Feeding is a dynamic process and diet can change based on shifts in ecologic conditions. These changes can affect growth, survival, and demographics of assemblages, and therefore affect the transfer of energy across the systems (Brodeur et al., 2017). Therefore, understanding changes in the trophic structure of fish assemblages caused by human-induced alterations can help to better understand such impacts and aid in the development of conservation and management strategies for freshwater environments (Pompeu, Godinho, 2003; Braga et al., 2012; Lima et al., 2020; Loch et al., 2020). Since a large number of dams are under construction or planned in the near future (Zarfl et al., 2019), expanding hydropower is one of the main threats to fluvial ecosystems (Reid et al., 2019). Damming alters the trophic ecology of rivers by causing fragmentation, changing the natural flow regime, and creating new reservoirs.

Fragmentation affects the fundamental processes and functions of healthy rivers, altering the natural flow of energy and nutrients (Dudgeon et al., 2006; Nestler et al., 2012). Impoundments also affect the natural flow regime, which is a key driver of river ecology (Bunn, Arthington, 2002) by determining channel morphology, habitat diversity, and substrate stability (Power et al., 1995; Nilsson, Svedmark, 2002). Alterations in natural flow regime change the availability of basal resources (detritus and phytoplankton) that provide energy to support food webs (Angradi, 1994), eventually compromising the quality of food resources downstream (Baxter, 1977; Abujanra et al., 2009). Reservoir creation causes a drastic transformation from a lotic to lentic habitat, consequently changing the physical, chemical, and geomorphological characteristics of the area (Agostinho et al., 2007).

These modifications influence the distribution, abundance, and diversity of organisms within these ecosystems (Poff, Zimmerman, 2010) by selecting species with functional traits related to the new conditions (Agostinho et al., 2016). Rheophilic and migratory species are more likely to be negatively affected since they are often associated with rapidly flowing water and depend on the natural flow regime to complete their life cycles (Arantes et al., 2019). The changing conditions also affects the different feeding guilds in distinct ways.

Shortly after filling, the reservoir undergoes a trophic upsurge period, consisting of a high nutrient input resulting from the decomposition of inundated vegetation and increased primary production (Kimmel et al., 1990; Agostinho et al., 2016). Detritivores, omnivores, herbivores, and insectivores increase in abundance exploring these resources, which also leads to increased prey availability for piscivores (Agostinho et al., 2016; Arantes et al., 2019). Over time, the reservoir progresses to a phase of trophic depletion marked by low productivity. At this point, fish abundance in the reservoir tends to decrease as species adjust to the new environment (Agostinho et al., 2016). Diet and trophic structure shifts due to reservoir creation have also been frequently documented in the literature (Albrecht, Pellegrini-Caramaschi, 2003; Luz-Agostinho et al., 2006; Delariva et al., 2013; Pereira et al., 2016).

Although the reservoir area experiences the most pronounced changes after dam construction (Baxter, 1977), the trophic ecology upstream of the reservoir (Dias et al., 2020) and downstream of the dam (Power et al., 1996) can also be affected. Increases in piscivorous, decreases in detritivorous biomass (Mérona et al., 2001; Peressin et al., 2016), and an overall dominance of omnivores and generalist feeders (Gandini et al., 2012; Sá-Oliveira et al., 2015; Peressin et al., 2016) have been observed downstream of dams after construction. Upstream of the reservoir, although these stretches of river tend to be considered control areas in relation to changes caused by the dam, changes have also been reported, related to populations that could migrate from the reservoir (Dias et al., 2020).

Studies aiming to analyse and compare trophic responses of fish assemblages, both upstream and downstream dam construction, are scarce (e.g., Sá-Oliveira et al., 2015; Monaghan et al., 2020). Therefore, the aim of our study is to compare between the fish assemblage trophic structures located upstream of the reservoir and downstream of the dam, both before and after the construction of the highest hydropower plant in Brazil. We expected downstream changes to be more pronounced due to flow regulation effects, and predicted that such changes will be observed both in species substitution and in changes to their relative biomass.

Study area. The Jequitinhonha River basin drains an area of approximately 70,315 km² across the states of Minas Gerais and Bahia, Brazil. Mean annual rainfall averages around 600 to 1,600 mm and are irregularly distributed throughout the year, although rainfall mostly occurs between October and March (IBGE, 1997). The Jequitinhonha River rises into the Espinhaço mountain range and runs 1,086 km northeast into the Atlantic Ocean (Guerrero, 2009).

