Seasonal changes in membrane structure and excitability in retinal neurons of goldfish (Carassius auratus) under constant environmental conditions

Many organisms have evolved adaptations that allow them to survive extreme changes in their environment. In the central nervous system (CNS) of anoxia-tolerant vertebrates, the matching of ATP supply to demand and the maintenance of ionic gradients across the plasma membrane occur even when oxygen is limiting (Bickler and Buck, 2007). Some species of fresh-water turtle (Chrysemys picta and Trachemys scripta), frogs (Rana spp.), the goldfish (Carassius auratus) and crucian carp (Carassius carassius) are remarkably tolerant to long periods of anoxia, especially at low temperatures (Herbert and Jackson, 1985; Jackson et al., 1984; van den Thillart et al., 1980; Lutz and Nilsson, 2004). Cellular homeostasis under such conditions may be accomplished by large-scale reductions in metabolic rate (Hochachka et al., 1996), increased tissue storage of fermentable substrate (Jackson and Ultsch, 2010) or ion channel arrest (Pamenter et al., 2008; Zivkovic and Buck, 2010; Rodgers-Garlick et al., 2013). These adaptations help to prevent the onset of anoxic injury to cells and tissues.

Long-term changes in temperature can induce modification in the phospholipid composition of the plasma membrane through homeoviscous adaptation (HVA), where a relative increase in saturated fatty acid composition is associated with an increase in environmental temperature (Hazel, 1995; Ernst et al., 2016). HVA was first described in bacteria (Sinensky, 1974) and later in cells derived from animals, including the goldfish brain (Cossins et al., 1981; Hazel and Williams, 1990). It is particularly important in poikilothermic organisms, as it leads to the synthesis of phospholipids to maintain constant viscosity or fluidity within the membrane as temperatures change. Phosphatidylethanolamines (PEs) are a structural component of membrane bilayers in all eukaryotes and comprise approximately 45% of total membrane phospholipids in the brain (Vance and Tasseva, 2013). PEs are particularly important in regulating membrane fluidity (Hazel, 1988, 1995; Pruitt, 1988). Importantly, changes in membrane structure may have profound effects upon integral membrane protein activity. In vertebrates, the degree of polyunsaturation in membrane phospholipids is positively correlated with Na⁺/K⁺-ATPase activity and maintenance of ionic gradients (Cossins et al., 1981; Hulbert and Else, 1999, 2005). Moreover, changes in lipid bilayer composition, thickness, viscosity and curvature can affect the activity of other membrane proteins, including ion channels and G-protein-coupled receptors (Lee, 2004; Andersen and Koeppe, 2007; Rusinova et al., 2014; Fried et al., 2021).

The above adaptive chemical and physiological changes typically occur as a response to environmental cues, such as changes in temperature or oxygen availability that may occur during acclimatization or seasonal transitions. Interestingly, no studies have investigated whether seasonal changes at the level of the plasma membrane might occur independently of environmental cues, and what impact such changes may have upon the activity of central neurons. Based on our preliminary observations that the electrical properties of retinal neurons isolated from goldfish kept in our animal facility differed throughout the calendar year, we designed a study carried out over multiple years to test the hypothesis that neurons isolated from animals held under constant environmental conditions during the summer versus winter months would display membranes with different biochemical and biophysical properties. We isolated horizontal cells from the goldfish retina, a part of the CNS, and used mass spectrometry, atomic force microscopy, electrophysiology and microspectrofluorometry to probe the structural and functional characteristics in neuronal membranes. We discovered significant changes in membrane phospholipid composition, stiffness of the membrane–cytoskeletal complex, electrical excitability and action potential profile. These results indicate significant seasonal modifications in neuronal membrane structure and function in the absence of seasonal cues in an animal that has adapted to survive extreme environmental change.

Adult common goldfish, Carassius auratus (Linnaeus 1758), of mixed age and sex weighing 10–75 g were used in the present study. Animals were obtained from commercial suppliers (Mirdo Importations Canada, Montréal, QC, Canada; or Aquality Tropical Fish Supply Inc., Mississauga, ON, Canada). Goldfish were originally sourced from sites in the southern USA, where they were bred and raised in outdoor ponds on a natural photoperiod. Upon arrival, animals were quarantined for 6 weeks and housed for up to 18 months in the aquatic facilities of the Laboratory for the Physiology and Genetics of Aquatic Organisms at the University of Ottawa, where they were maintained in 170 l tanks fitted with a flow-through system of fresh, aerated and dechloraminated water (Mar Cor Purification filter, Cantel Medical Corp., Little Falls, NJ, USA) at a constant temperature of 18°C. For electrophysiological experiments, goldfish were obtained from the Aquatron Laboratory (Dalhousie University, Halifax, NS, Canada) and placed in a 400 l tank at 15°C. In both facilities, photoperiod was kept at a constant cycle of 12 h light:12 h dark and all environmental conditions were kept constant year-round. Goldfish were generally taken from our facilities for cell or tissue isolation during the summer (June to September) or winter (January to March), as described in the sections below. Before dissection and cell isolation, goldfish were dark adapted for approximately 1 h to facilitate removal of the retina, euthanized by rapid decapitation and pithed. Procedures for animal care and use were approved by the University of Ottawa Animal Care and Veterinary Services, and by the University Committee for Laboratory Animals at Dalhousie University. All procedures were implemented in accordance with regulations of the Canadian Council on Animal Care.

