Hedgehog signaling controls segmentation dynamics and diversity via msx1 in a spider embryo

A genome-wide study of Hedgehog signaling reveals a previously unknown mechanism of segmentation in a chelicerate arthropod.


INTRODUCTION
Hedgehog (Hh) signaling is one of the most important signaling pathways in animal development. It mediates the formation of polarity and patterning in cellular fields of various animal tissues, such as Drosophila segments and imaginal discs (1)(2)(3)(4) and the vertebrate neural tube and limb buds (5,6). The Hh signaling pathway is composed of tens of components that function as complex molecular networks with positive and negative feedbacks. The main components of this pathway include the secreted signaling protein Hh, its receptor Patched (Ptc), which also acts as a negative regulator, the membrane signal transducer Smoothened (Smo), and the signaling effector transcription factor Cubitus interruptus (Ci, Gli in vertebrates) (7). Gradients of concentration-and time-dependent Hh signaling activities are thought to provide positional values to cells in the entire tissue field, leading to the establishment of discrete domains and repetitive stripes (8)(9)(10)(11). In this process, transcriptions of genes respond to different thresholds of the graded signaling activities, which specifies various spatial regions in the tissue field (10,11). The concentration-dependent effects of diffusible Hh on cell fate specification support the categorization of this molecule as a morphogen, which is assumed in the French flag model (12). However, the time dependency of the signaling effects and the positive and negative feedbacks reflect the mechanistic complication of the system (10,11). Further complications arise from the dynamics of tissue fields, where Hh signaling functions for patterning. Cell proliferation, cell lineages, cell behaviors, and spatial organization may have limited the understanding of the patterning activities of Hh signaling.
The early embryo of the spider Parasteatoda tepidariorum (synonym, Achaearanea tepidariorum) (13) forms a radially symmetrical germ disc in a monolayer at the spherical surface, where concentric gene expression patterns occur, reflecting the polarity of the anterior-posterior (AP) axis. These patterns, which are French flag-like, are established by a mechanism mediated by Hh signaling (14) (Fig. 1A). The spider hh homolog, Pt-hh (aug3.g4322, formerly called At-hh), is zygotically expressed in cells of the future anterior that is located opposite to the embryonic pole (the center of the forming germ disc) (14) and the rim of the formed germ disc (Fig. 1B). Following a symmetry-breaking event (15,16), the germ disc undergoes a relatively simple planar remodeling and is transformed into an axially growing germ band, in which spatially repetitive stripes of gene expression associated with body axis segmentation occur in three different modes depending on the region of the body (Fig. 1B) (17). The repetitive stripes fail to be formed in Pt-hh knockdown embryos (14). Pt-hh is one of the earliest genes that are expressed in repetitive stripes. The anterior Pt-hh stripe, which originates from the germ disc rim and travels posteriorly in the nascent germ band, undergoes repeated bi-splitting in the head region (18), whereas oscillatory Pt-hh expression waves sweep successively in the opisthosoma that is derived from the central region of the germ disc. Striped Pt-hh expression in the thoracic region, however, appears substantially later. Instead, Pt-noto1 (aug3.g15243) stripes simultaneously appear in this region at the earliest timing (17). The stripes of Pt-hh formed in the germ band (19) are presumably associated with the formation of segment polarity. Thus, the early P. tepidariorum embryo provides a simple cell-based platform where the patterns reflecting the AP polarity and the repetitive stripes associated with the body segments are formed through Hh signalingmediated mechanisms. To date, little is known about mechanisms that link the initial Hh signaling activity with the periodic stripe pattern formation.
The segmental stripe pattern along the AP axis is achieved by several distinct modes in arthropods (20,21); however, the functions of Hh signaling in the earliest step of AP patterning remain unclear in species other than Parasteatoda. In the long germ insect Drosophila, the hierarchical interactions of maternal, gap, and pairrule genes, most of which encode transcription factors, simultaneously subdivide the syncytial embryonic field into spatially repetitive units, wherein the polarity is further specified by segment polarity genes including hh (1)(2)(3). This segment polarity function seems a conserved feature of arthropods, since the hh genes are expressed in the posterior part of each segment in a wide range of arthropods (22,23). Although the expression of hh in the anterior region of the 1 early embryo has been reported in some arthropod species (1,23), the functional significance of the early hh expression remains unclear except that observed in the spider Parasteatoda.
Here, we used the spider embryo to investigate spatiotemporal changes in gene expression downstream of Hh signaling. Subsequently, we characterized mechanistic links between formation of the Hh signaling-mediated field polarity and the dynamics and diversity of segmentation-related gene expression.

