Environmental constraints on the production and removal of the climatically active gas dimethylsulphide (DMS) and implications for ecosystem modelling

Species composition

As is evident from the standard deviations in Table 1, the variability within the groups is high. In fact, the haptophytes are the only group where all the species tested were observed to produce DMSP. In all other groups, one or several clones did not produce DMSP. Of the groups depicted in Table 1, the prochlorophytes/cyanophytes and diatoms produced the least DMSP. Some exceptions have been observed, but these are not usually found in open ocean areas, but rather at the fringes: in estuaries (e.g., Melosira nummuloides in Keller et al. 1989), in ice algal communities (Baumann et al. 1994; Kirst et al. 1991; Levasseur et al. 1994) and in benthic microbial mats (van Bergeijk et al. 2002). Such data have been excluded from the calculations summarised in Table 1.

The nano/picoplankton are a very diverse group of species that belong to several taxonomic groups and, hence, vary considerably in their DMSP production. For instance, almost all cyanophytes (e.g., Synechococcus species, Trichodesmium), prochlorophytes and cryptophytes tested did not produce DMSP, whereas other species that belong to the chrysophytes and prasinophytes do produce DMSP (Table 1). There are only a few field studies that report on the production of DMSP in picophytoplankton (<2 μm diameter). Corn et al. (1996) showed that picoplankton can contribute up to 25% of depth-integrated total DMSP in oligotrophic waters of the subtropical Atlantic. Within this fraction, the picoeukaryotes were the main DMSP producers, whereas the prokaryotes contributed less than 1% of picoplanktonic DMSP. In contrast, Wilson et al. (1998) investigated DMSP and DMS production in mesocosm enclosures and suggested that Synechococcus can be a significant producer of DMSP under nutrient-replete conditions.

Physiological condition

The effect of abiotic parameters (light, nutrients, temperature and salinity) on the physiological condition of algal cells and subsequently on the production of DMSP, has been reviewed by Stefels (2000). DMSP is a multifunctional compound and there is no doubt that it has a role as a compatible solute in cell metabolism, but the regulation of its internal concentration is still unresolved. Because many abiotic parameters appear to have an effect to some extent, Stefels (2000) hypothesised that the production of DMSP might also serve as an overflow mechanism for excess reduced sulphur under conditions of unbalanced growth, when carbon and nitrogen flows are out of tune. The continued production and possible loss of DMSP would serve as a sink for excess carbon and at the same time regenerate intracellular nitrogen from methionine, which can then be used for synthesis of other amino acids. Although such a mechanism seems wasteful, the benefits are the continuation of the metabolic machinery. In this respect, it is comparable to the commonly observed exudation of carbohydrates by cells at high light and low nutrient concentrations. If indeed DMSP production is connected to overflow metabolism, this requires that DMSP is mainly located in the cytosol and that the intracellular equilibrium concentration is regulated by its degradation or loss from the cell rather than by its production. Transport of DMSP out of the cell would then be facilitated by the extracellular cleavage of DMSP (see also ‘Maintenance of intracellular DMSP concentration: algal DMSP-lyase activity’). Although this hypothesis can explain many of the observed changes in DMSP content presented in the literature, direct evidence is still lacking.

Another hypothesis on the physiological function of DMSP was presented by Sunda et al. (2002). It suggests that DMSP and its breakdown products DMS, acrylate, dimethylsulphoxide (DMSO) and methane sulphinic acid (MSNA) together form a cascade of radical scavengers that may serve as an efficient anti-oxidant system that would need to be regulated in part by the enzymatic cleavage of DMSP. If true, one would expect the production of DMSP and its enzymatic cleavage to take place in the chloroplast, where most reactive oxygen species (ROS) are produced. This indeed seems to be the case with respect to the production of DMSP in higher plants (Trossat et al. 1996), but there is no conclusive evidence for this in marine algae, which use a different biochemical pathway for DMSP production (Gage et al. 1997; Summers et al. 1998). Another complicating factor is that with the common techniques for DMS(P) analysis it is impossible to measure the fluxes through this cascade of compounds. Sunda and co-workers suggested the anti-oxidant hypothesis on the basis of elevated concentrations of intracellular DMSP under stress conditions. In the process of radical scavenging, however, DMSP would be converted into one of its breakdown products. Therefore, a loss of DMSP would be expected, unless the stress reaction results in increased de novo synthesis (up-regulation) of DMSP. Only in those cases, a subsequent overshoot production may lead to increased intracellular concentrations of DMSP and/or one of the downstream products. A method for the measurement of de novo synthesis of DMSP is clearly warranted.

