Risk of Parkinson Disease in Carriers of Parkin Mutations:  Estimation Using the Kin-Cohort Method FREE

Mutations in the Parkin gene (PARK2; GenBank AB009973)^1- 2 are associated primarily with early-onset Parkinson disease (PD), defined as age at onset (AAO) ranging from 45 years or younger to 55 years or younger,^3- 7 but have also been described in PD cases with an AAO older than 70 years.^7- 10 In PD cases with an AAO of 45 years or younger with a mode of inheritance consistent with autosomal recessive transmission, the frequency of Parkin mutations may be as high as 49%,³ whereas in cases without a family history of PD the range is 15% to 18%.^4,6 Age at onset is inversely correlated with the frequency of Parkin mutations in both familial³ and sporadic⁶ cases.

Several studies have compared the AAO of PD in heterozygous, compound heterozygous, and homozygous Parkin mutation carriers^3- 10 and found that heterozygous cases, both familial and sporadic, have an older AAO. Heterozygous Parkin mutation carriers are more frequently reported among sporadic than familial cases.⁷

Information on the risk of PD in individuals who carry Parkin mutations in either the homozygous, compound heterozygous, or heterozygous state (or penetrance) is essential for genetic counseling. The penetrance of Parkin mutations has only been reported for isolated families.⁷ Most of the previous study designs sampled PD cases based on family history of PD, which would bias penetrance estimates upwards.^11- 12 To obtain an unbiased estimate of risk, a population-based random sample would be desirable, but Parkin mutations are so rare in the population that such a sample would have to be extremely large to obtain sufficient precision in penetrance estimates.

To obtain unbiased estimates of the risk of PD in Parkin carriers despite the low population frequency of Parkin mutations,¹³ we used a kin-cohort study design^11- 12 applied to participants in the Genetic Epidemiology of Parkinson's Disease (GEPD) study.¹⁴ The kin-cohort design is highly efficient for estimating penetrance because the relatives' mutation status is not required for the analyses, thus reducing costs for genetic analysis.¹⁵

Details of the GEPD study have been previously described.^14,16- 17 Cases were ascertained based on AAO of motor signs of 50 years or younger (early-onset PD) or older than 50 years (late-onset PD). In this study, we included all 247 PD cases with an AAO of 50 years or younger. All cases were recruited from the Center for Parkinson's Disease and Other Movement Disorders at Columbia University, and early-onset PD cases were oversampled.¹⁴ From 412 controls in the GEPD study, 105 controls were randomly selected for complete sequencing of the Parkin gene.¹⁸ The majority of the controls were recruited by random-digit dialing, with frequency matching by age, sex, ethnicity, and area code/exchange. An additional sample of 40 controls of Hispanic descent, 11 African American controls, and 170 white controls participating in the GEPD study were used to examine Parkin variants that have not been previously described. All PD cases and control probands were seen in person and underwent an identical evaluation¹⁴ that included a medical history review and Unified Parkinson's Disease Rating Scale¹⁹ and videotape assessment.

Information on the family history of PD in first-degree relatives was obtained by conducting a reliable, validated interview with each case, control, and first-degree relative. For relatives who were deceased or otherwise unavailable for interview, the family history was obtained by interviewing the most knowledgeable informant.¹⁶ An algorithm was created to generate a final diagnosis for PD in each first-degree relative based on the family history interview and the direct interview with the relative. For relatives diagnosed as having PD, a level of certainty was assigned as definite, probable, possible, uncertain, and unlikely. For first-degree relatives who met the criteria for uncertain, possible, probable, or definite PD, we tried to obtain additional information in the form of an examination, medical records, or an independent interview by a neurologist. A best-estimate diagnosis of PD was assigned for each relative.¹⁶ Family history information including AAO of PD was available for 1330 relatives of 224 PD cases and 638 relatives of 103 controls included in the penetrance analysis. The institutional review board at the College of Physicians and Surgeons, Columbia University, New York, New York, approved this study.

