Microorganisms oxidize glucose through distinct pathways in permeable and cohesive sediments

Abstract In marine sediments, microbial degradation of organic matter under anoxic conditions is generally thought to proceed through fermentation to volatile fatty acids, which are then oxidized to CO2 coupled to the reduction of terminal electron acceptors (e.g. nitrate, iron, manganese, and sulfate). It has been suggested that, in environments with a highly variable oxygen regime, fermentation mediated by facultative anaerobic bacteria (uncoupled to external terminal electron acceptors) becomes the dominant process. Here, we present the first direct evidence for this fermentation using a novel differentially labeled glucose isotopologue assay that distinguishes between CO2 produced from respiration and fermentation. Using this approach, we measured the relative contribution of respiration and fermentation of glucose in a range of permeable (sandy) and cohesive (muddy) sediments, as well as four bacterial isolates. Under anoxia, microbial communities adapted to high-energy sandy or bioturbated sites mediate fermentation via the Embden–Meyerhof–Parnas pathway, in a manner uncoupled from anaerobic respiration. Prolonged anoxic incubation suggests that this uncoupling lasts up to 160 h. In contrast, microbial communities in anoxic muddy sediments (smaller median grain size) generally completely oxidized 13C glucose to 13CO2, consistent with the classical redox cascade model. We also unexpectedly observed that fermentation occurred under oxic conditions in permeable sediments. These observations were further confirmed using pure cultures of four bacteria isolated from permeable sediments. Our results suggest that microbial communities adapted to variable oxygen regimes metabolize glucose (and likely other organic molecules) through fermentation uncoupled to respiration during transient anoxic conditions.


Introduction
In the absence of oxygen, microbial oxidation of organic matter is initiated by hydrolysis of macromolecules into smaller constituents (e.g.sugars, fatty acids, amino acids), which are then fermented to dissolved inorganic carbon (DIC), molecular hydrogen (H 2 ), alcohols, and volatile fatty acids (VFAs).Heterotrophic bacteria largely use one of three fermentation pathways for degradation of glucose: the Embden-Meyerhof-Parnas (EMP), pentose phosphate (PP) or Entner-Doudoroff (ED) pathways (Fig. 1) [1].EMP fermentation is generally most widespread and active [1], with recent studies indicating at least 90% of obligate and facultative anaerobes use the EMP pathway [2].After fermentation, the reduced compounds produced (VFAs, alcohols, and H 2 ) are rapidly oxidized by respiring bacteria, i.e. terminal respiration coupled with fermentation (Fig. 1).This paradigm has been relatively well developed and studied in cohesive sediments (i.e.muds and silts), which typically have stable physical and redox regimes that allow a close coupling between fermenting and respiring bacteria, e.g.Schulz and Zabel [3].
In contrast to cohesive sediments, microorganisms in permeable and bioturbated sediments experience dynamic redox regimes.In permeable sediments (i.e.sands and gravels), wave oscillations, sediment movement, and currents drive advective pore water exchange, which can lead to shifts between oxic and anoxic conditions on timescales of minutes to hours [4,5].In bioturbated sediments, similar shifts between oxic and anoxic conditions have been observed in relation to faunal pumping [6].These short-term redox variations select for metabolically f lexible microbes, adapted to varying electron donor and acceptor availability.Indeed, the dominant bacteria in permeable sediments are facultative anaerobic bacteria from the families Flavobacteriaceae and Woeseiaceae, which are capable of aerobic respiration, anaerobic respiration, and fermentation [7][8][9].In turn, these communities appear to oxidize organic carbon under anoxic conditions through distinct pathways to those in cohesive sediments.Specifically, they mediate organic carbon fermentation, but do not fully reoxidize derived end-products through respiration.Evidence for this comes from lower-thanexpected accumulation of end-products of anoxic respiration such as nitrogen (N 2 ), iron (Fe 2+ ), and sulfide (H 2 S) compared to DIC production, accompanied by the accumulation of hydrogen observed in f low-through reactors in anoxic permeable sediments Figure 1.Different carbon degradation pathways from the six-carbon glucose molecule; glucose can be fermented into: acetate and CO 2 by the EMP pathway; CO 2 , acetate, and lactate by the PP pathway; or CO 2 and acetate by the ED pathway; the oxidation of VFAs produced during fermentation (acetate, lactate, etc.) is typically coupled to the reduction of anoxic terminal electron acceptors; respiration can also occur directly using the glucose with no prior fermentation; R n values of each pathway, derived from 13 CO 2 ratios of different glucose isotopologues (see Calculations), are represented underneath; if fermentation remains uncoupled from respiration, VFAs are not consumed and R n values remain that of whichever fermentation occurs.[9,10].However, to date all inferences of fermentation have been indirect, and it remains unresolved through what pathways organic carbon is oxidized in anoxic permeable or bioturbated sediments.
Here we used a differentially labeled 13 C glucose assay to compare oxidation pathways in sandy and muddy sediments.Nine sandy and three muddy sediments from Australia and Denmark were incubated with position-specific 13 C-labeled glucose isotopologues.Labeled either on the first carbon (1-13 C), second carbon (2-13 C), third carbon (3-13 C), or all six carbons ( 13 C 6 ) (Fig. 1), the ratio of 13 CO 2 produced in each treatment (R n ) was then used to determine the dominant carbon oxidation metabolism.Although this broad approach has long been used by biochemists for determining fermentation pathways of cultivated microbes [11,12], this is the first study to apply this approach quantitatively and to a range of sediments with complex microbial communities for direct comparison.We hypothesized that microbial respiration would dominate over fermentation of glucose in cohesive sediments, compared to permeable sediments, except if heavily bioturbated.Moreover, we expected permeable sediments to show increased coupling of fermentation to respiration after prolonged anoxic incubations of days to weeks in line with previously observed shifts in microbial communities under extended anoxia [9].To support these inferences, we also extended these assays to four bacterial isolates from permeable sediments, including members of the dominant family Flavobacteriaceae.