Since unbalanced growth and the production of ROS often co-occur under high irradiance and/or nutrient-limited conditions, it is difficult to test the two hypotheses individually without detailed investigation of the physiological condition of the cells and fluxes through the relevant biochemical pathways. Moreover, the two hypotheses do not necessarily need to be mutually exclusive, since a function in oxidative stress management does not exclude additional functions in cell metabolism. However, for a better understanding, one should be aware of the fundamental differences in operating principles and value the published data accordingly. Here, we present only an update of the current knowledge and quantify the relationships whenever possible.

Light

Most experiments on the effect of light on DMSP production by algae have been carried out with Emiliania huxleyi, a species that is unusual in its ability to grow at very high light (>1,000 μmol m⁻² s⁻¹) intensities (Paasche 2001). In order to make the data from the various publications comparable, we have converted intracellular DMSP quota to DMSP:C ratios. From these data, a good correlation between increasing DMSP:C-ratio and increasing irradiance can be observed (Fig. 2). For comparison, data for Phaeocystis antarctica are included in the figure. A comparable increase with irradiance can be observed, but with an overall higher offset. This offset might be the result of additional temperature effects.

Open image in new window

Getting information from field data on the controls on cellular DMSP is not straightforward but there is some evidence from a Lagrangian experiment in the Southern Ocean, which tends to suggest that light is an important factor at the daily scale (Fig. 3; S. Turner and I. Peeken, unpublished data). Figure 3 shows vertical profiles of DMSP_p, normalised to 19′-hexanoyloxyfucoxanthin (19′Hex), an accessory pigment which is indicative of Prymnesiophytes. Since the changes in the ratios are dominated by variation in DMSP, the results suggest that there is a diel cycle, in which DMSP_p is consumed during the day. This may be supportive of an antioxidant function, but also shows that short-term effects can be opposite to long-term effects as presented by Sunda et al. (2002) and as shown in Fig. 2. Clearly, better methods are warranted to measure fluxes through the DMSP_p pool in field samples.

Open image in new window

Temperature

DMSP has been found to be a compatible solute for cell metabolism under cold conditions (Karsten et al. 1996; Nishiguchi and Somero 1992, reviewed in Stefels 2000). The observation that DMSP is present in many ice algae, including the diatoms, and in many pelagic algae from polar regions (Matrai and Vernet 1997) may be indicative of its functionality under cold conditions. Surprisingly, only two studies report on the acclimatisation of the intracellular DMSP concentration at various temperatures (Sheets and Rhodes 1996; van Rijssel and Gieskes 2002). Converting the Van Rijssel and Gieskes (2002) data on Emiliania huxleyi to DMSP:C ratios gives a correlation with temperature as shown in Fig. 4. The added data points for Phaeocystis are for P. globosa at 10°C and for P. antarctica at 4°C (J. Stefels unpublished data) and fall close to the E. huxleyi relationship. However, further data are needed to establish whether the same relationship holds for DMSP-producing species that do not belong to the haptophytes.

Open image in new window

Nutrients

After Challenger (1951) had noticed the structural analogy between DMSP and glycine betaine (GBT), many have suggested that DMSP could replace GBT as an osmoregulator under nitrogen-limited conditions. Indeed, there are several reports of increased cellular DMSP content under N-limited growth, but there are also reports of the contrary (Stefels 2000 and refs. therein). As a whole, the effects of nutrient limitations on DMSP production are still enigmatic.

From a physiological point of view, the intracellular concentrations of organic solutes are most relevant, as it is this concentration that affects enzymatic processes. Shifting from unlimited towards limited growth in a batch culture, cell volume often reduces under nitrogen and iron limitation and stays constant under phosphate limitation. Therefore, an increased intracellular DMSP concentration under N or Fe limitation is at least partly due to a reduction in cell volume (Bucciarelli and Sunda 2003; Keller et al. 1999a; Stefels and van Leeuwe 1998).

