Limited response of a spring bloom community inoculated with filamentous cyanobacteria to elevated temperature and pCO₂

Experimental setup and sampling procedure

The outdoor experiment was conducted between 23 March and 4 April 2011. The water was collected at the Baltic Sea Proper monitoring station B1 on 17 March 2011 (58°49′N, 17°38′E), with a salinity of 6.6 and a sea surface temperature of ca. 1°C. The collected water was gently sieved through a 200-μm mesh to exclude large grazers. Cultures of the following diazotrophic filamentous cyanobacterial strains were obtained from the Kalmar Algae Collection, Linnaeus University, Kalmar, Sweden: Aphanizomenon sp. (A. Morren ex É. Bornet and C. Flauhault 1886‚ 1888‘), Nodularia spumigena (Mertens ex Bornet and Flauhault 1888) and Dolichospermum spp. ((Ralfs ex Bornet and Flauhault) P. Wacklin, L. Hoffmann and J. Komárek 2009) (KAC16, formerly known as Anabaena sp.), isolated from the Baltic Sea and kept at 10°C, were inoculated into the natural community. The cyanobacteria were pre-cultured for 1 week under nutrient replete conditions, 6°C and ca. 20 μmol photons m⁻² s⁻¹. The proportion of inoculated cyanobacteria to the natural community was based on in situ conditions at the nearby monitoring station B1 in June with a ratio of 1:10 between cyanobacteria and diatoms+dinoflagellates (from the database of the Swedish Meteorological and Hydrological Institute, for June 2011), hereby adding ca. 0.1 mm³ l⁻¹ of the three cyanobacteria species all together. Biovolume (mm³ l⁻¹) is recommended by HELCOM as a proxy for Baltic Sea phytoplankton biomass (Olenina et al. 2006).

After filtration and inoculation, the water was divided into 12×4-l Plexiglas aquaria. Triplicates of four treatments were established by combining ambient temperature (1°C) and elevated temperature (4°C) with ambient pCO₂ (390 μatm) and elevated pCO₂ (970 μatm) in an orthogonal design. The elevated temperature and CO₂ levels are according to the predictions by HELCOM (2013) and IPCC (2013) for the coming 100 years. The aquaria were randomly placed in two temperature controlled water baths, recorded with two loggers per basin (HOBO Pendant, Onset Computer Corporation, USA). The basins were exposed to sunlight and covered with green plastic mesh to reduce direct solar irradiance including ultraviolet radiation, mimicking photosynthetically active radiation (PAR, 400–700 nm) at 1 m water depth at station B1. PAR was measured with a LI-1000 datalogger equipped with a Li-Cor UWQ5201 PAR sensor (Li-COR, Lincoln, USA). A Li-COR radiometer equipped with a 2π PAR sensor was used to record irradiances under the mesh throughout the experiment.

The two pCO₂ treatments were established by connecting each aquarium with a tube, constantly providing synthetic air (AGA Gas, Linköping, Sweden). The gas was dispersed to the water by plastic air diffusers at a flow rate of ~15 ml min⁻¹. The aquaria were sealed with Plexiglas lids, where small holes were maintained for gas outlet to prevent backpressure build-up. In addition, each aquarium was provided with a submerged tube connected to an external syringe, which was used to transfer subsamples from the aquaria without opening the lids and thus disturbing the pCO₂ of the headspace (ca. 5 cm).

Each of the aquaria were subsampled for phytoplankton growth and species composition, bacterial abundance and production, dissolved inorganic nutrients, and carbonate system parameters at around 08:00 h on days 0, 3, 6, 9 and 12. On days 0 and 12, samples for photosynthetic pigment analysis and stoichiometry were collected. After analysis of in situ conditions, additional PO₄³⁻ of ca. 4 μm of was added at day 1 to avoid limitation. For each of the four treatments, one additional aquarium containing only 0.2-μm filtered seawater was sampled for pH, total alkalinity, temperature and inorganic nutrients, to control for non-organismal effects on the carbonate system.

