Biotic and abiotic degradation of the sea ice diatom biomarker IP25 and selected algal sterols in near-surface Arctic sediments

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.


Introduction
Over the past decade, the Arctic sea ice diatom biomarker IP 25 (Ice Proxy with 25 carbons atoms; Belt et al., 2007) has emerged as a useful proxy for the past occurrence of seasonal (spring) sea ice when detected in Arctic marine sediments (for a review see Belt and Müller, 2013). Consistent with its origin (i.e., sea iceassociated or sympagic diatoms; Brown et al., 2014), IP 25 is a common component of surface sediments across the Arctic (Müller et al., 2011;Stoynova et al., 2013;Xiao et al., 2013Xiao et al., , 2015Navarro-Rodriguez et al., 2013;Ribeiro et al., 2017;Köseoglu et al., 2018), while its variability in downcore abundance is generally believed to reflect temporal changes to spring sea ice cover, especially when its concentration profile is considered alongside those of other biomarkers indicative of open-water or ice-edge conditions (e.g., Müller et al., 2009Müller et al., , 2011Belt et al., 2015), through a combined IP 25 -phytoplankton biomarker index (PIP 25 ) (Müller et al., 2011), or a multivariate biomarker approach (Köseoglu et al., 2018). To date, however, the majority of IP 25 -based studies have focused either on surface sediment analysis or on long-term (multi-centennial or longer) records. Thus, surface sediment analyses have addressed aspects of proxy calibration, generally by comparison of IP 25 and other biomarker content with satellitebased measurements of sea ice conditions (Müller et al., 2011;Navarro-Rodriguez et al., 2013;Stoynova et al., 2013;Xiao et al., 2013Xiao et al., , 2015, while temporal studies have concentrated mainly on the reconstruction of sea ice conditions on a multi-centennial scale during the Holocene (e.g., Vare et al., 2009;Belt et al., 2010;Müller et al., 2012;Berben et al., 2014Berben et al., , 2017 Stein et al., 2017), recent glacial/interglacial intervals (Müller and Stein, 2014;Hoff et al., 2016), the Mid-Pleistocene Transition (Detlef et al., 2018), and even longer timeframes extending back to the Pliocene/Pleistocene boundary and the late Miocene Knies et al., 2014;Stein et al., 2016).
One key attribute of IP 25 as a sea ice proxy is its apparent relative stability in sediments. Indeed, the identification of IP 25 in sediments several million years old (Knies et al., 2014;Stein et al., 2016) has been attributed, in part, to such stability, and is supported by laboratory-based investigations, where it has been shown to be significantly less reactive towards degradation process such as photo-oxidation and autoxidation, at least compared to other common phytoplanktonic lipids (Rontani et al., 2011(Rontani et al., , 2014. As such, sedimentary signals have been interpreted as reflecting climatic (sea ice) conditions rather than diagenetic artefacts, although the possibility of some diagenetic over-printing of the environmental signal has been noted (e.g., Belt and Müller, 2013;Polyak et al., 2016). In contrast, temporal investigations covering recent decades or centuries are less common, although some studies from North Iceland (Massé et al., 2008;Andrews et al., 2009), East Greenland (Alonso- García et al., 2013;Kölling et al., 2017), the Barents Sea Köseoglu et al., 2018), northern Baffin Bay (Cormier et al., 2016) and the Chukchi-Alaskan margin (Polyak et al., 2016) have been reported. Such studies are somewhat different from those carried out on surface sediments (typically 0-1 cm) or longer timeframe investigations generally conducted on material from gravity/piston cores since, in some cases, at least, they likely result from analysis of material that spans the oxic/anoxic (redox) sediment boundary. However, such boundary layers are not generally identified (reported), even though they are likely found in the upper few centimetres of box cores or multi-cores, which reflect accumulation over decades or recent centuries for many Arctic Shelf regions (e.g., Stein and Fahl, 2000;Darby et al., 2006;Mudie et al., 2006;Maiti et al., 2010;Vare et al., 2010). On the other hand, in the central Arctic Ocean, such a layer may reflect substantially longer-term accumulation due to much lower sedimentation rates (e.g., Stein et al., 1994a,b).
The rate and extent of degradation of sedimentary organic compounds is strongly dependent on the molecular structure of the substrate, protective effects offered by association of organic matter with particle matrices, and the length of time accumulating particles are exposed to molecular oxygen in sedimentary pore waters (Henrich, 1991;Hartnett et al., 1998). The main degradative processes in the oxic layer of sediments are aerobic biodegradation and autoxidation. Numerous organisms, including bacteria, fungi and micro-and macrofauna, are responsible for the aerobic biodegradation of organic carbon in sediments (Fenchel et al., 1998) and almost all of these organisms have the enzymatic capacity to perform a total mineralization of numerous organic substrates (Kristensen, 2000). Although autoxidation of organic matter involving spontaneous free radical reaction of organic compounds with O 2 has been rather under-considered in the marine realm, it is now known that autoxidative processes can act very intensively on vascular plant debris in Arctic sediments . This high autoxidation efficiency likely reflects the enhanced photooxidation of senescent vascular plants in the region (thus yielding high amounts of hydroperoxides), together with high lipoxygenase activity (a potential source of radicals; Fuchs and Spiteller, 2014). Indeed, the latter mechanism has recently been observed in sinking particles dominated by ice algae  and in particles discharged from the Mackenzie River (Galeron et al., 2018).
The principal aim of the current study, therefore, was to investigate whether we could provide evidence for oxidative degradation processes acting on IP 25 in near-surface Arctic sediments and thus, potentially, on any resultant palaeo sea ice reconstructions. To achieve this, we first carried out laboratory-based oxidation of purified IP 25 and carried out product identification using high resolution mass spectral analysis. Since IP 25 was shown previously to be relatively resistant to oxidation (Rontani et al., 2014), more powerful oxidizing conditions were used. We then investigated the occurrence of the same oxidation products in sediment samples taken from box cores retrieved from three regions of the Canadian Arctic. To complement the IP 25 -based findings, we also measured the ratios of two common algal sterolsepi-brassicasterol and 24-methylenecholesterol -to provide further evidence of different oxidative pathways under oxic and anoxic conditions. Geochemical analysis of the box cores revealed variable redox boundary depths, which provided further context for interpreting the biomarker data.

