Newshot

Hypothalamic volume and cell-type changes in ageing and Alzheimer’s disease

Summary

Nature, Published online: 18 November 2026; doi:10.1038/s41586-026-11305-w HypoAD, an atlas of the human hypothalamus in ageing and Alzheimer’s disease that combines high-resolution magnetic resonance imaging from 458 individuals with single-nucleus RNA sequencing of 614,403 nuclei, reveals that metabolism- and circadian-regulating subregions, and their inhibitory neurons and microglia, are particularly vulnerable.

Original text(via Nature)

Main

Alzheimer’s disease (AD) is an age-associated neurodegenerative disease affecting millions worldwide, with a prevalence of roughly 10–30% among individuals older than 65 years3. Although its cardinal feature is cognitive impairment, AD manifests as a systemic condition involving profound disruptions in sleep, circadian rhythms, metabolic homeostasis, stress responses and neuroendocrine functions1. The canonical neuropathological hallmarks are extracellular amyloid-β plaques and intracellular neurofibrillary tangles (NFTs), which alongside widespread neuroinflammation and volumetric brain changes, also intensify with ageing1,4. Because cognitive decline has traditionally guided both diagnosis and research, investigation has focused mainly on the hippocampus and cortex5. As the effects of AD on other brain regions remain poorly understood, investigating these understudied areas is crucial for elucidating the full spectrum of disease mechanisms and developing new therapeutics6.

The hypothalamus, a small but complex brain region, orchestrates a wide range of physiological and neuroendocrine processes frequently dysregulated in ageing and AD2,7. Emerging evidence suggests that hypothalamic dysfunction not only correlates with but also actively exacerbates AD pathology and cognitive decline1. Specifically, pathology within hypothalamic subregions, such as the suprachiasmatic nucleus (SCN)8, arcuate nucleus (ARC)1,9 and mammillary body10, has been associated with prominent sleep disturbances, metabolic dysfunction and memory deficits observed in patients with AD. These systemic disruptions, in turn, are proposed to create a vicious cycle that further accelerates disease progression and cognitive decline1.

Despite its central role in homeostasis and brain ageing2,11, the hypothalamus remains an understudied region in the context of human ageing and AD. In particular, we have a limited understanding of the volumetric, cellular and molecular changes within the human hypothalamus during these processes. Although studies in animal models have highlighted hypothalamic involvement in ageing and neurodegeneration2,11,12, human studies remain scarce. This may be in part because the hypothalamus is more evolutionarily conserved than the cortex and is critical for survival, and thus it may show less overt vulnerability to ageing and prevalent diseases13. This paucity is especially pronounced for investigations that combine high-resolution imaging with single-cell analyses, complementary approaches essential for dissecting the anatomical and cellular heterogeneity of this complex brain region. This knowledge gap hinders our ability to determine how specific hypothalamic subregions and cell populations contribute to ageing and AD, and why they show differential vulnerability.

Here we present HypoAD, a multimodal atlas combining two powerful, cross-scale approaches. First, we performed high-resolution volumetric analysis of clinical imaging data from 458 individuals across the adult lifespan with varying cognitive and amyloid-β pathological statuses. Second, we generated a complementary single-cell transcriptomic atlas, comprising 614,403 nuclei from the hypothalami of 17 post-mortem donors, including young and older individuals with or without clinical and neuropathological diagnoses of AD, in which sampling was specifically enriched for the SCN, ARC and mammillary body. This dual-modality approach reveals the subregion-specific and cell-type-specific alterations of the human hypothalamus in ageing and AD.

Volumetric changes with age, sex and AD

Neuroimaging studies have reported changes in hypothalamic volumes associated with ageing and cognitive decline14. Previous analyses were low-resolution or limited to major areas, masking disease-relevant subregions such as the ARC and SCN14. To overcome this limitation, we leveraged a newly developed volumetric atlas of the human hypothalamus15 to delineate and measure 13 hypothalamic and 12 extrahypothalamic structures (Supplementary Table 1). Our analysis included a clinical imaging dataset of 458 individuals across two previously described cohorts: the Adult Metabolism & Brain Resilience (AMBR) study16 (n = 202) and the Connectomics in Brain Aging and Dementia (CBA) study17 (n = 256) (Extended Data Fig. 1a).

Pearson correlation analysis revealed that whereas total hypothalamic volume showed a weak correlation with age (r = 0.15, adjusted P = 1.5 × 10 −3, Fig. 1a, Supplementary Table 2), its constituent subregions showed stronger age-related correlations. Specifically, the lateral hypothalamus volume was negatively correlated with age (r = −0.22, adjusted P = 5.2 × 10−6), suggesting age-related atrophy. By contrast, the volumes of the SCN (r = 0.26, adjusted P = 2.8 × 10−8), ARC (r = 0.32, adjusted P = 1.7 × 10−11), paraventricular nucleus (r = 0.35, adjusted P = 6.9 × 10−14), periventricular nucleus (r = 0.34, adjusted P = 8.8 × 10−13) and dorsomedial hypothalamic nucleus (DMH) (r = 0.23, adjusted P = 1.8 × 10−6) all increased with age across all 458 individuals in two studies: AMBR and CBA (Fig. 1a,b and Supplementary Tables 2 and 3). A few subregions (SCN, mammillary body, dorsal periventricular nucleus and tuberomammillary nucleus) showed lower test–retest reliability (intraclass correlation coefficient (ICC) < 0.4; Extended Data Fig. 1b), probably reflecting their small size and limited tissue contrast; volumetric results for these subregions should therefore be interpreted with caution. In the AMBR cohort the positive correlation between DMH volume and age was not detected in cognitively normal, amyloid-β negative individuals (Extended Data Fig. 1c,d and Supplementary Tables 2 and 3), but it remained present in the larger harmonized two-study cohort (n = 458; Extended Data Fig. 1c,d). To test whether this age-related increase in DMH volume itself differs by disease status, an age × group interaction model (group: cognitively normal and amyloid-negative versus all others; Methods) showed no significant interaction (effect size/estimate ± standard error (s.e.) = −0.041 ± 0.025, adjusted P = 0.12, Supplementary Table 4), indicating that age-related increase in DMH volume does not differ significantly between groups. This suggests that the DM–age association probably reflects a general feature of hypothalamic ageing in this subregion—which is implicated in sleep–wakefulness18 and body-weight regulation19—rather than a process specific to neurodegeneration.

Fig. 1: Sex-specific alterations in hypothalamic subregion volumes with ageing and AD.
Fig. 1: Sex-specific alterations in hypothalamic subregion volumes with ageing and AD.

a, Pearson correlation heatmap of subregion volumes and age at MRI. Blue and red indicate positive and negative correlations, respectively. Boxes mark five clusters (complete linkage on 1 − r). All subregions were analysed. Benjamini–Hochberg correction covers unique pairwise correlations (adjusted P < 0.05, adjusted P < 0.01, adjusted P < 0.001). b, Age–volume Pearson r in pooled (black, n = 458), CBA (orange, n = 256) and AMBR (blue, n = 202) analyses. n denotes independent participants. Filled circles indicate Benjamini–Hochberg-adjusted P < 0.05 across 26 correlations per analysis. Open circles indicate non-significance. Two-sided Pearson tests in a,b. c, Age (left) and female-sex (right, male reference) effects from Gaussian identity-link models of log-volume, adjusting for age, sex, cognitive and amyloid-β status, their interaction, ICV and study (n = 316 independent participants with complete covariate data, 197 female, 119 male, CBA 209, AMBR 107). Points show coefficients. Bars show unadjusted 95% confidence intervals (estimate ±1.96 × s.e.). Age effects are per 11.61 years (full-cohort s.d.). Two-sided t-tests use 308 residual degrees of freedom, with Benjamini–Hochberg correction across eight coefficients, including the intercept, per model. Colours indicate adjusted P. Supplementary Table 4 supplies estimates, intervals, t statistics and raw and adjusted P. d, Three-dimensional MRI reconstructions from three representative women. e, PCA of raw subregion volumes, coloured by age, sex, amyloid-β and cognitive status on PC1 and PC2. Each point represents one participant. f, Female age–volume scatter plots of selected subregions and the total hypothalamus (mm3, n = 277 independent participants), coloured by cognitive and/or amyloid-β status. Each point represents one participant. Black LOESS curves estimate conditional means. Shading shows pointwise 95% confidence intervals. AHA, anterior hypothalamic area; DPEH, dorsal periventricular nucleus; HYPO, hypothalamus; LH, lateral hypothalamus; MM, mammillary body; MPO, medial preoptic nucleus; NS, not significant; Pa, paraventricular nucleus; Pe, periventricular nucleus; PH, posterior hypothalamus; s.d., standard deviation; SO, supraoptic nucleus; TM, tuberomammillary nucleus; VM, ventromedial nucleus.