Determination of the carbonate system

Total alkalinity (A_T) and pH were established according to Dickson et al. (2007). Samples for pH determination were 0.2-μm filtered and analyzed spectrophotometrically (Shimadzu UV-2100) at 19.0°C on the total scale (pH_T), using a 2-mm solution of m-cresol purple as indicator (Clayton et al. 1995). The pH of the indicator solution was measured and the perturbation of the sample pH by the addition of indicator was corrected according to Chierici et al. (1999).

A_T was determined using a potentiometric titration with 0.05 m HCl (Haraldsson et al. 1997). The equivalence point was estimated using Gran evaluation (Gran 1952) and manual titration with a dosimat (876 Dosimat plus, Metrohm) and a pH sensitive electrode (Aquatrode with Pt1000 Thermistor, Metrohm). The accuracy was tested using certified reference material (CRM) batch 97 supplied by A. Dickson, San Diego, USA (Dickson et al. 2007), with an average value of 2206±5.3 μmol kg⁻¹.

Calculations of pCO₂, [HCO₃⁻], [CO₃²⁻] and [CO₂] were performed with the chemical speciation model CO2SYS (Pierrot et al. 2006). Measured data for pH at 19.0°C, A_T, [PO₄], [Si(OH)₄], salinity, and temperature were used in the calculation together with the constant for HSO₄⁻ determined by Dickson (1990) and the total hydrogen ion scale (pH_T). The dissociation constants used for carbonic acid (K₁ and K₂) were determined as described by Mehrbach et al. (1973) and refitted by Dickson and Millero (1987).

Stoichiometry and analyses of dissolved inorganic nutrients

To determine the concentration of particulate organic carbon (POC) and nitrogen (particulate organic nitrogen, PON), 150-ml samples were filtered onto pre-combusted (400°C, 4 h) GF/C filters (25 mm). The filters were frozen at −20°C, freeze-dried, ground (MM301, Retsch, Haan, Germany), and analyzed on an EA 1108 (CHNS-O, Fisons Instruments, Thermo Fisher Scientific, Waltham, MA, USA) applying 2,5-bis-[5-ert.-butyl-bensoaxzol-2-yl]-thiophen as standard. To determine the concentration of dissolved inorganic NO₃⁻+NO₂⁻ (dissolved inorganic nitrogen, DIN), PO₄³⁻ (dissolved inorganic phosphorus, DIP) and Si (dissolved inorganic silicate, DISi), 20 ml of sample was filtered through a 0.2-μm pore-size filter and stored at −80°C, until analyzed within 3 weeks on an autoanalyzer (Grasshoff et al. 1999).

Photosynthetic activity

Photosynthetic activity was estimated by variable chlorophyll fluorescence measurements in photosystem II (PSII), with a WATER-PAM (Pulse amplitude modulated fluorometer) equipped with red light emitting diodes (Walz Mess- und Regeltechnik, Effeltrich, Germany). Samples were dark adapted at in situ temperature for a minimum of 20 min before measurements. Minimum fluorescence (F₀′) was determined by applying a low level of light and the maximum fluorescence (F_m′) by exposing the sample to a short saturation pulse of light (>4000 μmol photons m⁻² s⁻¹ for 0.6 s). Variable fluorescence (F_v=F_m′–F₀′) and maximum quantum yield (F_v/F_m) were determined for all samples. Rapid light curves (RLCs) were performed by measurement of ΔF/F_m′ ((F_m′−F₀)/F_m′) of quasi-adapted (15 s) cells at nine levels of actinic light (0, 51, 76, 110, 165, 250, 368, 571 and 851 μmol photons m⁻² s⁻¹) in the emitter-detector unit. Relative electron transport rate (rETR) was calculated by effective quantum yield (ΔF/F_m′) multiplied by PAR irradiance. Photosynthetic parameters (rETR_max, E_k, and α_PSII) were calculated according to Jassby and Platt (1976), fitted by the Nelder-Mead method in the R package phytotools (Silsbe and Malkin 2015, R Core Team 2016).

