Contrasting seasonal and spatial distribution of native and invasive Codium seaweed revealed by targeting species‐specific eDNA

Abstract Aim Codium fragile, an invasive seaweed, has spread widely during the last century, impacting on local seaweed communities through competition and disturbance. Early detection of C. fragile can help on its control and management. Environmental DNA (eDNA) has proved successful for early detection of aquatic invasive species but its potential use for seaweed remains understudied. We used a species‐specific eDNA qPCR approach to investigate the spatial distribution, abundance, and coexistence of the invasive C. fragile and three native Codium species (Codium vermilara, Codium tomentosum, and Codium decorticatum) in the Cantabrian Sea. Location Bay of Biscay, Northern Atlantic Coast of the Iberian Peninsula; two ports, a beach and a rocky cliff. Methods We designed species‐specific primers in barcoding regions targeting short fragments of the rbcL gene for the invasive Codium species, and the elongation factor Tu (tufA) gene for the native species, to assess their spatial and seasonal distributions using quantitative real‐time PCR in samples collected during summer, autumn, and winter. Results We found seasonal differences in the presence of the invasive Codium fragile and two of the native Codium species, but did not detect C. decorticatum at any point. Species distribution patterns produced with qPCR targeting species‐specific eDNA coincided with the known distribution based on previous conventional sampling, with a seasonal alternance of C. fragile and C. vermilara, and a marked dominance of invasive C. fragile in ports, which are known hotspots for invasive species. Main conclusions Our results demonstrate the utility of using eDNA for early detection and monitoring of invasive seaweed. Native and invasive Codium spp. displayed significant seasonal and spatial differentiation that needs to be taken into account in risk management. Regular monitoring of ports and adjacent areas using eDNA should help to assess the potential expansion of invasive Codium and the need for management interventions to avoid the displacement of native seaweed.

Cryptic invasion of morphologically similar invasive and native species (Provan, Booth, Todd, Beatty, & Maggs, 2008) has been identified as the most plausible cause for the competition between C. fragile and the native Codium spp. (García et al., 2018). Due to the wide physiological adaptations of C. fragile and its preference for higher temperatures during the reproductive season (Hanisak, 1979), new potential niches for its settlement are proliferating under current climatic conditions (Zanolla & Andreakis, 2016). Spatio-temporal information of native and invasive Codium spp. is crucial for evaluating whether patterns of competitive displacement or coexistence take place in Cantabrian Sea, where rising sea-surface temperatures have favored the spread of warm-water nonindigenous species over the past three decades (Díez, Muguerza, Santolaria, Ganzedo, & Gorostiaga, 2012).
Until now, knowledge on the spatial and seasonal distribution of seaweed has relied on traditional sampling methods (García et al., 2018), based on physical specimen collection and taxonomic identification, either based on morphological traits or molecular sequencing. These methods are typically limited by the feasibility of collecting specimens depending on the tides and weather conditions (Rojo et al., 2014), as well as by the multiple reproductive patterns of the different species (Schmidt & Scheibling, 2005). In addition, the taxonomic identification of different Codium spp. based on phenotypic traits is particularly challenging (Zanolla & Andreakis, 2016), often requiring molecular identification. Therefore, a more rapid and accurate detection tool is needed to monitor and/or control the distribution of invasive seaweed, which is less weather and tide dependent and incorporates the advantages of molecular identification.
Early detection of spatial and temporal changes in the distribution of Codium spp. is essential for assessing the potential displacement of native seaweeds in the Bay of Biscay. We evaluated the extent of seasonal and spatial overlap between native and non-native intertidal Codium seaweed. We also investigated whether invasive Codium was more frequent in ports than in natural coastal locations, in order to identify potential areas for targeted containment management.

