Experimental setup: Two complementary experiments were conducted in late 2018 (E1) and 2020 (E2), each rearing newly fertilized sand lance embryos to hatch over the course of 32-65 days. Founder adults were sampled at Stellwagen Bank National Marine Sanctuary (SBNMS) (42° 9' 58.26" N, 70° 18' 44.19" W) at the peak of their narrow, local spawning window on November 15th (E1) or 27th (E2), using a 1.3 × 0.7 m beam trawl (6 mm mesh) towed over ground at 3 knots for 15 min. On deck, all flowing-ripe males and females were strip-spawned together (at 10°C, E1: Nmale/female = 29/13; E2: Nmale/female = 50/46) and their progeny transported to the University of Connecticut's Rankin Seawater Lab. There, exposure experiments commenced within 8 hours post fertilization by placing a volumetrically measured random sample of 600 (E1) or 1,200 embryos (E2) into each replicate rearing container.

Experimental seawater was drawn from subsurface eastern Long Island Sound (~ 31 psu), filtered to 1 µm, and UV-sterilized before use. Oxygen levels were maintained at ~100% saturation, while the photoperiod was 11L:13D.

Seawater chemistry: Realized pCO₂ conditions and other seawater chemistry parameters (Table 1 of Baumann et al., MEPS (in review)) were estimated in CO2SYS (V2.1, Pierrot et al. 2006) based on samples taken every 10 days and measured for temperature, pHNIST, salinity (refractometer) and total alkalinity (AT, μmol kg-1). Seawater samples were filtered to 10 µm, stored in 300 ml borosilicate bottles at 3°C, and within days measured for AT using endpoint titration (Mettler Toledo® G20 Potentiometric Titrator) with an accuracy of ±1% (Murray et al. 2019; verified and calibrated using Dr. Andrew Dickson’s certified reference material for AT in seawater; Scripps Institution of Oceanography, Batch Nr. 162 & 164).

Experimental designs: During E1, we tested factorial combinations of two static temperatures and three target pCO₂ levels, thereby encompassing contemporary thermal conditions on Stellwagen Bank between late fall (10°C) and early winter (6°C), as well as current ambient (400 µatm, pH~8.12), predicted end-of-century (1,000 µatm, pH~7.76), and maximum open ocean pCO₂ benchmarks (2,000 µatm, pH~7.48; Caldeira & Wickett 2003, Salisbury & Jönsson 2018). At 10°C, three additional, pCO₂ levels below 1,000 µatm (570, 690, 890 µatm, Table 1) were included to better describe near future CO₂ sensitivities of sand lance embryos. The replication level for each of the 9 treatments was N = 5. Another 50 embryos per replicate were subsampled 90-190 ddpf and preserved in buffered (sodium tetraborate) 5% formaldehyde-in-freshwater solution. Those embryos sampled just before hatching began (170 ddpf, one random replicate per treatment) were later submitted for sectioning and staining (H&E stain; Horus Scientific, Worcester, MA), and later imaged for analyses of chorionic thickness (Nikon SMZ-1000 with Luminera® Infinity2-2 camera and ImagePro Premier v9.0, Media Cybernetics®).

During E2, we again tested target pCO₂ levels of 400, 1,000, and 2,000 µatm, first at an intermediate static temperature of 7°C and second at a dynamic temperature of 10°C decreasing to 5°C at a rate of 0.2°C d⁻¹ (105°C). The latter was chosen to approximate the seasonal decline in bottom temperatures experienced by sand lance embryos on Stellwagen Bank. The two treatments reached thermal equivalence at 32 dpf (224 ddfp) – just after hatching had started. To better describe sand lance upper CO₂-sensitivities (1,000 - 2,000 µatm), we added two intermediate pCO₂ levels at 7°C (~1,300; ~1,700 µatm) and one at 105°C (~1,300 µatm). The initial replication level for each of the 9 treatments was N = 6. However, to disentangle potential pCO₂ effects on embryonic development vs. effects on hatching itself, we switched three random replicates from each extreme pCO₂ treatment per temperature with the opposite pCO₂ treatment (i.e., 3× ~400 --> ~2,000 µatm and 3× ~2,000 --> ~400 µatm). The switch happened at 175 ddpf (25 dpf at 7°C; 22dpf at 105°C) just before hatching started.

Response traits: From 90 ddpf onwards, rearing containers were monitored daily until hatching commenced; then, the number of hatchlings per replicate was recorded daily until hatching ceased. All hatchlings were immediately preserved in buffered 5% formaldehyde/freshwater solution for later morphological measurements. At the conclusion of E1, unhatched remains were imaged at 4× magnification, allowing the later distinction between (a) early arrested embryos (no or only amorphous cell masses visible), (b) partially developed embryos (unpigmented eyes visible, body not fully wrapped around the egg), and (c) fully developed embryos (pigmented eyes, body clearly visible and more than 1× wrapped around; Fig.S2). In E2, we continued daily monitoring for 7 more days after hatching had ceased; then examined the remains microscopically for embryos still alive (i.e., with beating hearts). Absolute hatching numbers were transformed to daily relative frequencies via dividing by the initial number of embryos that was adjusted for subsampling (E1, N = 500 per replicate) or reduced fertilization success (E2, N = 873 per replicate, based on examining independent post-fertilization subsamples). Relative frequencies were then summed to yield cumulative hatching success (HS, %) for each replicate. For E1, we additionally calculated the proportions of (a) fully developed but unhatched embryos and (b) all other arrested embryos combined. The latter also included decayed stages that were no longer detectable at the conclusion of E1.

To characterize hatching phenology, we recorded the day of first hatch (dpf), day of peak hatch (= dpf with the highest relative hatch frequency), and the total hatching period (d) for each replicate. Following Murray et al. (2019), a large number of hatchlings were imaged at 4× magnification (E1: Ntotal = 3,923; E2: Ntotal = 2,659) and then individually measured (ImagePro) for three morphological traits, i.e., standard length (SL, nearest 0.01 mm), yolk sac area (nearest 0.001 mm²), and the size of the remaining oil globule inside the yolk sac (nearest 0.001 mm²). The latter two traits are proxies for endogenous energy reserves after hatching, but they were strongly correlated (N = 5,552; R = 0.62, p < 0.001). Hence, we used PCA to extract the first principal component (explaining 73% (E1), 81% (E2) of variability) and then used the PC1 scores as the new variable, hereafter referred to as ‘Endogenous Energy Reserves’ (EER). Histological sections of fully developed, pre-hatch embryos from E1 were imaged at 20× magnification to measure the thickness of the chorion (ImagePro). Chorion thickness was measured at 10 randomly selected locations around the circumference of the sectioned embryo, with measurements averaged subsequently for each embryo. Unfortunately, fewer than expected embryos were sectioned well enough for quality measurements, ranging from 2-7 per treatment.