BIOLOGY Pu N -Sv D Fo* Th* Fiih«R*« Society of Tke British I$Ki

Journal of

BIOLOGY Pu N -sV d fo* tH* Fiih«r*« Society of tKe British i$ki

Journal of Fish Biology (2013) 82, 165-188 doi:10.1111/jfb.12001, available online at wileyonlinelibrary.com

Genetic and morphological studies of Nothobranchius (Cyprinodontiformes) from Malawi with description of Nothobranchius wattersi sp. nov.

E. N o’oMA*f, S. V aldesalici|, K. Reichwald* and A. C ellerino*§

*Leibniz Institute for Age Research - Fritz Lipmann Institute, Beutenbergstrafie 11, D-07745 Jena, Germany, %Via Ca Bertacchi 5, 42030 Viano (RE), Italy and §Neurobiology Laboratory, Scuola Normale Superiore, c/o Area di Ricerca del CNR, via Moruzzi 1, 56124 Pisa, Italy

(Received 4 March 2012, Accepted 19 September 2012)

Molecular and morphological data were used to explore evolutionary differentiation among popu lations of Nothobranchius in the Lake Malawi-upper Shire River and the Lakes Chilwa-Chiuta drainage systems in Malawi. The aim of the study was to test the hypothesis that Nothobranchius of the Malawi-Shire system constitute a separate evolutionary group from Nothobranchius kirki. Mitochondrial and nuclear sequence data show a strongly supported phylogenetic split into two monophyletic groups separating the Lake Malawi basin fish from N. kirki. Unlike N. kirki, Lake Malawi-Shire fish do not deviate from neutrality and express an excess of rare haplotypes and mutations in terminal branches, characteristic of recently expanded populations. Further, the two groups significantly differ in morphology. Two body characters (dorsal-fin base length and pre- pelvic-pre-anal distance) are significantly different between the two species in both sexes. Several other characters are significantly different in either male or female comparisons with respect to both standard and head lengths, and robust morphological differentiation is detected by multivariate analysis. The two groups are readily distinguished on the basis of male colouration, especially in scale centres and the caudal fin. On the basis of this differentiation at the molecular and morpholog ical levels, in addition to colouration, the Lake Malawi-Shire fish are hereby formally recognized as constituting a new species, Nothobranchius wattersi. This distinction is in agreement with the geomorphologic and recent climatic history in the region. © 2012 The Authors Journal of Fish Biology © 2012 The Fisheries Society of the British Isles

Key words: divergence; male colouration; molecular divergence; new species.

INTRODUCTION The killifish genus Nothobranchius Peters 1868 occurs in tropical and sub-tropical eastern Africa with a wide distribution from South Africa to southern Sudan, and from Chad to Zanzibar and the Mafia islands in Tanzania. All known species are annual fishes, living principally in temporary pools and swamps formed during the rainy season (Seegers, 1997; Wildekamp, 2004).

fAuthor to whom correspondence should be addressed. Tel.: +49 176 627 80867; email: enoch@ fli-leibniz.de 165 © 2012 The Authors Journal of Fish Biology © 2012 The Fisheries Society of the British Isles 166 E . N G ’OMA E T A L .

Detailed accounts of Nothobranchius in Malawi in the contexts of geology, distri bution of present Nothobranchius biotopes, documentation and collection for estab lishment in captivity, soil and climate are presented by Watters (1991a, b, c, 2009). Two species are presently known in Malawi including the spotted killifish Notho branchius orthonotus (Peters 1844) in the lower Shire River area in the southern part of the country, and Nothobranchius kirki Jubb 1969 in the depression of Lakes Chilwa and Chiuta c. 100 km south-east of Lake Malawi. A third species, the bluefin notho Nothobranchius rachovii Ahl 1926 long suspected in the Malawian section of the lower Shire based on its consistent sympatric occurrence with N. orthonotus throughout the distribution range of the latter (Watters, 1991c), was recently con firmed as present in the Mozambican lower Shire (Shidlovskiy et al., 2010). A further population cluster of Nothobranchius occupying the fluvial plains of the upper Shire River and the Lake Malawi basin as far north as Chia Lagoon in central Malawi has traditionally been assigned to N. kirki (Jubb, 1981). For example, specimens of Nothobranchius collected from Salima were referred to by Jubb (1975) as N. cf. kirki. Nothobranchius of the Lake Malawi basin have, however, later been argued to constitute a separate group based on colouration, morphology (Watters, 1991a, c) and karyotypic characteristics (Scheel, 1990; Watters, 1991c). This view is in fact supported by the geomorphological features of the region that suggest sepa ration of the Lake Malawi basin from the Chilwa-Chiuta depression by the early Pleistocene or before (Watters, 1991c; Delvaux, 1995). Lakes Chilwa and Chiuta, currently isolated from each other, are believed on the other hand to have constituted a single water body draining eastwards through the Lugenda-Ruvuma River system (Lancaster, 1981). These observations raise an important question whether there is genetic divergence among fishes from the Chilwa-Chiuta and those from Lake Malawi-upper Shire basins. In this study, the morphological and genetic differentiation of Nothobranchius from these regions was assessed.

MATERIALS AND METHODS

FISHES Fish materials used for sequence analysis were collected in April 2009 from four repre sentative locations in the plains of Lakes Chilwa and Malawi [Figs 1 and 2(a), (b)]. Chilwa fishes were sampled from the western side of Lake Chilwa at the Njala Rice Scheme, Kachulu (KCH) and from Ntaja (NTJ) on the northern fringes of the lake (Table I). At both Chilwa sites (c. 90 km apart), specimens were collected from rice paddies that form part of the con tinuous seasonal marginal swamps of the lake. On the Lake Malawi plain, specimens were collected some 40 km south of Salima (SAL) town, and from Golomoti (GOL) some 60 km further south of Salima. Fishes were collected with either a dip-net, a two-man seine, or a traditional fish trap [Fig. 2(c)] and euthanized with an overdose of clove oil. Specimens were preserved in formaldehyde after removal of pectoral fins for DNA analysis. All materials for DNA were preserved in 96% ethanol. Specimens for morphological description were collected between 1988 and 2009 from 18 locations (Nothobranchius wattersi) (Table I), representing the entire known distribution range of Nothobranchius in the Lake Malawi plains. Compar ative materials of N. kirki were collected from five locations in the Lake Chiuta and Lake Chilwa catchments.

33 ° 34 ° 35°

Fig. 1. Map of parts of central and southern Malawi (Africa insert) with known distribution of Nothobranchius

wattersi (•, ^, type locality) and Nothobranchiuskirki ^ ) . Sites for populations of each species analysed for molecular divergence , N. wattersi; @ , N. kirki) are indicated.

MOLECULAR METHODS Genomic DNA was isolated from fin material of ethanol preserved samples by the phe- nol-chloroform method (Kingsley Lab Protocols, 2004) and used to examine parts of the mitochondrial gene cytochrome c oxidase subunit I (coxl) and of five nuclear genes. Degen erate primers for coxl were defined based on conserved areas of a coxl sequence alignment of Nothobranchius furzeri Jubb 1971, N. rachovi and Fundulusoma thierryi Ahl 1924 [this last species is retained in Fundulusoma based on Murphy & Collier (1998) whose analysis of mitochondrial DNA supports the paraphyly of Fundulusoma from Nothobranchius; molecular analyses (unpubl. data) support the results of Murphy & Collier (1998)]. For nuclear loci, Li et al. (2007) defined degenerate primers for 10 genes in ray-finned fishes. Five of these genes were selected for this study including myh6, glyt, zic1, sh3px3 and gpr85 based on robustness of polymerase chain reaction (PCR) amplification in most Nothobranchius species (unpubl. data). Two nesting sets of primers (Table II) were used for each nuclear locus. PCRs were performed in 25 |il final volumes, each with 2-5 |il x10 PCR buffer, 1-5 |il 25 mM MgCl2, 0-5 |il each of 10 mM deoxynucleotide triphosphate (dNTP) mix, 10 |iM for ward primer, 10 |iM reverse primer, 0-25 |il 5 U |il-1 Taq Polymerase (Qiagen; www.qiagen. com) and 100-150 ng of genomic DNA. PCR conditions were 94° C, 2 min initial denaturation followed by 35 cycles of 94° C, 30 s denaturation; 55° C (57° C for nesting), 30 s annealing; 72° C, 30 s extension and a final extension step of 1 min at 72° C. Nested PCR was carried out with 1 |il of 1:1000 dilutions of respective PCR product as template.

Fig. 2. Representative localities where specimens were collected: (a) rice fields accessed by canoe in the Group Village Headman Mkhuzumba on the northern fringes of a marsh near Ntaja town on Lake Chilwa, (b) near rice fields in Ngwimbi village, Golomoti, some 2-5 km east of the M5 road and (c) the traditional fish trap (m ono) belonging to local fishermen among whose catches many specimens for this study were collected.

