This study used Aquil medium due to its wide application in trace metal experiments (Price et al., 1989). The medium is composed of artificial seawater in which Milli-Q 18.2 MΩ cm-1 grade water is mixed with ultra-pure salts to reproduce the major ion composition of seawater (Pausch et al., 2019). The medium was filtered through a 0.2 µm pore size filter and sterilized in a microwave for a total of 11 min in acid-cleaned polycarbonate bottles (2 L) (Price et al., 1989). This artificial seawater is further enriched with the elements necessary for algal growth, such as vitamins, macronutrients (nitrate (NO3-) = 100, phosphate (PO43-) = 10, and silicate (SiO32-) = 100 µmol L-1), and various essential trace metals such as iron and manganese (Table A1). The trace metals were buffered with 100 µmol L-1 ethylenediaminetetraacetic acid (EDTA). A total of 17 Aquil media were produced that differed in the amount of Ni that was added, as will be described in more detail in the next section. Media preparation was done in a trace-metal-clean laminar flow hood. The salinity and pH (NBS scale) of Aquil media were 35 and 8.1 respectively.

Phytoplankton species were cultivated in acid-cleaned (10 % HCl for at least 24 h) polycarbonate tubes (30 mL, Nalgene™). These polycarbonate tubes (one tray with 40 tubes) were filled with Milli-Q water and then sterilized in the microwave for 8 min. The Aquil media were transferred from 1 L bottles into empty polycarbonate tubes under the clean bench under the trace-metal-clean laminar flow hood. Phytoplankton were added to the medium once it had reached chemical equilibrium (see next section).

The cultures were grown in a light chamber at 17 ∘C. All polycarbonate tubes were mounted onto a self-made “phytoplankton disc”, which rotated at 0.8 revolutions per minute (Fig. 1a). The phytoplankton disc ensured that equal light intensity was provided to all cultures and that phytoplankton cells were kept in suspension. The light was provided on a 14 to 10 h daily cycle (cool white fluorescent light) in which light intensities were 58 µmol photons m-2 s-1 (14 h) and 23 µmol photons m-2 s-1 (10 h). This unusual light cycle was due to some lights in the room being plugged-in energy sources, which had separate light–dark cycle setups linked to the computer system at the Institute of Marine and Antarctic Studies. Initially, we were not aware of this additional cycle and only realized the issue during the experiment. Therefore, we continued with this light cycle to maintain comparability between experiments. However, this issue does not affect the interpretation of the results as all species and replicates received the same amount of light throughout the experiment. The light intensity was the average light intensity at each of the 88 spots on the phytoplankton disc measured with a LI-COR light meter.

Aquil media were enriched with different concentrations of NiCl2: 0, 5, 10, 20, 30, 50, 70, 100, 150, 200, 300, 400, 500, 700, 1000, 10 000, and 50 000 nmol L-1. Unless otherwise noted, “Ni concentration” refers to the total added dissolved Ni concentration. For illustration and discussion of the data, concentrations were negatively log10 transformed: 1pNi=-log⁡10(Ni), where Ni is the total dissolved concentration of Ni (in mol L-1). This kind of transformation is also used to convert hydrogen ion concentrations to pH and is commonly used in studies investigating trace metal sensitivities to better visualize data when trace metal concentrations vary over orders of magnitude (Dupont et al., 2008).

Media were allowed to equilibrate chemically for at least 24 h before being inoculated with phytoplankton. To acclimate the phytoplankton strains, stock cultures were first transferred into Aquil medium without Ni enrichment. They were then cultivated for at least three batch cycles (i.e., transferred from one polycarbonate tube to the next one) before being transferred to polycarbonate tubes with the different Ni treatments. This ensured that the phytoplankton species were acclimated to Aquil medium before the Ni experiment commenced.

EDTA binds with metal ions and helps the dissolution of metal ions to create a nutrient-replete medium. Due to the addition of the ligand EDTA to the Aquil media, the “free Ni” ion concentrations (i.e., Ni2+) were substantially lower than the total dissolved Ni concentrations calculated with the chemical speciation software Visual MINTEQ 3.1 (Gustafsson, 2011).

