The Borbecker Mühlenbach is a small urban stream categorized as German stream type 6 (fine-grained, carbonic mountain streams; Sommerhäuser and Pottgiesser, 2005; Figure 1A) that has its source in Essen, North Rhine-Westphalia, Germany. Like most streams within the Emscher catchment, the Borbecker Mühlenbach was used as an open sewer system for wastewater disposal from the 19th century on. Recently, the stream was ecologically restored to a large extent (mainly carried out in 2011, finished in 2014) and today consists of both piped underground sections and restored above-ground sections. This study focuses on a 1050 m long above-ground section of the Borbecker Mühlenbach that represents a young freshwater ecosystem due to its recent restoration. A loose stone dam is located within the studied stream section, which was deemed to represent a potential dispersal barrier for the Rhine sculpin as it is approximately 40 cm high and blocks the entire width of the stream (Figure 1B). Further upstream of the stone dam, the stream Kesselbach enters the Borbecker Mühlenbach. After the studied stream section had been deemed suitable for reintroducing the Rhine sculpin post restoration, 118 Rhine sculpin individuals taken from the river Boye (Bottrop, Germany) were released into the stream at sampling sites 3–5 (Figure 1C) on August 23, 2017. To prevent individuals from drifting downstream into underground pipes, a net was installed in front of the pipes.

All lab work was carried out in an eDNA laboratory under sterile PCR hoods, and the whole lab, as well as the individual PCR hoods, were UV-light sterilized for 45 min between separate work cycles. DNA was extracted from filters using a modified salt extraction protocol [original protocol by Sunnucks and Hales (1996); for modified protocol see Weiss and Leese (2016)]. For the purification, either the NucleoSpin Gel and PCR Clean-up kit (MACHEREY-NAGEL GmbH & Co. KG, Düren, Germany) or the MinElute Reaction Cleanup Kit (QIAGEN Inc., Germantown, MD, United States) were used, following the manufacturer’s instructions. During each extraction and purification step, it was ensured that only one sample was opened at a time to avoid cross-contamination. The final DNA extracts were eluted into 20 μL of water. The success of all extractions and purifications was verified using agarose gel electrophoresis.

Primers targeting the hypervariable mitochondrial 12S rRNA gene of the Rhine sculpin were designed using a reference database containing 272 sequences of the 12S rRNA gene from 57 fish species (Hänfling et al., 2016). These included all native fish species resident in the Borbecker Mühlenbach and Kesselbach. The complete mitochondrial genome of the Rhine sculpin was downloaded from NCBI GenBank (MF326941/NC_036147, both identical), which was representative for the study population in the 12S rRNA gene sequence, and added to the downloaded 12S rRNA dataset [for final Mafft (Katoh and Standley, 2013) alignment of the 12S rRNA gene sequences, see Supplementary Appendix A]. The newly designed primers C_12S_377F (5′–AGGCCCAAGTTGACAAACAC–3′) and C_12S_731R (5′ – GGCGGGTAAAACAAGGAAGG–3′) amplify a 344 bp long region of the Rhine sculpin’s 12S rRNA, which is located within the 12S rRNA fragment targeted by the universal fish 12S rRNA primers 12S_30F and 12S_1380R (Hänfling et al., 2016). The application of a nested PCR assay design (nPCR) was chosen as this method is known to significantly improve detectability compared to conventional PCR (Davison et al., 2019). The designed primers are universal for 20 out of 24 species of the Cottus complex with available 12S sequences on GenBank including Baikal sculpins, according to PrimerBlast (NCBI)¹. No other Cottus species has been reported from the Emscher catchment (Gunnar Jacobs, personal communication), therefore, it is extremely unlikely that other Cottus species rather than the re-stocked C. rhenanus would be the source of captured eDNA.

The presence of Rhine sculpin DNA within the eDNA extracts was tested with end-point nPCRs using the QIAGEN Multiplex PCR Plus Kit (QIAGEN Inc., Hilden, Germany). We included negative PCR controls as well as field blank controls in the PCR runs to check for a clean PCR setup and cross-contamination. During each PCR step, it was ensured that only one 8-sample PCR microtube strip was opened at a time to avoid cross-contamination. For the first PCR step, 2 μL of the extracted eDNA, 0.25 μL of 100 μM universal fish 12S rRNA gene primers 12S_30F and 12S_1380R (Hänfling et al., 2016), 25 μL of the Multiplex MasterMix, and 22.5 μL water were used per reaction. PCR conditions consisted of an initial incubation at 95°C for 15 min followed by 30 cycles of 95°C for 30 s, 50°C for 30 s, and 72°C for 50 s, and final elongation at 72°C for 10 min. For the second PCR step, 1 μL template of each PCR product from step 1, 0.25 μL of 100 μM newly designed primers C_12S_377F and C_12S_731R, 12.5 μL of the Multiplex MasterMix, and 11.25 μL water were used per reaction. PCR conditions consisted of an initial incubation at 95°C for 15 min followed by 30 cycles of 95°C for 20 s, 63°C for 30 s, and 72°C for 1 min, and final incubation at 72°C for 5 min.

