Whole water column distribution and carbon isotopic composition of bulk particulate organic carbon, cholesterol and brassicasterol from the Cape Basin to the northern Weddell Gyre in the Southern Ocean

The combination of concentrations and δ 13 C signatures of Particulate Organic Carbon (POC) and sterols provides a powerful approach to study ecological and environmental changes both in the modern and ancient ocean, but its application has so far been restricted to the surface area. We applied this tool to study the biogeo- 5 chemical changes in the modern ocean water column during the BONUS-GoodHope survey (Feb–Mar 2008) from Cape Basin to the northern part of the Weddell Gyre. Cholesterol and brassicasterol were chosen as ideal biomarkers of the heterotrophic and autotrophic carbon pools, respectively, because of their ubiquitous and relatively refractory nature. 10 We document depth distributions of concentrations (relative to bulk POC) and δ 13 C signatures of cholesterol and brassicasterol from the Cape Basin to the northern Weddell Gyre combined with CO 2 aq. surface concentration variation. While relationships between surface water CO 2 aq. and δ 13 C of bulk POC and biomarkers have been previously established for surface waters, our data show that these remain valid in deeper 15 waters, suggesting that δ 13 C signatures of certain biomarkers could be developed as proxies for surface water CO 2 aq. Our data suggest a key role of zooplankton fecal aggregates in carbon export for this part of the Southern Ocean. We observed a general increase in sterol δ 13 C signatures with depth, which is likely related to a combination of particle size e ﬀ ects, selective feeding on larger cells by Our results suggest that in the upper 500 m the major contributor to the sterol pool is phytoplankton, whereas between 500 and 1000 m it becomes zooplankton. Finally, these two components reach stable proportions below 1000 m depth. These results suggest a key role of faecal aggregates in routing carbon to the deep ocean, as highlighted by Ebersbach et al. for the Australian of the Southern Ocean. However, further data would be of great to enlarge the resulting observations the of BONUS-GoodHope transect. dataset on combined δ 13 C and concentrations measurements of both bulk organic C and speciﬁc sterol markers throughout the water column shows the promising potential of analyzing δ 13 C signatures of individual ma- 20 rine sterols to explore the recent history of plankton and the fate of organic matter in the SO.

chemical changes in the modern ocean water column during the BONUS-GoodHope survey (Feb-Mar 2008) from Cape Basin to the northern part of the Weddell Gyre. Cholesterol and brassicasterol were chosen as ideal biomarkers of the heterotrophic and autotrophic carbon pools, respectively, because of their ubiquitous and relatively refractory nature. 10 We document depth distributions of concentrations (relative to bulk POC) and δ 13 C signatures of cholesterol and brassicasterol from the Cape Basin to the northern Weddell Gyre combined with CO 2 aq. surface concentration variation. While relationships between surface water CO 2 aq. and δ 13 C of bulk POC and biomarkers have been previously established for surface waters, our data show that these remain valid in deeper 15 waters, suggesting that δ 13 C signatures of certain biomarkers could be developed as proxies for surface water CO 2 aq. Our data suggest a key role of zooplankton fecal aggregates in carbon export for this part of the Southern Ocean. We observed a general increase in sterol δ 13 C signatures with depth, which is likely related to a combination of particle size effects, selective feeding on larger cells by zooplankton, and growth 20 rate related effects Additionally, in the southern part of the transect south of the Polar Front (PF), the release of sea-ice algae is hypothesized to influence the isotopic signature of sterols in the open ocean. Overall, combined use of δ 13 C and concentrations measurements of both bulk organic C and specific sterol markers throughout the water column shows the promising potential of analyzing δ 13 C signatures of individual ma-

