Joological Journal of.the Linnean Society (1984),80: 4 2 1 4 3 5 . With 6 figures Evolutionary relationships within three Tilapiine genera (Pisces: Cichlidae) B. J. McANDREW AND K. C. MAJUMDAR Institute o f Aquaculture, University OJ' Stirling, Stirling FK9 4 L A , Scotland Recezued March 1983, accepted Jar Publzcation August 1983 Species of thc cichlid genera Tilapia, Sarotherodon and Oreochromis were studied by electrophoresis at 25 enzyme loci. T h e levels of similarities within and between genera, and the distribution of genetic identities at individual loci are similar to those found in other fish species. T h e evolutionary relationships between thc species remains equivocal, and it is still not possible to decide between the hypotheses of Trewavas (1980) or that of Peters & Berns (1978, 1982). Further work needs to be undertaken on additional Tilapia and Sarotherodon species. The unexpected relationship of 0.jipe to the other Oreochromis species is probably the strongest argument against Trewavas' hypothesis. But the close similarity of the maternal mouth brooders to each other, and the closer relationship of S. galilaeus to the Oreochromis species than to the Tzlapza does not favour the hypothesis of Peters & Berns (1982). T h e difficulties in obtaining pure species from the wild due to widespread introduction of non-endemic species are likely to hinder progress in evolutionary and aquacultural studies. KEY WORDS:--Evolution Oreochromis ~ Sarotherodon - electrophoresis Tzlapia. - - dendograms - genetic distance enzymes - CONTENTS Introduction . . . . . . . . . . . . . . Materials and methods. . . . . . . . . . . . Results . . . . . . . . . . . . . . . . Genetic identity ( I n ) and genetic distance (Dn) . . . . . Unweighted pair group method and arithmetic means (UPGMA) . . . . . . . . Fitch & Margoliash method . . . Wagner trees of Farris trees using frequency coded data . Conventional Wagner procedure using binary coded data . . . . . . . . . . . . . . . Discussion. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 I 423 425 425 425 427 427 428 429 433 434 INTRODUCTION In recent years one group of fish within the Cichlidae has been receiving much renewed attention. The tilapias are a group of species which have immense potential in tropical aquaculture, and from their natural range within Africa they have been transplanted so that one species or another is to be found in most tropical and subtropical countries of the world. Despite the wide environmental tolerances of various species, the 70 species show relatively little + 0024-4082/84/04042 1 + I5 $03 00/0 4'2 I 01984 The Linnean Soctet\ of London 422 B. J. McANDREW AND K. C . MAJUMDAR morphological diversity and this has led to much confusion in species identification (McAndrew & Majumdar, 1983) and their phylogenetic rela tionships. The whole group was initially placed in the genus Tilapia but since the 1920s there have been numerous attempts by taxonomists to divide them into a number of smaller groups. (Regan, 1920; Thys van den Audenaerde 1968, 1971; Trewavas, 1973, 1980, 1982; Peters & Berns, 1978, 1982). Initially these divisions were based on structural characteristics such as the number of gill rakers, jaw structure and the teeth on the pharyngeal bone. More recently as information about the fishes’ biology has become known, subdivisions have been based on differences in breeding and brooding behaviour. The most recent classification by Trewavas (1982) divides the tilapias into three genera: Tilapia which consists of substrate spawners, Sarotherodon which includes paternal and biparental mouthbrooders and Oreochromis which is restricted to maternal mouthbrooders, the latter genus being divided into four subgenera. This classification follows that given by Thys van den Audenaerde (1968) which gave a different rank to the groups, leaving Tilapia as the genus and Sarotherodon and Oreochromis as subgenera. These ideas clash with those of Peters & Berns (1978, 1982) who believe that any of these subdivisions are not justified and that the various forms should all be called Tilapia based on their ideas of the evolution of the group. It is generally agreed that mouthbrooders evolved from a substrate-spawning ancestor. Peters & Berns (1978, 1982) (Fig. 1A) suggest that substrate spawners have given off mouthbrooding branches from time to time, and it is the time since this branch occurred which determines the state of advancement of mouthbrooding characteristics. The most ancient branch (‘older’ mouthbrooders) are represented by the maternal mouthbrooders, compared to the more primitive breeding characteristics of the more recently branched ‘younger’ mouthbrooders. Trewavas ( 1980) puts forward an alternative hypothesis (Fig. 1B) in which a Tilajia-like ancestor gave rise to a mouthbrooding branch which itself quickly divided into two: one which remained conservative regarding breeding behaviour but not other characteristics (Sarotherodon) and a more progressive branch which led to the Oreochromis. I n recent years data on molecular variation have proved very useful in tackling systematic problems and it is generally found that phylogenetic trees derived from molecular data are similar to those derived from anatomical and morphological characteristics (Mickevich & Johnson, 1976). A