Glycosylated Hydroxytryptophan in a Mussel Adhesive Protein from Perna viridis*

The 3,4-dihydroxyphenyl-l-alanine (Dopa)-containing proteins of mussel byssus play a critical role in wet adhesion and have inspired versatile new synthetic strategies for adhesives and coatings. Apparently, however, not all mussel adhesive proteins are beholden to Dopa chemistry. The cDNA-deduced sequence of Pvfp-1, a highly aromatic and redox active byssal coating protein in the green mussel Perna viridis, suggests that Dopa may be replaced by a post-translational modification of tryptophan. The N-terminal tryptophan-rich domain of Pvfp-1 contains 42 decapeptide repeats with the consensus sequences ATPKPW1TAW2K and APPPAW1TAW2K. A small collagen domain (18 Gly-X-Y repeats) is also present. Tandem mass spectrometry of isolated tryptic decapeptides has detected both C2-hexosylated tryptophan (W1) and C2-hexosylated hydroxytryptophan (W2), the latter of which is redox active. The UV absorbance spectrum of W2 is consistent with 7-hydroxytryptophan, which represents an intriguing new theme for bioinspired opportunistic wet adhesion.


MATERIALS AND METHODS
Protein Isolation from Mussel Feet-Green mussels (P. viridis) were collected from Tampa harbor in Florida. The feet were severed from about 100 freshly shucked mussels and stored at Ϫ80°C. Protein extraction from mussel feet was adapted from a previous report (12). Frozen mussel feet (in lots of 10 g of wet weight) were thawed and depigmented by scraping with a scalpel. The depigmented feet were homogenized in a glass tissue grinder with 50 ml of 5% acetic acid and two protease inhibitors (pepstatin A and leupeptin, both 30 mM). The homogenate was centrifuged (15,000 ϫ g, 4°C, 30 min), and the recovered supernatant (S1) was chilled in an ice bath. 70% perchloric acid was added dropwise with stirring to a final perchloric acid concentration of 1.4% (v/v). The mixture was further stirred and centrifuged (15,000 ϫ g, 4°C) for 30 min. The supernatant (S2) was collected, dialyzed in 500 volumes of 5% acetic acid for 12 h at 4°C, freeze-dried at Ϫ80°C, and resuspended in 0.2 ml of 5% acetic acid.
P. viridis foot protein-1 (Pvfp-1) was purified in three stages: gel filtration, reversed phase HPLC, and finally, again, gel filtration. Initial isolation was achieved by gel filtration chromatography on a Shodex KW-803 column (5 m, 8 ϫ 300 mm). The column was equilibrated and eluted with 5% acetic acid at a flow rate of 0.2 ml/min. A maximum volume of 200 l of S2 was loaded onto the Shodex column/run. The eluted volume was monitored at 280 nm, and the fractions under the Pvfp-1 peak were pooled (1-1.5 ml) and further resolved by C8 HPLC (Brownlee Aquapore RP-300, 7 m, 4.6 ϫ 250 mm) at a flow rate of 1 ml/min with an acetonitrile gradient described by Ref. 12. The collected fractions were assayed by redox cycling after acid urea gel electrophoresis (12) and amino acid analysis as described below to identify Pvfp-1 containing fractions. Pvfp-1 fractions were pooled and lyophilized at Ϫ80°C. After resuspending in 200 l of 5% acetic acid, Pvfp-1-positive fractions were run once more on Shodex KW-803 (above conditions). Fractions eluting between 33 and 43 min were redox active and had amino acid compositions consistent with previous studies (12). Purified Pvfp-1 fractions were recovered by freeze-drying. The N terminus of purified Pvfp-1 was sequenced by automated Edman degradation on a microsequencer (model 2090; Porton Instruments, Tarzana, CA) with a modified gradient program (14).
