The Role of a Zinc-Based, Serum-Borne Sulphide-Binding Component in the Uptake and Transport of Dissolved Sulphide by the Chemoautotrophic Symbiont-Containing Clam Calyptogena Elongata

Vesicomyid clams are found in many deep-sea reducing environments including hydrothermal vents, hydrocarbon seeps and anoxic sediments. These clams all appear to have endosymbiotic sulphur-oxidizing chemolithoautotrophic bacteria in their enlarged gills, and their ability to feed on particulate material appears to be quite reduced (Boss and Turner, 1980; Turner, 1985; Fiala-Médioni and Métivier, 1986; Fiala-Médioni, 1988). The endosymbionts appear to be the primary nutritional source for these clams (Felbeck et al. 1981; Rau, 1981; Cavanaugh, 1983; Fiala-Médioni and Métivier, 1986; Fisher et al. 1988a; Fisher, 1990). The best-known vesicomyid is the giant hydrothermal vent clam Calyptogena magnifica Boss and Turner (Vesicomyidae), which is an obvious and abundant member of the fauna that surrounds deep-sea hydrothermal vents on the Galapagos Rift and East Pacific Rise (Corliss et al. 1979; Hessler and Smithey, 1983; Hessler et al. 1985). The closest living relative of C. magnifica is C. elongata Dall (Boss and Turner, 1980), which also apparently has sulphur-oxidizing endosymbionts in its gills (Vetter, 1985).

Like all animals with sulphur-oxidizing endosymbionts, Calyptogena elongata is faced not only with the formidable problem of surviving in an environment containing highly toxic H₂S, but also of supplying sulphide or another reduced sulphur compound to its intracellular symbionts (Childress and Fisher, 1992). In addition, the clam must remain active in this environment to supply its own O₂ requirements as well as the O₂ and CO₂ requirements of its endosymbionts. The adaptations involved in supplying these needs have morphological, ecological and physiological aspects which set Calyptogena elongata and other vesicomyids apart as having a unique way of dealing with the problems and opportunities of the symbiosis with sulphur-oxidizing bacteria (Fisher, 1990; Childress and Fisher, 1992).

Calyptogena elongata apparently lives with its foot inserted deeply into the mud and its siphons extended into the overlying water. In the Santa Barbara channel at the depths at which C. elongata lives (500m), O₂ levels are quite low in the water (18–20 μmol l⁻¹) and total H₂S and S₂O₃^2-concentrations in the mud are less than about 10 μmol l⁻¹ in the top 10cm (Cary et al. 1989). Thus, the supply of O₂ to the endosymbionts seems obvious, but the mode of uptake of sulphide is more obscure.

The foot which Calyptogena elongata extends into the mud is very extensible and highly vascularized, like that of the vent clam C. magnifica (Arp et al. 1984). The blood of C. magnifica (about 34% of wet tissue mass; Fisher et al. 1988a) has an intracellular haemoglobin of moderate oxygen-affinity and an extracellular sulphide-binding component that can bind sulphide with high affinity (Terwilliger et al. 1983; Arp et al. 1984; Childress et al. 1991b). Arp et al. (1984) have suggested that this species takes sulphide into its blood through its foot. The high affinity of the sulphide-binding component apparently enables C. magnifica to concentrate sulphide from the environment and transport it to the endosymbionts while protecting its tissues from sulphide toxicity (Arp et al. 1984; Powell and Somero, 1986; Childress et al. 1991b). This contrasts with other symbiont-containing clams which lack blood sulphide-binding components and control sulphide toxicity to their tissues by oxidizing sulphide to S₂O₃^2-(a relatively non-toxic substance), which they then transport to their endosymbionts (Anderson et al. 1987; Childress, 1987; Vetter et al. 1989). Because Calyptogena magnifica has substantial sulphide oxidase activity in its foot tissue (Powell and Somero, 1986), vesicomyids may also supply S₂O₃^2-as well as sulphide to their endosymbionts. Although some endosymbiotic sulphur bacteria, such as those of Riftia pachyptila, require sulphide (Childress et al. 1991a; Fisher et al. 1988b, 1989; Wilmot and Vetter, 1990), others, such as those of Bathymodiolus thermophilus, apparently only utilize S₂O₃^2-(Belkin et al. 1986). The endosymbionts of C. magnifica are apparently able to use both sulphide and thiosulphate which the clam supplies to them (Childress et al. 1991b).

The experiments described in this paper document the general properties of this symbiosis, determine the sulphur substrate used by the symbiosis, investigate the properties of the sulphide-binding component, and test the role of the blood in gas uptake and transport in intact, living Calyptogena elongata. In particular, we (1) evaluate the role of the serum sulphide-binding component in concentrating sulphide from the medium, (2) determine whether sulphide, S₂O₃^2-or both are used as energy substrates by the symbiosis, (3) determine whether a metal is involved in the binding component, and

(4) examine the impact of sulphide and endosymbiont autotrophy on blood CO₂ pools and pH. This is the first comprehensive study of the symbiosis of a vesicomyid clam which lives on a mud bottom instead of the rocky substratum of hydrothermal vents.

Throughout this paper sulphide and inorganic carbon refer to these substances without specifying the chemical species involved. Total H₂S and total CO₂ refer to the amounts of these gases analyzed from acidified samples using the analytical methods described below. They are measures of the sum of the various chemical forms in which these substances are found. H₂S, HS⁻, S^2-, S⁰, CO_2, HCO₃⁻and any other chemical formulae refer only to the chemical species symbolized. ‘Free’ refers to that fraction of a substance in the body fluids that is not bound.

The clams used in our studies (Calyptogena elongata Dall) were collected from the bottom of the Santa Barbara Channel using a small clam dredge towed by an oceanographic vessel. The clams were found along the southern side of the channel at depths between 494 and 503m. Within this narrow depth range, the bottom is composed entirely of soft mud as determined both from the dredge samples and from observation from the DSV Alvin. Once at the surface, the clams were quickly transferred to cold sea water (5°C). They were then taken to the University of California at Santa Barbara, where they were maintained in buckets of soft mud supplied with chilled flowing sea water (7°C). The clams usually buried the posteriors of their shells in the mud and assumed a vertical position with their siphons extended from the anterior end. To isolate the clams from sulphide, they were placed in floating baskets in flowing sea water in tanks with no sediments.

