Assemblage organization over time prevents invasion success in phytoplankton

Invasion plays a central role in long-term species coexistence, sometimes rejuvenating and, at other times, collapsing diversity. Community resistance to invasions depends partly on the community’s species richness and structure, both of which change with community organization over time. Here, we study the resistance of phytoplankton assemblages to invasions based on varied initial species richness and structure, as well as assemblage organization over time. We hypothesize that invasion success will be greater in less organized assemblages and lower in more organized assemblages. To test this, we first experimentally constructed phytoplankton assemblages by mixing natural assemblages from regional lakes and manipulated richness along with a dilutional gradient as part of a large-scale mesocosm experiment. Phytoplankton assemblages from these mesocosms of varying dilution rate were selected and mixed to create unorganized assemblages of  high and low richness. We considered these constructed phytoplankton assemblages unorganized because their structures were not the result of ecological interactions. For an invader, we used the green alga Golenkinia radiata, a taxon historically observed in the region but absent from the natural assemblages at the time of sampling. We conducted four sequential invasion experiments, each lasting seven days, initiated with the increasingly organized assemblages of the mesocosms. We found that assemblages, while still unorganized early in succession, were vulnerable to invasions. However, as the assemblages organized with time, they became resistant to invasions. Assemblage richness, which ranged from 23 to 29 in the high-richness mesocosms and from 15 to 24 in the low-richness mesocosms, had only a marginal effect on invasion success. Instead, declining resources, notably phosphorus, reduced diversity, and the emerging dominance of strong competitors near the niche of the G. radiata invader explained the decreased success of invasions as the assemblages organized with time. In the more organized assemblages, G. radiata encountered formidable competitors in established populations of the diatom Nitzschia acicularis and the chrysophyte Synura sp., both of which had higher affinities for phosphorus than G. radiata. Our study highlights that assemblage organization with time plays a fundamental role in phytoplankton species coexistence, including a stronger resistance to invasion as assemblages mature.

Methods
Invasion experiments were initiated at different times using water samples collected from an ongoing mesocosm experiment. That mesocosm experiment had manipulated initial species pool richness from natural phytoplankton assemblages. Those constructed assemblages then organized over time, allowing for these invasion experiments (Figure 1). In the sections below, this is described in detail.

Collection of phytoplankton assemblages

Integrated water column samples were gathered from the euphotic zones of three sub-Alpine lakes in Lower Austria during August 2022. The euphotic zones were estimated using 2.5 times the Secchi depth (Nõges et al. 2010). The lakes included Lake Lunz (N 47°85’37”, E 15°05’23”, dimictic, maximum depth 32 meters), Lake Erlaufsee (N 47°79’30”, E 15°27’08”, dimictic, maximum depth 30 meters), and Lake Purgstall (N 48°09’61”, E 15°13’62”, polymictic, maximum depth 3.5 meters). Samples were sieved through a 100 µm mesh in the field to remove large zooplankton and were then transported to a mesocosm facility at the Biological Station of WasserCluster Lunz, Lunz am See. Lake assemblages were mixed and further sieved through a 55 µm mesh to eliminate microzooplankton. Consequently, an unorganized and species-rich assemblage was formed, which was then distributed into several mesocosm containers.

Making of unorganized assemblages, high and low richness, in mesocosms

The mesocosms were situated outdoors, with a volume of 320 liters, and constructed from food-safe polyethylene (ARICON Kunststoffwerk GmbH, Solingen, Germany). They rested on a gravel bed in an unshaded meadow (N 47°85’45”, E 15°06’76”). The water in each mesocosm was continuously, but gently, circulated using an airlift. The mesocosms were insulated with mineral wool to mitigate daily temperature fluctuations and covered with a 250 µm-mesh net to minimize the introduction of particles and other species.

To create unorganized assemblages of varying species richness, mesocosms were first maintained for a three-week period hosting the initial regional species pool inoculum at five different dilution rates in duplicate. The dilutions spanned a gradient of 0.066 d-1 to 0.505 d-1. The approach excluded slower-growing and rare taxa from the initial species rich pool in response to  dilution rates (Hammerstein et al., 2017), and successfully created a richness gradient. From assemblages of varied richness at the end of the three-week period, mixtures of varied proportions were used to further enhance the species richness gradient. By creating mixtures, any assemblage organization that had occurred during the previous three weeks was undone. These remixed assemblages were monitored for four weeks (see Smeti et al. In Preparation). It is during this four-week period that the invasion experiments were conducted. Two remixed assemblages —one with the highest and another with the lowest species richness— were utilized for our invasion experiment.

Design of the invasion experiments

Invasion experiments were conducted in 2.5-liter transparent plastic spouted bags (PET/NY/LDPE material, DaklaPack®), that were started on days 0, 7, 14, and 21 of the four-week mesocosm experiment described in Smeti et al. (In Preparation), with each lasting seven days. These experiments were initiated using assemblages collected from the high- and low-richness mesocosms, resulting in each invasion experiment being conducted with a progressively organized assemblage. The invader species used was Golenkina radiata Chodat, a species common to the region but absent from the mixed assemblages in the mesocosms. Two invasion propagules were explored, comprising 1% and 10% of the resident assemblage’s biomass, determined fluorometrically. Treatments and controls, with no added G. radiata, were performed in triplicate. Phytoplankton enumeration, which included determining population densities and cellular volumes for all taxa, was performed at the start of each invasion experiment, along with measurements for chlorophyll a (Chl-a), total phosphorus (TP), soluble reactive phosphorus (SRP), nitrate (NO3), nitrite (NO2), and ammonium (NH4). Cell counts of the G. radiata culture were also performed at these times.

