Variable ageing and storage of dissolved organic components in the open ocean

Seawater dissolved organic matter (DOM) is the largest reservoir of exchangeable organic carbon in the ocean, comparable in quantity to atmospheric carbon dioxide. The composition, turnover times and fate of all but a few planktonic constituents of this material are, however, largely unknown. Models of ocean carbon cycling are thus limited by the need for information on temporal scales of carbon storage in DOM subcomponents, produced via the ‘biological pump’, relative to their recycling by bacteria. Here we show that carbohydrate- and protein-like substances in the open Atlantic and Pacific oceans, though often significantly aged, comprise younger fractions of the DOM, whereas dissolved lipophilic material exhibits up to ∼90 per cent fossil character. In contrast to the millennial mean ages of DOM observed throughout the water column, weighted mean turnover times of DOM in the surface ocean are only decadal in magnitude. An observed size–age continuum further demonstrates that small dissolved molecules are the most highly aged forms of organic matter, cycling much more slowly than larger, younger dissolved and particulate precursors, and directly links oceanic organic matter age and size with reactivity.

Seawater dissolved organic matter (DOM) is the largest reservoir of exchangeable organic carbon in the ocean, comparable in quantity to atmospheric carbon dioxide 1,2 . The composition, turnover times and fate of all but a few planktonic constituents of this material are, however, largely unknown 3,4 . Models of ocean carbon cycling are thus limited by the need for information on temporal scales of carbon storage in DOM subcomponents, produced via the 'biological pump', relative to their recycling by bacteria 3,4 . Here we show that carbohydrate-and protein-like substances in the open Atlantic and Pacific oceans, though often significantly aged, comprise younger fractions of the DOM, whereas dissolved lipophilic material exhibits up to ,90 per cent fossil character. In contrast to the millennial mean ages of DOM observed throughout the water column, weighted mean turnover times of DOM in the surface ocean are only decadal in magnitude. An observed size-age continuum further demonstrates that small dissolved molecules are the most highly aged forms of organic matter, cycling much more slowly than larger, younger dissolved and particulate precursors, and directly links oceanic organic matter age and size with reactivity 3,5 .
Seawater DOM consists of analytically identifiable biochemicals such as carbohydrates, proteins and lipids, as well as operationally defined and long-lived geomacromolecules (for example, humic and fulvic substances 5,6 ). In order to resolve some of the key details of DOM sources and cycling in the oceans, major organic components were extracted from high-molecular-weight ultrafiltered DOM 5 (DOM HMW , .1,000 daltons) collected from 1,000-3,000 l of sea water, and analysed for both D 14 C and d 13 C isotopic signatures. Samples were collected from surface mixed-layer (3-20 m), mesopelagic oxygen-minimum (850-900 m), and abyssal (1,500-1,800 m) depths in the central North Pacific (June 1999) and the Sargasso Sea region of the North Atlantic (June 2000) oligotrophic ocean gyres. The contributions of solvent-extractable lipids, protein-like and carbohydrate-like organic matter (OM), as well as different molecular-weight fractions, to the overall age structure of seawater DOM, were thus established.
By far the most highly aged DOM component was the lipid extract (6.4-17.1 kyr before present, BP; Table 1), with 14 C ages in the deep Pacific representing the greatest yet observed for any component of seawater OM. The lipid extract was considerably older by ,5-13 kyr than the total DOM HMW and unfractionated, bulk DOM (SDOM) pools (Tables 1 and 2; Fig. 1a). Furthermore, at all mesopelagic and abyssal depths, the lipid extract and DOM HMW were older in the Pacific than in the Atlantic, similar to the oceanocean offsets observed for SDOM 6,7 (Table 1) and presumably due to cumulative ageing during deep water-mass transit 8 . Conversely, mixed-layer lipid extract, DOM HMW and SDOM were all older in the Atlantic than in the Pacific (Tables 1 and 2), suggesting possible aged North American continental or atmospheric inputs there 9 . The highly d 13 C-depleted signatures of lipid extracts (Table 1, Fig. 1b) are consistent with isotopic fractionations during cellular lipid letters to nature synthesis 10 . Thus, these fossil dissolved lipids arise either from extensive ageing within the oceans, with concomitant recycling over many ocean circulation times 8 , or from inputs of one or more pre-aged lipophilic precursors 11 . Concurrent elemental ratios (Table 2) and lipid biomarker data 12 of total DOM HMW suggest, however, that the lipid extract may be dominated by planktonic, rather than petrogenic, material.
