In Vivo Interaction between the Polyprenol Phosphate Mannose Synthase Ppm1 and the Integral Membrane Protein Ppm2 from Mycobacterium smegmatis Revealed by a Bacterial Two-hybrid System* □ S

Dolichol phosphate-mannose (Dol- P -Man) is a mannose donor in various eukaryotic glycosylation pro-cesses. So far, two groups of Dol- P -Man synthases have been characterized based on the way they are stabilized in the endoplasmic reticulum membrane. Enzymes belonging to the first group, such as the yeast Dpm1, are typical integral membrane proteins harboring a transmembrane segment (TMS) at their C terminus. In contrast, mammalian Dpm1, enzymes of the second group, lack the typical TMS and require the association with the small hydrophobic proteins Dpm3 to be properly stabilized in the endoplasmic reticulum membrane. In Mycobacterium tuberculosis , the Polyprenol- P -Man synthase Mt Ppm1 is involved in the biosynthesis of the cell wall-associated glycolipid li-poarabinomannan. Mt Ppm1 is composed of two domains. The C-terminal catalytic domain is Bacterial Strains and Growth Conditions— All cloning steps were performed in E. coli XL1-Blue (Stratagene, La Jolla, CA). M. smegmatis mc 2 155 was a generous gift from NY and was transformed as described previously Recombinant clones were selected on Middlebrook 7H10 agar supplemented with oleic acid-albumin-dextrose-catalase enrich-ment (Difco, Detroit, MI) containing 25 (cid:1) g/ml kanamycin (Sigma). Liquid cultures of recombinant M. smegmatis were grown at 37 °C in Luria-Bertani (LB) broth (Difco) supplemented with 25 (cid:1) g/ml kanamy- cin and 0.05% Tween 80. Liquid cultures of E. coli (pUC8) and E. coli (pUC8- Mt Ppm1/D2) were grown in LB broth at 37 °C with 100 (cid:1) g/ml ampicillin to an optical density (OD) at 600 nm of 0.4 and induced for 4 h with 1 m M isopropyl- (cid:2) - D -thiogalactopyranose. Large scale cultures of bacteria were grown as described above, harvested by centrifugation, washed with phosphate-buffered saline, and stored at (cid:2) 20 °C until further use. E. coli DHP1 is an adenylate cyclase deficient ( cya ) deriv-ative extracted as described previously (13). Thin-layer chromatography us- ing 10% of the reaction mixture were conducted on aluminum-backed plates of Silica Gel 60 F 254 (Merck) using CHCl 3 /CH 3 OH/NH 4 OH/H 2 O (65/25/0.4/3.6). Autoradiograms were obtained by exposing chromato-grams to Kodak X-Omat AR films at (cid:2) 70 o for 4–5 days. In parallel, autoradiograms were exposed to a PhosphorImager screen, and radioactivity was quantified using a PhosphorImager detector (Storm, Am- ersham Biosciences).

creased the transferase activity of the C-terminal domain without displaying catalytic activity by itself (13).
Surprisingly, in related mycobacterial species, such as Mycobacterium leprae, Mycobacterium avium, and Mycobacterium smegmatis, orthologs of the two domains are encoded by two distinct open reading frames organized as an operon. This observation suggests that MtPpm1 has resulted from the fusion of two ancestral neighboring open reading frames, still separated in some related mycobacterial species. According to the "Rosetta stone theory" (14,15), the presence in the M. smegmatis genome of adjacent genes encoding MsPpm1 and MsPpm2 that are both homologs of MtPpm1 encoded by a single gene in M. tuberculosis suggests that MsPpm1 and MsPpm2 interact with each other to exert a function similar to that of MtPpm1.
In this report, we show that MsPpm2 is an integral membrane protein and, using a bacterial two-hybrid system (16), we demonstrate that the synthase MsPpm1 binds to MsPpm2 in vivo. As observed with mammalian Dol-P-Man synthases, MsPpm1 also lacks the characteristic hydrophobic C terminus. Thus, the MsPpm1-MsPpm2 interaction is reminiscent of that of the mammalian Dpm1 with Dpm3 (8,9). In contrast, MsPpm1 is functionally active when expressed in Escherichia coli, similarly to the S. cerevisiae group of Dol-P-Man synthases (7,17). As a consequence, MsPpm1 may constitute a new intermediate group of Polyprenol-P-Man synthases.
