Longitudinal analysis of a dominantly inherited Alzheimer disease mutation carrier protected from dementia

Abstract

We conducted an in-depth longitudinal study on an individual carrying the presenilin 2 p.Asn141Ile mutation, traditionally associated with dominantly inherited Alzheimer’s disease (AD), who has remarkably remained asymptomatic past the expected age of clinical onset. This study combines genetic, neuroimaging and biomarker analyses to explore the underpinnings of this resilience. Unlike typical progression in dominantly inherited AD, tau pathology in this case was confined to the occipital region without evidence of spread, potentially explaining the preservation of cognitive functions. Genetic analysis revealed several variants that, although not previously associated with protection against AD, suggest new avenues for understanding disease resistance. Notably, environmental factors such as significant heat exposure and a unique proteomic profile rich in heat shock proteins might indicate adaptive mechanisms contributing to the observed phenotype. This case underscores the complexity of Alzheimer’s pathology and suggests that blocking tau deposition could be a promising target for therapeutic intervention. The study highlights the need for further research to identify and validate the mechanisms that could inhibit or localize tau pathology as a strategy to mitigate or delay the onset of Alzheimer’s dementia.

Dominantly inherited Alzheimer’s disease (DIAD) is caused by highly penetrant mutations in either Amyloid Precursor Protein (APP), Presenilin 1 (PSEN1) or Presenilin 2(PSEN2) that alter APP and amyloid-β processing1 . DIAD also has a highly predictable symptomatic age at onset (AAO) based on the specific mutation or family AAO history^2,3. Using the expected AAO and studying families with DIAD has been instrumental in identifying fundamental mechanisms of AD pathophysiology^1,4–6 that are now guiding the development of treatments advancing into clinical trials.

The Dominantly Inherited Alzheimer Network (DIAN) was established to create a global registry of DIAD family members and to determine clinical, cognitive and imaging/fluid biomarker changes, as well as identify genetic modifiers of AD⁷. The DIAN has enrolled >530 participants since inception in 2008 and has followed these individuals longitudinally to further determine onset of AD changes and longitudinal rates of change and to compare to sporadic AD. DIAN studies have meticulously analyzed the AAO and biomarker rates of change by gene and specific mutations, revealing a correlation of approximately 0.7 between mutation and the timing of symptom emergence. Despite this correlation, there is an observed variability in the AAO, which indicates a typical variation in the onset within these populations. However, it is extremely rare that an individual carrying a pathogenic mutation does not develop dementia within the expected age range^8–10. Previously, two individuals carrying the Colombian PSEN1 p.Glu280Ala mutation have been reported to escape dementia for decades after the expected AAO^8,10. Whole-exome sequencing studies in these two cases suggest that a rare homozygous APOE variant (APOE3-Christchurch (p.Arg136Ser)) and a rare RELN variant (p.His3447Arg) conferred strong protection against the development of DIAD in these two PSEN1 p.Glu280Ala carriers^8,10,11. These exceptional individuals who escape clinical onset of dementia (whom we term ‘exceptional resilience mutation carriers’) suggest there are genetic or environmental protective factors at play that can mitigate the strong and predictable effects of DIAD genetic mutations. These remarkable resilient individuals, who defy the typical genetic fate despite carrying pathogenic mutations, offer invaluable insights into natural protective mechanisms and resilience, potentially revealing new strategies for prevention and therapy in a broader AD population.

Here, we report the longitudinal clinical in vivo neuroimaging and biofluid biomarkers follow-up of an exceptional resilience mutation carrier who inherited a known pathogenic mutation in the PSEN2 gene (p.Asn141Ile)^12–16 but is 18 years past the expected AAO. Relative to previous reports, our study provides longitudinal data over a decade, exploring genetic underpinnings, biomarker and omics changes that allows us to uniquely examine this individual’s response to Alzheimer’s pathophysiology relative to other mutation carriers.

Results

Case report

Family history.

The participant is part of a large extended family carrying the pathogenic PSEN2 p.Asn141Ile variant. His mother, as well as 11 of her 13 siblings, developed AD at or close to the age of 50 (ref. 17). The family pedigree (Fig. 1) shows the expected dominantly inherited pattern with an average family symptomatic AAO of 49.3 years (range 44–53). Using the family-specific AAO, the individual is 22 years past the expected onset. Considering the latest known AAO within this kindred (58 years), the individual is 15 years beyond the expected AAO.

Fig. 1 |. Schematic representation of the pedigree for the DIAN family carriers of the pathogenic mutation PSEN2 p.Asn141Ile.

