Mimivirus collagen is modified by bifunctional lysyl hydroxylase and glycosyltransferase enzyme

: Collagens, the most abundant proteins in animals, are modified by hydroxylation of proline and lysine residues and by glycosylation of hydroxylysine. Dedicated prolyl hydroxylase, lysyl hydroxylase, and collagen glycosyltransferase enzymes localized in the endoplasmic reticulum mediate these modifications prior to the formation of the collagen triple helix. Whereas collagen-like proteins have been described in some fungi, bacteria, and viruses, the post-translational machinery modifying collagens has never been described outside of animals. We demonstrate that the L230 open reading frame of the giant virus Acanthamoeba polyphaga mimivirus encodes an enzyme that has distinct lysyl hydroxylase and collagen glycosyltransferase domains. We show that mimivirus L230 is capable of hydroxylating lysine and glycosylating the resulting hydroxylysine residues in a native mimivirus collagen acceptor substrate. Whereas in animals from sponges to humans the transfer of galactose to hydroxylysine in collagen is conserved, the mimivirus L230 enzyme transfers glucose to hydroxylysine, thereby defining a novel type of collagen glycosylation in nature. The presence of hydroxylysine in mimivirus proteins was confirmed by amino acid analysis of mimivirus recovered from A. polyphaga cultures. This work shows for the first time that collagen post-translational modifications are not confined to the domains of life. The utilization of glucose instead of the galactose found throughout animals as well as a bifunctional enzyme rather than two separate enzymes may represent a parallel evolutionary track in collagen biology. These results suggest that giant viruses Results: a collagen lysyl-hydroxylase/ glucosyltransferase


SUMMARY
Collagens, the most abundant proteins in animals, are modified by hydroxylation of proline and lysine residues, and by glycosylation of hydroxylysine. Dedicated prolyl-, lysyl hydroxylase and collagen glycosyltransferase enzymes localized in the endoplasmic reticulum mediate these modifications prior to the formation of the collagen triple helix. Whereas collagen-like proteins have been described in some fungi, bacteria, and viruses, the post-translational machinery modifying collagens has never been described outside of animals. We demonstrate that the L230 open reading frame of the giant virus Acanthamoeba polyphaga Mimivirus encodes an enzyme including distinct lysyl hydroxylase and collagen glycosyltransferase domains. We show that Mimivirus L230 is capable of hydroxylating lysine and glycosylating the resulting hydroxylysine residues in a native Mimivirus collagen acceptor substrate. Whereas in animals from sponges to humans the transfer of galactose to hydroxylysine in collagen is conserved, the Mimivirus L230 enzyme transfers glucose to hydroxylysine, thereby defining a novel type of collagen glycosylation in nature. The presence of hydroxylysine in Mimivirus proteins was confirmed by amino acid analysis of Mimivirus recovered from Acanthamoeba polyphaga cultures. This work shows for the first time that collagen post-translational modifications are not confined to the domains of life. The utilization of glucose instead of the galactose found throughout animals, as well as a bifunctional enzyme rather than two separate enzymes, may represent a parallel evolutionary track in collagen biology. These results suggest that giant viruses may have contributed to the evolution of collagen biology.

INTRODUCTION
Collagens are the most abundant animal proteins, acting not only as a scaffold for tissue but also as regulators of many biological processes, including cell attachment, proliferation, and differentiation (1). Mainly linear proteins, collagens are characterized by domains composed of Gly-X-Y repeats, with Pro and Lys often present at positions X and Y. Nascent pro-collagen polypeptides are posttranslationally modified beginning with the hydroxylation of selected Pro (2) and Lys (3) residues. Some hydroxylysine (Hyl) residues are subsequently modified by addition of carbohydrates forming the disaccharide Glc-1,2-Gal-1,O-Hyl (4)(5). The extent of glycosylation can vary with tissue distribution and the type of collagen. The presence of 4hydroxyproline (Hyp) is essential for collagen triple helix formation and thermal stability (6). Lys hydroxylation produces substrates for glycosylation reactions and is also important for the cross-linking of collagen fibrils. The biological significance of collagen posttranslational modifications is reflected in the diseases associated with defective collagen modifications. Mutation of the lysyl hydroxylase genes PLOD1, PLOD2 and PLOD3 cause Ehlers-Danlos type-VI (7), Bruck syndrome (8), and a form of skeletal dysplasia (9), respectively.
