Pax6 homologs are required for patterning both visual systems of the daddy-longlegs Phalangium opilio Authors: Ethan M Laumer, Sophie M Neu, Benjamin C Klementz, Pranil Panda, Emily V W Setton, Prashant P Sharma, Ilya Ruvinsky Published: 2026 ABSTRACT AbstractThe evolution of visual systems has compelled numerous investigations of developmental processes underlying eye patterning across Bilateria. It is well-established that homologs of the transcription factor Pax6 play a highly conserved role in eye fate specification and are at the top of the retinal determination gene network (RDGN) hierarchy. Curiously, Pax6 homologs do not appear to maintain this function in well-studied chelicerate models. It was recently proposed that the gene Pax2 (shaven; sv) may have subsumed the role of Pax6 in eye fate specification in chelicerates. However, no functional data are available for any chelicerate Pax homologs. We examined the incidence of Pax family genes across chelicerate phylogeny, and interrogated the expression and function of Pax2 and Pax6 homologs in the daddy-longlegs Phalangium opilio, which bears a highly plesiomorphic arrangement of visual systems. We show that the Pax6 homologs eyeless (ey) and twin of eyeless (toy) are expressed early in the developing head lobes of P. opilio, whereas sv is not. Gene silencing of ey, toy, and sv individually had no discernible effect on eye development. By contrast, double knockdown of ey and toy resulted in defects or loss in all three eye pairs of P. opilio. These data are consistent with a conserved role for Pax6 homologs in patterning both median and lateral visual systems across arthropods. Our results suggest that heterochronic shifts in expression, rather than changes in function, underlie the atypical dynamics of Pax genes in derived arachnid groups such as spiders.Keywords: morphogenesis, eye development, paralogy, vestigial organs, retinal determination network Introduction Eye fate specification is an intricate process that is essential for initiating the development of this complex organ. Across Bilateria, homologs of the transcription factor Pax6 play a central role in specifying eye fate, suggesting deep evolutionary conservation of the eye patterning program (Kozmik 2005). Broadly, the Pax family of transcription factors is characterized by the presence of a conserved 128-amino acid domain (PRD), which in turn is comprised of two helix-turn-helix (HTH) DNA-binding subdomains, PAI and RED (Friedrich 2017). Abrogation of Pax6 function in mammalian models results in failure of eye formation and brain morphogenetic defects (Quinn et al. 1996; Tzoulaki et al. 2005). Similarly, loss-of-function experiments for arthropod homologs of Pax6 in insect models elicit defects of the two visual systems (ocelli or median eyes; and compound or lateral eyes), as well as head defects (Quiring et al. 1994; Czerny et al. 1999; Yang et al. 2009; Friedrich 2017; Gutiérrez Ramos and Pick 2025). The functional significance of Pax6 sequence conservation is underpinned by seminal experiments showing that exogeneous Pax6 coding regions from a diversity of species can induce ectopic initiation of eye development in both the fruit fly Drosophila melanogaster and the clawed frog Xenopus laevis (Halder et al. 1995; Tomarev et al. 1997; Onuma et al. 2002). The accumulation of expression and functional data for Pax6 homologs in different metazoan clades has broadly supported their central role in eye morphogenesis (Yang et al. 2009; Yoshida et al. 2014; Klann and Seaver 2019; Gao et al. 2020), a function that may predate the origin of Bilateria (Suga et al. 2010). Recent work in the fruit fly has also suggested that the partitioning of the two visual systems of arthropods is associated with differential expression levels of the two arthropod Pax6 homologs, eyeless (ey) and twin-of-eyeless (toy), with ey playing a more prominent role in the patterning of the compound (lateral) eyes and toy of the ocelli (median eyes) (Jean-Guillaume et al. 2025). Both paralogs play synergistic roles in patterning the lateral eyes, whereas toy has a more prominent role in patterning the median eyes. Altering the relative expression of these two homologs can lead to a spectrum of eye defects, whereas altering the expression of downstream targets such as eyes absent (eya) can induce homeotic transformation of ocelli to compound eyes (Jean-Guillaume et al. 2025). While functional datasets from other insect and crustacean models are not as experimentally advanced, these broadly support the conservation of the retinal determination gene network (RDGN) in D. melanogaster, with emphasis on the prominent role of Pax6 genes in eye morphogenesis (Yang et al. 2009; Gao et al. 2020; Gutiérrez Ramos and Pick 2025). An odd exception to this understanding occurs in Chelicerata, the sister group of the remaining arthropods. Like in insects, two visual systems (median and lateral) occur in chelicerates, though one or both are reduced or completely lost in certain chelicerate orders (e.g. Pycnogonida; Solifugae; Ricinulei; Palpigradi) (Friedrich 2022). The best studied visual systems in chelicerates are those of horseshoe crabs and spiders, which differ markedly from each other (Fahrenbach 1975; Morehouse et al. 2017) (Fig. 1). The most prominent eyes of horseshoe crabs are the lateral facetted eyes, which bear over 1,000 ommatidia and are adapted for changes to vision between day and night (Harzsch et al. 2005, 2006). Given the early appearance of facetted eyes in Cambrian arthropods, the structure and organization of the horseshoe crab facetted eye has featured prominently in reconstructing the evolution of arthropod vision (Harzsch et al. 2006; Strausfeld et al. 2016). Posterior to the lateral faceted eye is a second rudimentary lateral eye that is formed during early development in the trilobite larva. Additionally, a trio of eyes occurs in the dorsal midline, comprising a pair of median eyes (ocelli) and an endoparietal eye, formed by fusion of a larval median eye pair; the median eyes play important roles in detection of UV light and the establishment of lunar cycles. A pair of ventral eyes is located near the mouth, but their function is unknown (Harzsch et al. 2006). By contrast, spiders bear a pair of median eyes and (typically) three pairs of disaggregated lateral eyes, though the spatial arrangement and number of eyes varies markedly between families (Morehouse et al. 2017; Chong et al. 2024). The lateral eyes are distinguished from the median eyes by the presence of a reflective crystalline layer called the tapetum, as well as the arrangement of the rhabdoms with respect to the photoreceptor cell bodies (inverted in the lateral eyes with respect to the median eyes) (Morehouse et al. 2017). In spiders with highly developed vision (e.g. Salticidae; Lycosidae), visual acuity is typically associated with the median eyes and the anterior pair of lateral eyes (Jakob et al. 2018). Organization of arthropod visual systems, with emphasis on Chelicerata. Red icons denote homologs of the lateral visual system, which includes the faceted eyes of insects and horseshoe crabs (Xiphosura), as well as the disaggregated ocelli of spider lateral eyes (Araneae). Blue icons denote homologs of the median visual system. Diagrams below line drawings represent simplified schematics of visual systems in each lineage. Note the incidence of two pairs of median eyes in sea spiders (Pycnogonida), as well as Opiliones (one pair vestigial). In Xiphosura, the third median eye shown in the schematic is the result of the fusion of paired larval median eyes. Additional eyes of Xiphosura are not shown for simplicity. Despite these structural and functional differences, horseshoe crabs and spiders are united by the absence of Pax6 expression in the eye primordia during embryonic stages when eye fate specification is thought to occur. Embryonic expression data are limited for horseshoe crabs, but a previous survey of one of the five Pax6 homologs (the result