Cypriniform fish in running waters reduce hyporheic oxygen depletion in a eutrophic river

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2020 The Authors. Freshwater Biology published by John Wiley & Sons Ltd Dirk Hübner and Madlen Gerke should be considered joint first author. 1Bürogemeinschaft für fischund gewässerökologische Studien, Marburg, Germany 2Institute for Integrated Natural Sciences, University of Koblenz-Landau, Koblenz, Germany 3Department of Ecology, Animal Ecology, University of Marburg, Marburg, Germany 4Bürogemeinschaft für fischund gewässerökologische Studien, Frankfurt, Germany


| INTRODUC TI ON
The hyporheic zone is the connecting ecotone between river and groundwater ecosystems (Brunke, 1999). An intact hyporheic zone provides key ecological functions: it acts as a filter that mediates the exchange of water, nutrients, organic matter, and contaminants; it plays a crucial role in biogeochemical cycling and buffers against physical and chemical influences (Brunke & Gonser, 1997;Findlay, 1995;Stanford & Ward, 1988). It also serves as an important habitat and refuge for benthic invertebrates and the early developmental stages of gravel-spawning fish (Baxter & Hauer, 2000;Brunke & Gonser, 1997;Findlay, 1995;Williams & Hynes, 1974). Due to the importance of the hyporheic zone for ecosystem functioning and biodiversity, efforts aimed at stream restoration must also include restoration of the hyporheic zone (Boulton, 2007;Hester & Gooseff, 2010).
In standing waters, the enhancement of zooplankton grazing by biomanipulation is a well-established technique to control eutrophication effects (Benndorf, 1990;Hansson et al., 1998;Shapiro & Wright, 1984). However, whether the active enhancement of grazing might be similarly used to control the effects of eutrophication in running waters is unknown. Given the potentially strong impact of the top-down control of periphyton by grazing in rivers, we expected that enhancing stocks of herbivorous and omnivorous fish would lead to increased indirect control of the eutrophication effects in rivers, especially those in which a drastic reduction in nutrient input and the restoration of flood dynamics are not feasible.
Indirect top-down effects cascading through three trophic levels, from zoobenthivorous fish to invertebrate grazers and then to periphyton, have been demonstrated in several small-scale experiments (e.g. Dahl, 1998;Gerke et al., 2018;Pagnucco, Remmal, & Ricciardi, 2016) and under near-natural conditions in stream ecosystems (Winkelmann et al., 2014). In two mesocosm experiments conducted in a Northern California river, predatory and omnivorous fish were shown to induce a four-level trophic cascade down to periphyton, by releasing grazing invertebrates from predation pressure (Power, 1990;Wootton & Power, 1993). In Central European rivers, one of the most widespread, common and at least partially piscivorous fish species is the large omnivorous cypriniform European chub Squalius cephalus (Linnaeus, 1758) (Leuciscidae: Leuciscinae).
While it partially feeds on periphyton, especially on filamentous algae (Balestrieri, Prigioni, Remonti, Sgrosso, & Priore, 2006;Hellawell, 1971), and benthic invertebrates, with progressing age and size it increasingly preys on small zoobenthivorous fish such as minnows, Phoxinus phoxinus (Linnaeus, 1758) and bullheads, Cottus gobio (Linnaeus, 1758) (Hellawell, 1971;Mann, 1976). Thus, we predicted that, in addition to its potential direct top-down effects on periphyton by feeding on filamentous algae, large chub would indirectly reduce periphyton biomass, and therefore eutrophication effects as well, via a four-level trophic cascade. Specifically, in response to a reduction in the local density of zoobenthivorous fish through either chub predation or displacement, benthic invertebrates would be released from predation pressure, thus increasing invertebrate grazing and reducing periphyton biomass.
To assess whether the top-down effects of herbivorous nase and omnivorous chub lead to a reduction in eutrophication effects in the hyporheic zone, we conducted a mesocosm-based field experiment using cage enclosures in a eutrophic river. The top-down control of periphyton by nase and chub was expected to reduce biological clogging, ultimately resulting in an increased oxygen supply in the hyporheic zone. Accordingly, we hypothesised that enclosures stocked with either nase or chub would show (1) higher oxygen concentrations within the hyporheic zone and (2) a lower periphyton biomass than in enclosures without fish stocking. Along with a reduction of periphyton biomass, we expected grazing-induced changes in the structure of the periphyton community, especially regarding the proportion of green algae within the periphyton. The feeding activity of nase was expected to directly reduce periphyton biomass and that of omnivorous chub to directly and/or indirectly reduce periphyton biomass via the above-described four-level trophic cascade.