Horizontal cells were acutely isolated from goldfish retina following the procedures of Jonz and Barnes (2007). The retina is a convenient source of central neurons (Dowling, 1987), and horizontal cells are neurons whose membrane and physiological properties are well characterized (e.g. Tachibana, 1981; Lasater et al., 1984; Cunningham and Neal, 1985; Malchow et al., 1990; Akopian et al., 1991; DeVries and Schwartz, 1992; Jonz and Barnes, 2007; Sun et al., 2017; Country et al., 2019). Unless otherwise stated, all reagents and chemicals were sourced from Sigma-Aldrich (Oakville, ON, Canada). Eyes were removed and placed in cold Ca²⁺-free Ringer's solution, containing (in mmol l⁻¹): 120 NaCl, 2.6 KCl, 1 NaHCO₃, 0.5 NaH₂PO₄, 1 sodium pyruvate, 4 Hepes and 16 glucose at pH 7.8. Whole retinas were placed in hyaluronidase (100 U ml⁻¹, cat. no. H-3506) in L-15 solution for 20 min at room temperature. L-15 solution was composed of 70% L-15 (Leibovitz's medium) and 30% Ca²⁺-free Ringer's solution. Retinas were washed 3 times for 3 min in L-15 solution and then placed in L-15 solution containing 7 U ml⁻¹ papain (cat. no. 3126, Worthington Biochemical Corporation, Lakewood, WI, USA) for 40 min. Papain was activated by 2.5 mmol l⁻¹l-cysteine. Retinas were rinsed 3 times in L-15 solution. Small (∼4 mm²) sections of retina were removed and mechanically dissociated by gentle trituration in L-15 solution. The resulting cell suspension was plated and allowed to settle for up to 20 min. Horizontal cells were used within 6 h of isolation.

Lipidomics was performed at the Carleton Mass Spectrometry Centre, Carleton University, Ottawa, Canada. For sample preparation, whole retinas were removed from three goldfish in July 2017 and three goldfish in January 2017, placed individually in PBS in centrifuge tubes and frozen immediately in liquid N₂. For each season, three retinas were then combined in a 15 ml centrifuge tube (Corning) and 1 ml of 2% acetic acid in methanol (AcMeOH) was added. Tissue was sonicated on ice using a 130 W ultrasonic processor for 3 min with the pulse on for 20 s and pulse off for 10 s using 50% amplitude.

Lipid extraction was accomplished using a method modified from Bligh and Dyer (1959). Briefly, each sonicated sample was transferred to a 10 ml glass Kimble tube containing 3.2 ml of 0.22 µm-filtered 0.1 mol l⁻¹ sodium acetate via triplicate rinses with 1 ml of 2% AcMeOH; 60 μl of a 1 mmol l⁻¹ C13:0 lysophosphatidylcholine solution in ethanol was added as an internal standard. Then, 3.8 ml of chloroform was added and the tubes were vortexed for 30 s followed by 2 min centrifugation at 2000 rpm and 4°C. The bottom phase was collected with a Pasteur pipette and transferred to a new 10 ml glass Kimble tube. Extractions on the aqueous phase were repeated 3 more times in the same manner using 2 ml of chloroform and combined. The combined chloroform extract was evaporated under a stream of N₂ gas until a dark yellow oily residue remained. The residue was then resuspended in 600 μl of 10 mmol l⁻¹ ammonium acetate in absolute ethanol. The samples were then transferred to an amber glass tube, flushed with N₂ gas and stored in a −20°C freezer until analysis within a week. Samples requiring longer storage times were stored at −80°C.

Glycerophospholipid samples were separated via high performance liquid chromatography (HPLC) and analysed by mass spectrometry (MS) and tandem mass spectrometry (MS²). C₄ reversed phase microflow chromatography columns were constructed in-house as previously described (Canez et al., 2016). Lipids were separated using a ternary gradient programme consisting of 10 mmol l⁻¹ ammonium acetate in 30% methanol/70% MilliQ water (solvent A), 10 mmol l⁻¹ ammonium acetate in 75% isopropanol (IPA)/25% methanol (solvent B), and methyl tert-butyl ether (MTBE) (solvent C). The gradient began at 100% A, increased linearly to 100% B by 41.4 min, then to 100% C by 48 min before returning to 100% A between the 59 and 60 min time points. An average flow rate of 135 µl min⁻¹ was used and each analytical run was followed by a full blank run to further clean the system and ensure no lipid carryover occurred.

For quantitative analysis, 12.5 μl lipid extract was diluted with 37.5 μl of 10 mmol l⁻¹ ammonium acetate, separated using a Dionex Ultimate 3000 HPLC (Thermo Fisher Scientific, Waltham, MA, USA) in the manner described above and introduced into a QTRAP 4000 hybrid triple quadrupole linear ion trap mass spectrometer (AB Sciex, Framingham, MA, USA) using electrospray ionization. Triplicate analyses for both the winter and summer samples were performed in positive ion mode according to the parameters outlined in Table S1.

In order to increase the granularity of our dataset, vinyl-ether-containing lipids (plasmalogens) were distinguished from alkyl-ether-containing lipids via reaction with formic acid, as described by Fhaner et al. (2012). Formic acid-treated lipid samples were redissolved in 10 mmol l⁻¹ ammonium acetate in ethanol for MS analysis, as described above. To reveal greater qualitative detail, individual lipids with statistically significant concentration differences between summer and winter were re-analysed at high resolution using a 1290 HPLC coupled to a 6550 Q-TOF mass spectrometer (Agilent, Santa Clara, CA, USA). The HPLC solvents and conditions were the same as described above and the MS was run in positive ion mode for high mass accuracy determination, and in negative ion mode using MS² with a collision energy of 35 eV for fatty acyl group identification.

For statistical analyses, Analyst 1.5.1 (AB Sciex) was used to generate extracted ion chromatograms, which were inputted into MultiQuant 2.1.1 (AB Sciex) to calculate peak areas. To standardize data, the area sum for every extracted ion was calculated for winter and summer samples and normalized to each other. The individual peak areas were then standardized to the area of the internal standard. Technical replicates enabled the average area with standard deviation to be calculated for each ion. To identify significant alterations in abundance between the winter and summer samples, fold-changes were calculated and a Student's t-test was performed (α=0.05). To control the false discovery rate, the Benjamini–Hochberg method was applied with α=0.1533 (Benjamini and Hochberg, 1995). P-values from the Student's t-test and Benjamini–Hochberg method were compared. High resolution MS data were analysed using MassHunter Qualitative Analysis B.07.00 (Agilent). MS files were converted to MGF format using Proteowizard's MSConvert tool (Chambers et al., 2012) using the Threshold Peak Filter to select the 100 most intense peaks by counts; mzML format conversion without filters was also performed. The resulting MGF file was searched against the LipidBlast database (Kind et al., 2013) using NIST MSPepSearch software with the following settings: Library LipidBlast-neg, MinMF 100, HITS 3, MzLimits 1-2000, MinInt 1, OnlyFound, OutPrecursorMZ, OutDeltaPrecursorMZ, OutSpecNum, HiPri. Hits were then filtered by m/z to each of the five targeted lipid masses and compared with their extracted ion chromatograms using MZmine 2 (Pluskal et al., 2010). Annotating the targeted masses was performed by selecting database matches to PE species with retention times corresponding to the elution profile observed in the chromatograms of target masses. Annotations were confirmed using the NIST MS Search 2.0 program for head-to-tail comparisons to the LipidBlast database.

