Visualizing Vascular Bone Marrow Niche Alterations in Diabetes

Abstract

BACKGROUND:

Diabetes is characterized by chronic hyperglycemia that leads to systemic vascular complications. Hyperglycemia impairs endothelial function and promotes vascular inflammation, resulting in leukocytosis, altered hematopoiesis, and cardiovascular complications. Bone marrow endothelial cells play a pivotal role in regulating myeloid progenitor cells and leukocyte trafficking. However, the effects of diabetes on the structure and function of bone marrow vasculature remain poorly understood. To address this, we used a multiscale imaging approach integrating intravital microscopy, dynamic contrast-enhanced magnetic resonance imaging, and multispectral optoacoustic tomography to investigate diabetes-induced vascular changes in the bone marrow.

METHODS:

Diabetes was induced in C57BL/6J female mice using streptozotocin. Flow cytometry and histology characterized bone marrow myeloid progenitors, blood leukocytes, and bone marrow endothelial cell populations, as well as hypoxia. Intravital microscopy was used to visualize vascular density, permeability, and sprouting angiogenesis in the calvarial marrow. Dynamic contrast-enhanced magnetic resonance imaging quantified vascular density and permeability in the femoral marrow, while multispectral optoacoustic tomography assessed hemoglobin oxygenation in the calvarial marrow.

RESULTS:

Hyperglycemia significantly increased myelopoiesis, leading to elevated leukocytosis driving diabetic inflammation. Flow cytometry and histology revealed increased bone marrow endothelial cell numbers (P=0.0006), while intravital microscopy showed elevated vascular permeability (P=0.0095) and sprouting angiogenesis (P=0.0095). Dynamic contrast-enhanced magnetic resonance imaging confirmed greater vascular density (P=0.019) and leakiness (P=0.0062), and multispectral optoacoustic tomography detected reduced hemoglobin oxygenation and increased hypoxia in the diabetic marrow (P=0.0065), reflecting a hypoxic niche favorable to hematopoietic stem and progenitor cells. This likely drives angiogenesis and contributes to inflammatory hematopoiesis in diabetes.

CONCLUSIONS:

This study demonstrates that diabetes induces profound vascular remodeling and hypoxia in the bone marrow, reshaping the hematopoietic niche and driving myelopoiesis and leukocytosis. By validating dynamic contrast-enhanced magnetic resonance imaging and multispectral optoacoustic tomography as noninvasive translational tools, coupled with intravital microscopy, we provide a comprehensive framework for exploring novel therapies targeting bone marrow vasculature to mitigate inflammation-driven outcomes in diabetes.

Highlights.

• Diabetes induces extensive vascular remodeling in the bone marrow, marked by increased endothelial cell density, heightened vascular permeability, and sprouting angiogenesis, as revealed by intravital microscopy, dynamic contrast-enhanced magnetic resonance imaging, and multispectral optoacoustic tomography.

• Diabetic bone marrow exhibits disrupted vascular integrity and neovascularization, facilitating increased leukocyte trafficking and the influx of proinflammatory signals into the hematopoietic niche.

• Multispectral optoacoustic tomography imaging noninvasively detected reduced hemoglobin oxygenation in the diabetic bone marrow, a finding corroborated by pimonidazole staining, establishing a clear link between hyperglycemia and marrow hypoxia.

• Vascular dysfunction in diabetes drives emergency myelopoiesis and leukocytosis, highlighting a key mechanism underlying systemic inflammation and immune dysregulation.

• This study validates dynamic contrast-enhanced magnetic resonance imaging and multispectral optoacoustic tomography as powerful, noninvasive modalities for assessing bone marrow vascular pathology, with translational potential for monitoring diabetes-associated inflammatory complications.

Diabetes is a chronic metabolic disorder characterized by the body’s inability to regulate blood sugar levels. This condition arises from either insulin resistance in peripheral tissues or impaired insulin production due to pancreatic β-cell dysfunction. Type 1 diabetes, characterized by autoimmune destruction of pancreatic β-cells,¹ and type 2 diabetes, driven by insulin resistance and eventual β-cell failure,² both contribute to systemic vascular complications.^3,4 These vascular complications contribute to an increased risk of cardiovascular disease, affecting major organs such as the heart, kidneys, eyes, and nerves.^3,4 In research, the streptozotocin (STZ)-induced diabetic mouse model is commonly used to mimic type 1 diabetes, as STZ selectively destroys pancreatic β-cells, enabling the study of hyperglycemia’s effects on endothelial function and systemic vascular health.^3,4 Hyperglycemia-induced vascular damage underlies many of the long-term complications of diabetes, significantly increasing the risk of developing or worsening cardiovascular disorders.^5,6

Chronic hyperglycemia drives heightened vascular inflammation, involving key mediators such as reactive oxygen species and a diverse range of cells, including endothelial cells and myeloid cells like neutrophils, monocytes, and macrophages.^2,3,7,8 Hyperglycemia-induced endothelial dysfunction compromises vascular barrier integrity, reduces oxygenation, and promotes vascular remodeling, including angiogenesis and hypoxic cell death.³ In addition to impairing endothelial function,³ diabetes disrupts the production of inflammatory myeloid cells,^2,8 leading to elevated leukocytosis—a hallmark of diabetes. This process is driven by enhanced production of myeloid cells from common myeloid progenitors (CMPs),^9,10 resulting in increased levels of circulating myeloid-derived white blood cells, which are strongly linked to a higher risk of coronary artery disease.¹¹ Myeloid cells originate from hematopoietic stem and progenitor cells (HSPCs) within the bone marrow niche, which resides in close proximity to the microvascular endothelium.¹² Bone marrow endothelial cells play a pivotal role in regulating hematopoiesis through secreted factors and cell-surface signaling, directly influencing HSPC activity.^13,14 Diabetes is well known to drive systemic structural and functional vascular alterations,^2,3,7,8 and our previous work demonstrated that the vascular architecture of the bone marrow changes in response to acute inflammation and cardiovascular disease, in conditions such as hypertension, atherosclerosis, and ischemia.^15–17 While alterations in the bone marrow endothelial cell signaling have been observed,^10,18 the impact of diabetes on the structure and function of the bone marrow vasculature—a key regulator of HSPC activity and leukocyte trafficking—remains insufficiently understood. To address this knowledge gap, we aim to explore the hypothesis that diabetes drives vascular changes within the bone marrow, thereby affecting hematopoiesis and immune cell trafficking.

