Shifting roles of Drosophila pair-rule gene orthologs: segmental expression and function in the milkweed bug Oncopeltus fasciatus

Mechanisms directing the formation of the basic segmented body plan have been unraveled for the model insect, Drosophila melanogaster (reviewed by Wieschaus and Nüsslein-Volhard, 2016). This study identified a set of pair-rule mutants, characterized by absence of alternate body segments, revealing that patterning of single segments is preceded by pre-patterning of a double-segment-wide unit that is repeated along the anterior-posterior axis of the embryo at half the frequency of segment number. Most of the pair-rule genes (PRGs) responsible for this pre-pattern are expressed in seven stripes in the Drosophila blastoderm, with PRG expression foreshadowing the corresponding mutant phenotype for individual PRGs (pair-rule stripes, Fig. 1). For example, even-skipped (eve) and fushi tarazu (ftz) are expressed in complementary seven-stripe patterns, each in the primordia of the alternate parasegments missing in eve or ftz mutants (Lawrence and Johnston, 1989). Other PRGs are expressed in similar complementary patterns, with the combined, staggered expression of the full set of seven-striped PRGs generating unique ‘double-segment’ codes to direct the formation of body segments (Gergen et al., 1986; Graham et al., 2019; Scott and Carroll, 1987). Many of the Drosophila PRGs transition to segmental expression as development proceeds but mutant phenotypes reveal the earliest roles of these genes in PR patterning: roughly half-sized mutant embryos missing alternate segments (reviewed by Wieschaus and Nüsslein-Volhard, 2016).

As all insects are segmented, the gene regulatory logic underlying segmentation might be wholly conserved. However, Drosophila are long-germ insects with all parasegments patterned more or less simultaneously at blastoderm. This mode of development is derived and found only among holometabolous insects, where it independently arose multiple times (Davis and Patel, 2002; Liu and Kaufman, 2005b). In contrast, most insect groups add segments sequentially after the blastoderm stage (‘sequential segmentation’), from the posterior end of the germband, a region known as the growth zone or segment addition zone (SAZ) (reviewed by Davis and Patel, 2002; Liu and Kaufman, 2005b). Thus, the Drosophila-like PR-patterning of a double-segment unit might be restricted to simultaneously segmenting species. However, PR-like expression patterns – defined as stripes of gene expression in the primordia of alternate (every-other) segmental units – have been observed in one or more sequentially segmenting species for orthologs of each of the nine Drosophila PRGs (PRG orthologs): ftz, fushi tarazu factor-1 (ftz-f1), eve, odd skipped (odd), runt (run), hairy (h), odd paired (opa), paired (prd) and sloppy paired (slp). Within holometabolous insects, the expression and function of the complete set of orthologs of Drosophila PRGs has been examined in two species, the sequentially segmenting beetles Tribolium castaneum and Dermestes maculatus (Choe and Brown, 2007; Xiang et al., 2015, 2017). These analyses, together with studies of selected PRG orthologs in a handful of other holometabolous insects (Grbić and Strand, 1998; Kraft and Jäckle, 1994; Nakao, 2010, 2015; Rosenberg et al., 2014), suggest that a role for some PRG orthologs in Drosophila-style PR patterning is shared among holometabolous insects, even those with sequential segmentation (Fig. 1). However, other members of this gene set have changed in expression and/or function within Holometabola (Choe et al., 2017; Clark and Peel, 2018; Heffer et al., 2013a,b). For example, ftz-f1 is expressed ubiquitously in Drosophila but in stripes in beetles (Heffer et al., 2013b; Xiang et al., 2017) and several PRG-orthologs are expressed in the segment addition zone (SAZ) in sequentially segmenting species as components of a vertebrate-like clock-and-wave mechanism, in addition to being expressed in PR stripes (El-Sherif et al., 2012; Sarrazin et al., 2012).

In contrast to studies in holometabolous insects, there has been less focus on PRG expression or function in hemimetabolous insects. PR-like expression of eve was observed in a cricket (Mito et al., 2006, 2007), whereas in grasshoppers, eve and ftz have distinctly non-PR-like expression, both being expressed in the SAZ (Dawes et al., 1994; Patel et al., 1992). Also different from holometabolous species, h is expressed segmentally in a cockroach (Pueyo et al., 2008). Interestingly, PR-like expression of PRG orthologs has been observed in evolutionarily distant, non-insect arthropods. For example, striped expression at half the frequency of segmental stripes (sometimes referred to as ‘double segment periodicity’) has been observed for several PRG orthologs in a centipede (Chipman and Akam, 2008; Chipman et al., 2004; Green and Akam, 2013) and the expression of prd in spider mites is suggestive of modulation by a PR-like regulator (Dearden et al., 2002). These findings suggest two evolutionary hypotheses: PR expression of PRGs arose independently in holometabolous insects and myriapods (Fig. 1, blue); or it was ancestral and lost in some hemimetabolous species (Fig. 1, red). If PR-like expression of PRGs was not ancestral (Fig. 1, blue), segmental expression of the PRG orthologs may be the ancestral state. This scenario is supported by the observation that expression of five Drosophila PRGs evolves from a seven-stripe PR pattern to a fourteen-stripe segmental pattern, either by stripe splitting or by de novo addition of a second set of seven stripes (Fig. 1). In either case, these results suggest extensive rewiring of segmentation networks in arthropods. As it is clear that individual members of the PRG set can vary in function without loss of PR patterning per se (see above), it is necessary to examine the whole set of PRGs in diverse taxa before making broader conclusions about gain or loss of this patterning mechanism.

