Structural and Functional Changes in a Lake Bacterial Community Exposed to Multiple Stress Regimes

(1)

Master's Thesis

Structural and Functional Changes in a Lake Bacterial

Community Exposed to Multiple Stress Regimes

Submitted by Philipp Siegel

Stockholm, 2015

(2)

I Supervisors: 1. Dr. Ingo Fetzer

2. Ass. Prof. Dr. Silke Langenheder

(3)

II Statement of Authorship

I hereby declare that I wrote this master's thesis myself without sources other than those indicated herein.

Stockholm, 31.05.2015 ____________________________

Philipp Siegel

(4)

III

List of Figures

Figure 1: Two-dimensional ball-in-cup model, which is commonly used for visualizing alternative stable states. The "ball" represents a system or community and the "cups" the environments in which the system can be placed in. Alternative stable states in microbial communities have not been found yet. Adapted from Beisner et al., 2003. ...3 Figure 2: Map of Sweden & Lake Ekoln, yellow arrow indicates sampling site at 59.76°N 17.58° E.

Source of the satellite image: LandsatLook Viewer (Data available from the U.S. Geological Survey). ...7 Figure 3: Temperature set-ups in the 6 different treatments. a) Temperature in treatments 1 and 2

("Control" and "Control Salinity") was constant at 15°C throughout the experiment. b) Temperature in treatments 3 and 4 ("Gradual Temperature Increase" and "Gradual Temperature Increase Salinity") was increased from 15°C by 1°C each day until 30°C and then decreased by 1°C until 15°C again. c) Temperature in treatments 5 and 6 ("Shocks" and

"Shocks Salinity") was increased from 15°C by 1°C each day until 30°C and then decreased by 1°C until 15°C again. Additionally the treatments were exposed to 5 temperature shocks. ...8 Figure 4: Bacterial productivity per cell in the six treatments over time. Linear models were fitted

onto the curves for the time between day 1 and day 16, as well as between day 16 and day 32 coinciding with the periods when stress was increased and decreased respectively, but also when sampling regime was more and less frequent. Blue line at day 16 in both "Gradual Temperature Increase" and both "Shocks" treatments indicates the time point after which temperature was stepwise decreased. Grey bars in the "Shocks" treatment plots indicate the periods of temperature shocks. ... 15 Figure 5: ß-glucosidase activity per cell in the six treatments over time. Linear models were fitted

onto the curves for the time between day 1 and day 16, as well as between day 16 and day 32 coinciding with the periods when stress was increased and decreased respectively, but also when sampling regime was more and less frequent. Blue line at day 16 in both "Gradual Temperature Increase" and both "Shocks" treatments indicates the time point after which temperature was stepwise decreased. Grey bars in the "Shock" treatment plots indicate the periods of temperature shocks. ... 17 Figure 6: Cellobiohydrolase activity per cell in the six treatments over time. Linear models were

fitted onto the curves for the time between day 1 and day 16, as well as between day 16 and day 32 coinciding with the periods when stress was increased and decreased respectively, but also when sampling regime was more and less frequent. Blue line at day 16 in both

"Gradual Temperature Increase" and both "Shocks" treatments indicates the time point after which temperature was stepwise decreased. Grey bars in the "Shock" treatment plots indicate the periods of temperature shocks. ... 19 Figure 7: EEMs of the water samples in the different treatments from the first day and the last day

of the experiment. Color scale is from 0 (dark blue) to 3 (dark red) Raman Units. ... 21 Figure 8: Cell abundances per ml in the six treatments over time. Linear models were fitted onto

the curves for the time between day 1 and day 16, as well as between day 16 and day 32 coinciding with the periods when stress was increased and decreased respectively, but also when sampling regime was more and less frequent. Blue line at day 16 in both "Gradual Temperature Increase" and both "Shocks" treatments indicates the time point after which

(7)

VI temperature was stepwise decreased. Grey bars in the "Shock" treatment plots indicate the periods of temperature shocks. ... 23 Figure 9: NMDS of BCC in the six treatments over time. Single profiles of each treatment (colored

numbers) were combined for an average BCC per day and treatment. Average BCC from each day and treatment are written out and connected via colored lines. Numbers 2, 3, 4, and 5 behind the treatment names correspond to the 9th, 15th, 25th and 32nd day of the experiment respectively. Single profiles of the starting community from all treatments were averaged to obtain NMDS scores for a "Day 1" BCC. ... 25 Figure 10: Environmental fitting of measured community functionalities, DOM parameters, time,

salinity, and cell abundance Environmental fit based on identical results as Fig. 9. Significant linear trends (p < 0.01) fitted onto all the measured community profiles. Arrows that point in the same direction have a positive correlation. Arrows that point in opposite directions have a negative correlation, and arrows that stand at a 90° angle to each other have no correlation. ... 26 Figure 11: Rank clocks of dominant (>0.1 relative abundance) OTUs over time. The rank clocks show

the increase and decrease of OTUs over time and have to be read like a clock: The starting and end point of the experiment (days 1 and 32 respectively) lie at "12 o'clock". The half-time of the experiment (day 16) lies at "6 o'clock". Direction of the progression is clock-wise. The numbers 10, 20 and 30 on the outer circle are the x-axis labels and correspond to days during the experiment. ... 28 Figure 12: Location of five primary fluorescence peaks in an EEM. The white area in the upper left

corner of each EEM is ... 47 Figure 13: Progression of the main fluorescence Peaks A, C and M over time. ... 48 Figure 14: Progression of the main fluorescence Peaks B and T, and of the Humification Index over

time. ... 48 Figure 15: Relative fractions of the five main fluorescence peaks in the total pool of fluorescent

DOM over time. ... 49 Figure 16: Progression of FI and BIX in the treatments over time ... 50 Figure 17: Progression of SUVA 254, A350, and A440 in the different treatments over time ... 50 Figure 18: Progression of S275 - 295, S350 - 400, and the Slope Ratio in the different Treatments over time51

(8)

VII

List of Tables

Table 1: Descriptions of each treatment during the experiment ... 8

Table 2: Sampling days in the first half of the experiment ... 42

Table 3: Sampling days in the second half of the experiment ... 42

Table 4: Coefficients of variation for BP time series ... 43

Table 5: Regression coefficients for linear models in BP time series ... 43

Table 6: Coefficients of variation for ß-glucosidase activity time series ... 43

Table 7: Regression coefficients for linear models in ß-glucosidase activity time series ... 43

