Hormone sensitive lipase ablation promotes bone regeneration

Abstract

There is an inverse relationship between the differentiation of mesenchymal stem cells (MSCs) along either an adipocyte or osteoblast lineage, with lineage differentiation known to be mediated by transcription factors PPARγ and Runx2, respectively. Endogenous ligands for PPARγ are generated during the hydrolysis of triacylglycerols to fatty acids through the actions of lipases such as hormone sensitive lipase (HSL). To examine whether reduced production of endogenous PPARγ ligands would influence bone regeneration, we examined the effects of HSL knockout on fracture repair in mice using a tibial mono-cortical defect as a model. We found an improved rate of fracture repair in HSL-ko mice documented by serial μCT and bone histomorphometry compared to wild-type (WT) mice. Similarly, accelerated rates of bone regeneration were observed with a calvarial model where implantation of bone grafts from HSL-ko mice accelerated bone regeneration at the injury site. Further analysis revealed improved MSC differentiation down osteoblast and chondrocyte lineage with inhibition of HSL. MSC recruitment to the injury site was greater in HSL-ko mice than WT. Finally, we used single cell RNAseq to understand the osteoimmunological differences between WT and HSL-ko mice and found changes in the pre-osteoclast population. Our study shows HSL-ko mice as an interesting model to study improvements to bone injury repair. Furthermore, our study highlights the potential importance of pre-osteoclasts and osteoclasts in bone repair.

1. Introduction

There is a strong inverse relationship between bone mass and bone marrow adipose tissue with aging and osteoporosis. In addition to hematopoietic stem cells that give rise to blood cell lineages, the bone marrow contains mesenchymal stem (or stromal) cells (MSCs) that give rise to cartilage, adipocytes, osteoblasts, and fibroblasts. A number of studies have shown an inverse relationship between the differentiation of MSCs along either an adipocyte or osteoblast lineage such that agents that stimulate adipocyte differentiation simultaneously inhibit osteoblast differentiation and, conversely, stimulation of osteoblast differentiation simultaneously inhibits adipocyte differentiation [1,2]. The key transcription factors that direct differentiation of MSCs are Runx2 (runt-related transcription factor 2) for osteoblastogenesis and PPARγ (peroxisome proliferator activated receptor γ) for adipogenesis. We and others have shown that secretory products generated directly from adipocytes suppress osteoblast differentiation and that fatty acids and their metabolites, which act as ligands for PPARγ, are the principal effectors of this action. The release of fatty acids from adipocytes is a highly regulated process, which is ultimately dependent on specific lipases that are responsible for mediating the breakdown of triacylglycerol stores within lipid droplets within cells. One of the major lipases responsible for triacylglycerol hydrolysis and generation of PPARγ ligands is hormone-sensitive lipase (HSL).

Fatty acids (FAs) and their metabolites are known to be endogenous ligands for activation of the PPAR family of nuclear hormone receptors [3]. The role of lipase-mediated FA release in this signaling was uncovered when analyzing HSL-knockout (ko) mice [4], where changes in adipose tissue gene expression suggested that abnormalities in the generation of FAs or FA metabolites might prevent complete adipose differentiation due to deficiency of ligands for PPARγ [5]. Adipocytes from normal mice were shown to provide ligands for activation of PPARγ and that activation is further boosted following lipolytic stimulation, whereas adipocytes from HSL-ko mice displayed attenuated activation of PPARγ, with no change following lipolytic stimulation [6]. These observations were confirmed and extended by other investigators showing that both HSL and ATGL generated FA metabolites that were needed for signaling through both PPARα and PPARγ [7]. Indeed, lipidomics studies revealed that >50 FA metabolites are generated by HSL-mediated lipolysis [8]. Extending these observations to bone, we observed that HSL-ko mice maintain increased bone density with age and that MSCs from HSL-ko mice display increased osteoblastogenesis [9].

PPARγ, besides involvement with MSC differentiation, also functions within the hematopoietic lineage [10]. In particular, osteoclastogenesis and osteoclast function have been shown to be regulated by PPARγ based on ligand availability [11–13]. Furthermore, osteoclasts and chondroclasts (cartilage degrading osteoclasts) are important in several aspects of fracture repair involving initial bone fragment resorption, cartilage matrix degradation, and bone remodeling [14,15]. Taken together, all of these studies suggest HSL-ko may also affect bone repair possibly by alteration to the osteoclast and chondroclast populations. The current studies were undertaken to directly address these possibilities.