In a generalized linear model (GLM) controlling for biological sex, intracranial volume (ICV), cognitive status and amyloid-β pathology, lateral hypothalamus volume decreased with age (effect size/estimate ± s.e. = −0.021 ± 0.0064, adjusted P = 0.0019). The SCN (0.059 ± 0.018, adjusted P = 0.0029), ARC (0.058 ± 0.012, adjusted P = 4.2 × 10−6), paraventricular nucleus (0.14 ± 0.019, adjusted P = 1.8 × 10−11), periventricular nucleus (0.092 ± 0.019, adjusted P = 6.4 × 10−6), DMH (0.055 ± 0.016, adjusted P = 0.0011) and total hypothalamic volumes (0.022 ± 0.0071, adjusted P = 0.0048) showed a positive association with age (Fig. 1c, Extended Data Fig. 1e and Supplementary Table 4). Biological sex was also a significant factor: females showed smaller volumes in several subregions, most prominently in the supraoptic nucleus (−0.097 ± 0.019, adjusted P = 1.7 × 10−6) (Fig. 1c, Extended Data Fig. 1e and Supplementary Table 4). This aligns with previous studies emphasizing the sexually dimorphic nature of the hypothalamus, including the supraoptic nucleus2,14. As expected, most subregion volumes were positively correlated with ICV (Extended Data Fig. 1e and Supplementary Table 4). For comparison, volumes of most extrahypothalamic structures showed age-related reductions, smaller sex differences than the hypothalamus and positive correlations with ICV (Extended Data Fig. 2 and Supplementary Table 4). Representative subregions were reconstructed in three dimensions (Fig. 1d).

To holistically evaluate volumetric relationships, we performed principal component analysis (PCA) on volumetric data from all 25 delineated hypothalamic and extrahypothalamic structures (Fig. 1e and Extended Data Fig. 3a–e). The first two principal components (PCs) were primarily associated with age, sex and global scaling factors (intracranial and total brain volume), along which individuals separated (Fig. 1e and Extended Data Fig. 3a). Amyloid-β and cognitive status showed weaker associations with these components, and after residualizing for age, sex and global scaling factors, neither remained significantly associated with the leading PCs (Fig. 1e and Extended Data Fig. 3a–e). This indicates that disease status does not define an independent dominant axis of macrostructural variation, and that the more sensitive disease-associated signal may instead be captured at the molecular level.

Examination of volumetric trajectories by means of scatter plots stratified by cognitive and amyloid-β status revealed subregion-specific patterns that also differed by sex (Fig. 1f and Extended Data Fig. 3f). Notably, individuals with AD (defined as cognitive decline and amyloid-β positivity) showed reduced total hypothalamus and lateral hypothalamus volumes compared with their age-matched non-AD counterparts, consistent with the overall age-related decline of these regions (Extended Data Fig. 3g). By contrast, ARC, SCN, paraventricular nucleus, periventricular nucleus and DMH volumes were increased in individuals with AD relative to age-matched non-AD individuals (Extended Data Fig. 3g). This pattern did not reach statistical significance, potentially reflecting the insensitivity of volumetric measures to the relatively early disease stage and the limited sample size (Extended Data Fig. 3g). Subregion volumes classified cognitive status, amyloid-β status and AD better than total hypothalamic volume (Extended Data Fig. 3h,i), suggesting that subregion-specific volumetric changes capture disease-relevant variation that is obscured when the hypothalamus is treated as a single structure.

Thus, ageing is associated with subregion-specific volumetric changes that persist after controlling for amyloid-β and cognition, whereas AD-linked changes did not reach significance, motivating a complementary, higher-resolution molecular modality.

Single-cell analysis of AD hypothalamus

Volumetric data provide anatomical context but cannot reveal cell-type-specific molecular changes. We therefore performed single-nucleus RNA sequencing (snRNA-seq) on post-mortem hemi-hypothalamus tissue from 14 individuals (Fig. 2a,b and Extended Data Fig. 4a): seven with mostly early-stage AD (n = 6 sporadic, n = 1 familial) and seven age-matched, cognitively and neuropathologically normal control participants. The familial AD sample (AD_44, ID 7A in Fig. 2b; APOE4/4) was excluded from all differential analyses and retained as an extreme example of advanced pathology.

Fig. 2: Single-cell analysis of the AD hypothalamus.
Fig. 2: Single-cell analysis of the AD hypothalamus.

a, Schematic of the snRNA-seq workflow. Frozen post-mortem hypothalamic tissue from 14 individuals (7 non-AD control participants and 7 patients with AD (6 age-matched sporadic and 1 younger familial)) was macrodissected to enrich for the ARC, SCN and mammillary nucleus (MN) subregions. Nuclei were isolated, and libraries were prepared and sequenced on a NovaSeq platform. The resulting data were integrated with 133,879 publicly available nuclei from three young, healthy control participants, yielding a final dataset of 614,403 nuclei across young control, older control (non-AD) and AD groups for comparative analysis. b, Overview of the 14 individuals included in the snRNA-seq study, detailing sample IDs, clinical condition (non-AD control, early-stage AD or familial AD), age at death and Braak stage. Each individual contributed tissue from three hypothalamic subregions (ARC, SCN and MN), totalling 42 tissue samples processed across seven experimental batches. c, UMAP plot of the integrated 614,403 nuclei. Distinct clusters represent identified main cell types, including neurons, astrocytes, oligodendrocytes, microglia, endothelial cells, ependymal cells, fibroblasts, epithelial cells, OPCs and leukocytes. Each cell type is annotated and highlighted with a unique colour. d, Dot plot of selected marker genes. Dot size corresponds to the fraction of cells within a given cell type expressing the marker; colour intensity indicates the average normalized expression level. e, Normalized expression of key canonical marker genes mapped onto the UMAP plot from c. Representative markers include GFAP (astrocytes), FLT1 (endothelial cells), CFAP299 (ependymal cells), TTR (epithelial cells), C1QA (microglia), PTPRC/CD45 (leukocytes), SYT1 (neurons), MBP (oligodendrocytes), VCAN (OPCs) and COL1A1 (fibroblasts). Illustrations in a created in BioRender; Yu, D. https://biorender.com/zt57wnj (2026).

Donors with AD predominantly had sporadic, clinically confirmed cognitive decline with pathology (Braak stage IV; Fig. 2b), a stage optimal for detecting transcriptional changes while neuronal populations remain20. Hierarchical clustering grouped samples by pathology status rather than by confounders such as time of death or quality metrics (Extended Data Fig. 4a).

For annotation, we integrated our dataset with 133,879 previously published, high-quality human hypothalamic nuclei from three younger individuals (29, 42 and 50 years of age)21,22 using the single-cell variational inference algorithm23. Main cell types were identified through the expression of canonical marker genes (Fig. 2c–e and quality control in Extended Data Fig. 4b–g).

Machine learning features in HypoAD

To identify transcriptomic features distinguishing hypothalamic cells from patients with AD versus control individuals, we trained Elastic Net models on 36,600 genes from 480,524 cells. The dataset spanned seven independent batches, each containing one donor with AD and one control donor (Fig. 3a). This balanced design enabled a leave-one-batch-out cross-validation that yielded strong classification performance for both the all-cell model and cell-type-specific models for the main cell types (Fig. 3b).

Fig. 3: Machine learning models identify shared and cell-type-specific predictive features of AD.
Fig. 3: Machine learning models identify shared and cell-type-specific predictive features of AD.

a, Nested leave-one-batch-out cross-validation. Outer folds held out one batch for testing. Inner folds selected hyperparameters using the remaining batches. Each of seven nonoverlapping batches contained two distinct donors (one control, one AD): 480,524 cells × 36,600 genes. b, Receiver-operating-characteristic curves for Elastic Net models using all nuclei (all-cell model) and each of the five most abundant cell types. Area under the curve reflects pooled performance across seven outer folds. The dashed line indicates chance (area under the curve of 0.5). c, Top 20 AD-predictive genes per model ranked by absolute model weight within each outer fold. Bars show mean rank across folds, ordered from highest importance (lowest rank). Error bars show s.d. (upper side shown), and dots show individual fold ranks. Seven all-cell ranks greater than 300 appear as triangles at the axis limit. Fully coloured and pale bars indicate positive and negative weights, respectively. n = 7 outer folds per model, with overlapping training sets. Folds are resampling units, not biological replicates. Each held-out batch contains two independent donors, totalling 14 donors. Models used 480,524 nuclei (all-cell), 227,283 oligodendrocytes, 103,815 neurons, 57,437 astrocytes, 52,390 microglia and 22,004 OPCs from these 14 donors. d, Top 35 shared features. Bars show mean rank across all five cell-type-specific models, with lower ranks indicating higher importance. e, Top 15 features per cell type, selected by gene rank in the target versus other cell types. Bars show specificity scores. f, Selected significantly enriched KEGG pathways from GSEA of genes ranked by signed model weights, for all-cell and cell-type-specific models. The x axis shows absolute normalized enrichment score, dot size indicates pathway gene count and colour indicates significance (−log10[FDR]). Astro., astrocyte; AUC, area under the curve; NES, normalized enrichment score; Oligo., oligodendrocyte.

