Effects of the invasive macroalgae Gracilaria vermiculophylla on two co-occurring foundation species and associated invertebrates

Mads S. Thomsen*, Peter A. Stæhr, Lars Nejrup and David R. Schiel 1 Marine Ecology Research Group, School of Biological Sciences, University of Canterbury, Private Bag 4800, Christchurch, New Zealand 2 UWA Oceans Institute and School of Plant Biology, University of Western Australia, Hackett Drive, Crawley 6009 WA, Australia 3 Aarhus University, Department of Bioscience, Frederiksborgvej 399, Box 358 DK-4000 Roskilde, Denmark 4 Orbicon A/S, Ringstedvej 20, DK-4000 Roskilde, Denmark E-mail: mads.solgaard.thomsen@gmail.com (MST), pst@dmu.dk (PAS), lbne@orbicon.dk (LN), david.schiel@canterbury.ac.nz (DRS) *Corresponding author


Introduction
Invasive non-native species typically have a negative impact on the abundance and diversity of native species and community structure (Vilà et al. 2011).However, non-native species can also be foundation species (Dayton 1972) that increase biodiversity by creating and modifying habitats for other organisms (Rodriguez 2006;Wallentinus and Nyberg 2007).Impacts of nonnative foundation species are likely to depend on the habitat type that is invaded.For example, non-native foundation species that invade habitats that lack functionally similar native foundation species are likely to cause positive effects on associated invertebrates, because these organisms may use the non-native species as food, habitat, for stress amelioration, and protection from grazers and predators (Rodriguez 2006;Wallentinus and Nyberg 2007;Thomsen et al. 2010).
However, most non-native foundation species probably invade habitats that are already dominated by one or more foundation species (Staehr et al. 2000;Levin et al. 2002;Ward and Ricciardi 2010;Hammann et al. 2013), and net impact on the local communities in these cases are more complex.For example, if the invasive and native foundation species have negative effects on each other (e.g., competing for limited resources), facilitation of organisms that depend on the invasive species could be offset by loss of organisms that depend on the negatively-affected native foundation species.Predicting invasion impacts on native foundation species and associated communities will be even more complicated in habitats where multiple native foundation species co-exist.For example, in mixed seagrass-mollusc habitats, these native foundation species may have both positive and negative effects on each other, depending on densities, environmental conditions, and species involved (Reusch and Chapman 1995;Reusch and Williams 1998;Peterson and Heck 2001a;Peterson and Heck 2001b;Vinther et al. 2008;Vinther et al. 2012).Currently, few field experiments have tested how marine invaders affect multiple foundation species and their associated communities, making it difficult to understand and predict invasion impacts in many natural systems.
Gracilaria vermiculophylla (Ohmi) Papenfuss (hereafter Gracilaria) is a coarsely-branched red alga that originates from the northwest Pacific.This species, like most other seaweeds, modifies the local abiotic environment (e.g., sedimentation, anoxia, light levels), provides habitat for numerous sessile and mobile species (Thomsen et al. 2010), and can therefore be considered a foundation species.Gracilaria has spread to shallow-water wave-protected estuaries and coastal lagoons along 1000s of km of coastline in the East Pacific, West Atlantic, East Atlantic, and Mediterranean Sea, making it one of the world's most successful marine invasive species (Kim et al. 2010;Sfriso et al. 2010).Within these ecosystems, Gracilaria is common on 'barren sediments' (Nejrup and Pedersen 2010;Sfriso et al. 2012) as well as habitats occupied by native foundations species, like salt marshes (Thomsen et al. 2009), seagrass meadows (Cacabelos et al. 2012;Hernández Cordero et al. 2012), fucoid seaweed beds (Weinberger et al. 2008;Hammann et al. 2013), polychaete reefs (Thomsen and McGlathery 2005;Byers et al. 2012) and bivalves reefs (Thomsen and McGlathery 2006).Impacts of Gracilaria have been documented on some of these native foundation species and/or their associated communities, typically demonstrating contextdependency.For example, accumulations of unattached Gracilaria have negative effects on oysters (and sessile species living on the oyster reefs) (Thomsen and McGlathery 2006), but appear to have less impact on functionally similar mussels, as Gracilaria is often found attached to the byssal threads of the blue mussel Mytilus edulis (hereafter Mytilus) (Weinberger et al. 2008;Thomsen et al. 2010).Furthermore, Gracilaria has no effect on the seagrass Zostera marina (hereafter Zostera) in low densities and at low temperature, but negative effects when occurring in high densities under high temperatures (Hoeffle et al. 2011).However, we are not aware of any studies that have tested for effects of Gracilaria on two native foundation species, and potential cascading effects on the communities that are associated with native foundation species.
We therefore tested, in a factorial field experiment, the hypotheses that Gracilaria has (1) negative effects on the co-occurring native foundation species Zostera and Mytilus, but (2) positive effects on invertebrate communities by providing more and/or different habitat structure in mixed Zostera-Mytilus beds.