| Study sites
Water samples were collected in July, October, and December 2017 at four different stations in Asturias (N. Spain) including a

| Ex situ experiment
An ex situ experiment was designed focusing on C. tomentosum to validate primer efficiency based on eDNA copy number with species density. The experimental setup consisted of six presterilized glass bottles with 1 L of marine water to which different densities (5, 10, 20, 40, 80, 160 g) of C. tomentosum were added and one control containing only seawater. Specimens were collected at Cabo de Peñas in October 2017 and brought in a cooling bag back to the laboratory, where they were identified morphologically following Provan et al. (2008), gently dried, and weighted before being added to the experimental 1-L water bottles ( Figure 1b). The increase of C. tomentosum biomass was based on doubling the previous weight to test for a correlation between eDNA quantity assessed by qPCR (Ct values) and species biomass. The marine water for the experiment was collected at a location with no known presence of C. tomentosum.
Water temperature was kept between 16 and 17.5°C. C. tomentosum specimens were kept in bottles for 36 hr and removed afterward. The water from the bottles was filtered using the same eDNA filtering procedure as described below for each bottle separately.
The negative filtration control using sterile nuclease-free water was filtered first, followed by filtration of marine water only, and then the rest of the bottles containing C. tomentosum in order of concentration, starting by the lowest. The DNA was extracted using the same protocol as for the collected eDNA water samples from field described below, including an additional negative extraction control, with extractions being stored at −20°C.

| Environmental DNA collection, filtration, and extraction
Three replicates of water samples (1 L of each) were collected with sterile bottles approximately 30 cm under the surface at all sampling F I G U R E 1 (a) DNA sampling locations from west to east side: Concha de Artedo, small port of Cudillero, rocky intertidal platform Cabo de Peñas, and international port of Gijón; (b) collection of C. tomentosum specimens and layout of the eDNA mesocosm experiment. The selected images of natural localities and ex situ experiment belong to authors, and the images of ports were collected from the Google marked with permission for reuse and modifications sites at consistent sampling points for each of the three sampling periods ( Figure 1a). All four sites were sampled either on the same day or in two consecutive days. Nitrile gloves were used while collecting the water. A cooling bag was used for the transportation to the laboratory where filtration took place immediately after returning from the field. Filtering in the laboratory took place in a dedicated eDNA room, where steps were taken to avoid eDNA contamination following Goldberg et al. (2016). A filter funnel was used for vacuum filtering in combination with sterile Supor1-200 Membrane Disc Filter (Pall Corporation) with 0.2 μm pore size. Water flow was 70 kPA.
For each of the sampling replicates, a maximum of two filters were used and stored together in a separate tube from other replicates at −20°C until the next day when DNA extraction was processed.
A negative control sample was filtered using sterile nuclease-free water between filtration samples from different sampling locations.
DNA was extracted on the following day of filtrations using the PowerWater ® DNA Isolation Kit Sample (Qiagen GmbH) following the manufacturer's recommendations with a modified last step of 50 µl for DNA elution. DNA extraction took place in a pressurized fume hood dedicated solely to eDNA handling. Sampling triplicates were extracted individually, including all five negative filtration controls with an additional negative control extraction sample for each of the sampling seasonal periods. DNA extractions were stored at −20°C before further processing.