Study sites
Between March 2019 and March 2021, nine sandy sediments and three muddy sediments were sampled at sites across Australia and Denmark (Fig. S1).Sites spanned temperate to tropical locations and included both silicate and carbonate sediments.The biogeochemistry of each site has been previously studied, and references to these studies are provided in Table S1.Sandy sites were selected to encompass a range of hydrodynamic regimes and included three high-energy surface sands and three low-energy surface sands.Sites were grouped into high or low energy based on the observation of colored depth profiles of sediment cores and wave height and frequency.Sites with greater wave action displayed yellow oxidized sediments at greater depths (∼15 cm deep) with anoxic grey layers below this.At lower energy sites, the anoxic sediment (5-10 cm deep) was much darker.Higher-energy locations included Werribee Beach and Melbourne Beach, both enclosed within a large bay, yet still frequently exposed to medium to large waves.Heron Reef was also deemed high energy, located on a coral cay in the Great Barrier Reef.Sites deemed lower energy included Melbourne Harbour, located further up the beach from Melbourne Beach and partially shielded from oncoming waves by a harbor break wall.Low-energy sediments were also collected in Denmark from Faellesstrand, a shallow marine lagoon on the northeast coast of Fyn, and Hjerting Beach, which is shielded from open ocean by a barrier spit complex known as Skallingen.In addition, deeper lower energy sediments were collected at Melbourne Beach, Melbourne Harbour, and Faellesstrand.Both surface (0-15 cm) and deep (15-30 cm) sediments were collected from the subtidal zone.Surficial muddy sediments were sampled from estuaries around Victoria (Yarra, Gippsland, Patterson).Patterson sediment was heavily infaunated, which is typical for this site [13], while no significant fauna was observed at other sites.

Slurry incubations
Sandy sediments were sieved (2-mm mesh) to remove large debris and fauna, and seawater was filtered (1.6-11 μm) to remove most pelagic species potentially present in overlying water.Muddy sediments remained unsieved.Slurries were prepared with 20 g wet sediment and 35 ml filtered seawater in 60 ml incubation vials before sealing with a butyl rubber stopper (Sigma-Aldrich) and Wheaton closed-top seals (Sigma-Aldrich).Anoxic treatments were purged with nitrogen gas to exclude O 2 , while the headspace of the oxic treatment remained as unamended air, corresponding to a dissolved oxygen concentration of ∼230 μM.Melbourne Beach and Melbourne Harbour deep sediments (15-30 cm) were prepared in an N 2 -filled anoxic chamber to avoid exposure to oxygen.Faellesstrand deep sediments were not prepared in an anoxic chamber and hence were momentarily exposed to oxygen during set up.After a designated preincubation time (explained below), 13 C-labeled glucose isotopologues (1-13 C, 2-13 C, 3-13 C, or 13 C 6 , Cambridge Isotope Laboratories Inc., 98%-99%) were added (50 μM) in parallel with three replicate slurries each (Fig. S2) and placed on the orbital shaker (150 rpm, dark, 20 • C) for 4 h.Subsequently, these slurries were left to settle for 5 min, then opened and subsampled.
To determine organic carbon oxidation pathways at four distinct time points, we employed preincubation periods of 0, 1, and 7 days before addition of 13 C-labeled glucose followed by a 4-h incubation (Fig. S2).These preincubation periods were chosen because previous studies have shown that sulfide production in permeable sediment commences after several days of anoxia and that there is a shift in the microbial community to sulfate reducers over this period [9].We therefore expected to see a distinct temporal sequence from fermentation to increased respiration over this timeframe.For the 0-day preincubation, glucose was added immediately after the preparation of the slurries.For 1and 7-day preincubations, slurries were put on an orbital shaker for gentle agitation (150 rpm) in a light-proof box at 20 • C for the designated period before glucose addition.After the preincubation period, the 13 C-labeled glucose isotopologues were added, while three initial slurries (t 0 ) were opened and subsampled simultaneously, as described below, to determine the background concentrations of DIC and 13 CO 2 (Fig. S2).Slurries with added 13 C-labeled glucose were then replaced on the orbital shaker.
Note that experiments were not set up to mirror the constantly f luctuating redox conditions; instead prolonged oxic and anoxic incubations enabled observation of otherwise transient processes.

