Unique arylalkylamine N-acetyltransferase-2 polymorphism in salmonids and profound variations in thermal stability and catalytic efficiency conferred by two residues

Biochemical, physiological and behavioral processes exhibit daily and annual rhythms. Photoperiod is the key proximate driver of the broad seasonal phasing of major biological functions of teleosts such as growth, reproduction and migration (Björnsson et al., 2011; Falcón et al., 2010; Migaud et al., 2010). While light provides an unambiguous temporal signal, temperature is suggested to have a modifying role, for example, enabling reproductive cycles to be locally tuned to inter-annual variations in thermal conditions (Pankhurst and King, 2010). In the context of a changing environment, temperature increase due to global warming is likely to alter the way ectotherm vertebrates decode time, in addition to more general effects on metabolism, physiology and behavior. Specifically, it is believed that the production of melatonin, the time-keeping hormone of vertebrates, will be impacted. Melatonin from the pineal gland is a hormonal messenger of daily and annual day–night cycles (photoperiod), and is released at night into the blood and cerebrospinal fluid. Melatonin is produced from serotonin in two enzymatic steps: arylalkylamine N-acetyltransferase (AANAT; EC 2.3.1.87) catalyzes the conversion of serotonin into N-acetylserotonin, which is then methylated by hydroxyindole-O-methyltransferase (HIOMT; EC 2.1.1.4) to produce melatonin (Falcón, 1999; Falcón et al., 2007; Klein et al., 1997). Variations in pineal AANAT activity are responsible for the rhythmic secretion of melatonin. Teleosts possess two AANAT subfamilies, AANAT1 and AANAT2, which are preferentially expressed in the retina and the pineal gland, respectively (Falcón, 1999; Falcón et al., 2007). The existence of multiple AANATs in teleosts is likely to result from a whole genome duplication that occurred shortly after the divergence of teleosts from the main tetrapod lineage (Volff, 2005).

Pineal AANAT2 activity of teleost fish is regulated by both light and temperature. In most species investigated, Aanat2 gene expression is controlled by intrapineal circadian clocks (Appelbaum and Gothilf, 2006; Coon et al., 1999; Gothilf et al., 1999), while light and temperature modulate AANAT2 protein amount and activity (Falcón et al., 2001; Falcón et al., 1987; Falcón et al., 1989). The result of these interactions shapes the pineal melatonin oscillations: photoperiod dictates the duration of the nocturnal surge, while temperature controls its amplitude (Falcón, 1999). Within this context, salmonids occupy a special position. Species of the family Salmonidae are migratory teleosts and their perception and anticipation of seasonal changes in the environment are crucial. In contrast to other teleost species, they lack a functional circadian clock in the pineal gland (Falcón, 1999; Iigo et al., 2007), yet in vitro and in the wild they exhibit daily melatonin rhythms that mirror exactly the prevailing day length throughout the year (Gern and Greenhouse, 1988; Strand et al., 2008). Temperature modulates the amplitude of the nocturnal surges in AANAT2 activity (Falcón, 1999) and melatonin secretion (Max and Menaker, 1992), and has also been suggested to be a driving signal of salmonids' biological rhythms because the seasonal changes in the amplitude of the nocturnal melatonin surge correlate with the ambient temperature changes (Strand et al., 2008). With this background, we recognized the need to increase our knowledge on the time-keeping system of salmonids. We chose two species: the rainbow trout, Oncorhynchus mykiss (Walbaum 1792), a temperate species that has been thoroughly studied, and the Arctic charr, Salvelinus alpinus (Linnaeus 1758). The latter is the northernmost living salmonid and, in the Arctic, is subjected to extreme temperature and light conditions. We investigated the impact of temperature on Arctic charr melatonin secretion; we also bring new information on the diversification of AANAT2 proteins in teleosts, their structure and kinetic relationships in salmonids, paying special attention to the role of temperature.

We used S. alpinus reared at the Tromsø Aquaculture Station (University of Tromsø, Norway). The fish were 2-year-old offspring of Arctic charr originally caught at the high-Arctic Svalbard archipelago (80°N) in 1990 and reared under natural conditions. All experiments were performed according to the European Union regulations concerning the protection of experimental animals.

Two different experiments were conducted to assess the effect of temperature on Arctic charr pineal gland melatonin production in vitro.