2. Material and methods

2.1. Chemicals and reagents

Non-esterified fatty acid HR Series NEFA-HR (2) from Wako Diagnostics, Wako Life Sciences, Inc., Mountain View, CA, USA; F10 media, dispase and collagenase from Thermo Fisher scientific. TRIzol reagent and SuperScript II from Invitrogen (Carlsbad, CA, USA); RNeasy kit from QIAGEN (Valencia, CA, USA); SyBr green Taqman PCR kit from Applied Biosystems (Foster City, CA, USA). PDGFRα-APC, Sca1-PB, CD31-PE, CD45-FITC from BioLegend (San Diego, CA, USA); anti-osteocalcin antibody (cat# ab93876), anti-Runx2 antibody (cat# ab192256), and anti-Plin1 antibody (cat# ab3526) from Abcam (Waltham, MA, USA); anti-osteopontin antibody (cat# 967802) and anti-collagen II antibody (cat# 967803) from R&D Systems (Minneapolis, MN, USA); donkey anti-goat IgG cross-adsorbed secondary antibody, Alexa Fluor 488 (cat# A11055), donkey anti-sheep IgG cross-adsorbed secondary antibody, Alexa Fluor 488 (cat# 11015), donkey anti-rabbit IgG highly cross adsorbed secondary antibody, Alexa Fluor 555 (cat# A31572) from Thermo Fisher Scientific (Waltham, MA, USA); leukocyte acid phosphatase staining kit (cat#386A-1KT) from Sigma-Aldrich (St. Louis, MO, USA). BAY 59–9435 (4-isopropyl-3-methyl-2-([(3 s)-3-methylpiperdin-1-yl]carbonyl)isoxazol-5(2H)-one) was synthesized by Atomax Chemicals Co., Ltd. (Shenzhen Guangdong, China).

2.2. Animals

Animal studies were performed in accordance with National Institutes of Health guidelines and all procedures were approved by the institutional animal care and use committee of VA Palo Alto Health Care System. HSL^−/− mice were generated by homologous recombination, as described previously [4], and backcrossed 5 times with C57/BL6J mice. Mice were maintained in the animal facility at the VA Palo Alto Health Care System on a 12-h light-dark cycle. For breeding experiments, mice heterozygous for the deleted HSL allele were used to generate homozygous HSL^−/− mice and HSL^+/+ (WT) littermates. Genotyping was performed by a single-step PCR using three primers, as described previously [9]. Both male and female mice were used in these studies. Some mice were sacrificed at different time points for immunohistochemical evaluation of bone healing and neovascularization, as well as bone histomorphometry, at the sites of injury. Body composition was determined by dual-energy X-ray absorptiometry (DXA) using a Discovery model DXA scanner adapted for rodent imaging (Hologic, Bedford, MA). Calibration was performed using human spine phantom and small step for rodents before each set of measurements. The animals were anesthetized with ketamine/xylazine before scanning, and data were obtained according to manufacturer’s protocols.

2.3. Bone surgery

WT and HSL −/− animals were age and gender matched for each experiment. Animals were anesthetized using katamine/xylazine, and prepped for surgery with hair removal and sterilization of area for surgery. Before surgery, animals were given a dose buprenorphine. For the mono-cortical defect model [16], a 1 mm pinhole was created using a dental drill in the anterior tibial plateau, penetrating through only a single tibial cortex. After creation of bony defects with the drill, small bone pieces were washed off with saline and the surgery site stitched together. For the calvarial bone injury model [17], a 2 mm circular defect was created in the center of the parietal bone using a diamond-coated trephine bit without disturbing the underlying dura mater. After creation of bony defects with the drill, small bone pieces were washed off with saline and bone marrow grafts were placed in the center and the surgery site stitched together. Animals were injected with buprenorphine 30 min before the surgery and three days post surgery, and monitored daily for their wellbeing. Three-dimensional imaging was performed the day after the surgery to document the baseline defect and weekly thereafter to evaluate bone healing. For the tibial injury model: WT, n = 6, 3 males (6 months) and 3 females (8 months); HSL-ko, n = 7, 4 males (6 months) and 3 females (8 months). For the calvarial bone injury model: recipients were all WT mice 8 months old, gender matched with the donor. Donors of the bone marrow grafts were WT, n = 5, 3 males (4 months) and 2 females (4 months); HSL-ko, n = 5. 3 males (4 months) and 2 females (4 months).

2.4. Three-dimensional μCT imaging

The mice were sedated using isoflurane and subjected to in vivo three-dimensional μCT imaging the day after the surgery to document the baseline defect and weekly thereafter to evaluate bone healing. Scan was performed using a Scanco vivaCT 40 μCT scanner (Scanco Medical AG, Basserdorf, Switzerland). Scanning was started from 0.5 mm above the injury site, and a total of 200 consecutive 10-μm-thick sections were scanned, and 150 sections were analyzed, encompassing a 1.5-mm length of the injury site. Cortical bone was excluded from the region of interest with semiautomatically drawn contours. The segmentation values were kept constant at 1.0/1/276. Relative bone volume/total volume (BV/TV), trabecular number (TbN), trabecular thickness (TbTh), and trabecular separation (TbSp) were calculated by measuring 3-dimensional distances directly in the trabecular network and taking the mean over all voxels.

2.5. Bone staining and histomorphometry

Tibias and calvaria were harvested at the specified time points and fixed in 4% paraformaldehyde (PFA) at 4 ^◦C for 16 h. Samples were decalcified in 19% EDTA, dehydrated in a graded ethanol series, embedded in paraffin, and sectioned at 8 μm thickness. Aniline blue and pentachrome staining were performed. For aniline blue staining, sections were treated with a saturated solution of picric acid followed by a 5% phosphotungstic acid solution and staining in 1% aniline blue. Pentachrome-staining was performed as described [18]; pentachrome-stained tissues reveal nuclei as blue to black colour, cytoplasm stains red, collagen stains yellow to greenish yellow, and fibrous tissue stains an intense red. Immunohistochemistry was performed as described [19]; tissue sections were permeabilized with 0.5% TritonX-100 followed by overnight incubation with primary antibodies (osteocalcin, Runx2) followed by goat anti-rabbit secondary antibody. Tartrate-resistant acid phosphatase activity (TRAP) was observed using a leukocyte acid phosphatase staining kit according to the manufacturer’s instructions. Image J was used to quantify the positive signals for various staining. The total and positive pixels to each specific staining in the region of interest (ROI) were measured. The % of positive pixels to total pixels in the region of interest were calculated and presented.

