Aquaculture 530 (2021) 735938 Contents lists available at ScienceDirect Aquaculture journal homepage: www.elsevier.com/locate/aquaculture Performance evaluation of Nile tilapia (Oreochromis niloticus) improved T strains in Ghana Trong Quoc Trinha,⁎, Seth Koranteng Agyakwahb, Hooi Ling Khawa, John A.H. Benziea,c, Felix K.Y. Attipoeb a WorldFish, Jalan Batu Maung, 11960 Batu Maung, Penang, Malaysia b CSIR-Water Research Institute, Aquaculture Research & Development Center (ARDEC), P.O. Box 139, Akosombo, Ghana c School of Biological Earth and Environmental Sciences, University College Cork, Cork, Ireland A R T I C L E I N F O A B S T R A C T Keywords: Growth performance and survival to harvest of the Akosombo strain generation 10 and the GIFT strain gen- Nile tilapia eration two (derived from GIFT generation 11 in Malaysia) were evaluated in Ghana. The fish were from 96 Akosombo strain families of the Akosombo selected line (AKOS), 30 of the Akosombo control line (AKOC), and 22 of the GIFT GIFT strain (GIFT). In total, 11,812 tagged fingerlings were stocked at a density of 3 fish.m−2 into two ponds, each Strain comparison 2000 m2. Fish were fed 38% crude protein pelleted feed, twice daily at a rate of 5% body weight, for 120 days. Harvest weight Survival Harvest weight of GIFT (150.1 ± 58.5 g) was 2.2 times larger than AKOC (67.6 ± 28.4 g) and 1.8 times larger than the AKOS (85.2 ± 32.0 g) strains. The least-squares mean weight of male GIFT were 98.6 g more than AKOC and 82.7 g more than AKOS. The corresponding ratios for females were even greater, with female GIFT weighing 106.5 g more than AKOC, and 89.9 g more than AKOS. The AKOS grew larger than the AKOC, 15.9 g larger for males and 16.6 g for females. All differences were highly significant (P < 0.001). Male fish were significantly larger than females for all three strains (P < 0.001). Overall, individual survival rate at harvest was 53.6% and not significantly different between strains, with GIFT survival being 55.2%, AKOS 54.5% and AKOC 50.0%. The results of the present study demonstrated: 1) marked differences in the growth performance of GIFT and Akosombo strains, and 2) the strong growth performance of GIFT in an equatorial West African en- vironment, with similar survival to that of the local strain. 1. Introduction in growth rate and has out-performed other strains in a variety of farming systems in Asia (Bentsen et al., 2017). Several studies have Tilapias comprise one of the most important groups of aquaculture identified the socioeconomic benefits arising from farming of GIFT, species. In 2018, of the 82.1 million tonnes (MT) of aquaculture food-fish including improved rural income and employment (Dey et al., 2000; production, 5.5 MT was accounted for tilapia production (FAO, 2020), Gupta and Acosta, 2004; Yuan et al., 2020). According to the Asian 81% of which was Nile tilapia (Oreochromis niloticus) (FAO, 2020). Sev- Development Bank (2005) “the economic internal rate of return on eral genetic improvement programs have been developed for Nile tilapia investments in GIFT development and dissemination was more than (Komen and Trong, 2014). These include the Genetically Improved 70% over a period from 1988 to 2010, with an estimated net present Farmed Tilapia (GIFT) developed in the 1980s by ICLARM (now known value of US$ 368 million in constant 2001 prices”. In 2001, the base as WorldFish) with other institutional partners and the Akosombo strain population of the Akosombo strain was composed of fish from three developed at the Aquaculture Research and Development Centre agro-ecological zones (Nawuni in the Northern Region, Yeji in the (ARDEC) of the Water Research Institute (WRI) of the Council for Sci- Brong Ahafo Region and Kpando in the Volta Region) and one farmer's entific and Industrial Research (CSIR), in Ghana, since 2000. stock (a fish farm at Nsawam in the Eastern Region) in Ghana (Attipoe The base population for selection of GIFT was composed from four et al., 2013; CSIR - Water Research Institute, 2013). The collected wild African strains (including a Ghanaian strain) and four farmed specimens were held in a quarantine facility at ARDEC and used to form Asian strains (of which three originated from Ghana) (Eknath et al., the base population for a full-pedigreed breeding program. After the 1993; Bentsen et al., 2017). GIFT has shown a remarkable genetic gain first two generations of selection (2002 to 2004), heritability for harvest ⁎ Corresponding author. E-mail address: T.QuocTrinh@cgiar.org (T.Q. Trinh). https://doi.org/10.1016/j.aquaculture.2020.735938 Received 25 June 2020; Received in revised form 14 September 2020; Accepted 14 September 2020 Available online 19 September 2020 0044-8486/ © 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). T.Q. Trinh, et al. Aquaculture 530 (2021) 735938 weight was found to be 0.039 ± 0.014, and estimated cumulative and control AKOC line) families were produced in immediately ad- response to selection ranged from 4.36 to 5.34% (Attipoe et al., 2013). jacent facilities outside the quarantine area. These facilities had larger After a decade of selection, the Aksombo strain was reported to have hatchery capability, resulting in larger numbers of families produced as 30% faster growth than other farmed tilapia strains in the West African part of the routine selection program. sub-region, resulting in improved tilapia production in Ghana (CSIR - The two replicate grow-out ponds available for the experiment were Water Research Institute, 2013). The Akosombo strain enabled farmers each 2000 m2. For testing growth performance in ponds under normal to shorten harvest time from eight to six months and was also reported farming conditions with standard stocking density of 3 fish.m−2, a total to have higher survival rate than other unimproved strains (CSIR - of 12,000 fish (6000 per pond) was required. Given the limited number Water Research Institute, 2013). of fry available with 22 GIFT and 30 AKOC families, 96 AKOS families In 2004, the Nairobi Declaration institutionalized the WorldFish were needed to provide the overall stocking density of 3 fish.m−2. (then WorldFish Center) policy of non-introduction of GIFT to African countries where Nile tilapia is indigenous (Gupta, 2002; Gupta and Acosta, 2004). Instead, WorldFish helped these countries to develop 2.2. Strains used their own selective breeding programs for indigenous Nile tilapia using the GIFT methodology. However, in 2007, WorldFish Center (2007) The GIFT strain (1852 fingerlings from 60 families of generation 11) approved the ‘Policy on the Transfer of GIFT from Asia to Africa’, re- was imported from WorldFish in Malaysia to ARDEC in Ghana during cognizing that the genetic risk of introducing GIFT is comparable to May 2012. Broodstock from both the AKOC and AKOS lines from those associated with the genetic improvement of indigenous Nile ti- generation nine of the Akosombo strain, and of the first GIFT generation lapia strains. in Ghana were used to produce experimental fish for this study. Critical data required for assessing risks and benefits of introduction Therefore, the performance of generation 10 of the Akosombo strain include that on performance of the strain to be introduced with that of (both AKOS and AKOC lines, of which the control line was made up of the local strains. As part of a project on aquaculture in the Volta basin male and female breeders with average estimated breeding values) and (i.e. the ‘TIVO project’), the performance of Nile tilapia genetic re- generation two of GIFT in Ghana were compared. sources was assessed in Benin, Burkina Faso, Côte d'Ivoire, Mali, and Togo (Lind et al., 2019). The GIFT strain (generation 11 in Malaysia) 2.3. Production of fish was introduced to Ghana (CSIR - Water Research Institute, 2013) on an experimental basis, with the purpose of comparing its growth perfor- Eggs or fry from both strains were collected in 11 batches from 26th mance with the locally improved Akosombo strain. December 2013 until 8th April 2014 (Table 1) and were incubated in The present paper reports the results of an experiment comparing down-welling jars until hatching. Fry were initially stocked at a density pond growth performance and survival to harvest of the Akosombo of 200 fry.m−2 in 1-m2 (1 × 1 m) nursing hapas (two per family) and strain generation 10 and the second generation of the GIFT strain in fed a fine powdered feed (38% crude protein) for six weeks. Thereafter, Ghana. This work on the two Nile tilapia strains, which have been ge- they were transferred into 3-m2 (3 × 1 m) hapas until the fingerlings netically improved under equatorial conditions, provides rare bench- reached approximately 5 g. Fish were from 96 families of generation 10 marking information on performance of these strains in West Africa. of AKOS, 30 of AKOC, and 22 of GIFT. All the fish were tagged with Passive Integrated Transponder (PIT) tags in May 2014. Prior to tag- 2. Materials and methods ging, the fish were anesthetized with Tricaine mesylate (Tricaine me- thanesulfonate, MS-222) at a concentration of 0.33 g.l−1 and placed in 2.1. Experimental facilities and experimental design freshwater to recover after tagging. Tagged fish were conditioned in 1- m2 (1 × 1 m) hapas for 10 days. Dead fish were replaced with newly The animals used in the experimental set up from breeding popula- tagged fish from the same family. tions at ARDEC, Ghana, were humanely handled and managed in con- The two 2000-m2 ponds (pond 1 and pond 2) located in the bio- formity with national guidelines on ethics and animal welfare (https:// secure facility at ARDEC