Oncopeltus fasciatus belongs to the order Hemiptera, a close outgroup of the Holometabola (Misof et al., 2014; Yeates et al., 2012). PR-like expression was observed in Oncopeltus embryos for the gene E75A and RNAi resulted in fusion of neighboring segments, demonstrating the existence of an underlying PR-like pre-patterning mechanism in this species (Erezyilmaz et al., 2009). However, E75A does not have PR-like expression or function in Drosophila (Bialecki et al., 2002; Buszczak et al., 1999; Segraves and Hogness, 1990). In Oncopeltus, eve is expressed in stripes in every segment (‘segmental expression’) and in the SAZ (Liu and Kaufman, 2005a). Here we have isolated and examined the expression of all nine orthologs of the Drosophila PRGs in Oncopeltus (Of-PRG orthologs) and seven paralogs of these genes, and have compared their expression to the only known Of-PRG: E75A. Despite the fact that Of-PRG orthologs are all expressed during the stages at which Oncopeltus specifies segments, only one (Of-run) is expressed in a pattern reminiscent of Drosophila PRGs. Most others are expressed in segmentally reiterated patterns, either in nascent segments in the anterior SAZ or in mature segments of the germband. In keeping with this, PR-like defects were not seen after RNAi-mediated knockdown of Of-PRG-orthologs. These results suggest that, although PR-patterning per se is retained in Oncopeltus, extensive re-writing has occurred such that the genes responsible for this pre-pattern are different from those in Drosophila.

Orthologs of the nine Drosophila pair-rule genes (PRGs), referred to throughout as Of-PRG orthologs, were isolated and gene structures determined (Fig. S1), combining experimental data with information from the Oncopeltus genome (Panfilio et al., 2019). Because the search criteria were designed to identify all potential orthologs, we identified multiple gene family members in most cases. All matches for each PRG were subjected to phylogenetic analysis to determine which, if any, were the ortholog of interest (Fig. S2). This analysis identified orthologs of odd-skipped family members odd, sob and bowl; one opa ortholog; prd/Pax3/7 orthologs prd and gooseberry; four Runt domain family members, run, lozenge, runxA and runxB; the h family members h and deadpan; one slp ortholog with 65% and 56% identity in the forkhead domain to Dmel-Slp1 and Dmel-Slp2; and a single copy of ftz-f1, with genomic sequence on four scaffolds that were merged after experimental verification.

For Of-ftz, three sequences encoding a homeodomain were isolated (Figs S1, S3): Of-ftz-A (788 bp), Of-ftz-B (443 bp) and Of-ftz-C (965 bp). One of these, Of-ftz-A, was also found by RNA-seq (Ewen-Campen et al., 2011) and confirmed by us using 5′RACE. These sequences overlap in a region encoding a full-length Ftz-family homeodomain. Upstream of the homeodomain, only one of these sequences, Of-ftz-C, appears to have a complete open reading frame; for this sequence, an HDWM appears to replace the YPWM motif seen in Hox proteins and homeotic-type Ftz proteins (Johnson et al., 1995), and no LXXLL motif, required for interactions with Ftz-F1, was found (Heffer et al., 2010; Yussa et al., 2001). Furthermore, the characteristic Ftz N-terminal arm (Heffer et al., 2010; Telford, 2000) differs from other Ftz proteins in arthropods. Of note is the substitution at position 4 of the homeodomain: Of-ftz encodes a lysine at this position, while all other arthropod Ftz proteins examined share a serine or threonine (Heffer et al., 2010; Telford, 2000). The crystal structure of the Drosophila Engrailed homeodomain suggests that arginines at homeodomain N-terminal arm positions 3 and 5 make direct contact with DNA; these residues are conserved in Of-Ftz (Kissinger et al., 1990). For the shorter sequences (Of-ftz-A and Of-ftz-B), neither includes a start codon, and stop codons are present in all three reading frames (Fig. S1 and S3). In addition, Of-ftz-A appears to include a 300 bp unprocessed intron; canonical GT-AG splice sites were found flanking an unaligned region of the sequence (Fig. S3). The sequences of all three Of-ftz homeoboxes match 100% at the nucleotide level and align to scaffold 2747 of the genome assembly. The region of Of-ftz-C 5′ of the homeobox aligns to scaffold 1144. Despite the Hox complex being greatly fragmented in the current Oncopeltus genome assembly, a partial Of-Scr sequence was annotated on this scaffold. Future experiments will determine whether the three Of-ftz sequences isolated are different isoforms of one Of-ftz gene or products of distinct Of-ftz paralogs.