Studies involving the maintenance of the clams at known sulphide concentrations were carried out in a 2l flask with a continuous flow of sea water pumped through it at 13mlmin⁻¹ by a peristaltic pump. This sea water was taken from a flowing chilled seawater system. Anaerobic sulphide stock solution (5, 10 or 20mmol l⁻¹ sodium sulphide in sea water) was added continuously by another peristaltic pump to achieve stable sulphide concentrations in the flask. The pH of the water in the flask was maintained at various values between 7.3 and 8.7 by titrating the sulphide stock to an appropriate pH. Water from the experimental flask was periodically sampled with a 0.5ml syringe and the gases were analyzed by gas chromatography (Childress et al. 1984). The pH of the effluent water was measured with a double-junction electrode.

Pieces of gill from freshly captured individuals were fixed in 3% glutaraldehyde in 0.1mol l⁻¹ phosphate-buffered 0.35mol l⁻¹ sucrose, pH7.3, and stored in this fixative at 4 °C for up to 2 weeks. The tissues were then washed in buffered sucrose, post-fixed in 1% osmium tetroxide (on ice) for 1h, dehydrated through a graded ethanol series, and embedded in Spurrs embedding medium. Thin sections were cut using an LKB Ultratome V, stained with uranyl acetate and lead citrate, and examined with a Phillips 300 transmission electron microscope.

Blood samples were taken by quickly removing the clams from the experimental flask, cutting their adductor muscles, opening their shells and withdrawing blood. Blood was drawn from beneath the mantle or the region of the heart using microlitre syringes. This process took less than a minute. All blood analyses were made on whole blood. Gill total H₂S was determined by quickly removing a gill from the clam, weighing it, homogenizing it in a known volume of N₂-equilibrated distilled water and then analyzing 0.5ml of the homogenate by gas chromatography.

Gas chromatographic methods similar to those described by Childress et al. (1984) were used to analyze gases in blood, serum, sea water and other fluids. Briefly, fluid samples were acidified with phosphoric acid and gases were stripped from them using a glass and Teflon extractor, in-line with a thermal conductivity gas chromatograph. This system allowed the analysis of the O₂, CO₂, H₂S and N₂ concentrations in fluid samples of 0.2–1.0ml. The limit of sensitivity for these gases was between 5 and 10 μmol l⁻¹, depending on the gas and the sample size. Throughout this paper, the terms total H₂S and total CO₂ refer to the amounts measured using this analytical method without regard for the chemical species present at the very different pH values and conditions in the clams.

Elemental sulphur (S⁰) in the gills was quantified by gas chromatography according to the method of Richard et al. (1977) as modified by Fisher et al. (1988b). Entire gills (0.15–0.46g wet mass) were dried for 18h in a drying oven at 100°C, and then extracted for 24h with cyclohexane in a micro-Soxhlet apparatus. The sulphur was detected and measured using a thermal conductivity detector. The detection limit for elemental sulphur was approximately 0.2μg atoms g⁻¹ wetmass of the gill (depending somewhat on sample size).

Blood pH was measured by introducing a subsample into a Radiometer glass capillary electrode (Radiometer America G298A) used in conjunction with a reference electrode (Radiometer K171) in a temperature-controlled chamber. Precision buffers (Radiometer S1500 and S1510) were used to calibrate the electrode.

Determinations of thiosulphate in the blood were made by HPLC analysis of monobromobimane-derivatized samples using the methods of Newton et al. (1981) and Fahey et al. (1983) as modified by Vetter et al. (1989). Derivatives were separated on a 15cm long C-18 reversed-phase column and detected using a 235nm filter for excitation and a 442nm filter for detection of fluorescence. The eluent flow rate was 1.5mlmin⁻¹, using an increasing hydrophobic gradient of HPLC-grade methanol and 2% acetic acid. This method was not adequate to detect the bound sulphide in the clam blood since it detected only about 2% of the total H₂S detected by the gas chromatographic method.

Zinc in the blood serum was determined by flame atomic absorption spectrophotometry (Mushak, 1979), using the manufacturers recommended operating procedures, after 250-fold dilution in Milli-Q (18 MΩ) water. A National Institute of Standards and Technology (NIST, formerly NBS) standard reference material (SRM 1643c) run along with the samples was within 3% of the accepted value. The instrument used was a Varian model SpectrAA-400P, equipped with a model PSC-56 autosampler. The protein content of blood serum was estimated by the method of Bradford (1976) using bovine serum albumin as the standard.

The Calyptogena elongata used in these experiments were all collected from a single trawl taken during a cruise in July 1984. All of the clams were of the same size class (1.2–2.1 g wet tissue mass). The incubations were conducted on four different days: 7, 11, 14 and 28 days after collection. All clams were maintained in submerged buckets of mud at 7°C before the preincubations.

Clams were ‘preincubated’ under the incubation conditions for 2–6 days depending on the treatment. Thiosulphate-stimulated animals were preincubated in open buckets with daily changes of chilled sea water (5l) containing 1mmol l⁻¹ thiosulphate. Animals incubated with H₂S were preincubated in flowing sea water in Erlenmeyer flasks with

H₂S (20mmol l⁻¹ stock, pH7.5) mixed into the incoming stream at a rate that produced the desired concentrations of H₂S in the outflowing stream. Control animals were preincubated in floating plastic mesh trays in flowing sea water at 7°C.

Incubations with NaH¹⁴CO₃ were conducted in separate closed vessels containing 400ml of sea water and added substrate, with the exception of the anaerobic clams, which were incubated in their preincubation medium (approximately 1l) with added NaH¹⁴CO₃. Concentrations of the dissolved gases in both preincubation and incubation media were determined by gas chromatography. Specific activities of the NaH¹⁴CO₃ in the incubations were calculated from measured concentrations of total CO₂ and total radioactivity (determined by triplicate counts of media stabilized by hyamine hydroxide) in the incubation media.

All incubations lasted 6.0–6.8h (exact dissection times were noted for each individual and these times were used in the calculations), because pilot trials had indicated considerable individual variation in behaviour and absolute fixation rates in shorter incubations. At the end of the incubation period, clams were removed one at a time, partially drained, weighed, dissected and separated into three tissue classes: gills, blood and remains. These tissues were placed directly into preweighed scintillation vials containing 0.5moll⁻¹ trichloroacetic acid (TCA) (5ml for each gill or remains sample and 1ml for each blood sample). Tissue masses were determined by reweighing the vials. Samples were ground in the TCA and replicate 0.1ml samples were placed in scintillation vials, allowed to stand open for several hours and degassed under a stream of N₂ for 1 min. 10ml of scintillation cocktail was added to each vial and the samples were allowed to stand for 24h before counting. Disintegrations per minute were determined on the basis of the counting efficiency (88–92%) and measured background in similarly prepared samples with unlabelled tissues.

