Article Text

Original research
Assessment of endothelialization of aneurysm wall over time in a rabbit model through CD31 scoring
  1. Praveen Kolumam Parameswaran1,
  2. Daying Dai1,
  3. Yong-Hong Ding1,
  4. Tina Gunderson2,
  5. David F Kallmes1,
  6. Ramanathan Kadirvel1
  1. 1Department of Radiology, Mayo Clinic, Rochester, Minnesota, USA
  2. 2Division of Biomedical Statistics and Informatics, Mayo Clinic, Rochester, Minnesota, USA
  1. Correspondence to Dr Daying Dai, Department of Radiology, Mayo Clinic, Rochester, MN 55901, USA; dai.daying{at}


Background Intracranial aneurysms represent a significant health concern and are poorly understood despite decades of research. Our study focused on understanding temporal patterns of endothelial cell distribution in different spatial locations within the aneurysm early after creation in a rabbit model.

Methods Elastase induced saccular aneurysms were created in rabbits and harvested on day 1 (n=3) and after 2 (n=5), 4 (n=4), 8 (n=5), and 12 (n=6) weeks. Sham operated controls (n=3) were harvested on the same day. Aneurysm and control tissue samples were subjected to en face whole mount CD31 staining for endothelial cells. Semiquantitative scoring was performed on the basis of endothelial coverage of the vessel wall (proximal, middle, and distal portions of the aneurysm dome). Mixed effects models were used to assess the effect of time and aneurysm section on endothelial coverage.

Results Aneurysmal segments were near completely de-endothelialized at 4 and 8 weeks but had re-endothelialized by 12 weeks. Compared with controls, aneurysms at all time points showed decreased endothelialization, but the difference was only significant compared with the 4 and 8 week groups. Both time (P=0.03) and aneurysm section (P=0.07) were significantly associated with the degree of endothelialization. Proximal locations showed increased endothelialization compared with distal locations (P=0.03).

Conclusion In experimental aneurysms of rabbits, endothelial cells regress during the first month after creation, followed by ascending re-endothelialization that stays incomplete. These findings suggest that re-population of endothelial cells comes from resident cells in the adjacent parent artery and that deranged hemodynamics may affect full reconstitution of endothelial cells long term.

  • aneurysm
  • vessel wall

Statistics from


Intracranial aneurysms (ICAs), prevalent in 3–4% of the population, are a relatively common vascular disorder.1 2 ICAs have varied wall structure, cell content, and clinical outcome. The healing mechanism of an aneurysm continues to be enigmatic, and despite treatment with endovascular devices, the recurrence rates are still high (20–25%).3–5 Studies have shown that endothelialization forms the critical component in arterial hemostasis and in achievement of complete occlusion of the aneurysm neck.6 7 Moreover, incomplete endothelialization at the aneurysm neck has been shown to account for the increase in recurrences following coil embolization.8 9 Therefore, it is essential to understand the pattern of distribution of endothelial cells over time after the formation of an aneurysm. This study is an attempt to fill the gaps in knowledge about aneurysm healing, and explain the reason behind incomplete endothelialization noted after treatment with endovascular devices.

Since the wall represents the primary pathologic factor, indeed the actual site of enlargement and rupture, we believe that a comprehensive exploration of the aneurysm wall can provide valuable insight that will improve patient treatment. Several studies have shown that CD31 is a reliable marker for identification of endothelial cells, and hence scoring based on CD31 expression is a better way to understand the distribution of endothelial cells.10 11

The elastase induced aneurysm model of the rabbit has been used as an established clinical model for studying aneurysms and has a hemodynamic status comparable with human aneurysms.12–15 Several studies using the rabbit model have demonstrated the shear stress patterns and device testing of the aneurysms.14–16 However, no study has highlighted the distribution pattern of endothelial cells over time after aneurysm formation. In our study, we attempted to characterize the pattern of de-endothelialization and re-endothelialization of the vessel wall over time at different spatial locations in the aneurysm in our rabbit model.


