Semaglutide reduces vascular inflammation investigated by PET in a rabbit model of advanced atherosclerosis.

BACKGROUND AND AIMS: The objective of this study was to investigate the effects of semaglutide, a long acting glucagon-like peptide-1 receptor agonist, on atherosclerotic inflammation and calcification using a multimodality positron emission tomography and computed tomography (PET/CT) approach.
METHODS: Atherosclerotic New Zealand White rabbits were randomized to an intervention- (n = 12) or placebo group (n = 11) receiving either semaglutide or saline-placebo. PET/CT imaging was done before and after 16-weeks of intervention. Three different radiotracers were used: [64Cu]Cu-DOTATATE for imaging of activated macrophages, [18F]FDG imaging cellular metabolism and [18F]NaF PET visualizing micro-calcifications. Tracer uptake was quantified by maximum standardized uptake value (SUVmax) and target-to-background-ratio (TBRmax). Animals were euthanized for autoradiographic imaging and histological analyses.
RESULTS: A reduction in activated macrophage tracer-uptake was observed in the semaglutide group (SUVmax: p = 0.001 and TBRmax: p = 0.029). When imaging cellular metabolism, an attenuation of SUVmax and TBRmax was observed in the semaglutide group (p = 0.034 and p = 0.044). We found no difference in uptake of the micro-calcification tracer between the two groups (SUVmax: p = 0.62 and TBRmax: p = 0.36). Values of macrophage density in the vessel wall were significantly correlated with SUVmax values of the activated macrophage (r = 0.54, p = 0.0086) and cellular metabolism tracers (r = 0.51, p = 0.013).
CONCLUSIONS: Semaglutide decreased vascular uptake of tracers imaging activated macrophages and cellular metabolism but not micro-calcifications compared to a saline placebo. This supports the hypothesis that semaglutide reduces atherosclerotic inflammation by means of decreased activated macrophage activity.

Introduction

Atherosclerosis is a chronic inflammatory vascular disorder. It is a hallmark of coronary artery disease and stroke, the leading causes of death in the western part of the world [1,2]. Atherosclerotic disease progression involves a multitude of complex pathological processes such as imbalance of cholesterol homeostasis and chronic low-grade arterial inflammation with macrophage accumulation, ultimately leading to endothelial dysfunction [3,4]. An increasing amount of evidence suggests that incretin based therapies reduce the risk of major adverse cardiac events (MACE) in high risk diabetic populations [5]. Analogues of the human incretin hormone glucagon-like-peptide-1 (GLP-1) are now routinely used in patients with type 2 diabetes for glycemic control. Semaglutide, a long acting GLP-1 receptor agonist (GLP-1RA), has demonstrated reduction of MACE in a large cardiovascular outcome trial [6]. This risk reduction effect is, in part, speculated to be driven by anti-inflammatory mechanisms, independent of blood glucose lowering effects and weight reduction.

Multimodality molecular imaging using positron emission tomography combined with computed tomography (PET/CT) is a promising in vivo hybrid modality in the study of atherosclerosis, making simultaneous anatomic localization of regions of interest and quantification of disease activity possible [[7], [8], [9]] [[7], [8], [9]] [[7], [8], [9]]. Multiple different radio-ligand tracers have been investigated in relation to atherosclerotic disease. The most frequently used tracer continues to be fluorine-18-labeled fluorodeoxyglucose ([18F]FDG) despite its inherent limitations, such as lack of cellular specificity [10,11]. One of the emerging radio-ligand tracers in the field of atherosclerosis imaging is the somatostatin receptor-subtype 2 (SSTR2) targeting tracer [1,4,7,10-tetraazacyclododecane- N,N′,N’’,N’’’-tetraacetic acid]-D-Phe1,Tyr3-octreotate (DOTATATE), routinely used in neuroendocrine tumor imaging [12,13]. Somatostatin receptor imaging (SRI) utilizes the up-regulation of SSTR2 on the membrane of activated M1 macrophages, which is an essential part of the inflammatory atherosclerotic process [14,15]. When coupled with either copper-64 or gallium-68, DOTATATE has been demonstrated to offer a complementary role to [18F]FDG in both the pre-clinical and clinical evaluation of atherosclerosis [16]. 18F-sodium fluoride ([18F]NaF) is another radio-tracer previously used for the detection of bone metastases but has also shown to be able to localize active pathological microcalcifications in vascular tissue [17].

