Cerebrospinal Fluid Aβ₄₂ Levels and APP Processing Pathway Genes in Parkinson’s Disease

Abstract

Background

Of recent interest is the finding that certain CSF biomarkers traditionally linked to Alzheimer’s disease (AD), specifically amyloid beta protein (Aβ), are abnormal in PD CSF. The aim of this exploratory investigation was to determine if genetic variation within the amyloid precursor protein (APP) processing pathway genes, correlate with cerebrospinal fluid (CSF) Aβ₄₂ levels in Parkinson’s disease (PD).

Method

PD (n=86) and control (n=161) DNA were genotyped for 19 regulatory region tagging single nucleotide polymorphisms (SNPs) within nine genes (APP, ADAM10, BACE1, BACE2, PSEN1, PSEN2, PEN2, NCSTN and APH1B) involved in the cleavage of APP. SNP genotypes were tested for their association with CSF biomarkers and PD risk while adjusting for age, gender, and APOE ε4 status.

Results

Significant correlation with CSF Aβ₄₂ levels in PD was observed for two SNPs, (APP rs466448 and APH1B rs2068143). Conversely, significant correlation with CSF Aβ₄₂ levels in controls was observed for three SNPs (APP rs214484 and rs2040273 and PSEN1 rs362344).

Conclusion

The results of this exploratory investigation suggest that an APP SNP and an APH1B SNP are marginally associated with PD CSF Aβ₄₂ levels in APOE ε4 non-carriers. Further hypotheses generated include that decreased CSF Aβ₄₂ levels are in part driven by genetic variation in APP processing genes. Additional investigation into the relationship between these findings and clinical characteristics of PD, including cognitive impairment, compared to other neurodegenerative diseases, such as AD, are warranted.

Introduction

Amyloid precursor protein (APP) is an integral membrane protein that is best known as the protein whose processing generates amyloid beta (Aβ), a peptide that is the primary component of the amyloid plaques characteristic of Alzheimer’s disease (AD). Aβ peptides of varying sizes are found in cerebrospinal fluid (CSF). Low CSF Aβ₄₂ levels are a promising biomarker for AD¹ and have been described in Parkinson’s disease (PD)².

Many proteins are involved in the post-translational cleavage of APP into Aβ. An α-secretase, ADAM10, cleaves APP in one processing pathway that does not form an amyloidgenic Aβ peptide^3–7. In another processing pathway, APP cleavage by β-secretases BACE1 or BACE2 produces a peptide, which subsequently can undergo γ-secretase cleavage to produce an amyloidgenic Aβ peptide^8–12. The γ-secretase protein complex is composed of several subunits including presenilin (either PSEN1 or PSEN2), nicastrin (NCSTN), APH1B and PEN2^13–15, and is uniquely involved in APP cleavage according to tissue specificity, or the stability of the protein complex^16–20.

Rare autosomal-dominant forms of early-onset familial AD, are caused by mutations in the PSEN1, PSEN2 or APP²¹. The majority of the rare coding mutations in APP alter processing of APP so that the relative levels of Aβ₄₂ are increased^22–24. Triplication of the APP gene, due to chromosome 21 trisomy in Down’s Syndrome, is associated with increased APP expression and early amyloid plaque formation^25–28. In addition, APP promoter polymorphisms have been associated with AD^29,30.

Aβ peptides of varying sizes are normally present in both the brain and CSF. Low CSF Aβ₄₂ levels are associated with increased Aβ deposition in the brain^31–33, age and APOE ε4 genotype in cognitively normal adults³⁴, and in AD and mild cognitive impairment^35–37.

Some studies, but not all, report reduced CSF Aβ₄₂ associated with PD or cognitive decline in PD^38–40. In addition, the APOE ε4 allele appears to be a risk factor across the Lewy body disease (LBD) spectrum, including PD⁴¹.

