Vascular and metabolic responses to elevated circulating PDGF-BB in mice: A multiparametric MRI study

Abstract

Elevated circulating platelet-derived growth factor-BB (PDGF-BB) has been implicated in various aged-related pathologies and, therefore, is recognized as a potential pro-aging factor. Despite extensive studies on the pathological characterizations centering PDGF-BB/PDGFRβ signaling alterations, there is limited research on the neurofunctional responses to elevated circulating PDGF-BB, primarily because in-vivo measurements are generally required for studying neurofunction. To address this knowledge gap, we characterized the vascular and metabolic responses to elevated circulating PDGF-BB in vivo using multiparametric non-invasive non-contrast MRI techniques in a conditional Pdgfb transgenic mouse model (Pdgfb^cTG) at 6 months of age. We found that Pdgfb^cTG mice exhibited decreased cerebral blood flow (P = 0.025), elevated oxygen extraction (P = 0.002), and increased metabolic rate of oxygen (P = 0.035), which replicated the changes observed in human aging. Compared to naturally aged mice, the rate of change in vascular and metabolic measurements was prominently higher (≥ 200.3%). Our study provides neurofunctional evidence that elevated circulating PDGF-BB accelerates neurovascular aging.

1. Introduction

Platelet-derived growth factor-BB (PDGF-BB) is increasingly recognized as a key factor in age-related pathologies and a potential driver of brain aging.^1–2 Normally, PDGF-BB promotes the proliferation and migration of cells during wound healing and tissue regeneration.³ The PDGF-BB/PDGFRβ signaling pathway mediates crosstalk between pericytes and endothelial cells, playing an important role in the establishment and maintenance of pericyte coverage on cerebral microvasculature.^4–5 Clinical data have shown elevated levels of soluble PDGFRβ, a marker of pericyte degeneration, in the cerebrospinal fluid of patients with mild cognitive impairment and Alzheimer’s disease (AD),^6–7 suggesting that dysregulation of the PDGF-BB/PDGFRβ signaling pathway is associated with disease pathologies.

Increases in serum/plasma PDGF-BB concentrations by 2–3 folds have been consistently observed in aged mice and humans compared to their young controls.^{1–2, 8} Skeletal TRAP⁺ preosteoclast-secreted PDGF-BB is a major contributor to the pathological elevation of circulating PDGF-BB, which mediates arterial stiffening¹ and calcification². Persistently high PDGF-BB levels induce pericyte loss in the brain by promoting enzymatic shedding of its receptor, PDGFRβ, from the pericyte membrane. This leads to reduced PDGF-BB/PDGFRβ signaling, and, consequently, blood-brain barrier (BBB) leakage.⁸ In addition to these vascular impairments, a recent human study found that circulating PDGF-BB levels were associated with white-matter dysfunction,⁹ a known risk factor for adverse neurovascular aging.

Understanding the dual role of PDGF-BB as both a necessary growth factor and a potential pro-aging agent is crucial for developing future therapies aimed at mitigating aging and halting age-related neurodegenerative diseases, such as AD or vascular dementia.^10–12 To further investigate the impact of elevated circulating PDGF-BB on brain neurofunction, this study aims to characterize the vascular and metabolic responses using a transgenic Pdgfb^cTG model^{8, 13} that overexpresses PDGF-BB. A multiparametric MRI study, combined with immunofluorescence analysis of brain tissue sections, was conducted to explore the neurofunctional profile of the Pdgfb^cTG mouse model.

2. Material and methods

2.1. General procedures

The experimental protocols for this study were approved by the Johns Hopkins Medical Institution Animal Care and Use Committee and conducted in accordance with the National Institutes of Health guidelines for the care and use of laboratory animals. Data reporting complied with the ARRIVE 2.0 guidelines. A cohort of 5 Pdgfb^cTG mice (age: 6 months; body weight: 29–35 g; males) and 5 littermate wild-type (WT) controls (age: 6 months; body weight: 26–34 g; males) was used in this study. No significant difference in body weight was observed between the WT and Pdgfb^cTG mice (unpaired t-test: degree of freedom = 8, t = 2.07, P = 0.072). All mice were housed in a quiet environment with a 12-h day/night cycle and had ad libitum access to food and water.

The experimenter performing the MRI scans was blinded to the group assignments. Mice were scanned in a randomized order, following a previously reported scheme.¹⁴ Briefly, mice were preassigned consecutive numbers starting from one. A set of pseudorandom numbers was generated using the MATLAB (MathWorks, Natick, MA, USA) “rand” function, and the ranks of these numbers (from largest to smallest) determined the experimental order. For instance, if the first pseudorandom number ranked third, the mouse preassigned to number 3 was the first to be scanned.

