Evolution of eumalacostracan development—new insights into loss and reacquisition of larval stages revealed by heterochrony analysis

Within Malacostraca (Crustacea), direct development and development through diverse forms of larvae are found. Recent investigations suggest that larva-related developmental features have undergone heterochronic evolution in Malacostraca. In the light of current phylogenetic hypotheses, the free-swimming nauplius larva was lost in the lineage leading to Malacostraca and evolved convergently in the malacostracan groups Dendrobranchiata and Euphausiacea. Here we reconstruct the evolutionary history of eumalacostracan (Malacostraca without Phyllocarida) development with regard to early appendage morphogenesis, muscle and central nervous system development, and determine the heterochronic transformations involved in changes of ontogenetic mode. Timing of 33 developmental events from the different tissues was analyzed for six eumalacostracan species (material for Euphausiacea was not available) and one outgroup, using a modified version of Parsimov-based genetic inference (PGi). Our results confirm previous suggestions that the event sequence of nauplius larva development is partly retained in embryogenesis of those species which do not develop such a larva. The ontogenetic mode involving a nauplius larva was likely replaced by direct development in the malacostracan stem lineage. Secondary evolution of the nauplius larva of Dendrobranchiata from this ancestral condition, involved only a very small number of heterochronies, despite the drastic change of life history. In the lineage leading to Peracarida, timing patterns of nauplius-related development were lost. Throughout eumalacostracan evolution, events related to epidermal and neural tissue development were clearly less affected by heterochrony than events related to muscle development. Weak integration between mesodermal and ectodermal development may have allowed timing in muscle formation to be altered independently of ectodermal development. We conclude that heterochrony in muscle development played a crucial role in evolutionary loss and secondary evolution of a nauplius larva in Malacostraca.