The Irapé Hydroelectric Plant started its operation in 2006, and is the tallest dam in Brazil with 205 m high (Cachapuz, 2006). The reservoir has an area of 142.95 km², an installed power capacity of 360MW, and holds up to 595,488 m³ of water (Cachapuz, 2006). Four sampling points were placed: two in lotic stretches upstream and two downstream of the Irapé Hydroelectric Power Plant (16°44’S 42°34’W) on the Jequitinhonha River (Fig. 1). Upstream of the reservoir, the hydrological patterns have remained the same after the hydropower construction, while downstream of the dam, the effects of the power plant on the natural flow regime were evident (Fig. 2).

FIGURE 1

FIGURE 2

Data sampling. Fish sampling occurred biannually both prior to the construction of the dam from 2003 to 2005, and after construction from 2011 to 2017 (Tab. 1), comprising both rainy and dry seasons. For fish collection, we used gillnets with mesh sizes 2.4, 3, 4, 5, 6, 7, 8, 10, 12, 14 and 16 cm between opposite knots. All net lengths were 30 m, except for the 2.4 mesh net which was 20 m long. All nets were exposed for 12 h overnight. Fish were captured exclusively using gillnets. Almost all individuals were dead when nets were removed the next day, so euthanasia was unnecessary. We fixed the collected individuals in formaldehyde and preserved them in alcohol in the laboratory. Fish were identified with comparisons to collections in the Museu de Ciências Naturais (MCNIP) at the Pontifícia Universidade Católica de Minas Gerais. Specimens were deposited at the Ichthyological Collection of MCNIP under the catalog numbers: MCNIP 0348, 0333, 0340, 0346, 0050, 0347, 0049, 0336, 0184, and 0051.

TABLE 1

Period	Sampling point	Location	Number of samplings
Before	1	17º18'44.8"S 43º12'14.9"W	6
	2	16º35'56.4"S 42º50'07.9"W	5
	4	16°37'30.04"S 42°15'54.58"W	6
After	1	17º18'44.8"S 43º12'14.9"W	7
	2	16º35'56.4"S 42º50'07.9"W	14
	3	16º44'20.6"S 42º34'02.3"W	27
	4	16°37'30.04"S 42°15'54.58"W	11

In the laboratory, fish were measured, weighted, and had their stomachs removed for subsequent analysis. Under a stereomicroscope, stomach contents were analysed to the lowest possible taxonomic level. Identification of food items was made with the aid of a taxonomic bibliography (Ward, Whipple, 1918; Costa et al., 2006; Mugnai et al., 2010). The wet weight of the items was determined using a precision scale (Ohaus Adventurer, AR 2140 - 0,0001g).

Data Analysis. We classified the species into trophic guilds through a bibliographic review (S1) and stomach content analysis. For species without dietary descriptions in the literature, we analysed up to 10 stomachs. To determine the importance of each food item for each species, the food index (IAi) (Kawakami, Vazzoler, 1980) was calculated, but the volume of the items was replaced by their weight (Bennemann et al., 2006) as shown in the following formula:

I A i = F i x P i ∑ i = 1 n F i x P i × 100

where IAi = Alimentary Index, Fi = Frequency of occurrence (%) of food item i, and Pi = Weight (%) of food item i. Based on the IAi, the consumption of ≥ 50% (Bennemann et al., 2011) of the same food item was used to classify the species in each guild (detritivore, herbivore, omnivore, piscivore, and insectivore). Species that consumed food items from animal and plant sources on similar proportions were classified as omnivores. When data of bibliographic review and IAi were available, we used whichever classification was based on a higher number of individuals. To determine trophic structure, we summed the biomass of the species individuals belonging to each guild.

To evaluate whether trophic structure changed with site (upstream and downstream) or with dam construction, the fish assemblages were compared considering four groups according to the location (upstream or downstream) and time (before or after) relative to the dam construction. Data was standardized by the total biomass in each sample. These groups were sorted by non-metric multidimensional scaling analysis (NMDS) (Legendre, Legendre, 2012) using the Bray-Curtis similarity index and compared by Analysis of Similarities (ANOSIM) (Clarke, 1993). The similarity percentage analysis (SIMPER) (Clarke, 1993) was used to evaluate which guild contributed most to the differentiation of these groups.

