Expression of genes involved in brain GABAergic neurotransmission in three-spined stickleback exposed to near-future CO2

We report here the first comprehensive gene expression analysis of GABAA receptor subunits and of genes involved in brain GABAergic transmission in fish exposed to near-future CO2 conditions. Altogether, 56 mRNA transcripts were quantified in brains of three-spined stickleback (Gasterosteus aculeatus) kept in control or elevated pCO2 for 43 days.


Introduction
The ongoing increase of CO 2 levels in the atmosphere and the resultant changes in the ocean chemistry are leading to what is commonly referred to as ocean acidification. In their most recent assessment report, the Intergovernmental Panel on Climate Change (IPCC) predicted an increase in the atmospheric CO 2 concentration from the present level of 400 μatm to 800-1150 μatm within this century (Collins et al., 2013). These changes in the atmosphere can then lead 1 to a decrease in average ocean pH of up to 0.32, with severe consequences for marine ecosystems (Doney et al., 2009;Ciais et al., 2013).
Studies using an antagonist (gabazine) or an agonist (muscimol) of the γ-aminobutyric acid receptor A (GABA A receptor) have indicated that an altered function of this inhibitory neurotransmitter receptor underlies these behavioural abnormalities. In particular, gabazine has been found to restore much of the altered behaviours (Nilsson et al., 2012;Chivers et al., 2014;Chung et al., 2014;Hamilton et al., 2014;Lai et al., 2015). The GABA A receptor is an ion channel with conductance for Cl − and HCO 3 − , and these are the same two ions that are involved in pH regulation in fish exposed to elevated CO 2 . Thus, when fish are exposed to high CO 2 levels, the reduction in blood pH is countered by accumulation of HCO 3 − in blood and tissues (Ishimatsu et al., 2008;Brauner and Baker, 2009), accompanied by a release of H + and Cl − over the gills into the ambient water. This led Nilsson et al. (2012) to suggest that pH-regulatory changes in fish exposed to high CO 2 alter the gradients of Cl − and HCO 3 − over neuronal membranes in a way that renders some GABA A receptors depolarizing (i.e. excitatory) rather than hyperpolarizing (i.e. inhibitory).
The GABA A receptor is the major inhibitory neurotransmitter receptor in the vertebrate brain, and~30% of all synapses respond to GABA (Bloom and Iversen, 1971;Kaila, 1994;Somogyi et al., 1996;Sieghart and Sperk, 2002). It is expressed throughout the central nervous system, and its role has been linked to important processes such as brain development, neural migration and excitability, network interaction in the cerebral cortex, memory, learning, cognition, vigilance and behaviour (Sieghart and Sperk, 2002;Makkar et al., 2010;Luscher et al., 2011).
The GABA A receptor is a ligand-gated ion channel composed by pentameric assemblies of subunits, arranged to form a central selective anion channel (Bormann et al., 1987). To date, a total of 19 genes have been found to encode for GABA A receptor subunits in mammals, namely α1-6, β1-3, γ1-3, δ, π, ε, θ and ρ1-3 (reviewed by Farrant and Kaila, 2007). However, information on GABA A receptor composition in fish is very scarce. An immunochemistry analysis has confirmed a widespread distribution of the receptor in Atlantic salmon (Salmo salar) brain (Anzelius et al., 1995). Ellefsen et al. (2008) surveyed mRNA transcripts of GABA A subunits in the anoxia-tolerant crucian carp (Carassius carassius), quantifying the effect of anoxia on the mRNA expression of subunits α1-6, β2, γ2 and δ1-2. Cocco et al. (2016) recently profiled the expression of GABA A subunits in zebrafish (Danio rerio) brain, showing α1, β2, γ2 and δ to be the most prominently expressed subunits.