were prepared for stocking of the tagged fish. csir.org.gh/index.php/r-d/research-ethics-intellectual-property-rights/ For both strains (including the AKOC), 80 tagged fingerlings per family irb-iacuc/item/432-iacuc-standard-operating-procedure-sop). were stocked equally into two ponds (40 each) on 17th July 2014. A The experiment was conducted at quarantine facilities that had been total of 12,043 tagged fingerlings were stocked at a density of ap- developed to maintain the introduced GIFT strain at ARDEC, proximately 3 fish.m−2. Fish were fed with 38% crude protein pelleted Akosombo, Ghana from December 2013 to December 2014. Facilities feed twice daily and at a rate of 5% body weight for 120 days. A sample for fry production were limited, as was space within the quarantine of fish was weighed every two weeks, and the amount of feed given was facility for fish rearing, and only two ponds were available for the ex- adjusted accordingly. periment. This meant that family production had to be carried out over Fish from pond 1 were harvested from 27th November to 2nd an extended period (Table 1). The number of available GIFT broodstock December 2014, while those from pond 2 were from 9th to 16th (58 males and 116 females, allowing 58 matings at a ratio of 1 male to 2 December 2014. They were transferred to hapas (5 × 2 m) and females) limited the number of families produced and fry subsequently maintained for seven days. Thereafter individual PIT tag code, sex, and tested for growth performance. The Akosombo strain (selected AKOS harvest weight were recorded for each fish. Table 1 Numbers of families collected from the three strains (11 batches). Strain1 Year Total 2013 2014 26/12 14/1 17/1 26/1 28/1 11/2 12/2 25/2 11/3 12/3 8/4 AKOC 14 7 1 5 1 1 1 30 AKOS 62 20 5 5 2 1 1 96 GIFT 2 5 11 3 1 22 Total 78 27 5 1 10 6 11 3 2 3 2 148 1 AKOC = Akosombo control line, AKOS = Akosombo selected line, GIFT = GIFT strain. 2 T.Q. Trinh, et al. Aquaculture 530 (2021) 735938 Table 2 Number of fish, mean stocking and harvest weights ± standard deviations, and coefficient of variation (CV) by pond, strain, and sex. Pond Strain1 Sex # of fish Stocking weight (g) Harvest weight (g) Stocking⁎ Harvest Mean ± SD CV Mean ± SD CV Pond 1 AKOC Female 304 8.3 ± 4.4 53.1 52.6 ± 14.5 27.6 Male 403 8.7 ± 4.2 48.8 92.6 ± 22.8 24.6 Total 1180 707 8.5 ± 4.3 50.6 75.4 ± 28.0 37.1 AKOS Female 1203 8.3 ± 3.5 42.1 75.1 ± 21.1 28.1 Male 1310 9.8 ± 4.0 40.9 112.4 ± 27.4 24.4 Total 3872 2513 9.1 ± 3.8 42.2 94.5 ± 30.8 32.6 GIFT Female 245 6.0 ± 3.6 60.5 165.1 ± 43.6 26.4 Male 179 6.4 ± 3.5 54.5 191.1 ± 50.3 26.3 Total 851 424 6.2 ± 3.6 57.9 176.1 ± 48.3 27.4 Total pond 1 5904 3644 8.6 ± 4.0 46.4 100.3 ± 43.5 43.3 Pond 2 AKOC Female 161 8.7 ± 4.5 51.9 36.7 ± 22.0 59.9 Male 313 9.4 ± 4.8 51.4 65.9 ± 20.1 30.5 Total 1184 474 9.2 ± 4.7 51.6 56.0 ± 25.0 44.6 AKOS Female 630 8.2 ± 3.5 43.0 50.4 ± 16.9 33.6 Male 1064 10.5 ± 4.6 43.7 83.9 ± 26.8 32.0 Total 3865 1694 9.6 ± 4.4 45.3 71.4 ± 28.6 40.1 GIFT Female 260 6.0 ± 3.8 63.1 111.1 ± 48.5 43.6 Male 260 7.3 ± 3.8 51.5 146.6 ± 60.5 41.3 Total 860 520 6.7 ± 3.8 57.6 128.8 ± 57.6 44.7 Total pond 2 5909 2688 9.0 + 4.5 50.0 79.8 ± 43.3 54.3 2 ponds AKOC 2364 1181 8.8 ± 4.5 51.2 67.6 ± 28.4 42.0 AKOS 7737 4207 9.3 ± 4.1 43.7 85.2 ± 32.0 37.6 GIFT 1711 944 6.4 ± 3.7 57.8 150.1 ± 58.5 39.0 Total 11,812 6332 8.8 ± 4.2 48.1 91.6 ± 44.6 48.7 1 AKOC = Akosombo control line, AKOS = Akosombo selected line, GIFT = GIFT strain. ⁎ Sexes cannot be identified at stocking. 2.4. Data analysis and 148 dams) harvested from the two ponds, of which pond 1 had more fish (3644, 58% of the total number of fish in the two ponds) than Data were analyzed using R software version 3.6.3 (R Core Team, pond 2 (2688, 42%) (Table 2). 2020). The different weights of fish at the start of the experiment The mean harvest weight of the GIFT strain over both ponds (Table 2) were accounted for by including the total age at harvest (total (150.1 ± 58.5 g) was 2.2 times larger than that for AKOC (67.6 ± age from spawning until harvesting) as a covariate. Harvest weight was 28.4 g) and 1.8 times larger than that for the AKOS strain analyzed using the following mixed-effects model (Model 1) using (85.2 ± 32.0 g), and this trend was consistent for both ponds package ‘nlme’ (Pinheiro et al., 2020) (Table 2). The mean harvest weight of GIFT in pond 1 (176.1 ± harwijklm = μ + β1 × harageijklm + (strain × sex)i + pondj + 48.3 g) was 37% larger than that in pond 2 (128.8 ± 57.6 g). For the sirek + daml + eijklm (Model 1) where, harwijklm is harvest weight of Akosombo lines, it was similar: AKOC fish in pond 1 (75.4 ± 28.0 g) the mth fish, μ is the population mean, β1 × harageijklm is the fixed were larger than in pond 2 (56.0 ± 25.0 g), and for AKOS it was regression on total age at harvest, harage, (strain × sex)i