Structure, function and evolution of topologically associating domains (TADs) at HOX loci

FEBS Letters xxx (2015) xxx–xxx

journal homepage: www.FEBSLetters.org

Review

Structure, function and evolution of topologically associating domains (TADs) at HOX loci Nicolas Lonfat a,1, Denis Duboule a,b,⇑ a b

School of Life Sciences, Ecole Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland Department of Genetics and Evolution, University of Geneva, 1211 Geneva, Switzerland

a r t i c l e

i n f o

Article history: Received 17 March 2015 Revised 16 April 2015 Accepted 17 April 2015 Available online xxxx Edited by Wilhelm Just Keywords: Hox genes Topologically associating domains (TADs) Enhancers Evolution Pleiotropy Architecture Regulation External genitalia Genital tubercle (GT) Intestinal cecum Digits Limbs

a b s t r a c t Hox genes encode transcription factors necessary for patterning the major developing anterior to posterior embryonic axis. In addition, during vertebrate evolution, various subsets of this gene family were co-opted along with the emergence of novel body structures, such as the limbs or the external genitalia. The morphogenesis of these axial structures thus relies in part upon the precisely controlled transcription of specific Hox genes, a mechanism involving multiple long-range enhancers. Recently, it was reported that such regulatory mechanisms were largely shared between different developing tissues, though with some specificities, suggesting the recruitment of ancestral regulatory modalities from one tissue to another. The analysis of chromatin architectures at HoxD and HoxA loci revealed the existence of two flanking topologically associating domains (TADs), precisely encompassing the adjacent regulatory landscapes. Here, we discuss the function of these TADs in the control of Hox gene regulation and we speculate about their capacity to serve as structural frameworks for the emergence of novel enhancers. In this view, TADs may have been used as genomic niches to evolve pleiotropic regulations found at many developmental loci. Ó 2015 Published by Elsevier B.V. on behalf of the Federation of European Biochemical Societies.

1. Introduction During vertebrate embryonic development, Hox genes are expressed in a collinear manner leading to the sequential patterning of the main body axis in time and space [1,2]. In addition to the implementation of this ancestral function, vertebrate Hox genes are also involved in the development of secondary axial structures such as the limbs or the external genitalia, or of different organs like the intestinal cecum or hair follicles [3–8]. The emergence of these various vertebrate evolutionary novelties is thought to derive from mechanisms of genes neo-functionalization, a process that was facilitated by the two rounds of genome duplication that occurred at the root of the vertebrate radiation [9–11]. In this context, both HoxD and HoxA cluster genes have been heavily studied

⇑ Corresponding author at: School of Life Sciences, Ecole Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland. E-mail address: denis.duboule@epfl.ch (D. Duboule). 1 Present address: Department of Genetics and Department of Ophthalmology, Harvard Medical School, Boston, MA 02115, USA.

and can be considered as a paradigm to witness the co-option and evolution of complex regulations [12–16]. The functional recruitment of Hox genes in novel body structures, distinct from the main anterior to posterior axis, occurred preferentially via multiple enhancer sequences located outside the gene clusters [17,18], as demonstrated thus far for HoxA and HoxD. In both cases, such long-acting enhancer sequences are found within ca. one megabase (Mb) from the clusters, either within flanking gene deserts (HoxD; [14,19–22]) or interspersed into neighboring genes (HoxA; [12,15,23]). These large DNA intervals thus generally contain several enhancers, often acting not only over their specific target genes, but also over unrelated genes residing at proximity (by-stander effects) and are referred to as ‘regulatory landscapes’ [18]. Such developmental enhancers in series can either display a similar specificity (e.g. [22,24]) or complement each other to elaborate a final transcription pattern [21] and hence regulatory landscapes can modulate target transcription via enhancer integration, both qualitatively and quantitatively (see [25]). The implementation of chromosome conformation capture (3C) approaches [26] (see also [27,28]) to embryonic limb and brain

http://dx.doi.org/10.1016/j.febslet.2015.04.024 0014-5793/Ó 2015 Published by Elsevier B.V. on behalf of the Federation of European Biochemical Societies.

Please cite this article in press as: Lonfat, N. and Duboule, D. Structure, function and evolution of topologically associating domains (TADs) at HOX loci. FEBS Lett. (2015), http://dx.doi.org/10.1016/j.febslet.2015.04.024

2

N. Lonfat, D. Duboule / FEBS Letters xxx (2015) xxx–xxx

tissues [21] has revealed that some regulatory sequences upstream the HoxD cluster interact with their target promoters even in the absence of transcriptional activity, suggesting that the adjacent one Mb large gene desert was a poised regulatory structure, also present in negative cellular contexts. Upon activation and transcription of target genes, however, additional contacts were scored within this pre-formed architecture [21]. Such structural units, referred to as topologically associating domains (TADs) were reported to exist genome wide [29] and to represent an intermediate level of chromatin organization [30–33], a series of contiguous domains wherein enhancers-promoters and/or enhancer-enhancer interactions are privileged (see Refs. in [34–36]). Such an organization of the genome in TADs is both conserved in mouse and human and globally consistent across different cell types [30,31,37–40]. This complex three-dimensional architecture of chromatin is key for proper transcriptional control (see [41]) and a number of recent studies have suggested that TADs have a supporting role in longrange gene regulation [42–45]. Below, we discuss the functional importance of TADs in the complex regulation of Hox genes in embryo, during the development of secondary body axes and other vertebrate-specific organs. We suggest that these chromatin structures may have been instrumental in the emergence of novel enhancers by providing opportunities for factors to hi-jack a powerful and rather generic transcriptional read-out. In turn, such regulatory co-optations may have solidified and re-enforced the spatial architecture associated with these regulatory landscapes. 2. Bimodal regulation in developing limbs The early development of vertebrate limbs is accompanied by two successive waves of Hox gene transcription [3,46] (Refs. in [47]), from both the HoxA and HoxD clusters. These two gene clusters indeed cooperate to build the limbs, as shown by their combined functional inactivation, which lead to severe limb truncation [48]. In the case of the HoxD cluster, an initial phase of activation (phase I) involves the sequential transcription, in both time and space, of anterior and central Hoxd genes (from Hoxd1 to Hoxd11) in the growing limb bud. This sub-group of genes is necessary for the development of the proximal part of the limb, including the future arm and forearm [49,50]. Soon after, a second yet partially overlapping sub-group of Hoxd genes (from Hoxd8 to Hoxd13), located toward the other (posterior) half of the cluster become transcribed in the distal domain of the limb bud (phase II) to drive the emergence of the most distal limb segment, which will eventually give rise to digits [13] (Refs. in [51]). These two successive expression phases, during which different groups of neighboring genes are transcribed in concert, are controlled by distinct regulatory landscapes located on either sides of the HoxD gene cluster and corresponding to the two large gene deserts that flank this locus [19]. Initially, series of enhancers located within the telomeric gene desert (on the Hoxd1 side) are activated and control phase I expression, as demonstrated both by 4C-seq experiments and by genetic approaches involving targeted inversions and deletions of this regulatory landscape [20]. Hoxd9, for example, is robustly transcribed during this early phase of expression. When this gene was used as a bait (a viewpoint) in a 4C approach using limb bud cells dissected out of E9.5 embryos, i.e. at a stage when the second phase has not yet initiated, very significant interactions were scored telomeric to the HoxD cluster (Fig. 1A and B). These contacts occurred within regions particularly enriched for H3K27ac marks [20], which are generally associated with active genes and enhancers [52,53]. Likewise, the 4C interaction profile established by Hoxd11, another gene transcribed during this early phase, also involved contacts with the telomeric landscape, extending over the same genomic distance than for Hoxd9.