We were interested to see if the response of phytoplankton to Ni may be different in other growth media where no EDTA was added. Therefore, we prepared a batch of natural seawater medium with water sampled from 15 m in the Southern Ocean (58.02∘ S, 141.17∘ E). There was little information about concentrations and types of Ni-binding organic ligands in the Southern Ocean because these ligands occur at very low concentrations within a highly complex mixture of organic matter (Boiteau et al., 2016). If we take Fe-binding organic ligands as examples, the characterized types of Fe-binding organic ligands were different in various studies due to the diverse measuring protocol, and the concentrations of these ligands in the Southern Ocean varied from 0.72 to 12.3 nmol L-1 (Nolting et al., 1998; Boye et al. 2001; Buck et al., 2010). Therefore, the Southern Ocean seawater we used in the experiment can be considered to have much lower organic ligands than the Aquil media (100 µmol L-1 EDTA). This natural seawater was filtered through an acid-cleaned 0.2 µm filter and sterilized in the microwave. The same amount of macro-nutrients (N, P, and Si) and vitamins were added as in the Aquil medium (mentioned above). The trace metal additions to the Southern Ocean seawater (no Ni included) were adjusted to a similar free trace metal concentration (nutrient-replete) as in Aquil medium (Table A1). For the experiment with natural seawater, we set up a dissolved Ni gradient with 17 concentrations: 0, 1, 2, 5, 10, 20, 30, 50, 70, 100, 150, 200, 300, 400, 500, 700, and 1000 nmol L-1. The extremely high Ni concentrations designed for the Aquil medium were avoided as we assumed the organic ligand concentrations in natural seawater to be much lower than the concentration of EDTA added in Aquil medium and therefore the concentration of free Ni2+ to be higher. We used P. tricornutum (CS-29) for this experiment. Phaeodactylum tricornutum was transferred from the stock cultures into natural seawater medium for three batch cycles prior to the experiment with different Ni treatments as described for the Aquil medium above.

The total ion concentrations of each trace metal in natural seawater and Aquil media before additions were measured using a seaFAST system and inductively coupled plasma mass spectrometry (ICP-MS). The free ion concentrations were calculated with Visual MINTEQ 3.1 based on the total ion concentration, together with the added concentration (Table 1).

Growth rate measurements were conducted according to the methods described by Andersen (2005). Briefly, the chlorophyll fluorescence of the cells was recorded daily at the same time of the day with a Turner model 10AU fluorometer. During the measurements, polycarbonate tubes did not have to be opened because they fit inside the sample chamber of the fluorometer. This reduced the risk of contamination as the polycarbonate tubes remained closed throughout the batch cycles. Fluorescence signals of samples were measured after 20 min of dark acclimation. The fluorescence values were ln transformed and plotted as a function of incubation days. A linear regression was fitted during the exponential phase of phytoplankton growth with the specific growth rate (μ; d-1) represented by the slope of the linear regression (Fig. 1b and c). We only used fluorescence values up to 13 (arbitrary unit) for our growth rate calculations so that the biomass in the incubation bottles remained relatively low and consistent with the dilute batch culture principle (LaRoche et al., 2010).

Reliable estimates of exponential growth rates in dilute batch cultures require multiple serial transfers of cultures (all performed while the strain is still in exponential growth) to allow the time for cultures to acclimate to the experimental conditions (Brand et al., 1981; Andersen, 2005). Therefore, the phytoplankton species were transferred into new polycarbonate tubes containing fresh medium during their early exponential stage for three batch cycles prior to recording growth rates shown in the results. This meant that cultures were usually growing in their respective treatment conditions for at least 3 weeks.

