Heme b distributions through the Atlantic Ocean: evidence for “anemic” phytoplankton populations

Heme b is an iron-containing cofactor in hemoproteins that participates in the fundamental processes of photosynthesis and respiration in phytoplankton. Heme b concentrations typically decline in waters with low iron concentrations but due to lack of field data, the distribution of heme b in particulate material in the ocean is poorly constrained. Here we report particulate heme b distributions across the Atlantic Ocean (59.9°N to 34.6°S). Heme b concentrations in surface waters ranged from 0.10 to 33.7 pmol L−1 (median = 1.47 pmol L−1, n = 974) and were highest in regions with a high biomass. The ratio of heme b to particulate organic carbon (POC) exhibited a mean value of 0.44 μmol heme b mol−1 POC. We identified the ratio of 0.10 µmol heme b mol−1 POC as the cut-off between heme b replete and heme b deficient (anemic) phytoplankton. By this definition, we observed anemic phytoplankton populations in the Subtropical South Atlantic and Irminger Basin. Comparison of observed and modelled heme b suggested that heme b could account for between 0.17–9.1% of biogenic iron. Our large scale observations of heme b relative to organic matter provide further evidence of the impact of changes in iron supply on phytoplankton iron status across the Atlantic Ocean.

Iron constitutes one of the most important nutrients for phytoplankton 1 because it is a component of biomolecules, termed cofactors, that participate in fundamental metabolic processes such as photosynthesis, respiration, nitrogen and sulphate assimilation 2,3 . The iron-containing cofactors are grouped based on the coordination and chemical bonds of iron with other elements within the molecule, and one major group of iron cofactors are hemes 4 .
Hemes are iron-containing porphyrins of different structures that act as cofactors (i.e. prosthetic groups) in hemoproteins 5 , and are produced via the tetrapyrrole synthesis pathway 5 . Ambient iron concentrations control heme biosynthesis and thus intracellular heme concentrations 4 . Heme b (iron protoporphyrin IX) is considered the most common heme structure within an organism 6 and is a constituent of the b type cytochromes, catalases, peroxidases, cytochrome P450, globins and nitrate reductase 7,8 . Hence, heme b is involved in electron transport and catalysis of hydrogen and other peroxides as well as oxygen control, oxygen-storage and oxygen-transport 8 . Some marine bacteria are known to use dissolved hemes in seawater as a direct source of iron [9][10][11] . In cyanobacteria, rhodophytes and cryptophytes, hemes are further metabolised to phycobilins via heme oxygenase 12 . Phycobilins function as chromophores in the light-harvesting phycobiliproteins and the photoreceptor phytochrome 12,13 .
Hemoproteins make up approximately 40% of the intracellular iron pool in phytoplankton, with hemoproteins containing heme b contributing towards approximately half of this amount 14 . Honey, et al. 15 demonstrated that heme b accounts for between 1 to 40% of the total biogenic iron pool in marine phytoplankton (mean 18 ± 14%). At low iron concentrations in culture media (≤0.50 nmol L −1 ), 6 to 26% of the available iron was utilized for production of heme b cofactors for species from temperate ocean areas 15 . This percentage ranged from 0.2 to 16% for species from high latitude ocean regions 16 . These findings suggested that, similarly to iron-use efficiencies 14 , the heme b quota required for growth is also variable amongst species and growth conditions 15,16 .
Iron deficiency in higher organisms is expressed by decreases in hemoprotein levels (e.g. hemoglobin). This condition is commonly described as "anemia" which literally refers to heme-iron deficiency. In a similar manner, reduced iron supply in the surface waters in the ocean leads to anemic phytoplankton. Hence, in both cultured and field phytoplankton populations, the concentrations of heme b tend to fall below 1.0 pmol L −1 under low iron (≤0.5 nmol L −1 ) conditions 15,16,35,36 . However, certain eukaryotes (Phaeocystis, Chaeotoceros, Rhizosolenia spp.) appear able to maintain growth despite the low heme b concentrations 16,36 . These species are considered to regulate and reduce intracellular heme b concentrations by allocating the available iron away from the hemoprotein pool in order to maintain other metabolic processes 16,36 . Heme b regulation is considered a response of phytoplankton to declining ambient iron concentrations that depends on species-specific requirements in iron and thus may be observed in the field during the iron-induced shifts 37 from larger-to smaller-sized phytoplankton populations 36 .