Averaging absolute gene weights across the seven outer cross-validation folds gave the top 20 features per classifier (Fig. 3c), many previously implicated in AD pathology, lending biological support to our model’s predictions. The all-cell model’s top features included arginine–vasopressin (encoded by AVP), a circadian neuropeptide elevated in AD24. Cell-type-specific models likewise highlighted AD-linked genes: MCF2L (differentially methylated in AD; oligodendrocyte)25, ARL17B (an inhibitory neuron expression quantitative trait locus elevated in AD cortex; neurons)26, CHI3L1 (astrocyte-secreted neuroinflammation marker and a cerebrospinal fluid (CSF) biomarker for early AD diagnosis; astrocytes)27, RNF150 (a cis-expression quantitative trait locus, microglia)28 and ITGA8 (linked to demyelination; oligodendrocyte precursor cell (OPC))29 (Fig. 3c).

Ranking genes by mean rank across the five most abundant cell types identified HP1BP3 as the top shared feature (Fig. 3d), a candidate modulator of hippocampal cognitive ageing affected by metformin30,31 but underexplored in the hypothalamus. A per-gene specificity rank (Fig. 3e) revealed strong enrichment of long noncoding RNAs (for example, LINC02646, LINC00551, LINC01250, LINC00937 and LINC02389) among top neuronal features (Fig. 3e), providing a cell-type-specific resource of known and new AD-associated genes. Gene-set enrichment analysis (GSEA) on the ranked predictive genes revealed a significant enrichment for circadian rhythm and inflammatory pathways (Fig. 3f, Extended Data Fig. 5 and Supplementary Table 5), connecting our cell-level classifiers to established hallmarks of neurodegeneration.

Cell-type-specific expression changes

We next performed pseudobulk differential expression analysis (edgeR likelihood-ratio test (LRT), by cell type and sample) between early-AD and control samples (Fig. 4a and Supplementary Table 6). Notably, microglia showed the largest number of differentially expressed genes (DEGs), highlighting their substantial transcriptional response even in early AD.

Fig. 4: Cell-type-specific transcriptional changes and the microglial ageing and disease trajectory in HypoAD.
Fig. 4: Cell-type-specific transcriptional changes and the microglial ageing and disease trajectory in HypoAD.

a, Significant DEGs for main cell types. One pseudobulk per donor and subregion. AD and older control (CTL) samples: main cell types, 18 and 21 (from 6 and 7 donors); ependymal, 13 and 15; epithelial, 2 and 3 (untested). DEGs (FDR < 0.05): edgeR LRTs (1 degree of freedom), Benjamini–Hochberg-adjusted (Supplementary Table 6). b, Early-AD non-neuronal Hallmark enrichment (fGSEA, Benjamini–Hochberg-adjusted P < 0.1). c, Senescence scores: young (3 donors), CTL, AD, AD_44 (1 donor). AD–CTL sample means: two-sided Wilcoxon rank-sum, Benjamini–Hochberg-adjusted across cell types, non-significant (Supplementary Table 8). Counts as in a. Violins, per-nucleus scores. Boxes, median, interquartile range (IQR), 1.5 × IQR whiskers. d, Microglial UMAP (51,420 nuclei): pseudotime (top), condition (bottom). e, Top, young, CTL and AD nuclear fractions per pseudotime bin (22 bins, 47,948 nuclei). Spearman ρ (95% CI), two-sided unadjusted P: young −0.77 (−0.90 to −0.51), 3.7 × 10−5, CTL −0.36 (−0.69 to 0.09), 0.10, AD 0.97 (0.93 to 0.99), 1.0 × 10−13. Groups 1–4, illustrative stages. Bottom, mean UCell scores per bin (51,420 nuclei), LOESS lines, 95% CI bands. f, Polynomial pseudotime fits (left), expression (right): young (3,213 nuclei), CTL (24,648), AD (20,087), AD_44 (3,472). AD–CTL FDRs (a): CSF1R 3.1 × 10−7, CX3CR1 9.7 × 10−6, P2RY12 7.1 × 10−3, AXL 9.6 × 10−3, CLEC7A 1.7 × 10−2. FDR < 0.05, FDR < 0.01, FDR < 0.001, ****FDR < 0.0001. Violins, per-nucleus expression. Boxes as in c. g, t-DEGs (13,104, Moran’s I test, q < 0.01), shown as four co-expression time-course clusters. Pseudotime bar (below plot) range as in d (0–5). h, Gene Ontology terms per cluster (Enrichr, one-sided Fisher’s exact test, Benjamini–Hochberg-adjusted P < 0.01). Size, combined score. Colour, −log10 [P].

For orthogonal validation, we leveraged a human CSF proteome dataset from healthy control participants and individuals with AD. Among the significantly differentially abundant proteins (Supplementary Table 7), five corresponded to significant DEGs across main cell types, three in endothelial cells, a primary component of the blood–brain barrier (Extended Data Fig. 6a): increased hypoxanthine-guanine phosphoribosyltransferase (HPRT) and decreased brain-derived neurotrophic factor and neuropeptide Y. These align with blood–brain barrier dysfunction as an early AD feature32, reduced brain-derived neurotrophic factor and neuropeptide Y reflecting loss of neuroprotection33,34 and elevated HPRT suggesting the purine-salvage stress response35. Microglial PGK1 upregulation indicates a glycolytic shift characteristic of pro-inflammatory microglia36,37. PTPRN points to an altered OPC state38. GSEA on microglial DEGs revealed enrichment for G2M checkpoint, mTORC1 signalling and cholesterol homeostasis. These findings suggest that in early AD, microglia may enter a complex, supportive state. The mTORC1 signature points to metabolic reprogramming, and the cholesterol homeostasis signature is consistent with enhanced cholesterol efflux linked to amyloid-β phagocytosis, together suggesting an early supportive microglial response (Fig. 4b).

Because cellular senescence contributes to ageing and neurodegeneration39, we scored three senescence signatures—Canonical senescence, Senescence responses and Senescence initiators (Fig. 4c)—alongside three overlapping NFT signatures39 (Extended Data Fig. 6b–d and Supplementary Table 8), across cell types. Microglia, leukocytes, fibroblasts and endothelial cells showed the highest scores, elevated with both ageing (control) and AD (Fig. 4c and Extended Data Fig. 6b–d), whereas neuronal NFT was most pronounced in the patient with APOE4/4 familial AD (AD_44) (Fig. 4c and Extended Data Fig. 6b–d). These findings suggest a dichotomy: senescence in non-neuronal cells may contribute to microenvironmental dysfunction in AD, whereas neuronal NFTs reflect a distinct, tau-driven pathology that escalates with disease progression or genetic susceptibility. This highlights the potential for different, cell-type-specific therapeutic targets. To ask whether these transcriptomic changes could relate to the subregion volume changes seen by magnetic resonance imaging (MRI), we examined pathways regulating cell size and tissue architecture. A curated mTOR signalling gene set (core regulators MTOR, RPTOR, RHEB, TSC1/2 and DEPTOR; effectors RPS6KB1, EIF4EBP1 and SREBF1/2) showed an increased module score in AD, most prominently in non-neuronal cells (Extended Data Fig. 6e and Supplementary Table 9). Because mTORC1 regulates cell growth and size40, this raises cellular hypertrophy as one speculative contributor to the volume increases observed at the macrostructural level. Conversely, an extracellular matrix (ECM) gene set (collagens COL1A1 and COL4A1; proteoglycans ACAN, BCAN and VCAN; glycoproteins SPARC,and TNC) showed a decreased module score in AD (Extended Data Fig. 6f and Supplementary Table 9). These shifts suggest cell-size regulation and extracellular remodelling as candidate molecular processes potentially relevant to the subregion volume changes.

Microglia trajectory in HypoAD

Microglia are central to brain ageing and neurodegeneration11,41,42. We applied trajectory inference43,44 to microglia from all conditions (young, older control participants and AD), inferring a continuous ageing-to-AD trajectory (Fig. 4d, Extended Data Fig. 7a and Supplementary Table 10). The pseudotime continuum was categorized into four stages: ageing phases of microglia from young participants (stage 1) and older control participants (stage 2), a late ageing phase enriched for older control participants (stage 3, representing a later state for these cells along the trajectory) and AD (stage 4) microglia (Fig. 4e and Extended Data Fig. 7a–e). A subset of microglia from older control individuals appeared at earlier trajectory points, whereas some microglia from young individuals persisted into later stages, suggesting that certain microglia may retain youthful properties whereas others show heightened susceptibility to early pathological changes.

