Evolutionary transformation of mouthparts from particle-feeding to piercing carnivory in Viper copepods: Review and 3D analyses of a key innovation using advanced imaging techniques

Background Novel feeding adaptations often facilitate adaptive radiation and diversification. But the evolutionary origins of such feeding adaptations can be puzzling if they require concordant change in multiple component parts. Pelagic, heterorhabdid copepods (Calanoida) exhibit diverse feeding behaviors that range from simple particle feeding to a highly specialized form of carnivory involving piercing mouthparts that likely inject venom. We review the evolutionary history of heterorhabdid copepods and add new high-resolution, 3D anatomical analyses of the muscular system, glands and gland openings associated with this remarkable evolutionary transformation. Results We examined four heterorhabdid copepods with different feeding modes: one primitive particle-feeder (Disseta palumbii), one derived and specialized carnivore (Heterorhabdus subspinifrons), and two intermediate taxa (Mesorhabdus gracilis and Heterostylites longicornis). We used two advanced, high-resolution microscopic techniques — serial block-face scanning electron microscopy and two-photon excitation microscopy — to visualize mouthpart form and internal anatomy at unprecedented nanometer resolution. Interactive 3D graphical visualizations allowed putative homologues of muscles and gland cells to be identified with confidence and traced across the evolutionary transformation from particle feeding to piercing carnivory. Notable changes included: a) addition of new gland cells, b) enlargement of some (venom producing?) glands, c) repositioning of gland openings associated with hollow piercing fangs on the mandibles, d) repurposing of some mandibular-muscle function to include gland-squeezing, and e) addition of new muscles that may aid venom injection exclusively in the most specialized piercing species. In addition, live video recording of all four species revealed mandibular blade movements coupled to cyclic contraction of some muscles connected to the esophagus. These behavioral and 3D morphological observations revealed a novel injection system in H. subspinifrons associated with piercing (envenomating?) carnivory. Conclusions Collectively, these results suggest that subtle changes in mandibular tooth form, and muscle and gland form and location, facilitated the evolution of a novel, piercing mode of feeding that accelerated diversification of the genus Heterorhabdus. They also highlight the value of interactive 3D animations for understanding evolutionary transformations of complex, multicomponent morphological systems. Electronic supplementary material The online version of this article (10.1186/s12983-019-0308-y) contains supplementary material, which is available to authorized users.

tooth of typical particle-feeders (e.g., Disseta and Microdisseta) is unspecialized and similar to that of other particle feeding calanoid copepods (Fig. 1). In intermediate taxa, the ventral tooth is enlarged (e.g., Mesorhabdus and Heterostylites; Fig. 1). In one intermediate taxon the elongate ventral tooth possesses a fine groove (Heterostylites; Fig. 1). The ventral tooth in carnivorous taxa bears a massive groove or is partly tubular in some taxa (Hemirhabdus and Neorhabdus; Fig. 1). In the most diverse and ecologically widely distributed Viper copepods (Paraheterorhabdus and Heterorhabdus) the ventral mandibular tooth forms a completely enclosed tube (Fig. 1).
The effectiveness of piercing carnivory depends not only on mandible form, but also on a) glands that secrete substances to facilitate prey capture and ingestion, and b) muscles that move the mandibular gnathobase. Three sets of gland openings in the upper lip (labrum) are associated with secretory cells in all heterorhabdid copepods [7,9]. Each set is thought to be associated with a different set of glands. However, details of gland structure are known only for the carnivore Heterorhabdus [7], and nothing is known about the muscles that control mandibular motion in any heterorhabdid copepod.
To better understand the fine structure and spatial relations among glands and muscles within the mouthparts of Viper copepods, we utilized two advanced, highresolution imaging methodsserial block face scanning electron microscopy (SBF-SEM) and two-photon excitation microscopyto produce 3D nanometer-scale reconstructions of the external and internal morphology of the labrum (=upper lip) and paragnath (=lower lip) of heterorhabdid species from four genera: Disseta palumbii Giesbrecht, 1889, Mesorhabdus gracilis Sars, 1907, Heterostylites longicornis (Giesbrecht, 1889), and Heterorhabdus subspinifrons Tanaka, 1964. We also video-recorded mouthpart movement in live specimens of all four taxa to clarify muscle function. Finally, to reconstruct the evolutionary history of this innovative feeding mode, we compared putative homologues of component elements (muscles, glands, gland openings) among all four genera and discuss character variation across the phylogenetic tree of heterorhabdid copepods.

