GLOBAL PATTERNS OF NET RESPIRATION IN SUBSEAFLOOR SEDIMENT

Microbial life is abundant in subseafloor sediment. Rates of respiration in subseafloor sediment appear to be very slow, but are generally poorly constrained. This thesis aims to address this problem by quantifying respiration rates in subseafloor sediment at sites across the globe and determining geographic patterns of depthintegrated respiration rates and average per-cell respiration rates. First, we compiled pore water profiles and descriptive data from scientific coring and drilling sediment cores deposited in publicly accessible databases. Using these profiles, we quantified the global distribution of net respiration rates in subseafloor sediment (>1.5 meters below seafloor) using (i) a reaction-transport algorithm, (ii) sediment physical properties, and dissolved oxygen, sulfate, and inorganic carbon concentration data from interstitial waters of sediment cores, and (iii) correlations of net respiration rates to sedimentation rate and sea-surface chlorophyll-a concentration. Our results indicate depth-integrated net global respiration in subseafloor sediment to be 2.6 x 10 ± 0.25 x 10 mol e/yr. Comparison to the global rate of organic-matter burial in marine sediment indicates that most organic matter in subseafloor sediment (>1.5 mbsf) goes unconsumed. Respiration rates per unit area vary between abyssal regions, where aerobic respiration dominates and rates are lowest, and continental margin regions, where anaerobic respiration dominates and rates are generally higher. Regional areal net respiration can vary by as much as six orders of magnitude. Mean per-cell respiration rates are much slower in subseafloor sediment than in the ocean, although both environments have equivalent numbers of prokaryotic cells. Per-cell respiration rates in subseafloor sediment are generally higher in regions dominated by aerobic respiration than in regions of anaerobic respiration. When combined with a recent estimate of radiolytic chemical production in subseafloor sediment, the potential total respiration rate in the sediment increases to 1.3 x 10 mol e/yr, but is 10,000 times lower than the electron-equivalent rate of carbon fixation in the ocean. This work contributes to the growing collection of studies showing cells in subseafloor sediment live at rates much slower than those seen in the ocean and surface world.

When a recent estimate of radiolytic chemical (electron donor and electron acceptor) production in subseafloor sediment  is considered, total respiration in subseafloor sediment may be as high as 1.3*10 13 mol e -/yr -a factor of five higher than the global net (organic-fueled) respiration rate. Even with this potential increase, the global respiration rate of subseafloor sedimentary life is 10,000 times slower than the global rate of photosynthetic carbon fixation in the ocean (Behrenfeld and Falkowski, 1997).