Oxygen consumption rates were measured at 101.3kPa using methods similar to those described previously (Childress, 1971; Mickel and Childress, 1982). Briefly, clams were placed in sealed, temperature-controlled glass chambers with 50mg l⁻¹ streptomycin sulphate added at 7.5°C. Oxygen partial pressure was monitored with a Clark-type O₂ electrode. The water in the chambers was mixed and the stirring needs of the electrodes were accommodated by enclosing the tips of the electrodes in a plastic vial with a magnetic stirring bar and having only a few holes for exchanging water with the rest of the chamber. The chambers were sufficiently large in relation to the clams that 24–48 h was required to exhaust the oxygen.

After each trial, the clam was removed and the chamber was resealed. After adding the minimum quantity of water necessary, a control rate of oxygen consumption was measured. This control rate was always very low (less than 5% of the total rate) and was subtracted from the total measured rate. The rates reported are those recorded after the clam had been in the respirometer for several hours and before O₂ became limiting. The average rate of O₂ consumption at O₂ partial pressures between 4.00 and 9.33kPa was taken as the representative metabolic rate. All rates are presented standardized to wet tissue mass, measured when the specimens were removed from the respirometers.

Serum was separated from whole blood by low-speed centrifugation, which caused settling of the haemoglobin-containing erythrocytes. The serum was then frozen (-70°C) and stored until used. Gill sulphide-binding was determined on samples of homogenized fresh gill tissue.

Sulphide binding was determined by dialysing a sample (either diluted serum or homogenized gill tissue) against a deoxygenated 30mmoll⁻¹ citric acid phosphate buffer (pH7.5 at 5°C), which was brought to the desired sulphide concentration by the addition of solid Na₂S (Childress et al. 1984). Sulphide-binding capacity was determined by dialysing against 2mmol l⁻¹ total H₂S, although for studies of sulphide affinity several total H₂S concentrations were used. The dialysis membrane used had a cut-off of 10000 relative molecular mass. After approximately 18h of dialysis at 5°C, both sample and buffer were analyzed for total H₂S content. The difference between the sample total H₂S and the buffer total H₂S was taken to be the amount of sulphide bound.

Freshly thawed serum was chromatographed on a gel filtration column 2.5cm in diameter and 45cm in length with a total bed volume of 220ml. The filtration medium used was Sepharose 4b, which separates over the fractional relative molecular mass range of 60X10³–20X10⁶. The saline buffer used is described in Terwilliger et al. (1980) and the columns were run at 10°C. Calibrants included Dextran Blue (M_r 2X10⁶), thyroglobulin (M_r 670X10³), gamma globulin (M_r 158X10³) and ovalbumin (M_r 44X10³). The eluent absorbance was measured on a spectrophotometer at 280nm for protein absorbance. Peak absorbances were identified and elution volumes were calculated. Sulphide binding, zinc concentration and protein content were determined on fractions that showed significant absorbance at 280nm as well as on some other fractions.

Enzyme assays were conducted on samples frozen at −70°C for less than 6 months. The samples were homogenized with a ground-glass tissue grinder in 0.2mmol l⁻¹ Tris–HCl (pH7.5) with 1% Triton X-100 (1:9 tissue:buffer). The homogenate was not centrifuged before use in assays, because a substantial portion of the enzyme activity was found to be associated with the pellet. This crude homogenate was assayed for ribulose-bisphosphate carboxylase (RuBp) (EC 4.1.1.39) activity by the ¹⁴C incorporation method (Wishnick and Lane, 1971) as modified by Felbeck (1981). Two types of controls were used: without tissue and without added substrate (RuBp). There was no activity in the incubations without tissue, and the substrate-free control rates (heterotrophic fixation) were subtracted from the experimental rates before calculation of RuBp carboxylase activity.

The stable carbon isotope composition (δ¹³C) of the tissues was determined from tissue samples taken from live clams. The samples were freeze-dried and acidified to remove possible carbonate contamination prior to combustion. Tissue samples were prepared in a Craig-type combustion system with CO ₂ determination on a Finnigan MAT 251 isotope ratio mass spectrometer. Values are reported as δ¹³C ‰ relative to the PeeDee belemnite standard.

The endosymbionts are contained within bacteriocytes in the gills of the clams (Fig. 1). These bacteriocytes are bounded by an exterior surface covered with microvilli on the outer side and by the haemal spaces on the inner side (Fig. 1A). The bacteriocytes are densely packed with endosymbionts, which are abundant throughout the cells except at the inner surface. Bodies that appear to be lysosomes are visible within the bacteriocytes near their inner surfaces. The endosymbionts themselves are polymorphic, about 1 μm in maximum dimension, and often occur several to a vacuole (Fig. 1B). There are numerous electron-lucent inclusions, which are areas where S⁰ deposits (Vetter, 1985) in the periplasmic spaces of the endosymbionts were extracted during tissue processing. These anatomical features of the endosymbionts and the clams are generally similar to those described for other vesicomyids, except that this small clam species appears to have much smaller bacteriocytes (about 10 μm between exterior and blood) than those in larger species such as Calyptogena magnifica (about 50 μm between exterior and blood) (Fiala-Médioni and Métivier, 1986; Fiala-Médioni and Felbeck, 1990; Fiala-Médioni and Le Pennec, 1988).

Morphometric measurements were made on five clams. The gills accounted for 15.7±0.021% (S.E.) of the live tissue mass (total mass drained of water held in mantle cavity minus shell mass). The blood accounted for 39.6±0.03% of the live tissue mass.

Ribulose-bisphosphate carboxylase activities were determined on a number of clams, both freshly captured and after maintenance in mud or sea water in the laboratory. One collection of 10 clams was subdivided into three groups, one of which was analyzed after being held in mud for 6 days. These clams had a mean RuBp carboxylase activity of 0.287±0.15 μmol g⁻¹ min⁻¹ (S.D.; N=4). Clams from this same collection kept in mud for 15 days after capture had a mean RuBp carboxylase activity of 0.285±0.129 μmol g⁻¹ min⁻¹ (N=3). Clams from this same collection kept in sea water without access to mud for 15 days after capture had a mean RuBp carboxylase activity of 0.045±0.0008 μmolg⁻¹ min⁻¹ (N=3). Clams from a different collection, which had been kept for 60 days in mud, had a mean RuBp carboxylase activity of 0.0139±0.012 μmol g⁻¹ min⁻¹ (N=4). These data indicate that this species has a moderate RuBp carboxylase activity and that this activity is lost during captivity in mud and even more rapidly when the clams are kept without access to reduced sulphur compounds. δ¹³C ratios were determined for two Calyptogena elongata individuals. The values for the gills, foot and mantle of the two individuals were −37.9 and −37.1, −37.6 and −37.3, and −36.8 and −37.8‰ respectively. Several shells combined gave a δ¹³C of −1.3‰.