In both cases, the 4C interactions domains had similar extents and precisely matched the position of a TAD, as defined by using Hi-C in ES cells [30], i.e. a cell type where these Hoxd genes are either silent or transcribed at very low level. The second wave of Hoxd gene transcription in limb buds initiates within a small group of posterior-distal cells, likely in response to signals emanating from the Apical Ectodermal Ridge (AER) in E10 limb bud. These cells strongly transcribe the Hoxd13 gene as well as Hoxd12 to Hoxd8, with a progressively weaker activity. This linear decrease in transcriptional efficiency, initially referred to as ‘quantitative collinearity’ [4] reflects the strong affinity of Hoxd13 for the relevant enhancers [13] and explains why this latter gene is transcribed slightly more widely than other Hoxd genes in presumptive digits, a phenomenon sometimes qualified as ‘reverse collinearity’ [46]. Similar to the early phase I, this second wave is controlled by multiple enhancer sequences, but in this case located centromeric to the HoxD cluster (on the Hoxd13 side) [14,16,18,21] (Fig. 1B). The interaction profile of Hoxd13 in these distal cells indeed has revealed at least seven blocks (islands) of regulatory sequences spanning this gene-poor DNA interval. Transgenic and genetic approaches have shown that these sequences cooperate to elicit the final transcriptional outcome in presumptive digits cells (see [54]). As for the telomeric landscape, this ‘regulatory archipelago’, defined by the 4C approach, precisely matches the extent of another TAD [30], located immediately adjacent to the former (Fig. 1A and B). As a consequence of this bimodal regulatory strategy involving global enhancers flanking both sides of the target locus, the HoxD cluster is positioned at the border between two distinct TADs (Fig. 1C), with different genes interacting toward different directions depending on their respective positions within the gene cluster. A related situation seems to occur at the HoxA cluster [12,15,30]. 3. Specific and semi-constitutive interactions During phase I, in the corresponding dissected proximal limb cells, the telomeric TAD is fully active whereas the centromeric TAD is inactive, as shown by the analysis of histone marks [20]. However, despite the negative ‘regulatory status’ of the centromeric TAD, robust constitutive contacts are observed between Hoxd13, Hoxd12 and several islands of this TAD, indicating that a poised architecture sequesters both genes thus preventing them to interact with the active telomeric TAD (Fig. 1D) [20,55]. This sequestration mechanism is essential to prevent the transcriptional activation of Hoxd13 at times and in places where it is not desirable, i.e. under the control of the telomeric TAD, due to its dominant negative effect over other Hox genes [56]. An example of these deleterious effects was provided by mice carrying the Ulnaless (Ul) mutation. In this mutant stock, Hoxd13 is ectopically expressed during phase I of limb development due to an inversion of the HoxD locus [18], leading to a severe shortening of the proximal limb [57,58]. Subsequently, in response to some signals, which may include Hox13 proteins themselves starting to accumulate in distal cells as a result of phase II initiation [59], the telomeric TAD becomes inactive in these most distal cells, loosing its H3K27Ac marks and accumulating H3K27me3 marks [20], i.e. a read-out of the polycomb repressive system. Here again, however, Hoxd genes located near the telomeric desert such as Hoxd1 or Hoxd4 will maintain strong constitutive interactions with the telomeric TAD, even after telomeric enhancers will have terminate operating, for example during digit development. At about the same time when the telomeric TAD finishes to control transcription, the centromeric TAD becomes fully active in the emerging presumptive digit cells. This transition can be visualized

Please cite this article in press as: Lonfat, N. and Duboule, D. Structure, function and evolution of topologically associating domains (TADs) at HOX loci. FEBS Lett. (2015), http://dx.doi.org/10.1016/j.febslet.2015.04.024

N. Lonfat, D. Duboule / FEBS Letters xxx (2015) xxx–xxx

3

Fig. 1. A regulatory switch between two adjacent TADs underlies the bimodal regulation occurring at the HoxD locus during limb development. (A) Schematic distribution of TADs over 2.8 Mb on mouse chromosome 2, including the HoxD locus [data from [30]]. The color code (yellow for the telomeric TAD and blue for the centromeric TAD) is the same for all panels and corresponds to the early (yellow) and late (blue) phases of limb development. (B) 4C smoothed interactions profiles established by Hoxd13 (orange bar) in E12.5 digits (top profile) and by Hoxd9 (orange vertical bar) in E9.5 forelimb buds (bottom profile) [data from [14] (top) and [20] (bottom)]. The majority of contacts established by Hoxd13 in digits outside the cluster itself are found within 1 Mb centromeric to the cluster, corresponding to a gene desert and matching the centromeric TAD (blue). In forelimb buds, the interaction profile is opposite, with most contacts observed telomeric to the cluster within the adjacent telomeric TAD (yellow). In both cases, some weaker contacts are also observed outside, restricted to next adjacent TADs (gray shaded area). (C) The HoxD cluster (purple) is localized right at the border between the two TADs. (D) During the early phase of limb development, which corresponds to the outgrowth of the future arm and forearm, a telomeric part of the HoxD cluster (from Hoxd1 to Hoxd11) interacts with the telomeric TAD (yellow shaded) where it contacts enhancers sequences specific for this early phase of limb outgrowth (gray and yellow are semi-constitutive and specific factors, respectively). (E) Subsequently, during the late phase of limb development, which will give rise to the distal part of the limbs (the digits), the centromeric TAD will be activated (blue factors) and will interact with Hoxd13 and Hoxd12. Concomitantly, more centrally located Hox genes (from Hoxd11 to Hoxd8) will switch their interactions from the telomeric to the centromeric TAD. These genes will all fall under the regulation of centromeric enhancers and participate to the making of hands and feet. The schematics of limb buds in the right hand side depict the domains where the corresponding TADs are active. Adapted from [20,84].

