The Musculotendinous System of an Anguilliform Swimmer: Muscles, Myosepta, Dermis, and Their Interconnections in Anguilla Rostrata

JOURNAL OF MORPHOLOGY 269:29–44 (2008)

The Musculotendinous System of an Anguilliform Swimmer: Muscles, Myosepta, Dermis, and Their Interconnections in Anguilla rostrata

Nicole Danos,1* Nina Fisch,2 and Sven Gemballa2

1Department of Organismic and Evolutionary Biology, Harvard University, Cambridge, Massachusetts 02138 2Department of Zoology, University of Tu¨ bingen, Auf der Morgenstelle 28, Tu¨ bingen D-72076, Germany

ABSTRACT Eel locomotion is considered typical of the efficient force transmission. Though the main modes of anguilliform swimming mode of elongate fishes and has undulatory locomotion (anguilliform, subcarangiform, received substantial attention from various perspectives and carangiform) have recently been shown to be very such as swimming kinematics, hydrodynamics, muscle similar with respect to their midline kinematics, we are physiology, and computational modeling. In contrast to able to distinguish two morphological classes with the extensive knowledge of swimming mechanics, there respect to the shape and tendon architecture of myo- is limited knowledge of the internal body morphology, septa. Eels are similar to subcarangiform swimmers including the body components that contribute to this (e.g., trout) but are substantially different from carangi- function. In this study, we conduct a morphological anal- form swimmers (e.g., mackerel). This information, in ysis of the collagenous connective tissue system, i.e., the addition to data from kinematic and hydrodynamic stud- myosepta and skin, and of the red muscle fibers that ies of swimming, shows that features other than midline sustain steady swimming, focusing on the interconnec- kinematics (e.g., wake patterns, muscle activation pat- tions between these systems, such as the muscle-tendon terns, and morphology) might be better for describing and myosepta-skin connections. Our aim is twofold: (1) the different swimming modes of fishes. J. Morphol. to identify the morphological features that distinguish 269:29–44, 2008. Ó 2007 Wiley-Liss, Inc. this anguilliform swimmer from subcarangiform and carangiform swimmers, and (2) to reveal possible pathways KEY WORDS: myoseptal tendons; red muscle; skin; of muscular force transmission by the connective tissue connective tissue; Anguilla rostrata in eels. To detect gradual morphological changes along the trunk we investigated anterior (0.4L), midbody (0.6L), and posterior body positions (0.75L) using microdissections, histology, and three-dimensional reconstruc- All swimming vertebrates share a similar axial tions. We find that eel myosepta have a mediolaterally musculoskeletal morphology, consisting of a com- oriented tendon in each the epaxial and hypaxial regions pression–resistant vertebral column or notochord (epineural or epipleural tendon) and two longitudinally surrounded by axial musculature. The skeletal oriented tendons (myorhabdoid and lateral). The latter and muscular components interact through a com- two are relatively short (4.5–5% of body length) and plex three-dimensional system of connective tissue. remain uniform along a rostrocaudal gradient. The skin In fishes the muscular system is organized as a and its connections were additionally analyzed using series of three-dimensionally folded segments, the scanning electron microscopy (SEM). The stratum com- myomeres. Adjacent myomeres are separated by pactum of the dermis consists of 30 layers of highly ordered collagen fibers of alternating caudodorsal and sheets of connective tissue, the myosepta, into caudoventral direction, with fiber angles of 60.51 6 which the muscle fibers insert (Alexander, 1969; 7.058 (n 5 30) and 57.58 6 6.928 (n 5 30), respectively. Gemballa and Vogel, 2002). Medially, the series of Myosepta insert into the collagenous dermis via fiber myomeres and myosepta inserts on the collagenous bundles that pass through the loose connective tissue of vertical septum and the bony axial skeleton. Later- the stratum spongiosum of the dermis and either weave ally, the whole system is wrapped by the skin. In into the layers of the stratum compactum (weaving fiber all basal groups of notochordates, including all fish bundles) or traverse the stratum compactum (transverse fiber bundles). These fiber bundles are evenly distributed along the insertion line of the myoseptum. Red Contract grant sponsors: Society for Experimental Biology, Journal muscles insert into lateral and myorhabdoid myoseptal of Experimental Biology, Sigma-Xi, Wilhelm-Schuler-Stiftung and Strukturfond Baden-Wu¨rttemberg at the University of Tu¨bingen. tendons but not into the horizontal septum or dermis. Thus, red muscle forces might be distributed along these *Correspondence to: Nicole Danos, Department of Organismic tendons but will only be delivered indirectly into the and Evolutionary Biology, Harvard University, 26 Oxford Street, dermis and horizontal septum. The myosepta-dermis Cambridge, MA 02138. E-mail: [email protected] connections, however, appear to be too slack for efficient force transmission and collagenous connections between Published online 21 September 2007 in the myosepta and the horizontal septum are at obtuse Wiley InterScience (www.interscience.wiley.com) angles, a morphology that appears inadequate for DOI: 10.1002/jmor.10570

Ó 2007 WILEY-LISS, INC. 30 N. DANOS ET AL. taxa, the skin includes a multilayer system of col- provided input to dynamic models that describe lagenous connective tissue, the stratum compac- the swimming movements of subcarangiform or tum (Gemballa and Bartsch, 2002). carangiform swimmers (Long et al., 2002). These A coordinated and sequential activation of the examples show that the study of fish swimming myomeric musculature in most fishes generates can benefit significantly from further morphologi- axial undulatory swimming characterized by cal information. waves of propulsive body undulations that travel One of the most characteristic ways of perform- down the body towards the caudal fin while gradu- ing undulatory swimming, anguilliform swimming, ally increasing in amplitude. However, we still is displayed by many elongate or eel-like fishes. have little understanding of how muscular activity The anguilliform mode was described as a distinct is translated into undulatory swimming waves in swimming mode characterized by body waves with most fishes. This lack of understanding is mainly relatively short wavelengths and large amplitudes due to the complexity of the components that (Lindsey, 1978). Most of the work on anguilliform form the swimming apparatus. Muscle fibers are swimming was carried out on the genus Anguilla oriented in complex three-dimensional patterns and has not only covered kinematics but also the within myomeres (Alexander, 1969; Gemballa and hydrodynamics, muscle dynamics, whole body Vogel, 2002) and myosepta consist of a three- mechanics, and modeling (Gray, 1933a,b,c, 1968; dimensional network of tendinous structures Gillis, 1996, 1998a,b; Long, 1998; D’Aout et al., (Gemballa et al., 2003a). In addition, the body 2001; Mu¨ ller et al., 2001; Tytell, 2004a,b; Tytell envelope with its highly ordered layers of collagen and Lauder, 2004; Kern and Koumoutsakos, 2006). fibers in the stratum compactum has been posited However, recent work suggests that the dorsal to affect whole body mechanics by acting as a ten- midline swimming profiles, from which kinematic don transmitting forces produced by the axial variables such as body wave wavelength and am- musculature to the tail, by stiffening the body plitude are calculated, do not differ substantially when under tension, or by storing elastic energy between anguilliform and subcarangiform or even during axial undulations (Wainwright et al., 1978; carangiform swimmers (Donley and Dickson, 2000; Hebrank, 1980; Hebrank and Hebrank, 1986; Lauder and Tytell, 2006). In contrast to the large Alexander, 1987; Long et al., 1996; Brainerd and body of knowledge in these experimental fields, we Patek, 1998). Similar hypotheses have been pro- know surprisingly little about the underlying mor- posed for the other connective tissues, such as the phology of elongate fishes. If the kinematic profiles horizontal septum and myoseptal tendons (e.g., of fishes so different in external morphology are so Westneat and Wainwright, 2001; Gemballa et al., similar, how does their internal locomotory anat- 2006). omy compare? Because of the complexity of the individual com- In this study, we address several aspects of the ponents, attempts to understand the function of morphology of the swimming apparatus of the the whole system have mostly dealt with just one American Eel, Anguilla rostrata. Our morphologi- component at a time, focusing on the myosepta, cal analysis includes the red muscle fibers, the vertebral column, segmented musculature, or part of the muscular system that powers steady skin. However, any functional analysis or model swimming, and the collagenous connective tissue should consider information on the morphology system, i.e., myosepta and skin, that transmits of as many components of the swimming appara- muscular forces to the backbone. In particular we tus as possible, including the interconnections bet- pay attention to the interconnections between ween each component (e.g., muscle-tendon associa- these systems, such as the muscle-tendon and tions or connections between myosepta and skin). myosepta-skin connections. Our goal is twofold. For example, it is critical to know the associations First, we aim to identify the morphological fea- of muscle fibers and myoseptal tendons if one is to tures that distinguish this anguilliform swimmer analyze the transmission of muscular forces. A from subcarangiform and carangiform swimmers. recent study that has linked muscle and tendon Such findings would facilitate the generation of morphology to swimming biomechanics by concur- testable hypotheses on the locomotory function of rently analyzing swimming kinematics, muscle dy- particular structures. Second, we aim to reveal namics, and myoseptal anatomy has led to an possible pathways of muscular force transmission improved understanding of how muscle activity is by the connective tissue in eels. Both skin and convergently translated into the uniform swim- myosepta have been implicated in force transmis- ming in lamnid sharks and tunas (Donley et al., sion in some fishes including eels (Wainwright 2004, 2005; Gemballa et al., 2006). Other studies et al., 1978; Hebrank, 1980; Wainwright, 1983; have provided further morphological information Hebrank and Hebrank, 1986; Westneat and Wain- of the geometry of myoseptal tendons or of mus- wright, 2001). By describing the insertions of red cle-tendon associations (Gemballa and Vogel, muscle fibers to collagen fiber tracts within myo- 2002; Gemballa and Treiber, 2003; review: Shad- septa and skin, and the interconnections between wick and Gemballa, 2006). Such information has myoseptal tendons and the dermis, we aim to