Genes playing roles in PR patterning are expected to be expressed first at the blastoderm stage and then throughout germband elongation, as segments are specified. Based on SYTOX green nuclear staining (Fig. S4), genes involved in segmentation should be expressed at 24-48 h after egg laying (AEL), which includes late blastoderm through early germband elongation. RT-PCR spanning the first 5 days of Oncopeltus development was used to determine whether Of-PRG orthologs are expressed at the right time to be involved in segmentation (Fig. S5). For all experiments, Of-actin was simultaneously amplified as an internal positive control (Fig. S5). Owing to its verified role as a PRG in Oncopeltus (Erezyilmaz et al., 2009), the expression profile of Of-E75A – with a clear peak at 24-48 h (AEL) – served as a guide for PR-like expression (Fig. S5A). Expression of Of-eve was detected at 0-24 h AEL, with highest levels at 24-48 h AEL and slightly lower levels after this (Fig. S5B). Of-odd and Of-slp expression were highest at 24-72 h AEL, from the blastoderm stage through germband extension (Fig. S5C,D), whereas Of-slp expression continued through the fifth day of embryonic development. Of-run expression was also highest at 24-48 h AEL, with attenuated expression for 2 more days (Fig. S5E). Of-ftz-f1 showed fairly consistent expression for all time points, including at 0-24 h AEL (Fig. S5F), possibly reflecting maternal deposition, seen for Drosophila ftz-f1 (Guichet et al., 1997; Yu et al., 1997). Of-ftz expression appears to be highest in 0-24 h AEL embryos and expression fades thereafter (Fig. S2G). Time courses for Of-h, Of-prd and Of-opa expression were similar, with consistent expression detected 24-120 h AEL (Fig. S5H-J). Their continued expression after germband extension suggests additional roles later in development. In sum, all orthologs examined were detected at 24-48 h AEL, when highest expression of Of-E75A was also observed, consistent with roles in segmentation, although distinct profiles were seen for each gene.

To determine whether Of-PRG-orthologs are expressed in PR-like spatial patterns, i.e. in stripes in the primordia of every other body segment, whole-mount in situ hybridization was carried out. Of-E75A, which is expressed in PR stripes (Erezyilmaz et al., 2009), serves as a positive control. Of-E75A was initially expressed in two stripes straddling the middle of the blastoderm-stage embryo, the anterior stripe being broad and diffuse, and the posterior stripe much narrower (Fig. 2A). Slightly later, three stripes were observed; this third stripe was added from the posterior, at first diffuse and later narrowing (Fig. 2B,C). A fourth stripe was observed just before gastrulation (Fig. 2D). As gastrulation proceeded, all four blastoderm-stage stripes moved toward the posterior (Fig. 2E). In early germbands, a broad stripe was observed in the SAZ with a narrower stripe just anterior (Fig. 2F). In later germbands, a pair of stripes was seen in the anterior region of the SAZ, with another weak stripe present in the segmental primordia (Fig. 2G).

Of all the Of-PRG orthologs examined, Of-run expression was the most PR-like (Fig. 2H-S′). Of-run was first detected in blastoderm-stage embryos, in two diffuse stripes in the center of the embryo (Fig. 2H), as well as in a posterior ‘cap’ (arrowhead, Fig. 2H), similar to run expression in beetles (Choe et al., 2006; Xiang et al., 2017). At the beginning of gastrulation, stripes 1 and 2 split, while a third primary stripe resolved from the posterior cap generating five stripes prior to gastrulation (Fig. 2I,J). The presence of two thick stripes, each of which split, is reminiscent of PR-gene expression for prd in Drosophila, and eve, h and prd in Dermestes (Kilchherr et al., 1986; Xiang et al., 2017). However, this pattern differs from Of-E75A, which displays strict, alternate segment, PR-like expression, with stripes never splitting. To compare expression of these genes, embryos were bisected and halves were examined for either Of-run or Of-E75A expression. At late blastoderm, when two Of-E75A stripes were detectable, four run stripes were present (Fig. 2K).

During germband elongation, the Of-run posterior SAZ ‘cap’ persisted (Fig. 2L-N). A stripe just anterior to this SAZ expression was also observed (Fig. 2M′,N′ arrowhead). As abdominal segments were added, dynamic expression in the SAZ was evident (Fig. 2L′-S′). In some germbands, only one stripe was observed in the anterior SAZ, and Of-run expression in the posterior SAZ clearly extended to the posterior edge of the germband. In others, two stripes were evident in the anterior SAZ, and a broad stripe, which appeared to span the width of two segments, was observed just posterior (Fig. 2O,P,R), while expression was no longer present in the posterior extreme of the germband. Expression was also seen in later germbands in head lobes, and in segmentally reiterated pairs of dots around the central midline in thoracic and abdominal segments (Fig. 2Q-S). In sum, Of-run appears to initiate expression in a PR-like register in blastoderm embryos, with stripes splitting to generate a segmentally reiterated pattern. Of-run is also expressed dynamically in the SAZ, similar to run expression in other sequentially segmenting species. We classify Of-run expression as PR-like because of the initial broad stripes that split at blastoderm, as well as the appearance of stripes two-segments wide in the germband. However, a clear set of PR alternate-segment stripes, as seen for Of-E75A, was not observed.

Segmental expression of Of-eve was reported by others but is included here for completeness (Liu and Kaufman, 2005a). Of-eve expression was first seen in a broad domain covering about 40-100% egg length (0%, anterior pole) (Fig. 3A). This domain then split into five stripes and a posterior cap, presumably by loss of transcripts from the inter-stripe regions (Fig. 3B). At the start of germband invagination, very weak stripes were observed, with a possible sixth stripe at the far posterior (Fig. 3C). In early germbands, tightly packed stripes in and around the SAZ were observed (Fig. 3D). Six blastoderm stripes are expected for genes expressed in the primordia of every segment, as is the case for Of-invected (Of-inv), a segmentally expressed gene. [Of-inv was originally thought to be Of-engrailed (en) (Genbank accession number AY460340.1) (Liu and Kaufman, 2004); it has since been recognized as invected (Auman et al., 2017).]

