Population genomic analysis reveals cryptic population structure in the commercially important Lake Malawi cichlidCopadichromis mloto

22 23 Fish is an important source of animal protein for many people living around Lake Malawi. The evaluation of 24 population structure and genetic diversity can yield useful information for management and conservation of fish 25 species but is complicated in Lake Malawi by the close genetic relatedness of species of the recent cichlid adaptive 26 radiation. In this study, we analysed whole-genome sequencing data of "true utaka", a group of cichlids previously 27 common in fisheries, but that has faced strong decline due to overfishing. Our analysis of 223 individuals collected 28 from fishermen’s catches along the western shoreline of Lake Malawi confirmed that Copadichromis mloto (C. sp. 29 "virginalis kajose") is the true utaka most targeted by fisheries. Genetic principal component analysis, phylogenetic 30 inference, and admixture analysis revealed complex patterns of population structure. The presence of at least three 31 geographically widespread genetic clades that have remained separate despite gene flow and partial sympatry hints 32 at the presence of currently undescribed, cryptic species diversity in C. mloto. This result leads us to suggest that, 33 despite the lack of obvious habitat barriers, benthic and pelagic species of Malawi cichlids might harbour 34 unidentified species diversity and calls for further genetic and taxonomic research to define appropriate conservation 35 units. 36 37

Population genomics; Conservation genetics; Cichlid fishes, Lake Malawi, Fishing pressure; 38 Copadichromis sp.; Utaka 39 40 41 42 43 44 45 46 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint 47 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint

48 49 Cichlid fishes of the Great Lakes of East Africa, particularly Lakes Victoria, Malawi, and Tanganyika are significant 50 study systems in evolutionary biology because of their rapid adaptive radiation (Anseeuw et al. 2008; Brawand et al. 51 2015; Rick et al. 2022). At the same time, Malawi cichlids are a source of food for millions and provide a livelihood 52 for thousands of rural and urban Malawians who participate in harvest and postharvest activities including fishing, 53 fish processing, transportation to the markets, and marketing (Weyl et al. 2004). Stocks of Malawi cichlids have 54 been dramatically declining over the past 40 years, probably as a consequence of this exploitation (Van Zwieten et 55 al. 2011; Ngochera et al. 2018; Chavula et al. 2023), calling for sustainable fisheries management strategies. 56 However, monitoring of exploited populations has been severely constrained by difficulties in species identification, 57 a particular challenge for the Lake Malawi cichlid radiation with its ~1000 closely related and taxonomically poorly 58 resolved species (Svardal et al. 2020). Cryptic species diversity can mask changes in community composition, 59 interactions between different fishery sectors and gears and render studies of population structure unreliable, as 60 apparent genetic differentiation may instead represent non-random sample mixes of different species collected at 61 different locations. 62 63 Previous studies on the genetic structure of Malawi cichlids suggest that geographic distance and habitat barriers 64 contributed to population structure in rock-dwelling species such as the species-rich clade of Mbuna (Van Oppen et 65 al. 1997; Danley et al. 2000) or Protomelas (Pereyra et al. 2004), but found no or only weak population structure 66 within the more continuously distributed species living over sandy bottom or in the open water (Shaw et al. 2000; 67 Taylor & Verheyen, 2001; Anseeuw et al. 2011), which constitute the main target of fisheries. However, traditional 68 genetic markers as used in these studies (mtDNA and microsatellites) may lack power to identify cryptic species 69 (Malinsky et al. 2018), which can lead to misleading population structure estimates. Conversely, recent whole 70 genome studies on Malawi cichlids have proven powerful in resolving species relationships (Malinsky et al. 2018; 71 Masonick et al. 2022; Scherz et al. 2022), but were limited to one or few samples of each species and thus not able 72 to resolve geographic structure within species. As such, we lack a detailed understanding of the patterns of 73 population structure of Malawi cichlid species. 74 75 The semi-pelagic utaka ( Copadichromis sp. (Turner, 1996) have traditionally been among the economically most 76 important fish species in Lake Malawi and Lake Malombe, a highly productive and heavily fished lake at the 77 outflow of Lake Malawi (Tweddle et al. 1995; Weyl et al. 2004; Kanyerere et al. 2018). Among the most common 78 Utaka in fishery catches are a group known as the ‘pure utaka’ which lack dark spots on their flanks, described as C. 79 mloto (slender-bodied) and C. virginalis (deeper bodied), but long recognised as posing considerable taxonomic and 80 identification issues (Iles, 1960; Eccels & Trewavas, 1989). The ‘pure utaka’ most common in the fisheries of both 81 lakes, previously often referred to as C. sp. 'virginalis kajose', has recently been re-identified as C. mloto (Turner et 82 al. 2022). This species seems to be ecologically differentiated from C. virginalis , (previously referred to as C. 83 'virginalis kaduna'; C. ilesi; C. sp. ‘firecrest mloto’) by its habit of breeding over the open sandy or muddy substrate, 84 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint rather than on rocky coast (Turner et al. 2022). The stocks of C. mloto have dramatically declined in Lakes Malawi 85 and Malombe over the past decades, likely as a result of overfishing (Banda et al. 2001; Weyl et al. 2004; Anseeuw 86 et al. 2011; Chavula et al. 2023). 87 88 In this study, we analysed whole genome sequences of 203 specimens of C. mloto collected with a research trawler 89 and from fisheries catches along the Malawian waters of Lake Malawi and from Lake Malombe, covering a 90 latitudinal extent of more than 480 kilometres. We confirm the monophyly of C. mloto with respect to C. virginalis. 91 Analysis of population structure revealed the presence of three genetically distinct but geographically overlapping 92 clades within C. mloto, hinting at the presence of hitherto cryptic taxonomic diversity, yielding important 93 information for the conservation and management efforts of these populations. Gene flow analysis confirmed 94 historic genetic exchange both between C. virginalis and C. mloto and among the different C. mloto clades, the latter 95 showing correlation with geographic proximity. Furthermore, we found similar levels of genetic diversity across C. 96 mloto populations and a strong excess of rare genetic variants consistent with historic population growth but could 97 not detect any discernible effect of fishing pressure on population genetic statistics. Finally, we discuss tentative 98 evidence for differences in male breeding colouration between the observed genetic clades. 99 100