Introduction
The intensity of organic matter export combined with the efficiency of deep water heterotrophic reprocessing of this material sets the sequestration's efficiency of the oceanic biological carbon pump (Honjo et al., 2008;Boyd and Trull, 2007;Battle et al., 2000). However, it appears difficult to balance the organic C demand by twilight 5 zone heterotrophs (currently defined as 100-1000 m) by the export flux from the upper mixed layer (Burd et al., 2010;Steinberg et al., 2008;Reinthaler et al., 2006;Arístegui et al., 2005). Therefore it is essential to better understand the processes controlling the export and reprocessing of exported OM (Boyd and Trull, 2007). Gaining information about the sources and fate of sinking and suspended biogenic particles is necessary 10 to improve our knowledge on what is happening below the euphotic layer, where the attenuation of the export flux is the strongest. Wakeham et al. (2009) state that one way to improve our understanding about mechanisms controlling interrelated biogeochemical processes involved in particle sinking and decomposition is to determine the chemical compositions of those particles sinking rapidly through the water column as 15 well as of the suspended fine particles with longer residence time.
While several studies have examined the δ 13 C of the bulk particulate organic carbon (POC) in the Southern Ocean (Lourey et al., 2004;Trull and Armand, 2001;Popp et al., 1999;Bentaleb et al., 1998;Dehairs et al., 1997;Rau et al., 1997;Kennedy and Robertson, 1995;François et al., 1993), fewer studies have focused on specific com- 20 pounds. O' Leary et al. (2001); Tolosa et al. (1999) and Popp et al. (1999) report on the factors controlling the carbon isotopic composition of phytoplankton based on the study of δ 13 C POC , δ 13 C sterols and δ 13 C phytol . Both these studies, however, were mainly limited to an investigation of the surface particles, so the distribution of these biomarkers in deeper waters remains unknown. To the best of our knowledge, no information proxy for zooplanktonic herbivory (Grice et al., 1998). Schouten et al. (1998) propose cholesterol as a general biomarker for the eukaryotic marine community, thereby avoiding the complexity inherent to species-specific sterols that have different isotopic offsets relative to phytoplankton δ 13 C. A general biomarker averages out the effects due to variability of biosynthetic pathways, cell size, geometry, and growth rate between individual species. Grice et al. (1998) show via controlled mesocosm experiments that grazers convert different algal precursor sterols into cholesterol without significant isotopic fractionation. Moreover, Chikarahishi (2006) in accordance with Grice et al. (1998) suggest that no substantial carbon isotopic fractionation occurs during either heterotrophic sterol assimilation or de novo synthesis. 15 Schouten et al. (1998) state that the refractory nature and ubiquitous character of cholesterol as an eukaryotic marker should favor the integrity of the δ 13 C cholesterol signatures when particles sink to greater depth, thereby preserving the information acquired in surface waters. Brassicasterol (28∆ 5,22 : 24-Methylcholesta-5,22E-dien-3βol) is reported in a large number of algal classes (Volkman, 2003;1986) and is consid-20 ered as a strict phytosterol, i.e. it cannot be biosynthesized by zooplankton. For these reasons, it is commonly used as an indicator of marine algae and of diatoms in particular in the environment. Recently, Rampen et al. (2010) stressed that when diatoms dominate the phytoplankton community, sterols, and brassicasterol in particular (abundant in the pennate diatoms) provide useful information on the type of diatoms that are 25 present. Therefore, for the Southern Ocean where diatoms are dominant but where other phytoplankton groups contribute to the primary production, we will consider brassicasterol as a strict phytoplankton indicator. Because of their ubiquitous and relative refractory nature (e.g. Volkman, 1986Volkman, , 2003, cholesterol and brassicasterol were thus the Cape Basin to the northern Weddell Gyre. Combined with data on CO 2aq. concentrations, our results furthermore enable to verify earlier surface waters observations about CO 2 substrate concentration and C isotopic composition of bulk POC and biomarkers. They also document to what extent this substrate-dependent C isotopic signature, acquired in the surface is preserved deeper in the water column. This is 10 important since particles may, or may not, become isotopically homogenous at depth as a result of mixing, depending on magnitude of sinking velocity relative to advection. This comes to answering the question whether the system is operating in 1-D (strong surface to deep links) or 3-D (strong mixing and homogenization in the deep ocean) configuration.
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | salt water from the Indian and Pacific oceans. The eastward flowing Antarctic Circumpolar Current (ACC), the South Atlantic Current and NADW meet with the westward flow of Indian waters carried by the Agulhas Current, leading to water mass exchanges through jets, meanders, vortices, and filament interactions. Indeed, the Agulhas Current is the major western boundary current of the Southern Hemisphere (Lutjeharms, and a key component of the global ocean "conveyor" circulation controlling the return flow to the Atlantic Ocean (Gordon, 2003;1986). Unusual dynamics pervade the motion of this warm-water current -as it moves west around the Southern tip of Africa, it is reflected back east by the ACC. Not all waters are captured by this sudden diversion of course -parts of the Agulhas Current leak away into the South Atlantic Ocean 10 (BGH cruise report). However, the ACC fronts ( Fig. 1) represent almost impermeable barriers delimiting zones with relatively constant hydrological and biogeochemical properties, and lower baroclinic transport (Sokolov and Rintoul, 2007). Cross frontal exchanges occur but mainly locally close to sharp topographic features. Water particles spend several years in each zone before to be advected northward by Ekman 15 pumping, and subsequently take part in several winter convective mixings (S. Speich, personal communication, 2009). Amongst the sites studied during BGH, five sites had sufficient station occupation time (about 48 h) to enable whole water column suspended matter sampling with High Volume In Situ Filtration Systems (HVFS) because this operation is time consuming (6 20 to 8 h per deployment). Up to a maximum of 12 HVFS units were deployed covering the entire water column from the surface to the deep ocean. The five sites were selected to cover the major zonal systems framed by fronts (Fig. 1). The complete salinity and temperature sections during BGH are available in Fig. 1 chon et al., this issue), filters were partitioned among the different end-users, using sterile scalpels and a custom-build INOX steel support for the PETEX screens and a plexiglass punch of Ø = 25.3 mm for the QMA filters. These operations were conducted under a laminar flow hood. Whenever possible, both size fractions (>53 µm and 53 1 µm) were sampled for biomarkers analyses. This was in general the case for the 10 upper ocean mixed layer. Below the mixed layer, the large particle fraction (>53 µm) could not be recovered because screens carried too little material. Aliquots dedicated to compound specific isotope analysis (CSIA) were packed in cryotubes and stored at -80 • C till processing in the home-laboratory. Aliquots dedicated to the analysis of δ 13 C POC were dried at 50 • C and stored in Petri dishes at ambient temperature till 15 processing in the home laboratory.

