Roger I. Jones, Department of Biological and Environmental Science, PL 35, FIN-40014 University of Jyväskylä, Jyväskylä, Finland. E-mail: email@example.com
1. It has long been known that substantial amounts of methane are produced in anoxic lake sediments, and the components of the methane cycle in lakes have been well described. At oxic–anoxic interfaces, methane-oxidising bacteria (MOB) convert methane to microbial biomass and can be highly productive. However, only recently has methane been recognised as a potentially important carbon and energy source for lake food webs, and some instances have also been reported of methane contribution to river food webs. Stable isotope analysis (SIA) has provided compelling evidence in this respect and has been supplemented by other lines of evidence.
2. In the benthic food webs of lakes, profundal chironomid larvae appear to be the main conduits for trophic transfer of biogenic methane via grazing on MOB. The mode of feeding of these larvae and the microhabitats they generate both promote larval ability to exploit MOB production. Support to chironomid larvae from methane is rather widespread, but its degree is highly variable; estimates suggest that in some lakes methane-carbon might contribute more than 60% of chironomid carbon biomass.
3. Evidence of crustacean zooplankton in lakes deriving part of their carbon from methane is currently more limited. Reports from some lakes have indicated Daphnia with a substantial (>50%) contribution of methane-carbon in their biomass. However, for this to happen, an oxic–anoxic interface where sufficient MOB production can occur needs to be within the range of vertical migrations by zooplankton, which may only rarely be the case. Hence, a significant methane subsidy of pelagic food webs in lakes is probably much less widespread than for benthic food webs.
4. There is also recent and currently very limited evidence that some stream benthos derives biomass carbon (reported values up to 30%) from methane. This can occur in stagnant backwater pools where conditions can be analogous to those in lake sediments. However, groundwater aquifers can also supply water supersaturated with methane to some rivers, providing a basis for a microbially-mediated transfer of methane-carbon to river benthos.
5. Evidence for significant transfer of methane-derived carbon to higher trophic levels is still very limited. Within some lakes, those fish species that feed extensively on chironomid larvae can derive a substantial part (perhaps up to 20%) of their carbon biomass from methane. It is also likely that methane-carbon produced in lakes or rivers is exported to riparian ecosystems when emerging chironomids or other insects are eaten by invertebrate or avian predators.
6. We argue that conceptual models of freshwater food webs, and especially those for lakes, need to be modified to enable incorporation of biogenic methane as a carbon and energy source. For some types of lakes, carbon and energy budgets certainly need to take account of the production and utilisation of biogenic methane, and the accumulating evidence indicates that this is a more widespread phenomenon that has generally been acknowledged hitherto.
A major paradigm shift in limnology over recent years has been the recognition that external (allochthonous) inputs of organic matter represent important carbon and energy subsidies to supplement autochthonous primary production in lake ecosystems, and especially for pelagic food webs (Jones, 1992; Hessen, 1998; Jansson et al., 2007; Reynolds, 2008). These allochthonous carbon inputs are mainly mediated under oxic conditions by heterotrophic bacteria which provide biomass for zooplankton grazers, but which also generate CO2 so that most lakes are now thought to be net heterotrophic (e.g. Cole et al., 2007). However, anoxic conditions prevail in freshwater sediments and also in the hypolimnion in many stratified lakes, and under these conditions organic matter of autochthonous or allochthonous origin is a substrate for different microbial metabolic pathways, one of the most important of which is methanogenesis (e.g. Rudd & Taylor, 1980). In fact, lake sediments in particular are known to be “hot spots” of methane production and lakes can be a significant source in global methane budgets (Bastviken et al., 2004). Part of this methane production, especially in shallow sediments and water columns, is exported by ebullition to the atmosphere, but much of the methane produced in deeper sediments probably travels upward by diffusive flux and water column mixing (Bastviken et al., 2004; Juutinen et al., 2009) and is readily oxidised by methane-oxidising bacteria (MOB) once it reaches the oxic sediment or water column (Rudd & Taylor, 1980; Bastviken, Ejlertsson & Tranvik, 2002; Whalen, 2005; Kankaala et al., 2006a). Although the essential chemical and microbiological elements of the methane cycle in lakes have long been known (see review by Rudd & Taylor, 1980), the possibility that methane might serve as a significant carbon and energy source for freshwater food webs has only quite recently attracted attention.
The first report of very low δ13C values from freshwater organisms indicating possible use of biogenic methane appears to have been by Bunn & Boon (1993) for chironomid larvae in small Australian billabongs. Around the same time, we had measured unexpectedly low δ13C values for chironomid larvae from Loch Ness, Scotland. A stable isotope ‘map’ of the Loch Ness food web represented by a bi-plot of δ15N against δ13C (Fig. 1) shows that most food web components in the lake cluster within the expected confines of the presumed food chain trajectories from two alternative basal resources (autochthonous phytoplankton and allochthonous terrestrial organic matter), assuming widely accepted trophic fractionation factors of 3.5‰ for nitrogen and 1‰ for carbon. Salmon (Salmo salar) lie clearly to the right of this cluster reflecting the marine origin of their accumulated biomass, while trout (Salmo trutta) parr lie a little to the right reflecting their growth in inflow streams rather than the lake itself (Grey, 2001). However, chironomid larvae collected from the lake profundal lie way to the left of the cluster, exhibiting much lower δ13C values and somewhat lower δ15N values than would have been expected if they had been simply feeding on some mixture of sedimented phytoplankton and terrestrial plant detritus. Clearly, the diets of these chironomid larvae must have included some food material that is strongly 13C-depleted relative to the sediment detrital organic matter; MOB with their known low δ13C values (see above) are an obvious candidate for this isotopically light food.
