[¹⁸F]FDG PET/CT is a quantitative tool to probe vascular inflammation and immune system metabolic activation in atherosclerotic nonhuman primates

Abstract

Background

Atherosclerosis is a major global health concern, necessitating advanced research models to better understand pathophysiology and inform potential interventions. Due to their physiological similarities to humans, nonhuman primates (NHPs), particularly crab-eating macaques, are valuable models to study the development and progression of atherosclerosis. This retrospective study utilized quantitative 2-deoxy-2-[fluorine-18]fluoro-D-glucose ([¹⁸F]FDG) positron emission tomography–computed tomography (PET/CT) to characterize a cohort of crab-eating macaques with diet-induced atherosclerosis. We evaluated [¹⁸F]FDG uptake in the vasculature, bone marrow, spleen, and adipose tissues from macaques fed an atherogenic diet compared with those fed a regular chow diet.

Results

[¹⁸F]FDG uptake was significantly higher in the vasculature, bone marrow, and adipose tissues of macaques on the atherogenic diet, along with increased vascular calcification. Correlation analyses demonstrated moderate to strong associations among vascular inflammation, calcification, and metabolic activity in the bone marrow and adipose tissues.

Conclusions

We provide a robust methodology for quantitative PET/CT imaging analysis of atherosclerosis in an established NHP model. Our work adds to the research toolkit and highlights the translational potential of imaging in NHP models to investigate systemic cardiovascular disease, including of fundamental pathobiology and of targeted interventions.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13550-026-01449-1.

Background

Atherosclerosis, a complex inflammatory condition involving primarily the vasculature but also other organs and tissues, can lead to clinically significant cardiovascular disease (CVD) and fatal cardiovascular and cerebrovascular events [1]; indeed, atherosclerosis is one of the main causes of morbidity and mortality worldwide [2].

Crab-eating macaques (Macaca fascicularis Raffles, 1821) are used as a well-characterized nonhuman primate (NHP) model of atherosclerosis that has contributed broad insight into the process of atherogenesis in humans [3–16]. Compared with smaller, non-primate mammals, such as laboratory mice, atherosclerotic NHPs develop vascular disease that more closely resembles human disease phenotypes, including in atherosclerotic plaque location, composition, and lipoprotein expression [17–21]. NHPs are also more naturally suited for studies that investigate the impact of known host and environmental contributors to human atherogenesis, such as biological sex [22–24], social status and associated behavior [25, 26], and stress [23, 27], on cardiovascular health.

Multimodal noninvasive imaging plays an important role in the clinical evaluation of patients with atherosclerosis [28]. Imaging is also widely used in both preclinical and clinical cardiovascular research to better understand the pathogenesis of atherosclerosis. Although imaging has been extensively used to characterize animal models of atherosclerosis [29–31], noninvasive imaging of atherosclerosis in NHPs, particularly using quantitative molecular imaging disease markers, has been much less common than in smaller animal models [32–41]. The development and application of robust, quantitative (molecular) imaging methodologies in NHPs may further our ability to use the NHP model of atherosclerosis as a tool to improve our understanding of CVD and human health.

Molecular imaging with 2-deoxy-2-[fluorine-18]fluoro-D-glucose ([¹⁸F]FDG) positron emission tomography–computed tomography (PET/CT) is widely applied in atherosclerosis research [42–47]. [¹⁸F]FDG is a glucose analog that is internalized but not metabolized by metabolically active cells, including the inflammatory cells found within atherosclerotic plaques. Once internalized, the tracer becomes trapped inside these cells, resulting in its accumulation and subsequent possible detection by PET/CT imaging. [¹⁸F]FDG vascular uptake quantified by PET/CT has been validated as a noninvasive quantitative imaging marker of atherosclerotic plaque inflammation, an important feature of vulnerable atherosclerotic plaques in both animals and humans [48, 49]. [¹⁸F]FDG PET/CT imaging also serves as a validated biomarker of inflammation and metabolic activation in both the vasculature and in other tissues implicated in atherogenesis (e.g., bone marrow, spleen, and subcutaneous and visceral adipose tissues [SAT and VAT, respectively]) [50–60]. Furthermore, [¹⁸F]FDG PET/CT is used as a surrogate imaging endpoint to evaluate therapeutic compounds in preclinical and clinical trials of atherosclerosis [61–65].

In this retrospective study, we used quantitative [¹⁸F]FDG PET/CT to evaluate a cohort of crab-eating macaques with diet-induced atherosclerosis. We first established a robust methodology to characterize and quantify [¹⁸F]FDG PET/CT uptake in the vasculature, bone marrow, spleen, VAT, and SAT of crab-eating macaques fed an atherogenic diet. We then systematically compared [¹⁸F]FDG uptake quantification of the atherogenic diet group with a control group fed a regular chow diet. Finally, we explored the correlation among quantitative [¹⁸F]FDG uptake in the vasculature, lymphoid organs, and adipose tissues within our dataset.