Our group has reported an association between CSF biomarker levels and APP processing genes in AD⁴². However, to the best of our knowledge, the genetic influences of APP processing related genes on AD-associated CSF biomarkers have not been studied in PD. Therefore, we hypothesized that genetic variation within regulatory regions of APP processing genes, would correlate with CSF Aβ₄₂ levels in PD according to APOE ε4 status. Specifically, the aim of this investigation was to determine if genetic variation in common transcriptional regulatory regions of APP, ADAM10, BACE1, BACE2, PSEN1, PSEN2, PEN2, NCSTN and APH1B correlate with PD CSF Aβ₄₂ levels. A total of 19 single nucleotide polymorphisms (SNPs) were analyzed while taking into account age, gender, and APOE ε4 status.

Methods

Subjects

All procedures were approved by the institutional review boards of the participating institutions. Following informed consent, all PD subjects (n=86) underwent evaluation that consisted of medical history, family history, physical and neurologic examinations, and laboratory tests. All PD subjects fulfilled criteria for a diagnosis of PD⁴³.

All control subjects (n=161) underwent thorough clinical and neuropsychological assessment as prescribed by the Alzheimer Centers’ uniform data set⁴⁴. All control subjects had a Clinical Dementia Rating (CDR) scale score of 0, and underwent consensus conference review.

Cerebrospinal Fluid

All CSF samples were collected in the morning after participants fasted overnight. CSF samples were collected as previously described^34,45. Results reported here are from assays run from comparable lumbar puncture fractions to limit variability from rostrocaudal concentration gradients. Concentrations of Aβ₄₂ in the 5th to 10^th ml of collected CSF Aβ42 were measured using the INNO-BIA AlzBio3 kit obtained from Innogenetics (Gent, Belgium) following the manufacturer’s instructions except that the CSF samples were diluted 1:4 before performing the assay. CSF Aβ₄₂ was measured using multiplexed Luminex reagents from InnoGenetics according to manufacturer’s instructions and as previously described⁴⁶. All CSF samples were analyzed using a LiquiChip Luminex 200TM Workstation (Qiagen, Valencia, CA). Intra-assay coefficient of variation was <10 % for all assays. Assays were performed blind to clinical diagnosis.

Genes and SNP selection

The nine studied genes were chosen for their biologically characterized role in encoding proteins that are involved in the processing of APP. SNPs were chosen within these genes according to the following criteria; (1) the SNP was located within a known or putative regulatory region of the gene. Tagging SNPs were chosen to capture regulatory regions when the actual regulatory region SNP was not available; (2) the SNP had a minor allele frequency (MAF) of ≥ 0.1 in HapMap Northern European population (CEU) and a minor genotype frequency in our study sample of ≥ 0.01; and (3) the SNP genotyping assay was commercially available. Based on these criteria, a total of 19 SNPs were selected (Table 2). Additionally, SNPs rs429358 and rs7412 were genotyped to determine APOE ε4 status.

Table 2.

Genotype description and frequency distribution.

