From the pond to the creek many years ago: life-history dynamics of feral poeciliids in Brazil

.


Introduction
In Brazil, aquatic invasions have been recognized as one of the most serious environmental problems since the last two decades (Pelicice et al., 2017;Vitule et al., 2019;Geller et al., 2020).Invasive aquatic organisms have spread over different habitats, causing multiple ecological disturbances and costs to Brazilian society (Adelino et al., 2021;Magalhães et al., 2021).
One of the major pathways by which invasive aquatic species are introduced and spread throughout Brazil is the aquarium trade (Magalhães & Jacobi, 2017;Vitule et al., 2019;Geller et al., 2020).With the growing pursuit of fish keeping, several aquarium fish species have recently been documented in inland waters, all of which are directly associated with aquarium dumping (Magalhães et al., 2017 and references therein) or escapes from fish farms (Magalhães et al., 2019(Magalhães et al., , 2020(Magalhães et al., , 2021)).Some records are diagnosed as occasional, e.g., the translocated blackline tail tetra Moenkhausia costae, bristlemouth catfish Ancistrus multispinis (Magalhães et al., 2019); whereas some others, including the exotic zebra fish Danio rerio and oriental wheather loach Misgurnus anguillicaudatus, have established viable populations in several wild-type phenotypes suggests establishment of populations in the study area.We hypothesize that the introduction of both species has occurred several years ago, and the continued exposure to predation seems to have molded their life-history traits and phenotype.
(i.e., phenotypic changes) of feral populations in response to shifts in ecological conditions (e.g., predation), studies are non-existent.
In this paper, we investigated ecological aspects of P. reticulata and X. maculatus introduced into a small creek in the region of Muriaé Ornamental Aquaculture Center, Minas Gerais State, with a special focus on life-history aspects of feral populations.In particular, this study investigated: (i) the trophic and reproductive dynamics of these fish (i.e., developmental stages, population structure for each developmental stage, gonadosomatic index and the association with environmental variables), (ii) the sex ratio, (iii) the possible establishment/ evolution of feral populations (i.e., presence of young and alteration of phenotype ornamentaltype to wild-type), and (iv) suggested management strategies to avoid new introductions of feral guppies and platies in the area affected by the largest ornamental aquaculture center in Brazil.

Study area
This study was conducted at an ornamental aquaculture center in the Muriaé region, located in the "Zona da Mata" of the Minas Gerais State, Brazil.The Muriaé Ornamental Aquaculture Center, the largest establishment in Brazil since 1979[sensu Vidal Júnior & Costa (2000)], comprising 350 fish farms and 4,500 production ponds bordering small creeks located in an Atlantic Forest region.Most of the ponds are drained eight times a year, resulting in more than 30,000 release events per year (Magalhães et al., 2020).The majority of these ponds do not have devices to prevent the escape of non-native fish.As a result, fish escapes have been routine, and several small-sized ornamental fish have invaded and established populations in different creeks of the region (Magalhães & Jacobi, 2013a, 2017;Magalhães et al., 2020Magalhães et al., , 2021)).
Fish were collected every two months, from January to December 2017, in Lopes Creek (20 o 50'16.5"S;42 o 13'48.9"W),municipality of Vieiras, Minas Gerais State, southeastern Brazil (Figure 2).Juveniles (P.reticulata: 91, X. maculatus: 165) and adults (P.reticulata: 658 females, 402 males, X. maculatus: 364 females, 36 males) were collected from margins and channel bed, with a rectangular sieve measuring 95 cm long × 25 cm high and 0.3 mm mesh.The sieve was hauled 50 times along a 100 m-long transect during 2 h in Lopes Creek at daytime.The fishes were anesthetized on an ice slurry (IACUC, 2002) and soon after, sacrificed by immersion in Eugenol (Griffiths, 2000): then, the material was sorted by site collection, packed in plastic bags, fixed in 10% formalin, and subsequently transferred to 70% alcohol.Voucher specimens are deposited under their respective catalogue numbers in the Universidade Federal de Sergipe fish collection (CIUFS): P. reticulata CIUFS 2244, X. maculatus CIUFS 2260.The samplings were carried out under authorization from the Chico Mendes Institute for Biodiversity Conservation-ICMBio, granted through the System of Authorization and Information on Biodiversity-SISBIO, nº 61740.In the laboratory, all fish were measured (total and standard lengths, cm), weighted (g) and eviscerated to remove the stomach and gonads.