δ 13 C POC : sample preparation and analysis
POC concentrations and δ 13 C POC were analyzed via elemental analyzer -isotope ratio mass spectrometer (EA-IRMS). Prior to this, inorganic carbon (carbonates) was removed by exposing the filters to concentrated HCl vapor inside a closed-glass con-20 tainer during 4 h (Lorrain et al., 2003). After drying at 50 • C the samples were packed in silver cups and analyzed with a Thermo Flash 112 elemental analyzer configured for C analysis and coupled on-line via a Con-Flo III interface to a Thermo-Finnigan Delta V IRMS. Acetanilide and IAEA-CH-6 reference materials were used for calibrating concentrations and isotopic composition, respectively. The bulk POC dataset for Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 2.3 δ 13 C sterols : sample preparation and analysis Samples were processed following the method described in Boschker (2004). Briefly, total lipids were extracted using a modified method from Bligh and Dyer (1959) with chloroform / methanol / Milli-Q water (v:v, 15:15:6 mL) mixture. Usually, 70 to 80 % of the chloroform phase containing the total lipid fraction was recovered, implying that 5 true concentrations are 20 to 30 % underestimated (Boschker, 2004).
Since the final standard error on the δ 13 C measurements is estimated to range between 1.1 and max 2.0 ‰, and because this extraction step implies permanent sample shaking to homogenize and optimize recovery of the extract, we assume that isotopic bias due to incomplete recovery is negligible compared to the actual accuracy and 10 precision of the measurement. The total lipid extract was then separated into neutral, glyco-, and polar lipids on silica chromatographic column (0.5 g Kieselgel 60; Merck): (i) the neutral phase was eluted with chloroform (7 mL), (ii) the glycolipidic phase with acetone (7 mL), and (iii) the polar phase with methanol (10 mL). The glycolipidic and the polar phases were stored 15 at -20 • C while the neutral phase was immediately processed for analysis.
Neutral phases were completely dried under gentle inert N 2 flow (to avoid degradation) while the glyco-lipids and the polar phases were stored at -20 • C for further analyses; and a known quantity of squalane used as internal standard (IS, i.e., a stable compound which does not co-elute with natural compounds present in the samples) 20 was added. To estimate compound concentrations in the samples, we use as reference the GC-c-IRMS peak area of squalane since the added IS amount is known.
1675 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | All the TMS-derivatized neutral fractions (+ IS) were analyzed using a Trace GC Ultra coupled to Trace Plus MS (Interscience) for compound identification and a Thermo Finnigan GC equipped with a combustion furnace (CuO/NiO/Pt reactor at 940 coupled to a DeltaPlus XL isotope ratio mass spectrometer (IRMS), for δ 13 C and concentration measurements. Compound identification was achieved by (i) retention time 5 matching between standard mix and sample, (ii) GC-MS mass spectrometry characterization (matching with Goad and Akihisa, 1997). Both gas chromatographs were equipped with similar capillary columns [DB-5-type (J&W Scientific), 30 m × 0.32 mm i.d. with 0.25 µm film thickness]. Identical temperature programming was used for both GC ovens starting from 50 • C (2 min) to 300 • C (15 10 min) with a ramp of 4 • C min −1 . Carrier gas flow (He) was 2 mL min −1 , and injection occurred in splitless mode at 300 • C. Blanks (hexane) and mixtures of standards were regularly measured in between sample analyses (standard bracketing) to check the stability of the systems and absence of possible contaminants. The carbon isotopic composition of squalane (internal standard), treated similarly as 15 the samples (extracted neutral phase), was used to check reproducibility and accuracy of the GC-c-IRMS system. Aliquots of 0.5 mg squalane powder are weighed into tin capsules (IVA, 3.3 × 5 mm) and analyzed via EA-IRMS (Thermo) using IAEA-CH 6 as reference material. A good agreement is observed between the mean δ 13 C squalane obtained via EA-IRMS (δ 13 C squalane = -20.2 ± 0.1 ‰, 1σ, n = 3) and GC-c-IRMS 20 (δ 13 C squalane = -21.2 ± 0.9 ‰, 1σ, n = 69).
Due to the addition of a trimethylsylil group to each individual compound (TMSderivatization), the δ 13 C of cholesterol (27∆ 5 ; cholest-5-en-3β-ol) and brassicasterol (28∆ 5,22 ; 24-Methylcholesta-5,22E-dien-3β-ol) obtained from GC-c-IRMS need to be corrected. Five standards: 5α-cholestane, cholesterin-α24, stigmasterol-α22, β-25 sitosterol, 7-dehydrocholesterol are used for this purpose. Comparison between mean δ 13 C measured by GC-c-IRMS on derivatized (n = 3) and non-derivatized (n = 3) standards' mix is achieved using a functional relationship estimation by maximum likelihood 1676 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | (FREML) (Fig. 2a) via Eq. 1: Where δ 13 C sterol is the measured carbon isotopic signature of non-derivatized standard sterol, n is the number of replicates, δ 13 C TMS−sterol is the measured carbon isotopic signature of derivatized standard sterol and δ 13 C BSTFA is the measured carbon isotopic 5 signature of BSTFA-toluene used to derivatize sterols. Since no certified reference materials for compound-specific stable isotope measurements exist, we furthermore compare mean δ 13 C measured by GC-c-IRMS and EA-IRMS of the same non TMS-derivatized compounds (n = 3) (Fig. 2B) and include it in the δ 13 C sterol correction following a FREML equation as described above (equa-10 tion 1 applied for non-derivatized standard sterols analyzed either via EA-IRMS or via GC-c-IRMS).
Final δ 13 C values for each compound are thereby corrected for isotopic deviation from TMS-derivatization, as well as possible isotopic deviation from GC-c-IRMS compared to EA-IRMS. The uncertainty on final δ 13 C of cholesterol (cho) and brassicast-15 erol (bra) (±1σ; Tables 2 and 3) is calculated by propagating standard deviations from triplicate measurements (i.e. 1σ) and correction from derivatization.