Hessen & Nygaard (1992) had earlier estimated for a small humic lake in Norway that methane sustained a bacterial production exceeding the net production by heterotrophic bacteria and had speculated that a methane-carbon pathway to pelagic consumers might be of special relevance for humic lakes. In a study of 12 small forest lakes in southern Finland, Jones et al. (1999) related crustacean zooplankton δ13C values to those of particulate organic matter (POM), dissolved organic matter (DOM) and, where possible, separated phytoplankton. POM and DOM in these lakes is predominantly of terrestrial origin, and their δ13C values of around −28‰ reflected this, whereas phytoplankton δ13C was generally lower. They had hypothesised that if zooplankton in coloured, high DOM lakes rely more on allochthonous carbon sources, zooplankton δ13C values should increasingly approach those of POM and DOM as the water colour of the lakes increased. In fact, an opposite trend was found of zooplankton δ13C decreasing with increasing water colour. Moreover, the lowest zooplankton δ13C values recorded of −46‰ were the most negative reported in the literature at that time, although Bunn & Boon (1993) had reported values of −37‰ for calanoid copepods from Australian billabongs. Jones et al. (1999) suggested incorporation of isotopically light methane via grazing on MOB to be a plausible explanation for these observations.
Following these early observations, further reports of isotopically light consumers have been steadily accumulating from lake studies and, accompanied by other lines of evidence, are supporting the view that some freshwater consumers are including appreciable, and in some cases very substantial, contributions from MOB in their diets. Here, we summarise this accumulating evidence for both benthic and pelagic freshwater food webs. Our review focuses on metazoan consumers in lakes and rivers and hence does not encompass the considerable work on wetlands (including rice paddies) where the focus has been more on microbial food web structure. We also attempt to evaluate the potential wider significance of methane in lake food webs, for which most evidence is available, by considering the conditions in lakes which promote methane-carbon entry into benthic and pelagic food webs, and we present a modified conceptual framework for lake ecosystems that incorporates methane as a carbon/energy source.
Benthic food webs
Benthic macroinvertebrates (zoobenthos) comprise a wide range of taxonomic groups ingesting diverse food material by various feeding strategies. Ingestion of MOB is only expected to be important for those zoobenthos classified as detritivores, ingesting POM and the microorganisms associated with it. Chironomid larvae are among the most important detritivorous zoobenthos in lakes, especially in the profundal zone where they often dominate the zoobenthos community. Most profundal chironomid larvae have been considered to exploit the rain of detrital organic matter, especially from phytoplankton, settling down through the water column to the sediment surface (Jónasson, 2004), so have been considered important to the process of pelagic–benthic coupling in lake food webs. Moreover, chironomid larvae can be important food for macroinvertebrate predators, fish and birds (Strayer & Likens, 1986; Lindegaard, 1994). Most evidence for exploitation of methane-carbon by zoobenthos via MOB relates to profundal chironomid larvae.
What evidence is there that profundal chironomid larvae exploit methane-carbon?
Using data from 87 lakes from the arctic to the equator and from both hemispheres, Jones et al. (2008) reported that δ13C values of chironomid larvae varied widely among lakes, with individual taxa ranging from −8‰ to as low as −72‰. At least one-third of the lakes yielded chironomid larvae with δ13C values sufficiently below those of the corresponding surface sediment organic matter to be strongly indicative of MOB in their diets. Thus, some contribution of methane-carbon to chironomid larvae biomass and hence to lake food webs is probably a rather widespread phenomenon. Almost one-third of the lakes contained chironomid larvae with an estimated 40% or more of their carbon biomass derived from methane, and as much as 70% in some circumstances. Jones et al. (2008) also found that δ13C values of chironomid larvae varied greatly between taxa. Very marked 13C-depletion was evident in Chironomus plumosus, C. anthracinus and C. tenuistylus, which are all taxa characteristic of eutrophic or dystrophic lakes in which late-summer hypolimnetic oxygen depletion near the sediment surface is prevalent. When chironomid δ13C values were related to the late-summer hypolimnetic oxygen concentration near the sediment surface for the 87 lakes, a clear relationship was evident; marked larval 13C-depletion was only found where the oxygen concentration dropped below around 2 mg O2 L−1. A relationship between δ13C of chironomids and oligochaetes and dissolved oxygen concentration was also reported by Hershey et al. (2006) from small arctic lakes in Alaska. Thus, it appears that the low δ13C values of these consumers are associated with low redox conditions, that is with oxic–anoxic interfaces where methane produced under anaerobic conditions can then be oxidised by MOB.
The importance of redox conditions for chironomid larvae feeding on MOB is further illustrated by detailed studies from particular lakes. Low δ13C values were not recorded for C. plumosus larvae from two polymictic, fully oxygenated lakes, Großer Binnensee (Grey et al., 2004b) and Lough Neagh (Kelly, Grey & Jones, 2004), nor from C. anthracinus larvae from Schöhsee (Grey et al., 2004b) or Rostherne Mere (Kelly et al., 2004) which are both summer-stratified but, being only mesotrophic, do not develop pronounced hypolimnetic oxygen depletion. In contrast, both these taxa from four summer-stratified, eutrophic lakes showed strong 13C-depletion, which was more marked for C. plumosus (Kelly et al., 2004) and in two of the lakes, Esthwaite Water and Wyresdale Park, was more pronounced in larvae from greater depths where oxygen depletion developed earlier in the year (Grey, Kelly & Jones, 2004a; Grey et al., 2004b). Moreover, larvae from these two lakes exhibited pronounced seasonal variations in δ13C values which were at least partly linked to the seasonal stratification and oxygen conditions; δ13C values decreased sharply when autumn overturn of the water column restored oxygenated conditions to the deeper sediment surface and presumably stimulated oxidation of accumulated methane by MOB.
