Aberrant cell segregation in craniofacial primordia and the emergence of facial dysmorphology in craniofrontonasal syndrome

Craniofrontonasal syndrome (CFNS) is a rare X-linked disorder characterized by craniofacial, skeletal, and neurological anomalies and caused by mutations in EFNB1. Heterozygous females are more severely affected by CFNS than hemizygous male patients, a phenomenon called cellular interference that is correlated with cell segregation resulting from EPHRIN-B1 mosaicism. Efnb1 heterozygous mutant mice also exhibit more severe phenotypes than Efnb1 hemizygous males as well as cell segregation, but how craniofacial dysmorphology arises from cell segregation is unknown and CFNS etiology therefore remains poorly understood. Here, we couple geometric morphometric techniques with temporal and spatial interrogation of embryonic cell segregation in mouse models to elucidate mechanisms underlying CFNS pathogenesis. By generating ephrin-B1 mosaicism at different developmental timepoints and in specific cell populations, we find that ephrin-B1 regulates cell segregation independently in early neural development and later in craniofacial development, correlating with the emergence of quantitative differences in face shape. Whereas specific craniofacial shape changes are qualitatively similar in Efnb1 heterozygous and hemizygous mutant embryos, heterozygous embryos are quantitatively more severely affected, indicating that Efnb1 mosaicism exacerbates loss of function phenotypes rather than having a neomorphic effect. Notably, tissue-specific disruption of Efnb1 throughout neural development does not appear to contribute to CFNS dysmorphology, but its disruption within neural crest cell-derived mesenchyme results in phenotypes very similar to widespread loss. Ephrin-B1 can bind and signal with EphB1, EphB2, and EphB3 receptor tyrosine kinases, but the signaling partner(s) relevant to CFNS are unknown. Geometric morphometric analysis of an allelic series of Ephb1; Ephb2; Ephb3 mutant embryos indicates that EphB2 and EphB3 are key receptors mediating Efnb1 hemizygous-like phenotypes, but the complete loss of EphB1-3 does not recapitulate CFNS-like Efnb1 heterozygous severity. Finally, by generating Efnb1+/-; Ephb1; Ephb2; Ephb3 quadruple knockout mice, we determine how modulating cumulative receptor activity influences cell segregation in craniofacial development and find that while EphB2 and EphB3 play an important role in craniofacial cell segregation, EphB1 is more important for cell segregation in the brain; surprisingly, complete loss of EphB1-EphB3 does not completely abrogate cell segregation. Together, these data advance our understanding of the morphogenetic etiology and signaling interactions underlying CFNS dysmorphology. Author Summary Craniofacial anomalies are extremely common, accounting for one third of all birth defects, but even when the responsible genes are known, it often remains to be determined exactly how development has gone wrong. Craniofrontonasal syndrome (CFNS), which affects multiple aspects of craniofacial development, is a particularly mysterious disorder because it is X-linked, but affects females more severely than males, the opposite situation of most X-linked diseases. The responsible gene has been identified as EFNB1, which encodes the EPHRIN-B1 signaling molecule that regulates cellular position. Why EFNB1+/- heterozygous females exhibit severe stereotypical CFNS phenotypes is not well understood, but it is related to the fact that X chromosome inactivation generates mosaicism for EPHRIN-B1. Using mice harboring mutations in the Efnb1 gene in different embryonic tissues, and in receptor genes Ephb1-3, together with quantitative methods to measure craniofacial structures in developing embryos, we establish the tissue-specific contributions of ephrin-B1 mosaicism to craniofacial dysmorphology. We also examine when ephrin-B1 regulates cellular position during different stages of craniofacial development and which EphB receptors are involved. Our results reveal the specific cellular context and signaling interactions that are likely to underlie CFNS, and provide new understanding of how EPHRIN-B1 may regulate normal craniofacial development.

is not biased by EFNB1 mutation [4,37], suggesting that loss of gene function does not impact cell survival.

