NUCLEAR LAMINA EROSION-INDUCED RESURRECTION OF ENDOGENOUS RETROVIRUSES UNDERLIES NEURONAL AGING

Bạn đang xem bản rút gọn của tài liệu. Xem và tải ngay bản đầy đủ của tài liệu tại đây (12.81 MB, 31 trang )

Trang 1<div class="page_container" data-page="1">

Nuclear lamina erosion-induced resurrection of endogenous retroviruses underlies neuronal aging Graphical abstract

d

Multi-omics profiling of primate frontal lobe (FL) aging

d

Neuronal-specific nuclear lamina erosion as a hallmark and driver of FL aging

d

Consequent ERV activation induces neuronal senescence and inflammation

d

The NRT inhibitor mitigates human neuronal senescence and mouse brain aging

Hui Zhang, Jiaming Li, Yang Yu, ..., Jing Qu, Weiqi Zhang, Guang-Hui Liu

Zhang et al. establish a granular view of the molecular alterations underlying primate FL aging through multi-omics profiling, neuropathological examination, andin vitro modeling. It reveals that, during neuronal aging, nuclear lamina erosion induces the reactivation of ERVs and contributes to neuronal

Zhang et al., 2023, Cell Reports42, 112593 June 27, 2023ª 2023 The Authors.

</div>Trang 2<div class="page_container" data-page="2">

Nuclear lamina erosion-induced

resurrection of endogenous retroviruses underlies neuronal aging

Hui Zhang,1,6,29Jiaming Li,3,6,29Yang Yu,11,12,29Jie Ren,3,4,6,29Qiang Liu,13,29Zhaoshi Bao,15,16,29Shuhui Sun,1,4,7

Xiaoqian Liu,2,4,7Shuai Ma,1,4,7Zunpeng Liu,2,6Kaowen Yan,1,4,7Zeming Wu,1,4,7Yanling Fan,3,6Xiaoyan Sun,3,6

Yixin Zhang,2,6Qianzhao Ji,1,6Fang Cheng,6,17Peng-Hu Wei,20,21,22Xibo Ma,21,22Shiqiang Zhang,23

(Author list continued on next page)

The primate frontal lobe (FL) is sensitive to related neurocognitive decline. However, the aging-associated molecular mechanisms remain unclear. Here, using physiologically aged non-human pri-mates (NHPs), we depicted a comprehensive landscape of FL aging with multidimensional profiling en-compassing bulk and single-nucleus transcriptomes, quantitative proteome, and DNA methylome. Conjoint analysis across these molecular and neuropathological layers underscores nuclear lamina and heterochromatin erosion, resurrection of endogenous retroviruses (ERVs), activated pro-inflamma-tory cyclic GMP-AMP synthase (cGAS) signaling, and cellular senescence in post-mitotic neurons of aged NHP and human FL. Using human embryonic stem-cell-derived neurons recapitulating cellular agingin vitro, we verified the loss of B-type lamins inducing resurrection of ERVs as an initiating event of the aging-bound cascade in post-mitotic neurons. Of significance, these aging-related cellular and molecular changes can be alleviated by abacavir, a nucleoside reverse transcriptase inhibitor, either through direct treatment of senescent human neuronsin vitro or oral administration to aged mice.

The frontal lobe (FL) of the primate brain evolved to control ex-ecutive functions and cognitive skills.1 The FL is one of the brain regions with the most decreased volume with aging,2,3 and its neuroanatomic and neurophysiological changes un-derlie the development of frontotemporal dementia and Alz-heimer’s disease (AD).4However, cognitive aging presumably occurs years before the onset of neurodegenerative diseases is diagnosed in elderly individuals, presenting challenges to early diagnosis and therapeutic development.5,6 Therefore, an in-depth understanding of the cellular and molecular changes associated with FL aging may help uncover check-points that can be targeted therapeutically, either at earlier

stages or preemptively, to delay the progression of neurocog-nitive decline.

In the gray matter (GM) of FL, as well as embedded within the white matter (WM), billions of highly interconnected and func-tionally diverse neurons and large populations of glial and other non-neuronal cells constitute a vastly heterogeneous cell popu-lation.7–10Moreover, a growing number of studies have reported that extensive and complex cellular structural and functional al-terations are involved in FL aging and neuronal degenera-tion.11,12Coupled with the diversity of regulatory factors involved at the epigenetic, transcriptional, and translational levels, our un-derstanding of the cellular and molecular drivers of FL aging, especially in primates, remains very limited. To this end, recently developed single-nucleus RNA sequencing (snRNA-seq) makes

1State Key Laboratory of Membrane Biology, Institute of Zoology, Chinese Academy of Sciences, Beijing 100101, China

2State Key Laboratory of Stem Cell and Reproductive Biology, Institute of Zoology, Chinese Academy of Sciences, Beijing 100101, China 3CAS Key Laboratory of Genomic and Precision Medicine, Beijing Institute of Genomics, Chinese Academy of Sciences and China National Center for Bioinformation, Beijing 100101, China

4Institute for Stem Cell and Regeneration, CAS, Beijing 100101, China

5Advanced Innovation Center for Human Brain Protection, and National Clinical Research Center for Geriatric Disorders, Xuanwu Hospital Capital Medical University, Beijing 100053, China

6University of Chinese Academy of Sciences, Beijing 100049, China

7Beijing Institute for Stem Cell and Regenerative Medicine, Beijing 100101, China

8Aging Translational Medicine Center, International Center for Aging and Cancer, Beijing Municipal Geriatric Medical Research Center,

(Affiliations continued on next page)

</div>Trang 3<div class="page_container" data-page="3">

it possible to study the cellular and molecular mechanisms of heterogeneous FL aging with high accuracy,13–17and, combined with multi-layered omics approaches,18it will serve as a valuable resource to thoroughly delineate the intricate regulatory mecha-nism of FL aging.

