No evidence of light inhibition on aerobic methanotrophs in coastal sediments using eDNA and eRNA

Abstract

It is estimated that up to half of global methane (CH₄) emissions are derived from microbial processes in aquatic ecosystems. However, it is not fully understood which factors explain the spatial and temporal variability of these emissions. For example, light has previously been shown to both inhibit and stimulate aerobic methane-oxidizing bacteria (i.e., methanotrophs) in the water column. These contrasting results indicate that the mechanisms that light has on CH₄ oxidation are not yet clearly known, even less so for benthic aerobic methanotrophs. Here, we tested whether light reaching the seafloor can inhibit methanotrophic activity on the sediment surface. We sampled and distributed over 40 intact sediment cores from two coastal sites (illuminated 10 m, and a dark site at 33 m water depth) into 0, 50, and 100 PAR light treatments. After 10 days, we found no difference between treatments for each site in pore-water CH₄ concentrations, relative abundance of aerobic methanotrophs, or the number of RNA transcripts related to methane oxidation. Our results suggest that light attenuation in coastal waters does not significantly affect aerobic methanotrophs in coastal sediments.

INTRODUCTION

Methane (CH₄) is an important greenhouse gas estimated to contribute to approximately 20% of Earth's radiative forcing (Kirschke et al., 2013). Its atmospheric concentration has rapidly increased since the last 40 years, from 1645 ppb (year 1984) to 1908 ppb measured in August 2022 (Lan et al., 2022; Nisbet et al., 2019; Stocker et al., 2013). Microbially mediated CH₄ emission is the dominant driver of the post-2006 increase in atmospheric CH₄ concentrations (Lan et al., 2021). In fact, half of global CH₄ emissions come from highly variable aquatic ecosystem sources, such as from sediments in wetlands and lakes (Rosentreter et al., 2021). Shallow coastal waters (<50 m water depth) constitute ~3% of the oceans but are estimated to account for 13%–32% of the oceanic yearly CH₄ emission (Weber et al., 2019). Within these shallow waters previous field studies have shown that inshore areas contain higher surface water CH₄ concentrations compared with deeper offshore areas (Bange, 2006; Borges et al., 2016; Humborg et al., 2019; Osudar et al., 2015). However, the CH₄ dynamics and factors driving spatial and temporal variability in CH₄ emissions in coastal waters are complex and not yet fully understood (Rosentreter et al., 2021; Weber et al., 2019). Aerobic CH₄ oxidation by specialized bacteria (so-called methanotrophs) in the sediment and water can potentially remove up to half of all CH₄ in shallow coastal waters (Mao et al., 2022). Understanding how and when these organisms oxidize CH₄ is therefore essential to comprehend the dynamics of CH₄ emissions from coastal zones.

In aquatic ecosystems, large parts of CH₄ is produced by methanogenic archaea in deep anoxic sediment (Enzmann et al., 2018). When this CH₄ diffuses upwards in the sediment, it can be oxidized by methanotrophs to CO₂. Methanotrophs typically consist of anaerobic methane-oxidizing archaea (ANME) and aerobic methane-oxidizing bacteria (Egger et al., 2018). ANME is found in the anoxic sediment, typically in the sulfate–methane transition zone, while aerobic methanotrophs are found in the oxic sediment surface and water column (Egger et al., 2018). Methanotrophs in and on the sediment surface therefore act as the last barrier before CH₄ diffuses upward into the water column. The abundance and composition of aerobic methanotrophs are regulated by the availability of CH₄ and O₂ (Knief, 2015), pH and salinity (Knief, 2015), and temperature and nutrient concentrations (Nijman et al., 2021). The enzymes particulate methane monooxygenase (pMMO, used to oxidize CH₄) and ammonia monooxygenase (AMO, used to oxidize ammonia) are structurally similar and evolutionary-related (Holmes et al., 1995). In addition, some aerobic methanotrophs utilize a soluble methane monooxygenase enzyme (sMMO) (Kalyuzhnaya et al., 2019). Interestingly, light can inhibit the activity of ammonia-oxidizing microorganism (Guerrero & Jones, 1996; Lu et al., 2020), which could indicate that aerobic methanotrophs are also inhibited by light. However, very little is known about how light intensity affects the activity of aerobic methanotrophs. Previous studies investigating the effect of light are scarce, and in some cases show both inhibition and stimulation of methane oxidation. For example, both Dumestre et al. (1999) and Murase and Sugimoto (2005) observed an increase in CH₄ concentrations under light treatments, and Morana et al. (2020) found a positive relationship between CH₄ oxidation inhibition and incident light in lake water. These findings indicate an inhibition of methane oxidation in the water column. However, such findings might not only be explained by methanotrophic inhibition as it is today known that oxic CH₄ production is possible by photosynthetic cyanobacteria and phytoplankton (Bizic, 2021; Bižić et al., 2020). In contrast, Oswald et al. (2015); Savvichev et al. (2019) measured potential stimulation of methane oxidation in illuminated lake water; however, it is unknown if this is an effect of direct methanotrophic stimulation or an indirect effect of stimulated oxic CH₄ production in the water column. Only two studies showed a potential stimulation of methanotrophy after saturating the sediment with light to enhance photosynthesis (King, 1990; King et al., 1990). Inhibition of CH₄ oxidation might be an important driver explaining variability in coastal CH₄ emissions. However, as far as we know, there are no studies that have used modern molecular tools to investigate any potential light inhibition on aerobic methanotrophs that inhabit the sediment surface.

Shallow coastal waters have higher CH₄ emissions (Weber et al., 2019) and lower abundance of methanotrophs than deeper offshore waters (Broman, Sun, et al., 2020). It is therefore possible that light illumination has a negative effect on the growth and activity of benthic aerobic methanotrophs. However, how the relative abundance and metabolic activity of methanotrophs are influenced by light remains unknown. Resolving this question is important as climate change and related anthropogenic pressures indirectly alter light availability in coastal ecosystem. For example, warmer waters enhance phytoplankton growth in the water column, which in turn, decreases light availability reaching the seafloor (Rabalais et al., 2009). Climate change can also increase water turbidity due to riverine-derived terrestrial dissolved organic carbon (Andersson et al., 2018), or wind-driven sediment resuspension (Mi et al., 2019). Therefore, it is possible that climate-induced alterations in coastal water illumination can impact CH₄ cycling through less explored mechanisms, not only in the water but also in the sediment. Furthermore, considering the high diel variability of CH₄ emissions from inland and coastal waters (Roth et al., 2022; Sieczko et al., 2020), it is necessary to elucidate the role of light in regulating aerobic methanotrophic activity.

