The Daphnia carapace and the origin of novel structures

Understanding how novel structures arise is a central question in evolution. Novel structures are often defined as structures that are not derived from (homologous to) any structure in the ancestor1. The carapace of the water flea Daphnia magna is a bivalved “cape” of exoskeleton that has been proposed to be one of many novel arthropod structures that arose through repeated co-option of genes that also pattern insect wings2–4. To determine whether the Daphnia carapace is a novel structure, we compare the expression of pannier, araucan, and vestigial between Daphnia, Parhyale, and Tribolium. Our results suggest that the Daphnia carapace did not arise by co-option, but instead derives from an exite (lateral lobe) that emerges from an ancestral proximal leg segment that was incorporated into the Daphnia body wall. The Daphnia carapace therefore appears to be homologous to the Parhyale tergal plate and the insect wing5. Remarkably, the vestigial-positive region that gives rise to the Daphnia carapace appears to be present in Parhyale6 and Tribolium as a small, inconspicuous protrusion. Similarly, the vestigial-positive regions that form thoracic tergal plates in Parhyale appear to be present in Daphnia, even though Daphnia does not form thoracic tergal plates. Thus, rather than a novel structure resulting from gene co-option, the Daphnia carapace appears to have arisen from a shared, ancestral tissue (morphogenetic field) that persists in a cryptic state in other arthropod lineages. Cryptic persistence of unrecognized serial homologs may thus be a general solution for the origin of novel structures.

