Shelf humic substances as carriers for basin-scale iron transport in the North Pacific

Iron is one of the key elements controlling phytoplankton growth in large areas of the global ocean. Aeolian dust has traditionally been considered the major external source of iron in the North Pacific. Recent studies have indicated that sedimentary iron from the shelf region of the Sea of Okhotsk has a strong impact on the iron distribution in the North Pacific, while the mechanism supporting its long-distance transport remains poorly understood. Here, we report that refractory shelf humic substances, which complex and carry dissolved iron, are transported conservatively at least 4000 km from the shallow sediments of the Sea of Okhotsk to the subtropical North Pacific with the circulation of intermediate water. This result indicates that shelf humic substances are probably one of the key factors shaping the distribution of dissolved iron in the ocean interior.

Iron (Fe) is one of the essential elements for marine life and has low solubility in oxic seawater 1,2 ; therefore, external inputs of Fe influence ocean primary productivity 3,4 . Aeolian dust, shelf sediments, and hydrothermal vents are major external sources of dissolved Fe (Fe d ), and the mechanisms that make Fe soluble and contribute to long-distance transport are vital to connecting external sources with primary productivity in remote ocean areas 3,5 . Although aeolian dust has traditionally been considered the major external source of Fe to the ocean 6,7 , shelf sediments have been noted to be much more important than aeolian dust or hydrothermal vents in terms of the percentage of the Fe d inventory in the ocean and its role in fueling the biological carbon pump 5 . The chemical species of Fe d contributing to long-distance transport from sediments are thus critical information to understand not only marine Fe cycle but also global carbon cycle.
Organic ligands increase the capacity of Fe to dissolve in seawater by complexing with Fe and possibly contribute to long-distance transport through protecting Fe d from being scavenged [1][2][3][4] . Siderophores, saccharides, and humic substances have been considered probable Fe-binding organic ligands in marine environments 4,8,9 . Among these substances, refractory humic substances are probably the most important Fe d carriers in the subsurface ocean because siderophores and saccharides are microbiologically labile 10,11 . Humic substances, which are complex and heterogeneous mixtures of organic molecules that form during the decay and transformation of biogenic remains, are highly functionalized and are generally characterized by their color due to their ultraviolet-visible (UV-Vis) absorbance 12 . As a consequence of their absorbance characteristics, humic substances exhibit fluorescence properties commonly referred to as humic-like fluorescent dissolved organic matter (FDOM H ) 13,14 . It is well established that FDOM H is universally distributed over the Earth's surface, namely, from streams to deep oceans 15,16 . Fe-binding ligands 4,17 and FDOM H [18][19][20] have been reported to be released during the microbial degradation of sinking particles, and a linear relationship was found between FDOM H and Fe(III) solubility (dissolution capacity of Fe) in subsurface waters [21][22][23][24] ; thus, FDOM H is very likely a major Fe-binding organic ligand in the dark ocean. A global ocean Fe biogeochemical model also successfully applied autochthonous FDOM H as the main organic ligand to reproduce the Fe d distribution in the ocean 25 . However, oceanographic linkage between Fe d and FDOM H has not been explored with the basin-scale.
Here, we present the distribution of Fe d together with FDOM H along a section in the western North Pacific (Fig. 1a) where basin-scale transport of sedimentary Fe d from the shelf region of the Sea of Okhotsk has been reported 26,27 (Fig. 1b). The diapycnal tidal mixing at the deep sill of the Bussol' Strait (2200 m) is known to be important to determine the physical and chemical properties of the intermediate water [28][29][30] . The levels of Fe d in the North Pacific Intermediate Water (NPIW; 26.6-27.5σθ) 31 , which is characterized by a salinity minimum in subtropical regions ( Supplementary Fig. 1a), were higher than those in the upper/deeper water masses. It has been suggested that the Fe d derived from shelf sediments in the Sea of Okhotsk is transported to the basin region by the Okhotsk Sea Intermediate Water (OSIW; 26.6-27.0σ θ ) 32 and then spreads through the circulation of intermediate water, including the NPIW, in the North Pacific 26,27 .
The lowest level of FDOM H was observed in surface waters, which was likely due to the photobleaching of FDOM H 33-35 , except in the shelf region of the Sea of Okhotsk (Fig. 1c). The levels along the transect generally increased with depth in the mesopelagic layer (200-1000 m) and then slightly decreased with depth in the deep layer (>1000 m). The distribution pattern of FDOM H was almost identical to that of AOU ( Supplementary  Fig. 1b), as previously reported [18][19][20]36,37 .