Journal of Morphology DOI 10.1002/jmor CONNECTIVE TISSUE SYSTEM IN ANGUILLA ROSTRATA 31

TABLE 1. List of Anguilla rostrata specimens used in this study

Body region Specimen no. [TL] Preparation examined Investigation

1 [290 mm] Skinned, cleared and double stained 0.2–0.9L 3D-shape of myosepta, collagenous architecture 2 [274 mm] for cartilage (Alcian Blue 8GX) of myosepta, measurement of overall and bone (Alizarin Red) myoseptal length 3 [295 mm] Horizontal sectionsa of: Computer based 3D reconstruction of myosepta Anterior body 0.42–0.48L from histological sections, measurement of Midbody (right side) 0.56–0.63L overall myoseptal length and muscle fiber Posterior body 0.71–0.78L angles, insertion of RM into myoseptal Sagittal sectionsa (left side) 0.56–0.63L tendons 3 [295 mm] Paraserial sectionsa 0.52–0.54L Interconnections between skin and myosepta, Interserial sectionsa 0.52–0.54L insertion of RM into myoseptal tendons 4 [300 mm] Formalin fixed 0.55–0.58L 3D l-CT scan for bony elements 5 [250 mm] Unskinned, cleared and stained approx. 0.42–0.52L Organization of skin, SEM interconnections for bone (Alizarin Red) between skin and myosepta 6 [370 mm] Formalin fixed 0.18–0.25L Measurements of collagen fiber angles in skin 0.48–0.56L 0.81–0.87L 7 [165 mm] Formalin fixed 0.17–0.24L Measurements of collagen fiber 0.54–061L angles in skin 0.89–0.93L 8 [210 mm] Formalin fixed 0.19–0.25L Measurements of collagen 0.49–0.53L fiber angles in skin 0.85–0.89L

Body region is given relative to tip of snout (0.0L) and tip of caudal fin (1.0L). See text for further details. aFor histology all body specimens were embedded in paraffin following standard histological protocols. Section thickness was 10 lm in all cases. All sagittal and horizontal sections were stained according to the Azan-Domagk protocol (Romeis, 1986). In addition, some of the interserial sections were stained according to the Fra¨nkel protocol (Romeis, 1986). RM, red muscle; WM, white muscle. reevaluate the probability of these functional fiber tracts. A few myosepta were retained in their native axial hypotheses. position. Their 3D-shape including attachment to the vertebral axis was drawn using a camera lucida. In addition to microdissections, we used digital 3D reconstructions (Amira v. 3.1) of three myosepta from horizontal or MATERIALS AND METHODS sagittal histological sections (anterior body position at 0.45L; Material Examined midbody position at 0.60L; posterior body position at 0.75L;Ta- ble 1; Specimen 3). After manual alignment of a series of sec- The results of this study are based on a collection of techni- tions with the alignment tools of this software we built a 3D- ques applied to eight specimens of Anguilla rostrata Le Sueur, model of a complete myoseptum. We are confident that digital 1817. Table 1 gives a list of specimens and summarizes the reconstructions from histological sections were as reliable as investigations applied to each. our traditional microdissection technique outlined above since both provided comparable results from different specimens. In addition to the shape of myosepta we recorded the overall Investigation of Myoseptal Shape and myoseptal length. This was defined as the distance between the Architecture tip of the anteriorly-pointing and the posteriorly-pointing myoseptal cones (Shadwick and Gemballa, 2006). In the cleared and The techniques for studying myosepta and their architecture stained specimens (Table 1; Specimens 1, 2) this length was follow the procedures described elsewhere (e.g., Gemballa and measured to the closest tenth of a millimeter using a caliper Hagen, 2004) and are only briefly described here. The only tech- gauge. In the computer generated 3D-model (Table 1; Specimen nical addition made here is the use of a computer-based 3D- 3), we measured the overall myoseptal length by relating the reconstruction as an alternative source of three-dimensional scale bar of the original microscopic image to the Amira meas- data for myoseptal architecture and length measurements of uring tool. myoseptal tendons. Collagen fiber tracts of excised myosepta from Specimens 1 Data on the 3D-shape of myosepta were obtained from and 2 were visualized and photographed by spreading them out cleared and double stained (for bone and cartilage according to on glass plates under polarized light (Zeiss Polarizer S and ana- Dingerkus and Uhler, 1977) specimens of Anguilla rostrata that lyzer A53 adapted to stereomicroscope; see Gemballa and had been skinned prior to clearing (Table 1; Specimens 1, 2). Af- Hagen, 2004). In addition, collagen fiber tracts were recorded ter completion of the staining procedure, the specimens were under incident light directly from the myosepta that were transferred stepwise into pure ethanol in order to better visual- retained in the cleared and stained specimens. These tracts ize the connective tissue. For the following microdissections we were included in the camera lucida drawings of myosepta (see used fine iris spring scissors (FST Vannas Mini) and removed above). each myoseptum close to its insertion line along the vertebral To visualize the spatial orientation of postcranial bony ele- axis and along the horizontal and vertical septa, to examine the ments, particularly the ossified myoseptal tendons (i.e., inter- collagen tracks and tendons within each myoseptum. We muscular bones) and their relation to vertebrae, we ran high recorded the anteriormost position of the insertion line on the resolution l-3D X-ray CT scans using a RayScan 200 system vertebral column of each myoseptum and defined this position (Wa¨lischmiller GmbH Germany; ARGE Metallguss at Steinbeis as the axial position of the myoseptum (axial position 0.0L is Transferzentrum at Fachhochschule Aalen, Germany). Scans of tip of snout; 1.0L is tip of caudal fin). Excised myosepta were a midbody segment of one specimen (Table 1; Specimen 4) were kept separately in ethanol for further investigation of collagen obtained at a resolution of 25 lm (417 optical sections per verte-

Journal of Morphology DOI 10.1002/jmor 32 N. DANOS ET AL.

both transmitted light and polarized light (Zeiss Axioplan 2 with polarized light device). The polarizer was used to highlight the collagenous interconnections between skin and myosepta. The stratum compactum fiber angles, defined as the angle of the serial fibers to the long axis of the body, were recorded from Specimens 6, 7, and 8 (Table 1) after carefully removing a rectangular piece of skin from two body positions. These positions were at 0.17–0.25L (anterior) and 0.49–0.61L (midbody). Both of these body regions have a nearly circular body cross-section which should allow us to interpret our morphological measurements with respect to models of organisms with circular cross- sections and fiber reinforcement in their skin arranged in two helices of opposite handedness around their long axis (e.g., Alexander, 1987). The rectangular pieces of skin were then photographed under a Leica dissecting microscope using a Nikon coolpix digital camera. Photographs were imported to ImageJ (ImageJ, U. S. National Institutes of Health, Bethesda, MD) and the fiber angles measured at five locations for each direction, caudoventral and caudodorsal, using the remnants of the Fig. 1. SEM specimen preparation. A flap of epaxial skin horizontal septum attachment to the skin as a reference for the was excised from a cleared and stained eel. The skin was pulled long axis of the body. back to reveal the two main dermis layers: the stratum spongiosum (red) with its irregularly oriented collagen fibers, and the stratum compactum (white) with its highly-oriented, crossed Association of Myoseptal Tendons and collagen fibers. Myoseptal attachments (dotted line) can be seen on the medial side of the excised skin. White arrows indicate Red Muscle Fibers the perspective of SEM images. We used the sagittal, paraserial, and interserial sections (Ta- ble 1; Specimen 3) to detect specific associations of red muscle fibers and myoseptal tendons. In these sections myoseptal tendons could easily be detected because they are thicker than any bral segment). 3D images of the scanned region were generated of the remaining parts of myosepta (Gemballa et al., 2003a). using VG Studio Max. The lCT datasets from VG Studio Max Muscle fibers that insert into these tendons are either fast were combined with the myosepta 3D data sets from Amira into white muscle fibers or slow red muscle fibers. Both types of a single file using Cinema 4D software (.wrl- file format). This fibers can be distinguished by their appearance (e.g., Videler, allowed us to combine in a single three-dimensional figure data 1993): red muscle fibers are slender and are oriented longitudi- from two sources, information on connective tissues from histol- nally whereas white muscle fibers are thicker and oriented at ogy and information on skeletal elements from l-CT, and exam- an angle to the long axis of the body. We paid particular atten- ine the spatial relationships of the two types of tissues. The tion to the insertion of red muscle fibers into the tendons in combined model was rotated until a perspective showing all order to reveal putative pathways along which red muscle the relevant detail was obtained. Images of the skeletal and the forces might be delivered during steady swimming. combined model were exported into Adobe Photoshop (.tif file format). Finally, the original unsmoothened myoseptum model was replaced by greytone artwork using Photoshop airbrush RESULTS tools. 3D-Shape and Architecture of Myosepta and Horizontal Septum Investigation of the Skin and its The terminology we use here to describe the dif- Connections to Myosepta ferent parts of a myoseptum follows Gemballa The skin and its connections to the myosepta were examined et al. (2003a). Myosepta are bisected into an epax- histologically (Table 1; Specimen 3) and by microdissection and ial and a hypaxial half at the midhorizontal level Scanning Electron Microscopy (SEM; Table 1, Specimen 5). In by the horizontal septum and exhibit the typical the latter case we used an unskinned cleared and stained speci- W-shape of most other fishes. The epaxial half of men from which we removed a square piece of skin 2.5 by 2.5 cm with some of the underlying connective tissue (myosepta, the myoseptum (Fig. 2A) bears an anteriorly-point- endomysium, and stratum spongiosum) from the left epaxial ing cone, the dorsal anterior cone (DAC), and a side of the body immediately caudal to the anus, where the posteriorly-pointing cone, the dorsal posterior cone cross-section of the body was closest to being circular (see Fig. (DPC). The myoseptal part connecting these two 1). From this section, smaller specimens were excised. Skin specimens underwent critical point drying and spattered with cones is termed the epaxial sloping part (ESP). Gold Palladium in preparation for SEM. The organization The dorsal most part that links the DPC to the of the skin and its interconnections with myosepta were cap- dorsal midline is termed the epaxial flanking part tured by a Cambridge Stereoscan 250 MK-2 scanning electron (EFP). The hypaxial half is almost the mirror microscope. image of the epaxial half with correspondent cones Skin-myoseptal interconnections were also examined histologically (Table 1; Specimen 3). Sections were taken in two oblique (ventral anterior cone VAC, ventral posterior cone transverse planes. One plane is parallel to the craniodorsal ori- VPC) and correspondent myoseptal areas (hypaxial entation of the collagen fibers in the skin, which is also the ori- sloping part HSP, hypaxial flanking part HFP). entation of scale rows; therefore we term these sections parase- Three collagen fiber tracts are visible in each of rial sections (see Gemballa and Bartsch, 2002). The other plane is parallel to the caudodorsal orientation of the collagen fibers the epaxial and hypaxial parts of the myoseptum in the skin, intersecting the scale rows and is therefore termed (Figs. 2 and 3). One tract of fibers, the myorhab- interserial. Sections were examined under a microscope with doid tendon (MT), spans the entire flanking part of