101 102 Sample collection 103 Specimens of Copadichromis mloto (n = 202) and C. virginalis (n = 21) were collected along the western shoreline 104 of Lake Malawi, its southeast and southwest arms, and Lake Malombe (Supplementary Table 1, Fig. 1) during field 105 expeditions in 2002 (Anseeuw et al. 2011), 2016, and 2017 (Supplementary Table 1). The samples were caught by 106 local fishermen using a variety of methods, including beach seine nets, open water seines (Chir imila nets in Lake 107 Malawi), purse seines (Nkacha nets in Lake Malombe) and trawl ships (both commercial and the Ndunduma 108 research trawler of the Malawi Department of Fisheries) (Supplementary Table I). The specimens' right pectoral fins 109 were preserved in pure ethanol and stored in a freezer. A subset of whole fish specimens was fixed in 10% formalin 110 and preserved in ethanol; specimens are being kept at the Royal Museum for Central Africa in Tervuren, Belgium 111 (2002 samples) or the Zoology Museum Cambridge (2016-2017 samples). However, we note that for the specimens 112 stored in the Royal Museum for Central Africa no 1:1 match to fin clips is available. 113 114 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint 115 116 Figure 1. Sampling information. (a) Location of Lake Malawi on the African continent (left) and zoom in of Lakes117 Malawi and Malombe (right), showing the number of samples and sampling locations of Copadichromis mloto118 (circles) and C. virginalis (triangle). Location names are given in Fig. 2A and Supplementary Table 1. (b )119 Representative photos of males in breeding colouration from the locations Nkhotakota, Monkey Bay Chenga120 trawler, Msaka, Malembo and Lake Malombe (top to bottom). 121 122 Sequencing, variant calling and filtering 123 DNA was extracted and libraries were prepared at the Wellcome Sanger Institute using standard protocols. Whole -124 genome sequencing was performed on an Illumina HiSeq platform to individual coverages ranging from 7.2- fold125 and 42.1-fold. Resulting short read sequences were aligned to the Astotilapia calliptera reference genome126 fAstCal1.2 ((GCA_900246225.3 https://www.ncbi.nlm.nih.gov/assembly/GCF_900246225.1) using BWA- MEM127 (BWA version 0.7.17). Furthermore, bcftools (version 1.14) was used to detect variant sites (Danecek et al. 2021)128 Sites were masked if the overall mapping quality was less than 50 or if more than 10% of the mapped reads had a129 mapping quality of zero, diagnostic of non-unique mapping. In addition, sites for which mapping quality was130 significantly different between the forward and reverse strand (P97.5 percentile) or low (20) were set to missing , representing between 0.02% and 0.15% of134 es to ) ga - ld me M . a as for e). ce of .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint heterozygous sites per individual. Additionally, sites with excess heterozygosity (dataset-wide inbreeding coefficient 135 20% missing genotypes were excluded from analysis (Supplementary Figs VI, VII and VIII). 136 Only biallelic single nucleotide polymorphisms (SNPs) were retained for further investigation, while recently 137 discovered inversion regions on five chromosomes (total 104 megabase pairs [Mbp]) were excluded from further 138 analysis (Supplementary Table II, Blumer et al. in preparation). Furthermore, we excluded ten samples because they 139 showed abnormally high heterozygosities reminiscent of cross contamination, or were inferred to be close relatives. 140 141 Population genetic structure 142 For this analysis, we only considered SNP variants with a minor-allele count of at least 3 and less than 25% missing 143 genotypes filtered for linkage-disequilibrium in PLINK (version 1.9) with r 2 > 0.2 in sliding windows of 50 Single 144 Nucleotide Polymorphism (SNPs), with 10 SNP overlap. We used PLINK to run a principal component (PC) 145 analysis calculating eigenvectors and eigenvalues separately for the whole dataset, including C. virginalis and C. 146 mloto, and for C. mloto only (Purcell et al. 2007). R (version 4.2.2) (R Core Team, 2021) was used to plot the 147 eigenvectors. 148 149 Population admixture analysis 150 We used the LD pruned sites mentioned above to create binary PLINK (.bed) files. The .bed files were used to 151 perform admixture analysis using Bayesian clustering in ADMIXTURE v1.3.0 (Alexander et al. 2009) with values 152 of K ranging from 1 to 5. We selected the cluster K value with the lowest corresponding CV error. Python (version 153 3) (Van Rossum & Drake, 2009) was used to plot the results of the admixture analysis. The admixture analysis 154

were plotted and visualised using the Pong workflow analysis (https://github.com/ramachandran-lab/pong). 155 156 Neighbour-joining tree construction 157 To construct a neighbour-joining (NJ) tree of genetic distances, we calculated total pairwise genetic differences on 158 the genome-wide SNP loci of the samples using a custom script based on Scikit allel and the Biophyton 1.79 phyllo 159 package for tree construction (https://github.com/feilchenfeldt/pypopgen3 ). The tree was rooted using an ancestral 160 sequence inferred by whole genome alignment of the reference genome and outgroup species (Camacho et al. in 161 prep.). 162 163 Population genetic statistics 164 Genetic diversity statistics were calculated for populations of C. mloto and for the single population of C. virginalis 165 from Nkhata Bay. The inbreeding coefficient (F) per individual, and nucleotide diversity per site and Tajima’s D in 166 windows of 100kbp were calculated using vcftools (version 0.1.16) (Danecek et al. 2021). Furthermore, the allele 167 frequency spectra were calculated for the population of each C. mloto clade with the largest sample size using a 168 custom script. 169 170 Gene flow test in Copadichromis species 171 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint To investigate potential gene flow between C. virginalis and C. mloto and between different C. mloto populations, 172 we computed four population tests such as Patterson’s D and the f4 (admixture) ratio ( also known as the ABBA-173 BABA statistics) as implemented in D-suite (Malinsky et al. 2021), setting the above-mentioned ancestral sequence 174 as outgroup. We defined as populations all the samples of a given sampling location assigned to one of the identified 175 genetic clades (mloto A/B/C, putatively admixed, C. virginalis). For analysis, we selected all f4 ratio tests comparing 176 populations from specific clades, e.g., P1 = population from clade mloto A; P2 = population from clade mloto B; P3 177 = C. virginalis, etc. Finally, to assess geographic signal in f4 ratio tests we obtained sampling coordinates either from 178 available field records or, where not available, by locating reported sampling sites on online maps, and then 179 calculated geographic distances using the function distance.geodesic implemented in the Python package geopy 180 (version 2.3.0). 181 182