Live and fixed cell images were acquired with a Nikon (Tokyo, Japan) TiE A1-R high-speed resonant laser scanning confocal microscope using a 60×/NA1.2 objective lens. For cell fixation, 3.5% w/v paraformaldehyde was used in 2% w/v sucrose in Dulbecco's phosphate-buffered solution (DPBS, 311-425-CL, Wisent, St-Bruno, Québec, QC, Canada) by incubating cells for ∼15 min at room temperature, followed by a wash with DPBS 3 times. Cells were incubated with 0.5% v/v Triton X-100 in permeabilization buffer (20 mmol l⁻¹ Hepes, 300 mmol l⁻¹ sucrose, 50 mmol l⁻¹ NaCl, 3 mmol l⁻¹ MgCl₂, diluted in distilled water) for 3 min. Cells were blocked with wash buffer (5% fetal bovine serum in DPBS) for a minimum of 1 h prior to incubation with antibodies. Monoclonal anti-α-tubulin produced in mouse (cat. no. T6074, Sigma-Aldrich; Antibody Registry ID AB_477,582) labelled microtubules and was used at a concentration of 1:200 and prepared in wash buffer. Following 3 wash steps with 5 min each in wash buffer, cells were stained with Alexa Fluor 488 rabbit anti-mouse IgG secondary antibody (cat. no. A11059, Invitrogen, Mississauga, ON, Canada) at 1:200 dilution. Actin staining was simultaneously performed by incubation with Alexa Fluor 546 phalloidin (cat. no. A22283, Invitrogen) at 1:100 dilution. DAPI (4′,6-diamidino-2-phenylindole, D1306, Invitrogen) staining was performed by incubation at 1:500 for 10 min in DPBS, following further rinse steps. All fixed cells were first seeded onto 35 mm dishes coated with Cell-Tak (Corning, Tewksbury, MA, USA). Stained images were captured as z-stacks and maximum intensity projections were generated using ImageJ (Schneider et al., 2012).

Stiffness (or elasticity) was measured in isolated horizontal cells during the summer (June to September) and winter (January) months over the period of two consecutive years (2011−2013). Cell suspensions were allowed to adhere to 35 mm glass-bottom dishes (MatTek, Ashland, MA, USA) for at least 10 min. Live cells were first imaged by confocal microscopy, followed by nano-indentation using a NanoWizard II (JPK Instruments, Berlin, Germany) atomic force microscope (AFM). Cantilevers (MSCT-AUHW, Veeco, Plainview, NY, USA) with a tip radius of 10 nm were calibrated as previously described (Haase and Pelling, 2013). Localized nano-indentation was performed by recording a minimum of 10 force curves for each cell above the most central nuclear region (approach rate 2 Hz, 10 µm s⁻¹, set-point of 1 nN). Force–distance curves were fitted to the first 200 nm of deformation using the modified Hertz model for a conical tip using PUNIAS software (Carl et al., 2001) to provide a measure of Young's modulus (Pa). At 200 nm, the mechanics of the plasma membrane will be linked with the underlying cytoskeletal cortex, and so measurement of Young's modulus will include local elastic responses of both membrane and cytoskeleton (Haase and Pelling, 2015). In some experiments, cells were pre-treated with Hoechst 33342 (Invitrogen) in order to label DNA so that nano-indentation measurements could be compared between measurements performed on-nucleus (N=11) and off-nucleus (N=11) for the same cell. Mean values for Young's modulus were used for each cell (N>10 force curves each) and overall statistics were calculated using Student's t-test with significance indicated by P<0.05.

Whole-cell, voltage-clamp recordings were obtained from horizontal cells during the summer (June to September) and winter (January to March) months of 2005 and 2006 at Dalhousie University in Halifax, Canada. Electrodes were made from capillary glass (cat. no. 2502, Chase Scientific Glass, Inc., Rockwood, TN, USA) and pulled on a vertical puller (Model 730, David Kopf Instruments, Tujunga, CA, USA). Electrodes had a tip resistance of 5–8 MΩ when filled with intracellular recording solution containing (in mmol l⁻¹): 120 KCl, 10 NaCl, 0.5 CaCl₂, 2 Mg-ATP, 5 EGTA, 10 Hepes, pH adjusted to 7.4 with KOH. Extracellular recording solution contained (in mmol l⁻¹): 120 NaCl, 5 KCl, 2.5 CaCl₂, 2 MgCl₂, 10 Hepes, 10 glucose (pH adjusted to 7.8 with NaOH). With these solutions, the calculated equilibrium potential for K⁺ was −80.5 mV. Liquid junction potentials (V_L) of 4.2 mV were calculated using pCLAMP software (Axon Instruments, Sunnyvale, CA, USA) and subtracted from pipette potentials (V_p) to determine actual membrane potential (V_m), according to the equation: V_m=V_p−V_L. Current–voltage data were corrected for V_L.

Voltage-clamp protocols were performed using an Axopatch-1D amplifier (Axon Instruments) and pCLAMP software. Recorded signals were converted using a DigiData 1200 interface (Axon Instruments). Cells were plated in 35 mm plastic Petri dishes (Nunclon, Nunc A/S, Roskilde, Denmark). From a holding potential of −60 mV, currents were evoked by changing V_m to a series of test potentials by a voltage-ramp protocol, in which the membrane potential was continuously changed (0.18–0.22 mV ms⁻¹) over a period of 1 s. Signals were filtered at 2 kHz and digitized at a rate of 10 kHz. Membrane capacitance (C_m) was measured using pCLAMP.