Pineal glands of six animals were sampled directly after euthanization and were cultured individually in 1 ml RPMI 1640 medium (Sigma-Aldrich, St Louis, MO, USA) in a 48-well plate (Nunclon Surface, NalgeNunc, Roskilde, Denmark). They were acclimated for 24 h at 5°C under an 8 h:16 h light:dark photoperiod (Termaks 8182 incubator, Termaks AS, Bergen, Norway). The day of the experiment, each organ was placed in the dark and subjected to increasing temperature challenges from 0 to 45°C in steps of 5°C. Each challenge lasted for 90 min and was preceded by 30 min incubation at 5°C. At each step the medium was collected and stored at −20°C until melatonin was quantified using a validated radioimmunoassay as described previously (Strand et al., 2008).

Pineal glands from 48 individuals were excised and cultured as described above. Groups of six pineal glands were incubated individually in 1 ml RPMI 1640 medium in wells of a 48-well plate. Each plate was then placed in the dark at temperatures ranging from 0 to 35°C. The media were sampled after 3 h and frozen at −20°C for radioimmunoassay measurement of melatonin concentration as indicated above.

Pineal glands from 10 individuals from the same high-Arctic strain were placed in RNAlater (Ambion, Austin, TX, USA) and archived at −80°C. Total RNA was extracted using Trizol (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocol. cDNA synthesis was achieved using the SMART RACE cDNA Amplification Kit (Clontech, Mountain View, CA, USA). The first portion of the AANAT2 open reading frame (ORF) was obtained by nested PCR using degenerated outer and inner primers (Table 1) and Advantage 2 Polymerase Reaction Mix (Clontech, Mountain View, CA, USA). The amplified fragment was used to design specific primers to obtain the full AANAT2 mRNA by RACE (Clontech, Mountain View, CA, USA). After successful amplification, PCR products were cloned using pGEM-T Easy (Promega, Madison, WI, USA) and DH5α Escherichia coli, purified by Miniprep (Sambrook and Russell, 2001) and sequenced. Two variants of a full AANAT2 sequence were obtained, which differed at positions 114 and 160. One had two valines (VV variant) and the other had an isoleucine and an alanine (IA variant).

The two variants were aligned with the O. mykiss AANAT2 sequence (GenBank, NM_001124257.1). Again, the differences were seen at positions 114 and 160. The trout exhibited an isoleucine at position 114 and a valine at position 160 (IV variant). In order to explore the catalytic properties of these enzymes and to relate kinetics with structure, we designed an AANAT2 mutant sequence corresponding to the fourth possible combination presenting a valine at position 114 and an alanine at position 160 (VA mutant). For this purpose, a chimera sequence was designed from both variants of the S. alpinus sequences. The N terminus of the AANAT2_VV variant and the C terminus of the AANAT2_IA variant from S. alpinus were used to obtain the AANAT2 VA mutant. Specific primers were designed for both variants and used in two PCR rounds (Table 1). The product of the last PCR was cloned, purified and sequenced as indicated above. The same strategy was employed to obtain the O. mykiss AANAT2 (IV variant), but using the first part of the IA variant and the last part of VV variant.

To estimate the intra- and inter-individual distribution of AANAT2 messenger variants expressed by the Arctic charr, we cloned and sequenced the ORF from nine individuals of the Svalbard strain raised at the Tromsø Aquaculture Station. Twelve clones per individual were sent for sequencing. Each Arctic charr was finally characterized by an average of 8.3 AANAT2 clones. Sequences were corrected by eye using BioEdit (Ibis Biosciences, Carlsbad, CA, USA). Nucleotide polymorphism was assessed in DNAsp V5 (Librado and Rozas, 2009).

The AANAT2 sequences of O. mykiss, S. alpinus IA and VV variants, and the VA mutant were transferred into a pGEX4T1 production plasmid (Novagen, EMD Chemicals, Philadelphia, PA, USA) containing a glutathione S-transferase (GST) tag. Plasmids were transformed in E. coli BL21 Rosetta bacteria (Merck KGaA, Darmstadt, Germany). Bacteria were grown in 2×500 ml Luria-Broth medium at 37°C. Protein production was induced with 0.5 mmol l⁻¹ IPTG (isopropyl β-D-1-thiogalactopyranoside), overnight at 20°C. Subsequent steps were carried out at 4°C. Bacterial cultures were centrifuged for 10 min at 1800 g. The bacterial pellet was then washed with 1× Tris buffered saline (TBS; 50 mmol l⁻¹ Tris and 150 mmol l⁻¹ NaCl, pH 8) and centrifuged for 10 min at 5000 g. Pellets were suspended in 50 ml lysis buffer (1× TBS + Complete protease inhibitor cocktail tablet, Roche, Basel, Switzerland) and sonicated on ice for 30 min by alternating 10 s on and off pulse episodes. The lysate was centrifuged for 30 min at 25,000 g after addition of 0.1% Tween and 1% Triton X-100.

