Abstract
Family and twin studies suggest that a substantial genetic component underlies individual differences in craniofacial morphology. In the current study, we quantified 444 craniofacial traits in 100 individuals from two inbred medaka (Oryzias latipes) strains, HNI and Hd-rR. Relative distances between defined landmarks were measured in digital images of the medaka head region. A total of 379 traits differed significantly between the two strains, indicating that many craniofacial traits are controlled by genetic factors. Of these, 89 traits were analyzed via interval mapping of 184 F2 progeny from an intercross between HNI and Hd-rR. We identified quantitative trait loci for 66 craniofacial traits. The highest logarithm of the odds score was 6.2 for linkage group (LG) 9 and 11. Trait L33, which corresponds to the ratio of head length to head height at eye level, mapped to LG9; trait V15, which corresponds to the ratio of snout length to head width measured behind the eyes, mapped to LG11. Our initial results confirm the potential of the medaka as a model system for the genetic analysis of complex traits such as craniofacial morphology.
CRANIOFACIAL morphology is a complex but interesting trait that reveals individual phenotypic differences within a species as well as morphological divergence among species. Multiple genetic factors and environmental variables account for the large degree of variability in human craniofacial morphology. The heritability of human craniofacial morphology has been thoroughly investigated in twins and families. A genetic component has been reported for 60–90% of craniofacial traits, including facial height, position of the lower jaw, and cranial base dimensions (Savoye et al. 1998; Johannsdottir et al. 2005). Further analysis of human linkages is difficult, due in part to sample heterogeneity, limited sample numbers, and a significant impact of environmental factors on craniofacial phenotypes. To control for these effects, animal models are often used to genetically dissect the developmental pathways that operate in concert to govern multifactorial traits.
Single gene mutation analyses in mice and zebrafish have shown that several common signaling cascades function during vertebrate craniofacial development, highlighting the potential of these animals as models. A loss of Sonic hedgehog (Shh) signaling severely reduces the size of the craniofacial skeleton and produces marked craniofacial defects, including complete or partial cyclopia, in both mice and zebrafish (Brand et al. 1996; Chiang et al. 1996; Chen et al. 2001). These data suggest that Shh signaling is crucial for the development of craniofacial components. Similarly, mouse embryos lacking Endothelin-1 (Et-1) show severe defects in Meckel's cartilage in the mandibular arch and the ventral cartilage in the hyoid arch (Kurihara et al. 1994). The mandibular and hyoid arch ventral cartilages are also affected by mutation of sucker (suc), the zebrafish Et-1 ortholog (Miller et al. 2000). Thus, the craniofacial phenotypes of homozygous mouse Et-1 mutants and zebrafish suc mutants are essentially identical. These findings suggest that the molecular mechanisms underlying craniofacial development are largely conserved in vertebrate species.
Quantitative trait locus (QTL) analysis has been used to successfully identify chromosomal regions affecting the quantitative traits, including obesity, bone density, and cerebellum size (Beamer et al. 1999; Brockmann et al. 2000; Airey et al. 2001). The jaw apparatus, one of the most thoroughly investigated craniofacial components, has recently been analyzed as a quantitative trait, and QTL analysis has identified several genomic regions responsible for the size and shape of the mandible in mice (Klingenberg et al. 2001, 2004; Dohmoto et al. 2002). Furthermore, the shape of the oral jaw in the east African cichlid fish has been analyzed using quantitative genetics, and associated QTL have been identified (Albertson et al. 2003); however, there are few quantitative studies on gross craniofacial morphology.
The medaka is a small, egg-laying freshwater fish found in eastern Asia. It is suitable for genetic studies of both Mendelian and complex traits because it has a short life cycle and is very fertile. In addition, several inbred strains are available (reviewed in Naruse et al. 2004a). Inbred strains of medaka differ in terms of various quantitative traits, including body shape, behavior, and susceptibility to chemical or radiation-induced tumorigenesis (Hyodo-Taguchi 1990). Furthermore, gross brain morphology varies greatly in five inbred strains with different genotypes (Ishikawa et al. 1999). The availability of a dense and accurate genetic linkage map (Naruse et al. 2004b; Kimura et al. 2005), in combination with a number of inbred strains that display diverse quantitative traits, makes it possible to identify QTL for complex traits in the medaka.
