Tetrapod-teleost conserved GLI3-intronic enhancers identified by comparative sequence analysis
Multi-species alignment of human GLI3 genomic sequence with orthologous intervals from other vertebrate species localized 12 intronic conserved non-coding elements, showing at least 50% identity over a 60 bp window down to Fugu. These elements are distributed across almost the entire GLI3 interval (Figure 1A and 1B), with 2 elements in each of introns 2, 3, 4, and 10 and one in each of introns 1, 6, and 13 [25]. The GLI3 specific gene regulatory functions of 11 of these putative human enhancers had previously been determined using human cell lines (Figure 1C). The elements which could activate reporter gene expression in cell cultures functioned likewise in zebrafish embryos [25]. Additionally, the spatiotemporal aspects of one ultraconserved element, CNE2, were analyzed in mouse embryos [26]. However, the spatiotemporal functionality of other GLI3 associated enhancers in a mammalian model remained to be defined.
To determine the tissue specific role of CNEs 1, 2, 6, 9, 10, and 11, which had acted as enhancers in cell cultures and zebrafish embryos, we generated transgenic mice driving lacZ reporter gene expression under the control of each CNE (Figure 1 and Additional file 1: Table S1). The boundaries of the selected subset of enhancer regions were defined bearing in mind that full scale enhancer activity is determined by a combination of core sequences conserved between human and teleosts (Fugu) and flanking tetrapod-specific sequences [28].
CNE10 mediated lacZ expression was largely confined to foregut derivatives, eye and mammary placodes, and will be dealt with in detail elsewhere. Here, we focus on the potential of CNEs 1, 2, 6, 9, and 11 to replicate endogenous Gli3 expression pattern during development of the limbs and the central nervous system (CNS).
Enhancer elements from GLI3introns directing expression in the mouse limb bud
The Gli3 expression patterns within the nascent limb bud are highly dynamic (Figure 2). Initially, Gli3 is expressed broadly in the mesenchyme of the emerging limb bud. At later stages, the genetic antagonism between Gli3 and Shh results in exclusion of the Gli3 expression domain from the posterior limb mesenchyme [29].
This expression pattern is in agreement with limb specific anomalies in Gli3 mutants, in particular the anteroposterior patterning of distal limb elements, i.e. the autopod [30]. Gli3 likewise plays a critical role in regulating the patterning of proximal and intermediate skeletal elements of limbs (stylopod/zeugopod patterning) at very early stages of development [21]. Importantly, the Gli3 functions in stylopod/zeugopod skeletal patterning are independent of its role in the anteroposterior patterning of the handplate [21]. However, as Gli3 is broadly present throughout the developing limb (Figure 2), its expression within proximal mesenchymal condensations (cartilage condensations of stylopod and zeugopod elements) can easily be overlooked [21]. Nevertheless, recently, through Western analysis and in situ hybridization, Gli3 was detected in the cartilage of developing limb elements [31] and plays a critical role in regulating proliferation during endochondral bone formation [32].
In transgenic mouse embryos, the spatiotemporal regulatory activity of two distinct enhancers, CNE6 and CNE11, reflects several of the known aspects of endogenous Gli3 expression within cartilaginous and non cartilaginous mesenchyme of embryonic limbs (Figure 3). CNE6-directed lacZ expression coincides with the emergence of the limb bud, continues towards anterior, and concentrates at E13.5 at the prospective mesenchyme condensations in the digits (Figure 3B and 3C). This spatiotemporal activity overlaps with Gli3 function during formation of proximal skeletal elements, stylopod/zeugopod [33].
In contrast, CNE11 directs reporter expression specifically within proximal regions of the limb bud from stage E12 on, once the mesenchyme starts to condense and form precartilage (Figure 3B and 3C). This spatiotemporal activity overlaps with Gli3 function in patterning of proximal skeletal elements, stylopod/zeugopod [21]. Thus CNE6 and CNE11 elements showed non-redundant regulatory activities. Since the sites of stable transgene insertion in the mouse lines have not been determined, this conclusion awaits confirmation by a larger number of independent transgenic mouse embryos. Multiple sequence alignment coupled with pattern recognition computer programs identified conserved binding sites for number of transcription factors in CNE6 and CNE11 intervals (see Additional file 2 & Additional file 3: Figures S1 & S2), many of which are among the core set of limb regulators and are known to be co-expressed with Gli3 during early limb patterning and growth [34]. Their role in the control of Gli3 expression by interaction with CNE6 and CNE11 will be tested experimentally.
