Persistent equatorial Pacific iron limitation under ENSO forcing

Projected responses of ocean net primary productivity to climate change are highly uncertain1. Models suggest that the climate sensitivity of phytoplankton nutrient limitation in the low-latitude Pacific Ocean plays a crucial role1–3, but this is poorly constrained by observations4. Here we show that changes in physical forcing drove coherent fluctuations in the strength of equatorial Pacific iron limitation through multiple El Niño/Southern Oscillation (ENSO) cycles, but that this was overestimated twofold by a state-of-the-art climate model. Our assessment was enabled by first using a combination of field nutrient-addition experiments, proteomics and above-water hyperspectral radiometry to show that phytoplankton physiological responses to iron limitation led to approximately threefold changes in chlorophyll-normalized phytoplankton fluorescence. We then exploited the >18-year satellite fluorescence record to quantify climate-induced nutrient limitation variability. Such synoptic constraints provide a powerful approach for benchmarking the realism of model projections of net primary productivity to climate changes.

Projected responses of ocean net primary productivity to climate change are highly uncertain 1 . Models suggest that the climate sensitivity of phytoplankton nutrient limitation in the low-latitude Pacific Ocean plays a crucial role [1][2][3] , but this is poorly constrained by observations 4 . Here we show that changes in physical forcing drove coherent fluctuations in the strength of equatorial Pacific iron limitation through multiple El Niño/Southern Oscillation (ENSO) cycles, but that this was overestimated twofold by a state-of-the-art climate model. Our assessment was enabled by first using a combination of field nutrient-addition experiments, proteomics and above-water hyperspectral radiometry to show that phytoplankton physiological responses to iron limitation led to approximately threefold changes in chlorophyll-normalized phytoplankton fluorescence. We then exploited the >18-year satellite fluorescence record to quantify climate-induced nutrient limitation variability. Such synoptic constraints provide a powerful approach for benchmarking the realism of model projections of net primary productivity to climate changes.
Model projections of global marine net primary productivity (NPP) for the year 2100 range from increases to declines by as much as 20% (ref. 1). These divergences largely centre on differences at low latitudes and result from inter-model variations in the underlying mechanisms regulating NPP [1][2][3] . Such uncertainties are important as they will lead to cascading impacts on ecosystems, ocean biogeochemistry and carbon cycling in models. Climate-regulated NPP changes of several per cent already occur naturally at interannual scales, driven by the impact of the El Niño/Southern Oscillation (ENSO) in the equatorial Pacific [5][6][7] . These changes in NPP are believed to be driven by variations in the upwelling rate of deep-water nutrients that controls their supply to phytoplankton 5,6,8 , with uncertainties in future projections strongly linked to changes in growth-limiting nutrients 3 .
Assessing limiting nutrients, their geographic boundaries and their variation over ENSO cycles is key to understanding contemporary ocean NPP and informing climate model projections 4 . In that way, historical variability can act as an 'emergent constraint' on Earth system model projections 7 . Available methods for determining nutrient limitation have so far centred on ship-based observations, including simple environmental predictions based on nutrient concentrations, biochemical signals linked to nutrient stress, and growth limitation tests using biomass changes following nutrient amendment [9][10][11][12] . These different approaches deliver unique information from population to community levels that are not directly comparable 9,13 . Furthermore, even with coordinated programmes and the highest throughput approaches 10,12 , all are ultimately restricted to documenting snapshots in time and space that are of limited utility in probing large-scale temporal changes. Attempts have been made to scale nutrient-related phytoplankton fluorescence properties to global remote sensing [14][15][16] , which would offer unparalleled advantages in spatial and temporal data availability, but a range of uncertainties have restricted usage 15,[17][18][19][20][21] . Here we tackle these challenges for the tropical Pacific by building an observational dataset of nutrient limitation spanning physiological to community-level scales. Moreover, by mechanistically connecting these ecophysiological signals with coincident measurements of radiometric quantities, we directly link nutrient limitation to satellite observations, to assess large-scale variability in nutrient limitation and constrain ocean models.