Journal of Morphology DOI 10.1002/jmor CONNECTIVE TISSUE SYSTEM IN ANGUILLA ROSTRATA 33

Fig. 2. A: Graphic representation of the epaxial part of an eel myoseptum at 0.60L. 3D-reconstruction of skeletal elements is based on a lCT scan, myoseptum was reconstructed from serial histological sections, and its attachment to axial structures was obtained from microdissections of cleared and stained specimens. The myoseptum is shown semitransparent. Gray lines indicate collagen fiber tracts. The lateral attachment line (lal) along which the myoseptum inserts into the skin is indicated as a white line enframed by fine dark lines; the white dotted line marks the insertion of the myoseptum to the vertebral axis. The area dorsal to the DPC is called EFP; this area contains the MT. The area between dorsal anterior cone (DAC) and DPC is called ESP; this area contains the LT and the ENT. Note that the proximal part of the ENT is ossified (ENB). ENT fibers diverge from this ENB. B: Detail of proximal end of ENB. The white circle shows the insertion of the ENB into the neural arch. fr, fin rays; hs, horizontal septum; pt, pterygiophores; v, vertebral column; vs, vertical septum. each myoseptum in an almost longitudinal orienta- tendons (ENT, EPT) are the only myoseptal tendons tion from its anterior tip towards the posterior that insert to the vertebral column. Their insertion cone. However, this tendon is only present in the points, as revealed by microdissections and l-CT lateral region of the flanking parts. The medial scans (Fig. 2B), lie on the neural arch in the case of part is devoid of longitudinal collagen fibers and the ENT and on the hemal arch in the case of the remains thinner (see also section on ‘‘Insertions of epipleural tendon. Immediately lateral to this inser- red muscles to myoseptal tendons’’). tion on the vertebral column, the tendons are ossi- The second and third distinct collagen fiber tracts fied along most of their way towards the skin lie in the sloping parts (ESP and HSP). The lateral (termed epineural and epipleural bones ENB, EPB; tendon (LT) connects the tips of the anterior and the Fig. 2A). posterior myoseptal cones, spanning the whole slop- The insertions of the ENT and EPT lie along the ing part (ESP or HSP) of a myoseptum. As with the medial insertion line of each myoseptum along the MT, this tendon is only present in the lateral region vertebral column. This myoseptal insertion onto of the myoseptum and thick longitudinal collagen the vertebral column spans three vertebral seg- fibers are absent from the medial part of the myo- ments (Fig. 2A). It starts at the anterior margin of septum. The epineural (ENT), or epipleural in the the first vertebra and runs caudally across the hypaxial region (EPT), tendon inserts into the verte- neural arch of this vertebra. After passing this bral column and runs caudolaterally towards the neural arch (including the insertion of the ENT) skin in the sloping parts of the myoseptum. It the insertion line passes two more neural arches intersects the LT at an acute angle in the lateral until it aligns with the neural spine of the third part of the myoseptum. Epineural and epipleural vertebra. Having reached the end of this spine the

Journal of Morphology DOI 10.1002/jmor 34 N. DANOS ET AL. mens examined. Myoseptal shape was similar throughout the trunk and the tendons described, including their ossifications, appeared similar in all body regions. Moreover, the overall length of myosepta (i.e., the rostrocaudal span reaching from the tip of the anterior to the tip of the posterior myoseptal cone), a feature used to compare myosepta along a rostrocaudal gradient, remains constant along the whole myoseptal series. The overall myoseptal length equals the span of the LT parallel to the long axis of the body from the tip of the anterior to the tip of the posterior cones. We obtained overall myoseptal length values between 4.4% and 5.1% of body length (Table 2). The horizontal septum consists of a system of crossing collagenous fibers (Fig. 4A). One set of fibers forms distinct tendon-like structures, the epicentral tendons, which run caudolaterally from their insertion at the vertebral column to the skin. Epaxial and hypaxial myosepta insert onto the horizontal septum along an epicentral tendon. The second direction of fibers is represented by evenly distributed caudomedial fibers that are termed posterior oblique fibers. In the posterior body only, these fibers are more prominent in the medial part of the horizontal septum and might be better termed here posterior oblique tendons (POTs). However, these POTs are only present close to the vertebral column and are absent in the lateral part of the horizontal septum (Fig. 4B).

The Dermal Connective Tissue System and its Connections to the Myosepta Organization of the dermis. Using SEM we were able to visualize the two thickest layers of the dermis, the stratum compactum with its highly ordered collagen fibers and the stratum spongiosum with its randomly oriented fibers (see Fig. 5). The stratum compactum from the lateral midbody section of the specimen photographed (TL 5 25 cm) comprises 30 collagen lamellae, with each lamella made up of only one layer of parallel collagen fibers (Fig. 5A). The histological micrographs Fig. 3. Eel myoseptum of posterior body region (0.75L) show only those lamellae that have parallel fiber spread out under polarized light. Tendinous fiber tracts are visi- orientations to the sectioning orientation and the ble as white strands. The membranous ossification (ENB, see mean number of visible lamellae was 13.58, as Fig. 2) cannot be distinguished from the collagenous fiber tract counted in nine micrographs from the midbody of the ENT. Note that epaxial and hypaxial parts are symmetri- 5 cal around the horizontal septum (gray dashed line). Prefix e- region (s.d. 1.33). This number is approximately stands for epaxial, h- for hypaxial; EPT, epipleural tendon; see half of the number measured in the SEM section Figure 2 for further legends. The medial attachment of the in which lamellae with fibers in both directions myoseptum to the vertebral axis and vertical septum is outlined can be distinguished, suggesting that histological with a white dotted line. Mediolateral from left to right, dorso- preparation artifacts allow only fibers parallel to ventral from top to bottom. the plane of section to be clearly viewed. The mean stratum compactum collagen fiber insertion line takes a turn in a craniodorsal direc- angles, the angle at which the collagen fibers tion at which it remains until its end at the dorsal within a lamella run parallel to the long axis of midline (Fig. 2A). the body, were 57.58 6 6.928 for the caudodorsal The described features of myosepta do not direction (n 5 30) and 60.51 6 7.058 for the caudo- change along a rostrocaudal gradient in the speci- ventral direction (n 5 30). Of the three specimens

Journal of Morphology DOI 10.1002/jmor CONNECTIVE TISSUE SYSTEM IN ANGUILLA ROSTRATA 35

TABLE 2. Rostrocaudal span (given as % of body length) of epaxial and hypaxial lateral tendons at three axial positions (0.4L, 0.6L, and 0.75L) in three specimens of Anguilla rostrata

Specimen 1 Specimen 2 Specimen 3 Axial position [290 mm TL] [274 mm TL] [295 mm TL] Mean