Dissolved protein-like and carbohydrate-like fractions were similar in 14 C age, ranging from modern to ,3-4 kyr BP (Table 1). Modern to near-modern ages in surface waters indicate that both fractions are derived from recent, post-bomb (that is, after ,1955) marine production, and contain little aged or recycled material. In deeper waters, however, these components, which are quite reactive in surface waters 4 , are deduced to have escaped degradation over many ocean circulation times 8 and to have aged extensively. The protein-and carbohydrate-like components were younger by as much as ,13-14 kyr compared to the corresponding lipid extract (Table 1), and by as much as ,1 kyr (Fig. 1a) compared to DOM HMW , supporting the contention that 'old' seawater SDOM 1,6,7 is actually composed of components having a spectrum of ages and reactivities. Transport-based ageing of protein-and carbohydrate-like DOM is also suggested at mesopelagic and abyssal depths (Table 1), similar to the SDOM 6,7 , total DOM HMW and lipid extract (Tables 1 and 2).
Dissolved forms of all three major organic fractions were significantly older than their particulate counterparts 13,14 (Fig. 2). Dissolved lipid extracts were ,6-17 kyr older than lipid extracts of sinking particulate OM (POM; 3,500 m depth) 13,14 , suggesting that dissolved and particulate lipids cycle on dramatically different timescales or arise from dissimilar sources. The latter possibility is supported by correspondingly lower d 13 C of dissolved lipids (about 229 to 228‰; Table 1) compared to particulate and sedimentary lipids (225‰ to 222‰; refs 13, 14; Fig. 2). Alternatively, particulate lipids may contain both a highly aged component, similar to the dissolved pool, and a recently derived component from contemporary marine production 15 , accounting for their intermediate ages.
Deep dissolved protein-and carbohydrate-like fractions were also significantly older by ,3-4 kyr than particulate forms 13,14 , whereas the relative abundances of these two fractions are reversed in the dissolved (Table 1) compared to particulate phases 13,14 , suggesting dissimilar sources or preservation. Therefore, although the dissolved fractions are far more abundant, they are also far longer-lived . Isotopic signatures of MUC were estimated by isotopic mass balance, as: where X is the D 14 C or d 13 C signature of each organic fraction, f is the relative contribution of each to the total DOMHMW pool, and fLE þ fPL þ f CL þ f MUC ¼ 1:0: ‡ Assumed from average of all observed values in that group for purposes of inclusion of D 14 C-d 13    and are thus deduced to be less reactive to bacteria than the younger particulate fractions. Dissolved carbohydrates have further been found to contain specific sugars of modern 14 C age 16 , indicating that even within a given organic fraction individual molecules may have unique cycling times, similar to findings for individual sedimentary lipids 11 . The highly modified, acid-insoluble fraction of DOM HMW (analogous to the molecularly uncharacterized, MUC, component of OM 17 ) was estimated to comprise 39-57% of DOM HMW carbon (Table 1). These large amounts of MUC-like DOM and its estimated ages (0.2-5.6 kyr BP; Table 1) suggest that much of the SDOM is composed of structurally modified biopolymers and geomolecules 17 , probably derived from combinations of diagenetically altered younger proteins and sugars and older lipid components. Based on their disparate D 14 C and d 13 C signatures (Table 1, Fig. 2), however, the dissolved forms of MUC and lipid extract are unlikely to share a common origin as has been suggested for POM 14 . Instead, dissolved MUC in deep waters is isotopically more similar to, and thus may arise from, dissolved protein-and carbohydrate-like and humic materials 6 (Fig. 2), or from a precursor common to each. This is further supported by NMR studies demonstrating that DOM HMW contains significant acyl polysaccharide, a carbohydrate-derived biopolymer 18 .
