Evolution of a Novel Appendage Ground Plan in Water Striders Is Driven by Changes in the Hox Gene Ultrabithorax

Water striders, a group of semi-aquatic bugs adapted to life on the water surface, have evolved mid-legs (L2) that are long relative to their hind-legs (L3). This novel appendage ground plan is a derived feature among insects, where L2 function as oars and L3 as rudders. The Hox gene Ultrabithorax (Ubx) is known to increase appendage size in a variety of insects. Using gene expression and RNAi analysis, we discovered that Ubx is expressed in both L2 and L3, but Ubx functions to elongate L2 and to shorten L3 in the water strider Gerris buenoi. Therefore, within hemimetabolous insects, Ubx has evolved a new expression domain but maintained its ancestral elongating function in L2, whereas Ubx has maintained its ancestral expression domain but evolved a new shortening function in L3. These changes in Ubx expression and function may have been a key event in the evolution of the distinct appendage ground plan in water striders.


Introduction
The diverse appendage morphologies found in insects constitute an important model for studying the developmental genetic mechanisms underlying morphological novelties [1][2][3]. Water striders are derived semi-aquatic bugs (Hemiptera, Gerromorphae, Gerridae), which possess a remarkable diversity of leg lengths and shapes among species and between sexes. We have a good understanding of the evolutionary forces that shape this diversity, including both adaptation to locomotion on the water surface [4,5], and adaptations associated with mating [6,7]. The combination of a striking diversity and an understanding of the forces shaping this diversity suggest that water striders provide an important context for understanding the developmental genetic basis of appendage diversification. Yet, there have been no developmental genetic studies of this group. Here we investigate the mechanisms underlying the distinctive appendage size ground plan in water striders. In most insects, the hind-legs (L3) are longer than the mid-legs (L2) and forelegs (L1), representing an L3.L2.L1 appendage size ground plan ( Figure 1). Water striders have evolved a novel appendage plan where L2 are longer than L3 (L2.L3.L1; Figure 1). This ground plan has most likely evolved as a consequence of adapting to locomotion on the water surface [4]. L2 are disproportionately elongated and function as oars for propulsion, while L3 are shorter and function as rudders [7,8]. L1 are the shortest among the three pairs, functioning primarily in prey handling.
In insects, including other hemipterans (e.g., Oncopeltus fasciatus), appendages differentiate from limb buds that are specified and elongated during embryonic development [2]. The final phase of appendage development consists of refining the allometric properties of each pair according to its segmental position and biological function [9]. The Hox gene Ultrabithorax (Ubx) is known to play multiple roles in defining specific morphological differences among the segments along the anteriorposterior body axis in arthropods, including appendage size, shape, and function [10][11][12][13][14][15][16][17][18]. In several hemimetabolous insect species, the spatial and temporal expression of Ubx correlates with the relative enlargement of the hind legs L3 [19][20][21]. Ubx expression in L3 segments causes their differential growth compared to those of L1 and L2, and the earlier Ubx is expressed in these segments, the more enlarged they become [21]. Furthermore, Ubx in Drosophila is expressed in both L2 and L3 during larval and pupal development, where it is required for establishing different patterns of trichome features within the femur [15,22]. Ubx is also required for elongating the size of these legs as the loss of Ubx function causes a significant decrease in the size of L3 and a subtle decrease in the size of L2 [16]. We therefore tested whether Ubx plays a role in regulating the relative sizes of L2 and L3 legs in G. buenoi, and thus in the evolution of this derived ground plan within the Gerridae.