Mass Spectrometry-The mass of intact Pvfp-1 was determined by matrix-assisted laser desorption and ionization with time-of-flight (MALDI-TOF) mass spectrometry (Voyager DE, AB Biosystems, Foster City, CA) in the positive ion mode with delayed extraction. The MALDI matrix was prepared by dissolving sinapinic acid (10 mg/ml) in 30 vol % acetonitrile. Purified Pvfp-1 was dissolved in this matrix solution to give a final concentration between 1 and 10 pmol/l. About 1 l of sample was then spotted to the target plate and allowed to dry under vacuum (500 microns). Sample spots were irradiated at 337 nm using an N 2 laser (LSI, Inc., Cambridge, MA) with a pulse width of 8 ns and a frequency of 5 Hz. Soft ionization generated singly, doubly, and triply protonated ions of the Pvfp-1 protein variants. The singly and doubly protonated ions of bovine serum albumin were used as molecular mass calibrants at m/z 66,430 and 33,215, respectively).
Cloning and Sequencing Pvfp-1 cDNA-To obtain the partial cDNA sequence of Pvfp-1, total RNA was first isolated from the phenol gland at the tip of P. viridis foot tissue using an RNase plant mini kit from Qiagen. A single freshly dissected foot was pulverized under liquid nitrogen using a mortar and pestle, after which the mini kit protocol was adopted verbatim. Following that, the first strand cDNA was synthesized using Superscript II reverse transcriptase at 50°C with an Adapter primer, 5Ј-GGCCACGCGTCG ACTAGTACT(T) 16 -3Ј (Invitrogen). The product of the reverse-transcribed reaction was used as a template for subsequent PCRs.
Assuming repetition of the known trypsinized internal peptide sequence APPPAX 1 TAX 2 K (where X denotes an unknown amino acid residues) from a previous study (12), a degenerate oligonucleotide (sense, 5Ј-AARGCNCCNCCNCCNGC-3Ј, which corresponds to the sequence KAPPPA) was designed and combined with an abridged universal amplification primer (antisense, 5Ј-GGCCACGCGTCGACTAGTAC-3Ј; Invitrogen) to amplify the 3Ј end of Pvfp-1 from reverse transcriptase templates. The cloned 3Ј cDNA end provided the C-terminal sequence of Pvfp-1 along with some 3Ј-untranslated region.
Once the cDNA sequences encoding the mature Pvfp-1s were cloned, a GeneRacer kit (Invitrogen) was used to obtain the 5Ј end untranslated region of the cDNA of Pvfp-1 from full-length transcripts by 5Ј rapid amplification to cDNA ends. PCR was performed with gene-specific primer (antisense, 5Ј-GCTT-TCCATGCAGTCCATGCAGG-TGGATG-3Ј (corresponding to HPPAWTAWK) coupled with a GeneRacer 5Ј sense primer from Invitrogen. Generally, PCR was carried out in 25 l of 1ϫ Buffer B (Fisher) and 5 pmol of each primer, 5 mol of each dNTP, 1 l of first strand reaction, and 2.5 units of Taq DNA polymerase (Fisher) for 35 cycles on a Robocycler (Stratagene). Each cycle consisted of 30 s at 94°C, 30 s at 50°C, and 1 min at 72°C, with a final extension of 15 min. The PCR products were subjected to 1% agarose gel electrophoresis, purified, and cloned into a PCR TA vector (TOPO TA cloning kit; Invitrogen) and transformed into competent "Top10" cells (Invitrogen) for amplification, purification, and sequencing.  Byssal Thread Analysis-Fifteen to twenty threads were severed from the proximal end of each byssal stem, washed in 500 volumes of MilliQ water, and blotted dry. The dry threads were weighed and hydrolyzed by one of two methods: 6 M HCl for all amino acids except for Trp and 4 M NaOH for Trp (15). Acid hydrolysis was done with reusable hydrolysis vials (Pierce) at 110°C for 24 h and flash evaporated under high vacuum, whereas NaOH hydrolysates were titrated to pH 3-4 with glacial acetic acid. Purification of Trp-derived amino acids was achieved by C-18 HPLC (Brownlee Aquapore A OD-300 , 7 m, 4.6 ϫ 250 mm) over an acetonitrile gradient of 0 -10%. The eluate was monitored continuously at 220 and 280 nm, and all of those peaks absorbing at both wavelengths were collected for analysis by electrospray ionization and tandem mass spectrometry.