is the fixed 94.5 ± 30.8 g in pond 1 and 71.4 ± 28.6 g in pond 2. For all the three effect of the i combination of strain (the two Akosombo lines and the groups, males were significantly heavier than females, most noticeably GIFT) and sex (male and female), pondj is random effect of pond j, sirek with AKOC (males 72.3% larger) and AKOS (males 49.5% larger), but is random effect of sire k, daml is random effect of dam l, and eijklm is less so in GIFT (20.0%). the random residual term. Overall survival rate of the three groups was 54%, with 55% for GIFT, Individual survival at harvest (alive = 1, dead = 0) was analyzed 54% for AKOS, and 50% AKOC, and the differences were not statistically using the following generalized (logistic regression) mixed-effects significant. The number of surviving fish in pond 1 was higher than in model (Model 2) using package ‘lme4’ (Bates et al., 2015) pond 2, but relative survival among strains differed between ponds. yijkl = μ + straini + pondj + sirek (Model 2) where, yijkl is the However, the overall percentages of the numbers of fish in each strain at logit link function of survival at harvest of individual l (alive = 1, harvest (65% AKOS, 20% AKOC, and 15% GIFT) were similar to those at dead = 0), μ is the population mean, straini is fixed effect of the i strain stocking (66% AKOS, 20% AKOC, and 14% GIFT) (Table 3). (the two Akosombo lines and the GIFT), pondj is random effect of the j pond (2 ponds), and sirej is random effect of sire j. 3.2. Fixed effects and covariate For harvest weight, p-values in type III Sum of Square ANOVA were calculated using package ‘lmerTest’ (Kuznetsova et al., 2017). For both For harvest weight, two fixed effects (strain and sex), their inter- harvest weight and individual survival at harvest, least squares means actions, and one covariate (total age at harvest) were all significant (LSM) for strains and sexes, and their pairwise comparisons of means (Table 4). For individual survival at harvest, strain as the only fixed were calculated using package ‘emmeans’ (Lenth et al., 2020). For in- effect was not significant (P = 0.08). dividual survival at harvest, all values were transformed from the logit scale to the response scale. 3.3. Strain comparisons 3. Results GIFT outperformed AKOC and AKOS by large margins for growth. The analysis took account of the differences in stocking weight by the 3.1. Descriptive statistics inclusion of age at harvest as a covariate (Table 2). The LSM of male GIFT individually was 98.6 g more than AKOC and 82.7 g more than In total, there were 6332 fish from 148 full-sib families (148 sires AKOS. The corresponding differences for females were greater; where 3 T.Q. Trinh, et al. Aquaculture 530 (2021) 735938 Table 3 Table 6 Numbers (and percentages) of surviving fish at harvest by strain. Least-square means of survival to harvest for each strain (on the diagonal, 1 value = logit scale = estimate ± standard error) and pairwise comparison (off Strain Pond 1 Pond 2 Total diagonal, value = response scale). The pairwise comparisons indicate the dif- AKOC 707 (19%) 474 (18%) 1181 (19%) ferences of the estimate for the given strain and sex in the column relative to AKOS 2513 (69%) 1694 (63%) 4207 (66%) that on the diagonal (e.g., AKOS estimates were 0.838 those of AKOC). GIFT 424 (12%) 520 (19%) 944 (15%) Total 3644 (100%) 2688 (100%) 6332 (100%) AKOC AKOS GIFT 1 AKOC 0.003 ± 0.251 0.838 (P = 0.09) 0.802 (P = 0.13) AKOC = Akosombo control line, AKOS = Akosombo selected line, AKOS 0.180 ± 0.243 0.957 (P = 0.89) GIFT = GIFT strain. GIFT 0.224 ± 0.255 Table 4 ANOVA for fixed effects (strain and sex), interaction, and covariate (total age-at three groups means that there will be direct competition for food and -arvest) on harvest weight. space between them. Selection for growth in GIFT has been found to in- Effect DF (num) DF (den) F-value p-value crease competitiveness (e.g., for feed) or aggressiveness (e.g., the ability to suppress growth of other individuals) (Khaw et al., 2016). Observations of (Intercept) 1 6037 14.9 < 0.0001 the Akosombo and GIFT fingerlings in separate hapas reported common Strain 2 287 471.1 < 0.0001 gathering of GIFT fish to the surface with mouths gaping open at feeding Sex 1 6037 505.9 < 0.0001 Strain × sex interaction 2 6037 7.9 0.0004 times, but rarely by Akosombo strain. Given those observations, finger- Total age-at-harvest 1 6037 72.1 < 0.0001 lings were sampled during the pond experiments. The identity of the tagged fish showed that the majority (80–90%) near the feeders at the start of the feeding periods were GIFT fingerlings. GIFT success in direct female GIFT individually weighed 106.5 g more than AKOC, and 89.9 g competition for food and space, may have contributed to its being able to more than AKOS. The AKOS performed better than AKOC, 