Similar to humans, the non-human primate (NHP) frontal cor-tex also increases in size and evolves into an extremely elaborate neocortical region during phylogenetic development.19,20 In addition, NHPs are highly similar to humans in neuroanatomical, physiological, and neuropathological aspects and experience neuronal deterioration and cognitive impairment in later life, similar to humans,15,21–23thus serving as a clinically relevant model to investigate aging-associated mechanisms underlying FL dysfunction.

Here, we revealed that nuclear lamina attrition, heterochro-matin erosion, and consequent resurrection of ERVs are intrinsic to aged neurons, thereby initiating innate immune responses and ultimately neuron degeneration in aged NHP and human FL

in vivo, as well as in an in vitro model. Notably, we demonstrated

that pharmacological treatment with the nucleoside reverse tran-scriptase (NRT) inhibitor abacavir, either directly supplemented to the human neuron model or orally administered to physiolog-ically aged mice over a 12-month time course, inhibits neuronal senility.

Multifaceted phenotypes of neurological degeneration in the aged primate FL

We first obtained FL tissues from young (4–6 years old, equiva-lent to 16–20 years of human age) and aged (18–21 years old, equivalent to65–70 years of human age) cynomol-gus monkeys without apparent morphological anomalies (Figures 1A and S1A). The overall anatomical structures and the integrity of cortical stratification in the aged FL were compa-rable with those in the young counterparts (Figures S1B and S1C).

However, distinct from their younger counterparts, we found that a spectrum of neurological degeneration indicators accu-mulated in the aged FL, especially in the GM of the FL. First, in the aged FL, we observed increased areas with senescence-associated b-galactosidase (SA-b-Gal) staining, a classic senes-cence marker24(Figure 1B). Consistent with earlier work docu-menting the loss of proteostasis in the aged cortex,25,26we found increased protein aggregates and amyloid-b (Ab) de-posits, marked by Ab (4G8), the major component of senile pla-ques, in the GM of aged FL (Figures 1C and 1D). In addition, we detected increased lipofuscin pigment in aged FL, which usually accumulates progressively with age (Figure 1E). Consistent with

Zhengwei Xie,24Yuyu Niu,25,28Yan-Jiang Wang,26,27,28Jing-Dong J. Han,23,28Tao Jiang,14,15,16,28

Guoguang Zhao,18,19,20,28Weizhi Ji,25,28Juan Carlos Izpisua Belmonte,10,28Si Wang,5,8,9,*Jing Qu,2,4,6,7,*

Weiqi Zhang,3,4,6,*and Guang-Hui Liu1,4,5,6,7,8,30,*

Xuanwu Hospital, Capital Medical University, Beijing 100053, China 9The Fifth People’s Hospital of Chongqing, Chongqing 400062, China 10Altos Labs, Inc., San Diego, CA, USA

11Beijing Key Laboratory of Reproductive Endocrinology and Assisted Reproductive Technology and Key Laboratory of Assisted

Reproduction, Ministry of Education, Center of Reproductive Medicine, Department of Obstetrics and Gynecology, Peking University Third Hospital, Beijing, China

12Clinical Stem Cell Research Center, Peking University Third Hospital, Beijing, China 13Department of Neurology, Tianjin Medical University General Hospital, Tianjin 300052, China 14Beijing Neurosurgical Institute, Beijing 100070, China

15Department of Neurosurgery, Beijing Tiantan Hospital, Capital Medical University, Beijing 100070, China 16Chinese Glioma Genome Atlas Network & Asian Glioma Genome Atlas Network, Beijing 100070, China

17National Laboratory of Biomacromolecules, CAS Center for Excellence in Biomacromolecules, Institute of Biophysics, Chinese Academy of Sciences, Beijing 100101, China

18Department of Neurosurgery, Xuanwu Hospital Capital Medical University, Beijing 100053, China 19Clinical Research Center for Epilepsy Capital Medical University, Beijing 100053, China

20Beijing Municipal Geriatric Medical Research Center, Beijing 100053, China

21MAIS, State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China

22School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing 100049, China

23Peking-Tsinghua Center for Life Sciences, Academy for Advanced Interdisciplinary Studies, Center for Quantitative Biology (CQB), Peking University, Beijing 100871, China

24Peking University International Cancer Institute, Peking University Health Science Center, Peking University, Beijing 100191, China 25State Key Laboratory of Primate Biomedical Research, Institute of Primate Translational Medicine, Kunming University of Science and Technology, Kunming, Yunnan 650500, China

26Department of Neurology, Daping Hospital, Third Military Medical University, Chongqing 400042, China

27State Key Laboratory of Trauma, Burn and Combined Injury, Institute of Surgery Research, Daping Hospital, Third Military Medical University, Chongqing 400042, China

</div>Trang 4<div class="page_container" data-page="4">

Figure 1. Phenotypic insights into aging-related alterations in the primate FL

(A) Experimental scheme of aging phenotype analysis, sequencing approaches, and mechanistic studies.

(B) SA-b-Gal staining of the GM and WM sections of FL from young and aged monkeys. The SA-b-Gal-positive area is quantified as fold changes (aged vs. young).(C) Aggresome staining of the GM and WM sections of FL from young and aged monkeys. The number of positively stained cells is quantified as fold changes(aged vs. young).

(D) Ab (4G8) immunofluorescence staining of the GM and WM sections of FL from young and aged monkeys. The number of Ab (4G8)-positive cells is quantified asfold changes (aged vs. young).

(E) Lipofuscin accumulation in the GM and WM sections of FL from young and aged monkeys. The number of lipofuscin-positive cells is quantified as fold changes(aged vs. young).

(F) gH2A.X immunohistochemical staining of the GM and WM sections of FL from young and aged monkeys. The signal intensity of gH2A.X is quantified as foldchanges (aged vs. young).

(G) MMP-9 immunofluorescence staining of the GM and WM sections in FL from young and aged monkeys. The number of MMP-9-positive cells is quantified asfold changes (aged vs. young).

