Comparative gene array analyses of severe elastic fiber defects in late embryonic and newborn mouse aorta

Abstract

Elastic fibers provide reversible elasticity to the large arteries and are assembled during development when hemodynamic forces are increasing. Mutations in elastic fiber genes are associated with cardiovascular disease. Mice lacking expression of the elastic fiber genes elastin (Eln^−/−), fibulin-4 (Efemp2^−/−), or lysyl oxidase (Lox^−/−) die at birth with severe cardiovascular malformations. All three genetic knockout models have elastic fiber defects, aortic wall thickening, and arterial tortuosity. However, Eln^−/− mice develop arterial stenoses, while Efemp2^−/− and Lox^−/− mice develop ascending aortic aneurysms. We performed comparative gene array analyses of these three genetic models for two vascular locations and developmental stages to determine differentially expressed genes and pathways that may explain the common and divergent phenotypes. We first examined arterial morphology and wall structure in newborn mice to confirm that the lack of elastin, fibulin-4, or lysyl oxidase expression provided the expected phenotypes. We then compared gene expression levels for each genetic model by three-way ANOVA for genotype, vascular location, and developmental stage. We found three genes upregulated by genotype in all three models, Col8a1, Igfbp2, and Thbs1, indicative of a common response to severe elastic fiber defects in developing mouse aorta. Genes that are differentially regulated by vascular location or developmental stage in all three models suggest mechanisms for location or stage-specific disease pathology. Comparison of signaling pathways enriched in all three models shows upregulation of integrins and matrix proteins involved in early wound healing, but not of mature matrix molecules such as elastic fiber proteins or fibrillar collagens.

INTRODUCTION

Elastic fibers provide reversible elasticity to the large, elastic arteries in vertebrate animals. During elastic fiber assembly, soluble tropoelastin is secreted from smooth muscle cells (SMCs) in the arterial wall. Tropoelastin coacervates into larger structures on the cell surface and then interacts with microfibrils composed of fibrillin proteins. Fibulin proteins serve to limit the coacervate size and assist in elastic fiber assembly (80). Lysyl oxidase enzymes cross-link tropoelastin to form mature insoluble elastin. Up to 30 different proteins are implicated in elastic fiber assembly (3), but only a few of them are absolutely required. The expression of elastin (Eln), fibulin-4 (Efemp2), and lysyl oxidase (Lox) is required for postnatal survival in mice. Eln^−/− mice have no arterial elastin and die soon after birth from overproliferation of SMCs that occludes the aortic lumen (60). Efemp2^−/− (32, 72) and Lox^−/− (31, 70) mice have severely disrupted arterial elastic fibers and die soon after birth with ruptured aortic aneurysms. In all three genetic knockout models, there is significant thickening of the aortic wall and a loss of reversible elasticity (47). Elastin mutations are associated with aortic stenoses (luminal narrowing) in humans (61), while fibulin-4 (19) and lysyl oxidase (28) mutations are associated with thoracic aortic aneurysms (diameter expansion).

Studies in mice and humans demonstrate similarities (i.e., wall thickening) and differences (i.e., stenoses vs. aneurysms) in the arterial response to severe elastic fiber defects due to absence of or mutations in elastin, fibulin-4, and lysyl oxidase. We reasoned that comparative gene array analyses of Eln^−/−, Efemp2^−/−, and Lox^−/− aorta may provide clues to the gene expression and signaling pathways associated with the common and divergent arterial phenotypes. Additionally, the phenotypes in mice and humans vary depending on the vascular section examined. For example, the ascending aorta (AA) in Efemp2^−/− (46) and Lox^−/− (100) mice is dilated compared with wild-type (WT) controls, but the descending aorta (DA) is not. In humans, ~60% of thoracic aortic aneurysms occur in the AA (35). SMCs in the AA and DA come from different embryonic origins, and this may play a role in disease susceptibility (69). Hence, we analyzed AA and DA regions separately in our genetic mouse models to determine the role that vascular location specific gene expression plays in the arterial response to severe elastic defects.

Elastin expression begins in late embryonic development, peaks in adolescence, and returns to low baseline levels in the adult mouse aorta (43). The short time period of elastin expression is possible because of the 74 yr half-life of the mature protein (95). Total elastin expression during embryonic and postnatal development correlates with increasing hemodynamic forces (blood pressure and flow), suggesting that the aortic wall adjusts elastin amounts to meet mechanical requirements (111). Eln^−/− mice show a minimal cardiovascular phenotype at embryonic day (E) 18 (108), but severe cardiovascular defects at postnatal day (P) 1 (109). We have suggested that the cardiovascular phenotype in Eln^−/− mice at P1 is partly due to increasing hemodynamic forces during late embryonic development (13). To investigate the role of hemodynamics and developmental changes in severe elastic defects, we analyzed aortas from all three genetic models at E17.5 and P1. We also examined arterial morphology and wall structure of P1 mice for all three genetic knockout models to confirm the expected arterial phenotypes and elastic fiber defects.

MATERIALS AND METHODS

Animals and tissue collection.

All animal protocols were approved by the Institutional Animal Care and Use Committee. Eln^+/− (60), Efemp2^+/− (32), and Lox^+/− (31) mice were mated to produce the associated wild-type (WT) (^+/+) and knockout (KO) (^−/−) pups to analyze expression patterns for three different genetic models of severe elastic fiber defects. Heterozygous mice were not used. Eln and Efemp2 mice are in the C57BL/6 background, while Lox mice are in the B6 albino background. The morning after mating was considered E0.5. Pups were euthanized by CO₂ inhalation and collected at either E17.5 or within 12 h of birth at P1 for analysis of gene expression differences by developmental stage. The thoracic aorta was removed and cut into two segments: 1) ascending aorta (AA) from the aortic valve to the left common carotid artery and 2) descending aorta (DA) from the ductus arteriosus to the diaphragm to analyze differences between vascular locations. The experimental design is shown in Table 1. The aortic segments were flash-frozen in liquid nitrogen and stored at −80°C.

Table 1.

Experimental design

Genetic Model	Genotype	Vascular Location	Developmental Stage	Aortic Segments/Sample	Total Data Sets
Eln	KO	DA	E17.5	8	48
Efemp2	WT	AA	P1	Samples/Group	Total Mice Used
Lox				2	384

Three different genetic models of severe elastic fiber defects (Eln, Efemp2, and Lox) were compared by genotype [knockout (KO), wild type (WT)], vascular location [descending aorta (DA), ascending aorta (AA)], and/or developmental stage [embryonic day (E)17.5, postnatal day (P)1]. Eight aortic samples were pooled per sample, and two samples were run for each group. Each genetic model was run on the same gene array chip, and comparisons were only made within genetic models to avoid batch effects.

Morphology and fluorescence microscopy.