15.9 g better grow faster and depress the growth of the Akosombo strains relative to for males and 16.6 g for females. All differences in terms of harvest situations in which the strains were reared separately. weight between different groups of fish were highly significant One way to test whether or not these effects occurred is to compare (P < 0.001). Male fish were found to be significantly larger than fe- the relative performance of the strains when reared separately. This was males for all three strains (P < 0.001). For AKOC, the difference done by comparing the growth of the fish used in the experiment with (35.7 g) was similar to that of AKOS (35.0 g). The difference between that of the previous generation, accessing data from the breeding pro- male and female GIFT was smaller (27.8 g) (Table 5). gram where the GIFT and Akosombo lines were reared separately. The odds for AKOC survival was 83.8% compared to AKOS and Given that the period over which growth was measured was similar 80.2% compared to GIFT, while the odds of survival for AKOS was (approximately 230 days), the comparison is most simply made using 95.7% compared to GIFT. However, all pairwise differences in LSM for the daily growth rate for each strain (Table 7). These results showed survival were not significant (P = 0.09 to 0.89) (Table 6). slower growth rate in the pond experiment relative to those in the breeding program for all strains. This may be due to a number of dif- 4. Discussion ferences in the environmental circumstances between the two sets of measurements. However, the pertinent issue is comparison of the re- The objective of this study was to compare growth performance and lative growth rates among strains. The reduction in growth rate in the pond experiment relative to the breeding program was greater for male survival of the Akosombo strain developed in Ghana with those of the GIFT strain developed in Asia, both of which had been selected for per- AKOS (35%) and AKOC (39%) than for male GIFT (22%) suggesting an effect of competition between the males accounting for around 10–15% formance in equatorial environments. In the present study, GIFT grew faster than the Akosombo strain and AKOS faster than AKOC. GIFT grew to of the growth difference between the Akosombo strains and GIFT in the pond experiment. Since AKOS was more abundant than both AKOC and a larger size, despite being smaller in size when stocked. However, some aspects of the experimental design may have differentially biased the es- GIFT, these results suggest no obvious effects of density differences in the pond experiment. Although the pattern of change was somewhat timates of growth of the different strains. First, the numbers of fish in each strain were unequal throughout the experiment, with the AKOS strain different for females, there was no evidence that any growth differences were related to density. Growth rates for females were reduced to a having more families and individual fish than the GIFT and AKOC strains (Tables 2 and 3). Where there are marked differences in abundance of two lesser degree than males, with AKOC having a greater decline (29%) than either AKOS (16%) or GIFT (18%), between which there was little groups, assuming no restriction for food, the rarer group might have a growth advantage as a result of reduced numbers (or density) relative to difference. Thus, the female data suggested no competitive effect. Overall, these results provide limited evidence for an effect of direct the more common group (Brett, 1979), although this is debated as being dependent on the natural schooling and feeding relations of the species or competition, perhaps restricted to males, but no effect of differences in relative density in the current experiment. groups (Yamagishi, 1963, 1969). The second issue is the co-rearing of the Table 5 Least-square means of harvest weight for each sex in each strain (on the diagonal, value = estimate ± standard error) and pairwise comparisons (off diagonal). Differences on the diagonal are in grams. The pairwise comparisons indicate the differences of the estimate for the given strain and sex in the column relative to that on the diagonal (e.g., AKOC female estimates were 35.7 g less than AKOC males). All pairwise comparisons were significant (P < 0.001). AKOC, female AKOC, male AKOS, female AKOS, male GIFT, female GIFT, male AKOC female 43.2 ± 15.8 −35.7 −16.6 −51.6 −106.5 −134.3 AKOC male 78.9 ± 15.8 19.1 −15.9 −70.9 −98.6 AKOS female 59.8 ± 15.7 −35.0 −89.9 −117.7 AKOS male 94.8 ± 15.7 −54.5 −82.7 GIFT female 149.7 ± 15.9 −27.8 GIFT male 177.5 ± 15.9 4 T.Q. Trinh, et al. Aquaculture 530 (2021) 735938 Table 7 based BLUP calculation for breeding values and a mating scheme designed Daily growth rate (g day−1) of three strains (GIFT, AKOS, and AKOC) in the to maintain inbreeding at an acceptable level of ~0.5% per generation). current experiment and previous generation of the breeding program in Ghana, There were factors that might have hindered the response to selection in the and the difference in growth rate between the two expressed as a percent of the Akosombo strain. Although three wild stocks were collected from three breeding program growth rate. The period of growth for both the experimental ecological zones in Ghana, the fourth was from a fish farm at Nsawam and the breeding-program fish was approximately 230 days. Daily growth where the fingerlings were introduced by the Institute of Aquatic Biology in rate = harvest weight/total age from spawning until harvest. 