Aneurysm creation

The Mayo Clinic Institutional Animal Care and Use Committee approved all of the protocols of the study. Elastase induced saccular aneurysms were created in 26 New Zealand white rabbits, and aneurysm tissues were harvested on day 1(n=3), and after 2 (n=5), 4 (n=4), 8 (n=5), and 12 (n=6) weeks. In the control group (n=3, day 0), similar procedures of aneurysm creation were followed but instead of the elastase, saline incubation was done in the right common carotid artery (RCCA). Detailed procedures for aneurysm creation have been described elsewhere.17–19 In brief, rabbits (3–4 kg) were induced with intramuscular injection of ketamine and xylazine (35 and 5 mg/kg, respectively) and maintained with 1–3% isofluorane. Using a sterile technique, the RCCA was exposed and ligated distally. A 5 F sheath (Cordis Endovascular, Miami Lakes, Florida, USA) was advanced retrograde in the RCCA to a point approximately 3 cm cephalad to the origin of the RCCA. A roadmap image was obtained by injection of contrast through the sheath retrograde in the RCCA, to identify the junction between the RCCA and the subclavian and brachiocephalic arteries (Advantx; General Electric, Milwaukee, Wisconsin, USA). Through the indwelling sheath, a 3 F Fogarty balloon (Baxter Healthcare Corporation, Irvine, California, USA) was advanced to the origin of the RCCA at its junction with the right subclavian artery. The balloon was inflated with just enough iodinated contrast material to achieve flow arrest in the RCCA. Porcine elastase (Worthington Biochemical, Lakewood, New Jersey, USA) mixed with iodinated contrast material was incubated in the dead space of the RCCA, above the inflated balloon, through a microcatheter (Tracker 10; Target Therapeutics, Fremont, California, USA). Balloon position and shape were documented with the use of fluoroscopic spot imaging, which was saved and printed. After incubation of the elastase solution, the balloon and sheath were removed, and the RCCA was ligated below the sheath entry site. The harvested aneurysm tissue was dissected along the midline into the dorsal and ventral pieces. The dorsal piece was processed for whole mount en face immunostaining with CD31 for endothelial cells.

Whole mount en face immunostaining

The freshly harvested dorsal piece of aneurysm wall was pinned to a dish coated with Sylgard (Dow Corning Corp) to expose its lumen side and was completely immersed in 10% neutral buffered formalin to fix for 2 hours at room temperature (RT). The sample was washed with Tris buffered saline (TBS) for 15 min in three washes. It was then blocked with 5% donkey serum in TBS Tween for 1 hour at RT, followed by incubation with primary antibody (CD31, 1:20–30 in 0.3% Triton X-100 (Sigma-Aldrich Co, LLC) in TBS) for 1 hour at RT, overnight at 4°C. The samples were washed again with TBS and incubated with secondary antibody (Cy3 conjugated donkey antimouse IgG, 1:200 in TBS) for 2.5 hours at RT and counterstained with Sytox green (ThermoFisher Scientific) (1:1000). Finally, the sample was scoped with confocal laser microscope.

Each sample was divided into three portions for scoring: distal (apical one-third far away from the parent artery); proximal (basal one-third close to the aneurysm neck); and middle (between the proximal and distal areas). Five fields from each portion were randomly selected for evaluation to achieve five scores for each portion. Scoring of endothelial cells was performed as stated below (figure 1).

  • Score 0=complete lack of coverage of CD31 positive endothelial cells.

  • Score 1=scattered, sparse cells positive for CD31, covered the lumen side of the wall.

  • Score 2=small, isolated, patchy cell clusters are positive for CD31.

  • Score 3=the cell clusters which are positive for CD31 are larger than that seen in 2.

  • Score 4=majority of the lumen side of the wall is covered with CD31 positive cells, with the exception of some small gaps among cells.

  • Score 5=the lumen side of the wall is completely covered with CD31 positive cells.

Figure 1

CD31 stained endothelial cells categorized into five scores based on the distribution of endothelial cells. Score of 5=highly confluent endothelial cells; score 0=no endothelial cells.

