Labilization and diversiﬁcation of pyrogenic dissolved organic matter by microbes

. CE1 CE2 With the increased occurrence of wild-ﬁres around the world, interest in the chemistry of pyrogenic organic matter (pyOM) and its fate in the environment has increased. Upon leaching from soils by rain events, signiﬁcant amounts of dissolved pyOM (pyDOM) enter the 5 aquatic environment and interact with microbial communities that are essential for cycling organic matter within the different biogeochemical cycles. To evaluate the biodegrad-ability of pyDOM, aqueous extracts of laboratory-produced biochars were incubated with soil microbes, and the molecu-10 lar changes to the composition of pyDOM were probed using ultrahigh-resolution mass spectrometry (Fourier transform– ion cyclotron resonance–mass spectrometry). Given that

Abstract. CE1 CE2 With the increased occurrence of wildfires around the world, interest in the chemistry of pyrogenic organic matter (pyOM) and its fate in the environment has increased. Upon leaching from soils by rain events, significant amounts of dissolved pyOM (pyDOM) enter the 5 aquatic environment and interact with microbial communities that are essential for cycling organic matter within the different biogeochemical cycles. To evaluate the biodegradability of pyDOM, aqueous extracts of laboratory-produced biochars were incubated with soil microbes, and the molecu-10 lar changes to the composition of pyDOM were probed using ultrahigh-resolution mass spectrometry (Fourier transformion cyclotron resonance-mass spectrometry). Given that solar irradiation significantly affects the composition of py-DOM during terrestrial-to-marine export, the effects of pho- 15 tochemistry were also evaluated in the context of pyDOM biodegradability.
Ultrahigh-resolution mass spectrometry revealed that many different (both aromatic and aliphatic) compounds were biodegraded. New labile compounds, 22 %-40 % of 20 which were peptide-like, were bio-produced CE3 . These results indicated that a portion of pyDOM has been labilized into microbial biomass during the incubations. Fluorescence excitation-emission matrix spectra revealed that some fraction of these new bio-produced molecules is associated with 25 proteinaceous fluorophores. Two-dimensional 1 H-1 HCE4 total correlation nuclear magnetic resonance (NMR) spectroscopy identified a peptidoglycan-like backbone within the microbially produced compounds. These results are consistent with previous observations of peptidoglycans within the 30 soil and ocean nitrogen cycles where remnants of biodegraded pyDOM are expected to be observed.
Interestingly, the exact nature of the bio-produced organic matter was found to vary drastically among samples indicating that the microbial consortium used may produce dif-35 ferent exudates based on the composition of the initial py-DOM. Another potential explanation for the vast diversity of molecules is that microbes only consume low molecularweight compounds, but they also produce reactive oxygen species (ROS), which initiate oxidative and recombi-40 nation reactions that produce new moleculesCE5 . Some of the bio-produced molecules (212-308 molecular formulas) were identified in surface and abyssal oceanic samplesCE6 , and 81-192 of these formulas were of molecular composition attributed to carboxyl-rich alicyclic molecules (CRAM). 45 These results indicate that some of the pyDOM biodegradation products have an oceanic fate and can be sequestered into the deep ocean. The observed microbially mediated diversification of pyDOM suggests that pyDOM contributes to the observed large complexity of natural organic matter ob-50 served in riverine and oceanic systems. More broadly, our research shows that pyDOM can be substrate for microbial growth and be incorporated into environmental food webs within the global carbon and nitrogen cycles.TS1

Introduction
Pyrogenic organic matter (pyOM), the carbonaceous solid residue that is left after biomass burning (e.g., wildfires, biochar production), has been gaining attention in recent years as an important active component of the global biogeochemical cycles. Compositionally, pyOM is mainly comprised of condensed aromatic compounds (ConACCE7) of various degrees of condensation and functionalization (Masiello, 2004;Schneider et al., 2010;Wagner et al., 2018;. ConAC have been found in various 10 environmental matrices such as soils and sediments (Schmidt and Noack, 2000;Skjemstad et al., 2002;Reisser et al., 2016) and atmospheric aerosols (Wozniak et al., 2008;Bao et al., 2017). In the terrestrial environmental matrices, particularly in soils and sediments, ConAC were originally thought to 15 be highly stable ("recalcitrant") due to their condensed character (Goldberg, 1985;Masiello and Druffel, 1998). However, more and more studies report the presence of pyrogenic molecules in different aquatic environments (Hockaday et al., 2006;Dittmar and Paeng, 2009;Roebuck et al., 20 2017; Wagner et al., 2017;Li et al., 2019). These studies support the proposition that pyOM can be solubilized upon rain events and be leached as pyrogenic dissolved organic matter (pyDOM) resulting in large annual riverine fluxes of pyDOM from global riverine systems to the open ocean 25 Jaffé et al., 2013;Wang et al., 2016;Marques et al., 2017;Jones et al., 2020). During export, py-DOM is likely altered by various processes resulting in its degradation and alteration of its physico-chemical characteristics (Masiello, 2004;Coppola et al., 2019;Wagner et al., 30 2019). Using laboratory-prepared biochars and conservative assumptions, Bostick et al. (2018) approximated that > 85 % of the leached pyDOM is degradable (e.g., mineralizable to CO 2 ), which indicates that pyDOM is a very active component within the global carbon cycle, as previously suggested 35 (Druffel, 2004;Lehmann, 2007;Riedel et al., 2016).
In sunlit aquatic environments, photodegradation is the most significant sink for the fraction of ConAC CE8 of py-DOM . The photochemistry of ConAC and pyDOM has been studied utilizing either laboratory- 40 prepared pyDOM (Ward et al., 2014;Fu et al., 2016;Li et al., 2019;Bostick et al., 2020; or natural organic matter rich in ConACCE9 (Stubbins et al., 2010Wagner and Jaffé, 2015). These studies have reported that ConAC are exceptionally photola- 45 bile and they degrade through a series of oxygenation, ring opening, and decarboxylation reactions leading to a pool of smaller aliphatic by-products. Additionally, pyDOM photochemistry has been associated with the production of high fluxes of reactive oxygen species (ROS), important transients 50 involved in the photo-transformation and photodegradation of pyDOM (Fu et al., 2016;Li et al., 2019;. These studies have contributed to a better understanding of the biogeochemical cycling of py-DOM in the presence of sunlight in the environment. Micro-55 bial (biotic) pathways are another degradative pathway with high potential for altering and/or mineralizing pyDOM, but these pathways are far less understood.
Biotic reworking of organic molecules is a key mechanism for producing the diverse molecular composition of nat-60 ural organic matter (Lechtenfeld et al., 2015;Hach et al., 2020). Due to the highly condensed character of pyOM, it is often regarded as bio-recalcitrant, though several studies have shown that a fraction of it (about 0.5 % to 10 %) is indeed biodegradable (Kuzyakov et al., 2009(Kuzyakov et al., , 2014Zimmer-65 man, 2010;Zimmerman et al., 2011). PyOM is mainly comprised of ConAC (Bostick et al., 2018;, which contributes to its low biodegradability (Zimmerman, 2010). By contrast, pyDOM is highly heterogeneous , and, in addition to ConAC, it contains nu-70 merous low molecular-weight (LMW) species (e.g., acetate, methanol, formate; Bostick et al., 2018; as well as various pyrogenic aliphatic compounds and inorganic nutrients (Hockaday et al., 2007;Mukherjee and Zimmerman, 2013;. 75 The high solubility of pyDOM is imparted by the greater abundance of polar functional groups, which would also allow for greater microbial accessibility. To date, there is no study that evaluates the molecular-scale biodegradability of pyDOM. It is unknown whether and how (e.g., mechanis-80 tic pathways, kinetic rates) the different compound groups of pyDOM are biodegraded and/or biotransformed.
In addition to the unexplored biodegradability of pyDOM, there are concerns that biochar leachates may be toxic due to the presence of condensed and ligninaceous aromatics. It 85 has been shown that cellulose-and pinewood-derived biochar water extracts (i.e., pyDOM) inhibit the growth of cyanobacteria, while pyDOM of lignin-derived biochar has no inhibitory effects (Smith et al., 2016). The toxicity of pyDOM has been mainly attributed to polysubstituted phenols present 90 in the cellulose-and pinewood-derived biochars. In natural systems, however, it is likely that other pyDOM components also play a role in controlling the biodegradability and toxicity of pyDOM. An important very recent finding is that pyOM and pyDOM contain organochlorine com-95 pounds (both aliphatic and aromatic; , which may enhance the toxicity of pyDOM. Thus, biotic incubations of pyDOM are needed to reveal if microbial growth can be sustained in pyDOM and environments rich in ConACCE10 .