It is of note that Gli3 transcriptional activity appears to be important for the separation of individual fingers [2], and also endogenous Gli3 is known to be expressed strongly in the interdigital mesenchyme at E12.5 (Figure 2 and [35]). However, the mesenchyme in between the prospective digit rays is an area where we did not observe this much expression with both CNE11 and CNE6 enhancer elements. Previously, we had detected reporter gene activity in the interdigital mesenchyme at E13.5 in transgenic mice employing the CNE2 enhancer element [26]. Earlier interdigital Gli3 expression appears to be governed by other enhancer regions. The VISTA enhancer browser http://enhancer.lbl.gov lists as element_1585 a Gli3 intragenic enhancer activating reporter gene activity at E11.5 distally in mouse embryo limbs. Study of the temporal and spatial activity of this enhancer element, which did not meet the inclusion criteria of our study, must be awaited to determine if it acts in a complementary fashion or if it overlaps the activity of CNE6 or CNE11.
Enhancer elements from GLI3introns directing expression in the chicken limb bud
Expressing a reporter controlled by human GLI3-CNEs in transgenic zebrafish embryos could not identify one of them reliably as fin-specific enhancer [28]. To determine, if limb bud specificity similar to the results obtained in mice was attributed to individual GLI3-CNEs in birds, we analysed if CNEs 1, 6, 9, 10, and 11 act as enhancers of Gli3-specific expression in the chicken limb buds. These CNEs were cloned into a GFP reporter construct under the control of a β-globin promoter (Figure 4A)[36] and co-electroporated together with an RFP reporter, to control for electroporation efficiency, into the chicken wing bud at stage HH 19/20. The developing chicken limbs were assessed for RFP and GFP expression 48 hours following electroporation when the embryos had reached approximately stage HH 26. At this stage, Gli3 has been reported to be expressed throughout the proximal region and at the distal anterior edge of the chicken wing bud (Figure 4C) [37]. Figure 4B shows the results of the electroporation experiments. The upper row shows control bright field photographs of the electroporated limb buds analyzed in the next two rows. The middle row shows RFP expression in the limb bud indicating the extent of electroporation. The bottom row shows GFP expression in the same limb bud. CNEs 1, 9, and 10 gave no GFP expression despite RFP being expressed throughout the limb. CNE6 and CNE11 had weak GFP expression in 3/12 and 2/6 cases, respectively, indicating slight enhancer activity. The distribution of Gli3 mRNA in the limb at stage HH26 is shown below via in situ hybridisation (Figure 4C). Gli3 highly expressed distally but also proximally at the posterior margin and therefore reporter activity appears to be within the region of the limb expressing Gli3. At earlier stages, Gli3 is more highly expressed throughout the anterior of the limb bud, and therefore we might expect that electroporation of the putative enhancer constructs at an earlier stage would provide a better test for enhancer activity. In a second set of experiments we therefore electroporated CNE11 into the presumptive limb mesenchyme [38] at stage HH14 of 5 embryos, and then looked for enhancer activity 48 hours post electroporation at approximately stage HH23. RFP expression was found throughout the anterior region of the limb, however, no GFP expression was seen in any of the cases examined (data not shown), although the construct had been successfully electroporated into the region of the limb bud, which would be expressing Gli3.
Consistent with data from mice, CNE11 and CNE6 were able to drive reporter expression in developing chicken limbs at stage HH26 (Figure 4B, arrows), while CNEs 1, 9, and 10 were not. We have previously demonstrated that conserved non-coding elements downstream of the homeobox gene SHOX have enhancer activity using the same assay [36]. In that case three out of the eight CNEs tested showed enhancer activity indicating that some but not all conserved non-coding elements act as enhancers. The remaining CNEs may act to regulate the gene in another way, for example, as repressors. Another difference between the present study and previous experiments with SHOX is that Gli3 has a more restricted expression pattern and lower level of expression than SHOX at later stages. This may reduce the chance of introducing a putative enhancer construct into a Gli3 expressing region of the limb bud and therefore make it more difficult to detect enhancer activity. However, introduction of CNE11 into the limb at an earlier stage when Gli3 expression is more widespread did not show any enhancer activity. This observation is in line with the data obtained in mouse embryos where CNE11 started to enhance reporter gene activity at E12.5 and was inactive at earlier stages.
