Evidence for the temporal regulation of insect segmentation by a conserved sequence of transcription factors

ABSTRACT Long-germ insects, such as the fruit fly Drosophila melanogaster, pattern their segments simultaneously, whereas short-germ insects, such as the beetle Tribolium castaneum, pattern their segments sequentially, from anterior to posterior. Although the two modes of segmentation at first appear quite distinct, much of this difference might simply reflect developmental heterochrony. We now show here that, in both Drosophila and Tribolium, segment patterning occurs within a common framework of sequential Caudal, Dichaete and Odd-paired expression. In Drosophila, these transcription factors are expressed like simple timers within the blastoderm, whereas in Tribolium they form wavefronts that sweep from anterior to posterior across the germband. In Drosophila, all three are known to regulate pair-rule gene expression and influence the temporal progression of segmentation. We propose that these regulatory roles are conserved in short-germ embryos, and that therefore the changing expression profiles of these genes across insects provide a mechanistic explanation for observed differences in the timing of segmentation. In support of this hypothesis, we demonstrate that Odd-paired is essential for segmentation in Tribolium, contrary to previous reports.


Fig. S2. Pair-rule gene expression in Dichaete mutant embryos.
Embryos are at late phase 2. In all cases, pair-rule periodicity is still present in the mutants, but stripes are irregular in width and intensity. Expression patterns are broadly consistent across stage-matched embryos. Arrowheads point to a weak/delayed slp stripe 4. Embryos are at late phase 2. Expression patterns of repressors (magenta) are shown relative to those of their target genes (green). (For a description of the pair-rule network, see Clark (2017).) In most cases (e.g., eve versus ftz/odd/slp, or runt versus slp), the relative phasing of the stripes is preserved, suggesting that cross-regulatory interactions are operating normally. Only the phasing of runt expression relative to hairy and odd is clearly abnormal. Scale bars = 50 µm.

