Expression patterns of BMP DV signalling genes in insects
Our expression analysis for DV patterning genes in C. albipunctata reveals an unexpected amount of upstream variation despite highly conserved target gene output (Fig. 6). Perhaps most surprising is the previously reported restriction of dpp expression to two ventral polar domains [25] (Fig. 2a, b), which stands in stark contrast to the dorsal expression domain observed in D. melanogaster [15]. This represents an extreme case of the very widespread expression variability for BMP ligands across insect species. In A. gambiae, dpp expression is much broader than in D. melanogaster, expanding into the ventral–lateral region of the embryo [20]. In A. mellifera, dpp is expressed in an anterior and a posterior domain that show no obvious DV polarity [55]. In N. vitripennis, dpp shows very faint ubiquitous expression before gastrulation [22]. Finally, in early embryos of T. castaneum, dpp expression is uniform along the DV axis, with a small posterior cap expression, only becoming restricted dorsally after gastrulation [3, 56]. Outside of the holometabola, in blastoderm embryos of the hemipteran O. fasciatus dpp is not detected [50], but is later expressed at the posterior pole, along the dorsal edge of the site of germband invagination [57]. These expression differences may reflect underlying differences in dpp regulation. RNAi knock-down of sog in C. albipunctata leads to expansion of dpp expression into the dorsal posterior pole region and ectopic expression in the middle of the embryo (Fig. 5d). Moreover, it leads to abnormalities in the abdominal and posterior patterning as well as dorsal closure (Fig. 5a–c). A transcriptional effect of sog on dpp has also been described in D. melanogaster [54] and in T. castaneum [3]. The expansion of dpp expression along the AP axis could suggest an implication of the BMP signalling pathway in the AP patterning, as seen in A. mellifera [55] and O. fasciatus [50]. Such an AP role is further supported by the compression of the abdominal segments in the cuticle phenotypes, although an analysis of the mechanism by which this could occur is beyond the scope of the present study.
A similar amount of variation can be observed for sog expression. In C. albipunctata, we detect a single ventral expression domain (Fig. 2b–d, g) similar to that seen in T. castaneum and A. gambiae [3, 20]. In O. fasciatus, early blastoderm embryos have ubiquitous expression of sog, and only at mid-blastoderm does its expression become ventral, similar to the expression described here [50]. These expression patterns differ from the two medio-lateral domains observed in D. melanogaster [58], which therefore may be evolutionarily derived. Interestingly, sog appears to have lost its ancestral role in DV patterning in hymenoptera: in A. mellifera it is only expressed after gastrulation [4], while in N. vitripennis it appears to be completely absent [21].
Other components of the DV patterning system also show variation in expression patterns. In C. albipunctata, we observe expression of the ligand tkv in a dorsal midline domain, covering the posterior-most ~ 25% of the embryo (Fig. 2e, f). In D. melanogaster, tkv is largely, but not exclusively, restricted to the dorsal half of the embryo along the entire antero-posterior axis [59]. In A. gambiae, it is initially expressed throughout the dorsal ectoderm, later to be excluded from the presumptive serosa [20]. In N. vitripennis, it is faintly expressed ubiquitously [22]. In A. mellifera, it is restricted to the posterior third of the embryo, with no DV polarity [55]. This indicates that the localisation of tkv expression cannot explain the dorsal localisation of pMAD, as suggested for D. melanogaster [60, 61].
Finally, the protease tld shows a very dynamic expression pattern in C. albipunctata. At the early blastoderm stage, it shows a dorsal expression domain (Fig. 2g), whereas later this expression has completely shifted to the ventral side of the embryo (Fig. 2h). This surprising switch is not seen in any other insect described so far, but expression variations for tld in other species are also very striking. Whereas tld expression is strictly dorsal in D. melanogaster [62], in N. vitripennis it is expressed in a small anterior dorsal domain and does not have any function in DV patterning [5]. In A. gambiae, tld expression is limited to lateral regions and is excluded from the dorsal ectoderm [20]. In T. castaneum, tld expression occurs in the presumptive germ rudiment and shows the highest levels of expression in a broad anterior domain [24]. In O. fasciatus, tld is expressed uniformly across the embryo circumference [50].
Taken together, the available evidence reveals a surprisingly large amount of expression variation among upstream signalling factors in the DV patterning system of insects. It is striking that the localisation and extent of expression along the DV axis are not particularly conserved for any of these factors.
