LRRK2 G2019S in Families with Parkinson Disease Who Originated from Europe and the Middle East: Evidence of Two Distinct Founding Events Beginning Two Millennia Ago

Abstract

The leucine-rich repeat kinase 2 (LRRK2) G2019S mutation is the most common genetic determinant of Parkinson disease (PD) identified to date. It accounts for 1%–7% of PD in patients of European origin and 20%–40% in Ashkenazi Jews and North African Arabs with PD. Previous studies concluded that patients from these populations all shared a common Middle Eastern founder who lived in the 13th century. We tested this hypothesis by genotyping 25 microsatellite and single-nucleotide–polymorphism markers in 22 families with G2019S and observed two distinct haplotypes. Haplotype 1 was present in 19 families of Ashkenazi Jewish and European ancestry, whereas haplotype 2 occurred in three European American families. Using a maximum-likelihood method, we estimated that the families with haplotype 1 shared a common ancestor 2,250 (95% confidence interval 1,650–3,120) years ago, whereas those with haplotype 2 appeared to share a more recent founder. Our data suggest two separate founding events for G2019S in these populations, beginning at a time that coincides with the Jewish Diasporas.

Parkinson disease (PD [MIM 168600]) is the second-most common neurodegenerative disorder and is characterized by bradykinesia, resting tremor, rigidity, and postural instability.^1,2 Although approximately one in five patients with PD reports a family history of the disease, pedigrees that demonstrate clear Mendelian inheritance are rare.^3,4 Linkage analysis of extended pedigrees has implicated five genes in autosomal recessive (PARK2 [MIM 600116 and 602544], PINK1 [MIM 605909], and PARK7 [MIM 606324]) or dominant (SNCA [MIM 163890] and LRRK2 [MIM 609007]) parkinsonism that closely resembles PD.^5–10 However, with the exception of LRRK2, mutations in these genes rarely result in typical late-onset PD. A single LRRK2 mutation (G2019S) accounts for 0.7%–1.6% of sporadic and 2.8%–6.6% of familial PD cases of European origin.^11–16 More-recent studies indicate a remarkably high prevalence of G2019S in Ashkenazi Jews (13.3% sporadic; 29.7% familial) and North African Arabs (40.8% sporadic; 37.0% familial) with PD.^17,18 Patients heterozygous for G2019S display clinical characteristics that are indistinguishable from those of patients with PD in the general population.^11–18

Data from five studies of 90 unrelated G2019S-bearing subjects of European or Middle Eastern–North African (MENA) origin revealed that all shared the same haplotype, consistent with a common founder.^11,16–19 The high prevalence of the mutation in Ashkenazi Jews and North African Arabs has led to the hypothesis that the mutation originated in the Middle East.¹⁸ Lesage and colleagues¹⁹ used a likelihood-based haplotype approach²⁰ in the study of six families of North African or European origin carrying G2019S and estimated that these individuals shared a common founder ∼725 years ago.¹⁹ Given the fact that G2019S is widely distributed across Europe and occurs at high frequency among Ashkenazi Jews and North African Arabs, this relatively recent estimated date is difficult to reconcile with established patterns of human migration.^21,22 Therefore, we collected genotype data on a larger sample of subjects, to examine the hypotheses that (1) patients with PD who are of European and MENA origin share a single ancestral haplotype and (2) the most recent common ancestor lived in the 13th century.

The study population was derived from 1,611 unrelated patients with PD and 1,647 control individuals enrolled at seven movement-disorder clinics in three U.S. cities (Albany, NY; Portland, OR; and Seattle). All patients met standardized diagnostic criteria for PD, as determined by a movement-disorder specialist.²³ Controls had no history of parkinsonism, and most were spouses of patients with PD. The institutional review boards at each participating site approved the study, and all subjects gave informed consent. Among these subjects, we identified 21 patients (1 homozygous, 20 heterozygous) and 1 control individual with G2019S (19 of these subjects have been reported elsewhere).^12,24 For the present study, we included these 22 subjects, 23 of their family members (6 affected and 17 unaffected), and 30 unrelated G2019S-negative European American controls. Of the G2019S-positive families, 13 were of (non-Jewish) European origin, 3 defined themselves as Ashkenazi Jews, and 6 were of unspecified Jewish ancestry. However, the probands from those six families all indicated that their ancestors emigrated from central/eastern Europe or Russia and were thus likely descendants of the Ashkenazim.

