Introduction

Adaptations and counter-adaptations create a tightly coupled coevolutionary feedback loop during host–parasite evolution1. As a consequence, dependent organisms such as parasites, commensals, and mutualists are at risk of coextinction when their associated partners decline, potentially triggering extinction cascades and rapid biodiversity loss2. Ongoing large-scale changes in biodiversity and community structure may prevent many parasitic transmission stages from encountering suitable hosts3. The abundance of one partner in a co-evolving system can be influenced by factors external to the system, such as interactions with other species. When habitat functionality deteriorates, community composition may become simplified4, disrupting the favourable-to-unfavourable host species ratio.

It has been demonstrated that local biodiversity can reduce infection risk across a range of freshwater host–parasite systems5, an effect known as the “dilution effect”6, which is also referred to in parasitology as the “decoy effect”7. This mechanism posits that multiple species may distract free-living parasite stages from suitable hosts, leading to unsuccessful transmission events involving so-called “dead-end” hosts. While the dilution effect is often associated with high biodiversity, it may also arise in species-poor communities if a dominant fish species acts as a dead-end host, thereby reducing transmission success and negatively affecting parasite fitness.

Freshwater mussels of the order Unionida (hereafter referred to as naiads) exhibit a specialized life cycle that includes an obligatory parasitic larval stage. Females deposit eggs into brood sacs located in modified gill tissue called marsupia, where fertilization occurs and the embryos develop into parasitic larvae known as glochidia. These are released into the water column and must attach to a suitable fish host to complete development into the pediveliger stage. Following transformation, the juvenile mussel detaches and begins its benthic life as a free-living individual8. The primary ecological role of fish hosts is thought to be the dispersal of mussel larvae over greater distances9. Because mussels exist as metapopulations, host-mediated dispersal enables range expansion and colonization of new habitats within river systems10.

Parasitism begins when glochidia are released into the water. Given that the fitness of naiads is entirely dependent on their fish hosts—without a reciprocal dependency—numerous adaptations have evolved to enhance host infection. In Unio crassus, the only known European species that actively attracts fish, females move toward the riverbank, immerse their inhalant siphon, draw in water, and then contract their valves to increase internal pressure. Upon opening the exhalant siphon, they expel jets of water containing glochidial conglutinates. These jets are directed onto the water surface and visually attract fish, which become infested while attempting to consume the apparent prey item11, including the larvae themselves12.

The number of glochidia released by adult mussels is generally consistent and size-dependent13, with individuals typically reaching an asymptotic size in a given habitat14. Glochidia are released in intermittent spurts—approximately every 91 s—for 3 to 6 h from mid-morning to mid-afternoon. During this period, the marsupium is emptied, releasing up to 1,127 glochidia, which can disperse over approximately 0.5 m² of the water surface11. Each spurt attracts small fish, typically fry (mean = 6.4, SD = 5.8, N = 32). Although the total number of glochidia in the marsupium is substantial (1.47 g in a 70 mm mussel; T. Zając, unpubl. data), their availability in the water column is limited in time. Glochidial conglutinates sink to the bottom, and their viability decreases significantly after 12 h11.

Importantly, mussels are adapted to infect specific fish species; glochidia cannot successfully metamorphose on all potential hosts15,16. Therefore, when glochidia encounter multiple fish species—including non-hosts—scramble competition may occur. This process, irrespective of whether fish actively consume glochidia or passively remove them via surface attachment, can lead to a strong dilution effect. The strength of this effect depends on the relative abundance of host versus decoy fish species.

In this study, we aim to test whether a dilution effect driven by decoy species occurs within the habitat of the threatened freshwater mussel Unio crassus, to assess the magnitude of this effect, and to identify the factors that influence it. We hypothesize that the infestation rate of the main fish host decreases in the presence of infested decoy species.

Results

Dynamics of fish community structure and Unio crassus abundance

In the studied river, the dominant species were Phoxinus phoxinus (54%), Gobio gobio (19%), and Barbatula barbatula (17%), which together accounted for 90% of all captured individuals—representing the majority of species available for infestation by Unio crassus. The remaining species comprised a marginal fraction of the catch (Table 1); however, 8 out of the 11 recorded species were capable of intercepting glochidia.