of lineage-specific whole-genome duplications) that occurs in Limulus polyphemus found no evidence that Pax6 was expressed in the eye primordia (Blackburn et al. 2008). Comparative data are more readily available for several spider species, including exemplars of different degrees of visual acuity. Initially, it was thought that the Pax6 homologs ey and toy were altogether absent from developing eye primordia, suggesting a Pax6-independent mechanism underlying eye fate specification across chelicerates (Samadi et al. 2015; Schomburg et al. 2015). However, subsequent explorations of earlier stages of development revealed that both copies were expressed in a broad domain demarcating the anterior head in the spider Parasteatoda tepidariorum, with subsequent diminution of transcription by the stages where the expression of downstream RDGN target genes eyes absent (eya) and sine oculis (so) is initiated (Leite et al. 2024). These expression patterns were later corroborated in several other spider species (Baudouin Gonzalez et al. 2025). It was therefore postulated that ey and/or toy may serve the function of establishing eye precursors indirectly and relatively early in development through the activation of eya and so, but are not required for the maintenance of these downstream targets (Friedrich 2022; Leite et al. 2024). While this explanation is plausible, no functional data exist supporting any role for Pax6 homologs in spider eye development, due to a combination of inefficiency of RNAi in this species (Pechmann et al. 2017; Gainett et al. 2020; Setton et al. 2024) and limited gene editing tools in spiders (Anon 2025; Santiago-Rivera and Scheibel 2025). In addition, it was recently shown that the same early expression domains of ey and toy occur in the anterior margin of the head in the blind oribatid mite Archegozetes longisetosus, suggesting that ey and toy plausibly retain a conserved role in head patterning as previously reported for insects (Joyce and Barnett 2025). An entirely separate hypothesis for the patterning of chelicerate eyes is the recent proposition that Pax2 (shaven; sv) may have taken on the function of eye fate specification in arachnids (Janeschik et al. 2022). Two copies of sv (distinguished as Pax2.1 and Pax2.2) occur in spider genomes, owing to an ancient genome duplication in the common ancestor of spiders and scorpions (Arachnopulmonata) (Sharma et al. 2014; Schwager et al. 2017; Ontano et al. 2021). In embryos of two spider species, it was found that Pax2.1 was expressed in the lateral eye primordia, whereas Pax2.2 was expressed elsewhere in the brain and in developing sense organs throughout the body (Janeschik et al. 2022). Based upon these gene expression patterns, the authors proposed an intriguing model wherein Pax2 replaced at least the lateral eye fate specification function of Pax6 in spiders (Pax2.1 is not associated with the median eyes), if not all chelicerates. The inference that Pax2.1 alone represents a key player in chelicerate lateral eye fate specification was later undermined by the observation that both sv ohnologs are expressed in both the median and the lateral eyes of the scorpion Centruroides sculpturatus (Gainett et al. 2024). Moreover, the single-copy homolog of sv in the daddy-longlegs Phalangium opilio exhibits a highly similar expression pattern to the two scorpion paralogs (Gainett et al. 2024), suggesting that the restriction of spider Pax2.1 to the lateral eyes represents a spider-specific phenomenon and cannot be generalized to all arachnids (Sharma and Gavish-Regev 2025). More generally, given that whole genome duplications are not characteristic of all chelicerates, the dynamics of duplicated genes in groups like spiders and horseshoe crabs may not be representative of the entire clade. Still, the role of sv (like Pax6 homologs) in arachnid eye development remains entirely untested because no functional data exist supporting either hypothesis. An investigation of Pax gene function in chelicerate eye development would best be served by investigating a model with (i) both visual systems, (ii) single-copy homologs of RDGN genes (due to the possibility of functional redundancy incurred by whole genome duplications), and (iii) the availability of robust functional tools. The single best candidate for such an investigation is the daddy-longlegs P. opilio (Sharma et al. 2013; Gainett et al. 2021, 2022). Though it was long thought that extant Opiliones only bear a single pair of eyes, we recently showed that three pairs of eyes develop in embryos of P. opilio (median eyes, vestigial median eyes, and vestigial lateral eyes), with two pairs of these persisting into adulthood (median and vestigial lateral pairs) (Gainett et al. 2024). Opiliones therefore exhibit a more plesiomorphic organization of visual systems than either spiders or scorpions, and more closely resemble the horseshoe crab and pancrustacean visual system arrangement. Here, we investigated the function of sv and both Pax6 homologs (ey and toy) in developing embryos of P. opilio using RNA interference (RNAi). We show that Pax6 homologs are expressed in early stages, consistent with the hypothesized role in early eye fate specification, whereas sv is not expressed until after the initiation of so and eya expression. We also show that double knockdown of ey and toy results in defects of the eyes and loss of opsin expression in all three eye pairs, whereas sv RNAi has no effect on any eye type, paralleling experimental results in insect models. These first functional datapoints for Pax genes in Chelicerata suggest retention of a deeply conserved role for Pax6 genes in the specification of both arthropod median and lateral visual systems. Results ey and toy are retained across Arthropoda A recent survey of Pax genes in insects reported some ambiguity in the presence of both ey and toy in chelicerates and myriapods, the two taxa subtending Pancrustacea in arthropod phylogeny (Gutiérrez Ramos and Pick 2025). The challenging orthology of chelicerate Pax6 genes is thought to result from the lack of two diagnostic sites in the Pax6 sequences of chelicerates (these two sites are understood to be specific to Pancrustacea), the paucity of information content in multiple sequence alignments restricted to the PRD domain, and limitations in taxon sampling (Friedrich 2017). We therefore performed a survey of the Pax family, sampling 11 out of the 14 orders of Chelicerata. Broadly, the number of Pax homologs per species varied more in chelicerates than previously reported in insects, a result consistent with multiple waves of whole-genome duplication in two separate branches of the chelicerate tree of life (Fig. 2a). The largest complement of Pax genes in genome assemblies was found in the horseshoe crab Carcinoscorpius rotundicauda (n = 33), whereas the smallest was found in the acariform mite Tetranychus urticus and the pseudoscorpion Cordylochernes scorpioides (n = 8). Pax gene complements across Chelicerata reveal lineage-specific patterns of paralog retention in Arachnopulmonata. a) Distribution of Pax homologs mapped onto a simplified phylogeny of panarthropods. Note the ancestral duplication of Pax6 in the common ancestor of Arthropoda. Incidence of Pax homologs is indicated by colored boxes (following Gutiérrez Ramos and Pick 2025); numbers inside boxes denote the number of gene copies. Circles on branches represent whole-genome duplication events. b) Multiple sequence alignment of the PRD domain for selected exemplars of Panarthropoda, indicating diagnostic sites for ey and toy in the linker region between the PAI and RED subdomains (red arrowhead). Design for panel (b) follows Friedrich (2017). In a phylogenetic analysis inferred from multiple sequence alignment of the PRD domain, two Pax6 homologs were recovered in all arthropods except Tetranychus urticae (Acariformes; ey only), Gluvia dorsalis (Solifugae; toy only), and Phrynus marginemaculatus (two copies of toy only). As previously noted, the incidence