In this study, we investigated how the relative abundance and activity of aerobic methanotrophs in the sediment surface are affected by photosynthetically active radiation (PAR) light. For this purpose, we sampled sediment cores from a shallow inshore (10 m) and an offshore deep coastal site (33 m) and investigated how different light intensities (0, ~50, and ~100 PAR μmol m⁻² s⁻¹) affected benthic methanotrophic communities and their activity for each site (Figure 1a). We sequenced environmental DNA (eDNA) and metabarcoded 16S rRNA gene amplicons. We also sequenced environmental RNA (eRNA) of the total RNA pool (rRNA plus mRNA) to analyze 16S rRNA and RNA transcripts. Furthermore, we used quantitative reverse transcription RT-qPCR, measured pore-water CH₄ concentrations, and organic carbon content (Table 1). We hypothesized that for both sampling sites, light would: (1) increase CH₄ pore-water concentrations due to inhibited methane oxidization in the sediment surface; (2) decrease the relative abundance of methanotrophs, and (3) decrease the number of detected RNA transcripts related to methane oxidation.

FIGURE 1. (a) Sediment cores were collected during September 2021 from two stations: a shallow illuminated site (10 water depth) and a deep dark site (33 m) on the coast of Finland. TSZ denotes the location of the Tvärminne Zoological Station. The map layer is © OpenStreetMap contributors. (b) Water column temperature, salinity, and CH4 concentrations (n = 24–26 measurements per water depth). (c) Sediment profiles of pore-water CH4 concentrations. Note that the 0–1 cm data are based on sliced sediment in the field (n = 7 per site), while pore-water at the other depth layers were collected using cut-off syringes inserted into the side of the core (n = 1 per layer).

TABLE 1 Overview of the experiment and number of replicates for each analysis.

Site	Dataset	Dark	Medium	High
Shallow	16S rRNA gene amplicon	7	7	7
	Total RNA-seq	3	3	3
	RT-qPCR	6	4	5
	Porosity + OM (%)	7	7	7
	Pore-water CH4	7	7	7
Deep	16S rRNA gene amplicon	5	6	8
	Total RNA-seq	3	3	3
	RT-qPCR	5	6	7
	Porosity + OM (%)	7	7	7
	Pore-water CH4	7	7	7

Note: Collected sediment cores from an illuminated shallow (10 m water depth) and dark deep site (33 m) were distributed between three light treatments for 10 days in the laboratory: dark (n = 7 cores per site); medium (~50 PAR μmol m⁻² s⁻¹, n = 7 per site); and high light intensity (~100 PAR μmol m⁻² s⁻¹, shallow site n = 7, deep site n = 8). In the text, the treatments are abbreviated shallow dark (SD), shallow medium light (SM), shallow high light (SH), deep dark (DD), deep medium light (DM), and deep high light (DH). The table shows the number of replicate samples (independent sediment cores) for each measured variable per site for each treatment.

MATERIALS AND METHODS Field sampling

Sediment samples were collected using a box corer (surface area: 1000 cm², model 80.100-50, KC Denmark) on board R/V Augusta in the Storfjärden bay area (Finland) in 2021. Samples were taken from a deep dark site on September 27 during mid-day (33 m water depth; Lat 59.8559, Long: 23.26695), and a shallow illuminated site on September 28 (10 m water depth; Lat 59.8521, Long: 23.24495) (Figure 1a). The two sites have previously been characterized during 2018 (station labels: 10 and 12, respectively), with the shallow site having less phosphate and ammonium pore-water concentrations and lower sediment organic matter content (Broman, Bonaglia, et al., 2020; Broman, Sun, et al., 2020). Two box-cores were collected at the deep site and three at the shallow site. The sediment from the boxes was subsampled using acrylic sediment cores (inner diameter 4.6 cm, length 40 cm). The cores contained approximately 20 cm of sampled sediment with overlying bottom water. They were closed at the bottom and top with rubber stoppers and kept at 12°C until the start of the experiment. At both sites, a CastAway CTD (SonTek, USA) equipped with an RBRsolo light logger (RBR, Canada) was used to measure water profiles of temperature, salinity, and PAR light. Dissolved O₂ was measured using a ProODO probe (YSI, USA) probe in the bottom water overlying the sediment inside one core per station. A total of 43 sediment cores (shallow site = 21, deep site = 22 cores) were collected for the light incubation experiment, with an additional seven cores per site sliced (top 0–1 cm sediment surface) directly on the boat to be able to compare our incubations to field conditions.

The top 1 cm sediment layer slice was transferred into a flat 215 mL polypropylene container (207.0215PP, Noax Laboratory); 2 mL were immediately transferred by means of a cut-off 3-mL syringe (Henke-Ject) into a 14 mL gas tight glass vial containing 4 mL 1 M NaOH that was crimped and stored for later GC-FID analyses; the sediment in the jar was then homogenized and 2–3 mL transferred into a 15 mL centrifuge tube (Sarstedt) that was flash frozen in liquid nitrogen (later stored at −80°C and used for RNA extraction); the remaining sediment was transferred into a 50 centrifuge tube (Sarstedt) that was stored at −20°C for DNA extraction, measurement of porosity, and loss on ignition (LOI) analysis to determine % of organic matter.

In addition, a GEMAX gravity corer was used to collect one acrylic core (inner diameter: 80 mm, height: 80 cm), pre-drilled with 1.5 cm diameter holes 2.5 cm apart, for sediment CH₄ profiles at each station. Using cut-off syringes, 10 mL of wet sediment was extracted from each of the holes and immediately transferred to 65 mL glass bottles filled with supersaturated NaCl solution. The bottles were capped with butyl rubber septa and metal screw caps and stored upside down. Within 2 h of sampling, 10 mL of N₂ was injected into the vials to create a headspace with a second needle in place to allow the same volume of sediment slurry to be ejected from the vial. A 1 mL subsample of the equilibrated headspace was measured using a gas chromatograph equipped with a flame ionization detector (Agilent Technologies 7890B). In addition, at each station, water column CH₄ concentrations were measured in real-time using a pump–CTD-system connected to a Water Equilibration Gas Analyzer System (WEGAS), enabling water sampling at high resolution along a vertical profile. WEGAS consists of an automated gas equilibrator and cavity ring-down spectrometer (model G2201-i, Picarro) and has previously been described in more detail (Humborg et al., 2019; Roth et al., 2022).

Experimental design and chemical analyses

The light incubation experiment started on September 29 by distributing the 43 sediment cores from the two sites into two identical incubation chambers (21 in one chamber and 22 in the second one). The cores of each site were evenly distributed between the two chambers. Each chamber was filled with ~95 L unfiltered Storfjärden surface water (6.08 salinity) close to the Tvärminne Zoological Station (Lat 59.8454, Long 23.25160). The cores were fully submersed and aerated with air pumps with connected hoses and air stones. In addition, the water inside each core was aerated to avoid stratification and hypoxia. The chambers were placed in two constant climate rooms and the water bath inside the chambers was kept at 12.5°C throughout the experiment. The oxygen in the water overlying the sediment inside each core was confirmed to be fully saturated throughout the experiment using a ProSolo O₂ probe (YSI, USA). Programmable lights were used to incubate the sediment cores in different light intensities, with two lamps placed above each incubation chambers (Aquarius 120 plant, Aqua Medic, Germany). The lamps were set to a diurnal cycle of 14 h light and 10 h dark, and the light colors white, blue, royal blue, and red were all turned on during light periods (i.e., covering the full PAR spectrum of 400–700 nanometers).