Table 1 Abundance and temporal changes in the studied species—fish, lamprey (*), and Unio crassus—in the Warkocz River. For fish and lamprey, values are expressed as the number of captured individuals and their percentage of the total catch; for U. crassus, values represent mean density derived from three cross-channel transects, each surveyed at three separate localities. Fish data are pooled across the two sampling years (2015 and 2016) and, in subsequent monitoring cycles, across all study plots. Species observed to intercept glochidia (regardless of whether successful metamorphosis occurred) are shown in bold.
Full size table

Infestation of Gobio gobio in comparison to Phoxinus phoxinus and Barbatula barbatula

Among the dominant fish species, 37.3% of P. phoxinus individuals (N = 1304) were found to be infested with glochidia. In contrast, infestation prevalence was only 9.8% in G. gobio (N = 461) and 1% in B. barbatula (N = 413). For the remaining fish species combined, the prevalence was 8.5%.

The mean intensity of infestation among infected P. phoxinus individuals was 3.5 glochidia per fish—twice as low as in G. gobio (6.8), and more than twice as high as in B. barbatula (1.5). The variation in infestation was much greater in G. gobio (SD = 11.26) than in P. phoxinus (SD = 3.20). Maximum infestation levels further highlight these differences: the highest number of glochidia on a single P. phoxinus was 23, whereas in G. gobio it reached 57. In comparison, only six glochidia were found across all B. barbatula specimens, indicating the negligible role of this species in glochidial attachment. Therefore, subsequent analyses focused exclusively on P. phoxinus and G. gobio.

When comparing the total number of glochidia found on both species, P. phoxinus accounted for 82.1% of all glochidia observed on fish, while G. gobio accounted for only 14.6% (Table 2). The maximum relative contribution of G. gobio to the glochidial pool occurred in 2015 at plot 3, where it carried 33.8% of all observed glochidia.

Table 2 Total number of glochidia found on individuals of each species per site and year, along with the percentage contribution to the total pool of attached glochidia, and observed range (min–max) of glochidia per fish.
Full size table

Gobio gobio as a dead-end host

On day five after artificial infestation, a total of 57 glochidia were observed attached to the fins, body, and opercular regions of ten G. gobio individuals. However, none of these larvae successfully metamorphosed into the pediveliger stage (the next juvenile phase) during the standard excystation period of 44 days. In contrast, the same protocol applied to ten P. phoxinus individuals yielded 389 attached glochidia and 249 viable, metamorphosed juveniles within the same period (Fig. 1A).

Fig. 1
figure 1

The role of G. gobio as a decoy species. (A) Results of the infestation experiment: mean number of glochidia attached to fish five days after artificial infestation and mean number of successfully metamorphosed juveniles recovered over 40 days for P. phoxinus and G. gobio. Whiskers show observed ranges. (B) Spatial co-occurrence of U. crassus and G. gobio in Poland. Each square represents a 10 × 10 km grid cell. White squares indicate sites where U. crassus was recorded; light-blue squares denote cells where U. crassus and G. gobio occurred within < 5 km along the river; dark-blue squares mark cells where records of both species were within < 200 m. (C) Number of surveys in which the seven most frequent fish species occurred sympatrically with U. crassus (matched within < 200 m). (D) Mean percentage of these seven species in the catch across the same set of surveys.

Full size image

Co-occurrence

In the Warkocz River we analysed 234 buffer zones (5 m segments), of which 66 contained no fish, to investigate whether the number of glochidia found on P. phoxinus within each buffer was affected by the number attached to G. gobio. Results (Table 3) indicate that an increase in glochidia attached to G. gobio was associated with a decrease in the number attached to P. phoxinus. Moreover, a significant interaction was found between the number of U. crassus mussels present and the number of glochidia attached to G. gobio.