From a modeller’s point of view, it is more relevant whether or not the DMSP production can be related to primary production. A complicating factor is that under nutrient-limited conditions the concentrations of DMS and dissolved DMSP often increase, either due to cell lysis or to active exudation (Laroche et al. 1999). The question is how the total production of DMS and DMSP is related to algal growth. Unfortunately, total pools of DMS(P) are rarely presented in the literature. In Fig. 5, a compilation of experiments with Phaeocystis globosa is presented (unpublished data, J. Stefels), in which it is clearly shown that the specific DMS + DMSP production is coupled to cell growth and compares well under a variety of conditions: a range of salinities, low or high light conditions and either nitrogen or phosphorus limitation. The fact that the regression coefficient deviates from 1 reflects the observation that under unlimited (high) growth rates, cells tend to divide faster than they grow in terms of carbon, which results in a cell-size reduction during exponential growth. The positive Y-intercept indicates that at limited (low) cell growth, DMSP production continues under all conditions, even when cell numbers decline (negative growth). Whether this production is due to a few healthy cells that are still growing amidst a majority of inactive or dead cells, or because of a reaction to stress is unknown. This compilation shows that there is no increased production under stress conditions as suggested in the anti-oxidant hypothesis and that a modeller’s practice of coupling DMSP production to cell growth is an appropriate approach.

Open image in new window

A potential pitfall in carbon-based models is, however, the decoupling between cell growth and carbon growth, which would result in shifts of the DMSP:C ratios. Direct evidence for such a decoupling between the carbon and sulphur cycles is still lacking, but indirect evidence can be sought in calculated DMSP:C ratios relative to growth rates. In Fig. 6, a compilation of available data is given. In those cases where only cell volume data were available, cell carbon is calculated according to Menden-Deuer and Lessard (2000). This may result in an underestimation of cell carbon under N or Fe limitation (and thus an overestimation of the DMSP:C ratio), since, under those conditions, cells often become carbon denser, i.e., an increase in cell carbon per cell volume (Stefels and van Leeuwe 1998). In the low-DMSP producing diatom Thalassiosira pseudonana a clear effect of nutrient limitation on DMSP:C ratios could be found, with highest ratios under N limitation and all other conditions comparable. However, in all other high-DMSP-producing species, changes of the ratio are negligible (see figure legend for description and references). This suggests that cells with a high DMSP content do not respond to nutrient limitation—or at least that a response in the DMSP production is too low to affect the DMSP:C ratio—whereas cells with low DMSP content do react. Modellers may implement this by assigning different behaviour towards nutrient limitation to different plankton groups in their model.

Open image in new window

Maintenance of intracellular DMSP concentration: algal DMSP-lyase activity

Algae can adjust the intracellular concentration of DMSP through the biosynthetic (anabolic) or the degradation (catabolic) pathways. DMSP-lyase enzymes facilitate the degradation pathway, in which DMSP is cleaved to DMS, acrylate and a proton. What controls the activity of DMSP-lyases in phytoplankton is still unknown. Stefels (2000) suggested that in the case of the production of DMSP as an overflow mechanism, the intracellular equilibrium concentration has to be regulated by its degradation rather than by its production. Since DMSP is a zwitter ion, this could be achieved by actively transporting DMSP out of the cell. Subsequent removal of DMSP from the transporter site by extracellular cleavage would facilitate the release, since concentration gradients are kept maximal. Such a mechanism would necessitate a membrane-bound extracellularly located DMSP-lyase. This is particularly relevant for organisms with a thick boundary layer such as Phaeocystis colonies. The idea of such a role for DMSP-lyase was instigated by the observations that Phaeocystis exhibits high in vivo lyase activities when dissolved DMSP is added to the culture medium and that extracellular inhibitors can repress this activity (Stefels and Dijkhuizen 1996). In addition, it has been observed that acrylate, one of the products of DMSP degradation, accumulates in the mucus layer of Phaeocystis colonies (Noordkamp et al. 2000), which is in agreement with an extracellular location of the enzyme. Sunda et al. (2002) suggested that DMSP-lyase is involved in the scavenging cascade that detoxifies cells of harmful oxygen radicals, as it facilitates the production of DMS and acrylate, two products that are highly efficient ROS scavengers. This hypothesis would favour an intracellular and possibly chloroplastic location of the lyase.

For modelling purposes two aspects of algal DMSP-lyases are relevant. Firstly, the activity of an extracellularly located enzyme may contribute to the conversion of the (ambient) dissolved DMSP pool. In this case, the apparent enzyme efficiency (α; i.e., the initial slope of a V/S plot showing DMSP-lyase activity versus DMSP concentration) at ambient conditions obtained during in vivo assays is relevant. Secondly, the activity of intra- or extracellular enzymes may be crucial for the direct conversion of intracellular DMSP, after cells break up due to autolysis, viral attack or grazing (see below for a description of these processes). In these cases, one can envisage that the lyase, formerly separated from its substrate, is exposed to elevated DMSP concentrations, resulting in high production rates of DMS. The conversion rate of this DMSP might then be related to the V _max of the enzyme. In Table 3, we tried to make the available published data comparable by recalculating them on a cell-carbon basis.