2.2. Multiparametric MRI experiments

All MRI experiments were performed on an 11.7T Bruker Biospec system (Bruker, Ettlingen, Germany) with a horizontal bore and an actively shielded pulse field gradient (maximum intensity of 0.74 T/m). Images were acquired using a 72-mm quadrature volume resonator as the transmitter, and a four-element (2×2) phased-array coil as the receiver. Magnetic field homogeneity over the mouse brain was optimized using global shimming (up to the second order) based on a pre-acquired subject-specific field map. Inhalational isoflurane, delivered with medical air (21% O₂ and 78% N₂) at a flow rate of 0.75 L/min, was used to minimize stress and motion in the mice. Anesthesia induction was achieved with 1.5–2.0% isoflurane for 15 minutes. At the 10^th minute under induction, the mouse was placed onto a water-heated animal bed with temperature control and positioned with a bite bar, a pair of ear pins, and a custom-built 3D-printed holder before entering the magnet. After induction, the isoflurane concentration was reduced to 1.0% for maintenance during the MRI scans.

Each mouse underwent a set of MRI scans as outlined in Figure 1, including fast spin echo (FSE) MRI to measure brain volume, phase-contrast (PC) for global cerebral blood flow (gCBF), pseudo-continuous arterial spin labeling (pCASL) for regional cerebral blood flow (rCBF), T₂-relaxation-under-spin-tagging (TRUST) for oxygen extraction fraction (OEF), and diffusion-weighted imaging (DWI) for apparent diffusion coefficient (ADC).^15–18 Cerebral metabolic rate of oxygen (CMRO₂) was calculated using the Fick principle.¹⁹

Figure 1.

Schematic diagram of MRI scans. A set of multiparametric MRI techniques was utilized, including fast spin echo (FSE) for assessing brain volume, phase-contrast (PC) for gCBF, pseudo-continuous arterial spin labeling (pCASL) for rCBF, T₂-relaxation-under-spin-tagging (TRUST) for OEF, and diffusion-weighted imaging (DWI) for ADC. CMRO₂ was calculated based on gCBF and OEF using the Fick principle.

The parameters for the T₂-weighted FSE MRI were as follows: repetition time (TR)/echo time (TE) = 4000/10.0 ms, field of view (FOV) = 15 mm [rostral-caudal] × 15 mm [left-right], matrix size = 128 × 128, slice thickness = 0.5 mm (without inter-slice gap), echo spacing = 5.0 ms (4 spin echoes per scan), 35 axial slices, and scan duration = 2.1 min.¹⁴

PC MRI covered the four major feeding arteries: left internal carotid artery (LICA), right internal carotid artery (RICA), left vertebral artery (LVA) and right vertebral artery (RVA), in separate scans to obtain corresponding through-plane velocity maps.¹⁶ Prior to the PC scans, a coronal time-of-flight (TOF) angiogram was performed (TR/TE = 45/2.6 ms, FOV = 25 mm × 16 mm, matrix size = 256 × 256, 9 slices without inter-slice gap, slice thickness = 0.5 mm, scan duration = 2.6 min) to visualize the feeding arteries. Then, a sagittal TOF (TR/TE = 60/2.5 ms, FOV = 16 mm × 16 mm, matrix size = 256 × 256, single slice tilted to cover the targeted artery identified from the coronal TOF images, slice thickness = 0.5 mm, scan duration = 0.4 min) was applied to visualize the in-plane trajectory of the targeted artery. Based on the reference TOF images (coronal and sagittal), the PC MRI was positioned and performed using the following parameters: TR/TE = 15/3.2 ms, FOV = 15 mm × 15 mm, matrix size = 300 × 300, slice thickness = 0.5 mm, number of average = 4, dummy scan = 8, receiver bandwidth = 100 kHz, flip angle = 25º, encoding velocity = 20 (for LICA/RICA)/10 (for LVA/RVA) cm/s, partial Fourier acquisition factor = 0.7, and scan duration = 0.6 min per artery.¹⁶

A two-scan pCASL method, which was designed to minimize the influence of magnetic-field inhomogeneity, was utilized.^{18, 20} First, a pre-scan was performed to optimize the phases of the labeling pulses in both the control and labeled scans. Then, the scan focusing on regional perfusion was performed with the following parameters:^21–22 TR/TE = 3000/13.1 ms, labeling duration = 1800 ms, FOV = 15 mm × 15 mm, matrix size = 96 × 96, slice thickness = 0.75 mm, labeling-pulse width = 0.4 ms, inter-labeling-pulse delay = 0.8 ms, flip angle of labeling pulse = 40°, post-labeling delay = 300 ms, two-segment spin-echo echo-planar imaging (EPI) acquisition, partial Fourier acquisition factor = 0.7, number of average = 25, and scan duration = 5.0 min.