The aim of this study was to evaluate the potential effects of the GLP-1RA semaglutide on the atherosclerotic and inflammatory processes in the aorta using in vivo multiparametric molecular PET/CT imaging with [64Cu]Cu-DOTATATE, [18F]FDG and [18F]NaF in a non-diabetic rabbit model of advanced atherosclerosis. Furthermore, we utilized immunohistochemical analysis to quantify the degree of macrophage infiltration in the aortic-wall.

Materials and methods

Data availability statement

Data and analytical methods can be made available upon reasonable request to the authors.

An extended Methods section can be found in the Supplementary Data.

Experimental design

All animal experiments were performed with appropriate approval from the Animal Research Committee of the Danish Ministry of Justice and in accordance with the guidelines in Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes.

Twenty-four female SPF New Zealand White rabbits (11 weeks old) were acquired from Lidkjöping Kaninfarm (Lidkjöping, Sweden) and put on an ad libitum high cholesterol diet (Research Diets, New Brunswick, USA). Two weeks after initiation of the high cholesterol diet, all animals underwent denudation of the abdominal aorta. The denudation procedure was repeated 3 weeks later (Fig. 1A). The denudation procedures were performed under anesthesia with intramuscular Xylazine (5 mg/kg) and Ketamine (35 mg/kg).

Fig. 1

Study timeline, body weight and fasting blood glucose.

(A) Timeline of the experimental procedures. (B) High cholesterol diet induced weight gain was significantly reduced over time in the semaglutide group as compared to the placebo group. Data are presented as group means, Whiskers indicate SEM. (C) There was no difference in the fasting blood glucose between the two groups. Mixed effect model with repeated measures: ***p < 0.001. Analysis of co-variance (ANCOVA) between the two groups at follow-up: ns = non-significant, p = 0.18. CT = computed tomography; HF = high fat; PET = positron emission tomography; [64Cu]DOTATATE = copper-64-labeled DOTATATE; [18F]FDG = fluorine-18-labeled fluorodeoxyglucose; [18F]NaF = fluorine-18-labeled sodium-fluoride.

One rabbit was euthanized after the second denudation procedure due to post-operative complications.

Four months after the second denudation procedure, the animals were randomly allocated to either an intervention group (n = 12) or a placebo group (n = 11). Animals received bi-weekly subcutaneous injections of either semaglutide (Novo Nordisk, Måløv, Denmark) or isotonic saline solution for 16 weeks. The dosing regimen in the intervention group was set up as a gradual escalation of the semaglutide dose, reaching a maximal dose of 44 μg/kg/week, to avoid potential early termination of the animals due to side effects and rapid weight loss (Supplementary Figure I). Total cholesterol was measured before initiation of the intervention, and after the total intervention period, on a Reflotron® Plus system (Roche Diagnostics, Risch-Rotkreuz, Switzerland).

PET/CT imaging and analysis

PET/CT imaging procedures were performed on a Siemens Inveon® Small Animal Scanner (Siemens Medical Systems, PA, USA).

All animals underwent baseline PET imaging with [18F]FDG and [64Cu]Cu-DOTATATE, 3 and 2 days prior to randomization, respectively. Follow up scans with [18F]FDG, [18F]NaF and [64Cu]Cu-DOTATATE were performed after the 16 weeks of intervention on day 1, 3 and 5, post intervention, for both groups. PET/CT with [18F]NaF was only carried out at follow-up due to the lack of availability at the time of the baseline PET/CT scans.

For all imaging procedures, anesthesia was induced with intramuscular injection of Xylazine (5 mg/kg) and Ketamine (35 mg/kg). The animals were then kept anaesthetized with 2–3% Sevoflurane mixed with 35% O2 and ambient air.

Reconstructed images were analyzed with Inveon Research Workplace software (Siemens Medical Systems, PA, USA). Image analysis was conducted by drawing regions of interest (ROIs) on axial plane CT images of the aorta from the right renal artery to the iliac bifurcation, with 5 mm intervals. On average, 38 axial ROIs were drawn on each aorta. All axial ROIs were then interpolated to form a single volume of interest (VOI), covering the aorta and a single maximum standardized uptake value (SUVs) were calculated from the interpolated VOI. Five consecutive ROIs were drawn in the inferior vena cava to allow for target-to-background (TBR) estimations. Image analyses was conducted blinded to group allocation.