Another important question is how algal lyase activity compares to bacterial activity. Although the latter will be discussed in the section ‘Microbial consumption of DMSP’, several studies on field samples indicate that algal lyase activity can be as important, if not more, as bacterial lyase activity. Stefels et al. (1995) found a strong correlation between Phaeocystis abundance and in vitro lyase activity in Dutch coastal waters. No lyase activity was found in size fractions <10 μm. During sampling of a bloom of E. huxleyi in the North Atlantic, Steinke et al. (2002a) found >74% of the in vitro lyase activity associated with particles >10 μm and suggested that dinoflagellates were responsible for this activity. Niki et al. (2000) calculated that the algal lyase pathway is as important as the bacterial lyase pathway, in samples from Tokyo Bay. Also other size fractionation experiments have shown that lyase activity is often associated with the large size fractions (Cantin et al. 1999; Scarratt et al. 2000), although it cannot be excluded that attached bacteria are partly responsible for this activity. In a modelling study of the seasonal evolution of DMS in the Southern Bight of the North Sea, van den Berg et al. (1996) showed that the presence of an algal DMSP-lyase associated with the occurrence of Phaeocystis was essential to properly describe the DMS spring peak.

These results suggest that for modelling purposes it might be beneficial to distinguish between phytoplankton groups with and without DMSP-lyase. The limited data we have so far suggest that the dinoflagellates and part of the haptophytes have DMSP-lyase. This has been incorporated in our recommendation on plankton groups and their characteristics (section ‘Conclusions’; Table 6).

Mechanisms of release to the dissolved fraction

Pathways through which particulate DMSP is released into the dissolved phase are active exudation, cell lysis due to senescence or to viral attack and grazing by zooplankton. In the many publications that describe these processes, this release is presented either in the form of dissolved DMSP (DMSP_d) or in one of its degradation products. For models, however, an important question is whether this release goes through the dissolved DMSP pool, with subsequent conversion by bacterial and algal enzymes, or that release and conversion of the particulate DMSP pool is intrinsic to the release process. In the vicinity of or within a bursting cell, high-affinity enzyme systems will result in the rapid conversion of DMSP to DMS, whereas in the case of no or only low-affinity enzyme systems the same process may result in dissolved DMSP alone. The subsequent fate of DMSP_d will then be consumption by bacteria with a DMS yield depending on the development of the bacterial community. The same release process may thus result in completely different DMS yields and therefore needs to be modelled differently for the different plankton groups.

A complicating factor in the evaluation of published data on the fractionation between dissolved and particulate DMSP is the potential for overestimations of DMSP_d, due to release from the cells during filtration (Kiene and Slezak 2006). DMSP_d is taken to be the filter fraction of the less than 1 μm fraction, but for fragile, DMSP-containing flagellates such as Phaeocystis, filtration can easily result in operational release of dissolved organic material from the cells. Any use of published data should, therefore, involve a thorough check of the analytical methods used for the quantification of DMSP_d. The following sections address the different mechanisms of release of DMSP-sulphur from algal cells.

Microbial consumption of DMSP

The chemical half-life of DMSP in seawater is >8 years (Dacey and Blough 1987), which results in high abiotic stability under natural conditions (moderate temperatures and pH). Therefore, most of the DMSP removal is through enzymatic processes. In the microbial food web, dissolved DMSP has many fates and several recent reviews on the microbial pathways and involved mechanisms have been published (Bentley and Chasteen 2004; Kiene et al. 2000; Lomans et al. 2002; Yoch 2002). They all show that DMSP can be readily used in a complex network of enzymatic conversions. This versatility indicates that this single compound is of major importance for the nutrition of the bacterial community. Indeed, several studies have shown that DMSP alone can contribute 1 to 15% of the total bacterial carbon demand in surface waters. Moreover, DMSP assimilation can satisfy most, if not all the, sulphur demand of marine bacteria (Kiene and Linn 2000; Simo et al. 2002; Zubkov et al. 2001). Since the focal point of this section is the quantification of DMSP removal, only the overall effects of the main pathways originating from DMSP (Fig. 1) will be discussed here.

