The pathogenicity of PSEN2 variants is tied to Aβ production and homology to PSEN1

Abstract

INTRODUCTION

Though recognized as a potential cause of autosomal dominant Alzheimer's disease, the pathogenicity of many PSEN2 variants remains uncertain. We compared amyloid beta (Aβ) production across all missense PSEN2 variants in the AlzForum database and, when possible, to corresponding PSEN1 variants.

METHODS

We expressed 74 PSEN2 variants, 21 of which had known, homologous PSEN1 pathogenic variants with the same amino acid substitution, in HEK293 cells lacking presenilin 1/2. Aβ production was compared to age at symptom onset (AAO) and between PSEN1/2 homologs.

RESULTS

Aβ42/40 and Aβ37/42 ratios correlated with AAO across all PSEN2 variants, strongly driven by the subset of PSEN2 variants with PSEN1 homologs. Aβ production across PSEN1/2 homologs was highly correlated. PSEN2 AAO correlated with AAO in PSEN1 homologs but was an average of 18.3 years later.

DISCUSSION

The existence of a PSEN1 homolog and patterns of Aβ production are important considerations in assessing the pathogenicity of previously reported and new PSEN2 variants.

Highlights

• There were associations between the patterns of amyloid beta (Aβ) production across presenilin 2 (PSEN2) variants and age at symptom onset (AAO).

• PSEN2 variants for which there is a known, corresponding variant in presenilin 1 (PSEN1) are more likely to have abnormal Aβ production patterns that strongly correlate with AAO, compared to those that lack a known PSEN1 counterpart (“non‐homologous PSEN2 variants”).

• Most PSEN2 variants lacking PSEN1 counterparts had Aβ42/40 ratios close to those of wild‐type PSN2, arguing against their pathogenicity.

• Homologous PSEN1 and PSEN2 variants had correlated Aβ42/40 and Aβ37/42 ratios, indicating that the corresponding amino acid substitution in each presenilin may have largely similar biochemical effects on γ‐secretase processivity.

1. INTRODUCTION

More than 250 pathogenic variants leading to autosomal dominant Alzheimer's disease (ADAD) have been identified in PSEN1, PSEN2, and amyloid precursor protein (APP), ¹ , ² , ³ with an additional 200+ variants reported as likely non‐pathogenic or of uncertain significance. Among pathogenic variants, the age at symptom onset (AAO) within families and between variants varies widely, spanning 50+ years. ⁴ , ⁵ Most pathogenic variants described are in PSEN1, accounting for 70% to 80% of ADAD cases. ¹ Presenilin 1 (PSN1), encoded by PSEN1, forms the catalytic core of the γ‐secretase complex, directly producing longer, aggregation‐prone amyloid beta (Aβ) peptides relative to shorter, non‐aggregating peptides. ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ Fewer ADAD‐causing pathogenic variants have been identified in PSEN2, ² similar to PSEN1, its encoded protein, presenilin 2 (PSN2), also forms the catalytic core of a γ‐secretase complex. ¹⁵ , ¹⁶ , ¹⁷ Despite many similarities, PSN2 and PSN1 are thought to have subtle variations regarding sub‐cellular localization ¹⁸ , ¹⁹ and overall expression levels. ²⁰ , ²¹ , ²² , ²³ Intriguingly, the average AAO in individuals bearing PSEN2 variants is substantially older than that seen in carriers of PSEN1 variants, ²⁴ , ²⁵ though this striking dissimilarity is not well understood. Furthermore, assessing the pathogenicity and penetrance of PSEN2 variants can be challenging, as AAO for PSEN2 variants can overlap with sporadic AD, and detailed pedigrees are often not available. For these reasons, the majority of PSEN2 variants reported in the AlzForum database are considered unclassified or of uncertain significance. This lack of clarity underscores the need to better characterize PSEN2 variants individually, especially changes that substantially alter the γ‐secretase. Recent studies have compared the Aβ production profiles between the wild‐type PSN1 and PSN2, ²⁶ and compared conserved mutations across PSEN1 and PSEN2 on a limited number of samples. ²⁴ , ²⁷ , ²⁸ However, there has yet to be a comprehensive analysis of the full set of conserved mutations.