Interestingly, however, the FDOM H -AOU relationships in the mesopelagic layer and the deep layer were different (Fig. 2a). The FDOM H levels in the mesopelagic layer were higher than those in the deep layer, thus showing deviations from the linear regression line obtained for the deep layer. Similar but smaller deviations in mesopelagic FDOM H from the deep linear regression line have also been observed in the central North Pacific 18 . Because AOU represents the amount of oxygen consumed by respiration after the subduction of a water mass, the deep linear regression line has been attributed to the in situ FDOM H produced by microbes during the oxidation of organic matter [18][19][20] . Thus, the autochthonous fraction of FDOM H in the mesopelagic layer corresponds to the linear portion of the regression between AOU and FDOM H (determined for the deep layer); then, the contribution of allochthonous FDOM H , which is defined here as FDOM H *, can be estimated quantitatively (see Methods).
The distribution pattern of FDOM H * was distinctly different from that of FDOM H (Fig. 1c,d). The highest level was observed in the shelf region of the Sea of Okhotsk. The levels of FDOM H * in the OSIW and the upper intermediate water (26.6-27.0σ θ ) 31 were higher than those in the upper/deeper water masses from the Sea of Okhotsk to the south as far as 20 °N in the subtropical North Pacific gyre, corresponding to the southernmost region of the NPIW distribution 31 . FDOM H * accounted for 37 ± 7% (n = 4) and 12 ± 4% (n = 9) of the bulk FDOM H in the OSIW and the upper NPIW at 20-30 °N, respectively ( Supplementary Fig. 2). A negative linear relationship was evident between FDOM H * and salinity in the intermediate water (adjusted R 2 = 0.78, Fig. 2b), even though FDOM H was not linearly related to salinity in the intermediate water (adjusted R 2 = 0.003). Because the OSIW, which is influenced by the dense shelf water that forms in the coastal polynya through sea-ice formation involving the interaction with sediments 32 , greatly contributes to the formation of the upper intermediate water 31 (Fig. 3a). The majority of the flux of Fe d from sediments to the water column has been considered to be dominated by organic-Fe(III) complexes 43,44 . Thus, the relationship (Fig. 3a) indicates that a specific fraction of Fe d from shelf sediments occurs as Fe d complexed with FDOM H * and is transported across the North Pacific with the conservative spread of FDOM H *. This  23 . The Fe(III) solubility was obtained by measuring Fe in the soluble fraction (<0.025 µm) [21][22][23] . Because the molecular weight of FDOM H is reported to be less than 1.8 kDa 34 , quite smaller than 0.025 µm, soluble Fe can form complexes with humic substances, as indicated by FDOM H . These pieces of evidence indicate that excess Fe d compared with the solubility derived from bulk FDOM H can occur as colloidal Fe (0.025-0.22 µm), which is not complexed with FDOM H . Although the size fractionation was not determined in this study, the substantial occurrence of colloidal Fe has been observed in the intermediate water of the western subarctic Pacific 45 , which is the same water mass observed in this study.

Discussion
FDOM H *, namely, allochthonous FDOM H , is most likely supplied from sediments as stable complexes with Fe d since major forms of sediment-derived Fe d are organic complexes 43,44 . Assuming that the other Fe d preferentially forms complexes with autochthonous FDOM H in the intermediate and deep waters, the spatial distribution of Fe d concentrations (Fig. 1b) can be separated into three groups ( Fig. 4 and Supplementary Fig. 3). High concentrations of allochthonous FDOM H -Fe complexes and colloidal Fe occur in the shelf region of the Sea of Okhotsk and spread to the western North Pacific through circulation of intermediate water, including the NPIW. The allochthonous FDOM H -Fe complexes and colloidal Fe are mainly distributed in the upper and lower intermediate waters, respectively, implying that the allochthonous FDOM H -Fe complexes can make more important contributions to primary production in remote areas due to intrusion into the upper layer. Interestingly, a shift in  The transport mechanism of sedimentary Fe d reported in this study can be applied to the western Arctic Ocean, where high levels of Fe d and FDOM H are evident in dense shelf water 49,50 . It has been documented that hypoxic shallow sediments are an important source of Fe d and labile particulate Fe through the supply of Fe(II) from the sediments, oxidation to Fe(III), and chelation of Fe(III) with organic ligands or formation of inorganic Fe(III) to labile particles 43 . It has also been reported that FDOM H is produced in marine sediments even under anoxic conditions 51 . Therefore, it can be concluded that FDOM H are primary organic ligands contributing to the long-distance transport of sedimentary Fe for the whole ocean, although stable transport is limited to the dark ocean where photodegradation of FDOM H is inhibited. An application of the method used in this study to other intermediate water systems will clarify the generality regarding with the relationship between Fe(III) solubility and FDOM H as well as the role of FDOM H as a carrier of sedimentary Fe.