The pedigree illustrates the inheritance pattern of AD within the studied family. Subfamilies are indicated by a number (1–14). Within each generation, individuals are numbered left to right (1, 2, 3 and so on). Affected individuals are marked with filled shapes (for example, solid black), and unaffected individuals are depicted with open shapes (for example, outlined in black). Carriers of the PSEN2 mutation are indicated with a plus sign, and non-carriers are indicated with a minus sign. Participants highlighted in blue have biological data available. Individuals included for the dominant and recessive model approaches are indicated with a green rectangle, and participants included in the de novo model are indicated with a yellow rectangle. The exceptional resilience mutation carrier is represented as a green-filled shape (10.7). AAO and age at last assessment (ALA) are provided when available. Gray diamonds indicate possible presymptomatic family members (mutation carriers but younger than expected age of onset). Fam, family.

The p.Asn141Ile variant^3,12,13 has a mean mutation symptomatic AAO of 53.7 (range 39–58) years, and its origin can be traced to people originally living in two small adjacent Volga German villages^12,13,17. The mean mutation AAO was derived from known symptomatic cases reported across literature, the DIAN-OBS study and the DIAN expanded registry, encompassing data from ~125 individuals. This study report uses the variant mean symptomatic AAO (53.7 years) as a reference for the expected years from onset, which is estimated with greater variability and is well-documented across individuals from different families carrying this variant³. Using the mean mutation AAO, the participant is 18 years past the expected AAO.

Clinical manifestations.

The exceptional resilience mutation carrier has not shown signs of clinical or cognitive impairments since enrollment in the DIAN study in 2011 at age 61. At enrollment, the Clinical Dementia Rating (CDR) score was 0 and the CDR sum-of-boxes was 0, with a Mini-Mental State Examination (MMSE) score of 30. These scores remained within the expected normal range (CDR sum-of-boxes 0 and MMSE 30) for the subsequent four (2014, 2016, 2018 and 2021) assessments performed over 10 years, including the most recent evaluation in 2021 at age 71.

During the interviews at each visit, the participant and his collateral source (longtime spouse (49 years)) denied memory or thinking problems; he recalls appointments and events, including details; no repetitions of statements or questions; normal calculations; no changes or problems in sleep, mood or personality; and no visuospatial problems. On his last visit in 2021, it was noted that he has ‘more hesitation when parking (at hotels)’ but no clear navigational changes. He has had a postural right-hand tremor since 2017. He has shown stable mood, no change in gait or slowness and no behavioral symptoms. Longitudinal assessments using the Neuropsychiatric Inventory Questionnaire, Geriatric Depression Scale and Functional Assessment Questionnaire have remained within the normative range, indicating no evidence of impairment. Details of neurocognitive testing scores are provided in Table 1 and Supplementary Fig. 1. As shown, his performance on delayed story recall (Wechsler Memory Scale—Revised Logical Memory) (35th percentile at baseline, 74th percentile at Visit 10) was within the expected range based on age-matched norms^18,19; however, these estimates may be biased by the pronounced practice effect observed on paragraph recall tests. Performance in processing speed and executive function based on the Digit Symbol Substitution Test (baseline = 18th percentile, Visit 10 = 3rd percentile), trails A and B (baseline = 57th percentile, Visit 10 = 10th percentile; and baseline = 48th percentile, Visit 10 = 21th percentile, respectively) and animal fluency (baseline = 24th percentile, Visit 10 = 8th percentile) were below age-matched expectations. Clinical and psychological characteristics of the DIAN cohort have been previously published (<9th percentile is impaired and 9th–24th is low average)¹⁸. Table 1 shows the performance of asymptomatic (CDR = 0) non-carriers relative to the exceptional resilience mutation carrier.

Table 1 |.

Summary of demographic, clinical-cognitive and biomarker results during DIAN study visits

