Structure and Function of the XpsE N-Terminal Domain, an Essential Component of the Xanthomonas campestris Type II Secretion System*

Secretion of fully folded extracellular proteins across the outer membrane of Gram-negative bacteria is mainly assisted by the ATP-dependent type II secretion system (T2SS). Depending on species, 12-15 proteins are usually required for the function of T2SS by forming a trans-envelope multiprotein secretion complex. Here we report crystal structures of an essential component of the Xanthomonas campestris T2SS, the 21-kDa N-terminal domain of cytosolic secretion ATPase XpsE (XpsEN), in two conformational states. By mediating interaction between XpsE and the cytoplasmic membrane protein XpsL, XpsEN anchors XpsE to the membrane-associated secretion complex to allow the coupling between ATP utilization and exoprotein secretion. The structure of XpsEN observed in crystal form P43212 is composed of a 90-residue α/β sandwich core domain capped by a 62-residue N-terminal helical region. The core domain exhibits structural similarity with the NifU-like domain, suggesting that XpsEN may be involved in the regulation of XpsE ATPase activity. Surprisingly, although a similar core domain structure was observed in crystal form I4122, the N-terminal 36 residues of the helical region undergo a large structural rearrangement. Deletion analysis indicates that these residues are required for exoprotein secretion by mediating the XpsE/XpsL interaction. Site-directed mutagenesis study further suggests the more compact conformation observed in the P43212 crystal likely represents the XpsL binding-competent state. Based on these findings, we speculate that XpsE might function in T2SS by cycling between two conformational states. As a closely related protein to XpsE, secretion ATPase PilB may function similarly in the type IV pilus assembly.

For most pathogenic Gram-negative bacteria, the secretion of extracellular hydrolytic enzymes and toxins involves two translocation steps (for reviews see Refs. [1][2][3][4][5]. The secreted proteins that possess typical N-terminal signal sequences are first transported across the cytoplasmic membrane to the periplasmic space by the highly conserved Sec pathway. Following the removal of signal peptides and folding into their native conformation, the type II secretion system (T2SS) 3 or general secretion pathway is then engaged to direct the translocation of matured exoproteins across the bacterial outer membrane. Depending on species, around 12-15 distinct proteins are usually required for the functioning of the T2SS.
Although the T2SS serves as an outer membrane translocase, essential components of this system are found in four distinct subcellular locations. Together, these proteins form a trans-envelope complex that spans the entire length of periplasm and penetrates both the outer and the cytoplasmic membranes. Embedded in the outer membrane is the gated secretion pore composed of GspD (also known as secretin) (6,7). A pilus-like structure comprising one major (GspG) and four minor (GspH, -I, -J, -K) pseudopilins (8,9) extends between the cytoplasmic and the outer membrane (10). This presumably retractable assembly of pseudopilins might facilitate protein transportation by acting as a piston (11). Three integral cytoplasmic membrane proteins, GspL, GspM, and GspC, form a hierarchically structured ternary complex GspL-M-C (12). Being able to mediate multiple protein/protein interactions, this inner membrane ternary complex may function as an energy-or signaltransmitting module to couple the various molecular events required for exoprotein secretion. Direct interactions have been demonstrated between GspC and GspD (13) as well as between GspC and the major pseudopilin GspG (14). In addition, GspL is known to interact with the cytosolic secretion ATPase GspE (15,16). It was suggested that the GspE/ GspL interaction is indispensable for secretion mediated by the T2SS.
GspE and its orthologues belong to the GspE-VirB11 secretion NTPases superfamily that is essential for the T2SS (GspE family) and the type IV secretion system (VirB11 family) (17); the GspE family also includes putative NTPases required for type IV pilus biogenesis (18). Members of the GspE-VirB11 superfamily are characterized by the presence of nucleotide-binding Walker motifs in the C-terminal portion of the polypeptide. The integrity of these motifs is crucial for GspE activity because the T2SS function is greatly perturbed by mutations in these sequences (15,16,18,19). As nucleotide binding and hydrolysis are commonly exploited to power or control the action of translocases (20), a potential function of GspE is to serve as an energy-generating or regulatory component for the T2SS. Although isolated GspE does not associate with the membrane, it is recruited to the membrane-anchored secretion complex by interacting with GspL. This interaction couples the cytosolic ATP binding/hydrolysis activity of GspE to key secretory events, such as the gating of the secretion pore and the assembly/retraction of the pilus-like structure.
