Toward a Mechanistic Understanding of Marine Invertebrate Behavior at Elevated CO2

Elevated carbon dioxide (CO2) levels can alter ecologically important behaviors in a range of marine invertebrate taxa; however, a clear mechanistic understanding of these behavioral changes is lacking. The majority of mechanistic research on the behavioral effects of elevated CO2 has been done in fish, focusing on disrupted functioning of the GABAA receptor (a ligand-gated ion channel, LGIC). Yet, elevated CO2 could induce behavioral alterations through a range of mechanisms that disturb different components of the neurobiological pathway that produces behavior, including disrupted sensation, altered behavioral choices and disturbed LGIC-mediated neurotransmission. Here, we review the potential mechanisms by which elevated CO2 may affect marine invertebrate behaviors. Marine invertebrate acid–base physiology and pharmacology is discussed in relation to altered GABAA receptor functioning. Alternative mechanisms for behavioral change at elevated CO2 are considered and important topics for future research have been identified. A mechanistic understanding will be important to determine why there is variability in elevated CO2-induced behavioral alterations across marine invertebrate taxa, why some, but not other, behaviors are affected within a species and to identify which marine invertebrates will be most vulnerable to rising CO2 levels.


INTRODUCTION
Human activity is resulting in unprecedented amounts of carbon dioxide (CO 2 ) being released into the atmosphere. Since the Industrial Revolution, atmospheric CO 2 levels have increased by over 45%, from approximately 280 ppm (Joos and Spahni, 2008) to over 410 ppm today (Dlugokencky and Tans, 2019), higher than any time in the past several million years (Masson-Delmotte et al., 2013). In the worst case scenario, following the business-as-usual representative concentration pathway (RCP) 8.5, atmospheric CO 2 levels will increase to over 900 ppm by the end of this century. Even if substantial efforts are made to curb global CO 2 emissions to keep warming below 2 • C, atmospheric CO 2 levels will still likely exceed 600 ppm by 2100 (Betts and McNeall, 2018). The ocean has absorbed 20-30% of anthropogenic CO 2 emissions since the mid-1980s (Bindoff et al., 2019), causing a reduction in seawater pH referred to as ocean acidification. Furthermore, CO 2 in the surface ocean is increasing at the same rate as in the atmosphere (Bindoff et al., 2019); therefore, marine organisms will need to cope with higher CO 2 levels as well as declining seawater pH. Finally, due to a decrease in the ocean's buffering capacity as CO 2 content rises, natural CO 2 fluctuations in the ocean are projected to amplify dramatically at future higher CO 2 levels (Shaw et al., 2013;McNeil and Sasse, 2016). Natural diel (Hofmann et al., 2011;Santos et al., 2011;Shaw et al., 2012) and seasonal CO 2 fluctuations (McNeil et al., 2007;Feely et al., 2008) will be amplified by up to three times in the future (McNeil and Sasse, 2016;Gallego et al., 2018), meaning that marine organisms will experience elevated CO 2 levels for certain periods of time (daily or seasonally) much earlier than predictions based on atmospheric CO 2 alone.
Elevated CO 2 has been found to affect a range of processes in marine organisms, including altering calcification (Ries et al., 2009;Kroeker et al., 2013), growth and survival (Fabry et al., 2008;Kurihara et al., 2008), and behavior Clements and Hunt, 2015;Nagelkerken and Munday, 2015). Projected future CO 2 levels were first found to alter animal behavior in orange clownfish Amphiprion percula larvae reared in a partial pressure of CO 2 (pCO 2 ) of ∼1,050 µatm (Munday et al., 2009). In laboratory experiments, the olfactory discriminatory abilities of 11-day-old clownfish larvae were tested in a twochannel flume. Most strikingly, larvae reared in control seawater (∼390 µatm pCO 2 ) avoided the side of the flume with chemical cues from pungent tree leaves compared to the seawater control side. However, larvae reared in elevated CO 2 spent nearly all their time in the side with these odors. Larval clownfish reared in ∼1,050 µatm CO 2 were also unable to discriminate between the odor of parents and non-parents, whereas control larvae avoided the odor of their parents (Munday et al., 2009). Elevated CO 2 has since been found to affect a variety of behavioral traits in a wide spectrum of fishes, including tropical and temperate reef species, eels, salmon and sharks . Behavioral alterations at elevated CO 2 have also been demonstrated in a variety of marine invertebrates, including cnidaria, polychaetes, echinoderms, arthropods and mollusks, from a range of environments, including the intertidal zone, coastal and offshore waters and the deep-sea (Clements and Hunt, 2015;Nagelkerken and Munday, 2015;Wang and Wang, 2019). Marine invertebrates exhibit alterations in a range of behaviors at elevated CO 2 , including activity levels (Rosa and Seibel, 2008;Ellis et al., 2009;Spady et al., 2014), feeding rates (Saba et al., 2012;Vargas et al., 2014), settlement and metamorphosis behaviors (Albright et al., 2010;Doropoulos et al., 2012;Guo et al., 2015), burrowing behaviors (Green et al., 2013;Clements and Hunt, 2014), shelter selection (de la Haye et al., 2011), predatory behaviors (behaviors related to finding and eating prey) (Kim et al., 2015;Queirós et al., 2015;Spady et al., 2018) and predator avoidance (Bibby et al., 2007;Manríquez et al., 2013Manríquez et al., , 2014aSpady et al., 2014;Watson et al., 2014). Behavioral categorization is often ambiguous as one behavior may actually include multiple behaviors or decisionmaking processes. For example, predator avoidance behaviors include multiple decisions such as mode of avoidance (including crypticity versus escape), flight-initiation distance and mode of escape (Lima and Dill, 1990). In this review, we use the behavioral category that was reported in the corresponding research paper.
Since the review by Clements and Hunt (2015) at least 61 additional papers have assessed the impact of elevated CO 2 on marine invertebrate behaviors (Supplementary Table S1).
Research has continued to focus on mollusks, arthropods and echinoderms, however, cnidarian settlement and metamorphosis (Foster et al., 2015;Olsen et al., 2015;Viyakarn et al., 2015;Fabricius et al., 2017;Yuan et al., 2018), the settlement behavior and swimming activity of a bryozoan (Pecquet et al., 2017), and settlement of an annelid (Nelson et al., 2020) have also been studied. In addition to continuing to assess the impact of elevated CO 2 on the range of behaviors previously studied (above), a few new behaviors have also been investigated. For example, the first study assessing the effect of elevated CO 2 on marine invertebrate reproductive behavior was recently published (Borges et al., 2018). Exposure of male amphipods Gammarus locusta to elevated CO 2 (800 µatm pCO 2 ) for two generations disrupted the chemosensory detection of potential mates (Borges et al., 2018). A light/dark test on swimming crabs Portunus trituberculatus exposed to control (485 µatm pCO 2 ) or elevated (750 µatm and 1,500 µatm pCO 2 ) CO 2 was the first to assess the impact of elevated CO 2 on anxiety-like behavior in a marine invertebrate. Crabs exposed to elevated CO 2 levels spent significantly more time in the dark zone (Ren et al., 2018).