For P1 mice, yellow latex (Ward’s Natural Science) was injected into the vascular tree to visualize morphology of the thoracic aorta (24). Morphology after latex infusion was similar to that observed without latex (108) or with a dye used instead of latex (109). Frozen sections of P1 AA were also prepared to visualize the aortic wall structure (58). Alexa Fluor 633 Hydrazide (Life Technologies) was used for elastin staining (16, 98). CNA35 (kindly provided by Magnus Hook, Texas A&M) labeled with Oregon Green 488 (Life Technologies) was used for collagen staining (55). Hoechst 34580 (Life Technologies) was used for nuclear staining. Images were taken on a Zeiss confocal microscope at ×40 magnification.

Gene expression studies.

Aortic segments were pooled in groups of eight, and two pooled samples were run for each group (Table 1). RNA was isolated with the RNeasy Plus Mini Kit (Qiagen). Only samples with 260/280 nm absorbance ratios of 1.8–2 were used for further analysis. RNA was processed for use on the Affymetrix mouse gene 2.0 array by the Genome Technology Access Center (GTAC) at the Washington University School of Medicine. RNA for each genetic model (Eln, Efemp2, Lox) was run on the same gene chip. Expression Console (Affymetrix) was used by GTAC for quality control and to generate gene expression values with robust multiarray average background correction, median polish summarization, and quantile normalization. Resulting files were uploaded into Partek Genomics Suite version 7.0 (Partek) or Gene Set Enrichment Analysis (GSEA, Broad Institute) for further analyses. All gene array data have been deposited in National Center for Biotechnology Information’s Gene Expression Omnibus (GEO) and are accessible through GEO Series accession number GSE120465. Statistical comparisons were only made within each genetic model (Eln, Efemp2, Lox) to eliminate batch effects.

Gene array data for Efemp2 (46) and Lox (100) P1 aortic segments were previously published, and genes of interest from those studies were confirmed by quantitative (q)PCR. To further confirm expression of genes of interest in the current study, qPCR was performed for 28 of the 48 total data sets (Table 1) for which additional RNA from the gene array studies (n = 2 each of WT and KO for Lox E17.5 AA, Lox E17.5 DA, Lox P1 AA, and Lox P1 DA) or additional aortic samples (n = 2 each of WT and KO for Efemp2 E17.5 DA, Efemp2 P1 DA, and Eln P1 DA) was available. For the additional aortic samples, eight segments were pooled/group, and RNA was isolated as described above for the gene array studies. RNA was transcribed with a High Capacity cDNA Reverse Transcription Kit (Applied Biosciences), and qPCR was performed on a QuantStudio 12K machine (Applied Biosystems) with TaqMan Fast Advanced Master Mix (Applied Biosystems). TaqMan Gene Expression Assays (Life Technologies) for Col8a1, Igfbp2, Thbs1, and B2m were used for primers, and all experiments were run in triplicate. The C_t from each triplicate was averaged, and the ∆C_t was calculated from the average expression level of B2m. Fold change for KO vs. WT was calculated via the 2^−ΔΔCt method (66), and the two pooled samples were averaged.

Individual gene expression analyses.

Gene array data for each genetic model were analyzed using a three-way ANOVA for genotype (WT/KO), developmental stage (E17.5/P1), vascular location (AA/DA), and all two-way interactions between independent variables using Partek software. Three-way interactions had minimal effects and were not included in the ANOVA statistical models. Lists were generated for genes that were differentially regulated by a fold-change of > 2 and a p-value with false discovery rate (FDR) step up < 0.05 for each comparison. Unidentified genes were removed from the gene lists. Gene lists that were differentially regulated in all three models by genotype, developmental stage, or vascular location were evaluated for overlap in hallmark and canonical pathways in the Molecular Signatures Database (MSigDB) at the website http://software.broadinstitute.org/gsea/msigdb/compute_overlaps.jsp to determine pathway changes that are associated with differential expression of the individual genes.

GSEA.

To further identify sets of genes in the array data that may be jointly regulated in genetic models of elastic fiber defects, GSEA was performed according to Subramanian et al. (101) and Mootha et al. (75). GSEA was run with 1,000 permutations, with a phenotype permutation type, and a weighted enrichment statistic separately for each genetic model. Gene sets were limited in size to 15–500 members. Enrichment of hallmark gene sets in KO vs. WT aorta was determined using a P value < 0.05 and FDR < 0.25 for all developmental stages and vascular locations combined (n = 8 in each group). The 50 hallmark gene sets represent biological processes that display coordinated expression and were curated from over 4,000 original overlapping gene sets in MSigDB (63). Enrichment of canonical pathways was also determined for each genetic model with a P value of < 0.05 and FDR < 0.1. The 1,321 canonical gene sets in MSigDB were curated from BioCarta, KEGG, Matrisome Project, Pathway Interaction Database (PID), Reactome, Sigma Aldrich, Signal Transduction KE, Signaling Gateway, and Superarray SABiosciences. Different FDRs were used to identify enrichment of hallmark and canonical gene sets due to the different size, nature, and redundancy of the gene sets. The leading edge genes from significantly enriched gene sets were compared across genetic models.

RESULTS

Arterial tortuosity, stenoses or aneurysms, and disorganized SMCs are characteristic of severe elastic fiber defects.

We visualized the morphology of P1 thoracic aorta by injecting yellow latex into the vasculature (Fig. 1). WT aorta has a consistent diameter along its length and the DA travels in a straight line down the spine. In contrast, Eln^−/− AA and DA have stenotic (narrow) regions, while Lox^−/− and Efemp2^−/− AAs are dilated (larger). The morphology results are consistent with previous quantitative reports showing a smaller diameter AA in Eln^−/− mice and a larger diameter, aneurysmal AA in Lox^−/− and Efemp2^−/− mice (47). The DA in Lox^−/− (100) and Efemp2^−/− (46) mice does not dilate. All three genetic models of severe elastic fiber defects show tortuosity (deviation from a straight line) in the DA and often in the major arterial branches. Wall structure of the AA was visualized with fluorescence microscopy (Fig. 2). Eln^−/− AA is lacking elastic fibers, while Lox^−/− and Efemp2^−/− AAs have fragmented elastic fibers. The collagen staining is inconsistent but may show more signal in the media when elastic fibers are absent or fragmented. The cell nuclei are more disorganized in the genetic models of severe elastic fiber defects, especially near the lumen. The wall structure results are consistent with previous reports showing wall thickening and disorganized cells in Eln^−/− (60, 109), Lox^−/− (31, 100), and Efemp2^−/− (32, 46) mouse aortas.

Fig. 1.

Gross aortic morphology for genetic models of severe elastic fiber defects in newborn mice. Wild-type (WT) thoracic aorta has a constant diameter, and the descending aorta (DA) follows a straight path (A). Eln^−/− mice have stenotic (narrow) regions in the ascending aorta (AA) and DA (arrowheads) and tortuosity (does not follow a straight path) in the DA (arrow) and major arterial branches (B). Lox^−/− and Efemp2^−/− mice have dilated (larger) AAs (arrowheads) and tortuous DAs (arrows) (C, D). Lox^−/− mice often have patent ductus arteriosus (*).The heart was removed for easier visualization in all but B. Scale bar is 500 µm. Images representative of n = 6–8 mice examined per group.