In this study we present an extensive dataset of heme b abundance in the Atlantic Ocean. We synthesise previously published 15,16,36 and new field data covering areas from the subpolar North Atlantic to the subtropical South Atlantic. Since the factors driving the heme b distribution in the natural environment are still uncertain, our aim was to examine physical, chemical and biological processes that potentially influence heme b. The Atlantic Ocean is a good study region for this purpose as it consists of several provinces that exhibit contrasting abundances of iron and macronutrients. In particular, the Atlantic Ocean is an area of scientific interest as it is subject to large spatial variability in iron supply 38,39 with potential impacts on primary productivity, nitrogen and phosphorus cycling [40][41][42][43][44][45] , and carbon export 46 . Furthermore, low iron supply in the Atlantic Ocean has been connected to reductions in primary productivity 47,48 , oceanic CO 2 uptake rates by phytoplankton 48 and nitrogen fixation rates 41 with potential consequences for the global climate 49 .
Finally, taking into account the laboratory experiments that showed that heme b makes up a significant portion of the total biogenic iron pool in marine phytoplankton, heme b could potentially provide an assessment of iron utilization in situ. Hence, the second part of our study examines (1) the comparability of field heme b concentrations to predicted heme b from the total biogenic iron pool using a global biogeochemical model, and (2) the potential utility of heme b as an indicator of the magnitude of the biogenic iron pool in the field.

Results and discussion
Study region. We compiled newly analysed (unpublished) and previously published data from the Atlantic Ocean collected over a time span from 2005 to 2016. We present here new data (total of 306 data points for heme b) from the research cruises M121 (2015) and M124 (2016) in the South Atlantic. We further obtained published data (total of 668 data points for heme b) for the research cruises CD173 16 , D346 15 , D361 50 and GEOVIDE 36 . Details on the expeditions are listed in Table 1 in chronological order and a map of the sampling areas and stations is shown in Fig. 1. These research expeditions and thus datasets cover several biogeochemical provinces of the Atlantic Ocean. For statistical analysis, data interpretation and discussion, we grouped the sampled stations into sub-regions according to the geographical location and the chlorophyll a (chl a) distribution. Here we list these regions in geographical order (north to south); Labrador Sea, Irminger Basin, Iceland Basin, Celtic Sea, Subtropical North Atlantic Gyre, Tropical North Atlantic, Coastal Tropical North Atlantic, Tropical South Atlantic, Angola Current, Subtropical South Atlantic Gyre and Benguela Current. particulate organic carbon and chlorophyll a. Particulate Organic Carbon (POC) concentrations ranged from 0.14 to 63.5 μmol L −1 (median = 3.1 μmol L −1 , n = 1137) for all oceanographic regions from the surface down to 220 m depth. The highest concentrations of POC were determined in the high latitude North Atlantic Ocean (Labrador Sea, Irminger Basin, Iceland Basin; median = 5.6 μmol L −1 , n = 203) and the coastal areas (Celtic Sea, Coastal Tropical North Atlantic, Angola Current, Benguela Current; median = 8.0 μmol L −1 ,   Fig. S1) indicated that heme b was typically higher in the SML (SML: median = 1.23 pmol L −1 , n = 502) and decreased with depth (Wilcoxon rank sum test, p < 0.01) except for the cases where a DCM was present (Labrador Sea, Celtic Sea and Benguela Current). The median concentration of heme b in the SML is shown in Fig. 2D. Statistically significant differences were observed in the spatial distribution of heme b in the SML for the various oceanographic regions (Kruskal-Wallis, H = 206, p < 0.01) (Fig. 3C). Significantly higher concentrations were observed in the SML of the coastal regions (Celtic Sea, Coastal Tropical North Atlantic, Angola Current, Benguela Current), ranging from 0.10 to 33.7 pmol L −1 (median = 4.1 pmol L −1 , n = 85). These regions include two upwelling areas (Coastal Tropical North Atlantic and Benguela Current) and the Congo river plume located in the Angola region; hence, heme b was highest in the Benguela area (median = 6.2 nmol L −1 , n = 14) (Fig. 4) followed by the Coastal Tropical North Atlantic (median = 4.2 nmol L −1 , n = 14) ( Table 2). In addition, heme b was also enhanced in the highly productive subpolar areas (Iceland Basin, Greenland Shelf and Labrador Sea) which were sampled during the spring bloom in 2014 (range 0.16-4.6 pmol L −1 , median = 2.2 pmol L −1 , n = 48) 36 . However, the Irminger Basin deviated from this    Fig. S1) suggest that both the spatial and the vertical distribution of heme b concentrations in the Atlantic Ocean were strongly influenced by biomass, as observed previously 15 . However, heme b was shown in previous studies to vary due to phytoplankton regulation strategies that occur as adaptation responses to iron-limitation 16,36 , thus weakening the overall correlation of heme b relative to POC.
