Hdac3-deficiency increases senescence-associated distention of satellite DNA and telomere-associated foci in osteoprogenitor cells

Abstract

Histone deacetylase 3 (Hdac3) is an epigenetic regulator of gene expression and interacts with skeletal transcription factors such as Runx2. We previously reported that conditional deletion of Hdac3 in Osterix-Cre recombinase-expressing osteoprogenitor cells (Hdac3 CKO_Osx) caused osteopenia and increased marrow adiposity, both hallmarks of skeletal aging. We also showed that Runx2+ cells within osteogenic cultures of Hdac3-depleted bone marrow stromal cells (BMSCs) contain lipid droplets (LDs). Cellular senescence, a nonproliferative metabolically active state, is associated with increased marrow adiposity, bone loss, and aging. In this study, we sought to determine if Hdac3 depleted Runx2+ pre-osteoblasts from young mice exhibit chromatin changes associated with early cellular senescence and how these events correlate with the appearance of LDs. We first confirmed that BMSCs from Hdac3 CKO_Osx mice have more Runx2 + LD+ cells compared with controls under osteogenic conditions. We then measured senescence-associated distention of satellite (SADS) DNA and telomere-associated foci (TAFs) in Hdac3 CKO_Osx and control BMSCs. In situ, Runx2+ cells contained more SADS per nuclei in Hdac3 CKO_Osx femora than in controls. Runx2+ BMSCs from Hdac3 CKO_Osx mice also contained more SADS and TAFs per nuclei than Runx2+ cells from age-matched control mice in vitro. SADs and TAFs were present at similar levels in Runx2 + LD+ cells and Runx2 + LD− cells from Hdac3 CKO_Osx mice. Hdac inhibitors also increased the number of SADS in Runx2 + LD+ and Runx2 + LD− WT BMSCs. Senolytics reduced viable cell numbers in Hdac3 CKO_Osx BMSC cultures. These data demonstrate that the depletion of Hdac3 in osteochondral progenitor cells triggers LD formation and early events in cellular senescence in Runx2+ BMSCs through mutually exclusive mechanisms.

Graphical Abstract

Graphical Abstract.

Introduction

Bone marrow adipose tissue (BMAT) is functionally unique from other fat tissues and impacts physiology in many ways.¹ For example, BMAT is a source of energy for cells, but it can also release inflammatory cytokines into the circulation.^2,3 It is well-documented that the accumulation of BMAT is associated with poor bone quality with advancing age.⁴ The amount of marrow fat is inversely related to BMD in both children and adults⁵ and directly related to the incidence of skeletal fractures,⁶ but this correlation can vary depending on the gender, species, and pathological conditions.^1,7–10 BMAT may have evolved as a mechanism for storing surplus energy and protecting bone from free fatty acids, which can impede hematopoiesis and osteoblast differentiation due to their lipotoxicity,¹¹ but its purpose and function are not fully known. BMAT levels are regulated by exercise,¹² Wnts,¹³ bone morphogenic proteins,¹⁴ Hedgehog factors,¹⁵ and glucocorticoids.¹⁶

The cellular origin of BMAT is complex and incompletely understood. Because bone marrow adipocytes (BMAd) and osteoblasts originate from mesenchymal progenitors, a prevailing view is that adipogenesis is a default program when osteogenesis is interrupted. Infiltration of adipocytes from nonosseous tissues and trans-differentiation of osteoblasts into BMAd have also been proposed.⁴ We and others showed that osteoprogenitor cells (characterized by expression of lineage-determining transcription factors Runx2 or Sp7/Osx1) can contain lipid droplets (LD).^13,17–19 This could be interpreted as evidence of trans-differentiation of preosteoblasts to early BMAd or as evidence of intracellular changes in pre-osteoblasts that promote lipid storage as a means of protection from free fatty acids.^4,13,18,19 Understanding the mechanisms responsible for LD formation in osteoprogenitors could lead to therapies that increase the number of healthy osteoprogenitors that are dedicated to forming bone during a lifespan.

Histone deacetylases (HDAC) regulate cell fate determination of mesenchymal progenitor cells by removing acetyl and other acyl groups from lysine residues in histones and other proteins,²⁰ thereby altering chromatin structure, gene expression, and the functioning of various biochemical and signal transduction pathways.²¹ HDACs influence BMSC lineage specification by interacting with cofactors and DNA-binding transcription factors. Hdac3, a functionally unique class I HDAC that interacts with the nuclear receptor co-repressor (NCOR) complex, is expressed at all stages of differentiation in osteoblasts and binds to Runt-related transcription factor 2 (Runx2), the master osteoblast transcription factor.²² Hdac3 levels decline in mice and humans during aging.^17,23 Conditional Hdac3 deletion with Osx1-Cre, Collagen2 (Col2)-Cre, and Osteocalcin (OCN)-Cre causes osteopenia, osteoblast dysfunction, and higher susceptibility to fracture.^17,24,25 Interestingly, increased BMAT accumulation is only observed in mice where Hdac3 was deleted in osteo- and chondro- progenitors (ie, Osx1- or Col2-Cre expressing cells).^17,25 Deletion of Hdac3 also increases LD formation in Runx2+ BMSC-derived bone forming cells in vitro;¹⁷ however, the mechanisms through which Hdac3 deletion in osteo-progenitor cells increases LD accumulation remain elusive.