INTRODUCTION
Bacteria and archaea are abundant (ca. 2.9*10 29 cells globally) (1) and active (2,3) in marine sediment, but per-cell respiration rates in subseafloor sediment are extremely low (4)(5)(6). Based on estimates of net respiration, microbes in subseafloor sediment catabolize 10 4 to 10 6 times more slowly than microbes in pure culture (4), with biomass turnover times of hundreds to thousands of years (3). The slow growth rates are limited by low bioavailable energy fluxes in the subseafloor (4, 6-10).
The past five decades of scientific ocean drilling and coring expeditions have generated chemical, biological, physical, and geological data for subseafloor sediment at hundreds of sites across the globe. Analyses of these data have greatly increased understanding of subseafloor life (e.g., ref. 4-8, 11, 12) and contributed to major advances in climatic and oceanographic sciences (13,14). In oxic marine sediment, oxygen (O2) is the predominant electron acceptor used to oxidize organic matter (4, 7), while in anoxic marine sediment, sulfate (SO4 2-) is the predominant external electron acceptor (4,7,(15)(16)(17). Therefore, depth-integrated rates of SO4 2and O2 reduction are direct measures of anaerobic and aerobic net respiration in subseafloor sediment (4,11,17). Dissolved inorganic carbon (DIC = CO2 + HCO3 -+ CO3 2-) is the primary product of net respiration in subseafloor sediment (4). At sites where SO4 2is completely reduced within the sediment, DIC is utilized as an electron acceptor at greater depths, where other net electron acceptors are rare (8).
Several previous studies have used chemical concentration data and sediment physical properties to quantify SO4 2reduction rates or O2 reduction rates at individual sites (e.g., ref. 5-7, 11, 18, 19). Such studies have yielded key insights into subseafloor life; for example, they have provided the primary evidence that per-cell respiration rates and per-cell energy fluxes are extraordinarily slow in subseafloor sediment (e.g., ref. 8,12). Other studies have used chemical concentration data and sediment properties from many sites (alone or in combination with radiotracer measurements of potential activity) to calculate global rates of SO4 2reduction in marine sediment (e.g., ref. 17,20). The latter studies placed important bounds on the global sulfur cycle and its relationship to the marine carbon cycle.
Another source of energy in subseafloor sediment are hydrogen (H2) and oxidants produced through radiolysis, or the splitting of water (21,22). A recent study showed that radiolysis is a principle source of electron donors and acceptors for microbial communities in subseafloor sediment from the surface down to the depth of basement or 0.1% sediment porosity (23). It is not yet clear how much subseafloor life could be living off the products of radiolysis.
No previous studies have explicitly focused on the total organic-fueled respiration of subseafloor sedimentary life or on broad patterns of subseafloor organicfueled respiration throughout the entire ocean. Furthermore, no previous studies have quantified the total rate or geographic distribution of subseafloor aerobic respiration, or its contribution to total organic-fueled respiration in subseafloor sediment.
To advance understanding of subseafloor life, this study has the following objectives. The first objective is to quantify how net (organic-fueled) respiration in subseafloor sediment varies from site to site and region to region. The second objective is to quantify the global rate of net respiration by the subseafloor sedimentary ecosystem, and place it in the context of other global metabolic rates. The third objective is to quantitatively assess the potential relative dependence of subseafloor sedimentary life on (i) organic oxidation (net respiration) and (ii) radiolytic redox chemicals produced within the same sediment.
Net respiration rates are lowest (minimum rate 6.0*10 -7 mol e -/m 2 /yr) in abyssal regions and highest (maximum rate 5.2*10 -1 mol e -/m 2 /yr) in continental margin regions and in other areas with high sedimentation rates. Sulfate reduction dominates along continental shelves and at the equator, where net respiration rates are high (2.5*10 -3 to 5.2*10 -1 mol e -/m 2 /yr ). Oxygen reduction dominates in the sediment beneath ocean gyres and near mid-ocean ridges, where net respiration rates are low (6.0*10 -7 to 2.5*10 -3 mol e -/m 2 /yr) ( Figure 1). Summation of the rates from both regions yields global net respiration in subseafloor (>1.5 mbsf) sediment of 2.6*10 12 ± 0.25*10 12 mol e -/yr. Although the regions of subseafloor sediment dominated by O2 reduction cover about 35% of the total seafloor, aerobic respiration accounts for only around 6% of the global net respiration in subseafloor sediment.
