Experiment
Previous studies using QTL mapping of mandible size and shape in mouse have relied typically on 2D landmarks and sparse sampling of the genome using microsatellite markers (Klingenberg et al. 2001, 2004) or a few hundred SNPs (Leamy et al. 2008). The primary purpose in this study was to validate QTL responsible for the variation in mandibular shape observed in mice using 3D phenotyping with denser genotyping than previously attempted, and assess the gains (if any) from the consideration of the third dimension.
Since none of the previous studies had associated 3D data for their samples, a new mouse backcross between A/J and C57BL/6J (Maga et al. 2015) was used. Mandibular shape genetics were reassessed using this new dataset and 3D phenotyping to evaluate the gain of information from the third dimension, given that mandible flattening seemed at first an acceptable compromise for large-scale study. As a secondary goal, the phenotypic coverage of the mandible was increased using a template of semilandmarks that was tied into the already collected expert landmarks to assess whether any further benefit was obtained from dense phenotypic coverage.
In the Maga 2015 study skulls of 433 (A/JC57BL/6J)A/J 28-day-old individuals were microCT scanned at 18 mm spatial resolution, and genotyped from liver tissue at 882 informative autosomal SNPs using the Illumina medium density linkage panel. After phenotyping and removing six incomplete specimens a full generalized Procrustes analysis was performed on the 3D landmarks using the R/Morpho package and then multivariate shape QTL mapping was done using the R/shapeQTL package of R statistical software.
Table 1: The mandible shape QTL that had already been assessed in the several studies using 2D landmarks on F2 mice from a LG/JSM/J intercross and on F3 mice from the same cross were kept for comparisons. A genome-wide association study using the first generation of wild-caught mice from a hybrid zone was also included in comparisons. It differs from the three other studies by its use of outbred hybrid mice, 3D data, and a smaller sample size.
A template of uniformly distributed semilandmarks was generated using poisson-disk sampling on the closest individual to the mean shape of 3D landmarks.
579 semilandmarks were retained in addition to 13 expert-annotated landmarks. The template was transferred onto new samples by thin-plate splines based on the 13 landmarks.
The effects of identified loci were estimated using a multivariate linear model. A forward/backward algorithm was used for multiple QTL model searching. The multivariate approach for mapping multiple QTL is very similar to the one developed for the mapping of function-valued traits, with pLOD based on the sum of the residual sum of squares.
Bayes credible intervals of QTL were computed from the 10LOD(0) profile. The closest proximal and distal markers given for the confidence intervals in earlier studies were converted to the current genetic map (Cox et al. 2009) using the Jackson Laboratorys Marker Query Tool. The genomic positions of loci discovered were converted from the GRCm38 coordinates to the Cox Map using the same tool.
In all three cases (2D, Manual 3D landmarks, and Semilandmarks), the three covariates (log of the centroid size, gender, and direction-of-cross) were included in the QTL mapping as additive covariates. Multivariate QTL mapping identified between 17 QTL for 2D landmarks, 19 QTL for 3D manual landmarks, and 23 QTL for the 3D semilandmark analysis. The median widths of the confidence intervals were 9 cM for both 3D landmarks and their 2D projections, and 5 cM for the semilandmark dataset, and three quarters of the confidence intervals were smaller than 17.3 cM, 20 cM, and 9.2 cM, respectively, for these three datasets [Table 1, Fig 4].
Table 2 : 23 QTL identified, p<0.0001, in 3D semilandmark analysis of (A/JC57BL/6J)A/J backcrossed mice :
Curator Note: Since a different mapping population was used here than in either of the comparison studies , (LG/J x SM/J and outbred mice respectively, we consider the current experiment a novel study and have assigned official nomenclature to each of the 23 QTL identified.
QTL Manh26 (mandible shape 26) mapped to Chromosome 1 peaking at 43.62 cM, nearest marker gnf01.075.385 in a confidence interval between 42.33 and 44.33 cM. This QTL overlaps QTL previously identified in both the comparison studies using both 2D and 3D analysis.
QTL Manh27 (mandible shape 27) mapped to Chromosome 2 peaking at 51.09 cM, nearest marker rs3722345 in a confidence interval between 51.09 and 60.54 cM. This QTL overlaps QTL previously identified in both the comparison studies using both 2D and 3D analysis.
QTL Manh28 (mandible shape 28) mapped to Chromosome 3 peaking at 20.01 cM, nearest marker rs6274061 in a confidence interval between 12.01 and 21.01 cM. This QTL was not previously identified.
QTL Manh29 (mandible shape 29) mapped to Chromosome 3 peaking at 65.01 cM, nearest marker rs3676039 in a confidence interval between 59.01 and 77.01 cM. This QTL was not previously identified. Within the confidence interval candidate gene Lef1 is annotated for mandible in the MGI database .
QTL Manh30 (mandible shape 30) mapped to Chromosome 4 peaking at 52.20 cM, nearest marker rs3711477 in a confidence interval between 52.10 and 53.01 cM. This QTL overlaps QTL previously identified in the 3D study.
QTL Manh31 (mandible shape 31) mapped to Chromosome 4 peaking at 83.01 cM, nearest marker UT4132.137715 in a confidence interval between 81.01 and 84.01 cM. This QTL was not previously identified. Candidate gene Rere mapped within the confidence interval of Manh31 with either nonsynonymous or splice variants between A/J and C57BL/6J.