Bacterial Strains and Growth Conditions-All cloning steps were performed in E. coli XL1-Blue (Stratagene, La Jolla, CA). M. smegmatis mc 2 155 was a generous gift from W. R. Jacobs, Albert Einstein College of Medicine, Bronx, NY (18) and was transformed as described previously (19). Recombinant clones were selected on Middlebrook 7H10 agar supplemented with oleic acid-albumin-dextrose-catalase enrichment (Difco, Detroit, MI) containing 25 g/ml kanamycin (Sigma). Liquid cultures of recombinant M. smegmatis were grown at 37°C in Luria-Bertani (LB) broth (Difco) supplemented with 25 g/ml kanamycin and 0.05% Tween 80. Liquid cultures of E. coli (pUC8) and E. coli (pUC8-MtPpm1/D2) were grown in LB broth at 37°C with 100 g/ml ampicillin to an optical density (OD) at 600 nm of 0.4 and induced for 4 h with 1 mM isopropyl-␤-D-thiogalactopyranose. Large scale cultures of bacteria were grown as described above, harvested by centrifugation, washed with phosphate-buffered saline, and stored at Ϫ20°C until further use. E. coli DHP1 is an adenylate cyclase deficient (cya) derivative of DH1 (F Ϫ , glnV44(AS), recA1, endA1, gyrA96 (Nalr), thi1, hsdR17, spoT1, rfbD1) (16). Protein-protein interactions leading to the cytoplasmic production and assembly of a functional adenylate cyclase in E. coli DHP1 were detected by the ability to ferment maltose. Thus, clones were checked for their ability to form red colonies at 30°C on freshly prepared MacConkey agar plates containing 1% maltose and supplemented with 25 g/ml kanamycin, 25 g/ml chloramphenicol, and 100 g/ml ampicillin when appropriate.
A similar strategy was used for Msppm1 and Msppm2. Primers N°200 (5Ј-AAGGTACCGAGCGTCCCAGGTGAACGTGA-3Ј) and N°201 (5Ј-CAAAGCTTACCACGCCCCTGGCCCGGTCGA-3Ј) were used to amplify Msppm1 from M. smegmatis chromosomal DNA. The 805-bp PCR product was cleaved by KpnI and HinDIII and inserted into pT18 digested previously with the same enzymes. The resulting plasmid was termed pT18-MsPpm1. Primers N°198 (5Ј-CCCTGCAGTCACCGAT-GACGACCCCCTC-3Ј) and N°199 (5Ј-GGGTCGGGCGGACGGTG-GCAGGA-3Ј) were used to amplify a 1853-bp fragment of Msppm1. Subsequently, the DNA fragment was cut with PstI and BamHI and then inserted into pT25 cut with the same enzymes to yield pT25-MsPpm2. All inserts were verified by DNA sequencing.
Alkaline Phosphatase Activity Assay-Alkaline phosphatase activity was assayed in M. smegmatis and in E. coli by measuring the rate of p-nitrophenyl-phosphate (Sigma) hydrolysis in intact cells as described previously (22,23). Enzymatic reactions were performed in triplicate in the dark at 37°C. Reactions were stopped with 100 l of 1 M KH 2 PO 4 , 0.1 M EDTA (pH 8.0), and the OD at the appropriate wavelength was measured. The activity is expressed in arbitrary units. Units of activity for M. smegmatis ϭ A 420 nm ϫ mg of protein Ϫ1 ϫ min Ϫ1 (23), units of activity for E. coli ϭ (A 420 nm Ϫ (1.75 ϫ A 550 nm )) ϫ 10 3 ϫ min Ϫ1 ϫ A 600 nm Ϫ1 ϫ ml of culture Ϫ1 (24).

