Aging-disturbed FUS phase transition impairs hematopoietic stem cells by altering chromatin structure

Aged hematopoietic stem cell (HSC) exhibits impaired reconstitution capacity and differentiation bias toward myeloid lineage, wherein the molecular mechanism remains largely unknown.^1,2 Several studies have shown that aged HSCs exhibit epigenetic drift, which results in transcriptome alteration and compromised function.^3-7 Genome organization is the process of compacting the long linear genomic DNA into a 3-dimensional (3D) micrometer-sized nucleus, which is regulated by a variety of molecular mechanisms.⁸ Phase separation, which provides a venue for biochemical reactions through the dynamic changes of membrane-less compartments,⁹ has been shown to be involved in the regulation of chromatin structure.^10-12 The internal connection between them has not been investigated.

Here, we investigate the functional role of FUS in HSC aging and, to our knowledge, for the first time demonstrate that aberrant FUS condensate promotes HSC aging by reorganizing chromatin structure. We also point out that this mechanism might be generalized to aging and aging-related diseases.

To generate Fus reporter mice (Fus-gfp mice), the TAG stop codon was replaced with the 3xEAAAK-EGFP cassette.

Peripheral blood (PB) cells and bone marrow (BM) cells were harvested and stained with antibodies as mentioned in supplemental Table 1, available on the Blood website.

CDS were cloned into the pRRL-PPT-SF-newMCS-IRES2-EGFP vector, and Fus shRNAs were cloned into the SF-LV-miRE-mCherry vector. Lentiviruses were produced in 293T cells.

HSCs (CD45.2) were injected into lethally irradiated recipient mice (CD45.1/2 or CD45.2) together with competitor cells (CD45.1).

Samples were resolved on sodium dodecyl sulfate-polyacrylamide gel electrophoresis, followed by transferring onto a polyvinylidene fluoride membrane. Membranes were blocked and then probed with the indicated primary antibodies. All antibodies are listed in supplemental Table 2.

Freshly isolated HSCs were lysed in Trizol. RNA was extracted by phenol-chloroform and reversely transcribed to complementary DNA by PrimeScript RT reagent Kit (Takara). PowerUp SYBR Green mix was used for reverse transcriptase polymerase chain reaction.

The transcriptome of HSCs was conducted by the Illumina Hiseq platform with a 150-bp paired-end (Berry Genomics, Beijing, China).

The tagHi-C was conducted according to the published protocol.¹³ Hi-C data were analyzed with HiC-Pro (version 3.0.0).¹⁴ 

The CUT&Tag was performed according to the protocol of CUT&Tag Assay Kit (TD903, Vazyme).

The CUT&RUN assay was conducted according to the published protocol.¹⁵ Data analysis was performed with the CUT&RUN Tools pipeline.¹⁶ 

293T cells transferred by SFB-tagged and Myc-tagged plasmids were collected, and the lysis supernatant was incubated with S-protein agarose beads. Beads were collected and washed with NETN buffer for 3 times. Then, 2× loading buffer was added to the beads and boiled for western blot.

293T cells transferred with SFB-tagged and Myc-tagged plasmids were collected, and the supernatant was incubated with streptavidin sepharose beads and biotin M1 motif. Beads were collected for western blot as described above.

Statistical analysis was made using the Prism software. Data are shown as mean ± standard error of the mean. The student t tests (2-tailed unpaired) were used for comparisons, and the Wilcoxon rank-sum tests were used for Hi-C comparisons. All experiments were repeated 2 or 3 times independently.

We observed that FUS protein, not Fus mRNA, is significantly increased in aged HSCs but not ST-HSC, MPP, common myeloid progenitor (CMP), GMP, or megakaryocyte-erythrocyte progenitor (MEP) cells (Figure 1A; supplemental Figure 1A-B). To confirm this observation, we found that the protein level of FUS is indeed increased in aged HSCs in a published proteomics database¹⁷ (supplemental Figure 1C).

To determine whether the increase of FUS promotes HSC aging, we cloned the complementary DNA of mouse FUS into a lentiviral vector, which exhibits comparable expression of FUS to aged HSCs (Figure 1B; supplemental Figure 1D). Competitive transplantation assay shows that enforced FUS significantly impairs the reconstitution capacity of HSCs and promotes HSC differentiation bias toward myeloid lineage (Figure 1C-E; supplemental Figure 1E-F). The secondary transplantation result reveals that the FUS-carrying cells were almost undetectable at the end of the 16th week (supplemental Figure 1G), indicating that enforced FUS severely impairs the self-renewal capacity of HSCs.

