A Structural Domain Mediates Attachment of Ethanolamine Phosphoglycerol to Eukaryotic Elongation Factor 1A in Trypanosoma brucei

Ethanolamine phosphoglycerol (EPG) represents a protein modification that so far has only been found in eukaryotic elongation factor 1A (eEF1A). In mammals and plants, EPG is covalently attached to two conserved glutamate residues located in domains II and III of eEF1A. In contrast, Trypanosoma brucei eEF1A contains a single EPG attached to Glu362 in domain III. The sequence and/or structural requirements for covalent linkage of EPG to eEF1A have not been determined for any organism. Using a combination of biosynthetic labelling of parasites with tritiated ethanolamine and mass spectrometry analyses, we demonstrate that replacement of Glu362 in T. brucei eEF1A by site-directed mutagenesis prevents EPG attachment, whereas single or multiple amino acid substitutions around the attachment site are not critical. In addition, by expressing a series of eEF1A deletion mutants in T. brucei procyclic forms, we demonstrate that a peptide consisting of 80 amino acids of domain III of eEF1A is sufficient for EPG attachment to occur. Furthermore, EPG addition also occurs if domain III of eEF1A is fused to a soluble reporter protein. To our knowledge, this is the first report addressing amino acid sequence, or structure, requirements for EPG modification of eEF1A in any organism. Using T. brucei as a model organism, we show that amino acid substitutions around the modification site are not critical for EPG attachment and that a truncated version of domain III of eEF1A is sufficient to mediate EPG addition.


Introduction
Eukaryotic elongation factor 1A (eEF1A) is one of the most abundant cytosolic proteins in all eukaryotic cells. It represents an essential component during (poly-) peptide chain elongation by binding aminoacyl-tRNAs in a GTP-dependent reaction to the acceptor site of ribosomes. After codon-anticodon match between the aminoacylated tRNA and the ribosomal acceptor site followed by GTP hydrolysis to GDP, eEF1A dissociates from the ribosome in its GDP-bound form and interacts with nucleotide exchange factor eEF1Ba, which replaces GDP by GTP to reactivate eEF1A (reviewed by [1]). Crystal structures of eEF1A in complex with components of eEF1B show that yeast eEF1A consists of three clearly distinct domains [2,3]. Domain I harbors the GTP/GDP binding site, while binding of eEF1Ba occurs to both domains I and II [2,4]. In addition, domain II also contains the binding site for aminoacyl-tRNAs [5][6][7], whereas in yeast domain III represents the binding site for elongation factor 3 (eEF3) [8].
Apart from its pivotal role during polypeptide synthesis, eEF1A is involved in many other cellular processes. Most notably, eEF1A binds to cytoskeletal components and modulates their interactions [9][10][11], is involved in signal transduction events [12] and mediates nuclear export of proteins [13] and mitochondrial tRNA import [14]. Interactions of eEF1A with components of the cytoskeleton, which occur independent of its translation elongation function, have been assigned to domains II and III [15,16]. Such nonconventional roles for eEF1A may not be surprising since eEF1A amounts to 1-3% of total cytosolic protein [17] and, thus, is present in excess compared to its ligands in peptide synthesis. Since eEF1A is one of the most highly conserved proteins between eukaryotic species, it likely adopts a similar tertiary structure in all eukaryotes, as revealed by programs to predict the threedimensional structures of proteins. Thus, the sites of interaction with its ligands are likely conserved in many other organisms as well.
The biological activity and cellular distribution of a given protein is often modulated by covalent modifications of amino acid side chains. Several protein modifications, such as phosphorylation [18,19], lysine methylation [20,21], and C-terminal methylesterification [22], have been reported to affect the activity of eEF1A during peptide synthesis in several species, yet their precise functions remain unknown (reviewed by [23]). In addition, mammalian and plant eEF1A is modified by ethanolamine phosphoglycerol (EPG) moieties attached to two conserved glutamic acid residues in domains II and III [21,[24][25][26]. The function of this modification, which so far has not been reported in any other protein except eEF1A, is completely unknown. Very recently, we have shown that eEF1A of the ancient eukaryote, Trypanosoma brucei, is also modified with EPG, however, in contrast to mammalian and plant eEF1A, T. brucei eEF1A contains a single EPG modification only. In addition, we showed that the phospholipid phosphatidylethanolamine (PE) represents the biosynthetic precursor of EPG in T. brucei [27]. The enzymes involved in EPG attachment to eEF1A, or the amino acid sequence or structure requirements for EPG modification of eEF1A, have not been characterized. Interestingly, Saccharomyces cerevisiae represents the only eukaryote so far where eEF1A has been reported not to be modified with EPG [28].
The present study was aimed at identifying sequence, or structure, motifs in T. brucei eEF1A necessary for EPG addition. By replacing (conserved) amino acids at, or around, the EPG modification site and by deleting individual domains of eEF1A in T. brucei, we found that attachment of EPG is strictly dependent on glutamate as the acceptor site and that (part of) domain III of eEF1A is sufficient to confer EPG modification, even if this domain is attached to a reporter protein.

