A Stable Thoracic Hox Code and Epimorphosis Characterize Posterior Regeneration in Capitella teleta

Regeneration, the ability to replace lost tissues and body parts following traumatic injury, occurs widely throughout the animal tree of life. Regeneration occurs either by remodeling of pre-existing tissues, through addition of new cells by cell division, or a combination of both. We describe a staging system for posterior regeneration in the annelid, Capitella teleta, and use the C. teleta Hox gene code as markers of regional identity for regenerating tissue along the anterior-posterior axis. Following amputation of different posterior regions of the animal, a blastema forms and by two days, proliferating cells are detected by EdU incorporation, demonstrating that epimorphosis occurs during posterior regeneration of C. teleta. Neurites rapidly extend into the blastema, and gradually become organized into discrete nerves before new ganglia appear approximately seven days after amputation. In situ hybridization shows that seven of the ten Hox genes examined are expressed in the blastema, suggesting roles in patterning the newly forming tissue, although neither spatial nor temporal co-linearity was detected. We hypothesized that following amputation, Hox gene expression in pre-existing segments would be re-organized to scale, and the remaining fragment would express the complete suite of Hox genes. Surprisingly, most Hox genes display stable expression patterns in the ganglia of pre-existing tissue following amputation at multiple axial positions, indicating general stability of segmental identity. However, the three Hox genes, CapI-lox4, CapI-lox2 and CapI-Post2, each shift its anterior expression boundary by one segment, and each shift includes a subset of cells in the ganglia. This expression shift depends upon the axial position of the amputation. In C. teleta, thoracic segments exhibit stable positional identity with limited morphallaxis, in contrast with the extensive body remodeling that occurs during regeneration of some other annelids, planarians and acoel flatworms.


Introduction
Regeneration, the ability to replace lost tissues and body parts following traumatic injury, is present in representatives of most metazoan phyla [1,2]. In some animals, this ability is limited to the replacement of a particular cell type, tissue or structure (e.g. the limb of axolotls, the fins of zebrafish), while others can rebuild their entire bodies from a single piece of tissue (e.g. planarians, cnidarians). Although variation in regeneration ability exists, the widespread phylogenetic distribution of some form of regenerative capability in the Metazoa suggests an evolutionary ancient origin of regeneration with subsequent multiple losses across many lineages.
To enable comparisons when discussing regeneration in diverse animals, T.H. Morgan coined the terms epimorphosis and morphallaxis. Epimorphosis occurs when cell proliferation leads to formation of new tissue, while morphallaxis is characterized by re-patterning of existing tissue, in the absence of cell proliferation [3]. Epimorphic regeneration can involve the production of a blastema, a group of cells of variable potency that appear at the wound site following wound healing. Much of the current regeneration research is devoted to studies of epimorphosis, and focuses on identifying the source of cells that form the regenerated tissue, and determining the mechanisms by which these cells regenerate the appropriate tissues (reviewed in [2,4]). Morphallaxis is primarily known from invertebrate species, and a number of striking cases have been described. For example, to reconstitute the original complement of tissues along the main body axis following transverse amputation, Hydra respecifies and rearranges the remaining tissue (reviewed in [5]), the oligochaete annelid Enchytraeus respecifies its gut [6], the sabellid annelid, Sabella changes identity of some of its segments [7], the acoel Hofstenia regains its transverse pigment stripes [8] and planarians reform their pharynx (reviewed in [9]). The broad phylogenetic distribution of documented examples of morphallaxis suggests that this might be a more widespread phenomenon than previously appreciated.
Annelids are the segmented worms, and have long been studied as models of regeneration, due to their impressive regenerative capabilities. Most species display robust posterior regeneration, a trait that is likely ancestral for annelids [10]. Some species that can regenerate posterior tissues can also regenerate their anterior segments and head. Regeneration in many annelids involves a combination of both epimorphosis (cell proliferation resulting in formation of new tissue) and morphallaxis (reorganization of pre-existing tissue) (e.g. [6,11,12]).
The annelid Capitella teleta can regenerate posterior but not anterior segments following transverse amputation [13,14]. There is little published data on whether posterior regeneration in C. teleta proceeds via epimorphosis, morphallaxis or a combination of the two. A blastema forms within 1-3 days following transverse amputation of adults [14], however thus far, morphallaxis during regeneration of C. teleta has not been reported. C. teleta has a number of morphological features uniquely distributed along the anterior-posterior axis that facilitate studies of patterning during regeneration. There are two distinct body regions composed of nine thoracic segments, and a variable number of posterior abdominal segments. In addition, many segments have unique structures. For example, male reproductive organs are present in segments 7 and 8, and female reproductive organs are present in most abdominal segments [15], the gut is highly regionalized along its length (pharynx, esophagus, midgut and hindgut) [16], a pair of FMRFamide immunoreactive neurons are restricted to segment 5 [17], and there are distinct chaetal morphologies between the thoracic and abdominal segments (simple or hooded hook chaetae, respectively) [18].
A previous study of C. teleta Hox genes-a family of transcription factors that define regional identities along the anterior-posterior axis during development in a wide variety of animalsrevealed molecular differences along the anterior-posterior axis [19]. In C. teleta, there are 11 distinct paralogous groups, each containing a single gene, and at least 8 Hox genes are linked in a single cluster within the genome. Hox genes in C. teleta larvae are expressed in broad ectodermal domains, and in juveniles most Hox genes are expressed specifically in the ganglia of the ventral nerve cord and in the posterior growth zone. Most genes have stable, discrete anterior and posterior expression boundaries, and exhibit spatial and temporal colinearity, with staggered anterior boundaries in the thoracic and anterior abdominal segments ( [19]; see Fig  1B). In the nine thoracic segments of juveniles, there are eight unique combinations of Hox genes expressed, suggesting that almost every thoracic segment has a unique molecular identity. In contrast, all abdominal segments express the same three Hox genes.
We hypothesized that both epimorphosis and morphallaxis occur during posterior regeneration in C. teleta. If epimorphosis occurs, we would expect cell proliferation to contribute to the formation of the regeneration blastema. Furthermore, if Hox genes are involved in anterior-posterior patterning of the blastema, we propose that they will be expressed in staggered domains with temporal and spatial co-linear expression characteristics during its outgrowth ( Fig 1A). For example, following amputation, we would expect CapI-pb, which is positioned at one end of the cluster, to be expressed earlier and in a more anterior location than CapI-Post2, which is positioned at the other end of the Hox cluster. If morphallaxis occurs in C. teleta, we hypothesize that amputation of a posterior piece of the animal would result in an anterior shift in expression of Hox genes in the pre-existing tissue (Fig 1B). These shifts would serve to reestablish all positional identities in the remaining tissue in an animal with a reduced number of segments.
We investigated the dynamics of regeneration following transverse amputation at various axial positions in C. teleta juveniles at 2 weeks post-metamorphosis, a stage at which juveniles have approximately 22-28 segments. Analysis of Hox gene expression in the context of regeneration allowed us to examine two distinct aspects of regeneration: morphallaxis of pre-existing tissue, and patterning in the blastema. To facilitate our analyses, we developed a staging system to describe the first 7 days of posterior regeneration. We describe changes in nervous system innervation and gut ciliation, reformation of the pygidium and posterior growth zone (PGZ), and quantify cell proliferation by EdU incorporation. Our results demonstrate that epimorphosis occurs during posterior regeneration of C. teleta, resulting in the formation of a regeneration blastema. Seven out of ten Hox genes are expressed within the blastema at different stages of its outgrowth. However, they do not exhibit spatial or temporal co-linearity, suggesting that they do not play a role in anterior-posterior patterning of the blastema. We also examined Hox gene expression in pre-existing tissue following amputation at multiple axial positions. Specifically, animals were amputated between segments 7 and 11, within the expression domains of CapIpb, CapI-Scr, CapI-lox5, CapI-Antp, CapI-lox4, CapI-lox2 and CapI-Post2, and the expression pattern of each Hox gene in pre-existing segments was compared to uncut controls. Most Hox genes have stable expression patterns in pre-existing tissue following removal of posterior pieces of the animal. However, three Hox genes (CapI-lox2, CapI-lox4 and CapI-Post2) shift their expression boundary in an anterior direction in pre-existing segments, and these shifts depend upon the axial position of amputation. The shift in expression of CapI-lox2 and CapI-lox4 includes a subset of cells in the ganglia. Our results demonstrate that segments exhibit a stable identity, and only limited morphallaxis occurs during C. teleta regeneration.