Table 3 Generalized linear model (GLZ) of the relationship between the number of glochidia found on P. phoxinus within each 5 m buffer (Poisson distribution, log link) and the sum of glochidia on G. gobio, number of U. crassus individuals, site, and season.
Full size table

At a broader spatial scale, among 158 fish surveys conducted in 49 rivers with known U. crassus localities, G. gobio was recorded in 129 surveys (81.6%) within 5 km of mussel sites. When the distance threshold was reduced to 200 m, co-occurrence was detected in 47 out of 52 surveys (90.4%) (Fig. 1C). Considering the quantitative structure of fish assemblages, G. gobio accounted for 23.8% of individuals among the seven most abundant species at sites located within 5 km of U. crassus localities, and for 17.4% at sites within 200 m (Fig. 1D).

Discussion

In all Unionidae, fish infestation constitutes an obligatory parasitic phase in their ontogeny and is under strong selective pressure. This has driven the evolution of numerous behavioural adaptations, such as releasing glochidia in confined areas of still water11 or employing species-specific lures17 to direct the larval suspension toward attracted hosts. Consequently, any dilution of the glochidial concentration in the environment is likely to be detrimental to parasitic success.

Glochidia lack the ability to actively select hosts18. Instead, they behave like particulate matter temporarily suspended in water, with their concentration measurable in quantitative terms19. Under such conditions, infestation can be viewed as a continuous depletion of glochidia in the water column due to three mechanisms: (1) sinking to the bottom, (2) passive, random attachment to fish fins or gills, and (3) active consumption by fish20,21,22. A similar process occurs when glochidia settle on the bottom, where foraging fish reduce their density on the substrate.

In all these cases, the number of infested fish and the number of glochidia attached per individual should be proportional to the local glochidial density per unit of water volume or benthic area (e.g., within a buffer zone), and should correlate across species. In such a framework, the dilution effect is expected to be directly proportional to the relative abundance of dead-end hosts within a given spatial unit.

In the present data, P. phoxinus was not only the dominant species in terms of abundance, but its individuals also showed a higher prevalence of infestation compared to G. gobio, the second most abundant species. However, the mean number of glochidia per fish was higher in G. gobio, suggesting that this species may remove a disproportionately large share of the glochidial pool despite its lower relative abundance. In some cases, such as plot 3 in 2015, G. gobio comprised 26.8% of all fish but carried 33.8% of the total glochidia load, indicating a substantial potential for impact under favourable ecological conditions. These data indicate that infestation probability decreased over time, as the abundance of P. phoxinus—the primary host in the river—declined, while the abundance of the dead-end host G. gobio increased (Table 1).

Glochidial availability is also expected to depend on mussel density, as it serves as the larval source. If certain fish species co-occur with mussels at sites of high density, while others are more common at sites with sparse mussel populations, these differences should be reflected in glochidial prevalence and intensity. A necessary condition for the dilution effect to manifest is the spatial co-occurrence of both competent and dead-end hosts, implying the potential for scramble competition23.

To test this, we assumed that within a defined spatial unit (e.g., a buffer), a finite pool of glochidia—produced by a local assemblage of spurting females—was partitioned between P. phoxinus and G. gobio. Our results (Table 3) revealed not only that glochidial attachment to G. gobio reduced infestation of P. phoxinus within the same spatial unit, but also that this interaction was modulated by mussel density. Specifically, at high mussel densities (and therefore high glochidial concentrations), G. gobio had no measurable effect on the infestation of P. phoxinus. However, when mussel density within a buffer fell below 19 individuals, the impact of G. gobio became significant (Fig. 2A).

This finding is particularly relevant in the context of declining populations of U. crassus, where low mussel densities may render the species especially vulnerable to the dilution effect—posing an additional threat and potentially contributing to an Allee effect24. Recent monitoring data from Poland (2023–2024) confirms that in 34.4% of the 64 surveyed sites across 34 rivers, fewer than 19 individuals were found per 5 m buffer (Zając K., unpubl. data). The decline of U. crassus was also confirmed in the Warkocz River in later years (Table 1), including evidence of increased mortality and reduced glochidia production25. This has direct implications for conservation planning: extensive low-density areas may contribute little to species demography compared to highly concentrated populations.