Study site
The study was conducted in a seagrass bed at 2 m depth in the northern part of Odense Fjord, Denmark (55.52659 N, 10.531254 E).Mytilus, Zostera and Gracilaria are all common in this estuary, but Mytilus and Gracilaria are sparse at this specific site making it easier to conduct 'addition-experiments' and avoid colonization from settling or drifting Mytilus and Gracilaria into control plots.During the experiment, water temperature varied from 16 to 20ºC.At the study site salinity varies seasonally between 13 and 22, and turbidity is high with Secchi depth varying between 1-6 m but reduced to near 0 m following periods of strong east-southeast winds (abiotic data from the Danish National Aquatic Monitoring and Assessment Program; DNAMAP, https://oda.dk).

Field methods
We conducted a 3-factorial orthogonal experiment to test for reciprocal effects between 'Gracilaria' (G±), 'Zostera' (Z±) and 'Mytilus' (M±).Each of the three foundation species was manipulated as a 'presence' ('+') vs. 'absence' ('-') treatment in a 2×2×2 design.The abundance of each foundation species was manipulated in 40 0.4×0.4m plots, i.e. with 5 replicates for each of the eight treatment-combinations.We did not use cages, thereby avoiding cage-artefacts, including changes to hydrodynamics, light conditions, sediment/ seaweed-trapping and attracting animals.
Zostera biomass was manipulated by removing all above-ground biomass from the 20 Z-plots by cutting all leaves with scissors at the sediment surface (Herkül and Kotta 2009).Seagrass in the Z+ plots were disturbed with hands simulating scissors disturbances without removing any biomass.
Mytilus biomass was manipulated by adding four live individuals to the 20 M+ plots (average size = 37.0 ± 8.1 g wet weight or 6.3 ± 0.5 cm shell length; based on 20 randomly chosen specimens; all reported values are means ± SE).The mussels were collected from a nearby site (<1 km away).We gently scraped off large attached sessile species, e.g., barnacles, as well as seaweed fragments incorporated into byssal threads.The mussels were then carefully added to the sediment surface around the seagrass stems without breaking any leaves.
Gracilaria biomass was manipulated by adding c. 3 kg wet weight (WW) m -2 to the 20 G+ plots.This Gracilaria was collected from the nearby Holckenhavn Fjord (5517.8´N,1046.2´E)because the alga at this site has little epiphyte cover and few clinging invertebrates.The collected Gracilaria was brought ashore, shaken to release the few mobile macro-invertebrates, and further inspected for clinging invertebrates which were then removed by hand.Gracilaria was fixed to the substratum in the 20 G+ plots by inserting 10 u-bent thin metal pegs flush with the sediment surface (Thomsen et al. 2012).A similar number of pegs were inserted into the 20 Gplots flush with the sediment surface to control for peg-artefacts (so that any peg-induced disturbances were similar between plots).
The experiment was initiated on 28 August 2012 and ran for 4 weeks.This is a common time period for such experiments where unattached seaweeds persist in a specific seagrass patch and a common time interval to run seaweed-seagrass impact studies (e.g., Nelson and Lee 2001;Holmer and Nielsen 2007;Martínez-Lüscher and Holmer 2010;Höffle et al. 2011;Holmer et al. 2011).The experiment was conducted in late summer/early fall because key invertebrate species produce recruits during this period and could be facilitated by the three foundation species (Thomsen 2010).