and expression of ey and toy have been described in the acariform mite A. longisetosus, suggesting that the absence of one copy in T. urticae may be a lineage-specific loss. Similarly, the lack of ey in G. dorsalis is likely attributable to the incompleteness of this genome; both ey and toy were found in the developmental transcriptome of a second Solifugae species (T. salinarum). By contrast to arthropods, only one Pax6 copy was found in both Onychophora and Tardigrada genomes, supporting the inference that ey and toy are the result of an arthropod-specific duplication (Friedrich 2017). As observed previously, the relationships between ey and toy are challenging to decipher using phylogenetic analysis of the PRD domain alignment alone, which did not result in mutually monophyletic clusters corresponding to ey and toy. To augment information content, we performed a separate analysis of only Pax6 homologs using full length sequences, which was able to resolve ey and toy as mutually monophyletic clusters (Figures S1 and S2). Comparison of sequence data also corroborated the utility of a diagnostic site in the linker between the PAI and RED subdomains that reliably distinguishes ey from toy (Fig. 2b). The recent survey of Pax genes in insects also reported some ambiguity in the homology of Pax3/7 homologs (in D. melanogaster: gooseberry, gooseberry-neuro, and paired) (Gutiérrez Ramos and Pick 2025). We performed a similar ancillary analysis using full-length sequences of Pax3/7 homologs, but recovered little phylogenetic structure within the gene tree and negligible nodal support for evolutionary relationships between Pax3/7 homologs (Figure S3), recapitulating the results of Gutiérrez Ramos and Pick (2025). Given that several chelicerate orders with unduplicated genomes retain more than one Pax3/7 homolog, we infer that a duplication of Pax3/7 genes likely predated the diversification of extant arthropods. Taken together, our results suggest that ey and toy homologs were present in multiple chelicerate and pancrustacean exemplars, and by extension, across Arthropoda. Duplicates of ey and toy in arachnopulmonate genomes are attributable to whole genome duplication A whole genome duplication in the common ancestor of Arachnopulmonata is evidenced by systemic paralogy across genetic elements, such as gene families, gene clusters (e.g. the Hox and Six genes), and microRNAs (Leite et al. 2016, 2018; Schwager et al. 2017; Ontano et al. 2021; Klementz et al. 2025: 20). By contrast, previous surveys of Pax6 homologs across seven spider species have consistently reported both ey and toy as single-copy (in contrast to other RDGN members, such as dachshund [dac], sine oculis/Six1 [so], Optix/Six3, and orthodenticle [otd]) (Samadi et al. 2015; Schomburg et al. 2015; Baudouin Gonzalez et al. 2025). We observed that four arachnopulmonates in our survey harbored two copies of ey (Cordylochernes scorpioides [Pseudoscorpiones], Aphonopelma hentzi [Araneae], Phrynus marginemaculatus [Amblypygi], and Mastigoproctus giganteus [Uropygi]). Two others retained a single copy of each (Centruroides sculpturatus [Scorpiones] and Parasteatoda tepidariorum [Araneae]). To test the hypothesis that the third Pax6 copy may reflect the process of whole-genome duplication (albeit with the shuffling of gene order and loss of duplicated genes in one or the other chromosome), we assessed the identity of genes neighboring ey and toy. We examined the identities of the 25 genes upstream of toy homologs and the 25 downstream of ey homologs in P. opilio, as well as four arachnopulmonates. We then circumscribed genes that were shared in this window in P. opilio and any arachnopulmonate. In the unduplicated genome of P. opilio (a non-arachnopulmonate arachnid), ey and toy occur on a single scaffold (ptg000048l_1) and are separated by a single gene (transmembrane 7 superfamily member 3), a homolog of which occurs upstream of a single ey homolog in M. giganteus (scaffold 12) (Figure S4; Table S1). Seven other genes occur in this 53-gene chromosomal window that are also found in at least one other arachnopulmonate scaffold containing a Pax6 homolog: glucose dehydrogenase (scaffold 8 of M. giganteus), transposable element Tc1 transposase (scaffold 7 of A. hentzi), ETS homologous factor (scaffold 7 of A. hentzi; scaffolds 8 and 12 of M. giganteus; scaffold 4 of C. sculpturatus; scaffold NC_092205.1 of P. tepidariorum), DNA polymerase delta 1, catalytic subunit (scaffold 12 of M. giganteus), ankyrin repeat and SAM domain-containing protein 1A (scaffolds 7 and 8 of A. hentzi; scaffolds 8 and 12 of M. giganteus; scaffold 4 of C. sculpturatus; scaffold NC_092205.1 of P. tepidariorum), general transcription factor IIH subunit 1 (scaffold 12 of M. giganteus), and buccalin (scaffold 4 of C. sculpturatus). Comparing within the arachnopulmonates for similarly sized windows (51-55 genes; 25 genes upstream and downstream of Pax6 homologs), we discovered 17 other cases of homologs shared between at least 2 scaffolds. We reasoned that the paucity of shared genes detected in the linkage group containing Pax6 homologs in the P. opilio genome (8 genes) could reflect gene shuffling and chromosomal reorganization over time, as P. opilio is more distantly related to any arachnopulmonate than they are to each other. We therefore searched all five genomes for 19 genes that we identified as present in at least two species, and mapped these with respect to the Pax6 genes, regardless of the number of intervening genes. This broader search supported the inference of a shared linkage group. Fifteen out of 19 genes were present on the same scaffold as the Pax6 homologs in P. opilio, whereas the remaining three were retained on the same scaffold as a Pax6 homolog only in the arachnopulmonates (Fig. 3; Table S2). As measured by the number of genes retained on the same scaffold as a Pax6 homolog, the organization of the duplicated linkage group was better retained in A. hentzi and M. giganteus compared to C. sculpturatus and P. tepidariorum. While the two single-copy homologs of ey and toy both occurred on a single chromosome in the latter two species, we nevertheless found smaller fragments of the duplicated cluster on other scaffolds in C. sculpturatus and P. tepidariorum. Conserved gene linkages support arachnopulmonate whole genome duplication as the mechanism underlying Pax6 paralogy. Chromosomal gene complements within 25 genes upstream and downstream of Pax6 homologs. From top to bottom: P. opilio (Opiliones), M. giganteus (Uropygi), C. sculpturatus (Scorpiones), A. hentzi (Araneae, Mygalomorphae), and P. tepidariorum (Araneae, Araneomorphae). Identified genes are shared between at least 2 taxa. Ribbons represent homologs present in adjacent taxa. Gene identities: 1—zinc-cadmium resistance protein; 2—BLOC-2 complex member HPS5/Hermansky-Pudlak syndrome 5 protein; 3—ETS Homologous Factor; 4—twin of eyeless; 5—eyeless; 6—ankyrin repeat/SAM domain-containing protein 1A; 7—Buccalin; 8—diuretic hormone class 2; 9—syntaxin-12; 10—prothoracicostatic peptide; 11—troponin I; 12—troponin T; 13—sodium-dependent proline transporter; 14—myosin kinase light chain, smooth muscle; 15—ras associated domain-containing protein 10; 16—fuzzy planar cell polarity protein; 17—anaphase-promoting complex subunit 4; 18—choline kinase alpha; 19—serine/arginine-rich splicing factor 4. Taken together, the incidence of shared homologs in the same linkage group across arachnid orders, as well as the duplication of the linkage group in arachnopulmonates, is consistent with the scenario of a shared genome duplication in the common ancestor of Arachnopulmonata, in tandem with lineage-specific patterns of chromosome reorganization and gene loss. ey and toy, but not sv, are expressed in the head lobes during early P. opilio development In contrast to spiders and horseshoe crabs, ey and toy are known to be expressed in the eyes of P. opilio throughout eye morphogenesis, with emphasis on lentigenic cells in later eye development (Gainett et al. 2024). However, expression of Pax genes in P. opilio is not known prior to stage 8. In the present study, HCR assays of earlier developmental stages detected ey and toy expression in the head lobes of stage 6 embryos (Fig. 4a). The two homologs are co-expressed in two bilaterally symmetric domains in the protocerebral (first head) segment, with toy expression extending toward the ventral midline to form a continuous domain traversing the full segment (Fig. 4a). ey is expressed in a smaller territory that co-expresses toy. At stage 7, toy is expressed discontinuously with respect to the ventral midline, with thin projections of toy expression still detected flanking the ventral midline (Fig. 4b, 4c). As in stage 6, ey is co-expressed within a subdomain of toy. At stage 8, expression of both ey and toy broadens in the head lobes, expanding dorso-laterally. Additional bilateral regions of ey expression appear outside the head in each of the following body segments (Fig. 4d). At this point, toy still retains a medial expression domain in the anterior-most region of the head that lacks co-expression of ey. Expression of Pax homologs is consistent with a role for ey and toy (but not sv) in early head patterning. Top: HCR in situ hybridization for Po-ey (yellow) and Po-toy (magenta) in wild type embryos. a′–j′) Expression patterns of targeted genes; a–j) counterstaining of same embryos with Hoechst (cyan). a) Stage 6. b) Stage 7, view of ventral ectoderm. c) Stage 7, view of head lobes. d) Stage 8. e) Stage 9. Bottom: HCR in situ hybridization for Po-sv (green). f) Stage 6. g) Stage 7, view of ventral ectoderm. h) Stage 7, view of head lobes. i) Stage 8. j) Stage 9. Note the early expression of Po-ey and Po-toy in the head lobes by stage 7, whereas Po-sv remains restricted to ventral neurectoderm at the same stage. Sample sizes indicate the number of embryos imaged with confocal microscopy that exhibited the expression patterns shown, for each stage. Total counts of embryos exhibiting these expression patterns are typically >25 per stage. Abbreviations: pc, protocerebral segment; hl, head lobe; ch, chelicera; pp, pedipalp; L1, first walking leg; me, median eye primordia; le, lateral eye primordia. Additionally, at stage 8, ey and toy expression in posterior segments appear as bilaterally symmetrical domains in the ventral ectoderm flanking the midline. At stage 9, expression of the two Pax6 homologs remains localized to the same regions as stage 8, but regions of heterogeneous expression strength result in a more complex expression pattern (Fig. 4e). Surveys of sv reveal a lack of expression in the ocular segment at stage 6 (Fig. 4f). Instead, only bilaterally symmetrical bands of expression in the ventral ectoderm of posterior segments are observed. At stage 7, expression of sv intensifies as bilaterally symmetrical patches along each body segment but is absent from the protocerebral segment (Fig. 4g, 4h). At stage 8, sv begins to be weakly expressed in small discontinuous territories in the lateral head lobes (Fig. 4i). Expression in the ventral ectoderm of posterior segments resolves into broad, bilaterally symmetrical domains. Additional domains are observed through the middle of each limb bud. At stage 9, sv expression in the limb buds resolves into clearly defined tracts of expression, corresponding to presumptive nerves (Fig. 4j). Strong expression is observed in paired domains in the ventral ectoderm of each segment. In the head, sv is strongly expressed in the territory corresponding to the presumptive lateral eyes, as well as two domains at the anterior of the head lobes. In summary, surveys of three Pax homologs' expression suggest that (i) ey and toy exhibit conserved expression dynamics with respect to other arthropods, and (ii) sv is not expressed in the protocerebrum until after the downstream RDGN genes eya and so have already commenced expression (stage 8; Gainett et al. 2024). ey and toy are required for the development of all three eye pairs in P. opilio No functional data exist for arthropod Pax genes outside of Pancrustacea. To redress this gap, we performed embryonic RNAi against ey and toy in P. opilio. To test the hypothesis that sv has gained a taxon-specific role in chelicerate eye patterning, we performed RNAi against the single-copy sv homolog of P. opilio. Phenotypes were scored in embryos at stage 17, by which point both well-formed eyes are present and the ozopores (repugnatorial gland pores) are first observable (Gainett et al. 2022). RNAi against ey elicited only wild type morphology (n = 154/225; 68.4%) or dead embryos (n = 71/225; 31.6%) (Fig. 5a, 5b; Table S3). Likewise, RNAi against sv elicited wild type eye development (n = 205/234; 87.6%) or dead embryos (n = 29/234; 12.4%). RNAi against toy resulted in a small number of embryos exhibiting subtle asymmetry of eye size, with one eye smaller than the other, but without evident defects in the morphology of the smaller eye (n = 14/271; 5.2%) (Table S3). Penetrance of knockdown was measured using RNA sequencing, which supported statistically significant reduction of target gene expression level in all single-gene experiments (Fig. 5c; Table S4). Double knockdown of Po-ey and Po-toy disrupts median eye patterning in P. opilio. a) Locations and lengths of cDNA fragments amplified from ey, toy, and sv transcripts for synthesis of dsRNA in RNAi experiments. b) Distribution of phenotypic outcomes following RNAi experiments. c) Relative expression of sv, ey, and toy transcripts for control and RNAi embryos. d) Brightfield image of stage 17 wild type P. opilio embryo showing normal eye development. d′) Same embryo as in (d) with magnification of median eyes. e–g) Brightfield images of spectrum of eye defects in double Po-ey + Po-toy RNAi embryos. e) Class I RNAi embryo at stage 17 with one small median eye and one defective median eye. f) Class II RNAi embryo at stage 17 with two defective median eyes. g) Class III RNAi embryo at stage 16 with egg tooth, but no obvious eye structures. Note that most Class III embryos underwent developmental arrest by stage 16. e′–g′) Same embryos as in (e–g), with magnification of median eyes. Abbreviations: et, egg tooth; me, median eye; me* median eye with developmental defect. Scale bars: 100 µm. It has previously been shown that ey and toy act synergistically in the patterning of the fruit fly and beetle compound eye; joint knockdown of both Pax6 genes induces a stronger effect than single-gene knockdown, which is inferred to result from functional redundancy between the two homologs (Friedrich 2017; Jean-Guillaume et al. 2025). We therefore performed double-knockdown of ey and toy, which elicited a spectrum of defective eye phenotypes. RNAi embryos exhibited eyes that were smaller, heterogeneously pigmented, and missing components of the cuticular lens (n = 148/268; 88.1%) (Fig. 5e–5g; Table S3). Of the 148 specimens exhibiting eye defects, we classified embryos with one small median eye and one partially formed median eye as Class I phenotypes (n = 41/148) (Fig. 5e). Embryos with two defective median eyes were classified as Class II phenotypes (n = 95/148) (Fig. 5f). Embryos with the most severe eye defects (Class III; n = 12/148) exhibited little pigmentation, a smaller ocularium (the eye mound that bears the median eyes), and anophthalmia; such embryos typically underwent developmental arrest by stage 16 and did not survive to later stages of development (Fig. 5g). Our recent discovery of vestigial eye pairs in P. opilio facilitates an investigation of Pax gene function with respect to the two visual systems of chelicerates. We surveyed the expression of three opsins that are differentially expressed in the three eye pairs, as readouts of proper eye morphogenesis: peropsin (median eyes), Rh1 (vestigial median and vestigial lateral eyes), and Rh3 (vestigial median and vestigial lateral eyes) (Gainett et al. 2024). Expression assays showed a spectrum of phenotypes, with opsin expression