183 Variant detection 184 Variant calling yielded approximately 41 million single nucleotide polymorphisms (SNPs) across 21 C. virginalis 185 and 203 C. mloto specimens of which ~31 million ( 74.7%) passed all variant filtering criteria. The accessible 186 genome size excluding recently discovered inversion regions (see methods) was 545 Mbp corresponding to a SNP 187 density of 75 SNPs per kbp. 188 189 Population genetic structure provides evidence of three geographically widespread C. mloto clades 190 Performing principal component (PC) analysis and constructing a neighbour-joining (NJ) tree of pairwise 191 differences of all samples, we confirmed that all C. mloto formed a monophyletic clade with respect to the samples 192 identified as C. virginalis (Supplementary Fig. V). The first PC clearly separated C. mloto samples from all 193 sampling locations from the single C. virginalis population from Nkhata Bay (10 % variance explained). Since our 194 collections from fisheries catches did not initially distinguish between the two species, this result suggests that C. 195 mloto is the dominant pure utaka species in commercial and artisanal fisheries of Lakes Malawi and Malombe. For 196 further analyses, we focused on the 203 specimens belonging to C. mloto. 197 Structure/admixture analysis of C. mloto samples revealed an interesting pattern in that identified ancestries did not 198 perfectly reflect geographic proximity. Although most individuals of the same sampling location were attributed to 199 the same cluster, the different clusters spanned a wide geographic range with individuals from non-adjacent 200 sampling locations being attributed to the same cluster, while geographically intermediate populations were 201 attributed to different clusters (Fig. 2AThis trend was consistent for different choices of ancestral populations (K) 202 (Supplementary Fig. IV). For several populations, some individuals were inferred to draw some or all of their 203 ancestry from a cluster different from the majority of individuals, consistent with the presence of distinct local 204 populations and, in the case of partial ancestry, gene flow, which will be further investigated below. 205 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint A PC analysis of C. mloto individuals broadly supported the results of the admixture analysis in that it identified 206 genetic clustering inconsistent with a simple isolation by distance pattern (Fig 2B). Instead, the first PC separated a 207 clade consisting of most Lake Malombe individuals, all individuals from Nkhudzi Bay on the western shoreline of 208 the southeast arm of Lake Malawi, all individuals from Nkhotakota on the central western shore of Lake Malawi, 209 some individuals from Namiasi Palm Beach at the southernmost tip of the south east arm of Lake Malawi as well as 210 a single individual from Malembo in the south west arm of Lake Malawi. These individuals formed a monophyletic 211 sister clade to all other individuals in an NJ tree of C. mloto specimens (Fig. 3). In the following, we refer to this 212 genetic cluster as mloto A clade. 213 The second PC separated samples more gradually, with individuals from the northern most population at Chilumba 214 and most individuals from the Namiasi Palm Beach population at the southern tip of Lake Malawi at the two 215 extremes and the mloto A clade that separated along PC1 at values close to zero. However, the clustering of other 216 populations did not reflect a clear geographic trend: individuals with a wide range of geographic origins clustered 217 next to the Namiasi Palm Beach cluster, including samples from Chiweta, Nkhata Bay, and Senga Bay along the 218 western shore of Lake Malawi as well as individuals from the southeast and southwest arm populations. Conversely, 219 individuals from other south west arm populations, as well as from Makanjila on the south eastern coast of Lake 220 Malawi, and a single sample from Namiasi Palm Beach, clustered in proximity to the northern most samples from 221 Chilumba. In the NJ tree, the described spread along PC2 corresponds to two reciprocally monophyletic clades, 222 which we refer to as clades mloto B and mloto C, in the following, for the clades including samples from Chilumba 223 (negative PC2) and Nkhata Bay (positive PC2), respectively. An exception to this are two samples from Msaka on 224 the eastern shores of the southwest arm that clustered basally to both mloto B and mloto C and intermediate along 225 PC2. This is consistent with admixed ancestry between the clades, as also inferred by Admixture (Fig. 2A), and 226 further investigated below. 227 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint 228 Figure 2. Relationships among the Copadichromis mloto populations visualised by population Admixture and229 Principal Component (PC) analysis. (a) Population structure patterns of C. mloto as inferred by Admixture230 (Alexander et al. 2009) assuming K = 3 ancestral populations; for the other K values see Supplementary Fig. IV. (b)231 Principal Component (PC) analysis of C. mloto populations. 232 nd re ) .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint 233 234 Figure 3. Neighbour-joining tree constructed from pairwise distances showing the three genetic clades of C. mloto235 The branches corresponding to the three genetic clades are coloured in red ( mloto A), green ( mloto B), and purple236 (mloto C), respectively. Black branches correspond to the ancestral sequence used for rooting and to putatively237 admixed individuals. Relative branch length is fixed for better visibility and thus not proportional to genetic distance238 (See Supplementary Fig. XII for a tree with distances). Representative photos of breeding males from each genetic239 clade are given (Fig. 6). 240 241 Differential genetic exchange between Copadichromis virginalis and C. mloto populations 242 243 To test whether the evolutionary separation between C. virginalis and C. mloto corresponded to a clean bifurcation244 or whether C. virginalis and the identified subclades and populations of C. mloto continued to exchange genetic245 material, we calculated the f4 admixture ratio – a measure of excess allele sharing of a clade P2 with a cl ade P3,246 relative to P2's sister clade P1. Specifically, we estimated a potential contribution of C. virginalis to a given C. mloto247 population compared to another C. mloto population, where we defined populations as all samples of a specific248 sampling location that fall into the same C. mloto clade as identified above. The test revealed strong (5- 25%) and249 highly significant excess allele sharing of all mloto C populations with C. virginalis compared to all mloto A250 . ple ly ce tic on tic 3, to fic nd A .