Contrasting nutrient-limited physiology
Field observations were conducted along an approximately 5,000-km-long transect across the tropical Pacific Ocean ( Fig. 1 and Extended Data Fig. 1). Nutrient-addition bioassay experiments and nutrient-stress biomarker proteins both revealed a consistent trend from iron (Fe) to nitrogen (N) limitation, which matched the prevailing nitrate gradient (Fig. 1a-d). Specifically, where nitrate concentrations were elevated (Fig. 1a,b), Fe amendment stimulated chlorophyll a biomass increases (sites 2 and 3; one-way analysis of variance (ANOVA) α = 0.05, Tukey's honestly significant difference test). Conversely, where nitrate was low at sites 4 and 5, supply of N stimulated small chlorophyll a biomass increases ( Fig. 1c and Extended Data Fig. 2), and the in situ abundances of several proteins reflecting N stress (urea ATP-binding cassette transporters, the signal transduction protein P-II, the global N regulator NtcA, and aminotransferase) 22 for Prochlorococcus-an important contributor to phytoplankton biomass (Extended Data Figs. 3 and 4)-all increased ( Fig. 1d and Supplementary Tables 1-3). In the bioassay experiments, an array of diagnostic pigments responded similarly to chlorophyll a, suggesting that most of the phytoplankton community experienced the same nutrient-limitation regime (Fig. 1e). A level of N-Fe co-stress was also identified at the predominantly N-limited sites 4 and 5, where the Fe stress biomarker flavodoxin was detected 23 , N + Fe supply further boosted phytoplankton pigment biomass accumulation in comparison to N alone (Figs. 1c,e and Extended Data Fig. 2), and net accumulation of Prochlorococcus cells and increases in chlorophyll a per cell were both largest following N + Fe treatment at sites 4 and 5 ( Fig. 1f and Extended Data Fig. 5).
A major shift in diel cycles of active fluorescence properties was observed across the nutrient limitation transition ('active' fluorescence denoting stimulation by high-intensity blue light flashes; Fig. 2; refs. 10,24,25). Active variable fluorescence (F v ) normalized to maximum fluorescence (F m ) showed both lower dawn and dusk values at Fe-limited sites and pronounced daytime and night-time reductions, with the latter matched by synchronous reductions in functional absorption cross-sections of photosystem II (PSII), σ PSII (Fig. 2a,b). By contrast, under N limitation, dawn and dusk F v /F m was elevated and night-time reductions were smaller (for F v /F m ) or eliminated (for σ PSII ). Such changes in diel active fluorescence properties in this region are well documented 10,24-26 and have been suggested to be due to declines in Fe-rich photosystem I (PSI) and cytochrome b 6 f components relative to PSII, which contains less Fe ( Fig. 2c; refs. 27-29). Such changes in photosynthetic components have not previously been demonstrated in the field, but would probably lead to a more reduced night-time plastoquinone pool in cyanobacteria, in turn triggering night-time energetic decoupling of pigments from PSII (Supplementary Discussion 1; refs. 10,24). Our metaproteomic analyses for field populations of Prochlorococcus demonstrated that proteins associated with PSII showed minor changes across the Fe-to-N limitation transition, whereas those associated with PSI and cytochrome b 6 f were mostly undetectable at the Fe-limited sites 1-3 but abundant at the predominantly N-limited sites 4 and 5 (Fig. 2c,d). Our data therefore provide in situ confirmation of a predicted, major nutrient-driven stoichiometric adjustment in photosynthetic components 28 and its impact on phytoplankton fluorescence properties 10 (Fig. 2c, We found that chlorophyll a-normalized active fluorescence (F /Chl active ) was an even clearer diagnostic of the nutrient limitation transition than σ PSII and F v /F m (Fig. 3a,b). Night-time F /Chl active values, calculated by normalizing in vivo night-time active fluorescence to solvent-extracted chlorophyll a concentrations, were on average 3.2-fold higher under the elevated-nitrate, Fe-limiting conditions in comparison to the low-nitrate zone (Fig. 3b). The 'additional' fluorescence per chlorophyll a produced under elevated-N, Fe-limiting conditions can derive both from a higher ratio of fluorescent PSII to minimally fluorescent PSI (Fig. 2c,d) and from Fe-stress-related pigment protein    Article complexes, which in addition to associating with PSI 30 , may accumulate to the point at which a proportion have weak or no energetic connectivity to either reaction centre 29,31-33 . Rapid (44-45 h), threefold reductions in F /Chl active following Fe supply at the higher-nitrate sites (sites 2 and 3) provided strong evidence that this signal was Fe regulated (Fig. 3a). Notably, these F /Chl active reductions following Fe supply occurred irrespective of the restricted community restructuring (Figs. 1e and 3a and Extended Data Fig. 6). Alongside indistinguishable light absorption properties of bulk phytoplankton communities between the two limitation regimes (Extended Data Fig. 7), both observations suggested that this shift was not due to possible changes in phytoplankton communities with intrinsically different fluorescence