10
Supporting the idea that mosaicism for ephrin-B1 expression results in more severe dysmorphogenesis, rare

41
Ephrin-B1 has a significant effect on embryonic facial shape from E11.5 to E14.5 that mirrors CFNS 42 Robust quantitative methods are required to investigate when the effects of mosaic expression of 43 ephrin-B1 on facial morphology first appear, whether the earliest facial shape effects parallel later facial shape 44 effects, how these change in severity over time, and whether phenotypic severity varies between heterozygous 45 females and hemizygous males. We therefore quantified mouse embryo facial shape at progressive stages 46 between E11.5 and E14.5 using geometric morphometrics analysis of landmarks collected on micro-computed 47 tomography (µCT) derived facial surfaces of Efnb1 +/Δ and Efnb1 Δ/Y embryos as well as a pooled control sample 48 of Efnb1 +/lox and Efnb1 lox/Y embryos that we refer to as Efnb1 wt . To determine the significance and relative 49 contribution of facial size (estimated as centroid size) and Efnb1 genotype in determining facial shape, we 50 carried out a Procrustes ANOVA analysis on E11.5 embryos using a published landmark set [45]. Facial size 51 and Efnb1 genotype both contribute significantly to facial shape of E11.5 embryos ( Table 1), explaining 52 approximately 23% and 11% of the facial shape variation, respectively. The significant genotype effect 53 indicates that ephrin-B1 mosaicism or loss influences facial shape as early as E11.5. Genotype-specific effects 54 on facial shape were interrogated to pinpoint specific regions where differences occur. Landmark-specific 55 shape change vectors for both mutant genotypes indicate increased facial width and decreased facial height, 56 with maxillary prominences more posterior in relation to vault landmarks (Fig. S1). Overall, there is evidence of 57 reduced anterior outgrowth of and greater lateral distance between the facial prominences in mutant mice.

58
Given a significant effect of the Efnb1 genotype on facial shape at E11.5, we performed morphometric 59 analysis on E12.5-E14.5 embryos to determine whether there was a change in the severity or type of facial 60 dysmorphology as the face outgrows. We used a novel landmark set that better captures facial shape at these 61 specific stages (Fig. S2). A Procrustes ANOVA analysis with facial size (estimated as centroid size), 62 embryonic age, and Efnb1 genotype as factors indicated that each contributes significantly to facial shape 63 ( Table 2). Additionally, the interaction between age and genotype has a significant effect on facial shape. As 64 expected for a sample covering multiple embryonic days, facial shape variation correlated with size (i.e., 65 allometry) explained 77% percentage of facial shape variation. The significant effect of Efnb1 genotype 66 explained almost 7% of facial shape variation. Visualization of landmark vectors illustrating genotype-specific 67 shape effects indicate overall similarities in the effects of Efnb1 Δ/Y and Efnb1 +/Δ genotypes on facial shape at 68 E14.5 (Fig. 1A-H). Both mutant genotypes display hypertelorism, represented by an increased relative width 69 between anterior eye landmarks. They also have a relatively inferior-posterior nose, anterior ear, and latero-70 posterior lip corners. Whereas Efnb1 Δ/Y embryos exhibited shorter faces, the degree of facial shortening was 71 more extreme in Efnb1 +/Δ embryos, as seen by longer vectors at the ear and nose landmarks (Fig. 1H).

72
Altogether, these shared patterns of dysmorphology indicate hypertelorism and facial shortening in both male 73 hemizygotes and female heterozygotes.

74
Similarities between E12.5-E14.5 and E11.5 mutant genotype effects suggest a continuity of shape 75 dysmorphology between E11.5 and E14.5. However, it was important to verify that effects at different 76 embryonic ages remain parallel after accounting for normal facial growth across this developmental period.

77
Given that 77% of facial variation of the E12.5-E14.5 sample was explained by size, it was not surprising that 78 the first principal component (PC) of a principal component analysis (PCA) of facial shape separates 79 specimens in this sample by embryonic age (Fig. 1I). A multivariate linear model was used to estimate the 80 allometric component of shape variation that is common across the sample regardless of genotype (Fig. 1J).