Preparation of Enzyme Fractions and Polyprenol Phosphate Mannose
were grown, washed, and sonicated as described previously (13). For further fractionation, the lysate of E. coli(pT18-MsPpm1) was centrifuged at 27,000 ϫ g for 20 min at 4°C. The membrane fraction was obtained by further centrifugation of the 27,000 ϫ g supernatant at 100,000 ϫ g for 1 h at 4°C. The supernatant was carefully removed, and the membranes were gently resuspended in buffer 50 mM MOPS (adjusted to pH 8.0 with KOH), 5 mM ␤-mercaptoethanol, 10 mM MgCl 2 at a protein concentration of 20 mg/ml. Protein concentrations were determined using the BCA protein assay reagent kit (Pierce). Reaction mixtures for assessing [ 14  Exogenous lipid monophosphate substrates (C 95 , dolichol monophosphate) were added to the reaction mixtures at a final concentration of 0.125 mM in 0.25% CHAPS. The reaction mixtures were then incubated at 37°C for 30 min. The reaction was terminated, and the lipids were extracted as described previously (13). Thin-layer chromatography using 10% of the reaction mixture were conducted on aluminum-backed plates of Silica Gel 60 F 254 (Merck) using CHCl 3 /CH 3 OH/NH 4 OH/H 2 O (65/25/0.4/3.6). Autoradiograms were obtained by exposing chromatograms to Kodak X-Omat AR films at Ϫ70 o for 4 -5 days. In parallel, autoradiograms were exposed to a PhosphorImager screen, and radioactivity was quantified using a PhosphorImager detector (Storm, Amersham Biosciences).

Computer-predicted Topology of Mycobacterial Polyprenol-P-
Man Synthases-Consistent with the previously shown membrane association of the Polyprenol-P-Man synthase activity in M. tuberculosis (13), MtPpm1 contains TMS located in its Nterminal domain (D1). Since MtPpm1 uses GDP-Man as a substrate (11), its catalytic domain (D2) would be expected to be cytoplasmic where GDP-Man is readily available. To predict the topology of MtPpm1, we used seven different programs available on the internet (TMHMM, HMMTOP, DAS, TMPRED, MEMSAT2, TOPPRED, and PHDHTM) since the reliability of topology predictions increases substantially when different prediction methods are compared (25). When applied to MtPpm1, all seven programs converged to the presence of several TMS, most of which was located in the N-terminal half of the protein (D1), strongly suggesting that MtPpm1 is an integral membrane protein (Fig. 1A). However, the algorithms did not converge on the precise number of TMS and on the orientation of the catalytic domain (D2) of MtPpm1.
In other mycobacterial species, such as M. leprae, M. avium, and M. smegmatis, the orthologs of the two MtPpm1 domains, referred to as D1 and D2, are encoded by separate genes. The orthologs of the catalytic domain (D2) of MtPpm1 are named Ppm1, and the orthologs of the N-terminal non-catalytic domain of MtPpm1 (D1) are named Ppm2. Application of the topology prediction algorithms on Ppm2 from M. smegmatis (MsPpm2) (Fig. 1B), M. leprae (MlPpm2), and M. avium (MaPpm2) (data not shown) strongly suggests that these proteins, similar to the MtPpm1/D1 domain, are anchored in the bacterial membrane. In contrast, none of the seven methods predicted the presence of TMS in MsPpm1, MlPpm1, or MaPpm1, nor were signal sequences predicted in the N-terminal part of these proteins. These results suggest a cytoplasmic localization for these proteins and, by analogy, a cytoplasmic localization for the catalytic domain of MtPpm1.