We systematically selected markers, to determine whether our subjects carrying G2019S shared a common haplotype, using data from an ongoing study in which we are screening the entire LRRK2 coding region and flanking intronic sequence for new mutations. Complete sequence data on all 51 exons were available for one subject homozygous and for three subjects heterozygous for G2019S. We compared these four individuals at 90 polymorphic sites and found five (rs28903073, rs2404834, rs10784522, rs10878405, and ss52051244) at which one subject (IPD584) differed from one or more of the others (IPD289, IPD452, and/or PD82801) at both alleles. We genotyped these five SNPs, 18 markers included in previous studies,^{11,16,17,19,24} and two additional microsatellites (D12S331 and D12S1592) in all 75 subjects in our study population. The markers (12 SNPs and 13 microsatellites) spanned 9 Mb across the PARK8 region. SNP genotyping was performed by sequencing with the Applied Biosystems BigDye Terminator v3.1 Cycle Sequencing Kit. Microsatellites were amplified by PCR with use of fluorescently labeled F primers, were run on an ABI PRISM 3130 Genetic Analyzer, and were analyzed using GeneMapper 4.0 software (Applied Biosystems).

We estimated the age of the most recent common founder using the program Estiage.²⁰ This maximum-likelihood method uses information on the recombination fractions between the mutation and each marker, the frequencies of the shared allele at each marker, and the position of the first marker in each direction that is no longer shared, to calculate the number of generations (with 95% CI) elapsed since the most recent common ancestor introduced the mutation into the population. We defined a marker as shared among families if a single (best-call) allele was included in all disease haplotypes or in at least half and at significantly greater frequency (Fisher’s exact test α=.05) than in 60 inferred control haplotypes (collapsing all other alleles into a single bin). Genetic-map positions for each marker were derived from the linkage-mapping server MAP-O-MAT, and physical positions were taken from the National Center for Biotechnology Information (NCBI) human genome assembly Build 35.²⁵ Because many of the markers used in the present study were tightly linked, genetic distances between them could not be accurately estimated from available genetic maps. Instead, we used the genetic length and physical distance between the extreme markers (D12S2080 and D12S1301) to calculate an average of 0.30 cM per Mb across the entire analyzed region and then computed recombination fractions using the Kosambi mapping function.

In instances where marker phase could not be unambiguously resolved by pedigree data, we used PHASE v2.1.1²⁶ (Matthew Stephens' Web site) to infer haplotypes. Since the use of inferred haplotypes can introduce additional uncertainty into age estimates, we first divided PHASE calls for each allele at a given marker into three categories by probability: best (P>.5), plausible (.05<P<.5), and unlikely (P<.05). We then performed four separate runs, in Estiage, to estimate the age of the most recent founder by using haplotypes from only the families in which phase could be unambiguously resolved across shared markers or with the most likely (best call only), longest plausible, or shortest plausible haplotypes inferred by PHASE (Matthew Stephens' Web site) in all families.