Fig. S4. Expression of Tc-cad, Tc-Dichaete, and Tc-opa relative to a common segment marker, Tc-wg, in
Tribolium castaneum germband stage embryos. (A-U). Sets of three Tribolium castaneum germband stage embryos that have been stage matched using Tc-wg expression patterns (Tc-wg expression brown in all panels). Stage-matched germband embryos increase in age from A to U. In each set of embryos, the left-hand embryo is also stained for Tc-cad expression, the middle embryo is stained for Tc-Dichaete expression and the right-hand embryo is stained for Tc-opa expression (all blue stains). In the double in situ hybridizations for Tc-cad & Tc-wg (left-hand embryos) the mandibular (Mn), prothoracic (T1), 1 st abdominal (A1), 4 th abdominal (A4), 7th abdominal (A7) and/or 10 th abdominal (A10) stripes of Tc-wg expression have been labeled. Note how the relative position of the expression domains of these three genes is remarkable conserved across progressive germband elongation stages. Consult Fig. 5 for a clear comparison across different developmental stages, rather than between Tc-cad, Tc-Dichaete & Tc-opa expression patterns. Double in situ hybridization for Tc-cad (blue) and Tc-Dichaete (brown) in embryos of increasing age from left (A) to right (F). (G-L). As for panels (A-F), but this time Tc-Dichaete DIG and Tc-cad FITC RNA probes were used instead of Tc-cad DIG and Tc-Dichaete FITC RNA probes such that the colours are reversed. Note the stripe of Tc-Dichaete expression that is observed anterior to the Tc-cad domain in some, but not all, embryos. (M-R). Double in situ hybridization for Tc-cad (blue) and Tc-opa (brown) in embryos of increasing age from left (M) to right (N). Panels (M-R) show higher magnification images of the regions in M-R where Tc-cad and Tc-opa expression overlaps. These data suggest that as posterior germband cells move anteriorly relative to the posterior tip of the elongating embryo due to convergent extension cell movements, they experience a drop in Tc-cad expression levels as Tc-opa expression levels increase. (S-X). Double in situ hybridization for Tc-Dichaete (blue) and Tc-opa (brown) in embryos of increasing age from left (S) to right (X). Black arrows points to late Tc-opa segmental stripes that overlap strong segmentally-reiterated Tc-Dichaete expression domains that are limited to the medially positioned neuroectoderm. Colour-coded lines on the right-hand side of the embryos indicate our interpretations of the expression patterns in (A-X). (A) At gastrulation, cad is transiently expressed in weak pair-rule stripes (white arrowheads). These stripes have previously been observed at the protein level during germband extension (Macdonald & Struhl 1986). (B) Weak pairrule stripes of Tc-cad (blue arrowheads) are sometimes observed anterior to the broad posterior domain. The domain corresponding to the lower arrowhead in the left panel has been reported previously (Schulz & Tautz 1995). In both Drosophila and Tribolium, these pair-rule cad stripes are located in the posterior of even-numbered parasegments, overlapping with even-numbered wg stripes. (C) During germband extension, ventral opa expression transitions to a segmental pattern. opa stripes posteriorly abut each en stripe, but are excluded from the cell row anterior to each en stripe. (D) Tc-opa exhibits an equivalent pattern in the segmented germband, posteriorly abutting each Tc-en stripe (images from Fig. S10).  (A-T) Double in situ hybridization for Tc-prd (blue) and Tc-Dichaete (brown) in embryos of increasing age from youngest (A) to oldest (T). Colour-coded lines on the right-hand side of the embryos indicate our interpretations of the expression patterns in (A-T). Note how the primary pair-rule stripes of Tc-prd first appear and form within the posterior-most Tc-Dichaete domain (see where blue lines overlap brown lines). In contrast, segmental stripes of Tcprd expression resolve by splitting anterior to this domain (see where blue lines lie anterior to the brown line). While dissecting and cleaning the embryos we noted that Tc-prd expression remains on stronger and longer in the overlying amnion compared to the underlying ectoderm; this is particularly apparent in panels (K-M), where the Tc-prd stained amnion has been ripped away while cleaning the embryo of yolk to reveal ectoderm free from Tc-prd expression (asterisks). Amnion-related expression can be seen down the lateral margins of many of the embryos where some amnion cells survived dissection and cleaning.   eRNAi experiments, displaying similar local segment fusion phenotypes. The frequency of these phenotypes was between 15% and 37% across the four different RNAi experiments. The relative frequency of cuticles exhibiting local segment fusions, and the number of fused segments per embryo, was higher in eRNAi compared to pRNAi (see Supplementary Tables 1-3 and text for further details). (C) Representative cuticles from 3' pRNAi, 3' eRNAi and 5' eRNAi, experiments displaying similar strong segmentation phenotypes. Less then 1% of cuticles exhibited these phenotypes in 3' pRNAi, and none were observed in 5' pRNAi, whereas their number and frequency was higher following 3' & 5' eRNAi (11-15%). Three representative cuticles are shown for each eRNAi experiment to illustrate the consistent 'pair-rule-like' appearance of these phenotypic cuticles; i.e. T1 & T2 legs fused, mandibular and labial appendages often lost, and only 4 abdominal segments obvious.  (e), abnormal leg(s) and T2 leg bifurcation(s) larval phenotypes observed in the 3' pRNAi, 5' pRNAi, 3' eRNAi and 5' eRNAi experiments, compared with sham injection controls. NB. An equivalent graph for head phenotypes is shown in Fig. 7. Refer to Tables S1-3 for exact details of the relative frequency of these phenotypes across the RNAi experiments and their controls. (E) Representative cuticles from each RNAi experiment exhibiting either one or two antennae that are abnormally twisted backwards (white arrowheads). (F) Representative cuticles from each RNAi experiment showing abnormalities in leg development (white arrowheads); note that these deformities included one or more of the following: twisted leg, short leg (absorbed into body wall), fused leg segments, bifurcated leg. (G) Representative cuticles from each RNAi experiment exhibiting asymmetric T2 leg bifurcations (white arrowheads); note that the proximodistal position of these bifurcations varied (from femur to claw). This phenotype represents a common subclass within the 'abnormal leg ( We suspect that disruption of the early blastoderm wedge shape domain shown in Fig. 8, which covers the future antennal segment, is linked to the broad head phenotypes (H, I), whereas disruption of the later domains of Tc-opa expression within the antennal segment is responsible for twisted antenna(e)). (iii) Reduced Tc-opa expression at the base of, and/or surrounding, developing appendages (black arrowheads in N; compare to the stage-matched Tc-opa RNAi embryo in O). Note that Tc-opa expression within and/or surrounding some gnathal appendages (i.e. mandibles; Mn & maxillae; Mx in N) is much stronger than that seen in/around leg appendages, and remains relatively strong in RNAi embryos (O), perhaps explaining why gnathal appendages were refractory to our Tc-opa RNAi. (iv) A patch of ectopic Tc-wg expression on the left side of the T2 segment (white arrowhead in O), is likely associated with the T2 leg bifurcations we observe in cuticles (G); note that the knockdown of Tc-opa seems quite efficient in the T2 segment (O), perhaps resulting in the derepression of Tc-wg (see Discussion).

Fig. S13. The Tc-opa RNAi blastoderm phenotype.
(A) The percentage of eggs that had reached the germband stage in early 48-hour (30°C) egg collections taken from 3' and 5' Tc-opa parental RNAi (pRNAi) females and their parallel control (buffer) injected females. Both 3' and 5' Tcopa pRNAi results in a drop in the percentage of germband stage embryos relative to controls, however this reduction is much more dramatic with 5' Tc-opa pRNAi. (B) A random sample of 50 germband-less eggs was taken from the same 5' Tc-opa pRNAi egg collection as shown in (A) and stained with DAPI. Despite being up to 48-hours old, only one egg (2%) had formed a blastoderm; this egg is shown in panel (B'''). The majority of eggs (68%) exhibited cleavage nuclei within the yolk, suggesting that in most of these eggs embryogenesis had started, but development had stalled prior to blastoderm stage (one of these eggs is shown in B''). The remaining 30% of eggs showed no sign of cleavage nuclei (although a polar body nuclei was clearly present in some cases; example shown in B'). However, it cannot be ruled out that some of these eggs possessed early cleavage nuclei undetected deeper within the yolk.