Hh signaling components in the Parasteatoda genome
We performed Basic Local Alignment Search Tool (BLAST) searches against the Parasteatoda genome and annotated genes for Hh sig-naling components using the amino acid sequences from Drosophila. For those that are not present in the Drosophila genome, the sequences of the human were used. The spider genes that had hits with substantial e-values were then used reciprocally in BLAST searches against Drosophila or human proteins to examine whether the spider genes showed the highest hits to the original proteins. These searches revealed that the Parasteatoda genome encodes all core components of the Hh signaling pathways that have been described in Drosophila and vertebrates (table S1) (24,25). Some of these genes, including hh, are present in duplicate. Of the two hh homologs, only Pt-hh (aug3.g4322) is expressed at substantial levels during embryogenesis (26). The other relevant genes, such as patched (Pt-ptc, aug3.g14374; formerly known as At-ptc) (14), smoothened (Pt-smo, aug3.g8490; At-smo) (14), and cubitus interruptus (Pt-ci, aug3.g13321; At-ci) (18), are present as single copies.
Genome-wide searches for genes responding to Pt-hh and Pt-ptc knockdown Parental RNA interference (pRNAi)-mediated knockdowns of Pt-hh and Pt-ptc have been shown to produce contrasting polarity defects with the highest penetrance ( Fig. 1A) (14), where the jellyfish green fluorescent protein (gfp) gene served as the control. Combining these knockdowns with RNA sequencing (RNA-seq), we conducted genome-wide searches for genes that transcriptionally responded to altered states of Hh signaling at early stage 3 and late stage 5 to comprehensively understand Hh signaling-mediated patterning in the early spider embryo ( Fig. 2A). In early stage 3, the embryo presents morphological asymmetry and expresses Pt-hh on the future anterior side, and in late stage 5, the germ disc reveals region-specific gene expressions (14). RNA was extracted from eggs laid by the same females before and after double-stranded RNA (dsRNA) injection, and the RNA-seq, genome mapping, and comparison of gene expression levels were performed to identify the differentially expressed genes (DEGs; see Fig. 2A, data files S1 to S4, and Materials and Methods).
To validate the methods, we first looked at mapping data at the Pt-hh and Pt-ptc loci. At the Pt-hh locus, the number of mapped reads were drastically reduced in all Pt-hh pRNAi samples, and Pt-hh was consistently detected as a DEG at early stage 3 and late stage 5 (data file S3). On the other hand, in Pt-ptc pRNAi experiments, Pt-ptc itself was detected as a DEG at early stage 3 but not at late stage 5 (data file S4). As ptc is a target of Hh signaling and acts as a negative regulator of Hh signaling (7), Pt-ptc knockdown likely promotes Hh signaling and activates the transcription of Pt-ptc. Nonetheless, the previous study showed that Pt-ptc transcripts were less detectable in the cytoplasm in Pt-ptc pRNAi embryos, indicating that the transcription was active but transcripts were unstable in the cytoplasm (14). These data suggested that the state of Hh signaling was expectedly altered by Pt-hh and Pt-ptc pRNAi.
We also examined mapping data at the loci of known target genes (14,18). At the loci of Pt-otd (aug3.g9172), Pt-lab-A (aug3.g7954), and Pt-opa (aug3.g12202), numbers of mapped reads were all affected by Pt-hh and Pt-ptc pRNAi at late stage 5, as expected ( fig. S1). These genes were also ranked highly in the list of DEGs (data files S3 and S4), further validating our experiments. In contrast, numbers of reads mapped to the Pt-Delta (aug3.g25248) locus were not clearly altered (fig. S1), and this gene was not listed among DEGs in stage 5 experiments (data files S3 and S4), despite our previous observations that expression of Pt-Delta in the germ disc is regulated by Hh signaling ( Fig. 1A) (14). This discrepancy is because additional expression   , Pt-Delta (red), and Pt-noto1 (green) transcripts and DNA (shown only in right panels). Approximate times AEL (top left) and surface cell counts (bottom right). Right: Signal intensity profiles in the surface cell layer along the midline (white line, left; averaged across 20-m width); extracted images (80-m wide) are shown together. Yellow lines refer to the center of the germ disc and the likely corresponding position (39 to 45 hours) and the posterior limit of the Pt-Delta L4 stripe (45 to 51 hours). Blue arrows indicate the anterior Pt-hh stripe(s) that undergo traveling and splitting (45 to 48 hours), green dots indicate Pt-noto1 stripes that appear simultaneously in the thorax (48 and 51 hours), and white arrows mark Pt-hh stripes generated through oscillations for L4 and O1 segments (45 to 51 hours). Scale bars, 200 m.

of 15
in cells of the yolk side is independent of Hh signaling regulation (14,19). Together, these results indicated that our list of DEGs included a substantial number of genes that are regulated by Hh signaling.
At a false discovery rate (FDR) threshold of adjusted P value < 0.01 in both Pt-hh and Pt-ptc pRNAi experiments, 104 aug3 gene models were identified as differentially expressed at late stage 5 ( Fig. 2B and table S2). However, we found that the 104 gene models included at least six redundancies. Specifically, aug3.g4422 and g22667 sequences were overlapped by more than 750 bp, aug3.g416 and g.417 were derived from the coding region and the 3′ untranslated region of Pt-snail (formerly At.snail), respectively (27). Evidence from reverse transcription polymerase chain reaction (PCR) and rapid amplification of cDNA ends analyses suggested that pairs of gene models aug3.g1485-g1487, aug3.g24611-g27568, aug3.g27747-g27748, and aug3.g407-g22741 were derived from single genes. Thus, the DEG count of 104, for the experiments with late stage 5 embryos, was reduced to 98. On the other hand, only 19 DEGs were identified in early stage 3, and these were not among the 98 late stage 5 DEGs (Fig. 2B). These data suggested that early stage 3 and late stage 5 embryos are in different phases of polarity formation.