Feeding habits
Stomach items were analyzed under a stereoscopic microscope and, when possible, identified to the Acta Limnologica Brasiliensia, 2023, vol. 35, e30 lowest taxonomic level.The relative importance of each food item was calculated applying the Feeding Index (Kawakami & Vazzoler, 1980) described as: IAi = ((Fi × Vi)/(Ʃ Fi × Vi)) × 100, where i = 1, 2, . . .n are the food items, Fi is the frequency of occurrence of item i, Vi is the relative volume of item I, and Vi is the relative volume of item i.

Developmental stages
We adapted the methods of Haynes (1995) to determine the developmental stages of adult females.We classified as non-gravid those females without a conspicuously enlarged ventral region, absence of eggs, embryos, or gravid spot near the base of the anal fin.Females were classified as gravid when they had enlarged belly, presence of gravid spot, eggs or embryos.Developmental stages were then classified as follows: Non-gravid (stage 1 -ovarian tissue without ovum and/or embryos), Yolked ova (stage 2 -ovum in the process of yolking), Eyed embryos (stage 3 -eyes pigmented, head enlarged compared to trunk, caudal and pectoral fin buds present), Mature embryos (stage 4 -yolk sac mostly or completely absorbed, pectoral fins elongated, embryo resembling a small adult) (Figure 3).After determining the developmental stages, the bimonthly absolute and relative frequencies of females were calculated.

Fecundity
We calculated the bimonthly and total mean values of fecundity (number of embryos in stages 3 and 4 per female).Only embryos were used as a measure of fecundity because ova development and fertilization are not discernible with a dissecting microscope (Haynes, 1995).

Population structure for each developmental stage
The population structure associated with developmental stages was determined for females using six standard length classes (0.5 cm intervals) (Hojo et al., 2004).We calculated the number of fish within each length class for each developmental stage.

Gonadosomatic index and environmental factors
The gonadosomatic index (GSI) (Hernández et al., 2004) was calculated bimonthly for females using the formula: GSI = (GW/BW × 100), where GW is the gonad weight, and BW is the body weight.We grouped the bimonthly mean values of GSI of females which were then overlayed with the average bimonthly rainfall (mm), and the water level (cm) in the Lopes Creek for identification of potential trends between biotic (GSI) and abiotic factors (rainfall, water level/temperature).Rainfall data was obtained from the database of the weather station A517 located in the municipality of Muriaé (INMET, 2020), water level was measured using a 2-m-long pole marked every 10 cm and water temperature was obtained with a mercury-in-glass thermometer to the scale 0º to 50ºC.

Sex ratio
The sex ratio (Pope et al., 2010) was calculated from the bimonthly and total number of females and males of P. reticulata and X. maculatus collected in the Lopes Creek.Chi-square (χ 2 ) goodnessof-fit test was used to check whether significant differences were present in the sex ratio (Sokal & Rohlf, 1995).Differences were significant when p < 0.05.The analysis was performed using the Paleontological Statistics-Past software (3.25 version) (Hammer et al., 2001).

Establishment and feral status
The adult-to-juvenile relative frequencies (Pope et al., 2010), were calculated bimonthly for P. reticulata and X. maculatus collected in Acta Limnologica Brasiliensia, 2023, vol. 35, e30 the Lopes Creek.The phenotype of females and males of guppies and platies was also analyzed (i.e., in percentages) to verify variations in color patterns.Two color patterns were considered as basic phenotypic categories: (i) ornamental-type and (ii) wild-type.Ornamental-type are colorful forms cultivated on the ornamental fish farm that surrounds Lopes Creek: guppy females are plainer in color and have a smaller colorful tail, while males are brighter and have colorful tails with various shapes such as round, flag, delta, or half-moon.In southern platyfish, both females and males express bright color patterns including yellow/red pigmentation (Figure 4A, 4B, 4C, 4D).Wild-type guppy females are dull brown, males contained pale orange stains and small black spots on the anterior/posterior lateral part of the body and a translucent small tail.Southern platyfish females and males are beige with or without melanic/speckled patches on the body and presence or absence of black spots on the caudal peduncle (Figure 4E, 4F, 4G, 4H).Each captured fish was assigned as ornamental or wild-type.

Feeding habits
The guppy and southern platyfish fed primarily on detritus (unspecified organic and inorganic material), followed by aquatic insects (i.e., dragonfly nymphs, true fly larvae) and zooplankton.Both species also consumed food resources of allochthonous origin such as ants and flies (Table 1).