| Primer design and validation
We developed species-specific primers in barcoding regions (rbcL and tufA genes) for the seasonal and spatial assessment of the invasive species C. fragile in coexistence with native Codium spp. We targeted 364 bp of the rbcL gene chloroplast subunit for the invasive C. fragile based on reference nucleotide sequences from GenBank, as this gene has previously been used for species identification (Verbruggen et al., 2007). For the three native species C. tomentosum, C. vermilara, and C. decorticatum, 211-, 180-, and 249-bp short fragments of plastid elongation factor Tu (tufA) gene were targeted to design species-specific markers ( Table 1). The plastid tufA and rbcL markers are some of the most widely applied markers to taxonomically separate the green algae group (Saunders & Kucera, 2010;Škaloud, Kynčlová, Benada, Kofroňová, & Škaloudová, 2012). To test the species specificity of the primers, they were firstly tested in silico using Primer-BLAST (Ye et al., 2012) and afterward used to amplify and cross-amplify tissue samples of the individual species before being used on eDNA samples for PCR and qPCR. First, primers were optimized for PCR, then for use in qPCR, where detection limits were determined. Cross-species amplifications were tested on each individual species amplifying it with all four primer pairs. C. decorticatum primers could not be tested on this species as no specimens were found along the Asturian coast at the time of the research. Extraction mixtures contained several specimens of each individual species to account for intraspecies variability. Tissues were extracted using GeneMATRIX Plant and Fungi Purification Kit (GeneMATRIX Purification Kit, Roboklon GmbH). A 100-fold dilution of an initial 1 ng/µl of each tissue was used for cross-amplifications in order to mimic eDNA detection levels in the environment.
All specimens of C. fragile collected in the Bay of Biscay region were identified based on sequencing as the invasive subspecies C. fragile ssp. fragile (Rojo et al., 2014), confirming the primer specificity for the subspecies. Oligo Analyser 3.1 tool (Integrated DNA Technologies) was used for primer check on hairpins and primer dimers. To estimate the detection sensitivity of each specific primer pair, 10-fold serial dilutions, starting from 1 ng/µl down to 1:10,000,000, were used and limits of detection were defined by qPCR amplification using dilution triplicates, for all three species individually. The last detectable melt peak at each species-specific melt temperature was accounted as detection limit and reported as corresponded dilution level. Additionally, the same 10-fold dilution was applied for defining the qPCR standard curve.

| PCR amplification
PCR and qPCR were optimized to avoid cross-species amplification for each specific primer pair. PCR conditions were as follows: 7 min at 95°C, followed by 10 touchdown cycles of 95°C for 30 s, 68-58°C TA B L E 1 Species-specific PCR primers used for amplification of targeted chloroplast rbcL and tufA region, with reported sequence, amplicon size (including primers), annealing temperature, qPCR detection limit based on 10-fold dilution series, and specific PCR and qPCR running conditions

| eDNA absolute quantification
In order to compare the seasonal and spatial distribution between the three species, absolute quantification based on differences in eDNA copies was performed, calibrated by each specific qPCR run efficiency. Absolute quantification determines the input copy number by correlating PCR signal to a standard curve (Schmittgen & Livak, 2008 The DNA copy number was used for calculation of the initial concentration given for the standard curve. Each standard curve was performed by a linear regression of the plotted standards. The slope of each standard curve determines qPCR efficiency (E), calculated by the following equation Lee et al. (2006): From the copy number of each standard, we quantified each sample by relating Ct values to the standard curve (Yu, Lee, Kim, & Hwang, 2005). Each specific sample quantification was performed as in Gallup (2011): All eDNA copy numbers were estimated per microlitre of filtered water (eDNA copies/µl).

| Statistical analysis
We modeled presence/absence data and species density in relation numbers (based on Ct values) depending on C. tomentosum actual biomass (g/l). All statistical analyses were done with the program R, version 3.3.2, with "dplyr" and "ggplot2" package used for data representation.

| RE SULTS
In total, 132 eDNA qPCR triplicates, 11 filtering, and three extraction negative controls were used for qPCR quantification. In seven of the samples, not all three sampling replicates produced speciesspecific positive confirmations, five targeting C. tomentosum and two targeting C. fragile; thus, sampling duplicates were used for further analysis. Triplicates of 21 eDNA samples, two filtrations, and one extraction negative controls from ex situ experiment were additionally processed for individual assessment based on correlation between C. tomentosum eDNA copies/µl and species density (g/l). There was no in silico possible cross-contamination for any of the species (native or invasive), tested with the Primer-BLAST tool on NCBI page (Johnson et al., 2008). No cross-amplification was produced either in PCR or in qPCR for any of the three species tested with all four primer sets, using dilution series of the three target species C. tomentosum, C. fragile, and C. vermilara.
Negative controls produced no amplification in any cases. Both controls from the ex situ experiment, the marine water, and nuclease-free water did not amplify during PCR and qPCR tested with all four primer pairs. All positive controls confirmed the target species by accurate alignment to sequences from target species, using BLAST and BioEdit (Hall, 1999  F I G U R E 2 eDNA density (Ct values) correlated to C. tomentosum actual biomass (g/L) in the ex situ experiment collected from Cabo de Peñas sampling point