TRUST MRI was performed on the confluence of sinuses using the following procedure.^{17, 23} To visualize the confluence of sagittal sinuses, an axial TOF scan was first performed under: TR/TE = 20/2.7 ms, FOV = 15 mm × 15 mm, matrix size = 256 × 256, 5 axial slices, slice thickness = 0.5 mm, and scan duration = 0.3 min. The TRUST scan followed the reported protocol:^{17, 24} TR/TE = 3500/6.5 ms, FOV = 16 mm × 16 mm, matrix size = 128 × 128, slice thickness = 0.5 mm, EPI factor = 16, inversion-slab thickness = 2.5 mm, post-labeling delay = 1000 ms, eTE = 0.3, 20, 40 ms, echo spacing of eTE = 5.0 ms, 3 repetitions, and scan duration = 8.4 min.

Key parameters of DWI MRI were: TR/TE = 2500/18.0 ms, FOV = 15 mm × 15 mm, matrix size = 128 × 128, slice thickness = 0.75 mm, encoding direction = 6, b = 650 s/mm², receiver bandwidth = 300 kHz, 20 axial slices, and scan duration = 4.5 min with 4-segment spin-echo EPI acquisition.

Vital physical signs of respiration rate and heart rate were recorded. Respiration rate was monitored using a MR-compatible system (SA Instruments, Stony Brook, USA). Heart rate was measured non-invasively (to avoid potential physiological perturbation due to electrocardiogram needle penetrations) using an ultra-short-evolution-time (UTE) MRI sequence. This sequence repeatedly acquired the center k-space to generate a time course of MR signal intensity that reflected the R-R interval. The experimental parameters were as follows:²⁵ TR/TE = 8/0.3 ms, FOV = 25 mm × 25 mm, matrix size = 96 × 96, slice thickness = 3.0 mm, and scan duration = 0.4 min. An axial imaging slice was positioned over the heart by reference to a scout image targeting the abdomen.

At the end of the MRI sessions, blood samples (one or two drops) were collected from each mouse using the submandibular bleeding method, and hemoglobin levels were measured with a HemoCue device.

2.3. Immunofluorescent staining of brain tissue sections

Collection and preparation of brain sections were performed accordingly to the previously reported protocols.^{2, 8} Briefly, mice were perfused transcardially with PBS, followed by phosphate-buffered 4% paraformaldehyde. The brains were extracted, fixed in phosphate-buffered 4% paraformaldehyde for 24 hours at 4°C, divided into two hemispheres from the midsagittal line, and dehydrated with 30% sucrose. Frozen brain tissue sections at the thickness of 50 μm were obtained with a cryostat. After blocking and antigen retrieval procedures, the brain sections were incubated with primary antibodies for 3 days at 4°C. Subsequently, sections were incubated overnight with matched fluorescence-linked secondary antibodies (Jackson ImmunoResearch Laboratories). The primary antibodies used were CD13 (1: 200, MCA2183, Bio-Rad) and CD31 (1:200, FAB3628G, R&D Systems). Fluorescent-tagged images were scanned using a confocal microscope (Zeiss LSM780 FCS, Leica SP5).

2.4. Data processing

Custom-written MATLAB scripts and graphic-user-interface tools were used following the procedures described briefly below.

For PC MRI, the artery of interest was first manually delineated on the complex-difference image, which showed excellent contrast between the vessel and surrounding tissue.¹⁶ The mask was then applied to the velocity map, and the integration of arterial voxels yielded blood flow through the artery in ml/min. The total blood flow to the brain was calculated by summing the blood flow values across the four major feeding arteries. To account for brain-size differences and obtain unit-mass CBF values, the total blood flow was divided by the brain weight, calculated as the product of brain volume and tissue density (1.04 g/ml²⁶).The gCBF value was reported as milliliters per 100 grams of brain tissue per minute (ml/100g/min).

The processing of pCASL data followed established procedures.¹⁴ Pair-wise subtraction between control and labeled images (i.e., $M_{ctr}-M_{lbl}$) was first performed to yield a difference image, which was then divided by an M₀ image (obtained by scaling the control image¹⁸) to provide a perfusion index image: $CB{F}_{index}=\frac{M_{ctr}-M_{lbl}}{M_0}$. The perfusion index maps were co-registered and normalized to a mouse brain template²⁷, resized to recover the original acquisition resolutions, and rescaled by reference to the gCBF values (from PC MRI) to obtain absolute values: $rCBF=\frac{CB{F}_{index}\cdot gCBF}{Mean\left(CB{F}_{index}\right)}$. Regions of interest (ROIs) were drawn on the averaged control images to encompass the cerebellum, midbrain, isocortex, hippocampus, thalamus, hypothalamus, striatum, and olfactory area by reference to the mouse brain atlas (https://atlas.brain-map.org/). Voxel-wise CBF values within each ROI were averaged to estimate the corresponding perfusion levels.

