Sediment sampling and set-up
The sediment surface layer (approximately the first 2 cm) was collected during a summer low tide at Alcochete intertidal flats, Tagus estuary, Portugal (38°44'45''N, 08°59'04''W). Sediment was transported in refrigerated containers to the laboratory, homegeneized and placed inside microcosms in a flow-through experimental life support system (ELSS), forming a layer of 6 cm.
Induction of MPB cell distribution within the sediment profile was achieved by exposing the sediment to an irradiance of 70 μmol photons m−2 s−1 for ca. 24 h. Establishment of the MPB surface biofilm was assessed by measuring the normalized difference vegetation index (NDVI, see below). Once the MPB surface biofilm was established, all microcosms were subjected to the initial conditions of temperature and pH (18°C, pH 8.0). After 24 h at these conditions, four different treatments were started and the experiment run for 11 days: 1) 18°C and pH 8.0; 2) 24°C and pH 8.0; 3) 18°C and pH 7.4; and 4) 24°C and pH 7.4. Four microcosms were used for each treatment (with a total of 16 microcosms being used in the whole experiment).
The temperatures were chosen within the summer variation range of the study site and corresponded to mean high tide (18°C) and mean diurnal low tide (24°C) sediment temperatures . The pH of the sediment interstitial water was 8.0, while a pH drop of 0.6 units (pH 7.4) was chosen on the basis of the Intergovernmental Panel on Climate Changes  maximum projections for the change in global ocean surface pH (~0.4 units) in 2100, together with possible increased acidification caused by upwelling of anthropogenic CO2-enriched water in coastal systems .
Experimental life support system (ELSS)
A flow-through ELSS was used, as described in detail by Coelho et al. . The ELSS consisted of 16 independent microcosms (glass tanks - 28 cm length x 25 cm height x 12.4 cm width) with a maximum functional water volume of approximately 7 L (see Additional file 1: Figure S1). The ELSS was equipped with 4 full spectrum fluorescent tubes (AquaLight, T5/54 W/10000K, Bramsche, Germany) and set to 6 h light–18 h dark cycle with an irradiance at sediment surface of 70 μmol photons m−2 s−1.
The ELSS was operated with one daily tide. Saltwater was prepared in two reservoirs (230 L each) by mixing freshwater purified by a reverse osmosis unit (Aqua-win RO-6080) with a commercially available marine salt mixture (Tropic Marin Pro Reef salt – Tropic Marine, Germany) to a final salinity of 30. The water for tidal cycles was prepared 24 h before use. To simulate high tide, saltwater was pumped from the respective reservoir using a submersible pump (Aquabee UP 3000) through an independent pipeline system of polyvinyl chloride (PVC) tubes into each microcosm. The saltwater flow rate was manually controlled by a PVC valve located above each microcosm. The saltwater input was stopped when the water layer reached ca. 15 cm. High tide started after 15 min of the onset of the dark period. To simulate low tide, outflow submersible pumps (Rena flow 400 C) were used in each microcosm, operated using digital timers. These pumps were positioned inside a PVC cylinder and protected with a mesh screen to avoid clogging. The water was discharged using a common outflow pipe. Low tide started 15 min before the period of light exposure.
The microcosms in the ELSS were partially submerged into two main water-bath tanks. One tank was set to 18°C, the water was continuously pumped by a canister filter pump (SunSun HW-302) through a cooler equipped with a thermostat (Teco TR10) with a flow rate of 1000 L h−1. The other tank was equipped with two submersible 200 W heaters with thermostats (Rena Cal 200) set to increase water temperature to 24°C.
Water pH was manipulated by acidifying the water stocked in the saltwater reservoirs by bubbling CO2 through a diffuser. The diffuser operated with a water pump (Aquabee UP 3000) to maximize CO2 gas mixing in saltwater. CO2 addition was controlled with a feedback system that included a combination of a pH electrode connected to a controller (V2 control pH controller, Tropical Marine Centre, Bristol, UK) and a pressure regulator with an integrated solenoid valve (V2 pressure regulator pro, Tropical Marine Centre, UK). The digital display of the controller allowed visualization of actual pH in the saltwater reservoir and pH monitoring with the pH electrode. The controller opened the solenoid valve whenever pH rose above the set value; CO2 was then injected until water pH returned to the pre-set value.
