lineages were investigated at least up until 1750, and when possible traced back to the 16th century. Birth-places of ancestors born between 1650 and 1950 were noted on regional maps to assess geographical clustering, over time.
Mutation age and prevalence estimations
The age of a mutation can be inferred by assessing the frequency of an allele, or the decay secondary to mutations and/or recombination of an ancestral haplotype over the generations in a sample of probands sharing mutations identical by descent [11, 12]. In p.R518X index families sharing a common haplotype, the distance (in generations) from the included probands to the most recent common ancestor (an approximation of mutation age) as well as the associated prevalence of founder descendants (as a function of mutation age) were estimated, including 95% confidence intervals (95% CI), using the ESTIAGE software  and DMLE freeware (available at http://www.dmle.org).
The mutation age estimate, as calculated by the ESTIAGE software, included haplotype data, allele frequencies from healthy controls, and recombination frequencies for the microsatellite markers. The haplotype data included the extent of shared alleles among probands, counting from the gene and outwards, excluding shared alleles distant of any marker with discordant alleles. The allele frequencies from healthy controls included the proportion of the founder allele, per marker, in a sample of 168 control chromosomes of northern Swedish origin. The recombination frequencies were derived from the physical distances between the mutation and the microsatellite markers, calculated using the standard correspondence 1 cM = 106 base pairs. Separate estimates were performed in order to assess the potential impact of the assumed mutation rate (10-6 to 10-4) as well as the mutation model used (stepwise or equal).
The approximation of p.R518X prevalence as a function of mutation age, as calculated by the DMLE software, included haplotype data from families sharing a common haplotype, regional population growth rates and an estimate of the proportion of population sampled. The haplotype data included the full haplotype of both p.R518X families and controls for all analysed markers. The regional population growth rate was analysed as a discrete variable = e^(ln [end population/start population]/number of generations) -1), calculated using population demographics data available from Statistics Sweden (http://www.scb.se). The proportion of population sampled was viewed as the unknown variable and iterations were performed over the interval 0.0001-0.5, within a range of possible values for mutation age (defined as the overlap between the 95% confidence intervals of the ESTIAGE and DMLE mutation age estimates). The upper limit of acceptable values for the variable proportion of population sampled was corrected for the number of ascertained mutation-carriers in the population.
Study population and p.R518X-associated clinical phenotype
Clinical characteristics of the Swedish R518X/ KCNQ1 study population
Age at last follow-up, years
29 ± 23, 28
34 ± 21, 37
Non-medicated follow-up, years
16 ± 20, 9
31 ± 21, 33
Experience of first cardiac event
Age at onset, years
2 ± 1, 3
18 ± 15, 12
Experience of ACA/SCD
Number of events
Non-medicated life-years, n
Annual incidence rate before therapy,%c
Triggers of symptoms,%
46/ 12/ 2/ 20
44/ 2/ 22/ 32
ECG (% recorded off therapy)
576 ± 61, 560
462 ± 34, 459
Heart rate, bpm
77 ± 23, 75
73 ± 20, 69
Age at therapy start, years
8 ± 15, 2g
25 ± 19, 17
JLNS cases (n = 15, QTc 576 ± 61 ms, range 461-697 ms) presented with congenital hearing loss and a severe cardiac phenotype (early symptoms debut and a high frequency of life-threatening cardiac events, including three sudden deaths, Table 1). A more severe phenotype for homozygous as compared to compound heterozygous JLNS cases was suggested in this limited material, but did not reach statistical significance (syncope 100% vs. 57%, p = 0.077; QTc 622 ± 64 ms vs. 550 ± 45 ms, p = 0.089).
LQTS cases (n = 86, QTc 462 ± 34 ms, range 397-545 ms) presented with a relatively benign overall phenotype (Table 1). Fifteen LQTS cases (17%) had experience of syncope during a mean follow up of 31 ± 21 years (median 33) before therapy. One case died suddenly and there were no aborted cardiac arrests reported (cumulative incidence of life threatening events 1.2%, annual incidence rate before therapy 0.04%). This corresponds to a low incidence of life-threatening cardiac events, as compared to a large LQT1 population  (0.04% vs. 0.3%, p = 0.007). The sudden cardiac death occurred in a previously asymptomatic adult female in family JLN2 (age 56 years) with a QTc of 506 ms without prophylactic beta-blocker therapy, in relation to physical exercise and hypokalaemia (probably diet-induced).