2.6. Isolation, culture, and differentiation of primary bone marrow cells

Femurs and tibiae from 4- to 5-mo-old WT and HSL^−/− mice were dissected and rinsed with 75% ethanol and diethylpyrocarbonate water to eliminate any surrounding soft tissue. Bone marrows were flushed out using a 25-gauge needle and digested with F10 media containing 1unit/ml of dispase and 100 units/ml of collagenase for 20 min before termination with wash media (F10+ 10% horse serum). Cell mixtures were passed through 5 ml pipets twice and filtered through a 100 μm cell strainer. Cells were pelleted at 1000 x g for 10 min before lysing the red blood cells with sterile distilled water for 10 s and immediately adding 1/10 the volume of 10XPBS buffer; cells were washed 2 times with wash media. The bone marrow stromal cells (BMSCs) were used for immunostaining for FACS or were plated in α-minimum essential medium (α-MEM) with 10% FBS and grown until confluent before being trypsinized and plated in 96 wells for studies of differentiation. Adipogenesis was induced with insulin, dexamethasone, and isobutylmethylxanthine, as described previously [20]. Chondrogenesis and osteoblastogenesis were induced with growth medium containing chondrogenic and osteogenic cocktails from R&D system, respectively. In addition to differentiation cocktail, some WT BMSCs were incubated in the absence or presence of 10 nM BAY 59–9435, an HSL inhibitor [21]. Media were changed every 3 days, and cells fixed on day 8 post induction of differentiation and used for immunohistochemical staining using specific antibodies and control serum.

2.7. Isolation of bone marrow cells for Dropseq and 10× genomic ScRNAseq analysis

At day 5 post injury, the injury sites were dissected to make sure the newly formed callus was still attached to the injury site. The bones were cut 2 mm on both sides of the injury site, and bone marrow, together with the callus, was flushed out and digested in digestion media. For the samples before injury, the length and the same location of the bone were dissected, and bone marrow were flushed out and digested in digestion media. Cell mixtures were passed through 5 ml pipets twice and filtered through a 100 μm cell strainer. Cells were pelleted at 1000 x g for 10 min before lysing the red blood cells with sterile distilled water for 10 s and immediately adding 1/10 the volume of 10×PBS buffer; cells were washed 2 times with 1× PBS containing 5% horse serum. Cells were counted and transported on ice for Drop-seq and/or 10× genomic scRNA-seq analysis.

2.8. Drop-seq of bone marrow cells

Drop-seq was carried out as previously described with a few alterations in the microfluidics step [22]. Cells were spun down at 1000 rpm for 3 mins. Two washes were done using PBS-BSA (0.01% BSA, ThermoFisher Scientific Cat#: 15260037). Finally, cells were pipetted through a 35 μm cell strainer into FACs tubes (Fisher Scientific, Cat#: 352235). Cells were counted using the Countess automated cell counter (ThermoFisher). Cells were loaded in PBS-BSA at a concentration of 311 Cells/μl. Drop-seq beads (Chemgenes Cat#: Macosko-2011–10) were loaded into another syringe at 311 beads/μl in lysis buffer. Cells, beads, and oil were run through a Dolomite scRNAseq chip (Dolomite Cat#: 3200583) with cells and beads at 40 μl/min and the oil at 200 μl/min as recommended in the product literature. Following the microfluidics step, library generation was conducted consistent with the standard Drop-seq protocol.

2.9. 10× chromium controller 3^′ counting

Cells were spun down at 1000 rpm for 3 mins. Two washes were done using DMEM (Life Technologies, Cat#: 10569–044) + 10% FBS (Millipore Sigma Cat#: F0926). Cells were then pipetted through a 35 μm cell strainer into FACs tubes and counted using the Countess automated cell counter. Cells were aliquoted for a final concentration of 1000 cells/μl. Cells were subjected to 10× Chromium Controller 3^′ counting V3 at the Stanford Functional Genomics facility consistent with standard protocols with an aim for 10,000 cells per run.

2.10. Sequencing

Each Drop-seq sample was subject to sequencing on the HiSeq 4000 (at the Stanford functional genomics facility) using a custom read 1 primer specified in the Drop-seq protocol [22]. Reads were trimmed to 25 bp on read 1 and 50 bp on read 2. 10× Chromium libraries were multiplexed onto a NovaSeq 6000 system for sequencing at Fulgent. Reads were trimmed to 28 bp on read 1 and 98 bp on read 2.