The combination of different subunits in the pentameric GABA A receptor can give rise to diverse receptor subtypes, with distinct physiological and pharmacological properties (Herd et al., 2007). Generally, a combination of the two most highly expressed subunits, α and β, is sufficient to form a functional GABA A receptor, while the presence of a third subunit is also often observed (reviewed by Farrant and Kaila, 2007). Indeed, the most predominant GABA A receptor stoichiometry among mammals is a heteromeric receptor composed of two α, two β and one γ subunit, with the most common combination being α1, β2 and γ2 subunits (Fritschy et al., 1992;McKernan and Whiting, 1996;Pirker et al., 2000;Sieghart and Sperk, 2002;Whiting, 2003;Benke et al., 2004). In other GABA A receptors, the γ subunit is replaced by δ, π or ε (forming αβδ, αβπ or αβε), whereas the θ subunit might replace the β subunit (Sieghart and Sperk, 2002;reviewed by Farrant and Kaila, 2007).
Activation of the receptor takes place when two GABA molecules bind to the extracellular domains between the α and β subunit, triggering a rapid conformational change in the transmembrane region that allows movement of Cl − and HCO 3 − through the channel (Bormann et al., 1987). Intracellular and extracellular [Cl − ] and [HCO 3 − ] are important for setting the E GABAA reversal potential. In most mature mammalian neurons, GABA A receptor activation reduces the excitatory neurotransmission through membrane hyperpolarization caused by a net influx of negatively charged Cl − ions into the neuron, with a smaller component of HCO 3 − flowing out (reviewed by Farrant and Nusser, 2005). However, in the fetal mammalian brain, and in some conditions of neuronal overactivity, such as in epilepsy, anion gradients are reversed as a result of increased intracellular [Cl − ] and/or intracellular [HCO 3 − ] linked to a different or altered expression of ion transporters. The Cl − gradient across neuronal membranes has been shown to depend largely on two ion-exchange mechanisms (Delpire, 2000). The K + -Cl − cotransporters (KCC) are responsible for K + -coupled Cl − outward transport in central neurons. In contrast, the Na + -K + -2Cl − cotransporter (NKCC) family is responsible for transporting Cl − into cells through a Na + -K + -coupled Cl − inward transport.  (Delpire, 2000).
The function of the GABAergic transmission is also affected by the timing of GABA release and clearance in the extracellular space. Extracellular GABA in not subject to enzymatic degradation, but its turnover relies on diffusion and uptake by specific GABA transporters, GAT1-3 (reviewed by Scimemi, 2014). GABA is susequently processed in the neurons by GABA aminotransferase (GABAT) and glutamate decarboxylases (GAD1-2, also known as GAD67 and GAD65; Delpire, 2000). The clustering, targeting and degradation of the GABA A receptor in the post-synaptic area is regulated by GABA A receptor-associated proteins (GABARAP and GABARAPL; Nemos et al., 2003). Both proteins belong to a microtubule-associated protein family.
Interestingly, in some fish the behavioural dysfunctions observed in hypercapnia set only in after several days of exposure to high CO 2 and then persist for several days after normal CO 2 levels have been restored . This led Lai et al. (2015) to propose that gene transcription may be involved. This could include the expression of GABA A receptor genes and the genes encoding for proteins responsible for establishing Cl − and HCO 3 − ion gradients over neuronal membranes.
The three-spined stickleback (Gasterosteus aculeatus) should provide a good model for investigating the effects of CO 2 on gene expression because its genome has been sequenced and annotated (Kingsley, 2003). Importantly, sustained high-CO 2 exposure has been shown to alter threespined stickleback behaviour Näslund et al., 2015), and this impairment can be reversed by treatment with the GABA A antagonist gabazine . Thus, as in other fishes, the neural effects of high-CO 2 exposure on three-spined stickleback appear to depend on altered GABA A receptor function.
We hypothesized that the proposed ion disturbances leading to altered GABA A receptor function in brains of hypercapnic fish lead to alterations in the expression of genes related to the function of these systems. Consequently, we have quantified the mRNA transcription levels of 56 genes involved in GABAergic transmission and anion regulation in brains of three-spined stickleback exposed to present and predicted future CO 2 levels. Our analysis included the expression of 28 genes encoding for the GABA A receptor subunits in threespined stickleback, six for GABA transporters (GAT1-3), GABA aminotransferase (GABAT), three for glutamate decarboxylases (GAD1-2), three for GABA A receptor-associated protein and protein-like (GABARAP and GABARAPL), 14 for ion cotransporters (KCCs, NKCCs, ClC2s, AE3 and NDAE) and two for carbonic anhydrases (CAII and CAVII).