Ex vivo analyses

Following the last imaging procedure, the rabbits were euthanized by intravenous injection of sodium-pentobarbital (120 mg/kg), and the aortas were harvested for further analysis. Autoradiography of the abdominal aorta was performed for all three tracers (n = 2 per group for each tracer).Formalin-fixed, paraffin-embedded sections of the aortas (n = 23) were stained as described earlier and using manufacturer's protocols with hematoxylin and eosin stain (HE), Masson's Trichrome (MT), Alizarin Red and immunohistochemical (IHC) staining for RAM-11 [18].The deep learning segmentation software Zen Intellesis (Carl Zeiss Microscopy, Oberkochen, Germany) was used to quantify the ratio (%) between macrophage-rich area and total vessel area from RAM-11 stained cross-sections of the aorta.

Statistical analysis

Normally distributed data are presented as mean ± SEM, if not stated otherwise. Statistical analyses were performed using the software environment R ver. 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria) with Rstudio ver. 1.3.1056 (RStudio, PBC, Boston, MA, USA) and GraphPad Prism ver. 9.2.0 (GraphPad Software, San Diego, USA).

Uptake values from PET scans using [64Cu]Cu-DOTATATE, [18F]FDG, and fasting blood glucose values at follow-up were compared between the two groups using analysis of co-variance (ANCOVA), with baseline uptake values and group-allocation as co-variates. Hypothesis testing between the two groups on image data from PET scans at baseline and [18F]-NaF at follow-up, as well as RAM-11 quantification of histological sections, was performed using un-paired Student's t-test. Difference in weight between the placebo and the semaglutide group during the 16-week treatment period was analyzed with the use of a linear mixed model with a repeated measures approach, where time and group were specified as fixed effects, and “rabbit” as a random effect. A sample size calculation based on a 2-sample unpaired t-test (2-sided) was conducted prior to the initiation of the study using a power of 80% and an α level of 5%, based on an assumed TBRmax value of 1.48 ± 0.23 from a previous study investigating the effects of a statin on aortic atherosclerosis using [18F]FDG PET, yielding a group-wise sample size of 11 to detect a 20% difference [19].

Reproducibility of SUVmax and TBRmax measurements was assessed using the intraclass correlation coefficient (two-way mixed effects model with absolute agreement) with corresponding 95%-confidence intervals (95% CI), and Bland-Altman plots. Thirty randomly selected PET/CT scans (∼26% of the total dataset) were used for reproducibility measurements, and assessed blinded to group allocation.

Correlation analysis was performed using simple linear regression and Spearman's correlation coefficient.

For all hypothesis testing, a two sided p-value < 0.05 represents statistical significance.

Results

Body weight, blood glucose and cholesterol

During the 16-week intervention period, the animals in the semaglutide group gained weight at a slower rate compared to the placebo group (p < 0.0001) (Fig. 1B).Fasting blood glucose in these non-diabetic rabbits did not differ between groups at baseline (p = 0.5) or follow-up (p = 0.18) (Fig. 1C).

Fasting total cholesterol levels were measured in both groups at baseline and follow-up. Almost all values in both groups were above the measuring range of the used equipment of 12.9 mmol/L (Supplementary Figure II).

Histological analysis of the aorta

Presence of atherosclerotic disease was confirmed by histological evaluation of the abdominal aorta at follow-up. HE stains showed aortic wall thickening with cholesterol clefts, and in some areas of necrotic tissue. Masson's trichrome stains revealed clear deposition of extra cellular matrix components in the intima media of the aortic wall. Staining aortic sections with Alizarin Red showed abundant calcification. Immunohistochemical staining with RAM-11 showed heavy build-up of macrophages in the atherosclerotic lesions of both groups (Fig. 2A). The above characteristics confirm that progression of atherosclerotic hallmarks was present in both groups with resemblances to the atherosclerotic changes observed in humans. When measuring macrophage density (%-total area) on RAM-11 stained cross-sections of the aorta, we observed a trend towards less deposition of macrophages in the semaglutide group (semaglutide: 11.49% vs. placebo: 15.90%, p = 0.093) (Fig. 2B).

Fig. 2

Histological characterization of atherosclerotic changes of the aorta.

