Neurometabolic rewiring in squid (Sepioteuthis lessoniana) optic lobes drives behavioral plasticity and visual integration under environmental acidification

Abstract

Ocean acidification’s impacts on marine animal behavior have substantial implications for ecosystem stability. Understanding how key predators respond to acidification is crucial for predicting future ocean food web dynamics, yet the underlying neural mechanisms remain poorly understood. Here, we show that prolonged exposure to projected year 2100 acidification conditions substantially impairs predatory behavior in bigfin reef squid (Sepioteuthis lessoniana), a key invertebrate predator. Chronic acidification exposure reduces expression of acetylcholine receptors in optic lobes and alters systemic HCO₃⁻ levels and metabolic rates. Using custom electroretinogram recordings, we find that while basic visual processing remains intact, behavioral impairments likely stem from changes in downstream neural integration pathways. Transcriptomic expression analysis reveals broad reductions in energy metabolism and synaptic signaling under acute exposure, while chronic exposure induces compensatory upregulation of cellular maintenance pathways. Our findings demonstrate that while squids maintain visual capabilities through adaptive mechanisms, the energy-intensive processes of neural integration and behavioral execution are compromised. These results highlight the complex physiological trade-offs marine predators face under ocean acidification, with implications for understanding future shifts in marine ecosystem structure and function.

Ocean acidification disrupts squid hunting behavior by inducing neurometabolic rewiring in the optic lobes. Changes in cholinergic transmission and metabolism make predators less successful, showing climate change impacts on marine predator neurobiology.

Introduction

Approximately 30% of atmospheric carbon dioxide (CO₂) dissolves into oceans, forming carbonic acid, which dissociates into protons (H⁺) and bicarbonate ions (HCO₃⁻)¹. Rising anthropogenic CO₂ emissions are rapidly increasing ocean surface acidity, a process known as ocean acidification (OA)². Climate models predict that the ocean’s pH will drastically increase in acidity compared to pre-industrial values by 2100 and 2300, to 7.7 and 7.1, respectively³. Hypercapnic acidic environments induce respiratory acidosis in water-breathing animals, typically counteracted by accumulating extracellular bicarbonate (HCO₃⁻) and eliminating protons (H⁺)^4,5. While these compensatory changes in ion transport processes allow some organisms to maintain acid–base homeostasis⁶, the necessary energy reallocation may limit traits including growth, reproduction, calcification, and activity^7–10.

Ocean acidification has also been suggested to affect the behavior and decision-making abilities of water-breathing species, inducing what is often described as anxiety-like symptoms^11,12. While behaviors may be altered through several mechanisms, the potential for acid-base compensatory changes in ionic gradients to influence ligand-gated ion channel function has received the most attention. This focus particularly concerns gamma-aminobutyric acid (GABA) type A receptors (see review by Thomas et al. ¹³, see also related controversies^14–16). When GABA binds at the α/β subunit interface, GABA_A receptors undergo conformational changes that open a central chloride-permeable channel, allowing influx of Cl⁻ ions and, to a lesser extent, efflux of HCO₃⁻ ions (relative permeability ratio ~0.2–0.4)^17–19. Under physiological conditions, this results in membrane hyperpolarization and reduced neuronal excitability. The accumulation of HCO₃⁻ under hypercapnic conditions results in an equimolar reduction of extracellular Cl⁻ to maintain electroneutrality^20–22. Based on these ionic changes, Nilsson et al. proposed the “GABA switch hypothesis”²³, suggesting that altered electrochemical gradients favor outward Cl⁻ flow through GABA_A receptors, changing their function from inhibitory to excitatory. This hypothesis has received empirical support from studies showing CO₂-induced anxiety in rockfish linked to altered GABA_A receptor activity¹¹ and behavioral changes in cephalopods associated with ligand-gated chloride channel dysfunction under elevated CO₂²⁴. Similar behavioral disruptions have been observed across various marine species, including tropical damselfish^25,26, temperate sticklebacks²⁷, and marine mollusks such as cone snails²⁸ and sea hares²⁹. Pharmacological experiments support the mechanisms behind the GABA switch hypothesis. Treatment with GABA_A receptor antagonists, especially gabazine, reverses OA-induced behavioral abnormalities in fish^23,27,30, while exposure to the GABA_A agonist muscimol replicates these impairments in control animals¹¹. However, incomplete behavioral rescue by gabazine and the persistence of gabazine-insensitive but picrotoxin-sensitive responses indicate involvement of non-GABA_A chloride channels²⁴. Since picrotoxin blocks both GABA_A and GABA_C receptors through action at the chloride channel pore, the gabazine-insensitive but picrotoxin-sensitive responses likely reflect GABA_C receptor involvement. Additional gabazine- and picrotoxin-insensitive responses suggest further complexity, potentially involving other ligand-gated chloride channels such as glycine receptors or invertebrate-specific channels, though their specific identities and pharmacological profiles remain to be determined. Moreover, transcriptomic studies reveal that hypercapnia affects gene expression across multiple neural pathways. Differential expression has been documented for genes involved in neurotransmission and synaptic function^31,32, as well as olfactory signaling pathways³³. In mollusks, pteropods exhibit altered expression of neural development and synaptic function genes under OA conditions^34,35, while physiological studies in cephalopods demonstrate disrupted ligand-gated chloride channel function²⁴. Collectively, these findings indicate that OA-induced behavioral alterations arise from system-wide neural disruption extending beyond isolated GABAergic dysfunction, highlighting the complexity of CO₂ effects on marine nervous systems.

Behavioral abnormalities have the strong potential to alter the stability and structure of an ecosystem and its food web by changing predator-prey dynamics and recruitment^{25,26,36–41}. Although there are fewer studies on invertebrates, they indicate that ocean acidification impairs foraging tactics⁴², locomotion⁴³, and decision-making^8,44,45. Coleoid cephalopods are particularly voracious predators with rising global populations^46,47. Highly active squid exert a particularly strong top-down influence on prey populations while also being important prey themselves for large teleost fishes and marine mammals^48,49. Aside from their athleticism and rapid growth rates, their success as predators is largely attributed to their advanced neural systems, especially their visual acuity^50–52. While these features make them successful predators, constraints on their physiology and energetic strategies mean that even slight environmental challenges can be fatal^52–55. Previous studies have suggested that the decision-making processes of squid species are sensitive to ocean acidification during predator-prey interactions; however, the persistence and cause of the symptoms are unclear^8,13,45. We hypothesized that since decision-making processes are largely controlled by the central nervous system, ocean acidification may negatively impact the central nervous system of cephalopods. Given their reliance on vision to secure prey, we first employed electroretinogram (ERG) measurements to evaluate basic visual function and temporal resolution. Following this initial assessment, we investigated potential changes in the optic lobes, which serve as crucial neural centers for visual information processing, visuomotor coordination, and as secondary retinas^51,56. To test this hypothesis, we used RNA-Seq and gene set enrichment analyses to examine transcriptomic changes in the optic lobes of lab-reared bigfin reef squid (Sepioteuthis lessoniana) under two distinct acidification scenarios: those exposed to acute OA for 7 days after development under modern-day conditions (termed “acute-OA”), and those that developed under acidified conditions for 90 days since hatching (termed “chronic-OA”) (Fig. 1A). These results were assessed in conjunction with parallel behavioral observations, visual processing, and the systemic state of the whole animal’s metabolism and extracellular acid-base status to link behavioral, sensory, and physiological changes.

Fig. 1. Physiological responses of S. lessoniana to ocean acidification.

A Schematic illustration of the experimental design that employed lab-reared 90-day-old bigfin reef squid (S. lessoniana) in control seawater (pH 8.20) or year 2100-like levels (pH 7.80) of ocean acidification (chronic-OA). Water parameters were measured weekly and are reported in Supplementary Table 5. Acute exposures involved the transfer of control squid to acidification for 7 days prior to making measurements (acute-OA). Routine metabolic rates (B, n = 8–12), whole animal ammonia excretion rates (C, n = 5–9), and systemic acid-base parameters (n = 7–9), including the D pH, E P_CO2, and F HCO₃⁻ concentration, of squid reared under control seawater for 90 days or CO₂-acidified seawater for 7 or 90 days. A box-and-whisker plot is employed to represent data. The median is indicated by the horizontal line within the box, which spans the 25th to the 75th percentile of the data. The box represents the interquartile range (IQR). Whiskers extend to the minimum and maximum values that are within 1.5 times the IQR. The distribution of the data points for each treatment group is illustrated by individual colored dots, which represent the examined value of each squid. Groups sharing the same letter are not significantly different from each other, while groups with different letters differ significantly (p < 0.05). Significant differences, denoted by letters, were detected using one-way further probing using Tukey’s multiple comparisons tests (metabolic features; Supplementary Table 1) or Welch ANOVAs further probed using Dunnett’s T3 multiple comparisons tests (blood parameters; Supplementary Table 2). Complete statistical details, including exact p-values and test statistics, are provided in Supplementary Tables 1 and 2.