All amplification reactions were performed with an Eppendorf thermocycler (Mastercycler ep gradients; www.eppendorf.com). The cox1 gene was sequenced in 69 specimens (Table III) representing two sub-populations of N. kirki (KCH and NTJ) and two sub-populations of the new species N. wattersi (GOL and SAL). Nuclear genes were sequenced in six individuals each of N. kirki and N. wattersi selected randomly. Direct sequencing of PCR products using PCR primers was performed using the BigDye Terminator v3.1 Cycle Sequencing Kit (ABI; www.appliedbiosystems.com), followed by separation on ABI 3730xl capillary sequencers. After quality clipping, sequences were assembled based on overlaps using the GAP4 module of the Staden Sequence Analysis Package (Staden, 1996). Visual inspection and manual edit ing of sequences were undertaken using GAP4. Sequences have been submitted to GenBank and accessioned JF444810-JF444933.

MORPHOLOGICAL METHODS Morphological measurements and counts were taken as described in Amiet (1987). Addi tional measurements were taken as in Garavello & Shibatta (2007) and Reichard (2010). The measurements are as follows: length of the caudal peduncle, from the posterior end of the anal-fin base to the posterior end of the hypural plate; pre-pelvic fin to pre-anal fin distance, from the anterior margin of the pelvic-fin base to the anterior margin of the anal-fin base; caudal peduncle width, measured as the thickness of the peduncle at the posterior origin of the anal-fin base; postorbital-dorsal-fin distance, from the posterior rim of the orbit to the anterior margin of the dorsal-fin base; width at dorsal origin, measured as the body thickness at the

EVOLUTIONARY DIVERGENCE IN MALAWI KILLIFISHES O VO 29 (° (° C) between Temperature (|iS cm-1 ) Conductivity Nothobranchius kirki and 30 60 25 70 140 31 50 100 35 40 80 27 TDS (mg r 1) 6-9 6-9 Nothobranchius wattersi 619 509 Altitude (m.a.s.l.) pH E E 504 E 5-4 40 24 E E 7-7 9-60 10 70 20 140 27 34 E E E 7-4 50 40 110 80 24 28 E 1988 and 2009 37' 37' 37' 37' 37' 3V 34° 35-455 34° 26-841' 35° 10-175/ 35° 49; E 35° 49; E 6-9

S S 35° 15; E S 35 34-978' E S S 34° S 34° S 34° 17; ES 34° 17; E 7-3 160 195 300 390 28 27

4-^

2T 26' 21' 21' 00

o o 15° 15° 04; 13° 13° 46; S 34 33; E 7-1 80 14° 44; S 160 31 14° 14° 26; S 34° 46; E 7-2 90 180 20-849' 51-116/ code Latitude; longitude MW 91-1 MW 88-9 Location MW 91-19 Chia MW 91-3 13° 13; S 34° 18; E 6-4 5 10 30 name Hoba MZMW 09-6 14° 55-783' S Ntaja NTJ 09 15° Benga MW 88-12 13° Benga MW 91-5 13° Chiuta Chiuta MW 91-15 14° 44; S SalimaSalima Salima MW 88-1 SAL 09 13° 46; S 13° 34 33; E 55 110 27 Chilwa MW 88-10 15° 24; S 35° 32; E 60 120 30 Kasinje MW 91-20 14° 28; S 34° 44; E 6-8 60 130 28 Kachulu KCH 09 15° 23-351' S 35 31-952' E 623 Liwonde MW 88-6 Location GolomotiGolomoti MW 94-7 GOL 09 14° 14° GolomotiGolomotiGolomoti MW 91-18Golomoti MW 88-4 MW 91-17 MW 92-7 14° 26; S 34° 14° 23; S 14° 23; 34° S 34° 14° ChinganjiChinganji MW 88-13 14° 26; S 34° 46; E 110 220 34 Mtakataka MW 92-8 14° 10; S 34° I. I. Geographic coordinates and habitat characteristics of 23 collection sites of

a b l e N. wattersi N. wattersi N. wattersi N. wattersi Species N . wattersi N. wattersi N. wattersi N. wattersi N. wattersi N. wattersi N. wattersi N. wattersi N. wattersi N. wattersi N. wattersi N. wattersi N. wattersi N. wattersi m.a.s.l., m above sea level; TDS, total dissolved solids. T N. kirki N. kirki N. kirki N. kirki N. kirki

©2012 The Authors Journal of Fish Biology © 2012 The Fisheries Society of the British Isles, Journal of Fish Biology 2013, 82 , 165-188 o N G ’OMA ET A L. number GeneBank accession JF44491 5 -JF444922 JF44481 0 -JF444880 Inner JF444903-JF444914 Inner JF444881-JF444890 Inner JF444891-JF444902 Inner Inner JF444923-JF444933 Outer Outer Outer Outer Outer direction

' 3 5 ; GGAAGTG GTGTTCTCTCC GAATTTTYAGRT CTCGTA TGTTRAAGAT ATCCAGTTGAA AGTTGAACAT TTGTGRATYTT AGTASAGGAG TGGTGCT ATGTTCTC

Nothobranchiusfurzeri. II. Primers used for PCR and sequencing listed in GACGTTCCCGGCWAAAATGAT AT CATCTCYCCGATGTT T a b l e 532-1299 (2007) (2007) 559-1562 (2007)ACMTACCACTGTCMAAGAT G GG 577-1464 (2007) CCCAAGAGGTTCT ACATGGTACCAGTATGGCTTTGT 459-1325 (2007) CATMTTYTCCATCTCAGATAATGC 507-1322 GTAAGGCATATAS AC GCTCKGTAT C ART ATC ATTCTCACCACCA AG GG CTCACCACCATCC (2007) 16-963 GGACCGCAGTATCCCACYMT GTGTGTCCTTTTGT (2007) 9-967 GGACGCAGGACCGCARTAYC CTGTGTGTGTCCTT (2007) 10-1094GGCGAACTAYAGCCGC AT AT CTGGATTTTCTGC (2007) 27-1082(2007)AGGGGACC 461-1303GC TAMC ATAC AGAYTGGTSAGGA GGCAC GTAT CAAACAKCTCYCCG CAGTASAGGAGCG (2007). et al. et al. et al. et al. et al. et al. et al. et al. et al. et al. et al. NC_011814-1* 555-1198 GACCCAGCTGGWGGAGG AGA AGCCTA CCTGCTA Li Li Li Li Li Li Li Li Li 1 1 Li in Li sreb2 *GenBank code for a mitochondrial sequence of Gene Reference Amplicon Forward primer sequence Reverse primer sequence PCR step t sh3px3 PCR. polymerase chain reaction. coxl myh6 myh6 sh3px3 gpr85 gpr85 g ¥ g ¥ zicl zicl

T a b l e III. Numbers of Nothobranchius specimens (n) sequenced for one mitochondrial and five nuclear loci, with summaries of variable sites (PI, parsimony informative sites) used in subsequent analyses

Gene n Sites Variable PI cox1 69 583 28 16 myh6 12 638 17 12 glyt 9 761 3 2 zicI 7 615 3 1 gpr85 11 867 1 1 sh3px3 12 659 1 0

anterior origin of the dorsal-fin base; body depth at dorsal-fin origin, at the anterior origin of the dorsal fin; pectoral-fin length, from the base to the posterior tip; pelvic-fin length, distance from the anterior pelvic girdle to the posterior tip of the pelvic fin; caudal-fin length, from the posterior end of the hypural plate to the furthest reach of the caudal fin; pectoral girdle width, measured across the cleithra at the base of the pectoral spine; snout length measured from the tip of the snout to the anterior margin of the eye; eye diameter, measured between the anterior and posterior orbital rim; head depth, at the mid line of the occiput vertically downward (i.e. just behind the eye); head width, greatest dimension horizontally between the opercules in normal position; upper jaw length, from the tip of the snout to the posterior of the mouth in front of the eye; lower jaw length, from the tip of the lower jaw to the posterior edge of the mouth; mandible length, from the tip of the lower jaw to the fold near the ventral opercular origin. Measurements were made with digital callipers (to the nearest 0-1 mm), using a dissecting microscope to observe small structures. All measurements and counts were taken from the left side. All visible rays of the dorsal and anal fins were counted. The count of scales on the median longitudinal series is the number of scales between the superior junction of the opercular membrane and the hypural plate; the scales on the base of the caudal fin were counted separately. Terminology for the cephalic neuromast series follows Scheel (1968) and for the frontal squamation as described by Hoedeman (1958). Type and comparative materials are deposited at the Museo di Storia Naturale e del Territo- rio di Calci University of Pisa, Italy (Catalogue number Pe0155-Pe0159). Additional material was deposited in Musee Royal de l’Afrique Centrale (MRAC), Tervuren, Belgium, and South African Institute for Aquatic Biodiversity (SAIAB), Grahamstown, South Africa.