Imaging the bone marrow poses significant challenges due to light scattering within the bone,¹⁹ which limits the effectiveness of optical imaging techniques and hinders broader imaging applications. While advanced cryo-sectioning methods address this limitation for ex vivo studies,¹⁹ time-lapse intravital microscopy (IVM) is restricted to the calvarial bone marrow. Despite its capacity to visualize dynamic endothelial processes such as vascular leakiness,^14–16 its depth penetration remains limited. In contrast, noninvasive magnetic resonance imaging (MRI) overcomes these depth-related constraints, offering high spatial resolution and comprehensive anatomic insights across larger anatomic regions. When combined with gadolinium-based contrast agents, MRI allows for the visualization of vascular dynamics, including vascular density and permeability, enabling the assessment of marrow vascular function on an organ-wide scale.^15,16 Additionally, multispectral optoacoustic tomography (MSOT) provides a unique capability to visualize vascular oxygenation—a critical parameter that remains challenging to assess with other imaging modalities. MSOT’s noninvasive nature and clinical potential make it particularly valuable for studying bone marrow vasculature in vivo.^19,20 By integrating IVM, MRI, and MSOT in a multiscale imaging approach, we aim to overcome the individual limitations of each modality, thereby enhancing our understanding of disease-relevant changes in hematopoietic tissues. This integrated approach supports the development of novel cardiovascular immunomodulatory therapies and advances both preclinical research and clinical translation.

Diabetes is known to amplify myeloid cell production and significantly disrupt endothelial health throughout the body.^2,3,7,8,10 However, the interplay between the bone marrow niche endothelium and HSPC activity, and its subsequent impact on systemic leukocyte levels, remains poorly understood. To investigate this relationship, we used a multiscale imaging approach to examine how STZ-induced hyperglycemia alters bone marrow vasculature. This comprehensive strategy integrated IVM for high-resolution imaging of endothelial function in the calvaria, dynamic contrast-enhanced (DCE) MRI to assess vascular density and permeability in the femur, and MSOT to measure hemoglobin oxygenation in the calvarial marrow. By leveraging the unique strengths of these complementary imaging modalities, we provide a detailed framework for understanding the effects of hyperglycemia on bone marrow vascular structure and function. These findings offer critical insights into the role of diabetes-induced vascular remodeling in promoting systemic vascular inflammation and highlight potential targets for therapeutic intervention.

Materials and Methods

Data Availability

The main data supporting the results in this study are available within this article. The raw and analyzed data are available for research purposes from the corresponding author on request.

Mice

Wild-type C57BL/6J mice were purchased from Envigo (Israel), Apln^tm1^{.1(Cre/ERT2)Bzsh} (Apln^CreER) mice were generously provided by Dr Ralph Adams and Dr Bin Zhou,²⁰ and B6.Cg-Gt(ROSA)26Sort^m6^{(CAG-ZsGreen1)Hze/J} (Rosa26^ZsGreen) mice were purchased from The Jackson Laboratory. Wild-type C57BL/6J and Apln^CreER;Rosa26^ZsGreen/+ female mice (8–12 weeks old) were randomly assigned to control or STZ-induced diabetes (STZ) groups using simple randomization, ensuring equal probability of group allocation. A total of 73 female mice were used across all experiments: 19 mice for Figure 1 (flow cytometry of blood and bone marrow), 14 mice for Figure 2 (flow cytometry and immunostaining of endothelial cells), and 12 mice for Figure 3 (IVM of vascular permeability and angiogenesis); this included 10 mice from the intravital +2 extra mice only for ex vivo analysis for Figure 4 (ex vivo confocal imaging of femoral vasculature), 13 mice for Figure 5 (DCE-MRI of femoral vascular density and permeability), and 15 mice for Figure 6 (MSOT imaging and pimonidazole-based hypoxia staining), with group sizes and quantification parameters detailed in the respective figure legends. Body weight and blood glucose measurements shown in Figure 1 were collected longitudinally from the same animals included in the above experimental groups, up to the point at which significant differences were observed. All experiments in this study were conducted using female mice to reduce biological variability and ensure consistent interpretation of vascular and hematopoietic outcomes, given known sex-related differences in immune and metabolic responses. Diabetes was induced via intraperitoneal injection of STZ at 50 mg/kg body weight for 5 consecutive days to establish a hyperglycemic state. For the activation of the genetic labeling system in Apln^CreER;Rosa26^ZsGreen/+ mice, 75 mg/kg of tamoxifen dissolved in corn oil (Sigma-Aldrich) was administered intraperitoneally for 3 consecutive days, allowing for the specific green fluorescent labeling of sprouting angiogenic ECs.²⁰ Mice received isoflurane (3%–4% induction; 1%–2% maintenance) unless stated otherwise. All animal procedures were reviewed and approved by the Technion Institutional Animal Care and Use Committee (protocol IL-131-08-21) and were conducted in full compliance with institutional guidelines and national ethical regulations for animal research. A completed ARRIVE (Animal Research: Reporting of In Vivo Experiments) checklist summarizing the animal experiment design, reporting, and ethical considerations is summarized in the Supplemental Material.

Figure 1.

Enhanced myeloid cell production in streptozotocin (STZ)-treated diabetic mice. A, Timeline for diabetes induction by injecting STZ for 5 consequent days (50 mg/kg). B, Glucose level changes between before and 4 weeks after diabetes induction in control (CTR; n=9) and STZ-induced (n=9) female mice. C, Gating strategy for blood leukocytes, with quantification of (D) monocytes and (E) neutrophils in CTR (n=8) and STZ (n=11) female mice. F, Representative gating for bone marrow stem and progenitor cells: focusing on common myeloid progenitors (CMPs; Lin⁻c-Kit⁺Sca-1⁻CD16/32^intCD34⁺) and granulocyte-macrophage progenitors (GMPs; Lin⁻c-Kit⁺Sca-1⁻CD16/32⁺CD34⁺), with quantification of (G) GMPs and (H) CMPs in CTR (n=7) and STZ (n=8) female mice. Each dot represents an individual mouse; all data from 3 independent experiments are displayed as mean±SEM; Mann-Whitney U statistical test. CD indicates cluster of differentiation; LK, lineage-negative, c-Kit-positive cells; LSK, lineage-negative, Sca-1-positive, c-Kit-positive cells; Ly6G, lymphocyte antigen 6 complex; Sca-1, stem cell antigen-1; and SSC, side scatter.

Figure 2.

Streptozotocin (STZ)-induced diabetes triggers endothelial cell expansion in femoral bone marrow. A, Experimental setup. B, Flow cytometric quantification of endothelial cells in control (CTR; n=7) and STZ-induced diabetic (n=7) female mice. C, Representative flow cytometric gating for bone marrow endothelial cell population of CTR and STZ-induced diabetic female mice. D, Representative confocal microscopy images showing EMCN (endomucin) expression detected by immunofluorescence in the femoral metaphysis of CTR and diabetic STZ-injected female mice. E, Quantification of EMCN-positive area in the femoral metaphysis of CTR (n=5) and diabetic STZ-injected (n=4) female mice. Each dot represents an individual mouse; Mann-Whitney U statistical test. CD indicates cluster of differentiation; and Sca-1, stem cell antigen-1.

Figure 3.