Results: Timing of 33 developmental events from the different tissues was analyzed for six eumalacostracan species (material for Euphausiacea was not available) and one outgroup, using a modified version of Parsimov-based genetic inference (PGi). Our results confirm previous suggestions that the event sequence of nauplius larva development is partly retained in embryogenesis of those species which do not develop such a larva. The ontogenetic mode involving a nauplius larva was likely replaced by direct development in the malacostracan stem lineage. Secondary evolution of the nauplius larva of Dendrobranchiata from this ancestral condition, involved only a very small number of heterochronies, despite the drastic change of life history. In the lineage leading to Peracarida, timing patterns of nauplius-related development were lost. Throughout eumalacostracan evolution, events related to epidermal and neural tissue development were clearly less affected by heterochrony than events related to muscle development.
Conclusions: Weak integration between mesodermal and ectodermal development may have allowed timing in muscle formation to be altered independently of ectodermal development. We conclude that heterochrony in muscle development played a crucial role in evolutionary loss and secondary evolution of a nauplius larva in Malacostraca.
However, comparatively few investigations focus on heterochrony in invertebrate evolution [12][13][14][15][16]. Invertebrates, such as the crustaceans (the potential paraphyly of crustaceans has no impact to our study), display an enormous disparity in development, exemplified by the multitude of larval forms and life histories found throughout this group. Here we set out to explore the impact of sequence heterochrony on life history evolution of the crustacean group Malacostraca ('higher crustaceans').
Malacostraca represents a large and morphologically highly disparate taxon within crustaceans. Although malacostracans also have a rich fossil record, including larvae, we refer throughout our study to recent taxa only because musculature and nervous system is hardly known for fossil larvae or embryos. The plesiomorphic developmental mode generally accepted for crowngroup Crustacea (or crown-group Tetraconata if crustaceans are paraphyletic in relation to Hexapoda) comprises hatching of a free-swimming, planktonic larva with conserved morphology, called nauplius [17][18][19]. The nauplius larva (i.e., orthonauplius) bears three pairs of appendages (first antenna, second antenna, and mandible), which are used for feeding and locomotion. In Malacostraca, such a larva is found only in two groups ( Figure 1): Dendrobranchiata and Euphausiacea [20][21][22][23]. Moreover, despite controversies concerning the phylogenetic relationships within Malacostraca, Dendrobranchiata and Euphausiacea are always placed at nested positions within the tree [24][25][26][27][28][29][30][31][32]. In this light, the nauplius larva in Malacostraca has evolved secondarily from ancestors, which either showed direct development or hatched as a more advanced larval stage with a higher number of segments ( Figure 1).
Leptostraca, Stomatopoda, Caridea (Decapoda), Reptantia (Decapoda), Anaspidacea, Bathynellacea, and Thermosbaenacea lack a nauplius larva but pass through a characteristic embryonic stage known as egg nauplius [59]. In the egg nauplius, the first antennal, second antennal, and mandibular buds (naupliar appendage buds) appear prior to the posterior (postnaupliar) appendage anlagen [44,52,53,[60][61][62][63][64][65][66]. Timing of naupliar and postnaupliar appendage bud formation is separated by a distinct gap. Scholtz [59] suggested that the egg nauplius is formed as part of a recapitulated developmental program originally involved in formation of a free-swimming nauplius larva. This egg nauplius concept has drawn our attention to the question how exactly transitions between larval and embryonic development are achieved in evolution. The presence of a larval developmental program in the malacostracan ground pattern can help to explain the secondary (and potentially independent) origin of the dendrobranchiate and euphausiacean nauplius larva [59]. Though this egg nauplius concept is of great value for understanding malacostracan evolution, it does not sufficiently consider developmental timing. For example, it treats the egg nauplius and the free nauplius larva as two alternative situations. Yet nauplius larvae are themselves preceded by embryonic stages which show three pairs of appendage buds ( Figure 2). They differ from the egg nauplius stages of direct developers or species with zoea-like larvae, only by the lower amount of yolk and the more lateral position of the limb buds [67]. We prefer a more inclusive definition of the term 'egg nauplius' which applies also to all crustacean representatives with nauplius larvae. In our view, the egg nauplius represents a part of an ancestral developmental program which is shared between species with and without a free-swimming nauplius larva, before two different paths can be taken in development: (i) development of postnaupliar tissues, leading to a larger number of functional segments at hatching (Figure 2a) or (ii) differentiation and early functionality of the naupliar segments and hatching of a nauplius larva (Figure 2b).
The evolutionary scenario of recapitulated nauplius larva development in the egg nauplius [68] is largely based on gross external morphology of epidermal limb buds. Development of the nervous system or mesodermal tissue, such as musculature, has not played an important role, and only one publication discusses neurogenesis of a malacostracan species in an egg nauplius context [69]. Recently, we have found that anlagen of musculature develop in the naupliar segments only after the egg nauplius stage in several malacostracan representatives [67] (Figure 2a). The dissociated timing of mesodermal and ectodermal development suggests that retention of the ancestral larval developmental program does not occur in all tissues likewise and underwent heterochronic change in evolution. Here we will apply a developmental sequence approach to malacostracan development considering different tissue types (epidermis, nervous tissue, and muscle tissue) to gain a more fundamental understanding of the evolutionary changes to developmental timing which caused loss and reacquisition of larvae in malacostracan evolution.
Following Alberch [68], heterochrony affects only particular features of the organism, never the whole. In the case of the egg nauplius, we want to determine the evolutionary changes of developmental timing in different body regions and tissue types. Thus the relation between the modular organization of the developmental program and heterochronic evolution in Malacostraca is at the heart of our study. A conservative view on malacostracan developmental evolution would assume that all embryonic naupliar tissues develop in species without a nauplius larva by the same timing pattern as they would in nauplius larvabearing species. In this case, in the malacostracan last common ancestor, a developmental path would be taken that accelerates tissue development in the postnaupliar segments relative to the naupliar segments after the egg nauplius stage. We will refer to the initial part of the developmental sequence, in which only developmental events of the naupliar segments occur (but none of the postnaupliar segments), regardless of the respective tissue type, as 'egg nauplius phase.' Transition of developmental events from the egg nauplius phase to later positions in development, would prevent formation of a viable nauplius larva. Such changes would have had to be reversed during secondary evolution of the nauplius larva in Dendrobranchiata and Euphausiacea.
The morphological features to be investigated here in terms of developmental timing were chosen in a manner Figure 1 Overview of malacostracan phylogeny. Simplified representation of malacostracan phylogeny, following [28]. The major malacostracan monophyla suggested by these authors (Eumalacostraca, Caridoida, Decapoda, Pleocyemata, Xenommacarida, Peracarida) are marked with horizontal brackets. The taxa Anaspidacea and Bathynellacea are shown together as Syncarida. The peracaridan subtaxa Lophogastrida, Spelaeogriphacea, Mictacea, Tanaidacea, and Cumacea are excluded. Therefore, Amphipoda and Isopoda appear as sister groups. The developmental mode of the taxa is indicated by symbolic drawings at the bottom. The developmental mode is color coded to the branches and the most parsimonious character states of the ancestral lineages are shown. Outgroups are not depicted. Color coding: Direct/pseudodirect development (black); nauplius larva as hatching stage (blue); zoea-like larva as hatching stage (green).
that allows comparison between tissue types, as well as between the germ layers ectoderm and mesoderm. Also, we rely on a large number of segmentally repeated features, namely appendage buds, ganglion anlagen, and muscle precursors to allow detection of timing differences also between segments. Such features are recorded for the head segments and the first trunk segment, the last trunk segment, and the telson (Figure 3). This allows us to record heterochronic changes in patterning of the naupliar and postnaupliar segments and to draw conclusions on their relation to loss or gain of a nauplius larva. Also features without obvious segment affiliation but with relevance to the evolution of nauplius larva development are included, such as the anlage of the nauplius eye, muscle anlagen of the stomodeum, and hatching from the egg envelope.
Other features are included which are potentially relevant for formation of a zoea-like larva and will serve to determine heterochronic changes that relate to this larval form: Formation of appendage buds, ganglion anlagen, and muscle precursors in the sixth pleonal segment, formation of a posterior longitudinal muscle precursor in the telson, and offset of segment formation. Offset of segment formation describes the point in development at which generation of body segments from the posterior growth zone terminates, and the full set of trunk segments is present as anlagen [67]. In Malacostraca, mesoderm and ectoderm of the trunk segments are formed by repeated asymmetric cell divisions of stem-like cells, the mesoteloblasts and ectoteloblasts, which are located in the growth zone in the posterior part of the embryo and can be recognized by the specific arrangement of stained nuclei. We use the mesodermal segment anlagen as reference for over all body segmentation here.
Timing of the first appearance (onset) of the specified features is recorded for six eumalacostracan representatives and one outgroup: Gonodactylaceus falcatus (FORSKÅL, 1775) (Stomatopoda), Sicyonia ingentis (BURKENROAD, 1938) (Dendrobranchiata), Neocaridina heteropoda (KEMP, 1918) (Caridea), Procambarus fallax forma virginalis (Astacidea), Neomysis integer (LEACH, 1814) (Mysidacea), Parhyale hawaiensis (DANA, 1853) (Amphipoda), and Artemia franciscana (KELLOGG, 1906) (Anostraca). G. falcatus hatches as a zoea-like larva (pseudozoea), S. ingentis hatches as nauplius larva, while the remaining species develop directly. A decapod representative that hatches as a zoea-like larva was not available. However, the late embryonic stages of N. heteropoda differ only little in morphology from other caridean zoea larvae. Thus, in terms of developmental timing, N. heteropoda can be considered a legitimate representative of zoea-bearing decapods. A representative of Euphausiacea could not be sampled for this study, because our methods demand fresh or appropriately fixed material and these animals are difficult to obtain. Thus we focus on the lineage leading to Dendrobranchiata to infer heterochrony related to evolution of a nauplius larva. N. integer differs from most other Peracarida in that it shows pseudodirect development. In this species, hatching occurs early in development, but the inert larva (nauplioid) remains in the brood pouch until juvenile morphology is established. Based on the timing data, we apply a dynamic programming approach using the software Parsimov-based genetic inference (PGi) to trace evolution of the developmental sequences. The method used was first introduced by Harrisson & Larsson [70] and applied successfully in analyses of heterochrony since [4,8,71]. We have chosen the malacostracan phylogeny proposed by [28] and [72] as framework for the reconstruction of developmental evolution. It is in our view still the best supported one. We are aware that other suggested phylogenies [31] would give different results. The reconstruction of the ancestral developmental sequences of Eumalacostraca, Caridoida, Decapoda, Pleocyemata, and Peracarida, and inference of the heterochronic events that occurred along the different branches, will allow us to shed light on the evolutionary transformations of the segment and tissue-specific developmental processes that were involved in alteration of the developmental mode.
The following questions will be addressed (all taxon names refer to the respective crown-groups): How did the last common ancestor (LCA) of Eumalacostraca develop and to what degree did it show larval developmental patterns? Which heterochronies were involved in evolution of the nauplius larva of Dendrobranchiata? Which changes of the developmental sequence caused the emergence of zoea-like larval forms? In which way were developmental sequences altered in the lineage leading to Peracarida?

