Spatially Distinct Bone Marrow Sites Are Asymmetrically Impacted by Inflammatory Cardiovascular Disease

Visual Abstract

Highlights

• •Bone marrow drives cardiovascular inflammation, but studies focus on femoral marrow
• •MarrowMet whole-body 18F-FDG reveals skull and sternum as disease-specific sites
• •Femoral marrow does not reflect systemic activity; profiling must be region-specific

Summary

Cardiovascular disease, a leading cause of mortality globally, is increasingly recognized to involve complex bone marrow-driven inflammatory mechanisms, yet the impact on spatially distinct bone marrow sites and comorbidities remains poorly understood. To address this, we developed MarrowMet, a methodology for whole-body, site-specific quantification of bone marrow activity. The approach involves intravenously injecting the metabolic tracer 18F-fluorodeoxyglucose (¹⁸F-FDG) in mice, followed by bone excision to quantify site-specific bone marrow activity, with values then superimposed on a whole-body mouse atlas. After establishing that ¹⁸F-FDG bone marrow uptake strongly correlated with inflammatory activity, we applied MarrowMet to map site-specific activation patterns across diverse cardiovascular pathologies, including mouse models of inflammatory atherosclerosis, acute ischemic events, acute respiratory distress syndrome, metabolic syndrome, and aging. MarrowMet guided the selection of bone marrow regions of interest for in-depth mass cytometric analyses, with the skull and sternum emerging as critical sites exhibiting distinct immune and metabolic profiles in cardiovascular disease. These results challenge the prevailing view that femoral marrow represents systemic activity. Together, this work lays a foundation for whole-body exploration of bone marrow heterogeneity, yielding critical insights into cardiovascular disease and associated inflammatory responses, and MarrowMet can be readily adopted to profile other immune mechanisms in a variety of pathologies, including cancer and autoimmune diseases.

Inflammation is an essential process in host defense, wound healing, and the maintenance of homeostasis. Unfortunately, inflammation also drives a broad spectrum of immune-mediated pathologies, most notably cardiovascular diseases (CVD), which are markedly exacerbated by aging. As the global population's average age continues to rise, CVD and its acute complications are expected to remain among the leading causes of mortality worldwide.¹ Increasing evidence implicates that the innate immune system and its bone marrow progeny play a pivotal role in the genesis and progression of inflammatory CVD.2, 3, 4, 5 Simultaneously, as humans age, hematopoiesis becomes more myeloid-biased, resulting in low-grade chronic inflammation, a phenomenon known as inflammaging.^6,7 Together with clonal and accelerated hematopoiesis, which are a direct result of inflammatory atherosclerosis,⁸ these bone marrow-originated processes lead to decreased immune cell diversity and compromise immune function, which further exacerbates ongoing pathology, but also fuels comorbidities.⁹ The bone marrow inflammaging framework in cardiovascular pathologies is predicated on the heterogeneity of this process.

In the past 2 decades, preclinical mouse studies have overhauled the way we look at inflammatory CVD.^4,10, 11, 12 As our understanding of bone marrow behavior grows, we acknowledge that distinct regions are asymmetrically impacted by ongoing pathology.13, 14, 15, 16, 17 For example, the skull bone marrow is the most diverse among the studied bones in health and neurologic disorders, in mice and humans.¹⁶ In a different study, Koh et al¹⁴ uncovered specialized properties of skull bone marrow, revealing this site to be largely protected against hallmarks of aging. However, in mice, typically the easy-to-access femoral bone marrow is analyzed, thereby creating a bias towards this specific site and disregarding region-specific bone marrow activity. The emerging insight that the hematopoietic system is heterogeneous creates a pressing need to analyze multiple bone marrow sites simultaneously to fully comprehend their role in health and disease. An important focus in studying inflammation is immunometabolism, which dictates immune cell function. Therefore, novel quantitative immunoprofiling methods must be developed to longitudinally study spatially distinct bone marrow sites in CVD and other inflammation-mediated conditions.

Here, we introduce MarrowMet, a novel method to quantitatively evaluate bone marrow cell metabolism at the whole mouse level. It involves the administration of 18F-fluorodeoxyglucose (¹⁸F-FDG)—a radiolabeled glucose analog—followed by ex vivo gamma-counting to measure this radiotracer's uptake across different bones. Relative uptakes are captured in a heatmap and projected onto a mouse schematic to visualize bone marrow metabolism across different sites. After establishing the methodology's rigor, we applied MarrowMet to several cardiovascular and related inflammatory pathologies, including atherosclerosis and its acute ischemic events, acute respiratory distress syndrome (ARDS), metabolic syndrome (MetS), and aging. We used MarrowMet to guide us to bone marrow regions of interest for in-depth analyses, uncovering unanticipated asymmetric immune processes in these spatially diverse sites.

Methods

Animals

All animal experiments conducted in the Netherlands were approved by the Radboud University Medical Center's Dierexperimentencommissie (DEC) (CCD: AVD10300-2021 14707) and complied with European and Dutch guidelines according to the care and use of laboratory mice. Female C57BL/6J or apoE^−/− mice were purchased from Charles River Germany (or United Kingdom for aged mice) and cohoused at ambient temperature (22-24 °C) and 45%-65% RH (relative humidity). Mice were housed and allocated randomly, and experiments were performed in a blinded manner.

Animal experiments performed in the United States were in accordance with the Icahn School of Medicine at Mount Sinai Institutional Animal Care and Use Committee (IACUC). Female C57BL/6J or apoE^−/− mice were purchased from Jackson Laboratory and cohoused at ambient temperature (22-24 °C) and 45%-65% RH. Mice were housed and allocated randomly, and experiments were performed in a blinded manner. All mice were at 12 weeks of age at endpoint unless otherwise indicated.

¹⁸F-FDG experiments

Mice were injected intravenously with ¹⁸F-FDG (0.375-0.75 MBq g⁻¹ body weight) and euthanized 1 hour or later. For FDG experiments, mice were fasted overnight prior to experiments and maintained under isoflurane anesthesia from the time of ¹⁸F-FDG administration until euthanization. Mice were euthanized using CO₂ inhalation, and organs of interest and femur, tibia and fibula, humerus, radius and ulna, scapula, pelvis, lumbar 6, lumbar 6, thoracic 13, cervical 7, sternum, and skull were collected, thoroughly cleaned, weighed, and the emitted ɣ radiation was measured on a gamma counter (2480 WIZARD2 Automatic Gamma Counter, PerkinElmer). Radioactivity values were corrected for decay and normalized to tissue weight to express radioactivity concentration as percentage of injected dose per gram (%ID/g).

Single-cell suspension from bone marrow

Mice were euthanized via CO₂ inhalation. Femurs, skull, and sternum were dissected and thoroughly cleaned of surrounding tissue.

For femoral bone marrow, femurs were flushed with either Maxpar cell staining buffer (CSB; for cytometry by time-of-flight (CyTOF) only; #201068, Standard BioTools) or flow buffer (phosphate-buffered saline [PBS] with 0.5% BSA, 2 mM ethylenediaminetetraacetic acid) and filtered through a 70 μm cell strainer.