DIAN follow-up year/visit	2011/Visit 0	2014/Visit 3	2016/Visit 5	2018/Visit 7	2021/Visit 10
Demographic
Age	61	64	66	68	71
DIAN EYO	8	11	13	15	18
Clinical-cognitive performance (reference values mean (s.d.))
CDR global (0.0 (0.0))	0	0	0	0	0
CDR sum of box (0.0 (0.1))	0	0	0	0	0
MMSE (29.1 (1.3))	30	30	30	30	30
Logical memory (13.5 (4.1))	11	14	10	14	15
Digit symbol (62.0 (11.0))	44	34	34	29	24
Word list (delayed) (3.2 (2.0))	3	2	2	NA	NA
Animal fluency (22.4 (5.1))	17	20	17	14	12
Trail A (21.9 (6.2))	27	40	39	30	53
Trail B (55.2 (18.4))	77	50	91	94	128
Digits span forward (6.9 (1.1))	7	10	8	9	7
Digits span backward (5.1 (1.2))	8	5	6	5	5
Neuroimaging
MR hippocampal volume left/right, mm3 (Z-scorea)	3,934 (−0.2)/4,221 (0.2)	3,761 (−0.4)/4,186 (0.3)	3,821 (−0.1)/4,016 (−0.0)	3,563 (−0.6)/4,023 (0.2)	3,484 (−0.5)/3,789 (−0.1)
MR precuneus surface volume left/right, mm3 (Z-scorea)	6,842 (−2.4)/8,060 (−1.3)	6,548 (−2.7)/8,070 (−1.1)	6,570 (−2.6)/7,872 (−1.3)	6,782 (−2.2)/7,808 (−1.3)	6,450 (−2.5)/8,164 (−0.7)
MR lateral ventricles volume, mm3 (Z-scorea)	12,834 (0.5)/12,396 (0.6)	12,826 (0.2)/12,593 (0.4)	16,483 (0.7)/15,699 (0.9)	19,566 (1.1)/17,724 (1.0)	22,870 (1.1)/20,062 (1.1)
FDG PET precuneus	1.605	1.584	1.565	NA	NA
PiB cortical mean SUVR	3.312	3.659	3.616	3.693	3.667
Tau PET precuneus	NA	NA	1.680	1.826	1.822
Tau PET cortical mean	NA	NA	1.661	1.737	1.950
Biofluids
LUMIPULSE amyloid-β42	430	327	382	361	359
LUMIPULSE amyloid-β42/amyloid-β40	0.048	0.039	0.044	0.041	0.046
LUMIPULSE tau	265	275	326	321	462
LUMIPULSE pTau181	34.1	44.7	49.4	47.1	60.6
MS amyloid-β42	0.78	0.85	0.76	0.78	-
MS amyloid-β42/amyloid-β40	0.067	0.069	0.066	0.067	-
MS pT217/T217	10.00	NA	11.05	7.95	7.83
MS PT181/T181	33.49	NA	38.25	32.15	34.16
MS pT205/T205	0.509	NA	0.549	1.043	1.085
Serum NfL (pgml−1)	30.5	28.6	NA	NA	NA

Reference values on clinical-cognitive performance are provided relative to non-carriers (CDR = 0). EYO, estimated years to symptom onset.

^aNormalized values based on a cognitively unimpaired cohort. NA, not available; MR, magnetic resonance; PET, positron emission tomography; SUVR, standardized uptake value ratio; NfL, neurofilament light chain; MS, mass spectrometry.

On neurologic examination, findings were within normal limits with the exception of right-hand essential or postural tremor in 2014, 2016, 2018 and 2021 and mild hypomimic facies noted in 2016. His Unified Parkinson’s Disease Rating Scale Motor Exam has consistently been within normal limits, except for slight and infrequent present right-hand tremor.

His medications are atorvastatin 10 mg, aspirin 81 mg and ibuprofen 200 mg daily —no other relevant clinical or medical history. His general blood panel was unremarkable.

Environmental exposures.

A semistructured interview and questionnaires were administered to obtain a lifetime history of environmental influences, including exercise, drug/smoking/toxin exposures, cognitive activity and others. One unusual and potentially relevant aspect was high exposure to high heat for years while working as a mechanic in a diesel engine in a naval ship for many hours a day, requiring him to be hosed down with water to prevent overheating. No other exposome factors were unusual.

Genetic analysis.

Our genetic analyses confirmed that the exceptional resilience mutation carrier was a heterozygous carrier of the PSEN2 p.Asn141Ile mutation and was homozygous for APOE3. Figure 1 shows the family pedigree of the exceptional resilience mutation carrier.

We further investigated for presence of the APOE3-Christchurch (APOECh-p.Arg136Ser) and RELN-COLBOS (RELN p.His3447Arg) mutations, but those previously described protective variants were absent in the exceptional resilience mutation carrier or any other family member. To identify potential protective variants, we performed segregation analyses addressing three potential modes of inheritance: dominant, de novo and recessive (Fig. 1). For all models, only missense and splicing, non-sense, stop-loss, start-gain, start-loss variants and small insertion/deletions frameshift (LOF) were included.

The dominant model assumed the presence of the protective heterozygous variant only in the exceptional resilience mutation carrier (10.7). In other words, any rare variant present in any of the symptomatic mutation carriers (4.2, 4.3, 7.1) or offspring of symptomatic mutation carriers when the whole-genome sequence (WGS) of the latter was unavailable (6.1, 7.4, 7.5, 10.11, 10.12, 10.13) was removed. To prioritize variants with potential strong effect size, we then restricted our analysis to variants absent from the non-Finnish European individuals included in gnomAD, absent from AD patients included in Alzheimer’s Disease Sequencing Project R4 (ADSP). We identified 23 rare variants with a Combined Annotation Dependent Depletion (CADD) prediction score above 20, of which eight were predicted LOF (Table 2). These variants were identified in the following genes: FAM151B (p.Ala108fs), CHST15 (p.Arg456fs), CEP290 (p.Gln981fs), GPCPD1 (p.Ser310fs), PRICKLE3 (p.Glu32fs) and MUC3 (p.Ser1679fs; Table 3). Within the de novo model, we assumed that in addition to fulfilling the segregation and minor allele frequency (MAF) criteria for the dominant model, the variant had to be absent from the non-carrier siblings of the exceptional resilience mutation carrier (10.1, 10.2, 10.3, 10.4, 10.5, 10.6; Fig. 1). We identified 11 rare variants with a CADD score above 20 and five de novo LOF variants (Table 2). These variants included the FAM1515B, CEP290, GPCPD1, PRICKLE3 and ANKRD36C variants described above.

