DFe concentration ([DFe]) in the fluid end-member was 15-fold higher than in the surface end-member. [DFe] followed quite similar dynamics in all fluid-enriched treatments ( Figure 2 ). They increased until day 1, peaking at 3.2-19.5 nmol L^-1, before gradually decreasing until the end of the experiment.

In the following section, the results will be discussed according to two selected periods of the experiment: period 1 (P1) from the start of the experiment to day 2 and period 2 (P2) from day 2 to the end of the experiment.

The exported material from the control, +3.6% and +10.9% fluid treatments showed similar Fe concentrations ([Fe]; ~3.67 mg g^-1; Figure 5A ). [Fe] from other fluid-enriched treatments doubled relative to the control. The highest [Fe] (x4 relative to the control) was observed in the most fluid-enriched treatment. Nevertheless, no significant correlation was found between the exported [Fe] and the percentage of fluid added (R² = 0.24, p = 0.57). High concentration of Cu ([Cu]) were exported in the most fluid-enriched treatments relative to the control (x2.4; Figure 5B ). However, low [Cu] was exported in the +5.5% and +9.1% fluid treatments relative to the large volume of fluid added. Nevertheless, a positive relationship between the [Cu] exported and the percentage of fluid added was observed (R² = 0.6, p = 0.03). POC concentrations ([POC]) exported in the fluid-enriched treatments were higher than the control (x1.4), except for the +3.6% fluid treatment that exported lower [POC] ( Figure 5C ). The highest [POC] (x2 relative to the control) were exported in the least fluid-enriched treatment. No correlation between the [POC] and the percentage of fluid added was found (R² = 0.27, p = 0.52).

In the control minicosm, a small increase of biomass associated with picocyanobacteria, haptophyceae, diatoms and dinoflagellates was observed during P1. This increase lasted for only two days, probably due to nutrient limitation, especially in nitrate (< 0.1 µmol L^-1, Figure S7 ; Bonnet et al., in rev.). The biomass associated with those groups then decreased rapidly until the end of the experiment. Nevertheless, this brief development at the start of the experiment resulted in fairly significant cumulative NCP and POC export into the traps. In contrast, N₂ fixation rates remained stable during this period in the control, and throughout the experiment.

Large differences were observed between the control and fluid-enriched treatments regarding community composition, plankton abundance and biological fluxes. Indeed, at the start of P1, only picocyanobacteria (mainly Synechococcus) developed after the mixing and at much higher abundances than in the control. Yet, the fluids provided a small amount of macronutrients that could have supported the growth of phytoplankton (+0.04-0.28 and +0.003-0.023 µmol L^-1 of NO 3 − and PO 4 3 − added by fluids, respectively, relative to the control; Bonnet et al., in rev.). Furthermore, active diazotrophs also likely provided new N sources to the system. Compared to the control, the large non-cyanobacterial species did not develop during P1 and dinoflagellate mortality (i.e., decreased abundance) was even observed. One possible hypothesis is that the observed patterns were the result of an initial toxicity of some elements supplied by the fluids. Such an effect would have decreased after 4 days of experiment, as species whose growth was stagnant finally started to grow. This implies that the studied waters had likely been detoxified, consistent with the dynamics of thiols at P1, as these compounds have been reported to selectively complex certain toxic trace metals (Grill et al., 1989; Ahner and Morel, 1995; Zenk, 1996). This mechanism will be explored in the following section.

Environmental detoxification may occur through the excretion of metal-binding ligands by communities (e.g., Leal et al., 1999; Morel and Price, 2003; Dupont and Ahner, 2005; Hoffmann et al., 2012). Two types of organic ligands capable of detoxifying the environment were measured during the experiment: FDOM, protein- and humic-like fluorophores, and thiols, a pool of non-fluorescent matter (Chen et al., 2013). In this section, an in-depth exploration of thiol implication in environmental detoxification will be conducted. FDOM fluorophores will be discussed in accordance with their role as strong Cu ligands (Wong et al., 2019).