The metabolic rates of nine Calyptogena elongata were determined at 7.5°C. Their masses ranged from 1.3 to 4.6 g wettissue with a mean of 3.0g. Their rates of oxygen consumption tended to be cyclic, probably reflecting opening and closing of the valves. The mean metabolic rate was 0.47±0.076 μmol O₂ g⁻¹ wettissuemass h⁻¹ (S.E.M.). The mean point at which regulation of oxygen uptake failed occurred at 1.41±0.37kPa

(S.E.M.) with a range of 0.63–2.17kPa.

The data in Table 1 are measurements of the rates of carbon fixation by intact Calyptogena elongata under a variety of conditions. As is typical of bivalves, there is a low rate of fixation in the absence of added sulphur substrate, perhaps representing utilization of stored substrates or nonautotrophic fixation. Thiosulphate does not significantly stimulate fixation. Initial total H₂S concentrations of 50 and 280 μmoll⁻¹ doubled the fixation rates, suggesting that sulphide stimulated autotrophy. Anaerobic conditions stimulated fixation into the entire animal by a factor of about 10, suggesting that anaerobic metabolic pathways had been activated. Clams which had been kept for 26 days in the laboratory and incubated with either 50 μmoll⁻¹ total H₂S or no initial sulphide showed rates of fixation that were neither significantly different from each other nor significantly higher than those of the zero-sulphide clams measured after 5 days. The rates of fixation into their gill tissue followed a similar pattern to their whole-body fixation rates, with between 21 and 65% of the total fixed carbon being found in the gills at the end of the incubation period.

Fixation in the body relative to the gills can provide further information. At 50 μmol l⁻¹ total H₂S, amounts of label in the gills and body were both increased by a factor of about 2, and 46% of the total label was found in the gills. In contrast, the anaerobic conditions without sulphur substrate increased the amount of label in the gills by a factor of about 6, whereas the amount of label in the body increased by a factor of 16, with 31% of the label being in the gills. We suggest that the greater incorporation of label into body tissues under anaerobic conditions supports the contention that the anaerobic fixation is achieved by clam metabolic pathways. The fixation under anaerobic conditions clearly indicates that the measured fixation does not represent net fixation of inorganic carbon because C. elongata under similar conditions show a net production of total CO₂ (see Table 5). At 280 μmol l⁻¹ total H ₂S, the gill fixation rate was increased by only a factor of 1.5 while the body fixation rate was increased by a factor of 2.5, suggesting that this sulphide concentration was beginning to stimulate anaerobic pathways in these clams. Thus, it is apparent that the stimulation of anaerobic pathways due to low-oxygen or high-sulphide conditions in incubations may seriously bias the results of ¹⁴C uptake experiments with sulphur-oxidizing symbioses.

Serum separated from the whole blood was initially cloudy in appearance. Upon freezing, a precipitate often formed. Dialysis (relative molecular mass cut-off of 10000) against buffer at pH7.5 resulted in the disappearance of the precipitate. The remaining sample retained all of the sulphide-binding activity.

The sulphide-binding capacities of the sera of three clams and the gills of four clams were determined by dialysis against 1mmol l⁻¹ total H₂S in a deoxygenated citric acid–phosphate buffer at pH7.5. The values for sera ranged from 12.0 to 18.4mmol l⁻¹ total H₂S, with a mean of 15.1mmol l⁻¹. The values for gills ranged from 8.5 to 13.2mmol l⁻¹ total H₂S, with a mean of 10.4mmol l⁻¹. The pooled sera of a different collection of five clams had a capacity of 10.8mmoll⁻¹ (Table 2).

To determine a sulphide-binding affinity curve for the sulphide-binding component, a serum sample of 6.5ml pooled from four clams was diluted 1:9 with the dialysis buffer and then divided into 3ml samples, which were frozen. Each sample was then dialysed for 18h with a different total H₂S concentration between 0.3 and 1060 μmol l⁻¹ total H₂S (Fig. 2). This diluted serum started out at about 0.5mmol l⁻¹ total H₂S, and we were only able to lower the sulphide level slightly by prolonged dialysis or flushing with N₂. These techniques quickly remove sulphide from the sulphide-binding haemoglobin of Riftia pachyptila (Childress et al. 1984). Thus, although the sulphide binding appears reversible in vitro at physiological pH, the dissociation appeared to be quite slow when the binding component is well below saturation. However, when saturated with sulphide, it readily releases sulphide when dialysed against sulphide-free buffer. Other results presented here indicate that sulphide binding is reversible in vivo at pH values near neutrality at all levels of saturation (see Table 4) and that there is little or no residual sulphide associated with the binding component (starting conditions in Fig. 3). Treatment with acid, pH<3, did release sulphide in vitro, but we were concerned that this might alter the binding affinity. The binding data presented in Fig. 2 show the total H₂S in the dialysed samples equilibrated with the different concentrations in the buffer. Because we could not start at a concentration near zero, we consider this to be a preliminary description of the binding properties of this component. These data indicate that this sulphide-binding component has a high affinity for sulphide, reaching about half (1mmol l⁻¹ total H₂S in the diluted serum) of its capacity at external total H₂S concentrations of about 0.02mmol l⁻¹.

In one experiment, native serum was eluted from the Sepharose 4b gel filtration column in only two peaks with absorbance at 280nm. Peak A voided from the column indicating a M_r greater than 20X10⁶. Peak B is approximately 850X10³ in relative molecular mass. These two peaks, along with the original serum, were tested for sulphide binding according to the standard protocol. In this experiment, the diluted native serum accumulated 4.7mmol l⁻¹ total H₂S, combined fractions containing peak A accumulated 2.1mmol l⁻¹ total H₂S and the fractions containing peak B accumulated no sulphide at all. This experiment demonstrates that sulphide-binding ability in this serum is associated with a very high molecular mass substance or a large molecular mass aggregate of smaller subunits.