Experimental animals
One hundred marine female three-spine sticklebacks weighing 1.24 ± 0.07 g were caught in Fiskebäckskil, Sweden, during July-August 2012 and were randomly distributed into ten 25 litre glass aquaria of 10 individuals each in Sven Lovén Centre for Marine Sciences, Kristineberg, Sweden. The aquaria were constantly supplied with water at 17.6 ± 1.2°C (SD) and salinity 24.2 ± 3.4 PSU (SD). Chemical parameters such as salinity, oxygen saturation, temperature and pCO 2 were measured daily, and alkalinity was measured weekly. Further details are given by Jutfelt et al. (2013), who published behavioural data from the same groups of fish.
The fish were divided into two experimental groups (distributed in duplicate aquaria for each group), where one group was exposed to increased pCO 2 (991.3 ± 56.6 μatm), while the other served as a control and was exposed to present-day CO 2 levels (333.0 ± 30.0 μatm pCO 2 ; Jutfelt et al., 2013). Fish were kept in a 14 h-10 h light-dark cycle and fed ad libitum twice daily with frozen Artemia nauplii. The exposures lasted for 43 days. Upon termination of exposure and behavioural studies, 12 individuals weighing 2.06 ± 0.14 g from the control group and 12 individuals weighing 1.58 ± 0.18 g from the CO 2 group were killed using an overdose of 2-phenoxyethanol in seawater. For the gene expression analysis, the whole brains were rapidly dissected, snap-frozen in liquid nitrogen and stored at −80°C until further use. Prior to downstream experiments, samples were transferred on dry ice to the Department of Biosciences, University of Oslo, Norway.
Animal experiments were carried out in accordance with national regulations and were approved by the ethical committee on animal experiments of Gothenburg, Sweden UV-Vis Spectrophotometer (Thermo Fisher Scientific, Rockland, DE, USA) and a 2100 BioAnalyzer with RNA 6000 Nano Lab Chip Kit (Agilent Technologies, Palo, Alto, CA, USA) were used to assess the quantity and quality of the extracted total RNA. Prior to cDNA synthesis, 1 µg of total RNA was treated with TURBO DNase using TURBO DNAse-free kit (Ambion Applied Biosystems, Foster City, CA, USA) to avoid any remnants of genomic DNA. Subsequently, cDNA was synthesized in duplicate from each total RNA sample using SuperScript III reverse transcriptase (Invitrogen) and oligo(dT) 18 in a total reaction volume of 20 µl. All procedures were carried out in accordance with the manufacturer's protocols.

Real-time RT-PCR primer design
To our knowledge, expression analyses of the GABA A subunits or genes linked with the GABA A activity studied here have previously not been described in stickleback. Therefore, a total of 56 gene-specific real-time rt-PCR (qPCR) primer pairs were designed from stickleback gene sequences retrieved from the Ensembl database (http://www.ensembl.org/index.html; see Table 1 for accession numbers). For each transcript, a minimum of three primer pairs were initially designed for each nucleotide sequence using Primer3 (http://primer3.ut.ee) and synthesized by ThermoScientific (Ulm, Germany). Emphasis was put on designing primers spanning exon-exon junctions to avoid amplification of any remnant genomic DNA. All primers were analysed for crossing point (Cp) values, primer efficiencies (E) and melting peaks, and their products were sequenced by GATC (Cologne, Germany), ensuring amplification of a single amplicon. The primer pairs showing the highest efficiency, lowest crossing point value and a single melting peak curve were selected for qPCR and are listed in Table 1.
For genes with known paralogues or splice variants, efforts were made to design transcript-specific primers when possible, in order to discriminate between closely related transcripts. A comparison aiming at determining identities between genes was carried out using a global alignment (NCBI-Needleman-Wunsch Global Align Nucleotide Sequences; blast.ncbi.nlm. nih.gov), which is a sequence alignment method based on the Needleman-Wunsch algorithm (Needleman and Wunsch, 1969) used to find the best optimal alignment along two sequences (Table S1).