Materials and Methods
Cloning of C. teleta CapI-Hox3, CapI-lox4 and CapI-lox5 genes Although a previous publication reported the cloning and expression patterns of all Hox genes also used in this study [19], it was necessary to reclone three Hox genes. We verified that these clones produced expression patterns identical to previously published results. Primer design was based on transcript sequences previously submitted to GenBank (CapI-Hox3 EU196539; CapI-lox4 EU196543; CapI-lox5 EU196542). Amplified fragments were 1283 bp (CapI-Hox3), 983 bp (CapI-lox4) or 1506 bp (CapI-lox5) and were cloned into pGEM-T easy vector (Promega) before sequencing at the University of Hawaii or Macrogen Corp (Maryland). Schematic representation of hypothesized Hox gene expression patterns in the blastema and pre-existing tissue during regeneration. At the top of the figure, black lines depict two scaffolds that contain 10 of the C. teleta Hox genes. Colored rectangles depict the coding sequence of each Hox gene. Color coding in (A) and (B) is the same as that used in the schematic depicting the genomic organization of Hox genes. (A) Schematic representation of the posterior end of regenerating juveniles at Stage II, III, IV and V of regeneration. Black dotted lines depict the amputation site. Colored areas within the blastema indicate regions of gene expression, with each color corresponding to a different Hox gene. If C. teleta Hox genes exhibited spatial co-linearity in the blastema during regeneration, genes at one end of the cluster (e.g. CapI-lab; see red band depicting expression at Stage V) would be expressed more anteriorly in the blastema than genes at the other end of the cluster (e.g. CapI-Post2; see light purple band depicting expression at Stage V). If the C. teleta

Animal husbandry and amputations
A C. teleta colony was maintained in the laboratory at 19°C according to published culture methods [20]. Larvae were allowed to emerge naturally from the maternal brood tube, and animals were maintained in bowls with filtered seawater (FSW) and previously frozen and sieved estuarine mud as a food source. Regeneration experiments were performed on juveniles at 2 weeks post-metamorphosis. Animals were removed from the mud and placed in 0.5% cornmeal agar:FSW plates supplemented with 60 μg/mL penicillin plus 50 μg/mL streptomycin. Prior to amputation, animals were immobilized in 0.5% cornmeal agar supplemented with MgCl 2 (1:1 FSW:0.37 M MgCl 2 ) for a maximum of 30 minutes, and then placed in a drop of 0.37 M MgCl 2 :FSW (1:1) on a platform of black dissecting wax (American Educational Products). Animals were amputated at the posterior edge of the target segment using a microsurgery scalpel (Feather; 15 degree blade), and then returned to 0.5% cornmeal agar:FSW dishes for up to 24 hours. For analysis of longer periods of regeneration, animals were then placed into bowls of FSW and sieved estuarine mud for the desired length of time. Prior to fixation, animals were recovered from mud and placed in 0.5% cornmeal agar:FSW dishes for 1-6 hours to remove mud and debris, immobilized in 0.5% cornmeal agar with MgCl 2 for 15 minutes, and fixed in either 3.7% paraformaldehyde:FSW at 4°C for 16 to 24 h for in situ hybridization, or at room temperature for at least 30 minutes for immunohistochemistry. Following fixation, animals were processed for either whole-mount in situ hybridization or immunohistochemistry (see below).

Detection of cell proliferation
The Click-iT EdU Alexa Fluor 488 Imaging Kit (Life technologies C10337) was used to label cells undergoing DNA synthesis, following the manufacturer recommendations. Briefly, juveniles were exposed to 5'-ethynyl-2'-deoxyuridine (EdU) at a final concentration of 3 μM for 1 hour or 8 hours, depending upon the experiment. Animals were then placed in 1:1 0.37 M MgCl 2 :FSW for 15 minutes before fixation overnight at 4°C in 3.7% paraformaldehyde:FSW. For some experiments, in situ hybridization was performed on samples prior to detection of EdU. Briefly, animals were rinsed several times with PBS and exposed to PBS+0.5% Triton-X100 before the EdU detection reaction was performed, following the manufacturer recommendations. Following the EdU detection reaction, antibody labeling was performed as described below.

Immunohistochemistry
Following EdU incorporation and detection, juveniles were washed several times in PBS + 0.1% Triton (PBT), before being treated with block solution (PBT + 10% normal goat serum, Sigma G9023) for 45-60 minutes. Mouse anti-acetylated α-tubulin antibody (Sigma T6743) was diluted to 1:300 in block solution, and animals were incubated for 12-18 hours at 4°C. The following day, animals were washed twice in PBT, followed by four PBT washes of 20-30 minutes each. Donkey anti-mouse-546 secondary antibody (Invitrogen A21203) was diluted to 1:250 in block solution, and animals were incubated for 2-4 hours at room temperature. Following two rinses in PBT, four PBT washes of 20-30 minutes each were conducted. Finally, animals were equilibrated in 80% glycerol:PBS plus 0.125 μg/μL Hoechst 33342 (Life Technologies, H3570) overnight, before being imaged and analyzed as described below.

Nuclei Counts and Statistical Analyses
To quantify EdU-positive cells, specimens were counterstained for total nuclei with Hoechst 33342, and then confocal z-stacks (see Microscopy section below) were rendered into 3D images and cropped according to which area of the animal was desired for analysis (8 th , 9 th , 10 th segment or regenerating tissue). Each z-stack was created from focal planes extending from the ventral ectoderm to the lumen of the gut. A segment was defined as the region of tissue extending from the posterior edge of the ganglion of the preceding segment, to the posterior edge of the ganglion of the segment of interest. Following amputation, new tissue was defined as tissue posterior to the posterior-most pre-existing ganglion (see S1A Fig). Once a particular area was defined, EdU-positive nuclei and total nuclei were digitally identified using Imaris Software (Bitplane, Switzerland), using a size threshold of 2uM, and a quality score of >15. Each area was examined to ensure digital identification of nuclei was accurate. The number of EdU-positive nuclei were divided by the total number of nuclei to generate a ratio for a particular area. Cell counts were taken from at least 5 individuals, and comparisons between different segments, or between segments and the newly formed tissue analyzed using a Student's one-tailed t-test. Differences between samples with a p value <0.05 were considered statistically significant.

Microscopy and Imaging
Specimens which had undergone in situ hybridization were imaged using an Axioskop 2 motplus compound microscope (Zeiss, Gottingen, Germany), coupled with a SPOT FLEX digital camera (Diagnostic Instruments, Inc., SterlingHeights, MI). Images were captured using SPOT imaging software and analyzed using Adobe Photoshop CS6 (version 13.0). Multiple DIC focal planes were merged for some images using Helicon Focus (Helicon Soft Ltd., Kharkov, Ukraine), as noted in the figure legends. Following immunohistochemistry or EdU labeling, animals were imaged using a Zeiss LSM 710 confocal microscope (Zeiss, Gottingen, Germany). Z-stack projections were generated using ImageJ (NIH). All figures were created in Adobe Photoshop CS6 (version 13.0).