Methods

Study design

This study retrospectively analyzed PET/CT imaging data from 27 crab-eating macaques (Macaca fascicularis Raffles, 1821) that were used in multiple prospective severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) exposure studies. To avoid potential confounders from infection-related effects, only imaging data acquired before viral exposure was used for the purpose of these retrospective analyses. Out of the 27 crab-eating macaques, 15 were fed a high-fat atherogenic diet (Table S1) for a minimum of 21 mo to induce atherosclerosis, as previously established (heretofore, the “atherogenic diet group”) [66–68]. The remaining 12 animals were fed regular monkey chow (LabDiet 5045 High Protein Monkey Diet) (heretofore, the “chow diet group”, “control group”, or “controls”)” (Fig. 1 and Table S2). The primary endpoint of this study was to compare relative [¹⁸F]FDG uptake in the vasculature, bone marrow, spleen, VAT, and SAT in the atherogenic diet group (n = 15) versus the chow diet group (n = 12). The presence of vascular calcification in the two groups was also evaluated by CT. As a secondary endpoint, correlations among [¹⁸F]FDG uptake in the vasculature, bone marrow, spleen, VAT, and SAT and vascular calcification by CT were also evaluated.

Fig. 1.

Study design and PET/CT image analysis. The study included 27 crab-eating macaques, divided into two groups based on their diet: an atherogenic diet group (n = 15) and a chow diet group (controls; n = 12). Whole-body 2-deoxy-2-[fluorine-18]fluoro-D-glucose ([¹⁸F]FDG) positron emission tomography–computed tomography (PET/CT) imaging was conducted on all animals. Image analysis quantified [¹⁸F]FDG uptake in vasculature (abdominal aorta [ABD], infrarenal aorta [IR], suprarenal aorta [SR], partial aortic section near the renal arteries [P2_R]), lymphoid organs (spleen and bone marrow [BM]), and fat tissues (subcutaneous adipose tissue [SAT], visceral adipose tissue [VAT]), as well as vascular calcification (VC) in the ascending aorta and arch

Clinical and pathological evaluation

Body weight and white blood cell (WBC) count data were collected in all animals. White blood cell counts were determined using the Procyte DX Hematology Analyzer (IDEXX Laboratories) (Table S3) [69]. In the atherogenic diet group, atherosclerosis was verified by histopathological analysis of vascular specimens using hematoxylin and eosin staining, as well Masson’s trichrome and Verhoeff van Gieson elastin staining.

PET/CT image acquisition

Prior to imaging, animals were anesthetized by intramuscular injection of ketamine (15 mg/kg) and glycopyrrolate (0.06 mg/kg). Blood glucose levels were measured in all animals before imaging (Table 1 and Table S2). During imaging, anesthesia was maintained using a constant rate of intravenous infusion of propofol (0.3 mg/kg/min). Vital signs were monitored throughout the acquisition. Whole-body PET/CT was performed on a Gemini TF PET/CT scanner (Philips Healthcare), with the animals in a supine head-out/feet-in position. CT imaging was performed in helical scan mode with the following parameter settings: ultra-high resolution, 120–140 kVp, 230 mAs per slice, 1–2-mm thickness, 0.5-mm increment, 0.688-mm pitch, and 16 × 0.75 collimation. No CT contrast agent was administered, and animals breathed freely during imaging. Whole-body PET was performed approximately 1 h (median time: 55.83 min, IQR: 55.33–56.33) after [¹⁸F]FDG intravenous injection (median net injected dose: 86 MBq, interquartile range [IQR]: 82.0–89.9) (Table 1 and Table S2), with the following parameter settings: total body scan covering from the head to the middle of the thighs; six or seven bed positions (depending on the animal’s size); and a 5-min scan per position with 50% overlap.

Table 1.

Study characteristics

Parameter	Chow diet group (controls; n = 12)	Atherogenic diet group (n = 15)	Total (n = 27)
Age (yr)	4.1 (3.9–4.3)	6.6 (6.4–7.7)	5.5 (4.1–7.0)
Sex, Female/Male (n, %)	7 (58.3)/5 (41.7)	0 (0)/15 (100)	7 (26)/20 (74)
Body weight (kg)	3.7 (3.5–4.6)	6.7 (5.9–7.6)	5.3 (4.0–6.7)
Pre-scan blood glucose (mg/dL)	63.0 (60.3–71.0)	56.0 (51.0–59.0)	59.9 (56.3–63.5)
Fasting duration (min)	180.0 (67.5–262.5)	231.0 (127.0—306.0)	202.0 (172.5—231.5)
Tracer circulation time (min)	53.5 (52.3–57.5)	57.0 (55.0–57.0)	55.8 (55.3–56.3)
Net injected radioactivity (MBq)	90.2 (85.3–95.1)	80.9 (78.2–84.3)	86.0 (82.0–89.9)

Data are presented as the median with interquartile range (IQR)

PET/CT image reconstruction

CT images were reconstructed using a 512 × 512 matrix size for a 250-mm transverse field of view, resulting in a pixel size of 0.488 mm. PET images were reconstructed in three-dimensional mode using a 128 × 128 matrix size with a 256-mm transverse field-of-view, generating a voxel size of 2 × 2 × 2 mm (8 mm³). All reconstructed PET images were corrected for radioactive decay, uniformity, random coincidences, and attenuation and scattering of PET radiation in situ. Further details on image acquisition and reconstruction have been previously published [69].

