Large differences in collateral blood vessel abundance among individuals arise from multiple genetic variants

Abstract

Collateral blood flow varies greatly among humans for reasons that remain unclear, resulting in significant differences in ischemic tissue damage. A similarly large variation has also been found in mice that is caused by genetic background-dependent differences in the extent of collateral formation, termed collaterogenesis-a unique angiogenic process that occurs during development and determines collateral number and diameter in the adult. Previous studies have identified several quantitative trait loci (QTL) linked to this variation. However, understanding has been hampered by the use of closely related inbred strains that do not model the wide genetic variation present in the "outbred" human population. The Collaborative Cross (CC) multiparent mouse genetic reference panel was developed to address this limitation. Herein we measured the number and average diameter of cerebral collaterals in 60 CC strains, their 8 founder strains, 8 F1 crosses of CC strains selected for abundant versus sparse collaterals, and 2 intercross populations created from the latter. Collateral number evidenced 47-fold variation among the 60 CC strains, with 14% having poor, 25% poor-to-intermediate, 47% intermediate-to-good, and 13% good collateral abundance, that was associated with large differences in post-stroke infarct volume. Collateral number in skeletal muscle and intestine of selected high- and low-collateral strains evidenced the same relative abundance as in brain. Genome-wide mapping demonstrated that collateral abundance is a highly polymorphic trait. Subsequent analysis identified: 6 novel QTL circumscribing 28 high-priority candidate genes harboring putative loss-of-function polymorphisms (SNPs) associated with low collateral number; 335 predicted-deleterious SNPs present in their human orthologs; and 32 genes associated with vascular development but lacking protein coding variants. Six additional suggestive QTL (LOD > 4.5) were also identified in CC-wide QTL mapping. This study provides a comprehensive set of candidate genes for future investigations aimed at identifying signaling proteins within the collaterogenesis pathway whose variants potentially underlie genetic-dependent collateral insufficiency in brain and other tissues.

Keywords: Collaborative Cross; collateral blood vessels; leptomeningeal anastomoses; quantitative trait locus; stroke.

Figure 1.

Wide variation in collateral number and diameter among Collaborative Cross strains. (a) Collaterals between the right and left MCA and ACA trees of 3 strains (stars; representative collateral shown in lower panels; bars = 1 mm in upper figures, 40 µm in lower figures). (b) Strain names and number of animals per strain given below each bar ( $\ddot{\text{x}}$  = 7.9 ± 0.2 per strain; 537 mice total). Strain CC044 has the largest number, CC031 the largest diameter, and CC036 the smallest number and diameter. CC049, CC055, CC053, CC036 were used in F2 crosses (Figure 4), CC036, CC032, CC016 in AAV9-expression assays (Figure 6). CC039 has number and diameter similar to B6 mice. Among the 60 CC strains: fold ranges and heritability values are given, distribution of collateral number is shown (c), and diameter correlates with number (r² = 0.23, p < 0.001, Supplemental figure I). Values are mean ± SD for this and subsequent figures.

Figure 2.

Among selected CC strains: (a) Stroke severity is inversely associated with collateral number and diameter; (b) PCA-to-MCA collaterals (stars) have same relative strain-dependent differences in abundance as ACA-to-MCA collaterals; (c, d) Collaterals (stars) in skeletal muscle and intestine have the same relative strain-dependent differences in abundance as in brain (Figure 1(a)). (a) Infarct volume determined 24 h after permanent distal M1-MCA occlusion. (b) Number of PCA-to-MCA collaterals of both hemispheres in the 6 CC strains with the highest and 5 strains with the lowest collateral number shown in Figure 1 (n = 5–11 per bar). Diameter of PCA-MCA collaterals for CC021 and CC036 (19.3 ± 1.0, 9.3 ± 0.8) do not differ from diameter of their ACA-MCA collaterals shown in Figure 1 (18.4 ± 0.5, 11.2 ± 1.3). c, d Angiographic images of abdominal wall skeletal muscle viewed from peritoneal side, and small intestine. Bars = 1 mm for (b,c) and 100 um for (d).

Figure 3.

QTL mapping of collateral number and diameter in CC mice. LOD plots show the genomic position of QTL peaks for collateral number and diameter data in Figure 1. Relative chromosome length is shown by shading across the abscissae. Thresholds are for genome-wide significance based on 1000 permutations in this and subsequent figures. Six peaks exceed LOD 4.5 for number and 6 for diameter (see also “Data for QTL for collateral number with LOD > 4.5 shown in Figure 3” in Supplement); none is centered on the same marker SNP for both traits. Red arrows, peaks coincide with location of Rabep2.

Figure 4.

F1 crosses of 12 CC strains. Strains were chosen to give progeny with heterozygous genomes except at Rabep2, since variants of Rabep2 were previously shown to cause large effects on collateral number and diameter in a panel of classical strains. (a) F1 lines CC049 x CC053 and CC055 x CC036 were selected, among the 8 crosses (a, b), to generate two F2 populations (c) based on their a priori power ranking (see Methods) and minimum evidence of dominance. N-sizes in base of bars. ##, ### p < 0.01, <0.001 vs. 1^st bar, ***, p < 0.001 vs. 2^nd bar; 1-sided t-tests.

Figure 5.

Mapping of CC049 × CC053 F2 progeny identifies a significant QTL on chromosome 1. (a) LOD plot. (b) Recombination blocks, peak, and confidence interval (CI) of Canq5. (c) The Canq5 allele accounts for 15% of the difference in collateral number between CC0049 and CC053 shown in Figure 4 (4.7/31 collaterals). Supplemental table I gives additional characteristics of Canq5. Mapping of CC055 × CC036 F2 progeny identifies 7 significant QTL. (d) LOD plot. (e, f) Recombination blocks, peaks, confidence intervals (CI) and allele effects of Canq7 and Canq6. Their alleles account for 19% and 20%, respectively, of the difference in collateral number between CC055 and CC036 shown in Figure 3. Supplemental table I gives additional characteristics of Canq6-12.

Figure 6.

In vivo analysis of candidate genes. AAV9-expression constructs for Rabep2 (panel a, control for the assay) and selected high-priority candidates underlying Canq5 (b), Canq6 (c) and Canq7-10 (d) were injected iv on postnatal day-zero before blood brain barrier closure into Rabep2^−/− (A), CC016 (B), CC032 (C) and CC036 (D) mice. Number of collaterals (Supplemental figure III gives diameter) between the MCA and ACA trees of both hemispheres were measured at 6 weeks-age. Naïve, no injection; Control/GFP, injection of AAV9 construct for enhanced green fluorescent protein only. Number of animals for each bar: Panel A: 7,16,20; B: 8,15,17,14,14,13; C: 8,16,13,12; D: 6,14,10,10,8,8,10,12,8. *p < 0.05, **p < 0.01, ***p < 0.002 vs. Control/GFP by 1-sided t-tests.