We scored three microglial signatures with UCell45 across pseudotime: (1) homeostatic (CSF1R and CX3CR1); (2) stage 1 disease-associated microglia (DAM1) (APOE and B2M), which are typically activated by ageing and amyloid plaques and (3) stage 2 DAM (DAM2) (AXL and CLEC7A), which are dependent on TREM2 signalling46,47. The scores showed a transition from homeostatic towards DAM1 with a modest DAM2 rise (Fig. 4e,f), some of which were validated in the CSF proteomics data (Extended Data Fig. 7f). This observed shift from homeostatic-to-DAM states underscores the critical role of microglial functional changes in AD progression and highlights the potential for therapeutic interventions aimed at preserving homeostatic functions or preventing the full transition to detrimental DAM states.

Moran’s I test identified 13,104 trajectory-associated DEGs (t-DEGs) (Supplementary Table 11 and Extended Data Fig. 7g), smoothed with RVAgene48 and clustered (Fig. 4g). Gene Ontology biological process analysis of these t-DEGs revealed that the ageing phase shifts from cilium assembly and mitochondrial morphogenesis towards stress responses, including amyloid precursor protein catabolic processes, and the disease phase shifts from cell morphogenesis and synaptic modulation to viral response pathways and protein targeting to the endoplasmic reticulum (Fig. 4h).

To test generalization beyond our cohort and the hypothalamus, we compared against the PsychAD single-nucleus atlas of the dorsolateral prefrontal cortex49,50. AD-associated expression was directionally concordant across non-neuronal cell types and strongest in glia, whereas endothelial cells showed more region-specific changes consistent with the distinct vascular microenvironment of the hypothalamus (Extended Data Fig. 8a and Supplementary Table 12). In two independent PsychAD cohorts (ageing, 208 donors49; AD, 585 donors)50, DAM signatures rose with age and AD status, recapitulating our homeostatic-to-DAM progression (Extended Data Fig. 8b–g and Supplementary Table 13). Together, these cross-cohort, cross-region comparisons indicate that the main transcriptomic signatures identified in the HypoAD hypothalamus are reproducible in independently generated datasets.

Neuronal diversity across subregions

Neurons in the human hypothalamus show extensive cell-type diversity13,21,22. We analysed 213,469 neuronal nuclei, classifying them into three neurotransmitter-defined categories: inhibitory and/or γ-aminobutyric acid-ergic (GABAergic) (GAD2), excitatory and/or glutamatergic (SLC17A6) and histaminergic neurons (HDC) (Fig. 5a–c). Hierarchical Leiden clustering yielded five levels of clustering (5 to 254 subclusters). Anatomical origin was inferred from curated subregional markers21 (Extended Data Fig. 9a), with the dominant subregion listed first for subclusters expressing several markers (Extended Data Fig. 9a). A dendrogram related subtypes across resolutions, with nodes and leaves labelled by distinguishing markers. The 254 leaf subclusters were annotated by relative abundance within the total neuronal population, their distribution across young participants, older control participants and patients with AD, and inferred subregion (Fig. 5d).

Fig. 5: Neuronal diversity and the transcriptional signatures in HypoAD across hypothalamic subregions.
Fig. 5: Neuronal diversity and the transcriptional signatures in HypoAD across hypothalamic subregions.

a, UMAP of 213,469 hypothalamic neurons, colour-coded by main neurotransmitter category: GABAergic (GAD2-expressing; blue), glutamatergic (SLC17A6-expressing; orange) and histaminergic (HDC-expressing; green). b, Feature plots overlaid on the UMAP, illustrating the normalized expression of GAD2 (GABAergic), SLC17A6 (glutamatergic) and HDC (histaminergic) across the neuronal populations. c, UMAP of neuronal nuclei, coloured by inferred subregional identity on the basis of marker gene expression (colour key in d), highlighting the regional distribution of neuronal clusters within the hypothalamus. d, Circular dendrogram illustrating the relationships among the 254 identified hypothalamic neuron subtypes. For each subtype, the dendrogram is annotated with its inferred subregional identity (on the basis of marker genes) and its relative contribution from young participants, older control participants and patients with AD. e, Significant DEGs identified within the top 30 transcriptionally altered neuronal subtypes. DEGs are grouped by whether the subtype belongs to the GABAergic, glutamatergic or histaminergic category. f, Heatmap showing significantly enriched Reactome pathways within AD-altered GABAergic neurons. g, UMAP showing the 30 neuronal subtypes with the most significant transcriptional changes in early AD. Subtypes are coloured by their inferred hypothalamic subregional origin. h, Heatmap showing significantly enriched Reactome pathways within AD-altered neuronal subtypes. Highlighted pathways include those related to vasopressin-like receptors, GPCR signalling and immune and/or cytokine signalling. AD, patients with AD; CTL, non-AD individuals; GABA, GABAergic; GLU, glutamatergic; HIST, histaminergic; R, subregion annotations; Y, young individuals from ref. 22.

AD signatures in neuronal subtypes

The hypothalamus comprises distinct neuronal populations with specialized homeostatic roles, yet their subtype-specific contributions to AD remain poorly characterized. To address this, we performed pseudobulk differential gene expression (edgeR LRT) across the 254 neuronal subtypes, prioritizing the top 30 by DEG count, grouped by neurotransmitter category (Fig. 5e). Notably, GABAergic and/or inhibitory neurons showed the most pronounced changes, with enriched Reactome pathways for peptide ligand-binding receptors and G-protein-coupled receptor (GPCR) signalling (Fig. 5f), suggesting a potential dysregulation of neuropeptide signalling and synaptic function in the early-AD hypothalamus that may contribute to metabolic disturbances characteristic of AD pathogenesis.

The top 30 subtypes were mapped by inferred subregional identity in uniform manifold approximation and projection (UMAP) and spatial locations (Fig. 5g,h, Extended Data Fig. 9b–d and Supplementary Table 14), with other identities validated against the Visium spatial transcriptomics from nine human donors13,51 (Extended Data Fig. 10). This analysis highlighted neurons within the SCN as particularly vulnerable: a prominent SCN cluster (C254-58) revealed altered GPCR and vasopressin-related receptor pathways. Vasopressin, encoded by AVP (Extended Data Fig. 11a), is a key neuropeptide mainly synthesized and secreted by SCN neurons in a circadian manner52 and is crucial for maintaining circadian homeostasis. These findings indicate that early AD disrupts SCN neuronal function, potentially contributing to the circadian dysregulation often observed in patients.

Region, modality and species comparisons

We validated our findings in the PsychAD cohort at the neuronal level, comparing cortical and hypothalamic glutamatergic and GABAergic lineages. AD-associated changes in glutamatergic neurons were more concordant across regions than in GABAergic neurons (Extended Data Fig. 11b and Supplementary Table 15), with cross-region concordance weaker for neurons than for glia, consistent with greater regional specialization of neuronal identity.

We next asked how the molecular signal relates to MRI volume within matched subregions, computing per-modality area under the receiver-operating characteristic curve (AUROC) for AD status (amyloid-positive and cognitively impaired versus all others; subregion volume for MRI; pseudobulk PC1 for snRNA-seq; Extended Data Fig. 12a–c). The modalities were concordant for some subregions and divergent for others, with no single modality uniformly superior. The SCN was more discriminative transcriptomically than by volume, consistent with its neuronal signatures described above. Imaging and transcriptomics thus provide non-redundant, regionally specific windows into hypothalamic AD pathology and combining them offers a more complete picture than either alone.

For a controlled cross-species comparison, we profiled the hypothalamus in the 5xFAD mouse model. We performed snRNA-seq on hypothalamic tissue from 8-month-old female 5xFAD mice and age-matched wild-type controls (Extended Data Fig. 13), a relatively early stage in the disease progression. The 5xFAD hypothalamus recapitulated the main DAM marker changes and immune response enrichment (Extended Data Fig. 13 and Supplementary Table 16). Neuronal subtypes and subregions were annotated with HypoMap53 (Extended Data Fig. 14). Differential expression analysis (5xFAD versus wild-type) across subregions highlighted alterations in GPCR signalling and ligand–receptor interactions within the SCN, alongside transcriptional changes in the DMH, lateral hypothalamus, ARC and paraventricular nucleus subregions (Extended Data Fig. 14 and Supplementary Table 16).