Results
Posterior regeneration in C. teleta A brief description of posterior regeneration in C. teleta has been published [14]; however, the previous study focused on reproductive adults (8 weeks post-metamorphosis), and here we provide a careful analysis of cell division patterns and blastema formation. To examine blastema formation, cell proliferation dynamics and ventral nerve cord architecture during posterior regeneration in C. teleta, we used nuclear staining, EdU incorporation and anti-acetylated α-tubulin reactivity, to compare the posterior end of 2 week old juveniles at various stages of regeneration following transverse amputation, with uncut controls. For quantification of proliferating cells, the ratio of EdU-positive nuclei to total nuclei were calculated in the three segments closest to the amputation site (segments 8, 9 and 10), at each stage of regeneration. The ratios for each segment were compared to the ratio of EdU-positive to total nuclei in the corresponding segment in uncut controls (S1A Fig).
The posterior end of unamputated juveniles contains mature abdominal segments (aqua bar in Fig 2A-2A"), a subterminal posterior growth zone (PGZ; green bar in Fig 2A-2A"), and a terminal pygidium (white bar in Fig 2A-2A"). The ventral nerve cord of each mature abdominal segment contains a ganglion (Fig 2A and 2A", white arrowhead marks one in each panel), with three peripheral nerves extending laterally (Fig 2A", open arrowheads showing three peripheral nerves). In uncut juveniles, we define the PGZ as the area anterior to the pygidium, but posterior to the posterior-most segment containing a well-defined ganglion. Segments immediately anterior to the PGZ tend to be smaller than segments further anterior, consistent with them undergoing significant growth after segment boundaries initially form. In the posterior end of the animal, the lumen of the hindgut contains cilia that can be visualized by antiacetylated α-tubulin (Fig 2A", white circle; S2A Fig). The pygidium of C. teleta is simple in morphology and lacks terminal projections such as anal cirri. Neurites extend from the posterior-most ganglion in a posterior direction into the pygidium. The lateral-most of these have a Y-shaped pattern as they extend through the PGZ, and innervate the pygidium with a diagonal orientation with respect to the anterior-posterior axis of the body (Fig 2A", white arrows).
Transverse amputation at the boundary of segment 10 and 11 (Stage 0), removes 13-18 abdominal segments, the PGZ and pygidium (Fig 2B-2B", aqua bar indicates the remaining segments; white arrowheads in B and B" mark the ganglion of segment 10). In addition, the ciliated hindgut is removed (S2B Fig), and the axons of the ventral nerve cord are abruptly terminated (Fig 2B"). In amputated animals, EdU-positive cells are scattered throughout the body, including in segments anterior to the amputation site, which is typical for actively growing two-week old juveniles (Fig 2B'). Stage I of regeneration (approximately 24 hours post-amputation (p.a.)), is characterized by several features. Wound healing is initiated within approximately 4-6 hours post-amputation, and occurs by contraction of the severed edges of the body wall. At Stage I, an epithelium covers the amputation site. In addition, the pattern of EdU-positive cells changes relative to uncut animals. EdU-positive cells are still scattered throughout the body (Fig 2C'), however there is a reduction in EdU-positive cells near the cut site. When the ratio of EdU-positive cells to total cells are quantified in the three segments proximal to the cut site (segment 8, 9 and 10), there is a significant decrease in the proportion of dividing cells  (Fig 2C").
Stage II of regeneration begins at approximately 2 days p.a. At this stage, there is no longer a significant difference in the ratio of EdU-positive cells to total cells in the three segments closest to the wound site compared to uncut controls (segment 8, p = 0.288; segment 9, p = 0.165; segment 10, p = 0.172; S1B Fig). A small blastema containing numerous dividing cells is now visible immediately posterior to the amputation site ( Fig 2D-2D", orange bars). Coincident with the appearance of a blastema containing dividing cells, neurites are dispersed throughout the blastema (Fig 2D"). The morphology of the longitudinal nerves enable the location of the amputation site to clearly be seen; the longitudinal nerves are thicker immediately anterior to the amputation site, and there is an abrupt transition at the amputation site, with thinner nerves posterior to the amputation site (compare longitudinal nerves to the left and right of the dotted lines in Fig 2D", 2E" and 2F"). This transition is clearly visible from Stage II-IV of regeneration (approximately 2-5 days p.a.). These appear to extend from severed axons of the ventral nerve cord, and form a random 'web' throughout the blastema. The hindgut appears very similar to the hindgut at a similar position in uncut animals (S2C Fig).
During Stage III of regeneration (approximately 3 days p.a.), there is more organization within the blastema, both in the distribution of proliferating cells and the pattern of axonal projections. Proliferating cells are densely packed within the nascent tissue, forming two bilateral clusters, which extend anterior of the boundary between the blastema and the pre-existing segmented tissue (Fig 2E', white arrows; Fig 2E-2E", orange bars). At Stage III, there is also a significant increase in the ratio of EdU-positive cells to total nuclei in the 10 th and 9 th segments (but not the 8 th segment), when compared to uncut controls (segment 8, p = 0.176; segment 9, p = 0.046; segment 10, p = 0.025; S1B Fig). Numerous axonal projections in the blastema are now organized into multiple axon tracts, oriented along the anterior-posterior axis, and each comprised of multiple neurites (Fig 2E", blue arrows; compare Fig 2D" and 2E"). Some preparations were observed to have a few cilia in the endothelial lining of the lumen of the gut within the blastema at stage III, however most preparations did not (S2D Fig).
At Stage IV (approximately 5 days p.a.), the blastema contains numerous dividing cells, organized in bilateral clusters that extend anterior to the cut site, and are less dense at the posterior end of the nascent tissue (Fig 2F'). Stage IV is marked by a characteristic pattern of neuronal projections into the blastema. Approximately 4 lateral axon tracts extend from preexisting lateral nerves (2 on each side; Fig 2F", red arrows). There are five medial axon tracts that extend from the ventral nerve cord into the posterior end of the regenerating tissue ( Fig  2F", blue and white arrows). Notably the two lateral-most neurites extending from the ventral nerve cord form a Y-shaped pattern, reminiscent of the Y-shaped branching pattern seen at the indicate approximate position of amputation, and all tissue to the right of dotted lines is regenerated tissue. The panels in each row are from a single individual. The regenerative process is described as progressing through different stages (left of rows), and specific stain, chemical or antibody is indicated at the top of columns. (A-G) Hoechst 33342 staining showing nuclei; (A'-G') EdU incorporation marking dividing cells; (A"-G") anti-acetylated α-tubulin labeling neurites. Scale bars in A"-G" are 50 μm. White circle in A", F" and G" shows cilia of the hindgut. White arrowheads show mature ganglia in A, A", B, B", G and G". Open arrowheads show peripheral nerves in A" and G". White arrows in A", F" and G" mark the Y-shaped neurites which extend into the pygidium. Blue arrows in E" mark nerve tracts. Blue arrows in F" mark medial axons that branch from the ventral nerve cord and red arrows mark axons that branch from lateral nerves. White arrows in E' indicate bilateral clusters of EdU-positive cells. Chaetae are autofluorescent and are visible in A', B', C' and D'. posterior end of uncut animals (Fig 2F", white arrows). In light of neurite morphology and the decreased density of EdU-positive cells at the posterior end, we postulate that Stage IV marks the beginning of the reformation of the PGZ and pygidium ( Fig 2F-2F", green and white bars, respectively). At this stage, a ciliated hindgut has also reformed (S2E Fig). Similar to Stage III, at Stage IV there is a significant increase in cell proliferation in the 9 th and 10 th segments, compared to corresponding segments in uncut controls. Unlike Stage III, at Stage IV this increase is also seen in the 8 th segment when compared to the corresponding segments in uncut controls (segment 8, p = 0.023; segment 9, p = 0.0049; segment 10, p = 0.00007; S1B Fig It should be noted that the rate of regeneration varies somewhat among individuals, and preliminary data suggests it may be due, at least in part, to variation in nutrient intake. Therefore days post-amputation are approximate for each stage described.