Table 8: Coefficients of variation for cellobiohydrolase activity time series ... 44

Table 9: Regression coefficients for linear models in cellobiohydrolase activity time series ... 44

Table 10: Coefficients of variation for cell abundance time series ... 44

Table 11: Regression coefficients for linear models in cell abundance time series ... 44

Table 12: Correlation of cell abundances (CA), bacterial productivity (BP) per cell, glucosidase activity (GA) per cell and cellobiohydrolase activity (CBH) per cell. Significant correlations are marked bold. ... 45

Table 13: Correlation of cell abundances (CA) with total glucosidase (GA) and cellobiohydrolase (CBH) activities, and with total bacterial productivity (BP). Significant correlations are marked bold. 45 Table 14: Summary of the various measured DOM parameters and their interpretation ... 46

Table 15: Correlation and p-values of the environmental fitting ... 52

Table 16: Evar-Index of all OTUs from this study over time ... 53

Table 17: Shannon-Index of all OTUs from this study over time ... 53

(9)

1

Abstract

Bacterial community resilience and resistance may play an important role in a world affected by climate change and increasing environmental disturbances. In this study we examined the resilience of a bacterial lake community from Lake Ekoln, Sweden, under increased environmental stress. The community was exposed to various stress regimes of altered temperature and salinity to answer the questions if and how higher amounts of stress influence the bacterial community in terms of its community structure and functionality. We measured bacterial productivity, ecto-enzyme activities and changes of fluorescent dissolved organic matter to evaluate the community's functional responses under stress and used terminal-restriction length polymorphism (TRFLP) to assess the community's structural characteristics over time. Although bacterial community composition (BCC) changed constantly under changing environmental conditions, community functionality remained high and the community performed well under the various stress regimes. Our results suggest that the Lake Ekoln bacterial community is highly resilient considering the investigated stresses, and its adaptive capacity is high due to the community's inherent adaptability and redundancy.

1. Introduction

In ecosystems, microorganisms often perform viable system functions such as nitrogen fixation, primary production, remineralization or nutrient cycling (Madigan, 2012). These processes form the first and most essential step in the global biochemical cycle. Microbial communities shape the underlying processes of an ecosystem they live in (Schimel et al., 2007), but are in turn heavily influenced by their surrounding world (Cutler et al., 2013; Fuhrman et al., 2006). In the face of global climate change it is therefore important to investigate how microbial communities will be influenced by predicted environmental changes because they are often the first to be affected by disturbances in an ecosystem (Begon et al., 2006; Shade et al., 2012a; Walther et al., 2002). They have fast turnover rates and respond quickly to environmental changes (e.g. Schmidt et al., 2007). This circumstance makes them very adaptive on one side but can bring an undesirable shift quickly, that could potentially deprive the system of a valuable ecosystem service.

Environmental disturbances will become even more intense and frequent in the near future than they are today (Planton et al., 2008). Especially lake ecosystems will underlie alterations (Adrian et al., 2009). Lake surface temperatures at the latitudes of Sweden or Canada for example are predicted to reach >25°C in some future climate scenarios (Keller, 2007) and some European lakes are also predicted to underlie salinisation (Moss et al., 2009;

Smith et al., 2008). Until now the effects of climate change on microbial community structure and function in lakes have not been fully investigated (e.g. Adrian et al., 2009).

Knowledge about the dynamics of microbial lake communities under stress will improve our ability to initiate adequate management actions in the future.

(10)

2 1.1 The theoretical background: Ecological resilience and changes in ecosystem functioning

Adaptive changes in communities towards environmental factors may occur linear or can be abrupt, with sudden loss of essential functionality (Scheffer et al., 2001). While the former allows both easy detection and predictability on the system’s response, the latter does not allow estimation on the system’s state and prediction of time points when functionality is lost. Complex systems such as lake ecosystems or coral reefs have been shown to drastically change from one stable state to another when gradual changes in underlying variables accumulate to a critical threshold (Folke et al., 2004; Scheffer et al., 2001). Most systems will have a threshold that should not be surpassed if a desired functioning of that system is to be ensured (Steffen et al., 2015). Scheffer and his colleagues (2009) showed that an approaching transition could be detected because of an increase of variance, greater fluctuations in ecosystem variables, longer recovery times after perturbations as well as a greater autocorrelation.

It is assumed that the general response of a community depends on the amount of species and amount of functional redundancy within the community (Walker, 1992). Former studies have shown that species-rich communities are ecologically more resilient and able to keep up their functions when exposed to change than less diverse communities (e.g. Allison, 2004;

Elmqvist et al., 2003; Folke et al., 2004; Walker et al., 1999). Although research on complex systems and their behavior is increasing and several key components of critical transitions between two stable states could be identified in the past, the actual time point of transition is still hard to predict and poses a challenge (Scheffer et al., 2009). Resolving this issue will allow for much more effective systems management.

The theory of ecological resilience builds on the assumption that an ecosystem can exist in at least two alternative stable states with feedbacks that keep the system from alternating freely between the states (Beisner et al., 2003; Folke et al., 2004; Holling, 1973; May, 1977;

Walker et al., 2004). For microbial systems, no clear evidence of multiple stable states exist so far (Fig. 1) (e.g. Allison and Martiny, 2008; Baho et al., 2012; Fuhrman et al., 2015; Shade et al., 2012a). A system can be altered through press disturbances (e.g. slow gradual temperature increases as in the case of global warming) or pulse disturbances (e.g extreme weather events such as storms or droughts). How these types of disturbances influence a system depends on the system's characteristics. Disturbances can affect fast-changing variables (e.g. stock size of a population) or slow-changing variables (e.g. organic matter content in lakes) (Carpenter et al., 2001; Carpenter and Brock, 2006; Sirota et al., 2013).

(11)

3 1.2 Resilience and stability in bacterial communities

Over the past decades microbiological methods have been developed that make it possible to analyze the abundance, evenness, distribution and functionality of bacterial taxa in communities (Madigan, 2012). This methodological advance enabled investigation of bacterial community resilience and resistance over time and under the influence of external stresses.

Bacterial communities pose a good model system for observing actual transitions and changes in a complex system due to fast turnover rates and the option to observe changes under controlled conditions (e.g. Sirota et al, 2013). However, bacterial communities are very sensitive to alterations. Changes and reactions to stresses occur quickly so that current temporal resolution data might not be sufficient to detect these changes, even though different regimes might exist. Whether specific environmental disturbances or a combination thereof can cause a bacterial community to shift into a new alternative stable regime, remains to be seen (Fig. 1).

