4.2 Role of different biological carriers in the export of TE.
In the following, we consider the 9 elements (P, Cd, Ba, Mo, Cu, Ni, V, Y, Mn) for which Fxs are positive throughout the season (Figure 7). Among these, 7 elements (P, Mo, Cd, Cu, Ni, V, Mn) have known biological functions and can therefore be directly associated with biological carrier phases. Overall, this is confirmed by the seasonal dynamics of their Fxs that presented 1 or 2 maxima corresponding to the cups that collected sinking material during the first (cups# 3, 4, 5) or second (cups# 9, 10) export event. To go a step forward, we took advantage of the detailed description of biological matter export provided by microscopic observations (Rembauville et al. 2015a) in the same cup material. An important aspect was to quantify the carbon content of exported diatoms and all types of fecal pellets. Microscopic observations revealed that diatoms dominated the phytoplankton community (Blain et al. 2020), and that 12 different taxa contributed significantly (>1% of total biomass) to both the surface carbon biomass and carbon export. However, the concentrations of TE with a biological role certainly varied throughout the season in surface waters due to intense uptake and remineralisation as observed for Fe above the Kerguelen plateau (Blain et al. 2008; Bowie et al. 2015). Similarly, TE quota are likely to vary over time in surface diatoms, with consequences on TE composition of the fecal pellets. The absence of data on seasonal changes in TE concentrations in the water column and the large uncertainty of the TE transfer efficiency between phytoplankton and zooplankton led us to make a rather conservative choice of only three biological carriers, vegetative cells, spores and fecal pellets. Additionally, we considered the total particle mass, POC, PON, POCdiat, and used CaCO3as a tracer of calcifying organisms.
We first investigated the role of these different biological carriers using PCA (Figure S3) based on Fxs and the different biological carriers mentioned above. However, this approach did not prove informative on the association of a given TE with a biological carrier, except for V, which was strongly associated with vegetative cells and/or calcifying organisms exported during the second bloom. The strong association of Mn with the first bloom, as revealed by the PCA, is not meaningful, because Fxs of Mn is high only in cup #3 albeit the export of this bloom is collected by cups #4 and #5 as well (Blain et al. 2020). For Ba and Y, the PCA does not provide any clues on their association with a particular biological carrier. Co-linearity between the different biological descriptors may have hampered the emergence of more significant relationships for other elements.
We have therefore analysed the data set using a different statistical tool, the Partial Least Square Regression (PLSR), also referred to as Projection of Latent Structure Regression (Abdi 2010). This method considers a set of predictors (X) and descriptors (Y) and extracts a single set of scores from both simultaneously. The method can be seen as a simultaneous PCA on X and Y which achieves the best relationships between X and Y. The method is efficient even when the variables are possibly correlated and when the number of variables is large compared to the number of observations. This method has been successfully applied to determine the ecological vectors associated with sinking carbon flux (Rembauville et al. 2015), to predict the partitioning of carbon within plankton assemblages based on bio-optical properties (Rembauville et al. 2017) or to link biological diversity and carbon fluxes (Guidi et al. 2016). To apply PLSR we considered the total export flux of the 15 elements (descriptors) and the different biological vectors (predictors) mentioned above, and we considered Ti as an overall predictor of lithogenic material. It is important to note that with this approach the search for relationships between elements and the lithogenic carrier phase does not require the use of an elemental ratio. To summarize the results of the PLSR analysis we present the projections of both descriptors and predictors in a three-dimensional space defined by the three first latent variables (Figure 8) which represent 57.5%, 22.1% and 8.6 % of the covariance, respectively. The three corresponding 2D dimensional projections in the latent vectors space are provided in Figure S4. Three different groups of TE emerge from this analysis.
TEs associated with the lithogenic carrier phase . The PLSR, clearly identifies a group of TEs (Al, Zr, Cr, Fe, Th, Co, Mn and Y) for which the seasonal dynamics are strongly related to the lithogenic carrier phase, represented by Ti. This result is in line with the conclusions of the PCA and TExs analysis (Figure 6). The PLSR provides novel information for the groups Cd, P, Ba and V, Ni, Mo, Cu.