Table 2 |.

Number of variants identified from segregation analyses with a putative protective effect

Filters applied	Dominant	De novo	Recessive
All	105,649	76,669	266,625
Coding variants	804	134	352
Coding variants, not in ADSP cases, less than 0.1 NFE	252	88	22
Coding variants, not in ADSP cases, less than 0.005 NFE	119	57	4
Coding variants, not in ADSP cases, less than 0.005 NFE, CADD > 20 or missing	58	18	1
Coding variants, not in ADSP cases, not in NFE, CADD > 20 or missing	23	11	0
LOF, not in ADSP cases, not in NFE, CADD > 20 or missing	8	5	0

NFE, MAF in gnomAD non-Finnish European is reported.

Table 3 |.

Candidate variants and genes from segregation analyses with a putative protective effect

Position	Ref allele	Alt allele	Effect	Gene	cDNA change	Protein change	MAFgnomAD NFE	Model
chr7:100956813	G	GAA	Frameshift	MUC3A	c.5035_5036insAA	p.Ser1679fs	0	D
chr15:89753039	CAAA	CAA	Splicing	MRPL15P1	n.450-2delA	·	·	D
chr10:124021236	CGG	CGGGG	Frameshift	CHST15	c.1363_1364dupCC	p.Arg456fs	·	D
chr5:80519697	GC	G	Frameshift	FAM151B	c.323delC	p.Ala108fs	·	D/DN
chr12:88077271	TG	T	Frameshift	CEP290	c.2941delC	p.Gln981fs	·	D/DN
chr20:5575485	G	GA	Frameshift	GPCPD1	c.928dupT	p.Ser310fs	·	D/DN
chrX:49184658	TCC	T	Frameshift	PRICKLE3	c.93_94delGG	p.Glu32fs	·	D/DN
chr2:95939016	C	T	Splicing	ANKRD36C	c.1532–1G>A		0	D/DN
chr1:170995396	A	T	Missense	MROH9	c.1202A>T	p.Gln401Leu	1.35×10−3	R

cDNA, complementary DNA; MAF gnomAD NFE, MAF in gnomAD non-Finnish European is reported; d, dominant; DN, de novo; R, recessive; Ref Allele: reference allele; Alt Allele, alternate allele.

Finally, using the recessive model, we assumed the presence of the potential protective homozygous alternative variant only in the exceptional resilience mutation carrier (10.7) and homozygous alternative genotype absent from 4.2, 4.3, 7.1, 6.1, 7.4, 7.5, 10.11, 10.12 and 10.13 (Fig. 1). We restricted our analysis to variants with MAF less than 0.005 in non-Finnish European individuals included in gnomAD and with homozygous state absent from AD patients included in ADSP R4, allowing for the presence of heterozygous genotypes among symptomatic ADSP participants.

We identified four coding variants fulfilling these criteria (Table 2), from which only one had a CADD score above 20, in the MROH9 gene (p.Gln401Leu; Table 3). Note that no rare coding variants in RELN fulfilling any of the three models were detected.

A burden test among the candidate genes was performed on the ADSP R4 dataset, but no significant association was identified for the 13 candidate genes (Supplementary Table 1).

In addition to the three potential modes of inheritance model approach for variant identification, we aimed to identify any coding variants in the previously reported AD risk genes. The exceptional resilience mutation carrier harbored a TREM2 p.Asp87Asn variant (rs142232675) whose association with AD remains unclear^20–23. The exceptional resilience individual is also a carrier of a reported protective MAPT variant p.Tyr441His (rs2258689)²⁴, which is on the 17q21.31 MAPT-H1 haplotype background. Other variants in the vicinity of the MAPT gene and exclusive to the exceptional resilience mutation carrier are described in Supplementary Table 2. We also identified the CD33 p.Gly304Arg (rs35112940) with a putative protective effect, because this variant has an odds ratio = 0.53 (P = 0.03) in the UK Brains for Dementia Research.

Imaging and fluid biomarkers.