Scale bars, 50 mm and 10 mm (zoomed-in images) in (B); 20 mm and 10 mm (zoomed-in images) in (C)–(G). White arrowheads indicate the corresponding positive-staining cells. Young, n = 8; aged, n = 8 monkeys. Data are represented as the mean± SEM. Two-tailed t test.

</div>Trang 6<div class="page_container" data-page="6">

the notion that genomic instability is a hallmark of brain ag-ing,27,28 we observed an elevated DNA damage response (marked by gH2A.X foci formation) in aged FL (Figure 1F). Further, we detected evidence of brain inflammation, a critical player in the development of cognitive dysfunctions, as two important pro-inflammatory matrix metalloproteinases (MMPs), MMP-9 and MMP-3,29,30 both accumulated in the aged FL (Figures 1G andS1D).

Cortical neurons are prominently vulnerable in primate FL aging

To dissect relationships between the multifaceted phenotypes of FL aging and underlying molecular mechanisms, we analyzed tissue-level RNA sequencing (RNA-seq) and quanti-tative proteomic analysis of young and aged NHP FL. By per-forming differential expression analysis, we identified a total of 497 upregulated and 524 downregulated differentially ex-pressed genes (DEGs), as well as 120 upregulated and 39 downregulated differentially expressed proteins (DEPs), dur-ing FL agdur-ing (Figures S2A–S2C; Table S1). Joint analysis of the aging-related DEGs and DEPs identified pathways involved in neuron functions, including synapse assembly and transsynaptic signaling, that were downregulated at both the RNA and protein levels in the aged FL (Figures 2A andS2D). Upregulated aging-related DEGs and DEPs were mainly related to chronic inflammation, such as leukocyte activation and granulocyte chemotaxis pathways, indicative of prominent neuroinflammation in the aged FL (Figures 2A andS2D). Furthermore, a subset of upregulated proteins in aged FL also overlapped with the upregulated DEPs that were reported to accumulate in the plasma of elderly individ-uals,31including galectin-1, a neuroinflammation-related

pro-tein encoded by LGALS132(Figure S2E), and therefore may serve as a circulatory indicator for FL aging.

To further investigate cell-type-specific alterations during FL aging, we conducted snRNA-seq and obtained 111,698

high-quality single-nucleus transcriptomes from FL tissues of young and aged monkeys (Figures 2B,S2F, and S2G). Based on this dataset, we annotated 10 cell types with specific classic markers and gene expression signatures reflecting biological functions for each cell type, including neurons (inhibitory neuron [InN]; excitatory neuron [ExN]), glial cells (microglia; oligodendrocyte [OL]; immune OL [ImmuOL]; astrocyte), pro-genitor cells (OL propro-genitor cell [OPC]; committed OPC [COP]), vascular cells (endothelial cell [EC]), and meningeal cells (meninge) (Figures 2B, S2H, and S2I;Table S2). Among these cell types, we noticed that ExNs and InNs harbored more DEGs detectable at the population level, indicating their high susceptibility and critical importance to FL aging ( Fig-ure 2C;Table S1). Indeed, neuronal functions, such as transsy-naptic signaling, ion transmembrane transport, and neuron pro-jection development, might be compromised during aging, as shown by enriched Gene Ontology (GO) terms of downregu-lated aging-associated DEGs (cluster 11) (Figures 2C, 2D, and S3A). Conversely, upregulation of neuron-specific aging DEGs was enriched for cytokine-cytokine receptor interactions

(IL15, LTB, TNFRSF25, GDF15, CXCL14) (clusters 4 and 5),

implying an intrinsic activation of inflammatory pathways in the neurons themselves (Figures 2C and 2D).

We also noticed that aging-related degenerative (i.e., accumu-lation of aggresomes and Ab deposits) and pro-inflammatory (i.e., MMP-9 escalation) features were overrepresented in the neuron-enriched GM region of the FL (Figures 1C, 1D, and 1G), consistent with the molecular profiling data. Indeed, aggresome staining was primarily detected in NeuN-positive neuronal cells, which was increased in the aged FL (Figure S3B). Similarly, over 90% of SA-b-Gal-positive cells were NeuN-positive neuronal cells (Figure 2E). In particular, morphological analyses by Golgi staining revealed decreased dendritic length, arborization, and spine density in the aged FL (Figures 2F and 2G), indicative of a functional deficit and validating that FL neurons are particularly vulnerable to aging.

Figure 2. Integrative analysis of transcriptomic, proteomic, and single-nucleus transcriptomic datasets underscores neuroinflammation during primate FL aging

(A) Network plots showing the enriched pathways for differentially expressed genes (DEGs) and proteins (DEPs) (aged vs. young). Shapes of the nodes indicatethat components of the pathways are dysregulated at both the RNA and protein levels or in an RNA/protein-specific manner. Edge colors from light to red/blueindicate the Jaccard index from low to high. Node sizes from small to large indicate the number of DEGs or DEPs in the enriched terms from low to high. Left,upregulated pathways; right, downregulated pathways.

(B) t-SNE plot showing the annotated cell types in the single-nucleus RNA-seq (snRNA-seq) data of the monkey FL. Samples were collected from young and oldmonkeys (n = 8 monkeys per group).

(C) Heatmap showing the averaged gene expression of the upregulated and downregulated DEGs (aged vs. young) detected in different cell types in the monkeyFL. Each row represents one cell type, and each column denotes the expression of one DEG. The color key from blue to red indicates gene expression (scaled bycolumn) from low to high.

(D) Bar plots showing the enriched pathways of upregulated (clusters 4 and 5 in C) and downregulated (cluster 11 in C) DEGs in neurons (InN and ExN) of themonkey FL. Color keys from light to blue/red indicate theLog10P from low to high.

(E) SA-b-Gal and NeuN immunostaining in FL from young and aged monkeys. Upper right, the number of SA-b-Gal-positive neurons in the GM sections isquantified as fold changes (aged vs. young). Lower right, pie plots showing the percentages of neurons and other cells to total SA-b-Gal-positive cells in the GM ofyoung and aged monkey FL. Black arrowheads indicate SA-b-Gal-positive neurons.