Methods

Animal handling and rearing protocols

Wildtype adult bigfin reef squid (S. lessoniana) obtained from local markets were housed in concrete 90,000-L breeding aquaria fed continuously with natural seawater and used exclusively as breeding stock. All experimental animals were their laboratory-reared offspring, with complete life-history control from hatching. Eggs were removed from the breeding tanks within 24 h of their appearance and transferred to nets floating within 1000-L aquaria fed continuously with natural seawater. Hatchlings were transferred to floating nets in similar aquaria fed with natural seawater or mildly hypercapnic seawater within 12 h of hatching. Hatchlings were reared under these conditions and progressively transferred out of the nets and eventually into 37,000-liter aquaria based on their size and density. Each treatment condition (control pH 8.2 and acidified pH 7.8) consisted of four independent replicate tanks (37,000 L each), totaling eight experimental systems. Each tank was equipped with its own CO₂ injection system governed by an independent electronic controller (MACRO, Guangdong, China), ensuring that tanks were true experimental units rather than pseudoreplicates. All aquaria were maintained at approximately 25 °C, continuously aerated, and kept on a 10/14-h light-dark cycle. Squids were fed a mixture of live locally sourced freshwater cherry shrimp (Neocaridina denticulate) and whiteleg shrimp (Litopenaus vannamei) in a size-dependent fashion at least three times per day. After 90 days of development, squid reached an average fresh mass of 87.5 ± 3.3 g and mantle lengths of 9.9 ± 0.2 cm.

The pH and temperature were measured using a temperature-sensitive WTW SenTix 81 electrode attached to a WTW 340i m (Xylem Analytics, Weilheim, Germany) calibrated to standardized NIST buffers of pH 4.01, 7.00, and 10.01 (Xylem Analytics). While the use of NIST/NBS standardized buffers is common, they can introduce an inherent ~0.05 pH unit uncertainty in high ionic strength solutions, which can be mitigated by using Tris synthetic seawater calibration buffers or an appropriate indicator dye. Please note that the electrodes and meters used to monitor water parameters are independent of those used by the electronic controllers regulating the acidity of the aquaria. The total inorganic carbon was measured using an automatic dissolved inorganic carbon analyzer (AS-C3, Apollo Sci Tech, USA). The total alkalinity of the seawater was measured spectrophotometrically using a plate assay, modified from ref. ⁵⁷. Salinity remained stable at 32–33 ppt throughout the experimental period and was measured weekly using a calibrated refractometer. The carbonate equilibrium of seawater was determined mathematically by entering these metrics into CO2SYS software, using the NBS pH scale and the constants of Mehrbach et al.⁵⁸, refit by Dickson and Millero⁵⁹. Ammonia, nitrate, and nitrite levels were not measured but assumed to be negligible, given the high exchange rate of natural seawater within the holding systems.

To minimize handling stress and exercise during capture, the depth of the holding aquaria was reduced to ~1 m. The squid were then corralled towards a corner of the arena by slowly advancing a large net spanning the width of the aquarium. Squid were then allowed to swim into additional handheld nets, raised to the water surface and transferred by the flow of water into an insulated cooler. The water within the cooler was then exchanged with aerated seawater taken from the aquaria spiked with 2% MgCl₂·6H₂O to lethally anaesthetize the animals^60,61 while maintaining relatively constant environmental conditions. After about 10 min the animals ceased all movement and failed to respond to physical touch, allowing their weight and mantle length to be recorded. To prevent the premature loss of blood, physical confirmation of termination was performed after blood sampling but before tissue collection by using scissors to sever the connections between the head and mantle.

Although Taiwan did not have specific legislation on cephalopod research at the time of publication, we implemented comprehensive welfare protocols based on international standards. All housing conditions and experimental procedures were reviewed voluntarily by an internationally represented panel of experts. Our protocols follow guidelines set by Fiorito et al.⁶¹ and Andrews et al.⁶², including using 2% MgCl₂·6H₂O for humane euthanasia, reducing handling stress with water depth capture methods, and maintaining species-appropriate housing conditions. This voluntary review aimed to establish robust welfare frameworks for cephalopod research at our institution and to ensure compliance with ARRIVE guidelines for reporting animal research.

Metabolic parameters

The routine metabolic rates of squid were measured using a closed-system respirometry system consisting of custom-made airtight 3000 ml glass chambers fitted with externally housed fiber-optic oxygen sensors (PreSens Sensor Spots type PSt3, PreSens, Regensburg, Germany) attached to an OXY-4 Mini multichannel fiber-optic oxygen transmitter (PreSens). This system allowed four chambers to be operated simultaneously, one of which always contained no animals to monitor microbial influence over the dissolved oxygen content during the experiments. The respirometry system was calibrated to 100% and 0% dissolved oxygen using air-saturated seawater and a 1% Na₂(SO₄)₃ seawater solution, respectively. The seawater used during the experiments and for calibration was UV-sterilized overnight to minimize microbial activity and subsequently pre-adjusted to 25 °C and the appropriate pH and CO₂ tension for the experiment.

Individual squids were carefully removed from their holding tanks using a rectangular net as a ‘false wall’ to slowly trap the animals and raise them to the surface without provoking alarm responses (e.g., inking, excessive jetting). Once at the surface, the animal was gently transferred into an adjacent aquarium containing the respirometry chamber and clean seawater that had been UV-sterilized, aerated overnight, and pre-adjusted to 25 °C and the appropriate pH and P_CO2 levels. The lids were sealed once the squid had been guided into the chamber, and the linear reduction in dissolved oxygen concentration was then monitored until reaching 80% of air-saturated levels. Squids were then removed from the chamber, their weight and mantle length recorded, and subsequently transferred to a non-experimental holding tank. The routine respiration rate of the squid was then calculated by dividing the rate at which the molar amount of dissolved oxygen changed during the experiment by the fresh mass of the squid and the duration of the experiment, after accounting for microbial influence, as described by Hu et al.⁶³. pH and total inorganic carbon were not measured at the end of respirometry trials; thus, CO₂ accumulation within chambers during measurements could not be quantified.

To measure whole-animal ammonium excretion rates, individual squid were transferred from their holding tank into 6000 ml aquaria filled with continuously aerated seawater that was UV-sterilized and pre-adjusted to 25 °C and the appropriate pH and CO₂ tension for the experiment. Note that these individuals are distinct from those used for measuring oxygen consumption rates. Water samples were collected 30 and 60 min after squid were transferred into the aquaria and subsequently frozen at –20 °C. Squid were then removed from the chamber, quickly dabbed free of excess water, and their fresh weight and mantle lengths were recorded before being returned to a non-experimental holding system. The ammonium concentration of the water samples was later determined fluorometrically using the amino acid insensitive method of Holmes⁶⁴ as modified by Hu et al.⁶³. In brief, 25 µl samples were mixed with 100 µl of a working reagent composed of orthophthaldialdehyde, sodium sulphite, and sodium borate in a black and opaque 96-well plate and incubated in the dark at room temperature for 2 h. The fluorescence of the samples was measured using a SpectraMax i3 multi-mode microplate reader (Molecular Devices, San Jose, California, USA) with excitation at 360 nm and emission at 420 nm. The molar difference of ammonium concentrations between 60 and 30-min samples was then calculated and divided by the fresh weight of the animal and the duration of the experiment to calculate the ammonia excretion rate of the squid.

Systemic acid–base parameters

Approximately 300–500 µl of blood was drawn from the central ventral vein and branchial arteries of lethally anaesthetized animals after bisecting the ventral side of their mantle. Blood pH was immediately recorded using an In-Lab Micro pH probe (Mettler-Toledo, Columbus, Ohio, USA) connected to a WTW 340i meter (Xylem Analytics) calibrated to standardized buffers of pH 4.01, 7.00, and 10.01. The total carbon within the blood sample was then measured by using a gas-tight syringe (Hamilton Company, Lincoln, Nebraska, USA) to inject a 10 µl aliquot into a spargine chamber filled with 0.01 N HCl that is continuously purged with N₂ gas (75 ml N₂ min^–1; GFC17, Aalborg, Orangeburg, New York, USA). The purged gas carried the total carbon within the sample as gaseous CO₂ through a Li-850 infrared CO₂/H₂O gas analyzer (LI-COR, Lincoln, Nebraska, USA), where LI-8 ×0 software (v1.0.4, LI-COR) recorded the data as a pulse of CO₂. Total CO₂ was calculated by integrating the recorded CO₂ pulse and comparing it to those of a series of known NaHCO₃ standards. Due to the proteinaceous nature of cephalopod blood, the HCl within the chamber was flushed regularly, although CO₂ pulses remained stable so long as the HCl: injected volume ratio remained constant.