Methods
Specimen preparation, staining, and imaging procedure Collection of embryo and larva material, fixation, fluorescent staining, and confocal microscopy followed by 3D image processing was performed in a previous investigation [67]. The respective methodology applied to P. fallax forma virginalis is described in [73]. Visualization of muscle tissue was performed on larval stages L4, L6, and L9 [74] of A. franciscana. The fixation protocol was previously described in [75]. Larvae of A. franciscana were incubated with Phalloidin-ALEXA 561 overnight to visualize muscle tissue by f-actin labeling. Imaris software Version 6.1 (Bitplane AG) was used to adjust image quality in projections of the volume data and to reconstruct and highlight single muscle precursors. Confocal image stacks of S. ingentis embryos and nauplius larvae labeled with BODIPY-FL-phallacidin were kindly provided by Phillip Hertzler for detection of nervous system development [76]. Embryos of 13h, 15h, 17h, and 20h after fertilization were analyzed, as well as nauplius stages 1 and 4. Immunohistochemical labeling of developing nervous tissue was performed on G. falcatus, N. heteropoda, P. fallax forma virginalis, and N. integer by application of an antibody against anti-acetylated α-tubulin (clone 6-11 B-1, Sigma T6793) which labels neurites, even at early developmental stages. For this the same preparation and staining, protocols as for muscle precursor labeling were used. Also histochemical staining with phalloidin-ALEXA488 (Molecular Probes, A12379) was applied to visualize developing ganglion anlagen of early embryonic stages. Graphics were drawn and image tables were assembled using CorelDRAW Graphics Suite X3 (Corel Corporation, Ottawa).

Developmental sequence data
The ontogeny of an individual organism can be viewed as an array of semaphoronts [77]. Hennig's concept of the semaphoront has recently been revived to improve morphology-based phylogenetic inference on Pancrustacea/Tetraconata [78]. A semaphoront, in the sense of Hennig is '[…] the individual at a certain, theoretically infinitely small, period of its life' [77, p6]. Sequences of developmental events always refer to series of semaphoronts. We will speak of semaphoronts instead of developmental stages throughout this paper, because (i) staging systems are not established for all of the species we investigate and (ii) staging systems rely on specific criteria that limit their resolution. For the present work however, we must allow timing to be recorded even within stages and based on new criteria, thus defining new operational stages that we refer to as semaphoronts. Developmental event sequences [79] were recorded for each species from the semaphoront series. Every event refers to the first appearance of a morphological feature (listed in Table 1). We restrict the present investigation to such 'onset events' to limit the size of the data set. The majority of events in our data set were coded from our previously published comparative study of malacostracan muscle development [67]. A substantial part of the event data was acquired from new observations presented in the results section (Figures 4,5,6,7,8,9). In cases where data could not be provided by our own investigations, events were coded using the literature. An overview of these events and a list of the publications used are given in Table 2. A detailed overview of semaphoronts and events is given in Additional file 1. Early appendage morphogenesis of Astacidea and Amphipoda was coded from [80,81], early neurogenic events for Stomatopoda, Astacidea, Amphipoda, and Artemia sp. were coded from [82][83][84], and early myogenic events for Dendrobranchiata and Artemia sp. were coded from [76,84]. Features which were not recorded to be formed in development but are reported to be present in adults were coded as events at the end of the sequence. This is an effort to avoid missing data where possible, even if this results in artificial simultaneity between late events. This was the case for developmental sequences of all groups [85][86][87][88][89][90][91], except for Artemia sp. Following the position of the events in the sequence, ranks were assigned to every event (Table 3). An overview of the event series for all species, also showing stage specifications from the literature, rank values, and literature sources used for specific events, is provided in Additional file 1. Ranks ranged from 9 to 11, depending on the total number of semaphoronts described for each species.
The table gives a list of all events coded for comparative analysis of malacostracan development. The events specify the first appearance of a specific morphological structure (e.g.) appendage primordium or property (e.g.) hatched larva. Muscle precursor terminology was adapted from [58] for myogenic events. Abbreviations of muscle precursor groups are given in italics. The morphological features are sorted by the following categories: Epidermal appendage development, segmentation, myogenesis, neurogenesis, and Hatching. Events are numbered from 1-33 and this order is maintained for the analysis. Descriptions are given for every event, as well as abbreviations. Abbreviations of developmental events are given in brackets and used consistently throughout the paper.

