Landmark-based morphometric analysis of the body shape of two sympatric , Ctenopharynx pictus and Otopharynx sp. "heterodon nankhumba" (Teleostei: Cichlidae), from Lake Malawi

Daud D. Kassam1*, Tetsu Sato2, and Kosaku Yamaoka1

1 Laboratory of Aquatic Ecology, United Graduate School of Agricultural Science, Ehime University, B 200 Monobe, Nankoku, Kochi 783-8502, Japan (e-mail: DDK, [email protected]) 2 WWF Japan, Nihonseimei Akabanebashi Building, 6 Fl., 3-1-14 Shiba, Minato-ku, Tokyo 105-0014, Japan

Received: May 17, 2001 / Revised: May 22, 2002 / Accepted: June 22, 2002

Abstract Morphological differences in body shape of two sympatric benthophagous species Ichthyological from Lake Malawi, Ctenopharynx pictus and an undescribed species, Otopharynx sp. “heterodon Research nankhumba,” were investigated using geometric morphometric methods. From digitized data of land­ mark points on lateral profiles of fishes, the shape of each species was compared by the thin-plate spline ©The Ichthyological Society of Japan 2002 method. Statistical analyses revealed significant variation in both uniform and nonuniform compo­ nents of shape between the two species. From the splines generated, it was revealed that most of the Ichthyol Res (2002) 49: 340-345 significant variation between the two species occurs in the head region. Specifically, C. pictus has a longer and deeper head than Otopharynx sp. In addition, the mouth of C. pictus is larger than that of Otopharynx sp. In the trunk region, C. pictus has a shorter abdominal cavity, which may indicate possession of shorter intestines than Otopharynx sp. The variation in gross head morphology and intestinal length may reflect interspecific differences in trophic ecology, possibly facilitating the coexi­ stence of the two species through resource partitioning.

Key words Geometric morphometric • Sympatric species • Ctenopharynx pictus • Otopharynx sp. “heterodon nankhumba” • Trophic specialization

he adaptive radiation and explosive speciation of However, not much research has been conducted on the the cichlid fishes of the Great Lakes of Africa (Lakes body shape of Lake Malawi’s , and most research has Victoria,T Tanganyika, and Malawi) has fascinated ecol­ focused on feeding apparatuses only (Reinthal, 1989, 1990). ogical, behavioral, and evolutionary biologists world­ Body shape has been investigated by the multivariate analy­ wide. Speciation in Lakes Malawi and Victoria is believed sis of distance measurements, the shortfall of this method to be relatively recent when compared to that of Lake being the restriction to the directions sampled by the mea­ Tanganyika. For example, Snoeks (2000) reported that Lake surement scheme (Bookstein, 1991). No study has yet exam­ Malawi contains about 800 cichlid species, but the lake itself ined the relationship between body form and feeding habit is reportedly less than 1 million years old (Meyer et al., in these cichlids using the novel landmark-based technique 1990). of geometric morphometrics (GM). Adaptive radiation of the cichlids of Lake Bookstein (1991) defined morphometrics as the statistical Malawi has resulted in the coexistence of extremely diverse study of biological shape and shape change. Morphometric assemblages of closely related species (Fryer and Iles, 1972). methods are needed whenever one wants to describe or Such adaptive radiation has mostly been attributed to mor­ to compare shapes of organisms or of particular structures. phological adaptations, in particular that of the feeding ap­ Such descriptions are useful in the understanding of growth, paratus (Fryer and Iles, 1972; Greenwood, 1974, 1981; Liem, experimental treatments, or evolution. Geometrically, the 1974; Barel et al., 1977; Witte, 1981; Yamaoka, 1987). shape of objects is studied well after removal of the effects Although several studies have shown that morphological of location, size, and orientation. Marcus et al. (1996) re­ differences in the feeding apparatus have directly contrib­ ported many applications of the GM method in different uted to the success of the cichlids (Fryer and Iles, 1972; biological fields including fisheries. GM data are the ho­ Greenwood, 1974, 1981; Liem, 1974; Barel et al., 1977), mologous landmark points (anatomic points with biological examining the whole organism’s body shape may contribute label) in either two-dimensional or three-dimensional to a greater understanding of species diversity. Body shape forms. GM has several merits compared to traditional is a highly variable characteristic among cichlid fishes and morphometrics (i.e., classical morphometrics of distance determines function, behavior, and habitat (Fryer and Iles, measurements), hence our choice of the former approach. 1972). First, the GM approach is much more effective in capturing Geometric morphometric study of Lake Malawi cichlids 341