QTc in LQTS cases showed variability between males and females (443 ± 29 ms vs. 473 ± 33, p < 0.001) as well as within genders (range 397-521 ms in males; 402-545 ms in females). Based on age- and gender adjusted QTc levels , 46% of LQTS cases had a prolonged QTc while 54% had a borderline (28%) or normal QTc (27%). Among LQTS males, the corresponding proportions were 34%, 28% and 38%. Among symptomatic LQTS cases (n = 15), QTc ranged between 402-545 ms and 33% (five cases, whereof two males) had a QTc below 440 ms. In symptomatic LQTS cases (n = 15, whereof five males) phenotypic variability was evident regarding age at onset (1.5-56 years), cardiac event frequency (1- >10) and QTc (402-545 ms). Five symptomatic cases (33%) had a QTc below 440 ms (whereof two males).
No significant indicators of risk for cardiac events were found when assessing the distribution of the following variables across the categories symptomatic and asymptomatic heterozygous p.R518X carriers; QTc ≥500 ms (p = 0.061), gender (p = 0.56), heart rate (p = 0.112) and scalar QTc prolongation (p = 0.622).
Symptomatic LQTS cases (n = 15) were equally found in families with JLNS and LQTS probands (7 vs. 8), i.e. several LQTS probands were investigated for other reasons than previous experience of syncope, such as palpitations, dizziness or chance findings of a prolonged QTc on the electrocardiogram.
With regards to therapeutic interventions, these were performed earlier and more frequently in JLNS cases as compared to LQTS cases. Four JLNS cases (all born before 1950) died prior to beta-blocker therapy, at the age of 20, 27, 37 and 59 years, respectively. Among the 11 JLNS cases treated with beta-blockers, two cases were treated with left cardiac sympathetic denervation (age 2, 24 years) and four cases, including the previous two, received implantable cardioverter defibrillators (age 5, 10, 19 and 32 years), due to recurrent syncope in spite of therapy. All cases with implantable cardioverter defibrillators (n = 4) have experienced appropriate shocks, according to their medical records. In LQTS cases beta-blocker therapy (43% of cases, Table 1) was associated with only one case experiencing a first syncope while on therapy, and no recurrences in 11 previously symptomatic cases with adequate dosage, while compliant.
A common origin established using genealogy and haplotype analysis
Mutation age and prevalence estimate
ESTIAGE mutation age estimates results, in generations including 95% confidence intervals
Estimation of proportion of population sampled and corresponding prevalence estimates for the p.R518X founder mutation
Probands sharing haplotype
Population growth rates,%
In this study, we identified 101 Swedish cases (15 JLNS, 86 LQTS) in 19 index families segregating the R518X/KCNQ1 mutation, and revealed that the common occurrence of this specific mutation in the Swedish population is related to founder effects.
A benign phenotype and remaining variability
While the p.R518X-associated phenotype in JLNS cases was expectedly severe (resulting from a near-complete loss of Kv7.1 function), an unexpectedly benign phenotype was seen in LQTS cases. When in the heterozygous form, the p.R518X nonsense mutation has been shown to cause Kv7.1 haploinsufficiency, in vitro . As for p.R518X, it is common for heterozygous nonsense mutations to present with relatively mild phenotypes, as the resultant protein products cannot co-assemble with wild type subunits and therefore seldom cause dominant-negative effects . However, as compared to 169 heterozygous carriers of KCNQ1 mutations causing haploinsufficiency, described by Moss et al, the clinical phenotype of the p.R518X founder heterozygotes still appear to be less severe (aborted cardiac arrest 3% vs. 0, sudden death 2.4% vs. 1.2%). The annual incidence rate of life-threatening cardiac events before therapy in the p.R518X founder heterozygotes is comparable to that of another unexpectedly benign Swedish founder population (0.04% vs. 0.05%), segregating the dominant negative p.Y111C mutation [5, 6]. The benign phenotypes of the LQTS river valley populations in Sweden remain unexplained.