The crystal structures of an N-terminal truncated form of EpsE (GspE of the Vibrio cholerae T2SS) were determined in the unliganded and nucleotide-bound states (21). This study not only revealed the general architecture of the ATPase domain for members of the GspE subfamily,  but crystallographic packing analysis also suggested that EpsE may assemble into a hexamer. Based on the structures of the hexameric secretion NTPase HP0525 of the Helicobacter pylori type IV secretion system (22), Robien et al. (21) also constructed a ring model of the EpsE hexamer. The proposed hexameric arrangement of EpsE has been supported by analytical gel filtration analysis (23). Despite these observations, the action of GspE during exoprotein secretion has remained obscure, largely because of the lack of structural information on its N-terminal domain (Fig. 1A). Various biochemical analyses have shown that the GspE N-terminal domain is required for its interaction with GspL (15,16). Removing the N-terminal domain eliminates the link between the GspE ATPase domain and the rest of the secretion complex and consequently abolishes exoprotein secretion. 4 To understand the structure and function of this crucial bridging domain, we have determined the crystal structures of the 21-kDa XpsE N-terminal domain (XpsE N ; residues 1-152) of the Xanthomonas campestris pv. campestris T2SS in two crystal forms, P4 3 2 1 2 and I4 1 22. The structure of XpsE N observed in the P4 3 2 1 2 crystal is composed of a 62-residue N-terminal helical region and a 90-residue ␣/␤ sandwich core domain. This core domain is structurally similar to the N-terminal domain of EpsE (EpsE N ; residues 1-96) published during the preparation of this manuscript (24), suggesting that the N-terminal domains of all GspE family members may share a common fold. As the XpsE N core domain exhibits structural similarity with the ATPase-stimulating NifU-like domain, XpsE N might also be involved in the regulation of XpsE ATPase activity. Most interestingly, comparison of the two independently determined XpsE N structures revealed that the N-terminal 36 residues of XpsE N can undergo a large structural rearrangement. Data obtained from deletion analysis indicates that these residues are required for exoprotein secretion by allowing the XpsE/XpsL interaction. We also provide evidence supporting that the more compact conformation of the XpsE N N-terminal helical region observed in the P4 3 2 1 2 crystal is likely to be functionally significant.

MATERIALS AND METHODS
Bacterial Strains, Plasmids, and Antisera-Two strains of X. campestris were used in this study. The secretion-positive parental strain XC1701 is a rifampicin-resistant mutant spontaneously derived from a wild-type isolate XC17 (25). XC1723 is a secretion-negative mutant strain with chromosomal xpsE deletion. Overexpression of T7 promoter-driven genes (cloned in pET16b) and tac promoter-driven genes (cloned in pMAL-c2X) were achieved in Escherichia coli BL21(DE3) and DH5␣, respectively. The plasmid pCY3 encoding the XpsE N (residues 1-152 of XpsE) fused to an N-terminal His 6 tag was obtained by cloning a PCR fragment in pET16b. The plasmid pMT37 encoding the fulllength XpsE fused to a C-terminal Strep tag (SNWSHPQFEK) that specifically binds to Strep-Tactin-Sepharose (IBA GmbH) was obtained by cloning a 1.7-kb PCR fragment into pMT31, a pET29a-based vector, containing a linker encoding the Strep tag followed with a stop codon. The plasmid pCY4 encoding the MBP-XpsL N fusion protein was obtained by cloning a PCR fragment encoding residues 1-215 of XpsL (XpsL N ) in pMAL-c2X. Both pBBR1MCS-5 (26) and pCPP30 5 are broad host-range vectors. The pCPP30-based plasmid pMT38 that encodes the full-length XpsE fused to the C-terminal Strep tag is positive in complementing XC1723. Antiserum against maltose-binding protein (MBP) was purchased from New England Biolabs.