All sequence variants identified in cases and controls were genotyped in ethnically matched controls. Sequence variants that were observed in controls, but at a frequency of higher than 1%, were classified as rare variants. The remaining variants were classified as mutations and included in the analysis based on the following 3 criteria: (1) the mutation is absent in controls; (2) the mutation is recurrent, has been reported in PD cases (unrelated) in more than 1 study, or changes an amino acid that is evolutionarily conserved and was predicted to effect protein function; (3) the mutation is located in the coding region and was predicted to change the amino acid sequence.

Mutation screening was performed in 247 cases and 105 controls to detect point mutations and exon deletions and duplications in the Parkin gene. In a previous study, we sequenced all Parkin exons and screened for exon deletions and duplications by semiquantitative multiplex polymerase chain reaction (PCR) in 101 cases and 105 controls.¹⁸ One control was subsequently found to have a putative splice variant IVS6-14 C>G that had not been reported in cases or controls. We considered this a rare variant and not a mutation and excluded it from the analysis. In this study, we report data on an additional 146 PD cases from the GEPD study who were screened for Parkin mutations by denaturing high-performance liquid chromatography (WAVE; Transgenomic Inc, Omaha, Nebraska), which has 100% sensitivity and specificity. Primers and denaturing high-performance liquid chromatography conditions used for analysis of the Parkin gene have been described previously.²⁰ Amplicons were either directly sequenced (n = 126) or analyzed using a Parkin genotyping array (n = 20)²¹ in DNA samples with abnormal elution profiles. The genotyping array has excellent sensitivity and specificity for detection of sequence variants compared with the gold standard of sequencing.²¹ The primers used for PCR amplification of Parkin exons 1 to 12 and intron-exon boundaries and sequencing have been described previously.²² Cycle sequencing was performed on the purified PCR product as per the manufacturer's instructions (BigDye; Applied Biosystems, Foster City, California). Products were analyzed on an ABI3700 genetic analyzer (Applied Biosystems). Chromatograms were viewed using Sequencher (Gene Codes Corp, Ann Arbor, Michigan) and sequence variants determined. All sequence variants identified in cases and controls were confirmed by analysis in a separate PCR followed by bi-directional sequencing. To identify genomic deletions and exon rearrangements in Parkin, semiquantitative multiplex PCR was performed as previously described.¹⁸

In addition to screening for mutations in Parkin, we genotyped 5 LRRK2 (GenBank NM_198578) mutations (G2019S, L1114L, I1122V, R1441C, and Y1699C) in 247 cases and 104 controls. The results of the analysis of LRRK2 in all participants in the GEPD study, including both early- and late-onset PD cases, and all controls have been reported.²³

Demographic and clinical characteristics were compared between cases and controls and between mutation carriers and noncarriers. The Fisher exact test for categorical characteristics and the independent (unpaired) t test for continuous characteristics were used to assess statistical significance. The penetrance of Parkin mutations was estimated using the kin-cohort method.¹¹ In this method, the genotypes of the relatives are first estimated using Mendelian principles and by the relationship of the relatives to the proband. Then the observed disease occurrence in the relatives is evaluated in relation to these estimated genotypes. This method assumes that, although the PD probands were sampled according to their AAO and thereby increasing the proportion of carriers among cases and their first-degree relatives, the relatives of these PD cases are representative of randomly chosen individuals with certain genotypes. Hence, familial influences on the relatives' PD risk other than the Parkin genotype are assumed to be negligible.

First, we computed the probability that each relative carries a mutation. Second, we used the kin-cohort method to estimate genotype-specific disease rates, using the consensus diagnoses of PD in the first-degree relatives.^16- 17 Third, we used Kaplan-Meier survival analysis to compute the cumulative risk of PD in the first-degree relatives of control probands. Kaplan-Meier analysis cannot be directly applied to estimate genotype-specific cumulative incidence in the relatives of cases because the carrier status in these relatives is unknown; however, the kin-cohort method allows calculation of cumulative incidence through a method similar to Kaplan-Meier analysis. Confidence intervals (CIs) for the penetrance and the cumulative incidence were computed using log-log transformation to ensure that the lower limits of the 95% CIs were positive.²⁴

Demographic and clinical characteristics of the cases and controls are given in Table 1. The mean (SD) AAO of the 247 cases was 41.8 (6.8) years, mean (SD) disease duration was 10.7 (7.6) years, and the mean (SD) total motor score (Unified Parkinson's Disease Rating Scale part III) was 20.3 (12.8). Of the 224 cases for whom family history information was available, 19 (8.5%) had a diagnosis of PD in a first-degree relative.

