A novel genetic locus modulates infarct volume independently of the extent of collateral circulation

Mice.

Both inbred strains of mice were obtained from the Jackson Laboratory (Bar Harbor, ME) and then bred locally at Duke University. All mice were fed a regular chow diet (PicoLab Mouse Diet 20, LabDiet). All mice were age-matched for collateral vessel traits measurement (p21 and 12 ± 1 wk) and for measurement of infarct volume after MCAO (12 ± 1 wk). Although we have shown that these phenotypes do not vary greatly according to sex, both sexes of mouse were used. Two independent cohorts of the F1 and F2 progeny of C57BL/6J (B6) and C3H/HeJ (C3H) were generated by a reciprocal breeding scheme. The first cohort was used for the measurement of infarct volume, whereas the second independent cohort was used to measure collateral vessel anatomy. To generate the F1 and F2 progeny, we used both males and females of each parental strain to intentionally examine the potential effect of sex chromosomes (or mitochondrial genomes) on phenotypes. Animal studies were performed under protocols approved by Animal Care and Use Committee of Duke University.

Middle cerebral artery and pial collateral circulation measurements.

We used a previously described dye-enhanced technique (43) with modification to stain the cerebral circulation, followed by collateral number and diameter measurements. As the collateral vessel traits are determined by 3 wk of age and remain constant for many months (8), we measured collateral phenotypes at age p21. Briefly, animals were first anesthetized with ketamine (100 mg/kg) and xylene (5 mg/kg). The left ventricle was cannulated retrograde, and the animal was well perfused with freshly made buffer composed of adenosine (1 mg/ml), papaverine (40 μg/ml), and heparin (25 mg/ml) in phosphate-buffered saline to wash out blood. The dorsal calvarium was removed to expose the brain before dye infusion. The left cardiac ventricle was cannulated retrograde through the previous cannulation site and polyurethane (PU4ii, VasQtec) with adequate viscosity to avoid capillary filling (1:1 resin/2-butanone) was slowly infused, and the cerebral circulation was visualized under stereomicroscope during infusion. Then, the brain surface was topically rinsed with 10% formalin and 20 min was allowed for the dye to solidify before overnight storage in 10% formalin. Each sample was imaged 24 h later for further analysis. The number and diameter of collateral vessels connecting the anterior cerebral artery and middle cerebral artery (MCA) on both hemispheres were counted and measured as previously described (41) with modifications. Briefly, collateral number was directly measured under stereomicroscope, whereas collateral diameter was measured from images using the image-processing package, Fiji (http://fiji.sc). The vasculature of each brain was imaged with a Nikon SMZ-2T stereomicroscope (Melville, NY) equipped with a Leica DFC425 camera (Buffalo Grove, IL). For diameter analysis, the central area of each brain was imaged, and three to six spots of the midzone of the collateral vessel were measured after optimal calibration. Collateral conductance was calculated as the product of the number multiplied the diameter of the collateral vessels (collateral number × diameter) (41). The number of primary MCA branches was determined from images covering the whole brain surface.

Surgical procedure of MCAO.

Focal cerebral ischemia was induced by distal MCA cauterization as described previously (24). Briefly, mice were first anesthetized with ketamine (100 mg/kg) and xylene (5 mg/kg). Then, the right MCA was exposed by a 0.5 cm vertical skin incision midway between the right eye and ear. After the temporalis muscle was split, a 2-mm burr hole was drilled at the junction of zygomatic arch and the squamous bone. The right MCA was visualized under a stereomicroscope and electrocauterized with a microcauterizer (F.S.T., Foster City, CA). The cauterized MCA was further transected distally with a lancet to verify the permanent MCA occlusion was complete. The surgical site was closed layer by layer with 6-0 sterile nylon sutures, and 0.25% bupivacaine was topically applied for local pain control. Animals were maintained at 37°C during and after surgery until fully recovered from anesthesia, when they were returned to their cages with free access to water and food. All mice were housed in an air-ventilated room with ambient humidity with room temperature maintained at 24 ± 0.5°C.