Apart from macronutrients, the chemical framework of the Fe cycle in the ocean has not been well established because Fe has extremely low solubility in modern seawater. The chemical properties of Fe control input from external sources and its residence time, which shape the Fe distribution in the ocean. Although organic ligands have been considered a major factor increasing Fe solubility, the exact role of organic complexation in the Fe cycle, and in fact the very nature of the ligands that stabilize soluble Fe, have been incompletely characterized. This study clearly indicates that FDOM H is a factor that controls the residence time of Fe d , at least sedimentary Fe d . Although aeolian dust has traditionally been considered a major source of Fe for phytoplankton growth in the western North Pacific, the episodic inputs of aeolian dust may not be sufficient to sustain primary productivity in the region 52  Allochthonous and autochthonous FDOM H , as ligands of Fe d , are able to be determined by salinity and AOU in the western North Pacific (Fig. 2). A global ocean Fe biogeochemical model successfully parameterized autochthonous FDOM H as the main ligand 25 . A parameterization of allochthonous and autochthonous FDOM H in the biogeochemical models may result in the accurate reproduction of the modern ocean Fe cycle and consequently ocean ecosystems and carbon cycling, which will have implications for the appropriate estimation of how climate change will affect ocean productivity 3,4 . Salinity and temperature were measured using a conductivity-temperature-depth (CTD) sensor, and dissolved oxygen (DO) concentrations were measured using an oxygen sensor connected to a CTD. The DO concentrations were also measured on board by the Winkler titration method, and the DO concentrations measured by the sensor were calibrated using the concentrations determined by the Winkler method. The oxygen solubility was calculated using the function of Weiss (1970) 53 , and apparent oxygen utilization (AOU) was then calculated as the difference between the solubility and the measured DO concentration. Seawater from the surface to bottom layers (16-29 depths) was collected with acid-cleaned Teflon-coated 10-or 12-L Niskin-X bottles that were mounted on the CTD with a carousel multi-sampling system during the R/V Hakuho Maru and R/V Professor Multanovskiy cruises. The sampling method used for seawater from the two stations (C3 and B5) located in the shelf region of the Sea of Okhotsk during the R/V Professor Khromov cruise has been described elsewhere 46 . Dissolved iron. Concentrations of dissolved iron (Fe d ) in the shelf region of the Sea of Okhotsk measured during the R/V Professor Khromov cruise were derived from previously reported data 46 . To collect a subsample from the Niskin-X sampler during the R/V Hakuho Maru (KH-12-3) cruise, the sampler was transported in a clean air bubble (filled with air that had been passed through a high-efficiency particulate air filter) and a 0.2-μm Acropak filter (Pall Corporation) was connected to the Niskin-X spigot; the filtrate was then collected in acid-cleaned 125-mL low density polyethylene (LDPE) bottles (Nalgene Co., Ltd). To collect a subsample from the Niskin-X sampler during the R/V Professor Multanovskiy cruise, the sampler was placed in a clean tent and a 0.22-μm Millipak filter (Millipore Corporation) was connected to the Niskin-X spigot; the filtrate was then collected in acid-cleaned 125-mL LDPE bottles (Nalgene Co., Ltd). We confirmed that there were no significant differences between the Fe d concentrations measured using the Acropak filter and the Millipak filter.

Methods oceanographic observations. Observations in the western
The filtrate (<0.22 μm) was adjusted to pH <2 by the addition of ultrapure HCl (Tamapure AA-10, final HCl concentration of the sample was 0.024 M) and then allowed to remain for one to three months at room temperature in the onboard clean room. Each sample was then adjusted to pH 3.2 just before its measurements by the addition of ammonium solution and a formic acid (10 M)-ammonium (2.4 M) buffer. Fe d , defined as the leachable Fe in the filtrate at pH <2, was then analyzed in the onshore laboratory using a flow injection analysis (FIA) chemiluminescence detection system 54 . All sample treatments were performed under laminar flow in the onboard or onshore clean air laboratory.
The Fe d measurements and reference seawater analyses in this study were quality controlled using SAFe (Sampling and Analysis of Iron) cruise 55 reference standard seawater (obtained from the University of California Santa Cruz for an inter-comparison study). We measured a SAFe reference sample during every sample measurement run of the FIA instrument performed in the onboard and onshore laboratories. The consensus values for Fe(III) in the SAFe reference standard seawater are 0.093 ± 0.008 nM (S) and 0.933 ± 0.023 nM (D2) (May 2013, www.geotraces.org), and we obtained values of 0.098 ± 0.010 nM (n = 12) (S) and 0.976 ± 0.101 nM (n = 10) (D2) using our method. This good agreement demonstrates that our data quality was high and that our data are comparable with the global GEOTRACES dataset. The detection limit (three times the standard deviation of the Fe(III) (2020) 10:4505 | https://doi.org/10.1038/s41598-020-61375-7 www.nature.com/scientificreports www.nature.com/scientificreports/ concentration of purified seawater (0.036 nM) that had been passed through an 8-quinolinol resin column three times to remove Fe) was 0.020 nM.

