Long-Read Sequencing Reveals Novel Epigenetic Mechanisms in Alzheimer’s Disease
DINGLIHUAThe allele-specific epigenetic regulatory mechanisms of APOE—the strongest genetic risk factor for Alzheimer’s disease (AD)—have long remained elusive.
In June 2026, a collaborative team from the National Institute on Aging (NIA) at the U.S. National Institutes of Health (NIH), the University of California, Santa Cruz, Northeastern University, and other institutions published a breakthrough study in npj Dementia, a Nature Portfolio journal, titled “Haplotype-Resolved DNA Methylation at the APOE Locus Identifies Allele-Specific Epigenetic Signatures Relevant to Alzheimer's Disease Risk.” Using Oxford Nanopore long-read sequencing technology, the team generated, for the first time, an allele-specific methylation map of the human brain APOE locus at haplotype resolution. They identified 18 novel differentially methylated CpG sites (DMCs), opening a new dimension for epigenetic research and precision risk assessment in AD.[1]
I. Background
APOE is the strongest genetic risk factor for Alzheimer’s disease (AD). Among the three common alleles, ε2 (rs7412 C>T) confers a protective effect, while ε4 (rs429358 T>C) significantly increases AD risk—by 3- to 4- fold in heterozygotes and over 10- fold in homozygotes. In addition, the ε3- linked rs769455[T] variant has been shown to synergize with ε4 in individuals of African ancestry, further elevating AD risk. All of these coding variants reside within a CpG island in APOE exon 4, and 5- methylcytosine (5mC) methylation at CpG sites is a key mechanism of gene expression regulation. This gave rise to a central hypothesis: beyond altering protein function, different APOE alleles may regulate gene expression through epigenetic pathways by modifying the methylation status of CpG sites—constituting a novel “genetic variation → epigenetic alteration → disease risk” axis.
However, allele- specific methylation maps had long remained unresolved due to three major technical limitations of conventional detection methods:
1.Sparse CpG coverage – The Illumina 450K methylation array covers only 46 CpG sites across the APOE region, leaving the vast majority of sites in a “blind zone.”
2.Inability to phase haplotypes – Conventional methods return only the average methylation level across both alleles, completely masking the independent methylation patterns of the maternal and paternal chromosomes in heterozygous individuals.
3.Bisulfite treatment exacerbates DNA degradation – Postmortem brain tissue DNA is already highly fragmented; the chemical conversion step further compromises data quality.
Leveraging the NIH CARD Long- Read Initiative, the study employed the Oxford Nanopore PromethION long- read sequencing platform to overcome these bottlenecks in a single stroke: it detected 1,556 CpG sites in the same region (versus 46 with the array)—a 33.8- fold increase in coverage; it directly detected 5mC modifications on native DNA at single- molecule resolution without bisulfite conversion; and it captured both SNV genotype and CpG methylation status on individual reads, enabling native allele phasing.
Figure 1. Comparison of long-read sequencing and conventional array-based detection workflows, and overview of the APOE gene cluster.
II. Study Cohort Design
The study comprised two independent cohorts:
1.NABEC cohort (North American Brain Expression Consortium) : 201 individuals of European ancestry, yielding 402 haplotypes, including 48 ε2 alleles and 58 ε4 alleles.
2.HBCC cohort (Human Brain Core Collection) : 131 individuals of African and African admixed ancestry, yielding 262 haplotypes, including 25 ε2 alleles, 64 ε4 alleles, and 7 copies of the African- ancestry- specific rs769455[T] risk allele.
Across both cohorts, the study included 332 postmortem brain samples and 664 independent haplotypes (73 ε2, 122 ε4), representing the largest haplotyperesolved methylation dataset at the human brain APOE locus to date.
III. Major Findings
3.1 CpG Coverage Increased 34- Fold
The study directly compared CpG site detection across the same genomic region (chr19:44889556–44953378, covering the TOMM40, APOE, APOC1, and APOC4- APOC2 gene cluster) using the two detection platforms. In 697 European- ancestry samples from the ROS/MAP (Religious Orders Study / Memory and Aging Project) cohort, the Illumina 450K methylation array detected only 46 CpG sites. In contrast, Oxford Nanopore PromethION long- read sequencing of 205 European- ancestry samples from the NABEC cohort covered 1,556 CpG sites in a single run—33.8 times the number detected by the array. The vast majority of ONT- unique CpG sites had never been detected before—effectively “dark matter” of the epigenome—and multidimensional annotations from the UCSC Genome Browser revealed that these sites fall precisely within known regulatory element clusters, suggesting their potential functional importance.