The cores were divided into six treatments, two sites with each having three different light intensities: dark, medium, and high, which corresponded to 0, ~50, and ~100 PAR μmol m⁻² s⁻¹ reaching the sediment surface, respectively (see Table 1 for an overview of the treatments and the number of replicates). In the following, the treatments are abbreviated as: shallow dark (SD), shallow medium light (SM), shallow high light (SH), deep dark (DD), deep medium light (DM), and deep high light (DH). Total absence of light in the dark treatment was achieved by covering the cores in aluminum foil and black plastic, and was confirmed to be 0 PAR μmol m⁻² s⁻¹ by measuring light intensity with a RBRsolo light logger. The light intensities used in the experiment were within the range of euphotic conditions (i.e., 1% of sea surface PAR) previously reported for large areas of the seafloor in the south-west coast of Finland (euphotic water depth range 2.8–18 m) (Luhtala et al., 2013). Surface water PAR ranged between 1000 and 1700 during sample collection in our study. Simulated euphotic conditions were also indicated by the stimulation of photosynthesis in our light treatments based on RNA transcript data (more details in the discussion). The shallow site bottom water had 8 PAR μmol m⁻² s⁻¹ and the deep site 0 PAR μmol m⁻² s⁻¹. The light intensity reaching the sediment surface for each incubated core was measured by placing the RBRsolo light logger at each core location in the chambers at the end of the experiment. The data showed some variability in light intensity within the light treatments, but the light levels in the three treatments remained distinctly different (i.e., 0, ~50, and ~100 PAR μmol m⁻² s⁻¹; Figure S1).

After 10 days, the experiment was ended and the top 1 cm sediment layer was sliced for each core and handled as described above for the field samples, except that 20 mL glass vials with butyl septa were used for the pore-water CH₄ samples (to fit the autosampler of the gas chromatograph), and porosity and LOI analyses were conducted without freezing the samples. Because the DH treatment had one more replicate than the other treatments, that core was sliced and used to obtain one additional sample for DNA and RNA extraction. Porosity was determined from 1 mL bulk sediment by weighing wet and dry sediment (48 h at 60°C), and LOI analysis was conducted by weigh difference after igniting the dry sediment at 550°C for 5 h. The concentration of CH₄ in the pore-water samples were quantified by headspace analysis on a gas chromatograph (GC Trace 1300, Thermo) equipped with an autosampler (TriPlus RSH, Thermo), a nonpolar PLOT column (TracePLOT TG-BOND Q, Thermo), and a flame ionization detector. For calibration, certified standards of 1.86 ppm and 49.82 ppm CH₄ (Air Liquide Gas) were injected. Using the ideal gas law, the ppm concentrations were converted into molar concentrations taking into account sediment porosity.

Nucleic acids extraction, PCR, and sequencing

DNA was extracted from ~0.25 g sediment, including a blank sample containing extra lysis buffer instead of sediment as input material, using the DNeasy PowerSoil Pro Kit (Qiagen) following the provided protocol. DNA was normalized to 10 ng/μL for each sample and the partial 16S rRNA gene was amplified in a first PCR using the primers 341f and 805r (Herlemann et al., 2011) and multiplexed with barcode indexes in a second PCR according to Hugerth et al. (2014) with a few modifications described in Lindh et al. (2015). The final protocols and programs for PCR 1 and 2 are available in Bunse et al. (2016), with the exception that here we used Q5 Hot Start High-Fidelity 2X Master Mix (New England Biolabs) as mastermix. Cleaning of PCR 1 products was conducted using Thermosensitive Alkaline Phosphatase and Exonuclease I (Promega and New England Biolabs, respectively), while PCR 2 products were cleaned using AMPure XP (Beckman Coulter) as previously described in Broman et al. (2019). The samples were then mixed with equimolar DNA concentrations into a library pool and sequenced on a MiSeq flow cell with a 2 × 300 bp setup at SciLife, Stockholm, Sweden.

RNA was extracted from ~2 g sediment using the Rneasy PowerSoil Total Kit (Qiagen) following the provided protocol. DNase treatment was conducted using the TURBO DNA-free kit (Invitrogen) on the eluted RNA to remove any leftover DNA. This was followed by gel electrophoresis to ensure genomic DNA had been removed. A total of three purified RNA samples were sent to SciLifeLab for sequencing. Libraries were prepared with the TruSeq-Stranded mRNA kit (Illumina) excluding the poly-A selection step, and sequenced on a NovaSeq 6000 S4 lane with a 2 × 150 bp setup.

Quantitative reverse transcription PCR (RT-qPCR)

Purified RNA was normalized to 10 ng/μL and reverse transcribed into cDNA using the AccuScript High Fidelity 1st Strand cDNA Synthesis kit (Agilent) with the supplied random hexamer primers. Not all purified RNA samples had sufficient amount of volume left for normalization and cDNA synthesis and therefore only a portion of the samples were analyzed for RT-qPCR (n = 4–7 per treatment; Table 1). The cDNA was then used to investigate if there was a difference in pmoA RNA transcripts between the treatments. RT-qPCR reactions consisted of 1 μL cDNA, 7 μL sterile RNase/DNase-free water, 1 μL of each primer, and 10 μL SYBR Green (LightCycler 480 SYBR Green I Master kit, Roche). Primers and the qPCR protocol used has been described previously (Broman, Sun, et al., 2020). In brief, degenerate pmoA primers targeting a 153 bp long fragment consisted of pmoA_13144881_3-1700181_5_Fw (GAGYGCATCTCAATCAGCTGTACG) and pmoA_13144881_3-1700181_5_Rv (GTCCAGAAATCCCAGTCACCRC) which were originally designed and named based on metagenome assembled contig regions classified as pmoA genes (for more details see Broman, Sun, et al., 2020). The pmoA Ct values were analyzed as relative to 16S rRNA Ct values, and the 16S rRNA gene primers 515F and 805R were used with qPCR (Herlemann et al., 2011; Parada et al., 2016). The qPCR program consisted of an initial denaturation of 95°C for 5 min, and 45 cycles of denaturation 95°C 30 s, annealing 60°C 30 s, and elongation 72°C for 15 s. The ramp rate was 4.40°C/s except for the annealing step which was 2.20°C/s. The qPCR setup also included DNase treated RNA control samples (i.e., not cDNA synthesized) to further confirm there was no genomic DNA contamination. The final Ct values were normalized against 16S rRNA by calculating ΔCt (pmoA Ct – 16S rRNA Ct) as well as using the 2^−ΔΔCt method by Livak and Schmittgen (2001) with the SM treatment considered as a control for the shallow site and the DD treatment as a control for the deep site.