DATA ANALYSIS Molecular analysis Sequence alignments were made in MEGA v4 (Tamura et al., 2007) and MUSCLE (Dereeper et al., 2008). Phylogenetic and molecular evolutionary analyses were done in MEGA, PhyML (Dereeper et al., 2008) and MrBayes v3.1 (Huelsenbeck & Ronquist, 2001; Ronquist & Huelsenbeck, 2003). Arlequin (Excoffier et al., 2005) was used to compute haplotype diversity and molecular evolutionary estimates within and between groups. Sequence variation of the cox1 gene was evaluated by several algorithms including neighbour-joining (NJ) (Saitotu & Nei, 1987), maximum parsimony (MP) (Eck & Dayhoff, 1966; Nei & Kumar, 2000), minimum evolution (ME) (Rzhetsky & Nei, 1992), unweighted pair-group method using arithmetric averages (UPGMA) (Sneath & Sokal, 1973) and Bayesian analyses. A second data set was constructed by concatenating corresponding sequences of four nuclear loci: glyt, myh6, zic1 and gpr85. A third data set was made by joining the nuclear concatenation with corresponding mitochondrial (cox1 ) sequences. Finally, each nuclear gene was analysed by itself. All data sets involving nuclear genes were analysed with MrBayes.

Joined data sets of nuclear-only and nuclear-plus-mitochondrial genes consisted of 2880 and 3462 bp, respectively. In distance and parsimony-based analyses of the cox1, the complete deletion option for gaps and missing positions was used resulting in 466 of 583 sites in the final data set. In all analyses, F. thierryi and N. furzeri were chosen as outgroups. A 498 bp sequence of Chilwa N. kirki cox1 gene was obtained from GenBank (accession number AF002575) and included for comparison with the sequence data in this study. Confidence probability (interior branch length) for optimal trees from NJ, ME and UPGMA were computed by bootstrap test of 1000 replicates (Rzhetsky & Nei, 1992; Dopazo, 1994). The evolutionary distances were estimated using the maximum composite likelihood method (Tamura et al., 2004) and are in the units of the number of base substitutions per site. The rate variation among sites was modelled with a gamma distribution (shape parameter = 1). The differences in the composition bias among sequences were considered in evolutionary comparisons (Tamura & Kumar, 2002). The MP tree was calculated using the close-neighbour- interchange algorithm with branch lengths calculated using the average pathway method (Felsenstein, 1985; Nei & Kumar, 2000) and tested by 500 bootstrap replicates. For Bayesian analysis, variance was partitioned either for individual genes (four partitions for nuclear genes and five partitions including the mitochondrial gene) or for nucleotide position within the codon (three partitions with genes concatenated in frame). A general time reversible model with gamma distribution and invariable sites (GTRr+I) was selected for each partition. The number of generations was set to 1 000 000 with sampling of every 100 generations using two parallel runs with four chains. Default settings of priors and temperature were used and two chains were run in parallel. The first 25% generations were discarded as burn-in. Convergence was checked by s .d . of splits and the simulation was stopped when s.d . of splits was < 0-01. Indices of molecular diversity including haplotype diversity (h), nucleotide diversity (n) and estimates of neutrality to detect signatures of selection in the cox1 locus were evaluated using Arlequin.

Morphological analysis The final data set of morphological characters comprised 164 specimens (102 males and 62 females). The males comprised 77 N. wattersi and 25 N. kirki; females included 60 N. wattersi and two N. kirki. Descriptive statistics (mean, s.d., s.e., range and median), normality tests (Shapiro-Wilk, skewness and kurtosis) and test of statistical difference in means (t-test) were computed in STATA 11 (StataCorp., 2009) on appropriately transformed non-normal data samples using STATA’s ladder and gladder commands. Permutation tests were conducted in PAST (Hammer et al., 2001) set to 2000 runs.

RESULTS

D N A SEQUENCE DIVERGENCE Evolutionary history was inferred from sequence data of one mitochondrial gene (cox1), extensively used for molecular barcoding (Ratnasingham & Hebert, 2007), and four nuclear genes (myh6, glyt, zic1 and gpr85) which are a sub-set of genes broadly used for phylogenetic studies in teleosts (Li et al., 2007, 2010; Kawahara et al., 2009; Hulsey et al., 2011). The NJ tree inferred from a sequence alignment consisting of 583 bp of the cox1 gene from 69 specimens (and two additional sequences of F . thierryi and N. furzeri as outgroups) is presented in Fig. 3. Fur ther analyses using MP, ME and UPGMA methods show similar topologies. The NJ topology is presented with interior branch test values. Bootstrap values of MP are also shown mainly for nodes separating clades in which the two species are found. Both methods show very high support in favour of the new species N. wattersi as a separate

T a b l e IV. Sequence variability in Nothobranchius spp. coxl gene: haplotype-gene diversity (h), nucleotide diversity (n) and estimates of Tajima’s D and Fu’s Fs. Significant deviations from neutrality are indicated (*P < 0-05; **p < 0-01)

Species h mean ± s.d . n mean ± s.d . Tajima’s D Fu’s Fs Nothobranchius wattersi 0-6750 ± 0-1174 0-004271 ± 0-002881 -0-78893 -0-85188 Nothobranchius kirki 0-6575 ± 0-0673 0-001499 ± 0-001200 -1.74404* -6-27997**

taxonomic group. Further, there appears to be higher variability within the N. wattersi clade, particularly between fish from Salima and those from the Golomoti area. Four nuclear gene fragments show parsimony informative variations from one to 12 (Table III). A fifth nuclear gene sh3px3 did not show sequence variation and was excluded from subsequent analyses. Bayesian analysis was used to partition variation in joined data sets so that evolutionary models are computed separately for each gene and codon. As with the larger data set, analyses either by gene or by codon favour the separation of the new species N. wattersi from N. kirki, with larger diversity signalled in the population of N. wattersi [Fig. 4(a), (b)]. This distinction persists even when some nuclear loci are analysed singularly. Single gene trees for myh6, glyt and zicl are shown in Fig. 4(c)-(e). It can be noted that myh6 [Fig. 4(c)] generally shows similar separations as in co xl (Fig. 3) and the two joined data sets [Fig. 4(a), (b)]. The genes glyt and zic1 [Fig. 4(d), (e)] with two and one informative sites, respectively, are also able to resolve the N. kirki and N. wattersi clades from the polytomy, although the Golomoti population of N. wattersi is misplaced in the topology for zic1 [Fig. 4(e)]. Analysis of sequence divergence confirms higher diversity within N. wattersi as compared to N. kirki, with N. wattersi showing almost three-fold higher nucleotide diversity (0 0043 v. 0 0015) (Table IV). Consistent with the phylogenetic analysis, a much larger value of divergence (0 021) over sequence pairs was observed between N. kirki and N. wattersi (Table V). Finally, N. wattersi does not deviate from neu trality (Table IV). On the other hand, N. kirki shows significantly negative values of Tajima’s D and Fu’s F statistics which indicate deviation from neutrality and over representation of rare haplotypes and mutations in terminal branches, characteristic of recently expanded populations. In summary, the data presented here show that the two species have significant differences in their genetic structure, although more detailed analysis would be necessary to describe their demographic history.

MORPHOLOGICAL DIVERGENCE Specimens representing almost the entire known distribution of Nothobranchius in the plains around Lakes Chilwa, Chiuta and Malawi were analysed morphologically. Twenty body and 10 head characters expressed as per cent of standard length (LS) or per cent of head length (LH), respectively, were examined. Several significant differences (t-test, P < 0 01) were observed in comparisons of both sexes among N. kirki (from Lakes Chilwa to Chiuta) and N. wattersi (from Lake Malawi to the Upper Shire). With respect to LS, dorsal-fin base length and pre-pelvic-pre- anal distances differ significantly between males and between females from the two geographical regions (Table VI). The dorsal-fin base (as % LS) is characteristically

N. kirki NTJ (N=5) N. kirki NTJ48 N. kirki NTJ42 N. kirki NTJ34 N. kirki NTJ30 N. kirki NTJ27 N. kirki KCH21 N. kirki KCH16 N. kirki NTJ11 N. kirki KCH04 N. kirki KCH02 N. kirki KCH05 N. kirki NTJ12 N. kirki KCH17 N. kirki KCH22 N. kirki NTJ28 N. kirki NTJ31 N. kirki NTJ35 N. kirki NTJ44 N. kirki NTJ50 N. kirki NTJ55 - N. kirki KCH14 N. kirki CHILWA N. kirki KCH03 N. kirki NTJ08 N. kirki NTJ13 N. kirki KCH19 N. kirki NTJ25 N. kirki NTJ29 N. kirki NTJ32 N. kirki NTJ38 N. kirki NTJ47 N. kirki NTJ51 N. kirki NTJ56 - N. kirki KCH20 -| N. kirki (N=7) 65/77 L N. kirki NTJ24 -I N. kirki (N=5) N. kirki KCH18 98/100 N. wattersi GOL86 74 N. wattersi GOL87 N. wattersi GOL85 74 N. wattersi SAL88 H- N. wattersi SAL102 74/73 f- N. wattersi SAL96 N. wattersi SAL89 N. wattersi SAL90 72 N. wattersi SAL93 N. wattersi SAL94 N. wattersi SAL95 72 N. wattersi SAL97 N. wattersi SAL98 N. wattersi SAL99 N. wattersi SAL100 N. wattersi SAL103 _ N.furzeri-GRZ - F. thierryi GH-06-5

0-02

Fig. 3. Neighbour-joining (NJ) analysis of 69 sequences of the coxl gene of Nothobranchius kirki (53) and Nothobranchius wattersi (16). The second number at nodes bearing two values is a bootstrap estimate for the maximum parsimony tree. All the other numbers represent the probability of that branch in per cent (interior branch test) of the optimal NJ tree. Fundulusoma thierryi and Nothobranchius furzeri were included as outgroups in all analyses.

