Missing MDA5 may slow blood cell aging by changing inflammation and protein balance in mice

We then wanted to check whether chromatin accessibility for inflammatory motifs is different in Mda5^-/- HSCs. To explore the chromatin accessibility landscape of aged WT and Mda5^-/- HSCs, we performed an assay for transposase-accessible chromatin followed by sequencing (ATAC-seq). We isolated HSCs using EPCR (endothelial protein C receptor) SLAM markers (EPCR + CD150 + CD48−), as EPCR has been proven to be a stable marker that is insensitive to stress signals^41,42. Indeed, more than 97% of the EPCR SLAM HSCs are also Lin-cKit + Sca+ (LSK) CD150 + CD48− (LSK SLAM) (Supplementary Fig. 1E). Similar results were obtained for middle-aged and young HSCs (Supplementary Fig. 1F, G). Therefore, in this text, “HSC” will henceforth refer to EPCR SLAM HSCs, unless stated otherwise. Clustering of samples by age was observed (Supplementary Fig. 1H). We compared the accessible chromatin regions between young or aged WT and Mda5^-/- HSCs. In the comparison of the young WT and Mda5^-/- HSCs, we found 1.723 regions that were mostly accessible in WT (reduced peaks), and 10.694 mostly accessible in Mda5^-/- HSCs (induced peaks) (Fig. 1D and Supplementary Data 1–4). Aged Mda5^-/- HSCs on the other hand have less accessible chromatin than the WT, with 11.904 sites induced in WT and 5.336 sites induced in Mda5^-/- HSCs (Fig. 1E and Supplementary Data 1–4). Both induced and reduced peaks were distributed in similar proportions across transcription start sites (TSS), exons, introns, and intergenic regions (Fig. 1D, E). Only the reduced peaks in young Mda5^-/- HSCs were found mostly around the TSS (Fig. 1D). Additionally, when we compared young and aged WT or Mda5^-/- HSCs, we observed an opening of chromatin in the aged populations that was distributed similarly in different genomic regions (Fig. 1F, G). Comparison of young and aged samples together revealed that most of the common induced regions were between the aged WT and Mda5^-/- HSCs, and thus the aged Mda5^-/- HSCs do not resemble younger WT HSCs (Fig. 1H).

Known motif analysis revealed that inflammatory motifs like IRF3 (IFN regulatory factor 3) were mainly enriched in aged Mda5^-/- reduced peaks (p-value 1e-21 in reduced and 1e-05 in induced sites), but not young WT or Mda5^-/- regions (Fig. 1I). In detail, in the young WT versus Mda5^-/- regions, we found NFY (nuclear transcription factor Y,1e-49), KLF1 (Krueppel-like factor 1, 1e-47), SP1 (specificity protein 1, 1e-38), FLI1 (friend leukemia integration 1,1e-21) in the reduced regions and ZSCAN1 (zinc finger and SCAN domain containing 1), P53 (both 1e-02) in the induced regions (Fig. 1I). In aged WT versus Mda5^-/- regions we found CTCF (CCCTC-binding factor, 1e-911), ETV2 (Ets variant 2, 1e-248), ERG (ETS transcription factor ERG, 1e-232), IRF3 (1e-21), PU1 (1e-133), JUN-AP1(1e-72) in the reduced regions and CTCF (1e-139), ETV2 (1e-69), ETS1 (1e-55), PU1 (1-36), RUNX (1e-31), IRF3 (1e-05) in the induced regions (Fig. 1I). Thus, the inflammatory IRF3 motif is less present in induced Mda5^-/- regions. The young WT HSCs harbor motifs for Pax7 (1e-2) while the induced regions in the aged WT HSCs harbor CTCF (1e-1965), ETV2 (1e-1063), ERG (1e-1032), FLI1 (1e-993), PU1 (1e-577), AP1 (activating protein-1, 1e-300), and IRF3 motifs (1e-131) (Fig. 1I). In contrast young Mda5^-/- HSCs harbor ZSCAN (1e-02) motifs while for the induced regions in aged Mda5^-/- HSCs we observed CTCF (1e-1177), ETV2 (1e-1057), ERG (1e-1003), FLI1 (1e-1015), PU1 (1e-552), AP1 (1e-211), IRF3 (1e-86) motifs (Fig. 1I).

We performed digital footprinting analysis to identify motif occupancy for transcription factors IRF3, NF-κB (p65), and STAT1 in aged compared to young WT and Mda5^-/- HSCs. We observed a substantial increase in occupancy between young and aged WT or Mda5^-/- footprints, although the total number of footprints was lower in the knockout. Additionally, we did not detect many footprints for inflammatory factors in young HSCs (Supplementary Fig. 1I). To further investigate this, we compared aged WT and Mda5^-/- footprints, and we observed less occupancy of the WT footprints in the Mda5^-/-samples (Supplementary Fig. 1J).

Taken together, these results further support that inflammatory signaling activation, a hallmark of aging, is significantly dampened in Mda5^-/- HSCs and bone marrow.

Next, we evaluated the MPP (multipotent progenitor) composition in the bone marrow of young, middle-aged and aged mice. We observed an expansion of the MPP1 population in both aged WT and Mda5^-/- animals, but to a lesser degree in the aged Mda5^-/- (Supplementary Fig. 2G). MPP2 cells also accumulated with age, to a similar extent between WT and Mda5^-/- animals (Supplementary Fig. 2H). The frequency of MPP3 cells, which represent myeloid-biased progenitor cells⁵⁰, increased in aged WT mice compared to aged Mda5^-/- animals (Supplementary Fig. 2I). Lastly, the MPP4 population, comprised of the lymphoid-prone progenitor cells⁵⁰, equally declined with age in WT and Mda5^-/- animals (Supplementary Fig. 2J). Taken together, these results indicate a myeloid bias within the WT progenitor compartment during aging, which is ameliorated by Mda5 deficiency.