number of workers have used molecular data on the tilapiines as an aid to species identification (Chen & Tsuyuki, 1970; Avtalion, Pruginin & Rothbard, 1975; Avtalion et al., 1976; McAndrew & Majumdar, 1983). The only previous study which has compared evolutionary relationships in this group was by Kornfield et al. (1979) which confirmed the broad taxonomic relationships derived from morphological data, although only utilizing one species from each of the three genera. In this paper we will consider the evolutionary trees produced by three different dendrogram construction methods, using 25 enzyme loci in nine different tilapiine species. At least one species from each of the three genera proposed by Trewavas (1982) are included. The results are discussed with 427 TILAPIINE EVOLUTIONARY RELATIONSHIPS Tflapfa Substrate spawners Genus Mput h brooders ‘Younger Older’ Tilapa Sarotherodon Oreochromis Substrate spawners Paternal and biparental Maternal mouthbrooders Figure IA, B. Diagrams to show the differences between the hypotheses of Peters & Berns I 1978, 1982) and ‘Irrwavas (1980). ‘The diagrams have been modified from those of Trewavas (1980’ to include the latest generic status of the different mouthbrooders (Trewavas 1982). reference to similar studies on other groups of organisms. Attention is paid to the way these results compare with existing ideas about evolution within this group. MATERIALS AND M E T H O D S T h e preparative procedures and electrophoretic techniques are given by McAndrew & Majumdar (1983). Table 1 gives the species, their sources a n d denotes populations which were sympatric. T h e allele frequency d a t a a n d relative mobilities of the loci studied have already been published by McAndrew & Majumdar ( 1983). Genetic identity and distance was calculated from the allele frequencies using the formulae given by Nei (1972). Both the genetic identity (In) and genetic are given in Table 2. Standard errors for Dn are also given (Nei, distance (Dn) Table 1 . T h e table shows the species studied a n d their sources and denotes populations which were sympatric. Species with (Bamburi) after their name came from the collection a t Baobab Farm, Bamburi (see Acknowledgements) -~ Speciri ~~~~ Source ~~~~~ ‘Tilapia zillii .Caratherodon galilaeus Oreoihromi~Subgcnus Oreoihromis andersonii Oreochromis (O., Oreoihromis ( 0 . aureus Oreochromis ( 0 .j j i p e Oreochromis 0. mossambicus Oreoihromis O., niloltcuJ Oreochromis ( 0 ., spilurus OreochromiJ Xjasalapia Oreochromis (J+’. , macrochir $8 Egypt (Lake Manzala) Kenya (Lake Turkana) (Bamburi Botswdna Okavango) (Bamburi Egypt (Lake Manzala) Kenya (Lake Jipe) (Barnbun) Aquanst stock Egypt (Lake Manzala) Kenya ( R T a n a ) (Bamburi) Botswana [Okavango) (Bamburi * t t * * B. J. McANDREW AND K. C. MAJUMDAR 424 Table 2. Genetic distances (on) and their standard errors (below diagonal); genetic identity (In) (above diagonal) for nine tilapiine species analysed for 25 enzyme loci 0 . ander. 0 . andersonii 0. aureur S.galilaeus 0 .jipe 0. macrochir 0 . mossambicus 0. niloticur 0. spilurur T. zillii 0 . aur. - 0.297 f0.119 0.429 f0.149 0.931 +0.253 0.350 50.132 0.171 f0.084 0.262 k0.108 0.311 k0.123 0.641 +0.194 0.732 - 0.304 fO.121 0.651 f0.194 0.249 f0.107 0.263 fO.111 0.115 f0.068 0.152 - 0.080 0.542 +0.173 + S . galil. 0.651 0.738 ~ 0.671 0 .jipe 0 . macro. 0 . moss. 0 . nil. 0 . spil. 7.zillii 0.394 0.521 0.705 0.780 0.873 0.769 0.768 0.891 0.732 0.859 0.526 0.582 0.511 0.768 0.687 0.806 0.768 0.532 0.517 0.429 0.564 0.517 0.391 0.768 0.843 0.897 0.566 - 0.799 0.784 0.486 - 0.925 0.668 - 0.647 0.436 k0.151 - ~ f0.199 0.263 0.659 ~ kO.111 f0.196 0.375 f0.137 0.216 f0.098 0.264 +0.111 0.631 +0.192 0.846 0.262 +0.234 f0.109 0.572 f0.178 0.660 k0.197 0.938 +0.254 0.171 0.224 0.085 f 0.100 0.108 0.243 k0.066 f0.103 0.569 0.721 f0.179 fO.210 0.077 50.052 0.403 f0.142 1971). Three different methods were used to produce dendrograms from the electrophoretic results. This was because there is still some doubt as to the best procedure for this type of data (Prager & Wilson, 1978; Felsenstein, 1982). The three different methods were the unweighted pair group arithmetic average clustering method (UPGMA), the phylogenetic tree construction method of Fitch & Margoliash (1967) (F-M) and the conventional Wagner procedure after Farris (1970). The UPGMA utilized the distance matrix in Table 2 and was calculated using the Clustan B computer package. The UPGMA method does not indicate the amount of change along individual lineages and assumes homogeneity of rates of evolution among the lineages compared. The F-M procedure was calculated with a programmable calculator. The choice of the best tree from those produced was the one with the least difference from the input data. The tree chosen was that with the smallest F value (Prager & Wilson, 1978). F is defined as F= lOOf/I where f is the sum of the absolute value of the differences between output and input values and I is the sum of the input values. The Wagner procedure of Farris (1970) utilized the original allele frequency data (McAndrew & Majumdar, 1983) both binary coded (if allele present equals 1, if absent equals 0) as well as frequency coding to one decimal place. The allele frequency data were used in preference to the distance matrix because it