Materials
Unless otherwise specified, all reagents were of analytical grade and were from Merck (Darmstadt, Germany), Sigma-Aldrich (Buchs, Switzerland) or ICN Biomedicals (Tä gerig, Switzerland). DNA polymerase was obtained from Qiagen (Hombrechtikon, Switzerland). All restriction enzymes were purchased from Roche Diagnostics (Rotkreuz, Switzerland) or New England Biolabs (Ipswich, MA

Generation of Expression Vectors and Site-Directed Mutagenesis
All plasmids containing hemagglutinin-tagged eEF1A (HA-eEF1A, Table S1) were derived from the trypanosome expression vector pCorleone [31]. The open reading frame of T. brucei eEF1A (GeneDB Tb10.70.5670), or parts of it, were amplified from T. brucei genomic DNA using appropriate primer pairs (Table S2). PCR products with flanking HindIII and BamHI restriction sites were subcloned into the pCRH2.1-TOPOH vector (Invitrogen) following the manufacturer's instructions and cloned between the HindIII and BamHI sites of pCorleone. Site-specific mutations of HA-eEF1A were introduced into the parental vector pE-GhaEF (12449) (Table S1) using the QuikChangeH Site-Directed Mutagenesis kit (Stratagene, Basel, Switzerland) and the appropriate primers (Table S3). To generate fusion proteins containing domain III of T. brucei eEF1A linked to Alba 1 (HA-Alba-III) or PTP (PTP-III), the PCR product corresponding to amino acids 315-449 of eEF1A was subcloned between the EcoRI and BamHI sites of pBluescript, resulting in pBS (3152449) . The open reading frame of Alba 1 (GeneDB, Tb11.02.2040) was amplified from T. brucei genomic DNA, and a PTP-tag was generated using plasmid pN-PURO-PTP as template (GenBank, DQ172901; Schimanski, 2005) and the appropriate primers (Table S2). Resulting PCR products were inserted between the HindIII and EcoRI sites of pBS (3152449) . The internal BamHI restriction site in the PTP sequence was removed by site directed mutagenesis as described above (see Table S2 for primers). The resulting cassettes HA-Alba-III and PTP-III were cloned between the HindIII and BamHI sites of pCorleone, resulting in plasmids pEGhaAlba (3152449) and pEGptp (3152449) ( Table S1). All DNA sequences were verified by sequencing (Microsynth AG, Balgach, Switzerland). Prior to transfection into T. brucei Dprocyclin#1, the vectors were linearized with NotI and SalI.

In Vivo Labeling, Extraction and Immunoprecipitation
Trypanosomes during exponential growth (0.5-0.8610 7 cells ml 21 ) were labeled with 1 mCi ml 21 [ 3 H]Etn for 18 h [32,27]. Subsequently, parasites were harvested and washed twice with icecold TBS (10 mM Tris/HCl, pH 7.4, 144 mM NaCl) to remove non-incorporated label. Proteins from 2610 8 cells were extracted with 300 ml of lysis buffer (50 mM Tris, pH 7.5; 150 mM NaCl; 0.1% (v/v) Nonidet P40), containing complete protease inhibitor cocktail tablets (Roche Diagnostics), for 15 min on ice. The cell lysate was homogenized by passing three times through a 27-gauge needle and cleared by centrifugation at 16,000 g for 30 min. The supernatant was supplemented with either anti-HA Affinity Matrix (Roche Diagnostics) or IgG Sepharose 6 Fast Flow matrix (GE Healthcare, Otelfingen, Switzerland) and incubated on a rotating device overnight at 4uC. The beads were spun down and washed three times with lysis buffer. Tagged proteins bound to the matrix were released by boiling in electrophoresis sample buffer.