The GLT25D1 and GLT25D2 genes encoding collagen galactosyltransferase (ColGalT) enzymes were only recently identified (10). Whole genome RNA interference studies in Caenorhabditis elegans suggest that loss of ColGalT is associated with severe phenotypes like slow growth, abnormal locomotion, and sterility (11). Interestingly, nonfibrillar proteins with collagen domains such as the hormone adiponectin (12), the mannosebinding lectin (13), and the acetylcholine esterase complex (14) also contain glycosylated Hyl. The collagen domains of these proteins are involved in protein folding and oligomerization making it likely that the glycan chains are involved in this process as well.
Collagens and collagen-like proteins are not confined to animals. Some fungi such as Metarhizium anisopliae (15), and bacteria such as Streptococcus pyogenes (16)(17) express collagen-like proteins. However, apart from a few proteins in bacteriophages (18)(19)(20), no collagen-like proteins in viruses have been reported. Little is known about collagen-like proteins in bacteria, fungi, and phages, and none of these proteins has been characterized for the common collagen post-translational modifications that are necessary for proper triple-helix formation. The first evidence suggesting viruses might code for their own glycosyltransferase genes rather than relying solely on host cell machinery was described in the Paramecium bursaria chlorella virus (21). More recently, reported genome sequences of viruses such as the shrimp White Spot Syndrome Virus (22), Lymphocystis Disease Virus (23), the Acanthamoeba polyphaga Mimivirus (24), and two virophages called Sputnik (25) and Organic Lake Virophage (26), suggested that viruses possess collagen-like proteins as well as collagen modifying enzymes.
Northern blots -Acanthamoeba polyphaga cells were infected with Mimivirus and RNA was isolated at 0, 4, 8, 16, and 24 hours post infection. For each sample, 2.5 g of RNA was separated on a 1% formaldehyde agarose gel and transferred to a nylon Hibond-N membrane (GE Healthcare Biosciences AB, Uppsala, Sweden). The probes were amplified by PCR using Mimiviral genomic DNA with the primers shown in Supplemental Table 2. Probes were labelled with [ 32 P]dCTP (Hartmann Analytic, Braunschweig, Germany) by random priming (Agilent, Basel, Switzerland). The membranes were incubated for 2 h at 80°C and prehybridized for 1 h at 64°C with QuikHyb hybridization solution (Agilent, Basel, Switzerland) containing 100 g/mL ultra pure herring sperm DNA (Invitrogen, Basel, Switzerland). Hybridization was performed overnight at 64°C using 5 × 10 5 cpm of labelled probe per mL of hybridization solution. Membranes were washed in 0.1 × SSC, 0.1% SDS and incubated in a 75°C water bath until the temperature of the solution reached 60°C. The washed membranes were exposed on Kodak BioMax XAR film (Sigma-Aldrich Chemie, Buchs, Switzerland) for 24 h at room temperature.
Baculovirus expression -Six well tissue culture plates were seeded with 2× 10 6 Spodoptera frugiperda 9 (Sf9) cells per well. Cells were allowed to attach to the plate and 500 L of Grace's Insect Cell Media containing 9.1% fetal calf serum was added to each well. A 160 L aliquot of virus stock was added to each well to be infected and incubated for 1 h at 28°C. Media was removed and replaced with 2 mL of fresh media. Cells were incubated at 28°C for 3 days. After 3 days, cells were loosened from the plate with a cell scraper and moved to a microcentrifuge tube. Cells were pelleted and the supernatant removed and retained as a viral stock. The cell pellet was washed with 150 L of TBS (50 mM Tris-HCl, 138 mM NaCl, 2.7 mM KCl) pH 7.4. The cells were lysed with 150 L of TBS + 1% Triton X-100. The cell pellet was incubated on ice for 30 min with the lysis solution. Cell debris was removed by centrifugation and the supernatant retained as an enzyme stock.
