Acidified waters are impacting commercial oyster production in the U.S. Pacific Northwest, and favorable carbonate chemistry conditions are predicted to become less frequent. Within 48 h of fertilization, unshelled Pacific oyster (Crassostrea gigas) larvae precipitate roughly 90% of their body weight as calcium carbonate. We measured stable carbon isotopes in larval shell and tissue and in algal food and seawater dissolved inorganic carbon in a longitudinal study of larval development and growth. Using these data and measured biochemical composition of larvae, we show that sensitivity of initial shell formation to ocean acidification results from diminished ability to isolate calcifying fluid from surrounding seawater, a limited energy budget and a strong kinetic demand for calcium carbonate precipitation. Our results highlight an important link between organism physiology and mineral kinetics in larval bivalves and suggest the consideration of mineral kinetics may improve understanding winners and losers in a high CO2 world.
 Several years of oyster seed production failures in the U.S. Pacific Northwest have been linked to wind-driven upwelled waters with elevated PCO2 [Barton et al., 2012]. Recent measurements and future projections of carbonate chemistry in the California Current Ecosystem highlight the rapid acidification of coastal waters [Feely et al., 2008; Gruber et al., 2012; Hauri et al., 2012] that are strongly coupled to U.S. Pacific Northwest estuaries [Hickey and Banas, 2003; Barton et al., 2012]. In coastal environments, carbonate chemistry may vary widely both spatially and temporally [Hinga, 2002; Hofmann et al., 2011; Waldbusser et al., 2011] and can result in transient conditions (in excess of global averages) that are unfavorable for sensitive species [Gruber et al., 2012; Hauri et al., 2012]. Many laboratory-based studies have documented the sensitivity of larval bivalves to deleterious carbonate chemistry [Kurihara et al., 2007; Talmage and Gobler, 2010; Miller et al., 2009], and more thermodynamically soluble mineral phases in larval shell is often cited for this sensitivity [e.g., Weiss et al., 2002]. However, sensitivities vary strongly as a function of life-history stage and food availability [Green et al., 2009; Waldbusser et al., 2010; Melzner et al., 2011], or on previous exposure of reproductive adults and larvae to acidified waters [Barton et al., 2012; Hettinger et al., 2012; Parker et al., 2012]. A model of organism sensitivity to ocean acidification based on CaCO3 polymorph solubility alone cannot adequately explain these results. Thus, an integrative examination of larval development, mineral dynamics, and early life-history is required.
 In adult bivalves, the calcifying fluid, where calcium carbonate is precipitated from, may chemically resemble ambient seawater [Thomsen et al., 2010] or deviate significantly [Crenshaw, 1972]. Bivalves are osmoconformers [Shumway, 1977] but do have limited control of their calcifying fluid [Nair and Robinson, 1998]. Because mineral precipitation consumes alkalinity and generates protons, the continued formation of calcium carbonate requires the removal of protons from the calcifying fluid. Active modification of the calcifying fluid is also evidenced in the significant contribution of metabolic carbon to precipitated calcium carbonate in some marine bivalves [Gillikin et al., 2007]. In Crassostrea spp., the unshelled swimming trochophore develops within 6–10 h of fertilization. An outer organic shell layer called the periostracum begins to form immediately following the trochophore stage and mineral calcification is initiated well in advance of its completion [Kniprath, 1981]. A fully formed and calcified larval shell (the prodissoconch I or D-hinge) is completed within 24–48 h [His and Maurer, 1988]. Larval shell formed after the D-hinge stage is distinct and called prodissoconch II shell (and also aragonite).
 In supersaturated seawater, calcium carbonate precipitation is thermodynamically favored and forms at a rate relative to the saturation state, modeled as
 where r is the mineral precipitation rate, k is the rate constant, Ω is the saturation state with respect to the given mineral, and n is the reaction order [reviewed in Morse et al., (2007)]. We can assume equation (1) provides the fundamental kinetic constraint on biocalcification. Most marine bivalve shell is however mineralized at rates that are much faster than abiotic precipitation in super saturated seawater [Zuddas and Mucci, 1998]. Calcifying organisms may overcome this kinetic constraint by the use of organic macromolecules [Elhadj et al. 2006] and by membrane/cellular transport pumps [Cohen and McConnaughey, 2003]. These strategies must also be physiologically and energetically constrained by the degree to which organisms can effectively increase k or Ω.