Sediment and sea ice algal sampling
Investigations of in situ degradation processes were performed on sediment material and sea ice algal aggregates. Sediment material was obtained from three locations within the Canadian Arctic Archipelago (CAA) on board the CCGS Amundsen in 2005Belt et al., 2010. Sampling locations correspond to Barrow Strait (STN 4), Viscount Melville Sound (STN 308) and the western Amundsen Gulf (STN 408) (Fig. 1). In each case, box cores were collected, sectioned on board, with subsamples (1 cm resolution) then being freeze-dried before storage between -20°C and +4°C prior to analysis. Regular monitoring of IP 25 concentration in these sediments stored under such conditions (since their collection) has not revealed any significant degradation (i.e. < 10%; . Previous reports of sedimentation rates from the study area (e.g., 0.15 cm/yr for the Barrow Strait (STN 4) core ) and preliminary additional 210 Pb data (S. Schmidt, personal communication) suggest that box cores (ca. 20 cm) from the region typically represent decades to centuries of accumulation. A sample of floating sea ice algal aggregates was obtained from Resolute Passage (western Barrow Strait) in 2012 as described by Brown et al. (2014).
Redox boundary layers in each of the box cores were identified using the change (reduction) in Mn content as described previously Brown, 2011 and References cited therein). Using this approach, redox boundaries were identified at ca. 2 cm in the box core from Barrow Strait (STN 4) and at ca. 11 cm and ca. 8 cm in box cores from Viscount Melville Sound (STN 308) and the western Amundsen Gulf (STN 408), respectively (L. Vare, personal communication).

IP 25 isolation
A sample of IP 25 (ca. 99%) was obtained by extraction of a multikg quantity of sediment from Barrow Strait in the Canadian Arctic (STN 4; Fig. 1) and purification by a combination of open column chromatography (SiO 2 ; hexane) and Ag + HPLC as described previously in detail by Belt et al. (2012).
The structures of all IP 25 oxidation products are shown in Appendix A. Due to the very low amounts of IP 25 available, compounds 1, 2, 6 and 7 could not be produced in sufficient amounts to permit quantification, although comparison of their mass spectra and retention times with compounds detected in sediments confirmed their identification.