Four animals died during the procedure and were excluded from further analysis, giving a postsurgical survival rate of 98.3%. An additional 16 animals were excluded from analysis on the basis of predetermined exclusion criteria. Five were excluded because of postmortem evidence of incomplete surgical occlusion of the MCA, and 11 animals were excluded because of inadvertent cauterization of a branching vessel from the MCA, and not the MCA proper.

Measurement of infarct volume of mouse brain after MCAO.

Infarct volume was measured at 24 h after MCAO, as previously described (24). In brief, the brain was removed after proper euthanasia. Body and brain weights of each mouse were measured. Each brain was inspected under a stereomicroscope and imaged to identify the MCA branching pattern and location of cauterization and to verify the success of permanent occlusion. Since animals harboring aberrant branches from the MCA might alter the infarct volume, these animals were excluded for future analysis. Subsequently, the brain was sliced into nine 1-mm coronal sections using a brain matrix. Slices were stained with 2% triphenyltetrazolium chloride as previously described (24). We determined infarct volume by measuring infarct area on separate slices, multiplying area by slice thickness, and then summing all slices (24). The edematous change of brain slices, if present, was corrected accordingly (16, 24).

Genomic DNA isolation and genotyping.

Genomic DNA was purified from tails treated by overnight proteinase K digestion at 55°C, followed by isopropanol precipitation. Genomic DNA samples were genotyped using the GoldenGate genotyping assay (Illumina, San Diego, CA), consisting of 377 genome-wide mouse single nucleotide polymorphisms (SNP). For fine mapping of the linkage peaks, additional markers were genotyped for Civq4 through Civq7. The additional markers for Civq4 (chromosome 8) were rs30481888, rs33198371, rs33154273, and rs32709547; for Civq5 (chromosome 18), D18mit156, D18mit22, D18mit135, and D18mit178; for Civq6 (chromosome 5), D5mit355, D5mit398, D5mit18, D5mit91, D5mit431, and D5mit138; and for Civq7 (chromosome 10), rs3696307. Genetic map positions are reported based on the SNP database (build 37.1) of the National Center for Biotechnology Information (NCBI) website (http://www.ncbi.nlm.nih.gov/projects/SNP/).

Linkage analysis.

Genome-wide linkage analysis was performed using the J/qtl mapping program, version 1.3.3 (http://churchill.jax.org/software/jqtl.shtml). Suggestive (P = 0.63) and significant (P = 0.05) thresholds for each phenotype were determined by 1,000 permutation tests using all informative markers (24, 26). The 95% confidence interval (CI) of each peak was determined using the CI function in J/qtl. The percentage of total trait variance attributable (effect size) to each locus was determined using the Fit QTL function provided within the J/qtl program. Effect size is defined as the percentage of phenotypic variation explained by a given locus. A genome-wide analysis of interaction (epistatic or additive) between loci was performed. QTLs reaching a significant threshold value were named in accordance with the International Committee on Standard Genetic Nomenclature for Mice, and our previous work (24).

Candidate gene identification using sequence variation and gene expression.

We utilized two criteria to assess candidate genes located in the 95% CI region of Civq4 (38–51.1 Mb). The causative gene(s) might either alter the amino acid sequence of the encoded protein or change the mRNA level of the gene. First, we used the Mouse Phenome Database (http://phenome.jax.org), dbSNP (http://www.ncbi.nlm.nih.gov/SNP/) and Mouse Genomes Project database (http://www.sanger.ac.uk/cgi-bin/modelorgs/mousegenomes/snps.pl) to identify nonsynonymous coding SNPs (NS cSNP), stop codon gain/losses, splicing site gain/losses, indel, and structural variants between B6 and C3H strains. These data were cross-checked using the Ensembl Genome Browser (http://useast.ensembl.org) and, when required, Sanger sequencing in our laboratory. Candidate genes were also filtered by the requirement to be expressed in brain tissue (but not exclusively) using the BioGPS (http://biogps.org) and the EMBL-EBI (http://www.ebi.ac.uk) databases. To determine the potential functional impact of these NS cSNPs, we utilized PolyPhen-2 (http://genetics.bwh.harvard.edu/pph2/) (1). The gene expression analysis was performed on mouse cerebral cortex of B6 and C3H mice, collected at two ages: p1 and 11 wk. Total RNA of cerebral cortex of both strains was purified using RNeasy Kit followed by DNase treatment (Qiagen, Valencia, CA). Samples from each time point and strain were collected in triplicate. Microarray analysis was performed on an Affymetrix Mouse 430 2.0 Array (Affymetrix, Santa Clara, CA). Microarray data analysis was performed using the Partek Genomics Suite (Partek, St. Louis, MO). Significant expression differences between strains were verified by a quantitative real-time PCR (qRT-PCR) assay from tissue isolated from both parental strains and, when possible (i.e., when a strain-specific SNP was present in the transcript), by allelic-specific gene expression analysis using the SNaPshot Multiplex Kit (Life Technologies, Grand Island, NY) using tissue isolated from F1 (B6/C3H) animals (Table 1).