MPB biomass was estimated daily and non-intrusively in each microcosm by calculating NDVI. Daily measurements of spectral reflectance as well as Pulse Amplitude Modulated (PAM) fluorescence (see below) were done in all microscosms during low tide, starting after 90 min of light exposure to ensure that the sediment was in the same conditions regarding diatom migration and biofilm establishment. Reflectance spectra were measured over a 350–1000 nm bandwidth with a USB4000 (Ocean Optics, USA) with a VIS-NIR optical configuration connected to a 400 μm diameter fiber optic (QP400-2-VIS/NIR, Ocean Optics, USA). The light spectrum reflected from the sample was normalized to the spectrum reflected from a clean polystyrene plate. A reflectance spectrum measured in the dark was subtracted to both spectra to account for the dark current noise of the spectrometer. The fiber optic was positioned perpendicularly to the sediment surface and both sample and reference spectra were measured under a constant irradiance of 70 μmol photons m−2 s−1. NDVI was calculated as (R750 − R675) / (R750 + R675), where R750, R675 and R636 represented the average diffusive reflectance in the intervals of 749.73–750.39 nm, 674.87–675.55 nm and 635.71–636.40 nm, respectively .
Additionally, MPB biomass was calculated using HPLC chlorophyll a (Chl a) analysis at the beginning (T0) and at the end of the experimental period (T11). Invasive sampling for Chl a determination was done because previous studies have indicated NDVI saturation for high MPB biomass [24,25]. Sampling for Chl a was performed after spectral reflectance and PAM fluorescence measurements. For Chl a analysis, one sediment minicore (diameter 1.1 cm) was collected per microcosm at the beginning of the experiment (T0) using a plastic corer. The sediment surface (0 – 2 mm) was pooled in groups of 4 to obtain 4 mixed sediment samples. At the end of the experiment (T11), three minicores were collected per microcosm and the sediment pooled to obtain a total of 16 samples, one per microcosm. Sediment samples were immediately frozen in liquid nitrogen and stored at −80°C. Before analysis, samples were freeze-dried and extracted with 95% cold buffered methanol (2% ammonium acetate) for 15 min at −20°C, in the dark. Samples were sonicated (1210, Bransonic, USA) for 30 s at the beginning of the extraction period. Extracts were filtered (Fluoropore PTFE filter membranes, 0.2 μm pore size) and immediately injected in a high performance liquid chromatographer (HPLC; LC10AVP, Shimadzu, Japan) equipped with a photodiode array (SPD-M10AVP) detector . Chromatographic separation was carried out using a C18 column for reverse phase chromatography (Supelcosil; 25 cm long; 4.6 mm in diameter; 5 mm particles) and a 35 min elution programme. The solvent gradient followed Kraay et al.  with a flow rate of 0.6 mL min−1 and an injection volume of 100 μL. Chl a was identified from absorbance spectrum and retention time and concentrations calculated from the signals in the photodiode array detector. Calibration of the Chl a peak was performed using a commercial pigment standard from DHI (Institute for Water and Environment, Denmark).
MPB photosynthetic parameters
Measurement of MPB photosynthetic parameters were carried out in each microcosm using a Diving-PAM Fluorometer (Walz, Effeltrich, Germany). The distance between the fluorometer fiber optic and the surface of sample was kept constant at 2 mm during all measurements. Maximum quantum yield of photosystem (PS) II (F
m) was determined daily in each microcosm by calculating (F
m – F
m, where F
m and F
o are, respectively, the maximum and the minimum fluorescence of dark-adapted samples . F
m gives a robust indication of the maximum efficiency of photosynthesis. Dark adaptation period was restricted to 2 min to reduce the possibility of inducing downward vertical migration of the epipelic MPB .