Gene ID	Chromosome Location	SNP	NCBI Gene Location	Alleles	PD Genotype Frequency			Control Genotype Frequency			PD vs Controls
											Unadjusted Genotype	Adjusted Genotype	Unadjusted Collapsed Genotype	Adjusted Collapsed Genotype
					EE	EF	FF	EE	EF	FF	p-value	p-value	p-value	p-value
Amyloid Precusor Protein
APP	21q21.3	rs466448	5' region	G/a	0.19	0.52	0.29	0.25	0.49	0.26	0.324	0.122	0.578	0.136
		rs2830101	intron 1	G/a	0.08	0.46	0.46	0.13	0.41	0.46	0.601	0.394	0.990	0.728
		rs214484	intron 17	C/g	0.07	0.47	0.46	0.11	0.42	0.47	0.697	0.909	0.917	0.987
		rs2040273	3' region	A/g	0.13	0.51	0.36	0.16	0.43	0.41	0.888	0.661	0.491	0.545
α-Secretase
ADAM10	15q22.1	rs514049	5' region	C/a	0.13	0.46	0.41	0.22	0.45	0.34	0.089	0.191	0.237	0.298
		rs2305421	Intron 13	A/g	0.04	0.21	0.75	0.04	0.27	0.70	0.414	0.778	0.345	0.933
β-Secretase
BACE1	11q23-24	rs573801	5' region	G/a	0.12	0.28	0.60	0.03	0.42	0.55	0.631	0.466	0.478	0.695
		rs11601511	intron 0	G/c	0.01	0.22	0.76	0.02	0.24	0.74	0.556	0.463	0.661	0.697
		rs638405	exon 5	C/g	0.14	0.53	0.33	0.16	0.50	0.34	0.929	0.449	0.848	0.948
BACE2	21q22.3	rs734757	intron 1	C/t	0.08	0.49	0.42	0.19	0.44	0.37	0.097	0.102	0.438	0.434
		rs12149	exon 9	C/t	0.22	0.47	0.31	0.25	0.47	0.28	0.557	0.252	0.665	0.417
γ-Secretase
PSEN1	14q24.3	rs362344	3' region	C/t	0.05	0.35	0.60	0.05	0.35	0.60	0.999	0.900	0.970	0.956
		rs362408	3' region	G/a	0.04	0.18	0.79	0.02	0.23	0.75	0.762	0.736	0.520	0.608
PSEN2	1q42.2	rs1295652	5' region	A/g	0.05	0.33	0.62	0.07	0.37	0.56	0.295	0.194	0.330	0.224
		rs2802268	3' region	T/g	0.04	0.28	0.68	0.06	0.40	0.55	0.048	0.083	0.040	0.078
PEN2 (PSENEN)	19q13.1	rs2293688	intron 2	C/g	0.11	0.47	0.42	0.12	0.48	0.39	0.570	0.964	0.625	0.925
NCSTN	1q22-q23	rs7540865	intron 5	G/t	0.07	0.48	0.45	0.11	0.41	0.48	0.978	0.414	0.577	0.223
APH1B	15q22.2	rs35408871	Intron 4	G/a	0.01	0.25	0.74	0.02	0.24	0.73	0.748	0.984	0.889	0.820
		rs2068143	intron 4	G/a	0.16	0.35	0.48	0.07	0.34	0.60	0.020	0.081	0.088	0.152

Adjusted p-values are based on a logistic regression model with diagnosis as the dependent variable and SNP as a predictor variable along with the covariates gender, age and APOE ε4 status.

P-values < 0.05 are in bold but do not remain significant after adjustment for multiple comparisons.

SNP Genotyping

Genomic DNA (10 ng) was genotyped using TaqMan allelic discrimination detection on 384 well plates as previously described⁴⁷. Briefly, for each reaction, SNP TaqMan Assay (Applied Biosystems), TaqMan Universal PCR Master Mix (Applied Biosystems) and DNA were pipetted into each well. PCR was carried out using a 9700 Gene Amp PCR System (Applied Biosystems). Plates were then subjected to an end-point read on a 7900 Real-Time PCR System (Applied Biosystems). The results were first evaluated by cluster variations; the allele calls were then assigned automatically before being integrated into the genotype database.

Seven PD patients with an age-at-onset of younger than 40 years were screened for Parkin (PARK2) mutations using a method previously described⁴⁸. One patient was found to be a compound heterozygote and was excluded from the analysis leaving 85 PD subjects for analysis (Table 1).

Table 1.

Population description.

	PD	Controls	χ2 p-value
	n=85	n=161
% Female	24	58	< 0.0001
% White	98	92	0.170
% APOE ε4+	25	37	0.060
Mean Age (Minimum-Maximum)	67 (48–83)	68 (52–88)	0.454
Mean Age-at-motor symptom onset (Minimum-Maximum)	57 (28–76)
MMSE (Minimum-Maximum)	27.6 (10–30)

Statistical Analysis

Logistic regression was used to compare SNP genotype (FF major homozygotes, EF heterozygotes, EE minor homozygotes) or collapsed genotype (EF, EE compared to FF; minor allele positive compared to minor allele negative) between PD and controls both with and without adjusting for age, gender and APOE ε4 status. SNP genotype frequencies were tested for Hardy-Weinberg equilibrium (HWE) using the chi-squared test with one degree of freedom⁴⁹ (Table 2). All subsequent analyses involving SNPs were based on the collapsed genotype group.