Cellular senescence is a hallmark of aging in both animals and humans. Senescent cells residing in the bone microenvironment may promote the switch of osteoprogenitor cells toward a BMAd fate, thereby depleting osteoprogenitor cells with aging.^26–28 The accumulation of DNA damage and other cellular stresses triggers senescence in actively dividing cells^29,30 as well as in terminally differentiated cells that no longer undergo division.³¹ Senescent cells retain metabolic activity and exhibit enhanced survival capabilities, rendering them resilient and resistant to apoptosis. In late stages of senescence, leakage of nuclear and mitochondrial DNA into the cytoplasm activates the cGAS-STING pathway, which triggers the production and release of proinflammatory cytokines, chemokines, and extracellular matrix proteins, collectively generating a detrimental microenvironment known as the senescence-associated secretory phenotype (SASP).^32–34 The SASP likely plays a role in the increased accumulation of senescent cells, changes in tissue composition, and the exacerbation of tissue dysfunction.

In this study, we sought to determine how deletion of Hdac3 in osteoprogenitor cells contributes to early phases of cellular senescence as defined by the appearance of nuclear senescence-associated distention of satellite (SADS) (distensions of satellite DNA at centromeres)³⁵ and telomere-associated foci (TAFs) (DNA damage at sites of telomeres)³⁶, and if these phenotypes affect LD formation in Runx2-expressing cells. Our data demonstrate that Runx2+ BMSCs from Hdac3 CKO_Osx mice exhibit accelerated early senescence phenotypes, as evidenced by more SADS and TAFs per nuclei, than control Runx2+ BMSCs do under osteogenic conditions. However, these early hallmarks of senescent cells were present at similar levels in Runx2 + LD+ cells and Runx2 + LD− cells from Hdac3 CKO_Osx mice. These data suggest that the depletion of Hdac3 triggers LD formation and early signs of cellular senescence in Runx2+ BMSCs through mutually exclusive mechanisms.

Materials and methods

Mice

All animal research was conducted according to guidelines provided by the National Institute of Health and the Institute of Laboratory Animal Resources, National Research Council. The Mayo Clinic Institutional Animal Care and Use Committee approved all animal studies. Animals were housed in an accredited facility under a 12-h light/dark cycle and provided water and food (PicoLab RodentDiet20; LabDiet, St. Louis, MO, USA) ad libitum. Mice were maintained on a C57BL/6 J background and genotyped as described previously.²⁴ Mice with two copies of the Hdac3 allele carrying loxP sites in introns surrounding exon 7 were crossed with mice expressing Cre recombinase under control of the Osterix (Osx1) promoter,³⁷ eventually yielding two groups of progeny that were studied: Hdac3-conditional knockout animals (Hdac3 CKO_Osx = Hdac3^fl/fl; Osx1-Cre+) and control littermates (Control = Hdac3^fl/fl).²⁴ Mice were euthanized at age 4–6 wk by carbon dioxide asphyxiation. Long bones (tibias, femurs) were aseptically harvested immediately after euthanasia. We previously reported that both male and female Hdac3 CKO_Osx mice exhibit the same phenotypes of MAT accumulation, osteopenia, osteoblast dysfunction, and increased vulnerability to fractures.¹⁷ Since there are no sex differences in these phenotypes, the assays reported within this report were all performed in male mice. Female mice were needed to maintain the colony.

Isolation of mouse BMSCs and Cytospin preparations

Bone marrow was flushed from femurs and tibias of male mice at specified ages. Bone marrow isolates were incubated with red blood cell lysis buffer (Miltenyi Biotec, Auburn, CA, USA) to remove erythrocytes. The samples were then depleted of cells expressing hematopoietic lineages using Magnet Assisted Cell Sorting (MACS) and mouse anti-CD45 and anti-Ter-119 microbeads (Miltenyi Biotec). CD45- and Ter-119- depleted BMSC suspensions were immediately transferred to slides using a Shandon Cytospin 3 Centrifuge (Thermo Fisher Scientific, Waltham, MA, USA) set at 800 Rcf for 10 min.

