Architecture and composition of the muscles that drive stomatopod eye movements

Stomatopods, commonly known as mantis shrimps, are a group of crustaceans which are frequently found in shallow well-lit marine habitats, such as reef tops (Caldwell and Dingle, 1976). These predatory animals have many visually guided behaviours and the peculiar nature of their eyes has long been noted (Exner, 1891). The full complexity of their optical system has, however, only recently been elucidated. Many species have an array of six ommatidial rows, the so-called mid-band, which runs across each eye. This is apparently highly adapted for the extraction of spectral and polarization information (Marshall, 1988; Marshall et al. 1991a,b). The mid-band is sandwiched between two hemispheres, which function like conventional apposition compound eyes. There is a great deal of optical overlap between these three eye regions.

The structure of the eye suggests that, while the hemispheres work best when fixating and tracking objects, the mid-band needs to be swept slowly across the field of view (that is ‘scanned’) if it is to extract information. Land et al. (1990) and Cronin et al. (1991) have employed interactive computer/video equipment to examine the stomatopod’s complex repertoire of eye movements. Their work has shown that the eye movements extend over a wide range around all three rotational axes and that each component can change independently of the other two, although they often act together. Furthermore, the movements can conveniently, if not exclusively, be placed in one of two classes; either small slow scans or larger faster targetting and tracking movements. It might therefore be reasonable to expect this behavioural division to be reflected in the musculature.

The muscular arrangements of some of the other stalk-eyed crustaceans have been studied (Burrows and Horridge, 1968; Mellon, 1977; Neil, 1982), but no investigation has yet been made of the remarkable stomatopod eye joint. Consequently, no analysis of the types of muscle fibre present has been made. This paper sets out to remedy this and to provide a basis for further investigations of how the stomatopod’s complex repertoire of eye movements is generated.

This study primarily concerns the large Indo-Pacific gonodactyloid Odontodactylus scyllarus (Linnaeus), specimens of which were either bought from tropical marine fish supply houses or caught in the wild. The animals were held in standard tropical marine aquaria and fed live shrimps or frozen fish food until used.

Animals were humanely killed, by cutting their nerve cords, and were dissected either fresh or after fixation in a 4% solution of formaldehyde in stomatopod saline (Watanabe et al. 1967). Sometimes Methylene Blue was used in fresh dissections to aid the differentiation of structures (Kunze, 1981).

For histological reconstruction, tissue samples were processed and stained according to standard protocols (Bancroft and Stevens, 1990). The stomatopod eyecup frequently proved to be difficult to section, but the use of chelating agents was avoided as they often left the muscles damaged or led to inconsistent staining patterns.

The eye movements of restrained animals were videotaped using an adaptation of the apparatus designed by Land et al. (1990). The adaptation consisted of a second camera, which shared a time signal with the first, and recorded the animal from the dorsal aspect. Sequences of suitable movements were then selected and the eye’s centre of rotation was estimated from overlays.

Whole eyes were suspended by their stalks and shock frozen in liquid nitrogen. After orientation in Tissue-TEK mounting medium (Miles, Elkhart, USA), 20 μm serial sections were cut in a cryomicrotome at -30°C. The resulting series were then tested for their total myofibrillar ATPase activity, the pH lability of myosin ATPase (mATPase) activity and succinate dehydrogenase activity using protocols described by Fowler and Neil (1992). The ATPase final incubation reactions were performed at pH9.4 as this provides the optimal conditions for phosphate precipitation (Günzel et al. 1993). Fibres could thus be identified and followed through the series.

Sections were either stained with a 1% solution of Toluidine Blue or mounted as lightly fixed unstained tissue. Cross-sectional area analysis was then performed on histochemically identified fibres. The areas were calculated by cutting up photographic enlargements of the muscle cross sections, weighing the photographs of the fibres and then transforming these weights into areas by comparison with a standard. The latter was generated using a microscope stage graticule.

When describing structure, it is always important to use clearly defined anatomical axes and this is especially true in the case of a highly mobile eye. A slightly artificial, but reproducible, system of axes based around a fixed eye condition is therefore employed here. In this system, the mid-band is regarded as being horizontal and the larger hemisphere is dorsal. Left and right always refer to the animal’s perspective. This is compatible with the terms previously used to describe the stomatopod retina (Marshall et al. 1991a).

The large prominent eyecups of Odontodactylus scyllarus are mounted on small stalks. Although the proximal end of the stalk can move relative to the head, the majority of the eye’s motion originates from the distal junction between the stalk and cup. It is this joint that is considered here. Various aspects of the assembly, which is described below, are illustrated in Figs 1 and 2.