Characterization of genes identified in the stage 5 pRNAi and RNA-seq experiments
The 98 late stage 5 DEGs were categorized into four classes based on their responses to the Pt-hh/Pt-ptc pRNAi as follows: class I, down-/ up-regulated; class II, up-/down-regulated; class III, down-/ down-regulated; class IV, up-/up-regulated by the respective pRNAi. Expression patterns of all corresponding transcripts were examined in wild-type (WT) late stage 5 embryos by whole-mount in situ hybridization (WISH). These experiments demonstrated that approximately half of the genes (48 genes) had region-specific expression in the germ disc epithelium, and many of their expression domains exhibited concentric circular patterns (Fig. 2, C to E, table S2, and data file S5). Expression trends were correlated between the class of genes and the region of gene expression, although no specific expression patterns were detected for class IV genes. A total of 18 genes, of the 34 class I genes, were expressed in the peripheral region of the germ disc (e.g., Pt-lab-A expression; Fig. 2E), and 14 of the 23 class II genes were expressed in the central region (e.g., Pt-msx1 and Pt-gas1 expression; Fig. 2E). Further, 10 of the 26 class III genes were expressed in the intermediate region (e.g., Pt-frizzled4 expression; Fig. 2E). Notably, most expression domains were diminished or expanded by Pt-hh and Pt-ptc pRNAi, as seen by the quantified responses of transcript expression levels ( Fig. 2, C to E, table S2, and data file S5). Simultaneous detection of transcript expression of multiple genes by fluorescence in situ hybridization (FISH) revealed that the expression domains of these genes were abutted and overlapped, with peak areas located at varied distances from the germ disc rim, where Pt-hh is expressed (Fig. 2F). These expression patterns were reminiscent of the "French flag" (12), although with less clear spatial boundaries. These results indicated that a variety of cell states was specified by Hh signaling, according to their positions from the germ disc rim.
In addition, 10 genes from classes I to III were expressed at the cumulus (Fig. 2C, table S2, and data file S5), where the cumulus mesenchymal (CM) cells, a cell cluster that functions in dorsalventral axis formation, are migrating underneath the surface epithelial cells (15,16). Hh signaling was found to be involved in regulating the gene expression in the epithelial cells associated with the CM cells, as was evident from the effects of Pt-hh and Pt-ptc pRNAi on aug3.g17729 (Pt-noggin-D) expression (Fig. 2E). In contrast, altered Hh signaling did not seem to affect the gene expression in the CM cells ( Fig. 2C and data file S5). Furthermore, Pt-fgf8 is expressed rather stochastically, but the transcripts tended to be detected near the cumulus (Fig. 2E).

Identification of Pt-msx1 as a segmentation gene
Class I genes that were positively regulated by Hh signaling included Pt-otd (aug3.g9172) and Pt-opa (aug3.g12202). These genes have been previously characterized as Hh signaling targets that mediate stripe-splitting dynamics in head segmentation (14,18,28). In contrast, class II genes were negatively regulated by Hh signaling and have not been characterized previously. To investigate novel aspects of patterning mechanisms, we performed expression analysis of class II genes ( fig. S2) and a pilot functional screen using pRNAimediated knockdown (29). From these analyses, we explored one class II gene aug3.g15189 (Pt-msx1), which encodes a homeodomain transcription factor and is a homolog of the Drosophila muscle segment homeobox (msh) gene and vertebrate Msx genes (30).
Expression analysis of class II genes revealed Pt-msx1 as one of the genes expressed in the central region of the germ disc at late stage 5 (fig. S2, A and B). Its expression was expanded to almost the entire region in the Pt-hh pRNAi germ disc and was missing in Pt-ptc pRNAi (Fig. 2E). Among class II genes, Pt-msx1 and Pt-ci were the only genes that showed spatially repetitive stripes of expression in WT embryos by 50 hours after egg laying (AEL; fig. S2, A to C). In the presumptive opisthosomal region of the nascent germ bands, Pt-msx1 expression exhibited a dynamic pattern similar to that of Pt-hh (Fig. 3A), indicating that it could behave as oscillatory traveling waves, whereas Pt-ci and most other class II genes did not ( fig. S2). The complementary patterns of Pt-hh and Pt-msx1 expression were observed at the earliest stage of oscillation ( Fig. 3B) and were eventually established throughout the entire germ band (Fig. 3C).
Knockdown experiments revealed that Pt-msx1 pRNAi led to failure to form germ bands and repetitive structures, despite the normal appearance of the germ disc ( Fig. 4A and movie S1). Pt-msx1 pRNAi embryos initiated the differentiation of the extraembryonic region in response to the arrival of the CM cells at the germ disc rim but did not fully expand it and remained in a fan-like shape without elongating the emerging AP axis, even at the stage corresponding to the limb-forming stage (stage 9; Fig. 4A and movie S1). The caudal part of the embryonic region of Pt-msx1 pRNAi embryos started to show thickening after the beginning of stage 7; although the surface epithelial cells appeared morphologically normal, in the interior, some cells appeared to be dying (Fig. 4B). Pt-msx1 pRNAi specificity was carefully confirmed ( fig. S3), and the phenotypes of Pt-msx1 pRNAi embryos were then assessed at the molecular level. In Pt-msx1 pRNAi germ discs, the expression of Pt-hh and Pt-ptc and formation of AP polarity were relatively normal (Fig. 4C), in contrast to Pt-hh pRNAi and Pt-ptc pRNAi germ discs ( Fig. 1A and data file S5). In the Pt-msx1 pRNAi embryos at later stages, the primary Pt-hh stripe traveled normally from the rim of the germ disc, but in the nascent germ bands, the stripe failed to enter the splitting phase, and the expression of Pt-hh and Pt-otd became obscure in the head (Fig. 4D). In the thorax of the nascent WT germ band, Pt-noto1 and Pt-hairy are expressed in stripes (17,31). In the Pt-msx1 pRNAi embryos, Pt-noto1 expression was detected in the putative thoracic region, disappearing from the most posterior part. In contrast, Pt-hairy expression was persistent in the posterior region. Both genes, however, did not show the spatially periodic expression (Fig. 4D). Notably, in the opisthosoma of the Pt-msx1 pRNAi embryos, oscillatory expression of genes-including Pt-hh, Pt-noto1, and a homolog of Drosophila pair-rule gene, Pt-evenskipped (Pt-eve)-did not occur (Fig. 4D). Moreover, serially arranged expression of hox genes, Pt-lab-A, Pt-Dfd-A, and Pt-antp-A (25), and Pt-caudal was not properly developed in Pt-msx1 pRNAi embryos (Fig. 4D); the anterior Pt-lab-A expression became reduced. The thoracic gene, Pt-Dfd-A, was expressed in the wide area of the embryonic region without showing any indication of stripes, while the expressions of the posterior genes, Pt-antp-A and Pt-caudal, were detected in only a few cells. In addition, Pt-Ap2-A (aug3.g23531) was expressed in a broad posterior region (Fig. 4D).
The normal germ disc displays a "salt-and-pepper" pattern of Pt-Delta expression at the central area reflecting opisthosomal mesoderm versus ectoderm fate decisions (19). As the germ disc transition toward the germ band, this Pt-Delta pattern disappears from the germ disc center, followed by segmental striped expression of Pt-Delta in the opisthosomal ectoderm ( Fig. 1B and fig. S2D) (14,19). In Pt-msx1 pRNAi embryos, the salt-and-pepper pattern of Pt-Delta expression was normal but persistent and failed to shift to the segmental mode (Fig. 4E). Similarly, the expression pattern of Pt-ci was affected by Pt-msx1 knockdown, with neither central attenuation nor transition to stripes (Fig. 4E). These observations suggested that Pt-msx1 is not necessary for forming graded positional information in the cell field but is essential for generating spatial repetitions of gene expression.