Expression of Hox genes in regenerating tissue following transverse amputation
To investigate the possible involvement of Hox genes in formation, patterning and outgrowth of the blastema, we analyzed the expression of each Hox gene following transverse amputation between segment 10 and 11 (Fig 3). Expression was examined at four stages of regeneration, as defined in the preceding Results section "Posterior regeneration in C. teleta". The stages examined were Stage II (approximately 2 days p.a.), when a blastema with proliferating cells becomes morphologically visible (see Fig 2D and   CapI-pb expression is somewhat variable between samples. In a subset of cases (n = 16/25), it is in ectodermal cells that form two diffuse bilateral clusters immediately posterior to the wound site ( Fig 3C; black arrows), and in the other cases it is not detectable (data not shown). CapI--Hox3 is expressed in a broader domain than CapI-pb, in bilateral clusters in both ectodermal and mesodermal cell populations (Fig 3H, black and white arrows, respectively). As in Stage II, the CapI-Hox3 expression domain is localized close to the cut site, with the posterior end of the blastema free of expression. CapI-lox5 is expressed in two small bilateral clusters in the ventral ectoderm of the blastema, posterior to the cut site ( Fig 3R, black arrows). The expression pattern of CapI-lox5 broadly resembles that of CapI-pb, although CapI-lox5 expressing cells form more defined clusters than those that express CapI-pb (compare Fig 3C and 3R). CapI-lox4 is expressed in a diffuse domain in the ventral ectoderm (Fig 3W), and covers a larger area of the blastema compared with the expression domain of CapI-pb ( Fig 3W; compare Fig 3C and 3W). CapI-Post2 is expressed in ectodermal tissue, extending from the cut site to the posterior-most end of the blastema in two bilateral clusters. The medial boundary of these two clusters is close to the ventral midline but there is a small gap in expression between the two domains ( Fig 3GG).
Hox All seven genes described are expressed in the blastema at Stage IV. At Stage IV, the blastema has increased in size, and a PGZ and pygidium have formed, although no mature segments are visible ( Fig 2F-2F"; green and white bars in Fig 3D, 3I, 3N, 3S, 3X, 3CC and 3HH indicate the PGZ and pygidium, respectively). CapI-pb is expressed in a small subset of cases (n = 4/21), in two bands of ectodermal cells; one immediately adjacent to the wound site, and a second one immediately anterior to the posterior end ( Fig 3D,  cases do not show blastemal expression). As in Stage III, CapI-Hox3 is expressed in both ectodermal and mesodermal cell populations. Expression is approximately halfway between the wound site, and the posterior end of the blastema (Fig 3I, black and white arrows indicate ectodermal and mesodermal cell populations, respectively). CapI-Dfd appears for the first time in the blastema at Stage IV in approximately half the animals assayed (n = 7/13). In these animals, CapI-Dfd is expressed in groups of ectodermal cells that radiate from the midline in approximately four transverse rows on each side (Fig 3N, black arrows). CapI-lox5 is evident in approximately three bilateral cell clusters (Fig 3S, black arrows), beginning immediately adjacent to the cut site, and extending approximately three quarters of the length of the blastema along the anterior-posterior axis. This pattern broadly resembles the repeated pattern of CapI-Dfd expression, although there are some differences. Adjacent CapI-lox5 expressing cell clusters are situated further apart from each other than the transverse bands of CapI-Dfd expression, and there are fewer bands of CapI-lox5 expressing cells (compare Fig 3N and 3S). CapI-lox4 is detected in a single, narrow band of ectodermal cells with a medial-lateral orientation, at the boundary of the cut site (Fig 3X, black arrows). CapI-lox2 expression is initiated at this stage, in multiple medial-lateral bands of ectodermal cells in the blastema (Fig 3CC, black arrows). Compared to its single broad expression domain at Stage III, CapI-Post2 is now expressed in transverse bands, with expression domains that extend ventro-laterally, but do not meet at the midline ( Fig 3HH, black arrows).
At Of the ten Hox genes surveyed, seven are expressed in regenerating tissue, and three of these seven are also expressed in the PGZ of growing animals. CapI-Dfd and CapI-lox5 are expressed in ectodermal cells of the PGZ in uncut animals and ectodermal cells of the blastema during regeneration (Fig 3K and 3P; see aqua, green and white bars denoting mature segments, posterior growth zone, and pygidium, respectively), and CapI-Hox3 is expressed in both ectodermal and mesodermal cells of the PGZ, and also in ectodermal and mesodermal cells in the blastema (Fig 3F). While most genes show a similar pattern between the posterior portion of uncut animals and Stage V of regenerates (for example, compare Fig 3A and 3E and Fig 3EE and 3II), the expression of CapI-Dfd differs. At Stage V, CapI-Dfd is expressed in a broader domain along the medial-lateral axis than that seen in uncut animals (compare Fig 3K and 3O), and is in individual cells in a lateral position (Fig 3O, open arrowhead shows an example). Our results show that most Hox genes are expressed in the blastema following transverse amputation, with different spatial and temporal characteristics. By Stage V of regeneration, the expression pattern of the Hox genes in regenerating tissue has returned to that of uncut animals.
Stable expression of Hox genes in pre-existing segments following amputation at the segment 10/11 boundary To determine the effect of amputation on the expression of Hox genes in pre-existing segments, we studied juveniles at 2 weeks post-metamorphosis (approximately 22-28 segments) and conducted cuts between the boundary of segments 10 and 11. This position is posterior to the first abdominal segment. Gene expression was examined at Stage III, IV and V of regeneration (approximately 3 days, 5 days and 7 days post-amputation), and compared to age-matched uncut controls. We observed stability of Hox gene expression boundaries in uncut juveniles irrespective of length, and even following amputation. A single exception is CapI-Post2, which shifted its anterior expression boundary in pre-existing segments following amputation (Fig 4).
In uncut 2 week post-metamorphic juveniles, the anterior class Hox genes, CapI-lab and CapI-pb, are expressed in thoracic segments 1-7 and 5-8, respectively, and CapI-Hox3 (paralogy group 3 (PG3)), is expressed exclusively in the PGZ. CapI-lab expression is limited to the ganglia of the VNC, whilst CapI-pb expression is also apparent in the epidermis, in a striped pattern (Fig 4A, 4E and 4I). Although the anterior and posterior boundaries of these genes are invariant, CapI-lab expression is strongest in segments 2 and 3, and weaker in the surrounding segments, while CapI-pb shows strongest expression in segments 6 and 7, with weaker expression in segments 5 and 8. Following amputation, the anterior and posterior boundaries of expression of each gene were unchanged (Fig 4B-4D and Fig 4F-4H). In the case for CapI--Hox3, amputation removes the original expression domain, and CapI-Hox3 expression was not observed in pre-existing segments during any time point examined. CapI-Hox3 expression is initiated in the newly formed tissue at Stages III-V of regeneration (Fig 4J-4L; see also Fig 3G-3J and Fig 5).
The six C. teleta central class Hox genes, CapI-Dfd, CapI-Scr, CapI-lox5, CapI-Antp, CapI-lox4 and CapI-lox2, are expressed in staggered domains in uncut animals, and together their expression encompasses almost the entire anterior-posterior body axis (see Fig 1B). In intact animals, most exhibit expression restricted to the ganglia of the VNC and growth zone. CapI-Dfd is expressed in the ganglia of segments 2-7 and the PGZ (Fig 4M). Although the anterior and posterior boundaries of expression are invariant, relative expression levels between segments differ for CapI-Dfd; expression is strongest in segments 3 and 4, but weaker in segments 2 and 5, 6 and 7. CapI-Scr is expressed in the ganglia of segments 3-7 with expression strongest in segments 5 and 6, but weaker segments 3, 4 and 7 (Fig 4Q). In addition, CapI-Scr is also expressed in the paired genital ducts located between segments 7 and 8, which are male reproductive structures ( [15]; Fig 4Q, black arrows). CapI-lox5 is expressed in the ganglia of segments 4-8 and the PGZ. Expression is strongest in segments 5 and 6, and weaker in segments 4, 7 and 8 ( Fig 4U). CapI-Antp is expressed in the ganglia of segments 5-8, and expression appears greatest in segments 6 and 7 (Fig 4Y). CapI-lox4 and CapI-lox2 have different anterior expression boundaries, otherwise they both are expressed in the ganglia of all abdominal segments, regardless of the length of the animal. CapI-lox4 has an anterior boundary in segment 7, and CapI-lox2 has an anterior boundary in segment 10. The expression of CapI-lox4 appears strongest in segments 7-9, and uniformly weaker in all abdominal segments, while the expression of CapI-lox2 appears uniform across all segments (Fig 4CC and 4GG). Following amputation between segments 10 and 11, the expression boundaries of each central class Hox genes do not change, even though amputation disrupts their normal expression domain (Fig 4N-4P, 4R-4T, 4V-4X, 4Z, 4AA and 4BB, 4DD-4FF, 4HH-4JJ). For example, although the expression domain of CapI-lox4 extends from segment 7 to the posterior end, amputation at the 10 th segment does not change the CapI-lox4 anterior boundary, and expression remains in the ganglia of the pre-existing segments 7-10 at Stages III-V (Fig 4DD and  4EE). At Stage V, when new abdominal segments have formed, expression is also seen in all of the new ganglia (Fig 4FF).
Unlike each of the other Hox genes, which have stable anterior expression boundaries, the anterior expression boundary of CapI-Post2 is variable in growing worms, and we observe a transient anterior shift in pre-existing tissue following amputation between segments 10 and Therefore, the anterior expression boundary is unrelated to the total length of the animal. When amputated at the boundary of segments 10 and 11, the anterior boundary of CapI-Post2 expression shifts to include thoracic segment 9 by Stage III, and the posterior boundary of expression remains in the posterior-most segment (segment 10) (Fig 4LL). This expression represents a shift anterior by one segment relative to any expression seen in normally growing animals. By Stage IV, the anterior boundary of expression has shifted to the first abdominal segment, segment 10 ( Fig 4MM). Similarly at Stage V, the CapI-Post2 anterior boundary is in segment 10, and is also expressed in all newly formed abdominal ganglia ( Fig  4NN).

CapI-Hox3 is expressed in a subset of dividing cells
CapI-Hox3 expression is distinct from the spatial-temporal expression patterns of the other Hox genes, and is localized to the PGZ but is absent from the pygidium (Fig 5A). We conducted EdU incorporation experiments in combination with in situ hybridization to determine the relationship between the CapI-Hox3 expression domain and dividing cells in the PGZ of uncut animals, and in the blastema during regeneration. For experiments involving regenerating animals, transverse amputations were conducted at the boundary of segment 10 and 11, and anterior pieces were collected at Stage II, III, IV and V of regeneration. Following collection of animals at each time point, a one hour pulse of EdU was administered, followed by fixation and in situ hybridization using CapI-Hox3 as probe. For experiments involving uncut animals, either a one hour pulse or an eight hour pulse of EdU was administered, followed by fixation and in situ hybridization using CapI-Hox3 as probe.
Following a one hour pulse of EdU, the relationship between the position of the CapI-Hox3 expression domain and that of dividing cells was analyzed. The CapI-Hox3 expression domain is immediately posterior to an area with a high-density of proliferating cells, and resides in an area with fewer dividing cells directly anterior to the pygidium (Fig 5A' and 5A"). There are a few EdU-positive cells within the CapI-Hox3 domain (Fig 5A' and 5A"; white arrows show a single example). Similarly, after an eight hour pulse of EdU, the CapI-Hox3 expression domain is in an area with few dividing cells compared with the zone directly anterior to it (Fig 5F-5F"). At Stage II of regeneration, when a small blastema becomes apparent, the anterior boundary of CapI-Hox3 is adjacent to the cut site, and expression is absent from the posterior end of the nascent tissue (Fig 5B-5B"). Numerous EdU-positive cells are distributed throughout the blastema, including some within the CapI-Hox3 expression domain (Fig 5B' and 5B"; white arrows show a single example). Similarly, at Stage III of regeneration, CapI-Hox3 is expressed in the anterior portion of the blastema within the field of EdU-positive cells, which are distributed throughout the blastema (Fig 5C-5C Fig 2G-2G"), CapI-Hox3 is expressed in a pattern resembling that in uncut animals. Expression is directly anterior to the pygidium, and demarcates a subdomain within the PGZ that has a low density of dividing cells (Fig 5E-5E"). At this stage there are also some EdU-positive cells within the CapI-Hox3 expression domain (white arrows in Fig 5E' and 5E").