There appears to be large variability in behavioral responses to elevated CO 2 ; across taxonomic groups, the same behavior can respond differently to elevated CO 2 , and within a species, some behaviors but not others can be affected (Nagelkerken and Munday, 2015) (Supplementary Table S1). As the phylogenetic variation among invertebrate taxa is enormous, it may account for a large component of the variability in behavioral responses across taxonomic groups; some taxa may be more tolerant to elevated CO 2 than others. At the same time, various behaviors are likely associated with different processes, such as specific circuits in the nervous system. These processes may be affected differently by elevated CO 2 , accounting for the effects of elevated CO 2 on some, but not all, behaviors within a species. It must also be noted that variability may be due to differences in experimental techniques and conditions. Animal behavior is, to a large extent, the functional output of the nervous system (Simmons and Young, 1999); therefore, behavioral changes induced by elevated CO 2 are likely caused by neurobiological mechanisms. In the nervous system, simplistically, the pathway that produces behavior involves sensory receptors that detect environmental stimuli (e.g., chemical cues, light waves) and internal stimuli (e.g., spatial orientation of the body). The received information is transduced into electrical impulses and neurotransmission relays these electrical impulses between neurons. Neurotransmission must be rapid to produce timely behavioral responses, and this is achieved via ligand-gated ion channel (LGIC) mediated neurotransmission (Dent, 2010).
LGICs are transmembrane protein complexes that, upon binding of a specific neurotransmitter, allow ion flow which results in excitation or inhibition of neuronal firing depending on the ion charge and direction of flow (Tovar and Westbrook, 2012). When the sensory information arrives at higher centers of the nervous system, this information is processed and a behavioral output is produced (Blom, 1978;Kreher et al., 2008) (Figure 1).
Elevated CO 2 could induce behavioral alterations through a range of mechanisms in the nervous system that disturb different FIGURE 1 | Simplistic pathway of how the nervous system produces behaviors and potential ways elevated CO 2 could alter marine invertebrate behavior. An external or internal stimulus is detected by sensory receptors located on a sense organ. A physical stimulus binding to a receptor, e.g., a chemical cue binding to the corresponding receptor (chemoreception), is depicted. However, stimuli and receptors may range from photoreceptors detecting light energy to stretch receptors detecting body movement or hair cells detecting vibrations. The detected sensory information is then transduced into electrical impulses that are relayed between neurons to higher centers of the nervous system. Here, the information is processed which includes using external and internal contextual factors to make behavioral choices. The information is transmitted between neurons to the motor system and a behavioral response is produced. Throughout this process, neurotransmission is used to relay the electrical impulses from neuron to neuron. Elevated CO 2 may alter behavior by interfering at multiple points along this pathway. Sensation may be disrupted by changes to sensory stimuli at elevated CO 2 through structural change of chemical cues, increased sound transmission and altered sensory output from animals. Sensation may also be disrupted on the receiving end via altered structure of sensory receptors or physical change to sensory organs. Elevated CO 2 may change the context within which decision-making is carried out, thus influencing behavioral choices. The change in ion gradients across neuronal membranes, due to acid-base regulation at elevated CO 2 , may disrupt LGIC-mediated neurotransmission, glial-neuronal signaling and the role of Cl − as a signaling effector. Numbers represent references providing evidence for each mechanism in marine invertebrates, while a * indicates this mechanism is based on theory with no experimental evidence in marine invertebrates. 1 Roggatz et al. (2016), 2 Maneja et al. (2011), 3 de la Haye et al. (2012), 4 Bibby et al. (2007, 5 Chan et al. (2011), 6 Dissanayake and Ishimatsu (2011), 7 Li and Gao (2012), 8 Peng et al. (2017), 9 Wang et al. (2018), 10 (Rich et al., 2018), 11 Watson et al. (2014, 12 Charpentier and Cohen (2016), 13 Moya et al. (2016), 14 Clements et al. (2017, 15 Ren et al. (2018), 16 Zlatkin and Heuer (2019).
components of the pathway that produces behavior , as described in Figure 1]. (1) Sensation may be disrupted via changes to sensory stimuli at elevated CO 2 , such as structural alteration of chemical cues, disturbed transmission of acoustic cues and altered sensory output from animals experiencing behavioral changes (Roggatz et al., 2016;Nagelkerken et al., 2019). (2) Alternatively, elevated CO 2 may disrupt sensation via physical change to sensory organs or structural alteration of sensory receptors (Maneja et al., 2011;Briffa et al., 2012;Bignami et al., 2013). (3) Morphological and respiratory changes at elevated CO 2 may alter the context within which decision making is carried out, influencing behavioral responses (Bibby et al., 2007;Chan et al., 2011;Peng et al., 2017;Rich et al., 2018).
(4) Elevated CO 2 may induce change in ion gradients across neuronal membranes, due to acid-base regulation to prevent acidosis at elevated CO 2 , which may disrupt LGIC-mediated neurotransmission via the γ-aminobutyric acid (GABA) type A receptor (GABA A R) (Nilsson et al., 2012). These mechanisms are not necessarily mutually exclusive and may interact to alter behavior at elevated CO 2 .
Despite the growth in literature demonstrating elevated CO 2 -induced behavioral alterations in marine invertebrates, the mechanisms responsible for these behavioral alterations are still poorly understood. Due to the diversity of invertebrate nervous and neurobiological systems, it is likely that a suite of different processes underlie these behavioral changes. Here, we discuss the potential mechanisms by which elevated CO 2 may alter marine invertebrate behavior; (1) disturbed sensation, (2) altered context within which behavioral choices are made, and (3) disrupted LGIC-mediated neurotransmission. Since the prominent hypothesis for altered LGIC-mediated transmission is the GABA hypothesis proposed in fish by Nilsson et al. (2012), here we discuss evidence for the GABA hypothesis in marine invertebrates and propose research to expand our understanding of the GABA hypothesis in marine invertebrates. We demonstrate that the effects of altered GABA A receptor function are likely to be widespread, including non-behavioral effects. Finally, we identify other neurobiological mechanisms that should be affected if the GABA hypothesis is correct, propose alternative neurobiological mechanisms by which behavior could be altered by elevated CO 2 and suggest techniques to be utilized for future study of elevated CO 2 -induced behavioral alterations.