Remote sensing nutrient limitation
Sunlight-stimulated chlorophyll fluorescence signals are observed at global, high-frequency scales by the Aqua Moderate Resolution Imaging Spectrometer (MODIS-Aqua). Although they offer unparalleled observational potential, these passively stimulated fluorescence signals are challenging to directly compare to the actively stimulated fluorescence previously discussed, because of known, high biophysical complexity of chlorophyll fluorescence under variable measurement conditions 14,15,18,35,36 . To circumvent this complexity, we additionally measured light naturally emanating from the surface ocean at hyperspectral resolution along the transect, using radiometers fitted to the bow of the research vessel. Fluorescence line height (the upwelling radiance peak height at the fluorescent wavelength, about 680 nm; Methods) and concentrations of chlorophyll a were calculated from these radiance signals, and from this, passively stimulated, shipboard F /Chl was derived (F /Chl passive ship ). We found that F /Chl passive ship was irradiance dependent, increasing with excitation energy at low light levels and then saturating at >500 μmol photons m −2 s −1 (Fig. 3c). Such a response confirms that for the high-light conditions characterizing satellite observations of sunlight-stimulated chlorophyll fluorescence (about 500-2,500 μmol photons m −2 s −1 ) the impact of increasing excitation irradiances, which will increase energy available for fluorescence, is largely balanced out by dynamic non-photochemical quenching processes, which protect phytoplankton Article from light damage but reduce fluorescence yields in the process 14,19 . These two major components of F /Chl passive thus seem to compensate above the >500 μmol photons m −2 s −1 irradiance threshold 14,15,19,37 . Subsequently, midday, light-saturated F /Chl passive ship was 3.5-fold higher in the Fe-limited part of the transect in comparison to the low-nitrate zone, quantitatively matching the changes in F /Chl active (Fig. 3a,b) and therefore independently monitoring the nutrient limitation transition (Fig. 3d).
Although various forms of satellite fluorescence yields have previously been calculated, their interpretation has been a major challenge owing to limited field constraints of F /Chl passive in the context of a myriad of potential drivers [14][15][16][17][18][19][20][21] . As for F /Chl active and F /Chl passive ship , values of a relative satellite-derived sunlight-stimulated chlorophyll fluorescence yield, F /Chl passive sat (see Methods for derivation), showed a similar approximately threefold decrease across the observed Fe-to-N limitation transition (Fig. 3d) and a consistent geographic pattern of higher values within the Fe-limited equatorial upwelling region and lower values outside ( Fig. 4a and Extended Data Fig. 7). In addition to reduced fluorescence yields under low-N conditions, nutrient-replete conditions also decrease the pool of energetically uncoupled pigment protein complexes 14,29 . Consequently, whereas elevated F /Chl passive sat was a general characteristic of the equatorial Pacific (Fig. 4a), the broader-scale pattern revealed by the satellite data was that of distinctly lower values in the core of the upwelling zone nearer the Equator, where supply rates of both N and Fe to surface waters are elevated 38 and reduced phytoplankton Fe stress has been observed 25 (Fig. 4a,b). with chlorophyll a anomalies was weaker than that with ΔSST (R 2 = 0.38, P = 9.7 × 10 −25 ), implying some decoupling between Fe-limitation levels and chlorophyll a biomass, presumably due to the latter being under additional grazing and/or photoacclimation control. Overall, for 2002-2021, we found that greater upwelling supplied more Fe (ref. 38), reducing Fe limitation 25 and increasing chlorophyll a biomass. Notably, no evidence was found for ΔF /Chl passive sat declines at the highest ΔSST, which would be expected to accompany an overall switch into N-limited conditions under stronger stratification. Our analysis of satellite observations therefore further showed that Fe limitation in the Niño3 region remained robust to the ENSO extremes encountered over the >18-year MODIS observational period. In the coming decades, ΔF /Chl passive sat will enable observation of the nutrient-limitation response to possible stronger ENSO events comparable to the 1997/1998 El Niño 8 as well as climate change [1][2][3][4][5][6][7] .

Fe limitation persists over ENSO
In addition to providing a synoptic view of ocean nutrient limitation to upscale field observations in the tropical Pacific, the parallel variations of SST and our F /Chl passive sat metric provide an emergent constraint for nutrient limitation in ocean and climate models. When the satellite Fe-limitation diagnostic was compared with output from a model driven by realistic ENSO dynamics (Methods), we found that spatial patterns in Fe limitation were well reproduced (Fig. 4e). However, the model demonstrated a twofold oversensitivity in the magnitude of Fe-limitation fluctuations to ENSO-driven SST changes (Fig. 4f). This suggests that the natural strength of phytoplankton Fe limitation in the equatorial Pacific is more stable in response to physical changes than models predict. This mismatch may highlight the importance of both Fe recycling 39,40 and restructuring of a diverse phytoplankton community 26,41 in modulating NPP responses to altered upwelled Fe supply rates. For example, a period of progressively lower Fe supply rates probably selects for phytoplankton with intrinsically lower Fe requirements and/or optimized Fe uptake, thus maintaining productivity to a greater extent than predicted within models lacking this flexibility. Better representation of such processes within ocean models in the future, benchmarked by synoptic observations of nutrient limitation such as those presented here, will lead to more realistic projections of ocean NPP to climate change.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-023-06439-0.