Intravital microscopy (IVM) reveals enhanced permeability, increased vascular density, and neovascularization in the diabetic bone marrow. A, Experimental setup of IVM of tamoxifen-treated Apln^CreER;Rosa26^ZsGreen/+ of control (CTR; n=6) and streptozotocin (STZ)-injected (n=4) diabetic female mice using a dorsal skin flap. In Apln^CreER;Rosa26^ZsGreen/+ reporter female mice, sprouting endothelial cells and their progeny express ZsGreen when Apln (apelin) is expressed (Apln ZsGreen). B, In vivo IVM imaging of the calvaria following albumin-Cy5 leakage administered intravenously at time point 0 and followed over time. C, Representative time-lapse IVM quantification of albumin-Cy5 leakage over time. D, Albumin-Cy5 leakage, CD31⁺ vessels, and sprouting ZsGreen⁺ endothelial cells inside bone marrow niches of the calvarium, with images at t=2 minutes after injection of fluorescently labeled albumin-Cy5 displaying vessel permeability in calvaria of CTR and diabetic STZ mice. E and F, Quantification of the percent area of (E) albumin-Cy5, (F) anti-CD31, and (G) sprouting blood vessels (Apln Zsgreen) in CTR (n=6) and diabetic STZ female mice (n=4). H, Angiotool quantification of the junction density in CTR mice (n=17 fields of view [FOVs] in n=6 mice) and diabetic STZ mice (n=13 FOVs in n=4 mice) and (I) representative images of Angiotool analysis showing the difference in the junction density of sprouting vessels in both groups. Each dot represents an individual mouse, except for H, which represents FOVs; Mann-Whitney U statistical test. CD indicates cluster of differentiation; and Cy5, Cyanine5.

Figure 4.

Ex vivo confocal microscopy highlights increased permeability and neovascularization in the diabetic bone marrow. A, Representative whole mount fluorescence microscopy of the femur focusing on the metaphysis (dotted square). B, Ex vivo confocal microscopy of albumin-Cy5⁺, endothelial cells, and sprouting ZsGreen in femoral metaphysis of control (CTR) and diabetic (STZ) female mice. C through E, Quantification of the percent area of (C) albumin-Cy5, (D) CD31-PE antibody, and (E) sprouting blood vessels (Apln [apelin] Zsgreen) in CTR (n=6) and diabetic STZ female mice (n=6). Each dot represents an individual mouse; Mann-Whitney U statistical test.

Figure 5.

Albumin-based dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) shows that diabetes enhances bone marrow endothelial density and leakiness on an organ level. A, Femoral parametric maps of fractional blood volume (fBV) and permeability×surface area product (PS) in control (CTR) and diabetic (streptozotocin [STZ]) female mice. Regions of interest used for quantification in the distal metaphysis are indicated by orange dotted lines and arrows. B, Mean albumin-GdDTPA (gadolinium-diethylenetriamine pentaacetic acid) accumulation in the consecutive DCE-MRI images for fBV (intercept at time 0) and PS (=slope) in CTR (n=8) and diabetic (STZ; n=5) female mice. C and D, Quantification of fBV and PS values for both groups. Each dot represents an individual mouse; Mann-Whitney U statistical test.

Figure 6.

In vivo multispectral optoacoustic tomography (MSOT) imaging and ex vivo confocal microscopy reveal hypoxia in the hematopoietic bone marrow of streptozotocin (STZ)-injected diabetic mice. A, Representative vascular images and Hbo₂ maps for control (CTR) and diabetic (STZ) female mice. B, Noninvasive MSOT imaging was performed on shaved skin with air inhalation to compare Hbo₂ levels between healthy control and STZ-injected diabetic mice. C, Quantification of Hbo₂ values in the calvarial bone marrow of CTR (n=9) and STZ-injected (n=6) mice, obtained via MSOT imaging. Real-time data acquisition parameters: 530- to 584-nm wavelength sweep at 2 nm intervals, sound speed of 1480 m/s, field of view (FOV) of 10×10 mm, and frequency cutoff of 0.1 to 6.0 MHz. D, Representative maximum intensity projection (MIP) images showing pimonidazole staining in bone marrow sections, indicating hypoxia in CTR and STZ-injected female mice. E, Quantitative analysis of pimonidazole-positive areas in the femoral metaphysis (n=3 slices per mouse; analyzed FOV, 250×250 µm) in CTR (n=6) and STZ-injected (n=6) mice. F, Correlation between EMCN (endomucin)-positive area (from Figure 2) and pimonidazole-stained hypoxic areas in the bone marrow in CTR (n=5) and STZ-injected (n=4) female mice. Statistical analysis was performed using the Mann-Whitney U statistical test and Spearman r for correlation; each dot represents 1 mouse (5 regions of interest averaged per mouse). DAPI indicates 4′,6-diamino-2-phenylindole.

Flow Cytometry

For peripheral blood leukocytes, blood was collected from the heart into EDTA tubes using a needle inserted into the left ventricle, specifically targeting the apex of the heart. The pellet was resuspended in lysis buffer to lyse red blood cells. To stain for blood leukocytes, cells were incubated with specific surface antibodies on ice (BV711 CD45 [cluster of differentiation; clone 30-F11], APC/Cy7 CD19 [clone 6D5], BV220 RA3-6B2, APC/Cy7 NK1.1 [clone PK136], APC CD11b [clone M1/70], PE [phycoerythrin] CD3 [clone 17A2], FITC [fluorescein isothiocyanate] Ly-6G [clone 1A8], and BV421 CD115 [clone AFS98]). Then cells were washed and fixed in 1% paraformaldehyde. The samples were stored overnight at 4 °C for subsequent analysis using LSR Fortessa with the FACS (Fluorescence-Activated Cell Sorting) Diva software (BD), and data were analyzed with the FlowJo 10 software (Tree Star). To stain HSPCs, murine femora were extracted and flushed with sterile PBS, followed by filtration through 40 µm filters. The cells were then centrifuged at 340g for 7 minutes at 4 °C, and the supernatant was discarded. Subsequently, the cells were incubated with biotin-conjugated lineage markers (listed below) for 30 minutes at 4 °C. After incubation, the samples were washed with 1 mL of FACS buffer and centrifuged. The cells were then stained with antibodies targeting c-Kit (clone 2B8), Sca-1 (stem cell antigen-1; clone D7), CD16/32 (clone 2.4G2), CD34 (clone RAM34), CD115 (clone AFS98), CD150 (clone 9D1), and CD48 (clone HM48-1). For bone marrow endothelial cells, bone marrow cells were isolated by flushing the femurs with PBS and digested with 1 mg·mL⁻¹ collagenase IV (Sigma) and 2 mg·mL⁻¹ dispase II (Gibco) in DMEM without phenol red (Sigma) at 37 °C for 30 minutes. Red blood cells were lysed with red blood cell lysis buffer. To stain for endothelial cells with brilliant violet 605, Ter119/erythroid (clone TER119; Biolegend), PerCP (peridinin-chlorophyll-protein complex)/Cy5.5 (Cyanine dye) anti-mouse CD41(clone MWReg30; Biolegend), anti-mouse CD45.2-super bright 436 (clone 104; Invitrogen), Alexa Flour 488 anti-mouse CD31 (clone MEC13.3; Biolegend), and PE/Cy7 anti-mouse Ly-6A/E (Sca-1; clone D7; Biolegend) were used in this study. Debris and dead cells were excluded by FSC (forward scatter), SSC (side scatter), and DAPI (4′,6-diamino-2-phenylindole; fluorescent stain for DNA; FxCycle Violet Stain; Life Technologies) staining profiles. Samples were recorded on the LSRII flow cytometer equipped with the FACS Diva 6.1 software (BD), and data were analyzed with the FlowJo 10 software (Tree Star). A series of gating steps were performed to identify bone marrow Ter119⁻CD45⁻CD31^high endothelial cells. The number of cells per femur was measured as a percentage of live cells identified within the FACS gate.