The observed interaction can be interpreted by considering the microhabitat preferences of U. crassus, which favours still water at channel margins and fine sediments14,26,27—conditions indicative of reduced flow. Glochidia expelled near the surface sink within seconds to minutes (unpubl. data), depending on water depth, while their overall lifespan ranges from 1 to 3 days, depending on temperature11. This implies that glochidia are accessible to pelagic fish for only a brief period but remain available to benthic fish for a significantly longer time.

P. phoxinus is known to forage in both the water column and on the substrate28,29, whereas G. gobio, a benthivorous species with an inferior mouth and sensory barbels30, is specialised for foraging on the bottom. At high mussel densities, glochidial concentrations in the water column are sufficient for P. phoxinus to forage effectively in both strata (Fig. 2B). In contrast, at low mussel densities (< 19 individuals per buffer), glochidia become scarce in the water column but remain abundant on the substrate. This shifts the zone of competition to the benthic habitat, where G. gobio, being better adapted, may outcompete P. phoxinus (Fig. 2C). This explains how 26.8% of the G. gobio population may intercept up to 33.8% of the available glochidia.

Fig. 2
figure 2

Conceptual model of “foraging-based” dilution: (A) graphical representation of the interaction between U. crassus density and glochidia load on the two dominant fish species; (B) at high mussel densities, large quantities of glochidia are released over a limited water surface, P. phoxinus consumes them in the water column, while remaining larvae settle and might be intercepted by G. gobio; (C) at low mussel densities, fewer glochidia are intercepted in the water column, forcing P. phoxinus to forage on the bottom, where it competes with the better-adapted G. gobio.

Full size image

Some studies directly report that fish may actively pursue and consume glochidia11,12. Although little is known about glochidial release in other naiad species, our observations of Pseudanodonta complanata suggest that they passively expel glochidia, which sink quickly and become available to benthic fish. Other species, such as U. pictorum and U. tumidus, release mature glochidia attached to mucous threads resembling a pearl necklace. These threads may reach up to 15 cm in length, which is substantial compared to the larval thread of U. crassus, the entire length of which fits within the field of view of a microscope. A very long thread carrying approximately 500 glochidia could be highly effective for infecting bottom-dwelling fish inhabiting the sub-habitats used by those mussel species31. However, a species that evolved a strategy to release glochidia higher in the water column, prolonging their suspension time, would provide more opportunity for pelagic fish—such as the abundant and competent P. phoxinus—to become hosts.

The data collected from across Poland indicate that G. gobio co-occurs with U. crassus and constitutes a significant proportion of fish assemblages at sites occupied by the mussel. This suggests that the changes observed in the fish community of the Warkocz River may also occur in other rivers, potentially affecting U. crassus populations at larger spatial scales. However, both infestation success and the strength of the dilution effect will depend on the local environmental context27: it is the complete set of available fish—both hosts and dead-end hosts—that determines infestation success, which implies that other bottom-foraging species, such as the common nase (Chondrostoma nasus), may substantially reduce the dilution effect exerted by bottom-dwelling dead-end hosts. Recent changes in fish community composition due to restocking efforts32,33, biological invasions34, and other anthropogenic pressures35 may profoundly influence host–parasite dynamics. While declines in freshwater mussels are often attributed to reductions in host fish populations36, alterations in the structure of non-host communities may also play a role by reducing the availability of glochidia to competent hosts. Although the importance of host species composition in completing the complex unionid life cycle is well established37, our findings suggest that host behaviour—and in particular the behavioural interactions between competent and dead-end species—is also important.

If a small subset of the fish community can intercept and remove a large share of glochidia due to specific foraging behaviours, then such behavioural traits become a key determinant of parasitic success and, consequently, of conservation outcomes for endangered mussels in low-density populations. Emerging evidence from other systems supports this view: in trematodes, the invasive snail Potamopyrgus antipodarum intercepts miracidia passively through its benthic behaviour, reducing infection in native snails3. Similarly, Schistosoma mansoni larvae may be diverted toward resistant or evasive snail species7. In terrestrial systems, the Virginia opossum (Didelphis virginiana) eliminates large numbers of ticks via grooming, acting as an ecological sink38.

These examples highlight the need to consider not only species richness and host competence, but also host behaviour and spatial ecology when investigating parasite transmission dynamics and developing conservation strategies for parasitic taxa such as Unio crassus.