We conducted photo-physiological measurements at the end of each batch cycle. A fast repetition rate (FRR) fluorometer (FastOcean Sensor FRRf3, Chelsea Instruments Group) was used to measure the maximum quantum yield, Fv/Fm, and the functional absorption cross-section of photosystem II (σPSII; nm2 per reaction center (RC)). These measurements were done with cultures directly after they had been used to inoculate the subsequent batch cycle (hence avoiding contamination of ongoing cultures). Cultures were kept in the dark for 20 min before the measurements. For each treatment and species, 5 mL phytoplankton samples were added to the FRR fluorometry cuvette, which was temperature-controlled at 17 ∘C. Filtered Aquil media (or natural seawater media) were used at the beginning of the measurement for blank calibration. Throughout the experiment, FRR fluorometry was used with an acquisition sequence of 100 saturation flashes for 200 µs and 40 relaxation flashes for 2.4 ms, while the flash duration was set to 100 µs (Schallenberg et al., 2020). In each acquisition sequence, three channels with different light wavelengths were used: channel A with 450 nm light, channel B with 450 and 530 nm light, and channel C with 450 and 624 nm. The FRR fluorescence results from channel A (450 nm) were used to analyze diatoms, haptophytes, and dinoflagellates photosynthetic performance due to the presence of chlorophyll a in their cells, while channel C (450 and 624 nm) results were used to analyze the photosynthetic performance of cyanobacteria because of the presence of the phycobilin which is commonly present in cyanobacteria (Roy et al., 2011). At least 10 acquisitions were measured for each sample and used to calculate the average value of Fv/Fm and σPSII. Fv/Fm is usually lower under nutrient or light stress (summarized by Suggett et al., 2009), while σPSII describes the ability of light to promote a photochemical reaction in PSII (Falkowski and Raven, 1997). The value of Fv/Fm and σPSII are known to vary among algal taxa (Suggett et al., 2009). Typically, cells growing in batch cultures at the exponential growth phase exhibit a constant value of Fv/Fm and σPSII (Parkhill et al., 2001).

Every strain was able to grow in all Ni concentrations in Aquil media for at least three batch cycles. The data from the third batch were used for analyses. The growth rate and photo-physiological response of phytoplankton were analyzed using generalized additive models (GAMs) and plotted in RStudio (R packages “mgcv” and “ggplot2”) (RStudio team, 2020). For the GAM analyses, we assumed that growth rates, Fv/Fm, and σPSII of phytoplankton would show an optimum curve in response to the wide range of Ni concentrations: Ni limitation at the lower extremes, Ni inhibition at the upper extremes, and an optimum at some intermediate Ni concentration. GAMs were fitted to plots to assess the presence of a relationship between Ni concentration, growth rates, and photo-physiological responses. P values of the smooth terms of GAM models greater than 0.05 indicated that there was no statistically significant trend in the response variable (μ, Fv/Fm, or σPSII) in response to the wide Ni gradient (i.e., the smooth term was not significantly different from a horizontal line and therefore no statistically significant relationship between Ni and the measured parameter present). The general GAM equation is 2Ym=mI0m+Sm(pNi)m+e, where Y is the response variable (μ, Fv/Fm, and σPSII), I0 is the intercept, S(pNi) is the non-parametric smooth function according to pNi, and e the error. The k value (basis dimension) of GAM formula in RStudio was set to the minimum k value that fitted the curve and explained the data points without fitting random noise. The function “gam.check” in the package “mgcv” was used to assess the appropriateness of the selected k value following Wood (2022). The selection of k involved ensuring the P value was greater than 0.05. This ensured the selected k value was sufficiently small to not fit random noise (i.e., overfitting) but still be statistically appropriate.

Most trace metals in seawater are partially bound by organic ligands, and their bioavailable “free” concentrations are lower than the total dissolved ion concentrations (Van Den Berg and Nimmo, 1987). The thermodynamic equilibrium concentrations of the -log10-transformed “free Ni concentrations” (pNi2+) and total dissolved Ni concentrations (pNi) in the different media (see Eq. 1 and Table 1) correlate linearly (R2>0.99). Thus, both can be displayed as separate x axes on the same plot (Figs. 2, 3, and 4). For the Southern Ocean seawater media we assumed the ligand concentration to be 0 and thus that the differences between pNi and pNi2+ are very small. In Aquil, however, the differences between pNi and pNi2+ are very large due to the presence of EDTA.

A total of 6 out of the 11 strains displayed statistically significant growth rate changes in response to Ni sensitivity (Fig. 2, Table 2). These strains were Synechococcus sp. (CS-205), A. carterae (CS-740), E. huxleyi (CS-1185), A. glacialis (CS-135), N. closterium (CS-5), and P. tricornutum (CS-29). Among these strains, N. closterium (CS-5) and P. tricornutum (CS-29) had consistent increasing growth rates when pNi increased (Fig. 2j and k). Other strains displayed optimum curve response patterns, although variations in growth rates between the low, high, and optimum concentrations of pNi and these trends were below 36 % (Table A3). Most of their optimal growth rates were in the range of pNi 8–7 (10 to 100 nmol L-1). Growth rates of the other strains (Geitlerinema sp. (CS-897), Oscillatoria sp. (CS-52), Cricosphaera sp. (CS-1183), I. galbana (CS-186), and P. parvum (CS-659)) were not significantly affected by different Ni concentrations.

