A novel skeletal-specific adipogenesis pathway defines key origins and adaptations of bone marrow adipocytes with age and disease

Bone marrow adipocytes (BMAs) accumulate with age and in diverse disease states. However, their age- and disease-specific origins and adaptations remain unclear, impairing our understanding of their context-specific endocrine functions and relationship with surrounding tissues. In this study, we identified a novel, bone marrow-specific adipogenesis pathway using the AdipoqCre+/DTA+ ‘fat free’ mouse (FF), a model in which Adipoq-Cre drives diphtheria toxin-induced cell death in all adiponectin-expressing cells. Adiponectin is highly expressed by BMAs, peripheral adipocytes, and a subset of bone marrow stromal progenitor cells with preadipocyte-like characteristics. Consistent with this, FF mice presented with uniform depletion of peripheral white and brown adipose tissues, in addition to loss of BMAs in canonical locations such as the tail vertebrae. However, unexpectedly, a distinctly localized subset of BMAs accumulated with age in FF mice in regions such as the femoral and tibial diaphysis that are generally devoid of bone marrow adipose tissue (BMAT). Ectopic BMAs in FF mice were defined by increased lipid storage and decreased expression of cytokines including hematopoietic support factor Cxcl12 and adipokines adiponectin, resistin, and adipsin. FF BMAs also displayed resistance to lipolytic stimuli including cold stress and β3-adrenergic agonist CL316,243. This was associated with reduced expression of adrenergic receptors and monoacylglycerol lipase. Global ablation of adiponectin-expressing cells regulated bone accrual in an age- and sex-dependent manner. High bone mass was present early in life and this was more pronounced in females. However, with age, both male and female FF mice had decreased cortical thickness and mineral content. In addition, unlike BMAs in healthy mice, expansion of ectopic BMAs in FF mice was inversely correlated with cortical bone volume fraction. Subcutaneous fat transplant and normalization of systemic metabolic parameters was sufficient to prevent ectopic BMA expansion in FF mice but did not prevent the initial onset of the high bone mass phenotype. Altogether, this defines a novel, secondary adipogenesis pathway that relies on recruitment of adiponectin-negative stromal progenitors. This pathway is unique to the bone marrow and is activated with age and in states of metabolic stress, resulting in expansion of BMAs specialized for lipid storage with compromised lipid mobilization and endocrine function within regions traditionally devoted to hematopoiesis. Our findings further distinguish BMAT from peripheral adipose tissues and contribute to our understanding of BMA origins and adaptation with age and disease.


SUMMARY 1
Bone marrow adipocytes (BMAs) accumulate with age and in diverse disease states. However, their age-2 and disease-specific origins and adaptations remain unclear, impairing our understanding of their context-3 specific endocrine functions and relationship with surrounding tissues. In this study, we identified a novel, 4 bone marrow-specific adipogenesis pathway using the Adipoq Cre+/DTA+ 'fat free' mouse (FF), a model in 5 which Adipoq-Cre drives diphtheria toxin-induced cell death in all adiponectin-expressing cells. Adiponectin 6 is highly expressed by BMAs, peripheral adipocytes, and a subset of bone marrow stromal progenitor cells 7 with preadipocyte-like characteristics. Consistent with this, FF mice presented with uniform depletion of 8 peripheral white and brown adipose tissues, in addition to loss of BMAs in canonical locations such as the 9 tail vertebrae. However, unexpectedly, a distinctly localized subset of BMAs accumulated with age in FF 10 mice in regions such as the femoral and tibial diaphysis that are generally devoid of bone marrow adipose 11 tissue (BMAT). Ectopic BMAs in FF mice were defined by increased lipid storage and decreased expression 12 of cytokines including hematopoietic support factor Cxcl12 and adipokines adiponectin, resistin, and 13 adipsin. FF BMAs also displayed resistance to lipolytic stimuli including cold stress and β3-adrenergic 14 agonist CL316,243. This was associated with reduced expression of adrenergic receptors and 15 monoacylglycerol lipase. Global ablation of adiponectin-expressing cells regulated bone accrual in an age-16 and sex-dependent manner. High bone mass was present early in life and this was more pronounced in 17 females. However, with age, both male and female FF mice had decreased cortical thickness and mineral 18 content. In addition, unlike BMAs in healthy mice, expansion of ectopic BMAs in FF mice was inversely 19 correlated with cortical bone volume fraction. Subcutaneous fat transplant and normalization of systemic 20 metabolic parameters was sufficient to prevent ectopic BMA expansion in FF mice but did not prevent the 21 initial onset of the high bone mass phenotype. Altogether, this defines a novel, secondary adipogenesis 22 pathway that relies on recruitment of adiponectin-negative stromal progenitors. This pathway is unique to 23 the bone marrow and is activated with age and in states of metabolic stress, resulting in expansion of BMAs 24 specialized for lipid storage with compromised lipid mobilization and endocrine function within regions 25 traditionally devoted to hematopoiesis. Our findings further distinguish BMAT from peripheral adipose 26 tissues and contribute to our understanding of BMA origins and adaptation with age and disease. Bone marrow adipose tissue (BMAT) is a unique fat depot located within the skeleton. BMAT acts as an 2 endocrine organ and energy storage depot and has the potential to contribute to the regulation of 3 metabolism, hematopoiesis, and bone homeostasis (reviewed in (1)). The development and subsequent 4 regulation of bone marrow adipocytes (BMAs) varies between skeletal sites (2-5) and current work 5 suggests that BMAs are functionally unique within the context of their niche (6). Specifically, the constitutive 6 BMAT (cBMAT) begins to form in distal regions at or slightly before birth, followed by rapid expansion and 7 maturation early in life (3,5). By contrast, the regulated BMAT (rBMAT) develops later and expands with 8 age, generally in areas of active hematopoiesis (3,5). Recent studies in rodents and humans have also 9 highlighted the heterogeneous metabolic properties of BMAs (2,7), suggesting that their capacity for 10 functional support of surrounding cells may change, particularly with age and in states of systemic disease.