Figure 1: Two-dimensional ball-in-cup model, which is commonly used for visualizing alternative stable states.

The "ball" represents a system or community and the

"cups" the environments in which the system can be placed in. Alternative stable states in microbial communities have not been found yet. Adapted from Beisner et al., 2003.

Fuhrman and his colleagues (2015) concluded in a review on aquatic community dynamics that microbial communities are in a perpetual state of change, thus implying that stable states in such assemblages never exist. However, it has been shown that there seems to be a base bacterial community that constantly changes due to outside influences but is stable over longer time periods (Shade et al., 2012a). These reoccurring patterns can be used to predict abiotic conditions (e.g. Eiler et al., 2012; Fuhrman et al., 2006) since the composition of the base community seems to be governed by external factors such as seasonality or the type of carbon source available (e.g. Eiler et al., 2012; Shade et al., 2012a). Despite showing a pronounced seasonal and temporal pattern in community succession, Bizic-Ionescu and

(12)

4 her colleagues (2014) also identified that short-lived bacterial blooms and drastic shifts in community composition can occur in lakes. These blooms, that last only a couple of weeks at the most, can cause altered ecosystem functioning for short periods of times.

Functional resilience in bacterial communities is ensured due to high species diversity (Strickland et al., 2009). A large share of bacterial community members exists in a dormant stage that can potentially be reactivated once environmental conditions change (Lennon and Jones, 2011; Pedrós-Alió, 2012). These so-called seed banks foster the resilience and resistance of a community and make it possible to adapt to new environmental conditions by increasing redundancy in disturbed surroundings due to similar functional performance of the single community members (Shade et al., 2012a). Under stress, certain bacterial members can possibly disappear from a community due to maladaptation to new environmental conditions. In turn, formerly inactive members can be reactivated as they encounter more favorable conditions. Higher species diversity in bacterial communities was also found to improve ecosystem functioning (Wohl et al., 2004).

1.3 Influence of temperature on bacterial communities

Average temperatures in lakes have been shown to increase in concert with rising air temperatures caused through climate change all over Europe (Smith et al., 2008).

Temperature influences on lake ecosystems include shifted patterns of seasonality, changes in ice-cover duration, as well as changed stratification regimes.

It has been shown that temperature changes can have unexpected and profound functional effects on a biological community (Petchey et al., 1999). Some of these effects could not have been foreseen by a mere straight forward relation between physiology and temperature as many ecosystem responses are non-linear (Petchey et al., 1999;

Smith et al., 2008). Due to this complex relation, various ways of how temperature affects BCC have been found in the past.

In a study of 30 Wisconsin lakes, Yanarell and Triplett (2005) found that temperature was not amongst the main factors of structuring bacterial community composition if compared to for example pH or turbidity. However, since community structure changes with season and therefore with differing temperature regimes, it seems to be a significant factor on larger timescales (Eiler et al., 2012; Yannarell and Triplett, 2005). It has also been found that bacterial diversity is strongly influenced and positively correlated with temperature (Fuhrman et al., 2008; Yannarell and Triplett, 2004). Rather than influencing bacterial communities directly, temperature seems to determine community composition and activity through more indirect effects. Temperature can influence certain enzyme activities or rates of metabolite exchange amongst bacteria (e.g. Shade et al., 2012a), which in turn can lead to compositional changes by for example not providing sufficient substrate for a certain community member anymore.

(13)

5 1.4 Influence of salinity on bacterial communities

Rising temperatures, as the ones predicted by future climate change scenarios, were also shown to cause drier grounds and a salinisation of freshwater ecosystems through various processes (Moss et al., 2009; Nielsen et al., 2003; Smith et al., 2008). Likewise, changes in watershed morphology through construction work, waterlogging, dumping of saline and polluted mining water, and other physical alterations of a freshwater system can cause water conditions to become more saline within a short period of time (e.g. Policht- Latawiec, 2014; Singh and Afroz, 1985; Weston et al., 2006). It is therefore important to investigate the influence of salinity on freshwater bacterial communities because weather conditions in several parts of the world will become drier in the future and industrial development and landscape alterations will increase with a future expanding human society (e.g. Smith et al., 2008, Moss et al., 2009). Swedish climate is expected to become wetter, especially through increased precipitation in the winter time (Andréasson et al., 2004).

Summers on the other side are predicted to get drier and hotter, potentially leading to increased salinity in lakes.

Microbial lineages seem to have problems crossing salinity boundaries such as the marine- freshwater boundary. Salinity influences metabolic pathways, increases energetic cost and most likely leads to a different taxa distribution within microbial communities when conditions become more saline (Logares et al., 2009; Oren, 2001). It has however been found that freshwater bacterial communities harbor a share of marine bacterial taxa in their seed banks which can become more dominant when environmental conditions turn more advantageous for them (Comte et al., 2014; Comte and del Giorgio, 2010).

The exact way salinity influences BCC has not been determined yet since other factors such as the composition of organic matter in the water and positioning of the water body in the landscape create combined effects that exert their influence on a lake ecosystem (Blenckner, 2005). Nonetheless, most studies investigating a salinity influence on BCC came to the conclusion that it plays an important role in structuring the composition and functioning of a community (Bouvier and del Giorgio, 2002). Lozupone and Knight (2007) even suggested that "the major environmental determinant of microbial community composition is salinity rather than extremes in temperature, pH, or other physical and chemical factors […]". Langenheder et al. (2003) showed that even small changes in salinity can have profound direct and indirect effects on community composition and their performance, for example the degradation of dissolved organic carbon (DOC). All these findings suggest that it is of utmost importance to know how bacterial communities behave under increased stress regimes and what happens when salinity stress is combined with other stresses such as increased temperatures.

(14)

6 1.5 Project description

In this study we investigated the resilience and resistance of a lake bacterial community under various combinations of external stresses. The aim was to assess whether bacterial community composition (BCC) would profoundly change over time and if potential changes encompass shifts in bacterial productivity (BP) or ecto-enzyme activity. Additionally we investigated how this would affect the pool of fluorescent dissolved organic matter (DOM) that is utilized by bacterial communities. Stress was exerted in the form of gradual temperature changes (between 15 - 30 °C), temperature shocks (5°C above the current ambient temperature for 24 hours), as well as altered salinity of 3 psu. These stresses were assumed to be a significant disturbance to the bacterial community from Lake Ekoln and were chosen to simulate predicted increasing temperatures and heat waves, as well as increased salinities as a consequence of higher evaporation rates.

