Distinct profiles of size-fractionated iron-binding ligands between the eastern and western subarctic Pacific

Iron (Fe) is well known as a limiting factor to control primary productivity especially in high-nutrient and low chlorophyll area such as the subarctic Pacific. The solubility of Fe is believed to be controlled by its complexation with natural organic ligands, while the distribution of organic ligands is poorly understood. Here, we report that dissolved (< 0.2 µm) organic ligands were unevenly distributed between the western and eastern stations in the subarctic Pacific. The concentration of dissolved organic ligands around the lower part of subarctic Pacific intermediate water was higher in the western station, suggesting that Fe complexation with these organic ligands supports a lateral transport within the water mass. However, a more detailed size-fractionated treatment indicated no significant difference in the soluble (< 1000 kDa) ligands’ distribution between the western and eastern stations. These results suggest that organic and inorganic colloid formations are potentially essential for Fe transport mechanisms in the subarctic Pacific.


Results and discussion
Hydrography of the study area. Stations CL2, CL5, and CL16 were located in the WSG, central North Pacific (CNP), and the AG (Fig. 1). Generally, the subarctic Pacific is characterized by low salinity surface waters that result in significant stratification between surface waters and intermediate waters 6 . In surface water, concentrations of nutrients and chlorophyll a in the western station were relatively higher than those in the eastern station ( Supplementary Fig. S1, Table S1), consistent with the previous observation 6 . During our observation, the subarctic intermediate water was distributed from 150 to 1250 m depths (σ θ = 26.6-27.5) (Supplementary Table S1). A rapid decrease in the dissolved oxygen (DO) and increase in the apparent oxygen utilization (AOU) was observed below the surface mixed layer. In general, the concentration of nutrients below the surface mixed layer reflected the regeneration of organic matters ( Supplementary Fig. S1, Table S1). At Stn. CL2, relatively high concentration of nutrients was observed in the upper NPIW compared to the other stations. Considering phosphate ([PO 4 3− ]) as an example for nutrients, the regenerated and preformed phosphate could be evaluated as follows: where the fixed value (1/170) is the constant stoichiometric Redfield ratio between phosphorus production and oxygen consumption 11 ; the [PO 4 3− ] Regenerated follows the same trend as that of AOU ( Supplementary Fig. S1). In the upper NPIW density range at Stn. CL2, the [PO 4 3− ] Preformed (1.36-1.42 μM) was similar to that at Stn. CL5, but slightly higher than that at Stn. CL16 (1.19-1.33 μM). There was not clear difference in [PO 4 3− ] Preformed between sampling stations in the lower NPIW density range. These results suggested that the preformed phosphate in the upper NPIW, which included transported phosphate from the Okhotsk Sea, may have been higher in the west compared to the eastern region during our observation period.
Fe and its speciation in the dissolved fraction (< 0.2 μm). In the upper 100 m at all three stations, the [D-Fe] ranged from 0.02 to 0.09 nM as low as Fe limitation for phytoplankton growth could occur (Fig. 2, Supplementary Table S1). An increase in depth from 100 m resulted in a corresponding increase in [D-Fe], which reached a maximum toward the lower NPIW density range (1000-1250 m). The highest [D-Fe] was recorded at Stn. CL2 in the WSG. Below the maximum layer, [D-Fe] gradually decreased with depth at Stns. CL2 and CL5. However, this trend was not observed at Stn. CL16 ( Fig. 2 and Supplementary Fig. S1). Generally, the vertical profile of dissolved Fe below the surface mixed layer reflects the influence of external suppliers such as atmospheric Fe and laterally transported D-Fe (external Fe). Internally regenerated D-Fe (hereafter internal Fe) from processes such as remineralization and desorption from organic particles in the water column are also reflected in this vertical profile 6 . The previous study 6 applied the Fe* concept 12 to distinguish between external Fe and internal Fe as follows: where R Fe:P is the ratio of [D-Fe] to [PO 4 3− ] 6 . In this study, the fixed R Fe:P value (0.16 nM/μM) was applied to determine the relative values Fe* and [Internal Fe] which in turn helped to compare the influence that lateral transport has on [D-Fe] between the sampling locations; the fixed R Fe:P value was the average value of the intermediate water data from