Thiol production by phytoplankton is triggered by the presence of various trace metals such as Cd, Cu, Hg, Pb and Zn (Grill et al., 1989; Ahner and Morel, 1995; Zenk, 1996). These metals, essential (Cu, Zn) or not (Cd, Hg, Pb), can cause numerous deleterious effects on communities at various concentrations (e.g., Sunda and Huntsman, 1998; Echeveste et al., 2012; Le Faucheur et al., 2014). During our experiment, communities immediately responded to fluid addition through striking thiol production (i.e., up to +823% [TA] relative to control produced during the 12 h after the mixing). Two types of thiol compounds were detected: TA and GSH. TA is actively exuded by phytoplankton (Leal et al., 1999) and precipitates trace metals, rendering them non-bioavailable to communities (Gharabaghi et al., 2012). GSH is the most common thiol ligand in surface waters (Le Gall and Van Den Berg, 1998; Leal et al., 1999; Tang et al., 2000) and is a precursor of phytochelatins (PC; Grill et al., 1989), a polypeptide that forms complexes with trace metals ultimately transported and degraded in the vacuole and/or chloroplasts (Cobbett, 2000; Worms et al., 2006). Thus, the decrease in [GSH] to near-zero levels reflects a high PC production (Romano et al., 2017). Previous studies reported that GSH levels in the presence of high metal concentrations could reach undetectable levels within a few hours (2-4 h; Scheller et al., 1987; Coppellotti, 1989; De Vos et al., 1992; Rijstenbil et al., 1994; Satoh et al., 1999), in agreement with the levels observed in minicosms 12 h after the mixing (< DL). After the probable environmental detoxification, [GSH] accumulated in minicosms. This result was expected as GSH also plays a homeostasis and environmental sensing role in phytoplankton (May et al., 1998; Noctor et al., 1998). Thus, these two thiol compounds had a key role in mitigating the deleterious effects of toxic trace metals added to the environment by fluids ( Figure 6 ). The possible trace metals released by hydrothermal fluids will be explored in this section in accordance with their ability to trigger thiol production (i.e., Cd ≥ Cu > Pb > Hg; Grill et al., 1989; Ahner and Morel, 1995).

The question of which players produce thiols capable of detoxifying the environment is relevant. In the present experiment, picocyanobacteria could be the potential thiol producers (Moffett et al., 1990; Moffett and Brand, 1996; Croot et al., 2000). In fact, the first organisms to grow were picocyanobacteria such as Synechococcus and Prochlorococcus to a lesser extent, consistent with the striking increase of thiols that occurred 12 h after the mixing ( Table 3 ). In line with these observations, Prochlorococcus has been reported to be more sensitive to trace metal addition than Synechococcus (Mann et al., 2002). However, Synechococcus has previously been reported as one of the most sensitive species to trace metal addition due to its large surface-to-volume ratio (Quigg et al., 2006). This was not observed in this experiment, as in several other studies, where Synechococcus was the most resistant species to trace metal addition while eukaryotic species were the most sensitive (Joux-Arab et al., 2000; Stauber and Davies, 2000; Le Jeune et al., 2006; Levy et al., 2007). Numerous studies have demonstrated the existence of Synechococcus ecotypes adapted to toxic trace metal concentrations, either genetically or physiologically (Huckle et al., 1993; Mann et al., 2002; Palenik et al., 2006; Stuart et al., 2009; Stuart et al., 2013). These ecotypes are primarily from environments where trace metal concentrations may be chronically high, such as coastal waters or near hydrothermal vents, and are adapted to a fast-changing environment (Mann et al., 2002; Stuart et al., 2009). They may also be related to their previous exposure to trace metals (Levy et al., 2007). Low abundance of cyanobacteria was measured in the fluid end-member (0.0035 µg L^-1). It is therefore possible that the thiol-producing Synechococcus adapted ecotypes found in fluid-enriched treatments originated from the environment near the shallow hydrothermal source and allowed for environmental detoxification. Molecular analyses would be necessary to confirm such a hypothesis. For all these reasons, it is very likely that Synechococcus, known to produce strong binding ligands, including thiols (Moffett et al., 1990; Moffett and Brand, 1996; Croot et al., 2000), was the major producer of these ligands that were needed to inhibit toxicity of ionic trace metal, allowing other species to grow. Moreover, Brand et al. (1986) demonstrated that Synechococcus growth rates subjected to high concentrations of Cu or Cd were inhibited in the absence of strong chelators.