Bacterial expression -The pET16bbased expression vector was transformed into chemically competent E. coli BL21(DE3) (Merck, Geneva, Switzerland) using a heat shock and a 1 h recovery step in 1 mL of antibiotic free lysogeny broth (LB) (27) with shaking at 220 rpm at 37°C. The next day, a fresh colony was inoculated into 50 mL of LB supplemented with 100 g/mL ampicillin (Sigma-Aldrich) (LBamp + ) and incubated overnight at 37°C with shaking at 220 rpm. The next morning, 10 mL of the overnight culture was used to inoculate a 1 L culture of LBamp + which was incubated at 37°C with shaking at 200 rpm until an optical density at 600 nm (OD 600 ) of approximately 0.4 was reached, at which point the temperature was lowered to 32°C. When the OD 600 approached 0.6, protein expression was induced with the addition of isopropylthio--D-galactopyranoside to a concentration of 1 mM. The culture was incubated for a further 3 h after which the bacteria were pelleted at 6000 × g at 4°C for 30 min, and then resuspended in 30 mL of ice cold MCAC10 buffer (20 mM Tris-HCl pH 7.4, 500 mM NaCl (Sigma-Aldrich), 10 mM imidazole (Sigma-Aldrich), 10% v / v glycerol (ERNE surface AG, Dällikon, Switzerland)) prior to freezing at -20°C.
Protein purification -One 30 mL frozen pellet of transformed and induced E. coli BL21(DE3) was thawed and lysed under icecold conditions using an Emulsiflex C5 Frenchpress (Avestin, Mannheim, Germany). The lysed bacteria were clarified by two consecutive 30 min centrifugations at 13,000 × g at 4°C followed by filtration through a 0.22 m filter (Millipore, Zug, Switzerland) under vacuum. The protein was injected over a 1 mL HisTrap FF crude column (GE Healthcare) on an Äkta FPLC (GE Healthcare) at 1 mL/min. The column was washed with 10 column volumes of MCAC10 buffer containing 1 M NaCl, followed by 5 column volumes of MCAC10 to reequilibrate. The column was then washed with a 5 column volume 10 mM to 100 mM gradient of imidazole at 1 mL/min. The column was eluted with a 20 column volume 100 mM to 500 mM gradient of imidazole at 1 mL/min. Eluates were immediately concentrated to 1 mL or less and buffer exchanged into MCAC10 using Amicon Ultra Centrifugal Filters (Millipore, Zug, Switzerland) with a nominal molecular weight cut-off of 30,000 Da, at 4000 × g in a swinging bucket rotor in a Heraeus Cryofuge 6000i centrifuge (Thermo-scientific, Geneva, Switzerland) at 4°C. The protein was stored in MCAC10 buffer at 4°C in the cold room for up to 1 week until needed. Samples for SDS-PAGE were prepared in 4 × loading buffer (200 mM Tris-HCl pH 6.8, 400 mM DTT (Sigma-Aldrich), 8% w / v SDS (Sigma-Aldrich), 40% v / v glycerol, 4 mg/mL bromophenol blue (Merck). Ten L of sample in 4 × loading buffer was subjected to 10% SDS-PAGE. The gel was then stained with Coomassie blue R250 to visualize protein bands.