| Experimental site
The experiment was conducted in the hyporhithral zone of the river Nister (Rhineland-Palatinate, Germany, 50°43′24 N, 7°44′24 E), a small gravel-bed river with a catchment area of 246 km 2 . The average mean discharge is 6.3 m 3 /s in winter and 2.4 m 3 /s in summer (measured at Heimborn, ID 2724030100; data supplied by the State Office for Environment of Rhineland-Palatinate). The land use type in the catchment area is dominated by forestry, pasture and agriculture. At the experimental site, the Nister flows from east to west and the land use type is forest on its right bank (north) and pasture on its left bank (south). Therefore, the river is only partly shaded, on the right bank side. The river's width here averages 15 m.
The river bed is relatively homogenous and mainly consists of cobbles (6.3-20 cm). Due to phosphate emissions in the catchment area from several minor municipal wastewater treatment plants and diffuse emissions from local agriculture, the river is highly eutrophic, with the effects including oxygen oversaturation and an extremely high pH during the daytime, especially during the spring algal bloom (Gerke et al., 2018). The hyporheic zone in the river is impaired by clogging especially during summer, as indicated by high proportions of fine sediments in the substrate and low hyporheic oxygen concentrations (unpublished data, see Appendix S1 for details). Moreover, high differences of electrical conductivity and pH-values between surface water and hyporheic water suggest reduced subsurfacesurface exchange (Geist & Auerswald, 2007).
The benthic algal and cyanobacterial community in the river is dominated by adnate and loosely attached diatoms, but filamentous cyanobacteria (e.g. Homoeothrix sp.) and filamentous green algae (Ulothrix sp., Cladophora sp.) can become dominant during late spring and summer. The experiment described herein was conducted in early summer, after both the spring peak of the algal bloom and the breakdown of periphyton biomass. The invertebrate community is dominated by chironomid larvae and scraping grazers, such as the mayfly Ephemerella ignita (Poda, 1761) and the snail Ancylus fluviatilis (Müller, 1774).
The obligate herbivorous common nase dominates the community of large fish, followed by the omnivorous European chub (estimated local biomass densities of fish >15 cm assessed in a 550 m river stretch directly upstream of the experimental site: common nase: 75 g/m 2 , European chub: 18 g/m 2 ). On a mesohabitat scale, much higher natural fish densities can be expected in run sections, as they are the preferred feeding habitats of nase (Huber & Kirchhofer, 1998). In contrast, due to high predation risk in the river from cormorants, Phalacrocorax carbo sinensis (Linnaeus, 1758), shallow riffles are usually avoided by large fish. Among the small zoobenthivorous fish, bullhead, common minnow, and stone loach, Barbatula barbatula (Linnaeus, 1758), occur in high densities (Gerke et al., 2018).

| Experimental setup
The enclosure experiment was conducted using in situ fish cages (8 m 2 ) and three treatment groups: (1) enclosures stocked with common nase; (2) enclosures stocked with European chub; and (3) control enclosures without fish stocking. The fish biomass density (mean ± standard deviation) was 308.6 ± 0.3 g/m 2 for enclosures stocked with nase and 312.5 ± 1.7 g/m 2 for enclosures stocked with chub, achieved by stocking each enclosure with six to nine nase individuals (total length of 28.0-34.5 cm, mean total length 30.8 cm) or four chub individuals (total length of 28.5-43.0 cm, mean total length 37.3 cm). The fish used for the experiment were caught in the vicinity of the cage locations by electrofishing (using EFGI 650, Bretschneider Spezialelektronik). Electrofishing was approved by the fisheries department of the local environmental agency, the SGD Nord (Rhineland-Palatinate, Germany).

| Sampling
The concentration of hyporheic dissolved oxygen was sampled weekly beginning 1 week prior to the start of the experiment (directly before the installation of the enclosures on 31 May 2017) and continuing until the end of the experiment (5 July 2017). The multilevel probes (Lenk et al., 1999)  Sampling was performed using a Surber sampler (0.08 m 2 , mesh size 500 µm) positioned at the upstream end of each enclosure. The invertebrate samples were rinsed over a 500 µm-sieve and stored in 70% ethanol.