The two major pathways of bacterial DMSP degradation are cleavage to DMS and acrylate and demethylation/demethiolation, with or without prior uptake into the cells. Quantitatively, the most important degradation pathway of DMSP, is demethylation. This pathway does not yield DMS, but 3-methiolpropionate (MMPA) and, after a second demethylation, 3-mercaptopropionate (MPA). MMPA can also yield methanethiol (MeSH) after a demethiolation reaction. Next to DMS, MeSH is another important volatile sulphur compound. It is metabolised rapidly and appears to be the major sulphur source for the production of sulphur-containing amino acids and proteins in bacterioplankton (Kiene et al. 1999). Members of the Roseobacter group, a ubiquitous group of marine bacteria, are well known MeSH producers (Zubkov et al. 2001) and in waters rich in these bacteria, consumption of DMSP via demethylation can be tenfold higher than DMS production (Kiene et al. 2000).

The cleavage of DMSP in aerobic DMS-producing bacteria is thought to be similar to the cleavage by algae and differentiating between the two in field samples is thus a challenge. As far as this has been investigated, it appears that acrylate is further used as a carbon source, leaving DMS untouched (Yoch 2002 and references therein). The current three conceptual models for the uptake and metabolism of DMSP (Yoch 2002) suggest considerable diversity of DMSP-lyases amongst the bacteria. The presence of DMSP and DMSP-cleaving enzymes is a prerequisite but cannot be used as proxies for DMS production. Measurements in the Northeast Atlantic Ocean showed maximum in vitro DMSP-lyase activity (which indicates the potential for DMS production from both bacterial and algal enzymes) on the order of 6–185 nM DMS h⁻¹ (Steinke et al. 2002b), whereas in vivo DMS production (from short-term (8 h) bottle incubations) resulted in only 1.2–14.4 nM DMS h⁻¹ (Simo and Pedros-Alio 1999b).

A more extensive screening of 15 Roseobacter strains showed that the cleavage pathway was present in all strains and that the demethylation pathway co-occurred within five of these strains (Gonzalez et al. 1999). In addition, several strains were able to degrade DMS and MeSH. This versatility of sulphur metabolism within a single genus of bacteria indicates that identification of the community structure alone is insufficient to determine the dominating degradation pathways. Furthermore, the fact that several isolates were also able to consume DMS means that the difference between DMSP loss and DMS production in natural waters not necessarily indicates a dominance of the demethylation pathway. While considering the total myriad of conversion pathways, Kiene et al. (2000) proposed a hypothetical model that helps us to understand the relative importance of these pathways. These authors suggested that bacterioplankton will prefer the demethylation/demethiolation over the lyase pathway at low DMSP_d concentrations. This is because this pathway provides more energetic benefits and it is a relatively economic way to assimilate reduced sulphur. In fact, it was proposed that the total sulphur demand of bacteria can be derived in this way and as such, this pathway can be directly linked to bacterial production. At higher DMSP_d concentrations, the DMSP that is not assimilated is then available to the cleavage pathway. In other words, the fraction that is converted to DMS depends on the biomass and growth of the bacterial community. Kiene et al. (2000) also stipulated that it is not necessary for the DMSP_d concentration to be high, but that it is the bacterial sulphur demand relative to the DMSP_d availability that is critical. For instance, if sulphur demand is low due to nutrient limitation or UV stress, the demethylation pathway is minor and DMS yield may increase. Such a DMSP-availability hypothesis is consistent with the shifts in the relative contribution of the demethylation- and cleavage pathway, as observed by Simo and Pedros-Alio (1999a), who found DMS yields from consumed DMSP ranging between 5% and 100%. These shifts were correlated to the depth of the mixed water layer, with higher DMS yields in shallow mixed layers, and could be related to UV stress on the bacterial population. In addition to this bacteria-oriented hypothesis, it appears that at high DMSPd concentrations, either in blooms or in microenvironments around algal cells or aggregates, the algal cleavage pathway may overrule the bacterial cleavage. For instance, the peak of the DMS concentration in Phaeocystis blooms is often associated with the younger parts of the bloom and not with the senescent part (van Duyl et al. 1998). This may be related to the fact that the bacterial community of younger blooms is often still developing and therefore their sulphur demand low. A relatively large proportion of DMSP_p may then be rapidly converted to DMS by algal DMSP-lyases. During the senescent stages of the bloom, the mature bacterial community may divert DMSP into the demethylation pathway and/or consume a large proportion of the DMS. In terms of enzyme kinetics, one can envision that a plankton bloom diverts from a low affinity but high capacity cleavage system to a high-affinity but low-capacity demethylation/demethiolation system (Kiene et al. 2000).