For the sternal and skull bone marrow, the target bone was placed in a prechilled mortar and pestle and mechanically crushed to release marrow. The resulting mixture of marrow and bone fragments was filtered twice through a 70 μm cell strainer.

All bone marrow samples were then centrifuged at 400 × g for 5 minutes at 4 °C, and the supernatant was discarded. Samples were lysed using 1 mL red blood cell lysis buffer (#555899, BD Pharm Lyse) and incubated for 1 min at room temperature. The lysis reaction was stopped by quenching with CSB or flow buffer, followed by centrifugation at 400 × g for 5 minutes at 4 °C.

Total cell counts were obtained using a CASY cell counter (OMNI Life Science)

For bone marrow–derived macrophage (BMDM) experiments, the same procedure was followed as described above, but in sterile conditions using sterile buffers.

¹⁸F-FDG cellular uptake experiments

Single-cell suspension and cell counts were obtained, and the emitted ɣ radiation was measured on a gamma counter. Samples were then subjected to CD11b magnetic-activated cell sorting (#130-049-601, Miltenyi Biotec) according to the manufacturer’s protocol. Both CD11b+ (myeloid cells) and CD11b-fractions were gamma counted, and cell counts were obtained. The CD11b-fraction was then subjected to a second magnetic-activated cell sorting (CD45, #130-052-301, Miltenyi Biotec) according to the manufacturer’s protocol to obtain the CD45+CD11b-fraction (lymphocytes) and the CD45-CD11b-fraction (nonimmune cells). Both fractions were gamma-counted, and cell counts were obtained.

Flow cytometry

Single-cell suspensions were transferred to a Corning V-bottom 96-well plate and incubated with 50 μL Viakrome 808 (#C36628, Beckman Coulter) diluted 100x in PBS for 5 minutes in the dark at room temperature. After washing, Fc receptors were blocked with anti-CD16/32 antibodies Fc-block (101,302, BioLegend) for 5 minutes on ice (not for progenitor panel). Cells were then stained with antibody cocktails diluted in Brilliant Stain Buffer (Invitrogen, total volume 50 μL) and incubated on ice in the dark for 30 minutes. Cells were then washed and analyzed. For intracellular staining of glucose transporter 1 (GLUT1), cells were additionally stained using the True-Nuclear Transcription Factor Buffer Set (#424401, BioLegend) according to the manufacturer’s protocol. The antibodies and gating strategy used for all flow cytometry panels are listed in the Supplemental Methods. All data were acquired using a 21-color CytoFLEX LX (Beckman Coulter). Data were analyzed using FlowJo v10.10.0 (BD Biosciences) or OMIQ (Dotmatics).

CENCAT

Single-cell suspensions were obtained, and cellular energetics through noncanonical amino acid tagging (CENCAT) was performed as described in structured, transparent, accessible, reporting (STAR) protocol¹⁸ using 1 x 10⁶ bone marrow cells per condition. The antibodies used for the flow cytometry panels are listed in the Supplemental Methods.

6AzGal cellular uptake experiments

Glucose analog uptake experiments were performed as described in Tsuchiya et al¹⁹ Mice were injected intravenously with 6-azido-6-deoxygalactose (6AzGal; #712752, Sigma - 20 mg kg⁻¹ body weight) and euthanized 1 hour later. Single-cell suspensions were obtained as described and were transferred to a Corning V-bottom 96-well plate and stained for flow cytometry as previously described. Additionally, after antibody staining, cells were incubated with 100 μL of 100 nM boron-dipyrromethene fluorescein dibenzocyclooctyne (BDP FL DBCO, #BP-23473, BroadPharm) for 15 minutes at room temperature. Cells were washed twice before analysis on the CytoFLEX LX. The antibodies used for the flow cytometry panels are listed in the Supplemental Methods.

Cytometry by time-of-flight

Single-cell suspensions were obtained as described earlier and were treated with 1 μL nuclease (Thermo Fisher Scientific, 25 kU; 1:10,000 dilution) and incubated for 5 min at room temperature. A total of 2 × 10⁶ cells per sample were transferred to 5 mL fluorescence-activated cell sorting tubes and centrifuged at 400 × g for 5 min at 4 °C. Pellets were resuspended in 1 mL rhodium intercalator (#201103A, Standard BioTools) diluted 500 times in Maxpar PBS (#201058, Standard BioTools) and incubated for 20 minutes at room temperature. Cells were washed twice with 2 mL CSB followed by centrifugation at 400 × g for 5 minutes at 4 °C, discarding the supernatant after each wash. Samples were barcoded using the Cell-ID 20-Plex Pd Barcoding Kit (#201060, Standard BioTools) according to the manufacturer’s protocol. Briefly, pellets were resuspended in 1 mL Fix I buffer and incubated for 10 minutes at room temperature. Following fixation, cells were permeabilized in 1 mL barcode perm buffer, centrifuged at 800 × g for 5 minutes at 4 °C, and washed once in the same buffer. Each sample was resuspended in 800 μL barcode perm buffer. Barcodes were resuspended in 100 μL barcode perm buffer and added to the respective samples. Samples were gently mixed, incubated for 30 minutes at room temperature with vortexing halfway through, and then washed twice in staining buffer. Postbarcoding, samples were resuspended in 100 μL CSB. Barcoded skull, sternum, and femur samples were pooled into 3 composite tubes (1 per target bone). Tubes were centrifuged at 800 × g for 5 minutes at 4 °C, supernatants removed, and surface antibody cocktail was added according to titrated staining volumes (see overview in the Supplemental Methods). After a 30-minute room-temperature incubation, cells were washed twice in CSB. Cells were then fixed and permeabilized in 1 mL Fix/Perm buffer (#00-5523-00, Invitrogen) for 1 hour at 4 °C, followed by washing twice in FoxP3 permeabilization buffer. Intracellular antibody cocktail was added and incubated for 1 hour at room temperature. Cells were again washed twice with FoxP3 perm buffer. Lastly, samples were incubated overnight at 4 °C in 1 mL iridium intercalator 1.6% paraformaldehyde solution (#201192A, Standard BioTools). The following day, samples were prepared for CyTOF acquisition.

CyTOF acquisition and data analysis

On the day of acquisition, cells were washed twice in CSB supplemented with benzonase, washed once in Maxpar Cell Acquisition Solution Plus (#201244, Standard BioTools), resuspended in cell acquistion solution containing 1/10 EQ 4 Element Calibration beads (#201078 - Standard BioTools), and acquired at a rate of <300 events/second on a Helios Mass Cytometer. Data were in the.fcs file format. All.fcs files in the experiment were normalized using the calibration beads contained in each sample and then debarcoded using tools within the Helios software. Normalized and debarcoded.fcs files were then uploaded to OMIQ for analysis.

Lipopolysaccharide

Mice were injected intraperitoneally with lipopolysaccharide (LPS) (#L2880, Sigma-Aldrich - 0.5 mg kg⁻¹ body weight) and euthanized at the indicated time points using CO₂.