Regulation of periodic expressions of genes by Pt-msx1
To examine gene regulation by Pt-msx1, we performed RNA-seq analysis of Pt-msx1 pRNAi embryos, followed by WISH-based gene expression screening of 40 selected genes (Fig. 5A, table S3, and data file S6). These included nine genes that were also identified as DEGs in the Pt-hh and Pt-ptc pRNAi and RNA-seq experiments at late stage 5 (Fig. 5Aa). Of these nine genes, the intermediate expressions of three class III genes, aug3.g8295 (Pt-mirr4), aug3.g21549 (Pt-oddskipped related-B), and aug3.g13482 (Pt-rhomboid), were expanded into the central region of the germ disc by Pt-msx1 knockdown; in contrast, the expressions of two class II genes, aug3.g9575 (Pt-gas1) and g27670, were reduced within the central region (Fig. 5B). These results suggested that the germ disc patterning involved regulatory interactions among Hh signaling targets.
Wave-like expression of Pt-msx1 preceding stripe pattern formation Next, we characterized events forming the striped expression of Pt-msx1 at cellular resolution. Transcriptomic profiling by RNAseq showed that Pt-msx1, similar to Pt-hh, started to be transcribed at stage 2 (12 to 14 hours AEL) (26). By FISH, Pt-msx1 transcripts were not detectable at 24 hours AEL but were detected stochastically at 28 and 30 hours AEL among surface cells near the embryonic pole or the center of the nascent germ disc ( fig. S5), which is on the opposite side of the egg relative to the position of initial Pt-hh transcription (14).
We then fixed sibling embryos, serially from the early germ disc stage, and first examined the spatial distributions of cells expressing Pt-msx1, relative to cells expressing another class II gene, Pt-AP2-A (aug3.g23531; Fig. 6 and fig. S5B). Observations from these experiments revealed three successive phases of Pt-msx1 expression. During the first phase, cells expressing these two genes were initially found only near the embryonic pole but covered up to more than half of the diameter of the germ disc, by 36 hours AEL. Extirpation of the embryonic polar region by laser irradiation at 24 hours AEL failed to prevent expansion of Pt-msx1 expression in cells in the broad central area of the later germ disc (Fig. 7A), suggesting that Pt-msx1 expression is independent of the potential signaling center at the    embryonic pole (34). In contrast, Pt-ptc knockdown completely abolished Pt-msx1 expression (Figs. 2E and 7B), and Pt-hh knockdown caused overexpansion of Pt-msx1-expressing cells even at earlier germ disc stages (33 hours AEL; Fig. 7B). Hence, expansion dynamics of Pt-msx1 expression were regulated by Hh signaling.
During the second phase, Pt-msx1, and not Pt-AP2-A, expression started to be attenuated in cells in the central area of the germ disc (39 hours; Fig. 6). Cells expressing both Pt-msx1 and Pt-AP2-A decreased in number (42 hours) and more peripheral (anterior) cells exhibited Pt-msx1-biased expression. As revealed by cell labeling experiments, Pt-msx1 expression was attenuated in a static field of cells (Fig. 7C). This indicated that a large-scale wave of Pt-msx1 expression developed from around the embryonic pole, which corresponded to the future posterior pole, and swept toward the thorax. Notably, the cells in which the Pt-msx1 expression was attenuated initiated Pt-hh expression (Fig. 3B). Furthermore, Pt-ci knockdown embryos were defective in the central attenuation of Pt-msx1 expression (Fig. 7D), as in the case with Pt-ci expression in Pt-msx1 knockdown embryos (Fig. 4E). These observations suggested that Parasteatoda embryonic segmentation involved mutually opposing regulations of Pt-msx1 and Hh signaling through gene networks that have not yet been characterized. Similar wave-like behaviors of gene expression patterns, starting from around the embryonic pole in the germ disc stage, were observed for eight more class II genes and the Pt-Delta gene (14,19), with varied timing and expansion range ( fig. S2, A and D). Here, these early wave-like gene expressions were referred to as the primary waves. With remodeling of the field into a germ band, the Pt-msx1 and Pt-AP2-A expression domains were spatially separated, with the exception of a de novo Pt-msx1 peak for the first opisthosomal (O1) segment (45 to 51 hours; Fig. 6).
During the third phase, the primary Pt-msx1 wave, after passing through the future opisthosoma and arriving at the future thorax, split into three stripes in a short time frame (48 and 51 hours; Fig. 6). These stripes corresponded to anterior portions of the second to fourth leg-bearing thoracic (L2 to L4) segments. This "tri-splitting" can be compared with "bi-splitting" of the Pt-hh stripe during head segmentation (18), but the former was temporally nonrepetitive, and the latter was temporally repetitive.
We reconstructed the tri-splitting process by staining sibling embryos for Pt-msx1 and Pt-noto1 (Fig. 8, A to D, and fig. S6). Pt-noto1 is one of the earliest indicators of thoracic segmentation (Fig. 1B) (17) and is expressed ubiquitously in the presumptive thoracic region of the germ disc in early embryos ( fig. S2D). In this reconstruction, the Pt-msx1 primary wave merged with Pt-noto1 expression   in the thoracic field, and, subsequently, complementary striped distribution of Pt-msx1-and Pt-noto1-expressing cells was generated, as double-positive cells disappeared. Considering previously described cell tracking data (17), this tri-splitting was likely elicited through transcriptional regulation in the epithelium, with limited contribution from cell sorting. To confirm the roles of Pt-msx1 in tri-splitting, we introduced Pt-msx1 embryonic RNAi (Pt-msx1 eRNAi ) cell clones into the presumptive thoracic region. These clones locally impaired the patterning event (15 of 17), whereas gfp eRNAi cell clones did not (0 of 10; Fig. 8E). Pt-msx1 eRNAi cell clones did not develop spatially organized, complementary expressions of Pt-msx1 and Pt-noto1. These data confirmed the functional importance of the Pt-msx1 wave in thoracic segmentation.