We wanted to observe whether a community exposed to higher stress (magnitude and frequency) would be less stable and change its functional performance more than less stressed communities. We were furthermore interested in estimating structural changes in BCC and whether functional responses would occur gradually or if they would show drastic shifts. Moreover, we investigated whether BCC would return to its initial state after environmental conditions were restored.

1.6 Hypotheses

We hypothesized that the more stresses a bacterial community is exposed to, the less resilient and stable it will be in terms of its community structure and its functional responses (bacterial productivity and ecto-enzyme activity). We also hypothesized that communities exposed to higher stress will exhibit a community structure that changed more from initial BCC than those that have been exposed to less stress. In addition we hypothesized that salinity will pose a major disturbance to a freshwater-adapted bacterial community.

We furthermore hypothesized that changes in bacterial community structure will cause and are correlated with changes in functionality, as well as changes in DOM patterns.

2. Materials & Methods

2.1 Sample site & sample collection

Lake water for this study has been taken from Lake Ekoln (59.76°N 17.58° E), near Uppsala, Sweden (Fig. 2), on September 9th, 2014 and on September 22nd, 2014. 20 L and 25 L canisters for sample collection were prepared by acid washing them with a 1M HCl solution, to remove residual carbon compounds, followed by rinsing them 3 times with MilliQ Water.

(15)

7

Figure 2: Map of Sweden & Lake Ekoln, yellow arrow indicates sampling site at 59.76°N 17.58° E. Source of the satellite image: LandsatLook Viewer (Data available from the U.S. Geological Survey).

Water from the first sampling date was prepared for being used as a replacement medium during the experiment. This water has been filtered twice with the help of a Masterflex L/S, Digital Modular and Pump Drive (Cole-Parmer Instrument Company, USA), first through a 3 µm filter, followed by a filter step through an 0.2 µm filter (Supor-200, 142mm diameter, PALL Life Sciences, USA) to remove bacteria and other organisms. The filtered water has been furthermore autoclaved with a Steam Sterilizer OT 90L (Nüve, Turkey), and stored in darkness at 4°C until usage. Before the replacement medium could be added to the treatments the pH had to be adjusted because it changed from 7.9 to 8.9 due to the autoclaving process. Medium water pH was adjusted with 0.1M HCl.

The water from the second sampling day harbored the bacterial community to be investigated. This water was filtered three times through a 20 µm filter net to exclude bigger plankton particles and predators. On September 22nd, ambient surface temperature in the lake was 15.1°C at the time of sampling. The water was stored in a regulated temperature room at 15°C until the beginning of the experiment on September 25th, 2015.

2.2 Treatments & experimental process

At the start of the experiment, 18 1L-blue cap glass bottles were filled with 950 ml of the sampled water. The experiment had 6 treatments with biological triplicates which were held in darkness throughout the duration of the experiment. The details for each treatment are stated in Table 1 and the temperature set-up can be seen in Figure 3.

(16)

8

Table 1: Descriptions of each treatment during the experiment

1 - Control incubated at 15°C throughout the

experiment

2 - Control Salinity incubated at 15°C throughout the experiment but with salinity of 3 psu 3 - Gradual Temperature Increase temperature constantly increased by 1°C

every 24 hours (±2 hours), starting at 15°C until 30°C, then constantly decreased by 1°C every 24 hours (±2 hours) until 15°C

4 - Gradual Temperature Increase Salinity as Treatment 3 but with salinity of 3 psu

5 - Shocks temperature constantly increased by 1°C

every 24 hours (±2 hours), starting at 15°C until 30°C, every third day treatments were exposed to 24-hour-temperature shocks (5°C higher than the current temperature);

after the shock, treatments were exposed to same temperatures as Treatments 3 & 4;

after the 5th shock temperature was

constantly decreased from 30°C by 1°C every 24 hours (±2 hours) without any additional temperature shocks until 15°C

6 - Shocks Salinity as Treatment 5 but with a salinity of 3 psu

Figure 3: Temperature set-ups in the 6 different treatments. a) Temperature in treatments 1 and 2 ("Control" and

"Control Salinity") was constant at 15°C throughout the experiment. b) Temperature in treatments 3 and 4 ("Gradual Temperature Increase" and "Gradual Temperature Increase Salinity") was increased from 15°C by 1°C each day until 30°C and then decreased by 1°C until 15°C again. c) Temperature in treatments 5 and 6 ("Shocks" and "Shocks Salinity") was increased from 15°C by 1°C each day until 30°C and then decreased by 1°C until 15°C again. Additionally the treatments were exposed to 5 temperature shocks.

(17)

9 On any given sampling day before a temperature shock and during the second half of the experiment (see appendix for sampling scheme), 150 ml were extracted from each treatment for further community analysis. The extracted water was replaced with 150 ml replacement medium. 20 ml of the extracted 150 ml were taken out with a syringe and filtered through a 0.2 µm GF/C-Glass microfiber filter (GE Healthcare Life Sciences Whatman, UK) and stored in 20 ml vials in darkness at room temperature. These 20 ml were kept for DOM-spectrometric analysis. The remaining 130 ml were used for bacterial productivity and enzyme activity measurements, flow cytometric measurements and T-RFLP analysis of community composition.

On sampling days after a temperature shock, 150 ml were taken out of the bottles for analysis of community functionalities, however no sample water was filtered for DOM and T- RFLP analysis.

2.3 Bacterial functionality 2.3.1 Bacterial productivity (BP)

Measurements of bacterial productivity give an indication of the protein synthesis of a bacterial community, and indirectly hint towards the carbon demand (Smith and Azam, 1992). Bacterial production was measured throughout the experiment on sampling days by measuring the communities' incorporation of radioactively labeled leucine with the method of Smith & Azam (1992).

For the analysis, 1.7 ml of sample water were extracted from each bottle on sampling days.