Recent work from our group and others has shown that the clinical and biomarker changes seen across the course of ADAD in PSEN1 variant carriers may be due in large part to the extent to which individual variants impact the production of aggregation‐prone Aβ species (Aβ42 and 43) compared to shorter, less aggregation prone species (Aβ37 and 38). ²⁴ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ Indeed, quantitative analysis of APP processing by individual PSEN1 variants reveals strong associations between the ratio of Aβ42/Aβ40 and/or Aβ37/42 production with the AAO and rate of cognitive decline across > 160 examined PSEN1 variants. ³² Beyond AAO, we have recently shown strong associations between the ratio of short‐to‐long Aβ species $(\frac{\mathrm{short}\mathrm{A}\beta 37+38+40}{\mathrm{long}\mathrm{A}\beta 42+43})$ and a wide set of in vivo imaging and biofluid markers of neurodegeneration, tau, and Aβ pathology in PSEN1 pathogenic variant carriers. ²⁹ , ³⁵ Broadly, the findings across PSEN1 variants strongly suggest that variant‐level differences in γ‐secretase dysfunction are critical determinants of the clinical and biomarker course of ADAD.

The associations among γ‐secretase enzymatic abnormality, Aβ production, and pathogenicity of ADAD have not been fully established for PSEN2 variants. Besides informing clinical care for people with suspected ADAD‐causing variants and the conduct of ongoing clinical trials in ADAD, understanding the link between altered γ‐secretase function and the pathogenicity of PSEN2 variants is particularly critical as many PSEN2 variants are reported as questionably pathogenic, unclassified, or of uncertain significance. In addition to AD, variants in PSEN2 have also been linked to other neurological disorders not classically associated with PSEN1 variants (as reviewed by Cai et al., ¹⁶ ) such as dementia with Lewy bodies, ³⁶ frontotemporal dementia (FTD), ³⁷ and Parkinson's disease dementia (PDD). ³⁸ , ³⁹ Here, the systematic evaluation of PSEN2 variants may help to clarify the pathological significance of individual variants in clinical and research settings, particularly when limited family history is available. In this context, we provide a comprehensive examination of γ‐secretase function and Aβ production across all non‐synonymous, missense PSEN2 variants currently listed in the AlzForum database, elucidating the differential pathogenicity of similar variants in PSEN1 compared to PSEN2 and to inform clinical evaluation for people bearing PSEN2 variants of uncertain pathogenicity.

RESEARCH IN CONTEXT

1. Systematic review: While variants in presenilin 2 (PSEN2) can lead to autosomal dominant Alzheimer's disease (ADAD), the pathogenicity of many variants remains uncertain. At a group level, the age at symptom onset (AAO) for PSEN2 pathogenic variants is generally later than presenilin 1 (PSEN1), though reasons for this dissimilarity are not well understood, and studies directly comparing conserved mutations in PSEN1 and PSEN2 are lacking.

2. Interpretation: Amyloid beta (Aβ)42/40 and 37/42 ratios were associated with AAO across PSEN2 variants, strongly driven by PSEN2 variants with a known PSEN1 pathogenic variant counterpart containing the same amino acid substitution (“homologs”). Aβ ratios from PSEN1/2 homologs were correlated, suggesting a similar mechanism of γ‐secretase dysfunction.

3. Future directions: These results support the consideration of the existence of a PSEN1 homologous variant when assessing the pathogenicity of PSEN2 variants, and provide a comprehensive assessment of Aβ production from PSEN2 missense variants that can inform inclusion criteria for ADAD clinical trials.