Confocal Microscopy

Intravital Confocal Microscopy

IVM of calvaria (skull marrow) was performed using a confocal LSM 880 attached to an upright Axio Examiner.Z1 microscope (Zeiss, Jena, Germany). Each mouse underwent imaging of 3 to 5 distinct regions of interest in the calvaria. First, mice were shaved at the skull and positioned in a stereotaxic skull holder. Next, a skin incision at the scalp was performed to reveal the calvarial bone marrow niches, and PBS was applied to prevent the tissue from drying before imaging. To outline the vasculature, we used PE-conjugated 30 μL anti-CD31 (clone 390; Biolegend, San Diego, CA). The microscopy was performed in 3 channels: 633 nm excitation for BSA-Cy5 (2.5 mg per mouse in 50 μL), 561 nm for anti-CD31 PE, and 488 nm for ZsGreen (green fluorescent protein used in genetic labeling) expression with a water dipping objective (20×; numerical aperture [NA], 1.0; working distance, 1.8 mm) covering a field of view of 425.10×425.10 µm, at 512×512 pixels. Z-stack images were acquired at 2 µm steps. Time-lapse imaging was performed in the 3 channels with a time resolution of 1.89 frames per second.

Ex Vivo Confocal Microscopy Imaging

After the IVM session, the animals were euthanized, and femurs were dissected and immersed in 4% paraformaldehyde for 12 to 18 hours and for 6 hours in a 15% sucrose solution and finally embedded in optimal cutting temperature compound and frozen in −80°. Femurs were shaved 300 µm with a cryostat at −25 °C to expose the bone marrow. These long bones were then thawed stained with DAPI and imaged by spinning disk confocal microscope (CSU-W1; Nikon) 10× (dry; NA, 0.45; working distance, 4 mm) and 20× (dry; NA, 0.70; working distance, 2.30 mm) equipped with 4 solid lasers, 405 nm for DAPI, 488 nm for ZsGreen, 561 nm for CD31, and 640 nm for albumin Cy5.

Confocal Image Processing

All confocal images were analyzed and processed using Fiji. Maximal intensity projections of Z stacks were reconstructed. The signal intensity derived from anti-CD31-PE, albumin-Cy5, and Apln (apelin)-ZsGreen maximal intensity projections of 30 slices was used to generate a thresholded image of the bone marrow vasculature and quantify, using the tracer, the percent area occupied in the field of view. For analysis of fractional blood volume, the first frame after albumin injection represented the amount of functional blood vessels, or microvascular density, in the bone marrow niche. The extravasated albumin-Cy5 at t=3 minutes post-injection showed vessel permeability×surface area product (PS) in control and STZ-treated animals. To assess vascular leakiness, we generated thresholded images of the bone marrow vasculature and compared the signal intensity derived from the first and last images after albumin-Cy5 injection.

MRI Acquisition

In vivo macromolecular DCE-MRI was performed on a horizontal bore 9.4T MRI system (Bruker Biospec, Ettlingen, Germany), using a cylindrical volume coil (86 mm inner diameter) for radiofrequency transmission and a surface coil (20 mm diameter) for detection. Mice were anesthetized (induction, 3%; maintenance, 1.25% in oxygen [0.7 L/min]) and placed in a custom-designed mouse bed compatible with MRI positioned on the right side while warmed by using circulating hot water and monitored with respiratory gating (SA Instruments). Following a series of variable flip angle precontrast T1-weighted fast low-angle shot sequences, a bolus of albumin-GdDTPA (gadolinium-diethylenetriamine pentaacetic acid; 10 mg/mouse; relaxivity, 71.8 mmol/L per s; Symo-Chem, the Netherlands) was injected intravenously through an indwelling tail vein, and sequential imaging was conducted to measure the endogenous precontrast R1. The albumin-GdDTPA’s dynamics were tracked every 1.7 minutes for a duration of 8.5 minutes postcontrast administration using longitudinal 3-dimensional fast low-angle shot imaging across the entire femur. Imaging parameters were as follows: precontrast flip angles, 2°, 5°, 15°, 25°, 30°, and 40°; postcontrast flip angle, 25°; TR, 10.424 ms; TE, 4.64 ms; number of averages, 4; 25×25×5 mm; matrix size, 512×256×16 with zero filling in the phase-encoding direction to 512×512×16. The fast low-angle shot sequence was performed with radiofrequency and gradient spoiling enabled, using Bruker ParaVision’s default radiofrequency spoiling phase increment of 117°.

MRI Data Analysis

The method used for albumin-GdDTPA concentration mapping is based on variable flip angle T1-weighted imaging, which assumes that the signal is primarily governed by longitudinal relaxation (T1) and that transverse relaxation (T2*) and spin density (M₀) remain constant before and after contrast injection. This assumption is generally valid for macromolecular contrast agents like albumin-GdDTPA,⁸ which predominantly alter T1 without significantly affecting T2* at the low concentrations and short TE (4.64 ms) used in our study. T2 is not explicitly included in the signal equation because fast low-angle shot sequences are designed for T1-weighted imaging and minimize T2 contributions through short echo times and gradient spoiling. This approach has been validated in multiple vascular permeability studies.^3,7,9,10 Concentrations of albumin-GdDTPA were calculated on a pixel-by-pixel basis using the Matlab R2021a software (Mathworks, Natick, MA) for a single representative slice located in the metaphysis of the femur, selected based on anatomic landmarks of the distal femoral metaphysis and maximal marrow signal. This slice corresponded to the region of interest used consistently across all animals to ensure comparability as previously described.¹⁵ Mean femoral R1 values (R1_pre; R1=1/T1) precontrast and postcontrast were calculated by fitting variable flip angle (α) data to Equation 1 using a nonlinear best-fit approach:

$${\displaystyle \begin{array}{l}{\displaystyle \mathrm{I}=\frac{{\mathrm{M}}_0\mathrm{s}\mathrm{i}\mathrm{n}\alpha (1-{\mathrm{e}}^{-\mathrm{T}\mathrm{R}{\mathrm{R}}_{1\mathrm{p}\mathrm{r}\mathrm{e}}})}{1-\mathrm{c}\mathrm{o}\mathrm{s}\alpha {\mathrm{e}}^{-\mathrm{T}\mathrm{R}{\mathrm{R}}_{1\mathrm{p}\mathrm{r}\mathrm{e}}}}}\end{array}}$$ (1)

where I is the signal intensity as a function of the pulse flip angle α, TR is the repetition time (10.424 ms), and the preexponent term M0 includes the spin density and T2 relaxation effects, assumed to be unaffected by the contrast agent. Postcontrast R1 values were calculated from signal intensities pre- and post-contrast (I_pre and I_post; Equation 2):

$${\displaystyle \begin{array}{l}{\displaystyle \frac{{\mathrm{I}}_{\mathrm{p}\mathrm{r}\mathrm{e}}}{{\mathrm{I}}_{\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{t}}}=\frac{{\mathrm{M}}_0\mathrm{s}\mathrm{i}\mathrm{n}\alpha (1-{\mathrm{e}}^{-\mathrm{T}\mathrm{R}{\mathrm{R}}_{1\mathrm{p}\mathrm{r}\mathrm{e}}})/(1-\mathrm{c}\mathrm{o}\mathrm{s}\alpha {\mathrm{e}}^{-\mathrm{T}\mathrm{R}{\mathrm{R}}_{1\mathrm{p}\mathrm{r}\mathrm{e}}})}{{\mathrm{M}}_0\mathrm{s}\mathrm{i}\mathrm{n}\alpha (1-{\mathrm{e}}^{-\mathrm{T}\mathrm{R}{\mathrm{R}}_{1\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{t}}})(1-\mathrm{c}\mathrm{o}\mathrm{s}\alpha {\mathrm{e}}^{-\mathrm{T}\mathrm{R}{\mathrm{R}}_{1\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{t}}})}}\end{array}}$$ (2)

Last, concentrations of the contrast agent were derived from the measured relaxivity (R) of albumin-GdDTPA (R=71.8 mmol/L per s; Equation 3):

$${\displaystyle \begin{array}{l}{\displaystyle [\mathrm{a}\mathrm{l}\mathrm{b}\mathrm{u}\mathrm{m}\mathrm{i}\mathrm{n}-\mathrm{G}\mathrm{d}\mathrm{D}\mathrm{T}\mathrm{P}\mathrm{A}]=\frac{1}{\mathrm{R}}({\mathrm{R}}_{1\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{t}}-{\mathrm{R}}_{1\mathrm{p}\mathrm{r}\mathrm{e}})}\end{array}}$$ (3)

Fractional blood volume was determined by extrapolating the linear regression of normalized concentration data to the time of contrast administration. The slope of these normalized concentration values provided the rate of contrast accumulation, expressed as PS (min⁻1), obtained through linear regression over time. PS values measured with albumin−GdDTPA indicated albumin extravasation from the blood vessels and its subsequent accumulation in the tissue. Parametric fractional blood volume and PS maps of the femur were created to display mean values as previously described.¹⁵

Multispectral Optoacoustic Tomography

MSOT was performed using a customized tomographic system²¹ as detailed in the Supplemental Material. Optical illumination was provided by the SpitLight DPSS EVO 150 OPO laser (Innolas Laser GmbH), with optoacoustic signals captured by a spherical matrix array transducer (Imasonic SaS) comprising 512 piezoelectric elements. In vivo imaging was conducted on anesthetized mice secured in a stereotaxic device with continuous monitoring of physiological parameters. Imaging parameters included a 530- to 584-nm wavelength range (2 nm intervals) and lateral field of view of 10×10 mm. Custom Matlab scripts were used for image reconstruction and spectral unmixing of oxygenated (HbO₂) and deoxygenated (Hb) hemoglobin, enabling the generation of oxygenation maps to visualize changes in bone marrow oxygenation. Full technical details are provided in the Supplemental Material.

Ex Vivo Micro-Computed Tomography Imaging

For accurate mapping of calvaria, the mice skull specimens were collected and fixed in 4% paraformaldehyde overnight and moved to PBS to prevent dehydration. The first micro-computed tomography (CT) scan was performed (Skyscan 1276; Bruker, Kontich, Belgium) with 55 kV source voltage, 72 μA source current, applied 0.25 mm aluminum filter using a 0.2° rotation step, and 8 µm isotropic voxel size. All the resulting projection images were reconstructed using the NRecon software (v.1.7.4.5, Bruker microCT; Bruker) with postalignment, beam hardening corrections (41%), and a ring artifact reduction. Analysis was performed with CTAn Software (Skyscan, version 1.17.7.2) used for segmentation and for 3-dimensional visualization.

CT/MSOT Coregistration

Calvaria were imaged by both MSOT and micro-CT. Proper alignment of the skull was ensured using anatomic fiducials of the calvaria. Subsequently, MSOT images were manually aligned with the micro-CT images to establish the correspondence of hemoglobin signals from MSOT images with the calvarian bone marrow cavities. We followed an established method for noise threshold on MSOT images in the analysis.²²

Histology and Immunofluorescence Staining

For histological analysis, the hypoxia marker pimonidazole (HP2-200Kit; Hypoxyprobe, Inc) was administered via an intraperitoneal injection at 60 mg/kg, 2 hours before euthanasia (before MSOT imaging). Following this, femurs were transferred into 4% paraformaldehyde solution and embedded in paraffin, decalcified for 4 days using MoL (Museum of London) decalcifier (EDTA-based decalcifying solution; Milestone, Bergamo, Italy). Once soft and pliable, skulls were paraffinized. Longitudinal sections of the femur with 4-µm thickness were cut and mounted on slides. Following antigen retrieval, sections were stained overnight with antibodies against EMCN (endomucin; anti-goat; Novus Biologicals; AF4666) and ki67 (anti-rat; Abcam; ab15580) and separately with FITC-conjugated anti-pimonidazole (HP2-200Kit; Hypoxyprobe, Inc). Secondary antibodies from Thermo Fisher Scientific (donkey anti-rat IgG AF488, A21208; donkey anti-goat IgG AF680, A21084) were used to amplify the signal, and nuclei were counterstained with DAPI. Confocal imaging was performed using the CSU-W1 spinning disk confocal microscope (Intelligent Imaging Innovations, Denver, CO) equipped with a 50-µm disk, 10-position emission filter wheel, and near-infrared capability for 685- to 785-nm illumination. The system was integrated with the LaserStack v4 launch housing 4 solid-state lasers at 488 nm (150 mW), 561 nm (150 mW), 638 nm (200 mW), and 785 nm (300 mW). Our histological slides were scanned ×10 (dry; NA, 0.3; working distance, 2.60 mm) and ×20 (dry; NA, 0.5; working distance, 2.60 mm) magnification.