Heterochrony analysis
Ranks were coded as character states for the respective events in a matrix of 7 × 33 cells and exported as a NEXUS file (Additional file 2) together with the phylogeny from [28], using the open source software package Mesquite 2.75 [92]. The tree was simplified by excluding all taxa that are not represented in our sampling. We use A. franciscana as an outgroup. Since a free-swimming nauplius larva is found throughout the 'entomostracan' crustaceans, it must also have been present in the linage leading to the crown-group Malacostraca. We added the developmental sequence of A. franciscana twice to the matrix to allow optimization of this condition in the analysis. Analysis of heterochrony was performed using a modified version of PGi [69], kindly provided by Luke Harrisson. The method uses a dynamic programming approach which treats the event sequence as a single complex character. Therefore, it avoids the assumption of event independence which is inherent to event pair-based methodology of heterochrony analysis [93]. PGi uses a simplified genetic algorithm-based heuristic on the event sequence, and Parsimov event pairing [94] is used as edit cost function. The program runs in the ape package [95] and was carried out using the open source statistics environment 'R' (version 3.0.1) [96]. We performed three runs with the following parameters for the PGi simplified genetic algorithm: 100 cycles of selection per node, 200 sequences per cycle of selection, and a maximum of 100 ancestral developmental sequences to be retained at each node. For each run, the most parsimonious solutions of equal cost are collected by the algorithm and used to calculate a pseudoconsensus tree. Heterochronies that occur in the equally parsimonious solutions are included in the pseudoconsensus tree if they fulfill the 50% majority rule criterion and the percentage of each heterochrony is given as bootstrap support [69]. The pseudoconsensus method was set to 'semi exhaustive' and the limit of evaluated solutions of equal score was set to 3,000. The pseudoconsensus trees of the three independent runs were combined to a superconsensus tree (Additional file 3). The ancestral  sequences are constructed by PGi from mean ranks that are calculated from the multiple equally parsimonious solutions of the pseudoconsensus trees [69]. Because of the high variation in the data set, the reconstructed ancestral sequences can show slight differences in event position that are not given as heterochronies by the analysis. Calculations were carried out on a Dell Optiplex790-computer with an i3-2100 CPU@3.1GHz and 8GB RAM, running 64Bit Microsoft Windows 7. Graphic representations of the superconsensus tree and transformation of developmental sequences were edited using CorelDRAW Graphic Suite X3. For all events, heterochrony rates were calculated. The heterochrony rate of an event in our case represents the number of heterochronic changes recovered by PGi for that event in the superconsensus tree, multiplied by its mean bootstrap value. Tissue-specific heterochrony rates were calculated which represent the mean heterochrony rate per event for all events specific to epidermis, neural tissue, or muscle tissue development. These values were also used to determine mean heterochrony rates for the germ layers ectoderm and mesoderm. Likewise heterochrony rates were compared between segments and tabulated. For this, events with problematic segment affiliation (FS, st, NEA, and HAT) were excluded. Two events represent combinations of several segment-specific events ([A1/A2], [NGA]). In these cases, the heterochrony rate of the combined event is used for each of the single segments, because simultaneity is observed in each of the investigated species and can thus be assumed also for the ancestral sequences. Where multiple events are affiliated with the same segment (myogenic events), the mean heterochrony rate of these events is used.

Recorded events
Event sequences were assembled for A. franciscana, G. falcatus, S. ingentis, N. heteropoda, P. fallax forma virginalis, N. integer, and P. hawaiensis by combination of our previous descriptions, literature data, and new observations presented here. Table 2 gives an overview of the literature sources used and of the events coded from them. The new observations are depicted in Figures 4,5,6,7,8,9. Furthermore, specifications of literature sources and reference to the corresponding figures depicting the new observations in the present work are shown for each event in Additional file 1. Information on development of other species is used in several cases to complete the semaphoront sequence. In Additional file 1, these species are also shown for the respective events.
Epidermal morphogenetic events (1-6): The events recorded for epidermal morphogenesis represent the appearance of distinct appendage buds of the first antenna, second antenna, mandible, first maxilla, second maxilla, first thoracopod, and sixth pleopod. An event is scored when the appendage anlage is recognizable as protuberance in the epidermal layer. Formation of first and second antenna is scored as a single (event 1) [A1/A2] because they are always observed to occur simultaneously.
Myogenic events (8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25): First appearance of the muscle precursors described in [67] is recorded here as myogenic events. For the present investigation, we reduced the total number of muscle precursors by combining some precursors to groups ( Table 1): The stomodeal muscle precursors are combined to one group (event 8). The same applies to the medial extrinsic appendage muscles of the first and second antenna, mandible, first and second maxilla, and first thoracopod, respectively (events 9,11,13,15,18,21), as well as the lateral extrinsic appendage muscles of the same body segments (events 10,12,14,16,19,22). First appearance of a group is registered as a developmental event, if any of the muscle precursors of one group is seen. Furthermore, the first appearance of longitudinal muscle precursors in the first maxilla, the second maxilla, and the first thoracopod segment (events 17, 20, 23) are recorded, but longitudinal muscle precursors of the mandible segment are excluded, as they occur only transiently in G. falcatus and P. fallax forma virginalis. Longitudinal muscle precursors of the sixth pleomeres (event 24) are recorded. They are recognizable as metameric, noncontinuous muscle precursors in the sixth pleopod segment. The posterior longitudinal muscle primordium (event 25) represents a portion of the longitudinal muscle strand that extends posteriorly into the growth zone and telson anlage. This event is lacking in Peracarida and is coded as absent for N. integer and P. hawaiensis.
Neurogenetic events (26)(27)(28)(29)(30)(31)(32): Data that are lacking in the published material for G. falcatus, P. fallax forma virginalis, as well as data on neurogenesis of N. heteropoda, N. integer, and P. hawaiensis were obtained by the methodology described above. We specified eight events for development of the nervous system (Table 1), six of which relate to the first appearance of ganglion anlagen. Ganglion anlagen are defined here as metameric cellular arrangements in the neuroectoderm, with a developing central neuropile [97]. Developing nerve fibers of commissures, connectives and lateral nerves can be present. Ganglion anlagen which are preceded by longitudinal neurite bundles originating from the posterior pioneer neurons are recognizable as spindle-shaped regions formed by these longitudinal fibers as shown by [82]. We specified the first appearance of ganglion anlagen in the naupliar, the first maxillary, second maxillary, first thoracic segment (events [26][27][28][29], and in the sixth pleomere (event 30).
Formation of the naupliar ganglia (protocerebrum, deutocerebrum, tritocerebrum, mandibular ganglion) are scored as a single event [NGA] because they are always observed to occur simultaneously. Developing ganglia are observed also posterior to the sixth pleomeres, e.g., a seventh pleonal ganglion in N. heteropoda or N. integer. For our purpose, we will record only the emergence of the sixth pleonal neuromere. Furthermore, we record the presence of the anlage of the nauplius eye, a feature commonly present in crustacean nauplius larvae (event 31), and of posterior pioneer neurons (event 32).
Overall development and segmentation (7, 33): We specified two features which are relevant for over all segment formation and differentiation: Offset of segment formation (event 7), recognizable by the presence of mesodermal segment anlagen of all thoracic and pleonal segments, and hatching from the egg membrane, which is at the same time the end of embryogenesis, is coded (event 33).