Stn. CL16, where the intermediate water may not be strongly influenced by lateral external Fe input 6 . The relationship between σ θ , AOU, [internal Fe], and Fe* indicated that dissolved Fe regeneration started to occur below the upper NPIW (Fig. 3). By contrast, an increase in Fe* occurred prominently near the lower NPIW density range, suggesting that the Fe-rich western water in the NPIW can be explained by external , which was not complexed with dissolved organic ligands (see "Methods" section), was 0.01 to 10 pM. Consequently, > 99.3% of dissolved Fe was estimated to be complexed with these natural organic ligands. High excess ligand concentrations (> 3.9 nM) with low D-K′ Fe′L (< 10 11.0 ) were detected in the lower NPIW density range at Stn. CL2. The complexation capacity of organic Fe-binding ligands (α FeL′ ), which was calculated from the product of the concentration of excess ligands ([L′], see "Methods" section) and K′ Fe′L , for dissolved fraction (D-α FeL′ ) in the water column ranged from 95 to 4850, within the reported values range (one class ligand, 0-501187) 13 . At Stn. CL2, relatively low D-α FeL′ (95-200) was found in the lower NPIW density range, resulting in high [D-Fe′] (Fig. 3). This suggests that excess ligands in the water mass do not contribute directly to increasing [D-Fe]. However, this trend was not observed at Stns. CL5 and CL16.
Previous studies have also reported excess ligands in the Pacific Ocean water column [14][15][16][17][18][19][20][21][22][23][24][25][26] . An excess of [D-Fe] relative to [D-L] has been demonstrated in deep waters (> 2000 m) 18,20 , suggesting a variation in dissolved Fe speciation between the Pacific Ocean's upper and deep waters. This variation is possibly the result of the difference in the competitive ligand exchange-adsorptive cathodic stripping voltammetry (CLE-ACSV) method used in the studies. The previous studies 18,20 used 2-(2-thiazolylazo)-p-cresol (TAC) as the competing ligand instead of salicylaldoxime (SA) used in this study (see "Methods" section). TAC is also recognized as one of the common competing ligands used to determine organic Fe-binding ligands in seawater 27,28 . However, it was suggested that using the CLE-ACSV method with TAC is inefficient for detecting Fe complexation with humic substances 29 . Indeed, dissimilarity in the CLE-ACSV method results has been reported for deep water in the Pacific Ocean 16 and the Arctic Ocean 30 . Generally, the CLE-ACSV method using 10 μM of TAC reagent underestimates ligand concentrations.  Table S1). The negative value was observed in the samples from the lower NPIW density range at Stn. CL2 and in the deep water at Stn. CL16.
On an average, 73% of the dissolved Fe was partitioned into the soluble fraction throughout the water column at Stns. CL2 and CL16 (Supplementary Table S1). For the organic ligands, the ratio of the complexation capacity between the soluble and dissolved fractions (S-α FeL′ /D-α FeL′ ) showed a broad range (16 to > 100%) (Fig.  S2). Particularly for the samples from the lower NPIW density range at Stn. CL2, S-α FeL′ overwhelmed D-α FeL′, indicating that there were no colloidal ligands with a capacity to bind new Fe in these samples (Supplementary Table S1). The saturation of organic ligands by Fe in the colloidal fraction was also detected in samples from the www.nature.com/scientificreports/ Southern Ocean 31 and the Atlantic Ocean 32,33 . Although details such as the reagent type and the applied detection window in the CLE-ACSV method were different for these studies, several parts of colloidal ligands might not be detected by our method. This study applied a relatively low detection window to increase the peak current's sensitivity 21 . Therefore, it is possible that the added Fe in the CLE-ACSV titration was absorbed onto the natural inorganic Fe colloids instead of SA in the sample during the equilibration. If so, this would result in an overestimation of the ligand concentration. Further research is required to validate the colloidal ligand concentrations obtained using the CLE-ACSV method.  33 . There was no clear relationship between the size-partitioning dissolved Fe and the ligands' complexation capacity for the study area ( Supplementary Fig. S2), consistent with the results from North Atlantic samples 33 . In this study, an anomalously high S-α FeL′ /D-α FeL′ ratio was obtained in the samples from the lower NPIW density range at Stn. CL2, reflecting the low D-α FeL′ . These results suggested that α FeL′ alone cannot explain the dissolved and soluble Fe distributions in the North Pacific and the North Atlantic.

