Specimens of the semi-terrestrial amphipod species Orchestia cavimana were collected from beaches of the Tegeler See (Berlin). The animals were maintained in a terrarium at 18–20°C and fed with carrots and oatmeal. To receive eggs in relevant stages gravid females were caught and isolated. The egg-bearing females where carefully anaesthetized in mineral water containing CO₂. In their ventral brood pouch – the marsupium – the purple colored eggs are easy to recognize. The eggs were flushed out with a Pasteur pipette and transferred to a saline solution that mimics the osmotic milieu in the marsupium (details described in Wolff and Scholtz 39).

Embryos in relevant stages (shortly before the first ectodermal rows appear) were mounted on microscopic slides. A ring of Vaseline was formed on the slide and this small "basin" was filled with amphipod saline and covered with small cover slips. By a careful movement of the cover slips the embryos can be arranged in a suitable position and arrested with a little pressure on the slip's top.

The fundamentals of 4D-microscopy are described by Schnabel et al. 38. The image stacks were analyzed with the software SIMI°BioCell (SIMI, Germany). The data are illustrated as 3D-representations with color coded spheres. About 15 embryos were observed during the first row formation up to the row-like formation of the region E(0).

The cell labeling was done with an inverse microscope equipped with a micromanipulator (Leica DMIRB). The eggs in the relevant stages were put on microscopic slides under small cover slips that were equipped with plasticine feet at the corners. The eggs could be brought into the desired position by carefully shifting the cover slip. The eggs were held for the injection with soft pressure on the cover slip.

To get suitable needles for the injection, pipettes (Hilsberg, diameter 1.0 mm, thickness 0.2 mm) were pulled (KOPF Puller 720). After pulling the tips of the needles are closed and had to be open and sharpened with a horizontal grinder (Bachofer). The angle of the cutting edge varied between 20 and 30 degrees. The fluorescent marker was sucked into the injection-needle. DiI (Molecular Probes) was used as a vital marker. DiI is a lipophilic fluorescence-dye that binds to the cell membrane. This guarantees that the dye is exclusively restricted to the daughter cells. After a defined period of development the labeled eggs were observed with a fluorescence-microscope (Zeiss Axiophot1) using blue light or green light (strongest stimulation of DiI), and the results were documented with a digital camera (Nikon D1).

For fixation and documentation on the laser scanning microscope (Leica SP2) the embryos were dissected in PBS-buffered 4% formaldehyde-solution, counterstained with nuclear staining dye (Hoechst) and mounted in DABCO-Glycerol (25 mg DABCO (1,4 diazabicyclol-2,2,2-octane, Merck) in 1 ml PBS to 9 ml glycerol), which is an anti-bleaching-detergent. The image stacks produced by the laser scanning microscope were analyzed with the software Imaris 5.0.1 (Bitplane AG). The 3D-reconstruction of the counter staining (Hoechst) and the clones of the in-vivo labeled cell have the advantage of very high resolution with respect to morphological data. The feature "Volume" in the program module "Surpass" created a three dimensional object, which can be magnified and moved in all directions. For better visualization of the objects, movies (AVI-files) were created in the program module "Animation".

For the following description we adopt the common nomenclature for malacostracan crustacean cell lineages which was modified for amphipods by Scholtz 36. In addition we introduce a nomenclature for the ectoderm row E(0). The anterior row of E(0) is named E(0)a and the posterior row E(0)p. If after the subsequent divisions a cell lies anterior it is again labeled with an "a". The posterior sister cell is labeled correspondingly with "p". As in more posterior rows the cells in positions to the midline (these cells and their early descendants are designated as columns) are numbered consecutively from the middle towards lateral (e.g. E(0)p1).