Microplanktonic composition and specific growth rates

Samples (50 ml) were preserved with alkaline Lugol’s solution, stored in darkness, and analyzed within 3 months using the Utermöhl method according to HELCOM (2014a). Half of the chamber was viewed at 100× magnification (Zeiss Axiovert 40CFL, Germany) and organisms >30 μm were counted and identified to species level or group (e.g. pennate diatom). Lengths and widths were measured and biovolume (mm³ l⁻¹) was calculated according to Hillebrand et al. (1999). At 200× and 400× magnification, a transect diagonally across the chamber bottom (1000 and 500 μm wide, respectively) was analyzed and organisms >8 μm were counted. The biovolume (mm³ l⁻¹) of organisms >8 μm was thereafter converted into POC based on Menden-Deuer and Lessard (2000). The calculated POC based on microscopy analysis was compared to the measured POC analyzed on GF/C filters, in order to estimate the fraction of organisms or organic material not identifiable by microscopy (ca. 1–8 μm). Specific growth rate was calculated for diatoms, dinoflagellates, and each cyanobacterial species separately, for days 6–12. The specific growth rate (μ d⁻¹)=(ln DB–ln DA)/(tB–tA), where DA is the biovolume (mm³ l⁻¹) at the first day of a period and DB the biovolume at the end of the period, and tA is day A and tB is day B.

Heterotrophic bacterial abundance and productivity

1.5 ml of sample were fixed with 1% glutaraldehyde and stored at −80°C. The samples were thawed, stained with SybrGreen (Invitrogen) and counted on a FASCanto II flow cytometer (Becton Dickinson; Gasol and del Giorgio 2000) using fluorescent beads (True counts, Becton Dickinson) to calibrate the flow rate. Bacterial productivity was measured by [³H]-thymidine incorporation (Fuhrman and Azam 1982) as modified for microcentrifugation by Smith and Azam (1992). Duplicate 1.7 ml aliquots were incubated with [methyl-³H]-thymidine (20 nm final conc., GE Healthcare) in sterile 2.0-ml capacity polypropylene tubes for ca. 1 h at in situ temperature. Duplicate blanks (killed control) were prepared by adding 5% trichloracetic acid prior to the addition of isotopes. Productivity was calculated using 1.4×10¹⁸ cells mole⁻¹ thymidine incorporated (average calculated from published Baltic Sea data, SE=0.1×10¹⁸ cells mole⁻¹ thymidine, n=73; HELCOM 2014b) and 20 fg carbon cell⁻¹ (Lee and Fuhrman 1987).

Statistical analyses

All data were analyzed for effects of elevated pCO₂ and temperature by linear mixed effects (LME) modeling using the package nlme in R (R Core Team 2016, Pinheiro et al. 2017). All models were fitted using restricted maximum likelihood, where sampling day was included as a fixed, categorical factor, and aquarium was included as a random factor (random intercept) to account for repeated measurement over time. Response variables were biovolume of the cyanobacterial species, diatoms, dinoflagellates, ciliates, POC, PON, heterotrophic bacterial abundance and productivity and photosynthetic activity (F_v/F_m, rETR_max, E_k, and α_PSII). Multiple comparisons of significant factors were performed by Tukey’s HSD test using the lsmeans package (Lenth 2016) in R. Parametric assumptions were visually verified using boxplots and Q-Q plots (Quinn and Keough 2002) and, if necessary, data were either log (POC, PON and POC:PON) or square root [cell-specific bacterial productivity (CSP), rETR_max, E_k] transformed. In addition, regression analysis was performed between bacterial abundance and biovolume of phytoplankton groups, POC:PON, POC and PON, using Pearson correlation for day 12 (SPSS). Significance was set to p<0.05. Additional multivariate analyses were performed to detect differences between treatments or time on species level and are included as Supplementary material.