Localised pMAD activity in insect embryos
Compared to the expression patterns of upstream signalling factors, the localisation of pMAD activity is much more conserved across holometabolan insect species. In C. albipunctata, pMAD is localised in a narrow stripe along the dorsal midline (Fig. 3a–d), similar to pMAD distributions in D. melanogaster (Fig. 3e–g; also see Dorfman and Shilo [63]) and N. vitripennis [5]. In A. gambiae the dorsal domain of pMAD activity is broader [20]. In T. castaneum, pMAD covers the dorsal 50% of the serosa in the anterior, narrowing posteriorly to cover a region of about 20% of the DV axis along the dorsal midline of the germ rudiment [3]. In addition to these similarities, C. albipunctata shows a peculiar deviation from the canonical dorsal holometabolan pattern of pMAD. Its DV domain of localisation expands at both anterior and posterior poles of the embryo, to dorsally mirror the ventral polar domains of dpp expression (Fig. 3b–d). This polar broadening of the pMAD domain is not seen in any other insect studied so far, although much more subtle polar expansions, predominantly around the posterior pole, are also seen in Drosophila and Megaselia [40, 64]. The complementary polar patterns of dpp and pMAD suggest a correlation between the two. We discuss this observation in the context of a potential BMP ligand shuttling mechanism in C. albipunctata below.
Expression patterns of DV target genes in insects
In C. albipunctata, as in other insects, twi and sna have conserved overlapping ventral domains of expression in the blastoderm embryo [4, 20, 21, 65, 66] (Fig. 4a, b). In D. melanogaster, these genes mark the mesodermal anlage [41, 42].
In D. melanogaster, DPP signalling represses expression of brk, while brk in turn is a transcriptional repressor of other DPP target genes [43]. The dorsal morphogen gradient activates brk expression in the ventral neurogenic ectoderm, which restricts dpp expression to the dorsal half of the embryo [67]. This results in opposite activity gradients of pMAD and brk [2]. In C. albipunctata, we observe a conserved brk expression pattern in two ventral–lateral domains (Fig. 4c), very similar to D. melanogaster, A. gambiae, and N. vitripennis [20, 21, 43]. In contrast, brk is not expressed in the early embryo of T. castaneum and sog alone is responsible for restricting dpp expression to the dorsal side of the embryo [3]. It is not clear whether brk was recruited independently into DV patterning in hymenopterans and dipterans, or whether it was lost in the coleopteran lineage, although the evidence slightly favours the former scenario (see below).
In D. melanogaster, pnr and Doc are activated by DPP signalling along the dorsal midline of the blastoderm and are involved in dorsal closure and the specification of the extraembryonic amnioserosa, respectively [2, 45, 46]. In C. albipunctata, Doc is expressed in a dorsal domain excluding the serosa (Fig. 4d), plus a head stripe, similar to D. melanogaster, A. gambiae, and late blastoderm embryos of N. vitripennis [20, 21]. In A. gambiae, Doc expression is restricted to the amnion, but repressed in the serosa [20]. In T. castaneum, Doc plays a role in extraembryonic tissue morphogenesis but not specification; it is expressed early in a dorsal anterior domain, then though the entire serosa, but most strongly in its dorsal region [3, 68]. This indicates some variability in the role of Doc for determining extraembryonic tissues, which reflects the rapid evolution of these tissues among holometabolan insects [69].
In D. melanogaster, N. vitripennis, and T. castaneum, pnr is expressed in a broad dorsal domain during the blastoderm stage [3, 21, 70]. In C. albipunctata, it is not detectable until after gastrulation, exhibiting a pattern in the dorsal epidermis and head lobes of the embryo (Fig. 4e). This pattern is similar to that found in A. mellifera, where pnr shows post-gastrulation expression at the ventral edges of the amnion, resembling post-gastrulation expression in D. melanogaster and N. vitripennis as well [21]. This indicates that the onset of pnr expression seems to vary between species, while its post-gastrulation expression pattern is strongly conserved.
Columnar genes vnd, ind, and msh are lateral markers of the neurogenic ectoderm: vnd is required for the specification of ventral column neuroectoderm, ind for the specification of intermediate column neuroectoderm, and msh labels all of the dorsal column neuroectoderm [47]. Interaction between DPP and SOG is necessary to restrict the expression of these genes to their respective expression sites [48]. In N. vitripennis, like in D. melanogaster, these genes are expressed in ventral–lateral stripes [21, 47,48,49]. In C. albipunctata, ind and msh show similar expression patterns as in other species (Fig. 4g, h). In contrast, vnd can only be clearly detected in the head region, with potential additional expression (although faint and diffused) at the posterior pole and the ventral side of the embryo (Fig. 4f). Unlike in N. vitripennis, D. melanogaster, and A. gambiae, we can only detect expression of all three columnar genes in C. albipunctata after gastrulation.
Despite some interesting differences in target gene expression between species, it is evident that downstream DV genes are much more conserved than the upstream signalling factors. In general, we observe a trend towards increasing conservation of expression and localisation patterns as we move downstream in the DV patterning cascade. This parallels the situation in the antero-posterior patterning system, where the most downstream tier of the segment-polarity gene network shows the highest degree of conservation [71,72,73].
BMP shuttling in C. albipunctata and other insect species
The complex nature of the post-translational shuttling mechanism for BMP ligands poses a challenge for the mechanistic interpretation of our evidence. Comparisons to vertebrate DV patterning suggest that this shuttling mechanism is extremely conserved [10, 17, 74]. What do our data reveal about shuttling of BMP ligands in C. albipunctata? The discrepancy between ventral dpp expression and dorsal pMAD localisation strongly suggests that some sort of ligand transport must be involved in DV pattern formation in this species.