Here, we demonstrate that craniofacial variation has a genetic basis in the medaka and estimate the heritability for many of the traits. QTL mapping performed using F2 analysis from two inbred medaka strains, HNI and Hd-rR, demonstrates the potential of this species for genetic analysis of craniofacial morphology. This is the first study using the medaka to dissect complex traits by QTL analysis.
MATERIALS AND METHODS
Strains, pedigree design, and fish maintenance:
HNI and Hd-rR strains (Hyodo-Taguchi 1996) are inbred strains of medaka established from a northern Japanese population and a southern Japanese population, respectively. Three pairs of HNI males and Hd-rR females and three pairs of Hd-rR males and HNI females were crossed to generate F1 progeny. A total of 184 F2 progeny were obtained by intercrossing eight pairs of the F1 fish. Fish were maintained in an in-house facility at 26° in a constant recirculating system on a 14 hr light/10 hr dark cycle.
Phenotypic analysis:
We analyzed 100 HNI fish, 100 Hd-rR fish, 50 F1 progeny (HNI × Hd-rR), and 184 F2 progeny obtained by intercrossing F1 fish. The F2 progeny were from the same set we used in a previous study (Kimura et al. 2005). Adult fish (4 months old) were anesthetized and mounted in 2% methylcellulose. Lateral, dorsal, and ventral head images of each medaka were then captured with a CCD camera (Fuji digital camera HC-2500) mounted on a Leica MZ12.5 stereo-microscope. Images were collected at a constant magnification with a resolution of 1280 × 968 pixels. Craniofacial landmarks were identified on each image as shown in Figure 1. We calculated the distances between different landmark combinations from their coordinates. Ratios of different craniofacial distances were then calculated to generate phenotypic values. The quantified traits we obtained are listed in supplemental Table 1 (http://www.genetics.org/supplemental/). The t-test was used to determine whether males and females within the same strain differed significantly and whether the two strains differed significantly. Step-down Holm-Sidak correction (Ludbrook 1998) was applied when testing for significance between the two strains. Sex-linked traits should be mapped using sorted samples by gender. Therefore, no correction was carried out when testing for gender differences so as to reduce the errors of analyzing sex-dependent traits in a gender-mixed sample.
Genetic analysis:
The broad-sense heritability of each quantified trait was estimated from the phenotypic variance of F1 and F2 progeny (Falconer 1989). The genetic correlation was estimated from the phenotypic variance and covariance of F1 and F2 progeny (Ukai 2002). A total of 304 markers, composed of 48 M-marker 2003 (Kimura et al. 2004) and 256 newly mapped markers (Kimura et al. 2005), were used to map the quantified craniofacial traits. The genotype data are those described in Kimura et al. (2005). Both genotype and phenotype data were analyzed by interval mapping using a MAPMAKER/QTL 1.1 program (Lander et al. 1987, DOS version, used in Windows 2000). The algorithms upon which MAPMAKER/QTL are based assume that the quantitative trait values are normally distributed across the population. We examined whether the phenotype data were normally distributed by using a chi-square test for goodness of fit. When the data did not fit well with a normal distribution, it was corrected using the renouncement method in the program STSS/EXCEL 5.5. The significance threshold of the logarithm of the odds (LOD) score was 2.8, based on the average intermarker distance in this study and the genome size of the medaka. This indicates that the chance of a false positive occurring anywhere in the medaka genome is at most 5% (Lander and Botstein 1989). Traits that were mapped with a maximum LOD score of 5.0 or greater were analyzed by the composite interval mapping (Zeng 1993, 1994) using the Windows program QTL Cartographer 2.0 (in Windows 2000).
RESULTS
Gender differences in craniofacial traits:
We first examined the craniofacial traits of 100 medaka (50 males and 50 females) from HNI and Hd-rR, the two major inbred strains from northern and southern Japanese populations, respectively. These two medaka populations differ in terms of both their genetic (Sakaizumi et al. 1983; Takehana et al. 2003) and morphological traits (Ishikawa 2000). These differences were likely generated by their geographic separation. In this study, 444 traits were quantified: 158 lateral view traits, 157 dorsal view traits, and 129 ventral view traits. Gender differences in teeth size—one of the craniofacial phenotypes—have previously been reported in the medaka (Egami 1956). We tested the two strains for the presence of other gender-dependent morphological differences (supplemental Table 1 at http://www.genetics.org/supplemental/). Gender differences were observed in 304 of 444 traits (68.5%, Table 1). Interestingly, the fraction of gender-specific trait differences was extremely large in ventral view (82.2%) and was the smallest in dorsal view (57.3%, Table 1). The landmarks in the ventral view were mostly on the pharyngeal structures, such as the lower jaw (Figure 1C), indicating that many of these structures may develop under the control of sex-dependent factors.