Purification was performed using affinity chromatography glutathione sepharose 4B beads (GE Healthcare, Little Chalfont, UK). All of the following steps were conducted at 4°C and beads were pelleted by a 2 min centrifugation at 500 g in between each step. The beads were then washed with 10 ml phosphate buffered saline (PBS) and mixed with 10 ml TBSTT (1× TBS pH 8, 0.1% Tween 20, 1% Triton X-100) and 1% powdered milk. Beads were washed two times in 10 ml TBSTT before incubation for 1 h with the supernatant recovered at the end of production step. Beads were then pelleted and the supernatant was removed. After three TBSTT washes and one 1× TBS wash, the GST-AANAT2 fusion proteins were recovered by 10 min incubation in 1 ml elution buffer (Tris 50 mmol l⁻¹, glutathione 10 mmol l⁻¹); this step was repeated six times. Protein amount was measured using the Bradford colorimetric assay and bovine serum albumin (BSA) as a standard.

All assays were conducted on GST-AANAT2 fusion proteins as it was shown that the tag had no influence on AANAT kinetics (De Angelis et al., 1998). Enzymatic activity was measured adapting a previously published methodology (De Angelis et al., 1998) for the four produced recombinant AANAT2 enzymes. The enzymatic assays were run in triplicate, in 96-well plates (Greiner BioOne, Monroe, NC, USA) in a 100 μl reaction mix containing 0.1 mol l⁻¹ PBS pH 6.8, 0.025 g l⁻¹ BSA, 1 mmol l⁻¹ acetyl coenzyme A (AcCoA), 0.5 μg of enzyme and variable concentrations of substrate depending on the experiment. The mix without enzyme was first held for 10 min at the desired temperature before enzyme was added. After 10 min incubation, 150 μl of a stop solution {6 mol l⁻¹ guanidine hydrochloride, 12 mmol l⁻¹ EDTA, 4 mmol l⁻¹ DTNB [5,5′-dithiobis- (2-nitrobenzoic acid)], 0.1 mol l⁻¹ Tris pH 6.8} was added to the reaction mix and incubated for 5 min before reading on a spectrophotometer plate reader. The choice of buffer pH and molarity, and AcCoA and enzyme concentrations, was dictated by previous validation studies, making sure that initial velocity could be calculated. Saturation curves were performed with eight different concentrations of substrate at the desired temperatures, as indicated in the Results and figure legends. The impact of enzyme denaturation was investigated as described above using 20 mmol l⁻¹ tryptamine and a 20°C incubation temperature, and an enzyme solution that had been previously incubated at 45 or 65°C for 5, 10, 20, 40 and 60 min.

All data were analyzed and displayed using Prism 5.00 software (GraphPad Software, La Jolla, CA, USA).

For melatonin quantification, statistical analysis of the data was performed using ANOVA followed by Tukey's comparison of means when appropriate.

For enzymatic assays, the best-fitted equation was calculated for each saturation curve using the extra sum of squares F-test, implemented in Prism 5.00. These equations were used to estimate the Michaelis constant, K_M, the substrate concentration for which half of maximal velocity (V_max) is reached, giving an idea of substrate affinity, the turnover number k_cat (k_cat=V_max/[enzyme]), i.e. the number of substrate molecules each enzyme site converts into product per time unit, and the catalytic efficiency (k_cat/K_M). The 95% confidence intervals (CI) were calculated for both K_M and k_cat. The 95% CI of the ratio k_cat/K_M was calculated using an online calculator implemented on the GraphPad website (www.graphpad.com). The 95% CI is displayed in the graphs for each kinetic constant in order to simplify direct comparison, and values that exhibited no overlap of their CIs were considered significantly different (Cumming et al., 2007).