2.11. Drop-seq and data analysis

Drop-seq libraries were converted to a matrix using a combination of the Drop-seq core computational protocol (referred to as the Drop-seq Alignment Cookbook) and hisat2 [23–25]. After generation of an expression matrix, datasets were analyzed primarily using Seurat [26]. A limit of 10% mitochondria expressed genes and 400 minimum genes per cell were used as a quality control filter on cells.

2.12. 10× chromium scRNA-seq and data analysis

10× chromium controller generated datasets were converted to expression matrices using Cellranger v3.1 (https://support.10xgenomics.com/singlecellgeneexpression/software/pipelines/latest/installation). Cellranger aggr was used to combine V2 (Replicate 1) and V3 (Replicate 2) datasets. Cellranger uses STAR for alignment of reads to the genome. Data was primarily analyzed using Seurat [26]. A limit of 10% mitochondria expressed genes was used as a quality control filter on cells.

2.13. qRT-PCR

qRT-PCR was conducted using the Kapa sybr fast one-step qRT-PCR kit (Kapa biosystems Cat#: KK2602). 100 ng of RNA was used as input for each gene. Each gene was tested with 3 technical replicates per sample. Reactions were run on a CFX384 Touch Real-Time PCR Detection System (Bio-rad Cat#:1855485). Primers were ordered from IDT and are listed in Table S1. Gene expression was normalized to Gapdh. Significance of difference between WT and HSL-ko was determined by an unpaired t-test in Prism.

2.14. Fluorescence-activated cell sorting (FACS) of BMSC

Two million isolated cells were stained with fluorochrome-tagged antibodies against cell surface markers (Sca1, CD31, CD45, PDGFRα) for 20 min at 4 ^◦C, followed by washes with wash media before resuspension in fresh wash media for FACS analysis. Fluorescence minus one (FMO) control was performed for cell preparations from each animal. Data were collected on an LSRII cytometer using FACSDiva (BD Biosciences, San Jose, CA) software, followed by FlowJo (Tree Star, Ashland, OR) analysis. Plots are represented using the biexponential transformation function to allow visualization of events close to or below the axes.

3. Results

3.1. Accelerated bone repair after injury in hormone-sensitive lipase (HSL) ablated mice

To examine the influence of fatty acids (FAs) on bone regeneration, we used HSL ablated mice, where PPARγ activation has been shown to be attenuated due to deficiency of PPARγ ligands. A 1 mm mono-cortical defect was introduced on one side of the tibia of WT and HSL-ko mice. μCT images of a WT and an HSL-ko mouse are shown in Fig. 1A. One day following injury (D1), the 1 mm defect was visible and comparable in both animals, documenting that a similar injury was created. There was healing of the injury in the HSL-ko mouse by D14 that appeared complete by D21, while the bone defect in the WT mouse, though showing some healing, was incomplete even at D21. By plotting the ratio of bone volume (BV) to total volume (TV) as a ratio of the initial BV/TV determined at D1 as a measure of the rate of bone regeneration (Fig. 1B), the bone regeneration rate was higher in HSL-ko than WT mice at D14 and returned to normal at D21. As another measure of the healing, the bone thickness at the site of injury was measured and was greater in the HSL-ko than WT mice at D21 (Fig. 1C). Aniline blue and pentachrome staining revealed the presence of more dense bone formation in the new bony bridge area in HSL-ko mice at D21 (Fig. 1D). In addition, immunohistochemical staining showed increased expression of both osteocalcin and Runx2 in HSL-ko mice (Fig. 1E), consistent with greater bone growth. Analysis of TRAP staining failed to show any significant differences between WT and HSL-ko mice, due to large individual variations (Fig. S1).

Fig. 1.

Accelerated bone repair after injury in HSL ablated mice. A. 1 mm mono-cortical defect was introduced on one side of the tibia of WT and HSL-ko mice. μCT images showing the bone regeneration during the period of twenty-one day of a WT and an HSL-ko mouse are shown in panel A. B. Bone regeneration rate calculated as the ratio of bone volume to total volume as a ratio of the initial BV/TV. C. Bone thickness at the site of the injury. D. Aniline blue and pentachrome staining of the bone section of the injury site. E. Immunohistochemical analysis of osteocalcin and Runx2 staining of the bone section of the injury site. Image J was used to quantify the staining. Data are representative of five independent experiments with n = 5–7 for each group. * P < 0.05.

To evaluate the effect of bone marrow from WT and HSL-ko mice on bone regeneration, in a separate animal model, a 2 mm circular defect was introduced using a dental drill in the center of the parietal bone of normal WT C57BL/6 mice; the bony defect was implanted with a bone graft harvested from the bone marrow of either normal WT or HSL-ko mice. The animals underwent weekly in vivo three-dimensional μCT imaging. As shown in Fig. 2, after 6 weeks of recovery, WT mice implanted with bone grafts obtained from WT bone marrow showed some bony bridging, but still had substantial bony defects. In contrast, WT mice implanted with bone grafts obtained from HSL-ko bone marrow showed almost complete bone replacement of the defect (Fig. 2A). Aniline blue (Fig. 2B) and pentachrome (Fig. 2C) staining revealed the increased presence of new osteoblasts in the bony bridge area of mice implanted with a bone graft from HSL-ko mice.

Fig. 2.