The natural abundance carbon isotope data provide persuasive but only circumstantial evidence for incorporation of methane-carbon into chironomid larvae; carbon isotopes alone could lead to ambiguous interpretations, for example when in-lake photosynthesis is based on isotopically light respired carbon generating organic matter that has rather low δ13C values. However, other analyses have provided additional and independent evidence. Deines, Wooller & Grey (2009) were equally able to distinguish between chironomid larvae that had incorporated methane and those that had not from their δ2H values as from their δ13C values. Moreover, the hydrogen and carbon isotope values from chironomid tissue seemed to reflect rather well the source of methane, either acetoclastic methanogenesis (acetate precursor) or hydrogenotrophic methanogenesis (H2/CO2 precursor) according to models of biogenic methane production (Whiticar et al., 1986; Whiticar, 1999; Conrad, 2005). These results were also consistent with evidence from molecular community analysis of methanogens from some of the same lakes (Eller et al., 2005).
Field data have been backed up by evidence from laboratory experiments. Using 13C-labelled methane additions to lake sediments in experimental cylinders with introduced C. plumosus larvae, Deines, Bodelier & Eller (2007a) were able to demonstrate unequivocally that the larvae assimilated methane-derived carbon through MOB. Moreover, phospholipid fatty acids (PLFAs) diagnostic for MOB and significantly 13C-enriched from the labelled methane were detected in the larval tissue. Parallel experiments showed that larvae could also obtain carbon from 13C-labelled type II MOB introduced into the water overlying the sediment. These results confirm that C. plumosus larvae can exploit different food sources by a combination of deposit feeding on sedimented material and filter feeding on particles in the overlying water and help to explain the high intraspecific variation in individual larval δ13C values that is often reported (e.g. Grey et al., 2004a). Other species may also exhibit versatility in feeding reflected in intraspecific variability in δ13C values (Grey et al., 2004a), while differences between species may partly reflect different predominant feeding strategies (Kelly et al., 2004). However, to date, other species have not been investigated as thoroughly as C. plumosus and C. anthracinus.
Why do not other zoobenthos show equivalent utilisation of methane?
Low δ13C values linked to the use of methane have been reported from various zoobenthos taxa, including oligochaetes (Hershey et al., 2006), coleopteran larvae (Kohzu et al., 2004) and trichopteran larvae (Trimmer et al., 2009). However, really marked 13C-depletion has so far only been reported for lake profundal chironomid larvae, which may relate to their mode of life within the tubes that they construct and inhabit in the sediment. Chironomid larvae bioirrigate their tubes by pumping water through them to provide oxygen and food in the form of suspended particles (Walshe, 1947; Jónasson, 2004). Thus, the tubes provide oxygenated microzones penetrating into otherwise anoxic sediment layers, and these oxygenated microzones affect a variety of chemical conditions within the burrows (Lewandowski, Laskov & Hupfer, 2007). MOB can flourish in small-scale opposing gradients of methane and oxygen (Bussmann, Rahalkar & Schink, 2006). Thus, methane produced in the anoxic sediment and which diffuses laterally down concentration gradients can be oxidised within the tubes where suitable conditions for MOB prevail on the tube walls (Brune, Frenzel & Cypionka, 2000), as well as being oxidised at the sediment surface as it diffuses upwards (Fig. 2). In this way, it is easy to understand how the environment around, and particularly within, a chironomid larval tube can provide localised hotspots of MOB production that are then extensively exploited by the larvae. Methane oxidation rates and abundance of MOB have been shown to be higher in chironomid tubes in flooded rice paddies than in the surrounding soil compartments (Kajan & Frenzel, 1999), but to the best of our knowledge this has not yet been shown directly for lake sediments even though there is every reason to expect it to be the case. In fact, chironomid larvae might be considered as ‘constant gardeners’ cultivating MOB food within their tubes, somewhat analogous to Tinodes waeneri caddisfly larvae gardening biofilms on their galleries (Ings, Hildrew & Grey, 2010).
The available evidence is that marked 13C-depletion, indicative of incorporation of methane-carbon, occurs in offshore zoobenthos but not in shallower littoral zoobenthos (e.g. Kiyashko, Narita & Wada, 2001; Grey et al., 2004a; Deines & Grey, 2006; Hershey et al., 2006); this presumably relates to differences in redox conditions and hence in the availability of suitable microhabitats for MOB. However, even in offshore zoobenthos communities, it appears that chironomid larvae often exhibit low δ13C values when other co-occurring taxa do not (Fig. 1; Kiyashko et al., 2001; Hershey et al., 2006; Syväranta, Hämäläinen & Jones, 2006), and this apparent paradox needs to be explained. Some taxa, such as large bivalve molluscs, are essentially filter feeders that rely predominantly on POM sedimenting from the pelagic and will not access MOB in sediments; indeed, large profundal mussels have been promoted as surrogate measures of phytoplankton δ13C (Cabana & Rasmussen, 1996). Smaller bivalves, such as Pisidium spp., feed parallel to the surface and are reputedly bacterial feeders (Jónasson, 2004), but perhaps miss the depth layers where MOB are most prevalent. However, sediment-dwelling tubificid oligochaete worms commonly co-occur with chironomid larvae in the profundal zoobenthos and ingest degraded organic matter from 5 to 10 cm depth within the sediment (Jónasson, 2004), yet according to the limited reports available to date these organisms do not show the marked 13C-depletion commonly reported for chironomid larvae (e.g. Kiyashko et al., 2001). While the reasons for these apparent differences between taxa are still unclear, we suggest that chironomid larvae may become selectively 13C-depleted in mixed zoobenthos communities, because only the chironomid larvae ‘garden’ MOB within bioirrigated tubes or graze particularly at the oxic–anoxic interface on the sediment surface.
