Iron Depletion in the Deep Chlorophyll Maximum: Mesoscale Eddies as Natural Iron Fertilization Experiments

In stratified oligotrophic waters, phytoplankton communities forming the deep chlorophyll maximum (DCM) are isolated from atmospheric iron sources above and remineralized iron sources below. Reduced supply leads to a minimum in dissolved iron (dFe) near 100 m, but it is unclear if iron limits growth at the DCM. Here, we propose that natural iron addition events occur regularly with the passage of mesoscale eddies, which alter the supply of dFe and other nutrients relative to the availability of light, and can be used to test for iron limitation at the DCM. This framework is applied to two eddies sampled in the North Pacific Subtropical Gyre. Observations in an anticyclonic eddy center indicated downwelling of iron‐rich surface waters, leading to increased dFe at the DCM but no increase in productivity. In contrast, uplift of isopycnals within a cyclonic eddy center increased supply of both nitrate and dFe to the DCM, and led to dominance of picoeukaryotic phytoplankton. Iron addition experiments did not increase productivity in either eddy, but significant enhancement of leucine incorporation in the light was observed in the cyclonic eddy, a potential indicator of iron stress among Prochlorococcus. Rapid cycling of siderophores and low dFe:nitrate uptake ratios also indicate that a portion of the microbial community was stressed by low iron. However, near‐complete nitrate drawdown in this eddy, which represents an extreme case in nutrient supply compared to nearby Hawaii Ocean Time‐series observations, suggests that recycling of dFe in oligotrophic ecosystems is sufficient to avoid iron limitation in the DCM under typical conditions.

Definitive evidence of Fe limitation in surface waters ultimately demanded the upscaling of Fe addition experiments from liter-sized bottles to the ecosystem scale via in situ fertilization experiments (Boyd et al., 2007), and from parallel studies of natural iron fertilization events: coastal Fe input from islands Martin et al., 1994;Pollard et al., 2009), or atmospheric deposition from dust storms (Bishop et al., 2002) and volcanic eruptions (Achterberg et al., 2013;Hamme et al., 2010). At present, there are significant logistical (not to mention ethical [Strong et al., 2009]) challenges facing would-be Fe fertilization experiments in the lower euphotic zone: the DCM cannot be observed by satellite and is out of reach of most underway sampling systems. Natural Fe fertilization events to the oligotrophic DCM (if they can be observed) represent an alternate approach.
Here, we show how perturbations to light, nutrient and iron supply caused by mesoscale eddy activity can be used to examine the vulnerability of phytoplankton in the DCM to iron limitation. This scheme is then applied to two adjacent eddies, a cyclone and an anticyclone, observed in the North Pacific Subtropical Gyre during Summer 2017. Along with several recent studies (Browning et al., 2021;Ellwood et al., 2020;Sedwick et al., 2020), our measurements also show that local Fe cycling is strongly perturbed by mesoscale eddies.

Expedition Summary
Sampling and experiments were conducted on the R/V Kilo Moana as part of the MESO-SCOPE expedition (26 June to 15 July 2017) in the North Pacific Subtropical Gyre. Satellite altimetry was accessed from the Copernicus Marine Environmental Monitoring Service and processed as described by Barone et al. (2019) to remove the seasonal cycle and interannual trend. The centers of the two mesoscale eddies sampled in this study were identified as maxima and minima of corrected sea level anomaly (SLA corr ): a cyclonic eddy centered at 24.9°N, 158.7°W, and an anticyclonic eddy centered at 26. 4°N, 158.0°W. An initial survey of the subsurface density structure across the upper water column was performed with an underway CTD (Teledyne), followed by hydrographic sampling for metals and nutrients, and deployment of autonomous Wirewalker drifting profilers near the eddy centers. This initial sampling was followed by Lagrangian, multi-day occupations of the eddy centers following Surface Velocity Program (SVP) drifters tracking near surface currents at 15 m depth, deployed near the centers of both eddies. Two Wirewalker drifting profilers were deployed near the SVP drifters to obtain vertically resolved observations of the upper 400 m at high temporal resolution. A more detailed description of sampling activities can be found in a companion manuscript (Barone et al., Under Revision). During the Lagrangian period, three sunrise to sunset (∼14 hr) primary production experiments were conducted in each eddy using surface tethered arrays following established methods for the Hawaii Ocean Time-series (HOT). These were accompanied by longer term (89-hr) Fe amendment experiments on a separate array.
For all hydrographic parameters, sampling depths were converted to a mean isopycnal depth specific to each eddy center, which was calculated based on multi-day averages of potential density profiles determined from the Wirewalkers. Daily integrals of downwelling photosynthetically active radiation (PAR) were calculated as the product of continuous measurements of surface PAR from a shipboard sensor , and the attenuation of PAR with depth, measured daily with a Hyperpro II optical profiler (Sea-Bird). Reported PAR values at the DCM are averaged from 5 m above and below the mean isopycnal depth of the DCM for each day, and then averaged across the 3-day Lagrangian sampling period.

Nutrient, Chlorophyll, and Flow Cytometry Analyses
Nutrient samples were collected using a rosette sampler and frozen immediately. Concentrations of nitrate + nitrite were analyzed by the high sensitivity chemiluminescent method described by Foreman et al. (2016), with a detection limit of 1 nM. Individual samples near the DCM of both eddies were also analyzed for nitrite using chemiluminescence (Foreman et al., 2016). Dissolved silicate was measured colometrically using a SEAL AA3 auto-analyzer (Strickland & Parsons, 1972;Wilson et al., 2019). Chlorophyll a was measured by the fluorometric method after acetone extraction (Lorenzen, 1967). Abundance of chlorophyll a, b, and c and divinyl chlorophyll a and b were measured by high performance liquid chromatography mass spectrometry (HPLC-MS) following procedures recently described by Becker et al. (2021). Abundance of picoeukaryotes, Prochlorococcus, Synechococcus, and heterotrophic bacterial cells were determined from samples preserved in 0.24% paraformaldehyde, flash frozen at −80°C and analyzed by flow cytometry on an Influx flow cytometer (BD). Phytoplankton populations were identified by fluorescence and scattering properties. Heterotrophic (non-pigmented) populations were analyzed after staining with SYBR Green I, with phytoplankton contributions subtracted.

