Structural basis for substrate selection by the translocation and assembly module of the beta-barrel assembly machinery

Abstract

The assembly of proteins into bacterial outer membranes is a key cellular process that we are only beginning to understand, mediated by the β-barrel assembly machinery (BAM). Two crucial elements of that machinery are the core BAM complex and the translocation and assembly module (TAM), with each containing a member of the Omp85 superfamily of proteins: BamA in the BAM complex, TamA in the TAM. Here, we used the substrate protein FimD as a model to assess the selectivity of substrate interactions for the TAM relative to those of the BAM complex. A peptide scan revealed that TamA and BamA bind the β-strands of FimD, and do so selectively. Chemical cross-linking and molecular dynamics are consistent with this interaction taking place between the first and last strand of the TamA barrel domain, providing the first experimental evidence of a lateral gate in TamA: a structural element implicated in membrane protein assembly. We suggest that the lateral gates in TamA and BamA provide different environments for substrates to engage, with the differences observed here beginning to address how the TAM can be more effective than the BAM complex in the folding of some substrate proteins.

Graphical abstract

INTRODUCTION

A remarkable characteristic of bacteria is their ability to adapt to new environments. In Gram-negative bacteria, outer membrane proteins (OMPs) are often responsible for sensing environmental changes and facilitating the downstream modulation of gene expression that this adaptation requires (Lin et al., 2002, Vogel & Papenfort, 2006). In order to achieve their functional form in the outer membrane, nascent OMPs must first traverse the ~210 Å distance (Matias et al., 2003) across the periplasm in a largely unfolded state to ensure passage through the peptidoglycan layer (De Geyter et al., 2016, Noinaj et al., 2017). Subsequent folding and assembly of the OMPs into their functional form relies heavily on free energy changes through chaperone binding and the catalytic properties of the β-barrel assembly machinery (BAM). Two elements of that machinery are the core BAM complex and the translocation and assembly module (TAM), both of which promote effective and efficient integration of OMPs into the outer membrane (Selkrig et al., 2014, Fleming, 2015, Noinaj et al., 2017).

In Escherichia coli the BAM complex is composed of two integral membrane proteins, BamA and BamC, and three peripheral outer membrane proteins, BamB, BamD and BamE (Selkrig et al., 2014, Noinaj et al., 2015, Noinaj et al., 2017). A lateral gate positioned between the first and last β-strand in the barrel domain of BamA has been identified in crystal structures, and current models for the mechanism of BAM complex action focus on the alternating conformations in this gate assisting OMP folding and integration (Selkrig et al., 2014, Noinaj et al., 2017). This region of BamA narrows the surrounding lipid bilayer and decreases lipid order, both of which could enhance the environment for substrate folding (Noinaj et al., 2013, Plummer et al., 2015). Molecular dynamics simulations reinforced by biochemical experiments suggest that the presence of an exit pore on the top face of the barrel domain may also assist BAM function (Gu et al., 2016, Noinaj et al., 2014). Crystal structures depict several static conformational states of the BamA POTRA domains, but recent work using nuclear magnetic resonance spectroscopy and chemical cross-linking shows that conformational flexibility is important for the function of BamA, highlighting the dynamic nature of the BAM complex (Noinaj et al., 2015, Warner et al., 2017, Noinaj et al., 2017).

In contrast to what is understood of the BAM complex, the TAM is more enigmatic. The TAM is composed of two integral membrane proteins, TamA and TamB: TamA is integrated in the outer membrane, while TamB has a signal-anchor sequence embedded in the inner membrane (Selkrig et al., 2012, Shen et al., 2014), and the interaction of these two subunits therefore depends on TamB penetrating through the peptidoglycan layer. TamA is found throughout the Proteobacteria, and TamB has an even broader distribution through Gram-negative bacterial lineages (Heinz et al., 2015). Recent studies with Borrelia burgdorferi suggest that in organisms that lack a TamA, TamB interacts instead with BamA (Iqbal et al., 2016, Stubenrauch et al., 2016b). Phylogenetic analyses suggested a model wherein the TAM evolved originally from a complex of BamA and TamB, with a subsequent gene duplication of bamA giving rise to a second Omp85 sequence in the genome, ultimately evolving to become TamA (Heinz et al., 2015). Despite their evolutionary relationship, biochemical and structural analyses underscore significant differences between TamA and BamA. The five POTRA domains of BamA organize four outer membrane lipoproteins, and can move to open or close access to the central chamber in the BamA barrel-domain (Gu et al., 2016, Noinaj et al., 2014). TamA has three POTRA domains that act as a substrate-activated lever to push against its partner protein TamB (Shen et al., 2014, Selkrig et al., 2015), a partner that while anchored to the inner membrane, spans the periplasm to reach TamA (Selkrig et al., 2012).