(A) Sections of the infra-renal abdominal aorta explanted from both the semaglutide and the placebo group and stained with HE, Masson's trichrome (extracellular matrix and smooth muscle cells), Alizarin red (calcifications) and RAM-11 (macrophages). (B) Quantitative analysis of macrophage area of the aortic wall based on segmentation of RAM-11 stained sections (n = 23). Colors in the exemplified segmented image represent: RAM-11 staining (red), vessel-wall (teal) and background (orange). (C) Scatter-plots show correlations of aortic inflammation measured as [64Cu]Cu-DOTATATE and [18F]FDG uptake versus the %-macrophage content of the vessel wall quantified on histological RAM-11 sections. HE = hematoxylin and eosin, other abbreviations as in Fig. 1. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Correlations of tracer uptake and aortic macrophage content

Aortic [64Cu]Cu-DOTATATE and [18F]FDG SUVmax values versus %-macrophage content of the vessel wall by RAM-11 immunohistochemical staining showed a moderate positive correlation with SUVmax ([64Cu]Cu-DOTATATE: r = 0.54; 95% CI: 0.14–0.78; p = 0.0086, and [18F]FDG: r = 0.51; 95% CI: 0.11–0.77; p = 0.013) (Fig. 2C). The same was observed for correlations using TBRmax, albeit r-values were smaller for [64Cu]Cu-DOTATATE (r = 0.43; 95% CI: 0.01–0.72; p = 0.04) than for [18F]FDG (r = 0.61; 95% CI: 0.26–0.82; p = 0.001) compared to correlations using SUVmax (Supplementary Figure III).

Inflammatory macrophage activity using [64Cu]Cu-DOTATATE uptake

PET-uptake values are summarized in Table 1. For all three tracers, the intraclass correlation coefficient for SUVmax (0.99; 95% CI: 0.992–0.998) and TBRmax (0.87; 95% CI: 0.723–0.942) measurements was excellent and good, respectively. Bland-Altman plots of agreement also showed good reproducibility with only a small fixed bias between both SUVmax and TBRmax of the three different tracers, although the bias was largest for the TBRmax measurements (Supplementary Figure IV).

Table 1

Positron emission tomography uptake values.

		SUVmax, g/ml				TBRmax
		Baseline	p-value, baseline	Follow-up	p-value, follow-up	Baseline	p-value, baseline	Follow-up	p-value, follow-up
[64Cu]Cu-DOTATATE	Semaglutide	7.59 ± 0.48	0.13	5.83 ± 0.24	0.001**	6.98 ± 0.56	0.22	4.71 ± 0.34	0.029*
	Placebo	6.69 ± 0.28		7.10 ± 0.33		6.04 ± 0.51		5.96 ± 0.45
[18F]FDG	Semaglutide	2.63 ± 0.12	0.29	2.49 ± 0.13	0.034*	3.61 ± 0.25	0.57	4.44 ± 0.26	0.044*
	Placebo	2.86 ± 0.19		2.99 ± 0.15		3.39 ± 0.29		5.09 ± 0.13
[18F]NaF	Semaglutide	–	–	4.15 ± 0.30	0.62	–		4.36 ± 0.45	0.36
	Placebo	–	–	3.92 ± 0.34		–		4.94 ± 0.42

Mean SUVmax and TBRmax ± SEM, determined by PET/CT for all three tracers at both baseline and follow-up. p-values between the two groups for [64Cu]Cu-DOTATATE and [18F]FDG uptake-values at baseline are calculated using Student's t-test and using analysis of co-variance at follow-up with the baseline value as a covariate. For [18F]NaF, p-values are calculated using Student's t-test.

[64Cu]Cu-DOTATATE = copper-64-labeled DOTATATE; [18F]FDG = fluorine-18-labeled fluorodeoxyglucose; [18F]NaF = fluorine-18-labeled sodium-fluoride; SUVmax = maximum standardized uptake values; TBRmax = maximum target-to-background-ratio.

At baseline, the [64Cu]Cu-DOTATATE uptake in the aorta measured as SUVmax and TBRmax showed a similar degree of macrophage inflammation in the semaglutide group and the placebo group (Fig. 3A) (SUVmax: p = 0.13 and TBRmax: p = 0.22).

Fig. 3

Effect of semaglutide on aortic wall uptake of [64Cu]DOTATATE, [18F]FDG and [18F]NaF.

