Samples were collected from Løgstør Bredning in Limfjorden close to the village of Aggersund (57°00′04.4″N 9°17′18.6″E). The place is subject to changing water levels where the sediment may be either covered by water or exposed to air for hours or days. Due to these harsh conditions the mats can develop without being grazed by larger invertebrates. Sampling was done by coring with Plexiglas cylinders having inner diameters of 3.6 cm and a length of 8 cm. Sampling was done in November 2014, January 2015, June 2015, September 2016, and December 2016. During each sampling campaign ∼10 samples were collected and selected cores with well-developed cyanobacterial communities were profiled.

Samples were kept moist and stored at 5°C under a halogen lamp (100 μmol photons m^-2 s^-1) with a timer mimicking a diurnal cycle until experiments could be conducted in an aquarium kept at 20°C. Experiments were conducted between 1 and 17 days after sampling.

A combined H₂ and H₂S microsensor was applied for measuring H₂ and H₂S profiles in the mat. This type of sensor has been developed very recently (Maegaard et al., 2017, submitted) and allowed for perfectly aligned profiles of H₂ and H₂S, as both species enter the sensor through the same membrane-covered opening at the 20–50 μm wide tip. The physical design is identical to the STOX oxygen sensor (Revsbech et al., 2009), with a Clark-type H₂ microsensor (obtained from Unisense A/S) housed in an outer sensor capillary with a tip silicone membrane. A gold anode placed in front of the H₂ microsensor and bathed in a ferricyanide containing electrolyte (Jeroschewski et al., 1996) serves to quantify the H₂S entering the sensor assembly and also prevents H₂S interference on the H₂ signal. The H₂ part of the sensors applied had a sensitivity of 1.2–7.7 pA μmol^-1 L^-1 and the H₂S part of the sensors had a sensitivity of 16.3–32.5 pA μmol^-1 L^-1. The calibration of the sulfide part of the sensors was done in 0.1 M HCl where O₂ had been removed by flushing with N₂. A 20 mmol L^-1 Na₂S solution was prepared from a washed crystal of Na₂Sx9H₂O from which fresh calibration standards of 0, 50, 100, and 200 μmol L^-1 H₂S were made. No pH profiles were recorded in the mat, and the total pool of dissolved sulfide thus cannot be calculated. The pKa of H₂S is about 7 (varying somewhat with salinity (Millero et al., 1988) and the total pool of dissolved sulfide was thus substantially higher than the measured H₂S concentration. The calibration of the H₂ part of the sensor was performed in water from the sampling site to which demineralized water saturated with H₂ was added to final concentrations of 0, 50, 100, and 200 μmol L^-1. The concentration of the H₂ saturated stock solution was calculated according to Wiesenburg and Guinasso (1979) to contain 805 μmol H₂ L^-1. In addition to the combined H₂ and H₂S sensor, an O₂ microsensor (Revsbech, 1989) with a tip diameter of 10 μm was applied. The O₂ sensor was calibrated assuming O₂ saturation (252 μmol L^-1 at 20°C and 20‰ salinity) in the aerated aquarium. The second calibration point for the linear calibration curve was the zero reading obtained in deep anoxic sediment. Further, a combined H₂ and O₂ sensor was constructed according to the same principle as the combined H₂ and H₂S sensor, but where the H₂S anode and ferricyanide electrolyte were substituted with a gold O₂ cathode and bicarbonate electrolyte (Revsbech, 1989). The sensitivity was 5.7–21.8 pA μmol^-1 L^-1 for O₂ and 0.7–1.0 pA μmol^-1 L^-1 for H₂. The large current for O₂ is associated with a large stirring sensitivity, amounting to an about 10% signal difference between stirred and stagnant water. Oxygen data obtained with such sensors should thus not be used for calculation of fluxes and rates near mat-water interfaces. The H₂ and O₂ calibrations were performed as described above. Solubility of O₂ in water and diffusion coefficients for O₂, H₂ and H₂S in water were obtained from the tables at http://www.unisense.com/files/PDF/Diverse/Seawater & Gases table.pdf.