Primary murine BMSC cultures

Flushed bone marrow cells were seeded into 6-well plates at 1 × 10⁷ cells per well or onto sterile poly D lysine-coated glass coverslips in 12-well plates at 4 × 10⁶ cells per well in basal culture medium (a-MEM, 20% FBS, 1% antibiotic/ antimycotic [Invitrogen, Carlsbad, CA, USA; #15240–062], 1% nonessential amino acids). BMSCs were selected for their ability to adhere to the plate over the first 3 d in culture. Cultures were supplemented with suberoylanilide hydroxamic acid (SAHA, obtained from MedchemExpress, Monmouth Junction, NJ, USA), RGFP966 (MedchemExpress), dasatinib (D, from Sigma-Aldrich, St. Louis, MO, USA) and quercetin (Q, from Sigma-Aldrich), or vehicle (0.1% DMSO) as indicated in figures. Cultures lasted for 24 h with 0.2 μM D + 20 μM Q and for 48 h with 1 μM SAHA and RGFP966. Media were then changed to osteogenic medium (basal culture medium +50 μg/mL ascorbic acid, 10 mM beta glycerol phosphate, ±10⁻⁷ M dexamethasone) to induce osteoblastic differentiation, with fresh media changes every other day for up to 21 d.

RNA isolation and real-time PCR

Total RNA was extracted from cells using TRIzol (Invitrogen, Carlsbad, CA, USA) and chloroform and 1 μg was reverse transcribed using the iScript cDNA Synthesis Kit (Bio-Rad Laboratories, Hercules, CA, USA). Resulting cDNAs were used to assay gene expression via real-time PCR with gene-specific primers. Fold changes in gene expression were calculated using the delta–delta comparative threshold cycle algorithm (2^−ΔΔCt) method relative to control after normalization of gene-specific Ct values to Gapdh Ct values.

Gene-specific primer sequences are as follows: Gapdh_Forward (F): 5′-GGGAAGCCCATCACCATCTT-3′, Gapdh_Reverse (R): 5′-GCCT CACCCCATTTGATGTT-3′; Hdac3_F: 5′-CCCGCATCGAGAATCAGA AC-3′, Hdac3_R: 5′-TCAAAGATTGTCTGGCGGATCT-3′; Pparγ_F: 5′-CCCACCAACT TCGGAATCAG-3′, Pparγ_R: 5′-AATGCGAGTG GTCTTCCATCA-3′; Plin1_F: 5′-TGCTGCA CGTGGAGAGTAAG-3′, Plin1_R: 5′-TGGGCTTCTTTGGTGCTGTT-3′; Cidec_F: 5′-TCCAAGCCCTGGCAAAAGAT-3′, Cidec_R: 5′-CGGAG-CATCTCCTTCACGAT-3; Lipe_F: 5′-AGAAGGATCGAAGAACCGCA-3′; and Lipe_R: 5′-GTGTGAGAACGCTGAGGCTTT-3′; Runx2_F: GGCACAGACAGAAGCTTGATG, Runx2_R: GAATGCGCCCTAAATCACTGA; Bglap_F: CCTGAGTCTGACAAAGCCTTCA, Bglap_R: GCCGGAGTCTGTTCACTACCTT; Alpl_F: CACAGATTCCCAAAGCACCT, Alpl_R: GGGATGGAGGAGAGAAGGTC; P16_F: GAACTCTTTCGGTCGTACCC, P16_R: AGTTCGAATCTGCACCGTAGT; P21_F: GAACATCTCAGGGCCGAAAA, P21_R: TGCGCTTGGAGTGATAGAAATC.

Cell confluency and viability assays

Primary BMSCs (1 × 10⁵ cells/well) were cultured in a 24-well plate for 3 d and then D + Q or vehicle was added as described above. Cell confluency was detected in real-time with the IncuCyte S3 Live Cell Analysis System (Roche Applied Sciences, Indianapolis, IN, USA), with 9 captures per well every 2 h for 24 h. For viability assay, cells (2 × 10³ cells/well) were cultured in 96-well plates for 24 h with D + Q. CellTiter 96® Aqueous One Solution Cell Proliferation Assays (Promega, Madison, WI, USA) were performed according to manufacturer specifications. Briefly, 20 μL of a solution containing the compound 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS) was introduced into each well and left to incubate for 30 min at a temperature of 37 °C. Following this incubation, measurements were made using a spectrophotometer at an optical density of 490 nm.

IF and LD staining

Cells were fixed onto coverslips with 2% PFA for 10 min in room temperature (RT) and then permeabilized using a 0.5% solution of Triton X-100 in PBS for 10 min. Next, they were incubated in immunofluorescence (IF) blocking buffer (10% normal goat serum, 1% BSA in TBS) for 1 h at RT. Primary antibodies to Runx2 (Rabbit, 1:100) or Hdac3 (mouse, 1:100, Cell signaling, Danvers, MA, USA) were applied in IF blocking buffer for 2 h at RT, followed by a 1-h incubation at RT with Alexa fluor-647 or -488-conjugated goat secondary antibodies (Life Technologies, Grand Island, NY, USA). After 3 washes with TBS, cells were stained with a neutral lipid stain dyes (HCS LipidTOX, Invitrogen, Waltham, MA, USA). Coverslips were mounted on glass slides with Vectashield-DAPI containing mounting media (Vector Laboratories Inc, Newark, CA, USA). Images were obtained using inverted laser scanning confocal microscope (LSM 780; Carl Zeiss Microscopy, Jena, Germany) with in-depth Z stacking.

