Neoatherosclerosis development following bioresorbable vascular scaffold implantation in diabetic and non-diabetic swine

Background DM remains a risk factor for poor outcome after stent-implantation, but little is known if and how DM affects the vascular response to BVS. Aim The aim of our study was to examine coronary responses to bioresorbable vascular scaffolds (BVS) in swine with and without diabetes mellitus fed a ‘fast-food’ diet (FF-DM and FF-NDM, respectively) by sequential optical coherence tomography (OCT)-imaging and histology. Methods Fifteen male swine were evaluated. Eight received streptozotocin-injection to induce DM. After 9 months (M), 32 single BVS were implanted in epicardial arteries with a stent to artery (S/A)-ratio of 1.1:1 under quantitative coronary angiography (QCA) and OCT guidance. Lumen, scaffold, neointimal coverage and composition were assessed by QCA, OCT and near-infrared spectroscopy (NIRS) pre- and/or post-procedure, at 3M and 6M. Additionally, polarization-sensitive (PS)-OCT was performed in 7 swine at 6M. After sacrifice at 3M and 6M, histology and polymer degradation analysis were performed. Results Late lumen loss was high (~60%) within the first 3M after BVS-implantation (P<0.01 FF-DM vs. FF-NDM) and stabilized between 3M and 6M (<5% change in FF-DM, ~10% in FF-NDM; P>0.20). Neointimal coverage was highly heterogeneous in all swine (DM vs. NDM P>0.05), with focal lipid accumulation, irregular collagen distribution and neointimal calcification. Likewise, polymer mass loss was low (~2% at 3M, ~5% at 6M;P>0.20) and not associated with DM or inflammation. Conclusion Scaffold coverage showed signs of neo-atherosclerosis in all FF-DM and FF-NDM swine, scaffold polymer was preserved and the vascular response to BVS was not influenced by diabetes.