Radiocarbon ages for total DOM HMW ranged widely, from modern to 4.6 kyr BP (Table 2), and were younger than SDOM 1,6 from the Atlantic and Pacific (Table 2). Therefore, SDOM must by definition also contain an older, low-molecular-weight (LMW) component in order to balance the younger DOM HMW (ref. 19; Table 2). In addition to being the oldest size fraction yet identified for seawater OM (,1.9-6.2 kyr BP), DOM LMW is also the most abundant, comprising 77-95% of SDOM. The d 13 C signatures of this older LMW material further suggest that it arises directly from recycling of younger DOM HMW ( Table 2). Observations of old DOM LMW , intermediate-aged DOM HMW , and young POM 7,20 in the oceans reveal a pronounced size-age relationship among the major forms of seawater OM (Table 2; Fig. 2). The two main forms of POM (that is, sinking and suspended) consistently contain bomb 14 C (D 14 C range: about 2100‰ to þ160‰ for suspended POM, and about 230‰ to þ35‰ for sinking POM) throughout the Pacific and Atlantic water columns 7,20 . However, only surface DOM HMW is similarly enriched, while LMW material contains no apparent bomb 14 C at any depth ( Table 2; Fig. 2). That is, 14 C age increases consistently in the size sequence from sinking POM (the youngest, largest fraction), to suspended POM, to DOM HMW and finally DOM LMW (the oldest, smallest fraction; Fig. 2), and most probably arises from sequential hydrolysis, dissolution and/or degradation of larger forms of OM to successively smaller fractions. This also suggests that the relative rate of OM respiration slows as highly aged LMW material accumulates in the latter stages of degradation.
An important corollary of this size-age continuum is that it coincides with a previously observed OM size-reactivity continuum 3,5 , ranging from structurally complex and recently produced sinking POM and DOM HMW (most reactive to bacterial degradation) to structurally simple but highly reworked DOM LMW (least reactive to bacterial degradation). The proposed size-age model for seawater OM is therefore consistent with higher utilization rates of DOM HMW than DOM LMW in oceanic and coastal waters 3,5 as well as with the presence of younger, presumably more reactive subcomponents 3,16 in DOM HMW (Fig. 1a). Studies of SDOM degradation further support the presence of specific bioavailable subfractions of this large, heterogeneous pool 4 . Besides the 14 C evidence, highly elevated elemental ratios of DOM HMW (Table 2; ref. 21) compared to recently produced 'Redfield' OM (C:N:P ¼ ,106:16:1), together with the presence of bacterial fatty acids in oceanic DOM HMW (ref. 12), further support the contention that dissolved components may have undergone extensive recycling compared to POM (Fig. 2). An alternative to the proposed sizedependent ageing model for seawater OM is that there exist one or  Table 1 for assumed d 13 C values. more allochthonous or autochthonous sources of pre-aged DOM in the oceans that are largely independent of plankton-dominated OM formation and degradation. While aged allochthonous sources of DOM such as submicrometre fossil black carbon 22 are probably minimal on the basis of the observed OM size-age distributions (Fig. 2), other 'pre-aged' inputs may include natural hydrocarbon seepage 23 in certain ocean regions and atmospheric deposition 9 . Soluble forms of seawater OM are further predicted to escape degradation and undergo ageing either within the deep ocean, or during surface-to-deep ocean transport, by mechanisms different from those such as sorptive preservation 24 identified for particulate and sedimentary forms.