Based on our experimental data, the largest non-cyanobacterial species (haptophyceae, diatoms and dinoflagellates) were probably not the major thiol producers, especially since eukaryotes have been reported to have more affinity for thiols-Cu complexes, in contrast to prokaryotes (Walsh et al., 2015) and to produce weaker ligands (Croot et al., 2000). Indeed, diatoms and haptophyceae only developed 2 and 4 days after the mixing, respectively, not coinciding with the thiol peaks measured 12 h after mixing. However, it is interesting to note that diatoms – known to be more tolerant to the addition of certain trace metals, such as Cu (Croot et al., 2000) – were the first to develop after environmental detoxification. Dinoflagellate mortality (i.e., decrease in biomass) was observed from the start of the experiment, rendering these species poor candidates for strong ligand production. This result is not surprising as dinoflagellates are known to be very sensitive to Cu, even at low concentrations, and to other trace metals (Anderson and Morel, 1978; Brand et al., 1986; Taylor, 1987; Lage et al., 2001). Encystment strategies have been developed by dinoflagellates to protect themselves from a harmful environment (Anderson and Morel, 1978; Pinto et al., 2003), allowing the persistence of plasma membrane and chlorophyll integrity (Lage et al., 2001). Recovery from trace metal exposure has been observed for numerous dinoflagellates species within 24 to 96 h after exposure (Lage et al., 1994; Okamoto et al., 1999). This excystation process may explain the delayed growth of dinoflagellates (6 days after the mixing) relative to other species.

Abiotic factors such as pH and temperature and biotic factors such as uptake pathways and internal detoxification processes can influence trace metal tolerance. When mixing the two end-members, addition of one (or a cocktail of) trace metal(s) and H₂S-rich, low-pH and oxygen-poor fluid (Tilliette et al., 2022) temporarily altered the surface water chemistry and may have exacerbated the toxic effect of fluid addition. Nevertheless, due to the significant production of thiols by picocyanobacteria, especially by resistant Synechococcus ecotypes, potential toxic elements supplied by hydrothermal fluids have been modified into less harmful forms. This environmental detoxification allowed the larger species to grow and a late positive effect proportional to the fluid addition could be observed at the end of the experiment (see the following section).

As discussed above, the addition of hydrothermal fluids to surface waters first inhibited biological activity until the fluid chemistry was modified by the production of thiols making the mixed water less harmful or even fertilizing ( Figure 6 ). Indeed, growth of non-cyanobacterial species was observed 4 days after the mixing, resulting in enhanced cumulative NCP rates (up to x3.3 relative to the control) at the end of the experiment. Moreover, an increasing fertilizing effect depending on the addition of hydrothermal fluid was observed, as the maximum cumulative NCP values were proportional to the percentage of fluid added ( Figure 4B ). This positive effect was also observed for N₂ fixation, which was up to 2.5-fold higher in fluid-enriched treatments relative to the control. Consequently, an increase of POC export was measured in the fluid-enriched treatments (x1.4 on average and up to x2.2 relative to the control). Such a positive effect on NCP and N₂ fixation is likely due to the introduction of high concentrations of fertilizing elements, such as Fe, along with rather low concentrations of macronutrients, so the system was likely still nitrate-limited. During the TONGA cruise (Guieu and Bonnet, 2019), the same stocks and fluxes studied in the minicosms were measured in-situ along a 6100-km transect through the Tonga volcanic arc. Cruise results show that N₂ fixation rates were the highest above hydrothermal sources and primarily supported by large (> 10 µm) species such as Trichodesmium (Bonnet et al., in rev.). In contrast, in-situ primary production was high a few kilometers from the source, but lower just above it (Bonnet, unpublished), likely reflecting the toxic effect of fluids evidenced during our experiment.