Linear regression models were used to compare PD and control CSF Aβ₄₂ protein levels with and without adjusting for gender, age, and APOE ε4 (Figure 1, Panel A) and to evaluate the relationship between APOE ε4 status and CSF Aβ₄₂ levels while taking into account gender and age (Figure 1; Panels B).

Figure 1.

Cerebrospinal fluid (CSF) levels stratified by disease status and APOE status. There is not a significant difference in CSF Aβ₄₂ levels between Parkinson’s disease (PD) and cognitively normal control subjects (Controls) (Panel A). There is a significant difference in CSF Aβ₄₂ levels in controls between APOE ε4+ and APOE ε4− but not PD (P-values are Bonferroni corrected for multiple comparisons: Panel B). Bars represent interquartile range for CSF Aβ₄₂ levels. P-values are adjusted for covariates gender and age. P-values in parentheses are not adjusted for covariates and bar graphs are not adjusted for covariates. The horizontal line within the quartiles represents the median. The vertical lines represent minimum and maximum CSF Aβ₄₂ levels. Circles represent outliers.

Linear regression main effect models both with a single SNP or all SNPs were used to examine the relationship between SNP (for collapsed genotype) and CSF Aβ₄₂ levels with and without adjusting for gender, age, and APOE ε4 status (Figure 2). Additional regression models included a SNP-by-APOE ε4 status interaction term or a term for group (1) ε4+ and SNP allele +, (2) ε4+ and SNP allele −, (3) ε4− SNP and allele +, and (4) ε4− and SNP allele −, both with and without adjusting for age and gender (Figure 3).

Figure 2.

Linear regression beta-coefficients for cerebrospinal fluid (CSF) Aβ₄₂ levels in cognitively normal control subjects (Controls: grey circle) or Parkinson’s disease patients (PD: black square) for each collapsed genotype. A confidence interval that does not cross the vertical line at zero indicates that the difference in genotype is significant (p < 0.05) before adjusting for multiple comparisons as indicated by asterisk (*). A beta coefficient to the left of the vertical line indicates lower CSF Aβ₄₂ levels for the collapsed genotype group (EE, EF) than for the major homozygote genotype (FF). Beta coefficients, 95% confidence intervals and p-values in parentheses are not adjusted for covariates.

Figure 3.

PD CSF levels stratified by APOE and SNP. There is a significant difference in CSF Aβ₄₂ levels between APOE ε4+ rs466448 G+ and APOE ε4− rs466448 G− (p=0.030) as well as between APOE ε4− rs466448 G+ and APOE ε4− rs466448 G− (p=0.035) (Panel A). There is a significant difference in CSF Aβ₄₂ levels between APOE ε4+ rs2068143 A− and APOE ε4− rs2068143 A+ (p=0.027) as well as between APOE ε4− rs2068143 A− and APOE ε4− rs2068143 A+ (p=0.021) (Panel B). Bars represent interquartile range for CSF Aβ₄₂ levels and width of bars a adjusted for number of subjects. P-values are adjusted for covariates gender and age. P-values in parentheses are not adjusted for covariates and bar graphs are not adjusted for covariates. The horizontal line within the quartiles represents the median. The vertical lines represent minimum and maximum CSF Aβ₄₂ levels. Circles represent outliers. All p-values are Bonferroni corrected for multiple comparisons.

Diagnostic plots of model residuals were inspected to assess any major departures from normality or homoscedasticity. Statistical analyses were performed in R (version 3.0.2; R Core Team, 2013; http://www.R-project.org.) or SPSS (Version 22). When correcting for multiple comparisons, the Holm (1979) was used.

Results

Sample Population

Subjects were 161 healthy cognitively normal control subjects with a mean age of 68 years, and 85 PD patients with a mean age of 67 years and a mean age-of-motor-onset symptom of 57 years (Table 1). All normal control subjects were reviewed by a consensus panel and designated as neurologically and cognitively normal based on cognitive testing and examination. All PD subjects fulfilled criteria for PD, based on history and examination⁴³. All but two PD cases had a mini-mental status examination (MMSE), with a mean MMSE score of 27.6 (+/− 3.5) for these cases. Only thirteen cases had a MMSE below 26.⁵⁰ The percentage of female subjects was significantly higher in control subjects compared to PD subjects (Table 1). Out of 86 PD patients, there were 7 PD patients with an age-at-onset of younger than 40 years (Table 1). These 7 patients were screened for Parkin (PARK2) mutations. One patient was found to be a compound heterozygote. This patient was excluded from the analysis leaving 85 PD patients for the analyses.