Statistics

Statistical analyses were performed using the GraphPad Prism software (GraphPad Prism 10). Results are reported as mean±SEM. Data were first tested for normality using the Shapiro-Wilk test. As no data sets met the assumptions of normality, nonparametric tests were applied throughout. Differences between groups were compared using 2-tailed Mann-Whitney U tests. Correlations between variables were examined using the nonparametric Spearman correlation test. A P value of <0.05 was considered statistically significant. All statistical tests and group sizes are detailed in the figure legends.

Results

Elevated Myelopoiesis Drives Diabetic Inflammation

In mice, type 1 diabetes can be modeled using STZ, an alkylating agent that selectively targets insulin-producing pancreatic β-cells. STZ administration induced significant hyperglycemia compared with healthy control female mice (Figure 1A and 1B), accompanied by weight loss, as previously reported (Figure S1). To assess the impact of diabetes on leukocyte populations, flow cytometry analysis of peripheral blood was performed. The gating strategy (Figure 1C) identified blood myeloid cells, including monocytes (NK1.1⁻B220⁻CD19⁻CD11b⁺CD115⁺Ly6G⁻), and neutrophils (NK1.1⁻B220⁻CD19⁻CD11b⁺CD115⁻Ly6G⁺). By 4 weeks post-STZ injection, diabetic female mice exhibited significantly elevated levels of circulating monocytes and neutrophils compared with controls (Figure 1C through 1E). Moreover, glucose levels positively correlated with the percentage of monocytes (Spearman r=0.52; P=0.049) and neutrophils (Spearman r=0.057; P=0.014; Figure S1B and S1C), suggesting a link between hyperglycemia and increased leukocyte production. To further investigate the source of this increase, we analyzed bone marrow myeloid progenitors using flow cytometry. The diabetic bone marrow exhibited significantly higher counts of granulocyte-macrophage progenitors (Lin⁻c-Kit⁺Sca-1⁻CD16/32⁺CD34⁺) and CMPs (Lin⁻c-Kit⁺Sca-1⁻CD16/32^intCD34⁺) compared with controls (Figure 1F through 1H). Correlation analysis revealed that glucose levels were associated with an increased percentage of granulocyte-macrophage progenitors (Spearman r=0.56; P=0.017) and CMPs (Spearman r=0.50; P=0.031; Figure S1D and S1E), further supporting the link between hyperglycemia and myeloid progenitor expansion. These findings demonstrate that diabetes-induced hyperglycemia promotes low-grade inflammation, which activates myelopoiesis in the bone marrow. This activation results in the overproduction of myeloid cells and elevated leukocyte trafficking, potentially driving systemic inflammation and immune dysfunction.

Diabetes-Induced Elevation of Bone Marrow Endothelial Cells

Systemic vascular alterations are a hallmark of diabetes, with well-documented changes in endothelial signaling across various tissues.^2,3,7,8 While endothelial remodeling has been observed in the bone marrow during acute inflammation and cardiovascular diseases,^10,15,16 the specific impact of diabetes on the structure and function of the bone marrow vasculature remains insufficiently characterized. Given the critical role of endothelial cells in regulating HSPC activity and leukocyte trafficking, we investigated the femoral bone marrow vasculature in diabetic mice. To quantify endothelial cell changes, we isolated bone marrow endothelial cells from the femurs of STZ-injected diabetic mice and analyzed them using flow cytometry. At 4 weeks post-STZ administration, diabetic mice displayed a significant increase in the number of endothelial cells compared with control female mice (Figure 2A through 2C). Correlation analysis revealed a strong association between glucose levels and the percentage of endothelial cells in the bone marrow (Spearman r=0.74; P=0.0052; Figure S1F), suggesting a direct link between hyperglycemia and endothelial expansion. We further examined endothelial remodeling at the structural level using histological analysis of the femur. By immunostaining for the sialoglycoprotein Endomucin, a marker of bone marrow endothelium, we visualized and quantified vascular changes. STZ-treated mice exhibited a marked increase in the number and density of EMCN⁺ blood vessels in the femoral bone marrow compared with controls (Figure 2D and 2E). These findings indicate that diabetes induces both cellular and structural changes in bone marrow vasculature, with hyperglycemia driving endothelial expansion and increased vascular density. Such changes may enhance the capacity for leukocyte trafficking and contribute to inflammatory hematopoiesis in diabetes.

Cellular Imaging Reveals Endothelial Leakiness and Vascular Sprouting in the Diabetic Bone Marrow

Diabetes induces significant adaptive changes in bone marrow vasculature, prompting us to investigate how these modifications impact vascular function and promote angiogenesis. To explore this, we assessed endothelial barrier integrity and angiogenic activity using fluorescently labeled albumin and Apln^CreER;Rosa26^ZsGreen/+ reporter mice, in which sprouting endothelial cells and their progeny express ZsGreen (Figure 3A). Concurrently, vessel lumens were visualized using a fluorescently labeled CD31 antibody. Intravital time-lapse imaging revealed a pronounced increase in albumin leakage from the vasculature of the diabetic bone marrow, indicating impaired barrier function (Figure 3B through 3E). This leakiness facilitates the trafficking of cells and blood-borne signals across the bone marrow endothelium, a process that likely contributes to HSPC activation via circulating alarmins and reactive oxygen species. Furthermore, IVM showed an increase in CD31⁺ vessel density in the calvarial marrow of diabetic mice (Figure 3D and 3F), alongside a rise in newly formed ZsGreen⁺ sprouting vessels (Figure 3D and 3G). Analysis of CD31⁺ vessels in the calvaria further revealed heightened vascular junction density, which may enhance intercellular interactions and facilitate leukocyte trafficking (Figure 3H and 3I). Consistent with these in vivo findings, ex vivo spinning disk confocal microscopy of femoral bone marrow confirmed similar vascular alterations (Figure 4A through 4E). Diabetic bone marrow displayed increased albumin leakage (Figure 4B and 4C), higher density of CD31⁺ vessels (Figure 4B and 4D), and more newly formed ZsGreen⁺ sprouting vessels (Figure 4B and 4E). These results highlight that vascular permeability and sprouting are not confined to the skull marrow but also extend to the femoral bone marrow, indicating a systemic vascular adaptation within the bone marrow in response to diabetes. Collectively, these findings suggest that diabetes induces remodeling of bone marrow vasculature, affecting multiple skeletal sites. This widespread response likely supports emergency leukocytosis and enhanced immune cell trafficking, thereby reshaping the hematopoietic niche and potentially driving inflammatory myelopoiesis.