Parameter measurements

To determine the volume of culture needed to produce our target invader propagules in the invader experiments, phytoplankton biomass estimates were made from the mesocosms and in the G. radiata culture. We used chlorophyll a as a proxy for biomass. We measured chlorophyll a in vivo using autofluorescence. A hand-held AquaPen-C AP-C 100 PSI (Drásov, Czech Republic) fluorometer, along with a 20-minute dark adaptation period, was employed for measurements of autofluorescence.

An analytical method for measuring Chl-a samples was also employed, conducted twice per week in the mesocosms, with one of the weekly measurements coinciding with the start times of the invasion experiments. In these instances, water was passed through GF/F filters (Whatman, pore size: 0.7 μm), extracted in acetone, stored at -20 °C until analyses, and then analyzed using a fluorescence method without phaeophytin correction (Arar and Collins, 1997).

For nutrients, TP was measured using an ascorbic acid colorimetric method (Hansen and Koroleff, 1999) after persulfate digestion (Clesceri et al., 1999). The detection limit of this method was 0.2 µg liter-1. A FLOWSYS Continuous Flow Analyzer (CFA, SYSTEA, Italy) was used for the measurements of NO3, NO2, NH4, and SRP with detection limits of 100 µg-N liter-1, 1 µg-N liter-1, 2 µg-N liter-1, and 0.04 µg-P liter-1, respectively. Nutrients were measured at least twice per week in the mesocosms, with one of the weekly measurements coinciding with the start times of the invasion experiments.

Samples for phytoplankton enumeration were collected twice weekly from the mesocosms and preserved in Lugol’s iodine solution, with sampling coinciding with the start times of the invasion experiments. Phytoplankton samples were analyzed according to Utermöhl method (1958) and Lund et al. (1958), using an Axio Observer 7 (Zeiss, International) inverted microscope at 400x magnification. Counts included 400 sedimentation units. Taxon-specific median biovolume for each species per sample was calculated using approximate geometrical forms (Hillebrand et al., 1999), based on at least 20 measurements for dominant taxa. Biovolume was converted to biomass by assuming a density of 1 g ml-1.

Calculations and Analyses

In this study, species richness refers to the number of taxa identified in our microscopic enumeration. Using the number of taxa and the biomass densities, diversity and evenness were determined with the Shannon Index (Shannon 1948) and Pielou's Index (Pielou 1966), respectively. Resource use efficiency (RUE) was calculated as ln(Chl-a/TP) following Ptacnik et al. (2008). The productivity of the phytoplankton assemblage was calculated as the total biomass change in the mesocosm divided by the initial total biomass for seven-day periods coinciding with the invasion experiments. The invasion success was assessed by the change in G. radiata biomass density in a bag divided by the biomass density of the added G. radiata. Large proportional changes were considered high invasion success (here in the 200-300% range), lower proportional changes were considered lower invasion success (here in the ~100% range), and decreases in the proportional changes were considered as unsuccessful invasion outcomes (here in the -50% to -100% range).

Invasion success into increasingly organized assemblages was assessed using one-way ANOVAs with Tukey’s post hoc tests and smooth spline models. Four data arrangements were explored with these tests based on the t0, t7, t14, and t21 invasion experiments that began with high- and low-species richness, and at 1% and 10% invader propagules. These statistics were performed using the Statistics Toolbox in the MATLAB programming language.

Relationships between invasion success, measures of assemblage structure (richness, evenness, diversity), performance (biomass, RUE, production), and nutrient availability (SRP, TP, NO3, NO2, NH4) at the 1% and 10% invader propagule levels were explored using multiple linear regression models. To achieve this, we created regression models representing all combinations of the measured parameters, omitting models that had diversity included with either richness or evenness, as diversity is calculated from those. So, the regression model that included all considered parameters (omitting diversity) took the form:

Invasion =       a0 + a1Richness + a2Evenness + a3*Biomass                               (1)

+ a4RUE + a5Production + a6SRP + a7TP

+ a8NO3 + a9NO2 + a10*NH4

and the regression model that included all considered parameters (omitting richness and evenness) took the form:

Invasion =       a0 + a1Diversity + a2Biomass + a3RUE + a4Production          (2)

+ a5SRP + a6TP + a7NO3 + a8NO2 + a9*NH4

where a0 through a10 were coefficients of the regression models. The root mean square errors (RMSEs) of all models were then used to determine the best models, with the lowest RMSE indicating the optimal model. We explored these regression models using both raw data and z-scored data. These regression models were performed using the Statistics Toolbox in the MATLAB programming language.

The relationships between invasion success and species composition of assemblages at 1% and 10% invader propagule levels were investigated using Principal Component Analysis. There are many taxa observed in these experiments. The multivariate analysis, here PCA, shows which taxa negatively associate with invasion success, suggesting which taxa are likely bringing about poor invader outcomes. To achieve this, a covariance matrix was created from the z-scored species biomasses and invasion success data from the invasion experiments using a 1% propagule, and another covariance matrix was generated from the invasion experiments using a 10% propagule. PCAs were conducted using the Statistics Toolbox in the MATLAB programming language.

The distribution of P-affinities was used to explore if lumpy coexistence was occurring in these assemblages, as this is a factor important to invasion success. The P-affinities can also be used to identify like competitors. Distribution of the taxa in each assemblage along the P-affinity gradient was estimated using log10(cell volume), based on the assumption that body size distributions represent niche differentiations (Scheffer and van Nes 2006, Segura et al. 2011, 2013, Muhl et al. 2018, Graco‐Roza et al. 2021). Nutrient affinities for each taxon were estimated using the bioinformatics-driven online tool Phyto-PhlyoPars (Bruggeman et al. 2009, Bruggeman 2011). Our focus was on phosphorus affinity (P-affinity), since nitrogen was less influential in these experiments (to be shown).