Results/Discussion
Differential leg sizes are established during embryonic development in Gerris buenoi Size differences between the three pairs of legs are established during embryogenesis and can be visualized in late embryos (Figure 2A). At this stage, L2 is over one and a half times longer than L3 ( Figure 2D). L2 and L3, due to their excessive length, extend in a stereotypic pattern along the body axes of late embryos ( Figure 2A). L2 extend from the ventral towards the dorsal side (ventral-to-dorsal arrangement), whereas L3 extend from one lateral to the opposing lateral side of the embryo (lateral-to-lateral arrangement; arrows in Figure 2A). In first instar larvae ( Figure 2B), the difference in size between L2 and L3 legs is comparable to the difference found between these two appendages in adults ( Figure 2C).
Insect appendages are generally subdivided into five segments, from proximal to distal: coxa, trochanter, femur, tibia, and tarsus. Elongation of the three distal leg segments, the femur, tibia, and tarsus, starts during G. buenoi embryogenesis and continues throughout subsequent developmental stages ( Figure 2). In late embryos, the tibia and the tarsus of L2 are not significantly different in size, but both are significantly longer than the femur ( Figure 2D). In contrast, the tibia in L3 is slightly but significantly shorter than the tarsus, and both tibia and tarsus are significantly shorter than the femur ( Figure 3D). Together, these results show that the novel appendage size ground plan L2.L3.L1 is established early during Figure 1. Ground plan of appendage morphology in insects. The common ancestor of Hemiptera most likely presents a universal insect ground plan where L3 is longer than L2, which is in turn longer than L1. The Lygaeoidae, which are terrestrial bugs such as Oncopeltus, represent the ancestral ground plan. Most semi-aquatic bugs (Gerromorpha) including the Veliidae, such as Ocellovelia, share a similar ground plan, which is associated with a mode of locomotion on water by alternating leg movements, similar to the terrestrial mode of locomotion. The Gerridae, such as Gerris, and some Veliidae, such as Husseyella, have evolved a derived ground plan where L2 is longer than L3, which is in turn longer than L1. This ground plan is an adaptation to a derived mode of locomotion on the water surface by means of oars (L2) and rudders (L3). doi:10.1371/journal.pgen.1000583.g001