In common with the crab (Burrows and Horridge, 1968), the joint has no fixed condyles but consists of a sheet of flexible arthrodial membrane with the eyecup suspended by its muscles. This membrane is ruched to provide the slack necessary when eye movements are made. It is also supported by a triangular fold of cuticle and a piece of strengthened membrane, which articulates onto the distal end of the stalk.

On the lateral side of the cup, an infolding of the membrane contains a curved, scalloped triangle of cuticle which articulates with the stalk and is henceforth called the lateral sclerite.

The two eyecups are mirror-symmetrical about the animal’s mid-line and each contains eight individual muscles. These have been named using an arbitrary nomenclature that reflects their gross position and appearance in the eyecup. Muscle 1 has its origin on the lateral sclerite and runs dorsally and slightly medially to its insertion on the eyecup. It runs almost parallel to the posterior face of the cup. Muscle 2 is the shortest in the eyecup. It originates from the ventral edge of the lateral sclerite and runs ventrolaterally to insert on the eyecup. Its insertion is more distal than that of muscle 1. Muscle 3 originates on the lateral sclerite and runs laterally and distally to its insertion. Muscle 4 also originates from the lateral sclerite and runs medially to terminate on the eyecup wall. Muscle 5a is a distinctive penniform muscle that runs from the mediodorsal edge of the eyestalk projection to a very broad insertion. Its site of termination is a little distal to its origin.

Muscle 5b has its origin slightly medial to that of muscle 5a and runs alongside it. It is slender, with a long tendon that passes dorsally around some of the stalk. It is also the only muscle in the eyecup to be contained within a membranous sheath. Muscle 6a runs from a medial point on the eyestalk and has a fan-like structure similar to that of muscle 5a. Muscle 6b shares its origin with muscle 6a and runs ventrolaterally.

The preferred ocular position adopted by ‘resting’ individuals of Odontodactylus is one in which the mid-bands make an angle of about 30° to the animal’s horizontal plane, with their medial ends depressed. This situation presumably represents a balance point where all of the muscles in the sling are under the minimum amount of imposed stretch (although this was not measured directly in this study). The effect of the contraction of individual muscles can be approximated by taking the midline of each one as an estimate of its line of action (Knox and Donaldson, 1991); if this is done with the eyes at rest, muscles 1 and 2 are seen to cause movement in the pitch plane (latitude) while muscles 3 and 4 alter the yaw position (longitude).

The function of the obliquely running muscle pairs 5 and 6 is not quite as clear, but it appears possible that the fan-shaped muscles 5a and 6a may act primarily as antagonists for muscles 6b and 5b, in order to bring about movements in the pure rotational plane (bearing). This is actually more complicated than it may at first seem, because, although it is relatively easy for a linear motor such as a muscle fibre to cause a translational movement, extra constraints are needed to achieve rotation. Since the eye’s centre of rotation is not fixed by a condyle, it is not possible to generate rotation in a fashion analogous to a bow-drill. Instead, as muscles 5b and 6b contract, fibres must be recruited to oppose the translation that they would otherwise cause. In Fig. 3 a model is proposed by which the fan-shaped muscles 6a and 5a could fulfil this role. Providing that the muscle pairs are appropriately yoked, this arrangement would limit any translational component to their combined action. This also explains the observation that the wide arc of the insertion of muscles 5a and 6a seems to be similar to the range of observed rotational movement.

No consistent centre of rotation was found either in complex movements or in those that appeared to be confined to a single plane. Instead, the virtual pivot seemed to move around the proximal portion of the eyecup in the region where the muscles are located. This is what one would expect from an eyecup which has no fixed hinge (Burrows and Horridge, 1968).

The results discussed here relate to muscle 1, but similar data have been obtained for the other muscles.

The total mATPase test revealed the existence of two populations of fibre which stained either darkly or lightly (Fig. 4A). This test is routinely used to distinguish between slow (type 1) and fast (type 2) fibres, with the latter staining more darkly. This is because the activity of mATPase can be related to a fibre’s maximum speed of contraction (Bárány, 1967; Müller et al. 1992).

The tests for pH stability showed there to be at least two isoforms of myosin ATPase. If the sections were exposed to an alkaline (Fig. 4B) or acidic (Fig. 4C) preincubation medium, then the picture which developed was complementary to that of the total reaction test. This result, that the slower fibres possess the mATPase isoform with the widest pH tolerance, is in accordance with data obtained from the crab (Rathmayer and Maier, 1987). The reciprocity of these results also allows the experiments to act as their own controls. At extremes of preincubation pH, acidic values of 4 and alkaline values of 11, all mATPase activity was extinguished (see Fig. 4D for pH4).