(F) Golgi staining in the GM sections in FL from young and aged monkeys. Left, representative images and trajectory tracking of dendrites. Middle, the averagelength of dendrites is quantified as fold changes (aged vs. young). Right, Sholl analyses showing reduced dendritic intersections (ranging from 60 to 140 mm) ofaged neurons.

(G) Golgi staining in the GM sections of FL from young and aged monkeys. The apical dendritic spine density is quantified as fold changes (aged vs. young). Redarrowheads indicate the dendritic spines.

Scale bars, 20 mm and 10 mm (zoomed-in images) in (E); 50 mm in (F); and 5 mm in (G). Young, n = 8; aged, n = 8 monkeys. Data are represented as the mean± SEM.Two-tailed t test.

</div>Trang 8<div class="page_container" data-page="8">

Disrupted nuclear lamina and derepressed retrotransposable elements in the aged FL

In human mitotic cells, we and others have shown in vitro that

disruption of nuclear lamina organization and augmentation of heterochromatin erosion can lead to the derepression of retro-transposons therein and enhance inflammatory responses.33–40 Next, we asked whether aging-associated nuclear lamina erosion and epigenetic changes might underlie the dysfunction

and inflammation of post-mitotic neurons in vivo. By analyzing

the snRNA-seq data, we found that transcript levels for two

ma-jor structural proteins of the nuclear lamina, LMNB1 and LMNB2,

were decreased in neurons (Figure 3A). As expected, we observed discontinuous morphology of the nuclear lamina with lower expression of Lamin B1 and Lamin B2, as well as reduced LAP2b, a heterochromatin-related inner nuclear membrane pro-tein predominantly in cortical neurons (but not in glial cells) in the aged primate FL, while other nuclear envelope proteins remained unchanged (such as Emerin, SUN1, and SUN2) (Figures 3B–3D andS4A–S4J). Moreover, the 3D conformation of nuclear morphology based on the H3K9me3 and LAP2b sig-nals and transmission electron microscopy (TEM) visualization of the nuclear structure revealed compromised integrity of the nuclear lamina along with diminished heterochromatin under-neath the nuclear membrane in the aged neurons (Figures 3D– 3F andS4K).

Next, we sought to dissect changes in genome-wide DNA methylation, an important epigenetic modification that locks down heterochromatin to prevent aberrant transcription.41,42We analyzed aging-related DNA methylation changes in both pro-tein-coding regions and regions harboring various repetitive ele-ments encapsulated in dense heterochromatin, such as ERVs

belonging to retrotransposable elements.36,37,39,43–45 Although no apparent alterations in DNA methylation levels of protein-cod-ing regions were observed (Figure S4L), we revealed reduced methylation levels of ERV in the aged FL (Figure 3G). In accor-dance, nearly 20% of repetitive elements detected were transcrip-tionally upregulated, with ERVs as the top-ranking families, while very few repetitive elements were downregulated (Figure 3H;

Table S3). Accordingly, elevated ERV-Env levels, the protein prod-ucts of ERV transcripts, were present in the aged FL and specif-ically detected in cortical neurons but not non-neuronal cells (Figures 3I andS4M). It is noteworthy that, consistent with the neuron-specific nuclear deformation and ERV activation in aged monkey FL, a similar upregulated expression of ERV-Env proteins is observed in the GM, but not the WM, of the FL of monkeys with Hutchinson-Gilford progeria syndrome (HGPS) (Figures S4N and S4O), an early-onset aging disorder characterized by carrying a

heterozygous mutation of the LMNA gene.39We also found a higher DNA content of ERVs in the aged FL, implying that their reverse transcripts increased and/or underwent active genomic transposition (Figure 3J). Additionally, an increased number of retrovirus-like particles (RVLPs) was detected in the aged FL ( Fig-ure 3K), which is consistent with our recent observation that RVLPs appear in senescent mitotic cells.39We further calculated the co-expression gene network and identified DEGs whose expression fluctuation synchronized with that of ERVs. These DEGs were highly relevant to viral infection pathways and antigen processing-cross presentation pathways (Figures 3L–3N;

Table S4). These results raise the possibility that insufficient nu-clear lamina proteins lead to epigenetic instability with resurrected ERV retrotransposons that may cause aging-related pro-inflam-matory phenotypes in the NHP FL.

Figure 3. Epigenetic landscape reveals disruption of nuclear architecture and derepression of retrotransposable elements in the aged primate FL

(A) Violin plots showing the expression levels of LMNB1 and LMNB2 in snRNA-seq data of neurons (InN and ExN) from the indicated group of the monkey FL.

(B) Immunofluorescence staining of Lamin B1/Lamin B2 and NeuN in FL from young and aged monkeys. The fluorescence intensity of Lamin B1/Lamin B2 inneurons in the GM sections is quantified as fold changes (aged vs. young). Young, n = 8; aged, n = 8 monkeys.

(C) 3D reconstruction of Lamin B1 (top) and Lamin B2 (bottom) immunofluorescence images in neurons.

(D) Immunofluorescence staining of LAP2b and H3K9me3 in the GM sections of FL from young and aged monkeys. Left, 3D reconstruction of LAP2b andH3K9me3 immunofluorescence images in neurons. Right, the number of cells with abnormal nuclear lamina in the GM sections is quantified as fold changes (agedvs. young). White arrowheads indicate the abnormal nuclear lamina. Young, n = 8; aged, n = 8 monkeys.

(E) TEM analysis of the heterochromatin architecture at the nuclear periphery in the GM sections in FL from young and aged monkeys. The percentages of neuronswith heterochromatin loss at the nuclear periphery are presented below the images. Red arrowheads indicate heterochromatin-loss regions. Young, n > 200 cells;aged, n > 200 cells.

(F) Immunofluorescence staining of H3K9me3 and NeuN in FL from young and aged monkeys. The fluorescence intensity of H3K9me3 in neurons in the GMsections is quantified as fold changes (aged vs. young). Young, n = 8; aged, n = 8 monkeys.