81
The residuals of this regression are interpreted as facial shape after accounting for size related shape 82 variation. The first PC of a PCA of these facial shape residuals represents a common axis of facial shape 83 covariation that separates genotypes (Fig. 1K), suggesting major similarities in mutant genotype effects on 84 facial shape across embryonic ages. Although individual PCs illustrate patterns of facial shape covariation, 85 they each represent only part of overall covariation. Therefore, we calculated Procrustes distances between 86 mean control and affected genotype facial shapes to confirm the significance of mean facial shape differences 87 between genotypes and to estimate the relative severity of facial shape dysmorphology. There were significant 88 differences in mean facial shape between control and each mutant genotype at all embryonic ages (Table 3).

89
In addition, within each age, the mean facial shapes of Efnb1 +/Δ embryos were always more different from 90 Efnb1 wt controls than were Efnb1 Δ/Y facial shapes. Finally, the facial shape of both mutant genotypes is more 91 different from controls at E14.5 than at E12.5, indicating an increase in severity of dysmorphology over this 92 embryonic period.

93
Based on our analysis, Efnb1 +/Δ and Efnb1 Δ/Y mice display similar types of dysmorphology, with 94 Efnb1 +/Δ females displaying quantitatively greater severity. Similarly, after accounting for normal growth 95 processes, the major axis of facial shape variation separates genotypes across embryonic ages, indicating 96 strong similarities in genotype effects that increase in severity across this period of growth. While these general 97 similarities across age and genotype exist, there are some noted differences in Efnb1 +/Δ and Efnb1 Δ/Y genotype 98 effects ( Fig. 1A-H). For example, Efnb1 +/Δ embryos display increased relative width of the posterior whisker 99 margins and a posterior-inferior corner of the whisker region whereas Efnb1 Δ/Y embryos do not. This suggests 00 a larger increase in relative width of the midfacial region in the female heterozygotes that is not matched by the 01 male hemizygotes. In addition, the female heterozygotes display a reduced length of the midline connection 02 between the whisker pads, that appeared as a midline notch in the upper lip, possibly analogous to a 03 shortened human filtrum (Fig. 1A, C, E, G). These results demonstrate that increased midfacial expansion is 04 exacerbated in Efnb1 +/Δ embryos compared with Efnb1 Δ/Y embryos, rather than resulting from distinct effects on 05 additional craniofacial structures.

07
Ephrin-B1-mediated cell segregation occurs in post-migratory neural crest-derived craniofacial mesenchyme 08 Cell segregation has been proposed to underlie increased severity in heterozygous female CFNS 09 patients with ephrin-B1 mosaicism. We have previously shown that cell segregation first occurs in the headfold 10 of E8.5 Efnb1 +/Δ embryos prior to NCC emigration [44], suggesting the possibility that early segregation of NCC 11 progenitors might result in the cellular distribution patterns we observe at later stages. Alternatively, later 12 segregation within post-migratory NCC-derived populations could result in increased CFNS severity. To 13 determine when and where cell segregation was occurring, we utilized a ubiquitously expressed X-linked GFP 14 (XGFP) transgenic allele to monitor normal patterns of X chromosome inactivation (XCI) at distinct stages of 15 development [44, 46,47]. We generated NCC-specific ephrin-B1 mosaic Efnb1 +XGFP/lox ; Sox10-Cre Tg/0 embryos 16 and examined them for segregation at E10.5, after NCC migration has populated the craniofacial mesenchyme. Sox10 is expressed throughout NCCs prior to their emigration, and we observed robust 18 recombination throughout the post-migratory NCCs including the maxillary process (MXP) and the frontonasal 19 prominence (FNP) in Sox10-Cre Tg/0 ; ROSA26 mTmG/+ reporter embryos (Fig. S3A, B). Notably, Efnb1 +XGFP/lox ; 20 Sox10-Cre Tg/0 embryos did not exhibit cell segregation in the MXP at E10.5 (Fig. S3E) and instead resembled 21 control Efnb1 +XGFP/lox embryos (Fig. S3C, D), indicating that cell segregation in migratory NCCs, if it occurs,