that the Daphnia carapace emerges from the dorsal posterior region of the head and, like insect 35 wings, is composed of a bilayered sheet of ectodermal cells and expresses and requires the "wing" genes wingless (wg), vestigial (vg) and scalloped (sd) 2 . They propose that the Daphnia carapace and other flat, lateral lobes in arthropods arose by multiple instances of co-option of wing patterning genes.
However, the analyses in Bruce and Patel 2020 5 and Bruce and Patel 2021 9 suggest an 40 alternative hypothesis. These analyses, drawing on over a century of morphological and embryological studies as well as gene expression and loss-of-function studies, generated a model for understanding the homologies of arthropod leg and body wall structures (Fig. 1). In this model, most arthropods have incorporated one or two ancestral proximal leg segments into the body wall 5,[9][10][11][12][13] (leg segments 7 and 8, counting from the terminal claw), but the division 45 between "true" body wall (tergum) and the incorporated leg segments that now function as lateral body wall can be distinguished by the expression of pannier (pnr) and Iroquois complex genes such as araucan (ara), respectively. In the embryos of Drosophila (fruit fly), Tribolium (flour beetle) Parhyale (amphipod crustacean), and Acanthoscurria (tarantula), ara expression brackets the hypothesized incorporated 8 th leg segment, while pnr is expressed in the dorsal-most 50 tissue and marks the true body wall 9 . Bruce and Patel 2020 5 showed that the insect wing and Parhyale tergal plate are both derived from exites -multi-functional lobes that emerge from proximal leg segments and that are patterned by "wing" genes such as vg and sd 6,14,15 . They showed that the insect wing and Parhyale tergal plate emerge from the ancestral leg segment 8 that was incorporated into the body wall (Fig. 1B). 55 Based on this model, the morphological and molecular data in Shiga 2017 suggest that the Daphnia carapace did not arise by co-option, but instead represents an exite on an incorporated 8 th leg segment of the head. The Daphnia carapace would therefore be homologous to the Parhyale tergal plate 5,6 and the insect wing 5 .
To test the proximal-distal register of the Daphnia carapace, the expression of pannier, 60 an Iroquois gene, and the wing/exite patterning gene vg was examined in embryos of Daphnia magna, Tribolium castaneum, and Parhyale hawaiensis using in situ hybridization chain reaction (HCR) version 3.0 16,17 . A single pnr gene was identified in Daphnia which was the reciprocal best blast hit of Drosophila, Tribolium, and Parhyale pnr 5 . A single Iroquois complex gene was identified in Daphnia which was the reciprocal best blast hit of Drosophila, Tribolium, and Parhyale ara. This Daphnia gene is hereafter referred to as ara 5 . Daphnia vg was identified previously by Shiga 2017 2 .
If the Daphnia carapace is the exite of the incorporated 8 th leg segment, then our model predicts that pnr will be expressed in a narrow stripe dorsal to the carapace and the ara domain adjacent to pnr will extend into the carapace. Alternatively, if the carapace is a dorsal, non-leg-70 derived structure, then pnr expression should extend into the carapace, and the two domains of ara will be located ventral to the carapace. In either case, vg will be expressed along the edge of the carapace 2 .
Consistent with the hypothesis that the carapace is an exite on leg segment 8, Daphnia vg is expressed along the edge of the carapace, pnr is restricted to a narrow, dorsal stripe above the 75 carapace, and the ara domain adjacent to pnr extends into the carapace (Figs. 2 and 3).
Interestingly, ara expression on leg segment 6 appears to mark the position of the exopod. In a uniramous Parhyale leg (Figs. 3e) that lacks an exopod, the leg segment 6 ara domain forms a circular patch on the lateral side of the leg. However, in the biramous (split) legs of Daphnia (Figs. 2d and 3f, S1) and Parhyale (Fig. S2) where an exopod emerges on the lateral side of leg 80 segment 6, the ara patch is expanded, and in Parhyale, the patch encircles the base of the exopod (see Fig. S1 for explanation of Daphnia leg identification).
If the Daphnia carapace is the exite of the incorporated 8 th leg segment of a mouthpart (modified leg) on the head, this exite may still exist on the head appendages of arthropods that do not form a carapace. In support of this hypothesis, vg is expressed in the head of Tribolium 85 dorsal/proximal to the mouthparts (Fig. S3A). This vg domain is bracketed by ara expression, just like the insect wing, the Parhyale tergal plate, and the Daphnia carapace. This region is therefore presumably homologous to the 8 th leg segment. Notably, there is no obvious structure associated with the mouthpart vg domain. In Parhyale, vg patterns the flange-like protrusion that protects the mouthparts, because the flange is reduced when vg is knocked out (Fig. S3B, C 6 ). 90 This flange emerges from the incorporated 8 th leg segment, because it is bracketed by ara expression. Given that the arthropod head is composed of several modified legs fused into a head capsule 18 , the head flange likely represents several adjacent exites. Thus, rather than new, co- Rather than de novo co-options, these morphogenetic fields are always there, persisting in a dormant, truncated, or highly modified state, and de-repressed in various lineages to form 155 apparently novel structures. Cryptic persistence of morphogenetic fields may therefore provide a mechanistically satisfying explanation for the origin of novel structures.
In fact, cryptic persistence is hinted at in the definition of crustacean exites. In crustacean morphology, exites are defined as lateral lobes that may occur on any of the three proximal leg segments (leg segments 6 -8, "protopod") and on any post-mandibular leg along the anterior-160 posterior axis 14,19 . However, it is widely appreciated that not all possible exites occur on all proximal leg segments on all legs in all crustacean taxa. Exites occur variably across different crustacean taxa: they may appear in some lineages but not in others, on some legs and not others, and on some leg segments but not others 14,19,32 . These observations imply that there is some latent ability to form a structure at these locations: sometimes the pathway is activated and 165 sometimes it is repressed, yet all these structures are considered exites, rather than repeatedly novel structures. The concept of cryptic persistence makes this implicit assumption explicit, and furthermore provides a mechanistic, molecular developmental explanation for the variable occurrence of exites.