Atherosclerosis

Female apoE^−/− mice were put on a standard chow diet or a 42 kcal% fat diet (#TD.88137, Inotiv) for 8 weeks. For ARDS experiments, mice were intratracheally injected with LPS (5 mg kg⁻¹ body weight). Mice were euthanized at the indicated time points using CO₂.

Myocardial infarction with ischemia-reperfusion

Twelve-week-old female C57BL/6J mice were subjected to temporary ligation of the left anterior descending coronary artery (LAD) to induce myocardial infarction (MI). Mice were treated with a subcutaneous injection of buprenorphine (0.1 mg kg⁻¹ body weight) before surgery for perioperative analgesia. Anesthesia was induced via intraperitoneal injection of ketamine (100 mg kg⁻¹ body weight) and xylazine (10 mg kg⁻¹ body weight), and maintained with isoflurane (1%-2%) following endotracheal intubation. Body temperature was maintained at 37 °C using a heating pad throughout the procedure, and eye ointment was applied to both eyes to prevent corneal dryness. Sterile conditions were maintained throughout, with operating stations disinfected using chlorhexidine or betadine before and after each case. After surgical prep and shaving of the left thoracic area, a left-sided thoracotomy was performed at the fourth intercostal space. A small retractor was used to keep the surgical field exposed. Pericardiotomy exposed the heart, and the LAD was visualized and ligated with an 8-0 Prolene suture. A small polyethylene tube was placed between the ligation suture and the LAD, and the occlusion was maintained for 30 minutes. Successful LAD occlusion was confirmed by pallor or blanching of the myocardium distal to the ligature. Following reperfusion, the thoracotomy was closed in layers with appropriate sutures. Mice were extubated postoperatively and monitored until fully recovered. Incisions were closed with interrupted sutures, and buprenorphine was administered every 12 hours postoperatively.

Stroke (transient middle cerebral artery occlusion)

Twelve-week-old female C57BL/6J mice received a subcutaneous injection of Carprofen (5 mg kg⁻¹ body weight) before surgery for perioperative analgesia. Anesthesia was induced with 4% isoflurane in a heated induction chamber, after which the mouse was positioned on a heating mat. While maintained under anesthesia (∼1.5% isoflurane in oxygen), the ventral neck (right side), the skull above the right ear, and the periocular region were shaved. Eye ointment was applied to both eyes throughout the procedure to prevent corneal dryness. Sterile conditions were maintained throughout, with operating stations disinfected using chlorhexidine or betadine before and after each case. A midline scalp incision was made between the ears and extended toward the right eye, exposing the skull by separating skin and periosteum. The Laser-Doppler probe and holder were placed 3.5 mm lateral and 2.5 mm caudal to the bregma, and the value was verified for baseline blood flow. A ventral midline neck incision was made from the sternum to just below the chin. The thyroid and sternohyoid muscles were separated and reflected laterally with tissue hooks to expose the common carotid artery (CCA). Three silk sutures were placed under the CCA (caudal, middle, and rostral), after which the caudal and rostral knots were tied (firmly and loosely, respectively). A small arteriotomy was performed in the CCA, and a silicone-coated filament was inserted and advanced to the third (rostral) knot. Opening the knot allowed further advancement through the internal carotid artery into the Circle of Willis to occlude the middle cerebral artery. Occlusion was confirmed by a ≥70%-80% drop in cortical cerebral blood flow measured by Laser-Doppler. Throughout the 30-minute occlusion, anesthesia and temperature were maintained, the area was moisturized with NaCl, and cerebral blood flow stability was monitored. After 30 minutes, the filament was slowly withdrawn to the second knot, which was then tightened. The third knot was tied and all knots trimmed. The wound was irrigated, muscles repositioned, and the skin was closed with interrupted 6-0 sutures. The Laser-Doppler probe was removed, the skull cleaned and irrigated, and the scalp closed. Mice were transferred to a heated recovery cage for at least 30 minutes. Carprofen was administered every 12 hours postoperatively.

Metabolic syndrome

Four-week-old female C57Bl/6 mice were either put on a control diet (#D12450 K, Research Diets) or a 60 kcal% fat (#D12492, Research Diets) diet for a total of 16 weeks, with 2 cohorts undergoing a midway switch. At week 15, mice were fasted overnight, and glucose tolerance was determined by measuring blood glucose levels before (t = 0) and at several time points after intraperitoneal glucose administration (1,000 mg kg⁻¹ body weight). At week 16, mice were euthanized using CO₂.

Bone marrow–derived macrophages

Single-cell suspensions were obtained under sterile conditions as described earlier. Subsequently, bone marrow cell suspensions were centrifuged at 400 × g for 5 minutes at 4 °C. Pellets were resuspended in 3 mL BMDM medium (Roswell Park Memorial Institue 1640 supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin, and 2 mM L-glutamine, and with 15% L929 cell-conditioned medium.

Cells (3 mL per dish) were seeded in 60 mm tissue culture-treated Petri dishes (Corning) and incubated at 37 °C/5% CO₂. On day 3, 1.5 mL fresh BMDM medium was added to each dish. After another 3 days, medium was replaced with 3 mL fresh BMDM medium per dish. On day 7, dishes were washed once with PBS, and macrophages were detached with 1.5 mL TrypLE (Gibco). Cells were counted using a CASY cell counter and centrifuged at 400 × g for 5 minutes.

For cell stimulations, pellets were resuspended to a density of 1 × 10⁶ cells per mL in BMDM medium at 100 μL per well in a 96-well flat-bottom TC-treated plate (Sarstedt) and incubated for at least 3 hours. A 100 μL volume of stimuli was then added, and cultures were incubated for 24 hours. After 24 hours, supernatants were harvested for tumor necrosis factor and IL-6 ELISA (#430916 & #431316, BioLegend). Stimuli: TLR3L = Poly-IC (#tlrl-pic, InvivoGen - 5,000 ng mL⁻¹ end concentration), TLR4L = LPS (#L2880, Sigma-Aldrich – 10 ng mL⁻¹ end concentration), TLR7L = R848 (#tlrl-r848-1, InvivoGen - 1,000 ng mL⁻¹ end concentration).

For phagocytosis experiments, BMDMs were diluted to 1 × 10⁵ cells mL⁻¹ in BMDM medium. Cells were seeded at 100 μL per well (1 × 10⁴ cells per well) in flat-bottom 96-well tissue culture plates (Sarstedt). Plates were then placed in a humidified 37 °C, 5% CO₂ incubator for ≥1 hour before adding pHrodo zymosan bioparticles (#P35365, Invitrogen – 5 μg per well). Plates were briefly centrifuged to ensure bioparticles settled at the bottom of wells. Real-time imaging was performed on the Incucyte Live-Cell Analysis System (Sartorius) with acquisition every 15 minutes for the first 2 hours, then hourly for a total of 24 hours.