Flow through reactor experiment
Surface sand (0-10 cm) from Melbourne Beach was sieved (1 mm mesh), homogenised, and packed into 6 cylindrical reactors (4.2 cm length, 4.8 cm inner diameter) as previously described [9,10,14].Filtered seawater (0.7 μm, GF/F) amended with 1-13 Cor 13 C 6 -labeled glucose (50 μM) was pumped through f low through reactors (FTRs) using a peristaltic pump at a f low rate of ∼45 ml h −1 with residence time of the pore water ∼1 h.Both labeled seawater reservoirs were contained in 10-L high-density polyethylene carboys, which were replaced every ∼48 h.Due to the large volume of the reservoirs, 1-13 C and 13 C 6 labeled glucose were most cost-effective so were selected over other 13 C glucose isotopologues.O 2 levels were monitored from the FTR outlet using a f low-through O 2 -sensitive probe (PyroScience FireSting).Glass syringes collected seawater from the outlets and reservoirs for DIC, 13 CO 2 , and VFA analysis every few hours.Both seawater reservoirs were bubbled with ambient air at the first sampling point (0 h), before transitioning to anoxia by purging with 800 ppm CO 2 in pure N 2 using a digital gas mixer (Vögtlin).

Isotopic and biogeochemical measurements
Samples for measurement of DIC concentration (3 ml) and 13 CO 2 concentration (12 ml) were collected in gastight glass vials (Labco Exetainer), preserved with 10 μl 6% HgCl 2 and stored with no headspace.Prior to 13 CO 2 analysis, 4 ml of sample in the 12-ml vial was replaced with helium.Phosphoric acid (12.5 mM) was added to the sample to convert DIC to CO 2 before being analyzed on a Hydra 20-22 Continuous Flow Isotope Ratio Mass Spectrometer (CF-IRMS; Sercon Ltd., UK).Total DIC concentration was measured using a DIC analyser (Apollo SciTech).Nitrate and sulfate concentrations were not measured; however, we know from previous measurements background that nitrate is likely to be <10 μM [10], and sulfate will be in the order of seawater concentrations at 28 mM.Sediment grains were sized using a Beckham Coulter LP13 320 particle sizer after soaking in sodium hexametaphosphate for 24 h and sonicating for 10 min.FTR samples for VFA analysis (2 ml) were collected in 4 ml ashed borosilicate glass vials with Tef lon lined lids (Sigma Aldrich) and frozen.Upon analysis, samples were thawed and derivatized as previously described (Albert and Martens, 1997), before injection into reverse phase High Performance Liquid Chromatography (HPLC) combined with preconcentrator and guard column (Agilent SB-C8 4.6 × 12.5 mm) and analytical column (Agilent SB-C8 4.6 × 250 mm).

Isolation, cultivation, and sequencing of bacteria
We designed a strategy to culture facultative anaerobes representative of the permeable marine sediment communities of Port Philip Bay, Victoria.Sediment and seawater from Melbourne Beach were collected in November 2020 and combined in 180 ml vials to form a slurry.The slurry was supplemented with 1 mM glucose and sealed with a butyl rubber stopper before being made anoxic by purging with N 2 .The vial was then placed on an orbital shaker (150 rpm, dark) at room temperature (20 • C).After a 14-day incubation, plates of Marine Agar 2216 medium (Difco) supplemented with 1 mM glucose were prepared in Petri dishes and inoculated with a portion of the slurry and incubated aerobically at 30 • C for 3 days.Following this incubation, individual bacterial colonies were repeatedly transferred to fresh Marine Agar 2216 + 1-mM glucose plates for purification and identification.16S rRNA genes from each bacteria were amplified using colony PCR and visualized via gel electrophoresis [15], before extraction (Isolate II PCR & Gel Kit, Bioline) and whole-genome sequencing.Cellular DNA was sent to MHTP Medical Genomics Facility, Hudson Institute of Medical Research for whole-genome sequencing.The library preparation was performed using the Illumina Nextera XT DNA library prep kit with unique dual Indexing, which was then passed to 150 bp paired end sequencing on a NextSeq2000 platform (Illumina).Raw shotgun sequences were subjected to quality filtering using the BBDuk function of the BBTools v38.80 (https://sourceforge. net/projects/bbmap/), which sequentially removed contaminating adapters (k-mer size of 23 and hamming distance of 1), PhiX sequences (k-mer size of 31 and hamming distance of 1), bases from 3 ends with a Phred score below 20, and resultant reads with lengths shorter than 50 bp.Quality-filtered reads were assembled using Unicycler v.0.4.7 (-mode normal, -keep 0) to obtain draft genomes [16].The purity of the isolates was corroborated by the absence of foreign DNA and the presence of a sole 16S rRNA gene in the assembled genomes.The taxonomy of the newly isolated bacteria was assigned by GTDB-Tk v1.4.0 [17] with reference to the Genome Taxonomy Database (GTDB) R06-RS202 [18].Annotation of the genomic features and metabolic capabilities was performed using DRAM v1.2.4 [19], with dbCAN2 database, MEROPS peptidase database, and KEGG protein database (accessed 22 November 2021).We additionally searched the genomes against our custom databases (doi: 10.26180/c.5230745)for the presence of key metabolic marker genes involved in using various electron acceptors and donors using DIAMOND v.2.0.11[20], with cut-offs reported previously [21].