Anterior shift of Hox gene expression boundaries following transverse amputation
To determine if the axial position of amputation has any effect on Hox gene expression in preexisting segments, we conducted cuts at various positions along the anterior-posterior axis, and assayed expression of Hox genes at 24 hours p.a. (Fig 6 and S4 Fig). This time point was chosen since we observed the anterior shift in CapI-Post2 within 2 hours following amputation (see below and Fig 7L), and we reasoned that a shift in expression of other Hox genes would be apparent by 24 hours. Most Hox genes do not show a shift in expression boundary following cuts at the boundary between 7 and 8, 8 and 9, 9 and 10 or 10 and 11 at 24 hours post-amputation (data not shown and Fig 6). For example CapI-Antp is expressed in the ganglia of segments 5 to 8 (Fig 6A-6D; black arrowheads), following amputation at any of these positions, the same pattern as observed in uncut juveniles. This result is similar to amputations conducted at the boundary of segments 10 and 11 (Fig 4).
However, three genes, CapI-lox4, CapI-lox2 and CapI-Post2, show a shift in their anterior boundary of expression, and this shift depends upon the axial position of amputation. In unamputated juveniles, CapI-lox4 is expressed in the ganglia of segment 7 to the posterior end of the animal (see Fig 1B and Fig 4CC). However, when juveniles are amputated between segments 7 and 8 or 8 and 9, the anterior expression boundary of CapI-lox4 is now the ganglia of segment 6, in approximately half the cases (n = 4/7 and n = 9/18 respectively; the other cases show an anterior boundary at segment 7), and extends to the posterior end of the animal (segment 8 or 9; Fig 6E, 6E', 6F and 6F'). When the amputation is conducted at the boundary of segment 9 and 10, or segment 10 and 11, an anterior shift in expression does not occur, and CapI-lox4 is expressed from segment 7 to the posterior-most ganglia in almost all individuals (n = 10/11 and 8/8 respectively; one case shows an anterior boundary at segment 6; Fig 6G and  6H). In unamputated juveniles, CapI-lox2 is expressed in the ganglion of segment 10, and in posterior abdominal segments (see Fig 1B and Fig 4GG). If juveniles are amputated between either segment 7 and 8, or 8 and 9, and analyzed at 24 hours post-amputation, no expression of CapI-lox2 is detected. This is expected as the original expression domain has been removed tissue. All panels are confocal images. Images shown in panels A-F were generated from a subset of slices from a z-stack, whereas panels A'-F' were generated from maximum projections of z-stacks. ( Fig 6I and 6J). However, if amputation is at the boundary of segment 9 and 10, or 10 and 11, the anterior boundary of expression shifts to include the ganglion of segment 9 in almost all individuals examined (n = 12/14 and n = 8/10 respectively. The remaining cases show either no expression (n = 2/14) or expression in segment 10 only (n = 2/10); Fig 6K, 6K', 6L and 6L'). For amputations conducted at the boundary of segment 10 and 11, this anterior shift in expression is no longer apparent at 2 days p.a. (data not shown).
The anterior boundary of CapI-Post2 expression in uncut juveniles is variable, but does not extend anterior to the ganglion of segment 10 (see S5 Fig). As expected, if transverse amputation is conducted between either segment 7 and 8, or 8 and 9, no expression of CapI--Post2 is detectable, due to the original expression domain being removed (Fig 6M and 6N). However, as shown in Fig 4LL-4NN, if amputation is conducted at the boundary of segment 10 and 11, the anterior expression boundary shifts to include segment 9 in the majority of cases, and this shift occurs within 24 hours post-amputation (Fig 6P and 6P'; n = 8/10; two cases show expression in segment 10 only). In addition, amputation between segments 9 and 10 induces a shift in the anterior boundary of expression, so that CapI-Post2 is expressed in the ganglion of segment 9 (n = 16/17; 1 case shows no expression; Fig 6O and 6O').

Timing of the shift in anterior boundary of expression of CapI-lox4, CapI-lox2 and CapI-Post2
Since amputation at specific axial positions induced a new anterior boundary of expression for CapI-lox4, CapI-lox2 and CapI-Post2, we sampled animals at several time points following amputation, to determine when this shift occurs. Five time points within 24 hours were examined following amputation (Fig 7). When amputated between segments 8 and 9, expression of CapI-lox4 is seen in the ganglia of segment 7 onwards for the first 4 hours following amputation (Fig 7A-7C), similar to what is seen in unamputated juveniles of similar age (Fig 4CC). However, at 16 hours post-amputation, ectopic expression in segment 6 is first detected (n = 4/ 6; 2 cases show expression in segment 7 and 8 only; Fig 7D) and it persists until at least 24 hours p.a. (Fig 7E; n = 9/18; 9 cases show expression in segment 7 and 8 only). A higher magnification image of this expression shows that approximately 6 cells in the posterior side of ganglia 6 express CapI-lox4 (Fig 7D', black arrows). The pattern of expression is identical at both 16 and 24 hours p.a.
We amputated juveniles at the boundary of segment 9 and 10, and analyzed onset of expression of CapI-lox2 in the ganglion of segment 9. The original expression domain was removed and no expression is observed within the first 2 hours of amputation (Fig 7F and 7G). However, at 4 hours p.a., expression of CapI-lox2 in segment 9 is observed (n = 14/15; one case showed expression in segment 8 and 9; Fig 7H and 7H'), and it persists until at least 24 hours p.a. (Fig  7I and 7J; 16 hours, n = 11/13, two cases showed no expression; 24 hours, n = 12/14, two cases showed no expression). In segment 9, CapI-lox2 is expressed in multiple cells at the midline and the posterior portion of the ganglion (Fig 7H'), and this expression pattern is similar at 4 hours, 16 hours and 24 hours p.a.
Immediately following amputation at the boundary of segment 10 and 11, no CapI-Post2 expression is detectable (Fig 7K). However, by 2 hours p.a., there is expression of CapI-Post2 in the ganglion of segments 9 and 10 (Fig 7L; n = 7/15, seven cases showed expression in segment 10 only, one case showed no expression). This expression pattern is present at 24 hours p.a. (Fig 7M-7O; at 24 hours two cases showed expression in segment 10 only), and persists until 3 days p.a. (Fig 4LL). Expression in the ganglion of segment 9 extends from the anterior to the posterior boundary of the ganglion, and is similar to that in segment 10 across all time points analyzed.

Discussion
We present a comprehensive study of Hox gene expression during regeneration in the polychaete annelid, C. teleta, and interpret these expression data in the context of descriptions of cell proliferation and nervous system dynamics during the regenerative process. We show that C. teleta regenerates via cell proliferation and blastema formation (i.e. epimorphosis), a process concomitant with expression of a subset of Hox genes in the regenerating tissue. A blastema with proliferating cells is formed by Stage II (approximately 2 days p.a.), and new segments are formed by Stage V (approximately 7 days p.a.). Quantification of proliferating cells in the three segments closest to the wound site show that levels initially fall (by Stage I), recover to levels comparable to controls (by Stage II), and then rise over subsequent days (up to Stage V). From Stage 0-IV (the first 5 days of regeneration), neurites extend from pre-existing segments into the blastema, always maintaining contact with the posterior end of the new tissue. New ganglia appear by Stage V. During growth of the blastema, we observe unique expression of most of the Hox genes, suggesting a role in regeneration and patterning of new tissue. We also show that there is general stability of Hox expression domains in pre-existing segments following amputation, and that only three out of ten Hox genes shift their anterior boundaries of expression in the ventral nerve cord of pre-existing segments. This shift only occurs at some amputation positions, and is suggestive of limited morphallaxis.

Cell proliferation and blastema formation during posterior regeneration of Capitella teleta
We investigated cell proliferation and blastema formation during posterior regeneration of C. teleta, using EdU incorporation as a marker of dividing cells. A morphologically distinguishable blastema appears by Stage II, and it is populated by numerous proliferating cells during all stages of outgrowth. This demonstrates that C. teleta undergoes posterior regeneration by epimorphosis. This is not a surprising finding, since the regeneration of most annelids involves localized cell division in the blastema [6,[22][23][24][25]. Cells involved in regeneration of new tissue are generally assumed to be derived from one or more different sources: from division of resident stem cells, by dedifferentiation and subsequent division of differentiated cells, via transdifferentiation of pre-existing cell types (either in the presence or absence of proliferation), or from cells that migrate from sites distant to the wound [2]. In annelids, cell migration from sites distant to the wound have been reported, although these studies have not used lineage tracing methods to track cells in live preparations [26][27][28][29][30][31][32][33][34][35]. Currently, we do not know the source of dividing cells in the blastema.
We quantified cell proliferation in pre-existing segments closest to the amputation site, and show there are statistically significant differences in the number of dividing cells in the three segments closest to the wound site compared to the corresponding segments in uncut controls. These differences change over time. Following amputation at the boundary of segment 10 and 11, our data shows that by Stage I (approximately 24 hours p.a.), cell proliferation significantly decreases in the three segments proximal to the cut site (segment 8, 9 and 10; see Results and S1 Fig). It is not currently known whether this shutdown of cell proliferation extends to include other regions of the body, or how rapidly following amputation this phenomenon occurs. A recent study in the naid annelid Pristina leidyi, reports a whole body shut down of cell proliferation within 30 minutes of amputation, that persists for several days [23]. The authors propose that long-range signals produced by the wound drive cessation of cell division, perhaps due to an increased need for resources at the expense of other processes dependent on dividing cells. In contrast to the persistent low levels of proliferation in Pristina, cell proliferation in C. teleta recovers to levels comparable to uncut controls by Stage II of regeneration (approximately 48 hours p.a.), concomitant with formation of a morphologically distinguishable blastema. Subsequently, the proportion of dividing cells in segments 9 and 10 significantly increases compared to uncut controls during Stage III and IV. At Stage IV, the proportion of dividing cells in segment 8 is also significantly greater. It is not currently known when (or indeed if) this increased number of proliferating cells return to pre-amputation levels. The proliferating cells within these segments may be a source of cells for the blastema, and could either be born within these segments, or migrate from a distant site.
By Stage V, addition of segments occurs via a posterior growth zone, similar to the addition of segments during growth. However, the zone of proliferating cells in the posterior end of regenerating individuals is broader along the anterior-posterior axis relative to the posterior end of uncut individuals (compare Fig 1A' and 1G'). This is most likely due to rapid segment addition and growth in these new segments following transverse amputation. In Platynereis dumerilii, addition of segments occurs much faster after caudal amputation than during posterior growth. During growth, one posterior segment is added approximately every five days, whereas more than two segments per day are added during post-regeneration growth [25,36,37].