After gastrulation, the blastopore is closed. The germ disc consists of about 250 cells (Fig. 2A). The descendants of the macromeres B (left) and D (right) form the anterior parts of the germ disc by migration towards the middle (for details see Wolff and Scholtz 39). At this time there is no regular cellular pattern in the ventral ectoderm. The first recognizable transverse ectoderm row occurs in a germ disc with about 400 cells (Fig. 2B). This row marks the border between the naupliar (head lobes, the segments of the first and second antennae and of the mandibles) and the post-naupliar (the segment of the first maxillae up to the telson anlage) regions of the developing embryo. This first recognizable cell row forms the genealogical unit E(1) which provides most of the material for the segment of the first maxilla (Fig. 2C). The cells anterior to E(1) show a typical spatial configuration. According to Scholtz 36 we named this region E(0), which forms part of the prospective mandibular segment (Fig. 2C). The cells of E(0) are somewhat smaller and rounder than the more posterior cells. They are arranged in two rows of at least 6 cells on either body half (Fig. 2C, 3A–B). The anterior row is designated as E(0)a and the posterior row as E(0)p (Fig. 3A,B). It is not clear whether these two rows originate from one transverse ectoderm row by longitudinally oriented cell divisions. A distinct unpaired column of midline cells as is typical for the post-naupliar region (Fig. 2E) does not exist in the naupliar part of the germ band. In the median area of the posterior naupliar region E(0) we found about 10–15 smaller cells in a V-shaped arrangement (Fig. 3A,B).

At about this stage, cells of the anterior row E(0)a divide more or less synchronously (mitotic wave) in a longitudinal direction. The products are the anterior row E(0)aa and the posterior row E(0)ap (Fig. 3A,B). The following division pattern of cells in row E(0)aa is not clear in detail but the cells divide symmetrically in each body half (see supplemental material). In row E(0)ap a detailed pattern could not completely be described but some features of the more medially lying cells of this row could be traced. The cells E(0)ap1 und E(0)ap2 are the first to divide in a longitudinal direction whereas E(0)ap3 divides a bit later in a horizontal direction (Fig. 4E). The cell E(0)ap4 shows a delayed division. In relation to the division cycles in row E(0)a, the cells of E(0)p are delayed in their division.

During the second mitotic wave in row E(0)aa, E(0)p2 is the first cell that divides in row E(0)p. It divides in a longitudinal direction. More or less at the same time the midline cell E(0)p0 divides also in a longitudinal direction. Subsequently the cells E(0)p3, E(0)p5, and E(0)p6 divide likewise in a longitudinal direction E(0)p2 (Fig. 4B,C). At the same time the relatively large cell E(0)p1 divides in a horizontal direction (Fig. 4C). At last, the cell E(0)p4 undergoes a division, again in a horizontal direction (Fig. 4D). In this developmental stage small buds of the naupliar appendages (first and second antennae, mandible) become visible (Fig. 4F). This invariant cleavage pattern in the prospective mandibular region eventually produces a reproducible arrangement of cells in the posterior region of E(0)(Fig. 4E).

This is an important pre-condition for our detailed lineage study of the mandibular segment. Unfortunately, we could not reconstruct the complete lineage pattern of the more anterior lying region E(0)a.

Identified cells of the transverse ectoderm rows around the boundary between the naupliar and post-naupliar regions were labeled in-vivo during the early germ band stage. The identification of cells is the prerequisite for the analysis of further cell fates and the clonal composition of morphological structures. Since a single cell approach was not feasible throughout, we divided the region E(0) from median to lateral into three sections. The median area, area I, begins lateral to the midline and ends before E(0)p2. The area II starts from E(0)p2 and comprises cell E(0)p3. The most lateral area III reaches from E(0)p4 to the end of the visible row formation (Fig. 4E). The successful in-vivo markings could be assigned to the corresponding areas.

The median region of the mandibular segment does not consist of an unpaired column as is known for the post-naupliar ectoderm (the so-called midline, see 4041). From the onset, about 10 smaller cells are more sunken into the yolk than are the surrounding symmetrically arranged cells. Thereby, the median cells form a typical V-shaped region (Fig. 3A). During ongoing development a clear separation (in comparison with surrounding lateral cells) is recognizable (Figs. 3C, 4F). The surface of this median region loses its cellular character and has a smooth appearance.

