INTRODUCTION

Top Abstract Introduction Procedures Results & Discussion References

The pathophysiology of acquired immune deficiency syndrome is usually understood as multifactorial (1, 2). There is a general consensus as to the important role played by an overactivation state of the immune system observed in HIV¹-infected individuals. It is known that modifications of glycosylation are associated with activation of immune cells (3-5). We have described an altered glycosylation status due mainly to hyposialylation of the two major lymphocyte molecules CD43 and CD45 in HIV-1-infected T cells (6, 7), and autoantibodies directed to CD43 with an altered glycosylation status have been found in HIV-1⁺ individuals (8, 9). Although the level of hyposialylation might be very important in HIV-1-infected T cells and seems to mostly affect O-glycans (10), the activity of Gal1-3GalNAc -2,3-sialyltransferases types I and II (hST3Gal I and hST3Gal II), which are the two principal sialyltransferases that act on O-glycan cores, appears to be conserved (11).

Among other pathways examined to explain the mechanism(s) of hyposialylation, secretion of two O-glycosylated and sulfated proteins (P47 and P40) has been observed in culture supernatants of latently HIV-1-infected CEM_LAI/NP cells (6, 7, 11).

Here, we report molecular cloning and initial characterization of the P47 protein, which we propose to name LSLCL (lymphocytic secreted long form of C-type lectin).

	EXPERIMENTAL PROCEDURES

Top Abstract Introduction Procedures Results & Discussion References

Cell Lines and Tissue Culture-- CEM, MOLT-4, HUT 78, Jurkat, SUP-T1, Raji, Daudi, HL-60, THP-1, U-937, and MRC5 cells were obtained from the American Type Culture Collection (Rockville, MD) and maintained in RPMI 1640 or basal medium Eagle (Life Technologies, Inc.) containing 5% fetal calf serum (Whittaker M.A. Bioproducts), supplemented with glutamine (2 mM), penicillin (100 units/ml), and streptomycin (50 µg/ml). CEM_LAI/NP are CEM cells infected by HIV-1 LAI strain. They harbor HIV-1 provirus but are latently infected and virus-nonproducing (6). The 7H1 clone was obtained as follows²: hypoxanthine aminopterin thymidine-sensitive CEM_LAI/NP cells were first selected as cells resistant to 8-azaguanine (15 µg/ml) after mutational treatment with ethyl methanesulfonate. Using polyethylene glycol, resistant cells were fused with human embryonic fibroblasts MRC5. Hybrid clones were selected as hypoxanthine aminopterin thymidine-resistant cells.

Preparation of Rabbit Serum Anti-P47-- Rabbits were immunized with the oligopeptide ARGAEREWEGGWGGAQEE (Arga), coupled to KLH protein (Neosystem Laboratoire, Strasbourg, France). The sequence of the peptide was determined by NH₂-terminal microsequencing of purified P47 as described below. A rabbit serum (Argara) was shown to precipitate two proteins (P47 and P40) from lysates and culture supernatants of secreting cells.

³⁵SO₄² Metabolic-labeling Assay and Precipitations-- Cells (4 × 10⁶/ml) were washed and incubated for 18 h in SO₄²-free RPMI 1640 medium containing SO₄² (ICN Biomedicals) (500 µCi/ml) and 5% dialyzed fetal calf serum. Cells were sedimented, and culture supernatants added with phenylmethylsulfonyl fluoride (1 mM, final concentration) (Sigma), aprotinin (10 mg/ml) (Sigma), and leupeptin (10 µg/ml) (Boehringer Mannheim). Supernatants were then mixed with 10 µl of immobilized streptavidin beads (Pierce) previously coated with 2 µg of the various biotinylated lectins, PNA (peanut agglutinin from Arachis hypogaea), Jacalin (from Artocarpus integrifolia), or GNA (from Galanthus nivalis), all purchased from EY Laboratories. 10 µl of protein A-Sepharose 4B beads (Sigma) precoated with 10 µl of rabbit serum were added for immunoprecipitations. After 2 h of agitation, precipitates were collected by brief centrifugation. The pellets were washed three times in 50 mM HEPES buffer, pH 7.5, containing 1% Nonidet P-40, 150 mM NaCl, 1 mM EGTA, and the antiprotease mixture as described above, and washed once with the same buffer containing 1 M NaCl. They were resuspended in SDS-polyacrylamide gel electrophoresis sample buffer (2% SDS, 1.7 M 2-mercaptoethanol) and boiled for 3 min. Samples were resolved by SDS-polyacrylamide gel electrophoresis (7.5% acrylamide). After intensification with Amplify (Amersham), the gels were vacuum dried and exposed to XAR5 Kodak films.