Landmarks in the digital images of the medaka head region. Dots indicate odd-numbered landmarks, which were denoted this way for convenience. (A) Lateral view, (B) dorsal view, and (C) ventral view. Bar, 3 mm.
Characteristics of craniofacial traits quantified in medaka
Strain differences in craniofacial traits:
We next examined whether craniofacial traits differed between the two inbred medaka strains, HNI and Hd-rR. Gender-independent traits were examined in a gender-mixed population of 100 individuals from each strain, whereas gender-dependent traits were tested in male and female subpopulations. Most of the craniofacial traits quantified in this study differed significantly between HNI and Hd-rR (Table 1). Of the 140 gender-independent traits, 125 (89.3%) differed between these two strains, whereas 254 of the 304 gender-dependent traits (83.6%) differed in either male or female medaka (Table 1). Of these, 145 (47.7%) differed in both sexes, 59 (19.4%) differed only in males, and 50 (16.4%) differed only in females (supplemental Figure 1 at http://www.genetics.org/supplemental/). Thus, craniofacial phenotypes in this organism are controlled genetically, at least to a certain extent.
Heritability of craniofacial traits in the medaka:
Craniofacial traits that differed significantly between strains HNI and Hd-rR were further quantified in 50 F1 and 184 F2 progeny. The broad-sense heritability of each trait was estimated from the variance in the isogenic F1 population and heterogeneous F2 samples. Supplemental Table 2 (http://www.genetics.org/supplemental/) shows the mean values for gender-independent craniofacial traits in HNI, Hd-rR, F1, and F2 populations, as well as the heritability of these traits in medaka. Many of the craniofacial traits were highly heritable. The greatest heritability for a lateral view trait was 0.768 for L37, for a dorsal view trait was 0.870 for D34, and for a ventral trait was 0.893 for both V1 and V4. A high degree of heritability was evident in many of the lateral view traits; 67.6% showed a heritability value > 0.5. Only 43.2% (38 of 88 traits) of the dorsal and ventral view traits showed such a high degree of heritability. The mean values and heritability of gender-dependent traits are shown in supplemental Table 3 (http://www.genetics.org/supplemental/).
QTL mapping:
Craniofacial traits that fulfilled the following conditions were analyzed by QTL mapping: (1) no gender difference; (2) differed between strains HNI and Hd-rR; and (3) the F2 population was normally distributed, indicating polygenic regulation of the traits. F2 samples should be analyzed by each gender in order to map gender-dependent traits. Because the resulting sample size (<100) was too small to detect QTL, gender-dependent traits were not dealt with in this study but will be analyzed in the near future as we collect additional F2 medaka. Eighty-nine of the target traits met the criteria described above: 31 of the lateral view traits, 35 of the dorsal view traits, and 23 of the ventral view traits. Six of the lateral view traits and 30 of the dorsal view traits were not normally distributed; 4 of these lateral view traits involved the length between landmarks 13 and 47 (L13:47), and most of these dorsal view traits contained either D19:29 or D29:31 (non-normally distributed traits are marked by asterisks in supplemental Table 2). An interval mapping genomic screen was performed by genotyping all 184 F2 medaka using 304 primers that cover the entire genome with an average intermarker distance of 5.9 cM (supplemental Table 4 at http://www.genetics.org/supplemental/, Figures 2–4⇓⇓). We detected QTL for a total of 66 traits (66/89; 74.2%): 17 lateral view traits (17/31; 54.8%), 34 dorsal view traits (34/35; 97.1%), and 15 ventral view traits (15/23; 65.2%).
QTL locations for the craniofacial traits in the lateral view. Thick lines in the chromosomes represent the chromosome parts constructed by Kimura et al. (2005). The thin lines were added following analysis of M-marker 2003 (Kimura et al. 2004). The LGs are listed according to the numbering used by Naruse et al. (2004b). The numbers on the left side of each chromosome show the genetic distances in Kosambi centimorgan (cM). Markers are indicated on the right side of each chromosome. Colored bars to the left of the genetic distances indicate the regions of 1-LOD support interval of linkage.
QTL locations for the craniofacial traits in the dorsal view. See details in the legend for Figure 2.
QTL locations for the craniofacial traits in the ventral view. See details in the legend for Figure 2.