The test for oxidative capacity of the fibres, which is a measure of their fatigue resistance, relied upon the activity of the enzyme succinate dehydrogenase (SDH). Individual fibres showed the Formazan dye product to be concentrated in their periphery. This is characteristic of crustacean muscles (Ogonowski and Lang, 1979) and is due to the sub-sarcolemmal position of their mitochondria. The general pattern of staining was similar to that shown by the ATPase reactions, with the proposed faster fibres expressing lower levels of SDH (Fig. 4E). The agreement between sections, however, was never as complete as that between the tests for total mATPase activity and the pH stability of the mATPases. The stain intensity also varied so that a range of fibre types, designated ‘strong’, ‘medium’ and ‘weak’, could be subjectively identified (Fig. 4E and more clearly in Fig. 4F, which shows a section of muscle 4). Comparable groupings were obtained by both knowledgeable and naïve sorters, but this consistency broke down if further subdivisions were attempted.

Control sections run with the enzyme substrate excluded from the reaction medium showed no, or only faint background, staining.

Fig. 5A shows the raw data for fibre cross-sectional area pooled from histological sections of five different examples of muscle 1. Fig. 5B shows the same data as Fig. 5A, but standardised to eliminate any error induced through oblique sectioning angles. The standardisation procedure simply consisted of expressing each fibre’s area as a proportion of its parental muscle’s cross-sectional area. Fig. 5B has more peaks than Fig. 5A. Fig. 5C shows the absolute data from another muscle 1, which was cut accurately perpendicular to its long axis. The abscissae of Fig. 5B,C are equivalent with respect to absolute fibre area and it can be seen that the graphs have similar profiles. Both graphs give a strong indication that the fibres can be divided into three or even four groups on the basis of their size.

Qualitatively it can be seen from Fig. 6A that there is a slight tendency for the fibres that stain positively with the total mATPase reaction, the labile type 2, to be larger than the negatively staining type 1. This ties in with observations on lobster leg muscles in which fast followers are larger (Jahromi and Atwood, 1971). In the stomatopod eye, however, there is an additional peak of small, fast, type 2 fibres.

The SDH groupings were also found to be correlated to size, with stain intensity and cross-sectional area being inversely related (Fig. 6B). Here again, there is an apparently anomalous peak of weakly staining fibres with a small cross-sectional area that corresponds to the ATPase histogram. As with the mATPase data, the peaks are graded into one another, which may indicate the existence of other sub-populations that are indiscernible to the naked eye (Reichmann and Pette, 1982).

From the histochemical analysis, it is clear that there is no straightforward dichotomy between fast and slow fibre types in stomatopod eyes. Instead, there exists a number of overlapping populations, members of which may be only partially delineated by a single test. This is because the oxidative and glycolytic systems are independently variable with no obligate coupling between them (Nemeth and Pette, 1981). Many fibre type combinations are therefore potentially possible. Using these two pathways as markers for the stomatopod eye, four classes are identifiable, with at least one more being apparent on the basis of cross-sectional area. Since many workers are able to employ the relative intensities of cobalt deposits to distinguish fibre types even further (Günzel et al. 1993), and as there are other physiological parameters which could be examined, we should begin to wonder whether these really are discrete classes or merely parts of a continuum (see Discussion).

Given the conservative nature of vertebrate extraocular design (Young, 1981), one might expect there to be little variety within the stalk-eyed crustaceans. It may therefore seem odd that the stomatopod has an arrangement rather different from those of the crab (Burrows and Horridge, 1968) or crayfish (Mellon, 1977). However, even closely related species of crayfish have rather different ocular muscular plans (Robinson and Nunnemacher, 1966). The stomatopods diverged from the main malacostracan lineage about 400 million years ago (Schram, 1969). Their large eyes are different, not only in their need for scanning movements but also because of their comparatively slender stalks. The size of the eyecup precludes the possibility of a protective withdrawal reflex, while the stalk adds to the extent of available movement. It is therefore not surprising that the stomatopods have rather idiosyncratic eye muscles.