Statistical analysis.

All the data are presented as means ± SD. Statistical analyses of collateral vessel traits, number of primary branch from MCA, and infarct volume were performed by one-way ANOVA. χ² analysis was to statistically analyze the genotype distribution for each phenotype.

Collateral diameter but not number differs between B6 and C3H.

Collateral number has been proposed as a major factor determining infarct volume in the mouse model of ischemic stroke (43). However, we have observed a few inbred mouse strains (B6, C3H, and LP/J) that exhibit similar collateral number but that differ significantly in infarct volume. We surmised that crosses between these strains might enable us to identify collateral vessel-independent genetic determinants of infarct volume. And since the Civq1 (infarct volume) and the Canq1 (collateral number) loci appear to be one and the same, most crosses between a “small infarct/high collateral number” inbred strain with a “large infarct/low collateral number” inbred strain will invariably reidentify the same large effect Civq1/Canq1 locus, as we have previously shown (24). With the phenotypic effects from this robust locus eliminated, we hoped to uncover additional loci that are otherwise masked by the overwhelming phenotypic effects of Civq1/Canq1.

We first reinvestigated the collateral vessel phenotypes between inbred strains B6 and C3H at p21 and 12 ± 1 wk old, the latter being the age that we routinely perform MCAO in our stroke model (Fig. 1, A–D). We observed no statistically significant difference between B6 and C3H in collateral number at either the p21 or 12 ± 1 wk age (Fig. 1E). Thus, the collateral number phenotype is similar between these two strains, while the C3H strain exhibits a fourfold larger infarct volume after MCAO than that seen with B6 (19.3 vs. 4.8 mm³, P < 0.01) (Fig. 2). This suggests that the difference in infarct volume between B6 and C3H might be determined by mechanisms other than collateral number. In each parental strain, the infarct volume does not vary according to sex (data not shown).

In addition to collateral number, collateral diameter has also been proposed as a factor modulating infarct volume, although diameter alone (in the absence of a difference in collateral number) is not a robust predictor of infarct volume, especially for strain C3H (41, 43). Collateral diameter differs between B6 and C3H, with the average collateral diameter at p21 of 22.2 μm in B6 and 16.9 μm in C3H (P < 0.01) (Fig. 1F). Collateral conductance, calculated from collateral number and diameter, is also lower in C3H compared with B6 (Fig. 1G).

In addition to collateral vessel traits, the infarct volume will be most obviously influenced by the size of the area supplied by the culprit artery, in this case the MCA, but the MCA territory is nearly identical between B6 and C3H (43). We also investigated whether the number of primary branches of MCA might contribute to the observed differences in infarct volume between strains. However, B6, C3H, and their F1 progeny do not differ in this trait (data not shown).

Four novel loci modulate infarct volume differences between B6 and C3H.

We generated an initial cohort of 230 F2 animals, and MCAO was performed on the entire cohort at age 12 ± 1 wk. After applying predetermined inclusion criteria (see materials and methods), we found that 210 F2 animals remained for QTL mapping. The distribution of infarct volume in this 210 F2 cohort was quite broad, ranging from 0.36 to 35.4 mm³ (Fig. 3A), providing more than sufficient phenotypic variation for the linkage analysis. The number of males and females were distributed nearly equally, and infarct volume in the F2 animals did not vary according to sex (Fig. 3B).

