Plaques were obtained from 42 consecutive patients undergoing carotid endarterectomy. Patients were symptomatic (experiencing a transient ischemic attack (TIA)) and asymptomatic referring to department of Vascular Surgery at A. Gemelli Hospital during 2013–2015. Clinical data, obtained as part of standard patient care, and samples were retrospectively examined with the approval of the local ethical committee Board (Comitato Etico della Fondazione Policlinico Universitario A. Gemelli, reference number: 26089/16). Degree (%) of stenosis was assessed by echocolordoppler. Patients were stratified by 70, 75, 80, 85, and 90% stenosis, according to the ECST criteria (33) and guidelines of Italian Society of Vascular and Endovascular Surgery (34). Only patients with stenosis ≥70% were subjected to surgery and therefore included in this study (34). Patients were excluded from surgery and therefore from this study when the operation risks were considered too high.

Patients’ details, including therapies and comorbidities, are reported in Table 1. None of the patients had any autoimmune disease.

Immediately after surgery specimens were snap-frozen. The plaques were examined macroscopically to identify the site of the maximal plaque thickening and then transversely dissected into two segments along the longitudinal axis. The plaque segment showing the largest plaque burden as determined by visual assessment was defined as the zero (0) segment (35). The rationale is that the segment of a plaque with the largest plaque burden is generally, where the largest lipid core and the more extensive inflammation are present. The 0 segment was the reference segment and the adjacent 1-mm-thick segments at both sides were called ±1, ±2, … The plaque region cut and used for histological analysis ranged from segments −1 to +1, being defined as the segments with the greatest plaque burden. Plaque segment cutting was performed on dry ice. Segments embedded in optimal cutting temperature medium were then transferred to a cryostate and cut in 20 μm-thick slices. Slices were laid on a gelatinized glass for subsequent histological stainings. Before stainings, sections were thawed by 5 min washing with 0.05M triphosphate buffer saline (TBS) at room temperature (RT) and then post-fixated by 15 min incubation with 4% paraformaldheyde.

Blood samples were obtained one day after surgery. Clotting and complement activation were prevented by collecting samples in 10 mM of ethylenediaminetetraacetic acid (EDTA). Plasma was processed at 2,000 g for 15 min at 4°C and stored −80°C before analysis.

Hematoxyline and eosin (H&E) staining was performed according to standard protocols. H&E stained slices were acquired at 20× using a Nikon Eclipse Ni-E virtual stage microscope so to have complete stitching of the whole plaques at high resolution (1 pixel = 0.345 μm). A digital repository of stained plaques was created and used for subsequent quantifications. As a first evaluation, an investigator blinded to patients’ clinical data identified plaques with signs of rupture according to the classification proposed by Virmani et al. (36). For quantifications, images were scaled to microns. The area of intra-intimal lipid-rich core with hemorrhage and cholesterol clefts was calculated. Lipid core area was outlined on the plaque by manual operator selection (Figure 1A). In the lipid core area, cholesterol clefts were identified by gray scale thresholding and selected (Figure 1B). Hemorrhagic area was identified by digital color thresholding based on RGB pixel values (Figure 1C). Hemorrhagic content, lipid core, and cholesterol cleft areas were calculated as μm² and expressed as percentage of total plaque area. Plaque thickness was measured manually along the traces outlined in Figure 1D. Total tunica thickness was measured at the thickest point at the border of the vascular lumen and the media layer, excluding adventitial cutting artifacts from analysis. The media thickness was measured from the intimo-medial interface to the external border. The intra-intima infiltrated content of cellular debris and cholesterol (from here on referred to as “infiltrated content”) was defined as the ratio between total tunica thickness and media thickness.

After thorough washings with PBS 0.01M, 20-μm coronal sections were incubated with blocking solution (10% normal goat serum, 0.3% Triton X-100) for 1 h at RT and then with primary antibodies in the same solution overnight at 4°C. Primary monoclonal antibodies used were anti-human MBL (1:50, Abcam, USA), anti-human ficolin-3 (1:50, Hycult Biotechnologies, NL), and anti-human ficolin-2 and ficolin-1 (both 1:50) (37).