The participant’s ¹¹C-Pittsburgh compound-B (PiB) positron emission tomography (PET) images showed elevated signal corresponding to positive amyloid at his baseline visit, 8 years past his expected AAO, and subsequent follow-up visits with PiB cortical summary measures of ~3.6 (abnormal > 1.42) SUVR (Table 1). The regional profile of amyloid burden was typical for an individual carrying a pathogenic DIAD mutation at least 5 years past their AAO (Fig. 2 and Supplementary Fig. 2). Tau PET signal of the participant showed focal asymmetric tau deposition solely in the occipital lobe, which is not consistent with a typical DIAD profile (Fig. 2). ¹⁸F-fluorodeoxyglucose (FDG) PET glucose metabolism showed visible hypometabolism colocalized with the focal tau deposition but not affected in typical parieto-occipital regions. Brain structures such as the hippocampus, measured on volumetric magnetic resonance imaging (MRI), showed longitudinal decrease (for example, left hippocampus was 3,934 mm³ at baseline versus 3,484 mm³ 10 years later) consistent with age (for example, left hippocampus Z-score between −0.2 and −0.5; Table 1). The participant also showed markers of small-vessel disease as white-matter hyperintensity consistent with age and had one microhemorrhage in the left periventricular parietal area visible at all visits from 2011 to 2021 and one incident microhemorrhage in the right subcortical temporal lobe at the 2021 visit.

Fig. 2 |. Z-scored voxel-wise PET images of the exceptional resilience mutation carrier and of averaged symptomatic mutation carriers relative to non-carriers.

Amyloid (PiB, top row), metabolism (FDG, middle row) and tau (AV1451, bottom row). SUVR values were Z-scored relative to non-carriers aged 50 years old and above. ^aThe FDG color scale was inverted for visualization so red consistently indicates abnormality across all measures. Left, the most recent available images for the exceptional resilience mutation carrier (MC). Middle, average of symptomatic mutation carriers at 5 years past their onset for PiB and FDG and of all symptomatic carriers for AV1451. Right, difference between the exceptional resilience mutation carriers and typical mutation carriers. Red indicates greater pathology, and blue indicates less pathology in the exceptional resilience mutation carrier relative to the average symptomatic mutation carriers.

The exceptional resilience mutation carrier had five study visits with biofluid collection. AD fluid biomarker levels by study visit are described in Table 1. The longitudinal biomarker profile relative to the matched mutation carrier is shown in Supplementary Fig. 3. Cerebrospinal fluid (CSF) proteomic biomarker measures and imaging measures in the exceptional resilience mutation carrier relative to mutation carriers and non-carriers are shown in Fig. 3 and Supplementary Figs. 3–10. In summary, the exceptional resilience mutation carrier exhibited CSF amyloid-β levels and amyloid-β42/40 ratios comparable to those seen in other symptomatic mutation carriers. No targeted proteomic markers we previously measured in DIAN²⁵ were identified as outliers compared to other ADAD mutation carriers or PSEN2 carriers (Supplementary Figs. 7–11). Although most markers of synapses, sugar metabolism, immune response and acute phase processes were at non-carrier levels, phosphorylated tau markers at positions 181, 217 and 205 (Table 1) were found to be intermediate between carrier and non-carrier levels (Fig. 3).

Fig. 3 |. Exceptional resilience mutation carrier biomarker measurements at EYO 13.

CSF proteomic biomarker measures, imaging measures and cognitive and functional measures were modeled in mutation carriers and non-carriers across the disease course by EYO. The estimates at EYO 13 are shown here. Each measure was normalized between the minimum and maximum estimate value across the entire EYO range. Vertical lines indicate the mean estimate in carriers (red) and non-carriers (blue). Color shading indicates the 99% credible intervals. The yellow diamond indicates the exceptional resilience mutation carrier measure. Colored boxes indicate the proteomic brain module that maps to the CSF proteomic measurement, as previously described²⁵. ^aMeasures decreased in carriers are inverted for visualization purposes. MAPK, mitogen-activated protein kinase; PSD, post-synaptic density; ECM: extracellular matrix.

In addition, we explored untargeted proteomic and metabolomic profiles of the exceptional resilience mutation carrier relative to other mutation carriers in the family (Supplementary Fig. 2). The exceptional resilience mutation carrier is not an overall proteomics outlier, but compared to other PSEN2 p.Asn141Ile carriers in the DIAN cohort, he exhibits a lower abundance for almost 200 proteins and higher abundance for almost 300 proteins (Supplementary Fig. 3). The highly abundant proteins are enriched in pathways affecting protein folding and include multiple members of the heat shock protein family (Supplementary Fig. 2b). Similarly, there was not a significantly different metabolomics profile; but 28 metabolites have different levels than other PSEN2 p.Asn141Ile carriers. Combining the differentially abundant metabolites with differentially abundant proteins, we identified three pathways altered in the exceptional resilience mutation carrier that play important roles in protein synthesis, anti-aging and anti-oxidative mechanisms.