100
To explore these questions, we incubated aqueous biochar leachates (i.e., pyDOM) with a soil-derived microbial consortium and evaluated the compositional changes to pyDOM using numerous analytical techniques. Laboratory-produced biochars can be considered model pyrogenic substances as 105 they are similar to what is produced during natural wildfires (Santín et al., 2017) but have not experienced environmental aging which would have impacted their physico-chemical properties (Ascough et al., 2011). We have used oak wood because most riverine dissolved organic matter (DOM) is exported from forested catchments . We used two pyrolysis temperatures (400 and 650 • C) representative of forest fire temperatures (Santín et al., 2015(Santín et al., , 2016. As photochemistry has been shown to increase the 5 bio-lability of various types of DOM (Kieber et al., 1989;Lindell et al., 1995;Wetzel et al., 1995;Benner and Biddanda, 1998;Moran and Covert, 2003;Qualls and Richardson, 2003;Obernosterer and Benner, 2004;Abboudi et al., 2008;Chen and Jaffé, 2014;Antony et al., 2018), we also 10 incubated pyDOM that had been photo-irradiated. Previous studies showed that significant compositional changes occur to pyDOM during photo-irradiation, which certainly implies different biodegradability . 15 In a parallel study of the same incubations (Bostick et al., 2021), we quantified the total organic carbon (TOC)CE11 loss, respired CO 2 quantities and the changes to the bulk structural composition of pyDOM as determined by onedimensional 1 H nuclear magnetic resonance (NMR) spec-20 troscopy. Additionally, in that study, benzenepolycarboxylic acid (BPCA) molecular markers were used to quantify the changes specific to the fraction of ConACCE12 pyDOM. It was found that the pyDOM leachates derived from the biochar of the higher pyrolysis temperature (650 • C) were 25 less biodegradable than the pyDOM leachates from the lower temperature (400 • C) biochar. As expected, photo-irradiation increased the bio-lability of pyDOM. Over the 96 d incubation, up to 48 % of the carbon was respired to CO 2 . The degradation followed first-order kinetics, with LMW com-30 pounds (e.g., acetate, formate, methanol) being preferentially degraded. To elucidate the molecular-level changes taking place during the bio-incubation of pyDOM and assessCE13 the various molecules that are being degraded or produced by soil biota, we employed ultrahigh-resolution mass spec-35 trometry (Fourier transform-ion cyclotron resonance-mass spectrometry, FT-ICR-MS), two-dimensional NMR and fluorescence spectroscopy. The collective results from these two studies improve our understanding of the degradative pathways of pyDOM and ConAC in the environment and allow 40 us to better interpret observations pertaining to terrestrial-tomarine transfers and global cycling of organic matter.

Preparation of pyDOM samples
Two biochars were prepared by heating laurel oak wood 45 (Quercus hemisphaerica) under N 2 atmosphere at 400 and 650 • C for 3 h. After grinding and sieving to particles of uniform size (0.25-2.00 mm), the biochars were leached in 18.1 m MilliQ laboratory-grade water (5 g in 500 mL) over 50 h on a shaker table. The obtained pyDOM leachates, here-50 after referred to as "Oak 400 Fresh" and "Oak 650 Fresh", were filtered using 0.2 µm Millipore GSWP mixed cellulose ester filters. Physico-chemical characteristics of similarly produced solid biochars and their leachates were reported in several previous studies (Zimmerman, 2010;Mukherjee 55 et al., 2011;Bostick et al., 2018;. A fraction of each leachate was also subjected to photoirradiation for 5 d in a custom-made solar simulator equipped with Q-Lab Corporation UV-A lamps (295-365 nm, λ MAX = 340 nm, 40 W) equivalent to natural photo-irradiation of 60 12 d. Photo-transformation rates, structural changes, photoirradiation apparatus design and other relevant information have been published previously . Photo-irradiated pyDOM samples will be hereafter referred to as "Oak 400 Photo" and "Oak 650 65 Photo". The four samples (Oak 400 Fresh, Oak 650 Fresh, Oak 400 Photo, Oak 650 Photo) were diluted to a uniform TOC concentration of 4.7 mgC L −1 prior to inoculation.

Microbial incubations of pyDOM
Microbial incubations were performed using a soil-derived 70 microbial consortium as an inoculum. Soil from the Austin Cary Memorial Forest (Gainesville, FL) was chosen because this area is frequently subjected to prescribed burns (Johns, 2016), and its soil microbes likely interact with pyOM and pyDOM on a regular basis. Taxonomic details of the soil 75 used have been published previously (Khodadad et al., 2011). The collected soil was treated to remove roots and detritus, and its water extract was centrifuged to obtain a pellet. The pellet was then dissolved in 10 mL MilliQ laboratory-grade water to obtain an inoculate, 100 µL of which was used to 80 spike 50 mL of each pyDOM substrate. Additionally, microbial nutrients (KH 2 PO 4 and (NH 4 ) 2 SO 4 ) were provided following Zimmerman (2010) to support a healthy growth medium. Samples were incubated in gas-sealed amber vials on a shaker table at 28 ± 5 • C for 10 d in the dark. Using 85 a double-needle assembly, CO 2 -free air (Airgas, Zero) was flushed through the samples on days 0, 2, 5 and 10, which oxygenated the samples and removed dissolved inorganic carbon for its measurement (reported by Bostick et al., 2021). A procedural blank and control samples were prepared in 90 the exact same way but were poisoned with HgCl 2 immediately following the mixing of the different components (py-DOM, inoculate, nutrients). Additionally, a solution of sucrose (0.5 g C 12 H 22 O 11 in 40 mL MilliQ laboratory-grade water) was also incubated in the same manner to serve as 95 a positive control. All incubated samples were poisoned with HgCl 2 to terminate microbial activity before shipment to Old Dominion University (Norfolk, VA) for analysis. Prior to spectroscopic analysis (see Sect. 2.3 and 2.5 below) or spectrometric analysis (see Sect. 2.4 below), samples were 100 filtered using acid-washed 0.1 µm Teflon (PTFE) syringe filters. Further details about sample preparation can be found in the parallel study (Bostick et al., 2021). Chromophoric DOM (CDOM) measurements were performed on a Thermo Scientific Evolution 201 ultravioletvisible (UV-Vis) spectrophotometer operated in a double-5 beam mode. A matched Starna quartz cuvette with MilliQ water was used as a reference during all spectral measurements. Spectra were recorded from 230-800 nm using a 1 nm step, 0.12 s integration time and 500 nm min −1 scan speed. In addition to the double-beam referencing, the average noise 10 in the 700-800 nm spectral region was subtracted from the spectra to correct for any instrument baseline drifts, temperature fluctuations, as well as scattering and refractive effects (Green and Blough, 1994;Helms et al., 2008). After consecutive procedural-blank corrections, the spectra (kept 15 in decadic units) were normalized to the cuvette path length (1.0 cm) and TOC content (in mgC L −1 ) to convert them to specific absorbance spectra (L mgC −1 cm −1 ; Weishaar et al., 2003). CDOM was quantified by integrating the spectra from 250-450 nm (Helms et al., 2008), and CDOM quantity is re-20 ported in L mgC −1 cm −1 nm units. Fluorophoric DOM (FDOM) measurements were performed on a Shimadzu RF-6000 spectrofluorometer operated in 3D acquisition mode. Samples were analyzed without dilution as no sample yielded absorbance at 230 nm above 25 0.07 (Miller et al., 2010). Samples were excited from 230-500 nm (5 nm step) and emission was recorded over 250-650 nm (5 nm step) to obtain excitation-emission matrices (EEMs). Additionally, five replicate water Raman scans were acquired on MilliQ water in 2D emission mode by exciting 30 the sample at 350 nm, and fluorescence intensity was monitored over 365-450 nm (0.5 nm steps). All measurements were done with 5 nm slit widths of the monochromators, with 600 nm min −1 scan speed and in high-sensitivity mode.
EEMs were processed in MATLAB using the drEEM 35 toolbox (version 0.4.0.) using previously published routines (Murphy et al., 2010(Murphy et al., , 2013. Briefly, using the FDOMcorrrect function, the raw EEMs were adjusted for instrumental bias, blank-corrected using an EEM of the procedural blank and scaled to adjust for any inner-filter effects using the raw 40 UV-Vis spectra (Kothawala et al., 2013). This function also normalized the EEMs to Raman units (RU) after the area of the water Raman peak (peak maximum at 397 nm) had been determined by the ramanintegrationrange function (Murphy, 2011) on the averaged water Raman spectrum. The EEMs 45 were then processed using the smootheem function to remove first-and second-order Rayleigh signals and Raman scattering. EEMs are visualized and difference plots are generated using an in-house MATLAB script. Procedural-blank, control and incubated samples were loaded onto solid-phase extraction cartridges (Agilent Technologies Bond Elut Priority Pollutant (PPL), 100 mg styrene divinyl copolymer) as previously described (Dittmar et al., 55 2008). Cartridges were eluted with methanol (Fisher Scientific, Optima LC-MS grade) and infused into an Apollo II electrospray ionization (ESI) source interfaced with a Bruker Daltonics Apex Qe FT-ICR-MS operating at 10 T and housed in the College of Sciences Major Instrumentation Cluster 60 (COSMIC) facility at Old Dominion University (Norfolk, VA). The instrument is externally calibrated daily with a polyethylene glycol standard, and a surrogate laboratory py-DOM standard was analyzed before and after the analytical sequence to verify for the lack of instrumental drift. Addi-65 tionally, an instrumental blank of methanol was analyzed between samples to verify for the absence of sample carryover. Samples were analyzed in negative ionization mode. ESICE14 spray voltages were optimized for each sample to ensure consistent spray currents among all samples. For each sample, 70 300 transients with a CE15 time domain were collected and co-added, and the resultant free induction decay was zerofilled and sine-bell apodized. After fast Fourier transformation, internal calibration of the resultant mass spectra was performed using naturally abundant fatty acids, dicarboxylic 75 acids and compounds belonging to the CH 2 -homologous series as previously described (Sleighter et al., 2008). Then, using an in-house MATLAB script, salt, blank and isotopologue ( 13 C, 37 Cl) peaks were removed. Molecular formulas within ±1 ppm error were assigned to FT-ICR-MS spec-80 tral peaks (S/N ≥ 3) using the molecular formula calculator from the National High Magnetic Field Laboratory (Tallahassee, FL). Formula assignments were restricted to elemental composition of 12 C 5−∞ , 1 H 1−∞ , 14 N 0−5 , 16 O 0−30 , 32 S 0−2 , 31 P 0−2 , and 35 Cl 0−4 and were refined using previ-85 ously established rules (Stubbins et al., 2010). Any ambiguous peak assignments were refined by inclusion within homologous series (CH 2 , H 2 , COO, CH 2 O, O 2 , H 2 O, NH 3 , HCl) following Kujawinski and Behn (2006) and Koch et al. (2007). For all samples, at least 80 % of the mass spectral 90 peaks were assigned, and they accounted for at least 93 % of the mass spectral magnitude. Molecular composition was evaluated by plotting the molecular formulas on van Krevelen (vK) diagrams, scatterplots of the formulas' hydrogen-to-carbon (H/C) versus 95 oxygen-to-carbon (O/C) ratios (Van Krevelen, 1950;Kim et al., 2003). Formulas were further categorized using the modified aromaticity index (AI MOD ), a proxy for the aromatic character of molecules Dittmar, 2006, 2016), and calculated as shown in Eq. (1).