Of-prd was first detected at the blastoderm stage, in a dark narrow stripe at ∼40% egg length (Fig. 3E). Two or three very light stripes just posterior to this were barely visible. This first stripe likely corresponds to the mandibular segment; the appearance of the mandibular prd stripe before others has been observed in other insects (Choe and Brown, 2007; Osborne and Dearden, 2005; Xiang et al., 2015). At later blastoderm, six Of-prd stripes were observed; the posterior-most two stripes appeared much lighter than the presumably older, more anterior stripes (Fig. 3F). The stripes moved posteriorly as the germband invaginated (Fig. 3G). In early germbands, dots of expression were seen along the midline, indicating prd expression in central nervous system (CNS) development as in other arthropods (Davis et al., 2005; Osborne and Dearden, 2005). A group of four new tightly packed Of-prd stripes arose in the anterior SAZ similar to Of-eve, but with expression notably absent from the posterior SAZ (Fig. 3H). The register of these nascent stripes in the germband, along with the presence of six closely spaced stripes at blastoderm, demonstrate that Of-prd is expressed segmentally and not in a PR-like fashion.

Of-odd was first detected at blastoderm as a broad stripe at ∼40-80% egg length, with two or three stripes resolving from this domain (Fig. 3I). Five narrow stripes appeared as the earlier diffuse expression cleared (Fig. 3J). Occasionally, a strong-weak alternation of these stripes was observed, possibly reflecting modulation of expression by a regulator expressed in a PR pattern. An analogous alternate strong-weak pattern of stripes is seen for Drosophila en, likely reflecting regulation by different PRGs in alternate sets of stripes (DiNardo et al., 1985). To further test whether Of-odd stripes have segmental or PR register, expression of Of-odd and Of-E75A were compared in bisected embryos. It is clear that there were two Of-odd stripes (bottom half of embryo) for every Of-E75A stripe (top half of same embryo) at this stage (Fig. 3K). In slightly older embryos after the germband started to invaginate, a sixth stripe arose at the posterior (Fig. 3L). In mature segments of the germband, expression in the presumptive mandibular through T2 segments gradually restricted to dots along the midline and one stripe in the newest mature segment (Fig. 3M). In the anterior SAZ, a group of four stripes, similar to Of-prd, was observed. In later germbands, a similar pattern was observed in the SAZ, but expression had faded in the segmented germband (Fig. 3N).

In summary, Of-prd, Of-odd and Of-eve are expressed segmentally in blastoderm and germband stage embryos. For Of-prd and Of-odd, a cluster of stripes was observed in the anterior SAZ. The posterior cap of odd expression seen in beetles was not observed for Of-odd (Choe et al., 2006; Xiang et al., 2017). Overall, no hint of expression in a PR-like register was seen for Of-prd, Of-odd or Of-eve.

Of-opa expression was first detected in late blastoderm-stage embryos as six stripes on the lateral plates (Fig. 4A). In early germbands, before all six blastoderm stripes had invaginated, Of-opa stripes appeared in each segment (Fig. 4B). In later germbands, the six most anterior stripes had invaginated and persisted in each segment with weak expression just anterior to the SAZ (Fig. 4C). This persistent expression is different from Of-odd or Of-prd, where the six segmental stripes in the blastoderm eventually fade after those segment primordia become part of the germband.

Of-slp was first detected at blastoderm in a broad domain covering ∼0-40% egg length, in a pattern complementary to that of Of-eve (Fig. 4D). In slightly later embryos, two stripes had emerged at the posterior boundary of this broad domain as the most anterior expression started to clear (Fig. 4E). This anterior expression had completely cleared by later blastoderm stages, leaving two close stripes at ∼30% egg length (Fig. 4F). A total of seven stripes were observed at blastoderm stage (Fig. 4G,H). In early germbands, double staining revealed Of-slp expression spanning the mediolateral width of each segment, anterior to each inv stripe (Fig. 4I). As germband elongation continued, a broad stripe of Of-slp expression persisted in each mature segment (Fig. 4J-O). These segmental stripes have clearly defined posterior borders but more-diffuse anterior boundaries.

In summary, both Of-opa and Of-slp are expressed in segmental stripes in the blastoderm, which persist through germband elongation. These genes differ from Of-eve, Of-odd and Of-prd in that Of-opa and Of-slp were not detected in the SAZ and their striped expression persisted in the elongated germband, similar to slp in D. maculatus (Xiang et al., 2017).

Of-h expression was first detected later than other PRG orthologs, at late blastoderm stage, in three faint stripes (Fig. 5A). These stripes were never observed to split, and this pattern was observed at much later stages than the three-stripe pattern of Of-E75A (note the invagination pore in Fig. 5A). These stripes are likely the earliest manifestation of the germband expression seen slightly later. In germbands, Of-h stripes were observed posterior to each inv stripe (Fig. 5B). Notably, expression of Of-h appeared to lag behind that of the other genes examined, as it was observed in older, more mature, segments, rather than in the younger segments being generated from the SAZ, suggesting coordinate activation of Of-h stripes by gene(s) already expressed segmentally (Fig. 5C-E). Auman and Chipman (2018) also found h expression in two stripes in the anterior SAZ. In later germbands, new stripes were seen in the abdominal segments, in addition to expression in antennal segments (Fig. 5F). In fully extended germbands, expression in the abdomen faded away, and new expression was seen in labial through T3 segments (Fig. 5G). Later, two or three dots in thoracic appendage primordia were observed (Fig. 5H).