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint populations and most mloto B populations (Fig. 4a). Furthermore, the mloto C populations from Chilumba, Mbenji 251 Islands, and Namiasi Palm Beach showed 9-10% excess allele sharing with C. virginalis compared to other mloto C 252 populations. Finally, mloto B populations showed 4-10% excess allele sharing with C. virginalis compared to mloto 253 A populations. Overall, these results suggest differential exchange of genetic material between C. virginalis and 254 different C. mloto populations, with the highest levels of C. virginalis contributions in some mloto C populations 255 and the lowest in mloto A. 256 257 Evidence for local genetic exchange between C. mloto clades 258 Next, we wanted to test whether the three C. mloto clades are fully reproductively isolated entities or whether gene 259 flow has occurred between the clades after their initial sp lit. In the former case of clean splits between the clades 260 without subsequent gene flow, we would expect that individuals from two given clades are equally closely related to 261 each other irrespective of their geographic origin, while in the latter case of cross-clade genetic exchange, we would 262 expect cross-clade excess allele sharing of the populations involved, meaning that different populations of one clade 263 show variation in their genetic distance to an outgroup population. We first checked for signals of excess allele 264 sharing of the two samples from Msaka which clustered basally to the mloto B and C clades and found that these 265 samples showed highly elevated f4 admixture ratios of up to 42% with mloto A populations relative to most mloto B 266 and C populations, while also generally closer to mloto B than to mloto C populations. This provides further 267 evidence for the admixed status of these two samples with ancestries related to mloto A and mloto B clades as 268 already suggested by the Admixture analysis. 269 270 Overall, we found strong evidence for genetic exchange between the different C. mloto clades with 49% of f4 ratio 271 tests significant above multiple testing (Bonferroni FWER < 0.05). If this pattern were due to relatively recent gene 272 flow, we would expect a trend in which individuals from different clades are relatively more closely related to each 273 other if they originate from nearby sampling locations. To investigate this, we tested whether excess allele sharing 274 between populations of different clades depended on their relative geographic distance. We found that most clade 275 comparisons showed a significantly positive correlation between relative geographic proximity and signatures of 276 excess allele sharing (Pearson's r = 0.25-0.38) except for comparisons that tested excess allele sharing of different 277 mloto A or mloto C populations with mloto B, which showed no significant correlation with geographic location 278 (Fig. 4.b). Taken together, our results are consistent with the presence of three widely distributed groups of C. mloto 279 that have (occasionally) been exchanging genetic material in places where they were in contact, but st ill retain their 280 separate genetic identities. 281 282 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint 283 Figure 4. Dtrios for gene flow test in Copadichromis mloto and C. virginalis populations. (a) f4 admixture ratio tests284 of the form f4 (P1, P2, P3, Outgroup) in which P1 and P2 are C. mloto populations and P3 C. virginalis, segregated285 by C. mloto clade adherence of P1 and P2. (b) f4 admixture ratios plotted against relative geographic proximity286 among the Mloto clades. Points are coloured by C. mloto clade adherence of P1, P2, P3 populations in tests f4 (P1,287 P2, P3, Outgroup) (See Methods). 288 sts ted ity 1, .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint 289 Figure 5. Population genetic summary statistics (a) Nucleotide diversity ( ) per bp calculated in windows of290 1Mbp. (b) Average inbreeding Coefficient (F) per population (with n>1). (c ) Tajima’s D in windows of 100kb per291 population (with n>1). Observed allele frequency spectra for (d) the Lake Malombe population of clade mloto A; (E)292 the Namiasi palm beach population of clade mloto B , and (f) the Msaka population of clade mloto C . Black293 dots/lines show the respective expectations for neutrally evolving populations of constant size. 294 295 C. mloto populations show strong excess of rare genetic variants 296 297 Investigating population genetic summary statistics, we found relatively similar levels of nucleotide diversity among298 populations and clades (Fig. 5A, supplementary Table XI). The relatively largest variation is seen among mloto C299 populations with the relatively smallest value of 0.1336% ± 0.0139 for the Chilumba population and the largest300 value of 0.1414% ± 0.0149 for the Nankhwali trawler population. Unexpectedly, despite its peripheral geographic301 location and recent demographic changes due to fishing, the mloto A population from Lake Malombe (0.1398% ±302 of er E) ck ng C est hic ± .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint 0.0141) showed one of the highest values of genetic diversity. With 0.1365% ± 0.0149, the estimated nucleotide 303 diversity for C. virginalis was at the lower end of values measured for C. mloto populations. 304 305 Inbreeding coefficients, measuring a deficiency of heterozygote genotypes relative to Hardy-Weinberg expectations, 306 were moderately positive in all populations (Fig. 5B), consistent with either moderate degrees of (historic) mating 307 between related individuals or population substructuring. However, the fact that relatively high inbreeding 308 coefficients coincide with relatively low nucleotide diversity – the highest values most variable along the genome 309 being seen in the least diverse Chilumba mloto C – suggests that inbreeding coefficients reflect within population 310 demographic events rather than residual population structure. C. virginalis showed the second largest inbreeding 311 coefficient with little variation of it along the genome, a signal which might point to a historic population bottleneck. 312 313 To gain further insight into how utaka genetic diversity patterns have been shaped by past demographic events, we 314 computed Tajima's D, a summary of the distribution of allele frequencies (Fig. 5C). For the population of each clade 315 with largest sample size, we also computed full site frequency spectra (SFS) and compared them to expected spectra 316 under neutral evolution (Fig. 5D-F). These analyses revealed an excess of rare genetic variants (negative Tajima's D) 317 compared to neutral expectations in all populations, with strongly negative Tajima's D values close to -2 being found 318 for populations across the C. mloto clades and also for C. virginalis. Such an excess of rare genetic variants can for 319 example be caused by historic population expansion or by directional selection. That said, the presence of negative 320 values of Tajima's D at a genome-wide scale, without large genomic variation (small error bars in Fig. 5C), suggests 321 that the observed strong excess of rare genetic variants is mainly driven by strong population expansion in the time 322 frame of the (largely shared) coalescent history of present day populations, rather than by positive selection, which 323 would lead to genomically more localised allele frequency changes. 324 325 Colour of male breeding dress varies with genetic clade membership 326 327 To investigate potential morphological differences between the different clades, we qualitatively assessed variation 328 in male breeding dress across C. mloto clades based on photographs (Fig. 6). Male breeding dress is a key trait in 329 reproductive isolation of Malawi cichlids that often differs between sister species (Maan and Sefc, 2013). 