The mat was placed in an aerated aquarium filled with seawater from the sampling site (20‰ salinity) with the temperature kept at 20°C. The water depth above the sample was ∼2 cm. The depth position of the sensor tips was controlled by a motorized micromanipulator connected to a PC with the data acquisition software Sensor Trace Pro (Unisense A/S). The program also served to collect the A/D converted signals from the sensors. The signal currents from all types of sensors were measured by custom-made picoammeters connected to an 8-channel A/D converter (Unisense A/S). The sensor signals were recorded for every 50 μm depth interval. The recording of a H₂/H₂S profile took about 5 min and recording of a H₂/O₂ profile took about 15 min. The times for recording profiles mentioned in the following are for the start of a profile.

The surface of the sediment cores was illuminated at an irradiance of 600 μmol photons m^-2 s^-1 in the 400–700 nm range using a halogen lamp equipped with a heat filter. This light intensity corresponds to about a third of maximum sunlight intensity at noon during mid-summer. For all H₂ accumulation experiments the samples received 3 h of light before onset of darkness. Profiles of O₂, H₂ and H₂S were recorded both during illumination and during darkness. O₂ profiles were only recorded at steady-state light and dark conditions, whereas combined H₂/H₂S profiles and H₂/O₂ profiles were measured after 3 h of illumination and at about 20 min intervals throughout the dark period following illumination. For clarity of illustration only some of the measured profiles are shown on figures.

Profiles of reaction rates of O₂ and H₂ were obtained by use of the diffusion-reaction program “Profile” (Berg et al., 1998). The porosity in this type of mat can be assumed to be 0.75 according to Glud et al. (1995). We estimated the effective diffusion coefficient in the mat (D_s) asuming D_s= φ D (Berg et al., 1998), where φ is porosity and D the diffusion coefficient in water. For the values of D for O₂ and H₂ we used the Unisense table values of 2.0 and 3.9.10^-5 cm^-2 s^-1, respectively. The boundary conditions were a H₂ or O₂ flux of zero in the water column and below the consumption zone. Calculation of activity distributions from H₂ and O₂ profiles were only possible when the profiles were relatively stable, and we were thus unable to do calculations on data obtained just after a light-dark shift. The presence of iron minerals oxidizing and precipitating sulfide and the lack of pH data prevented modeling of H₂S production profiles. It should be stressed that the recording of a H₂/H₂S profile required about 5 min, which further stresses the need of near-steady state of modeled profiles. Fick’s first law of diffusion (Crank, 1975) was applied for calculation of water – mat diffusional exchange of O₂, assuming that the steepest gradient was in a loose surface layer with diffusion coefficient for O₂ close to the one in water.

The sampled mats were 1–2 mm thick rigid, dark-green structures formed on top of sandy sediment. Microscopic examination showed that Microcoleus sp. were dominating the oxygenic photosynthetic community, but other filamentous cyanobacteria and diatoms were present as well. The underlying sediment was black indicating sulfidic conditions. During illumination O₂ bubbles were formed on top of the mat. A white layer of colorless sulfur bacteria was observed at the surface after several hours of dark incubation, indicating that the O₂/H₂S interface was at the very surface of the mat (Jørgensen and Revsbech, 1983).

Oxygen concentrations peaked at ∼0.1 mm during illumination and O₂ penetrated to a depth of ∼1 mm. Modeling of the profile shown in Figure 1 indicated a net O₂ production zone from 0 to 0.2 mm and a net consumption zone from 0.2 to 1.1 mm. The net O₂ production was calculated to be 0.59 nmol cm^-2 s^-1 and the net O₂ consumption below 0.2 mm depth was 0.17 nmol cm^-2 s^-1 (Figure 1). Oxygen profiles were measured at 3 additional positions of the mat, and all showed O₂ penetrations of 1.0–1.1 mm and comparable O₂ productions in the zones with net O₂ production. The exact figure for net O₂ production is, however, very dependent on the diffusivity at the mat/water interface. If we assume that the steepest O₂ gradient is in the overlying water, which will be true with a diffuse water-sediment interface, then the estimate of O₂ production in the zone with net production (Figure 1) increases to 0.79 nmol cm^-2s^-1, and the average from all 4 profiles was 0.86 ± 0.16 (SD) nmol cm^-2 s^-1.