Statistical methods

Each rabbit had five fields scored from each aneurysm portion (proximal, middle, and distal). Scoring consistency between fields was tested using Krippendorf’s α. Field scores were averaged to create a composite score for each aneurysm location within the study rabbit. A mixed effects model was used to test for association of score with time and location, as shown in table 1. Denominator degrees of freedom were corrected using the Kenward–Rogers method. Model diagnostics were performed. Overall variable significance in the model and model based means are presented. Because of the preliminary nature of the investigations, significance was set at α=0.10. Analyses were performed with statistical software (SAS Institute Inc).

Table 1

Estimate of group means from the mixed effects model


Aneurysms harvested on day 1 were excluded from the analysis due to the presence of fresh, unorganized thrombus within the dome, which prevented us from processing the samples for en face CD31 immunostaining.

Sham operated control samples showed the highest level of endothelialization (figure 2), but the difference was only significant compared with the 4 and 8 week groups (table 2). Compared with the control group, the groups at 4 weeks and 8 weeks showed decreased endothelialization (−2.18 (90% CI −3.39 to 0.974), P=0.006; −1.65 (90% CI −2.81 to–0.497), P=0.023). Of the remaining comparisons, the 2 week and 12 week groups showed significantly greater endothelialization than the 4 week group (2 week–4 week: 1.53 (90% CI 0.468 to 2.59), P=0.022; 2 week−12 week − 1.34 (–2.36, –0.317), P=0.036); none of the other pairwise comparisons between different times of sacrifice were significantly different (table 2). Thus our results showed a trend of decrease in CD31 scores starting from day 0 to 4 weeks followed by an increase in CD31 scores at 8 and 12 weeks. However, the scores at 12 weeks did not match the scores for the controls(day 0).

Figure 2

Comparison of CD31 scores and time points of sacrifice at different aneurysm section locations. Each point represents the composite score for one rabbit. The center line on each box represents the median; the limits of the box, and the interquartile range. Endothelialization is greatest in the control group (mean 2.47) followed by de-endothelialization in week 2 (mean 1.81) and week 4 (mean 0.283). Re-endothelialization is noted again at week 8 in the proximal and middle portions of the aneurysms. Re-endothelialization is noted in all three regions at week 12 (mean 1.62).

Table 2

Mixed model pairwise comparisons between levels of each variable

The coefficient of reliability among the five raw CD31 positive scores at each animal x location combination was α=0.755. Descriptive statistics for the composite scores per rabbit were calculated and figure 2 shows the box and whisker plots for composite CD31 scores at each time point and aneurysm section location. Group mean estimates and CI from the linear mixed model are shown in table 1. Both location and time of sacrifice were significant at the P=0.1 level (location P=0.078, time P=0.032).

The interaction effect (time x location) was not significant and was dropped from the model. Both middle and proximal locations had higher endothelialization compared with the distal location, although only the proximal location was significantly different (mean difference (90% CI) 0.417 (–0.027 to 0.862), P=0.12; 0.600 (0.16 to 1.04), P=0.028, respectively). Middle locations were also decreased compared with proximal locations, but these differences were not significant (−0.183 (–0.627 to 0.262), P=0.49) (table 2).


In our study, moderate de-endothelialization was noted in the sham operated control in the aneurysm wall. It was followed by further de-endothelialization at 2 and 4 weeks and re-endothelialization at 8 and 12 weeks. Endothelialization was not complete near the 12 week point, probably because of deranged hemodynamics that altered the blood flow within the aneurysm.

Aneurysm formation occurs because of weakening in the vessel wall due to stripping of the internal elastic lamina and tunica media, resulting in an outpouching of only tunica intima and adventitia. It was conventionally believed that endothelial dysfunction leading to denudation occurs during aneurysm formation, followed by re-endothelialization and collagen deposition.20 In our rabbit aneurysm model, we observed initial denudation of endothelial cells associated with advancement of sheath and balloon inflation during aneurysm creation. We were not able to clearly delineate the endothelial injury attributed to the elastase injection during aneurysm creation owing to the unorganized thrombus formation within the aneurysm 24 hours after creation. However, a pattern of early de-endothelialization at aneurysm creation and further de-endothelialization at 2 and 4 weeks could be attributed to the altered flow pattern within the aneurysm. This finding is clinically relevant because de-endothelialization and hypocellular vessel wall are important contributors in aneurysm rupture, as reported by Frosen et al.21 We believe the wall is at increased risk of rupture around 4 weeks after aneurysm creation. Our rabbit model has histologic characteristics identical to human aneurysms.22 This finding also sheds light on the reason for the rupture of small aneurysms in certain patients. 22–24

