Parkin-deficient mice are not a robust model of parkinsonism

Abstract

Mutations in the human parkin gene cause autosomal recessive juvenile parkinsonism, a heritable form of Parkinson's disease (PD). To determine whether mutations in the mouse parkin gene (Park2) also result in a parkinsonian phenotype, we generated mice with a targeted deletion of parkin exon 2. Using an extensive behavioral screen, we evaluated neurological function, motor ability, emotionality, learning, and memory in aged Parkin-deficient mice. The behavioral profile of Parkin-deficient mice on a B6;129S4 genetic background was strikingly similar to that of control mice, and most differences were not reproducible by using coisogenic mice on a 129S4 genetic background. Moreover, catecholamine levels in the striatum, olfactory bulb, and spinal cord of Parkin-deficient mice were normal. In contrast to previous studies using independently generated Parkin-deficient mice, we found no evidence for nigrostriatal, cognitive, or noradrenergic dysfunction. Understanding why Parkin-deficient mice do not exhibit robust signs of parkinsonism could advance knowledge and treatment of PD.

Parkinson's disease (PD) is a devastating neurodegenerative disorder that affects >500,000 people in the United States alone (1). The age-related and progressive signs of PD classically include resting tremor, muscular rigidity, abnormal gait, and slow movement; however, other signs and symptoms associated with parkinsonism can include somatosensory deficits, impaired cognitive function, and psychiatric disturbances (2-4). Many PD signs result from the degeneration of dopaminergic neurons in the substantia nigra pars compacta. Dopamine replacement medications often succeed in managing parkinsonism; however, these treatments lose effectiveness, have undesirable effects, and do not prevent the underlying neurodegeneration. Understanding the molecular mechanism of dopaminergic neuron degeneration would facilitate the development of therapies that prevent PD.

The identification of mutations in single genes that result in familial parkinsonism will help advance our understanding of the molecular mechanism of PD. Mutations in the human parkin gene are responsible for autosomal recessive juvenile parkinsonism (AR-JP), a heritable disease that resembles PD (5). The Parkin protein is a widely expressed, E3 ubiquitin-protein ligase that is thought to target specific proteins for proteasomal degradation (5-7). Presumably, in the absence of Parkin function, these protein targets accumulate to toxic levels and cause dopamine neuron degeneration.

To investigate how mutations in parkin lead to parkinsonism, we generated mice with a targeted disruption of Park2. We hypothesized that Parkin-deficient mice would recapitulate the behavioral signs and pathology of AR-JP. During our investigation, several other groups independently generated and described mice with various targeted deletions of the parkin gene (8-11); however, results from these studies have raised many new issues. First, the behavioral phenotypes of these different lines of mice are inconsistent; understanding the reasons for these discrepancies will be critical for planning and interpreting studies using Parkin-deficient mice. Second, the effect of genetic background on the phenotypes of Parkin-deficient mice was not investigated in previous studies; phenotypes modified by genetic background could be valuable for understanding AR-JP. Third, previous behavioral studies were limited primarily to investigating motor function. A more detailed behavioral profile could reveal deficits in additional behaviors relevant to parkinsonism or other neurological disorders. To address these inconsistencies and limitations, we provide an extensive and rigorous report of general neurological function, motor ability, emotionality, learning, memory, and neurochemistry in aged Parkin-deficient mice.

Methods

Generation of Parkin-Deficient Mice. The mutant parkin allele lacking exon 2, Park2^tm1Rpa, was generated by homologous recombination in mouse AK18.1 ES cells (129S4/SvJaeSor). About 4 kb of parkin genomic sequence, including exon 2, was replaced with the neo^r gene driven by the Polr2a gene promoter as described in Fig. 3, which is published as supporting information on the PNAS web site. Chimeras were generated by injecting targeted ES cells into C57BL/6 (B6) blastocysts. Park2^tm1Rpa/Park2⁺ mice were established on a mixed B6;129S4 genetic background by crossing chimeras to B6 mice. Male chimeras were also bred to female 129S4/SvJaeSor mice to establish the Park2^tm1Rpa mutant allele on a coisogenic background.