| Spatial and seasonal variation
We evaluated the seasonal and spatial representation of C. fragile, C. tomentosum, and C. vermilara by qPCR quantification ( Figure 3). Overall, the most predominant two species were C. fragile and C. tomentosum, the latter accounting for the highest abundance of eDNA copies of all the species, with an average of 6.079 × 10 5 eDNA copies/µl in the two Western sampling points and 2.201 × 10 5 eDNA copies/µl at the Eastern sampling side. C. fragile was predominantly found on the east with an average of 5.629 × 10 5 eDNA copies/µl and a more even distribution between the three localities with species occurrence (±6.653 × 10 4 eDNA copies/µl), without spatially predominant patterns of C. vermilara eDNA presence ( Figure 3). We did not find C. fragile at Concha de Artedo, the most Western sampling point, whereas the highest eDNA presence was found at both ports, Cudillero with an average of 32.956 ± 1.78 Ct values corresponding to 4.780 × 10 5 ± 4.945 × 10 5 eDNA copies/µl, and Gijon with 32.733 ± 2.348 Ct values, corresponding to 7.929 × 10 5 ± 6.323 × 10 5 eDNA copies/µl. We detected the highest average density of C. fragile in the summer, but the highest single eDNA detection was measured in October in the port of Gijon with 3.192 × 10 6 eDNA copies/µl. The only locality where we found eDNA of C. fragile at all seasons was at the port of Cudillero, whereas in the port of Gijon we only detected it in the autumn sampling.
C. tomentosum eDNA presence was detected at all four stations, with a highest coverage in the summer and winter periods (Figure 3).

C. tomentosum exhibited the overall highest presence in summer
and winter compared to other two species, whereas the abundance of C. fragile was high in summer and autumn and declined in winter ( Figure 3). The highest abundance of C. tomentosum was detected in July at Concha de Artedo with 4.922 × 10 6 ± 9.515 × 10 5 copies/µl (24.814 ± 0.288 Ct value). eDNA from C. vermilara had been also found at all four stations with the highest representation in the winter, where on the average the eDNA copy number was 11.390% higher compared to autumn period (Figure 3). In the summer, we only detected at port of Cudillero with 32.023 ± 1.113 corresponding to 5.082 × 10 3 ± 3.380 × 10 3 eDNA copies/µl.
Seasonal and spatial presence/absence of species assessment using binary logistic regression, testing for an interaction between species, sampling location, and season, indicated high variation between species (