We also assessed the frequency of committed progenitors, including common myeloid progenitors (CMP), granulocytic monocytic progenitors (GMP), megakaryocytic-erythroid progenitors (MEP) and common lymphoid progenitors (CLP), as well as differentiated lineage populations in the bone marrow. While all the population frequencies were comparable in young animals, CMPs and MEPs were decreased while CLPs were increased in middle-aged and aged Mda5^-/- compared to age-matched WT mice (Fig. 2G–L).

In conclusion, HSCs tend to accumulate less, and there is less myeloid bias during aging in Mda5^-/- animals.

Next, we assessed the repopulation capacity in non-competitive transplantations. In these experiments, we transplanted the volume of whole bone marrow containing 300 HSCs from aged WT or Mda5^-/- animals (CD45.2) into lethally irradiated recipients (CD45.1). The overall chimerism of Mda5^-/- HSCs was significantly higher compared to their WT counterparts, both in primary and secondary transplantations (Fig. 3C–J). However, this reconstitution advantage was lost in competitive transplantations, where the volume of whole bone marrow containing 300 HSCs of either aged WT or Mda5^-/- animals (CD45.2) was transplanted with 0.5 million bone marrow cells from a young donor (CD45.1) into lethally irradiated recipients (CD45.1/.2) (Supplementary Fig. 3C–J). Thus, aged Mda5^-/- HSCs have better repopulation capacity in a non-competitive setting, but they lose this advantage when competing with young HSCs.

To examine the capacity of WT and Mda5^-/- HSCs to repopulate lethally irradiated recipients, we sorted and transplanted 100 LSK-SLAM HSCs from young or aged WT or Mda5^-/- mice together with aged-matched CD45.1 competitors into CD45.1 lethally irradiated recipients, followed by whole bone marrow competitive secondary transplantations. Overall, we observed similar total chimerism in the primary competitive transplantations of young and aged WT and Mda5^-/- HSCs (Fig. 3K–O and Supplementary Fig. 3K–O). However, the chimerism of young Mda5^-/- HSCs was higher in the secondary competitive transplantations when compared to young WT HSCs (Supplementary Fig. 3K), with a significantly higher contribution of CD45.2 donor-derived cells to B cell, T cell, monocyte, and granulocyte populations (Supplementary Fig. 3L–O). One potential explanation for the significantly increased chimerism of the secondary transplants of the young Mda5^-/- is the increase in the frequency of the LSKs and HSCs in the primary transplant (not significant due to the transplanted animals with very low chimerism). In contrast, the total and lineage-specific chimerism of aged Mda5^-/- HSCs was comparable to the WT counterpart (Fig. 3K–P).

In line with these results, analysis of the bone marrow of primary recipients revealed no significant differences between WT and Mda5^-/- in the overall chimerism and the chimerism of lineage-specific and hematopoietic progenitor populations (Supplementary Fig. 4A, C). Yet, the contribution of CD45.2 donor-derived cells in the bone marrow of secondary recipients was higher for young and aged Mda5^-/- HSCs (Supplementary Fig. 4B, D). This confirms that both have better reconstitution capacity upon secondary transplantations and indicates that Mda5^-/- HSCs show diminished differentiation, potentially due to their more quiescent nature.

Taken together, these results show that aged Mda5^-/- are more quiescent and have a better reconstitution capacity in a non-competitive setting, as well as in the bone marrow of secondary competitive transplantations.

The master regulator of heat shock response and proteostasis is the transcription factor HSF1, which has been found to increase its concentration in the nucleus of aged and middle-aged HSCs³³. We performed GSEA analysis using the transcriptomic signature from the study by Kruta et al.³³ (HSF1 signature), and we observed enrichment in middle-aged WT HSCs when compared to young WT or middle-aged Mda5^-/- HSCs (Fig. 4J). Additionally, upstream regulator analysis from Ingenuity pathway analysis revealed HSF1 as an upstream regulator in all ages in WT and Mda5^-/- HSCs (Fig. 4K), suggesting that it is an important regulator of HSC aging. We then performed Ingenuity pathway analysis for all the different comparisons. The categories for “eukaryotic translation elongation” or “EIF2 signaling” were enriched especially in Mda5^-/- transcriptomes and at advanced ages. Finally, we observed enrichment of PKR signaling in middle-aged Mda5^-/- HSCs (Supplementary Fig. 5A).

Thus, translation and cellular response to heat stress, both maintaining proteostasis, are important in Mda5^-/- HSCs.

To assess whether we could observe similar results in the multipotent progenitor populations, we clustered the MPP3 and MPP4 top 500 genes (Supplementary Fig. 6E, F and Supplementary Data 8–12). We observed 23 upregulated genes in aged Mda5^-/- versus WT, while 276 were up and 49 were downregulated when comparing young Mda5^-/- and WT MPP3 populations. Similarly, 20 genes were up, and 12 were downregulated in aged Mda5^-/- versus WT, and 162 were up, and 66 were downregulated in young Mda5^-/- and WT MPP4 populations (Supplementary Data 8–12). Ingenuity pathway analysis revealed an enrichment of “EIF2 signaling”, “PKR-mediated signaling”, and “unfolded protein response” categories in both HSC/MMP1 and MPP3, but not MPP4, aged populations. Additionally, “aggrephagy” and “cell cycle checkpoints” were enriched in young HSC/MPP1, MPP3, and MPP4 populations. Lastly, HSF1 was identified as an upstream regulator of deregulated genes in young HSC/MPP1, MPP3, and MPP4 and aged MPP3 (Fig. 6F).

Taken together, these results are in line with our bulk RNA-seq analysis and further show that the MPP3 population may undergo transcriptional changes similar to the HSC population, especially in terms of proteostasis and protein translation.