Author Summary
Water striders are derived semi-aquatic bugs that possess a remarkable diversity of leg lengths and shapes among species and between sexes, and the selective forces shaping this diversity are well studied. The transition to living on the water surface was accompanied by dramatic changes in the size and function of their legs. The mid-legs are disproportionately long and function as oars, whereas the hind-legs are shorter and function as rudders. We present evidence demonstrating that changes in the pattern of expression and function of the Hox gene Ultrabithorax are responsible for establishing the relative size differences between mid-and hind-legs in the water strider Gerris buenoi. These changes in Ubx expression and function may have been a key event in the evolution of the distinct appendage ground plan in water striders. . Data are expressed as means6SD. Note that the Tarsus (Ta), Tibia (Ti), and Femur (Fe) in L2 an L3 are the segments that are the most elongated. The average size variation is significantly different between the three appendages (F 2,27 = 5866.264, P,0.05). The dynamics of growth of leg segments changes throughout ontogenesis (not shown). The tarsus and tibia of L2 are not significantly different in size (F 1, 18 = 0.367, P = 0.552) and are both significantly longer than the femur (F 1, 18 = 2526.094, P,0.01). In the adult, the femur of L2 is significantly longer than the tibia (F 1, 18 = 15.533, P,0.025; Figure 2D), which is in turn significantly longer than the tarsus (F 1, 18 = 50.522, P,0.001; Figure 2D). In L3 however, the tarsus is significantly longer than the tibia and significantly shorter than the femur throughout G. buenoi ontogenesis (F 2, 27 = 55.308, P,0.01). doi:10.1371/journal.pgen.1000583.g002 G. buenoi development, and has evolved mostly through modifications of the three distal segments in L2 an L3.
Ubx is expressed in both L2 and L3 in differential and dynamic patterns The expression pattern of Ubx mRNA or both Ubx and Abdominal-A proteins (UbdA) in trunk segments of G. buenoi is conserved relative to other insects [17,[23][24][25]. The domain of Ubx expression expands anteriorly from abdominal segment A1 to the posterior of the second thoracic segment T2, whereas Abd-A is restricted to abdominal segments A2-A8. UbdA staining first appears in the abdominal segments only, strongly in A1 and weakly in A2 through A8 of early embryos ( Figure 3A and 3B; Figure S1). Later in development, we observed a faint UbdA staining in the third thoracic segment T3 ( Figure   In addition to the ancestral expression of Ubx in the hind legs L3 of hemimetabolous insects [19,21,25], we discovered a novel expansion of Ubx expression to L2 legs. The antibody detects Ubx in both L2 and L3 throughout embryonic development ( Figure 3). This result suggests that Ubx may play a role in regulating the relative sizes of both appendages, and thus, may be responsible for the development of the derived appendage ground plan of water striders. Previous studies have shown a correlation between the spatial/temporal expression of Ubx and the enlargement of L3 in other insects [16,20,21]. We therefore expected a similar correlation between Ubx expression and the relative sizes of both L2 and L3, as well as the size of segments within these two legs in G. buenoi. However, we found Ubx spatial/temporal expression pattern to be consistent with some but not all aspects of leg size development within and between L2 and L3. Furthermore, we found that the levels of Ubx expression between L2 and L3 are not consistent with the relative sizes of these two legs. Between L2 and L3, Ubx expression is first detected uniformly throughout L2 but not in the distal part of L3 limb buds ( Figure 3A, Figure 3B1 and 3B2), which is consistent with the larger size of L2 relative to L3. This suggests that the earlier timing of Ubx accumulation contributes to elongating L2 size relative to L3. However when Ubx appears in L3, its levels are greater than in L2 throughout the remainder of development ( Figure 3C-3G), even though L2 is one and a half times longer than L3 ( Figure 2D).
Within L2, the spatial/temporal expression of Ubx is not consistent with the size differences among the three distal segments. Ubx expression appears simultaneously and uniformly in all these three distal segments ( Figure 3B and 3B9), even though the tarsus and the tibia are both longer than the femur ( Figure 2D). Interestingly, L2 legs curve slightly in the embryo (curved arrow in Figure 3F and 3G) most likely through faster growth in the anterior compared to the posterior compartment. This differential growth in L2 compartments is likely to direct this leg to wrap around the embryo in a ventral-to-dorsal arrangement (Figure 2A). We uncovered higher levels of Ubx accumulation in the posterior relative to the anterior compartment of L2 legs throughout most of embryo development ( Figure 3C-3F). Therefore, Ubx may also be involved either early in the differential growth of the two L2 compartments to direct the curving, or later in correcting growth differences between the two compartments so that L2 legs become straight at the end of embryo development.
Within L3, Ubx first accumulates in a strong stripe corresponding to the future tibia, and a faint stripe corresponding to the future femur ( Figure 3C). It is only shortly after this developmental stage that a stripe corresponding the future tarsus appears ( Figure 3D). This timing of Ubx expression in L3 is not consistent with the size differences between these segments, as the tibia is the shortest in this leg compared to both tarsus and femur. The three stripes of Ubx expression in L3 expand and become uniform in the whole leg including the two proximal segments, coxa and trochanter ( Figure 3E and 3F), even though these two proximal segments are not elongated. Furthermore, the levels of Ubx accumulation within L3 are higher in the tibia and in the tarsus, but lower in the femur ( Figure 3C and 3D), despite the tarsus and tibia being significantly smaller than the femur in both late embryos and adults ( Figure 2D). Altogether, these results suggest that Ubx may specify the development of G. buenoi appendage ground plan by playing multiple and distinct roles in regulating the size differences between L2 and L3, and those of the segments within each of these legs.
Ubx functions to elongate L2 and to shorten L3 in G. buenoi To determine the function of Ubx in the water strider appendage ground plan, we used parental RNA interference. We injected adult females with Ubx double stranded RNA (ds-Ubx) to knockdown Ubx expression, and separately injected females with yellow fluorescent protein double stranded RNA (ds-YFP) as a negative control (Text S1 and Figure S1). Surprisingly, Ubx knockdown resulted in shorter L2 but longer L3 compared to their control counterparts.
In L2, Ubx depletion results in a 20% shortening relative to the control embryos (F 1, 18 = 205.326, P,0.001). The sizes of the tarsi, tibias, and femurs within L2 are significantly reduced (F 1,18 = 78.093, P,0.001; Figure 4G), although the tarsi and the tibias are affected to a much greater extent than the femur. Furthermore, L2 legs in ds-Ubx first instar larvae are curved towards the posterior ( Figure 4F). This result may be accounted for by the higher levels of Ubx accumulation we observed in the posterior relative to the anterior compartment of this leg ( Figure 3C-3F). During embryonic development, as L2 legs are elongating they curve dorsally and wrap around the embryo in a stereotypic ventral-to-dorsal arrangement (Figure 2A). This dorsal curvature may occur through faster growth of the anterior relative to the posterior compartment as L2 elongates (curved arrow in Figure 3F and 3G). After L2 has curved towards the dorsal side of the embryo and before the nymph emerges, L2 must straighten while it continues to elongate. The processes of straightening and elongation of L2 may be affected by the stronger Ubx expression we observed in the posterior, causing this compartment to now grow faster relative to the anterior. Thus, in the absence of Ubx, the growth difference between the two L2 compartments is not corrected resulting in L2 that are curved instead of straight ( Figure 4F), and that are shorter relative to the controls ( Figure 4D). Therefore, Ubx is necessary for straightening the initial curvature as well as elongating the size of L2 legs.
In contrast to our results for L2, we observed a 22% elongation of L3 in ds-Ubx embryos compared to control embryos (F 1, 18 = 865.620, P,0.001). Within L3, the tarsus and the tibia are the most elongated relative to the femur in ds-Ubx animals ( Figure 4G), and these are the two segments with the higher levels of Ubx expression in wild type ( Figure 3C and 3D). Therefore, Ubx functions to reduce the size of L3, primarily through reducing the sizes of the tibia and the tarsus ( Figure 4G). This is opposite to the function of Ubx in L2.
Unlike in the second thoracic segment, where Ubx primarily elongates the size of L2, Ubx plays multiple and specific functions to establish the distinct identity of the third thoracic segment, including the reduction of L3 size (Figure 4; Text S1 and Figure  S2). These functions include: (1) establishing morphological differences of the spiracles found on the base of L2 and L3, because both legs in ds-Ubx animals now exhibit a spiracle at their base that is characteristic to wild type L2 legs only ( Figure 4E and Figure S2D); (2) directing L3 to wrap in a lateral-to-lateral arrangement, because in ds-Ubx embryos both L3 and L2 legs are now arranged ventral-to-dorsal as in wild type L2 but not L3 legs ( Figure 4D and Figure S2C, S2D); and (3) orienting the L3 femurs to point posteriorly, because L3 femurs in live ds-Ubx first instar larvae now point anteriorly, which is the posture for L2 legs in wild type (compare arrow directions in Figure 4C and 4F).