Construction of xpsE Mutants-The deleted xpsE(⌬1-36) gene in pET29aE was obtained by PCR amplification of the full-length xpsE gene and subcloning in pET29a at the NdeI and XhoI sites. The plasmid pIC4 was constructed by ligating a 1.6-kb XbaI-XhoI fragment from pET29aE to pCPP30 to allow the expression of XpsE(⌬1-36) in X. campestris. The plasmid pCY6 encoding the variant XpsE(L40D) was obtained by performing site-directed mutagenesis on the full-length xpsE gene cloned in pBBR1MCS-5 (pCH14) using the QuikChange mutagenesis protocol (Stratagene). A 0.5-kb NdeI-BamHI PCR fragment encoding the XpsE N (residues 1-152 of XpsE) was subcloned to pET16b (pCY3) for its overexpression in E. coli, and to pBBR1MCS-5 (pJL6) for introducing into X. campestris. The plasmid pCY5 encoding XpsE N (⌬1-36) was constructed by ligating a PCR fragment encoding residues 37-152 of XpsE to pBBR1MCS-5.
Protein Purification-E. coli containing plasmid pCY4 or pMT37 was grown in LB medium containing 200 g/ml ampicillin or 50 g/ml kanamycin at 37°C to an A 600 of 0.8 -1.0. Isopropyl-␤-D-thiogalactopyranoside was added to a final concentration of 0.4 mM for induction, and the culture was grown at 25°C for 6 h. The cells harvested by centrifugation at 7000 rpm for 15 min at 4°C were suspended in 10 ml of Buffer A (20 mM Tris-HCl, pH 8.0, 200 mM NaCl, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride). Cell disruption was accomplished by passing twice through a French press at 12,000 p.s.i. The fusion protein MBP-XpsL N immobilized on amylose resin through the MBP moiety was washed thoroughly with Buffer A before use. XpsE-Strep, XpsE(⌬1-36), and XpsE(L40D) were purified with Strep-Tactin-Sepharose. To minimize the possible effects from contaminating free ATP in the protein preparation, we treated the affinity-purified XpsE-Strep with hexokinase plus glucose (27) before loading it on a Superdex HR200 gel filtration column that was pre-equilibrated with Buffer B (20 mM Tris-HCl, pH 8.0, containing 200 mM NaCl). Resin-bound MBP-XpsL N and XpsE variants were quantified by using the BCA protein assay reagent kit (Pierce).
Preparation and Crystallization of Selenomethionyl XpsE N -Expression of selenomethionyl XpsE N was achieved at 20°C after inducing for 16 h with 1.0 mM isopropyl-␤-D-thiogalactopyranoside in strain B834(DE3) in M9 medium supplemented with 40 g/ml seleno-L-methionine (28). Purification of selenium-labeled XpsE N was performed using the protocols established for the native protein except that 1.0 mM dithiothreitol was used in all buffers to replace ␤-mercaptoethanol as the reducing agent (29). Selenomethionyl XpsE N was crystallized in two different tetragonal crystal forms using the hanging drop vapor diffusion method. The conditions for growing the P4 3 2 1 2 crystals have been described previously (29). The I4 1 22 crystals of XpsE N were obtained using reservoir solution containing 0.15 M ammonium sulfate, 0.015 M MgCl 2 , 0.05 M MES, pH 5.6, and 20% polyethylene glycol 8000. The unit cell parameters of the selenomethionyl XpsE N crystals are summarized in TABLE ONE.