Systemic P_CO2 and [HCO₃^-] were mathematically derived by inputting the measured pH and total carbon into rearrangements of the Henderson-Hasselbach equation (Eqs. 1 and 2, respectively). The solubility of CO₂ (α) and the first dissociation constant of carbonic acid (pka1) used for calculations were temperature and salinity-adjusted constants of green shore crab (Carcinus maenas) hemolymph⁶⁵.

$$P_{{CO}_2}=TC{O}_2/\left[\left(1+antilog\left(pH-pk{a}_1\right)\right)*\alpha C{O}_2\right]$$ 1

$$\left[HC{O}_3^{-}\right]=TC{O}_2-\left(P_{{CO}_2}*\alpha C{O}_2\right)$$ 2

Behavioral assays

To reduce transportation stress, squid were moved to a 20-L aquarium containing free-flowing control seawater or pre-adjusted hypercapnic water, depending on the state of their holding tank. Squids were given 7 days to acclimate to the aquarium and could be quickly and gently transferred to the behavioral arena. The behavioral arena was divided into three zones: a “main” zone measuring 60 cm length × 20 cm width × 22 cm water depth, a “shrimp” zone measuring 20 cm length × 20 cm width × 22 cm water depth, and a “buffer” zone of identical dimensions to the “shrimp” zone. These zones were separated by perforated partitions. The partition separating the “main” zone from the “shrimp” zone was attached to an air-suspended pulley system, allowing the partition to be removed without the animal seeing an operator. The buffer zone was inaccessible to the animals and was treated as a reservoir to prevent changes in water quality during the 30-min experiment. The experimental arena was drained and scrubbed after each experiment to eliminate hormones from the prey or predator that could influence subsequent assays.

Two cameras were used during the behavioral recordings, with one being positioned at 90° angle directly above the arena and the other being positioned directly in front of the arena to record the full 80 cm length of the “main” and “shrimp” zones simultaneously, permitting paired 2-dimensional analyses of the squid’s locomotion and positioning within the arena. The footage was recorded at five frames per second. EthoVision XT 11.5⁶⁶ was used to determine the X and Y coordinates of the squid, specifically where the head meets the mantle, within the arena, using the media-recorded front-facing recording. The media recorded by the camera held above the arena was similarly used to determine the Z coordinate of the squid within the arena, based on matching X coordinates from both recordings. Coordinate transformation: Raw tracking data from EthoVision XT 11.5 uses image coordinate conventions with the origin (0,0) at the upper-left corner and Y values increasing downward. To improve biological interpretability and match standard spatial analysis, all positional data were transformed to Cartesian coordinates: X coordinates remained the same, Y coordinates were inverted (Y_cartesian = 22 cm–Y_image), and Z coordinates were measured upward from the arena bottom. This transformation ensures that “upper zones” (zones A&B) correspond to higher Y values and positions near the water surface. All subsequent statistical analyses and heat map visualizations were performed using these transformed coordinates. The resulting X, Y, and Z coordinates were recorded for every 0.1 s of the recordings and then compiled using OriginPro (OriginLab, Northampton, Massachusetts, USA) to visualize the positional data of the squid collectively in multiple dimensions.

Animals were gently transferred from the nearby 20-liter aquarium into the “main” zone of the arena and given 10 min to acclimate to the surroundings. Ejection of ink during the 10-min acclimation period was considered a sign of excessive transfer stress. If ink was ejected, squids were returned to the nearby aquaria, the arena cleaned, and an attempt was made with a different individual. After acclimating to their surroundings, the footage recording the locomotory and positional behavior of the squid was collected. Eq. 3, a 3-dimensional equivalent of Pythagoras’ theorem, was used to determine the total distance traveled by individual squid throughout their trial by summing the sequential change in x, y, and z coordinates of two points separated by 0.1 s throughout the duration of the trial. The total distance was then divided by the duration of the experiment to determine the mean velocity of the squid during the experiment—both of which were normalized to the mantle length of the individual which was recorded after the trial concluded.

$$TotalDistance=\Sigma \surd \left[{\left(x_2-x_1\right)}^2+{\left(y_2-y_1\right)}^2+{\left(z_2-z_1\right)}^2\right]$$ 3

The “main” zone was divided into upper/lower and left/right regions to objectively define the positional behaviors of the squid. The upper/lower divide of the front-facing recording (regions AB and CD, respectively) was separated by the halfway point of the arena’s water depth. The left/right divide (regions AC and BD, respectively) defined the halfway point of the “main” zone’s length. In the recording made from above the arena, zones EF and GH defined the “front zone” closer to the front-facing camera and the “back zone” furthest from the camera, respectively.

After 10 min of recording, an out-of-view operator used the pulley system to gently remove the partition separating the “main” and “shrimp” zones to reveal an appropriately sized live whiteleg shrimp (ca. 3 cm) that had been previously placed within the arena. Squid were given a maximum of 10 min to secure their prey. Successful captures were defined as squids that were able to immobilize the shrimp, strike attempts were considered as moments when the squid extended their arms towards the prey and/or rapidly accelerated toward the prey, and the hunt duration was considered the time elapsed until a successful capture was made since the partition was removed. The velocity and positional behavior of the squids were not analyzed during this time, allowing the analysis to be made using 2-dimensional recordings from the front-facing recording. Given that predator-prey interactions were not dependent on the dimensionality of the recording, the predator-prey interaction data includes recordings made using the 3-dimensional tracking system as well as prior measurements made solely using a front-facing camera prior to our realization that locomotory and positional data of cephalopods would not be comprehensively investigated by such a system. The velocity of the prey used in the experiments was measured using the 2-dimensional system and EthoVision XT 11.5 to determine if acute exposure to the hypercapnic seawater induced any significant effects on their locomotion.

Electroretinogram (ERG) recording and analysis

To assess the visual function of S. lessoniana under different environmental conditions, ERG recordings were conducted on squids exposed to control (pH 8.2) and ocean acidification conditions (pH 7.8, either acutely for 7 days or chronically for 90 days). The squid was immobilized in a custom-built ERG chamber, which included continuous perfusion of aerated seawater over the gills to maintain normal physiological conditions.

A white LED light source (300–700 nm) was used to deliver flicker stimuli at varying frequencies, while the light intensity was controlled using neutral density filters. The light stimuli were presented in a dark, scotopic environment to minimize interference from ambient light. The stimuli duration lasted for 200 ms, followed by a 200 ms rest period, for a total of 5 s per frequency. Recording tungsten electrodes (573510, A-M SYSTEMS Inc, WA, USA) were placed on the surface of the squid’s eye, while reference electrodes were positioned on nearby tissue to ensure differential recording. Vaseline was applied around the eye to prevent water-electrode contact, maintaining electrical isolation. Electrical signals were amplified (×10,000) using an amplifier (DP-301, Warner Instruments, Holliston, MA, USA) and digitized using an analog-to-digital (AD) converter (PowerLab 4/20, ADInstruments Pty. Ltd., Dunedin, New Zealand) (sampling rate: 4 kHz), and data were continuously recorded on a computer for further analysis. The ERG traces for different stimulus frequencies (20 Hz, 30 Hz, 40 Hz, and 50 Hz) revealed a clear attenuation in response amplitude with increasing frequency.

The critical flicker fusion frequency (CFF) was determined by analyzing the maximum response amplitude across different stimulus frequencies. Response amplitudes were averaged for each frequency. The CFF threshold, defined as the maximum response frequency divided by the stimulus frequency where the response value was greater than or equal to 0.9, was used to assess temporal resolution. The recorded signals were processed using a custom Microsoft Excel script to subtract baseline noise and improve the signal-to-noise ratio (S/N). The analysis also included subtracting the no-stimulus period spectra from the stimulus period spectra to isolate the visual response (Fig. S4B). Statistical comparisons between control, acute-OA, and chronic-OA groups were made using one-way ANOVA followed by Tukey’s multiple comparisons test, with significance set at p < 0.05. All experiments were repeated on 3–6 individuals per group.

RNA extraction

After developing for 90 days under their acclimatory conditions, squids were lethally anaesthetized in aerated seawater containing 2% MgCl₂. Once all movement ceased and the squid failed to respond to physical touch, their optical lobes were dissected, frozen in liquid nitrogen, and stored at –80 °C until total RNA could be extracted using the RNeasy Plus Mini Kit (Qiagen, Hilden, Germany). In brief, 50 mg of frozen tissue was submerged in QIAzol lysis reagent and homogenized with 2 mm steel beads using a TissueLyzer II (30 Hz, 2 min; Qiagen). The homogenates were then treated with gDNA Eliminator (Qiagen) before using chloroform and centrifugation (12,000 × g, 15 min, 4 °C) to achieve phase separation. The RNA-rich top layer of the supernatant was collected and mixed with 70% ethanol before being passed through the RNeasy spin column, where the samples were washed before being eluted with RNase-free water. The quantity and quality of the extracted RNA were monitored using a NanoDrop spectrophotometer (ThermoFisher, Waltham, Massachusetts, USA) before being sent to third-party commercial entities (NGS High Throughput Genomics Core, Academia Sinica, Taipei, Taiwan, Genomics Co. COMPANY INFO) for library preparation and sequencing.