2. METHODS

2.1. Generation of human PSEN1 and PSEN2 expression vectors

The coding sequence of the wild‐type human PSEN1 and PSEN2 were subcloned into the pcDNA3.1 vector, which was then used to generate vectors expressing specific mutations. Polymerase chain reaction (PCR) amplified the template into two DNA fragments with an overlapping sequence containing the target locus to introduce mutations. Overlap PCR was done to generate the whole open‐reading frame containing the mutations. PCR products were subcloned into the same pcDNA3.1 vector for expression under a human cytomegalovirus promoter. Vectors were sequenced from 5′ and 3′ ends to confirm successful mutagenesis.

Using the AlzForum database, ² we identified all missense PSEN2 variants listed (a total of 74 at the start of this study in September 2023), including all of those listed as pathogenic, benign, not classified, associated with FTD or PDD, or of uncertain significance (VUS; see Table 1 for a list of variants). Due to the model system used, PSEN2 variants in which the wild‐type amino acid was not altered (i.e., synonymous mutations) or in which a frameshift mutation was present were not included in the present study. We identified 21 PSEN2 variants for which the same amino acid substitution was present in a previously reported, pathogenic variant in PSEN1. For the purposes of discussion, we refer to similar substitutions at consonant residues of PSN1 and PSN2 as homologous mutations in PSEN1 and PSEN2. Conversely, PSEN2 variants without a known PSEN1 counterpart will be referred to as non‐homologous.

TABLE 1.

Description of PSEN1 and PSEN2 variants characterized.