Please cite this article in press as: Lonfat, N. and Duboule, D. Structure, function and evolution of topologically associating domains (TADs) at HOX loci. FEBS Lett. (2015), http://dx.doi.org/10.1016/j.febslet.2015.04.024

4

N. Lonfat, D. Duboule / FEBS Letters xxx (2015) xxx–xxx

both by the appearance of active epigenetics marks and through a slightly different contacts profile within the TAD. When using Hoxd13 as a viewpoint in 4C-seq for example, a robust contact with island III [21] is established, an interaction specifically and systematically observed upon active transcription of the target Hoxd genes in vivo, while virtually absent from all other biological contexts assayed thus far. This specific interaction seems to be sensitive to an in vivo physiological context since it is absent either from brain or ES cells (e.g. [21,30]), or from cultured immortalized limb bud cells [60]. This illustrates the necessity to study these interactions in the proper developing embryo rather than in less physiological systems. In the latter case, however, the constitutive interactions associated with the presence of the centromeric TAD were faithfully revealed by using a 5C approach [60]. A similar observation applies to island II-1, where contacts are significantly enriched in digit cells where Hoxd genes are fully active, when compared to the genital bud (see Fig. 2; [14]). Such an internal reorganization in the interaction patterns within the same TAD upon target gene activation, which was assimilated to a kind of chromatin ‘allosteric transition’ (see below and [20]), was also observed at the HoxA locus [12]. Of note, interactions appearing specifically when the centromeric TAD becomes transcriptionally active do not necessarily correspond to strong autonomous enhancer sequences. While the island II-1 sequence could drive minimal promoter expression right into the Hoxd digit expression territory, thus demonstrating its genuine enhancer potential [14,21], the digit-specific island III was unable to elicit a similar regulatory activity in the same conditions [21]. These observations suggest that the activation of a TAD into a ‘transcription-positive’ mode, via exposure to tissue-specific factors, may impact its internal configuration, leading to the detection of novel structural interactions involving target genes, yet not necessarily associated with an enhancer sequence. We refer to such specific interactions lacking an autonomous regulatory potential as ‘semi-constitutive’. 4. A regulatory switch from TAD to TAD As a result of the regulatory switch from an active telomeric TAD/inactive centromeric TAD to an active centromeric TAD/inactive telomeric TAD, some of the genes located at the border between both domains (mostly Hoxd11, Hoxd10 and Hoxd9) will change their interaction tropism along with time. Initially, they participate to the morphogenesis of the proximal limb by responding to the telomeric regulation. Therefore, they closely interact with the telomeric TAD. Subsequently however, this regulation will terminate, likely leading to a weakening of these interactions, making such genes located centrally in the cluster less tightly involved into this TAD. At the same time, in most distal cells, the centromeric TAD will start operating on Hoxd13 and Hoxd12, which are naturally interacting with this TAD, in all conditions (see above). The strengthening of these interactions, however, may drag along the centrally located genes, which have now weakened their contacts with the telomeric TAD (Fig. 1D and E). This topological switch of a few Hoxd genes from one regulatory landscape to the other exactly corresponds to the developmental transition between the proximal (arm and forearm) and the distal (digits) parts of the limb bud (Fig. 1). Because the second wave of transcription likely responds to signals emanating from the AER, which delivers growth signals to the underlying distal cells of the early limb bud [61,62], the early and late Hoxd expression territories will segregate and not abut to one another. As a consequence, a cellular zone where Hoxd genes are not transcribed will appear (see [20,51,63] and schematics in Fig. 1D and E). This particular cellular territory will generate the mesopodium (the wrist and ankle), as shown by analyzing mutant genetic conditions in mice and human [64,65].

This remarkable bimodal and sequential regulatory strategy not only gives a mechanistic explanation to the collinear distribution of Hoxd expression domains during limb development but also supports the idea of a clear evolutionary distinction in both the times and mechanisms of occurrence, between our ‘proximal’ limbs and the most distal part of our appendages, the digits. The existence of two adjacent TADs flanking the HoxD cluster was nevertheless also observed in teleost fishes [15] suggesting that such a regulatory organization during limb development evolved in tetrapods by the co-option of a pre-existing TAD along with the emergence of digits. The regulatory function (if any) of this ‘ancestral’ TAD is elusive but may relate to the implementation of the collinear mechanism at work during the elongation of the major body axis, a process that naturally predated the emergence of limbs. 5. Tissue-specific structural transitions within TADs Recent studies of both the HoxA and HoxD clusters have revealed that other organs or tissues controlled by these genes make use of the same regulatory architectures to implement long-range transcriptional controls (Fig. 2A and B). For instance, both posterior Hoxa and Hoxd genes are strongly transcribed during the development of the external genital organs. Like limbs, mammalian genitalia develop following a distal outgrowth. The incipient genital tubercle (GT), which will give rise to the future penis or clitoris [66–69] follows morphogenetic mechanisms globally similar to those operating during digits formation involving the same signaling pathways and major developmental regulators (see [62,66,70]), thus suggesting that similar genetic programs are implemented [71]. Accordingly, the same Hox genes are patterning these embryonic structures (Fig. 2C) [4,13,72] and mice inactivated for both Hoxa13 and Hoxd13 functions are lacking digits and external genitals [73–75]. Genetic and biochemical analyses have shown that for both HoxA and HoxD clusters, the long range regulations associated with GT development rely upon the same regulatory landscapes than those controlling expression in digits and are consequently associated with the same TADs (Fig. 2A and B) [14]. However, when carefully examined, the 4C interaction profiles established either by Hoxa13 or by Hoxd13 in GT and digits cells, though very similar, still displayed some significant differences. When using GT cells, Hoxd13 contacted at least two DNA regions, always within the same TAD, which were not as clearly detected when using digit cells. In addition, these regions had strong enhancer activity in the developing genitalia. Conversely, an interaction peak present when using digits was not scored with the GT material (Fig. 2B). Similar tissue-specific differences in intra-TAD interactions were also observed at the HoxA locus [14]. Therefore, it appears that within the same general TAD structure, slightly different regulatory architectures are formed either in GT or in digit cells, i.e. in two cases where a strong positive transcriptional outcome is produced. This transcriptional response is virtually identical in both conditions [13] and hence these minor structural differences may not impact upon the readout of the system. Instead, they simply illustrate various possibilities for transcription factors to interact within such pre-existing chromatin structures to elicit comparable responses, yet in different contexts. Two kinds of structural differences can thus be observed within these TADs during Hox gene regulation in vivo, at both HoxA and HoxD loci. First, a transition occurs from a transcriptionally silent structure to an active configuration. This transition involves a clear internal re-organization in contacts, even though the global extent of the TAD remains the same [12,21]. In addition, more subtle structural differences exist between the same TAD when active in distinct tissues, such as in GT and digits cells. In this case, most contact regions are shared, indicating a comparable backbone, with