0.40L (epaxial) 5.05% 4.88% 4.66% (at 0.45L) 4.86% 6 0.16a 0.40L (hypaxial) 3.42% 3.49% 4.05% (at 0.45L) 0.60L (epaxial) 4.98% 4.81% 4.85% 4.85% 6 0.16 0.60L (hypaxial) 4.65% 5.09% 4.71% 0.75L (epaxial) 4.41% 4.47% 4.58% 4.54% 6 0.09 0.75L (hypaxial) 4.51% 4.64% 4.62%

aHypaxial myosepta from the anterior body generally differ in shape (e.g., less pronounced cones and tendons; e.g., Gemballa and Ro¨der, 2004) from myosepta of other body regions due to the presence of the outbulging peritoneal cavity. Hence, only epaxial myosepta were used for the calculation of the mean value at 0.40L. measured the angles at anterior and midbody dif- to the dorsal anterior-pointing cone. The myosep- fered neither between body regions nor in the cau- tum was cut close to its attachment with the doventral and caudodorsal directions (Student’s t skin,hencethefreefiberendsinFigure5B,C. test, a50.05). In contrast, the collagen fibers in The stratum spongiosum was peeled back from the stratum spongiosum are arranged irregularly one myoseptum toward the other to expose its but densely, forming a collagenous sheath (Fig. attachments to the stratum compactum. The 5B,C). Observations from gross dissections of 10 myosepta connect to the stratum compactum via individuals suggest that the stratum spongiosum discrete fiber bundles that run all the way from varies in thickness between individuals and at dif- the myoseptum through the stratum spongiosum ferent body locations. to the stratum compactum. Both SEM and histo- Myosepta-skin interconnections. For the logical investigation confirmed that two kinds of sample depicted in Figure 5B, the stratum spon- connections are present: transverse fiber bundles giosum was peeled away from half of the area (tfb; Fig. 5D,F,G) or weaving fiber bundles (wfb; between two adjacent myosepta, cross-sectioned Fig. 5E,F). The transverse fiber bundles cut parallel to the direction of the myoseptum and transversely through all the stratum compactum lifted up. This exposed the attachments of the layers and insert in the basal lamella, the layer myoseptum to the stratum compactum. We verified between the dermis and the epidermis. Trans- that the observed attachments were of the myosep- verse fiber bundles are more often observed tum and not of the stratum spongiosum, by pulling directly lateral from the myoseptum (Fig. 6B,C). on myoseptal fibers and noticing that the attach- The weaving fiber bundles on the other hand ment fibers following the same trajectory came branch anteriorly and posteriorly from the myo- under tension. septal-stratum spongiosum interface and weave Figures 5C–G depict a section cut along the into stratum compactum collagen fibers at multi- base of two myosepta, from the horizontal septum ple lamellae, beginning with the medial-most

Fig. 4. Polarized light micrographs of the horizontal septum. The crossing fiber directions of the epicentral tendon (ECT) and the posterior oblique fibers/tendons (POF/POT) are indicated by the arrows. A: Midbody region, top view of left body side. Medial insertion to vertebral column on top, cranial to the left. B: Posterior body region, top view of left body side. Medial insertion to vertebral column on top, cranial to the left. The horizontal septum is cut along the anteriormost epicentral tendon (white dotted line). The collagenous fibers to the left of this dotted line belong to the myoseptum (MS) that inserts into the horizontal septum. Note that fibers of two directions (asterisk and double-asterisk) run from the myoseptum into the horizontal septum.

Journal of Morphology DOI 10.1002/jmor 36 N. DANOS ET AL.

Fig. 5. SEM images of Anguilla rostrata skin. A: Cross-section of the stratum compactum, illustrating the plywood arrangement of alternating collagen fiber layers at approximately 908 to each other. Section plane is parallel to one fiber orientation and perpendicular to the other. B: Mediolateral view of skin, showing the attachment of an epaxial myoseptum (ms) to the skin. The stratum spongiosum (ss), which separates the muscle fibers from the skin has been pulled back, cross sectioned and lifted up to expose the connections of the myoseptum with the stratum compactum. The irregular arrangement of collagen fibers in the stratum spongiosum can be seen in contrast to the regular arrangement in alternating fiber layers on the medial surface of the stratum compactum (sc). C: The stratum spongiosum (ss) is partially removed to expose its connections to the stratum compactum. D: Magnification of the white square in (C). The two kinds of connections attaching the stratum spongiosum and the myoseptum to the stratum compactum, as seen under a lifted subdermal layer. E: Magnification of the white square labeled ‘‘E’’ in (D). The weaving fiber bundle extends from the stratum spongiosum and joins the weave of the stratum compactum. Fibers from the bundle run in both stratum compactum directions and dive into more lateral layers of the stratum compactum. F: Magnification of the white square labeled ‘‘F’’ in (D). The transverse fiber bundle originates more medially in the stratum spongiosum. No apparent fibers from this bundle follow the orientation of either stratum compactum layer but instead cross the stratum compactum perpendicularly. G: Ultrastruc- ture of a transverse fiber bundle. Collagen fibers dive into the stratum compactum. Each fiber consists of smaller fibrils. Fat glob- ules can be distinguished on the stratum compactum. Ms, myoseptum; sc, stratum compactum; ss, stratum spongiosum; tfb, transverse fiber bundle; wfb, weaving fiber bundle.

Journal of Morphology DOI 10.1002/jmor CONNECTIVE TISSUE SYSTEM IN ANGUILLA ROSTRATA 37 lamella. These branches from the stratum spongiosum to the stratum compactum are never observed to cross the branches from an adjacent myoseptum, suggesting that there are regions of the myomere that are not directly attached to the skin (Fig. 6A). A third type of connection is obvious from some high-resolution histological micrographs (Fig. 6B, yellow arrow). This type connects two adjacent stratum compactum layers and the distribution of this type of connections is staggered and evenly disbursed throughout the stratum compactum.

The Musculotendinous System: Insertions of Red Muscles to Myoseptal Tendons Red muscles in eel form a superficial layer that extends far epaxially and hypaxially (Fig. 7A) and only extends medially towards the vertebral column along the horizontal septum (‘‘deeper’’ red muscle). In these regions it can be as deep as one third to one half of the distance from the skin to the vertebral column. Red muscle not only covers the two inner legs of the ‘‘W’’ shape of myosepta (i.e., the epaxial and HSP) but also part of the dorsal- and ventral-most flanking parts. For most parts of its dorsoventral extension the red muscle cover remains a thin layer. The thickness of this layer is 55.8 6 4.7 lm(n 5 13) in the flanking parts and 89.7 6 13.6 lm(n 5 17) in the sloping parts. We describe the spatial associations of red muscle fibers, myosepta, and the horizontal septum from the histological sections (horizontal, sagittal, para- and interserial sections; see Table 1). Red muscle fibers insert into the lateral-most parts of epaxial and hypaxial myosepta in the sloping and flanking parts; they do not insert into the horizo- nal septum. However, the ‘‘deeper’’ red muscle fibers that run along the horizontal septum insert into the myoseptum which itself attaches to the horizontal septum. Therefore, in order to best describe red muscle insertions into myosepta and the red muscle-tendon associations we need to address two aspects: insertions of red muscle fibers Fig. 6. Micrographs of histological specimens. A: Myoseptum to myoseptal tendons of the lateral-most flanking at the ESP level inserting into the stratum compactum. The stratum compactum is comparable in thickness to the epider- and sloping parts, and insertions of red muscle mis. Two thick fiber bundles, stained blue, extend from the fibers to those myoseptal parts that are close to myoseptum through the stratum spongiosum. One fiber bundle the horizontal septum. inserts into the stratum compactum in this section via a weav- The lateral-most area of the flanking parts is ing fiber bundle (left fiber) while the fiber bundle on the right looks like a transverse fiber bundle. B: Higher resolution of formed by the MT. In histological sections, we myoseptum-skin connection from a HSP region. There is an detected these tendons by their thickness (40.4 6 example of a weaving fiber bundle and an example of a trans- 12.4 lm; n 5 17) that is significantly different verse fiber bundle. Yellow arrow points to the interlamellar con- from more medial areas of the flanking parts (8.9 nections. C: Same view as 6B but under polarized light. Parallel 6 3.8 lm; n 5 13, see also Fig. 7B). Red muscles collagen fibers appear in bright yellow. All micrographs are of cross sections along one stratum compactum fiber direction. Ep, insert into these robust MT (Fig. 7B). This muscle- epidermis; ms, myoseptum; rm, red muscle; sc, stratum compac- tendon association is present in the epaxial as well tum; scl, scale; ss, stratum spongiosum; tfb, transverse fiber as in the HFP. bundle; wfb, weaving fiber bundle. As in the flanking parts, myoseptal tendons also form the lateral-most area of the myoseptal

Journal of Morphology DOI 10.1002/jmor 38 N. DANOS ET AL.