Reversal potential (V_rev) was measured directly from current–voltage (I–V) curves of winter neurons as an estimate of their resting membrane potential (see Discussion). In summer neurons, however, it was not possible to obtain accurate values of V_rev because of the flat slope conductance, i.e. high input resistance (R_in), within the physiological range in these cells (see below). To compare membrane activity of winter versus summer neurons, we therefore measured ion channel current in winter neurons at V_rev (which was zero) and compared this with current measured in summer at the same membrane potential. In some experiments, the inverse slope of the I–V relationship was used to calculate R_in from each cell by measuring the change in current from a stable region (of up to 40 mV) within the voltage range of −80 mV to −15 mV. R_in was then calculated using Ohm's law: R_in=V/I. All data were analysed using pCLAMP software and figures were arranged using Prism 5.0 (GraphPad Software Inc., La Jolla, CA, USA). Student's t-tests were performed using Prism 5.0.

Data were obtained from horizontal cells during the summer (August and September) and winter (January and February) months between 2016 and 2020. Relative changes in free intracellular Ca²⁺ concentration ([Ca²⁺]_i) were measured by microspectrofluorometric imaging following Country et al. (2019). Isolated cells were plated in 35 mm dishes (Corning) fitted with perfusion chambers (Warner Instruments Inc., Hamden, CT, USA, cat. no. RC-33DL). Dishes were pre-coated with 0.01% poly-l-lysine (cat. no. A-005-C). Cells were protected from light and incubated in extracellular solution containing (in mmol l⁻¹): 120 NaCl, 5 KCl, 2.5 CaCl₂, 2 MgCl₂, 10 Hepes and 10 glucose, with pH adjusted to 7.8 with NaOH. Added to this solution was 5 µmol l⁻¹ membrane-permeant Fura-2 (Fura-2-LeakRes AM; Teflabs, Austin, TX, USA) and 0.1% (v/v) of a 10% (w/v) Pluronic F-127 solution for 30 min at room temperature. Cells were washed 3 times to remove remaining esterified products. The chamber was continuously perfused at ∼1 ml min⁻¹ by gravity-fed recording solutions at room temperature. Recording solution was removed from the chamber at the same rate using a variable-flow pump (Thermo Fisher Scientific).

Fluorescence imaging was performed with a Lambda DG-5 wavelength changer (Sutter Instruments, Novato, CA, USA) and a Chroma 79001 filter set (340 nm and 380 nm band-pass filters for excitation, 510 nm bandpass for emission; Chroma Technology, Bellows Falls, VT, USA). Excitation and emission light was passed through a 40× water-immersion objective lens (MRF07420 CFI Fluor, Nikon) optimized for UV transmission on an upright microscope (FN-1, Nikon). Images were collected with a CCD camera (QImaging) by focusing on a region of interest surrounding a soma. Excitation wavelength was changed between 340 nm and 380 nm and emission intensity was recorded for both excitation wavelengths every 2 s with NIS Elements software (Nikon). Imaging data were logged in Excel (Microsoft Corp., Redmond, WA, USA).

Horizontal cells isolated from goldfish retina were routinely identified by their large, stellate morphology and thick dendrites. We identified all subtypes of horizontal cells (Fig. 1A–D) based on morphological descriptions from previous studies (Dowling et al., 1985; Tachibana, 1981; Country et al., 2021) and observed no apparent differences in morphology between summer and winter. Plotting the distribution of the ratio of dendritic field area to soma area provided an estimate of the subtypes present in the dissociation (Country et al., 2021). Calculation and plotting of the r_d/s values of 80 neurons showed that the distribution of cells isolated during the summer and winter was similar, with median r_d/s values of 7.2 and 6.2, respectively (Fig. 1E). Analysis with the Mann–Whitney U-test indicated that r_d/s distributions were not significantly different from each other (P=0.20). Moreover, during both seasons, the dominant subtypes observed were H1 and H2 with r_d/s values of less than 11, while H3 and H4 subtypes, with higher r_d/s values, were fewer in number (Fig. 1E). Subsequent experiments were therefore primarily performed on H1 and H2 subtypes.

Mean membrane capacitance (C_m) of neurons was measured from electrophysiological experiments to estimate cell size (Fig. 1F). Biological membranes have a specific capacitance of 0.01 pF µm⁻² (Hille, 2001). C_m was therefore used to estimate total membrane surface area. C_m was 26.85±1.6 pF in summer (N=46) and 26.90±1.1 pF in winter (N=81), indicating that neuronal membrane surface area did not differ between seasons (P=0.49, Student's t-test).

Analysis of low-resolution MS data yielded a total of 153 lipid species in goldfish retina that were above the limit of quantification, calculated using established protocols (Canez et al., 2016). Of these, 90 were identified as phosphatidylcholine (PC), 37 as PE, 19 as sphingomyelin (SM) and 7 as phosphatidylserine (PS) (Fig. 2A). Measurement of standardized area values from MS data and analysis using Student's t-test with correction for false discovery rate using the Benjamini–Hochberg method revealed that five PE species had significantly higher concentrations in winter (P<0.05; Fig. 2B). These PE species are listed in Table 1 along with their proposed identities based on their nominal masses using IUPAC-IUBMB lipid nomenclature (total number of acyl chain carbons:total number of double bonds). PEs were unsaturated and detected at concentrations 1.8–2.7 times higher in the winter compared with the summer, as illustrated in Fig. 2B.

In order to further characterize the lipids that increased in abundance during winter, high-resolution MS was conducted on these five PE species in positive ion mode to obtain accurate mass, and in negative ion mode to elucidate the specific fatty acyl composition of each via MS². It was immediately discovered, based on accurate mass measurements, that one of the PE masses contained co-eluting diacyl- and ether-linked species that were indistinguishable in the quantitative results. In order to further characterize the ether-linked lipid, formic acid hydrolysis was performed (Fhaner et al., 2012) to determine whether it was a plasmanyl (vinyl-ether-containing) species or plasmenyl (alkyl-ether-containing) species. It was noted that both the plasmanyl and plasmenyl species were present at that mass and elution time at a ratio of 3:1, respectively. Table 2 shows the accurate masses of the PE species that increased in the winter months and reveals the relative percentages of the plasmanyl and plasmenyl lipids.