To further validate this observation from the opposite direction, we conducted a competitive transplantation assay by knocking down Fus in young and aged hematopoietic stem and progenitor cells (HSPCs) with an efficient shRNA (Figures 1F-G). The PB and BM analyses reveal that knockdown of FUS ameliorates the reconstitution capacity of aged HSCs and has no effect on young HSCs (Figure 1H, supplemental Figure 1H-J), but does not rescue the differentiation skewing (Figure 1I; supplemental Figure 1K). These data suggest that aging-increased FUS is a pro-aging factor in HSCs.

Because abnormal aggregation of FUS protein has been identified in multiple neurodegenerative diseases,^18,19 we generated Fus-gfp mice to investigate whether FUS protein in aged HSCs exhibits aberrant mobility (supplemental Figure 1L; see details in Methods). The immunofluorescence result shows that FUS extensively colocalized with green fluorescent protein (GFP) in non-DAPI (4′,6-diamidino-2-phenylindole)–dense regions (supplemental Figures 1M-N), suggesting that the fusion of GFP at the C-terminal of FUS does not perturb the location of endogenous FUS. It is notable that we observed that the GFP signal is significantly increased in HSCs of aged Fus-gfp mouse (supplemental Figure 1O), indicating that FUS is increased upon aging, which is consistent with western blotting result (Figure 1A). Lineage composition and hematopoietic reconstitution of Fus-gfp mice are comparable to those of the control (supplemental Figure 1P, 1-1A-C), indicating that it is an ideal model to investigate the behavior of FUS in HSCs.

Time-lapse imaging reveals that FUS forms dynamic and reversible assemblies upon sodium arsenite (ARS) treatment in HSPCs, which are featured as undergoing frequent fusion and fission events (cytoplasm, supplemental Figure 1-1D; nucleoplasm, supplemental Figure 1-1E). The fluorescence recovery after photobleaching (FRAP) assay reveals that FUS exchanges rapidly for recovery within 30 seconds (supplemental Figure 1-1F). These data suggest that FUS undergoes liquid-liquid phase separation (LLPS) in HSPCs. To further investigate whether aged HSCs exhibit abnormal dynamic behavior of FUS, the FRAP assay shows that the recovery rate of FUS condensates of aged HSCs is significantly compromised compared with young HSCs (cytoplasm: Figures 1J-K; supplemental Video 1; nucleoplasm: supplemental Figures 1-1G and 1-1H; supplemental Video 2). In addition, the mobile fraction is drastically reduced and the half-maximal recovery time (T-half) is prolonged in aged HSCs (Figure 1L; supplemental Figure 1-1I), suggesting that the mobility of FUS in aged HSCs is significantly compromised. Notably, knockdown of FUS in aged HSCs rescues the dynamic behavior of FUS (cytoplasm: Figure 1M-O; nucleoplasm: supplemental Figures 1-1J-L).

To examine the influence of enforced FUS on its mobility in young HSCs, the FRAP assay shows that enforced FUS indeed compromises the dynamic behavior of FUS condensates (cytoplasm: supplemental Figures 1-1M-O; nucleoplasm: supplemental Figures 1-1P-R). These data indicate that the expression level of FUS in HSCs is the key determinant of its dynamic behavior.

To determine whether the increase of FUS in aged HSCs is homogenous or heterogeneous, we observed that HSCs (CD34^– Flt3^– CD150⁺ LSK) are fractionated into 2 subpopulations: FUS^high and FUS^low HSCs (Figure 2A) and the percentage of FUS^high HSCs is significantly increased with aging (Figure 2B). Further analysis reveals that the percentage of FUS^high cells exhibits a conspicuous increase in LT-HSC but not in ST-HSC, MPP, CMP, GMP, MEP, or common lymphoid progenitor (CLP) with aging (Figures 2C-D; supplemental Figure 2A). These results indicate that FUS is heterogeneously expressed in HSCs, and the increase of FUS with aging attributes to the increase of FUS^high HSCs.