Together, these analyses reveal selective transcriptional dysregulation in specific hypothalamic neuronal subtypes and subregions in AD, with a particular impact on GABAergic and/or inhibitory neurons and the SCN. The pronounced alterations in neuropeptide signalling (for example, vasopressin) and GPCR pathways within the SCN suggest a mechanistic link to circadian rhythm disruption, a hallmark of AD that can precede cognitive decline. These findings position the hypothalamus as a critical, vulnerable hub in AD pathogenesis, advocating for therapeutic strategies that target neuropeptide-GPCR signalling, circadian entrainment or subregion-specific resilience mechanisms to potentially mitigate preclinical AD symptoms.

Discussion

Our findings provide a multimodal view of how the hypothalamus is altered in ageing and early AD, illustrating region-specific and cell-type-specific changes that have not been observed previously. Neuroimaging studies have suggested that hypothalamic volume correlates with both ageing and cognitive decline, albeit with a relatively small effect size compared with other well-studied vulnerable brain regions54. This relative volumetric stability may reflect the region’s evolutionarily conserved nature and its critical role in survival. Here, by using a recently developed, high-resolution atlas of the human hypothalamus15, we detected specific volumetric changes in 13 distinct hypothalamic subregions that would be masked by more coarse-grained analyses with low-resolution imaging data.

Specifically, we found that although certain nuclei (for example, the lateral hypothalamus) diminish in size with age, others (for example, the SCN and ARC) can show an increase in volume among older adults, even when controlling for amyloid-β pathology and cognitive impairment. Together, these findings suggest that hypothalamic nuclei undergo differential volumetric changes with ageing, even in the absence of AD. However, in comparison to ageing and sex, differences in the volumes of hypothalamic nuclei in the context of amyloid-β deposition and cognitive impairment were more modest or absent. This could be related to lower early AD-related neurodegeneration in the hypothalamus compared with other brain regions or, instead, a poor sensitivity of volumetric analysis in detecting AD-related effects on the hypothalamus with limited sample size. In addition, compared with more vulnerable brain regions such as the hippocampus, the whole hypothalamus and other autonomic regions seem to remain stable with ageing, suggesting potential compensatory mechanisms conferring its resilience as it is evolutionarily conserved and vital for survival55.

To investigate the cellular and molecular effects of AD in individual hypothalamic nuclei, we performed snRNA-seq on post-mortem hypothalamic tissue from control donors and donors with early-stage AD, revealing key molecular features related to the inflammatory response and circadian rhythms that distinguish cells from control donors versus those from donors with AD by machine learning. Further analysis with trajectory inference and differential expression identified striking vulnerabilities in microglia, GABAergic and/or inhibitory neurons and specifically circadian rhythm and sleep-regulating neurons predicted to be from SCN and DMH. Consistent with other brain regions, microglia, which are widely recognized mediators of brain ageing and neurodegeneration2,41,42, emerged as particularly prone to transcriptional reprogramming in the hypothalamus in early AD, although our unpaired, cross-sectional data cannot establish a causal or directional relationship.

The observed alignment of volumetric and molecular vulnerability in the SCN points to the circadian circuitry of the hypothalamus as a potentially vulnerable node in ageing and AD, placing our findings within a broader literature on circadian dysfunction in AD. Circadian disruption, including fragmented sleep–wake cycles, sundowning and altered melatonin secretion, is a well-recognized clinical feature of AD that often precedes cognitive symptoms and may accelerate disease progression8, and the SCN is the central regulator that governs these rhythms. However, because the SCN is among the subregions with lower volumetric reliability (together with mammillary body, dorsal periventricular nucleus and tuberomammillary nucleus), we interpret its volumetric association with particular caution; its involvement in our conclusions rests mainly on the orthogonal molecular evidence. We further note that resolving volume changes in these small nuclei approaches the limits of the 3T MRI-based methods used here; emerging techniques such as ultra-high-field MRI may substantially improve the reliability of subregion volumetry in vivo56. Nevertheless, our study adds converging volumetric and transcriptomic evidence that the SCN and DMH are particularly affected in ageing and AD, through dysregulated vasopressin signalling in SCN neurons, microglial inflammatory remodelling and pronounced vulnerability of GABAergic neurons implicated in circadian regulation.

Returning to the volumetric findings, our molecular data may offer candidate, although speculative, explanations for the subregion volume increases we observed with age and AD. Although these changes are small, the single-cell analyses point to several cellular mechanisms that may underlie them. Main cell-type composition in the 5xFAD whole-hypothalamus mouse data was comparable between early-AD and control animals, consistent with earlier studies with limited sample sizes in humans57,58, suggesting that the observed volumetric effects are unlikely to reflect frank gain or loss of cells, at least at this disease stage. Instead, our data implicate changes in cell size and state: we observed enrichment for ECM remodelling and mTOR signalling, which regulates cell growth, alongside inflammation signatures in non-neuronal cells. A likely contributor is reactive gliosis, which develops in the ageing hypothalamus and can manifest as apparent volume increases on structural MRI even without neuronal loss, as also reported for the hypothalamus in obesity59. Together with previous reports of neuronal hypertrophy in the human hypothalamus with age60, these changes suggest a possible link between cellular remodelling and macrostructural volume. We emphasize, however, that this link is a hypothesis rather than an established mechanism: given the large spatial-scale gap between single-nucleus transcriptomic signatures and macroscopic MRI volume, and the transcriptional noise inherent to snRNA-seq, it cannot be inferred from the transcriptomic data alone and would require spatially resolved protein or histological validation. More fundamentally, fully unifying the two modalities remains challenging. Because the volumetric and molecular data derive from separate cohorts without paired, coregistered measurements in the same individuals, we did not attempt coupled statistical integration, such as multimodal embedding or the prediction of molecular state from imaging features. Instead, we compared the AD-predictive signal of each modality at matched anatomical subregions, revealing that imaging and transcriptomics provide non-redundant and regionally specific windows into hypothalamic pathology (Extended Data Fig. 12a–c). True integration, in which volumetric and molecular features are jointly modelled in a shared statistical framework, will require future studies that acquire both modalities from the same individuals, ideally with spatially coregistered sampling.

In summary, our combined volumetric and single-cell atlas of the human hypothalamus provides a high-resolution framework for elucidating how ageing and AD pathology differentially affect specific hypothalamic subregions and cell types. We highlight notable vulnerabilities in GABAergic and/or inhibitory neurons, SCN and DMH circadian regulators, and inflammatory microglial cascades, underscoring their potential roles in instigating early hypothalamic dysfunction.

Methods

Volumetric analysis of imaging data

The brain MRI scans used in this study were obtained from two separate datasets. The first was from the AMBR study16. A second dataset from the CBA study17 was added to improve rigour and reproducibility. Both studies obtained 0.8-mm isotropic T1 magnetization-prepared rapid gradient echo and T2 sampling perfection with application-optimized contrasts using different flip-angle evolution sequences for segmentation and volumetric analysis of specific hypothalamic nuclei as described in ref. 15. Although several other high-resolution brain MRI datasets exist, the AMBR and CBA studies also obtained assessments of brain amyloid-β deposition and cognitive impairment. Differences in other aspects of the AMBR and CBA studies, such as cohort characteristics, were seen as valuable to improve the robustness and generalizability of this study. Test–retest reliability of the segmentation pipeline was assessed using the ICC ((2,1); two-way random effects, absolute agreement, single measurement) in eight participants with repeat scans in the CBA study. Following ref. 61, we classified ICC < 0.40 as poor and 0.40–0.59 as fair; we note that other schemes, such as ref. 62, place the poor–moderate boundary at 0.50. For structures with poor ICC, we later corroborate the volumetric findings with a complementary molecular modality. All associations reported here are based on bilateral (combined left and right) volumes; hemisphere-specific measures, which are inherently noisier, were examined only descriptively and did not form the basis of our conclusions.

To specifically assess the effect of AD on subregion volumes, participants were categorized on the basis of their amyloid-β and cognitive status. Amyloid status was determined from amyloid positron emission tomography imaging or CSF biomarker assessments when available using study-specific criteria: in AMBR, labelling was based on Knight Alzheimer Disease Research Center (ADRC) cut-offs for [carbon-11]-Pittsburgh compound B (C-11 PiB) standardized uptake value ratio (SUVR), [fluorine-18]-florbetapir (F-18 AV-45) SUVR, by visual read when SUVR data were not available, or by CSF test results. In the CBA study, labelling was based on provided centiloid values thresholded at 16.4, corresponding to the Knight ADRC C-11 PiB mean cortical SUVR cut-off of 1.42 (ref. 63). If amyloid-β status was positive any time before or up to 1 year after the MRI, it was labelled as positive. Conversely if amyloid-β status was negative any time after or up to 1 year before the MRI, it was labelled as negative. Otherwise amyloid-β status was considered not available (NA).