330 Unfortunately, we only had photos or preserved specimens available for a subset of the individuals, namely 17 males 331 of clade mloto A (all except one from Lake Malombe), 20 males of clade mloto C, and only two males of mloto B . 332 Although the present sample size is too small for quantitative conclusions, the data suggests that the general pattern 333 of black body and yellow dorsal fin colouration of breeding males common to many utaka shows clade specific 334 variation. Specifically, all examined males of mloto A showed clear yellow colouration in the dorsal fin covering the 335 whole fin at the anterior end and extending relatively far towards the posterior end at the upper fin margin (while the 336 lower margin has a blackish colour). At the same time, mloto A males generally featured a yellow blaze on the 337 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint forehead above the eyes, a pattern that can be hard to spot on photographs but is very apparent in examined338 specimens. Therefore, we tentatively refer to the mloto A breeding dress morphotype as "Yellow Head Yellow339 Dorsal" (YHYD). Breeding males of mloto C showed a dorsal fin colouration generally similar to mloto A, albeit340 with more variation in the relative amounts of black and yellow, but they generally do not show any discernible341 yellow blaze on their forehead, a morphotype to which we tentatively refer as "Black Head Yellow Dorsal"342 (BHYD). Finally, the two mloto B in male breeding dress, one from Nankhwali and one from Chiweta, showed a343 strong yellow blaze on their forehead but no or only very little yellow in the ir dorsal fin, wherefore we refer to their344 morphotype as "Yellow Head Black Dorsal" (YHBD). In summary, we found tentative evidence for clade- specific345 variation in male breeding dress, but a more comprehensive set of matched sequence and phenotype data wil l be346 necessary to scrutinise morphological differentiation among genetic clades. 347 348 349 350 351 Figure 6. Representative males in breeding colouration in the three cryptic clades of Copadichromis mloto. A)352 Copadichromis mloto Yellow Head Yellow Dorsal (YHYD) “mloto A clade”, B) Copadichromis mloto Yellow Head353 ed w eit ble al" a eir fic be A) ad .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint Black Dorsal (YHBD) “ mloto B clade”, C) Copadichromis mloto Black Head Yellow Dorsal (BHYD) “ mloto C 354 clade”. 355 356

357 An accurate understanding of genetic relationships, geographic distribution, and evolutionary history among 358 populations plays a crucial role in the management of species of conservation concern. The pure utaka were recently 359 assessed as near threatened due to a 40% decrease in catch rates between 1999 and 2016 likely due to intense fishing 360 pressure (Konings, 2019). As we confirm here, this assessment probably mainly concerns C. mloto (previously also 361 referred to as C. sp. 'virginalis kajose') – the pure utaka most common in artisanal and commercial catches and the 362 main focus of this study. Here, we assess the population structure of C. mloto based on an almost lake-wide sample 363 of whole genome sequences derived from fisheries catches and find – despite the general close relatedness typical 364 for members of the Lake Malawi cichlid adaptive radiation – clear evidence for the presence of three genetically 365 distinct groups. All three groups seem widely distributed with largely overlapping ranges and local populations that 366 show evidence for variable levels of genetic exchange both with C. virginalis and with geographically proximate 367 populations of other C. mloto groups. Together, this suggests that evolutionary relationships in the utaka are even 368 more complicated than previously thought, with C. mloto alone deserving the recognition of at least three separate 369 taxonomic entities. 370 371 The fact that, despite overlapping ranges, most specimens collected from a given fisheries catch pertained to the 372 same C. mloto clade may indicate either a tendency to school according to species or that there is some kind of 373 habitat partitioning, but the presence of exceptions to this trend and of apparently admixed individuals corroborates 374 at least some degree of sympatry and occasional interbreeding. C. mloto are known to have defined breeding seasons 375 when they aggregate in sandy or muddy habitats where males build and defend mating platforms (bowers), while 376 outside of breeding season they follow a semi-pelagic lifestyle mainly feeding on plankton in inshore open-water 377 habitats (Turner, 1996; Konings, 1999; Anseeuw et al. 2008). A difference in t iming or location of breeding 378 aggregations might contribute to the maintenance of differentiation between the clades, but further research will be 379 needed to better understand ecology and life history of C. mloto groups – an endeavour timely and urgent given the 380 worrisome conservation status of this species and continuing exploitation. 381 382 Despite the relatively isolated geographic position of Lake Malombe, being connected to Lake Malawi through the 383 ~15km long Upper Shire river, its C. mloto population clustered together with populations from Nkhudzi Bay in the 384 south east arm and Nkhata Bay at the central east coast of Lake Malawi, together forming clade mloto A. Given that 385 Lake Malombe had completely dried out and only refilled in the 1930s (Tweddle et al. 1995), it is possible that C. 386 mloto from Lake Malombe are relatively recent migrants from Lake Malawi, which is consistent with observations 387 that the Upper Shire River is used by different species to migrate between the lakes (FAO, 1993). However, the fact 388 that, with the exception of a single specimen, the population from Namiasi Palm Beach collected less than four 389 kilometres from Lake Malawi's drainage into the Upper Shire belongs to a completely different genetic group (clade 390 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint mloto B) suggest that migration through the Upper Shire is presently not a very common event or is specific to one 391 clade. Also consistent with previous work (Anseeuw et al. 2011), we did not find any evidence for a population 392 bottleneck in the founding of the Lake Malombe population – on the contrary, nucleotide diversity was relatively 393 high and inbreeding coefficient at the low end of all populations. The strongly negative values of Tajima's D in this 394 population could reflect recent selection which would be consistent with the extreme fishing pressure reported for 395 Lake Malombe (Chavula et al. 2023). However, further analysis will be necessary to disentangle evolutionary 396 changes due to (fisheries-induced) selection from those brought about by demographic changes. 