SADS assay

Cells (on coverslips) were fixed with 2% PFA for 10 min at 37 °C, washed 3 times with PBS, and stored at −80 °C until the staining was performed. For in vivo analysis, the decalcified femurs of mice were fixed, embedded in paraffin, and sectioned for use. For fluorescent in situ hybridization (FISH) staining, cells were crosslinked with 4% PFA for 30 min, washed 3 times in PBS for 5 min each time, and dehydrated in ice-cold graded EtOH as follows: 70% EtOH (for 1 min), 90% EtOH (for 1 min), and 100% EtOH (for 1 min). Cells were then briefly air dried and incubated for 20 min at 85 °C in hybridization buffer (70% formamide [Sigma-Aldrich, Burlington, MA, USA], 25 mM MgCl₂, 0.1 M Tris [pH 7.2], 5% blocking reagent [Roche, Basel, Switzerland]) containing 1.0 μg/mL Cy3-labeled (F3002), CENPB-specific (ATTCGTTGGAAACGGGA) peptide nucleic acid (PNA) FISH probe (Panagene Inc., South Korea). Cells were moved to a dark space at RT for an additional 2 h. Cells were then washed with 70% formamide in 2× saline-sodium citrate (SSC) (pH 7.0), followed by 2 washes (for 10 min each) in 2× SSC (pH 7.0) and 1 wash in PBS (for 10 min). Finally, cells were stained with Hoechst (Thermo Fisher Scientific), mounted with Vectashield DAPI-containing mounting media. For SADS visualization, images were obtained using inverted laser scanning confocal microscope (LSM 780; Carl Zeiss Microscopy, Jena, Germany) with in-depth Z stacking. The number of decondensed peri-centromeric satellites per cell was assessed by the quantification of decondensed/elongated centromeres.³⁵ Nuclei of at least 50 cells were analyzed for each sample.

TAFs assay

TAF assays were performed on PFA fixed cells by telomere fluorescence FISH using Cy3-conjugated PNA probe specific to the telomere repeat TTAGGG (Panagene Inc.) and immunohistochemistry using antibody to γH2AX (Anti-phospho-Histone H2A.X [Ser139] mouse monoclonal, 05-636, Sigma-Aldrich) and Alexa Fluor 488 as a secondary antibody. Images were obtained using inverted laser scanning confocal microscope (LSM 780; Carl Zeiss Microscopy, Jena, Germany) with in-depth Z stacking. At least, 50 nuclei were examined for each sample, and the numbers of TAFs (co-localized foci of telomeric DNA and γH2AX) were counted.³⁶

Statistical analysis

Data are shown as mean ± SD and values were considered statistically significant at P < .05 using Student’s t-test, one-way ANOVA, or two-way ANOVA with appropriate post-hoc tests for multiple comparisons. Analyses were performed in GraphPad Prism (version 8; GraphPad Software, Inc., La Jolla, CA, USA).

Results

Hdac3 deficiency decreases the expression of osteogenic genes in BMSCs

To determine how Hdac3 deletion in osteo-progenitor cells impacts osteogenic related gene expression, primary BMSCs from Hdac3 CKO_Osx and control (Hdac3^fl/fl) mice were cultured in osteogenic medium in vitro. The mRNA level of Hdac3 was 40% lower in Hdac3 CKO_Osx BMSC cultures compared with control cultures (Figure 1A). Levels of mRNA encoding Runx2, a master regulator for osteogenesis, were also 40% lower in Hdac3 CKO_Osx cells compared with control cultures (Figure 1B). Transcripts for osteoblast differentiation markers (Bglap and Alpl) were reduced by 70%–75% (Figure 1C, D). Reductions in Hdac3 and Runx2 were confirmed at the protein level in freshly isolated CD45- and Ter-119-depleted BMSCs from Hdac3 CKO_Osx mice (Figure 1E–G). There were no differences in the expression of transcripts for cell cycle inhibitors, p16 and p21, under these conditions (Figure 1 H, I). We also compared the in vivo expression of Runx2 in the Hdac3 CKO_Osx and control from paraffin-embedded mice femur tissue sections. Consistent with in vitro data, Runx2+ cell numbers were significantly lower (4.7-fold) in femurs from Hdac3 CKO_Osx compared with control mice (Figure 1J, K). These data show that Hdac3 deletion reduces the ability of BMSCs to undergo osteogenic differentiation.

Figure 1.