Contents of cholesterol, brassicasterol and POC-bulk
Results for POC (large and small particles), brassicasterol and cholesterol concentra- 20 tions in the small (1 53 µm) and large (>53 µm) particle size fractions and are listed in Tables 1, 2 and 3, respectively. Brassicasterol and cholesterol contents are of similar magnitude. Cholesterol was usually slightly less abundant than brassicasterol, but was detectable over the entire water column, while brassicasterol was undetectable 1677 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | (i.e. concentrations are below the GC-c-IRMS/GC-MS detection limit) in deeper water at the most northward stations S1 and S2. Depth profiles of large particle POC (LP-POC) are shown relative to total POC (T-POC = LP-POC + SP-POC, where SP-POC is small particle POC) to highlight specific features of the profiles (Fig. 3, upper panel). To that purpose we also plotted the pa-5 rameters vs. density rather than vs. depth. Depth profiles of small particles cholesterol (SP-cholesterol) and brassicasterol (SP-brassicasterol) are shown relative to SP-POC (Fig. 3, lower panel) following the same rationale.

General features
Relative contents of LP-POC and SP-sterols changed considerably in surface waters 10 along the studied transect. Overall, the contributions of LP-POC ranged from 0.8 to 27.0 % of total POC, with highest values at S3 (27.0 %) in the Polar Frontal Zone (PFZ), and S5 (25.8 %) in the northern Weddell Gyre. LP-POC values often displayed a maximum in mesopelagic to deep waters. This maximum was clearly visible at S1 (750 m), S3 (450 m) but less prominent at S2 (250 m) and absent at S4 (Fig. 3, upper panel). At 15 S5, an LP-POC maximum was present at 1500 m, below the mesopelagic layer. With the exception of station S4, a systematic increase of LP-POC over T-POC ratio was apparent near the seafloor (nepheloid layer), ranging between 2.7 % (station S1) and 12.3 % (station S2). The large error bars (up to 100 %) in Fig. 3 reflected the large variability between close-by sampled depths (see also Tables 1, 2) reflecting both het-20 erogeneity of the system (as an example, close-by depths were often sampled with an interval of 24h) and variability due to analytical method.
Profiles of relative SP-cholesterol and SP-brassicasterol contents (Fig. 3, lower panel) were also characterized by mesopelagic maxima, except at S1 where brassicasterol below 750 m was under detection limit and at S4 where no mesopelagic 25 maximum was present. The depths of mesopelagic maxima for cholesterol and brassicasterol were rather similar between sites, and ranged between circa 200 and 800 m ( Fig. 3 lower panel). The highest brassicasterol contents were observed at S1 and S5 1678 Introduction in good accordance with the presence of the diatom pigment fucoxanthin in surface water (J. Ras, personal communication, 2009). Below the mesopelagic maximum the relative contents of cholesterol and brassicasterol decreased toward the seafloor (S1, S2, S3; Fig. 3 lower panel), or remained rather unchanged (S4, S5). In contrast to relative contents of LP-POC, no increase of brassicasterol and cholesterol contents was 5 observed near the seafloor.

Site specific features
The surface waters at station S1 showed the highest (relative) SP-brassicasterol content and also the highest concentration (65.3 ng L −1 ) of the whole dataset ( Fig. 3 and Table 2). This is in accordance with the observed high concentrations of Chl-a 10 and phaeopigments (Le Moigne et al., 2012). The profiles of relative SP-cholesterol and LP-POC contents at S1 were quite similar, with pronounced surface and broad mesopelagic maxima: the latter reached circa 100 % of the surface content for cholesterol, and circa 40 % for LP-POC. Station S2, which was also located in the area of high Chl-a and phaeopigment 15 concentrations, north of the Sub-Antarctic Front (SAF) (Le Moigne et al., this issue), showed LP-POC contents relatively similar than at S1. Here surface water brassicasterol reached its second highest value (46.8 ng L −1 ; Table 2) Table 1) of the entire cruise. A second LP-POC maximum was present at 1500 m (Fig. 3). The relative contents of brassicasterol and cholesterol showed a sharp mesopelagic maximum ranging between 300 to 500 m. Brassicasterol showed a second smaller maximum at 1500 m, coinciding with the LP-POC maximum.