Description of developmental sequences
For convenience, we use specific font style for event and semaphoront abbreviations throughout this paper. Semaphoront abbreviations are given in bold letters and contain a two-letter code for the species name, followed by a roman number specifying the position of this semaphoront in the sequence, or by an abbreviation of an established stage name. The semaphoront abbreviations are adapted from [67]. For hatching individuals, the abbreviation 'HAT' is used instead of the roman number. Abbreviations that refer to specific developmental stages that were coded from the literature are given in italics in brackets. The abbreviations for ontogenetic events are given in square brackets throughout this paper. In the following section, the coded semaphoronts of the investigated species are described for G. falcatus, S. ingentis, N. heteropoda, P. fallax forma virginalis, N. integer, P. hawaiensis, and A. franciscana, together with the respective developmental events.     (Figure 5b). Nh I+ also shows a differentiated circumesophageal nerve ring, an anlage of the nauplius eye, and posterior pioneer neurons with elongate longitudinal neurite bundles (Figure 5c,d). Since the anlagen of the naupliar ganglia must be formed between Nh EN and Nh I+, we assign the event [NGA] to  The table shows rank values assigned to the events coded for heterochrony analysis. Events are ordered by number and sorted into groups as in Table 1 and Table 2.   [83] and is assigned to semaphoront Ph E3. These ganglion anlagen appear rapidly in anterior posterior progression, with a slight gap between naupliar and postnaupliar segments. However, since these events all occur within a single stage and since single events do not coincide with different events of our series, we assign them to one semaphoront and treat them as simultaneous in the sequence.  (Figure 10a). In the case of extrinsic appendage muscle precursor formation, the lateral muscle precursors show a higher heterochrony rate than the medial muscle precursors. Comparison of segment specific heterochrony rates (Figure 10b) shows that myogenic events in the naupliar segments (A1, A2, Md) have been altered in evolution significantly more often than appendage bud and ganglion formation. The first and second maxilla segments show no heterochrony in neurogenesis but slightly higher rates of heterochrony in appendage bud formation than in myogenesis. Between tissue types the mean heterochrony rates differ strongly, as reflected by mean rates of 1.32, 0.91, and 1.72 changes per event for epidermis (appendage buds), nervous tissue and musculature, respectively (Table 4). This results in considerably differing heterochrony rates between germ layers, namely 1.12 in the ectoderm and 1.72 in the mesoderm.
In the following, the ancestral developmental sequences and heterochronies represented in the PGi superconsensus tree are presented.

The branchiopod/malacostracan last common ancestor
The analysis revealed an ancestral sequence for the bran-

S. ingentis (Dendrobranchiata)
Reconstructed ancestral event sequences of Decapoda, Pleocyemata, and the terminal developmental sequence of S. ingentis are given together with the heterochronic changes in Figure 12

Peracarida
The reconstructed ancestral sequences of the caridoid and peracaridan LCA, as well as the developmental sequence of P. hawaiensis are shown in Figure 13 In the peracarid LCA sequence, the egg nauplius phase now only consists of epidermal appendage bud formation. It is followed by rapid formation of the postnaupliar appendage buds. The first part of the developmental sequence is dominated by formation of appendage buds and ganglion anlagen which all occur in strict anteroposterior progression. The majority of myogenic events is concentrated in the second half of the sequence and shows no trace of an anteroposterior gradient in development.

P. hawaiensis (Amphipoda)
P. hawaiensis shows loss of the longitudinal muscle precursors [lmp-mx1] and [lmp-mx2] in the first and second maxilla segments. Formation of extrinsic appendage muscle precursors [md-m], [md-l], and [mx1-l] is shifted late while formation of the first thoracic and sixth pleon ganglion anlagen [t1-g] and [p6-g] are shifted to earlier positions. As a result, appendage bud and ganglion anlage formation are even more concentrated at the beginning of the sequence. A temporal gap between formation of naupliar and postnaupliar features is shown only for appendage buds, but not for neurogenic or myogenic events.