Thirty-five gene sequences for GABA A receptor subunits were retrieved from the Ensemble stickleback database. Diverse paralogue sequences exist in the GABA A subunit families, except for the δ subunit, which has only one known gene variant (see Table 1A). The majority of these sequences showed a distant relationship (identitites ranging from 37 to 78%; Table S1A-E), and qPCR primers were directed at conserved regions. In contrast, paralogues belonging to the subunits β3, γ2 and ρ1 showed a close identity (51-100%), and qPCR primers were directed at poorly conserved regions and analysed as single transcripts (Table S1B, C and E). Among all paralogues, Ensembl presents alternative splice variants for β2 1 , β3 1 and β3 2 (Table 1A). In the γ family, γ2 is the only subunit that splices for three alternative transcripts (Table 1A), and for the ρ subunits, two alternative variants are known to be present for ρ1 2 and ρ3a: ρ1 2i , ρ1 2ii and ρ3a i , ρ3a ii (Table 1A). Altogether, a total of 28 qPCR primers were designed for the gene expression analysis of subunits (some sequences showed too much similarity to allow for the design of specific primers).
For the genes involved in GABA turnover, 11 different gene paralogues were found to encode for GAT, three for GAD, one for GABAT, two for GABARAP and four for GABARAPL. We designed a total of 13 qPCR primers, of which some will work for more than one transcript. Moreover, four different genes encoding for ClC2, seven for KCCs and four for NKCC1 are found in the stickleback genome, whereas AE3, NDAE and CAII have only one variant (Table 1). A total of 15 qPCR primers were designed for the ion cotransporter analysis. Effort was made to design primers able to detect CAVII, but we were unsuccessful in detecting this transcript.

Quantitative PCR
Quantitative PCR was carried out in duplicates using 1:30 diluted cDNA (3 μl), LightCycler 480 SYBR Green I Master Mix (5 μl; Roche Diagnostics, Basel, Switzerland), primers (1 μl; 5 μM) and nuclease-free water (1 μl; Ambion Applied Biosystems). The reaction mix and samples were loaded onto 384 multiwell plates (Roche Diagnostics) using an Agilent Bravo robot (Agilent Technologies, USA) The following qPCR program was used: (i) 95°C for 10 min; (ii) 95°C for 10 s; (iii) 60°C for 10 s; (iv) 72°C for 13 s; and (v) repeat steps (ii) to (iv) 42 times. A melting curve analysis was performed for each amplicon after the qPCR program. Ubiquitin (ubc) and ribosomal protein L13A (rpl13A) were used as reference genes for normalization, as they have previously been demonstrated to be the most stably expressed genes in the three-spined stickleback (Hibbeler et al., 2008) (Table 1). The geometric average of their expression was used to normalize the data sets, because this method has been shown to be a prerequisite for an accurate qPCR expression analysis leading to the possibility of studying small expression differences (Vandesompele et al., 2002;Hellemans et al., 2007).  The Cp values and priming efficiencies for each reaction were calculated using the second derivative maximum method (Roche Lightcycler 480; Rasmussen, 2001) and the LinRegPCR software (Ruijter et al., 2009), respectively. Subsequently, relative mRNA expression levels were calculated using the following formula: Where ga is the geometric average of the two reference genes; tar is the gene of interest, E is priming efficiency and Cp is the crossing point.
Given that duplicate cDNA syntheses were performed, and each of these were analysed in duplicates in the qPCR analyses, four data points were present for each original sample for each primer pair used, and their means were used in the mRNA expression calculations.