Statistical analysis

Continuous data are expressed as the mean ± SD. The unpaired t-test with Welch's correction was used to compare 2 groups, whereas >2 groups were compared using 2-way analysis of variance with Sidak's or Dunnett's post hoc test for multiple pairwise comparisons. Data were analyzed using GraphPad Prism version 10.0.0 (GraphPad Software), and a P value < 0.05 was considered statistically significant. Shapiro-Wilk test was used to test for normality before analysis of variance or t-tests. For MarrowMet experiments, all P-values are listed in the supplemental table.

Results

MarrowMet allows quantitatively profiling spatially distinct bone marrow sites at whole-mouse level

In the clinic, ¹⁸F-FDG positron emission tomography (PET) imaging is the diagnostic workhorse, primarily applied for cancer staging. Every year, 3-5 million ¹⁸F-FDG PET scans are acquired worldwide.²⁰ At the beginning of this century, ¹⁸F-FDG PET was adopted for quantitative inflammation imaging in CVD.²¹ Inspired by anecdotal observations of heterogeneous ¹⁸F-FDG accumulation in different bones, we set out to establish a method to accurately measure ¹⁸F-FDG uptake in multiple bones in mice. Whereas PET imaging is a valuable tool for determining ¹⁸F-FDG distribution in humans, its inherently low spatial resolution does not allow studying different bone marrow sites in mice in vivo. Therefore, we developed an ex vivo, whole-body mouse profiling method called MarrowMet. It involves intravenously administering ¹⁸F-FDG to mice, which is allowed to circulate for 60 minutes, before sacrificing and meticulously extracting 11 bones, including the femur, humerus, scapula, different vertebrae, sternum, and skull. Each individual bone is subsequently weighed and gamma-counted to obtain decay- and weight-corrected ¹⁸F-FDG uptake values.

Using a well-established model for systemic inflammation, we first evaluated MarrowMet's robustness in mouse femurs (Figure 1A). Eight hours following intraperitoneal LPS administration, which activates the immune and hematopoietic system,²² we measured elevated ¹⁸F-FDG uptake in the femoral bone marrow (Figure 1B). In addition to immune cells, bone marrow contains stem cells, adipocytes, and stromal cells, which also take up ¹⁸F-FDG. Therefore, we set out to delineate which cells are responsible for the enhanced ¹⁸F-FDG uptake in response to LPS-induced inflammation. After sacrificing mice, we sorted immune cells (myeloid and lymphoid) and nonimmune cells from the femoral bone marrow. We found that immune cells, mainly myeloid cells, contributed to ¹⁸F-FDG's enhanced uptake during LPS-induced inflammation, whereas ¹⁸F-FDG uptake in non-immune cells remained constant (Figure 1C, Supplemental Figure 1A).

Figure 1.

MarrowMet Allows Quantitative Profiling of Spatially Distinct Bone Marrow Sites at A Whole-Mouse Level

(A) Schematic depicting LPS-induced inflammation experiments: 12-week-old female C57BL/6J mice (n = 5) received intraperitoneal administration of lipopolysaccharide (LPS) (red, dose = 0.5 mg/kg) or phosphate-buffered saline (PBS) (gray). (B) Weight-corrected 18F-fluorodeoxyglucose (¹⁸F-FDG) uptake in the femur. (C) Cell count-corrected ¹⁸F-FDG uptake in myeloid cells (top) and nonimmune cells (bottom). (D) gMFI of 6AzGal-BDP FL in myeloid cells (top) and nonimmune cells (bottom). (E) Glycolytic capacity determined by CENCAT in myeloid cells (top) and nonimmune cells (bottom) (n = 3). (F) Representative flow cytometry plots showing gating hierarchy used for determining glucose transporter 1 (GLUT1)+ cells. (G) GLUT1 gMFI (y-axis) and GLUT1+ cells in nonimmune cells and myeloid cells following (n = 5). (H) Correlation graph with Pearson correlation coefficients between Log₂ fold change in GLUT1 expression and 6AzGal uptake after LPS administration relative to control mice (n = 5). (I) Graphical representation of MarrowMet's workflow: mice are euthanized 60 minutes after ¹⁸F-FDG injection; 11 bones are then collected, cleaned, weighed, and gamma-counted. (J) Weight-corrected ¹⁸F-FDG uptake across different bones following PBS administration (n = 6). (K) Weight-corrected ¹⁸F-FDG uptake across different bones following LPS administration (n = 6). (L) Heatmap showing Log₂ fold change of ¹⁸F-FDG uptake across different bones of LPS vs control mice, which is projected on the graphical representation of a mouse. All the Log₂ fold change values and P-values from MarrowMet can be found in the supplemental table. Continuous data are expressed as the mean ± SD. Uptake for femur was analyzed by unpaired t-test with Welch's correction. Statistically significant differences in uptake are indicated by: ∗P < 0.05 ∗∗P < 0.01. ∗∗∗P < 0.001, BDP FL = boron-dipyrromethene fluorescein; gMFI = geometric fluorescence mean fluorescence intensity; ns = not significant.

Flow cytometric profiling of femurs using a fluorescent glucose analog (6-azido-6-deoxygalactose–BODIPY-FL; 6AzGal-BDP FL¹⁹) corroborated the gamma-counting results, showing the metabolic dye's enhanced uptake in myeloid cells (Figure 1D) and certain progenitor subsets (Supplemental Figure 1B). As an orthogonal method for cell metabolism, we employed CENCAT¹⁸ to determine glycolytic capacity (Figure 1E, Supplemental Figure 1C) and fatty acid and amino acid oxidation (Supplemental Figure 1D) in bone marrow-resident cells. In line with ¹⁸F-FDG uptake experiments, we found that increased metabolic activity in LPS-induced inflammation is myeloid cell-driven, which should be accompanied by increased expression of glucose transporters. Indeed, flow cytometric analysis showed that GLUT1 expression was elevated in innate immune cells (myeloid cells and natural killer (NK) cells; Figures 1F and 1G, Supplemental Figure 1E), which correlated with increased 6AzGal uptake (Figure 1H).

After determining that enhanced ¹⁸F-FDG accumulation in femoral bone marrow is driven by immune cells, we deployed MarrowMet in naive mice, measuring ¹⁸F-FDG uptake across 11 distinct bone marrow sites (Figure 1I). We found the coefficient of variation within the bone marrow samples to be 4%-10%, which allows distinguishing differences ≥25%, indicative of MarrowMet's robustness, rigor, and sensitivity (Figure 1J). LPS treatment resulted in significantly increased ¹⁸F-FDG uptake across all bone marrow sites, showing widespread systemic bone marrow activation (Figure 1K, Supplemental Figure 2). For visualization purposes, MarrowMet captures these differences (as Log₂ fold change) in a heatmap and subsequently projects this heatmap on a graphical image of a mouse (Figure 1L). This map serves as an intuitive guide for subsequent profiling of bone marrow sites of interest.