(G) Metaplots showing the loss of CG methylation (mCG) levels (mCG/CG) at ERVs in the monkey FL during aging.

(H) Top, pie and bar plots showing the percentage of different types of differentially expressed repetitive elements (aged vs. young). Gray indicates unchangedrepetitive elements. Bottom, distribution density of the log2FC value of all ERV members in aged monkey FL.

(I) Immunofluorescence staining of ERV-Env and NeuN in FL from young and aged monkeys. Left, schematic diagram of ERV-RVLP production process. Middle,representative images. Right, the fluorescence intensity of ERV-Env in neurons in the GM sections is quantified as fold changes (aged vs. young). Young, n = 8;aged, n = 8 monkeys.

(J) qPCR analysis of the relative ERV genomic DNA content in the FL from young and aged monkeys. Young, n = 8; aged, n = 8 monkeys.

(K) TEM analysis of the putative RVLPs in the GM sections of FL from young and aged monkeys. The number of RVLPs per cell is quantified. The red arrowheadindicates the putative RVLP. Young, n = 80 cells from four samples; aged, n = 80 cells from four samples.

(L) Dot plots showing the log2FC of DEGs (aged vs. young) and their Pearson’s correlation coefficients with the expression levels of ERV retrotransposableelements. DEGs with high positive or negative correlations are shown in red (correlation coefficient >0.7) or blue (correlation coefficient <0.7), respectively.(M) Bar plot showing the percentage of upregulated and downregulated DEGs (aged vs. young) to total DEGs correlated with ERV expression level.(N) Bar plot showing the enriched pathways of upregulated DEGs correlated with ERVs. The color key from light to dark indicates theLog10P from low to high.Scale bars, 20 mm in (B); 2 mm in (C) and (D); 2 mm and 200 nm (zoomed-in images) in (E); 20 mm and 10 mm (zoomed-in images) in (F) and (I); 200 nm and 100 nm(zoomed-in images) in (K). Data are represented as the mean± SEM. Two-tailed t test.

</div>Trang 9<div class="page_container" data-page="9">

(legend on next page)

</div>Trang 10<div class="page_container" data-page="10">

Downstream activation of inflammatory responses intrinsic to aged neurons in both monkey and human FL Following the prediction by the co-expression analysis, we asked whether derepressed ERVs might trigger activation of innate immune responses mediated by the DNA-sensing recep-tor cyclic GMP-AMP synthase (cGAS),39and whether such a scenario occurs in a cell-type-specific manner. Indeed, we observed more cGAS-positive neurons in the aged FL than in the younger counterparts, but not in non-neuronal cells (Figures 4A, 4B, andS5A), along with increased cGAS enrich-ment on the cytoplasmic ERV DNA as detected by immunopre-cipitation analysis in aged FL (Figures 4C and 4D). Consistently, 2030-cGAMP, an endogenous second messenger produced by cGAS to activate TBK1 and downstream innate immune path-ways, was also found to accumulate in the lysates of aged FL (Figures 4E, 4F, andS5B–S5E). Along this molecular cascade, the protein levels of nuclear factor kB (NF-kB)/RelA were increased and an exaggerated inflammation response score was detected in cortical neurons of the aged primate FL (Figures S5C–S5E). Consequently, downstream NF-kB target

genes and pro-inflammatory cytokines, including IL6, IL1, and

IL15, classified as senescence-associated secretory phenotype

(SASP) factors,46,47 were also escalated (Figures 4G, S5F, and S5G).

Similar to aged monkey FL neurons, we also observed dimin-ished B-type lamins in aged human FL tissues (Figures 4H–4J), accompanied by heterochromatin erosion, unlocked expression of ERVs, accumulation of RVLPs, and activation of the innate im-mune pathway, relative to younger counterparts (Figures 4K–4O and S5H). Likewise, in cerebrospinal fluid (CSF) samples obtained from young and aged donors, expression of ERV-Env protein, release of 2030-cGAMP, and accumulation of the pro-in-flammatory factor IL-6 were markedly increased with age

(Figure 4P). Altogether, we demonstrate that destabilization of the nuclear lamina and heterochromatin, derepressed ERV retro-transposable elements, with concurrently augmented cGAS signaling detected in senescent FL neurons serve as prominent features of both NHP and human FL aging.

Deficiency of Lamin B1 and Lamin B2 elicits aging in human neurons

We established a human embryonic stem cell (hESC)-derived neuronal model that recapitulates ‘‘aging in a dish’’ to investigate the mechanism that triggers neuronal aging (Figures 5A and

S6A).48–50Notably, B-type lamin proteins exhibited decreased expression and discontinuous distribution along the nuclear pe-riphery during prolonged culture of human neurons (hNeurons) (Figures 5A–5C and S6B–S6E), followed by a reduction of H3K9me3, derepression of ERV retrotransposons, accumulation of RVLPs, increased cGAS enrichment on the cytoplasmic ERV DNA, and activation of the cGAS pathway and downstream genes related to interferon signaling (Figures 5A, 5B, 5D–5I, and S6F–S6N;Table S5). More importantly, we also detected the manifestation of classic aging phenotypes, such as augmented SA-b-Gal activity and accumulation of aggresomes and Ab deposits in senescent hNeurons (Figures 5G, 5H, and

To test whether such early events can drive neuronal aging by activating ERV retrotransposons and downstream innate im-mune pathways, we silenced Lamin B1 and Lamin B2 in young hNeurons using small interfering RNAs (siRNAs) (Figures S7A– S7D). Lamin B1 and Lamin B2 knockdown led to a loss of hetero-chromatin, ERV activation, accumulation of RVLPs, increased cGAS enrichment on the cytoplasmic ERV DNA, activation of the cGAS pathway, increased numbers of p-TBK1-positive cells, and increased RelA levels (Figures 5J–5M and S7E–S7I).