Pure culture 13 C isotopologue glucose assay and volatile fatty acid measurements
Marine Broth 2216 medium (Difco) was prepared in 180-ml vials, sealed with butyl rubber stoppers, and the headspace kept under anoxic (purged with N 2 ) and oxic (purged with air) conditions.Each vial was then inoculated with a bacterial strain to an OD 600 of 0.02 before addition of 50 μM 13 C-labeled glucose isotopologues (1-13 C, 2-13 C, 3-13 C, or 13 C 6 ) as outlined previously.Simultaneously, three initial incubations (t 0 ) that were not treated with glucose were opened and subsampled to determine background concentrations of 13 CO 2 .Vials with added 13 C-labeled glucose were then placed on an orbital shaker (150 rpm, 30 • C) for 4 h before final sampling of 13 CO 2 .Samples for measurement of 13 CO 2 concentration were collected and analyzed as described in Isotopic and biogeochemical measurements section.

Bacterial incubations for volatile fatty acid analysis
Incubations were set up to monitor growth (OD 600 ) of both bacteria under oxic and anoxic conditions as well as VFA production.About 50 ml Marine Broth 2216 medium (Difco) supplemented with 1 mM glucose was prepared in 180-ml vials before bacteria was inoculated to an OD 600 of 0.05 and vials sealed with butyl rubber stoppers.Incubations were made anoxic by purging with N 2 or constantly f lushed with ambient air to ensure the headspace remained oxic.Vials were then placed on an orbital shaker (150 rpm) at 30 • C for a week.About 1 ml sample was removed each day and the optical density at 600 nm (OD 600 ) measured (1-cm cuvette; Eppendorf BioSpectrometer basic).Based on the anoxic growth curves of each bacteria, these cultures were incubated for 2 and 4 days, which correlates to early and midstationary phase.At these time points, the culture was transferred to centrifuge tubes and centrifuged at 4500 × g.Samples of the supernatant (3 ml) were filtered (0.2 μm) and frozen in 4-ml ashed borosilicate glass vials with Tef lon lined lids (Sigma-Aldrich) for VFA analysis.Media controls without bacteria were also run.Analysis of VFA samples proceeded as described in Isotopic and biogeochemical measurements section.

Slurry incubations
Depending on the 13 C-labeled glucose isotopologue added (1-13 C, 2-13 C, 3-13 C, 13 C 6 ) and the metabolism taking place, carbon in different positions will be converted to 13 CO 2 , and a 13 CO 2 ratio (R n ) can be calculated (Fig. 1).
The excess concentration of 13 CO 2 produced in incubations amended with glucose labeled at the 1C, 2C, and 3C positions was derived using the difference in r from the beginning (t 0 ) to the end of the 4-h incubations and DIC concentration such that 13 where r is the ratio of masses 45/44 and n = 1, 2, 3 (position of labeled carbon atom).The same is calculated for the excess concentration of 13 CO 2 produced in incubations amended with glucose labeled on all six carbon atoms. 13 13 CO 2 ratios (R n ) for each labeled carbon position are then derived by normalizing against the 13 C 6 treatment (Fig. S3, Table S2), such that where n = 1, 2, 3. and For example, respiration of 3-13 C glucose results in a 13 CO 2 ratio of 0.17, as only one out of six carbons becomes 13 CO 2 (Fig. 1).
When respiration is inactive, R n will resemble that of a fermentation pathway (EMP, PP, ED).
For example, when bacteria undergo EMP fermentation using 3-13 C glucose, one of two CO 2 produced will be labeled, such that Alternatively, if bacteria undergo PP fermentation using 1-13 C glucose, the only CO 2 produced will be labeled, such that Total carbon oxidation was then distributed into fractions as respiration-driven (f resp ), EMP fermentation-driven (f EMP ferm ), ED fermentation-driven (f ED ferm ), or PP fermentation-driven (f PP ferm ).
A best fit for the contribution of the metabolisms (respiration, EMP fermentation, ED fermentation, PP fermentation) to the total rate of CO 2 production was estimated from all R n values (Fig. 2 and Table S3).This fit was performed by minimizing the total sum of squares of the error between observed and theoretical R n values using the R package Rsolnp [22] and subject to the constraints that all contributions are positive and add up to 1 [23].Carbon oxidation for each incubation is often distributed as more than one fraction and, in these instances, we defined the dominant metabolic pathway as that responsible for the largest fraction of 13 CO 2 production.It should be noted that some fermentations do not produce CO 2 [24], and therefore, this approach will not quantify these pathways.However, we observed less than or equal to the expected accumulation of VFAs (see results) based on the calculated fractions of fermentation, and it is therefore unlikely such pathways are occurring at significant rates.