This protocol was also followed to extract total RNA from samples of control and acidification exposed adult squid brain (B), esophagus (ESO), intestine (Int), muscle (Mus), posterior salivary gland (PSG), retina (Rn), the nerve plexus between the retina and optic lobes (RnS), skin (Sk), caecum (SV) and stomach (St) as well as 1-day and 1-month post-hatch whole squid. These samples were used to develop the de novo assembly that was used to reconstruct the expression of transcripts collected from the optical lobes.

Transcriptome sequencing

Poly-A RNA-Seq libraries were constructed using the TruSeq Stranded RNA Sample Prep Kit (Illumina, San Diego, California, USA). Initially, 2 µg of total RNA was utilized per sample for oligo-dT bead enrichment. The RNA bound to the beads underwent heat fragmentation at 94 °C for 7.5 min, followed by cDNA synthesis using random priming. The resulting double-stranded cDNA was purified using AmPure beads (Beckman Coulter, Brea, California, USA) and subjected to end-repair, A-tailing, and barcoded adapter ligation. Ligation products were treated with USER enzyme (New England BioLabs, Ipswich, Massachusetts, USA) and amplified by PCR for 10 or 12 cycles, depending on the yield. Libraries were purified with AmPure beads, and their concentration was determined using Qubit (ThermoFisher) and their profile using a BioAnalyzer HS DNA Assay (Agilent, Santa Clara, California, USA). Effective molar concentrations were normalized using the KAPA Library Quantification Kit for Illumina® Platforms (Roche, Basel, Switzerland). Equal molar concentrations of libraries were pooled and subjected to deep sequencing on the HiSeq 2500 Sequencer (Illumina) in a paired-end 2 × 151 nt format. On average, 21–23 million fragments (42–45 million reads) were obtained per RNA-Seq sample.

Data preprocessing and bacterial sequence removal

The quality of all the obtained pair-end reads was estimated by fastQC (v0.12.1). Based on the fastQC reports, the adapter sequence and quality lower than Q30 reads, and the first 15 bases in the front/the last base in the tail were removed by fastp (v0.23.2)⁶⁷. The high-quality sequences were then aligned to the SILVA database (SSU and LSU Parc 138.1) by Bowtie 2 (v2.5.3) to identify the bacterial sequences with the option --very-sensitive-local parameter⁶⁸. The sequences that had no alignment to the SILVA database were retained and utilized as the input for further analyses.

RNA-seq de novo assembly and transcript quantification

The de novo assembly was performed by running Trinity (v2.15.1) in a container with the default settings aside from setting the minimum k-mer to 2 (--min_kmer_cov 2) and output 902,413 transcripts⁶⁹.

To assess the number of full-length transcripts in the assembly output, blastx (BLAST v2.14.0 + ) was used to search the SwissProt knowledge base protein database using a threshold of e-value = 1e-20⁷⁰. The search results indicated that there were 8238 proteins in the database whose alignment coverage was higher than 80% in comparison to our assembled transcripts. The assembly quality was assessed using expression-dependent N50 (ExN50), following the guidance provided in the Trinity wiki (https://github.com/trinityrnaseq/trinityrnaseq/wiki). BUSCO (v5.4.3) was used to determine if the assembled transcripts were likely to represent the transcriptome of bigfin reef squid. The BUSCO identified that the Mollusca were the shortest animal lineage from the assembled transcripts⁷¹. Within the Mollusca_db10 dataset, a total of 5295 genes were identified from the assembled transcripts, including 4,759 complete genes (1477 single and 3272 duplicated), 138 fragmented genes, and 407 missing genes.

The abundance of the transcripts from each optical lobe sample was then estimated using the align_and_estimate_abundance.pl script running with the -trinity_mode, -seqType fq, -est_methid RSEM and -aln_method bowties2 options within the trinity container.

Functional annotation of the assembled transcripts

The coding region of the assembled transcripts was identified by TransDecoder (v5.5.5) following the program’s website (https://github.com/TransDecoder/TransDecoder). The sequences translated by TransDecoder were further utilized for the hmmscan command within HMMER (v.3.3.2) with the default options to search the Pfam database to identify protein domains (http://hmmer.org/). Meanwhile, blastp and blastx were used, respectively, to determine if the assembled transcripts were homologous to the SwissProt knowledge base protein database. Mollusca protein sequences within the TrEMBL of UniProt’s database were also utilized to identify homologous sequences within the assembled transcripts using blastx. All BLAST searches were performed using a set threshold of e-value 1e-3 and the -max_target_seqs 1 option (https://www.uniprot.org/)⁷⁰. The gene ontology was predicted by EggNog-mapper (v2.1.11) with options–target_taxa bilateria, --sensmode ultra-sensitive, and –go_evidence all⁷². The output results of the above analyses were integrated, and the names and gene ontology terms were extracted using Trinotate (v4.0.1) with an e-value threshold of 1e-3 (https://github.com/Trinotate/Trinotate/wiki/Software-installation-and-data-required)⁷³.

For the annotation of the assembled transcripts, the gene ID, protein name, and gene ontology of SwissProt knowledge base proteins and Mollusca proteins in the TrEMBL of Uniprot’s database were collected. The Trinotate extracted and the information downloaded from UniProt were combined and merged using R (v4.3.2) in RStudio (v.2023.12.0 build 369). Finally, the data containing transcript IDs, protein names, and their gene ontology terms of the assembled transcripts were utilized to establish the database using AnnotationForge (v1.44.0; https://github.com/Bioconductor/AnnotationForge).

Differential expression and functional analysis

The differential expression of the assembled transcripts was performed by DESeq2 (v1.42.0)⁷⁴. Transcripts that had expression counts less than 10 and were expressed in less than 3 samples were excluded from the downstream analysis, leaving 69,368 transcripts after filtering. The batch effect between sample batches was estimated and corrected by sva (v3.50.0)⁷⁵. The data was then transformed by a variance stabilizing transformation, and the batch effect detected from sva was removed⁷⁶. The log fold change of differential expression transcript outputs was shrunk by a normal shrinkage estimator⁷⁴ (Supplementary data: DET Results). The shrunken log fold change of differential expression transcripts was then ranked and utilized as the input for GSEA analysis using the clusterProfiler package (v4.10.0) with the options -ont “ALL”, -minGSSSize 10, -maxGSSSize 500, -nPermSimple 1000, -pAdjustMethod “BH,” and -pcalueCutoff 0.05⁷⁷ (Supplementary data: GSEA results).

Transcriptomic data visualization

The Venn diagram was made using the ggVennDiagram package (v1.4.9)⁷⁸. The ridgeplot and treeplot were made using the clusterProfiler package (v4.10.0)⁷⁷. The heatmap and the PCA plot were made by their respective functions in the DESeq2 package (v.1.42.0)⁷⁴.

Statistics and reproducibility

In each study, biological replicates are measurements taken from distinct individual squids. Critical measurements, such as ERG recordings and metabolic rate determinations, were subjected to technical replicates, and the mean value was employed for statistical analyses. Previous investigations of similar physiological and behavioral parameters in cephalopods^8,63 determined sample sizes. To address potential pseudoreplication concerns, all biological replicates were randomly sampled across the four replicate tanks within each treatment. Sample sizes reported (n) represent individual squids from different tanks, with efforts made to balance sampling across tanks to capture tank-to-tank variation within treatments. The Shapiro-Wilk test was used to determine whether the data’s residuals represented a Gaussian distribution. The residuals of data to be inspected with a 1-way ANOVA were inspected for heteroscedasticity using the Browne-Forsyth test. Data that met parametric assumptions were screened for outliers using ROUT (Q = 1%); however, no significant outliers were detected. If residuals did not meet the assumptions for parametric analyses even after attempting transformations, non-parametric tests were used, and no outlier screening was attempted. Blood parameters were statistically assessed using Welch ANOVAs, whereas routine metabolic and ammonia excretion rates were assessed using ordinary one-way ANOVAs and Tukey post-hoc tests. Logistic regression analysis (GLM, binomial family, logit transformation) was used to model the likelihood that environmental pH and duration of exposure affected the proportion of squid that attacked prey during behavioral assays. The strength of the model was probed using a receiver operator characteristic curve, which had an underlying area of 0.6992. Other behavioral parameters were investigated using ordinary one-way ANOVAs or Kruskal-Wallis tests. The potential impact of treatments on the velocity of prey was investigated using a two-way ANOVA where the independent factors were environmental pH and exposure time. Logistic regression was performed using R, whereas other statistical tests and plotting were completed using GraphPad Prism (v10.2.3).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Results

Metabolic and homeostatic performances

Results presented here are based on individuals sampled across multiple replicate tanks per treatment (see “Methods” for tank replication details). Squid reared in natural control seawater and subsequently transferred to acidified conditions for 7 days (acute-OA) showed routine metabolic rates of 16.9 ± 4.1 and 18.9 ± 7.0 µmol O₂ g_FM^–1 h^–1, respectively (Fig. 1B). Compared to control squid, squid being chronically exposed to acidified conditions for 90 days (chronic-OA) showed a 41% increase in metabolic rate, reaching 23.9 ± 2.1 µmol O₂ g_FM^–1 h^–1 (Fig. 1B and Supplementary Table 1). Ammonia excretion rates were also similar between control (3.3 ± 0.7 µmol T_Amm g_FM⁻¹ h⁻¹) and acutely exposed squids (2.9 ± 1.5 µmol T_Amm g_FM⁻¹ h⁻¹) (Fig. 1C). However, chronically exposed squids had significantly lower ammonia excretion rates, at 1.6 ± 0.5 µmol T_Amm g_FM^–1 h^–1, about 52% and 45% reduction compared to control and acutely exposed squids, respectively (Fig. 1C and Supplementary Table 1).