Humic-like fluorescent dissolved organic matter.
To determine the level of humic-like fluorescent dissolved organic matter (FDOM H ) in the samples obtained during the R/V Hakuho Maru and R/V Professor Multanovskiy cruises, the seawater samples from the Niskin-X sampler were poured directly into pre-combusted, triple-rinsed glass vials with Teflon-lined caps. The glass vials were thoroughly washed with Milli-Q water for their next use on board the ship. Just after sampling, the seawater was allowed to stand until reaching room temperature without undergoing any filtration procedure, and fluorescence measurements were performed with a spectrofluorometer (RF-1500, Shimadzu) with a 1-cm quartz cell. The fluorescence intensity of the FDOM H was determined at excitation and emission wavelengths of 320 nm and 420 nm, respectively, according to Yamashita et al. 37 . It was reported that the observed differences in FDOM H levels with and without filtration using GF/F glass fiber filters were negligible for the open ocean samples 37 .
Seawater samples collected at two stations located in the shelf region of the Sea of Okhotsk during the R/V Professor Khromov cruise were filtered with a 0.22-μm Millipak filter (Millipore Corporation) connected to the Niskin-X spigot and poured into acid-cleaned fluorinated high-density polyethylene (HDPE) bottles (Nalgene Co., Ltd). The filtrate was stored frozen in the dark until analysis. The frozen samples were thawed and allowed to stand until reaching room temperature; fluorescence measurements were then conducted as described above.
After the analysis, the fluorescence intensities were corrected to the area under the water Raman peak of Milli-Q water (excitation = 320 nm), which was analyzed daily with freshly prepared Milli-Q water and calibrated to Raman Units (RU 320 ) 56 . Because the instrument-specific response 57 of the spectrofluorometer (RF-1500, Shimadzu) was not corrected commercially, the instrument-specific response was corrected with the comparison of FDOM H fluorescence intensity in RU 320 obtained by an instrument-specific response-corrected spectrofluorometer (FluoroMax-4, Horiba) 58 . The conversion factor from RU 320 to commonly used Raman Units (RU; fluorescence intensity corrected by peak area of Raman scatter at 350 nm) 56 [18][19][20]36,37 . In this study, the linear relationship between FDOM H and AOU was also evident in the deep layer (>1000 m) along the 160 °E transect (FDOM H = 1.54 × 10 -5 × AOU + 2.17 × 10 -3 , n = 46, adjusted R 2 = 0.93, p < 0.01). However, many samples in the mesopelagic layer (200-1000 m) did not follow the linear relationship observed in the deep layer but exhibited deviations from this linear relationship at high levels of FDOM H (Fig. 2a) 21,23,41,42 . Such differences in these relationships are likely due to the occurrence of organic ligands (e.g., siderophores and saccharides) other than FDOM H in surface waters. Thus, using a previously published dataset 21 , the linear regression between Fe(III) solubility and FDOM H fluorescence intensity in quinine sulfate units (QSU) was determined for the deep waters (>1000 m) of the western subarctic Pacific gyre and the basin of the Sea of Okhotsk collected in 2000 during the R/V Mirai cruise (MR00) (Supplementary Fig. 4). Because the instrument-specific response of the spectrofluorometer used in the previous study 21 was not corrected, the regression equation in Supplementary Fig. 4 could not be directly applied to this study. Therefore, to determine the calibration factor between the two fluorescence units, namely, the previously reported QSU 21 and the RU 320 used in this study, we compared the FDOM H fluorescence in QSU and RU 320 using samples in the deep layer. For this comparison, two stations located in the western subarctic Pacific gyre and in the basin of the Sea of Okhotsk were selected from each cruise (Supplementary Fig. 5a). Although the observations in this study (MU14) were conducted 14 years after those of the previous study (MR00), the vertical profiles of AOU in the deep layer were almost identical between the two observations ( Supplementary Fig. 5b,c). Additionally, the linear relationship between the AOU values in the deep layer of the two cruises is evident, with a slope of almost one ( Supplementary Fig. 5d), indicating that the water mass was observed to have the same biogeochemical characteristics in both cruises, thus allowing one to make a calibration factor between RU 320 and QSU using the relationship between the FDOM H values of the deep layer observed in both cruises ( Supplementary  Fig. 6).
The conversion factor from FDOM H with units of RU 320 to Fe(III) solubility with units of nM was achieved using the slope (± a standard deviation) of two relationships, namely, H 320 The estimated value (96.2 ± 9.7) was applied as the conversion factor from FDOM H [RU 320 ] to Fe(III) solubility [nM] in this study. The bulk Fe(III) solubility and allochthonous Fe(III) solubility (Fe(III) solubility*) were estimated using the conversion factor with the fluorescence intensity of bulk FDOM H and FDOM H *, respectively.