Induction of autoxidation in solvent
Autoxidation experiments were performed under an atmosphere of air in 15 ml screw-cap flasks containing IP 25 (10 mg), tert-butyl hydroperoxide (200 ml of a 6.0 M solution in decane), di-tert-butyl nitroxide (1.2 mg) and hexane (2 ml). After stirring, the flask was incubated in the dark at 65°C. A relatively high temperature was selected in order to accelerate the autoxidation reactions. Aliquots (200 ml) were withdrawn from the reaction mixture after incubation for different times. Each sub-sample was evaporated to dryness under a stream of nitrogen and analyzed by gas chromatography-electron ionization quadrupole time of flight mass spectrometry (GC-QTOFMS) after NaBH 4 reduction (see Section 2.5) and derivatization (see Section 2.8) for identification of hydroxylated oxidation products.

Reduction of oxidation products
Hydroperoxides resulting from IP 25 oxidation were reduced to the corresponding alcohols by reaction with excess NaBH 4 in diethyl ether:methanol (4:1, v:v, 10 mg/mg of residue) at room temperature (1h). After reduction, a saturated solution of NH 4 Cl (10 mL) was added cautiously to remove any unreacted NaBH 4 . The pH was then adjusted to 1 with dilute HCl (2N) and the mixture shaken and extracted with hexane:chloroform (5 ml, 4:1, v:v; Â3). The combined extracts were dried over anhydrous Na 2 SO 4 , filtered and evaporated to dryness under a stream of nitrogen.

Aerobic biodegradation of phytoplankton sterols
Aerobic biodegradation of phytoplankton cells was performed using the upper layer (0-1 cm) of Arctic sediments collected in July 2016 from Davis Strait (70°29 0 55.56 00 N, 59°31 0 30.24 00 W) during the GreenEdge cruise on board the CCGS Amundsen as bacterial inoculum. Enrichment cultures were incubated in the dark in 250 ml Erlenmeyer flasks containing 50 ml portions of an enrichment medium consisting of LB medium (20 ml) and phytoplankton suspension (10 ml) (10 mg dry weight) as carbon source. Samples were maintained at 2°C (a temperature close to that of Arctic waters) and agitated using a reciprocal shaker for different times. The amounts of 24-methylenecholesterol and epi-brassicasterol in the sediment inoculum were negligible relative to those in the phytoplankton suspension. These phytoplankton cells (mainly composed of diatoms) were collected in Commonwealth Bay (East Antarctica, 66°56 0 S; 142°27 0 E) during the IPEV-COCA2012 cruise in January 2012 as described previously (Rontani et al., 2014). After incubation, phytoplankton material was recovered by filtration on GF/F filters and saponified as described in Section 2.7.

Sediment and sea ice algal treatment
Sediments from box cores (i.e., STN 4, 308, 408) or sea ice algae (19.3 mg dry weight) were placed in MeOH (15 ml) and hydroperoxides were reduced to the corresponding alcohols with excess NaBH 4 (70 mg, 30 min at 20°C). Following the reduction step, water (15 ml) and KOH (1.7 g) were added and the mixture saponified by refluxing (2 h). After cooling, the contents of the flask were acidified with HCl to pH 1 and extracted three times with dichloromethane (DCM) (30 ml). The combined DCM extracts were dried over anhydrous Na 2 SO 4 , filtered and concentrated to give a total lipid extract (TLE). Since IP 25 oxidation product content was quite low relative to other lipids, accurate quantification required further separation of the TLE using column chromatography (silica; Kieselgel 60, 8 cm Â 0.5 cm i.d.). IP 25 was obtained by elution with hexane (10 ml) and its oxidation products by subsequent elution with DCM (10 ml). Additional elution with MeOH (10 ml) was carried out to recover the more polar lipid compounds. Relative IP 25 content was determined using the method of Vare et al. (2009) and Belt et al. (2010) and some uncalibrated data (STN 308) were presented previously by Brown (2011). Here, all previous GC-MS data were re-analyzed and converted to absolute concentrations using instrumental response factors derived from solutions of known IP 25 concentration (Belt et al., 2012). Biomarker data were further normalised to total organic carbon (TOC) to accommodate possible changes in burial efficiency. TOC data were obtained following removal of inorganic carbonate from sediment material according to the method of Berben et al. (2017).

Derivatization of hydroxyl-containing products
In order to analyze for hydroxylated products (i.e. alcohols and carboxylic acids), DCM-and MeOH-eluted fractions were derivatized by dissolving them in 300 ml pyridine/bis-(trimethylsilyl)tri fluoroacetamide (BSTFA; Supelco; 2:1, v:v) and silylated (50°C, 1 h). After evaporation to dryness under a stream of N 2 , the derivatized residue was re-dissolved in 100 ll BSTFA (to avoid desilylation of fatty acids), together with an amount of co-solvent (ethyl acetate) dependent on the mass of the TLE, and then analyzed using GC-QTOFMS.