1982 and had not been replenished for 24 years according to anecdotal data Strain Daily growth rate (g day−1) in Attipoe et al. (2013). After the first two generations of selection, Attipoe et al. (2013) reported that the heritability was 0.039 ± 0.014. Despite Males Females Experiment (reared together) experimental difficulties that may have affected accuracy of heritability GIFT G2 Ghana experiment 0.56 0.47 estimation, such as tag loss and high mortalities (Attipoe et al., 2013), this AKOS G10 experiment 0.30 0.21 value indicated that the genetic variation for growth was low in the Ako- AKOC G10 experiment 0.25 0.15 sombo strain. These data also suggest a fundamental genetic difference with Breeding program (reared separately) GIFT G1 Ghana 0.72 0.57 respect to growth rates between the Akosombo and GIFT strains. AKOS G9 0.46 0.25 It is well known that sexual dimorphism exists in Nile tilapia popu- AKOC G9 0.41 0.21 lations, with males growing faster than females (Beveridge and Differences (%) of experiment compared to McAndrew, 2000). The same trend was reported in GIFT (Bentsen et al., breeding program 2017). It has been assumed that growth of both males and females of the GIFT −22 −18 AKOS −35 −16 GIFT strain was simultaneously improved, although the difference be- AKOC −39 −29 tween two sexes over time caused by long-term selection has not yet been quantified. In from the present work, the difference between GIFT male and female fish was smaller than those of AKOS and AKOC, and that of This interpretation agrees with the first assessment from data cal- AKOS significantly smaller than of AKOC (Tables 2 and 5), indicating culated in 2012, where GIFT was reported to grow 1.6 times faster than that selection improved growth for both sexes and at the same time re- Akosombo strain after 122 days (CSIR - Water Research Institute, duced the difference between the two sexes. 2013). As noted, the present experiment may have overestimated the Individual survival at harvest was not significantly different be- differences between GIFT and AKOS, but it is clear that GIFT grows 1.6 tween the Akosombo strains and GIFT (Tables 3 and 6). For GIFT, to 1.9 times faster than AKOS. survival remained the same after nine generations of selection (Ponzoni It is pertinent for this discussion to consider the results of studies during et al., 2011). Given that both programs used the same breeding ap- the initial testing of populations used to create the base population of the proach, the same trend would be expected for the Akosombo strain and GIFT strain. The wild Ghanaian strain tested in the 1980s (220 fingerlings is suggested by the similar survival rates observed in the experiment. collected from the Volta River System in October 1988) was the poorest There are debates about the use of indigenous or imported elite strains performer during the inter-strain comparison that preceded to establish- with respect to balancing conservation and development, and differences in ment of the base population of GIFT (Eknath et al., 1993). In addition, the views of the extent to which these strains are adapted to local ecosystems and farmed strains from Israel, Singapore, and Taiwan, all of which originated to ecological and their economic sustainability (Ansah et al., 2014; Anane- from Ghana, were intermediate- to slow-growing (Eknath et al., 1993, Taabeah et al., 2019). These issues will be debated in respect of the choice of 2007). All four were the slowest-growing strains in all environments (cage, strain for use in the development of the aquaculture industry in Ghana and pond, and rice-fish cultures) in which the fish were tested, thus, all strains the Volta basin at large. The results of the present study demonstrated marked derived from Ghana contributed only 2.6 to 8.5% each to the founding differences in the performance of GIFT and Akosombo strains relative to population of GIFT. The representation of all these strains was reduced over growth, and strong growth performance of GIFT in an equatorial West time following selection for faster growth in the presence of other superior African