11
BMAT expansion occurs in diverse conditions including anorexia, obesity, aging, osteoporosis, 12 hyperlipidemia, estrogen deficiency, and treatment with pharmacotherapies such as glucocorticoids and 13 thiazolidinediones (1,8). Many of these conditions are associated with increased fracture risk. Thus, 14 understanding the context-specific origins and function of BMAs has important implications for development 15 of clinical and pharmacologic strategies to support skeletal and metabolic health. 16 Genetic causes of lipodystrophy have provided clues about the molecular differences between BMAT and 17 white adipose tissues (WAT) (reviewed in (5)). Congenital generalized lipodystrophy (CGL) is a disorder 18 characterized by complete loss of peripheral adipose tissues and is associated with secondary 19 complications including hypertriglyceridemia, osteosclerosis, insulin resistance, diabetes, and hepatic 20 steatosis (5,9-11). Patients with CGL uniformly lack WAT, however, BMAT is selectively preserved in those 21 with CGL resulting from mutations in CAV1 (CGL3) or PTRF (CGL4) but not AGPAT2 (CGL1) or BSCL2 22 (CGL2) (5). Similarly, all BMAT is retained in Cav1 knockout mice and cBMAT is present in Ptrf knockouts 23 (3). These results in humans and mice suggest that, unlike WAT, BMAT has unique compensatory 24 mechanisms that promote its preservation. In this study, to define the cellular basis for this observation, we 25 examined the formation and regulation of BMAT in the 'fat free' Adipoq Cre+/DTA+ mouse (FF), a novel genetic 26 model of CGL (10,12). 27 In the FF mouse, any cell that expresses adiponectin (Adipoq-Cre+) will express diphtheria toxin A (DTA), 28 leading to DTA-induced cell death (13,14). Adiponectin is a secreted adipokine that is expressed by all 29 brown, white, and BMAT adipocytes in healthy mice, independent of sex (15). Expression of adiponectin 30 also defines the major BMA progenitor, termed 'Adipo-CAR' cells (adipogenic CXCL12-abundant reticular) 31 or 'MALP' (marrow adipogenic lineage precursor) (16-18). In the FF mouse, we hypothesized that ablation 32 of adiponectin-expressing cells would promote activation of alternate, adiponectin-negative skeletal 33 progenitors to form adipocytes in vivo in times of systemic metabolic demand. To test this hypothesis, we 34 performed adiponectin lineage tracing of bone marrow stromal cells and BMAs. We also analyzed age-and 35 sex-associated changes in bone, BMAT, and peripheral adipose tissues in control and FF mice. In addition, 36 we defined the impact of adrenergic stimulation and peripheral fat transplantation on the formation and 37 regulation of BMAT in the FF model. This work refines our understanding of the origins and adaptation of 38 BMAT with age and disease and defines compensatory pathways of adipocyte formation that are unique to 39 the bone marrow and emerge in states of compromised progenitor function and altered lipid load. 40

RESULTS 41
Adiponectin is expressed by BMAT adipocytes and a subset of stromal progenitor cells. 42 As described previously (15), Adipoq Cre+/mTmG+ lineage tracing reporter mice were used to localize 43 adiponectin-expressing cell lineages within the skeletal niche. In this model, any cell having expressed 44 adiponectin (Adipoq-Cre+) at any time during its genesis will change plasma membrane color from red to 45 green (mT→mG, (19)). Cross-sections of the proximal tibia and tail vertebrae were imaged at 3-and 16-46 weeks of age in both males and females after immunostaining for green fluorescent protein (GFP), red 47 fluorescent protein (RFP), and perilipin 1 (PLIN1). Adipoq Cre-/mTmG+ littermates were used as a negative 48 control. In Adipoq Cre+/mTmG+ male mice, this work confirmed that membrane-localized GFP expression was 49 present in all PLIN1+, rBMAT adipocytes within the proximal tibia (Fig.1A). Prevalent GFP labeling of 50 reticular stromal cells and bone lining cells was also noted (Fig.1A). GFP expression was absent in 1 hematopoietic cells, chondrocytes, and osteocytes (Fig.1A). Similarly, the cells lining the endosteal bone 2 surface were predominantly GFP/Adipoq negative (Fig.1A). In negative controls, all cells within the bone, 3 including adipocytes and bone-lining cells, stained positive for RFP and negative for GFP (Fig.1B). 4 Comparable patterns of GFP expression were observed in the proximal tibia of Adipoq Cre+/mTmG+ female 5 mice at both 3-and 16-weeks of age ( Fig.1C and data not shown). In the tail vertebrae, Adipoq-Cre traced 6 all PLIN1+ cBMAT adipocytes independent of sex or age, as indicated by GFP (Fig.1D). 7 8 To determine if adiponectin was expressed by the BMAT progenitor cell, we isolated primary bone marrow 9 stromal cells from the femur and tibia of 16-week old male Adipoq Cre+/mTmG+ mice for colony-forming unit 10 (CFU) assays. After 2-weeks of expansion ex vivo, an average of 79.8±9.0% of CFUs were completely 11 positive for adiponectin, as indicated by expression of membrane-bound GFP in 100% of the fibroblast-12 appearing progenitor cells within the colony, 16.5±9.1% of CFUs were negative (RFP+ only) and 3.7±0.3% 13 were mixed, containing both GFP+ and RFP+ fibroblasts ( Fig.2A,B). Within these transitional colonies, cells 14 with RFP+ membranes, indicative of their lack of adiponectin expression, routinely contained GFP+ 15 cytoplasmic granules (Fig.2C). This was often near to cells that had already become fully GFP+ (Adipoq-16 Cre+), suggesting that adiponectin expression is activated at later stages of stromal progenitor maturation. 17 Small, RFP+, myeloid-lineage cells were commonly present, particularly around the edges of the plates 18 ( Fig.2A) vivo, PLIN1+ lipid droplets were only present in GFP+ cells in vitro (Fig.2D,E). In negative controls, RFP+ 23 stromal cells and PLIN1+ adipocytes were observed without the presence of GFP+ (Fig.2F). Together, 24 these results suggest that all BMAs and their progenitor cells express adiponectin in healthy conditions. 25

Global ablation of adiponectin-expressing cells causes sex-and age-dependent regulation of bone. 26
Male and female FF mice were analyzed at 4-months and 8-months of age relative to Adipoq Cre-/DTA+ control 27 littermates (Con) to isolate sex and age-related changes in body mass, bone, and bone marrow adiposity absence of fat was accompanied by secondary sequelae including pronounced liver enlargement and 31 steatosis (Fig.3B,D) and elevated blood glucose (Fig.3E). Bone size, as indicated by tibia length, was 32 reduced by 3-7% in FF male and female mice relative to controls (Fig.3F). Body mass was unchanged at 4-33 months. However, from 4-months to 8-months of age, male FF mice resisted age-associated gains in body 34 mass relative to controls (Fig.3G). By contrast, female FF mice were 9-13% heavier than controls at both 35 ages examined and did not exhibit age-associated restriction (Fig.3G).