Of the 247 cases, 25 (10.1%) had a Parkin mutation; 5 (20.0%) were homozygous, 2 (8.0%) were compound heterozygous, and 18 (72.0%) were heterozygous (Table 2). Twenty-two of the mutations have been previously described in other studies.^2,21- 22 Three new mutations were identified that have, to our knowledge, not been previously published and that were not detected in any of our control samples (Iso298Leu, Asp18Asn, and Pro153Arg). Eleven different point mutations (c.81G>T, Gly319Gly, Arg42Pro, Arg275Trp, Met192Leu, Cys253Tyr, Asp280Asn, Iso298Leu, Arg366Gln, Asp18Asn, and Pro153Arg) and 4 different exon rearrangements (exon 5 deletion, exon 3-4 deletion, 40–base pair [bp] exon 3 deletion, and exon 2 deletion) were identified (Table 2). Point mutations included 9 missense mutations (Arg42Pro, Arg275Trp, Met192Leu, Cys253Tyr, Asp280Asn, Iso298Leu, Arg366Gln, Asp18Asn, and Pro153Arg), 1 synonymous substitution (Gly319Gly), and a noncoding 5′UTR mutation (c.81G>T). Exon deletions were found in 4 different exons (exons 2, 3, 4, and 5). Of the variants identified in 25 cases, 10 (40.0%) were found in exons encoding functional domains including the ubiquitin domain (exon 2) and RING1 domain (exon 7). Six cases carried the 40-bp exon 3 deletion, 5 cases carried Arg275Trp, and 2 carried Arg42Pro. We previously reported that the synonymous substitution, Leu261Leu, is a variant rather than a disease-associated mutation and thus have not included carriers of Leu261Leu in our estimates.¹⁸ Among the 25 carriers, 3 (12.0%) had an AAO of 20 years or younger (2 of 3 heterozygotes); 4 (16.0%) had an AAO of 21 to 30 years (3 of 4 heterozygotes); 6 (24.0%) had an AAO of 31 to 40 years (5 of 6 heterozgygotes); and 12 (48.0%) had an AAO of 41 to 50 years (8 of 12 heterozgyotes). Among all 247 early-onset PD probands, carriers represented 75.0% (3/4) of those with an AAO of 20 years or younger, 36.4% (4/11) of those with an AAO of 21 to 30 years, 8.2% (6/73) of those with an AAO of 31 to 40 years, and 7.5% (12/159) of those with an AAO of 41 to 50 years.

Demographic and clinical features of the 25 mutation carriers and 222 noncarriers are given in Table 3. The mean (SD) AAO of PD was significantly younger in carriers (36.5 [10.3]) than in noncarriers (42.4 [6.1]) (P = .01) but did not differ between heterozygotes compared with compound heterozygotes and homozygotes combined. Other comparisons of clinical features were not significant.

The clinical characteristics of first-degree relatives stratified by the probands' mutation status are given in Table 4. Information on the family history of PD in first-degree relatives was available for 23 of 25 carriers and 201 of 218 noncarriers. Of 23 carrier probands, 1 (4.4%) had a family history of PD. This proband was homozygous for a 40-bp deletion in exon 3 and had 2 affected siblings (AAO, 26 and 30 years).

The probability of a relative being a carrier, regardless of whether he or she was actually diagnosed as having PD, stratified by the proband's carrier status, is given in Table 5. In the calculation, the population frequency of Parkin mutations (p) was assumed to be 0.03%.^3,9 We have run a sensitivity analysis by taking p to be 2.8% (the upper limit of the 95% exact CI of the mutation frequency estimated from the controls) and the results did not change, suggesting that the estimates are robust to misspecification to the population frequency of Parkin mutations.