Fig. 2. Landmarks collected from the left side of the fish. 1, tip of the premaxilla; 2, 3, anterior and posterior insertion of the dorsal fin; 4, 6, Fig. 1. The two species used: A Ctenopharynx pictus; B Otopharynx sp. upper and lower insertion of caudal fin; 5, posterior extremity of the “heterodon nankhumba” (Courtesy of Dr. Ad Konings) lateral line; 7, 8, posterior and anterior insertion of the anal fin; 9, insertion of the pelvic fin; 10, insertion of the operculum on the profile; 11, posterior extremity of the gape; 12, upper insertion of pectoral fin; 13, posterior extremity of the operculum information about the shape of an organism, as the geom­ etry of the whole organism is taken into consideration. The approach also has the capability to show shape variation Materials and Methods graphically, as deformations in a manner similar to D ’ Arcy Wentworth Thompson’s transformation grids (Thompson, Sample collection.—From September through November 1917), which is easier to interpret than the tables of numeri­ 1999, and in April 2000, 37 specimens of Ctenopharynx cal coefficients of the traditional morphometrics approach pictus fUsa Kochi University; UKU 387001001-387001037; (Rohlf and Marcus, 1993). 38.2-111.3mm standard length, SL) and 37 specimens The two study species (Fig. 1), Ctenopharynx pictus of Otopharynx sp. fUKU 387001038-387001075; 39.7­ Trewavas and Otopharynx sp. “heterodon nankhumba” 116.7 mm SL) were collected from West Thumbi Island in (sensu Konings, 1990), are both benthic feeders (Ribbink et the Cape Maclear region of Lake Malawi, East Africa al., 1983; Konings, 1990) and are sympatric (T. Sato, personal (14°00' S, 34°50' E). Specimens were collected by scuba observation). The breeding behavior of C. pictus is interest­ divers using hand nets and gill nets. Immediately after cap­ ing in that it relies on mutualistic association with the catfish ture, fish were killed and placed in 10% formalin solution. Bagrus meridionalis, locally called Kampango. Research has The body cavity of each specimen was also injected with the shown that predation of catfish fry has been significantly formalin solution. After fixation, specimens were trans­ reduced in broods that incorporate C. pictus fry, whereas ferred to 70% ethanol until examination. C. pictus fry benefit by feeding on the eggs produced by Morphometric and statistical analyses.—Images were adult catfish (McKaye, 1985). However, the presence of taken for each specimen using a digital camera (Fuji Finepix Otopharynx sp. at this locality needs to be further investi­ 500). Thirteen landmarks taken from the left lateral side of gated, as it is not clear whether this species is a parasite as each specimen (Fig. 2) were digitized using computer soft­ well. ware, TPSDIG version 1.19 (Rohlf, 1996). The two species look morphologically similar, except that All 74 specimens were superimposed using the general­ Otopharynx sp. has spots at the base of the dorsal fin. As ized least squares (GLS) method (Rohlf and Slice, 1990). both species are benthic feeders, there may be intense com­ This method removes nonbiological variation by scaling petition for food resources. Alternatively, differences in all specimens to unit size, translating them to a common morphology could have resulted in trophic specializations location, and rotating them so that their corresponding that enable the species to coexist. Neither detailed ecologi­ landmarks line up as closely as possible. From this superim­ cal nor morphological study has been done of these two position, the reference configuration was computed that was species to understand their relationship, hence our first at­ used in later analyses. The thin-plate spline algorithm was tempt. This study aims to describe variation in body shape used to calculate the uniform and nonuniform components between C. pictus and Otopharynx sp. by utilizing the GM of shape for each specimen (Bookstein, 1989, 1991). approach, and subsequently to explain the mechanism for Superimposition of specimens and the calculation of the coexistence of the two species. both uniform and nonuniform shape components were 342 D.D. Kassam et al. done using computer software TPSRELW version 1.20 anterad shift of landmark 1 versus the posteroventrad shift (Rohlf, 1997). To determine if shape varied significantly of landmark 11 (Fig. 3a,d). In contrast, Otopharynx sp. has a between species, a multivariate analysis of variance smaller mouth as shown by an opposite shift in the same (MANOVA, computed with JMP statistical package, ver­ landmarks, with landmark 1 shifting posterad against the sion 3.2; Sall et al., 1999) was performed on both the uni­ anterad shift of landmark 11 (Fig. 3b,e). form and nonuniform shape components. The predicted Another variation is evidenced through the comparison mean shapes of each species were then generated using of landmarks 1, 10, and 13 in both species. In C. pictus, these TPSRELW, by performing GLS on specimens within that landmarks move in opposite directions (i.e., landmark 1 species only. The mean shapes for each species were then anteriorly, 10 ventrally, 13 posteriorly), indicating C. pictus compared to the reference configuration to depict shape has a longer and deeper head (Fig. 3a,d) than Otopharynx differences between the species. sp. (Fig. 3b,e), in which such landmarks move against each On relative warp analysis, we set a scaling parameter, a = other. 0, as recommended for taxonomic studies (for more details The placement of landmark 8 (which is also the determi­ on a values, see Rohlf, 1993); this is because when a is 0, it nant of the anus, because in cichlids the anterior insertion of weights all partial warp scores equally so that neither large- the anal fin is very close to the anus) differs between the two scale nor small-scale differences are emphasized (Rohlf, species. In C. pictus, landmark 8 is shifted anterad whereas 1993; Adams and Funk, 1997). Principal component analysis in Otopharynx sp. it is posterad displaced. This result may was performed on a covariance matrix of the partial warp indicate that the anus is positioned more anteriorly in C. scores to summarize the variation in shape in as few dimen­ pictus than in Otopharynx. sp. sions as possible. Such analysis yields what is called relative The caudal peduncle region also shows variation between warps, of which the first relative warp (Bookstein, 1991) the two species. In C. pictus (Fig. 3a,d), the anterad displace- depicts the direction of the maximum nonuniform shape variation among all specimens (Rohlf, 1993). Intestinal morphology.—In addition to the landmark data, gross intestinal morphology was also analyzed. Indi­ vidual specimens were dissected and the intestines re­ moved. Intestine length (IL) was measured as the length from the esophagus to the anus. The relative gut length (R), which is the ratio of intestine length to standard length (SL, following Hubbs and Lagler, 1958), was calculated. The sta­ tistical difference in intestinal length was assessed by a t test on the means of the R value (using JMP statistical software, version 3.2; Sall et al., 1999). To determine the degree of complexity, a value for the intersecting point (IP) was calcu­ lated following Yamaoka (1985).