 We therefore propose two hypotheses to explain why oyster larvae appear particularly sensitive to ambient carbonate chemistry during the development of the initial prodissoconch I (D-hinge) shell: (1) the calcifying fluids and surfaces are more exposed to ambient carbonate chemistry conditions until the D-hinge shell is completed and (2) the kinetic demand (or rate) of shell accretion far exceeds possible abiotic precipitation rates even under thermodynamically favorable conditions, thus, mineral precipitation requires significant energy input at a life history stage when energy is limited.
 We measured the isotopic composition of carbon in larval shell, larval tissue, food, and holding tank water during growth and development of a commercial larval Crassostrea gigas cohort in May 2011. We coupled these measurements to bulk measures of larval biochemical composition to estimate energetic status and demands for the initial shell development.
2.1 Setting and Experimental Design
Crassostrea gigas larvae were raised according to commercial spawning protocols at Whiskey Creek Shellfish Hatchery in Netarts Bay, Oregon, as previously described in detail in [Barton et al., 2012]. Briefly, successfully fertilized Willapa Bay oyster larvae from several conditioned broodstock oysters were added to a 22 m3 tank an hour after fertilization at an estimated density of approximately 5 × 107 ind. m−3. Water was changed in culture tanks every 2–3 days during cohort development. The pH of fertilization water was 8.15 (measured with a YSI 30 on the NBS scale), and developing embryos were reared under a PCO2 of 396 µatm and aragonite saturation state of 2.38. Samples for PCO2 and TCO2 of culture tanks were collected at each tank filling and analyzed following Bandstra et al.  with an uncertainty of less than 5% and 0.2% for PCO2 and TCO2, respectively. The rest of the carbonate system was calculated using standard dissociation constants as previously described [Barton et al., 2012]. Conditions in culture tanks at filling had an average PCO2 (±1 S.D.) of 364 (±40) µatm aragonite saturation state with 2.58 (±0.33) at the culture temperature of 25 °C, and a salinity of 26.5 (±2.29). Larvae were fed a diet of Isocrysis galbani cultured in a continuous culture system during the first week following fertilization, and this was switched to a mixed algae diet consisting primarily of Chaetocerous gracilis, Thalassiorsira spp., and Isochrysis galbana cultured in batches. Estimated cell densities were 5–7 × 1010 cells m−3.
 During each tank change, all larvae were captured on an appropriately sized sieve, and a primary sample of the entire cohort was randomly collected with a clean stainless steel spatula. The mass of each primary sample varied from ~0.5 to 1.0 g representing an estimated 2.8 × 106 to 5.6 × 106 two day old larvae or 2.3 × 105 to 4.6 × 105 19 day old larvae. All 400 million larvae in the cohort were reared collectively in one tank, so replicate primary samples would have been pseudo-replicates of larvae from the same tank.
 The larval samples were quickly rinsed with deionized water, immediately frozen, and freeze dried within 2 weeks with a Labconco FreeZone 6 Freeze Drier. Subsamples from each primary sample were then used for subsequent analyses. Approximately 100s to 10s of larvae for stable isotope analyses of shell carbonates from day 2 and 19 samples were needed, and lipid extractions required roughly 104 to 105 larvae for day 2 and 19, respectively. Analyses of stable isotopes and elemental analyses of tissue required approximately 103 to 104 larvae. Variable volumes (200–800 ml) of algal cultures were vacuum filtered onto pre-combusted onto glass fiber filters (GF/F 47 mm, Whatman) to achieve a similar analytical mass for all analyses.
2.2 δ13C Measurements
 All stable carbon isotope samples were analyzed in Oregon State University's CEOAS Stable Isotope Laboratory and details for standard analyses may be found here (http://stable-isotope.coas.oregonstate.edu/). All runs were calibrated and checked with NIST certified (NBS 19, NBS 20 UQ6) and in-house calibrated standards, with precision and accuracy noted below.