397 398 An important open question is whether the genetic differentiation of C. mloto clades is also reflected in 399 morphological differences. Answering this question is complicated by the fact that only a fraction of the specimens 400 used in this study have been preserved, that no 1:1 match of specimens and DNA samples is available for the 2002 401 collection, and that the large variation in adult specimen body size across locations (which might reflect variation in 402 fishing pressure, see below) makes morphological examination tricky for the current sample set. That said, focusing 403 on specimens in male breeding colouration that could be matched to DNA sequences we found some tentative 404 evidence that variation in male breeding dress correlates with adherence to a genetic clade. This is worth further 405 examination given that assortative mating based on male breeding colouration has been shown to be an important 406 factor in the maintenance of species barriers in cichlids (Blais et al. 2009; Madeleine et al. 1998). 407 408 Previous analysis of the population structure of C. mloto based on microsatellites and mtDNA of a partly 409 overlapping sample set found low but significant degrees of population substructuring of C. mloto, but was not able 410 to identify more than a single genetic cluster (Anseeuw et al. 2008). Thus, while overall consistent, our results 411 exemplify the increased power of whole genome data in identifying geographic substructuring among evolutionary 412 recent lineages. While previous studies mainly reported population structure within species restricted to the rocky 413 habitat (e.g., mbuna and Protomelas (Van Oppen at al., 1997; Arnegard et al. 1999; Pereyra et al. 2004), our results 414 emphasise that sufficiently dense data, both in terms of genetic markers as well as geographic coverage, may reveal 415 intricate patterns of population structure and previously unknown taxonomic diversity, also in sand dwe lling and 416 (semi) pelagic species. 417 418 Both decreasing catch rates and an observed decrease in the size of breeding adults suggest that fishing has had a 419 profound effect on C. mloto (Anseeuw et al. 2011), affecting individual survival and potentially leading to fisheries-420 induced evolution (Heino et al. 2015). Here, we found that estimates of genetic parameters do not yet reflect 421 demographic and life history changes brought about by fishing: C. mloto shows relatively higher diversity and 422 generally lower inbreeding coefficients compared to the probably less fished C. virginalis , and relatively little 423 geographic variation in genetic diversity despite strong observed variation in local fishing pressure. On the contrary, 424 the remarkably strong bias towards rare genetic variants in all examined C. mloto and C. virginalis populations (and 425 consequently strongly negative Tajima's D) points to a strong historic population expansion. This signal might 426 reflect a rapid expansion of Utaka into open water habitats at the time such habitats became available after the lake 427 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint recovered from almost complete desiccation ~75 kya (Ivory et al. 2016; Salzburger et al. 2014). This is consistent 428 with studies suggesting that wet-dry transitions in Lake Malawi over the past 1.2 million years and resulting habitat 429 variability have shaped the rapid diversification of Malawi cichlids through changing niche availability and 430 colonisation (Ivory et al. 2016). Combining detailed demographic modelling with paleobiological data (e.g., on lake 431 level fluctuations) will yield a more precise understanding of utaka's evolutionary history and diversification 432 (Danley et al. 2000; Sturmbauer et al. 2001; Nevado et al. 2013; Ivory et al. 2016). 433 434 The lack of population genetic signal for fisheries-induced demographic changes is not unexpected given that the 435 onset of fishing pressure is relatively recent on evolutionary scales and that the mentioned population statistics 436 summarise demographic patterns over long timescales. Indeed, theoretical research suggests that although fishing 437 pressure strongly reduces the current effective population size, the effects of 100 years of intense fishing on genetic 438 diversity are minor (Marty et al. 2015). Earlier studies of fisheries-induced evolution in marine systems have not 439 been able to find genetic evidence for fisheries-induced evolution, attributing observed life history changes to 440 phenotypic plasticity (Jørgensen et al. 2009; Enberg et al. 2012). We suggest that larger sample sizes and bespoke 441 statistics will be necessary to assess the within-population effect of fishing on patterns of genetic diversity. That 442 said, spatially variable fishing pressures are also expected to engender changes in the relative abundance of different 443 species and genetic clades, a signal that might be easier to detect than within-population demographic and 444 evolutionary effects. Specifically, fishing could affect the relative abundance of C. mloto populations pertaining to 445 different genetic clades. Sufficiently fine-scaled geographic and temporal genetic monitoring of C. mloto 446 populations and other commercially important Malawi cichlid species more generally will be crucial for the timely 447 detection of such changes and appropriate management responses. 448 449

450 We investigated population genetic structure in the Malawi cichlid species C. mloto based on whole-genome data of 451 a relatively large and geographically widespread sample set. We found that what was previously thought to be a 452 single species of utaka – in itself a complex of semi-open water, planktivorous cichlid species – shows unexpected 453 patterns of population structure, pointing to the presence of at least three widely distributed, genetically distinct C. 454 mloto clades. Generalising from this, our results indicate that taxonomic diversity in the large and understudied 455 group of benthic and pelagic Lake Malawi cichlid species might be higher and more complex than previously 456 thought, despite generally wide distribution ranges and seemingly continuous habitats. Such diversity might often 457 not be easily discernible by morphological examination and calls for broad genomic investigation and monitoring. 458 This is especially urgent in the context of increasing commercial exploitation of most benthic and pelagic species. 459 460 Funding 461 This study was financially supported by a grant from the Flemish University Research Fund (BOF) awarded to 462 Hannes Svardal (H.S) Department of Biology as approved by the Chair of the Bureau of the Research Board 463 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint (BOZR) on 21/12/2020. The 2017 sample collection was funded by a grant from the Alborada Cambridge-Africa 464 Foundation awarded to H.S. and B.R. 465