The particulate DMSO pathway

In comparison to DMS and DMSP, the nonvolatile dimethyl sulphoxide (DMSO) is a poorly understood component of the marine sulphur cycle. In addition to DMSO being ubiquitous in seawater in the dissolved (or filterable DMSO_d) phase, evidence is also gathering for the direct production of particulate DMSO (GF/F retained DMSO_p) by cultures and natural assemblages of marine microalgae. DMSO has been the subject of three reviews (Hatton et al. 2004; Lee and de Mora 1999a; Lee and de Mora 1999b) and whilst we aim not to be overly repetitive here, it is nonetheless important to underline some pertinent information regarding particulate DMSO. This compound is remarkable in its ability to permeate intact biological membranes and it is well known as an effective radical scavenger. Indeed, it is widely used medically to deliver drugs through the skin and as a cryoprotectant for the storage of cells at freezer temperatures. This ability to transfer across membranes and the DMSP antioxidant cascade hypothesis put forward by Sunda et al. (2002) could explain the marked increase in DMSO_d seen in dinoflagellate cultures as they approach and enter stationary phase (Simo et al. 1998). DMSO_p production has been noted for cultures of the dinoflagellate Amphidinium carterae and the coccolithophorids Pleurochrysis carterae and E. huxleyi (Simo et al. 1998), suggesting that a wide range of phytoplankton may produce DMSO_p. Exponential phase cultures of A. carterae and E. huxleyi gave mean DMSP_p:DMSO_p molar ratios of 25 and 8, respectively. In a study of the Peruvian upwelling system Riseman and DiTullio (2004) found a strong positive correlation though with ratios of 0.7 to 4.2, and they noted that concentrations of both sulphur pools and the antioxidant β-carotene increased under low-iron conditions which might be consistent with the production of free radicals due to iron deficiency. These authors provide a table that includes DMSP_p:DMSO_p values from five other studies that span the range 0.02–100. More recently Simo and Vila-Costa (2006) discussed geographic and temporal patterns of DMSO_p distribution using the largest data set for DMSO_p and DMSP_p compiled to date. They report that the ratio of DMSP_p:DMSO_p varied from 1 to 13 with an average of 5.2, and noted a trend towards higher proportions of DMSO_p in phytoplankton in warmer seas. Further research is needed to better evaluate whether these variable ratios result from differences in phytoplankton speciation, bloom stages or prevailing growth conditions.

Significance of vertical flux as a sink for particulate DMSP

The significance of vertical flux as a sink for particulate DMSP was recently reviewed and reassessed using new data gained in coastal waters of northern Europe (Belviso et al. 2006 and references therein). Special attention was paid to the many biases that can affect estimates of downward fluxes of DMSP (catchment efficiency of traps, releases of particulate material to the dissolved pool, etc). A method was also suggested to correct DMSP fluxes from biological losses during the sedimentation process. Indeed, the lability of DMSP during the sedimentation process has been shown to be comparable to that of chlorophyll-a (Cailliau et al. 1999). The chlorophyll-a degradation products resulting from grazing and senescence are, in order of least to most degraded: phaeophytin-a, phaeophorbide-a and pyrophaeophorbide-a. These phaeopigments can all be detected by high-performance liquid chromatography (HPLC). Assuming that phaeopigments trace the degradation products of DMSP, the following equation can be used to restore DMSP fluxes and to revise the daily vertical loss rates for DMSP:

Bacterial consumption

In incubation studies with water samples from the equatorial Pacific, Kiene and Bates (1990) demonstrated that bacterial degradation of DMS dominated over sea-to-air gas exchange. The major microbial degradation pathways of DMS are consumption via DMS monooxygenases and methyltransferases and oxidation via DMS dehydrogenase. DMS can be oxidised to DMSO by a variety of sulphur and ammonia oxidisers, methylotrophs and phototrophs, which suggests that this is a versatile worldwide process (reviewed by Bentley and Chasteen 2004). The reverse reaction—reduction of DMSO to DMS via DMSO reductases under aerobic and anaerobic conditions is well known from laboratory studies, but as far as we are aware, this pathway has not been documented for any natural assemblage. The enzyme responsible for the conversion of DMS to MeSH and formaldehyde has long been thought to be DMS monooxygenase. However, both the enzyme and the metabolic pathway are enigmatic (Bentley and Chasteen 2004). Detailed biochemical and molecular level studies are needed to improve understanding of these important microbial processes. DMS can also be assimilated as a sulphur source although Zubkov et al. (2002) found this to be a minor pathway in the northern North Sea. However, it is interesting to note that Fuse et al. (2000) isolated a γ-proteobacterium Marinobacterium sp. that assimilated DMS at wavelengths of 380–480 nm via the production of heat-stable photosensitisers. A light-requiring mechanism like this would not operate under the dark incubation conditions that are often used for incubation experiments.