To assess the effects of location and damming on species composition and occurrence, we sorted groups of species by two NMDS analysis: one based on the biomass of each species using the Bray-Curtis similarity index and the other based on the presence/absence of species using Jaccard similarity index. Then, we conducted ANOSIM between pairs of groups by location (before upstream x before downstream and after upstream x after downstream) and period (before upstream x after upstream and before downstream x after downstream) to determine differences between those groups. The similarity percentage analysis was used to evaluate which species contributed most to the differentiation of these groups. All ANOSIM analyses were conducted with 999 permutations. NMDS, ANOSIM, and SIMPER tests were performed in Primer 6 using the PERMANOVA+ package (Clarke, Gorley, 2006).

Trophic structure. A total of 4,055 individuals belonging to 31 species were sampled. Sixteen species were recorded both before and after dam closure, while three species were captured only before dam construction and 12 only after. Regarding the feeding habits, 10 species were classified as detritivores, 2 as herbivores, 3 as insectivores, 10 as omnivores, and 6 as piscivores (S1, S2).

Before the construction of the dam, there was no difference in the trophic structure between the river segments upstream and downstream. After the construction of the dam, however, the two regions differed from the previous state, and they also became different from each other (Fig. 3; Tab. 2).

FIGURE 3

TABLE 2

Groups	Trophic Structure		Species composition
			Species substitution		Relative biomass
	R statistical	p value	R statistical	p value	R statistical	p value
BU, AU	0.483	0.001	0.450	0.001	0.378	0.002
BD, AD	0.615	0.001	0.482	0.004	0.547	0.001
AU, AD	0.477	0.001	0.424	0.001	0.39	0.001
BU, BD	0.139	0.102	0.013	0.402	0.077	0.214

After the dam implementation, the area sampled upstream experienced an increase in biomass of detritivores and a decrease of piscivores. Downstream, on the other hand, showed an exactly opposite response for these guilds, presenting an increase of omnivore biomass and causing a differentiation between the two assemblages (Fig. 3A; Tab. 3). Due to the sharp variations in biomass downstream after the dam construction, we performed the analysis without site 3, which had the higher number of samples, to assess its impact on fish biomass. Analyses without site 3 followed the same tendencies of when it was included (S3).

TABLE 3

Trophic Guilds	Groups / Average biomass				Contribution %	Cumulative %
	Before Upstream	Before Downstream	After Upstream	After Downstream
BU x AU Average dissimilarity = 53.39
Detritivore	30.67		59.97		34.56	34.56
Piscivore	35.9		10.37		29.82	64.38
Omnivore	20.53		14.77		18.14	82.52
Herbivore	7.8		10.7		11.02	93.54
BD x AD Average dissimilarity = 71.33
Detritivore		60.93		13.9	36.38	36.38
Piscivore		4.21		40.86	25.83	62.21
Omnivore		11.8		32.62	19.3	81.51
Insectivore		16.43		4.42	12.47	93.98
AU x AD Average dissimilarity = 60.50
Detritivore			59.97	13.9	39.38	39.38
Piscivore			10.37	40.86	27.57	66.95
Omnivore			14.77	32.62	20.33	87.29
Herbivore			10.7	8.2	7.36	94.65

Species composition. The observed trophic structure changes occurred due to both species composition and changes in relative biomass (Fig. 4; Tab. 2). After dam construction in the upstream region, the migratory detritivorous Prochilodus hartii Steindachner, 1875, increased its biomass while the sedentary predators from genus Hoplias Gill, 1903, had lower captures. Downstream of the dam, Prochilodus hartii decreased in biomass, while the omnivorous Wertheimeria maculata Steindachner, 1877, and native (Hoplias spp.) and the non-native (Serrasalmus brandtii Lütken, 1875) piscivorous increased (Tab. 4).