Research cruise and sample collection
Observations and experiments were conducted on the RV Sonne (cruise SO267/2) from 28th January to 14th February 2019. Near-surface (about 2 m) seawater was collected through a towed, trace-metal-clean sampling system equipped with acid-washed tubing and a peristaltic pump. All sampling was conducted using trace-metal-clean techniques within a laboratory environment that was maintained over-pressurized with HEPA-filtered air.

Nutrient and trace element concentration analyses
Samples for determination of dissolved inorganic macronutrients (nitrate, phosphate and silicic acid) and dissolved Fe were 0.2 μm filtered (AcroPack1000 0.8/0.2 μm filter capsule, Pall). Macronutrient samples were collected into acid-washed 15-ml polypropylene vials and stored frozen at -20 °C until analysis in a laboratory on land. Samples were thawed over 24 h in a refrigerator before determination of concentrations using an autoanalyser (QuAAtro, Seal). Samples for dissolved Fe analyses were collected into acid-washed 125-ml low-density polyethylene bottles (Nalgene) and acidified under a laminar flow hood using 150 μl concentrated hydrochloric acid (10 M; Fisher Optima grade). After >6 months, samples were analysed following pre-concentration using inductively coupled plasma mass spectrometry (Element XR), adhering to the procedure of ref. 43 except that a Preplab instrument (PS Analytical) was used for pre-concentration and standard addition was used for determination of concentrations. Analysis of GEOTRACES intercalibration standard GSP91 in the same analytical run yielded concentrations matching previous determinations (Fe mean ± standard deviation = 0.166 ± 0.048, n = 6; https://www.geotraces.org/ standards-and-reference-materials/) 44,45 .

Chlorophyll a and phytoplankton community structure
Samples for chlorophyll a analysis (100 ml) were filtered onto glass-fibre filters (Macherey-Nagel), extracted in 90% HPLC-grade acetone and measured on a calibrated fluorometer (Trilogy, Turner Designs) following the method of ref. 46. Samples for diagnostic phytoplankton pigments (2-3.5 l; experimental samples from pooled treatment replicates) were filtered onto glass-fibre filters (Macherey-Nagel) and stored frozen at -80 °C. Extraction, analysis and peak picking (Chromeleon, Thermo Fisher Scientific) followed the procedure described in ref. 26. Diagnostic phytoplankton pigments were converted to estimated contributions of different phytoplankton types using CHEMTAX 47 , using starting pigment ratios from ref. 48. Samples for flow cytometry analysis (2 ml) were fixed with paraformaldehyde (1% final concentration) and stored at -80 °C before analysis using a FACSort flow cytometer (Beckton-Dickinson) and gated and enumerated using CellQuest Pro software (Beckton-Dickinson) following procedures described in ref. 49 (see Supplementary Fig. 1 for an example of the gating strategy).  Fig. 3b).

Underway active fluorometry and absorption measurements
The absorption properties of discrete seawater samples were determined using a point-source integrating-cavity absorption meter. Samples were collected from both the towed trace-metal-clean sampling system and underway flow-through system. The measurement procedure was identical to that described in ref. 51 except that the calibration was made using a solid standard instead of the nigrosin solution 52 . The total absorption of sampled water was measured for unfiltered and filtered (0.2 μm) samples. The values for spectrally resolved (400-710 nm) absorption from filtered samples was subtracted from those for unfiltered samples to yield particulate absorption. Line height absorption was calculated and converted to approximate chlorophyll a concentration following ref. 53, which was then used to normalize particulate absorption spectra (Extended Data Fig. 7).