Collections and laboratory methods
At the end of the experiment we collected a 290 cm 2 circular core (with sharp edges) from each plot centre.A mesh-bag (1-mm mesh size) covered the top of the core to ensure mobile animals did not escape.We approached each plot slowly before inserting the core over the centre.The core was hammered through the seagrass rhizomes and 10 cm into the sediment.A small shovel was used to dig up sediments into the mesh-bag attached to the core, to ensure all infaunal animals were collected.All core content was pushed into the bottom of the attached mesh-bag together with a plot-marker tag.The mesh-bag was then closed with a string and detached from the core.A new mesh-bag was attached to the core before approaching the next plot.Two of the 40 plots were lost (probably from fishing, vandalism from local snorkelers, or storms) resulting in two of the treatments (Z+M-G-; Z+M-G+) having 4 replicates.The 38 meshbags were shaken gently in the field to remove sediments and to 'stress' the collected invertebrates to avoid animal loss due to predation during transport to the laboratory.
Mesh-bags were kept cool until arrival at the laboratory where we immediately separated Gracilaria, Zostera, Mytilus and macroinvertebrates from any remaining sediments by sieving through a 2-mm mesh sieve.Foundation species and invertebrates were then separated from each other, and Zostera biomass was further separated into above-(leaves) and belowground (root and rhizomes) biomass.The biomass of each foundation species was measured after drying at 60 C until no further biomass loss occurred (g DW per core).Invertebrates were immediately conserved in 70% alcohol and, over the following weeks, identified to the lowest practical taxonomic level (usually species) and counted.Amphipods were grouped together as a single taxonomic unit.Sedentary polychaetes were omitted from the analysis because many of these small fragile animals were lost or broken through sieving (data were therefore deemed unreliable).

Data analysis
Tests were conducted as factorial permutationbased ANOVA ('PERANOVA') on univariate responses and permutation-based MANOVA ('PERMANOVA') on multivariate community structure, where Gracilaria, Mytilus and Zostera treatments were considered fixed factors (Anderson et al. 2008).Both univariate and multivariate analyses were conducted with the PERMANOVA add-on to Primer v6 software package, using 4999 permutations (Clarke and Gorley 2006;Anderson et al. 2008).We also compared sum of square values to discuss what test factors explained most of the data variability (Levine and Hullet 2002).
We conducted univariate factorial analyses on the above and below ground biomass of Zostera, biomass of Gracilaria, biomass of Mytilus, on densities of gastropods, bivalves, crustaceans, echinoderms, errant polychaetes (these taxa together constituted >95% of sampled individuals) and total invertebrates density, as well as on invertebrate richness, diversity (Shannon index), and evenness (Pielou's index).All these univariate analyses were conducted using Euclidian distances on untransformed data under a reduced model (Anderson et al. 2008).Most univariate variables had homogeneous variances, although variances for invertebrate densities for the Gracilaria treatment were slightly heterogeneous (Levines test, p > 0.005).However, we did not transform these responses, in part because ANOVA is relatively robust to variance heterogeneity in balanced replicated designs (Underwood 1997).
We did not include the G-and M-plots in the analysis of Gracilaria and Mytilus biomass, respectively, because these species had virtually no chance of colonizing their control plots (i.e., we found zero biomass in these controls).Gracilaria and Mytilus biomass were therefore analyzed with 2-factorial tests (Appendix 1C, D, Figure 1C, D).By contrast, biomass of Zostera were (like the invertebrate analyses) analyzed with 3-factorial tests because the below-ground biomass was not manipulated and because the above-ground biomass in the Z-treatments could potentially recover through (a) horizontal growth from seagrass adjacent to the plots, (b) vertical growth from cut leaves within the plots, and (c) seed germination (Herkül and Kotta 2009).We conducted tests on both total Mytilus biomass and Mytilus density (2-4 shells per core).The statistical results between Mytilus biomass and Mytilus density were similar and we therefore only present the biomass data here (this biomass can then be compared directly to the biomass of the two other foundation species).
We conducted multivariate 3-factorial analysis on the invertebrate species-sample matrix, using Bray Curtis similarity coefficient, and squareroot transformed densities to downplay the importance of the dominant taxa (Anderson and Ter Braak 2003;Clarke et al. 2006).Note, however, that we found similar results when data were analyzed without transformation and with a more severe log (x+1) transformation (unpubl.data).The multivariate pattern was visualized with a 2D-PCO plot (Anderson et al. 2008).