SNP Genotype Frequency

All 19 APP processing gene SNPs passed the HWE test after correcting for multiple comparisons. PSEN2 rs2802268 and APH1B rs2068143 frequency was significantly different between PD and controls. Significance did not remain after Holm multiple comparison correction (Table 2).

CSF Aβ₄₂ Levels by Disease Group and APOE ε4 Status

CSF Aβ₄₂ levels were not significantly different between PD and controls (Figure 1; Panel A). In the control sample, but not in PD, APOE ε4 carriers had significantly lower CSF Aβ₄₂ levels compared to non-carriers both with and without adjusting for gender and age (p = 0.006; p=0.012, respectively) (Figure 1; Panel B).

SNP Effect on CSF Aβ₄₂ Levels

The effect of SNP (collapsed genotypes) on CSF Aβ₄₂ levels was analyzed for each group (controls or PD) both with and without adjusting for the covariates; gender, age and APOE ε4-status. The results are presented graphically as beta coefficients and 95% confidence intervals without adjusting for covariates.

Within the PD group, Aβ₄₂ levels were significantly lower for minor allele carriers of the APP rs466468 (p = 0.014 adjusted, p = 0.015 unadjusted), the rs214484 (p = 0.032 adjusted, p = 0.036 unadjusted),and significantly higher for minor allele carriers of the APH1B SNP rs2068143 (p = 0.002 adjusted, p = 0.003 unadjusted). The rs2068143 adjusted value remained significant after Holm correction for multiple comparisons (p = 0.040). The rs466468 and rs2068143, but not rs214484, remained significant in main effect models containing all SNPs and covariates (p = 0.049, p = 0.011, p = 0.106, respectively).

Within the control group, Aβ₄₂ levels were significantly lower for minor allele carriers of the APP rs214484 (p = 0.002 adjusted, p = 0.007 unadjusted) and rs2040273 (p = 0.017 adjusted, p = 0.039 unadjusted), and for the PSEN1 rs362344 (p = 0.023 adjusted, p = 0.040 unadjusted). The APP SNP rs214484 adjusted value within controls remained significant after Holm correction (p = 0.040) and after including all the SNPs and covariates in the model (p = 0.044) (Figure 2). When a SNP-by-APOE ε4 status interaction term was included in the model it was not significant within controls and not significant within PD patients, while APOE ε4 and SNP remained significant, suggesting an additive effect.

CSF Aβ₄₂ Levels by APOE ε4 Status and SNP status

The nature of this additive effect for 3 SNPs positive in the control group was evaluated further using linear regression models described in the methods section. For the one SNP that remained significant after correcting for multiple comparisons there was a significant difference in CSF Aβ₄₂ levels between APOE ε4 carriers with the rs214484 G allele and APOE ε4 non-carriers without the rs214484 G allele with (p = 0.0002) and without adjusting for gender and age (p = 0.0001). There was a significant difference between APOE ε4 carriers with the rs2040273 G allele and APOE ε4 non-carriers without the rs2040273 G allele with (p = 0.002) and without adjusting for gender and age (p =0.001). There was a significant difference in CSF Aβ₄₂ levels between APOE ε4 carriers with the rs362344 T allele and APOE ε4 non-carriers without the rs362344 T allele with (p =0.001) and without adjusting for gender and age (p =0.001). There was also a significant difference in CSF Aβ₄₂ levels between APOE ε4 carriers with the rs362344 T allele and APOE ε4 non-carriers with the rs362344 T allele with (p = 0.027) and without adjusting for gender and age (p = 0.020) (data not shown).