Lysyl hydroxylase assays -Lysyl hydroxylase assays were performed substantially as described in Kivirikko et al. (28). All solutions were kept on ice. Initiation of assays and all subsequent manipulations were performed in a fume-hood to prevent exposure to 14 CO 2 . Fresh stocks of 2 mM FeSO 4 (Sigma-Aldrich), 20 mM ascorbate (Sigma-Aldrich), and 6 mM 2-oxoglutarate (Sigma-Aldrich) were prepared. Enzyme and acceptor substrate (either His 10 -tag purified, E. coli expressed L71 Mimivirus collagen in MCAC10 buffer, or collagen-like peptide acceptors (GenScript, Piscataway, NJ, USA) dissolved in ddH 2 O) were added to each reaction tube. A master-mix of the remaining assay components was prepared, and aliquots of this were used to initiate each assay. The assay contained 50 nCi of 2-[ 14 C]oxoglutarate (Perkin Elmer, Schwerzenbach, Switzerland), 300 M 2-oxoglutarate, 100 M FeSO 4 , 1 mM ascorbate, 50 mM Tris-HCl pH 7.4, and 100 M DTT. When peptide acceptor substrates were used, the assay contained 600 g/mL of peptide. Total assay volume was 100 L with the master-mix component comprising no less than half the total volume. Briefly, a small rectangular filter paper was soaked in NCS II Tissue Solubilizer (GE Healthcare) and suspended from a small hook in a rubber stopper. The top was cut from the microcentrifuge tube containing the enzyme and acceptor substrate which was then carefully lowered into a 30 mL scintillation vial (VWR, Dietikon, Switzerland). The assay was initiated by addition of the master mix, and the vial was immediately closed with the stopper allowing the soaked filter paper to absorb any radioactive 14 CO 2 produced. The vial was incubated at 37°C for 1 h. The assay was stopped with 100 L of ice cold 1 M KH 2 PO 4 administered into the reaction tube by a syringe and needle inserted through the stopper. The stopped assay was incubated for 30 min at room temperature, at which point the rubber stopper was removed and the filter paper transferred to a fresh scintillation vial. The filter paper was vortexed for approximately 5 s with 10 mL of IRGA-Safe Plus scintillation fluid (Perkin Elmer) and then measured in a Tri-Carb 2900TR scintillation counter (Perkin Elmer).
Collagen glycosyltransferase assays -Glycosyltransferase assays were performed with bovine collagen type I or peptides as acceptor substrates substantially as described in Schegg et al. (10). Briefly, 20 nCi of UDP-[ 14  with a total UDP-sugar concentration of 240 M were incubated with enzyme and substrate for 30 min at 37°C. The assay was stopped by incubation on ice with 5% phosphotungstic acid/5% trichloro-acetic acid for 30 min. The precipitate was recovered on filters using a vacuum manifold, and then scintillation counted. Total assay volume was 100 L. Assays measuring the acceptor substrate specificity contained a total UDP-sugar concentration of 60 M, and utilized baculovirus produced L230 and GLT25D1 enzyme. The enzymes for this study could not be produced in E. coli as GLT25D1 will not express as a functional protein in that system. Baculovirus produced enzyme was prepared by lysing the cells and centrifuging the lysate to remove cellular debris. Double activity (lysyl hydroxylase plus collagen glucosyltransferase activity) assays utilized 20 nCi of UDP-[ 14 C]Glc without additional UDP-Glc added (approximately 660 nM UDP-Glc). E. coli produced, His 10 -tag purified Mimivirus L71 protein was used as the acceptor substrate. Assays were incubated overnight at 4°C.
Amino Acid Analysis -Aliquots of 10 g of Mimivirus proteins were hydrolyzed in 4 M NaOH for 48 h at 100°C. Amino acids were derivatized with Fmoc as described by Bank et al. (29) and analyzed by fluorescence detection after HPLC separation (10). Up to 100 L of sample was injected over an ODS Hypersil 150 mm × 3 mm column with a 3 m particle size (Thermo-scientific). Underivatized amino acids were also analyzed using the Dionex AAA-Direct, high pH anion exchange chromatography system (Thermo-scientific).
Samples were diluted up to 300-fold prior to injection of 25 L over an AminoPac PA-10 column (Thermoscientific) using the following gradient: 0-8 min, Collected fractions were then analyzed by direct infusion mass spectrometry using a TriVersa NanoMate HD system (Advion BioSciences) connected to a quadrupole-TOF mass spectrometer (QTOF G2, Waters).

Mimivirus L230 is Similar to Human Lysyl-Hydroxylase and Collagen
Galactosyltransferase Enzymes -The Mimivirus ORF L230 was identified as a potential collagen modifying enzyme because of sequence similarity to the human ColGalT GLT25D1 at the amino terminus and homology to a human lysyl hydroxylase PLOD1 at the carboxy terminus (Fig. 1). The L230 amino terminus shares 27.7% amino acid identity and 61% similarity with GLT25D1 over a stretch of 285 amino acids. At the carboxy terminus, L230 has 35% amino acid identity and 68% similarity with PLOD1 over a stretch of 388 amino acids (Fig. S1).