Variant		AAO		Aβ 42/40		Aβ 37/42		ACMG pathogenicity classification
PSEN2	PSEN1	PSEN2	PSEN1	PSEN2	PSEN1	PSEN2	PSEN1	PSEN2	PSEN1
WT	WT			0.15	0.15	0.44	0.41	–	–
A85V	A79V	65.5	53	0.17	0.18	0.29	0.26	AD: Not Classified, DLB: Not Classified	AD: Likely Pathogenic
T122R	T116R	57	35	0.25	0.51	0.23	0.10	Atypical Dementia: Not Classified	AD: Not Classified
P123L	P117L	57	30.25	0.18	0.34	0.28	0.16	AD: Not Classified	AD: Pathogenic
E126K	E120K	54	39	0.28	0.38	0.21	0.12	AD: Not Classified	AD: Pathogenic
N141D	N135D	59	35	0.20	0.26	0.29	0.19	AD: Not Classified	AD: Pathogenic
N141S	N135S	52	33	0.29	0.43	0.19	0.10	AD: Not Classified	AD: Pathogenic
N141Y	N135Y	43	32	0.32	0.40	0.17	0.13	AD: Likely Pathogenic	AD: Not Classified
V148I	V142I	76	53	0.18	0.15	0.32	0.40	AD: Benign	AD: Likely Pathogenic
I149T	I143T	60.5	35	0.25	0.32	0.26	0.21	AD: Not Classified	AD: Pathogenic
F183S	F177S	46	30	0.30	0.24	0.23	0.22	AD: Not Classified	AD: Pathogenic
G212V	G206V	62.5	30	0.24	0.24	0.22	0.20	AD: Not Classified	AD: Pathogenic
H220Y	H214Y	65	41	0.19	0.15	0.37	0.21	AD: Not Classified	AD: Pathogenic
L225P	L219P	42	49.5	0.19	0.15	0.32	0.22	AD: Not Classified	AD: Likely Pathogenic
I235F	I229F	56.9	33	0.27	0.26	0.24	0.22	AD: Not Classified	AD: Not Classified
A237V	A231V	87	65	0.19	0.15	0.32	0.41	AD: Uncertain Significance	AD: Not Classified
L238P	L232P	54	45	0.24	0.18	0.27	0.28	AD: Not Classified	AD: Not Classified
L238F	L232F	49	59	0.23	0.17	0.31	0.31	AD: Uncertain Significance	AD: Not Classified
M239T	M233T	52	38	0.27	0.30	0.26	0.22	AD: Uncertain Significance	AD: Pathogenic
M239V	M233V	54.3	31	0.35	0.38	0.20	0.13	AD: Pathogenic	AD: Pathogenic
M239I	M233I	49.6	28	0.28	0.30	0.23	0.17	AD: Pathogenic	AD: Not Classified
F369S	F386S	51	48	0.27	0.20	0.21	0.19	AD: Not Classified	AD: Pathogenic
T18M		62		0.16		0.40		PD: Not Classified
R29H		86		0.15		0.40		AD: Uncertain Significance
G34S		64		0.15		0.40		AD: Benign
R62C		67.3		0.15		0.42		AD: Benign
R62H		61.6		0.15		0.42		AD: Benign, FTD: Not Classified
C65Y		52		0.16		0.37		svPPA: Not Classified
P69A		75		0.16		0.42		AD: Benign
R71W		62.4		0.17		0.36		AD: Benign
L79P		52		0.16		0.36		AD: Not Classified
K82R		53		0.14		0.37		AD: Not Classified
V101M		–		0.16		0.40		AD: Not Classified
T122P		51.8		0.25		0.22		AD: Likely Pathogenic
S130L		65.2		0.15		0.40		AD: Uncertain Significance
L135R		42		0.16		0.35		FTD: Not Classified
V139M		76		0.14		0.45		AD: Benign
N141I		55.7		0.31		0.20		AD: Pathogenic
L143H		–		0.15		0.36		AD: Not Classified
I146T		64.5		0.15		0.37		AD: Uncertain Significance
S147N		60		0.14		0.38		AD: Not Classified
V150M		64		0.14		0.40		AD: Not Classified
T153S		60.5		0.15		0.41		AD: Not Classified
K161R		65		0.14		0.45		AD: Not Classified
R163C		73		0.17		0.28		AD: Uncertain Significance
R163H		39		0.14		0.42		AD: Benign
H169N		63		0.14		0.44		AD: Uncertain Significance, FTD: Not Classified
M174I		53		0.15		0.39		AD: Not Classified
M174V		52.2		0.14		0.39		AD: Benign
S175C		61.5		0.20		0.29		AD: Not Classified
S175F		53		0.30		0.20		AD: Not Classified
Y195C		42		0.15		0.39		FTD: Not Classified
V214L		56.7		0.14		0.41		AD: Benign
Q228L		60		0.18		0.39		AD: Not Classified
Y231C		52		0.19		0.40		FTD: Not Classified
A252T		–		0.17		0.42		AD: Benign
A258T		–		0.16		0.46		AD: Benign
A258V		–		0.16		0.44		AD: Not Classified
R284G		57.5		0.22		0.25		AD: Likely Pathogenic
M298T		57.5		0.16		0.39		AD: Likely Pathogenic
T301M		60		0.15		0.38		AD: Benign
P334A		–		0.17		0.34		AD: Benign
P334R		80		0.17		0.39		AD: Benign
P348L		65		0.16		0.40		AD: Not Classified
I368F		60		0.32		0.18		AD: Not Classified
A377V		–		0.16		0.37		AD: Benign
A379D		55		0.18		0.36		AD: Not Classified
T388M		45		0.16		0.44		bvFTD: Not Classified
C391R		74		0.17		0.44		MCI: Not Classified
V393M		50		0.17		0.36		AD: Benign
A415S		59		0.16		0.38		AD: Not Classified
T421M		55		0.16		0.35		AD: Benign
T430M		54.5		0.15		0.46		AD: Not Classified
P436L		52		0.17		0.42		Dementia: Not Classified
D439A		52		0.15		0.41		AD: Likely Benign

Abbreviations: AAO, age at symptom onset; Aβ, amyloid beta; ACMG, American College of Medical Genetics and Genomics; AD, Alzheimer's disease; bvFTD, behavioral variant frontotemporal dementia; FTD, frontotemporal dementia; MCI, mild cognitive impairment; svPPA, semantic variant primary progressive aphasia; VUS, variant of uncertain significance; WT, wild type.