Please cite this article in press as: Lonfat, N. and Duboule, D. Structure, function and evolution of topologically associating domains (TADs) at HOX loci. FEBS Lett. (2015), http://dx.doi.org/10.1016/j.febslet.2015.04.024

N. Lonfat, D. Duboule / FEBS Letters xxx (2015) xxx–xxx

5

Fig. 2. TADs as genomic niches to evolve long-range pleiotropic regulations. (A) Schematized distribution of TADs over ca. 2.8 Mb including the HoxD locus (in purple) [data extracted from [30]]. (B) 4C smoothed interaction profiles established by Hoxd13 (orange bar) either in E15.5 genital tubercle (GT) cells (top), or in E12.5 digit cells (bottom) [data from [14]]. In both contexts, the majority of interactions occurs within one Mb, centromeric to the cluster, corresponding to a gene desert and matching the centromeric TAD (blue shaded area). Weaker interactions are scored at more centromeric positions (gray shaded area) and also within the telomeric TAD, which contains the forelimb regulatory modalities (yellow). Although many interaction peaks are shared between the two tissues, some of them show a clear preference for one given tissue and are found either in GT or in digits (arrows). (C) Hoxd13 is transcribed in both digits and GT [4], adapted from [14]. (D) The centromeric TAD (blue) can thus be found in two similar yet not identical configurations, leading to the same high transcriptional activity in both digits (bottom left) and GT (bottom right). The slight differences in conformations are likely induced by various tissue-specific factors (yellow or blue), whereas other factors may be shared between the two tissues (yellow and blue) or be of semi-constitutive nature (gray). The existence of a TAD as a preformed, poised regulatory architecture may have served as a backbone to evolve long-range pleiotropic regulations, offering niches to potential enhancers and an appropriate transcriptional read-out. Adapted from [14].

only few tissue-specific interactions (Fig. 2D). These observations suggest that most of the TAD architecture is conserved in various regulatory active situations, though with punctual differences likely induced by specific transcription factors [14]. Therefore, at least at Hox loci, strong constitutive regulatory structures exist, which can be slightly – yet specifically – modified in various ontogenic contexts to elicit a common transcriptional response (Fig. 2D). Such a transition from an inactive TAD toward several possible active conformations suggests that a fixed number of distinct chromatin states can be formed within the structural constraints imposed by the intrinsic TAD architecture, a process somewhat similar to allosteric transitions [76], yet applied to large chromatin structures [14,77]. In this view, TADs can be seen as permissive regulatory systems, poised to respond to external signals. Accordingly, tissue-specific transcription factors may have little

instructive role in the building of this structure [42]. Whether or not such regulatory architectures are fluctuating in their internal organization, with a high dynamics in contacts (see [78]) is difficult to evaluate. In any case, the 4C profiles reveal that some configurations are largely favored over others. The direct visualization of TADs in embryonic cells will be informative in this respect. 6. Integration and evolution of multiple enhancers in cis Interestingly, TADs often contain range of enhancer sequences, which can have either the same, related or distinct functional specificities (see e.g. [44,79]) (Fig. 3A and B). Here again, longrange regulation at Hox clusters can be used as a paradigm to understand both the functional reason underlying this organization and its evolutionary origin. For example, the digit regulation

Please cite this article in press as: Lonfat, N. and Duboule, D. Structure, function and evolution of topologically associating domains (TADs) at HOX loci. FEBS Lett. (2015), http://dx.doi.org/10.1016/j.febslet.2015.04.024

6

N. Lonfat, D. Duboule / FEBS Letters xxx (2015) xxx–xxx

at both HoxA and HoxD clusters is controlled by series of enhancers displaying complementary specificities [12,21] (Fig. 3B). The final transcription pattern is delivered via the combined action of all enhancers present in the same TAD and partial deletions within this TAD generates both qualitative and quantitative phenotypes [21]. Conversely, Hoxd transcription during the budding of the intestinal cecum relies upon a collection of enhancers displaying comparable expression specificity [22], similar to the case of the Fgf8 gene regulation in the limb bud ectoderm [24] (Fig. 3A). This series of cecum enhancers is also found restricted within a single TAD, telomeric to HoxD, which also contains the proximal limb regulatory sequences. Thus, TADs can include enhancers of unrelated functionalities, such as those for the cecum and the proximal limbs (HoxD telomeric TAD) or for the digits and external genitalia (HoxD centromeric TAD). These different cases illustrate how complex regulatory systems can use their modularity to generate both robustness and flexibility [21,44]. They also document the integrative function of TADs (see [24] and Refs. in [25]). The existence of constitutive regulatory structures [21,29–31] provides a framework to understand the evolutionary emergence of multiple enhancers in cis, displaying similar or comparable specificities. It was indeed proposed that the presence of a strong enhancer sequence may have favored the evolution of similar regulatory sequences at its vicinity (‘regulatory priming’ [16,80]). Within a constrained chromatin backbone, a particular factor targeted to a specific enhancer may have a higher probability to