Hdac3 deletion reduces osteogenic gene expression in BMSCs under osteogenic conditions. BMSCs from 6-wk-old male Hdac3 CKO_osx or control mice were cultured in osteogenic medium for 7 d. The abundance of mRNA transcripts for the indicated gene was determined by RT-qPCR. (A–D) mRNA levels of Hdac3 and osteogenic genes. (E) BMSCs from 6-wk-old male Hdac3 CKO_Osx mice or control were depleted of CD45 and Ter-119 positive cells using MACS and stained with Runx2 (AF647) or Hdac3 (AF488). (F, G) The number of Hdac3+ cells and Runx2+ cells was counted, respectively. (H, I) mRNA levels of cell cycle inhibitor genes. (J) Representative images of IF staining for Runx2 in cortical bone in the diaphysis of femurs from 8-wk-old Hdac3 CKO_Osx and control mice. The growth plate and cortical bone are outlined with dashed, white lines in recognition of bone structures. (K) Quantification of the percentage of Runx2+ cells in Hdac3 CKO_Osx mice and control mice. Data were collected from 100 cells in each sample. Data represent means ± SD of 3 replicates. Statistically significant differences were determined with the Student’s t test.

Deletion of Hdac3 promotes lipid storage in osteoblastic cultures of BMSCs

LD formation relies on the expression of molecules that facilitate fatty acid storage (eg, Plin1 and Fsp27/Cidec), whereas LD lipolysis is controlled by lipases and autophagic mechanisms.³⁸ Consistent with previous studies,¹⁷ substantial increases in lipid storage genes (Figure 2A, B) were observed in Hdac3 CKO_Osx cultures, whereas there were no changes in Lipe or Pparγ, an adipogenic transcription factor, compared with control BMSCs (Figure 2C, D). We next examined single cells in osteogenic cultures for Runx2 and LD expression. Consistent with our previous observations,¹⁷ Hdac3 CKO_Osx BMSCs cultured with osteogenic medium for 14 and 21 d contained Runx2+ cells, but they constituted a lower percentage of the population than in control cultures (Figure 2E, F). LD started to appear in both control and Hdac3 CKO_Osx BMSCs around day 10 (data not shown) and by day 14 were present in about 40% of cells in both cultures (Figure 2G). At day 21, the percentage of cells containing LD in control cultures did not change; however, the percentage of cells containing LDs increased in Hdac3 CKO_Osx cultures. When Runx2+ cells were examined in these cultures, we found that the number of Runx2+ LD+ cells was ~2- to 3-fold higher in cultures from Hdac3 CKO_Osx compared with control cultures (Figure 2H).

Figure 2.

Cytosolic LDs exist in Runx2+ cells. BMSCs from 6-wk-old male Hdac3 CKO_Osx or control mice were cultured in osteogenic medium for 7 d. The abundance of mRNA transcripts for the indicated genes was determined by RT-qPCR. (A, B) mRNA levels of lipid storage genes. (C, D) mRNA levels of adipogenic commitment co-factor and lipase. (E) Representative images of IF staining for Runx2 (AF647) and LDs (AF488) in BMSCs from 4-wk-old male Hdac3 CKO_Osx mice. (F, H) Percentage of Runx2+ cells, LD+ cells, and Runx2 + LD+ cells, respectively, in osteogenic BMSC cultures. Data were collected from at least 50 cells in each sample. Data represent means ± SD (n = 3 per group). Statistically significant differences were determined with Student’s t test (A–D) or two-way ANOVA with Tukey’s post-hoc test (F–I).

Runx2+ cells in Hdac3 CKO_Osx BMSC cultures have more LDs, SADS, and TAFs per nuclei than Runx2+ cells in control BMSCs

To investigate whether Runx2+ BMSCs exhibit early characteristics of cellular senescence, we measured SADS (ie, the unraveling of DNA a centromeres) in both Runx2 + LD+ and Runx2 + LD− subpopulations. We found significantly more SADS per nuclei using the CENBP-Cy3 FISH probe in Runx2+ BMSCs from Hdac3 CKO_Osx mice after 14 or 21 d in osteogenic medium compared with BMSCs from control mice (Figure 3A–C). The increases in SADS were not correlated with the presence of LD in Runx2+ Hdac3 CKO_Osx BMSCs as similar numbers were found in Runx2+ cells that did not contain LDs as those that contained LDs (Figure 3D). We also measured SADS in Runx2+ cells within tissue sections of mouse femurs. Consistent with the data from the in vitro BMSC cultures, Runx2+ cells in femurs from Hdac3 CKO_Osx mice had more SADS per cell compared with cells in control femurs (Figure 3E, F).

Figure 3.

Runx2+ BMSCs from Hdac3 CKO_Osx display marked distension of pericentromeric satellite DNA. (A) More senescence-associated distension of satellites (SADS, indicated by arrows) were observed in 14- and 21-d osteogenic cultures of BMSCs from male Hdac3 CKO_Osx than BMSCs from control mice. (B–D) Quantification of the percentage of SADS number per nuclei in Runx2+ cells, Runx2 + LD+ cells, and Runx2 + LD- cells, respectively. (E) More SADS numbers per nuclei were observed in Runx2+ cells within cortical bone in the diaphysis of femurs from male Hdac3 CKO_Osx than control mice. (F) Quantification of the SADS number per nuclei in Runx2+ cells from cortical bone diaphyseal sections. Data were collected from at least 50 cells in each sample. Data represent means ± SD (n = 3 per group). Statistically significant differences were determined by two-way ANOVA with Tukey’s post-hoc test (B–D) or Student’s t test (F).