MECHANISMS FOR ELEVATED CO 2 -INDUCED BEHAVIORAL CHANGES
Altered Sensory Stimuli at Elevated CO 2 Elevated CO 2 may influence behavior by altering an animal's ability to sense the environment Draper and Weissburg, 2019). A range of sensory stimuli may be disrupted at elevated CO 2 via differing mechanisms, thereby affecting associated behaviors (Figure 1). Structural alteration of chemical cues at elevated CO 2 may affect chemoreceptionthe detection of chemical cues by binding to sensory receptors, e.g., odor molecules binding to olfactory receptors (Tierney and Atema, 1988). Impaired chemo-responsive behavior was first shown to be due to structural alteration of the chemical cue at low pH in a freshwater system (Brown et al., 2002) and the same mechanism has since been demonstrated in a marine invertebrate; the shore crab Carcinus maenas. Near-future pH levels altered the structure and charge of signaling molecules that mediate egg ventilation behavior in the shore crab C. maenas and the threshold of signaling molecule concentration required to induce egg ventilation behavior in this species increased when tested at pH 7.7 compared to pH 8.1 (Roggatz et al., 2016). Receptor alteration, such as change in ionization state, could also conceivably occur at low pH disrupting chemoreception (Tierney and Atema, 1988), and may be an additional explanation for behavioral changes observed in the shore crab (Roggatz et al., 2016). However, to date receptor structure has never been directly tested at different CO 2 levels in conjunction with a behavioral assay.
Changes in ocean chemistry associated with rising CO 2 levels will directly affect acoustic cues, potentially altering auditory driven behaviors. Sound absorption, in the low frequency range of ∼0.01-10 kHz, is reduced by decreasing pH due to shifts in the chemical reactions of sound absorbing compounds (e.g., magnesium sulfate, boric acid and carbonate ions) in seawater (Hester et al., 2008). Sound absorption (decibels per kilometer) is predicted to decrease by over 20% and almost 40% with a pH drop from 8.1 to 7.95 and 7.8, respectively (Hester et al., 2008). Thus, as CO 2 levels rise, transmission of low-frequency sounds will increase and ecologically relevant acoustic cues, within this affected frequency range, will be transmitted further at elevated CO 2 levels. For example, compared to off reeflocations, oyster reefs have higher acoustic energy levels within the frequency range of 1.5-20 kHz. This acoustic signature of reefs is used as a settlement cue by larval oysters (Lillis et al., 2013). As CO 2 levels rise, oyster larvae may thus be able to detect appropriate settlement habitats from greater distances, however, it remains unknown whether the magnitude of change is sufficient to be of biological relevance. Elevated CO 2 will also increase the transmission of abiotic sounds produced naturally (e.g., waves, raindrops) and by human activity (e.g., shipping, sonar and construction) (Ilyina et al., 2010). This will create a noisier environment in which it is harder for marine invertebrates to detect ecologically relevant sounds, such as those used for communication (Popper et al., 2001;Buscaino et al., 2011) as well as navigation and habitat selection for settlement (Jeffs et al., 2003;Stanley et al., 2009;Vermeij et al., 2010;Lillis et al., 2013). In a coral reef fish, predatory behavior decreased when exposed to boat noise or elevated CO 2 (925 µatm pCO 2 ), however, there was no additive effect when these stressors co-occurred (McCormick et al., 2018). Studies in marine invertebrates to determine how increased transmission of biologically relevant cues and background noise will interact as CO 2 levels rise, and if this will be biologically relevant will be important.
Behavioral changes induced by elevated CO 2 may alter the sensory output of an animal, affecting whether and how this animal is sensed by other animals (Draper and Weissburg, 2019). For example, increased activity of a prey animal could enhance how much or how often sound, visual and mechanosensory cues are produced, strengthening predatory sensory detection of the prey and increasing the chance of predation (Draper and Weissburg, 2019). Elevated CO 2 reduced the intensity and frequency of snaps produced by snapping shrimp (Rossi et al., 2016). As these snaps are commonly present in the soundscapes used by marine invertebrate larvae as settlement cues (Stanley et al., 2009;Vermeij et al., 2010;Lillis et al., 2013), altered snapping behavior of snapping shrimp at elevated CO 2 may in turn alter marine invertebrate settlement behavior.

Physical Change of Sensory Organs
Sensation may also be disrupted by physical change of sensory organs at elevated CO 2 . Due to lower saturation states of seawater with respect to calcium carbonate at elevated CO 2 , animals can have difficulty maintaining calcium carbonate structures (Orr et al., 2005) which may damage sensory organ structures . Alternatively, active acid-base regulation to maintain a steady internal pH may alter the concentrations of ions that are fundamental for the formation of calcified sensory organs (Grosell, 2019). For example, many marine invertebrates use statocysts, which contain mineralized statoliths, to detect gravity to maintain orientation (Cohen, 1960;Clarke, 1978;Spangenberg, 1986), as well as vibrational stimuli for hearing in cephalopods (Mooney et al., 2010). Statocysts are also involved in motor programs that underlie hunting behavior in mollusks (Levi et al., 2004). Statolith size was reduced and morphology altered in cephalopods exposed to ∼1,300 µatm (Zakroff et al., 2019), 2,200 µatm (Kaplan et al., 2013), and 4,000 µatm (Maneja et al., 2011) pCO 2 . Abalone exposed to ∼700 and ∼1,000 µatm pCO 2 also exhibited decreased statolith size compared to control conditions (Manríquez et al., 2014b). Conversely, statolith size was increased, and chemical composition altered, in squid exposed to 850 and 1,500 µatm pCO 2 (Lacoue-Labarthe et al., 2011). Cuttlefish exposed to 4,000 µatm pCO 2 exhibited reduced statolith calcification, altered statolith microstructure and decreased prey capture efficiency compared to squid in control conditions (700 µatm pCO 2 ) (Maneja et al., 2011). Furthermore, computer modeling showed that an increased statolith mass, similar to that seen in the otoliths of fish exposed to 2,500 µatm pCO 2 , would alter cephalopod hearing below 10 Hz (Zhang et al., 2015). However, squid exposed to ∼1,300 µatm pCO 2 had smaller statoliths with an altered morphology (Zakroff et al., 2019) but no impairment in swimming orientation (Zakroff et al., 2018). Therefore, elevated CO 2 -induced alteration of statoliths may disturb hearing but not gravity detection in cephalopods, impacting auditory-driven behavioral outputs but not the ability to maintain orientation. Decapod crustaceans possess calcified antennules, housing chemoreceptors, which are used for long range chemoreception. Rapid antennule flicking is used to gather chemical cue information, much in the way sniffing increases our ability to determine smells (Schmitt and Ache, 1979;Koehl, 2005). Hermit crabs with disrupted chemo-sensory responses at extremely high levels of CO 2 (c. >12,000+ µatm pCO 2 ) showed no damage to their antennules , suggesting that other mechanisms must be responsible for the observed response.