FIGURE 4

TABLE 4

Species	Groups / Average biomass				Contribution%	Cumulative %
	Before Upstream	Before Downstream	After Upstream	After Downstream
BU x AU Av.Diss = 73.09
Prochilodus hartii	17.92		35.14		19.19	19.19
Hoplias malabaricus	13.54		4.43		10.62	29.8
Hoplias brasiliensis	12.39		4.48		9.42	39.22
Hypostomus sp.	12.68		12.85		8.73	47.94
Megaleporinus garmani	7.22		9.86		7.76	55.7
Hypomasticus steindachneri	10.68		1.18		7.67	63.37
Delturus brevis	0		8.14		5.57	68.94
Psalidodon cf. fasciatus	0.19		7.45		4.99	73.93
Pseudoplatystoma sp.	6.99		0		4.78	78.71
Megaleporinus elongatus	5.11		3.56		4.53	83.25
Wertheimeria maculata	4.31		0.49		3.21	86.45
Trachelyopterus galeatus	3.3		2.04		2.93	89.38
Pareiorhaphis sp.	0		3.24		2.21	91.59
BD x AD Av.Diss = 87.12
Prochilodus hartii		29.36		7.73	16.87	16.87
Wertheimeria maculata		0.93		16.68	9.43	26.3
Gymnotus carapo		16.43		0.06	9.43	35.73
Steindachnerina elegans		15.84		0.17	9.06	44.79
Hypostomus sp.		15.73		1.7	8.53	53.32
Hoplias brasiliensis		1.13		14.22	8.1	61.42
Serrasalmus brandtii		0		12.76	7.32	68.74
Hoplias malabaricus		2.91		11.82	6.51	75.26
Trachelyopterus galeatus		8.92		4.8	5.62	80.87
Astyanax brevirhinus		5.67		4.14	3.78	84.66
Hypomasticus steindachneri		1.34		4.42	2.8	87.46
Megaleporinus elongatus		0		4.35	2.5	89.96
Megaleporinus garmani		0.96		4.06	2.38	92.34
AU x AD Av.Diss = 80.47
Prochilodus hartii			35.14	7.73	18.8	18.8
Wertheimeria maculata			0.49	16.68	10.35	29.15
Hoplias brasiliensis			4.48	14.22	9.26	38.41
Serrasalmus brandtii			0.78	12.76	7.99	46.41
Hoplias malabaricus			4.43	11.82	7.83	54.24
Hypostomus sp.			12.85	1.7	7.53	61.77
Megaleporinus garmani			9.86	4.06	5.12	66.89
Delturus brevis			8.14	0	5.06	71.94
Psalidodon cf. fasciatus			7.45	3.83	4.43	76.38
Megaleporinus elongatus			3.56	4.35	3.79	80.17
Trachelyopterus galeatus			2.04	4.8	3.27	83.43
Hypomasticus steindachneri			1.18	4.42	3.04	86.47
Astyanax brevirhinus			0.84	4.14	2.52	89
Pareiorhaphis sp.			3.24	0	2.01	91.01

Before the dam construction, trophic structure and species composition were similar between the regions upstream and downstream of the future hydropower plant. After the dam construction, however, both aspects of the assemblages changed. The two regions became distinct and changed when compared to the previous conditions. Downstream of the dam, three trophic groups were affected: piscivores and omnivores increased in biomass, and detritivores decreased. Upstream of the reservoir, detritivores increased their biomass while piscivores decreased. It is important to highlight that this study focused only on the lotic areas upstream the reservoir and downstream of the dam, and did not evaluate intraspecific changes in trophic guilds by comparisons of stomach content analyses before and after the impoundment. We focused on assessing the changes in fish assemblage structure regarding trophic groups.

The observed changes in trophic structure and species composition downstream of the Irapé dam can be linked to its impacts as a barrier, on the water quality, and on the natural flow regime. The blockage of migration routes can lead to the accumulation of fish downstream of dams (Mérona et al., 2001), increasing the availability of fish as a food resource (Mérona et al., 2001; Delariva et al., 2013; Peressin et al., 2016). Alterations in water level and transparency can also make fish more vulnerable to predation (Granzotti et al., 2018), favoring visual predators. We observed an increase in the relative biomass of piscivores following the construction, represented by Hoplias spp., and Serrasalmus brandtii, an invasive species to the Jequitinhonha River basin (Andrade et al., 2018). Besides belonging to the same feeding guild, these species use different strategies to obtain food, possibly avoiding the overlap of resource consumption. Serrasalmus brandtii presents mutilating piscivorous habits, feeding on fish scales, fins, and flesh (Pompeu, 1999; Alvim, Peret, 2004; Andrade et al., 2018) whereas Hoplias malabaricus (Bloch, 1794) is a specialist piscivore (Hahn, Fugi, 2008), swallowing whole prey (Peretti, Andrian, 2008). These species are sedentary and are capable of spawning multiple times throughout the year, and therefore do not depend on the natural flow regime to complete their life cycle (Prado et al., 2006; Honorato-Sampaio et al., 2009).