In a second experiment, 0.5ml of native serum was applied to the column and the effluent was collected in 3ml fractions through a volume of 400ml. In this trial, peak A (the forefront peak) was the only elutant that absorbed at 280nm. The absorbance spectrum of the fractions containing peak A showed very strong absorbance below 220nm, with absorbance increasing below 290nm. Those fractions containing peak A, as well as 20 additional samples taken from other points in the run, were analyzed for sulphide binding, protein content and zinc content (Table 2). The sulphide binding was again found only in peak A, with recovery of 96.5% of the activity present in the native serum in the fractions from peak A (Table 2). Further, zinc was also found only in the fractions of peak A and 91.1% of the zinc applied in the native serum was recovered in these fractions (Table 2). When protein was assayed by the Bradford (1976) dye-binding method, about 100% of the protein in the native serum was recovered in the fractions from peak A. It should also be noted that neither the Lowry nor the bicinchonic acid methods of protein analysis react with fractions from peak A, suggesting that, if this is a protein, it is an unusual one. Further, given the well-known variability in response of the Bradford assay to different proteins, the data presented in Table 2 should be considered to be relative and not absolute. These data indicate that sulphide-binding and zinc concentration are solely associated with a very large molecular mass substance(s) in the serum of these clams. Further, the molar quantities of zinc and bound sulphide appear to covary, with a ratio of 1.20 total H₂S per zinc being found in the native serum. The average ratio of total H₂S per zinc in eight serum samples dialyzed against 2mmol l⁻¹ total H₂S was 1.14±0.29 (95% confidence interval). The similarity of molar concentrations strongly suggests that zinc is involved in the binding of sulphide.

In another set of experiments, we tested the involvment of zinc in sulphide binding (Table 3). In the first experiment, a sample of native serum was diluted 1:9 with buffer and equilibrated with sulphide. This sample was then split into two subsamples. One of these had dry EDTA added to a concentration of 30mmol l⁻¹. This resulted in the immediate evolution of H₂S gas, which could be detected by its strong, distinctive smell. Both samples were then dialyzed against 2mmol l⁻¹ total H₂S. Absorbance spectra were also measured for both samples. The EDTA-treated sample had lost most of its sulphide-binding capacity and most of its zinc compared to the untreated sample, but both still had about the same amount of ultraviolet-absorbing material (Table 3). This indicates that the EDTA removed only the zinc and the concomitant removal of sulphide-binding capacity strongly suggests that the active site contains zinc. In the second experiment, we treated a similar diluted serum sample with the same concentration of EDTA and then split the sample. Both subsamples were dialyzed for 24h against zinc-free buffer. Dry ZnCl₂ was then added to one to a concentration of 60mmol l⁻¹. Upon addition of the ZnCl₂, a large amount of preciptate was formed and quickly settled. Zinc acetate produced similar results (data not shown), so this is not an effect of the chloride. Virtually all of the ultraviolet-absorbing material was removed by this treatment. Both subsamples were then dialyzed against zinc-free buffer for 72h before being dialyzed against 2mmol l⁻¹ total H₂S for 18h. The subsequent dialysis did not reduce the amount of precipitate, restore the ultraviolet-absorbing material to solution or restore the sulphide-binding activity (Table 3). The precipitate contained a high concentration of zinc even after repeated washing with distilled water. It appears that the preciptate is the high molecular mass substance and that it does bind very strongly to Zn²⁺. The peculiar solubility properties of this substance may well be caused by the tendency to link the molecules together through zinc.

To evaluate the rate at which C. elongata can concentrate sulphide into its blood, clams that had low blood total H₂S concentrations were equilibrated with a low sulphide concentration and killed at intervals (Fig. 3). A group of 15 Calyptogena elongata were kept without exposure to sulphide for 6 days. After this, there was little sulphide left either in the blood (<20 μmoll⁻¹) or gills (300 and 450 μmol l⁻¹) of the two clams which were analyzed before the introduction of sulphide (time zero data in Fig. 3). The remaining clams were placed in a flowing-water system with 65 μmoll⁻¹ total H₂S at pH8.1. These clams opened and extended their siphons and feet quickly. They were removed and their blood was collected for gas chromatographic analysis over the next 12h. The blood sulphide concentration increased very rapidly, greatly exceeding the environmental concentration by the first time point (15min) and stabilizing at 13mmol l⁻¹ total H₂S after 90min (Fig. 3). The blood sulphide level remained stable in this range for the rest of the experiment.

In a similar manner, the gill sulphide concentration also increased rapidly at the beginning of the experiment and stabilized after about 2h (Fig. 3). In contrast, there was no apparent change in blood total CO₂ during this experiment. This experiment demonstrates the ability of these clams to concentrate sulphide rapidly into their blood and gills from their medium.

To test the reversibility of the sulphide accumulation in the blood as well as the dependence of blood thiosulphate and gill elemental sulphur on environmental sulphide, another group of 19 Calyptogena elongata was incubated for 24h in a flowing-water aquarium containing a total H₂S concentration of 65 μmoll⁻¹ and then maintained for varying periods in the absence of sulphide (Table 4). After the incubation with sulphide, the clams had high levels of blood sulphide (5.9mmol l⁻¹), blood thiosulphate (1.5mmol l⁻¹) and gill S⁰ (81 μg atoms g⁻¹ wetmass). In contrast, blood total CO₂ (2.1mmol l⁻¹) was similar to that of sea water, and blood pH was low (7.3).

The data show highly significant declines in blood sulphide, blood thiosulphate and gill S⁰ after removal from exposure to sulphide. In contrast, blood pH rose dramatically to 8.1 after 6 days and total CO₂ declined slightly. The rapid decline in blood total H₂S shows the in vivo reversibility of serum sulphide-binding. The initially high concentration of thiosulphate indicates that the clams oxidize some sulphide to thiosulphate. The decline in thiosulphate level over time suggests that the endosymbionts are able to use this material to support their own metabolism. The decline in gill S⁰ indicates that the endosymbionts are able to accumulate this element when sulphide is available and oxidize it in the absence of sulphide. The change in blood pH indicates, as suggested elsewhere in this report, that there is a dramatic, inverse relationship between blood pH and blood total H₂S.