PET image analysis

PET images were analyzed using MIM software version 7.1.6. Regions of interest (ROIs) were drawn in images of target tissues (aorta, bone marrow, spleen, VAT, and SAT) to quantify mean and maximum standardized uptake values (SUV_mean and SUV_max, respectively). Before quantitative analysis, an experienced PET imaging analyst (H.Y.) used SUV_max values to identify regions with significant spillover (especially for the aorta) as a means of quality control; SUV_max values are more susceptible to spillover artifacts and were intentionally not used for downstream quantitative analyses. Target organ and tissue radiotracer uptake was also normalized to background activity in the inferior vena cava (IVC) and expressed as mean target-to-background ratio (TBR_mean, calculated as target SUV_mean divided by background SUV_mean). Details on PET image analysis for target and background organs and tissues are provided below.

Vasculature

The whole aorta (from the ascending aorta to the iliac bifurcation) was contoured by manually placing ROIs around the outer vessel wall contour identified on CT images. ROIs were drawn every three to four CT slices and then automatically interpolated to in-between slices by the image analysis software. ROIs were transferred from CT images to corresponding PET images. To explore potential regional differences in [¹⁸F]FDG uptake [70], PET ROI SUV_mean and SUV_max were recorded for each slice and averaged within four aortic sections: (i) ascending aorta (AS), delineated from the sinotubular junction of the aortic root to the end of the aortic arch; (ii) thoracic aorta (T), delineated from the end of the aortic arch to the diaphragm; (iii) suprarenal abdominal aorta (SR), delineated from the diaphragm to the top of the left renal artery; (iv) infrarenal abdominal aorta (IR), delineated from the left renal artery to the iliac bifurcation; and (v) abdominal aorta (ABD), defined as a combination of SR and IR sections. Three partial aortic sections were also contoured at specific locations where atherosclerotic plaques are usually observed, namely: (i) the aortic arch (P1_A), defined as the full aortic arch area; (ii) the area around the two renal arteries (P2_R), defined as an area of approximately 1.2 cm in length from the top of the left renal artery to the bottom of the right renal artery; and (iii) the iliac bifurcation (P3_IB), defined as an area of approximately 1.2 cm in length covering the terminal part of the aorta and the proximal part of both common iliac arteries. Vascular [¹⁸F]FDG uptake was quantified as average section/subsection SUV_mean and normalized to IVC background activity to calculate TBR_mean.

Lymphoid organs

For analysis of bone marrow, the vertebral bodies of five lumbar vertebrae were contoured on CT images. Two HU thresholds (for the control group: lower = 300, upper = 600; for the atherogenic diet group: lower = 260, upper = 600) were then applied to the contour to segment the marrow-rich area, which was then transferred to the co-registered PET image to calculate SUV_mean [71, 72]. For analysis of the spleen, a previously developed, in-house deep learning (DL) automated segmentation algorithm [73, 74] was used to segment the macaque spleen in CT images. This algorithm relies on a feature pyramid network (FPN), a type of convolutional neural network (CNN), which has been previously trained and tested on a similar dataset [75]. Spleen contours identified by the DL algorithm were manually reviewed for consistency before being applied to PET images to extract whole-spleen SUV_mean. Bone marrow and spleen radiotracer uptake were also normalized to IVC background activity to calculate TBR_mean.

Adipose tissues

Both VAT and SAT were included in the analysis. ROIs were drawn on CT images and consisted of three spheres for each tissue (5 mm in diameter), located below the kidneys and in the loin area between the third and fifth lumbar vertebrae, respectively. ROIs were transferred from CT images to corresponding PET images, and SUV_mean was recorded for each ROI. SUV_mean for VAT and SAT were also normalized to background activity in the IVC to calculate TBR_mean.

Background activity quantification

Radiotracer background activity for organ SUV_mean normalization was quantified in the IVC. IVC SUV_mean was calculated as the average of three spherical ROIs of 4 mm each in the hepatic area.

CT image analysis

The presence or absence of vessel wall calcification was identified in the ascending aorta and arch, where calcifications were identifiable at a visual inspection of CT images. First, an ROI encompassing the whole ascending aorta was delineated for each animal in the control group. Average ROI HUs were recorded and averaged across all animals in the chow diet group (controls) to establish a reference HU value (mean [HU_ref]), as well as its standard deviation (SD [HU_ref]). For each animal in either the atherogenic diet or control group, vascular calcification (VC) was then defined as the set of voxels (in the ascending aorta ROI) whose HU exceeded mean (HU_ref) + 2 × SD (HU_ref). While this data-driven threshold differs from established cut-offs used to identify vascular calcification in humans, it allowed for adequately powered downstream statistical comparisons in our dataset. For each animal, VC HU_mean was calculated by averaging HU values of all voxels identified as calcification, while HU_max represented the voxel with the highest HU in the identified set. Three macaques in the control group, in which no voxel exceeding the predefined threshold to identify calcification could be identified, were excluded from further quantitative analyses. Bone marrow, spleen, VAT, SAT, and liver HU_mean were calculated from the same ROIs used for PET analysis.

Statistical analysis

Continuous variables were expressed as median IQR or mean ± SD, and categorical variables as n (%). Statistical analyses were performed using GraphPad Prism (version 10.2.0). To compare imaging endpoints between the atherogenic diet group and the control group, variables were classified as primary or secondary (Table 2). Group comparisons were performed using a nonparametric, unpaired Mann–Whitney test with multiple comparison correction using the Holm-Sidak method. To compare [¹⁸F]FDG uptake among different aortic segments in the animals of the atherogenic diet group, a paired nonparametric Friedman test with Dunn’s multiple comparison correction was used. A Spearman rank test was performed to evaluate correlation of multiple variables. A p-value of < 0.05 (after multiple comparison correction) was considered statistically significant for all tests.