Three-dimensional structures of the four AANAT2 analyzed by enzyme assays were generated by homology modeling. The sequence alignment used for homology modeling was prepared in ClustalX available in the BioEdit program and is visualized by ESPript2.1 (Gouet et al., 2003). The models were generated using the Swiss Model server (Arnold et al., 2006) using the truncated ovine, Ovis aries, AANAT as a template [oaAANAT30-195; 1KUV (Wolf et al., 2002)]. It is important to note that the crystallographic structure of AANAT has previously only been achieved on a truncated form of the enzyme (Hickman et al., 1999a; Hickman et al., 1999b; Obsil et al., 2001; Wolf et al., 2002). The homology models obtained were subsequently subjected to energy minimization in 500 steps using the Macromodel option (Version 9.8) of Schrödinger software (http://www.schrodinger.com) in order to avoid steric clashes between modeled amino acids. The models were in addition minimized in the presence of the bisubstrate analog, the coenzyme A-S-acetyltryptamine, i.e. the substrate and the acetyl co-A covalently attached (Hickman et al., 1999a; Wolf et al., 2002).

A marked effect of temperature was seen on melatonin production (Fig. 1). There was a strong increase in melatonin secretion when the pineal glands were incubated at 20, 25 and 30°C compared with when they were cultured at higher or lower temperatures. The values obtained at 20, 25 and 30°C did not differ significantly (P>0.05). Excluding the 40 and 45°C values, those obtained at all other temperatures tested (0 to 15°C) were not significantly different from the 5°C control values, which remained the same throughout the experiment up to 35°C. No melatonin production was seen during incubation at 40 and 45°C.

Melatonin concentration increased from 0 to 15°C and decreased from 15 to 35°C (Fig. 2).

A total of 74 successful clones presenting the entire AANAT2 ORF were obtained for the nine sampled Arctic charr. Eight out of the nine individuals possessed the two IA and VV variants, but only six clones were sequenced in the individual missing the IA variant. The proportion differed with a significant overrepresentation of the VV isoform compared with the IA isoform (paired t-test, P<0.001; Fig. 3). Both variants presented the same length (627 bp) and 70 sites were polymorphic including 19 parsimony informative sites, corresponding to an average number of nucleotide differences of 7.07. Intra- and inter-variant nucleotide diversity was also assessed. The 49 VV and 25 IA variants possessed, respectively, 39 and 20 polymorphic sites corresponding to an average number of nucleotide differences of 1.75 and 1.67. Most of the polymorphic sites in one variant were not polymorphic in the other and vice versa. The average number of nucleotide differences between the two variants reached 13.4.

The O. mykiss enzyme (omAANAT2_IV), the S. alpinus variants IA (saAANAT2_IA) and VV (saAANAT2_VV) and the VA mutant were successfully produced and purified. All kinetic constants, K_M, k_cat and k_cat/K_M (Fig. 4), were calculated from the saturation curves performed at the different temperatures (not shown). The K_M values were of the same order of magnitude for the four enzymes at all temperatures tested (Fig. 4A). However, the VA mutant displayed the lowest K_M compared with the other AANAT2 enzymes. Temperature did not significantly affect K_M values, but saAANAT2_IA, saAANAT2_VV, mutant VA and omAANAT2_IV were substrate inhibited at 45°C (respective inhibition constants, K_i, were 100, 115, 174 and 15 mmol l⁻¹; not shown). The non-negligible substrate inhibition at this temperature explained the observed larger 95% CI. For saAANAT2_IA, the VA mutant and omAANAT2_IV, the k_cat values increased slightly with temperature, within the range 3.5–8.5, 7–14 and 6–26 s⁻¹, respectively; the increase was from 18 to 63.7 s⁻¹ with temperatures increasing from 5 to 45°C for saAANAT2_VV (Fig. 4B). The catalytic efficiency k_cat/K_M was relatively constant with temperature, and only saAANAT2_VV exhibited an increase between 5 and 25°C, after which it remained constantly high until 45°C (Fig. 4C).

The four enzymes displayed a notably different response in the thermal stability experiments (Fig. 5). omAANAT2_IV was totally inactivated at both 45 and 65°C. Both saAANAT2_VV and the VA mutant maintained approximately half of their normal activity (40 and 65%, respectively) after incubation at 45°C but less than 10% activity was detected after incubation at 65°C. After denaturation at 45°C, saAANAT2_IA was able to resume its normal activity, but not after 65°C incubation, where 90% of its activity was lost.