The nature of this additive effect for 2 SNPs positive in the PD group was evaluated further using the linear regression models described in the methods section. There was a significant difference in CSF Aβ₄₂ levels between APOE ε4 carriers with the rs466448 G allele and APOE ε4 non-carriers without the rs466448 G allele with (p = 0.030) and without adjusting for gender and age (p = 0.031) (Figure 3; Panel A). There was also a significant difference in CSF Aβ₄₂ levels between APOE ε4 non-carriers with the rs466448 G allele and APOE ε4 non-carriers without the rs466448 G allele with (p = 0.035) and without adjusting for gender and age (p = 0.036) (Figure 3; Panel A). There was a significant difference in CSF Aβ₄₂ levels between APOE ε4 carriers with the rs2068143 A allele carriers and APOE ε4 non-carriers without the rs2068143 A allele with (p = 0.027) and without adjusting for gender and age (p = 0.025) (Figure 3; Panel A). There was also a significant difference in CSF Aβ₄₂ levels between APOE ε4 non-carriers with the rs2068143 A allele and APOE ε4 non-carriers without the rs2068143 A allele with (p = 0.021) and without adjusting for gender and age (p = 0.016) (Figure 3; Panel A).

Discussion

This exploratory investigation demonstrates that genetic variation within regulatory regions of APP processing pathway genes correlate with CSF Aβ₄₂ levels in PD and normal controls.

There was not a significant difference in CSF Aβ₄₂ levels between PD and controls (Figure 1, Panel A). Some, but not all, previous studies describe PD CSF Aβ₄₂ levels as lower than in controls^51–57. A more consistent finding is an association between lower CSF Aβ₄₂ levels and cognitive impairment in PD^38,56,57. Upon stratification by APOE ε4, control, but not PD, APOE ε4 carriers had significantly lower CSF Aβ₄₂ levels, both with and without age adjustment, compared to APOE ε4 non-carriers (Figure 1, Panel B). These results are supported by previous reports that show cognitively normal subject CSF Aβ₄₂ levels decrease with increasing age in APOE ε4 carriers³⁴. In addition, a lack of association between PD CSF Aβ₄₂ levels and APOE ε4 has been described^58,59.

One APP rs214484 was significantly associated with CSF Aβ₄₂ levels in cognitively normal controls after correcting for multiple comparisons and after adjusting for gender, age and APOE ε4 (Figure 2). APP rs2040273 and PSEN1 rs362344 nominally correlated with CSF Aβ₄₂ levels in controls and APP rs466448 and APH1B rs2068143, nominally correlated with CSF Aβ₄₂ levels in PD (Figure 2). To our knowledge this is the first investigation to find an association between SNPs in the APP processing pathway and CSF Aβ₄₂ levels in PD. However, in support of these findings, a recent report suggests that disturbed APP processing has an effect on the variable rate of motor and functional decline in PD⁶⁰. Coding mutations in the APP gene that influence the processing of APP and increase the production of Aβ cause a rare early onset form of AD and emphasize the importance of the APP processing pathway in AD cognitive decline^29,6121. Our group previously reported an association between soluble CSF biomarkers and APP processing genes in AD⁴². Specifically, a SNP allele located at the α-secretase ADAM10 locus was associated with CSF APPα levels in AD and was significantly different between AD and controls. However, there was no association between APP processing genes and CSF Aβ₄₂ levels in AD⁴². Taken together, a lack of association between APP processing genes and CSF Aβ₄₂ levels in AD in the previous study, but an association in PD in the present study, suggests that a different biological factor may play a role in the pathobiology of amyloid pathology in PD. Furthermore, the association between an APP 3’ region SNP in controls (rs214484) and an APP 5’ region SNP in PD (rs466448), generates further hypotheses regarding CSF Aβ₄₂ levels and APP genetic variants as therapeutic targets or markers of disease specific neurodegeneration.