Table 2.

Quantitative positron emission tomography–computed tomography (PET/CT) imaging measurements in various regions of interest in crab-eating macaques in the atherogenic diet group and chow diet group (controls)

Region of Interest (ROI)	Measurement	Chow diet group (n = 12)	Atherogenic diet group (n = 15)	Adjusted p-value
Primary variables
Abdominal aorta (ABD)	TBRmean	1.05 (0.90–1.29)	1.55 (1.42–1.65)	0.000283
Suprarenal aorta (SR)	TBRmean	1.01 (0.88–1.17)	1.42 (1.34–1.57)	0.000283
Infrarenal aorta (IR)	TBRmean	1.08 (0.92–1.31)	1.52 (1.44–1.75)	0.000831
Partial aorta near renal arteries (P2_R)	TBRmean	1.11 (0.95–1.30)	1.50 (1.46–1.59)	0.000134
Bone marrow (BM)	TBRmean	2.19 (1.94–2.48)	3.56 (3.12–3.92)	0.000179
Spleen	TBRmean	1.36 (1.21–2.03)	1.53 (1.38–1.58)	0.448216
Visceral adipose tissue (VAT)	TBRmean	0.56 (0.47–0.68)	0.73 (0.57–1.21)	0.021015
Subcutaneous adipose tissue (SAT)	TBRmean	0.47 (0.39–0.67)	0.70 (0.47–0.92)	0.049840
Vascular calcification (VC)	HUmax	117.00 (113.00–138.50)	164.00 (124.00–208.00)	0.011453
Secondary PET variables
Abdominal aorta (ABD)	SUVmean	1.02 (0.95–1.12)	1.22 (1.12–1.48)	0.034818
Suprarenal aorta (SR)	SUVmean	1.02 (1.00–1.04)	1.12 (1.03–1.38)	0.076352
Infrarenal aorta (IR)	SUVmean	1.01 (0.92–1.15)	1.28 (1.17–1.54)	0.033383
Partial aorta near renal arteries (P2_R)	SUVmean	1.04 (0.96–1.11)	1.17 (1.09–1.45)	0.017951
Bone marrow (BM)	SUVmean	2.22 (2.01–2.46)	3.17 (2.44–3.43)	0.002762
Spleen	SUVmean	1.46 (1.05–1.88)	1.27 (1.16–1.38)	0.419877
Visceral adipose tissue (VAT)	SUVmean	0.51 (0.47–0.59)	0.69 (0.47–1.14)	0.246207
Subcutaneous adipose tissue (SAT)	SUVmean	0.47 (0.39–0.52)	0.58 (0.42–0.73)	0.246207
Inferior vena cava (IVC)	SUVmean	0.89 (0.82–1.12)	0.83 (0.75–0.94)	0.246207
Liver	SUVmean	1.18 (1.12–1.32)	1.06 (0.96–1.20)	0.154262
Secondary CT variables
Vascular calcification (VC)	HUmean	113.29 (107.39–116.65)	124.17 (116.92–132.22)	0.022149
Bone marrow (BM)	HUmean	461.34 (437.28–494.73)	428.44 (399.69–445.98)	0.022149
Spleen	HUmean	72.38 (70.52–75.88)	73.10 (70.88–74.60)	0.904561
Visceral adipose tissue (VAT)	HUmean	-79.07 (-85.91–(-72.21))	-75.03 (-85.42–(-59.67))	0.703916
Subcutaneous adipose tissue (SAT)	HUmean	-75.72 (-81.91–(-63.27))	-62.56 (-75.25–(-49.37))	0.331746
Liver	HUmean	84.35 (78.73–88.06)	66.07 (57.95–69.66)	< 0.000001

Data are presented as the median with interquartile range (IQR). p-values were determined using the Mann–Whitney test with multiple comparison correction using the Holm-Sidak method; Bold data values indicate statistical significance (p < 0.05). TBR, target-to-background ratio; SUV, standardized uptake value; HU, Hounsfield unit

Results

Animal characterization

Out of the 27 macaques used in this study, 7 were females and 20 were males. The macaque age range was 3.7–8.8 yr (median age: 5.55 yr, interquartile range [IQR]: 4.13–6.96) and the weight range was 3.14–8.04 kg (median weight: 5.34 kg, IQR: 3.98–6.70) (Table 1). Compared to the control group, animals in the atherogenic diet group had a significantly higher frequency (but not absolute counts) of neutrophils in (p = 0.0023). Both the absolute lymphocyte count and frequency were significantly lower (p = 0.0058 and p = 0.0014, respectively) in the atherogenic diet group compared to the chow diet group. The total white blood cell count, and the absolute numbers and frequency of monocytes, eosinophils, and basophils were similar between groups (Table S3). In the atherogenic diet group, significant atherosclerotic lesions were present systemically in the large elastic and muscular and medium and occasionally small muscular arteries of all animals. By Masson’s trichrome staining, abundant mature and nascent fibrosis was detected in the vascular wall, markedly expanding the intima and frequently replacing the media. By Verhoeff van Gieson elastin staining, frequent fragmentation, splitting, degeneration, and loss of medial elastic laminae, with variable elastosis of the (neo)intima, was noted. In many aortic sections, the (neo)intima was as thick or markedly thicker than the media (Figure S1).