Longitudinally profiling bone marrow sites in aging mice

After validating MarrowMet's analytical rigor in an LPS model, we extended its application to longitudinally profile bone marrow activity in aging mice. Aging and CVD are inextricably connected, fueling a vicious cycle of mutual aggravation.23, 24, 25, 26 During aging, bone marrow-driven impaired immune responses (immunosenescence) and chronic inflammation (inflammaging) predispose patients to infections and cardiovascular pathologies.^6,27 To delineate cell responsiveness and phagocytic capacity during aging, we analyzed femoral BMDMs from mice of different ages (Figures 2A and 2B). BMDMs from aged mice displayed heightened tumor necrosis factor production upon toll-like receptor (TLR) stimulation, indicative of inflammaging, whereas BMDMs from young mice (<5 weeks old) exhibited blunted responses, indicative of an underdeveloped immune system. Similarly, the phagocytic capacity of BMDMs was attenuated in both young mice and old mice, indicative of impaired effector function and immunosenescence as a hallmark of marrow aging. Using the established correlation between ¹⁸F-FDG uptake and inflammation, we quantified ¹⁸F-FDG uptake in the femur, revealing that metabolic activity decreased with age (Figure 2C) and inversely correlated with BMDM responsiveness (Figure 2D). Extending this analysis, we mapped ¹⁸F-FDG uptake across distinct bone marrow sites (Figure 2E, Supplemental Figure 3B), uncovering a metabolic trajectory during aging: juvenile (3 weeks) mice displayed hypermetabolic marrow, which dipped in middle age (48 weeks) and re-emerged in aged (91 weeks) mice, suggestive of late-life chronic inflammation. When additionally profiling BMDMs from the skull and sternum (Figure 2F, Supplemental Figure 3D), we found the sternum and femur to undergo aging-associated shifts, whereas BMDMs from the skull appeared to be less impacted in old mice. These data agree with prior findings that skull bone marrow is uniquely protected from several hallmarks of aging.¹⁶ Collectively, our findings underscore the profound influence of aging on bone marrow activity, with metabolic and functional reprogramming of immune (progenitor) cells that may critically shape susceptibility to cardiovascular pathologies throughout life.

Figure 2.

Longitudinally Profiling Bone Marrow Sites in Aging Mice

(A) Femoral bone marrow–derived macrophages (BMDMs) of 3-week, 5-week, 12-week, 48-week, or 91-week-old female C57BL/6J mice were stimulated with (TLR3 ligand (Poly-IC), TLR4 ligand (lipopolysaccharide), or TLR7 ligand (R848) for 24 hours, and tumor necrosis factor (TNF) levels in cell supernatant were determined using ELISA. (B) Phagocytotoxic capacity of BMDMs of mice of 5 age-groups, measured by quantification of the uptake of zymosan-coated fluorescent particles using live-cell imaging. (C) Weight-corrected 18F-fluorodeoxyglucose (¹⁸F-FDG) uptake in the femur in mice of 5 age-groups. (D) TNF levels (left y-axis) from (A) and ¹⁸F-FDG uptake from (C) as a function of age of mice (x-axis) for all 3 TLR ligands. (E) MarrowMet was used to quantify ¹⁸F-FDG uptake in different bones displayed as a Log₂ fold change (relative to 12-week-old mice) heatmap, which is projected on the graphical representation of a mouse. All the Log₂ fold change values and P-values from MarrowMet can be found in the supplemental table. (F) TNF levels in cell supernatant of BMDMs from all 5 age-groups derived from femur (left), skull (middle), and sternum (right) after 24-hour stimulation with TLR3 ligand, TLR4 ligand, or TLR7 ligand. For all experiments with aged mice, 6 mice per group were used. Continuous data are expressed as the mean ± SD. TLR3 = toll-like receptor 3.

Femoral bone marrow is minimally impacted in metabolic syndrome

Following experiments in aging mice, we studied bone marrow heterogeneity in a mouse model of MetS. MetS, defined as the co-occurrence of obesity, hypertension, impaired glucose tolerance, hypertriglyceridemia, and low HDL-cholesterol, and strongly associated with chronic low-grade inflammation, constitutes a leading risk factor for CVD, more than doubling the likelihood of adverse events and mortality.^23,28 In a population-based cohort, bone marrow ¹⁸F-FDG uptake was strongly associated with the presence of MetS and its components.²³ As inflammation drives this process and fuels disease progression,²⁹ we used MarrowMet to study (the onset of) bone marrow activation in MetS.

To induce MetS, mice were fed a 60 kcal% fat or control diet for 16 weeks, with 2 additional crossover cohorts undergoing a mid-course switch (Figure 3A). As expected, prolonged high-fat feeding induced obesity and impaired glucose tolerance (Figures 3B and 3C). After 8 weeks, pronounced weight gain and glucose intolerance had already manifested, but this could still be almost completely reversed by shifting to a healthy diet. At first glance, unchanged ¹⁸F-FDG uptake in the femur suggested no obvious bone marrow activation following the various diet regimens. However, MarrowMet allowed us to evaluate the bone marrow activation of 11 distinct sites, revealing that MetS induces significant bone marrow activation throughout the body (Figure 3D). Bone marrow activation was detectable as early as 8 weeks after starting a 60 kcal% fat diet, suggesting inflammation occurs before advanced disease onset (Figure 3E, Supplemental Figure 4A). Strikingly, although switching back to a healthy diet reversed MetS clinical symptoms (Figures 3B and 3C), it only partially attenuated bone marrow activation, pointing toward more irreversible rewiring. This “obesogenic memory” mirrors observations in adipocytes and hematopoietic stem cells, where epigenetic imprints persist after weight loss and contribute to relapse in metabolic disease.^30,31 Thus, bone marrow reprogramming emerges as an early hallmark of MetS, shaping systemic immunity before manifestation of clinical diseases.

Figure 3.

Femoral Bone Marrow is Minimally Impacted in Metabolic Syndrome

(A) Schematic depicting metabolic syndrome experiments: 4-week-old female C57BL/6J mice were either put on a control diet or a 60 kcal% fat diet for a total of 16 weeks. After 8 weeks, a diet switch from 60 kcal% fat to a control diet (dark blue, n = 6) or a diet switch from control to a 60 kcal% fat diet (light blue, n = 6) was applied. Two separate groups continued for 16 weeks on the same diet without interventions, either control (gray, n = 12) or 60 kcal% fat (red, n = 12). (B) Body weight of mice following different regimens. (C) Blood glucose levels before (t = 0) and after intraperitoneal glucose (1,000 mg/kg) administration after 15 weeks of diet to determine glucose tolerance. (D) MarrowMet was used to quantify weight-corrected 18F-fluorodeoxyglucose (¹⁸F-FDG) uptake in the femur (left) and cumulative uptake in all bones (right). Uptake was analyzed by Welch's analysis of variance (ANOVA), comparing the groups with the control group only. Statistically significant differences in uptake are indicated by: ∗P ≤ 0.05. (E) ¹⁸F-FDG uptake in different bones displayed as a Log₂ fold change (relative to control) heatmap, which is projected on the graphical representation of a mouse. All the Log₂ fold change values and P-values from MarrowMet can be found in the supplemental table. Continuous data are expressed as the mean ± SD. Weight and GTT data are expressed as the mean ± SEM. Uptake was analyzed by one-way ANOVA with Dunnett's post hoc test, comparing the diet groups with the control group only. Statistically significant differences in uptake are indicated by: ∗P < 0.05, GTT = glucose tolerance test; ns = not significant.