Metaproteomics
Sample collection. Samples for microbial metaproteomics were collected through direct filtration of 100-120 l of seawater diverted from the trace-metal-clean sampling system. Sample collection was always at local night-time and directly after collection of bioassay experimental seawater. Seawater was pumped through two in-line 142-mm filters (3 μm Versapor and 0.2 μm Supor 200 housed in Sartorius filtration holders). Following filtration, both filters were immediately folded, placed in separate 10-ml cryovials, soaked in RNAlater (Sigma-Aldrich) and stored at -80 °C.
Protein extraction. Analyses followed methods described in ref. 54. Proteins were extracted using a modified magnetic bead method from ref. 55. Filter sections were placed in 15 ml of protein extraction buffer (50 mM HEPES pH 8.5, 1% SDS in HPLC-grade water). Samples were heated at 95 °C for 10 min and shaken at room temperature for 30 min. Filters were removed and protein extracts were filtered through 5.0 μm Millex low-protein-binding filters. Samples were then centrifuged; supernatant was removed from the pellet and 6-7.5 ml of each sample was transferred to a Vivaspin 5K MWCO ultrafiltration unit. Protein extract was concentrated to approximately 350 μl, washed with 1 ml of lysis buffer and transferred to a 2 ml ethanol-washed microtube. Vivaspins were rinsed with small volumes of protein extraction buffer to remove all concentrated protein, and all samples were brought up to 430 μl with extraction buffer. A 30 μl volume was set aside for total protein quantification and DNA analysis.
Protein quantification. Standard curves were generated using albumin standard (Thermo Scientific). Total protein was quantified after extraction and after purification with 2 μl of sample in duplicate using the bicinchoninic acid method (Thermo Scientific Micro BCA Protein Assay Kit). Absorbance was measured on a Nanodrop ND-1000 spectrophotometer (Thermo Scientific). Protein purification and digestion. SpeedBead magnetic carboxylate-modified particles (GE Healthcare) were prepared according to ref. 55. A 20 μl volume (20 μg μl −1 ) of magnetic beads was added to 400 μl of extracted protein sample. Samples were heated at 37 °C periodically to avoid precipitation. Samples were acidified to pH 2-3 by adding 50 μl of 10% formic acid. Twice the volume (1,100 μl) of acetonitrile was immediately added. Samples were incubated at 37 °C for 15 min and then at room temperature for 30 min. Samples were placed on a magnetic rack, incubated for 2 min, and supernatant was removed and discarded. Samples were washed two times removing and discarding supernatants with 1,400 μl of 70% ethanol for 30 s on the magnetic rack. A 1,400 μl volume of acetonitrile was added to each sample for 30 s on the magnetic rack. Supernatant was removed and discarded. Samples were air dried for approximately 4 min until acetonitrile had just evaporated. Samples were removed from the magnetic rack, and beads were reconstituted in 90 μl of 50 mM HEPES pH 8.0. Purified protein was quantified as described above. Trypsin (Promega) dissolved in HEPES pH 8.0 at a concentration of 0.5 μg μl −1 was added to samples at a 1:25 trypsin-to-protein ratio and incubated at 37 °C overnight.
Peptide recovery and preparation. Acetonitrile was added to digested peptides at a concentration of ≥95% and incubated for 20 min at room temperature. Samples were then placed on the magnetic rack for 2 min and supernatant was removed and discarded. A 1,400 μl volume of acetonitrile was added to samples on the magnetic rack for 15 s. Supernatant was removed and discarded. Samples were air dried for approximately 4 min, just until acetonitrile was evaporated. Beads were reconstituted in 90 μl of 2% dimethylsulfoxide and incubated off the rack at room temperature for ≥15 min. Samples were centrifuged slowly and briefly at a relative centrifugal force of 900 to remove liquid from the tube walls. Samples were incubated on the magnetic rack for 15 min and supernatant containing peptides was transferred to a new ethanol-washed 1.5-ml microtube. This step was repeated to ensure removal of all magnetic beads. Then, 1% trifluoroacetic acid was added to samples for a final concentration of 0.1%. Samples were zip tipped with Pierce C18 tips (Fisher) according to the manufacturer's protocol with a final resuspension in 25 μl of 70% acetonitrile, 0.1% formic acid. Samples were evaporated to approximately 10 μl in a DNA110 Speedvac (ThermoSavant). Samples with lower protein concentrations were further evaporated to minimize acetonitrile percentage in final resuspension-zip tip product to be less than 30% of total final buffer B volume. Samples were finally resuspended to a peptide concentration of 1 μg μl −1 in buffer B (2% acetonitrile, 0.1% formic acid).
Mass spectrometry. Metaproteomic samples were analysed by liquid chromatography-mass spectrometry (Michrom Advance HPLC coupled to a Thermo Scientific Fusion Orbitrap mass spectrometer with a Thermo Flex source). A 1 μg quantity of each sample (measured before trypsin digestion) was concentrated onto a trap column (C18 Reprosil-Gold, Dr. Maisch GmbH) and rinsed with 100 μl of 0.1% formic acid, 2% acetonitrile, 97.9% water before gradient elution through a reverse-phase C18 column (C18 Reprosil-Gold, Dr. Maisch GmbH) at a flow rate of 500 nl min −1 . The chromatography consisted of a nonlinear 100-min gradient from 2% to 95% buffer B, with buffer A being 0.1% formic acid in water and buffer B being 0.1% formic acid in acetonitrile (all solvents were Fisher Optima grade). The mass spectrometer was set to carry out mass spectrometry scans on the Orbitrap (240,000 resolution at 200 m/z) with a scan range of 380 m/z to 1,280 m/z. Tandem mass spectrometry was carried out on the ion trap using data-dependent settings (top speed, dynamic exclusion 10 s, excluding unassigned and singly charged ions, precursor mass tolerance of ±3ppm, with a maximum injection time of 150 ms).
Metaproteomic data analysis. The raw mass spectra files were searched using SEQUEST HT within Thermo Proteome Discoverer 2.2 software using a parent ion tolerance of 10 ppm and a fragment tolerance of 0.6 Da and allowing up to 1 missed cleavage. Oxidation and acetyl dynamic modifications and carbamidomethyl static modifications were included. Percolator peptide spectral matching was used within Proteome Discoverer with a maximum Delta Cn of 0.05 and a decoy search validation based on posterior error probabilities. Processed files were then loaded into Scaffold 5.0 (Proteome Software Inc.) using prefiltered mode with a protein threshold of 1.0% false discovery rate, a peptide threshold of 0.1% false discovery rate and a minimum of 1 peptide for analysis.
The Prochlorococcus nutrient stress and photosystem proteins shown in Figs. 1d and 2d were extracted from the resultant annotated metaproteomics dataset; details are provided in Supplementary Table 1. In cases for which there was more than one detection of the same protein function attributed to Prochlorococcus (same annotation) in the total exclusive spectral counts (probably reflecting mapping of measured tryptic peptides to distinct metagenome contigs with sequence variants), the sequence with the highest counts was used. Only proteins with ≥3 exclusive spectral counts for at least one of the sites were used. A subsequent peptide-level analysis was conducted using METATRYP V2.0 (https://metatryp.whoi.edu; ref. 56), and demonstrated that the peptides found in each protein were largely unique to the Prochlorococcus taxonomic classification (Supplementary Table 1). The data reported here are in unnormalized total exclusive spectral counts, which as described in ref. 57, can be useful in avoiding biases associated with changes in biological diversity across gradients (note that a subsequent normalization of the data for visualization in Figs. 1d and 2d was applied to scale variability in protein abundances across all sites to between 0 and 1). Normalizations according to total Prochlorococcus exclusive spectral counts did not alter the interpretation of these large metaproteome signals. In addition to total exclusive spectral counts, total non-exclusive spectral counts for each of the proteins were inspected to assess for sequence variations across the transect, which could map onto a different annotation despite having the same function (Supplementary Table 2). This analysis showed that total spectral count trends for the highest count proteins showed either much lower values or largely the same trends as the exclusive spectral count trends. To further investigate whether the low abundance or absence of photosystem I and cytochrome b 6 f proteins found in the high-nitrate, Fe-limited region (sites 1-3) was due to strain-level differences in sequences that could have restricted their detection relative to low-nitrate strains, we carried out a further peptide analysis using METATRYP V2.0 (https:// metatryp.whoi.edu; ref. 56) to assess for Prochlorococcus strain matches for peptide sequences (Supplementary Table 3). This analysis showed that all peptides matched to Prochlorococcus strains isolated in the equatorial Pacific, in addition to strains isolated in other regions.