While it is clear from genetic and biochemical analyses that both TamA and BamA mediate OMP folding and assembly, it has been unclear whether TamA and BamA act in parallel but independent pathways, or collaborate together as a composite β-barrel assembly machinery. Recent findings showed that assembly of the usher protein FimD is mediated by both the TAM and the BAM complex, but the available data can not discriminate between parallel-independent functions or a collaborative function (Stubenrauch et al., 2016a). It has also been unclear whether there is a lateral gate in the barrel domain of TamA: there are two crystal structures for TamA (pdb 4n74 and 4c00), and in both the last β-strand is in a “tucked” conformation with no evidence of a gate (Gruss et al., 2013). Nevertheless, this region of the TamA barrel-domain does have a very thin aromatic girdle (Selkrig et al., 2015), as is seen around the lateral gate of BamA, and may therefore disorder the lipids in its immediate vicinity (Noinaj et al., 2013, Plummer et al., 2015).

Here we used the substrate protein FimD as a model to address the role of the barrel domain in substrate binding, and to assess the selectivity of substrate interactions for TamA relative to those of BamA. Chimeric constructs between TamA and BamA expressed in vivo show that the barrel domains are not functionally interchangeable. A peptide scan of the substrate FimD revealed that the barrel domains of TamA and BamA selectively bind the β-strands of FimD: while many of the strands can interact equally well with the barrel domains of either BamA or TamA, some of these strands were shown to be specifically selected by either BamA or TamA. Chemical cross-linking and molecular dynamics suggest that this interaction takes place between the first and last strand of the TamA barrel domain, providing the first experimental evidence that TamA possesses a lateral gate. The conformations of TamA that we observe here, tucked and open, are equivalent to those seen in structures and molecular dynamics simulations of BamA (Noinaj et al., 2013, Gu et al., 2016, Han et al., 2016, Bakelar et al., 2016, Noinaj et al., 2017). However, while broadly similar these conformations are not the same, raising the prospect that the lateral gate in TamA may provide a different environment for substrates to engage. The differences observed here begin to address how the TAM can be more effective than the BAM complex in the folding of some substrate proteins.

RESULTS

Structural basis for BamA and TamA substrate selection

In E. coli the BAM complex has four lipoprotein partners attached to BamA, while the TAM consists of a single membrane protein partner, TamB, attached to TamA (Fig. 1A). As expected, structural overlays of BamA and TamA reveal substantial similarities in the two Omp85 proteins (Fig. 1B). Despite the overall similarities, there are three key aspects in which TamA differs from BamA. First, the relative placement of the first and last β-strands with respect to each other differs in TamA where five residues of the last β-strand (Q⁵⁶⁸ to G⁵⁷²) are in a zipped position, able to form hydrogen bonds with several residues of the first β-strand (Fig. 1C). Secondly, the more perpendicular pitch with which the β-strands in the TamA barrel domain (shear number S=20(Liu, 1998, Schulz, 2002) cross the membrane means shorter strands are required and, together with shorter loops 4 and 7 between β-strands, means that the size of the barrel domain in TamA (35 kDa) is smaller than the BamA barrel domain (43 kDa). This size difference is reflected in their solvent accessible area, with BamA having a larger surface area compared to TamA (Fig. 1D). Third, analysis of the electromagnetic landscape across the lumenal surface of the two proteins revealed that the barrel domain of TamA is highly-charged relative to that of BamA. In particular, β-strands 1–8 of TamA display many charged residues to the lumenal surface of the barrel. By contrast, BamA displays only a small patch of charged residues on β-strands 1–8, isolated away from the lateral gate (Fig. 1E).

Figure 1. Structural comparison of BamA and TamA.