Two clearly distinct disease haplotypes (referred to hereafter as “haplotype 1” and “haplotype 2”) became evident on inspection of the data (fig. 1). Haplotype 1 included a core region of 17 consecutive markers spanning 243 kb (D12S2514 to D12S2519) that was shared by 19 families, with the exception of one marker, D12S2515. Our group and others have argued that D12S2515 is unstable and that the allelic differences previously observed among disease haplotypes at this marker were likely due to recurrent mutation rather than recombination.^11,24 Data from the current study further support this idea; all 19 families with haplotype 1 shared alleles at three markers upstream from D12S2515, including the A allele of rs28903073, which was observed elsewhere in only 1 of 906 European American control chromosomes (fig. 1).²⁴ This suggests that the A allele is identical by descent for haplotype 1 carriers and that any discordant alleles at intervening markers (between rs28903073 and G2019S) in these individuals arose via new mutations. Thus, D12S2515 was excluded from the data set used to reconstruct haplotypes and was not used in subsequent common founder-age estimates. Haplotype 2 was present in three families of European descent (fig. 1) and could be differentiated from the core region of haplotype 1 by five intragenic SNPs and one extragenic microsatellite (D12S2519). These included the two closest flanking markers, which were located 5 kb upstream (rs2404834) and 6 kb downstream (rs10784522) from G2019S. Haplotype 2 extended a minimum of 6 Mb across the PARK8 region, from the 5′-most marker assayed (D12S2080) to D12S1592. Haplotypes 1 and 2 appear to be rare in populations of European ancestry, since the frequency of haplotype 1 can be no greater than that of the A allele at rs28903073 (0.1%),²⁴ and haplotype 2 was not observed in our sample of 60 control chromosomes. Taken together, these data strongly suggest that the disease chromosomes in our study population originated from two separate founders.

Figure 1. .

Comparison of G2019S-containing haplotypes across the LRRK2 region. Five intragenic markers (bold type) delineate two distinct haplotypes in which complete allele sharing is indicated by blue (haplotype 1) or yellow (haplotype 2) shading. For haplotype 1, significant but incomplete allele sharing is denoted by light-blue shading. For markers for which phase could not be unambiguously determined, both alleles are shown with the best call, as inferred by PHASE (Matthew Stephens' Web site) listed first (alleles of probability <.05 are in brackets). The proband of family IPD289 was homozygous for G2019S; thus, haplotypes for the paternal (^p) and maternal (^m) lineages are provided separately. D12S2515 (shown in italics) is unstable and was not included in haplotype analyses.

We then estimated the age of the most recent common founder among the 19 families with haplotype 1, designating D12S2194 and D12S1048 as the boundaries of the shared haplotype as determined by the aforementioned criteria (table 1). Our first approach, which was the one most comparable to that used by Lesage and colleagues,¹⁹ restricted the analysis to families with phase-known data within the shared region. This reduced the data set to five families (IPD289, IPD452, PD82801, PN184, and PN196) representing six disease chromosomes and produced an age estimate of 65 (95% CI 36–121) generations. After incorporating the remaining 14 families for which phase was inferred, the best-call and longest haplotypes were identical and yielded an estimate of 75 (95% CI 55–104) generations, whereas the shortest haplotypes estimate was 84 (95% CI 61–117) generations. With use of the most widely published intergenerational interval, 25 years,²⁷ our best-call data indicate that these families shared a common ancestor 1,875 (95% CI 1,375–2,600) years ago. However, several recent studies have concluded that 30 years is a better approximation of generation time for the period of human history in question.^27–29 A 30-year interval with best-call haplotypes increases the age of the founding event to 2,250 (95% CI 1,650–3,120) years ago. We did not perform age estimates for the three families with haplotype 2, because the sample size was small and we were not able to define the 5′ limit of the shared region. However, since the genetic length of the shared region is inversely proportional to the age of the founding event, these three families likely shared a common ancestor more recently than did the haplotype 1 families.

Table 1. .

Data Used in Estimating the Age of the G2019S Haplotype 1 Common Founder^[Note]