To further confirm this observation, we performed a serial transplantation assay by using FUS^low HSCs (supplemental Figure 2B), which is a classical approach to recapitulate the aging process by inducing replication stress.^20-22 The result reveals that the percentage of FUS^high HSCs is increased at the end of the first transplantation (36.1%), further increased at the end of the second transplantation (90.8%), and kept stable at the end of the third transplantation (94.0%) (Figure 2E-F; supplemental Figure 2C).

To investigate the dynamic behavior of FUS condensates in these 2 populations, we performed FRAP assays for FUS^high and FUS^low HSPCs from young Fus-gfp mice. The result shows that FUS^high HSPCs exhibit compromised mobility of FUS condensates compared with FUS^low cells (cytoplasm: Figure 2G-I; supplemental Video 3; nucleoplasm: Figure 2J-L; supplemental Video 4).

Briefly, these results reveal that FUS is heterogeneously expressed in HSCs, and FUS^high HSPCs display compromised mobility of FUS condensates. The percentage of FUS^high HSCs is significantly increased in response to physiological aging and serial transplantation–induced replication stress.

We conducted competitive transplantation assays for FUS^high and FUS^low HSCs (Figure 3A), and observed that the reconstitution capacity of FUS^high HSCs is impaired, including myeloid, B, and T cells (Figure 3B). In addition, FUS^high HSCs exhibit differentiation bias toward myeloid lineage (Figure 3C), which recapitulates the phenotype of aged HSCs. In a biological repeated experiment (supplemental Figure 3A), we observed a significant decrease in donor-derived HSPCs in the FUS^high group (supplemental Figure 3B-C), indicating that the impaired reconstitution of FUS^high HSCs is due to compromised proliferation capacity.

We then performed RNA sequencing to identify the transcriptional difference between FUS^high and FUS^low HSCs, and the resemblance between young and aged HSCs. Principle component analysis (PCA) shows that FUS^high HSCs are close to aged HSCs, and FUS^low HSCs are close to young HSCs (supplemental Figure 3-1A). The differentially regulated genes between FUS^high and FUS^low HSCs on the PC1 axis are mainly enriched in hemopoiesis and myeloid differentiation pathways (supplemental Figure 3-1B; supplemental Table 1). Gene set enrichment analysis (GSEA) reveals that aging, myeloid differentiation, DNA repair, DNA damage response, cell cycle, and DNA replication–related genes are enriched in FUS^high HSCs (Figure 3D; supplemental Figure 3-1C), whereas HSC fingerprint genes and autophagy-related genes are enriched in FUS^low HSCs (Figure 3E; supplemental Figure 3-1D; supplemental Table 2). In addition, dGSE analysis reveals that the canonical HSC aging genes²³ are enriched in both aged HSCs and FUS^high HSCs (supplemental Figure 3-1E; supplemental Table 2).

To further investigate whether knockdown of FUS ameliorates aged HSCs at the transcriptomic level, we conducted RNA sequencing for shFUS-carrying and shCtl-carrying HSCs freshly isolated from the recipients (Figure 1F). The result shows that knockdown of FUS reforms the transcriptional profile of aged HSCs on the dimension of PC1 axis (supplemental Figure 3-1F). Enrichment analysis reveals that knockdown of FUS extricates aged HSCs from aging-related genes, whereas it does not influence the myeloid differentiation–related genes (supplemental Figure 3-1G), which is consistent with the result that knockdown of FUS does not rescue the differentiation skewing (Figure 1I).

In brief, the above data show that FUS^high HSCs resemble aged HSCs both functionally and transcriptionally. Knockdown of FUS reforms the transcriptional alteration of aged HSCs.

Because FUS^high HSCs of young mice exhibit compromised reconstitution capacity, we then wondered whether the FUS^low HSCs of aged mice exhibited rejuvenated function. The result of the competitive transplantation assay shows that the reconstitution capacity of FUS^high HSCs and unfractionated HSCs of aged mice is comparable (Figures 3F-G), and both of them exhibit differentiation bias toward myeloid lineages (Figure 3H). It is notable that the reconstitution capacity of FUS^low HSCs from aged mice is comparable to those of young HSCs (Figure 3G), and they exhibit balanced differentiation potential from B/T/myeloid lineages (Figure 3H). This result suggests that the expression of FUS is inversely proportional to the function of HSC, and FUS^low HSCs of aged mice exhibit youthful function.