The onset of oscillatory expression in the caudal region
The field through which the primary Pt-msx1 wave passed initiated cyclic expression of segmentation clock genes ( fig. S4). In particular, the propagation and attenuation dynamics of the primary Pt-msx1 expression wave were closely associated with the attenua-tion of Pt-ci expression and the initiation and propagation of expression of serial clock genes, including Pt-hh, Pt-prd2, and Pt-krü-1 (Fig. 9, A and B). All these expression dynamics were affected in Pt-msx1 pRNAi embryos (Figs. 4 and 5). To confirm the roles of Pt-msx1 in regulating these dynamics, we examined Pt-msx1 eRNAi cell clones that were introduced to the central region of the germ disc. These cell clones, but not gfp eRNAi cell clones, prevented both Pt-ci attenuation and Pt-krü-1 initiation (6 of 6; 0 of 5 for gfp eRNAi ; Fig. 9C). In the representative Pt-msx1 eRNAi cell clone, most cells remained Pt-ci positive and did not initiate Pt-krü1 expression despite the progression of Pt-ci attenuation outside of the cell clone (Fig. 9C). This was in contrast with the gfp eRNAi cell clone, which included cells expressing Pt-ci, Pt-krü1, or both according to the progression of Pt-ci attenuation (Fig. 9C). These results suggested that the primary Pt-msx1 wave is among the key mechanisms that underlie the initiation of the genetic oscillations associated with opisthosomal segmentation. In addition, as mentioned above, Pt-msx1 attenuation was abolished in Pt-ci pRNAi embryos (Fig. 7D), suggesting that interdependent negative regulation of Pt-msx1 and Hh signaling is a prerequisite for initiating the patterning cycles.   , which mark Pt-msx1 expression for the O1 segment. Yellow lines refer to the center of the germ disc and the likely corresponding position (33 to 45 hours) and the posterior limit of the Pt-msx1 L4 stripe (45 to 51 hours). Gray dots, three Pt-msx1 stripes (L2 to L4) resulting from tri-splitting of the primary Pt-msx1 wave; blue arrows, Pt-msx1 stripes at the bi-splitting sites (A and C). Anterior is to the left (36 to 51 hours). Germ disc orientation is arbitrary at 33 hours. Scale bar, 200 m.

Regulation of bi-splitting dynamics by Pt-msx1 in the head segmentation
In the presumptive head region, temporally repeated bi-splitting of Pt-hh stripes gives rise to head segments (18). Expression analysis revealed a nascent stripe of Pt-msx1 expression in the head region ( Figs. 6 and 8), and the Pt-msx1 pRNAi embryos exhibited reduced expressions of head genes (Figs. 4D and 5C). Simultaneous detection of Pt-hh and Pt-msx1 expression in sibling embryos revealed that Pt-msx1 was dynamically expressed in association with the bi-splitting of Pt-hh stripes (Fig. 9D). In particular, cells expressing Pt-msx1 were observed within the Pt-hh stripe preceding each bi-splitting event, and Pt-msx1 expression expanded to occupy the space between the splitting stripes. We found that Pt-msx1 eRNAi cell clones failed to abolish Pt-hh expression between the splitting stripes (15 of 15; 0 of 9 for gfp eRNAi ; Fig. 9E). Pt-msx1 eRNAi cell clones did not induce Pt-hh expression outside of the Pt-hh stripes. These findings suggested that Pt-msx1 acts as a negative regulator of Pt-hh expression to mediate the generation of bi-splitting dynamics for head segmentation.