Within the oceanic DOM pool, various organic and size fractions persist on radically different timescales (Tables 1 and 2). The turnover time (TOT) of undifferentiated bulk pools such as SDOM is equivalent to its 14 C age if substances in the bulk pool are uniform in age. However, the presence of discrete organic and size fractions having different 14 C ages (and therefore different rates of turnover) results in a weighted mean TOT that departs from the SDOM age 25 . On the basis of the heterogeneity in the organic and size fraction ages (Tables 1 and 2), the weighted mean TOT for surface ocean SDOM is estimated to be ,60-90 yr, increasing to ,3,700-6,000 yr in Atlantic and Pacific deep waters, respectively (see Supplementary Table S1). Surface ocean differences between the weighted mean TOT (decadal) and SDOM 14 C ages (millennial) result from surface DOM being dominated by young protein-and carbohydrate-like fractions that are recycled rapidly compared to the balance of the DOM. This contrasts with deep waters, where SDOM ages and TOT converge owing to the uniform low reactivity and turnover of all subcomponents. A

Sample collection and organic fraction separation
Large-volume (1,000-3,000 l) water samples were collected using 30-l rosette-mounted Niskin bottles. Samples were pre-filtered (0.2-mm) and transferred to an Amicon DC-10L tangential flow ultrafiltration system equipped with spiral wound filter cartridges with a 1,000 dalton molecular-weight cut-off. Samples were reduced to ,1 l and frozen until processing. For analyses, samples were thawed, diafiltered to remove salts, and lyophilised, followed by sequential extraction for solvent-extractable lipids, protein-like and carbohydrate-like organic fractions. Total lipids were extracted from the lyophilised sample using a modified Bligh-Dyer extraction with dichloromethane:methanol (2:1 v/v) and a Dionex accelerated solvent extractor 12 . Residue from the lipid extraction was divided by weight into two portions for protein-like and carbohydrate-like extraction and isotopic analyses (adapted from ref. 13). One portion was hydrolysed (6 N HCl at 100 8C for 19 h) and eluted with 1.5 N NH 4 OH through a cation exchange column to trap the protein-like fraction. The other portion was hydrolysed for the carbohydrate-like fraction (72% H 2 SO 4 for 2 h, then 0.6 M H 2 SO 4 at 100 8C for 2 h). The solution was neutralized with Ba(OH) 2 8H 2 O, adjusted to pH 6-7 with 1.5 N NH 4 OH, and eluted with Nanopure water through a cation/anion exchange column to trap the carbohydrate-like fraction. Concentrations and carbon isotope signatures of known standard compounds were measured before and after the extraction procedure to evaluate extraction efficiencies and potential isotopic fractionation caused by processing. The lipid standard was a mixture of a wax ester (C 14 alcohol, C 20 fatty acid), C 19 alcohol, C 19 fatty acid and androstanol. The amino-acid and sugar standards were L-leucine and D-glucose, respectively. Procedural blanks for protein and carbohydrate extraction methods were processed through the hydrolysis, neutralization, and elution steps using only the reagents. Blanks (2-3 each) needed to be combined to yield enough carbon for 14 C analyses. No lipid-extract blanks were measured isotopically because of their extremely low (#1 mg) carbon yields. Mean standard recoveries were 76^33% (n ¼ 4) for the lipid extract, 66^12% (n ¼ 4) for the protein-like fraction and 93^17% (n ¼ 4) for the carbohydrate-like fraction.

Carbon isotope analyses
Total DOM HMW and organic fractions (lipid extract, protein-and carbohydrate-like) were dried in vacuo and acidified with 1% H 3 PO 4 overnight to remove any carbonates or dissolved inorganic carbon. Acidified samples were then dried in vacuo and the organic carbon oxidized to CO 2 at 850 8C in evacuated sealed quartz tubes containing CuO and Ag metal 26 . The CO 2 was purified and the yield quantified using a calibrated Baratron absolute pressure gauge (MKS Industries) on a vacuum extraction line. Samples were split into sealed Pyrex tubes for D 14 C (,90% of total CO 2 ) and d 13 C (,10% of total CO 2 ) determinations. D 14 C is defined here as the ‰ deviation of the 14 C/ 12 C ratio ( 13 Cnormalized and corrected to AD 1950) for the sample relative to the 14 C/ 12 C ratio of the absolute international standard (95% of the AD 1950 activity of NBS Oxalic Acid I, normalized to d 13 C ¼ 219‰) 27 . Sample CO 2 for D 14 C analyses was reduced to elemental graphite using H 2 over Co catalyst 28 . D 14 C analyses were performed by accelerator mass spectrometry (AMS) and d 13 C measurements were made using a Finnigan Delta S isotope ratio mass spectrometer. The D 14 C and d 13 C values of standard lipid, amino-acid and sugar compounds following extraction procedures were not significantly different from those of the pure compounds, indicating that both blank and fractionation effects were minimal. Errors (^1j) were considered to be the larger of either standard compound analyses or sample replicates (n ¼ 2-3), and were^9‰ for DOM HMW ,^49‰ for the protein-like fraction and^75‰ for the carbohydrate-like fraction for D 14 C measurements, and^0.9‰ for all d 13 C measurements. Owing to the small sample sizes, replicate lipid extractions were not possible. Errors for lipid-extract D 14 C measurements were^4-7‰ based on AMS analytical uncertainties. The 14 C ages of SDOM from this study were identical within measurement error to those determined previously at these two sites, and had not changed significantly over the 10-15 yr between the first and second occupations of the Pacific and Atlantic stations (J.E.B., unpublished data).