For routine amino acid analysis, the purified Pvfp-1s were hydrolyzed in one of three ways: 1) 4 M methanesulfonic acid with 0.5% 3-methyl-indole (16), 2) 6 M HCl with 5% phenol in vacuo at 110°C for 24 h, or 3) 4 M NaOH in vacuo at 110°C for 24 h (15). The first method was dropped because of extremely low tryptophan derivative recovery. The HCl hydrolysate was flash evaporated at 50°C under vacuum and shaken to dryness with Յ0.5 ml of MilliQ water followed by methanol; NaOH hydrolysates were titrated to pH 4. Amino acid analysis was performed according to conditions described earlier with a Beckman System 6300 auto analyzer (14) on which an authentic C 2 -(␣-D-mannopyranosyl)tryptophan (C-ManTrp) standard (17) eluted just after Met with a run time of 31.5 min.
Trp-containing Peptides-Preparation of tryptic peptides from Pvfp-1, including purification by C-18 HPLC (Brownlee, A 300 , 4.6 ϫ 260 mm) was done exactly as described by Ohkawa et al. (12). Proteolysis was stopped by acidification to pH 4 with glacial acetic acid before HPLC. Three peak fractions corresponding to peptides d, e, and j described by Ohkawa were selected for analysis by electrospray ionization mass spectrometry followed by tandem mass spectrometry with collision-induced decomposition using the PE Sciex QStar quadrupole/time-of-flight tandem mass spectrometer (PerkinElmer Life Science) in the University of California at Santa Barbara Mass Spectrometry Facility. Interpretation of fragments produced from the peptides by collision-induced decomposition was assisted by comparison with an authentic C-ManTrp standard and by mock fragmentation of various model sequences using Protein Prospector, version 5.1.4, on ExPASy Tools.

RESULTS
A previous investigation by Ohkawa et al. (12) reported the following attributes for Pvfp-1: 1) an apparent mass of 89 kDa based on SDS-PAGE, 2) marked quinone-like redox cycling activity without Dopa, 3) significant carbohydrate content, particularly with respect to mannose, N-acetylglucosamine and fucose, and 4) an Edman-derived primary sequence dominated by two closely related consensus repeats: AOOOAX 1 TAX 2 K and APOKOX 1 TAX 2 K, in which O denotes trans-4-hydroxyproline. The X 1 and/or X 2 positions were consistent with the presence of an aromatic amino acid but could not be reconciled with any known modification of Tyr or Dopa.
In the present investigation, purified Pvfp-1 was shown to consist of at least two variants (supplemental Fig. S1) with masses of 50 and 64 kDa observed by MALDI-TOF mass spectrometry (supplemental Fig.  S2). The N-terminal sequence of the isolated protein was found to be (A)VY*HPPSX 1 TAX 2 IAOK (supplemental Fig. S1), where X 1 and X 2 represent Edman cycles devoid of detectable phenylthiohydantoin derivatives. Y* denotes Dopa and was the only Dopa detected in Pvfp-1.
To further explore the chemistry of X 1 and X 2 , we first deduced the complete sequence of Pvfp-1 from its corresponding cDNA prepared by standard cloning procedures. Two Pvfp-1 variants, probably related by alternative splicing, were found with the shorter differing from the longer by exactly 15 decapeptide repeats. In the longer variant 1 (calculated mass, 61 kDa), APPPAWTAWK and ATPKP-WTAWK consensus repeats occur 14 and 29 times, respectively ( Fig. 2  and supplemental Fig. S3). Other intriguing features in Pvfp-1 are a distinct collagen-like sequence with 18 tripeptide Gly-X-Y repeats, followed by a short interval (14 amino acids long), before returning to a Trp-rich C-terminal sequence. It must therefore be concluded that in Pvfp-1 some modification of tryptophan, not tyrosine, provides the basis for the unknown aromatic amino acids in the decapeptide repeats.