466 We would like to thank the Department of Fisheries in Malawi for permission to conduct fieldwork for the 2016 and 467 2017 sample collections and Mexford Mulumpwa, Milan Malinsky, Alix Tyers, Mingliu Du, Gregoire Vernaz, Karl 468 Svardal, and Richard Zatha for their assistance during fieldwork. We sincerely thank Eric Miska and Richard Durbin 469 (University of Cambridge and Wellcome Sanger Institute) for funding sample collection and sequencing for this 470 study. We thank Jos Snoeks (Royal Belgian Museum for Central Africa), Erik Verheyen (Royal Belgian Institute of 471 Natural Sciences) and Dieter Anseeuw for sharing fin clips and extracts of the 2002 C. mloto collection, and 472 Maarten Van Steenbergen for help with locating samples. 473 474 Authors contributions 475 W.S. and H.S. conceived the study. A.H. generated and prepared variant call format (VCF) data. W.S. analysed 476 genomic data with contributions from A.H and H.S. M.G., B.R. and H.S.2 facilitated sample collecti on. S.G. 477 processed sample metadata and performed DNA extractions. G.F.T undertook comparison of male breeding dress 478 against clade membership. W.S. and H.S. wrote the manuscript with input from A.H. and G.F.T. All authors read 479 and approved the manuscript. 480 481 Conflicts of interest 482 The authors declare no conflict of interest. 483 484 Data availability 485 Supporting data is made available in the online supplementary on an open-access basis for research use only. ENA 486 Accession numbers of raw sequencing data will be added before final publication. Data was collected under 487 appropriate ethical and sampling permits and genetic material and sequences are subject to an Access and Benefits 488 Sharing (ABS) agreement with the Government of Malawi. Any person who wishes to use this data for any form of 489 commercial purpose must first enter into a commercial licensing and benefit-sharing arrangement with the 490 Government of Malawi. 491 492

493 494 Alexander, D.H., J. Novembre & K. Lange, 2009. Fast model-based estimation of ancestry in unrelated individuals. 495 Genome Research: 19, 1655–1664. https://doi.org/10.1101/gr.094052.109 496 Anseeuw, D., G.E. Maes, P. Busselen, D. Knapen, J. Snoeks & E. Verheyen, 2008. Subtle population structure and 497 male-biased dispersal in two Copadichromis species (Teleostei, Cichlidae) from Lake Malawi, East Africa. 498 Hydrobiologia: 615, 69–79. https://doi.org/10.1007/s10750-008-9565-z 499 Anseeuw, D., J.A.M. Raeymaekers, P. Busselen, E. Verheyen & J. Snoeks, 2011. Low Genetic and Morphometric 500 Intraspecific Divergence in Peripheral Copadichromis Populations (Perciformes: Cichlidae) in the Lake 501 Malawi Basin. International Journal of Evolution Biology: 2011, 1–11. https://doi.org/10.4061/2011/835946 502 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint Arnegard, M.E., J.A. Markert, P.D. Danley, J.R. Stauffer, A.J. Ambali & T.D. Kocher, 1999. Population structure 503 and colour variation of the cichlid fish Labeotropheus fuelleborni Ahl along a recently formed archipelago of 504 rocky habitat patches in southern Lake Malawi. Proceedings of the Royal Society of Biological Sciences: 266, 505 119–130. https://doi.org/10.1098/rspb.1999.0611 506 Banda, M.C., J. Chisambo, R.D. Sipawe, K.R. Mwakiyongo & O.L.F. Weyl, 2001. Fisheries Research Unit 507 Research Plan. Government of Malawi, Department of Fisheries Fisheries Bulletin. 1–54. 508 Blais, J., M. Plenderleith, C. Rico, M.I. Taylor, O. Seehausen, C. Van Oosterhout & G.F. Turner, 2009. Assortative 509 mating among Lake Malawi cichlid fish populations is not simply predictable from male nuptial colour. BMC 510 Evolutionary Biology: 9, 1–12. https://doi.org/10.1186/1471-2148-9-53 511 Blumer, M. et al. Chromosome-scale inversions contribute to species divergence and sex 512 determination in the Malawi cichlid adaptive radiation. Manuscript in preparation. 513 Brawand, D., C.E. Wagner, Y.I. Li, M. Malinsky, I. Keller, S. Fan, O. Simakov, A.Y. Ng, 514 Z.W. Lim, E. Bezault, J. Turner-Maier, J. Johnson, R. Alcazar, H.J. Noh, P. Russell, 515 B. Aken, J. Alföldi, C. Amemiya, N. Azzouzi, J.F. Baroiller, F. Barloy-Hubler, 516 A. Berlin, R. Bloomquist, K.L. Carleton, M.A. Conte, H. D’Cotta, O. Eshel, L. Gaffney, 517 F. Galibert, H.F. Gante, S. Gnerre, L. Greuter, R. Guyon, N.S. Haddad, W. Haerty, 518 R.M. Harris, H.A. Hofmann, T. Hourlier, G. Hulata, D.B. Jaffe, M. Lara, A.P. Lee, 519 I. MacCallum, S. Mwaiko, M. Nikaido, H. Nishihara, C. Ozouf-Costaz, D.J. Penman, 520 D. Przybylski, M. Rakotomanga, S.C.P. Renn, F.J. Ribeiro, M. Ron, W. Salzburger, 521 L. Sanchez-Pulido, M.E. Santos, S. Searle, T. Sharpe, R. Swofford, F.J. Tan, 522 L. Williams, S. Young, S. Yin, N. Okada, T.D. Kocher, E.A. Miska, E.S. Lander, 523 B. Venkatesh, R.D. Fernald, A. Meyer, C.P. Ponting, J.T. Streelman, K. Lindblad-Toh, 524 O. Seehausen, Di & F. Palma, 2015. The genomic substrate for adaptive radiation in 525 African cichlid fish. Nature: 513, 375–381. https://doi.org/10.1038/nature13726 526 Camacho, J. et al. Pervasive evolutionary adaptation of visual pathways in deep-water dwelling 527 Lake Malawi cichlids. Manuscript in preparation. 528 Chavula, G.M.S., M.S. M'balaka, E. Gondwe, M. Ngochera, J.S. Halafo, A.H. Shechonge, N.P. Ga 529 bagambi, A.S. Mbonde, D. Banda, N. Muyumbana, T.J. Lawrence, M.Z. Ajode, 530 S. Smith, C. Jonasse, G.Z. Kanyerere, & B.P. Ngatunga, 2023. Lake Malawi/Niassa/Nyasa 531 basin: Status, challenges, and research needs. Journal of Great Lakes Research: 49, 102241. 532 https://doi.org/10.1016/j.jglr.2023.102241. 533 Danecek, P., J.K. Bonfield, J. Liddle, J. Marshall, V. Ohan, M.O. Pollard, A. Whitwham, T. Keane, S.A. McCarthy, 534 R.M. Davies, & H. Li, 2021. Twelve years of SAMtools and BCFtools. Gigascience:10. 535 https://doi.org/10.1093/gigascience/giab008 536 Danley, P.D., J.A. Markert, M.E. Arnegard, & T.D. Kocher, 2000. Divergence with Gene Flow in the Rock-537 dwelling cichlids of Lake Malawi. Evolution: 54, 1725. 538 Duftner, N., K.M. Sefc, S. Koblmüller, B. Nevado, E. Verheyen, H. Phiri, & C. Sturmbauer, 2006. Distinct 539 population structure in a phenotypically homogeneous rock-dwelling cichlid fish from Lake Tanganyika. 540 Molecular Ecology: 15, 2381–2395. https://doi.org/10.1111/j.1365-294X.2006.02949.x 541 Eccles, D.H., & E. Trewavas, 1989. Malawian cichlid fishes. The classification of some 542 Haplochromine genera. Lake Fish Movies, Herten, Germany, 335 pp. 543 Enberg, K., C. Jørgensen, E.S. Dunlop, Ø. Varpe, D.S. Boukal, L. Baulier, S. Eliassen, 544 & M. Heino, 2012. Fishing-induced evolution of growth: Concepts, mechanisms and the 545 empirical evidence. Marine Ecology: 33, 1– 25. https://doi.org/10.1111/j.1439-0485.2011.00460.x 546 Food and Agriculture Organisation of the United States (FAO)., 1993. Report of the expert consultation on 547 utilization and conservation of aquatic genetic resources, Grottaferrata, Italy, 9-13 November, 1992. FAO 548 Fish. Report 491. Rome, FAO. 58p. 549 Heino, M., B. Díaz Pauli, & U. Dieckmann, 2015. Fisheries-Induced Evolution. Annual Review of Ecology, 550 Evolution and Systematics: 46, 461– 480. https://doi.org/10.1146/annurev-ecolsys-112414-054339 551 IIles, T.D., 1960. A group of zooplankton feeders of the genus Halpochromis (Cichlidae) in Lake 552 Nyasa. Annals and Magazine of Natural History 2, 1959: 257-280. 553 Ivory, S.J., M.W. Blome, J.W. King, M.M. McGlue, J.E. Cole, & A.S. Cohen, 2016. Environmental change explains 554 cichlid adaptive radiation at Lake Malawi over the past 1.2 million years. Proceedings of the National 555 Academy of Sciences of the United States of America:113, 11895–11900. 556 https://doi.org/10.1073/pnas.1611028113 557 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint Jørgensen, C., B. Ernande, & Ø. Fiksen, 2009. Size-selective fishing gear and life history evolution in the Northeast 558 Arctic cod. Evolutionary Applications: 2, 356–370. https://doi.org/10.1111/j.1752-4571.2009.00075.x 559 Kanyerere, G.Z., D. Kaonga, O. Mponda, B. Nkhoma, & E. Ngulande, 2018. Status of the large-scale commercial 560 trawl fisheries in Southern Lake Malawi. Aquatic Ecosystem Health and Management: 21, 121– 131. 