Results

Body shape morphometries. Body shape differences between the two species were highly significant for both uniform and nonuniform shape components (Table 1). From the splines generated (Fig. 3), based on both uniform and nonuniform shape components, the most pronounced shape variation between the two species occurs in the head region. Through inspection of the splines, it is clear that Ctenopharynx pictus has a larger mouth as evidenced by the

Table 1. Results on MANOVA, including all 74 specimens, for both uniform and nonuniform shape components

Source Wilk’s X Fs dfn dfd P

Uniform 0.4496 43.4536 2 71 <0.0001 Fig. 3 Visualizations of shape deformations for the two species: a Nonuniform 0.1781 12.2305 20 53 <0.0001 mean shape of Ctenopharynx pictus; b mean shape of Otopharynx sp.; MANOVA, multivariate analysis of variance; Wilk’s X, test statistic; Fs, e reference configuration. a and b are grid forms; d and e represent f ratio; dfn, degrees of freedom of the numerator; dfd, degrees of vector forms of a and b, respectively. Deformations are slightly exagger­ freedom of the denominator; P probability values ated to enhance contrast Geometric morphometric study of Lake Malawi cichlids 343

Fig. 5. The relationship between standard length and intersecting point value. Solid circles, Ctenopharynx pictus; open circles, Otopharynx sp. Fig. 4. Scatterplot depicting ordination of the 74 specimens on relative warps 1 and 2. The two relative warps accounted for 44.3% of the total variance in the sample. Solid circles, Ctenopharynx pictus; open circles, Otopharynx sp. Discussion