Negative ion MS² was conducted on all PE masses in Table 3 in order to obtain their identities. Lipid MS² spectra were searched against the LipidBlast database using NIST MSPepSearch software and annotations were confirmed using the NIST MS Search 2.0 program via head-to-tail comparisons with the LipidBlast database. In negative ion mode, fatty acids are lost from lipid ions as both charged and neutral fragments, revealing the sn structural chemistry. The fatty acyl composition was deduced for the PE species by computationally analysing the MS² spectra and all PE species were identified in this manner with the exception of m/z 702.5069 because of interfering peaks in its MS² spectrum. Lipids were identified based on the top scoring hit that was obtained and verified through manual inspection, as illustrated and described in Fig. 3. Fig. 3 includes the assumed positions of the double bonds and sn regiochemistry based on lipids reported in the NIST database. Unfortunately, determining sn regiochemistry or the position of the double bonds on the alkyl chains is beyond the capabilities of our analytical methodologies and the degree to which we are able to confidently identify the PE species is accurately reported in Table 3.

We used AFM to investigate whether membrane–cortical stiffness (or elasticity) differed between cells isolated during the summer versus winter months. Labelling of the cytoskeleton and nucleus was first performed to assess the morphology and relative location of these structures in neurons for AFM experiments. Microtubules and actin filaments were labelled with anti-α-tubulin and phalloidin, respectively, whereas the nucleus was labelled with DAPI (Fig. 4A). Using confocal microscopy, we found that neurons were often binucleated, with nuclei off-centre within the soma. Neurons possessed an extensive cytoskeletal system with a higher concentration of α-tubulin localized to the soma, and phalloidin labelling localized primarily to the dendrites.

AFM was employed to measure changes in apparent stiffness of neurons across sequential summer and winter seasons. Isolated neurons were allowed to settle onto glass-bottomed dishes, then imaged using a laser-scanning confocal microscope prior to nano-indentation via AFM. Force–indentation measurements were performed by positioning the probe over the largest central region of neurons (Fig. 4B). By initiating a low set-point value of 1 nN (maximum force applied), a conical tipped AFM cantilever was used to deform cell membranes at a frequency of 2 Hz (Fig. 4B,C; see Movie 1). Force–indentation curves were acquired and fitted (the first 200 nm indentation) to a modified Hertz contact model in order to interpret the Young's modulus (apparent stiffness) of the cells (Fig. 4D). Mean±s.e.m. Young's modulus for neurons dissociated in summer (2537±125 Pa, N=30) was significantly less than that in winter (3041±104 Pa, N=44), as determined by Student's t-test (P=0.003; Fig. 4E). An additional experiment employed Hoechst 33342 (a live-cell nuclear dye) to confirm that there was no difference in stiffness between measurements made above or adjacent to nuclei. Measurements were performed on- and off-nucleus on the same cells and resulted in no difference in measured stiffness (P=0.36, Student's paired t-test, N=11).

Voltage-clamp recordings were obtained from 23 cells isolated during the winter and 33 cells isolated during the summer. An I–V curve (Fig. 5A) was generated from these currents and displayed a similar profile to that described in previous studies (Tachibana, 1983; Shingai and Christensen, 1986; Jonz and Barnes, 2007). Currents recorded from neurons were characterized by prominent inward rectification at negative potentials carried by inwardly rectifying K⁺ channels, a sustained inward Ca²⁺ current through L-type channels that was activated near −40 mV, and outward rectification at potentials above +40 mV (Fig. 5A). These currents and ion channels have been described elsewhere (Tachibana, 1983; Shingai and Christensen, 1986; Jonz and Barnes, 2007).

The reversal potential (V_rev), where no net current was recorded, in winter neurons was measured as an averaged value from all traces and was −79.8 mV (Fig. 5A inset, arrow). This value was near the calculated equilibrium potential for K⁺, which was −80.5 mV. In summer neurons, V_rev was approximately −50 mV but could not be accurately determined because of the relatively flat slope conductance across the voltage axis. We therefore measured current density at −80 mV (to approximate V_rev in winter neurons) to compare differences in membrane activity between seasons. Compared with current measured in winter neurons, mean current density at −80 mV in cells isolated during the summer was significantly inward at −1.4±0.3 pA pF⁻¹ (Fig. 5B; Student's t-test, P=0.0004). This change in current density corresponded to the positive (depolarizing) shift in V_rev in cells isolated during the summer months (Fig. 5A). Current density otherwise remained relatively unchanged between seasons at more negative potentials, and at potentials at which voltage-gated Ca²⁺ channels were active. At +5 mV, the approximate voltage at which peak Ca²⁺ current was recorded, there was no significant difference in current density between winter and summer neurons (Fig. 5B; Student's t-test, P=0.67).

From voltage-clamp data summarized in Fig. 5A,B, we identified two different I–V phenotypes: (1) those from cells in which I–V curves displayed a negative slope conductance within the physiological range and an often hyperpolarized V_rev, and (2) those from cells that did not display a negative slope conductance, had a relatively depolarized V_rev, and an inward current at −80 mV. Grouping cells according to these phenotypes, and analysis using Fisher's test, demonstrated a significant association (P<0.0001) between phenotype and season, such that cells with a negative slope conductance and hyperpolarized V_rev were more often observed in the winter (72%) compared with the summer (28%), when the other phenotype was more common (Fig. 5C). In the retina, horizontal cells maintain a membrane potential in darkness of −35 to −20 mV and hyperpolarize when photoreceptors are exposed to flashes of light (Shingai and Christensen, 1986; Yang et al., 1988; Thoreson and Mangel, 2012; Sun et al., 2017). We therefore measured R_in from currents evoked between membrane potentials that correspond to resting levels in the intact retina. Cells isolated during the winter had a significantly lower mean R_in of 3.7±0.5 GΩ compared with 8.8±1.2 GΩ in the summer (Student's t-test, P=0.0007; Fig. 5D).