No blastoderm expression of Of-ftz-f1 was detected. In germbands, two stripes were observed anterior to the SAZ during early abdominal segment addition; similar to h, these stripes were weaker along the central midline (Fig. 5I). Later, only one ftz-f1 stripe near the SAZ was observed (data not shown). Double in situ hybridization with Of-inv revealed that Of-ftz-f1 and Of-inv segmental stripes are out of register with each other (data not shown).

No localized expression pattern was observed for Of-ftz during segmentation, using an Of-ftz-C-specific probe or a probe that would detect all three isoforms (Fig. S1). At later stages, segmental expression of Of-ftz-C was observed in groups of internal cells, presumably CNS, as has been seen for ftz in other species (Dawes et al., 1994; Heffer et al., 2013a) (Fig. 5J).

As functional divergence following gene duplication can lead to subfunctionalization of ancestral protein functions, it is possible that, for any of the Of-PRGs, the ancestral gene encoded a protein with pair-rule function, and that this function was relegated to a different paralog in the lineage leading to Oncopeltus than that leading to Drosophila. To investigate whether Of-PRG paralogs are expressed in PR-like patterns, and thus retain potential to perform PR functions, the timing of expression of paralogs was determined by RT-PCR. gooseberry-neuro (gsb-n) was not found in the O. fasciatus genome or in the H. halys genome (Fig. S2C), suggesting loss of this gene in Pentatomomorpha. All other paralogs investigated were located in the genome and gene identity was determined by phylogenetic analysis (Fig. S2). runxB could not be amplified using mixed-stage 0-120 h AEL cDNA, nor could it be identified in the Oncopeltus embryonic transcriptome (Ewen-Campen et al., 2011); sister of odd and bowl (sob), brother of odd with entrails limited (bowl), gooseberry (gsb), runxA, lozenge (lz) and deadpan (dpn) were all found to be expressed within this time frame. As segmentation occurs during the first 3 days of embryogenesis, runxA, which was found to be expressed 72-120 h AEL, was excluded from further analysis.

Expression patterns of sob, bowl, lz, dpn and gsb were determined by in situ hybridization on 24-72 h AEL embryos. No patterned expression was observed for lz and bowl. odd paralog Of-sob was observed in an Of-odd-like pattern at blastoderm in six segmental stripes and later in stripes in the anterior SAZ (Fig. S6A-D). This Of-sob expression pattern was also found by Auman and Chipman (2018), who showed by comparison with Of-eve that Of-sob and Of-odd stripes overlap. Fully elongated germbands showed Of-sob expression in stripes in the appendages, suggesting a role in appendage patterning, as was shown for sob in Tribolium (Angelini et al., 2012). Expression of Of-dpn, an Of-h paralog, was observed in the head lobes and later along the midline, possibly in CNS primordia (Fig. S6F,G). Of-gsb, prd paralog, was observed in one stripe at blastoderm similar to early Of-prd expression (compare Fig. S6H with Fig. 3E), and later in each mature segment of the elongating germband. Thus, none of the paralogs are expressed in PR-like patterns.

Parental RNAi (pRNAi) was performed to assess the function of each Of-PRG. qPCR was used to determine relative expression and verified knockdown of each gene targeted (Fig. S7). Expression of Of-inv in elongating germbands was used to assay segmentation phenotypes as loss of alternate inv stripes would be expected after loss of PR function. RNAi knockdown of odd or prd resulted in severe defects (Fig. 6B,C). No thoracic or abdominal segmentation was apparent in these shortened embryos, suggesting overall loss of segmentation, more similar to that seen in Drosophila segment polarity mutants than in Drosophila pair-rule mutants. Consistent with this, in different RNAi embryos, inv was undetectable, detectable in partial stripes at only very low levels (Fig. 6B) or unpatterned throughout the embryo (Fig. 6C). These effects were seen throughout the embryo, without PR-like register, indicating that Of-odd and Of-prd impact inv stripes in all segments. Although the Of-prd knockdown phenotype was fairly consistent (22/22 embryos, Fig. 6C, Fig. S8B), slightly more variation in phenotype was observed in Of-odd^pRNAi offspring. These differences appeared to broadly stratify with different replicates, indicating that changes in dsRNA integrity or amount of dsRNA injected may cause this variation. In some odd^pRNAi embryos, every inv stripe was nearly lost in every segment (15/29 embryos, Fig. 6B, Fig. S8A7-16) while in others, inv expression was less severely affected (14/29 embryos, Fig. S8A1-6). Auman and Chipman (2018) found fusion of maxillary and labial segments, as well as in the first and second thoracic segments after odd RNAi, which we did not observe. Unhatched odd^pRNAi offspring showed severe segmentation defects, often with all thoracic appendages fused (Fig. 6H), while unhatched prd^pRNAi offspring developed only heads (Fig. 6I). In contrast to Of-odd and Of-prd knockdown, striped inv expression was largely maintained in Of-slp^pRNAi embryos, although five rather than six stripes were consistently observed in the gnathal/thoracic region (23/23 embryos, Fig. 6D and Fig. S8C). Stripes appeared expanded, especially along the midline. As Of-slp stripes are offset from inv stripes (Fig. 4I), this expansion suggests that Of-slp represses inv but without PR-like register. These embryos showed no clear morphological segmentation and failed to develop appendages (Fig. 6J), as found by Auman and Chipman (2018).