GLI3enhancer activity reflects evolutionary advances of limb specification
Our finding that enhancer elements within GLI3 act differentially during mouse limb patterning corroborates the current view of limb evolution. Despite the morphological and functional diversity of fish fin and mammalian limbs, development of these structures is regulated by a similar and related set of genes [39]. Evolution of regulatory components was proposed to be key for the origin and subsequent morphological diversification of the vertebrate fin/limb skeleton [40]. The spatiotemporal regulatory role of CNEs 6 and 11 in zebrafish was more redundant with that of other GLI3 enhancers and not preferentially used for fin/limb patterning as now seen in mice [26, 28]. During early embryonic development of the tetrapod limb, GLI3 plays a double role: SHH dependent anteroposterior patterning of the autopod and SHH independent specification of skeletal elements along the proximodistal axis from the stylopod up to the distal margin of the zeugopod [19, 21]. In accordance with these two distinct roles, this study has defined distinct enhancer regions residing in introns of GLI3 that independently regulate expression in the evolutionary ancient stylopod and zeugopod or in modern skeletal structures of the limb autopod, respectively. This suggests that redeployment of ancient cis-regulatory elements to direct GLI3 expression in distinct limb domains might have been instrumental in diversifying the vertebrate limb skeleton during the course of evolution.
Enhancer elements from GLI3introns directing expression in the mouse CNS
The spatiotemporal activities of CNE1, CNE2, and CNE9 complemented each other in the control of reporter expression reflecting part of the GLI3-specific pattern in the brain, spinal cord and craniofacial structures (Figure 1D, Figure 5). The mouse embryo expression patterns governed at E11.5 by CNE1 and CNE2, respectively, are independently reported in the Vista enhancer browser for sequence elements 1213 and 111 which include the sequences employed here http://enhancer.lbl.gov, adding credibility to the notion that the CNEs studied are bona fide GLI3 enhancers.
At day E11.5, CNE1-controlled lacZ was strongly expressed in the dorsal brain and spinal cord, and less prominently in hypaxial buds of the thoracic somites, proximal muscle masses in the forelimb bud, dorsal root ganglion, and in the facial mesenchyme (Figure 5A & 5B). At E12.5, stronger reporter activity was seen in the cerebellum and nerves innervating the dorsolateral trunk region and forelimbs, and extended more rostrally in the head mesenchyme (Figure 5C). In the midbrain, transgene expression was present in the roofplate, dorso-lateral portion of the alar plate, and confined to a marginal layer of the basal plate (Figure 5D). In the spinal cord, X-gal signal was present in the roofplate, in a central region presumably covering the progenitor domains of dorsal interneurons dl5-dl6 and the ventro-lateral progenitor domains Vp0-Vp1, as well as in the ventral most mantle zone of the spinal cord occupying the post-mitotic V3 interneurons (Figure 5E). Additionally, lacZ activity was detected in the medial and lateral nasal processes, precartilage primordium of nasal capsule, Meckel's cartilage, lateral palatine process, and in the dental lamina (data not shown). The role of CNE2, a highly conserved non-coding element, has been outlined previously [26]. Whereas the reporter expression driven by CNE2 was present throughout the walls of telencephalic vesicle, CNE1 activity was confined medially. Thus, there is a partial overlap in the activities of these two enhancers within the anterior domain of the forebrain. At E11.5, reporter activity induced by CNE9 was detected in the dorso-lateral aspects of the anterior and posterior midbrain regions and in ventral portions of the hindbrain and spinal cord up-to the level of forelimb region (Figure 5F and 5G, arrows). At E12.5, CNE9 driven transgene expression was also demonstrated in the medial ganglionic eminence (Figure 5H, arrow). In transverse sections, CNE9 driven lacZ expression was observed in the ventral midline of caudal midbrain (presumptive dopaminergic neurons, Figure 5I, arrow heads) and confined to the dorso-lateral marginal layer (Figure 5I). Reporter activity induced by CNE9 in the CNS overlapped with CNE1 only in the dorso-lateral marginal tissue of the midbrain (Figure 5I).