Mimivirus L230 and Mimivirus Collagen Transcripts are Expressed Together During Infection -The possible role of L230 as a collagen modifying enzyme is relevant considering that the Mimivirus genome encodes seven collagen-like genes (24). In fact, Northern blot analysis of Mimivirus gene expression in infected Acanthamoeba polyphaga culture confirmed the co-expression of L230 with the seven Mimivirus collagen genes L71, R196, R239, R240, R241, L668, L669 by the late phase of the lytic infection (Fig. 2). This finding indicates that L230 is produced at the same stage of the viral life cycle as its putative substrates.
L230 is a Bifunctional Lysyl Hydroxylase and Collagen Glycosyltransferase -The L230 protein was expressed in E. coli with an N-terminal His 10 -tag and purified by Ni 2+ affinity chromatography (Fig. S2). To determine if L230 is enzymatically active, lysyl hydroxylase activity toward the peptide acceptor substrate (GDK)4 (Table S1) was assayed. L230 showed substantial activity toward this peptide confirming indications from homology shared with the PLOD1 enzyme that L230 is a lysyl hydroxylase (Fig. 3A). Similarly, to assess the putative collagen glycosyltransferase activity of L230, we assayed the transfer of carbohydrates to bovine collagen type I as an acceptor substrate. This experiment confirmed the activity of L230 as a glycosyltransferase. Unlike the animal ColGalT enzymes GLT25D1 and GLT25D2, L230 preferentially catalyzed the transfer of Glc to collagen (Fig. 3B) identifying L230 as a glucosyltransferase.
L230 is a Collagen O-Glucosyltransferase Defining a Novel Collagen Glycoform -The product of the L230 glycosyltransferase reaction was further analyzed after alkaline hydrolysis to determine the structure of the resulting glycoconjugate. The corresponding amino acid analysis of modified bovine collagen type I by HPLC pointed to the presence of [ 14 C]Glc-O-Hyl, indicating that L230 attaches Glc directly to Hyl (Fig. 4). The core Gal-O-Hyl structure is found in all animal collagens, from sponge (30), to birds (31) to mammals (4,32). Therefore, the Glc-O-Hyl core represents a novel form of collagen glycosylation.

L230 is Substrate Promiscuous Toward Both Viral and Human Collagen Substrates -
The acceptor substrate specificity of L230 was further investigated in a lysyl hydroxylase assay using six collagen-like peptides as acceptor substrates. In addition to the (GDK)4 peptide (Fig. 3), we tested a standard (GIK)4 peptide (33) and four peptides derived from the Mimivirus collagen ORFs L71 (labeled L71a), R196, R240 and L669 (Table S1). The highest activity was observed with the R240 peptide (Fig. 5A) featuring a mix of hydrophobic, hydrophilic, positively and negatively charged residues. More work is required to delineate the exact substrate preference of L230 since at present we cannot discern any obvious pattern of acceptor substrate specificity. The collagen glucosyltransferase (ColGlcT) activity of L230 was also assayed on three Hyl containing peptides (Table S1). The peptide labeled L71b was from the Mimivirus ORF L71 collagen-like protein, while peptides Col2A1, and ADPQ were from human collagen 2A1 and adiponectin, respectively. Mirroring the result with the lysyl hydroxylase activity, L230 exhibited substrate promiscuity, as shown by the ability of the enzyme to glycosylate Hyl in both Mimiviruslike and human-like sequences (Fig 5B). By comparison, the human collagen galactosyltransferase GLT25D1 showed similar activities toward the three acceptor peptides (Fig. 5C). However, the human enzyme had a somewhat lower tolerance for the Mimivirus acceptor peptide featuring GDK repeats, which are not common in animal collagens.