Introduction
Patients with diabetes mellitus (DM) are generally at risk for worse outcome after stentimplantation than patients without DM. [1] Due to the complex and multifactorial nature of the disease process, including metabolic abnormalities and vascular dysfunction, the vascular response to stent-implantation is generally impaired, complicating current stenting strategies. [2][3][4] The everolimus-eluting bioresorbable vascular scaffold (BVS) may offer advantages. It elutes everolimus in the first 3 to 6 months after implantation, inhibiting excessive neointimal growth [5] and starts losing structural integrity 3 months after implantation, [5] potentially enabling vascular function restoration. [6] Histology in healthy swine demonstrated that struts are covered at 28 days and resorbed around 3 years with minimal calcification and inflammation. [7,8] In selected patients from the ABSORB Cohort A (BVS 1.0) and B (BVS 1.1) trials, excellent results for treatment of coronary artery lesions were observed. [9,10] However, only 3%-20% of the study population suffered DM and no studies were performed in diabetic animals. Thus, little is known about the effect of DM on the vascular response to BVS. DM may cause inflammation, which could influence scaffold degradation by disregulated acid-base balance or body-temperature. [11] Animal models reflecting the impact of atherosclerosis and DM can be useful, as they allow us to study vascular responses and scaffold degradation in a more complex setting. [12] Moreover, swine can be rendered diabetic and in combination with an atherogenic diet they develop atherosclerosis comparable to humans. [13] Scaffolds can be placed in coronary arteries and in-vivo sequential intracoronary imaging can be performed by optical coherence tomography (OCT), polarization-sensitive (PS)-OCT and near-infrared spectroscopy (NIRS). After sacrifice, histology and gel permeation chromatography (GPC) can be performed to assess scaffold coverage and degradation of the polymer.
We examined the mechanistic and morphological aspects of the coronary response to BVS1.1 in DM and non-DM swine fed a fast-food diet (FF-DM, FF-NDM respectively) using longitudinal intracoronary imaging and histology.
for a detailed description). DM was induced by streptozotocin (140mg/kg iv, single dose) in 8 randomly selected male crossbred swine. [15] During streptozotocin injection, the swine were anesthetized with intramuscular azaperone (2 mg/kg, Stressnil, Janssen, Tilburg, The Netherlands), followed by intravenous thiopental (15 mg/kg, Nesdonal, Rhone Merieux, Lyon, France). The swine were housed in metabolic cages and were fed two fast-food (FF) meals a day during which they had access to food for one hour. The FF-diet is a diet containing 10% sucrose, 15% fructose, 25% (swine) lard, 1% cholesterol and 0.7% sodiumcholate (bile salts). The food intake was monitored for each animal separately and titrated to maintain growth at 1.5 kg/week. After 9 months, all 8 FF-DM and 5 FF-NDM received single 3.0x18.0mm Absorb BVS1.1 implants in 2, and 2 FF-NDM received single Absorb BVS1.1 implants in all 3 coronary arteries to ensure an even amount of scaffolds in FF-DM (N = 16) and FF-NDM (N = 16) (see S1 File for details about the BVS1.1). One day prior to BVS1.1-implantation the swine received 300 mg acetylsalicylic acid and a loading dose of 300 mg clopidogrel (Plavix, Sanofi). After an overnight fast, the swine were sedated using ketamine/ midazolam (20 mg/kg / 1 mg/kg i.m.) and atropine (1mg/30kg i.m.). After induction of anesthesia with thiopental (15 mg /kg i.v.; Nesdonal, Aventis), the swine were connected to a ventilator that administered a mixture of oxygen and nitrous oxide (1:2 [vol/vol]). Vascular access was obtained with an 8F vascular sheath in the carotid artery, 10.000 IU heparin was administered initially and thereafter 5000 IU of heparin was administered every hour. Anesthesia was maintained using 0.5-2.5 vol% isoflurane (Florence, Abbott Laboratories) as guided by hemodynamics and pain reflexes to ensure adequate analgesia and sedation. Antibiotic prophylaxis was administered by an intramuscular injection of 8 mL 200 mg/mL procaine-benzylpenicillin and 250 mg/mL streptomycin. After BVS-implantation, all swine were treated with clopidogrel (75mg) and acetylsalicylic acid (300mg) daily, until the end of the study. The latter also functions as analgesia during the post-operative recovery.
Sequential coronary imaging included QCA and OCT pre-, immediately, 3 and 6 months (M) post-implantation and NIRS pre-, 3M and 6M post-implantation. PS-OCT was performed in N = 3 FF-DM and N = 4 FF-NDM BVS at 6M. 3M imaging was included when pre and/or post-implantation imaging were available and 6M imaging was included when pre-and/or post-and 3M imaging were available. After the 3M imaging assessment, 3 FF-DM and 2 FF-NDM were sacrificed and after the 6M imaging assessment the remaining swine were sacrificed. After sacrifice, hearts were removed, the coronary tree dissected free and coronary arteries containing BVS randomized to histological (3M N = 5/10, 6M N = 12/22) or GPC analysis (3M N = 5/10, 6M N = 10/22) (Fig 1).
Fasting blood samples were obtained at baseline, 3M and 6M to measure glucose, total, low-and high-density lipoprotein cholesterol (TC, LDL, HDL) and triglyceride levels. Furthermore, in the FF-DM swine, glucose levels were assessed weekly by 24-hour urine samples. When glucose appeared in undiluted urine samples, venous glucose and ketone levels were checked via ear vein puncture and a handheld reader. When glucose levels were high (>20 mmol/L), in combination with ketone production, subcutaneous insulin (approx. 5-15 units once daily) was given to eliminate detectable ketone production while maintaining hyperglycemic state.