The ability of Calyptogena elongata to concentrate sulphide from their environment was studied by measuring the steady-state distribution of sulphide in the clams’ blood after 24h of exposure to a given set of conditions in a flowing-water aquarium. Much of the inter-individual variation in sulphide bound under a given set of conditions is probably due to variation in the sulphide-binding capacity of their blood, as this can vary from less than 11mmol l⁻¹ to greater than 20mmol l⁻¹ total H₂S. The data in Fig. 4 clearly demonstrate that the serum sulphide-binding component enables these clams to concentrate H₂S from the environment by a factor of 1–2 orders of magnitude in total H₂S under normoxic conditions at pH8.1. Data collected at pH7.3–7.4 at two sulphide concentrations lie on this line, indicating that this concentration process is not very sensitive to the molecular species of sulphide available externally, because the relative amounts of H₂S and HS⁻vary greatly over this pH range. Data collected at one total H₂S concentration at pH8.6 show slightly higher blood total H₂S concentrations: however, this is within the range of variation about the line and is probably not significant. The blood total H₂S concentrations for the two clams exposed to 1.4mmol l⁻¹ total H₂S appear to indicate that the serum is saturated with sulphide (16.4 and 18.7mmoll⁻¹ total H₂S). All the other data collected under steady-state aerobic conditions show, by comparison with the data for in vitro sulphide-binding (Sulphide-binding component properties section and Fig. 2), that the blood total H₂S concentration in these clams is considerably less than the concentration that would be in equilibrium with the external total H₂S concentration. Thus, the internal free total H₂S concentration would be expected to be well below the external concentration, serving to protect the clams from the toxic effects of sulphide. The depression of the internal total H₂S concentration is probably primarily the result of sulphide oxidation by the endosymbionts or the hosts. This is supported by the observation that, in the one experiment conducted under anoxic conditions, sulphide concentrations in the blood were much higher than under aerobic conditions. Without oxygen, the endosymbionts were apparently unable to oxidize sulphide and, thus, the serum approached equilibrium with the external water. The two Solemya reidi individuals that were tested have sulphide concentrations slightly lower than the external concentration, indicating that they lack a sulphide-binding component in their blood.

The blood pH varied strongly in relation to the external total H₂S concentration, being depressed by up to 1 unit at total H₂S concentrations above 0.2mmol l⁻¹ (Fig. 5). This effect is shown to be a function of blood total H₂S concentration (Fig. 6). There is a pronounced decline in blood pH at blood total H₂S concentrations above 1mmol l⁻¹. At a blood total H₂S of 10mmol l⁻¹ (about 50% saturated with sulphide) the pH is as low as 7.2. Such a decline has previously been reported in Calyptogena magnifica (Childress et al. 1991b).

external total H₂S (Fig. 5), indicating that blood total CO₂ transport is unlikely to be a limiting step in the uptake of inorganic carbon by the endosymbionts. Blood total CO₂ is also independent of blood pH (Fig. 6). Thus, blood total CO₂ does not appear to have any role in the control of blood pH under these conditions.

Sulphide is also concentrated into the gills of Calyptogena elongata from the environment (Fig. 4). The sulphide-binding capacity of the gills appears to be about 10mmol l⁻¹ total H₂S and there appears to be a residual acid-labile sulphide content of between 0.3 and 2.5mmol l⁻¹ total H₂S. Between environmental total H₂S concentrations of 0.02 and 1.4mmol l⁻¹, the gill total H₂S concentration appears to increase as the environmental concentration increases, reaching saturation at about 1.4mmol l⁻¹ external total H₂S. The gill sulphide-binding component concentrates sulphide into the gills at far higher concentrations than are found in the external water. This binding capacity, like the serum sulphide-binding component, approaches saturation at much lower total H₂S concentrations under anoxic conditions that prevent oxidation of sulphide. The two Solemya reidi again contrast strongly with the pattern for C. elongata in that the S. reidi individuals do not concentrate sulphide into their gills, although these clams do have substantial sulphide concentrations in their gills.

Preliminary observations suggested that Calyptogena elongata had a very substantial anaerobic capacity and that sulphide seemed to be produced during anaerobiosis. In contrast, other bivalves with chemoautotrophic sulphur-oxidizing endosymbionts (Bathymodiolus thermophilus, Solemya reidi and Lucinoma aequizonata) do not show consistent, substantial production of sulphide under anaerobic conditions (J. J. Childress, unpublished studies). We placed four groups of three Calyptogena elongata in 125ml Erlenmeyer flasks filled with nitrogen-bubbled water and then sealed (Table 5). The clams used had been in captivity for only about 3 days so they probably still had some S⁰ in their gills. The flasks were sequentially opened and the blood total CO₂ and total H₂S, gill total H₂S and flask total CO₂ and total H₂S were analyzed. The clams survived the longest period tested (164h) with no sign of death or tissue decay. During the anoxia, there was a steady and highly significant increase in all of the measured variables. The blood total CO₂ increased in parallel with that of the sea water in the flasks, with which it appeared to be roughly in equilibrium. The blood and gill total H₂S concentrations increased to nearly 5mmol l⁻¹ during the experiment and appeared to be roughly in equilibrium with each other. The total H₂S concentration in the flask also increased dramatically during the experiment, but it remained much lower than the concentrations in the blood and gills as one would expect from the high-affinity sulphide-binding activities of those tissues. In fact, significant total H₂S concentrations were not measurable in the water until the blood total H₂S concentration approached about 25% saturation with sulphide.

The source of the sulphide in this experiment is of particular interest because it could be produced either by the endosymbionts or by free-living sulphate-reducers. The sulphur compound that is being reduced is also in question because S⁰, S₂O₃^2-and SO₄^2-are all present in the clams, and sulphate is present in the sea water as well. Given the quantity of sulphide produced, it must come from the reduction of either sulphate or S⁰. This can be illustrated in the following example. If we assume that each of the three clams in each flask had a wet mass of 2g (a typical value for the size of individuals used), contained 81 μg atoms g⁻¹ elementalsulphur in gills that weighed 0.3g, and contained 1.5mmol l⁻¹ S₂O₃^2-in 0.8ml of blood, then the total of these two sulphur sources in each flask would be 76.5 μmol, with 95% coming from the S⁰. Using the same morphometric assumptions and the 164h data in Table 5, one can estimate that 53.8 μmol of sulphide was produced in the 164h flask. The uncertain factor here is the S⁰ content of the gills, which has the potential to be a major source but was likely to be quite reduced because the clams had been in captivity for 3 days without exposure to sulphide before the beginning of the experiment. The source of reducing power is clearly the organic material within the bodies of the clams, because the sea water used was clean, the clams’ shells had been cleaned and the clams produced no faeces.