Conclusions
Our results show that Ubx establishes the appendage ground plan in water striders through elongating L2, and through multiple functions that establish the identity of the third thoracic segment, including the shortening of L3. In other insects that present the common L3.L2.L1 appendage ground plan such as O. fasciatus and Acheta domesticus, Ubx is expressed in L3 only and functions to elongate its size [20]. Therefore in the novel L2.L3.L1 ground plan of water striders, Ubx has evolved a new expression domain but maintained its ancestral elongating function in L2, whereas in L3, Ubx has maintained its ancestral expression domain but evolved a new shortening function.
The gain of Ubx expression in L2 may have evolved through the ability of Ubx to respond to one or several upstream regulators in this leg. One possibility to explain the evolution of this pattern may be that L2 has gained the expression of a factor(s) that can either promote the activation or alleviate the repression of Ubx in this leg [26,27]. Another possibility may be that Ubx cis-regulatory elements have been modified such that these elements are now accessible for an activator or inaccessible for a repressor in L2 [27][28][29][30]. When these changes in Ubx expression began to randomly appear, they resulted in longer L2, which may have enhanced or promoted the specialized function of this pair as oars.
In addition to elongating L2, the evolution of the L2.L3.L1 appendage ground plan may have involved a shortening of L3 relative to L2. This is supported by our observation that in G. buenoi the function of Ubx is to shorten L3 legs. One possibility may be that Ubx in L3 has gained new downstream targets [18,31], or may have gained the ability to associate with different co-factors [14,32,33], so that the outcome of these interactions is opposite to that in L2, resulting in the development of shorter legs. Another possibility could be that the differential timing and levels of Ubx expression [13,21,34] in L3 relative to L2 may be responsible for this opposing function of Ubx in regulating the size of these two legs. As in L2, this new Ubx function in shortening L3 may have enhanced or promoted the function of L3 as rudders.
Although most insects possess the common L3.L2.L1 plan, multiple independent transitions to L2.L3.L1 appear to have occurred in the Veliidae, a group that is basal to the Gerridae [4,35] ( Figure 1). These transitions may have involved repeated parallel changes in the expression and function of Ubx to elongate L2 and shorten L3. It is only in the derived Gerridae where the L2.L3.L1 plan becomes a general feature, which may suggest that this ground plan has been a key adaptation during the evolution of this group [4,35]. Future studies on semi-aquatic bugs should focus on whether transitions to the derived appendage ground plan are generally associated with parallel changes in Ubx expression and function, as well as on elucidating the molecular mechanisms underlying the opposite effect of Ubx in L2 compared to L3.