Autophagy has been implicated in the maintenance of protein homeostasis in response to stress^62,63. Additionally, PERK signaling has been shown to lead to an induction of autophagy³⁵. Recently, it has been shown that quiescent HSCs have a high autophagic flux, while aging has also been associated with impaired autophagy^37,39,64. Consistent with these results, we observed a significant increase in the autophagic flux in aged Mda5^-/- HSCs compared to WT (Fig. 8E). Taken together, these data indicate that aged Mda5^-/- HSCs retain enhanced proteostasis and autophagic activity.

As a master regulator of protein homeostasis, HSF1 is activated under stress conditions and translocates into the nucleus to trigger a transcriptional response^30,31. To evaluate HSF1 localization and abundance, we performed immunofluorescence staining on isolated young, middle-aged, and aged WT and Mda5^-/- HSCs. HSF1 levels were similar in young WT and Mda5^-/- HSCs, increased in middle-aged WT but not Mda5^-/- HSCs, and significantly increased in aged Mda5^-/- HSCs compared to age-matched WT counterparts (Fig. 8F, G). As we did not observe changes in HSF1 transcript levels upon aging (Supplementary Data 5), our results agree with Kruta et al.³³, in which they described HSF1 upregulation, at least at the protein level, upon aging and already in middle-aged cells. To further strengthen these results, we performed a similar staining using an antibody against phosphorylated HSF1 (pSer326), which marks HSF1 activation. Similarly, we observed reduced accumulation of phospho-HSF1 in middle-aged Mda5^-/- HSCs. Increased levels of phospho-HSF1 were detected in aged Mda5^-/- HSCs, albeit this result was not significant (Supplementary Fig. 6E, F).

To investigate whether MDA5 can retain HSF1 in the cytoplasm, we overexpressed MDA5 in HEK293T cells and stained for HSF1 to assess its localization. In HEK293T cells, HSF1 is localized in the nucleus. We observed that overexpression of full-length MDA5 was enough to keep a portion of HSF1 in the cytoplasm (Supplementary Fig. 7G, H). However, HSF1 was not tethered in the cytoplasm upon overexpression of a mutant that lacks both CARD domains (Supplementary Fig. 7G, H). Lastly, proximity ligation assay revealed that MDA5 and HSF1 are found in proximity (Fig. 8H), suggesting a potential interaction between these two proteins. To our surprise, we observed both nuclear and cytoplasmic staining for the proximity ligation assay, hinting at the fact that MDA5 may also be in the nucleus.

Because HSF1 protein levels were increased and localized in the nucleus in Mda5^-/- HSCs, we reasoned that HSF1 activation could lead to the alteration of Mda5^-/- and WT phenotypes, respectively. HSF1A is a small molecule inhibitor of TRiC chaperonin (chaperonin that keeps HSF1 in the cytoplasm) and functions as HSF1 activator (HSF1A), by preventing TRiC binding to HSF1, thus permitting its trimerization and translocation to the nucleus^33,65. Therefore, we reasoned that the HSF1A inhibitor could induce aged WT HSCs to behave like aged Mda5^-/- HSCs. Indeed, HSF1A treatment enhanced aged WT HSC quiescence by decreasing their ability to divide in culture (Fig. 8I). In addition, we performed serial colony-forming unit (CFU) assays by plating HSCs in methylcellulose and replating a similar number of cells every week. This assay evaluates the in vitro repopulation capacity of HSCs. Indeed, the inhibitor of HSF1 that made the cells more quiescent led to the enhanced capacity of aged WT HSCs to produce colonies after the third replating, suggesting that they can keep their in vitro self-renewal capacity (Fig. 8J).

These data suggest that a lack of Mda5 can delay the aging process in the HSC compartment by retaining effective proteostasis through HSF1 engagement and reduction of protein translation.

All mouse experiments were carried out in accordance with the guidelines of the Federation of European Laboratory Animal Science Association and following legal approval of the Regierungspräsidium Freiburg and the Animal Care and Use Committee of the French authorities-Ciepal Azur (35-9185.81/G-18/127, 35-9185.81/G-22/058, PEA845 or apafis 2022062916507603_v5). All of the animals were maintained at the animal facility of the Max Planck Institute of Immunobiology and Epigenetics or at the IRCAN facility under specific pathogen-free conditions in individually ventilated cages with a light-dark cycle of 12–12 h at 20–24 °C under 45–65% humidity. For all genotypes, gender matched female or male mice were used in the experiments, and experimental and control mice were co-housed in the same facility rooms. Mice were euthanized with cervical dislocation or CO₂ followed by cervical dislocation. HSC transplantations were performed at St. Jude Children’s Hospital. The transplantation procedure was performed according to protocols approved by the St. Jude Children’s Research Hospital Institutional Animal Care and Use Committee, under animal protocol #3096.

Mda5^-/- mice or B6.129X1(C)-Ifih1^tm1.1Cln/J were purchased from Jackson Laboratory Strain #:015812. CD45.1 BL6 mice were bred in MPI-IE unless otherwise stated (Jax Strain #:002014). CD45.1 JAXBoy mice were purchased from Jackson Laboratory (C57BL/6J-Ptprc^em6Lutzy/J, JAX stock #033076).