Fourier transform-ion cyclotron resonance-mass
Please note the remarks at the end of the manuscript.
Formulas were classified as follows: condensed aromatic compounds (ConAC, AI MOD ≥ 0.67, number of C-atoms ≥ 15), aromatic (0.67 < AI MOD ≤ 0.50), olefinic / alicyclic (0 < AI MOD < 0.50) and aliphatic (AI MOD = 0). Additionally, N-containing formulas falling 5 in the ranges of 1.5 ≤ H/C ≤ 2 and 0.1 ≤ O/C ≤ 0.67 were classified as peptide-like. Statistical evaluation of means using one-way analysis of variance (ANOVA) was performed in MATLAB using the anova1 function. Post hoc Scheffé's assessments were performed using the multcompare func-10 tion. Pearson correlations were performed using the corrcoef function in the same software. A confidence level of 95 % was used for all statistical assessments. For the Kendrick mass defect (KMD) series analysis (described later in the paperCE16 ), Kendrick mass (KM) was 15 first calculated using the molecular weight of each compound (i.e., calculated mass from its molecular formula) following Eq. (2). Then, the Kendrick nominal mass (KNM) was calculated as the integer (no decimals) of the KM as shown be-lowCE17 . The KMD is the difference between KM and KNM, 20 i.e., the decimals (Eq. 3). This analysis was performed for oxygen (O), carbonyl (CO) and carboxyl (COO) series (S). preparation and acquisition of 1D 1 H spectra are published in the companion study (Bostick et al., 2021). To obtain ultraclean NMR spectra, NMR tubes were soaked with aqua regiaCE18 , rinsed extensively with ultrapure water and individually tested as blanks to verify that no background peaks 45 are present. While 1 H spectra were originally processed using an exponential multiplication function (line broadening) of 5 Hz to obtain higher signal-to-noise for a more accurate and precise integration (Bostick et al., 2021), here they were re-processed using a multiplication function of 1.5 Hz to bet- 50 ter observe the splitting (multiplicity) patterns of the peaks of interest. TOCSY spectra were acquired using the phasesensitive gradient-enhanced mlevgpphw5 pulse program. It utilizes a 17-step Malcolm Levitt (MLEV-17) composite scheme (Bax and Davis, 1985) for magnetization transfer be-55 tween any coupled nuclear spins and a W5-WATERGATE element for water suppression (Liu et al., 1998). Both shortrange and long-range spin-spin couplings were observed using 30 and 100 ms mixing times, respectively. The data were then zero-filled to a 4096×1024 matrix and then fitted with a 60 π/2-shifted (SSB = 2) sine-squared window function. Linear prediction to 256 points was used in the F 1 dimension. All spectra were internally calibrated to the sharp distinguishable methanol singlet at 3.34 ppm (Gottlieb et al., 1997), and then spectra were phased and baseline-corrected. T 1 -noise 65 removal was performed by calculating the positive projection of rows with no resonances, and the summed projections were subtracted from all rows in the spectrum (Klevit, 1985). The same procedure was performed for all columns (i.e., the F 2 dimension). 70 3 Results

Molecular changes to pyDOM after microbial incubation
Ultrahigh-resolution mass spectrometric analysis of the bioincubated and corresponding control pyDOM leachates re-75 vealed significant changes in molecular composition after the 10 d incubation (Fig. 1). The identified molecular formulas were classified into one of three groups using a presenceabsence approachCE19 (Stubbins et al., 2010;Sleighter et al., 2012). This approach identifies any common formulas 80 among the two samples being compared (control and bioincubated), as well as any formulas that are unique to each sample. It is important to note that the electrospray ionization (ESI) source is prone to biases, and the analytical window of FT-ICR-MS depends most critically on it. Thus, it may not 85 identify compounds that are present if they are not ionizable (Stenson et al., 2002;Patriarca et al., 2020). Therefore, our observations are influenced by the limited analytical window, and it is essential that observations by FT-ICR-MS are always paired with supplementary quantitative techniques (op-90 tical analyses, NMR, etc.) in order to determine if the identified trends are real or an artifact of ESI charge competition (D'Andrilli et al., 2020). In all samples, nearly a third of the formulas (23 %-31 %) present in the control samples were not observed after the 95 biotic incubations. This is somewhat proportional to the organic carbon losses observed over the 10 d incubation by Bostick et al. (2021). The organic carbon loss was also found to be equivalent to mineralized CO 2 (± 4 %, Bostick et al., 2021) indicating that microbial respiration had occurred al-100 though CO 2 mineralization can happen abiotically as well. Using the number of formulas lost as a proxy for bio-lability Formulas are classified as bio-labile (formulas only found in the "killed" control pyDOM leachates) and bio-produced (formulas only found in the bio-incubated samples). Formulas that are present in both the control and bio-incubated samples are operationally classified as bio-resistant and not shown for clarity. These three classes of molecules are separately plotted on vK diagrams and shown in Sect. S2 of the Supplement (Figs. S2-S4). The number of formulas found in each of these pools is listed in the legends along with corresponding percentages (relative to total number of formulas in the two samples being compared). The carbon losses quantified by Bostick et al. (2021) are listed under the legends. The black lines indicate modified aromaticity index cutoffs (AI MOD ; Dittmar, 2006, 2016), and the red box indicates the peptide region (valid only for N-containing formulas). TS2 here, it appears that Oak 400 Fresh (1646 bio-labile formulas, 16 % carbon loss) is more bio-labile than Oak 650 Fresh (1364 bio-labile formulas, 15 % carbon loss). This was expected because of the richness of Oak 400 Fresh in smaller less aromatic compounds . 5 Upon photo-irradiation, both Oak 400 Fresh and Oak 650 Fresh experience significant changes in their molecular composition as previously described in detail by . The photo-transformed pyDOM is much more aliphatic and richer in nitrogen and LMW compounds which 10 render CE21 pyDOM much more biologically labile . Surprisingly, it was found that Oak 400 Fresh (1646 bio-labile formulas) is more bio-labile than its photo-irradiated counterpart (Oak 400 Photo, 1242 bio-labile formulas). However, this observation using molecular data 15 does not agree with quantitative carbon loss results for the 10 d incubation (Oak 400 Fresh: 16 % carbon loss; Oak 400 Photo: 25 % carbon loss). The observed discrepancy is because LMW compounds contribute to a large fraction of the degraded carbon in the Oak 400 pyDOM systems and LMW 20 species are not observed following the employed PPL sample preparation and FT-ICR-MS detection. A similar discrepancy is observed when comparing Oak 400 Photo (1242 biolabile formulas, 25 % carbon loss) and Oak 650 Photo (1410 bio-labile formulas, 23 % carbon loss). In contrast, Oak 650 25 Fresh (1364 bio-labile formulas) was observed to be less bio-labile than Oak 650 Photo (1410 bio-labile formulas) via both FT-ICR-MS and the observed quantitative carbon losses (Oak 650 Fresh: 15 % carbon loss; Oak 650 Photo: 23 % carbon loss). LMW species are less abundant in the Oak 650 pyDOM systems resulting in consistent trends between the analyses. Biodegradability trends derived from FT-ICR-MS molecular data match those from the UV-Vis data from chromophoric pyDOM (Fig. S1 in the Supplement) revealing a similar inability of UV-Vis to detect LMW compounds which do not absorb UV-Vis light. In summary, we observe a degradation of a variety of different molecular classes as well as a production of many molecules that appear to be of high biological lability. However, we caution that there are observed discrepancies among carbon loss and molecular/chromophoric CE22 data for the Oak 400 pyDOM systems, an observation that highlights the need to clearly understand methodological analytical windows when interpreting molecular and spectroscopic data.
Interestingly for all leachates, the degraded (bio-labile) molecules were not from a specific area of the vK diagrams 20 but rather represent a broad range of H/C and O/C ratios and compound types (see Fig. S2). This variety of compound characteristics among bio-labile molecules suggests that the degradation pathway may not be from microbial consumption alone. It would be unlikely for the soil microorgan- 25 isms to utilize organic-matter compounds as food indiscriminately. Most interestingly, it is evident that large numbers of aromatic (AI MOD ≥ 0.50) formulas are lost, in agreement with observed losses in CDOM (Fig. S1) and losses in aryl functional groups (measured by 1 H NMR) reported in the 30 parallel study (Bostick et al., 2021). ConAC were found to be resistant to biodegradation (Bostick et al., 2021) and therefore losses of ConAC observed via FT-ICR-MS are considered an artifact due the low ionizability of ConAC and competition processes in the ESI source (Stenson et al., 2002;35 Patriarca et al., 2020). However, the agreement between FT-ICR-MS and other quantitative data (UV-Vis, NMR, TOC) confirms the interpretation of bulk pyDOM degradation.
Approximately half of the formulas (37 %-56 %) in the original pyDOM leachates are classified as bio-resistant us- 40 ing the presence-absence approach (observed before and after biotic degradation). These formulas are located in all areas of the vK diagrams (Fig. S3) showing variable oxygenation and aromaticity. The relative peak magnitudes of these formulas did not change significantly following the incuba-45 tions (R 2 > 0. 95, Fig. S9;Sleighter et al., 2012), also suggesting that a wide variety of pyDOM molecules are resistant to microbial degradation. Using the available molecular data, it is not possible to attribute the observed recalcitrance to any molecular property. Therefore, it is likely that some 50 of these bio-resistant molecules are still bio-labile and would have degraded in due time if the incubations were sampled at later time points. Longer time series should be conducted in future studies to fully differentiate among bio-labile and bio-resistant pyDOM molecules.