Despite the emphasis on lake profundal chironomid larvae, there is evidence that other freshwater zoobenthos can sometimes exhibit low δ13C values suggestive of incorporation of methane-derived carbon. Kohzu et al. (2004) reported that macroinvertebrates collected from backwater pools in a Japanese stream showed large variations in δ13C values. Some taxa, like gammarids and tipulids, exhibited δ13C values close to those of photoautotrophic resources, but taxa collected from inside the detritus accumulation (Helodes sp. and Agabus sp.) showed unusually low δ13C values close to those measured for methane gas. They hypothesised that these low δ13C values reflected methane-derived carbon, with the methane being produced in the anaerobic conditions prevailing in detritus accumulations in stagnant water or in the hyporheic zone. Another case of stream macroinvertebrates with low δ13C values was reported by Trimmer et al. (2009), who found trichopteran larvae (Agapetus fuscipes and Silo nigrocornis) in a southern English chalk stream to be consistently 13C-depleted (less than −40‰) compared to other taxa and their putative basal resources. The river water was supersaturated with methane originating both from groundwater aquifers and from production in fine sediments, and appreciable methane oxidation could be measured from biofilms on stream substrates as well as on the caddisfly cases. Trimmer et al. (2009) estimated that grazing these biofilms including the MOB component provided the caddis larvae with up to 30% of their carbon, although this probably largely represents an external subsidy from an ancient groundwater carbon source, rather than utilisation of biogenic methane produced within the system boundaries.
Does methane-carbon transfer from zoobenthos to higher trophic consumers?
Because chironomids feature prominently in the diets of fish and even birds (Strayer & Likens, 1986; Lindegaard, 1994), and if some chironomid larvae derive an appreciable part of their carbon biomass from methane, it might be expected that chironomids can act as a vector for passage of methane-derived carbon up food chains to higher trophic levels and across ecosystem boundaries. However, to date, there have been surprisingly few attempts to evaluate this, and evidence for methane-derived carbon as a detectable part of higher consumers is almost lacking. Deines & Grey (2006) suggested that the probability of 13C-depleted chironomid larvae being consumed in situ by benthic predators might be low because fish generally do not forage in the oxygen-depleted water layers in which such larvae can become abundant. They proposed that methane-derived carbon is more likely to be transported as a pulse to the terrestrial ecosystem when isotopically light chironomids pupate and emerge as imagos en masse. Certainly many adult aquatic insects move to riparian areas where they are fed upon by terrestrial consumers such as birds, spiders and beetles (Murakami & Nakano, 2002; Paetzold, Schubert & Tockner, 2005; Gratton, Donaldson & Vander Zanden, 2008). Indeed, web-building Araneid spiders became even more 13C-depleted relative to their typical mixed terrestrial and aquatic prey (predominantly aphids, lacewings and caddisflies), following peak emergences of Chironomus spp. from a small, stratifying lake (Fig. 3; J. Grey, unpub. data). Chironomid larvae and imagos during the peak of emergence had a mean δ13C value of −49‰, with individuals down to −72‰ (Deines et al., 2007b). Output from a simple two-source mixing model suggests the spiders inhabiting the immediate reed-fringed riparian zone comprised approximately 18% methane-derived carbon in early June, representing a significant subsidy from the aquatic to the terrestrial ecosystem.
Although 13C-depleted chironomid larvae are most abundant in lake zones where oxygen concentrations are low, some benthic fish species, such as bream (Abramis brama), are able to forage temporarily in hypoxic conditions (Malinin, Kijaško & Vääränen, 1992). Harrod & Grey (2006) reported rather low δ13C values for large bream from a small German kettle lake and estimated that 10–20% of the diet of the fish might have to comprise isotopically light chironomid larvae to account for the low bream δ13C values. Another common and widespread European fish species, the ruffe (Gymnocephalus cernuus), is known to be well adapted to foraging in the conditions prevailing at greater depths in lakes (Bergman, 1988; Holker & Thiel, 1998), potentially enabling greater exploitation of the profundal zoobenthos. Rather low δ13C values for ruffe were reported by Sierszen, Keough & Hagley (1996) and by Tarvainen, Vuorio & Sarvala (2008). More recently, Ravinet et al. (2010) studied ruffe sampled from the littoral, sub-littoral and profundal areas of Jyväsjärvi, Finland, a lake known to contain chironomid larvae with very low δ13C values indicative of methane exploitation. Mixing models indicated the percentage of methane-derived carbon in chironomid biomass to be zero in the littoral, 9% in the sub-littoral and 28% in the profundal. The diets of ruffe from all depths were dominated by chironomid larvae, and a progressive decrease in ruffe δ13C (and δ15N) with depth was observed. Models suggested that in Jyväsjärvi some 12% of the carbon biomass of ruffe caught from >12 m depth, and around 17% of the total ruffe biomass in the lake is ultimately derived from methane. Other fish species sampled from the lake did not show δ13C values low enough to suggest significant incorporation of methane-derived carbon.