SDS-Polyacrylamide Gel Electrophoresis (SDS-PAGE) and Immunoblotting
Glycine-SDS-PAGE was performed under reducing conditions using 12% polyacrylamide gels [33]. Proteins with predicted masses smaller than 15 kDa were separated by Tricine-SDS-PAGE (Schä ger, 2006) in the presence of 6 M urea using 16% polyacrylamide gels. For fluorography of 3 H-labeled proteins, gels were fixed (10% methanol, 7% acetic acid), soaked in Amplify (GE Healthcare), dried, and exposed to Kodak MBX films at 270uC. For immunoblotting, proteins were transferred at 2.5 mA/cm 2 for 75 min onto polyvinylidene difluoride membranes (Millipore, Bedford, MA) using a semidry blotting apparatus. The blotted membranes were blocked overnight with 5% (w/v) milk powder in TBS containing 0.05% (v/v) Tween 20 and probed with primary antibody. Mouse monoclonal antibodies against HA (a-HA; Covance, Berkeley, CA) and eEF1A (a-EF; Upstate, Lake Placid, NY) were used at a dilution of 1:3000 and 1:5000, respectively. Primary antibodies and the PTP-tag were detected with secondary rabbit anti-mouse IgG antibody conjugated to horseradish peroxidase (Dako, Baar, Switzerland) at a dilution of 1:5000, followed by enhanced chemiluminescence (Pierce, Lausanne, Switzerland).

Mass Spectrometry
Immunoprecipitated HA-eEF1A was analyzed by SDS-PAGE and blotted as described above. After staining with Coomassie Brilliant Blue, proteins were cut out and exposed to on-membrane reductive alkylation and trypsin digestion [34]. The tryptic peptides were analyzed by liquid chromatography tandem mass spectrometry (nano-LC-MS/MS) using a Rheos2200 pump, equipped with GROM-SIL MB6 C8 or Magic-C18 columns (75 mm, 7 cm), coupled to an LTQ-orbitrap XL mass spectrometer. The peptide-resolving part of the nano-LC gradient was from 5% to 35% acetonitrile in 0.1% formic acid over 20 min at ,400 nl min 21 . The two most abundant ions were subjected to collision-induced dissociation (CID) fragmentation using a mass list comprising all expected 2+/3+ charge states of the modified and non-modified peptides. CID was triggered first in normal mode and second with the multistage activation option on when the prominent neutral loss fragment of 2172 Da was observed as most abundant ion. CID spectra were recorded in the orbitrap with resolution set to 7500 at m/z 400. CID spectra were interpreted by Phenyx software (Genebio SA, Geneva, Switzerland) using carboamidomethylation of cysteine as a fixed modification, and oxidation of methionine and EPG on glutamic acid as variable modification. All nano-LC-MS/MS chromatograms were manually controlled for the presence of the expected peptides.