QTL Manh32 (mandible shape 32) mapped to Chromosome 5 peaking at 15.71 cM, nearest marker rs13478154 in a confidence interval between 13.50 and 17.50 cM. This QTL overlaps QTL previously identified in both the comparison studies using both 2D and 3D analysis. Candidate genes Shh, Drc1 and Ift172 map within the confidence interval and are annotated for mandible in the MGI database.
QTL Manh33 (mandible shape 33) mapped to Chromosome 5 peaking at 52.50 cM, nearest marker rs13478154 in a confidence interval between 43.50 and 56.50 cM. This QTL overlaps QTL previously identified in both the comparison studies using both 2D and 3D analysis. Candidate genes Ambn,Prkg2, Dmp1, Fgfrl1 and Kctd10 map within the confidence interval and are annotated for mandible in the MGI database. Genes Fras1 and Mn1 map within the same interval with either nonsynonymous or splice variants between A/J and C57BL/6J.
QTL Manh34 (mandible shape 34) mapped to Chromosome 6 peaking at 43.00 cM, nearest marker CEL-6-86289708 in a confidence interval between 41.55 and 43.00 cM. This QTL overlaps QTL previously identified in both the comparison studies using both 2D and 3D analysis.
QTL Manh35 (mandible shape 35) mapped to Chromosome 6 peaking at 88.00 cM, nearest marker rs3658783 in a confidence interval between 84.00 and 89.28 cM. This QTL was not previously identified.
QTL Manh36 (mandible shape 36) mapped to Chromosome 7 peaking at 55.02 cM, nearest marker rs13479427 in a confidence interval between 43.05 and 55.02 cM. This QTL was not previously identified. Candidate genes Serpinh1 and Folr1 map within the confidence interval and are annotated for mandible in the MGI database. Genes Akap13 and Kif7 map within the same interval with either nonsynonymous or splice variants between A/J and C57BL/6J.
QTL Manh37 (mandible shape 37) mapped to Chromosome 8 peaking at 25.52 cM, nearest marker rs6386110 in a confidence interval between 22.38 and 27.52 cM. This QTL overlaps QTL previously identified in both the comparison studies using both 2D and 3D analysis.
QTL Manh38 (mandible shape 38) mapped to Chromosome 9 peaking at 44.47 cM, nearest marker rs3721056 in a confidence interval between 43.10 and 71.10 cM. This QTL overlaps QTL previously identified in both the comparison studies using both 2D and 3D analysis. Candidate gene Ryk mapped within the confidence interval and is annotated for mandible in the MGI database. Gene Atr mapped within the same interval with either nonsynonymous or splice variants between A/J and C57BL/6J.
QTL Manh39 (mandible shape 39) mapped to Chromosome 10 peaking at 3.18 cM, nearest marker rs3686911 in a confidence interval between 3.03 and 9.03 cM. This QTL was not previously identified.
QTL Manh40 (mandible shape 40) mapped to Chromosome 10 peaking at 70.03 cM, nearest marker mCV24217147 in a confidence interval between 67.03 and 71.12 cM. This QTL overlaps QTL previously identified in both the comparison studies using both 2D and 3D analysis.
QTL Manh41 (mandible shape 41) mapped to Chromosome 11 peaking at 16.08 cM, nearest marker rs3700830 in a confidence interval between 12.08 and 17.08 cM. This QTL was not previously identified.
QTL Manh42 (mandible shape 42) mapped to Chromosome 11 peaking at 49.08 cM, nearest marker rs13481127 in a confidence interval between 48.08 and 54.08 cM. This QTL overlaps QTL previously identified in both the comparison studies using both 2D and 3D analysis.
QTL Manh43 (mandible shape 43) mapped to Chromosome 11 peaking at 84.08 cM, nearest marker rs3672597 in a confidence interval between 82.08 and 86.08 cM. This QTL overlaps QTL previously identified in both the comparison studies using both 2D and 3D analysis.
QTL Manh44 (mandible shape 44) mapped to Chromosome 12 peaking at 7.99 cM, nearest marker rs13481321 in a confidence interval between 6.09 and 8.95 cM. This QTL overlaps QTL previously identified in the 3D analysis.
QTL Manh45 (mandible shape 45) mapped to Chromosome 13 peaking at 26.00 cM, nearest marker rs3693942 in a confidence interval between 25.00 and 26.52 cM. This QTL overlaps QTL previously identified in both the comparison studies using both 2D and 3D analysis.
QTL Manh46 (mandible shape 46) mapped to Chromosome 15 peaking at 13.68 cM, nearest marker CEL-15_36490596 in a confidence interval between 13.68 and 14.99 cM. This QTL overlaps QTL previously identified in both the comparison studies using both 2D and 3D analysis.
QTL Manh47 (mandible shape 47) mapped to Chromosome 16 peaking at 48.03 cM, nearest marker rs4204106 in a confidence interval between 33.03 and 53.31 cM. This QTL overlaps QTL previously identified in both the comparison studies using both 2D and 3D analysis.
QTL Manh48 (mandible shape 48) mapped to Chromosome 17 peaking at 18.14 cM, nearest marker rs6298471 in a confidence interval between 16.03 and 21.14 cM. This QTL was not previously identified.
Analysis Tools