After illumination ceased a new steady state O₂ profile was approached within 5 min, and O₂ penetrated to a depth of only ∼0.2 mm. The modeled O₂ consumption in the zone from 0 to 0.2 mm was 0.17 nmol cm^-2 s^-1 (Figure 1). With a similar argumentation about the steepest gradient found in a layer with a diffusivity close to the one in water, then the estimate of total diffusive uptake increases to 0.25 ± 0.04 (SD, n = 4) nmol cm^-2 s^-1. It should be stressed that the relatively large differences between surface gradient and profile modeling obtained estimates of reaction rates are due to the ill-defined boundary between mat and water. Modeling of rates in deeper layers are not subject to a similar uncertainty.

Many sets of combined H₂/H₂S profiles were measured in the Aggersund microbial mat, and an example is shown in Figure 2. The extent of H₂ accumulation was highly variable between samples with the maximum H₂ levels after darkening varying from 0 to 40 μmol L^-1 [average 12 ± 12 (SD, n = 13)]. We suspect that the variation was caused by differences in the species composition of the cyanobacterial community. The average depth of the H₂ peak was 0.5 ± 0.2 (SD, n = 12) mm. The average H₂ production rate was 0.031 ± 0.027 nmol cm^-2 s^-1 (SD, n = 6). Mats that appeared to be very similar could have very different levels of H₂ accumulation, but a prerequisite for substantial H₂ accumulation seemed to be a well-developed cyanobacterial community. The data in Figure 2 show H₂ accumulation higher than average, but the shapes and positions of the profiles are representative. After only 1 min of dark incubation H₂ concentrations of up to ∼15 μmol L^-1 were observed in the formerly oxic layer of the mat (Figure 2A). Hydrogen sulfide was present from a depth of ∼1.3 mm, and in this layer H₂ was consumed as well. Hydrogen and hydrogen sulfide profiles after 20 min of darkening are shown in Figure 2B. Here the H₂ accumulation peaked with a concentration of ∼40 μmol L^-1 and H₂S was present from a depth of ∼0.8 mm. Modeling of the H₂ profile showed a production zone from 0.5 to 1 mm and a consumption zone from 1 to 1.3 mm. The H₂ production in the zone with net production was 0.060 nmol cm^-2 s^-1 and the H₂ consumption in the layers with measurable H₂ and net consumption was 0.043 nmol cm^-2 s^-1, with the difference of 0.017 nmol cm^-2 s^-1 diffusing out of the mat. The peak H₂ concentrations had decreased to ∼20 μmol L^-1 after 120 min of darkening, and H₂S was present at the surface of the mat (Figure 2C). Modeling of the H₂ profile now showed a production zone from 0.5 to 0.8 mm and a consumption zone from 0.8 to 0.9 mm, where the net H₂ production was 0.032 nmol cm^-2 s^-1 and the H₂ consumption was 0.023 nmol cm^-2 s^-1, respectively. At 120 min, 0.09 nmol H₂ cm^-2 s^-1 diffused out of the mat. After 240 min (Figure 2D), there was no H₂ accumulation and H₂S was present to the very surface of the mat.

To investigate how O₂ affected H₂ metabolism, the mats were analyzed with combined H₂-O₂ microsensors. An example of perfectly aligned data for O₂ and H₂ obtained with such sensors is shown in Figure 3. During illumination, O₂ was produced in a zone from ∼0-0.4 mm and consumed in a zone from ∼0.4-1.0 mm, and no H₂ was detected (Figure 3). The set of profiles initiated 1 min after darkening showed that O₂ now only penetrated to ∼0.25 mm, and H₂ was already accumulating with maximum concentrations of 26 μmol L^-1 (Figure 3). It is not possible to model the rates of H₂ metabolism as the data were obtained 1–5 min after darkening and thus are far from a steady state. However, as seen in Figure 2 the peak of H₂ occurs in the layer that was oxic during illumination, and the linear H₂ profile through the ∼0.25 mm thick oxic zone indicates no net transformation in this layer. Due to sulfide poisoning of the sensor it was not possible to get H₂ profiles for extended periods of dark incubation.