Statistical analysis
All statistical analyses were performed using GraphPad Prism (GraphPad Software; version 6.0d; Mac OS X). Normality and homogeneity of variance were assessed using the D'Agostino & Pearson omnibus normality test and F-test. According to function, data were grouped into seven families as follows: (i) GABA A α subunits; (ii) GABA A β subunits; (iii) GABA A γ subunits; (iv) GABA A δ, π and ρ subunits; (v) GAT, GAD and GABAT; (vi) GABARAP and GABARAPL; and (vii) KCC, NKCC, ClC2, AE3, NDAE and CAVII. Two-way analysis of variance (ANOVA) followed by the Sidak post hoc test was used to examine differences in expression between the genes within the families and between the two treatment groups. A value of P < 0.05 was considered significant. All data are presented as means ± SEM, unless otherwise stated. Table 1A, GABA A subunits can be regrouped into six families: α(1-6), β(1-3), γ(1-3), δ, π and ρ(1-3). In contrast to mammals, no genes encoding for ε and θ subunits are present in the stickleback genome. All GABA A paralogues retrived on the Ensembl database were found to be expressed in the three-spined stickleback brain (Fig. 1A-D). Expression within each gene family was analysed using twoway ANOVA, with subunit and CO 2 treatment as the two

8
variables. Not surprisingly, the mRNA transcripts levels differed significantly for the different subunits ( Fig. 1; two-way ANOVA, P < 0.001). The most highly expressed GABA A subunits in the three-spined stickleback belonged to the α, β, γ and δ families (Fig. 1A-D), and within these families the expression was dominated by α1 1 and α1 2 , α6b, β1, γ1 (there was only a single isoform for the δ subunit). Among the ρ subunits, ρ1 1 was the most abundant (Fig. 1D). The π subunits showed the lowest expression levels (Fig. 1D). Exposure of three-spined stickleback to elevated pCO 2 (~990 μatm) resulted in significantly altered expression levels for relatively few of the GABA A subunits investigated. The α family   Figure 1: Messenger RNA expression levels of GABA A receptor subunits. Data were normalized to the geometric average of the reference genes ribosomal protein L13A (rpl13A) and ubiquitin (ubc) and grouped into four families as follows: α subunits (A); β subunits (B); γ subunits (C) and δ, π and ρ subunits (D). Each family was analysed by two-way ANOVA followed by Sidak post hoc test. Open and filled columns represent three-spined sticklebacks exposed to control water (n = 12) and high-CO 2 water (n = 12) for 43 days. Values are shown as means + SEM.
Of the genes involved in GABA turnover included in Ensembl Genome Browser (GAT, GAD, GABAT, GABARAP and GABARAPL), all were found to be expressed in threespined stickleback brain (Figs 2 and 3). Within families, there were significant differences in the mRNA abundance of the paralogue members (two-way ANOVA, P < 0.001). In the control group, the GAT1 paralogues (GAT1 1 and GAT1 2at ) were more abundant than GAT2 1-3 (Fig. 2). In the GAD family, GAD1b displayed higher mRNA expression levels than the GAD1a and GAD2 transcripts (Fig. 2). GABARAP was almost four times more highly expressed than GABARAPL1 and eight times more highly expressed than GABARAPL2 (Fig. 3).  Figure 2: Messenger RNA expression levels of GAT, GABAT and GAD genes. Data were normalized to the geometric average of the reference genes ribosomal protein L13A (rpl13A) and ubiquitin (ubc). Data were analysed by two-way ANOVA followed by Sidak post hoc test. Open and filled columns represent three-spined sticklebacks exposed to control water (n = 12) and high-CO 2 water (n = 12) for 43 days. Values are shown as means + SEM.  ubc). Data were analysed by two-way ANOVA followed by Sidak post hoc test. Open and filled columns represent three-spined sticklebacks exposed to control water (n = 12) and high-CO 2 water (n = 12) for 43 days. Values are shown as means + SEM.

10
In our experiment, none of the transporters or enzymes involved in GABA metabolism displayed significant alterations in expression in response to high-CO 2 treatment (Figs 2 and 3; two-way ANOVA, P > 0.05).
Of the genes involved in ion transport, the expression level of the transcripts differed between the members of the gene families (two-way ANOVA, P < 0.0001; Fig. 4). KCC2a, ClC2 1at and NKCC1 1i were the most highly expressed transcripts among the ion transporters (Fig. 4). Exposure to high CO 2 did not cause significant changes in mRNA expression levels for the ion transporter transcripts in high-CO 2 -treated fish compared with control fish (Fig. 4; two-way ANOVA, P > 0.05).