Nutrient-addition bioassay experiments
Bioassay experiments conducted at sites 2-5 directly followed previously published protocols 26 . Acid-washed polycarbonate 1-l bottles (Nalgene) were filled with trace-metal-clean seawater from the sampling system described previously. Initial bottles were used to characterize initial conditions, control bottles were incubated without treatment, and treatment bottles were spiked with full factorial combinations of the nutrients N, Fe and Co. Supplementary N+Fe+Zn and N+Fe+Si treatments were also conducted to assess for potential Zn or Si serial limitation or co-limitation alongside N and Fe. All controls and treatments were conducted in triplicate. The N spike was a combined treatment of nitrate + ammonium (final amended concentration of 2 μM nitrate + 1 μM ammonium). Subsamples from incubations used for determination of macronutrient concentrations indicated that treatments with supplied nitrate were not depleted to <1 μM nitrate in any experiment over the 48-h incubation duration. Fe, Co and Zn were each added to a final added concentration of 2 nM. Silicic acid was added to a final amended concentration of 2 μM. Macronutrient spikes were previously passed through a column of prepared cation exchange resin to remove contaminating trace elements (Chelex-100, BioRad). Trace element spikes were prepared from 99+% purity solid standards dissolved in 0.01 M hydrochloric acid (Fisher Optima grade diluted in Milli-Q deionized water). Treated and control bottles were Parafilm-sealed, placed in a clear sample bag, and incubated in on-deck incubators that were continuously flushed with near-surface seawater and screened with blue screening (Blue Lagoon, Lee filters) for about 48 h. Following incubation, experiments were ended and each individual replicate was subsampled for photophysiology (fast-repetition-rate fluorometer), chlorophyll a, flow cytometry and macronutrients. The remaining samples from each triplicate replicate of a treatment were combined and filtered for HPLC pigment analysis. Calculation of chlorophyll a-based net growth rates (μ Chl ) in Fig. 1c used the equation ln(Chl Treatment /Chl Control )/t, in which t is the duration of the experiment (2 days). The control rather than initial chlorophyll concentration was used for this calculation owing to the consistent chlorophyll reductions between initial and controls that was assumed to be a result of photoacclimation to a higher-light environment in the deck incubators over in situ conditions (Extended Data Fig. 2).