| D ISCUSS I ON
We used an eDNA approach to assess the spatio-temporal variation TA B L E 2 Evaluation of seasonal and spatial patterns of all three species using binary logistic regression for species presence/ absence assessment, identified with two models, first one based on species, sampling season, and location, and second one based on species, sampling season, and artificial/natural categories, including interactions between them TA B L E 3 Evaluation of seasonal and spatial patterns of all three species using linear models based on Gaussian distribution for species abundance estimation by eDNA copies/µl of C. fragile at both ports and its new detection at Cabo de Peñas confirms that this invasive species is spreading. The additional ex situ experiment of C. tomentosum contributed toward estimations of eDNA correlation with the relative density assessment in the field. eDNA density assessments using ex situ experiments have been used previously to estimate how relative abundance correlates with eDNA copies (Doi et al., 2015;Takahara, Minamoto, Yamanaka, Doi, & Kawabata, 2012;Wilcox et al., 2016), finding it as the most suitable measure for general biomass/density species-specific assessment being reflected in eDNA relative densities. We found no C. decorticatum in our eDNA sampling, confirming previous studies along the coast (García et al., 2018), despite having been occasionally reported (Cires Rodríguez & Rico Ordás, 2007). Tide-induced sampling limitations had been one of the potential causes proposed for finding no particular species during sampling events (Rojo et al., 2014), but our study suggests that this species was absent at the time of sampling as water sampling was not affected by the available shoreline sampling transect. C. decorticatum was not detected at all sampling events, as well as C. fragile was not detected at the most Western sampling point, which illustrates the usefulness of eDNA as a tool for seaweed monitoring. The east side higher density of C. fragile spread found with eDNA, overlapped with previous findings (Cires Rodríguez & Rico Ordás, 2007). Our results were also concordant with the previous surveillance at most western point of Concha de Artedo where in summer sampling events the majority of the specimens belonged to C. tomentosum with a small representation of C. vermilara and no confirmed presence of C. fragile (Rojo et al., 2014).