In relation to the more lateral adjoining cells (area II and III) the cells of area I of E(0) are retarded in cell division during segment formation. They form only a small portion (10–15 cells) of the early mandibular segment, and have their origin in the first column of region E(0) – E(0)a1 and E(0)p1.

In vivo labeling of cell E(0)a1 shows that its descendants form the medio-anterior part of the developing humps of the paragnaths. In addition, a median part of the mandibular sternite is formed by E(0)a1-descendants.

The descendants of E(0)p1 form postero-median parts of the paragnaths and adjacent areas of the sternite (Fig. 5A–C, Additional file 1). A few clones of E(0)p1 are found within the neuro-ectoderm in the posterior region of the mandibular segment and form nervous structures. Finally, after further development the ectodermal descendants are involved in the formation of postero-median parts of the paragnaths.

In general, area I forms the median mandibular sternite and median parts of the paragnaths. The clonal composition of a developed paragnath and developed parts of the sternite reflects the early a/p-arrangement of the rows.

The lateral adjoining area of area I is area II, which comprises the descendants of column 2 and 3. It forms lateral parts of the paragnaths and median parts of the mandible. Descendants of column 2 are found in more lateral parts of the paragnath and in antero-median parts of the mandible (Fig. 6A–C, Additional file 2). Interestingly, these median parts of the early mandibular bud are later differentiated into the molar process.

The cells of column 3 give rise to median parts of the mandible (Fig. 6D–E, Additional file 2). As in area I, the descendants of the anterior row E(0)a form more anterior parts and according to that the descendants of row E(0)p proliferate more posterior parts. As well, some cells of the sternite have their origin in cells of area II.

In general, area II is responsible for the formation of lateral parts of the paragnaths and more median parts of the mandibles.

Area III lies laterally adjacent to area II and comprises columns 4 and 5 of region E(0). They form lateral parts of the mandible and neighboring parts of the tergite. That is why the cell E(0)p4 produces cell material for the outer (postero-lateral) part of the mandible and lateral bordering tergite (Fig. 7A–C, Additional file 3). In contrast, descendants of row E(0)a in area III form more anterior parts of the outer (antero-lateral) mandible and parts of the adjacent tergite. Remarkably, the lateral portion of the mandibles differentiates by cell proliferation into the prospective pars incisiva.

Summarizing for area III, column 4 forms lateral parts of the mandibles and adjacent tergites and column 5 is not involved in the development of mandibles or paragnaths.

Like all malacostracan crustaceans studied in this respect (for a recent review see Dohle et al. 42), amphipods show a stereotyped cell divisions pattern in the post-naupliar region during growth, differentiation and segmentation of the germ band 364344. In contrast to this, the naupliar region does not exhibit an obvious stereotyped cell division pattern 3742. Only Scholtz 36 suggested that there might be a certain regularity in the divisions and arrangements of the posterior cells of the developing mandibular segment in the amphipod Gammarus pulex but with the methods then at hand the details were not resolvable With the technique of 4D-microscopy we have been able to provide the first evidence for an invariant cell division pattern in the naupliar region of another amphipod species, the freshwater beach hopper Orchestia cavimana. At least in the posterior part of the region of E(0) we recognized a relatively strict cell division pattern. This pattern has only superficial similarities to the post-naupliar cell division pattern of malacostracan crustaceans but it is not as elaborated in terms of timing of mitoses, cell size, and the spatial arrangement of the resulting cells. In E(0) the sequence of the individual divisions is not as strict and the direction of the mitotic spindles is more or less longitudinally oriented. It appears that the posterior border of the mandibular segment forms some kind of transition between the more irregular divisions and cell arrangements of the anterior naupliar region and the highly complex stereotyped post-naupliar patterns.