561 https://doi.org/10.1080/14634988.2018.1457388 562 Abundance, Spatial Distribution, Size and Species Composition of Fish Stocks in Lake Malombe. 563 Konings, A., 2019. Copadichromis virginalis (amended version of 2018 assessment). The IUCN Red List of 564 Threatened Species 2019: e.T60879A155042587. https://dx.doi.org/10.2305/IUCN.UK.2019-565 3.RLTS.T60879A155042587.en. Accessed on 16 February 2024. 566 Konings, A., 1999. Description of Three New Copadichromis Species (Labroidei; Cichlidae) from Lake Malawi, 567 Africa. Tropical Fish Hobbyist: 63–84. 568 Maan, E.M., K.M. & Sefc, 2013.Colour variation in cichlid fish: Developmental mechanisms, 569 selective pressures and evolutionary consequences. Seminars in Cell and Developmental 570 Biology:24, 516-528.https://doi.org/10.1016/j.semcdb.2013.05.003. 571 Madeleine J. H, H. Van oppen, G. Turner, L. Rosanna, J. Robinson, C. Deutsch, M. Genner, & G. M. Hewitt, 1998. 572 Assortative mating among rock-dwelling cichlid fishes supports high estimates of species richness from Lake 573 Malawi. Molecular Ecology: 7, 991–1001. 574 Malinsky, M., M. Matschiner, & H. Svardal, 2021. Dsuite - Fast D-statistics and related admixture evidence from 575 VCF files. Molecular Ecology Resources: 21, 584–595. https://doi.org/10.1111/1755-0998.13265 576 Malinsky, M., H. Svardal, A.M. Tyers, E.A. Miska, M.J. Genner, G.F. Turner, & R. Durbin, 2018. Whole-genome 577 sequences of Malawi cichlids reveal multiple radiations interconnected by gene flow. Nature Ecology and 578 Evolution: 2, 1940–1955. https://doi.org/10.1038/s41559-018-0717-x 579 Marty, L., U. Dieckmann, & B. Ernande, 2015. Fisheries-induced neutral and adaptive evolution in exploited fish 580 populations and consequences for their adaptive potential. Evolutionary Applications: 8, 47–63. 581 https://doi.org/10.1111/eva.12220 582 Masonick, P., A. Meyer, & C.D. Hulsey, 2022. Phylogenomic Analyses Show Repeated Evolution of Hypertrophied 583 Lips among Lake Malawi Cichlid Fishes. Genome Biology and Evolution:14, 1–15. 584 https://doi.org/10.1093/gbe/evac051 585 Nevado, B., S. Mautner, C. Sturmbauer, & E. Verheyen, 2013. Water-level fluctuations and metapopulation 586 dynamics as drivers of genetic diversity in populations of three Tanganyikan cichlid fish species. Molecular 587 Ecology: 22, 3933–3948. https://doi.org/10.1111/mec.12374 588 Ngochera, M., S. Donda, M. Hara, & E. Berge, 2018. Defragmenting resource management on the southeast arm of 589 Lake Malawi: Case of fisheries. Aquatic Ecosystem Health and Management: 21, 139–151. 590 https://doi.org/10.1080/14634988.2018.1475182 591 Pereyra, R., M.I. Taylor, G.F. Turner, & C. Rico, 2004. Variation in habitat preference and population structure 592 among three species of the Lake Malawi cichlid genus Protomelas. Molecular Ecology: 13, 2691–2697. 593 https://doi.org/10.1111/j.1365-294X.2004.02224.x 594 Purcell, S., B. Neale, K. Todd-Brown, L. Thomas, M. Ferreira, D. Bender, J. Maller, P. Sklar, P. De Bakker, M. 595 Daly, & P. Sham, 2007. PLINK: a toolset for whole-genome association and population-based linkage 596 analysis. American Journal of Human Genetics: 81. 597 R Core Team, 2021. R: A language and environment for statistical computing. R Foundation for 598 Statistical Computing, Vienna, Austria. URL https://www.R-project.org/ 599 Rick, J.A., J. Junker, I.A. Kimirei, E.A. Sweke, J.B. Mosille, C. Dinkel, S. Mwaiko, O. Seehausen, & C.E. Wagner, 600 2022. The Genetic Population Structure of Lake Tanganyika’s Lates Species Flock, an Endemic Radiation of 601 Pelagic Top Predators. Journal of Heredity:113, 145–159. https://doi.org/10.1093/jhered/esab072 602 Salzburger, W., B.V. Bocxlaer, & A.S. Cohen, 2014. Ecology and evolution of the African Great Lakes and their 603 faunas. Annual Review of Ecology, Evolution and Systematics: 45, 519–545. https://doi.org/10.1146/annurev-604 ecolsys-120213-091804 605 Scherz, M.D., P. Masonick, A. Meyer, & C.D. Hulsey, 2022. Between a Rock and a Hard Polytomy: Phylogenomics 606 of the Rock-Dwelling Mbuna Cichlids of Lake Malawi. Systematic Biology: 71, 741–757. 607 https://doi.org/10.1093/sysbio/syac006 608 Shaw, P.W., G.F. Turner, M.R. Idid, R.L. Robinson, & G.R. Carvalho, 2000. Genetic population structure indicates 609 sympatric speciation of Lake Malawi pelagic cichlids. Proceedings of the Royal Society B: Biological 610 Sciences:267, 2273–2280. https://doi.org/10.1098/rspb.2000.1279 611 Sturmbauer, C., S. Baric, W. Salzburger, L. Rüber, E. Verheyen, 2001. Lake Level Fluctuations Synchronize 612 Genetic Divergences of Cichlid Fishes in African Lakes. Molecular Biology and Evolution: 18, 144–154. 613 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint Svardal, H., F.X. Quah, M. Malinsky, B.P. Ngatunga, E.A. Miska, W. Salzburger, M.J. Genner, G.F. Turner, & R. 614 Durbin, 2020. Ancestral hybridization facilitated species diversification in the Lake Malawi cichlid fish 615 adaptive radiation. Molecular Biology and Evolution: 37, 1100–1113. https://doi.org/10.1093/molbev/msz294 616 Taylor, M.I., & E. Verheyen, 2001. Microsatellite data reveals weak population substructuring in Copadichromis sp. 617 `virginalis kajose’, a demersal cichlid from Lake Malawi, Africa. Journal of Fish Biology: 59, 593–604. 618 https://doi.org/10.1006/jfbi.2001.1670 619 Turner, G.F., D.A. Crampton, A. Van Huysduynen, & H. Svardal, 2022. Taxonomic investigation 620 of the zooplanktivorous Lake Malawi cichlids Copadichromis mloto (Iles) and C. virginalis 621 (Iles). bioRxiv pre-print. https://doi.org/10.1101/2022.03.10.483679 622 Turner, G., 1996. Offshore cichlids of Lake Malawi. Cichlid Press, Lauenau, Germany; 240 pp. ISBN 3-928457-623 33-0. 624 Tweddle, D., G.F. Turner, & M. Seisay, 1995. Changes in species composition and abundance as a consequence of 625 fishing pressure in Lake Malombe, Malawi. In: Pitcher, T.J. & Hart, P.J.B. (eds) The Impact of Species 626 Changes in African Lakes. London: Chapman and Hall, pp. 413–424. 627 Van Oppen, M.J.H., G.F. Turner, C. Rico, J.C. Deutsch, K.M. Ibrahim, R.L. Robinson, 628 & G.M. Hewitt, 1997.Unusually fine-scale genetic structuring found in rapidly speciating 629 Malawi cichlid fishes. Proceedings of the Royal Society and London B: 264, 1803-1812. 630 Van Rossum, & G., F.L. Drake, 2009. Python 3 Reference Manual, Scotts Valley, CA: CreateSpace 631 Van Zwieten, P.A.M., C. Béné, & J. Kolding, 2011. Review of tropical reservoirs and their fisheries-The cases of 632 Lake Nasser, Lake Volta and Indo-Gangetic Basin reservoirs. FAO Fisheries and Aquaculture Technical Paper 633 1–148. 634 Weyl, O.L.F., K.R. Mwakiyongo, & D.S. Mandere, 2004. An assessment of the Nkacha net fishery of Lake 635 Malombe, Malawi. African Journal of Aquatic Science: 29, 47–55. 636 https://doi.org/10.2989/16085910409503791 637 638 .CC-BY-NC-ND 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted April 28, 2024. ; https://doi.org/10.1101/2024.04.26.591058doi: bioRxiv preprint