15
At all stations the brassicasterol/cholesterol ratio for the small particle size fraction was the highest in the upper 100 m (Table 2). In the mesopelagic and deep ocean this ratio decreased systematically to reach a value close to 1. This condition reflected the relative variation of autotrophic organic material (brassicasterol) vs. heterotrophic material (cholesterol) contribution to the export flux of organic material with increasing 20 depth. For large particles the brassicasterol to cholesterol ratio stayed close to 1 with very surface waters having slightly larger ratios than underlaying waters (Table 3). Such low values reflected larger contributions of hetetrotrophs vs. autotrophs in this large size fraction of particles compared to the smaller sized particles.

25
The vertical profiles for δ 13 C LP−POC , δ 13 C SP−POC , δ 13 C SP−cholesterol and δ 13 C SP−brassicasterol are shown in Figure 4 (all dataset in Tables 1 and 2). Note that for S1 and S2 SP-brassicasterol data are limited to the upper water column, since 1680 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | concentrations fell below detection limit deeper in the deeper water column. The range of δ 13 C variation through the water column can reach up to 10 ‰ (see for example the δ 13 C SP−cholesterol profile at station S2; Fig. 4). At stations S1 and S2 δ 13 C profiles of sterols and bulk POC do cross over at depth, while at stations S3 to S5 these are clearly separated from each other. 5 At S1 and to a lesser extent at S2, δ 13 C LP−POC values decreased below 2750 m and converge toward the δ 13 C SP−POC values ( Table 2, Fig. 4). At these depths the water mass consisted of diluted North Atlantic Deep Water (NADW) from the eastern route (S. Speich, personal communication, 2009). This could be an indication of NADW transporting highly degraded organic matter constituted by isotopically light refractory 10 lipidic material since it has been shown that proteins and carbohydrates (higher δ 13 C isotopic signals than lipids, see Galimov, 2006) are more labile compared to lipids and are then removed preferentially from the particulate organic matter (POM) (Degens, 1969). Also, intense bacterial degradation of the more labile carbohydrates and proteins could explain this strong decrease in δ 13 C-POC observed for large particles at 15 S1 and S2. Stations S1 and S2 are defined by mixed phytoplanktonic community composed of dinoflagellates, chromophytes, nanoflagellates and cyanobacteria, and are defined as systems supported to a large degree by regenerated production (Joubert et al., this issue). Taking the time lag due to sinking speed into account, observed variations of δ 13 C SP−POC and δ 13 C LP−POC along the S1 and S2 depth profiles could indeed 20 be linked to high degradation processes occurring in the surface water. At S2 all δ 13 C values in the upper 600 m decreased with depth, whereas below 600 m they increased (note that SP-brassicasterol was not detected below 600 m). Below 1500 m LP-δ 13 C POC decreased again, as observed at S1 (see above). At S3 below the surface waters δ 13 C LP−POC and δ 13 C SP−cholesterol showed a gradual increase 25 with depth, while δ 13 C SP−POC and δ 13 C SP−brassicasterol remained rather unchanged over the entire water column (Fig. 4) Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | δ 13 C SP−cholesterol increased rather uniformly with depth while δ 13 C SP−POC barely changed with depth. δ 13 C SP−brassicasterol strongly increased from the surface water to 1000 m. At S5 all components showed a gradual δ 13 C increase with depth (Fig. 4).

Discussion
Since this study presents the first complete depth profiles of δ 13 C data on cholesterol 5 and brassicasterol in the Southern Ocean from Cape Basin to the northern Weddell Gyre, we focus the discussion on the observed variation below the surface layers, with the aim of using these proxies to provide insight into the fate of organic matter sinking out of the surface water where the photosynthetic signal is acquired.

Variation of brassicasterol and cholesterol content with depth
The increase of SP-sterol contents relative to SP-POC and (Fig. 3, lower panel) in the mesopelagic layer (100 to 1000 m) overlaps with the domain of excess 234 Th activity and excess particulate Barium during the BONUS-GoodHope expedition, discussed by Planchon et al. (2012). In that paper it is reported that excess 234 Th and particulate Barium reflect the processes of particle break-up and organic matter remineralisation.