37
To assess bone morphometry, tibiae were scanned by µCT. Consistent with a previous report in younger 38 males (10), trabecular bone in both male and female FF mice extended deeper into the diaphysis than 39 controls (Fig.4A). In the proximal tibial metaphysis, female FF mice had increased trabecular bone volume 40 fraction (BVF), number, thickness, and bone mineral density (BMD), with decreased trabecular spacing at 41 both 4-and 8-months of age ( Fig.4B-F). Increases in metaphyseal trabecular bone were less prominent in 42 the 4-month old male FF mice (Fig.4B). Unlike females, trabecular number was the only factor that was 43 increased significantly in males ( Fig.4C) with a comparable decrease in spacing at 4-months of age 44 (Fig.4E). By 8-months of age, metaphyseal trabecular BVF, BMD, number, and spacing in male FF mice 45 were comparable to controls (Fig.4B,C,E,F). In addition, unlike females, male FF mice had decreased 46 trabecular thickness relative to the control group at 8-months of age (Fig.4D). This reveals that ablation of 47 adiponectin-expressing cells is sufficient to promote sustained increases in trabecular bone in females, but 48 not in males.

50
Female FF mice at 4-months also had significantly higher cortical BVF and cortical thickness than controls 51 ( Fig.S1A-C). Increased cortical thickness was associated with decreased medullary area and no change in 52 total area, indicative of increased endosteal bone (Fig.S1D,E). In male FF mice at 4-months, increased 53 cortical BVF was also associated with decreased medullary area ( Fig.S1A-F). However, unlike in females, 1 the total area was also decreased (Fig.S1E), reflecting an overall decrease in bone cross-sectional size. 2 With age, both male and female FF mice exhibited a significant -19.0% and -19.1% decrease in tibial 3 cortical bone thickness, respectively (Fig.S1B). This was in direct contrast to control mice, where tibial 4 cortical thickness remained constant (male) or was increased by +16% (female) with age (Fig.S1B). 5 Changes in the bone mineral content (BMC) mirrored this result, with age-associated increases in controls, 6 but not in FF mice (Fig.S1G). There were no significant differences in predicted torsional bone strength by 7 polar moment of inertia at any of the ages examined (Fig.S1H). Overall, this demonstrates that ablation of 8 adiponectin-expressing cells promotes early gains in the amount and thickness of cortical bone, however, 9 these increases are not sustained and tend to be normalized or decreased relative to controls with age. 10

Global ablation of adiponectin-expressing cells drives ectopic expansion of BMAT. 11
Tibiae from the 4-and 8-month old FF and Con mice were decalcified and stained with osmium tetroxide for 12 visualization and quantification of BMAT. Unlike peripheral adipose tissues, the 3D-reconstructed images of 13 the osmium-stained tibiae indicated that BMAT was still present (Fig.5A,B). When quantified and expressed 14 relative to total bone marrow volume, the percentage total tibial BMAT was comparable to controls in 4-15 month old FF male and female mice and in 8-month old males (Fig.5C). In control mice, BMAT was 16 localized in the well-established pattern of concentration within proximal and distal ends of the tibia 17 ( Fig.5A,B) (3). By contrast, the BMAT in the FF mice was found predominantly in the proximal tibia and mid-18 diaphyseal region with few adipocytes in the distal tibia ( Fig.5A,B). Consistent with the 3D reconstructions, 19 regional sub-analyses revealed that retained BMAT adipocytes were primarily localized proximal to the 20 tibia/fibula junction (Fig.5D). Within the proximal tibia, BMAT increased by 2.2-fold in control males and 5.6-21 fold in control females from 4-to 8-months of age (Fig.5D). In FF mice, though the absolute volume of 22 BMAT was similar or less than controls (Fig.5D), proximal tibial BMAT increased by 6.9-fold and 23.2-fold 23 with age in males and females, respectively (Fig.5D). This included expansion within the mid-diaphysis, a 24 region in mice that is generally relatively devoid of BMAT (Fig.5B). In the distal tibia, control males and 25 females had a large volume of BMAT at 4-months that also increased by 2.7 and 2.3-fold with age ( Fig.5E).

26
By contrast, FF mice had very little BMAT in the distal tibia and, though minor increases with age were 27 noted, these changes were not significant (Fig.5E). Distal tibia BMAT often behaves similarly to constitutive 28 BMAT in regions such as the tail vertebrae (2,3). Consistent with this, BMAT adipocytes were generally 29 absent in the 8-month-old FF tail vertebrae, a region of dense cBMAT-like adipocytes in control mice 30 (Fig.5F). These findings demonstrate that BMAT persists in FF mice despite global ablation of adiponectin-31 expressing cells and, further, that these ectopic BMAT adipocytes expand with age primarily in regions 32 traditionally comprised of red bone marrow.

34
By histology, FF BMAT adipocytes in the tibia and femur were morphologically comparable to control BMAT 35 adipocytes (Fig.6A). FF BMAs contained a large, central PLIN1+ lipid droplet (Fig.6B,C) and were negative 36 for macrophage-marker CD68 (Fig.6C). Histologic sections also confirmed the DTA-mediated depletion of 37 the peripheral peri-skeletal adipose tissues in FF mice (Fig.6A). For example, control mice had infrapatellar 38 PLIN1+ adipocytes in the knee joint region (Fig.6A,C). By contrast, in FF mice, the infrapatellar adipocytes 39 were replaced with a population of foam-cell like, auto-fluorescent, PLIN1-, CD68+ macrophages 40 (Fig.6A,C). The same result was observed in the extra-skeletal adipose tissues surrounding the tail 41 vertebrae and the bones in the feet (Fig.S2). This confirms that adipocyte cell death occurs uniformly in the 42 peripheral fat tissues, with selective adipocyte preservation within the bone marrow of FF mice. FF BMAT 43 adipocytes were on average 13.7% and 42.9% larger than controls in male and female mice, respectively, 44 reflecting increases in lipid storage (Fig.6D). Purified FF BMAs also demonstrated a unique gene 45 expression profile. As expected, expression of Cre was elevated in FF mice ( Fig.6E) with paired decreases 46 in Adipoq (Fig.6F). Similarly, expression of cytokines including stromal cell-derived factor 1, also known as 47 C-X-C motif chemokine 12 (Cxcl12), adipsin (Cfd), and resistin (Retn) were significantly decreased (Fig.6F).