Relationship between
As mentioned above, our method may have miscalculated several portions of organic ligands. As a result, the negative value of C-α FeL′ in the lower NPIW density range at Stn. CL2 was calculated, indicating the existence of "not organically complexed" Fe in the colloidal fraction.

The concentration of colloidal inorganic Fe ([C-Fe′]) was calculated from the equation [C-Fe′] = [D-Fe′]-[S-Fe′] to evaluate Fe speciation in the colloidal fraction.
Interestingly, there was a positive correlation between [C-Fe′] and [C-Fe] in this study area (Fig. 4). Particularly, high [C-Fe′] and [C-Fe] were observed in the lower NPIW density range at Stn. CL2, suggesting that colloidal Fe in the lower NPIW density range at Stn. CL2 contained an inorganic colloidal form and organic complexes that were not detectable by our CLE-ACSV method. Furthermore, the behavior of [D-Fe′] was partially similar to Fe* (Fig. 3), suggesting that external Fe tended to form unstable colloidal Fe in the western region. These findings are consistent to the previous study which demonstrated higher [D-Fe] over the solubility of Fe(III) hydroxide (in this case, < 0.025 μm fraction) in the deep-water column of the western Pacific (165°E) 10 .
Judging from the Fe behaviors in the western subarctic Pacific, the east-west differences in the biological production and lateral transport via NPIW formation are essential factors that control Fe distribution and chemical speciation (Fig. 5). It has been suggested that high Fe supplies via vertical mixing with intermediate waters from the Okhotsk and Bering Seas and dust input results in high productivity in the western subarctic Pacific compared to the eastern area 7,34 . The distribution of organic ligands could be influenced by multiple biological sources including the release of extracellular polymeric substances by phytoplankton, and by degradation process such as photochemical reaction in the surface water 4,28 . As a result, the Fe speciation in the surface water was similar between Stns. CL2 and CL16 during our observation period despite the east-west differences in biological productivity 34 . The high productivity in the upper waters would cause a high flux of sinking particles and partially contribute to the high regeneration of macro-and micro-nutrients, including Fe and its organic ligands. In this study, the rapid increases of AOU and [PO 4 3− ] preformed were observed in the upper NPIW at Stn. www.nature.com/scientificreports/ CL2 ( Supplementary Fig. S1). It was demonstrated that humic substance-like fluorescent dissolved organic matter, which includes some part of Fe-binding organic ligands, supported the transportation of dissolved Fe and nutrients via the intermediate water from the Okhotsk Sea 9 , indicating the relatively high [D-Fe] in the upper NPIW at Stn. CL2 was derived from the Okhotsk Sea. In the lower NPIW density range, high Fe* and [D-L] were observed in the western station, suggesting an external source of Fe and organic ligands from the Bering Sea. Since some parts of these external Fe might exist as inorganic colloidal forms and organic complexes that were not detectable by our CLE-ACSV method 21 , the unstable colloidal Fe would be partially flocculated and scavenged by sinking particles in the deep water. However, the existence of excess ligands in both size fractions throughout the water column also indicated the potential dissolution of particulate Fe via complexation, suggesting the reversible exchange of Fe mediated by colloids between the soluble, colloidal and particulate phases, especially near the lower NPIW. Taken together, further research incorporating a multiple analytical windows analysis for the CLE-ACSV 22 is still required to clarify the detail of colloidal ligands' behavior and its influence on size-exchange of Fe in this area. However, our results suggest that dissolved Fe speciation in the NPIW density range in the western subarctic Pacific has unique features relative to those in the eastern area. Considering that the NPIW density range sources are derived from the Okhotsk Sea and the East Kamchatka Current, further high-resolution observation is required to clarify the source of organic ligands and the relationship between Fe speciation and the transport mechanism in the western subarctic Pacific.

