The primary aim of the present study involved identifying and comparing proteins and molecular pathways defining the two most commonly used mouse AAA models: AngII infusion into Apoe^−/− or Ldlr^−/− male mice, and the localized, luminal perfusion with elastase into the infrarenal aorta of male wildtype mice (Figure 1A). To capture a proteome correlated with a “developing” aneurysm as well as one that correlated with an “established” aneurysm, we chose two time-intervals for aortic tissue harvest: 2 and 4 weeks after initiation of AngII infusion, and 4 days and 2 weeks after elastase exposure (Figure 1B). The extent of aortic expansion was determined by ultrasound-based measurements of aortic diameter (Figure 1C; Supplementary Table 1), depicted as percent increase at each time interval (Figure 1D). Aortic diameter dilation of >50% (relative to baseline prior to aneurysm induction) was considered aneurysmal. Compared to elastase, expansions due to AngII infusion demonstrated greater variability (Figure 1D), consistent with previous reports (16, 22).

Histopathology demonstrated adventitial thickening, cellular infiltrate, and elastin layer tears or breaks common to both two models (Figures 1E,F). The difference between diameter of representative aortas from saline-infused from each model can be attributed to the suprarenal or infrarenal locality of the AngII and elastase models, respectively.

To capture the spatiotemporal protein expression patterns within and across experimental AAAs, we performed label-free proteomic profiling of the arch, thoracic, suprarenal, and infrarenal segments of the aorta (Figure 2A). The proteomics data reflected the variation in the aneurysmal expansion between and within the AAA models (Supplementary Figure 1; all mice studied). This variability, particularly notable in AngII-infused mice, could pose problems if the proteomes of mice with an aortic expansion <50% would dilute the aneurysm-specific variation in the proteome. We therefore selected a subset of mice from each group that would specifically capture the aneurysm-specific proteome, prioritizing the final analyses on data acquired from three mice, per procedure, with the most severe aneurysm pathology (Figure 2B) for subsequent analyses.

A total of 2,477 proteins from all three AAA groups and appropriate controls were quantified (Supplementary Table 2). A Venn diagram demonstrated the interconnectivity of these models (1,913 proteins were shared amongst all three groups), as well as the unique features resulting from the specific aneurysm model and/or the genetic background of each group (Figure 2C). Hierarchical clustering of protein abundances revealed three major patterns (Figure 2D). Pattern 1, a baseline pattern, contains predominately proteins associated with AngII model controls, or segments of artery not directly affected by surgery (i.e., artery segments 1 through 6) in the elastase model. Pattern 2, a general pathology pattern, contains proteins whose expression coincides with aneurysm locality within each model (suprarenal segments in the AngII groups, infrarenal segments in the elastase model); and Pattern 3 is populated by proteins with increased expression in the arch segments of Apoe^−/− and Ldlr^−/− aneurysmal mice.

The principal component analyses (PCA, Figure 2E) further delineated the intra-group sources of variability. Segments of the suprarenal and arch aorta revealed the greatest variation in the Apoe^−/− group, likely due to the aneurysmal and atherosclerotic plaque features, respectively. Several points representing the arch and suprarenal segments remain intermingled, possibly demonstrating the interference of atherosclerotic and aneurysmal features in these mice, even in the absence of a “Western” diet (Figure 2E). The Ldlr^−/− plot, on the other hand, showed the fewer intermingled suprarenal and arch segment points, possibly due to the expected absence of atherosclerotic plaques in these mice in the absence of a “Western” diet (23). In complete contrast, and as expected for the elastase group in wildtype mice, infrarenal fractions cluster closely together and provide the most notable variation (Figure 2E).

The PCA (Figure 2E) revealed that the protein abundance patterns of the eight segments can be simplified into three arterial segments of interest: arch (segments 1–3), suprarenal/thoracic (4–6), and infrarenal (7 and 8) (Figure 3A). Subsequent, in-depth analyses of the data was performed using these data-identified arterial segments. The multi-dimensional nature of our dataset, including model, treatment, time interval, and aortic section, required careful consideration to identify proteins contributing to aneurysm development and progression. Thus, we turned to a high dimensional dataset clustering strategy that combined protein abundance profiles from multiple datasets into a single dataset for clustering (21) (Figure 3B). The single 20-cluster output, comprising a combined 11,600 protein profiles pertaining to 2,477 unique proteins, depicted the variation across all six groups—the three aneurysms and their saline control counterparts (Figures 3B,C). Abundance profiles expected from aneurysm-specific responses for the AngII model (suprarenal, red-highlighted clusters) or the elastase model (infrarenal, green-highlighted clusters) were selected (Figure 3C). The pie charts depict group compositions for each cluster, with red-highlighted, red/pink dominated pie charts demonstrating a dominance of proteins from the AngII groups whereas green-highlighted pie charts show a dominance of proteins from the elastase group (green shades; Figure 3D).