34
The finding that segregation occurs in E11.5 craniofacial mesenchyme demonstrates that ephrin-B1 35 mediates this process after NCC migration is completed. We next wished to determine whether segregation   observed in neurofilament-expressing maxillary trigeminal ganglion nerve cells at E11.5 (Fig. S4A, B), 42 recombination in the anterior palatal mesenchyme was first apparent at E12.5 (Fig. S4C, D). Consistent with 43 this timing of Shox2 IresCre onset, we observed no segregation in either genotype at E11.5 (Fig. S4E, F) but 44 small patches of segregated ephrin-B1/GFP expression in E12.5 Efnb1 +XGFP/lox ; Shox2 IresCre/+ embryos ( Fig.   45   S4H) compared with Efnb1 +XGFP/lox control embryos (Fig. S4G). Ephrin-B1 is therefore a driver of segregation 46 not only in the headfold and NCC progenitor cells, but also in post-migratory craniofacial mesenchyme. These 47 data demonstrate that ephrin-B1-mediated cell movements continue through development of craniofacial 48 structures, and segregation within these structures may continually contribute to CFNS dysmorphology.

49
We have demonstrated that differences in facial shape are evident in female Efnb1 +/Δ heterozygous 50 embryos as early as E11.5, but these shape changes continue to develop over time and increase in severity 51 through E14.5. To investigate how segregation later in development correlates with changes to craniofacial 52 tissue morphology, we examined embryos with ephrin-B1 mosaicism in specific cell types at E13.5. Control 53 embryos have strong ephrin-B1 expression in the tips of the anterior palatal shelves and lateral FNP consistent 54 with the CFNS-like phenotypes we discovered by morphometric analysis, while XGFP is visible in a fine-55 grained mosaic pattern in each structure (Fig. 3A, E). In full Efnb1 +XGFP/Δ heterozygotes, large ephrin-B1/GFP 56 expressing and non-expressing patches correlated with aberrant ephrin-B1 expression boundaries, including 57 irregularities of palatal shelf shape (Fig. 3B) and apparent bifurcations of the nasal conchae (Fig. 3F). Neural 58 crest-specific mosaic Efnb1 +XGFP/lox ; Sox10-Cre Tg/0 embryos exhibited a similar correspondence between 59 ephrin-B1/GFP patches and local dysmorphology in both the secondary palatal shelves (Fig. 3C) and nasal 60 conchae (Fig. 3G). Interestingly, in palate mesenchyme-specific Efnb1 +XGFP/lox ; Shox2 IresCre/+ heterozygotes, 61 small ephrin-B1/GFP expressing and non-expressing patches were apparent in the E13.5 anterior palate 62 mesenchyme (Fig. 3D). These patches appeared somewhat smaller than those in full or NCC-specific mosaic 63 embryos, and the palatal shelves were overall not as dramatically dysmorphic as Efnb1 +XGFP/Δ heterozygotes, 64 though local bending occurred at ephrin-B1 expression boundaries with small bumps surrounding the boundary 65 ( Fig. 3B, D). No segregation was evident in the FNP of palate mesenchyme-specific Efnb1 +XGFP/lox ; 66 Shox2 IresCre/+ heterozygotes, with no local dysmorphology in the nasal conchae (Fig. 3H). In total, these data 67 demonstrate that ephrin-B1 mediates segregation in the post-migratory NCC-derived mesenchyme of two 68 structures key to CFNS pathology and that these boundaries correlate with tissue structure dysmorphology.