Fig. 7. Arrangement of red muscles in Anguilla rostrata and insertions of red muscle fibers to specific myoseptal tendons. A: Overview of the thin layer of red muscle (encircled by dashed line) in the left epaxial midbody region; paraserial section. Note that the superficial thin layer of red muscles is well separated from the deeper white muscles. Along the horizontal septum red muscles enter deeper regions of the trunk. B: Insertion of red muscles into MT in EFP. Paraserial sections of left body side. Note that the red muscle fibers and lateralmost white muscle fibers insert into the thick tendon whereas more medial white muscle fibers insert into the thin part of the myoseptum. C: Polarized light micrographs showing two examples of insertions of red muscles into LT in HSP. Paraserial sections of left body side. D: Insertion of red muscles into myoseptal part close to horizontal septum. Note that the red muscle fibers do not insert into the horizontal septum. Sagittal section of left body side; cranial to left. Ep, epidermis; hs, horizontal septum; ms, myoseptum; rm, red muscles; sc, stratum compactum; v, vertebral axis; vs, vertical septum.

sloping part. These LT were visible as relatively obtuse angles with the posterior oblique tendon thick structures of connective tissue in histologi- (POT) fibers of the horizontal septum (Fig. 4B). cal sections (34.2 6 11.5 lm; n 5 12, as opposed Therefore, any transmission of red muscle forces to thin non-tendinous regions with a mean thick- to the POT would be directed along the path of ness of 9.9 6 4.4 lm; n 5 11, see also Fig. 7C). In the POT to the vertebral column. The maximum this region too, red muscle fibers insert into these distance across which any force transmission via robust LT. the horizontal septum can occur is the distance Sagittal sections show that red muscle fibers spanned by the POT fibers. This span was deter- are not directly connected to the horizontal sep- minedtobearound1.2%L (1.23 6 0.12%L; n 5 9 tum. Rather, these muscle fibers run longitudi- measurements in Specimens 1 and 2). In contrast nally, connecting adjacent myosepta (Fig. 7D) to red muscle fibers, the more medially placed and inserting into collagenous myoseptal fibers white muscle fibers insert directly into collage- that run caudolaterally. The myosepta in turn nous myoseptal fibers on the horizontal septum insert into the horizontal septum. The myoseptal parallel to POTs that are found in the septum fibers to which red muscle fibers attach are at (Fig. 4B).

Journal of Morphology DOI 10.1002/jmor CONNECTIVE TISSUE SYSTEM IN ANGUILLA ROSTRATA 39 DISCUSSION form and thunniform swimmers usually span 0.2L. 3D-Shape and Architecture of Myosepta and Both of these features, the lack of ENT and EPT Horizontal Septum and the elongation of LTs, have been noted in car- Comparison of myoseptal architecture: angiform and thunniform swimmers from a wide Eels versus non-anguilliform swimmers. This taxonomic range (e.g., lamnid sharks, carangids, study is the first to describe the architecture, swordfish, basal scombrids, and tunas) but have 3D-morphology, and vertebral insertions of myo- never been reported for any subcarangiform septa in an anguilliform swimmer. Our data show swimmer (for review see Shadwick and Gemballa, the presence of one anteriorly- and one posteriorly- 2006). pointing cone and a set of three myoseptal tendons In contrast, LTs of subcarangiform swimmers in each the hypaxial and epaxial parts of the myo- are short and relatively uniform along the body. septum of Anguilla rostrata (Figs. 2 and 3). Both For example, tendon lengths lie between 0.050L the cones and the three tendons (myorhabdoid, lat- anteriorly and 0.056L posteriorly in catsharks, and eral and epineural in the hypaxial region or epi- between 0.072L anteriorly and 0.082L posteriorly pleural in the epaxial tendon) have been shown to in the perciform Channa (for corresponding data be present in a number of teleost and non-teleost in 11 species see Shadwick and Gemballa, 2006; fishes, including basal actinopterygians and repre- average for subcarangiform swimmers is 6.4 6 0.9 sentatives of all major clades of chondrichthyans anteriorly and 7.5 6 1.1 posteriorly) and between (Gemballa et al., 2003a; Gemballa and Hagen, 0.086L anteriorly and 0.094L posteriorly in the 2004; Gemballa and Ro¨der, 2004). Except for the perciform Lepomis gibbosus (Gemballa, unpub- MT which has been identified as an apomorphic lished data). A similar LT average length and feature of the Myopterygii (Petromyzontiformes 1 rostrocaudal length uniformity has also been dem- Gnathostomata) these myoseptal features repre- onstrated for the American Eel in this study sent apomorphic features of the Gnathostomata (around 0.045L to 0.05L anteriorly and posteriorly; (Gemballa et al., 2003a, 2006). All of the above Table 2). Thus, with respect to this first example myoseptal features have been retained in the anguilliform swimmers are morphologically similar American Eel (Figs. 2 and 3). The same is true for to subcarangiform swimmers but differ consider- the insertion of the myoseptum to the vertebral ably from carangifom and thunniform swimmers. column which in the eel as well as in other actino- We discuss the functional significance of this pat- pterygian fishes spans three subsequent vertebrae tern in the light of muscle dynamics and swim- (Fig. 2 of this study; Gemballa et al., 2003a). ming kinematics in the section titled ‘‘Characteri- In Anguilla rostrata, as in other teleosts, most of zation of Undulatory Swimming Modes.’’ the ENT and EPT tendons are partly ossified, Comparison of the horizontal septum: Eels forming the ENB and EPB (see Fig. 2). In teleosts versus non-anguilliform swimmers. We found such membranous ossifications have been observed a cross-fiber array in the horizontal septum of in all myoseptal tendons to various degrees. Anguilla rostrata (see Fig. 4) that is similar to Though they generally follow a rostrocaudal pat- that of other actinopterygians. Both fiber direc- tern of ossification, with the rostralmost tendons tions are present in most species studied so far being ossified more often, the degree of ossification (35 actinopterygians; Gemballa et al., 2003b). varies largely even between closely related species The epicentral tendons show little variation among (Patterson and Johnson, 1995; Gemballa et al., species, whereas fibers of the second direction may 2003a). Since there is currently no experimental be evenly distributed (i.e., posterior oblique fibers; evidence (e.g., dynamic bending experiments of the e.g., in the anterior body region of Anguilla)or structure) pointing to the function of these rod-like form distinct tendons (i.e., POTs; e.g., in the poste- ossifications and since their systematic distribu- rior body region of Anguilla). However, this variation is irregular, we do not at this time propose a tion does not correlate with swimming modes. functional explanation for the existence of such Independent of the swimming type, a species may intermuscular bones. have POTs or posterior oblique fibers or even lack Recent morphological studies have revealed the whole horizontal septum all together. exceptions to the gnathostome groundpattern of myoseptal architecture in fishes with differing swimming modes. Carangiform and thunniform The Skin Connective Tissue System and its swimmers lack a distinct ENT and EPT. Further- Connections to the Myosepta more, in these species the length of the LT (i.e., Organization of the skin. Our SEM images of the distance spanned between the anterior and the stratum compactum (Fig. 5) show that lamellae posterior cones) increases twofold to threefold from are composed of only one layer of collagen fibers, in tendons of the anterior body region to the posterior accord with the results from a previous histological body region (Gemballa and Treiber, 2003; Gem- study of Anguilla rostrata skin (Leonard and balla, 2005; Shadwick and Gemballa, 2006; Gem- Summers, 1976) but in contrast to a study of balla et al., 2007). The posterior LTs in carangi- Lepadichthys lineatus skin (Fishelson, 1972) and a

Journal of Morphology DOI 10.1002/jmor 40 N. DANOS ET AL. developmental study in Danio rerio (Le Guellec from the vertebral column and at an angle close to et al., 2004). The presence of more than one layer of 458 with the longitudinal axis of the body (Vogel, collagen fibers per lamella seems to be the most com- 2004). This is the angle at which we would expect mon condition among other vertebrates with a simi- fiber reinforcement of the body walls if we model larly ordered stratum compactum (Olsson, 1961; the eel body as a cylinder with circular symmetry Fujii, 1968; Fishelson, 1972; Hawkes, 1974; Mittal in its material properties as well as symmetrical and Banerjee, 1974). We counted 30 lamellae from motions generating torsion. However, we know the lateral side of the body from one SEM image that the cross-section of an eel is neither perfectly and 13.58 6 1.33 parallel lamellae from nine micro- circular nor symmetrical in its material properties. graphs. Other studies of fish integument have found Furthermore, the presence of fins and the kine- 15 to 30 lamellae in the goby Chasmichthys gulosus matics of eel swimming (Tytell, 2004a,b) are fac- (Fujii, 1968), between 8 and 12 in seven species of tors that introduce asymmetry to the pattern of sharks (Motta, 1977), 15 to18 in Neoceratodus and torsional forces experienced by the body. With a 17 to 25 in polypterids (Gemballa and Bartsch, cross-section that is oblong in the dorsoventral 2002). American eels and the goby, C. gulosus, have direction we are not surprised to find that fibers the highest number of lamellae of the species exam- are directed more in that direction. The overall ined, suggesting a thick and tough skin. Both of dermal collagen fiber arrangement allows for lat- these species experience high mechanical abrasion eral flexibility of the body while resisting torsional when locomoting in contact with the substrate in forces. shallow freshwater or tide pools or even on land. A The subdermis, or hypodermis as it is also some- tougher but smooth skin with diminutive scales em- times called, contains the stratum spongiosum, a bedded in the epidermis as in the American eel may sheath of irregularly arranged collagen fibers and provide mechanical protection and reduced friction lipid storage. The specimens examined, all in the against the substrate (Fishelson, 1996). migratory elver stage, had ample fat storages. Fiber angle, the angle at which a fiber lies with Other researchers have noticed large fat stores in respect to the long axis of the body, has been used the air-breathing Channa striata and suggested to predict the mechanical effects on bending of ply- that the fat may function as a water repellant pre- wood-like structures that wrap around animal venting water loss when under desiccation stress bodies in a helical pattern. Alexander (1987) identi- or as a shock-absorber when C. striata buries in fies an angle of 608 as the angle at which extensible the mud (Mittal and Banerjee, 1975a). Even fibers in the body walls of helically wound animals though eels regularly locomote on land or in shal- do not change length during bending. Under the low marshy fields, given the catadromous migra- same model, fiber angles less or more than 608 tend tory stage the animals in this study were in, it is to increase the elastic strain energy of the system more likely that the fat in the stratum spongiosum in the bent position, causing the body to spring was for energy storage. back to the straight position. Harris and Crofton Myosepta-skin interconnections. Vertical (1957) calculate that a cylinder has maximum vol- bundles of collagen fibers, originating in the myo- ume when enclosed by a plywood-like arrangement septa, traversing the stratum spongiosum and join- with a fiber angle of 558. Cylinders under torsion ing fibers of the stratum compactum have been tend to break at 458 to their long axis and Hebrank observed at varying densities in nearly all previous (1982) who measured fiber angles in Anguilla ros- histological studies of fish skin including sarcop- trata indirectly by measuring the angle between terygian fish as well as in studies of Amphioxus, scale rows and the long axis of the body found them anuran and snake skin (Chapman and Dawson, to be approximately 458. 1961; Olsson, 1961; Liem, 1967; Logan and Odense, The 95% confidence intervals of the fiber angles 1974; Jayne, 1988; Gemballa and Bartsch, 2002). measured in this study, 57.58 6 6.928 for the cau- These bundles terminate in the basal membrane, dodorsal direction and 60.51 6 7.058 for the caudo- the layer between the stratum compactum and the ventral direction, contain all the above critical epidermis, and do not pass into the epidermis angles. This suggests that there is the potential (Mittal and Banerjee, 1975b; Leonard and for elastic storage in the skin of the American eel, Summers, 1976). The bundles provide bonding but we cannot confirm that this is indeed occur- strength to the stratum compactum lamellae and ring during locomotion. Although our results differ add strength to the whole plywood configuration. from those of Hebrank (1982), we still consider re- The three types of perpendicular fibers described sistance to torsion to be a major function of dermal earlier (Figs. 5E,F and 6B), have also been collagen. Collagen is a strong tensile material described in detailed morphological studies of per- (Wainwright et al., 1982) and any tensile force pendicular fiber bundles in the skin, such as that of applied to the fiber, such as torsion around the Amphioxus (Olsson, 1961) and of Danio skin (Le vertebral column as experienced by the skin, will Guellec et al., 2004). The perpendicular bundles be resisted. Any torsional forces generated during were not present in a leptocephalus larva examined swimming will be maximal at the farthest distance in one study (Leonard and Summers, 1976).