We believe that the endosymbionts are the most likely sulphur reducers in this experiment because there was no large source of organic material outside the bodies of the clams and they did survive the experiments. Furthermore, similar experiments with other chemoautotrophic bivalves have failed to show similar results, suggesting that the endosymbionts of this species are different in this regard.

Vesicomyid clams with their endosymbionts are widely distributed chemoautotrophic symbioses. In addition to the well-known vent clam Calyptogena magnifica, there are other vent species and many other species in other deep-sea reducing habitats (Turner, 1985; Fisher, 1990). All of the adequately studied vesicomyid species appear to have sulphur-oxidizing endosymbionts in their gills (Brooks et al. 1987; Somero et al. 1989; Fisher, 1990). All of the species which have been studied appear to live in environments characterized by relatively low levels of sulphide and a significant spatial separation of the sulphide-containing waters or muds from the overlying oxic waters. The vesicomyid clams appear to bridge these different zones to supply both reduced sulphur compounds and oxygen to their endosymbionts (Arp et al. 1984). It has been suggested that the sulphide is taken up across the foot (which is embedded in the reducing environment), bound to a serum sulphide-binding component and transported to the gills, whereas oxygen is taken up across the gills directly from the ambient water (Arp et al. 1984; Childress et al. 1991b; Childress and Fisher, 1992). The present study of Calyptogena elongata provided an opportunity to evaluate the adaptations of a soft-bottom-living vesicomyid as well as to test some of the hypotheses concerning vesicomyid functioning.

The Calyptogena elongata symbiosis studied here appears similar to other vesicomyid symbioses in general properties. Their endosymbiotic bacteria are similar in size and appearance to those previously observed in other vesicomyids (Fiala-Médioni and Métivier, 1986; Fiala-Médioni and Felbeck, 1990; Fisher, 1990). The bacteria are located in similar bacteriocytes between the external medium and the clam’s blood, except that the bacteriocytes are considerably smaller in C. elongata (about 10 μm maximum dimension) than in larger species such as C. magnifica (where they are about 50 μm maximum dimension). The C. elongata endosymbionts accumulate elemental sulphur, as is typical of other vesicomyid endosymbionts and indicative of sulphur oxidation by the endosymbiotic bacteria (Brooks et al. 1987; Fisher, 1990; Fisher et al. 1988a; Somero et al. 1989; Vetter, 1985). The C. elongata endosymbionts also have significant RuBp carboxylase activities, indicating that they are autotrophic and employ the Calvin–Benson cycle, as do other vesicomyid endosymbionts. The rapid decline in the activity of this enzyme when the endosymbionts are deprived of a sulphide source is typical of what has been found in other sulphide-oxidizing symbioses (Anderson et al. 1987). The distinctive δ¹³C values of the gills, foot and mantle of this species (−36 to −37‰) are typical of those found in other vesicomyid species (−30 to −40‰) and strongly support the hypothesis that the primary carbon source for this symbiosis is carbon autotrophically fixed by the endosymbionts (Fisher, 1990). The small differences in δ¹³C among the tissues in C. elongata indicate that there is little input of other carbon into this symbiosis, as studies of gut contents have shown for other vesicomyids (Fiala-Médioni and Felbeck, 1990; Fiala-Médioni and Le Pennec, 1989).

Recent studies of C. magnifica have indicated that the endosymbionts in a crude gill homogenate can use either sulphide or thiosulphate as a substrate to support the fixation of inorganic carbon (Childress et al. 1991b). The present study of fixation by whole C. elongata indicates that low external sulphide concentrations apparently stimulate fixation of inorganic carbon but that externally supplied thiosulphate does not, suggesting that the only external sulphur species utilized by the symbiosis is sulphide. However, the high concentrations of thiosulphate found in the blood of C. elongata exposed to sulphide, and the rapid decline of these concentrations when the clams are deprived of sulphide, suggest that, like C. magnifica (Childress et al. 1991b), this species oxidizes some sulphide to thiosulphate and that the endosymbionts can use this blood-borne thiosulphate. The rapid decline in S⁰ in the gills of C. elongata kept without sulphide also parallels findings with C. magnifica and indicates that the endosymbionts oxidize this material when deprived of other reduced sulphur species (Childress et al. 1991b).

The serum sulphide-binding component of vesicomyid clams has a high affinity and capacity for sulphide, but appears to be an unusual molecule. It is not destroyed by boiling or treatment with proteases (Arp et al. 1984). It has unusual solubility properties, tending to precipitate when frozen and going back into solution upon dialysis against a variety of buffers. It absorbs light strongly in the ultraviolet, but primarily at shorter wavelengths than is characteristic of proteins. As shown here, it is a very large molecule which only reacts with some protein assays. This indicates that it is likely to possess some amino acids, but must have other components as well. The findings presented here that the serum contains extremely high concentrations of zinc (8.95mmol l⁻¹ in one C. elongata) and that this zinc is closely associated with the high molecular mass fraction, which also contains the sulphide-binding component, strongly suggest that the sulphide is bound to zinc within the component. The findings that treatment of the serum with EDTA results in the evolution of sulphide, the freeing of zinc and the removal of sulphide-binding activity without removing the characteristic ultraviolet-absorbing material indicate that the zinc is not only involved in the binding of sulphide but is also itself attached to the larger molecule. The co-precipitation of this large molecule with zinc when zinc is added indicates that zinc strongly interacts with this molecule and may be involved in its unusual solubility properties. The approximately 1:1 ratio of bound sulphide to zinc in the serum and purified fractions also supports the conclusion that the active site of the sulphide-binding component includes a zinc ion.

The highly concentrated zinc-based sulphide-binding component in the serum of this vesicomyid clam also explains the anomalously high zinc concentrations found in the tissues of Calyptogenamagnifica (Roesjadi and Crecelius, 1984; Roesjadi et al. 1985) but not in other vent molluscs (Smith and Flegal, 1989). The whole soft tissues of C. magnifica were found to contain 2.152mgzinc g⁻¹ drymass, far higher than the content of any other metal and approached only by iron, which is in the haemoglobin of this clam (Roesjadi and Crecelius, 1984). Assuming that 80% of the clam’s mass was water, we can calculate an approximate zinc concentration of 6.6mmol l⁻¹ in the living clams. This value is in reasonable agreement with the concentration in C. elongata. Roesjadi et al. (1985) further showed that the zinc in C. magnifica was associated with a large molecular mass fraction in their gel column separations. Thus, the high zinc concentrations in these two vesicomyids appear to result from the presence of the zinc-based sulphide-binding component in their blood serum.