Increased bone regeneration with bone graft implant from HSL-ko mice. Bone graft harvested from the bone marrow of either normal WT or HSL-ko mice were implanted after a 2 mm circular defect was introduced to the center of the parietal bone of normal WT C57BL/6 mice. A (WT), B (HSL-ko) show the μCT imaging after six weeks. C (WT), D (HSL-ko). Analine blue staining of the bone section of the injury site. E (WT), F (HSL-ko). Pentachrome staining of the bone section of the injury site. Image J was used to quantify the staining. Data are representative of five independent experiments with n = 3 for each group. *, P < 0.05.

3.2. Increased differentiation potential of bone marrow stromal cells with treatment of HSL inhibitor

Bone marrow stromal cells were isolated, cultured in vitro and induced for differentiation using adipogenic, osteogenic, or chondrogenic cocktails in the presence and absence of BAY 59–9435, an HSL inhibitor [21] (Fig. S2 shows the dose response of the inhibitor on HSL activity.). After eight days of differentiation, cells were fixed and stained using anti-Plin1 antibody for adipocytes, anti-osteopontin for osteoblasts and anti-collagen II for chondrocytes. As shown in Fig. 3, the bone marrow stromal cells showed increased potential for differentiation to osteoblasts and chondrocytes, as well as adipocytes, in the presence of the HSL inhibitor.

Fig. 3.

Increased differentiation potential of bone marrow stromal cells with treatment of HSL inhibitor. Bone marrow stromal cells were isolated, cultured in vitro and induced for differentiation using adipogenic, osteogenic, or chondrogenic cocktails in the presence and absence of an HSL inhibitor (10 nM). After eight days of differentiation, cells were fixed and stained using anti-Plin1 antibody for adipocytes (A, B), anti-osteopontin for osteoblasts (C, D) and anti-collagen II for chondrocytes (E, F). Image J was used to quantify the staining. Data are representative of three independent experiments with n = 3 for each group. *, P < 0.05, **, P < 0.01.

3.3. Increased MSCs in bone marrow cell preparation from HSL-ko mice

To evaluate potential differences in the cell population within bone marrow of WT and HSL-ko mice, bone marrow mononuclear cells were prepared from mice four days after injury. The bone marrows were flushed out and the calluses that formed at four days post injury were collected and mononuclear cells were prepared and stained for markers Sca1, CD31, CD45 and PDGFRα and subjected to fluorescent activated cell sorting (FACS) analysis. As shown in Fig. 4, with WT and HSL-ko paired for age and gender, four days after injury, there was a higher percentage of MSCs, defined as Sca1⁺, CD31⁻, CD45⁻ and PDGFRα⁺, in the mononuclear cells from HSL-ko mice, which reached statistical significance (P < 0.05). This change in MSC numbers at the fracture site suggested potential osteoimmunological changes in HSL-ko mice. MSC recruitment and survivability at the injury site is heavily regulated by immune populations, including neutrophils, macrophages, and CD8+ T cells [27,28]. We next investigated changes in baseline immune cells that could provide further explanation for the differences in bone repair.

Fig. 4.

Increased MSCs in bone marrow cell preparation from HSL-ko mice. Bone marrow mononuclear cells were prepared from WT (A) and HSL-ko (B) mice four days after injury and stained for markers Sca1, CD31, CD45 and PDGFRα and subjected to fluorescent activated cell sorting (FACS) analysis. There were ~15–20,000 cells recovered from the bone marrow of the callus and surrounding area. C. Data presented are a summary of five independent experiments with matching WT and HSL-ko mice. *, P < 0.05.

3.4. Single cell RNA sequencing of the mouse medullary cavity from the tibia and femur reveals population levels and gene expression of 26 bone marrow resident cells

Considering the increased MSC recruitment at day 4 in HSL-ko mice compared to WT, we hypothesized that there may be differences in local immune populations prior to the injury. This could lead to an altered initial inflammatory phase for the bone repair, allowing MSCs to localize and survive at the injury site sooner and in higher numbers. We used single cell RNA sequencing (scRNAseq) to simultaneously profile 26 cell types of the medullary cavity prior to injury and five days post-injury (Figs. 5A, S3, S5–S8). There were no significant differences in the population levels of cell types prior to injury (Fig. 5B). Differential gene expression (DGE) analysis showed that pre-osteoclasts had the most differentially expressed genes (DEG) prior to injury (Fig. 5C). Unexpectedly given their numbers in our FACs analysis, we found very few Pdgfrα and Sca1 double positive cells in both our 10 × 3^′ counting and Drop-seq datasets (Fig. S8).

Fig. 5.

scRNAseq of mouse medullary cavity from tibia and femur reveal gene expression changes among pre-osteoclasts prior to injury. A) Two dimensional UMAP plot of mouse medullary cavity cells from uninjured and post-injury mice. B) Cell type composition of the uninjured mouse medullary cavity. C) Count of differentially expressed genes (adjusted pval < 0.05, minimum log fold change > 0.25) between WT and HSL-ko uninjured mouse cell types.