Behavioral Testing. All procedures adhered to the Guide for the Care and Use of Laboratory Animals (12) and were approved by the University of Washington Institutional Animal Care and Use Committee. At 18 months of age, male Parkin-deficient mice (Park2^tm1Rpa/Park2^tm1Rpa, hereafter called “KO,” n = 30) and male control mice (Park2⁺/Park2⁺, hereafter called “WT,” n = 27) began a behavioral screen that lasted ≈4 months. Mice used in the behavioral screen were generated by crossing F₁ Park2^tm1Rpa/Park2⁺ mice on a mixed B6;129S4 genetic background. Mice were housed in groups containing both genotypes in a specific-pathogen-free facility, maintained on a 12-hr light cycle, and provided with free access to water and food (5053, LabDiet, Richmond, IN). Behavioral testing occurred during the light cycle (with one test performed per day) and was conducted by the same investigator without knowledge of genotypes. To achieve the large, age-matched sample size and to control for slight differences in prior behavioral testing, four sets of mice were tested (approximately seven WT and approximately seven KO). Initial statistical analyses included set as a factor, and data were pooled when the performance of WT and KO mice between sets was not significantly different (criteria for significance was P < 0.05). The sequence of behavioral tests was designed to minimize potential carryover effects and to characterize the following behavioral categories: general health and neurological function, motor ability, emotionality, learning, and memory (Table 2, which is published as supporting information on the PNAS web site). Detailed methods and references for all behavioral tests are available in Supporting Appendix 1, which is published as supporting information on the PNAS web site.

Neurochemical Analyses. HPLC with electrochemical detection was used to determine the concentrations of norepinephrine, dopamine, and the dopamine-related metabolites 3-methoxytyramine, 3,4-dihydroxyphenylacetic acid, and homovanillic acid in the striatum, olfactory bulb, and spinal cord of aged male mice. Details are available in Supporting Methods, which is published as supporting information on the PNAS web site.

Statistical Analyses. Behavioral data were typically analyzed by using a repeated measures ANOVA with genotype as a factor and trial or day as a repeated measure (statistica, StatSoft, Tulsa, OK). In the behavioral screen, 442 statistical hypotheses involving genotype were tested. If the traditional threshold for statistical significance (P < 0.05) is used, 22 statistically significant differences between WT and KO mice could be expected by chance. We addressed this issue by using three increasingly stringent approaches. First, all differences between genotypes with 0.01 < P < 0.05 or P < 0.01 were identified, thereby facilitating individual interpretation of the results. Second, we sought to confirm findings where P was <0.05 by using an independent set of mice on a 129S4 genetic background; however, this approach can only support a significant genotype effect if the findings are reproducible. Significant findings may not be reproducible across strains because of differences in modifier genes, genes linked to parkin, or baseline behavioral performance. Third, a formal approach was taken to control the false discovery rate by using the Benjamini-Hochberg and Benjamini-Liu procedures (13).

Results

Generation of Parkin-Deficient Mice. Exon 2 of parkin was deleted because it encodes most of the functionally important ubiquitin-like domain (14-16), mutations in exon 2 have been reported in patients with AR-JP (3, 17-19), and RNA splicing from exons 1 to 3, if successful, is predicted to change the reading frame and produce a 4-aa peptide (Met-Ile-Glu-Leu). Deletion of exon 2 was confirmed at the parkin genomic locus with Southern analysis and PCR (Fig. 3 B and C). Mutant parkin mRNA lacking exon 2 was detected in the brain of KO mice by using RT-PCR (Fig. 3D). Sequence analysis confirmed a frameshift in the mutant parkin cDNA predicted to encode a 4-aa peptide. The presence of mutant parkin mRNA is consistent with previous findings in other Parkin-deficient mouse models (8-10). Although very low levels of parkin splice variants have been previously detected in mice, including variants lacking exons 2, 3, or 7, these could reflect unsuccessful splicing events because the parkin gene spans ≈1.2 megabases and Parkin protein variants corresponding to these isoforms have not been detected (20). Using RT-PCR to amplify parkin cDNA corresponding to exons 1 and 6, we detected minor products with sizes consistent with parkin splice variants lacking exons 2 or 3 in WT mice (20). Moreover, we detected a faint RT-PCR product (≈350 bp) in WT and KO mice that could support the existence of a previously unrecognized parkin splice variant lacking both exons 2 and 3. Parkin protein (≈50 kDa) was absent in KO brains that were analyzed by using the PRK8 C-terminal Parkin antibody (Fig. 3E). When Western blots were overexposed, a faint (≈40-kDa) band was detected in WT and KO mice (21). Although this finding could reflect nonspecific binding of the PRK8 antibody, a parkin mRNA splice variant lacking both exons 2 and 3 could encode an ≈37-kDa protein missing the functionally important ubiquitin-like domain. This potential Parkin variant would not have been detected in previous studies using antibodies directed against the peptide region encoded by exon 3 (8, 20); however, if this were a real Parkin variant, it would have been detected in previous studies using C-terminal antibodies (9, 10).