TEs associated with fecal pellets and diatom spores. The export of Cd, P and Ba was strongly associated with POCfp and to a lesser extend to POCsp. For Ba, this result is not surprising considering that particulate Ba is largely found as authigenic mineral barite in the ocean (Dehairs et al. 1980), formed by precipitation from dissolved Ba in low oxygen environments. Such anoxic microenvironments are typically found in fecal pellets (Alldredge and Cohen 1987; Ploug 2001) or in aggregates like marine snow which in our study contained large quantities of spores (Blain et al. 2020). Strong correlations of Cd and P export fluxes have already been observed with sediment traps deployed in the upper water column (> 1500m) whereas this relationship vanished at greater depth (Ho et al. 2011; Conte et al. 2019). In the present study, the export of Cd was mainly driven by spores during the first bloom and by fecal pellets throughout the season, while vegetative cells and calcifying organisms present during the second bloom played a minor role for Cd export. Cd, but also Co, can substitute for Zn in the carbonic anhydrase (CA) enzyme, Cd-CA and Zn-CA, respectively (Morel et al. 2020). This has been demonstrated for diatoms under low Zn conditions (Lane and Morel 2000). Cd-CA is present in Thalassiosira antarctica , Chaetoceros dichaeta ,Proboscia alata and Proboscia inermis (Morel et al. 2020), species that were well represented in our sediment traps (Blain et al. 2020). Interestingly T. antarctica and C. dichatea are small and spore forming diatoms which dominated during the first bloom, while the genus Proboscia contains large diatoms exported as vegetative cells that thrived during the second bloom. Cd utilisation by different diatoms in surface waters could explain the seasonal variation of Cd export in the sediment traps. Cd can also be coincidentally taken up by the divalent transporter under Fe-limited conditions (Lane et al. 2008; Horner et al. 2013). At the beginning of the season, the reservoir of Zn and Fe was large above the Kerguelen plateau (Wang et al. 2019), but the rapid development of the massive bloom of small diatoms could lead to a rapid decrease in Zn to levels at which the substitution of Zn by Cd in CA occurred and/or Cd being taken up by the divalent transporter. No strong signal of particulate Cd was associated with the second bloom suggesting that the substitution of Zn by Cd in CA or divalent transport uptake are not dominant processes at the end of the productive season, either due to increased Zn or Fe concentrations provided by remineralisation after the first bloom or due to lower requirements of large diatom cells which do not need Cd for CA activities. Although calcifying organisms including coccolithophorids, present during the second bloom, have high Cd requirements (Ho et al. 2003; Sunda 2012), their contribution was likely hidden behind the large fluxes associated with fecal pellets.
TEs associated with lithogenic and biological carrier phases.V, Mo, Cu and Ni export fluxes are both driven by lithogenic and biological carrier phases. This is a consequence of both their significant contribution to Kerguelen basalt composition (Table 2) and their biological role in microorganisms. Using a similar approach to that used for Cd, we examine the seasonal dynamics of the export of these four metals by first summarizing a few recent insights on their biological role for microorganisms relevant for our study. We then discuss how these observations can provide clues to understand the seasonal dynamics of their export.
The main non-lithogenic V export event coincided with the flux of large vegetative diatoms and calcifying organisms after the second bloom (Figure 6). Due to the similar seasonal patterns of these biological carrier phases, it is not possible, based on PLSR, to make a clear preferential association with either of them. The current knowledge on the biological role of V is mainly related to diatoms, thus our discussion on the temporal changes of V export focuses on this phytoplankton group. V is a cofactor of haloperoxidase enzymes (VHPO) that produce organo-halogens (Moore et al. 1996; Murphy et al. 2000; Hill and Manley 2009). Haloperoxidase activity by diatoms could alter the quorum sensing of prokaryotes and therefore protect diatoms against algicidal prokaryotes (Amin et al. 2012). In contrast to the seasonal dynamics of all other elements, the export of V associated to the biological fraction was higher during the second than during the first bloom (Figure 7). Seasonal observations of diatom and prokaryotic communities in the surface layer revealed compositional changes and strong associations (positive and negative) between diatom species and prokaryotic taxa (Liu et al. 2020). Positive associations could result from interactions based on the exchange of metabolites between diatoms and prokaryotes for resource acquisition, but negative associations are more difficult to interpret. The seasonal dynamics of non-lithogenic particulate V, if related to VHPO activity, could suggest that some diatoms efficiently reduce the growth of targeted prokaryotic taxa with algicidal activity in the phycosphere.