Sections were then incubated with biotinylated secondary antibody (1:200, Vector Laboratories) for 1 h at RT and followed by fluorescent signal coupling with streptavidine TSA amplification kit (fluorescein, Perkin Elmer, USA). Sections were then incubated with True-Black quencher (1:20 in 70% Ethanol, Biotium, USA) to quench non-specific fluorescent signal. Appropriate negative controls without the primary antibodies were performed. None of the immunofluorescence reactions revealed unspecific fluorescent signal in the negative controls (Figure S1 in Supplementary Material). Immunofluorescence was acquired using a scanning sequential mode to avoid bleed-through effects by an IX81 microscope equipped with a confocal scan unit FV500 with three laser lines: Ar–Kr (488 nm), He–Ne red (646 nm), and He–Ne green (532 nm, Olympus, JPN), and a UV diode. Three-dimensional images were acquired over a 10 μm z-axis with a 0.23 μm step size and processed using Imaris software (Bitplane, CH) and Photoshop CS2 (Adobe Systems Europe Ltd.). Exclusion images on single focal planes were obtained by Fiji software image calculator protocol (“subtract” function).

A LP specific C3 deposition ELISA was performed to measure residual LP functional activity in patient plasma (38). A Maxisorp ELISA plate (NUNC™) was coated with 10 μg/mL mannan to test LP activation by MBL (38, 39), or 25 μg/mL acetylated bovine serum albumin (acBSA) to test LP activation by ficolins (21, 40) diluted in coating buffer (15 mM Na₂CO₃, 35 mM NaHCO₃, pH 9.6) and incubated overnight at 4°C. Residual protein binding sites were saturated by incubating the plate with 1% BSA-TBS blocking buffer (0.1% (w/v) BSA in 10 mM Tris–CL, 140 mM NaCl, 1.5 mM NaN₃, pH 7.4) overnight at 4°C. The plate was then washed with washing buffer (TBS with 0.05% Tween 20 and 5 mM CaCl₂). EDTA-plasma samples were thawed on ice and suspended in barbital buffered saline (BBS; 4 mM barbital, 145 mM NaCl, 2 mM CaCl₂, 1 mM MgCl₂, pH 7.4), to a final plasma concentration of 6%. Wells receiving only BBS buffer were used as negative controls. Plasma solutions were incubated on the coated plate at 37°C for 1 h 30 min (40 μL/well). The plate was washed and incubated for 1 h 30 min at RT with a polyclonal anti-human C3c antibody (Dako, A0062) diluted 1:5,000 in washing buffer. After washing, the plate was incubated with an alkaline-phosphatase labeled goat anti-rabbit IgG antibody (Sigma A-3812) diluted 1:5,000 in washing buffer for 1 h 30 min at RT. Following washing, the assay was developed by adding 100 μL substrate solution (Sigma Fast p-Nitrophenyl Phosphate tablets, Sigma). The absorption at OD405 nm was then measured using the Infinite M200 spectrofluorimeter managed by Magellan software (Tecan, CH).

Ficolin-1, ficolin-2, ficolin-3, and MBL in plasma were analyzed after plasma incubation on acBSA or mannan-coated plates prepared according to the procedure reported in the previous paragraph. The plates were incubated for 1 h 30min at RT with mouse polyclonal anti-human MBL (HM2061, Hycult Biotechnologies, The Netherlands), anti-human ficolin-1 (41), anti-human ficolin-2 (37), and anti-human ficolin-3 (HM2089, Hycult Biotechnologies, The Netherlands), all diluted 1:100 in washing buffer. After washing, the plates were incubated with an HRP labeled goat anti-mouse IgG antibody (Santacruz, CA, USA) diluted 1:1,000 in washing buffer for 1 h 30 min at RT. Following washing, the assay was developed by adding 100 μL substrate solution TMB (TMB Substrate Kit; code 34021; Thermo Scientific, MA, USA; 1:1 with H₂O₂ solution). The reaction was stopped by adding 100 μL H₂SO₄ 2M, and absorption at OD450 nm was measured as above reported. The specificity of ficolin binding to acBSA was shown by the lack of binding to BSA or to uncoated plates (Figure S2 in Supplementary Material).

Ficolin-2 stained sections were acquired at 10× (pixel size of 0.646 μm) with an Olympus BX-61 Virtual Stage microscope so to have complete stitching of the whole plaque for the two color channels (DAPI and ficolin-2). Analysis was done using Fiji software. Briefly, the region of interest was delineated within the necrotic area referring to the corresponding H&E image. After separation of the color channels, the ficolin-2 signal channel was normalized by imposing a 0 gray level value to background noise. To this purpose, three background areas were calculated for their mean gray value and this mean value subtracted to the whole image. Ficolin-2 signal was then calculated as integrated density within the region of interest.