Journal of Morphology DOI 10.1002/jmor CONNECTIVE TISSUE SYSTEM IN ANGUILLA ROSTRATA 41 Evaluation of Possible Ways of Red Muscle of Myoseptal Architecture: Eels vs. Non-Anguilli- Force Transmission form Swimmers’’). Second, in thunniform swimmers red muscle inserts into the anterior part of the There are currently no techniques to directly or elongated LT. Thus, red muscle forces will be trans- indirectly measure forces along tendons or tissues mitted over a long distance. Such red muscle-ten- in swimming fishes at such small scales. Thus, our don arrangements have been identified in tunas understanding of force transmission in swimming and lamnid sharks and for both groups and it has fishes is mainly inferred from morphology. Three been shown experimentally that bending occurs connective tissue structures have been hypothe- more caudally than muscle contraction (Shadwick sized to play a role in force transmission in fishes: et al., 1999; Katz et al., 2001; Donley et al., 2004, myoseptal tendons, POTs of the horizontal septum, 2005; Gemballa, 2005). In contrast, as a conse- and the stratum compactum of the skin (Wain- quence of force transmission over a short distance wright et al., 1978; Hebrank, 1980; Hebrank and caudally in non-thunniform species, we expect Hebrank, 1986; Westneat and Wainwright, 2001; muscle contraction to lead to body bending close to Gemballa and Treiber, 2003; Donley et al., 2004; the site of and in phase with muscle contraction. Gemballa et al., 2006). In the following sections we Indeed, this has been demonstrated for several evaluate whether our results in the American eel subcarangiform and carangiform swimming tele- are consistent with these hypotheses. osts and one shark (Coughlin et al., 1996; Shad- Direct observations of insertions of red muscle wick et al., 1998; Katz et al., 1999; Donley and fibers into myoseptal tendons in some species sug- Shadwick, 2003). On the basis of our morphological gest that these tendons play a role in force trans- data we predict that this will also hold true for mission (Gemballa and Treiber, 2003; Donley anguilliform swimmers. et al., 2004; Gemballa et al., 2006). The signifi- Although the horizontal septum has been cance of myoseptal tendons as force transmitters hypothesized to be part of the red muscle force was further supported by a computational model transmission system in scombrid fishes (Westneat that suggests tendons might act to transmit forces and Wainwright, 2001), we cannot extend this hy- and add body flexural stiffness, and by findings in pothesis to include the American eel. First, red a shark species in which the length of the LT muscles that are placed close to the horizontal sep- agreed with the distance of force transmission pre- tum (i.e., ventrally to the anterior cone) do not dicted from muscle dynamics and swimming kine- insert into the horizontal septum but into the myo- matics (Long et al., 2002; Donley et al., 2004, 2005). septum (Fig. 7D). As these muscle fibers contract In this study we show that in Anguilla rostrata they will pull on the myoseptum but these forces red muscles insert into the lateral and myorhab- are unlikely to be transferred into the horizontal doid tendons (Fig. 7B,C). Thus, red muscle forces septum because the collagenous fibers in the myo- might be transmitted posteriorly through these septum and horizontal septum are at obtuse tendons. These tendons, however, are remarkably angles that are not efficient for force transmission short (around 0.045–0.05L; see Table 2) and the from the myosepta to the horizontal septum (Fig. red muscle fibers do not insert at the anterior ends 4B). In contrast, medially placed white muscle but in the middle of the tendons. Hence, this fibers will pull on myoseptal fibers that are in line arrangement suggests that forces will be trans- with the POTs of the horizontal septum (POTs; ferred only a short distance (maximum of 0.025L). Fig. 4B). Thus, muscular forces might be trans- This insertion of red muscle fibers in the mid ferred via POTs to the vertebral column. This force region of a LT has also been identified in subcar- transfer, however, will only span a short distance angiform and carangiform swimmers (Gemballa (around 0.01L; Fig. 4B). and Treiber, 2003; Gemballa et al., 2006; Shadwick The idea that the stratum compactum is part of and Gemballa, 2006). Since subcarangiform and the force transmission system in fishes was first anguilliform swimmers are also similar in LT proposed for the lemon shark, Negaprion brevirost- lengths (see above section titled ‘‘Comparison of ris (Wainwright et al., 1978). In that study the Myoseptal Architecture: Eels vs. Non-Anguilliform authors identified a fiber angle at which the skin Swimmers’’) they appear to share a similar muscu- would become extremely stiff and thus act as a lotendinous design. This overall morphological tendon. This angle was also identified in the Amer- similarity is consistent with the view that the mid- ican eel and it was inferred that eel skin would line kinematics in these two swimming types is also act as an exotendon (Hebrank, 1980, 1982). more similar than previously thought (Lauder and These studies assumed that internal body pressure Tytell, 2006). was putting the skin under tension but several The musculotendinous system of carangiform and studies have found that there are different inter- thunniform swimmers differs from that of anguilli- nal pressures on contralateral sides of the body, form or subcarangiform swimmers in some ways. negating the simple transformation of internal First, carangiform and thunniform swimmers bear body pressure into body wall stress by modeling elongated LTs (see earlier section titled ‘‘Comparison the body as thin-walled cylinder (Westneat et al.,

