Sampling was conducted at Lake Rutamba (10°01'52"S, 39°27'44"E; Fig. 1) near Lindi in Tanzania during two field sampling events (22–24 October 2016; 1–2 November 2019). Lake Rutamba is a turbid vegetation-rich shallow lake, measuring approximately 2 km × 1 km, with an approximate maximum depth of 2.5 m. Reed beds surround the lake but during the summer months the water level drops and recedes away from the reeds reducing the potential for fish to shelter in vegetation. Predators of fishes observed to be associated with the lake include sharptooth catfish Clarias gariepinus, Nile crocodile Crocodylus niloticus and birds (Mycteria ibis, Microcarbo africanus, Ardea alba). The lake is exploited as a capture fishery, with approximately 25 active gill net fishers being recorded daily during the 2019 sampling season.

O. niloticus, O. korogwe and their hybrids were purchased from local fishers using gill nets (in 2016 and 2019) or collected using a survey seine net (used three times in 2019; dimensions 30 m × 1.5 m, 25.4 mm mesh, fine mesh cod-end). The individuals retained from surveying were selected based on phenotypic characteristics and all individuals retained were greater than 35 mm (Table 1). We aimed to collect a sample enabling comparisons O. niloticus and O. korogwe specimens, rather than a sample that reflected the relative population size of each species within Lake Rutamba. Samples collected using the survey seine net were euthanised using an overdose of anaesthetic (clove oil). Specimens were pinned to a polystyrene board, photographed, labelled, and stored individually in 100% ethanol before transport. Long-term storage was in 70% ethanol.

We determined the genetic composition of sampled individuals using microsatellite genotypes. DNA was extracted from fin clips following the Wizard Genomic DNA Purification Kit protocol (Promega, Madison, WI). DNA concentrations were measured using an N60 Touch NanoPhotometer (Implen, München, Germany), and diluted to 50 ng/µl. For the assay we selected six microsatellite loci (OMO219, OMO229, OMO391, OMO337, OMO129, OMO043) from Saju et al. (2010) and Liu et al. (2013), previously used to classify individuals as O. korogwe, O. niloticus or their hybrids (Blackwell et al. 2021). Polymerase chain reactions were performed using 10 µl solutions comprised of: 1 µl DNA, 0.2 µl of the six forward primers (each 10 µM), 0.2 µl of the six reverse primers (each 10 µM), 5 µl of Multiplex PCR Master Mix (Qiagen, Hilden, Germany) and 1.6 µl of distilled water. PCR conditions were an initial denaturation at 95 °C for 60s, followed by 35 cycles of 94 °C for 30s, 57 °C for 90s and 72 °C for 60s, before a final extension stage at 60 °C for 30 minutes. PCRs were conducted on a 3PRIMEX/02 thermal cycler (Techne, Staffordshire, UK). Amplicons were genotyped using a 3500 Genetic Analyser (Applied Biosystems, Waltham, MA) with a LIZ500 size standard, and scoring was conducted using Genemapper 4.1 (Applied Biosystems, MA). Amplification of one locus (OMO043) was unsuccessful, so analyses were conducted on the five remaining loci. Genetic composition of individuals were estimated using the admixture model in Structure v.2.3.4 (Pritchard et al. 2000), assuming two populations (K = 2; O. korogwe or O. niloticus), in 10 separate runs of 100,000 steps following a 100,000 burn-in. This resulted in an assignment probability of between 0 and 1 for each specimen. Following Blackwell et al. (2019) Individuals with < 0.1 were deemed to be O. korogwe, individuals > 0.9 O. niloticus, and individuals > 0.1 and < 0.9 to be hybrids.

To assess the recent growth rate of the Oreochromis specimens we followed an approach developed by Doyle et al. (1986). This method provides an instantaneous measurement of individual growth rate using scale circuli measurements, circumventing the need for the release and recapture of the same individual over a known length of time. Specifically, the authors demonstrated that the distance between marginal scale circuli in Oreochromis mossambicus × O. urolepis hybrids was significantly correlated with growth rate (Doyle et al. 1986). Experiments further demonstrated that the ratio between body length and the distance between scale circuli can be used to reliably compare the growth rate of different tilapia species and aquacultural strains (Matricia et al. 1989; Talbot and Doyle 1992; Martin 2012). This method has previously been used to compare relative growth of invasive O. niloticus and indigenous Tanzanian tilapia species (Bradbeer et al. 2020).