Dissolved Iron Sampling and Analysis
Trace metal sampling was conducted using a 12-position powder-coated "trace metal" rosette (SBE 32C with SBE 9plus CTD, Sea-Bird Electronics) mounted with 8 L externally sprung Niskin bottles (Ocean Test Equipment). The rosette was deployed on Spectra line using a metal-free block. All samples were processed in a HEPA-filtered, positive pressure trace metal clean "bubble" within a laboratory van. Bottles were filtered with an Acro-Pak 1500 cartridge filter (0.8 and 0.2 μm pore size) into 1 and 4 L LDPE bottles (Nalgene). All plasticware was rigorously cleaned by soaking in 2% Citranox overnight, followed by 1 week in a 10% hydrochloric acid (HCl) bath, and extensive rinsing with ultra-high purity water.

Siderophore Analysis by HPLC-ICPMS and HPLC-MS/MS
Iron-binding ligands and siderophores were extracted from 4 L of 0.2 μm filtered seawater onto solid phase extraction columns at a flow rate of 20 mL min −1 . Prior to extraction, 6 mL Bond-Elut ENV columns (1 g, Agilent) were activated by passing 9 mL of high-purity methanol (Fisher Optima), 3 mL of 10 mM HCl, and 6 mL ultra-high purity water. Samples were then rinsed with 6 mL water, frozen at −20°C, and returned to the laboratory at Woods Hole Oceanographic Institution for processing. Immediately prior to analysis, columns were thawed and ligands were eluted with 12 mL methanol. Extracts were concentrated at 35°C to ∼1.5 mL using a SpeedVac (Thermo Scientific). Aliquots were evaporated to dryness, and reconstituted in ultra-high purity water.
Iron ligands were analyzed using a Dionex Ultimate 3000 bioinert liquid chromatography system (Thermo Scientific) fitted with a C18 column (0.5 mm × 150 mm, 5 μm, Agilent). Compounds were eluted at a flow rate of 40 μL min −1 with a 20 min gradient from 5% to 90% solvent B, followed by a 10 min gradient from 90% to 95% solvent B, and a 5 min isocratic hold at 95% solvent B (solvent A: 5 mM aqueous ammonium formate, solvent B: 4 of 18 5 mM methanolic ammonium formate). Eluent was plumbed directly into a quadrupole ICP-MS (iCAP-Q, Thermo Scientific), using instrument settings described in Bundy et al. (2018). Instrument sensitivity was determined with a 4-point calibration curve of ferrichrome and ferrioxamine E standards. To identify siderophores, the liquid chromatography system was coupled to an Orbitrap Fusion mass spectrometer (Thermo Scientific) equipped with a heated electrospray ionization (H-ESI) source. Tentative identifications are made by comparison to known siderophore masses and retention times, as well as the presence of iron isotope pattern for siderophores bound to 54 Fe and 56 Fe. Ultra-high purity solvents and reagents were used throughout. Expanded details of this protocol are described in Boiteau and Repeta (2015).

Fe Amendment Incubations
Fe limitation of primary production was investigated with multi-day incubations conducted at in situ light and temperature conditions on a surface-tethered array. Seawater was collected from nighttime trace metal rosette casts and subsampled into cleaned 30 mL, 500 mL, and 2 L polycarbonate bottles (Nalgene) to determine bacterial production, primary production, and prokaryotic community structure (via 16S rRNA analyses), respectively. For Fe amended treatments (+Fe), 2 nM Fe was added directly to incubation bottles (as 5 μM FeCl 3 in a 10 mM HCl solution). At the DCM, additional treatments were also conducted using Fe bound to commercially available ferrioxamine B (Sigma) and amphibactin siderophores purified by HPLC from cultures of Vibrio cyclitrophicus 1F-53 following Boiteau et al. (2016). Siderophore-Fe complexes were added to final concentrations of 2 nM Fe with stock solutions stored at 4°C until use. The 500 mL and 2 L bottles were mounted on the array with custom-built acrylic frames and 30 mL bottles were placed in mesh bags and affixed to the array line. Dark incubations were conducted in vinyl dry-bags with pin-holes cut to allow exchange with surrounding waters with minimal light penetration. All processing was performed in the trace-metal-free bubble underneath HEPA-filtered workstations. Bottles were stored in coolers prior to deployment to minimize light and temperature perturbations during set up.
The arrays were deployed before sunrise and allowed to drift freely for 69 hr, when they were recovered (after sunset). At this point, 500 mL bottles were spiked with 14 C bicarbonate and 30 mL bottles were spiked with 3 H-leucine, prepared as described previously (Viviani & Church, 2017). Both radioisotope solutions were cleaned of metals using Chelex-100 resin (Bio-Rad; prepared according to Sunda et al., 2005) conditioned with unlabeled bicarbonate and leucine solutions at equal concentration. The 500 and 30 mL bottles were then re-affixed to the array, which was deployed prior to sunrise, ∼73 hr after the start of the experiment. Incubations to measure primary and bacterial production continued until sunset (89 hr, an incubation time of 14 hr for the 3 H-leucine and 14 C uptake assays) when the arrays were recovered and measured according to established protocols (Viviani & Church, 2017). This duration is equivalent to the standard HOT program productivity measurements that also span dawn-to-dusk incubation periods. The 2 L bottles were harvested after initial recovery of the array (at 69 hr), filtered onto 25 mm 0.2 μm polyethersulfone filters, and preserved with RNA-later and frozen at −80°C. 16S rRNA genes were amplified from extracted genomic DNA with primers targeting the V4-V5 hypervariable regions and sequenced on an Illumina MiSeQ. Amplicon sequence variants were generated in DADA2 (Callahan et al., 2016). Full description is provided in Supporting Information S1.
Following the deployment of the in situ array, additional seawater from the DCM was collected in 2 L bottles and placed in incubators on deck (Caron Model# 7001-10) with light and temperature set to conditions matching the DCM in each eddy. Filtered control experiments were conducted in parallel and samples for HPLC-ICPMS analyses were collected after 5 days of incubation.