Several traits were sometimes mapped to the same region of a linkage group (LG). Five lateral view traits (L3, L5, L7, L9, and L28) were mapped to LG4 (Figure 2). Four of these traits (L3, L5, L7, and L9) exhibited moderate to high phenotypic and genetic correlations in F2 medaka. L28 was not well correlated with the others (supplemental Table 5 at http://www.genetics.org/supplemental/). QTL for another 5 traits (L2, L3, L11, L13, and L17) were detected on LG8, and a similar, but different, combination of 5 traits (L2, L4, L11, L13, and L17) were found on LG23 (Figure 2). Most of these traits were highly correlated (supplemental Table 6 at http://www.genetics.org/supplemental/), and as many as 4 traits (L2, L11, L13, and L17) shared regions on both LG8 and LG23. All of these traits were associated with the position of the eye along the anteroposterior axis in the head region. The genetic factors for eye positioning along the anteroposterior axis within the head region are therefore likely located in the shared regions on LG8 and LG23. Overlapping QTL were also detected for dorsal traits. Fifteen traits (D1–10, D17, D21, D25, D32, and D35) were mapped to LG3, another 7 (D23, D24, D26, D27, D31, D33, and D34) were mapped to the upper distal region of LG4, and 9 (D6, D13, D14, D16, D20, D24, D27, D30, and D31) were located on LG12 (Figure 3). The 15 traits that were mapped to LG3 were highly correlated except for D6, which was only moderately correlated with the other 14 traits (supplemental Table 7 at http://www.genetics.org/supplemental/). High phenotypic correlations were observed between D23, D24, D26, and D27 and between D31 and D33, but correlations were low to moderate between the two groups of traits. D34 showed little phenotypic correlation and relatively low genetic correlation with the others (supplemental Table 8 at http://www.genetics.org/supplemental/). The traits that were mapped to LG12 could be roughly classified into four groups on the basis of the phenotypic correlations: (1) D6, D13, and D14; (2) D16 and D20; (3) D24 and D27; and (4) D30 and D31. Moderate to high correlations were observed between the traits within a group, but low or few correlations were found between traits from different groups (supplemental Table 9 at http://www.genetics.org/supplemental/). In contrast to the dorsal view traits, only one region was overlapped by 5 or more ventral view traits. Traits V1, V4, V5, V12, V13, and V17 were mapped to LG5 (Figure 4). High correlations were observed between V1, V4, and V5. V12 correlated well with V13 (supplemental Table 10 at http://www.genetics.org/supplemental/); however, V17 did not correlate well with the other traits from these groups.
Eight traits were mapped with a maximum LOD score of ≥5.0 (supplemental Table 4). Three of these were the lateral view traits, L23, L29, and L33. L23 was mapped to LG14 with a maximum LOD score of 5.2. L29 and L33 were mapped to LG9 with maximum LOD scores of 5.4 and 6.2, respectively. The QTL for L23 accounted for 13.0% of the phenotypic variation, and two Hd-rR alleles in the nearest marker of the QTL had longer snout-to-eye height than the other genotypes (Table 2). For L29 and L33, 13.1% and 15.9%, respectively, of the phenotypic variation could be explained by the QTL on LG9. The ratio of the head length to height was greater in homozygotes with the HNI allele near the given QTL (Table 2). The QTL for the dorsal view traits, D13 and D14, both localize to the same region of LG12, with the same maximum LOD score (5.3). Both of them account for 12.6% of the variance. Both D13 and D14 correspond to the ratio of snout length to premaxillary length in the dorsal view, and measurements are smaller in medaka that are homozygous for the Hd-rR allele in the nearest marker of the QTL (Table 2). Another dorsal trait, D29, was mapped to LG22 with a maximum LOD score of 5.0, accounting for 12.5% of the total variance. Fish homozygous for HNI alleles near the mapped locus had the shortest head-to-snout width; those homozygous for Hd-rR alleles had the largest head length to snout width. Heterozygotes were intermediate, suggesting no dominance between the alleles (Table 2). A region within LG11 was also highly associated with V14 and V15 with maximum LOD scores of 6.0 and 6.2, accounting for 15.8% and 16.2% of variance, respectively. An HNI allele in the nearest marker of the QTL enlarged the proportion of the head width to snout length (Table 2). The eight traits described above were further analyzed with composite interval mapping to confirm these results. Some QTL with low LOD scores on L29 and L33 could not be detected by composite interval mapping; however, all QTL that were identified with LOD ≥ 5.0 with interval mapping were confirmed (supplemental Table 11 at http://www.genetics.org/supplemental/). We did not find any traits with an extremely high phenotypic variance that could account for the total variance. The highest that we detected was 16.2% in V15, which was mapped to LG11.