| Laboratory analyses
Periphyton biomass was characterised based on the total periphyton biomass (estimated as ash-free dry mass, in mg AFDM/cm 2 ), as a measure of total organic matter, and the autotrophic periphyton biomass (estimated as the chlorophyll a concentration in μg Chl a/cm 2 ), as a measure of the mass of photosynthetically active algae. Both were quantified based on the total volume of the obtained periphyton suspension and the sampled area of the stones. The surface area of the stones was estimated by carefully wrapping each stone in aluminium foil; overlapping areas were cut off, and the foil was then weighed. After their total volumes were determined, the periphyton suspensions were homogenised using a magnetic stirrer to ensure comparable aliquots.
Ash-free dry mass was quantified by transferring 10-ml aliquots to pre-weighed ceramic crucibles and then drying the samples at  The supernatants were discarded, and the pellets stored at − 80°C.
Chlorophyll a was extracted and then analysed spectrophotometrically according to Mewes, Spielvogel, and Winkelmann (2017). In short, four 3-mm glass beads and 1.8 ml of 96% ethanol buffered with 1 g MgCO 3 /L was added to each pellet and the mixtures were then homogenised using a mixing mill (MM 400, Retsch Technology GmbH). Chl a was extracted for at least 3 hr at room temperature in the dark. Subsequently, the samples were centrifuged for 3 min at  (Baumgärtner & Rothhaupt, 2003;Benke, Huryn, Smock, & Wallace, 1999;Burgherr & Meyer, 1997;Gerke et al., 2018;Mährlein, Pätzig, Brauns, & Dolman, 2016;Meyer, 1989;Smock, 1980). To calculate the total biomass of invertebrate grazers in each sample, the biomass of each herbivorous and omnivorous species was weighted according to the average proportion of plant food in the diet of that species (Schmedje & Colling, 1996).

| Data analyses
To assess the effects of stocking the enclosures with nase or chub on the extent of biological clogging, we tested the effects of the fish treatments on hyporheic oxygen concentrations. The effect on periphyton biomass of stocking the enclosures with nase or chub was examined by measuring the effects of the fish treatments on AFDM and the Chl a concentration. Whether the fish treatments changed the proportion of green algae within periphyton was also determined. Potential effects of the fish treatments on invertebrate grazing were investigated by measuring the effects of the fish treatments on overall invertebrate biomass and specifically on invertebrate grazer biomass. In all cases, with the exception of the effect of chub on benthic invertebrates, the effects of the fish treatments were tested against the control treatment without fish stocking. The effect of chub on invertebrate biomass (and specifically on grazer biomass) was tested against the other two treatments combined by applying orthogonal contrasts (contrast I: chub versus nase and control; contrast II: nase versus control). In general, all univariate statistical analyses were calculated by applying a generalised linear mixed model (GLMM) in which the treatment was a fixed factor and the experimental block a random factor, in order to account for differences in the localities of the enclosures. In the analyses of the hyporheic oxygen concentration, the depth horizon was included as a fixed factor and the interaction treatment × depth was entered into the model. Parameters were estimated using the maximum likelihood method. Differences in invertebrate community composition were assessed using permutational multivariate analysis of variance (PERMANOVA) based on the Bray-Curtis dissimilarities calculated from the invertebrate biomass (fourth-root transformed).
A PERMANOVA was performed with 999 permutations stratified within the experimental blocks. For a detailed assessment of the potential effect of chub treatment on invertebrate community composition, in addition to the full model, the orthogonal contrasts chub versus nase and control (contrast I) and nase versus control (contrast II) were analysed as well.
Prior to the tests of the hypotheses, we examined whether the habitat quality at the different enclosure sites differed so as to cause a bias in the results, by testing for potential differences between the enclosures in the initial values (before installation of the enclosures and fish stocking) of the dependent variables used for hypothesis testing according to the future fish stocking treatment. The same statistical methods as described above were applied, except that, rather than the control treatment, the overall model was used as the testing framework. There were no significant effects of future treatment on the dependent variables prior to the experiment (Table B1 in Appendix S2). Consequently, the measurements obtained after termination of the experiment were used for hypothesis testing without a correction for the initial values.
Both the statistical analyses and graph plotting were performed using R version 3.6.1 (R Development Core Team, 2019). Generalised linear mixed models were applied using the lme function included in the R-package nmle (Pinheiro et al., 2018). PERMANOVA was performed using the adonis function from the R-package vegan (Oksanen et al., 2018).

| RE SULTS
In accordance with our first hypothesis, in which higher oxygen concentrations were predicted for fish-stocked enclosures within the hyporheic zone, at the end of the experiment the hyporheic oxygen concentration was significantly higher in enclosures stocked with either nase (p = 0.001) or chub (p = 0.02) than in control enclosures without fish stocking (GLMM, n = 4; Figure 2). Prior to the experiment, the hyporheic oxygen concentrations did not differ among the future treatments (p = 0.59, see Table B1 in Appendix S2 for details; F I G U R E 2 Mean oxygen concentration in the hyporheic zone (8, 13, and 23 cm depth) in enclosures stocked with nase or with chub and in control enclosures without fish stocking (n = 4). Left: before the installation of the enclosures (1 week prior to the start of the experiment), Right: at the end of the experiment. Error bars represent the standard deviations.