Altered Behavioral Choices
The physiological and ecological context, including external factors (e.g., presence of predators or temperature) and internal factors (e.g., hunger or reproductive state) can influence behavioral choices (Palmer and Kristan, 2011). Elevated CO 2 may alter both external and internal factors, changing contextual modulation of behavioral choice and resulting in altered behavioral output. Physical changes induced by elevated CO 2 may alter an animal's behavioral choice. For example, predator-induced shell thickening observed in control periwinkles Littorina littorea did not occur in periwinkles exposed to extremely high levels of CO 2 (c. >12,000+ µatm). However, predator avoidance behavior increased at elevated CO 2 conditions, compared to control, which suggests behavioral compensation for the lack of morphological defense at extremely high levels of CO 2 (Bibby et al., 2007). In another example, the swimming performance of larval sand dollars Dendraster excentricus was maintained at elevated CO 2 (∼1,000 µatm pCO 2 ) despite impaired arm and body morphology, likely due to a behavioral change in ciliary beat patterns (Chan et al., 2011). By contrast, both predator cue-induced byssal thread production and protective clustering behavior was decreased in mussels exposed to elevated CO 2 (1,100 µatm pCO 2 ) (Kong et al., 2019), indicating no behavioral compensation for the lack of morphological defense at elevated CO 2 .
The influence of elevated CO 2 on respiration, energy turnover and mode of metabolism (Pörtner et al., 2004) may also alter behavioral choice. Depressed metabolism at elevated CO 2 levels may reduce energy production, reducing the energy available to meet other demands and constraining performance of some behaviors. For example, metabolic scope and swimming ability were reduced in shrimp exposed to ∼1,000 µatm pCO 2 (Dissanayake and Ishimatsu, 2011) and oxygen consumption rate and digging depth were decreased in razor clams exposed to 1,900 and 3,000 µatm pCO 2 . Increased metabolism can indicate an increased energy demand at elevated CO 2 and may decrease the energy available for other costly processes. For example, crabs exposed to 1,200 and 2,300 µatm pCO 2 exhibited an increased metabolic rate but a decreased feeding rate . Alternatively, organisms may alter behaviors to meet the high energy demand. For example, respiration and feeding rate were increased in a copepod exposed to 1,000 µatm pCO 2 (Li and Gao, 2012) and a sea urchin exposed to 1,300 µatm pCO 2 (Rich et al., 2018). However, other studies show a change in metabolism with no associated behavioral change at 750 and 1,200 µatm pCO 2 in an echinoderm (Carey et al., 2016) and at 1,500 µatm pCO 2 in a mollusk (Benítez et al., 2018), or no metabolic change but altered behavior at 960 µatm pCO 2 in a mollusk  and at 1,000, 2,000, and 3,000 µatm pCO 2 in a crustacean (Menu-Courey et al., 2018). Therefore, it seems that altered metabolism in elevated CO 2 may be responsible for some instances of altered behaviors, but not others.

Altered Functioning of the GABA A Receptor
Elevated CO 2 has been found to alter a range of behaviors across different sensory modalities, as well as behaviors that involve higher order processing, such as decision making (de la Haye et al., 2011) and anxiety-like behaviors (Ren et al., 2018). This suggests that not only sensory detection, but also other neuronal processes are altered by elevated CO 2 . Neurotransmission is crucial to all components of the pathway producing behavior, from relaying electrical signals from sensory receptors to (and between) higher-order neurons for information processing, to motor neurons for the production of a behavioral response. Therefore, altered neurotransmission at elevated CO 2 may underlie a variety of behavioral disturbances.
The prominent hypothesis for altered neurotransmission at elevated CO 2 is the GABA hypothesis, proposed to occur in fish (Nilsson et al., 2012;Heuer and Grosell, 2014) and also suggested to apply to marine invertebrates . In vertebrates, GABA acts on the GABA type A receptor (GABA A R), a LGIC permeable to chloride (Cl − ) and bicarbonate (HCO 3 − ) ions, as the main inhibitory neurotransmitter in the central nervous system (DeFeudis, 1975;Bormann et al., 1987). Under normal conditions, binding of GABA opens the GABA A R channel allowing a net influx of negative charge resulting in hyperpolarization and inhibition of neuronal firing. Nilsson et al. (2012) proposed that the change in HCO 3 − and Cl − gradients across the neuronal membrane, due to acid-base regulation at increased CO 2 levels, results in a net efflux of negative charge from the GABA A R upon GABA binding (Figure 2A). This could cause depolarization and excitation of neurons, thereby altering behavioral responses. Pharmacological studies administering GABA A R antagonists and agonists (Nilsson et al., 2012;Hamilton et al., 2013;Chivers et al., 2014;Chung et al., 2014), and measurements of brain ion gradients (Heuer et al., 2016) have supported this hypothesis in fish. − and Cl − gradients across the neuronal membrane, due to acid-base regulation at elevated CO 2 , was proposed to reverse the net flow of negative charge through the GABA A R (Nilsson et al., 2012) and likely applies to the marine invertebrate iGABA R. Under normal conditions the net influx of negative charge is primarily carried by Cl − , while at elevated CO 2 the net efflux of negative charge may be primarily carried by HCO 3 − (see Heuer et al., 2019 for a detailed explanation). (B) A range of invertebrates also possess glutamate, acetylcholine, histamine and serotonin-gated chloride channels. The flow of Cl − through these channels may be similarly altered at elevated levels of CO 2 . Research is needed to determine whether these receptors are also permeable to HCO 3 − which could also contribute to the reversal of the net movement of negative charge. A net influx of negative charge will result in hyperpolarization and inhibition of neuronal firing, while a net efflux of negative charge will cause depolarization and excitation. Glu, glutamate; ACh, acetylcholine; 5-HT, serotonin.

THE GABA HYPOTHESIS IN MARINE INVERTEBRATES
GABA is the main inhibitory neurotransmitter in the invertebrate peripheral and central nervous systems (Lummis, 1990;Lunt, 1991), acting on the ionotropic GABA receptor (iGABA R), which is also permeable to Cl − and HCO 3 − ions (Kaila and Voipio, 1987). Thus, the GABA hypothesis likely applies to marine invertebrates. The GABA hypothesis has been tested in marine arthropods and mollusks using pharmacological studies Charpentier and Cohen, 2016;Clements et al., 2017), measurement of ion concentrations Charpentier and Cohen, 2016) and molecular studies (Moya et al., 2016;Ren et al., 2018) (Table 1).
One method of assessing the GABA hypothesis involves administering the GABA A R antagonist gabazine (SR-95531) (Heaulme et al., 1986). If GABA A R functioning is altered at elevated CO 2 , gabazine administration should reverse elevated CO 2 -induced behavioral alterations by inhibiting channel opening and thus blocking the altered ion flow (Nilsson et al., 2012). Indeed, impaired escape behavior caused by exposure to 961 µatm pCO 2 was restored to control levels by gabazine in the jumping conch snail Gibberulus gibbosus . In the soft shell clam Mya arenaria, burrowing behaviors altered by CO 2 sediment levels representing present day variation were restored by gabazine (Clements et al., 2017). By contrast, in Asian shore crab larvae Hemigrapsus sanguineus, the loss of chemical cue-induced photosensitive behavior at elevated CO 2 conditions (1,380 µatm pCO 2 ) was not restored by gabazine (Charpentier and Cohen, 2016). These contrasting results initially appear to suggest that the GABA hypothesis applies to some, but not other, marine invertebrate taxa. However, crustacean iGABA Rs are commonly insensitive to gabazine (El Manira and Clarac, 1991;Jackel et al., 1994;Pearstein et al., 1996;Barry, 2002), meaning that gabazine may be inadequate for testing the GABA hypothesis in crustaceans.