Bioinformatics

The sequenced 16S rRNA gene reads were demultiplexed by the sequencing facility and the delivered raw data were processed using R 4.1.1 and the DADA2 v 1.21.0 package (Callahan et al., 2016; R Core Team, 2021). To remove leftover primer sequences and quality trim the reads, the filtering step was run with the following options: maxEE = 2, truncQ = 2, maxN = 0, rm.phix = TRUE, truncLen = c(290, 210), and trimLeft = c(21, 22). That the quality filtered reads had sufficient length, quality, and that no Illumina adapters remained was confirmed by using FastQC 0.11.9 and MultiQC 1.11 to merge the quality reports (Andrews, 2010; Ewels et al., 2016). The error modeling was run with the settings: bases = 1e8; merging step with minOverlap = 10 and maxMismatch = 1; and chimera removal with method = “consensus” and minFoldParentOverAbundance = 4. The ASV sequences were annotated against the SILVA SSU NR 99 v 138.1 database (Quast et al., 2013). Read counts for ASVs sequences with the exact same taxonomic label were summed and are reported in Data S1. Chloroplast and mitochondrial read counts were removed from the dataset. A few samples had failed the sequencing and were removed from the dataset (<5000 counts; five samples: one shallow field, two DD, one DM, and the DNA extraction blank; see list of samples excluded in Data S1). Finally, the counts per sample were normalized as relative abundances (%) and analyzed in the software Explicet 2.10.5 (Robertson et al., 2013).

The RNA-seq yielded on average 85.1 million paired-end reads per sample (min: 61.1, max: 102.3). Illumina adapters were removed using SeqPrep 1.2 with default settings targeting the adapter sequences (St John, 2011). Any leftover PhiX control sequences were removed by mapping the data to the PhiX genome (NCBI Reference Sequence: NC_001422.1) using bowtie2 2.3.5.1 (Langmead & Salzberg, 2012). Quality trimming was conducted with Trimmomatic 0.39 using settings LEADING:20, TRAILING:20, and MINLEN:80 (Bolger et al., 2014). That the quality of the trimmed data was sufficient was verified using FastQC 0.11.9 (Andrews, 2010) to generate reports combined using MultiQC 1.12 (Ewels et al., 2016). After quality trimming each sample had on average 83.9 million paired-end reads (min: 59.9, max: 101.2), an average read length of 147 bp (min: 146, max: 148), and each read an average Phred score of 36 (min: 35, max: 36). Full details of which sediment cores were used for RNA sequencing and number of reads per sample throughout the bioinformatic analysis is available in Data S2.

Taxonomic classification of the RNA-seq data was conducted by extracting small subunit rRNA (SSU rRNA) reads from the quality trimmed reads (trimmomatic paired without unpaired (PwU) reads) using SortMeRNA 4.3.6 with settings --out2 --paired_out using the SILVA SSU dataset within the software supplied database (smr_v4.3_default_db.fasta) (Kopylova et al., 2012). These SortMeRNA settings extracts SSU rRNA sequences from both the R1 and R2 reads and only keeps the reads if both pairs classify as being SSU rRNA. The SSU rRNA reads were then taxonomically classified using Kraken2 2.0.9 (Wood et al., 2019) against the SILVA SSU database (download date: 2022, September 1). Kraken2 was run using default settings with the – paired setting to account for both R1 and R2 reads. The Kraken2 reports were then used with the software Bracken 2.7 (Lu et al., 2017) to be able to classify reads to genus level and compare the relative abundances (%). Bracken was run using the settings: −r 150 −l G −t 10, which denotes a read length of 150 bp, classification at genus level, and a minimum threshold of 10 counts per genus. The bracken reports for each sample were combined into biom format using the python package kraken-biom 1.0.1 with settings: – fmt hdf5 – max D – min G (Dabdoub, 2016), and converted to a tab delimited table using the python package biom format 2.1.7 (McDonald et al., 2012). Prokaryotic classifications (i.e., 16S rRNA) were extracted from the dataset and consisted on average of 30.4 million counts per sample (min: 20.5, max: 35.1). The final data were then analyzed and normalized as relative abundance (%) using the software Explicet 2.10.5 (Robertson et al., 2013).

The functional annotation of the RNA-seq data followed the outline of the SAMSA2 pipeline (Westreich et al., 2018). In more detail, the quality trimmed PwU reads were merged using PEAR 0.9.10 with default settings (Zhang et al., 2014). This yielded on average 54.9 million merged reads per sample (min: 40.9, max: 67.2) representing on average 64.6% merging rate, with an average read length of 215 bp (min: 210, max: 219). For each sample, the paired reads were concatenated with the PEAR notCombined forward reads as recommended by the SAMSA2 pipeline (Westreich et al., 2018). Ribosomal RNA reads were excluded from the data by using SortMeRNA 4.3.6 with default settings to extract only non-rRNA reads using the software supplied database (smr_v4.3_default_db.fasta) (Kopylova et al., 2012). The non-rRNA data consisted of ~3% of the quality trimmed data which represented on average 2 million reads per sample. This was followed by functional annotation using DIAMOND 2.0.14 against the NCBI NR database (download date: October 2, 2022) using an e-value threshold of 1e⁻¹⁰. The DIAMOND output.daa files were used to link results with the KEGG database (MEGAN database: megan-map-Feb2022) by using the daa-meganizer tool with default settings that is supplied with MEGAN 6 Ultimate Edition 6.24.1 (Bağcı et al., 2021; Huson et al., 2007). This was followed by merging the results into one combined file using the MEGAN tool compute-comparison (setting: absolute counts) which was imported into MEGAN and analyzed further. The MEGAN software was then used to extract all KEGG KO classifications and sample counts were normalized between samples as counts per million (CPM, i.e., relative proportion × 1 million). Finally, because of their similar sequence homology and shared function the KEGG database joins pmoABC and ammonia monooxygenase genes (amoABC) into the same KEGG KO classifications, we therefore also investigated to which proportion pmo transcripts were represented within these KEGG KOs. Reads classified as pmoA/amoA, pmoB/amoB, and pmoC/amoC (KEGG KOs 10944, 10945, and 10946, respectively) were extracted from the meganized DIAMOND.daa files using the MEGAN supplied tool read-extractor (options: -c KEGG -n 10944, 10945, 10946). The extracted reads were then annotated against the curated UniProtKB-SwissProt database (database date: August 7, 2022) using BLASTX (Blast version 2.12.0+) (Altschul et al., 1990) with an e-value threshold of 1e⁻¹⁰. Only results for pmoAB and amoAB were reported as only these pmo/amo protein sequences are available in the UniProtKB-SwissProt database. The number of blast hits per protein in each sample was used to calculate the proportion of pmoAB of pmoAB/amoAB.

