The cis-regulatory logic underlying abdominal Hox-mediated repression versus activation of regulatory elements in Drosophila

Arya Zandvakili, Juli Uhl, Ian Campbell, Yuntao Charlie Song, Brian Gebelein

Preprint posted on July 20, 2018

From sequence to function: a recent preprint highlights the importance of binding site orientation and spacing for determining transcription factor activity

Selected by Clarice Hong


How do transcription factors recognise their binding sites in vivo? Despite binding to very similar DNA sequences in vitro, the Hox family of transcription factors regulate distinct sets of genes which give rise to unique cell identities along the developing body in animals. Furthermore, Hox factors have been found to both activate or repress target genes depending on their context. One model that partially explains this phenomenon is the presence of other transcription factors binding sites close to the Hox binding sites, as cooperativity between other transcription factors and different Hox genes can modulate their specificity. However, it remains unknown what configurations of binding sites are required for the activity of different Hox factors. This preprint investigates how combinations of binding sites confer binding specificity of Hox factors, which is crucial in furthering our understanding of the cis-regulatory grammar that governs transcription factor binding.

Key findings

The authors focused on studying two Hox transcription factors, Ultrabithorax (Ubx) and Abdominal-A (Abd-A) in Drosophila that regulate the expression of distinct target genes. In order to distinguish between the activating and repressive activity of Hox factors, the authors also used two target cis-regulatory elements of Hox, the DCRE, which mediates repression in abdominal segments, and RhoBAD, which activates expression in a subset of abdominal sensory organ precursor cells (SOPs). A series of synthetic constructs containing the cis-regulatory elements in various configurations driving expression of a LacZ reporter were created and integrated to generate transgenic flies. The expression of LacZ was then measured in the abdominal segments of the developing fly to determine the activity of each construct.

The primary claim of the preprint is that at least part of the differential activity of the two Hox factors comes from the configuration of Hox binding sites. Interestingly, while the Hox binding site explored here comprises of Exd/Hth/Hox, it is not the configuration of these sites that matter, instead, it is the relative location of Exd/Hth/Hox to other binding sites in the genome. Both Ubx and Abd-A can repress expression through the DCRE, but only Abd-A can activate transcription through RhoBAD. However, the reason for Abd-A specificity for RhoBAD does not appear to be due to the configuration of binding sites in the RhoBAD element, and both Ubx and Abd-A appear to have similar binding affinities for RhoBAD. Furthermore, the authors showed the orientation of the Hox binding sites relative to another transcription factor binding site, FoxG, along with the orientation of the FoxG binding sites is crucial for DCRE-mediated repression, because reversing the orientation of the elements leads to loss of repression. Intriguingly, the authors also systematically altered the spacing between Hox elements and the FoxG site in the DCRE, and between Hox elements and the Pax2 site in the RhoBAD elements. This experiment clearly demonstrated the need for proper spacing between the Hox binding sites and its partners, as different spacing between the binding sites led to drastically different results.

What I liked about the preprint

What I liked most about this preprint is the use of Hox transcription factors to study cis-regulatory activity. Not only are the Hox transcription factors intriguing because of their uncanny ability to regulate so many different target genes in different cell types, they can also activate or repress genes in different contexts. Most studies thus far have focused on the ability of cis-regulatory elements to activate transcription, but not repression, and it is thus interesting to see what kind of grammar regulates repression. The authors also measured expression of the reporter gene in vivo, in the abdominal segments of the fly. This is interesting because most of the experiments studying the grammar of cis-regulatory elements are conducted in cell lines, which may not have the most relevant conditions for transcription such as the concentration of transcription factors in the cells. The experiments about spacing between binding sites affecting activity are also very intriguing. The authors attempted to alter the phasing between binding sites with their spacing, and showed that at least in the case of FoxG, when 5bp was added between FoxG and the Hox binding site (half a turn of the DNA helix), repression is completely lost, but when 10bp was added (1 turn of the DNA helix), the repression was partially rescued. This seems to suggest that the Hox factors and FoxG need to be on the same side of the helix for it to function. With this, the authors succeeded in using simple sequence manipulations to understand some biochemical basis of transcription factor cooperativity. Finally, I also feel that the discussion was extremely comprehensive and insightful, which made the paper altogether delightful to read.

Future directions and questions

Because of the unique ability to characterise both Hox activation and repression, it will be interesting to further characterise how Hox factors mediate repression in different contexts and how it distinguishes between activation and repression. Additionally, while many sequences were assayed in this preprint, it would be interesting to test more Hox factors and/or more configurations of binding sites to not only test the effect of orientation and spacing, but also things like binding site affinity and positioning of the binding sites. A comprehensive survey of different combinations using an assay like a massively parallel reporter assay could be very informative.