It is interesting to note that the action of gabazine differed across control animals; gabazine significantly altered the behavior of control crab larvae (Charpentier and Cohen, 2016), had a non-significant trend of altering the behavior of control snails , and did not alter control clam burrowing behavior (Clements et al., 2017). As gabazine also blocks ion flow under normal conditions, preventing inhibition, over-excitation and behavioral alterations should occur in control animals. As gabazine does not appear to affect crustacean iGABA Rs, the behavioral change observed in crabs held at control CO 2 levels may be through the action of gabazine on a different pathway, such as a different receptor type. Characterizing the pharmacology of gabazine in the studied species and using a range of GABA A R drugs will be important to confirm the GABA hypothesis is actually being tested (see the section "Pharmacological Considerations"). Control: 1 and 10 µM gabazine loss of chemical cue-induced photosensitive behavior. 0.1 µM gabazine no effect. CO 2 treatment: 10 µM gabazine no effect Mechanistic support for the GABA hypothesis in marine invertebrates also comes from recent studies indicating changes in ion concentration and altered behavior in the same species at elevated CO 2 ( Table 1). Hermit crabs Pagurus bernhardus exhibited impaired chemosensory responses to a food odor and increased hemolymph Cl − concentration ([Cl − ]) at extremely high (12,061 µatm pCO 2 ) compared to control (373 µatm pCO 2 ) conditions (de la Haye et al., 2012). Asian shore crab larvae had altered chemical cue-induced photosensitive behavior and increased extracellular osmolality, but similar extracellular [Cl − ] at elevated CO 2 (1,380 µatm pCO 2 ) compared to controls (461 µatm pCO 2 ) (Charpentier and Cohen, 2016). However, the [Cl − ] measurements were very close to the limit of detection, which may be why no difference was observed. The increased extracellular osmolality was suggested to be due to an increase in HCO 3 − concentration ([HCO 3 − ]), however, [HCO 3 − ] was not directly measured. In a more recent study, sea hares Aplysia californica exposed to elevated CO 2 (1,200 and 3,000 µatm pCO 2 ) showed a reduced antipredator response and increased hemolymph [HCO 3 − ] compared to control (400 µatm pCO 2 ) (Zlatkin and Heuer, 2019 Molecular studies also provide support for the GABA hypothesis in marine invertebrates ( Table 1). Transcriptomic analysis of the Mediterranean pteropod Heliconoides inflatus exposed to elevated CO 2 (617-720 µatm pCO 2 ) for 3 days showed upregulation of the transcript encoding a GABA A R subunit (Moya et al., 2016). However, RNA was extracted from the whole animal, potentially masking differential expression within the nervous system, and behavioral assays were not carried out. Ren et al. (2018) isolated the gene encoding the GABA A R associated-protein (GABARAP) in the crab Portunus trituberculatus. Real-time PCR found that GABARAP mRNA levels were significantly upregulated by 4.34-fold after 6 h at 750 µatm pCO 2 and by 2.89-fold after 3 h at 1,500 µatm pCO 2 , compared to control (485 µatm pCO 2 ). Showing a similar trend, average speed of the crab shoal's movement was significantly higher after 3 and 6 h at 750 µatm pCO 2 , and six and 12 h at 1,500 µatm pCO 2 , compared to control. The increase in the GABARAP gene was suggested to assist more GABA A Rs to cluster on neuronal membranes, which may exaggerate the impaired function of GABA A Rs at elevated CO 2 and lead to the altered behavior (Ren et al., 2018).

iGABA R Subtypes and Variability in Elevated CO 2 -Induced Behavioral Alterations
Inter-and intra-species variation in the behavioral effects of elevated CO 2 may be due to the presence and variability of iGABA R subtypes in invertebrates. The ion permeable pore of iGABA Rs is made up of five subunits (Olsen and Sieghart, 2008). There is large variation in gene structure and the number of genes encoding iGABA R subunits between invertebrate species. For example, 5 GABA R-like genes have been found in the sea-squirt Ciona intestinalis, 12 in the fruit fly Drosophila melanogaster and 39 in the roundworm Caenorhabditis elegans (Tsang et al., 2006). Differing subunit composition forms iGABA R subtypes (Olsen and Sieghart, 2008) which vary in a range of functional properties including GABA binding affinity, and Cl − and HCO 3 − permeability (Lee and Maguire, 2014). Differences in iGABA R subunits may account for the variability in behavioral alterations at elevated CO 2 observed between invertebrate species. Furthermore, subunit composition can vary between regions in the nervous system and cell types (Lee and Maguire, 2014). As different behaviors are driven by different nervous system regions, this may explain why some behaviors, but not others, are disrupted by elevated CO 2 within a species.

Pharmacological Considerations
Studies using gabazine have provided a useful starting point to understand the mechanisms underlying behavioral alterations at elevated CO 2 in marine invertebrates. However, the pharmacological profile of invertebrate iGABA Rs differs from that of vertebrate GABA A Rs (Rauh et al., 1990), and have not been characterized as extensively as in vertebrates. The majority of invertebrate research has been in non-marine invertebrates, with invertebrate iGABA R pharmacology being best studied in insects due to their potential target for insecticides (Hosie et al., 1995;Bloomquist, 2003). Gabazine inhibits a cloned planthopper iGABA R subunit expressed in a cell line (Narusuye et al., 2007) and two cloned fruit fly iGABA R subunits expressed in Xenopus laevis oocytes (Hosie and Sattelle, 1996). Gabazine also inhibits iGABA Rs in native neurons of locusts (Janssen et al., 2010) and moths (Satoh et al., 2005), but not in cockroaches (Aydar and Beadle, 1999). The few studies in other diverse invertebrates are conflicting; gabazine inhibits iGABA Rs in a freshwater hydrozoan (Concas et al., 1998), only weakly antagonizes GABA responses in a terrestrial nematode (Duittoz and Martin, 1991) and has no effect on crustaceans, including a freshwater crayfish (El Manira and Clarac, 1991;Pearstein et al., 1996) and a marine lobster (Jackel et al., 1994). These non-marine examples are more taxonomically relevant to marine invertebrates than comparisons with evolutionarily divergent marine vertebrate taxa. They indicate the wide variability in invertebrate iGABA R responses to gabazine. As iGABA R pharmacology can differ by subunit composition (Lee and Maguire, 2014;Sieghart, 2015) and there is large variation in subunit genes between invertebrates (Tsang et al., 2006) it will be useful to characterize the pharmacology of gabazine on iGABA Rs in the studied marine invertebrate species.