To further investigate the effect of FUS expression on HSC, we conducted a competitive transplantation assay by knocking down Fus in aged FUS^low and FUS^high HSPCs. The PB and BM analyses reveal that knockdown of FUS improves the reconstitution capacity of aged FUS^low HSCs but does not rescue the reconstitution and differentiation skewing of aged FUS^high HSCs (supplemental Figures 3-1H-K). These results indicate that the expression level of FUS controls the function of HSCs until HSCs are severely impaired.

Given that FUS undergoes phase transition in HSCs, we then set out to identify the key domain by generating FUS variants: FUS^ΔLC, FUS^ΔRGG1, FUS^ΔRGG2, FUS^ΔRGG3, FUS^ΔRGG123, and FUS^ΔNLS, wherein the indicated domains have been proven to modulate FUS mobility (Figure 4A).²⁴ Western blot assays reveal that these FUS variants are efficiently expressed (supplemental Figure 4A).

We treated HSPCs carrying FUS variants by sodium arsenite (ARS) to induce FUS condensates. The result shows that the FUS condensates are still generated in FUS^ΔLC and FUS^ΔNLS groups, but are significantly compromised in arginine-glycine-glycine (RGG)-truncated groups (cytoplasm, Figures 4B,D; nucleoplasm, Figures 4C,E), indicating that the RGG domain is essential for FUS to form condensates in HSPCs. To further confirm this observation, we purified protein of FUS variants, including FUS^ΔRGG1, FUS^ΔRGG2, and FUS^ΔRGG3. In vitro assays reveal that RGG-truncated groups exhibit significantly fewer condensates than full-length FUS protein (FUS^FL) at the same concentration (Figure 4F), demonstrating that the RGG domain indeed mediates FUS condensate formation.

We next set out to investigate whether the RGG domain mediates the inhibitory function of enforced FUS on HSCs in vivo. We conducted a competitive transplantation assay and observed that enforced FUS impairs the reconstitution capacity of HSCs, whereas the reconstitution capacity is restored in the FUS^ΔRGG1, FUS^ΔRGG2, FUS^ΔRGG3, and FUS^ΔRGG123 groups, including myeloid, T, and B cells (Figure 4G). The result suggests that the impairment of enforced FUS on the reconstitution capacity of HSCs is mediated by the RGG domain and perhaps the RGG–mediated phase transition.

To rule out the influence of the potential non-LLPS function of the RGG domain on HSCs, we generated FLL2^PLD-FUS^ΔRGG123 and ELF3^PLD-FUS^ΔRGG123 by complementing FUS^ΔRGG123 with the PLD domain of Arabidopsis ELF3 and FLL2, which are in vivo regulators of phase separation (Figure 5A).^25,26 Western blot assay reveals that these variants are efficiently expressed (supplemental Figure 4B). Imaging analysis reveals that the cells transfected by FUS^ΔRGG123 cannot form droplets, whereas the cells transfected by ELF3^PLD-FUS^ΔRGG123 and FLL2^PLD-FUS^ΔRGG123 form condensates efficiently in the nucleoplasm (Figure 5B) and cytoplasm (supplemental Figure 5A), indicating that the PLD domain of Arabidopsis ELF3 and FLL2 indeed restores the ability of FUS to form condensates.

To further investigate the dynamic behavior of these variants, we performed a FRAP assay and observed that FLL2^PLD-FUS^ΔRGG123 recovers moderately within 60 seconds in the nucleoplasm (Figures 5C-D), which exhibits a mild difference with FUS^FL in mobile fraction (53% vs 63%) and T-half (3.43s vs 2.68s) (Figure 5E). Meanwhile, ELF3^PLD-FUS^ΔRGG123 exhibits weak recovery in both the nucleoplasm (26.8%) (Figures 5C-E) and cytoplasm (28.6%) (supplemental Figures 5B-D). The result shows that the mobility of FLL2^PLD-FUS^ΔRGG123 is comparable with FUS, whereas the mobility of ELF3^PLD-FUS^ΔRGG123 is significantly compromised.