One important feature that both D. melanogaster and C. albipuncatata share is the complementarity of their dpp and sog expression patterns. In D. melanogaster, the border between the two domains occurs in the ventral–lateral region of the embryo along the entire length of the antero-posterior axis [58]. In C. albipunctata, it is restricted to an anterior and a posterior interface, where sog expression abuts, with a slight overlap, the ventrally localised polar dpp domains (Fig. 2b). Shuttling is likely to occur at these interfaces as there will be opposing protein gradients of DPP and SOG present at these sites. This suggests that a significant amount of DPP would be shuttled around the anterior and posterior poles of the embryo in C. albipunctata, involving transport along the antero-posterior as well as the DV axis, rather than straightforward ventral-to-dorsal transport throughout the embryo as observed in D. melanogaster. If our interpretation of the evidence is correct, this can explain the intensified and expanded domains of pMAD in the anterior and posterior polar regions of the C. albipunctata blastoderm embryo (Fig. 3b, c).
There are additional differences between the two species. One concerns the posterior-only expression of tkv in C. albipunctata (Fig. 2e, f). The discrepancy between tkv expression and pMAD localisation may indicate that other BMP receptors, such as PUT, may be required for signal transduction in the anterior of the embryo. Even more difficult to explain is the switch of tld expression from dorsal to ventral during the late blastoderm stage in C. albipunctata (Fig. 2g, h). Its function (if any) remains mysterious, but it either suggests that BMP ligand shuttling must be very dynamic in this species, or that all relevant protein cleavage by TLD and subsequent DPP signalling activity must happen before the dorsal-to-ventral transition in tld expression. Further experimental work including immunostaining against the relevant proteins and/or enzymatic activity assays will be required to gain further insight.
A final difference between species is the absence of scw in the genome of the non-cyclorrhaphan fly C. albipunctata [25]. Comparative analyses suggest that scw arose in the cyclorrhaphan lineage from a duplication of the ancestral gbb homolog [25, 75]. The SCW ligand is essential for DPP transport in D. melanogaster [7]. Is it possible that gbb is fulfilling its role in C. albipunctata? gbb is required for DPP signalling in N. vitripennis [5]. Moreover, it is expressed dorsally in the blastoderm embryo of the scuttle fly Megaselia abdita [76], and in a ubiquitous pattern excluding the poles in C. albipunctata [25], similar to that seen in O. fasciatus [50]. Despite this, our evidence indicates that gbb is not needed for DV patterning in C. albipunctata. Expression of the target gene brk shows no detectable defects in gbb knock-down embryos (Fig. 5g), indicating that dorsally localised DPP signalling is occurring correctly in the absence of GBB protein. Taken together, our evidence suggests that DPP is the only BMP ligand contributing to DV patterning in C. albipunctata.
Our evidence indicates that BMP shuttling is likely to occur in C. albipunctata, although the exact set of factors involved and the spatio-temporal dynamics differ compared to D. melanogaster. This further suggests that BMP ligand shuttling is a conserved phenomenon in dipteran insects. The situation is more complicated in other holometabolan taxa. In T. castaneum, dorsal localisation of DPP signalling activity depends on SOG and TLD as in flies [3] but does not involve tsg [24]. Still, BMP ligand shuttling is probably happening in this species. In contrast, sog is not expressed in embryos of N. vitripennis [21] and is only expressed at late embryonic stages in A. mellifera [4]. Yet, BMP signalling is still responsible for the patterning of the DV axis in N. vitripennis [5]. It has been proposed that maternal localisation of BMP receptors, combined with zygotic regulatory feedback, could take the role of dorsal shuttling in hymenoptera [22]. It is not entirely clear whether this condition is ancestral or derived, although the fact that BMP ligand shuttling has been proposed to occur in vertebrates [17] would favour the latter alternative.
Fundamental differences between DV patterning in dipterans and hymenopterans are further supported by the following evidence: RNAi knock-down of dpp in C. albipunctata leads to an expansion of brk expression to the dorsal midline of the embryo (Fig. 5f). This is similar to brk expression in dpp mutants of D. melanogaster [43], but very different to dpp knock-down in N. vitripennis, where brk is restricted to an antero-dorsal expression domain by an otherwise ubiquitous expansion of twi [5]. In A. mellifera, embryos treated with dpp RNAi also show dorsal expansion of twi [55], although to a lesser degree than in N. vitripennis. In contrast, C. albipunctata embryos treated with dpp RNAi show wild-type twi expression (Fig. 5i), similar to dpp mutants in D. melanogaster [5] (supporting the difference in the determination of the mesodermal fate by Toll or BMP signalling in dipterans versus hymenopterans). This indicates that dpp downregulation in dipterans induces dorsal expansion of the neurogenic ectoderm, while in hymenopterans it leads to an expansion of mesodermal markers. Such fundamental differences in the role of brk between dipterans and hymenopterans favour a scenario where brk was independently recruited into DV patterning in each lineage [5].