Comparison of SO4 2reaction rates to DIC production rates indicates that net SO4 2reduction approximates net organic-matter oxidation in anoxic subseafloor sediment. Dissolved SO4 2data and DIC data allow direct comparison of SO4 2reduction rates and DIC production rates in the subseafloor sediment of nine sites. For those sites, the mean relationship of SO4 2reduction rate to DIC production rate is roughly 1-to-1 (m = 1.3) (Supplemental Figure 3). This result reinforces the longstanding recognition that SO4 2is the dominant electron acceptor for degrading organic matter in anoxic marine sediment (4,7,(15)(16)(17)19). Comparison of O2 reduction rates to NO3production rates has similarly shown that net O2 reduction approximates net organic-matter oxidation in oxic subseafloor sediment.
To quantify global patterns of mean respiration per-cell in subseafloor sediment, we calculated mean per-cell respiration rates from (i) our depth-integrated  (Table 1) and sediment accumulation rate tends to co-vary with marine primary production.
Buried organic matter serves as the main electron donor for net respiration in subseafloor sediment. Where sedimentation rates are high, burial rates of organic matter and dissolved electron acceptors are rapid, sustaining relatively high net respiration in subseafloor sediment. In contrast, where sedimentation rates are low, most organic flux to the seafloor is consumed in very shallow sediment, leaving much less organic matter to support net respiration by the subseafloor ecosystem (6).
Sediment accumulation rates tend to be highest in regions where marine productivity is high (e.g., along continental margins, where upwelling of deep water sustains high photosynthetic biomass production and there is large terrigenous input of both sediment and organic matter) and lowest in regions where marine productivity is low (e.g., the abyssal sediment of mid-ocean gyres, where there is very little upwelling to sustain photosynthetic biomass production, very little planktonic biomineralization, and almost no input of terrigenous sediment). Sedimentation rate is even lower in sediment deposited below the carbonate compensation depth (where calcium carbonate particles secreted by marine plankton in the photic zone are dissolved in the deep water and near-seafloor sediment).
Although it explicitly includes both SO4 2reduction and O2 reduction, our global estimate of net respiration in subseafloor sediment (2.6*10 12 ± 0.25*10 12 mol e -/yr) is much lower than published estimates of total sulfate reduction in marine sediment [9.04*10 13 mol e -/yr (17) to 7.5*10 14 mol e -/yr (20)]. This difference between net subseafloor respiration and global SO4 2reduction may primarily result from the considerable rate of global SO4 2reduction in sediment shallower than 1.5 mbsf.
Additionally, the highest SO4 2reduction estimate (7.5*10 14 mol e -/yr) includes data from near-shore sites (which are under-represented in our data, and where most SO4 2reduction occurs at shallow sediment depths) and uses rates from radiosulfur experiments (potential rates) as a proxy for in situ rates.  Figure 2B).
Although they approximate gross organic-fueled respiration in subseafloor sediment, net rates of SO4 2and O2 reduction may not capture all significant electronaccepting activity in the subseafloor sedimentary ecosystem. Water radiolysis produces electron donors and acceptors that may also play a key role in sustaining microbial life in subseafloor sediment (18,21). Radiolytic H2 and radiolytic oxidants are produced in stoichiometric balance (18), and the mass balances described in the Results (net SO4 2reduction vs net DIC production, and net O2 reduction vs net NO3production) indicate that consumption of radiolytic products does not significantly affect net respiration in subseafloor sediment.
Electron equivalents of natural radiolytic production of H2 and oxidants in subseafloor sediment are each estimated to approximate 9.8*10 12 mol e -/yr (23). This is almost a factor of four larger than the global organic-fueled respiration rate we Addition of potential radiolysis-supported respiration to the net (organicfueled) respiration that we calculate has a relatively small effect on our estimates of mean per-cell respiration (compare Figure 2B to Figure 2D). sites. Sites located on seamounts breaching the 2000m water depth cutoff were removed from the regression analysis. We used SPSS Statistics software (v24.0) to run a multiparametric backstepping regression analysis to determine, via the R 2 value, that a linear relationship between sedimentation rate, sea surface chlorophyll-a, and respiration rates create a regression line that best fits the actual electron transport rate data for both margin regions (R 2 = 0.84) and abyssal regions (R 2 = 0.76) ( Table 1).
Grids of sedimentation rate and sea surface chlorophyll-a were manipulated according to the regression equations (Table 1) to develop a global map of respiration rate in subseafloor sediment ( Figure 1A). The residuals were plotted beneath a normal distribution curve calculated from the appropriate standard deviation to ensure a normal distribution of residuals ( Figure 1B).
The map of respiration through organic matter oxidation was directly     The relative difference between globally predicted radiolytic hydrogen production  and net respiration in subseafloor sediment. Net respiration exceeds hydrogen production in blue regions. Red and yellow regions denote where radiolytic hydrogen production most and least outweighs net respiration, respectively.