Behavioral Testing. To determine whether deletion of exon 2 from parkin leads to behavioral signs consistent with parkinsonism, large cohorts of 18- to 22-month-old WT and KO mice (on a B6;129S4 genetic background) were tested by using an extensive battery of 25 behavioral tasks designed to evaluate general neurological function, motor ability, emotionality, learning, and memory. Many of the motor function tests were also conducted on younger mice. For behavioral tests where a significant difference between genotypes was detected in the B6;129S4 mouse strain, we also tested, when possible, WT and KO mice on a coisogenic 129S4 strain to evaluate the effect of genetic background on the observed phenotype (typically >10 mice per genotype were used). Results from behavioral experiments are briefly described below and summarized in Fig. 1. Comprehensive results are available with extensive statistical analyses in Supporting Appendix 1.

Fig. 1.

Behavioral profile of Parkin-deficient mice (B6;129S4) at various ages. An interactive version of this figure with links to detailed methods, results, and statistical analyses is available in Supporting Appendix 1. Differences between WT and KO mice at the 0.01 < P < 0.05 level are indicated by a medium-gray box; differences between WT and KO mice at the P < 0.01 level are indicated by a black box. After controlling the false discovery rate at 0.10, none of these findings would be considered statistically significant. Light-gray boxes reflect tests in which no differences were detected (P > 0.05). Previously reported behavioral findings in Parkin-deficient mice that were not reproducible in the present study are indicated with an “X.” (A) KO mice weighed less on a B6;129S4 genetic background but not on a 129S4 genetic background. (B)KO mice demonstrated appropriate body-temperature regulation and nest-building behavior in response to cold exposure but consumed more food than WT mice. (C) KO mice exhibited a reduced visual placement response. (D) KO mice were less likely to display aggressive behavior. (E) KO mice displayed increased sensitivity in the acoustic-startle response; 129S4 mice did not display a startle response. (F) The reduction in daytime locomotor activity from Day 1 to Day 2 was more dramatic for KO mice. (G) Locomotor activity for KO mice was greater during the second dark cycle. (H) KO mice were more likely to grip onto the rotarod. (I and J) KO mice demonstrated increased endurance, but only for one trial; no differences were detected with 129S4 mice. (K) KO mice took longer to orient themselves; no differences were detected with 129S4 mice. (L and M) KO mice took less time to turn, but only for one trial. (N) KO mice spent more time immobile only during the third minute of the test. (O) KO mice displayed enhanced learning; 129S4 mice could not be tested because of long latencies to enter the dark chamber during training.

General Health and Neurological Assessment. We observed no difference in morbidity or mortality between WT and KO mice up to 24 months of age. The number of aged mice that did not complete the behavioral screen because of illness was also similar (WT, 3/27; KO, 5/30). Although abnormal behaviors such as excessive grooming or episodes resembling paroxysmal dystonia were occasionally observed during testing, there was no difference between WT and KO mice. Extensive neurological and physical examination of WT and KO mice by using the SHIR Phenotype Assessment primary screen (22) revealed no consistent differences.