was felt that more systematic information would be retained. The Wagner algorithm works on the principle that the best estimate of a monophyletic group is the tree with the smallest number of character transformations, and like the FM procedure does not assume homogeneous rates of evolution in the lineages concerned. The Wagner trees were calculated using the Wagner procedure p r o g r a m (WAGPROC) based on Swofford ( 1981) . This program produces an undirected Wagner network which is rooted at the mid point of the greatest patristic distance, and of the minimum evolutionary length. TILAPIINE EVOLUTIONARY RELATIONSHIPS 425 60- 50- 0 0:2 0.'4 1.0 Genetic identity Figure 2. Histogram showing the percentage of loci within a given range of genctir identity among nine species from three tilapiine genera (Trewavas, 1982). RESUL'I'S Genetic identity ( I n ) and genetic distance ( D n ) The calculated values of In and Dn with standard errors are given in Table 2. The values for Dn will be used in discussion where possible as these allow an estimate of the level of error. The most similar pair are 0. niloticus and 0. spilurus (Dn=0.077f0.056) and the least similar are 0 . j i p e and 7.zillii ( D n= 0.938 f0.254) although the values overlap between all other 0.jipe comparisons. In general, the Dn increased with increasing rank. The mean Dn within the Oreochromis was Dn=0.369 including 0 . j i p e but only Dn=0.217 if 0.j;Pe is removed. Average intergeneric differences in Dn between Oreochromis and Sarotherodon were Dn = 0.360, between Oreochromis and Tilapia it was Dn = 0.607. The difference between Sarotherodon and Tilapia was Dn = 0.63 1. The average Dn between the three genera was Dn = 0.533. The distribution of single locus In can be seen in Fig. 2 and is the typical Ushaped pattern which has been seen by many other such surveys (Avise & Smith, 1977). The majority of the loci studied (89.5'1;) have either identical allelic composition or are else completely distinct with unique alleles. I;nu,eighted pair group method with arithmetic means (C'PGMA) The product of the UPGMA agglomerative clustering procedure (Fig. 3 ) is based on the Dn measures in Table 2, and the dendrogram produced shows the majority of the mouthbrooders clumped together. However, the position of 0. jipe is totally unexpected being more remote from the other mouthbrooders than even 7. zillii. Similarly, the position of S. ga1ilaeu.r amongst the B. J. McANDREW AND K. C. MAJUMDAR 426 3 Nei: Gormaneta/: I I 15 2 I I I I Years x lo6 I 0m e TI///// S go/i/aeus 0 macrochir 0 spilurus 0 nl/ot/cus 0 oureus 0 rnossambicus 0 andersonii I 0.0 I 0.6 I 0.4 I 0.2 I 0 Genetic distance Figure 3. Phenogram showing relationships of nine species from three tilapiine genera (Trewavas, 1982) according to UPGMA method of cluster analysis using species distance matrix Table 2, with two different estimates of evolutionary time. mouthbrooders is unexpected if the Sarotherodon are held to be ancestral to the maternal mouth-brooding species. The scale on this Fig. 3 is in Dn and if one assumes a reasonably homogeneous evolutionary rate within the species, it may also be used as an evolutionary time scale. There are two different estimates as to the time scale which is most appropriate; a Dn value of 1 is equivalent to 5 x lo6 years (Nei, 1975) whereas other workers believe it to approximate to 18 x lo6 years (Gorman, Kim & Rubinoff, 1976). Both scales have been included in the figure but the latter seems to have proved to be close to the known geological L 0.5 I I 0.4 0.3 I 0.2 I 0.1 I 0 Average genetic distance Figure 4. Phenogram showing relationships of nine species from three tilapiine genera (Trewavas, 1982) according to the Fitch-Margoliash phylogenetic tree construction procedure using species distance matrix (Table 2). Each apex point is placed a t a n ordinate value representing the average of the sums of genetic distances in the lines from that apex. 'I'ILAPIINE EVOLUTIONARY RELATIONSHIPS 42 7 events in previous studies (Soule & Gorman, 1974; Gorman et al., 1976; Vawter, Rosenblatt & Gorman, 1980). Fitch & Margolaash method This tree (Fig. 4) was selected from 20 others as the one with the lowest Fvalue. In general the tree is very similar to the UPGMA and Wagner trees except that S. galilaeus is ancestral to the majority of the mouthbrooders. Olaochromis jipe is still very different to the other mouthbrooders and is further away than T. Zillii. Wugner trees or Farris trees using frequency coded data The two trees of the same minimum length were (Fig. 5A,B) produced by using the allele frequency data and both have the same minimum length and 0 andersonii a 0 mossarnbicus S galilaeus 0 rnacrochir 0 spilurus 0 aureas L 0niloflcus 0 andersonii 0 rnossamhicus 0macrochir 0 spilurus m, 0 aureus 0niloficus T ziJli/ CL 13.14 L 8.27 12.41 16.55 O/lm Distance from root Figure 5, A, B. Phenograrns showing relationship of nine species from three tilapiine gencra (Trewavas, 1982 according to Wagner tree procedure of Farris (19701. These trees have been constructed using the allelr frequency data from McAndrew & Majumdar ' 1983 . B. J. McANDREW AND K. C. MAJUMDAR 428 Q ondersonii 0.mossombicus Ad- 0 mocrochir 0 spilurus 0 oureus 0 nifoticus I 0. Iipe 0.ondersonii B 0 mossombicus 0 mocrochir 0 spilurus 0 oureus rl