Temperature and the carbonate system

As desired, our manipulation generated ambient and elevated temperature conditions from day 1 onwards (Figure 1A). The average pH was 7.70 in all treatments at day 0 (before starting the pCO₂ treatment), thereafter the elevated pCO₂ remained ca. 0.2 pH units below ambient pCO₂ (Table 1). The average total alkalinity (A_T) was 1527 μmol kg⁻¹ at day 0, and 1573 at day 12, with no difference between the treatments (Table 1). The pCO₂ at day 12 was lower in the treatments with organisms as compared to the controls without organisms.

Figure 1:

Environmental conditions during the experiment.

(A) Temperature (°C) at ambient (blue: 1°C) and elevated (red: 4°C) conditions. (B) Radiation conditions (PAR, 400–700 nm) during the experiment. Gray areas (A) and error bars (B) indicate standard deviation, n=2.

Stoichiometry and dissolved inorganic nutrients

There was a significant interaction between sampling day and temperature for POC and PON concentration [p=0.006 and 0.002, F_1,8=13.9 and 16.6, respectively, linear mixed effects (LME)]. The concentrations of POC and PON (Table 2) did not differ between temperatures at day 0 (p>0.132, Tukey’s test), but were significantly higher at ambient temperature as compared to elevated temperature at day 12 (p=0.010 and p<0.0001, respectively, Tukey’s test). The measured POC (>1 μm) was ca. 20–40 times higher than the calculated POC based on microscopy (>8 μm) analysis at day 0, and ca. 46–128 times higher at day 12 (Table 2). There was a significant interaction between sampling day and pCO₂ for the POC:PON ratio (p<0.018, F_1,8=8.9, LME). Thus, the interaction between sampling day and pCO₂ was reflected by a significant difference between pCO₂ treatments for POC:PON ratio at day 0 (p=0.005, Tukey’s test), but no difference at day 12 (p=0.999, Tukey’s test). Also, the POC:PON ratio was, overall, significantly higher at elevated temperature than at ambient temperature (p=0.002, F_1,8=21.0, LME, Table 2).

The concentration of DIN (NO₃⁻+NO₂⁻) was around 8 μm at day 0 and decreased to below the detection limit (<0.2 μm) at elevated temperatures at day 3, but with a concentration >6 μm at ambient temperature (Figure 2A). Thereafter the concentration of DIN decreased in all treatments, to an average of 0.31 μm at day 12. The concentration of DISi (Si) remained >10 μm in all treatments throughout the experiment (Figure 2B), with lower concentrations at elevated temperature at day 12 as compared to ambient. The concentration of DIP (PO₄³⁻) was <0.25 μm in all treatments day 0. From day 1, the concentration of PO₄³⁻ (DIP) remained >2 μm in all treatments throughout the experiment (Figure 2C). The addition of DIP changed the DIN:DIP ratio from an average of 38.66 at day 0 to an average of 0.03 at elevated temperature and 1.85 at ambient temperature at day 3 and, from day 3 and onwards, the average DIN:DIP ratio remained <0.14 in all treatments (Table 2).

Figure 2:

Inorganic nutrient concentrations in different temperature and pCO₂ treatments.

(A) Dissolved inorganic nitrogen (DIN; NO₃⁻+NO₂⁻). (B) Dissolved inorganic silica (DISi). (C) Dissolved inorganic phosphate (DIP). Error bars indicate standard deviation, n=3.

Photosynthetic activity and PAR levels

There was a significant interaction between sampling day and temperature on F_v/F_m (p<0.0001, F_4,32=28.9, LME). The F_v/F_m decreased from day 0 to day 6 (Figure 3A) and did not differ between temperature treatments at these sampling days (p>0.990, Tukey’s test). F_v/F_m increased from day 9, with a significantly higher ratio at increased temperature as compared to ambient at days 9 and 12 (p<0.0001, Tukey’s test). Interactions between temperature and sampling day were also observed for the RLC-derived photosynthetic parameters (Figure 3B–D), α_PSII (p=0.012, F_3,19=4.7, LME), rETR_max (p=0.040, F_3,19=3.4, LME) and E_k (p=0.004, F_3,19=6.1, LME). PAR was on average 150 μmol photons m⁻² s⁻¹ at midday during the initiation of the experiment and ranged from 100 to 250 photons m⁻² s⁻¹ on average until day 8, when PAR decreased to around 50 μmol photons m⁻² s⁻¹ until the end of the experiment (Figure 1B).