The following antibodies were purchased from BioLegend and used at a dilution 1:500 dilution unless stated otherwise: anti-CD45.2/Ly5.2 (Pacific Blue or FITC, 104); anti-CD45.1/Ly5.1 (Alexa Fluor 700 or Pe/Cy7, A20); anti-CD45 (FITC, 1:1000, 30-F11); anti-CD45R/B220 (BV650 or Alexa Fluor 700 or FITC or PE/Cy7 or biotin, RA3-6B2); anti-Ly6G/Ly6C (Gr1, BV650 or PE/Cy7 or APC, 1:1000, FITC or biotin, 1:1600, RB6-8C5); anti-CD11b (BV650 or PE/Cy7 or APC/Cy7, 1:1000, FITC or biotin, 1:1600, M1/70); anti-TER119 (BV650 or PE/Cy7, 1:1000, FITC or biotin, 1:1600, TER119); anti-CD3ε (BV650 or PE/Cy7, 1:1000, FITC or biotin, 1:1600, 145-2C11); anti-NK-1.1 (FITC, PK136); anti-CD19 (FITC, 1D3/CD19; anti-CD4 (PE/Cy5 or FITC, 1:1000, RM4-5); anti-CD8a (PE/Cy5 or FITC, 1:2000, 53-6.7); anti-CD117 (cKit, BV711, 1:1000, 2B8); anti-Ly-6A/E (Sca1, PE/Cy7 or APC/Cy7, 1:400, E13-161.7); anti-CD201 (EPCR, PE or APC, 1:200, RCR-16); anti-CD150 (SLAM, PE/Dazzle™ 594 or BV421, 1:400, TC15-12F12.2); anti-CD48 (PE/Cy7 or APC/Cy7, 1:400, or BV421, 1:1000, HM48-1); anti-CD34 (FITC, 1:50, SA376A4); anti-CD135/Flk2 (PE or APC, 1:200, A2F10); anti-Ki67 (FITC or PE, 1:100, 16A8); anti-CD16/32 (APC or PE, 1:1000, 93); anti-CD127 (IL-7Rα, APC or PE, 1:1000, A7R34); goat anti-mouse IgG, IgM (H + L) secondary (1:2500, Alexa Fluor 488, A-10680, Invitrogen) See also Supplementary Table. 1

Throughout the text, HSCs refer to EPCR SLAM cells (Lin⁻EPCR⁺CD150⁺CD48⁻ unless otherwise stated. LSK SLAM: Lin⁻Sca1⁺cKit⁺(LSK) CD150⁺CD48⁻; LT-HSCs: LSKCD150⁺CD48⁻CD34⁻CD135⁻; MPP1: LSKCD150⁺CD48⁻CD34⁺CD135⁻; MPP2: LSKCD150⁺CD48⁺CD34⁺CD135⁻; MPP3: LSKCD150⁻CD48⁺CD34⁺CD135⁻; MPP4: LSKCD150⁻CD48⁺CD34⁺CD135⁺.

Tibiae, femurs, and hip bones were isolated and crushed in PBS. Cells were washed with FACS buffer (PBS, 2% FCS, 1 mM EDTA) and counted Vi-cell XR counter (Beckman Coulter).

-For lineage quantification, HSPC characterization, 3 × 10⁶ cells were stained with antibody mix in FACS buffer for 30 min at 4 °C. Sample acquisition was performed using a Fortessa FACS analyzer (BD Biosciences). All data were analyzed using FlowJo (BD) software.

-For sorting, samples were enriched by lineage depletion using a biotin-conjugated lineage antibodies cocktail (CD3ε, CD11b, CD45R/B220, Ly6G/Ly6C, TER-119) for 20 min at 4 °C. Streptavidin nanobeads (MojoSort, 480016, BioLegend) were added for 20 min at 4 °C, with subsequent magnetic separation for 5 min at room temperature. Enriched samples were stained with an antibody mix in FACS buffer against EPCR, CD150, and CD48 for 30 min at 4 °C. Cells were washed and resuspended in 1 ml FACS buffer, and sorting was performed on a BD FACSAriaIII or a BD FACSAriaFusion (BD Biosciences).

HSCs or lineage-depleted cells were cultured in StemPro-34 SFM medium with 2.5% StemPro-34 Supplement (Gibco), 50 ng/ml mSCF, 25 ng/ml mTPO, 30 ng/ml mFlt3L, 1% penicillin-streptomycin, and 2 mM l-glutamine.

HSCs were single sorted on Terasaki microtest plates (654102, Greiner) in the medium described above. For HSF1 treatment, DMSO or HSF1A (8 μM final concentration, Axon 1890, Axon MedChem) was added to the medium before sorting. The frequencies presented in the single cell division assay plots were calculated for each Terasaki plate separately and independently as follows: Frequency of HSCs that did not divide = number of wells with one sorted HSC that did not divide/total number of wells in the Terasaki plate with live sorted HSCs Frequency of HSCs that divided at least once = number of wells with sorted HSC that did divide/total number of wells in the Terasaki plate with live sorted HSCs

Empty wells or wells containing dead cells were not included in the frequency calculations. Fisher’s exact test was performed considering the total number of wells across all different independent experiments.

Samples were isolated and enriched as described above. 3 × 10⁶ lineage-negative cells were stained with the antibody mix in FACS buffer for 30 min at 4 °C to identify HSCs. Then, cells were washed and resuspended in fixed intracellular Fixation Buffer (00-8222-49, ThermoFisher Scientific) for 10 min at 4 °C. Samples were then washed and resuspended in permeabilization buffer (00-8333-56, ThermoFisher Scientific) with anti-Ki67 antibodies for 2 h at 4 °C. Lastly, cells were washed and resuspended in PBS with Hoechst 33258 (H3569, Life Technologies) at room temperature for at least 15 min before acquisition. Sample acquisition was performed using a Fortessa FACS analyzer (BD Biosciences). All data were analyzed using FlowJo (BD) software.

Two hundred HSCs were sorted in 1 ml Mouse MethoCult GF medium (M3434, StemCell Technologies) and plated in 35 mm petri dishes. DMSO, HSF1A (8 μM final concentration, Axon 1890, Axon MedChem) was added to the medium before each plating. Colonies were counted after 5–7 days at 37 °C and 5% CO₂. For replating, cells were washed with PBS, counted, and replated (10,000 cells of each replicate) in MethoCult GF (M3434, StemCell Technologies). Colonies were quantified after 4–5 days, after the second and third replating.