Cognitive status was labelled as ‘impaired’ or ‘unimpaired’ using study-specific criteria. In the AMBR study, labelling was based on the Clinical Dementia Rating scale (CDR), with CDR = 0 indicating cognitively unimpaired and CDR > 0 indicating cognitive impairment. A subject was labelled as impaired only if assessed as impaired within ±1 year of the MRI, and as unimpaired only if assessed as unimpaired at any time after the MRI or within 1 year before (otherwise coded as NA). In the CBA study, labelling was based on the timepoint-specific diagnosis code, with code 1 indicating cognitively unimpaired; codes 2, 4, 5 and 6 indicating evidence of cognitive impairment and code 3 (subjective impairment) coded as NA.

In both studies, individuals meeting criteria for both cognitive impairment and positive amyloid-β status were designated as the AD group. Conversely, participants without cognitive deficits and with negative amyloid-β status were classified as the non-AD (control) group. To minimize potential confounding from intermediate pathology profiles, only individuals clearly falling into these AD or non-AD categories were included in these specific comparative analyses. The cohort had a mean age of 63.4 ± 11.6 years (mean ± standard deviation) and included 60.5% women. Moreover, our dataset includes information directly relevant to AD, including 30.6% with abnormal cognitive status and 13.8% who were amyloid-β positive (Extended Data Fig. 1a). The mean total hypothalamic volume for this cohort was 736.2 ± 88.2 mm3.

Given that the dataset did not have predefined group structures for all desired comparisons, we used a GLM to assess the associations between hypothalamic subregion volumes (summed across left and right hemispheres) and various demographic and clinical variables.

Specifically, in the AMBR study (n = 202), we modelled (denoted as ‘~’) the natural logarithm of each subregion’s volume (log[subregion volume]) as a function of scaled age at MRI, sex, cognitive (cogn.) status, amyloid-β deposition status and other relevant covariates, where ‘+’ denotes additive main effects, ‘:’ denotes an interaction, ‘×’ denotes crossed terms (main effects plus their interaction), and ‘scale()’ denotes z-score standardization of a covariate:

P values for individual variables from the GLM analyses were corrected for multiple comparisons using the Benjamini–Hochberg procedure to control the false discovery rate (FDR).

In the CBA study (n = 256), the metabolic and sleep covariates (body mass index (BMI)), HgbA1c and Epworth Sleepiness Score) were not available; we therefore modelled the natural logarithm of each subregion’s volume (log[subregion volume]) as a function of scaled age at MRI, sex, cognitive status, amyloid-β deposition status, their interaction and ICV:

In the harmonized two-study pooled cohort (AMBR + CBA, n = 458), we modelled the natural logarithm of each subregion’s volume as a function of scaled age at MRI, sex, cognitive status, amyloid-β deposition status, their interaction, ICV and study of origin. The metabolic and sleep covariates above were not included here as they were not available across both studies:

To test whether the association between age and subregion volume differed by disease status, we fitted a separate GLM that included an age × group interaction term, where group distinguished cognitively normal, amyloid-negative individuals from all others:

To compare subregion volumes between patients with AD and non-AD individuals while balancing demographic differences, we constructed a logistic regression model to estimate each participant’s propensity score for AD-group membership, using sex and age at MRI as covariates. Propensity-score matching was then performed using a nearest-neighbour algorithm, generating matched pairs of patients with AD and non-AD individuals with similar distributions of sex and age. This balanced these key covariates between groups for the subsequent volumetric comparisons.

To evaluate the predictive power of hypothalamic volumes in identifying AD and cognitive status, we conducted a series of binary classification tasks using receiver-operating-characteristic analysis. Three primary target conditions were defined: (1) cognitive status (impaired ‘A’ versus normal ‘N’), (2) amyloid status (positive ‘P’ versus negative ‘N’) and (3) a composite AD group (A+P+ versus all others). For each condition, we compared the diagnostic performance of two distinct feature sets: a multivariate model comprising 13 individual hypothalamic subregions (ARC, anterior hypothalamic area, dorsal periventricular nucleus, DMH, lateral hypothalamus, medial preoptic nucleusMPO, paraventricular nucleus, posterior hypothalamus, periventricular nucleus, SCN, supraoptic nucleus, tuberomammillary nucleus, ventromedial nucleus) and a univariate model using only the total hypothalamic volume. Classification was performed using logistic regression with balanced class weights to account for potential group size discrepancies. Model performance was validated using a fivefold stratified cross-validation approach. Features were standardized (z-score normalized) within each training fold before model fitting. The resulting area under the curve was calculated as the mean across cross-validation folds, with the standard deviation reported as a measure of model stability.

Given the central role of the hypothalamus in energy homeostasis and metabolic regulation, we examined whether BMI was associated with hypothalamic volume at the total and subregion level. BMI was not significantly associated with total hypothalamic volume (estimate ± s.e. = −0.005 ± 0.007, adjusted P = 0.505, Extended Data Fig. 2e and Supplementary Table 4) or with any bilateral subregion volume after FDR correction (all adjusted P > 0.086). In lateralized analyses, the left DMH showed a significant negative association with BMI (left DMH −0.042 ± 0.014, adjusted P = 0.014), although this was not observed on the right side (right DMH −0.010 ± 0.018, adjusted P = 0.664, Extended Data Fig. 2e and Supplementary Table 4).

Post-mortem human brain tissues

Unfixed, flash-frozen half-hypothalamus tissues were obtained from the Netherlands Brain Bank (NBB), Netherlands Institute for Neuroscience, Amsterdam (open access at www.brainbank.nl). All material has been collected from donors for or from whom a written informed consent for a brain autopsy and the use of the material and clinical information for research purposes had been obtained by the NBB.

This consent procedure was approved by the Ethical Committee of the VU University Medical Center (Vrije Universiteit, Amsterdam, the Netherlands), independently and before this study. Tissue blocks were stored at −80 °C.

The cohort comprised 14 donors: 7 control participants without dementia and 7 patients with AD (6 sporadic and 1 familial). Comorbidities were common in both groups and predominantly cardiovascular; control participants also included donors with malignancy (colorectal carcinoma, multiple myeloma, melanoma), whereas the AD group included one donor each with chronic obstructive pulmonary disease and bladder carcinoma. Causes of death included acute cardiac events and pulmonary embolism (control participants), euthanasia (2 control participants), end-stage malignancy (control participants) and respiratory insufficiency, aspiration pneumonia, cachexia, ileus and dehydration due to swallowing problems (AD patients). No donor had a documented intensive care unit admission; all deaths occurred in community, nursing home or palliative care settings. All key determinants of post-mortem gene expression quality (age, post-mortem interval, brain pH, RNA integrity number and agonal state estimated from clinical records) were comparable between groups (all P > 0.1 with Mann–Whitney U test). All samples showed high quality, with short post-mortem intervals of less than 9 h and RNA integrity number scores above 6.5. Full donor-level clinical and neuropathological details can be requested from the NBB using the NBB ID, which is available on request.

To overcome the input cell number limits of droplet-based single-cell sequencing (which typically captures tens of thousands of cells per sample) relative to the vast cellularity of the entire hypothalamus, we performed targeted macrodissection. In this initial study, we prioritized in-depth profiling of three subregions already implicated in ageing and AD over a shallower survey of the whole hypothalamus: the ARC, a hub for energy homeostasis housing pro-opiomelanocortin and agouti-related peptide neurons that integrate leptin and insulin signals64,65; the SCN, the master circadian pacemaker whose disruption underlies sleep–wake disturbances among the earliest non-cognitive symptoms of AD66,67; and the mammillary nucleus, a component of the Papez memory circuit disproportionately vulnerable to neurodegeneration in AD models10. Together, these subregions span the anterior–posterior extent of the hypothalamus and represent three of its main functional domains: metabolic control, circadian regulation, and memory circuitry. Focusing on three primary regions of interest, the ARC, SCN and mammillary nucleus, prevented the severe dilution of rare neuronal populations, ensuring sufficient sampling depth for downstream differential expression analysis. Macrodissection was performed on hemisected hypothalami within a Leica cryostat at the lowest temperature setting, with the tissue surrounded by dry ice. To maintain RNA integrity and prevent cross-contamination, the blade and all dissection tools were sanitized and treated with RNaseZap RNase Decontamination Solution between each sample.

Animals

Wild-type and heterozygous (8-month-old) 5xFAD-C57BL/6 female mice were obtained from The Jackson Laboratory. Mice were housed and used according to protocols approved by Brown University Institutional Animal Care and Use Committee and in accordance with institutional and national guidelines. Animals were housed at 21 ± 1 °C (70 ± 2 °F) with humidity from 50% to 70%. Animals were fed ad libitum Lab Diet 5010 chow and water. The light cycle was 12 h on 12 h off and lights were on between 7:00 and 19:00. Animals were exposed to male bedding 3 d before being euthanized to synchronize the oestrous cycle. Animals were euthanized at Zeitgeber time ZT2–ZT3. No statistical methods were used to predetermine sample size. Animals were not randomly allocated to experimental groups. Investigators were not blinded to group allocation during the experiments or outcome assessment.