Statistics

To estimate the difference in the number of methanotrophic taxa between treatments, genus richness was calculated for methanotrophs based on the lowest taxonomic level (down to genus, analysis based on the observed number of taxonomic groups) using Past 4.07b (Hammer et al., 2001). Because some samples (shallow site) had very few counts classified as methanotrophs, it was not possible to rarefy the data to the lowest sample size and instead the relative abundance (%) was used as an indication for presence or absence. Non-metric multidimensional scaling (NMDS) analysis were plotted on ASV level and based on the Bray–Curtis dissimilarity index, alongside PERMANOVA (9999 permutations) and pairwise-comparison tests with p-adjusted Bonferroni-corrected values, using the software Past. Additional statistics were conducted in R 4.1.1 (R Core Team, 2021) using ANOVA tests when data met assumptions (normal distribution analyzed with Shapiro–Wilk test and homogeneity of variance with Levene tests), in other cases the data were boxcox transformed using the MASS R 7.3-54 package (Ripley et al., 2013), or nonparametric Dunn tests using the R package dunn. test 1.3.5 (Dinno & Dinno, 2017). Linear regression models were constructed in R using the lm function. The model consisted of CH₄ pore-water concentrations (boxcox transformed) as a dependent variables, and site, OM %, light intensity, relative abundance of methanotrophs (boxcox transformed), and RT-qPCR pmoA ΔCT values (boxcox transformed) as independent variables (one light value per sample measured at the location of the core in the incubation chamber; Figure S1). The effect size of each independent variable was estimated using the eta_squared function of the effectsize package (Ben-Shachar et al., 2020) with the argument partial set to F. This function represents an estimate of how much variance in the response variables is accounted for by the explanatory variables (Ben-Shachar et al., 2020). This value was then interpreted according to the rules described in (Cohen, 1992). When the linear model was used to explore which variables influenced OM %, relative abundance methanotrophs, or pmoA ΔCT values they were interchanged with CH₄ as a dependent variable. The number of replicates (individual sediment cores) for each analysis is mentioned in Table 1. Values are reported as the mean ± one standard deviation.

RESULTS Field conditions

The deep site had a bottom water temperature of 12.2°C, 6.08 salinity (PSU), 9.98 dissolved O₂ mg L⁻¹, and received 0 PAR μmol m⁻² s⁻¹, while the shallow site bottom water had a temperature of 11.7°C, 5.98 salinity, 9.67 O₂ mg L⁻¹, and received 8.2 PAR μmol m⁻² s⁻¹. Temperature measurements in the water column at each site showed isotherm profiles with 11.9°C degrees in the water, and salinity was approximately 6 (PSU) throughout the water column (Figure 1b). CH₄ concentrations in the water column increased with water depth at both stations. Bottom water CH₄ concentrations reached ~40 nM at the shallow site and ~90 nM at the deep site (Figure 1b). CH₄ concentrations in the sediment surface (top 1 cm) were significantly higher at the deep site (8.6 ± 5.7 μM, mean ± SD) compared with the shallow site (1.3 ± 0.3 μM) (n = 7 per site, One-Way ANOVA of log transformed values, df = 13, F = 38.0, p < 0.0001; Figure 1c & Data S3). Sediment profiles of CH₄ concentrations showed an increase of CH₄ concentrations with sediment depth at both sites, and had the highest concentrations in the deep site (~5000 μM at 24 cm sediment depth compared with ~33 μM at 22 cm in the shallow site; Figure 1c). Organic matter (LOI %) content in the top 1 cm sediment surface was significantly higher at the deep site (13% ± 2%) when compared with the shallow site (10% ± 2%) (One-Way ANOVA, df = 13, F = 6.2, p < 0.05).

Effect of light on CH₄ concentrations and organic matter content

After 10 days of diurnal light and fully dark incubations, the results showed that there was no significant difference between the three treatments (i.e., dark, medium light, and high light) from each site for both CH₄ pore-water concentrations and organic matter (OM) % in the 0–1 cm sediment surface (Nonparametric Dunn tests; Benjamini-Hochberg adjusted p-values >0.05 for all pairwise comparisons between treatments within the same sampling site) (Figure 2a,b). Linear regression models showed that the factors “site,” that is, shallow or deep, and “OM %” were the only significant independent variables explaining the CH₄ pore-water concentrations in the sediment (site t-value −3.398, p < 0.01; OM % t-value −2.202, p < 0.05). CH₄ pore-water concentrations and OM % were significantly correlated to each other when tested for all samples (Pearson correlation, r = 0.86, p < 0.0001).

FIGURE 2. (a) CH4 concentrations and (b) organic matter (%) in the sediment surface (top 1 cm) after the light experiment (n = 7 per treatment). Light intensities denote: Dark = 0, Medium = ~50, and High = ~100 PAR μmol m−2 s−1. The middle line in the boxplots represents the median, and the top and bottom of the box shows the first and third quartiles, while the whiskers show the maximum and minimum values. The circles denote outliers (≥1.5 × box length).

Effect of light on aerobic methanotrophic communities

Type I and Type II methanotrophic lineages were extracted from the 16S rRNA gene metabarcoding dataset and analyzed separately. In addition, other known methanotrophic groups, such as the phyla Methylomirabilota (previously named NC10), the Verrucomicrobiota order Methylacidiphilales, and the Alphaproteobacteria family Methyloligellaceae, were included (Data S4). Methyloligellaceae are methylotrophic bacteria, but at least one known genus, Methyloceanibacter, can oxidize methane (Vekeman et al., 2016). Therefore, it could not be excluded that unclassified Methyloligellaceae were capable of methane oxidation. Type I and II methanotrophs belong to Gammaproteobacteria and Alphaproteobacteria, respectively. Gammaproteobacteria represented ~18% and Alphaproteobacteria ~4% of the whole microbial community (Figure S2). The relative abundance (%) of methanotrophs was higher in samples from the deep site compared with the shallow site (pairwise Dunn tests between treatments, p < 0.05; Figure 3). Specifically, the shallow site had 0.1%–0.7% methanotrophs, while the deep site had 0.5%–1.7% methanotrophs (Figure 3). Furthermore, there was no significant difference in the relative abundance of methanotrophs between light treatments for both the shallow and deep site (pairwise Dunn tests, p > 0.05). A linear model including the relative abundance of all methanotrophs (boxcox transformed) as a dependent variable showed that the factor “site” (t-value −6.906, p < 0.001) and RT-qPCR pmoA ΔCt values had a significant effect (t-value −2.445, p < 0.05, more details about the RT-qPCR results in the next section). These results indicate that light had no influence on the relative abundance of methanotrophs in the sediment surface throughout the experiment. The genus richness calculated on the aerobic methanotrophic groups (lowest taxonomic classification down to genus) showed that that the deep site had significantly more methanotrophic groups than the shallow site (10.7 ± 2.1 compared with 5.0 ± 1.3; One-Way ANOVA, df = 12, F = 32.75, p < 0.001; Figure S3), as well as when tested on ASV level for all methanotrophs (67.4 ± 20.1 compared with 21.5 ± 4.9; One-Way ANOVA, df = 12, F = 29.45, p < 0.001). However, there was no difference in the number of methanotrophic ASVs between the light treatments for each site (Pairwise Dunn tests between all treatments for each site, p > 0.05).