2.2. Pathogenicity classification and AAO of variants

Pathogenicity classification of PSEN2 and PSEN1 variants were determined using American College of Medical Genetics and Genomics–Association for Medical Pathology (ACMG‐AMP) guidelines and further sub‐classified into three groups: Pathogenic AD (AD: pathogenic or AD: likely pathogenic classification), Benign/VUS AD, and Not Classified. AAO was determined from literature search and was available for 67 of 74 PSEN2 variants and 21 of 21 of PSEN1 variants.

2.3. Tissue culture and transfection of adherent cells

Adherent human embryonic kidney 293 (HEK) cells were cultured in complete growth medium: Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 10 units/ml penicillin, and 10 µg/mL streptomycin. Adherent HEK cells were seeded in a 48‐well plate at a density of 200,000 cells per well for transfection. Transfection was carried out with PEI Max. Cells were incubated for 48 hours before medium were harvested for enzyme‐linked immunosorbent assay (ELISA).

2.4. Generation of PSEN1/2 dKO cell line with stable human APP expression

The HEK293 PSEN1/2 dKO cells ³² were transduced with a lentivirus expressing human APP‐695 with a C‐terminal green fluorescent protein (GFP) tag (OriGene). A single clone was isolated with serial dilution. Expression of APP‐695‐GFP was confirmed with ELISA.

2.5. Aβ immunoassay (Meso Scale Discovery)

Conditioned media from cultured cells was diluted with 1% bovine serum albumin (BSA) in the wash buffer (tris buffered saline [TBS] supplemented with 0.05% Tween). For Aβ 37/40/42 assays, each well of an uncoated 96‐well multi‐array plate (Meso Scale Discovery, #L15XA‐3) was coated with 30 µL of a phosphate buffered saline solution containing 3 µg/mL of m266 capture antibody to the mid‐region of soluble Aβ (original gift of P. Seubert, Elan, plc) and incubated at room temperature overnight. We prepared detection antibody solutions containing biotinylated monoclonal antibodies specifically recognizing the C‐terminal residue of each Aβ isoform, as well as 100 ng/mL Streptavidin Sulfo‐TAG (Meso Scale Discovery, #R32AD‐5), and 1% BSA diluted in the wash buffer. After overnight incubation with the capture antibody, 25 µL/well of the sample, followed by 25 µL/ well of detection antibody solution was incubated for 2 hours at room temperature with shaking at > 300 rpm, with washing of wells with wash buffer (TBS supplemented with 0.05% Tween) between incubations. The plate was read and analyzed according to the Meso Scale Discovery manufacturer's manual. Aβ peptide ratios of 37/42 and 42/40 were chosen a priori for analyses. 

FIGURE 1.

Schematic depicting the human PSN2 amino acid sequence (top row) and corresponding human PSN1 amino acid sequence (bottom). Homologous amino acid residues are bolded and underlined. Location of amino acids with known homologous variants in both PSN1 and PSN2 are bolded in red with the amino acid substitutions shown above. PSN, presenilin.

2.6. Statistical analysis

All statistical analyses were performed using R studio (version 2021.09.0). Pearson correlations were performed to examine (1) the association between AAO and Aβ42/40 or Aβ37/42 in PSEN2 variants, within the full PSEN2 sample with AAO available (N = 67) and separately in PSEN2 homologous variants (N = 21) and PSEN2 non‐homologous variants (N = 46); (2) the relationship between AAO for PSEN2 variants and the AAO of their PSEN1 homologs (N = 21 pairs); and (3) the relationship between Aβ42/40 or Aβ37/42 levels produced by homologous PSEN1 and PSEN2 variants (N = 21). t tests were performed to examine (1) the difference in AAO between PSEN1 (N = 21) and PSEN2 variants (N = 67); and (2) within PSEN2 variants, the difference in AAO among variants classified as AD: pathogenic or AD: likely pathogenic compared (N = 7) to variants with other pathogenicity classifications (AD: benign/AD: VUS and not characterized; N = 60). A linear regression model with the interaction of AAO and PSEN1/2 grouping interaction as the term of interest was performed to examine how the association between AAO and cell‐based Aβ42/40 or Aβ37/42 levels differ between homologous variants in PSEN1 versus PSEN2.