The L230 Lysyl Hydroxylase and Glucosyltransferase Activities are Separable -To determine whether the lysyl hydroxylase and glycosyltransferase domains of L230 can function independently from each other, we prepared separate mutants in each domain and assayed the corresponding lysyl hydroxylase and ColGlcT activity (Fig. 6A and B). Mutation of two conserved residues (H825S/D827A) in the lysyl hydroxylase domain (Fig. S1) abrogated the lysyl hydroxylase activity while having no effect on the ColGlcT activity (Fig. 6A). Many glycosyltransferases possess DXD motifs that chelate a metal ion cofactor. We began by attempting to mutate the putative DXD motif (10) in the glycosyltransferase domain (E106A/D107A/D108A, E106A, D107A) (Fig.  S1). Unfortunately, these mutant proteins did not express stably in E. coli. Most mutations in the N-terminal domain of L230, namely L78K, H80A, D97A/D99A, and D131A in addition to the aforementioned DXD mutants yielded unstable proteins when expressed in E. coli. The substitution D250A in the glycosyltransferase domain did not affect the expression level and stability of L230 but led to decreased ColGlcT activity while leaving the lysyl hydroxylase activity intact (Fig. 6B). L230 Sequentially Hydroxylates and Glucosylates Collagen Substrates in vitro -Next, we assessed whether the L230 enzyme is capable of performing both the lysyl hydroxylase and ColGlcT activities in the same assay. As acceptor substrate, we utilized His 10 -tag purified Mimivirus L71 collagen expressed in E. coli. As shown by the incorporation of [ 14 C]Glc into L71 collagen (Fig. 6C), L230 first hydroxylated Lys in L71 and then glucosylated the newly formed Hyl. By comparison, the H825S/D827A mutant enzyme which possesses wild-type ColGlcT activity (Fig. 6A) but lacks the lysyl hydroxylase activity was unable to produce glucosylated L71 in the double activity assay. Similarly, the reaction of L230 with L71 in the presence of UDP-[ 14 C]Glc but absence of 2-oxoglutarate failed to produce glycosylated L71. The lack of activity with the H825S/D827A mutant enzyme confirms the complete absence of Hyl in the E. coli produced L71 substrate prior to modification by L230. Thus, these experiments demonstrate that L230 is a bifunctional enzyme, capable of sequentially hydroxylating and glycosylating collagen.
Post-translationally Modified Collagen is Present in Mimivirus -Considering the dual activity of L230 and the expression of the L230 gene during Mimivirus replication, we investigated whether Hyl and Glc-O-Hyl could be identified in Mimivirus proteins. To this end, Mimivirus was isolated from the supernatant of infected A. polyphaga and the viral protein was subjected to alkaline hydrolysis and analyzed. Using the Dionex AAA-Direct amino acid analysis system with pulsed amperometric detection, we could readily detect Hyl but no Glc-O-Hyl in the virus extract (Fig. 7). Fractions covering the region of Glc-O-Hyl elution were collected and analyzed by mass spectrometry. However, we did not see any ion products corresponding to Glc-O-Hyl in these fractions. Taken together, our findings show that Mimivirus possesses the machinery to posttranslationally modify collagens in vivo, thereby establishing the existence of modified collagen in Mimivirus virions.

DISCUSSION
We have shown that Mimivirus L230 is a bifunctional lysyl hydroxylase and glycosyltransferase enzyme involved in the posttranslational modification of collagen. This is the first report of a collagen glycosylating enzyme in viruses and of hydroxylysine outside of animals. It is interesting to note that an animal collagen glucosyltransferase activity has been reported for human lysyl hydroxylase 3 (34). Importantly, the description of Mimivirus L230 activity reveals a novel type of collagen core glycosylation in nature.