Qualitative assessment and quality control of aortic PET ROIs

Before proceeding with further analysis, an experienced image analyst examined PET [¹⁸F]FDG uptake in all distinct aortic sections and subsections (from each animal) for the presence of artifacts; a total of 189 aortic sections were evaluated. The same examiner reviewed PET images and SUV_max for each segment to identify potential signal contamination from adjacent tissues with high [¹⁸F]FDG uptake. Significant signal contaminations (defined as SUV_max > 2.5) were frequently observed in a total of four aortic sections and subsections, namely AS, T, P1_A, and P3_IB, due to signal spillovers from neighboring cardiac muscles, lymph nodes, vertebrae, or bladder (data not shown); these four sections and subsections were excluded from further analyses. Overall, a total of 108 aorta segments derived from sections ABD, SR, and IR, as well as aortic subsection P2_R, were subsequently used for quantitative analysis.

[¹⁸F]FDG vascular uptake was significantly higher in macaques in the atherogenic diet group compared with those in the chow diet group (controls)

As compared to the control group, [¹⁸F]FDG vascular uptake (expressed as TBR_mean) was significantly higher in animals in the atherogenic diet group in aortic sections ABD, SR, and IR, as well as aortic subsection P2_R (Fig. 2 and Table 2).

Fig. 2.

2-deoxy-2-[fluorine-18]fluoro-D-glucose ([¹⁸F]FDG) positron emission tomography–computed tomography (PET/CT) imaging of crab-eating macaques in the atherogenic diet group and chow diet group (controls). (a) Representative [¹⁸F]FDG PET/CT images demonstrating increased [¹⁸F]FDG uptake in vascular (abdominal aorta [ABD]) and bone marrow (BM) tissues of the atherogenic diet group. The scale bar indicates standardized uptake value (SUV). (b) Representative CT images demonstrating increased vascular calcification (yellow arrows) in the aorta of atherogenic diet group. The scale bar indicates Hounsfield units (HU). LV, left ventricle. (c) Quantification of [¹⁸F]FDG uptake in various regions of interest: abdominal aorta (ABD), suprarenal aorta (SR), infrarenal aorta (IR), partial aortic section near renal arteries (P2_R), bone marrow (BM), spleen, subcutaneous adipose tissue (SAT), and visceral adipose tissue (VAT). (d) Quantification of vascular calcification (VC). Each point represents an individual animal with the median (solid lines) and quartiles (dotted lines). p-values were determined using Mann–Whitney test with multiple comparison corrections. Asterisks indicate statistical significance: *p < 0.05, ***p < 0.001

Median TBR_mean in section ABD was 1.55 (IQR 1.42–1.65) in the atherogenic diet group versus 1.05 (IQR 0.90–1.29) in controls (p < 0.001). Median TBR_mean in section SR was 1.42 (IQR 1.34–1.57) in the atherogenic diet group versus 1.01 (IQR 0.88–1.17) in controls (p < 0.001). Median TBR_mean in section IR was 1.52 (IQR 1.44–1.75) in the atherogenic diet group versus 1.08 (IQR 0.92–1.31) in controls (p < 0.001). Median TBR_mean in subsection P2_R was 1.50 (IQR 1.46–1.59) in the atherogenic diet group versus 1.11 (IQR 0.95–1.30) in controls (p < 0.001) (Fig. 2c and Table 2). Median SUV_mean in the background blood pool (IVC) was not significantly different between groups (Table 2 and Fig. S2a).

Median SUV_mean in section ABD was significantly higher in the atherogenic diet group (1.22, IQR 1.12–1.48) versus controls (1.02, IQR 0.95–1.12; p = 0.035). Median SUV_mean in section SR was 1.12 (IQR 1.03–1.38) in the atherogenic diet group versus 1.02 (IQR 1.00–1.04) controls (p = 0.076). Median SUV_mean in section IR was 1.28 (IQR 1.17–1.54) in the atherogenic diet group versus 1.01 (IQR 0.92–1.15) in controls (p = 0.033). Median SUV_mean in subsection P2_R was 1.17 (IQR 1.09–1.45) in the atherogenic diet group versus 1.04 (IQR 0.96–1.11) in controls (p = 0.018) (Table 2 and Fig. S2a).

In animals in the atherogenic diet group, [¹⁸F]FDG TBR_mean and SUV_mean were similar among most of the different vascular segment analyzed. Only in aortic segment SR were [¹⁸F]FDG TBR_mean and SUV_mean significantly lower compared with segment IR (TBR_mean = 1.42 [IQR 1.34–1.57] versus 1.52 [IQR 1.44–1.75], p = 0.028; SUV_mean = 1.12 [IQR 1.03–1.38] versus 1.28 [IQR 1.17–1.54], p = 0.028) (Table 2, Fig. S2c, and Table S4).

[¹⁸F]FDG uptake was significantly higher in bone marrow and adipose tissues of macaques in the atherogenic diet group compared with controls

Median TBR_mean was significantly higher in the bone marrow of animals in the atherogenic diet group compared with controls (3.56 [IQR 3.12–3.92] and 2.19 [IQR 1.94–2.48], respectively; p < 0.001) but was not significantly different in the spleen (p = 0.448) (Fig. 2 and Table 2). Similarly, bone marrow median SUV_mean was significantly higher in the atherogenic diet group compared with controls (3.17 [IQR 2.44–3.43] versus 2.22 [IQR 2.01–2.46], respectively; p = 0.003). Splenic median SUV_mean was not different between groups (p = 0.420) (Table 2 and Fig. S2a).