The mean body weight for KO mice was slightly less than that for WT mice at 6 months of age, but not at 3, 12, 18, or 22 months of age. Our results are partially consistent with previous studies that found a significantly lower body weight for mice with targeted deletions of parkin exon 3 (B6;129) starting at 1-2 months of age (8, 11); however, we detected no significant differences in body weight before 6 months of age. Moreover, no body weight differences were detected on a 129S4 genetic background. Body weight differences also were not detected in mice (B6;129) with a targeted deletion of parkin exon 7 (10).

The mean body temperature of aged KO mice was indistinguishable from that of WT mice over a 24-hr period or in response to a 2-hr cold challenge. Mean body temperature was also indistinguishable at 3 months of age. Our results are in disagreement with a previous study that found a decreased body temperature in Parkin-deficient mice (8).

Somatosensory function was evaluated by using the adhesive-removal test and the tail-flick test for nociception. In the adhesive-removal test, mice had to remove various sizes of adhesive labels affixed to their forehead. There were no differences between 19-month-old WT and KO mice in the latencies to remove the adhesive labels (Fig. 2A). This finding is in partial disagreement with a previous study that found deficits in the ability of younger Parkin-deficient mice to remove small adhesive labels (9). In the tail-flick assay for nociception, we found no differences between genotypes.

Fig. 2.

Intact neurological function, motor ability, emotionality, learning, and memory in 18- to 22-month-old Parkin-deficient mice. (A) Latency to remove labels in the adhesive-removal test for somatosensory function. (B) Exploratory locomotor activity over 4 h. (C) Latency to fall off the rotarod for four trials per day over 5 days. (D) Time spent immobile during the forced-swim test for depression-related behavior. (E) Latency to locate the platform in the Morris water maze test for learning and memory. (F) Locomotor activity for 2 h after a saline injection (Sal) or after repeated administration of amphetamine (A1 and A2) in 3-month-old mice. Normally distributed data are shown as mean and SEM or 95% confidence interval (CI); data that are not normally distributed are shown as median and quartiles (Q25, Q75). Full results are published in Supporting Appendix 1.

To evaluate hearing ability and auditory threshold, the acoustic-startle response was tested. Compared with WT mice, KO mice exhibited an increased flinch response to low-intensity auditory stimuli. We attempted to confirm our findings in 18- to 22-month-old WT and KO mice on a 129S4 genetic background, but neither exhibited a startle response, consistent with a previous report (23). Moreover, the maximum acoustic-startle responses can differ by nearly 1 order of magnitude between strains of mice, including the B6 and 129 substrains (23). The increased sensitivity we observed in KO mice is not consistent with a previous study that found a reduced startle response in Parkin-deficient mice on a B6;129 genetic background (10).

Motor Function. To identify disturbances in locomotor activity consistent with parkinsonism, we evaluated general locomotor activity in a novel environment and over a 48-hr period; no differences were detected between WT and KO mice at 3, 6, or 18 months of age (Fig. 2B). However, 12-month-old KO mice exhibited greater locomotor activity specifically during the second dark cycle. We also conducted detailed analysis of exploratory locomotor behavior in an open-field arena (24) by using a video-tracking system; no differences between genotypes were detected. Our results are in disagreement with a previous study that found decreased locomotor activity in Parkin-deficient mice (8).

To identify deficits in motor coordination or gait, mice were trained to traverse balance beams of various sizes, tested on the accelerating rotarod, and evaluated by using footprint gait analysis. No differences were detected between aged WT and KO mice on the balance beam. Our results are inconsistent with a previous study that found deficits in Parkin-deficient mice by using the balance-beam test (9). In the rotarod task, we found no deficits in learning or performance for KO mice at 3, 12, or 18 months of age (Fig. 2C); however, 6-month-old KO mice were more likely to grip the rotarod as it rotated. Gait analysis revealed no differences between aged WT and KO mice.

To evaluate motor strength, the hanging-wire-grip test was performed. The total time each mouse could hang suspended from an inverted wire grid was recorded for three trials. KO mice exhibited greater endurance limited to a single trial at 6 and 18 months of age, but not at 12 months of age. To confirm these findings, we tested mice on a 129S4 genetic background at 6 and 18-22 months of age; no differences in performance were detected between genotypes.