Transcriptome Searches
Metatranscriptome samples were collected at ∼4 hr intervals for three days from isopycnal surfaces within the DCM during the Lagrangian sampling of the cyclonic (25.24 kg m −3 ) and anticyclonic eddies (24.43 kg m −3 ; n = 18 for each eddy). Procedures for filtration, preservation, addition of quantitative standards, sequencing and assembly followed published protocols (Gifford et al., 2016;Wilson et al., 2017) and are described fully in Supporting Information S1. Prochlorococcus iron responsive genes were identified from the culture experiments of Thompson et al. (2011). The abundance of gene transcripts (as copies mL −1 ) were aggregated at the genus level and then normalized to total Prochlorococcus transcripts. The statistical significance of the relative change in expression was tested using a Kruskal-Wallis test in the SciPy Python package (Virtanen et al., 2020).

Analysis of Hawaii Ocean Time-Series Datasets
HOT data from Station ALOHA were accessed using the Hawaii Ocean Time-series data organization and graphical system (HOT-DOGS: hahana.soest.hawaii.edu/hot/hot-dogs/), except for dFe data, which are replotted from Fitzsimmons et al. (2015). Direct comparisons to the MESO-SCOPE expedition are restricted to May -September months, while comparisons with PAR are conducted with monthly averages throughout the year (Karl et al., 2021). Monthly mean PAR values at specific depths are calculated as the product of the monthly averages of daily integrated irradiance at the ocean surface and the fraction of surface PAR at "standard" HOT depths.

Mesoscale Eddies as Natural Iron Fertilization Experiments to the Lower Euphotic Zone: Concepts and Predictions
Mesoscale eddies form from instabilities in large scale currents and can isolate waters for several months while transporting them hundreds of kilometers from their origin (Chelton et al., 2011;Zhang et al., 2014). The rotation of mesoscale eddies leaves a characteristic distortion in the corrected sea-level anomaly (SLA corr ; see Barone et al., Under Revision). Anticyclonic motions are associated with SLA maxima and accumulation of surface waters. Conversely, cyclonic eddies are associated with SLA minima and a depletion of surface water near the eddy center. Although there are myriad perturbations that can be encouraged by eddy activity (McGillicuddy, 2016), the most consistent feature of mesoscale eddies is the vertical displacement of density strata according to eddy polarity: cyclonic eddies lift denser waters upward, while anticyclonic eddies depress isopycnal surfaces downward Siegel et al., 1999;Wunsch, 1997).
Mesoscale perturbation is of special relevance to the lower euphotic zone, where eddies alter the balance between light (largely a function of depth and light attenuation) and nutrient concentration (a function of density and vertical diffusivity) that influence phytoplankton growth in the DCM (Figure 1). Uplift of denser, nutrient-rich waters associated with cyclonic eddies can increase NO 3 − supply relative to a given light intensity (isolume), Note that surface dFe in the cyclone is predicted to be lower due to the uplift of low dFe waters at the initial DCM, but the changes in mixed layer dFe for both eddies depends on Fe sources from the atmosphere, as well as biological responses in the mixed layer affecting Fe uptake.
while anticyclones displace the nutricline downward, decoupling DCM isolumes from the diffusive supply of nitrate from below.
The effect of mesoscale perturbation can be separated into two conceptual stages: (a) the initial, physical perturbation and (b) the biological response to the physical perturbation. For the case of the cyclonic eddy, isopycnal uplift will shift the original DCM layer to shallower depths, where light absorption does not require the same degree of photoacclimation. Over time, the original DCM is expected to fade while a new DCM layer emerges underneath, near the 0.5 mol photon m −2 day −1 optimum that supported the DCM prior to perturbation (Letelier et al., 2004). For the anticyclone, the original DCM will be positioned too deeply for photosynthesis to remain viable, leading to the development of a new DCM in shallower waters that approach the 0.5 mol photon m −2 day −1 isolume. For either eddy type, biological responses in the mixed layer may further modulate the positioning of the DCM by increasing light attenuation in the overlying water column.
Because the original DCM overlaps with a minimum in dFe, with concentrations increasing both above and below, regrowth of the DCM within a mesoscale eddy will occur in waters with higher dFe (Figure 1). The cyclonic eddy DCM will regrow in an environment that now hosts higher dFe from the nutricline, in addition to higher NO 3 − . In contrast, an anticyclone DCM will regrow within lower density waters, containing elevated dFe originating from the surface but with potentially lower NO 3 − . Thus, for eddies that are sufficiently long-lived, re-equilibration of the lower euphotic zone ecosystem will occur in distinct nutrient regimes: the centers of anticyclones resemble an Fe addition without NO 3 − , while cyclones represent a simultaneous addition of Fe and NO 3 − ( Figure 1). From this, we hypothesize that both eddy types would relieve iron limitation at the DCM, if it occurred, and would increase primary production and carbon export relative to baseline values at similar isolumes. In contrast, prevalence of nitrogen limitation would favor a productivity increase within cyclonic eddies only. The ecosystem response to mesoscale forcing may therefore reveal the extent of iron limitation at the DCM in the mean state.