Three scales were collected from the right side of each specimen, from the first scale row dorsal to the lateral line and posterior to the pelvic girdle. To ensure consistency in scale type, scales were removed sequentially until three fully formed scales with tight foci were obtained (Fig. 2a). Scales were submerged in water and excess skin and debris were removed using forceps to ensure individual circuli were visible. Scales were then dried, coated with glycerol on a microscope slide and covered with a glass coverslip. Images of individual scales were then taken using a M205c stereo microscope (Leica Microsystems, Wetzlar, Germany) with a GXCAM HICHROME MET-M camera attachment (GT Vision, Newmarket, UK). Five measurements were recorded from each scale in micrometres (μm). First, the total width of the scale at its widest point (0.78× magnification), followed by four measurements of the distance between the five outermost circuli on four primary radii (5× magnification; Fig. 2b). Measurements, calibrated using an image of a graticule taken at the same magnification, were made using Image-J v.1.3.3 (Schneider et al. 2012).

Figure 2. 

Stereomicroscope images of a) whole scale and b) outer scale circuli on primary radii. Red arrows represent the measurements made of scales for this study.

All analyses were performed in R v.4.1.3 (R Core Team 2022). To investigate variation in growth rate between O. niloticus, O. korogwe and their hybrids a linear model was constructed with growth increment as the dependent variable, defined as the average distance in μm between the five outer circuli from four primary radii on three separate scales. We included species (O. korogwe, O. niloticus and O. korogwe × O. niloticus hybrids) as a fixed factor, alongside mean scale diameter (μm) as a covariable indicator of body size. We also included the interaction term of species × mean scale diameter, to investigate whether the association between growth increment and mean scale diameter varied among the groups of individuals. In a second linear model, also including growth increment as the dependent variable, we included species, standard length (SL, another indicator of body size) and an interaction between these variables. The growth rate, mean scale diameter and SL were log₁₀ transformed for analyses. The simulate.Residuals function in the package DHARMa v.0.4.6 (Hartig 2020) was used to generate a Q-Q plot of observed vs expected residuals, and a plot of residuals vs fitted values, to ensure normality of residuals and homogeneity of variances. The function Anova in the package car v.3.1-2 (Fox et al. 2013) was used to test for the significance of the fixed effects, employing a type II model due to unequal sample sizes across the three groups of individuals. Figures were produced using ggplot2 v.3.4.4 (Wickham 2016).

Nile tilapia has been widely introduced to natural water bodies across Tanzania since the 1950s, but the precise timing of the introduction into Lake Rutamba is unclear. We know that O. niloticus was fully established in the lake in 2013 (Blackwell et al. 2021). We also know that samples of O. korogwe from Lake Rutamba were accessioned to the Natural History Museum in London in 1982 (as Sarotherodon ruvumae), alongside Coptodon rendalli (as Tilapia rendalli), but no O. niloticus individuals were accessioned alongside them, suggesting Nile tilapia were absent at the time of collection (presumed to be early 1980s). We therefore estimate that O. niloticus was introduced to the lake between the early 1980s and the early 2010s. Evidence of the negative effects of O. niloticus across its native range (Canonico et al. 2005), including the extinction of a native tilapia species from the Hombolo reservoir in Tanzania (Turner et al. 2019), implies that O. niloticus may pose a major threat to native tilapia in East African freshwaters. Thus, the ability of O. korogwe to persist in the face of O. niloticus introduction for numerous generations is contrary to expectation and suggests that this species may possess traits which predispose them for resilience to O. niloticus invasion.

Higher growth rates leading to increased body size are linked to several competitive advantages. These include performance during interference competition for shelter, greater efficiency during exploitative competition for food resources, increased reproductive output in mature females, and an enhanced probability of success during interference competition for lekking spaces in males (Chifamba and Videler 2014; Barneche et al. 2018; Bradbeer et al. 2020). Previous studies have shown that O. niloticus is an aggressive competitor that can dominate competitive interactions and prevent access to shelter in subordinate species (Martin et al. 2010; Champneys et al. 2021). Typically, fish are most vulnerable to predation during juvenile stages (Sogard 1997), and in Lake Rutamba, survival is likely to depend on access to shelter resources such as the large reed beds located in the littoral regions. Elevated growth rates leading to increased body size could enable O. korogwe to avoid competitive dominance by O. niloticus and access shelter resources and preferred habitats despite competition. During sampling in 2019, O. niloticus and O. korogwe were both found in all three seine locations in the littoral areas of the lake (Fig. 1), suggesting strong habitat overlap and likely competition over preferred habitats. However, more accurate information about the habitat use and niche overlap of the two species within the lake would lead to more precise predictions about the likely prevalence and outcomes of competition between the species over shared resources.