Based upon the changes in membrane excitability observed in neurons between summer and winter, we hypothesized that action potentials generated by these cells would be subject to seasonal variation. Spontaneous changes in [Ca²⁺]_i were measured in a total of 144 cells. Cells isolated in the summer typically displayed a plateau phase of the Ca²⁺-based action potential shortly after reaching peak [Ca²⁺]_i, whereas cells isolated during the winter months displayed a profile characterized by a transient rise in [Ca²⁺]_i (Fig. 6).

To quantify seasonal changes in spontaneous [Ca²⁺]_i events, data from event amplitude, duration, AUC, frequency and time to peak were compiled from up to 208 spontaneous events from 76 cells in summer, and 209 spontaneous events from 68 cells in winter. Means±s.e.m. for summary data are shown in Fig. 7. Amplitude and AUC values were obtained from relative changes in the Fura-2 emission ratio (F₃₄₀/F₃₈₀) and are proportional to [Ca²⁺]_i. Parameters of spontaneous action potentials, such as amplitude and frequency, were not significantly different between seasons (Fig. 7A,D; Student's t-test). However, analysis of the duration of action potentials between seasons revealed that cells isolated during the summer had a significantly longer duration (56.1±7.2 s) compared with those isolated in the winter (34.9±3.4 s; P=0.0057, Student's t-test; Fig. 7B). In addition, AUC and time to peak were significantly greater in the summer compared with the winter (P=0.049 and P=0.042, Student's t-test; Fig. 7C,E), a trend that is consistent with longer duration action potential activity during the summer months.

The present study demonstrates that seasonal changes in membrane structure and function in goldfish horizontal cells occur when animals are held under constant environmental conditions, such as temperature and photoperiod. We integrated techniques of mass spectrometry, AFM and cell physiology to compare the chemical and biophysical properties of neurons isolated during the summer and winter months. Winter neurons displayed a greater proportion of unsaturated PEs and an increase in membrane stiffness. In addition, neurons isolated during the winter months exhibited reduced membrane excitability and shorter-duration Ca²⁺-based action potentials compared with summer neurons.

Our analysis of membrane chemistry in whole retina using mass spectrometry revealed 153 species of phospholipids. Of these, only five species of PEs were found to change in concentration throughout the year, where PEs were highest in the winter compared with the summer. In addition, the PEs that increased in concentration during the winter were unsaturated. The proportion of other phospholipids, such as PCs, did not exhibit seasonal changes. Remodelling of the plasma membrane is an important aspect of thermal adaptation. PEs are disordering lipids, as a result of their conical shape, and decrease the degree of close packing of lipids within the bilayer compared with other phospholipids (Reynolds et al., 2014). It has previously been demonstrated that a reduction in temperature produces an elevated proportion of PEs relative to PCs, and an increase in unsaturated fatty acids (particularly long-chain polyunsaturated fatty acids) in cold-tolerant animals (Hazel, 1988, 1995; Pruitt, 1988). This was more recently demonstrated in the congeneric crucian carp, in which a decrease in temperature was the primary factor for a greater proportion of unsaturated PEs in cell membranes from whole-brain samples in both wild-caught and laboratory-acclimated animals, although changes in PEs in laboratory animals were more pronounced (Käkelä et al., 2008). Such changes are important for HVA to maintain constant viscosity or fluidity within the membrane as temperatures change. To our knowledge, the present study is the first report of the occurrence of HVA in vertebrate neurons in the absence of external cues, such as would be experienced during acclimatization to lower temperatures. The seasonal changes in PEs that we observed in the goldfish retina represent only 3% of total phospholipid species, but suggest potentially important chemical modifications in the bilayer of retinal neurons that may occur before, or in parallel with, the temperature-induced changes described above. Because HVA induced by temperature acclimatization may take weeks to occur (Cossins et al., 1977; Sellner and Hazel, 1982), a mechanism of temperature-independent changes in PEs may ensure that structural modifications in the membrane bilayer, which are critical for HVA, begin before environmental temperatures begin to change. This may provide cold-tolerant animals, such as goldfish, with an additional advantage for surviving cold temperatures by reducing the time required for HVA.

We found that winter neurons displayed an increase in membrane–cortical stiffness, as determined by nano-indentation by AFM and subsequent measurement of Young's modulus. While it is important to note that AFM experiments were performed on isolated horizontal cells and mass spectrometry was carried out on whole retina, which includes multiple cell types, the increase in apparent stiffness in isolated horizontal cells during the winter correlates well with our observation of increased concentrations of PEs in the retina. Dawaliby et al. (2016) demonstrated that an increase in the PE:PC ratio in liposomes increased bilayer rigidity, thereby representing a potentially important mechanism for maintaining cell membrane fluidity. Moreover, in winter neurons, we recorded an increase in PE P-18:0_18:1, a plasmalogen that has been shown to form more rigid bilayers (Rog and Koivuniemi, 2016). In a similar manner, sterols maintain fluidity by inhibiting the transition of the membrane to the solid gel state, but sterols also increase the rigidity of fluid membranes by limiting the flexibility of neighbouring unsaturated acyl chains (Holthuis and Menon, 2014). Al-Rekabi and Contera (2018) recently confirmed that, at high cholesterol concentrations, model lipid bilayers had a higher Young's modulus and an increased membrane stiffness.

Whether PEs increase membrane stiffness whilst maintaining membrane fluidity in the same way as sterols has not yet been resolved and will require further attention. The present study did not assess the concentration of cholesterol in cell membranes; however, in wild crucian carp, no seasonal changes were detected in the proportion of neuronal membrane cholesterol to other phospholipids (Käkelä et al., 2008), suggesting that cholesterols likely did not contribute to the increase in stiffness observed in the present study. The mechanics of the plasma membrane and cortex, particularly actin filaments, are physically linked and so measurement of Young's modulus will include local elastic responses of both membrane and cytoskeleton (Haase and Pelling, 2015). However, local measurements of Young's modulus were taken over the soma, away from higher concentrations of actin filaments that were found primarily in horizontal cell dendrites. Although actin is present throughout somata and dendrites, a higher concentration of dendritic actin was previously described in horizontal cells of white bass (Roccus chrysops; Vaughan and Lasater, 1990). The location of the nucleus within the cytosol may also impact measurement of Young's modulus in whole-cell preparations (Haase and Pelling, 2015). In the present study, however, we performed nano-indentation experiments on the membrane over large central regions of the soma, away from the nucleus; and in control experiments, where we compared the Young's modulus between on- and off-nucleus, we found no differences in our results.