encounter a neighboring DNA sequence and start establishing a productive interaction, for example by tightening a ‘positive’ architecture and thus slightly modifying the final transcription read-out in an adaptive manner. In such a case, an optimization of this protein-DNA interaction could be selected leading to a new enhancer sequence [22] (Fig. 3A). Accordingly, an evolutionary virtuous circle may start, whereby the presence of multiple enhancers targeted by the same factors within a given TAD would increase factor promiscuity and hence the potential addition of yet additional regulatory sequences. 7. The evolution of developmental pleiotropic regulations Another major potential impact of TADs upon the evolution of regulations concerns genetic loci displaying high pleiotropic functions during development. In the case of Hox genes, pleiotropy can be considered at two levels: first, individual genes are transcribed at multiple sites during development, by using distinct regulatory sequences located in a close cis-proximity as revealed by classical transgenic approaches (e.g. [81]). Secondly, Hox gene clusters as entities or ‘meta-genes’ [82] also display pleiotropic regulations, leading to global cluster-specific expression controls (e.g. [83]). In the latter case, it was proposed that these long-range regulations targeting several Hox genes at once could have evolved following the two genome duplication events, which occurred at the root of the vertebrate phylum, much like single gene neo-functionalization

Fig. 3. TADs as genomic niches to evolve long-range enhancer sequences. A schematic view of how TADs can help evolve multiple long-range enhancers acting on two target genes (blue). TADs at Hox loci contain multiple enhancer sequences, which can display either comparable specificities (A), distinct specificities in the same tissue (B) or different tissue-specificity (pleiotropic effect, see Fig. 2). (A) Because of the general structural promiscuity induced by a TAD, a productive interaction with a moderate transcriptional outcome (left) may increase the probability to evolve a closely located low-affinity binding sequence (middle, blue arrow) into a strong binding site (right), due to the concomitant increase of the transcription read-out, which may have an adaptive value. In such a case, a quantitative effect is expected within the same cellular landscape (stronger blue in the limb buds). (B) Conversely, an enhancer sequence can be recruited only in part of a tissue, due to the lack of one particular factor, as exemplified here with the most posterior part of the digital plate, which does not express the two target genes (top). Because the TAD provides a poised regulatory configuration, another factor may compensate for this situation by interacting via different modalities and eliciting the same transcriptional read out in the posterior part of the digital plate only (bottom). In this latter example, various enhancers are required to bring about the final transcription pattern. While the two transcription patterns derived from A and B are similar, they are implemented by various regulatory strategies. As a consequence, targeted mutations of specific sequences within this TAD would lead to various phenotypic effects [21]. Adapted from [21,22].

Please cite this article in press as: Lonfat, N. and Duboule, D. Structure, function and evolution of topologically associating domains (TADs) at HOX loci. FEBS Lett. (2015), http://dx.doi.org/10.1016/j.febslet.2015.04.024

N. Lonfat, D. Duboule / FEBS Letters xxx (2015) xxx–xxx

processes [9,10], yet applied to gene clusters [80]. In this view, rather than evolving another gene-specific regulatory sequence at the vicinity of the target locus, global enhancers evolved to control sub-groups of neighboring genes in a coordinated manner. The association between the presence of a TAD on the one hand, and multiple enhancers of various tissue specificities, on the other hand, suggests a relation of causality. However, the nature of this causality is unclear. For instance it is possible that the presence of several global enhancers dispatched over one Mb lead to the formation of a TAD. Alternatively, an initial TAD may have triggered the emergence of various enhancer sequences of distinct specificities. We favor the second scenario whereby an initial TAD was in place before global regulations started to accumulate at Hox loci. This alternative however does not rule out the possibility that the starting point involved a pioneer productive enhancer-promoter interaction, instead of a mere constitutive architecture unrelated to transcriptional activity. In the latter scenario, TADs might be seen as evolutionary playgrounds for the co-option of novel regulations, due to the pre-existence of a poised chromatin structure associated with potent target genes and ready to elicit a generic transcriptional response. Accordingly, even a weak interaction between a transcription factor and a target sequence within a TAD could have been productive, leading to a situation where selection may start operate. Much like the addition of multiple enhancers with comparable specificities (see above), the evolutionary addition of distinct specific regulations may have re-enforced the organization of interactions within the TAD to further maximize its read-out. TADs could thus be seen as genomic niches where tissue-specific factors may hi-jack a pre-existing transcriptional read-out, providing a plausible mechanism to the emergence of long-range pleiotropic regulations in vertebrates [14]. This hypothesis implies that genetic loci coding for highly pleiotropic developmental factors (for example transcription factors or signaling molecules) should be embedded into tightly defined TADs, whereas such a chromatin organization may not be so easily identifiable around genetic loci with higher cellular specialization or for housekeeping genes [42]. In the case of the HoxD cluster, the presence of the two flanking TADs was also observed in teleost fishes, with a partition in the interactions established by Hoxd genes similar to those observed in mammals [15]. Even though these fishes do not represent an ancestral form of amniotes, they lack any obvious trace of digits and external genitalia. Therefore, the limb and GT enhancers described in tetrapods were likely not required for (or did not coevolved with) the existence of this particular TAD, suggesting that a pre-existing architecture was co-opted. This raises the question of the evolutionary origin of such structures and future analyses of early chordata genomes may be informative in this respect, as discussed above. The fact that both HoxA and HoxD clusters display the same general organization of their TADs [12,20] indeed suggests that this particular chromatin topology predated the genome duplications accompanying the emergence of vertebrates [14]. Acknowledgments Our laboratory acknowledges the generous funding from the Federal Institute of Technology (EPFL), the University of Geneva, the Swiss National Research Foundation (No. 310030B_138662) and the European Research Council (ERC) Grant ‘SystemsHox.ch’ (No. 232790). References [1] Kmita, M. and Duboule, D. (2003) Organizing axes in time and space; 25 years of colinear tinkering. Science 301, 331–333. [2] Krumlauf, R. (1994) Hox genes in vertebrate development. Cell 78, 191–201.