48
Expression of adipogenic transcription factor peroxisome proliferator-activated receptor gamma (Ppary) was 49 also decreased. By contrast, expression of CCAAT/enhancer-binding protein alpha (Cebpa), fatty acid 50 transporter Cd36, alkaline phosphatase (Alpl), and diphthamide biosynthesis 1 (Dph1) were comparable in 51 control and FF BMAs. Overall, this defines the FF BMA as a PLIN+, CD68-adipocyte with increased lipid 52 storage and decreased expression of cytokines including adiponectin, resistin, adipsin, and Cxcl12. 53 Ectopic BMAT in FF mice is not regulated by cold stress or β3-adrenergic stimulation. 1 Regulation of BMAT adipocytes by adrenergic stimulation has important implications for the functional 2 integration of BMAs with local and peripheral energy stores. FF mice lack WAT and BAT and have impaired 3 thermoregulatory capabilities. Thus, control and FF mice are bred and housed at thermoneutrality (30°C). 4 To assess the response of FF bone and BMAs to thermal stress, male control and FF mice were housed at 5 thermoneutrality (30 o C) or room temperature (22 o C) for 3-4 months, beginning at 4-weeks of age. Under 6 mild cold stress (22 o C), trabecular BVF was decreased by 33-57% relative to housing at thermoneutrality 7 (30 o C) in the tibia and femur of control and FF mice (Fig.7A). By comparison, cortical thickness was 9-13% 8 lower in control mice at 22 o C but remained unchanged in FF mice, regardless of temperature or analysis 9 site (Fig.7B). By osmium µCT, BMAT in the proximal tibia was 82% lower in control mice housed at 22 o C 10 than in mice housed at 30 o C (Fig.7C,D). However, similar to cortical bone, proximal tibial BMAT remained 11 unchanged with mild cold stress in FF mice (Fig.7C,D). Retention of BMAT in the FF mice was also 12 prevalent in the femur and presented with the same atypical pattern of accumulation in the mid-diaphysis. 13 Regulation of BMAT within the femur mirrored that observed in the proximal tibia, though it did not reach 14 statistical significance (Fig.7E,F). Unexpectedly, BMAT in both control and FF mice in the distal tibia 15 increased by 1.5-and 5-fold, respectively, in mice housed at 22 o C (Fig.7C,D) 8A). This response to CL316,243 was absent in FF mice (Fig.8A). After 10-days, β3-AR stimulation 28 decreased BMAT adipocyte cell area in the proximal tibia by 26% in control mice (Fig.9B,C). This reflects an 29 estimated 37% decrease in adipocyte cell volume (  =   4  3 3 , 3D, µm 3 ). BMAT size was 30 unchanged by β3-AR stimulation in FF mice ( Fig.9B,C). Expression of β3 adrenergic receptor (Adrb3) was 31 significantly decreased in FF mice (Fig.8D). Gene expression of adipose triglyceride lipase (Pnpla2) and 32 hormone-sensitive lipase (Lipe) were comparable in FF BMAs (Fig.8D). However, expression of 33 monoglyceride Lipase (Mgll) was significantly reduced (Fig.8D). Together, these results suggest that the 34 ectopic BMAT in FF mice is resistant to cold and β3-AR agonist-induced lipolytic stimulation. 35

Subcutaneous fat transplant prevents ectopic BMAT expansion in FF mice. 36
We hypothesized that BMAT expansion in FF mice occurs secondary to peripheral fat depletion and 37 hypertriglyceridemia. To test this hypothesis, male and female FF and control mice underwent sham 38 surgery or were transplanted subcutaneously with wild type adipose tissue at 3-to 5-weeks of age. After 39 surgery, mice were monitored for 12-weeks prior to euthanasia. There were no differences in the body mass 40 of the male mice over time regardless of genotype or transplant (Fig.9A). In females, consistent with 41 previous 4-and 8-month old cohorts (Fig.3E), the body mass of the non-transplanted FF mice was 14-16% 42 higher, on average, than controls (Fig.9A). Subcutaneous fat transplant reduced the body mass of the FF 43 female mice to control levels ( Fig.9A). Fat transplant was also sufficient to normalize the hyperglycemia 44 present in both the male and female FF animals ( Fig.9B). At the end point, the total mass of the 45 transplanted adipose tissue was significantly higher in the FF mice than in the controls (Fig.9C). As has 46 been reported previously (10), engrafted adipose tissue fragments resembled subcutaneous white adipose 47 tissue at the time of sacrifice (Fig.S3). There was also a significant rescue of liver enlargement and 48 peripheral hypertriglyceridemia by fat transplant in both male and female FF mice (Fig.9D,E). Fat transplant 49 did not substantially modify the cortical and trabecular bone phenotypes in the tibia (Fig.S4). However, the 50 3D rendering and quantification of tibial BMAT revealed that most of the BMAT present in FF mice was 51 eliminated after subcutaneous fat transplantation ( Fig.9F-H). An independent increase in tibial BMAT 1 volume was also observed in male WT fat transplanted mice (Fig.9F,H). The reason for this is unclear as no 2 differences were noted in BAT, iWAT, or gWAT mass after fat transplant in male or female control mice 3 ( Fig.S5A-C). Overall, these results reinforce the critical role of the peripheral adipose tissue as a lipid 4 storage depot that reduces the systemic burden of hypertriglyceridemia on peripheral tissues such as liver 5 and bone marrow. 6