Blood pH in squids that developed under chronic acidification was significantly more alkaline (7.17 ± 0.07) than in control squids (7.09 ± 0.02) (Fig. 1D; all calculations independently verified, see “Methods”). This alkalinization was accompanied by a significant 2.6-fold increase in systemic P_CO₂ from 1057.4 ± 229.5 to 2778.7 ± 637.2 µatm (Fig. 1E) and a 3.1-fold increase in bicarbonate concentration from 3.56 ± 0.87 to 11.08 ± 2.20 mmol l^–1 HCO₃⁻ (Fig. 1F) when reared under control and chronically acidified conditions, respectively. When squids were reared in control conditions and then transferred to acidified conditions for 7 days, they showed an intermediate blood pH of 7.15 ± 0.09, which did not differ significantly from control (p = 0.391) or chronic-OA squids (p = 0.754) (Fig. 1D and Supplementary Table 2). During this intermediate state, systemic P_CO₂ increased to 1583.8 ± 283.4 µatm, representing a 1.5-fold increase compared to control squids (Fig. 1E and Supplementary Table 2). Bicarbonate levels in acutely exposed squid increased to 5.90 ± 0.64 mmol l^–1 HCO₃⁻, representing a 1.7-fold increase over controls, indicating partial but incomplete compensation (Fig. 1F and Supplementary Table 2). All pairwise comparisons for P_CO₂ and [HCO₃⁻] showed statistically significant differences (P_CO: p ≤ 0.013; [HCO₃⁻]: all p = 0.001, Dunnett’s T3 test).

Squids behavioral paradigms

During 10-min exploratory experiments in the novel arena (Fig. 2A), the positional frequencies of squid were visualized using 3D-tracking-generated heatmaps (Fig. 2B) and subsequently analyzed across three spatial planes (XY, XZ, and ZY; Supplementary Fig. 1). Squids reared under control conditions for 90 days traveled a total of 132.9 ± 93.4 mantle lengths (Supplementary Fig. 2) with a mean velocity of 0.2 ± 0.16 mantle lengths s^–1 during the 10-min exploratory period within the behavioral arena. Total distance traveled was unaffected by acute (124.4 ± 56.7 mantle lengths) or chronic exposure (147.3 ± 97.3 mantle lengths) to acidified seawater (Supplementary Fig. 2 and Supplementary Table 3). Thereby, mean swimming velocity remained unchanged at 0.21 ± 0.1 and 0.25 ± 0.16 mantle lengths sec^-1 for acute and chronic exposure, respectively (Fig. 2C and Supplementary Table 3).

Fig. 2. Behavioral responses under acidification exposure.

A Schematic representation of the novel arena used for the 3-D exploratory and predator-prey interaction behavioral assays of squid reared in control seawater for 90 days (n = 15), exposed acutely to acidified conditions for 7 days (acute-OA, n = 8) or reared in acidified conditions for 90 days (chronic-OA, n = 11). A The coordinate system follows a Cartesian convention with the origin at the lower-left corner of the main arena, with X increasing to the right, Y increasing upward (toward the water surface), and Z increasing upward. Locomotory and positional behaviors were monitored in the “Main Arena”, which was extended to include the “Shrimp” zone during the predator-prey assay. The “Buffer Zone” was never accessible to animals, serving as a means to aerate the water and monitor its quality. For positional purposes, zones A and B represented the top-half of the main arena, whereas zones A, C, G, E represented the left side of the arena with respect to the front-camera. Zones E and F were recorded from an above-camera and respectively represented the regions closest to the “back” and “front” of the arena walls with respect to the front-facing camera. The experimenter was not visible during the experiment, but was able to operate the pulley system to lift a perforated door to reveal the shrimp from a distance. B Spatial occupancy heatmaps showing collective density patterns during 10-min exploration trials. Squid positions (X, Y, Z coordinates) were recorded at 0.1-s intervals, and all positional observations from individuals within each treatment group were pooled and spatially binned (Control, n = 15 individuals, 90,000 total observations; Acute-OA, n = 954,000 observations; Chronic-OA, n = 11, 66,000 observations). Tracking data were transformed from image coordinates to Cartesian coordinates matching (A). Color intensity represents collective density (cumulative positional observations at each location), ranging from 0 (blue, never occupied) to maximum occupancy (red). Higher density values indicate locations where squid spent more time during exploration. Scale bar: 20 cm. Complete positional data of squid has been separated into three separate planes (XY, ZX, and ZY; Supplementary Fig. 1), which are then merged to generate the relative differences between the experimental groups on normalized velocity (C), duration in upper zone (D), duration in left zone (E), and duration in front zone (F) as they explored a novel arena for 10 min. The willingness of squid to hunt prey (G) was measured using 46 control, 36 acute, and 22 chronically treated squids. Squids that attempted to capture prey at least one time (control, n = 29; acute, n = 8; chronic, n = 8; Data are presented as mean ± SEM) were used to determine the hunt duration (H) and strike attempts (I). All the squid that were willing to hunt succeeded in capturing their prey. These observations were collected using squid that partook in the 2-D and 3-D tracking system experiments, as the analysis was manually performed and independent of dimensionality. A box-and-whisker plot is employed to represent data, which is then combined with a swarm plot for individual data points. The median is indicated by the horizontal line within the box, which spans the 25th to the 75th percentile of the data. The box represents the interquartile range (IQR). Whiskers extend to the minimum and maximum values that are within 1.5 times the IQR. Each squid’s normalized velocity is represented by individual black dots that surround the box plots. The dots are arranged to avoid overlap for better visualization, reflecting the density and distribution of data points within each experimental group. The violin plot overlay provides a visual representation of the probability density of the data within each experimental group. Significant differences across treatments are denoted by letters and were investigated using one-way ANOVA and Tukey post-hoc tests or Kruskal-Wallis and Dunn’s post-hoc tests (Complete statistical details, including exact p-values are provided in Supplementary Table 3). Groups sharing the same letter are not significantly different from each other, while groups with different letters differ significantly (p < 0.05).

Further separation of the data into three planes showed that squid reared in control and chronically acidified seawater spent 30.0 ± 39.0% and 14.7 ± 17.2% of the trial in the upper arena depths, respectively (Fig. 2D). In contrast, acute acidified conditions substantially influenced the behavior of squid, as they spent 86.0 ± 34.6% of the trial in the upper half of the arena at a significantly shallower depth (Fig. 2D). Acidification did not influence the preference for the left or right side of the arena. Control squids spent 55.8 ± 31.8% of the trial on the left side, while acute and chronic exposure squids spent 54.1 ± 27.9% and 57.1 ± 22.4%, respectively (Fig. 3E and Supplementary Table 3). The control squids spent 58.3 ± 35.6% of the trial near the front wall, compared to 80.0 ± 15.2% and 66.1 ± 30.1% for acute and chronic exposure, respectively (Fig. 3F and Supplementary Table 3). Similarly, no significant difference was observed in front/back preference. The 3D heatmaps and the separate plane views effectively illustrate that acute exposure has a more pronounced impact on vertical positioning within the arena than chronic exposure.

Fig. 3. Visual temporal resolution assessment via electroretinogram (ERG).