3.5. Focused analyses of pre-osteoclasts reveal different states of osteoclastogenesis

To further elucidate the differences between uninjured WT and HSL-ko pre-osteoclasts, we focused specifically on these cells (Fig. 6A-B). Pre-osteoclasts expressed the expected general pre-osteoclast/osteoclast markers of Acp5, C1qa, C1qb, and C1qc [29,30] (Fig. 6C). We observed expression of genes known to be expressed in certain stages of osteoclastogenesis presented in subsets of the pre-osteoclasts (Fig. 6C). As an example, we observed expression of Ciita and Ccr2 in subsets of the pre-osteoclasts [31,32] (Fig. 6C). Finally, we assessed the expression of Pparγ, finding it only expressed among a subset of the pre-osteoclasts (Fig. 6C). Collectively this data suggests we have a heterogeneous group of pre-osteoclasts, potentially representing different stages of osteoclastogenesis, of which Pparγ is only present in a subset (Fig. 6C).

Fig. 6.

Examination of pre-osteoclasts from uninjured mice reveals distinct expression patterns of subsets tied to WT or the HSL-ko. A) Isolation of pre-osteoclasts from the total dataset. B) Two dimensional UMAP plots comparing pre-osteoclasts from uninjured WT and HSL-ko mice. Cells were subsetted to ensure equal cell numbers in plots. C) Feature plots showing gene expression within pre-osteoclasts. D) Biological GO terms generated from differentially expressed genes in favor of the uninjured HSL-ko pre-osteoclasts. No significant GO terms (adjusted pval <0.05) were found for differentially expressed genes in favor of the uninjured WT pre-osteoclasts. E) ChEA 2016 enriched transcription factors generated from differentially expressed genes when comparing WT uninjured pre-osteoclasts and HSL-ko uninjured pre-osteoclasts. F) Feature plot showing expression of Irf8 and Ncor2 (SMRT). The bottom panel shows violin plots comparing expression of Irf8 and Ncor2 in WT and HSL-ko uninjured pre-osteoclasts. N.S. = adjusted pval >0.05.

3.6. DGE analysis of uninjured pre-osteoclasts reveal a large number of genes with Irf8 promoter binding

We used Enrichr to further understand the differences between the uninjured pre-osteoclasts [33,34]. Only HSL-ko uninjured pre-osteoclast DEGs had a number of associated significant GO terms (Fig. 6D). These include pattern recognition receptor (PRR) signaling pathway, Toll-like receptor (TLR) signaling pathway, and neutrophil mediated immunity as the top three significant GO terms (Fig. 6D). We note that there was a lot of overlap in genes triggering these GO terms. We also looked at transcription factors with enrichment by chromatin immunoprecipitation (ChIP)-sequencing in promoters of genes queried by using the ChEA 2016 database within Enrichr [33,34]. We found Irf8 to be the most significantly enriched transcription factor for both WT and HSL-ko uninjured DEGs (Fig. 6E). Smrt (Ncor2) was also found enriched for the HSL-ko uninjured DEGs (Fig. 6E). Examination of Irf8 expression among the pre-osteoclasts revealed that only a subset of the cells expressed Irf8 although there was no significant difference between WT and HSL-ko uninjured expression (Fig. 6F). Irf8 has been noted as a transcription factor expressed only during part of osteoclastogenesis and downregulated in later stages [35]. This suggests the difference between WT and HSL-ko pre-osteoclasts may be due to difference in the stage of osteoclastogenesis. Finally, we asked, among uninjured pre-osteoclasts, what genes were consistently highly enriched in either the WT or HSL-ko setting. We report the top 10 scoring genes for both the WT and HSL-ko conditions as well as expression of some of them among the pre-osteoclasts (Fig. 6G). Among the WT enriched genes, Fabp4 in particular stood out. Fabp4 has been previously implicated in wound healing and inflammation, making its downregulation in HSL-ko pre-osteoclasts potentially relevant to the enhanced regeneration phenotype observed [36,37].

3.7. Examination of pre and 5-day post injury pre-osteoclasts reveal changes in pre-osteoclast maturity in both WT and HSL-ko populations

We next looked at post-injury pre-osteoclasts to ascertain differences between the injured and uninjured states as well as differences between the WT and HSL-ko after injury (Fig. 7). After normalization by cell counts, injured pre-osteoclasts were shown in a UMAP chart (Fig. 7B). We wanted to understand whether differences between the HSL-ko pre-osteoclasts and WT pre-osteoclasts could again be explained by differences in stages of maturation. We examined the expression of two early pre-osteoclast genes, Irf8 and Adgre1 (F4/80), as well as two late stage pre-osteoclast genes, Fcgr4 and Acp5 (Fig. 7C) [35,38–40]. WT pre-osteoclasts prior to injury showed lower expression of early pre-osteoclast genes and higher expression of late pre-osteoclast genes. Post-injury WT pre-osteoclasts showed an opposite trend. HSL-ko pre-osteoclasts prior to injury showed lower expression of late pre-osteoclast genes and higher expression of early pre-osteoclast genes. Post-injury HSL-ko pre-osteoclasts showed an opposite trend to their uninjured counterparts (Fig. 7C).

Fig. 7.