It is not clear whether our findings of a regular pattern in the posterior part of the mandibular segment holds true for other malacostracans as well, although some data from the isopod Porcellio scaber hint to that possibility 45. However, the cellular events in the corresponding region in Porcellio are much more irregular when compared with those in Orchestia. It has been even shown for Porcellio that some cells of the most anterior row of the post-naupliar segments can migrate into the posterior area of the mandible segment 45, a phenomenon that does not occur in Orchestia.

Interestingly enough, our results reveal that the posterior segmental boundary of the mandibular segment corresponds to the genealogical border between rows E(0) and E(1), i.e. E(1) does not contribute to the posterior part of the mandibular segment. This stands in contrast to all more posterior post-naupliar segmental boundaries which are formed within the descendants of one ectoderm row and thus do not match the genealogical borders (see Fig. 2), 42. Row E(1) forms a kind of transition between these segmentation modes because its posterior region follows the typical post-naupliar pattern in that it contributes to the anterior portion of the segment of the first maxillae whose posterior part is formed by anterior descendants of the next adjacent row E(2) 3637. These differences indicate that the parasegmental organization (i.e. a frame-shift between the early metameric anlagen and the resulting morphological segments) of the post-naupliar germ band (see 42) is not found in the naupliar region with a transition in the first maxillary segment.

Based on our knowledge of the cell division pattern of the early developing mandibular region (see above) we were able to look at its morphogenesis at a very high level of resolution. By means of single cell labeling with the fluorescent dye DiI we were able to reconstruct and analyze the clonal composition of the mandibular region from the beginning of ectodermal proliferation up to the differentiation of the mouthparts. The cell labeling reveals that the paragnaths have their origin in the area I and area II which comprises columns 1 to 3 of region E(0). The mandibles originate from cells of the areas II and III (columns 2 to 4). Areas I and II contribute also to the sternal region and the mandibular ganglia whereas area III forms parts of the tergites as well. In more posterior segments, columns 1 and 2 mainly contribute to the formation of segmental ganglia and probably sternites, and columns 3 to 5 mainly give rise to limbs 3646. Hence, when compared with clonal composition of the post-naupliar segments it is evident that the columns that form the ganglia and sternites in these segments correspond to those that give rise to the paragnaths in the mandibular segment. In addition, the mandibular buds are formed in a comparable position to the other limbs. This clearly reveals that paragnaths of Orchestia are processes of the mandibular sternal region.

Further evidence for the claim that crustacean paragnaths belong to the mandibular segment is based on gene expression data. For instance, the segment polarity gene engrailed is expressed in a regular stripe in the posterior region of the mandibular segment, as in all other segments, comprising cells that form the posterior part of the paragnaths in amphipods, isopods and decapods 141837. Moreover, the expression of the Hox-gene Deformed (Dfd) is mainly found in the mandible segment of hexapods 4748, myriapods 4950, and crustaceans 1416, and in the latter case expression of Dfd comprises the buds of the paragnaths 1416.

All these data reveal that paragnaths are part of the mandibular segment and that they are not limb derivatives as has been suggested by several authors. This confirms previous ideas on the origin and nature of paragnaths based on embryological and larval evidence 17. With our clonal analysis we can definitely rule out the possibility that paragnaths indicate the existence of an additional head segment as has been claimed for example by Casanova 29, Chaudonneret 26, Denis 34, and Hansen 28. Furthermore, the limb-bud like early appearance of the paragnaths in Orchestia and other crustaceans is no indication for a limb-related nature of these structures but is only a superficial similarity that does not represent a genealogical relation. Our results show the power of cell lineage and clonal analyses for inferences on the nature, origin and thus homology of morphological structures. With this kind of investigation morphological and gene expression data can be complemented.