A Schematic representation of the ERG recording system used in this study. The system includes a water reservoir tank connected to an irrigation tube for gill perfusion, ensuring proper aeration during the experiment. The light stimulus is delivered via an LED light source, controlled through a computer system, with neutral density (ND) filters to adjust intensity. Recording and reference electrodes were positioned on the squid’s eye and nearby tissue, respectively, connected to an amplifier and an analog-to-digital (AD) converter for real-time data acquisition. B Photograph showing the ERG recording setup in an immobilized squid before scotopic (low-light) conditions. A small tube inserted into the mantle continuously irrigates the gills with aerated seawater, while Vaseline is applied around the eye to prevent contact between the electrode and the water. This arrangement ensures stable differential recording during visual stimulation. C Frequency response characteristics showing the relationship between stimulus frequency (Hz) and normalized maximum response frequency (maximum response frequency/stimulus frequency). Each symbol represents the response of one individual squid at the indicated stimulus frequency (control, n = 8; acute-OA, n = 8; chronic-OA, n = 9). Lines connect mean values across frequencies. Error bars represent standard deviation. Each animal was tested at multiple stimulus frequencies using a repeated measures design. Normalized response values decrease markedly at higher stimulus frequencies (>40 Hz). D A box-and-whisker plot is employed to represent data. The box represents the interquartile range (IQR), with the median indicated by the horizontal line within the box, which spans the 25th to the 75th percentile of the data. Whiskers extend to the minimum and maximum values that are within 1.5 times the IQR. Each dot represents an individual squid, and the horizontal red lines represent group means of critical flicker fusion (CFF) frequencies. Statistic analyses were performed by one-way further probing using Tukey’s multiple comparisons tests (n = 3–6 per group).

Logistic regression showed that acidified conditions significantly reduced the proportion of squid that attempted to attack prey (Fig. 2G, p = 0.0422), with no significant effect of exposure duration (p = 0.2461). After 7- and 90-days of acidification, squid were 64.7% and 42.3% less likely to attack prey compared to controls (Fig. 2G). Acute treatments significantly increased hunt duration, requiring 262.2 ± 193.0 s compared to 106.7 ± 125.2 s for controls (2.5-fold change, p = 0.03; Fig. 2H and Supplementary Table 4). Chronic exposure did not affect hunt duration. In addition, while monitoring the attempted attacks, the majority of examined squids were able to strike at prey successfully within 1–2 attempts (Fig. 2I). The mean velocity of prey was unaffected by acidification (Supplementary Fig. 3).

Temporal visual processing in squids

Given our observations of decreased predatory willingness and increased hunt duration in acidified conditions, we attempted to determine whether these prey-drive behavioral changes could be attributed to changes in visual processing. In order to evaluate the effects of acute and chronic acidification on the visual temporal resolution of squid, we created a custom ERG recording apparatus (Fig. 3A) that measured ERG responses to flicker light stimuli. Typical ERG waveforms showed distinct responses to various stimulus frequencies between 20 and 50 Hz (Supplementary Fig. 4A), and visual responses were distinguished from background noise using Fast Fourier Transform analysis (Supplementary Fig. 4B). This setup allowed for reliable light stimulation and ERG recordings while maintaining steady physiological conditions (Fig. 3B).

Under scotopic conditions, ERG recordings were conducted using tungsten electrodes positioned on the eye surface to access the CFF measured in control, acute- and chronic-acidified exposed squids to assess potential impacts on the temporal resolution of vision. The association between stimulus frequency and maximum response frequency was investigated. As the stimulus frequency increased, the ratio of maximal response frequency to stimulus frequency decreased in all groups, suggesting a diminished capacity to resolve rapidly flickering light at higher frequencies (Fig. 3C). Nevertheless, the CFF average values showed no significant difference between treatments (control: 50.8 ± 1.4 Hz; acute: 50.8 ± 6.3 Hz; chronic: 47.7 ± 5.5 Hz; Fig. 3D), though variance increased markedly under acidification, with 4- to 5-fold higher standard deviations in treated groups compared to controls. This heterogeneous response pattern, although not statistically significant, indicates variable individual sensitivity to acidification and warrants consideration when interpreting visual function outcomes.

RNA-seq profiling in the optic lobes

Given the critical role of optic lobes in visual processing and behavior, we conducted RNA-Seq analysis to explore transcriptional responses to acidification. Due to the absence of a reference genome, we performed de novo transcriptome assembly, achieving high-quality transcripts with mapping rates exceeding 97% (average 99.51%, Bowtie2 v2.5.3)⁶⁸. The output showed that the highest 78% of expressed transcripts had an N50 of 2038 bp, and that the in-total N50 of the assembled transcripts is 476 bp (Supplementary Fig. 5A). An initial analysis after batch-effect removal showed that acute-OA samples clustered closer to controls than chronic-OA samples (Supplementary Fig. 5B). This pattern was further supported by principal component analysis, where acute-OA samples remained near controls along PC1 (35% variance), while chronic-OA samples showed significant divergence along both PC1 and PC2 axes (15% variance), indicating more substantial transcriptional reprogramming under chronic exposure (Fig. 4A).

Fig. 4. Transcriptomic analysis of optic lobes reveals distinct responses to acute and chronic acidification.

A Principal Component Analysis illustrating transcriptomic variability among samples (Control, n = 4; Acute-OA, n = 5; Chronic-OA, n = 3), where samples cluster along PC1 (35% variance) and PC2 (15% variance). Acute-OA samples cluster closer to controls than chronic-OA samples. B Venn diagram quantifying shared and unique differentially expressed GO terms (adjusted p < 0.05) between treatments, showing distinct transcriptional responses. C Ridge plots comparing normalized enrichment scores (NES) of significantly altered GO terms between acute (left) and chronic (right) OA versus control. Color gradient indicates statistical significance (adjusted p-value). Hierarchical clustering of down-regulated gene sets after D 7-day acute and E 90-day chronic acidification versus control. Tree plots display functional gene clusters, with node size proportional to gene count and color intensity (red) reflecting statistical significance (adjusted p-value). Functional annotations show distinct pathway enrichment patterns between acute and chronic exposure. Heatmaps of upregulated GO terms under chronic acidification, displaying clustered and simplified terms for F cellular components (CC, 82 gene sets), G biological processes (BP, 308 gene sets), and H molecular functions (MF, 128 gene sets). Diagonal blocks (red) indicate high functional similarity between terms. The size of terms in legends is proportional to their enrichment within clusters, with larger text indicating greater functional significance in the acidification response. Color scales: D, E show adjusted p-values (0–0.8, optimized for significant results); F–H show GO term similarity scores (0–1, full mathematical range). Analysis performed using simplifyEnrichment v1.12.0.

Gene Set Enrichment Analysis (GSEA) was subsequently conducted to compare the effects of acute and chronic acidification on the optic lobes, with each group compared to controls. The Venn diagram (Fig. 4B) illustrates that only 2 out of 36 GO gene sets were down-regulated in both acute and chronic acidified conditions: organelle inner membrane and mitochondrial inner membrane, as indicated in the intersect GSEA ridgeplot (Fig. 4C). Acute acidification significantly suppressed multiple metabolic functions, including cellular lipid metabolism, carboxylic acid metabolism, and organic acid metabolic processes. Protein metabolism was notably impaired, evidenced by suppressed transferase and endopeptidase complexes. The transcriptional suppression of protein metabolic pathways in optic lobes under acute exposure contrasts with maintained whole-animal ammonia excretion rates (Fig. 1C), suggesting that either optic lobes contribute minimally to overall nitrogen metabolism or that compensatory responses in other tissues maintain systemic homeostasis. The ribosomal subunit and ribosome-related GO terms were considerably downregulated, while cell-cell adhesion pathways showed significant upregulation, emphasizing maintained cellular structural integrity under acute stress.

The GSEA tree plot revealed extensive suppression of biological process clusters under acute treatment (Fig. 4D). The key downregulated pathways included mitochondrial and energy metabolism components, particularly respirasome cytochrome, electron transport chain complex IV, and mitochondrial envelope. Fatty acid biosynthesis and elongase activity pathways showed substantial suppression, while downregulation of organelle and protein localization pathways indicated disrupted intracellular organization. The suppression of ncRNA ribonucleoprotein and 3’-end RNA processing suggested impaired RNA regulation mechanisms. Under chronic exposure, the GSEA tree plot demonstrated distinct downregulation patterns (Fig. 4E), particularly affecting neurotransmission and synaptic activity. This included reduced expression in postsynaptic membrane, synaptic membrane, transmitter-gated channel activity, and neurotransmitter receptor activity terms. Energy metabolism was compromised through suppressed NAD(P)H dehydrogenase activity and respiratory electron transport chain processes. Under chronic exposure, this downregulation of oxidative metabolism genes in optic lobes contrasts markedly with systemic physiological changes: whole-animal oxygen consumption increased by 41% (Fig. 1B) while ammonia excretion decreased by 52% (Fig. 1C). This divergence between localized neural transcriptomics and whole-organism physiology indicates fundamental metabolic reorganization, suggesting differential tissue responses to chronic acidification. Concurrent downregulation of cell adhesion molecules and prefoldin complex functions further suggests potential compromise of cellular structural integrity under prolonged acidification stress.