Traits mapped with an LOD ≥ 5 for each of the three genotypes of the nearest marker to the QTL
DISCUSSION
We have analyzed the craniofacial traits of two inbred medaka strains, HNI and Hd-rR, and their F1 and F2 progeny. Our phenotypic survey shows that 304 of 444 quantified traits are gender dependent and 379 traits differ significantly between the two inbred strains. Thus, many craniofacial traits are highly heritable in medaka. Furthermore, 66 of 89 traits were localized in the medaka genome with interval mapping. This is the first demonstration that the QTL of craniofacial traits in medaka can be identified by genetic analysis of F2 progeny.
To measure the craniofacial traits, we collected magnified digital images and extracted the quantitative traits from the images using a computer. Direct measurements are difficult when the object is as small as medaka. The use of magnified digital images reduces measurement errors. Furthermore, individual traits can be quantified repeatedly using the images, thereby reducing human measurement errors. Finally, the same set of images can be used to test quantification methods, including length or ratio calculations with different landmark combinations, identification of novel landmarks, area determinations, and mathematical computations of various indices. To successfully detect a QTL, it is vital to extract trait data that are highly heritable and without much environmental influence. There is currently no standard efficient method for measuring craniofacial traits, however. Different laboratories utilize various techniques and refine these methods as needed (for example, Klingenberg et al. 2001, 2004; Albertson et al. 2003; Nishimura et al. 2003). The present method for determining craniofacial traits is effective and facilitates repeated trials on a two-dimensional display.
We calculated the ratios of two linear distances to generate phenotypic values and identified at least a QTL for 74.2% of traits, with the highest LOD score of 6.2. This is a simple, classic method that quantifies proportions of a shape independent from absolute size. There is no guarantee that variations in shape are captured completely, but quantification of ratios makes it possible to directly quantify differences that a researcher finds through careful observation. This simple technique has been utilized for genetic studies of craniofacial morphology in humans (Savoye et al. 1998; Johannsdottir et al. 2005) as well as quantification and mapping of wing shape traits in Drosophila (Zimmerman et al. 2000). Another more sophisticated approach is known as geometric morphometrics, which reflects the entire diversity of spatial patterns, thus allowing identification of QTL affecting all features in a single analysis (Klingenberg et al. 2001, 2004; Albertson et al. 2003). Some QTL in these studies have higher LOD scores than ours; however, it is difficult to directly compare the results. There are many factors that affect the power to detect QTL, including the type of samples, sample size, and statistical procedures. To clarify whether geometric morphometrics is a better method for detecting QTL in medaka, we plan to quantify our samples using this technique and directly compare the results with the current mapping results. Consistent mapping result of wing shape traits have been obtained by either geometric morphometrics or ratios (Zimmerman et al. 2000). Thus, at the very least, the ratio method we used is unlikely to be inferior to geometric morphometrics.
In this study, several craniofacial traits were frequently mapped to an overlapping region of the medaka genome (Figures 2–4⇑⇑). Shared QTL for several craniofacial traits have been reported for cichlids, East African fish species known for their divergent morphology (Albertson et al. 2003). In cichlids, the shape of the oral jaw apparatus, the craniofacial skeleton, and the teeth were analyzed. Statistically significant QTL clusters were found in the cichlid genome. Thus, craniofacial development may be pleiotropically regulated or controlled by several closely linked genes. Alternatively, traits on a common locus may share a common feature. This latter possibility is more likely in this study, because a feature can be represented by several ratios and because a given distance is used to calculate several ratios. For example, the traits mapped on both LG8 and LG23 were all associated with the position of the eye along the anteroposterior axis in the head region. Another clear example is the region on LG3 that contains 15 dorsal view traits (Figure 3). With the exception of D6, all of these traits are proportional measurements calculated with the same denominator, D7:13 (supplemental Table 1). The values were highly correlated in F2 medaka (supplemental Table 7). The numerators used to calculate these 14 traits are various lengths along the anteroposterior and left-right axes. Thus, these mapping results probably reflect characteristics specific to the D7:13 length. The D7:13 measurement, not each ratio, is the trait controlled by the region on LG3. When D7:13 was subjected to interval mapping, it was mapped to the same region on LG3 (data not shown). This example suggests that analysis of each distance may provide additional insight into the genetic structure of craniofacial morphology. We are preparing such an analysis of the distances corrected for the size of the entire head.