Fig. 2.

AA wall structure for genetic models of severe elastic fiber defects in newborn mice. Elastic fibers are visible throughout the medial wall thickness in WT AA (A) and are lacking in Eln^−/− AA (B) or fragmented in Lox^−/− (C) and Efemp2^−/− AAs (D). Collagen staining was inconsistent across and within individual samples (i.e., E, medial region) but generally showed high signal in the adventitial (top) and intimal (bottom) portions of the wall and less signal in the medial region (E–H). Cell nuclei are more disorganized, especially near the intima in AAs with elastic fiber defects (I–L). Composite images are shown in M–P. AAs with elastic fiber defects have thicker walls. Scale bar is 10 µm. Adapted from original images in Kim et al. (47) with permission from Elsevier. Images representative of n = 3–5 aortic sections examined per group.

Differentially regulated genes for three different genetic models of elastic fiber defects.

To determine the effects of severe elastic fiber defects on aortic gene expression in developing mice, we examined differentially regulated genes for each genetic model. There are 18–25 differentially regulated known genes in KO vs. WT aorta for each group (Fig. 3, Table 2). The top differentially regulated gene by fold change in each case is the model gene (Eln, Efemp2, or Lox), as expected. Three genes, Col8a1, Igfbp2, and Thbs1, are upregulated in KO vs. WT aorta in all three models and are members of the Naba core matrisome and Naba matrisome gene sets, indicating common extracellular matrix (ECM) remodeling in response to the loss of Eln, Efemp2, and Lox. Relative expression levels in KO vs. WT aorta for Col8a1, Igfbp2, and Thbs1 were confirmed by qPCR in 28 of the 48 total data sets for which additional mRNA or aortic segments were available (Fig. 4). One gene, Itga11, is upregulated in KO vs. WT aorta for Eln and Efemp2 models only. Six genes are differentially regulated in KO vs. WT aorta for Eln and Lox models only. Pi15 and Clca3a2 are downregulated, while Itgbl1, Gna14, Wnt16, and Adamts8 are upregulated. Three genes, Ptgs2, Col11a1, and Serpine1, are upregulated in KO vs. WT aorta for Efemp2 and Lox models only.

Fig. 3.

Venn diagram of differentially regulated known genes in knockout (KO) vs. wild type (WT) for all three models. Differentially regulated genes were determined for each genetic model by three-way ANOVA in Partek software with P < 0.05 and a fold change > 2. The targeted gene (Eln, Lox, or Efemp2) is differentially regulated for each model, as expected. Three genes, Col8a1, Igfbp2, and Thbs1, are differentially regulated in all three models. A summary of the P values and fold changes are given in Table 2. n = 2 of eight pooled aortic samples for each group.

Table 2.

Differentially regulated known genes for KO vs. WT for each genetic model

Eln			Efemp2			Lox
Gene Symbol	P Value	Fold-change	Gene Symbol	P Value	Fold-change	Gene Symbol	P Value	Fold-change
Eln	8.65E-09	−73.2	Efemp2	7.38E-09	−9.26	Lox	6.92E-07	−6.26
Pi15	9.50E-06	−2.74	Col8a1	1.43E-07	7.66	Clca3a2	3.08E-05	−2.31
Gm22137	5.12E-06	−2.87	Serpina3n	2.50E-06	6.39	Cdr1	2.00E-05	−2.13
Clca3a2	1.74E-05	−2.90	Saa3	3.70E-05	6.10	Pi15	0.000119	−2.00
Col8a1	1.02E-05	4.95	Igfbp2	1.13E-05	4.89	Col8a1	1.78E-07	5.70
Igfbp2	4.30E-08	4.21	Thbs1	8.44E-06	4.27	Ptgs2	0.000166	2.89
Itgbl1	4.16E-06	3.42	Hmox1	3.58E-06	3.81	Serpine1	4.57E-07	2.78
Chad	2.87E-05	3.15	Serpine1	5.80E-06	3.67	Igfbp2	2.61E-07	2.77
Crabp2	1.88E-06	3.11	Ptgs2	2.89E-05	3.45	Thbs1	2.43E-05	2.58
Gna14	2.94E-06	3.10	Clec4d	4.30E-06	2.97	Esm1	0.000277	2.56
Moxd1	3.03E-06	2.82	Col11a1	6.56E-07	2.74	Egr1	4.88E-05	2.56
Itga11	3.21E-06	2.63	Ctss	4.70E-06	2.71	Gna14	6.15E-07	2.48
Tenm3	0.000159	2.43	Cd109	1.69E-05	2.58	Wisp2	2.11E-05	2.47
Ltbp2	6.67E-06	2.30	Itga11	2.34E-05	2.47	Gm5416	0.000271	2.30
Wnt16	1.04E-05	2.29	Col12a1	1.78E-07	2.09	Col11a1	4.26E-06	2.24
Adamts8	4.73E-06	2.27	Svep1	7.08E-07	2.09	Wnt16	2.03E-07	2.13
Ndp	6.78E-06	2.26	Figf	3.08E-05	2.07	Itgbl1	9.99E-07	2.08
Glp1r	2.70E-05	2.18	Plaur	1.18E-05	2.03	Pyhin1	0.00018	2.07
Uchl1	0.000129	2.14				Adamts8	2.95E-07	2.02
Dsp	0.000164	2.13				Akap5	0.000115	2.00
Nell2	3.79E-05	2.07				Ier3	6.45E-05	2.00
Ctgf	4.06E-05	2.07
Cck	3.52E-06	2.06
Fjx1	0.000141	2.06
Thbs1	1.45E-05	2.05

Differentially regulated genes were determined by three-way ANOVA for genotype, vascular location, and developmental stage with P < 0.05 and a fold change > 2 for each comparison. P values and fold-changes with KO vs. WT for each model are shown. The targeted gene (Eln, Lox, or Efemp2) is the top differentially regulated for each model. Overlaps between different models are shown in Fig. 3; n = 2 of eight pooled aortic samples for each group.

Fig. 4.

Comparison of gene array and quantitative (q)PCR results for the three genes differentially regulated in all three models. Average fold change for KO vs. WT was determined for each combination of genetic model (Eln, Efemp2, and Lox), vascular location (AA, DA), and developmental stage (E17.5, P1) for the gene array and available qPCR data (individual points) for Col8a1 (A), Igfbp2 (B), and Thbs1 (C). The composite average and standard deviation (bars) for the fold changes are consistent between the gene array and qPCR data. n = 2 of eight pooled aortic samples for each group. E, embryonic day; P, postnatal day.

Differentially regulated genes by developmental stage and vascular location.