15
The observation that in some cases the contribution of large particles to total POC (Fig. 3, upper panel) also increases in mesopelagic waters may reflect local zooplanktonic activity, possibly feeding on the vertical particle flux thereby synthesizing cholesterol. Our present data do confirm the significance of mesopelagic waters as a region where the flux of material exported from the surface undergoes major changes.
20 Figure 5a shows the depth variation of brassicasterol over cholesterol ratios (bra:cho) with depth (see also Furthermore, the lowest value of bra:cho (0.5 at station S2, 608 m; 0.4 at station S4, 749 m; Table 2) are observed at intermediate depths, between 500 and 1000 m. As already highlighted in the Results section, the general decrease of bra:cho ratios reflects the relative decrease of autotrophic organic material (brassicasterol -phytoplankton) vs. heterotrophic material (cholesterol -zooplankton) with increasing depth and its 5 stabilization around 1 in the deep water. These observations lead to the hypothesis that the role of zooplankton in the carbon export composition increases with depth compared to the surface water and shows a maximum role between 500 and 1000 m in the mesopelagic layer. Deeper, the bra:cho ≈ 1 is probably due to an equilibrium reached between the 2 components and their most refractory fraction below the twilight zone where most of the attenuation of the carbon export flux occurs.
To evaluate the importance of such processes at different depth layers we propose in Fig. 6 (panel B is a zoom of panel A) the linear regressions obtained from the correlation between bra and cho concentrations, taking into account that the entire dataset can be split in two groups (group A from surface to 500 m, group B from 500 m to the 15 deep ocean). We obtain a general image for the BONUS-GoodHope transect covering 5 complete water column sampling stations located in five different Southern Ocean regions (see Sect. 2.1.). Regression slopes are 0.5 ± 0.1 (p-value < 0.05) and 1.5 ± 0.2 (p-value < 0.05) for the upper 500 m and the water column below 500m, respectively.
Our results suggest that in the upper 500 m the major contributor to the sterol pool 20 is phytoplankton, whereas between 500 and 1000 m it becomes zooplankton. Finally, these two components reach stable proportions below 1000 m depth. These results suggest a key role of faecal aggregates in routing carbon to the deep ocean, as highlighted by Ebersbach et al. (2011)

Variation of δ 13 C signatures with depth
We observed that in general the δ 13 C SP−sterols but also δ 13 C LP−POC (S1, S2 excepted; see above Sect. part 3.2. and Fig. 4) increase with depth, whereas δ 13 C SP−POC remains stable throughout the water column profile. Figure 5b shows depth variations of calculated values of ε cholesterol and ε brassicasterol for the complete dataset. This ε is 5 based on a simple calculation (ε sterol (‰) = δ 13 C SP−sterol − δ 13 C SP−POC ) allowing us to estimate the apparent offset between the 13 C isotopic signal of SP-POC and studied sterols. Indeed ε brassicasterol ranges between -13.6 ‰ and -4.1 ‰ in the upper 500 m and then stabilizes around -7 ‰ below 500 m. On the other hand, ε cholesterol shows a pronounced increase from the surface water (ε cholesterol ranges between -12.3 ‰ and 10 -5.2 ‰) to the deep ocean where it reaches a value of ∼ +5‰. The stabilization of ε brassicasterol around -7 ‰ below 500 m is in accordance with previous laboratory experiments estimating a ε sterol of -7 ‰ (Bidigare et al., 1997;Schouten et al., 1998; see also Popp et al., 1999 for an application to Southern ocean surface water data analysis). However, previous studies were performed for laboratory experiments on growth 15 rate of various phytoplankton groups or surface water samples, while the present observation concerns samples below the euphotic layer where no more biosynthesis of brassicasterol is possible since it is only synthesized by phytoplankton. However, we previously observed that δ 13 C SP−POC is stable through the entire water column, and thus hypothesize that the stable ε brassicasterol of -7 ‰ below 500m indicates that this 20 sterol is biosynthesized in the surface water and not below. On the other hand, the increase of ε cholesterol with depth could reflect several factors which will be described below.
Although it can not be excluded that the observed increase of δ 13 C-cholesterol is partly due to the Suess effect reflecting the invasion of the upper ocean by isotopically 25 light anthropogenic CO 2 reflected till the secondary producers, this would imply a large age difference between deep and surface ocean suspended matter (the former much older than the latter; Suess, 1980  (McDonnell and Buesseler, 2010) it is evident that the time lag between the bloom events that generate the deep ocean particles and those that generate the particles in the upper layers is much inferior to the period covering the bulk of the anthropogenic CO 2 emissions (weeks, months vs. years). Therefore, it is probable that the observed lower δ 13 C values for surface ocean SP-sterols compared 5 to deep ocean values are due to factors other than the Suess effect. Below, we discuss several possible mechanisms leading to isotopically heavier organic carbon and also sterols.

Growth rate related effect
Fry and Wainright (1991) report that fast-growing diatoms are 13 C-enriched compared 10 to more slowly growing cells. Highest growth rates occur during the early season when there is no nutrient limitation, while later in the season growth rates decrease due to the onset of nutrient limitation conditions (see e.g. Arrigo, 2002), the latter leading to more pronounced isotope fractionation and thus more 13 C-depleted phytoplankton (e.g., Popp et al., 1999). During the end of the growing season, we would thus ex-15 pect relatively 13 C-depleted phytoplankton in the upper water column, while the 13 Cenriched cells from the start of the growing season would have reached the deeper water column. Such mechanism could explain the observed enrichment in δ 13 C cholesterol with depth in the case of BONUS-GoodHope, which took place during the austral summer, i.e., the end of the productive season.