Pelagic food webs
In contrast to the situation with benthic food webs, there is still rather little evidence that methane may contribute appreciable carbon to pelagic food webs in lakes. This is not really surprising, as in most lakes the pelagic zone is well oxygenated, methane concentrations are generally very low, and the likelihood that MOB will be sufficiently abundant to provide a significant part of the diets of zooplankton must be considered small. However, there are some circumstances in which these generalisations do not apply and when some contribution of methane to pelagic food webs might be possible. These favourable circumstances can be expected to arise in strongly stratified lakes in which the hypolimnion becomes anoxic, owing to oxidation of copious amounts of organic matter as a consequence of eutrophy or dystrophy. Under such conditions, methane concentrations in the hypolimnion can be high, and considerable methane oxidation can occur at the oxic–anoxic interface in a rather narrow depth layer around the metalimnion where high biomass of MOB may be found (e.g. Bastviken et al., 2004; Kankaala et al., 2006a, 2007b). If some zooplankton are able to access this layer of MOB, for example by vertical migrations in the water column, it could make an appreciable contribution to their diet. A surge in methane oxidation rates can occur during the autumnal overturn, when methane accumulated in the anoxic hypolimnion is mixed with oxygenated water (Kankaala et al., 2006a, 2007b), and then more abundant MOB may become temporarily available to a wider range of zooplankton.
What evidence is there that crustacean zooplankton exploit methane-carbon?
Although lake zooplankton are typically considered as herbivorous consumers of phytoplankton, it is known that bacteria can constitute a substantial part of zooplankton diets (Kankaala, 1988), and in this way carbon of terrestrial origin can be incorporated into zooplankton biomass (Jones et al., 1998; Grey, Jones & Sleep, 2001; Carpenter et al., 2005). Kankaala, Eller & Jones (2007a) reported that, in laboratory experiments with water from a small, polyhumic lake, densities of MOB and rates of methane oxidation were reduced in the presence of the large bacterivorous cladoceran, Daphnia longispina, providing indirect evidence that Daphnia do graze on MOB. Jones & Lennon (2009) also showed that rates of methane oxidation in laboratory experiments were 25% lower in the presence of Daphnia than in control treatments from which Daphnia were absent. Kankaala et al. (2007a) and Jones & Lennon (2009) speculated that, in some circumstances, grazing on methanotrophic bacteria by bacterivorous zooplankton might significantly suppress methane oxidation in situ, and thus that grazer density could influence CH4 efflux from lakes to the atmosphere. However, this has yet to be demonstrated in practice, and the quantitative importance of zooplankton grazers in modifying microbially-mediated ecosystem functions and biogeochemical cycles, such as methanotrophic activity and CH4 effluxes, needs further study under field conditions.
Nevertheless, these ‘alternative’ food sources may not provide such high quality food as phytoplankton, nor support such rapid growth and reproduction of zooplankton, partly because they may lack essential fatty acids that cannot be synthesised de novo by zooplankton (e.g. Brett et al., 2009). In this context, it is reasonable to ask first if MOB can provide suitable food for zooplankton. Evidence to answer this question is still extremely sparse, but Kankaala et al. (2006b) measured growth rates of D. longispina in replicate laboratory cultures fed microbial suspensions with or without enrichment by biogenic methane. They found that the δ13C values of Daphnia indicated consumption of 13C-depleted MOB, while growth rates, survival and reproduction of Daphnia in cultures enriched with methane were equal to or greater than those in non-enriched cultures. They concluded that Daphnia growth was principally a function of food quantity and that food quality was not markedly reduced by inclusion of a high proportion of MOB relative to algae in the diet. Of course, whether Daphnia or other zooplankton taxa could grow as well, or even at all, on a diet of MOB alone remains uncertain, but in any case such a scenario would be unrealistic in situ.
Low δ13C values, suggestive of grazing on MOB have been reported from field studies (e.g. Jones et al., 1999; Bunn & Boon, 1993). However, it is important to recognise that rather low δ13C values of zooplankton might arise through other pathways than consumption of MOB. For example, in lakes with high DOC concentrations, a substantial part of the dissolved CO2 may originate from respiration of terrestrial organic matter and have low δ13C (−15 to −20‰: Lennon et al., 2006; Kankaala et al., 2010). The magnitude of photosynthetic carbon fractionation by phytoplankton with respect to CO2 is uncertain and extremely variable, but in lakes appears to be commonly in the range 0–15‰ with values near the upper end of the range probably most widespread (Bade et al., 2006). Hence, δ13C values for lake phytoplankton of less than −30‰ are quite possible, and zooplankton feeding selectively on these 13C-depleted phytoplankton can be expected to exhibit low δ13C values (Pel, Hoogveld & Floris, 2003). Lennon et al. (2006) sampled north-eastern U.S.A. lakes representing a gradient of terrestrial-derived DOC during late summer (July to early September). They used naturally occurring δ13C values of CO2, POM and crustacean zooplankton, as well as gas measurements and culture-independent assessments of microbial community composition, to make inferences about the flow of terrestrial carbon in the lake food webs. Stable isotope ratios of POM and zooplankton decreased with DOC and were often depleted in 13C relative to terrestrial carbon, indicating the importance of an isotopically light carbon source. Lennon et al. (2006) found only weak evidence for the hypothesis that low δ13C values for zooplankton were attributable to incorporation of biogenic methane; they instead concluded that the observed low δ13C values for zooplankton could be adequately explained by consumption of 13C-depleted phytoplankton, which increased their use of heterotrophically respired and 13C-depleted CO2 with increasing concentrations of terrestrial derived DOC. Jones & Lennon (2009) did find MOB (3% of the bacterial community) in the water column of a humic lake in Michigan, but found no evidence that these were grazed by Daphnia. They therefore concluded that the moderate 13C-depletion of zooplankton in the lake might be explained by CO2 recycling or by consumption of protists that had fed on MOB. Mohamed & Taylor (2009) also found significant correlations between CO2 concentration and epilimnetic δ13C-DIC, δ13C-POC and δ13C-zooplankton in Canadian lakes with variable DOC concentrations. It should also be remembered that oxidation of strongly 13C-depleted biogenic methane generates CO2 that is isotopically light relative to CO2 in lakes derived from other sources, so even consumption of 13C-depleted phytoplankton may, in some instances, constitute an indirect pathway from methane-carbon to zooplankton. Oxidation of methane can account for a high proportion of the excess inorganic carbon accumulation in the hypolimnion of stratified lakes (Houser et al., 2003), but is perhaps less likely to be an important source of the epilimnetic CO2 that is primarily used by phytoplankton. Nevertheless, in view of these uncertainties it is probably unsafe to assume that zooplankton with δ13C as low as −35 or even −40‰ have necessarily been consuming MOB, although of course that remains a plausible explanation. Alternative lines of evidence will be necessary to resolve these arguments.
