A  bimodal  chromatin  structure  at  the  snake  HoxD  locus

The reallocation of mesodermal enhancers together with the fact that snakes are limbless, could have resulted in the absence of a bimodal chromatin structure at the snake HoxD locus. However, previous work has shown that a bimodal partitioning that is necessary for the differential expression of Hoxd genes is present at the zebrafish locus and therefore predates tetrapod evolution (Woltering et al., 2014). Our results show that the same conformation is maintained in a limbless tetrapod such as the snake.

Our whole embryo 4C-seq data, however, revealed important changes in DNA-DNA contacts between mouse and snake tissues. Perhaps the most striking one is perceived when Hoxd9 is used as a viewpoint. Indeed, in the snake, the Hoxd9 contact profile strongly resembles that of Hoxd4, owing mainly to a lack of interactions with the second part of the 3’

gene desert. This region of the HoxD regulatory locus has been shown to specifically increase in interaction with Hoxd9-Hoxd11 genes during proximal limb development (Andrey et al., 2013). Therefore, the snake Hoxd9 contact profile could reflect the absence of important regulatory sequences located in this particular region of the 3’ regulatory gene desert.

Our results also show a more homogeneous profile of interaction for all snake viewpoints used in this analysis, compared to the mouse. This could reflect either a specialization of specific key enhancers in the mammalian regulatory landscape, or the disuse of long-range enhancers in the snake. 4C-seq profiles of other vertebrate species would be necessary to address this issue. The same holds true for the presence of a larger number of CTCF binding sites in the snake HoxD locus compared to the mouse counterpart. However, no functional explanation for this observation was found and whether mammals have discarded CTCF binding sites that were unnecessary or if extra CTCFs at this locus would have been beneficial for the snake is unknown.

Interestingly, though, the extra CTCF binding events observed in the 3’ region of the corn snake HoxD cluster (Hoxd3-Hoxd1 region) correlate with the increased size that characterize squamate clusters. Indeed the orientation of these binding sites would allow for interactions mostly with either the 5’ gene desert or in the Hoxd11-Hoxd8 region of the cluster. A hypothesis could therefore arise whereby the larger size of the corn snake cluster would make CTCF-mediated interactions within the cluster necessary for correct Hoxd gene expression.

3.3.1 Divergent regulatory strategy for genital bud patterning

Profound differences in regulatory strategies at the snake HoxD locus were not restricted to main body axis expression. Indeed, the snake cluster BAC line was the only one to be able to elicit expression in the genital bud and limbs. This result is reminiscent of mouse isolated mesodermal enhancer reporter assays that often activate reporter gene expression in secondary structures (Charité et al., 1995; Kwan et al., 2001; Sharpe et al., 1998). Therefore, it would seem that cis-regulatory regions inside the cluster confine Hox gene expression to the trunk, thus preventing its expansion to the developing limb bud. This mechanism would allow for a separate regulatory modality to be employed in the developing limb. In light of these observations, it would seem that the corn snake cluster is unable to prevent Hoxd genes from

being expressed in limbs and external genitalia. Furthermore, we find that this lack of repressive regulatory activity is also present in the snake endogenous context.

GT regulatory elements described in the mouse, although conserved in the snake, do not show the expected increase in interaction with the snake cluster. Expression analyses indicate that the majority of Hoxd genes that are expressed in the snake HP are regulated in the same way as in the main body axis at that same anterior-posterior position. However, Hoxd13 and perhaps Hoxd11 appear to be specifically activated in the snake genital bud.

Considering the lack of specific interactions arising when using Hoxd13 or Hoxd11 as viewpoints in the snake HP tissue, the regulatory strategy might be based on the use of pre-existing contacts that likely contain genital bud enhancers (Figure 26). We have shown this to be the case in two instances. Both IslandI and Prox are conserved constitutive contacts and their snake versions have the ability to drive reporter gene expression in the mouse GT. The Prox in particular was one of the few interactions that appeared to be reinforced with the cluster in the snake HP. This distinct regulatory strategy operating in the snake to activate Hox gene expression in the developing genitalia is not entirely surprising since recent work has revealed that mammals and squamate external genitalia have different developmental origins (Tschopp et al., 2014).