The prevalence of non-lithogenic particulate V during the second bloom could be related to seasonal changes of the bioavailability of Fe. Haloperoxidase can contain Fe-heme as prosthetic group instead of V. Fe-heme containing enzymes could dominate the haloperoxidase activity of diatoms when the bioavailable Fe stock is high such as at the beginning of the season. However, as biological uptake during the first bloom consumed a large part of the bioavailable Fe, haloperoxidase activity of diatoms dominating during the second bloom may have switched to VHPO, which requires the uptake of vanadate, an anion that is always present at non-limiting concentrations in seawater. V can also be found in nitrogenase (nif ) involved in the fixation of dinitrogen (N2) where it substitutes Mo. A recent study illustrated that Mo/Fe containing nif genes are overexpressed by prokaryotic communities on marine particles (Debeljak et al. 2021). Therefore, the association of V or Mo with vegetative diatoms could partly be explained by N2 fixing prokaryotes attached to particles and their downward transport could be a biological carrier phase for Mo and V.
The dominant biological carrier phase for Cu was different to that of V and Mo. Cu export was mainly related to diatom spores and to a lesser extend to fecal pellets whereas no clear association with vegetative cells and CaCO3 was observed (Figure 8 and Figure S4). Cu is a co-factor of a large number of oxidative enzymes involved in different metabolic pathways including Fe acquisition (Maldonado and Price 2001) and nitrogen cycling (Kuypers et al. 2018). Another noticeable feature of these Cu proteins is that most of them are located outside eukaryotic cells or in the periplasm of prokaryotes (Silva and Williams 2001). A possible consequence can be that Cu enzymes are prone to rapid degradation and release of Cu following cell death. Fecal pellets or spores could provide a protected environment during export, which could explain our observations.
The biological carrier phases for Ni were mainly diatoms (spores or vegetative cells) and fecal pellets had a minor role. Among the many biological pathways, Ni is involved in the assimilation of urea (Oliveira and Antia 1984) and is also the cofactor of an enzyme of the superoxide dismutase (SOD) family which can substitute for Fe-superoxide dismutase in low Fe environments (Dupont et al. 2010; Cuvelier et al. 2010). These requirements for Ni likely lead to high Ni quota of diatoms relative to other phytoplankton groups (Twining et al. 2012). It was, however, also noted that 50% of Ni contained in diatoms is associated with the frustule with an unknown function. These Ni dependent enzymes suggest that diatom spores, vegetative cells and fecal pellets that contained mainly diatoms are all potential biological vectors of Ni export. If true, the lack of a marked difference between the first and the second bloom dominated by spores and vegetative cells, respectively, is surprising. A larger contribution of diatoms to Ni export would be expected during the second bloom for two reasons. First, the assimilation of urea is likely only noticeable when the switch from NO3- to NH4+ uptake has occurred, thus after the first bloom. Second, since Fe bioavailability was lower during the second bloom, Fe-SOD is likely to be replaced by Ni-SOD. We suggest an additional process to significantly contribute to the biological export of Ni. Methanogenic Archaea utilise different enzymes belonging to the hydrogenase, reductase or CO dehydrogenase families where Ni is present as co-factor (Mulrooney and Hausinger 2003). MethanogenicArchaea have been detected in different marine particles like marine snow or fecal pellets (Maarel et al. 1999) where they could thrive within anoxic niches (Alldredge and Cohen 1987; Ploug 2001). A time series of the composition of the particulate matter in the surface layers would certainly provide new data required to decipher between these different hypotheses.