The cyanobacterium Oscillatoria sp. (CS-52) tended to aggregate during culturing, and the fluorescence signals were more variable on a day-to-day basis. This made the growth rate calculation less accurate, indicated by lower R2 values in linear regression when fitting ln-transformed data over time to calculate the growth rate.

We were interested if we could trust singular data points at the extreme ends of the optimum curves as they often drove trends in our data (e.g., Synechococcus in Fig. 2 at pNi < 7.5, total dissolved Ni <30 nmol L-1). Therefore, we did an additional experiment with Synechococcus sp. (CS-205) in which we replicated the lowest added Ni treatment (0; 0.07 nmol L-1 including background Ni) and the optimum Ni concentration (20 nmol L-1) (Table 3). The results confirmed the trend in the optimum curve, with the added 20 nmol L-1 Ni resulting in significantly enhanced growth rates (Table 3).

The FRR fluorescence data were largely consistent with the growth rate data in that no strong trends within the Ni range tested were observed for most of the species. The σPSII and Fv/Fm measurements across the Ni gradient revealed minimal trends, with generally little variation between treatments (Figs. 3 and 4). A few exceptions to this general pattern of results are mentioned below.

Synechococcus sp. (CS-205) had higher Fv/Fm values in the mid-pNi range (pNi 8–6, 10–1000 nmol L-1), and the lowest Fv/Fm value was in the Aquil medium without any Ni addition. In contrast, Geitlerinema sp. (CS-897) had lower Fv/Fm values in the mid-pNi range, but the variation between maximum and minimum Fv/Fm values was small. These two strains, however, exhibited little change in σPSII over the range of Ni treatment. Some species (e.g., A. carterae (CS-740), P. parvum (CS-659), and A. glacialis (CS-135)) had slightly lower σPSII values at the highest Ni concentrations (> 10 000 nM), suggesting some reduction in light harvesting capacity at high Ni concentrations. The small P values of the smooth terms (Table 2) are likely driven by these low σPSII values in the high Ni concentrations. In general, most of the tested strains appeared photosynthetically healthy across the tested Ni gradient.

The most pronounced effect of Ni was observed in Fv/Fm for Synechococcus sp. (Fig. 3). Fv/Fm was considerably lower at pNi = 10.2 (0.07 nmol L-1) than at the optimum concentrations (approximately pNi = 7.7, 20 nmol L-1). Our additional experiment with Synechococcus, in which we replicated the pNi 10.2 and 7.7 treatments three times, did not confirm this trend (Table 3). Neither Fv/Fm nor σPSII values were significantly different between the two Ni concentrations (Table 3).

Phaeodactylum tricornutum (CS-29) growing in the natural Southern Ocean seawater media (see Sect. 2.2) showed no significant trend (Fv/Fm) or particularly strong changes (growth rate, σPSII) across the experimental Ni concentration gradient. This result was very similar to the result of P. tricornutum (CS-29) grown in Aquil media. The average growth rate of P. tricornutum growing in the Southern Ocean seawater media was 0.83 d-1, which was very similar to the growth rates of the cultures growing in Aquil media (0.86 d-1) (Fig. 2k). Absolute numbers were also very similar for the σPSII and Fv/Fm data (Figs. 3 and 4).