DISCUSSION 7
It has previously been assumed that all adipocytes, including bone marrow adipocytes, express the 8 adipokine adiponectin (15,21,22). And, conversely, that adiponectin is not expressed by cells that are not 9 adipocytes. However, recent lineage tracing and single-cell RNAseq studies, including the data presented 10 here, challenge this paradigm and demonstrate that adiponectin is expressed by a subset of bone marrow 11 stromal progenitor cells. These adiponectin-expressing progenitors overlap with CAR cells (17,18,23) and 12 have been more recently termed MALPs (16). They are largely positive for PDGFRβ and V-CAM-1 and 13 have a unique gene expression pattern that mimics known features of pre-adipocytes (17,18,23). 14 Adiponectin-expressing stromal progenitors appear after birth (P1+), matching the known development of 15 BMAT which also occurs primarily postnatally (3,5). Consistent with this and likely due also to the high 16 expression and secretion of adiponectin by healthy BMAs (22), classic depots of rBMAT and cBMAT failed 17 to form in the Adipoq Cre+/DTA+ FF mouse (3). However, instead, an ectopic population of FF BMAs developed 18 with age in regions of the skeleton such as the diaphysis that are generally devoid of BMAT. 19 This ectopic BMA population did not rescue circulating adiponectin in the FF mice and had decreased 20 expression of Adipoq, reinforcing the efficacy of the DTA. The location of the FF BMAs aligns with known 21 sites of arteriolar entry and distribution within the femur and tibia (24,25). These cells also expressed 22 Cxcl12, though this was decreased relative to control BMAs. Arterioles have recently been defined as a site 23 of Osteo-CAR cells, a subpopulation of Cxcl12-expressing cells that are enriched for osteogenic progenitors 24 while Adipo-CAR or MALP cells are primarily localized to the venous sinusoids (16,18). Alternate 25 adiponectin-negative, Cxcl12-negative mesenchymal progenitor populations also exist within bone (26-28). 26 Peri-sinusoidal Adipo-CAR/MALP progenitor cells are generally primed to undergo adipocyte differentiation, 27 however, they are also recruited to undergo differentiation into trabecular bone osteogenic cells with age 28 (~35% at 6-months) and into cortical osteoblasts and osteocytes during injury-induced skeletal repair 29 (17,18) ( Fig.10). Our results mirror these findings and support an inverse model whereby the depletion of 30 peri-sinusoidal, adiponectin-expressing MALP/Adipo-CAR progenitor cells drives the preferential 31 differentiation of adiponectin -/lo , Cxcl12 -/lo progenitors into adipocytes in states of local and systemic 32 metabolic demand, as occurs in CGL (Fig.10). This secondary adipogenesis pathway is unique to the bone 33 marrow and is absent in peripheral adipose tissue depots including WAT and BAT, helping to explain the 34 relative preservation of BMAs relative to WAT in clinical states of CGL and reinforcing their likely importance 35 to maintaining the local homeostasis of the skeletal and hematopoietic microenvironment. 36 The ectopic BMAT in FF mice expands with age and has a larger volume in females than in males, which is 37 a general trend that exists in normal BMAT (29). The 87% decrease in Cxcl12 expression in FF BMAs also 38 aligns with what has been previously reported in aged BMAs (30). Specifically, a 46% decrease in Cxcl12 39 expression was observed in BMAs from 18-month old mice relative to BMAs isolated at 6-months (re-40 analyzed microarray data from (30)). Decreased BMA-specific expression of Cxcl12 also occurs in obese 41 mice fed with high-fat diet relative to controls (-24 to -41%, re-analyzed microarray data from (31)). 42 Decreased expression of Adipoq has also previously been highlighted as a feature of aged BMAs (30). 43 Thus, we propose that expansion of an adiponectin -/lo , Cxcl12 -/lo BMA population is a conserved adaptation 44 with age and in states of metabolic stress (Fig.10). Functionally, decreases in stromal and BMA-derived 45 Cxcl12 may contribute to decreased focal support of hematopoiesis (32), helping to explain the well-defined 46 pattern of bone marrow atrophy and BMA expansion that occurs with age and disease (1,5,30). In addition 47 to Cxcl12 and Adipoq, FF BMAs had significant decreases in expression of adipokines including adipsin and 48 resistin, suggesting that these cells may have limited endocrine functions relative to controls. 49 The deficient response to lipolytic agonists including cold exposure and β3-adrenergic receptor stimulation, 1 larger cell size, and sustained expression of fatty acid transporter Cd36 also suggests that FF BMAs have 2 decreased capacity to serve as a local fuel source for surrounding hematopoietic and osteogenic cells, 3 preferring instead to take up and to store lipid. This result provides insight into recent conflicting studies on 4 rodent and human BMA lipolysis (2,7). Specifically, purified BMAs from healthy rodents are capable of 5 responding to lipolytic agonists such as forskolin (2). However, purified adipocytes from older humans are 6 not (7). In humans, this was found to be due to a selective decrease in expression of key lipase Mgll (7), a 7 serine hydrolase that catalyzes the conversion of monoacylglycerides to free fatty acids and glycerol. 8 Similar to aged human BMAs, we found that FF BMAs have decreased expression of Mgll with comparable 9 expression of lipases Lipe (Hsl) and Pnpla2 (Atgl) relative to controls (Fig.8E). This suggests that rodent 10 BMAs can undergo the same adaptations as are present in aged humans, contributing to their resistance to 11 lipolysis. However, unlike humans (7), we also observed decreased expression of Adrb3 in mice. 12 Systemic abnormalities including hepatic steatosis, hyperglycemia, and hypertriglyceridemia in FF mice 13 were rescued by subcutaneous fat transplant. Similarly, fat transplant prevented expansion of ectopic BMAs 14 (Fig.9). This supports the existing paradigm whereby excess circulating lipids contribute to the development 15 and expansion of BMAT, and, conversely, that the decrease of these factors in circulation may reduce the 16 development of ectopic BMAs in bone. It is unclear if this could be accomplished in relatively healthy mice 17 or humans to limit age-associated increases in BMAs in regions of hematopoietic bone marrow. And, 18 beyond this, if such a strategy would have benefits to bone or hematopoiesis. The alteration of other 19 circulating factors such as estradiol, leptin, adiponectin, insulin, corticosterone, and catecholamines may 20 also contribute to BMAT expansion in FF mice. Future work is needed to address these questions. 21 Regarding bone, µCT scans of the FF tibia revealed a general enhancement of bone formation in young FF 22 mice (Fig.4,5), as has been reported previously (10,16). However, this was paired with a mild, but significant 23 decrease in tibial length in FF mice, suggesting that the depletion of adiponectin-expressing cells limits 24 bone elongation, potentially through the suppression of chondrocyte differentiation and/or proliferation within 25 the growth plate. This process may also be mediated by reduced levels of growth hormone or growth 26 factors in circulation or within the bone microenvironment. Consistent with this observation, short stature is 27 consistently reported in patients with CGL3, whose BMAT is also preserved (33). In addition, µCT scans of 28 the FF tibia showed a sex-dependent regulation of bone formation in FF mice. Specifically, trabecular bone 29 formation in FF mice was more pronounced in females with significant differences in age-related trabecular 30 maintenance. It has previously been reported that these Adipoq Cre+/DTA+ FF mice have low levels of estradiol 31 (34), which may contribute to the sex-dependent changes in bone that were observed in these mice. Historically, BMAT expansion has been linked to bone loss and osteoporosis (8). In recent years, this 41 assumption has come into question as multiple models have shown that both high bone mass and high 42 BMAT can occur simultaneously (3,5,8). In addition, a recent study using Prx1-cre mediated knock-out of 43 PPARγ demonstrated that BMAT expansion was not necessary for age-associated bone loss in female 44 mice, though it was sufficient to intensify age-dependent cortical porosity (35). Our study was not designed 45 to isolate the direct relationship between bone and BMAT. However, some observations can be made in the 46 FF model. First, the initial onset of the bone phenotype was independent of the formation of ectopic BMAs, 47 as demonstrated in the fat transplant experiment, suggesting that these early phenotypes are independent. 48 When considered across experiments (n=32 control, 29 FF), trabecular bone volume fraction in the tibia 49 was inversely correlated with metaphyseal BMAT volume in control mice (Slope -2.0, R 2 0.137, p=0.037). 50 This is consistent with previous reports in aged mice and humans. However, this association was lost in FF 51 mice (Slope +0.7, R 2 0.073, p=0.155). Instead, diaphyseal BMAT was inversely correlated with cortical bone 1 volume fraction in FF mice (Slope -2.1, R 2 0.306, p=0.002), but not in controls (Slope +1.8, R 2 0.044, 2 p=0.252). In healthy mice, the majority of BMAs are derived from Adipoq+ marrow adipocyte lineage 3 progenitors (16) (Fig.10). MALP cells form a peri-sinusoidal network throughout the bone marrow and were 4 recently shown to secrete factors such as RANKL, providing active suppression of trabecular, but not 5 cortical, bone mass (36). Ablation of these cells in the FF mice may help to explain the absence of a 6 correlation between FF BMAT and trabecular bone. Instead, this new, ectopic adipocyte population appears 7 to be more closely linked to cortical bone. Though we cannot presume causation, we hypothesize that this 8 relationship may be due to the origination of these BMAs from key adiponectin-negative/low progenitor 9 populations such as Osteo-CAR cells that localize to the peripheral arterioles and endocortical surfaces (18) 10 (Fig.10). Future clarification of this point will undoubtedly help to reveal the niche-specific regulation of and 11 associations between bone, blood, and BMAT. 12