Shipborne radiometric quantities
Two sets of hyperspectral radiometer systems were used to collect radiometric quantities on the research cruise. Each set consisted of a TriOS RAMSES-ACC hyperspectral cosine irradiance radiometer that measured incoming solar irradiance E D (λ) and two TriOS RAMSES-ARC hyperspectral radiance meters. The radiance radiometers measured bulk ocean surface-leaving radiance L sfc (θ sfc , Φ, λ) and sky-leaving radiance L sky (θ sky , Φ, λ). A custom-made frame held the radiance radiometers at fixed viewing angles at the bow of RV Sonne. The zenith viewing angle of the sky-facing radiometers was θ sky = 45°, and the nadir viewing angle of the sea-surface-facing radiometer was θ sfc = 45°. Each set was separated by an azimuthal angle of Φ = 45° from the ship's heading. Irradiance meters were attached to the railing at the top of the mast at least 1 m apart pointing directly upwards. Hyperspectral radiometric quantities were recorded at 5-min intervals from the ultraviolet (320 nm) to near-infrared (950 nm) regions. A correction for sea-surface-reflected glint was applied to derive water-leaving radiance (L W ) and remote-sensing reflectance (R RS ) as proposed in previous studies 58,59  Chlorophyll a (Chl) concentrations were derived from R RS using the OC3 algorithm 60 . Although variability in chlorophyll a concentrations was relatively minor throughout the cruise track, comparing shipboard radiometry-derived OC3 and filtered and extracted chlorophyll a concentrations collected at similar time points yielded a good relationship between both methods (R 2 = 0.57, P = 2.8 × 10 -7 , n = 33; Supplementary  Fig. 2). FLH values were normalized to chlorophyll a values to generate F /Chl passive ship . Dots in Fig. 3c represent 25-min rolling averages, and in Fig. 3d, data were extracted for 12:00-15:00 local time for each day to approximately match MODIS-Aqua overpass timing (solid line indicates the mean and dots represent individual observations).
Satellite remote sensing MODIS-Aqua level 3 normalized FLH (nFLH), chlorophyll a concentration (OCX algorithm) and SST data products were downloaded at about 9 km (1/12 of a degree) spatial resolution as standard mapped images from the NASA (National Aeronautics and Space Administration) Ocean Colour website (https://oceancolor.gsfc.nasa.gov). The nFLH product is calculated using normalized water-leaving radiances, with the normalization accounting for the viewing geometry of the ocean surface with respect to incident solar radiation 14,21,61 . Passive chlorophyll a fluorescence signals were observed to be at fully stimulated plateaux at the incident irradiances at the time of the satellite overpass (>500 μmol photons m −2 s −1 ; Fig. 3c). Therefore, we followed ref. 21 and carried out a first-order correction of nFLH values to remove the normalization, by multiplying nFLH values by the cosine of the solar zenith angle calculated at each pixel latitude for the mid-time point of the respective composite L3 image 21 . This conversion led to little change in trends of chlorophyll-normalized fluorescence values in the low-latitude equatorial Pacific cruise region focused on here (however, it completely changed distributions and seasonal trends at higher latitudes; Extended Data Fig. 8). It is important to note that the generality of our FLH versus irradiance observations made in the low-latitude Pacific remains to be tested in other regions, such that the latter satellite observations should be treated with caution (for example, any regions away from the tropical Pacific in regions within Fig. 4e and Extended Data Fig. 8). We also note that a conversion of nFLH back to FLH is essentially equivalent to multiplying the nFLH values by the instantaneous PAR (iPAR) at the time of image capture, as carried out in ref. 14 to correct for the predicted (now verified here for the tropical Pacific) impact of non-photochemical quenching on nFLH 14 .
Satellite observations of FLH were divided by satellite-derived chlorophyll a concentrations to retrieve F /Chl passive sat . We also excluded pixels with chlorophyll a concentrations <0.1 or >0.4 mg m −3 , which bounds the estimated lower FLH detection limit 18,62,63 as well as more uncertain chlorophyll a concentration retrievals 64 and upper levels where the relationship between chlorophyll a and phytoplankton light absorption become increasingly nonlinear 65,66 . Within this chlorophyll a range, the relationship between chlorophyll a concentrations and phytoplankton light absorption is indistinguishable from linear 66 ; therefore, an attempt to convert chlorophyll a to absorption is not necessary in this case (and could possibly introduce error due to the impact of high chlorophyll a values on chlorophyll a versus absorption equations developed using datasets from the global ocean 66 ). For the time series analyses of equatorial Pacific F /Chl passive sat (that is, Fig. 4c,d,f), we used the Niño3 box where F /Chl passive sat was validated, and where chlorophyll a consistently falls within the applied chlorophyll a thresholds (in >90% images <5% pixels fall outside this range; including these pixels does not change trends). In Fig. 3d, F /Chl passive sat pixels were averaged over 100-km-radii circles around each ship track sampling point, to reduce noise in signals derived from individual pixels 63 . To generate a F /Chl passive sat range comparable to the Fe limitation term in biogeochemical ocean models in Fig. 4e,f, the range in F /Chl passive sat was scaled to range between 0 (no Fe limitation) and 0.78 (maximum Fe limitation predicted by the used model; see next section). The upper value of F /Chl passive sat used to scale to the model limitation range was taken as the F /Chl passive sat threshold where the fraction of pixels above this value from a global-scale monthly composite reached an asymptote (F /Chl passive sat = 0.7 Wm −2 sr −1 μm −1 [mg Chl m −3 ] −1 ; varying this threshold by 10% had <10% impact on the twofold difference in slopes between the model and observations; Fig. 4f).