Since introduced O. niloticus are typically descendants of fish selected for high production yields, including fast growth rates, the increased growth rate observed in the native O. korogwe contrasts with our expectations. One explanation is that introduced O. niloticus are relatively poorly adapted to the local environment in comparison to the native O. korogwe. Unlike native species, which have had a long evolutionary timeframe in which to adapt to environmental conditions, introduced species are faced with novel conditions to which they must rapidly adapt to become established (Flores-Moreno et al. 2015). Studies have shown that introduced species can be successful through rapid adaptation to novel environments via behavioural, phenological and morphological changes (Thompson 1998; Lambrinos 2004). However, such rapid adaptation is not ubiquitous, and slow rates of phenotypic evolution following introduction are also reported (Mooney and Cleland 2001; Sakai et al. 2008). Growth rate is crucially influenced by access to food resources and a relatively poor ability to locate and consume food items could result in the relatively higher growth rates in O. korogwe, which may be better adapted to exploiting these resources.

Given that the relatively high growth rate of O. korogwe was most exaggerated in larger individuals, it is possible that O. korogwe and O. niloticus diverge in their feeding strategy as they grow. Fish commonly shift their diet upon reaching larger sizes (Juncos et al. 2015), and while both species are microphagous generalists (M. Genner pers obs.), resources in offshore habitats may enable the species to diverge into planktivorous and detritivorous feeding strategy, with one providing nutrition which facilitates higher growth rates. Such a separation of dietary niches may not be feasible in inshore nursery habitats, and this could explain the size-dependent differences in growth rate between the two species. Further investigation into ontogenetic dietary overlap of the two species using gut content or stable isotope analysis may help to clarify trophic niches of these species, while surveys of distributions of the species in the lake will provide corresponding insight into any differences in habitat use.

Fishing pressure is high within Lake Rutamba, with at least 25 fishers active during the 2019 sampling period. Fishers were using gill nets which target larger individuals in the central areas of the lake. During both sampling periods, catches were observed to be dominated by O. niloticus, which suggests that fishing pressure may be reducing the population of O. niloticus disproportionately relative O. korogwe, potentially removing breeding adults from the population. This pattern was evident within our sampling, where the largest O. niloticus was only 23 cm total length — roughly half the maximum size of O. niloticus observed in Tanzania (~45 cm total length; Genner et al. 2018). By reducing population sizes of O. niloticus, fishing likely reduces competition for resources, and in turn potentially prevents the formation of interspecific dominance hierarchies over shared resources.

Predation is likely to have a strong effect on the population sizes of both O. niloticus and O. korogwe within Lake Rutamba, especially during summer months when the water level recedes from littoral vegetation, reducing the availability of shelter. Predation by sharptooth catfish Clarias gariepinus, Nile crocodile Crocodylus niloticus and piscivorous birds (Mycteria ibis, Microcarbo africanus, Ardea alba) could disproportionately affect introduced O. niloticus due to poorer anti-predatory adaptations within the introduced species resulting from generations of selection in aquaculture, or a poorer adaptation to the local environment. Evidence of the dietary composition of predator populations within Lake Rutamba would be necessary to expand on this hypothesis.

When ecosystems are affected by multiple stressors there can be antagonistic interactions between them, with one stressor offsetting the other (Berlarde et al. 2016). Few studies have considered how other stressors, such as fishing pressure, can interact with the impacts of invasive species to exacerbate or reduce their negative effects. Further research may be able to able to confirm if there is an antagonistic interaction between fishing pressure and the invasive O. niloticus that may contribute to the persistence of the relative fast-growing O. korogwe in a heavily modified environment.

We found clear evidence of hybridization between O. korogwe and O. niloticus in our samples, in accordance with the findings of Blackwell et al. (2021). The hybrid individuals within our sample were of a range of body sizes and showed a growth rate closer to the faster growing O. korogwe. This result suggests that there is no strong ecological selection against hybrid genotypes, and hybrids may persist within the population. Considering our findings, it is possible that hybrids possessing a faster growing phenotype may have competitive advantages over the parental species, if the elevated growth rates of O. korogwe are combined with traits which may benefit O. niloticus such as a maximum large body size. Parental species are thought to be most threatened when hybrids possess fitness advantages. Where such heterosis is present, genetic swamping can lead to the formation of hybrid swarms (Hwang et al. 2012). Further work is needed to establish the reproductive preferences of parental species and their hybrids within Lake Rutamba, which is important given genomic evidence of fertile hybrids and introgression between the species (Blackwell et al. 2021). Moreover, information on the abundance and habitat use of purebreds and hybrids within this population would be useful for accurate prediction of the consequences of hybridization for these fish populations.