Conclusions 13
With age, BMAs demonstrate gene changes associated with the inflammatory response, mitochondrial 14 dysfunction, and lipid metabolism (30). The unique nature of mature BMA metabolism is not replicated in 15 BMAs differentiated from progenitor cells in vitro (7). Consistent with this, our work demonstrates that the 16 spatially defined progenitor cell, the systemic metabolic profile, and the local microenvironment are 17 necessary regulators of BMA expansion and adaptation in vivo. In addition, we present evidence for a 18 conserved secondary adipogenesis pathway that is unique to the bone marrow and is activated during times 19 of metabolic stress (Fig.10). The resulting adipocytes express low levels of hematopoietic-and metabolic- Tukey's multiple comparisons test. ANOVA results as indicated. *p≤0.05. Data presented as mean ± SD. 1 WT and FF mice were housed at 30 o C on a 12h/12h light/dark cycle.

. BMAT is present in Adipoq Cre+/DTA+ fat free (FF) mice and expands with age. (A,B) 4
Representative μCT images of osmium-stained tibiae for both male and female Adipoq Cre+/DTA+ fat free (FF) 5 and Adipoq Cre-/DTA+ control (Con) mice at (A) 4-months and (B) 8-months of age. Bone marrow fat is in dark 6 grey and bone is in light grey. (C) Quantification of total tibial BMAT volume as a percentage of total bone 7 marrow volume. (D) Regional analysis of BMAT within the proximal end of the same tibiae as in (C), 8 expressed as the total volume of osmium-stained lipid from the proximal end of the tibia to the tibia/fibula 9 junction. (E) Regional analysis of BMAT within the distal end of the same tibiae as in (C), expressed as the 10 total volume of osmium-stained lipid from tibia/fibula junction to the distal end of the bone. Consistent with this and likely due also to the high expression and secretion of adiponectin by healthy 33 BMAs, classic depots of bone marrow adipose tissue (BMAT) failed to form in the Adipoq Cre+/DTA+ FF mouse. 34 Instead, we observed age-dependent expansion of a BMA population with reduced expression of 35 adiponectin (Adipoq -/lo ) and Cxcl12 (Cxcl12 -/lo ) in regions of the skeleton such as the diaphysis that are 36 generally devoid of BMAT. FF BMAs were resistant to cold challenge and β3-adrenergic stimulation and 37 had decreased expression of β3-adrenergic receptors and monoacylglycerol lipase (Mgll), suggesting that 38 they have decreased capacity to serve as a local fuel source for surrounding hematopoietic and osteogenic 39 cells. We hypothesize that these cells originate from progenitors including the peri-arteriolar Osteo-CAR 40 population, similar to previous work showing that Adipo-CAR cells are capable of undergoing osteogenic 41 differentiation with age and after acute injury. We propose that expansion of this BMA population is a 42 conserved adaptation with age and in states of metabolic stress and, furthermore, that this is a unique 43 adaptation of the bone marrow that is not present in peripheral adipose depots. Functionally, decreases in 44 stromal and BMA-derived Cxcl12 may contribute to decreased focal support of hematopoiesis, helping to 45 explain the well-defined pattern of bone marrow atrophy and BMA expansion that occurs with age and 46 disease.  the Adipoq-Cre promoter lack both white and brown adipose tissues and were bred and housed at 11 thermoneutral temperature (30 o C). All transgenic mice were maintained on a C57BL/6J background (Strain 12 #000664). Body mass was recorded with an electronic scale and blood glucose was monitored by tail prick 13 with a glucometer (Contour Next). For end points requiring tissue mass measurements, mice were 14 euthanized with carbon dioxide followed by cervical dislocation. Tissues were collected and weighed using 15 an electronic scale. For end points requiring histology and immunostaining, mice were anesthetized with 16 ketamine/xylazine cocktail (100mg/kg ketamine; 10 mg/kg xyalzine) and perfused through the left ventricle 17 of the heart with 10 mL phosphate-buffered saline followed by 10 mL 10% neutral buffered formalin (NBF, 18 Fisher Scientific 23-245684). When indicated, tibia and femur lengths were determined using a digital 19 caliper (iKKEGOL). For all experiments, collected tissues were post-fixed in 10% NBF for 24-hours. For 20 western blot and serum assays, as detailed below, blood was collected through capillary action from the 21 lateral tail vein and serum was isolated by centrifugation at 1500 x g for 15 minutes after clotting on ice. 22