Discussion

Our study performed a comprehensive deep analysis of genetic, imaging and biomarker profiling of an individual resilient to a DIAD pathogenic variant who has not exhibited the typical clinical onset of dementia compared to related family members or mutation carriers. Although we have not identified any previously described protective factors, our analysis has revealed several putative protective genetic variants. To date, no functional studies have been conducted to confirm or refute the role of these variants in conferring resilience to AD. Our extensive clinical, cognitive, biomarker and genetic datasets provide in-depth phenotyping of this individual. Existing biological samples from this study offer valuable resources for ongoing and future research in identifying therapeutic targets. We have established a platform at DIAN for interested investigators to access these data, propose studies and suggest collaboration ideas aimed at understanding how this individual has resisted the clinical onset of AD. Identifying the underlying mechanisms may present significant opportunities for advancing research and developing new treatments for AD.

In this remarkable resilience case, amyloid deposition was consistent with individuals carrying a DIAD mutation²⁶; however, the evidence of tau pathology (as measured by tau PET), primarily to the left occipital region, without evidence of spreading contrasts markedly with typical AD progression, where tau typically spreads to multiple brain regions, correlating with symptom progression and cognitive decline^27–31. This localized tau deposition also contrasts with more typical cases of DIAD, where tau is known to correlate with symptom progression and cognitive decline³².

Notably, although the neuroimaging findings of tau pathology distribution in the occipital cortex may resemble those observed in posterior cortical atrophy, this patient did not exhibit clinical features consistent with posterior cortical atrophy syndrome, such as visual or spatial impairments at neurological exam or neurocognitive testing. As a limitation, our study did not include highly sensitive visuospatial tasks that could detect subtle visual impairments that might not be captured by standard clinical–cognitive assessments. Highly sensitive visuospatial tasks might have detected subtle changes that could correlate with the observed tau distribution.

Evidence from previous studies in DIAD highlights the critical role of tau aggregation in triggering clinical onset; independent of amyloid-β pathology, tau accumulation, particularly in the medial temporal lobe, closely precedes and predicts symptom onset, suggesting that tau aggregation may serve as a primary driver of AAO in these individuals^32,33. Thus, factors limiting tau propagation in this individual may have conceivably decoupled the genetic risk of Alzheimer’s from its typical phenotypic manifestation.

This pattern of restricted tau deposition is similar to previously reported extreme resilience cases, where individuals carrying the APOECh and RELN-COLBOS mutations showed remarkably limited tau pathology despite a high amyloid-β plaque burden^8,10. Postmortem findings from the extremely resilient ADAD PSEN1 E280A carrier, who was homozygous for the APOE3Ch variant, confirmed restricted tau deposition. This case exhibited an unusual regional distribution of pathology, with sparing of the frontal cortex, severe involvement of the occipital cortex, a distinctive chronic inflammatory microglial profile and reduced APOE expression³⁴.

This containment of tau pathology, a common feature across currently known extremely resilient individuals, might be a critical factor in delaying the onset of Alzheimer’s symptoms by preventing the neurodegenerative processes typically driven by widespread tau deposition. In cases of localized (no spreading) tau deposition, the neurodegenerative cascade that leads to typical Alzheimer’s symptoms might be interrupted or significantly slowed, suggesting a new form of disease modification that warrants further exploration.

The mechanisms by which tau spread is restricted in this individual may provide crucial insights into potential therapeutic targets for preventing or slowing the progression of AD. The connection between amyloid-β and tau pathology appears to be disrupted in this individual, given the intermediate levels of pTau species that are typically associated with amyloid-β plaque formation, which may suggest a unique plaque microenvironment that is less conducive to downstream neurofibrillary tangle formation. Our study identified several genetic variants in the exceptional resilience mutation carrier, notably near the MAPT haplotype, which, although not reaching statistical significance, provide intriguing insights into potential mechanisms of disease resistance. These variants could subtly influence neurodegenerative pathways, perhaps mitigating the pathological impact of tau and limiting tau spread. The exceptional resilience mutation carrier’s unique environmental exposure—significant heat exposure due to professional circumstances—suggests a possible biological adaptation that might confer neuroprotection. Chronic exposure to such stress could activate cellular pathways involved in heat shock response and protein folding, which have been implicated in cellular resilience in neurodegenerative diseases. This leads us to speculate that the observed enrichment of heat shock protein family members and other protein-folding-related proteins in the exceptional resilience mutation carrier could be a potential mechanism in mediating this protective effect.

In term of genetics, we performed extensive segregation analyses encompassing several modes of inheritance for identification of the potential protective variant. Even though we used stringent filtering criteria, we could not objectively pinpoint one variant as being associated with the protective phenotype observed in this participant. Interestingly, the upregulation of GPCPD1, an enzyme involved in choline production, was recently associated with cognitive resilience in AD patients³⁵. The GPCPD1 frameshift variant may escape non-sense-mediated decay, potentially leading to a gain of protein function, but additional work is needed to assess its effect. Finally, we identified variants in CD33 with potential protective effects that may play a role in the phenotypic presentation. Overall, more work is needed to narrow down the list of candidate variants, especially additional family members and expanding on structural variant detection.

