Aspolin, a Novel Extremely Aspartic Acid-rich Protein in Fish Muscle, Promotes Iron-mediated Demethylation of Trimethylamine-N-oxide*

Trimethylamine-N-oxide (TMAO) is abundant in marine fish. Formaldehyde synthesis by TMAO demethylation during storage markedly deteriorates fish meat. In the present work, we cloned the extremely aspartic acid-rich proteins from skeletal muscle of a commercially important species, walleye pollack, in the course of molecular identification of trimethylamine-N-oxide demethylase (TMAOase). One of the cDNAs, designated as aspolin1, encodes an extremely aspartic acid-rich protein of 228 amino acids which is converted to the TMAOase after processing between Ala42 and Asp43. Mature aspolin1/TMAOase protein contains 179 Asp in 186 total amino acids. The other cDNA, designated as aspolin2, has a common nucleotide sequence with aspolin1 in the 5′ part and encodes a protein which has an additional Asp polymer and a C-terminal cysteine-rich region. The amino acid sequence of the C-terminal cysteine-rich region of aspolin2 is highly homologous to the mammalian histidine-rich Ca2+-binding protein. Aspolin1/TMAOase and aspolin2 mRNA was most abundant in the skeletal muscle. A lower level of the mRNA was also detected in kidney, heart, spleen, and brain. Synthetic Asp polymer showed marked TMAOase activity in the presence of Fe2+, whereas a monomer and oligomers did not. Purified TMAOase protein bound to Fe2+ with low affinity, which may be responsible for the catalytic activity. Poly aspartic acid-Fe2+ complex generated after death would be involved in formaldehyde synthesis by the demethylation of TMAO during the storage of fish meat.

The occurrence of formaldehyde markedly deteriorates the fish meat by the crosslinking of muscle protein. As a result, fish fillet loses elasticity and develops a sponge-like texture that is unsuitable for surimi products (seafood items that are made of fish-meat gel and sometimes look like crab, scallop, etc.) owing to the decrease in the solubility of myofibril at high-salt concentrations (5,6).
TMAO demethylase (TMAOase, or trimethylamine-N-oxide formaldehyde-lyase, EC 4.1.2.32) has been proposed to be responsible for the formaldehyde synthesis by TMAO demethylation during frozen storage of fish meat (3). TMAOase activity has been detected in kidney, spleen, pyloric caecum, liver, dark muscle, and ordinary muscle of gadoid and non-gadoid fish (3). Some cofactor systems have been developed for the separation of the enzyme; flavin-NAD(P)H and Fe 2ϩ -ascorbatecysteine with or without methylene blue under aerobic or anaerobic conditions (3). Despite the efforts to purify the enzyme, it has not been possible to identify the detailed component(s) of the enzyme, primarily because of the insolubility and the complex cofactor requirement of this enzyme (7)(8)(9).
Walleye pollack is caught in the northern Pacific and is the most important source of surimi products. Recently, TMAOase was detected and purified from the myofibrillar fraction of walleye pollack (10,11). The walleye pollack enzyme was characterized as a single acidic protein with an apparent molecular mass of 25 kDa. This enzyme required Fe 2ϩ for its activity and showed an optimum pH at 7.0; the K m value for TMAO was 30 mM. The activity of this enzyme was stable in the presence of SDS. Furthermore, the enzyme activity was retained after heating at 80°C for 30 min, suggesting an extremely high thermal stability (12).
In this paper, we report the novel extremely aspartic acidrich proteins found in the course of the molecular identification of walleye pollack muscle TMAOase. Our cDNA cloning revealed that this enzyme is one of the two homologous proteins containing the extremely long Asp polymer, an uninterrupted stretch (except for a single Glu residue) of 171 * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AB117517 (aspolin1) and AB117518 (aspolin2).
Asp residues. We demonstrate that the TMAO demethylation activity of this protein arises from the interaction of poly Asp and Fe 2ϩ .

EXPERIMENTAL PROCEDURES
Materials-Walleye pollack Theragra chalcogramma caught off the northern Hakodate coast was used for protein and RNA preparation. Aspartic acid monomer, oligomer, and polymer were obtained from Sigma. Poly glutamic acid was obtained from ICN. TMAO was obtained from Aldrich.