The seasonal changes in membrane chemistry and stiffness reported in the present study are important because they may have a corresponding impact on the function of integral membrane proteins, such as ion channels. Membrane lipids modulate ion channels in one of two ways: through non-specific interactions – such as by changes in membrane curvature, thickness or fluidity – or through specific binding of select lipids to protein domains. Lipid–protein interactions may therefore lead to significant effects on ion channel activity and membrane excitability (Tillman and Cascio, 2003; Poveda et al., 2014; Rusinova et al., 2014). A change in the proportion of phospholipids with saturated versus unsaturated acyl chains, as occurs during HVA, may also impact the function of membrane proteins. Polyunsaturated fatty acids can modify gating and activation of voltage-gated ion channels (Moreno et al., 2012). In vertebrates, the degree of polyunsaturation in membrane phospholipids is also correlated with Na⁺/K⁺-ATPase and Ca²⁺-ATPase activity and maintenance of ionic gradients (Cossins et al., 1981; Hulbert and Else, 1999, 2005; Arnold et al., 2015). Importantly, naturally occurring seasonal acclimatization, or laboratory-based thermal acclimation, have been shown to affect activity of ion channels in excitable membranes of fish (Hassinen et al., 2007, 2008; Rohmann et al., 2013; Kubly and Stecyk, 2015; Vornanen, 2016; Filatova et al., 2019), although these have not yet been linked specifically with structural changes in the phospholipid bilayer.

We found that neurons isolated during the winter had a reduced membrane excitability compared with summer neurons. We measured the parameters V_rev and R_in as indicators of membrane excitability. V_rev is the voltage at which there is no net current measured across the membrane and provides an estimate of the resting membrane potential, which is not measured directly in voltage-clamp experiments. A shift in V_rev in the hyperpolarizing (i.e. leftward) direction on an I–V curve would represent a decrease in membrane excitability, as the resting potential moves further away from the activation threshold of voltage-dependent ion channels and generation of an action potential becomes less likely. R_in is inversely proportional to membrane conductance and is a convenient measure of ion channel activity. A membrane with a greater number of open channels would have a lower R_in and, according to Ohm's law, would be less excitable because a given inward current would have only a small depolarizing effect on the membrane potential. In the CNS and other excitable tissues, R_in and membrane excitability are controlled by K⁺ channels that are open at rest, such as the inwardly rectifying K⁺ (Kir) channels and leak K⁺ channels (Nichols and Lopatin, 1997; Reimann and Ashcroft, 1999; Goldstein et al., 2001; Lesage, 2003; Hibino et al., 2010).

Winter neurons had a V_rev of −80 mV, which occurred at a segment of the I–V curve that displayed a positive slope conductance as it crossed the voltage axis. This indicates a region of high membrane stability, where a shift in membrane potential in either direction would be expected to elicit an opposing current that would return the membrane potential to this resting state (Shingai and Christensen, 1986). Moreover, because V_rev was approximately equal to the calculated equilibrium potential for K⁺ in our experiments, it was evident that ionic permeability of membranes in winter was due primarily to K⁺ channels. Both voltage-gated and inwardly rectifying K⁺ channels are present in teleost horizontal cells, although only Kir is substantially activated at resting potentials (Tachibana, 1983; Lasater, 1986; Jonz and Barnes, 2007). By contrast, summer neurons displayed inward (i.e. depolarizing) current at −80 mV and had correspondingly depolarized values of V_rev. This suggests that neurons in summer maintained a relatively right-shifted resting membrane potential as a result of a relatively greater influence of other ion conductances on V_rev.

In the retina, horizontal cells receive glutamatergic input from pre-synaptic photoreceptors. As a consequence, horizontal cells maintain a relatively depolarized resting membrane potential in vivo, broadly estimated to be in the range of −35 mV to −20 mV, and hyperpolarize when photoreceptors are exposed to flashes of light (Shingai and Christensen, 1986; Yang et al., 1988; Thoreson and Mangel, 2012; Sun et al., 2017). Our experiments were performed on isolated cells, where resting membrane potential would be relatively hyperpolarized in the absence of glutamate. Our results therefore indicate that seasonal modulation of membrane excitability must be an intrinsic property of horizontal cells. To address the issue of whether horizontal cells might respond differently to excitatory inputs in winter versus summer, we measured R_in over a range of voltages that correspond to resting potentials in the intact retina. We found that summer neurons had a higher R_in compared with those isolated during the winter. This suggests a higher level of membrane excitability in summer, as resting potentials would be closer to the activation potential of voltage-dependent Ca²⁺ channels (i.e. above −45 mV; Tachibana, 1983; Lasater, 1986; Jonz and Barnes, 2007).

Ca²⁺-based action potentials in horizontal cells have been described in a number of species, and are present in isolated cells, tissue slices and intact retina (Johnston and Lam, 1981; Tachibana, 1981; Shingai and Christensen, 1986; Lasater et al., 1984; Murakami and Takahashi, 1987; Dixon et al., 1993; Takahashi et al., 1993; Kreitzer et al., 2012; Country et al., 2019, 2021). Although a specific role for action potentials in horizontal cells has not been firmly established, they may be involved in Ca²⁺-dependent regulation of pH (Kreitzer et al., 2012), or be important for light-to-dark transitions in membrane potential, or play a role in feedback inhibition of pre-synaptic photoreceptors (Country et al., 2019). When monitoring intracellular Ca²⁺ dynamics, we found that spontaneous Ca²⁺-based action potentials in summer neurons were of longer duration and greater AUC (an integrated measure of intracellular Ca²⁺ activity over time). The increase in Ca²⁺ activity is consistent with a depolarized resting state and higher membrane excitability in summer neurons, where activation of voltage-dependent Ca²⁺ channels is more likely. We previously showed that increasing membrane excitability by injection of current to depolarize the membrane potential increased action potential activity in isolated horizontal cells (Country et al., 2019). As predicted by seasonal measurements of V_rev, our Ca²⁺-imaging data therefore point to maintenance of a more depolarized resting membrane potential in summer neurons.