Each animal of the F2 cohort was genotyped with a genome-wide mapping panel of which 264 SNP markers were informative. We then performed whole genome linkage analysis for the phenotype of infarct volume (Fig. 4). As we had predicted, the robust Civq1 locus mapping to chromosome 7 was not identified in this cross, and in fact, the linkage profile for chromosome 7 is nearly flat. Thus, the predominant locus (Civq1/Canp1) affecting infarct volume in other pairwise crosses (24, 41) was rendered inert by our judicious choice of parental strains that exhibit large variation in infarct volume with minimal to no differences in collateral vascular anatomy.

In the absence of the usually overwhelming effects of Civq1 on infarct volume, four new linkage peaks reached genome-wide significance. These were designated as Civq4 through Civq7 based on the magnitude of the LOD (logarithm of the odds) score and our previous work that identified three distinct loci for this trait (24). The new significant loci are Civq4 (chromosome 8, LOD 9.8), Civq5 (chromosome 18, LOD 4.2), Civq6 (chromosome 5, LOD 3.7), and Civq7 (chromosome 10, LOD 3.5). We also examined the genotype distributions of the F2 mice to identify any deviations from Mendelian segregation, and the genotypes at all four loci followed the expected 1:2:1 distribution of B6/B6:B6/C3H:C3H/C3H alleles.

Among the four loci, Civq4 is the strongest, contributing 21% of the phenotypic variance of infarct volume. To better define the linkage interval, we genotyped mice for four additional markers mapping within the linkage peak. The resulting 95% CI for Civq4 is from 38.0 to 51.1 Mb (Fig. 5). In contrast to the C3H parental strain that exhibits a larger infarct volume than B6, the C3H allele at Civq4 renders animals more resistant to infarction, whereas the B6 allele renders them more sensitive to infarction. These alleles that operate in the opposite direction of the parental strain phenotypes are termed transgressive alleles.

Civq5, mapping to chromosome 18, contributes 10% of the observed phenotypic variance in the cross. After fine mapping, the resulting 95% CI for Civq5 lies between 9.6 and 46.6 Mb (Fig. 6). Civq6, mapping to chromosome 5, contributes 8% of the observed variance of infarct volume (Fig. 7). The resulting 95% CI after genotyping with six additional markers in the vicinity of the linkage peak is from 67.0 to 119.8 Mb. After fine mapping, the observation of two peaks in Civq6 is persistent, strongly suggesting that Civq6 is in reality the result of at least two genes mapping on the same chromosome but spaced at least a considerable distance apart. Civq7, mapping to chromosome 10, contributes 6% of the infarct volume variance in the cross; the 95% CI for this locus is between 28.2 and 68.4 Mb after fine mapping (Fig. 8). At both Civq5 and Civq7 loci, the B6 allele exhibits a protective effect on infarct volume that is consistent with the phenotype of the parental B6 strain. Similar to Civq4, the C3H alleles at both peaks of Civq6 also represent protective alleles for infarction, with the B6 alleles as susceptibility alleles. None of the Civq loci yields epistatic interactions with other regions of the genome. The characteristics of all four loci are summarized in Table 2.

Since traits associated with sexual dimorphism, such as body or brain size, might affect infarct volume, we performed QTL mapping in the same 210 animal cohort using brain weight, body weight, and ratio of brain weight/body weight as quantitative traits. However, none of the Civq loci identified as infarct volume loci were identified in this analysis after adjustment for sex effects (data not shown). Additionally, in this F2 cohort, the strain origin of the X and Y chromosomes as well as the mitochondrial genome exhibit no effect on infarct volume.

Collateral vessel-dependent and -independent loci identified in the cross.

To further investigate whether the new loci we have mapped modulate infarct volume in a collateral vessel-independent manner, we employed QTL analysis using collateral number, diameter, and conductance as quantitative traits in a second, independent F2 (B6 × C3H) cohort consisting of 105 animals. We reasoned that if the new loci that we mapped for infarct volume are determined by collateral phenotypes, then we should map these same loci for the collateral traits as well. The distribution of collateral number per hemisphere in the F2 cohort ranged from 6 to 14.5 with the vessel diameter ranging from 12.0 to 24.7 μm (Fig. 9, A and B). The calculated collateral conductance per hemisphere (41) ranged from 90.1 to 311.1 (Fig. 9C). This cohort was genotyped for 264 informative makers across the genome for QTL mapping.