However, chronic exposure also triggered extensive compensatory responses, revealed through GO term analysis. Cellular components (CC) showed enhanced vesicular, endosomal, and Golgi apparatus pathways (Fig. 4F). Biological processes (BP) exhibited upregulated purine biosynthesis and carbohydrate catabolism (Fig. 4G), while molecular functions (MF) displayed elevated ribonucleoprotein activity and ubiquitin-related processes (Fig. 4H). Enhanced oxidoreductase activity suggested improved oxidative stress management, complemented by elevated hydrolase and transporter activities for metabolic byproduct removal.

Discussion

Laboratory-reared bigfin reef squid with well-controlled ontogeny stages exhibited significantly impaired predatory behavior after 7-days of exposure to ca. 1000 µatm P_CO₂, requiring nearly 2.5 times longer to capture prey. This behavioral alteration aligns with observations in pygmy squid (Ideosepius pygmaeus) after 5-day hypercapnic exposure⁸, and parallels their findings that bigfin reef squid exposed for 28 days showed significantly increased strike times, though the 14% reduction in attack willingness was statistically insignificant. However, despite these delays in prey capture, the initial strike success rate remained unaffected across treatments, indicating preserved motor capabilities under acidified conditions. While our findings focus on visual processing, the reduced predatory drive may potentially involve other sensory modalities, such as olfaction, whose contribution might be examined using chemical cue-based behavioral experiments.

Our implementation of 2D and 3D behavioral monitoring revealed that acute acidification induced distinct spatial preferences, with squid occupying significantly higher positions in the experimental arena. This spatial response presents an interesting contrast to typical anxiety-like behaviors observed in vertebrate models such as zebrafish, where acidification exposure typically induces “bottom-dwelling” responses in novel environments^12,79. The preference for shallower depths could be explained by impaired pressure sensing or environmental awareness processes acting as a protective reaction, though more mechanosensory study is needed to corroborate this idea. While cephalopod behaviors cannot be directly compared to vertebrate responses due to species-specific patterns⁸⁰, the observed shift in depth preference, combined with reduced prey interaction despite their normally voracious predatory nature, suggests complex behavioral alterations beyond simple anxiety responses. This interpretation is further supported by recent field studies on Loligo forbesi, which have documented natural depth preferences, with squid occupying lower depths during daylight hours and ascending during evening periods⁸¹. The disruption of these natural depth-regulation behaviors under acute acidification suggests fundamental alterations in spatial orientation mechanisms, potentially affecting their ecological fitness in natural environments.

Unlike previous reports of altered swimming patterns in wild S. lessoniana under similar CO₂ conditions⁸, we observed no changes in locomotory capacity in our laboratory-reared specimens. This discrepancy might reflect intrinsic differences between wild and laboratory-reared specimens, potentially due to geographical variations in S. lessoniana populations, which likely represent cryptic species complexes⁸². The diversity of behavioral responses to acidification among cephalopod species further emphasizes the complexity of these adaptations. While wild S. lessoniana and I. pygmaeus show increased activity under ~1000 µatm PCO₂^8,45, jumbo squid (Dosidicus gigas) exhibit substantially reduced activity (45% decrease) under similar conditions⁸³. These varied responses highlight the species-specific nature of behavioral adaptations to ocean acidification and underscore the importance of considering both developmental history and ecological context when interpreting behavioral changes in marine organisms. We acknowledge that the 7-day habituation in smaller chambers represents a methodological constraint. Nevertheless, preserved feeding behavior and consistent within-group responses, combined with our randomized design that distributed any confinement effects equally across treatments, support that the observed differences primarily reflect acidification rather than housing artifacts, though larger arena studies would provide valuable validation.

The switch of GABAergic neurotransmission from an inhibitory to an excitatory state is considered to be a core contributor to the anxiety-like behaviors observed following an animal’s exposure to ocean acidification^11,23. While gabazine has successfully attenuated OA-induced behaviors in pygmy squid²⁴, thermodynamic models that support the electrochemical change in HCO₃⁻ and Cl⁻ concentrations as a causative agent have only been calculated in teleosts exposed to P_CO2 levels that are nearly two times higher than most behavioral studies^5,84. Laboratory-reared bigfin reef squid in this study showed modest HCO₃^- accumulation within their blood during a 7-day acute acidification exposure, with a 1.7-fold increase (from 3.56 to 5.90 mmol l^–1). However, this partial compensation was insufficient to fully counteract the acidosis, as evidenced by the P_CO₂ increase (1.5-fold) and the persistent behavioral impairments. In contrast, chronically exposed squid achieved more complete compensation with a 3.1-fold increase in HCO₃^- (11.08 mmol l^–1), though this came at a significant metabolic cost (41% increase in oxygen consumption). The 2.34 mmol l^–1 HCO3⁻ elevation observed during our acute exposure exceeds the threshold (<1 mmol l^–1) previously suggested as insufficient to disrupt neuronal ion channels⁸⁵. This suggests that the behavioral impairments we observed may indeed result from HCO₃⁻-mediated neuronal effects, supporting the GABA hypothesis even with partial compensation. Moreover, while previous studies suggested branchial NH₄⁺ excretion as a key mechanism for acid-equivalent elimination^63,86, we observed no compensatory increase in NH₄⁺ excretion rates during acute acidification, despite baseline rates nearly double those reported in Loligo forbesi (1.7 µmol ammonia gFW⁻¹ h⁻¹) and Illex illecebrosus (1.43 µmol ammonia gFW⁻¹ h⁻¹). While the stability of these systemic and whole animal level traits cannot entirely rule out a potential dysregulation of GABAergic neurotransmission within the central nervous system, we did not observe changes in optic lobe transcripts associated with acid–base or ion regulatory processes that would foreseeably coincide with compensatory demands.

Chronic exposure to acidified conditions caused significant changes in transcripts related to synaptic function, including the downregulation of several nicotinic acetylcholine receptor subunits (alpha-7, alpha-10, and beta-4; Supplementary data_GSEA_results). These findings support growing evidence that cholinergic systems, along with GABAergic pathways, respond to hypercapnic stress across marine species. Previous research has shown varied cholinergic responses to ocean acidification. Upregulation of acetylcholine-related transcripts occurs in pteropods^34,35 and cephalopod neural tissues after short-term exposure²⁴. Similar compensatory responses have been observed in teleost fishes^31,32,87. However, downregulation also occurs, as seen in European sea bass olfactory tissue, where chrna7 expression decreases under elevated CO₂³³. Our 90-day exposure, representing about 50% of the bigfin reef squid’s lifespan, reveals that chronic acidification elicits responses fundamentally different from those seen in acute exposures. While most studies examine responses within 7 days, our extended exposure shows a shift from potential compensation to dysfunction in cholinergic signaling.

This temporal shift is particularly notable given that cephalopod cholinergic networks are dominated by nicotinic rather than muscarinic receptors^88,89, and are essential for prey capture⁹⁰, memory and learning^91,92. Recent pharmaceutical experiments have shown that multiple neural pathways contribute to behavioral changes in pygmy squid under acidification²⁴, which aligns with our observation of broad neuromodulatory effects through cholinergic dysfunction. Moreover, evidence from vertebrate studies suggests potential interactions between cholinergic and GABAergic systems. Reduced nicotinic acetylcholine receptor activity can prevent GABAergic neuron maturation⁹³, while certain nicotinic receptors directly stimulate GABA release in the central nervous system⁹⁴. Though these vertebrate-derived mechanisms must be cautiously applied to cephalopods, they suggest that cholinergic downregulation might indirectly contribute to GABAergic dysfunction under ocean acidification.

Chronic acidification induced complex metabolic responses in bigfin reef squid. Despite increased whole-animal oxygen consumption under chronic acidification, transcriptomic analysis revealed significant downregulation of mitochondrial function in the optic lobes, particularly affecting genes involved in oxidative phosphorylation. These observations suggest that the capacity for the optic lobes to perform aerobic metabolism is negatively impacted by chronic acidification, despite an overall greater oxygen consumption rate by the whole body, supporting a degree of tissue-specific metabolic responses to environmental stress as demonstrated in other marine species⁹⁵. The observed increase in routine metabolic rate under chronic conditions, previously hypothesized to relate to reduced predatory behavior⁸, may represent a compensatory response to maintain basic neural functions despite compromised cellular energy efficiency. This 41% metabolic elevation coincides with the substantial HCO₃⁻ accumulation achieved under chronic exposure (3.1-fold increase to 11.08 mmol l⁻¹), suggesting that the energetic cost of maintaining acid-base homeostasis contributes considerably to the overall metabolic burden. The sustained elevation in oxygen consumption, despite downregulated mitochondrial transcripts in the optic lobes, indicates that acid-base regulation and cellular maintenance processes demand considerable energy reallocation, potentially at the expense of complex behaviors such as predation.