Proteins from the five AngII- and three elastase- dominated clusters were further filtered to identify those with a profile unique to the aneurysm, i.e., the counterpart saline profile did not fall within any of the selected cluster patterns. These filtered protein lists were refined by identifying proteins with a fold change ≥ 2 or ≤0.5 as compared to their localized saline controls, p-value ≤ 0.05, and false discovery rate (FDR) ≤ 0.05. This high dimensional clustering and filtering step reduced the proteins of interest to include 159 and 158 proteins in AngII-infused Apoe^−/− and Ldlr^−/− mice, respectively, and 173 proteins in the elastase mouse group (Figure 3E; Supplementary Table 3). The elastase model contains 127 unique proteins and shares 31 and 26 proteins with AngII-infused Ldlr^−/− or Apoe^−/− mice, respectively; the two groups in the AngII model share 56 common proteins.

The list of eleven proteins shared across all groups includes three proteins of particular interest: regulator complex protein Lamtor 3 (LAMTOR3), procollagen C-endopeptidase enhancer 1 (PCOLCE), and collagen alpha-1 (XII) chain (COL12A) (Figure 3F; Supplementary Table 3). LAMTOR3 participates in mitogen activated kinase 1 (MAPK1/ERK2) signaling (24, 25), a protein studied previously in aneurysm (26, 27) and enriched in AngII Ldlr^−/− and wildtype elastase aneurysms. PCOLCE interacts with THBS1 (28) and MMPs such as MMP2 (29), two proteins identified in both AngII groups and implicated previously in mouse aneurysm models as well as human aneurysms (26, 27, 30–33). In addition to PCOLCE, the presence of COL12A in all three aneurysm groups indicates a role for fibrosis and/or ECM interactions (34). Other proteins listed in Figure 3F include those with relevance to the networks shown further below. A full list of filtered aneurysmal proteins is presented in Supplementary Table 3.

Western blot analyses and tissue immunostaining validated select proteins (Supplementary Table 4). For the later interval in each model, we readily detected and verified an increase in signal for CORO1A (AngII-filtered protein), CTSB (elastase-filtered) and HMGB1 (AngII and elastase-commonly filtered) across all three groups (Figure 3G). Although the presence of STAT3 and MAPK1 could be confirmed in each model using Western blot analysis, the expression was not increased in aneurysmal tissues as compared to control (Figure 3G). While immunostaining revealed FERMT3/KINDLIN3 to be increased moderately in AngII Apoe^−/− tissues as compared to control, the difference did not reach significance (Supplementary Figure 2).

Due to the multi-step, stringent filtration process chosen for this proteomics dataset, there are proteins ultimately listed as associated with only one AAA model (e.g., CORO1A in AngII, and CTSB in elastase), which are in actuality increased in the respective aneurysm region for both models. CORO1A increased in elastase-administered tissues but was excluded from the elastase model proteome because corresponding saline-perfused tissues exhibited a similar expression pattern (Figure 3H). AngII-infused Apoe^−/− mice excluded CTSB due to its elevated expression in both supra- and infra- renal segments (Cluster #11 in Figure 3C). This retrospective examination of the proteomes reflected the pros and cons of any given protein filtering strategy.

Our study aimed to monitor the proteomes of early (developing) and late (established) aneurysms, providing a means to address a potential mechanism and investigate factors in disease progression. Considering each model separately, Venn diagrams provided an overview of aneurysm-related proteins in early and late time intervals (Figure 4A). With the aim to visualize potential relationships across different AAA proteomes and identify the molecular mediators between early and late intervals, we built subnetworks for each model by mapping the temporally labeled (early, late, or both) proteins on a fixed coordinate system on the protein-protein interaction (PPI) network (Figures 4B–D). The PPI network was built from human interaction databases, and thus the human orthologs for proteins are indicated in the subnetworks. Inspecting the connections within each model's subnetwork revealed the following trends: (1) the largest PPI subnetwork for AngII mice comprised ribosomal proteins that primarily associate with the early interval for Ldlr^−/− and the late interval for Apoe^−/− (Figures 4B,C); and (2) the largest PPI subnetwork for the elastase model comprises a distinct group of ribosomal proteins that remain connected to EGFR and MAPK subnetworks primarily associated with the late interval (Figure 4D). This MAPK node connected to LAMTOR3 and WASF2, two proteins identified in all three aneurysm groups. These proteins were associated with the late stage/established AAA in AngII Ldlr^−/−, but with early/developing, late, or both stages for AngII Apoe^−/− and elastase-perfused mice. Other proteins that were relatively close to the MAPK node include THBS1, MMP2 in both AngII groups, and PCOLCE in all three groups (Figures 4B–D). Particularly in the context of the integrins found in proximity to this subnetwork in the late stage AngII groups (ITGAM/CD11b, ITGB2/CD18, ITGA2B, ITGB3) (35–37), THBS1 (30, 38) and WASF2 (39) may have contributed to adhesion and transmigration activities. The overall topology of the fixed coordinate PPI confirmed, as expected, that the AngII Apoe^−/− and Ldlr^−/− models share more similarities to each other than to the elastase model; but a closer look at subnetworks demonstrated that peak expression times vary across all three groups.