Gene expression data in each genetic model were also analyzed for differences between vascular location, developmental stage, and interactions between independent variables. Developmental stage caused the largest change in gene expression with 687–1,211 known genes differentially regulated in E17.5 vs. P1 aorta. Vascular location caused differential regulation of 38–122 known genes in DA vs. AA. Genes that are differentially regulated by developmental stage or vascular location in all three genetic models in the same direction are listed in Tables 3 and 4, respectively. Thirty-eight genes are differentially regulated in the same direction by developmental stage in all three models. Three of the differentially regulated genes, C3, C4B, and Cfd, are members of the BioCarta complement pathway and the KEGG complement and coagulation cascade, indicating alterations in the thrombotic pathway that depend on developmental stage regardless of genetic defects in the elastic fibers. Eleven genes are differentially regulated in the same direction by vascular location in all three models. Five of the differentially regulated genes, Figf, Ccl11, Hhip, Chrdl1, and Ptn, are members of the Naba secreted factors, Naba matrisome associated, and Naba matrisome gene sets, indicating regulation of the ECM remodeling response that depends on vascular location. Only Ptn is regulated by both developmental stage and vascular location in all three models. Four known genes are differentially regulated by interactions between independent variables in a single genetic model. Vps28 and Arap2 expression shows significant interactions between genotype and vascular location for the Eln group. P4ha1 and Mustn1 show significant interactions between genotype and age for the Lox group (Table 5).

Table 3.

Differentially regulated known genes for E17.5 vs. P1 common to all three genetic models

	Eln		Efemp2		Lox
Gene Symbol	P Value	Fold-change	P Value	Fold-change	P Value	Fold-change
Bst2	5.97E-05	−2.25	0.001919	−2.35	1.55E-07	−2.91
C3	3.33E-07	−5.60	8.94E-06	−2.47	1.27E-07	−4.12
C4b	0.000181	−2.68	0.00031	−2.20	3.81E-10	−2.78
Car3	5.66E-06	−9.90	0.005458	−3.11	1.69E-08	−9.17
Cd14	5.35E-05	−2.20	0.004719	−2.06	2.98E-05	−2.34
Cfd	0.001355	−4.26	0.000537	−2.67	0.00189	−2.53
Cox8b	2.94E-05	−6.18	0.005716	−2.96	6.02E-05	−2.98
Fmo2	7.36E-05	−2.023	0.000538	−2.29	1.85E-06	−3.15
Fabp7	0.00198	3.08	0.000132	3.99	6.89E-05	2.77
Gm23666	3.32E-05	3.44	0.001206	2.08	1.40E-05	2.1
Gm26080	1.45E-05	2.46	0.000186	2.13	0.000617	2.27
Gm26289	7.15E-06	4.23	0.003806	2.07	6.26E-07	2.16
Hba-x	4.39E-05	2.51	9.53E-05	4.014	3.42E-06	3.39
Ibsp	0.006093	2.61	0.000391	2.61	0.000316	2.73
Ifi27l2a	1.12E-08	−7.19	7.96E-05	−4.23	1.04E-05	−3.87
Iigp1	0.000257	−2.80	0.00339	−2.36	2.12E-05	−2.39
Ly6a	0.000435	−2.19	1.34E-05	−2.18	1.15E-05	−2.63
Mir142hg	0.003426	−2.64	0.000272	−5.08	2.34E-05	−6.54
Mir434	0.001561	2.44	4.26E-06	3.59	4.05E-06	3.49
Mir680–1	0.001535	−4.08	0.000427	−11.38	0.000622	−3.02
Mir98	3.51E-06	2.77	0.000137	2.13	0.000103	2.01
Mt1	1.41E-06	−2.74	2.09E-05	−2.39	2.23E-08	−3.80
Mt2	9.83E-06	−3.74	1.34E-05	−3.46	4.36E-06	−5.29
mt-Ta	4.23E-05	−8.13	1.76E-07	−7.66	0.014892	−4.35
mt-Tf	1.96E-05	−4.57	0.001107	−2.98	1.25E-08	−2.80
mt-Tk	0.006654	−2.24	0.002285	−5.18	0.003794	−2.21
mt-Tt	0.000383	−4.19	0.000125	−5.34	5.52E-05	−4.58
mt-Tv	1.31E-06	−8.63	9.98E-05	−4.13	6.30E-05	−2.58
Nefm	0.008916	2.48	0.000151	2.34	0.000255	2.03
Olfr726	4.95E-06	7.70	0.001217	3.16	4.48E-07	3.07
Pdk4	2.07E-05	−2.76	3.67E-07	−2.65	1.57E-08	−5.52
Plin4	2.43E-06	−2.01	0.000599	−2.16	4.87E-07	−2.43
Ptn	1.21E-06	2.62	5.12E-06	2.20	1.26E-08	2.21
Saa3	2.31E-06	−8.95	0.006383	−2.35	0.005682	−3.72
Serpina3n	0.001038	−4.01	8.83E-05	−3.29	6.46E-05	−5.56
Snora28	0.000297	−6.63	0.004147	−2.47	0.00064	−2.69
Ucp1	0.003894	−7.23	0.003193	−2.87	0.003229	−2.48

Differentially regulated genes were determined by three-way ANOVA for genotype, vascular location, and developmental stage with P < 0.05 and a fold change > 2 for each comparison. P values and fold-changes for E17.5 vs. P1 in each model are shown for the 37 known genes in alphabetical order that are differentially regulated in the same direction for all three genetic models; n = 2 of eight pooled aortic samples for each group.

Table 4.

Differentially regulated known genes for DA vs. AA common to all three genetic models

	Eln		Efemp2		Lox
Gene Symbol	P Value	Fold-change	P Value	Fold-change	P Value	Fold-change
Ccl11	3.10E-05	3.14	7.14E-06	4.21	1.34E-06	2.93
Chrdl1	0.00017	2.78	1.33E-06	3.51	4.97E-07	3.26
Elmod1	0.000158	2.36	6.57E-05	2.21	3.91E-05	2.98
Figf	4.62E-05	3.11	1.47E-06	2.87	1.15E-06	2.63
Hhip	1.74E-05	3.02	2.73E-06	4.01	2.11E-08	3.53
Hoxb6	0.000138	4.55	9.24E-07	3.18	1.06E-06	7.50
Hoxb5	2.57E-05	2.42	4.71E-05	2.31	2.00E-07	3.70
Mira	2.95E-05	10.75	1.83E-09	15.0	2.01E-10	17.9
Osr2	2.91E-06	−2.92	1.97E-06	−3.51	1.18E-07	−3.56
Ptn	5.02E-07	−2.91	7.15E-08	−3.68	4.06E-10	−3.21
4833423E24Rik	9.67E-05	2.80	2.54E-05	2.335	1.29E-06	3.46

Differentially regulated genes were determined by three-way ANOVA for genotype, vascular location, and developmental stage with P < 0.05 and a fold change > 2 for each comparison. P values and fold-changes for DA vs. AA in each model are shown for the 11 known genes in alphabetical order that are differentially regulated in the same direction for all three genetic models; n = 2 of eight pooled aortic samples for each group.