Regeneration of the ventral nerve cord
We analyzed regeneration of the ventral nervous system and associated peripheral neurons using an anti α-acetylated tubulin antibody to label neurites. At all stages of blastemal outgrowth, neurites extend from pre-existing segments into the blastema, such that contact is always maintained with the posterior end of the newly forming tissue. Initially, the neurites form a web-like pattern that cover the surface of the blastema. Over time neurites gradually become more organized into five longitudinal nerves that straddle the midline. This is the same organization of longitudinal nerves present at the posterior end of uncut animals. By Stage V, multiple new ganglia appear along these longitudinal nerves, and peripheral nerves are also apparent.
The presence of neurites that innervate the wound epidermis has long been known in annelids from classical morphological studies (see [29]), and from more recent studies that utilize fluorescent markers of cell bodies and axons [22,[38][39][40][41][42]. Innervation of the blastema during posterior regeneration occurs during a similar timeframe in different annelid species. For example, as in C. teleta, in Platynereis dumerilii, the first indications of nerve fiber regeneration are visible within 2 days post-amputation, and by the following day, innervated anal cirri buds have formed in P. dumerilii. Intersegmental nerves of the peripheral nervous system are visible within 5 days post-amputation [42], and in C. teleta, intersegmental nerves appear between Stage IV and V (approximately 5 days and 7 days post-amputation, respectively). Similarly in the oligochaete, Enchytraeus japonensis, nerve fibers begin to expand posteriorly from the remaining ventral nerve cord by approximately 3 days post-amputation, and by 5 days postamputation, a network around the new pygidium has formed [38].
In contrast to similarities in timing of neurite expansion into the blastema, there appear to be differences in the origin of these neurites. In C. teleta, neurites extend from both the ventral nerve cord and the peripheral nerves near the wound site (Fig 2F", white, blue and red arrows). Similarly, in the oligochaetes Enchytraeus fragmentosus, Stylaria lacustris, and Pristina leidyi, the nerve network that covers the blastema partly originates from the peripheral nervous system [22,41]. In contrast, in the closely related E. japonensis, neurites extend from the ventral nerve cord only [38], suggesting that in different species, regenerating neurites are recruited from different sources.
There is evidence that regeneration might depend on the presence of neurons. In 1902, T.H. Morgan described the inhibition of regeneration in the denervated adult annelid Allolobophora foetida [43]. Since then, many other annelids have been shown to require nerves to regenerate [29,35,44,45]. Nerve dependence has also been shown for many other invertebrate and vertebrate species (reviewed in [46]). In vertebrates, it is thought that secreted factors from nerves activate cell proliferation and differentiation in the blastema [47]. While this may be the case in annelids [35,45,48], it is intriguing to consider that nerves may also act as "highways" for the migration of neoblasts (annelid stem cells) to the wound site, as has previously been suggested [28,49].
The rapid appearance of 6-8 new ganglia with segmentally-iterated peripheral nerves between Stage IV and Stage V, suggests that multiple new ganglia and segments form simultaneously during regeneration. This has been previously observed in other regenerating annelids [40]. However, examination of a finer time scale between Stage IV and Stage V is required to reveal if there is a currently unappreciated anterior to posterior progression of regenerating peripheral nerves. If there is no anterior to posterior temporal progression, this would represent an example of a key difference between regeneration and development. During development of C. teleta, the segmentally-iterated peripheral nerves and ganglia form with an anterior to posterior progression [17].

Most Hox genes are expressed in the regeneration blastema
We investigated expression of C. teleta Hox genes at multiple time points following amputation, during the formation and outgrowth of the blastema. Of the ten Hox genes surveyed, seven are expressed in the new tissue between Stage II and Stage V of regeneration. However, the onset of Hox gene expression within the regenerating tissue does not follow any temporal co-linearity. That is, genes at one end of the Hox cluster are not expressed earlier in regeneration than those at the opposing end of the cluster (Fig 3, genomic organization to the left of panels; see Fig 1A and S3 Fig). In addition, there is no evidence for spatial co-linearity of Hox gene expression within the blastema, since there is no simple correspondence between expression boundaries of the Hox genes and their relative position within the Hox cluster (see Fig 1A and S3 Fig). This result contrasts with Hox gene expression patterns during development; Hox genes positioned at one end of the genomic cluster are expressed earlier and more anteriorly in C. teleta larvae than genes located at the other end of the cluster [19]. Thus, Hox gene expression patterns in the blastema are not consistent with a role in patterning the anterior-posterior axis of the blastema, and likely do not specify the identity of new segments being formed. Interestingly however, Hox genes are expressed in the ventral ectoderm of the blastema in repeated patterns (Stage IV; for example, see Fig 3N, 3S and 3HH), before the appearance of ganglia and peripheral nerves (Stage V). This suggests that the Hox genes may have roles in early patterning events or differentiation of cellular subtypes in newly forming segments.
Dividing cells are present at all stages of blastemal outgrowth, and Hox genes are expressed during this same timeframe. Proliferating cells first appear in the blastema at Stage II, the same time CapI-Hox3 expression is initiated. Proliferating cells are distributed throughout the blastema at Stage III and IV, when expression of six more Hox genes is detected. The spatial expression patterns of Hox genes do not precisely mirror the pattern of EdU-positive cells for any of the Hox genes expressed in the blastema, and therefore Hox genes are not expressed exclusively in dividing cells. It is likely that Hox genes are expressed in a subset of dividing cells (see also Fig 5 and Discussion section, "CapI-Hox3 demarcates a subdomain within the blastema and posterior growth zone). Hox genes have been shown to regulate proliferation in multiple cell types in other animals including in Drosophila neuroblasts [50], in the vulva of C. elegans [51] in human breast cancer cells [52], and in mouse neural crest cells [53]. Expression of Hox genes in a subset of proliferating cells in the blastema of C. teleta is consistent with the possibility that they are involved in controlling proliferation during regeneration. Whether they act to promote or inhibit proliferation would require knock down of each Hox gene followed by analysis of the number of dividing cells at multiple time points during regeneration, and comparisons to wild-type controls.
There are a number of genes expressed in the regeneration blastema of annelids [14,37,42,[54][55][56]. Thus far, it has been observed that genes expressed in the blastema are also expressed in the PGZ, indicating a link between regeneration and segment addition [57]. However, we show that of the seven Hox genes expressed during outgrowth of the blastema, only three are also expressed in the PGZ of uncut juveniles (CapI-Hox3, CapI-Dfd and CapI-lox5), suggesting that at least in C. teleta, the process of regeneration and segment addition may not be closely linked. Unlike their dynamic expression in the blastema during early stages of regeneration (Stages II-IV), by Stage V when new segments appear, the expression of each Hox gene closely resembles that seen in uncut controls. One exception is CapI-Dfd, which is expressed in a broader expression domain at Stage V than in uncut controls (Fig 3K and 3O). This broad expression domain may reflect the function of CapI-Dfd. For example, although new segments and mature ganglia have formed, CapI-Dfd may be required for the differentiation of late forming cell types. Presumably the expression domain of CapI-Dfd eventually returns to that of uncut controls.