A series of experiments were conducted where H₂ was added to the overlying water, and an example of data from these experiments is shown in Figure 4. Illumination resulted in O₂ being produced in a zone from 0 to 0.55 mm at a rate of 0.94 nmol cm^-2 s^-1 and O₂ being consumed in a zone from 0.55 to 0.65 mm at a rate of 0.38 nmol cm^-2 s^-1. The main H₂ consumption zone was from 0.5 to 0.7 mm, and H₂ was consumed at a rate of 0.03 nmol cm^-2 s^-1 (Figure 4). There was no indication of any H₂ consumption in the oxic photosynthetic zone, but a high rate of H₂ consumption at the oxic-anoxic interface. This consumption could be due to both photosynthetic (Drutschmann and Klemme, 1985) and non-photosynthetic prokaryotes. Oxygen was consumed in a zone from 0 to 0.15 mm at a rate of 0.17 nmol cm^-2 s^-1 during darkness. Hydrogen was consumed down to 0.5 mm at a uniform low rate integrating up to 0.05 nmol cm^-2 s^-1 (Figure 4). It is not possible to judge if the oxic zone actually had lower activity than the anoxic zone, but modeling of the profile with zero activity in the oxic zone gave just as good a curve fit (data not shown) as the activity distribution shown in Figure 4B.

Substantial H₂ concentrations persisted for a long time in the mat when sulfate reduction was inhibited by molybdate (Figure 5). After 120 min the H₂ accumulation thus peaked with a concentration of ∼13 μmol L^-1. This concentration is much lower than the levels shown in Figure 2, but this may be due to a less developed mat. After 210 min the maximum H₂ concentration had decreased to ∼6 μmol L^-1, and the methanogensis inhibitor BES was added. Inhibition of methanogenesis resulted in the maximum H₂ concentration increasing to ∼10 μmol L^-1. Hydrogen sulfide profiles are not shown since sulfide forms a complex with molybdate (Newport and Nedwell, 1988). A positive effect of BES on H₂ accumulation was also seen in two similar experiments with an average increase in maximum H₂ concentrations of 53 ± 9% (SD, n = 3).

Hydrogen accumulation after darkening also occurred when the photosynthetically active part of the mat was placed on agar (Figure 6A). The H₂ accumulation peaked after ∼30 min and H₂ was barely detectable after 450 min. No H₂S was detected at any time (data not shown). Oxygen profiles were measured at several locations and all data showed that the mat was oxic in all layers during the illumination but became anoxic below about 0.2 mm after 5 min of darkening (data not shown). In another experiment with the surface ∼1 mm of the mat placed on agar, the H₂ accumulation peaked at ∼15 μmol L^-1 after 20 min of darkness (Figure 6B). When the maximum concentration had decreased to ∼4 μmol L^-1 at 165 min, molybdate was added to inhibit sulfate reduction. This resulted in a sizable stimulation of the H₂ accumulation so that a maximum concentration of 27 μmol L^-1 was reached. Sequential addition of inhibitor was also tested on a new slice of mat by adding molybdate 2 h before darkening and adding BES after 210 min of darkening. The maximum level of H₂ of 13 μmol L^-1 in the molybdate inhibited mat occurred after 20 min of darkening, after which H₂ decreased to a maximum of 5 μmol L^-1 at 195 min. When BES was added, it led to an increase of the maximum H₂ concentration to 15 μmol L^-1 at 315 min (data not shown).

The microbial mat was oxic down to about 1 mm during illumination, but after darkening O₂ only penetrated to about 0.2 mm. When darkened, H₂ accumulated only within the formerly oxic zone of the intact sediment cores, where photosynthates formed during illumination were available for fermentation. The zone of net H₂ formation as measured by the combined H₂ – H₂S sensor was in anoxic layers below 0.4–0.5 mm depth and thus relatively far below the depth of maximum net O₂ production as measured by a separate O₂ sensor. When the combined H₂ – O₂ sensor was used we observed, on the other hand, a perfect alignment between the concentration peak of H₂ and photosynthetic layers below 0.2 mm that became anoxic by darkening. It thus seems that our surface definition was about 0.2 mm wrong when we estimated the surface from combined H₂-H₂S data. This illustrates the problems of aligning profiles from separate sensors and the strength of using sensors with dual chemical sensing as used here. By use of the two types of sensors with dual chemical sensing it was possible to determine at very high resolution how H₂ production and consumption were associated with the presence of both O₂ and H₂S. The problem of aligning profiles from different sensors was in our case probably aggravated by the very different tip shapes and dimensions of O₂ and combined sensors, where the O₂ sensor had a slimline-tip of 10 μm and the combined sensors had blunt tips of 50 μm. Insertion of a microsensor changes the flow pattern above the mat and also the thickness of the diffusive boundary layer (Glud et al., 1994), and this effect is very dependent on sensor dimensions and shape.