Regarding adipose tissues, median TBR_mean of VAT was significantly higher in atherosclerotic animals compared with controls (0.73 [IQR 0.57–1.21] versus 0.56 [IQR 0.47–0.68], p = 0.021). A significant increase was also observed for SAT (0.70 [IQR 0.47–0.92] versus 0.47 [IQR 0.39–0.67], p = 0.0498) (Fig. 2 and Table 2). No significant differences were found in median SUV_mean for either VAT or SAT (p = 0.246 for both) (Table 2 and Fig. S2a).

CT identified consistent vascular calcification in macaques in the atherogenic diet group but not in the chow diet group (controls)

Qualitative visual inspection of CT images of the whole aorta consistently identified calcification in the ascending aorta of animals in the atherogenic diet group; similar calcification was not immediately evident in CT images of controls. For quantitative analysis, vascular calcification was defined as voxels (in the ascending aorta ROI) exceeding a HU threshold of 2 SDs above the mean HU of the ascending aorta of the chow diet group (controls; HU = 110). Using this criterion, 100% (15/15) animals in the atherogenic diet group had vascular calcification compared with 75% (9/12) of controls. In calcified areas, median HU_max was significantly higher in the atherogenic diet group compared with the controls (164.00 [IQR 124.00–208.00] versus 117.0 [IQR 113.0–138.5]; p = 0.011) (Fig. 2 and Table 2). Quantitative CT image analysis also showed that tissue density (expressed as HU_mean) of bone marrow was significantly lower in the atherogenic diet group (428.44 [IQR 399.69–445.98] versus 461.34 (IQR 437.28–494.73), p = 0.021) compared with the control group (Table 2 and Fig. S2b). Liver tissue density was also significantly lower in the atherogenic diet group (66.07 [IQR 57.95–69.66]) compared with the control group (84.35 [IQR 78.73–88.06]; p < 0.0001) (Table 2 and Fig. S2b). These results indicate increased fat deposition in the bone marrow and livers of animals in the atherogenic diet group compared with the control group. Mean tissue density in the spleen and adipose tissues was not significantly different between groups (Table 2 and Fig. S2b).

Vascular [¹⁸F]FDG uptake was moderately to strongly associated with CT-quantified vascular calcification and [¹⁸F]FDG uptake in bone marrow and adipose tissues, but not in the spleen

Using pooled data from all animals, associations among vascular inflammation, vascular calcification, and metabolic activity of bone marrow, spleen, and adipose tissues were investigated using a Spearman rank test. Correlations with 0 ≤ r_s <|0.3| were deemed to be of low strength; correlations with |0.3|≤ r_s <|0.7| were deemed to be of moderate strength; and correlations with|0.7|≤ r_s ≤|1| were deemed to be strong. Using these criteria, vascular inflammation in all four evaluated aortic segments, quantified as [¹⁸F]FDG TBR_mean, was moderately correlated with vascular calcification (Fig. 3a and b; r_s = 0.66 with ABD, r_s = 0.58 with SR, r_s = 0.60 with IR, and r_s = 0.58 with P2_R). Bone marrow TBR_mean was moderately to strongly correlated with inflammation and calcification in evaluated aortic segments (Fig. 3a and c; r_s = 0.75 with ABD, r_s = 0.72 with SR, r_s = 0.74 with IR, and r_s = 0.75 with P2_R, r_s = 0.64 with VC [measured by HU_max], and Fig. S3a); spleen [¹⁸F]FDG TBR_mean was only moderately correlated (Fig. 3a and Fig. S3b). Both VAT and SAT TBR_mean were found to be moderately associated with aortic [¹⁸F]FDG uptake (Fig. 3a and d–e; VAT: r_s = 0.59 with ABD, r_s = 0.44 with SR, r_s = 0.68 with IR, and r_s = 0.49 with P2_R; SAT: r_s = 0.50 with ABD, r_s = 0.44 with SR, r_s = 0.58 with IR, and r_s = 0.55 with P2_R) and with aortic calcification (Fig. 3a and Fig. S3a; VAT: r_s = 0.44; SAT: r_s = 0.36). Moderate correlations with metabolic activity were also observed in bone marrow (Fig. 3a and Fig. S3c; VAT: r_s = 0.63; SAT, r_s = 0.45), but only very weak correlations in the spleen. [¹⁸F]FDG TBR_mean of aortic segments were strongly correlated with each other, as were [¹⁸F]FDG TBR_mean of VAT and SAT (Fig. S4a and c). TBR_mean of bone marrow and spleen were only moderately correlated (Fig. S4b).

Fig. 3.