DISCUSSION
The present genome-wide analyses of P. tepidariorum, using a combination of Pt-hh and Pt-ptc pRNAi and RNA-seq, have provided comprehensive evidence that Hh signaling mediates the formation of a polarized French flag-like pattern in a cell-based field of the germ disc in the spider embryo (Fig. 2). This pattern comprised concentric circular expressions of class I to III genes, which were achieved through differential transcriptional responses to the initial Hh signaling activities [ Fig. 10 (1)]. Transcription of class I genes was positively regulated by Hh signaling, that of class II genes was negatively regulated, and that of class III genes required both Pt-hh and Pt-ptc. Pt-hh is expressed at the periphery of the germ disc corresponding to the future anterior ( Figs. 1 and 2). The radius of the germ disc is about 300 m, which is comparable to the effective range of Hh molecules shown in vertebrate studies (35,36). Although any protein products of Hh signaling components have not been examined in the spider embryos, our previous observation regarding the changing pattern of Pt-ptc transcript expression (14), which is generally considered to be the indicator of Hh signal reception (7), supports the notion that the Hh signaling network operates in the full spatial range of the spider germ disc. Our extensive gene expression analyses combined with the gene knockdown experiments revealed the direct involvement of Hh signaling in generating spatially organized diverse cell states in the entire germ disc before the emergence of spatially periodic patterns.
The French flag-like pattern in the spider germ disc seems to be comparable with the expression pattern of Drosophila gap genes in the blastoderm, which is regulated by the graded concentration of Bicoid (37). Despite the differences in the nature of the patterning fields (syncytial versus cellular) and in the type of the master molecules (the transcription factor, Bicoid, versus the secreted signaling protein, Hh), similar expression patterns of a large spectrum of genes are displayed in both static fields. At first, Drosophila gap gene expression domains overlap, but sharp boundaries are established depending on mutually repressive interactions between gap genes (37). In the spider, clear boundaries of gene expression domains were not observed in the germ-disc stage (Fig. 2), but interactions among target genes appeared to occur. For example, the expression patterns of some of class II and class III genes were reduced and expanded, respectively, by knockdown of a class II gene, Pt-msx1 (Fig. 5). The periodic stripe pattern of Drosophila pair-rule genes develops in the static field (20). In contrast, spatial repetition in the spider embryo became apparent only after transition of the germ disc toward the germ band began (Figs. 1, 6, and 8).
A prominent aspect of Hh signaling-based patterning, which is directly associated with periodic stripe formation, was revealed by our findings on the expression and function of the transcription factor gene Pt-msx1, which is a class II gene. As summarized in Fig. 10, Pt-msx1 exhibited expression associated with three types of waves and mediated the formation of spatially repeated stripes in all the body regions. Hence, Pt-msx1 is a key mediator of the initial Hh signaling activity and downstream patterning system. The primary Pt-msx1 wave arose from around the center of the germ disc (corresponding to the posterior pole of the later embryo) [ Fig. 10 (2  propagated to the future thoracic region in the nascent germ band [ Fig. 10 (3 and 4)], and then split into three stripes complementary to those of Pt-noto1 (tri-splitting) [ Fig. 10 (4 and 5)], thereby specifying the thoracic segments [ Fig. 10 (5)]. In the future opisthosomal region, where the primary Pt-msx1 wave passed away due to attenuation, oscillatory expressions of segmentation clock genes, including Pt-hh and Pt-msx1, followed, which temporally behaved as traveling waves [ Fig. 10 (4 and 5)]. In the future head region, the Pt-hh stripe initiated at the rim of the germ disc traveled posteriorly and underwent two rounds of bi-splitting, each of which is preceded by an onset of Pt-msx1 expression in the Pt-hh stripe [ Fig. 10 (2) to (5)]. Collectively, the dynamics of Pt-msx1 expression under the control of Hh signaling are distinct depending on the body regions of the spider embryo, and all are related to segmentation. The primary expression waves of Pt-msx1 and certain class II genes started around the site corresponding to the posterior pole and then propagated toward the future thorax (fig. S2A). These wave propa-gations were concurrent with that of Pt-Delta, which was described in the previous studies (14,19). This type of phenomenon, where expression of negatively regulated targets of Hh signaling starts at the side opposite to the Hh signaling source in a large field and dynamically propagates in the field, has not been reported in other organisms. The laser ablation experiment suggested that cells at and around the expression start site made little contribution to the expansion of Pt-msx1 expression (Fig. 7A). Instead, involvement of Hh signaling in the regulation was suggested by the pRNAi experiments (Figs. 2E and 7B). Our previous model, explaining the expansion dynamics of Pt-Delta expression, incorporated the role of Pt-ptc in altering the effective distance of Hh molecules from longer to shorter (14) and may be applicable to the expanding expressions of class II genes. However, because neither Hh protein distribution nor Hh receptor behavior has been examined, the ideas remain hypothetical. For the attenuation dynamics, data from Pt-ci and Pt-msx1 knockdown suggested the presence of mutually repressive interactions between Pt-msx1 and Hh signaling (Figs. 4E, 7D, and 9C). Overall, the dynamic response and wave-like behavior of the Hh signaling activities that occur in the surface epithelial cell sheet of the germ disc and germ band are easily accessible, and the simple features of the spider embryonic field may be advantageous for the study of Hh signaling-mediated patterning dynamics. The tri-splitting of the primary Pt-msx1 wave presented the segmentation dynamics in the thoracic region described in this study (Figs. 6 and 8). The requirement of Pt-msx1 for this phenomenon was shown by the eRNAi experiment (Fig. 8E). Although expression of Pt-hh transcripts in this region started later than tri-splitting, it appeared at the cells expressing Pt-noto1 and eventually formed a complementary pattern to the Pt-msx1 stripes (Fig. 3). Ahead of, or simultaneous with, the propagation of the Pt-msx1 wave, the thoracic field displayed a French flag-like pattern, which was likely to be under the control of Hh signaling (Fig. 2). It is possible that genes expressed to form the French flag-like pattern contribute to the tri-splitting phenomenon cooperatively with Pt-msx1. Although the Pt-hairy gene, described in a previous study (31), presents expression dynamics rather similar to those of Pt-msx-1 in the thoracic and opisthosomal regions, its expression pattern development was found to depend on Pt-msx1 (Fig. 4D). Interactions between Hh signaling downstream activities should be addressed in future studies to better understand the Hh signaling-based patterning systems. Oscillation is the segmentation dynamics observed in the opisthosomal region. FISH staining showed that expressions of multiple clock genes, including Pt-msx1 and Pt-hh, repeatedly oscillated in the cells of the opisthosoma, exhibiting repeated wave-like propagation in one segment periodicity in this opisthosomal field ( fig. S4). The first oscillatory waves followed the primary waves for Pt-msx1 and other class II genes that propagated toward the thorax, and they did not occur in the Pt-msx1 knockdown embryos (Figs. 4, 5, and 9). Oscillation is a common strategy to establish a spatially periodic stripe pattern, as seen in somitogenesis of vertebrates and segmentation of short germ insects (20,33). However, the initiation mechanisms for the first oscillatory waves had been rarely studied. The spider embryo thus provides a unique opportunity to analyze the onset of oscillation. Moreover, the clock genes identified in the Pt-msx1 pRNAi and RNA-seq experiments were family genes of a Drosophila gap gene (Pt-krü-1), a pair-rule gene (Pt-prd2), and basic helix-loophelix-type vertebrate oscillatory genes (Pt-hairy-D and Pt-asl-H) (Fig. 5). Studies of spider segmentation may contribute to filling the gap in our understanding of the periodic stripe-forming mechanisms among animals.
Further, we unexpectedly found that Pt-msx1 was involved even in the repeated bi-splitting of Pt-hh stripes, associated with the head segmentation (Fig. 9). It is most likely that Pt-msx1 acts cooperatively with region-specific factors for head segmentation because Pt-msx1 knockdown cell clones prevented the down-regulation of Pt-hh expression between the splitting Pt-hh stripes but did not induce ectopic Pt-hh expression outside the areas of the splitting stripes. The candidate genes include Pt-otd and Pt-opa, which have been shown to play roles in the bi-splitting event (18,28). Interactions of Pt-msx1 with these genes are subjects of future studies.
Collectively, our findings have demonstrated that Hh signaling controls all the segmentation dynamics via Pt-msx1 in a spider em-bryo. Pt-msx1 was found to be one of the most sensitive genes to the altered states of Hh signaling in the germ disc stage, and mutually exclusive expressions and repressive interactions between Pt-msx1 and Hh signaling components were observed during the bi-splitting and at the onset of the oscillatory waves. We suggest a deep origin of involvement of msx1 and hh in segmentation because striped expression of a msx1 homolog is conserved in harvestman embryos (30) and that of hh homologs is in a wide range of arthropods (14,22,23). This study provides a new paradigm, beyond the Drosophila-based knowledge framework, of segmentation and lays the foundation for comparative investigations of segmentation mechanisms across the arthropod phylum.
Evolvability of Hh signaling-based patterning systems has attracted increasing attention in terms of animal form diversification (38)(39)(40)(41). Mathematical models have previously proposed that a minimal mechanism of Hh signaling coupled with a regulatory network of transcription factors confers various pattern-forming capabilities (42). This proposal might be potentially applicable to spider embryos. It is required to investigate whether Pt-msx1 cooperates with region-specific transcription factor genes to achieve segmentation diversity.