C S golifoeus I 0.nifoticus ’ - , 0 ondersonii 0.mossombicus 0 mocrochir 0 spifurus S.gofifaeus ri 0 oureus 0 niloficus T zifhi 0 jipe Distance from root Figure 6A, B, C. Phenograms showing relationships of nine species from three tilapiine genera (Trewavas, 1982) according to the Wagner tree procedure of Farris (1970). These trees have been constructed using the allele frequency data from McAndrew & Majumdar (1983) coded in a binary fashion. goodness of fit. The trees produced differ in the relative positions of S. galilaeus to the 0. spilurus, 0. macrochir branch. Conuentional Wagner procedure using binary coded data In all, six trees of the same minimum length and goodness of fit were produced. Only three have been reproduced, Figs 6A, B & C since the other ‘I‘ILAPIINE EVOLUTIONARY RELATIONSHIPS 129 trees showed only minor positional changes of some of the arms. One tree had a branching pattern and arm lengths for all species identical to the tree in Fig. 6A except that the positions of 0. aureus and 0. niloticus were reversed. The tree in Fig. 6B had a variant in which the 0. aureus and S. galilaeus branches changed position with 0. niloticus. The final tree had the positions of 0. aureus and 0. niloticus reversed in comparison with the tree shown in Fig. 6C. In general the two Wagner methods give similar results in that it is the position of S. galilaeus within the mouthbrooders which is being varied. Only one tree, Fig. 6A, shows S. galilaeus ancestral to the maternal mouthbrooders other than 0.jipe. I n the frequency analysis 0. aureus and 0 . niloticus are always ancestral to S. galilaeus and it is the position of the 0. macrochir, 0. spilurus branch that changes position with S. galilaeus. I n the binary coded data (apart from the tree in Fig. 6A) S. galilaeus is somewhere close to 0. aureus and 0 . niloticus and always ancestral to the 0. macrochir, 0. spilurus pair. DISCUSSION It is generally considered that phylogenetic trees based on morphological and allozyme characters produce similar results, despite the fact that the underlying genetic bases for the two sets of characters are likely to be quite different. T h e relative ease and reliability of electrophoresis for assessing the levels of intraspecific and inter-specific variation has meant that i t has been used widely in such studies. However, the technique does have a number of limitations. Electrophoresis will always underestimate differences since only a percentage (27%) of the codon changes will actually produce detectable differences (King & Wilson, 1975). The number of loci used in systematic studies is also very important. The U-shaped distribution of the percentage of loci with a given In in Fig. 2 clearly shows that the majority of the loci are in fact identical or unique for the different species. The main implication of this is that the validity of the systematic information lies in the number of loci studied. More information will be available from increasing the number of loci rather than the number of individuals within a species. Similarly increasing the number of loci will also decrease the size of the standard error associated with calculating Nei’s coefficients and lead to a more accurate interpretation (Nei, 1971; Thorpe, 1982). T h e U-shaped distribution is very similar to others produced from studies on other groups of sibling species, species, and genera, and appears to be a common pattern for most outcrossing sexual organisms. The levels of In and D n both within and between genera are similar to other studies of fishes (Kirpichnikov, 1981: 237). The only other evolutionary study on tilapiines was by Kornfield et al. (1979) which studied a number of cichlids endemic to the Sea of Galilee. Kornfield et al. ( 1 9 7 9 ) obtained In=O.92 between 0. aureus and S. galilaeus, In = 0.48 between T. zillii and S. galilaeus and In = 0.49 between T. Zillii and 0 . aureus. These compare with In=O.74, In=O.52 and In = 0.58 respectively in this study. These two independently derived sets of similarities are interesting in themselves since they underline the variation inherent in using electrophoresis as a systematic tool. The differences are probably a result of the different loci compared in the two studies. Sarich (1977) 430 B. J . McANDREW AND K. C. MAJUMDAR noted that the choice of loci can dramatically affect the estimate of similarity. Comparisons between loci common to both studies will be informative and may show the level of divergence which has occurred since these populations of tilapia have become reproductively isolated. Of the 12 loci which can be identified as common between the two sets of data, five show some difference between one or all of the species involved. Sarotherodon galilaeus is polymorphic at the Pgi-1 & 2 loci but is fixed for the more common allele at both Pgi loci and a faster 6 Pgdh allele in the study of Kornfield et al. (1979). Tiliapia eillii is fixed for a slower Ak allele and appears to be fixed for a different Est-2 allele in Kornfield’s study. Finally 0. aureus is fixed for the more common of the two Ak alleles which Kornfield et al. (1979) found in their study. A number of differences occur at the Est-2 loci; it appears that the Israeli population of S. galilaeus and 0. aureus have more alleles a t this loci but no direct comparisons