It is also important to note that antagonists are commonly not completely specific. For example, in insects gabazine partially and bicuculline (a GABA A R antagonist) fully inhibits locust nicotinic acetylcholine receptors (Jackson et al., 2002), and the GABA A R antagonist bicuculline inhibits and the GABA A R agonist muscimol activates a model of insect GABA-gated cation channels (Gisselmann et al., 2004). Less research has studied the off-target effects of GABA drugs in marine invertebrates, though bicuculline and picrotoxin both inhibit acetylcholinegated chloride channels in the sea hare Aplysia californica (Yarowsky and Carpenter, 1978). To ensure the low affinity, alternative effects of drugs do not occur, careful consideration of concentration administered must be made. Using a range of GABA A R drugs with differing side effects will provide further evidence for or against the role of iGABA Rs in behavioral alterations at elevated CO 2 .

Marine Invertebrate Acid-Base Regulatory Mechanisms and the GABA Hypothesis
Acid-base regulatory mechanisms in marine invertebrates indicate that extra-and intra-cellular [HCO 3 − ] and [Cl − ] will alter at elevated CO 2 , providing theoretical support for the GABA hypothesis. Such mechanisms have best been studied in crustaceans [see reviews by Henry and Wheatly (1992) and Wheatly and Henry (1992)]. The primary mechanism to maintain extracellular pH (pH e ) is via ion exchange with the external water environment (Cameron, 1985;Wheatly and Henry, 1992;Pörtner et al., 1998), including HCO 3 − influx in exchange for Cl − efflux (Truchot, 1983;Wheatly and Henry, 1992). Indeed, in all marine invertebrates studied so far, [HCO3 − ] e in the blood/hemolymph increases upon exposure to elevated seawater CO 2 . Exposure to pCO 2 of 15 and 30 mm Hg (∼2,000 and 4,000 µatm pCO 2 ) increased hemolymph [HCO 3 − ] and decreased hemolymph [Cl − ] compared to control conditions in the blue crab Callinectes sapidus (Cameron and Iwama, 1987). At a pCO 2 of 45 mm Hg (∼6,000 µatm pCO 2 ) in the same species, hemolymph [HCO 3 − ] also increased, however, hemolymph [Cl − ] increased (Cameron and Iwama, 1987), and 12,061 µatm pCO 2 increased hemolymph [Cl − ] compared to control (373 µatm pCO 2 ) in the hermit crab Pagurus bernhardus . In palemonid shrimps exposure to 0.3 kPa CO 2 (∼3,000 µatm pCO 2 ) decreased hemolymph [Cl − ] in the high shore Palaemonidae elegans, but increased hemolymph [Cl − ] in the subtidal Palaemonidae serratus (Dissanayake et al., 2010). Thus, changes in both [HCO 3 − ] e and [Cl − ] e at elevated CO 2 support the role of compensatory acid-base regulation as a key part of the GABA hypothesis, although studies at CO 2 levels more relevant to future scenarios, such as ∼1,000 µatm CO 2 , would be valuable.
Intracellular pH (pH i ) is also regulated via ion exchange (Walsh and Milligan, 1989), including Na + dependent Cl − /HCO 3 − exchange , as seen in muscle fibers of the sipunculid worm (Pörtner et al., 2000), crayfish (Galler and Moser, 1986) and barnacle (Boron, 1977;Boron et al., 1981), as well as in crayfish neurons (Moody, 1981) and the squid giant axon (Russell and Boron, 1976;Boron and Russell, 1983). This suggests that [HCO 3 − ] i increases and [Cl − ] i decreases in order to maintain pH i . To date, neural [HCO 3 − ] i has not been measured in a marine invertebrate exposed to elevated CO 2 . However, in the sipunculid worm Sipunculus nudus muscle [HCO 3 − ] i significantly increased over 96 h in 1% CO 2 (∼10,000 µatm pCO 2 ) in air (Pörtner et al., 1998). A net efflux of Cl − is observed from the squid giant axon at a pH i of 6.5 reached by intracellular acid administration (Boron and Russell, 1983), and in crayfish isolated abdominal ganglia the resting [Cl − ] i (35 mM) decreased by 3-5 mM when exposed to Ringer's solution equilibrated with 5% CO 2 (∼10,000 µatm pCO 2 ) (Moody, 1981). Therefore, changes in [HCO 3 − ] i and [Cl − ] i occur in marine invertebrates exposed to extremely high levels of CO 2 . Again, studies using CO 2 levels more relevant to future scenarios will be important to understand the theory underlying the GABA hypothesis.
It is unknown whether the above changes in [Cl − ] i/e and [HCO 3 − ] i/e are sufficient to alter iGABA R functioning. For elevated CO 2 to disrupt GABA functioning, it is not simply altered [HCO 3 − ] and [Cl − ], but a difference in ion gradients across neuronal membranes that will alter ion flow through the iGABA R, i.e., [Cl − ] i and [HCO 3 − ] i within neurons must change by a different amount to [Cl − ] e and [HCO 3 − ] e present in the fluid bathing the neurons (Nilsson and Lefevre, 2016). A useful way to determine whether the changes in [HCO 3 − ] and [Cl − ] can alter ion flow through the iGABA R is by determining the GABA reversal potential (E GABA ) and comparing it to the neuronal resting membrane potential (see Tresguerres and Hamilton (2017) and Heuer et al. (2019) for a detailed explanation). This approach has previously been employed to demonstrate that altered [Cl − ] and [HCO 3 − ] at elevated CO 2 could change GABA A R function in the spiny damselfish Acanthochromis polyacanthus (Heuer et al., 2016).
Calculating E GABA requires knowledge of the HCO 3 − /Cl − permeability ratio of the iGABA R, and [ − ] e and [Cl − ] e at both control and elevated CO 2 conditions. The HCO 3 − /Cl − permeability ratio is estimated to be between 0.2 and 0.6 in crayfish muscle fibers (Kaila and Voipio, 1987;Kaila et al., 1989;Farrant and Kaila, 2007), indicating that the iGABA R is more permeable to Cl − than it is to HCO 3 − . However, this permeability ratio is unknown for other invertebrates. It is important to ensure ion concentration measurements are taken in the correct fluids. In most marine invertebrates no blood brain barrier is present (Cserr and Bundgaard, 1984) and measuring extracellular ion concentration in the hemolymph may be adequate. However, structural organization of nervous tissue may provide some regulation of the neuronal microenvironment (Cserr and Bundgaard, 1984). Cephalopods have a blood brain barrier separating the blood from the brain (Cserr and Bundgaard, 1984) and extracellular measurements should be made in the brain interstitial fluid that bathes the neurons. It is also vital intracellular measurements are done on neuronal cytoplasm, as pH i regulatory mechanisms can differ between different cell types . Studies measuring these parameters and calculating E GABA in a marine invertebrate exposed to elevated CO 2 will be useful to understand if altered [Cl − ] and [HCO 3 − ] could change iGABA R function.