Statistics and reproducibility

No statistical methods were used to predetermine sample sizes, but our donor numbers are comparable to those of established snRNA-seq studies of the human brain in AD, which typically profiled 3–10 donors per group68,69,70. Individual nuclei were excluded from analysis on the basis of quality control metrics (feature count and mitochondrial read count). Data collection and analysis were not performed blind to the conditions of the experiments.

snRNA-seq

Nuclei extraction was performed using the Nuclei PURE Prep kit (Millipore Sigma) according to the manufacturer’s instructions. All procedures were performed in the cold room before the 45-min spin down. Tissue was transferred using a transfer pipette into a refrigerated Dounce homogenizer with 5 ml of lysis solution following kit instructions. Tissue was homogenized with the Dounce B and the lysate was transferred into a 15-ml Falcon tube through a 40-μm filter. The sucrose purification step was performed with the modifications that half the volume of all reagents was used to account for the small tissue sample sizes, an SW34 rotor was used and samples were spun for 45 min at 30,000g (13,000 r.p.m.) at 4 °C.

Nuclei were counted using a hemocytometer and 10,000 nuclei per sample were loaded onto the Chromium Single Cell 3′ Chip (10x Genomics) and processed with the Chromium Controller (10x Genomics). Samples were sequenced at GENEWIZ on an Illumina NovaSeq, with a target of 50,000 reads per sample.

Quality control, data processing and analysis

Sequence alignment for human snRNA-seq data (including the dataset from ref. 22) was performed against the GRCh38 (2020) reference genome. These alignments were executed using Cell Ranger (10x Genomics; v.7.1.0 for human) with the --include-introns flag, essential for capturing nuclear RNA.

Following alignment, systematic background noise, such as ambient RNA, was removed from the count matrices using CellBender (v.0.3.0)71. For quality control, outlier cells were identified on the basis of established criteria. Specifically, cells were flagged as outliers if they exceeded 5 median absolute deviations for any of the following quality control metrics: log1p_total_counts, log1p_n_genes_by_counts or pct_counts_in_top_20_genes. In addition, cells with a mitochondrial transcript percentage of 10% or more were excluded.

Potential doublets (two or more cells captured as a single droplet) were then identified and filtered out using scDblFinder72. After these quality control steps, data from different individuals were integrated using single-cell variational inference23 to correct for batch effects. Subsequent data visualization and downstream analyses were performed using tools from the Seurat73 and Scanpy74 ecosystems.

Subregional samples were never pooled. Each donor contributed three tissue samples obtained by macrodissection enriching for distinct hypothalamic subregions (ARC, SCN, MN), yielding 42 samples across the 14 older donors. Each sample was collected into a separate tube and processed through an independent 10x Chromium channel, producing independent library preparations and therefore constituting a biologically and technically independent observation capturing distinct cellular compositions and independent sources of technical variation.

Differential expression analysis was performed using a pseudobulk approach with the LRT in edgeR75. Genes were significant if the Benjamini–Hochberg-adjusted P value was below 0.05. The three young reference donors22 were included only to anchor cell-type annotation and the microglial ageing trajectory; all AD-versus-control differentially expression analysis was restricted to the HypoAD cohort, excluding the familial AD (AD_44) donor.

GSEA was performed using the fgsea package76 with the hallmark gene-set lists and Kyoto Encyclopedia of Genes and Genomes (KEGG) and Reactome gene-set list from Molecular Signatures Database76,77. For GSEA input, genes within each cluster were ranked by their log2 fold change (log2[FC]) values, determined from edgeR LRT differentially expressed analysis. The GSEA was then performed using the fgseaMultilevel() command with default settings. Gene sets were considered significantly enriched if the adjusted P < 0.1.

Embedding analysis with UCE

To contextualize our data within a broader reference, we used a universal cell embedding (UCE) foundation model. UCE projects single-cell gene expression profiles into a universal embedding space that is agnostic to technical covariates such as batch effects or protocol differences78. We implemented the 33-layer version of UCE as specified in its official repository (https://github.com/snap-stanford/UCE), using Python (v.3.9.0) and PyTorch (v.2.1.1), with the publicly available pretrained model weights.

We then used UCE to project our control and AD snRNA-seq datasets into a shared embedding space alongside a publicly available, community-curated human brain atlas containing more than 25 million cells78. The resulting embeddings were visualized in a two-dimensional UMAP space to benchmark our newly generated data against this large-scale public reference.

Model training and feature ranking

To identify predictive features for the classification task distinguishing cells from AD versus control individuals, we developed and applied a machine learning pipeline using Python (v.3.10.9) and the scikit-learn library (v.1.6.1). The full pipeline is available through GitHub (https://github.com/Webb-Laboratory/HypoAD).

Data subsetting

For cell-type-specific models, the full dataset was filtered to include only cells corresponding to a single target cell type (for example, Neuron, Astrocyte), on the basis of predefined annotations.

Nested cross-validation

We used a leave-one-batch-out nested cross-validation strategy to ensure robust hyperparameter tuning, prevent data leakage and give an unbiased performance evaluation. For each of the seven outer folds, an inner sixfold cross-validation was used to tune the model’s hyperparameters. This approach generated seven distinct models, from which we ranked gene-associated weights to identify the most predictive features (Fig. 3a). Specifically, the data for all cells, as well as for each of the top five most abundant cell types, were split into seven outer folds on the basis of the sample batch of origin (leave-one-batch-out). Each batch comprised two donors (one control and one AD), and no donor appeared in more than one batch. For each outer fold, the test set consisted of nuclei from a single held-out batch, such that both donors in that batch were entirely unseen during training. The remaining six batches were used for an inner leave-one-batch-out cross-validation to select the optimal hyperparameters for an Elastic Net logistic regression model. The 5 × 5 hyperparameter grid included a logarithmic search for the regularization parameter C (from 10−2 to 102) and a linear search for the L1 ratio (from 0 to 1). The model with the best average inner-fold AUROC was then retrained on all six training batches and evaluated on the held-out test batch. We first trained a model on all cells, which achieved predictive performance (average AUROC = 0.75). We then trained models specific to the five most abundant cell types. All cell-type-specific models also performed well, with the microglia model showing the highest accuracy (average AUROC = 0.91). The comparatively lower performance for neuronal models may be attributable to the complexity and heterogeneity of neurons in the hypothalamus (Fig. 3b).

Feature weight extraction and ranking

After completing the outer cross-validation, we extracted the feature weights (coefficients) from the seven models, each trained on one outer fold. Two feature ranking lists were generated for each cell type. The ‘raw-weight’ rank was created by ranking all features on the basis of their mean weight across the seven folds, preserving the sign to indicate positive or negative prediction association. The ‘absolute-weight’ rank was created by ranking features on the basis of the mean of their absolute weights to measure overall predictive importance regardless of direction.

Shared and cell-type-specific features

To identify features consistently important across all cell types, we defined a ‘shared score’ for each gene as the mean of its absolute-weight rank across the five cell types analysed, and coloured by its direction (that is, whether the feature is on the top or bottom of the raw-weight rank list). Cell-type-specific ranks for features were defined by inquiry_rank/mean_rank_others.

GSEA

To identify biological pathways associated with our model’s predictions, we performed preranked GSEA using the gseapy (v.1.1.9) Python package. For each cell type, the full list of features was ranked using the mean raw weight. This ranked list was tested for enrichment against the KEGG pathway database (v.2021), Hallmark (v.2020) and Gene Ontology Biological Process (2021) from the Molecular Signatures Database. Significance was determined using a permutation-based test (1,000 permutations), and pathways with FDR q < 0.05 were considered significantly enriched.

Trajectory analysis using Monocle3

To infer the ageing process for the microglia cluster (n = 51,420 nuclei) generated in Seurat, we applied Monocle3 (refs. 43,44). Monocle3 typically uses dimensionality reduction to position cells in a lower-dimensional space, can correct for batch effects using mutual nearest-neighbour alignment and subsequently connects cells to construct trajectories in a semi-supervised manner.

For our analysis of the microglia–macrophage cluster, we directly used the integrated Seurat object, bypassing any further batch correction or dimensionality reduction steps within Monocle3 itself, as these had been addressed in the previous Seurat processing. After subsetting the microglia cluster, the root of the trajectory was programmatically defined by selecting the principal graph node most enriched for cells originating from young individuals. The horizontal axis in Fig. 4e represents a data-driven transcriptomic pseudotime learned from the single-cell data, not a chronological timeline, and individual donors do not necessarily follow this ordering. Trajectory-based expression analysis along this inferred trajectory was then performed using Moran’s I test, as implemented in Monocle3. Genes with a q value (FDR-adjusted P value) of less than 0.01 were considered significantly trajectory dependent, resulting in the identification of 13,104 such genes.