Table 5.

Known genes differentially regulated by significant interactions between independent variables

Eln		Lox
Gene Symbol	P Value(genotype × location)	Gene Symbol	P Value(genotype × age)
Arap2	1.81E-06	Mustn1	4.49E-06
Vps28	1.62E-06	P4ha1	4.53E-06

Significant interactions between two independent variables (genotype, vascular location, and developmental stage) were determined by three-way ANOVA with P < 0.05 and a fold change > 2 for each comparison. Significant two-way interactions were found for genotype × location for two genes in the Eln group and for genotype × age for two genes in the Lox group; n = 2 of eight pooled aortic samples for each group.

Comparative gene set enrichment results for hallmark gene sets.

GSEA with P < 0.05 and FDR < 0.25 was performed to determine enrichment of hallmark gene sets in KO vs. WT aorta for each genetic model. Hallmark gene sets summarize thousands of overlapping gene sets and provide a basis for GSEA of biological states with coherent expression (63). A summary of the differentially regulated hallmark gene sets is presented in Fig. 5. Complete statistics are given in Supplemental Table S1. (The online version of this article contains supplemental material.) Four hallmark gene sets, IL2 STAT5 signaling, inflammatory response, estrogen response early, and coagulation, are differentially regulated in KO vs. WT aorta for all three groups. The leading edge genes in these gene sets that are common for all three genetic models are shown in Supplemental Table S2. Xenobiotic metabolism is differentially regulated in KO vs. WT aorta for Eln and Efemp2 groups. Fifteen of the 50 hallmark gene sets are differentially regulated in KO vs. WT aorta for Efemp2 and Lox groups. Four hallmark gene sets are differentially regulated in only Efemp2 KO vs. WT aorta, and nine hallmark gene sets are differentially regulated in only Lox KO vs. WT aorta. There are no hallmark gene sets that are uniquely regulated in Eln only or in Eln and Lox only groups.

Fig. 5.

Venn diagram of enriched hallmark pathways in KO vs. WT for all three models. Enriched hallmark pathways were determined for each genetic model with Gene Set Enrichment Analysis software with P < 0.05 and false discovery rate < 0.25 for all vascular locations and developmental stages combined. A summary of the enrichment statistics is given in Supplemental Table S1. There are four pathways enriched in KO vs. WT for all three genetic models. The leading edge genes common to these enriched pathways are listed in Supplemental Table S2. There are 15 pathways commonly enriched in KO vs. WT for Lox and Efemp2 models that may play a role in the aortic aneurysm phenotype observed in Lox^−/− and Efemp2^−/− mice. n = 2 of eight pooled aortic samples for each group.

Comparative gene set enrichment results for canonical pathway gene sets.

GSEA with P < 0.05 and FDR < 0.1 was performed to determine enrichment of canonical gene sets in KO vs. WT aorta for each genetic model. The 16 canonical gene sets that are enriched in all three genetic models are summarized in Table 6. All of the enriched pathways have at least one ECM or integrin gene in the list of leading edge genes and are divided into ECM- or integrin-associated groups in Table 6, depending on the number of each type in the leading edge list. Complete enrichment statistics are given in Supplemental Table S3. The leading edge genes in these gene sets that are common in all three genetic models are shown in Supplemental Table S4. Five ECM-specific pathways are enriched including Naba collagens, Naba core matrisome, Naba ECM glycoproteins, KEGG ECM receptor interaction, and Reactome ECM organization. PID Syndecan 1 and PID Syndecan 4 pathways are also enriched, highlighting a possible interaction between elastic fibers and these two heparan sulfate proteoglycans. The genes in the leading edge lists for the ECM-specific and syndecan pathways are associated with “early provisional” matrix molecules implicated in the initial stages of wound healing and hemostasis (i.e., Fga, Fgg, F2), rather than “late provisional” or mature matrix molecules (i.e., Fn1, Col1a1) (5) associated with later stages of wound healing and vascular remodeling in hypertension (34).

Table 6.

Summary of enriched canonical pathways in KO vs. WT for all three genetic models

ECM Associated	Integrin Associated
Naba collagens	KEGG arrythmogenic right ventricular cardiomyopathy ARVC
Naba core matrisome	KEGG ECM receptor interaction
Naba ECM glycoproteins	KEGG leukocyte transendothelial migration
PID syndecan 1 pathway	PID integrin A9B1 pathway
PID syndecan 4 pathway	PID integrin CS pathway
Reactome integrin cell surface interactions	PID integrin 1 pathway
Reactome platelet aggregation plug formation	PID integrin 3 pathway
	PID integrin 5 pathway

Enriched canonical pathways were determined for each genetic model with GSEA software with P < 0.05 and FDR < 0.25 for all vascular locations and developmental stages combined. Enrichment statistics are given in Supplemental Table S3. The enriched pathways are divided into extracellular matrix (ECM)- or integrin-associated groups depending on the number of ECM- or integrin-associated genes in the common leading-edge gene lists (Supplemental Table S4).

Six integrin-specific canonical pathways are enriched in KO vs. WT aorta in all three genetic models including Reactome integrin cell surface interactions, PID integrin A9B1, PID integrin CS, PID integrin 1, PID integrin 3, and PID integrin 5. Although KEGG arrhythmogenic right ventricular cardiomyopathy (ARVC) is not an integrin-specific pathway, seven of 11 of the leading-edge genes code for integrin subunits. For KEGG leukocyte transendothelial migration, three of the 16 leading edge genes code for integrin subunits, implicating an integrin-mediated inflammatory response due to severe elastic fiber defects. The physiological roles of the integrin subunits contributing to the leading-edge gene lists for these enriched canonical pathways are summarized in Table 7.

Table 7.

Summary of integrin subunits present in the leading edge gene lists for commonly regulated canonical pathways (Supplemental Table S4)

Gene Symbol	Binding Partners and Function
Itga1	pairs with β1 and is receptor for fibrillar collagen and laminin-1
Itga11	pairs with β1 and is receptor for collagen
Itga4	pairs with β1 and is receptor for fibronectin, VCAM-1, and ICAM-4
Itga5	pairs with β1 and is receptor for fibronectin, and LAP-TGF-β
Itgav	pairs with β1, β3, β5, β6, β8 to act as receptors for wide variety of proteins
	αVβ1 = receptor for fibronectin, vitronectin, and LAP-TGF-β
	αVβ3 = receptor for fibronectin, vitronectin, fibrinogen, vWF, thrombospondin, collagen, LAP-TGF-β, fibrillin-1, tropoelastin
	αVβ5 = receptor for fibronectin, vitronectin, LAP-TGF-β, tropoelastin
Itgam	pairs with β2 to form MAC-1 (also known as CR3) allowing adherence of neutrophils and monocytes to stimulated endothelium
Itgal	pairs with β2 to form LFA-1 expressed in leukocytes
Itgb2	pairs with αL or αM to modulate adhesion of immune cells
Itgb3	pairs with αIIb in platelets and acts as a receptor for fibrinogen, vWF, and fibronectin
Itgb5	pairs with αV and is receptor for fibronectin, vitronectin, LAP-TGF-β, tropoelastin

The gene symbol for each integrin subunit is given along with the partner(s) and function(s).