The results shown in Figures 3, 4 were obtained by use of combined H₂-O₂ sensors. Microbial mat overlaid with air-saturated water was analyzed in Figure 3, and the dark profile initiated 1 min after darkening showed no sign of H₂ consumption in the 0.2-mm thick oxic surface layer. Further proof for the very low H₂ consumption potential is provided by the data in Figure 4, obtained after saturating the overlying water with a mixture of ∼10% H₂ gas and air. No H₂ consumption could be seen in the oxic 0.7-mm layer of the illuminated mat (Figure 4A). When the mat had been dark incubated for 2 h there was a uniform low rate of H₂ consumption throughout the formerly oxic zone (Figure 4B). These data are important as it has previously been suggested that sulfate reduction may occur at high rates in oxic cyanobacterial mats (Canfield and Des Marais, 1991). The absence of any significant H₂ oxidation potential in oxic mat counter-indicates that there are active sulfate reducers in the oxic mat, as sulfate reducing bacteria generally use H₂ as electron donor (Jørgensen, 1982).

The data obtained with combined H₂ – H₂S sensors shortly after darkening, but sufficiently long after darkening to approach diffusional equilibrium, all have almost linear profiles between the H₂ production zone and the overlying water (e.g., Figure 2B). This indicates that not only oxic mat layers, but also mat layers that have a recent history of being oxic are characterized by very little H₂ consumption as illustrated by Figure 4B. A low H₂ oxidation potential of oxic cyanobacterial mats was also suggested by Hoehler et al. (2002).

A high H₂ consumption potential was found in the layer that formed the oxic-anoxic interface during illumination (Figure 2). This indicates that anaerobic bacteria are the main hydrogenotrophs in the mat, and this is in accordance with the findings of Lee et al. (2014) who identified sulfate reducing bacteria as the main hydrogen consumers. The zone of intense H₂ consumption moved 0.2 mm upward from 20 to 120 min after darkening (Figures 2B,C), indicating that the anaerobic bacteria were able to migrate. At 120 min there was also a low rate of H₂ consumption in the surface layers that were now highly sulfidic, indicating that some anaerobic bacteria had reached these layers by migration, or that dormant bacteria permanently inhabiting these layers had become active. The presence of sulfate reducing bacteria in oxic microbial mat has previously been reported by Minz et al. (1999). The question of migration of bacteria versus revival of inactive hydrogenotrops in the layer that was oxic during the light period was addressed by slicing off the top 1 mm and placing the slice on agar (Figures 6A,B). When molybdate was added to the sliced mat after 165 min of dark incubation (Figure 6B) there was a rapid and significant increase in H₂ concentration, indicating that resident sulfate reducing bacteria had become active although all layers of the mat slice were oxic during the preceding light period. The effect of molybdate could also be explained by change in the fermentative pattern, but we are not aware of reports indicating such an effect of molybdate. So after >1 h of dark incubation it seems that both migration from below and revival of resident anaerobic hydrogenotrophic microorganisms caused hydrogen consumption in the surface layers of intact mats.