| D ISCUSS I ON
Biological clogging of the hyporheic zone is a severe consequence of eutrophication in running waters, as it causes oxygen depletion Ibisch et al., 2009), which in turn strongly reduces habitat quality in this important ecotone (Geist & Auerswald, 2007;Hübner et al., 2009;Malcolm et al., 2003). Our mesocosm experiment The positive effects of herbivorous nase and omnivorous chub on hyporheic oxygen availability were caused by different mechanisms. Initially, we expected that nase, as a specialised periphyton feeder (Corse et al., 2010;Freyhof, 1995), would directly reduce periphyton biomass accrual, thereby reducing biological clogging.
Our results support this expectation, as judged by the significantly lower total periphyton biomass (AFDM) measured at the end of the experiment in enclosures stocked with nase. Several other studies have similarly shown that herbivorous fish can reduce periphyton, both on a small scale (e.g. Flecker et al., 2002;Martin et al., 2016;Schneck et al., 2013;Veach et al., 2018;Wootton & Oemke, 1992) and on a mesohabitat scale (Gelwick & Matthews, 1992;Power et al., 1985Power et al., ,1989Stewart, 1987). However, compared to the mean hyporheic oxygen concentration, which was more than twice as high in enclosures stocked with nase than in those without fish stocking, the reduction in the mean AFDM was rather small. Thus, even a small reduction of periphyton biomass is sufficient to cause a large increase in hyporheic oxygen availability. Because hyporheic oxygen availability is mainly controlled by water transport and heterotrophic respiration rates (Brunke & Gonser, 1997), it might be affected by two different processes associated with biological clogging: (1) the reduced permeability of the hyporheic zone as a consequence of both external clogging by benthic algae (Battin & Sengschmitt, 1999;Ibisch et al., 2009) and internal clogging by infiltrated algal cells (Kloep & Röske, 2004); and (2) an increase in biological oxygen demand in the hyporheic zone due to the enhanced decomposition of decaying algae (Rode, Hartwig, Wagenschein, Kebede, & Borchardt, 2015). In this respect, the relatively large effect of herbivorous nase on hyporheic oxygen availability may have reflected the reduced input of organic matter into the hyporheic zone by fish grazing, resulting in a decreased biological oxygen demand but only a slight reduction in the extent of external clogging. TA B L E 2 Estimated proportion (mean ± standard deviation) of higher periphyton taxa contributing to the algal and cyanobacterial communities in the three treatments (n = 4) at the end of the experiment