Detailed analysis of stomatopod eye movements has shown that if they are decomposed into their component parts then changes in latitude, longitude and bearing are independently variable. This is impressive because it means that, as with the vertebrate oculomotor system, there is no mechanical ‘cross-talk’ between the rotational axes. This strongly suggests that each degree of rotational freedom has its own ‘neural driver’ (Land et al. 1990). The myology of the eyecup further supports and extends this hypothesis, as it gives good candidates for dominant prime mover muscle pairs (Neil, 1982) in each axis. Sub-division of the system in this way may facilitate the superposition of the different sorts of movement. However, it should be remembered that the eye movements must result from the concerted action of several muscles, rather than the exclusive influence of just one pair (Burrows and Horridge, 1968).

The animal must also have an internal map of the world in head or body coordinates, so that its many visually guided behaviours can be directed. If the animal is to use retinal information to construct such a map, then the eye movements must be compensated for. Although Sandeman (1964) found proprioceptive organs in the antennules of Squilla mantis (a member of the squilloid superfamily of stomatopods), he could find none associated with the eyecup. Likewise, none have been seen during this investigation of gonodactyloid eyes. While each eye may be watching the other to monitor its position in space (Cronin et al. 1991), a far more plausible theory is that the control system is based on efference copy, that is a signal related to the shrimp’s oculomotor output. Coordinate prime movers would simplify this task in much the same way that latitude and longitude aid mariners. Indeed, it has been shown that the orthogonal components of barn owl head movements are each controlled by their own distinct neural circuit (Masino and Knudsen, 1990). In the case of the bird, the coordinate map is abstract relative to the neck anatomy but the authors feel that coordinate systems may prove to be a general feature of sensorimotor organisation. They even cite the extraocular muscles of vertebrates and the generation of saccades as another likely example. The analogies between stomatopod and vertebrate extraocular structure are clear.

The behavioural data suggest that there are at least two types of eye movement in stomatopods: fast saccades and slower scans (Land et al. 1990; Cronin et al. 1991). However, it is important to remember that the so-called stomatopod saccades do not have a predictable time course, as they do in humans, but are much more variable in their velocity profiles. Since the eye is negatively buoyant and in a turbulent medium, the musculature must also fulfil a postural role, which implies very slow movements. Some of the optokinetic movements would also fall into this class. As there appears to be a pair of prime movers for each axis, one might reasonably expect there to be just three physiologically distinct types of fibre.

A classification based on the SDH test gives three classes of fibre with a range of oxidative capacities which would suit the posture/scan/saccade scenario. The matching of fibre types to movements in this way serves as a useful Aunt Sally from which to begin if one wishes to attempt a functional classification of fibres. However, the other tests give results which do not easily fit into this scheme and one is left with the conclusion either that there exist more types of fibre than there are eye movements or that a classification into discrete types is not really possible.

Of course, it must not be forgotten that the phenotypic identity of muscle fibres represents just one element of the neuromuscular system and so the movement spectrum cannot be attributed to them alone. Instead, it is clear that central and peripheral neuronal mechanisms, which have not been considered here, play an important role in the generation of variable movements (Govind and Atwood, 1982). This is especially true for animals, such as crustaceans, which have only a relatively small number of nerve and muscle cells at their disposal (Atwood, 1973).

Any single fibre can only have a fairly narrow working window over which it is ideally active and so a variety of interleaved types is required if a wider activity range is to be obtained (Rome et al. 1988). While it is possible to break stomatopod eye movements down into two populations, saccades and scans, these groups overlap and represent movements with a continuous, rather than a stepwise, distribution of velocities, sizes and durations. If the eyes were not able to move over a continuum, then there would be compensatory actions which they could not perform and targets which they would be unable to track.

The situation is thus rather different from that in locomotory systems. The gaited nature of locomotion means that muscular systems are often geared, with a few more or less exclusive fibre pools being used. Consider, for example, the red, pink and white muscle blocks of some fish (Johnston, 1981) or the two-gear three-speed system of Clione limacina (Satterlie et al. 1990). Even in geared systems, there is often a continuum of skeletal muscle fibre types present (Staron and Pette, 1988). However, this lack of precise phenotypic identity exists at the molecular level and one should be cautious of its value when attempting to correlate muscular physiology with behavioural function. That is not to say that the molecular differences are not crucial to the biology, but rather that it is not necessary to invoke them to generate a meaningful behavioural classification. The large number of fibre types in stomatopod eye muscles should therefore be seen as a reflection of the wide extent and variety of movements the eyes can make, rather than as a way of classifying different movement types.

I cannot thank Mike Land enough for the endless support, encouragement and sound advice which he has provided throughout the course of this work. Thanks also to Michael O’Shea, Justin Marshall and Douggie Neil for advice and discussion at various times. This work was supported by a grant from the SERC (UK).