Examination of post-injury pre-osteoclasts reveals changes in maturation and gene expression. A) Isolation of pre-osteoclasts from the total dataset. B) Two dimensional UMAP plots comparing pre-osteoclasts from injured WT and HSL-ko mice. Cells were subsetted to ensure equal cell numbers in plots. C) Violin plots comparing expression of osteoclastogenesis stage specific markers prior and after injury for WT and HSL-ko mice. D) Heatmap clustering the top 500 variable genes for all pre-osteoclasts. Genes presented are also seen in at least 10% of pre-osteoclasts. Each pre-osteoclast condition had its cells’ expression averaged. Conditions were clustered for similarity in gene expression. Genes were split into four groups based on clustering. Bioplanet 2019 pathways and GO biological processes for each group are shown.

To better clarify how these different states of pre-osteoclasts compare to each other, we did hierarchical clustering on the top 500 variable genes seen in at least 10% of pre-osteoclasts (Fig. 7D). WT uninjured and HSL-ko injured pre-osteoclasts clustered the closest while WT injured pre-osteoclasts and HSL-ko uninjured pre-osteoclasts clustered closer together (Fig. 7D). Genes were put into groups based on their position in the hierarchical gene clustering. Genes were partitioned into four groups based on clustering and the top three pathways found by BioPlanet 2019 are shown. Additionally, the top three GO Biological processes are shown for each gene group (Fig. 7D). Interleukin-2 (IL-2) signaling was an enriched pathway from gene groups 1 and 2. IL-2 signaling has been previously shown to be beneficial to osteoclast function [41,42]. Despite gene groups 2 and 4 triggering the same inflammation related GO biological process terms, they were associated with distinct pathways. Furthermore, while WT uninjured pre-osteoclasts have low expression of genes from group 2 and mid to low expression of group 4 genes. Post-injury WT pre-osteoclasts maintain mid to low expression of group 4 genes while highly expressing group 2 genes (Fig. 7D). This could suggest group 4 genes are more general inflammation related genes expressed by pre-osteoclasts while group 2 are responsive inflammation related genes. Interestingly, prior to injury, HSL-ko pre-osteoclasts express high levels of both group 2 and group 4 genes (Fig. 7D).

3.8. No detectable change in inflammation levels of the uninjured HSL-ko mice

Given the increased inflammation seen in the adipose tissue in HSL-ko mice, the downregulation of FABP4 in the HSL-ko pre-osteoclasts prior to injury, and the better repair at the injury site, we tested for changes in inflammation levels between uninjured WT and HSL-ko mice prior to injury [36,43]. We first examined a number of inflammation related genes in our scRNAseq study (Fig. S4). We found very small fold changes for the various inflammation related genes, although some were considered significantly different (Fig. S4A). We further tested whether inflammation was altered by following up with these genes with qRT-PCR. We found no significant difference in inflammation markers (Fig. S4B). These inflammation related genes we tested are expressed by a number of different cell types analyzed in our study (Fig. S4C). This indicates that prior to injury there does not seem to have been a general difference in the inflammation of the medullary cavity. Furthermore, as these markers largely correspond with granulocyte and monocyte/macrophage cells, this provides further evidence (corroborating the scRNAseq data) that there were no significant differences in population levels of these early injury hematoma cells (Fig. S4C).

3.9. Day 5 post-injury differences in medullary cavity cells

We also measured the cellular composition of the medullary cavity using Drop-seq and 10 × 3′ counting (Fig. S5–7). Drop-seq had been used initially in the study as a pilot before focusing on the pre-injury states of the medullary cavity. We had difficulty running mouse medullary cavity samples using Drop-seq during the initial stages of the study and this led us to switch to using 10 × 3′ counting for the study. While we do note differences in population levels of cell types and gene expression differences, we suspect these differences are inherent to differences in stages of injury repair rather than being due directly to the HSL-ko (Figs. S5B-C, S7).

4. Discussion

In this work, we studied the effects of HSL-ko on bone injury repair in mice. We had initially hypothesized that this would lead to increased osteo/chondrogenesis of MSCs, thus facilitating improved bone repair. Using a combination of μCT, tissue staining, and transplantation experiments, we were able to confirm the pro-reparative effects of the HSL-ko on bone repair. We then investigated potential mechanisms behind the improved repair. We found that HSL inhibition led to improved MSC differentiation in general in vitro. We also found that MSC recruitment to the injury site was improved in the HSL-ko mice. Finally, considering the improved MSC recruitment, we considered osteoimmunological implications by conducting scRNAseq on cells flushed from the tibia and femur medullary cavities prior to injury and 5 days post injury. Here we found substantial changes in the pre-osteoclast population, though significant changes in TRAP staining were not observed.

Once MSCs are recruited to the injury site, it is imperative that they differentiate to chondrocytes and osteoblasts to help facilitate construction of cartilage and bone [14]. Studies examining osteogenic Wnt signaling in mice have shown that inhibition of the signaling leads to an impairment in fracture repair [44–46]. This impairment in signaling led to a decrease in osteoblast numbers and higher MSC numbers indicating impaired differentiation. This evidence highlights the importance of osteoblast numbers to fracture repair. Therefore, the increased MSC differentiation capacity of the HSL inhibition is well in line with improving bone repair (Fig. 3). Indeed, we also showed in an earlier study that 14 month old HSL-ko mice have higher numbers of osteoblasts under normal homeostatic conditions and that differentiated bone marrow MSCs from HSL-ko mice express higher alkaline phosphatase activity [9].