To test for akinesia or bradykinesia, the negative-geotaxis test and catalepsy test were performed. Mice were placed in a head-down position on a vertical wire grid, and the latency to orient in the preferred head-up position was recorded for three trials. KO mice took longer to orient to a head-up position at 18 months of age, but not at 6 or 12 months of age. Aged WT and KO mice on a 129S4 genetic background were also tested, and no differences between genotypes were detected. In the catalepsy test, no differences between WT and KO mice were found at 6, 12, or 18 months of age.

To detect motor deficits due to nigrostriatal dysfunction, the pole test was conducted. Mice were placed at the top of a vertical pole, and the latency to descend was scored for three trials. No differences were observed in the latencies to descend halfway or completely down the pole. However, KO mice displayed enhanced turning performance limited to a single trial at 6 and 18 months of age, but not at 3 months of age.

Emotionality. To evaluate depression-related behavior, the forced-swim test and tail-suspension test were performed. No differences were detected between aged WT and KO mice in the forced-swim test at 12 or 20 months of age (Fig. 2D). In the tail-suspension test, aged KO mice spent more time immobile only during the third minute of the test. There were no significant differences between genotypes at any other time points or in the total time spent immobile during the final 4 min of the test. To evaluate anxiety-related behavior, the light-dark exploration test and elevated-plus maze test were conducted. No differences were detected between genotypes.

Learning and Memory. The ability to explore and recognize a familiar object was tested by using the novel-object-recognition test; no differences between genotypes were detected. To evaluate spatial working memory and exploratory behavior, multiple T-maze spontaneous alternation tasks were performed; no differences between genotypes were detected on either a B6;129S4 or 129S4 genetic background. Although there were days when >80% of WT or KO mice (B6;129S4) exhibited alternation, the mean alternation rate for individual mice from either genotype was not significantly greater than chance. WT 129S4 mice also failed to display spontaneous alternation. Although B6 mice exhibit spontaneous alternation, some 129 substrains do not (25). We were unable to verify a previous report of decreased T-maze spontaneous alternation in Parkin-deficient mice on a B6;129 genetic background (8).

To evaluate various aspects of learning and memory, mice were trained to swim through various water mazes of increasing complexity to locate an escape platform. No differences in performance were detected between aged WT and KO mice in the straight-alley-swim escape task, black-white-discrimination-swim task, or challenging Lashley III water maze. To evaluate spatial learning and memory, the Morris water maze task also was performed. Mice were trained to locate an escape platform hidden beneath the surface of a circular pool for four trials per day over 6 days; on Day 7, the platform location was moved to evaluate reversal learning. WT and KO mice demonstrated equivalent spatial learning as determined by using multiple measurements such as the latency to locate the platform (Fig. 2E).

To evaluate associative learning and memory, the passive-avoidance test was performed. In this task, mice had to learn and remember an association between the preferred dark chamber and an aversive foot shock. There was no difference between 21-month-old WT and KO mice in the latency to enter the dark compartment on the training day and 24 hr or 7 days after training. However, 12-month-old KO mice exhibited enhanced learning by taking longer to enter the dark compartment 24 hr after training compared with WT mice. We attempted to confirm this finding by using aged mice on a 129S4 genetic background; however, the long latency to enter the dark chamber for WT and KO mice during the training session precluded further testing.

Amphetamine-Induced Locomotion. To screen for disturbances in dopaminergic neurotransmission, we measured the locomotor response to repeated amphetamine treatments (4 mg/kg of body weight) in 3-month-old male mice (B6;129S4). No differences were detected between genotypes. Our results disagree with a previous report that found a decreased amphetamine-induced locomotor response in Parkin-deficient mice (8).

Neurochemical Analyses. Results from neurochemical analyses are summarized in Table 1. The levels of dopamine and metabolites in the striatum of 22-month-old WT and KO mice were indistinguishable. There were also no differences when the data were analyzed as ratios that indicate dopamine turnover. We detected no differences in olfactory bulb or spinal cord norepinephrine levels between 18- to 22-month-old WT and KO mice. Our results are inconsistent with previous studies that found decreased norepinephrine in the olfactory bulb and spinal cord (10) or an elevated 3,4-dihydroxyphenylacetic acid/3-methoxytyramine ratio in the striatum (8) of Parkin-deficient mice.