Aged WT and Mda5^-/- mice (CD45.2⁺/Ly5.2) were used as donors, young WT mice (CD45.1⁺/Ly5.1) were used as competitors in competitive transplantations and as recipients in non-competitive transplantations. Young WT mice (CD45.1⁺/CD45.2⁺) were used as recipients in competitive transplantations. We calculated the HSC frequency of the donors and intravenously injected the volume of bone marrow containing 300 HSCs into lethally irradiated (9.5 Gy) recipients. In competitive transplantations, we additionally injected 500,000 competitor-derived bone marrow cells. Peripheral blood chimerism was checked every 4 weeks after transplantation for 16 weeks. Briefly, 20 μl of blood was obtained from the submandibular vein using a lancet. Erythroid cells were lysed in ammonium-chloride-potassium (ACK, BP10-548E, Lonza Bioscience) for 5 min at room temperature. Cells were washed with PBS and stained for 30 min at 4 °C with anti-CD45.1/Ly5.1, anti-CD45.2/Ly5.2, anti-CD4, anti-CD8a, anti-CD11b, anti-Ly6G/Ly6C, anti-CD45R/B220, and anti-Ter119 antibodies. After washing and resuspending in FACS buffer, samples were acquired on a Fortessa FACS analyzer (BD Biosciences). All data were analyzed using FlowJo (BD) software.

Transplantations were performed using the CD45.1/CD45.2 congenic system. Recipient CD45.1 JAXBoy mice (C57BL/6J-Ptprc^em6Lutzy/J, JAX stock #033076) (10 weeks old) were lethally irradiated with 11 Gy delivered in a split dose of 5.5 Gy each, at least 3 h apart, using a gamma irradiator with a 137Cs source. Recipients were treated with Sulfamethoxazole/Trimethoprim in drinking water for 7 days before the transplant and for 21 days post-transplant. On the day of the transplant, 100 FACS-sorted donor HSCs (Lin-Sca1⁺c-kit⁺CD135⁻CD48⁻CD150⁺) were injected into the tail vein along with 2 × 10⁵ of whole bone marrow cells isolated from age-matched CD45.1 JAXBoy mice. Hematopoietic reconstitution in peripheral blood was monitored by flow cytometry every 4 weeks, and the primary recipients were euthanized 16 weeks post-transplant. For the secondary transplantation, bone marrow of primary recipients injected with HSCs from the same donor mice was pooled, and 3 × 10⁶ of pooled whole bone marrow cells were delivered via tail vein injections into lethally irradiated CD45.1 JAXBoy secondary recipients. Hematopoietic reconstitution in peripheral blood was monitored by flow cytometry every 4 weeks. The secondary recipients were euthanized 16 weeks post-transplant. The following cell populations were analyzed in the peripheral blood and bone marrow of transplant recipients using flow cytometry: granulocytes (Mac1⁺Gri^hi), monocytes (Mac1⁺Gri^lo), B cells (CD19⁺), T cells (CD3⁺), LSK (Lin⁻Sca1⁺c-kit⁺), SLAM LSK (Lin⁻Sca1⁺c-kit⁺CD135⁻CD48⁻CD150⁺), MPP2 (Lin⁻Sca1⁺c-kit⁺CD135⁻CD48⁺CD150⁺), MPP3 (Lin⁻Sca1⁺c-kit⁺CD135⁻CD48⁺CD150⁻), MPP4 (Lin⁻Sca1⁺c-kit⁺CD135⁺CD150⁻). Lineage (Lin) cocktail contained the following antibodies: CD19, B220, CD3, Mac1 (CD11b), Gr1 (Ly-6G/Ly-6C), CD11c, and Ter119. The following antibodies were used for flow cytometry: APC anti-mouse/human CD11b (Biolegend, 101212, M1/70, 1/250), APC anti-mouse Ly-6A/E (Sca-1) (Biolegend, 108112, D7, 1/250), APC/cyanine7 anti-mouse CD19 (Biolegend, 115529, 6D5, 1/250), brilliant violet 421 anti-mouse CD117 (c-Kit) (Biolegend, 105828, 2B8, 1/250), brilliant violet 421 anti-mouse CD3ε (Biolegend, 100341, 145-2C11, 1/250), brilliant Violet 605 anti-mouse CD150 (SLAM) (Biolegend, 115927, TC15-12F12.2, 1/1000), brilliant violet 605 anti-mouse CD19 (Biolegend, 115539, 6D5, 1/250), brilliant violet 605 anti-mouse CD45.1 (Biolegend, 110737, A20, 1/1000), brilliant violet 605 anti-mouse CD45.2 (Biolegend, 109841, 104, 1/250), eBioscience Fixable Viability Dye eFluor 780 (Thermofisher, 65-0865-14, 1/2000), FITC anti-mouse CD11c (Biolegend, 117306, N418, 1/250), FITC anti-mouse CD19 (Biolegend, 152404, 1D3/CD19, 1/250), FITC anti-mouse CD3ε (Biolegend, 100306, 145-2C11, 1/250), FITC anti-mouse CD45.2 (Biolegend, 109806, 104, 1/250), FITC anti-mouse Ly-6G/Ly-6C (Gr-1) (Biolegend, 108406, RBC6-8C5, 1/500), FITC anti-mouse TER-119 (Biolegend, 116206, TER-119, 1/250), FITC anti-mouse/human CD11b (Biolegend, 101206, M1/70, 1/250), FITC anti-mouse/human CD45R/B220 (Biolegend, 103206, RA3-6B2, 1/250), PE anti-mouse CD135 (Biolegend, 135306, A2F10, 1/500), PE anti-mouse Ly-6G/Ly-6C (Gr-1) (Biolegend, 108408, RBC6-8C5, 1/2000), PE/cyanine7 anti-mouse CD45.2 (Biolegend, 109830, 104, 1/250), and PerCP/cyanine5.5 anti-mouse CD48 (Biolegend, 103422, HM48-1, 1/500). See also Supplementary Table 1.