can be drawn as no standard exists between the two sets of data. If an Zn value is calculated for each of the three species based on the common loci, values of Zn=O.99, 0.96 and 0.92 are found for 0. aureus, S. galilaeus and 7. zillii respectively. These values are interesting and are similar to the differences which have been found between populations in other fish species studied (Kirpichnikov, 1981). However, these results must be treated with some caution because of the inherent errors resulting from the small number of loci and possible differences in the electrophoretic techniques in the two laboratories. The results do show that some divergence has occurred since these populations were isolated when changes in the River Nile drainage caused Lake Turkana and the northern extension of the rift valley into Israel and Syria to be left with no direct contact with the main river. The construction of dendrograms from electrophoretic and sequencing data has been receiving increasing attention. However, there is still some doubt as to the best method of analysing such data. Prager & Wilson (1978) have shown that there are distortions caused by the different methods and suggested that the Fitch-Margoliash method was the best choice to use with electrophoretic distance data. The UPGMA method was said to be on average better than the Farris method but not quite as good as the F-M method. The different methods also make different assumptions about the homogeneity of evolutionary rates within individual lines. The F-M and Farris methods, although different procedures, make no assumption about the rate of evolution within any line, whereas the UPGMA method produces a phenogram which can only be assumed to represent an evolutionary tree if the rate of evolution for the different proteins is constant in all lineages. More recently Felsenstein (1982) suggested that the different methods for inferring phylogenies should be evaluated using statistical criteria, as this might give a clearer impression of how accurately these methods estimated the phylogeny. The various methods make different sets of assumptions concerning the characters being used. Felsenstein ( 1982) thought that a standard statistical approach such as maximum likelihood would produce a method, particularly for electrophoretic data, whose properties are known and which would give some degree of statistical confidence. Despite this, little work has been done to develop statistical models which resect the complexity of the evolutionary processes. Because of the doubts on their relevance, all three of the commonly-used TILAPI INE EVOLUTIONARY RELATIONSHIPS 43 1 dendrogram construction techniques have been used in this study, and in general they show very similar patterns of evolution. Some species, 0.j;Pe and T. zillii and the 0. andersonii and 0. mossambicus pairing maintain their relative position no matter which method is used. T h e majority of the differences between the various methods occur with the relative grouping within Oreochromis and the relationship of S. galilaeus to this genera. It is generally agreed that mouthbrooding species have evolved from substrate spawning ancestors. Sarotherodon species also exhibit characteristics intermediate between substrate spawners and maternal mouthbrooders, particularly in their reproductive behaviour and in the regression of substrate-spawning characteristics in their larvae such as the adhesive layer on the eggs and adhesive glands on the larval head. (Peters, 1965; Peters & Berns, 1978, 1982). However Peters & Berns (1982) believe that mouthbrooding has occurred a number of times from substrate spawners possibly from different ancestors and at different times. These authors demonstrate that in breeding characters today’s Sarotherodons are closer to the Tilapia than the Oreochromis and propose that Sarotherodons have only recently split from the Tilapia ancestor. Trewavas (1980) put forward an alternative theory in which Tilapia gave rise to a mouthbrooding branch which soon divided into two: a more conservative branch, leading to the ,Sarotherodon, and a progressive one leading to the Oreochromis. I t should be possible to differentiate between the two hypotheses. From Peters & Berns (1978) explanation one might expect S. galilaeus to be closer to the substrate spawning T. zillii, although this may not be its direct ancestor. Peters & Berns (1978) hypothesis does not anticipate that all maternal mouthbrooders will be closely related as they may have evolved from different ancestors or even from the same ancestor at different times. This should be evident in the electrophoretic analysis because once reproductive isolation has taken place, mutations accumulate at a regular if stochastic rate. Selection may act on some of these new mutations but is unlikely to cause any major distortions when a large number of loci are studied. The Peters & Berns hypothesis would predict larger differences between the mouthbrooders than would be expected by a monophyletic model of evolution. The lack of availability of more species of Tzlapia and Sarotherodon does not make comparisons within the genera possible. However, it appears that S. galilaeus is much closer to the majority of Oreochromis species than it is to 7.. zillii. This might suggest that Peters & Berns [ 1978) hypothesis is n o t correct in this case. The position of 0 . j i p e in all the various trees is unexpected as it is known to be a maternal mouthbrooder. The fact that 0 . j i p e is so different might suggest that this species has evolved from a different ancestral substrate spawner, and may be a candidate for one of Peters’ ‘Older mouthbrooders’. O . j $ e was first described by Lowe (1955) as one of three new species (and one subspecies) of the 0. mossambicus complex from Lake Jipe and the Upper Pangani River. Oreochromisjipe was morphologically similar to one of the other species 0. girigan but had quite different pharyngeal teeth, more similar to S. galilaeus than to the ‘mossambicus’ complex in which these other species were thought to belong. Trewavas (pers. comm. 1983) considers 0. hunteri, 0.j p e , 0. girigan, and 0. pangani to be a little group apart from the rest of the Oreochromis in that they have a higher vertebral number (31-34, mode 33) as compared with the rest of 43 2 B. J. McANDREW AND K. C. MAJUMDAR the genera (27-32, mode 30), and a characteristic body coloration. The extreme position of 0. jipe would not be expected from Trewavas’ ( 1980) hypothesis and could be explained by this species having become isolated at an early stage from the other mouthbrooders, with local conditions causing the accumulation of a large number of unique alleles within this species, but there are no apparent reasons to suspect that this might be the case. Hybridization is well known amongst the tilapiines. The differences in 0.jipe could well be the result of a combination of new alleles from a number of species. The majority of the differences between 0.jipe and the other Oreochromis are a t uniquely fixed enzyme loci. The heterozygosity of 0. jipe is similarly only the second highest within the species studied. This would not suggest a recent hybridization event with anything but a very closely related species. A number of authors have noted that hybrid populations often contain alleles not present in either of the parental populations (Hunt & Selander, 1973; Sage & Selander, 1979). Two mechanisms have been proposed to explain the presence of these new alleles. The first is that there might be an increased mutation rate in the hybrid (Thompson & Woodruff, 1978) or that they might result from intragenic recombination between different alleles of the parental populations (Watt, 1972; Morgan & Strobeck, 1979; Tsuno, 1981). Recently Golding & Strobeck (1983) have produced a model which shows that intragenic recombination at heterozygous loci can generate new unique alleles in such hybrid populations. However, this is likely to increase the level of heterozygosity still further in the population. It is possible that any increase in heterozygosity has been subsequently reduced because of the loss of alleles due to selection and or allelic repression (see Whitt, Philipp & Childers, 1973, 1977). There is little evidence to show that 0.j;Pe is of hybrid origin. The fact that 0 . j i p e is polymorphic for anal spine number (3 or 4) (Lowe, 1955) is unusual amongst the tilapias; the only other species which show this are 0. hunteri from Lake Chala and 0. korogwe from the Pangani river. However, this is in no way confirmatory that any of these species are of a hybrid origin. It is obvious that a much more detailed study of 0 . j i p e and its closely related species is needed to determine whether they are a discrete group apart from the other Oreochromis or whether O.jt$e is an unusual case caused by some hybridization event with an unknown species. These facts will be needed before any firm conclusions about the relationship of 0.jipe and its close relatives with the other Oreochromis are formed. The overall picture from the various trees must be treated with some caution, because the trees produced cannot give an exact picture since many species are missing from the analysis. It is obvious that some species essential for critical consideration are missing. These omissions will lead to differences between the trees particularly when each analysis used the data in a slightly different way. The small number of species considered will affect the results for the ‘mossambicus’ complex because only two close members of this group are present, 0. mossambicus and 0. spilurus, and a more distant relative, 0. andersonii. This complex is endemic to the rivers of the East African coast and comprises a number of localized populations from 0. spilurus (Kenya) in the north to 0. mossambicus (South Africa) in the south. It is thought that the fish have moved along the coast from estuary to estuary especially during times of flood. TILAPIINE EVOLUTIONARY RELATIONSHIPS $33 The apparent differences between 0. mossambicus and 0. spilurus could well be caused by the fact that they are at opposite ends of a range. The differences could also be Iarger than is actually the case in the wild because the 0. mossambicus stock was very inbred, coming from aquarist stock, and was monomorphic for most loci. This could have decreased similarities because any polymorphic loci may have had shared alleles. Similarly, the alleles fixed in this population may not all be the common wild allele, because sampling error and bottlenecking of the populations in captivity may have