Factors of interactions
C. fragile are reproductively more successful in warmer waters with maximum growth at 24°C (Hanisak, 1979) compared to the two native ones with lower temperature preferences (Yang, Blunden, Huang, & Fletcher, 1997). This could explain the higher densities of C. fragile on the east side of Cantabrian coast due to higher summer temperatures modifying seaweed assemblages (Díez et al., 2012). Our results confirmed species-specific seasonal and spatial overlap with previously defined distribution (García et al., 2018). C. vermilara's optimum growth occurs at 18 µmol/mol of photon irradiance (Yang et al., 1997) and averaged quarter and half of the averaged photon irradiance of other five Codium spp., making it an ideal candidate species to shift its reproductive cycle toward colder seasons. C. fragile becomes a dominant canopyforming species once established as dense meadows in new environments (Scheibling & Gagnon, 2006) and could force C. vermilara to shift toward winter growth preferences. Similar coexisting acclimatization of two native and invasive kelp species in the same environment has been previously evidenced, where habitat preferences were identified through specific gene expression in correlation with temperature shifts (Henkel & Hofmann, 2008). The results show that C. fragile was the predominant species during autumn sampling, whereas previously it had been predominantly found in the summer period (Rojo et al., 2014). Colder spring and summer temperatures in the year of the eDNA sampling, with additional warmer temperatures in autumn (only 1°C degree difference from summer sampling), could have postponed C. fragile reproductive season toward autumn, and the corresponding increase in release of gametes (Bohmann et al., 2014) might be correlated to the eDNA density increase in that particular autumn. With the increasing temperatures along the N Spanish coast (Gómez-Gesteira, Decastro, Alvarez, & Gómez-Gesteira, 2008), a range shift in the relative abundance of seaweed species (Duarte et al., 2013;Voerman, Llera, & Rico, 2013) and the potential increase of C. fragile toward the west could be expected. Years with increased coastal upwelling at the Central Cantabrian Coast could potentially increase the seaweed distribution, as observed for the planktonic phase of local barnacle populations (Rivera et al., 2013). High chlorophyll concentration levels in summer have also been observed to have a positive effect on settlement of another invasive seaweed, the Asian kelp Undaria pinnatifida along the Cantabrian coast (Báez et al., 2010). Both the upwelling and increased chlorophyll levels seem to be the result of prevailing northeast winds during summer (Botas, Fernández, Bode, & Anadón, 1990), which result in thermal stratification, that could have prolonged the seasonal persistence of C. fragile.
A high eDNA density of invasive C. fragile was detected in both ports, with potential displacement of the native species.
Colonization of C. fragile subspecies on artificial marine structures is a regular occurrence around the globe (Bulleri & Airoldi, 2005;Campbell, 1999;Trowbridge, 1995), where artificial structures facilitate its spread. eDNA-based methods could be used for invasive green seaweed monitoring, by integration with port baseline surveys (David, Gollasch, & Leppäkoski, 2013) for ballast water management or implementation within Marine Strategy Framework Directive (Borja, Elliott, Carstensen, Heiskanen, & Bund, 2010;Directive, 2008). Despite the apparent noncompetitive status of C. fragile in the Cantabrian Sea due to their clear seasonal reproductive segregation with native species (García et al., 2018), there is no potential reduction in its introduction rates, which depends on multiple vectors (Boudouresque & Verlaque, 2010) such as shipping routes through ports.
Substantially higher eDNA copy numbers were identified for C. fragile and C. tomentosum in comparison with C. vermilara, which had 100-fold lower detection limits compared to the other two species. This could be explained by differences in qPCR efficiencies and variation in species-specific detection limits (Ludwig & Schleifer, 2000), where for between species primer calibration identical sequences between the primer targets are required. It has also been shown that qPCR primers efficiencies do not vary between different species or strains (Matsuki, Watanabe, Fujimoto, Takada, & Tanaka, 2004), indicating high precision of the method used for the interspecific comparison of Codium spp. in the present study. PCR assays may not vary greatly, depending on the species or strains. By using copy numbers per µl of DNA, it should be possible to compare the results between species, research facilities (Whelan et al., 2003), provided the same chemistry is used (Dhanasekaran, Doherty, & Kenneth, 2010), instead of comparing Ct values, which lack species-specific quantification precision. The ex situ experiment identified an upper limit of detection due to the selection of 1 ng/µl as the highest level of the standard curve dilution series in order to be able to detect low eDNA densities, representative for the values found in environment. Yet, the upper limit of detection was reached at concentrations that were unlikely to be found in the natural environment (Drouin, McKindsey, & Johnson, 2012;Scheibling & Gagnon, 2006). Primer specificity is important for successful detections of species (Mächler et al., 2014;Wilcox et al., 2013), but comparison of species densities using species-specific primers should be interpret with caution, comparing eDNA efficiency with the traditional abundance estimates (Agersnap et al., 2017). Thus, as the ex situ experiment was only conducted on C. tometosum, it is possible that different upper or lower detection limits applied to the other two species, C. fragile and C. vermilara, resulting in different detection levels, as eDNA quantification can vary even for the same species under different conditions (Klymus, Richter, Chapman, & Paukert, 2015). An internal inhibition control to monitor for PCR inhibitors (Henke-Gendo et al., 2012), was not added to each individual sample but could benefit toward qPCR run efficiency assessment.
Early detection of seaweed species in the aquatic environment can significantly improve AIS management and potential eradication (Jerde et al., 2013), with more efficient monitoring and containment of its spread (Tréguier et al., 2014), predicting its dispersal through spatial distribution models (Muha, Rodríguez-Rey, Rolla, & Tricarico, 2017), or influencing management and policy decisions (Kelly et al., 2014). As we have demonstrated here, eDNA can be used to assess the spatial and seasonal distribution patterns of invasive and native green seaweed algae.

| CON CLUS IONS
Our results on the distribution of native and invasive Codium species largely confirm those from more traditional surveillance methods, indicating that species-specific eDNA qPCR is an efficient and effective tool for monitoring seaweed seasonal and spatial patterns.
We found seasonal and spatial segregation in the presence of the invasive and two of the native Codium spp., potentially explaining the establishment success of the non-native species.

ACK N OWLED G M ENTS
We would like to thank the laboratory group at Department of

CO N FLI C T O F I NTE R E S T
The authors declare that there is no conflict of interest regarding the publication of this article. Researchers. The work is a product of their PhD research activities.

DATA ACCE SS I B I LIT Y
The data used for the following study are allocated in Table S1.