We first noticed that only modest linkage “peaks” resulted in this analysis for any of the collateral vessel traits, with none of these potential loci reaching genome-wide statistical significance (Fig. 10). This outcome was not unexpected. Since collateral number does not differ between the B6 and C3H strains, we did not expect this cross to yield genetic loci for this trait. Although collateral diameter does differ between the two strains, the difference is much smaller than is seen between other strain pairs (43), again suggesting that robust loci for this trait might not be operating in this cross. Indeed, this pairwise cross would have been a rather poor choice to map collateral vessel traits, unless one wished to formally rule out any effects of collateral phenotypes on infarct volume, which was our stated objective. But the question still remains as to whether any of these rather modest linkage peaks for collateral traits are nonetheless related to any of the four new infarct volume loci.

Civq4, a very strong locus (21% effect size, LOD score of 9.8) for infarct volume mapping to chromosome 8, did not appear for any of the collateral vessel phenotypes. No portion of chromosome 8 even reached the modest “suggestive” alpha threshold of 0.63 (LOD score of 2.1). Thus, Civq4 modulates infarct volume in a collateral-independent manner. To date, this is the first infarct volume locus identified that clearly operates in a collateral-independent manner.

The mechanism underlying Civq5 mapping to chromosome 18, the next strongest of our novel infarct loci, is less clear. A rather modest peak at the same region on this chromosome is found for the traits of collateral number and conductance (mathematically derived from vessel number). Thus, this nascent collateral trait locus, although not reaching statistical significance, may suggest that the gene underlying Civq5 modulates infarct volume via effects on collateral vessel anatomy. The same region on chromosome 5 harboring the infarct locus Civq6 also shows a modest peak for the all three collateral vessel phenotypes. Although this nascent collateral vessel locus does not reach statistical significance, the possibility also remains that Civq6 exerts its effects on infarct volume via modulation of collateral vessel anatomy.

Finally, Civq7 mapping to chromosome 10, appears to operate in a collateral vessel-independent manner. No portion of chromosome 10 even reached the modest “suggestive” alpha threshold of 0.63 (LOD score of 2.1). Indeed the linkage profile of this chromosome for all collateral phenotypes is the lowest or among the lowest of all other chromosomes. Thus, Civq7, like Civq4, modulates infarct volume in a collateral-independent manner. Our rationale that the B6 × C3H intercross would not only preclude the overwhelming effects of the Civq1/Canq1 locus but also by this enable the mapping of novel, collateral independent loci, was validated.

Civq4 is a novel locus modulating infarct volume in a collateral-independent manner.

As stated above, Civq4 mapping to chromosome 8 harbors transgressive alleles for both B6 and C3H. These alleles modulate the infarct volume phenotype in the opposite direction to that observed in the parental strains. Previous QTL mapping studies of infarct volume and collateral phenotypes using the B6 × BALB/c inbred strains identified what appears to be nonoverlapping loci on chromosome 8 (24, 41). These two loci, Civq3 for infarct volume and Canq4 for collateral artery number, also harbor transgressive alleles, in these cases for the B6 and BALB/c alleles. The current evidence suggests that Civq4 mapped in this study is distinct from either of these two loci. First, and most importantly, the Canq4 was identified as collateral number-determining locus, whereas Civq4 does not share this phenotypic effect. Secondly, the peaks of the Civq3 and Civq4 loci are 26 Mb apart. However, since both Civq3 and Canq4 were identified in the presence of the overwhelming phenotypic effects of Civq1/Canq1, their contribution to their respective phenotypes was quite low, leading to comparatively low effect sizes and broad linkage peaks that cover large portions of chromosome 8. The linkage profile for Civq4 by contrast, intentionally mapped in the absence of effects from Civq1/Canq1, is much more amenable to eventual gene identification (13) because of the effect size of Civq4 is >20%, and the 95% CI covers a comparatively smaller region of the chromosome.

Candidate genes for Civq4.