Methods
Sample collection and treatment. Seawater samples were obtained from three stations in the subarctic Pacific Ocean onboard R/V Hakuho Maru KH-17-3 (GEOTRACES GP02) cruise between June 23 and August 7 in 2017. Samples were collected using acid-cleaned Teflon coated 12-L Niskin-X bottles attached to a CTD-Carousel Sampling system suspended on a Vectran cable (Fig. 1). After the recovery of Niskin-X bottles, these bottles were placed in a clean-air booth. Seawater samples were filtered through an acid-clean AcroPak 200 Capsule filter with a 0.2 μm pore-size Supor Membrane (Pall) attached directly to the spigot with silicon tubing. Compressed clean air was then used to obtain the dissolved Fe concentrations of the samples. The seawater sam- www.nature.com/scientificreports/ ples from Stns. CL2 and CL16, which were measured for soluble Fe and natural organic Fe(III)-binding ligands, were additionally filtered using an acid-cleaned polyethylene hollow fiber filter (< 1000 kDa) 35 . These filtered seawaters were collected in acid-cleaned 500-mL fluorinated high-density polyethylene bottles (Nalgene) and low-density polyethylene bottles (Nalgene) for the analyses of organic Fe-binding ligands and Fe concentrations, respectively. Samples collected for organic ligand analysis were stored at -20 °C until analysis. Fe concentration analysis samples were acidified to pH < 1.7 with 20% quartz-distilled HCl (Tamapure AA-100, Tama Chemicals). indicates the concentration of excess ligands that were not complexed with Fe. Because K′ Fe′L occasionally differs between the dissolved and soluble fractions, we did not evaluate the ligand concentration in the colloidal fraction. Instead, we calculated α FeL′ in the colloidal fraction (C-α FeL′ ) from C-α FeL′ = D-α FeL′ -S-α FeL′ to evaluate the capacity of colloidal ligands' complexation. Furthermore, the concentration of inorganic Fe ([Fe′]), which was not complexed with natural organic ligands, was estimated from concentrations of Fe and organic Fe-binding ligands, and the K′ Fe′L in each fraction (D-Fe′ and S-Fe′).

Fe concentrations.
Other parameters. Concentrations of chlorophyll a and nutrients (NO 3 − + NO 2 − , PO 4 3− and SiO 2 ) were also sampled and measured on-board. Seawater samples for chlorophyll a analysis were immediately filtered through GF/F filters (Whatman). Chlorophyll a was extracted in 6-mL aliquots of N, N-dimethylformamide, stored at − 20 °C for over 24 h, and analyzed using a fluorometer (10-AU, Turner Design Inc.). For nutrient analysis, seawater was collected into a 10-mL polyethylene tube. The nutrients were determined by a continuous flow system (SWAAT, BLTEC Japan). DO data were obtained from a CTD sensor. The DO concentration was calibrated using automatic titrator data (DOT-15X, Kimoto Electric Co.). AOU was calculated from the dissolved oxygen, temperature, and salinity using the program Ocean Data View (https ://odv.awi.de/).

Data availability
The datasets presented in the current study are available from supplementary Table S1. All [D-Fe] data in this study is cited from Nishioka et al. 7 to calculate organic ligands data (https ://www.pnas.org/conte nt/117/23/12665 ).