Sternal bone marrow activity dominates in inflammatory atherosclerosis

After employing MarrowMet in aging and MetS mice, 2 major comorbidities for CVD, we applied MarrowMet to uncover novel aspects of bone marrow site heterogeneity in inflammatory atherosclerosis. To this end, apoE^−/− mice were either given a standard chow diet or an 8-week atherogenic diet (composition 42 kcal% fat; Figure 4A). MarrowMet showed widespread bone marrow activation in most bones in atherosclerotic mice (Figure 4B), recapitulating observations from ¹⁸F-FDG imaging studies in atherosclerosis patients.²³ Interestingly, sternum bone marrow showed the strongest activation (P = 0.014) as compared to other bones, such as the typically analyzed femur, while the skull showed decreased ¹⁸F-FDG uptake, highlighting heterogeneity in bone marrow activation during atherosclerosis.

Figure 4.

Sternal Bone Marrow Activity Dominates in Inflammatory Atherosclerosis

(A) Schematic depicting atherosclerosis experiments: 4-week-old female apoE^−/− mice (n = 6) were either put on a conventional (gray) or 42 kcal% fat (purple) diet for 8 weeks. (B) MarrowMet was used to quantify weight-corrected 18F-fluorodeoxyglucose (¹⁸F-FDG) uptake in different bones, displayed in a bar graph and Log₂ fold change heatmap, which is projected on the graphical representation of a mouse. All the Log₂ fold change values and P-values from MarrowMet can be found in the supplemental table. (C) Glucose transporter 1 (GLUT1) expression in leukocytes in femoral, skull, and sternal bone marrow, as determined by cytometry by time-of-flight (CyTOF). (D) Total number of leukocytes in 3 bone marrow sites of interest in a 42 kcal% fat diet vs control group as determined by CyTOF. Leukocyte counts were analyzed by unpaired t-test with Welch's correction. Statistically significant differences in uptake are indicated by: ∗P < 0.05. (E) Cells analyzed with CyTOF concatenated from femoral (left), skull (middle), and sternal (right) bone marrow were clustered with tSNE (top) on 35 cellular markers (see Supplemental Figure 3A). FlowSOM clustering identified 28 populations, which are visualized on the tSNE plot (bottom). Cluster proportions that were significantly different between the conventional and 42 kcal% fat diet are circled on the plots. (F) Myeloid cell clusters from FlowSOM in bone marrow sites of interest, displayed as a percentage of all CD45+ cells. (G) Lymphoid cell clusters from FlowSOM in bone marrow sites of interest, displayed as a percentage of all CD45+ cells. (H) Stem and progenitor cell clusters from FlowSOM in bone marrow sites of interest, displayed as percentage of all viable cells. Continuous data are expressed as the mean ± SD. Cluster abundances were analyzed by 2-way analysis of variance with Šidák's post hoc test. Statistically significant differences in uptake are indicated by: ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ns = not significant; tSNE = t-distributed stochastic neighbor embedding.

The MarrowMet findings highlight the femur, skull, and sternum as bone marrow sites of interest for further profiling. We therefore applied cytometry by time-of-flight with 45 antibodies targeting cellular, epigenetic, and metabolic markers. While GLUT1 expression on leukocytes remained unperturbed across sites (Figure 4C), CyTOF analysis revealed clear shifts in overall immune cell abundance (Figure 4D). The number of CD45+ cells significantly increased by almost 40% in the sternum and was slightly decreased in the skull, corroborating MarrowMet's changes in ¹⁸F-FDG uptake. Overall, our data implicate the sternum as the predominant site orchestrating atherogenic inflammation.

To untangle cellular shifts within CD45+ leukocytes, we applied unsupervised t-distributed stochastic neighbor embedding (tSNE) and FlowSOM clustering to the CyTOF data (35 markers, Supplemental Figure 5A). tSNE visualization mapped 28 FlowSOM clusters, several of which exhibited statistically robust shifts during atherogenesis (Figure 4E, Supplemental Figure 5A). Within the leukocyte pool, we detected increases in the proportion of myeloid cells—at the expense of lymphocytes—among the femur, skull, and sternum. The inflammatory landscape featured marked expansion of myeloid subsets, especially PD1+ macrophages/monocytes, Ly6C^hi inflammatory monocytes, and neutrophils, which was most pronounced in the skull and sternum (Figure 4F). These findings are in accordance with clinical evidence linking monocytosis, neutrophilia, and metabolic reprogramming to atherosclerosis progression and disease severity.^3,32 Contrastingly, expansion of proatherogenic myeloid cells was associated with a loss of lymphocytes, including B cells, a high cytotoxic T-lymphocyte associated protein 4-expressing subset, and FoxP3+ and glucocorticoid-induced TNFR-related protein (GITR) + T cells (Figure 4G), which are all mediators of immune tolerance and atheroprotection.33, 34, 35, 36, 37, 38, 39 Beyond mature cells, CyTOF analysis of the stem cell pool revealed a significant reduction in common lymphoid progenitors in the skull (Figure 4H), further underlining the shift towards myelopoiesis.

Acute ischemic events trigger localized bone marrow immune activation

MI and stroke remain unrivaled as the leading contributors to global mortality.^40,41 Mapping of site-specific bone marrow immune responses in atherosclerosis warrants further investigations into these clinical consequences. Whereas atherosclerosis is accompanied by chronic, low-grade inflammation, ischemic events result in acute inflammation and emergency hematopoiesis, in turn fueling atherosclerosis progression and ischemic recurrence.42, 43, 44 Is spatially diverse activation of bone marrow regions relevant in the context of MI and stroke?

To study the role of the bone marrow following MI or stroke, we occluded the LAD coronary artery or middle cerebral artery for 30 minutes, followed by reperfusion (Figure 5A). Two days postocclusion, MarrowMet uncovered robust bone marrow activation in all 11 measured bones after MI (Figure 5B), while stroke triggered heterogeneous bone marrow responses (Figure 5C). Most striking was the sternum: ¹⁸F-FDG uptake increased 80% post-MI, but paradoxically decreased 30% poststroke. The skull, together with the femur and C7 vertebra, emerged as a preferential site of bone marrow activation after stroke, further illuminating the asymmetric impact on diverse bone marrow sites. The preferential ¹⁸F-FDG uptake in the skull after stroke is of particular interest when considering that the skull preferentially contributes neutrophils to the ischemic brain,⁴⁵ likely reflecting the fast and predominant activation of marrow directly adjacent to the ischemic injury.

Figure 5.