We next measured TAFs, another early marker of cellular senescence associated with DNA damage where phosphorylated H2AX co-localizes to telomeres that are identified with TelC-Cy3 probe. Runx2+ cells from Hdac3 CKO_Osx BMSCs had more TAFs per cell compared with control cells (Figure 4A, B). Taken together, these data show that Runx2+ cells in Hdac3 CKO_Osx BMSC cultures exhibit early hallmark features of senescence under these in vitro osteogenic conditions and these changes occur irrespective of LD formation.

Figure 4.

Hdac3 CKO_Osx BMSCs cultures display more TAFs per nuclei in Runx2+ cells. (A) Runx2+ cells within 14- and 21-d osteogenic cultures of BMSCs were examined for TAFs (co-localization of γH2AX; a DNA damage protein) and telomeres (indicated by arrows). (B) Quantification of the number of TAFs per nuclei in Runx2+ BMSCs from male Hdac3 CKO_Osx versus control mice. Data were collected from at least 50 cells isolated from 3 mice per group. Statistically significant differences were determined by two-way ANOVA with Tukey’s post-hoc test.

HDAC inhibitors induce SADS in mouse BMSCs

To further investigate the role of HDAC inhibition in the development of SADS, the pan-HDAC inhibitor SAHA (ie, Vorinostat) or Hdac3-selective inhibitor RGFP966 was added BMSC cultures. SAHA suppresses osteogenic colony formation, decreases osteoblastic gene expression, induces cell cycle arrest, and causes DNA damage in mouse BMSCs during the initiation of osteogenic conditions.³⁹ WT BMSCs were seeded in osteogenic medium, incubated for 3 d to allow for cell adherence, and then switched to osteogenic medium supplemented with SAHA or RGFP966 (1 μM) or vehicle. SADS assays were performed after 10 d of osteogenic differentiation. Neither SAHA nor RGFP966 affected Runx2+ cell numbers or LD+ cell numbers in the cultures (Figure 5A–C). Consistent with observations from Hdac3 CKO_Osx BMSCs, the number of SADS per nucleus significantly increased in both Runx2+ and Runx2+/LD+ cells after SAHA or RGFP966 was added to osteogenic cultures. However, the number of SADS per nucleus increased irrespective of the presence of LD.

Figure 5.

HDAC inhibitors SAHA and RGFP966 induce SADS in BMSCs under osteogenic conditions. (A) WT BMSCs were incubated with vehicle, SAHA (1 μM), or RGFP966 for 48 h, followed by 10 d in osteogenic medium. SADS expression in a Runx2+ cell is shown. (B–C) Percentage of Runx2+ cells and LD+ cells, respectively, in osteogenic BMSC cultures from 3 mice per group. (D–F) Quantification of the number of SADS per nucleus in Runx2+ cells, Runx2 + LD+ cells, and Runx2 + LD− cells, respectively. Data were collected from at least 50 cells isolated from 3 mice per group. Statistically significant differences were determined by one-way ANOVA with Tukey’s post-hoc test.

Senolytics (Dasatinib + quercetin; D + Q) reduce SADS and TAFs in Runx2+ Hdac3 CKO_Osx BMSCs

To determine if the early signs of senescence observed in Runx2+ cells within Hdac3 CKO_Osx BMSC cultures can be reversed, a senolytic cocktail comprised of dasatinib and quercetin (D + Q) was added to the BMSC cultures. D + Q clear senescent cell by suppressing senescence-associated anti-apoptotic pathways.⁴⁰ D + Q reduce senescent cells and improve aged bone tissue functions⁴¹ and osteogenic potential of aged BMSCs.⁴² BMSCs cultures were established from Hdac3 CKO_Osx and control mice, respectively, and were incubated with vehicle (Veh) or D + Q (0.2 μM + 20 μM) for 24 h.⁴⁰ D + Q reduced the number of cells Hdac3 CKO_Osx BMSCs cultures by ~40% as compared with vehicle. Control BMSC cultures showed only a 10% reduction in cell number following incubation with D + Q (Figure 6C, D). We next measured single cells in osteogenic conditions after D + Q incubation for Runx2 and LD expression. D + Q did not impact Runx2+ cell numbers or LD+ cell numbers in the cultures (Figure 6E–G). These results indicate that D + Q selectively eliminated certain cell populations within BMSCs cultures; however, was not specific to Runx2+, LD+, or Runx2+/LD+ cells.

Figure 6.