To examine the effect of additional PLD domain grafting on the mobility of FUS, we generated FLL2^PLD-FUS^FL and ELF3^PLD-FUS^FL by grafting the PLD domains of FLL2 and ELF3 on FUS^FL (supplemental Figures 4B and 5E). Imaging analysis reveals that the cells transfected by FLL2^PLD-FUS^FL exhibit more condensates, whereas the cells transfected by ELF3^PLD-FUS^FL form FUS aggregation (supplemental Figure 5F-G).

To further investigate the dynamic behavior of FUS variants, we performed a FRAP assay on HSPCs carrying ELF3^PLD-FUS^FL and FLL2^PLD-FUS^FL, respectively. The result shows that additional FLL2^PLD grafting indeed compromises the mobility of FUS^FL (59% vs 69% in the nucleoplasm and 49% vs 63% in the cytoplasm), and ELF3^PLD grafting results in the formation of FUS aggregation (cytoplasm, supplemental Figure 5H-J; nucleoplasm, supplemental Figure 5K-M).

To further test whether enforced FUS variants impair the reconstitution capacity of HSCs in vivo, we conducted a competitive transplantation assay and observed that enforced FLL2^PLD-FUS^ΔRGG123 exhibits comparable chimera with FUS^FL (Figure 5F), which is consistent with the result that the mobility of FLL2^PLD-FUS^ΔRGG123 exhibits no difference with FUS^FL (Figure 5E). We observed that enforced FLL2^PLD-FUS significantly impairs the reconstitution capacity of HSCs (Figure 5F), which is consistent with the result that the mobility of FLL2^PLD-FUS is compromised (supplemental Figure 5I,L).

Moreover, we performed the same assay for ELF3^PLD-FUS^ΔRGG123 and ELF3^PLD-FUS^FL. The result reveals that enforced ELF3^PLD-FUS^ΔRGG123 and ELF3^PLD-FUS^FL severely impair HSCs, whereas Arabidopsis ELF3^PLD domain exhibits no inhibitory effect on the reconstitution capacity of HSCs (Figure 5G; supplemental Figure 5N).

The above results show that compromised mobility of FUS induced by heterografting of Arabidopsis FLL2^PLD and ELF3^PLD domains indeed impairs HSCs. The worse the dynamic behavior of FUS is, the more it impairs HSCs.

Previous studies have shown that senescent cells undergo multiscale, 3-dimensional genome reorganization.²⁷ Combining with the results that FUS^high HSCs resemble aged HSCs functionally and transcriptionally (Figure 3G; supplemental Figure 3-1A), we tentatively assume that the destructive role of enforced FUS on HSCs may be achieved by altering chromatin structure. To test this hypothesis, we reanalyzed the RNA-sequencing data of FUS^high and FUS^low HSCs and observed that the upregulated genes are enriched for nuclear division and chromosome segregation signaling (supplemental Figure 6A), which is confirmed by GSEA analysis (supplemental Figure 6B; supplemental Table 2). To further test this hypothesis, we profiled genome-wide chromatin interactions for FUS^high, FUS^low, and aged HSCs. The result exhibits sufficient sequencing depth and high reproducibility (supplemental Figure 6C-D). Analysis of FUS^high and FUS^low, young and aged HSCs reveals a different global chromatin organization (the Hi-C data of young HSC is obtained from a published data set¹³).

On the genomic level, we observed that the interaction matrixes of FUS^high HSCs exhibit enhanced proximal interactions and weakened distal interactions compared with FUS^low HSCs (Figure 6A). Similarly, aged HSCs exhibit the same alteration (supplemental Figure 6E). The mean intrachromosomal contact probability shows that FUS^high HSCs and aged HSCs display increased contact probabilities at <6 Mb and decreased contact probabilities at >6 Mb distances (Figure 6B). To further exhibit the difference in contact probability between FUS^high and FUS^low, young and aged HSCs, we observed that PvD is significantly higher in FUS^high and aged HSCs (Figure 6C), indicating that the chromosome structure is reorganized.

At compartment scale, we observed that the interaction between A and B compartments is strengthened in FUS^high HSCs (AA and AB) (Figure 6D) and aged HSCs (AA) (supplemental Figure 6F), whereas BB is compromised in both FUS^high and aged HSCs (supplemental Figure 6G).