3. RESULTS

3.1. Pathogenicity classifications of PSEN2 variants

There was a wide range of ACMG‐AMP classifications across the 74 PSEN2 variants (Table 1 and Table S1 in supporting information). All homologous PSEN1 variants characterized were classified as either AD: pathogenic/AD: likely pathogenic (67%) or AD: not classified (33%). Though the large majority of PSEN2 variants were not definitively classified, we observed that PSEN2 variants with an identified homologous PSEN1 variant, as illustrated in Figure 1, were more likely to be classified as pathogenic (i.e., listed as “AD: pathogenic” or “AD: likely pathogenic” variants in the AlzForum database) compared to those with no known PSEN1 homologous variant (14% and 8%; Figure S1 in supporting information). Additionally, PSEN2 variants classified as AD: pathogenic had earlier AAO compared to other variants (t [12] = 2.68 and P = 0.020).

3.2. Aβ production patterns are associated with AAO in PSEN2 variants

We observed an association between AAO and the cell‐based production ratios Aβ42/40 (r [65] = –0.272 and P = 0.026; Figure 2A) and Aβ37/42 (r [65] = 0.251 and P = 0.040; Figure 2B) across  PSEN2 variants. However, examining these data more closely, the associations between Aβ ratios and AAO were completely driven by PSEN2 variants with a known PSEN1 homolog (“homologous variants”; Aβ42/40: r [19] = –0.530 and P = 0.014; Aβ37/42: r (19) = 0.490 and P = 0.024; Figure 2C,D). No such relationship was observed in PSEN2 variants that lacked known PSEN1 homologs (Aβ42/40: r [44] = –0.105 and P = 0.487; Aβ37/42: r [44] = 0.147 and P = 0.328; Figure 2E, F). Further, we observed that a majority of PSEN2 variants lacking PSEN1 homologs had Aβ42/40 and Aβ37/42 ratios that were similar to that observed from wild‐type PSEN2 (e.g., Aβ42/40 close to 0.15), arguing against the pathogenicity of these variants.

FIGURE 2.

Correlation between Aβ42/40 and Aβ 37/42 cellular production ratios and AAO across all PSEN2 variants (A, B), across only PSEN2 variants with a known PSEN1 homolog (C, D; orange), and across only PSEN2 variants that lack a PSEN1 homolog (Non‐Homologs; E, F; blue). Pathogenicity classifications, presented in Table 1, for each of the PSEN2 variants are depicted as triangles (Pathogenic AD), squares (Benign/VUS AD), or circles (Not Characterized). Wild‐type PSEN2 ratios for cell‐based Aβ42/40 and Aβ37/42 are depicted as black dashed horizontal lines. The shaded area each of the linear fit lines represents the 95% confidence interval. AAO, age at symptom onset; Aβ, amyloid beta; AD, Alzheimer's disease; VUS, variant of uncertain significance.

FIGURE 3.

A, Relationship between AAO in PSEN1 (red), in PSEN2 variants with a known PSEN1 homolog (blue), and in PSEN2 variants lacking a PSEN1 homolog (orange). The boxes map to the median, 25th and 75th percentiles, and the whiskers extend to 1.5 × interquartile range (IQR). The raincloud plots illustrate kernel probability density (i.e., the width of the shaded area represents the proportion of the data located there). Lines connect PSEN2 variants with their PSEN1 homologs to show relative shifts in AAO. Pathogenicity classifications, presented in Table 1, for each of the PSEN variants are depicted as triangles (Pathogenic AD), squares (Benign/VUS AD), or circles (Not Classified). B, Relationship between AAO of PSEN2 variants and AAO of corresponding homologous PSEN1 variants. Correlation between cell‐based Aβ42/40 (C) and Aβ37/42 (D) ratios and AAO in PSEN2 (orange) and homologous PSEN1 (red) variants. Relationship between AAO of PSEN2 homologous variants and cell‐based Aβ42/40 (E) and Aβ37/42 (F) ratio. In panels comparing Aβ ratios of PSEN1/2 homologs (E and F), the PSEN2 pathogenicity classification is depicted. The shaded area around the linear fit line represents the 95% confidence interval. AAO, age at symptom onset; Aβ, amyloid beta; AD, Alzheimer's disease; VUS, variant of uncertain significance.