Neural processes other than behavior may also be affected by altered iGABA R functioning. In vertebrates, GABA can act as a trophic factor (a molecule supporting cell survival) through the GABA A R, influencing cell proliferation, migration and differentiation (Owens and Kriegstein, 2002;Sernagor et al., 2010). In a fish, the three-spined stickleback Gasterosteus aculeatus, genes involved in neurogenesis and neuroplasticity were upregulated after exposure to ∼1,000 µatm pCO 2 compared to control (∼330 µatm pCO 2 ) (Lai et al., 2017). Similarly, GABA induces cellular differentiation and proliferation in abalone larvae (Morse et al., 1980). Thus, elevated CO 2 may alter neurogenesis in marine invertebrates.
GABA can also have effects in non-neural tissue, playing an important role in the vertebrate immune system (Barragan et al., 2015;Wu et al., 2017). Invertebrate GABA also appears to play an immunomodulatory role. The iGABA R associated protein is implicated in the immune response of the abalone (Bai et al., 2012), GABA in the immune response of an oyster (Li et al., 2016a) and mussel (Nguyen et al., 2018), and a homolog of the glutamic acid decarboxylase (a rate limiting enzyme in GABA production) in immune regulation of an oyster (Li et al., 2016b). Thus, altered iGABA R functioning at elevated CO 2 may have widespread effects.
Cross-talk between neurotransmitter receptors may also result in widespread effects of altered iGABA R functioning. Different types of neurotransmitters can be co-released from the same nerve terminal, resulting in simultaneous activation of their specific receptors, co-localized at the same post-synaptic site. This simultaneous activation can result in cross-talk between the receptors, modulating signal transmission. For example, negative cross-talk occurs when two different neurotransmitters simultaneously bind to their specific receptors, resulting in a current smaller than the sum of the currents of these two neurotransmitters acting separately (Li et al., 2003). Crosstalk involving GABA A Rs is well documented in mammals. For example, GABA A R activation suppresses the function of a dopamine receptor (de la Mora et al., 1997) and negative cross-talk occurs in both directions between GABA A Rs and glycine Rs (Trombley et al., 1999;Li et al., 2003) and GABA A Rs and the adenosine-triphosphate R P2X (Karanjia et al., 2006;Toulmé et al., 2007). Neurotransmitter crosstalk is yet to be studied in an invertebrate, but co-release of neurotransmitters occurs in the marine invertebrate nervous system, such as proctolin and GABA in crabs (Blitz et al., 1999), and dopamine and GABA in the sea hare Aplysia californica (Díaz-Ríos et al., 2002;Díaz-Ríos and Miller, 2005;Svensson et al., 2014). Furthermore, GABA has been found to postsynaptically increase dopamine currents in the sea hare Aplysia californica (Svensson et al., 2014). If cross-talk is present between invertebrate iGABA Rs and other neurotransmitter Rs, altered functioning of the iGABA R at elevated CO 2 may alter crosstalk mechanisms. Thus, altered iGABA R function at elevated CO 2 would not only alter the GABAergic pathway, but other pathways as well.

ALTERNATIVE MECHANISMS FOR BEHAVIORAL CHANGE AT ELEVATED CO 2
If elevated CO 2 does alter HCO 3 − and Cl − gradients across neuronal membranes, it is likely that functioning of LGICs other than the iGABA/GABA A R, that are also permeable to these ions, will also be disrupted (Figure 2). In vertebrates, altered glycine receptor functioning at elevated CO 2 has been suggested due to its similarity to the GABA A R (Tresguerres and Hamilton, 2017). Many invertebrates lack glycine receptors (Tsang et al., 2006); however, invertebrates possess a larger variety of LGICs than vertebrates (Dent, 2010). Glutamate-gated chloride channels have been found in mollusks (Kehoe and Vulfius, 2000;Kehoe et al., 2009) and crustaceans (Marder and Paupardin-Tritsch, 1978), and are suggested to be the invertebrate equivalent of vertebrate glycine Rs (Vassilatis et al., 1997;Kehoe and Vulfius, 2000). There is also evidence for acetylcholine-gated chloride channels in mollusks (Kehoe, 1972). Moya et al. (2016) found a range of nervous system transcripts differentially expressed at elevated CO 2 in the pteropod Heliconoides inflatus, including genes encoding for the LGICs (and associated proteins) of cholinergic, GABAergic, glutamatergic and glycinergic-like synapses (Moya et al., 2016) (Table 1). Thus, altered [Cl − ] can conceivably disrupt functioning of a range of LGICs (Figure 2). Furthermore, taxa specific differences in the presence of specific LGICs (Dent, 2010) may explain the variability of the effects of elevated CO 2 on marine invertebrate behavior.
Elevated CO 2 may not only disrupt neurotransmission, but also neuronal-glial and glial-glial signaling. Glia are nonneuronal cells present in the vertebrate and invertebrate nervous system (Pentreath, 1989;Laming et al., 2000). Initially thought to be restricted to supporting neurons, the role of glia is now understood to include active participation in nervous system functioning, thus contributing to behavior (Laming et al., 2000;Jackson and Haydon, 2008). Like many other cell types, glial cells regulate pH i via ion exchange, including HCO 3 − /Cl − exchange (Deitmer and Rose, 1996). GABA A Rs are present in vertebrate glial cells (Butt and Jennings, 1994;Fraser et al., 1994). Less research has been carried out on invertebrate glia, with no research on the presence of iGABA Rs on marine invertebrate glial cells. However, leech glial cells reportedly respond to GABA [unpublished work reported in Deitmer and Rose (1996)]. If marine invertebrates are found to express iGABA Rs, elevated CO 2 may also affect information processing through glial cells. Furthermore, fluxes in H + ions have been found to contribute to neuron-glia signaling (Deitmer and Rose, 1996;Laming et al., 2000), which may be disrupted by exposure to elevated CO 2 (and the resultant increase in H + ions).
Changes in [Cl − ] due to acid-base regulatory mechanisms at elevated CO 2 may affect LGIC-mediated neurotransmission through a different mechanism, as well as having alternative effects on the nervous system. Cl − has a role as a signaling effector, with changes in [Cl − ] i affecting a range of processes including gene expression, protein activity and cell proliferation (Valdivieso and Santa-Coloma, 2019). Mammalian work has supported the role of Cl − as a signaling effector in the nervous system, including [Cl − ] i regulation of GABA A R expression (Succol et al., 2012) and growth of neuronal processes (Nakajima and Marunaka, 2016). The role of Cl − as a signaling anion has also been observed in bacterial cells, suggesting a conserved function (Valdivieso and Santa-Coloma, 2019). Thus, invertebrate Cl − is also likely to act as a signaling effector. If altered [Cl − ] i has similar effects in marine invertebrates, elevated CO 2 may impact nervous system functioning not only by altered iGABA R function but also changes in iGABA R expression, as well as altered growth of neuronal projections. Thus, altered [Cl − ] i at elevated CO 2 may have effects additional to altered LGIC function, having widespread consequences.