7

[3] Dolle, P., Izpisua-Belmonte, J.C., Falkenstein, H., Renucci, A. and Duboule, D. (1989) Coordinate expression of the murine Hox-5 complex homoeoboxcontaining genes during limb pattern formation. Nature 342, 767–772. [4] Dolle, P., Izpisua-Belmonte, J.C., Brown, J.M., Tickle, C. and Duboule, D. (1991) HOX-4 genes and the morphogenesis of mammalian genitalia. Genes Dev. 5, 1767. [5] Haack, H. and Gruss, P. (1993) The establishment of murine Hox-1 expression domains during patterning of the limb. Dev. Biol. 157, 410–422. [6] Zacchetti, G., Duboule, D. and Zakany, J. (2007) Hox gene function in vertebrate gut morphogenesis: the case of the caecum. Development 134, 3967–3973. [7] Godwin, A.R. and Capecchi, M.R. (1998) Hoxc13 mutant mice lack external hair. Genes Dev. 12, 11–20. [8] Yokouchi, Y., Sakiyama, J. and Kuroiwa, A. (1995) Coordinated expression of Abd-B subfamily genes of the HoxA cluster in the developing digestive tract of chick embryo. Dev. Biol. 169, 76–89. [9] Holland, P.W., Garcia-Fernandez, J., Williams, N.A. and Sidow, A. (1994) Gene duplications and the origins of vertebrate development. Dev. Suppl., 125–133. [10] Ohno, S. (1970) Evolution by Gene Duplication, Springer-Verlag, Heidelberg. [11] Wagner, G.P., Amemiya, C. and Ruddle, F. (2003) Hox cluster duplications and the opportunity for evolutionary novelties. Proc. Natl. Acad. Sci. USA 100, 14603–14606. [12] Berlivet, S., Paquette, D., Dumouchel, A., Langlais, D., Dostie, J. and Kmita, M. (2013) Clustering of tissue-specific sub-TADs accompanies the regulation of HoxA genes in developing limbs. PLoS Genet. 9, e1004018. [13] Montavon, T., Le Garrec, J.-F., Kerszberg, M. and Duboule, D. (2008) Modeling Hox gene regulation in digits: reverse collinearity and the molecular origin of thumbness. Genes Dev. 22, 346–359. [14] Lonfat, N., Montavon, T., Darbellay, F., Gitto, S. and Duboule, D. (2014) Convergent evolution of complex regulatory landscapes and pleiotropy at Hox loci. Science 346, 1004–1006. [15] Woltering, J.M., Noordermeer, D., Leleu, M. and Duboule, D. (2014) Conservation and divergence of regulatory strategies at hox Loci and the origin of tetrapod digits. PLoS Biol. 12, e1001773. [16] Gonzalez, F., Duboule, D. and Spitz, F. (2007) Transgenic analysis of Hoxd gene regulation during digit development. Dev. Biol. 306, 847–859. [17] Spitz, F., Gonzalez, F., Peichel, C., Vogt, T.F., Duboule, D. and Zákány, J. (2001) Large scale transgenic and cluster deletion analysis of the HoxD complex separate an ancestral regulatory module from evolutionary innovations. Genes Dev. 15, 2209–2214. [18] Spitz, F., Gonzalez, F. and Duboule, D. (2003) A global control region defines a chromosomal regulatory landscape containing the HoxD cluster. Cell 113, 405–417. [19] Lee, A.P., Koh, E.G., Tay, A., Brenner, S. and Venkatesh, B. (2006) Highly conserved syntenic blocks at the vertebrate Hox loci and conserved regulatory elements within and outside Hox gene clusters. Proc. Natl. Acad. Sci. USA 103, 6994–6999. [20] Andrey, G., Montavon, T., Mascrez, B., Gonzalez, F., Noordermeer, D., Leleu, M., Trono, D., Spitz, F. and Duboule, D. (2013) A switch between topological domains underlies HoxD genes collinearity in mouse limbs. Science 340, 1234167. [21] Montavon, T., Soshnikova, N., Mascrez, B., Joye, E., Thevenet, L., Splinter, E., de Laat, W., Spitz, F. and Duboule, D. (2011) A regulatory archipelago controls Hox genes transcription in digits. Cell 147, 1132–1145. [22] Delpretti, S., Montavon, T., Leleu, M., Joye, E., Tzika, A., Milinkovitch, M. and Duboule, D. (2013) Multiple enhancers regulate Hoxd genes and the Hotdog LncRNA during cecum budding. Cell Rep. 5, 137–150. [23] Lehoczky, J.A. and Innis, J.W. (2008) BAC transgenic analysis reveals enhancers sufficient for Hoxa13 and neighborhood gene expression in mouse embryonic distal limbs and genital bud. Evol. Dev. 10, 421–432. [24] Marinic, M., Aktas, T., Ruf, S. and Spitz, F. (2013) An integrated holo-enhancer unit defines tissue and gene specificity of the Fgf8 regulatory landscape. Dev. Cell 24, 530–542. [25] Schwarzer, W. and Spitz, F. (2014) The architecture of gene expression: integrating dispersed cis-regulatory modules into coherent regulatory domains. Curr. Opin. Genet. Dev. 27, 74–82. [26] Dekker, J., Rippe, K., Dekker, M. and Kleckner, N. (2002) Capturing chromosome conformation. Science 295, 1306–1311. [27] de Laat, W. and Dekker, J. (2012) 3C-based technologies to study the shape of the genome. Methods 58, 189–191. [28] de Wit, E. and de Laat, W. (2012) A decade of 3C technologies: insights into nuclear organization. Genes Dev. 26, 11–24. [29] Jin, F., Li, Y., Dixon, J.R., Selvaraj, S., Ye, Z., Lee, A.Y., Yen, C.A., Schmitt, A.D., Espinoza, C.A. and Ren, B. (2013) A high-resolution map of the threedimensional chromatin interactome in human cells. Nature 503, 290–294. [30] Dixon, J.R., Selvaraj, S., Yue, F., Kim, A., Li, Y., Shen, Y., Hu, M., Liu, J.S. and Ren, B. (2012) Topological domains in mammalian genomes identified by analysis of chromatin interactions. Nature 485, 376–380. [31] Nora, E.P., Lajoie, B.R., Schulz, E.G., Giorgetti, L., Okamoto, I., Servant, N., Piolot, T., van Berkum, N.L., Meisig, J., Sedat, J., Gribnau, J., Barillot, E., Bluthgen, N., Dekker, J. and Heard, E. (2012) Spatial partitioning of the regulatory landscape of the X-inactivation centre. Nature 485, 381–385. [32] Sexton, T., Yaffe, E., Kenigsberg, E., Bantignies, F., Leblanc, B., Hoichman, M., Parrinello, H., Tanay, A. and Cavalli, G. (2012) Three-dimensional folding and functional organization principles of the Drosophila genome. Cell 148, 458– 472.