Journal of Morphology DOI 10.1002/jmor 42 N. DANOS ET AL. 1998; Danos, 2005). Therefore, the only other way thermore, amplitudes in the whole posterior half that the skin can be put under tension and thus of the body appear to be similar in all swimming act as a force transmitter would be by direct physi- types (Lauder and Tytell, 2006). cal connection of the force-producing elements, the If the parameters from traditional 2D-midline muscle fibers, to the force transmission tissue, the kinematics do not adequately discriminate bet- myosepta and skin. But the results of our present ween undulatory swimming types, parameters study demonstrate that neither red nor white from other fields are needed to either support the muscles directly insert into the skin. On the basis traditional classification or redefine new types. We of our morphological findings presented here, the consider the morphology of the locomotory system, only way of transmitting muscular forces to the muscle dynamics, and hydrodynamics to be inte- skin would be through the transverse or weaving gral features that affect swimming styles of fishes. fiber bundles (tfb, wfb; Figs. 5 and 6). These con- Since recent studies, including this study on the nections, though, do not appear very robust when American Eel, have provided comparative data in compared with the LTs or MT (Figs. 6 and 7). each of these fields we are able to update the cur- Moreover, they do not form continuous intercon- rent discussion on classification of swimming nections between myoseptal tendons and skin but modes in fishes. are present as single fiber bundles that are evenly Anguilliform and subcarangiform swimmers are distributed along a myoseptum (Fig. 5B,C). morphologically very similar (eels: this study; tele- Studies of body stiffness at different phases rela- osts: e.g., Gemballa and Treiber, 2003; Gemballa tive to the strain cycle of the muscles in electri- et al., 2003a; Gemballa and Ro¨der, 2004; Gemballa cally stimulated eels suggested that net positive et al., 2006; Shadwick and Gemballa, 2006). They power in the posterior muscle of eels is, in part, share the same 3D-shape of myosepta and the produced by an elastic strain energy mechanism same set of myoseptal tendons. This arrangement (Long, 1998). At that point too little was known differs substantially from carangiform and thunni- about the musculotendinous system in eels to find form swimmers that lack ENT and have elongate the morphological correlate for the serial elastic LTs in posterior body (see section titled: ‘‘Compari- elements that were hypothesized to account for son of Myoseptal Architecture: Eels vs. Non- elastic energy storage in eels. From the morpholog- Anguilliform Swimmers’’; Shadwick and Gemballa, ical findings presented here we hypothesize that 2006). Thus, swimming types are well divided into the lateral and MT might contribute substantially carangiform and non-carangiform by morphologi- to the body’s mechanical properties. However, as cal features. the predictions derived from the analysis of body Recent comparative hydrodynamic studies have stiffness only partly match the results of in vivo identified substantial differences in the wake pat- studies of muscle activity (Gillis, 1998a), the idea terns of eels and other teleosts that can be used to of an elastic strain energy mechanism remains divide swimming types into anguilliform and non- equivocal. anguilliform types (e.g, Tytell and Lauder, 2004; Lauder, 2006; Lauder and Tytell, 2006). These differences in wake patterns are thought to occur Characterization of Undulatory mainly due to differences in body shape. Non- Swimming Modes anguilliform swimmers (especially carangiform In the classical review of fish swimming modes, swimmers) exhibit a marked narrowing of the lon- Lindsey (1978) distinguished three main types of gitudinal shape anterior to the tail that is not axial undulatory swimming: anguilliform, sub- present in anguilliform swimmers. carangiform, and carangiform. These swimming The length-specific propulsive wavelength (k/L) modes were thought to form a continuum with the is one of the classical kinematic parameters used subcarangiform mode representing an intermedi- by Lindsey (1978) that still seems to be valid for ate mode. The differentiation of these swimming defining undulatory swimming types. Recent types is based on two parameters that are derived studies confirmed the classical view that k/L grad- from 2D-midline kinematics, the length-specific ually increases from anguilliform to carangiform body amplitude as a function of body position (A/ swimmers (0.6 to >1.2; e.g., Videler, 1993; Wardle L), and length-specific wavelength of the propul- et al., 1995; Donley and Dickson, 2000; Donley and sive wave (k/L). However, recent comparative stud- Shadwick, 2003; Dowis et al., 2003; Tytell and ies revealed that the midline kinematics of Lauder, 2004). Furthermore, the length-specific the anguilliform mode is unexpectedly similar to propulsive wavelength is correlated with red mus- other modes, especially the subcarangiform mode cle activation patterns. Fishes that swim with (Lauder and Tytell, 2006). During slow swimming great k/L have long burst durations (i.e., the time (<2 L/s) the specific tail beat amplitude lies a certain muscle segment is turned on within a between 0.08 and 0.12L in anguilliform swimmers tail-beat cycle) anteriorly and a large decrease and between 0.07 and 0.13L in subcarangiform (17–19%) in burst duration between anterior and swimmers (Videler, 1993; Gillis, 1997, 1998b). Fur- posterior segments (review: Gillis, 1998a). As a

Journal of Morphology DOI 10.1002/jmor CONNECTIVE TISSUE SYSTEM IN ANGUILLA ROSTRATA 43 consequence, in carangiform swimmers anterior Brainerd EL, Patek SN. 1998. Vertebral column morphology. C- sites still remain active (long-duration anterior start curvature, and the evolution of mechanical defenses in tetraodontiform fishes. Copeia 1998:971–984. bursts) while the wave of activity reaches posterior Chapman GB, Dawson AB. 1961. Fine structure of larval anu- sites. Thus, muscle activity in carangiform ran epidermis, with special reference to figures of Eberth. swimmers is characterized by a long block of ipsi- J Biophys Biochem Cytol 10:425–436. lateral muscle activity. In contrast, anguilliform Coughlin DJ, Valdes L, Rome LC. 1996. Muscle length changes swimmers have relatively short burst durations during swimming in scup: Sonomicrometry verifies the ana- tomical high-speed cine technique. J Exp Biol 199:459–463. anteriorly with almost no decrease (0–5%) towards Danos N. 2005. Biomechanics of the skin during swimming in posterior sites (Gillis, 1998a). Subcarangiform the American Eel, Anguilla rostrata. Masters Thesis. swimmers are at intermediate values (10–14%). Amherst: University of Massachusetts, 47 p. Interestingly, the long block of ipsilateral muscle D’Aout K, Curtin NA, Williams TL, Aerts P. 2001. Mechanical properties of red and white swimming muscles as a function activity in carangiform swimmers is associated of the position along the body of the eel Anguilla anguilla. with long myoseptal tendons in the posterior body J Exp Biol 204:2221–2230. (e.g., tendon lengths of 0.16–0.20L; Shadwick and Dingerkus G, Uhler DL. 1977. Enzyme clearing of Alcian blue- Gemballa, 2006) whereas in anguilliform and sub- stained whole small vertebrates for demonstration of carti- carangiform swimmers short blocks of ipsilateral lage. Stain Technol 32:229–231. Donley JM, Dickson KA. 2000. Swimming kinematics of juve- muscle activity are associated with short tendons nile Kawakawa Tuna (Euthynnus affinis) and Chub Mackerel in all body regions (e.g., around 0.05L in eels, this (Scomber japonicus). J Exp Biol 203:3103–3116. study; 0.005–0.009L in other fishes, Shadwick and Donley JM, Shadwick RE. 2003. Steady swimming muscle dy- Gemballa, 2006). However, a causal relationship namics in the leopard shark Triakis semifasciata. J Exp Biol that explains how these differences in musculoten- 206:1117–1126. Donley JM, Sepulveda CA, Konstantinidis P, Gemballa S, Shad- dinous design and muscle function are translated wick RE. 2004. Convergent evolution in mechanical design of into different swimming modes (e.g., different pro- lamnid sharks and tunas. Nature 429:61–65. pulsive wavelengths) has not yet been established. Donley JM, Shadwick RE, Sepulveda CA, Konstantinidis P, These examples from morphology, hydrodynam- Gemballa S. 2005. Longitudinal patterns of red muscle strain and activation and body kinematics during steady swimming ics, and muscle physiology demonstrate that differ- in a lamnid shark, the shortfin mako (Isurus oxyrinchus). ences across the classical swimming types can be J Exp Biol 208:2377–2387. found when parameters like wake patterns, pro- Dowis JD, Sepulveda CA, Graham JB, Dickson KA. 2003. pulsive wavelengths, red muscle activity patterns, Swimming peformance studies on the eastern Pacific bonito and LT lengths are compared. Thus, the tradi- Sarda chiliensis, a close relative of the tunas (family Scom- bridae). II. Kinematics. J Exp Biol 206:2749–2758. tional classification still appears to be justified Fishelson L. 1972. Histology and ultrastructure of the skin of although the various swimming types are very Lepadichthys lineatus (Gobiesocidae: Teleostei). Mar Biol 17: similar in 2D-midline kinematics. However, data 357–364. from the various fields mentioned earlier, except in Fishelson L. 1996. Skin morphology and cytology in marine eels adapted to different lifestyles. Anat Rec 246:15–29. the case of eels, were usually not collected on the Fujii R. 1968. Fine structure of the collagenous lamella underly- same species and thus the picture drawn may rep- ing the epidermis of the goby, Chasmichthys gulosus. Annot resent a trend rather than a clear picture. A Zool Jpn 41:95–106. clearer redefinition of swimming types will need Gemballa S. 2005. Locomotory specializations in aquatic envi- further support from comparative studies that ronments: Morphological differences in the axial musculotendinous system of cruising specialists and accelerators. Comp obtain all relevant data from multiple species. Biochem Physiol A 141:S163. Gemballa S, Bartsch P. 2002. Architecture of the integument in lower teleostomes: Functional morphology and evolutionary implications. J Morphol 253:290–309. ACKNOWLEDGMENTS Gemballa S, Hagen K. 2004. The myoseptal system in Chi- maera monstrosa and its evolution in the gnathostome stem We would like to thank E.L. Brainerd for her lineage. Zoology 107:13–27. generous guidance and support. Our work bene- Gemballa S, Ro¨der K. 2004. From head to tail: The myoseptal fited from the fine technical assistance of Monika system in basal actinopterygians. J Morphol 259:155–171. Meinert (Tu¨ bingen) for histology and Charlie Gemballa S, Treiber K. 2003. Cruising specialists and accelera- Hellmer for SEM imaging (Tu¨ bingen). 3D-recon- tors—Are different types of fish locomotion driven by differ- ently structured myosepta? Zoology 106:203–222. structions using Amira were done under the Gemballa S, Vogel F. 2002. Spatial arrangement of white mus- instruction of Thomas Schmelzle (Tu¨ bingen). The cle fibers and myoseptal tendons in fishes. Comp Biochem manuscript benefited from comments by M. Dean Physiol A 133:1013–1037. and E.M. Standen on earlier versions. Gemballa S, Ebmeyer L, Hagen K, Hoja K, Treiber K, Vogel F, Weitbrecht GW. 2003a. Evolutionary transformations of myoseptal tendons in gnathostomes. Proc R Soc Lond B 270:1229–1235. LITERATURE CITED Gemballa S, Hagen K, Ro¨der K, Rolf M, Treiber K. 2003b. The horizontal septum: Structure and evolutionary transforma- Alexander RMcN. 1969. The orientation of muscle fibers in the tions in vertebrates. J Evol Biol 16:966–975. myomeres of fishes. J Mar Biol Assoc UK 49:263–290. Gemballa S, Konstantinidis P, Donley JM, Sepulveda C, Shad- Alexander RM. 1987. Bending of cylindrical animals with heli- wick RE. 2006. Evolution of high-performance swimming in cal fibres in their skin or cuticle. J Theor Biol 124:97–110. sharks: transformation of the musculotendinous system from