Spider egg collection
The animal experimentation was conducted according to the protocol reviewed and approved by the Institutional Animal Care and Use Committee of JT Biohistory Research Hall (2005-5). We used laboratory stocks of the spider P. tepidariorum (synonym, A. tepidariorum), which were kept at 25°C with a 16-hour light/8-hour dark cycle. Developmental stages have been described previously (15). Developmental time AEL was estimated from the beginning of stage 2 (10 hours AEL; the nuclei reach the surface of the egg) or the beginning of stage 5 (30 hours AEL; internalization of the CM cells is completed). All mated females were evaluated by monitoring development of their embryos to exclude data derived from nonreliable embryos; morphology of germ discs at stage 5 and that of germ bands at stages 7 and 9 was examined. In experiments using earlier embryos, development of their siblings was examined. In pRNAi experiments (see below), embryos were checked whether they form morphologically normal germ discs.
We prepared 12 independent (derived from different egg sacs) groups of sibling embryos fixed at multiple time points. These embryo groups were used for (i) Figs fig. S2 were derived from a total of seven egg sacs and were examined by Pt-msx1 staining to confirm that the developmental stages were consistent among them.

Head
Thorax Opisthosoma (1)  At_eW_020_B15 (LC379630.1). In pRNAi experiments, dsRNA was injected at a concentration of 2.0 g/l. All females were given four injections of dsRNA solution (1 to 2 l each) at 2-to 3-day intervals. Spiders lay eggs repeatedly every about 5 days, and the effect of ds-RNA injections usually appears in the eggs laid after 10 days from the first injection and in the peak around 20 days after the first injection. Eggs that were laid at 13 days or later from the day of the first dsRNA injection were used as pRNAi samples.
To verify the specificity of RNAi for Pt-msx1, we confirmed that injections of Pt-msx1-1 and Pt-msx1-2 dsRNAs produced the same morphological phenotypes with complete penetrance. The defects of transition to the germ band were observed in 7 of 7 females injected with Pt-msx1-1 dsRNA and in 3 of 3 females injected with Pt-msx1-2 dsRNA (fig. S3A). We also confirmed that the same gene expression phenotypes were obtained with both dsRNAs in WISH analyses of the transcripts shown in Figs. 4 and 5. In this paper, the presented Pt-msx1 knockdown data, including those of RNA-seq, were generated using Pt-msx1-1 dsRNA unless stated otherwise.
eRNAi experiments were performed as previously described (18, 28) using msx1-1 dsRNA and gfp dsRNA. Briefly, a mixture of dsRNA (0.4 g/l), Rhodamine B isothiocyanate (RITC)-dextran [0.8 g/l; weight-average molecular weight (M w ) of 70,000; Sigma-Aldrich, R9379], and biotin-dextran (0.8 g/l; M w of 10,000; Sigma-Aldrich, B9139) was microinjected into single blastomeres of 32-or 64-cell stage embryos, which were then allowed to develop until stage 6 or 7 and were then fixed for staining. RITC-dextran was used to visualize injected cell clones in living embryos, whereas biotin-dextran was used to visualize cell clones in the fixed samples.