For decades, researchers have been finding that many exciting structures on the legs and 170 body wall of many insects express the same genes (the "wing gene network"). These were interpreted to be wing serial homologs or partial wing homologs, and have often been interpreted as evidence of gene co-option. This has contributed to a widespread narrative that co-option is a common mechanism for the generation of novel structures. However, given that crustacean exites may occur on any body segment 14,32 , and that insects have incorporated two leg segments 175 into the body wall, each with an exite 5 , and all of these insect structures may be exites inherited from their crustacean ancestors. Thus, the lack of a molecular, developmental model from the ancestral crustacean group gave the appearance that insect structures had repeatedly arisen de novo, and spurring a mistaken foundation for one of the most widely discussed notions in evo devo. This is a clarion call for a diversity of research model organisms in every major clade, so 180 that structures no longer appear to be novel simply because the ancestral state has not yet been investigated.
In addition to the phylogenetically unequal distribution of research organisms, a second reason for the dominance of the gene co-option model in explaining the origin of novel structures may be the poor naming of "master regulator". In many discussions surrounding the origin of 185 novel structures, authors propose that a novel structure came about when an entire genetic pathway was co-opted 2,21 . The underlying assumption seems to be that this structure and its genetic pathway was not present at that location before, and that the tissue doing the co-opting is unrelated to the tissue it is co-opting from, i.e. they are different morphogenetic fields, such as eye vs heart. The related assumption seems to be that it is relatively straightforward for a master 190 regulator to become misexpressed in an unrelated tissue (i.e. outside of that master regulator's morphogenetic field at serially homologous locations, i.e. a tissue not previously primed with dozens of other TFs) in order to activate the entire structure in an entirely novel, non-serially homologous location. If this were true, we would expect misexpression of the heart gene tinman in the late ectoderm to form an ectopic heart or misexpression of eyeless in the late gut to form 195 an eye. One might posit that such outcomes would not be expected, because the identities of different germ layers are extensively restricted at multiple levels (chromatin accessibility, which suites of transcription factors and their cofactors are expressed, etc). But the same is probably true for morphogenetic fields within the same germ layer, for example leg vs true body wall (tergum): each has been extensively programmed and restricted such that misexpression of even 200 a "master" regulatory gene from the other's pathway would not have an effect, because it would not be able to talk to or interact with the other pathway.
A better model for the origin of apparently novel structures may be the Hox genes, where serially homologous tissues may switch between alternate identities. In this model, misexpression of "master regulators" or selector genes at previously unappreciated serially 205 homologous locations would be expected to transform tissue towards the structures it patterns.
Unique phenotypes may arise due to the selector gene being expressed in a different Hox or proximal-distal regulatory environment, or because the structures have not been expressed at this body position for millions of years and will thus communicate differently with the genetic pathways at this position. 210 Criteria should be formed to distinguish structures that have been de-repressed at a serially homologous location. For example, de-repressed serial homologs would be expected to appear in the same developmental register, to be composed of the same germ layer and similar tissue type (which in itself may be difficult to determine with certainty), and to express most of the genes in the pathway. In contrast to whole pathway co-option, single gene co-option may be 215 plausible, and would be expected to take genes piece-meal from different pathways, like butterfly eyespots, rather than entire pathways, which would require co-opting multiple regulatory genes simultaneously.
Many lateral, leg-associated structures across arthropods may be exites. Examples include the lateral organs of arachnids, glands of pycnogonid larvae, trachea of myriapods and 220 insects, paranota of polydesmid millipedes, abdominal gills of insects, gin traps of insects, thoracic styli of jumping bristletails, scarab beetle support structures, and oostegites of amphipod crustaceans. Rather than being repeated co-options of "wing" genes, or "partial wing homologs", these structures may simply be different kinds of exites, and the observed differences in gene expression profiles would be due to differences in the type and therefore function of the exite 225 (i.e. respiratory gill/trachea, secretory gland, rigid plate, etc).
Morphogenetic fields are related to vestigial structures in the sense that vestigial structures are generated from morphogenetic fields. However, a morphogenetic field can exist in a dormant, repressed state where it does not form a structure at all, vestigial or otherwise, and yet it is still capable of forming a structure in the future if it becomes de-repressed. Cryptic 230 persistence of morphogenetic fields therefore reframes the discussion to focus on the molecular networks that generate structures (or the Character Identity Networks of Wagner 2007 34 ). The concept therefore encompasses vestigial structures, when the field forms a structure, but expands the idea by pointing out that not all fields generate an actual structure, and also explains why structures can re-appear at serially homologous locations. Morphogenetic fields occur at serially 235 repeated locations, and not all of these locations will form a structure, vestigial or otherwise. But these locations should be identified and examined in order to discover dormant fields so that homologies with structures of interest may be established. Identifying all relevant serially homologous morphogenetic fields within an organism is the only way to know for sure whether a structure is expressed at a novel or a serially homologous position. 240 The work presented here calls into question whether gene co-option should be invoked to explain the origin of novel structures as often as it is. On a final note, it is interesting to consider that if cryptic persistence is commonplace, and most "novel" structures evolved from existing structures, then many familiar morphogenetic fields (for example, legs) may be far more evolvable, and have far more ancient origins, than currently thought. 245