This study revealed substantial morphometric differences, ment of landmarks 3 and 7 versus the posterad shift of of which some have functional significance, between the landmarks 4 and 6 signifies a longer caudal peduncle than in two sympatric species Ctenopharynx pictus and Otopharynx Otopharynx sp. (Fig. 3b,e). sp. The differences are expressed at both higher and From the relative warp analysis, the first three relative lower spatial scales. Most of the localized shape changes warps explained about 57.5% of the total variation. The two occur in the head region. Compared to body shape, Fryer species can be easily discerned from the scatterplot of non­ and Iles (1972) observed that the head region is more uniform morphometric variation for all individuals along diverse, particularly the mouth, in cichlid fishes. This obser­ the first two relative warps (Fig. 4). Together these two vation is concordant with the results of the present study, warps accounted for 44.3% of the total variance in shape. because most of the pronounced variation between the two The first relative warp loaded highly for the horizontal study species is in the head region. This result may imply (x) component of nonuniform variation corresponding to that an examination of the trophic morphologies of many partial warps 4x, 7x-10x, and Ux (x component of uniform cichlid fishes may lead to a greater understanding of the variation). Loadings on the second relative warp were high­ diversity of the group. Such morphological diversity is a est for the vertical (y) component of nonuniform variation very important factor in adaptive radiation because indi­ corresponding to partial warps 2y, 7y, 9y, 10y, and warp Uy (y viduals can utilize all available feeding niches (Fryer and component of uniform variation). Although separate inter­ Iles, 1972). pretation of each partial warp is complex and not prefer­ Ctenopharynx pictus feeds on copepods that are sieved able, such patterns result from pronounced variation in the from the loose sediments by the long and numerous gill head and caudal peduncle regions, as evidenced from the rakers (Ribbink et al., 1983; Konings, 1990). The large gape splines generated. of C. pictus is a feeding-related adaptation that enables the Intestinal morphology. The t test performed on the fish to take a large volume of loose sediment into the mouth, means of relative gut length revealed that Otopharynx sp. in a manner similar to that of Cyathopharynx furcifer in has longer intestines than C.pictus (mean ± SD = 5.4 ± 0.5 Lake Tanganyika (Yamaoka, 1987) and other zooplankton for Otopharynx sp., and 2.5 ± 0.7 for C. pictus; t = 14.5, feeders (Fryer and Iles, 1972). In contrast, Otopharynx sp. df = 38, P < 0.0001). Correlated with such difference in has a small gape and feeds on other cichlids’ fry, as evidenced relative gut length are the intersecting point (IP) values by the presence of ctenoid scales in the stomach, as well as (Fig. 5). Ctenopharynx pictus has lower IP values than particulate matter including algae (D.D. Kassam, personal Otopharynx sp. with 11 as the mode and 17 as the maximum observation). According to Konings (1990), Otopharynx sp. value observed. In contrast, Otopharynx sp. has the higher does not dig into the sand when feeding, instead picking its IP values, with 19 as the mode and the maximum value food from rock crevices or the sand surface. Otopharynx sp. observed as 21. There seems to be no relationship between therefore does not need to open its mouth as wide as SL and IP value (Fig. 5), but this may be attributed to the Ctenopharynx pictus. In other words, a small gape is suited to fact that the specimens used in this study were all adults. It picking small fry and particulate matter individually, which is clear from such results that there are no species-specific is later processed by specialized oral and