Figures show representative sagittal and coronal fused PET/CT, CT-only and autoradiographic images of the aorta in both groups using (A) [64Cu]Cu-DOTATATE, (B) [18F]FDG and (C) [18F]NaF together with scatter plots of both SUVmax and TBRmax uptake values for both groups at baseline and follow-up. Hypothesis testing on data from [64Cu]Cu-DOTATATE and [18F]FDG PET was done using analysis of co-variance, and using Student's t-test on [18F]NaF data: *p < 0.05, **p < 0.01. SUV-bw = standardized uptake value corrected by body weight; SUVmax = maximum of the standardized uptake value; TBRmax = maximum target-to-background ratio; other abbreviations as in Fig. 1, Fig. 2.

After the 16-week intervention, we observed that the semaglutide group had a significant decrease in the uptake of [64Cu]Cu-DOTATATE compared to the placebo group when controlling for baseline values (SUVmax: p = 0.001) (Table 1). This corresponded to a decrease in [64Cu]Cu-DOTATATE uptake, measured as SUVmax, in 11 out of 12 animals in the semaglutide group and a mean decrease in uptake of 23% when comparing to baseline. This was opposed to the placebo group where the SUVmax increased in 7 out of 11 animals, corresponding to a mean increase of 6% (Fig. 3A).

When calculating TBRmax, we also observed a decrease in uptake in 8 out of 12 in the semaglutide group compared to only 3 out of 11 in the placebo group (p = 0.029) (Table 1). In the semaglutide group, this corresponded to a mean 33% decrease from baseline. In the placebo group, there was a mean increase of 1% from baseline to follow-up (Fig. 3A). The autoradiographic images showed a uniform distribution of [64Cu]Cu-DOTATATE throughout the explanted aortas with streaks of intense uptake in both groups (Fig. 3A).

Vascular inflammation imaging using [18F]FDG uptake

The [18F]FDG uptake at baseline was comparable in the two groups, when measuring SUVmax and TBRmax (SUVmax: p = 0.29 and TBRmax: p = 0.57) (Fig. 3B).

At follow-up, [18F]FDG uptake was attenuated in the semaglutide group, where we observed a minor, but significant, decrease in uptake in 7 out of 12 animals (Fig. 3B). This was compared to a minor increase in 7 out of 11 animals in the placebo group, when measuring SUVmax (p = 0.034).

When calculating TBRmax, there was an increase in both groups when comparing baseline to follow-up of 23% and 50% in the semaglutide and placebo group, respectively (Fig. 3B). The difference in uptake between the groups was significant at follow-up when controlling for baseline values (p = 0.044).

Microcalcification visualized with [18F]NaF at follow-up

The [18F]NaF scans detected no difference in the degree of microcalcification of the aorta between the two groups at follow-up, neither when measuring the SUVmax (p = 0.62) nor the TBRmax (p = 0.36) (Fig. 3C). Autoradiography revealed a patchy distribution of the radiotracer in the aortas of both groups (Fig. 3C).

Discussion

The reduction in MACE seen in the large cardiovascular outcome trials after treatment with GLP-1RA has been suggested to be mediated by mechanisms reducing the extent of atherosclerosis and atherogenic inflammation [5,6,20]. In this study, we found that treatment with the long acting GLP-1RA, semaglutide, affects atherosclerotic inflammation in the aorta of a rabbit model with induced atherosclerosis. This was for the first time done using a multi tracer in vivo PET imaging approach that can readily be translated into human studies.