Bacterial cultures
13 CO 2 ratios of pure culture glucose assays were calculated similarly to above (Fig. S4, Table S2); however, DIC was not accounted for.Instead, the excess concentration of 13 CO 2 produced in incubations amended with glucose isotopologues (1-13 C, 2-13 C, 3-13 C, 13 C 6 ) was derived using only the difference in r from the beginning (t 0 ) to the end of the 4-h incubations such that 13 where r is the ratio of masses 45/44 and n = 1, 2, 3 (position of labeled carbon atom).The same is calculated for the excess concentration of 13 CO 2 produced in incubations amended with glucose labeled on all six carbon atoms. 13CO 2 13C 6 = r (10) 13 CO 2 ratios (R n ) for each labeled carbon position and a best fit for the fractional contribution of the four metabolisms are then derived as above (Equations ( 3)-( 8)).The standard deviation of each metabolism is determined by propagating the standard deviations of the raw ratios through the calculation.

Flow through reactor experiment
The excess concentration of 13 CO 2 produced in the reactor eff luent amended with glucose labeled at the 1st and on all six carbon atoms is also calculated using Equations ( 1) and ( 2) except the difference in r is determined between the reservoir and sedimentpacked reactors.As only 1-13 C and 13 C 6 labeled glucose were used, R 1 is derived using Equations ( 3) and ( 4) (Tables S2 and S5).
Given that we only used two isotopologues for this experiment, we assumed that respiration and EMP fermentation dominated (consistent with the four isotopologue experiment), and total carbon oxidation was distributed as respiration-driven (f resp ) or fermentation-driven (f EMP ferm ) based on the prominence of R 1 values: where R 1 is determined following Equation ( 3) and 1/6 and 0 follow from the theoretical R n values (see Fig. 1).
Solving Equations ( 11) and (12) gives Equation ( 13), which predicts the proportion of carbon oxidation driven by EMP fermentation at each time point.Following Equation ( 11), the remaining carbon oxidation is then assumed to be the fraction driven by respiration (f resp ).Again, EMP fermentation or respiration is then determined as the dominant metabolic pathway when the fraction is >0.5.

Fermentation dominates anoxic glucose degradation in permeable sediments
We distinguished glucose oxidation of each sediment as respirationdriven (f resp ), EMP fermentation-driven (f EMP ferm ), ED fermentationdriven (f ED ferm ), or PP fermentation-driven (f PP ferm ) through the pathways summarized in Fig. 1 (see "Calculations").Across all sediments, glucose metabolism was typically dominated by either respiration or EMP fermentation, though ED fermentation did contribute in some incubations (Figs 2 and S3, Tables S2 and S3).
In line with the classical redox cascade, respiration is dominant in low-energy, unperturbated muddy sediments (Yarra, Gippsland) across both oxic and anoxic incubations up until 7 days.In contrast, fermentation was the dominant glucose mineralization pathway in most permeable sediments.Microbial communities in surface sands from multiple sites all undertook EMP fermentation under anoxia.EMP fermentation typically dominated (f EMP ferm > 0.5) within the first hours of anoxia (Anoxic 0) for nearly all surface sediments and persisted for up to 24 h (Anoxic 1).Respiration typically increased following 3 and 7 days of anoxic incubations.While some of the deep sediments show fractions of EMP fermentation under initial anoxia, they are all more respiration-driven than the surface sands of the same location.The observations that respiration is higher in deeper sands or following long-term incubations are consistent with more stable redox conditions favoring activity of sulfate-reducing bacteria.Surprisingly, substantial fractions of both EMP and ED fermentation were also observed in oxic incubations, particularly in highenergy surface sands including Werribee and Melbourne Beach.
Comparing the total fraction of fermentation (f EMP ferm + f ED ferm + f PP ferm ) with respiration (f resp ) shows surface and deep sands have similar average fractions under initial anoxia, though large differences are seen in muds (Fig. 3A).The total fraction of fermentation on the first day of anoxia (Anoxic 0-1) for each sediment was then compared against median grain size as a measure of hydrodynamic energy (Fig. 3B).High-energy sites with larger grain sizes such as Werribee Beach had the highest fraction of fermentation, while fine-grained muddy sediments such as Yarra Table 1.The amount of CO 2 and acetate accumulated in the reservoir and FTR in μM C equivalents as per Fig. 5; CO 2 accumulation in the reservoir was based on acetate accumulation assuming a 2:1 acetate:CO 2 C equivalent for fermentation C 6 H 12 O 6 + 2 H 2 O 2 CH 3 COOH + 2 CO 2 + 4 H 2 ; acetate accumulation in the FTR was based on the difference between acetate concentrations in the reservoir and FTR outlet; CO 2 production in the FTR was based on measured 13 CO 2 production in the FTRs.f EMP ferm is calculated as per Equation ( 9); calculated acetate production in the FTR is based on 13  Estuary had fermentation fractions close to 0. Altogether, this produced a strong positive correlation (R 2 = 0.60, P = .001)when the notable outliers of highly bioturbated muds (Patterson) and large grain size carbonate sands (Heron) are excluded (Fig. 3B).Average total fractions of fermentation of each sediment were also plotted against DIC production, but no correlation was observed (data not shown).