Purification and Analysis of Walleye Pollack TMAOase-The TMAOase was purified from the myofibrillar fraction of walleye pollack skeletal muscle according to the method reported previously (11). After dialysis against distilled water, the protein was lyophilized. Phosphate was detected by staining with Coomassie Brilliant Blue and aluminum nitrate (13). Carbohydrate was detected by the phenol-sulfuric acid method, periodic acid Schiff staining, or staining with a GelCode glycoprotein staining kit (Pierce). Amino acid composition analysis of the purified enzyme was performed by o-phthalaldehyde post-column labeling method using high pressure liquid chromatography. The aminoterminal amino acid sequence was analyzed with a Hewlett-Packard G1005A protein sequencing system. Molecular mass of the purified protein was measured by electrospray-ionization mass spectrometry using a Sciex API300 triple quadrupole mass spectrometer (PerkinElmer Life Sciences).
cDNA Cloning and Sequencing-Poly(A) ϩ RNA was isolated from walleye pollack skeletal muscle using Quik-Prep micro mRNA purification kit (Amersham Biosciences). Double-strand cDNA was synthesized by using the SuperScript system (Invitrogen) and ligated to Ex-Cell vector (Amersham Biosciences). An oligonucleotide corresponding to a stretch of Asp hexamer (5Ј-GAYGAYGAYGAYGAYGAYG-3Ј) was 5Ј end-labeled with [␥-32 P]ATP and used for plaque hybridization. Positive phage clones were excised according to the manufacturer's instructions by using an Escherichia coli NP66 in vivo release system. The clones in pExCell plasmid were sequenced with an ABI 3700 automated DNA sequencer.
Measurement of TMAO Demethylation Activity-For measurement of TMAO demethylation activity at room temperature, the method of Kimura et al. (11) was used with slight modification. Peptides or amino acids were mixed with 10 mM Tris acetate, pH 7.0, containing 20 mM TMAO and 0.2 mM FeCl 2 . The reaction was terminated by the addition of 5% trichloroacetic acid. DMA in this solution was quantified by the copper-dithiocarbamate procedure (14). Peptide content was determined by the biuret method using bovine serum albumin as a standard. For the measurement of TMAO demethylation activity in the frozen state, 20 mM HEPES-NaOH, pH 6.9, containing 10 mM TMAO and 0.2 mM FeCl 2 , was used. After thawing the reaction mixture on ice, DMA was immediately quantified as described above.
Iron Binding Assay-Purified protein (15 g/ml) was mixed with 20 -500 M FeCl 2 in 10 mM Tris acetate buffer, pH 7.0, containing 5 mM ascorbic acid and placed at 25°C for 30 min. Free iron was separated by ultrafiltration using a Microcon-3 column (Millipore). Iron concentration of the initial mixture and the through fraction of ultrafiltration were measured by a bathophenanthrolinedisulfonic acid method using an Fe test kit (Wako). Bound iron was calculated by subtraction of filtrated iron from the initial iron content.

Molecular Properties of Walleye Pollack
TMAOase-To investigate the primary structure of purified TMAOase, the N-terminal amino acid sequence was analyzed. Fifteen amino acids from the N terminus were determined as DDDDDDDDDAGD-DDD (Table I). Because the purified protein was resistant to digestion with trypsin and lysylendopeptidase, attempts to analyze internal amino acid sequence were unsuccessful. The results of the N-terminal amino acid sequencing indicate the processing of the protein after translation. Amino acid composition analysis revealed that an ϳ90% molar ratio of amino acids of purified protein was Asp or Asn (Table II). As for other amino acids, His, Glu or Gln, Gly, Ala, Leu, and Lys were also detected at Ն 0.4% molar ratio. Ser, Thr, Pro, Val, Ile, and Arg were detected at trace levels. Cys, Met, Tyr, and Phe were not detected at all. To estimate the number of each amino acid in a single protein molecule, the molecular mass of the protein was measured by mass spectrometry. As shown in Table I, the molecular weight of the purified protein was 21,389, which is a little smaller than that estimated from SDS-PAGE (11). On the other hand, neither phosphate nor carbohydrate was detected in the purified protein (data not shown). Because it seems that the purified protein is not subject to major post-translational modification other than the processing of peptide, the number of amino acids in the protein can be estimated. Based on the molecular weight of the aspartic acid residue (115) and the purified protein, about 180 amino acids are contained in the  purified protein. According to this assumption, the number of each amino acid was estimated as shown in Table II. cDNA Cloning of Walleye Pollack TMAOase-The results of amino acid sequencing and composition analysis suggested that a large part of the amino acid sequence of the protein contains an Asp repeat. Therefore, we carried out the screening of the cDNA library of walleye pollack skeletal muscle using an oligonucleotide probe corresponding to the amino acid sequence, DDDDDD. As a result, two species of clone, designated as aspolin1 and aspolin2, containing open reading frames encoding extremely Asp-rich proteins, were obtained. To confirm the number of GAY repeats in these clones, RT-PCR was performed, and amplified products were cloned and sequenced (data not shown). Finally, the nucleotide sequence of two species of cDNA was determined (Fig. 1). As shown in Fig. 1, the 5Ј noncoding region and the former part of coding region of two clones have a common nucleotide sequence. Deduced amino acid sequences of aspolin1 and aspolin2 are aligned in Fig. 2. Amino acid sequences of aspolin1 and aspolin2 contain 182 and 225 amino acids of Asp polymer, respectively, with a few inserts of other amino acids.