Table 1. Neurochemical analyses.

Region and measurement	WT	KO
Striatum (B6; 129S4)	n = 12	n = 12
Dopamine	46,000 ± 3,100	53,000 ± 4,300
Norepinephrine	2,600 ± 380	2,400 ± 430
3-methoxytyramine	3,200 ± 230	3,800 ± 330
3,4-dihydroxyphenylacetic acid	7,000 ± 230	7,200 ± 420
Homovanillic acid	8,000 ± 390	8,600 ± 470
3,4-dihydroxyphenylacetic acid/3-methoxytyramine	2.3 ± 0.15	2.0 ± 0.15
Olfactory bulb (129S4)	n = 11	n = 11
Norepinephrine	2,500 ± 67	2,300 ± 170
Spinal cord (129S4)	n = 11	n = 11
Norepinephrine	4,600 ± 280	4,800 ± 400

The units of measurement are pg/mg of protein (except for ratio data); data are expressed as mean ± SEM.

Histopathological analyses also failed to detect any differences between the brains of aged WT and KO mice by using the following stains: cresyl violet, hematoxylin/eosin, Congo red, or Gomori's iron stain. Immunohistochemistry using antibodies for tyrosine hydroxylase, ubiquitin, and α-synuclein also revealed no differences between genotypes. Specifically, substantia nigra pars compacta dopamine neurons were intact in KO mice. Extensive necroscopic examination of most organ systems in a large number of aged WT and KO mice also revealed no consistent differences in the presence of gross pathology such as tumors (data not shown).

Analysis of Statistical Results. In the behavioral analyses, we identified 25 differences out of 442 statistical tests involving genotype where P was <0.05; however, 22 statistically significant differences could be expected by chance alone. The behavioral tests associated with these statistical findings are summarized in Fig. 1. We sought to minimize the likelihood of false positives among these findings by controlling the false discovery rate at a liberal level of 0.10 by using either the Benjamini-Liu step-down procedure or the Benjamini-Hochberg step-up procedure (13). Both procedures suggest that none of the behavioral findings meet criteria for statistical significance after controlling the false discovery rate.

Discussion

Parkin-deficient mice generated in our laboratory demonstrated no consistent deficits in neurological function, emotionality, learning, or memory that were suggestive of parkinsonism. After controlling for the large number of behavioral tests we conducted, we were unable to detect any behavioral abnormalities in Parkin-deficient mice greater than what could be expected by chance. At a neurochemical level, striatal dopamine was normal in 22-month-old KO mice. Moreover, substantia nigra pars compacta dopamine neurons were intact, and no pathology was associated with aged KO mice. Based on these findings, we conclude that Parkin-deficient mice do not recapitulate signs central to AR-JP.

Previous reports indicated that Parkin-deficient mice display nigrostriatal, cognitive, or noradrenergic dysfunction (8-10). A summary of inconsistent findings in various Parkin-deficient mouse models can be found in Table 3, which is published as supporting information on the PNAS web site. The inconsistent phenotypes observed by different laboratories could be a consequence of genetic background differences between the mouse strains tested. Although these previous studies also used Parkin-deficient mice on a B6;129 hybrid genetic background, 129 substrains are not genetically identical (26, 27), and genes that modify the phenotypes of Parkin-deficient mice may differ among mouse strains. For example, Parkin-deficient mice with a targeted deletion of exon 7 (B6;129) have fewer norepinephrine neurons in the locus coeruleus and exhibit a concomitant decrease in norepinephrine levels in the olfactory bulb and spinal cord (10). In contrast, quaking mutant mice (Qk^qk), which are Parkin-deficient because of a spontaneous deletion of parkin, parkin co-regulated gene, and dysregulation of the nearby Qk gene, exhibit the opposite phenotype, an increase in the number of norepinephrine neurons in the locus coeruleus (28-30). We found that 129S4 mice with a targeted deletion of parkin exon 2 exhibit normal levels of norepinephrine in the olfactory bulb and spinal cord. Therefore, the genomic context of the parkin deletion could lead to striking differences in the integrity of the noradrenergic system. It is possible that additional genes could modify other phenotypes that have been observed in Parkin-deficient mice.