Porewater chemical data
Dissolved chemical measurements (SO4 2-, O2, DIC) from ocean drilling cruises and coring cruises were downloaded and compiled from the Scientific Earth Drilling Information Service (SEDIS), the Biological and Chemical Data Management Office (BCO-DMO), and the Janus Database. We also collected relevant metadata from each drillsite or coring site (e.g., latitude, longitude, water depth, depth cored, available age models).

QA/QC of dissolved chemical measurements
We plotted and examined the dissolved O2 and SO4 2measurement profiles from each site to ensure the quality of the reaction-rate results used to create our global model. Through this exercise, we culled any sites that contain profiles characterized by one or more of the following criteria: (i) fewer than 5 total data points; (ii) missing data from the top 10 meters of sediment; (iii) anomalously high variability in downcore measurements; (iv) evidence of upward advection; or (v) location on a seamount or near evidence of turbiditic activity. To determine the latter, we used cruise report site descriptions and GeoMapApp (v3.6.6, 2017) to analyze the seafloor morphology near each study site.

Global grids of environmental parameters
Each site has unique environmental characteristics, including basement age, sea-surface chlorophyll-a, sedimentation rate, distance from land, and sea-surface compiled archived data to create ten-year average grids.

Creation of the sedimentation rate grid
Using variable tension 2D spline with the mask applied to the initial sedimentation rate grid.

Reaction rate analysis
We used the MatLab program (version 8.3.0) and numerical procedures from We accounted for the fluxes from the omitted 1.5m section by extrapolating the values at 1.5mbsf to the sediment water interface.
To directly compare aerobic respiration rates to anaerobic rates, we converted all reaction rates (e.g., mol O2/m 2 /yr) to electron equivalents (mol e -/m 2 /yr) by assuming the transfer of 4 electrons per O2 molecule reduced, and 8 electrons per

Multivariate regression analysis
After converting all respiration rates to electron-equivalent rates, we used the SPSS Statistics software (v24.0) for a backstepping multiparametric regression analysis. This analysis determined which independent variables (e.g., sedimentation rate, sea surface chlorophyll-a, distance from land, etc.) best capture the variability in respiration rate. The backstepping regression determined the linear best fit with sedimentation rate and sea surface chlorophyll-a, with sedimentation rate being the predominant driver of variance in the model. Distance from land also describes some of the variance, however the addition of this variable to the regression model does not increase the total amount of variance in respiration rate the model can explain.
Additionally, using a logarithmic relationship between respiration rate and chlorophyll-a does not significantly impact the R 2 value or the model values.
We also identified a difference in the data trends for sites in shallow water relative to sites in deep water, especially when it came to sedimentation rate.
Therefore, we experimentally split the sites into two regions, using various depth cutoffs between 1000 and 4000m water depth. We discovered that the regression analysis worked best when the data was split into "margin" sites, with <2000m water depth, and abyssal sites, with water depth >2000m. Sites on seamounts in water shallower than 2000m were removed from the regression analysis. Of the examined variables, linear fits to mean sedimentation rate and sea surface chlorophyll-a exhibit the highest correlation to respiration rates in both margin regions (R 2 = 0.84) and abyssal regions (R 2 = 0.76), with mean sedimentation rate being the predominant driver of variation in this model. When non-linear fit possibilities were examined, no non-linear fit options exhibited a higher R 2 than the linear fits.

Creation of the Global Model
The output of the above statistical analysis resulted in a regression surface for each region (Equations S1 and S2), where SR is sedimentation rate and CHL is sea surface chlorophyll-a.
The margin and abyssal model grids were combined to create a final grid encompassing the entire subseafloor at 0.66 degree resolution.
We then plotted the global model grid and the location of each study site using GMT. We color-coded each site based on its calculated net respiration rate, using the same color scale as the global model. This allowed us to compare how well the model estimates match the site-specific rates.
We calculated the residuals of the model (measured value -modeled value) and standard deviation of 0.05. We then plotted as a histogram of the residuals and the corresponding normal distribution curve for the appropriate standard deviation. The residuals are normally distributed, and all sites fall within the range of sedimentation rate/sea surface chlorophyll-a combinations that naturally occur in the world ocean ( Figure 1). Finally, using grdvolume we calculated a net global respiration rate.