Figure 4. Activation of inflammatory responses intrinsic to aged neurons in both monkey and human FLs

(A) Schematic diagram of the innate immune response through the cGAS-STING pathway.

(B) Immunofluorescence staining of cGAS and NeuN in FL from young and aged monkeys. The number of cGAS-positive neurons in the GM sections is quantifiedas the fold change (aged vs. young). Young, n = 8; aged, n = 8 monkeys.

(C) Schematic diagram of the ChIP-qPCR strategy for measuring the level of cGAS-immunoprecipitated cytoplasmic ERV cDNA in FL from young and agedmonkeys.

(D) Immunoprecipitation assay showing cGAS enrichment in cytoplasmic ERV cDNA in FL from young and aged monkeys. Quantitative data of the levels ofcytoplasmic ERV cDNA immunoprecipitated by cGAS are presented as the means± SEMs. Young, n = 8; aged, n = 8 monkeys.

(E) ELISA analysis of 2030-cGAMP levels in the FL from young and aged monkeys. The 2030-cGAMP levels were normalized to the total protein concentration.Young, n = 8; aged, n = 8 monkeys.

(F) Immunofluorescence staining of p-TBK1 and NeuN in the GM sections in FL from young and aged monkeys. The number of p-TBK1-positive neurons in the GMsections is quantified as fold changes (aged vs. young). Young, n = 8; aged, n = 8 monkeys.

(G) qRT-PCR analysis of the relative IL1, IL6, and IL15 mRNA levels in the FL from young and aged monkeys. Young, n = 8; aged, n = 8 monkeys.

(H) FL sample information from young and aged donors.

(I–K) Immunofluorescence staining of Lamin B1 (I), Lamin B2 (J), H3K9me3 (K), and NeuN in FL from young and aged humans. The fluorescence intensity of LaminB1/Lamin B2/H3K9me3 in NeuN-positive neurons in the GM sections is quantified as fold changes (aged vs. young). Young, n = 200 cells from five individuals;aged, n = 200 cells from six individuals.

(L) Immunofluorescence staining of ERV-Env and NeuN in FL from young and aged humans. The fluorescence intensity of ERV-Env in NeuN-positive neurons inthe GM sections is quantified as fold changes (aged vs. young). Young, n = 5 individuals; aged, n = 6 individuals.

(M) TEM analysis after immunogold labeling of FL from young and aged humans using ERV-Env antibody. The number of RVLPs per mm2

is quantified. The regionhighlighted by the red dashed circle indicates putative RVLPs. Young, n = 60 cells from two individuals; aged, n = 60 cells from two individuals.

(N and O) Immunofluorescence staining of cGAS (N), p-TBK1 (O), and NeuN in FL from young and aged humans. The number of cGAS/p-TBK1-positive neurons inthe GM sections is quantified as fold changes (aged vs. young). Young, n = 5; aged, n = 6 individuals.

(P) Left, CSF sample information from young and aged donors. Young, n = 3; aged, n = 3 individuals. Right, ELISA analysis of ERV-Env protein, 2030-cGAMP andIL-6 levels in the CSF from young and aged individuals. Young, n = 3; aged, n = 3 individuals.

Scale bars, 20 mm and 10 mm (zoomed-in images) in (B), (F), (L), (N), and (O); 10 mm in (I)–(K); 200 nm and 100 nm (zoomed-in images) in (M). White arrowheadsindicate the corresponding positive-staining neurons. Data are represented as the mean± SEM. Two-tailed t test.

</div>Trang 12<div class="page_container" data-page="12">

Concurrently, deficiency of Lamin B1 and Lamin B2 accelerated cellular senescence along with an accumulation of Ab and aggresomes in hNeurons (Figures 5N,S7J, and S7K), similar to the phenotypes in hNeurons after prolonged culture or in

the aged primate FL in vivo (Figures 1B–1D,5G, 5H, andS6O– S6Q). At a global level, the simultaneous silencing of Lamin B1 and Lamin B2 led to a reshaping of the transcriptome, resem-bling that of aged cultured hNeurons, specifically in upregulated

genes related to interferon signaling (e.g., DRB1,

HLA-DRB5, and OAS2) and cellular aging and inflammation (e.g.,ICAM1, MME, and PRELP), consistent with the phenotypic

changes we observed in the in vitro aging model (Figures 5O andS7L–S7N;Table S6).

Inhibiting the activation of ERVs and the cGAS pathway alleviates human neuronal aging

Next, we asked whether we could block the activation of ERVs or the cGAS pathway to inhibit the accelerated senescence of Lamin B1- and Lamin B2-deficient hNeurons. To test this pos-sibility, we performed siRNA-mediated silencing of ERVs, cGAS and STING in Lamin B1- and Lamin B2-deficient hNeur-ons (Figures 6A–6G). As expected, the induction of pro-in-flammatory NF-kB/RelA and SASP factors was repressed by the downregulation of ERVs, cGAS, or STING in Lamin B1-and Lamin B2-deficient neurons (Figures 6H–6J). More impor-tantly, inhibition of ERVs or the cGAS pathway in the absence of B-type lamins in hNeurons rescued aging phenotypes, including SA-b-Gal activity along with an accumulation of Ab and aggresomes (Figures 6K–6P). These findings highlight that reactivation of ERVs and the cGAS-STING molecular cascade are critical targets in alleviating human neuronal aging.

Treatment with the NRT inhibitor abacavir attenuates neuronal aging

We next explored the potential of our human neuronal aging-in-a-dish model as a robust platform for discovering and evaluating pharmaceutical aging interventions.51–53We found that abaca-vir, a potent ERV reverse transcriptase inhibitor,39,54 when directly added to the culture medium, can reduce ERV DNA con-tent, indicative of diminished reverse transcripts or retrotranspo-sition events, in both prolonged culture of wild-type hNeurons and the Lamin B1- and Lamin B2 double-knockdown cells (Figures 7A–7C). Notably, supplementation with abacavir repressed cellular senescence and inflammatory phenotypes in the prolonged culture of hNeurons (Figures 7D, 7E, 7G, and 7I). Similar effects were also observed in abacavir-treated Lamin B1- and Lamin B2-deficient hNeurons (Figures 7F, 7H, 7I, andS7O).