The popular notion that founder mutations would by their nature be benign is negated by the severe phenotype of the A341V/KCNQ1 mutation that segregates within a South African founder population, and is associated with a staggering 30% cumulative incidence of life-threatening cardiac events . As in the case with the Boer progeny , a substantial intra-familial phenotypic variability regarding clinical phenotype remained in the p.R518X founder population, and the symptomatic phenotype in p.R518X heterozygotes did not correlate with gender, heart rate nor, surprisingly, QTc.
The commonly occurring p.R518X mutation
In the international context p.R518X is commonly described as a hotspot mutation, reported as a common cause of JLNS [8, 9, 16], as well as one of the five most common mutations in Northern American LQTS probands . In Sweden, the p.R518X mutation has previously been revealed as the major cause of JLNS, contributing to the JLNS genotype in 9/12 identified JLNS index families with ascertained genotype (12/24 alleles) . The importance of the p.R518X allele also with regards to the Swedish LQTS mutation spectrum was indicated by the finding that p.R518X was the second most common mutation identified in 200 Swedish index cases referred for LQTS diagnostics at the laboratory of Clinical Genetics, Umeå University Hospital, between 2006 and 2009 . The cohort included LQTS index cases (excluding JLNS probands) originating from all six Swedish health care regions, whereof 78% from without the northern region. Together with p.Y111C (the most commonly identified mutation) these two mutations accounted for over 25% of index cases with ascertained genotype (n = 102) .
The major LQTS genes were not analysed in 4/10 LQTS probands, negating the possibility of identifying additional mutations contributing to phenotype in these cases.
The p.R518X prevalence estimate of ~1:2000-4000 presented in this study is based on extrapolations from genetic, genealogical and epidemiological data, and as such should be interpreted with caution. Being derived from data on the founder population the estimates do not include calculations regarding p.R518X cases secondary to hotspot effects (risk of underestimation). Also, regional differences in prevalence with regards to distribution of mutation-carriers are to be expected (i.e. higher prevalence in the northern region and the major urban regions in the south that have received the majority of the 20-21th century migration). The p.R518X prevalence estimate of ~1:2000-4000 is supported by the high frequency of both founder (n = 14) and non-founder JLNS cases (n = 1) with p.R518X mutations identified in the population, the high contribution of the specific p.R518X allele to the Swedish JLNS mutation spectrum , as well as congruent prevalence estimates in the Norwegian population .
The common occurrence of the p.R518X mutation among Swedish probands with recessive and dominant type LQTS is mainly secondary to a founder effect. Our findings suggest a high prevalence of the p.R518X founder mutation in the Swedish population. In the clinical setting, due to the low penetrance of clinically identifiable markers in p.R518X heterozygotes, molecular genetics diagnosis of probands and cascade-screening of first-degree relatives remains imperative in order to identify individuals at risk of developing preventable arrhythmia.
AW is a medical doctor in Umeå, Sweden. In 2012 she published her thesis on LQTS founder effects and associated cardiac phenotypes in the Swedish population. AR is a pediatric cardiologist and the principle investigator of the LQTS research group. Together with clinical geneticist ELS, cardiologist SJ and biomedical analyst UBD she founded the LQTS Family Clinic in Umeå, Sweden, where LQTS families including several generations of carriers have been counselled and treated in a multi-disciplinary setting since 2005.
We thank Susann Haraldsson, Department of Medical Biosciences, Medical and Clinical Genetics, Umeå University, for expert technical assistance. Original artwork and figures were constructed using open source software (Inkscape vector graphics editor and GNU image manipulation program) by illustrator Erik Winbo (erikwinbo.artworkfolio.com).