55
The use of hydrogen-to-carbon ratio (H/C) versus molecular-weight (MW) plots has also been useful in interpreting ultrahigh-resolution mass spectrometry data (e.g., Gonsior et al., 2018;Powers et al., 2019;Valle et al., 2020). Such graphs are presented using the presence-absence ap-60 proach in Figs. S5-S8 in Sect. S3. These graphics help evaluate how different types of compounds (aliphatic versus aromatic) change relative to their MW. For both Oak 400 leachates, it is clear that large aromatic molecules (H/C < 1.5, MW > 550 Da) are removed during the biotic 65 degradation, and smaller (300 < MW < 550) aromatic compounds are produced. The consumption of large molecules indicates that microbes utilize extracellular enzymes to produce ROS, which degrade larger molecules into smaller substrates (Billen et al., 1990). The large aromatic molecules 70 that are being degraded into smaller ones are mainly ligninaceous and not ConAC, in agreement with the insignificant changes in BPCA data published by Bostick et al. (2021). With regards to the aliphatic molecules (H/C > 1.5), it is clear that molecules of a wide range of MW are degraded 75 and produced during the incubations suggesting that MW is not a critical factor in their bio-lability CE23 . This is in apparent disagreement with the general knowledge that microbes preferentially consume LMW substrates (e.g., Søndergaard and Middelboe, 1995). Bostick et al. (2021) also concluded 80 that LMW substances are preferentially degraded in the incubations of pyDOM. The observed production of higher MW aliphatics suggest that the microbial incubations were still very active at the point of sampling (10 d). This additionally suggests that future studies need to evaluate the molecular 85 composition of biotically incubated pyDOM over a longer timescale.

Composition of bio-produced organic matter
The bio-produced organic compounds can be evaluated in various ways to examine the processes that may have oc-90 curred during the incubations. Using a presence-absence approach (Sleighter et al., 2012), the bio-produced formulas of each sample are compared with those of the other samples (Table 1). No significant overlap was found (2-320 formulas in common, 0 %-12 %) among the molecules produced in 95 the incubated pyDOM samples. Furthermore, no significant match was found between the bio-produced formulas of incubated pyDOM and those of theCE24 sucrose control sample (63-94 formulas in common, 3 %, Table 1). These observations indicate that the products of the incubations were vastly 100 different for each sample and likely depend on the starting substrate. An alternative explanation is that bio-produced formulas were further altered post-exudation by ROS to result in their molecular diversification.
A significant fraction of the bio-produced organic 105 matter was characterized as peptide-like (N-containing, 1.5 ≤ H/C ≤ 2.0, 0.1 ≤ O/C ≤ 0.67). This indicates that microbes convert a part of pyDOM into labile DOM (Moran 8 A. I. Goranov et al.: Labilization and diversification of pyrogenic dissolved organic matter by microbes Table 1. Overlap of bio-produced molecular formulas among samples. The number of formulas corresponds to the formulas in common between the two samples being compared, and the percentage is relative to the total number of formulas in the two formula sets.

Sample
Oak 400 Vorobev et al., 2018), a process hereafter referred to as "microbial labilization". Given that the pyDOM samples used in this study were poor in organic nitrogen, the microbes must have used the inorganic nitrogen (NH + 4 ) that was provided as a nutrient and converted some or all of 5 it into microbial biomass. Peptide-like formulas comprised 22 %-40 % of the bio-produced formulas (Table S2). The results of the comparative analyses (Table 1) imply that these proteinaceous formulas are of highly variable composition. Their molecular diversity is additionally evaluated using one-10 way analysis of variance (ANOVA) reported in Sect. S6. This statistical analysis revealed high molecular variability supporting the findings by the presence-absence comparisons presented earlier (Table 1). Collectively, these findings conclude that the microbial incubations of pyDOM created pools 15 of new, very diverse molecules, a process hereafter referred to as "microbial diversification". As FT-ICR-MS was performed with soft electrospray ionization with no fragmentation, the structure of the observed molecules is inferred from the elemental composition of the assigned molecular formu-20 las. Another possibility for these N-containing molecules is that they were formed by coupling reactions among pyDOM molecules with the NH + 4 nutrient that was added to support microbial growth (e.g., via Michael addition reactions; Mc-Kee et al., 2014). 25 To confirm that the bio-produced formulas were associated with proteinaceous structures and are not just N-containing compounds that coincidentally plotted in the peptide region, spectrofluorometric analysis was performed to obtain excitation-emission matrices (EEMs) of the pyDOM sam- 30 ples before and after bio-incubation (Fig. 2). The data for Oak 650 Photo is not reported as the produced EEM spectra were of questionable quality, and as there were very limited amounts of the sample, analytical replication and quality assessment were not possible. 35 Proteinaceous organic matter has a highly characteristic fluorophoric signature due to the distinguishable signals of the aromatic amino acids tyrosine and tryptophan. The short Stokes' shifts of these fluorophores allow them to spectroscopically separate on the EEM plot allowing for the iden- 40 tification of related labile substances (Wünsch et al., 2019). Other amino acids, namely histidine and phenylalanine, are also fluorophoric but are not easily identified in EEM data of complex matrices. A practical approach to evaluate the change after the bio-incubation is to use difference plots 45 (e.g., Hemmler et al., 2019). For all samples, strong proteinaceous signals evolve after biotic incubations indicating that molecules of autochthonous (microbial) origin are produced (Coble, 1996;Coble et al., 2014). This indicated that peptidelike molecules observed using FT-ICR-MS are not artificially 50 produced by coupling reactions growth (e.g., via Michael addition reactions; McKee et al., 2014) or evolve due to less charge competition in the ESI source. Thus, the protein-like formulas are truly bio-produced, validating the findings of the presence-absence analysis.