CapI-Hox3 demarcates a subdomain within the blastema and posterior growth zone
To more precisely define the spatial position of CapI-Hox3 within the posterior growth zone, we combined EdU incorporation and in situ hybridization in uncut and regenerating juveniles, using CapI-Hox3 as a probe. Surprisingly, following a one hour EdU pulse in uncut animals, CapI-Hox3 is expressed in a heretofore unknown region of the PGZ, which contains a low density of dividing cells. This domain lies directly anterior to the pygidium and posterior to a high density of proliferating cells (Fig 5A-5A"). This result is our first observation of molecular and cellular differences within the PGZ of C. teleta that define distinct subdomains. During early stages of regeneration (Stage II-III; Fig 5B-5C"), CapI-Hox3 is expressed in the blastema posterior to the wound site, within the field of dividing cells. At later stages of regeneration (Stage IV-V; Fig 5D-5E"), CapI-Hox3 is expressed in the PGZ, in an area with a low density of dividing cells, reminiscent of its expression in uncut juveniles.
The expression of CapI-Hox3 in the posterior end of uncut juveniles is similar to the expression reported for Hox3 orthologs in juvenile stages of the annelids Platynereis dumerilii (Pdu-hox3; [56]), and Alitta virens (Nvi-Hox3; [54]). In both species, Hox3 is expressed in the segment addition zone (equivalent to the PGZ in C. teleta), between the pygidium and the most posterior segment. The Pdu-hox3 expression domain in Platynereis includes ectodermal teloblast-like stem cells, thought to contribute to new segments during posterior segment addition [56]. BrdU pulse-chase experiments suggest that these cells divide synchronously and more slowly than adjacent anterior cells [37]. The CapI-Hox3 expressing cells in C. teleta may have similar characteristics. As we failed to observe EdU incorporation after a short exposure of one hour in most CapI-Hox3 expressing cells, we hypothesized that they are simply slowly dividing cells. However, a longer pulse of EdU (eight hours) revealed a similar result, and does not support the hypothesis that the CapI-Hox3-expressing cells are synchronously dividing.
The expression of CapI-Hox3 in the posterior end of uncut juveniles is broadly similar to the expression reported for Hox3 orthologs in juvenile stages of the annelids Platynereis dumerilii (Pdu-hox3; [56], and Alitta virens (Nvi-Hox3; [54]). In both species, Hox3 is expressed in the segment addition zone (equivalent to the PGZ in C. teleta), between the pygidium and the most posterior segment. The Pdu-hox3 expression domain in Platynereis includes ectodermal teloblast-like stem cells, thought to contribute to new segments during posterior segment addition [56]. BrdU pulse-chase experiments suggest that these cells divide synchronously and more slowly than adjacent anterior cells [37]. In C. teleta, for both the one and eight hour pulses of EdU, there are only a few EdU-positive cells within the CapI-Hox3 expression domain, and the longer pulse of EdU failed to reveal a pattern of synchronous cell divisions. Although one cannot eliminate the possibility that the CapI-Hox3 expression domain contains a quiescent stem cell population, the current data suggest that there are differences between species in the characteristics of the Hox3-expressing cells in the PGZ.
The expression of CapI-Hox3 in C. teleta juveniles is a departure from the spatial co-linearity observed for the other genes in the Hox cluster, and suggests that CapI-Hox3 does not play a role in anterior-posterior patterning (see Fig 1B). Larval expression of Hox3 from the basally branching annelid, Chaetopterus sp., has been interpreted as evidence that Hox3 plays a role in axial patterning in this species [58]. However, the expression patterns of the Hox3 orthologs in the polychaetes P. dumerilii and A. virens resembles that observed in C. teleta, and suggests that Hox3 acquired a novel function within annelids. The loss of an axial patterning function for Hox3 likely occurred prior to the split of the Errantia from the Sedentaria, but following the split of the Chaetopteridae [59]. Interestingly, the loss of a segmental identity function of Hox3 occurred at least one other time in protostome evolution; in insects the Hox3 ortholog zen, does not specify segmental identity, but rather has roles in development of extraembryonic tissue [60,61].