GC-QTOFMS analyses
Accurate mass spectra were obtained with an Agilent 7890B/7200 GC-QTOFMS System (Agilent Technologies, Parc Technopolis -ZA Courtaboeuf, Les Ulis, France). A cross-linked 5% phenyl-methylpolysiloxane (Macherey Nagel; Optima 5-MS Accent) column (30 m Â 0.25 mm, 0.25 lm film thickness) was employed. Analysis was performed with an injector operating in pulsed splitless mode at 280°C and the oven temperature programmed from 70°C to 130°C at 20°C/min, then to 250°C at 5°C/min and then to 300°C at 3°C/min. The carrier gas (He) was maintained at 0.69 Â 10 5 Pa until the end of the temperature program. Instrument temperatures were 300°C for transfer line and 230°C for the ion source. Accurate mass spectra were recorded across the range m/z 50-700 at 4 GHz with nitrogen as collision gas (1.5 ml/min). The QTOFMS instrument provided a typical resolution ranging from 8009 to 12,252 from m/z 68.9955 to 501.9706. Perfluorotributylamine (PFTBA) was utilized for daily MS calibration. Structural assignments were based on interpretation of accurate mass spectral fragmentations and confirmed by comparison of retention times and mass spectra of oxidation products with those of authentic compounds, when available.

IP 25 concentrations (Supplementary
Next, we aimed to identify IP 25 autoxidation products in the DCM fractions of the TLEs of different sediments by comparison of accurate mass fragmentations and retention times with the oxidation products characterised during the thermal incubation reactions. Using this approach, we detected compounds 1, 3, 4 and 5 in sediments from Barrow Strait (STN 4), which also contained the highest concentrations of IP 25 (Fig. 4). The combined relative abundance of these compounds (estimated on the basis of similar TOFMS responses to that of IP 25 ) reached 8.8% of the amount of IP 25 in the 1-2 cm layer and then decreased rapidly to 1.2% in the 3-4 cm horizon. In addition, 2,6,10,14-tetramethyl-7-(3-methylpenten-4-yl)-pentadecan-6-ol (8), which was absent in the incubation experiments, was also identified, albeit tentatively (Fig. 4B). In contrast, since the mass spectrum of the TMS derivative of the saturated tertiary C 25 HBI alcohol (C-7) had already been reported (Robson, 1987), we were able to investigate if the corresponding mono-unsaturated oxidation product was also present; however, no characteristic fragmentation ions corresponding to oxidation at C-7 of IP 25 could be identified. Analysis of extracts by GC-QTOFMS did, however, enable us to detect 3,9,13-trimethyl-6-(1,5-dimethyl hexyl)-tetradecan-1,2-diol (6) in sediments from Barrow Strait (STN 4) and the western Amundsen Gulf (STN 408) (Fig. 5), while traces of 2,8,12-trimethyl-5-(1,5-dimethylhexyl)-tridecanoic acid (7) could be identified in Barrow Strait (STN 4) and Viscount Melville Sound (STN 308) sediments (Fig. 6). These two compounds were formally identified by comparison of their accurate mass spectra (Fig. 7) and retention times with those of standards. On the other hand, we failed to detect compounds 1-8 in floating sea ice algal aggregates from Resolute Passage despite the presence of relatively large amounts of IP 25 within these samples .