K_M, k_cat and k_cat/K_M (Fig. 6) values were calculated from saturation curves obtained by assaying the four enzymes with four different substrates: two indolethylamines (tryptamine and serotonin) and two phenylethylamines (PEA and dopamine). All four enzymes exhibited a higher K_M and a lower k_cat for phenylethylamines than for indolethylamines. The resulting lower k_cat/K_M for phenylethylamines was never higher than 2×10³ s⁻¹ mol⁻¹. Moreover, as these enzymes poorly acetylated phenylethylamines, measurements were more difficult, which explains the larger 95% CI. saAANAT2_VV exhibited an at least twofold higher catalytic efficiency for indolethylamines compared with the other enzymes. For all four enzymes the best catalytic efficiency was observed with serotonin.

In an attempt to understand the mechanisms underlying the differences observed between enzyme activity assays, homology models were generated for the four enzymes saAANAT2_IA, saAANAT2_VV, omAANAT2_IV and the VA mutant using the truncated ovine AANAT crystal structure [Protein Data Bank (PDB): 1KUV] as template. Based on the ovine enzyme/ligand structures (PDB 1KUV and 1CJW), the three loops illustrated with blue dashes in Fig. 7, with residues 55–57, 60–65 and 182–189, as well as residues 122–127, 131–138 and 159–171, could be considered to form the cofactor and substrate binding sites. In addition, residue 110 is also suggested to be involved in forming the serotonin/tryptamine binding site. As described for the ovine enzyme, the structures of the homology models displayed a central twisted seven stranded beta sheet surrounded by a total of five alpha helices on both sides (Fig. 8A). Differing in amino acid composition at only two positions, 114 and 160, where the first is located on the enzyme surface and the latter within the crevice constituting the substrate binding site, the molecules could be superimposed on each other with an average xyz displacement value of 0.2–0.4 Å for all atoms, thus demonstrating the similarity of the models. The overall tendency was that when considering position160 the enzymes containing Val were similar and those having Ala were similar. The only significant differences in main chain positions were found for regions 99–108, ~114, 157–163, those containing the mutations, and 185–190 (Fig. 8B). There was no obvious explanation for the differences between regions 99–108/114/157–163 and regions containing the mutations, but the differences between the latter and region 185–190 appeared to be most likely due to the introduction of a Val instead of an Ala at position 160. The bulkier Val160 restricts the conformational space of Phe189, resulting in a slightly different conformation of the Phe189 side chain in enzymes containing Val or Ala at position 160 (Fig. 9A). Val160 also causes slight displacements of the main chain atoms of residues 157–163 and 185–195. These displacements may in turn affect the conformation of the side chain of Glu191, such that the ionic interactions between Glu191 and Arg104 have different hydrogen binding patterns in the enzymes containing Val and Ala (Fig. 9B). In addition, Glu191 is almost completely buried [according to the Dictionary of Protein Secondary Structure, a program of pattern recognition of hydrogen-bonded and geometrical features (Kabsch and Sander, 1983)]. Thus, the negative electrostatic potential of Glu191 may also influence the conformation of the hydrophobic Leu105, perhaps explaining the xyz displacement in this region. It is interesting to note that minimization suggests that Gln190 and Glu176 had different conformations in the enzymes containing Val or Ala at position 160.

This work brings new information on the melatonin-generating system in salmonids, in particular on the enzyme that generates the melatonin rhythm and on the impact of temperature.