(a) (b) ------F. thierryi GH06-5 N. furzeri GRZ F. thierryi GH06-5 - N. wattersi SAL102 96 " N.furzeri GRZ • N. wattersi SAL98 100 92 N. wattersi GOL85 - N. wattersi SAL94 N. wattersi SAL94 100 L N. wattersi GOL85 100 N. wattersi SAL98 92 N. wattersi SAL102 65i— N. kirki KCH20 95 N. kirki NTJ13 100 N. kirki KCH19 N. kirki KCH19 L N. kirki NTJ13 N. kirki KCH20

(c) (d) F. thierryi GH06-5 N. furzeri GRZ 83 N. kirki NTJ41 — N. kirki KCH19 F. thierryi GH06-5 99 N. kirki NTJ13 N. furzeri GRZ — N. kirki NTJ06 N. wattersi SAL102 72 N. kirki KCH20 N. wattersi SAL98 “ i------N. kirki KCH01 N. wattersi SAL94 N. wattersi SAL102 N. wattersi GOL87 N. wattersi SAL98 N. wattersi GOL85 98 N. wattersi SAL94 r N. kirki KCH19 N. wattersi GOL87 N. kirki NTJ13 N. wattersi GOL86 - N. kirki NTJ06 N. wattersi GOL85 N. kirki KCH01

(e) F. thierryi GH06-5 N. furzeri GRZ 79 N. wattersi SAL102 N. wattersi SAL98 -E N. wattersi SAL94 64 N. wattersi GOL85 N. kirki NTJ41 N. kirki KCH20 ----- N. kirki KCH19

Fig. 4. Bayesian analysis of Fundulusoma thierryi, Nothobranchius furzeri, Nothobranchius wattersi and Nothobranchius kirki with posterior probabilities of joined data sets. (a) Concatenation of four nuclear genes glyt, myh6, zic l and gpr85 with variance partitioned by codon. (b) coxl concatenated with nuclear genes, variance partitioned by gene. (c-e) Bayesian analysis and posterior probabilities of single nuclear genes: (c) myh6, (d) glyt and (e) zicl. shorter in male N. wattersi (17-6-29-9%) than in male N. kirki (22-1-32-7%), and 18-4-27-7% v. 23-9-29-0% for female N. wattersi compared to female N. kirki. The difference between mean values is more pronounced between females (22-1%

T a b l e V. Estimates of average evolutionary divergence over sequence pairs (d) within and between groups. The number of base substitutions per site from averaging over all sequence pairs within and between groups is shown. The rate variation among sites was modelled with a gamma distribution (shape parameter = 0-7). n = 69 sequences (i.e. 53 Nothobranchius kirki and 16 Nothobranchius wattersi)

Diversity index d s .e . Divergence over sequence pairs within N. kirki 0-001 0-001 Divergence over sequence pairs within N. wattersi 0-003 0-002 Divergence over sequence pairs between groups 0-021 0-006 Overall mean divergence over all sequences 0-008 0-002 Coefficient of evolutionary differentiation 0-623 0-002

23-9-29 11-4-11-9 34-5-39-1 47-0-49-1 Female range 6-9 6-9-6-9 = 2 14-4 13-0-15-7 30-2 29-2-31-2

mean

n Female

0 00 Male 8-4-16-9 range 4-5-14-3 11-7 28-1-33-9 53-2-65-2 64-1 62-5-65-6 Nothobranchius kirki and Nothobranchius 8-9 4-2 3-0-6-6 5-2 3-7-6-6 13-7 10-5-15-3 12-1 11-0-13-1 29-8 24-2-33-1 26-1 24-0-28-1 58-9 = 25 Male mean n range Female 3 4 -8 -3 9-9-13-9 10-6-17-4 13-1 10-6-16-6 14-0 13-6-14-4 20-8-27-9 60 = 13-9 < 0-01) are marked between males ("), between females (") and between 22-1 18-4-27-7 27-2 22-1-32-7 26-5 23-9-29-0 46-4 41-6-55-5 45-3 40-5-50-0 48-1

mean Female n P ( both sexes (*) N. wattersi N. kirki N. kirki Male range 3-9-8-7 6-1 17-6-29-9 37-7-51-9 23-5-32-5 24-7 and 11 5-9 19-8 15.9-24-4 19-6 16-2-23-7 19-0 144-23-7 14-5 13-1-15-8 27-9 29-3 24-4-34-8 24-8 20-2-30-7 31-0 27-1-36-1 26-5 44-6 42-7 22-3-47-5 46-4 41-3-51-6 44-0 39-5-48-1 45-1 42-9-47-2 Male = mean n N. wattersi 8-1 10-8 3-6-16-0 10-78-7 6-4-14-1 8-2 6-8-10-1 7-8 6-1-10-2 8-8 4-8 15-1 14-6 10-8-18-6 19-1 14-012-1 14-6 13-1 11-0-18-5 10-6-15-3 16-8 11-8 13-2-22-7 13-6 38-1 40-6 30-7-53-4 36-1 24-9-47-5 34-1 27-0-43-3 36-8 30-5 32-1 28-7-35-5 31-2 25-1-34-2 30-4 22-527-6 22-3 25-4 15-1-28-6 16-3 114-24-7 23-3 19-4-26-6 16-5 16-2-16-7 20-1 17-1 12-8-21-4 16-829-0 28-9 12-7-22-1 19-4 154-23-4 17-4 14.3-20-4 44-9 43-9 124-2 119-5 111-9-127 118-8 109-6-126-3 121-8 117-2-125-2 117-3 113-9-120-7 Holotype MW 88-1 Ls expressed as a percentage of standard length (Ls) and head length (Lh)- Character means were tested with f-test and characters that are VI. Morphometric measurements (minimum and maximum values of all specimens studied) for

(mm) h a b l e Ls significantly different (boldface) between Caudal-fin length" Caudal peduncle width" Caudal peduncle depth" wattersi, Total length" T Pelvic-fin length" Caudal peduncle lengthWidth at dorsal origin" Postdorsal 20-7 distance 20-1Depth at dorsal-fin origin" Depth 15-6-25-8at pelvic fin" Pectoral-fin lengthPectoral-girdle width" 21-0 15-9-25-7 19-4 18-2 16-0-23-3 17-5 14-4-21-5 18-8 16-4 18-3-19-2 12-2-19-8 17-1 13-6-19-2 9-3 9-3-9-3 In % of Predorsal lengthPre-anal lengthPre-pelvic length Dorsal-fin base length* 58-6 61-0 57-8 58-8 51-9-63-4 51-9-64-5 61-0 63-4 47-4-66-2 56-6-81-5 58-4 53-9-63-0 59-2 57-0-61-4 Postorbit-dorsal-fin distance" Pre-pelvic-pre-anal distance* Anal-fin base length" L

© © 2012 The Authors Journal of Fish Biology © 2012 The Fisheries Society of the British Isles, Journal of Fish Biology 2013, 82, 165-188 EVOLUTIONARY DIVERGENCE IN MALAWI KILLIFISHES 16-2-21-6 23-5-38-9 Female range 9-5 6-6-11-4 19-1 mean n = 2 Female Male range 7-8-13-9 25 15-1 11-0-20-5 9-7 5-1-15-5 20-1 14.7-24-8 17-7 13-25-5 59-8 52-5-65-7 60-4 52-1-69-7 Male mean n = range Female 8-3-19-9 16-1-29-9 32-5-48-8 43-6 36-8-50-6 38-0 31-6-44-3 77-4-97-1 88-6 78-3-100-0 89-4 83-8-97-0 46-7-71-9 Continued VI. VI. 11-2 7-9-16-0 10-4 88-9 58-0 48-3-69-3 63-6 51-2-78-7 60-8 54.7-67-0 40-9 = 60 mean Female n T a b l e N. wattersi N. kirki 74.6 - 7 . Male range 9-2—17-1 54 87-4 77-9-99-3 Male mean n = 11 11-6 13-0 17-1 16-2 11-6-22-0 15-5 26-1 24-0 16-2-31-1 22-1 24-3 25-1 16-9-34-3 22-7 14-9-29-3 22-6 14-0-34-0 27-3 30-7 20-0-48-5 28-2 18-3-38-3 35-1 16-7-43-8 28-3 20-6-35-8 90-9 51-2 52-3 39-7-57-2 51-1 43-3-68-0 47-4 34-2-59-8 48-2 42-1-53-0 46-3 43-7 36-2-51-6 Holotype MW 88-1 h L (mm) h , , sample size. In % of Eye diameter 25-0 24-8 20-0-33-2 26-0 20-2-32-2 26-4 21-2-34-0 28-0 Head depthHead widthInter-orbital length Snout length* 64-3 60-7 59-2 64-8 51-1-69-4 58-6 Upper jaw length Lower jaw length* Upper head length Postorbital distance" Mandible length" n L