The N-terminal amino acid sequence of purified protein corresponds to the peptide stretch from Asp 43 to Asp 57 in aspolin1 and aspolin2. Assuming that the processing of the peptide occurs between Ala 42 and Asp 43 , the molecular weight of the mature aspolin1 protein is calculated as 21,383 (Table I). This value is almost identical to the molecular weight of the purified protein obtained by mass spectrometry, 21,389. Furthermore, the number of each amino acid in mature aspolin1 is almost identical with that estimated from amino acid analysis,  (Table II). Thus, we concluded that aspolin1 encodes the precursor of the purified walleye pollack TMAOase.
Primary Structure of Aspolin1/pre-TMAOase and Aspolin2-Proteins encoded by aspolin1/pre-TMAOase and aspolin2 cDNAs have 228 and 347 amino acids, respectively. It seems that both proteins contain 42 amino acids of pre-peptide at the N terminus. Mature aspolin1/TMAOase protein consists of a 182-amino acid stretch of Asp polymer and a C-terminal of 4 amino acids, His-Glu-Glu-Leu. In the Asp polymer region, there are three amino acids other than Asp, namely Ala 52 , Gly 53 , and Glu 81 . The deduced amino acid sequence of aspolin1/ pre-TMAOase did not show homology to any known proteins in the data base. On the other hand, putative mature aspolin2 protein consists of a 225 amino acid stretch of Asp polymer and a C-terminal cysteine-rich region. The Asp polymer region of aspolin2 contains Ala 52 , Gly 53 , and Glu 81 , as well as aspolin1, and a His-Glu-Ala-Gly sequence at the amino acid position 225. The C-terminal region contains 14 Cys residues, and each of the Cys residues is spaced with 1-3 amino acid(s) other than Cys. The data base search revealed that the Cys-rich region of aspolin2 is highly homologous to mammalian histidine-rich calcium-binding protein (HRC) (Fig. 3; Ref. 15).
Tissue Distribution of Aspolin1/TMAOase and Aspolin2 mRNA-The tissue distribution of aspolin1/TMAOase and as-polin2 mRNAs was analyzed by RT-PCR (Fig. 4). Both aspo-lin1/TMAOase and aspolin2 mRNAs were most abundant in skeletal muscle. Moderate amounts of both mRNAs were detected in kidney, and traces of the mRNAs were also detected in heart, spleen, and brain. Neither mRNA was detected in liver, stomach, pyloric caecum, intestine, and fin. None of the amplified band was detected in the control without reverse transcription (data not shown).