only in vestigial lateral eyes (n = 8/12) or lost in all three eye pairs (n = 4/12), with the latter associated with embryos that bore drastically reduced ocularial height (Fig. 6). Consistent with the high degree of mosaicism in eye morphology of RNA embryos at stage 17 (Fig. 5b), we observed mosaicism in HCR surveys of stage 9 embryos, with one side visibly more strongly affected than the other in 2/11 embryos surveyed (Figure S5). RNAi against Po-ey + Po-toy disrupts development of both median and lateral visual systems. a–e) Hoechst counterstaining (gray) and multiplexed expression of Po-peropsin (green), Rh1 (magenta), and Rh3 (yellow) in P. opilio embryos. a) Wild-type embryo at stage 13. a′) Same embryo as in (a) with isolated opsin expression. Note that the expression of Po-peropsin is restricted to the median eyes, whereas Rh1 and Rh3 are restricted to the vestigial median and vestigial lateral eyes. b) Stage 15 negative control embryo. b′) Same embryo as in (b) with magnification of the median eyes. c) Stage 15 Po-ey + Po-toy RNAi embryo. (d) Stage 17 or 18 negative control embryo. d′) Same embryo as in (d) with magnification of median eyes. e) Stage 17 or 18 Po-ey + Po-toy RNAi embryo. Note the loss of Po-peropsin expression in the median eyes, and loss of r-opsin and Rh3 in the vestigial median eyes in RNAi embryos. Filled circles represent wild-type expression of opsins in each eye pair. Empty circles represent complete absence of opsin expression. Half-filled circles represent mosaic phenotypes (retention of wild-type expression on only one side). Abbreviations: me, median eyes; vme, vestigial median eyes; vle, vestigial lateral eyes. Scale bars: 100 µm. To address the possibility of differential roles for ey, toy, or sv in subdividing the patterning of different eye types, we performed the same HCR assays for single-gene knockdown experiments. However, all embryos assayed in these experiments exhibited wild type expression of all three opsins (ey RNAi, n = 22; toy RNAi, n = 25; sv RNAi, n = 18). Beyond eye patterning defects, the insect Pax6 loss-of-function phenotypic spectrum includes defects of the head that include microcephaly, as Pax6 plays an additional role in head and brain development. For example, a maternal RNAi experiment in the flour beetle Tribolium castaneum showed that reduction of dachshund expression in the compound eye in ey + toy RNAi embryos is the result of a deletion of that territory of the developing head lobe (Yang et al. 2009). CRISPR experiments in a honeybee have also shown that toy knockout results in eye defects and microcephaly (Hu et al. 2024). To assess the possibility that loss of opsin expression in the vestigial median eyes in P. opilio ey + toy RNAi phenotypes could have resulted from secondary effects of head deletion, we assayed RNAi embryos for the RDGN members dac, so, and eya. ey + toy RNAi embryos exhibited diminution of eya in the head lobes, consistent with a known anophthalmia phenotype previously described for eya RNAi in P. opilio (Fig. 7). These embryos also exhibited diminution of so and dac, which is consistent with known roles of these genes in patterning both visual systems of a spider and the lateral visual system in P. opilio, respectively (Gainett et al. 2020, 2024). HCR is challenging for interpreting the degree of knockdown of transcription factors; we therefore examined quantitative expression of these genes, using an RNA-Seq dataset. Consistent with the imaging data, normalized counts of these RDGN members showed significant knockdowns in all three genes relative to their negative controls (independent two-sample, two-tailed t-tests; P << 0.05). Nevertheless, we detected no qualitative differences in the expression patterns of these three genes, disfavoring the interpretation that the loss of opsin expression is a secondary effect of head deletions in ey + toy RNAi embryos (Fig. 7). This result is consistent with the implementation of our embryonic RNAi experimental design; the stages injected with dsRNA in this study have already formed the head lobes (stage 6 of P. opilio; Gainett et al. 2022), and thus ey + toy RNAi should not affect earlier head-patterning functions, in contrast to maternal RNAi experiments in T. castaneum. Embryonic Po-ey + Po-toy RNAi phenotypes are not attributable to deletions of head territories. HCR assays for dac, so, and eya. Gray: Hoechst counterstain. a) Negative control embryo 72 h after injection (stage 11). a′–a′′′) Same embryo as in (a), single-channel counterparts. b) Po-ey + Po-toy RNAi embryo 72 h after injection (stage 11). b′–b′′′) Same embryo as in (a), single-channel counterparts. Note the diminution of dac, so, and eya expression, though the morphology of the embryo does not support the deletions of head regions, with emphasis on the lateral eye territory (le). c) Gene regulatory network showing present understanding of RDGN regulatory interactions in insects, after Yang et al. (2009). d) Normalized read counts of surveyed genes 48 h after injection. P values correspond to independent two-sample, two-tailed t-tests. Only statistical comparisons between RNAi and control counterparts are shown; raw data are reported in Table S4. Abbreviations as in Fig. 4. Taken together, our results are consistent with the understanding that ey and toy act synergistically to pattern both visual systems of an arachnid, closely paralleling the results of insect experiments (Friedrich 2017). By contrast, we found no role for sv in patterning any of the eyes, consistent with its later onset of expression in the head in stages after the onset of so and eya expression. Discussion A conserved role for Pax6 homologs in the patterning of the 2 arthropod visual systems Despite long-standing interest in the evolution of the arthropod visual system, the developmental genetic literature harbors a prominent dearth of functional data from phylogenetically significant groups, such as chelicerates, myriapods, and Onychophora (the sister group to Arthropoda). Part of this gap is attributable to the complete absence of functional tools in exemplars of Myriapoda and Onychophora, but this limitation does not apply to chelicerates. RNA interference has been available as a tool for the study of developmental genetics in spider models for 25 years, though its efficiency is variable across studies (Schoppmeier and Damen 2001; Akiyama-Oda and Oda 2006; Khadjeh et al. 2012; Pechmann et al. 2017; Setton et al. 2024). For example, our previous RNAi study on so-1 in the spider P. tepidariorum was able to elicit a spectrum of anophthalmia phenotypes, albeit in only 9.5% of embryos (Gainett et al. 2020). Despite this seemingly promising beginning, we and others have also previously performed RNAi against every RDGN gene duplicate in P. tepidariorum as single-gene knockdowns; none of these experiments yielded eye defects (Gainett et al. 2020; Schacht et al. 2020). As there are no functional data corresponding to any other spider eye-patterning genes beyond so-1, most of the present understanding of spider eye patterning (e.g. RDGN ohnolog dynamics; Hedgehog and Wnt signaling) is based on interpretation of gene expression patterns in tandem with extrapolations of function derived from the fruit fly literature (Baudouin-Gonzalez et al. 2022; Medina-Jiménez et al. 2024; Baudouin Gonzalez et al. 2025). Advanced functional tools (e.g. maternal CRISPR-mediated knockout and knock-in) do exist for some chelicerate models, such as the mite Tetranychus urticae and the tick Ixodes scapularis, but these models either lack eyes (I. scapularis) or ongoing work on these systems has mostly been directed toward advancing study of biological control in support of agricultural and biomedical interests (Dermauw et al. 2020; Sharma et al. 2022; De Rouck et al. 2024). The recent establishment of a maternal CRISPR technique in the spider P. tepidariorum (Li et al. 2026) is similarly not promising without further development, due to intractably low penetrance, inconsistent