	Data Used for Family
Marker	IPD289(p)a	IPD289(m)a	IPD483	PN65	PN371	PN196	FP0079	FP0341	IPD452	FP0353	PN718	PD82801	PN278	IPD374	PN961	IPD197	FP0021	PN184	PN5664	IPD620	Distanceb	θc	Frequencyd
D12S331	181	181	181	181	181	181	181	181	181	181	181	177	177	177	177	179					1.473	.0045
D12S2194	257	257	257	257	257	257	257	257	257	257	257	257	257	257	257	257	261	261	261	253	.282	.0009	.20
D12S2514	291	291	291	291	291	291	291	291	291	291	291	291	291	291	291	291	291	291	291	291	.147	.0004	.53
rs10878245	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	.102	.0003	.60
rs28903073	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	.081	.0002	.001
rs7966550	T	T	T	T	T	T	T	T	T	T	T	T	T	T	T	T	T	T	T	T	.046	.0001	.85
D12S2516	254	254	254	254	254	254	254	254	254	254	254	254	254	254	254	254	254	254	254	254	.031	.00009	.63
rs1896252	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	.020	.00006	.62
rs1427263	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	.020	.00006	.68
rs11176013	G	G	G	G	G	G	G	G	G	G	G	G	G	G	G	G	G	G	G	G	.020	.00006	.62
rs11564148	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	.020	.00006	.39
rs2404834	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	.005	.00002	.91
G2019S	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	0	0
rs10784522	T	T	T	T	T	T	T	T	T	T	T	T	T	T	T	T	T	T	T	T	.006	.00002	.40
rs10878405	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	.008	.00002	.40
D12S2518	154	154	154	154	154	154	154	154	154	154	154	154	154	154	154	154	154	154	154	154	.014	.00004	.80
ss52051244	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	A	.023	.00007	.88
rs3761863	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	C	.024	.00007	.67
D12S2519	132	132	132	132	132	132	132	132	132	132	132	132	132	132	132	132	132	132	132	132	.096	.0003	.30
D12S2520	260	260	260	260	260	260	260	260	260	260	260	260	260	260	260	260	260	260	257	260	.100	.0003	.20
D12S2521	359	359	359	359	359	359	359	359	359	359	359	359	359	359	359	359	359	355		359	.108	.0003	.08
D12S1048	214	214	214	214	214	214	214	214	214	214	214	214	214	214	214	214	214			211	.292	.0009	.21
D12S1592	254	256	256	256	254	254	256	254	256	254	256	254	254	254	254	254	254				1.711	.0052

Note.— The interval between markers D12S2194 and D12S1048 was defined as shared and, within this segment phase, was inferred for 14 of 20 disease chromosomes. Marker D12S2515 is unstable and was excluded from the analysis.

^aThe proband of family IPD289 was homozygous for G2019S; thus, haplotypes from the paternal (p) and maternal (m) lineages are provided separately.

^bPhysical distance (Mb) between G2019S and each marker, from ^NCBI Build 35.

^cRecombination fraction between G2019S and each marker, calculated using the Kosambi mapping function and an average of 0.30 cM per Mb across the entire region.

^dFrequencies of the shared allele were taken from Kachergus et al.¹⁶ for microsatellites and 30 International HapMap Project CEPH trios for SNP markers, except for rs28903073 (which was from a previous report²⁴) and ss52051244 (which was derived from 60 European American control chromosomes genotyped in this study).

Our study is the first to provide evidence that G2019S-positive patients of European ancestry originated from more than one founder. In our sample, haplotypes 1 and 2 could be adequately distinguished only by inclusion of five intragenic SNPs added on the basis of extensive sequence comparisons performed on a subset of probands. Only one of these five SNPs (rs28903073) has been included in similar analyses by other groups,¹¹ so it is possible that both haplotypes were present among the 90 G2019S-bearing families of European or MENA origin reported elsewhere.^11,16–19 However, we suspect that the majority of these families will prove to carry haplotype 1, on the basis of allele sharing at four extragenic microsatellites (D12S2519, D12S2520, D12S2521, and D12S1048) assayed in four of five studies^11,16,17,19 and at rs28903073, which was genotyped in a single study of 18 (largely Italian) families.¹¹ This is consistent with our data and suggests that haplotype 1 is predominant in European Americans and Ashkenazi Jews—and likely in North African Arabs as well—carrying G2019S, which would be explained by a common Middle Eastern founder. The geographic origin of haplotype 2 is less certain, but all three families in which it occurred originated primarily from Western Europe.