Results
Amino Acid Sequence Requirement for EPG Attachment eEF1A shows a very high amino acid sequence identity over the entire polypeptide chain between different organisms. Interestingly, however, the residues around the EPG modification site, Glu362 in T. brucei [27], are less well conserved (Fig. 1). To study the importance of conserved amino acids within this stretch of the primary sequence, i.e. between Cys358 and Ile367, for EPG attachment to eEF1A, we introduced point mutations at, or around, Glu362 in ectopically expressed, HA-tagged full-length eEF1A (HA-eEF1A) in T. brucei procyclic forms. Attachment of EPG to eEF1A was monitored by incubating trypanosomes in culture with [ 3 H]Etn, which has been used before to label eEF1A in mammalian cells [24,25], plants [26], and T. brucei [27]. Since this procedure also results in labeling of abundant GPI-anchored surface proteins in trypanosomes (see, e.g., [35,36,32,37]), we chose the T. brucei procyclic cell line, Dprocyclin#1, which lacks the expression of the major GPI-anchored proteins, EP and GPEET procyclins [29], for all transfection experiments. Using this cell line, labeling with [ 3 H]Etn results in a single radioactive band at 49 kDa after SDS-PAGE and fluorography, allowing easy identification of wild-type or HA-tagged eEF1A ( [27]; see also below). Expression of the various HA-tagged forms eEF1A had no effect on growth of the parasites in culture. The amino acid sequence of T. brucei eEF1A between residues 347 and 373 is compared to the corresponding eEF1A sequences from Daucus carota (GenBank, BAA02205) and Mus musculus (GenBank, NP_034236) using ClustalW. A grey background indicates amino acid identity between the sequences. The arrows in indicate the conserved residues at, or around, the EPG modification site of T. brucei eEF1A (Glu362) that were mutated in this study (E362Q, E362A, E362D, F360T, A361D, K366S). The box marks the tryptic fragments containing the EPG modification site (indicated by the asterisk). doi:10.1371/journal.pone.0009486.g001  Immunoblotting experiments showed that HA-eEF1A proteins carrying the point mutations E362Q, E362A and E362D were expressed in T. brucei procyclic forms and could be immunoprecipitated using a-HA antibodies (Fig. 2, upper panels). As expected, replacement of the glutamate residue representing the EPG attachment site by glutamine (E362Q), or alanine (E362A), resulted in complete loss of [ 3 H]Etn labeling of HA-eEF1A (Fig. 2, lower left and middle panels). In addition, substitution of glutamate by aspartate (E362D) also completely abolished labeling of HA-eEF1A (Fig. 2, lower right  Since the conserved glutamate residue in eEF1A from S. cerevisiae is not modified with EPG [28], we converted the tryptic peptide of T. brucei eEF1A, FAEIESK, carrying the EPG modification site, to the corresponding tryptic peptide from yeast eEF1A, FDELLEK (Fig. 3A), and studied its potential modification in T. brucei. Our results show that replacement of all four amino acids around the EPG attachment site to mimic the yeast sequence had no effect on [ 3 H]Etn labeling of HA-eEF1A in T. brucei (Fig. 3B).

Mass Spectrometry Analysis of EPG Attachment to Mutated HA-eEF1A
Although incorporation of [ 3 H]Etn into eEF1A is a helpful and convenient marker for EPG labeling, it provides no information on the exact site of modification, or if the entire EPG moiety-and not only ethanolamine (see [38]) -is attached. To obtain this information, purified (immunoprecipitated) HA-tagged proteins were reduced, alkylated, digested with trypsin, and analyzed by nano-LC-MS/MS. All tryptic peptides recovered matched the deduced amino acid sequence of wild-type T. brucei eEF1A (results not shown), except for the peptides carrying the point mutations (see below). HA-tagged wild-type eEF1A revealed a tryptic fragment with m/z 1020. 465 Table 1 and [27]). MS/MS analysis of the parental ion at m/z 510.736 identified Glu362 in the sequence of the tryptic peptide FAE*IESK (with E* indicating EPG-modified glu) as the site of EPG attachment (Fig. 4A) (Fig. 4B). Similarly, only the unmodified tryptic peptides FAQIESK and FAAIESK were detected in the E362Q and E362A mutants, respectively (Table 1). In contrast, point mutations at the conserved residues Ala361 and Lys366 had no effect on EPG attachment to T. brucei  [11,39].
Together, these results demonstrate that the attachment of EPG to eEF1A only occurs if Glu362 is strictly conserved, but is not affected by substitution of conserved amino acids in the vicinity of the modification site. Based on these observations, and our failure to identify an amino consensus sequence in eEF1A that could mediate EPG attachment, we investigated if a structural domain may determine EPG modification of eEF1A.