55
There are subtle differences among the EEMs of all control and bio-incubated samples indicative of the high variability in fluorophoric content of these samples. This agrees with the observed variability in molecular composition described earlier. An interesting observation is that in the 60 two Oak 400 pyDOM incubations, tyrosine-like fluorescence (peaks B 1 and B 2 ) decreases after biotic incubation, whereas tryptophan-like fluorescence (peaks T 1 and T 2 ) increases. In contrast, the tryptophan-like fluorophores are degraded and tyrosine-like ones are produced after biotic incubation of Oak 65 650 Fresh pyDOM. CE25 It must be noted that there are proteinaceous fluorophores (and peptide-like formulas) in the control samples resulting from the addition of the microbial inoculate, but the associated fluorophores were present in low amounts. Thus, proteinaceous fluorescence signals 70 in the control samples are not unexpected. However, a decrease in proteinaceous fluorophores is the opposite of what is expected after significant microbial growth. Proteinaceous compounds are highly bio-labile and aromatic compounds are susceptible to oxidation by ROS. Therefore, it is pos-75 sible that tyrosine-like fluorophores (in Oak 400 pyDOM) and tryptophan-like fluorophores (in Oak 650 pyDOM) are still actively participating in biodegradation processes as the incubations were still active at the time of sampling (10 d).
The loss of tyrosine-like fluorophores in the Oak 400 sam-80 ples and loss of tryptophan-like fluorophores in the Oak 650 Fresh sample are indicative of different microbial physiology and exudates. The complexity of these EEM spectra and the compound-specific changes observed here indicate that proteomic and/or metabolomic analyses (e.g., Nalven et 85 al., 2020) are necessary in future microbiological studies in Figure 2. Fluorescence excitation-emission matrices (EEMs) of control (left panels) and bio-incubated (middle panels) pyDOM samples. Difference spectra are shown in the right panels. The black box indicates the region where compounds of proteinaceous origin fluoresce (Coble, 1996;Coble et al., 2014), with tyrosine-like (B 1 and B 2 ) and tryptophan-like (T 1 and T 2 ) peaks labeled on the difference plots. order to fully understand the changes to the molecular composition of pyDOM during biotic incubations.
To determine if the bio-produced formulas are from true proteins or from compounds with residual proteinaceous fluorophores, the formulasCE26 were evaluated in the con-5 text of possible combinations of amino acids that would be singly charged. Given that microbes exude large proteins (MW > 30 kDa) such as lignin peroxidases, manganese peroxidases and laccases (Higuchi, 2004), the peptide-like formulas observed by FT-ICR-MS (analytical window of 200-10 1000 Da) may have resulted from hydrolysis of the abovementioned enzymes (or other proteinaceous exudates). If that is the case, the hydrolysates would likely have had a simple oligomeric composition. To test this, the bio-produced peptide-like formulas in each sample were compared to 15 a library of 888 009 possible combinations of 20 amino acids (oligomeric sequences of two to seven residues). Only a small number of oligopeptides were identified (5-18 oligopeptides of 2-5 amino acids, Tables S2 and S3) which is counter to the proposed idea that hydrolysis of 20 microbial exudates produced these newly observed peptidelike formulas. Therefore, the observed bio-produced formulas may represent compounds with residual proteinaceous fluorophores and are not true oligopeptides. The lack of iden- Figure 3. Two-dimensional 1 H-1 H total correlation spectroscopy (TOCSY) NMR spectra of the bio-incubated Oak 650 Fresh sample. Short-and long-range couplings were allowed to evolve during mixing times (τ ) of 30 (blue) and 100 ms (red), respectively. The 1D 1 H spectrum is shown as a projection on top (black). tified oligopeptides also calls into question the idea that microbial processes were solely responsible for the high variability of the bio-produced organic matter observed after the microbial incubation of pyDOM.
To further elucidate the composition of these bio-produced 5 N-containing substances, we re-evaluated the previously published 1 H NMR data of these samples (Bostick et al., 2021) in greater detail. Additionally, the connectivity between previously observed functional groups was assessed using two-dimensional 1 H-1 H total correlation NMR spectroscopy (TOCSY) on a select sample. Figure 3 shows the TOCSY spectra of the bio-incubated Oak 650 Fresh sample.
There are three groups of resonances representing an alicyclic structure, a β-hydrogen to a heteroatom and a methylene group that were found in all samples, even in the controls 15 (although with small contributions relative to the total spec-tral signal). These resonances have not been previously observed in the 1 H NMR spectra of these pyDOM samples (Bostick et al., 2018; indicating that they represent by-products of the microbial incubations, likely mi-20 crobial biomass. In the control samples, the compounds associated with these resonances must be from the soil inoculant that was added. The three resonances are observed to be in the same coupling network indicating that they are a part of the same or similar structures. Due to the very low 25 concentration of these samples (3.5-4 mgC L −1 ), the NMR analysis did not allow for a high-resolution structural elucidation, but some distinct signatures were nonetheless observed. The deshielded aliphatic peaks at δ = 2.1 − 2.3 ppm have a complex multiplicity pattern, a characteristic feature 30 of alicyclic structures. These are likely residual carbohydrate moieties which have lost most of their O-containing groups through various cleavage processes, and their backbone C alicyclic -H resonances have been shifted upfield. The peak at 1.55 ppm is from β-hydrogens to a heteroatom (H-35 C β -C α -X, where X = O, N, S), and these are known to be associated with peptidoglycans (Spence et al., 2011). The TOCSY analysis was performed with two different mixing times (τ = 30 and τ = 100 ms) in order to evaluate shortrange (2-3 bond) and long-range (4-6 bond) connectivities. 40 Based on the couplings the observed resonances are vicinal to each other (three bonds away). This indicates that these functional groups are closely bound in the peptidoglycan substances they likely represent.
All of the assessments described above conclude that the 45 observed biochemical processes in these pyDOM incubations are complex and difficult to unambiguously interpret. Based on our findings, we summarize that the bio-produced formulas (Fig. 1) can originate from three possible sources: 1. exoenzymes, which microbes use to extracellularly de- 50 grade larger molecules into smaller ones (Hyde and Wood, 1997;Higuchi, 2004); 2. peptidoglycans, which likely leached into solution after bacterial death and cell lysis (Yavitt and Fahey, 1984); and 55 3. other metabolites and exudates involved in the physiology of the different microbes in the consortium used (e.g., signaling compounds).
The significant degradation of pyDOM and production of these biological compounds indicates that microbes success-60 fully converted the presumably carbon-rich recalcitrant pyrogenic molecules into more labile substances, a process we define as microbial labilization. However, the fact that the observed bio-produced labile molecules are not identifiable as simple oligopeptides and are present in significantly dif-65 ferent composition among the four samples suggests that this molecular diversity may not be caused by predictable biotic reactions but by random radical-driven processes. Further evidence for the random radical-driven processes comes from the observed degradation of molecules across the whole vK space (Figs. 1 and S2), which is unusual because microbes preferentially consume smaller aliphatic species (Berggren et al., 2010a, b;Kirchman, 2018).

Radical oxygenation as a potential source of molecular diversity
Microbial physiology has been associated with the production of ROS, which have been shown to be important in the degradation of various types of organic compounds (e.g., Scully et al., 2003;McNally et al., 2005;Porcal et al., 10 2013; Trusiak et al., 2018;Xiao et al., 2020). Recent studies showed that radicals can degrade various types of ligninaceous molecules (Waggoner et al., 2015 suggesting that microbially induced radical reactions can target a variety of pyDOM molecules as 15 well. While there were no ROS measurements made in this study, we have performed Kendrick mass defect (KMD) analysis of the FT-ICR-MS data (Kendrick, 1963;Hughey et al., 2001) to seek evidence for radical processes. The KMD analysis identifies formulas that differ by any repeating struc-20 tural moiety (e.g., -CH 2 -). To identify potential products of radical attacks, we have evaluated the FT-ICR-MS data in the context of oxygenation, i.e., searched the mass lists for formulas differing by one oxygen atom (addition of hydroxyl group), carbonyl group (addition of aldehydes or ketones) 25 and carboxyl group (Fig. 4). The mathematics behind the KMD analysis (see Sect. 2.4) convert the mass of the molecular formula (also known as the IUPAC mass) to a "Kendrick" mass placing the formula on a scale that is specific for the selected struc-30 tural moiety. In Fig. 4a, an example is shown with the KMD analysis for molecules differing by one oxygen (-O-). On the regular (IUPAC) mass scale, such formulas would differ by 15.994915 Da, but on the Kendrick "O" mass scale, they differ by 16 Da. The difference be- 35 tween the Kendrick mass (e.g., KM = 408.2876 Da) and the Kendrick nominal mass (e.g., KNM = 408 Da) is the Kendrick mass defect, KMD (i.e., KMD = 0.2876 Da). Formulas with the exact same KMD differ by one or more oxygens and lie on a KMD series. Visually these formu- 40 las would plot on horizontal lines on the KMD plot as indicated by the dashed lines in Fig. 4b. Taking the series of KMD = 0.4174 Da as an example, the KMD evaluation shows that there are five formulas in this particular KMD series that differ in the number of oxygens (C 24 H 40 O 5-10  gens can be formed via oxygenation by hydroxyl radical ( q OH) attacks (Waggoner et al., 2015. This ROS can abstract a hydrogen from C-H bonds and the hydrogen is substituted with an OH group, resulting in the formation of alcohols (C-OH) as shown in 55 Fig. 4c. This is the suggested pathway of how the oxygenation products shown in Fig. 4a and b have formed. Evidence for such reactions will be found on the KMD plots as the evolution of new molecules within the same KMD series but with a different number of oxygens. Further radical attacks 60 would produce polyols (Fig. 4c). In the case of formation of geminal diols (two alcohol groups on the same carbon atom), they can rearrange to aldehydes or ketones via keto-enol tautomerism (Fig. 4d). Further radical attacks would produce carboxyl groups, which can also be radically cleaved, and 65 pyDOM radicals would be formed. PyDOM radicals (as well as any other radical intermediate in this pathway) can be then further paired with hydrogen radicals ( q H) from the solution, with other q OH radicals or with other radicalized pyDOM or proteinaceous species. 70 Using KMD analysis, formulas that could have been produced by oxygenation were identified and plotted individually (Fig. 5). It is assumed that the smallest molecule in each series is the substrate and any molecules with increasing number of oxygens are oxygenation products.