To expand the abacavir-mediated geroprotective effect to

the in vivo aging scenario, we treated 18-month-old aged

mice with abacavir dissolved in daily drinking water for 12 months,39 after which FL biopsies were subjected to pathological analysis (Figure 7J). In FL tissues from abaca-vir-treated mice, we found that the activated inflammatory response was attenuated compared with that in untreated mice of the same age (Figures 7K and 7L). Notably, we found that neuronal aging was mitigated, as manifested by reduced Ab deposits and levels of protein aggregates (Figures 7M and 7N). Altogether, we demonstrated that nuclear architectural aberration and the concomitant reacti-vation of ERVs fuel neuronal senescence and inflammation that is mediated by the cGAS-STING pathway, which was antagonized by abacavir, which blocks ERV reverse transcription.

Figure 5. Targeted depletion ofLMNB1 and LMNB2 results in senescence of hNeurons

(A) From left to right, immunofluorescence staining of Lamin B1, H3K9me3, and ERV-Env in MAP2-marked hNeurons during prolonged culture. White arrowheadsindicate the putative RVLPs. Scale bars, 20 mm and 5 mm (zoomed-in images).

(B) Quantitative data for fluorescence intensity of Lamin B1, H3K9me3, and ERV-Env in hNeurons as shown in (A). n = 3 biological samples.

(C) Violin plot showing the expression levels of LMNB1 (left) and LMNB2 (right) in hNeurons at D7 and D35 detected by bulk RNA-seq. n = 3 biological samples.

The p value was calculated with the Wilcoxon test.

(D) TEM analysis after immunogold labeling of the cultured hNeurons at D7 and D35 using ERV-Env antibody. The dashed circles indicate putative RVLPs withdiameters spanning from 80 to 120 nm, which are composed of at least 10 colloidal gold particles, the same below. The red arrowheads indicate dispersed signalsafter immunogold labeling. Scale bars, 200 nm and 100 nm (zoomed-in images). The number of RVLPs per cell is quantified. D7, n = 60 cells; D35, n = 60 cells.(E) ELISA analysis of ERV-Env protein levels in the supernatant of cultured hNeurons at D7 and D35. The ERV-Env protein concentration was normalized to the cellnumber. n = 3 biological samples.

(F) ELISA analysis of 2030-cGAMP levels in the supernatant of cultured hNeurons at D7 and D35. The 2030-cGAMP level is normalized to the cell number. n = 3biological samples.

(G) Immunofluorescence staining of p-TBK1 (left) and SA-b-Gal staining (right) in MAP2-marked hNeurons during prolonged culture. White arrowheads indicatep-TBK1-positive neurons. Scale bars, for SA-b-Gal, 50 mm; for p-TBK1 staining, 20 mm and 5 mm (zoomed-in images).

(H) The number of p-TBK1- (left) and SA-b-Gal-positive (right) neurons, as shown in (G), quantified as fold changes. n = 3 biological samples.

(I) ELISA analysis of IL-6 levels in the supernatant of cultured hNeurons at D7 and D35. The IL-6 level was normalized to the cell number. n = 3 biological samples.Data are represented as the mean± SEM. Two-tailed t test.

(J and K) Immunofluorescence staining of H3K9me3 (J)/ERV-Env (K) in MAP2-marked neurons transfected with NC or siRNA duplexes against LMNB1 and

LMNB2. White arrowheads indicate the putative RVLPs. Scale bars, 20 mm and 5 mm (zoomed-in images). n = 3 biological samples.

(L) TEM analysis after immunogold labeling of the hNeurons transfected with NC or siRNA duplexes against LMNB1 and LMNB2 using ERV-Env antibody. The

dashed circles indicate putative RVLPs. Red arrowheads indicate dispersed signals after immunogold labeling. Scale bars, 200 and 100 nm (zoomed-in images).The number of RVLPs per cell is quantified. si-LB1&2, n = 20 cells; si-NC, n = 20 cells.

(M) Immunofluorescence staining of p-TBK1 and MAP2 in the neurons transfected with NC or siRNA duplexes against LMNB1 and LMNB2. White arrowheads

indicate p-TBK1-positive neurons. Scale bars, 20 mm and 5 mm (zoomed-in images). n = 3 biological samples.

(N) SA-b-Gal staining in neurons transfected with NC or siRNA duplexes against LMNB1 and LMNB2. Scale bars, 50 mm. n = 3 biological samples.

(O) ELISA analysis of IL-6 levels in the supernatant of cultured neurons transfected with NC or siRNA duplexes against LMNB1 and LMNB2. The IL-6 level was

normalized to the cell number. n = 3 biological samples. Data are represented as the mean± SEM. Two-tailed t test.

</div>Trang 13<div class="page_container" data-page="13">

Figure 6. ERV, cGAS, or STING knockdown abrogates the inflammation and aging of hNeurons caused by depletion of Lamin B1 and Lamin B2

(A) Schematic diagram of ERV, cGAS, or STING knockdown in LMNB1- and LMNB2-knockdown hNeurons using siRNA transfection.

(B, D, and F) Verification of knockdown efficiency by qRT-PCR at 48 h after transfection with NC or siRNA duplexes against ERVs/cGAS/STING (si-ERVs/si-cGAS/si-STING) in hNeurons transfected with siRNA duplexes against LMNB1 and LMNB2. 18S rRNA was used as a loading control.

(C, E, and G) Verification of knockdown efficiency by immunofluorescence staining of ERV-Env/cGAS/STING at 72 h after transfection with NC or siRNA duplexes

against ERVs/cGAS/STING in hNeurons transfected with siRNA duplexes against LMNB1 and LMNB2.

(H) Immunofluorescence staining of RelA and MAP2 in LMNB1- and LMNB2-knockdown hNeurons after transfection with NC or siRNA duplexes against ERV,

cGAS, or STING.