The LEGENDplex mouse inflammation panel (740150, BioLegend) was used according to the manufacturer’s instructions. Briefly, after crushing the bones, the cell suspension was centrifuged at 1500 rpm for 5 min at 4 °C. The bone marrow supernatant was then collected in a separate tube and stored at −80 °C until further use. Samples were diluted 1:1 in PBS, incubated with beads conjugated with the respective antibodies, and analyzed on a Fortessa FACS analyzer (BD Biosciences). All data were analyzed using FlowJo (BD) software.

2 × 10⁶ lineage-depleted cells were stained with the antibody mix in FACS buffer for 30 min at 4 °C to identify HSCs. Cells were washed and resuspended in 1% PFA for 10 min at room temperature. Samples were then washed and incubated overnight with 1:500 primary pan-ubiquitin antibody (clone FK2, 04-262, Sigma) in permeabilization buffer (00-8333-56, ThermoFisher Scientific). Then, cells were washed and resuspended with 1:2500 goat anti-mouse IgG1 secondary antibody in permeabilization buffer at 4 °C for at least 2 h. After washing and resuspending in FACS buffer, samples were acquired on a Fortessa FACS analyzer (BD Biosciences). All data were analyzed using FlowJo (BD) software. For measurement of misfolded proteins in G0 HSCs cells were washed and resuspended with 1:2500 goat anti-mouse IgG1 secondary antibody in permeabilization buffer at 4 °C for 2 h, along with anti-Ki67 antibody at 4 °C for at least 2 h (H3569, Life Technologies). After washing and resuspending in FACS buffer, samples were acquired on a Fortessa FACS analyzer (BD Biosciences). All data were analyzed using FlowJo (BD) software.

For total protein concentration measurement, 1000 EPCR SLAM cells were sorted from young and aged WT and Mda5^-/- mice. Cells were lysed in RIPA buffer (89900, ThermoFisher) with protease inhibitor (1x) for 10 min at 4 °C. Samples were centrifuged at 13,000 × g for 15 min to pellet the cell debris. The protein concentration in the supernatant was measured by Qubit 3.

Tetraphenylethene maleimide^39,61 (TPE-MI, HY-143218, MedChemExpress) was reconstituted according to the manufacturer’s instructions, diluted in PBS (50 mM final concentration), and added to each sample. 2 × 10⁶ lineage-depleted cells were incubated for 45 min at 37 °C. Samples were washed and stained with the antibody mix in FACS buffer for 30 min at 4 °C to identify HSCs. Cells were washed and acquired on a Fortessa FACS analyzer (BD Biosciences). All data were analyzed using FlowJo (BD) software.

2 × 10⁶ lineage-depleted cells were resuspended in complete StemPro-34 SFM medium with either MitoTracker Green (MTG, 50 nM final concentration, M7514, ThermoFisher Scientific) or tetramethylrhodamine methyl ester perchlorate (TMRM, 100 nM final concentration, I34361↗, ThermoFisher Scientific), supplemented with verapamil (V4629, Sigma Aldrich). After 30 min incubation at 37 °C, samples were washed and stained with the antibody mix in FACS buffer with verapamil for 30 min at 4 °C to identify HSCs. Cells were washed and acquired on a Fortessa FACS analyzer (BD Biosciences). All data were analyzed using FlowJo (BD) software.

2 × 10⁶ lineage-depleted cells were stained with the antibody mix in FACS buffer for 30 min at 4 °C to identify HSCs. Then, cells were resuspended in complete StemPro-34 SFM medium (10639011ThermoFisher) with OP-Puro at 50 μM and incubated at 37 °C for 1 h. Samples were then washed, fixed with 1% PFA in PBS for 10 min at 4 °C, and permeabilized with permeabilization buffer (00-8333-56, ThermoFisher Scientific) for 20 min at 4 °C. The azide-alkyne cycloaddition was performed using the Click-iT Plus OPP Alexa Fluor 488 Kit (C10456↗, ThermoFisher Scientific), according to the manufacturer’s instructions. After the 30 min reaction, the cells were washed with PBS and acquired on a Fortessa FACS analyzer (BD Biosciences). All data were analyzed using FlowJo (BD) software.

This assay was performed using the CYTO-ID autophagy detection kit (ENZ-KIT175-0050, ENZO Life Sciences) according to manufacturer’s instructions. Briefly, 2 × 10⁶ lineage-depleted cells were resuspended with CYTO-ID green detection reagent diluted in assay buffer and incubated for 30 min at 37 °C. Samples were then stained with the antibody mix in FACS buffer for 30 min at 4 °C to identify HSCs. After washing, cells were resuspended in FACS buffer and acquired on a Fortessa FACS analyzer (BD Biosciences). All data were analyzed using FlowJo (BD) software.

For low-input targeted metabolomics (of polar metabolites), 5000 HSCs were isolated as described above. Samples were then washed with 1 mL of 9 g/L NaCl and filtered through a 30 mM cell strainer before proceeding immediately with sorting using 2 g/L NaCl in deionized water as sheath liquid. Cells were sorted directly into a 96-well plate containing 25 μL of extraction solution, which consisted of 0.1% ¹³C internal standard stock (1 aliquot of U-¹³C yeast extract (Isotopic solutions, Cat# ISO-1) in 10 mL of 25% v/v methanol in water) in acetonitrile (LC-MS grade). To assess background levels, 5000 cell debris (low light-scattering signals) were collected. Targeted metabolite quantification by LC-MS was carried out as previously described⁹⁰, and data processing was performed using the R package automRm⁹¹. Metabolites were only included if they were detected above the cell debris detection levels, and the retention time was identical to the ¹³C yeast standard qualifier peak. The metabolic differences were calculated by determining the area under the curve (AUC). Data were normalized to the mean of every detected metabolite in the aged WT population per experiment.