Data Collection and Structure Determination-The P4 3 2 1 2 and I4 1 22 crystals of selenomethionyl XpsE N were stabilized in mother liquor containing 17% polyethylene glycol 400 plus 20% ethylene glycol and 22% polyethylene glycol 8000 plus 30% ethylene glycol, respectively, before flash-freezing in liquid N 2 . For each crystal form, diffraction data sets were collected from a single selenium-labeled crystal at three different wavelengths at SPring8, Japan (beamline SP12B2, ADSC Quantum 4 charge-coupled device detector, 100 K) and processed using HKL2000 (30). Data from each wavelength were indexed according to the same crystal orientation matrix but integrated and scaled independently. Scaled data sets from each of the three wavelengths were scaled together from 30.0 Å to the corresponding high resolution limit using Scaleit (31). The structures of XpsE N were determined by the multiwavelength anomalous diffraction method. Specifically, the selenium substructure was determined with the program Shake-and-Bake (32). The refinement of selenium positions, calculation of phases, density modification, and building of initial models were all performed by using SHARP/ autoSHARP (33). The programs O (34) and REFMAC5 (31) were used for rounds of manual model rebuilding and refinement. Data collection and refinement statistics are summarized in TABLE ONE. The structural superposition was performed using the least-square fitting algorithm implemented in the Swiss-PdbViewer (43). Because of large structural differences, residues 1-36 were excluded from superposition. E, secondary structure elements of XpsE N are indicated above (P4 3 2 1 2 structure) and below (I4 1 22 structure) the sequence. The disordered regions are represented as dashed lines.
Pull-Down Assay-XpsE or its variants (at final protein concentration of 0.6 M) were incubated with 1 mM AMPPNP and 10 mM MgSO 4 on ice for 30 min followed by incubation with resin-bound MBP-XpsL N at 4°C for 1.5 h. The supernatant collected from centrifugation at 6000 rpm was saved as unbound fraction. The amylose resin with bound protein was subsequently washed three times with Buffer B before being suspended in SDS-polyacrylamide gel sample buffer.
Assays for ␣-Amylase Secretion-␣-Amylase secretion activity was assayed on starch plate as described previously (25). In brief, freshly grown colonies were transferred with toothpicks in triplicate onto a minimal medium supplemented with starch (starch at 2%, w/v) plate and incubated at 28°C overnight. XC1701 and XC1723 were included as secretion-positive and -negative controls, respectively. To get a semiquantitative estimation of the secretion activity, we assayed the ␣-amylase activity in the extracellular fraction by following the procedure of Lee et al. (35).

RESULTS
Structures of XpsE N -XpsE N , the N-terminal domain of XpsE (residues 1-152) (Fig. 1A), was crystallized in two tetragonal crystal forms, P4 3 2 1 2 and I4 1 22, and the corresponding crystal structures were determined by using multiwavelength anomalous diffraction phasing with selenomethionine-labeled protein crystals and refined to 2.0-and 2.2-Å resolution, respectively (TABLE ONE). Ribbon diagrams for the two independently determined XpsE N structures are shown in Fig. 1, B and C. The final P4 3 2 1 2 model contains 1,136 protein atoms, a cacodylate ion, and 152 solvent molecules. Presumably because of disorder in the crystal, the electron density for two solvent-exposed loop segments (residues 31-33 and 80 -81) and the C-terminal tripeptide (residues 150 -152) was not observed, and these residues were not included in the model. In addition, residues 34 and 35 were modeled as alanines because of poor electron density. Although the structure of the I4 1 22 crystal form was determined at a slightly lower resolution, all residues except the C-terminal tripeptide have well defined electron density, and 213 solvent molecules were included in the final model. The stereochemis-try of both models is reasonable (TABLE ONE); no residue of either model falls within the disallowed region of the Ramachandran plot.