Please cite this article in press as: Lonfat, N. and Duboule, D. Structure, function and evolution of topologically associating domains (TADs) at HOX loci. FEBS Lett. (2015), http://dx.doi.org/10.1016/j.febslet.2015.04.024

8

N. Lonfat, D. Duboule / FEBS Letters xxx (2015) xxx–xxx

[33] Hou, C., Li, L., Qin, Z.S. and Corces, V.G. (2012) Gene density, transcription, and insulators contribute to the partition of the Drosophila genome into physical domains. Mol. Cell 48, 471–484. [34] Nora, E.P., Dekker, J. and Heard, E. (2013) Segmental folding of chromosomes: a basis for structural and regulatory chromosomal neighborhoods? BioEssays 35, 818–828. [35] Gorkin, D.U., Leung, D. and Ren, B. (2014) The 3D genome in transcriptional regulation and pluripotency. Cell Stem Cell 14, 762–775. [36] Sexton, T. and Cavalli, G. (2015) The role of chromosome domains in shaping the functional genome. Cell 160, 1049–1059. [37] Naumova, N., Smith, E.M., Zhan, Y. and Dekker, J. (2012) Analysis of long-range chromatin interactions using Chromosome Conformation Capture. Methods 58, 192–203. [38] Zuin, J., Dixon, J.R., van der Reijden, M.I., Ye, Z., Kolovos, P., Brouwer, R.W., van de Corput, M.P., van de Werken, H.J., Knoch, T.A., van IJcken, W.F., Grosveld, F.G., Ren, B. and Wendt, K.S. (2014) Cohesin and CTCF differentially affect chromatin architecture and gene expression in human cells. Proc. Natl. Acad. Sci. USA 111, 996–1001. [39] Sofueva, S., Yaffe, E., Chan, W.C., Georgopoulou, D., Vietri Rudan, M., MiraBontenbal, H., Pollard, S.M., Schroth, G.P., Tanay, A. and Hadjur, S. (2013) Cohesin-mediated interactions organize chromosomal domain architecture. EMBO J. 32, 3119–3129. [40] Dixon, J.R., Jung, I., Selvaraj, S., Shen, Y., Antosiewicz-Bourget, J.E., Lee, A.Y., Ye, Z., Kim, A., Rajagopal, N., Xie, W., Diao, Y., Liang, J., Zhao, H., Lobanenkov, V.V., Ecker, J.R., Thomson, J.A. and Ren, B. (2015) Chromatin architecture reorganization during stem cell differentiation. Nature 518, 331–336. [41] Bickmore, W.A. and van Steensel, B. (2013) Genome architecture: domain organization of interphase chromosomes. Cell 152, 1270–1284. [42] de Laat, W. and Duboule, D. (2013) Topology of mammalian developmental enhancers and their regulatory landscapes. Nature 502, 499–506. [43] Shen, Y., Yue, F., McCleary, D.F., Ye, Z., Edsall, L., Kuan, S., Wagner, U., Dixon, J., Lee, L., Lobanenkov, V.V. and Ren, B. (2012) A map of the cis-regulatory sequences in the mouse genome. Nature 488, 116–120. [44] Symmons, O., Uslu, V.V., Tsujimura, T., Ruf, S., Nassari, S., Schwarzer, W., Ettwiller, L. and Spitz, F. (2014) Functional and topological characteristics of mammalian regulatory domains. Genome Res. 24, 390–400. [45] Smallwood, A. and Ren, B. (2013) Genome organization and long-range regulation of gene expression by enhancers. Curr. Opin. Cell Biol. 25, 387–394. [46] Nelson, C.E., Morgan, B.A., Burke, A.C., Laufer, E., DiMambro, E., Murtaugh, L.C., Gonzales, E., Tessarollo, L., Parada, L.F. and Tabin, C. (1996) Analysis of Hox gene expression in the chick limb bud. Development 122, 1449–1466. [47] Zakany, J. and Duboule, D. (2007) The role of Hox genes during vertebrate limb development. Curr. Opin. Genet. Dev. 17, 359–366. [48] Kmita, M., Tarchini, B., Zakany, J., Logan, M., Tabin, C.J. and Duboule, D. (2005) Early developmental arrest of mammalian limbs lacking HoxA/HoxD gene function. Nature 435, 1113–1116. [49] Tarchini, B. and Duboule, D. (2006) Control of Hoxd genes’ collinearity during early limb development. Dev. Cell 10, 93–103. [50] Davis, A.P., Witte, D.P., Hsieh-Li, H.M., Potter, S.S. and Capecchi, M.R. (1995) Absence of radius and ulna in mice lacking hoxa-11 and hoxd-11. Nature 375, 791–795. [51] Woltering, J.M. and Duboule, D. (2010) The origin of digits: expression patterns versus regulatory mechanisms. Dev. Cell 18, 526–532. [52] Creyghton, M.P., Cheng, A.W., Welstead, G.G., Kooistra, T., Carey, B.W., Steine, E.J., Hanna, J., Lodato, M.A., Frampton, G.M., Sharp, P.A., Boyer, L.A., Young, R.A. and Jaenisch, R. (2010) Histone H3K27ac separates active from poised enhancers and predicts developmental state. Proc. Natl. Acad. Sci. USA 107, 21931–21936. [53] Rada-Iglesias, A., Bajpai, R., Swigut, T., Brugmann, S.A., Flynn, R.A. and Wysocka, J. (2011) A unique chromatin signature uncovers early developmental enhancers in humans. Nature 470, 279–283. [54] Montavon, T. and Duboule, D. (2013) Chromatin organization and global regulation of Hox gene clusters. Philos. Trans. R. Soc. Lond. B Biol. Sci. 368, 20120367. [55] Tschopp, P. and Duboule, D. (2011) A regulatory ‘landscape effect’ over the HoxD cluster. Dev. Biol. 351, 288–296. [56] Duboule, D. and Morata, G. (1994) Colinearity and functional hierarchy among genes of the homeotic complexes. Trends Genet. 10, 358–364. [57] Herault, Y., Fraudeau, N., Zakany, J. and Duboule, D. (1997) Ulnaless (Ul), a regulatory mutation inducing both loss-of-function and gain-of-function of posterior Hoxd genes. Development 124, 3493–3500. [58] Peichel, C.L., Prabhakaran, B. and Vogt, T.F. (1997) The mouse Ulnaless mutation deregulates posterior HoxD gene expression and alters appendicular patterning. Development 124, 3481–3492. [59] Sheth, R., Bastida, M.F., Kmita, M. and Ros, M. (2013) ‘‘Self-regulation’’, a new facet of Hox genes’ function. Dev. Dyn. 243, 182–191.