increased the frequency of less common alleles so that some of them became fixed. The geological history of this group of fish is little known. Trewavas (1937) described a fossil Tilapiine with four anal spines from a Pleistocene deposit (2.5 million years old) at Rawe near Lake Victoria and concluded that it belonged to 0. spilurus niger. Greenwood (1951) discussed a fossil tilapiine with four anal spines from Miocene deposits (7-26 million years ago) and considered they were most closely related to 0. mossambicus. This tends to suggest that relatively little morphological change has occurred since these times. Figure 3 shows the UPGMA dendrogram and two estimates of evolutionary time. The estimate of Gorman et al. 1’1976) (Dn= 1 equivalent to 18 million years) seems to be most appropriate as it appears to be backed by some geological evidence (Yang et al., 1974; Gorman et al., 1976; l’awter et al., 1980). Using this time scale it would appear that the split between Tilapia and the mouthbrooders occurred some 10 million years ago and that the main speciation within the maternal mouthbrooders occurred around 5 million years ago. These estimates seem conservative when the fossil record already shows a mouthbrooding species as long ago as 7 million years. Indeed the use of such estimates of evolutionary time are fraught with many errors. Temporal divergence estimates assume that one is dealing with a constant linear function. Corruccini et al. (1980) have demonstrated that the relationship between Dn and time is distinctly non-linear. The errors involved in this type of comparison are probably too great to draw any meaningful conclusions, particularly as there are no known geological events to act as an independent estimate of evolutionary time. It is obvious that many more species need to be analysed before a reliable evolutionary tree within this group can be developed. However, the ability to collect pure samples of any tilapiine species is becoming more and more difficult as uncontrolled introductions of new species and man-made disturbances lead to a widespread introgression of new genes into pure tilapiine stocks. This has obvious consequences for taxonomic and evolutionary studies, in that it will become progressively more difficult to separate endemic pure species from hybrid populations. I t also has a profound long-term effect on man because unique gene pools are being lost before anyone can identify their potential for exploitation directly or as a source of new variation within existing farmed stocks. ACKNOWLEDGEMENTS The authors would like to express their thanks to all colleagues who assisted in the collection of the different species. These include Dr I. Payne and M r J. Balarin, and particular gratitude is expressed to M r R. Haller of Baobab Farm, Bamburi, Mombasa, Kenya, who supplied a number of species from his collection (Table 1 ). 20 434 B. J. McANDREW AND K. C. MAJUMDAR Dr E. Trewavas gave useful advice on suitable areas for collection of pure species as well as making constructive comments on the manuscript. The authors would also like to thank I. Kornfield, R. D. Ward, G. Hulata and A. Abreu-Grois for their useful comments on an early draft of this manuscript. The work was supported at Stirling by the Overseas Development Administration. Kshitish Majumdar was also funded by the Commonwealth Scholarship Commission. REFERENCES AVISE, J. C. & SMITH, M. H., 1977. Gene frequency comparisons between sunfish (Centrarchidae) populations at various stages of evolutionary divergence. Systematic zoology 26: 3 19-335. AVTALION, R. R., PRUGININ, Y. & ROTHBARD, S., 1975. Determination of allogeneic and xenogeneic markers in the genus. Tilapia I. Identification of sex and hybrids in tilapia by electrophoretic analysis of serum proteins. Barnidgeh, 27: 8-13. AVTALION, R. R., DYCZMIER, M., WODJANI, A. & PRUGININ, Y., 1976. Determination of allogeneic and xenogeneic markers in the genus of Tilapia 11. Identification of T. aurea, T. vulcani and 7.nilotica by electrophoretic analysis of their serum proteins. Aquaculture, 7 : 255-265. CHEN, F. Y. & TSUYUKI, M., 1970. Zone electrophoretic studies on the proteins of Tilapia mossarnbicus and T. hornorum and their F1 hybrids, 7. zillii and T. melanopleura. journal of the Fisheries Research Board of Canada, 27: 2167-2177. CORRUCCINI, R. S., BABA, M., GOODMAN, M., CIOCHON, R. L. & CRONIN, J. E., 1980. Nonlinear macromolecular evolution and the molecular clock. Evolution, 34: 1216-1 2 19. FARRIS, J. S., 1970. Methods of computing Wagner trees. Systematic ~ o o l o g y ,19: 83-92. FELSENSTEIN, J., 1982. Numerical methods for inferring evolutionary trees. Quarterly Review of Biology, 57 ( 4 ) : 379-404. FITCH, W. H. & MARGOLIASH, E., 1967. Construction of phylogenetic trees. Science, 155: 279-284. GOLDING, G. B. & STROBECK, C., 1983. Increased number of alleles found in hybrid populations due to intragenic recombination. Evolution, 37: 17-29. GORMAN, G. C., KIM, Y. J. & RUBINOFF., 1976. Genetic relationships of three species of Bathygobius from the Atlantic and Pacific sides of Panama. Copeia, 2: 361-364. GREENWOOD, P. H., 1951. Fish remains from Miocene deposits of Rusinga Island and Kavirondo Province, Kenya. Annals of the Magazine of Natural History, ( 1 2 ) 4: 1192. HUNT, W. G. & SELANDER, R. K., 1973. Biochemical genetics of hybridization in European house mice. HerediQ. 