For each TRUST dataset, subtraction between the control and labeled images was performed to obtain difference images.¹⁷ An ROI was manually drawn on the difference image to encompass the confluence of sinuses. Four voxels exhibiting the highest difference signals were automatically selected for spatial averaging. Venous blood signal intensities at different eTE values were fitted into a monoexponential function to obtain venous T₂. Finally, T₂ was converted into Y_v using a T₂-Y_v calibration plot reported by Li and coworkers.²⁸

CMRO₂ was computed from gCBF and Y_v using the Fick principle,¹⁹ i.e., $CMR{O}_2=C_a\cdot \left(Y_a-Y_v\right)\cdot gCBF$, where C_a denotes the molar concentration of oxygen in a unit volume of blood and was assumed to be 882.1 µmol O₂/100 ml blood based on previous literature²⁹ and $Y_a$ was assumed to be 0.99.³⁰ CMRO₂ was written in the unit of µmol oxygen per 100 grams of brain tissue per minute (µmol O₂/100g/min).

Analogous to the pCASL data processing, ADC maps were co-registered and normalized to the brain template. ROIs were drawn on the averaged M₀ images to estimate regional ADC values.

ImageJ (National Institute of Health, USA) was used to quantify the percentage area of CD31⁺ in different brain regions.^{2, 8} The scales were set based on known dimensions within the microscope images. ROIs, including the cortex, hippocampus, striatum, and olfactory area, were labeled by referring to the mouse brain atlas consistently across all microscope images. For each ROI, images were displayed in grayscale, and a threshold value was applied to isolate CD31⁺ positive regions. The “Measure” function was then used to calculate the “% Area” of CD31⁺ staining for each region. To assess pericyte coverage, CD13 and CD31 channels were processed with the “colocalization 2” plugin in ImageJ, which calculates the degree of overlap between the two markers. Colocalization thresholds were applied to isolate and quantify areas where CD13⁺ pericytes were in proximity to or overlapping with CD31⁺ endothelial structures. The plugin provided an output metric indicating the colocalized area as a percentage of the CD31⁺ area, representing pericyte coverage.

2.5. Statistical analyses

An unpaired Student’s t-test was used to assess group-wise differences in brain volume, hematocrit, and ADC, and reported with t-statistics and degree of freedom (DF) in the format of t(DF). A linear regression model was applied to examine the group effects in functional measurements (OEF, gCBF, rCBF, and CMRO₂), including respiration rate and heart rate as co-variates. These vital signs were used as markers to account for variations in anesthetic depths across the mice.^{25, 31} When presenting effect estimates, the coefficients of effect and 95% confidence intervals (CI) were provided. A P value of < 0.05 was considered statistically significant, and non-significant differences were marked with n.s.

3. Results

3.1. Vascular dysfunction in Pdgfb^cTG mice

Brain volume was not significantly different between WT and Pdgfb^cTG mice (Figure 2A, t(8) = −0.380, P = 0.713), suggesting the absence of brain atrophy or swelling. Figure 2B shows the velocity maps of PC MRI. The Pdgfb^cTG mouse exhibited reduced blood-flow velocity or arterial area in its major feeding arteries. At the group level, Pdgfb^cTG mice had significantly reduced gCBF (Figure 2C, coefficient = −72.5 ml/100g/min, CI = [−132.7, −12.4], P = 0.025). Based on the perfusion maps obtained from pCASL (Figure 2D), we confirmed the existence of regional vulnerability to perfusion deficiency. There was no significant rCBF impairment in the cerebellum (Figure 2E, coefficient = −61.8 ml/100g/min, CI = [−143.5, 19.9], P=0.113), midbrain (Figure 2F, coefficient = −107.3 ml/100g/min, CI = [−222.7, 8.1], P=0.063), isocortex (Figure 2G, coefficient = −73.035 ml/100g/min, CI = [−161.1, 15.0], P = 0.088), hippocampus (Figure 2H, coefficient = −18.6 ml/100g/min, CI = [−112.0, 74.7], P = 0.642), thalamus (Figure 2I, coefficient = −35.6 ml/100g/min, CI = [−112.3, 41.2], P = 0.300), or hypothalamus (Figure 2J, coefficient = −43.203 ml/100g/min, CI = [−130.8, 44.4], P = 0.273). In contrast, rCBF was significantly impaired in the striatum (Figure 2K, coefficient = −89.8 ml/100g/min, CI = [−137.5, −42.1], P = 0.004) and olfactory area (Figure 2L, coefficient = −66.2 ml/100g/min, CI = [−123.8, −8.5], P = 0.031).

Figure 2.