In order to examine the changes of heme b relative to biomass as indicated by POC and chl a, we calculated the ratios of heme b:POC and heme b: chl a 16,35 . The heme b:POC ratio represents the heme b per unit carbon derived from the total particulate organic matter and previous field studies showed that this ratio varied as a result of iron availability 15,16,35,36 . The ratio heme b: chl a corresponds to organic matter originating from phytoplankton, but because of intracellular chlorophyll variability, can also be influenced by light-and nutrient-driven changes in the photosystem.
In this study, heme b: POC ranged overall from 0.01 to 6.2 μmol mol −1 (median = 0.39 μmol mol −1 , n = 907). However, the difference in values between the two layers (above and below the SML) was statistically significant (Wilcoxon Rank Sum test, p < 0.01), with heme b: POC significantly higher in the DCM. The depth profiles ( Supplementary Fig. S1) showed that the increase in heme b: POC in the DCMs was accompanied by an increase in the chl a:POC suggesting photoacclimation of phytoplankton to low light and higher investment in the photosynthetic proteins 15,54,55 . Statistically significant differences were observed for the SML for the various oceanographic regions in heme b:POC (Kruskal-Wallis, H = 153, p < 0.01) (Fig. 3D). The lowest values of heme b:POC were determined in the Irminger Basin (SML: median = 0.03 μmol mol −1 , n = 17) and the Subtropical South Atlantic Gyre (SML: median = 0.08 μmol mol −1 , n = 27) suggesting a decoupling of heme b from POC in these regions (Fig. 2E). Similarly, we observed statistically significant differences among the oceanographic regions for the heme b:chl a ratio (Kruskal-Wallis, H = 146, p < 0.01) (Fig. 3E). Heme b:chl a ratios ranged overall from 0.07 to 327 mmol mol −1 (median = 7.3 mmol mol −1 , n = 942) and followed a different pattern of spatial distribution in the SML compared to heme b:POC (Fig. 2F). Hence, heme b: chl a was lowest in the SML of the Irminger Basin (SML: median = 0.20 mmol mol −1 , n = 15) and highest in the Subtropical South Atlantic Gyre   www.nature.com/scientificreports www.nature.com/scientificreports/ high iron concentrations (≥0.30 nmol L −1 ), indicating iron limitation resulting from the high iron requirements of the extant phytoplankton population.
In this study, we report a similar pattern of low heme b concentrations (<1 pmol L −1 ) and low heme b:POC ratios (<0.1 μmol mol −1 ) in the eastern and in the central Subtropical South Atlantic Gyre, which point to iron-limited phytoplankton communities in these areas. Our observations are confirmed by bioassay experiments that showed the eastern boundary of the Subtropical South Atlantic Gyre was nitrate-iron co-limited during the period of the cruise 45 . The good agreement between the heme b measurements and the bioassay experiments of Browning, et al. 45 reinforces previous comparisons with other approaches for mapping iron-limited phytoplankton communities 36 that included the dissolved iron:nitrate ratio 56 , modified Si * tracer 57-59 and satellite-derived quantum yield of fluorescence Φ sat 60,61 . Heme b measurements thus appeared to successfully map iron limited phytoplankton by depicting the momentary condition of the phytoplankton cells in situ.