The structure of XpsE N observed in the crystal form P4 3 2 1 2 is composed of an ␣/␤ core domain (residues 63-149) capped by a helical region (residues 1-62) at one end (Fig. 1B). The N-terminal helical region folds into four ␣-helices (labeled as ␣1-␣4 in Fig. 1B) to enclose a hydrophobic core filled predominantly with Leu residues. Although the arrangement of these helices is reminiscent of an antiparallel four-helix bundle, individual helical axes tilt considerably away from the superhelical axis. As a result, a structural similarity search performed by using the DALI server (36) only revealed marginal similarity (DALI Z scores around 2) with other four-helix bundle proteins. Unexpectedly, the highest degree of structural similarity was identified between this region and a number of six-helix bundle proteins, including the caspase recruitment domain of Apaf-1 (Z ϭ 4.2) and the ICEBERG protein (Z ϭ 3.6) (37,38). Structural superposition revealed that four of the caspase recruitment domain helices (␣2-␣5) can be aligned with the XpsE N helical region with a root mean square difference in C␣ atoms of 2.2 Å over 47 structurally equivalent residue pairs ( Fig. 2A). Although in caspase recruitment domain these four helices form a concave surface to accommodate two additional helices (␣1 and ␣6), the equivalent surface of the XpsE N helical region interacts with its own ␣/␤ core domain. The six-helix bundle proteins are known as structural scaffolds for molecular recognition (39,40). This structural similarity may thus have a functional significance because it can be inferred that the XpsE N helical region may mediate XpsE/XpsL interaction. An exposed hydrophobic patch formed between helices ␣1 and ␣2 in XpsE N may serve this function.
The core domain of XpsE N has an ␣/␤ sandwich fold, with a highly twisted three-stranded mixed ␤-sheet surrounded by three ␣-helices (Fig. 1B). This domain resembles the NifU-like domain of mouse Hirip5 protein (Protein Data Bank code 1veh) with a DALI Z score of 2.6 ( Fig.  2B). Superposition of the two structures yields a root-mean-square difference of 2.7 Å in 57 structurally equivalent C␣ atom pairs. Although the biological functions of NifU-like domains remain to be better characterized, it is known that IscU, a NifU-like Fe/S escort protein, binds to  (24). Although XpsE N and EpsE N differ significantly in length and share only limited sequence homology, the overall structure of the shorter EpsE N is similar to that of the XpsE N core domain. This observation strongly suggests that the N-terminal domains of all GspE family members may share a common fold.
Although the structures of the XpsE N core domain are essentially identical, foldings of the N-terminal helical region are very different in the two crystal forms (Fig. 1, B-D). Instead of packing against helices ␣3 and ␣4 to form a helical bundle as observed in crystal form P4 3 2 1 2, the first 36 residues of XpsE N , which form helices ␣1 and ␣2, extend away from the XpsE N structural core to contact three neighboring crystallographic symmetry-related molecules in crystal form I4 1 22. This observation reveals that the XpsE N N-terminal helical domain can undergo a large structural rearrangement and can exist in at least two distinct conformational states that are in equilibrium. Because the structure observed in crystal form P4 3 2 1 2 is more compact, this conformation is referred as the "closed conformation" of XpsE N as opposed to the "open conformation" observed in crystal form I4 1 22. The conversion between the two conformational states is accompanied by a loop-to-helix secondary structure-switching event. Helix ␣2 is composed of only six residues (Asp 22 -Arg 27 ) in the closed conformation but spans 11 residues (Leu 24 -Gln 34 ) in the open form (Fig. 1E).
The First 36 Residues of XpsE Are Required for ␣-Amylase Secretion-Comparison of the two XpsE N structures clearly shows that the first 36 residues of XpsE N are able to undergo a large structural rearrangement (Fig. 1, B-D). To determine whether this particular region is involved in XpsE function, a truncated XpsE mutant devoid of the N-terminal 36 residues XpsE(⌬1-36) was constructed, and ␣-amylase secretion assay was performed to examine whether this mutant can support exoprotein secretion in an xpsE knock-out strain (XC1723). As shown in Fig. 3A, in sharp contrast to wild-type XpsE-containing cells that secreted ␣-amylase at a high level, only background levels of ␣-amylase secretion were detected for the cells harboring XpsE(⌬1-36). Although this observation strongly suggests that the deleted region is required for T2SS function, it is possible that folding of the remaining sequence of XpsE N (residues 37-152) was seriously perturbed by the truncation; therefore, the phenotype exhib-ited by XpsE(⌬1-36) simply resulted from structural instability of this key domain of XpsE. To rule out the latter possibility, we took advantage of the dominant negative activity of XpsE N on exoprotein secretion and tested in vivo whether this activity can be retained in a truncated form of XpsE N , XpsE N (⌬1-36), in which 36 residues were deleted from its N terminus. If the correct folding of the XpsE N core domain relies on an intact N-terminal helical region, then XpsE N (⌬1-36) should be structurally unstable and would no longer inhibit protein secretion. As XpsE N (⌬1-36) remains dominant negative (Fig. 3A, bottom), it can be concluded that the first 36 residues of XpsE are most likely involved in the normal functioning of XpsE. This result also suggests that the core domain of XpsE N along with helices ␣3 and ␣4 is sufficient for its dominant negative activity.