Fig. S14. Increased frequency of cuticle ball and cuticle fragment phenotypes following Tc-opa RNAi.
(A) The percentage of cuticles scored as 'cuticle balls' and/or 'cuticle fragments' following pRNAi or eRNAi and associated parallel injection controls. Cuticle balls/fragments were observed in higher numbers in our second 5' pRNAi experiment (see discussion associated with Table S1), and in 3' and 5' eRNAi experiments, when compared to injection controls. (B-C) Examples of eRNAi eggs containing cuticle balls and/or cuticle fragments arranged in two highly speculative phenotypic series. Note that in each of the eight images a fully developed hindgut (Hg) is present, suggesting that this aspect of development proceeded as normal. The speculative phenotypic series in row (B) begins on the left with a cuticle that would have been classified as a strong head phenotype (note the lone pair of maxillae and almost complete abdomen) had its thoracic and/or anterior abdominal segments(?) not collapsed and shriveled up into a bristle lined cylinder. Numerous cuticles assigned to this phenotypic class were bristle-lined cylindershaped cuticles (with an absence of other discernible features); increasingly severe examples are shown along row (B). In contrast, the speculative phenotypic series in row (C) begins on the left with a large cuticle ball that could be interpreted as an extreme segmentation phenotype, with gnathal appendages perhaps present but indecipherable, evidence of only extremely short legs and less than 4 clear abdominal segments. Numerous cuticles assigned to this phenotypic class were smaller cuticle balls, attached to -or alongside -a fully developed hindgut, whereas other eggs exhibited cuticles that appeared to have broken up, with some recognizable structures remaining (e.g. a fully developed leg); examples of these are arranged in order of increasing severity along row (C). These cuticles are almost impossible to interpret, since no two are entirely alike, and examples are also observed following embryonic control injections. However, given their increased relative frequency in 3' and 5' Tc-opa eRNAi compared to controls, and their observation in a 5' pRNAi experiment (i.e. arguing against injection artifacts being solely responsible), it is possible that at least some of these cuticles are the direct, or indirect, result of strong Tc-opa RNAi knockdowns, and represent extreme head and/or segmentation phenotypes. This may explain why in the Tc-opa eRNAi experiments only 10-15% of eggs exhibit clear and interpretable pair-rule-like phenotypes.

Supplementary Table 1
The percentage of cuticles that exhibited each class of egg or cuticle phenotype in each of the parental RNAi experiments and their corresponding parallel injection controls.  In the first round of pRNAi experiments 3' and 5' dsRNA was injected into adult females on different days, each time alongside parallel injection controls, such that there is a control group associated with each dsRNA fragment. The same population (box) of animals was used, and subsequent egg collections were made at the same times in relation to the day of injection. In the second round of pRNAi experiments, 3' and 5' dsRNA was injected on the same day, alongside one set of injection controls. In the first 5' pRNAi experiment, cuticle preparations were made before all eggs would have had the opportunity to secrete cuticles. It is notable that the control eggs possessed a significant number of embryos that were in the process of secreting cuticle (23/480), and therefore "Not scorable" as wildtype or phenotypic cuticles (fifth line of table). In contrast, 5' pRNAi eggs possessed very few developing embryos (1/392), consistent with the higher level of empty eggs in this experiment. The second round of pRNAi experiments was therefore carried out partly to gain a more accurate measure of the frequency of empty eggs, but also acted as an experimental repeat. In order to gain a more accurate comparison between the frequencies of empty eggs in pRNAi  145: doi:10.1242/dev.155580: Supplementary information vs. eRNAi experiments, in the second pRNAi experiments eggs from injected females were lightly bleached and lined up on slides, as they would be for embryonic injection.
We note that the second set of 3' and 5' pRNAi experiments appear to have resulted in stronger knockdowns. Importantly, the same classes of phenotype were observed across all pRNAi experiments. However, the frequency of phenotypes, and/or the frequency of stronger phenotypes relative to weaker ones, was generally higher in the second round of pRNAi experiments. Of particular note is the higher number of cuticle balls/fragments observed in the second 5' pRNAi experiment. There are a number of potential explanations for this, none mutually excusive: (i) 5' pRNAi eggs (i.e. stronger knockdowns) were more sensitive to the mechanical manipulation associated with lining eggs up on slides; this might also explain the high number of cuticle balls/fragments seen in eRNAi experiments. (ii) Cuticle ball/fragment phenotypes represent a stronger knockdown than strong head phenotypes (see Fig. S14). This is supported by the observation that although the frequency of head phenotypes was lower overall in the second 5' pRNAi experiment, there was a higher proportion of strong head phenotypes (60% vs. 40%). (iii) The lower number of cuticles obtained and scored for the 5' pRNAi experiments (e.g. 43 compared to 232 for 3' pRNAi), means that the data are more sensitive to random variations.

Supplementary Table 2
The percentage of cuticles that exhibited each class of egg or cuticle phenotype in each of the embryonic RNAi experiments and their corresponding parallel injection controls.  The percentage of cuticles that exhibited each class of egg or cuticle phenotype in each of the parental and embryonic RNAi experiments.