We constructed a histogram of all our SML heme b:POC data in order to investigate overall trends in Atlantic Ocean phytoplankton populations. The distribution of the heme b:POC data ( Supplementary Fig. S3A) was skewed to the right with a median value of 0.44 μmol mol −1 (n = 468) and a mean value of 0.47 ± 0.42 μmol mol −1 (n = 468). Furthermore, both the Irminger Basin and the Subtropical South Atlantic Gyre exhibited an exponential distribution and their medians (0.03 μmol mol −1 and 0.09 μmol mol −1 respectively) deviated from the median of the distribution (Supplementary Fig. S3A). Taking into account laboratory studies and field observations of heme b: POC ratios and their relationship to iron limitation, we suggest that 0.10 µmol mol −1 is a reasonable estimate for defining an anemic, iron limited phytoplankton community in the Atlantic Ocean. Hence, the data appearing in the left tail (break 0.10 μmol mol −1 , n = 50) (Supplementary Fig. S3A) of the distribution point to iron limited phytoplankton communities. Thus, by this definition, the southern stations of Tropical South Atlantic were iron limited and iron (co-)limitation is predicted to extend to the central parts of the subtropical gyre (Fig. 2, cruise M124, sampling February-March 2016).
In contrast, the western boundaries of the Subtropical South Atlantic Gyre were characterized by slightly higher heme concentrations (1.36 pmol L −1 ) and heme b:POC ratios (0.37 μmol mol −1 ) compared to the eastern and central Subtropical South Atlantic Gyre. These results suggest that this area was not iron limited; indeed Rijkenberg, et al. 38 reported iron concentrations up to 6.1 nM in the upper 800 m off shore of Brazil in the Subtropical Shelf Front (STSF), which is formed by the southward flowing Brazil Current and the norward flowing Malvinas Current 62 . The iron enrichment was attributed to aeolian deposition and transport by the STSF, and to offshore export of iron from Brazilian shelf waters and the Rio de la Plata river 38,62 .
The heme b:POC values of the open ocean stations in the Tropical North Atlantic, (SML; median = 0.26 μmol mol −1 ) do not point towards iron-limited phytoplankton communities, even though heme b concentrations were low (SML; median = 0.59 pmol L −1 ); we attribute this trend to low biomass at the time of sampling. In general, the location of the Intertropical Convergence Zone (ITCZ) defines the dust derived iron supply to surface waters 49 and thus influences the biogeochemical status and bloom progression between the northern and the southern oligotrophic waters around the Equator 63,64 . In February-March 2011, when sampling was performed, Schlosser, et al. 63 reported that the ITCZ was at ∼1°N and the Tropical North Atlantic was receiving significant amounts of atmospheric deposition whilst the Tropical South Atlantic had low atmospheric dust concentrations. In addition, Snow, et al. 64 reported increased nitrogen fixation rates due to the abundance of Trichodesmium sp. between 15°N and 7°S which have high iron requirements relative to non-nitrogen fixing phytoplankton.
Heme b regulation strategies?. Similar to the heme b:POC ratio, the distribution of the population for the heme b:chl a ratio ( Supplementary Fig. S3B) was also skewed to the right with median value of 7.77 mmol mol −1 and a mean value 8.98 mmol mol −1 (n = 491). However, we observed a contrasting behaviour in the ratios of the two iron limited areas Irminger Basin (median SML = 0.20 mmol mol −1 , n = 20) and Subtropical South Atlantic Gyre (median SML = 12.7 mmol mol −1 , n = 28), despite the similar trend in heme b:POC. The medians of heme b:chl a for these two areas deviated from the overall median of the distribution and located either on left (Irminger Basin) or the right (Subtropical South Atlantic Gyre) tail. Heme b exhibited the same trend for the two areas which implies that changes in chl a quotas drove the differences in the ratio. Hence, we ascribed this contrast to different extant phytoplankton groups, to different photoacclimation and nutrient status.
The low heme b:chl a observed in Irminger Basin implied that the heme b containing proteins of the photosynthetic apparatus decreased whilst chl a was conserved. In culture, several diatoms and prymnesiophytes exhibited decreased heme b:chl a ratio under low iron conditions (<0.50 nmol L −1 ) 15,16 implying allocation of iron away from the hemoprotein pool 16 . This behaviour was considered an adaptation strategy of phytoplankton that would allow a reduction of the overall iron requirements and a more efficient utilization of the available iron in order to sustain growth 16 . Here, this pattern was also observed in the field in the Irminger Basin (GEOVIDE, May-June 2014) for diatom populations that likely employed heme b regulation 36 in order to adapt to declining iron concentrations during bloom progression. For example, allocating the iron away from the hemoproteins would lead to decline of the heme b-containing cytochromes b 6 f and b 559 of the PSII apparatus 8,18,28 , which in turn would be accompanied by increases in the chl a:PSII ratios. Indeed this was observed by Macey, et al. 65 in the post-bloom iron-limited Iceland Basin. Particularly for the eukaryotes, another strategy of reducing iron requirements is switching from nitrate to ammonium utilization [66][67][68] which induces the reduction of heme b-containing nitrate reductase.