environment, with survival similar to that of the local strain. genetic material, and after five generations of selection provided only 0.6 to Significant gains achieved in the Akosombo genetic improvement 5.3% of the GIFT population (Bentsen et al., 2017). The extent to which program over ten years are less than those achieved in the GIFT program, bottlenecks and consequent inbreeding may have affected the Ghanaian even after taking account of the longer period of selection for GIFT. The strain(s) introduction into Asia, and into GIFT, is unknown. However, the data suggest that it would require more than 20 years of selection for the numbers of individuals included in the GIFT program from the strains in- Akosombo strain to achieve similar levels of growth to the GIFT strain cluding the Ghanaian strains were comparable to that for other sources. generation tested, during which time continued selection of GIFT would Therefore, these results suggest a fundamental genetic difference with re- result in even greater improvement in performance. Use of GIFT would spect to growth rates between the Ghanaian strain and the other source be expected to provide a huge boost in commercial tilapia production in populations of GIFT compared more than 30 years ago. Ghana and the sub-region. In the absence of a formal and transparent This study, however, does not suggest that Ghanaian stocks will not process for responsible introduction and management of improved respond to selection. It is clear that the Akosombo program has suc- strains, the possibility of the risk of illegal or uncontrolled introduction cessfully produced, over 10 years of selection, a strain that grows about and mismanagement of stocks of unverified origins with potential dete- 30% faster than unimproved Ghanaian stocks (CSIR - Water Research riorative effects can be expected. Genetic characterization of tilapia Institute, 2013). The present experiment found AKOS grew 26% faster purchased in Ghana (Anane-Taabeah et al., 2019) suggested that there than AKOC, with males 22% and females 42% faster. However, the GIFT may have been informal introductions to Ghana contrary to Ghanaian strain was reported to show a 67% improvement in growth compared to policy at the time. There could be significant consequences for the de- the base population or 88% improvement compared to the control group velopment of the aquaculture industry in Ghana and the sub-region de- after only five generation of selection (Bentsen et al., 2017). pending on choice of strains and corresponding responsible management. When transferred to Ghana in 2012, GIFT had been selected for seven generations in the Philippines and thereafter 11 generations in Malaysia, a total of 18 generations of selection, with Ponzoni et al. (2011) reporting Declaration of Competing Interest more than 10% improvement per generation sustained over more than six generations. While there were likely differences in the management of The authors declare that they have no known competing financial Akosombo and GIFT programs in terms of number of families per genera- interests or personal relationships that could have appeared to influ- tion, the approaches in these two programs were the same (i.e., pedigree- ence the work reported in this paper. 5 T.Q. Trinh, et al. Aquaculture 530 (2021) 735938 Acknowledgements Dey, M.M., Eknath, A.E., Sifa, L., Hussain, M.G., Thien, T.M., Hao, N.V., Aypa, S., Pongthana, N., 2000. Performance and nature of genetically improved farmed tilapia: The authors thank Ms. Lily O. Boateng and Mr. Godfred Yeboah for a bioeconomic analysis. J. Aquacult. Econ. Manag. 4, 83–106. Eknath, A.E., Tayamen, M.M., Palada-de Vera, M.S., Danting, J.C., Reyes, R.A., Dionisio, their technical assistance. This study was conducted under the frame- E.E., Capili, J.B., Bolivar, H.L., Abella, T.A., Circa, A.V., Bentsen, H.B., Gjerde, B., work of the Tilapia Volta (TIVO) project, coordinated by UN FAO, with Gjedrem, T., Pullin, R.S.V., 1993. Genetic improvement of farmed tilapias: the growth Chris Nugent as the Chief Technical Advisor. Funding was provided by performance of eight strains of Oreochromis niloticus tested in different farm en- vironments. Aquaculture 111, 171–188. the Spanish government, and technical support from WorldFish. The Eknath, A.E., Bentsen, H.B., Ponzoni, R.W., Rye, M., Nguyen, N.H., Thodesen, J., Gjerde, analysis for and writing of this report by WorldFish staff was supported B., 2007. Genetic improvement of farmed tilapias: composition and genetic para- by European Commission-IFAD Grant Number 2000001539, the meters of a synthetic base population of Oreochromis niloticus for selective breeding. Aquaculture 273, 1–14. International Fund for Agricultural Development (IFAD) and the CGIAR FAO, 2020. The State of World Fisheries and Aquaculture 