Global biogeochemical ocean model
The Fe limitation term (L Fe ) from a state-of-the-art global biogeochemical model NEMO-PISCES 67 was compared with spatial patterns and seasonality in F /Chl passive sat . PISCES has a relatively complex consideration of the ocean Fe cycle, including the dynamics of phytoplankton Fe uptake, storage and limitation, as well as the parallel processes of zooplankton recycling and scavenging 3,68,69 . The L F term is the fractional limitation of phytoplankton maximum growth rate (range 0-1), calculated from dynamic phytoplankton Fe quotas relative to emergent minimum and optimum quotas 70 . The model limitation terms were extracted from the surface layer at monthly resolution for diatoms and nanophytoplankton (the phytoplankton groups in PISCES). Weighting these terms by the carbon biomass of each phytoplankton type generated an average L Fe . In Fig. 4e,f and Extended Data Fig. 8 this average L Fe was subtracted from 1 to generate an Fe limitation term (Fe Lim = 1 - L Fe ) , so that higher values indicate greater Fe limitation (and vice versa). The model results were drawn from an online ocean model simulation forced by atmospheric reanalysis product JRA55 from 1958 to 2019 under the OMIP2 protocol. The model is forced at a 6-h time frequency over three cycles between 1648 and 2019 71 .

Statistics
Statistics and calculations were conducted using R. Type II linear regressions (ranged major axis) in Fig. 4c,d,f were conducted using the lmodel2 package.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
Biogeochemical data have been deposited in Zenodo (https://doi. org/10.5281/zenodo.8059552) 72  Reporting Summary Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.
nature portfolio | reporting summary

April 2023
Data Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A description of any restrictions on data availability -For clinical datasets or third party data, please ensure that the statement adheres to our policy Note that full information on the approval of the study protocol must also be provided in the manuscript.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Ecological, evolutionary & environmental sciences study design
All studies must disclose on these points even when the disclosure is negative.

Study description
Field sampling and experiments were conducted onboard the RV Sonne in January/February 2019 (SO267/2). Experimental: Seawater was collected under trace-metal-clean conditions using a towed water sampling device (~2 m depth) and filled in 1 L acid-washed polycarbonate bottles (Nalgene). Triplicate amendments of nutrients (see Methods section for full details) were performed and were incubated for 2 days. Additionally, three bottles were incubated with no amendment (controls) and three were sampled for initial conditions. Nutrient, trace element and phytoplankton community structure samples were collected alongside experimental seawater. Following incubation, bottles were sub-sampled for chlorophyll-a concentrations, flow cytometry cell counts, fast repetition rate fluorometry and diagnostic pigments (pooled samples from triplicate replicates). Underway sampling: Seawater was collected under trace-metal-clean conditions using a towed water sampling device (~2 m depth) for macronutrients, dissolved iron, chlorophyll-a concentrations, flow cytometry cell counts, fast repetition rate fluorometry, diagnostic pigments, and metaproteomics (5 sites only). Radiometric quantities were recorded continuously via hyperspectral radiometers at the bow of the ship.