Femoral MRI Reveals Increases in Endothelial Density and Leakiness in the Diabetic Bone Marrow

To complement cellular-level vascular imaging, we quantified vascular changes at a larger scale using DCE-MRI of the femur. This approach enables in vivo assessment of vascular alterations in the myelopoietic niche. Using a validated MRI protocol,¹⁵ we administered gadolinium-labeled albumin (albumin-GdDTPA) to control and diabetic female mice. DCE-MRI revealed a significant increase in vascular density, measured as fractional blood volume, in the diabetic femoral marrow compared with controls (Figure 5A through 5C). In addition to the increase in vascular density, DCE-MRI also demonstrated heightened permeability, quantified as the PS, in the femoral marrow of diabetic mice (Figure 5A, 5B, and 5E). This increase in permeability reflects enhanced vascular leakiness, consistent with impaired endothelial barrier integrity. These observations align with cellular-level findings from IVM, reinforcing that diabetes induces both structural and functional vascular changes within the bone marrow.

Diabetes Reduces Calvarial Bone Marrow Oxygenation, as Visualized by Noninvasive Optoacoustic Imaging

To evaluate the impact of hyperglycemia on bone marrow oxygenation, we used MSOT to compare healthy control with age-matched female mice subjected to STZ-induced hyperglycemia. Real-time in vivo MSOT imaging revealed significantly reduced hemoglobin oxygenation levels in the calvarial bone marrow of diabetic mice compared with controls (HbO₂ control, 52.05±2.10%; HbO₂ STZ, 39.89±2.61%; Figure 6A through 6C). Color-coded oxygenation maps, coregistered with CT scans of the skull, visually highlighted reduced oxygenation levels in the calvarial bone marrow of hyperglycemic mice (Figure 6A). These findings validate the capacity of MSOT to detect subtle oxygenation changes, emphasizing its potential as a powerful, noninvasive tool to assess metabolic alterations in the hematopoietic stem cell niche. To further explore these bone marrow oxygenation changes, we mapped hypoxia levels in the metaphyseal bone marrow of the femur ex vivo using pimonidazole staining, a hypoxia marker. Elevated hypoxia was confirmed in the femoral marrow of STZ-treated mice, as evidenced by increased pimonidazole-positive areas (Figure 6D and 6E). Moreover, a significant correlation was observed between the EMCN⁺ blood vessel area and pimonidazole-positive regions, suggesting a close link between angiogenesis and hypoxia in the bone marrow microenvironment (Figure 6F). Collectively, these results reveal that STZ-induced hyperglycemia leads to reduced oxygenation in the calvarial bone marrow and heightened hypoxia in the femoral bone marrow. This shift in oxygenation and hypoxia may drive angiogenic remodeling and metabolic reprogramming of the bone marrow niche, with implications for hematopoietic stem cell activity and myelopoiesis.

Discussion

Our study demonstrates that diabetes induces significant vascular remodeling in the bone marrow, leading to profound changes in the hematopoietic niche that promote myelopoiesis and systemic inflammation. Using a multiscale imaging approach that integrates IVM, DCE-MRI, and MSOT, we reveal that diabetes drives structural, functional, and metabolic changes in bone marrow vasculature. This comprehensive approach enabled us to identify critical shifts in vascular density, permeability, angiogenesis, and oxygenation, which collectively contribute to enhanced myeloid cell production and leukocytosis.

Our findings support the concept that hyperglycemia promotes hematopoietic activation and enhances myelopoiesis.^8,10,23 This is evidenced by elevated counts of circulating neutrophils and monocytes in diabetic mice, which strongly correlate with blood glucose levels. Flow cytometry revealed a significant increase in granulocyte-macrophage progenitors and CMPs in the bone marrow of diabetic mice, indicative of upstream activation of myeloid progenitor cells.¹⁰ This result aligns with previous studies showing that diabetes drives the proliferation of myeloid precursors through pathways involving S100A8/S100A9 signaling via RAGE (receptor for advanced glycation end products) and IL (interleukin)-1β released from adipose macrophages.^23,24 These signaling pathways converge to promote the expansion of granulocyte-macrophage progenitors and CMPs, thereby fueling leukocytosis.^8,10,23,24 Our findings further establish that diabetes-induced activation of myelopoiesis is sustained by an increase in bone marrow endothelial cell populations and vascular adaptations, which collectively enhance leukocyte trafficking from the bone marrow to peripheral tissues.

Endothelial cells play a pivotal role in regulating HSPC activity and leukocyte trafficking.^13,14 Our study shows that diabetes induces a marked increase in bone marrow endothelial cells, as confirmed by flow cytometry and immunohistological analysis of EMCN-positive vasculature. The observed increase in endothelial density may provide an expanded surface for leukocyte trafficking and vascular interactions. Consistent with our findings, previous studies have demonstrated that bone marrow endothelial cells regulate HSPC function through secreted factors and surface-bound ligands, such as SCF (stem cell factor) and CXCL12 (C-X-C motif chemokine ligand 12).^13,14 Notably, endothelial cell expansion has also been reported in other pathological states, including cardiovascular disease and acute inflammation, where it supports emergency myelopoiesis.^15,16

Angiogenesis is a hallmark of vascular remodeling, and our study demonstrates that diabetes induces sprouting angiogenesis in the bone marrow. Using Apln^CreER;Rosa26^ZsGreen/+ reporter mice,²⁰ we observed an increase in newly formed ZsGreen⁺ sprouting vessels in the bone marrow of diabetic mice. Previous studies in other vascular regions have linked hyperglycemia to the production of proangiogenic factors, which drive the formation of new blood vessels.^2,25–27 Bone marrow angiogenesis in diabetes may represent a compensatory response to microvascular damage caused by hyperglycemia-induced oxidative stress and inflammation.^25,28 This result suggests that hyperglycemia triggers neovascularization, possibly in response to hypoxia or oxidative stress.²⁹ While angiogenesis is typically associated with tissue repair, uncontrolled vascular expansion can promote chronic inflammation.³⁰ Analysis of CD31⁺ vessels further revealed increased vascular junction density, supporting the notion that diabetes alters vessel architecture, facilitating enhanced leukocyte trafficking and infiltration. Together, these structural changes likely contribute to the heightened immune response observed in diabetes.

The increased vascular leakiness observed in diabetic bone marrow may exacerbate systemic inflammation.^10,16 Using IVM and DCE-MRI, we identified a significant increase in vascular permeability in the calvarial and femoral bone marrow. The leakage of albumin-Cy5 from blood vessels, visualized in real-time with IVM, indicated a loss of endothelial barrier integrity. DCE-MRI further confirmed this observation by quantifying PS in the femoral bone marrow, revealing heightened permeability in diabetic female mice compared with controls. This increased permeability is consistent with endothelial barrier dysfunction seen in other vascular beds during diabetes.^31,32 Damage to tight junctions and the glycocalyx layer may underlie this dysfunction, as observed in endothelial cells exposed to hyperglycemia.¹⁵ The resulting loss of vascular integrity allows for the extravasation of blood-borne alarmins, cytokines, and immune cells into the bone marrow, potentially driving HSPC activation, myeloid progenitor proliferation, and leukocytosis.^14,16,33