Comparisons of vascular function between WT and Pdgfb^cTG mice: (A) brain volume, (B) velocity maps of PC MRI, (C) gCBF, (D) averaged regional perfusion maps obtained using pCASL MRI with ROIs overlaid on the averaged control image, (E) cerebellar rCBF, (F) midbrain rCBF, (G) isocortical rCBF, (H) hippocampal rCBF, (I) thalamic rCBF, (J) hypothalamic rCBF, (K) striatal rCBF, and (L) olfactory rCBF.

As shown above, normalization by PC-based gCBF values was implemented to provide rCBF maps with absolute values. To confirm that the normalization did not affect the statistical findings, we further analyzed the pCASL data without normalization, yielding consistent results. There was significant rCBF impairment in the striatum (Supplementary Figure 1H, coefficient = −1.07%, CI = [−2.00, −0.13], P = 0.031) and olfactory area (Supplementary Figure 1I, coefficient = −0.80%, CI = [−1.52, −0.08], P = 0.034), but no significant impairment in other brain regions (see Supplementary Figure 1).

3.2. Metabolic dysfunction in Pdgfb^cTG mice

Figures 3A-B show the TRUST images of representative WT and Pdgfb^cTG mice. Difference images were obtained by pair-wise subtraction between control and labeled images. Venous signals at the confluence of sagittal sinuses were fitted and converted into Y_v using the T₂-Y_v calibration plot reported by Li et al. (Figure 3C)²⁸. Pdgfb^cTG mice exhibited significantly higher OEF (Figure 3D, coefficient = 6.7%, CI = [3.5, 9.9], P = 0.002) and CMRO₂ (Figure 3E, coefficient = 80.5 µmol/100g/min, CI = [7.9, 153.1], P = 0.035). Since hematocrit levels can be a potential confounding factor for T₂-based oxygenation measurements,³² we measured hemoglobin levels and found no significant difference (Figure 3F, t(8) = 1.43, P = 0.191). Therefore, the elevated OEF and CMRO₂ are more likely outcomes of metabolic perturbation rather than hematopathy.

Figure 3.

Comparison of brain metabolism between WT and Pdgfb^cTG mice. (A) and (B) show representative TRUST images (control, labeled, and difference) at different effective TE (0.3, 20, and 40 ms) for WT and Pdgfb^cTG mice, respectively. (C) shows the T₂-Y_v calibration curve reported by Li et al. (NMR Biomed 2020; 33: e4207). (D, E, and F) present the comparisons of OEF, CMRO₂, and hemoglobin levels, respectively.

3.3. Characterizations of diffusion properties in the Pdgfb^cTG model

Regional ADC maps were compared between WT and Pdgfb^cTG mice (Figure 4A). A significant difference was observed in the olfactory area (Figure 4B, t(8) = 2.88, P = 0.021), but not in other regions (Figure 4B, |t(8)| ≤ 1.99, P ≥ 0.082). Additionally, quantitative relaxation measurements (T₂ and T₁) were compared between the WT and Pdgfb^cTG mice with no significant differences observed in the cortex, hippocampus, thalamus, and hypothalamus (Supplementary Figure 2).

Figure 4.

Comparison of microstructural features between WT and Pdgfb^cTG mice. (A) shows averaged ADC maps for WT and Pdgfb^cTG groups, along with corresponding ROIs. (B) presents regional comparisons of ADC between WT and Pdgfb^cTG mice.

Characterizations of diffusion and relaxation properties indicate that Pdgfb^cTG mice are not associated with significant microstructural abnormalities in major brain regions of the parietal lobe. Additionally, the olfactory area of the frontal lobe in Pdgfb^cTG mice exhibited abnormal diffusion features.

3.4. Abnormalities of vessel density and pericyte coverage in Pdgfb^cTG mice

Figure 5 summarizes the results of immunofluorescence staining on vessel density (CD31⁺/region area) and pericyte coverage (CD13⁺/CD31⁺). Representative microscope images of WT and Pdgfb^cTG mice exhibited similar qualities (Figures 5A-D). Zoomed-in images in Figures 5A-D reveal tight attachment of CD13⁺ pericytes to CD31⁺ vascular endothelial cells across all regions in WT mice, indicating robust pericyte coverage of the capillaries (white arrows in Figures 5A-D). In contrast, many CD13⁺ pericytes in Pdgfb^cTG mice exhibited detachment from the CD31⁺ vessels in the cortex, hippocampus, and olfactory area (yellow arrows in Figures 5A, B, and D). At the group level, vessel densities were preserved in the cortex (Figure 5E, t(4) = 1.33, P = 0.253) and striatum (Figure 5I, DF = 4, t = 0.89, P = 0.426), but reduced in the hippocampus (Figure 5G, t(4) = 2.82, P = 0.048) and olfactory area (Figure 5K, t(4) = 3.78, P = 0.019). Meanwhile, pericyte coverages were significantly altered in the cortex (Figure 5F, t(4) = 3.90, P = 0.018), hippocampus (Figure 5H, t(4) = 4.45, P = 0.011), and olfactory area (Figure 5L, t(4) = 2.95, P = 0.042), but not in the striatum (Figure 5J, t(4) = −0.37, P = 0.727).