Particle size related effects
Pancost et al. (1997) report a particle size-dependency for δ 13 C brassicasterol , with a 13 C enrichment in suspended particles >20 µm relative to the <20 µm fraction. This suggests that surface-area to volume ratios control the fractionation of carbon isotopes by diatoms, as confirmed by Popp et al. (1998). Therefore, a relatively increased con-25 tribution of larger phytoplankton to the export flux could explain the observed δ 13 C 1685 enrichment with depth. Selective feeding of zooplankton on larger cells and active transport of this material to deeper layers during migration (Steinberg et al., 2008) might cause the translocation of isotopically heavier material to deeper waters, while leaving isotopically lighter material -associated with smaller particles-in the surface layers.  , 1996;Gibson et al., 1999;Villinski et al., 2000). Melting of seaice with release of sea-ice phytoplankton occurs during the growth season, so these isotopically heavy particles, if sinking out of the surface waters, can be expected to be found deeper in the water column. For the most southern stations (S4, S5), which 15 are influenced by the Seasonal Ice Zone (SIZ) in winter, sinking of sea-ice algae could contribute to 13 C enrichment of organic C at depth.

Other processes potentially affecting variations of δ 13 C POC
Several further processes different from those listed above could affect the isotopic composition of deep ocean POC. These processes include: (i) preferential degradation 20 of the more labile carbohydrate and protein material which is 13 C-enriched compared to lipids (Degens, 1969;Galimov, 2006); (ii) 13 C enrichment of sinking particles due to selective grazing on the larger 13 C-enriched cells (see discussion above) followed by fecal pellet production; (iii) loss of 13 C-depleted carbon associated with methane production by prokaryotes within zooplankton digestive tracts (Freeman, 2001; Hayes, Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | to transfer of organic carbon to higher trophic levels (∼1-2 ‰ enrichment per increasing trophic level due to respiratory effects; DeNiro and Epstein, 1978;Fry, 1988;Fry and Sherr, 1984). We note, however, that δ 13 C SP−POC did not show an increase with depth. We hypothesize that in this case there may be a balance between a δ 13 C SP−POC decrease 5 due to preferential loss of 13 C-enriched, more labile carbon phases (carbohydrates, proteins; see above) and a δ 13 C SP−POC increase due to the trophic level succession effect, inducing an increase ∼1-2 ‰ keeping δ 13 C SP−POC rather unchanged through the deep water column. Alternatively, it might simply be that small particles are indeed representative of organic matter biosynthesized in the surface water. This would imply 10 then a fast transfer mechanism for these particles. Recent work by McDonnell and Buesseler (2010) did reveal that sinking velocity is largest not only for the largest particles they studied (up to 1 mm), but also for the small sized particles (<50 µm). Clearly, any realistic scenario for 13 C enrichment of biomarkers with depth needs also to comply with the observation that POC of the <53 µm size class does not show a significant 15 13 C enrichment with depth.
In summary, the observed isotopic increase for SP-cholesterol with depth would need a cell size or growth rate related impact on isotopic fractionation and a mechanism separating isotopically heavy from light cells. This mechanism could be (1) a separation in time between sinking of isotopically heavier cells (early season) and light cells (late 20 season) and/or (2) selective zooplankton grazing on larger isotopically heavier cells and their transfer to the deep sea via fecal pellets. The first hypothesis implies an age difference between deep and surface particles, while the second one does not. However, the absence of a significant δ 13 C variation of SP-POC with depth requires that the above processes affecting δ 13 C SP−cholesterol are somehow neutralized for SP-POC 25 (via a balance between preferential degradation of isotopically heavy compounds such as carbohydrates and proteins and isotopic enrichment due to trophic translocation of organic matter). As an additional hypothesis, we also stress a possible effect of high pressure on cholesterol biosynthesis occurring below the surface water: cholesterol is a component of cell membrane and the idea is that 13 C-13 C covalent bond is stronger than 12 C-12 C covalent bond. Such specificity could be of great interest to resist high pressure effect and would imply an increase of δ 13 C cholesterol with depth. 5 Our observations stress the interest of studying both δ 13 C POC and δ 13 C sterol in sinking and suspended particles since it allows gaining information on the fate of organic matter in the whole water column via cross-comparison.  (1982), that the carbon isotopic composition of phytoplankton is primarily determined by (i) the isotopic composition of the source of inorganic carbon; (ii) the isotopic fractionation during transport into and out of the cell; (iii) isotopic discrimination during carboxylation and the degree of leakage of intracellular inorganic carbon. When focus is on the isotopic composition of the source (marine algae take up either CO 2 aq. or bicarbonate HCO − 3 , or both), François et al. (1993)  1985) and in case phytoplankton incorporate bicarbonate it would tend to increase δ 13 C POC . In contrast, δ 13 C CO 2 aq. systematically decreases from ∼-7.5 ‰ in warm water to ∼ -10 ‰ in cold Southern Ocean surface water. François et al. (1993) conclude that changes in the isotopic composition of CO 2 aq. can contribute significantly to the observed latitudinal trend in δ 13 C POC (a circa 2.5 ‰ decrease contribution). For
As observed for δ 13 C POC , the decrease in δ 13 C LP and SP−cholesterol , and δ 13 C LP and SP−brassicasterol in the surface water coincides with an increase in CO 2 aq. Introduction  (Table 4). However, the average (LP and SP) decrease reaches -11.1 ± 1.9 ‰ and -14.7 ± 0.9 ‰ for cholesterol and brassicasterol, respectively (from Table 4), which is larger than for small and large particle POC. This observation is in agreement with Popp et al. (1999) who report that the individual sterol isotopic compositions in surface waters (WOCE SR3 transect, 145 • E, between 45 • 55 S to 5