A shift from fermentation to respiration during long-term anoxic incubations
We investigated the effect of extended anoxic conditions on the transition between fermentation and respiration using FTR experiments where we measured fermentation, respiration, and VFA production (Fig. 4).This experiment showed the dominance of respiration during the initial oxic conditions from 0 to 18 h (Fig. 4A and B, Tables S4 and S5), then transitioned to a dominance of EMP fermentation at 46 h (after 28 h anoxia).This was followed by a return to a dominance of respiration at 100 h where it remained until the experiment finished at 160 h.We measured the concentrations of various VFAs to determine whether fermentative end products were excreted (Fig. 4C, Table S7).At the onset of anoxia, acetate and lactate production dominated (10 μM mean concentration for both), after which acetate production dominated reaching a maximum concentration of 40 μM at 160 h.Formate was consistently consumed by the FTRs.In order to account for fermentation products, we undertook a mass balance of CO 2 and acetate produced in the reservoir and FTR at 46 h (anoxic fermenting) and 160 h (anoxic respiring, Table 1).At 160 h, we were able to account for 102% of the glucose respired as acetate and CO 2 ; however, at 46 h, we were only able to account for 66% of the glucose in the form of CO 2 and acetate.To compare the observed accumulation of acetate with that expected from fermentation, we applied the fraction of fermentation from the 13 C glucose assay to the measured amount of 13 CO 2 produced and calculated an expected acetate accumulation equivalent.At 160 h, we were able to account for 110% of the expected acetate accumulation, but we were only able to account for 12% of the expected acetate at 46 h indicating an unknown sink for fermentation products.

Bacterial cultures validate occurrence of oxic and anoxic fermentation
To further compare respiration, fermentation, and VFA production dynamics in bacterial cultures under oxic and anoxic conditions, we isolated species Lutibacter sp., Vibrio sp., Tropicimonas sp., and Maribacter sp. from marine sediments and quantified the fractions of fermentation and respiration under short-term oxic and anoxic conditions (Figs 5A and S4, Tables S2 and S6).All species showed a dominance of EMP fermentation (f EMP ferm > 0.5) under anoxic conditions, though we also observed a distinct fraction of PP fermentation in Vibrio (f PP ferm = 0.39).Under oxic conditions, metabolism was surprisingly also dominated by fermentation through the EMP pathway for most species (f EMP ferm > 0.5), with the exception of Tropicimonas for which respiration comprised half of CO 2 production (f resp = 0.50) (Fig. 5A).Tropicimonas and Maribacter, both capable of respiration under anoxic conditions, possess genes for nitrous oxide reduction (Fig. S5).Additionally, Tropicimonas, which displayed the greatest fraction of anoxic respiration, has genes for nitrate and nitric oxide reduction (Fig. S5).It should be noted that the different pathways are integrated over the 4-h experiment and that it is likely that they occur sequentially as the glucose is consumed and thermodynamics change.In extended incubations to measure VFA accumulation, both Lutibacter and Maribacter reached stationary phase under oxic conditions within 1 day of incubation (Fig. 5B and C).Lutibacter was able to grow under anoxic conditions, albeit at a much slower rate and to a lower yield than under oxic conditions, reaching stationary phase after ∼3 days.Maribacter showed no growth under anoxic conditions.Under oxic conditions, both bacteria produced VFAs including acetate, formate, propionate, isobutyrate, butyrate, succinate, isovalerate, and valerate (Fig. 5D and E), with concentrations highest after 4 days under oxic conditions in the Maribacter culture.Under anoxic conditions, the highest concentrations of VFAs were observed after 4 days of anoxia, including acetate, propionate butyrate, and succinate.