Notably, however, a cluster of PSEN2 variants lacking a PSEN1 counterpart demonstrated Aβ42/40 and Aβ37/42 ratios that were similar to those previously observed in PSEN1 variants that were clearly pathogenic. This suggests that this small set of non‐homologous PSEN2 variants (T122P, S175C, S175F, N141I, and I368F) may be pathogenic despite the absence to date of a known, corresponding pathogenic mutation in PSEN1.

3.3. Comparing homologous PSEN1 and PSEN2 variants

We next compared AAO between PSEN1 and PSEN2 variants. We observed an average AAO of 58.4 years old, across all known PSEN2 variant carriers, whereas PSEN1 variant carriers bearing homologous mutations had an earlier average AAO of 40.1 years (difference of 18.3 years; t [32] = –6.9835 and P = 6.841e‐08; Figure 3A). There was a strong positive correlation between the AAO of PSEN2 variants and the AAO of their PSEN1 homologs (r [19] = 0.473, P = 0.030; Figure 3B), with a median shift of 21.6 years (range: –10, 32.5) later AAO in PSEN2 variants compared to homologous variants in PSEN1. Consistent with this finding, for a given Aβ42/40 or Aβ37/42 ratio, we observed that PSEN2 variants have a later AAO compared to homologous PSEN1 variant (Aβ42/40: B [standard error (SE)] = 0.004 [0.002] and P = 0.045; Aβ37/42: B [SE] = –0.004 [0.001] and P = 0.018; Figure 3C, D).

3.4. Relationship between secreted Aβ levels across PSEN1/2 homolog pairs

Last, we compared how homologous changes in PSEN1 and PSEN2 impacted the production of Aβ species in our cell culture system. We observed that Aβ42/40 and Aβ37/42 production ratios were correlated between PSEN1 variants and their PSEN2 homologs (Aβ42/40: r = 0.580, P = 0.006 and Aβ37/42: r = 0.678, P = 0.0007; Figure 3E, F), suggesting that PSEN2 variants had similar biochemical effects on γ‐secretase conformation and thus proteolytic function as did the homologous mutation in PSEN1.

4. DISCUSSION

Though it is widely recognized that numerous pathogenic variants leading to ADAD exist in PSEN2, ² , ¹⁶ , ²⁴ the pathogenicity of many PSEN2 variants remains uncertain, ⁴⁰ as does the extent to which corresponding changes between the highly similar PSN1 and PSN2 proteins may lead to similar AAO and patterns of Aβ production. Here we examined a comprehensive set of PSEN2 pathogenic variants, including variants with uncertain pathogenicity, to elucidate these issues. Broadly, we observed that, similar to recent work characterizing PSEN1 variants, ²⁴ , ²⁹ , ³² , ⁴¹ there were significant associations between the patterns of Aβ production across PSEN2 variants and AAO. However, follow‐up analyses clearly suggested that PSEN2 variants for which there is a known, corresponding variant in PSEN1 are more likely to have abnormal Aβ production patterns that strongly correlate with AAO. Aside from a small subset of PSEN2 variants without a known PSEN1 homologous variant that had Aβ42/40 ratios likely within a pathogenic range, most PSEN2 variants lacking PSEN1 counterparts had Aβ42/40 ratios close to those of wild‐type PSEN2, arguing against their pathogenicity. Together, these results support consideration of whether a PSEN1 homolog variant is known when assessing the pathogenicity of a PSEN2 variant of uncertain significance, though clinical information and available family history remain critical to definitively classify individual variants. Interpreted in another way, these results also suggest that there may be qualitative biochemical differences between PSEN2 variants that have a known PSEN1 counterpart (“homologous PSEN2 variants”) and those that lack a known PSEN1 counterpart (“non‐homologous PSEN2 variants”).