Purification and Microsequencing of the P47 Glycoprotein-- The culture of 7H1 cells was amplified up to one liter of cell suspension in RPMI 1640 medium containing 5% fetal calf serum. Then the cells were resuspended in RPMI 1640 without fetal calf serum and maintained at 37 °C for two days. After centrifugation of cells, supernatants were added with the mix of antiproteases as described above and with streptavidin-immobilized beads (1 ml/liter) precoated with biotinylated Jacalin. After 2 h of agitation, precipitates were collected by brief centrifugation. The pellets were washed 3 times in lysis-buffer and once with the same buffer containing 1 M NaCl. They were resuspended in SDS-polyacrylamide gel electrophoresis sample buffer (2% SDS, 1.7 M 2-mercaptoethanol), boiled for 3 min, and resolved by SDS-polyacrylamide gel electrophoresis (7.5% acrylamide). For NH₂-terminal sequencing, proteins were transferred to Problott membranes (Applied Biosystems). The peptides were detected by staining with 0.003% amido black, then the P47 band was carefully excised and treated on an Applied Biosystems model 470A sequencer. Internal sequences were obtained from peptides separated by high performance chromatography following cleavage of purified P47 with trypsin.

cDNA Cloning and Sequence Analysis-- A CEM_LAI/NP cell line cDNA library was constructed using pcDNA3 vector plasmid and a cDNA synthesis kit (both from Invitrogen) with an oligo(dT) (NotI) primer according to the manufacturer's recommendations. Approximately 4 × 10⁶ colonies of Escherichia coli DH5 transfected with this library were isolated and screened by using the oligonucleotide probe 5-GAGTGGGAGGGIGGITGGGGIGGIGCICAGGAGGAG-3, where inosine were incorporated to limit degeneracy. The sequence of this probe was based on the peptide sequence EWEGGWGGAQEE established by NH₂-terminal sequencing of P47. This probe was 5-end radiolabeled using T4 polynucleotide kinase (CLONTECH) and [-³²P]ATP (ICN Biomedicals). Membrane prehybridization was carried out at 60 °C for 3 h in 20 mM phosphate buffer, pH 7.5, containing 5 × SSC, 7% SDS, 10 × Denhardt's solution, and 1% salmon sperm DNA. Hybridization was achieved overnight at 60 °C. Then membranes were washed twice in 2 × SSC, 2% SDS solution for 20 min at 20 °C, once in 0.1 × SSC, 0.1% SDS solution for 20 min at 20 °C, and once in 0.1 × SSC, 0.1% SDS solution for 30 min at 48 °C, and thereafter exposed to XAR5 Kodak film for 1 day at 70 °C. Several types of recombinants were selected and sequenced on both strands by the dideoxynucleotide chain-termination method (12), using an ABI PRISM^TM dye terminator cycle sequencing ready reaction kit with AmpliTaq^® DNA polymerase, FS (Perkin Elmer), and an ABI PRISM 310 genetic analyzer (Perkin Elmer). Analysis was started on each strand using SP6 and T7 primers (Invitrogen) and extended with successive overlapping primers (20-base oligonucleotides) that were synthesized using an Applied Biosystem 381A DNA synthesizer. Sequences were analyzed using MacDNASIS (Hitachi Software).

Northern Blot Analysis-- Total RNA were extracted according to the guanidine isothiocyanate technique, described by Davis et al. (13). Poly(A)⁺ RNA were selected from 200 µg of total RNA by incubating with 100 mg oligo(dT) cellulose (Pharmacia Biotech Inc.) in 5 ml of TL buffer containing 20 mM Tris, pH 7.5, 0.5 M LiCl, 1 mM EDTA, and 0.1% SDS at room temperature with gentle shaking for at least 30 min. The beads were collected and washed twice with TL buffer and twice with the same buffer containing 0.15 M LiCl. The beads were then deposited on a Spin-X column (Costar, Cambridge, MA) for extensive washings with TL buffer containing 0.15 M LiCl, and thereafter the poly(A)⁺ RNA were eluted with 200 µl of 10 mM Tris buffer, pH 7.5 (containing 1 mM EDTA, 0.05% SDS) and precipitated with 2.5 vol of ethanol.