Discussion
Most carbohydrates within the sediment exist in a polymeric form, which are hydrolyzed to monomeric sugars such as glucose.Our addition of 50 μM glucose was most likely higher than natural concentrations of ∼1 μM in permeable sediments [25].Of critical importance here is the concentration of glucose relative to the saturating concentration for glucose uptake (fermentation), which can range from 0.2 to 23 μM for particle-associated bacteria [26].At the lower end of this range, our glucose addition is unlikely to have increased the rate of glucose uptake, while at the higher range, this could have greatly stimulated glucose uptake (fermentation) leading to an accumulation of fermentation products before respiring bacteria were able to assimilate them [27].It is therefore possible that the "fermentation" proxy we are measuring with the isotopologues represents a temporal decoupling between fermenting and respiring bacteria.If this were occurring, then we would expect to see a substantial release of VFAs after glucose addition, which was not observed (Table 1, Fig. 4C see also discussion below).It should therefore be recognized that fermentation (or some fraction of it) may have been induced by the glucose tracer addition.
Fermentation of glucose dominated in most permeable sediments, with EMP fermentation being the predominant carbon oxidation pathway (f EMP ferm > 0.5, Figs 2 and 4B).Within permeable sediments, it is likely that redox conditions are constantly f luctuating due to periodic incorporation of oxygen and other electron acceptors into the sands by sediment resuspension and advective pore water f low [5,28].The microbial community in these high-energy sands is known to be dominated by metabolically f lexible generalist bacteria of families Flavobacteriaceae and Woeseiaceae that ferment under anoxic conditions [7][8][9]29], and the culture experiments confirmed this with all species showing a dominance of fermentation under anoxic conditions (Fig. 5A).Metabolism in these high-energy sandy sediments trended toward respiration between 1 to 7 days of anoxia (Fig. 2).In agreement, FTR incubations show the evolution of fermentation in Melbourne Beach sediments, with fermentation dominating upon initial anoxia before switching back to respiration after ∼2 days of anoxia (Fig. 4A and B).This is consistent with previous observations that after 4-10 days anoxia, there is an enrichment of sulfate reducers of families Desulfobacteraceae and Desulfobulbaceae, which leads to a recoupling between fermentative and anaerobic respiratory bacteria [7][8][9]30].
Comparing the fractions of fermentation within the first day of anoxia (Anoxic 0-1) against median grain size for each site showed a positive correlation (P = 0.001), with two notable outliers (Fig. 3B).Hydrodynamic energy at each site controls both grain size and the frequency of sediment re-oxygenation [28,31].Coarse-grained sediments will be more oxygenated and likely dominated by facultative aerobes, and finer grain sizes more anoxic and dominated by coupled anaerobic fermentative and respiratory bacteria.This relationship is therefore consistent with our hypothesis that fermentation uncoupled to respiration will dominate in more dynamic higher-energy settings.
The two noted outliers that do not conform to this relationship are a highly bioturbated mud and coarse grain carbonate sediments.In the case of the bioturbated mud, there was much more fermentation than would be expected based on the grain size alone and fermentation dominated immediately after the onset of anoxia (Fig. 2).The exact mechanism driving the fermentation is unclear, but we suggest it could be linked to the activity of benthic fauna that can oxygenate sediments on a time scale comparable to that of advective transport [32][33][34][35].Similar to sandy sediments, pore waters f luctuate between oxic and anoxic conditions, which may select for metabolically versatile bacteria that can survive in both oxic and anoxic conditions [7][8][9].Previous studies of infaunated sediments have shown that microbial communities indeed differ between bioturbated and nonbioturbated zones [36,37], with dynamic and re-worked bioturbated sediments featuring generalist fermenters such as Flavobacteriaceae [37].The occurrence of fermentation at the Patterson River suggests that our hypothesis of f lexible generalist communities adapting to dynamic conditions by fermenting extends beyond sandy sediments to other temporally oxic/anoxic sediments.In the case of the coarse-grained carbonate sediments, there was much more respiration than would be expected from grain size.Carbonate sediment grains are known to be porous and highly biologically active and hence harbor permanently anoxic zones within the grains, which can enhance anoxic processes such as denitrification [38].As such, it is highly likely that even though this sediment type is highly oxygenated around the grains, there is a community of coupled fermentative and respiring bacteria within the grains that fully respire the added glucose tracer.
As noted previously, the 50 μM glucose addition may have stimulated fermentation and hence led to a temporal decoupling of respiration and fermentation.If this is the case here, then an alternative explanation for our observations is that cohesive and less disturbed sediments have a higher capacity to assimilate fermentation products or a lower saturation concentration for glucose fermentation.Further studies with direct glucose measurements and lower glucose tracer additions are required to definitively validate this hypothesis.
The observation that significant rates of fermentation occurred in the high-energy sediments under oxic conditions (Fig. 2) was unexpected.These observations were consistent with the results from the pure culture experiments, which showed that facultatively aerobic bacteria isolated from Melbourne Beach including Lutibacter, Vibrio, and Maribacter undertook fermentation (f EMP ferm > 0.5) during oxic conditions (Fig. 5A).Vibrio and Tropicimonas undertook significant fractions of ED fermentation as well (f ED ferm = 0.15 and 0.31, respectively).The production of VFAs under oxic conditions in both pure culture experiments is also consistent with oxic fermentation.Fermentation under oxic conditions at high glucose concentrations (>10 mM range) has been well documented in the literature and is known as the "Crabtree" effect or overf low metabolism in bacteria [39][40][41].Although fermentation yields less energy per mol of glucose than oxidative phosphorylation (complete oxidation to CO 2 ), it can proceed much more rapidly and therefore yield more energy [42] as well as maximizing growth relative to proteome formation [43].Within the context of cancer cells, this metabolism is known as the Warburg effect, and it has been argued the reason for this is that it rebalances the Krebs cycle in proliferating cells to provide anabolic (biomolecules) as opposed to catabolic (CO 2 , ATP) products required for cell growth [44].Fermentation under oxic conditions may therefore convey an ecological advantage in dynamic environments such as permeable sediments supporting rapid growth when oxygen and organic matter is often transiently available [45].The fact that we were able to observe fermentation under oxic conditions in both cultures and natural sediments exposed to lower glucose concentrations (50 μM for the isotope assays) suggests that oxic fermentation may be a common phenomenon in bacteria even at low glucose concentrations and warrants further investigation.In addition, the increased predominance of the ED fermentation pathway under oxic conditions at some sites and cultures (e.g.Maribacter and Tropicimonas) is consistent with the paradigm that it can be a major catabolic pathway for glucose under oxic conditions [46].
Our results also support previous observations that there is very little acetate production relative to CO 2 in anoxic permeable sediments [10].A mass balance for CO 2 and acetate produced in the FTR at 46 and 160 h provides further insight into the fate of glucose under anoxic fermenting (46 h) and respiring (160 h) conditions.At 160 h, we were able to account for 102% of the glucose respired as acetate and CO 2 (Table 1).However, at 46 h, we were only able to account for 66% of the glucose in the form of CO 2 and acetate.Furthermore, if we apply the fraction of fermentation calculated using the 13 C glucose assay to the measured amount of 13 CO 2 produced, we can also calculate an expected acetate accumulation equivalent (Table 1).At 160 h, the calculated acetate accumulation was 74 μM C equivalents, compared to 82 μM C equivalents measured (110% of that expected).At 46 h, the calculated acetate accumulation was 145 μM C equivalents compared to a measured value of 18 μM C equivalents (12% of that expected).This suggests that the fermentation uncoupled to respiration observed here is not as a result of a transient accumulation of VFAs stimulated by glucose addition [27,47].It is possible that the missing acetate was assimilated by polyhydroxyalkanoate-accumulating organisms, which have been observed in permeable sediments [48] or stored as lipids [10,30].Consistent with this, the culture experiments showed that Lutibacter and Maribacter produced negligible VFAs after 2 days anoxia.Furthermore, Lutibacter was able to increase its biomass under anoxic conditions, which suggests Lutibacter may be able to assimilate some of the fermentation products into storage molecules, which have recently been suggested to comprise an under-appreciated form of biomass [49].The exact nature of this process remains to be determined and requires further investigation.