HSCs were isolated and sorted as previous described. Cells were loaded into cytospin (700 rpm, 5 min) to allow them to adhere to glass slides. Fixation was performed with 4% PFA in PBS for 10 min at room temperature, slides were air-dried and kept at +4 °C until the next step. Permeabilization and blocking steps were done with 0.1% Triton X-100 and 10% goat serum (Sigma G9023), respectively. Indirect immunofluorescent analysis was conducted using primary anti-HSF1 antibodies (Enzo Life Sciences ADI-SPA-901-D, 1:500 or StressMarq SMC-118D for Rat Ab 1:500), anti-dsRNA (Sigma Aldrich MABE1134, 1:500), phosphorylated HSF1 (Enzo Life Sciences [pSer326]HSF1 ADI-SPA-902-D 1:500) and phosphorylated EIF2a (cell signaling phospho-EIF2a (Ser51) 9721S 1:500) and secondary anti-rabbit or anti-mouse antibodies conjugated with AlexaFluor-488 dye (ThermoFisher Scientific,1:1000). Samples were mounted with Fluroshield mounting solution containing DAPI (Sigma F6057). Images were acquired using an LSM880 confocal microscope (Carl Zeiss, Germany).

HEK293T cells are cultured in IMDM (10% FBS, 2% penicillin, streptomycin, and glutamine). 20.000 cells were plated in each well of 12-well plates, on coverslips. pFLAG_WT MDA5 (pFLAG_WT MDA5 was a gift from Sun Hur (Addgene plasmid # 216798; http://n2t.net/addgene:216798↗; RRID:Addgene_216798) and pFLAG_MDA5∆2CARD (pFLAG_MDA5∆2CARD was a gift from Sun Hur (Addgene plasmid # 216800; http://n2t.net/addgene:216800↗; RRID:Addgene_216800) were transfected using Lipofectamine overnight according to the manufacturer’s instructions. Fixation was performed with 4% PFA in PBS for 20 min at room temperature. Permeabilization and blocking steps were done with 0.1% Triton X-100 and 1% BSA, respectively. Indirect immunofluorescent analysis was conducted using primary anti-HSF1 antibodies (Enzo Life Sciences ADI-SPA-901-D,1:1000) and AlexaFluor-488 dye (ThermoFisher Scientific,1:1000). Samples were mounted with Fluroshield mounting solution with DAPI (Sigma F6057). Images were acquired using an LSM880 confocal microscope (Carl Zeiss, Germany).

Human BJ fibroblasts, untreated or treated with 24 h at 10 μg/mL PolyI:C, were fixed with 4% PFA for 20 min at 37 °C and permeabilized with 2% Triton X-100 for 10 min at room temperature. PLA was performed according to the manufacturer’s protocol (Sigma Aldrich DUO94104). Primary antibodies were used at the following dilutions: anti-HSF1 (Santa Cruz sc-17757, 1:100), anti-MDA5 (Abcam ab79055, 1:200). Cells were mounted with DAPI-containing mounting medium (Sigma Aldrich DUO82040). Interaction between HSF1 and MDA5 was visualized in situ using a Carl Zeiss LSM880 confocal microscope (Carl Zeiss). Presented images represent the maximum intensity projection of the z-stack.

Two thousand five hundred EPCR SLAM cells were sorted from young, middle-aged, and aged WT and Mda5^-/- mice. RNA was then isolated using the Arcturus PicoPure RNA isolation Kit (KIT0204, Applied Biosystems). Low-input RNA-seq libraries were prepared using SMART-Seq Stranded Kit (634442, Takara Bio). First, random priming allowed the generation of cDNA from all RNA fragments in the sample. Next, a first round of PCR amplification (5 cycles) added full-length Illumina adapters (Indexing Primer Set HT for Illumina v2), including barcodes. Then, the ribosomal cDNA (originating from rRNA) was cleaved. The cDNA fragments were enriched via a second round of PCR amplification (14 cycles) using primers universal to all libraries. Libraries were cleaned up using AMPure XP beads (A63881) following the manufacturer’s protocol. Libraries were sequenced at Vanderbilt.

Paired-end 300 bp reads were generated using the Illumina NovaSeq 6000 system and sequenced at a depth of 48–70 M. Raw fastq read files were assessed for quality with FastQC v0.11.9 (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/↗) and trimmed of Illumina sequencing adapters with Trim Galore v0.3.7 (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/↗). Alignment to the reference mouse genome (GENCODE GRCm38.p6 release M23) was done using STAR v2.7.10a⁹² with default parameters. Aligned bam files were indexed with samtools v1.12⁹³, and gene count matrices for each sample were generated from the alignment bam files using HTSeq-count v0.12.4⁹⁴ with the parameter –intersection-non-empty, and –stranded = “reverse” was specified for RNA libraries prepared with the Takara directional kit. Samples were hierarchically and PCA clustered to determine sample congruence. Normalization, scaling, and differential expression analysis were performed in R v4.1.3 (R Core Team, 2022) using the edgeR v3.34.1⁹⁵ qlf test for each pair-wise comparison. Genes were considered significantly and differentially expressed between conditions if they met FDR and fold-change cutoffs of 0.05 and 1.5, respectively. Plotting of results was performed using heatmap.2 function of the gplots package v3.1.1 and ggplot2 v3.3.5⁹⁶ packages in R.

Read generation and data pre-processing were performed as described above for RNA-seq analysis of genes. Alignment to the reference mouse genome (GENCODE GRCm38.p6 release M23) was done using STAR v2.7.10a⁹² with the following parameters to retain multi-mapping reads: --outFilterMultimapNmax 100 --winAnchorMultimapNmax 100 --outMultimapperOrder Random --outSAMmultNmax 1. Alignment BAM files were indexed with samtools index⁹³. Expression of transposable element families was quantified by sample using TECounts v2.0.3⁹⁷ with parameter –mode = multi to estimate transposable element abundances from multimapped alignments, and –stranded reverse to properly process the stranded reads, using the pre-generated GENCODE GRCm38 TE GTF RepeatMasker file from the Hammel lab FTP site (https://labshare.cshl.edu/shares/mhammelllab/www-data/TEtranscripts/TE_GTF↗) and the GENCODE GRCm38 v23 annotation GTF. Counts were assembled into a single table in R, were normalized to counts per million (CPM), and differential expression was assessed in R v3.1.4 (R Core Team, 2022) on the combined transposable element /gene counts using the edgeR qlf test, as described above, determining TEs to be differentially expressed if they met the cutoffs of FC 1.5 and FDR 0.05.