A representative subset of significantly enriched pathways derived from AAA proteomes at early and late time intervals of each group are presented in a heatmap (Figure 4E; Supplementary Figure 3 for the complete list of pathways). Signaling pathways related to translation and transcription, inflammation and inflammatory cell activities, and coagulation pathways were associated with all three groups, primarily the early and late interval proteomes of both AngII models, and the late interval proteome of the elastase model (Figure 4E). The early interval proteome for the elastase model uniquely associates with apoptosis-related pathways and phagocytosis (Fc gamma R-mediated phagocytosis), while oxidative stress (ROS, RNS production in phagocytes) is identified in the early proteome of the AngII Ldlr^−/− group. The early and late interval proteome of the AngII Apoe^−/− group share several features including cell damage and apoptosis (apoptosis-induced DNA fragmentation, phagosome), and features that may indicate endoplasmic reticulum (ER) stress (retrograde and anterograde transport, ER associated degradation, and cargo concentration in the ER) (Supplementary Figure 3; Figure 4E). The late interval proteomes of the AngII model shared pathways related to ECM and collagen formation and leukocyte infiltration, while the late interval of the elastase model included pathways of nuclear factor kappa B (NF-κB)-driven inflammation and shared pathways of vasculogenesis (vascular endothelial growth factor, VEGF), cell proliferation, and mitogen-activated protein kinase (MAPK) signaling with the late interval AngII Ldlr^−/− proteome (Figure 4E).

The proteomic signal from the thrombus could potentially supersede that of the arterial wall at the site of aneurysm expansion. We used laser capture microdissection (LCM) to separate the thrombus from the surrounding arterial wall for proteomics (Supplementary Figure 4; Supplementary Table 5). From six Apoe^−/− and three Ldlr^−/− aneurysm LCM samples, we quantified a total of 2,311 proteins. A two-group comparison between the arterial wall and thrombus proteomes was performed for each genotype (Supplementary Figure 4; Supplementary Table 5). Most differentiating proteins within each group were increased in the arterial wall. Since the wall and thrombus proteomes largely overlap, the thrombus itself did not account for the differences between aneurysm and control tissues.

To further substantiate the clinical relevance of the mouse models, we performed a proteomics study on human diseased tissues. Nine human infrarenal AAA samples were acquired, and eight were separated into thrombus and AAA tissues, leaving one sample with no distinct thrombus as an intact sample. Non-aneurysmal, infrarenal aortic tissues from 10 autopsy cases served as controls (Figure 5A; Supplementary Table 6). A total 2,485 proteins were quantified, however, initial analyses showed control tissues to contribute 43 unique proteins as compared to 6 from AAA tissues and none from thrombus tissues (Figure 5B).

Hierarchical clustering of the protein abundances yielded three major clusters that showed not only a marked difference between the control and aneurysmal tissues, but also a subdivision within the aneurysmal aortic wall samples (Figure 5C). Specifically, five AAA samples clustered closer to the thrombus samples, hence designated thrombus-like AAA (T-AAA), whereas the remaining four aneurysmal aortic wall samples formed a distinct cluster and were designated non-thrombus AAA (nT-AAA, Figure 5D). The heatmap also demonstrated that, when compared to the control tissues, aneurysmal tissues contain a low abundance of most detected proteins, likely reflecting the ECM breakdown, medial degeneration, and acellular composition of these tissues. Only a few proteins comprised the tissue-specific proteome signal, defined primarily by Cluster 1 for nT-AAA samples and Cluster 3 for T-AAA and thrombus samples (Figure 5C). The full list of the protein abundances within each tissue type is represented in Supplementary Table 6.