phenotypes of putative G0 crispants within experiments, and the inability of the embryos to survive to hatching in the injection medium, precluding the establishment of transgenic lines or full characterization of structures that develop postembryonically, including the eyes (Gainett et al. 2020). This gap in our understanding of arachnid eye development has been bridged by the daddy-longlegs P. opilio, both due to the efficiency of RNAi in this species as well as the recent discovery of a plesiomorphic visual architecture featuring two sets of vestigial eyes in Opiliones (Gainett et al. 2024). We therefore brought this system to bear upon the question of Pax gene function in Chelicerata. Both the expression pattern and the lack of an eye-patterning function of sv disfavor the hypothesis that Pax2 has superseded Pax6 function in eye fate specification (either median or lateral). This is consistent with the known function of sv in D. melanogaster, wherein sv is necessary for the differentiation of cone and primary pigment cells and regulating crystallin production (Fu and Noll 1997). Mutants for sv exhibit defects in ommatidial cell development, which include cell fate transformations and a disruption in the formation of proper cell structures. These known functions accord with the onset and expression patterns of Pax2 homologs in embryonic daddy-longlegs, spiders, and scorpions (Janeschik et al. 2022; Gainett et al. 2024). Our previous surveys have also shown that the subdivision of expression patterns of Pax2 ohnologs in spiders is taxon-specific; in a scorpion (two sv ohnologs) and a harvestman (single-copy sv), sv homologs are expressed in both the median and lateral eye primordia. These expression surveys tentatively support a series of derived events in some lineage subtending spiders (e.g. crown-group Araneae; Tetrapulmonata), where that lineage lost Pax2.1 expression in the median eyes and Pax2.2 expression in both the median and lateral eyes (Sharma and Gavish-Regev 2025). These spider-specific expression dynamics are therefore intriguing, but their significance remains unknown. The availability of new genomic and developmental resources for exemplars of Pedipalpi (the sister group of Araneae) at least offers a potential means to pinpoint this evolutionary sequence of expression domain changes across the arachnopulmonates (Gainett and Sharma 2020). Like sv RNAi, single-gene knockdown experiments against ey and toy did not elicit an eye defect phenotype (though a minority of toy RNAi embryos exhibited mild asymmetry of eye size). By contrast, double-knockdown of ey and toy elicited a consistent spectrum of eye defects with >85% efficiency. In these double RNAi experiments, expression assays of differentially expressed opsins corroborated defects to the vestigial median and vestigial lateral eyes as well, suggesting that ey and toy act synergistically to pattern all three eye pairs of P. opilio. The significance of this result is that it closely parallels Pax6 knockdown experiments in insects, with single-knockdown RNAi experiments eliciting limited phenotypic spectra and double knockdown of ey + toy eliciting a range of microphthalmic and anophthalmic phenotypes with higher penetrance (Friedrich 2017). The requirement that both ey and toy are silenced to obtain this phenotypic spectrum is consistent with the understanding that these two Pax6 homologs can reciprocally rescue one another's loss of function and that they compete for binding sites of target genes (Jean-Guillaume et al. 2025). In addition, the evolutionary significance of these results is that they support an ancestral requirement of ey and toy for patterning both median and lateral visual systems of insects and arachnids, and by extension (given the phylogenetic placement of chelicerates as the sister group of the remaining arthropods), across the phylum. Recent advances in the understanding of Pax6 function in the fruit fly have leveraged the availability of functional tools to demonstrate how precise expression levels of ey and toy establish ocellar and compound eye fates in D. melanogaster. A remarkable component of these experiments achieved the homeotic transformation of the ocelli to compound eyes through overdriving expression of eya, a downstream target of Pax6, in the anterior head via a GAL4 driver (Jean-Guillaume et al. 2025). The significance of this study is that it connects the dissimilar morphologies of ocelli and facetted eyes as part of an evolutionary transformation series, underpinned by a shared developmental program, despite hundreds of millions of years since the separation of the two visual systems. It has been suggested that this mechanism may be conserved across arthropods as well, a possibility that could be tested by altering levels of Pax6 homolog expression in the median versus lateral eyes of spiders (Jean-Guillaume et al. 2025). Due to the unavailability of advanced functional tools in spider and daddy-longlegs models, overdriving expression in specific tissues is not yet feasible in these models. Nevertheless, a potential route to testing conservation of eye-patterning mechanisms with available tools may be to trial RNAi against groucho (gro), a broadly expressed transcriptional co-repressor that has been shown to mediate the choice between ocellar and compound eye fate in D. melanogaster (Jean-Guillaume et al. 2025). Specifically, gro mutants and GAL4-driven gro RNAi lines exhibit hyperproliferation of ocellar field or transformation of the anterior head epidermis into ocelli. This effect is additive; GAL4-driven overexpression of eya in the anterior head in tandem with gro RNAi results in both homeotic ocelli-to-compound eye transformations, as well as homeosis of the head epidermis in between ectopic compound eyes into ocelli (Jean-Guillaume et al. 2025). Future investigation of eye-patterning in P. opilio should therefore trial gro RNAi, with the prediction that abrogation of gro expression levels should incur hypertrophy of the median eye fields. Genome architecture of the linkage group containing ey and toy supports an arachnopulmonate WGD event A recent survey of RDGN genes in spiders reported single-copy homologs of both ey and toy in all surveyed taxa, which included the theraphosid Acanthoscurria geniculata (Baudouin Gonzalez et al. 2025). We were therefore surprised by the occurrence of three Pax6 copies in the genomes of the theraphosid A. hentzi, as well as two other arachnopulmonate species (M. giganteus and C. scorpioides). A possible explanation for this third copy is that it represents the signature of the whole genome duplication (WGD) event shared in the common ancestor of Arachnopulmonata; the requirement of precise expression levels of Pax6 during eye fate specification may have selected against the retention of duplicate copies of ey and toy, resulting in widespread purging of their duplicates across arachnopulmonate genomes. However, one study has refuted the incidence of a WGD in arachnids altogether (Thomas et al. 2024). The basis for this conclusion was that (i) distributions of interparalog distances (Ks plots) in selected arachnopulmonates were not consistent with WGD, (ii) only a subset of genes such as the Hox genes exhibited systemic paralogy, (iii) self-synteny plots did not support patterns of duplication within arachnopulmonate genomes, and (iv) arachnopulmonate genomes did not exhibit syntenic patterns supporting WGD, as inferred from blocks of contiguous genes with conserved gene order. As discussed in our previous analysis of the M. giganteus genome (Kulkarni et al. 2025), (i) interparalog distances inferred using simplistic nucleotide substitution model corrections are deeply sensitive to model choice and lose power to detect ancient WGD events as a function of WGD event age (Tiley et al. 2018); (ii) systemic paralogy in arachnopulmonate genomes is widely known to extend well beyond the Hox genes and is substantiated by analyses of various gene families, gene expression patterns, and