Stability of Hox gene expression during growth
We previously described the expression of each of the C. teleta Hox genes during larval development and in 3 day post-metamorphic juveniles [19]. Comparison of expression patterns in 3 day old juveniles (approximately 14-18 segments) from the previous study, with Hox gene expression in uncut 2 week post-metamorphic juveniles (approximately 22-28 segments) from the current study, allowed us to draw conclusions regarding the expression of Hox genes during growth (i.e. posterior segment addition). During growth, the anterior and posterior expression boundaries of each Hox gene remains stable, even as posterior segments are added. For example, CapI-Antp is expressed in the ganglia of segments 5-8 in both 3 day and 2 week post-metamorphic juveniles. In another example, CapI-lox2 is expressed from segment 10 to the final posterior segment (segment 15) in 3 day post-metamorphic juveniles, and is expressed from the ganglia of segment 10 to the final posterior segment in 2 week post-metamorphic juveniles (Fig 4GG shows an animal of 22 segments in length).
However, there are some differences between the expression reported in 3 day post-metamorphic juveniles and in 2 week post-metamorphic juveniles. At 3 days post-metamorphosis, the only Hox gene reported to be expressed in the PGZ was CapI-Hox3, but in the present study the three genes, CapI-Hox3, CapI-Dfd and CapI-lox5, are expressed in the PGZ of 2 week old juveniles. We believe these differences in expression are due to nutritional status differences rather than due to growth or segment addition. In the previous study, animals were starved for at least 24 hours prior to fixation, while the present study denied animals food for a maximum of 6 hours (see Materials and Methods, "Animal husbandry and amputations"). We have recently become aware of the close relationship between nutritional status, cell division and gene expression levels in juveniles. Nutritional status has been shown to affect growth patterns in other annelids [23]. C. teleta continuously feeds as a juvenile, and there is a significant decrease in the number of dividing cells in the PGZ within 1-2 days of food removal (data not shown). Each of the three genes expressed in the PGZ at 2 weeks post-metamorphosis also showed expression in the PGZ when their patterns were re-analyzed in unstarved 3 day postmetamorphic juveniles (data not shown).
One other difference between the present study and the results reported in Frobius et al. (2008), is the position of the anterior expression boundary of CapI-Post2. Previously it was reported that the anterior boundary of CapI-Post2 in 3 day post-metamorphic juveniles is fixed at segment 10; however, data collected in this study shows that the anterior boundary is variable among individuals (see Results "Stable expression of Hox genes in pre-existing segments following amputation at the segment 10/11 boundary" and S5 Fig). Re-analysis of the expression of CapI-Post2 in 3 day post-metamorphic juveniles revealed a variable anterior boundary of expression also. In all cases for 3 day and 2 week old juveniles, the anterior-most segment in which CapI-Post2 is expressed is the ganglion of segment 10 (the first abdominal ganglia). Our results may differ from the results previously reported due to larger sample sizes analyzed in this study. In total, we analyzed expression for 55 individuals over four independent experiments (S5 Fig). In 3 day post-metamorphic juveniles, the anterior expression boundary differed by as many as 4 segments, while in 2 week juveniles the anterior expression boundary differed by as many as 11 segments among individuals. It should be noted that 3 day post-metamorphic juveniles have only 5-9 abdominal segments, and 2 week post-metamorphic juveniles have 13-19 abdominal segments. Older juveniles may have greater flexibility in the extent to which their anterior expression boundary can vary, since there are more abdominal segments present, and CapI-Post2 is expressed in abdominal segments. It is notable that in Alitta virens, most Hox genes also exhibit variable anterior expression boundaries [54], whereas only CapI-Post2 exhibits a variable anterior expression boundary in C. teleta.
Hox gene expression in pre-existing segments following amputation We previously hypothesized that following amputation, Hox gene expression in the remaining segments would be re-organized to account for the reduced number of segments of the animal.
That is, the remaining body would express the complete suite of Hox genes, but each gene would have different anterior-posterior expression boundaries when compared to uncut controls. Surprisingly, we found that expression of most Hox genes remained stable following amputation at multiple axial positions, suggesting that at least the ventral nerve cord does not undergo substantial re-patterning. However, amputation at different axial positions resulted in an anterior shift of the anterior expression boundary of CapI-lox4, CapI-lox2 and CapI-Post2. The occurrence of the shift depends upon the axial position of the cut, and appears to be related to the original expression domain of the gene (see Fig 6, Fig 7 and S4 Fig). CapI-lox4 is expressed from the ganglia of segment 7 to the posterior end of the animal in uncut animals, and amputation between segments 7 and 8 or 8 and 9 cause the anterior boundary of expression to shift to segment 6. Amputations further posterior (between 9 and 10 or 10 and 11) do not cause the anterior expression boundary to change. CapI-lox2 is expressed from the ganglia of segment 10 (the first abdominal segment) through all posterior segments in uncut animals. Amputation at the boundary of segment 9 and 10 or 10 and 11 causes a shift in expression to segment 9; however, no shift in expression occurs following cuts between segment 7 and 8 or 8 and 9. Therefore, CapI-lox4 shifts its anterior expression boundary by one segment when it is amputated one segment posterior of its original anterior expression boundary, while CapI-lox2 shifts its anterior boundary of expression by one segment when it is amputated one segment anterior or posterior of its original anterior expression boundary. When amputated at the boundary of segment 9 and 10 or 10 and 11, CapI-Post2 shifts its anterior expression domain to the ganglia of segment 9. CapI-Post2 has a variable anterior boundary, so this shift encompasses at least one, but possibly more segments. Notably, for CapI-lox2 and CapI-Post2, amputation can cause the anterior expression domain to cross the thoracic/abdominal boundary. This boundary may be significant, as eight unique combinations of Hox genes are expressed in the nine thoracic segments, whereas there is a single Hox code for all abdominal segments.
Each of the three genes show an axial shift of expression by 24 hours after amputation, and the timing of the shift in expression varies among each of the three Hox genes. CapI-Post2 changes its anterior boundary within 2 hours, CapI-lox2 within 4 hours and CapI-lox4 within 16 hours of amputation. Data for CapI-lox2 and CapI-Post2 suggests that it can take up to 5 days until expression domains return to resemble uncut controls. Following amputation between segments 10 and 11, the expression of CapI-lox2 reverts to that of uncut controls between 24 hours and 3 days post-amputation. In contrast, when amputated at the same position, the anterior boundary of CapI-Post2 remains in the ganglia of segment 9 until Stage IV (5 days post-amputation), when it reverts to that of segment 10 (Fig 4LL-4NN). From Stage IV until the final day of our analyses (Stage V, approximately 7 days p.a.), the anterior boundary of CapI-Post2 expression remains at segment 10. We do not know if or when a variable anterior expression boundary in abdominal segments is re-instated for CapI-Post2.
Due to the fact that we observe only slight shifts in the anterior boundaries for three Hox genes, our data provides only limited evidence for neural morphallaxis in C. teleta. Morphallaxis has been observed in a number of metazoans, and a number of striking examples have come from studies in annelids. Neural morphallaxis has been shown to occur in the annelid Lumbriculus variegatus, where regeneration induces re-organization in the nervous system [11,[62][63][64]. Intact animals display anterior-specific (head withdrawal) and posterior-specific (tail-withdrawal) responses that are mediated by distinct neural pathways. Within days following transverse amputation, posterior segments exhibit anterior-specific responses, a change later accompanied by transformation of neural morphology. A similar change is observed in anterior segments, where posterior escape reflexes are adapted following transverse amputation. It is possible that in C. teleta, transient changes in Hox gene expression could induce a structural and/or physiological shift in the nervous system, although we currently have no data to support this. Whether morphallaxis occurs in other tissues in C. teleta is currently unknown, although there are precedents for remodeling of the gut in the oligochaetes Enchytraeus japonensis [6,55] and Pristina leidyi [22].
There are broad similarities among the expression patterns of the Hox genes in A. virens, P. dumerilii and C. teleta. For example, in both the nereids and in C. teleta, most Hox genes are expressed in the ganglia of mature segments, and in the blastema during regeneration. In A. virens, nine out of the ten Hox genes (Nvi-Lox5, Nvi-Lox2, Nvi-Post2, Nvi-Hox1, Nvi-Hox4, Nvi-Lox4, Nvi-Hox7, Nvi-Hox2 and Nvi-Hox3) are expressed in the blastema following transverse amputation. In C. teleta, orthologs of each of these genes are also expressed in the blastema, with the exception of CapI-lab (Nvi-Hox1 ortholog) and CapI-Antp (Nvi-Hox7 ortholog). In P. dumerilii, seven of the nine identified Hox genes are expressed in the blastema following transverse amputation (Pdu-Hox2, Pdu-Hox3, Pdu-Hox1, Pdu-Hox4, Pdu-Lox5, Pdu-Post2 and Pdu-Post1). Orthologs of each of these genes are also expressed in the blastema during C. teleta regeneration, with the following exceptions: while Pdu-Hox1 is expressed in the blastema, the C.teleta ortholog CapI-lab is not, and while CapI-lox2 is expressed in the C. teleta blastema, the P. dumerilii ortholog, Pdu-Lox2 is not. There are some similarities in expression in the blastema between C. teleta, A. virens and P.dumerilii. For example, Hox3 is expressed in the PGZ in all three species, and Post2 is broadly expressed early in the blastema in all three species. The fact that most Hox genes are expressed in the regeneration blastema for all three species is consistent with the idea that Hox genes have an ancestral role in patterning the regeneration blastema of annelids.
Comparisons of Hox gene expression patterns in pre-existing tissue between A. virens and C. teleta hint at some important functional differences for Hox genes following amputation in these two annelids. In A. virens juveniles, Hox genes are expressed in overlapping domains with broad anterior to posterior gradients in the ventral nerve cord [54]. Each gene displays variable anterior boundaries, a fixed boundary at the very posterior of the animal, and are not considered to exhibit spatial or temporal colinearity [54]. Following amputation, there are changes in expression of Nvi-Lox2, Nvi-Post2, Nvi-Hox2 and Nvi-Hox3 in the pre-existing tissue of A. virens. In each case, amputation causes loss of the entire expression domain, and is followed by re-initiation of expression at the 'new' posterior end within 10 hours of amputation. The anterior shift in expression of the A. virens Hox genes is quite dramatic, and contrasts with the 1-2 segment anterior shifts seen for three Hox genes in C. teleta. For example, in uncut A. virens juveniles, Nvi-Post2 is expressed in the four to six posterior-most segments and in the growth zone. Following transverse amputation, Nvi-Post2 is re-expressed in two to three segments at the new posterior end within four hours of amputation, an anterior shift of approximately ten segments. Similarly, Nvi-Hox3 is expressed in the growth zone of uncut juveniles, and following removal of the posterior ten segments, it is expressed in the new posterior end within ten hours of amputation. The variable axial position of amputation for Nvi-Lox2 makes expression data in regenerates difficult to interpret. In intact animals, the anterior boundary of Nvi-Lox2 expression is at approximately segment 8, and at 4 hours post-amputation, expression is initiated in two pre-existing segments adjacent to the amputation site, indicating a morphallactic shift in expression [54]. However, the specimen shown at 4 hours post-amputation contains 11 segments. This result shows the expected pattern when amputations are conducted at the 11 th segment. C. teleta orthologs of two of these genes, CapI-lox2 and CapI-Post2, show an anterior shift in expression following amputation. Expression patterns in pre-existing segments following amputation have not been analyzed in detail in P. dumerilii.
The differences in segmental identity between A. virens and C. teleta may explain the observed differences in Hox gene expression patterns. The body segments in nereids are all very similar to each other morphologically [36]. In A. virens juveniles, Hox genes are expressed in an anterior to posterior gradient, with high levels of expression in the posterior end and variable anterior boundaries. These features have led to the proposal that Hox genes function to maintain positional co-ordinates along the anterior-posterior axis in A. virens [54]. In contrast, C. teleta has distinct thoracic and abdominal regions, and the Hox genes have stable anteriorposterior expression domains, with staggered anterior boundaries. Eight of the nine thoracic segments express a unique combination of Hox genes (S4 Fig), although there are only morphological distinctions currently known for a subset of these segments. In the abdomen, three Hox genes are expressed in all abdominal segments, regardless of the length of the animal. This suggests that in C. teleta, most Hox genes function to specify segment identity in the thoracic region, while the others maintain positional information in the abdominal region. Notably each of the three Hox genes expressed in abdominal segments shift their anterior boundaries of expression following amputation. The variation in Hox gene expression patterns among the three annelid species examined thus far emphasizes the challenge in trying to define homologous segments across annelids, and furthermore, highlights the importance of comparative studies within phyla.

Conclusions
We characterize the contribution of tissue remodeling and formation of new tissue during posterior regeneration in C. teleta. Posterior regeneration proceeds via epimorphosis and formation of a blastema, and during the first week following transverse amputation, the blastema is populated with numerous proliferating cells. Multiple neurites from pre-existing segments extend into the blastema, and continuously maintain contact with the posterior end of the regenerating tissue. Newly formed ganglia are apparent by approximately 7 days post-amputation (Stage V). Seven Hox genes are expressed in the blastema during the first week following amputation, although these genes do not exhibit spatial or temporal colinearity, suggesting regeneration is not a simple recapitulation of development. While the identity of the nervous system is largely stable in pre-existing segments following amputation, we provide the first evidence for limited morphallaxis in the ventral nerve cord based on anterior shifts in expression domains for three Hox genes. Future studies examining the effect of amputation on other tissues will reveal whether changes in gene expression in C. teleta is restricted to neural components, or whether there is an integrated whole body response.
Supporting Information S1 Fig. Quantification of the ratio of proliferating cells to total nuclei in segments 8, 9 and 10, and in new tissue during regeneration. (A) Schematic representation of areas taken for EdU/nuclei counts. The 8 th , 9 th and 10 th segments were defined as the area from the posterior end of the ganglia of the preceding segment to the posterior end of the ganglia in the segment of interest. The 8 th segment is shown with a purple square, the 9 th segment by a blue square, and the 10 th segment by a red square. EdU/nuclei counts for the blastema/new tissue were taken from the posterior end of the ganglia of the last mature segment (segment 10) to the posterior end of the blastema, as indicated by a black square. Ganglia are represented by open horseshoe-shaped structures along the midline of the animal. Double-headed arrows represent pair-wise comparisons between the ratio of EdU-positive nuclei and total nuclei. Comparisons were made between a particular segment (segment 8, 9 or 10) in uncut animals, versus the corresponding segment in regenerating animals. (B) Graphical representation of average EdUpositive nuclei/total nuclei per segment and in the blastema in uncut animals, and at Stage 0, Stage I, Stage II, Stage III, Stage IV and Stage V of regeneration. Each line represents ratios of EdU-positive cells to total nuclei calculated from the 8 th (purple), 9 th (blue), 10 th (red) segment or from blastema/new tissue (black). Error bars represent standard deviations of the mean. Data were generated from at least 5 individuals for each time point.