To determine if these 5 APP pathway SNPs in combination with APOE ε4 have an additive effect on CSF Aβ₄₂ levels, all 5 SNPs were grouped by APOE ε4 in regression models. All 5 SNPs exhibited an additive effect where the carriers of both APOE ε4 and one of the APP processing SNP alleles had significantly lower CSF Aβ₄₂ levels compared to APOE ε4 non-carriers and APP processing SNP allele non-carriers. Interestingly, rs466448 and rs2068143 were the only SNPs that showed significantly lower PD CSF Aβ₄₂ levels within the APOE ε4 non-carriers, suggesting that these SNPs may influence PD CSF Aβ₄₂ levels, but not control levels, regardless of APOE ε4 status (Figure 3). Rs466448 is located in the 5’ region at position -1023 upstream of TSS of the APP promoter⁶² and is within an ENCODE described promoter associated histone mark (H3K4Me3)⁶³ suggesting that allele specific variation may play a mechanistic role in APP expression. In addition, rs214484 is located in the 3’ region of APP within an ENCODE described H3K27Ac mark often found near active regulatory elements⁶³ further suggesting that APP expression level may be influenced by genetic variation at these locations.

The rs2068143 is located in the 3’ region of APH1B in intron 4 in an ENCODE DNase I hypersensitivity cluster as well as an enhancer associated histone mark (H3K4Me1)⁶³ suggesting that allele specific variation may play a mechanistic role in APH1B expression. Inactivation of the Aph1B subunit of γ-secretase in a mouse model of AD has been described. In this model inactivation of Aph1B led to improvements in AD relevant phenotypic features⁶⁴. Given that the genetic variation of this gene has not been extensively investigated, and given that only 2 SNPs were analyzed in the present investigation, further analysis is needed to determine if rs2068143 is a surrogate marker for another genetic variant at the APH1B locus. Interestingly, there is a modest effect by all these described SNPs after multiple comparison adjustment. However, the information provided by this exploratory investigation may be important as it suggests that APOE ε4 non-carriers have a different APP pathway related association with CSF Aβ₄₂ compared to APOE ε4 carriers. This information suggests that it may be important, in future PD therapeutic strategies related to amyloid treatment of cognitive decline, to take into account APOE ε4 status.

An advantage of this investigation, in contrast to genome wide association studies, is that a small number of SNPs (n=19), within the context of a specific biological pathway were analyzed, and therefore increased the power to test the hypothesis that the APP processing pathway genes influence PD CSF Aβ₄₂ levels. A limitation of this investigation’s approach was that only a few APP processing genes were tested. Thus, many other contributors to APP processing may have been missed, such as genes related to the clearance or deposition of Aβ in the brain. In addition, a limited number of SNPs were used to capture putative regulatory genetic variation within and surrounding the genes of interest. Therefore, these results must be approached with caution since many important SNPs may have been missed. Another limitation of this investigation is that sAPPα and sAPPβ were not measured. Previously, our group described an association between CSF sAPPα and ADAM10 in AD⁴². Others have described CSF sAPPα, sAPPβ and various CSF Aβ₄₂ fragments in PD, Lewy body disease, and AD^60,65,66. Unfortunately, we did not have sAPPα and sAPPβ data available for this PD cohort.

Others have reported correlation between cognitive decline in PD and CSF Aβ₄₂⁶⁶. In the present investigation the relationship between cognition (MMSE) and SNP was evaluated (data not shown) and there was not a significant association for any of the SNPs or CSF Aβ₄₂ levels. However, we did not have cognitive data available for this cohort beyond MMSE, which has been shown to have low sensitivity to cognitive impairment in PD. Thus, additional cognitive testing would be necessary to adequately assess this issue. Given the results described here, future investigation of the relationship between CSF Aβ₄₂, APP pathway genes and cognition in PD may be warranted.

In conclusion, the main novel findings are that an APP SNP and an APH1B SNP are marginally associated with PD CSF Aβ₄₂ levels in APOE ε4 non-carriers. Further hypotheses generated by this exploratory investigation, include that decreased CSF Aβ₄₂ levels are in part driven by genetic variation in APP processing genes. This information may have important implications for therapeutic strategies related to amyloid treatment of cognitive decline that take into account APOE ε4 status. Investigation into the relationship between these genes and clinical characteristics of PD, such as cognitive impairment, compared to other neurodegenerative diseases, are warranted. Characterization of the functional influence of APP and APH1B genetic variation on expression may lead to a better understanding of PD, help identify novel PD biomarkers and help identify PD therapeutic targets.