Declaration of interests
The authors declare no competing interests.  carapace and the ara domain adjacent to pnr will extend into the carapace. If the carapace is an outgrowth of the body wall (c'), then pnr will extend into the carapace, and the two domains of ara will be located ventral to the carapace.

Animal care
Daphnia were kept in Daphnia culture medium 33 in cleaned pickle jars with loose-fitting glass dishes for lids and fed daily with 3 -6 drops of RGComplete (reefnutrition.com) depending on population size. To reduce overcrowding and resting egg production, all but the largest Daphnia were removed once every two weeks by pouring through two nets into a Tupperware, the first net with a 2mm pore size to catch the largest Daphnia, the second net with a fine pore size such 315 that no hatchlings went through to the Tupperware. The 2mm net was then placed upside down over the mouth of the pickle jar and the water in the Tupperware was poured through the 2mm net, releasing the largest Daphnia back into the pickle jar.
Fixation 320 From jars culled as above, the largest Daphnia were coaxed with light to one area of the pickle jar and removed with a plastic pipette with tip cut off to a Sylgard 184 (Dow Corning) dish.
Water was removed from Sylgard dish to immobilize Daphnia. Daphnia with embryos were picked up gently with forceps and placed in a medicine cup with culture medium. Once all animals with embryos had been gathered, animals were transferred to a 1.5mL Eppendorf tube, 325 water removed with pipette, then animals were fixed for 1-2 hours by adding 3.2% aqueous paraformaldehyde (Electron Microscopy Sciences) in Daphnia culture medium. Less fixation time seems to reduce background. Fixed animals were washed 3x5min with PBS-Tween then dehydrated stepwise into methanol, then stored at -20C.

330
In situ HCR In situ HCR performed as in Bruce et al 2021 17 . Embryos were removed from adults during final PBS-Tween wash prior to pre-hybridization step.

Imaging 335
Embryos imaged with Zeiss LSM880 confocal. Image processing done with Fiji-ImageJ. Fiji "Image Calculator > Subtract" method was used to remove high background from yolk autofluorescence. Figures processed using Adobe Photoshop 2020.

Supplemental information legends 340
Fig. S1. Daphnia adult dissected thoracic legs 1 -3. Daphnia legs are "phyllopodous", or leafshaped, and challenging to relate to walking legs in other arthropods such as Parhyale and Tribolium. However, a few key points facilitate comparisons. First, leg segment 6 can be identified because when crustacean legs are "biramous", i.e., split into a lateral exopod (exo) and 345 a medial endopod (endo), such as the Daphnia thoracic legs, then the exopod and endopod are carried on leg segment 6 (basis or basipod). Counting distally from leg segment 6, the endopod of leg 1 appears to have one segment and the endopod of leg 2 has at most 3 segments. Given that a complete endopod has 5 leg segments, and no terminal claw is apparent, Daphnia thoracic legs do not appear to express leg segment 1. Second, each Daphnia thoracic leg is identified by 350 its shape and setal patterns. For example, Leg 2 has a short comb (gnathobase) and an elongated exopod with two long setae, while Leg 3 and 4 have a wide comb and a paddle-shaped exopod with an array of six setae. Endopod in c is underneath comb and not clearly visible. Epi, epipod (a type of exite) of the coxa (leg segment 7). Pre-epi, pre-epipod of leg segment 8 35 .