TMAO Demethylation Activity of Poly Asp-To identify the region of mature aspolin1/TMAOase responsible for the TMAOase activity, we attempted to have aspolin1/TMAOase cDNA expressed in E. coli, COS-7 cells, and a wheat germ cell-free system. Probably because of the special characteristic of the primary structure of aspolin1/TMAOase, all of these trials were unsuccessful. Therefore, we analyzed the TMAO FIG. 4. Tissue distribution of the mRNA for aspolin1 and aspo-lin2. Total RNA was prepared from walleye pollack tissues, and RT-PCR was performed using specific primers designed at the 3Ј side of Asp polymer region of aspolins (see Fig. 1) and the coding region of actin. The expected sizes of the amplified bands of aspolin1, aspolin2, and actin were 970, 425, and 720 bp, respectively. demethylation activity of synthetic poly Asp. Surprisingly, synthetic poly Asp showed a marked TMAO demethylation activity. As shown in Fig. 5, the relative TMAO demethylation activity of poly Asp 8k was as much as that of purified TMAOase. The activity of poly Asp 35k was 1.7ϫ as much as that of purified TMAOase. To examine whether polymerization of Asp is required for TMAO demethylation activity, the activity of L-Asp monomer, dimer, and tetramer were measured. None of these small molecules, however, showed definite activity (Fig. 5). To examine whether carboxyl groups in the side chains of poly Asp are responsible for TMAO demethylation activity, the activity of poly Glu was measured. As shown in Fig. 5, poly Glu showed TMAO demethylation activity, although its level was relatively low.
Because TMAO demethylation in muscle of walleye pollack occurs during frozen storage, TMAO demethylation activity of poly Asp in the frozen state was measured. As shown in Fig. 6, poly Asp promoted TMAO demethylation both at Ϫ20°C and Ϫ50°C. The plateau level of DMA synthesis at Ϫ20°C was more than twice as high as that at Ϫ50°C.
Iron Binding of Purified TMAOase-A previous report (11) demonstrated that TMAO demethylation catalyzed by the walleye pollack TMAOase required an Fe 2ϩ as a cofactor. To examine the binding capacity of purified TMAOase to Fe 2ϩ , an iron binding assay was performed by an ultrafiltration-based method. Iron binding of purified TMAOase exhibited a sigmoidal dependence on Fe 2ϩ concentration (Fig. 7). The bound/free ratio of Fe 2ϩ was not more than 0.25. At the concentration of Fe 2ϩ used in the TMAO demethylation assay, bound Fe 2ϩ amounted to about 1 ⁄4 of the carboxyl groups of the Asp side chain of TMAOase. At higher concentrations of Fe 2ϩ , however, bound Fe 2ϩ continued to increase. DISCUSSION Nucleotide sequences of aspolin cDNAs contain long GAY repeats encoding poly Asp. In general, a correct analysis of such a repeat sequence is difficult because of the problem of sequencing chemistry. For this reason, we carefully analyzed nucleotide sequences to confirm the accuracy of the presented sequences. In the nucleotide sequences of the aspolin cDNAs, two kinds of codons for Asp, GAC, and GAU, appear randomly. Because this random mixture of GAC and GAU ensured the specificity of sequence and avoided slippage, the sequencing pattern of the GAY-repeat region was very stable (data not shown). On the other hand, cDNA in cloning vector and PCR product were rather unstable (data not shown). Therefore, we cloned the RT-PCR product containing the GAY region and sequenced several clones to confirm the correct sequences. As for aspolin1/TMAOase, the calculated molecular weight from deduced amino acid sequence and the actual value from mass spectrometry were almost identical (Table I), supporting that the idea that cDNA was derived from a functional mRNA.
Nucleotide sequences upstream from GAA, corresponding to Glu 226 , are common to aspolin1/pre-TMAOase and aspolin2 (Fig. 1). This sequence identity indicates that each mRNA is generated by alternative splicing. According to the N-terminal amino acid sequence analysis, aspolin1/pre-TMAOase preprotein is processed between Ala 42 and Asp 43 . If the processing of aspolin2 protein depends on the N-terminal amino acid sequence, aspolin2 is processed at the same position as aspolin1/pre-TMAOase.
The amino acid sequence of the C-terminal region of aspolin2 is very similar to that of mammalian HRC (Fig. 3). HRC is known as a Ca 2ϩ -binding protein of sarcoplasmic reticulum (16). Because aspolin2 mRNA is also abundant in skeletal muscle, aspolin2 may appear to be a fish homologue of HRC. However, the molecular mass of HRC, 165 kDa, is much higher than that of aspolin2, 35 kDa. Furthermore, HRC contains 9 tandem repeats of a 29-residue sequence consisting of a stretch of 10 -11 acidic amino acids and a histidine-rich sequence (15). Although both HRC and aspolin2 have C-terminal cysteinerich regions and are rich in acidic amino acids, the overall structures of these molecules are quite different.