The seven Mimivirus collagen proteins are dominated by Gly-X-Y repeats, but are, in contrast to mammalian collagens, poor in proline content. The in vitro activity of a viral prolyl hydroxylase toward peptides substrates has been reported in Paramecium bursaria Chlorella virus-1 (35). However, the presence of hydroxyproline in viral proteins has not yet been established, nor is it known whether this viral enzyme is a collagen modifying hydroxylase. While Hyp is generally required for thermal stability of animal collagen (36), there are instances, such as in worms from deep sea hydrothermal vents, where the typical role of Hyp is replaced by glycosylated Thr (37)(38). Whether or not Mimivirus collagens form triple helices, and whether triple-helix formation is Hyp dependent or is dependent on some other residue type is as yet unknown. Additionally, the role of collagen in Mimivirus biology remains an open question. It seems reasonable to speculate that there is a structural function for these proteins, possibly contributing to the dense layer of fibrils reported to cover the viral capsid (39). These fibrils could play a role in providing the virus resistance to the environment where viral interaction with host cells may be an infrequent occurrence.
The apparent ease with which giant viruses accumulate and potentially disseminate genes from various organisms (40)(41)(42)(43)(44)(45) begs the question of what role Mimivirus may have played in the evolution of collagen biology. The presence of collagens in animals and near absence in other organisms suggests that collagen biology was an animal adaptation. The disparity between animal collagen sequences and typical Mimivirus sequences, including a paucity of Pro in the viral collagens raises potential questions about the origin of collagen biology. The possibility exists that collagens did not originate with animals, but first emerged from a microbial source earlier in evolution. Nevertheless, the ability of viruses such as Mimivirus to transfer genes from one organism to another signifies a potential role in the evolution of collagen biology irrespective of the ultimate origins of collagens in nature.
There are other potential glycosyltransferase enzymes in the Mimivirus genome including L137, L193 and L363, which are homologous to bacterial glycosyltransferases. The L179 ORF is homologous to an archaea glycosyltransferase, and the L373, R654, and R655 ORFs are homologous to animal glycosyltransferases (24,(46)(47). Whether or not the Glc residue on Hyl in Mimivirus collagens is further elongated, possibly by one of these enzymes, is still an unanswered question. Indeed, our analysis did not identify Glc-O-Hyl in Mimivirus protein extracts whereas Hyl was readily detectable. It was recently reported that Mimivirus experiences a significant reduction in genome size after 150 passages through an amoebal host (48). While the ORFs of several putative Mimivirus glycosyltransferases and one of the collagen genes were either deleted or disrupted in this reduced genome, L230 was maintained as were the remaining six collagen genes.
Mimivirus ColGlcT is the very first microbial collagen glycosyltransferase described. It is also the first viral protein shown to be a bifunctional enzyme involved in the posttranslational modification of collagen. This work confirms the presence of post-translationally modified collagen domains in proteins outside the three currently accepted domains of life. enzymes respectively, is shown below. Identical residues are shaded in black. The protein sequence intervening between the two regions of homology is shown as a line connecting the two boxes. An alignment containing the complete sequences is available in Fig. S1.     (Table S1). Negative control assays (Neg) without acceptor substrate are shown as black bar. The error bars show the standard error of the mean of six assays from two independent experiments. (B) L230 ColGlcT activity toward three peptide acceptor substrates containing Hyl (Table S1). Mock assays lacking the recombinant enzymes are shown as black bars. The error bars show the standard error of the mean of four assays from two independent experiments. (C) GLT25D1 ColGalT activity towards the same three peptide acceptor substrates containing Hyl as in panel B (Table S1). Mock assays lacking the recombinant enzymes are shown as black bars. The error bars show the standard error of the mean of four assays from two independent experiments. shows decreased ColGlcT activity. The acceptor substrate is bovine collagen type I. In both panels, the standard error of the mean of six assays from two independent experiments is shown. (C) Combined lysyl hydroxylase and ColGlcT assays. The transfer of [ 14 C]Glc on Mimivirus L71 collagen substrate was measured in the presence of L230 or L230(H825S/D827A) lacking lysyl hydroxylase activity. The absence of 2-oxoglutarate in the assay buffer prevented the transfer of [ 14 C]Glc to L71 collagen. These data show the mean and standard error of the mean of four assays from two independent experiments.

FIGURE 7.
Hyl detection in Mimivirus protein extracts. Amino acids were detected by pulsed amperometric detection. Peaks are labelled with amino acid single letter codes. The retention time of Glc-O-Hyl is indicated by an arrow. The asterisk indicates an unidentified peak.