T a b l e VII. Multivariate analysis of body and head characters between Nothobranchius wat tersi and Nothobranchius kirki using the two-group permutation, Mahalanobis distance (dM) test

Comparison set d M P (same mean) Body characters, male 0-3133 < 0-001 Head characters, female 0-3061 < 0-001 Head characters, male 0-2254 < 0-001

in N. wattersi v. 26-5% in N. kirki) as compared to males (25-4% in N. wattersi v. 27 2% in N. kirki). On the contrary, the mean distance between the pelvic and anal fin is greater in female N. wattersi (16 8% v. 14 4% in female N. kirki) and in male N. wattersi (14 6% v. 13 6% in male N. kirki), with pronounced variability among the females (13-2-22-7% v. 13-0-15-7% for N. wattersi and N. kirki, respectively). Snout and lower jaw lengths (as % LH) are significantly longer in N. wattersi than in N. kirki (P < 0 01) (Table VI) giving N. wattersi a snout shape that curves upward slightly more than N. kirki. There are several other body characters that are significantly different between species (P < 0 01) in one sex but not the other. In males, these are as follows: maximum recorded total length, head length, anal-fin base length, pelvic-fin length, postorbit-dorsal-fin distance, caudal peduncle width, width at dorsal-fin origin, cau dal peduncle depth, depth at dorsal-fin origin and depth at pelvic fin and at pectoral girdle. In females, the only body character that is interspecifically variable is the postdorsal distance. Two head characters differ significantly between the two species (P < 0 01). These are postorbital distance and mandible length in males and inter orbital length and upper jaw length in females. Furthermore, the dorsal surface of the head in N. wattersi is slightly concave (i.e. curves inwards) as compared to N. kirki. The difference in the Lake Chilwa v. Lake Malawi fishes is further supported by a multivariate two-group comparison using permutation analysis, as the distributions of groups by sex are multivariate non-normal. Mahalanobis distance (2000 permuta tions; Table VII) in head and body characters for male and female comparisons are highly significant (P < 0 001). A test of body characters for the female was skipped because there were only two individuals in the N. kirki female group.

NOTOBRANCHIUS WATTERSI N. SP. [Fig. 5(c)-(f) and Table VI] Material listed below includes the following information: collection of catalogue number (field number), sex, LS, collection locality, collector and date of collection.

Holotype Pe0155 (field voucher no. MW 88-1); one male, 381 mm LS; Salima District, Malawi, temporary roadside pools 9-11 km ENE of Salima on road S29, 13° 46' S; 34° 33' E; B. R. Watters; 19 March 1988.

Paratypes Pe0156 (field voucher no. MW 91-1); six males, 35 1-46 0 mm LS; six females, 32 1-34 9 mm LS; Salima District, Malawi, roadside pools (including rice fields, ditches, dambos (permanent or seasonal wetlands; http://www.wetlands.org/Whatwe do/Actions/SustainabledambomanagementinMalawi/AboutthedambosinMalawi/tabid/ 2372/Default.aspx) and flooded grassy areas), 9-11 km ENE of Salima on road S29, 13° 46' S; 34° 33' E; B. R. Watters; 22 March and 10 April 1991. Non-type material examined: SAIAB 98799 (field voucher no. SAL 09); two males, 33-7-42-6 mm LS, 10-1-12-4 mm LH; six females, 35-0-41-5 mm LS, 10-1-10-9 mm LH; Salima Malawi, small pool at culvert under railroad c. 100 m east of M5 road and stream-side pool, 50 m west of M5 road, 13° 51116' S; 34° 26 841' E; E. Ng’oma, 9 April 2009. SAIAB 98800 (field voucher no. MW 91-18); six males, 40-6-50-8 mm LS, 14-0-15-3 mm LH; eight females, 36-8-44-1 mm LS, 11 -9-13-4 mm LH, Golo moti, Malawi, roadside pool at a culvert on east side of M17 road, 15 km SSE of Golomoti road junction, 14° 26' S; 34° 37' E; B. R. Watters, 7 April 1991. MRAC B1-14-P-46-47 (field voucher no. MW 88-4); one male, 43 4 mm LS, 131 mm LH; one female, 42 9 LS, 117 LH, Golomoti, Malawi, roadside pools, some rice fields, alongside road D81, 4 5 km NNE of Golomoti, 14° 23' S; 34° 37' E; B. R. Watters, 22 March and 5 April 1988. MRAC B1-14-P-19-31 (field voucher no. MW 91-17); eight males, 35-2-47-2 mm LS, 12-1-14-9 mm LH; five females, 32-1-37-9 mm LS, 10-2-11-9 mm LH, Golomoti, Malawi, roadside pools and some rice fields alongside road D81, 4 5 km NNE of Golomoti Centre, 14° 23' S; 34° 37' E; B. R. Watters, 7 April 1991. SAIAB 98801 (field voucher no. MW 92-7); one male, 48 4 mm LS, 16 3 mm LH; four females, 43-2-47-5 mm LS, 14-7-16-0 mm LH, Golomoti, Malawi, pool at a culvert on east side of M17 road, c. 3 5 km SSE of Golomoti road junction, 14° 27' S; 34° 37' E; B. R. Watters, 26 March 1992. MRAC B1-14-P-78-87 (field voucher no. MW 94-7); two males, 49-1-50-2 mm LS, 16-5-16-7 mm LH; three females, 35-0-40-6 mm LS, 11 -8-12-6 mm LH, Golomoti, Malawi, natural pool used for cultivation of rice on west side of M17 road, 2 km south of Golomoti road junction, 14° 26' S; 34° 37' E; B. R. Watters, 29 March 1994. MRAC B1-14-P-88-90 (field voucher no. GOL 09); two males, 40 02-41 01 mm LS, 12-30-12-32 mm LH; one female, 34 09 mm LS, 9 25 mm LH, at Ngwimbi village, Golomoti, Malawi, pool near rice fields some 2 5 km east of M5 road, 14° 20 849' S; 34° 35 455' E; E. Ng’oma, 7 April 2009. SAIAB 98802 (field voucher no. MW 88-6); one female, 411 mm LS, 119 LH, Liwonde, Malawi, small roadside pools in ditches, at a culvert, on either side of D221 road, 4 km east of Liwonde, 30 m east of junction with road to Liwonde National Park, 15° 04' S; 35° 15' E; B. R. Watters, 23 March 1988. SAIAB 98803 (field voucher no. MW 88-12); one male 39 7 mm LS, 13 2 mm LH, Benga, Malawi, pool in a ditch alongside road S33 in vicinity of Benga village, c.54 km NNE of Salima in direct line, 13° 21' S; 34° 17' E; B. R. Watters, 1 April 1988. MRAC B1-14-P-48-67 (field voucher no. MW 91-5); 13 males, 30-7-47-0 mm LS, 10-4-15-0 mm LH; seven females, 24-9-39-9 mm LS, 7-9-12-7 mm LH, Benga, Malawi, pool in a ditch alongside road S33 in vicinity of Benga village, c. 54 km NNE of Salima in direct line, 13° 21' S; 34° 17' E; B. R. Watters, 24 March and 9 April 1991. MRAC B1-14-P-45 (field voucher no. MW 88-13); one male, 44 9 mm LS, 14 7 mm LH, Chinganji, Malawi, roadside pool adjacent to a swampy river close to the village of Chinganji, c. 15 km NE of Kasinje

© 2012 The Authors Journal of Fish Biology © 2012 The Fisheries Society of the British Isles, Journal of Fish Biology 2013, 82, 165-188 180 E . N G ’OMA E T A L . on D116 road, 14° 26' S; 34° 46' E; B. R. Watters, 5 April 1988. MRAC B1-14- P-1-18 (field voucher no. MW 91-19); 13 males, 31-2—33-8 mm LS, 9-2-14-9 mm Lh; six females, 29-0-37-3 mm LS, 8-9-12-3 mm LH, Chinganji, Malawi, road side pool adjacent to a swampy river close to the village of Chinganji, c. 15 km NE of Kasinje on D116 road, 14° 26' S; 34° 46' E; B. R. Watters, 7 April 1991. MRAC B1-14-P-68-72 (field voucher no. MW 91-3); five males, 36-1-44-5 mm LS, 11-9-14-0 mm LH, Chia, Malawi, large shallow pool on east side of S33 road, 12 km south of bridge over Chia Lagoon, immediately north of village of Wabango at junction with road to Mwansambo, 71 km north of junction between S33 and M5 roads, 13° 13' S; 34° 18' E; B. R. Watters, 23 March and 9 April 1991. SAIAB 98804 (field voucher no. MW 91-20); five males, 32-2-44-7 mm LS, 11 -0-14-3 mm LH; seven females, 29-9-39-7 mm LS, 9-7-12-2 mm LH, Kasinje, Malawi, shallow circular pool, 9 km NE of Kasinje on north side of road D116, 14° 28' S; 34° 44' E; B. R. Watters, 7 April 1991. MRAC B1-14-P-73-77 (field voucher no. MW 92 8); two males, 47-7-51-6 mm LS, 15-6-15-8 mm LH; three females 42-3-44-0 mm LS, 12-2-13-9 mm LH, Mtakataka, Malawi, large pool on west side of M17 road, at Kamzati village, 17 km north of junction with M18 road, c. 7 km north of Mtakataka, 14° 10' S; 34° 31' E; B. R. Watters, 26 March 1992. SAIAB 98805 (field voucher no. MZMW 09-6); two males, 38 0-41 0 mm LS, 11 -2-13-7 mm LH; one female 29 8 mm LS, 9 2 mm LH, Hoba, Malawi, west side of M3 road from Liwonde to Ulonga, south of Lake Malombe near Hoba, upper Shire River floodplain, 14° 55 783' S; 35° 10175' E; P. Kearney & J. Jordaan, April 2009. See Appendix for measurements of N. kirki.