Impacts imposed by dams on downstream areas extend to other biological communities as well, such as macroinvertebrates. Changes in their abundance and density have been related to both flow regulation (Lobera et al., 2017) and daily flow peaking (Tupinambás et al., 2014), with associated responses in fish diet (Gandini et al., 2014). Therefore, an increase of omnivores is expected, related to their generalist feeding habits by preying on a broad spectrum of food resources (Gerking, 1994) and using resources according to their availability in the river. High levels of omnivory have been reported for communities subject to environmental disturbances (Wootton, 2017), and the same pattern was found on post-impoundment stretches on the Paraná River that presented increased numbers of fish with generalist feeding habits (Oliveira et al., 2018). Downstream of the Irapé dam, the relative biomass of Wertheimeria maculata, which is an opportunistic feeder (Vono, Birindelli, 2007), showed a noteworthy increase.

Changes on the natural flow regime can negatively affect detritivore species (Abujanra et al., 2009), as the input of allochthonous resources downstream also decreases with flow regulation (Mor et al., 2018). Moreover, dams trap sediments and nutrients in the reservoir (Baxter, 1977; Roberto et al., 2009), decreasing turbidity and nutrient load and therefore modifying water quality downstream (Agostinho et al., 2008). The decrease of detritivore relative biomass downstream of the dam observed in our study was also found downstream of the Tucuruí Dam (Mérona et al., 2001), and detritus was found to be the lesser consumed food item on the river downstream of 12 hydroelectric power plants (Peressin et al., 2016). In our study, Prochilodus hartii, a detritivorous migratory species, showed reduced biomass downstream of the dam. The species have terminal mouth and protractible lips, which allow them to obtain detritus from the bottom and other substrates. Their diet is composed of finely divided detritus and mud (Fugi et al., 2001), and therefore can be highly impacted by the oligotrophication caused by the dam. The Irapé reservoir has a water residency time of 457 days (Rodrigues, 2009), and longer water residence times increase the oligotrophication process by increasing nutrient and sediment retention (Santos et al., 2020). Impoundments also affect the reproductive aspects of migratory species (Agostinho et al., 2004); on the Jequitinhonha River, the migration blockage has had a negative effect on the populations of P. hartii (Abdo et al., 2018), and it can also be pointed out as one of the factors for its decrease in population numbers.

Upstream of the reservoir, changes in the trophic structure were not expected due to changes in resource availability, but could be explained by interferences in migratory fish movements caused by the hydropower plant (Agostinho et al., 2016; Wu et al., 2019), or from fish displacements from the reservoir (Dias et al., 2020; Monaghan et al., 2020). Since the post-dam data was collected five years after the reservoir was filled, the first explanation seems to be the most plausible. The most remarkable change was an increase in biomass of P. hartti. If the breeding sites of P. hartii are located upstream, the barrier represented by the reservoir (Pelicice et al., 2015) could be confining most of the population to the upstream region. Because they are able to complete their life cycle upstream of the reservoir, the lower portion of those lotic segments may have been used as the main feeding site available for the species.

The construction of the Irapé Hydropower Plant has led to changes in the trophic structure of lotic river stretches, both upstream of the reservoir and downstream of the dam. Therefore, impoundments will cause drastic changes, not only in the reservoir but also in the adjacent lotic remnants that were not affected by the transition from lotic to lentic regimes. Species composition and biomass changed in both upstream and downstream river stretches, leading to changes in trophic structure. As such, we emphasize that upstream lotic areas must be used with caution as control areas to evaluate changes in downstream/reservoir areas since they may not represent the assemblages before the impact occurred. Understanding the most affected trophic groups along the lotic stretches adjacent to dams is useful to determine conservation strategies for fish fauna. Downstream of dams, the use of artificial seasonal flows to emulate natural variation and reduce the effects of flow regulation downstream could constitute an alternative to minimize some of the observed changes. Upstream of the reservoir, the maintenance of long free flowing river stretches is possibly the best alternative to maintain the fish fauna trophic structure near the original one.