Functional enrichment analysis

Functional enrichment analysis of the 13,104 t-DEGs, grouped into individual expression groups identified through co-expression analysis, was performed using Enrichr (v.3.0)79. This analysis used the GO_Biological_Process_2018 gene-set database. Statistically significantly enriched Gene Ontology terms were identified on the basis of a Benjamini–Hochberg-adjusted P value (Padj) < 0.01. For visualization purposes, the top ten significantly enriched terms per group were shown in a dot plot.

To decode and visualize the dynamic expression patterns of the t-DEGs (n = 13,104) along the inferred pseudotime trajectory, we used the RVAgene package (v.1.0)48 in Python (v.3.9.6) with PyTorch (v.1.9.0). Before input into RVAgene, expression values for each gene were averaged within discrete pseudotime bins and subsequently rescaled to a range of [−1, 1]. The RVAgene neural network was configured with a symmetrical architecture comprising two hidden layers (48 nodes per layer) and two latent variable dimensions. The smoothed, reconstructed expression trajectories of the t-DEGs generated by RVAgene were then used to create the summary heatmap.

mTOR and ECM module scores

Two gene sets were assembled to score mTOR(PI3K-AKT-mTOR) signalling and ECM (matrisome) activity. Each set began from a curated list of canonical genes: for mTOR, core pathway components and downstream lipid–sterol effectors; for ECM, matrisome constituents, including collagens, laminins, proteoglycans, matrix metalloproteinases and their inhibitors (MMPs or TIMPs) and ADAMTS proteases. Each seed list was expanded by gene-symbol pattern matching (for example, COL[0–9] + A[0–9], MMP[0–9], LAM[ABC][0–9]) and filtered to remove unrelated ‘look-alike’ families (for example, mitochondrial import genes TIMM/TOMM, TIMELESS), then restricted to genes detected in the dataset, yielding 70 mTOR and 157 ECM genes (Supplementary Table 9, tabs mTOR_set and ECM_set).

Counts were normalized with Seurat’s NormalizeData (log-normalization, scale factor 10,000). Per-cell module scores were computed with Seurat’s AddModuleScore (default parameters: 24 expression bins, control pools of 100 randomly sampled genes per set gene), which subtracts the mean expression of a randomly selected control gene set from the mean expression of the target set. mTOR and ECM scores were calculated jointly (Score1 = mTOR, Score2 = ECM). For statistical testing, donors were grouped as young, control or AD. To avoid pseudoreplication, per-cell scores were aggregated to the sample level by taking the mean score across all cells of a given cell type within each sample, retaining only sample × cell-type combinations with ten or more cells. For each cell type and gene set, module scores were compared between control and AD samples using a two-sided Wilcoxon rank-sum test, and P values were adjusted for multiple comparisons across all cell type × gene-set tests using the Benjamini–Hochberg procedure. Analyses were performed in R with Seurat and rstatix.

Spatial mapping with cell2location

Cell2location is a Bayesian model that uses snRNA-seq cell-type signatures to infer cell types in Visium spatial transcriptomics by decomposing messenger RNA counts in each Visium voxel into cell types. We performed the three main steps in the cell2location workflow: estimating reference expression signatures of cell types using our dataset, mapping the learned cell-type signatures onto the slides, and performing downstream analysis. The code, model parameters and training evaluations can be found in the Jupyter notebooks in our GitHub repository. In brief, we used the default parameters to train the cell2location model.

Proteomic validation

Lumbar CSF samples from a cohort of 280 participants (including healthy control, AD and other conditions) were analysed by means of liquid chromatography–tandem mass spectrometry. Samples were processed on a Thermo EASY-nLC 1200 system coupled to a Thermo Orbitrap Eclipse Tribrid mass spectrometer. Detailed chromatography and data acquisition parameters follow the methods described in ref. 80. Quality control was maintained using inter-batch and inter-experiment samples across 96-well plates to evaluate and correct for technical effects. Protein quantification was performed by mapping raw spectra to an empirically corrected Prosit81 library (generated by means of gas-phase fractionated, narrow-window injections) using EncyclopeDIA, with peak integration and protein mapping finalized in Skyline.

Differential abundance analysis was performed to compare patients with AD (n = 19) against the healthy control group (n = 205) using the limma and eBayes framework82 on the preprocessed and log2 normalized protein quantification table in R. Statistical significance for differential protein abundance was defined by Padj < 0.05.

To validate cell-type-specific transcriptomic signatures at the protein level, significant CSF proteins were intersected with significant DEGs from single-cell pseudobulk analysis (Padj < 0.05). To address nomenclature inconsistencies between modalities, a ‘fuzzy-join’ logic was implemented. This approach matched protein symbols to RNA symbols on the basis of exact matches, prefix alignment and antisense transcript identification.

Validation of nine ageing-related and AD-related markers from the microglia pseudotime analysis (CSF1R, CX3CR1, P2RY12, APOE, B2M, CTSB, AXL, CLEC7A and TIMP2 in Fig. 4f) was performed by extracting their corresponding features from the CSF proteomics expression matrix using case-insensitive substring matching. To isolate age-related versus disease-related protein trajectories, the healthy control cohort was further stratified into ‘young’ (less than 65 years of age) and ‘aged’ (65 years or older) subgroups and visualized accordingly. Statistical analysis did not show significant differences between the groups, despite the protein trajectories’ showing trends similar to the gene expression data.

External validation in the PsychAD cohort

AD-associated transcriptomic signatures were validated against the PsychAD Consortium single-nucleus RNA-seq resource of the human dorsolateral prefrontal cortex (Mount Sinai Brain Bank cohort), obtained from CELLxGENE Discover (https://cellxgene.cziscience.com/collections/84ce6837-548d-4a1f-919f-0bc0d9a3952f). Two releases were used: an AD-enriched cohort for disease comparisons and a neurologically normal cohort spanning adulthood for the ageing analysis. For all analyses, Ensembl gene IDs were mapped to HUGO Gene Nomenclature Committee symbols using the feature_name field supplied with the dataset, and raw counts were normalized to 10,000 counts per cell and log1p-transformed.

For cross-cohort differential expression concordance, PsychAD cell types were mapped to their HypoAD counterparts at the main cell-type level. Non-neuronal populations (astrocytes, microglia, oligodendrocytes, OPCs and endothelial cells) were assigned direct one-to-one correspondences. Because neuronal subtypes are largely region-specific and the cortical subtypes annotated in PsychAD do not correspond one-to-one to hypothalamic populations, the finer PsychAD neuronal subtypes were aggregated into glutamatergic and GABAergic (GABA) lineages; for each gene, the subtype result with the lowest FDR was taken as the representative statistic for the lineage. Pseudobulk differential expression (AD versus control) was computed per cell type with edgeR and compared across datasets by plotting the paired log2[FC] of shared genes. Directional concordance of AD-associated changes was quantified with Spearman’s rank correlation (ρ) on the subset of genes significantly differentially expressed (FDR < 0.05) in both datasets.

For microglial DAM module scoring, module scores for homeostatic microglia (HEXB, CST3, CX3CR1, CSF1R, CTSS, TMSB4X, P2RY12, C1QB), DAM1 (TYROBP, CTSB, APOE, B2M, FTH1, TREM2) and DAM2 (TREM2, AXL, CST7, CTSL, LPL, CD9, CSF1, ITGAX, CLEC7A, LILRB4, TIMP2) were computed with scanpy.tl.score_genes (analogous to Seurat’s AddModuleScore). In the ageing cohort, scores were averaged per donor and then per age bin, correlated with age by Spearman’s ρ and shown with locally weighted scatterplot smoothing (bandwidth 0.6); individual representative genes were shown as scatter plots against donor age (locally weighted scatterplot smoothing bandwidth 0.5) and as violin plots comparing donors less than 65 years of age versus those 65 years of age or older. In the AD cohort, the same gene-level visualizations were produced across the four donor groups (young, control, AD and familial AD).

Ethics statement

AMBR-MRIs were obtained from studies involving positron emission tomography imaging that were approved by the Radioactive Drug Research Committee and the Institutional Review Board at Washington University in St. Louis. All participants provided written consent. CBA-MRI studies involving human participants were reviewed and approved by the Human Research Protection Office at the University of Pittsburgh. All participants provided written informed consent to participate17. All procedures and research protocols for post-mortem tissue were approved by the Institutional Review Board at Brown University. All research was performed in accordance with relevant guidelines and regulations. Informed consent for the use of post-mortem tissue was obtained from all donors or their next of kin.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

The text above was extracted automatically from the source page and may miss images, tables or code blocks. The original is authoritative.

Published by Nature
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