Toward the goal of gene identification for the most robust of our mapped loci, Civq4, we investigated candidate genes mapping within the 95% CI region of Civq4 (38–51.1 Mb, chromosome 8). Of a total of 70 genes falling within the nearly 13 Mb genomic interval 15 genes harbor NS cSNPs, with some genes harboring multiple cSNPs (Table 3). We employed the PolyPhen-2 algorithm (http://genetics.bwh.harvard.edu/pph2/) to determine the potential functional impact these NS cSNPs would have on the encoded proteins. Five genes contain NS cSNPs that were predicted to be either “possibly or probably damaging” their respective proteins: Mtus1 (Mitochondrial tumor suppressor), Dlc1 (Deleted in liver cancer 1), AI429214 (Expressed sequence AI429214), Msr1 (Macrophage scavenger receptor 1), and Adam39. According to the BioGPS (http://biogps.org/) and EMBL-EBI (http://www.ebi.ac.uk) databases, Mtus1 is mainly expressed in lung, cerebellum, liver, skeletal muscle, and placenta; Dlc1 is expressed in lung, adipose brown fat, adrenal gland; Msr1 is predominantly expressed in macrophages and microglia (22, 23); and AI429214 and Adam39 are almost exclusively expressed in testis.

One gene, Mtmr7 (Myotubularin related protein 7), was particularly interesting as a SNP is reported to cause a premature truncation codon (Mouse Genomes Project, http://www.sanger.ac.uk/cgi-bin/modelorgs/mousegenomes/snps.pl) and Ensembl Genome Browser (NCBIM37, http://useast.ensembl.org/Mus_musculus/Info/Index). Mtmr7 is expressed primarily in brain tissues (31) with two major Mtmr7 transcripts in the mouse, ENSMUST00000048898 and ENSMUST00000174205, which encode proteins of 660 and 498 amino acid residues, respectively. The C3H allele at SNP rs33076137 is listed to cause a stop codon in ENSMUST00000048898 that would result in a truncated transcript distinct from that of ENSMUST00000174205. By Sanger sequencing genomic DNA and cDNA generated from both B6 and C3H cortex, we validated the presence of this SNP in genomic DNA but found that it was not located in any known splice isoform of the Mtmr7 transcript (data not shown). Deeper inspection of the databases indicates that this SNP is indeed located within an intron of ENSMUST00000048898, although it is unclear then how this SNP became annotated as a stop-gain variant.

In addition to protein coding differences, sequence variation can cause differences in gene regulation, defined broadly as any variant that alters mRNA levels, regardless of the mechanism. Thus, we employed mRNA microarray analysis on cerebral cortex tissue from p1 and 11 wk old mice to identify differences in transcript levels between the inbred strains B6 and C3H. A twofold or greater expression difference between B6 and C3H was used as a cut-off point to filter candidate genes from the microarray results. Genes expressed at least twofold greater in B6 than in C3H are categorized as B6>C3H, whereas genes expressed twofold or greater in C3H are categorized as C3H>B6. In the B6>C3H category at age p1, there were 194 genes expressed at least twofold higher in B6 than in C3H. Among these 194 genes, eight are located within the 95% CIs of Civq4 to Civq7 (Table 4). At age 11 wk 169 genes are expressed at least twofold higher in B6 compared with C3H, with only seven located within the 95% CIs of the four new Civq loci (Table 4). The Mtmr7 gene located in the 95% CI of Civq4, shows 8- to 12-fold higher expression in B6 than in C3H at both ages, respectively. The expression difference of MTMR7 was further validated by both qRT-PCR and SNaPshot. The qRT-PCR and SNaPshot data both show that that although splice variant ENSMUST00000048898 is equally expressed between B6 and C3H (data not shown), isoform ENSMUST00000174205 is expressed more highly in B6 cerebral cortex (Fig. 11).

In the C3H>B6 category at age p1, there are 76 genes showing a strain-specific expression difference, whereas at 11 wk there are 82 genes differentially expressed. Seven genes map within the 95% CIs of Civq4 to Civq7 for p1 cortex and eight genes for the adult tissue (Table 4). However, none of these transcript level differences are as great as those found in the B6>C3H category. The microarray data was submitted to Gene Expression Omnibus at NCBI (accession number GSE46443).