How to align arthropod legs

How to align leg segments between the four groups of arthropods (insects, crustaceans, myriapods, and chelicerates) has tantalized generations of researchers, as this would answer over a century of speculation about the origins and homologies of the fascinating diversity of arthropod appendages and outgrowths. Here we compare the expression and loss-of-function phenotypes of leg patterning genes in crustaceans, insects, and arachnids using our own and previously published data. We find that all arthropod leg segments correspond to each other in a one-to-one fashion. This alignment suggests that chelicerates with seven leg segments incorporated a proximal leg segment into the body wall. In addition, this alignment suggests that insect and myriapod tracheae are convergent and homologous structures: each evolved via the independent internalization of an ancestral gill (respiratory exite) on the proximal-most leg segment of their shared ancestor. A framework for understanding the homologies of arthropod appendages opens up a powerful system for studying the origins of novel structures, the plasticity of morphogenetic fields across vast phylogenetic distances, and the convergent evolution of shared ancestral developmental fields.


Introduction 25
Arthropods are the most successful animals on the planet, in part due to the diversity of their appendages. How to align the legs of all arthropods has tantalized researchers for over a century [1][2][3][4][5][6][7][8] , as the solution would answer centuries of observation and speculation about arthropod structures. There are four groups of arthropods: chelicerates (spiders, etc.), myriapods (millipedes, etc.), crustaceans (shrimps, etc.), and insects (beetles, etc.). The difficulty in aligning 30 or homologizing arthropod leg segments is due to the different numbers, shapes, and names of leg segments. Chelicerates can have either 7 or 8 leg segments, myriapods have either 6 or 7, insects have 6, and crustacean have 7 or 8 leg segments (Fig. 1) 4, [8][9][10][11] . Since at least 1927, researchers have proposed many different theories to account for this variation, invoking leg segment deletions, duplications, and fusions to account for the different numbers of leg segments 35 between arthropod taxa (see for example 3,4,7,8 ). The incredible diversity of arthropod legs even contributed to some author's conclusions that the four arthropod groups arose independently, and therefore it is not possible to homologize and align their legs 12 . However, as molecular studies confirmed arthropod monophyly and mapped the topology of arthropod relationships with ever greater precision 13,14 , and as loss-of-function studies of leg patterning genes have been conducted 40 on more branches of the arthropod tree of life, this long sought model can now be brought to light.

Aligning the leg segments of crustaceans and insects
To align the leg segments of two arthropod groups, crustaceans and insects, Bruce and Patel 2020 15 compared the function of five leg patterning genes, Distalless (Dll), dachshund 45 (dac), Sp6-9, extradenticle (exd), and homothorax (hth), in the amphipod crustacean Parhyale hawaiensis to previously published results in insects. By aligning the leg segment deletion phenotypes for these five genes, they found that the six distal leg segments of Parhyale and insects (leg segments 1 -6, counting from the distal claw) aligned in a one-to-one fashion (Fig.   2). To align the proximal leg segments, they compared the expression of pannier (pnr) and the Iroquois complex gene araucan (ara) in Parhyale and insects 15 . They found that, in both Parhyale and insects, the expression of ara distinguishes two proximal leg segments (leg segments 7 and 8; Fig. 2), while expression of pnr marks the true body wall (tergum). These data suggested that insects had incorporated two ancestral proximal leg segments, 7 and 8, into the body wall 16 . This work demonstrated that crustacean and insect legs had 8 leg segments could be 55 homologized in a straightforward, one-to-one relationship. If insect and crustacean legs can be homologized, this model may extend to myriapods and chelicerates as well, in a generalizable model of appendages across all four groups of arthropods.