Autoxidation of IP 25
According to our product identifications, autoxidation of IP 25 involves hydrogen atom abstraction by peroxyl radicals on the allylic carbon C-22 and the tertiary carbon atoms C-2, C-10 and C-14. Subsequent oxidation of the resulting radicals together with hydrogen abstraction from other substrate molecules leads to the formation of various hydroperoxides (Fig. 8). These labile compounds were reduced to their corresponding alcohols (1-5) during NaBH 4 -reduction and silylated prior to analysis by GC-QTOFMS. The failure to detect any autoxidation product resulting from reaction with either of the tertiary carbons C-6 or C-7 is likely due to increased steric hindrance during hydrogen abstraction by the bulky tert-butylperoxyl radicals employed during the incubation.
Indeed, when comparing our data from laboratory and environmental samples, we note that the relative abundances of IP 25 oxidation products are very different in Arctic sediments (Fig. 4B) compared to those from incubations in solvent (Fig. 4A), likely reflecting the contrasting nature of the peroxyl radicals involved during autoxidation. For example, the bulky tert-butylperoxyl radical pertinent to the laboratory-based incubations probably favours the attack of the less hindered external carbon atoms of IP 25 (i.e. C-2 and C-14), while the unknown (structurally) peroxyl radicals acting in sediment seem to be less sensitive to such steric hindrance. ) showing the distribution of IP 25 oxidation products obtained after incubation in the presence of tert-butyl hydroperoxide and di-tert-butyl nitroxide at 65°C under darkness (A) and present in the 2-3 cm layer of the core sediment from the core from Barrow Strait (STN 4) (B). The peak labelled * represents a compound tentatively identified as 2,6,10,14-tetramethyl-7-(3-methylpenten-4-yl)-pentadecan-6-ol (8) trimethylsilyl derivative.
Finally, although each of 1-5 could be readily identified during the incubation reactions, they were only ever present in low abundances and none accumulated over time. We attribute this to the likely secondary oxidation of primary hydroperoxides to polar and oligomeric compounds (Fig. 8)  J the GC-QTOFMS method employed here. This kind of secondary oxidation was described previously for other HBIs (Rontani et al., 2014).

Degradation of epi-brassicasterol and 24-methylenecholesterol
Due to the different positions of the double bonds in their alkyl chains (see Appendix A), an enhanced autoxidative and photooxidative reactivity of epi-brassicasterol compared to 24-methylenecholesterol would be expected. Indeed, the C-H bond energy for allylic hydrogens is lower for internal double bonds than it is for terminal double bonds (77 kcal/mol vs 85 kcal/mol) (Schaich, 2005), thus making allylic hydrogen abstraction more favourable in epi-brassicasterol. Moreover, on the basis of degradation rates of singlet oxygen ( 1 O 2 ) with terminal and internal double bonds (4.0 Â 10 3 and 7.7 Â 10 3 M À1 s À1 , respectively; Hurst et al., 1985), Type II photosensitized oxidation of epi-brassicasterol should also be favoured compared to 24-methylenecholesterol. However, in natural settings, it was previously reported that autoxidation (Rontani et al., 2014) (Rontani et al., , 2016 processes act more intensively on 24methylenecholesterol than on epi-brassicasterol, at least in mixed phytoplanktonic assemblages. These differences in reactivity can be attributed to the involvement of intra-cellular compartmentalization effects, which may significantly modify the reactivity of lipids towards autoxidative and photooxidative processes according to their location in phytoplanktonic cells (Rontani, 2012). This enhanced reactivity of 24-methylenecholesterol towards autoxidation in phytoplanktonic cells suggests that an increase in the Bra/24-Me ratio may be a good indicator of autoxidation processes in sediments, especially as the main autoxidative products of these two sterols are unspecific and labile 7a/b-hydroperoxysteroids (Christodoulou et al., 2009;Rontani et al., 2009).
In contrast to autoxidation reactions, aerobic microbial degradation of D 5 -sterols involves two processes: side-chain elimination and ring opening (Rostoniec et al., 2009). The degradation is initiated by oxidation of the 3b-hydroxyl moiety and isomerization of the D 5 double bond to the D 4 position (Sojo et al., 1997). Further degradation of the resulting 4-steren-3-one proceeds via hydroxylation at C 26 to initiate side-chain degradation, or oxidation of rings A and B resulting in the cleavage of the ring structure (9,10-secopathway; Philipp, 2011). In the case of cholesterol, the degradation of the 26-hydroxylated alkyl chain may be carried out after oxidation to the corresponding acid by classical sequences of b-oxidation (Rostoniec et al., 2009). In contrast, in the case of epi-brassicasterol and 24-methylenecholesterol, due to the presence of a methyl or methylene group at C-24, the involvement of alternating b-decarboxymethylation (Cantwell et al., 1978) and b-oxidation sequences is needed (Fig. 9). The very close degradation rates of these two sterols observed after incubation of phytoplanktonic cells in the presence of sediment inoculum under oxic conditions (Table 1) may be attributed to the involvement of a 2,3-enoyl-CoA isomerase (Ratledge, 1994). Indeed, these widely distributed enzymes may catalyze the isomerisation of the methylidene double bond to the C24-25 position in the case of 24-methylenecholesterol (Fig. 9), thus permitting the involvement of a similar degradation process of the alkyl side-chain in the case of the two sterols.
Under anoxic conditions, ring cleavage of D 5 -sterols may be mediated by oxygen-independent enzymatic processes (Chiang et al., 2007). In the case of cholesterol, only hydroxylation of the side chain at C-25 has been shown to occur, with the resulting tertiary alcohol not oxidized further (Chiang et al., 2007). For sterols with more substituted or unsaturated side chains, such as sitosterol, fucosterol and isofucosterol, similar degradation rates were observed following incubation of cells of the microalga Nannochloropsis salina in anoxic sediment slurries (Grossi et al., 2001). This suggests that changes to the sterol side chain have little impact on the overall degradation rates under anoxic conditions. As such, in the absence of any reported experimental data, it is reasonable to propose similar anaerobic degradation rates for epibrassicasterol and 24-methylenecholesterol, especially given their common ring structure. Overall, therefore, aerobic and anaerobic bacterial degradation processes should not induce significant changes to the Bra/24-Me ratio in sediments.