At the topologically associating domain (TAD) scale, the contact probability of the TAD border is increased in aged HSCs (supplemental Figure 6-1A). Further analysis reveals that the insulation score is significantly increased and 250 TADs exhibit contact changes with an insulation score change greater than 0.5 in aged HSCs (Figure 6E; supplemental Figure 6-1B; supplemental Table 3), indicating the insulation of TAD of aged HSCs is dampened. Consistently, the insulation score is increased in FUS^high HSCs and 603 merged TADs are observed (Figure 6E; supplemental Figure 6-1C; supplemental Table 3). Given that aged HSCs are composed of FUS^high and FUS^low HSCs, it is reasonable that fewer merged TADs are observed in aged HSCs. We next explored TAD changes between FUS^high and FUS^high HSCs in details.

The contact probability of the intra-TAD score of FUS^high HSCs is significantly reduced (Figure 6F), indicating that the TADs are decompacted. The inter-TAD score is increased in FUS^high HSCs (Figures 6F-G), indicating the interaction between adjacent TADs is strengthened. The contact probability among 10 adjacent TADs shows that the interactions between a TAD and its 2 neighboring TADs are enhanced (supplemental Figure 6-1D). A representative example exhibits TADs at chr17: 85 Mb-87 Mb of FUS^low HSCs are merged with FUS^high HSCs because of the blurred boundary between them (Figure 6H).

In brief, the above data reveal that the interaction within TADs of FUS^high HSCs is compromised, whereas the interaction between adjacent TADs is strengthened, which is correlated with the blurred boundaries of TADs in FUS^high HSCs.

Given that CCCTC-binding factor (CTCF) is one of the most important insulators separating TADs,^28,29 we profiled CTCF binding sites for young and aged HSCs, FUS^high and FUS^low HSPCs. The result shows that the binding intensity of CTCF is significantly compromised in FUS^high HSPCs (Figure 7A) and aged HSCs (supplemental Figure 7A). A representative example exhibits TADs at chr3: 30 Mb-31.5 Mb of FUS^low HSCs merge into 1 TAD in FUS^high HSCs because of acute reduction of CTCF binding (Figure 7B).

Coimmunoprecipitation assay demonstrates that FUS exhibits strong interaction with CTCF (Figure 7C). Immunofluorescence assay shows that CTCF exhibits increased overlap with FUS in FUS^high HSPCs (Figure 7D-E), aged HSCs (supplemental Figure 7B-C), and HSPCs with enforced FUS (supplemental Figure 7D-E). Given that CTCF forms TADs boundary by binding with its motif, we then performed a preextraction procedure to leave the CTCF stably associated with chromatin intact. The result exhibits a reduced amount of chromatin-associated CTCF (Figure 7F-G), indicating that the binding of CTCF with chromatin is compromised in FUS^high HSPCs.

We next investigated the influence of FUS mobility on the binding of CTCF with its motif by performing a DNA pull-down assay. The result shows that FLL2^PLD-FUS^FL and ELF3^PLD-FUS^FL compromise it (Figure 7H), indicating that the binding of CTCF with DNA decreases as long as the mobility of FUS is deteriorated.

To confirm the observation, we performed a preextraction immunofluorescence assay for CTCF in aged HSPCs carrying Fus shRNA. The result shows that CTCF intensity is significantly increased compared with the control group (supplemental Figure 7F-G), indicating that the binding of CTCF with chromatin is increased in response to FUS reduction. Consistent with this result, we observed that the binding of CTCF with its motif is strengthened in response to Fus reduction (supplemental Figure 7H).

Briefly, these results show that aberrant phase–separated FUS limits the binding of CTCF with chromatin and furthermore results in the compromise of TAD insulation in aged HSCs.

Previous studies have shown that TADs’ fusion leads to the alteration of genes within them.^30,31 We then wondered whether the fusion of TADs in FUS^high HSCs results in transcriptome alteration. Firstly, we observed that 77.3% of the DEG (Figure 7I), including 77.1% of the upregulated genes (Figure 7J) and 77.5% of the downregulated genes (Figure 7K) between FUS^high and FUS^low HSCs, are located in merged TADs (hereafter, the DEG inside merged TADs are named DEGs^TAD, the upregulated genes inside merged TADs are named UPGs^TAD, and the downregulated genes inside merged TADs are named DWGs^TAD). The dGSE and GSEA of DEGs show that the genes involved in HSC aging, myeloid differentiation, chromosome segregation, DNA repair, DNA damage response, DNA replication, and cell cycle, but not HSC fingerprints or autophagy, are enriched in FUS^high HSCs (supplemental Figure 7-1A). Furthermore, gene ontology (GO) analysis reveals that genes within the differential CTCF binding region between FUS^low and FUS^high HSPCs are enriched in B-cell differentiation and the cell cycle pathway (supplemental Figure 7-1A).