65
• 24 S) generally decrease southward and exceed the variations in bulk δ 13 C POC .
Our data reveal for large particles that the difference in δ 13 C composition between LP-POC and LP-cholesterol is between -6.0 and -10.3 ‰ (Table 4), while the difference between LP-POC and LP-brassicasterol δ 13 C ranges between -6.8 and -13.9 ‰. For small particles the difference between POC and cholesterol ranges between -6.2 to 10 -9.3 ‰ and the difference between POC and brassicasterol between -5.8 to -10.9 ‰. Such offsets appear to increase southward. Tolosa et al. (1999) estimate the δ 13 C isotopic difference between bulk organic matter and cholesterol to be 5 ‰, and stress that such difference is typically observed between whole plant material and extractable lipids (Hayes, 1993). While for the northern part of the BGH transect our data are close 15 to the observations of Tolosa et al. (1999), in general we observe a larger δ 13 C difference between POC and cholesterol, and the discrepancy is enhanced southward. This enhanced discrepancy could be related to larger cells contribution (diatoms).

Mesopelagic and deep ocean
For the δ 13 C LP−POC regressions against CO 2aq. the slopes are quite variable (Fig. 7a, 20 surface and deep ocean and from the north to the south part of the Southern Ocean (Fig. 7d, Table 5). It is not unlikely (though not confirmed here) that this reflects an enhanced contribution of sea-ice diatoms south of the BGH transect (S4 and S5, see above Sect. 4.2.3.), although this cannot be unambiguously demonstrated by our data. These two proxies appear thus useful to trace fate of organic matter in the whole water 10 column.

Conclusions
Although the BONUS-GoodHope transect crossed several frontal structures and specific biogeochemical zones, our study showed several general patterns in brassicasterol and cholesterol concentrations, bra:cho ratios, and δ 13 C signatures of POC, 15 cholesterol and brassicasterol along depth and latitudinal gradients. First, the bra:cho ratio, which reflects the relative importance of autotrophs vs. heterotrophs, shows a general trend relevant for each of the five studied stations: in the upper 500 m depth the major component of organic material is phytoplankton (bra:cho ratios highly variable and >1) , whereas between 500 and 1000 m depth the major component becomes 20 zooplankton (bra:cho ratios reaching values <1 in the mesopelagic layer). Finally, these two components reach stable ratio below 1000 m depth (bra:cho ∼1). Our observations support a key role of fecal aggregates in the fate of organic matter and export to deep ocean. Second, depth variation of ε brassicasterol and ε cholesterol shows again general trends 25 valuable for the five stations. This indeed highlights (i) that brassicasterol is only biosynthesized in the surface water; thus showing a stable value close -7 ‰ below 500 m depth in accordance with Schouten et al. (1998) and Bidigare et al. (1997) increase of ε cholesterol with depth which is mainly defined to be related to growth rate related effect and/or particle size related effect and selective feeding by zooplankton. A role of sea-ice algae release south of the transect is also proposed to act on latitudinal and depth variations of δ 13 C brassicasterol and δ 13 C cholesterol . As an additional hypothesis, we also suggest that a high pressure effect on cholesterol biosynthesis below the sur-5 face water might induce an increase of δ 13 C cholesterol in deeper waters.
Third, the relationship between CO 2aq. concentration and δ 13 C SP−sterols is generally maintained throughout the deep ocean. This observation was not expected a priori, since circulation of deep waters carrying suspended matter formed in different regions may possibly dilute the isotopic signatures resulting from biological activity in local 10 surface waters. It thus seems that export of organic matter and fate of organic compounds in the deep sea closely fit to a 1D scheme (surface to deep ocean). In case further studies confirm this observation, the fact that regressions of δ 13 C cholesterol and δ 13 C SP−POC versus CO 2aq. are quite conservative with depth highlights the potential of these proxies for surface water CO 2 aq. concentration. Future field work is necessary 15 to confirm variability and trends observed during this study, and to adequately constrain the application of such this potential proxy.
While not conclusive, this first dataset on combined δ 13 C and concentrations measurements of both bulk organic C and specific sterol markers throughout the water column shows the promising potential of analyzing δ 13 C signatures of individual ma-20 rine sterols to explore the recent history of plankton and the fate of organic matter in the SO.        bottom water (crosses). Averages (± 1σ) are given in Table 5 Table 4) and δ 13 C POC (‰) in large (>53 µm) (A) and small (1 << 53 µm) (B) particles in surface water (0-100 m, closed diamonds), mesopelagic layer (100-1000 m, open square), deep ocean (>1000 m, closed triangles), and bottom waters (crosses). Lower panel: relationship between surface water CO2 aq. concentration and small particles (1 << 53 µm) δ 13 C cholesterol (C) and δ 13 C brassicasterol (D) in surface water (0-100 m, closed diamonds), mesopelagic layer (100-1000 m, open square), deep ocean (>1000 m, closed triangles), and bottom water (crosses). Averages (±1σ) are given in Table 5. Standard errors are not shown here for sake of clarity.