The current knowledge of endothelialization pertaining to aneurysm healing has been shown in relation to growth of cells on endovascular coils and flow diverter stents.15 25 The source of endothelialization after endovascular treatment continues to be controversial. Dai et al26 have shown that initial endothelialization following coil embolization occurs as a result of continuation of cells from the parent artery. Liu et al27 have shown that endothelial progenitor cells in circulation contribute to the endothelialization following coil embolization.28–30 On the basis of our initial endothelialization noted until 2 weeks of aneurysm creation, we believe that endothelial cells remain contiguous with the parent artery. Following this, we observed further de-endothelialization at 4 weeks. We believe this further denudation of endothelial cells may be due to an altered flow pattern within the aneurysm. Over time, re-endothelialization occurs, with the new flow pattern starting from the aneurysm neck, and reaches the dome around 12 weeks, as reported by findings in our study. This finding further confirms our theory of endothelial cell origin from the parent artery because re-endothelialization begins in the neck close to the parent artery at 8 weeks, whereas no endothelial cells were noted at the dome in that time frame. Moreover, endothelialization was noted to be greatest at baseline (day 0) of aneurysm formation and did not have a similar cell number around 12 weeks (figure 2). This outcome might be due to the altered hemodynamics within the aneurysm, leading to incomplete endothelialization, subsequently leading to recanalization of aneurysms.

Currently, no study has delineated the pattern of distribution of endothelial cells in the aneurysm wall and no scoring system is available to quantify endothelialization across different time points. Our scoring system not only quantitates spatial distribution of endothelial cells in aneurysms but also can help predict the rate of healing of aneurysms after treatment with endovascular devices. This information will enable us to identify devices that promote better healing of aneurysms. Future studies are warranted to explore new options in stabilizing the endothelial cells during the first month of aneurysm formation that might lead to better outcomes in aneurysm treatment.


Our study has limitations. The aneurysms produced in the elastase induced rabbit model continue to be stable after 1 month and do not enlarge and rupture. Thus endothelialization in enlarging ruptured aneurysms cannot be correlated in our model. Even though vascular smooth muscle cells are also critical in aneurysm formation, owing to the technical difficulty of scoring smooth muscle cells, we only analyzed endothelialization patterns. The presence of unorganized thrombus in the aneurysm sac 24 hours after aneurysm creation limits us in understanding the role played by elastase in the de-endothelialization in our model. With regards to the role played by altered flow pattern within the aneurysm, we can only postulate and not definitively state that deranged hemodynamics is responsible for the stripping of endothelial cells observed at 4 and 8 weeks following aneurysm creation.


In experimental aneurysms in rabbits, endothelial cells regress in all regions during the first month after creation, followed by ascending re-endothelialization that remains incomplete. These findings suggest that re-population of endothelial cells comes from resident cells in the adjacent parent artery, and that deranged hemodynamics may affect the full reconstitution of endothelial cells long term.


We would like to thank the Radiology Research Internal Grant Award program, Mayo Clinic, for the support provided to present the findings of the study at the Society of Neurointerventional Surgery conference.


View Abstract


  • Contributors PKP and DD contributed to tissue processing, slides staining, interpretation of the data, and drafting of the manuscript. Y-HD contributed to the aneurysm model creation. TG contributed to the statistics of the study. DFK and KR contributed to the conception and design of the study, and to revision of the article critically for important intellectual content.

  • Funding This work was supported by grants from the National Institutes of Health awarded to KR (R01.NS076491) and DFK (R21.NS088256). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Correction notice Since this article was first published online the author name Kadirvel Ramanathan has been updated. It has been corrected to read Ramanathan Kadirvel.

  • Presented at These data were presented at the 14th Annual Meeting of the Society of Neurointerventional Surgery, Colorado Springs, Colorado, July 23-28, 2017.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.