The experiments presented here provide new information and tests concerning the role of the serum and gill sulphide-binding components in the uptake and transport of sulphide in vivo. These experiments not only demonstrate the reversibility of the serum binding in vivo, by loss of sulphide from the blood when the clams are deprived of sulphide (as previously shown for C. magnifica, Childress et al. 1991b), but also demonstrate loss from the gill tissues as well as rapid uptake into the blood and gills when sulphide is supplied (Fig. 3). Thus, the sulphide-binding activities in both of these tissues appear to be readily reversible in vivo and therefore able to function in the transport of sulphide to the endosymbionts.

The experiments at steady-state sulphide conditions (Fig. 4) indicate that these binding activities are able to concentrate the sulphide greatly from the environment. This property is especially important in the low-sulphide environments occupied by most vesicomyid clams. This concentrating ability does not appear to be affected by the external pH. However, the availability of oxygen allows the clams to hold their internal total H₂S concentrations well below saturation of the sulphide-binding components (Fig. 4), enabling them both to concentrate sulphide from the environment and to protect their tissues and endosymbionts from sulphide toxicity (Childress et al. 1991b). The limited tolerance of vesicomyids for sulphide is demonstrated by the apparent stimulation of anaerobic metabolism at higher sulphide concentrations (Table 1), a situation similar to that documented for the clam Solemya reidi and its endosymbiotic sulphur-oxidizing bacteria (Anderson et al. 1990).

As in the case of C. magnifica, the in vivo binding of sulphide in the serum of C. elongata is associated with a decline in blood pH at higher blood sulphide concentrations (Figs 5 and 6). This is unlike the binding of sulphide by Riftia pachyptila haemoglobin (Childress et al. 1991a), which is not associated with a shift in blood pH, suggesting that different binding mechanisms are involved although the functions are apparently similar. An alternative possibility is that the animals regulate their blood acid–base balances differently in response to elevated sulphide concentrations. Another difference between the R. pachyptila symbiosis and vesicomyid symbioses is that the blood inorganic carbon concentration in the tubeworm is strongly affected by the autotrophic status of the symbiosis (Childress et al. 1991a), whereas in the vesicomyids this concentration appears to be independent of the autotrophic status (Figs 5 and 6). It has been suggested that this results from the greater variety of sources for inorganic carbon for the vesicomyid endosymbionts as well as their lower rates of demand (Childress et al. 1991b).

The data presented here and in the cited papers enable us to conclude that all of the known vesicomyid symbioses appear to function in similar ways. The present studies have provided some new tests which support previously presented models of these symbioses and add new information about their functioning. These clams typically live in areas where there is spatial separation between mildly reducing conditions in sediments or venting water and oxic conditions in the overlying water. They physically bridge this spatial separation to make reduced sulphur compounds and oxygen available to the endosymbionts in their gills. Sulphide is taken up across the foot and either bound to the serum sulphide-binding component or oxidized to thiosulphate. The reduced sulphur compounds are then transported to the gills, where they either diffuse or are transported to the endosymbionts. The gill sulphide-binding activity probably has a role in the transport of sulphide from the blood to the endosymbionts. Oxygen is apparently taken up directly into the bacteriocytes from the ambient water, which is moved over the gills by ciliary pumping. Its transport to the bacteria may be facilitated by tissue haemoglobin within the gills (Wittenberg, 1985). Inorganic carbon is probably taken up from the medium as well as from the blood, with the main sources being the ambient sea water, the clam metabolism and the interstitial or vent water. Vesicomyids appear to depend virtually entirely on their endosymbionts for reduced carbon.

This model contrasts sharply with those of other chemoautotrophic symbioses (Anderson et al. 1987; Childress and Fisher, 1992; Fisher, 1990) which, although facing common problems of substrate supply and toxicity avoidance, have evolved very different mechanisms for accomplishing these ends. Chemoautotrophic mussels, for example, seem to have evolved little specialization for particular endosymbionts or substrates, as they harbour either methanotrophic or sulphur-oxidizing symbionts, but oxidize sulphide to thiosulphate in both cases and lack specialized mechanisms for transporting sulphide to their symbionts (Childress et al. 1986; Childress, 1987; Fisher, 1990; Childress and Fisher, 1992). These mussels live at the interface between the reducing and oxidizing environments upon which they depend and apparently have minimal adaptations for bridging these environments.

The solemyid bivalves live in environments high in total H₂S concentration and are highly specialized for dealing with sulphide, but are organized very differently from the vesicomyids in that they do not concentrate sulphide from their environment and rely extensively on the oxidation of sulphide to thiosulphate, which can then be supplied to their endosymbionts (Powell and Somero, 1985, 1986; Anderson et al. 1987). The solemyids have tissue haemoglobins in their gills that may facilitate the movement of sulphide and oxygen to the endosymbionts (Doeller et al. 1988) but, as we show here, they do not appear to concentrate sulphide from the environment. The solemyids appear to bridge reducing and oxidizing zones by ventilating and moving in their burrows, which extend from the surface of the sediments to the highly reducing zones deeper in the sediments (Fisher, 1990).

Vestimentiferan tubeworms, in contrast, are highly specialized to supply only sulphide to their endosymbionts by using haemoglobins that bind sulphide (Arp and Childress, 1983; Childress et al. 1984, 1991a; Arp et al. 1987; Fisher et al. 1989; Wilmot and Vetter, 1990). This symbiosis also depends upon blood transport to supply oxygen and inorganic carbon to its endosymbionts, unlike the other symbioses studied to date (Childress and Fisher, 1992).

Vesicomyid symbioses appear to be specialized for sulphide-oxidizing endosymbionts and for habitats where mildly reducing environments are spatially separated from oxic environments.

This work was supported by NSF grants OCE-8609202 and OCE-9012076 to J.J.C. and OCE-8610514 to J.J.C. and C.R.F. We thank A. Anderson, D. Cowles, N. Sanders, J. Zande and M. Wells for assistance with some of the data collection. The stable carbon isotope analyses were performed at the Geological and Environmental Research Group at Texas AandM University, courtesy of J. Brooks. Thanks are also due to the captains and crews of the RV Sproul, RV Point Sur and RV Velero IV for helping us to collect the experimental animals. This manuscript has benefited from discussions with and comments by R. Kochevar.