This study requires cautious interpretation given the inherent limitations associated with generalizing findings from a single case to a broader population, especially in the context of an n = 1 study that restricts the ability to draw statistically significant conclusions. Resilience to AD is likely influenced by a multifaceted interplay of genetic, environmental and biological factors, complicating efforts to pinpoint precise putative factors responsible for disease modification. However, by contextualizing the findings within the broader DIAN cohort, we aim to highlight the unique aspects of this case, acknowledging that the insights gained are primarily descriptive and intended to generate hypotheses for future research.

Despite the limitations, our findings highlight the need for further studies to validate the protective effects in identified pathogenic variants and explore their biological mechanisms as potential avenues for therapeutic targets. Additionally, broader research into the balance between risk and protective factors in genetically predisposed populations is essential. Such studies could expand our understanding of AD and lead to the development of comprehensive strategies that incorporate both genetic and lifestyle factors to manage or prevent the disease. It is our hypothesis that the restriction of tau spread has contributed to the relative preservation of cognition in this individual, yet the exact factors that prevent tau spread in this case remain unclear. However, the observation of localized, non-spreading tau pathology provides an important proof of concept for therapeutic approaches targeting tau dynamics. This research could have broad implications for the development of treatments aimed at mitigating tau pathology in the wider AD population. Understanding the mechanisms that restrict tau spread in this individual could provide crucial insights into potential therapeutic targets for preventing or slowing the progression of AD. We invite researchers to join us in this search.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41591-025-03494-0.

Methods

Clinical assessment

Participants underwent a detailed clinical evaluation that included family history, personal medical history, current medication and a thorough neurological examination. Clinicians were blinded to participants’ genetic status and conducted the clinical evaluation of dementia status according to DIAN-OBS standard protocols. At all visits, participants underwent a neuropsychological examination evaluating delayed recall (Wechsler Memory Scale—Revised Logical Memory), category fluency (Animal Naming Test) and processing speed (Wechsler Adult Intelligence Scale—Revised Digit Symbol Substitution Test)¹⁸. Further, the MMSE, Geriatric Depression Scale and Neuropsychiatric Inventory were included in each assessment. An additional questionnaire exploring medical history and exposome factors included multipurpose health surveys from the National Health and Nutrition Examination Survey (wwwn.cdc.gov/nchs/nhanes/)³⁶.

Genetic and molecular studies

As part of the DIAN protective factor project, genetic analysis was performed comparing participating family members. The father of the exceptional resilience mutation carrier was one of 14 children, 11 of whom would go on to develop AD at or close to the age of 50 (Fig. 1). We obtained detailed phenotypic and biological data for the living individuals of four of these 14 children (children 4, 6, 7 and 10), here onward named subfamilies 4, 6, 7, 10. For segregation analyses, we utilized whole-exome sequencing (WES; n = 3) and WGS (n = 16) from blood for 16 family members: two individuals of subfamily 4, one individual of subfamily 6, three individuals of subfamily 7 and 10 individuals of subfamily 10 (including the exceptional resilience mutation carrier).

All raw files, regardless of whether they were generated by WES or WGS, were aligned individually to the reference genome GRCh38 following GATK’s Best Practices Workflow standards³⁷ with either BWA-MEM or NVIDIA’s ClaraParabricks optimizations for high-performance computing, providing the same results regardless of specific software. Details of the pipeline employed can be found on GitHub (NeuroGenomicsAndInformatics/WXS-Pipelines (github.com)). Briefly, the pipeline includes an optimized set of commands to go from sequencing reads (fastq, CRAM or BAM) to gVCF files. Samples with a freemix value greater than 0.05 or coverage less than 30× (for WES) or 20× (for WGS) are marked as failed.

We used WGS data from 14 family members and WES data from two family members, and all passed quality controls. To identify potential protective variants, we performed segregation analyses addressing multiple modes of inheritance: dominant, de novo and recessive. Only variants altering the protein coding sequence were included based on SnpEffect (PubMed IDentifier) annotation. The dominant model assumed the presence of the protective heterozygous rare variant in 10.7 and its absence from 4.2, 4.3, 7.1, 6.1, 7.4, 7.5, 10.11, 10.12 and 10.13. For the de novo model, we applied the same segregation strategy, and variants were further removed if present in 10.1, 10.2, 10.3, 10.4, 10.5, 10.6. For the recessive model, we selected homozygous alternative variants in 10.7 for which the homozygous alternative genotype was absent from 4.2, 4.3, 7.1, 6.1, 7.4, 7.5, 10.11, 10.12 and 10.13. ADSP R4 WGS and gnomAD v.3.1.2 non-Finnish European data were used to further filter out variants³⁸.