Although previous analyses of XpsE and its homologues have located their N-terminal domains as the region responsible for mediating the  GspE/GspL interaction (15,16), amino acid residues participating in this essential interaction have not been identified. Knowing the importance of the first 36 residues in XpsE function, we used a pull-down assay to address whether this region is required for the interaction between XpsE and the cytoplasmic domain of XpsL (XpsL N ). In contrast to the binding activity exhibited by the full-length XpsE toward amylose resin-immobilized MBPtagged XpsL N , no significant amount of bound XpsE(⌬1-36) was detected (Fig. 3B), indicating that XpsE(⌬1-36) can no longer associate with XpsL N . Because XpsL binding brings XpsE to the secretion complex, the in vitro observation shown in Fig. 3B is consistent with the loss-of-function phenotype of XpsE(⌬1-36) and strongly suggests that the XpsL N -contacting residues are located within the first 36 residues of XpsE.
L40D Mutation Decreases the Level of ␣-Amylase Secretion and Weakens the Interaction between XpsE and XpsL-Involvement of the XpsE N-terminal 36 residues in XpsL binding combined with the large structural rearrangement observed in this region led to an important question. Could the observed N-terminal structural change be functionally relevant? More specifically, would the XpsE/XpsL interaction be affected by the conformational state of this region? If so, which of the two observed structural states of XpsE N might interact more efficiently with XpsL? To provide clues to these questions, the Leu 40 of XpsE was mutated into Asp to produce XpsE(L40D), and the effects of this mutation on exoprotein secretion and XpsE/XpsL interaction were examined. As the isobutyl side chain of Leu 40 interacts with residues Ile 10 , Val 11 , and Leu 14 of helix ␣1 through van der Waals contacts in the closed form (Fig. 4A), it is reasonable to expect that the packing between helices ␣1 and ␣3 and consequently the closed conformation of the XpsE N N-terminal helical region would be perturbed by replacing the large and aliphatic Leu at this position with a small and negatively charged Asp. Results from the ␣-amylase secretion assay demonstrated that the L40D mutation decreased the secretion level by ϳ50% compared with wild-type XpsE-containing XC1723 cells (Fig. 4B), whereas similar amounts of wild-type and mutant XpsE proteins were detected in the corresponding cell lysates (data not shown). Given that the Leu 40 side chain-mediated packing contacts are almost exclusively within the N-terminal helical region, the structure of the XpsE N core domain should not be significantly affected by the L40D mutation. Therefore this dramatic reduction in ␣-amylase secretion most likely resulted from mutation-induced structural perturbations on the closed conformation of the XpsE N N-terminal helical region. Although this experiment did not address the functional relevance of the open conformation, it appears that an intact closed conformation is likely involved in XpsE function.
We next used a pull-down assay to assess whether the XpsE/XpsL interaction was affected by the L40D mutation (Fig. 4C). Unlike the binding activity displayed by the wild-type XpsE, the XpsE(L40D) mutant protein did not interact with XpsL N and was found in the unbound fraction. Because this interaction is required for T2SS function, this result not only provides a clear explanation regarding why the L40D mutation greatly reduces ␣-amylase secretion in vivo (Fig. 4B) but also implies that an intact closed conformation of the XpsE N-terminal helical region may be involved in exoprotein secretion by mediating the XpsE/XpsL interaction.