The ventral view traits, V14 and V15, were mapped to a shared region of LG11 with high maximum LOD scores (supplemental Table 4, Figure 4). These traits are associated with the ratio of head width to snout length, measured behind the eyes. Interval mapping of dorsal view traits localized two additional traits within the same region, D18 and D22 (see Figures 3 and 4). These traits also reflect the ratio of head width to snout length behind the eyes. Although there was little phenotypic correlation between the ratios in the ventral view and those in the dorsal view (r2 < 0.1, data not shown), medaka homozygous for the HNI allele at the nearest marker of the QTL had wider head-to-snout length ratios in all related traits (Table 2, data not shown). These findings strongly suggest that these ratios are regulated by a genetic factor(s) within a shared region of LG11. Future fine mapping studies are necessary to precisely identify the gene(s) involved in this regulation.
Non-normally distributed traits were excluded from this study. These traits might be controlled by a few major genes, however, making it relatively easy to detect QTL. Length of the pelvic spine in F2 progeny of threespine sticklebacks is non-normally distributed in two peaks and has been successfully mapped near the gene Pitx1 with an extremely high LOD score (82.8, Shapiro et al. 2004). We carefully examined the distributions of F2 progeny that were not normally distributed for such patterns, but no traits were distributed across a few clear peaks. We also tried to map a trait from each group classified by phenotypic correlations. No QTL with higher LOD scores were identified, however, and the percentage of phenotypic variance explained by the QTL did not increase (data not shown), indicating that we did not overlook traits controlled by a few major genes.
A number of studies have analyzed morphological phenotypes as quantitative traits in mice (Cheverud et al. 1997; Dohmoto et al. 2002; Nishimura et al. 2003; Lang et al. 2005). The LOD scores in our study of medaka are similar to those obtained in mice. When craniofacial traits were analyzed in >100 mice, a few QTL were identified for each trait with highest LOD scores of ∼5 (Dohmoto et al. 2002; Nishimura et al. 2003). When skeletal morphology was analyzed in 200 F2 mice, 7.3 was the highest LOD score and most traits were mapped to between 1 and 5 genomic regions (Lang et al. 2005). The ability to identify QTL is affected by a sample size. Increasing the samples size increases the probability of finding QTL with high LOD scores. When mandible shape was analyzed in 480 F2 mice, 21 traits were mapped to between 2 and 10 genomic regions. The highest LOD score in that study was 9.8 (Cheverud et al. 1997). We analyzed 184 F2 samples in this study, and mapped, at most, 5 genomic regions per trait (supplemental Table 4). The highest LOD score in our study was 6.2, just slightly lower than the LOD score of 7.3 obtained in the study of 200 F2 mice (Lang et al. 2005). Thus, the medaka can be an alternative animal model for the genetic analysis of morphological traits, which is less expensive and easier to prepare a large number of samples.
In conclusion, this study demonstrates that many of craniofacial traits in medaka are quantitative traits controlled by both environmental and genetic factors. It is possible to identify the genomic regions responsible for these traits using QTL analysis of F2 progeny from inbred medaka strains. Our genomewide study has identified many QTL in the medaka genome. Further analysis of these regions is necessary to identify the genes. We plan to utilize congenic strains carrying an individual QTL to resolve the genes. These congenic strains will take several years to construct, but these lines and the subcongenics derived from them will be very powerful tools for studying the effects of these genes and the interactions among them. By comparing our results with those from future genetic studies of craniofacial traits in other animals, common molecular mechanisms underlying craniofacial development may be identified.
Acknowledgments
We thank Keiko Yoshida and Yasuko Ozawa for their valuable assistance with this study, and Shuhei Mano, Tsuyoshi Koide, Juzoh Umemori, and Akinori Nishi for helpful discussions. The work presented here was supported in part by the Research and Study Program of Tokai University Educational System General Research Organization, by the Center of Excellence 21st Century Program, and by a Grant-in-Aid for Scientific Research on Priority Areas “Systems Genomics” and “Comparative Genomics” from the Ministry of Education, Culture, Sports, Science and Technology in Japan.
Footnotes
Communicating editor: N. Takahata
- Received February 5, 2007.
- Accepted October 8, 2007.
- Copyright © 2007 by the Genetics Society of America