Acute Ischemic Events Trigger Localized Bone Marrow Immune Activation

(A) Schematic depicting acute ischemia experiments: 12-week-old female C57BL/6J mice (n = 6) underwent left anterior descending artery (LAD)—myocardial infarction (MI)—or middle cerebral artery (MCA)—stroke—occlusion for 30 minutes, after which reperfusion was allowed. 48 hours after occlusion, mice were euthanized. (B) MarrowMet was used to quantify weight-corrected 18F-fluorodeoxyglucose (¹⁸F-FDG) uptake in different bones in control (gray) and MI mice (red), displayed in a bar graph and a Log₂ fold change heatmap, which is projected on the graphical representation of a mouse. All the Log₂ fold change values and P-values from MarrowMet can be found in the supplemental table. (C) MarrowMet was used to quantify weight-corrected ¹⁸F-FDG uptake in different bones in control (gray) and stroke mice (orange), displayed in a bar graph and a Log₂ fold change heatmap, which is projected on the graphical representation of a mouse. All the Log₂ fold change values and P-values from MarrowMet can be found in the supplemental table. (D) Cells analyzed with cytometry by time-of-flight (CyTOF) concatenated from femoral, skull, and sternal bone marrow were clustered with tSNE (top) on 35 cellular markers (see Supplemental Figure 7A). FlowSOM clustering identified 34 populations, which are visualized on the tSNE plot (bottom). Abundances of all FlowSOM clusters are displayed on a heatmap, separated by femoral (top), skull (middle), and sternal (bottom) bone marrow and ischemic event; control, MI, and stroke. Abundances that were significantly different (P < 0.05) between ischemic event and control are marked by a diamond (♦). Cluster abundances were analyzed by 2-way analysis of variance with Dunnett's post hoc test. Abundance values of each cluster can be found in Supplemental Figure 6E. (E) Heatmaps of the expression of the metabolic markers Glucose transporter 1 (GLUT1), G6PD, ASS1, Acetyl-Coenzyme A, and IDH2 in all significantly different cell clusters identified in (D). Each heatmap is separated by femoral (fe, left), skull (sk, middle), and sternal (st, right) bone marrow and ischemic event. (F) Heatmaps of the levels of NLRP3, DNTMT3A, H3K4me1, H3K27ac, and H3K27me3 in all significantly different cell clusters identified in (C). The levels of H3K4me1, H3K27ac, and H3K27me3 were corrected for total histone 3 levels in the respective sample and cluster. Continuous data are expressed as the mean ± SD. IDH2 = isocitrate dehydrogenase 2; tSNE = t-distributed stochastic neighbor embedding

Guided by MarrowMet's asymmetric marrow responses in MI and stroke, we performed a high-dimensional immune phenotyping strategy. Leveraging an established CyTOF panel, we systematically profiled femoral, skull, and sternal bone marrow from control, MI, and stroke cohorts, mirroring our approach in the atherosclerosis model. Consistent with expectations, GLUT1 was significantly upregulated in leukocytes at 48 hours postischemia, and absolute leukocyte counts correlated with MarrowMet's changes in ¹⁸F-FDG uptake (Supplemental Figures 6B and 6C). Using tSNE and FlowSOM, we resolved 34 clusters from our CyTOF data set (Figure 6D, Supplemental Figures 6D and 7A).

Figure 6.

Acute Respiratory Distress Syndrome Increases Bone Marrow Activation in Naive But Not Atherosclerotic Mice

(A) Schematic depicting lipopolysaccharide (LPS)-induced acute respiratory distress syndrome (ARDS) experiments: C57Bl/6 mice (n = 4-6) received intratracheal administration of LPS (dose = 5 mg/kg) and were euthanized 3 days or 4 weeks later. One group (n = 6) received 2 doses of LPS with 25 days between the 2 doses and was euthanized 3 days after the second dose. Another group of Apoe^−/− mice (n = 5) was put on a 42 kcal% fat diet for 8 weeks. Three days before the end of the diet period, they received intratracheal administration of LPS and were euthanized 3 days later. All mice were 12 weeks old at endpoint, and 18F-fluorodeoxyglucose (¹⁸F-FDG) was injected intravenously 1 hour before euthanasia. (B) Weight-corrected ¹⁸F-FDG uptake in the lungs. (C) MarrowMet was used to quantify weight-corrected ¹⁸F-FDG uptake in different bones, displayed in a bar graph (D). MarrowMet was used to quantify weight-corrected ¹⁸F-FDG uptake in different bones, displayed in a Log₂ fold change (relative to control) heatmap, which is projected on the graphical representation of a mouse. All the Log₂ fold change values and P-values (compared to control) from MarrowMet can be found in the supplemental table. Lung uptake was analyzed by 1-way analysis of variance with Šidák's post hoc test. Statistically significant differences in uptake are indicated by: ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ns = not significant.

Both myocardial and cerebral acute ischemic insults caused a significant drop in mature bone marrow neutrophils, likely reflecting their rapid egress and homing to injured tissue.45, 46, 47 Concomitantly, we observed an expansion of a more immature neutrophil subset (Neutrophil_1) and progenitor cells (lineage-, sca-1+, cKit+ (LSKs) and myeloid progenitors), indicative of augmented hematopoiesis. The release of neutrophils was accompanied by a surge in inflammatory monocytes, reflecting a key role for myeloid-driven inflammation in the early phase after ischemia.⁴⁸ While these shifts were apparent across the entire bone marrow landscape, a 30% reduction of NK cells emerged in the skull after stroke, but not MI. This suggests a unique role for NK cells in stroke, as they are recruited from adjacent bone marrow in the skull to subsequently infiltrate the infarct, where they play a dual role in tissue repair and exacerbation of neurovascular injury.⁴⁹ CyTOF analysis revealed more pronounced localized immune cell shifts, with mature B cells exhibiting a dramatic ∼50% increase in the sternum and skull after acute ischemia. In contrast, this cluster remained unaltered in the femur. These distinct spatial and temporal marrow immune profiles underscore the sophisticated, compartmentalized nature of bone marrow responses to acute cardiovascular injury.

While profiling immune cell populations provides valuable insight into bone marrow dynamics, it is essential to recognize that activation also manifests through metabolic and epigenetic modulation, 2 interconnected processes of cellular function and pathogenesis.⁵⁰ Metabolic and epigenetic rewiring collectively underpin long-term innate immune reprogramming (also termed “trained immunity”), which is implicated in many pathologies, including CVD.^3,51,52 To dissect these intricate networks, we leveraged targeted staining for pivotal metabolic effectors (GLUT1, G6PD, ASS1, acetyl-CoA, and IDH2), each representing a sentinel node within glucose transport, the oxidative pentose phosphate pathway, the urea cycle, and the tricarboxylic acid cycle. This integrative approach enabled resolving functional states within the bone marrow's metabolic landscape, with clustered heatmaps revealing that innate immune subsets are most affected by these metabolic perturbations (Figure 5E). Extending the analysis, we determined elevated inflammasome (NLRP3) expression alongside 4 key epigenetic markers intricately linked to innate immune responses and chronic inflammation. While these markers were less impacted by acute ischemic insult, the immune landscape shifts in the bone marrow trended towards mostly impacting myeloid cells (Figure 5F). Whereas cellular composition shifted markedly between the sites and ischemic insults, metabolic and epigenetic alterations seemed to occur without bias toward a specific bone marrow site, indicating a highly dynamic spatiotemporal response of the bone marrow in response to acute ischemic challenge.