Experimental Determination of the Cellular Localization of the Catalytic Domain of MtPpm1-In Gram-negative bacteria, the topology of integral membrane proteins has largely been studied using PhoA fusions since the sites at which alkaline phosphatase has high enzymatic activity normally correspond to periplasmic domains of the membrane protein. In a previous report, we have shown that PhoA fusions can be used in mycobacteria to determine the cytoplasmic or extracytoplasmic location of proteins or protein domains (22). To determine the cellular location of the catalytic domain of MtPpm1, we made use of PhoA fusions to various parts of the protein. The phoA gene was thus inserted into pMV261-MtPpm1, a plasmid previously shown to allow for overexpression of Mtppm1 in mycobacteria (13). The phoA gene was inserted in-frame with Mtppm1 either before (codon 465) or after (codon 600) the last predicted TMS (amino acids 509 -527) of MtPpm1 (Fig. 1A). The resulting constructs, named pMV261-MtPpm1/PhoA 465 and pMV261-MtPpm1/PhoA 600 , respectively, were introduced into M. smegmatis mc 2 155, and the recombinant clones were tested for their ability to hydrolyze X-phosphate. When compared with untransformed bacteria, M. smegmatis It is thus likely that MsPpm1 and MsPpm2 are able to directly interact with each other. Among the general methodologies to identify interactions between proteins, the yeast two-hybrid system represents the most powerful in vivo approach. However, the detection of the interaction between MsPpm1 and MsPpm2 using a yeast two-hybrid system may be impaired by the association of MsPpm2 with the membrane. Therefore, to test whether MsPpm2 can interact with MsPpm1, a recently described bacterial two-hybrid system (16) was used, which is suitable to study protein interactions even if one of the partners is membrane-associated. In this system, the interaction of two proteins results in functional complementation between two domains of the adenylate cyclase (CyaA) from Bordetella pertussis, leading to cAMP synthesis. As a soluble regulatory molecule, cAMP is then able to activate cAMP-dependent transcriptional events, which can be easily monitored (16).
The entire MsPpm1 was thus genetically fused to the 18-kDa domain of CyaA to produce pT18-MsPpm1, and MsPpm2 was fused to the 25-kDa domain of CyaA to produce pT25-MsPpm2. The cyclase-deficient E. coli strain DHP1 was co-transformed with the two constructs, and the resulting recombinant colonies were tested for their ability to metabolize maltose on McConkey agar plates supplemented with ampicillin and chloramphenicol. As shown in Fig. 2A, E. coli DHP1 containing both fusions were able to metabolize maltose and appeared red, whereas colonies containing pT18-MsPpm1 with pT25 or pT25-MsPpm2 with pT18 were not able to metabolize maltose and remained white on McConkey-maltose plates. These results demonstrate specific interactions between MsPpm1 and MsPpm2.
Interaction between the Two Domains of MtPpm1 Is Conserved-Since MsPpm1 and MsPpm2 correspond to domains D2 and D1 of MtPpm1, respectively, encoded by a single gene, we wanted to test whether D1 and D2 can also interact with each other, even in the absence of a covalent link between them. We therefore genetically disconnected the two domains and co-expressed them within the bacterial two-hybrid system. MtPpm1/D2 was fused to T18, and MtPpm1/D1 was fused to T25. After introduction of both constructs into E. coli DHP1, the colonies were found to be able to metabolize maltose on McConkey agar plates, in contrast to the cells containing pT18-MtPpm1/D2 and pT25 or pT25-MtPpm1/D1 and pT18 (Fig. 2B). This result demonstrates that the two domains D1 and D2 of MtPpm1 can specifically interact with each other, even if they are not covalently linked.
Trans-species in Vivo Interaction-We have previously shown that MtPpm1/D1 has no Dol-P-Man synthase activity but increases Dol-P-Man production when its gene is overexpressed in M. smegmatis, suggesting that the M. tuberculosis MtPpm1/D1 may interact with the M. smegmatis MsPpm1 enzyme and may thereby increase its enzymatic activity. To test this hypothesis, pT18-MsPpm1 and pT25-MtPpm1/D1 were both introduced into E. coli DHP1. The recombinant bacteria showed a red color on McConkey-maltose plates (Fig. 2C), demonstrating that MsPpm1 can interact with MtPpm1/D1. Vice versa, when pT18-MtPpm1/D2 and pT25-MsPpm2 were introduced into E. coli DHP1, the recombinant bacteria also displayed a red color on McConkey-maltose plates (Fig. 2C).