Previous investigations (Lee et al., 2014; Nielsen et al., 2015) indicate that sulfate reduction is the major hydrogenotrophic process in dark-incubated cyanobacterial mats, although methanogenesis may also occur in saline environments (Oremland and Taylor, 1978). We investigated the possible role of methanogens as hydrogenotrophs in the mat by sequentially adding molybdate and BES to both intact (Figure 5) and sliced mat (data not shown). Data from experiments with addition of inhibitors should always be interpreted with care as no inhibitor is totally specific. Molybdate thus complexes with H₂S (and thereby lowers the H₂S activity), and it also forms complexes with volatile fatty acids (VFAs) (Finke et al., 2007). BES addition caused a significant increase in H₂ concentration, indicating that significant numbers of active methanogens were present. As expected the effect of BES was smaller than the effect of molybdate, in accordance with the conclusions of Lee et al. (2014) that sulfate reducers are the main hydrogenotrophs in such mats. The high H₂ concentrations in a dark-incubated mat might, however, make methanogenesis locally and temporally favorable (Hoehler et al., 2002). The effect of BES in the sliced mat provides evidence that methanogens like sulfate reducers could survive long periods of oxic conditions in the mat, as previously described for a hot spring microbial mat (Revsbech et al., 2016) and soils (Angel et al., 2012).

It should be noticed that sulfide did not accumulate in the sliced 1-mm mat layer during the dark incubation (Figures 6A,B). That might seem to counter-indicate H₂ consumption by sulfate reduction, suggested by the increase in H₂ after addition of molybdate, but the data in Figure 2 may offer an explanation. Here H₂S was found in high concentrations in layers deeper than 1.1 mm after 1–6 min (the time to record a profile was about 5 min) of darkness, and the situation was basically unchanged after 20–26 min of darkness. The mean diffusion time for a H₂S molecule over a distance of 1 mm is less than 10 min, assuming a diffusion coefficient of 1 × 10^-5 cm²s^-1 within the mat (Jørgensen, 2000). By diffusion alone the whole mat should thus have been highly sulfidic already after a few minutes, and that points to extensive sulfide consumption within the previously oxic mat. The consumption of free sulfide may be due to oxidation and precipitation by iron and manganese minerals. The buffer capacity of the iron pool should be very high if it could consume the high sulfide flux from below in the intact mat for more than 20 min. The iron pool might thus be able to keep an isolated 1-mm top layer free of dissolved sulfide for a long time. An alternative explanation for the absence of free sulfide is oxidation by cable bacteria (Pfeffer et al., 2012). The mats from Aggersund have been found to be rich in cable bacteria (L.P. Nielsen, unpublished), and these may be able to oxidize H₂S within the upper few mm of the mat by transporting the electrons to the surface of mat. In intact mat they may not be able to oxidize all the sulfide diffusing up from below, but the much lower amount formed in a thin slice might all be oxidized.

The rates of net H₂ production deduced from Figure 2B were 0.06 nmol cm^-2 s^-1 when they were at the highest. This is about 10% of the rate of net O₂ production in the same mat (Figure 1). Even in the presence of molybdate there is some decrease in H₂ production with time (Figure 5), but the illumination periods were also limited, so the figures are roughly comparable. One O₂ corresponds to two H₂ molecules in terms of electron transfer, so about 5% of the reducing equivalents stored in organic compounds during the light period were liberated as H₂. Fermentation of glucose by a Clostridium that was studied because of a high rate of H₂ formation was found to yield up to 2 moles of H₂ per mole of glucose (Jo and Kim, 2016). Assuming a loss of glucose equivalents as fast as the rate of build-up during photosynthesis, the 0.6–0.9 nmol O₂ cm^-2 s^-1 corresponds to a formation rate of 0.2–0.3 nmol H₂ cm^-2 s^-1. The loss rate of glucose in the dark should actually be lower than the rate of glucose production by photosynthesis in the light, and we thus also see only 20–30% of the H₂ production calculated above. The mass balance considerations do indicate a substantial loss of photosynthates by fermentation during the dark period. This is in accordance with previous estimates of extreme rates of sulfate reduction is such mats (Jørgensen and Cohen, 1977), as the sulfate reducing bacteria utilize the fermentation products.

Based on our data we cannot rule out that there was H₂ production due to nitrogenase activity, which would invalidate the H₂ budget outlined above. However, Burow et al. (2012) and Lee et al. (2014) found no effect on H₂ accumulation in a similar mat when nitrogen fixation was inhibited. Both studies observed a production of other fermentation products as well as H₂, which support fermentation and not nitrogen fixation as the H₂ source. The immediate rise in H₂ after darkening observed in our study also points to fermentation as the major H₂ forming process, as de novo synthesis of the O₂ sensitive nitrogenase would require some time.