We observed very high heme b:chl a in the Subtropical South Atlantic Gyre which was driven by low concentrations of chl a in the area (SML median = 0.03 nmol L −1 ). Furthermore, the ratio chl a:POC was 7.68 μmol mol −1 (n = 32) in that region pointing to lower photoacclimation of the dominant phytoplankton groups. At the time of the study, haptophytes and the picophytoplankton species Synechococcus and Prochlorococcus were most dominant in the eastern parts of the gyre 45 . Similar community composition characterizes the Subtropical North Atlantic Gyre 69 . Therefore, we made a comparison of the trends in chl a:POC between these two regions; Scientific RepoRtS | (2020) 10:4551 | https://doi.org/10.1038/s41598-020-61425-0 www.nature.com/scientificreports www.nature.com/scientificreports/ in the Subtropical North Atlantic Gyre, the chl a:POC was higher compared to the Subtropical South Atlantic Gyre reaching a median value of 65.0 μmol mol −1 (n = 312). The contrasting behaviour in chl a:POC can be ascribed to differences in light climate, which is known to strongly influence chl a:POC ratios 70 . Higher light conditions in the Subtropical South Atlantic Gyre in comparison to the Subtropical North Atlantic Gyre, as a result of both higher incident irradiance during winter months 71 and shallower mixed layers ( Supplementary  Fig. S1), would be expected to result in the lower chlorophyll-to-carbon ratios observed. However, the low chl a concentrations, and thus the low chl a:POC values, could also be a result of general nutrient limitation in the area as shown previously 72,73 . Hence, our results implied a high variability in chl a due to a combination of photoacclimation, nutrient limitation, phytoplankton physiology and species specific modifications in the photosynthetic apparatus.
Our heme b data from the field appear to be broadly consistent with laboratory studies in terms of magnitude and relationship with POC. Nevertheless, a key challenge in the interpretation of relationships between heme b, chl a and POC is distinguishing between variability resulting from physiological responses to changes in light and nutrient regimes, and variability resulting from changes in the relative contributions of different carbon (C) pools 35 . The relationship between biomass and POC is also critical to the interpretation of optical parameters measured from space. Studies of interrelationships between chl a, POC and phytoplankton biomass in the Atlantic Ocean have concluded that chl a:POC ratios vary to a greater extent than C biomass:POC 70,74 , with the greater variability in chl a:POC attributed to changes in phytoplankton photophysiology driven by nutrient and light regimes. Thus in analogy, we have assumed throughout our discussion that variability between heme b, chl a and POC reported here will be more strongly influenced by physiological changes than by changes in POC composition. Further support for this assumption, or alternative methods of assessing phytoplankton biomass would be useful in future applications of heme b as an indicator of iron status in the field.

Comparison of field heme b concentrations with model-based predictions.
Both theoretical estimates and direct measurements suggest heme b represents an important amount of the total cellular iron pool in phytoplankton (mean 18 ± 14%) 14,15,75 thus the second part of our study aimed to examine to what extent the heme b field data were comparable to heme b estimated from the biogenic iron pool as predicted from a global biogeochemical model. Here we used the PISCES-v2 (Pelagic Interactions Scheme for Carbon and Ecosystem Studies volume 2) biogeochemical model to estimate the biogenic iron pool which is determined based on Michaelis Menten uptake kinetics, further regulated by a maximum iron-to-carbon (Fe:C) cellular quota and enhanced iron uptake under iron limitation [76][77][78] . Figure 5A illustrates the biogenic iron (Fe bio ) pool from the model which corresponds to sum of the diatom and nanophytoplankton iron pools extracted at the same geographic position and month as the field data. Values represent the monthly means for the upper most layer of the model (depth 10 m) and ranged from 9.20 to 285 pmol L −1 (median = 78.8 pmol L −1 , n = 254). Although a direct comparison of the absolute values is questionable due to the large uncertainties associated with model predictions of the ocean iron cycle 79 , there are similar trends between the Fe bio and heme b abundance in the Atlantic Ocean. In particular, both parameters were highest in the subpolar North Atlantic (Labrador Sea, Iceland Basin), along the continental margins (Celtic Sea, Angola Current) and in the upwelling areas (Coastal Tropical North Atlantic, Benguela Current), whilst the lowest values were determined in the oligotrophic Subtropical South Atlantic Gyre.