Degradation of IP 25 , epi-brassicasterol and 24methylenecholesterol in Arctic sediments
Due to the extremely low rate of autoxidation of IP 25 in solution, even at higher temperature (e.g., 65°C), it was suggested previously that it should be largely unaffected by autoxidative degradation processes in the Arctic, at least in comparison with lipids of similar structure such as other HBIs with greater unsaturation (Rontani et al., 2014). However, here we show that autoxidative degradation of IP 25 may occur under more 'forced' conditions and such processes may also take place in Arctic surface sediments. Indeed, due to recent evidence of strong lipoxygenase activity (a well-known source of radicals; Fuchs and Spiteller, 2014) in bacteria associated with ice algae  and in terrestrial particulate organic matter discharged from Arctic rivers (Galeron et al., 2018), autoxidative degradation reactions can even be dominant in Arctic sediments , despite the low temperatures. The autoxidation of IP 25 in sediments possessing a thick oxic layer, where the contact of ice algal detritus with oxygen may be relatively long, therefore represents a viable degradation pathway of this biomarker in near-surface sediments.
Consistent with this suggestion, the decrease in IP 25 concentration observed in the oxic layer of sediments from Viscount Melville Sound (STN 308) (between 3 and 10 cm) and the western Amundsen Gulf (STN 408) (between 0 and 3 cm) (Fig. 3) may potentially be attributed to the involvement of oxic degradation processes such as aerobic biodegradation (Robson and Rowland, 1988) or autoxidation, and this last suggestion is supported further by the increase of the Bra/24-Me ratio within the same sediments (Fig. 3). In contrast, the strong decrease in Bra/24-Me observed in the bottom of the oxic layer of sediments from the western Amundsen Gulf (STN 408) is potentially due to an input of fresh algal material (with a low Bra/24-Me ratio) during this period. This suggestion is supported by the observation of a 10-fold increase in phytoplanktonic sterol concentration in the 6-7 cm horizon compared to the 4-5 cm layer. Further, Brown (2011) proposed that rapid decreases in sedimentary IP 25 concentration in some other cores from the Canadian Arctic could potentially reflect degradation processes, more generally. In contrast, the more consistent concentration of IP 25 in anoxic sediments from Barrow Strait (STN 4) and the western Amundsen Gulf (STN 408) (Fig. 3) is likely indicative of enhanced resistance to oxidation under such conditions. Unfortunately, we were not able to detect the primary autoxidation products of IP 25 in sediments other than from Barrow Strait (STN 4), likely due to: (i) their further oxidation (as suggested from the incubation reactions), especially in the oxic layers of cores from Viscount Melville Sound (STN 308) and the western Amundsen Gulf (STN 408) and (ii) the detection limits of GC-QTOFMS analyses. However, despite the general resistance of IP 25 towards free radical oxidation, as reported previously (Rontani et al., 2011(Rontani et al., , 2014, the detection of compounds 1, 3, 4 and 5 (Fig. 4)   this HBI alkene can be susceptible to autoxidation in Arctic sediments, an environment where such processes have previously been shown to be enhanced for some other lipids . Further, this vulnerability towards autoxidation may be especially prevalent in cases where sequestered ice algal material experiences long residence times in the oxic layer.
Interestingly, compounds 6 and 7 could be detected in anoxic sediments from Barrow Strait (STN 4) and oxic sediments from Viscount Melville Sound (STN 308) and the western Amundsen Gulf (STN 408) (Figs. 5 and 6). We attribute the formation of such compounds to aerobic or anaerobic bacterial metabolism of IP 25 . In contrast, a mechanism involving autoxidative production (via epoxidation and subsequent hydrolysis; Schaich, 2005) is discarded on the basis of: (i) the detection of only one pair of enantiomers of compound 6 in sediments (Fig. 5) and (ii) the lack of compounds 6 and 7 observed during our in vitro autoxidation experiments. Aerobic bacterial degradation of IP 25 may be initiated either via attack on the double bond or by the same mechanisms associated with n-alkane metabolism (i.e., attack of terminal methyl groups; Morgan and Watkinson, 1994). Oxidation across the double bond in IP 25 can produce diol 6 via the corresponding epoxide 9 (Soltani et al., 2004) (Fig. 10). Previously, it was demonstrated that various pristenes and phytenes (also isoprenoid alkenes) can be rapidly biodegraded by sedimentary bacteria under anaerobic conditions, mainly by hydration reactions (Rontani et al., 2013). Enzymes that catalyze the addition of water to isolated and electron-rich carbon-carbon double bonds are termed hydratases and display a high degree of enantioselectivity (Resch and Hanefeld, 2015). In the case of IP 25 , addition of water to the C23-24 double bond results in the formation of 3,9,13-trimethyl-6-(1,5-dimethylhexyl)-tetradecan-2-ol (10) (Fig. 10), which subsequently oxidises to the corresponding ketone (11). Mechanisms involving hydration of the enol forms of the keto group have been proposed for the anaerobic metabolism of isoprenoid ketones by denitrifiers (Rontani et al., 1999(Rontani et al., , 2013. Hydration of the enol form under kinetic control of the ketone 11 affords the diol 6 ( Fig. 10). This diol may be subsequently cleaved to form 2,8,12-trimethyl-5-(1,5-dimethylhexyl)-tridecanal (12), which may then be fully metabolized via 2,8,12-trimethyl-5-(1,5-dimethylhexyl)tridecanoic acid (7) by alternating b-oxidation and b-decarboxymethylation sequences (Cantwell et al., 1978;Rontani and Volkman, 2003). These interesting results suggest that IP 25 may be also affected by bacterial degradation processes in Arctic sediments, although the extent to which this occurs remains to be determined.