DISCUSSION

We examined arterial morphology and wall structure in newborn mice to confirm that the absence of elastin, fibulin-4, or lysyl oxidase expression leads to arterial malformations and severe elastic fiber defects. Consistent with previous results, Eln^−/− mice have stenotic arteries, while Efemp2^−/− and Lox^−/− mice have dilated AAs. All three genetic knockout models have tortuous arteries and absent or fragmented elastic fibers in the arterial wall. We performed comparative gene array analyses of the three genetic models to examine the effects of severe elastic fiber defects, vascular location, and developmental stage on gene expression. Similarities and differences in the gene expression patterns may provide clues to mechanisms of elastic fibers disease, vascular location disease susceptibility, and regulation of aortic development.

Similarities and differences in gene expression for genetic models of severe elastic fiber defects.

The three genes, Col8a1, Igfbp2, and Thbs1, that are differentially regulated in KO vs. WT for all three genetic models (Figs. 3 and 4) may be important in elastic fiber assembly or for common vascular phenotypes, such as SMC overproliferation (31, 32, 60) and aortic wall thickening (47). Although all three genes are associated with alterations in SMC phenotype, only Col8a1 has previously been identified as an elastic fiber associated molecule (45). Col8a1 is a short chain collagen expressed by vascular endothelial cells (118) and SMCs (68) that localizes to the aortic intima and media (94). Col8a1 is upregulated in vascular injury and disease and promotes SMC migration (1). Igfbp2 is an IGF binding protein in the plasma. It is expressed by vascular endothelial cells (76) and enhances IGF-induced SMC migration and proliferation (97). Thbs1 associates with the ECM and is synthesized by fibroblasts, endothelial cells (ECs), and SMCs. Thbs1 modulates cell adhesion, EC proliferation, TGF-β1 activity, and signaling from inflammatory cells (65). The commonly regulated genes suggest that severe elastic fiber defects promote SMC and EC proliferation and migration.

The genes that are similarly regulated only among two models may be important for specific roles in elastic fiber assembly or phenotypes observed only in those models. For example, Itga11 is differentially regulated in Eln and Efemp2 KO vs. WT aorta, but not Lox (Fig. 3). Itga11 forms a complex with integrin beta 1 and is a collagen type 1 receptor in mesenchymal stem cells and a subset of fibroblasts (120). Itga11 mediates fibroblast-ECM-cardiomyocyte interactions in cardiac fibrosis (15) and collagen remodeling in periodontal fibroblasts (4). Hence, although fibrillar collagens are not differentially regulated in any of the genetic models, the interaction between cells and collagen may be altered through differential expression of Itga11 in Eln and Efemp2 deficiency.

The six genes differentially regulated in KO vs. WT aorta for Eln and Lox models, but not Efemp2 include Pi15, Clca3a2, Itgbl1, Gna14, Wnt16, and Adamts8 (Fig. 3). Pi15 is a trypsin inhibitor that is a candidate gene for the internal elastic laminar ruptures in the abdominal aorta of the Brown Norway rat (22). Clca3a2 is a calcium-activated intracellular chloride channel. Altered Clca3a2 expression is associated with closure of the ductus arteriosus (96), which also depends on inhibition of elastic fiber assembly through degradation of Lox (119). Itgbl1 is an ECM protein that has been implicated in ovarian (102), breast (62), and lung cancer (25). Gna14 encodes a member of the alpha q subfamily of G proteins. Mutations in Gna14 are associated with sporadic vascular tumors (64) and hypertension (49). Elastin insufficiency is associated with hypertension in humans (53) and mice (54). Wnt16 is a member of the WNT gene family that plays a role in oncogenesis and multiple developmental processes. Wnt16 has been implicated in chondrogenic transformation of SMCs in vascular calcification (7). Elastin insufficiency also modulates vascular calcification (44). Adamts8 encodes a member of the ADAMTS family that share common features including a disintegrin domain, zinc metalloproteinase domain, and a thrombospondin motif. Adamts8 is able to cleave aggrecan (17) and has been implicated in monocyte to macrophage differentiation in atherosclerosis (112) and cancer progression (11, 84). Commonly regulated genes in Eln and Lox models show that the absence of properly crosslinked elastic fibers stimulates expression of genes important for ECM remodeling in developmental processes, vascular disease, and cancer.

The three genes differentially regulated in Efemp2 and Lox models, but not Eln, include Ptgs2, Col11a1, and Serpine1 (Fig. 3). Differential expression of these genes may be important for expansion of the aortic diameter and aneurysm development that is associated with mutations in Efemp2 (93) and Lox (28), but not with Eln (20). Ptgs2, also known as Cox2, is responsible for prostanoid biosynthesis in inflammation and mitogenesis. Ptgs2 is differentially regulated in ECs by fluid shear stress (92) and in SMCs by substrate stiffness (52), suggesting that Ptgs2 is a mechanosensitive gene in multiple vascular cell types. Ptgs2 has been implicated in thoracic (99) and abdominal (30) aortic aneurysms. Col11a1 is a minor fibrillar collagen that is highly expressed in cartilage (8) and has been linked to cancer progression (88). Increased Col11a1 expression is observed in vascular calcification (10). Serpine1 encodes PAI1, which negatively regulates fibrinolysis and impairs the dissolution of clots. Serpine1 is upregulated in aortic dissection (48) and in aging-related pathologies (107). In the Efemp2 and Lox models, we find upregulation of two genes (Serpine1 and Ptgs2) that have previously been associated with aneurysm formation and one gene (Col11a1) that supports the hypothesis that chondrogenic differentiation of SMCs may be a phenotypic response to severe defects in elastic fibers.

Similarities and differences in gene expression by developmental stage and vascular location.

Developmental stage causes the largest change in gene expression and has the largest overlapping set of differentially expressed genes (37) in all three genetic models of elastic fiber defects (Table 3). For brevity, we will not discuss the individual genes that are regulated by developmental stage. However, three of the 37 genes are implicated in coagulation signaling pathways, indicating that thrombosis may be a common response to severe elastic fiber defects and that the thrombotic response may be enhanced with advancing developmental stage.