In contrast to these studies from North America, studies from Europe, and especially from Fennoscandia, have provided support for the hypothesis of incorporation of biogenic methane into pelagic food webs. Bastviken et al. (2003) reported zooplankton δ13C values around −40‰ from three small Swedish lakes and estimated that, even if phytoplankton had δ13C of −30 to −35‰, the low zooplankton values would indicate a fraction of 5–15% of MOB in their diets. Moreover, Sundh, Bastviken & Tranvik (2005) showed that in the same three lakes, type I MOB were often a large component of the pelagic bacterial community and represented a potentially important pathway for re-entry of carbon and energy into pelagic food webs that would otherwise be lost as evasion of CH4. Eller et al. (2005) reported a zone of aerobic methane oxidation (and additionally a deeper zone of anaerobic methane oxidation) in the water column of Plußsee in northern Germany, where high cell numbers of methane-oxidising organisms were also detected by fluorescence in situ hybridisation techniques. It was at this depth in the lake that Santer, Sommerwerk & Grey (2006) found that Diacyclops bicuspidatus consistently maintained highest abundance and exhibited δ13C values approximately 10‰ lighter than epilimnetic species sampled during the same time interval. Furthermore, Harrod & Grey (2006) reported that all zooplankton δ13C values in the same lake fell markedly in the autumn following water column overturn reaching values as low as −45‰, which may have been attributable to seasonal exploitation of MOB at the time of overturn when methane oxidation increased at the same time as phytoplankton production was declining under light limitation.
Probably the most compelling evidence has come from studies of small forest lakes in southern Finland. Taipale, Kankaala & Jones (2007) made additions of 13C-enriched bicarbonate to enclosures in Mekkojärvi, a small humic lake in southern Finland. They found that during summer the 13C-enriched DIC (δ13C up to +44‰) was soon taken up by phytoplankton (δ13C up to −5.1‰) and was transmitted to Daphnia (δ13C up to −1.7‰), demonstrating consumption of phytoplankton. In contrast, during autumn 13C-enriched DIC (δ13C up to +56‰) was not transmitted to Daphnia, whose δ13C became progressively lower (δ13C down to −45.6‰) concomitant with decreasing methane concentration. Although autumn primary production was negligible, zooplankton biomass persisted at the summer level. Outputs from a model suggested methanotrophic bacteria contributed 64–87% of Daphnia diet during autumn, while a calculated carbon mass balance indicated a contribution of 37–112%. Thus, it appeared that methanotrophic bacteria could supply virtually all the carbon requirement of Daphnia in this lake during autumn. Subsequently, Taipale et al. (2008) made sustained whole-lake additions of 13C-enriched bicarbonate to the same lake in spring, summer and autumn. During all three periods, the added inorganic 13C quickly increased the epilimnetic DIC δ13C by between 18 and 21‰ above the initial value of around −21‰, and this 13C-enrichment of DIC was rapidly transmitted to the POM which included photosynthetic phytoplankton. In spring and summer, δ13C of both adult and juvenile Daphnia increased by around 10‰, indicating that Daphnia utilised autochthonous carbon. However, this 13C-labelling of Daphnia was not so obvious during the autumn period, when their δ13C generally decreased. According to IsoSource model (Phillips & Gregg, 2003) outputs based on both δ13C and δ15N values, MOB always made a significant contribution to Daphnia diet in the lake and were probably the single most important food source (around 50% of diet) in autumn.
These labelling studies with 13C-enriched DIC are important because they demonstrate unequivocally that the observed low δ13C values for zooplankton were not attributable to selective grazing of phytoplankton using 13C-depleted DIC from respiration of terrestrial organic matter, as suggested by Lennon et al. (2006); incorporation of biogenic methane via grazing of MOB appears the only plausible explanation in this case. This argument is further corroborated by independent evidence from fatty acid analyses (Taipale et al., 2009), which showed that MOB-specific PLFAs in the seston in Mekkojärvi were highest in autumn when methanotrophic activity was also highest, were also abundant during summer, but made only a minor contribution to total seston PLFAs in spring. The highest magnitude of MOB-specific PLFAs was also recorded in both adult and juvenile Daphnia in autumn. Moreover, a strong relationship existed between the δ13C values of Daphnia reported previously and the proportion of MOB-specific PLFAs in Daphnia. Although most of this evidence derives from a single small humic lake, studies from five lakes in the same geographical location (Kankaala et al., 2010) indicate that MOB can generally make an appreciable contribution to zooplankton diets in these small lakes, especially in autumn, and even in lakes with relatively low DOC concentrations.