Ingenuity pathway analysis (IPA) for upstream regulators and canonical pathways was performed on log-transformed differential expression values for genes found to be significantly differentially expressed in edgeR, using the default settings of IPA v.24.0.2⁵⁵.

Normalized logCPM (counts per million) matrices for all genes of each comparison were extracted from the edgeR DGE object in R, generated as described above, and saved as a tab-separated text file for input into GSEA (v4.4.0 build 18)^98,99. GSEA was run using the gene_set permutation with the max and min set sizes specified at 1000 and 1, respectively. Gene sets tested were obtained from MSigDB (https://www.gsea-msigdb.org/gsea/msigdb↗)¹⁰⁰ and^85,101,102.

100,000 LSK cells from young and aged WT or Mda5^-/- mice were isolated and sorted as previously described. Samples were then barcoded using the 10x Genomics 3′ CellPlex Kit. Cell multiplexing was performed before sorting. Pooled gene expression and feature barcoding fastq files obtained via NovaSeq 6000 sequencing platform were de-multiplexed and aligned to the GRCm38/mm10 reference genome using cellranger count¹⁰³. The resulting gene expression, barcode, and hashtag matrices were imported into R v4.1.3, and analysis was performed using the Seurat package v4.1.0¹⁰⁴. Cells were filtered for nFeature_RNA > 1000 and a mitochondrial gene percentage of at most 5%. Hashtag data was used to trace pooled cells of a sample back to individual mice via the HMOdemux function in Seurat. Of 38,137 WT aged cells output by cellranger count, 21,241 were singlets, and 12,616 cells were retained after filtering for further analysis. Of 34,017 Mda5^-/- aged cells output by cellranger count, 18,248 were singlets, and 12,084 cells were retained after filtering for further analysis. Of 32,441 WT young cells output by cellranger, 18,788 were singlets, and 9986 cells were retained after filtering for further analysis. Of 34,636 Mda5^-/- young cells output by cellranger count, 14,007 were singlets, and 8178 cells were retained after filtering for further analysis. Samples were individually regularized using SCTransform V2 before Seurat anchor-based integration using 2000 anchors. Based on individual gene expression results, aged sample replicates CMO303 (WT) and CMO306 (Mda5^-/-) were found to be unfit for further analysis and were excluded. After PCA dimensionality reduction, an elbow plot was used to determine the optimal number of dimensions to proceed with (30), and SNN clustering was performed at several resolutions (0.1–1, increment of 0.1) for input into clustree¹⁰⁵ to determine the optimal clustering resolution. Differential expression between clusters was performed with the FindAllMarkers function, and cell populations were determined based on the expression of HSC and MPP markers⁵⁶. To perform a comparison at the sample level, pseudobulk analysis was carried out by collapsing cell population clusters by condition and differential expression analysis in edgeR qlf test⁹⁵. Functional enrichment, GSEA, and IPA analysis were performed on the results as described for RNA-seq analysis above.

For ATAC-seq (3000–5000 cells per sample, 3 biological replicates per condition), HSCs were sorted in FACS buffer. HSCs were spun down at 500 g, 4 °C for 10 min, and incubated with Illumina TAGMENT DNA enzyme and buffer (20034210, Illumina) or with tagmentase (Diagenode, C01070012) and the respective buffer (Diagenode C01019043) and Digitonin (0.01%) for 30 min at 37 °C on a thermomixer shaking at 700 rpm. Transposed DNA was purified using MinElute Cleanup (28004, QIAGEN). Adapter indices were ligated using the Nextera XT Index Kit (15055289, Illumina), and libraries were amplified using NEBNext High-Fidelity 2X PCR Master Mix (M0541S, NEB).

Paired-end 100 bp reads were generated using the Illumina Hiseq 3000 system and sequenced at a depth of 60–96 M. Fastq read files were trimmed of Nextera adapters with Trim Galore v0.6.7 (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/↗) and aligned to GRCm38/mm10 with Bowtie2 v2.4.4¹⁰⁶. The resulting BAM files had PCR duplicates, mitochondrial reads, and ENCODE blacklist¹⁰⁷ features removed before further analysis using Picard v3.4.0 (Broad Institute, 2019), samtools v1.13⁹³, and bedtools v2.30.0¹⁰⁸, respectively. For differential accessibility analysis, BAM files were merged by sample using Samtools before calling peaks with MACS v2.2.9.1¹⁰⁹. Induced/reduced peaks were determined by overlapping peaksets with bedtools intersect¹⁰⁸, after which gene TSSs overlapping peaks (−100 kb, +50 kb) were determined with bedtools closest¹⁰⁸, and motif enrichment was obtained with findMotifsGenome.pl script from the HOMER v5.1 package¹¹⁰. For visualization, filtered BAM files were converted to bigwig format via a bedgraph intermediate with bamCoverage and bedtobigwig, using scaling factors obtained from deeptools v3.5.6¹¹¹ multiBamSummary. These bigwigs, along with the previously obtained induced and reduced peaks in bed format, were input into deeptools computeMatrix, and the output was plotted using deeptools plotHeatmap reference-point.

Footprinting analysis to determine TF occupancy was performed on sample-merged bam files (samtools merge⁹³) and sample-merged narrowpeak output from MACS2 using the Wellington algorithm from the pyDNase package v0.3.0¹¹². Output bed regions were merged, then intersected among samples in order to plot profiles with dnase_average_profile.py.

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