The expected genotype distribution in the relatives and the prevalence of a history of PD in relatives predicted to be carriers or noncarriers are given in Table 6. To obtain these prevalence estimates, we first computed the probability that each of the relatives was a carrier based on the observed carrier status in the probands (Table 1) and then combined the information on the predicted genotypes in the relatives with the observed PD diagnoses in the relatives. There were 93 relatives expected to be carriers (homozygotes or heterozygotes), among whom 2 had PD (Table 6). Therefore, the prevalence of a history of PD in carrier relatives was estimated at 2.2% (2/93) (95% CI, 0.3%-7.6%). Among relatives expected to be noncarriers (n = 1237), 19 had PD (Table 6); thus, the prevalence of a history of PD in noncarrier relatives was estimated at 1.7% (95% CI, 0.9%-2.4%). These 2 prevalence estimates did not differ significantly.

The cumulative incidence of PD in 1330 relatives of case probands (without regard to carrier status) was estimated to be 1.7% (95% CI, 0.9%-3.3%) to age 65 years and 5.9% (95% CI, 3.7%-9.3%) to age 80 years. Estimates of the cumulative risk of PD in Parkin carriers and noncarriers and the cumulative risk of PD in 647 relatives of controls are given in Table 7. Among relatives expected to be carriers, the cumulative incidence of PD was 7.0% to age 65 years and remained so, up to age 80 years. Among relatives expected to be noncarriers, the cumulative incidence was 1.7% to age 65 years and 6.1% to age 80 years. The ratio of the cumulative incidence to age 80 years for carriers vs noncarriers was 1.1 (95% CI, 0.1-18.8). Estimates of risk in carriers and noncarriers did not differ significantly with our sample size. The cumulative incidence in relatives of controls was similar to that in relatives expected to be noncarriers (1.1% to age 65 years and 4.1% to age 80 years).

We have reported that LRRK2 mutations may be responsible for familial aggregation in both early-onset PD and late-onset PD in the GEPD study.²³ To separate out the effect of the LRRK2 G2019S mutation (the only observed LRRK2 mutation in the cases), we repeated our analyses using 214 case probands who did not carry a LRRK2 mutation. However, we cannot exclude the possibility that they carry “other” mutations in the LRRK2 gene because we did not completely sequence the LRRK2 gene but only genotyped 5 previously reported LRRK2 mutations. There were 1274 relatives included in the analysis. Since these relatives were family members of probands who did not carry any of the 5 previously identified LRRK2 mutations, it is unlikely that they carry any of these mutations. The cumulative incidence to age 80 years for relatives expected to be noncarriers of either LRRK2 mutations or Parkin mutations was 5.9%, compared with 6.1% for relatives expected to be noncarriers of Parkin mutations. The cumulative incidence of PD to age 80 years in non-Parkin, non-LRRK2 carrier relatives was not significantly different from controls.

Using the kin-cohort method, we have shown that the cumulative risk to age 65 years in a relative of a case with early-onset PD who is not estimated to carry a Parkin mutation is not significantly greater than in the general population risk among controls. We estimated a cumulative risk of PD in carriers of Parkin mutations (age-specific penetrance) of 7.0% (95% CI, 0.4%-71.9%) up to age 80 years. We were unable to examine risk of PD among heterozygotes separately. Parkin heterozgyosity may not be sufficient for the development of PD. A recent article suggested that Parkin variants were equally common in cases and controls in predominantly noncoding regions.²⁵

The sampling scheme of our study, in which probands were ascertained without regard to their family history of PD, differs markedly from that in other studies that included only families with multiple affected individuals. As noted in the context of estimating the penetrance of mutations in BRCA1 and BRCA2, penetrance estimates are inflated when based on samples selected though family history,^12,15 and hence we believe our estimates are more representative of those in the general population than are those derived from familial samples. While the 95% CIs are extremely wide, the observed low frequency of PD in first-degree relatives of Parkin mutation carriers, 72.0% of whom are heterozygotes, may be a reflection of the low penetrance in carriers.