Iron, Nutrient, and Phytoplankton Perturbations Along an Eddy Dipole
During July 2017, we examined the biogeochemistry of a cyclonic-anticyclonic eddy pair in the central North Pacific Subtropical Gyre (Figure 2). A more detailed description of the 2017 MESO-SCOPE expedition is presented by Barone et al. (Under Revision). Satellite tracking of both eddies suggested a similar point of origin northeast of Hawaiʻi, with estimated ages of 100 days for the cyclonic eddy and 140 days for the anticyclone. At the time of their sampling, these eddies were near their peak SLA: +24 and −15 cm for the anticyclone and cyclone, respectively ( Figure 2a). Eddy centers were associated with large perturbations to the density structure of the upper ocean, with an eddy-to-eddy difference of ∼120 m for the depth of the 25.0 kg m −3 isopycnal anomaly (σ θ ), which falls within the nitracline at Station ALOHA (1.9 ± 0.9 μM NO 3 − , 190 m mean depth; Figure 2b). In contrast, the position of the DCM between eddy centers varied by less than 20 m, corresponding to 25.2 kg m −3 in the cyclonic eddy (104 m depth) and 24.4 kg m −3 in the anticyclone (118 m). Multi-day integrals of PAR indicated that the DCM of both eddies occurred at a similar light flux: 0.46 ± 0.18 and 0.29 ± 0.02 mol photon m −2 day −1 for the cyclonic and anticyclonic eddies, respectively (Table 1). While these irradiance estimates fall slightly below the canonical value of 0.5 mol photon m −2 day −1 (Letelier et al., 2004), they suggest that the eddies sampled in 2017 were sufficiently mature and stable to allow for biological re-equilibration of the lower euphotic zone to an optimal irradiance.
The ecosystem response to eddy-driven perturbations left clear signatures in nutrient inventories. At 200 m depth (below the photosynthetic compensation irradiance), isopycnal uplift in the cyclonic eddy led to NO 3 − concentrations of 7.65 μM, an order of magnitude greater than observed in the anticyclonic eddy at the same depth (0.69 μM; Figure 3). Most of this difference in NO 3 − could be accounted for solely by the vertical displacement of water masses. As a result, variability decreased substantially when eddy center profiles were compared against density ( Figure 3). Similar depth offsets were observed for profiles of dissolved Ni, Cu, and Cd, and Mn, which also realigned when plotted against density ( Figure S1 in Supporting Information S1). However, between 25.0 and 25.3 kg m −3 , the cyclonic eddy contained 1-3 μM less NO 3 − than the anticyclonic eddy. This density range overlapped with a large population of small eukaryotic phytoplankton (picoeukaryotes; up to 5 × 10 3 cells ml −1 ) in the cyclonic eddy DCM and a decrease in Prochlorococcus abundance (Figure 4). Although Prochlorococcus is typically responsible for most of the productivity, biomass, and chlorophyll near the DCM in this region (e.g., at Station ALOHA; Rii et al., 2016), the Prochlorococcus-specific pigments divinyl chlorophyll a and b were exceeded by chlorophylls a, b and c in the cyclonic eddy DCM, the latter three representing ∼60% of total chlorophyll and primarily reflecting eukaryotic phytoplankton (Synechococcus and other cyanobacteria were much less abundant; Table 1). Among larger cells (>4 µm), haptophytes were the strongest contributors to the increase in eukaryotic biomass in the cyclonic eddy DCM, with smaller contributions from diatoms, dinoflagellates and chlorophytes (Barone et al., Under Revision). We note that isopycnal comparisons indicate little Si drawdown in the cyclonic eddy center ( Figure S2 in Supporting Information S1). In the anticyclone, Prochlorococcus accounted for 58% of total chlorophyll at the DCM. Together, these observations suggest that uplift of nutrient-rich waters into the lower euphotic zone enabled significant biological uptake of NO 3 − in the cyclonic eddy, primarily by eukaryotic phytoplankton, consistent with the predictions of the conceptual model.
In contrast, the depth profile of dFe in the cyclonic eddy resembled profiles in the anticyclonic eddy and under non-eddy conditions at Station ALOHA. Each of these profiles contained a minimum at the DCM (Figure 3d), despite the fact that the DCM of the cyclonic eddy occurred on a distinct isopycnal surface that is usually associated with elevated Fe from remineralization ( Figure 3h). Indeed, dFe at 25.25 kg m −3 -within the DCM of the cyclonic eddy-was 63 pM, compared to 170 pM in the anticyclone at a similar density (25.15 kg m −3 , 250 m depth). The latter value is similar to an average dFe of 152 ± 46 pM between 175-250 m depth at Station ALOHA (Fitzsimmons et al., 2015), corresponding to a range of σ θ between 24.9 and 25.5 kg m −3 . Thus, NO 3 − uptake in the cyclonic eddy coincided with dFe uptake on the order of 100 pM. While the anticyclone also had a minimum in dFe at the DCM, this concentration (∼136 pM) was roughly double the dFe measured in the cyclone and at Station ALOHA (Table 1), likely reflecting downwelling of dFe from surface waters.