Females injected with dsRNA targeting Of-ftz-f1 failed to lay eggs, and dissection revealed that the ovaries of injected females were abnormal, containing oocytes of about the same size along the length of the ovariole (Fig. S9), similar to those seen after Tribolium and Dermestes ftz-f1^RNAi (Heffer et al., 2013b; Xiang et al., 2017; Xu et al., 2010). RNAi knockdown of Of-PRGs ftz, h, run and opa did not impact Of-inv expression compared with gfp controls (47/47, 8/8, 27/27, 166/167 and 37/37 embryos examined, respectively). Although hatch rates were severely suppressed for all concentrations of Of-opa dsRNA tested, no effect was observed on inv stripes (Fig. S10). Examination of unhatched first-instar offspring revealed irregular fusions of thoracic appendages (Fig. 6L), suggesting that Of-opa is required to maintain segment boundaries later in embryogenesis. For Of-h, Of-ftz and Of-run, hatch rates were only slightly lower than controls (Fig. S10). Cuticular defects were observed after Of-ftz^pRNAi (Fig. 6K) in which the maxillary and mandibular segments appeared undifferentiated, similar to Of-Scr^pRNAi, suggesting possible off-target effects; however, the labium appeared similar to wild type, in contrast to the distinct labial to antenna or leg transformation observed after Of-Scr knockdown (Hughes and Kaufman, 2000). No defects in body patterning were observed in Of-run or -h pRNAi offspring (Fig. 6F,G). In most cases, more than one dsRNA sequence was tested at multiple concentrations for each gene. Levels of knockdown, as measured by qPCR for these genes, were similar to those for genes showing abnormal inv expression (Of-odd, Of-prd and Of-slp), suggesting that RNAi was effective (Fig. S7). However, functional inference from RNAi is suggestive but not definitive; future experiments will be required to generate genomic deletions of these genes to more conclusively assess their function. In summary, for those PRG orthologs showing RNAi defects in early embryos, all segments were affected equally. Thus, these genes do not appear to function in a PR-like manner in Oncopeltus.

We isolated and examined the expression patterns and functions of orthologs of the nine Drosophila PRGs, as well as their paralogs, in Oncopeltus. This is, to our knowledge, the first examination of the complete set of PRG orthologs in a hemimetabolous insect. Using this candidate gene approach, we found that only one ortholog, Of-run, is expressed in a striped pattern reminiscent of Drosophila PRGs. However, unlike Of-E75A, which shows PR-like expression, Of-run was not observed in a striped pattern in alternate segments in the blastoderm, but rather in two broad stripes that then split, and no pair-rule defects were observed after knockdown. Most Of-PRG orthologs were expressed segmentally during blastoderm and germband development. The finding that all inv stripes were impacted after Of-odd, Of-prd or Of-slp RNAi knockdowns suggests that these genes are involved in the specification or boundary formation of every segment. This function resembles that of Drosophila segment polarity genes. Among the Of-PRG-orthologs expressed segmentally, distinct temporal and spatial classes were observed. Of-eve, Of-odd and Of-prd were expressed in six segmental stripes in the blastoderm with segmental register in germbands in a transitional zone between the posterior SAZ and the segmented germband. This is of interest because segmental stripes appear to emanate directly from the SAZ. In contrast, Tribolium orthologs expressed in the SAZ emerge with PR periodicity. Of-opa and Of-slp were observed in a similar segmental pattern in the blastoderm and were later seen as persistent stripes in each mature segment of the germband, but were absent from the SAZ. Of-h and Of-ftz-f1 were observed in stripes only at later stages in the mature germband; these stripes appear to be more spatially restricted. Of-ftz-f1 was never seen in more than two stripes anterior to the SAZ; h was present in sets of stripes, either in anterior or in posterior segments, just posterior to Of-inv in germbands. The temporal differences in expression of these genes, along with the increasingly restricted spatial localization of the stripes, suggests a regulatory hierarchy among them. If indeed PR expression for this gene set was ancestral (Fig. 1, red), a coordinated shift in their expression patterns from PR-like to segmental could be explained by cross-regulatory interactions.

While this work was in progress, Auman and Chipman (2018) examined expression of Of-odd, Of-opa, Of-slp, Of-h and Of-run during germband elongation, and tested RNAi knockdown of Of-odd and Of-slp. Their results were largely similar to those shown here. Our study differs from their study in that we analyzed expression and function for the full set of Of-PRGs, including Of-ftz, ftz-f1 and prd, as well as paralogs. Furthermore, a major goal of our approach was to determine whether these genes were expressed in PR-like register. For this, we focused on examination of expression of each gene at the blastoderm stage, where stripe register can be determined clearly by comparison with the one known Oncopeltus gene expressed in a PR-like pattern: Of-E75A (Fig. 2). Comparison with Of-E75A expression, including examination of expression in bisected embryos (Figs 2K and 3K) allowed us to make strong conclusions about the mode of expression of Of-PRG orthologs.