The frequency of Parkin mutations was 10.1% (95% CI, 8.0%-16.4%) in 247 cases with an AAO of 50 years or younger, which is within the range of other series with primarily sporadic cases (15%-18%)⁶ or population-based series (9%).²⁶ The mean AAO of our case sample was 41.6 years, and 48% of the cases had an AAO between 40 and 50 years. The frequency of Parkin mutations has been reported to decrease with increasing AAO.^3,6 While the age-specific frequencies of carriers are similar to other reported series,⁶ the high percentage of older cases in the present study may explain why we report a frequency at the lower end of the spectrum.

There are 3 important limitations in this study. Most importantly, the number of relatives of probands who are estimated to carry Parkin mutations is limited, which results in cumulative risk estimates with wide 95% CIs. We have limited our study to only those who participated in the GEPD study for whom we have accurate information on vital status, a necessary criterion for the kin-cohort method. In addition, only 105 controls were completely sequenced for the Parkin gene, which may explain our inability to detect differences between relatives of PD probands and relatives of control probands. The sample size required for a typical kin-cohort study is usually large, owing to the low population prevalence of the mutation being studied. Second, the diagnosis in the relatives was a best-estimate diagnosis. We have previously reported high sensitivity (95.5%) and specificity (96.2%) of our family history questionnaire based on examination of 141 relatives. Third, we did not have actual genotype data on relatives. We used the kin-cohort method to estimate penetrance in the absence of these data.

Some relatives may have been too young to manifest PD. Our sampling plan was to include all probands with an AAO of 50 years or younger regardless of their family history of PD and AAO in the relatives. The relatives of these probands are likely to be younger than the relatives of randomly selected patients with PD. In a study of affected sibling pairs, the mean AAO of heterozygous Parkin carriers was reported to be 49.6 years.¹⁰ Of the relatives in our sample, 49% were younger than 49.6 years (Table 4). Our estimates of the prevalence of a history of PD in the relatives do not account for this young age distribution and thus underestimate prevalence of PD according to Parkin genotype in the general population. We accounted for the younger age distribution by computing age-specific cumulative incidence of PD according to genotype.

One assumption that is required for the kin-cohort analyses to be valid is that the risk of PD in relatives within the same family is independent, given the proband's Parkin genotype. The presence of other genetic and environmental risk factors that aggregate in the families may violate this assumption and bias the penetrance estimation.^15,27 Because of the possibility of additional unspecified familial risk factors in the relatives of early-onset PD probands, penetrance estimates obtained from our sample can be applied to the population of relatives of the early-onset PD probands but not to the entire population.

We are currently recruiting a larger multicenter sample of early-onset PD cases and examining and obtaining DNA in relatives of carrier probands. We hope to use this new sample to refine our estimates of penetrance in heterozygous, homozygous, and compound heterozygous carriers.

Correspondence: Karen Marder, MD, MPH, Sergievsky Center, Unit 16, Columbia University College of Physicians and Surgeons, 630 W 168th St, New York, NY 10032 (ksm1@columbia.edu).

Accepted for Publication: June 21, 2007.

Author Contributions:Study concept and design: Wang, Harris, Ottman, and Marder. Acquisition of data: Clark, Mejia-Santana, Cote, Waters, Andrews, Ford, Frucht, Fahn, and Marder. Analysis and interpretation of data: Wang, Clark, Louis, Ottman, Rabinowitz, and Marder. Drafting of the manuscript: Wang, Clark, Mejia-Santana, and Marder. Critical revision of the manuscript for important intellectual content: Wang, Louis, Harris, Cote, Waters, Andrews, Ford, Frucht, Fahn, Ottman, Rabinowitz, and Marder. Statistical analysis: Wang, Louis, Andrews, Ottman, and Rabinowitz. Obtained funding: Clark, Fahn, Ottman, and Marder. Administrative, technical, and material support: Harris, Andrews, Fahn, and Marder. Study supervision: Clark, Mejia-Santana, Cote, and Waters.

Financial Disclosure: None reported.

Funding/Support: This study was funded by grants NIH NS36630 and NIH RR00645 (Dr Marder) from the National Institutes of Health and by the Parkinson's Disease Foundation (Drs Marder and Clark).

Additional Contributions: Paul Greene, MD, participated in the acquisition of data.