Partial Determination of the Topology of T25-MsPpm2 in E. coli-MsPpm2 and MtPpm1/D1 were both predicted to con-tain many TMS, suggesting that they are integral membrane proteins. Moreover, PhoA fusions demonstrated that MtPpm1 is associated with the membrane. Here, PhoA fusions were used to determine the global topology of MsPpm2. TnTap (21) was randomly transposed in vitro into pT25-MsPpm2 using the EZ::TN transposase (Epicentre, Madison, WI), and the recombinant plasmids were introduced into E. coli CC118. Blue colonies obtained as a result of PhoA activity were selected on agar plates containing X-phosphate. The insertion positions of the mini-transposon were determined by DNA sequencing for 25 blue colonies that contained the phoA cassette in the Msppm2 gene. Activities were measured according to Manoil (24). The results confirm that MsPpm2 is an integral membrane protein and provide some insight into the topology of the protein (Fig. 3). The blue color obtained by insertion of phoA after codon 495 is similar to the results obtained after insertion of phoA in the corresponding region of MtPpm1, indicating the extracytoplasmic location of this region for both proteins. Four of the seven algorithms predicted the existence of a TMS at approximately residue 450 (Fig. 1B). However, the PhoA fusion data indicate that this TMS is unlikely since PhoA fusions upstream and downstream of residue 450 resulted in strong phosphatase activity. No TnPhoA insertions leading to a PhoApositive phenotype were obtained in the region between Ala 160 and Val 222 . However, based on the phosphatase activity of the PhoA insertions in the Gly 88 to Gly 116 region and the unanimous prediction of TMS Tyr 117 to Ser 134 and Ala 160 to Ala 139 , this region of MsPpm1 is likely to be periplasmic. It is possible that insertions of TnTap in this region lead to unstable hybrid proteins. One colony with a very low phoA activity (phoA activity ϭ 1.15 units) obtained after transposition revealed an insertion of TnTap in the region corresponding to Trp 239 to Ala 252 , confirming that this portion of the protein is cytoplasmic (Fig. 3). Interestingly, the topology determined by the PhoA fusions together with the TMS predicted by all the algorithms used indicate that only very few amino acids of MsPpm2 are located at the cytoplasmic side of the membrane and may thus be available for the interaction with MsPpm1 (Figs. 3  and 5).

DISCUSSION
Due to the lack of a transporter, GDP-Man, a widely used mannosyl donor, is generally unable to cross biological membranes. Consequently, the enzymes that use GDP-Man are predicted to be located in the cytosol. In eukaryotes, Dol-P-Man synthase (EC 2.4.1.83) catalyzes the transfer of Man from GDP-Man to dolichol monophosphate, forming Dol-P-Man, which is subsequently translocated through the endoplasmic reticulum membrane to be used as a mannosyl donor in the lumen of the endoplasmic reticulum. The Dol-P-Man synthases are usually associated with the cytoplasmic side of the membrane, and two different mechanisms have been proposed for this membrane association (7,17). In S. cerevisiae, U. maydis, and T. brucei Dpm1 contains a C-terminal hydrophobic domain proposed to anchor the enzyme into the lipid bilayer of the membrane. In contrast, mammalian Dol-P-Man synthases are associated with the endoplasmic reticulum membrane through specific interactions with the hydrophobic protein Dpm3, in turn, stabilized by Dpm2 (8,9). In mycobacteria, the biosynthesis of cell wall glycoconjugates, such as lipoarabinomannan, also requires the transfer of Man residues through the bacterial membrane. The Man carriers in mycobacteria are Polypre-nol-P-Man, isoprenoid derivatives that are shorter than Dol-P-Man and contain an unsaturated ␣-isoprene residue.