Because tryptophan is unstable to hydrolysis by HCl, purified Pvfp-1 and P. viridis byssal threads were hydrolyzed with 4 M NaOH. After hydrolysis, component aromatic amino acids were chromatographically separated by C-18 HPLC (Fig. 3) and analyzed by mass spectrometry to help deduce their relationship to tryptophan. Two ions with masses of 367 and 529 Da eluting between 22 and 24 min were detected; the latter decomposed readily to 367 through a loss of 162 Da (hexose) (supplemental Fig. S4), suggesting a cleavage that is typical of O-and N-linked hexoses (20,21). The 367-Da ion and its fragmentation, most notably the 120-Da loss to m/z 247, are identical with standard C 2 -[␣-D-mannopyranosyl]-tryptophan (22,23) (Fig. 4). The hexose linked to Trp in Pvfp-1 is likely to be mannose, but tandem mass spectrometry by itself is unable to distinguish different hexoses. Because C-ManTrp elutes just after Met in amino acid analysis, this method was used to estimate a C 2 -mannosyltryptophan content of 1-2 dry weight % in byssal threads and 3-4 dry weight % in Pvfp-1 following base hydrolysis (supplemental Table S1). Only trace levels of Trp could be detected in Pvfp-1; thus the Trp in byssus must come from other proteins.
To detect modified tryptophans more directly in the decapeptide sequences of Pvfp-1, we used tandem mass spectrometry with collision-induced decomposition of three peptides following trypsin digestion of Pvfp-1 (supplemental Fig. S5). Peptides d and e were selected because 75% of their sequences had already been established by Edman chemistry (12). The doubly charged parent ion (Mϩ2H) 2ϩ (m/z 953.8) for the tryptic decapeptide d readily decomposes to another ion (Mϩ2H) 2ϩ (m/z 771.25) after a neutral loss of 162 to m/z 873 (hexose) followed by another neutral loss of 204 to m/z 771 (N-acetyl hexose) (Fig. 5A). The oxonium ion of hexosyl-N-acetylhexose is evident at m/z 366. The 771.25 peak loses water to become m/z 762, which then undergoes the 120 Da (60 ϫ 2) neutral loss typical of C 2 -mannosylation; yЉ 6 or OW 1 TAW 2 K (m/z 1144) does the same. The internal fragment, OW 1 TA (m/z 616.1), indicates a 366-Da mass for X 1 (Fig. 5B). Notably, the 511.1 (yЉ 2 ) and 290 (yЉ 3 2ϩ -2H) ions corresponding to W 2 K and AW 2 K are larger than the comparable W 1 -containing fragments by 16 Da. C-ManTrp ϩ 16 is consistent with C 2 -hexosylated hydroxytryptophan (W 2 K) and the quinoid counterpart of C 2 -hexosylated hydroxytryptophan (AW 2 K) (Fig.  5B). A complete annotation of fragment masses is given in supplemental Fig. S6.
Mannose is plausible as the C 2 -hexose given that WX 1 X 2 W is a well accepted sequence signature for C 2 -mannosylation of tryptophan in proteins (20, 24 -26). The precise ring assignment of the hydroxyl group in hydroxytryptophan (OHTrp) is not possible from electrospray ionization tandem mass spectrometry, but a phenyl ring placement is consistent with the loss of two hydrogens to form a doubly protonated quinoid ion at m/z 290 and by the strong quinone-like redox cycling activity in Pvfp-1 and tryptic decapeptides (12).