Histology and Immunostaining 33
Paraffin immunostaining and imaging. Paraffin embedding, slide preparation, and H&E stains were 34 performed by the WUSM Musculoskeletal Histology and Morphometry core. Bones were fully decalcified in 35 14% EDTA (Sigma-Aldrich E5134), pH 7.4 prior to embedding. For immunostaining, 10 µm paraffin sections 36 were rehydrated in a series of xylene and ethanols prior to antigen retrieval with 10 mM sodium citrate 37 buffer (pH 6.0, 20-minutes, 90-95°C or overnight at 55 o C). Antibodies used for paraffin immunostaining are 38 detailed in Table S1. Paraffin Immunofluorescence: Retrieved sections were permeabilized for 10-minutes 39 in 0.2% Triton-X in PBS, blocked for 1-hour with 10% donkey serum (Sigma-Aldrich D9663) in TNT buffer 40 (0.1 M Tris-HCL pH 7.4, 0.15 M sodium chloride, 0.05% Tween-20), and incubated for 24-hr at 4°C with 41 primary antibodies followed by washing and secondary detection (Table S1). Secondary antibodies in TNT 42 buffer were applied for 1-hour at room temperature. Nuclei were counterstained in 1 μg/mL DAPI (Sigma-43 Aldrich D9542) for 5-min prior to mounting in Fluoromount-G (ThermoFisher, 00-4958-02). All washes 44 between steps were performed three times each in TNT buffer. Paraffin Immunohistochemistry: Tissue 1 sections were permeabilized for 10-minutes in 0.2% Triton-X in PBS, blocked for 1-hr in kit-specific blocking 2 reagent (ImmPRESS HRP Goat Anti-Rabbit IgG Polymer Detection Kit, Vector Laboratories, MP-7451), and 3 incubated for 24-hr at 4°C with primary antibody (Table S1). Sections were washed in TNT and endogenous 4 peroxidase activity was quenched in 0.3% hydrogen peroxide (Sigma-Aldrich 216763) in PBS for 30-5 minutes. Sections were then incubated with ImmPRESS polymer reagent for 30-minutes prior to 6 development with peroxidase substrate solution. Slides were counterstained with hematoxylin (Ricca 7 Chemical 3536-16) and dehydrated through a reverse ethanol gradient prior to mounting in Permount. 8 Images were taken using a Nikon Spinning Disk confocal microscope or a Hamamatsu 2.0-HT NanoZoomer 9 System with NDP.scan 2.5 image software. 10 Frozen immunostaining and imaging. Tissues were embedded in OCT mounting media (Fisher HealthCare 11 23-730-571) and cut at 50 μm on a cryostat (Leica). Bones were fully decalcified in 14% EDTA, pH 7.4 prior 12 to embedding. Sections were blocked in 10% donkey serum in TNT buffer prior to incubation for 48-h with 13 primary antibodies (Table S1). After washing, secondary antibodies in TNT buffer were applied for 24-hours 14 at 4°C (Table S1). The sections were then washed and incubated in DAPI for 5-min prior to mounting with 15 Fluoromount-G. Images were taken at 10x on a Nikon spinning disk confocal microscope. 16

Bone marrow stromal cell (BMSC) isolation, cell culture, and immunostaining. 17
Immediately after euthanasia by CO2, long bones were harvested under aseptic conditions. The ends of the 18 long bones were cut to allow flushing of marrow contents, as described previously (37). Cells were 19 suspended in MesenCult Expansion Medium (STEMCELL Technologies 05513) containing MesenCult 20 Basal Medium, MesenCult 1X Supplement, 0.5 mL MesenPure, 1X L-Glutamine, and 1X 21 penicillin/streptomycin. Primary bone marrow cultures were plated at a density of 2.0 x 10 6 cells/cm 2 and 22 incubated at 37°C, 5% CO2. After 48-hr, nonadherent cells were removed with subsequent media changes 23 occurring every 2-3 days. After 14 days, colonies were fixed with methanol prior to permeabilization with 1% 24 Triton X-100 (Sigma-Aldrich 9002-93-1) in PBS for 10-min at room temperature. Cells were blocked with a 25 solution containing PBS, 10% donkey serum, and 0.1% Triton X-100 for 30-min prior to incubation for 24-hr 26 at 4°C with primary antibodies (Table S1). Cells were then washed prior to application of secondary 27 antibodies in PBS and 0.1% Triton X-100 for 30-min at room temperature. Nuclei were stained with 1 µg/mL 28 DAPI for 5-min prior to mounting in Fluoromount-G. Images were taken at 4x and 20x using a Nikon 29 spinning disk confocal microscope. 30 1