A previous study has indicated an effect of season on the amplitude of plasma melatonin surge in the Arctic charr (Strand et al., 2008). This, together with in vitro studies indicating temperature impacts on pineal AANAT2 activity (Benyassi et al., 2000; Falcón et al., 1996) and melatonin secretion (Iigo et al., 2007; Masuda et al., 2003; Zachmann et al., 1992; Zachmann et al., 1991), suggests that temperature may act directly on the Arctic charr pineal gland to regulate melatonin secretion. The impact of temperature may be different depending on the output measured (molecule, cellular metabolism, physiology). Accordingly, although the preferred temperatures of Arctic charr are below 11.7°C in summer and 8.7°C in winter (Mortensen et al., 2007), we felt it was necessary to investigate a wider range of temperatures in culture, as we did for the recombinant enzymes (see below). The two in vitro experiments designed here indicated the Arctic charr pineal gland responds directly to temperature. However, different results were obtained depending on whether the temperature challenges were applied sequentially or continuously. In the former case, maxima were seen between 20 and 30°C; in the latter case a maximum was seen at 15°C. A peak at 15°C agrees with previous data obtained in cultured pineal glands of the rainbow trout in which AANAT2 activity was highest at temperatures ranging from 12 to 15°C after 6 h of static culture (Benyassi et al., 2000) whereas melatonin secretion peaked at 18°C (Max and Menaker, 1992) or 20°C (Iigo et al., 2007) in superfused pineal cultures (i.e. when the medium is continuously renewed). The differences observed in the Arctic charr pineal glands are probably due to the use of different experimental designs, because all the pineal glands were from individuals of the same population. It is possible that a sort of ‘acclimation effect’ (Currie et al., 1998) has played a role in the sequential experiment because the same pineal glands were gradually acclimated to increasing temperature (including intermediate 5°C control steps in between each tested temperature). Conversely, acute and longer acclimation to high temperatures, as in Experiment 2, may have induced deleterious effects. Previous studies have shown that the initial thermal restriction is set at the highest level of functional complexity, i.e. loss of organ function occurs first, due to limited oxygen supply, whereas the enzyme functions per se are maintained in a wider window of thermal tolerance (Pörtner, 2006). Temperature increase enhances oxygen demand through rising metabolic rates, while oxygen solubility decreases and hence results in a double challenge leading to hypoxia. In the first culture, organs experienced returns to 5°C between the different temperature challenges and thus may have had limited hypoxia and organ damage. In favor of this is the observation that Arctic charr recombinant AANAT2 isoforms display high stability at 45°C (see below), a temperature fatal to melatonin secretion irrespective of the experimental conditions. Whatever it may be, the time-keeping system of the Arctic charr displays a range of tolerance above what is experimented by the fish in its natural habitat, where temperatures above 15°C are seldom experienced and are far above the temperature that the strain of charr used in the present study experiences (0 to 5°C) at 80°N (Nilssen et al., 1997). On the contrary, it is noteworthy that melatonin secretion remained relatively high in Experiment 2 at the temperatures Arctic charr experience in the cold range, e.g. 0–5°C.

Previous studies have shown that an increase in temperature does not necessarily imply an increase in melatonin secretion, i.e. there is no linear correlation between temperature increase and melatonin secretion. Thus, in the cold-adapted fish Catostomus commersonii, melatonin secretion is higher at 10°C than at 20°C, while the opposite holds true in temperate or warm-adapted species, as is the case of pike, Esox lucius (Falcón et al., 1994; Zachmann et al., 1992). This implies that temperature has complex effects involving both cellular-mediated regulatory processes as well as species-specific enzyme adaptations. In favor of this are the following observations: (1) in cultured teleost pineal glands, the accumulation of cyclic AMP (an intracellular second messenger that promotes AANAT2 activation) and the activity of AANAT2 both display a similar response to temperature changes, with a peak response that depends on the species (12/15°C in trout, 20/25°C in pike) (Benyassi et al., 2000; Falcón, 1999); and (2) the activity of recombinant AANAT2 proteins varies with temperature in the same way as the native enzyme does (Coon et al., 1999; Zilberman-Peled et al., 2004). In other words, the maximal catalytic response of the recombinant AANAT2 protein varies between 0 and 37°C in a species-dependent manner. This is in marked contrast to the AANAT1 enzymes of teleosts and AANAT of tetrapods, whose activity increases linearly with temperature from 0 to 37°C. In order to obtain further insights into how temperature impacts S. alpinus AANAT2, we decided to clone and produce the enzymes for further structural and functional characterization.

The cloning strategy used allowed us to obtain two different AANAT2 transcripts. To our knowledge, this is the first time two different copies of AANAT2 have been detected in teleost fish. These two variants present no insertion or deletion; they differ by many synonymous changes at the nucleotide level, but display only two amino-acid differences, an Ile or a Val at position 114 and an Ala or a Val at position 160. There is good indication that both are expressed in the same pineal gland, discarding the possibility that they were representatives from two different fish strains. It is not yet known whether this resulted from gene duplication [as suggested for AANAT1/AANAT2 and AANAT1a/AANAT1b dichotomies (Coon and Klein, 2006)] or whether they are different alleles of the same gene. Most interestingly, the saAANAT2_IA and saAANAT2_VV variants displayed a high degree of conservation with the trout AANAT2, from which they differ by only one or two amino acids at the same positions. This provided a unique opportunity to study the impact of two residues that occupy key positions in the enzyme structure, and are important (at least position 160) for substrate selectivity (Zilberman-Peled et al., 2011). In order to further explore the role these residues play, we constructed a hybrid enzyme presenting a Val114 and an Ala160; this _VA version complemented the natural saAANAT2_IA, saAANAT2_VV and omAANAT2_IV combinations.