In-vivo QCA, (PS)-OCT and NIRS analysis
See S1 File for a detailed description of the imaging analyses. Coronary angiograms were obtained in two orthogonal views and QCA-analysis was performed (CAAS, version 5.9.2 Pie Medical Imaging BV). Mean (LD) and minimal lumen diameter (MLD), scaffold to artery (S/A) ratio, acute gain and late lumen loss (LL) were documented. Longitudinal matching of OCT pullbacks (C7XR Fourier-Domain, St. Jude Medical) was performed as described previously using dedicated CURAD analysis software (CURAD BV). [16] OCT parameters for vascular reaction, including lumen and scaffold dimensions and scaffold strut appearance, apposition, neointimal coverage and coverage morphology were assessed in 1-mm intervals using off-line OCT analysis software according to previously published methodology. [17] In pre-implantation lesions not exceeding the penetration depth of OCT, plaque burden (PB) was determined. Mean number of discernible struts were documented immediately post-implantation and at follow-up. Changes in strut appearances were categorised as preserved, open, dissolved bright, and dissolved black box. [18] Struts were scored as covered or uncovered and the morphology of the coverage-defined as [SA-LA] [16]-was described as homogeneous or heterogeneous. Heterogeneous coverage was furthermore described as lipid-laden, calcified, surrounding the struts or subluminal, or mixed (S1 Fig). [19] PS-OCT was performed using a prototype imaging system. PS-OCT provides a measure of tissue birefringence, an optical tissue property that describes the interaction with polarized light. It grossly relates to microscopic tissue organization, and enables characterization of collagen content and smooth muscle cell (SMC) density in atherosclerotic plaques. [20]  NIRS analysis (LipiScan, InfraReDx) was used for lipid core plaque (LCP) characterization. [21] Lipid-core burden index (LCBI) was documented, indicating high probability that LCP is present. To evaluate the agreement between OCT and NIRS for detection of lipid, we compared-per scaffold-the LCBI score to the percentage of OCT cross-sections with lipid-containing morphology.
Ex-vivo degradation analysis GPC was performed as described previously (see S1 File). [7] Degradation in our model was studied in relation to DM, time, inflammation, scaffold recoil, OCT-derived strut appearance and pre-implantation plaque burden.

Ex-vivo histological analysis
See S1 File for a detailed analysis. Proximal, middle and distal sections within each BVS were obtained. Tissue sections were stained by Hematoxylin-Eosin (HE) as an overview stain, Resorcin-Fuchsin for elastin, Alcian-Blue for proteoglycans, Oil-red-O (ORO) for lipids, Picrosirius Red (PSR) for collagen, von Kossa for calcium, and immunohistochemistry for smooth muscle cells (aSMA, clone 1A4, Dako, the Netherlands) and leukocytes (CD45, clone MCA 1447, AbD Serotec, UK). Polarization microscopy was performed to assess scaffold struts.
Histological analysis included neointimal healing and organization, collagen distribution, injury and inflammation score, lipid accumulation and presence of calcium classified as subluminal or surrounding struts (S2 Fig).

Statistical analysis
Statistical analysis (SPSS 20.0) entailed the Kolmogorov-Smirnov test for normality of the data. Normal distribution was expressed as mean ± standard deviation. Non-normally distributed data were presented as median with interquartile range. Comparison of in-vivo imaging between FF-DM and FF-NDM was performed by generalized estimating equations (GEE) modeling. GEE is a statistical method that accounts for the clustered nature of >1 scaffold analyzed from one swine, which might result in unknown correlations among measurements within these scaffold clusters. A linear response model was applied with an exchangeable structure for the within-cluster correlation matrix. For repeated measures, GEE modeling was performed using a linear response model with an autoregressive (AR(1)) structure for the withincluster correlation matrix. Comparison of ex-vivo GPC between FF-DM and FF-NDM swine was performed by independent samples t-test. To assess variable relations, the Spearman correlation coefficient was computed. All statistical tests were 2 tailed, and P<0.05 was considered statistically significant.