However, given that most evidence for a significant role of MOB in zooplankton diets derives from studies on small, steeply stratified, forest lakes, the possible wider generality of the findings is still uncertain, while there does not yet appear to be any firm evidence for significant onward transfer of methane-carbon through pelagic food webs to higher consumers, as is starting to become available for benthic food webs. Moreover, the apparent discrepancy between the evidence from European (and especially Fennoscandian) lakes and that from ostensibly equivalent lakes in north America is puzzling and needs to be resolved. If this discrepancy is real, it might reflect some systematic differences in the physical and chemical conditions in the water column of lakes from these different regions. For example, in the set of North American lakes studied by Lennon et al. (2006), high DOC lakes did not always exhibit hypolimnetic anoxia (see their Fig. 5) and hence may not have offered most favourable conditions for grazing on MOB by zooplankton, whereas the Fennoscandian lakes that have been studied invariably exhibit strong and persistent hypolimnetic anoxia.
The available evidence indicates that incorporation of methane-carbon into benthic food webs is widespread in lakes, but is highly variable between lakes and is most obvious for a few taxa. In contrast, incorporation of methane-carbon into pelagic food webs is probably much more restricted, although this current inference needs to be tempered by recognition that less attention has been paid to methane in pelagic food webs. In this context, we can ask whether it is possible to present any general guidelines as to the conditions that must prevail in lakes to promote a significant contribution of methane to food webs. Biogenic methane is universally produced in anoxic aquatic sediments, although it should be noted that the anaerobic production of methane is influenced by the presence of other electron acceptors, especially sulphate. Hence, in marine environments with generally high sulphate concentrations, the sulphur cycle dominates under anaerobic conditions; in contrast, the carbon-methane cycle is more important in freshwaters where sulphate concentrations are generally lower (e.g. Hobbie, 1988). However, in freshwaters rich in sulphate, the production of methane and the potential contribution of methane-carbon to food webs may be constrained well below what might otherwise be expected from the environmental conditions. In general, as methane enters lake food webs via MOB, which flourish where there are strong opposing gradients of methane and oxygen (at oxic–anoxic interfaces), appreciable contributions of methane-carbon to lake food webs should only be expected when suitable oxic–anoxic interfaces are established within the foraging capabilities of those taxa that have the potential to graze on MOB.
In polymictic lakes with a water column thoroughly oxygenated to the bottom and in stratified lakes that fail to develop hypolimnetic anoxia, methane is produced in the deeper sub-surface anoxic sediment layers. However, much oxidation by MOB of any methane not escaping by ebullition is then likely to take place below the sediment water interface, and only little methane will diffuse laterally into the tubes of chironomid larvae in the shallower sediment layers. In these circumstances (Fig. 4a), there will be no possibility that zooplankton can graze MOB and less opportunity for significant proportions of MOB to be included in the food material obtained by chironomid larvae from the sediment surface or from the inner walls of their tubes. Nevertheless, although both methane oxidation rates and chironomid larval δ13C values are lower in such lakes, methane-carbon can still contribute to larval biomass, perhaps as much as 20% (Deines et al., 2007b). Examples of such lakes with permanently oxic sediment surface layers in which methane-carbon has been shown to make only a limited or negligible contribution to chironomid biomass include Großer Binnensee and Schöhsee in north Germany (Grey et al., 2004a; Eller et al., 2005; Deines et al., 2009) and Lough Neagh and Rostherne Mere from the U.K. (Kelly et al., 2004).
In contrast, in more eutrophic or dystrophic lakes that stratify in summer and develop partial, intermittent hypolimnetic anoxia, the oxic–anoxic interface can develop at or above the sediment water interface (Fig. 4b) where high rates of methane oxidation can be recorded and dense populations of MOB develop that can be grazed by chironomid larvae. In such lakes, chironomid larvae can obtain substantial proportions of their biomass from methane-carbon, although this can vary much between taxa, as well as with depth, which controls the position of the benthic grazers relative to the oxic–anoxic interface and hence to the abundance of the MOB on which they graze (e.g. Grey et al., 2004a,b). In these lakes, the amount of methane diffusing or mixing to the upper oxic water column is small, so that densities of MOB in the upper water column are too low to make a significant contribution to zooplankton diets. Thus, even if some zooplankton in these lake types do graze MOB, this is unlikely to be detectable from their δ13C values, although it might be revealed by the presence of MOB-specific PLFAs in zooplankton.
Finally, some strongly dystrophic or highly eutrophic lakes develop such extensive, prolonged hypolimnetic anoxia that the oxygen conditions in the profundal may be too severe even for the most highly tolerant species of chironomid larvae. In such circumstances, these benthic taxa become restricted to the sub-littoral slopes where the oxic–anoxic interface occurs (Fig. 4c), and in this depth zone they can feed extensively on MOB and show very low δ13C values. Moreover, in these same conditions the oxic–anoxic interface may occur high within the water column, bringing the associated high densities of MOB within foraging range of vertically mobile zooplankton and allowing methane-carbon to enter the pelagic food web (Fig. 4c). Examples of such lakes can be found amongst the small forest lakes studied by Jones et al. (1999), Jones & Grey (2004) and Kankaala et al. (2010), or kettle lakes like Plußsee (Grey et al., 2004a; Harrod & Grey, 2006). Although these mostly tend to be rather small lakes, it should be borne in mind that small lakes are globally extremely abundant (Downing et al., 2006) and also have a high relative storage of methane available for oxidation (Bastviken et al., 2004).