FIGURE 3. Relative abundance (%) of all taxa within classified methanotrophic groups in the 16S rRNA gene data. The x-axis shows the data from the field samples as well as the light experiment. Shallow site n = field 6, dark 7, medium 7, high 7. Deep site n = field 7, dark 5, medium 6, and high 8.

Non-metric multidimensional scaling analysis based on Bray–Curtis dissimilarity index of 16S rRNA gene methanotroph data alongside PERMANOVA tests showed that the community composition of methanotrophs was different between the two sites and between light treatments of the deep site (Pseudo-F = 3.901, p = 0.0001), but not for the shallow site (Pseudo-F = 1.167, p = 0.219; Figure 4). The deep site treatments were tested with Pairwise PERMANOVA tests and showed that the deep high light (DH) treatment was different from the field samples and the two other treatments (field p = 0.0018, deep dark (DD) p = 0.0174, deep medium light (DM) p = 0.024).

FIGURE 4. The difference in community composition of methanotrophs in the field samples and the various light treatments for the 16S rRNA gene data. The NMDS is based on the Bray–Curtis dissimilarity index using the relative abundance of the methanotrophs as input data. PERMANOVA (9999 permutations) were conducted including all treatments for each site. Shallow site samples are colored yellow, while deep site samples are colored blue. The asterisk denotes that the DH treatment was significantly different compared with the other deep site treatments (pairwise PERMANVOA test).

Looking in more detail on the taxonomy and distribution of the various methanotrophic groups in the 16S rRNA gene data, it was found that dominant methanotrophs included unclassified taxa within the Type I methanotrophs Methylococcales family Methylomonadaceae (27.7% ± 16.9% of all methanotrophs, average of all samples), Type II Methyloligellaceae previously described methanotrophic genus Methyloceanibacter (37.4% ± 17.0%), and Type II unclassified Methyloligellaceae (29.1% ± 16.6%; Figure 5). Statistical testing of the relative abundance for each methanotrophic group showed that the shallow site had no differences between treatments (pairwise Dunn tests, p > 0.05 for all methanotrophic groups). In the deep site, dominant groups included Methyloceanibacter (35.1% ± 6.9% of all methanotrophs), unclassified Methyloligellaceae (20.5% ± 7.3%), and Methylococcales that included several classified genera (44.2% ± 27.0%; Figure 5). These genera included e.g. Crenothrix, Methylobacter, Methylomonas, and Methyloprofundus (Figure 5). Pairwise Dunn tests showed that the genera Crenothrix and Methylomonas had a higher relative abundance in DH compared with the DD treatment (Crenothrix 0.18% ± 0.06% vs. 0.10% ± 0.03% and Methylomonas 0.05 ± 0.02 vs. 0.02% ± 0.01%; values denote relative abundance of all prokaryotes; p < 0.05). However, these small changes in the 16S rRNA gene amplicon relative abundance data were not reflected in the RNA-seq data (more details below). The sediment cores sliced for RNA extraction and total RNA-seq (n = 3 per treatment) showed that there was no difference per treatment between the different methanotrophic groups for both sites (pairwise Dunn tests, p > 0.05, tested for each group shown in Figure 6). Similarly to the 16S rRNA gene amplicon data, Type I Methylococcales genera, such as Crenothrix, Methylobacter, and Methyloprofundus were dominant methanotrophs in the sediment for the deep site, as well as the deep site have significantly higher relative abundance of methanotrophs than the shallow site (5.36% ± 0.63% compared with 0.30% ± 0.03%, respectively; One-Way ANOVA, df = 17, F = 584.9, p < 0.0001, n = 9 per site; Figure 6 & Data S5). The RNA-seq showed that the methanotrophs in the deep site had a relative abundance of approximately three times higher compared with methanotrophs in the 16S rRNA gene amplicon data, indicating their high activity in these sediments. In contrast, shallow site methanotrophs instead had a relative abundance that was approximately six times lower in the RNA-seq data compared with the 16S rRNA gene data (Figure 3 compared with Figure 6). However, no significant effect of light was found on the relative abundance of methanotrophic in the RNA-seq data (more details below).

FIGURE 5. Relative abundance (%) of the methanotrophic groups for the 16S rRNA gene data. Each row shows the community composition from one individual sediment core, and the numbers on the y-axis denote the core replicate number. Bold text denotes taxa with a high relative abundance.

FIGURE 6. Relative abundance (%) of methanotrophic groups classified with the software combo Kraken2 + Bracken2 of the 16S rRNA sequences extracted from the RNA-seq data. Each row shows the community composition from one individual sediment core, and the y-axis shows the site, treatment, and sediment core number sliced and used for RNA-seq. Bold text denotes taxa with a high relative abundance.

Effect of light on methane monooxygenase transcripts

The RT-qPCR data showed that the deep site had a higher proportion of pmoA transcripts (normalized for 16S rRNA) when compared with the shallow site (7.13 ± 0.57 ΔCt compared with 14.13 ± 1.01 ΔCt, respectively; Table 2). This was also indicated with pmoA ΔCt values correlating negatively with both CH₄ pore-water concentrations and the relative abundance of methanotrophs which were both higher in the deep site samples (Pearson's r = −0.90 and −0.86, respectively, boxcox transformed data). Note that lower ΔCt values indicate earlier amplification on the qPCR instrument and therefore a higher proportion of pmoA transcripts. However, there was no difference in pmoA ΔCt values between the treatments for both of the sites (pairwise Dunn tests, p > 0.05; Table 2). A Linear model with pmoA ΔCt values (boxcox transformed) showed that site, relative abundance of methanotrophs (boxcox transformed), and light were significant explanatory variables (site t-value 4.082, p < 0.001; relative abundance of methanotrophs t-value −2.445, p < 0.05; light t-value 2.793, p < 0.01). However, site explained the large majority of variability in the model (as indicated by the effectsize Eta2 value and interpretation: site 0.95 large, relative abundance of methanotrophs 0.0079 very small, and light 0.0076 very small). Furthermore, in the linear model CH₄ pore-water concentration was not a significant variable, and light was not found to be correlated with pmoA ΔCt values (r = 0.11, p = 0.54). The linear model was also tested with treatment as a factor (instead of light as an independent continues variable) and only the deep site DH treatment was significant (t-value 1.819, p < 0.05), explained by an average 1.4%–2.9% increase in pmoA ΔCt values compared with the DM and DD treatments (Table 2). The RT-qPCR data were also used to calculate 2^−ΔΔCt values and Dunn tests indicated that there were no differences between treatments (pairwise Dunn tests, p > 0.05; Table 2). The 2^−ΔΔCt method reports the fold change compared with a control group, and here the shallow medium light (SM) was used as a control for the shallow site and the DD treatment as a control for the deep site. The results indicate that the proportion of pmoA transcripts in the treatments was not different when tested against the control groups.

TABLE 2 Results from the RT-qPCR analysis.