Figure 26 – Proposed model for the change in Hoxd gene regulation in snake genitalia

In the mouse, a number of sequences in the centromeric desert contact the cluster in all tissues, including the trunk region at the same anterior-posterior location as the GT. In the GT tissue, these constitutive contacts are still present but new specific contacts are also scored. In the snake, however, it would seem that in the HP tissue no new contacts are formed with the cluster but that instead constitutive contacts are used to elicit Hoxd gene expression in the HP. Black boxes are Hoxd genes, the tissue is either trunk tissue highlighted in green (top) or GT/HP tissue highlighted in red (bottom). Mouse (left) and snake (right) proposed interaction models are illustrated with structural proteins as blue

circles maintaining constitutive interactions while GT transcription factors, shown as red circles, are only present in GT/HP tissue.

The reallocation of regulatory elements to the inside of the cluster as well as the exploitation of regulatory modalities employed in other tissues and the use of enhancer-bearing constitutive contacts might all be regulatory adaptations to ensure proper Hoxd gene expression in the context of a large TE-rich cluster.

The loss of limbs and a highly distinct regulatory strategy driving Hoxd gene expression in snake genitals did not result in the lack of sequence conservation of mouse limb and/or GT enhancers. These observations were however in agreement with recent genome-wide results, which report that limb enhancers are often conserved in snakes (Infante et al., 2015). The same study found a significant overlap between limb and genital cis-regulatory mechanisms in agreement with the very similar molecular tool kit employed to generate the two structures. A Tbx4 enhancer with regulatory activity in both the GT and the limb of the mouse was shown to lose the ability to drive expression in the limb in snakes (Infante et al., 2015). Consistent with this observation we also find that the snake Prox specifically loses its activity in limbs while keeping a strong regulatory function in the GT. A more striking result was the finding of a limb-only enhancer in mouse (IslandI) that in the snake shows GT-only regulatory specificity. This could be a result of divergence from a dual function enhancer that became specialized for two different roles in two separate species. Alternatively, the GT enhancer function of IslandI could have been co-opted from an ancestral limb enhancer function. Investigating the enhancer activity of this sequence in different tetrapod species could clarify this issue. In addition, functional assays would also be necessary to establish if other mouse limb enhancers have been adapted to perform genital patterning functions in the snake.

Mouse proximal limb enhancers CNS39 and CNS65 where the sequences of more species were assayed, originated more conflicting results. While the snake CNS65 is unable to drive reporter gene expression in the mouse limb, a very reduced expression is also scored in limbed tetrapods such as the chicken and the lizard. In addition, the snake CNS39 sequence elicited an unexpected strong reporter gene expression in the mouse limb. CNS39 is not just a limb enhancer but also has an important structural function at the HoxD locus, probably owing to three CTCF sites that surround this sequence (Andrey et al., 2013) (Figure 9). This structural function could have exerted a strong selective pressure to preserve the CNS39 sequence in the snake, which would explain its ability to respond to the limb upstream effectors in a mouse context.

3.3.2 Intrinsic bimodal organization of the human HoxD cluster

The complex mechanisms behind the bimodal organization at the HoxD locus are not fully understood. Two mouse lines containing each a random integration of a full human HoxD cluster, showed that this structure is not intrinsic to the HoxD cluster itself but that it is dependent on the genomic neighborhood. In this case the site of integration determined whether there was a differential preference of interaction to one of the surrounding genomic landscapes.

The site of integration that elicited a bimodal organization from the cluster is a large gene desert. It is therefore possible that a gene poor region is necessary for this conformation to take place. Interestingly, none of the integrations occurred in a TAD with strong intradomain interactions and therefore the chromatin organization around the insertion in a

“stronger” TAD is unknown. Interaction with surrounding CTCF binding sites occurred in the two instances making this unlikely to be a factor in establishing a bimodal chromatin structure.