In addition to genetic background, the type of parkin mutation could explain the phenotypic differences observed in various Parkin-deficient mouse models. For example, the targeting strategy used to delete exon 3 produces a detectable mutant protein consisting of the Parkin ubiquitin-like domain fused to EGFP (9). Findings in these mice could be due to the loss of Parkin function or expression of a mutant Parkin protein. Moreover, alternative splicing of the parkin gene could produce many functionally important protein variants that are difficult to detect (20, 31). Depending on the exon targeted, specific protein variants could be differentially affected in various Parkin-deficient mouse models, thereby resulting in distinct phenotypes.

Although the phenotypes of Parkin-deficient mice could be modified by genetic background or the type of targeted mutation, carefully controlled studies are required to rule out the possibility that these phenotypes are not the result of parkin deletions but could reflect artifacts due to several technical confounds of gene targeting in mice (32, 33). First, some Parkin-deficient mice were generated by using the Pgk-neo^r cassette, which can affect the regulation of neighboring genes (34, 35) and could produce phenotypes that are not the result of the parkin deletion. Second, testing Parkin-deficient mice on a mixed B6;129 genetic background, as has been done in the current and previous studies, can lead to false-positive findings, depending on the segregation of 129- vs. B6-derived genes. Compared with B6 mice, 129 substrains could be considered parkinsonian because of reduced locomotor activity, poor cognitive ability, and passivity (25, 36-38). In addition to dramatic differences in behavior, including the acoustic-startle response (23), T-maze spontaneous alternation (25), and amphetamine-induced locomotion (39), there are significant differences in electrophysiological characteristics (40), markers for oxidative stress (41), and the number of locus coeruleus neurons between mouse strains (42). Third, although backcrossing to a B6 genetic background can help reduce the likelihood of false positives due to genetic heterogeneity, genes closely linked to parkin will remain 129-derived in Parkin-deficient mice but B6-derived in control mice because the targeted parkin deletions were generated by using ES cells derived from 129 mice (32, 33, 36). Even after 12 backcrosses to B6 mice, ≈16 cM surrounding the mutant parkin allele will still be 129-derived; however, in control mice, the corresponding region will be B6-derived (33, 43). Systematic, strain-dependent differences in genes linked to parkin could lead to phenotypes that are mistakenly attributed to the parkin deletion. For example, genes and quantitative trait loci closely linked to parkin are confounding because they are known to modify body weight, locomotor response to amphetamine, the acoustic-startle response, body temperature, muscle endurance, nigrostriatal signaling, proteasome function, mitochondrial transcription, oxidative stress, and susceptibility to dopaminergic neurotoxins. Moreover, polymorphisms in several of these closely linked genes are known to cause functional differences between mouse strains. Specific examples of confounding genes linked to parkin are identified in Supporting Appendix 2, which is published as supporting information on the PNAS web site.

The current study reinforces fundamental concerns regarding the use and interpretation of studies using gene targeting in mice to model PD. Although the confounds of gene targeting have been known for many years, it is clear from previous studies using Parkin-deficient mice that these confounds have not been adequately considered. Unless these issues are addressed, the potential contribution of gene-targeting studies to our understanding of PD and other neurological diseases could be severely limited. Strategies to overcome these confounds, such as transgenic rescue of phenotypes, alternative breeding strategies, or the use of coisogenic mice, should be adopted in future work (32, 33, 44). We partially addressed these issues by testing a large number of B6;129S4 mice and attempting to confirm significant differences by using 129S4 WT and KO mice that are expected to be identical at all loci except parkin. Ultimately, we cannot conclude that the phenotypes we observed are due to the parkin mutation because they were not reproducible with 129S4 mice.

Despite the inconsistent phenotypes that have been reported in Parkin-deficient mice and the potential confounds of gene targeting, it is important to emphasize the surprising absence of a robust parkinsonian phenotype. Identifying reasons why Parkin-deficient mice do not exhibit parkinsonism, such as redundant E3 ubiquitin-protein ligases, the absence of appropriate environmental triggers, compensation, or species-dependent Parkin function, could advance knowledge and treatment of PD.