71
The expression patterns of ephrin-B1 in the early neural plate, telencephalon and post-migratory 72 craniofacial neural crest, together with the finding that cell segregation can occur independently in each of 73 these contexts, led us to ask whether disruption in distinct tissues contributes to CFNS dysmorphology. We 74 have previously shown that ephrin-B1 mediates segregation in the neural plate neuroepithelium and that 75 segregation is apparent in the developing brain [44,52]. Apoptosis of neuroepithelial cells is observed together 76 with a reduction in cranial NCCs leading to abnormal craniofacial development in Tcof1 +/mutant embryos, a 77 model of Treacher Collins syndrome [53,54], and changes to the shape of the brain can indirectly cause 78 changes to facial shape [11,12]. We therefore wondered whether ephrin-B1 mosaicism in the brain could result 79 in changes to facial shape. Sox1 Cre mediates recombination in the neural plate as early as E8.5 [55], and 80 crossing to the ROSA26 mTmG reporter revealed widespread recombination throughout the brain at E13.5 (

91
Because neural-specific Efnb1 heterozygosity does not contribute to CFNS facial dysmorphology, we 92 quantified the gross facial shape effects of disrupted Efnb1 expression in NCC-derived tissues. Procrustes 93 ANOVA analysis indicated that Efnb1 +/lox ; Sox10-Cre Tg/0 genotype had a significant influence on facial shape 94 ( Table 5). Landmark-specific vectors of the facial shape effects indicated broadly similar directions of shape 95 change for Efnb1 lox/y ; Sox10-Cre Tg/0 hemizygotes and heterozygotes compared with control ( Fig. 5B, D). These 96 include hypertelorism, a relatively inferior rhinarium, and relatively anterior ear. The Efnb1 +/lox ; Sox10-Cre Tg/0 97 heterozygotes show increased width of the posterior whisker margins and a higher midline lip cleft when 98 compared to Efnb1 lox/y ; Sox10-Cre Tg/0 hemizygotes. As with the comparison of Efnb1 +/Δ and Efnb1 Δ/Y 99 genotypes, the severity of facial shape dysmorphology is lower in Efnb1 lox/y ; Sox10-Cre Tg/0 males than in 00 Efnb1 +/lox ; Sox10-Cre Tg/0 heterozygous females ( Fig. 5F; Table 6). Strong similarities in facial dysmorphology 01 are apparent between embryos with global disruption of Efnb1 and those with NCC-specific-loss. However, the 02 Procrustes distances between affected mice and wildtype mice are lower for the Sox10-Cre crosses (Table   03 3,6), suggesting a lower severity of facial dysmorphology when cell segregation occurs only in NCC-derived 04 structures. In summary, these morphometric results quantitatively demonstrate that neural-specific disruption 05 of Efnb1 has no effect on facial shape in CFNS dysmorphology, while NCC-specific disruption leads to facial 06 shape effects that are similar to, but slightly milder than those resulting from global disruption of Efnb1 07 expression.  Table 7). The proportion of facial 19 shape variation explained by variation in the Ephb1 null mutation is 1%, while Ephb2 genotype explains 6% 20 and Ephb3 genotype explains 10% (Rsq values). Specimens with more null alleles across all three receptors 21 tended to have facial shapes more similar to Efnb1 +/Δ and Efnb1 Δ/Y specimens, but each receptor contributed to 22 facial shape change to a different extent (Fig. 6A). For example, specimens that were homozygous null for 23 Ephb1 often had facial shapes similar to Efnb1 wt mice, while specimens that were homozygous null for Ephb2 24 usually had facial shapes more similar to Efnb1 Δ/y mice (Fig. 6B). So, while genotype of each receptor was 25 associated with a significant shape effect, the facial shape effect of Ephb1 genotype explained less facial 26 shape variation than Ephb2 or Ephb3 genotypes and was associated with less severe phenotypic effects.

27
Interactions between multiple Ephb receptor genotypes further explained facial shape variation across 28 this triple null series. For example, some of the variation across specimens that were homozygous null for 29 Ephb1 resulted from heterozygosity of other receptors. Ephb1 homozygotes with no other null Ephb alleles had 30 facial shapes like Efnb1 wt mice, indicating weak or no independent impact of Ephb1. Ephb1 -/-; Ephb2 +/-31 embryos also displayed wildtype-like phenotypes; however, Ephb1 -/-; Ephb2 +/-; Ephb3 +/exhibited phenotypes 32 more similar to Efnb1 Δ/Y mutant embryos (Fig. 6D). Ephb3 -/null mutants exhibited an intermediate facial 33 phenotype with the severity of dysmorphology increased by Ephb2 heterozygosity (Fig. 6E). While many 34 specimens that were homozygous null for one receptor gene showed wildtype-like facial shape, most 35 specimens that were homozygous null for two receptor genes displayed more severe dysmorphology (Fig.   36   6C). However, the embryos that were homozygous null for both Ephb1 and Ephb3 clustered into two groups 37 along major axes (PCs) of facial shape variation. This separation of specimens was based on whether these 38 specimens were also heterozygous for Ephb2 (Fig. 6F), indicating that having two wild-type copies of Ephb2 in 39 embryos without EphB1 or EphB3 function can lead to a notably milder facial phenotype.