The four enzymes exhibited the classical AANAT2 pattern of substrate preferences: they acetylated indolethylamines much better than phenylethylamines (Benyassi et al., 2000; Cazaméa-Catalan et al., 2012; Coon et al., 1999; Gothilf et al., 1999; Zilberman-Peled et al., 2004; Zilberman-Peled et al., 2011). The kinetic constants were within the range of those already reported for other AANAT2, except for saAANAT2_VV, which had a catalytic efficiency for indolethylamines at least twofold higher than that of the other enzymes. This difference seems mainly due to a higher k_cat for tryptamine and a lower K_M for serotonin. The kinetic constants of saAANAT2_IA, omAANAT2_IV and the VA mutant changed only slightly with temperature. In contrast, saAANAT2_VV exhibited a marked rise in k_cat with increasing temperature, resulting in an increase in k_cat/K_M. Previous studies have shown an increase in both k_cat and K_M with rising temperature in teleosts (Cazaméa-Catalan et al., 2012; Fields and Somero, 1998; Johns and Somero, 2004). Another interesting observation resides in the fact that a very similar primary structure, with only two amino-acid position changes, induced significant differences in thermal stability. Thus, saAANAT2_IA activity remained unaffected by incubation at 45°C, while om_AANAT2_IV activity was rapidly and totally inactivated; an intermediate situation characterized saAANAT2_VV, with 60% inactivation, while the hybrid enzyme displayed only 40% inactivation. All enzymes were inactivated after incubation at 65°C. This would suggest a role of Val114 and Val160 in thermal stability, as discussed below.

In order to obtain further insight into the role these residues play, we thus decided to investigate the 3D structure of the enzymes. The 3D structure of a protein is a fine balance between the minimum overall energy of folding and the steric, polar and hydrophobic effects. A slight change in this balance can induce drastic changes in activity. A domino effect can be observed when a single change induces a succession of other changes, which finally destabilizes the overall structure. However, the effect of a single change is difficult to predict, as some residues can take the place of others and the structure can rearrange in other ways to adopt an active folded protein. In the present case, we observed that none of the mutations, either natural or directed, entrained a total misfolding and inactivity of AANAT2, as indicated by the activity measurements. The superimposition of the modeled enzymes (Figs 8, 9) showed that the mutations in position 114 or 160 have different effects on the spatial position of the other residues, in their direct surroundings as well as in more distant positions. Residue 114 is located on the protein surface and replacing Val114 by an Ile appears to entrain changes affecting mainly the surrounding area. Position 160, however, is more buried in the active site cleft. A mutation in position 160 induces more dislocations in the rest of the structure, thus accounting for most of the conformational change (even far from position 160). The most obvious and significant changes were the high xyz displacement values seen for the regions containing residues 187–189 (substrate binding site) and 157–163 (cofactor site), which are affected by the Val160Ala mutation. The high xyz displacement values seen for residues 103–107 were less obvious, but these were also likely to be the result of perturbations caused by the Val160Ala mutation, as described above, rather than the mutation at position 114. The regions containing residues 103–107 and 187–189 have both been shown to be part of loops 2 and 3 involved in substrate binding in the ovine AANAT (Hickman et al., 1999a; Hickman et al., 1999b). Position 160 has already been suggested to be an important position for substrate selectivity in AANATs. In a recent study, the naturally occurring Ile160 in seabream (Sparus aurata) AANAT2 was mutated to methionine (Zilberman-Peled et al., 2011). The authors chose this mutation because of the natural occurrence of methionine at this position in seabream AANAT1 and ovine AANAT. The effect of this mutation on seabream AANAT2 enhanced acetylation, and decreased K_M, for both indolethylamines and phenylethylamines, confirming the important role this position seems to play in substrate preference. However, in our study the effect of a mutation at position 160 was difficult to characterize as enzymes having the same residues at this position, such as saAANAT2_VV and omAANAT2_IV, displayed different kinetic constants. So, the mutation at position 114 also contributes to the differences in catalytic properties. Therefore, the catalytic differences seem to result from a kind of synergistic effect of mutations at both sites.