Direct Tests of Fe Limitation in the Lower Euphotic Zone
To determine the influence of the eddy perturbation on ecosystem productivity, primary production was measured during a period of Lagrangian sampling in the centers of both eddies ( Figure 5). Consistent with isopycnal evidence for uptake of NO 3 − and dFe, rates of primary production in the DCM (by the 14 C method) were moderately higher (0.21 ± 0.03 μM C day −1 between an irradiance of 0.2-0.6 mol photon m −2 day −1 , n = 4) compared to rates at equivalent PAR in the anticyclonic eddy (0.10 ± 0.01, n = 2) or average values from Station ALOHA (0.12 ± 0.04 μM C day −1 , Table 1). This amounts to a doubling of productivity within the DCM, but the effect is small compared to the large increase in 14 C uptake at all sites with increasing light ( Figure 5). Similar relationships between irradiance and primary production in the anticyclone and at Station ALOHA suggest that the two-fold increase in dFe at the anticyclonic eddy DCM did not stimulate primary production.
Longer term incubations were conducted to evaluate the sensitivity of primary production to dFe. After 3 days of incubation in situ, unamended control incubations in the cyclonic eddy harbored an unusual maximum in 14 C uptake at 100 m depth, just above the DCM, equal to 0.43 ± 0.04 μM C day −1 (mean ± 1 standard deviation of 3 bottle replicates; Figure 6). This rate is more than double the productivity observed in shorter term experiments at 100 m (0.21 μM C day −1 ) and 3 standard deviations above mean May-September primary production at 100 m for Station ALOHA (0.21 ± 0.07 μM C day −1 ; nearly equal to the maximum rate of 0.55 μM C day −1 over the entire record at 100 m). The higher productivity in the longer term experiments could be evidence of reduced grazing pressure within the incubated bottles, allowing biomass to increase. Iron addition to bottles incubated at this

Table 1 Attributes of the Deep Chlorophyll Maximum (DCM) in the Cyclonic and Anticyclonic Eddies Observed During the 2017 MESO-SCOPE Expedition
depth increased primary production slightly to 0.54 ± 0.08 μM C day −1 , but not significantly (p > 0.05, Student's t-test). At the DCM of the cyclonic eddy (incubated at 110 m), primary production was lower: 0.25 ± 0.05 μM C day −1 in the control and 0.27 ± 0.07 μM C day −1 in the 2 nM Fe treatment, similar to the short term productivity measurements from the DCM (0.22 ± 0.04 μM C day −1 ). Equivalent primary production was also observed in control and +Fe treatments throughout the lower euphotic zone of the anticyclone ( Figure 6). Additions of Fe as Fe-amphibactin D or ferrioxamine B siderophores also did not increase DCM primary production relative to the unamended control in either eddy center ( Figure S3 in Supporting Information S1).
Although primary production did not change significantly following Fe addition, Fe did increase rates of 3 H-leucine incorporation at both 100 and 110 m in the cyclonic eddy. At 100 m, leucine incorporation under ambient  light increased from 15.7 ± 1.8 pM hr −1 (control) to 23.5 ± 2.2 pM hr −1 (+Fe; p < 0.05, Student's t-test; Figure 6b). A considerably smaller increase was observed in dark incubations at this depth (13.3 ± 1.2 vs. 15.4 ± 0.08, Figure S3 in Supporting Information S1). At the DCM (110 m), Fe also increased rates of 3 H-leucine incorporation ∼30% above unamended controls under ambient light. Analysis of the bacterial community at 100 and 110 m by amplification and sequencing of 16S rRNA genes indicated dominance of Prochlorococcus and Pelagibacter, but community composition in ambient light incubations did not differ significantly between controls and +Fe treatments ( Figure 6c). Compared to observations at Station ALOHA (Viviani & Church, 2017), rates of 3 H-leucine incorporation were high throughout the euphotic zone of the cyclonic eddy, possibly reflecting the same relief of grazing pressure that led to higher 14 C uptake in long term incubations. Alternatively, elevated rates could be explained by microbial succession following high initial productivity and export during the intensification phase of this eddy (Barone et al., Under Revision).
In the anticyclone, rates of 3 H-leucine incorporation were similar to observations at Station ALOHA, and lower throughout the water column compared to the cyclonic eddy. Leucine incorporation was not stimulated with Fe Figure 5. Intercomparison of nutrients, iron, productivity, and light. Cyclonic (blue) and anticyclonic (red) eddy center profiles of (a) chlorophyll a, (b) nitrate, (c) dissolved iron from Figure 3, and (d) rates of 14 C uptake from three separate 14-hr (sunrise to sunset) incubations at in situ temperature and light during the Lagrangian period. Each circle for 14 C uptake represents the mean of triplicate incubations. Monthly averaged data from Station ALOHA are plotted as gray circles. Note that these 14-hr primary production measurements are distinct from the multi-day experiments in Figure 6. addition at 100 m, nor at the DCM (120 m), consistent with greater dFe in the anticyclonic eddy (p > 0.05, oneway ANOVA). Despite distinct prokaryotic communities between eddies and between 100 m and the DCM (Figure 6c), no significant changes in community structure (based on rRNA gene analysis) were observed following Fe addition in the anticyclone.