Aligning the leg segments of crustaceans, insects, and chelicerates 60
To align Parhyale, insect, and chelicerate legs, the leg segment deletion phenotypes in Parhyale and insects were compared to previously published results in chelicerates. Functional experiments in chelicerates have been performed for Dll, Sp6-9, dac, and hth. Based on the leg segment deletion phenotypes of these genes, the six distal leg segments of Parhyale, insects, and chelicerates (leg segments 1 -6, counting from the distal claw) can be aligned in a one-to-one 65 fashion, as follows.
To elucidate the composition of proximal leg segments in chelicerates, the expression of 85 pnr, ara, and Dll was compared between Parhyale, Tribolium, and the tarantula Acanthoscurria geniculata (Fig. 5) 15 . Three orthologs of pnr were identified in Acanthoscurria that had closest homology to Drosophila, Tribolium, and Parhyale pnr (Fig. S1). However, only one of these was expressed at the stages examined and was therefore presumed to be pnr. An Acanthoscurria Iroquois gene was identified which was the reciprocal best BLAST hit to Drosophila, Tribolium, 90 and Parhyale ara. An Acanthoscurria Dll gene was identified which was the reciprocal best BLAST hit to Drosophila, Tribolium, and Parhyale Dll.
As in Parhyale and Tribolium, Acanthoscurria Dll was found to be expressed in leg segment 1 -5, and pnr expressed in the most dorsal tissue (Fig. 5). Thus, it appears that pnr marks the "true" body wall in all arthropods. In Parhyale and Tribolium, ara is expressed in 95 three domains: a dorsal armband on proximal leg segment 8 that is adjacent to the pnr domain; a second armband on proximal leg segment 7; and a dot of expression on the medial side of leg segment 6. Parhyale also expresses ara in the tip of the claw . In Acanthoscurria, at the embryonic stages examined, ara is expressed in three of these domains: a dorsal armband adjacent to the pnr domain; a second armband on proximal leg segment 7; and some expression 100 in the tip of the claw. The dot of ara expression in leg segment 6 was not observed. Perhaps this domain is expressed at embryonic stages that were not examined, or it is not expressed in the Acanthoscurria lineage. However, as predicted by the leg segment alignment model, the two armbands of ara expression in Acanthoscurria bracket a region proximal to leg segment 7 (spider coxa) and adjacent to pnr. This suggests that Acanthoscurria, like Parhyale and 105 Tribolium, also retains an ancestral, proximal 8 th leg segment.
To test this hypothesis, the expression of odd-skipped was examined in Acanthoscurria.
In Drosophila, the odd-skipped family of genes is expressed in the distal edge of each leg segment, where it induces cells to buckle and form the flexible joint 33 . An odd-skipped gene was identified in Acanthoscurria which was the reciprocal best BLAST hit to Cupiennius (spider) 110 odd-related 3 (odd-r3) 34 . This Acanthoscurria odd-r3 is expressed in the distal region of leg segments 1 -7 but also in an additional ring proximal to leg segment 7 (Fig. 6). This additional ring of odd-r3 notably occurs in the distal side of a leg-segment-like bulge. Given that oddskipped is expressed in the distal side of leg segments 33 , the ring of odd-r3 expression on the distal side of the leg-segment-like bulge suggests that it is a bona fide leg segment. Together, the 115 expression of pnr, ara and odd-r3 and the presence of a leg-segment-like bulge suggest that Acanthoscurria has an additional proximal 8 th leg segment.

Aligning the leg segments of myriapods
No functional data for leg patterning genes is available for myriapods. However, 120 morphological and embryological evidence suggests that myriapods, like insects, have incorporated proximal leg segment(s) into the body wall 4,35,36 . In the embryos of both myriapods and insects, the proximal part of the developing leg ("subcoxa" in insects or "limb base" in myriapods) broadens and flattens to form the adult lateral body wall 36-45 . In insects, this subcoxa was shown to correspond to the two proximal-most leg segments of crustaceans 15 , and the same 125 may be true for myriapods 35 . Incorporation of proximal leg segments into the myriapod body wall would bring their leg segment count to 8, in agreement with other arthropods.

Discussion
The expression and embryological data shown here, in conjunction with the expression 130 and functional data from previous publications, demonstrates that all arthropod legs can be aligned in a one-to-one fashion (Fig. 7). For example, the coxa of spiders, millipedes, and crustaceans are leg segment 7; the insect coxa, crustacean basis, and spider trochanter are leg Myriapods with 6 or 7 leg segments incorporated 2 or 1 leg segments into the body wall, respectively.

Chelicerate legs 145
Chelicerates with 8 free leg segments, such as sea spiders and solfugids, have not incorporated any leg segments into the body wall. However, in chelicerates with 7 leg segments, such as Acanthoscurria, a proximal leg segment is missing and must be accounted for. One hypothesis is that it was simply deleted. However, the expression of Acanthoscurria pnr, ara, and odd-r3 presented here suggests that the proximal-most 8 th leg segment of these chelicerates 150 was incorporated into the body wall, similar to how insects incorporated proximal leg segments into their body wall 3,15,39,45 . Embryological evidence also supports this conclusion: a leg-like proximal 8th leg segment can be observed in embryos of the tarantula spider Acanthoscurria dorsal to the coxa, which appear to articulate to the coxa with condyle joints (Fig. S2), that may be the remnant of the proximal 8 th leg segment.