Diagnosis Nothobranchius wattersi males similar to N. kirki and differing from all other species of the genus by the following combination of characters: red colouration on caudal fin and part of caudal peduncle, anal fin red with a blue-green proximal portion, dorsal fin with light blue margin; pectoral, caudal and anal fins with black margin. Nothobranchius wattersi is readily distinguished from N. kirki by light blue colouration in scale centres (v. blue-green), plain deep red colouration on caudal and distal portion of anal fin (v. orange hue on distal portion), vermiculated blue pattern on anal fin (v. less irregular and limited at fin base), dorsal fin blue-green (v. bright light blue), anal fin more rounded (v. slightly wedge shaped in N. kirki), dorsal-fin base shorter, pre-pelvic to pre-anal distance longer and head longer, with a more flattened to slightly concave shape on the upper head surface relative to N. kirki.

Description See Fig. 5(c)-(f) for overall appearance and Table VI for morphometric data of the type series. Robust, elongate Nothobranchius, maximum observed LS in males 53 4 mm. Dorsal profile slightly concave on head, convex from nape to end of dorsal-fin base. Ventral profile convex, slightly concave on caudal peduncle poste rior to dorsal and anal fin. Snout slightly pointed, mouth directed upwards, lower jaw longer than upper, posterior end of rictus at same level as or slightly above centre of eye. Branchiostegal membrane projecting posteriorly from opercle. Dor sal and anal fins located posterior to mid-body. Dorsal and anal fins rounded, tips

Fig. 5. Comparison of Nothobranchius wattersi from several locations in the Lake Malawi plain with Notho branchius kirki from Lakes Chilwa and Chiuta basins: (a) N. kirki, adult male c. 36-6 mm standard length (L S), not preserved; Malawi: MW 88-10: pools in a temporary stream system, at a bridge on road S69, 23 km east of Zomba and c. 7 km west of Kachulu, (b) N. kirki, adult male c. 37-4 mm Ls, not preserved; Malawi: MW 88-9, very shallow pool in ditch at roadside, c. 5 km west of northern end of Lake Chiuta, 3 km south-east of village of Nkhokwe on road to Mpakaka, (c) N. wattersi, adult male c. 39-5 mm Ls, not preserved; Malawi: MZMW 09-6A, west side of M3 road from Liwonde to Ulonga; south of Lake Malombe; near Hoba; upper Shire River floodplain, (d) N. wattersi, adult male c. 39-2 mm Ls, not preserved; Malawi: Mw 92-5, roadside pools (including rice fields, ditches, dambos and flooded grassy areas, 9-11 km east-north-east of Salima on road S29, (e) N. wattersi, adult male c. 37-6 mm Ls, not preserved; Malawi: MW 92-2, pool in a ditch alongside road S33 in vicinity of Benga village, c. 54 km north-north-east of Salima in direct line and (f) N. wattersi, adult male c. 41-0 mm Ls, not preserved; Malawi: MW 92-4, large shallow pool on east side of S33 road, 12 km south of Chia Lagoon bridge, immediately north of village of Wabango at junction with road to Mwasambo, 71 km north of junction between S33 and M5 roads.

with short filamentous rays. Anal fin with papillate contact organs on fin rays, dor sal with few rudimentary contact organs. Dorsal-fin tip not reaching caudal fin. Dorsal-fin rays 16-17; anal-fin rays 17-18. Pectoral fin approximately triangu lar, tip not reaching pelvic fin. Pelvic-fin tip reaching anal-fin base. Caudal fin subtruncate. Female smaller than male, maximum observed Ls 47 5 mm. Body less compressed and deep compared to males [depth at pelvic fin 20-2-30-7 (24 8)% v. 24-4-34-8 (29 3)% L s]. Dorsal fin rounded. Anal fin triangular with rounded tip. Anal fin positioned more posteriorly [56-6-81-5 (63 4)% v. 51-9-64-5% Ls (58 8)] and less deep caudal peduncle [9-9-13-9 (118)% v. 10-6-15-3 (131)% Ls] compared to males. Branchiostegal membrane not projecting from operculum.

Scales Scales cycloid, body and head entirely scaled, except for ventral surface of head. Scales in median lateral series 26-28 + 3-4 on caudal-fin base. Cephalic squamation pattern variable, some specimens presenting irregular G-type. Anterior neuromast series ‘open’ type. Central supra-orbital series in a shallow groove with two neuro masts. Posterior cephalic neuromast series curved with two, rarely three, neuromasts. One neuromast on each scale of median longitudinal series.

Osteology and dentition Basihyal bone triangular, six branchiostegal rays, lateral process of post-temporal present but reduced, single reduced anterodorsal process of urohyal, vomerine teeth present in a small patch. Total vertebrae 27-28. Premaxilla and dentary with many irregularly distributed unicuspid, slightly curved teeth of different size, a small num ber of larger ones on the outer row of upper and lower jaws.

Colour in life Males [Fig. 5(c)-(f)], body and head scales light blue to blue-green with dark red margin, creating a reticulated pattern on body and head; on posterior part of body and caudal peduncle scale margins wider than on anterior scales becoming completely red near caudal-fin base. Operculum with three red oblique stripes. Branchiostegal membrane light blue to whitish. Dorsal fin blue-green with irregular red spots and dots distally denser, elongate over fin rays and smaller. Complete light blue margin. Anal fin red, proximally to medially with light blue to blue-green vermiculated patterns of spots and dots, with a black rim. Pelvic fins light blue with red spots and with red tips. Pectoral fins hyaline, light blue distally with grey to black margin. caudal fin dark red with a black narrow margin. Iris golden, with faint black vertical bar through centre of eye. Females, body and head scales pale brown, with light blue iridescence on scale centres. Opercular region light blue to silver. Abdomen silver. All fins hyaline. Iris golden, with faint black vertical bar through centre of eye.

Distribution This species is currently known from ephemeral pools, swamps and ditches on the floodplains of rivers associated with central (as far as Chia in the north, c. 15 km from Nkhota to kota township; 13° 13' S; 34° 18' E) and southern sections of Lake Malawi, areas around Lake Malombe and the upper Shire River region in the Hoba and Liwonde areas close to the Chilwa and Chiuta systems (Fig. 1). The Lake Malawi-Shire system is separated from the Lakes Chilwa and Chiuta regions where N. kirki occurs along a linear stretch of hills east of Lake Malawi from Makanjira through Mangochi Hills to the Machinga-Zomba Mountain.

Habitat note Biotopes of Nothobranchius in Malawi are restricted to the relatively flat, low- lying alluvial flood plains underlain by calcimorphic soils which are alkaline or hydromorphic (water-logged) soils such as in the Chilwa-Chiuta Lakes (Watters,

1991c) (Fig. 2). Like most other African Nothobranchius (Watters, 2009), N. wat tersi occupies seasonal water pools that are briefly connected owing to seasonal flooding by summer rains and isolated later in the season before eventually desic cating. These pools are characterized by fine black or grey-coloured clay-rich soils typically comprising silt, mud and fine sand, roughly in that order of abundance, and eggs are laid in the uppermost layers of this substratum (Watters, 1991d). Much of the known range of N. wattersi and N. kirki is savanna that has now become settled and used for agriculture; pool water is highly inorganically and organically turbid from livestock stirring and defaecation, with total dissolved solids up to 195 mg l-1, pH 6 8-9 6. The general elevation of the plains around southern Lake Malawi where N. wattersi occurs is c. 500 m above sea level (m.a.s.l.), whereas the elevation around Lakes Chiuta and Chilwa areas is 600-630 m.a.s.l. Mean annual rainfall exceeds 2400 mm in the northern part of Lake Malawi and is between 800 and 1100 mm in Chilwa (Nicholson, 1998; Thomas et al., 2009) (c.1200 mm per annum in the area of Salima).

Etymology The species is named in dedication to B. R. Watters, who has studied Malawi Nothobranchius and their biotopes extensively, and has also made significant con tributions to the further understanding of the ecology of Nothobranchius fishes.