Diatoms [%] Green algae [%] Cyanobacteria [%]
Nase 64 ± 9 17 ± 6 20 ± 10 Chub 63 ± 8 30 ± 7 7 ± 7 Control 57 ± 5 34 ± 9 9 ± 11 F I G U R E 4 Mean total invertebrate biomass and mean biomass of invertebrate grazers in enclosures stocked with nase or with chub and in control enclosures without fish stocking (n = 4) at the end of the experiment. Error bars represent the standard deviations. In contrast to total periphyton biomass, autotrophic periphyton biomass (measured as Chl a) was not lower in the enclosures stocked with nase. One explanation for this result is the masking of grazing effects on Chl a by compensatory growth due to high rates of algal productivity. The autotrophic growth rate at the experimental site was presumably high, due to high nutrient and light availability (Lamberti, Gregory, Ashkenas, Steinman, & McIntire, 1989;Sturt et al., 2011). Moreover, fish grazing might even have stimulated algal growth. Benthic grazing reduces competition for nutrients and light by removing the upper layers of algae, thereby offering optimal growth conditions for new algae (Lamberti & Resh, 1983;McCormick & Stevenson, 1989,1991. In this respect, as an indicator of periphyton biomass accumulation over time, AFDM may be more appropriate than Chl a because it includes live, dead, and senescent algae (Lamberti et al., 1989).
Support for a stimulation of algal growth by fish grazing comes from our finding of a significantly lower proportion of green algae, especially filamentous Cladophora, in the nase-stocked enclosures, which indicated grazing-induced changes in periphyton community structure. The lower proportion of filamentous green algae in the nase-stocked enclosures may have additionally contributed to reducing the extent of clogging, given the potential of the enhanced trapping of sediments in the filamentous branches of Cladophora (Berger et al., 2003;Dodds, 1991) to increase external clogging of the hyporheic zone.
Among the possible mechanisms underlying the increased oxygen availability in the chub-stocked enclosures was the ability of these fish to reduce biological clogging by exerting indirect topdown effects on periphyton. We expected that chub negatively affect the densities of small zoobenthivorous fish either by predation (Hellawell, 1971;Mann, 1976) or by displacement (Magurran, 1989 (Power, 1990;Wootton & Power, 1993). The absence of a similar cascade down to periphyton in our experiment can be explained by the fact that chub is an opportunistic feeder (Balestrieri et al., 2006) that also feeds on large invertebrate grazers such as mayfly larvae. These are generally highly susceptible to predation while grazing on the stone surface (Kohler & McPeek, 1989).
The observed positive effect of chub on the oxygen availability in the hyporheic zone can be best explained by the enhancement of bioturbation. Benthic foraging by chub might have disturbed the river bed, thereby increasing permeability in the upper layer of the hyporheic zone. This scenario seems likely because chub and other benthic-feeding fish can act as zoogeomorphic agents, by increasing the mobility of sediments and thereby causing substrate coarsening (Pledger, Rice, & Millett, 2016Statzner, Sagnes, Champagne, & Viboud, 2003).
On larger spatial scales, the top-down effects of herbivorous nase that lead to an increased oxygen availability in the hyporheic zone might be more important than those of chub, due to the shoaling behaviour of nase. Specifically, adult nase typically form large single-species shoals of a dozen up to several hundred individuals that move actively within defined home ranges (Huber & Kirchhofer, 1998;Lusk, 1967). Within the feeding habitats of their home range, the densities and consequently the top-down effects of nase are presumably similar to those observed in our mesocosm experiment. The effects of chub might be less intense at larger scales because biomass in the enclosures exceeded those found in the river. Nonetheless, natural assemblages of river fish accommodate several species and rheophilic cypriniform fish such as nase and chub are important species in the hyporhithral and epipotamal zones of European rivers (Aarts & Nienhuis, 2003). As in our experiment both nase and chub positively affected hyporheic oxygen availability, under natural conditions the sympatric occurrence of the two species is likely to have synergistic effects. Hence, protecting and enhancing the stocks of both herbivorous and omnivorous fish may contribute to a reduction of eutrophication effects and a restoration of the hyporheic zone in running waters.
Although the effects of nase and chub on hyporheic oxygen concentration were not strong enough to impede or even reverse clogging, these fish may play an important role in the river's local biodiversity. In Central European rivers, the period during which fish eggs and larvae of several fish species develop within the hyporheic zone (intragravel period) coincides with the spring algal peak. Among these species are the salmonid grayling, Thymallus thymallus (Linnaeus, 1758) and the cypriniforms common barbel, Barbus barbus (Linnaeus, 1758), and nase (Britton & Pegg, 2011;Hübner et al., 2009;Kamler, Keckeis, & Bauer-Nemeschkal, 1998).
A low oxygen supply due to the reduced permeability of spawning gravels has been identified as a critical factor compromising the survival of salmonid eggs and larvae (e.g. Malcolm et al., 2003;Soulsby, Youngson, Moir, & Malcolm, 2001;Suttle, Power, Levine, & McNeely, 2004), and the survival of grayling eggs and larvae is negatively affected by eutrophication (Hübner et al., 2009).
Lithophilic cypriniforms such as nase also require a permeable and well-oxygenated hyporheic zone for successful development (Duerregger et al., 2018;Keckeis et al., 1996;Nagel, Pander, Mueller, & Geist, 2019). High densities of large herbivorous and omnivorous fish are therefore likely to substantially improve habitat quality in the short intra-gravel period of spring spawning fish such as grayling, barbel, and nase.

ACK N OWLED G EM ENTS
We especially wish to thank Christian Sodemann and Manfred Fetthauer for their technical advice and valuable support during field work. We also thank Michael Götten, Theresa Graf, Tobias Ebelhäuser, Mark Huffsmith, and all the colleagues and students who helped with field sampling and laboratory analyses. We are also grateful to Wendy Ran for linguistic support. This work was financially supported by the Federal Office for Agriculture and Food of Germany (grant number 2813BM011).

CO N FLI C T S O F I NTE R E S T
The authors have no conflict of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data supporting the findings of this study are available in the owncloud repository (Hübner, Gerke, Fricke, Schneider, & Winkelmann, 2020; https://cloud.uni-koble nz-landau.de/s/yjPsJ Ap57m mQTZL).