Impacts on foundation species
Not surprisingly, we found a highly significant effect of Zostera removals on its own aboveground biomass (Appendix 1A), with > 5 times more biomass in the Z+ plots (3.88  0.65 g DW core -1 ; all reported values are means ± SE) compared to the Z-plots (0.80  0.15 g DW core -1 ).Furthermore, adding Gracilaria significantly reduced Zostera above-ground biomass (G-= 2.85  0.73 vs. G+ = 1.68  31 g DW core -1 ) although graphical comparisons suggested that most of the negative effects occurred in the presence of Mytilus (Figure 1A; compare Z+M+G-vs.Z+M+G+).We found nearsignificant effects (p < 0.1) of Mytilus on Zostera above-ground biomass (indicating that mussels have a positive net effect on seagrass biomass) and on the Z×G interaction (indicating that negative effects of Gracilaria were smaller in Z-than Z+ plots, Figure 1A).
Effects were less pronounced on the belowground Zostera biomass (Appendix 1B, Fig. 1B); Mytilus had significant positive effects (M-= 6.78  0.44 vs. M+ = 8.83  0.66 g DW core -1 ) and there was a near-significant M×G interaction (indicating that positive effects of Mytilus on Zostera was stronger in G-than G+ plots).
As expected, we found Gracilaria in all the G+ plots but none in G-plots (Figure 1C).ANOVA did not detect any effects of Zostera or Mytilus treatments on Gracilaria biomass (5.97  1.35 g DW core -1 , all p-values > 0.35, Appendix 1C), even though there appeared to be more Gracilaria in the Z+M+ treatment.
We found 2-4 (all alive) Mytilus in the M+ but none in M-plots.'Lost' mussels from the M+ plots had probably re-positioned themselves outside of the plot center (we found no empty shells to indicate predation).However, there was no pattern regarding what treatmentcombinations had lost most mussels and this lack of pattern was reflected in the ANOVA on mussel biomass, which did not detect any effects of Zostera or Gracilaria treatments (53.53 g DW  3.44 core -1 , all p-values > 0.19, Appendix 1D, Figure 1D).
A comparison of the dry weight between the three foundation species (Figure 1A-C, Appendix 2) showed that Mytilus was dominant; however, Gracilaria and Zostera dominated 'visually' in the field (due to high volume-to-dry weight Figure 1.Effects of the three foundation species Gracilaria vermiculophylla (G±), Zostera marina (Z±), and Mytilus edulis (M±) on their own and each other's biomass.The experiment was conducted in a seagrass bed, removing the above-ground Zostera leaves and adding Gracilaria and Mytilus in all 2×2×2 treatment-combinations (N = 5 for most treatment-combinations, except N = 4 for Z+M-G-and Z+M-G+).We found no significant interaction effects; horizontal bars represent significant single-factor effects (see Appendix 1 and the result section for details).The G-treatments was excluded from analysis of Gracilaria biomass and the M-treatment from Mytilus biomass, because these foundation species had little opportunity to colonize control plots (i.e., with zero biomass at the end of the experiment).ratios of Gracilaria and Zostera; see Appendix 6 for examples of differences in species traits between the three foundation species).