Figure 5.

Comparison of immunofluorescence staining results between WT and Pdgfb^cTG mice. (A-D) show representative microscope images from the cortex, hippocampus, striatum, and olfactory area, respectively. The scale bar corresponds to 200 µm. Zoomed-in images (right two panels) highlight the co-localization of CD13 and CD31. White arrows indicate normal pericytes in different regions, while yellow arrows point to detached pericytes in the cortex, hippocampus, and olfactory area of Pdgfb^cTG mice. Vessel density was quantified as the ratio of the CD31⁺ area to the total regional area. Pericyte coverage was quantified as the ratio of the CD13⁺ area to the CD31⁺ area. (E), (G), (I), and (K) present comparisons of vessel density in the cortex, hippocampus, striatum, and olfactory area, respectively. (F), (H), (J), and (L) show comparisons of pericyte coverage in the cortex, hippocampus, striatum, and olfactory area, respectively.

The cortex was associated with normal vessel density, normal rCBF, but reduced pericyte coverage, implying that impairment of pericyte coverage alone is insufficient to drive perfusion deficiency. Moreover, the simultaneous loss of vessel and pericyte coverage was still insufficient to induce rCBF deficiency, as supported by the finding in the hippocampus, where reduced vessel density and pericyte coverage were accompanied by unchanged rCBF. On the other hand, the striatum, which showed no abnormalities in vessel density or pericyte coverage, exhibited reduced rCBF, indicating that impairment of vessel density or pericyte coverage is not required to drive perfusion deficiency.

Based on these results, regional perfusion is decoupled from vessel density or pericyte coverage, even although vessels serve as the primary structure to carry blood flow and pericytes regulate this flow.³³ One possible explanation is the existence of compensatory mechanisms,^34–36 which adds complexity to the investigation of the relationship between microscopic structure and macroscopic neurofunction.

4. Discussion

In this study, we characterized the microvascular responses to elevated circulating PDGF-BB using the Pdgfb^cTG mouse model with multiparametric and multimodal techniques. We found significant vascular and metabolic alterations in the Pdgfb^cTG mice, alongside pathological abnormalities in vessel density and pericyte coverage. These findings contribute to a better understanding of the effect of elevated circulating PDGF-BB on neurofunction.

Pdgfb^cTG mice are preosteoclast-specific Pdgfb transgenic mice with markedly high serum PDGF-BB concentrations from 1.5 month of age.⁸ PDGF-BB, produced by preosteoclasts in bone, can promote age-associated vascular impairment in the hippocampus, and the Pdgfb^cTG mice faithfully replicate age-associated BBB dysfunction in the hippocampus and subsequent cognitive decline.⁸ In the present study, we further confirm that Pdgfb^cTG mice exhibit vascular and metabolic dysfunctions. These findings support a unique association between bone aging and cerebrovascular disorders.

Perfusion imaging in this study revealed impaired cerebral blood flow (CBF) in the striatum and olfactory area, but not in other regions. This selective vulnerability to CBF impairment may be attributed to region-specific vascular characteristics, responses to elevated PDGF-BB, and compensatory mechanisms. Different brain regions have varying vessel densities,³⁷ which may contribute to their diversified vulnerability to CBF impairment. The expression of PDGF receptors, which can influence the local response to elevated PDGF-BB, may also vary across brain regions.³⁸ The hippocampus benefits from a mixed blood supply, which increases the reliability of its circulation.³⁹ In contrast, the striatum and olfactory area may lack such a compensatory mechanism, making them more vulnerable to vascular dysfunction. These factors, either alone or in combination, could contribute to the selective vulnerability to CBF impairment. Additionally, selective vulnerability to ADC impairment was observed. According to the literature,⁴⁰ the PDGF B-chain protein exhibits the strongest expression in the olfactory system, particularly in olfactory nerve fibers, and remains at high levels in this region. The ADC impairment in the olfactory area following elevated circulating PDGF-BB could be related to the reported high levels of PDGF B-chain protein in the olfactory system.