RNA sequencing
Poly(A) RNA was extracted from 50 to 120 embryos at early stage 3, late stage 5, and early stage 6. These embryos were produced by single females before dsRNA injection (the first egg laying) and at 13 to 24 days after the first dsRNA injection. DEGs were identified from comparisons of the former control samples and the latter test (pRNAi) samples. Comparative sets of RNA-seq libraries were generated for Pt-hh, Pt-ptc, and gfp pRNAi-treated embryos at early stage 3 and late stage 5 and for Pt-msx1 and gfp pRNAi-treated embryos at late stage 5 and early stage 6. These libraries were prepared with two or three biological replicates (different females). Poly(A) RNA extraction was performed using the QuickPrep Micro mRNA Purification Kit (GE Healthcare) or the Dynabeads mRNA DIRECT Micro Kit (Veritas, DB61012). mRNA samples were quantified using the Qubit fluorometric RNA HS assay (Life Technologies, Q32852), and whole amounts (20 to 50 ng) were used in subsequent steps. The mRNA was fragmented with the New England Biolabs (NEB)Next RNase III RNA Fragmentation module (NEB, E6146), and samples were then used for RNA-seq library construction using the NEBNext Ultra Directional RNA Library Prep Kit for Illumina (NEB, E7420) and the NEBNext Multiplex Oligos for Illumina (Index Primers Set 1; NEB, E7335). Quantitative PCR were performed using a KAPA or Takara Library Quantification kit (KAPA, 4824; Takara, Z8324N). Libraries were then sequenced in single-end runs in the antisense direction using the MiSeq reagent kit V3 (150 cycles; Illumina, MS-102-3001) on an Illumina MiSeq platform. Between 12 and 24 million reads were generated for each library. Information for constructed libraries and RNA-seq reads is summarized in data file S1.
Sequence read trimming, mapping on the genome, and quantification Raw reads of the MiSeq sequencing were trimmed in three steps using the CLC Genomics Workbench (v. 7.0.4). Adapter trimming, sequence primer trimming, and quality trimming with the following parameter settings: trim using quality scores, limit of 0.05; trim ambiguous nucleotides, maximum number of ambiguities 2; and filter on length, discard reads below length 50. Trimmed reads were mapped onto the P. tepidariorum genome sequence assembly Ptep_1.0 (GCA_000365465) using the alignment tool BLAT (v.34) (43) in the DDBJ Read Annotation Pipeline with default settings. Output alignments were filtered on the basis of quality, coverage, and uniqueness using a perl script filterPSL.pl, which is available from the AUGUSTUS 3.0.1 scripts folder (https://github.com/Gaius-Augustus/Augustus). The parameter settings were as follows: minimum coverage, 60%; minimum percent identity, 90%; and unique threshold, 0.96. Filtered alignments that were in the pslx format were then converted into the psl and sam formats with a perl script written in our laboratory. Using sam files, mRNA abundances against AUGUSTUS gene models (aug3) (25) were calculated using htseq-count v.0.5.4p5 (44) with -s reverse, -m union settings. The results of trimming, mapping, and htseq-counts are summarized in data file S1, and counts per gene are listed in data file S2. Mapping results were visualized in University of Tokyo Genome Browsers (45) using wig files that were converted from the psl files with a script aln2wig, which is available from the AUGUSTUS 3.0.1 scripts folder, and were normalized as counts per 10 million reads.

Identification of DEGs
We used the R software (v. 3.1.2) edgeR_3.8.5 (46) to identify DEGs. In these analyses, pairwise comparisons were performed with datasets from the same parent pair so that pseudo-positives due to individual variations were reduced. Genes with very low counts (less than 1 count per million reads in all samples) were filtered out before running edgeR. Genes with the adjusted P value (FDR) < 0.01 were listed as DEGs after excluding genes with FDR < 0.01 in experiments of gfp pRNAi and RNA-seq at the same developmental stage. In DEG detection analyses using Pt-hh and Pt-ptc pRNAi samples, EdgeR was run with comparisons between control samples and (i) test samples derived from eggs laid at days 13 to 17 after the first injection of dsRNA, (ii) test samples derived from eggs laid at days 18 to 22, and (iii) all test samples. Genes that were identified as differentially expressed (FDR < 0.01) in any of the comparisons were all considered as DEGs (data files S3 and S4), and genes identified as DEGS in both Pt-hh and Pt-ptc pRNAi were selected for further analyses (table S2).
In Pt-msx1 pRNAi experiments, we listed genes with FDR < 0.01 as DEGs (data file S6). In the stage 6 Pt-msx1 pRNAi experiments, we selected a total of 40 DEGs based on the following two criteria: (i) genes with FDR < 0.01 in both Pt-hh and Pt-ptc pRNAi experiments at stage 5 (nine genes) and (ii) genes with the 31 lowest FDR values in Pt-msx1 pRNAi experiments (FDR < 0.0003; table S3). We examined expression in germ disc epithelial cells for genes identified in i and in early germ bands for genes identified in ii (Fig. 5).

cDNA cloning
Full-length or partial cDNAs of identified DEGs were obtained from our laboratory stocks of EST clones (19) or were isolated in PCR. The resulting cDNAs were used to synthesize probes. The cDNA clones and the PCR primers were listed in table S4.