Methods

The study was conducted in 2015–2016 in the Warkocz River, located at the foothills of the Świętokrzyskie Mountains in central Poland. The Warkocz is a primary tributary of the Nida River (Vistula basin, Baltic Sea catchment). It is 17.5 km long and drains a 54 km² catchment consisting of forested and agricultural landscapes. The river retains a largely natural character along most of its length, exhibiting a pool–riffle and meandering structure, and is incised approximately 2 m into the surrounding terrain. Channel width ranges from 5 to 10 m, and water depth varies from 0.1 m in riffles to 1.5 m in pools during low-flow periods.

The riverbed is predominantly sandy, though patches of silt, fine sediment, loam, gravel, and rocks are also present. In-stream vegetation is generally sparse, with occasional occurrences of Batrachium spp., Sagittaria sagittifolia, and submerged bank-root systems; isolated individuals of Nuphar luteum were found in pools. The riparian zone is dominated by Alnus glutinosa stands, often adjacent to meadows. The Warkocz River is protected under the EU Habitats Directive as part of the Natura 2000 site “Dolina Warkocza” (PLH260021). The Unio crassus population within this site is monitored regularly as part of the national Natura 2000 monitoring programme and has been declining.

Three study plots (plot 1: 50°50’25.0”N, 20°45’27.0”E; plot 2: 50°50’22.9"N, 20°45’26.8"E; plot 3: 50°50’22.2"N, 20°45’22.3"E; all at 257.9–258.5 m a.s.l.) were selected within the Warkocz River to represent slightly distinct habitat types occupied by U. crassus. Each plot consisted of a river reach, 60–100 m in length, and the plots were separated by 100 m stretches that were not surveyed. In early March of each study year, immediately after ice melt, each plot was precisely mapped using a regular grid of measurement points (1 m intervals longitudinally, 0.2 m crosswise, and 2 cm vertical resolution for depth).

The Warkocz River supports a viable, though fluctuating, population of Unio crassus with evidence of active recruitment. As this is the only unionid species present, any glochidia observed on fish could be unequivocally attributed to U. crassus.

All U. crassus individuals were mapped within each plot. Each year, substrates and riverbanks were carefully hand-searched to locate mussels, which were marked on detailed plot maps. Beginning in early April, individuals were examined for the presence of glochidia in the marsupia, to determine their phenology according to39. Once fully swollen marsupia were detected, samples of 20–30 gravid females were examined every 1–2 weeks to determine whether glochidia were mature and ready for release (i.e., displaying “snapping” behaviour). This was used to determine the appropriate timing for fish sampling.

Fish sampling and data collection

Each year, fish were sampled during three or more events coinciding with the expected peak glochidial release period in U. crassus (May–June): 5 May, 5 June, 18 June, and 2 July 2015; 11 May, 9 June, and 24 June 2016. The mean water temperature recorded in the study plots during May–June was 14.7 °C. According to40, glochidia remain attached for 16–28 days at 17 °C. Thus, if release occurred in mid-May, glochidia would be detectable on fish until mid-June. Since the dilution effect was expected only under high glochidial densities, analyses were restricted to the first 10 days of June in both years.

Each sampling event consisted of a single upstream pass using a battery-powered backpack electrofishing unit (IUP-12, Radet, Poland). Fish were immediately passed to staff responsible for recording the number of glochidia attached to fins. Specimens were identified to species, measured to the nearest 0.5 cm to minimize handling time, and examined for glochidial attachment. All fish were released after full recovery from electronarcosis, in areas unaffected by the electric field.

Spatial location of each captured individual was recorded during sampling. Prior to data collection, when the earliest signs of infestation were detected, a subset of fish from the site (three G. gobio and three P. phoxinus) was sacrificed to verify the presence of glochidia under a microscope and confirm the identification of Unio larvae12. Whenever doubts arose, we verified them directly in the field using a pen microscope; however, this was done only rarely, as the extended procedure could potentially harm the examined fish.