DISCUSSION
The Interaction between XpsE and XpsL-The two similar yet distinct crystal structures of the XpsE N-terminal domain reported in this study reveal the possibility of functionally relevant structural rearrangement of its N-terminal residues. The functional significance of this region is illustrated by the abolition of exoprotein secretion upon its removal. A pull-down assay further demonstrated the requirement of these N-terminal residues for mediating the XpsE/XpsL interaction, which offered a straightforward explanation as to why the truncation mutant XpsE(⌬1-36) does not support the T2SS functioning. Taken together, these results strongly suggest that the XpsL-contacting region is mainly located within the first 36 residues of XpsE. Moreover, as the control experiment shows that the folding of the XpsE N ␣/␤ core domain is not affected by the truncation of these N-terminal residues, the finding that XpsE(⌬1-36) no longer interacts with XpsL N also indicates that the core domain is not sufficient for mediating the XpsE/XpsL interaction. To pinpoint the residues directly involved in XpsL binding would require additional mutagenic or XpsE N /XpsL N cocrystallization studies because the corresponding amino acid sequence of this region is less conserved among members of the GspE family of the secretion NTPases. This sequence divergence is in fact consistent with the notion that species-specific interactions between the GspE and GspL proteins are mainly determined by the GspE N-terminal domain (15).
The N-terminal domains of the GspE family members not only diverge somewhat in sequence but also vary considerably in length (Fig.  5). Although some GspEs can be aligned with XpsE over the entire length based on primary sequences and secondary structure predictions, others apparently have much shorter N-terminal sequences. The missing residues can be mapped to those involved in the formation of the XpsE N helical region, indicating the absence of a corresponding helical region in the shorter GspEs, such as EpsE of V. cholerae and in turn suggesting their utilization of different structural determinants for mediating the interaction between GspE and GspL. Recent findings (24) based on crystallographic study of the 90-residue EpsE N indeed show that it interacts with the EpsL cytosolic domain via helix ␣2 and strand ␤C, which are equivalent to helix ␣6 and strand ␤3 of the XpsE N core domain, respectively.
Functional Significance of the Closed Conformation of XpsE N -Having established the functional requirement of the XpsE N helical region, the existence of two distinctive XpsE N conformational states prompted us to ask whether the interaction between XpsE and XpsL can be affected by the structural state of this region. Although the functional relevance of the open conformation was not directly addressed in this study, the in vivo and in vitro properties of XpsE(L40D) in which the closed conformation is supposed to be considerably perturbed imply that an intact closed form is critical for the T2SS function by allowing the XpsE/XpsL interaction. Although a secretion-defective phenotype can also be expected if the L40D mutation seriously disrupts the overall structural integrity of XpsE, the observation that XpsE(L40D) remained partially active toward exoprotein secretion suggests that the mutation-induced perturbations are more likely localized in the vicinity of Leu 40 . A global misfolding is expected to completely abolish the function of XpsE. Consistent with this view, a Western blot analysis performed on the cell lysate revealed no enhanced degradation of XpsE(L40D) in vivo (data not shown).
The above discussion hints that the interaction between XpsE and XpsL is governed by the conformation of the XpsE N helical region. Given our observation that this region can adopt an alternative conformation, it is reasonable to speculate that XpsE might cycle between the two conformational states to support exoprotein secretion. By adopting the closed conformation, XpsE interacts efficiently with XpsL and is recruited to the membrane-anchored secretion machinery. Upon switching back to the open conformation, XpsE dissociates from XpsL and returns to its soluble cytoplasmic form. This hypothesis is consistent with the findings that XpsE and its homologues are present in two different forms during cellular fractionation, a soluble cytoplasmic form and a GspL-bound membrane-associated form (15). Additional support for this hypothesis comes from the recently established ATPase activity of V. cholerae EpsE, which implies that GspE can indeed exist in at least two states, a hexameric ATP hydrolysis-competent form and a monomeric hydrolysis-incompetent form (23). However, it remains possible that the XpsE N open conformation is merely a crystallization artifact.