Computed tomography and osmium staining 2
Bones were embedded in 2% agarose prior to scanning at 20 µm voxel resolution using a Scanco µCT 40 3 (Scanco Medical AG). Analysis was performed according to reported guidelines (38). For cancellous bone, 4 100 slices (2 mm) below the growth plate, beginning where the primary spongiosa was no longer visible, 5 were contoured and analyzed at a threshold of 175 (on a 0-1000 scale relative to a pre-calibrated 6 hydroxyapatite phantom). For cortical bone, 20 slices (400 µm) located 2 mm proximal to the tibia-fibula 7 junction were contoured and analyzed at a threshold of 260. To assess bone marrow adiposity, bones were 8 decalcified in 14% EDTA, pH 7.4 and incubated in a solution containing 1% osmium tetroxide (Electron 9 Microscopy Sciences 19170) and 2.5% potassium dichromate (Sigma-Aldrich 24-4520) for 48-hours (39). 10 After washing for 2-hours in running water and storage in PBS at 4°C, osmium-stained bones were 11 embedded in 2% agarose and scanned at 10 µm voxel resolution (Scanco μCT 40; 70 kVp, 114 µA, 300 ms 12 integration time). Regions of interest were contoured for BMAT quantification as detailed in the figure  13 legends. BMAT was segmented with a threshold of 400. 14

Bone marrow adipocyte cell size analysis 15
Tiled 10x images covering the femoral and tibial metaphyses were exported from the Nanozoomer scans of 16 H&E stained slides and processed in Fiji to estimate average adipocyte cell size (40). Based on previous 17 recommendations for adipocyte cell size analyses (41), a minimum of 100 adipocytes were analyzed for 18 each mouse. Briefly, the scale in Fiji was set to be consistent with the original scan. The image was then 19 converted to 8-bit and a threshold of 230 to 255 was applied to create a mask. Then the image was cleaned 20 up using the wand tool and the deletion command to eliminate non-adipocyte structures. The cleaned mask 21 was processed using the Fill Holes and the Watershed tools. The size of adipocytes was determined using 22 the "Analyze Particles" tool by setting the size to 200 to 4000 μm 2 and circularity to 0.40 -1.00. Histograms 23 were created in GraphPad Prism and the average adipocyte cell size was calculated using Excel.  collagenase and finely minced to liberate any residual BMAs. Bone and bone marrow preparations were 1 centrifuged at room temperature, 400g x 2 min and BMA-containing supernatant was decanted into a new 2 tube prior to re-centrifugation at 400g x 1 min. Infranatant and any residual pellet was removed using a 3 pulled glass pipet until only 1-2 mL of liquid was remaining. The adipocyte-containing liquid was serially 4 applied to a NucleoSpin® Filter Column (NucleoSpin RNA XS Kit, Takara, 740902) for on-column BMA lysis 5 and RNA extraction. Briefly, the filter column was centrifuged slowly at 50 g x 10 seconds to retain the BMA 6 cells while removing any residual liquid into the collection tube. The bottom of the column was then sealed 7 with parafilm and kit-supplied RNA lysis buffer was added with gentle agitation. BMA-enriched ('BMAe') 8 lysates were processed for RNA extraction using the kit-supplied protocol and reagents. 9 For qPCR, 100 ng of total RNA was reverse transcribed into cDNA using SuperScript IV VILO Master Mix 10 with ezDNase™ Enzyme (Thermo Fisher Scientific 11766050) according to the manufacturer's instruction. 11 SyGreen 2x Mix Lo-ROX (PCR Biosystems PB20.11-51) was used to perform the qPCR assay on a 12 QuantStudio™ 3 Real-Time PCR System (Thermo Fisher Scientific A28136). Gene expression of individual 13 targets was calculated based on amplification of a standard curve for each primer. Results were normalized 14 to the geometric mean of housekeeping genes Ppia and Tbp. Primer sequences are listed in Table S2. 15

Fat transplantation 17
Adipoq Cre-/DTA+ and Adipoq Cre+/DTA+ mice received subcutaneous fat transplant or sham surgery at 3-to 5-18 weeks of age. Mice were maintained for an additional 12-weeks prior to sacrifice (end age 15-to 17-weeks).

19
Donor preparation: wild type donor mice on the same background (C57BL/6J) ranged from 22-to 39-days 20 of age. Immediately after decapitation under anesthesia, bilateral inguinal WAT depots were dissected free 21 of surrounding tissues and placed into sterile PBS in a petri dish. The lymph node was removed and a 22 scalpel blade was used to mince the remaining iWAT into small pieces of ~0.5-1.0 mm 3 . The entire minced 23 iWAT from one donor mouse was transplanted to one recipient mouse. Recipient surgery: the recipient 24 mouse was anesthetized with isoflurane and the skin on the back was prepared (shaved and treated 2x 25 each with 70% ethanol and betadine) prior to making two 1 cm incisions along the midline, one over the 1 shoulder blades and one just above the level of the pelvis. Blunt dissection was used to create four pockets 2 just lateral to each incision, one on each side. The minced iWAT from the donor mouse was evenly 3 distributed into the 4 pockets. The incisions were closed and all mice received Buprenex SR at the time of 4 surgery for post-operative analgesia. Post-surgical monitoring and management were performed per DCM 5 guidelines, as approved in our animal protocol. 6 Serum glycerol and triglyceride assay 7 Serum glycerol and true triglyceride (TG) levels were determined using a Serum Triglyceride Determination 8 Kit (Sigma-Aldrich TR0100). In brief, free glycerol reagent and triglyceride reagent were prepared according 9 to the manufacturer's instruction. To measure serum glycerol, 10 µL serum/well was added to a 96-well 10 microplate on ice prior to addition of 150 μL of free glycerol reagent and incubation at 37°C for 10-minutes. 11 The absorbances of the standards and the samples at 540 nm versus blank (pure Free Glycerol Reagent) 12 were measured using a microplate spectrophotometer (BioTek). To determine serum true TG level, 38 μL of 13 Triglyceride Reagent was added to each well after the initial absorbance measurement for glycerol, followed 14 by an additional 10-minute incubation at 37°C. The absorbances of the standards and the samples at 540 15 nm versus blank were measured again using the microplate reader. A standard curve was utilized for the 16 calculation of serum-free glycerol and total TG concentrations. The serum true TG level was calculated by 17 subtracting the free glycerol level from the total TG level for each sample, as per manufacturer instructions. 18 All samples were assayed in duplicate. 19

Statistics 20
Statistical analyses were performed in GraphPad Prism including unpaired t-test, one-way, two-way, and 21 three-way ANOVA with multiple comparisons tests, applied as detailed in the figure legends. A p-value of 22 less than 0.05 was considered statistically significant. Quantitative assessments of cell size and µCT-based 23 analyses were performed by individuals that were blinded to the sample identity.