Figure 3:

Photosynthetic activity of phytoplankton in different temperature and pCO₂ treatments.

(A) Maximum quantum yield: F_v/F_m. (B) Initial slope of the rapid light curve (photosynthetic efficiency): α_PSII. (C) Minimum saturation irradiance: E_k (μmol photons m⁻² s⁻¹). (D) Maximum relative electron transport rate: rETR_max. Error bars indicate standard deviation, n=3.

Microplanktonic composition and specific growth rates

Shannon’s diversity index was significantly affected by time (p=0.001, F_4,32=6.3, LME) but not by temperature and pCO₂; the index was significantly higher at day 0 compared to the following days (p<0.01, Tukey’s test). The biovolume of both centric and pennate diatoms was also significantly affected by time (p<0.0001, F_4,32=77.8 and 46.5, respectively, LME). The biovolume of all diatoms (Figure 4B) significantly decreased between days 0 and 3 (p<0.0001, Tukey’s test) but then remained stable in all treatments until day 12. Neither the biovolume of pennate nor of centric diatoms showed any treatment effects at day 12 (Figure 4B). No treatment effects were found for dinoflagellates and ciliates day 12. No differences between the treatments were found for the specific growth rate of diatoms and dinoflagellates between days 6 and 12.

Figure 4:

Biovolume (mm³ l⁻¹) of phytoplankton in different temperature and pCO₂ treatments (a: 1°C 390 μatm, b: 1°C 970 μatm, c: 4°C 390 μatm, and d: 4°C 970 μatm).

(A) Biovolume of filamentous cyanobacteria. (B) Biovolume of ciliates, dinoflagellates, pennate diatoms and centric diatoms. Error bars indicate standard deviation, n=3.

The biovolume of the three inoculated filamentous cyanobacteria species (Nodularia spumigena, Aphanizomenon sp. and Dolichospermum sp.; Figure 4A) was significantly affected by time (p<0.0001, F_4,32=9.1, 1991.6 and 148.7, respectively, LME), and significantly decreased between days 0 and 6 (p<0.0001, Tukey’s test). Thereafter, the biovolume of N. spumigena slightly increased, but there was no significant difference between treatments at day 12. Aphanizomenon sp. and Dolichospermum sp. were almost non-detectable between days 3 and 9, and cells were only occasionally found at elevated temperature at day 12. For N. spumigena, an average specific growth rate of 0.10±0.13 d^-1 was observed between days 6 and 12, with no significant difference between the treatments. No specific growth rate could be calculated for Aphanizomenon sp. and Dolichospermum sp. due to lack of filaments at day 6.

Heterotrophic bacterial abundance and productivity

The bacterial abundance increased for all treatments on days 0–6, followed by a decrease until day 12 (Figure 5A). The bacterial abundance was significantly affected by sampling day (p<0.0001, F_4,29=312.4, LME) and pCO₂ (p=0.003, F_1,8=17.8, LME), with an interaction effect (p<0.0001, F_4,29=60.3, LME). There were also interaction effects for day and temperature (p<0.0001, F_4,29=76.0, LME), pCO₂ and temperature (p<0.0001, F_1,8=71.0, LME), and day, temperature and CO₂ (p<0.0001, F_4,29=44.8, LME). No significant correlations were found between bacterial abundance and groups of phytoplankton. However, significant correlations were found between bacterial abundance and POC:PON ratio (R²=0.41, p=0.024, n=12), POC (R²=0.35, p=0.042, n=12), and PON (R²=0.48, p=0.013, n=12), with high bacterial abundance at high POC and PON concentrations and low POC:PON ratio. The cell-specific productivity (CSP; Figure 5B) was significantly affected by sampling day (p=0.026, F_4,32=3.2, LME), gradually decreasing until day 9 (p<0.01, Tukey’s test), and remaining stable on days 9 and 12 (p=0.99, Tukey’s test).