 Water samples were analyzed by continuous mass flow spectrometry on a GasBench-Delta V system. The run had a precision of 0.02‰ and accuracy better than 0.06‰. Larval shell carbonate δ13C was measured by reacting the freeze-dried larvae with a viscous ~105% H3PO4 solution in a Finnigan MAT Kiel III at 70 °C for 5 min after which the isotopic composition of the evolved and purified CO2 was measured on a dual inlet mass spectrometer (MAT 252 IRMS). Each run had a precision and accuracy of 0.02‰ and 0.01 ‰, respectively. Repeatability of larval shell samples was within 0.02‰. To obtain measures of tissue δ13C composition of tissues, freeze-dried larvae were acid fumed with concentrated HCl to eliminate the shell carbonate signal. The remaining sample material was then analyzed by flash combustion (>1000 °C) using a Carlo Erba NA 1500 elemental analyzer interfaced by continuous-flow mass spectrometer (Delta-Plus XL). Repeatability of organic larval samples was 0.21‰. Algae samples, collected as noted above, were analyzed in the same way as larval tissue however without the acid fuming. Measurements of organic δ13C composition were made with a precision and accuracy of 0.17‰ and 0.12‰, respectively.
2.3 Larval Biochemical and Size Analyses
 Fractions of organic tissue and shell (% of dry weight) were determined by first measuring %C on bulk freeze-dried larval samples and then decalcified samples (via HCl fuming) on a Carlo Erba NA1500 elemental analyzer. The difference between whole and decalcified larvae and the molar proportion of C in calcium carbonate were used to estimate the percent calcium carbonate by weight, with the remainder being organic matter. These values for percent organic matter were correlated to (r2 = 0.89) loss on ignition (LOI) measurements, with LOI consistently over estimating percent organic matter. Total lipid percent was measured by extracting freeze-dried larvae with a methylene chloride: methanol solution (3:1) using a Dionex ASE 200 Accelerated Solvent Extractor and weighing an aliquot of the isolated total extractable lipid (TEL) fraction dissolved in CS2 gravimetrically. Larval size was evaluated on 100–150 larvae per sample, taken as a separate subsample during tank water changes and evaluated on digital micrographs using standard image analysis techniques (ImageJ 1.44p). A standard mass at length model for C. gigas larvae [Bochenek et al., 2001] was used to calculate a total individual mass. Measured proportional composition data were then applied to the mass at length estimate to compute per larva weights of shell, bulk organic, and lipid content.
 The δ13Cshell of C. gigas larvae shifted from +1.5 ‰ 48 hours after fertilization (immediately after the prodissoconch I stage) to −1 ‰ by 11 days (prodissoconch II), indicating a shift away from 13C-enriched seawater dissolved inorganic carbon (DIC) to 13C-depleted metabolically respired carbon in the larval shell (Figure 1). Metabolic carbon in shell carbonate doubled from ~7% on day 2 to ~14% on day 14 post-fertilization (Figure 1), calculated following [McConnaughey et al., 1997] as:
 where R is the proportion of metabolic contribution to shell, δ13CResp is the isotope value of respired carbon approximated from larval organic tissue; δ13CDIC is the isotope value of the ambient DIC; δ13Cshell is the isotope value of the shell, and Δ is the abiotic fractionation between aragonite and bicarbonate [Romanek et al., 1992].
 At fertilization, oyster embryos are entirely organic material and by 2 days post-fertilization, the shelled larvae have precipitated ~160 ng CaCO3 larva−1, constituting >90% of total larval mass. The mass-normalized shell precipitation rate in this first 2 days is 0.45 d−1, and decreases by an order of magnitude by day 5 to 0.04 d−1 (Figure 2a). Formation of the initial shell occurs with little to no new growth in tissue; organic mass remained constant from day 2 to 5 at 16 ng larva−1, and lipids do not begin to accumulate until 11 days post fertilization, following a minimum of 2.6 ng larva−1 (Figure 2b).