L-[4, 5-3H] Leucine (161 Ci mmol^-1, Perkin Elmer) was diluted with unlabeled L-Leucine (Sigma, St Louis, MO, USA) to a 15% hot leucine mixture. Hot leucine mixture (27 µL) was added to the 1.7 ml of water sample (final concentration 100 nM), and incubated for an hour. The incubation was then stopped with 90 µL of 100% TCA. The productivity of the bacteria was measured with a TRI-CARB 2100TR Liquid Scintillation Analyzer (Packard BioScience Company, USA).

Productivity per cell has then been calculated with the formula:

[ ] =

1.7 ÷

(18)

10 2.3.2 Enzyme activities

Enzymatic activities of ecto-enzymes in aquatic bacteria give an indication of the organisms' activity and growth (Overbeck and Chróst, 1990). In other words, it is an indicator of how well the bacteria are performing under the given environmental conditions. Enzyme activities were measured for the extracellular enzymes ß-glucosidase and cellobiohydrolase.

A final concentration of 0.6 mM methylumbelliferone (MUF)-linked artificial substrate (Sigma-Aldrich, USA) was used for the measurement of the enzyme activities. These substrates were added to three replicated samples of 1 mL for each treatment and incubated in the dark at 20°C for 2.5 hours. At the end of the incubation period, glycine buffer (pH 10.4) was added (1 : 1 v : v) to stop the reaction, and after loading the samples onto a 96-hole-blackwell plate, fluorescence for MUF was measured at 365/455 nm excitation/emission with an Ultra 384 plate reader (TECAN, Switzerland).

Enzyme activities per cell [nmol MUF h^-1 cell^-1] were calculated by subtracting the blank fluorescence values from the measured sample fluorescence values to correct for abiotic hydrolysis of the substrate or fluorescent substances in the medium (Ylla et al., 2013). MUF- fluorescence was then calculated from the standard curve and adjusted for the incubation time. Activity per cell was calculated by dividing the activity per ml by the number of cells per ml.

2.4 DOM-Spectrometry

Heterotrophic bacteria, like the ones investigated in this study, are major competitors for DOM and alter it through their metabolic activity (Overbeck & Chróst, 1990). In order to assess how bacteria influence the DOM of water, we measured changes in DOM quality over time using absorbance and fluorescence spectrometry techniques.

We obtained Excitation-Emission Matrices (EEMs) from fluorescence measurements. This method makes it possible to observe changes in DOM by measuring the absorbance of a water sample at different wavelengths and tracking the change in absorbance and emission over time. This qualitative method can pinpoint net changes for fluorescently active compounds without identifying the actual molecules that cause those changes (Kothawala et al., 2012).

Spectrometric analysis was done with a FluoroMax-4 Spectrofluorometer (JobinYvon Technology, HORIBA Scientific, Japan) and a Lambda 35 UV/VIS Spectrometer (PerkinElmer, USA) using a 1 cm quartz cuvette. 3D-EEMs were then created for each sample. EEM scans were run at 5 nm excitation increments between 250–445 nm, and at 4 nm emission increments between 300–600 nm. Manufacturer supplied correction factors were applied to correct excitation and emission intensities for instrument-specific biases. A water blank (MilliQ) EEM recorded under the same conditions was subtracted from each sample to

(19)

11 eliminate Raman scattering; the area underneath the water Raman scan was calculated and used to normalize all sample intensities (Lawaetz and Stedmon, 2009). Absorbance spectra of the samples were taken at a wavelength of 200 - 800 nm and used to correct the EEMs for inner filter effects according to Kothawala et al. (2014). These corrections were applied using the FDOMcorrect toolbox for MATLAB (Mathworks, Natick, MA, USA) following Murphy et al. (2010). The fluorescence intensities of the main fluorescent peaks A, B, C, M, and T, associated with DOM (Coble, 1996; Parlanti et al., 2000) were measured (Locations of the 5 main fluorescence peaks in an EEM can be seen in two example spectra in the appendix). We discarded the application of parallel factor analysis (Stedmon et al., 2003) because of the experimental characteristics of the data set, which impeded the validation of the model. We also calculated several spectral indexes, including the fluorescence index (Cory and McKnight, 2005), humification index (Zsolnay et al., 1999), biological index (Huguet et al., 2009), spectral slopes (Helms et al., 2008), the specific ultra-violet absorbance (Weishaar et al., 2003) and the absorbance of colored DOM at 350 and 440 nm wavelengths (Fasching et al., 2014).

2.5 Community structure 2.5.1 Flow cytometry

Flow cytometric analysis enables one to reliably count cell numbers of a given sample (e.g. del Giorgio et al., 1996) and enables to follow changes in cell abundances and physiological community structure changes over time (e.g. Kleinsteuber et al., 2006;

Koch et al., 2013).

Cell abundances have been obtained through flow cytometric analysis over the duration of the full experiment on every sampling day. A CyFlow space flow cytometer (Partec, Germany) has been used to obtain the counting data. For the analysis, 1.5 ml of sample water have been extracted from the treatments and fixed with sterilized formaldehyde (final concentration 2%) and stored at 4°C until the end of the experiment. When all necessary samples were collected, 190 µL of each sample have been stained with 10 µL Syto13 (1.25 M , Molecular Probes, Invitrogen, USA) dye and loaded onto a 96-well plate. Of this 200 µL mix, 50 µL were then taken up by the flow cytometer to perform the counting.

The total number of cells per ml were then calculated with the formula:

= 50 ∙ 1000

45.2

(20)

12 2.5.2 T-RFLP analysis

Terminal Restriction Fluorescence Length Polymorphism (T-RFLP) analysis is a method to retrieve a genomic fingerprint from a bacterial community by amplifying the 16S rRNA gene.

In T-RFLP, bacterial DNA is extracted from a sample, the 16S rRNA gene amplified and then digested by a restriction enzyme which creates DNA fragments of unequal length. It is assumed that each of these DNA fragments represents one bacterial operational taxonomic unit (OTU), thus giving information about the abundance of specific bacteria in a sample at a given time point (Liu et al., 1997).

On sampling days, around 130 ml of sample water from each sample have been collected for the analysis through filtration (Supor-200, 0.2 µm, 47 mm Membrane Filter, PALL Life Sciences, USA). The PowerSoil DNA Isolation Kit by MoBio Laboratories, Inc. (USA) was used to extract DNA from the collected samples. 25 µL DNA extract was used as template for PCR amplification of the 16S rRNA genes. The PCR mix furthermore included the EUB-8-forward primer, labeled with hexachlorofluorescein (HEX) (Thermo Scientific, USA) (final conc.