Figure 5:

Heterotrophic bacterial abundance and bacterial production in different temperature and pCO₂ treatments.

(A) Bacterial abundance (10⁹ cells l⁻¹). (B) Cell-specific production (CSP, 10⁻⁹ μg C cell⁻¹ d⁻¹). Error bars indicate standard deviation, n=3.

Natural spring bloom community response to elevated temperature and pCO₂

All PSII parameters (F_v/F_m, α_PSII, rETR_max and E_k) decreased together with the drop in DIN concentration and phytoplankton biovolume but recovered by the end of the experiment. The initial decrease in DIN concentration at elevated temperature can possibly be explained by a relatively higher affinity for NO₃⁻ in the phytoplankton (Reay et al. 1999). Also, the sudden drop in photosynthetic parameters suggests that the cells became nutrient stressed early on in the experiment (Beardall et al. 2001). F_v/F_m, α_PSII, rETR_max and E_k were significantly higher at elevated temperatures as compared to ambient levels after 12 days, indicating that even though the total biovolume of phytoplankton was not yet affected by temperature, the cells were more active. The slow increase in photosynthetic parameters may have been triggered by regeneration of N through the decomposition of the initial standing stock of phytoplankton, or potentially from N₂-fixation by Nodularia spumigena. This strain of N. spumigena was recently shown to perform high rates of N₂-fixation in culture (Olofsson et al. 2016). From days 9 to 12 we observed a decrease in Si concentration at elevated temperature, suggesting that diatoms were the main contributor to the increase in photosynthetic activity, but this was not reflected in the biovolume (or cell numbers) of diatoms. Silica-scaled chrysophytes have been observed in the Baltic Sea, but a more likely explanation may be that diatoms attached to the walls of the aquaria were missed out in the sampling. Also, the assimilation of Si by diatoms increases with temperature, possibly explaining the decrease from day 9 to day 12 at elevated temperature (Martin-Jézéquel et al. 2000).

No short-term effects of elevated pCO₂ were observed on the biovolumes of diatoms and dinoflagellates. Previously documented short-term effects on diatoms include enhanced photosynthesis and carbon production (Wu et al. 2010, Yang and Gao 2012, Liu et al. 2017), while long-term studies have reported decreased growth rate and increased rates of respiration (7 months incubation in Torstensson et al. 2015, 1800 generations in Li et al. 2017). In a large-scale Baltic Sea mesocosm experiment, no effect on primary productivity at elevated pCO₂ (ca. 1330 μatm) was observed, while a decreased respiration rate was detected after 31 days (Spilling et al. 2016). These previous studies point at the large variation between species, incubation times and environmental conditions affecting the outcome of experimental treatments. Our study highlights the resistance to short-term fluxes by the phytoplankton community, where no major effects by elevated pCO₂ were found after 12 days. However, minor effects on species levels were detected, where the pennate diatom (10–40 μm) and the dinoflagellate Peridiniella catenata (Levander) Balech 1977 were negatively affected by the elevated temperature and pCO₂, respectively (Supplementary Tables S1, S2). At day 12, there were higher concentrations of POC and PON at ambient temperature as compared to elevated, which may be related to organisms of 1–8 μm, like nano- and picoflagellates, present in the Baltic (Kuuppo 1994), which was not in focus in this study.

Response of filamentous cyanobacteria to elevated temperature and pCO₂

Filamentous cyanobacteria are important N-sources in the Baltic Sea during summer. Overall, Aphanizomenon sp. contributes the most to N₂-fixation due to its high biomass (Klawonn et al. 2016) and sustains high N₂-fixation rates already at 10°C during early spring (Svedén et al. 2015). In the present study, the DIN:DIP ratio was <3 from day 3 onwards, similar to late-spring conditions, which may stimulate growth of cyanobacteria (Larsson et al. 2001). Anyhow, the filamentous cyanobacteria in our study were not able to outcompete the natural spring bloom species at low temperatures. Recovery was occurring faster at higher temperature, which was interestingly not (yet) reflected in biovolume and specific growth rates. The early acclimation of PSII is most likely occurring faster than what can be observed in growth, as the light-dependent reactions generally respond rapidly and growth encompasses many more processes than photochemistry.