Intriguingly, we observed that homologous PSEN1 and PSEN2 variants had strongly correlated Aβ42/40 and Aβ37/42 ratios, indicating that the corresponding amino acid substitution in each presenilin may have largely similar biochemical effects on γ‐secretase processivity. However, comparing AAOs across PSEN2 and PSEN1 homologs, we observed a shift of 18.3 years later for PSEN2 variants compared to their PSEN1 counterparts. While this latter observation is consistent with what has previously been described in the literature, ²⁴ , ²⁵ the contrast between similar cell‐based patterns of Aβ production but substantially shifted AAO between PSEN1 and PSEN2 homologs underscores the need for further study regarding the differences between PSN1 and PSN2 under normal physiological conditions. Several possibilities exist to explain these differences, including subtly different conformations (folding) of PSN1 and PSN2 in the γ‐secretase complex, differential efficiency of incorporation of PSN1 versus PSN2 into the pool of active γ‐secretase complexes (as previously described in Liu et al., ⁴² ) less contribution of PSN2 to total γ‐secretase activity, ⁴³ differential trafficking of PSN1 and PSN2 within the cell, and potential differences in the half‐life of the PSN1 compared to the PSN2 protein. These and other biochemical differences may explain the observation that a PS2‐containing γ‐secretase complex is less proteolytically active than a PSN1‐containing γ‐secretase complex as an intramembrane protease and therefore less commonly the site of AD‐causing mutations. Notably, there are several variants (e.g., PSEN2 L225P), with younger AAO than their PSEN1 homologs, ²⁹ suggesting there are additional, as yet unknown, factors that contribute to the pathogenicity of these variants.

Because it relies on over‐expression, the cell model system used here is not well suited to assess the differential transfection efficiency of PSEN1 and PSEN2, a limitation of the present study. Most PSEN1 and PSEN2 mutation carriers are heterozygous, meaning our system may not accurately reflect the correct allele dosage, potentially leading to an overestimation of the effects of PSEN2 mutations. However, this experimental setup was chosen intentionally to emphasize the differences between PSEN2 variants and their homologous PSEN1 counterparts. Another important limitation is the relative lack of in vivo clinical and AD biomarker data available for PSEN2 carriers as a group. The relative infrequency of PSEN2 variants and the absence of a suitably large cohort of PSEN2 variants within the Dominantly Inherited Alzheimer's Network (DIAN) ⁴⁴ and similar studies makes it difficult to definitively assess the potential pathogenicity of most PSEN2 variants with respect to AD cognitive and pathological changes. This contrasts with prior work from our group characterizing a large number of PSEN1 pathogenic variants, in which we were able to correlate a diverse set of in vivo clinical, cognitive, biofluid, and neuroimaging data with cell‐based assessments of γ‐secretase function. ²⁹

Despite these limitations, the findings here suggest that the relatively simple cellular model system used is useful for characterizing newly discovered PSEN2 variants and making decisions about the inclusion of particular variants in ADAD clinical trials. Additionally, the characterization of a comprehensive set of PSEN2 variants in the same cellular system is valuable for comparing AAO and pathologic changes among PSEN2 variants. Most importantly, the present results strongly support the consideration of the existence of a PSEN1 homologous variant when assessing the pathogenicity and estimated AAO for a known or new PSEN2 variant.

CONFLICT OF INTEREST STATEMENT

D.J.S. is a director and consultant of Prothena Biosciences. L.L. is a consultant of KorroBio, Inc. and Dynamicure Bio, LLC. J.P.C has served on medical advisory boards for ExpertConnect. All other authors have nothing to disclose. Author disclosures are available in the supporting information.

CONSENT STATEMENT

No human subjects were studied in this work; therefore the statement is not necessary.

Supporting information

Supporting information

Supporting information

Supporting information

Liu L, Schultz SA, Saba A, et al. The pathogenicity of PSEN2 variants is tied to Aβ production and homology to PSEN1 . Alzheimer's Dement. 2024;20:8867–8877. 10.1002/alz.14339