Mesoscale Perturbation of Siderophore Distributions and Cycling
Most dFe in the upper ocean is strongly bound to organic ligands, including siderophores, which cycle at an unknown rate. Prior measurements of excess Fe ligand concentrations in the North Pacific Subtropical Gyre appear invariant over the upper 300 m (Fitzsimmons et al., 2015), but siderophores with unique depth distributions have been identified throughout the water column at Station ALOHA (Bundy et al., 2018). In the eddy dipole, HPLC-ICPMS analyses highlighted significant shifts in Fe speciation between eddies. Both eddies displayed an unresolved complex mixture of Fe ligands that appear as a hump in the chromatogram baseline (Figure 7a). Within each eddy, these unresolved ligands were roughly constant with depth, but were greater in the anticyclone (44 ± 5 pM) compared to the cyclone (17 ± 2 pM). Identified ligands in the anticyclone included the high-affinity and polar siderophore ferrioxamine-E and a suite of amphibactins, which are non-polar and have weaker binding affinity for Fe (Bundy et al., 2018). In the cyclonic eddy, ferrioxamine-E was the most abundant siderophore, but total siderophores averaged over 0-150 m were much lower (0.3 pM) compared to the anticyclonic eddy (2.6 pM). Indeed, siderophore concentrations in the cyclonic eddy DCM (0.13 pM) were 15-fold lower than at a corresponding density in the anticyclone (2.0 pM at 250 m), perhaps reflecting uptake of these compounds following uplift into the euphotic zone (Figure 7a). The differences in dFe and siderophore abundances in the cyclonic and anticyclonic eddies suggest that their turnover times in the upper ocean are fast relative to the lifetimes of these eddies (estimated to be 100 days or more).
To estimate the rate of siderophore cycling, parallel incubations were also performed on deck with 2 nM ferrioxamine B and Fe-amphibactin D added to DCM waters from both eddies. After 5 days of incubation at similar light and temperature, concentrations of ferrioxamine B and Fe-amphibactin D in waters from the cyclone DCM decreased by 61% and 86%, respectively (Figure 7b). However, ferrioxamine B added to a 0.2 μm filtered control experiment also decreased by 49%, indicating that part of the loss in the unfiltered experiments was due to abiotic factors (e.g., hydrolysis, chelate dissociation, or adsorption to bottle walls). In contrast, 95% of added Fe-amphibactin D was recovered in the filtered controls, indicating that its removal from unfiltered incubations was primarily due to biological activity. Similar levels of amphibactin D removal (88%) were also found in the anticyclonic eddy, where greater ferrioxamine B removal was observed (90%). These experiments provide evidence that amphibactins represent a readily bioavailable Fe source to the microbial community, which is consistent with the uptake inferred in the cyclonic eddy DCM based on isopycnal comparisons (Figure 7a). Yet, amphibactins may not be bioavailable to the entire microbial community: addition of 2 nM Fe-amphibactin D did not increase 3 H-leucine incorporation in the cyclonic eddy DCM, despite the increase observed with addition of unchelated Fe ( Figure S3 in Supporting Information S1). Overall, there is a need to constrain the bioavailability and turnover times of these siderophores, especially the roles of heterotrophic uptake and abiotic degradation that could limit their longevity below the euphotic zone.

Prochlorococcus Fe Limitation in the Cyclonic Eddy?
The physical perturbation of the anticyclonic eddy can be conceptualized as an Fe fertilization to the DCM (Figure 1). By the time this eddy was sampled, the DCM had repositioned near the 0.5 mol photon m −2 day −1 isolume, with residual dFe concentrations that were significantly greater than observed at Station ALOHA (Figures 3  and 5). Yet, this apparent increase in dFe was not matched by increased primary productivity (by 14 C uptake) or 3 H-leucine incorporation relative to measurements at Station ALOHA (Figures 5 and 6). Additional evidence for a significant biological response in the lower euphotic zone (e.g., O 2 accumulation or increases in sinking organic matter flux) is also lacking in this eddy (Barone et al., Under Revision). The simplest explanation for these observations is the absence of a bioavailable nitrogen source that would allow biomass and/or productivity to increase beyond typical values, but this implies that any effect of iron on DCM productivity first requires an adequate N supply, likely ruling out proximal Fe-limitation under non-eddy conditions.
Simultaneous N and Fe inputs did occur in the center of the cyclonic eddy, which hosted increased primary production near the DCM. Increased nutrient delivery to DCM isolumes is consistent with the emergence of picoeukaryotes, which can grow rapidly with a NO 3 − source, and the decline of Prochlorococcus (Figure 4), often considered to grow solely on reduced N sources (e.g., ammonia, urea, and amino acids; Moore et al., 2002). Despite this conception, genomic analyses have found that genes for nitrate reductase are widespread in LL1 Prochlorococcus ecotypes, whose abundance peaks near the DCM at Station ALOHA (Berube et al., 2016(Berube et al., , 2019Casey et al., 2007;Malmstrom et al., 2010). Thus, increased NO 3 − supply may still allow an increase in Prochlorococcus biomass.
Relatively high iron requirements for Prochlorococcus photosynthesis may make them less competitive when NO 3 − supply increases. Experiments under DCM conditions with a low-light adapted Prochlorococcus (MIT1214 strain, LL1 ecotype) have indicated that the onset of Fe limitation is associated with an Fe:C ratio of 30-40 μmol:mol (Hawco et al., 2021), which is similar to the HLI Prochlorococcus strain MED4 when grown under low light (∼45 μmol:mol; Cunningham & John, 2017;Shire & Kustka, 2015). In the center of the cyclonic eddy, removal of 100 pM dFe was associated with the uptake of 2 μM NO 3 − . Assuming a C:N ratio of 6.6, the corresponding Fe:C ratio, 8 μmol:mol, is below these Prochlorococcus Fe:C thresholds, suggesting that NO 3 -dependent growth in the cyclonic eddy would lead to Fe limitation. In contrast, some eukaryotic phytoplankton can grow at low irradiance with an Fe:C ratio below 5 μmol:mol (Maldonado & Price, 1996;Marchetti et al., 2006;Strzepek et al., 2012), likely because eukaryotic cells are able to support larger chlorophyll antennae than Prochlorococcus, decreasing the required number of photosynthetic reaction centers at low light (Bibby et al., 2003;Hawco et al., 2021;Strzepek et al., 2019). Lower photosynthetic Fe requirements among eukaryote phytoplankton could therefore allow NO 3 − drawdown without Fe limitation. While this reasoning applies mostly to the intensification stage of the cyclonic eddy where heightened export occurred (taking place prior to our observations; Barone et al., Under Revision), persistent and steep NO 3 − gradients imply that diffusive mixing maintained a similar Fe:NO 3 supply ratio during the mature phase of this eddy (∼7.6 μmol Fe: mol C when converted with a 6.6 C:N ratio; see Table 1), which would support the continued growth of well-adapted picoeukaryotes without forcing Fe limitation.