This work was supported by the Swedish Heart-Lung Foundation, the Heart Foundation of Northern Sweden, the medical faculty at Umeå University and the Northern County Councils Cooperation Committee.
- Shimizu W, Horie M: Phenotypic manifestations of mutations in genes encoding subunits of cardiac potassium channels. Circ Res. 2011, 109 (1): 97-109. 10.1161/CIRCRESAHA.110.224600.View ArticlePubMedGoogle Scholar
- Kapplinger JD, Tester DJ, Salisbury BA, Carr JL, Harris-Kerr C, Pollevick GD, Wilde AA, Ackerman MJ: Spectrum and prevalence of mutations from the first 2,500 consecutive unrelated patients referred for the FAMILION long QT syndrome genetic test. Heart Rhythm. 2009, 6 (9): 1297-1303. 10.1016/j.hrthm.2009.05.021.View ArticlePubMedPubMed CentralGoogle Scholar
- Stattin EL, Bostrom IM, Winbo A, Cederquist K, Jonasson J, Jonsson BA, Diamant UB, Jensen SM, Rydberg A, Norberg A: Founder mutations characterise the mutation panorama in 200 Swedish index cases referred for Long QT syndrome genetic testing. BMC Cardiovasc Disord. 2012, 12 (1): 95-10.1186/1471-2261-12-95.View ArticlePubMedPubMed CentralGoogle Scholar
- Winbo A, Stattin EL, Diamant UB, Persson J, Jensen SM, Rydberg A: Prevalence, mutation spectrum, and cardiac phenotype of the Jervell and Lange-Nielsen syndrome in Sweden. Europace. 2012, 14 (12): 1799-1806. 10.1093/europace/eus111.View ArticlePubMedGoogle Scholar
- Winbo A, Diamant UB, Rydberg A, Persson J, Jensen SM, Stattin EL: Origin of the Swedish long QT syndrome Y111C/KCNQ1 founder mutation. Heart Rhythm. 2011, 8 (4): 541-547. 10.1016/j.hrthm.2010.11.043.View ArticlePubMedGoogle Scholar
- Winbo A, Diamant UB, Stattin EL, Jensen SM, Rydberg A: Low incidence of sudden cardiac death in a Swedish Y111C type 1 long-QT syndrome population. Circ Cardiovasc Genet. 2009, 2 (6): 558-564. 10.1161/CIRCGENETICS.108.825547.View ArticlePubMedGoogle Scholar
- Tester DJ, Will ML, Haglund CM, Ackerman MJ: Compendium of cardiac channel mutations in 541 consecutive unrelated patients referred for long QT syndrome genetic testing. Heart Rhythm. 2005, 2 (5): 507-517. 10.1016/j.hrthm.2005.01.020.View ArticlePubMedGoogle Scholar
- Wei J, Fish FA, Myerburg RJ, Roden DM, George AL: Novel KCNQ1 mutations associated with recessive and dominant congenital long QT syndromes: evidence for variable hearing phenotype associated with R518X. Hum Mutat. 2000, 15 (4): 387-388.View ArticlePubMedGoogle Scholar
- Berge KE, Haugaa KH, Fruh A, Anfinsen OG, Gjesdal K, Siem G, Oyen N, Greve G, Carlsson A, Rognum TO, et al: Molecular genetic analysis of long QT syndrome in Norway indicating a high prevalence of heterozygous mutation carriers. Scand J Clin Lab Invest. 2008, 68 (5): 362-368. 10.1080/00365510701765643.View ArticlePubMedGoogle Scholar
- Priori SG, Schwartz PJ, Napolitano C, Bloise R, Ronchetti E, Grillo M, Vicentini A, Spazzolini C, Nastoli J, Bottelli G, et al: Risk stratification in the long-QT syndrome. N Engl J Med. 2003, 348 (19): 1866-1874. 10.1056/NEJMoa022147.View ArticlePubMedGoogle Scholar
- Slatkin M, Rannala B: Estimating allele age. Annu Rev Genomics Hum Genet. 2000, 1: 225-249. 10.1146/annurev.genom.1.1.225.View ArticlePubMedGoogle Scholar