31: 11-33. KING, M. C. & WILSON, A. C., 1975. Evolution at two levels. Molecular similarities and biological differences between humans and chimpanzees. Science, 188: 107-1 16. KIRPICHNIKOV, V. S., 1981. Genetic Bases of Fish Selection. Heidelberg: Springer-Verlag Berlin. KORNFIELD, I. L., RITTE, U., RICHLER, C. & WAHRMAN, J., 1979. Biochemical and cytological differentiation among cichlid fishes of the sea of Galilee. Euolution, 33: 1-14. LOWE, (McCONNELL) R. H., 1955. New species of Tilapia from Lake Jipe and the Pangani River. East African Bulletin of the British Museum (Natural History) .Zoology, 2 (12): 349-368. McANDREW, B. J. & MAJUMDAR, K. C., 1983. Tilapia stock identification using electrophoretic markers. Aquaculture, N: 249-26 1. MICKEVICH, M. F. &JOHNSON, M. S., 1976. Congruence between morphological and allozyme data in evolutionary inference and character evolution. Systematic ~ o o l o g y 25 , ( 3 ) : 260-270. MORGAN, K. & STROBECK, C., 1979. Is intragenic recombination a factor in the maintenance of genetic variation in natural populations? Nature, 277: 383-384. NEI, M., 1971. Interspecific gene differences and evolutionary time estimated from electrophoretic data on protein identity. American Naturalist, 105: 385-398. NEI, M., 1972. Genetic distances between populations. American Naturalist, 106: 7-10. NEI, M., 1975. Molecular Population Genetics and Evolution. Amsterdam: North Holland Publishing Company. PETERS, H. M., 1965. uber larvale Haftorgane bei Tilapia (Cichlidae, Teleostei) und ihre Riickbildung in der Evolution. ~oologischerjahresbericht Abteilung allgemein .Zoologic und Physiologie der Tiere, 71, 287-300. PETERS, H. M. & BERNS, S., 1978. uber die Vorgeschicht der maulbrutender Cichliden. Aquarien Magazin, 1978 ( 5 ) : 211-217 and ( 7 ) 324-331. PETERS, H. M. & BERNS, S., 1982. Die Maulbrutpflege der Cichliden Untersuchugen zur Evolution eines Verhaltensmusters. zeitschrqt f u r Zoologische Systemik und Euolutionsfrschung, 20: 18-52. PRAGER, E. M. & WILSON, A. C., 1978. Construction of phylogenetic trees for proteins and nucleic acids: empirical evaluation of alternative matrix methods. Journal of Molecular Evolution, 11: 129-142. REGAN, T., 1920. The classification of the fishes of the family Cichlidae. I. The Tanganyika Genera. Annals of the Magazine of Natural History Series. 9, 5 : 33-53. TILAPIINE EVOLUTIONARY RELATIONSHIPS 435 SAGE, D. E. & SELANDER, R. K., 1979. Hybridization between species of the Rana p2pens complex in central Texas. Euolution, 33: 1069-1088. SARICH, V. M., 1977. Rates, samples sizes and the neutrality hypothesis for electrophoresis in evolutionary studies. Nature London, 265: 24-28. SWOFFORD, D. L., 1981. On the utility of the distance Wagner procedure. In V. A. Funk and D. R . Brooks, (Eds) Advances tn Cladistics: Proceedings of the First Meeting of the Willi Hennzg Soczety: 2 5 4 3 New York Botanical Garden, Bronx, N.Y. THOMPSON, J. N. & WOODRUFF, R. C., 1978. Mutator genes- -pacemakers of evolution. Nature London, 274: 317-321. THORPE, J. P . , 1982. The molecular clock hypothesis: biochemical evolution genctic differentiation and systematics. Annual Review of Ecolou and Systematics, 13: 139-168. THYS VAN DEN AUDENAERDE, D. F. E., 1968. An annotated bibliography of Tzlapia (Pisces: cichlidae). Annales Musee Royal I‘Africainc Centrale <oologique. 406p. THYS VAN DEK AUDENAERDE, D. F. E., 1971. Some new data concerning the ‘Tilailapia species of the subgenus Sarotheroden (Pisces: Cichlidae). Revue <oologique Botanique Ajricaine, 84: 203-2 16. TREWAVAS, E., 1937. Fossil cichlid fishes of L. S. B. Leakey’s expedition to Kenya in 1934-35. Annals ofthe Magazine of Natural history. (10) 19: 38 1. TREWAVAS, E., 1973. On the cichlid fishes of the genus Pelmatochromir with proposal of a new genus for P. congicus; on the relationship hetween Pe/matochromis and Tilailapia and the recognition of Sarolherodon as a distinct genus. Buttetin of the British Museum (Natural History) ,Soology, 25: 1-26. TREWAVAS, E., 1980. Tilupiu and Sarotherodon. Buntbarsche Bulletin, 81: 1-6. TREWAVAS, E., 1982. Generic grouping of Tilapiini used in aquaculture. Aquaculture, 27: 79-81. TSUNO, K., 1981. Studies on mutation at esterase loci in Drosophila uirilzs. I. Spontaneous mutation rates and newly arisen variants. Japanese Journal of Genetics, 56: 155- 175. VAWTER, A. T., ROSENBLATT. R . & GORMAN, G. C., 1980. Genetic divergence among fishes of the eastern Pacific and Caribbran: support for the molecular clock. Evolution, 34: 7 0 5 7 1 1 . WATT, W. B., 1972. Intragenic recombination as a source of population genetic variability. American .Vaturalist, 106: 737--753. WHITT, G. S., CHILDERS, W. F. & CHO, P. L., 1973. Allelic expression at enzyme loci in an intertribal hybrid sunfish. Journal of Heredily, 64: 55-61. WIIITT, G. S., PHILIPP, D. P. & CHILDERS, W. F., 1977. Aberrant gene expression during the development of hybrid sunfishes. (Perciformes, leleostei). Dzfferentiation, 9: 97-109. YANG, S. Y., SOC‘LE, M. & GORMAN, G . C., 1974. Anolis lizards of the eastern Carribean: a rasr study in evolution. I . Genetic relationships, phylogeny, and colonization sequence of the roquet group. .$sternatic ?oology, 23: 387-399. 20 *
© Copyright 2025 Paperzz