The NifU-like Core Domain of XpsE N Might Be Involved in the Regulation of XpsE ATPase Activity-Although it has long been speculated that members of the GspE subfamily function as ATPases, unambigu- Amino acid sequences in each subclass were first aligned using ClustalW (44); optimal alignment between the two subclasses was achieved via structure-based alignment technique using the three-dimensional structures of XpsE N (this study) and EpsE N (24). The XpsE N helical region, XpsE N core domain, and the EpsE N are indicated by the yellow, red, and purple lines below the sequences. The more conserved residues are color-shaded according to the following scheme: aliphatic (gray), polar (green), negatively charged (cyan), positively charged (red), aromatic (light brown), and proline (light purple). Residue Leu 40 of XpsE N is marked with an asterisk. The residues for XpsE and EpsE are numbered. P. aeruginosa, Pseudomonas aeruginosa; T. maritima, Thermotoga maritima; N. meningitidis, Neisseria meningitidis. ous nucleotide hydrolysis activity was only detected very recently in EpsE (23). However, as the activity exhibited by EpsE is merely 2-3-fold higher than that of the non-functional EpsE K270A Walker mutant, Camberg and Sandkvist (23) suggested that the condition for optimal EpsE ATPase activity remains to be identified. Such a proposal agrees with the concept that exoprotein secretion via the T2SS is strictly gated; therefore, the ATPase activity of GspE is likely regulated through a highly coordinated network of protein/protein interactions and is activated only when extracellular proteins are presented to the secretion complex. Because the N-terminal domain of GspE couples its ATPase domain to the membrane-bound secretion machinery, the signal for ATP hydrolysis may be transmitted via this bridging domain. With the structure of its core domain resembling the ATPase-stimulating NifUlike domain, XpsE N appears particularly suitable for the proposed signal-transmitting role. It is conceivable that XpsL binding may modulate the interaction between XpsE N and the XpsE ATPase domain to activate its ATP hydrolysis activity. In agreement with this assumption, biochemical studies indicate that OutE (the GspE of Erwinia chrysanthemi) binds to and induces conformational change on OutL and vice versa (16). To gain further insights for the proposed allosteric interactions between these domains, we have initiated structural studies on the full-length XpsE as well as the XpsE-XpsL complex.
Two Subclasses of GspE, One of Which Includes NTPases Involved in Type IV Pilus Biogenesis-It has been pointed out earlier in the discussion that XpsE N and V. cholerae EpsE N interact with their respective GspLs by making use of different structural determinants. We also noted that the N-terminal sequence of XpsE N , shown to be required for the interaction between XpsE and XpsL, is missing not only in EpsE but also in other orthologues, such as ExeE (Aeromonas hydrophila), PulE (Klebsiella oxytoca), and OutE (E. chrysanthemi) (Fig. 5). Conversely, in addition to the closely related Xylella fastidiosa XpsE-XF, at least four other orthologues (including those involved in type IV pilus assembly (PilB)) are similar to XpsE in having an extended N terminus. Thus, we suggest grouping members of the GspE protein family into two subclasses, a long form (the XpsE subclass) and a short form (the EpsE subclass), based on the lengths of their corresponding N-terminal ends. According to our results and the structural findings by Abendroth et al. (24), members of the XpsE subclass interact with their respective GspLs by using the N-terminal helical region, whereas this interaction is mediated by the NifU-like domain for members of the EpsE subclass. Furthermore, because PilB, the NTPase involved in type IV pilus assembly, is categorized as a member of the XpsE subclass (Fig. 5), we predict that PilB may utilize a structural determinant similar to that in XpsE N to interact with an XpsL homologue. In agreement with such a prediction, an equivalent interaction has been demonstrated between the BfpD (GspE homologue) and BfpC (a likely GspL homologue) in the bundleforming pilus machinery of enteropathogenic E. coli (42).