EPG Attachment to eEF1A Deletion Mutants
The predicted three dimensional structure of T. brucei eEF1A shows three distinct structural domains (Fig. 5A, termed I, II and III) and closely matches the three dimensional structure of S. cerevisiae eEF1A as determined by X-ray crystallography [3]. The Glu362 residue modified by EPG in T. brucei eEF1A is located in a predicted b-sheet in domain III (Fig. 5A). To determine the structural requirements for the attachment of EPG to Glu362, several HA-tagged deletion mutants of T. brucei eEF1A were designed (Fig. 5B) and transfected into T. brucei procyclic forms. Expression of HA-eEF1A and incorporation of [ 3 H]Etn was monitored as described for the point mutations. Immunoblotting experiments using a-HA demonstrate that eEF1A proteins lacking domain II (construct 1-243/315-449) or domains I + II (construct 315-449), as well as HA-tagged full-length eEF1A (construct 1-449), were expressed (Fig. 6A). The minor band migrating below HA-tagged eEF1A lacking domain II likely represents a proteolytic degradation product of the protein construct. Interestingly, after incubation of trypanosomes with [ 3 H]Etn, the two truncated proteins were strongly labeled (Fig. 6B), indicating that domains I and II are not required for incorporation of [ 3 H]Etn into domain III of T. brucei eEF1A.
To identify a minimal structural unit of domain III needed for [ 3 H]Etn incorporation, we generated deletion mutants of domain III and analyzed their expression and labeling with [ 3 H]Etn. The results show that the truncated forms of domain III, after deletion of 33 amino acids at the N-terminus (construct 348-449) or 55 amino acids at the C-terminus (construct 315-394), were expressed (Fig. 6A) and, in addition, labeled with [ 3 H]Etn (Fig. 6B). In contrast, no expression was observed after truncating both the N-and the C-terminus of domain III, resulting in a peptide fragment consisting of amino acids 348-394 of eEF1A (results not shown). To confirm the site of attachment and the structure of the modification, the deletion mutants were analyzed by nano-LC-MS/MS. The results showed that the purified (immunoprecipitated) HAtagged eEF1A peptides 348-449 and 315-394 were modified at Glu362 and that the entire EPG structure was present (Table 1). Together, these results demonstrate that a peptide consisting of 80 amino acids of domain III of eEF1A is sufficient for EPG modification in T. brucei.

EPG Attachment to Chimeric Proteins
Having demonstrated that domain III of eEF1A alone is modified by EPG, we next studied if linking this domain to other proteins confers EPG modification to the chimeric proteins. For this purpose, we linked domain III of eEF1A to the small cytosolic protein, T. brucei Alba 1, or to the tandem affinity purification tag, PTP [40], resulting in the chimeric proteins HA-Alba-III or PTP-III with predicted molecular masses of 30 or 33 kDa, respectively. Immunoblotting experiments showed that both constructs were expressed in T. brucei procyclic forms (Fig. 7A, B, upper panels) and incorporated [ 3 H]Etn (Fig. 7A, B, lower panels). Nano-LC-MS/MS analysis of immunoprecipitated proteins confirmed the presence of the entire EPG moiety attached to Glu362 of domain III (Table 1). These data demonstrate that domain III of eEF1A is modified with EPG as independent structural unit, regardless if it is expressed as individual polypeptide or attached to other proteins.