- Rannala B, Bertorelle G: Using linked markers to infer the age of a mutation. Hum Mutat. 2001, 18 (2): 87-100. 10.1002/humu.1158.View ArticlePubMedGoogle Scholar
- Genin E, Tullio-Pelet A, Begeot F, Lyonnet S, Abel L: Estimating the age of rare disease mutations: the example of Triple-A syndrome. J Med Genet. 2004, 41 (6): 445-449. 10.1136/jmg.2003.017962.View ArticlePubMedPubMed CentralGoogle Scholar
- Andersson P, Lundkvist L: The Q-T syndrome–a family description. Acta Med Scand. 1979, 206 (1–2): 73-76.PubMedGoogle Scholar
- Goldenberg I, Moss AJ, Zareba W: QT interval: how to measure it and what is “normal”. J Cardiovasc Electrophysiol. 2006, 17 (3): 333-336. 10.1111/j.1540-8167.2006.00408.x.View ArticlePubMedGoogle Scholar
- Huang L, Bitner-Glindzicz M, Tranebjaerg L, Tinker A: A spectrum of functional effects for disease causing mutations in the Jervell and Lange-Nielsen syndrome. Cardiovasc Res. 2001, 51 (4): 670-680. 10.1016/S0008-6363(01)00350-9.View ArticlePubMedGoogle Scholar
- Wilde AA, Escande D: LQT genotype-phenotype relationships: patients and patches. Cardiovasc Res. 2001, 51 (4): 627-629. 10.1016/S0008-6363(01)00389-3.View ArticlePubMedGoogle Scholar
- Moss AJ, Shimizu W, Wilde AA, Towbin JA, Zareba W, Robinson JL, Qi M, Vincent GM, Ackerman MJ, Kaufman ES, et al: Clinical aspects of type-1 long-QT syndrome by location, coding type, and biophysical function of mutations involving the KCNQ1 gene. Circulation. 2007, 115 (19): 2481-2489. 10.1161/CIRCULATIONAHA.106.665406.View ArticlePubMedPubMed CentralGoogle Scholar
- Crotti L, Spazzolini C, Schwartz PJ, Shimizu W, Denjoy I, Schulze-Bahr E, Zaklyazminskaya EV, Swan H, Ackerman MJ, Moss AJ, et al: The common long-QT syndrome mutation KCNQ1/A341V causes unusually severe clinical manifestations in patients with different ethnic backgrounds: toward a mutation-specific risk stratification. Circulation. 2007, 116 (21): 2366-2375. 10.1161/CIRCULATIONAHA.107.726950.View ArticlePubMedGoogle Scholar
- Brink PA, Crotti L, Corfield V, Goosen A, Durrheim G, Hedley P, Heradien M, Geldenhuys G, Vanoli E, Bacchini S, et al: Phenotypic variability and unusual clinical severity of congenital long-QT syndrome in a founder population. Circulation. 2005, 112 (17): 2602-2610. 10.1161/CIRCULATIONAHA.105.572453.View ArticlePubMedGoogle Scholar
- Tranebjaerg L, Bathen J, Tyson J, Bitner-Glindzicz M: Jervell and Lange-Nielsen syndrome: a Norwegian perspective. Am J Med Genet. 1999, 89 (3): 137-146. 10.1002/(SICI)1096-8628(19990924)89:3<137::AID-AJMG4>3.0.CO;2-C.View ArticlePubMedGoogle Scholar
- Westin G, Olofsson SI: Övre Norrlands historia. D. 1, Tiden till 1600 /The History of The Upper Northern Regions. part 1, Time before 1600. 1962, Umeå: Norrbottens och Västerbottens läns landstingGoogle Scholar
- Einarsdottir E, Egerbladh I, Beckman L, Holmberg D, Escher SA: The genetic population structure of northern Sweden and its implications for mapping genetic diseases. Hereditas. 2007, 144 (5): 171-180. 10.1111/j.2007.0018-0661.02007.x.View ArticlePubMedGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2261/14/22/prepub
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.