Model modifications for comparison to global cell abundance and radiolytic chemical production
For our comparisons to global maps of cell abundance  and radiolytic production of electron donors and acceptors (Sauvage, 2018), we extended our net respiration model to the depth of (i) basement, (ii) 0.1% porosity, or (iii) the 122°C isotherm, as appropriate to match the environmental limits of the cellabundance and radiolytic production models. Therefore, for each grid cell, we determined the depths of these three parameters using grdtrack in GMT. We then used

Calculation of Per-cell Rates
We modified the Kallmeyer et al. (2012) global grid of cell abundance by imposing a cutoff at 1.5 mbsf to only includes cells in sediment below this depth, just as our global respiration model does. We also applied a cutoff to the model as necessary, to the depth of basement, 0.1% porosity, or the 122°C isotherm, whichever was shallowest. Using GMT, we converted the grid from cells/km 2 to cells/m 2 . We then divided our adjusted respiration grid (described above) and the calculated total respiration grid by the cell abundance grid to develop maps of global respiration percell (mol e -/cell/yr)( Figure 2B and 2D). It should be noted that this map is a rough approximation of global respiration per-cell, as both maps are based on models derived from different variables (e.g., sedimentation rate and distance from land  vs. sedimentation rate and sea surface chlorophyll-a).

Comparison to DIC production rates
We quantified net DIC production rates using the same methods as described above for net SO4 2and O2 reduction rates. These rates were converted to electron equivalents by assuming 4 electrons per mol of DIC produced. When compared to SO4 2reduction rates, the slope was 1.3 (Supplemental Figure 3). The approximate equivalence of the two rates indicates that net SO4 2reduction roughly equals net organic oxidation within the sediment. The slight deviation from a strict 1-to-1 relationship may in part result from (i) contribution to DIC production by dissolution of carbonate minerals, (ii) organic oxidation via metal (iron and/or manganese) reduction (e.g., Wang et al., 2010), (iii) DIC production through fermentation in the methanogenic zone, (iv) DIC production occurring below the depth of sulfate reduction, and/or (v) relative high errors and a very small number of sites with high respiration rates (4). We are unable to account for the above variables (i-v), and therefore we expect the slope to be greater than one.

Comparison to a previous estimate of global distributions of SO4 2and O2 reduction in subseafloor sediment
In a previous study (D 'Hondt et al., 2019), we used published SO4 2and O2 concentration data to estimate where SO4 2reduction and O2 reduction respectively dominate net respiration in subseafloor sediment. We directly compared our model output to this published distribution by using GMT to convert the two grids to a binary system denoting grid nodes of oxygen reduction or sulfate reduction. The two grids were then directly compared and color-coded depending on if the areas agreed (e.g., both predicted oxygen reduction at on grid node) or disagreed (e.g., one estimated oxygen reduction and the other sulfate reduction at the same grid node). We found 81% agreement in the predictions between the two models.

Comparison to global patterns of radiolytic H2 production
The global grid of radiolytic hydrogen production from Sauvage (2018) was cut off at the depth of basement, 0.1% porosity, or the 122°C isotherm, whichever was shallowest. This grid was converted to mol e -/m 2 /yr. The resulting radiolysis grid and our adjusted grid of respiration through organic matter oxidation (described above) were added using grdmath in GMT to produce a grid of total respiration ( Figure 2C). Within the black contour line is where aerobic respiration is predicted to occur. Anaerobic respiration is predicted to occur at all locations outside of the contour (A) Global distribution of dissolved oxygen and sulfate in subseafloor sediment. Oxygen reduction is expected to dominate in blue regions, while sulfate reduction is expected to dominate in white and yellow regions (B). A direct comparison between areas of predicted aerobic and anaerobic respiration in subseafloor sediment, with both models predicting aerobic respiration in orange regions and anaerobic respiration in blue regions. The two models have conflicting predictions in white regions (C).