40
We have previously demonstrated that loss of forward signaling through EphB2 and EphB3 resulted in 41 a loss of cell segregation in the neural plate of Efnb1 +/Δ embryos at E8.5. Because ephrin-B1 cell segregation 42 occurring within the post-migratory NCC-derived mesenchyme appears to drive CFNS dysmorphology, we 43 genetically tested which receptors were required for cell segregation in the secondary palate, FNP and brain.

44
We generated compound Efnb1 +/Δ mutant embryos also harboring loss of function of different combinations of 45 Ephb1, Ephb2 and Ephb3 alleles and analyzed cell segregation at E13.5 by ephrin-B1 immunostaining.

53
Even complete loss of EphB1, EphB2 and EphB3 was not sufficient to completely abrogate ephrin-B1-54 mediated cell segregation in the palate and FNP, suggesting that additional receptors may contribute to cell 55 segregation in this context. In the brain, a somewhat different priority of receptor requirement was observed.

56
Again, cell segregation was apparent in most Efnb1 +/Δ ; Ephb1-3 compound mutant embryos, though the extent 57 of intermixing and distribution of patches was different with different receptor combinations (Fig. S7). Notably,

63
From its description as a subgroup of frontonasal dysplasia that affects females more severely than

06
CFNS is not caused by defects in NCC migration as previously suggested, but rather reflects a role for Efnb1 07 in shaping the craniofacial primordia following migration. Notably, we found that Efnb1 +/Δ mutants exhibit 08 changes in tissue shape such as bending, folding and bifurcations in the secondary palate and FNP that 09 correlated with ectopic ephrin-B1 expression boundaries. How exactly local dysmorphology exacerbates 10 phenotypic severity is uncertain, but it may be that the ephrin-B1 expression pattern constrains the regions of 11 greatest dysmorphology which then leads to stereotypical CFNS face shape changes. Additionally, these 12 findings may suggest the existence of previously unappreciated tissue boundaries that exist in the craniofacial 13 mesenchyme that are lost in Efnb1 Δ/Y hemizygous males, but ectopically imposed in Efnb1 +/Δ embryos. Further 14 studies will be needed to determine how these aberrant boundaries and/or disruption of boundary maintenance 15 contribute to craniofacial phenotypes.

16
Although segregation occurs dramatically in neural precursor cells at the neural plate and is present in

99
Immunofluorescence. Embryos were fixed in 4% PFA in PBS, dehydrated through sucrose, embedded in OCT, 00 and frozen in dry ice/ethanol. 12m sections were cut using an HM550 (Thermo Scientific) or a CM1900 01 (Leica) cryostat. Slides were washed with PBS, blocked in 5% normal donkey serum (Jackson 02 ImmunoResearch) and 0.1% Triton-X-100 in PBS, incubated in primary antibody overnight at 4°C, washed with 03 PBS, and incubated in secondary antibody at room temperature (for antibody information, please refer to Table   04 8). Slides were counterstained in DAPI (Millipore) in PBS and coverslips were mounted on slides using 05 Aquamount (Thermo Scientific) for imaging. Images were obtained on an Axio Imager.Z2 upright microscope 06 using an AxioCamMR3 camera and AxioVision Rel.4.8 software (Zeiss).