Impact on associated invertebrates
A total of 22 taxa were identified in the 38 samples, dominated by gastropods and crustaceans (including 3 shrimp species -Crangon crangon, Palaemon adspersus and Hippolyte varians).The most common taxa were amphipods, the sea star Asterias rubens (recruits only), the crab Carcinus maenas (adults and recruits), the snails Rissoa membranacea and Littorina littorea (adults and recruits), and the isopod Idotea baltica.We also found 2 fish species (Pomatoschistus microps, Pholis gunnellus).

Individuals per core
The analysis on diversity was the only test with a significant interaction term.In this test, we found a significant G×M interaction on diversity (Figure 3B, Appendix 1L) as well as significant G and M single factor effects, where most sums-of-squares data variability was accounted for by Gracilaria.This G×M interaction showed that Mytilus had a positive effect on invertebrate diversity in the absence (M+G-= 1.35  0.08 vs. M-G-= 0.79  0.13) but not presence M+G+ = 1.52  0.11 vs. M-G+ = 1.54  0.07) of Gracilaria.
Evenness (Pielou's index) was the only invertebrate response not affected by any of the treatments (Figure 3C, Appendix M).
Finally, we found significant effects of Gracilaria and Mytilus on the multivariate community structure, and a near-significant G×M interaction, with Gracilaria treatments explaining three times more of the data variability than Mytilus (Appendix 1N).This multivariate results result was visualized with a 2D-PCO plot (Figure 4) showing that communities with Gracilaria were quite similar (as demonstrated by a relatively tight sample cluster).

Discussion
Non-native foundation species often invade habitats already occupied by native foundation species but little is known about how they interact with each other (Ward and Ricciardi 2010) and possible cascading effects on the organisms that depend on these foundation species (Thomsen et al. 2010).
Here we documented that the invasive seaweed Gracilaria had relatively low impact on two cooccurring native foundation species, and vice versa, but that it had a strong positive effect on habitat-associated invertebrates.Although our results are constrained by the experimental setup (i.e., a short time period, relatively small plot sizes, embedded in seagrass meadow, with relatively low invader density) we hypothesize that these findings are typical when invaders are structurally more complex than the native species and when they occur in localized patches over space and time.