Weighted mean turnover time calculation
Weighted mean turnover times (TOT) for bulk DOM (SDOM) having non-homogeneous component ages were estimated by calculating first-order turnover rate constants for each individual DOM HMW organic subcomponent (lipid extract, protein-like, carbohydratelike and MUC fractions) and DOM LMW as: where k i is the turnover rate constant in yr 21 for fraction i, F i is the relative contribution of fraction i to the SDOM pool, and A i is the 14 C age of fraction i (Tables 1 and 2,  Supplementary Table S1). The weighted mean TOT 25 of SDOM is then calculated as: Phenology, the study of annually recurring life cycle events such as the timing of migrations and flowering, can provide particularly sensitive indicators of climate change 1 . Changes in phenology may be important to ecosystem function because the level of response to climate change may vary across functional groups and multiple trophic levels. The decoupling of phenological relationships will have important ramifications for trophic interactions, altering food-web structures and leading to eventual ecosystem-level changes. Temperate marine environments may be particularly vulnerable to these changes because the recruitment success of higher trophic levels is highly dependent on synchronization with pulsed planktonic production 2,3 . Using long-term data of 66 plankton taxa during the period from 1958 to 2002, we investigated whether climate warming signals 4 are emergent across all trophic levels and functional groups within an ecological community. Here we show that not only is the marine pelagic community responding to climate changes, but also that the level of response differs throughout the community and the seasonal cycle, leading to a mismatch between trophic levels and functional groups.
The vast majority of documented phenology studies relating seasonal shifts in biology to climate have come from terrestrial and limnological sources (see refs 5, 6). Furthermore, most studies have solely reported phenological changes for a single species and have not explored trophic and ecological interactions 7 . In this study we investigated changes in marine pelagic phenology in the North Sea across three trophic levels using five functional groups. The major functional groups included diatoms and dinoflagellates separately (primary producers); copepods (secondary producers); non-copepod holozooplankton (secondary and tertiary producers) and meroplankton including fish larvae (secondary and tertiary producers). Inter-annual changes in a measure of the timing of the seasonal peak throughout the whole pelagic production season (the central tendency; see Methods and Fig. 1a, b) were calculated using data from the Continuous Plankton Recorder (CPR) 8 , one of the longest and most spatially extensive marine biological data sets in the world.
The x axis of Fig. 1c shows the timing of the seasonal peaks in 1958 of all 66 plankton taxa used in the analysis; this represents the classical view of succession in the temperate marine pelagic ecosystem. Using the linear slope of the time series of the timing of the seasonal peak, we calculated the change in timing of the seasonal cycle (in months) from 1958 to 2002 for each taxon (Fig. 1c; y axis). Substantial temporal modifications in seasonal successional peaks have occurred over the past few decades. In particular, seasonal peaks of meroplankton have moved significantly (P , 0.0001) forward (for example, the phylum Echinodermata has moved by 47 days (d)). By contrast, diatom peaks in spring and autumn have collectively remained relatively static, albeit with considerable