200 nM) and unlabelled 519-reverse primer (Invitrogen, USA) (final conc. 200 nM), BIOTAQ DNA Polymerase (BIOLINE,UK) (final conc. 0.05 u/µL), dNTP Mix (Thermo Scientific, USA) (final conc. 0.04 mM each dNTP), Q5 Reaction Buffer (New England Biolabs, USA) (final conc.

0.05x), 10x NH4 reaction buffer (BIOLONE, UK) (final conc. 1x), 50 mM MgCl2 solution (BIOLINE, UK) (final conc. 1.5 mM).

Thermo cycling was carried out with a MyGene MG96+ (LongGene Scientific Instruments Co., Ltd., China) protocol and the PCR conditions were an initial denaturation at 94°C for 3 min;

25 - 35 cycles of 94°C for 30 s, 52°C for 30 s, 72°C for 45 s; and a final extension at 72°C for 7 min. The PCR products were purified with the QIAquick PCR Purification Kit (Qiagen, Hilden, Germany) and later quantified with the Quant-iT PicoGreen dsDNA Assay Kit (Life Technologies, USA) according to a protocol by Logares & Feng (2010). Samples that had less than 4 ng/µL were speed vacuumed with a Scan Speed 32 (SCANVAC, LABOGENE, Denmark) connected to a MZ 2C NT Vacuum Pump (Vacuubrand, Germany) to increase the DNA concentration. The restriction digestion was performed after a modified T-RFLP protocol by Logue (2009). The restriction enzyme used was HaeIII (New England Biolabs, USA) and samples were digested for 18 hours at 37°C. The digested samples were then handed over to the SciLife Lab Uppsala for analysis of the digested DNA fragments. Analysis of the raw data has been conducted with GeneMarker (v. 2.6.4, Softgenetics, USA) and binning has been done manually in Windows Excel 2010 (Microsoft, 2010).

(21)

13 2.6 Statistical analysis

All statistical analysis was carried out with the statistical program R (R Core Team, 2014) and with Windows Excel, 2010 (Microsoft, 2010).

Linear models were used to analyze the overall trends of the time series for cell abundances, BPs and ecto-enzyme activities in the first and the second half of the experiment (Chambers, 1992). The reason for this was that although the progression of the individual time series displayed a lot of variation, general patterns of curve progressions could be found common to all treatments within one measured parameter.

NMDS with a Bray-Curtis-dissimilarity was performed to assess differences in BCC amongst the different treatments and over time (Faith et al., 1987). Environmental fitting of the measured DOM parameters, community functionalities, time, salinity, temperature and cell abundances has been conducted to find significant linear trends relating to BCC with the envfit function from the R package "vegan" (Oksanen, 2014).

Permutational Multivariate Analysis of Variance (PERMANOVA) was used to test for significant differences amongst non-saline and saline communities and to test if BCC changed significantly from the starting community on the first day of the experiment (Legendre and Anderson, 1999).

Rank clocks were created to display the change of abundant OTUs over time (Batty, 2006;

Collins et al., 2008). Abundant OTUs were defined as those that comprised 10% or more of the community on at least one day of the experiment in any of treatments (making them the 5% most abundant OTUs in the experiment). The Evar-Index (Smith and Wilson, 1996) and Shannon-Index (Spellerberg and Fedor, 2003) were calculated to observe changes in the evenness and abundance of OTUs in the six treatments over time.

(22)

14

3. Results

3.1 Bacterial productivity

In the first half of the experiment, all treatments showed higher BP per cell with a higher degree of variation than in the second half (Fig. 4). Starting BP was low in all treatments with initial BP values of 0.005 to 0.01 DPM/cell. After a short lag phase that lasted from day 1 to 3, BP values rose in all treatments continuously. Depending on the treatment this rise in BP lasted until day 4 to 10. After this initial rise, BP values in the treatments declined or leveled off with different degrees of variation until day 16. BP values in the second half were on average lower and more consistent and all treatments had a negative trend in their BP values between days 16 and 32.

Highest productivity values were reached in the "Control Salinity" treatment with 0.066 DPM/cell on day 10, after which BP continuously decreased until day 32. Rises and declines in BP in the two "Shock" treatments during the temperature shocks could not have been caused by the shocks since similar progression for the same time periods could be noted in the other treatments.

Coefficients of variation and regression coefficients of the linear models for the first and second half of the experiment can be found in the appendix.

(23)

15

Figure 4: Bacterial productivity per cell in the six treatments over time. Linear models were fitted onto the curves for the time between day 1 and day 16, as well as between day 16 and day 32 coinciding with the periods when stress was increased and decreased respectively, but also when sampling regime was more and less frequent. Blue line at day 16 in both "Gradual Temperature Increase" and both "Shocks" treatments indicates the time point after which temperature was stepwise decreased. Grey bars in the "Shocks" treatment plots indicate the periods of temperature shocks.

(24)

16 3.2 Enzyme activities

3.2.1 β-Glucosidase activity

All Treatments started out with their lowest β-glucosidase activities per cell on day 1. Apart from the first day, activity in all treatments differed depending on the stress regime (Fig. 5).

Fitting linear models onto the curves showed that in the "Control" treatment, enzyme activity was constantly rising throughout the experiment. In the "Control Salinity" treatment activity was on average the lowest of all treatments. Here, activity peaked on day 13, at around 2^-6 nmol MUF h^-1 cell^-1, and then slowly decreased towards initial activities until the end of the experiment. Both the non-saline and saline "Gradual Temperature Increase"

treatments showed similar magnitude in enzyme activities. Linear regression models showed a positive trend in enzyme activity during the time of gradual temperature increases, and a negative trend after temperatures were lowered again. Both "Shocks" treatments showed an overall pattern of constantly increasing activity in the first half of the experiment.

However, the "Shocks Salinity" treatment had lower average activities than the non-saline shock treatment. The linear regression models showed that enzyme activities in the two

"Shocks" treatments stayed relatively constant after changing the sampling regime and lowering the temperature. The first temperature shock between days 3 and 4 seemed to be a major disturbance to the two communities since β-glucosidase activities dropped severely in that 24 hour period, which did not happen in any of the treatments that did not receive a temperature shock. Especially the activity of the non-saline shock treatment was strongly decreased after having an initial spur in activity from days 1 to 3. What can be furthermore inferred from the time series (Fig. 5) is that especially in the two "Shocks" treatments, standard deviations in β-glucosidase activity were larger in the second half of the experiment than in the first half.