Computer Analysis-- Data base searches for nucleotides and amino acids sequences alignments were achieved using the BLAST program of the National Center for Biotech. Protein sequence analysis, motif search, and hydrophobicity were established with MacDNASIS programs.

	RESULTS AND DISCUSSION

Top Abstract Introduction Procedures Results & Discussion References

Since the alterations of sialylation that we have observed on HIV-1-infected CEM cells were principally involving O-glycans attached to the two major lymphocytic surface glycoproteins CD45 (6) and CD43 (7), we verified the presence of possible O-glycosylated proteins in the culture supernatants by using the lectins PNA, specific for the O-glycan core 1 nonsialylated disaccharides Gal1-3GalNAc (14) and Jacalin, specific for sialylated and nonsialylated O-glycans (15). In addition, an ³⁵SO₄² metabolic incorporation assay was used to display secreted glycoproteins. Jacalin only was able to precipitate a protein doublet (molecular mass, 47 and 40 kDa) (P47, P40) from supernatants of parental CEM cells culture (Fig. 1, lane 2), whereas this was achieved by both Jacalin or PNA from supernatants of HIV-1-infected CEM_LAI/NP cells culture (Fig. 1, lanes 7 and 8). This fact could be explained by hyposialylation status of CEM_LAI/NP cells that affects multiple O-glycosylated proteins, whereas parental cells are normally sialylated. The following has to be pointed out: (i) the nonreactivity of the lectin GNA (Fig. 1, lanes 3 and 9) that preferentially binds terminal mannose (16); (ii) a weak upward of the CEM doublet (Fig. 1, lane 2) probably due to the more numerous sialic acid molecules beared by the P40 and P47 proteins of these cells as confirmed by the nonreactivity of the lectin PNA (Fig. 1, lane 1); and (iii) a much more pronounced doublet P40-P47 from CEM_LAI/NP cells when compared with parental CEM doublet although equal quantities of total protein were deposited in each lane of the gel. This interesting result was further confirmed on several occasions and was also observed when culture supernatants from CEM_BCL cells (virus-producing CEM cells infected by HIV-1 strain BCL) (6) were compared with supernatants from parental CEM cells culture (not shown). This is currently being investigated in terms of increased activity of sulfotransferases and/or secretion regulation.

View larger version (41K): [in this window] [in a new window]   Fig. 1.   Precipitation of sulforadiolabeled P40-P47 by lectins. After 35SO42 incorporation for 18 h, precipitations were carried out from cell culture supernatants with the lectins PNA (specific for nonsialylated O-glycan core Gal1-3GalNAc), Jacalin (specific for sialylated or nonsialylated O-glycans), or GNA (specific for mannose residues), immobilized on agarose beads.

To obtain a sufficient material to allow sequencing of these proteins, hybrid cells were raised by fusion between CEM_LAI/NP cells and embryonic human fibroblasts MRC5 by using polyethylene glycol as described under "Experimental Procedures."² It was found that the clone 7H1 secreted very large amounts of P47 (Fig. 1, lanes 4 and 5) that was then purified using Jacalin lectin and then microsequenced. Thus, the sequences of an NH₂-terminal and two internal peptides were obtained (Fig. 2, underlined sequences). Thereafter, the cloning of P47 was achieved by screening a CEM_LAI/NP cDNA library constructed in pcDNA3 vector using a degenerate oligonucleotide based on the NH₂-terminal amino acid sequence. The longest isolated recombinant (1.4 kb) yielded a single open reading frame of 972 nucleotides encoding a protein of 323 amino acids (Fig. 2). Apart from an NH₂-terminal hydrophobic region of 21 residues located just upstream from the known microsequenced NH₂-terminal part of P47 (Figs. 2 and 3), no other potential hydrophobic transmembrane domain was found (Fig. 3). Thus this region very likely corresponded to the signal peptide of preprotein P47, which was purified from culture supernatants, and then appeared as a secreted protein with a Gly²¹Ala²² signal peptidase cleavage site. As shown on Fig. 2, several other domains were identified on P47: (i) a glutamic acid-rich NH₂-terminal domain localized between amino acids 22 and 90 with a stretch of 6 successive identical residues (Glu^84-89); (ii) an RGD triplet (Arg⁶¹-Gly-Asp⁶³) in the middle of the glutamic acid-rich domain; (iii) a leucine zipper consisting of six heptads, spanning Val¹⁰⁸-Ala¹⁴⁹; and (iv) a C-terminal C-type lectin domain that is known as a carbohydrate recognition domain (CRD) (17) spanning Gly¹⁷⁶-Phe³²³ and including most of the amino acids conserved for this type of functional domain, in particular the six cysteines (Fig. 2, closed arrows) found in the long form CRDs (18). In addition, a (Thr-Pro-Ser)-rich domain with 13 Thr/Pro/Ser out of 14 residues was found between amino acids 91 and 104 (Fig. 2).