Myriapod legs
When the morphological and embryological evidence in myriapods is incorporated into the above leg segment alignment model, two fascinating hypotheses become apparent: insect and 160 myriapod respiratory systems may in fact be homologous; and insect wings may be homologous to myriapod "wings" (polydesmid paranota).
Myriapod and insect respiratory systems are astonishingly similar -a small circular spiracle associated with each leg leads to internally branching trachea 46 (Fig. S3). This contributed to entrenched support for their sister relationship 46 . Notably, both respiratory systems appear to occur on the proximal-most leg segment (leg segment 8) 4,15,36,38 . Thus, while these two systems are not homologous as tracheae 13 , the evidence suggests that insect and myriapod tracheae each evolved via the independent internalizing an ancestral gill (respiratory exite) on the proximal-most leg segment of their shared ancestor 47,48 .
In support of this, the genes trachealess (trh) and ventral veins lacking (vvl), which are 170 expressed in and required for insect tracheal formation, are also expressed in the crustacean gill 49,50 , an exite on the leg. This is expected if insect tracheae are an invaginated exite on the incorporated 8 th leg segment. If insect tracheae are an invaginated exite on the 8 th leg segment, then perhaps the morphologically and functionally similar myriapod tracheae are as well.
Some myriapods (polydesmid and platydesmid millepedes) have many wing-like 175 "paranota" or "paratergites" (Fig. S3) along the side of the body. If these emerge from the proximal-most leg segment, then they would be positionally homologous to the Parhyale tergal plate and insect wing. Thus, millipedes and insects may have convergently evolved tracheae and "wings" from the same exites on the same leg segment. These structures would therefore be bizarrely both convergent and homologous. 180 To test these two hypotheses, the expression of pnr, Iroquois genes like ara, joint-   In Parhyale and insects, loss of hth deletes the proximal leg segments, leaving only the distal 2 leg segments intact. In harvestman, reduction of hth shortens and fuses the proximal leg segments, leaving the distal segments unaffected 32 . It is not clear from the figures or text how many distal leg segments are unaffected -the most 390 severely affected embryos did not survive to hatching and their cuticle was shriveled, thus obscuring what deformities are due to loss of hth and what are due to the embryo not developing fully before hatching. However, leg segment 1 and at least the distal half of leg segment 2 are clearly unaffected. Thus, in harvestman, insects, and Parhyale, hth appears to function in all but the distal 2 leg segments. In spiders, harvestman, and Parhyale, a weak dac2 phenotype causes 395 green leg segment 4 to be truncated and fused onto cyan leg segment 3. In harvestman, Parhyale, and Drosophila, a strong dac2 phenotype affects leg segments 3 -5. prior to dorsal closure. (i) dissected leg of Acanthoscurria embryo. In all three arthropods, leg segments 1 through 5 are identified by Dll expression (magenta) 61 . In all three arthropods, the two ara (green) armband domains bracket a region proximal to leg segment 7. In all three arthropods, pnr (red) marks the most dorsal domain and is adjacent and partially overlapping the dorsal ara domain. In g, h, i, ara is expressed in a smattering of ventral non-leg cells. Tribolium 410 larvae have a fused tibia and tarsus, the tibiotarsus, here labelled 2-3 37 . In Acanthoscurria, leg segment 7 is easily identified by the coxal endite (ce) that bulges medially. In both Tribolium (g) and Parhyale (h) legs, a muscle expressing pnr and ara that extends the length of leg segments 7 and 8 was masked to clearly show the ectodermal domains. Tp, tergal plate. Cp, coxal plate. G, gill. In Acanthoscurria, auto-fluorescent yolk cells (yc); ventral expression domain of ara (*). X 415 in c indicates distal leg that was accidentally removed.

Fig. 6.
Expression of odd-r3 in Acanthoscurria. odd-r3 is expressed in the distal region of each leg segment where the joint will later form. a, b Stage 12.5 Acanthoscurria embryos dissected 420 away from yolk mass. Leg segment 7 is readily identified by the presence of a medial bulge, the coxal endite (ce). Proximal to leg segment 7, there is a leg-segment-like bulge (white curly brace), which expresses odd-r3 in the distal region. pnr expression in this late stage embryo is reduced and obscured by other colors. DAPI in b is shown in cyan to better observe morphology. c, d dissected walking leg 1 from slightly earlier embryo, Stage 11.5, where leg segment 425 divisions haven't yet bulged out. odd-r3 encircles the distal region of each leg segment, including the hypothesized proximal 8 th leg segment.

Fig. 7.
Model of how to align all arthropod legs. A. Schematic of which genes function is related to (specific) leg segments. B. Morphology and homologies of arthropod leg segments based on leg gene function in insects, Parhyale, and chelicerates. Colors and patterns indicate proposed homologies. Exites (checker pattern); endites (stripe pattern). Drawings in B modified from Snodgrass 1952. 435 Fig. S1. Three orthologs of pnr were identified in Acanthoscurria with closest homology to Drosophila, Tribolium, and Parhyale pnr (Fig. S1-2). However, only one of these was expressed at the stages examined, Acanth_DN78099, and was presumed to be pnr. Consensus tree generated using Mafft, which gave similar topology to Clustal consensus tree. 440