Results
Both state-of-the-art SBF-SEM microscopy [14] and two-photon excitation microscopy [15], combined with associated image-analysis technologies, yielded full 3D perspectivesat nano-scale resolutionof the glands and muscles in the mouthparts of the heterorhabdid species studied. Although these two methods are based on different principles, and use different types of fixative, the results were similar for both (Fig. 2). Sections from two different individuals of Mesorhabdus gracilis (Fig. 2) show planes of four pairs of glands (dashed outlines), and planes of one pair of muscles, which correspond nicely between the two pictures. Both imaging methods clearly show the same spatial relationships of glands and muscles. Gland contents, however, appeared to differ somewhat between methods. For example, gland lg1C2 in the SBF-SEM scan (Fig. 2a) appeared to be filled with tiny and flattened disc-shaped granules, whereas in the two-photon excitation microscopy scan (Fig. 2b) the granules appeared to be rather big and more rounded in shape. Curiously, gland contents also appeared to differ between sides even within a single specimen (compare contents of gland lg1C2 on the left and right side of Fig. 2b).
Numerous muscles and glands are associated with the mouthparts examined. All of the descriptive terms used here to refer to morphological units of muscles and glands do not imply any homology hypothesis (see [29] for a discussion of homology-free terminology in morphological description). Homology hypotheses for these descriptive terms are outlined explicitly in Table 1 and discussed in detail in the Discussion.