View larger version (42K): [in this window] [in a new window]   Fig. 2.   Amino acid sequence of LSLCL (P47) and alignment with SCGF (25). Amino acids lacking in SCGF are marked with dashes. Regions of protein sequence determined by Edman degradation are underlined. The predicted signal peptide is boxed. The 6 Tyr residues are marked with an open arrow. The RGD domain is in bold characters. The Pro/Ser/Thr-rich domain is in brackets. The 6 Leu/Ile residues that compose the leucine zipper (at the structural position d) are marked , and the six Leu/Val intercalated (at the structural position a) by a filled circle. The C-type lectin domain comprises several conserved amino acids (18): six cysteines (closed arrow), and several other amino acid residues under scripted in bold characters, or marked with # (aliphatic residue),  (aromatic residue), (aliphatic or aromatic residue), Z (Glu or Gln), B (Asp or Asn), J (Glu, Gln, Asp, or Asn).

View larger version (24K): [in this window] [in a new window]   Fig. 3.   Hydrophobicity of LSLCL (P47). Hydropathy was calculated using the Kyte and Doolittle algorithm (37) with a span of 6 amino acid residues. The sole hydrophobic domain found on LSLCL corresponded to the first 21 amino acid residues of this protein.

The leucine zipper of P47 is particularly prone to form coiled coil structures (19), since it is composed of six heptads with a Leu/Val residue at structural position a and an Ile/Leu residue at position d (Fig. 4). The structure of P47 could thus be related to that of collectins, which are secreted proteins containing a CRD connected to a collagen-like -helix suitable to form coiled coils (17). The NH₂-terminal portion of P47, spanning Ala²²-Ala⁹⁰, is rich in glutamic acid (30% of residues) and has a predicted pI of 3.95. This is reminiscent of the secretory eosinophil major basic protein, which also contains a C-terminal C-type lectin and a glutamic acid-rich NH₂-terminal part with a pI of 3.9 (20) that is sequestered into secretory granules before release.

View larger version (19K): [in this window] [in a new window]   Fig. 4.   Schematic representation of the proposed amphipathic leucine zipper helix corresponding to the portion Val108-Ala149 of P47. In the sequence Val108-Ala149 of P47 shown at the bottom, Val108 is renumbered as Val1, and so on the aliphatic residues Leu/Ile/Val are renumbered according to their position in the analyzed portion. Residues are displayed on a helical wheel whose the angles (102.9°) are those found in -helices. Since there are 3.5 residues per turn in the helix within coiled coils, this allows the position of residues to repeat after two turns (or 7 amino acids). Here, aliphatic Leu/Ile/Val residues are found at the same position a and d and are thus aligned on the same face of a helix. The other faces b, c, e, f, and g are hydrophilic. Coiled coil structures are constituted in the manner of knobs-into-holes by packing of hydrophobic residues (knobs) from a helix into holes of the facing helix provided in the middle of 4 amino acid side chains (19).