The Val160Ala mutation might account for the thermal stability observed for saAANAT2_IA and the VA mutant. The models suggest that these two enzymes display different hydrogen binding patterns in the ionic interaction (salt bridge) between Glu191 and Arg104 compared with the two other AANAT2 enzymes. The salt bridge where the carboxy group of the glutamic acid has no water-accessible surface is likely to stabilize the interaction between loops 1 and 3. Salt bridges were previously suggested to be important for conferring thermal stability to proteins, although this importance has been debated (Saelensminde et al., 2009). These results should be interpreted with care; however, differences in the ionic interaction between Glu191 and Arg104 are the sort of changes conferred by the Val160Ala mutation that could contribute to rendering saAANAT2_IA and mutant VA more thermal tolerant. It has been shown that a single or a few amino-acid substitutions can affect the impact of temperature on enzyme kinetics when these appear in regions likely to influence the conformational mobility of regions involved in catalytic conformational changes (Holland et al., 1997; Pavlicek et al., 2008). However, the differences in thermal tolerance existing between saAANAT2_IA and the VA mutant on the one hand, and saAANAT2_VV and omAANAT2_IV on the other hand, suggest that the 114 position also plays a role. Here also the remarkable thermal tolerance of mutant VA seems due to a synergistic effect of both mutations. It is interesting to note that a similar pattern of thermal tolerance has recently been highlighted in O. sifontesi (Cazaméa-Catalan et al., 2012), which displays a Val at both the corresponding 114 and 160 positions, as is the case in saAANAT2_VV. However, as O. sifontesi AANAT2 also presents other mutations, the direct comparison and interpretation of the structures is difficult.

This study brings new information on the melatonin system of teleost fish. One novel and interesting finding of this study is that one pineal gland can expresses two forms of AANAT2. These forms differ by only two amino acid residues in the Arctic charr and show very high identity with trout AANAT2, from which they differ by two residues in the same positions. This indicates a high degree of conservation among salmonids. The two amino acid positions seemed to play a role in thermal stability and catalytic efficiency, but the functional significance of the presence of two forms in the Artic charr remains to be elucidated. This study also shows that the charr pineal gland responds directly to temperature in terms of melatonin secretion. The response was similar to the response obtained with the recombinant AANAT2 VV variant, providing the temperature challenge was of short duration and followed by a recovery in the cold. In contrast, continuous long-term incubations at different temperatures shifted the peak of the response curve towards cold temperatures. This would indicate that the mechanisms of response to temperature are complex and reflect more than just molecular adaptations of the AANAT2 protein; i.e. in addition to these molecular adaptations some cellular mechanisms are probably involved, which transduce the temperature effect and regulate melatonin secretion. One target might be the cyclic AMP-dependent pathway (Thibault et al., 1993). Previous studies have shown that the dark induced rise in cyclic AMP induces AANAT phosphorylation through the RRNH and RRNS sequences at both ends of the molecule (present in the two AANAT2 Arctic charr isoforms); this allows binding of the enzyme to a chaperone protein (14-3-3), providing higher stability and preventing degradation through the proteasome (Falcón et al., 2001; Ganguly et al., 2001; Gastel et al., 1998; Klein et al., 2003). In fish, temperature could impact AANAT2 activity directly and indirectly. The direct effects would result from the intrinsic properties of the enzyme, as shown in this report. The indirect effects would involve modulation of cyclic AMP production (Benyassi et al., 2000) and consequently of AANAT2 phosphorylation and stability.

Our thanks to Dr Jo J. Aarseth, who helped us with the melatonin experiment.

 
• AANAT

  arylalkylamine N-acetyltransferase

 
• AcCoA

  acetyl coenzyme A

 
• BSA

  bovine serum albumin

 
• CI

  confidence interval

 
• GST

  glutathione S-transferase

 
• HIOMT

  hydroxyindole-O-methyltransferase

 
• k_cat

  turnover number

 
• k_cat/K_M

  catalytic efficiency

 
• K_M

  Michaelis constant

 
• ORF

  open reading frame

 
• PBS

  phosphate buffered saline

 
• TBS

  tris buffered saline

 
• V_max

  maximum velocity