Age and Origin of the PRNP E200K Mutation Causing Familial Creutzfeldt-Jacob Disease in Libyan Jews

To the Editor:

Creutzfeldt-Jacob disease (CJD [MIM 123400]), the most prevalent of the human spongiform encephalopathies (HSEs), is a rapidly progressive neurodegenerative disease that manifests itself as a sporadic, transmissible, or familial disorder (Johnson and Gibbs 1998). Patients with CJD generally develop neurological dysfunction in midlife and die within 6–24 mo of onset. The largest cluster of CJD occurs among Libyan Jews, where the incidence (1/10,000) is ∼100 times higher than incidence worldwide. The origin of the higher incidence of CJD in this population is an intriguing problem that has not yet been resolved.

Although a few studies (reviewed in Meiner et al. 1997) had pointed out the familial predisposition to CJD, it was first speculated that its frequency in Libyan Jews could be explained by their habit of consuming lightly grilled sheep’s brains or eyeballs (Herzberg et al. 1974; Alter and Kahana 1976), reflecting a shared environmental risk (exposure to scrapie-infected meat) rather than any genetic factor. This hypothesis, based on the unrealistic assumption that scrapie was widespread in Libya and that a marked culinary difference between Jews of Libyan and other North African origins existed, suffered a strong blow when two cases of CJD among young Jews born in Israel to immigrants from Libya were discovered (Nisipeanu et al. 1990; Hsiao et al. 1991). They could not have been exposed to scrapie, since it does not exist in Israel. The most likely explanation soon became that CJD could be inherited genetically in a pattern similar to that of another HSE, Gerstmann-Sträussler-Scheinker disease (MIM 137440). Genetic linkage between a missense mutation (E200K: 598A→G; 200Glu→Lys) in the prion protein (PRNP) gene and CJD was established in Libyan Jews (LOD score >4.85 [Gabizon et al. 1993]). The E200K mutation, which accounts for >70% of cases of familial CJD, was first identified in a Polish family and subsequently in patients living in England, France, Austria, Slovakia, Chile, the United States, and Japan. The analysis of its geographic distribution (Goldfarb et al. 1991) suggested that the E200K mutation originated in Spain >5 centuries ago, possibly in a Jewish person, and spread to Mediterranean and continental countries after the expulsion of Sephardic Jews from Spain. The hypothesis of propagation through a limited number of successful migrants was supported by the discovery of a higher frequency, in patients with CJD and their unaffected relatives, compared with the general Libyan Jewish population, of PRNP’s 129M polymorphism in the normal allele. However, the same hypothesis was disputed (Gabizon et al. 1993; Korczyn 1994) on the basis of lack of evidence for the presence of the E200K mutation in Spain and among Jews living in other countries to which Sephardim have emigrated and because it seems to disregard the intermarriage rules followed by the local Jewish communities.

Recently, the E200K mutation has been discovered in other European countries—that is, in Italy, as well as in Spain itself. More decisively, Lee et al. (1999) reported in the Journal that Libyan, Tunisian, Italian (continental), Chilean, and Spanish families with CJD share a major haplotype on chromosome 20p12-pter, to which the PRNP gene has been mapped, whereas families with CJD from Germany, Sicily, Austria, and Japan bear different corresponding haplotypes, suggesting independent mutational events. The prominence of this study for the reinstatement of the “founder effect” hypothesis to explain the Libyan focus of familial CJD cannot be dismissed. However, the probational strength of haplotype data presented by Lee et al. (1999) can be even more convincing if they are quantitatively analyzed for linkage disequilibrium (LD) decay over time and the results compared with the Libyan Jewish population's history. To perform this, two different methods were used, both of which are based on the genetic clock equation lnP=-θg, relating the time (in generations, g) tracing back to the most recent common ancestor of mutant chromosomes, the frequency of recombination between the disease locus and the marker (θ), and the probability that a marker's allele on a disease chromosome is the ancestral one (P). Taking the frequencies on E200K-bearing (p_d) and normal (p_n) Libyan Jewish chromosomes for the alleles of three microsatellite markers flanking the PRNP locus and defining the putative common ancestral haplotype (D20S116–20, D20S482–14, and D20S895–18/19; table 2 of Lee et al. 1999), the LD measure, δ=(p_d-p_n)/(1-p_n), was calculated according to Bengtsson and Thomson (1981). Recombination fraction values (θ) were conveniently estimated from chromosome-20 physical mapping (LDB and the Sanger Centre) under the assumption of a genetic to physical distance ratio of 3.7 obtained from regression of centimorgan versus megabyte values for 20p12-pter loci. Applying the algorithm of Risch et al. (1995), g=logδ/log(1-θ), the estimated age (in generations, g) of the E200K mutation is 16.8±2.0 SD (95% confidence interval [CI] 11.7–21.8 g) (table 1). Using the iterative method of Reich and Goldstein (1999), which models the regeneration of the ancestral haplotype by the recombination and mutation process through a Markov transition matrix that gives the probabilities that any one haplotype will be transformed into any other one in a single generation, a somewhat lower age was obtained (14±2 SD; 95% CI 9–19 g; table 2). When supplemented by setting the genetic clock according to the Luria-Delbrück approach (Labuda et al. 1996), under the assumption of a mean population growth rate of .5, the estimate rises to 22.7±2.1 SD (95% CI 17.4–27.9 g). Given an intergenerational interval of 30 years (Tremblay and Vézina 2000) and an average year of birth of 1950 for all affected and carrier subjects, the present results would date the most recent common ancestor bearing the E200K mutation back to 1450–1530 or to the second half of the 13th century. This dating points to the origin of CJD in Libyan Jews at the time, or before, that Jewish families of Iberian origin settled in Libya after their expulsion from Spain in 1492 and from Portugal in 1497 (Barnavi 1992). Intrafamily marriages were a common practice among them (Gabizon et al. 1993), and a constant growth of this isolated population in the centuries after immigration has been documented (see, e.g., Goldberg 1971). However, other events, such as immigration of Jews from the neighboring island of Jerba at approximately the same time (Udovitch and Valensi 1984) or bottleneck effects, should be taken into consideration. Despite the methodological limits associated with LD-based allele age estimation, persuasive further evidence for the hypothesis of a “Spanish founder effect” in Libyan Jewish CJD genetic epidemiology can be drawn from the analysis of the haplotype data reported by Lee et al. (1999).