At the time we were cloning P47, there was no significant similarity to any other protein accessible from data banks. The best alignments were found to be 29.2% with tetranectin-like protein (21) and 24.8% with tetranectin (22), provided by the restricted CRD sequence Cys¹⁷⁷-Arg²²⁸ of P47 (Fig. 4). Tetranectin is a plasminogen kringle 4-binding protein with a CRD joined to an -helical coiled coil structure (23, 24). During preparation of this manuscript, Hiraoka et al. (25), using a differential technique with subtractive probe, published the cloning of SCGF, a protein that very closely matched P47. However, SCGF and P47 appear to be different because SCGF cDNA lacks 234 nucleotides encoding a peptide of 76 amino acids located in the CRD found on P47. In particular, the Cys²⁰⁴ of P47 is absent from SCGF (Fig. 2). Surprisingly, these authors did not mention any of the protein domains (RGD, leucine zipper, O-glycan sites, or CRD) that are present on SCGF and P47 (Fig. 2).

Sulfation of P47 was not likely to occur on tyrosine residues. Indeed, tyrosyl protein sulfotransferase is known to recognize tyrosines in exposed protein domains containing acidic amino acids (26, 27) that were not found in the surroundings of the 6 Tyr residues present on P47 (Fig. 2, open arrows).

P47 did not show any potential N-glycosylation sites (Asn-Xaa-Ser/Thr). This fact correlated well with the nonreactivity of P47 with the lectin GNA (Fig. 1). On the other hand, the (Thr-Pro-Ser)-rich domain is particularly suited for O-glycans attachment (28) and could explain the binding of P47 with the lectins Jacalin and PNA (Fig. 1). In this context, and since sulfation of P47 did not implicate tyrosine residues nor N-glycan, the O-glycans remained the most probable or even the sole possible sulfation sites, as it has been demonstrated on other proteins such as the cell surface mucin of mammary adenocarcinoma cells (29), gpMEL-14 (LECAM-1) (30), and leukosialin (CD43) (31, 32). The molecular mass of P47, calculated on the basis of its amino acid sequence subtracted the signal peptide, was 33,532 Da. We conclude that O-glycan side chains added with sulfate residues account for approximately 13,500 Da in the mature secreted form of P47, which thus could be regarded as a mucin-like protein.

It was tempting to compare P47 to a bifunctional molecule with an RGD domain located in the very hydrophilic NH₂-terminal portion and known to interact with integrins (33), and a C-terminal C-type lectin domain well known to bind glycoconjugates (17), separated from each other by a neck region consisting of an extended O-glycosylated hinge domain and a leucine zipper. In this context it seemed interesting to select, among numerous proteins harboring a CRD, the two proteins tetranectin (22) and low affinity receptor for IgE (CD23) (34) for alignment with P47. Indeed, these molecules showed a neck leucine-rich region, upstream of their CRD, suitable for coiled coil associations. As shown on Fig. 5, there are several conserved amino acids, especially the six cysteines and surrounding residues in CRD, but also numerous aliphatic amino acids in the portion corresponding to the leucine zipper of P47 that is aligned with the leucine zipper of CD23 as well as with the E2 -helix that governs the coiled coil formation of tetranectin (24).

View larger version (57K): [in this window] [in a new window]   Fig. 5.   Alignment of the amino acid sequences of CD23, LSLCL (P47), and tetranectin. The residues conserved for three proteins are under scripted in bold capital letters and in lightface capital letters when conserved for two proteins. The periodic alignment of aliphatic residues (especially Leu) in the neck region, which is marked by brackets, has to be noted. The sequence with the best homology between LSLCL (P47) and tetranectin is limited by arrows.

Northern blot analysis showed a single band at approximately 1.42 kb (Fig. 6, A and C), even when a better resolution was obtained with mRNA deposits in 1-cm large wells (Fig. 6B). Among the various cell lines tested, all T and B lymphoblastoid (CEM, MOLT-4, HUT 78, Jurkat, SUP-T1, Raji, Daudi) cells highly expressed P47, whereas monocytoid U-937 and THP-1 expressed very faintly, and embryonic fibroblasts MRC5 did not (Fig. 6A). Tissue-specific expression was analyzed using commercially prepared Northern blots. A very weak band was seen in most tissues (Fig. 6C, lanes 7-14). Lymphoid tissues, notably bone marrow, appeared to better express P47 (Fig. 6C, lanes 1-6). Since the sequence studied in this report is likely of lymphoblast/lymphocyte origin, it is quite possible that the weak Northern signals seen in essentially all tested tissues is the result of contaminating blood that it is difficult to entirely remove during preparation of tissues.