Are orthologs of Drosophila PRGs involved in an ancestral PR mechanism? Orthologs of Dmel-PRGs are expressed in PR-like patterns in a number of holometabolous insects (Aranda et al., 2008; Choe et al., 2006; Rosenberg et al., 2014; Wilson and Dearden, 2012; Xiang et al., 2017), although expression of individual orthologs in different species varies (Fig. 1), as mentioned in the Introduction. This variation among PRG orthologs in Holometabola may reflect an evolutionary reshuffling of functions within one level of a regulatory hierarchy. In hemimetabolous insects, the situation appears to be different: here, we have found absence of PR-like expression or function for most of the PRG orthologs and their paralogs. Although the full set of PRG orthologs has not been examined in other hemimetabolous insects, there is variability in the degree of conservation of expression and function for those orthologs that have been studied. For example, in grasshoppers, neither eve nor ftz is expressed in stripes (Dawes et al., 1994; Patel et al., 1992); however, eve is expressed in broad, PR-like stripes that split to generate segmental expression in the cricket (Fig. 1; Mito et al., 2007). In contrast, an ancestral role for PRG orthologs in patterning a double segment repeat has been suggested from studies of myriapods. In particular, in the centipede Strigamia maritima, orthologs of odd (Sm-odr1), eve, run and h are expressed in stripes that appear in alternate segmental intervals (Chipman and Akam, 2008; Chipman et al., 2004). There is also some evidence for PR-like expression of PRG-orthologs in the millipede Glomeris marginata, based on the finding that initial expression of a number of these genes (eve, run, h, slp and opa) is in double- or multiple-segment-wide domains, which then split into segmental stripes (Janssen et al., 2012). This differs from Drosophila, where no single PRG is expressed in a stripe spanning a double segment primordia; the widest PR stripes are those of prd, which are broader than other PRGs but do not span an entire double segmental unit (Gutjahr et al., 1993). The ‘double segment’ prepattern inferred for Drosophila PR patterning arises from the complementary expression of different PRGs. Thus, these wide PRG-ortholog stripes are not necessarily reflective of a Drosophila-like mechanism. However, an underlying PR mechanism is suggested by the fact that, as seen in the spider mite (Dearden et al., 2002), the Glomeris prd ortholog is expressed in segmental stripes that are stronger in alternate segments (Janssen et al., 2012).

These findings highlight the distinction between PR pre-patterning as a developmental mechanism, and the specific genes identified in Drosophila that control this process. PR-like pre-patterning occurs in Oncopeltus largely or wholly independently of the set of PRGs that play this role in holometabolous insects. E75A, which has no PR-like expression in Drosophila (Wilk et al., 2013), is expressed in the primoridia of alternate body segments in Oncopeltus (Erezyilmaz et al., 2009; Fig. 2), demonstrating that an underlying PR-type mechanism exists in Oncopeltus. Interestingly, PR-like expression in Oncopeltus is also seen for Toll-like genes, which are downstream targets of PRGs in Drosophila (Benton et al., 2016; Graham et al., 2019; Paré et al., 2014) (data not shown). As the Drosophila regulators of these Toll gene PR patterns are non-PR in Oncopeltus, the regulatory interactions controlling their expression must have been re-wired in one of these lineages. Thus, although there is little overlap between Oncopeltus and Drosophila in the sets of genes responsible for the initial subdivision of the embryo into double segment repeats, PR pre-patterning is part of the ‘regulatory logic’ by which the embryo is sequentially subdivided into increasingly specified units. The endpoint of this process – segment formation – is highly stable during evolution, but the regulatory pathways directing it have diverged, a phenomenon termed developmental systems drift (Haag, 2014; True and Haag, 2001), which we refer to as ‘phenotypic stability’.

We close with an historical note of interest. Before geneticist Peter Lawrence shifted the focus of his career to the Drosophila model system, he did extensive research on Oncopeltus segmentation (Lawrence, 1966, 1973; Lawrence and Green, 1975). Through clonal analysis of irradiated embryos, he showed that patterning of segments begins at the blastoderm stage of Oncopeltus. Given the complexity of PR patterning in Drosophila and other holometabolous insects, we expect that, in addition to E75A, other genes with PR function remain to be discovered in Oncopeltus. It is therefore not a stretch to speculate that if Lawrence had continued his studies of segmentation in Oncopeltus, we might be referring to a whole different set of regulatory genes when thinking about PR patterning in insects. This example is one of many that underscores the importance of studying diverse experimental systems to understand biodiversity at the regulatory level; limiting our studies to a handful of organisms inevitably biases our understanding of developmental processes.

Oncopeltus, originally purchased from Carolina Biological and maintained in our lab for several years, were reared in 30×18×20 cm plastic cages on a diet of water and raw organic sunflower seeds at 25±1°C, 50% relative humidity with a photoperiod of 16 h light:8 h dark. At the beginning of a collection period, a piece of cotton on top of a section of paper towel was placed in the cage; the towel barrier ensured that no older embryos already in the cage would be collected. At the end of the collection period, the cotton was removed from the cage and kept at 25°C until embryos reached the appropriate age.

PRG orthologs were isolated using standard methods (see supplementary Materials and Methods for further details). Embryo fixation and in situ hybridization were carried out as described by Liu and Kaufman (2004) with modifications made by Ben-David and Chipman (2010) and by our lab (see supplementary Materials and Methods for further details). Antisense RNA probes were synthesized using digoxigenin or fluorescein RNA labeling mix (Roche) and templates generated by PCR amplification of staged cDNA with reverse primers containing the T7 promoter sequence. The T7 RNA transcription reaction was carried out at 37°C for 2 h, followed by precipitation in ice-cold ethanol and LiCl. In situ hybridization was carried out as previously described (Ben-David and Chipman, 2010 or Liu and Kaufman, 2009), with the addition of a 1 h 5% BSA blocking step prior to blocking with 10% sheep serum. BCIP/NBT was used most often as the chromogenic substrate for alkaline phosphatase; in the case of double in situs, BCIP/INT was used to produce an orange product. As our original probes were designed to hybridize to conserved domains, we designed additional probes for odd, prd, run and slp that were gene specific to ensure that our results were not obscured by cross-hybridization with related genes. Gene specificity was confirmed by conducting a BLAT search of the Oncopeltus genome using the probe sequence as the query; if only the sequence of the target gene matched the probe sequence, we considered the probe to be gene specific. We confirmed with odd, prd, run and slp that the gene-specific probes (Fig. S1, labeled probe 2 or 3) gave the same results as the less specific probes (Fig. S1, labeled probe 1).