Gland morphology and arrangement
We adopted labral gland terms from Nishida and Ohtsuka [7], where gland cells were divided into three "Types" according to the arrangement of gland openings ( Fig. 3a-d). We use the same terminology here, but apply these terms differently except for Heterorhabdus subspinifrons. The arrangement of gland openings is essentially the same as reported previously [7], but we found an extra opening of labral gland Type 2 in Disseta palumbii: two openings were reported earlier [7], but we found a third (Fig. 3a).
Even though gland openings were readily identified and easy to homologize among taxa, the size, shape and configuration of gland cells differed considerably among the four genera. In the particle feeding D. palumbii, Fig. 1 Overview of evolutionary relations, feeding modes, species diversity, depth distributions, and mandible and ventral tooth form of Viper copepods (Heterorhabdidae, Calanoida). Phylogeny after Ohtsuka et al. [9] and (Hirabayashi et al. [10]), feeding modes from Ohtsuka et al. [9], species diversity from WoRMS (http://www.marinespecies.org/, July 2017), ocean depth ranges from Ohtsuka et al. ([9]; B-Bathypelagic, M-Mesopelagic, E-Epipelagic), SEM images from Ohtsuka et al. [9], drawings by TK. Asterisk: inferred origin of poison-injection system (Hirabayashi et al. [10]), Dashed circle: ventral tooth gland cells are located postero-ventrally in the labrum, and are not associated with muscles (Fig. 3a, e: see Additional file 1: Figure S1 for viewing instructions for the interactive 3D-pdf images). In Mesorhabdus gracilis (intermediate feeding mode), the labrum is almost fully packed with labral gland cells and parts of these cells intercalate between the muscles lab-eso.dM3 and u-l.labM2 ( Fig. 3b and f). In Heterostylites longicornis (intermediate feeding mode), labral gland cells are located at the posterior half of the labrum, and half of the cells are stacked between muscles lab-eso.dM3 and u-l.labM2 ( Fig. 3g and c). Significantly, in the piercing carnivore, H. subspinifrons, all of the labral gland cells are highly extended anteriorly: a) Type 3 gland cells are enveloped by three muscles u-l.labM2, lab-eso.dM1-4 and for-eso.dM ( Fig. 3h and d: click on the view "Labral Gland Type 3 and muscles" in the interactive 3D-PDF, Fig. 4d), b) Type 2 gland cells extend up to the posterior margin of the paragnath (Fig. 4d), and c) Type 1 gland cells are inflated, and posteriorly elongated into the paragnath (Fig. 4d). The total number of cells in gland Types 1-3 also differed among these genera (Table 1). Disseta palumbii has 15 pairs of cells, but M. gracilis, H. longicornis, and H. subspinifrons have only 8 pairs (Table  1). Type 1 and Type 3 glands were largest in the piercing carnivore, H. subspinifrons (Fig. 4d), but all three types were well-developed in the intermediate feeding-mode M. gracilis (Fig. 4b).
Our observations of cell numbers and orientation in each gland differ somewhat from Nishida and Ohtsuka [7]. They reported "Type 1 and 3 labral glands have two secretory cells...Type 2 labral glands and the paragnathal gland have one secretory cell" in Heterorhabdus abyssalis, H. pacificus, H. papilliger, and H. spinifrons. However, our observation of H. subspinifrons confirmed two cells in Type 1 glands, but revealed three cells in each of Types 2 and 3 (Fig. 4d, Table 1). Regarding cell structures, Type 2 gland cells were previously considered to be anteriorly elongate cells along the labral wall, and Type 3 gland cells as small cells located within the posterior side of the labrum [7]. However, our observations revealed that Type 2 gland cells extend toward the paragnath, and that the dramatically inflated Type 3 gland cells were directed anteriorly, reaching all the way to the forehead.
The arrangement of gland openings also differed between the carnivore Heterorhabdus and the non-carnivore taxa. The openings line up nearly in a straight line in D. palumbii, M. gracilis and H. longicornis, but the opening for Type 1 lies far off the line in H. subspinifrons ( Fig.  3a-d). Significantly, the opening for the Type 1 gland in H. subspinifrons lies directly at the posterior end of the hollow fang (Fig. 4d).
Secretory granules in the gland cells appeared to vary among taxa and among the three gland types (Fig. 5). Granules in homologous types of gland cells (based on location) were not similar in shape and size (e.g., compare "lg3c1"and "lg3c2" in Fig. 5a; "lg1c1" and "lg1c2" in Fig.  5b; "lg1c2" and "lg1c1" in Fig. 5f). However, granule form of homologous gland cells also differed between individuals of the same species (Fig. 2), and even between sides of the same individual (Fig. 2b). Therefore, these observations, combined with inconsistent resolution due to technical limitations of contrasting and resolution, greatly limited the utility of granule form as a tool for making any inferences about gland function or homology.
A small, fourth type of glandtermed here Epidermal Glandwas found by the ventral side of the epidermis, with the duct opening on the ventral side of  Table 1 for gland and muscle names and abbreviations). a Scan from SBF-SEM. b Scan from two-photon excitation microscopy. Note the significant differences in appearance of gland contents between these two individuals, which suggests that high-resolution images of gland contents may not be very informative phylogenetically Table 1 Homology hypotheses for muscles and glands and their formal descriptions, following the scheme adopted by [30]. Each morphological unit in the left-most column is inferred to be homologous among all four species but has spatial relations, connections and constituents as indicated under each species  Table 1 Homology hypotheses for muscles and glands and their formal descriptions, following the scheme adopted by [30]. Each morphological unit in the left-most column is inferred to be homologous among all four species but has spatial relations, connections and constituents as indicated under each species (Continued) both the labrum and paragnath in M. gracilis, H. longicornis, and H. subspinifrons (Fig. 4b, c and d: represented in yellow). No such cells were seen in D. palumbii. Cell numbers were lowest in M. gracilis (2 cells; but paragnath epidermal gland might have been overlooked because of the limited scanning field), greater in H. subspinifrons (7 cells), and highest in H. longicornis (14 cells). In addition, arrangement of the labral epidermal gland cells was erratic and not always symmetrical (e.g., Fig. 4c and d).
Epidermal gland cells in H. subspinifrons contained distinctive spindle-shaped secretory granules ("peg" cells and "leg" cells in Fig. 5g, h). Unfortunately, the contents of these epidermal gland cells were unclear in other genera due to limited contrast and resolution (Fig. 5c, d and e).

Muscle configuration and movement of mouthparts
Given the large differences in mandible form, the overall arrangement and attachment sites of muscles were surprisingly similar among the four genera examined (Fig.  4). These muscles are named based on their attachment sites or locations (Table 1). The only species-specific muscle we observed was in the highly derived carnivore Heterorhabdus subspinifrons, ("saggital labral muscle", Fig.  4d, Table 1). This muscle was located at the posterior side of the labrum: one end attached just beside the opening of labral gland Type 1 and the other end attached near the esophagus opening (Fig. 4d: click on the view "Sagittal Labral Muscle insertions" in the interactive 3D-PDF).

Discussion
Muscle homology among the four genera, and a novel muscle in carnivorous Heterorhabdus subspinifrons Because of the highly conserved arrangement of muscles among the four heterorhabdid genera, putative homologues could be readily identified throughout (Table 1). Therefore, muscle arrangement itself did not appear to play a major role in the evolutionary shift among feeding modes. However, one unique muscle was found in the piercing carnivore Heterorhabdus subspinifrons ("Saggital Labral Muscles (s.labM)", Fig. 4d). Labral gland Type 1 is assumed to be the main gland that charges venomous substances into the hypodermic needle-like mandibular fang [7]. This inference is supported by the position of the Type 1 gland opening, which is located exactly behind the charging pore of the mandibular fang (Fig. 4d). This structural arrangement of muscle and pores therefore suggests that muscle contraction enhances the efficiency of venom charging by adjusting the position of the pores relative to the base of the fang.