Dasatinib and quercetin (D + Q) reduce SADS/TAFs in Runx2+ cells from Hdac3 CKO_Osx BMSC cultures. (A, B) BMSCs were incubated with vehicle (Veh) or D + Q for 24 h, followed by 21 d in osteogenic medium. Runx2+/LD+ cells were examined for SADS and TAFs (indicated by arrows, respectively). (C, D) Cell confluency and viability of BMSCs incubated with either Veh or D + Q for 24 h. n = 3 biological replicates. (E–G) Percentage of Runx2+ cells, LD+ cells, and Runx2+/LD+ cells in osteogenic BMSC cultures following incubation with D + Q or Veh. n = 3 biological replicates. (H–J) Quantification of the number of SADS per nucleus in Runx2+ cells, Runx2+/LD+ cells, and Runx2+/LD− cells, respectively. (K) Quantification of the number of TAFs per nucleus in Runx2+ cells. For H–K, data were collected from at least 50 cells isolated from 3 mice per group. Statistically significant differences were determined by two-way ANOVA with Tukey’s post-hoc test.

To assess the impact of D + Q on the number of SADS per nucleus, we quantified SADS in both Runx2 + LD+ and Runx2 + LD− subpopulations. We observed a significant reduction of SADS per nuclei in Runx2+ BMSCs from Hdac3 CKO_Osx mice following the addition of D + Q to osteogenic medium, compared with BMSCs incubated with vehicle (Figure 6A, H). Although the observed effect does not reach the levels seen in the control cultures, it nonetheless suggests a discernible degree of amelioration (Figure 6H–J). In line with the reduction of SADS per nuclei, we also observed a significant decrease in the number of TAFs per nucleus in Runx2+ Hdac3 CKO_Osx BMSCs after D + Q exposure (Figure 6B, K). However, this effect does not fully revert to the control condition.

Discussion

Hdac3 is required for skeletal development, bone formation, and skeletal heath during aging. We previously reported that osteogenic cultures BMSCs from 14- and 26-mo-old mice had reduced Hdac3 levels, greater expression of lipid storage genes, and increased LD formation than BMSCs from young mice.¹⁷ Lower Hdac3 mRNA levels were also observed in bone marrow of aging C57BL/6 mice from the Atlas of Gene Expression in Mouse Aging Project²³ as well as in human bones from postmenopausal women.^24,43 We also reported that the conditional deletion of Hdac3 in osteo-progenitor cells increases mRNA expression of the cell cycle inhibitor, p21, in calvaria but not long bones, and increases phosphorylated H2AX levels, a sign of double-stranded DNA damage repair, in osteoblasts and osteocytes.^24,43 These cellular events align with the premature skeletal aging phenotypes of mice in which Hdac3 is depleted in osteoprogenitors, namely osteopenia and the accumulation of BMAT.^17,24,43 In this study, we aimed to determine if Hdac3 deletion in bone marrow-derived osteo-progenitor cells triggers chromatin changes associated with senescence and if these changes correlate with LD formation Runx2+ osteo-progenitor cells.

LD numbers and the relative expression of lipid storage genes, Cidec and Plin1, were 2- to 3-fold higher in Runx2+ cells from Hdac3-depleted BMSCs than in Runx2+ BMSCs from control mice. LDs were observed in Runx2+ BMSCs from control mice, but it was less than 10% of total cell population. These data are consistent with our previous study showing that Runx2+ osteoblast lineage cells have enhanced capacity to store lipids and that Hdac3 prevents LD formation in vitro.¹⁷ Hdac3 deletion in other cell types also promotes lipid storage. For example, Hdac3 null mice upregulated Plin2, leading to hepatic steatosis which could be improved by depleting Plin2.⁴⁴ A recent genome-wide CRISPR-Cas9 KO screening study in Plin2-GFP reporter cells revealed Hdac3 governs Plin2 levels in various cell types and different metabolic conditions.⁴⁵ Plin2 is ubiquitously expressed and participates in small LD generation and LD stabilization.⁴⁶ Depletion of Hdac3 in Plin2-GFP reporter cells increased GFP signals along with increased LD formation and Plin2 transcript levels.⁴⁵ While Plin2 was not examined in these experiments, the upregulation of other lipid storage genes (Cidec, Plin1) in Hdac3 deficient cells indicates that Hdac3 deletion primes cells, including bone marrow-derived osteoprogenitors, for LD formation.