Based on growth rates and FRR fluorescence results, we conclude that changes in dissolved Ni, within the range tested and under the experimental conditions, do not have a strong effect on the 11 phytoplankton species. Only four species showed significant trends in both growth rates and at least one photophysiological parameter. A total of 8 out of 11 species showed < 25 % and < 16 % change relative to the average values in growth rates and photo-physiological parameters, respectively (Table A3). An exception was Synechococcus sp. (CS-205), which showed a significant and quite pronounced growth rate enhancement of 74 % from the lowest to optimum Ni (20 nmol L-1) and then gradually declining growth rates towards the highest Ni. Likewise, growth rates of A. carterae (CS-740) showed a relatively pronounced Ni sensitivity, following an optimum curve with highest growth rates between a pNi of 8–7 (10–100 nmol L-1). The Ni sensitivity of growth rates in the other species where significant trends were detected were smaller, i.e., smaller than 25 % change relative to the average growth rate of the species (Table A3). However, we emphasize that even a small difference in growth rate can have a pronounced effect on population sizes during extended periods of growth due to the exponential nature of phytoplankton reproduction. For example, an increase in growth rate by 0.05 d-1 (as frequently observed in our data; Fig. 2) would lead to a ∼ 65 % larger population at the end of a 10 d growth period. Furthermore, even if a species is completely insensitive to Ni, it may still be affected indirectly within a competitive environment with multiple phytoplankton species present. This is because other species may benefit from, or be inhibited by, changing Ni concentrations, thereby altering the competition for nutrient resources. Therefore, small changes in growth rates should not be readily marginalized as they may still be of ecological and biogeochemical relevance.

The inhibition of growth rate or photosynthesis performance was evident in a few species when pNi reached 5 (10 000 nmol L-1) (i.e., Fig. 2a and d), but most species did not have growth inhibition in high Ni concentrations. The relatively small effects of high Ni on growth rates, Fv/Fm, and σPSII were surprising because we expected stronger species-dependent Ni sensitivity within the pNi range of 9–5, at least based on the available experimental evidence summarized by Glass and Dupont (2017). There are several potential reasons for the disagreement on the Ni sensitivity results from previous research. These will be discussed in the following subsections.

OAE can be achieved by distributing pulverized rocks on land and ocean surfaces, thereby accelerating chemical weathering rates and the generation of alkalinity. The environmental perturbation depends on the chemical composition of the applied rock minerals. If dunite is used as the source rock for OAE (an olivine-rich ultrabasic rock, often associated with vulcanism), the Ni perturbation could be particularly high as dissolution experiments with olivine powder found roughly a 3 µmol kg-1 increase in dissolved Ni for a ∼ 100 µmol kg-1 increase in alkalinity within approximately 50 d (Montserrat et al., 2017). However, it is difficult to estimate how the free Ni2+ concentration will change because it depends on the organic ligand concentration at the perturbation site. Furthermore, the optimal Ni concentration can vary considerably between phytoplankton species and as discussed above, can depend on the availability of N sources in the environment. Therefore, we need to consider not only the total inputs of Ni but also regional differences in organic ligand concentrations, nutrient availability, and phytoplankton community composition to evaluate the potential impact of Ni on phytoplankton.

Our results suggest that excess Ni has a limited toxic impact on most of the phytoplankton species tested in our study. As the tested species cover a relatively wide range of taxa, it may be assumed that our findings can be generalized more widely to natural communities of phytoplankton in temperate regions. However, great care must be taken when interpreting our results because we used EDTA, a strong organic ligand, in our experiments. EDTA binds large amounts of Ni so that using the total dissolved Ni concentration for inferring the absence of a toxic effect of high Ni on phytoplankton may not be valid. Although we confirmed the absence of a toxicity effect of Ni on P. tricornutum (CS-29) grown in Southern Ocean seawater media (which did not contain EDTA), we cannot rule out toxicity for all the other species tested here. Because P. tricornutum is a known “lab rat” that readily grows under a wide range of conditions, the absence of a toxicity effect in this species is not necessarily indicative of other species. Thus, we must emphasize that our results do not reject the possibility that high Ni concentrations invoked by OAE could inhibit the growth of phytoplankton.

If we assume that ligands mitigate the impacts of dissolved Ni by reducing the concentration of Ni2+, it raises an interesting question: could such a dependency be exploited for OAE implementation strategies? As we discussed in section 4.1, open-ocean ecosystems probably have lower organic ligand concentrations than many coastal or estuarine regions. Thus, under this assumption, a perturbation with Ni due to OAE would lead to a more pronounced increase in Ni2+ in open-ocean systems than the same perturbation in a coastal/estuarine region rich in organic ligands. Therefore, future research could investigate if regional differences in ligand concentration may be utilized to identify suitable spots to manage environmental impacts of OAE applications with Ni-rich minerals.