Acute respiratory distress syndrome increases bone marrow activation in naive but not atherosclerotic mice

The inflammatory myeloid-biased bone marrow phenotype in atherosclerotic mice and the established relation between CVD and respiratory infections,53, 54, 55 prompted us to induce ARDS in naive and atherosclerotic mice (Figure 6A). As expected, we measured increased ¹⁸F-FDG uptake in the lungs 3 days after intratracheal LPS administration in naive mice (Figure 6B), reflecting increased immune cell recruitment and activation to the site of injury.^56,57 Similarly, MarrowMet uncovered increased bone marrow activation in mice with ARDS (Figures 6C and 6D). Interestingly, no long-term effects (4 weeks) on lung and bone marrow activation could be detected, and a second insult yielded similar inflammation as the initial injury (Figures 6B to 6D). Surprisingly, inducing ARDS in atherosclerotic mice did not result in elevated bone marrow activity (Figures 6C and 6D), suggesting that the low-grade bone marrow-driven inflammation during atherosclerosis blunts the ability to mount an acute inflammatory response. Such a compromised immune responsiveness is a possible explanation for the higher mortality rate in ARDS patients with CVD.^58,59

Discussion

To advance the field of immunometabolic and trained immunity research, we developed MarrowMet: a novel methodology leveraging the metabolic tracer ¹⁸F-FDG to quantify and visualize whole-body bone marrow activity. After establishing a strong correlation between bone marrow tracer uptake and immune cell activation, we confirmed the rigor of our method involving the quantification of the ¹⁸F-FDG in 11 distinct bone marrow sites. Harnessing MarrowMet, we mapped site-specific marrow activation patterns across diverse cardiovascular pathologies, including atherosclerosis, acute ischemic events, ARDS, MetS, and aging. We used MarrowMet to guide us to bone marrow regions of interest for in-depth CyTOF analyses. This integrative approach identified previously overlooked sites, most notably the skull and sternum, as critical loci of immune modulation, challenging the conventional paradigm of the femur being representative of systemic bone marrow activity. This heterogeneity between different regions might be caused by site-specific differences in vascularisation⁶⁰ or sympathetic innervation,^2,61 which play key roles in immune activation after infarction.^2,61 Although our efforts centered on CVD, the implications of these findings are decidedly broader. Bone marrow is a principal driver in a variety of conditions, including cancer,62, 63, 64 infection,^65,66 neurologic dysfunction,^67,68 and autoimmune disease.⁶⁹

In conclusion, MarrowMet is a method that enables quantitative profiling of bone marrow sites across the entire mouse body. MarrowMet provides an unprecedented approach to study metabolic shifts across bone marrow sites, which closely parallel inflammatory processes. Although primarily descriptive, our findings underscore the limitations of restricting analysis to femoral marrow in CVD models and potentially other sterile inflammatory or infectious conditions. We demonstrate that atherosclerosis and acute ischemic events asymmetrically affect distinct bone marrow regions that are proximal to the injured and inflamed tissues. These findings, collected across the entire hematopoietic system, indicate a distinct heterogeneity in the bone marrow response to CVD. The data pose new questions (eg, how this heterogeneity arises) and opportunities (eg, region-specific therapeutic intervention). Importantly, MarrowMet serves as an effective tool for pinpointing regions that merit deeper investigation, using immune landscape profiling methods such as CyTOF in the current study. By comparing cellular and metabolic shifts in health and disease, future studies can delineate spatially resolved zones where inflammation and hematopoietic activation converge or diverge. Collectively, our study establishes a foundation for systematic, whole-body interrogation of bone marrow heterogeneity. These insights will have broad relevance across immune-mediated diseases, from cancer and autoimmunity to neurodegeneration and infections, while providing a platform to uncover new immune pathways and advance therapeutic strategies.

Perspectives.

COMPETENCY IN MEDICAL KNOWLEDGE: Inflammatory cardiovascular disease engages the bone marrow heterogeneously rather than uniformly. Distinct skeletal sites—most notably the skull and sternum—display immune and metabolic activation that differs from the commonly sampled femur, informing the interpretation of clinical 18F-FDG PET bone marrow signals in patients with cardiovascular disease and its comorbidities.

TRANSLATIONAL OUTLOOK: Whole-body, site-resolved profiling of bone marrow activity may refine imaging biomarkers of cardiovascular inflammation and enable region-specific immunomodulatory strategies. Extending MarrowMet toward human imaging, and to other inflammation-driven conditions such as cancer, infection, and autoimmune disease, represents a key next step toward translating bone marrow heterogeneity into diagnostic and therapeutic advances.

Author Contributions

W.J.M.M. conceived the study and supervised the work together with T.J.B. and E.L. R.vd.M., C.M., M.G.N., N.P.R., M.D., M.N., F.K.S., L.A.B.J., Z.A.F., A.J.P.B., M.M.Tv.L., and C.P.M. provided guidance and support for study components and experiments. J.D. performed and coordinated experiments together with Yv.E. and I.V. J.D., Yv.E., I.V., Y.C.T., C.G.R., M.U., W.W., and S.N. performed in vitro and in vivo experiments. J.D., Yv.E., I.V., and J.E.C. performed CyTOF experiments. K.D. and A.J.B.P. set up ¹⁸F-FLT radiosynthesis. W.J.M.M., J.D., Yv.E., I.V., and T.J.B. wrote the manuscript and prepared figures. All authors read the manuscript and provided feedback.

Funding Support and Author Disclosures

Dr Mulder was supported by an NWO Vici grant (no. 91818622), ERC Advanced grant TOLERANCE (no. 101019807), Leducq Foundation Network CHECKPOINT ATHERO, and NIH grants P01A168258 and P01HL131478. Dr Meel was supported by a Dutch Research Council (NWO) Vidi grant (no. 19681) and ENW XL grant Heal-sepsis. This work was partially supported by the National Institutes of Health (HL176359, NS136068). NWO (The Hague, Netherlands); ERC (Brussels, Belgium); Leducq Foundation (Boston, MA, USA / Paris, France); NIH (Bethesda, MD, USA). Dr Netea and Dr Mulder are scientific cofounders of and have equity in Trained Therapeutix Discovery. Dr Mulder is CSO of Trained Therapeutix Discovery. Dr Netea and Dr Mulder are scientific cofounders of and have equity in BioTrip. Dr Nahrendorf has received funds or material research support from Alnylam, Amgen, Biotronik, CSL Behring, GlycoMimetics, GSK, Medtronic, Novartis and Pfizer, as well as consulting fees from Biogen, Bitterroot, Boehringer Ingelheim, Gimv, GSK, IFM Therapeutics, Molecular Imaging, Moonlake Immunotherapeutics, Novartis, Novo Nordisk, Pfizer, Sigilon, and Verseau Therapeutics. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.