Poly Asp in aspolin is reminiscent of another Ca 2ϩ -binding protein of sarcoplasmic reticulum, calsequestrin. Skeletal muscle calsequestrins have a cluster of 10 -45 acidic residues, typically aspartates, at C terminus (17)(18)(19)(20)(21). Calsequestrin binds Ca 2ϩ with high capacity (40 -50 Ca 2ϩ /molecule) and moderate affinity (22,23). Negatively charged carboxyl groups of acidic residues are thought to be responsible for Ca 2ϩ binding. A recent study (24) showed that binding of Ca 2ϩ is highly dependent on the C-terminal Asp-rich region. It has been proposed that the length of the C-terminal Asprich region may be important for the Ca 2ϩ binding capacity of calsequestrin (22). According to this report, mature aspolin1/ TMAOase and aspolin2 are expected to have high Ca 2ϩ binding capacity. Therefore, aspolins may be a Ca 2ϩ -binding protein in fish muscle.
Although no aspolin ortholog was found in the protein data base, a similar nucleotide sequence was found in the genomic data base of tiger puffer (data not shown). We also found some aspolin-like nucleotide sequences in amphibian expressed sequence tag (GenBank TM /EBI accession numbers BQ731479, BJ073162, and BQ524571). However, none of the aspolin-like sequences was found in sequences of the other groups of vertebrate. Because fish and amphibian are poikilothermic animals, aspolins may contribute to the adaptation to lower temperature.
The present paper shows that poly Asp is the real component of walleye pollack TMAOase. TMAO demethylation catalyzed by Fe 2ϩ and small ligands were studied in terms of rearrangement of detoxication in liver (25)(26)(27). In these reports, it is shown that Fe 2ϩ and various chelating agents with carboxyl group, such as oxalic acid, malic acid, and aspartic acid, catalyze the TMAO demethylation at 80°C. Although these chelating agents are comparable with poly Asp in the present study, the reaction with poly Asp proceeded under much milder condition. Therefore, polymerization of the carboxyl group is important for the catalytic activity under mild conditions. Ferris et al. (26) proposed the model of TMAO demethylation cata-lyzed by Fe 2ϩ chelated with oxalic acid and suggested the importance of the two adjacent nonchelated sites on the iron. As shown in Fig. 7, affinity of poly Asp for Fe 2ϩ was relatively low. Labile interaction between Fe 2ϩ and carboxyl groups of poly Asp may be necessary for the catalytic activity. This assumption is consistent with the high K m value of walleye pollack TMAOase reported previously (11).
Whether aspolins catalyze TMAO in vivo or not is unknown. However, aspolins are not likely to act as TMAO catalytic enzymes in vivo for several reasons. Because the circular dichroism spectrum of purified TMAOase at the neutral condition suggested the random coil structure (13), it is unlikely that TMAOase has a single effective active center. The low affinity of aspolin1/TMAOase for ferrous ion suggests that catalytic activity would be very low in vivo because the concentration of free ferrous ion in live cells is kept at a very low level, 0.2-0.5 M (28).
In the present study, TMAO demethylation with Fe 2ϩ and poly Asp occurred even in the frozen state at Ϫ20°C and Ϫ50°C. It has been reported (29) that iron-catalyzed dealkylation of tertiary amine oxides proceeds around Ϫ10°C. This reaction feature is consistent with the TMAO demethylation in the walleye pollack muscle during frozen storage (5).
In conclusion, we have identified novel extremely aspartic acid-rich proteins, aspolin1 and aspolin2, in fish skeletal muscle and demonstrated that poly Asp in these proteins promotes iron-mediated demethylation of TMAO under frozen conditions as well as at room temperature. Aspolins may act as calciumbinding proteins and are unlikely to be TMAO catalytic enzymes in vivo. According to these results, a possible scenario of formaldehyde synthesis in fish meat during frozen storage might occur as follows. After the death of the fish, free ferrous ions are generated by the degradation of heme and the release from iron-binding proteins like ferritin and transferrin. Then, ferrous ions bind to the poly Asp region of aspolin and other proteins and exhibit the TMAO demethylating activity. The prevention of the formaldehyde synthesis based on the presented mechanism can contribute to a longer shelf life and safety of gadoid flesh.