Susceptibility to vascular diseases including aneurysms (27) and atherosclerosis (104) is dependent on vascular location, so we examined differential gene expression in DA vs. AA segments to highlight molecular mechanisms underlying these differences. Vascular location has 11 genes that are differentially regulated in all three genetic models of severe elastic fiber defects (Table 4). Five of these 11 genes have previously shown vascular location-specific expression patterns or been implicated in vascular location-specific diseases. Hoxb5 and Hoxb6 are sequence-specific transcription factors involved in development. Hoxb5 regulates differentiation of ECs (116), enhances EC and monocyte migration, and increases leucocyte infiltration in ischemic tissues (23). Eleven homeobox genes, including Hoxb6, are differentially expressed in the aortic arch compared with the descending thoracic aorta in adult mice (104). Ptn is the only gene differentially regulated by developmental stage and vascular location in all three genetic models. Ptn codes for heparin binding growth factor, which is secreted from ECs and affects cell growth, fibrinolytic activity, and angiogenesis (77). Ptn is expressed in the atheroprone coronary artery and not in the atheroresistant internal mammary artery (86). Ccl11, also known as eotaxin, is thought to be involved in eosinophilic inflammatory diseases. Circulating Ccl11 is increased in patients with calcified atherosclerotic plaques and increases SMC calcification in vitro (87). Elmod1 acts as a GTPase-activating protein toward guanine nucleotide exchange factors in the Arf family that regulate the actin cytoskeleton (36). Elmod1 mutations are associated with hearing and balance deficits in mice due to defects in the actin cytoskeletal dynamics of hair cell stereocilia (38). Primary cilia on ECs in the aorta are differentially distributed by vascular location and are most abundant in atheroprone locations, suggesting that Elmod1 may play a role in actin dynamics in EC cilia and disease susceptibility (106).

The other six genes differentially regulated by vascular location are associated with vascular development and angiogenesis but have not previously been associated with vascular location-specific phenotypes. Osr2 is a transcription factor expressed at sites of epithelial to mesenchymal interactions during development (56). Hhip is a vertebrate-specific inhibitor of hedgehog signaling that interacts with all three hedgehog family members. Hhip is abundantly expressed in ECs and is regulated during angiogenesis (79). Figf, also known as Vegfd, is a member of the platelet-derived growth factor/vascular endothelial growth factor (PDGF/VEGF) family and is important for angiogenesis, lymphangiogenesis, and EC growth. It also stimulates myofibroblast growth and collagen synthesis (121). Chrdl1 codes an antagonist of Bmp4. Chrdl1 is upregulated by hypoxia (42) and plays a role in cornea organogenesis (82). Little is known about the function of 4833423E24Rik or Mira. Six of the 11 genes differentially regulated by vascular location are part of the ECM gene set subgroup Naba secreted factors, indicating that elastic fiber defects stimulate signaling pathways that lead to vascular location-specific ECM remodeling and may be associated with location-specific diseases, such as atherosclerosis and aneurysms. Together these results support previous observations on vascular location-specific gene regulation, as well as suggest new avenues of investigation.

Four genes show significant interactions between genotype and vascular location (Vps28 and Arap2 in the Eln group) or between genotype and developmental stage (P4ha1 and Mustn1 in the Lox group) (Table 5). The interactions suggest that elastic fiber defects in combination with vascular location or developmental stage determine the gene expression level. Vps28 codes a subunit of an endosomal complex, Escrt1, required for transport and sorting of proteins (89). Arap2 associates with focal adhesions and regulates focal adhesion dynamics downstream of RhoA. Arap2 signals through Arf6 (which also interacts with Elmod1 discussed above) to control focal adhesion morphology and actin cytoskeleton arrangement (12). The changes in Arap2 suggest that the cellular cytoskeleton and connections with the ECM are modulated by elastic fiber defects in combination with vascular location. P4ha1 catalyzes the formation of 4-hydroxyproline, which is essential to the proper folding of newly synthesized procollagen chains. Although additional fibrillar collagen expression is not detected in Lox^−/− aorta, additional P4ha1 expression encourages formation of collagen fibers. Upregulation of P4ha1 is also observed in hypoxic fibroblasts and modulates ECM remodeling (26). Mustn1 is a musculoskeletal protein involved in development and muscle control and is upregulated in lengthening exercised muscle (51), prompting speculation that it may also play a role in SMC response to lengthening due to increasing wall stresses during development and reduced stress shielding by defective elastic fibers.

Commonly regulated hallmark gene sets.

Hallmark gene sets have been designed to reduce overlap between different canonical gene sets and highlight genes that have coherent expression patterns (63). Because there are a limited number of gene sets in the hallmark group and they have been rigorously curated, we used a higher FDR than we did with the canonical pathways analyses to generate hypotheses for signaling pathways that may be investigated in future research. Two of the hallmark pathways that are differentially regulated in all three genetic models (Fig. 5) are associated with inflammation (IL2 STAT5 signaling and inflammatory response), highlighting the role of inflammation that has traditionally been associated with abdominal aortic aneurysms but has only recently been appreciated in thoracic aortic aneurysms associated with elastic fiber defects (83). While adaptive immunity is considered immature before birth, the innate immune system, including activities by innate immune cells, such as macrophages, are active throughout development (57). Macrophages are abundant in the aortic adventitia (21) and intima (91). The immune signatures we observe changing may thus relate to macrophage activity. Differential regulation of the estrogen response pathway in all three genetic models indicates that estrogen and sex differences may play a role in elastic fiber disease. Previous studies have shown that sex has significant effects in cardiovascular diseases associated with elastic fiber insufficiency including arterial stiffening (18), hypertension (78), and aortic aneurysms (71). For this study, males and females were combined and we did not track how many of each sex were included in the pooled aortic samples. If unequal numbers of males and females were included in each group, our results could be skewed by sex rather than elastic fiber defects. However, for the effect of sex to be consistent in KO vs. WT comparisons for all three genetic models, the 24 KO data sets would have to be skewed by unequal numbers of males and females in the same direction compared with the 24 WT data sets, which is unlikely. The coagulation pathway is differentially regulated in all three genetic models and includes many genes necessary for ECM remodeling. It is unclear if differential regulation of this pathway is associated with coagulation differences, ECM remodeling, or both. We have observed hematomas and areas of blood pooling in some of the dissected aortas (100), which may reflect alterations in the coagulation pathway due to local defects in the arterial wall caused by compromised elastic fibers.

Although individual numbers of differentially regulated genes are approximately the same for each genetic model and for overlaps between two of the three models (Fig. 3), the hallmark gene sets highlight the differences in gene expression of the Efemp2 and Lox groups, compared with the Eln group (Fig. 5). Fifteen hallmark gene sets are differentially regulated in both Efemp2 and Lox groups, compared with four gene sets common to all three groups, one gene set common to Efemp2 and Eln only, and no gene sets common to Eln and Lox only. The commonly regulated hallmark gene sets for Efemp2 and Lox may provide clues for molecular mechanisms of aneurysm formation observed in these two genetic models.

Commonly regulated canonical gene sets.