10.1029/2021GB007112
13 of 18 The primary evidence for Fe stress within the cyclonic eddy comes from greater leucine incorporation when iron was added to multi-day incubations, especially in bottles receiving ambient light. 3 H-Leucine incorporation assays have a long history of tracing heterotrophic production, but rates in the lower euphotic zone can increase significantly when exposed to ambient light due to increased uptake and incorporation by phototrophs (Church et al., 2006). At Station ALOHA, Prochlorococcus can account for 60% of leucine incorporation under ambient light despite being outnumbered by heterotrophic bacteria. In dark incubations, the Prochlorococcus contribution decreases to ∼20% (Björkman et al., 2015). Therefore, the Fe stimulation of leucine incorporation in the cyclonic eddy could reflect enhancement of Prochlorococcus activity that was not apparent (or statistically resolvable) in bulk 14 C uptake. At 100 m, the relative Fe stimulation effect was much greater in the light (∼50%) than the dark (16%), consistent with the magnitude of Prochlorococcus-driven incorporation at Station ALOHA (Björkman et al., 2015). Because NO 3 − uptake and reduction via the nitrate reductase enzyme is expected to increase Fe requirements (Raven, 1988), Prochlorococcus leucine incorporation may be motivated by acquisition of reduced nitrogen rather than organic carbon (Björkman et al., 2015;Duhamel et al., 2018). Stimulation of Prochlorococcus metabolism is also consistent with their abundance in these incubations: 16S rRNA gene amplicon sequencing of +Fe and control treatments did not provide evidence for emergence of a new population of possibly Fe-limited heterotrophic bacteria (Figure 6c), suggesting that the leucine incorporation signal is driven by a population that was already dominant (e.g., Pelagibacter, Prochlorococcus).
However, there are also potential experimental artifacts for the multi-day incubations that can complicate comparisons to the surrounding seawater. At a fixed depth, these incubations provide an incomplete representation of the dynamic light fields caused by inertial waves and other physical processes, which result in vertical oscillations with periods of hours to days (Letelier et al., 1993). Bottled phytoplankton are also isolated from numerically rare and vertically migrating grazers. Relief from grazing pressure over several days could support increases in autotrophic biomass and explain why the large productivity maximum at 100 m in the multi-day incubations was not found in multiple 12-hr 14 C productivity measurements in the same eddy. We speculate that if isolation from protistan grazers and zooplankton allowed the biomass of some phytoplankton to increase, it would also have reduced the recycling of biomass Fe needed to sustain the growth of Prochlorococcus, enabling the observed Fe stimulation of leucine incorporation.
Indeed, we were unable to find strong evidence for Prochlorococcus Fe stress under background conditions. Like other phytoplankton, the onset of Fe limitation in Prochlorococcus in culture is associated with a decrease in the Chl:C ratio (Hawco et al., 2021;Sunda & Huntsman, 1997). In both eddies, there is a similar relationship between PAR and derived Prochlorococcus Chl:C, calculated as the sum of divinyl chlorophyll a and b, which are both specific to Prochlorococcus, using a uniform mol C cell −1 conversion (Figure 8). This comparison indicates that Prochlorococcus photoacclimation is not strongly impacted by the low dFe in the cyclonic eddy. We also searched for known Fe stress markers from transcriptomes sampled in the DCM throughout the Lagrangian observation period. Iron limitation is expected to up-regulate expression of isiB, which encodes the electron carrier flavodoxin that substitutes for Fe-containing ferredoxin in photosynthetic electron transport, the latter encoded by the petF gene (Bibby et al., 2003;Thompson et al., 2011). Expression of isiB was slightly increased in the cyclonic eddy compared to the anticyclonic eddy, while petF was slightly decreased ( Table 2). The resulting ∼2fold increase in the isiB:petF ratio in the cyclonic eddy follows the direction anticipated by Fe stress, but falls short of the >10-fold change expected from culture studies of both high-and low-light adapted Prochlorococcus strains (Thompson et al., 2011). Similarly, the small (17%) increase in expression of the idiA gene, encoding the periplasmic Fe binding component of the iron ABC transporter, in the cyclonic eddy matches the direction but not the magnitude identified in culture (>500%). Of these genes, only petF (ferredoxin) met significance criteria used to identify differential expression between the two eddies (p = 0.046, Table 2). Given the muted response of these genes in the cyclonic eddy DCM, it seems possible that Fe stress only emerged in Prochlorococcus during long-term incubations.
Overall, this analysis suggests that the relatively high Fe requirements of Prochlorococcus under low irradiance may make them more sensitive to Fe limitation in the DCM than eukaryotic phytoplankton, even if Prochlorococcus was not Fe-limited outside of the multi-day incubations. It is interesting that this scenario is opposite to that outlined in the surface waters of the Equatorial Pacific, where Prochlorococcus tend to be less vulnerable to Fe limitation than eukaryotic phytoplankton (Cavender-Bares et al., 1999;Mann & Chisholm, 2000;Price et al., 1994). There, the small size of Prochlorococcus makes them more competitive for diffusive supply of recycled Fe, which may be largely unchelated (Morel et al., 2008). Bioassays and molecular studies have suggested that eukaryotic phytoplankton may be able to access a much larger portion of the dFe pool than Prochlorococcus (Coale et al., 2019;Kazamia et al., 2018;Lis et al., 2015;Maldonado & Price, 1999. Low irradiance near the DCM also makes Fe-chelates less prone to photodegradation or Fe photoreduction than tropical surface waters (Barbeau et al., 2001), likely increasing reliance on specific transport mechanisms rather than uptake systems based on inorganic Fe. In the on-deck incubations under DCM conditions, added Fe amphibactin D decreased markedly after 5 days in cyclonic eddy waters (Figure 7b), indicating that these compounds were bioavailable and that they cycle rapidly in the lower euphotic zone. It is interesting to note that similar removal was also apparent in the relatively Fe-rich anticyclone. Compared to the cyclone, the anticyclone DCM contained a similar number of heterotrophic bacteria but fewer picoeukarytoes and more Prochlorococcus (Figure 4), suggesting that the observed drawdown of siderophores may not depend on phytoplankton alone.