Conclusion
Using isotopologues of 13 C labeled glucose, we show through slurry and FTR incubations that EMP fermentation occurs in transiently anoxic sandy sediments, remaining uncoupled from respiration processes for up to 160 h.Factors controlling the frequency of sediment re-oxygenation such as hydrodynamic energy and bioturbation exert a strong control on the extent to which glucose fermentation decouples from respiration at the onset of anoxia.Oxic fermentation was observed in both sands and bacterial cultures and suggests that this process may be significant in the environment.Our understanding of the extent to which this occurs with naturally present carbohydrates and the fate of the organic fermentation products remains incomplete and the question remains as to what the final products of fermentation are.

Figure 2 .
Figure 2.Fractions of respiration, ED, EM, and PP fermentation occurring at each site during each incubation, as determined by R package Rsolnp using13 CO 2 ratios (see Calculations); incubations include oxic (0), anoxic (0), anoxic (1), and anoxic(7); numbers in parentheses represent number of days of anoxia pretreatment before the glucose assay (4-h duration) was undertaken; sands and muds are arranged from high-energy sites to low-energy sites and are also grouped into surface and deep.

Figure 4 .
Figure 4. FTR experiments were run under oxic conditions for 18 h then transitioned to anoxia (dotted line, A) and remained anoxic for the rest of the experiment (grey shading); fractions of EMP fermentation (f EMP ferm ) and respiration (f resp ) were estimated from R 1 values (B, see calculations); respiration is dark grey and EMP fermentation is light grey; net concentration changes (reservoir values subtracted from FTR outlets) of acetate (•), formate ( ), propionate ( ) and lactate ( , C); acetate and formate concentration apply to the left y-axis while propionate and lactate values are on the right y-axis; error bars show the standard deviation.

Figure 5 .
Figure 5.A fractions of respiration, ED, EM, and PP fermentation occurring in oxic and anoxic incubations of bacterial species Lutibacter sp., Vibrio sp., Tropicimonas sp., and Maribacter sp. as determined by R package Rsolnp using 13 CO 2 ratios (see Calculations); growth curves of B Lutibacter and C Maribacter over 1-week incubation under both oxic and anoxic conditions; VFA concentrations in the cultures of D Lutibacter and E Maribacter after 2 and 4 days under oxic and anoxic conditions; error bars show the standard deviation.