 Our measurements of δ13Ctissue show the slow transition from endogenous energy sources derived from egg reserves to exogenous food sources for developing bivalve larvae by the large deviation of tissue from algal food sources during the first week following fertilization (Figure 1b). The deviation between food and tissue between days 14–16 is likely related to changes in feeding associated with a change in algal diet, which is concurrent with a decrease in growth and calcification (Figure 2a) and a slight shift in δ13Cshell as would be expected if calcification rate slowed [Gillikin et al., 2007]. δ13Cegg of C. gigas eggs from another cohort produced in the hatchery was −20‰. Assuming a simple two-member mixing model of food sources; egg and algae, and no trophic enrichment [McCutchan et al., 2003], the contribution of egg derived carbon to larval tissue growth is 100% on day 2 decreasing to 86%, 58%, and 0% on days 5, 8, and 11, respectively. These results highlight that initial shell production in oyster larvae is supported by the limited energy reserves of eggs.
4 Discussion and Conclusions
 Greater incorporation of DIC from ambient seawater during D-hinge shell formation (Figure 1a) and the developmental biology of bivalve larvae [Kniprath, 1981] suggest two non-exclusive posits: developing D-hinge larvae cannot isolate the calcifying space as easily as later stage larvae, and under high rates of shell precipitation, more DIC is needed. Both of these indicate that during initial shell formation, the calcification surfaces will be more exposed and dependent on seawater carbonate speciation. Adult bivalves are known to incorporate a small but significant fraction of metabolic carbon into their shells and this fraction generally increases with age, slowed growth/calcification, or increased food supply [Lorrain et al., 2004; McConnaughey and Gillikin, 2008; Lartaud et al., 2010]. Such ontogenic changes have not been previously documented in developing bivalve larvae, but it should be noted that our observational study is based on one cohort (from multiple parents). Ongoing studies of other larval cohorts show only minor variations from the δ13C patterns reported here and these appear diagnostic of larval fitness, confirming our findings. These results will be reported in a future manuscript.
 During the formation of the initial D-hinge shell, larval oysters must precipitate calcium carbonate at a rate greater than abiotic precipitation [Zuddas and Mucci, 1998] and an order of magnitude greater than later larval stages (Figure 2a). If we assume first order (n = 1) precipitation kinetics (e.g., r = k[Ώ − 1] as above), at the mass-normalized mineral precipitation rates (r = 0.45 and 0.04 d−1) for initial and later shell stages, respectively, an exponential increase in the difference between k to maintain each precipitation rate is found as aragonite saturation state (Ω) approaches 1 (saturation). It is important to note that in this formulation, precipitation rates would be better described as per surface area (rather than mass) [Morse et al., 2007], however, surface area estimates of calcification surfaces are lacking. Mollusks employ two strategies to overcome the kinetic constraint: (1) use of organic molecules that increase rates of mineral formation [Elhadj et al., 2006] and (2) increase of proton pumping [Ries2011] to elevate saturation state at the mineral formation sites. Both of these strategies require energy, and it is unclear to what extent each is employed during and following D-hinge shell formation. The initial shell appears to be formed with less ability to isolate the site of calcification (highlighted by the shift in δ13Cshell). Following D-hinge formation, this kinetic constraint is greatly reduced as the rate of shell formation decreases, while feeding and shell-forming organs develop providing means to reduce acidification stress [Waldbusser et al., 2010].
 The formation of calcium carbonate removes carbonate ions from seawater and results in the formation of protons within the calcifying fluids. The number of protons generated by calcium carbonate formation depends on the carbonate system speciation. With bicarbonate generally constituting 85-90% of seawater DIC, we can assume that the protons pumped to mineral precipitation ratio will be ~1 [Cohen and Holcomb, 2009] with an increasing ratio as metabolic CO2 incorporation into the shell increases. At 2 days post-fertilization, larval shell mass was estimated at 162 ng (Figure 2). Assuming a calcium-ATP linked proton pump [Cohen and Holcomb, 2009] (2 mol H+ = 1 mol ATP), 1 mol CaCO3 = 1 mol H+, and using an estimated available energy from converting ATP to ADP of 30 KJ mol−1, the estimated cost of precipitating ~160 ng of CaCO3 is ~24 μJ. If the calcifying fluid during D-hinge shell formation is in greater contact with seawater (as evidenced by the shell isotopes), then proton pumping may be a less important strategy to elevate calcification rate and k must be increased further (presumably by the use of macromolecules). The greater energetic cost is in fact protein synthesis for the shell organic matrix [Palmer, 1992]; based on an estimate of 1.9% organic material in the two day old larval shell, and a total shell mass of 162 ng larva−1, we estimate this cost to be 70 μJ larva−1(following Palmer ). The individual energetic cost associated with building the 2 day old shell in our measured cohort can be estimated to be almost 100 μJ larva−1.