Discussion
Tissue-related evolution of developmental timing Studies on developmental genetics of the fruit fly Drosophila melanogaster [98] and experimental developmental studies on P. hawaiensis [99] suggest that the ectoderm has a strong regulatory influence on the development of the mesoderm in Arthropoda, but not vice versa. This does not necessarily imply that heterochronies within the ectoderm are unlikely. Fritsch & Richter The table shows event numbers and heterochrony rates for the three tissue types from which the events were coded: Epidermis, muscle tissue, and nervous tissue, as well as the germ layers the tissues originate from (ectoderm, mesoderm). The heterochrony rate of an event is the event-specific number of heterochronic changes shown by PGi for the entire superconsensus tree, multiplied by the mean bootstrap value for the event. Tissue-specific heterochrony rates were calculated by forming the mean heterochrony rate per event for all events of epidermis neural tissue or muscle development. Mean heterochrony rates of the germ layers were calculated from these values.
[12] describe several instances of intraectodermal heterochrony in evolution of Branchiopoda. Yet tracing the evolutionary history of eumalacostracan, developmental timing by PGi in our study showed that the different tissue types (epidermis, nervous system, musculature) have taken different evolutionary paths depending on the germ layer they originate from. The mean heterochrony rates of ectoderm and mesoderm development that were calculated from the results of PGi analysis differ strongly (Table 4). Ectodermal development is generally less affected by heterochrony in malacostracan evolution than mesodermal development. Almost no heterochronic events appear in the naupliar region if only the ectodermal development is considered. Heterochrony of muscle precursor formation is far more extensive in the first antennal, second antennal, and mandibular segments than neural development and formation of appendage buds (Figure 10b). This suggests that the divergent evolution of developmental timing between mesodermal and ectodermal tissues reflects a modular property of the crustacean developmental system as would be expected due to findings from genetic and experimental developmental biology [98,99]. Reconstruction of the ancestral developmental sequence by computational heterochrony analysis with PGi suggests that the naupliar pattern, known as 'egg nauplius stage' , was present in the last common ancestor of Eumalacostraca as combination of epidermal and neurogenic, but not muscle developmental patterns. In the last common ancestor of Caridoida and Peracarida, the reconstructed sequence shows only an 'epidermal egg nauplius'. Formation of naupliar ganglion anlagen, however, occurs only slightly later in the sequence. The persistence of the very early timing of epidermal and neural naupliar events in these lineages could be explained as the result of a developmental constraint that limited the plasticity of developmental timing in ectodermal development compared to mesodermal development. This constraint would represent a modular property of the developmental regulatory system patterning the naupliar ectoderm, similar to the observations on other arthropods [98,99]. Evolutionary alterations of timing in ectodermal development of the naupliar segments would thus have had a stronger impact on the developmental outcome and viability of the organism and therefore would have been more likely eliminated by selection, than timing alterations in mesoderm development. Of course this explanation is based only on cross-species comparison of timing patterns and not on experimental investigations of the developmental systems. Nevertheless, our findings support the hypothesis that heterochronic change of muscle development played a major role in evolutionary loss and reacquisition of the nauplius larva.

Evolution of naupliar developmental patterns in Malacostraca
In arthropod development, commonly the material of a variable number of anterior segments is laid down in a different manner than following segments that are added posteriorly during development. This is reflected by the process of short germ development in embryogenesis or anamorphic postembryonic development [100], and references therein and represents a condition of the arthropod ground pattern. A naupliar developmental pattern, meaning that the material of the first antennal, second antennal, and mandibular segments is formed (more or less) simultaneously before the posteriorly following segments, can be understood as a specialized form of the arthropod developmental pattern. The plesiomorphic condition for crustaceans (or Tetraconata) including extinct representatives of the stem lineage (called Crustacea sensu lato in [101]) was a 'head' larva with functional first antennae followed by three pairs of appendages. A larva bearing three appendage pairs-the nauplius-is considered apomorphic for crowngroup Tetraconata [102] (Pancrustacea) (Eucrustacea or Crustacea sensu stricto in [101]) [17]. Our analysis reveals an ancestral developmental sequence for the branchiopod/ malacostracan clade with an extensive egg nauplius phase. However, the reconstructed sequence is not fully compatible with a developmental mode comprising a freeswimming nauplius larva because some naupliar myogenic events are formed only after the egg nauplius phase and hatching occurs only after of the second maxilla and first thoracopod bud are formed. Yet a large number of postnaupliar developmental events, such as formation of postnaupliar ganglion anlagen and muscle precursors, occur after hatching which is in line with nauplius larva formation. Since a nauplius larva is predominant throughout the (See figure on previous page.) Figure 11 Heterochronic changes and ancestral developmental sequences for major malacostracan nodes. Simplified phylogram of Malacostraca with ancestral developmental sequences from the PGi superconsensus tree. Sequences are shown as columns of downward pointing arrows. Each arrow represents a different semaphoront containing a single event or a group of simultaneous events (abbreviations listed in Table 1). The position of Leptostraca is indicated by a dotted line. Ancestral ontogenetic sequences of the branchiopod/malacostracan, the eumalacostracan, the caridoid, and the decapod LCA are shown. Events are color-coded corresponding to Figures 3 and 10a. Heterochronic changes are indicated by horizontal arrows that use the same color code as the respective events. The egg nauplius phase is indicated by brackets. A symbolic nauplius drawing marks the terminal taxa, which develop nauplius larvae. Arrows with a 'plus sign' mark events that are interpreted as new evolutionary acquisitions on the respective branch. Arrows with a 'minus sign' indicate evolutionary loss of a feature. Abbreviations: ENP egg nauplius phase. Event abbreviations are listed in Table 1.
remaining crustacean taxa, we suggest that this was also the condition in the last common branchiopod/malacostracan ancestor and that the late position of many naupliar events is an artifact caused at this basal node by the extensive variation in the data set.
The developmental sequence of the malacostracan ground pattern could not be reconstructed, because Leptostraca, the sister group of Eumalacostraca according to the phylogeny of Richter & Scholtz [28], is not present in our taxon sampling. Embryogenesis of the leptostracan Nebalia bipes has been described [52,53,103] but unfortunately not sufficiently to integrate this species into the analysis. It is known however that N. bipes lacks free-swimming larval phases. Also, an egg nauplius stage in appendage morphogenesis is described for this species. The developmental pattern of the malacostracan last common ancestor can therefore not be expected to differ much from the eumalacostracan last common ancestor in these respects.
The eumalacostracan last common ancestor, according to heterochrony analysis with PGi, possessed a developmental sequence with late position of the hatching event and thus major postnaupliar developmental events occurring in embryogenesis. The late shift of the hatching event to the end of the sequence, the early shift of postnaupliar muscle precursor, ganglion anlagen, and appendage bud formation suggests that in the lineage leading to Malacostraca a change of ontogenetic mode took place and the nauplius larva was lost. Our analysis suggests that only formation of naupliar appendage buds and naupliar ganglia remained part of the egg nauplius phase. The egg nauplius stage [67,69] was thus likely restricted to ectodermal tissues already in the eumalacostracan ground pattern.
Along the branches leading from the eumalacostracan to the caridoid and to the decapod last common ancestor, comparatively few heterochronies are recovered by our analysis. Formation of two naupliar muscle precursors is shifted to an earlier position while naupliar appendage bud formation and formation of naupliar ganglia are retained close to the beginning of the sequence. The developmental sequence reconstructed for the ground pattern of Decapoda shows an extensive egg nauplius phase, comprising all naupliar ectodermal events and formation of part of the naupliar muscle precursor group. Thus naupliar myogenic events must have been added to the egg nauplius stage in development before the emergence of a nauplius larva in the evolution of Decapoda.
The lineage leading to Dendrobranchiata represents a change in developmental mode and evolution of a freeswimming nauplius larva. Compared to the eumalacostracan stem lineage where the nauplius larva was lost, S. ingentis shows only few (five) heterochronies, of which only three are relevant for the reacquisition of the nauplius larva: early shift of the two muscle precursors of the first antenna and early shift of the hatching event. These changes were sufficient for the reacquisition of a free-swimming nauplius larva because the other necessary events were already in place in the developmental sequence of the decapod ground pattern. This refers to the naupliar appendage bud formation and formation of naupliar ganglia, which constitute the ectodermal egg nauplius stage in embryogenesis, as well as the early positions of naupliar myogenic events that are the result of heterochronic shifts in the lineages leading to Decapoda. Therefore, our results support the hypothesis formulated by Scholtz [67] that an embryonic egg nauplius served as a prerequisite for the secondary evolution of the dendrobranchiate nauplius larva.
Within Decapoda, in the lineage leading to Pleocyemata, an egg nauplius pattern in ectodermal development has been retained, together with a set of naupliar muscle precursors in the egg nauplius phase, according to our results. This pattern differs only minimally from the decapod ground pattern. Yet timing of naupliar myogenesis is altered within the Pleocyemata in the lineages leading to N. heteropoda and P. fallax forma virginalis. We should note that both species are direct developers, which is a derived condition within decapods. Yet the results of PGi suggest that the pleocyemate ground pattern is reminiscent of the decapod ground pattern and that alterations to myogenesis have occurred only within the group. Adding taxa with a more basal phylogenetic position and a zoea-like larva in future studies can be expected to uncover a similar condition.
It should be noted that the other malacostracan taxon with a nauplius larva, the Euphausiacea, is not represented in our study. Following the phylogeny used here, Euphausiacea are the sister group of Neocarida. Another popular phylogenetic hypothesis places Euphausiacea and Decapoda together in a monophylum called Eucarida [104,105]. Thus mapping timing data on these two alternative hypotheses could improve our understanding of malacostracan phylogeny and clarify whether a nauplius larva evolved once or twice within Malacostraca. It is unlikely however that inclusion of Euphausiacea will (See figure on previous page.) Figure 12 Heterochronic changes and ancestral developmental sequences within Decapoda. Phylogeny of Malacostraca as in Figure 11, but showing only Decapoda. Ancestral ontogenetic sequences reconstructed by heterochrony analysis using PGi are shown for the decapod and the pleocyemate LCA, as well as the terminal sequence of S. ingentis (Dendrobranchiata). Abbreviations: ENP egg nauplius phase. Event abbreviations are listed in Table 1. significantly change the reconstructed developmental mode in the eumalacostracan or malacostracan ground pattern, as this depends on the position of other not included taxa, such as Leptostraca, Amphionidacea, and Syncarida.