microRNAs (Leite et al. 2016, 2018; Nolan et al. 2020; Ontano et al. 2021; Aase-Remedios et al. 2023; Klementz et al. 2025); (iii) self-synteny has been reliably recovered in multiple arachnopulmonate (and particularly spider) genomes (Fan et al. 2021; Miller et al. 2022; Kulkarni et al. 2025); and (iv) a low discovery rate of syntenic patterns is closely tied to the use of conserved gene order as a defining criterion (Simakov et al. 2020). For example, a syntenic block may be contextually defined in a given study as a minimum of three adjacent genes that retain a conserved gene order (e.g. A–B–C; C–D–E). The problem with this definition is that even a single rearrangement within a small block of genes (e.g. A–B–C–D–E to A–B–E–D–C) will break the required minimum of three genes retaining the same gene order. Such a narrow definition of synteny for defining linkage groups works well for recent WGD events, but rapidly accrues a high false negative rate as the age of the WGD event increases. This is due to the accumulation of gene shuffling and genomic rearrangements over evolutionary time, particularly among small-bodied and rapidly evolving arthropod groups (Hoy et al. 2016; Klementz et al. 2025). Indeed, a return to the classic definition of synteny—gene linkages without consideration for gene order—has dramatically augmented the ability to detect ancient duplicated chromosomes in the metazoan (and especially vertebrate) tree of life (Simakov et al. 2020, 2022). The application of this concept of synteny to Arachnopulmonata and Xiphosura has accelerated the identification of duplicated ancestral chromosomes, though there is substantial variability in the cohesion of syntenic patterns across arachnopulmonate taxa (Klementz et al. 2025; Kulkarni et al. 2025). Specifically, lineages that exhibit longer generation times and lower evolutionary rates (e.g. mygalomorph spiders; Uropygi) tend to retain stronger signatures of WGD than lineages with short generation times and high evolutionary rates (e.g. small-bodied araneomorph spiders) (Klementz et al. 2025). Our examination of genome architecture in the context of the third Pax6 gene copy in A. hentzi and M. giganteus supported the inference of paralogy via WGD. When the criterion of gene order is imposed on arachnopulmonates, little evidence is available for syntenic blocks containing Pax6 homologs. When the criterion for gene order is relaxed, ancestral gene linkages harboring the Pax6 homologs are evident, as are conserved arrangements of genes within those linkage groups, albeit with additional genes interspersed between identified ohnologs. These results support the interpretation that various embryonic patterning genes and gene families may not faithfully retain duplicated copies stemming from the arachnopulmonate WGD (contra well-studied cases such as the homeobox and Wnt gene families; Leite et al. 2018; Harper et al. 2021; Aase-Remedios et al. 2023), possibly owing to different selection pressures acting upon different developmental processes. Moreover, the identification of numerous putative ohnologs that are unrelated to embryonic patterning (Fig. 3) suggests that the true scale of paralogy in arachnopulmonate genomes has not yet been fully explored. Ongoing establishment of new genomic resources for phylogenetically significant groups, in tandem with exploration of ancestral gene linkages specific to Chelicerata, is anticipated to improve the understanding of how ohnologs are retained or lost in lineage-specific patterns. A corollary of these observations is that the persistence of duplicated genes in arachnopulmonates may not reflect adaptive processes such as subfunctionalization or neofunctionalization, which are frequently invoked as explanatory vehicles for the diversification of arachnopulmonates. Specifically, divergence of expression patterns in arachnopulmonate ohnologs is often associated with divergence of function, morphological diversification, and/or the emergence of novel traits, though functional support for these associations is frequently lacking (Schomburg et al. 2015; Turetzek et al. 2016, 2017; Janeschik et al. 2022). A valid and competing hypothesis is that subsets of duplicated genes retained after WGD may simply be redundant, as exemplified by recent cases of posterior patterning in spiders (Benton et al. 2016; Li et al. 2026), or even in the process of gradual loss, a process that has received recent theoretical attention (Johri et al. 2022). Functional redundancy between persisting gene copies may also explain the recalcitrance of spider RDGN duplicates to RNAi. One potentially unexplored model for such investigations may be the horseshoe crabs, whose genomes retain a higher proportion of ohnologs, likely due to much younger genome duplications (Kenny et al. 2016; Shingate et al. 2020; Nong et al. 2021; Castellano et al. 2025). Investment in developing functional tools for Xiphosura may yield compelling insights about the incidence and persistence of functional redundancy between ohnologs, particularly in the context of ey and toy paralog dynamics during median versus compound eye development in this chelicerate order. Conclusion The role of Pax6 homologs in Chelicerata has long been a topic of interest, given expression data periodically generated for spiders and horseshoe crabs, but these discussions have been hampered by a lack of functional data. We were able to bridge this gap by generating a loss-of-function phenotype via double knockdown of ey and toy. This dataset, the first of its kind for Chelicerata, supports the inference that Pax6 genes play an evolutionarily conserved role in patterning both visual systems of an arachnid and insects, and by extension, across the arthropods. Next steps in deciphering the evolutionary history of arthropod eye evolution should prioritize testing the hypothesis that duplication and functional subdivision of ey and toy drove the origin of lateral and median visual systems in the common ancestor of Arthropoda. Materials and Methods Gene tree analysis and orthology inference To survey Pax homologs across Chelicerata, 13 chelicerate genomes or developmental transcriptomes were selected for comparative analysis, as well as six outgroup datasets. The 13 chelicerate genomes surveyed comprised Parasteatoda tepidariorum (Araneae), Aphonopelma hentzi (Araneae), Phrynus marginemaculatus (Amblypygi), Mastigoproctus giganteus (Uropygi), Centruroides sculpturatus (Scorpiones), Cordylochernes scorpioides (Pseudoscorpiones), Carcinoscorpius rotundicauda (Xiphosura), Gluvia dorsalis (Solifugae), Titanopuga salinarum (Solifugae), Tetranychus urticae (Acariformes), Ixodes scapularis (Parasitiformes), Phalangium opilio (Opiliones), Pycnogonum litorale (Pycnogonida). The outgroup genomes comprised Drosophila melanogaster and Daphnia pulex (Mandibulata), Euperipatoides kanangrensis and Epiperipatus broadwayi (Onychophora), and Rammazottius varieornatus and Paramacrobiotus metropolitanus (Tardigrada). BLASTp or tBLASTn searches were performed on these 19 datasets using the PRD domain of the ten D. melanogaster Pax genes as peptide queries. All hits with e-values <10−10 were retained. The resulting putative orthologs were verified and filtered using SMART BLAST searches and gene tree analysis, keeping only sequences that recovered a SMART BLAST hit to a D. melanogaster Pax family member. Multiple sequence alignment was conducted using CLUSTAL Omega (Sievers et al. 2011). Phylogenetic reconstruction consisted of maximum likelihood analysis with IQ-TREE v3 (Nguyen et al. 2015), with automated model selection and 1,000 ultrafast bootstrap resampling replicates (iqtree 3 -m MFP -bb 1000; chosen model: LG + G4). Three families of analyses were performed. First, a gene tree was inferred for all Pax genes using only the PRD domain in the multiple sequence alignment (ca. 125-peptide alignment; File S1), following the orthology inference strategy of Gutiérrez Ramos and Pick (2025). Next, to decipher relationships between