Impact on foundation species
The relatively small effects of the three foundation species on each other are perhaps not surprising.Previous laboratory experiments testing for impacts of Gracilaria on Zostera done over a similar time period also showed small negative effects (Martinez-Luscher and Holmer 2010;Hoeffle et al. 2011).These findings were supported by our field experiment as Gracilaria had negative effect on above-ground, but no effect on below-ground, biomass of Zostera.The negative effect of Gracilaria on Zostera is likely associated with reduced levels of light, nutrients, oxygen and water currents around the basal seagrass leaves and meristem in the presence of the seaweed (Holmer and Nielsen 2007;Holmer et al. 2011).
In contrast to Gracilaria, Mytilus facilitated Zostera, perhaps because bivalves through their filtering capacity and metabolic activities can increase nutrients and decrease turbidity (Reusch et al. 1994;Peterson and Heck 2001b), and their byssal threads may stabilize and protect the rhizomes (Reusch and Chapman 1995;Peterson and Heck 2001b).Interactions between bivalves and seagrasses can, however, be both positive and negative depending on spatial locations, eutrophication levels, wave exposure, and abundance (Reusch et al. 1994;Reusch and Chapman 1995;Vinther et al. 2008;Vinther et al. 2012).For example, Vinther et al. (2012) found correlative evidence for a threshold of coexistence, as Zostera was never found when Mytilus was present at more than 1.6 kg WW m -2 .
Although we did not find any effects of Mytilus on Gracilaria we noted that Gracilaria was incorporated into the mussels byssal threads in all M+ plots (a few Zostera leaves and rhizomes were also incorporated; see Appendix 4).We therefore expect that this bivalve can be important in stabilizing Gracilaria populations, as unattached Gracilaria is susceptible to removals during storms and by tidal currents (Reusch and Chapman 1995;Thomsen 2004).Therefore, if we had not used pegs to stabilize Gracilaria, it might have occurred in lower abundances in the M-than M+ plots.
Finally, the Mytilus itself was not significantly affected by either Gracilaria or Zostera, which is consistent with observations that this mussel often is found together with Gracilaria (Weinberger et al. 2008;Thomsen et al. 2010) and Zostera (Vinther et al. 2012).Again, it is possible that high densities of Gracilaria over spatially extensive areas may cause negative impacts on Mytilus through oxygen limitation, water current reductions, and by interfering with bivalve filtration capacity (Norkko and Bonsdorff 1996;Tyler 2007;Vinther et al. 2008;Vinther et al. 2012).From our experiment it appears that the three foundation species can co-exist when they occur in low to medium abundances and over small/short time scales.However, if Gracilaria and Mytilus are found in high abundance, in large areas, and over longer time frames, Zostera could be dramatically inhibited, particular under stressful conditions, such as high temperatures, excessive nutrient levels, and low light levels (Huntington and Boyer 2008;Vinther et al. 2008;Hoeffle et al. 2011;Vinther et al. 2012).

Impacts on associated invertebrates
The three foundation species differed in their habitat suitability for the associated invertebrates; Gracilaria provided better habitat than Mytilus, which was more important than Zostera (Figure 2-3, Sum-of-Squares in Appendix 1).
It may appear surprising that we, in contrast to established theory (Heck et al. 2003;Boström et al. 2006), found no clear facilitation from the seagrass on invertebrates.However, the experimental habitat (a seagrass bed) likely causes strong spill-over edge effects of seagrass-associated invertebrates into removal plots.Furthermore, above-ground removals were not 100% efficient and showed partial recovery causing all plots to have some seagrass above-ground biomass (Figure 1A).Finally, below-ground biomass was not manipulated, leaving a relatively large and similar below-ground biomass in all plots.We expect stronger facilitation of seagrass on invertebrates if we instead add Zostera to Mytilus or Gracilaria beds or to unvegetated sediments.
Perhaps more surprisingly, adding Mytilus only indicated weak (non-significant) positive effects on invertebrate densities.Again, Mytilus has, like Zostera, been shown to facilitate mobile invertebrates compared to 'barren' sediments (Markert et al. 2010).The lack of a significant effect on invertebrate densities by Mytilus could be caused by large data variability -perhaps being swamped by strong Gracilaria effects.Still, graphical inspection suggests that Mytilus do facilitate invertebrates because mean values were higher across the M+ than the Mtreatments for all taxonomic groups (Figure 2), thereby supporting the pattern observed by Markert et al. (2010).Indeed, we did find significant effects on multivariate community structures and increases in taxonomic richness and diversity (Appendix 1), leading us to conclude that adding Mytilus to seagrass beds increased invertebrate biodiversity through provision of additional and/or different habitat.
Gracilaria had a much stronger positive impact on invertebrate densities than did Zostera and Mytilus, and we found 38, 5, 5, 4, 4 and 2 times more crustaceans, total invertebrates, echinoids, bivalves, errant polychaetes and gastropods, respectively, when Gracilaria was present (= 'Magnification Ratios', see Thomsen et al. 2010).Furthermore, Gracilaria had positive effect on taxonomic richness and diversity, and modified the multivariate community structure causing invertebrate communities to be relatively more homogenous in the presence of Gracilaria.This strong across-the-board facilitation by Gracilaria is not surprising because its coarsely branched fronds provide a complex 3-dimensional habitat which is characterized by different-sized interstitial spaces for different species in different ontogenetic phases to occupy, a large attachment space for bivalves to recruit onto, and likely also high protection from predators.Similar facilitation of invertebrates has been found in other habitats (Nyberg et al. 2009;Byers et al. 2012;Johnston and Lipcius 2012;Hammann et al. 2013) and on different Gracilaria species (Thomsen et al. 2012), and many studies have found invertebrate densities to be high on algae with complex thalli as opposed to those with simples forms (Hacker and Steneck 1990;Taylor 1994;Chemello and Milazzo 2002;Wernberg et al. 2004).
In this study, we tested for impacts of foundation species using a planned presenceabsence approach because we wanted to test for interaction effects (having multiple densities of foundation species in 3-factorial designs requires a very high number of total plots).However, we found few interaction effects and we therefore supplemented the ANOVA with non-parametric correlation and step-wise multivariate linear regression analyses (Appendix 4 and 5).In these analyses we used the biomass per core of each foundation species as predictor variables.These tests confirmed the key importance of Gracilaria over Mytilus and Zostera (almost no additional variation was explained by adding Mytilus or Zostera to Gracilaria in the regression models) and suggested that the facilitation process was density-dependent for all the significant responses detected in the ANOVA.Similar types of correlation analysis have previously been used to show positive continuous effects of foundation species (Bishop et al. 2012;Byers et al. 2012;Gribben et al. 2013) and density-dependent impacts of invasive species appears to be universal across taxa and ecosystems (Parker et al. 1999;Thomsen et al. 2011).