Evolution of a zoea-like larva
In Malacostraca, both ectodermal and mesodermal tissues, from the second maxillary to the sixth pleonal segment, are formed sequentially by proliferation of stem-like cells (ectoteloblasts and mesoteloblasts) in the posterior growth zone [61,62,106,107]. This mechanism leads to an observable anteroposterior gradient of segment differentiation in the germ band. We have recently described a growth zone independent muscle precursor 'lmp-post' which likely plays a crucial role in development of a zoea-like larva [67,73]. lmp-post is formed in the telson and extends anteriorly while at the same time longitudinal muscle precursors form in the anterior postnaupliar segments. This way a continuous longitudinal muscle strand across all trunk segments is formed, even before the full set of trunk segments is differentiated. Zoea-like larval forms share a functional, movable trunk, consisting of the thoracic or pleonal segments and a paddle-shaped telson which can perform extension and flexion movements, and actively participate in swimming, e.g., by performing tail flip escape reactions. Activity of [lmp-post] allows trunk functionality before the posterior trunk segments are fully developed, as is the case in zoea-like larvae. Further common features of zoea-like larva development in terms of developmental timing are rapid formation of appendages in an anterior set of postnaupliar segments (first and second maxillae, thoracopods), late formation of pleonal segments [P6], [P6-g], [lmp-p6] (with the exception of the stomatopod pseudozoea), and late offset of segment formation in the germ band [FS]. Event data for the second thoracic to the fifth pleonal segment was not analyzed because the data could not be acquired for a sufficient amount of species and semaphoronts.
The scenario reconstructed by PGi suggests that a zoea-like larva likely evolved independently in the lineages leading to Stomatopoda and Decapoda and that the eumalacostracan last common ancestor developed directly. The posterior longitudinal muscle primordium [lmp-post] which we consider a necessary feature for zoea-like larval motility was acquired twice independently in the lineages leading to Stomatopoda and Decapoda and was not part of the eumalacostracan ground pattern. The appendage bud and longitudinal muscle precursor of the sixth pleonal segment are formed early in the sequence reconstructed for the eumalacostracan last common ancestor, suggesting that functionality of the trunk did not precede differentiation in the posterior pleon segments, and consequently that direct development rather than a zoea-like larva constituted the developmental mode. We point out that the conclusions on this early node should be treated with caution because variation in the data set is extensive and also because we cannot rule out the possibility of bias due to the strong representation of direct development in the analysis.
For the last common ancestor of Decapoda, a zoea-like larva as hatching stages appears well supported. In the developmental sequence of the decapod ground pattern, appendage bud formation and formation of the longitudinal muscle precursor in the sixth pleonal segment occur close to the end of the sequence. Offset of segment formation also occurs late while the anterior postnaupliar appendage buds are formed simultaneously and just after the egg nauplius phase. Also all anterior longitudinal muscle precursors and lmp-post are formed simultaneously. The same is true for the extrinsic appendage muscle precursors of the anterior postnaupliar segments. In the lineage leading to Dendrobranchiata, timing of postnaupliar events relevant to zoea-like larva formation remains nearly unchanged (with the exception of [t1-g]). The evolution of the novel developmental mode of Dendrobranchiata, involving the novel larval stages metanauplius, protozoea and mysis stages, from an ancestral condition with a more extensive embryonic period did not depend on changes in developmental timing of the analyzed morphogenetic events. It is the predisplacement of the hatching event that makes the actual difference between larval and embryonic development, while the sequence in which appendage buds, ganglion anlagen, and muscle precursor are generated remains largely unchanged. Certainly, acceleration of differentiation processes which follow the formation of appendage, ganglion, or muscle anlagen in the segments of a viable free-swimming larva must be assumed for the evolution of the dendrobranchiate ontogenetic mode. In the lineage leading to Pleocyemata, [FS] and [lmp-p6] are shifted to earlier positions. Both changes point toward loss of zoea-like larva formation, but the majority of relevant events is still in place. Both pleocyemate representatives (N. heteropoda and P. fallax forma virginalis) develop directly and lack larval stages. Therefore, we consider early placement of [FS] and [lmp-p6] as bias toward direct development in our data set, not necessarily as part of the pleocyemate ground pattern.
(See figure on previous page.) Figure 13 Heterochronic changes and ancestral developmental sequences within Peracarida. Phylogeny of malacostraca as in Figures 11  and 12. Only the caridoid LCA, the peracarid LCA, and the terminal sequence of P. hawaiensis are shown. Abbreviations: ENP egg nauplius phase. Event abbreviations are listed in Table 1.