The canonical pathways enriched in KO vs. WT for all three genetic models of severe elastic fiber defects show a general upregulation of ECM genes (Table 6). The upregulated ECM genes are associated with hemostasis and early wound healing responses (5) (Supplemental Table S4). These genes include F2, which encodes coagulation factor II that is cleaved to form thrombin, and Fga and Fgg, the alpha and gamma chains, respectively, of fibrinogen, which is converted by thrombin to fibrin to form a clot. Increased plasma fibrinogen is present in patients with abdominal aortic aneurysms (2), and activated coagulation state correlates with abdominal aneurysm size (113). Our results suggest that the hemostatic cascade is an early wound healing response common to severe elastic fiber defects in the vascular wall and is not specific to abdominal aortic aneurysms. There may be a bimodal effect of hemostasis on elastic fiber integrity as elastogenesis is encouraged in neonatal SMCs embedded in cardiovascular tissue equivalents composed of fibrin compared with collagen type I (67), but increasing intraluminal thrombus thickness correlates with elastin degradation and abdominal aorta aneurysm growth in adults (50), suggesting that elastic fiber formation and degradation are both linked to thrombus formation.

Fibrillar collagens (90) and prominent elastic fiber molecules (110) are not present in the leading-edge genes for the differentially regulated canonical pathways common to all three genetic models (Supplemental Table S4). The collagen genes that are included in the leading-edge groups include Col8a1 and Col11a1 (discussed above) and Col12a1. Col12a1 is a member of the FACIT (fibril-associated collagens with interrupted triple helices) family and is found in association with collagen type 1 (90). Col12a1 expression is induced by shear stress in ECs (37) and is rapidly and reversibly induced by tensile stress in fibroblasts (103), indicating that Col12a1 expression changes in this study may be related to altered mechanical forces on the vascular wall due to severe elastic fiber defects.

The three collagen genes are the only leading-edge genes for Naba collagens and PID syndecan 1 pathways. The PID syndecan 4 pathway is also enriched in all three genetic models of elastic fiber defects, but the matrix-associated genes in the leading-edge gene group are F2, Thbs1, and Tnc. Syndecans 1 and 4 are transmembrane heparan sulfate proteoglycans that are receptors for many types of proteins including cytokines, growth factors, enzymes, ECM glycoproteins, and collagens. They are connected to the actin cytoskeleton and can regulate cell adhesion and migration. They may also regulate stretch-activated ion channels (74). Heparan sulfate proteoglycans (9) bind tropoelastin and are associated with poorly cross-linked elastic fibers (81). Syndecan-1 plays a protective role in abdominal aortic aneurysm formation by modulating inflammation (117). Syndecan-4 limits neointimal formation after vascular injury (33). Syndecan 1 and 4 are upregulated by mechanical strain (40, 41). The enrichment of the PID syndecan I and 4 pathways supports the hypothesis that severe elastic fiber defects are sensed as a vascular wall injury and mechanical insult.

Most of the enriched canonical pathways have integrin subunits as members of their leading-edge genes, as summarized in Supplemental Table S4. Many of the integrin alpha subunits that are represented pair with integrin β1 to bind to collagen, laminin, fibronectin, vitronectin, and/or LAP-TGF-β (14). Integrin β1 itself is not represented in the leading-edge genes, suggesting regulation of specific integrin pairs for targeted remodeling of ECM molecules. Two of the integrin alpha subunits, Itgam and Itgal, pair with integrin β2, which is also represented in the leading-edge genes to modulate monocyte and neutrophil phagocytosis and adhesion (α_Mβ₂) or macrophage maturation and leukocyte adhesion (α_Lβ₂), consistent with upregulation of inflammatory pathways. One integrin alpha subunit in the leading-edge gene sets, Itgav, pairs with numerous integrin β subunits including 1, 3, 5, 6, and 8. Itgb3 and Itgb5 are both represented in the leading-edge gene sets. Integrin αVβ5 binds fibronectin, vitronectin, LAP-TGF-β (14), and tropoelastin (59). Integrin αVβ3 binds fibronectin, vitronectin, fibrinogen, vWF, thrombospondin, collagen, LAP-TGF-β (14), fibrillin-1 (39), and tropoelastin (6). Itgb3 also pairs with integrin α11b in platelets and acts as a receptor for fibrinogen, vWF, and fibronectin (85). Integrin β3 inhibition has been implicated as a therapy for SMC overproliferation and stenoses associated with elastin insufficiency. Reduced integrin β3 gene dosage in Eln^−/− mice extends their survival from P1 to P4 (73). Our results suggest that integrin β3 inhibition may also be an effective treatment strategy in Efemp2^−/− and Lox^−/− mice, and possibly for human diseases associated with severe elastic fiber defects. Several of the integrin subunits or pairs present in our leading-edge groups, including β1, β3 (115), α5 (29), αV (105), and αVβ5 (114), are upregulated or change configuration in response to strain, implicating mechanosensitive pathway regulation in genetic models of elastic fiber defects.

Conclusions

We used comparative gene array analyses to examine gene expression patterns and pathways for three different genetic models of severe elastic defects for two vascular locations and developmental stages in mice. We found that commonly associated elastic fiber proteins and fibrillar collagens are not upregulated in response to severe elastic fiber defects. Integrin subunits and ECM genes in early wound healing are upregulated in all three models, suggesting that severe elastic fiber defects are sensed as a vascular wall injury. Differences in gene expression pathways, such as the upregulation of multiple inflammatory pathways in Efemp2^−/− and Lox^−/− aorta, may provide clues to specific disease phenotypes, such as aneurysm formation.

Limitations and Future Work

Microarray analyses are not as comprehensive as RNA-Seq, so we may have missed differentially regulated genes or pathways in our current data. Expression levels of genes of interest were confirmed by qPCR in previous publications (46, 100) or in this study for 32 of the 48 data sets. Expression levels of additional genes of interest must be confirmed in the other data sets in future work. We did not track the sex of the mice included in each group, hence our results could be skewed by sex differences between groups rather than elastic fiber defects. Additional studies are needed to evaluate sex-specific differences. Lastly, differential gene expression identified in mouse models must be verified in human tissue samples. Despite these limitations, the genes and pathways identified here provide new avenues of investigation for better understanding of severe elastic fiber defects in cardiovascular disease. Future work must address the functional consequences of the observed changes in gene expression.

GRANTS

This study was partially funded by National Institutes of Health (NIH) Grants HL-115560 and HL-105314 (J. E. Wagenseil) and the Washington University Institute of Clinical and Translational Sciences Grant UL1TR-000448 from the National Center for Advancing Translational Sciences. This publication is solely the responsibility of the authors and does not necessarily represent the official view of the NIH.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M.C.S., A.C., J.P., and J.K. performed experiments; M.C.S., A.C., J.P., J.K., and J.E.W. analyzed data; M.C.S., A.C., J.P., J.K., and J.E.W. interpreted results of experiments; M.C.S., A.C., and J.E.W. prepared figures; M.C.S., A.C., J.P., J.K., and J.E.W. approved final version of manuscript; A.C. and J.E.W. drafted manuscript; A.C. and J.E.W. edited and revised manuscript.