Peptide e (m/z 757) is less glycosylated than peptide d but shows some trends similar to those in the 771 ion during fragmentation (Fig. 6). The yЉ 8 ion (OPAW 1 TAW 2 K) at m/z 1328 shows a neutral loss of 120 Da. The fragment ions implicate both W 1 and W 2 as C 2 -[␣-D-mannopyranosyl]-hydroxytryptophan; the yЉ 2 (W 2 K) and yЉ 4 ions (TAW 2 K) are particularly suggestive of W 2 ( Fig. 6; complete fragment annotation in supplemental Fig. S7). The fragmentation of peptide j ( Fig. 7 and supplemental Fig.  S8) was undertaken to corroborate the odd collagen-like primary structure predicted by the cDNA-deduced Pvfp-1 sequence (Fig. 2). Peptide j is 14 residues long, and its sequence corresponds to residues 496 -509 with Gly at every third residue and three of the five prolines converted to trans-4-hydroxyproline.
Neither hydroxytryptophan nor its C-hexosylated forms were detected after methansulfonic acid or NaOH hydrolysis of the byssus or Pvfp-1. We observed that whereas standard 5-or 7-hydroxytryptophan were stable to hydrolysis in methansulfonic acid, standard C-ManTrp was not. OHTrp standards hydrolyzed in the presence of sugars were not detected by amino acid composition. C 2 -[␣-D-mannopyranosyl]-hydroxytryptophan would thus be unlikely to survive methansulfonic acid. OHTrp recovery following alkaline hydrolysis is prevented by the formation of quinonimine at alkaline pH (27).
Tandem mass spectrometry supports the presence of OHTrp but is unable to specify the position of the hydroxy group. To further clarify this important point, Pvfp-1 and peptide e were reacted with nitrosonaphthol, which is specific for the 5-hydroxy position on the indole (18); both were negative. In addition, the UV absorbance spectrum of peptide e at pH 3.5 was compared with a series of related indole derivatives (28,29). In contrast to L-tryptophan and C-Man-Trp, peptide e has a max at 269 nm shifted to a lower wavelength ( Fig. 8 and supplemental Fig. S9), and an extinction coefficient of 18,000 M Ϫ1 cm Ϫ1 at 269 nm that is consistent with the presence of two modified tryptophans enhanced by mannosylation (28). Of the four hydroxy-indole derivatives tested, only 7-hydroxytryptophan shared the same max values as peptide e. Indeed, the two spectra are nearly superimposable between 250 and 300 nm. Although future studies need to further characterize the tryptophan chemistry of peptide e by proton NMR and electrochemistry, the identification W 2 as C 2 hexosyl-7hydroxy tryptophan in Pvfp-1 appears reasonable. A final test of hydroxy-indole reactivity with Arnow's reagent showed that only 7-hydroxy-tryptophan produced the same bright yellow product formed by Pvfp-1 (supplemental Fig. S9).

DISCUSSION
Pvfp-1 resembles other mussel coating proteins in its canonical, Pro-and Lys-rich repeats. It diverges significantly, however, in its reliance on modifications of tryptophan rather than tyrosine for its redox chemistry. Pvfp-1 consists of two closely related variants, the larger of which has predicted and observed masses of 61 and 64 kDa, respectively. Both variants show a highly repetitive sequence with four distinct domains: a Trprich decapeptide repeat (consensus: APPPAWTAWK and  ATPKPWTAWK) domain largely in the N-terminal twothirds, a collagen domain (Gly-X-Y repeats), a short hinge (residues 450 -459), and a predicted, short coiled-coil region before returning to Trp-rich repeats at the C terminus (Fig. 9). The collagen domain is reminiscent of other collagen-like proteins such as the macrophage scavenger receptor (30), Torpedo acetylcholinesterase (31), and complement C1q (32), in which a small collagen domain directs trimer formation. The proposed trimeric structure remains speculative for Pvfp-1 because only the single chain mass has been determined with any accuracy by MALDI-TOF.