[60] Williamson, I., Berlivet, S., Eskeland, R., Boyle, S., Illingworth, R.S., Paquette, D., Dostie, J. and Bickmore, W.A. (2014) Spatial genome organization: contrasting views from chromosome conformation capture and fluorescence in situ hybridization. Genes Dev. 28, 2778–2791. [61] Tabin, C. and Wolpert, L. (2007) Rethinking the proximodistal axis of the vertebrate limb in the molecular era. Genes Dev. 21, 1433–1442. [62] Zeller, R., Lopez-Rios, J. and Zuniga, A. (2009) Vertebrate limb bud development: moving towards integrative analysis of organogenesis. Nat. Rev. Genet. 10, 845–858. [63] Rodrigues, A.R. and Tabin, C.J. (2013) Developmental biology. Deserts and waves in gene expression. Science 340, 1181–1182. [64] Villavicencio-Lorini, P., Kuss, P., Friedrich, J., Haupt, J., Farooq, M., Turkmen, S., Duboule, D., Hecht, J. and Mundlos, S. (2010) Homeobox genes d11–d13 and a13 control mouse autopod cortical bone and joint formation. J. Clin. Invest. 120, 1994–2004. [65] Bruneau, S., Johnson, K.R., Yamamoto, M., Kuroiwa, A. and Duboule, D. (2001) The mouse Hoxd13(spdh) mutation, a polyalanine expansion similar to human type II synpolydactyly (SPD), disrupts the function but not the expression of other Hoxd genes. Dev. Biol. 237, 345–353. [66] Yamada, G., Suzuki, K., Haraguchi, R., Miyagawa, S., Satoh, Y., Kamimura, M., Nakagata, N., Kataoka, H., Kuroiwa, A. and Chen, Y. (2006) Molecular genetic cascades for external genitalia formation: an emerging organogenesis program. Dev. Dyn. 235, 1738–1752. [67] Seifert, A.W., Harfe, B.D. and Cohn, M.J. (2008) Cell lineage analysis demonstrates an endodermal origin of the distal urethra and perineum. Dev. Biol. 318, 143–152. [68] Perriton, C.L., Powles, N., Chiang, C., Maconochie, M.K. and Cohn, M.J. (2002) Sonic hedgehog signaling from the urethral epithelium controls external genital development. Dev. Biol. 247, 26–46. [69] Suzuki, K., Ogino, Y., Murakami, R., Satoh, Y., Bachiller, D. and Yamada, G. (2002) Embryonic development of mouse external genitalia: insights into a unique mode of organogenesis. Evol. Dev. 4, 133–141. [70] Cohn, M.J. (2011) Development of the external genitalia: conserved and divergent mechanisms of appendage patterning. Dev. Dyn. 240, 1108– 1115. [71] Cobb, J. and Duboule, D. (2005) Comparative analysis of genes downstream of the Hoxd cluster in developing digits and external genitalia. Development 132, 3055–3067. [72] Podlasek, C., Houston, J., McKenna, K.E. and McVary, K.T. (2002) Posterior Hox gene expression in developing genitalia. Evol. Dev. 4, 142–163. [73] Fromental-Ramain, C., Warot, X., Messadecq, N., LeMeur, M., Dolle, P. and Chambon, P. (1996) Hoxa-13 and Hoxd-13 play a crucial role in the patterning of the limb autopod. Development 122, 2997–3011. [74] Kondo, T., Zakany, J., Innis, J.W. and Duboule, D. (1997) Of fingers, toes and penises. Nature 390, 29. [75] Warot, X., Fromental-Ramain, C., Fraulob, V., Chambon, P. and Dolle, P. (1997) Gene dosage-dependent effects of the Hoxa-13 and Hoxd-13 mutations on morphogenesis of the terminal parts of the digestive and urogenital tracts. Development 124, 4781–4791. [76] Changeux, J.P. (2011) 50th anniversary of the word ‘‘allosteric’’. Protein Sci. 20, 1119–1124. [77] Doyle, B., Fudenberg, G., Imakaev, M. and Mirny, L.A. (2014) Chromatin loops as allosteric modulators of enhancer-promoter interactions. PLoS Comput. Biol. 10, e1003867. [78] Giorgetti, L., Galupa, R., Nora, E.P., Piolot, T., Lam, F., Dekker, J., Tiana, G. and Heard, E. (2014) Predictive polymer modeling reveals coupled fluctuations in chromosome conformation and transcription. Cell 157, 950–963. [79] Groschel, S., Sanders, M.A., Hoogenboezem, R., de Wit, E., Bouwman, B.A., Erpelinck, C., van der Velden, V.H., Havermans, M., Avellino, R., van Lom, K., Rombouts, E.J., van Duin, M., Dohner, K., Beverloo, H.B., Bradner, J.E., Dohner, H., Lowenberg, B., Valk, P.J., Bindels, E.M., de Laat, W. and Delwel, R. (2014) A single oncogenic enhancer rearrangement causes concomitant EVI1 and GATA2 deregulation in leukemia. Cell 157, 369–381. [80] Duboule, D. (2007) The rise and fall of Hox gene clusters. Development 134, 2549–2560. [81] Sharpe, J., Nonchev, S., Gould, A., Whiting, J. and Krumlauf, R. (1998) Selectivity, sharing and competitive interactions in the regulation of Hoxb genes. EMBO J. 17, 1788–1798. [82] Duboule, D. (1994) Temporal colinearity and the phylotypic progression: a basis for the stability of a vertebrate Bauplan and the evolution of morphologies through heterochrony. Dev. Suppl., 135–142. [83] Soshnikova, N., Dewaele, R., Janvier, P., Krumlauf, R. and Duboule, D. (2013) Duplications of hox gene clusters and the emergence of vertebrates. Dev. Biol. 378, 194–199. [84] Andrey, G. and Duboule, D. (2014) SnapShot: Hox gene regulation. Cell 156 (856–856), e1.

Please cite this article in press as: Lonfat, N. and Duboule, D. Structure, function and evolution of topologically associating domains (TADs) at HOX loci. FEBS Lett. (2015), http://dx.doi.org/10.1016/j.febslet.2015.04.024