pharyngeal teeth IP values in these two species. (Kassam et al., in preparation). Similarly, it can be argued 344 D.D. Kassam et al. that the larger head of C. pictus provides a larger buccal Literature Cited cavity volume than that of Otopharynx sp. If Liem’s (1991) suction model of buccal cavity is applicable, then the larger Adams DA, Funk DL (1997) Morphometric inferences on sibling spe­ buccal cavity in C. pictus may imply possession of an attenu­ cies and sexual dimorphism in Neochlamisus bebbinae leaf beetles: ated truncated cone, which is a characteristic of suction multivariate applications of the thin-plate spline. Syst Biol 46:180­ feeders. Such a cone is capable of generating very large 194 negative pressures required for suction feeding to occur Barel CDN, Van Oijen MJP, Witte F, Witte-Maas ELM (1977) An (Liem, 1991). However, typical suction feeders tend to have introduction to the and morphology of the haplochromine a smaller gape (Liem, 1991), which is contrary to that which cichlidae from Lake Victoria. Neth J Zool 27:333-389 C. pictus has, and we suggest that this species is not a pure Bookstein FL (1989) Principal warps: thin-plate splines and the decom­ suction feeder as Ribbink et al. (1983) suggested, but rather position of deformations. IEEE Trans Patt Anal Mach Intell 11:567­ combines ram and suction feeding. 585 The anterad displacement of landmark 8 in C. pictus com­ Bookstein FL (1991) Morphometric tools for landmark data. Cambridge University Press, New York pared to its posterad displacement in Otopharynx sp. sug­ Fryer G, Iles TD (1972) The cichlid fishes of the Great Lakes of Africa: gests that C. pictus may have a smaller abdominal cavity. their biology and evolution. Oliver and Boyd, Edinburgh As this study did not detect any variation in body depth Greenwood PH (1974) Cichlid fishes of Lake Victoria, East Africa: the between the two species, it is assumed that with similar body biology and evolution of a species flock. Bull Brit Mus (Nat Hist) depths, the displacement of the anus may be used to predict Zool Suppl 6:1-134 a volume for intestinal coiling. This idea is concordant with Greenwood PH (1981) Species-flocks and explosive evolution. In: the observed intestinal length, for which C. pictus has a Greenwood PH, Forey Pl (eds) Chance, change and challenge— mean relative gut length (R) of 2.5 as opposed to the mod­ the evolving biosphere. Cambridge University Press and British erately longer intestines in Otopharynx sp. whose mean Museum, London, pp 61-74 relative gut length (R) is 5.4. Considering the correlation Hubbs CL, Lagler KF (1958) Fishes of the Great Lakes region, 2nd edn. between intestinal length and feeding habits or diet as University of Michigan Press, Ann Arbor shown by other researchers (Fryer and Iles, 1972; Konings A (1990) Ad Konings book of cichlids and all the other fishes Greenwood, 1974; Zihler, 1982; Yamaoka, 1985; Reinthal, of Lake Malawi. T.F.H. Publications, Neptune City, CA 1989), the shorter intestines possessed by C. pictus seem Liem KF (1974) Evolutionary strategies and morphological innova­ to correlate well with its carnivorous habit as opposed to tions: cichlid pharyngeal jaws. Syst Zool 22:425-441 the longer intestines in Otopharynx sp, which is regarded as Liem KF (1991) Functional morphology. In: Keenleyside MHA (ed) omnivorous. These findings are in agreement with findings Cichlid fishes: behaviour, ecology and evolution. Chapman and Hall, of other researchers, which if categorized from a broad per­ London, pp 129-150 Marcus LF, Corti M, Loy A, Naylor GJP, Slice DE (1996) Advances in spective, state that omnivores have longer intestines than morphometrics. Plenum, New York carnivores (Fryer and Iles, 1972; Greenwood, 1974; Zihler, Meyer A, Kocher TD, Basasibwaki P, Wilson AC (1990) Monophyletic 1982; Yamaoka, 1985; Reinthal, 1989). The variation