Half of the extracted DNA of every sample (10 μL) was used as a template for the first PCR step to minimize stochasticity effects, leading to five PCR replicates per sample containing 2 μL of the extracted DNA each. The presence of the Rhine sculpin was confirmed if a band was present at the expected amplicon length using agarose gel electrophoresis (with 1% agarose gels run for 15 min at 80 V). In case the Rhine sculpin was not detected in the first half of the eluate, the remaining eluate was used as an input for further PCRs. If this approach also did not show any signal, then the Rhine sculpin was considered to be absent at the sampling site.

Samples from each sampling day were tested up to site 12. If the species was detected behind the barrier (sites 11 and 12), then sites 13 and 14 were also tested.

To validate that the newly designed primer pair amplified our target species C. rhenanus in field samples, PCR amplicons of 22 positive replicates were Sanger sequenced (Eurofins Genomics, Ebersberg, Germany). These included replicates from all positive samples upstream of the barrier, as well as all from samples that resulted in just one positive replicate (excluding sample 9 B, for which Sanger sequencing failed). All Sanger sequenced replicates confirmed the detection of the Rhine sculpin.

We assumed the detection of Rhine sculpin eDNA at a given site if at least one replicate of our end-point nPCR approach confirmed amplification. We did not have the option for more sensitive qPCR or ddPCR, yet our applied end-point nPCR approach is known to lead to reliable detections (Nathan et al., 2014; Davison et al., 2019). Furthermore, end-point PCRs can have low detection limits (e.g., Qu and Stewart, 2017), which is why we considered that our results were more impacted by low detection rates than cross-contamination.

Based on end-point nPCR eDNA detections, we showed that individuals dispersed at least 200 m within the first 5 days (positive sample 9 B), on average at least 40 m/day, which greatly exceeds daily moving distances of the Rhine sculpin in an established population observed by Ovidio et al. (2009; maximum of 5.51 m/day). This verifies our first hypothesis that the reintroduced individuals disperse faster from their reintroduction sites than expected from their known stationary behavior. An explanation for this pattern is that the initially high density of individuals at each reintroduction site led to high intraspecific competition for habitats. A density-dependent, exploratory behavioral mechanism to compensate for this high density is the dispersal of smaller individuals to sub-optimal habitats (Davey et al., 2005). Note that maximum moving distances of up to 395 m for the Rhine sculpin during 1 year and over 250 m for the closely related European bullhead during several months have been observed (Knaepkens et al., 2005, 2006; Ovidio et al., 2009). Furthermore, seasonal migration has been shown for the European bullhead (Crisp et al., 1984). However, none of these studies investigated the fish’s movement at a temporal scale comparable to the present study.

The Rhine sculpin was not detected in eDNA samples upstream of site 10 within the first 3 months and furthermore not detected with electrofishing after 1 year at these sites, indicating a dispersal limitation by the barrier before site 11. However, eDNA detections of the species upstream of the barrier after 1 year (detections at sites 11 G and 13 G) and moreover after one additional month (detections at sites 11 H, 11a H, and 11b H) indicate that at minimum one Rhine sculpin individual was present upstream of the barrier. This result was unexpected (see the second hypothesis) and not supported by our electrofishing survey. We based our second hypothesis on the finding that the European bullhead (which used to be a cryptic species including the Rhine sculpin) is unable to cross solid stream barriers of 18–20 cm height (Utzinger et al., 1998). However, the barrier in the present study consists of loose stones and is occasionally flooded during heavy rainfall events. It is possible that juveniles were able to swim through gaps in the loose stone dam or that individuals were able to cross the barrier after heavy rainfalls. Another explanation for positive eDNA detection upstream of the barrier is that predators (for example, gray heron) that preyed upon Rhine sculpin individuals defecated fish DNA into the upstream section. Human-induced translocation of individuals across the barrier (for example, by hand-netting) can also not be excluded, as the stream is urban and frequently visited by pedestrians. Although we consider predator- or human-induced translocation of Rhine sculpin individuals or DNA as unlikely, we cannot exclude it. Our results indicate a higher dispersal potential of the species than previously assumed because the loose stone dam was deemed to represent a substantial dispersal barrier for small, benthic fish, but this indication cannot truly be evaluated, due to the impossibility to rule out other dispersal mechanisms within the scope of this study.

The inconsistency between eDNA-based and traditional monitoring results for sites upstream of the barrier was unexpected. An explanation for the failure in detecting the species using electrofishing is that individuals were located upstream of the electrofishing sites and hence could not be detected via electrofishing. In that case, they were still releasing eDNA that was transported downstream and was collected, making the detection of the individuals possible. However, a more comprehensive electrofishing survey would be needed to confirm this explanation. In contrast to that, the positive detection of Rhine sculpin DNA could also be predator-induced, as earlier discussed. In that case, no actual individuals of the species would have been present upstream of the barrier, which would explain the failed electrofishing detection.