It is clear that bacterial activity is an important factor in reducing the quantity of DMS that is emitted to the atmosphere. Simo (2004) compiled data from several studies in different marine areas, which illustrates the linear relationship between biological DMS consumption and DMS production, and suggests that consumption accounts for between 50% and 80% of the production. However, the significance of bacterial DMS consumption also alters depending on the strength of other, competing loss processes. Since all of these can alter rapidly according to meteorological forcing it is difficult to derive a range of widely applicable quantification terms. This is best illustrated by a study in the subpolar North Atlantic which showed that: (1) photochemical degradation dominated under clear skies and shallow mixing conditions, (2) bacterial consumption was most important when skies were cloudy and/or when the water column mixed to a greater depth and (3) loss of DMS due to sea-to-air transfer was roughly equivalent to bacterial consumption during a storm (Simo and Pedros-Alio 1999a).

Photochemical oxidation of DMS

Brimblecombe and Shooter (1986) established that photochemical oxidation is an important loss process for DMS, that the reaction rate varies with DMS concentration and that photolysis can occur at visible wavelengths via photosensitisers. It is important to emphasise that DMS photolysis does not always result in DMSO production. However, when DMSO is produced there remains a possibility that a reduction pathway could operate to return DMS to the seawater pool. Kieber et al. (1996) showed that maximum photolysis occurred at wavelengths between 380 nm and 460 nm and in Pacific waters only 14% of the DMS was converted to DMSO. Hatton (2002a) found that DMS was removed from seawater by UVA/visible light (>315 nm) and UVB (<315 nm) though only UVA/visible wavelengths resulted in DMSO production and in northern North Sea waters ∼37% of total DMS was lost by this route under full natural sunlight. Toole and Siegel (2004) investigated the factors that influenced DMS cycling using a DMS time series for the Sargasso Sea and found that the UV radiation dose explained 77% of the variability in DMS concentrations. Brugger et al. (1998) demonstrated that the initial rate of DMS removal is proportional to initial DMS concentration (5–100 nM tested), irradiance intensity and DOC concentration. However, we note that photolysis is most likely to be mediated by the portion of DOC that is known as coloured dissolved organic matter (CDOM). The apparent quantum yield for DMS also doubles with a temperature increase of 20°C (Toole et al. 2003). Interestingly nitrate concentrations are also relevant, in that Toole et al. (2004) found that in high-nitrate Antarctic waters 35% of the observed DMS photolysis was related to nitrate photochemistry and photolysis rates increased linearly with added nitrate. This nitrate dependence was also observed by Bouillon and Miller (2004). Moreover, Bouillon and Miller (2005) found that nitrate-induced photolysis of DMS was strongly enhanced by the presence of bromide ions and to a lesser extent by bicarbonate/carbonate ions. Contrary to what was expected bicarbonate/carbonate-induced photolysis reduced with increasing pH.

By pooling data from different oceanic regions, Hatton et al. (2004) found a highly significant correlation between DMS and DMSO in (near-)surface waters, with the DMSO concentration comparing to 1.5 times the DMS concentration. This clearly shows that photolysis is an important DMS loss process and that accurate photolysis parameterisations are needed to improve our understanding of DMS cycling and DMS models. Some of the studies mentioned above give DMS photolysis rates that are highly variable: 0.03–0.07 h⁻¹ for the northern North Sea (Hatton 2002a), 0.12 h⁻¹ for the coastal Adriatic (Brugger et al. 1998), 0.04 h⁻¹ for the Pacific (Kieber et al. 1996), 0.026–0.086 h⁻¹ for the western Atlantic (Toole et al. 2006) and 0.16–0.23 h⁻¹ for the Antarctic (Toole et al. 2004). Given the previous discussion of CDOM, nitrate concentration, wavelength and temperature, all of which can vary with depth, we might conclude that an ideal parameterisation could be rather complex. For now modelling studies appear to use a variety of relatively simple approaches such that the DMS photolysis rate varies with depth and season (e.g., Archer et al. 2004; Lefevre et al. 2002).