A key limitation is the use of the STZ-induced diabetes model, which—while widely adopted for mechanistic studies—does not capture the full heterogeneity of human diabetes, including fluctuations in glycemic control and regulation through insulin therapy. Our model reflects a state of initial hyperglycemia, which allows the dissection of early pathological mechanisms underpinning diabetes-induced vascular bone marrow remodeling. While the inclusion of an insulin-treated group to explore the reversibility of these changes would be of significant interest, it falls outside the scope of the current study and will be considered in future investigations. Moreover, our study focused on a single time point—4 weeks after diabetes induction—selected based on prior studies showing consistent hyperglycemia and inflammatory activation at this stage.¹⁰ While this approach enables a clear snapshot of early vascular and hematopoietic remodeling, it does not capture the temporal dynamics of disease progression. It is plausible that both earlier and more prolonged durations of hyperglycemia could produce distinct effects on the bone marrow vasculature and hematopoietic output. Future studies incorporating multiple time points will be important to fully understand the evolution and potential reversibility of these changes over the course of diabetes. While our study centers on local changes in bone marrow vasculature, it is important to consider the role of systemic factors in diabetes. Elevated levels of circulating glucose, proinflammatory cytokines such as IL-6 and TNF-α (tumor necrosis factor-α), and altered lipid profiles may also drive endothelial activation and hematopoietic dysregulation.^11,16 These factors likely act in concert with local hypoxic cues to promote the vascular remodeling and inflammatory hematopoiesis observed in our model. Future studies incorporating metabolic profiling and cytokine analysis will help to delineate the relative contributions of systemic versus local drivers of bone marrow niche dysfunction in diabetes. Finally, whether similar vascular changes occur in the human bone marrow remains unknown and warrants further investigation.

A unique aspect of our study was the assessment of oxygenation in the bone marrow using MSOT. This noninvasive imaging technique enabled real-time monitoring of HbO₂ in calvarial bone marrow. Diabetic mice displayed significantly lower HbO₂ levels compared with controls, suggesting reduced oxygen availability in the hematopoietic niche. These observations were further validated by ex vivo pimonidazole staining, which revealed extensive hypoxia in the metaphyseal bone marrow of diabetic mice. The significant correlation between EMCN-positive vessel density and pimonidazole-positive hypoxic regions suggests a strong link between angiogenesis and hypoxia²⁹ in the diabetic bone marrow.

Hypoxia is known to play a key role in the regulation of HSPCs, promoting proliferation, and differentiation of hematopoietic cells niches.^15,34 Diabetic bone marrow exhibits a more hypoxic environment, which may drive HSPC activation, further enhancing myelopoiesis. Furthermore, hypoxia is a powerful stimulus for angiogenesis, mediated by HIF-1α (hypoxia-inducible factor 1α), which promotes the release of VEGF (vascular endothelial growth factor) to stimulate endothelial sprouting.³⁴ This feed-forward loop may explain the observed increase in sprouting angiogenesis and vascular density in the diabetic bone marrow.

Our study highlights the utility of DCE-MRI and MSOT as noninvasive tools for investigating vascular dynamics and oxygenation in the bone marrow. While IVM provides high-resolution insights into vascular leakiness and endothelial cell behavior, its application is limited to small animal models and confined to shallow regions, such as the calvarial marrow.^14–16 By incorporating DCE-MRI^14–16 and MSOT,^19,20 we bridge this gap, enabling quantification of vascular density, permeability, and oxygenation at an organ-wide scale. Together, these modalities enable in vivo imaging of molecular and structural processes across the hematopoietic marrow, addressing microscopy’s limitations while offering scalability for clinical applications. Although MSOT is highly sensitive to oxygenation changes, its resolution and depth are limited in the bone tissue due to light scattering.³⁵ However, the thin calvarial bone in mice allowed us to effectively image marrow oxygenation.²² Conversely, MRI, while providing unparalleled anatomic detail and spatial resolution, relies on contrast agents—in this study, albumin-GdDTPA—to visualize the vascular processes.³⁶ These imaging modalities have clear translational potential for monitoring vascular dysfunction in clinical settings. The ability to noninvasively measure marrow oxygenation (via MSOT) and vascular permeability (via DCE-MRI) offers new avenues for assessing hematopoietic stress, myelopoiesis, and inflammatory responses in patients with diabetes and cardiovascular disease.

While our findings point to vascular remodeling and hypoxia as key contributors to inflammatory hematopoiesis in diabetes, it is important to recognize that these processes also play vital roles in normal bone marrow homeostasis, immune surveillance, and tissue repair.^37,38 Endothelial sprouting, vascular permeability, and hypoxia-driven signaling (eg, HIF-1α) are central to stem cell maintenance, niche remodeling, and regeneration following injury.³⁴ Therefore, therapeutic strategies aimed at modulating these processes must be approached with caution. Broad inhibition of angiogenesis or vascular permeability could disrupt hematopoietic stem cell function, delay wound healing, or impair host defense. Future research should focus on identifying selective molecular targets or critical time windows that enable disease-specific modulation of vascular responses, both systemically and within the bone marrow niche.

Our study demonstrates that diabetes induces profound vascular remodeling and hypoxia in the bone marrow, reshaping the hematopoietic niche and promoting myelopoiesis and leukocytosis. Using a multiscale imaging approach, we identified key structural, functional, and metabolic changes in bone marrow vasculature, including endothelial expansion, enhanced permeability, angiogenesis, and hypoxia. These changes create a permissive environment for myeloid progenitor activation and heightened systemic inflammation. By validating DCE-MRI and MSOT as noninvasive imaging modalities, we present a novel framework for assessing diabetic vascular remodeling, with potential applications in preclinical research and clinical diagnostics. Targeting endothelial barrier function and oxygenation in the bone marrow may provide new therapeutic strategies to mitigate the proinflammatory effects of diabetes on hematopoiesis and systemic immunity.

Article Information

Acknowledgments

The authors thank the Levenberg Lab and the Rotenberg Lab for their support in initiating early-stage experiments. They are grateful to Amit Yaacobovich for assistance with MATLAB analysis and to Liron McLey, Ruth Pikovsky, Shirly Hagay, Daniel Razansky, and Amir Rosenthal for their guidance in the setup and analysis of multispectral optoacoustic tomography imaging. They also acknowledge the Biomedical Core Facility for technical support, and Yael Lupu-Haber and Yousef Mansour from the Lorry I. Lokey Interdisciplinary Center for their contributions. Special thanks to Amit Avrahami and Vita Zlobin from the Preclinical Research Authority at the Technion for their animal support and guidance. Some figures were created with BioRender.com. ChatGPT was used solely for language editing and clarity improvement, not for content generation or data interpretation.

Sources of Funding

This research was supported by the Technion Human Health Initiative, the Israel Science Foundation 446/21 and 660/21, and BIRAX Ageing–British Council 31BX21DSKV. A. Tiwari was supported, in part, by the Zeff Fellowship and by the Teva BioInnovator grant.

Disclosures

None.

Supplemental Material

Supplemental Materials & Methods

Figure S1

Major Resources Tables