Developmental stages
Females of P. reticulata (n = 525) and X. maculatus (n = 219) were found in several reproductive stages (i.e., stages 2, 3, 4) during the sampling period in Lopes Creek.Females in stages 3 and 4 were found in all months for P. reticulata, and between September and April for X. maculatus (Figure 5).

Fecundity
The mean fecundity of guppy and southern platyfish was greater than 27 and 30 embryos per brood respectively in each bimester; considering the total period, it was greater than 32 and 34 embryos per brood respectively (Figure 6).

Population structure for each developmental stage
We observed wide variation in the proportion of developmental stages within length classes.Virtually all stages were recorded in each length class; stages 2 and 3 predominated across length classes for P. reticulata, while stages 1 and 2 predominated for X. maculatus.In Lopes Creek, the smallest gravid female of guppy measured 1.4 cm TL (0.8 cm SL) and southern platyfish measured 1.5 cm TL (1.0 cm SL).The largest P. reticulata females were 3.9-4.4TL (3.3-3.8 cm SL) and X. maculatus 4.5-5.0TL (4.0-4.5 cm SL) (Table 2).

Gonadosomatic index and environmental factors
For both species, GSI of females showed peaks coinciding with the highest rainfall, and water level/ temperature values (Figure 7).Acta Limnologica Brasiliensia, 2023, vol.35, e30

Establishment and feral status
Juveniles of P. reticulata and X. maculatus were found during the entire sampling period in Lopes Creek, although in low proportions (Figure 8).
For both species, all individuals captured were classified as wild-type forms.