The radiotracer [64Cu]Cu-DOTATATE has previously been used to evaluate atherosclerosis in 2 cross-sectional studies where it was shown that uptake in carotid plaques correlated with markers of activated macrophages [21,22], but this is the first study to use a somatostatin receptor tracer for evaluation of atherosclerotic therapy. Several other groups have also used DOTATATE for plaque imaging but coupled to 68Galium (68Ga), where it was observed to be a promising cell specific marker of SSTR2 expression in activated M1 macrophages [16,23]. The 64Cu labelling offers a shorter positron range compared to 68Ga labeled DOTATATE, which essentially results in less blurring of the PET images due to better spatial resolution. This is crucial when imaging smaller structures e.g. coronary arteries. When investigating the effects of semaglutide on macrophage mediated atherosclerotic inflammation using [64Cu]Cu-DOTATATE, we found a significantly reduced uptake in the aorta of the semaglutide treated group compared to the placebo treated animals. This indicates that the amount of inflammatory M1 macrophages in the vessel wall is decreased, and corresponds somewhat to the results from the quantitative analysis of macrophage content on histological RAM-11 sections of the aorta. Analysis of macrophage polarization into either M1 or M2 sub-populations might have provided a more detailed characterization. A correlation between uptake of [64Cu]Cu-DOTATATE and macrophage content in the vessel wall was also observed. These findings are supported by evidence suggesting that treatment with other GLP-1RA hinders recruitment and adhesion of monocytes and macrophages in the vessel wall of atherosclerotic mice [24,25]. These studies have also shown a decrease in expression of VCAM-1 and ICAM-1 when treating ApoE-/- mice and human endothelial cells with a GLP-1RA. This is observed along with a reduction of pro-inflammatory markers like TNF-α, IL-6, NF-kB and MCP1 [[24], [25], [26]]. Anither study used a rabbit model of atherosclerosis in the investigation of the effect of a GLP-1RA (Lixisenatide). The authors found a decreased macrophage content and less necrotic tissue formation of the vessel wall in the GLP-1 RA group compared to a saline placebo [27]. In two sub-studies of a larger clinical trial, [64Cu]Cu-DOTATATE was used to investigate the effects of liraglutide on vascular inflammation in a diabetic population without prior MACE. In these studies, a decrease in uptake was observed in the treatment groups from baseline to follow-up, but no difference was found when comparing the treatment groups to the control groups [28,29]. These findings suggest that the effects of GLP-1RA on vascular inflammation might only be achievable in populations with high cardiovascular risk. Other antidiabetic drugs have also been investigated with regards to their potential benefit to cardiovascular risk and outcomes, and it is becoming apparent that both GLP-1RA and sodium-glucose cotransporter-2 inhibitors are superior in this regard to dipeptidyl peptidase-4 inhibitors [30]. [18F]FDG PET is a well-known reproducible measure of overall inflammation in the arterial wall, and is associated with known risk factors of CVD [31]. Although [18F]FDG PET is also known to be influenced by other contributing factors like glucose metabolism of other active cells in the arterial wall and local microcirculation, it has previously been used successfully in drug efficacy trials showing a reduction in uptake when treating with drugs known to reduce cardiovascular risk [32]. One study used [18F]FDG PET to assess the effect of the short-acting GLP-1RA liraglutide on vascular inflammation in a cohort of type-2 diabetics with low to moderate cardiovascular risk [33]. Here, no effect of liraglutide on vascular inflammation in that particular cohort of patients was found. It can be speculated that the discrepancy in outcomes between the aforementioned study and the present might be due to the fact that the cardiovascular risk profile of the included patients was very heterogenous, as opposed to the present study where all the animals displayed quite similar and advanced atherosclerosis.

The [18F]FDG PET scans in the present study showed a significant attenuation of the uptake measured as SUVmax in the wall of the aorta in the semaglutide group compared to placebo, reflecting a decreased glucose cell metabolism. When calculating TBRmax values, we observed an increase in both groups from baseline to follow-up, albeit with a smaller increase in the treated group (Fig. 3B). Since we did not observe the same increase in SUVmax values from baseline to follow-up, it can be speculated that the observed increase in TBRmax is due to decreased blood-pool activity at follow-up. This reduced blood-pool activity might be the result of multiple factors like change in e.g. glomerular filtration rate, difference in venous blood uptake and blood flow [34]. Time from [18F]FDG injection to PET imaging and injected activity (MBq) showed no significant difference between the groups and time points (Supplementary Table I).

We did not observe a reduction in plasma glucose in the semaglutide group compared to the saline placebo group at follow-up, but this is consistent with other studies using non-diabetic rabbit models to evaluate the effect of GLP-1RA [27,35].This also adds to the evidence of the anti-inflammatory effect being independent of the glucose lowering effect. An attenuation in body weight gain was observed throughout the treatment period in the semaglutide treated group (Fig. 1B). This was expected, as multiple prior studies have reported this effect previously in both humans and animals [26,36]. The groups in this study were not controlled according to weight, and it should therefore be mentioned that some of the observed effects might be attributable to the lack of weight gain in the semaglutide group. However, others have investigated the effects of GLP-1RA in a weight controlled setting and still observed an anti-atherosclerotic effect independent of the body-weight [25,26].