DISCUSSION Analysis of sequence data has revealed a clear evolutionary distinction between N. kirki and Nothobranchius from the Lake Malawi-upper Shire system, here described as N. wattersi. Phylogenetic inference by parsimony, likelihood and Bayesian meth ods has recovered the same highly supported topology for both nuclear and mito chondrial locus data treated either separately or together. These analyses cluster fishes from the two basins in separate monophyletic clades. Interestingly, this sepa ration is maintained even when a single nuclear gene, myh6, is examined on its own [Fig. 4(c)]. The geographic clades representing N. wattersi and N. kirki show fur ther differences in within-clade genetic diversity (Table IV). Nothobranchius wattersi from the plains along Lake Malawi shows higher nucleotide and haplotype diversity and is clearly not different from neutrality. Nothobranchius kirki, however, signifi cantly departs from neutrality (Tajima’s D, P < 0 05; and Fu’s Fs, P < 0 01). This is evidence for an excess of alleles, as would be expected from a recent population expansion or genetic hitch-hiking. Excess of haplotypes is likely to reflect the relative stability of the Lake Malawi flood plain compared with the Lake Chilwa plain with respect to dry conditions over geological times. Both Lakes Malawi and Chilwa catchments have historically experienced phases of reduced precipitation and fluctuations in water level (Owen et al., 1990; Nicholson, 1998). In a synthesis of historical and archaeological data of the two lake environments for the past 200 years, Nicholson (1998) found little correlation in their water levels, although drought periods tended to be synchronous. Precipitation in the two basins is affected by different systems (Indian Ocean mon soon for Chilwa v. eastern equatorial monsoon for the Lake Malawi basin) (Thomas et al., 2009). Although both lakes are characterized by small catchment areas, the

Chilwa catchment is relatively much smaller and tends to be more sensitive to local precipitation conditions (Delvaux, 1995). It could be argued though that N. kirki may not be restricted to this catchment nor is its absence proved from the upper Lugenda areas. Indeed a species of Nothobranchius was recently discovered in the Niassa Reserve in Mozambique and formally described as Nothobranchius niassa Valdesalici, Bills, Dorn, Reichwald & Cellerino 2012. The N. kirki and N. wattersi species pair, however, does not appear to be closely related to any other species of Nothobranchius. Nothobranchius niassa is indeed genetically unrelated to N. kirki and N. wattersi and is more similar to Nothobranchius kilomberoensis Wildekamp, Watters & Sainthouse 2005 (Valdesalici et al., 2012). The demographic histories of aquatic fauna in these areas are likely to reflect the episodes of precipitation. Unlike the Chilwa-Chiuta plain, the Lake Malawi plains represent a relatively stable system in terms of availability of moisture for Nothobranchius. Five incidences of recession in Lake Malawi are noted for the period before 25 000 years before present (b.p.) until now (Owen et al., 1990). None of these events, however, involved complete drying up of the lake, making it likely that suitable habitats for Nothobranchius were always available in the receding lake. In contrast, both Lakes Chilwa and Chi- uta completely dried out in the period before 1850; Chilwa desiccated again in the early 1900s, 1940s and more recently in 1968 and 1973 (Nicholson, 1998). Although detailed accounts of similar episodes in the Quaternary context for Chilwa are not available, it is likely that fauna in the Chilwa-Chiuta system has been exposed to more severe drought situations more often in the course of history. Such bottlenecks would explain allele deficiency observed in N. kirki and perhaps point to differential genetic abilities between N. kirki and N. wattersi to withstand prolonged dry spells in diapause. Morphologically N. wattersi and N. kirki are similar with almost all measurement ranges overlapping (differences are apparent only on average, see Table VI) such as meristics and similar frontal scalation. The two species are osteologically similar hav ing a well-developed lateral process of the post-temporal and a reduced anterodorsal process of the urohyal. The two species can, however, be clearly distinguished on the basis of male coloration [see Fig.5(a), (b) for N. kirki and Fig. 5(c)-(f) for N. wattersi ]. Fully coloured males of N. wattersi are distinctively blue-green on their scale centres, with deep red tails including posterior portions of the caudal peduncle. The anal fin is more rounded in N. wattersi compared to wedge-shaped anal fins of N. kirki (Watters, 1991c). Watters (1991c) further observed increasingly deep red tail colour tending towards northern locations. Specimens of both males and females of N. wattersi have relatively longer heads with a more flattened to concave dorsal surface. In general, characters that significantly differ in both sexes suggest larger mean differences between females of the two species. On the other hand, the males clearly have accumulated differences in far more characters than their female coun terparts. This observation is not unexpected for fishes of the Nothobranchius genus which show a clear sexual dimorphism. The two species are allopatric in distribution (see Fig. 1) with N. wattersi restricted to the Lake Malawi drainage system. It is likely that they have evolved separately from at least early Pleistocene times (Watters, 1991c); that is, at least 2 6 M b.p. The lake Malawi Basin is the southernmost section of the African Rift Valley formed through subsidence of the valley floor and rifting of the shoulders which started in the late Miocene (some 8 6 M b.p.), but probably retained water permanently since

0 4 M b.p. (Delvaux, 1995; Won et al., 2005). The Lakes Chilwa-Chiuta depression located some 100 km south-east of Lake Malawi is thought to have formed around the same time probably due to the uplift and eastward tilting of the eastern shoulder of the Lake Malawi rift. This resulted in drainage away from the Lake Malawi system into the Indian Ocean through the Lugenda and Ruvuma River systems (Lancaster, 1981; Watters, 1991c). It is therefore likely that Nothobranchius in the two depressions have diverged separately since the formation of the present major geomorphological features. The similarity of Nothobranchius in the Lake Chilwa and Lake Chiuta systems is expected in the context of the very short period the two lakes have been separated by a 30 km sand bar that runs in the east-west direction (Lancaster, 1981). These lakes were connected prior to the Pleistocene (Lancaster, 1981). Other authors date the isolation to have occurred much more recently. Thomas et al. (2009) estimate the detachment to have existed since 44 000 b.p. According to Nicholson (1998), the sand ridge contains two gaps near its western end through which the Nkonde and Mikoko Rivers may have flowed as late as the 1880s, providing connection between the two systems. Although specimens from the Lake Malawi plains probably represent the same species, there is considerably deep structuring within this group as revealed by mitochondrial DNA and also signalled by nuclear loci, with Golomoti and Salima specimens forming sub-clades. As a future perspective, it would be interesting to examine sequences from all known localities in the N. wattersi range. In particular, specimens from Liwonde and Hoba, the most southerly locations of N. wattersi, are prominently bluer in body colour. Despite specifically targeting these regions for collections of populations, no specimens were caught in April 2009 for molecular analysis. On the basis of colouration, Watters (1991c) has suggested that the Hoba- Liwonde populations might be placed as a subspecies to those occurring further north. To the north-west of these locations are the Phililongwe Hills that terminate in the high ground in the Chiripa and Bilila areas that separate the northward flow of the Bwanje, Kabudira and Lisangadzi Rivers into Lake Malawi, from the Livilidzi River system flowing into the upper Shire plain. This subtle topography may have provided partial isolation of the Malombe-Liwonde sub-system containing Hoba fish.

The authors wish to thank R. B. Watters for materials and images used in the morpho logical description, F. Fontana and R. Sonnenberg for the assistance with the first draft of the manuscript, K. Seitz for expert technical assistance in sequencing and sequence quality control and M. Platzer for generous access to the sequencing facility.

References Amiet, J. L. (1987). Faune du Cameroun. Fauna of Cameroon, Vol. 2, Le genre Aphyosemion Myers. The genus Aphyosemion Myers (Pisces: Teleostei: Cyprinodontiformes). Compiegne: Sciences Naturales. Delvaux, D. (1995). Age of Lake Malawi (Nyasa) and water level fluctuations. Musee royal de l’Afrrique Centrale, Tervuren, Rapport annuel du Departement de Geologie et de Mineralogie 1995-1996, 99-108. Dereeper, A., Guignon, V., Blanc, G., Audic, S., Buffet, S., Chevenet, F., Dufayard, J.-F., Guindon, S., Lefort, V., Lescot, M., Claverie, J.-M. & Gascuel, O. (2008). Phylogeny.