Discussion
Based on the published three-dimensional structure of yeast eEF1A [3,4] and the high amino acid sequence homology between eEF1A from different organisms, the EPG modification site in T. brucei eEF1A, Glu362, is predicted to be on the surface of a b-sheet in domain III. Using in vivo labeling with [ 3 H]ethanolamine and mass spectrometric analyses, we now showed that a peptide consisting of amino acids 315-394 of the primary sequence, which is located in domain III of eEF1A, is sufficient for EPG attachment to occur. This was demonstrated by expressing HA-tagged forms of eEF1A lacking domains I and/or II, or the N-and C-terminal parts of domain III. Similarly, EPG modification also occurred on chimeric proteins consisting of domain III of T. brucei eEF1A fused to soluble reporter proteins, demonstrating that domain III contains all necessary information to become EPG-modified. In addition, point mutations of conserved amino acids around the EPG attachment site didn't prevent EPG modification, indicating that a specific primary sequence does not appear to be involved in EPG attachment. Furthermore, while replacement of Glu362 by Ala or Gln was expected to block EPG attachment, replacement by Asp also completely abolished EPG addition, demonstrating that the enzyme catalyzing attachment of EPG, or its precursor molecule, is highly specific for Glu at this position and doesn't  immunoprecipitates (IP) were analyzed by SDS-PAGE using 12% acrylamide gels. SDS PAGE was followed either by immunoblotting (upper panels) or fluorography (lower panels). For immunoblotting a-HA antibody or a-IgG were used to detect HA-Alba-III or PTP-III, respectively. The lanes contain material from 1610 7 parasites, except for the lanes with the immunoprecipitates in the fluorographs, which contain material prepared from 1.8610 8 cell equivalents. doi:10.1371/journal.pone.0009486.g007 recognize a carboxyl group linked to a shorter amino acid side chain.
Mammalian and plant eEF1A contain two EPG residues attached to Glu in domains II and III [24][25][26]. This is in contrast to T. brucei, where eEF1A is modified with a single EPG moiety in domain III only ( [27] and this report), despite the fact that the second potential modification site, Glu289, is strictly conserved between trypanosomes, mammals and plants. Our results demonstrate that this residue remains unmodified in T. brucei even if EPG attachment to Glu362 is prevented by point mutations.
Until today, EPG attachment has only been described in eEF1A. Two other amino acid side chain modifications show a similar selectivity for single proteins. In analogy to eEF1A, these proteins, eukaryotic translation initiation factor 5A (eIF5A) and elongation factor 2 (eEF2), are also involved in polypeptide elongation. First, hypusine (Ne-(4-amino-2-hydroxybutyl)-lysine) represents a post-translationally modified lysine residue in eIF5A ( [41]; reviewed by [42]). It has been described in eukaryotes and archaeabacteria, but is absent from bacteria. Recent reports have demonstrated that eIF5A associates in a hypusine-dependent way with actively translating ribosomes, suggesting that eIF5A may have a role in polypeptide elongation rather than, or in addition to, translation initiation [43][44][45]. In addition, the hypusine modification may be involved in correct localization of eIF5A within the cytosol [46]. The biosynthetic pathway leading to hypusine formation has been established (reviewed by [42]). Furthermore, a minimal structural domain of eIF5A necessary for hypusine modification has been identified [47].
Second, diphthamide, 2-(3-carboxyamido-3-(trimethylammonio)propyl)-histidine represents a post-translationally modified histidine residue found only in eEF2 ( [48]; reviewed by [49]). Similar to hypusine, diphthamide has been conserved throughout eukaryotic and Archaea evolution. It represents the target for ADP-ribosylation by bacterial toxins, rendering eEF2 inactive and causing inhibition of protein synthesis. However, its physiological role is still unclear, as several studies have demonstrated that eEF2 function is not directly dependent on diphthamide formation (reviewed by [49]). More recently, evidence for a direct role of diphthamide in cell function has been obtained when yeast strains expressing diphthamide-deficient eEF2 were shown to have increased -1 frame-shifting [50]. In addition, diphthamide was demonstrated to protect mutant mammalian cells from the action of the ribosome-inactivating protein, ricin [51]. The biosynthetic pathway leading to attachment of diphthamide to eEF2 has been completely solved in S. cerevisiae (reviewed by [49]). In contrast to EPG addition to eEF1A, the conserved amino acid residues around the modification site are critical for diphthamide formation in eEF2 [50,52].
Similar to hypusine and diphthamide, EPG is absent in bacteria [25]. However, whereas hypusine and diphthamide seem to be present in all eukaryotes, EPG was reported to be absent in eEF1A from S. cerevisiae [28]. In line with this observation, we were unable to label yeast eEF1A using [ 3 H]ethanolamine in vivo (results not shown). To test if yeast eEF1A lacks EPG because the biosynthetic pathway is deficient or if structure or sequence differences in yeast eEF1A prevent EPG attachment, we tried to express a tagged form of T. brucei eEF1A in S. cerevisiae, and vice versa. Unfortunately, despite several attempts we were unable to obtain expression of HA-tagged T. brucei eEF1A in different yeast strains. In contrast, HA-tagged eEF1A from S. cerevisiae was readily expressed in T. brucei procyclic forms, however, it was not modified with [ 3 H]ethanolamine (results not shown). At present, it is unclear if yeast eEF1A was not modified in T. brucei because it is not properly folded. However, our results showing that replacement of the primary sequence around the EPG attachment site of T. brucei eEF1A by the yeast sequence didn't affect EPG attachment in T. brucei, demonstrate that the different amino acid sequence of yeast eEF1A per se is not responsible for the lack of EPG modification in S. cerevisiae.
The pathways involved in EPG attachment to eEF1A, and its biological functions, are completely unknown. Our observations that domain III of eEF1A, expressed either as separate entity or fused to tagged proteins in T. brucei, is modified with EPG suggest that it may be feasible to identify interaction partners of EPGmodified eEF1A using tagged domain III to advance the general knowledge on the synthesis and role of this enigmatic protein modification using trypanosomes as model system.