ey and toy, a gene tree was inferred using full-length sequences of all Pax6 homologs without trimming (File S2). Third, to decipher relationships between Pax3/7 homologs, a separate gene tree was inferred using full-length sequences of all Pax3/7 homologs without trimming (File S3). The corresponding maximum likelihood gene trees are provided as Files S4-S6. Analyses of synteny To test for patterns of conserved gene linkages, scaffolds containing ey, toy, or both were annotated to determine the identity of neighboring genes in the genome assemblies of P. opilio, A. hentzi, P. tepidariorum, M. giganteus, and C. sculpturatus. The P. opilio genome was used as a starting point, due to its unduplicated condition. On the scaffold containing ey and toy, the 25 genes upstream of toy and the 25 genes downstream of ey, as well as a single gene in between ey and toy, were all identified using best BLASTp hit against the ClusteredNR database. BLASTp hits with an e-value <1 × 10−5 were retained. BLASTp searches resulted in one of three outcomes: a match to a sequence with an available functional annotation (almost always from other arachnid genomes); a match that lacked functional annotation (e.g, “uncharacterized protein”; “hypothetical protein”); or no matches (“no significant similarity found”). When hits were recovered, functional annotations were inferred using the annotation available for the best BLASTp hit (i.e. either the gene name or “hypothetical protein”). The same approach was repeated for the remaining four species (25 genes upstream and downstream of ey and toy homologs). Genes with identical BLAST hits in any two scaffolds were identified as homologous cliques. To assess the possibility of more extensive shuffling of genes, BLASTp searches were then performed for each clique, for all five arachnid species. Locations of any BLAST hits matching scaffolds containing ey or toy were retained. The relative order of genes was established using GFF or GTF files for the corresponding genome. Collection of embryos and gene expression assays Embryos of P. opilio were collected using coconut fiber dishes, as previously described (Gainett et al. 2022). Stage 6-10 embryos were fixed and assayed for hybridization chain reaction. Staging of embryos and protocols for fixation follows a recent staging system (Gainett et al. 2022). Orthology inference for Po-ey, Po-toy, Po-sv, Po-eya, Po-so, and Po-dachshund, design of hybridization chain reaction (HCR) probes, and protocols for fluorescent detection of gene expression were previously reported in our prior work (Gainett et al. 2024). The same HCR probes were redeployed to survey earlier stages of P. opilio development. RNA interference Primers were designed for fragments of the coding regions of Po-ey, Po-toy, and Po-sv using Primer3 (Table S5). These fragments were amplified using standard PCR protocols and cloned using a TOPO TA Cloning Kit using One Shot Top10 chemically competent Escherichia coli (ThermoFisher) following the manufacturer's protocol. Amplicon identity was verified via Sanger sequencing with M13 universal primers. Double-stranded RNA (dsRNA) was synthesized following the manufacturer's protocol using a MEGAscript T7 kit (Ambion/Life Technologies) from PCR templates. The quality of dsRNA was assessed and concentrations adjusted using a NanoDrop ONE to 3.9-4.1 µg/µL. dsRNA was mixed with vital dyes for visualization of injections. For double knockdown, dsRNA for each gene was adjusted to a concentration of 6.0 µg/µL and combined in equal volume. For negative controls, we used dsRNA for an exogenous fragment of vector plasmid sequence amplified using M13 universal primers. For visualization during injection, dsRNA was mixed with vital dye. Microinjection under halocarbon-700 oil (Sigma-Aldrich) was performed at stage 6; approximately 2 nL of dsRNA were injected into the peri-vitelline space using a Narishige IM-300 picospritzer. Subsets of injected embryos were fixed for HCR; the remainder was developed at 26 °C until stage 17-19 for scoring eye phenotypes. Validation and differential gene expression analyses Sets of 30-35 embryos were injected with dsRNA for sv, ey, toy, ey + toy, or an exogenous vector sequence. Embryos were fixed 48 h after injection using TriZOL trireagent (ThermoFisher). First-strand cDNA was prepared using oligo(dT) primers and Super-Script III reverse trancriptase (ThermoFisher). qPCR was trialed using PowerUp SYBR Green Master Mix (Life Technologies) on a StepOne Plus Real-Time PCR instrument (Applied Biosystems) following the manufacturer's protocol and using primers for elongation factor 1α as the endogenous control. However, low Ct values despite high input cDNA (>35 cycles for target genes of control embryos in pilot assays) precluded this strategy for validation of the knockdown. As an alternative, RNA sequencing was performed using total RNA extractions of sets of 30-35 embryos at 48 h after injection. Library preparation was performed using an Illumina TruSeq kit, following manufacturer's protocols. Stranded mRNA sequencing was performed on an Illumina NovaSeq with 2 × 150 bp paired-end reads. Phalangium opilio genome assembly version 2 (GenBank accession no. GCA_054372755.1) was converted into a reference index using STAR v.2.7.11 (Dobin and Gingeras 2015), followed by sequential mapping of each read file to the genome. Analysis of differential gene expression was performed using DESeq2 v. 1.48.2 (Love et al. 2014). DESeq2 was run with cutoff values for identification of differentially expressed genes set to a P-value of <0.05 and log-fold change >1. Imaging Brightfield microscopy was performed using a Nikon SMZ fluorescence stereomicroscope mounted with a DSFi2 digital color camera and driven by Nikon Elements software. Images of P. opilio appendages were captured at varying focal planes and compiled into focused stacks with inbuilt software tools. Confocal laser scanning microscopy was performed using a Zeiss LSM 980 microscope driven by Zen software. Maximum intensity projections were assembled using inbuilt software tools in Fiji (Schindelin et al. 2012). Supplementary Material Acknowledgments Imaging infrastructure was provided by the Newcomb Imaging Center, Department of Botany, University of Wisconsin-Madison. Discussions with Jesús Ballesteros, Guilherme Gainett, and Markus Friedrich refined the ideas presented in this paper. Comments from three anonymous reviewers improved an earlier draft of the work. Contributor Information Ethan M Laumer, Department of Integrative Biology, University of Wisconsin-Madison, Madison, WI 53706, USA. Sophie M Neu, Department of Integrative Biology, University of Wisconsin-Madison, Madison, WI 53706, USA. Benjamin C Klementz, Department of Integrative Biology, University of Wisconsin-Madison, Madison, WI 53706, USA. Pranil Panda, Department of Integrative Biology, University of Wisconsin-Madison, Madison, WI 53706, USA. Emily V W Setton, The Whitney Laboratory for Marine Bioscience, Department of Biology, University of Florida, St. Augustine, FL 32080, USA. Prashant P Sharma, Department of Integrative Biology, University of Wisconsin-Madison, Madison, WI 53706, USA. Supplementary material Supplementary material is available at Molecular Biology and Evolution online. Funding This work was supported by National Science Foundation grant no. IOS-2016141 to PPS. EML was supported by a Hilldale undergraduate research scholarship award at the University of Wisconsin-Madison. Data availability All data associated with this manuscript are published in the supplementary materials. Raw read data for RNAseq experiments are available through the NCBI Sequence Read Archive (SRA) under project number PRJNA1506670. References Associated Data Supplementary Materials Data Availability Statement All data associated with this manuscript are published in the supplementary materials. Raw read data for RNAseq experiments are available through the NCBI Sequence Read Archive (SRA) under project number PRJNA1506670.