Except for the pathway, chromosome segregation, GO analysis, and dGSE analysis reveal that the UPGs^TAD are enriched in the DNA damage response pathway, including DNA repair, DNA replication, and double-strand break repair, but not in HSC aging genes (supplemental Figure 7-2A,B), whereas the DWGs^TAD are enriched in inflammatory signaling and HSC aging (supplemental Figure 7-2C,D). We conducted a conjoint analysis by using the database²³ and Top20 FUS^high HSC-specific aging signature genes are listed in supplemental Table 3, including Mt1 and CD86. Furthermore, it is notable that 122 genes within the DEGs^TAD (42 are upregulated and 80 are downregulated) have been demonstrated to modulate HSC function (supplemental Figure 7-2E; supplemental Table 3). For example, the gene Tgfbi is downregulated in FUS^high HSCs (supplemental Figure 7-2F) and a previous study reported that dysfunction of Tgfbi impairs the reconstitution capacity of HSCs.³² By exploring the proteomic database,¹⁷ we observed that TGFBI is indeed decreased in aged HSCs (supplemental Figure 7-2G).

Taken together, our study demonstrates that aged HSCs exhibit aberrant FUS mobility, which promotes HSC aging by blurring the boundary of TADs and then altering the expression of HSC function–related signaling.

Biomolecular condensates are an epoch-making biological phenomenon discovered in the last few decades. To our knowledge, our study demonstrates for the first time that the aberrant phase transition of FUS promotes HSC aging by altering chromatin structure and further points out that aberrant biomolecular condensate assembly may be one of the driving forces behind aging and aging-related diseases.

The authors thank Hong Zhang of Institute of Biophysics, Chinese Academy of Sciences and Pilong Li and Yi Lin of Tsinghua University for detailed discussion. They thank Qianwen Sun of Tsinghua University for offering Arabidopsis complementary DNA and Zhangbin Bao of Tsinghua University for offering purified protein of FUS^FL, FUS^ΔRGG1, FUS^ΔRGG2, and FUS^ΔRGG3. The authors also thank the Tsinghua-Peking Center for Life Sciences and the China Telecom Corporation Limited for facility and financial support.

This work was supported by grant numbers Z200022, 82250002, 92249305, 2018YFA0800200 and 81870118 to J.W., and 81890991 and 20222000298 to M.S. from the National Key R&D Program of China or the Beijing Municipal Science & Technology Commission or the National Natural Science Foundation of China and CapitalBio Technology Co, Ltd, and HH23KYZX0002 from Haihe Laboratory of Cell Ecosystem Innovation Fund.

Contribution: J.W. and M.S. contributed in conceptualization; B.T., X.W., H.H., R.C., G.Q., Y.Y., Z.X., L.W., and Q.D. contributed in methodology; B.T., X.W., H.H., R.C., G.Q., Y.Y., Z.X., L.W., Q.D., and J.Y. contributed in investigation; B.T., X.W., L.W., and Q.D. performed the formal analysis; J.W., J.Y., and M.S. contributed in resources; J.W. and M.S. contributed in writing; J.W. and M.S. contributed in funding and acquisition; and J.W., M.S., and M.Q.Z. supervised the research.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Michael Q. Zhang, The Ministry of Education Key Laboratory of Bioinformatics, Bioinformatics Division and Center for Synthetic and Systems Biology, Beijing National Research Center for Information Science and Technology, School of Medicine, Tsinghua University, Beijing, China; email: michael.zhang@utdallas.edu; Minglei Shi, The Ministry of Education Key Laboratory of Bioinformatics, Bioinformatics Division and Center for Synthetic and Systems Biology, Beijing National Research Center for Information Science and Technology, School of Medicine, Tsinghua University, Beijing, China; email: shiml79@tsinghua.edu.cn; and Jianwei Wang, School of Pharmaceutical Sciences, Tsinghua University, Haidian District, Beijing 100084, China;.