Table 1.

Estimation of the Age of the E200K Mutation in the PRNP Gene, Causing CJD in Libyan Jews (Risch et al.’s [1995] Method, without and with the Luria-Delbrück Correction for Population Growth Rate [Labuda et al. 1996])

	Distance from PRNP locusa				Haplotype datab			Estimated aged
Marker	Mb	cM	θ	Allele	pd	pn	LDc (δ)	g	g0	gc
D20S116	.57	2.11	.0211	20	.81	.29	.73	14.6	6.3	20.9
D20S482	.27	1	.001	14	.93	.59	.83	18.6	12.4	31
D20S895	.35	1.3	.013	18+19	.87	.35	.8	17.1	7.3	24.4

^aGenetic distances (in centimorgans) were calculated using a conversion factor of 3.7 cM/Mb obtained from regression of centimorgan vs. megabyte values for 20p12-pter loci (mapping data are from LDB).

^bp_d and p_n are the frequencies for the marker allele on disease-mutation–bearing and normal chromosomes, respectively. Source of data: table 2 of Lee et al. (1999).

^cLinkage disequilibrium index, calculated according to Bengtsson and Thompson (1981): δ=(p_d-p_n)/(1-p_n).

^dg and g_c are the estimated number of generations obtained by use of Risch et al.’s (1995) algorithm, without and with the Luria-Delbrück correction of the genetic clock (Labuda et al. 1996), respectively: g=logδ/log(-θ) and g_c=g+g₀, where g₀=-(1/d)ln(θ×f_d), assuming d=0.5 and f_d=1/d.

Table 2.

Estimation of the Age of the E200K Mutation in the PRNP Gene, Causing CJD in Libyan Jews (Reich and Goldstein’s [1999] Iterative Method, without and with the Luria-Delbrück Correction for Population Growth Rate [Labuda et al. 1996])

	Distance from PRNP locusa				Mutation datab		Haplotype datac		Estimated aged
Marker	Mb	cM	θ	Allele	μ	f	pd	pn	g	g0	gc
D20S116	.57	2.11	.0211	20	.00056	.32	.81	.29	14	6.3	20.3
D20S482	.27	1	.001	14	.0021	.29	.93	.59	12	12.4	24.4
D20S895	.35	1.3	.013	18+19	.00056	.09	.87	.35	16	7.3	23.3

^aSee footnote a to table 1.

^bμ is the assumed frequency of mutation at the marker locus (dinucleotide repeats [D20S116 and D20S482], μ≈.00056; tetranucleotide repeat [D20S482], μ≈.0021; Weber and Wong 1993) and f is the observed frequency of all one-mutant neighbors of the ancestral allele in the control population (source of data, table 1 of Lee et al. 1999).

^cSee footnote b to table 1.

^dg and g_c are the estimated number of generations obtained by use of Reich and Goldstein's (1999) algorithm, without and with the Luria-Delbrück correction of the genetic clock (Labuda et al. 1996), respectively. To provide a complete model for the haplotype's evolutionary process, according to the method of Reich and Goldstein (1999), a Markov transition matrix (K) for each marker was generated, which gives the probabilities that any one haplotype will be transformed into any other one in a single generation. K was calculated as the weighted sum of matrices corresponding to recombination (R), mutation (M), and no event occurring (I): K=θR+μM+(1-θ-μ)I, where μ is the frequency of mutation at the marker locus. The matrix R has the elements R₁₁=p_n, R₁₂=p_n, R₂₁=1-p_n, and R₂₂=1-p_n. Under the assumption of a stepwise mutation model for microsatellites (Goldstein and Pollock 1997) and a distribution of marker allele sizes on disease chromosomes that matches that seen in the control population, M has the elements M₁₁=0, M₁₂=f/2, M₂₁=1, and M₂₂=1-f/2, where f is the frequency of all one-mutant neighbors of the ancestral allele in the control population. With the parameters of K specified, the number of generations that have passed since the foundation event was estimated by multiplying the state vector (q, 1-q) by K iteratively, until the observed proportion of ancestral haplotypes was reached (q≈p_d). Iteration began at a frequency vector of (1,0), corresponding to the archetypal condition of only ancestral haplotypes. The number of times that K had been multiplied yielded an estimate of g (Reich and Goldstein 1999). The iterative procedure was implemented through the facilities of the SPSSv.9 matrix language. Iterations stopped at an average distance of ±5×10^-3 from the corresponding p_d value, which is the maximum attainable accuracy, being p_d values accurate to ±1×10^-2. The Luria-Delbrück-corrected age (Labuda et al. 1996) is given as g_c=g+g₀, where g₀=-(1/d)ln(θ×f_d), assuming d=0.5 and f_d=1/d.