Discussion
Changes in the function of GABA A receptors caused by altered ion gradients have been suggested as a general mechanism behind the behavioural disturbances seen in CO 2 -exposed fish (Nilsson et al., 2012;Chivers et al., 2014;Chung et al., 2014;Hamilton et al., 2014;Lai et al., 2015). Exposure to high CO 2 levels triggers acid-base adjustments in fish involving altered levels of Cl − and HCO 3 − (Brauner and Baker, 2009), which have been suggested to change neuronal membrane gradients of these ions, switching some GABA A receptors from being inhibitory to excitatory (Nilsson et al., 2012). Based on the scarce data available, calculations of GABA A equilibrium potentials of neurons in fish exposed to near-future pCO 2 are consistent in showing a possibility for a shift in the GABA A receptor equilibrium potential from causing hyperpolarization to depolarization (Heuer and Grosell, 2014;Nilsson and Lefevre, 2016;Heuer et al., 2016;Regan et al., 2016). Here, we hypothesized that an increase in CO 2 levels in the marine environment, triggering acid-base regulatory mechanisms in fish, could lead to changes in the expression of genes involved in regulating GABA A receptor function and neuronal ion distribution. Such molecular changes could be adaptive. However, the persistence of the behavioural disturbances reported in some experiments, and a lack of transgenerational acclimation (Welch et al., 2014), suggest that possible molecular responses are insufficient, or even maladaptive.
The present study is the first comprehensive expression analysis focused on genes involved in GABAergic transmission in fish exposed to elevated CO 2 . The fish in this study were the same individuals as those previously examined behaviourally , where the behavioural alterations, including reduced exploratory behaviour and lateralization, were characterized and found to persist for the whole experimental period. Of the 28 GABA A receptor subunits examined herein, there was a significant effect of the high-CO 2 treatment on the mRNA expression level for the α family subunits, all showing a tendency to be more highly expressed in the CO 2 group. This could suggest some subunit rearrangement of GABA A receptors in this group, assuming that gene expression is reflected in  Figure 4: Messenger RNA expression levels of KCCs, NKCC1, ClC2, NDAE, AE3 ion cotransporters and CAII enzyme. Data were normalized to the geometric average of the reference genes ribosomal protein L13A (rpl13A) and ubiquitin (ubc). Data were analysed by two-way ANOVA followed by Sidak post hoc test. Open and filled columns represent three-spined sticklebacks exposed to control water (n = 12) and high-CO 2 water (n = 12) for 43 days. Values are shown as means + SEM.  Farrant and Kaila, 2007). The α subunits play important roles on desensitization and deactivation of the receptor, because the GABA binding is presumed to take place at the α-β interfaces (Böhme et al., 2004). In the present experiment, changes at the mRNA expression level of the α subunit family after exposure to high CO 2 might indicate possible compensatory mechanisms used by three-spined stickleback to restore proper GABA A receptor function, but further investigations are required. However, if adaptive, the changes are apparently not sufficient, because the behavioural alterations detected in the same individuals by Jutfelt et al. (2013) remained, and if anything increased, during the 43 day exposure period. Also, we cannot exclude the possibility that the changes detected in this study are maladaptive rather than adaptive and contribute to the behavioural alterations.