DIRECTIONS FOR FUTURE RESEARCH
Mechanistic studies have used targeted approaches to assess the GABA hypotheses in marine invertebrates. These approaches will also be useful to assess potential alternative mechanisms by which elevated CO 2 may alter marine invertebrate behavior (Figure 3). The GABA hypothesis has been pharmacologically assessed by administering the GABA A R antagonist gabazine to marine invertebrates Clements et al., 2017). Likewise, the administration of pharmacological agents targeting different LGICs will be useful to assess whether altered functioning of alternative LGICs may underlie behavioral changes at elevated CO 2 conditions. Due to the diversity of invertebrates and the potential for off-target effects, it will be important to use a range of pharmacological agents, particularly those shown to work in the invertebrate taxa being studied. Furthermore, pilot studies to determine the optimal drug concentration to use will be important. For example, Charpentier and Cohen (2016) tested three gabazine concentrations. Measurement of ion concentrations at control and elevated levels of CO 2 in conjunction with behavioral tests has been carried out in crabs Charpentier and Cohen, 2016) and a mollusk (Zlatkin and Heuer, 2019). Future studies in other invertebrate taxa, with measurements made within the correct intra-and extra-cellular fluids, as well as determining the HCO 3 − /Cl − permeability ratio will be important to calculate E GABA at elevated CO 2 to theoretically assess the GABA hypothesis. Real-time PCR, measuring the expression level of a specific gene, has been employed to assess the GABA hypothesis in a crab (Ren et al., 2018) and will also be useful to assess alternative mechanisms, e.g., measuring the expression of genes encoding for alternative LGIC subunits, and genes involved in neuronal growth and proliferation.
These targeted approaches, however, may leave potentially relevant information unexplored. Omic technologies, such as transcriptomics and proteomics, provide a non-targeted approach in which a priori hypotheses are not required (Figure 3). Thus, data from omic approaches could unveil patterns leading to the development of novel hypotheses. For example, transcriptomics and proteomics have already been employed in fish nervous tissue, providing support for the GABA hypothesis as well as new avenues to pursue (Schunter et al., 2016(Schunter et al., , 2018Porteus et al., 2018;Williams et al., 2019).
In marine invertebrates exposed to control and elevated CO 2 levels, transcriptomic studies so far have analyzed the whole animal and have focused on pteropods (Koh et al., 2015;Maas et al., 2015;Moya et al., 2016;Thabet et al., 2017) and sea urchins (Todgham and Hofmann, 2008;Evans et al., 2013;Padilla-Gamiño et al., 2013;Clark et al., 2019). Likewise, proteomic studies have analyzed the whole body of tubeworm larvae (Mukherjee et al., 2013), oyster larvae (Dineshram et al., 2012;Dineshram et al., 2013), barnacle larvae (Wong et al., 2011), sea snail larvae (Di et al., 2019) and clam larvae (Timmins-Schiffman et al., 2019), and a metabolomics study analyzed the whole body of a crab (Trigg et al., 2019) exposed to control or elevated CO 2 . To understand the neurobiological mechanisms underlying behavioral changes, rather than the general molecular response to elevated CO 2 , it will be important to carry out omic techniques on the nervous tissue, as measurements at the whole body level may mask differential expression in the nervous system due to the heterogeneity and complexity of gene/protein expression. This is exemplified by Liu et al. (2019) who found region-specific regulation of neuropeptides in the nervous tissue of crabs Callinectes sapidus exposed to elevated CO 2 .
Omic technologies will provide a powerful, holistic approach to explore neurobiological mechanisms underlying behavioral change, potentially leading to the development of novel hypotheses. However, omic approaches can only determine correlational, and not causative, links between expression and behavior. Gene knockdown, in which the expression of a specific FIGURE 3 | Conceptual diagram illustrating the techniques that will be useful for future research to assess the neurobiological mechanisms underlying behavioral change at elevated CO 2 in marine invertebrates. Targeted approaches will test specific hypotheses and include pharmacological research administering drugs that target a specific receptor, measuring the expression of specific genes, measuring the concentration of HCO 3 − and Cl − ions in the relevant intra-and extra-cellular fluids, and knocking down the expression of specific genes. Omic techniques such as transcriptomics, proteomics and epigenomics will provide a non-targeted approach which does not require a priori hypotheses and will likely lead to the development of new hypotheses. Squid image taken by the author, JT. gene is reduced, will be a promising avenue for future research to determine a causative link between gene expression and behavioral change at elevated CO 2 . Gene knockdown is yet to be used in elevated CO 2 behavioral research, however, gene knockdown of a heat shock protein assessed the stress tolerance of the white leg shrimp Litopenaeus vannamei to high CO 2 (Aishi et al., 2019).

CONCLUSION
There is large variability in the effects of elevated CO 2 on marine invertebrate behavior, which is likely due to the incredible diversity of marine invertebrates. Elevated CO 2 likely alters behavior via a range of mechanisms that disrupt the nervous system pathway producing behavior; from sensory input to behavioral output. These mechanisms are not necessarily mutually exclusive, and interactions between mechanisms may account for the diversity in responses. Many of these mechanisms are based on theory and lack solid experimental evidence. Mechanistic research addressing these gaps will be important, for example linking altered sensation at elevated CO 2 to behavioral change. Mechanistic fish research has focused on altered neurotransmission via disrupted GABA A R functioning at elevated CO 2 . The GABA hypothesis, as well as altered functioning of other LGICs, likely applies to marine invertebrates. Further research into the ionic properties of the iGABA R and other LGICs, including whether they are also permeable to HCO 3 − , and measuring intra-and extra-cellular ion levels in the relevant fluids at near-future CO 2 levels will be beneficial for advancing our understanding of this mechanism. The diversity of LGIC subtypes between invertebrate species, and even between nervous system regions, may explain the variability in behavioral responses. Investigating the presence of LGICs on invertebrate glial cells, other modes of neuronal-glial and glial-glial transmission, and the role of Cl − as a signaling effector in invertebrates will help us understand the wider impact elevated CO 2 may have on nervous system functioning.
The interconnectivity of the nervous system, such as receptor cross-talk, suggests that even disruption of one component or pathway will have widespread effects. This will make understanding the neurobiological mechanisms underlying elevated CO 2 -induced behavioral change extremely complex. Omics approaches will be useful in providing an untargeted, holistic approach to understand the response of the nervous system to elevated CO 2 , provide support or opposition for proposed mechanisms, and likely provide new avenues to explore. Exploring the mechanisms underlying behavioral change at elevated CO 2 will help us to understand the variability in behavioral responses to elevated CO 2 and predict which marine invertebrates are likely to be the most vulnerable to rising CO 2 levels.

AUTHOR CONTRIBUTIONS
All authors contributed to the conception and design of the review. JT wrote the first draft of the manuscript. All authors contributed to manuscript revision, read and approved the submitted version.