Figure 2. Comparison of CpG site coverage at the APOE locus.
3.2 18 Novel Differentially Methylated Sites (DMCs)
Through linear regression analysis (BH- FDR corrected P < 0.05), the study identified CpG sites significantly associated with APOE alleles in both the NABEC and HBCC cohorts.
NABEC cohort (European ancestry) : Within the TOMM40–APOE–APOC1–APOC4- APOC2 gene cluster, multiple CpG sites showed methylation levels significantly associated with ε2 and ε4 alleles, with ε2 and ε4 exhibiting opposing methylation effects. Key sites, such as cpg_chr19_44914329 and cpg_chr19_44917997, showed methylation levels stratified by APOE genotype.
Figure 3. Differentially methylated CpG sites associated with APOE alleles in the NABEC cohort.
HBCC cohort (African and African admixed ancestry) : Eleven significant DMCs were identified, associated with ε2, ε4, and rs769455[T] respectively. This marked the first discovery of an independent methylation effect for rs769455[T]—an African- ancestry- specific AD risk variant—at sites such as cpg_chr19_44904817 and cpg_chr19_44896082.
Figure 4. Differentially methylated CpG sites associated with APOE alleles and rs769455[T] in the HBCC cohort.
Combining both cohorts, a total of 18 novel DMCs were identified, all located within the dense TOMM40–APOE–APOC1–APOC4APOC2 gene cluster, with associated alleles spanning ε2, ε4, and rs769455[T]. Several DMCs were significant in both ancestral backgrounds, suggesting that these methylation patterns may be broadly applicable across populations.
IV. Scientific Significance and Clinical Implications
Methodological level: “Single molecule, single haplotype” represents a core breakthrough of long- read sequencing in epigenetics. Conventional technologies return only population- average methylation, obscuring true inter- allelic differences. Long- read sequencing locks SNV genotype and CpG methylation status onto the same DNA molecule, achieving true haplotype resolution. This advantage translates directly into clinical value—higher diagnostic sensitivity, more precise risk stratification—and eliminates bisulfite conversion, greatly simplifying the experimental workflow.
Mechanistic level: This study reveals a novel “genetic variation → epigenetic alteration → disease risk” pathway: missense variants in the APOE coding region not only alter protein function but also reshape the CpG island methylation landscape, thereby influencing transcriptional regulation of neighboring genes. The 18 DMCs identified represent the epigenetic nodes along this axis. The ONT platform’s “one- run” sequencing simultaneously captures SNVs, SVs, 5mC, and haplotype phasing, providing a “panoramic” molecular report for AD risk assessment.
Precision medicine level: The independent methylation signal of rs769455[T] suggests that AD risk in different ancestral backgrounds may be mediated through distinct epigenetic pathways, offering critical clues for cross- ancestry precision risk assessment. Moreover, this study demonstrates that long- read sequencing outperforms conventional methods on low- quality clinical samples such as postmortem brain tissue, validating its feasibility for “real- world” applications.
Long- read sequencing technology is redefining the paradigm for understanding classic genetic loci. APOE has been a central focus of AD research for four decades, and the pathogenic mechanisms of its coding variants were long thought to be limited to alterations in protein function. By generating a haplotype- resolved methylation map of the human brain APOE locus, this study has uncovered an allele- specific epigenetic dimension that was systematically masked by conventional technologies. The discovery of 18 novel DMCs provides the first experimental evidence for the multi- layer pathogenic axis of “genetic variation → epigenetic regulation → disease risk.” The far- reaching implication of this paradigm shift is that a genomically informed interpretation with true clinical translational value must simultaneously address two questions: which variant is carried and how that variant is regulated.
[1] Rylee M. Genner, Melissa Meredith, Kensuke Daida, Abraham Moller, et al. Haplotype-resolved DNA methylation at the APOE locus identifies allele-specific epigenetic signatures relevant to Alzheimer’s disease risk. Nature. npj Dementia volume 2, Article number: 45 (2026).
https://www.nature.com/articles/s44400-026-00094-8