According to a previous report,⁸ abnormally elevated circulating PDGF-BB in aged mice is associated with capillary reduction, pericyte loss, and blood-brain barrier (BBB) dysfunction in the hippocampus. Further experiments using transgenic mice that overexpress and knockout Pdgfb support the mechanistic link between PDGF-BB and hippocampal vascular impairment.⁸ Moreover, prolonged exposure of brain pericytes to high concentrations of PDGF-BB upregulates matrix metalloproteinase 14 (MMP14), which promotes the ectodomain shedding of PDGFRβ from the pericyte surface. This process contributes to pericyte loss and capillary reduction. By contrast, MMP inhibitor treatment alleviates hippocampal pericyte loss and capillary reduction. These findings suggest that elevated PDGF-BB impairs hippocampal vasculature by inducing PDGFRβ shedding from pericytes, disrupting their function, and degenerating the BBB integrity.⁸ Following blood-brain barrier (BBB) disruption, toxins may extravasate from capillaries into the surrounding tissue, leading to increased neuroinflammatory activity.¹² The observed elevation in CMRO₂ could reflect an increase in mitochondrial oxidative phosphorylation, as a mechanism to generate more ATP in response to greater metabolic demands.⁴¹ Thereafter, the increase in OEF naturally follows as a compensatory response to the elevated CMRO₂ but impaired blood flow.³⁵ Further studies are needed to fully unravel the mechanistic relationship between hypermetabolism and elevated circulating PDGF-BB. Additionally, increased circulating PDGF-BB during aging leads to elevated soluble PDGFRβ in the cerebrospinal fluid (CSF).⁸ As increased soluble PDGFRβ in the CSF is recognized as a marker of pericyte degeneration and BBB impairment, we believe the elevated serum PDGF-BB could also serve as a marker of BBB dysfunction. However, we have not yet analyzed the correlation between soluble PDGFRβ in the CSF and PDGF-BB in the serum due to the technical challenges of obtaining sufficient CSF from mice. Future studies could address this question using samples collected from larger animals or human subjects.

With normal aging in humans, CBF decreases, while OEF and CMRO₂ increase.^42–45 CBF impairment is primarily attributed to chronic developments such as stenosis, arteriovenous malfunctions, atherosclerosis, vascular-smooth-muscle-cell loss, basement membrane thickening, and cardiac dysfunction - common vascular complications in the aging population.⁴⁴ An elevated CMRO₂ is believed to compensate neuronal deficiency, such as mitochondrial dysfunction,⁴⁶ and/or attributed to increased glial cell activity⁴⁷. Due to neurovascular coupling,³⁵ an increased OEF naturally follows from the rise in CMRO₂ when blood flow is reduced. A longitudinal aging study in C57BL/6 mice covering major portions of mouse lifespan (3–20 months) reveals consistent changes in OEF and CMRO₂.¹⁵ In contrast, CBF increases slightly with age at a rate of 1.93 ml/100g/min/month.¹⁵ The discrepancy between human and mouse studies suggests to two possibilities: first, CBF changes in mice are non-monotonic with a sharp decline at very advanced ages (>20 months); second, the typical mouse lifespan (2–3 years) may not be long enough to induce significant CBF deficiency. According to the longitudinal study of mouse aging, the rates of change for OEF and CMRO₂ are 0.20%/month and 6.70 µmol/100g/min/month, respectively.¹⁵ In comparisons, Pdgfb^cTG mice exhibit much more pronounced changes: −12.08 ml/100g/min/month for CBF, 1.12%/month for OEF, and 13.42 µmol/100g/min/month for CMRO₂. Clearly, the Pdgfb^cTG model is associated with accelerated vascular and metabolic aging. If vascular aging is of primary interest, Pdgfb^cTG mice would be more suitable for study than regular C57BL/6 mice. In the literature, the interleukin 10 (IL-10) knockout model develops vascular dysfunction, such as increased vascular stiffness and decreased vascular relaxation, at 9 months of age.⁴⁸ The dietary model of hyperhomocysteinemia, a mouse model for small vessel diseases, exhibits reduced CBF after approximately 2 months on a customized diet.⁴⁹ Dominant mutations in NOTCH₃ cause cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), leading to attenuated myogenic responses, impaired cerebrovascular autoregulation, and reduced CBF by 18 months of age.⁵⁰ Our finding of reduced CBF is consistent with the results observed in these mouse models, and the Pdgfb^cTG model presents a complementary option to existing models for vascular aging.

Our imaging and histological observations indicate that impaired pericyte coverage, with or without vessel loss, is insufficient to drive CBF deficiency. The presence of compensatory mechanisms may explain the decoupling between CBF, vessel density, and pericyte coverage. In the case of vessel loss, existing cerebral arteries and arterioles can dilate in response to hypoxia or chemical mediators (e.g., nitric oxide or carbon dioxide), increasing blood flow to the affected brain regions and compensating for reduced CBF.³⁴ Furthermore, the BBB can adapt to regulate nutrient transportations.³⁶ In response to reduced CBF, the BBB may become more permeable, allowing for faster exchange of vital substances. With these compensatory mechanisms in place, the regulation of neurofunction becomes a complex biological process. Therefore, changes in the foundational structure do not necessarily lead to changes in neurofunction, exemplifying the decoupling between structure and function. As a result, imaging-based functional measurements and histological evaluations complement each other in revealing the full landscape of disease pathophysiology. It is important to note that clinical diagnoses and therapeutic evaluations are often based on symptoms and functions. For example, ischemic stroke is characterized by significantly reduced CBF, a parameter of vascular function, and the efficacy of thrombolytic therapy can be assessed by perfusion imaging to map the CBF distribution.⁵¹ In vivo functional imaging bridges the biological mechanisms with clinical diagnoses and therapeutic evaluations. The integration of non-invasive imaging with invasive histology offers a powerful tool for multiscale characterization.