Site	Treatment	pmoA Ct	16S rRNA Ct	ΔCt	2−ΔΔCt	n
Shallow	Dark	28.7 ± 0.8	14.9 ± 0.7	13.8 ± 1.2	2.3 ± 1.8	6
	Medium	29.4 ± 1.2	14.7 ± 0.6	14.6 ± 0.8	1.1 ± 0.7	4
	High	28.7 ± 1.0	14.5 ± 0.5	14.2 ± 1.0	1.6 ± 0.9	5
Deep	Dark	23.2 ± 0.4	16.2 ± 0.4	7.0 ± 0.4	1.0 ± 0.2	5
	Medium	23.0 ± 0.4	15.9 ± 0.3	7.1 ± 0.5	1.0 ± 0.3	6
	High	23.2 ± 0.5	16.0 ± 0.3	7.2 ± 0.8	1.0 ± 0.5	7

Note: The table shows results based on relative quantification. ΔCt denotes normalized pmoA transcripts in relation to 16S rRNA (pmoA Ct – 16S rRNA Ct). The 2^−ΔΔCt shows the difference in fold change compared to a control group, and here SM was used as a control for the shallow site, and the DD treatment as a control for the deep site. The Ct values are based on the average of duplicate technical replicates, and n shows the number of biological replicates (i.e., individual sediment cores). Note that lower Ct values indicate a higher abundance of transcripts. The values show mean ± SD.

The functional annotation of the non-rRNA sequences from the total RNA-seq data showed that the number of transcripts classified as pmoABC/amoABC in the KEGG database (KEGG KOs 10944 (pmoA/amoA), 10945 (pmoB/amoB), and 10946 (pmoC/amoC)) did not significantly differ between the light treatments when tested for each site (One-Way ANOVA tests, p > 0.05; data normalized as counts per million (CPM); Figure 7a & Data S6). Because the RNA-seq data only consisted of three samples per treatment, the data were not included in a linear model. The deep site had significantly higher number of these transcripts when compared with the shallow site (18,313 ± 2423 compared with 1299 ± 723 CPM; Figure 7a). There was a negative correlation between pmoABC/amoABC transcripts and pmoA ΔCt values (r = −0.96, p < 0.001), and a positive correlation with relative abundance of methanotrophs (r = 0.91, p < 0.001) and CH₄ concentrations (r = 0.91, p < 0.001). However, light was not significantly correlated with pmoABC/amoABC transcripts (r = 0.02, p = 0.95). Based on classifying the pmoAB/amoAB sequences against the UniProtKB-SwissProt it was found that the deep site was dominated by sequences classified as pmoAB transcripts, that represented on average 96% of pmoAB/amoAB. In contrast, the shallow site sequences classified as pmoAB represented on average 57% of pmoAB/amoAB (Figure 7b & Data S6). Finally, sMMO transcripts were missing (K16158 mmoY, K16160 mmoB, K16159 mmoZ) or had low CPM values (K16157 mmoX, K16161 mmoC both, <10 CPM; Data S6) indicating that microorganisms carrying sMMO genes likely did not play a significant role in methane oxidation.

FIGURE 7. (a) shows results from the RNA-seq data of the number of RNA transcripts summed for KEGG KOs classified as pmoA/amoA, pmoB/amoB, and pmoC/amoC (KEGG KOs 10944, 10945, and 10946, respectively). The data show normalized counts as counts per million values (CPM). (b) shows the representation of pmoAB transcripts (%) within the KEGG KOs classified as pmoAB/amoAB.

DISCUSSION

Here, we show that the light intensities tested had little or no inhibitory effect on the relative abundance of methanotrophs, RNA transcripts related to methane oxidation, and did not change the CH₄ porewater concentrations in the sediment surface. These findings were observed for both the shallow site (10 m water depth) that naturally receives light to the seafloor and the deeper dark site (33 m) that is not illuminated. Our results bring some clarity regarding the role of light in mediating methanotrophic activity in coastal surface sediments.

The interplay between light and methanotrophs is, however, complex, as indicated by previous studies that have both found inhibitory (Dumestre et al., 1999; Murase & Sugimoto, 2005) and stimulatory effects on methanotrophs (King, 1990; Oswald et al., 2015; Savvichev et al., 2019). In studies that exemplifies the latter for sediments (King, 1990); King et al. (1990) found that illumination of sediment increased the oxygen penetration depth due to photosynthesis that supplied oxygen to the sediment, which in turn stimulated aerobic methane oxidation. However, it is uncertain how photosynthetically driven oxic CH₄ production, recently shown to be prevalent in aquatic ecosystems (Bizic, 2021; Bižić et al., 2020) might have contributed to measured changes in methane oxidation. In any case, such potential stimulatory effects were not found in our sediments, but King et al. (1990) used a high light intensity (700–1000 PAR) with the goal to saturate benthic photosynthesis, or incubated sediment covered with an algal mat being illuminated for a brief time (3–8 h) (King, 1990), while in our study, we were aiming for a more realistic scenario of the amount of light possibly reaching the sediment surface (including a diurnal light cycle and several days of incubation). For example, in a previous study, several stations in the same bay (water depth range: 10–45 m) never had measured PAR values >100 μmol m⁻² s⁻¹ in the bottom waters of the shallow stations (Broman, Sun, et al., 2020). Our light incubations were nevertheless, sufficient for photosynthetic stimulation as confirmed by our RNA-seq data, with RNA transcripts attributed to the KEGG category photosynthesis being 20% (SH treatment) and 45% (DH treatment) higher compared to the dark treatments (Data S6). As we could not detect a difference in the relative abundance or activity of methanotrophs between the treatments, it is unlikely any O₂ production by benthic microalgae had a significant role on the methanotrophs in our experiment. Likely a higher illumination is needed for such effects (King, 1990; King et al., 1990), which might be less relevant to what is found under natural settings (as discussed above). Interestingly, the linear model of the RT-qPCR data showed a small but significant effect of light on the transcription of pmoA, driven by the somewhat higher ΔCt values in DH treatment (on average 1.4%–2.9% higher compared with DM and DD). This minor inhibition, or lack thereof, was also suggested by all other variables analyzed, including RNA-seq, CH₄ pore water concentrations, and relative abundance of methanotrophs based on 16S rRNA gene amplicon sequencing.