Discussion
This study provided information about trophic, reproductive status, and the establishment of feral P. reticulata and X. maculatus in a small Atlantic Forest creek in Brazil.Both species are popular among fish farmers in the study region, and have been raised in ponds for decades [since 1979, see Vidal Júnior & Costa (2000)], which implies that they have been released in the creeks continuously (Magalhães & Jacobi, 2017;Magalhães et al., 2020Magalhães et al., , 2021)).Our results showed that feral guppies and platies dominate and colonize the studied creek, a probable result of the environmental selection that favors the survival and recruitment of more discrete phenotypes.Moreover, both species colonized successfully the environment, considering they fed on abundant resources and showed prolonged reproductive activity, including early maturity and the presence of young (i.e., recruitment).This is a particular case in which ornamental fish (i.e., colorful patterns) escape from production ponds and rapidly evolve to feral strains (i.e., discrete or cryptic patterns) in the environment.
The analysis of the trophic activity of guppy and southern platyfish demonstrated that detritus dominated the diet of these feral poeciliids in the Lopes Creek.The same tendency occurred with feral P. reticulata in streams located in São Paulo and  Paraná states and X. maculatus in an Indonesian lake (Green et al., 1978;Rocha et al., 2009;Silva et al., 2012).The opportunistic feeding on a variety of animal resources with a major component of detritus may be the common feeding pattern in feral poeciliids (Green et al., 1978).This is consistent with intestinal morphology, i.e., long and coiled, varying in length from one to two times the standard length (Dussault & Kramer, 1981), an adaptation for digesting detritus (Silva et al., 1980).The high *Indicates the significant values for sex ratio (Chi-square; p < 0.05, χ 2 0.05 = 3.84; d.f.= 1).Acta Limnologica Brasiliensia, 2023, vol.35, e30 consumption of detritus, which is a resource with low nutritional value, but highly abundant, can indicate a "compensatory feeding strategy" (Yeager et al., 2014, p. 2), a behavior that may facilitate the colonization of non-native poeciliids in new environments (Alves et al., 2016).The increased consumption of detritus by invasive poeciliids can impact nutrient dynamics in streams by mobilizing organic matter by feeding, and releasing nutrients in more liable forms via excretion.Increased nutrient inputs, especially nitrogen, can change stream food web structure and ecological functions such as nutrient cycling and primary production (Vanni, 2002;Holitzki et al., 2013).We found females of feral guppy and southern platyfish in reproductive activity during the 12 months of the study period.A long reproductive period has also been found for feral P. reticulata and X. maculatus in five headwater creeks at the Muriaé Ornamental Aquaculture Center and in an Australian stream (Milton & Arthington, 1983;Magalhães & Jacobi, 2017).Frequent reproduction over an extended breeding season is an opportunistic life-history strategy adopted by poeciliids that permit colonization in environments with strong predation pressure (Reznick & Endler, 1982), and this might be the case in the Lopes Creek, which is inhabited by different predators, i.e., aquatic spiders/insects, native and non-native fish (see Figure 1).The fecundity observed for the two feral species was high (> 30 embryos per brood).High fecundity was also found in X. maculatus introduced in an Australian stream subjected to high predation pressure (Milton & Arthington, 1983), and conversely, low fecundity was found in P. reticulata inhabiting a Brazilian urban reservoir with few predators (Oliveira et al., 2014).High fecundity is a specific trait adopted by native and non-native poeciliids to assure the maintenance of a viable population in an environment with high predation rates (Reznick & Endler, 1982;Sa-nguansil & Lheknim, 2010).Higher fecundity can also give non-native livebearers a competitive advantage over native species, because fry grow more quickly and become adult predators faster, completely dominating the invaded site (Magalhães & Jacobi, 2017).
Our observations showed that adult breeding females of feral P. reticulata and X. maculatus were found in small sizes (ca.1.0 cm SL), which can indicate early sexual maturation.This process may be related to size-selective predation performed by a suite of aquatic predators (Figure 1), including native fishing spiders, diving beetles, water boatmen, water striders, dragonfly nymphs, water bugs, native fish P. parahybae, D. janeiroensis, P. vivipara and nonnative fish originated from a nearby ornamental fish farm, such as cyprinids, danionids, cobitids, cichlids, characids, and other poeciliids.According to Endler (1982) and Reznick & Endler (1982), the larger the body size, the greater the chance to be visible to the predators and therefore, many animals -including poeciliids -compensate this risk by reducing their body size (and consequently sexual maturation) in order to avoid predation.This is probably the lowest size found in introduced populations of feral P. reticulata and X. maculatus.In other sites and environments throughout Brazil and the world, the smallest reproductive size of feral guppy and southern platyfish ranged between 1.2-2.0cm, and 1.7-2.0cm SL respectively (Milton & Arthington, 1983;Andrade et al., 2008;Montag et al., 2011;Kırankaya & Ekmekçi, 2021;Luduvice & Brito, 2023).In the present study, we also found smallest females of P. reticulata and X. maculatus measuring 4.4-5.0cm TL respectively, indicating a feralization process for both species.Previous studies have documented that ornamental strains of guppy and southern platyfish females living in predation free environments (i.e., aquariums) are usually larger, e.g., 6.0 cm TL (Sterba, 1983), than their feral counterparts, e.g., 4.0-5.1 cm TL in P. reticulata and X. maculatus (Favilli et al., 2018;Kırankaya & Ekmekçi, 2021), likely due to selecting forces such as predation acting upon the body size (Endler, 1982;Reznick & Endler, 1982).
The gonadosomatic index indicated that feral P. reticulata and X. maculatus females have the peak of reproduction activity occurring during the austral summer, characterized by warmer temperatures, rainy season and rising water level in Lopes Creek.The main breeding period for many non-native poeciliid species, including guppy and southern platyfish, occurs during periods of heavy rains which lead to the expansion of the littoral zone, thus increasing the availability of food resources for the young (Magalhães et al., 2002;Gómez-Márquez et al., 2016).In addition, warmer water temperatures may increase embryonic development rate and consequently, more broods produced by this group of fish (Snelson, 1989).
We also recorded a higher frequency of females.A similar pattern was found for feral P. reticulata and X. maculatus introduced into a stream located in Sergipe State and in Australia (Milton & Arthington, 1983;Luduvice & Brito, 2023).The prevalence of females may also be associated with high predation From the pond to the creek many years ago... Brasiliensia, 2023, vol. 35, e30 pressure, which may cause differential predation risk between females and males.In poeciliids, males of P. reticulata and X. maculatus perform prolonged courtship displays (e.g., sigmoidal body curves, lateral displays with erect fins), continuously chasing and mating with females without their consent (i.e., coercive mating), expending a large part of the day in these activities.These behaviors -along with conspicuous colors and ornaments such as larger tails/fins -may expose males to predators, as they become less careful in avoiding predator attacks, biasing the sex ratio toward females (Borowsky & Kallman, 1976;Stoner & Breden, 1988;Snelson, 1989 and references therein).