Materials and Methods
Animals G. buenoi females were collected from a pond in Toronto, Ontario, Canada. Water striders were reared in water tanks and fed with live Drosophila. Females lay eggs that are glued length-wise to floating pieces of Styrofoam by a gelatinous substance, which swells in water [36]. These eggs can be incubated at 25-27uC to allow embryo development.

Ubx cloning
Total RNA was extracted from late embryos and early prolarvae, and used as template in a first strand complementary DNA synthesis reaction (Invitrogen). This first strand cDNA pool was used as a degenerate PCR template to amplify a 500 base pairs (bp) G. buenoi Ubx fragment. The following forward and reverse Ubx degenerated primers were synthesized based on an Ubx sequence alignment from closely related insects: Forward: 59-TAYGCCGC KGTKGTGGCAGCCGC-39 from HQNGYAAV and reverse: 59-TTCATKCGY CGGTTTTGGAACC -39 from WFQNRRMK. G. buenoi Ubx sequence can be retrieved in Genbank under the accession number: FJ460166.

Leg measurements and statistical analyses
All measurements were performed on a sample size of 10 animals (i.e., n = 10) using a Zeiss dissecting scope and Axiovision software. We first performed a model II single classification analysis of variance (ANOVA; [37]) to determine whether or not there is a statistically significant difference between the mean sizes of (1) L1, L2 and L3; and between (2) tarsus, tibia and femur of L1; tarsus, tibia and femur of L2, and tarsus, tibia and femur of L3; and finally between (3) the tarsi between all three legs; the tibias between all three legs; and the femurs between all three legs. We then performed a two-way mixed model ANOVA [37] to determine whether or not there is a statistically significant difference between the mean sizes of (1) L1, L2 and L3 in control and ds-Ubx treatment; between (2) tarsi of L1, L2 and L3 in control and ds-Ubx treatment; between (3) tibias of L1, L2 and L3 in control and ds-Ubx treatment, and finally between (4) femurs of L1, L2 and L3 in control and ds-Ubx treatment. All statistical analyses were performed using SPSS (SPSS Inc).

Ubx in situ hybridization
Ubx in situ hybridization is performed using a DIG-labeled Ubx anti-sense RNA probe (Roche). Fixed embryos are rehydrated in decreasing methanol concentrations in PTW then washed several times in PBT (16 PBS; 0.3% Triton). The remnants of chorions are removed, manually using fine forceps, followed by a 2 min Proteinase K (50 mg/ml) treatment in PTW. Embryos are washed two times in PBT then incubated 2 min in 2 mg/ml Glycine in PBT to inactivate Proteinase K. Embryos are washed several times in PBT, post-fixed in 4% formaldehyde for 20 min, then washed several times again in PBT. Embryos are pre-hybridized one hour in hybridization solution (50% formamide, 56 SSC pH 6.5, 50 ug/ml salmon sperm DNA, 50 mg/ml Heparin and 0.1% Tween-20) at 58uC, then hybridized over-night with Ubx probe in 100 ml hybridization solution at 58uC. Embryos are washed in decreasing concentrations of hybridization solution in PBT at 58uC, then five times in PBT at room temperature followed by a one hour blocking step in PAT (16 PBS; 1% Triton-X 100 and 1% bovine serum albumin). Embryos are then incubated two hours at room temperature with the anti-DIG antibody, conjugated with Alkaline Phosphatase. Ubx expression is revealed using NBT/BCIP as substrates for the Alkaline Phosphatase.