Implications for palaeo sea ice reconstruction
The identification of some degradation pathways of IP 25 in some Arctic marine sediments raises potentially important questions regarding the use of this biomarker as a reliable proxy measure of past sea ice. However, the failure to investigate the occurrence of any of the degradation products described herein in previous studies, prevents a comprehensive evaluation of the importance of IP 25 degradation from being made at this stage. In the meantime, analysis of an extensive set of surface sediments from different Arctic regions has revealed excellent agreement between IP 25 content and known sea ice cover (e.g., Müller et al., 2011;Stoynova et al., 2013;Navarro-Rodriguez et al., 2013;Xiao et al., 2013Xiao et al., , 2015Belt et al., 2015;Köseoglu et al., 2018;Ribeiro et al., 2017), while IP 25 data obtained from several short core records (typically covering recent decades to centuries) show generally good agreement with known sea ice conditions derived either from historical records or satellite data (Alonso-García et al., 2013;Weckström et al., 2013;Cormier et al., 2016), including examples where IP 25 concentration increases with depth (e.g., Massé et al., 2008;Andrews et al., 2009;Vare et al., 2010;. However, in a recent study from the Chukchi-Alaskan margin, a decline in IP 25 abundance in near-surface sediments was suggested to indicate a combined influence of diagenesis and long-range sediment transport (Polyak et al., 2016). Further, the previously reported surface sediment datasets (and their relationship to known sea ice cover) might need re-examination in light of the evidence described herein for at least partial IP 25 degradation in some near-surface sediments.
Interestingly, although there is a clear decline in IP 25 concentration with depth in the box core from Viscount Melville Sound (STN 308) (Fig. 3), a similarly continuous negative trend was not apparent in the cores from either Barrow Strait (STN 4) or the western Amundsen Gulf (STN 408) (Fig. 3), despite the detection of IP 25 oxidation products in both cases (Fig. 5D). This suggests that climatic influences likely exceeded those from degradation, although the possible impact of bioturbation, a feature in some near-surface sediments, cannot be totally ruled out at this stage. However, preliminary 210 Pb data suggest that bioturbation is negligible in cores from Barrow Strait (STN4) and Viscount Melville Sound (STN308), and confined to the (at most) upper 2 cm in the core from the western Amundsen Gulf (STN408) (S. Schmidt, personal communication).
For longer records (i.e. those beyond recent centuries), a common feature in many IP 25 -based sea ice reconstructions has been a reduction in IP 25 concentration over time, especially during the Holocene (e.g., Vare et al., 2009;Belt et al., 2010;Fahl and Stein, 2012;Müller et al., 2012;Hörner et al., 2016;Kölling et al., 2017;Stein et al., 2017). Such changes have generally been interpreted as reflecting an increase in sea ice extent or duration from the warm early Holocene through neoglacial conditions towards present, an interpretation generally supported with other paleoclimatic proxy data. The often higher IP 25 concentrations observed in older sections of the same (or related) records, covering the Younger Dryas stadial (ca. 12.9-11.5 kyr BP) Müller and Stein, 2014;Belt et al., 2015;Méheust et al., 2015;Jennings et al., 2017) and the Last Glacial Maximum (LGM; e.g., Müller and Stein, 2014;Hoff et al., 2016) provide further evidence of substantial climatic overprinting within biomarker profiles.
Resolving the relative contributions of climatic influence and diagenetic alteration on downcore IP 25 (and other biomarker) distributions is likely to remain a challenge from an analytical perspective, however, not least because, on the basis of our new results described here, the oxidation products of IP 25 are unlikely to accumulate in sufficient amounts to enable their quantification (or even detection), especially since IP 25 content itself is often quite low in Arctic marine sediments. On the other hand, the measurement of certain biomarker ratios such as Bra/24-Me may prove useful for assessing such degradation processes, especially when used alongside IP 25 concentration profiles; however, the potential for changes in environmental conditions to also influence such ratios should also be considered. Further, the measurement of redox boundary layers in upper sections of sediment cores might also provide additional insights into the nature of different degradation processes.
Finally, it is interesting to note that we were not able to detect any IP 25 oxidation products in our sample of sea ice algae, which supports conclusions from previous studies that it is largely resistant to abiotic alteration in the host matrix (Rontani et al., 2014) and also in the water column soon after ice melt Rontani et al., 2016).