28
Procrustes superimposition was performed on landmarks to align each specimen and remove scale from 29 analysis. Procrustes ANOVA analysis, with permutation-based tests for significance, was used to determine 30 whether size (numeric; centroid size), genotype (factor; Efnb1 +/Δ , Efnb1 Δ/Y , Efnb1 wt ), age (numeric; 12.5, 13.5, 31 14.5) and their interactions have a significant influence on facial shape (α=0.05). We visualized the effects of 32 Efnb1 +/Δ and Efnb1 Δ/Y genotypes on facial shape by plotting differences between predicted genotype-specific 33 shapes estimated from the Procrustes ANOVA multivariate linear model (assuming E14.5 age and average 34 E14.5 centroid size). Given the strong changes in facial shape that normally occur between E12.5 and 14.5,

35
we completed a multivariate regression of facial shape on centroid size to estimate allometry and used the 36 rescaled residuals of that regression as "allometry-corrected" coordinates for further analysis. Principal 37 component analyses of coordinate values were completed both before and after "allometry correction" to 38 visualize patterns of specimen clustering along major axes of facial shape covariation within the sample.

39
Procrustes distances between mean control and affected facial shapes were calculated from residual landmark 40 coordinates at each age to determine whether genotypes displayed significantly different facial shapes.

41
Significance was determined by comparing Procrustes distances to 95% age-specific confidence intervals that 42 were estimated with 1000 permutations of distances between two randomly selected control groups of 15 43 specimens. Geometric morphometric analysis of the E11.5 sample was completed in the same way, except 44 without age as a factor in the Procrustes ANOVA analysis and without allometry correction, because only one

29
Generation of ephrin-B1 mosaicism specifically in neural progenitor cells using Sox1 Cre results in dramatic segregation throughout the brain of E13.5 Efnb1 +XGFP/Δ ; Sox1 Cre/+ embryos, visible as large patches of ephrin-31 B1 and GFP expression.    Estimate of the influence of genotype (as a factor) on facial shape.

41
c Rsq provides an estimate of how much facial shape variance a given covariate explains.

42
* indicates a significant effect on facial shape, as calculated using a permutation test.

51
* indicates a significant effect on facial shape, as calculated using a permutation test.   Estimate of the influence of genotype (as a factor) on facial shape.

70
c Rsq provides an estimate of how much facial shape variance a given covariate explains.

71
* indicates a significant effect on facial shape, as calculated using a permutation test.  Estimate of the influence of genotype (as a factor) on facial shape.

79
c Rsq provides an estimate of how much facial shape variance a given covariate explains.

80
* indicates a significant effect on facial shape, as calculated using a permutation test.

87
* indicates a significantly different facial shape than E14.5 Efnb1 wt controls used for comparison to Efnb1 +/Δ and 88 Efnb1 Δ/Y ; based on the 95% control confidence intervals produced by bootstrapping.

89
^ Although both Sox1 Cre controls and heterozygote facial shapes are significantly different than ß-actin-cre 90 controls, they are not significantly different from each other. Estimate of the additive influence of a specific EphB genotype (as a factor) on facial shape.

96
c Rsq provides an estimate of how much facial shape variance a given covariate explains.

97
*indicates a significant effect on facial shape, as calculated using a permutation test.  The ventral most midline point along the developing lip 3 (15) Dorso-caudal corner of the whisker field, taken on the skin right next to the plateau of the whisker field, rather than on the field itself 4 (16) Ventro-rostral tip of the plateau on the ventro-rostral member of the supra-orbital vibrissae pair that is found dorsal to the eye 5 (17) Rostral apex of the forming Medial Canthus of the eye 6 (18) Caudal apex of the forming Lateral Canthus of the eye 7 (19) Center of the infraorbital vibrissa found ventral to the eye 8 (20)

99 00
Point at the rostral base of the dorso-caudal portion of the developing pina of the ear 9 (21) Point at the rostral base of the ventro-rostral portion of the developing pina of the ear 10 (22) Point at the edge of the whisker margin between the second and third whisker rows, counting from the top. This point is frequently next to the second large mystacial vibrissa. 11 (23) Ventro-caudal corner of the whisker field, taken on the skin right next to the plateau of the whisker field, rather than on the field itself 12