According to phytoplankton culture experiments with different phytoplankton species, heme b accounts for between 1 and 40% of the particulate Fe (PFe) (averaging 18 ± 14%) 14,15,80 . We calculated the predicted heme b concentrations (in pmol L −1 ) from the modelled Fe bio of each station in the SML using three scenarios (values) 1%, 18% and 40% that represent the range and mean of heme b:PFe observed in laboratory cultures. We used the formula heme b = [Fe bio * [heme b ⁄ PFe] lab ]/100, where [heme b / PFe] lab denotes the proportion of heme b relative to PFe determined for cultured phytoplankton under various growing conditions. Predicted heme b concentrations were lowest for the 1% scenario and highest for the 40% scenario ( Fig. 5B) with the 1% scenario corresponding more closely to the field observations reported in this study. This trend was further supported by the ratio heme b:Fe bio calculated from the mean heme b concentrations (in the SML) and the model-based Fe bio which ranged from 0.17% to 9.1% (median = 1.8%, n = 234). Taken together with heme b:POC, these results suggest that, as far as heme b is concerned, field populations were comparable to laboratory observations for species with low heme b content. Furthermore, the ratio of 1% was determined in cultures where iron (0.50 nmol L −1 ) or nitrate supply had been exhausted, thus perhaps better representing the natural environment. Therefore, the model prediction of Fe bio and the in situ heme b concentrations were consistent with previous laboratory observations 15,16 . The differences in the proportion of heme b relative to Fe bio likely arose because of the microbial community composition of each region and the interspecific differences in hemoprotein processes.
Biogenic iron and iron-to-carbon (Fe:C) ratios are critical for linking the iron and carbon cycles in the ocean and constraining phytoplankton iron requirements within biogeochemical models 81 . Currently, the biogenic iron pool can be assessed by radioisotope uptake experiments 81,82 , single-cell synchrotron x-ray fluorescence 81,83,84 and determination of particulate iron after careful removal of or correction for lithogenic iron 81,[85][86][87] . Whilst all these techniques provide complimentary and useful information on particulate iron, single cell iron-to-phosphorus ratios or community wide Fe:C uptake rates, none provide a definitive estimate of average in-situ biogenic iron quotas within phytoplankton populations. Furthermore, none of the above methods were applied consistently on larger scale field expeditions. The good agreement of heme b observational and model-based data suggests that heme b could potentially serve as an indicator of the biogenic iron for field studies and provide an assessment of the proportion of iron used in the heme b and the hemoprotein pools in marine phytoplankton, although further comparison of heme b abundance with established methods of assessing biogenic iron are required.

conclusions
In this study, we suggest that the ratio heme b:POC can serve as a reliable indicator of iron limitation since it appeared not to be influenced by species-specific processes which potentially involve modifications in the hemoprotein and/or other protein pools. A comparison of the heme b method with already established protocols, such as bioassay experiments 43,56,88 and iron-stress biomarkers (e.g. flavodoxin, ferredoxin and IdiA) 25,31,34,89,90 , would validate this simple method for mapping iron limited phytoplankton in small or large scale field expeditions. Although determination of heme b in marine particulate material requires specialist analytical instruments, sample collection is straightforward and utilises the same approaches routinely applied for collection of pigment samples. Importantly, sampling for heme b does not require trace metal clean conditions, significantly widening the potential for assessment of iron limitation in the field. Furthermore, heme b is a ubiquitous biomolecule and is thus not dependent on the presence of particular phytoplankton species.