The results from the [18F]FDG and [64Cu]Cu-DOTATATE PET scans are somewhat in contrast to our results from the [18F]NaF PET scans visualizing active microcalcifications. Here, we found no difference in uptake of the radiotracer in the aorta of the 2 groups. It can be speculated that this might be because of the already chronic nature of the atherosclerotic changes of the aorta developed in this model with inherent inability of the radiotracer to penetrate into the deeper layers of the calcification, or because semaglutide has no effect on active micro-calcification [17]. Of course it should also be taken into consideration that unlike the other two tracers used, we did not perform a baseline [18F]NaF scan, which limits the interpretability of this result.

It is debated in the literature in exactly which cell- and tissue types the anti-inflammatory effect takes place. Studies concerning the expression of the canonical GLP-1 receptor (GLP-1R) in endothelial cells (EC) and vascular smooth muscle cells (VSMC), especially in humans and primates, found no expression of GLP-1R [37,38]. In rodents, most often ApoE-/- mice, the presence of GLP-1R in both cell types was observed [25,39], although some controversy exists on whether the assays for detection are specific enough. In vitro studies on both human and rodent EC and VSMC seem to confirm a direct anti-atherogenic effect of GLP-1RA via reduction of reactive oxygen species (ROS) and induction of eNOS in EC, thereby corroborating endothelial dysfunction [[40], [41], [42]] [[40], [41], [42]] [[40], [41], [42]].

The effect of GLP-1 RA on development of atherosclerosis and inflammation has largely been investigated in mouse models. These are most often used in a context of a short-term induction of atherosclerosis and a short treatment period, although a few groups have succeeded in establishing small animal models of long-term exposure to atherosclerosis [43]. The rabbit model used in the current study presents a more chronic and severe form of atherosclerosis compared to the majority of small animal studies. This approach has multiple advantages, among these are larger animal-size, which enables precise non-invasive imaging of vascular structures like the aorta, and a lipid-metabolism that shares a closer resemblance to that of humans, albeit still with some differences [44].

Limitations

This study has inherent limitations. First, spatial resolution of PET when imaging small structures like the rabbit aorta or human coronary arteries, even though the pre-clinical scanner used in this study has significantly better PET resolution compared to clinical PET-scanners. Second, [64Cu]Cu-DOTATATE still needs to be validated in prospective cohort studies to establish a definitive relationship between the uptake of this radiotracer and robust cardiovascular outcomes. We only investigated female rabbits, but we acknowledge the importance that sex plays in the atherosclerotic process.With [18F]NaF, the close proximity of the aorta to the vertebrae also poses a challenge when measuring aortic-wall uptake. Also, we did not measure baseline [18F]NaF uptake, as with the other two tracers. Even though this rabbit model of advanced atherosclerosis produces atherosclerotic changes of the arterial wall similar to those in humans, it should be noted that it does not fully resemble the intricate composition of human plaques. Also, analysis was limited to the abdominal portion of the aorta.

Conclusions

Multimodality non-invasive molecular imaging with different PET tracers allows for in vivo evaluation of atherosclerotic inflammation and mode of action studies in this rabbit model of advanced atherosclerosis. Our present data supports the hypothesis that the risk reduction demonstrated by semaglutide in cardiovascular outcome trials is in part mediated by modulation of the activated M1 macrophages, in the context of advanced atherosclerotic disease.

Declaration of competing interests

Professor Andreas Kjaer, one of the initiators of the study, is an inventor/holds IPR on human use of [64Cu]Cu-DOTATATE in neuroendocrine tumor patients, and has received consultancy fees from Novo Nordisk. Rasmus S. Ripa owns shares in Novo Nordisk. All the other authors state that they have no relevant conflicts of interest to disclose.

Financial support

The authors are supported by funding from the under grant agreements no. 670261 (ERC Advanced Grant) and 668532 (Click-It), the , the , the Innovation Fund Denmark, the , , the Neye Foundation, the , the (grant 126), the Research Council of the , the Danish Health Authority, the Andreas Kjaer is a Lundbeck Foundation Professor.

CRediT authorship contribution statement

Jacob K. Jensen: Conceptualization, Methodology, collection of the data, Formal analysis, Writing – original draft, Writing – review & editing. Tina Binderup: Data collection, Formal analysis, Writing – original draft, Writing – review & editing. Constance E. Grandjean: Data collection, Formal analysis, Writing – original draft, Writing – review & editing. Simon Bentsen: data collection, interpretation, Writing – original draft. Rasmus S. Ripa: Conceptualization, Methodology, Writing – original draft, Writing – review & editing. Andreas Kjaer: Conceptualization, Methodology, Writing – original draft, Writing – review & editing.