Pervasive effects of trans-acting variation on CTCF occupancy
3.3.4 Dominant inheritance affect cis-directed CTCF occupancy
Previous work has shown inherited TF binding sites to be expressed in an additive or non-additive (dominant) fashion[257, 542, 594]. Additive inheritance was observed when the combined binding intensity of the two F1 alleles is equivalent to the sum of the two parental F0 alleles, whereas dominant (non-additive) inheritance when the total allelic binding signal from the F1 alleles is equal to that of either F0 parent. Dominant inheritance; therefore, could also be sub-categorized as “high” if the signal from the F1 alleles equals that of the parent with the higher binding intensity, or “low” if that signal is similar to the one from the parent with the lower binding intensity signal (Figure 3.6a). Inheritance in this context is defined by the total allelic signal from each replicate, whereas regulatory categories discussed above were assigned based on the ratio of signal between the F1 alleles and the ratios of their F0 parent of origin. As with regulatory and lineage-specific category assignment, we fitted statistical models to test the three inheritance patterns outlined above, using Bayesian Information Criteria (BIC) to assess the outcomes in manner equivalent the one reported by Wong et al.[257] (see Methods). Of the CTCF binding sites under cis- influenced variants, 1021 passed the BIC difference minimum of 1 for inheritance pattern assignment, and 178 of the trans-acting variants were assigned an inheritance pattern with BIC > 1. In both cases of cis- and trans-acting variation, the predominant form was additive inheritance in which the total allelic signal from the F1 was equal to the sum of both parental allelic signals (55% and 43% respectively) (Figure 3.6b). The contributions of dominant inheritance of the inheritance observed in cis- and
trans-influenced CTCF sites; however, were not equal. Most dominantly inherited cis-
acting CTCF sites belonged to the dominant high variety of non-additive inheritance, in which the total allelic signal from the F1 was equal to that of the parent with the higher binding signal (34% vs 11% for dominant low). The same was observed, albeit to a much smaller scale, in trans-influenced CTCF sites (36% high vs 21% for low). When stratified by the F0 parent in with the higher median binding intensity (F0MAX),
the general distribution of inheritance modes did not differ between BL6 and CAST, and all trends in total, cis and trans were consistent in both mouse subspecies, and reflected the overall pattern (Figure 3.6b). A slight enrichment of sites inherited in dominant low form in BL6 in trans was observed, but owing to the small numbers involved, this might be a small number effect.
Figure 3.6: CTCF occupancy affected by cis-acting variation is show higher dominant effects
a A schematic model for assigning modes of inheritance for the cis- and trans-
influenced TF binding sites. F0MAX and F0MIN refer to the F0 parental subspecies
with the higher and lower median binding intensity, respectively. Binding intensities were summed across replicates in F0, and across alleles for F1. b The bar plots (top) show the proportion of cis- and trans-acting variants in CTCF binding sites based on their assigned mode of inheritance. The circle plot (bottom)
breaks down each mode of inheritance by the F0 father of origin with the higher median binding intensity (F0MAX), with the number of sites per subspecies
denoted inside the circles. The radius of each circle indicates the proportion of that mode of inheritance for the particular F0MAX parent for the same category
of cis/trans variation. c Pie charts showing the relative proportions of the three modes of inheritance for CTCF and 3 liver-specific TFs as determined using a statistical model to fit binding sites affected by cis or trans variation (see Methods). d A heatmap showing CTCF (left) binding events affected by cis- and
trans-acting variation. Different modes of inheritance were defined in a (see
Methods for the statistical model). The data from CEBPA assigned modes of inheritance (right) for both cis- and trans-acting variation was used for comparison (see Methods). Total F1 counts were individually scaled to 1 (yellow).
This observation, in the case of cis-acting variation, is significantly different from the pattern observed in other TFs, where although the most prevalent mode of inheritance was additive (Figure 3.6c), the contribution of dominant inheritance was much reduced (c2 test for pairwise comparison between CTCF and other TFs with
Bonferroni's correction, all p-values < 2.2e-16). Although non-additive inheritance was the predominant form for trans-acting variation in CTCF and other TFs, the contributions of additive and both forms of non-additive inheritance varied in TF- specific fashion. For example, the enrichment of the dominant high mode of inheritance in trans-acting variation observed in CTCF was not seen in HNF4A, in which the dominant low mode was more common (Figure 3.6c).
A close inspection of the ratios of the signal in CTCF cis and trans in comparison with CEBPA reveals that the overall effect of regulatory variations in additive inheritance tends to centre the ratios of F0MIN to F1 towards 1, with few sites showing
enrichments towards the lower ends (Figure 3.6d). Only dominant inheritance of the low variety shows a clear difference of the pattern in those ratios between F0MAX and
F0MIN in cis-influenced inheritance (Figure 3.6d). This indicates that in CTCF even
binding sites classified as influenced by cis-acting variation are under a clear dominant
low influence that skews the inheritance pattern from the expected additive mode,
which in turn could explain the increased effect of non-additive inheritance observed in the ratios of CTCF total binding signals compared to other TFs.