In-vivo QCA, (PS)-OCT and NIRS
All 32 BVS were successfully implanted with a mean S/A-ratio of 1.1±0.1 in FF-DM and FF-NDM (P = 0.20).
QCA findings are presented in Fig 2A + 2B and S1 Table. Pre-implantation lesions were mild and similar between FF-DM and FF-NDM (P = 0.33). In all swine, mean LD decreased from post-implantation to 3M (P<0.01) and remained fairly stable from 3M to 6M (P = 0.34 and P = 0.54, respectively).
PS-OCT was only qualitatively analyzed and showed a heterogeneous neointima with spots of elevated birefringence (Fig 3E and 3K). Rapid depolarization of the signal was often observed, focally ( Fig 3F2) and in areas with the appearance of lipid-rich plaque and inflammation (Fig 3F and 3L).
The association of LCBI-score and percentage of OCT cross-sections with a lipid-laden or mixed appearance was modest at 3M (Spearman's rho 0.397; P = 0.05) but became stronger at 6M (Spearman's rho 0.666; P<0.01).
Collagen poor regions were observed within the neointima. They contained leukocytes, were evident at sites with extracellular lipids and often coincided with calcifications (Fig 4). Collagen poor but SMC positive tissue generally contained lipid accumulation (P>0.10 for FF-DM vs. FF-NDM) (Fig 4G-4I). Mainly intra-and extracellular lipid deposits with few cholesterol crystals were observed and advanced necrotic cores were absent.
Neointimal and peristrut calcifications were observed in FF-DM and FF-NDM (P = 0.04) at 3M and 6M (Fig 3, Table 2), with varying size, shape and location between animals, suggesting an inter-animal difference. From 3M to 6M, lipid-accumulation remained (Fig 5) and

Discussion
The present study describes the coronary artery response to BVS1.1 in FF-DM and FF-NDM swine. A remarkable neotintima burden with a highly heterogeneous appearance was observed in all swine, with lipid accumulation and calcification, indicative for the formation of

Neoatherosclerosis development following BVS-implantation
Interestingly, considerable neointima formation with complete strut coverage was observed in all swine, independent of the presence of DM. This is relevant, as uncovered struts have been associated with adverse events like stent thrombosis. [23] However, the neointima was highly heterogeneous in all swine, with substantial lipid-and calcium accumulation and lack of intimal organization at 6M, consistent with neoatherosclerosis formation. This is of note, as previously published experimental studies in healthy swine and clinical studies in selected patients demonstrated a favorable vascular response with rather homogeneous coverage following BVS-implantation. [7,10,24] There are three main differences between our study and the previously published experimental study by Onuma et al., namely 1: the version of the BVS; 2: the species in which the majority of the study was performed (miniswine) and 3: the presence of diabetes and hypercholesterolemia; In addition, we expect that, given the size of the animals in our study, our stents were implanted more distally.
First, in the histological evaluation of the first-generation BVS, revision 1.0, implanted in the coronary arteries of healthy Yorkshire x Landrace or Yucatan miniswine of unspecified age, minor calcifications were observed around the scaffold struts in the majority of BVS as early as 28 days post-implantation. [7] After an initial increase, the presence of calcifications decreased to 17.2% between two and four years post-implantation. In our study, implantation  vascular wall barrier-function against excessive uptake of circulating lipids, resulting in neoatherosclerosis formation. [25,26] In humans, calcification is accelerated by young age and mechanical stress and this might explain why we found such a high incidence in our swine model. [27] While it is not completely clear to what extent our observations can be extrapolated to the clinical setting, our observation of considerable neoatherosclerosis formation under diet induced dyslipidemia might point at neoatherosclerosis as an important contributor to BVS failure at long term, similar to that described for DES and BMS. [28,29,30] It might also be in line with individual observations of asyptomatic neoplaque rupture after BVS-implantation and recently reported cases of BVS thrombosis. [10,17,24,31] Furthermore interesting is the fact that neoatherosclerosis development was observed in all swine, independent of the presence of DM. Kereiakes et al. demonstrated that diabetic patients who were receiving insulin treatment had a worse outcome after stent-implantation compared to those who were not receiving insulin. [32] In our study we did not observe a correlation between the amount of insulin given and the development of neoatherosclerosis. Of note, the swine receiving the most insulin, were not the swine that developed the worst neoatherosclerosis. Even in the two swine that received insulin throughout the entire study, who also received the greatest amount of food to ensure a similar growth pattern in all swine, neoatherosclerosis development was similar compared to all other FF-DM and FF-NDM swine. Factors such as duration of DM and hypertension may attribute to the severity of atherosclerotic disease and may therefore attribute to a more human-like evaluation of the coronary vascular healing response after BVS-implantation. Moreover, hypertension, not present in the current study, has been associated with adverse atherosclerosis-related events in DM patients. [33] Future studies assessing the coronary healing response after stent or scaffold-implantation in swine should consider using mature swine, and include risk factors such as hypertension to accurately evaluate the coronary vascular healing response to stent-implantation in a model that mimics human coronary atherosclerotic disease.
This study presents the first data of neoatheroasclerotic tissue organization characterized with PS-OCT. We observed enhanced tissue birefringence in areas with SMC alignment in the neointima, as well as in areas with inflammation. Macrophage recruitment in atherosclerosis has been associated with formation of cholesterol crystals [34], which are highly birefringent [35]. The contrast provided by PS-OCT, consisting of birefringence and depolarization, reflects tissue organization, which has an impact on structural plaque stability. A fuller understanding of the features highlighted by PS-OCT may complete our comprehension of neoatherogenesis and its impact on clinical sequelae.