Correlation of vascular positron emission tomography–computed tomography (PET/CT) measurements with metabolic activity of lymphoid organs, and adipose tissues in crab-eating macaques. (a) Correlation matrix illustrating Spearman correlation coefficients (r_s) among 2-deoxy-2-[fluorine-18]fluoro-D-glucose ([¹⁸F]FDG) uptake (measured by target-to-background ratio, TBR_mean) in selected regions of interest (abdominal aorta [ABD], suprarenal aorta [SR], infrarenal aorta [IR], partial aortic section near renal arteries, [P2_R], bone marrow [BM], spleen, subcutaneous adipose tissue [SAT], visceral adipose tissue [VAT], and vascular calcification [VC] measured by HU_max). Strong, |0.7|≤ r_s ≤|1|; Moderate, |0.3|≤ r_s <|0.7|; Weak, 0 ≤ r_s <|0.3|. (b–e) Pairwise correlation between vascular inflammation (ABD, SR, IR, and P2_R) and vascular calcification (VC, b), bone marrow metabolic activity (BM, c), and adipose tissue metabolic activity (visceral adipose tissue [VAT, d] and subcutaneous adipose tissue [SAT, e], respectively). Each point represents an individual animal. Solid lines indicate the linear regression fits, and dashed lines represent 95% confidence intervals. Asterisks indicate statistical significance: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Discussion

NHPs are well-studied models of CVD and atherosclerosis. Although experiments in NHPs are generally more costly and labor intensive than investigations in smaller non-primate mammals, the higher genetic, metabolic, and physiologic similarities between NHPs and humans make NHP models so far indispensable for the study of atherogenesis [76]. Among the various NHP models of atherosclerosis, crab-eating macaques have been most well-studied [3, 11, 13, 15, 21]. Reports of naturally occurring and diet-induced atherosclerosis in crab-eating macaques emerged in the late 1960s and early 1970s [15]. Since then, many studies [27] have used crab-eating macaques to investigate the effect of diet [16, 26], biological sex [25], social status and associated behavior [14, 23, 25], and stress [8, 23] on the development of atherosclerosis.

Noninvasive multimodal imaging is routinely used for the clinical evaluation of patients with atherosclerosis and CVD and as a valuable tool for cardiovascular research. In vivo imaging not only complements other common readouts (e.g., blood biomarkers or histopathology) but has several intrinsic advantages. Unlike histopathologic sampling (via directed biopsies or at necropsy, typically cross-sectional in nature), longitudinal noninvasive imaging of the same clinical or research subject enables reduction in the number of study subjects needed to achieve adequate statistical power and captures intra-individual variation over time. Compared with blood biomarkers, imaging can support the characterization of pathobiology in an organ- or tissue-specific manner simultaneously in different organs and tissues; measurement of a signal of disease in a point of and across space and time is particularly useful when evaluating complex multi-organ system diseases such as atherosclerosis. Despite these attractors, a robust methodology to quantify imaging biomarkers of atherosclerotic disease in an NHP model has not been established.

Noninvasive imaging characterization of NHP models has been less common than in smaller non-primate animal models of atherosclerosis. While ultrasound has been used more often in the NHP model (32—36, 41), very few studies have used CT, alone or in combination with PET, or magnetic resonance imaging (MRI), either to characterize vascular disease or to establish safety for new imaging probes or novel drugs [38, 39, 55, 77, 78]. To our knowledge, only two other studies in the literature have used [¹⁸F]FDG PET/CT in a crab-eating macaque model of atherosclerosis [37, 38]. In a total of 17 crab-eating macaques, some fed an atherogenic diet and some fed a regular chow diet, Iwaki et al. [38] used [¹⁸F]FDG PET/CT at either 7 and 12 (early) or 18 and 24 (late) mo after atherogenic diet initiation. In crab-eating macaques fed an atherogenic diet (n = 13), and compared with those fed a chow diet (n = 3), Di Cataldo et al. [37] used [¹⁸F]FDG PET and [¹¹C]PK11195 PET, MRI, and ultrasound to investigate the association between cardiovascular inflammation and immune system activation and/or neuroinflammation. Similarly to our results, these studies also observed an increase in [¹⁸F]FDG PET vascular uptake in NHPs fed an atherogenic diet (for a comparable amount of time) as compared to controls. Comparison to these studies reveals some limitations of this work, most notable in its secondary and retrospective nature and in the limited number subjects. For example, we were unable to carefully balance age and sex between the atherogenic and chow diet groups (Table 1 and Table S2). Notably, macaques in the atherogenic diet group were all male and comparatively older versus the chow diet group; it should be acknowledged that these recognized risk factors (older age, male sex) may confound the effect of diet alone. These differences obviously limit any causal inference/assumptions as well as more comprehensive analyses adjusting for covariates; however, even the “mixed” potential combination (of an atherogenic diet and known risk factors) is associated with an increase in [¹⁸F]FDG uptake in the atherogenic diet group. In addition, we were only able to collect a limited set of blood hematologic parameters (white blood cell counts and frequency, Table S3), as opposed to more comprehensive blood biomarkers (for example including lipids and hsCRP) included in the other studies. Finally, although we provide histopathological confirmation of atherosclerosis developing in the arteries of macaques fed an atherogenic diet, caution is needed in the interpretation of these results. In 12 out of the 15 macaques fed an atherogenic diet, specimens for histopathology were collected at end of virus exposure experiments, some after therapeutic intervention. While the relationship between virus exposure, treatment and development or acceleration of atherogenesis is not known, a similar degree of atherosclerosis was observed in almost all macaques fed an atherogenic diet (non-exposed versus exposed, and untreated versus treated); nonetheless, this confounder cannot be ruled out. Finally histopathological data were not available for the chow group and qualitative or quantitative comparisons of histopathological hallmarks of atherosclerosis between the two groups were not possible.