Kir is important in central neurons for setting resting membrane potential and controlling membrane excitability (Nichols and Lopatin, 1997; Reimann and Ashcroft, 1999; Hibino et al., 2010), and in horizontal cells Kir is also involved in accelerating the hyperpolarizing response to light (Dong and Werblin, 1995). In spiny neurons of the striatum, Kir hyperpolarizes the membrane potential to a ‘down state’ and reduces R_in, thereby producing a shunting effect that makes the cell relatively insensitive to small synaptic inputs (Nisenbaum and Wilson, 1995; John and Manchanda, 2011; Wilson, 2008). An analogous mechanism of shunting inhibition, though involving Cl⁻ conductance through GABA receptors, underlies suppression of action potentials in cortical neurons of anoxia-tolerant turtles (Pamenter et al., 2011). Moreover, a hyperpolarizing shift in V_rev by Kir was recently proposed to contribute to the temperature-dependent loss of electrical excitation in cardiac myocytes in fish, which may play a role in seasonal acclimatization (Vornanen et al., 2014; Vornanen, 2016; Badr et al., 2018). Kir channel expression has also been shown to be regulated by thermal acclimation (Hassinen et al., 2007, 2008; Vornanen, 2016). In addition, an increase in Kir current density and shortened action potential was reported in cardiac myocytes in the European shorthorn sculpin (Myoxocephalus scorpio) caught during the winter (Filatova et al., 2019).

We propose that, in a similar manner, Kir may modulate R_in and membrane excitability in goldfish horizontal cells during the winter to reduce action potential duration or depolarizing inputs from photoreceptors, thereby lowering energy requirements. Regulation of ionic gradients, such as following an action potential, consumes approximately 50% of neuronal ATP (Bickler and Buck, 2007; Howarth et al., 2012). Moreover, the unstimulated retina generally has high energy requirements as a result of photoreceptor dark current and high Na⁺/K⁺-ATPase activity (Ames et al., 1992; Country, 2017). Horizontal cells receive tonic stimulation from photoreceptors yet actively maintain transmembrane ionic gradients, regulate intracellular Ca²⁺ homeostasis and (in fish) support loading of synaptic vesicles with γ-aminobutyric acid (GABA) for Ca²⁺-dependent neurosecretion (Thoreson and Mangel, 2012; Country and Jonz, 2017) – all of which are mechanisms that consume energy in the form of ATP. Reduced activity in horizontal cells during the winter may, therefore, lower energy consumption in cold- and anoxia-tolerant animals, such as goldfish. In crucian carp, exposure to anoxia reduced electroretinogram (ERG) activity by 90% (Johansson et al., 1997). A reduction in ERG activity by anoxia was also reported in other anoxia-tolerant vertebrates (Stensløkken et al., 2008), and by low temperature in hibernating ground squirrels (Zhang et al., 2020), suggesting that an overall reduction in retinal activity may be an important adaptation for survival in vertebrates that are tolerant to low oxygen or temperatures. Further investigation is needed, however, to determine whether the seasonal changes observed in the present study are confined to Carassius spp. and other vertebrates adapted to environmental changes, and whether they may be present in other neurons of the retina or CNS.

In the present study, goldfish were imported at mixed age and from a natural photoperiod, but were maintained in our animal facilities at constant photoperiod, temperature and oxygen availability. Although we have not yet explored what might be directly regulating the observed seasonal patterns in membrane structure and excitability, we suspect that biological rhythms associated with seasonal or annual cycles were entrained before fish were imported into our holding facilities. Seasonal changes in photoperiod and temperature regulate many important physiological processes, such as growth, feeding, metabolism and reproduction (Peter and Crim, 1979; Zhang et al., 2009; Cowan et al., 2017; Nakane and Yoshimura, 2019). In goldfish, reproductive ability is maintained in captivity, where gene expression patterns in the brain, including genes involved in neurotransmission, change between May and August, which corresponds to a transition from prespawning to sexual regression (Zhang et al., 2009). Furthermore, there is evidence that reproductive transitions are entrained in fish and continue for multiple cycles without the need for environmental cues. Duston and Bromage (1991) found that 2 year old rainbow trout (Oncorhynchus mykiss) subsequently maintained in captivity at constant photoperiod and temperature continued to display seasonal rhythms of gonadal maturation and ovulation for up to 3 years. Further investigation is required to establish whether seasonal changes in gene expression that modify phospholipid composition and membrane excitability are entrained in neurons of vertebrates adapted to extreme environments, and to potentially determine the longevity of these cyclical patterns in wild or longer-living species. In addition, it will be of interest to explore whether regulation of specific phospholipid–protein signalling mechanisms might impact seasonal changes in membrane excitability. For example, the membrane phospholipid phosphoinositol 4,5-bisphosphate (PIP₂) is essential for maintaining activity of most Kir channels (reviewed by Hibino et al., 2010), and an increase in PIP₂ activity was associated with the hyperpolarizing response to light in horizonal cells of Xenopus laevis (Anderson and Hollyfield, 1981, 1984). Interestingly, seasonal changes in the regulation of ion channels by PIP₂ in dorsal root ganglion cells modulates thermal sensitivity in mice during adaptation to changing ambient temperatures (Fujita et al., 2013; Brenner et al., 2014). Whether direct interactions between PIP₂, or other membrane phospholipids, and ion channels are involved in controlling seasonal changes in membrane excitability in the goldfish retina will be an important avenue for further research. Future studies focused on identifying the role of seasonal patterns controlling membrane excitability in neurons will provide a better understanding of the evolutionary adaptations that allow some organisms to survive extreme environmental change.

The authors thank Cameron Jonz for colour modifications to Fig. 4B; and Dr Steven Barnes for access to animals and equipment at the Department of Physiology and Biophysics, Dalhousie University.