Notably, in the developing spinal cord CNE9 induced reporter expression appeared up to E11.5 in the motor neuron progenitor domain (pMN, Figure 5J). Thus, regulatory factors operating at different time intervals via CNE9 or CNE1 could activate GLI3 expression in motor neuron or interneuron territories, respectively.
According to current models, in the ventral spinal cord positional information encoded by a ventral to dorsal SHH gradient is transmitted by a GLI code, the interplay of activator or repressor functions of GLI proteins (reviewed by [41]). In mouse embryos, GLI1 functioning as transcriptional activator is expressed in the ventral neural tube whereas the expression pattern of Gli2 remains uniform along the dorsal-ventral axis of neural tube [42, 43]. Gli3 is expressed extensively in the intermediate and dorsal spinal cord regions. Consistent with its expression pattern, genetic studies with mice suggest that Gli3 repressor activity (proteolytically processed isoform) is essential for the normal patterning of at least six neuronal classes: V2, V1, V0, dI6, dI5, and dI4 neurons in the intermediate region of the spinal cord [29]. In addition to its repressor role in the intermediate spinal cord, Gli3 can transduce Hedgehog signaling as an activator. For instance, at the highest levels of Hh (ventral most region of spinal cord) all expression of the Hh target gene Gli1 is dependent on both Gli2 and Gli3. Unlike Gli2, however, Gli3 requires endogenous Gli1 for induction of floor plate and V3 interneurons [44]. Beyond the well established dorso-ventral patterning function through a Gli3-derepression mechanism, Shh and Gli3 activities are required to promote the timely appearance of motor neuron progenitors (MN) in the developing spinal cord [44, 45]. The weak activator functions of endogenous Gli3 observed by Bai and coworkers [44] near the source of Shh are compatible with a subtle expression of Gli3 protein in the three most ventral domains, FP, V3 and MN. The domains of Gli3 expression and Gli3 function in the developing spinal cord reported in these studies are mirrored perfectly by the sites of CNE1 and CNE9 action.
Electroporation of reporter constructs employing conserved Gli3-intronic sequences, which include CNE1 or CNE2, induced the strongest expression signals preferentially in the dorsal spinal cord [12]. However, time, amount, and location of expression governed by these enhancer elements, are analyzed in greater detail in transgenic mouse embryos.
Evolutionary conserved transcription factor binding sites (TFBSs) are predicted in CNE1, CNE2, and CNE9 intervals for multiple established developmental regulators (see Additional file 4 & Additional file 5: Figures S3 & S4, and [26]), many of which are known to be co-expressed with Gli3 during embryonic development of brain and spinal cord [34]. Their interaction with these enhancers remains to be determined experimentally.
Including CNE2, we have identified three independent GLI3-intronic enhancer regions that control reporter expression in developing neural tissues of the mouse embryo in a time- and position-specific complementary fashion. With multiple independent enhancers controlling early CNS patterning, Gli3 follows suit other key developmental genes with a high level of complexity in their genetic regulatory mechanisms governing neural tube patterning [12, 46, 47].
Multiple independently acting regulatory sequences herald the occurrence of higher levels of modularity in the body plans of modern vertebrates
It has widely been accepted that differences in morphological and anatomical traits among closely related species are correlated to changes in cis-acting sequences [48]. Our study on the spatiotemporal activity of independent, anciently conserved cis-regulatory modules, controlling expression of the evolutionary conserved developmental regulator gene GLI3 during limb (Figure 6A) and CNS patterning (Figure 6B and 6C), suggests that these enhancers dictate expression in discrete developmental compartments. Above that, cellular subpopulations within a given compartment, such as the motor neuron or interneuron territories in the spinal cord behave as semiautonomous units with respect to expression control of GLI3 (Figure 6). This subtle specification of enhancer functions corroborates the view that cis-acting regulatory networks of early developmental regulators are often modular, with multiple independent enhancers mediating the expression of associated gene in multiple embryonic compartments independently [49, 50]. Functional changes in one specific cis-regulator through mutations might alter the spatiotemporal distribution of the associated gene product in one developmental domain, whereas the rest of the expression pattern and the protein activity will largely remain un-interrupted. Thus changes in cis-acting sequences will have minimal cost on overall fitness and can serve as raw material for the evolution of morphological and anatomical diversification within and between species.