Perspectives, study limitations and future studies
Our study provides a rare experimental test of how an invasive foundation species can interact with multiple native foundation species with cascading effects on associated invertebrates.We suggest that this scenario -where invasive and native foundations species co-occur with positive effects on associated fauna -is a relatively common phenomenon.This process represent 'cascading habitat formation' (a habitat/facilitation cascade, Altieri et al. 2007;Thomsen et al. 2010) where primary/basal habitat-formers (or modifiers/facilitators -here Zostera or Mytilus) have positive effects on secondary/inter-mediate habitat-formers (or modifiers/facilitators -here Gracilaria that become entangled around stems and byssal threads) to indirectly facilitate endusers (here invertebrates) through trait-or density-mediated interactions (Cruz-Angon et al. 2009;Byers et al. 2010;Thomsen 2010;Bishop et al. 2012).
There are numerous opportunities to expand our experiment to identify mechanistic links between invasions and cascading impacts on communities and ecosystem function.We here documented that most invertebrate end-users were facilitated by the invasive foundation species, but future studies should also test how these invertebrates are facilitated.For example, endusers may benefit from habitat-forming foundation species by escaping enemies (e.g., competitors, parasites, predators) and environmental stress (e.g., waves, heating, desiccation), or by finding friends (e.g., mating partners, mutualist, schooling benefits, allee effects) and resources (e.g., nesting/resting space or food, but note that Gracilaria is a poor food resource for some grazers (Nylund et al. 2011;Nejrup et al. 2012;Rempt et al. 2012)).These mechanisms will likely differ between environmental conditions (Gracilaria may provide predation refugia in the subtidal zone but ameliorate desiccation stress in the intertidal zone), life histories (a juvenile crab may use Gracilaria to avoid predators whereas the adult crab may use it as feeding ground), and species characteristics (a bivalve may use it as substrate for attachment, whereas a snail may use it for grazing).
More specifically, we found strong facilitation of herbivorous end-users, like amphipods, isopods and snails, suggesting that this trophic level graze on Gracilaria (the secondary habitatformer) or on associated microscopic epiphytes.Many studies have shown that herbivores can have positive indirect effects on primary habitatformers by preferentially consuming secondary habitat-formers (e.g., Boström and Mattila 1999;Worm and Sommer 2000;Jones and Thornber 2010).This type of 'keystone consumption' (Thomsen et al. 2010) is thereby a mirror-process of cascading habitat formation/facilitation, but where research focus is top-down enemy-vs.bottom-up facilitation-processes, respectively.Understanding these direct and indirect mechanisms and how they vary in space, time and across invaders and invaded habitats is vital to provide better assessments of impacts from invasive foundation species.
Our experiment was, like all experiments, constrained in space, time, and across taxa.Future experiments should therefore explore different spatio-temporal conditions and different invasive vs. native foundation species and enduser taxa.We expect that the positive effects on invertebrates have cascading impacts on higherorder predators, in particular small fish, and, if the invasion occurs on larger scales, also on larger top-predatory fish, birds and mammals.Similarly, we focused on mobile species (including slow moving mollusks), but facilitation of sessile species that depend on hard substratum for attachment can also be important (Thomsen et al. 2010).For example, we noted that filamentous brown and red seaweeds were attached to Zostera, spirorbid polychaetes were attached to Gracilaria, and barnacles, bryozoa and hydrozoa were attached to Mytilus.Future studies should therefore also quantify sessile end-users (as well as smaller meiofauna and fragile sedentary polychaetes) to better understand facilitation cascades involving foundation species.
In addition, we conducted our experiment as a pulse treatment, that is, as a single initial manipulation.We purposely did not maintain treatment levels throughout the experiment because our first emphasis was to test if the foundation species had a strong impact on each other (which they did not).This approach also represents a typical local invasion scenario when invaders are seasonally and spatially patchy distributed, such as when unattached Gracilaria clumps drift into a seagrass patch and become entangled for 1-2 months before disappearing again (Sfriso et al. 2012).Indeed, changes in seaweed biomass over the course of a field experiment are common and can be caused by decay and fragmentation, growth, grazing, biotic disturbances, and hydrodynamic stress (e.g., Nelson and Lee 2001;Huntington and Boyer 2008).However, if the foundation species have strong effects on each other, invertebrate responses may co-vary with biomass changes of the foundation species.Thus, if stronger effects are expected between foundation species and the main aim is to quantify responses to associated species, 'press treatments' with repeated manipulations are more suitable.
In summary, effects of invasive foundation species on native foundation species and associated communities should be tested with pulse and press experiment, with and without stabilizing pegs/cages, with multiple densities, and in different seasons (to reflect temperature and recruitment patterns), locations (e.g. of different salinities), spatio-temporal scales (larger and longer), habitats (e.g., on 'neutral' sediments, in Mytilus reef, in Gracilaria bed), as well as including novel manipulations to control predation, resources, and abiotic stress (e.g., using mimics).