All histological and plasma quantifications were performed by investigators blinded to patients’ clinical information. Patients were stratified by the degree of carotid stenosis (≤75 vs. >75%) to assess whether the relationships between plaque components were modified by severity of atherosclerosis. Fisher’s exact test was used to define the association between hemorrhagic content and plaque rupture and whether patients with low stenosis (≤75%) had higher hemorrhagic plaques. Column analysis after patient stratification was done by Mann–Whitney test in case of non-normal distribution of data or unpaired t-test in case of normal distribution. Normal distribution was assessed by Kolmogorov–Smirnov test. A “vulnerability score” was obtained stratifying plaque components (hemorrhagic content, lipid core area, media thickness, and infiltrated content) into four groups according to their quartiles. The total score ranged from 4 (stable plaque) to 16 (vulnerable plaque). Correlation analysis was performed by computing a Pearson r or Spearman r depending on data distribution. The Forest plot was obtained using the odds ratios calculated stratifying patients for high plaque vulnerability (defined as a score ≥12) and high protein levels (defined as optical density ≥ median of each protein tested in the ELISA assay). Correlations of ficolin-2 integrated density were done by computing the Spearman r since ficolin-2 values did not have a normal distribution (Kolmogorov–Smirnov test for normality). Statistical analysis was performed using standard software packages GraphPad Prism (GraphPad Software Inc., USA, version 6.0). All tests were two-sided and p values lower than 0.05 were considered statistically significant. Details on the statistical analysis applied for each experiment are reported in figure legends.

Of the 42 patients undergoing endoarterectomy, 5 were excluded due to excessive plaque fragility hampering cryostat sectioning (4) or to unsatisfactory staining (1). We therefore analyzed a total of 37 patients (Table 1). The mean age of the patients was 73.03 ± 7.42. In line with the prevalence of atherosclerosis in the general population (42, 43), men outnumbered women (31 vs. 6). All patients but one received either statins or antiaggregants; therefore, we did not correct data for treatment. The cohort was homogeneous for conventional risk factors, with majority of either current or former smokers (73%), having hypertension (86%), and dyslipidemia (92%).

Only 30% (11 out of 37) of patients were symptomatic, e.g., had single or recurrent transient ischemic attacks, making them eligible for urgent surgery. Clinically, patients were classified according to stenosis degree (%) in five ranks: 70, 75, 80, 85, and 90%.

Seventeen out of 37 patients showed ruptured plaques according to the morphological evaluation proposed by Virmani et al. (36) and depicted in Figure 2A. Plaque rupture was associated with a high hemorrhagic content (>1.98%, threshold defined on the hemorrhagic content median value; Fisher’s exact test, p = 0.022, Figure 2B). We therefore used the measure of hemorrhagic content as a quantitative indicator of plaque rupture. Contingency analysis revealed that patients with less stenosis (70–75%) were more likely to have hemorrhagic plaques than patients with higher stenosis (80–85–90%, Fisher’s exact test, p = 0.006, Figure 2C). This suggests that in the cohort analyzed, patients with ≤75% stenosis had more ulcerated, rupture-prone plaques (44). On the contrary, stenosis was not associated with the occurrence of symptoms (Fisher’s exact test, p = 1.000, data not shown). The observation that in the analyzed cohort, patients with low degree of stenosis had ulcerated, rupture-prone plaques was used as a criteria to stratify patients according to the presence of vulnerable plaques (low stenosis, LS) vs. those with stable plaques (high stenosis, HS). To test the hypothesis that lower degree of stenosis may be due to the erosion of vulnerable plaques and to define a histology-based evaluation of plaque vulnerability, we next quantified parameters of plaque morphology.

Atherosclerotic plaques were quantified for the following morphological characteristics, representing parameters associated with plaque instability (45): hemorrhagic content, cholesterol clefts, lipid core area, media thickness, total tunica thickness, and infiltrated content (Figure 3).

Plaques from LS patients had higher hemorrhagic content than those from HS patients (mean ± SD: 3.69 ± 2.17 vs. 1.67 ± 1.48% of plaque area, p = 0.003). These two groups did not differ for either cholesterol cleft area or tunica thickness (media + intima thickness). Patients with LS had smaller lipid core area (31.04 ± 15.72 vs. 42.60 ± 12.12% of plaque area, p = 0.016) and less infiltrated content (3.57 ± 1.87 vs. 5.77 ± 3.73 total tunica/media, p = 0.048) than HS patients. In line with these data, LS patients had thicker tunica media (530.1 ± 266.0 vs. 372.5 ± 156.6 μm, p = 0.031) since either less infiltration exerts less compression on this area or lamina becomes thicker and flaky, therefore favoring instability.