View larger version (49K): [in this window] [in a new window]   Fig. 6.   Northern blot analysis of P47 expression in various cell lines and tissues. Poly(A)+ RNA were electrophoresed (5 µg/lane), transferred to nylon membranes, and then probed with a -32P-random-labeled fragment of 1.4 kb excised from the full-length pcDNA3-P47 recombinant. A, lane 1, parental CEM; lane 2, latently HIV-1-infected CEMLAI/NP; lane 3, HUT-78; lane 4, MOLT-4; lane 5, Jurkat; lane 6, SUP-T1; lane 7, Raji; lane 8, Daudi; lane 9, HL-60; lane 10, THP-1; lane 11, U-937; lane 12, MRC5. B, lane 1, parental CEM; lane 2, CEMLAI/NP. C, lane 1, fetal liver; lane 2, bone marrow; lane 3, PBL; lane 4, thymus; lane 5, lymph node; lane 6, spleen; lane 7, heart; lane 8, brain; lane 9, placenta; lane 10, lung; lane 11, liver; lane 12, skeletal muscle; lane 13, kidney; lane 14, pancreas.

The transcription rate of P47 appeared to be identical in both HIV-1-infected CEM_LAI/NP and parental CEM cells, as confirmed on several occasions (Fig. 6 A and B). This result is important to consider while compared with the apparent differential expression of P47 when examined after ³⁵SO₄² incorporation (Fig. 1). This phenomenon is currently being investigated in terms of increased activity of sulfotransferases and/or secretion regulation.

Several discrepancies were noted between P47 (this study) and SCGF (25). Whereas expression of SCGF was found in fibroblasts, P47 was not detected in similar cells (Fig. 6A). On the contrary, SCGF mRNAs were not detected in MOLT-4 cells, although we did find P47 in these cells (Fig. 6A). We showed two closely migrating sulfated proteins with a molecular mass of 47 and 40 kDa that were precipitated from culture supernatants by lectins (Fig. 1). The 40-kDa species is very likely a derivative of P47, since the two species were precipitated by rabbit serum Argara raised against an NH₂-terminal peptide of P47 (not shown). Although a single mRNA species (1.42 kb) was seen after probing with a full-length cDNA corresponding to P47 (Fig. 6), it is tempting to think that P40 could correspond to SCGF. However, SCGF was claimed to have a molecular mass of 29 kDa according to (i) calculation from its amino acid sequence and (ii) partial purification of SCGF monomer by chromatography as an elution peak of 29 kDa (25). Nevertheless, it has to be pointed out that SCGF contains the same Thr/Pro/Ser-rich domain as P47 and that it is very likely O-glycosylated, thus leading SCGF to a molecular mass >29,000 Da. In addition, it is difficult to consider SCGF as a spliced form of P47. Indeed, it can be deduced from comparison of gene structures of C-type CRDs (35) that the most probable exon that might be removed from P47 to generate SCGF could correspond to the internal peptide Glu²²¹-Phe²⁶² and not to the differential peptide Ala¹⁹⁸-Gln²⁷⁵ present on P47 only. Therefore, P40 is very likely a processed form of P47, and there is no obvious explanation for differential transcription of LSLCL (P47) and SCGF.

SCGF has been shown to exhibit burst promoting activities when associated with growth factors such as erythropoietin or granulocyte/macrophage colony-stimulating factor (25). The RGD domain present on P47 and SCGF might explain such properties, since this type of motif is known to interact with integrins that activate signal transduction pathways coordinated with responses to growth factors (36). However, Hiraoka et al. (25) claimed the activities of SCGF to be accounted for by a protein purified from culture supernatants as an elution peak of 29 kDa. They have cloned SCGF using a differential technique with subtractive probe and claimed this protein to be the active species of 29 kDa. However, analogy between P47 and SCGF is sufficient to allow similar molecular masses that might be of 40-47 kDa rather than 29 kDa. In additon, the various functional domains (C-type lectin, leucine zipper, and RGD) are reminiscent of collectins that are known as anti-infectious agents (17) rather than of growth factors. Hiraoka et al. (25) conclude their paper as follows: "It is presently unclear, however, whether KPB-M15-CM contains another SCGF-like factor." We think that cloning of SCGF might be fortuitous and that the growth factor activity of culture supernatants from KPB-M15 cells remains to be determined.

Since P47 is composed of several domains that allow for prediction of various interactions, counterparts of this complex molecule remain to be identified to understand its exact role(s).