Figure 2 .
Figure 2. Effects of the three foundation species Gracilaria vermiculophylla (G±), Zostera marina (Z±), and Mytilus edulis (M±) on the density of invertebrates.The experiment was conducted in a seagrass bed, removing above-ground Zostera leaves and adding Gracilaria and Mytilus in all 2×2×2 treatment-combinations (N = 5 for most treatment-combinations, except N = 4 for Z+M-G-and Z+M-G+).We found no significant interaction effects; horizontal bars represent significant single-factor effects (see Appendix 1 and the result section for details).

Figure 3 .
Figure3.Effects of the three foundation species Gracilaria vermiculophylla (G±), Zostera marina (Z±) and Mytilus edulis (M±) on the taxonomic richness, Shannon's diversity, and Pilou's evenness of invertebrates.The experiment was conducted in a seagrass bed, removing above-ground Zostera leaves and adding Gracilaria and Mytilus in all 2×2×2 treatment-combinations (N = 5 for most treatment-combinations, except N = 4 for Z+M-G-and Z+M-G+).Horizontal bars represent significant single-factor effects (see Appendix 1 and results for details).There was a significant G×M interaction on diversity; Mytilus had a positive effect in the absence (M+G-= 1.35 vs. M-G-= 0.79) but not presence (M+G+ = 1.52  vs. M-G+ = 1.54) of Gracilaria.