In this context, it is intriguing to note that one of these forest lakes in Finland, Halsjärvi, was the subject of a whole-lake artificial mixing experiment intended to lower the depth of the thermocline during the ice-free period for 2 years, and during the experiment the δ13C values of perch (Perca fluviatilis) in the lake became clearly more negative (Rask et al., 2010). The experimental mixing of Halsjärvi and the consequent lowering of the thermocline and expansion of the oxic part of the total lake volume are likely to have increased the spatial extent of sediment surface oxic–anoxic interfaces, increased the production and availability of MOB to chironomid and zooplankton consumers which were known to be markedly 13C-depleted in this lake (Jones & Grey, 2004; Jones et al., 1999), as well as increasing the oxic ecosystem space available to perch for foraging. All these factors would be expected to increase the potential transfer of isotopically light methane carbon to perch in Halsjärvi, leading to the observed significant reduction in perch mean δ13C values. No equivalent changes were seen in a nearby reference lake.
Thus, it is possible to predict quite logically the kinds of lakes in which an appreciable contribution of methane-carbon to benthic and/or pelagic food webs can be expected. In conclusion, we argue that the recent major paradigm shift in lake carbon and energy sources and their connection to trophic structure (Jansson et al., 2007; Reynolds, 2008) should be expanded to incorporate methane as an additional carbon and energy source (Fig. 5). Following the ‘energy mobilisation’ concept proposed by Jones (1992) and extended by Jansson et al. (2007), we can now recognise three carbon/energy sources: CO2 and solar energy, DOC, and CH4, respectively mobilised for lake food webs by photosynthetic plants (primarily phytoplankton), heterotrophic bacteria and MOB. [Note that this extension is not directly linked to the debate about ‘autochthony’ versus ‘allochthony’ in lakes, as biogenic methane in lakes can in principle originate equally well from autochthonous organic matter (in eutrophic lakes) as from allochthonous organic matter (in dystrophic lakes), and current stable isotope approaches do not allow the ultimate source of the methane to be identified with confidence.] Moreover, although this review has focused on lakes, we have also cited some recent evidence that methane produced in situ or entering via ground water can contribute to river food webs, so the wider importance of methane to freshwater ecosystems in general will need to be reassessed.
We have argued that the evidence to support an extension of the fundamental food web and its potential carbon/energy sources to include biogenic methane is now beyond dispute for lakes and may also need to be considered for some riverine cases. However, studies of the contribution of biogenic methane to aquatic food webs are still in an adolescent state. Among many issues that warrant further study, we highlight the following.
1 More case studies and synoptic surveys are needed to improve understanding of how frequently and in which circumstances methane-carbon makes a substantial contribution to fuelling aquatic food webs. This is particularly true for lake pelagic food webs, as current information derives almost entirely from small, intensely stratified, north-temperate lakes. Information about the role of biogenic methane in lotic food webs is even more limited. Although stable isotope analysis will be a useful general method of assessment, complementary use of specific biomarkers, such as PLFAs, will be necessary to resolve cases where 13C-depletion cannot be unambiguously attributed to incorporation of biogenic methane.
2 Improved estimates are required for the quantitative contribution of methane-carbon to the total carbon budgets and annual production of different lake and river types. Partly this will require more detailed case studies. However, current estimates of the quantitative contribution of methane-carbon to the carbon biomass of metazoan consumers rely on mixing models which are sensitive to the input values for isotopic ‘end members’ and to trophic fractionation of isotope values. More robust end member values for MOB are badly needed in this context. Reliable δ13C values for dissolved methane are available from a wide range of aquatic environments and are relatively easily measured. In contrast, δ13C values for MOB are calculated using carbon isotope fractionation factors between methane and MOB reported from a small number of laboratory experiments. More direct measures of MOB δ13C from aquatic environments would be of great value, although technically very challenging at present. Determining δ13C of MOB-specific PLFAs extracted from aquatic sediments is now possible, but the relations between these values and those of whole cells still need to be better established. In lieu of this, more laboratory studies of how carbon isotope fractionation between methane and MOB varies under different environmental and cell growth conditions would be helpful.
3 Most work to date has focused on the contribution of biogenic methane to selected primary consumers. Of particular interest will be more quantitative evaluations of how much methane-carbon is transferred up the food chain to higher consumers, either within the aquatic ecosystem or into adjacent riparian ecosystems.
4 Considerable inter-individual variation in δ13C, and hence presumably in exploitation of methane-carbon, has been reported for chironomid larvae and may also exist in other taxa. This raises the question of whether individual variability in exploitation of food sources like MOB is a fixed trait that is subject to selection, leading to different sub-populations associated with particular environmental conditions.
5Hardenbroek et al. (2010) recently confirmed that the 13C-depletion measured for whole chironomid larvae is also recorded in the chitin of the head capsules and demonstrated taxon-specific variation in δ13C of fossil chironomid head capsules from different lake sediments; the same is likely to be true for cladoceran carapaces. Both these tissues are well preserved in sediments and are widely used in palaeolimnological reconstructions. Therefore, it is possible that analysis of their carbon isotope ratio fluctuations through time could provide insight into past changes in the contributions of methane-derived carbon to freshwater food webs, and perhaps even to past changes in methane production in freshwaters and associated methane emission to the atmosphere.
We thank all those persons who have helped us with research on methane over recent years and who have contributed to the development of the views presented here, especially Paula Kankaala, Peter Deines and Sami Taipale. However, we emphasise that the authors accept responsibility for the specific views expressed in this paper. We thank P. Deines and Inter-Research for permission to reproduce Figure 2.