(25)

17

Figure 5: ß-glucosidase activity per cell in the six treatments over time. Linear models were fitted onto the curves for the time between day 1 and day 16, as well as between day 16 and day 32 coinciding with the periods when stress was increased and decreased respectively, but also when sampling regime was more and less frequent. Blue line at day 16 in both "Gradual Temperature Increase" and both "Shocks" treatments indicates the time point after which temperature was stepwise decreased. Grey bars in the "Shock" treatment plots indicate the periods of temperature shocks.

(26)

18 3.2.2 Cellobiohydrolase activity

For cellobiohydrolase activities per cell, the curve progressions over time were very similar to the β-glucosidase activities per cell and significantly correlated with them (see Fig. 5 & 6, and Table 12 in the appendix). The most prominent differences were that the linear regression in the "Control" treatment showed a negative trend in the first half of the experiment and that the linear regression of the non-saline "Gradual Temperature Increase"

treatment showed a positive trend throughout the whole experiment. Furthermore, cellobiohydrolase activity did not decrease in the "Shocks Salinity" treatment during the first temperature shock.

(27)

19

Figure 6: Cellobiohydrolase activity per cell in the six treatments over time. Linear models were fitted onto the curves for the time between day 1 and day 16, as well as between day 16 and day 32 coinciding with the periods when stress was increased and decreased respectively, but also when sampling regime was more and less frequent. Blue line at day 16 in both "Gradual Temperature Increase" and both "Shocks" treatments indicates the time point after which temperature was stepwise decreased. Grey bars in the "Shock" treatment plots indicate the periods of temperature shocks.

(28)

20 3.3 DOM-Spectrometry

DOM quality did not change much during the experiment. Changes in EEMs were existent but not pronounced over time (Fig. 7). However, significant linear trends (p < 0.01) for several of the measured parameters could be observed when fitted on top of an NMDS of BCC from the different treatments over time (Fig. 10, see also supplementary material on DOM changes over time in the appendix).

A more detailed description of the various measured DOM parameters' progression during the experiment and what they signify can be found in the appendix.

(29)

21

Figure 7: EEMs of the water samples in the different treatments from the first day and the last day of the experiment.

Color scale is from 0 (dark blue) to 3 (dark red) Raman Units.

(30)

22 3.4 Community structure

3.4.1 Flow cytometry

Cell abundances in all treatments showed the general pattern of being the highest at the beginning of the experiment and then declining over time (Fig. 8). All curves had similar magnitude and progression under the different temperature and salinity conditions regardless of whether the temperature was stable ("Control" treatments) or slowly increased with additional temperature shocks ("Shocks" treatments).

In all the non-saline treatments, cell numbers decreased drastically within the first three days and then recovered until day 7. The saline treatments did not show an immediate decline in cell numbers and kept high cell abundances at the beginning, dropping however sharply after day 4. After the recovery in the non-saline treatments and the drop after day 4 in the saline ones, cell numbers decreased continuously until day 16 in all treatments, the half-time point of the experiment. After day 16 all treatments kept low cell numbers until the remainder of the experiment. Only the "Control Salinity" treatment showed signs of recovering cell numbers after reaching the lowest cell abundance on day 13 (Fig. 8).

Fitting linear models onto the cell abundance time series showed that for all three saline treatments, a positive trend in cell numbers existed in the second half of the experiment.

However, in the "Gradual Increase Salinity" and "Shocks Salinity" treatments, the regression coefficient of the linear models is close to 0 (see appendix) and cell numbers are almost stable throughout days 16 - 32.

Coefficients of variation and regression coefficients of the linear models for the first and second half of the experiment can be found in the appendix. Furthermore the appendix contains correlation tables presenting the correlations of cell abundances with total BP and total ecto-enzyme activities as well as the correlations of cell abundances with BP per cell and ecto-enzyme activities per cell.

(31)

23

Figure 8: Cell abundances per ml in the six treatments over time. Linear models were fitted onto the curves for the time between day 1 and day 16, as well as between day 16 and day 32 coinciding with the periods when stress was increased and decreased respectively, but also when sampling regime was more and less frequent. Blue line at day 16 in both

"Gradual Temperature Increase" and both "Shocks" treatments indicates the time point after which temperature was stepwise decreased. Grey bars in the "Shock" treatment plots indicate the periods of temperature shocks.

(32)

24 3.4.2 T-RFLP analysis

The BCC in the different treatments changed throughout the duration of the experiment.

PERMANOVA showed that communities in the six treatments changed significantly from the starting community on day 1 (p < 0.001, with 999 permutations). The development of BCC over time was treatment-specific and depended on the type and amount of external stresses the communities were exposed to (Fig. 9).

On general, BCC of the non-saline treatments did not change as much from the initial community structure as the BCC in the saline treatments. Apart from the "Gradual Temperature Increase" treatment on day 32 and the non-saline "Shocks" treatment on day 25 and 32, BCC in non-saline treatments was always more similar to the starting community than in the saline treatments on any day from day 9 onwards (the second day on which T- RFLP profiles were obtained). PERMANOVA showed that the difference in BCC between non- saline and saline communities was significant (p <0.001, with 999 permutations). Salinity influenced communities changed more in their BCC between day 1 and day 9 than the three non-saline treatments. In the four treatments that underwent temperature changes, BCC did not show a tendency to return back to the initial community composition of day 1 after the temperatures had been decreased to 15°C again.

The community in the "Gradual Temperature Increase Salinity" treatment did barely change between day 15 and 25 (the 3rd and 4th days on which T-RFLP profiles were obtained). BCC during those two days was also very similar to the BCC of the "Shocks Salinity" treatment on day 15. On day 32, the last day of the experiment, the two "Shocks" treatments, which underwent the most complex combination of stresses, displayed a community structure that had changed more from the initial starting community than the BCC of the other four treatments on day 32.

OTU evenness (Evar-Index) and OTU diversity (Shannon-Index) of all OTUs displayed no clear trends relating to stress treatments or along a salinity/no-salinity divide. Although measures of evenness and diversity stayed relatively constant throughout the duration of the experiment, changes amongst OTUs in each treatment were highly dynamic. Tables summarizing Evar-Index and Shannon-Index values can be found in the appendix.