As mentioned, the three dominant subunits that make up the mammalian receptor are α, β and γ (reviewed by Farrant and Kaila, 2007), and our data show that these subunit families are also highly expressed in three-spined stickleback, but it is striking that the δ subunit is also expressed at a level similar to the most highly expressed α subunits (Fig. 1). Although the mammalian receptors are dominated by subunits α1 β2 and γ2 (Pirker et al., 2000), the most predominantly expressed subunits in three-spined stickleback were found to be α1, α6b, β1, γ1 and δ. Receptors that comprise γ2 in association with α1, α2 or α3 subunits are normally localized to post-synaptic membranes, where they mediate a phasic inhibition involving a rapid and brief inhibitory postsynaptic potential in response to GABA in the synaptic cleft (reviewed by Farrant and Nusser, 2005). In contrast, GABA leaking out of the synapse can activate extrasynaptic GABA A receptors made up of δ, α4 or α6 subunits, and these are mainly responsible for slower, but longer-lasting inhibitory post-synaptic potentials, causing tonic inhibition (reviewed by Farrant and Nusser, 2005). Consequently, the high expression of the extrasynaptic δ and the α6b subunits might indicate a more important role of extrasynaptic GABA A receptors in three-spined stickleback compared with mammals. Interestingly, previous studies on crucian carp (Carassius carassius) and zebrafish (Danio rerio) brain found a dominating expression of the δ subunits (Ellefsen et al., 2008;Cocco et al., 2016). In light of this high expression of δ subunits, it is tempting to suggest that extrasynaptic GABA A receptors causing tonic inhibition play more important roles in fish than in mammals. In contrast to both crucian carp and zebrafish, three-spined stickleback express alternative splice variants for β2, β3 1 , β3 2 , ρ1 2 and ρ3a. The only common subunit that exhibits alternative splicing in all three species is γ2 (Ellefsen et al., 2008;Cocco et al., 2016). The presence of numerous splice transcripts in three-spined stickleback brain could mean that GABA A receptor isoforms are particularly diverse in this species, and fishes differ in the degree to which alternative splicing is used to modulate GABA A receptor function (see also Cresko et al., 2003).
Perhaps surprisingly, exposure to high CO 2 did not result in significant changes in the mRNA expression levels for ion cotransporters. Heuer et al. (2016) showed that the increase in plasma partial pressure of CO 2 (in millimetres of mercury) in spiny damselfish (Acanthochormis polyacanthus) kept in high-CO 2 conditions was accompanied by increases in intracellular and extracellular HCO 3 − concentrations, with an assumed decrease in intracellular Cl − (Heuer et al., 2016). Based on their measurements, they calculated a positive deviation in the E GABAA resting potential in fish exposed to 1900 µatm CO 2 , consistent with a shift in the GABA action towards depolarization (excitation) rather than hyperpolarization (inhibition). Such alterations could be compensated for by changes in the expression of the Cl − transporters NKCC1 and KCC2 in fish exposed to high CO 2 . As mentioned, these transporters are known to play important roles in setting the reversal potential for Cl − (E Cl ) in the mammalian central nervous system, leading to a shift of the GABA A receptor function from excitatory in immature neurons to inhibitory in mature neurons (Delpire, 2000). In any case, this does not appear to happen in CO 2 -exposed stickleback in the present conditions, as we found no significant changes in the expression of NKCC1 and KCC2, or in other Cl − and HCO 3 − transporters.
Our gene expression results are in agreement with the recent findings of Schunter et al. (2016) on juvenile spiny damselfish (Acanthochromis polyacanthus). In a transcriptome and proteome analysis, they investigated the molecular responses of offspring of CO 2 -tolerant and CO 2 -sensitive parents reared in control or high-CO 2 conditions. The main molecular differences between the two groups were found among genes involved in circadian rhythm control, such as bmal1, clock, per1 and nr1d1 (Schunter et al., 2016). In contrast, the GABA A receptor genes were expressed at similar levels across treatments. The only possible change seen in the GABAergic system was at the protein level of an enzyme that may participate in GABA synthesis, aldehyde dehydrogenase 9 member 1 (Al9A1), which were more highly expressed in the offspring of the CO 2 -tolerant parents (Schunter et al., 2016).

Conclusions and perspectives
In general, the present findings show that exposure of threespined stickleback to elevated CO 2 resulted in only few and minor changes in the expression of genes involved in GABAergic neurotransmission in the brain. If these few adjustments reflect compensatory mechanisms they are apparently not sufficient, because the behavioural dysfunctions remained during the course of the 43 day high-CO 2 exposure . Thus, the present results, together with results reporting that aberrant behaviours displayed by fish exposed to elevated pCO 2 are persistent and not reduced even by transgenerational acclimation (Welch et al., 2014), lead to the worrying conclusion that fish might be incapable of adaptive responses to these new conditions. Given that globally sustained pCO 2 levels >500 µatm have probably not occurred on earth during the last 30 million years (Beerling and Royer, 2011), we may have to face the conclusion that many present-day fishes do not possess the genes and mechanisms necessary to cope with the projected near-future elevation of CO 2 levels.

Supplementary material
Supplementary material is available at Conservation Physiology online.