Brain physiology provides valuable information for the diagnosis, stratification, staging, and treatment monitoring of various diseases.^{42, 52–53} Among the physiological measurements, gCBF, rCBF, OEF, and CMRO₂ are promising imaging biomarkers and have been extensively used in previous studies. In a rat model of cardiac arrest, gCBF, OEF, and CMRO₂ recover during the acute phase after the return of spontaneous circulation (ROSC), and the extent of these recoveries at the early stage (~3 hours post-ROSC) can predict neurological outcomes at a later stage (24 hours post-ROSC).⁵⁴ Moreover, the hippocampus is the most vulnerable region to cardiac arrest, and restoring of hippocampal rCBF is a priority to promote neurological recovery.⁵⁵ Beyond acute physiological perturbations (e.g., cardiac arrest), physiological measurements are highly feasibility for monitoring pathologies with slow progression. Patients with AD exhibit reduced OEF and CMRO₂ before brain atrophy occurs,⁵³ and cognitively healthy carriers of the apolipoprotein E4 gene show diminished OEF independent of amyloid burden.⁵⁶ Furthermore, CBF has been confirmed as a sensitive marker for studying the co-pathogenesis of AD and sleep disruption.⁵⁷

Quantitative relaxation measurements in this study were limited to a single slice. Future studies employing fast, whole-brain quantitative T₂/T₁ mapping MRI techniques^58–59 will prove helpful in cross-validating the present region-specific findings by providing additional microstructural information. It is important to note that, in such studies, experimental imperfections, such as B₁ field inhomogeneity, compromised slice selection profiles, and suboptimal gradient performance, can affect the accuracy of quantitative relaxation measurements and should be carefully considered. We employed non-contrast, non-invasive MRI techniques to measure in vivo brain physiology, minimizing potential confounding from invasive procedures and contrast agents. However, integrating MRI with other optical imaging methods could enhance the specificity of measurements and improve the interpretation of MRI findings.^60–61 Furthermore, NMR-based metabolomics^62–63 could provide additional molecular insights into metabolic pathway perturbations, aiding in the understanding of causal and consequential relationships between neurofunction and metabolites.

We performed Monte Carlo simulations (20,000 iterations) based on the collected data to assess the statistical power of this study. The simulations revealed a power of 0.83 to detect differences in gCBF, 0.99 for OEF, and 0.72 for CMRO₂, which are the key findings supporting the study’s conclusions.

Results from the current study should be interpreted in the context of several limitations. First, while this study characterizes the vascular and metabolic responses to elevated circulating PDGF-BB, providing neurofunctional evidence for its pro-aging effect, it lacks mechanistic depth to fully elucidate the underlying causal relationships at the molecular level. Second, only male mice were included in this study. According to our previous study,² circulating PDGF-BB levels are differentially altered by age in male and female mice. Further investigations into sex differences, particularly the effects of menopause on the neurofunctional profile, will provide deeper insight into the neurofunctional responses to elevated circulating PDGF-BB. Thirdly, our study is limited to a mouse model. Given the species difference between humans and mice, future studies in humans repeating the present multiparametric observations will be necessary. It is worth noting that MRI techniques for measuring CBF, OEF, and CMRO₂ are readily applicable to human scanners,^{23, 64–66} suggesting that the clinical translation of our findings is straightforward. Additionally, our multiparametric observations suggest that the onset of abnormalities may occur over varying timelines. The current cross-sectional study is limited in its ability to capture the full landscape of multifaceted pathophysiological progression. A future study employing a longitudinal design with a larger sample size would be valuable to determine the cascade of different physiological and microstructural events over time. The current cross-sectional evidence can serve as a guide for future mechanistic studies to determine the causal relationships between different abnormalities, thereby identifying potential therapeutic targets.

In conclusion, hypoperfusion and hypermetabolism occur as a result of elevated circulating PDGF-BB in mice. Our study provides neurofunctional evidence supporting the notion that increased circulating PDGF-BB accelerates both vascular and metabolic dysfunction in brain.

Data Availability Statement

MRI and immunofluorescent staining data reported in this work are available upon request.