Journal of Morphology DOI 10.1002/jmor 44 N. DANOS ET AL.

subcarangidorm to thunniform swimmers. J Morphol 267: Long JH Jr. 1998. Muscles, elastic energy, and the dynamics of 477–493. body stiffness in swimming eels. Am Zool 38:771–792. Gemballa S, Kistner N, Sepulveda C, Bernal D. 2007. The mus- Long JH Jr, Hale ME, McHenry MJ, Westneat MWW. 1996. culotendinous system of pelagic fishes: How does swordfish Functions of fish skin: Flexural stiffness and steady swim- (Xiphias gladius) compare to thunniform and carangifom ming of longnose gar Lepisosteus osseus. J Exp Biol 199: swimmers? ICVM-8 abstract. J Morphol 268:1077. 2139–2151. Gillis GB. 1996. Undulatory locomotion in elongate aquatic ver- Long JH Jr, Adcock B, Root RG. 2002. Force transmission via tebrates: Anguilliform swimming since Sir James Gray. Am axial tendons in undulating fish: A dynamic analysis. Comp Zool 36:656–665. Biochem Physiol A 133:911–929. Gillis GB. 1997. Anguilliform locomotion in an elongate sala- Mittal AK, Banerjee TK. 1974. Structure and keratinization of mander (Siren intermedia): Effects of speed on axial undula- the fresh water teleost Notopterus notopterus (Notopteridae, tory movements. J Exp Biol 200:767–784. Pisces). J Zool (Lond) 174:341–355. Gillis GB. 1998a. Neuromuscular control of anguilliform locomo- Mittal AK, Banerjee TK. 1975a. Histochemistry and structure tion: Patterns of red and white muscle activity during swim- of skin of a Murrel, Channa striata (Bloch, 1797) (Channi- ming in the American Eel Anguilla rostrata. J Exp Biol formes, Channidae). I. Epidermis. Can J Zool 53:833–843. 201:3245–3256. Mittal AK, Banerjee TK. 1975b. Histochemistry and structure Gillis GB. 1998b. Environmental effects on undulatory locomo- of skin of a Murrel, Channa striata (Bloch, 1797) (Channi- tion in the American Eel Anguilla rostrata: Kinematics on formes, Channidae). II. Dermis and subcutis. Can J Zool water and land. J Exp Biol 201:949–961. 53:844–852. Gray J. 1933a. Studies in animal locomotion. I. The movement Motta PJ. 1977. Anatomy and functional-morphology of dermal of fish with special reference to the eel. J Exp Biol 10:88–104. collagen-fibers in sharks. Copeia 1977:454–464. Gray J. 1933b. Studies in animal locomotion. II. The relation- Mu¨ller UK, Smit J, Stamhuis EJ, Videler JJ. 2001. How the ship between waves of muscular contraction and the propul- body contributes to the wake in undulatory swimming fish: sive mechanism of the eel. J Exp Biol 10:386–390. Flow fields of a swimming eel (Anguilla anguilla). J Exp Biol Gray J. 1933c. Studies in animal locomotion. III. The propulsive 204:2751–2762. mechanism of the whiting (Gadus merlangus). J Exp Biol Olsson R. 1961. The skin of Amphioxus. Z Zellforsch 54:90–104. 10:391–400. Patterson C, Johnson GD. 1995. The intermuscular bones and Gray J. 1968. Animal Locomotion. New York: WW Norton & Co. ligaments of teleostean bones. Smithsonian Contrib Zool Harris JE, Crofton HD. 1957. Structure and function in the 559:1–83. nematodes—Internal pressure and cuticular structure in Romeis B. 1986. Mikroskopische Technik. Mu¨ nchen: Urban and Ascaris. J Exp Biol 34:116–130. Schwarzenberg. 697 p. Hawkes JW. 1974. Structure of fish skin. I. General organiza- Shadwick RE, Gemballa S. 2006. Kinematics and muscle dy- tion. Cell Tissue Res 149:147–158. namics in undulatory swimming. In: Shadwick RE, Lauder Hebrank MR. 1980. Mechanical properties and locomotor func- GV, editors. Fish Biomechanics. Fish Physiology, Vol. 23. San tions of eel skin. Biol Bull 158:58–68. Diego: Academic Press. pp 241–280. Hebrank MR. 1982. Functional morphology of fishes: Roles of Shadwick RE, Steffensen JF, Katz SL, Knower T. 1998. Muscle the backbone and skin in locomotion. Doctoral Dissertation, dynamics in fish during steady swimming. Am Zool 38:755– Duke University. 142 p. 770. Hebrank MR, Hebrank JH. 1986. The mechanics of fish skin— Shadwick RE, Katz SL, Korsmeyer K, Knower T, Covell JW. Lack of an external tendon role in 2 teleosts. Biol Bull 1999. Muscle dynamics in skipjack tuna Katsuwonus pela- 171:236–247. mus: Timing of red muscle shortening and body curvature Jayne BC. 1988. Mechanical-behavior of snake skin. J Zool during steady swimming. J Exp Biol 202:2139–2150. 214:125–140. Tytell ED. 2004a. The hydrodynamics of eel swimming. II. Katz SL, Shadwick RE, Rapoport HS. 1999. Muscle strain his- Effect of swimming speed. J Exp Biol 207:3265–3279. tories in swimming milkfish in steady as well as sprinting Tytell ED. 2004b. Kinematics and hydrodynamics of linear gaits. J Exp Biol 2002:529–541. acceleration in eels, Anguilla rostrata. Proc R Soc Lond B Katz SL, Syme DA, Shadwick RE. 2001. High speed swimming: 271:2535–2540. Enhanced power in yellowfin tuna. Nature 410:770–771. Tytell ED, Lauder GV. 2004. The hydrodynamics of eel swim- Kern S, Koumoutsakos P. 2006. Simulations of optimized ming. I. Wake structure. J Exp Biol 207:1825–1841. anguilliform swimming. J Exp Biol 209:4841–4857. Videler JJ. 1993. Fish Swimming. Fish and Fisheries Series Lauder GV. 2006. Locomotion. In: Evans DH, Claiborne JB, edi- 10. London: Chapman & Hall. 260 p. tors. The Physiology of Fishes, 3rd ed. Boca Raton: CRC Vogel S. 2004. Comparative Biomechanics. Princeton and Press. pp 3–46. Oxford: Princeton University Press. Lauder GV, Tytell ED. 2006. Hydrodynamics of undulatory pro- Wardle CS, Videler JJ, Altringham JD. 1995. Tuning in to fish pulsion. In: Shadwick RE, Lauder GV, editors. Fish Biome- swimming waves: Body form, swimming mode and muscle chanics. Fish Physiology, Vol. 23. San Diego: Academic Press. function. J Exp Biol 198:1629–1636. pp 425–468. Westneat MW, Wainwright SA. 2001. Mechanical design for Le Guellec D, Morvan-Dubois G, Sire JY. 2004. Skin develop- swimming: Muscle, tendon, and bone. In: Block BA, Stevens ment in bony fish with particular emphasis on collagen depo- ED, editors. Tuna: Physiology, Ecology, and Evolution. Fish sition in the dermis of the zebrafish (Danio rerio). Int J Dev Physiology, Vol. 19. San Diego: Academic Press. pp 271– Biol 48:217–231. 311. Leonard JB, Summers RG. 1976. Ultrastructure of integument Westneat MW, Hale ME, McHenry MJ, Long JH Jr. 1998. of American eel. Anguilla rostrata. Cell Tissue Res 171:1–30. Mechanics of the fast-start: Muscle function and the role of Liem KF. 1967. Functional morphology of integumentary respi- intramuscular pressure in the escape behavior of Amia calva ratory and digestive systems of synbranchoid fish Monopterus and Polypterus palmas. J Exp Biol 201:3041–3055. albus. Copeia 1967:375–388. Wainwright SA. 1983. To bend a fish. In: Webb PW, Weihs D, Lindsey CC. 1978. Form, function, and locomotory habits in editors. Fish Biomechanics. New York: Praeger Publishers. pp fish. In: Hoar WS, Randall DJ, editors. Locomotion. Fish 68–91. Physiology, Vol. 7. New York: Academic Press. pp 1–100. Wainwright SA, Vosburgh F, Hebrank JH. 1978. Shark skin— Logan VH, Odense PH. 1974. Integument of ocean sunfish (Mola- Function in locomotion. Science 202:747–749. Mola L) (Plectognathi) with observations on lesions from two Wainwright SA, Biggs WD, Currey JD, Gosline JM. 1982. Me- ectoparasites. Capsala-Martinierei (Trematoda) and Philortha- chanical Design in Organisms. Princeton, New Jersey: Prince- goriscus-Serratus (Copepoda). Can J Zool 52:1039–1045. ton University Press.

Journal of Morphology DOI 10.1002/jmor