One question that I had, however, is why Ubx is able to bind to the Exd/Hth/Hox binding sites in the context of DCRE to repress transcription but not in the context of RhoBAD to activate transcription. Are there sequences flanking the binding site in the RhoBAD element that are perhaps conducive to Abd-A binding, or impervious to Ubx binding? Additionally, given that the binding partners are so important in modulating their activity, is the expression of the partners (such as FoxG and Pax2) sufficient to fully explain the activity of its Hox partner in a given cell type? I would expect that other binding partners are going to be necessary for modulating Hox activity, and that their location relative to the Hox binding sites would have had an impact in these experiments. While care was taken to not introduce any binding sites in the spacing experiments, is it possible that binding sites were destroyed when altering the orientation/spacing of the relevant sites?





Posted on: 8th August 2018 , updated on: 30th October 2018

Read preprint (No Ratings Yet)

  • Author's response

    Arya Zandvakili and Brian Gebelein shared

    We would like to thank Clarice Hong for the very well written “Prelight” of our preprint article focused on how Hox factors regulate distinct outcomes based on cis-regulatory logic.

    Clarice highlights three important and largely unanswered questions:

    (1) Are there sequences flanking the binding site in the RhoBAD element that are perhaps conducive to Abd-A binding, or impervious to Ubx binding?

    It is certainly possible that flanking sequences contribute to the Abd-A selectivity of the RhoA sequence. What we found is that when the Hox-Hth-Exd binding sites were swapped from RhoA into the DCRE, Abd-A selectivity was lost, as it now was also regulated (repressed) by Ubx. This swap included a relatively large sequence to represent the Hox binding site (8 nucleotides, which is larger than the core binding site identified for Hox factors). Moreover, we previously showed that mutating all of these flanking sequences did not significantly impact RhoAmediated activation in abdominal SOP cells (Li-Kroeger et al 2012). Hence if additional sequences contribute to Abd-A selectivity to RhoA, it would require something beyond this DNA sequence and thus beyond the “typical” Hox binding site.

    (2) Is the expression of the partners (such as FoxG and Pax2) sufficient to fully explain the activity of its Hox partner in a given cell type?

    This is an important question that needs to be addressed to determine how Hox factors produce differential activities. Since many abdominal sensory organ precursor cells (SOPs) express both Abd-A and Pax2 and RhoA is only expressed in one specific subtype of these SOPs, it is highly likely that additional factors and/or post-translational modifications of Pax2 or Abd-A are required for the cell-type specific activation of RhoA. In contrast, the wild type DCRE sequence that contains FoxG and Hox sites can be repressed in all abdominal ectodermal cells that co-express FoxG (Slp2) and Ubx or Abd-A. However, we should point out that mutations in the DCRE that weaken the binding of these factors can result in a preferential loss of repression in the ventral-most FoxG and Ubx/Abd-A positive cells. This finding also suggests that additional factors and/or post-translational modifications could modify the DCRE-mediated-transcriptional response.

    Importantly, if partner TFs fully explained the differential activity of Hox factors (e.g. if Pax2 explained by Abd-A but Ubx regulates RhoA), then we could leverage the mere presence of partner TF binding sites near a Hox binding sites in a putative CRM to predict the activity of the CRM. An alternative model would be that the partner TFs must be arranged or positioned in a particular manner to distinguish between Hox factors. For example, while it appears that all Hox factors can form trimeric complexes with the transcription factors Extradenticle/Pbx (Exd) and Homothorax/Meis (Hth), cis-regulatory sequences with adjacent Hth-Hox binding sites are preferentially bound by posterior Hox-factors (Shen et al, 1997; Jolma et al, 2015). Our prediction is that a mix of both these models will be necessary to explain Hox-specificity of CRMs.

    (3) Is it possible that binding sites were destroyed when altering the orientation/spacing of the relevant sites?

    It is always possible that additional, unknown transcription factors bind and contribute to the regulation of cis-regulatory elements. What we can say is that EMSAs demonstrated that our sequence manipulations did not substantially affect TF binding of the known factors to the DCRE and RhoA variants. In addition, as mentioned above, the TF binding sites that we have mapped to the DCRE appear to be sufficient to explain the repressive activity of the DCRE in the abdomen. However, we do not have a sufficient set of TF binding sites to fully explain RhoA activity. Therefore, as we noted in the Discussion, it is possible that our manipulations of the RhoA sequence disrupted the activity of an unknown factor.

    Jolma A et al. Nature. 2015 Nov 19;527(7578):384-8. doi:10.1038/nature15518. PMID:26550823.
    Li-Kroeger D et al. Development. 2012 May;139(9):1611-9. doi:10.1242/dev.077842. PMID:22438572.
    Shen WF et al. Mol Cell Biol. 1997 Nov;17(11):6448-58. PMID: 9343407.

    Have your say

    Your email address will not be published. Required fields are marked *

    This site uses Akismet to reduce spam. Learn how your comment data is processed.

    Sign up to customise the site to your preferences and to receive alerts

    Register here