Revised homology hypotheses for the glands
Ohtsuka et al. [9] hypothesized that the Type 1 gland in Heterorhabdus subspinifrons is a specialized form of one of the two Type 3 glands that exist in other heterorhabdid species. To be consistent with our homology hypothesis, we revised this terminology, and adopt the term Type 1 for all species examined here (see Figs. 1 and 3). Our hypothesis of gland homology, consistent with traditional criteria for structural homology [28], is based on the unambiguous spatial relationships of gland openings: Type 2 are located at the lateral tip of the labrum and contain 2-3 openings; Type 1 are located beside the Type 2 opening complex; and Type 3 are located at the most medial part of the labrum compared to the other openings. Although we did observe differences among species in the contents of putatively homologous glands (Fig. 5), gland contents also differed significantly between individuals of the same species and sides of the same individual (Fig. 2). Therefore, apparent differences in gland contents among species in Fig. 5 are unlikely to be informative phyologenetically and do not impair our homology inferences.
As noted in the results, we did find a few differences in cell numbers and structures between our results and previous observations [7]. These differences might be due to the different species we used, but they are more likely due to the greatly enhanced spatial resolution of our 3-dimensional analysis.
Nonetheless, these differences do not affect our homology inferences.
Finally, the previously reported "paragnathal gland" (presumed to be associated with a "pore in the ventro-medial region of the posterior surface of each paragnath"; visible in Fig. 3a of Nishida and Ohtsuka [7]) we now consider to be a synonym of the paragnathal epidermal gland defined here.