Contrary to the above-mentioned studies that found a potential stimulatory effect on methanotrophy in sediment (King, 1990; King et al., 1990), Sieczko et al. (2020) reported mirrored patterns between PAR and lake water CH₄ emissions, with increasing PAR having higher CH₄ emissions. However, the authors were not able to find any clear correlations between light and CH₄ day/night flux ratios (Sieczko et al., 2020). Potentially, light inhibition of aerobic methanotrophs might be occurring close to the surface waters where illumination is high (>1000 PAR μmol m⁻² s⁻¹ in our field measurements), explaining why light inhibition of methanotrophs in the water column has been previously indicated (Dumestre et al., 1999; Morana et al., 2020; Murase & Sugimoto, 2005). However, it is uncertain how much this would contribute to overall CH₄ loss, as such inhibition is likely to occur close to the sea–air interface. We are aware that oxygen reaches a few mm in the sediment surface, and might thus harbor methanotrophs below the sediment surface exposed to light (Myllykangas et al., 2020). However, here we collected 0–1 cm slices which would include all methanotrophs in the oxic layer, and did not detect a significant difference in methanotroph abundance or activity between the treatments, including CH₄ pore-water concentrations. These findings indicate that light has no significant influence on methane oxidation or CH₄ pore-water concentrations in coastal sediments. This implies that the reported variability in diurnal CH₄ emission from shallow waters, such as inland and coastal environments, are explained by other factors than light inhibition of methanotrophs (Roth et al., 2022; Sieczko et al., 2020). For example, temperature is a strong driver of microbial CH₄ production (Yvon-Durocher et al., 2014), which has been confirmed by Roth et al. (2022) in coastal waters. However, the water temperature-CH₄ production relationship seems to be weaker or insignificant on a diurnal scale when measured in lakes (Sieczko et al., 2020). Further studies are therefore needed to deduce which factors drive CH₄ variability in aquatic systems.

The 16S rRNA gene amplicon data indicated that Methyloceanibacter and unclassified Methylomonadaceae were the most abundant methanotrophs at the shallow site. However, there was no difference between light treatments, read counts were low, and according to the RNA-seq these organisms contributed only little to the CH₄ oxidation. At the deep site the methanotroph relative abundance was higher, with Crenothrix, Methylobacter, and Methyloprofundus being more prominent, which has also been shown previously at the deep site (Broman, Sun, et al., 2020). However, similarly to the shallow site, there was no difference in relative abundance between the treatments in the RNA-seq data. These findings indicate that site (i.e., shallow vs. deep) had a stronger effect on the aerobic methanotrophic community rather than light availability. The data also showed that methanotrophs carrying sMMO had an insignificant role in CH₄ oxidation. These findings are in accordance with a previous field study that showed increased methanotrophic activity at the deep site compared with the shallow site (Broman, Sun, et al., 2020). Coastal waters (<50 m) have the highest CH₄ emissions in the marine environment (Weber et al., 2019). Even within these coastal waters there is a high variability in CH₄ water concentrations between shallow inshore and deeper coastal offshore waters (Borges et al., 2016; Broman, Sun, et al., 2020; Humborg et al., 2019; Osudar et al., 2015). Compared with shallow inshore waters, the findings of our study and above-mentioned further indicate that deep offshore coastal waters (>30 m) have higher CH₄ concentrations in the bottom water and sediment, as well as a higher methanotrophic activity in the sediment surface. Even though much of the CH₄ in these areas is biologically oxidized (Mao et al., 2022), large CH₄ emissions can occur during water column mixing events, when CH₄-rich bottom waters are brought to the surface (Bonaglia et al., 2022). Likely this lower benthic methanotrophic activity in shallow areas is related the lower concentration of pore-water CH₄ limiting methanotrophic growth. To relate these findings to light, we have shown that the low methanotrophic activity on the sediment surface of shallow coastal areas is unlikely to be an effect of light inhibition.

Coastal ecosystems are exposed to multiple anthrophonic pressures, such as eutrophication, pollution, and climate change (Howarth & Marino, 2006; Pan et al., 2013). Some of these pressures will decrease the light availability in the water column, such as increased temperature and nutrients that enhance phytoplankton growth (Rabalais et al., 2009); increased water brownification due to higher pluviosity and consequent higher riverine inputs of terrestrial-derived dissolved organic carbon (Andersson et al., 2018). Additionally, other factors like increased wind-driven sediment resuspension can enhance water turbidity (Mi et al., 2019). For such reasons, light penetration has decreased in the Baltic Sea during the last century, with an estimated 13%–17% light attenuation during summer being explained by enhanced algal blooms (Fleming-Lehtinen & Laamanen, 2012). Because our results showed that light did not increase CH₄ pore-water concentrations or inhibit methanotrophs in the sediment for both tested sites, it can be expected that further changes in light attenuation in both inshore and offshore coastal waters will not significantly affect aerobic methanotrophs in the sediment surface. However, eutrophication influences the CH₄ dynamics by, for example, supplying organic carbon and decreasing oxygen concentrations in the sediment (Wallenius et al., 2021), as well as driving community shifts in primary producers and enhancing blooms that could increase oxic CH₄ production (Bizic, 2021; Bižić et al., 2020). These findings help answer an unresolved question regarding the influence of light on CH₄ variability in aquatic environments.

AUTHOR CONTRIBUTIONS

EB conceptualized the idea, designed and coordinated the study, sampled in the field, performed the laboratory experiment, helped with molecular laboratory work, conducted bioinformatics plus data analyses, and drafted the manuscript. RB conducted molecular laboratory work, analyzed molecular data, and gave feedback on the manuscript. DD sampled in the field, conducted sediment methane profiling in the field, performed the laboratory experiment, conducted chemistry analyzes, and gave feedback on the manuscript. FR sampled in the field, measured water column parameters in the field including real-time methane concentrations, performed the laboratory experiment, conducted chemistry analyzes, and gave feedback on the manuscript. CH contributed to the design of the study, developed the method for real-time in situ methane measurements, and gave feedback on the mansucript. AN contributed to the design of the study, helped coordinate the study, sampled in the field, and gave feedback on the manuscript. TJ helped coordinated the study and perform the laboratory experiment, and gave feedback on the manuscript. SB helped design the study, conducted pore-water methane analyses, analyzed chemistry data, and gave feedback on the manuscript. FN helped design and coordinate the study, sampled in the field, performed the laboratory experiment, conducted data analyzes, and helped draft the manuscript.

ACKNOWLEDGMENTS

The study was funded by Walter and Andrèe de Nottbeck Foundation (application: 20220001) and FORMAS (grant no: 2020-02304) given to EB. We wish to thank Laura Kauppi for the support in the field plus laboratory, Jaana Koistinen for the support in the laboratory. We also thank Markus Olsson for initial testing and conducting a pilot experiment. The authors also acknowledge support from the National Genomics Infrastructure in Stockholm funded by Science for Life Laboratory, the Knut and Alice Wallenberg Foundation and the Swedish Research Council, and SNIC/Uppsala Multidisciplinary Center for Advanced Computational Science for assistance with massively parallel sequencing and access to the UPPMAX computational infrastructure. The bioinformatic analyses were enabled by resources in project SNIC 2022/22-391 provided by the Swedish National Infrastructure for Computing (SNIC) at UPPMAX, partially funded by the Swedish Research Council through grant agreement no. 2018-05973.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT

The data that support these findings are available in the manuscript and supplemental files. Supplementary Data files have been uploaded to figshare (see [dataset] Broman, 2023). The raw 16S rRNA gene sequencing data has been uploaded to the NCBI BioProject PRJNA799903 ([dataset] Broman, 2022a). The raw RNA-seq data has been uploaded to the NCBI BioProject PRJNA886485 ([dataset] Broman, 2022b).