On specific days (T0, T6 and T11), rapid light-response curves (RLC) were carried out in all microcosms to assess MPB photosynthetic activity over a wide range of ambient light intensities . For the construction of RLC, the samples were exposed to 8 intensities of actinic light increasing from 38 to 616 μmol photons m−2 s−1 (38, 55, 81, 122, 183, 262, 367 and 616 μmol photons m−2 s−1). Each irradiance step was 10 s; the saturation pulse intensity had duration of 0.6 s and an intensity of 8,000 μmol photons m−2 s−1. RLC were constructed by calculating, for each level of actinic light, the effective quantum yield of PSII (ΔF/F
m′) and the relative electron transport rate (rETR) from the delivered actinic irradiance (E) by rETR = E x ΔF/F
m′ . The light response was characterized by fitting the model of Platt et al.  to rETR vs E curves and by estimating the initial slope of the light curve α (light utilization coefficient), ETR
max (maximum rETR) and E
k (light saturation parameter), where E
k = ETR
max /α. The model was fitted iteratively using MS Excel Solver.
MPB community analysis
Surface sediment samples to determine the composition of the MPB community were collected as described for Chl a analysis and stored in a 2.5% glutaraldehyde solution at 4°C. Cells were extracted from the sediment following an isopycnic separation technique using silica sol Ludox HS-40 that separates the organic material from mineral particles and is, thus, able to remove microorganisms (e.g. MPB) from the sediment . Cell counts of MPB were made in a Sedgwick-Rafter cell counting chamber (50 μL of each extract) on an Olympus BX50 optical microscope (Olympus Corporation, Tokyo, Japan) at a 400x magnification. Between 3 and 9 horizontal transects (1300 – 8500 individual cells) were made, the cells counted separated into major MPB taxonomical groups (i.e. diatoms, euglenids, dinoflagellates and cyanobacteria) and the relative percentage determined.
Diatom analysis was conducted after cleaning the diatom valves of organic material. A subsample of 750 μL of extract was oxidized with 5–7 mL of hydrogen peroxide (30%) at 90°C for at least 4 h. Permanent slides, mounted in NaphraxTM (Northern Biological Supplies Ltd., Ipswich, UK), were made for each sample. Phase and differential interference contrast optical microscopy were used to identify and count diatoms at a magnification of 1,000x. For each slide, a minimum of 400 valves were counted and identified to the species level, following Ribeiro  and references therein.
The existence of significant differences was tested using two-way repeated measurements ANOVA (NDVI, F
m, and RLC parameters) or two-way ANOVA (Chl a and MPB major group relative abundance) for the effects of temperature (18 and 24°C) and pH (7.4 and 8.0). Multiple comparisons were performed using Tukey HSD. Bonferroni correction was applied to p values of multiple tests on correlated variables (NDVI and Chl a; ETR
max, α and E
k; relative abundance of diatoms and cyanophytes). Statistical analyses were carried out using Statistica 10 (StatSoft Inc., USA).
Diatom community structure was analysed with non-parametric multivariate tools using PRIMER® 6 software package (PRIMER-E, Plymouth, UK). The species abundance matrix was previously standardized and root-transformed and used in all multivariate routines. Bray-Curtis coefficients  were used to compute the similarity or dissimilarity distances between samples. A similarity-based ANOSIM permutation test, with a 2-way crossed layout , was performed to test if there were any statistically differences between groups of samples, namely, between temperature or pH treatments. A classification analysis (CLUSTER), which uses hierarchical agglomerative clustering of the samples and group-average linking , was also performed. During the dendrogram construction statistical significance of every cluster node was tested by the SIMPROF routine . The SIMPROF is an a posteriori permutation test of the null hypothesis that the set of samples below a given node does not show any multivariate structure, which are then represented by dashed lines. Species mainly responsible for possible differences between treatments were determined using SIMPER analysis .