Hdac3 removes acetyl groups from hyper-acetylated histones and thereby promotes chromatin condensation. Hdac3 is unique from other HDACs in that it associates with the nuclear receptor co-repressors NCOR1/2 and transcription factors such as Runx2 to suppress gene expression.²² In the absence of Hdac3, chromatin may be more relaxed, leaving DNA susceptible to nicks and breaks. In this study, we sought to determine how chromatin changes associated with early stages of cellular senescence correlated with LD formation. Using fluorescent probes and antibodies, we measured satellite distensions at centromeres and phosphorylated-H2AX localization at telomeres in Runx2+ cells within BMSCs cultures. Runx2+ BMSCs from Hdac3 CKO_Osx mice had substantially more SADS and TAFs than Runx2+ BMSCs from control mice. SADS were present in both Runx2+ LD+ and Runx2+ LD- BMSCs. We previously documented increased phosphorylated H2AX expression in osteoblasts and osteocytes within Hdac3 CKO_Osx mice.⁴³ Phosphorylated H2AX is a marker of double-stranded DNA damage and triggers repair processes. If DNA breaks are not repaired, cells may enter apoptosis. The number of Runx2+ cells was lower in Hdac3 CKO_Osx BMSC cultures than in control cultures, indicating that Hdac3-deficient osteoblasts either failed to thrive under these conditions, entered a senescence program, and/or died by apoptosis. Both Hdac3 CKO_Osx and control cultures contained equivalent levels of p21 and p16 (cell cycle inhibitors associated with senescent cells) as measured on a population level by RT-PCR. Additional single cell assays at earlier timepoints are needed to measure the expression of cell cycle inhibitors and SASP genes in Hdac3 CKO_Osx BMSCs.

Furthermore, we observed that SAHA and RGFP966 (pan-HDAC and HDAC3-selective inhibitors, respectively) induce SADS and TAFs formation in Runx2+ BMSCs regardless of whether they contained LDs. Neither SAHA nor RGF966 significantly increased the number of cells with LDs. SAHA (also known as vorinostat) has shown efficacy as a treatment for refractory cutaneous T cell lymphomas and other cancers, but it is not a front-line chemotherapy because of its side effects.⁴⁷ Its mechanisms of action include upregulation of tumor suppressor genes and relaxing chromatin structure to an extent that increases DNA damage within tumor cells. More recently, it was shown that low doses of HDAC inhibitors, including SAHA, prevented cellular senescence by removing blocks on the expression of nuclear-encoded mitochondrial oxidative phosphorylation genes, which improved mitochondrial function and suppressed the appearance of cytoplasmic chromatin fragments that induce transcription of the SASP through the cGAS-STING pathway.³⁴ We previously showed that daily injections of SAHA over the course of 4 wk reduced BMD in mice but did not increase BMAT.³⁹ In that study, osteoprogenitor cells declined in number after SAHA exposure, while mature and nondividing osteoblasts were more active, demonstrating an effect of cell differentiation and maturation state on susceptibility to HDAC inhibition. Bone remodeling markers returned to normal after cessation of SAHA exposure in these animals. Altogether, these data indicate that HDAC inhibition in osteo-progenitor cells increases the chromatin unraveling as centromeres and telomeres and that these events are not necessary for LD formation in osteoblasts.

Finally, we observed that a senolytic cocktail of dasatinib and quercetin (D + Q) reduced Runx2+ SADS and TAFs in Hdac3 CKO_Osx BMSC cultures, independent of LD formation. Dasatinib is anticancer drug and tyrosine kinase inhibitor that induces apoptosis. Quercetin is a phytochemical, and has antioxidant, anti-inflammatory, anticancer properties. When combined, D and Q target different cell survival mechanisms and effectively clear senescent cells to mitigate aging-related phenotypes.⁴⁰ In this study, we aimed not only to investigate the potential reversal of the senescent phenotypes in Hdac3 CKO_Osx BMSCs by administering D + Q but also to further test that LD formation in Runx2+ cells contributes to the induction of senescence. D + Q significantly reduced the number of SADS and TAFs; however, they did not affect the number of Runx2+ cells, Runx2 + LD+ cells, or LD+ cells alone. Altogether, these data suggest that LD formation is not necessary for inducing cellular senescence in osteo-progenitor cells. A limitation of this study is that D + Q experiments were performed only in vitro. D + Q could be delivered to newborn mice via gavage; however, we determined that administering D + Q after birth at the levels and frequency needed to reverse phenotypes would be too stressful for the Hdac3 CKO_Osx newborn mice and lead to premature death because they are very small and fragile. D + Q administration in utero could be problematic because D is a teratogen,⁴⁸ which would complicate data interpretations of the skeleton.

In conclusion, our data show that the depletion of Hdac3 in bone marrow-derived osteoprogenitor cells contributes to an increase in early senescence markers and LD formation in Runx2+ cells under osteogenic conditions. LDs form in Hdac3-deficient Runx2+ cells regardless of these chromatin changes. Altogether, our data demonstrate that Hdac3 inhibition induces DNA damage and LD formation in bone marrow osteoprogenitor cells via independent mechanisms.

Author contributions

Dongwook Yeo (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Visualization, Writing—original draft), Elizabeth Zars Fisher (Resources), Sundeep Khosla (Methodology, Resources, Validation, Writing—review & editing), Joshua Farr (Methodology, Resources, Validation, Writing—review & editing), and Jennifer Westendorf (Conceptualization, Funding acquisition, Investigation, Project administration, Supervision, Validation, Writing—original draft, Writing—review & editing)

Funding

This work was supported by grants from the Kogod Center on Aging at Mayo Clinic.

Conflicts of interest

The authors declare that they have no disclosures.

Data availability

The data underlying this article will be shared on reasonable request to the first author and the corresponding author.