Conclusions
This study represents the first attempt to evaluate, via oxidative product identification, the possible fate of IP 25 in Arctic sediments. Laboratory-based autoxidation of the Arctic sea ice diatom biomarker IP 25 results in the formation of a series of oxidation products that could be characterised using high resolution GC-MS methods. Some of the same oxidation products could also be identified in sediment material from the Canadian Arctic although their accumulation was very low, likely due to further oxidation. The detection of bacterial metabolites of IP 25 showed that this HBI alkene may also be affected by aerobic and/or anaerobic degradation processes in sediments. We suggest that complementary evidence for autoxidation and biodegradation processes may potentially be obtained from measurement of certain phytoplankton sterol ratios, although these may also be influenced by changes to the overlying climatic conditions.
Although degradation of IP 25 has, to date, not been considered in detail within IP 25 -based sea ice reconstructions, our initial overview of previous studies suggests that climatic contributions to sedimentary IP 25 distributions likely exceed the impact of sedimentary degradation, at least in the albeit still limited number of case studies thus far reported. On the other hand, oxidative degradation may have a significant impact on IP 25 concentration in some near-surface material, especially in cases where the oxic layer represents relatively long time intervals. In any case, we suggest that such degradation processes should be considered more carefully in future sea ice reconstructions based on IP 25 .

Acknowledgments
Financial support from the Centre National de la Recherche Scientifique (CNRS) and the Aix-Marseille University is gratefully acknowledged. Thanks are due to the FEDER OCEANOMED (N°1 166-39417) for the funding of the apparatus employed. We are grateful to Lindsay Vare for providing us with the redox boundary information used in this study and for carrying out some of the sediment extractions for IP 25 analysis. Patricia Cabedo-Sanz is acknowledged for conducting calibration experiments that enabled us to subsequently quantify IP 25 from archived GC-MS data. We also thank Thomas Brown for fruitful discussions regarding the potential degradation of IP 25 in Arctic sediments. We are grateful to Lindsay Vare, Guillaume Massé, André Rochon and the officers and crew of the CCGS Amundsen for help with obtaining box core sediment material. Finally, we thank the Associate Editor, Kirsten