Di s s e t a p a l u mb i i Me s o r h a b d u s g r a c i l i s He t e r o s t y l i t e s l o n g i c o r n i s He t e r o r h a b d u s s u b s p i n i f r o n s
despite their immense diversity and success in marine and aquatic systems, venom-assisted feeding appears to be exceedingly rare in Crustacea. The only definitive case occurs in remipedes [12]. Envenomation is suspected in two parasitic crustacean groups (branchiuran fish lice and siphonostomatoid copepods) as well as two free-living taxa (caprellid amphipods and heterorhabdid copepods) [12], but pharmacological confirmation is still required. In the piercing carnivorous copepod, Heterorhabdus subspinifrons, cells of the Type 3 labral gland are tightly enveloped by multiple muscles (for-eso.dM, lab-eso.dM 1-4 and u-l.labM2; Figs. 3h, 4d). This arrangement implies that muscular contraction squeezes the gland to eject secretions. Ejection of secretions via muscular contraction appears to be a common strategy in many animals: venom release in cone snails [19]; venom gland discharge in elapid and viperid snakes [20]; silk ejection from the antennal exopod in ostracods [21]; and venom ejection from the head of specialized soldier termites [22]. Furthermore, these muscles in H. subspinifrons all contract cyclically during experimentally induced mastication (Fig. 6d, Additional file 2: Movie SM1D), so presumed venom ejection likely occurs simultaneously with mastication.
Three of our observations suggest that the piercing carnivore Heterorhabdus subspinifrons injects a poison or venom into its prey via the hollow fang on its mandible. First, one gland type (Type 1) is greatly enlarged. Second, the opening to that enlarged gland shifted to lie at the base of the hollow fang (Fig. 3d). Third, secretions from the Type 3 gland in H. subspinifrons may have a specific role in carnivorous feeding, because the enveloping arrangement of muscles around the glands does not occur in the other heterorhabdid species examined (Fig.  4). However, pharmacological evidence would be required to show definitively that the secretions are a venom and not mucus or some other salivary secretion.
Minor morphological change supported a major radiation of feeding strategies Despite drastic functional changesfrom particle feeding with a mundane mandible to carnivorous feeding with a Fig. 4 Three-dimensional surface models of whole muscles and glands in the labrum and paragnath of all four heterorhabdid species: a) Disseta palumbii, b) Mesorhabdus gracilis, c) Heterostylites longicornis, D) Heterorhabdus subspinifrons. The PDF version of the paper contains interactive 3D content that can be activated by clicking on each figure panel in Adobe Reader. To view/exclude individual drawing elements: 1) click on a figure panel to activate it, 2) click on the "Toggle Model Tree" icon in the 3D tool bar to display viewing options, and 3) check/uncheck drawing elements to include/exclude specific elements. In any view, use the scroll function to zoom in/out and click/drag the cursor to rotate the view. To observe the specific views referred to in the text, select the named view from list of views in the "Model Tree" side bar (for a detailed explanation of interactive 3D viewing functions, see Additional file 1: Figure S1). Color codes as in Fig. 2, except for yellow-labral gland and paragnathal epidermal gland, and tan-esophagous. Note: the orientation of the X-Y-Z axis indicators are arbitrary for each panel and are not comparable among panels sophisticated piercing-injection systemthe overall morphological units and their arrangement are strikingly similar among the Viper copepod species examined here. This similarity implies that differential use of the mandible, for simple mastication or for venom injection, can be accomplished by a slight modification of cuticle structure and minor modification of muscle structure. Other examples of great functional innovation in pancrustaceans follow a similar principle, where minor morphological modifications facilitate significant functional change. First, in some highly derived snapping-shrimp genera, like Alpheus and Synalpheus, minor changes in muscle structure (e.g., subdivided claw-closer muscle) maximize the efficiency of the latch-releasing motion before snapping [23]. Second, a similar evolutionary sequence of muscle subdivision to control latch release is seen in Anochetus trap-jaw ants [24]. Third, muscles in the suction disc of adult parasitic branchiuran Crustacea are identical to those in the larva that control ordinary appendage-like motion (the larval mouthpart appendage is the anlagen of the suction disc), except for two newly acquired muscles -"circular sucker muscle" and "disc rim muscle"that both minutely adjust the shape of the sucker to attach it to the host surface in the most efficient way [25]. This evolutionary tendencyfor seemingly minor but functionally significant adjustments of form to a novel functionis comparable to acquisition of the novel "sagittal labral muscle" in Heterorhabdus subspinifrons. Because that muscle, which adjusts the gland opening to the pore of the mandibular fang, is the only newly acquired muscle associated with piercing carnivory, it may enhance efficiency of charging the fang with liquid. Since this efficiency may have critical role in the newly acquired "fang" function of the mandible, this small muscle may play an important role in the new carnivorous feeding strategy. The evolutionary shift of the Type 1 gland opening to lie at the proximal end of the mandibular fang in H. subspinifrons (unlike other taxa), also likely enhances the efficiency of injection.
Such drastic changes of this function-adaptation complex, enhanced by minor morphological change, may facilitate invasion of wholly new adaptive zones and potentially explosive diversification in harmony with body mituarization [26]. The remarkably high diversity of heterorhabdid copepods that utilize piercing carnivory ( [10], Fig. 1) implies that functional transformation of feeding structures may have greatly accelerated the rate of evolutionary diversification.
Three-dimensional visualization of small animals and the "renaissance of morphology" This study also illustrates the great power of new imaging tools, and sophisticated 3D visualization techniques, to help understand complex morphologies, particularly in the small creatures that make up the vast majority of animal diversity. These advances have led to a "renaissance of morphology" [16,17].
The difficulties of 3D imaging in small animals were overcome by using two advanced imaging techniques: serial block-face scanning electron microscopy (SBF-SEM) and two-photon excitation microscopy. SBF-SEM uses a robotic ultramicrotome-embedded within a scanning electron microscope. It is a major advance over confocal laser scanning microscopy (CLSM: appropriate specimen thickness roughly 10-150 μm) and micro-computed tomography (micro-CT: appropriate specimen size roughly 1 mm-20 cm) because it permits 3D reconstruction of meso-scale structures (roughly 100-1000 μm) at nanometer resolution [14]. Two-photon excitation microscopy also yields nanometer resolution of meso-scale structures up to one millimeter depth-of-field [15].
The 3D information contained in the high-resolution image stacks were made comprehensible and presentable by advanced 3D visualization techniques. First, each discrete morphological element (specific muscle or gland) can be segmented out of each plane of an image stack (e.g., see outlined regions in Fig. 5) so that it can be rendered in three dimensions and assigned an informative color and shading (e.g., Fig. 3e-h). But such 3D renderings can still be difficult to interpret from 2D perspective images where many component elements are involved (e.g., Fig. 3e-h). The limitations imposed by 2D  Table 1 for muscle names and abbreviations representations of 3D renderings are overcome entirely by interactive 3D models that can be incorporated directly in pdf files (e.g., Fig. 4). These 3D interactive models give the viewer extraordinary viewing power: 1) virtually unlimited zoom and pan capability, 2) the ability to look at one subset of structures at a time (e.g., only muscles or only glands), and specific items in each subset, via logically structured hierarchical groupings of elements (see Additional file 1: Figure S1 for viewing tips), 3) the ability to examine specific pairs or specific sets of structures in isolation (e.g., the relations of specific muscles (for-eso.dM, lab-eso.dM1-4 and u-l.labM2)) associated with the Type 3 labral gland in Heterorhabdus subspinifrons) by excluding all other structures. Such selective viewing was vital to understanding the 3D spatial relations of component parts in the piercing carnivore H. subspinifrons. It also allows readers to judge for themselves these relations free from any author prejudice. In preparation for observations by SBF-SEM, individuals were fixed with 2% glutaraldehyde and 2% paraformaldehyde in 0.15 M cacodylate sodium buffer with 2 mM CaCl 2 (pH 7.4) for 5 h at 4°C, then decalcified in 10% EDTA in water for 2 days at 4°C. The specimens were post-fixed with 2% osmium tetroxide and 1.5% potassium ferrocyanide in the same buffer for 2 h at room temperature. They were incubated in 1% thiocarbohydrazide for 30 min at room temperature, and fixed again with 2% osmium tetroxide in water for 1 h at room temperature. En bloc staining was performed with 1% uranyl acetate for 3 h at room temperature and then with Walton's lead-aspartate solution (20 mM, pH 5.5) for 60 min at 60°C. The specimens were washed with cacodylate buffer or distilled water between each step described above. Each specimen was a) dehydrated by a graded ethanol series (30-100%) at 4°C with 30 min for each step, b) transferred to 100% acetone for 1 h, and c) incubated in a graded Durcupan resin series (25, 50, 75, 100% using acetone as a solvent) in a vacuum chamber for 12 h at each step. The resin was allowed to polymerize at 60°C for 3 days. Trimmed resin blocks were glued onto an aluminum SBF-SEM rivet with conductive epoxy resin (SPI Conductive Silver Epoxy; SPI Supplies and Structure Prove, Inc., West Chester, PA, USA), and coated with gold using an ion coater. Scanning electron microscopes (SIGMA/VP and MERLIN, Carl Zeiss Microscopy, Jena, Germany), equipped with an in-chamber ultramicrotome system and a back-scattered electron detector (3View; Gatan Inc., Pleasanton, CA, USA), were used to slice and image each specimen as described previously [27]. The serial-section image stack was acquired in an automated fashion by using Gatan Digital Micrograph software.