75
KMD analysis revealed that about a third (34-748, 3 %-42 %) of the bio-produced formulas could be classified as products of oxygenation reactions, likely driven by ROS species such as the hydroxyl radical ( q OH). This is in agreement with previously observed cross-linking of microbial 80 compounds through oxidative processes (Sun et al., 2017). The majority of the bio-produced formulas, however, were not found to be products of oxidation as they did not lie on any of the evaluated KMD series (O, CO or COO). Therefore, the majority of the bio-produced formulas are likely 85 formulas of exudates which were resistant to radical attacks or are formulas of compounds which have already been radically coupled with other compounds to result in unrecognizable molecules by the KMD analysis.
Additional evidence for intense radical processes in 90 these systems is the evolution of bio-produced unsaturated aliphatic compounds (1 < H/C < 2, O/C < 2) on the vK diagrams ( Figs. 1 and S4). ROS can attack aliphatic and aromatic compounds, open aromatic and alicyclic rings, cleave oxygen-or nitrogen-containing functionalities, and produce 95 highly aliphatic molecules, as previously observed after photo-irradiation of pyDOM , ConAC (Zeng et al., 2000a, b), and radical-based degradation of lignin (Waggoner et al., 2015Khatami et al., 2019a, b). ROS can also attack any 100 of the proteinaceous exudates and peptidoglycans, cleaving them from many of their functional groups and converting them into the observed unsaturated aliphatic compounds. These produced aliphatic compounds could also contribute to the newly produced N-containing peptide-like compounds 105 observed by FT-ICR-MS if they are oxygenated by ROS post-formation. However, this seems unlikely as data from the supplementary fluorescence and NMR analyses support after a sequential attack by hydroxyl radicals ( q OH). Panel (c) shows possible chemical reactions that can cause an increase in the number of oxygens. Panel (d) shows further oxidative processes involving the formation of keto and carboxyl groups, processes which ultimately produce pyDOM radicals and CO 2 . The KMD plots for all samples are shown in Figs. S10-S12. the formation of microbial biomass. The KMD analysis shown here strongly suggests the presence of intense radical processes as formulas with increasing numbers of oxygen atoms are known to be formed following radical oxygenation (Waggoner et al., 2015. 5 However, it must be noted that this KMD analysis does not directly prove the existence of radical processes and the suggested radical processes are speculation CE28 based only on indirect observations. Future studies need to directly test for the presence of radical reactions by performing biotic incu-10 bations of pyDOM with radical quenchers as well as by quantifying radical fluxes in these microbiological systems. While FT-ICR-MS peak magnitudes are a function of molecular ionizability, making it generally impossible to quantify the different bio-labile and bio-produced com-15 pounds of our study, the ultrasensitivity of this technique ensures the detection of all compounds that are within the FT-ICR-MS analytical window. Here, the number of molecular formulas can be used as a quantitative measure for molecular diversity (e.g., Gurganus et al., 2015). Previously pub-20 lished liquid-state 1 H NMR data for the same samples provide a quantitative measure of functional group content (Bostick et al., 2021). Significant positive and negative correlations were observed between the numbers of bio-labile and bio-produced formulas and the percent NMR spectral signal 25 accounted for by olefinic functionalities and methanol ( Fig. 6 and Table S4). These correlations suggest that the diversity of biodegraded (bio-labile) and bio-produced molecules was related in some way with a process related to the availability Formulas not part of any of the oxygenation KMD series (O, CO or COO) are colored in gray. Formulas in dark green are substrates with their oxygenation products colored in light green. The number of formulas in each of these pools is shown in the legends (along with corresponding percentages). The black lines indicate modified aromaticity index cutoffs (AI MOD ; Dittmar, 2006, 2016). of methanol (CH 3 OH) and olefinic functionalities (C=C) in pyDOM.
Olefinic functionalities have been recently identified as important structural motifs in the composition of pyDOM and were observed to degrade in photochemical experiments 5 likely due to their high reactivity with ROS species . Although olefins are in low abundance in pyDOM (< 10 %), it is likely that they act as important intermediates in the degradative pathways of pyDOM . Olefinic bonds can be homolytically cleaved when 10 attacked by radicals and effectively act as radical accelerators that further propagate radical-mediated organic-matter transformations. Thus, the abundance of olefins can further increase the abundance of radicals and contribute to the elevated molecular diversity resulting in the significant correla-15 tion shown in Fig. 6.
The other significant correlation between molecular diversity and NMR data is observed to be with methanol (CH 3 OH), a very sharp highly distinguishable singlet at δ = 3.34 ppm in 1 H NMR spectra (Gottlieb et al., 1997). As it 20 is a common contaminant in NMR analysis, special precautions were taken to obtain ultraclean spectra (see Sect. 2.5). Methanol is a species that is naturally present in pyDOM (Bostick et al., 2018), and while it is generally considered to be toxic to microbes (Dyrda et al., 2019), there are methy-25 lotrophic bacteria and fungi (microbes of the families Methylococcaceae and Methylobacteriaceae) that can utilize it as a substrate (Chistoserdova et al., 2003;Kolb and Stacheter, 2013;Chistoserdova and Kalyuzhnaya, 2018). These species have been previously observed in the soil from the area 30 where the microbial inoculum was extracted (Khodadad et al., 2011), suggesting that the degradation of methanol may Figure 6. Pearson correlation analysis between the number of biolabile and bio-produced formulas detected by FT-ICR-MS and relative intensity (in %) of olefinic functionalities (C=C) and methanol (CH 3 OH) as measured by liquid-state 1 H NMR and reported by Bostick et al. (2021). No significant correlations were found between other functional groups and the number of bio-produced or bio-labile formulas (data shown in Table S4). be biotic. In fact, in these samples, methanol, along with the other two measured LMW substances, acetate and formate, was almost completely degraded over the 10 d incubation (Bostick et al., 2021).
The inverse relationship between the content of methanol 5 and molecular diversity (Fig. 6) can be interpreted in several ways. Firstly, methanol could be exhibiting toxicity to the microbes that assimilate pyDOM, as has been observed previously (Dyrda et al., 2019). This, however, is unlikely for the pyDOM systems studied here because the sample with the highest amount of methanol (Oak 400 Photo, ∼ 3.7 % CH 3 OH) was the second-most bio-reactive (Bostick et al., 2021). Instead, the observed significant negative correlation may be explained by the fact that methanol is a known radical scavenger (Múčka et al., 2013). If, as we propose, 15 the molecular diversity results from the activity of radical processes, an increasing concentration of methanol would quench ROS, thereby decreasing the radical activity and limiting the molecular diversity in these systems.

Multiple pathways for the alteration of pyDOM by microbes
Using a variety of analytical platforms in this and the parallel study (Bostick et al., 2021), significant quantitative and qualitative losses were observed when pyDOM was subjected to 25 incubation with a microbial consortium collected from a site impacted by wildfires. Additionally, labile and diverse compounds were produced during these incubations. Due to the high complexity of pyDOM, the changes are not straightforward, and there are at least two degradation pathways at play: 30 (1) degradation through microbial assimilation (consumption of pyDOM) and (2) degradation and/or transformation via radical-mediated reactions (e.g., oxygenation) by ROS produced from microbial exoenzymes. These two pathways are discussed in the context of degradation of pyDOM and the 35 formation of new labile and diverse molecules.

Molecular degradation of pyDOM
A surprising observation in this study is that there was a uniform loss of pyDOM molecules from all regions of the vK diagrams. Microbes, it is generally presumed, prefer-40 entially assimilate small non-aromatic substances such as carbohydrates, proteins and LMW acids (Berggren et al., 2010a, b;Kirchman, 2018). Thus, the aromatic fraction of pyDOM, mainly the ConAC, are generally considered to be bio-resistant (Goldberg, 1985;Masiello, 2004). In addition to 45 the condensed character of many of the molecules, there are significant numbers of potentially toxic organochlorine compounds, of both aliphatic and aromatic character, in pyDOM . Thus, the finding of major biological activity in these pyDOM systems and the significant amount 50 of carbon, including aromatic carbon, that was mineralized is a very significant finding for the wildfire biogeochemistry community (Bostick et al., 2021). Although pyDOM is highly heterogeneous , the observation of diverse molecular biodegra-55 dation is not unique to it. In a recent microbial degradation study of snow DOM, Antony et al. (2017) observed that both aromatic and aliphatic formulas were biodegraded. This is likely due to microbes evolving chemical mechanisms to thrive under extreme glacial conditions (Antony et al., 2016). 60 Analogously, as there have been previous prescribed fires in the area from which the microbes for this study were extracted (Johns, 2016), it is also possible that our microbes had adapted to the presence of pyOM CE29 by developing mechanisms for assimilating the carbon in large, complex 65 molecules, including ConAC (Judd et al., 2007).
While direct microbial assimilation of pyDOM compounds is certainly likely to have occurred, our molecular and spectroscopic findings suggest a second degradative pathway contributing to the extensive molecular alteration 70 and to the significant loss of carbon that was quantified in the parallel study (Bostick et al., 2021). While some microbial exoenzymes operate via hydrolytic pathways (amylases, lipases, proteases, cellulases, β-galactosidases, etc.), many other enzymes operate through oxidative (electron-75 withdrawing) pathways. Examples of such enzymes are the various lignin-modifying enzymes in the peroxidase (lignin peroxidases, manganese peroxidases, etc.) and phenol oxidase CE30 (e.g., laccases) families (Higuchi, 2004). Thus, ROS are usually produced and involved in the microbial 80 degradation of organic matter in the environment.
The bio-labile molecules in the studied pyDOM samples are of highly variable degrees of oxygenation, aromaticity and size. There were large MW compounds (MW > 550 Da) that were degraded indicating that microbial exoenzymes 85 producing ROS would have been needed to reduce the size of these large substrates into smaller units that could pass through microbial cell membranes and be consumed by the biota (Sinsabaugh et al., 1997;Fuchs et al., 2011;Burns et al., 2013). The presence of enzymatic compounds is confirmed by observation of peptide-like compounds (FT-ICR-MS analysis) and proteinaceous fluorophores (spectrofluorometric analysis). An important finding is that a preferential degradation of ConAC of smaller MW was observed (Bostick et al., 2021). As small ConAC (i.e., oxygenated poly-10 condensed aromatic hydrocarbons) are known to be toxic (e.g., Idowu et al., 2019), it is unlikely that they were directly consumed by the microbes. Oxygenated polycondensed aromatic hydrocarbons are highly susceptible to attacks by ROS, which is likely how they were degraded in these samples. 15 Thus, we speculate that microbes are most likely not directly consuming the smaller ConAC, but, rather, smaller ConAC are degraded indirectly via ROS oxidation. Furthermore, ROS can oxygenate pyDOM with various functional groups (e.g., hydroxy, aldehyde/keto, carboxyl), and can also cleave 20 functional groups (e.g., methoxy functionalities), open aromatic rings and completely mineralize compounds to inorganic carbon (CO, CO 2 , HCO − 3 TS3 and CO 2− 3 ) as shown in Fig. 4. ROS have been previously shown to be very important in pyDOM photochemistry (Ward et al., 2014;Fu et al., 25 2016; , and we speculate that they play an important role in the microbial degradation of pyDOM as well. More indirect evidence for radical species involvement is provided by the peptidoglycan molecules produced during 30 the pyDOM incubations. Peptidoglycan molecules are generally large (Vollmer et al., 2008) and would not be detected as singly charged ions using FT-ICR-MS (analytical window covering only m/z 200-1000). The hydrolytic products of peptidoglycans (small oligopeptides) would be observed. 35 Very few peptide sequences (5-18 oligopeptides of 2-5 residues) were identified among the bio-produced formulas, indicating that such hydrolysates did not exist in the samples at the time of measurement. However, if there were abundant radical reactions occurring in the system, as we suggest, it is 40 very possible that these hydrolysates were altered into unrecognizable organic structures that would still be classified as peptide-like but would have different molecular composition than the predicted linear peptide sequences. It is also possible that instead of peptidoglycan hydrolysis followed by consec-45 utive oxygenation, ROS directly cleaved the peptidoglycans into smaller non-linear substances of peptide-like molecular composition.
It must be noted that the results of our study were acquired using negative-mode ESI which is only effective for elec-50 tronegative (carboxyl-rich, hydroxyl-rich) compounds (Stenson et al., 2002;Patriarca et al., 2020). Thus, the trends of degradation and labilization are skewed to fit this criterion and do not provide a complete overview of all molecules that are bio-labile or bio-produced. Future studies should employ 55 positive-mode ESICE31 and/or different ionization sources (such as atmospheric pressure photoionization CE32 ) to better elucidate the molecular degradability of pyDOM.