The variability of the heme b to biomass ratios suggested that phytoplankton 1) employ diverse mechanisms to utilize the available iron in the hemoprotein pool, and 2) are able to regulate the hemoproteins under iron limiting conditions. Expanding the field and laboratory research in hemoprotein abundance and cycling would contribute to the identification of such iron utilization and regulation strategies. One aspect could be the transcriptomic and proteomic analyses of biomolecules involved in the heme b cycling pathways. These include the proteins associated with the utilization of the heme b cofactor, such as the components of the photosynthetic or nitrate assimilation apparatus, as well as the components of the biosynthesis and breakdown pathways. The information about iron-utilization in hemoproteins and about regulation of the hemoprotein pool would shed light on the molecular response and adaptation of phytoplankton to iron limitation. This information will be particularly useful in the future due to projected alterations in desert dust supply 49,91 and thus, to iron availability in the ocean, with potential impacts on carbon sequestration and climate. Finally, application of the heme b method in oceanic regions other than the Atlantic Ocean would enhance the estimates of the magnitude of the biogenic iron pool and expand our knowledge on phytoplankton physiological status in relation to iron.

Methods
Sampling. During all research cruises sampling of seawater was performed by a stainless steel CTD rosette equipped with Niskin bottles for the determination of heme b, POC and chl a. We sampled 279 stations and 4 to 6 different depths per station (typically down to 300 m deep) in order to study heme b abundance in the Atlantic Ocean.
Heme b. Marine particulate material (>0.7 μm) was collected after filtration of seawater on glass fiber filters (GF/F, pore size 0.7 μm, MF300, Fisherbrand) for the determination of heme b. The filters were stored at −80 °C prior to analysis. Heme b was quantified by High Performance Liquid Chromatography (HPLC) -Diode Array Detection (DAD) -Electrospray Ionisation (ESI) Mass spectrometry (MS) after extraction by 2.5% w-v Octyl β-D-glucopyranoside-OGP solution (Sigma-Aldrich, ≥98% GC Grade) 80,92 . Regarding the analytical procedure and the instruments used, we followed the protocol described by Gledhill (2007) for the samples from the CD173 cruise and Gledhill (2014) for the samples from cruises D346, D361 and M124. Good agreement was obtained www.nature.com/scientificreports www.nature.com/scientificreports/ between the two analytical methods 92 . Slight modifications were applied to the analysis of the samples from GEOVIDE and M121 as described in Louropoulou et al. 36 .
Quantification of heme b for D346, D361, GEOVIDE, M121 and M124 was performed from the MS data on a m/z ratio of 616.122, since the PDA detector is known to have interferences from other pigments co-eluting with heme b and absorbing at a wavelength of 401 nm 92 . The analytical detection limits were defined as three times the standard deviation of the lowest calibration standard for each of the methods followed and were 1.57 nM (CD173), 190 pmol heme b L −1 (D346, D361 and M124) and 32 pmol heme b L −1 (GEOVIDE and M121). Heme b could not be detected in the blank extraction solution in either utilised method. particulate organic carbon and chlorophyll a. Seawater was filtered through pre-combusted GF/F for the determination of POC and acidified either with sulphurous acid -H 2 SO 3 (cruises D346, D361, M121, M124) or with hydrochloric acid -HCl (GEOVIDE) for the removal of inorganic carbon following published protocols 93,94 . POC in samples and several blank filters was quantified by elemental NC analyser using acetanilide as the calibration standard 93 .
Chlorophyll a was determined for the cruises CD173, D346, D361 and M124 via fluorometry after extraction with 90% (v:v) acetone in the dark 95 . For the GEOVIDE and M121 cruises, chlorophyll a was determined after extraction with 100% methanol and sonication followed by High Performance Liquid Chromatography quantification 96 . Data handling. Data processing, statistical analysis and visualization were carried out using R Statistical Software 97 (R Core Team 2016, https://www.R-project.org). The SML was calculated from the CTD data (obtained by sensors on the rosettes) using the temperature-based criterion (ΔT = 0.5 °C) 98 . We checked the normality of the data distribution by applying the Shapiro-Wilk test which showed a non-normal distribution for all parameters (p < 0.05). Hence, we used the Kruskal-Wallis test for analysis of variance, the Wilcoxon rank sum test to identify differences in the distributions in and below the SML and the Spearman's rank correlation to check the association between parameters. Surface and section plots were produced using the Ocean Data View 99 v.4.7.9 (Schlitzer, R. Ocean Data View. 2018).

Data availability
Heme b and supporting data are available on the British Oceanographic Data Centre, http://www.bodc.ac.uk/.