Collection and imaging
In preparation for observations by the multiphoton microscope (Leica TCS SP8 MP), specimens were fixed in Bouin's solution, dehydrated in an isopropanol series, and then mounted on slides using a 2:1 mixture of benzyl benzoate and benzyl alcohol for clearing. Specimens were imaged using autofluorescence, so excitation wavelength, detected emission wavelength range, etc., were adjusted individually for each specimen to obtain maximum brightness and contrast.

3D visualization and videography
Image stacks from SBF-SEM were automatically aligned using the registration plug-in "Register Virtual Stack Slices" in Fiji/ImageJ software package (http://fiji.sc/Fiji). Surface and volume renderings of the scanned data were performed using IMARIS 7.0.0 (Bitplane AG). Objects for the 3D-pdfs were exported as vrml format. File sizes were reduced by MeshLab (http://www.meshlab.net/), and then exported as u3d format. Files were arranged using Deep Exploration (Right Hemisphere) and re-arranged by Adobe Acrobat Pro (Adobe) to create 3D-pdf files.
To video mouthpart motion, living copepods were briefly semi-dried and attached to a glass dish with cyanoacrylate glue on the dorsal side of the metasoma and the dish was then filled with seawater. Positioned copepods were stimulated to move their mouthparts using a needle. A video camera EX-F1 (CASIO, Japan) was used to record mouthpart behavior.

Additional files
Additional file 1: Figure S1. Instructions for how to use the viewing functions of the interactive 3D-pdf in Fig. 4. 1) Click any panel you want to view. 2) Click model tree icon (A) to reveal operation windows (B) and (C). 3) Window B shows the heirarchical tree diagram of morphological characters defined in this paper. Click the arrowhead to the left of each to reveal subcategories, and click each checkbox to hide/unhide each specific character. 4) Click "view"s (C) to view the specified perspective of the characters selected in (B) as described and instructed in the main text. (JPG 877 kb)