After 1 year, the Rhine sculpin was detected at every sampling site downstream of the barrier using eDNA and at every sampling site downstream of the barrier except for site 1 using electrofishing. This verified previous expectations and motivations for the reintroduction of the Rhine sculpin into the stream section, justified by the presence of suitable habitats for the species. Additional sampling following the Rhine sculpins’ next reproduction period is needed to investigate the fish’s further dispersal and mid-term reintroduction success.

The finding of juveniles during the electrofishing survey confirmed the reproduction of the species. The confirmed reproduction, together with the fact that we captured 113 individuals with electrofishing in a fraction of the stream habitat while 118 individuals were released in total, implies a substantial increase in population size. This increase is a likely source for the highly increased number of positive eDNA-based detections after 1 year, although the adsorption-release dynamics of eDNA might also have facilitated the detection, i.e., more eDNA being released in contrast to the initial phase (Spear et al., 2015; Shogren et al., 2017).

Our results confirm successful reintroduction, dispersal, and establishment of the species within the stream Borbecker Mühlenbach. Additionally, the species’ establishment indicates good water quality, habitat structure and passability of the stream and consequently indicates a successful stream restoration.

Our electrofishing observations suggest that the Rhine sculpin was not able to cross a minor barrier in the stream, which is congruent with previous observations for the closely related European bullhead (Utzinger et al., 1998). However, the results of our eDNA analysis indicate that the species was present upstream of the barrier, which would mean that the mobility of the Rhine sculpin and its capacity to cross instream obstacles is better than previously assumed. Greater electrofishing efforts, which are typically performed in standardized surveys, might have led to certainty about the presence or absence of the Rhine sculpin upstream of the barrier. Nevertheless, our observed trend is consistent with previous studies showing that overlooking individuals is the rule rather than the exception with electrofishing (Jerde et al., 2011; Pont et al., 2018). Thus, although analyzing eDNA cannot be used for estimating individual numbers or age structure, its application in biomonitoring presents a promising, non-invasive, and sensitive tool for species detection.

Nevertheless, the applicability of eDNA analysis as a monitoring tool is still limited, as many characteristics of eDNA are poorly understood yet. It is known that eDNA release can be affected by various conditions of an individual, for example, size, health, sex, fecundity, or diet (Goldberg et al., 2015; Klymus et al., 2015); the extent and variation between species and even individuals remain, however, unclear. Moreover, eDNA durability in aquatic environments is strongly influenced by degradation through temperature, pH value, conductivity, UV light, and microbes (Barnes et al., 2014), reducing eDNA concentration and hence causing additional bias. Furthermore, eDNA movement processes such as downstream transport, soil retention, and soil resuspension constantly affect eDNA concentration (Shogren et al., 2017). Apart from these ecological issues, methodological difficulties can lead to failure in detecting eDNA, caused by stochastic effects and PCR inhibition (Taberlet et al., 1996; Thomsen and Willerslev, 2015). Together, these limitations may lead to inexplicable patterns of eDNA detection in some studies [observed by Jane et al. (2015) and Laramie et al. (2015)].

eDNA can be transported downstream over long distances in flowing waters (Deiner and Altermatt, 2014; Jane et al., 2015; Deutschmann et al., 2019), which implies that Rhine sculpin eDNA was transported downstream from actual residence sites in our study. This effect was observed for sample 1 G, where the species was detected with eDNA even though electrofishing failed to detect individuals up to 20 m upstream of the site. Consequently, eDNA detections of the Rhine sculpin downstream of the uppermost detection site do not equal true local detection. Visual observation of individuals by traditional surveys is still necessary for the validation of local presence at sites downstream of actual residence sites. However, when validating the general presence of a fish species in flowing waters, this effect might be beneficial, as there is no need to sample the entire stream or river.

In this study, the Rhine sculpin was not detected at several sites despite the presence of the species further upstream on the same day. These occurrences represent false negatives, whose presence is consistent with comparable studies (Foote et al., 2012; Jane et al., 2015; Laramie et al., 2015) and can be explained by dilution of DNA by high flows, increasing distance from the DNA source, or PCR inhibition (Jane et al., 2015); eDNA dynamics in the system (e.g., retention to stream bottom; Shogren et al., 2017); and the less sensitive end-point nPCR approach we used (Turner et al., 2014). Inhibition is especially problematic in eDNA studies, as it can mask even high eDNA copy numbers (Jane et al., 2015) and hence can lead to false negatives although the target organism is present (Goldberg et al., 2015; Thomsen and Willerslev, 2015).

Our results show that false negatives can be reduced by both high numbers of technical replicates and periodical sampling. An increase from 5 to 10 technical replicates led to positive detection of the Rhine sculpin in several cases (Table 1, samples 7 B, 10 C, 7 F, 13 G, 11a H, 11b H), and periodical sampling enabled the detection of the species at sites where detection failed at previous sampling events (Table 1, samples 4 C, 6 C, 8 C, 4 E, 5 E, 6 E). However, addressing the previously mentioned challenges in eDNA studies, especially PCR inhibition, will further decrease the number of false negatives in future eDNA studies for target species detection.