Flux of DMS to the atmosphere

Until very recently there was no direct method for measuring the sea-air flux of DMS. Micrometeorological techniques are considered to be the best approach for determining fluxes (Businger and Delany 1990) but are a considerable technological challenge. Huebert et al. (2004) made the first measurements of DMS flux using the eddy correlation technique, on board ship. Their breakthrough was due to the development of the atmospheric pressure ionization mass spectrometer-isotopically labelled standard (APIMS-ILS) technique, which can measure DMS at high frequency. Another micrometeorological technique used recently for DMS is relaxed eddy accumulation (Zemmelink et al. 2004). Both methods require highly sophisticated equipment and specialist knowledge, so it is likely that the emission rates of DMS will continue to be determined by semi-empirical methods for some time to come.

The most widely used method for calculations of gas fluxes requires the concentration gradient between surface water and the atmosphere and a kinetic parameter known as the transfer or piston velocity (Liss and Slater 1974):

where F _DMS = net flux; k _(T) = transfer velocity; ΔC = concentration gradient across the air-sea interface i.e., C _w−C _a H ⁻¹, where C _w = concentration in water, C _a = concentration in air and H = Henrys law constant. Since C _a is very low relative to C _w, C _a is often taken to be zero, which generally may lead to an overestimation of about 6% or less which is not significant, relative to other uncertainties (Turner et al. 1996).

Over the last decade the results of oceanic multiple deliberate tracer experiments have refined existing estimates of k (Nightingale et al. 2000) and, including the findings of the micrometeorological studies, there is still disagreement. Estimates vary by factors of about two and there is generally greater uncertainty for k at higher wind speeds. What is agreed, however, is that wind speed alone is not adequate for the parameterization of k and that other factors, such as surfactants and breaking wave bubble generation are important (e.g., Frew et al. 2004). Quantification of these parameters in the field is not straightforward and future measurement and modelling efforts are required before improvements in the quantification of k can be achieved.

The concentration term of the flux calculation also has major uncertainties, mainly associated with the spatial and temporal scales of measurements. These include bias towards Spring and Summer measurements with many areas of the world oceans still sparsely represented (Kettle et al. 1999). More important, in the context of ecosystem modelling and validation, is the uncertainty in measured DMS concentrations arising from the depth at which the samples are taken: operationally defined as surface water, but can be as deep as 11 m. Vertical profiles of DMS concentration through the oceanic mixed layer are rarely homogenous and transient stratification (surface heating and freshwater input) can have marked affects (see Ward et al. 2004). At the smaller vertical scale of the sea-surface microlayer (∼60 μm thickness) lies further uncertainty for the DMS concentration term. Study of the microlayer is technologically and logistically difficult and the few reports suggest a wide range of enrichments of DMS, relative to subsurface concentrations (e.g., 0.38–2.94, mean 1.1; Yang and Tsunogai 2005). It is not possible to assess the uncertainty of the DMS concentrations in the literature and further work is required to determine how important the depth of sampling is for a variety of biological and physical drivers and hence for the concentration term in the air-sea flux equation.

In summary, there is still a great deal of uncertainty in the calculation of sea-to-air fluxes and, for the time being, we recommend use of the transfer velocity parameterisation of Nightingale et al. (2000), which appears to be intermediate between all other estimations.

Vertical and horizontal mixing of DMS

Although oceanic mixing and stirring processes do not constitute loss of DMS per se, in dynamic oceanographic systems, dispersion and homogenisation can have significant control on the amount of DMS measured in the surface layer over timescales of hours and days. Vertical profiles of DMS in the mixed layer generally show some structure, often with a subsurface maximum. However, during high-wind events increased mixing causes homogenisation of the upper water column and deepening of the mixed layer. Thus, in the course of a few hours the concentration of DMS at the surface can increase or decrease depending on the prior vertical profile structure and degree of enhanced turbulence. Post storms and during warm quiescent periods, when turbulence decreases, the mixed layer tends to shallow which causes DMS to be trapped below the new depth of the picnocline. This, in effect, reduces the size of the reservoir of DMS that is available for sea-to-air flux, inter alia.

Winds, currents and tides all contribute to horizontal mixing, which can also affect DMS concentrations. As an example, during a Lagrangian iron-addition experiment in the Southern Ocean a sulphur-hexafluoride-labelled patch of water spread from about 70–1,000 km² in 18 days which led to rapid dilution of the bloom with water containing low biomass and DMS (Turner et al., in prep). Vertical and horizontal mixing processes can thus be overruling loss processes after a local built-up of the DMS concentration, which warrants inclusion of these processes in models.