Comparison to these studies also reveals several strengths: our study used more animals than other studies (a total of 27 compared with 17 [38] and 16 [37] NHPs, and, most importantly, included methodologic development of robust quantification of [¹⁸F]FDG uptake in the macaque vasculature, bone marrow, spleen, and adipose tissues. We also provided a quantification of vascular calcification, an important hallmark of atherosclerotic plaques, using CT in this model. Acknowledging limitations, these data may provide a solid basis for future, adequately powered, prospective studies in NHP models of atherosclerosis. As the use of PET/CT is further validated in the NHP model, and improved by the application of optimized acquisitions and analysis protocols (i.e., gated PET or partial volume correction techniques), we predict utility for this methodology to be leveraged to investigate how other factors (such as age, sex, or stress) may influence cardiovascular inflammation and immune activation in NHP models, and how their deleterious cardiovascular effects may be mitigated by therapeutic interventions.

Conclusions

Our methodology constitutes a significant step forward and suggests the expansion of PET/CT-enabled quantification (with [¹⁸F]FDG or other, more specific tracers) of atherosclerosis in the macaque or other NHP models. Investigation of any association between PET/CT-imaging features and other recognized biomarkers (e.g., highly sensitive C-reactive protein) was not possible in this retrospective work; with further validation, these quantitative readouts may be used (alone or in combination with other multiparameter noninvasive imaging modalities, such as MRI) to complement recognized or novel biomarkers (both in space and across time) to better characterize vascular disease due to atherosclerosis in valuable NHP models of disease.

Abbreviations

[¹⁸F]FDG

2-Deoxy-2-[fluorine-18]fluoro-D-glucose

ABD

Abdominal aorta

AS

Ascending aorta

CNN

Convolutional neural network

CVD

Cardiovascular disease

DL

Deep learning

FPN

Feature pyramid network

HU

Hounsfield units

HU_max

Maximum HU value

HU_mean

Mean HU value

HU_ref

Reference HU value

IQR

Interquartile range

IR

Infrarenal abdominal aorta

IVC

Inferior vena cava

MRI

Magnetic resonance imaging

NHPs

Nonhuman primates

P1_A

Partial aortic section at aortic arch

P2_R

Partial aortic section near renal arteries

P3_IB

Partial aortic sections at iliac bifurcation

PET/CT

Positron emission tomography–computed tomography

ROIs

Regions of interest

SARS-CoV-2

Severe acute respiratory syndrome coronavirus 2

SAT

Subcutaneous adipose tissues

SD

Standard deviation

SR

Suprarenal abdominal aorta

SUV_max

Maximum standardized uptake values

SUV_mean

Mean standardized uptake values

TBR_mean

Mean target-to-background ratio

T

Thoracic aorta

VAT

Visceral adipose tissues

VC

Vascular calcification

Author contributions

H.Y., H.W., V.M., and C.C. performed the conceptualization, study design, data analysis, interpretation, and manuscript writing. R.B., C.B., and P.J.S. performed data acquisition. G.W., A.C., T.C.F., D.H.O., I.C., and J.H.K. contributed critical review of the data analysis and manuscript. All authors read and approved the final manuscript.

Funding

This research was supported in part by the Intramural Research Program of the National Institutes of Health (NIH), under Contract No. HHSN272201800013C. This work was supported in part through a Laulima Government Solutions, LLC, prime contract with the National Institute of Allergy and Infectious Diseases (Contract No. HHSN272201800013C). H.Y., H.W., R.B., C.B., P.J.S., G.W., A.C., and V.M. performed this work as employees or affiliates of Laulima Government Solutions, LLC. J.H.K. and C.C. performed this work as employees of Tunnell Government Services (TGS), a subcontractor of Laulima Government Solutions, LLC, under Contract No. HHSN272201800013C. This work was also supported in part with federal funds from the National Cancer Institute (NCI), National Institutes of Health (NIH), under Contract No. 75N910D00024 (I.C.). The contributions of the NIH authors were made as part of their official duties as NIH federal employees, are in compliance with agency policy requirements, and are considered Works of the United States Government. The views and conclusions contained in this document are those of the authors and should not be interpreted as necessarily representing the official policies, either expressed or implied, of the U.S. Department of Health and Human Services or of the institutions and companies affiliated with the authors, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval

All experiments were performed at the Integrated Research Facility at Fort Detrick (IRF-Frederick), which is accredited by AAALAC International. The IRF-Frederick is part of the National Institutes of Health (NIH) National Institute of Allergy and Infectious Diseases (NIAID). All animal procedures were approved by the NIAID Animal Care and Use Committee (ACUC) and conducted in compliance with the Animal Welfare Act regulations, Public Health Service policy, and the Guide for the Care and Use of Laboratory Animals (Eighth Edition). This study utilized animals that were subsequently assigned to unrelated severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) studies. Leveraging existing baseline (pre-exposure) imaging data from these animals adheres to the principles of reducing the number of animals used for research, as part of the "Three Rs” (Replacement, Reduction, and Refinement) [79].

Consent for publication

Not applicable.

Competing interests

DHO is a co-founder and Managing Member of Pathogenuity LLC, a consultancy whose services include providing expert guidance on nonhuman primate study design. In the last 24 months he has consulted for Moderna and University of Southern California. He has received unrestricted funding and in-kind equipment donations from Heart of Racing. He also receives funding from Inkfish, National Institutes of Health, and Centers for Disease Control and Prevention.