Ubx/Abdominal-A antibody staining
Ubx/Abdominal-A antibody staining is performed using the FP6.87 antibody that recognizes both Ubx and Abdominal-A proteins [19]. Fixed embryos are manually dissected to remove the remnants of the chorion, and then incubated in a blocking solution (16 PBS; 0.1% Triton-X 100; 0.1% BSA and 10% Normal Goat Serum) for one hour at room temperature. Blocked embryos are then incubated with the FP6.87 antibody at 1/5 dilution overnight at 4uC. Embryos are then washed five times in blocking solution then incubated with an anti-mouse secondary antibody conjugated to horse radish peroxydase for two hours at room temperature. Embryos are washed five times in PBT, then UbdA expression revealed in a reaction solution containing DAB, NiCl2 and 0.002% H2O2.

Ubx parental RNAi
The synthesis of Ubx double stranded RNA (ds-Ubx) was performed as described in [38]. The following Ubx forward 59-TAATACGACTCACTATAGGGAGACCACGTggcagccgcatgta agctatatt -39 and reverse 59-TAATACGACTCACTATAGG GAGACCACGTttggtacctcgtatatgtttgtc-39 primers containing T7 promoter sequence (capital letters) were used to clone a fragment that is flanked by T7 promoter from each side. This fragment was used as a template for in vitro transcription using T7 RNA polymerase, generating both sense and anti-sense transcripts, at 37uC. Complementary single RNA strands are automatically annealed into double stranded RNA (dsRNA) while the reaction progresses without any further treatment. dsRNA is then purified using Qiagen RNeasy purification kit and eluted in Spradling injection buffer [39]. G. buenoi adult females were anesthetized using carbon dioxide, immobilized on double sticky tape and injected with 1.5-2 ml Ubx dsRNA or YFP dsRNA (as negative control) at ,2 mg/ml concentration. Injected females were replaced on water tanks; embryos collected on floating Styrofoam and allowed to develop at room temperature. Embryos were screened for phenotypes morphologically by examining leg sizes and segment transformation. Figure S1 Specificity and phenotype frequency of Ubx RNAi in G. buenoi. (A) Embryo from a female injected with YFP doublestranded RNA as a control, stained for Ubx/Abd-A proteins. At this stage, Ubx is expressed in the hind-legs and strongly in the boundary between T3 and A1 (arrowhead), whereas Abd-A is expressed in the abdominal segments A2-A7. (B) Embryo of a similar developmental stage from a female injected with Ubx double-stranded RNA, also stained for Ubx/Abd-A. Note that AbdA expression persists in the abdominal segments, while Ubx is no longer expressed neither in the boundary between A1 and T3 (arrowhead) nor in the hind-legs. This suggests that our Ubx dsRNA is highly specific and does not interfere with AbdA expression. (C) Ubx RNAi phenotype count based on homeotic defects observed in the trunk segments and their corresponding legs of late embryos and early emerged larvae. G. buenoi RNAi efficiency was higher than 90%, while no Ubx-specific phenotypes were found in the YFP control. (B) Lateral view of the same embryo, showing spiracle that characterizes the base of L2 (arrowhead) as well as the ventral-todorsal stereotypic arrangement of L2 and the lateral-to-lateral arrangement of L3 appendages (arrows). Note that segment T3 also possesses a spiracle, which is smaller and distinguishable from that of L2. (C) Dorsal view of an Ubx-depleted embryo, showing an increase in the length of both T3 and A1 segments. (D) Lateral view of the same embryo, showing that both segments T3 and A1 now have developed the same T2-specific spiracle (arrowheads). Note in both (C) and (D) that L3 appendage now adopts a ventralto-dorsal arrangement characteristic to L2. (E) Comparison of the lengths of trunk segments T2, T3 and A1 between control and Ubx RNAi embryos. Both segments T3 and A1 exhibit a dramatic increase in length in Ubx-depleted embryos suggesting that these segments now exhibit morphological features that resemble T2.