(legend continued on next page)

</div>Trang 14<div class="page_container" data-page="14">

Senescent cells, defined by permanent cell-cycle arrest, natu-rally accumulate in various organs with aging.55–58Most cellular senescence studies in the brain have focused on glial cells that actively proliferate.59,60 In particular, epigenetic alterations, exemplified by the loss of heterochromatin, have recently been identified as contributors to senescence.61–64Again, however, most studies have been reported for actively dividing cells, which undergo repeated cycles of nuclear lamina disintegration and reformation, along with decondensation and re-establish-ment of heterochromatin, and accumulate DNA replication errors and DNA damage.65–71 However, the senescence of post-mitotic neurons is not affected by either DNA replication or cell division.72 Whether neurons, which are long-lived and highly specialized post-mitotic cells, also exploit mechanisms of nu-clear envelope-heterochromatin attrition to initiate senescence is still an open question. In this study, using both NHP and human FL samples, we revealed that most senescent programs in the aged cortex are actually present in post-mitotic neurons relative to non-neuronal cells. Furthermore, we systemically character-ized a wide range of aging-related features in aged neurons, such as the accumulation of (epi)genomic instability with hetero-chromatin erosion, along with the potential underlying mecha-nisms. Together with the aforementioned pioneering studies, our work expands the traditional scope of senescent cells and highlights that neuronal senescence per se is one of the most prominent characteristics during FL aging.

Efforts have been made to reveal molecular events that trigger aging-related dysfunctions in FL73–76; however, intrinsic alter-ations in neurons in response to aging are not completely under-stood. Using the neuronal aging-in-a-dish model derived from hESC, we genetically perturbed the expression of Lamin B1 and Lamin B2 in hNeurons and pinpointed that decreased B-type lamins drive the activation of ERV retrotransposons dur-ing neuronal agdur-ing, which leads to elevated senescence and

inflammation. Our study demonstrated that this in vitro human

neuronal model could mimic the aging-related phenotypes of neurons from primate brains and study cell-autonomous mech-anisms underlying primate neuronal aging.23,49,77Future

appli-cations of the in vitro model in intervention treatment and drug

screening might broaden our understanding of neuronal aging.78,79

ERV element activation has been reported to be associated with cellular senescence.39In addition, increased expression

of ERV elements was found to contribute to neurodegenerative diseases, such as amyotrophic lateral sclerosis.44,80,81However, the links between ERV derepression and physiological brain ag-ing have not been established. For the first time, our study re-vealed that ERV retrotransposable elements are derepressed in aged hNeurons, which activate cGAS signaling, exacerbating neuroinflammation, thus providing a vivid paradigm of how this cascade functions in a highly physiologically and pathologically relevant context, the aging brain. We previously found that the in-hibition of reverse transcription of the endogenous retrovirus can alleviate cartilage degeneration and aging-related inflamma-tion.39In this study, we further found that treatment with the NRT inhibitor abacavir can attenuate the augmented inflamma-tion and protein aggregates in hNeurons during prolonged cul-ture and in the neurons of FL from aged mice, indicating ERV tar-geting as a promising strategy to delay brain aging and extend health span.

Limitations of the study

Our study found that ERVs are activated in aged neurons in the primate FL. However, further research is needed to determine whether ERV activation also occurs in specialized neurons in different brain regions. While this study sheds light on the age-associated accumulation of RVLPs in the CSF of patients diagnosed with nervous system tumor diseases, further research is necessary to investigate changes in RVLPs in the CSF of healthy individuals during physiological brain aging. In addition, we also found that RVLPs can be secreted into the culture medium by senescent neurons. However, whether these RVLPs can infect young neurons, mediate aging infec-tivity, and spread between neuronal and non-neuronal cells de-serves further investigation. Furthermore, the reverse transcrip-tase inhibitor abacavir was shown to repress neuronal inflammation and aging in hNeurons during prolonged culture and in the FL of aged mice. However, whether ERV inhibition can extend the lifespan of aged mice and whether it has a similar effect on primates also deserve further exploration in the future. Finally, due to the limitations of technical means,

although the in vitro neuron model can simulate neuronal aging

and be applied for drug evaluation, it cannot fully recapitulate

the microenvironment of neurons in vivo. In the future, humanorganoid-based and in vivo research systems are needed to

better explore the regulatory mechanisms of these transpos-able elements and how to suppress their activation to protect neurons from aging-related degeneration.

(I) Quantitative data for the fluorescence intensity of RelA in hNeurons, as shown in (H).

(J) qRT-PCR analysis of the relative IL6 mRNA levels in LMNB1- and LMNB2-knockdown hNeurons after transfection with NC or siRNA duplexes against ERV,

cGAS, or STING.

(K) SA-b-Gal staining in LMNB1 and LMNB2 knockdown hNeurons after transfection with NC or siRNA duplexes against ERV, cGAS, or STING.

(L) Quantitative data for SA-b-Gal-positive neurons as shown in (K).

(M) Ab (4G8) and MAP2 staining in LMNB1- and LMNB2-knockdown hNeurons after transfection with NC or siRNA duplexes against ERV, cGAS, or STING. White

arrowheads indicate Ab (4G8)-positive neurons.

(N) The number of Ab (4G8)-positive neurons quantified as fold changes as shown in (M).

(O) Aggresome and MAP2 staining in LMNB1- and LMNB2-knockdown hNeurons after transfection with NC or siRNA duplexes against ERV, cGAS, or STING.

White arrowheads indicate the aggresome-positive neurons.(P) Quantitative data for aggresome-positive neurons as shown in (O).

Scale bars, 20 mm and 5 mm (zoomed-in images) in (C), (E), (G), (H), (M), and (O); 50 mm in (K). Data are represented as mean± SEM. n = 3 biological samples. Two-tailed t test.

</div>Trang 15<div class="page_container" data-page="15">

(legend on next page)

</div>