Preserved scaffold integrity
GPC and histology demonstrated preserved scaffold integrity up to 6M after BVS1.1-implantation in all swine, which was not affected by DM or inflammation. This is expected as the scaffold starts losing structural integrity at 3-6 months, and scaffold resorption is driven by hydration, rather than inflammation. Although, theoretically, other factors associated with inflammation such as deregulated acid-base balance or body-temperature could influence scaffold degradation, this was not seen in the present study. [11] Interestingly, OCT did demonstrate morphological changes at individual strut levels despite preserved scaffold integrity. The OCT classification of strut appearances was developed in the ABSORB Cohort A trial to characterize the optical changes of the struts during the process of bioresorption. [9] However, preclinical evaluation of BVS1.0 demonstrated full degradation of the scaffold struts by GPC, while OCT demonstrated the presence of so-called 'preserved black boxes' within the vascular wall. [7] As the OCT signal is arising from the interface of structures with different optical indices, OCT reflects changes of tissue surrounding the struts, rather than changes in strut morphology. This should be kept in mind when interpreting in-vivo clinical and preclinical OCT observations in BVS.

Methodological considerations
Sacrifice was planned for 1/3 of the swine at 3M, and thus the serial BL, 3M and 6M sample size was relatively small. The aim of our study, however, was to longitudinally examine mechanistic and morphological aspects of the coronary response to BVS1.1 in FF-DM and FF-NDM swine. To accurately assess the mechanistic aspects-e.g. scaffold resorption-at various time points, additional planned sacrifice at 3M was beneficial. Furthermore, atherosclerotic lesions that developed in FF-DM and FF-NDM swine before scaffold-implantation were relatively small. However, distribution and size of the lesions were similar in both groups, allowing for adequate comparison of vascular responses following BVS-implantation between FF-DM and FF-NDM swine.

Conclusions
Scaffold coverage showed signs of neo-atherosclerosis in all FF-DM and FF-NDM swine, scaffold polymer was preserved and the vascular response to BVS was not influenced by diabetes.  Table. Quantitative QCA and OCT analysis results. Normally distributed data are presented as mean ± SD, non-normally distributed data as median (interquartile range). FF-DM = fast-food fed diabetic swine, FF-NDM = fast-food fed non-diabetic swine, QCA = Quantitative coronary angiography, OCT = optical coherence tomography, BVS = bioresorbable vascular scaffold, post = post-implantation, 3M = 3 months follow-up, 6M = 6 months follow-up. Ã P-value for the comparison between FF-DM and FF-NDM swine, †P-value for the difference between post-procedure and 3M, ‡P-value for the difference between 3M and 6M. (DOCX) S2 Table. NIRS analysis results. NIRS = Near-infrared spectroscopy, LCBI = lipid core burden index. §P-value for the difference between pre-procedure and 3M. Remaining footnotes and abbreviations are as listed in Table 1.

Supporting information
(DOCX) S3 Table. OCT strut appearance. Percentages are calculated as mean from the total (100%). Footnotes and abbreviations are as listed in Table 1