Pvfp-1 exhibits extensive posttranslational decoration. Pro is targeted for hydroxylation throughout both decapeptide consensus repeats and in the collagen domains; Thr-2 in ATPKPWTAWK appears to be O-glycosylated with O-(N-acetyl)hexosyl-hexose, which agrees with the high levels of N-acetyl-glucosamine reported earlier (12) but differs from the previous designation of Pro for position 2 in peptide e. Possibly, the phenylthiohydantoin derivative of glycosylated Thr has the same elution time as Pro-phenylthiohydantoin on C-18 HPLC. Notably, Trp undergoes C 2 -hexosylation (probably by mannose) and hydroxylation. Whereas W 1 and W 2 are always hexosylated, W 2 seems favored for hydroxylation. Peptides d and e may be representative of other decapeptides in Pvfp-1 with respect to C-hexosylation and hydroxylation. At this stage, however, aside from ample evidence for 120-Da mass losses, i.e. C-hexosyl-Trp in other peptides, collision-induced decomposition of many peptides remains too complex for confident interpretation. Pvfp-1 has only a single Dopa located near the N terminus.
C-linked mannosylation of proteins is an unusual but not unprecedented modification of tryptophan. It was first reported in pancreatic ribonuclease and since then in over 47 other proteins including notably thrombospondin and complement proteins with the sequence motif WX 1 X 2 W (20,(22)(23)(24)(25)(26). Although the function of this modification remains elusive, mannosylation is known to render the tryptophan more polar and solvent-accessible (23). In complement proteins, C-ManTrp in the thrombospondin-like repeat domains has been proposed to have an adhesive function (33). The potential presence of up to 80 C-ManTrp residues in Pvfp-1, which functions as an adhesive and coating in byssus (9 -11), begs the hypothesis that manno- sylation makes tryptophan behave more like Dopa: sticky and prone to cross-link formation. This conjecture, however, is too simple because tryptophan needs to be hydroxylated before becoming redox active like Dopa. 5-Hydroxytryptophan in engineered proteins and as a free amino acid was reported to undergo oxidation resulting in the formation of p-quinonimine and di-Trp cross-links (27,34). Indeed, based on these studies, we expected to find C 2 -mannosylated 5-HOTrp in peptide e and Pvfp-1. UV spectra, however, support the rarer 7-OH rather than the 5-OH isomer and represent the first report of naturally occurring 7-hydroxytryptophan in proteins (35). 7-OHTrp seems a better mimic of Dopa than the other isomers. Upon oxidation, it forms an o-quinonimine (Fig. 10A), and the indolic N-H and phenolic OH groups are close enough to chelate metal ions (Fig. 10A), a common capability of mussel byssus (9,36).
If 7-hydroxylation makes Trp more like Dopa, why bother with mannosylation? Trp mannosylation was observed to block cleavage in the vicinity of modified residues by exo-and endoproteases, hence rendering proteins more resistant to degradation (28). In addition, Trp-rich proteins are quite hydrophobic and thus prone to aggregation in solution (37). C-mannosylation of Trp residues appears to make them fully water accessible while preventing aggregation (38).
In summary, the green mussel P. viridis, employs an intriguing alternative to Dopa in Pvfp-1, one of its byssal adhesive proteins. Although much bulkier, C-Man-7-OHTrp resembles Dopa in having attributes that contribute to both cohesive and adsorptive interactions necessary for adhesion (39).
With respect to cohesion, the two-electron oxidation of C-Man-7-OHTrp to an o-quinonimine (Fig. 10A) mimics Dopa-o-quinone, which is known to form covalent cross-links with other Dopa, cysteine, lysine, and histidine residues (Fig.  10B) (7,40). The coordination of metal ions via the o-hydroxyl groups in Dopa also contributes to cohesion, particularly in the byssal cuticle of Mytilus galloprovincialis (9). The ortho-placement of indolic-N and phenolic OH groups in C-Man-7-OHTrp seems well suited to metal ion chelation, but this has yet to be investigated.
The stickiness of adhesive molecules is determined by their adsorptive tendencies. Dopa complexation of metal hydroxides on surfaces provides strong and reversible interactions (Fig.  10B) (41). Again, given the ortho-placement of electronegative elements, such interactions also seem likely with C-Man-7-OHTrp. Future studies need to closely examine what adaptive advantages C-Man-7-OHTrp offers over Dopa.