In this study, we found that the synthesis of Polyprenol-P-Man in M. smegmatis is also mediated by two proteins, one being the catalytic MsPpm1 and the other being the "helper" protein MsPpm2. MsPpm1 and MsPpm2 interact in vivo when co-produced in E. coli, and MsPpm2 enhances the catalytic activity of MsPpm1. Thus, as for mammalian Dpm1, MsPpm1 lacks the C-terminal hydrophobic domain but interacts instead with the integral membrane protein MsPpm2. This finding suggests a role for the two-domain structure of Ppm1 from M. tuberculosis, in which the first domain would anchor the protein into the bacterial membrane. As shown by the bacterial two-hybrid assay, the two domains of MtPpm1 are able to interact with each other even in the absence of covalent linkage FIG. 5. Proposed comparative model for the stabilization of Dol-P-Man synthases and Polyprenol-P-Man synthases in membranes. As shown in A, the M. smegmatis Polyprenol-P-Man synthase MsPpm1 interacts with membrane-associated MsPpm2. LAM, lipoarabinomannan. As shown in B, the M. tuberculosis Polyprenol-P-Man synthase is a single-component enzyme harboring multiple TMS at the N terminus. As shown in C, the S. cerevisiae Dol-P-Man synthase is a single-component enzyme with a TMS near the C terminus (7). As shown in D, human Dol-P-Man synthase consists of three subunits. Dpm2 associates with Dpm3, which in turn stabilizes the catalytic subunit Dpm1 (8,9). between them. Moreover, the interaction between the two domains of MtPpm1 independently produced in E. coli results in enhanced Polyprenol-P-Man production. In addition, the capacity of interaction has been conserved between two mycobacteria that differ with respect to the genetic structure of their ppm1 genes, as illustrated by the capacity of MsPpm1 and MsPpm2 to interact with MtPpm1/D1 and MtPpm1/D2, respectively.
By analogy with the S. cerevisiae Dpm1 (ScDpm1), MtPpm1 contains a hydrophobic region responsible for its attachment to the bacterial membrane. However, in M. tuberculosis, the hydrophobic region is located in the N-terminal portion of the protein and is larger than the C-terminal hydrophobic domain of ScDpm1. On the other hand, MsPpm1 may be, to some extent, compared with the human Dpm1 as they both lack the C-terminal transmembrane domain and interact with a polypeptide (MsPpm2 versus human Dpm3) that is localized in the membrane. In contrast, MtPpm1/D2 and MsPpm1 are active when produced in E. coli, whereas human Dpm1 is not. Thus, as illustrated in Fig. 5, we propose to extend the family of Polyprenol-P-Man synthases by including two new members that use two original strategies of membrane association.
MsPpm1 is a cytoplasmic soluble protein, whereas MsPpm2 is a hydrophobic protein containing of 6 -8 predicted TMS. Due to the hydrophobic nature of MsPpm2, we have not been able to purify MsPpm2 to test physical interactions with MsPpm1 in vitro. However, the interactions could be studied in vivo by using a bacterial two-hybrid system based on the functional complementation between two domains of the cyaA gene from B. pertussis, which catalyzes the production of cAMP from ATP. Both the substrate and the product of the reaction are soluble in the cytosol, even if one of the partners of the two-hybrid system is anchored in the inner face of the membrane. This approach allowed us to demonstrate a specific interaction between the transmembrane protein MsPpm2 and its soluble catalytic partner MsPpm1. The bacterial two-hybrid system did not abolish the enzymatic activity of MtPpm1/D2 or MsPpm1, nor did it affect the enhancing effects of MtPpm1/D1 and MsPpm2 on the enzymatic activities of MtPpm1/D2 and MsPpm1, respectively.
Based on both the TMS prediction methods and the PhoA insertions, we deduced a topological model of MsPpm2 (Fig. 3). Surprisingly, the proposed model suggests that only few amino acids of MsPpm2 are located at the cytoplasmic side of the membrane and may thus be available for the interaction with MsPpm1. Mutagenesis of residues of the N-terminal tail and of the first and second intracellular predicted loops of MsPpm2 may help to identify amino acids implicated in the interaction with MsPpm1.
The physical interaction demonstrated between MsPpm1 and MsPpm2 and the existence of MtPpm1 in M. tuberculosis containing the two domains fused into a single protein are an illustration of the Rosetta stone theory (15). This theory proposes that interactions between a protein of unknown function and a well characterized protein suggest that the function of the former is somewhat related to that of the latter. In accordance with this theory, the implication of MsPpm2 in the Polyprenol-P-Man synthesis pathway may have been inferred from the architecture of MtPpm1 (14).
In conclusion, the interaction of MsPpm1 with MsPpm2 allows us now to understand how the Polyprenol-P-Man synthase activity of M. smegmatis is associated with the membrane in the absence of a TMS in MsPpm1. Therefore, M. smegmatis uses a membrane-targeting strategy similar to that of the mammalian Dol-P-Man synthases.