An earlier start of the summer season has recently been predicted in a future climate change scenario (Kahru et al. 2016). Both Aphanizomenon sp. and Nodularia spumigena are present in low concentrations during winter in the Baltic Sea, but so far very little is known about their wintertime activity (Suikkanen et al. 2010, Wasmund 2017). During winter, Aphanizomenon sp. appears in higher abundance than N. spumigena in the Baltic Sea and, thus, possibly exhibits different survival strategies in terms of initial biomass in a pre-bloom situation (Wasmund 2017). Even though the low temperature (≤4°C) is far from optimal growth temperatures for the filamentous cyanobacteria (Suikkanen et al. 2010), the increase in biomass of N. spumigena from day 6 onwards indicates rapid acclimation to the experimental conditions in this strain. Thus, we suggest that N. spumigena has a potential to grow during spring bloom conditions. The biovolume of Aphanizomenon sp. decreased at the initiation of the experiment, as also observed by Wulff et al. (2018), thus, no temperature effects could be observed.

Manipulations of the carbonate system using air bubbling to mimic future climate change scenarios have previously been used successfully (Gattuso et al. 2010, Karlberg and Wulff 2013, Torstensson et al. 2013, 2015, Wulff et al. 2018). A pH difference was established from day 3 onwards, even though specific pCO₂ values are difficult to control because pCO₂ also varies with photosynthesis and respiration rates (Wulff et al. 2018). Nevertheless, no effects of elevated pCO₂ were found on the filamentous cyanobacteria. These results are consistent with some previous studies (Paul et al. 2016, Wulff et al. 2018), while others show increased growth (Wannicke et al. 2012, Eichner et al. 2014). Also, filamentous cyanobacteria are known to cope with daily fluctuations in pH measured both in aggregates (Ploug 2008) and in situ conditions (Wulff et al. 2018).

Response of heterotrophic bacteria to elevated temperature and pCO₂

Growth of heterotrophic bacteria is limited by temperature during most of the year in the Baltic Sea (Hagström and Larsson 1984). Nevertheless, no response in bacterial abundance to higher temperature was observed in our experiment, which may indicate that in early spring an increase of 3°C would not stimulate bacterial growth. Consistent with earlier work (Burrell et al. 2017), bacterial abundance was not affected by the increase in pCO₂. The observed early increase in abundance and productivity in all treatments (up to day 6) may be due to the release of DOC from surrounding phytoplankton (Engel et al. 2013, Torstensson et al. 2015). After day 6, the decline in bacterial abundance may suggest that bacterial mortality agents (grazers or viruses) increased, especially at higher temperature. Indeed, the grazing by heterotrophic flagellates has been shown to increase with temperature in the Baltic Sea Proper (Riemann et al. 2008). In addition, other work suggests that virus-induced bacterial mortality might be particularly sensitive to temperature variations in cold environments (Vaqué et al. 2017). Hence, while we did not examine viruses and flagellates (<8 μm) in the present experiment, they may potentially play central roles for the biogeochemical cycling of nutrients and energy through the planktonic microbial loop, particularly at the low temperatures prevailing in early spring.

In conclusion, increased pCO₂ and temperature did not enhance filamentous cyanobacteria over the natural spring bloom species, but the overall higher photosynthetic activity at elevated temperatures during the latter part of the experiment indicated more active cells within the community. For the toxic Nodularia spumigena, a positive specific growth rate during the second half of the experiment reflects its successful acclimation to temperatures ≤4°C and indicates a putative ecological relevance also during wintertime in the Baltic Sea. Our results highlight the resistance of the natural spring bloom community to short-term variations in abiotic conditions and, as such, suggest only limited microplankton responses to the pCO₂ and temperature levels predicted.