Labilization and diversification of pyDOM
The production of labile unrecognizable biological sub-60 stances during these incubations correlates well with previous findings showing the formation of thousands of new biological compounds during biotic incubations unrelated to microbial metabolic pathways (Lechtenfeld et al., 2015;Wienhausen et al., 2017;Patriarca et al., 2021). There was 65 no significant overlap among the bio-produced formulas of the four pyDOM samples (2-320 formulas in common, 0 %-12 % overlap). Insignificant numbers of bio-produced formulas from pyDOM were also found in the bio-produced formulas of an incubation of sucrose with the same soil microbes 70 (63-94 formulas, 3 % overlap). This indicates that microbes diversified the composition of the incubated pyDOM.
The observed diversity of bio-produced formulas can be explained by a scenario wherein the microbes secreted molecules whose identities differed depending on the growth 75 medium and/or food source, yielding high variability among bio-produced formulas after the incubation of pyDOM. Additionally, it is possible that different microbial species (different bacteria, fungi, archaea, etc.) have proliferated in response to the sample-specific pyDOM composition, yield-80 ing different microbial populations growing during each different incubation, sequentially producing different bioproduced compounds (Fitch et al., 2018). Another possible explanation for the observed diversification is the presence of ROS-driven oxygenation processes. ROS are known to 85 be key species in the consumption of organic matter by microbes. Microbes can only consume small molecules that could pass through their cell membranes (Sinsabaugh et al., 1997;Fuchs et al., 2011;Burns et al., 2013). Thus, to utilize large molecules as food, microbes produce exoenzymes 90 which generate ROS extracellularly (Hyde and Wood, 1997;Higuchi, 2004). These ROS then degrade the large molecules into smaller ones that are utilized as food. Though not directly proven to exist in this study, many of the observed trends in FT-ICR-MS, NMR and fluorescence data suggest 95 the presence of radicals which diversify the composition of the bio-produced formulas.
Our finding of extreme molecular diversity contrasts with previous observations in a study by Lechtenfeld et al. (2015) evaluating the molecular composition of microbially pro-100 duced DOM. In their study, marine microbes were supplied with two different substrates (glucose and glutamic acid; and a mixture of oligosaccharides and oligopeptides), and a significant overlap (67 %-69 %) in the bio-produced organic matter was observed. The difference in observations 105 between the work presented in this paper and by Lechtenfeld et al. (2015) is likely caused by a large difference in the composition of the pyDOM substrates relative to those 16 A. I. Goranov et al.: Labilization and diversification of pyrogenic dissolved organic matter by microbes in the Lechtenfeld et al. (2015) study. While the four py-DOM samples used here are highly different to one another , the substrates by Lechtenfeld et al. (2015) were of much higher similarity (glucose, glutamic acid, oligosaccharides and oligopeptides). Another possible reason is that due to different physiology the soil microbes used here may be producing more diverse biomass than the marine microbes used by Lechtenfeld et al. (2015). It is likely that aquatic microbes have a very different degradation strategy than soil microbes. As 10 soils are far less rich in labile molecules, it is possible that soil microbes have evolved to produce much higher fluxes of ROS to degrade the more recalcitrant soil organic matter into consumable substrates, which can also explain the larger dissimilarity in bio-produced organic molecules after the in-15 cubations of pyDOM.
An important observation using the H/C versus MW plots (Fig. S5) was that the bio-produced compounds after incubation of pyDOM were of various MW. Thus, it is likely that the microbial biomass produced during the incubation is rad-20 ically coupled with ambient pyDOM molecules or their biochemical remnants. Radical coupling has recently been proposed as an important process in marine DOM cycling (Hach et al., 2020). In that study, when isotopically 13 C-labeled organisms were incubated with oceanic surface waters, mi-25 crobially produced compounds were quickly coupled to the ambient marine DOM molecules. This "recombination" process occurred within hours of the production of microbial exudates, followed by the observation of a highly diversified DOM pool. This process is likely driven by radical coupling 30 reactions, and such pathways have also been observed in incubations in the presence of sunlight (Sun et al., 2017). Another possible explanation is that chemically reactive species, such as quinones, reacted with microbially produced compounds or the NH + 4 nutrient via nucleophile-driven reactions 35 (such as the Michael addition; McKee et al., 2014) to produce highly diverse pools of molecules after each incubation. Our results are also compared to previous work by , where a ligninaceous sample was treated with three different ROS: hydroxyl radical ( q OH), singlet 40 oxygen ( 1 O 2 ) and superoxide (O − 2 q TS4 ). Each different radical degraded a specific pool of ligninaceous compounds, which showed that different ROS can degrade a variety of types of organic matter. However, there was a significant overlap observed between the three pools of molecules that 45 were degraded  indicating that degradation pathways solely based on ROS attacks are still ordered. Thus, because ROS on their own do not produce completely diversified molecular pools, the combination of the two possible pathways (consumption and ROS degradation) 50 must have occurred to produce the great variability in the bio-produced microbial biomass observed in our study.
Collectively, our results indicate that pyDOM can be both directly consumed by biota and secondarily degraded by ROS-driven processes. These pathways could not be ex-55 plored at a mechanistic level in the current study, and we suggest that future studies focus on employing more specialized analytical techniques (e.g., genetic sequencing, ROS quantification) for deconvoluting the complexity of the biodegradation of pyDOM. The present study provides a detailed evaluation of the compounds that microbes degrade and produce in samples mimicking pyDOM in hydrologically dynamic systems such as 65 riverine and groundwater systems. This work brings new knowledge regarding the properties and degradability of py-DOM and challenges the conventional idea that pyDOM is stable towards biotic degradation. Several studies have already shown that pyrogenic substances have soluble DOM 70 components (Hockaday et al., 2007;Mukherjee and Zimmerman, 2013;Wagner et al., 2017;Bostick et al., 2018) and that more soluble components are produced with environmental aging (Abiven et al., 2011;Ascough et al., 2011;Roebuck et al., 2017;Quan et al., 2020). The experiments presented here, 75 in parallel with Bostick et al. (2021), show that a large portion of pyDOM can be remineralized (biodegraded to CO 2 ) without priming (Guenet et al., 2010;Bianchi, 2011), which indicates that pyrogenic molecules may be far less resistant to biodegradation than previously presumed.