Implications for Iron Limitation in the Oligotrophic DCM
Taken together, these results highlight the potential for Fe stress within the microbial community in the cyclonic eddy center, which ultimately did not manifest in Fe limitation of primary production. It is noteworthy that dFe measured in the cyclonic eddy was almost identical to dFe in the lower euphotic zone at Station ALOHA (Table 1; Fitzsimmons et al., 2015). However, at similar irradiance, the cyclonic eddy supported a nearly two-fold increase in primary production relative to mean rates at Station ALOHA ( Figure 5), which increased further in the long term incubations ( Figure 6). Therefore, low dFe (50-80 pM) does not necessarily prevent increases in primary production under low light conditions. Fe limitation of eukaryotic phytoplankton was recently reported for the DCM in the Eastern North Pacific (Hogle et al., 2018), especially close to the California coast where shoaling of isopycnal surfaces leads to greater NO 3 − supply to the lower euphotic zone (2.9-9.1 μM at the DCM), but similar dFe (0.05-0.1 nM). Other incubations at similarly high NO 3 − , but higher dFe (>0.2 nM), have shown signs of Fe-light co-limitation, but no response to Fe addition alone (Hopkinson & Barbeau, 2008;Johnson et al., 2010). From this perspective, it is likely that the cyclonic eddy did not reach a critical rate of N supply to trigger Fe limitation for the entire phytoplankton community. Relative to eddy lifetimes on the order of 100 days or more, the short-term nature of sampling and experiments may limit this conclusion to the mature phase of these eddies. Depending on the relative kinetics  Thompson et al. (2011). b Shire and Kustka (2015). Compared to long term observations at Station ALOHA, however, the 2017 cyclonic and anticyclonic eddies were clearly anomalous, bracketing the extremes of sea-level anomaly and the displacement of isopycnal surfaces (Barone et al., Under Revision). The fact that isopycnal shoaling and enhanced NO 3 − supply in the 2017 cyclonic eddy was insufficient to induce Fe limitation of primary production means that the possibility for Fe limitation under non-eddy conditions should be relatively narrow. Similar conclusions are reached based on retrospective analyses of the HOT program Church et al., 2009). Episodes of negative sea-surface height anomaly at Station ALOHA coincide with negative isopycnal NO 3 − anomalies in the lower euphotic zone and with positive anomalies in chlorophyll, productivity and oxygen (Table S1 in Supporting Information S1; Barone et al., 2019). These anomalies are best explained by processes also observed in the MESO-SCOPE cyclonic eddy center in July 2017: isopycnal uplift into the euphotic zone enables primary production and NO 3 − removal. The population of picoeukaryotes also increases during these events . Thus, eddy-driven NO 3 − drawdown in the lower euphotic zone has been observed on several instances, suggesting that the amount of dFe that is brought into the euphotic zone is sufficient for NO 3 − uptake. Anticyclonic eddies and other episodes of positive SLA corr at Station ALOHA also support our observations from 2017: surface dFe injection into the lower euphotic zone following a depression of the thermocline does not lead to elevated productivity or O 2 concentration .

Conclusions
The pair of eddies sampled in the 2017 MESO-SCOPE project depict distinct NO 3 − and dFe supply regimes at the depths and isolumes that characterize the DCM ( Figure 5, Table 1). In the cyclonic eddy, prior removal of NO 3 − and dFe, presumably during the intensification stage, was evident in isopycnal anomalies. Low dFe:NO 3 − supply ratios from turbulent mixing also characterized the mature phase of the cyclone. The magnitude of these ratios are associated with Fe stress in Prochlorococcus, but may be sufficient for the picoeukaryote population found in this eddy. Compared to Station ALOHA, atypically high primary production near the 0.5 mol photon m −2 day −1 isolume in the cyclonic eddy was supported by elevated NO 3 − but similar dFe, implying that any regulating role for iron is secondary to nitrogen and light ( Figure 5). Although unconfirmed by direct measurements, population control by grazers also appears to play a key role in recycling dFe to sustain this elevated productivity throughout the mature phase of these eddies. However, there are still key aspects of Fe budgets in the lower euphotic zone that are not constrained or considered by this work and likely influence the ecosystem susceptibility to iron stress, especially the magnitude and lability of particulate Fe inventories and the bioavailability of dFe. For now, we can only highlight the apparently rapid alterations of the siderophore pool as an indicator for a fast and complex iron cycle near the DCM (Figure 7).
Meanwhile, the unusual abundance of dFe in the anticyclone DCM should have created conditions to relieve Fe-stress. In some ways, this conceptual model is consistent with elevated Prochlorococcus abundance in the anticyclonic eddy, and with signs of Fe stress in leucine uptake experiments in the cyclone, but not the anticyclone. However, the fact that 14 C primary production profiles in the anticyclone overlapped with Station ALOHA mean values implies that any Fe fertilization effect must be small in magnitude or constrained by depletion of other nutrients (e.g., nitrogen). To the extent that the biogeochemistry documented within the MESO-SCOPE eddies and at Station ALOHA represent the North Pacific Subtropical Gyre and oligotrophic regions elsewhere, our results suggest that Fe limitation is difficult to induce in the DCM. However, given its apparently higher Fe requirements, one might look for Prochlorococcus in the DCM as a bellwether of Fe stress that would precede the emergence of Fe limitation at the ecosystem scale.

Conflict of Interest
The authors declare no conflicts of interest relevant to this study.

Data Availability Statement
Data from this manuscript can be accessed at https://doi.org/10.5281/zenodo.5064292 and https://doi.org/10.5281/ zenodo.3750468. Amplicon sequence data can be accessed at the National Center for Biotechnology Information (NCBI) under sequence read archive PRJNA733670.