Our age estimates for the most-recent common founder for G2019S (haplotype 1) were two- to threefold larger than the one calculated by Lesage and colleagues.¹⁹ There were three potential differences of relevance between the two studies: the physical distance shared among disease chromosomes, our use of inferred haplotypes, and the estimation of recombination fraction between markers. We observed a minimum shared region of 243 kb flanking G2019S in our sample, whereas Lesage et al.¹⁹ reported a much shorter segment, 60 kb. We suspect that this was due largely to their using an unstable marker (D12S2515) to delineate the 5′ boundary of the shared segment and possibly unknowingly combining subjects with haplotypes 1 and 2 in the analysis. Slight differences in the marker sets genotyped in the two studies might also have had an effect. However, this would not explain the difference in age estimates, since our inclusion of longer disease haplotypes would have the opposite effect (i.e., predict a more recent ancestor). Likewise, our use of inferred haplotypes in some families does not explain the discrepancy, because, even when the analysis was restricted to chromosomes of known phase, as was done in the study of Lesage et al.,¹⁹ our estimate was still more than twice as large (65 vs. 29 generations). Furthermore, our inclusion of reconstructed haplotypes in 14 of 19 families allowed more-efficient use of the data set, with minimal added uncertainty in the estimate, as evidenced by the narrow range of ages calculated using the longest or shortest plausible haplotypes (75 vs. 84 generations). The only remaining explanation to account for the discrepancy between the two studies is that we must have used lower recombination fractions in the maximum-likelihood calculations (table 1). Lower recombination fractions correspond to a decreased genetic length of the segment shared among disease chromosomes and an increased estimated time since the last common ancestor. This general relationship is apparent when our estimate of 75 generations, based on a shared segment of <0.2 cM, is compared with two previous studies that used Estiage to estimate a founder age of 46 generations (0.7–12.6 cM shared) for the AAAS (MIM 605378) IVS14+1G→A mutation and 23 generations (7 cM shared) for the VAPB P56S mutation (MIM 605704 and 608627).^20,30 In contrast, Lesage and colleagues¹⁹ calculated an age of only 29 generations, using a shared physical distance less than what we used. We cannot determine the degree to which our estimated recombination fractions differed from theirs, because they have not published these data, although the genetic distances used in both studies were derived from the same source (MAP-O-MAT).

Our data are consistent with the hypothesis of Ozelius and colleagues¹⁸ that Europeans, Ashkenazi Jews, and North African Arabs with G2019S arose from a common Middle Eastern founder, if only those individuals with haplotype 1 are considered. Our best estimate suggests that the common ancestor lived 2,250 years ago, during the period of the Jewish Diasporas (586 bc to 70 ad), at a time when the ancestral Jewish population and some Arab communities existed in close proximity. Similar age estimates have been calculated for other disease alleles found at high frequency among Ashkenazi Jews, non-Ashkenazi Jews, and Arabs, including the factor XI deficiency type II F11 (MIM 264900) E117X mutation and the APC (MIM 175100) I1307K variant, which increases risk for colorectal cancer.^22,31 Thus, Lesage and colleagues’ conclusion that the G2019S founding event occurred in the 13th century¹⁹ appears implausible, since, by that time, Jews destined to become Ashkenazim had been separated from Arab and other Jewish communities for many centuries.

Our data provide an important framework for further investigation of the widespread expansion of G2019S over the past two millennia. Mapping the frequency of disease haplotypes 1 and 2 in European and MENA populations, with use of the marker set described above, will be critical in this endeavor, particularly in non–Ashkenazi Jewish and non-Jewish communities in the Middle East. Defining the age and origin of haplotype 2 might help to explain the variable prevalence of the mutation across Europe. Finally, G2019S was recently identified in an Asian population²⁴ on a haplotype background clearly distinct from the two observed here, which indicates that at least three founding events have occurred. It is becoming clear that the genetic and demographic processes that have shaped the current distribution of G2019S across the world are more complex than previously recognized.