Acta Limnologica
The proportion of juveniles was lower than adults in the studied populations, contrasting with the high occurrence of gravid females.Again, selective predation may be associated with this pattern.Massive predation events were observed on juveniles of feral poeciliids in the Lopes Creek, performed especially by visually oriented aquatic invertebrates and fish (A.L.B.Magalhães, pers.obs.) (see Figure 1).In spite of the low proportion of juveniles, the year-round presence of reproductive females and juveniles suggests that feral P. reticulata and X. maculatus are established in the Lopes Creek.The same pattern was found for guppy introduced in a stream in Serbia, and guppy and southern platyfish introduced in an Italian pool (Milenković et al., 2014;Favilli et al., 2018).A strong evidence of local establishment is the alteration of phenotype patterns, in the case, the shift from ornamentaltype phenotypes (Figure 4A,4B,4C,4D) Alley et al. (2023)].The presence of wild forms of feral P. reticulata and X. maculatus in Lopes Creek may be a result of sixteen years of intermittent escapes and environmental selection, especially shaped by the predation pressure imposed by fishing spiders, numerous aquatic insects and fish (see Figure 1), denoting a rapid selection of life-history traits.This assumption is supported by studies conducted by Endler (1978Endler ( , 1980) ) and Milenković et al. (2014) who showed that predation makes poeciliid populations to become cryptic by eliminating the most colorful individuals from the gene pools.Thus, feral populations evolve to become more closely matched with their environment as an anti-predation strategy.The focus of this study was to investigate trophic/reproductive traits and change of phenotypes related to predation.However, we recognize that other natural "forces" may also contribute to the shift in phenotypes observed here such as dominance of wild-type alleles (Fisher, 1930), and sexual selection: gradual decrease of ornamental-type individuals due to a presumably female natural preference for wild-type males (Houde, 1994).
Headwater creeks of Muriaé Ornamental Aquaculture Center are particularly vulnerable to invasion by ornamental non-native species, because this region has low native richness, while it has received intense and constant propagule pressure of ornamental non-native organisms escaping from production ponds (Magalhães & Jacobi, 2013a, 2017;Magalhães et al., 2002Magalhães et al., , 2019Magalhães et al., , 2020Magalhães et al., , 2021Magalhães et al., , 2022)).Negative impacts of non-native Poecilia and Xiphophorus have been reported elsewhere (Arthington et al., 1983;Courtenay et al., 1988;Englund, 1999;Magalhães et al., 2021 and references therein), but the ecological situation in the region seems especially complicated, considering that the ecological impacts of feral guppy and southern platyfish may be intensified by the invasion of other feral poeciliids, including black molly, sailfin molly, green swordtail, and variable platyfish.Impacts may be especially intense on aquatic insects, zooplankton, D. janeiroensis, P. parahybae, P. vivipara, food web structure and nutrient cycling/ primary production.
Modifying human behavior is the key to prevent or minimize the potential ecological impacts of ornamental aquaculture in Brazil [see Magalhães & Jacobi (2017);Lima Junior et al. (2018); Vitule et al. (2019); Luduvice & Brito (2023)].Therefore, the main management action should include awareness campaigns to educate people who work directly with the ornamental fish industry in the region (Magalhães et al., 2021).Without a change in the behavior of these workers, it will be impossible to avoid fish escaping (i.e., propagule pressure) from production ponds (Magalhães et al., 2019(Magalhães et al., , 2020(Magalhães et al., , 2022)).Parallel to education efforts, inspection by environmental authorities is elemental to prevent

Figure 2 .
Figure 2. Map showing the Lopes Creek, municipality of Vieiras, Minas Gerais State, Brazil, where feral populations of Poecilia reticulata and Xiphophorus maculatus were monitored.In the satellite image, the section of the Lopes Creek (yellow) is surrounded by more than 40 production ponds stocked with many varieties of guppies and platies.

Figure 4 .
Figure 4. Ornamental and wild-type phenotypes of Poecilia reticulata and Xiphophorus maculatus reared in an ornamental fish farm bordering Lopes Creek and collected in the aforementioned creek, Minas Gerais State, Brazil.(A, B) Ornamental-type female/male of P. reticulata (red mosaic round-tail variety), (C, D) ornamental-type males of X. maculatus (yellow/red Mickey Mouse varieties), (E, F) wild-type female/male of P. reticulata, (G, H) wild-type females of X. maculatus.Scale bar: 1 cm.

Table 1 .
Food items found in stomachs of feral poeciliids (values represent the feeding index, % IAi) captured in Lopes Creek, Minas Gerais State, Brazil.

Table 3 .
Bimonthly and total sex ratio of feral poeciliids collected in Lopes Creek, between January and December 2017, Minas Gerais State, Brazil.