Association between variants—defined as missense, splicing, non-sense, stop-loss, start-gain, start-loss variants, small insertion/deletions frameshift and CADD prediction score above 20—with AD were assessed by burden test. We used the same models as described earlier: dominant and de novo (absent from the non-Finnish European individuals included in gnomAD) and recessive (MAF less than 0.005 in non-Finnish European individuals included in gnomAD). We ran a segregation analysis within this family, and we comprehensively examined all rare (MAF < 1%) and common missense variants that were (1) exclusive to the exceptional resilience mutation carrier, (2) exclusive to the exceptional resilience mutation carrier and controls within the family and subfamily of the exceptional resilience mutation carrier and (3) present in any of the AD-associated loci (as per ref. 38, including MAPT); we also performed a search for new potential protective factors exclusive of the exceptional resilience mutation carrier. Association between variants—defined as missense, splicing, non-sense, stop-loss, start-gain, start-loss variants, small insertion/deletions frameshift and CADD prediction score above 20—with AD were assessed by burden test. We used the same models as described earlier: dominant and de novo (absent from the non-Finnish European individuals included in gnomAD) and recessive (MAF less than 0.005 in non-Finnish European individuals included in gnomAD).

In vivo neuroimaging

Details about DIAN MRI and PET protocols has been published elsewhere³⁹. In summary, MRI sessions consisted of volumetric T1-weighted images for the evaluation of atrophic processes, T2-star or susceptibility weighted images for the detection of microhemorrhages or other hemosiderin deposits and T2-FLAIR images for detection of markers of small-vessel disease. Amyloid burden was estimated with PiB, glucose metabolism with FDG and tau pathology with F18-AV1451 radiotracer. PET images were registered using a FreeSurfer region-of-interest-based approach, as previously described, with cerebellum cortex as the reference region⁴⁰. PET images from non-carriers over 50 years old without abnormal PiB deposition were registered to the Montreal Neurological Institute (MNI) space, and average and standard deviation (s.d.) images were created for each modality. PET images from the participant were co-registered to the MNI space and Z-scored in a voxel-wise approach using the average and s.d. non-carrier images. PET images from symptomatic mutation carriers were co-registered to the MNI space, averaged and also Z-scored relative to non-carriers for visual comparison of each modality. Note that due to the limited numbers of acquired tau PET images in the DIAN-OBS study, the Z-scored images were based on four non-carrier participants, and the Z-scored average symptomatic carriers utilized five scans. All T1-weighted MRIs were processed with FreeSurfer, and regions of interest were normalized against a Normal Aging Cohort for all visits as previously described.

Biofluids collection

Biofluids were collected according to the specific guidelines of DIAN study protocols. CSF was collected under fasting conditions by lumbar puncture, using an atraumatic Sprotte spinal needle, into two 13-ml polypropylene tubes and then flash-frozen on dry ice and subsequently aliquoted into polypropylene tubes before storage at −80 °C. Amyloid-β42, amyloid-β40, total tau and phosphorylated tau-181 (pTau181) were measured using the validated LUMIPULSE G1200 immunoassay (Fujirebio) according to standardized procedures. CSF microtubule binding region concentrations were measured using liquid chromatography–mass spectrometry (LC–MS)/MS. Full details of the CSF preparation and LC–MS/MS processing have been previously outlined in detail⁴¹.

CSF proteomic and metabolomic measurements have been previously described²⁵. CSF samples were analyzed using two platforms for proteomic and metabolomic profiling: the SOMAscan platform by Somalogic measured 7,584 protein aptamers corresponding to 6,179 unique proteins across 3,065 samples. Additionally, Metabolon, Inc. conducted untargeted Precision Metabolomics LC–MS analysis for metabolomics data generation. Targeted MS-based proteomic analysis was performed as previously described²⁵. Modeling of biomarkers in mutation and non-mutation carriers is described in Supplementary Appendix 1.

Statistical analysis

All data presented are expressed as mean values and standard deviations. Targeted MS and other biomarker data modeled with Bayesian statistics are presented as 99% credible intervals as described in Supplementary Appendix 1.

Ethics considerations

The study is supervised by the Institutional Review Board at Washington University in St. Louis, USA, and all study procedures were approved by the Human Research Protection Office and the Institutional Review Board at Washington University or the respective participating sites.

Data availability

Data and study samples (for example, CSF, plasma) supporting the findings of this study are available on request and will follow the policies of the DIAN (https://dian.wustl.edu), which comply with the guidelines established by the Collaboration for Alzheimer’s Prevention. De-identified whole-genome sequencing data will be made available through NIAGADS as part of the DIAN data-sharing process.