Evolution of developmental timing in Peracarida
Peracaridan development is derived in many respects relative to the malacostracan ground pattern, because of the advanced mode of brood care that is autapomorphic to this group [20,28,[108][109][110]. In Peracarida, females possess a ventral brood pouch (marsupium), in which eggs are reared. Nauplius or zoea-like larvae are not found in peracarids. Changes to the developmental sequence in the peracarid stem lineage comprise loss of the nauplius eye and the posterior pioneer neurons, as well as late shift of several naupliar and early shift of postnaupliar events. In the ancestral peracaridan developmental sequence, naupliar and anterior postnaupliar appendage buds are formed rapidly at the beginning of the sequence, while naupliar ganglia are formed late. Offset of segment formation and formation of the posteriormost pleonal appendage bud occur early while the majority of muscle precursors are formed late. These properties of the developmental sequence do not resemble zoea-like developmental timing patterns. The condition in the caridoid ground pattern is difficult to interpret in terms of developmental mode. The late position of pleonal events suggests zoea-like larva formation but the posterior longitudinal muscle primordium [lmp-post] is not formed in the sequence. The question whether direct development might be plesiomorphic for Peracarida can therefore not be answered at this point.
Within Peracarida, most likely development was consistently adapted to efficient formation of juvenile body morphology after the advent of the new developmental mode, which was constrained by the specialized mode of maternal brood care. In P. hawaiensis (Amphipoda) the adaptation to efficient formation of juvenile body structure is intensified. Here the events of the six anterior segments appear in an order corresponding to the tissues they belong to: appendage buds, followed by ganglion anlagen, followed by muscle precursors. Together with the early offset of segment formation, this suggests a strong acceleration of morphogenesis in this lineage, which resulted in more rapid anteroposterior progression of segment formation and an earlier onset of tissue differentiation.
In Mysidacea, an inert larval stage hatches, and remains in the marsupium, a situation we call pesudodirect development. The hatchling is termed 'nauplioid' [110,111]. The name is suggestive of a cryptic larval stage related to a nauplius larva. Also the early hatching event, presence of a solid cuticle with setation, and intramarsupial molting to the 'postnauplioid' stage suggest that a part of an ancestral larval developmental program is still active in Mysidacea. The evolutionary scenario reconstructed with PGi suggests that the mysid sequence is derived from an ancestor with only an epidermal egg nauplius phase and a late position of the hatching event. This may be an artifact of insufficient taxon sampling. However, Mysidacea show a unique timing pattern of appendage bud development with the first and second antennal bud being formed clearly before the mandible bud, early offset of segment formation, and finally late formation of naupliar ganglia and musculature. These observations suggest that the developmental pattern found in Mysidacea is not homologous to the egg nauplius pattern of Eumalacostraca.

Conclusions
Our reconstruction of developmental sequence evolution of Malacostraca revealed that development of musculature has played a crucial role in evolutionary transitions between larval and embryonic development. The following conclusions can be drawn from our analysis of heterochrony: The eumalacostracan last common ancestor has retained the developmental timing pattern of nauplius larva formation in epidermal appendage development and neurogenesis, but not in myogenesis. The ontogenetic mode using a nauplius larva was replaced most likely by direct development in the lineage leading to the Malacostraca by delay in naupliar muscle development. Secondary evolution of the dendrobranchiate nauplius larva involved only little heterochronic change, because the major features of naupliar development were present already in the decapod last common ancestor. The transition relied on early shift of naupliar muscle precursors. According to our analysis, convergent evolution of a zoea-like larva in the stomatopod and decapod lineage is more likely than a zoea-like larva in the eumalacostracan last common ancestor. The developmental sequence of the peracarid last common ancestor has lost the larva-related timing patterns in embryogenesis. Developmental timing was likely adapted to efficient formation of juvenile body structure under the constraint of specialized brood care within the Peracarida.
Some key taxa of Malacostraca have not been sampled here: Leptostraca, Anaspidacea, Bathynellacea, and Euphausiacea. Also inclusion of additional event data, considering the thoracic and pleonal segments, more advanced stages of tissue differentiation, or the formation of external cuticular structures would contribute to a more detailed picture of malacostracan developmental evolution. Such investigations have the potential to further clarify the evolutionary history of malacostracan development, but this is left to future studies.