Random buffers and spatial distribution in the Warkocz river

Using QGIS (v3.14 “Pi”, Open Source Geospatial Foundation), random points were generated within a shapefile representing the Warkocz riverbed. These were used to define 5 m buffers along the channel. Spatial data layers for U. crassus, G. gobio, and P. phoxinus were intersected with each buffer, resulting in a geospatial dataset assigning occurrence values to each buffer. In total, 50 random buffers were generated per plot, resulting in 150 buffers, repeated across two years (2015 & 2016), for a total of 300 records.

Large-scale spatial analysis for Poland and its data sources

Occurrence records of Unio crassus were obtained from a curated distribution dataset41, comprising 424 georeferenced records from 71 rivers. Fish assemblage data originated from the Polish State Environmental Monitoring programme (GIOŚ) and included 4,867 surveys conducted in 2,259 rivers between 2011 and 2024 under the EU Water Framework Directive. Each record contained precise site coordinates, species composition, and abundance.

Spatial matching of U. crassus records with fish survey sites was performed in QGIS (v3.14). Vector layers of mussel localities and fish monitoring sites were intersected with the national hydrographic network (MPHP). Fish surveys located within 5.0 km and 0.2 km along the river network from U. crassus records were retained, yielding 158 surveys from 122 sites in 46 rivers (< 5 km) and 52 surveys from 35 sites in 24 rivers (< 0.2 km). Records referring to different watercourses or exceeding distance thresholds were excluded.

Co-occurrence frequency of the seven most frequently associated fish species was calculated as the proportion of surveys in which a species was present. For each survey, the percentage contribution of individuals belonging to these seven species was computed, and mean values were derived for both spatial categories. Results are presented in Fig. 1.

Controlled laboratory experiment

To assess host suitability for U. crassus, artificial infestations were conducted. Fish were collected via electrofishing in May 2022 from the same study plots. Only individuals free of visible glochidia on fins or opercula were selected. Ten G. gobio and ten P. phoxinus were transported in oxygenated river water to the lab. Gravid female mussels were collected by hand from the same locations. Marsupia were punctured with syringes and inspected under a microscope for the presence of active, snapping glochidia39. Five gravid females were transported in aerated water to a 96-L tank without mechanical filtration. The tank was monitored daily for larval release.

Once conglutinates were visible, fish were introduced for inoculation. Water was vigorously aerated and hand-stirred to prevent sedimentation and encourage fish movement. Exposure lasted 20 min.

Post-inoculation, fish were placed individually in 13-L tanks within a recirculating aquaculture system, with surface overflow and filtered, aerated water. A 2 mm mesh grid at the tank bottom prevented fish from consuming settled juveniles. On day 5, fish were examined for attached glochidia on fins and opercula. Every 3–4 days, tanks were flushed through plankton nets and juveniles were counted under stereomicroscopes. Juvenile counts were recorded per individual. Sampling continued until no further juveniles were observed (day 44). Laboratory temperature was maintained at 16 °C. All animals were released at their collection sites upon experiment completion.

Statistical analysis

Differences in abundance between dominant fish species were tested using proportion Z-tests. To assess the effects of G. gobio infestation level, G. gobio abundance, and U. crassus density on glochidial load in P. phoxinus, a Generalized Linear Model (GLZ; Poisson distribution, log link) was constructed. The response variable was the number of glochidia attached to P. phoxinus individuals. Predictors included: the number of G. gobio, number of U. crassus, and the number of glochidia on G. gobio, plus their interactions. The best-fitting model was selected using Akaike’s Information Criterion (AIC).

The threshold at which the effect of G. gobio on P. phoxinus infestation became neutral was calculated analytically from fixed-effect estimates in the GLZ model. The model included a main effect of glochidial load on G. gobio and an interaction with U. crassus abundance. The value of U. crassus density (N) at which the effect was neutral was derived by solving the equation:

$$\beta Gobio + \beta {\text{interaction}} \times N = 0$$

Rearranged:

$${\text{N}} = - \beta {\text{Gobio}}/\beta {\text{interaction}}$$

This value represents the density of mussels at which increased glochidial supply fully offsets the negative effect of G. gobio on P. phoxinus, effectively neutralizing the dilution effect.