2020: Sustainability in Action. Research Program on Fish Agri-Food Systems (FISH) led by WorldFish. Food and Agriculture Organization of the United Nations, Rome. The latter program is supported by contributors to the CGIAR Trust Gupta, M.V., 2002. Genetic enhancement and conservation of aquatic biodiversity in Africa. NAGA WorldFish Center Q. 25, 48–53. Fund. Gupta, M.V., Acosta, B.O., 2004. A review of global tilapia farming practices. Aquacult. Asia Mag. IX (1), 7–12. References Khaw, H.L., Ponzoni, R.W., Yee, H.Y., Aziz, M.A., Bijma, P., 2016. Genetic and non-ge- netic indirect effects for harvest weight in the GIFT strain of Nile tilapia (Oreochromis niloticus). Aquaculture 450, 154–161. Anane-Taabeah, G., Frimpong, E.A., Hallerman, E., 2019. Aquaculture-mediated invasion Komen, J., Trong, T.Q., 2014. Nile tilapia genetic improvement: achievements and future of the Genetically Improved Farmed Tilapia (GIFT) into the lower Volta Basin of directions. In: The 10th International Symposium on Tilapia in Aquaculture - ISTA10, Ghana. Diversity 11 (10), 188. October 6–10, 2013, Jerusalem, Israel. Ansah, Y.B., Frimpong, E.A., Hallerman, E.M., 2014. Genetically-improved tilapia strains Kuznetsova, A., Brockhoff, P.B., Christensen, R.H.B., 2017. lmerTest package: tests in in Africa: potential benefits and negative impacts. Sustainability 6 (6), 3697–3721. linear mixed effects models. J. Stat. Softw. 82 (13), 1–26. Asian Development Bank, 2005. An Impact Evaluation of the Development of Genetically Lenth, R., Buerkner, P., Herve, M., Love, J., Riebl, H., Singmann, H., 2020. Package Improved Farmed Tilapia and their Dissemination in Selected Countries. Operations ‘emmeans’, Version 1.5.0. https://github.com/rvlenth/emmeans. Evaluation Department Asian Development Bank 124 pp. Lind, C.E., Agyakwah, S.K., Attipoe, F.Y., Nugent, C., Crooijmans, R.P.M.A., Toguyeni, A., Attipoe, F.Y., Blay Jnr., J., Agyakwah, S., Ponzoni, R.W., Khaw, H.L., Abban, E.K., 2013. 2019. Genetic diversity of Nile tilapia (Oreochromis niloticus) throughout West Africa. Genetic parameters and response to selection in the development of Akosombo strain Sci. Rep. 9 (1), 1–12. of the Nile tilapia (Oreochromis niloticus) in the Volta Basin, Ghana. In: Proceedings of Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., Core Team, R., 2020. nlme: Linear and the International Symposium on Tilapia in Aquaculture, Jerusalem, Israel, 6–10 Nonlinear Mixed Effects Models. R Package Version 3. pp. 1–149. https://CRAN.R- October 2013, https://doi.org/10.3382/ps.2009-00198. project.org/package=nlme. Bates, D., Mächler, M., Bolker, B., Walker, S., 2015. Fitting linear mixed–effects models Ponzoni, R.W., Nguyen, N.H., Khaw, H.L., Hamzah, A., Bakar, K.R.A., Yee, H.Y., 2011. using lme4. J. Stat. Softw. 67 (1), 1–48. https://doi.org/10.18637/jss.v067.i01. Genetic improvement of Nile tilapia (Oreochromis niloticus) with special reference to Bentsen, H.B., Gjerde, B., Eknath, A.E., de Vera, M.S.P., Velasco, R.R., Danting, J.C., the work conducted by the WorldFish Center with the GIFT strain. Rev. Aquac. 3, Dionisio, E.E., Longalong, F.M., Reyes, R.A., Abella, T.A., Tayamen, M.M., Ponzoni, 27–41. R.W., 2017. Genetic improvement of farmed tilapias: response to five generations of WorldFish Center, 2007. Policy on the transfer of Genetically Improved Farmed Tilapia selection for increased body weight at harvest in Oreochromis niloticus and the further (GIFT) from Asia to Africa by the WorldFish Center. In: The WorldFish Center impact of the project. Aquaculture 468, 206–217. Working Papers. WorldFish Center, Penang, Malaysia. Beveridge, M.C.M., McAndrew, B.J., 2000. Tilapias: Biology and Exploitation. Kluwer Yamagishi, H., 1963. Some observations on growth variation and feeding behavior in the Academic Publishers, Dordrecht, The Netherlands. fry of two races of Japanese crucian carp, Carassius carassius. L. Jap. J. Ecol. (Nippon Brett, J.R., 1979. 10 - environmental factors and growth. In: Hoar, W.S., Randall, D.J., Seitai Gakkaishi) 13, 156–161. Brett, J.R. (Eds.), Fish Physiology. Academic Press, New York, pp. 599–675. Yamagishi, H., 1969. Postembryonal growth and its variability of the three marine fishes R Core Team, 2020. R: A Language and Environment for Statistical Computing. R with special reference to the mechanism of growth variation in fishes. Res. Popul. Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org. Ecol. 11, 14–33. Council for Scientific and Industrial Research (CSIR) - Water Research Institute, 2013. Yuan, Y., Yuan, Y., Dai, Y., Zhang, Z., Gong, Y., Yuan, Y., 2020. Technical efficiency of Performance Comparison of GIFT and Akosombo Strains of the Nile Tilapia different farm sizes for tilapia farming in China. Aquac. Res. 51, 307–315. (Oreochromis niloticus). 19 pp. TIVO Project Technical Report. 6