fr: robust phylogenetic analysis for the non-specialist. Nucleic Acids Research 36, W465-W469. Dopazo, J. (1994). Estimating errors and confidence intervals for branch lengths in phyloge netic trees by a bootstrap approach. Journal of Molecular Evolution 38, 300-304. Eck, R. V. & Dayhoff, M. O. (1966). Atlas of Protein Sequence and Structure. Silver Springs, MD: National Biomedical Research Foundation. Excoffier, L., Laval, G. & Schneider, S. (2005). Arlequin (version 3.0): an integrated soft ware package for population genetics data analysis. Evolutionary Bioinformatics 2005, 47-50. Felsenstein, J. (1985). Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39, 783-791. Garavello, J. C. & Shibatta, O. A. (2007). A new species of the genus Pimelodus Lacepede, 1803 from the rio Iguacu basin and a reappraisal of Pimelodus ortmanni Haseman, 1911 from the rio Parana system, Brazil (Ostariophysi: Siluriformes: Pimelodidae). Neotropical Ichthyology 5, 285 -292. Hammer, 0., Harper, D. A. T. & Ryan, P. D. (2001). PAST: paleontological statistics soft ware package for education and data analysis. Palaeontologia Electronica 4, 9. Hoedeman, J. J. (1958). The frontal scalation pattern in some groups of tooth carps (Pisces, Cyprinodontiformes). Bulletin of Aquatic Biology 1, 23-28. Huelsenbeck, J. P. & Ronquist, F. (2001). MRBAYES: Bayesian inference of phylogenetic trees. Bioinformatics 17, 754-755. Hulsey, C. D., Keck, B. P. & Hollingsworth, P. R. Jr. (2011). Species tree estimation and the historical biogeography of heroine cichlids. Molecular Phylogenetics and Evolution 58, 124-131. Jubb, R. A. (1975). The Nothobranchius species of Malawi, Africa. Journal of American Killifish Association 8, 170-172. Jubb, R. A. (1981). Nothobranchius. Neptune City, NJ: T. F. H. Publications Inc. Kawahara, R., Miya, M., Mabuchi, K., Near, T. J. & Nishida, M. (2009). Stickleback phylogenies resolved: evidence from mitochondrial genomes and 11 nuclear genes. Molec ular Phylogenetics and Evolution 50, 401-404. Lancaster, N. (1981). Formation of the Holocene Lake Chilwa sand bar: Southern Malawi. Catena 8, 369-382. Li, C., Ortf, G., Zhang, G. & Lu, G. (2007). A practical approach to phylogenomics: the phylogeny of ray-finned fish (Actinopterygii) as a case study. BMC Evolutionary Biology 7, 44. Li, C., Ortf, G. & Zhao, J. (2010). The phylogenetic placement of sinipercid fishes (“Per- ciformes”) revealed by 11 nuclear loci. Molecular Phylogenetics and Evolution 56, 1096-1104. Murphy, W. J. & Collier, G. E. (1998). Phylogenetic relationships of African killifishes in the genera Aphyosemion and Fundulopanchax inferred from mitochondrial DNA sequences. Molecular Phylogenetics and Evolution 11, 351-360. Nei, M. & Kumar, S. (2000). Molecular Evolution and Phylogenetics. New York, NY: Oxford University Press. Nicholson, S. E. (Ed.) (1998). Fluctuations of Rift Valley Lakes Malawi and Chilwa during Historical Times: A Synthesis of Geological, Archaeological and Historical Information. Dordrecht: Kluwer Academic Publishers. Owen, R. B., Crossley, R., Johnson, T. C., Tweddle, D., Kornfield, I., Davison, S., Eccles, D. H. & Engstrom, D. E. (1990). Major low levels of lake Malawi and their impli cations for speciation rates in cichlid fishes. Proceedings of the Royal Society B 240, 519-553. Ratnasingham, S. & Hebert, P. D. N. (2007). BOLD: the barcode of life data system (www.barcodinglife.org). Molecular Ecology Notes 7, 355-364. doi: 10.1111/j.1471- 8286.2006.01678.x Reichard, M. (2010). Nothobranchius kadleci (Cyprinodontiformes: Nothobranchiidae), a new species of annual killifish from central Mozambique. Zootaxa 2332, 49-60. Ronquist, F. & Huelsenbeck, J. P. (2003). MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19, 1572-1574.

Rzhetsky, A. & Nei, M. (1992). A simple method for estimating and testing minimum- evolution trees. Molecular Biology and Evolution 9, 945-967. Saitotu, N. & Nei, M. (1987). The neighbor-joining method: a new method for reconstructing phylogenetic trees. Molecular Biology and Evolution 4, 406-425. Scheel, J. J. (1968). Rivulins of the Old World. Neptune City, NJ: Tropical Fish Hobbyist Publication. Scheel, J. J. (1990). Atlas of Killifishes of the Old World. Neptune City, NJ: T. F. H. Publi cations, Inc. Seegers, L. (1997). Killifishes of the World. Old World Killis II. Morfelden-Walldorf: Verlag A.C.S. Shidlovskiy, K. M., Watters, B. R. & Wildekamp, R. H. (2010). Notes on the annual killifish species Nothobranchius rachovii (Cyprinodontiformes; Nothobranchiidae) with the description of two new species. Zootaxa 2724, 37-57. Sneath, R. P. H. & Sokal, R. R. (1973). Numerical Taxonomy. San Francisco: Freeman. Staden, R. (1996). The staden sequence analysis package. Molecular Biotechnology 5, 233-241. StataCorp. (2009). Stata Statistical Software: Release 11. College Station, TX: StataCorp LP. Tamura, K. & Kumar, S. (2002). Evolutionary distance estimation under heterogeneous substi tution pattern among lineages. Molecular Biology and Evolution 19, 1727-1736. Tamura, K., Nei, M. & Kumar, S. (2004). Prospects for inferring very large phylogenies by using the neighbor-joining method. Proceedings of the National Academy of Sciences of the United States of America 101, 11030-11035. Tamura, K., Dudley, J., Nei, M. & Kumar, S. (2007). MEGA4: Molecular Evolutionary Genetics Analysis (MEGA) software version 4.0. Molecular Biology and Evolution 24, 1596-1599. Thomas, D. S. G., Bailey, R., Shaw, P. A., Durcan, J. A. & Singarayer, J. S. (2009). Late Quaternary highstands at Lake Chilwa, Malawi: frequency, timing and possible forcing mechanisms in the last 44 ka. Quaternary Science Reviews 28, 526-539. Valdesalici, S., Bills, R., Dorn, A., Reichwald, K., & Cellerino, A. (2012). Nothobranchius niassa (Cyprinodontiformes: Nothobranchiidae), a new species of annual killifish from northern Mozambique. Ichthyological Exploration of Freshwaters 23, 19-28. Watters, B. R. (1991a). The Nothobranchius of Malawi. Part I. Journal of American Killifish Association 24, 126-137. Watters, B. R. (1991b). The Nothobranchius of Malawi. Part II. Journal of American Killifish Association 24, 138-185. Watters, B. R. (1991c). The Nothobranchius of Malawi. Part III. Journal of American Killifish Association 24, 186-193. Watters, B. R. (1991d). Nothobranchius habitats in Malawi. Part IV. Journal of American Killifish Association 24, 194-201. Watters, B. R. (2009). The ecology and distribution of Nothobranchius fishes. Journal of American Killifish Association 42, 37-76. Wildekamp, R. H. (2004). A World of Killies. Atlas of the Oviparous Cyprinodontiform Fishes of the World. Elyria: The American Killifish Association. Won, Y.-J., Sivasundar, A., Wang, Y. & Hey, J. (2005). On the origin of Lake Malawi cichlid species: a population genetic analysis of divergence. Proceedings of the National Academy of Sciences of the United States of America 102, 6581-6586.

Electronic Reference Kingsley Lab Protocols (2004). Isolating Stickleback Tail DNA (Version 1.0; 01/27/04). Avail able at http://kingsley.stanford.edu/Lab_Protocols.html/ (accessed 20 April 2009).

APPENDIX COMPARATIVE MATERIAL EXAMINED SAIAB 98797 (field voucher no. NTJ 09); 10 males, 27 0-31 5 mm standard length (Ls), 7-8-9-7 mm head length (Lh), Ntaja, Malawi, in rice field accessed by canoe, Lake Chilwa marginal swamps that receed in dry season, 15° 00 848' S; 35° 34 978' E; east Ng’oma, 4 April 2009. Pe0157 (voucher no. MW 88-9); one male, 33 8 mm Ls, 10 4 mm Lh, Lake Chiuta, Malawi, very shallow pool in ditch at roadside, c. 5 km west of northern end of Lake Chiuta, 3 km south-east of village of Nkhokwe on road to Mpakaka, 14° 44' S; 35° 49' E; B. R. Watters, 24 March 1988. Pe0158 (field voucher no. MW 91-15); seven males, 28-5-43-3 mm Ls, 8-8-13-9 mm Lh , one female 34 5 mm Ls, 10 8 mm Lh , Ntaja, Machinga, Malawi, very shallow pool in ditch at roadside, c. 5 km west of northern end of lake Chiuta, 3 km south-east of village of Nkhokwe on road to Mpakaka, 14° 44' S; 35° 49' E; B. R. Watters, 6 April 1991. Pe0159 (field voucher no. MW 88-10); one male 35 0 mm Ls, 119 mm Lh , Zomba, Malawi, pools in a temporary stream system, at a bridge on road S69, 23 km east of Zomba and c. 7 km west of Kachulu, 15° 24' S; 35° 32' E; B. R. Watters, 27 March 1988. MRAC B1-14-P-91-97, (field voucher no. KCH 09); six males, 34-8-39-2 mm Ls, 9-3-12-1 mm Lh, one female 391 Ls, 114 mm Lh ; three females, 9-4-11-1 mm Lh, at Njala Rice Scheme, Kachulu Malawi, 15° 23 351' S; 35° 31 952' E; east Ng’oma, 1 April 2009 [Fig. 5(a), (b) and Table VI; Nothobranchius kirki (Jubb 1969)].