80
The involvement of pyDOM within the global carbon cycle is complex and, for many biogeochemical processes, poorly understood. There is a growing body of literature showing that significant amounts of ConAC are solubilized and exported to the global ocean Jaffé 85 et al., 2013;Wang et al., 2016;Marques et al., 2017;Jones et al., 2020) suggesting constant global leaching of pyDOM in riverine systems from pyOM in soils. The global riverine flux of pyDOM is estimated using the recently reported global flux of ConAC (18 Tg C yr −1 ) and scaled using a factor of 90 7.5 as proposed by Bostick et al. (2018) to be 135 Tg C yr −1 , a value that is much lower than the estimated annual pyDOM production and seepage flux of 1440 TgC yr −1 (Bostick et al., 2018). In addition to the implied 91 % loss of carbon during riverine export, a recent study also reported that the stable 95 carbon isotopic signature (δ 13 C) of oceanic ConAC is not terrestrial but rather marine-like (Wagner et al., 2019). This suggests that either all riverine-exported ConAC are being mineralized before they reach the global ocean or ConAC are chemically altered significantly to change their δ 13 C isotopic 100 signature (Jones et al., 2020). Microbial and photochemical processes have been found to transform DOM with characteristic terrestrial DOM composition (compounds with lower H/C and higher O/C ratios) into compounds having characteristics of marine-derived DOM (compounds with higher 105 H/C, lower O/C ratios; Rossel et al., 2013). Thus, pyDOM may simply be losing its diagnostic molecular and isotopic terrestrial-like fingerprints during riverine export due to a va-riety of degradative post-production processes, and the observed molecular transformations in our study are likely one of them.
The cycling of organic matter in the environment has always been an enigma, and there has been a long-standing 5 effort to explain the fate of terrestrial DOM (including py-DOM) in the global ocean . In a previous paper evaluating the photochemical transformation of pyDOM , we suggested that biotic consumption of photodegradation products of pyDOM ("small aliphatic compounds") could result in the formation of marine-like DOM. This hypothesis was tested by comparing our incubation products (the bio-produced formulas) to FT-ICR-MS formulas of an estuarine transect of the Elizabeth River, VA (Sleighter and Hatcher, 2008) and another 15 10 oceanic DOM samples (reported in Sect. S5). A significant number of common formulas was observed in these comparisons (193-308 common formulas, 8 %-18 % overlap) confirming the hypothesis that bio-incubation of py-DOM can produce marine-like DOM. The observed common 20 formulas were not condensed and 81-192 of them (42 %-70 % overlap) were of molecular composition attributed to carboxyl-rich alicyclic molecules (CRAM) per the definition of Hertkorn et al. (2006). These results indicate that biotic incubations of pyDOM (regardless of photo-irradiation) can 25 contribute to some of the molecules observed in oceanic environments. The fact that some of these molecules were observed in both surface and abyssal oceanic DOM indicate that some pyDOM biodegradation products may be sequestered into the deep ocean as refractory DOM. 30 The observed bio-produced labile formulas in our study do not appear to be commonly observed in other environmental samples. This is likely because these labile molecules are part of the fast-cycling, labile DOM pool per Hansell's model (Hansell and Carlson, 2015) and are quickly depleted 35 in the natural environment. This parallels the findings of a recently published study (Hach et al., 2020) observing that microbially produced molecules are extremely labile and are, within hours, broken down and recombined with ambient DOM molecules. The closed laboratory systems in our 40 study may have enabled the observation of these highly labile molecules, whereas in the natural environment they would have been quickly transformed, diluted, or mineralized to inorganic carbon resulting in their removal from analytical detection. The richness in nitrogen and peptide-like character 45 of the bio-produced molecules we observe suggest greater potential lability, and it is likely that the by-products of biotic degradation of pyDOM are readily incorporated into microbial food webs. This is consistent with the idea that terrestrial DOM is either mineralized to CO 2 or incorporated 50 into food webs (Berggren et al., 2010a;Ward et al., 2013;Fasching et al., 2014). It is also consistent with the suggestion that the majority of organic nitrogen in the oceans is derived from microbial peptidoglycans (McCarthy et al., 1997(McCarthy et al., , 1998Simpson et al., 2011) and with observations of nitrogen 55 from peptidoglycans in soil and sedimentary porewater systems (Schulten and Schnitzer, 1998;Hu et al., 2018Hu et al., , 2020. The production of highly variable and diverse bioproduced molecules is likely a contributing factor to the large complexity of natural organic matterCE33 (Hertkorn et 60 al., 2007;Hawkes et al., 2018). Our observed bio-produced molecules likely contribute to the highly variable microbial exometabolomes observed previously (Antón et al., 2013;Watrous et al., 2013;Romano et al., 2014) and stimulate further questions about pyDOM's function and fate within the 65 global carbon and nitrogen cycles. In this study, we have used soil microbes, as the corresponding degradation by-products can be observed in both soil and groundwater and partially in the upstream sections of rivers. Therefore, it would be critical to perform further studies with different microbial con-70 sortia (riverine, estuarine, marine, etc.) to fully understand the biological degradation of pyDOM in different environments. Additionally, the observed evidence for two possible degradative pathways (consumption and ROS degradation) indicates that these pyDOM incubations are extremely 75 complex systems. Future microbiological studies must aim to investigate these pathways further by designing radical quenching experiments (to test for the presence/absence of radical oxygenation pathways) as well as to employ bioanalytical techniques (e.g., genetic sequencing; Nalven et al., 80 2020) for assessing what microbes are responsible for the labilization and diversification of pyDOM.

Conclusions
This study probing the molecular changes occurring after biotic degradation of pyDOM revealed that soil microbes 85 can effectively recycle and transform a significant portion of pyDOM molecules into labile microbial biomass. After the 10 d incubations, it appears that a wide range of pyDOM molecules, both aromatic and aliphatic, were degradedCE34 , forming a highly diverse pool of compounds, including N-90 containing compounds with proteinaceous signatures and a peptidoglycan-like backbone. These observations are consistent with the previous identification of nitrogen from peptidoglycans in soils and oceans. These bio-produced compounds were highly specific for each pyDOM sample which 95 was concluded by observing very few common bio-produced molecular formulas among incubated samples. The observed molecular labilization and diversification have implications for the biogeochemistry of pyDOM as we show that microbial reworking of pyDOM can contribute to the large com-100 plexity and variability of natural organic matter. This study reveals that (1) pyDOM can be a medium for microbial growth and (2) previously considered recalcitrant pyrogenic molecules can be broken down and the carbon and nitrogen therein can be incorporated into microbial food webs. 105 This study suggests that pyDOM is a much more active component in the global carbon and nitrogen cycles and that 18 A. I. Goranov et al.: Labilization and diversification of pyrogenic dissolved organic matter by microbes some non-condensed pyDOM degradation products have an oceanic fate. Therefore, future studies need to further evaluate the biodegradability of pyDOM with microbial consortia of different environments as well as in the context of wetted soils, groundwater processes, cycling within the riverine and marine water columns, and other aspects of the global carbon and nitrogen cycles. Remarks from the language copy-editor CE1 We cannot make changes that go beyond the obvious correction of typos, etc., or that may affect the meaning of what is said without first obtaining the handling editor's approval. This also applies to title changes. Therefore, please give an explanation of why the title needs to be changed, which will then be forwarded to the editor. Other instances in which editor approval is required have been marked with comments in the text. Thanks.

CE2
Please note that a large number of changes could not be implemented as they conflicted with our standards or required further approval from the handling editor as outlined in the comment above. Furthermore, we do not usually accept stylistic changes at this stage of the production process, and so most of the changes that were deemed stylistic have not been made. The main reason for this is that any change always carries the risk of introducing new errors or typos and we wish to avoid a corrigendum at a later stage. In some instances, it was not possible for me to tell whether a change was stylistic or significantly affected the content and was therefore necessary but required editorial approval -such cases will also be sent to the editor for approval. In order to allow you to see quickly which copy-editing (i.e. language) changes have been implemented and which have not, Viola Zierenberg will include an edited version of the pdf file that you sent with your requested changes. All comments that have not been implemented have been deleted from this file. The comments that remain refer to the changes that have now been made. Please double-check that these changes have been implemented correctly. The following comments explain some of the changes that were not implemented, and further explanations on other points are also found as comments in the text (an outline of our house standards can also be found at https://www.biogeosciences. net/submission.html#english): (1) hyphens in place of en dashes -please note that when the dash indicates a relationship or correlation, an en dash rather than a hyphen is used (standard English punctuation); (2) plural verb with units of measure -a singular verb should be used here (see, for example, the ACS Style Guide); (3) indentation of paragraphs -no indentation is used following a heading as is standard practice, but if there are any inconsistencies that prompted your query, please mark them specifically in the pdf; (4) italicization of taxonomic divisions higher than genus -please note that we do not italicize these though we do capitalize them (a standard we also apply to Bacteria and Archaea), and the terms you highlighted appear to refer to families; (5) numbers changed from written-out form to numerals -we use numerals in mathematical contexts, but otherwise numbers lower than 10 are written out.

CE3
The requested change is not grammatical. Please provide an alternative.

CE4
Please clarify your comment here. You have requested a dash, but the symbol here is already a dash. How should this be displayed?

CE5
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE6
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE7
The addition of a plural s is standard practice in English and our house standard when an abbreviation is used both in the singular and the plural (as in this paper). However, I have reworded the four instances where this was used in the singular, so that the abbreviation could simply stand for the plural throughout. Please check these instances (I have marked them with comments) and let us know if you are happy with the compromise.

CE8
Wording rearranged to avoid singular use of ConAC.

CE9
Wording rearranged to avoid singular use of ConAC.

CE10
Wording rearranged to avoid singular use of ConAC (incorporating also the other change regarding the slash here).

CE11
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval (any other changes from TOC to DOC in the paper must also await the editor's response on this point). Thanks.

CE12
Wording rearranged to avoid singular use of ConAC.

CE13
"pyDOM and" was highlighted, but there was no comment. Please clarify whether anything needs to be changed here.

CE14
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE15
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE16
Is this a necessary change (i.e. is the statement incorrect)?

CE17
Is this a necessary change (i.e. is the statement incorrect)?

CE18
We only rarely italicise Latin terms used in English. In this case, the term has become fully incorporated into English (as witnessed by its inclusion in the dictionary; cf. https://www.merriam-webster.com/dictionary/aqua%20regia) and we would therefore not italicise it.

CE19
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE20
Two words were highlighted in this caption without any comment on the change requested. Please clarify what needs to be changed. Thanks.

CE21
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE22
Please verify all the slashes used throughout this paper. A slash can be used to mean "or", but if you meant "and" or "and/or" in some or all instances, the slash should be replaced. Please check all cases and indicate whether any instances should be changed.

CE23
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE24
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE25
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE26
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE27
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE29
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE30
Open compound according to the dictionary (https://www.merriam-webster.com/medical/phenol%20oxidase). We generally use dictionaries where possible and other resources, such as wikipedia, for more general information or if the term is too specialized to be contained in the dictionary. This is done to try and maintain consistency across our publications.

CE31
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE32
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE33
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE34
Please give an explanation of why this needs to be changed. We have to ask the handling editor for approval. Thanks.

CE35
Text inserted with minor edits (serial comma removed in keeping with the style in the rest of the paper).

Remarks from the typesetter TS1
Please note that the header of this article is a place-holder. The information will be updated automatically during publication. Similarly, the webpage of the preprint will also be updated automatically as soon as the final revised version is published. Furthermore, please note that according to our house standards we do not add spaces around en dashes unless they are a typographic symbol (separating parts of a sentence). Please note also that we abbreviate units in combination with numbers (days and hours).

TS2
The figures have been adjusted to our standards; this means, for example, adding spaces in front of the percent symbol, deleting (separation) lines between the panels if not needed, and adding opening parentheses to panel labels.

TS3
I have double-checked that these are minus symbols.

TS5
Unfortunately, it is not possible to provide the final DOI of the article before it is published. Please confirm that this DOI will be correct as no further changes can be made once the article has been published or consider using the DOI of the preprint to register the data set (doi:10.5194/bg-2021-23).

TS6
The updated supplement file has been uploaded. Please note that the supplement link will be updated automatically during the publication.

TS7
Please check DOI.

TS8
Not mentioned in the text.