Since the protein distribution signal varied greatly between control and aneurysmal tissues, we instead compared tissues using a percentile ranking of their averaged protein abundances. Each protein is plotted according to relative abundance within each tissue type, and each plot divided to five percentile ranks (pentiles) in order to identify proteins moving from lower percentile ranks in control tissues to higher percentile ranks in the aneurysmal tissues (Figure 5E). We mapped the locations of some proteins of interest as identified from the animal models such as FERMT3/KINDLIN3, CORO1A, THBS1, and STAT3, and observed interesting trends: FERMT3/KINDLIN3, CORO1A, and THBS1 move up in abundance rank between control and each aneurysmal tissue, with FERMT3/KINDLIN3 and CORO1A moving from the 60th to 80th percentiles, and THBS1 remaining in the upper most pencentile range (Figure 5E). STAT3, on the other hand, moved to lower abundance ranks in all three aneurysmal tissues, dropping to as low as the 40th percentile in the T-AAA tissues.

The increase in abundance rank for FERMT3/KINDLIN3 and CORO1A led to the examination of other proteins exhibiting a similar behavior (Supplementary Table 7). These analyses resulted in 122 (nT-AAA), 160 (T-AAA) and 192 (thrombus) proteins that were mapped to the PPI network (Figure 6). Larger font and arrows are used to highlight several topological features, proteins, and/or subnetworks shared among the three networks. For example, a subnetwork of proteins related to translation and transcription predominates the nT-AAA network (Figure 6A) but remained almost completely absent from the T-AAA and thrombus networks (Figures 6B,C). The latter two networks, while generally comprising a complementary topology, contain notable attributes. Specifically, the thrombus network contained a subnetwork populated by proteasome-related proteins such as proteasome subunit alpha 1 (PSMA1), 4 (PSMA4), 6 (PSMA6), and 7 (PSMA7), which further extends to both an erythrocyte subnetwork and a platelet subnetwork, including scavenger receptor CD36, platelet glycoproteins GP1BB and GP9, von Willebrand factor (VWF) (Figure 6C). The erythrocyte subnetwork, including erythrocyte membrane protein band 4.2 (EPB42) and/or protein 4.1 (EPB41), spectrins beta and alpha 1 (SPTB and SPTA1), and ankyrin-1 (ANK1) is present in each the T-AAA, nT-AAA, and thrombus network (Figures 6A,B). The interactions between this erythrocyte subnetwork and the proteasome and platelet subnetworks specifically in the thrombus PPI network could reflect inflammation and oxidative stress contributed by a thrombus (7, 10, 40, 41). Proteins shared among the three networks include matrix metalloproteinases MMP9 and MMP12, the metalloproteinase inhibitors TIMP1 and TIMP3, and myeloperoxidase (MPO) are identified in at least two the networks, and each protein has been identified previously in at least one proteomics-based aneurysm study (11). In addition, integrins including ITGB2 (CD18), ITGAM (CD11b), and ITGA2B are identified, supporting a role for leukocyte infiltration within aneurysmal tissues (35, 37, 42). The monocyte/macrophage marker CD14 is linked to LTF (lactoferrin) that has been shown to have inflammatory effects on macrophages (43–45).

We next combined all three sets of mouse and human aneurysmal proteomes, and created a combined consensus network (Figure 7; Supplementary Figure 5; Supplementary Table 8). The combined network contains a highly interconnected, central subnetwork comprised largely of ribosomal proteins and other transcription/translation-related proteins, and an outer subnetwork that comprises an extensive network of immunity, platelet activation, and integrin biology-related proteins (Figure 7). The proteome of the three mouse model datasets and the human nT-AAA contributed unique and overlapping nodes to the central subnetwork, whereas no contribution by human thrombus and T-AAA proteomes occurs (Figure 7; Supplementary Figures 5E–G). A more general assessment of this combined network leads to the following initial impressions: (1) a prominent deviation between the elastase model and any of the other networks due to its largest subnetwork localized to the right side of the network (Figure 4D); and (2) the consistent presence of inflammation and leukocyte migration-related pathways throughout all six networks.

To assess the degree of association between human tissue and mouse model subnetworks at the molecular level, we performed network closeness analyses (Figures 7B,C). Both analyses depict the highest level of similarity between all three aneurysm model groups and the nT-AAA human tissue type, likely attributed in part to the central transcription/translation subnetwork. In contrast to the human thrombus tissue, which has relatively high overlap with nearly all mouse groups, the disparity between T-AAA and all three mouse groups indicates this tissue type to be relatively unique to the human disease.