Tissue regeneration is a complex task in which tissue resident macrophages have recently garnered increased attention. Tissue resident macrophages serve to maintain tissue homeostasis, signal in case of injury or disease, and facilitate tissue repair [38,47]. Osteoclasts are a bone resident macrophage population known to be involved in bone remodeling, immune signaling, and fracture repair [14,15,48]. The most perturbed cell type from our scRNAseq analyses were the pre-osteoclasts. This was not unexpected as Pparγ has been extensively shown to regulate osteoclastogenesis and osteoclast function [11–13]. When starving Pparγ of ligands, as is the case in an HSL-ko, an alteration to the osteoclast and pre-osteoclast populations is foreseeable. One limitation of the scRNAseq analyses was cell size, and, therefore, observing fully mature osteoclasts was not possible due to their size. Despite this limitation, we still uncovered interesting differences between WT and HSL-ko pre-osteoclasts prior to injury. Additionally, we provided one of the most extensive transcriptomic mappings of osteoclastogenesis at the single cell level in mice.

A key question in finding the change to pre-osteoclast behavior is why this contributes to accelerated bone repair. A potential factor could be that subsets of pre-osteoclasts express different genes, and thus those genes show preferential expression in either the WT or HSL-ko mice. Two examples of this were the genes Fcna and Fabp4. Fabp4 has been linked to regulation of inflammation in multiple contexts and, therefore, could be particularly relevant to changes in bone injury repair during the inflammatory phase [49]. Fcna is a ficolin that works as a pattern recognition factor detecting microbes and subsequently activating the lectin complement system [50,51]. The complement system helps facilitate phagocytosis and removal of dead cells by marking them [52]. Phagocytosis of dead cells has been shown to lead to anti-inflammatory and pro-resolution factor expression in macrophages, as well as less professional phagocytic cells [53]. Additionally, members of the complement system have been implicated in fracture repair [54–56]. Although notably, components of the complement system seem to either have beneficial or detrimental effects in different contexts [54–56]. Increased expression of Fcna ahead of the injury could therefore potentially lead to better opsonization and clearance of dead cells while promoting anti-inflammatory behavior by phagocytic cells such as the recruited monocytes/macrophages. Follow-up focusing on these factors may prove fruitful for uncovering therapies to improve fracture and bone injury repair.

A limitation of our study was the use of mice with a total knockout of HSL, making it difficult to determine which changes were due to direct or indirect effects of the absence of HSL. Another limitation is the lack of stromal cells within our scRNAseq dataset. We did visualize some osteoblasts; however, their numbers were very limited, thus preventing extensive analysis of these cells. Furthermore, and unexpectedly, we found much fewer MSCs in our dataset than expected given earlier FACS data. One explanation is that these cells were present, but expression of typical surface markers used in FACS are not as clear in scRNAseq data. This would be consistent with the low number of Pdgfrα and Sca1 double positive cells we found in the scRNAseq data (Fig. S8). To address this, we created a reference dataset composed of previously published datasets containing various different bone marrow stromal and immune populations. MSCs from this reference set failed to label strongly any clusters in our scRNAseq dataset (data not shown). This indicates that MSCs did not make it into our final scRNAseq dataset despite being seen in decent numbers in the FACS data from HSL-ko mice (Fig. 4). We also assessed Pdgfrα and Sca1 double positive cells in the raw barcodes from our 10× datasets and still failed to see any more cells. As samples for FACS and scRNAseq were prepared in a similar manner up to the respective assay, this result is perplexing and requires further investigation.

5. Conclusions

In conclusion, we analyzed bone injury repair in an HSL-ko mouse compared to WT controls. We identified two potential effectors of improved injury repair. The first is the increased propensity for MSC differentiation following HSL inhibition. The second is the alteration in pre-osteoclast behavior. Moving forward, studies should investigate the increased localization of MSCs at the injury site in HSL-ko mice. This is suggestive of differences in osteoimmunological behavior. In particular, we need to better understand how changes in pre-osteoclasts may influence these differences in osteoimmunological behavior. Furthermore, studies taking a proteomic approach to comparing differences in cytokine levels (i.e. Cxcl12) will be useful for further fleshing out the mechanisms underlying the improved bone repair of HSL-ko mice.

Supplementary Material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.bbadis.2022.166449.

Abbreviations

ATGL

adipose triglyceride lipase

BMSC

bone marrow mesenchymal stem cells

BV

bone volume

ChIP

chromatin immunoprecipitation

DEG

differentially expressed genes

DGE

differential gene expression

DXA

dual energy X-ray absorptiometry

FA

fatty acid

FACS

fluorescently activated cell sorting

HSL

hormone sensitive lipase

KO

knockout

MSC

mesenchymal stem cell

PFA

paraformaldehyde

PPAR

peroxisomal proliferator activated receptor

PRR

pattern recognition receptor

scRNAseq

single cell RNA sequencing

TbN

trabecular number

TbSp

trabecular spacing

TbTh

trabecular thickness

TRAP

tartrate resistant acid phosphatase

TV

total volume

WT

wild type