Blastoderm-stage embryos were photographed in PBST. In blastoderm-stage embryos, the anterior pole is somewhat tapered; this morphological feature was used to orient embryos anterior-left. Otherwise, germbands were dissected out, transferred to glycerol and mounted on a slide. Imaging was carried out using an AxioCam MRc camera (Zeiss) mounted on a SteREO Discovery.V12 dissecting microscope (Zeiss) or Axio Imager.M1 compound microscope with DIC (Zeiss). When necessary, image stitching was performed in Fiji using the pairwise stitching plug-in (Preibisch et al., 2009) or Adobe Photoshop.

Double-stranded RNA for RNAi was synthesized using the Megascript T7 Transcription Kit (Invitrogen); dsRNA target regions are shown in blue in Fig. S1. The template for the RNA synthesis reaction was produced by PCR of cloned sequence fragments, using primers with the T7 promoter sequence at the 5′ ends. The RNA synthesis was allowed to proceed for 16 h at 37°C, then treated with TURBO DNase and annealed in a thermocycler. Each 20 μl reaction of double-stranded RNA was precipitated with 30 μl of 7.5 M lithium chloride and 250 μl ice-cold ethanol and then resuspended in injection buffer (5 mM KCl, 0.1 mM phosphate buffer pH 6.8). A 1:100 dilution of each dsRNA was run on an agarose gel to check that dsRNA was indeed present and the expected size.

Anticipating possible variability in effect between different dsRNAs, we tested three different concentrations per dsRNA (10, 15 and 20 μg), injecting two or three females with each concentration. Embryos were collected daily and the hatch rate was tracked for 3-4 weeks post-injection, or until injected females died (Fig. S10). One of the longest-living females, injected with 15 μg prd dsRNA, continued to lay clutches that failed to hatch 31 days after injection, a much longer penetrance than that reported previously in this species (Liu and Kaufman, 2004). The remaining experiments were performed with the lowest concentration of dsRNA that produced a noticeable effect on the hatch rate. For some dsRNAs (h, run, opa and ftz) no effect on hatch rate was observed even at the highest concentration tested. For these, additional experiments were performed with the highest concentration. Adult females were injected with dsRNA corresponding to each Of-PRG-ortholog about 1 week after molting from L4 to adult. Injections were carried out in triplicate, with 3-5 females per group for each gene. Double-stranded gfp RNA was injected as a control. One day after injection, an equal number of males was added to each cage. Embryos were collected daily and divided such that some were allowed to hatch, some were fixed at 48-72 h AEL for subsequent staining by in situ hybridization with an inv probe, and some were frozen in TRIzol at −80°C at 24-48 h AEL (odd, prd, slp, h, run, ftz and gfp) or 35-50 h AEL (opa and gfp). A tighter staging of opa RNAi embryos was necessary to ensure embryos used for RNA extraction were expressing opa. An additional dsRNA from a different part of the gene sequence was tested for prd, odd, slp, h, run and opa in a separate round of experiments (Fig. S1, red). Results were largely similar; defects seen for slp with this dsRNA were weaker; however, no defects were seen for Of-odd with this second dsRNA, which matched the odd 3′UTR. Thus, it is possible that defects shown above and by Auman and Chipman (2018) reflect knockdown of both Of-odd and Of-sob, which appear to be expressed in the same pattern (compare Fig. 3I-N to Fig. S6A-D).

qPCR was performed using 24-48 h AEL (prd, odd, slp, ftz, h and run) or 35-50 h AEL (opa) RNA from RNAi embryos. RNA was extracted using TRIzol (Invitrogen), DNase treated using the TURBO DNA-free kit (Invitrogen) and cDNA was transcribed using the iScript cDNA Synthesis kit (Bio-Rad). PCR was performed in a Roche LightCycler 480 real-time PCR machine, using the Luna Universal qPCR master mix (NEB). TATA-box binding protein (Tbp, found on scaffold 2359 of the Oncopeltus genome) was used as the reference gene as it was found to have the most stable expression of three candidate reference genes tested, and has been used as a reference gene in many other studies (Liang et al., 2014; Niu et al., 2014; Zhai et al., 2014). The 2^−ΔΔCT method was used to calculate fold change of gene expression relative to the gfp control. Statistical significance was determined by performing a Welch two-sample t-test in R with α=0.05. All primer sequences are listed in Table S2.

To conduct phylogenetic analyses of our candidate ortholog sequences, alignments of orthologous sequences were generated using MUSCLE (Edgar, 2004). Alignments were then trimmed using Aliview (Larsson, 2014). Finally, phylogenetic analysis was carried out in TOPALi v2.5 using the Bayesian algorithm MrBayes (Milne et al., 2009; Ronquist and Huelsenbeck, 2003). Additional formatting of trees was performed in MEGA6 (Tamura et al., 2013). See supplementary Materials and Methods for more details on determination of paralogous and orthologous gene relationships.

We thank Lakshmi Kirkire for keeping this project going after Y.L., Jie Xiang and Patricia Graham for technical advice and suggestions, and Jessica Hernandez for technical assistance.