 Our δ13Ctissue isotopic data showing sole utilization of maternally derived carbon during the formation of the initial D-hinge shell (Figure 1b) confirms previous findings on the role of endogenous energy sources for early larval bivalves [Gallager and Mann, 1986; Sanchez-Lazo and Martinez-Pita, 2012]. The depletion of larval energy reserves (primarily lipids) following D-hinge shell formation (Figure 2b and 2c), and the δ13Ctissue trend (Figure 1b) highlights that the energy available for embryogenesis and D-hinge shell growth is limited by the available maternally-derived energy. Previous measurements of total lipid content of C. gigas and other Crassostrea spp. eggs found a maximum lipid content of 10 ng egg−1 [Gallager and Mann, 1986; His and Maurer, 1988]; with 1 day old C. gigas non-structural lipids (triacylglycerides) comprising greater than half (3.8 ng larva−1) of the total lipid pool (5.9 ng larva−1) [Moran and Manahan, 2004]. Assuming a minimum structural (phospholipid) lipid concentration of ~2 ng larva−1 at 1–2 days old, the remaining 8 ng of lipids would therefore provide a total of ~316 μJ larva−1 (39.5 μJ per ng lipid). Shell formation of two-day old C. gigas therefore would require a minimum of 30% of the total available energy in egg reserves (~100 μJ larva from above), assuming lipid rich eggs. Crassostrea virginica larval survival and growth drop quickly when egg lipids falls below 6 ng egg−1 [Gallager and Mann, 1986] (or 158 μJ larva−1 in lipid energy).
 Environmental stressors, such as acidification, create additional energetic costs and lower the scope for growth, particularly if food is limited [Melzner et al., 2011]. For bivalve larvae, there are two critical stages when larvae must rely on endogenous energy: during embryogenesis (and prodissoconch I shell formation) and during metamorphosis from larvae to juvenile stages. Chronic acidification stress lowers lipid levels in bivalve larvae as they approach metamorphosis and results in greater failure of settling larvae [Talmage and Gobler, 2010, 2011], while calcification in newly settled juvenile hard clams is more sensitive to acidification than later stages [Waldbusser et al., 2010]. Success of bivalves in future higher CO2 waters will ultimately depend on the timing of these critical life-history stages relative to the frequency and duration of unfavorable carbonate chemistry conditions; chemical conditions that are variable [Hofmann et al., 2011], predicted to become less favorable rapidly in some systems [Gruber et al., 2012], and result from multiple drivers in coastal zones [Kelly et al., 2011]. The energetics and kinetics of biocalcification presented here also explains the seemingly unusual sensitivity of early larval calcification to ambient-water chemistry, that is thermodynamically favorable for precipitation (Ω > 1) [Barton et al., 2012]. We suggest that the predictions of winners and losers in a high CO2 world may be better informed by calcium carbonate kinetics, bioenergetics, ontogeny, and life-history characteristics than by shell mineralogy alone.
 The authors would like to thank S. Cudd, M. Wiegardt, and A. Brown of the Whiskey Creek Shellfish Hatchery for oyster larvae, time, and a platform for this work. The authors would also like to thank M. Sparrow, J. McKay, and A. Ross for their analytical expertise. A previous version of this manuscript was greatly improved by thoughtful and critical feedback by C. Gobler and one anonymous reviewer. This work was supported by the National Science Foundation OCE CRI-OA 1041267 to GGW, BH, CJL, and BAH.
 The Editor thanks Christopher Gobler and an anonymous reviewer for their assistance in evaluating this paper.