in intes­ origin of Lake Victoria cichlid fishes suggested by mitochondrial tinal length among species with different trophic specializa­ DNA sequences. Nature (Lond) 347:550-553 tions is believed to depend on the nutritional value of the McKaye KR (1985) Cichlid-catfish mutualistic defense of young in food ingested (Zihler, 1982). Lake Malawi, Africa. Oecologia (Berl) 66: 358-363 From what has been revealed through this research, we Reinthal PN (1989) The gross intestine morphology of a group of rock- can conclude that shape variation, especially concerning dwelling cichlid fishes (Pisces, Teleostei) from Lake Malawi Neth J trophic morphology, enables the two species to exploit Zool 39:208-225 different trophic levels. A situation in which two or more Reinthal PN (1990) Morphological analyses of the neurocranium of a species with different trophic specializations coexist will group of rock-dwelling cichlid fishes (Cichlidae, Perciformes) from definitely result in reduced interspecific competition for Lake Malawi, Africa Zool J Linn Soc 98:123-139 resources, especially food resources, hence enhancing Ribbink AJ, Marsh BA, Marsh AC, Ribbink AC, Sharp BJ (1983) A successful coexistence. As this study has shown, GM, preliminary survey of the cichlid fishes of rocky habitats in Lake through its capability to detect even subtle shape differ­ Malawi. S Afr J Zool 18(3):149-310 ences, is a robust technique for quantifying morphological Rohlf FJ (1993) Relative warp analysis and an example of its variation between C. pictus and Otopharynx sp. This mor­ application to mosquito wings. In: Marcus LF, Bello E, Garcia- phological variation may play a vital role in the coexistence Valdecasas A (eds) Contributions to morphometrics, vol 8. Mono- grafias del Museo Nacional de Ciencias Naturales (CSIC), Madrid, of the two species. pp 131-159 Rohlf FJ (1996) TPSDIG, version 1.19. Geometric morphometric soft­ Acknowledgments We thank Dr. Dean Adams for his constructive ware for the PC. http://www.life.bio.sunysb.edu/morph/software.html ideas on the GM methodology and the manuscript itself. Katherine Rohlf FJ (1997) TPSRELW, version 1.20. Geometric morphometric Martin, Dr. Andrew Rossiter, and Dr. Juanito Dasilao deserve our software for the PC. http://www.life.bio.sunysb.edu/morph/ heartfelt thanks also for their criticisms and good suggestions on this software.html manuscript. The Malawi government through its Fisheries Department Rohlf FJ, Slice DE (1990) Extensions of the procrustes method for the deserve also our thankfulness for permitting us to collect samples from optimal superimposition of landmarks. Syst Zool 39:40-59 Lake Malawi. Many thanks to Dr. Ad Konings who provided us with Rohlf FJ, Marcus LF (1993) A revolution in morphometrics. Trends images for the two species. Ecol Evol 8:129-132 Geometric morphometric study of Lake Malawi cichlids 345

Sall J, Ng K, Hecht M, Tilley D, Potter R, Corcoran B (1999) JMP Yamaoka K (1985) Intestinal coiling pattern in the epilithic algal- statistical software. SAS Institute, Cary, NC feeding cichlids (Pisces, Teleostei) of Lake Tanganyika, and its Snoeks J (2000) How well known is the ichthyodiversity of the large phylogenetic significance. Zool J Linn Soc 84(3):235-261 East African lakes? Adv Ecol Res 31:17-38 Yamaoka K (1987) Comparative osteology of the jaw of algal-feeding Thompson DW (1917) On growth and form. Cambridge, London cichlids (Pisces, Teleostei) from Lake Tanganyika. Rep Usa Mar Biol Witte F (1981) Initial results of the ecological survey of haplochromine Inst Kochi Univ 9:87-137 cichlid fishes from the Mwanza Gulf of Lake Victoria (Tanzania): Zihler F (1982) Gross morphology of digestive tracts of cichlidae breeding pattern, trophic and species distribution. Neth J Zool (Teleostei, Perciformes): phylogenetic and functional significance. 31(1):175-202 Neth J Zool 32:544-571