In many cases the origin of the paragnaths during crustacean embryonic and larval development has either not been specified (e.g. Ref. 515253) or it has been suggested that paragnaths develop from mandibular and/or maxillary segments (e.g. Ref. 54). A look at the corresponding figures in these articles with our results in mind reveals that it is possible in almost all examples to relate the paragnathal structures to the mandibular segment (e.g. Manton 51, plate 24, fig. 27; Manton 52, plate 25, fig. 23; Moeller 53, figs. 2, 7). Even the median lobe of the so-called lower lip of the raptorial cladoceran Leptodora kindtii might represent fused paragnaths 55. Accordingly, we tentatively conclude that the paragnaths lobes are homologous throughout Crustacea. Stein et al. 56 suggest that a pair of paragnaths humps is an apomorphy for a crustacean subgroup comprising Eucrustacea and Phosphatocopina (Labrophora). However, these authors did not consider similar structures in myriapods and hexapods which indicate a much more widespread occurrence among euarthropods. (see next chapter).

There are several reports of post-oral paired bud-like anlagen in myriapods (e.g. Chilopoda: Heymons 57, Progoneata: Tiegs 19) and Hexapoda (e.g. Ref. 2425585960). In dicondylian hexapods the situation is somewhat ambiguous. Larink 61 and Scholl 62 report for Lepisma and Carausius an undivided early hypopharynx anlage whereas Rohrschneider 24 for Periplaneta and Ibrahim 63 for Tachycines describe two separated buds. Whether these are factual differences is not clear. However, the presence of paired buds in a number of pterygotes as well as collembolans, diplurans, and archaeognathans allows the conclusion that these paired buds were present in the hexapod stem species. Nevertheless, all these buds have in common that they originate from the sternum of the mandible segment directly ventral to the mandibular ganglion anlagen and between the mandibular limb buds, after these have formed. This very much resembles the early stages of crustacean paragnaths (see above, 1417185152). These sternal buds of the mandibular segment give either rise to the superlinguae, paired lateral lobes of the hypopharynx, in Symphyla 19, Collembola 5859, Diplura, Archaeognatha 25, or to (another part of) the hypopharynx 222457 when superlinguae are not present as for instance is the case in many Pterygota and all Chilopoda 2264. However, the contribution of these mandibular sternal buds to the hypopharynx body (linguae) is interpreted controversially. The hypopharynx of Hexapoda and Progoneata is thought to be a composite structure formed by protrusions of the sternites of different numbers of gnathal segments (e.g. Ref. 1922256162). In contrast to this, Haget 65 and Wada 66 suggest, based on their experimental teratological studies that the entire hypopharynx originates from the mandibular sternites. According to Heymons 57, this is also the case in Chilopoda. The situation is further complicated by the fact that buds appearing in the intercalary segment of hexapods are called "hypopharyngeal lobes" (Hypopharynxhöcker) and have been suggested to form (part of) the hypopharynx, an obvious misinterpretation (see Ref. 236367). Reading the numerous articles dealing with this problem it is evident that investigations focusing on the differentiation of the hypopharynx in myriapods and hexapods using modern approaches are badly needed to solve this issue.

Authors such as Crampton 27, Snodgrass 68, and Bitsch and Bitsch 5 suggest that the paragnaths and the superlinguae are homologous. However, based on our data and what we find in the above discussed literature about the development of the hexapod and myriapod hypopharyngeal complex, we think that the conclusion of these authors is simplifying the matters because paragnaths cannot be homologous to superlinguae alone if the major part of the hypopharynx of myriapods and hexapods originates from the mandibular sternites as well. Accordingly, we modify the homology statement concerning crustacean paragnaths and the myriapod/hexapod superlinguae/hypopharynx by suggesting that the early anlagen of these structures are homologous, taking into account that the homology of early developmental stages does not necessarily mean that more advanced stages are also homologous 69. Since a comparable structure is absent in the corresponding segment of Chelicerata and Onychophora (first walking leg, see Ref. 12707172) it is likely that a pair of sternal buds in the mandibular segment is a shared apomorphy of Crustacea, Myriapoda, and Hexapoda. Accordingly these structures provide further support for the Mandibulata hypothesis (see also Ref. 2737475) which has been disputed based on molecular data (e.g. Ref. 767778).