These data indicate that, in the cohort of patients analyzed, eroded plaques (associated to LS and increased hemorrhagic content) can be histologically defined based on smaller infiltration and increased tunica media thickness.

We next sought to detect the presence of LP initiators within the plaques. We performed immunofluorescence for ficolin-1, ficolin-2, ficolin-3, and MBL, initiators of the LP (46). Ficolin-1 (Figures 4A,A′), ficolin-2 (Figures 4B,B′), and ficolin-3 (Figures 4C,C′) were present both in lipid core and tunica media, as further confirmed by the exclusion images showing ficolin selective signal in tunica. Ficolins were found in plaques with either low or high hemorrhagic content. MBL was present in lipid core and tunica media in plaques with high hemorrhagic content (>median 1.98%, Figures 5A,A′), while did not show any presence in plaques with low hemorrhagic content (<1.98%, Figures 5B,B′).

The presence of ficolin-1, -2, and -3 and MBL in the atherosclerotic plaques suggested that the LP is activated and involved in atherosclerotic events, therefore making it a candidate marker of pathology. To correlate LP activity to plaque vulnerability, a comprehensive vulnerability score was obtained based on quartile distribution of all four risk parameters (Figure 3): hemorrhagic content, lipid core area, media thickness, and infiltrated content. The total score ranged between 4 (stable) and 16 (vulnerable, Figures 6A,A′). Table 2 summarizes the histological information relative to each degree of stenosis.

A functional ELISA was used to measure residual LP activity in plasma based on C3 deposition on plates coated with the best-affinity ligands for MBL (mannan) or ficolins (acBSA) (38), as indicated by the data shown in Figure S3 in Supplementary Material. Vulnerability score significantly correlated with LP activation by MBL (C3 deposition on mannan, Spearman’s correlation test, p = 0.015), but not by ficolins (Figure 6B). In line with this, the LP activation by MBL was associated with vulnerable plaques (Figure 6C) showing lower lipid core, increased tunica media thickness, and lower infiltrated content (Figure 7A). In contrast, ficolin-2 was decreased in plasma from patients with vulnerable plaques, indicating its use (consumption) in pathway activation and possibly driving plaque instability (Figure 6C). However, ficolin-2 plasma level changes did not affect the functional assay of C3 deposition on acBSA (Figures 6C and 7B).

It is likely that an acute cardiovascular event (such as TIA, these patient symptom) affects the circulating levels of the LP proteins (17, 19). This implies that a central event may affect LP levels independently from the atherosclerotic process. In order to have a clear view of the events related to atherosclerosis, we have analyzed our data excluding the symptomatic patients (Figure 6D; Figure S4 in Supplementary Material). The associations between LP protein plasma levels and plaque vulnerability were confirmed analyzing only the non-symptomatic patients.

By analyzing the plasma levels of each LP initiator, stratifying patients using the median values of the four histological parameters associated with plaque vulnerability—hemorrhagic content, lipid core area, media thickness, and infiltrated content—we found that MBL and ficolin-3 were higher in plasma from patients with increased media thickness (Figures 7C,D). Ficolin-2 decreased in plasma of patients with high hemorrhagic content and low lipid core area (vulnerable plaques, Figure 7E). Ficolin-1 did not change throughout morphological stratifications (Figure 7F). These observations are in line with the Forest plot data presented in Figure 6C.

When we stratified patients for symptomatology, plasma levels of ficolin-1 were lower in symptomatic than in asymptomatic patients (Figure 8), and no other differences were found in the other LP initiators.

To define the correlation between LP protein deposition in the plaque and plaque vulnerability, we stained the plaques from all patients for ficolin-2, the only LP initiator protein changing in association with the vulnerability score (see Figures 6C,D). We acquired the whole plaque at high resolution and normalized the color channel with the ficolin-2 signal (Figure 9A). Ficolin-2 intraplaque levels (expressed as integrated density of fluorescent pixels in the region of interest), calculated in the necrotic core, were inversely correlated with the vulnerability score (Figure 9B), similarly to what is observed with circulating ficolin 2. Ficolin-2 intraplaque levels were directly correlated with the infiltrated content (Figure 9C).