Notably, chronic exposure to CO₂ led to significant decreases in NADPH dehydrogenase activity, indicating increased vulnerability to oxidative stress. The combination of mitochondrial dysfunction and reduced NADPH dehydrogenase activity suggests compromised antioxidant defenses, particularly given that cephalopods exhibit naturally lower antioxidative enzyme activity compared to other ectothermic animals⁹⁶. This reduced capacity for oxidative stress management may be especially critical in neural tissues like the optic lobes, where high metabolic demands make them particularly susceptible to oxidative damage. Recent studies in cuttlefish (Sepiella inermis) have reported increased apoptotic cells in eyes and optic ganglia following exposure to 1000–1800 µatm CO₂, correlating with reduced predatory success⁹⁷. While that study did not directly examine oxidative stress, our findings suggest that oxidative damage might be a key mechanism linking acidification exposure to neural dysfunction, particularly given the optic lobe’s role as a secondary retina⁹⁸.

Our transcriptomic analysis revealed extensive compensatory responses under chronic acidification. GSEA demonstrated coordinated upregulation across cellular components, BP, and MF in the optic lobes. RNA-seq analysis revealed coordinated upregulation of cellular maintenance pathways in the optic lobes, including enhanced vesicular transport, Golgi apparatus components, and protein metabolism. These compensatory mechanisms appear to impose energetic costs, as suggested by elevated routine metabolic rates occurring alongside reduced ammonia excretion to less than half of control levels. The concurrent enhancement of purine biosynthesis and carbohydrate catabolism pathways, together with increased oxidoreductase activity, indicates a potential shift from protein catabolism to alternative energy production mechanisms while maintaining oxidative stress management. In fact, the shift away from protein catabolism may not necessarily be restricted to the optic lobes, given that whole animal ammonia excretion rates were lowered by 52% when squid were exposed to acidified conditions for 90 days. This 52% reduction in ammonia excretion, occurring despite successful HCO₃⁻ compensation (11.08 mmol l⁻¹), indicates a fundamental shift in nitrogen metabolism rather than a failure of acid-base regulation. The preservation of elevated HCO₃⁻ levels while decreasing protein catabolism shows that chronic acidification causes metabolic reorganization beyond simple acid-base compensation. Although the calculation of atomic O: N ratios was not possible as oxygen consumption and ammonia excretion data were not paired to an individual, the concurrent increase in oxygen consumption (41%) and decrease in ammonia excretion (52%) strongly suggest a shift from protein-based to carbohydrate-based metabolism. Future studies employing intermittent-flow respirometry would enable paired measurements of oxygen consumption and ammonia excretion from the same individuals, allowing direct calculation of O: N ratios to quantify metabolic fuel switching. Such measurements would provide definitive evidence of the metabolic reorganization suggested by our separate measurements and could reveal whether the magnitude of this shift varies with exposure duration or among individuals.

These molecular adaptations suggest coordinated support for neural function under chronic stress, though the energetic cost remains to be quantified. The intensification of cellular maintenance processes was further evidenced by elevated hydrolase and transporter activities, facilitating efficient removal of metabolic byproducts. This comprehensive cellular response indicates a stress adaptation strategy where enhanced maintenance mechanisms may compensate for compromised energy production pathways, though direct causal relationships require further investigation. However, the elevated energy demands of these compensatory processes may explain the increased whole-animal oxygen consumption observed under chronic acidification. This complex interplay between impaired energy production, enhanced cellular maintenance, and compensatory mechanisms highlights the substantial physiological costs of adaptation to ocean acidification. While these responses may preserve essential neural functions, they likely contribute to the observed behavioral changes by altering the energy available for complex tasks such as predation. The distinct transcriptional profiles between acute and chronic exposure emphasize the temporal dynamics of adaptation, from initial stress response to long-term compensatory mechanisms in cellular maintenance and neurotransmission pathways.

Given the observed changes in predatory behavior and depth preferences, we initially hypothesized that visual impairment might underlie these behavioral alterations. Contrary to findings in fish, where ocean acidification directly impairs visual function^99,100, our ERG measurements revealed preserved temporal resolution across treatments. However, we noted a significant 4- to 5-fold increase in CFF variance under acidification conditions (control: 50.8 ± 1.4 Hz; acute: 50.8 ± 6.3 Hz; chronic: 47.7 ± 5.5 Hz), indicating that while population-level visual function seems maintained, individual responses to acidification vary widely. This increased variance may reflect different sensitivities to CO₂ stress among individuals, with some squid potentially experiencing subtle visual processing issues that are masked when averaging across populations. These varied responses match our transcriptomic results showing disrupted synaptic signaling in the optic lobes, which could cause diverse effects on sensory processing. This individual variation could have important ecological consequences, as differences in visual processing under acidification might affect prey capture ability and survival, potentially leading to population-level selection as ocean conditions change. Although basic visual function appears intact, probably because it’s vital for survival⁵¹, changes in the optic lobe transcriptomes might mean that interpreting visual stimuli is more sensitive to ocean acidification and could relate to behavioral shifts in predator-prey interactions. Besides the possible impact of changes in cholinergic neurotransmission on behavior, we also saw an overall increase in gene sets tied to cellular maintenance and metabolic reorganization within the optic lobes. Future studies exploring whether these transcript-level changes lead to actual increases in maintenance costs in the optic lobes would help clarify if energetic limits compromise neural functions. Such research could include metabolic flux analysis, ATP measurements, or live imaging of neural activity under acidified conditions.

Taken together, this study reveals that despite cephalopods’ capacity to regulate acid-base balance in acidified seawater^63,101, projected 2100 ocean acidification levels significantly impact metabolic and synaptic processes in bigfin reef squid optic lobes. While squids maintain fundamental visual capabilities, behavioral changes appear to stem from alterations in higher-order neural processing, particularly in the integration of visual information with motor control systems. The observed disruptions in cholinergic networks and redox balance likely underlie reduced prey attack behavior, suggesting parallels with vertebrate neuromodulation systems¹⁰².

Our findings emphasize the complexity of acidification responses across biological organization levels. While adaptive mechanisms preserve primary sensory functions, energy-intensive processes of neural integration and behavioral execution show significant impairment. Transcriptomic analysis reveals that under chronic acidification, enhanced cellular maintenance mechanisms attempt to compensate for compromised energy production, though at substantial metabolic cost. This is evidenced by increased oxygen consumption despite downregulation of mitochondrial processes and structural proteins, indicating fundamental alterations in energy utilization. The progression from partial acute compensation to more complete chronic compensation, as evidenced by our corrected carbonate chemistry data, demonstrates that while physiological adaptation is possible, it comes at considerable metabolic cost. From an ecological perspective, these adaptations may have significant implications. As key marine predators, altered hunting efficiency in squid could potentially reshape predator-prey dynamics and marine food webs, though field studies would be needed to confirm such ecosystem-level effects. The observed metabolic trade-offs suggest a possible compromise of growth and reproduction, functions already susceptible to climate change^63,103–105.

In conclusion, while demonstrating remarkable physiological plasticity, the long-term suppression of key metabolic and structural processes raises concerns about sustained adaptation to intensifying ocean acidification. This study underscores the intricate relationship among environmental stress, neurophysiological adaptation, and ecological function, highlighting urgent research needs regarding the long-term consequences of ocean acidification for marine ecosystems.

Author contributions

G.J.P.A.: Conceptualization, Methodology, Investigation, Formal analysis, Writing the original draft. J.J.Y.: Investigation, Data curation, Formal analysis, Visualization. P.L.K.: Investigation, Data curation, Formal analysis, Resources. O.H.: Investigation, Data curation, Formal analysis, Visualization. S.C.H.: Investigation, Data curation, Formal analysis, Visualization. M.Y.J.L.: Data curation, Formal analysis. K.A.: Investigation, Data curation. G.C.W.: Resources, Conceptualization, Methodology, Funding acquisition. Y.C.T.: Resources, Conceptualization, Methodology, Formal analysis, Writing the original draft, Supervision, Funding acquisition, Project administration.

Peer review

Peer review information

Communications Biology thanks Kirt L. Onthank and the other anonymous reviewer for their contribution to the peer review of this work. Primary Handling Editors: Michele Repetto and George Inglis. A peer review file is available.

Data availability

All data supporting the findings of this study are publicly available. Raw physiological data (metabolic rates, ammonia excretion rates, and blood acid-base parameters), behavioral tracking data, electroretinogram (ERG) recordings, carbonate chemistry calculations, verification procedures, along with all bioinformatics code availability, are deposited in the Dryad Digital Repository (10.5061/dryad.rjdfn2zqz). RNA-sequencing data have been deposited in the NCBI Sequence Read Archive (SRA) under BioProject accession PRJNA1204452. The assembled transcriptome is available in the NCBI Transcriptome Shotgun Assembly (TSA) database under accession GLBY00000000 (version GLBY01000000). Statistical analyses were performed using standard R packages (v4.3.2) and GraphPad Prism (v10.2.3) as detailed in the Methods. Custom scripts, if any, are included in the Dryad repository. All materials are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-025-09506-6.