Fe is an essential element for phytoplankton growth and metabolism (Behrenfeld and Milligan, 2013). It is a known fertilizing element for all phytoplankton species, having a direct influence on photosynthesis (Sunda and Huntsman, 1995) and thus an important role in carbon export and sequestration in the ocean interior (Martin, 1990; Pollard et al., 2007). Fe is also often a limiting element for diazotrophs, as in addition to photosynthesis, the N₂ fixation process also requires high Fe amounts (Raven, 1988). During the experiment, hydrothermal fluids supplied high Fe concentrations, both in particulate and dissolved forms ( Figure 4 ). Indeed, a 772- and 16-fold enrichment of particulate ( Table S2 ) and dissolved Fe, respectively, could be observed in the fluid end-member relative to the surface end-member. These high DFe concentrations introduced by the fluid end-member ([DFe] = 15.8 nmol L^-1) were likely responsible for the high N₂ fixation and NCP rates measured in the fluid-enriched treatments. This was confirmed (see Text S6) by the higher Fe uptake rates of picocyanobacteria (mainly Synechococcus and bacteria), that consumed up to 16% of the total minicosm DFe stock over the course of the experiment, relative to non-cyanobacterial species, that consumed up to 8% of the total stock, in agreement with the uptake rates measured by Lory et al. (2022) during the TONGA cruise (~4-11 pmol Fe L^-1 d^-1). Interestingly, consistent with these uptake rates, the DFe dynamics observed during the experiment were primarily driven by precipitation. This process can easily be explained by the presence of high H₂S concentrations in the fluid (Tilliette et al., 2022) and the gradual increase in pH, which was significantly decreased at the start of the experiment through the low pH fluid addition (pH of 6.5; Figure S1 ). Estimated precipitation rates in fluid-enriched treatments (66 ± 7%) were similar to those estimated in-situ above the shallow hydrothermal source (62-86%; Tilliette et al., 2022). As for Cu, only a small proportion of the particulate Fe was recovered in traps (< 10%), probably due to the small particle size that did not allow them to be exported.

The growth of the large non-cyanobacterial species occurred at days 2-4, when phosphate and nitrate concentrations were nearly depleted (Bonnet et al., in rev.; Figure S4 ). Nitrogen was made available to non-diazotrophic phytoplankton through N₂ fixation and the use of rather abundant DOP (~0.2 µmol L^-1 after 4 days, data not shown) as an alternative P-source may have allowed their growth. Indeed, microbial utilization of DOP through alkaline phosphatase activity can be Fe-limited (Browning et al., 2017) and the alleviation of Fe stress leads to DOP loss (likely as a consequence of enhance uptake) under low phosphate availability at the global scale (Fernández-Juárez et al., 2019; Liang et al., 2022). Thus, the introduction of high concentrations of DFe by the fluid end-member mitigated the Fe limitation of surface end-member waters. This DFe fertilization by fluids is of great importance as diazotrophs can exploit these high concentrations to build biomass, as long as phosphate concentrations are sufficient or if these organisms also use DOP (Dyhrman et al., 2006; Dyhrman and Haley, 2006). The fluid addition also allowed an important enrichment in Al (x27.5 of particulate Al relative to control; Table S2 ). Interestingly, Al has been reported to promote Fe uptake (Shaked et al., 2005; Mujika et al., 2011; Ruipérez et al., 2012), DOP utilization and N₂ fixation (Liu et al., 2018) and may explain, in addition to DFe availability, the fertilizing effects observed on phytoplankton communities over half of the experiment. Subsequent nitrogen provision by these organisms in this N-poor environment likely allowed the development of larger non-diazotroph species (i.e., haptophyceae, diatoms and dinoflagellates), which in turn likely also explains the observed significant impact on primary production (Berthelot et al., 2016). Hydrothermal fluid-induced DFe fertilization, which resulted in higher primary production rates and POC export, indicated that such fertilization has a positive impact on the biological carbon pump, in line with in-situ results obtained during the TONGA cruise (Bonnet et al., in rev.). It should be noted that the experiment had to be stopped for logistic reasons, during the exponential phase of non-cyanobacterial species. Thus, POC export was likely supported by diazotrophs, as observed in-situ (Bonnet et al., 2023), and cyanobacteria, as larger non-diazotroph phytoplankton developed belatedly and were still in exponential phase when the experiment was stopped. This suggests that the effect of fluids on POC export was likely underestimated in our experiment due to the limited incubation time considered.