(33)

25

Figure 9: NMDS of BCC in the six treatments over time. Single profiles of each treatment (colored numbers) were combined for an average BCC per day and treatment. Average BCC from each day and treatment are written out and connected via colored lines. Numbers 2, 3, 4, and 5 behind the treatment names correspond to the 9th, 15th, 25th and 32nd day of the experiment respectively. Single profiles of the starting community from all treatments were averaged to obtain NMDS scores for a "Day 1" BCC.

3.4.3 Environmental fitting

Several significant trends (p < 0.01) relating to BCC could be identified when fitting DOM parameters, community functionalities and independent variables onto the NMDS of all obtained T-RFLP profiles (Fig. 10). A detailed list of p-values for the environmental fitting of each parameter can be found in the appendix.

The linear trends of SR and the fractions of Peak T and Peak B were closely related to each other, and opposite to the trends of the slope measure S350-400 and the relative fraction of Peak C. The trends in β-glucosidase and cellobiohydrolase activity per cell were also highly correlated and opposite to the measures of SUVA254, the relative fraction of Peak M, and the Peak A:C ratio. Time had nearly opposite trends to cell abundances and the A440

(34)

26 measure. The freshness index BIX and the A350 were found to be unrelated to time and opposing in each other's trend. The trend for salinity was determined by whether treatments were saline or not. Productivity per cell and the relative fraction of Peak A displayed a similar relation to BCC as the trend for salinity.

Figure 10: Environmental fitting of measured community functionalities, DOM parameters, time, salinity, and cell abundance Environmental fit based on identical results as Fig. 9. Significant linear trends (p < 0.01) fitted onto all the measured community profiles. Arrows that point in the same direction have a positive correlation. Arrows that point in opposite directions have a negative correlation, and arrows that stand at a 90° angle to each other have no correlation.

3.4.4 Rank clocks

In all treatments combined, there was a total of 13 OTUs that surpassed the 0.1 relative abundance mark on at least one day during the duration of the experiment (Fig. 11).

At the beginning of the experiment, OTUs 72, 83 and 84 were present and surpassed the relative abundance of 0.1 in all six treatments. Whereas OTU 72 stayed relatively abundant in all treatments throughout the duration of the whole experiment, abundances in OTUs 83 and 84 declined quickly after day 1 in all treatments. The "Control" treatment had the most stable dynamics considering its dominant OTUs and was dominated mainly by OTUs 72 and 75 in the first half of the experiment and mainly by OTU 75 in the second half. OTUs 151, 154 and 166 became only prominent in the saline treatments and surpassed the 0.1 relative

(35)

27 abundance mark always during the second half of the experiment. OTU 151 was present in all three saline treatments, whereas 166 appeared only in the "Control Salinity" treatment and 154 only in the "Shocks Salinity" treatment.

The dominant OTUs in the "Shocks" treatments never became as prominent as OTUs in other treatments with only OTU 75 and 73 surpassing the 0.2 relative abundance mark on only one day each. In contrast, in each of the other treatments more than one OTU surpassed the 0.2 relative abundance mark and for longer periods than just one sampling day. Additionally, it can be seen that treatments exposed to salinity stress and/or the temperature shocks had 7 - 9 OTUs surpassing the 0.1 relative abundance mark during the experiment, whereas in the "Control" and "Gradual Increase" treatments only 5 OTUs surpassed 0.1 relative abundance.

(36)

28

Figure 11: Rank clocks of dominant (>0.1 relative abundance) OTUs over time. The rank clocks show the increase and decrease of OTUs over time and have to be read like a clock: The starting and end point of the experiment (days 1 and 32 respectively) lie at "12 o'clock". The half-time of the experiment (day 16) lies at "6 o'clock". Direction of the progression is clock-wise. The numbers 10, 20 and 30 on the outer circle are the x-axis labels and correspond to days during the experiment.

(37)

29

4. Discussion

4.1 Bacterial functionality: BP and enzyme activities

Initially it had been hypothesized that more stresses would lead to less resilience in the community's functional responses. Throughout the experiment, BP and enzyme activities changed in the six treatments. However, no decrease below starting values, even under the highest stress conditions was found.

For BP, higher average productivities with higher degrees of variation (see appenidx for coefficients of variation in BP time series) were observed in all treatments in the first half of the experiment than in the second half, regardless of the exerted stresses (Fig. 4). It is possible that the frequency of the exchange of medium water played a role in the observed patterns. An explanation for this could be that after medium water exchanges, the fresh medium was rapidly colonized by fast-growing OTUs, since it constituted new living space.

These fast growing OTUs have higher pronounced productivity rates than slower growing OTUs (del Giorgio and Cole, 1998) leading to the observed high BP values. In the second half of the experiment, sampling frequency was lower (see sampling time table in the appendix).

Spikes in productivity as observed in the first half of the experiment could have been present after medium exchanges in the second half as well (Fig. 4), but were possibly not observed due to longer time spans between sample takings by 1 or 2 days. It could be that slower growing OTUs might already have outcompeted those fast-growing OTUs by the time new samples were taken, and those would then display lower average productivity values due to slower growth rates. Therefore the removal and exchange of medium water seems to have constituted a major disturbance to the communities which we had assumed to not be the case when only removing 150 ml (around 20 %) from each bottle on a sampling day.

Initially it was also expected that salt-stressed communities would display lower productivities than communities in the non-saline treatments. This was however not the case (Fig.4). This might be related to the fact that the large seed bank of freshwater bacterial communities harbors marine taxa that can be recruited and influence BCC when lake conditions become more saline (Comte et al., 2014). When these salinity-adapted bacteria flourish, similar metabolic rates as in a community that is dominated solely by freshwater taxa can be achieved (Hart et al., 1991; Hobbie, 1988). Therefore it is not surprising that even in salinity-stressed communities BP values had similar rates and ranges as the non- salinity influenced communities.

Langenheder et al. (2012) found that functional overcompensation, the increase in functionality during times of increased stress, can occur by dominating species under environmental change. In this study, fast rising enzyme activities could be observed in the

"Control" treatment and in the treatments exposed to temperatures above 15°C (Fig. 5 & 6).

A possible explanation for this could be that these higher enzyme activities arose in all the temperature stressed treatments due to a coupling of functional overcompensation by the community members and a boost of enzymatic activities due to increased temperature. This

Structural and Functional Changes in a Lake Bacterial Community Exposed to Multiple Stress Regimes