A. Schematic of the BAM complex and the TAM involved in outer membrane protein assembly. In E. coli, the BAM complex contains five subunits (pdb 5d0o) and the TAM contains two subunits, TamA and TamB (TamA crystal structure pdb 4c00). No crystal structure is available for TamB, but its elongate shape has been established by atomic force microscopy (Shen et al., 2014). B. Structural alignment of BamA (pdb 5d0o chain A, aqua) and TamA (pdb 4c00, purple). The different shear numbers for barrel domains of TamA (S=20) and BamA (S=22) (Noinaj et al., 2015) are distinguishable in the overlay, as the β-strands in the TamA barrel domain are closer to being perpendicular to the membrane. C. According to current definitions (Noinaj et al 2015), the crystal structure of TamA is in a tucked conformation, with β-strand 16 spanning five residues (Q⁵⁶⁸ to G⁵⁷²). D. The barrel domains of TamA (pdb 4c00) and BamA (pdb 4c4v) were rendered and the surface area of each was calculated in the PyMOL software package(Schrodinger, 2010). The surface areas are documented in Ų. E. Comparison of the TamA (pdb 4c00) and BamA (pdb 4c4v) barrel lumen. Each barrel domain is divided in half to examine the amino acid composition inside the barrel lumen. Internal facing residues that are positively charged are shown in blue and negatively charged in red.

The POTRA domains of TamA are necessary and sufficient to bind TamB

To delineate the specific functions of the barrel domain of TamA, a series of BamA:TamA chimeric proteins were constructed (Fig. 2A). The POTRA domains of BamA were fused to the barrel domain of TamA (BamA_POTRA:TamA_BARREL) or the POTRA domains of TamA were fused to the barrel domain of BamA (TamA_POTRA:BamA_BARREL). Control plasmids encoding full-length BamA or TamA were also included in this study. Initially, the constructs were expressed in an in vivo system in which BamA is depleted but where cell viability is maintained (Supplementary Fig. S1A). In this system, following an initial 4-hour depletion then sub-culture into fresh media, 3 hours of culture in the presence of glucose was established as a time-point at which BamA is undetectable by western blot while approximately two-thirds of the cell population remains viable (Dunstan et al., 2015). Western blots of cell extracts from cultures grown in parallel revealed equivalent steady-state levels of either the BamA, TamA, BamA_POTRA:TamA_BARREL or TamA_POTRA:BamA_BARREL chimeric proteins (Supplementary Fig. S1B), and confirmed that there were no indirect effects that would change the steady-state levels of the partner proteins of BamA or TamA (Supplementary Fig. S1C).

Figure 2. Characterization of the BamA-TamA chimeras in BamA shutdown strain.

A. The constructs as cartoons. B. Plasmid-borne production of BamA, BamA_POTRA:TamA_BARREL, TamA_POTRA:BamA_BARREL and TamA in the bamA shutdown strain (see Supplementary Fig. S1). Following a four-hour depletion, cells were sub-cultured in fresh LB media supplemented with glucose and after three hours of further bamA repression, total membrane extracts were prepared and analyzed by BN-PAGE. Filters were probed with anti-sera against BamB, BamC, BamD, BamE or the POTRA domains of BamA. The position of the BAM complex is shown, it runs in two conformational states of similar but not identical size, but each having all five subunits. The position of the BamAB module is shown. C. Production of BamA, BamA_POTRA:TamA_BARREL, TamA_POTRA:BamA_BARREL and TamA in a ΔtamA strain. Total membrane extracts were prepared as in B and analyzed by BN-PAGE. Filters were probed with anti-sera raised against the POTRA domains of TamA, or against TamB. The position of the TAM is shown and the population of free TamA is indicated (see Supplementary Fig. S2).

Blue native PAGE (BN-PAGE) analysis was used to monitor the interaction of lipoprotein partners: BamB, BamC, BamD or BamE, to the chimeric constructs. The BamA_POTRA:TamA_BARREL is capable of binding to BamB, but not the other lipoproteins: neither the BamA_POTRA:TamA_BARREL nor TamA_POTRA:BamA_BARREL proteins could maintain interaction with BamC, BamD or BamE (Fig. 2B). Thus, despite the fact that in crystal structures the lipoprotein partner proteins make contact primarily with the POTRA domains, in a membrane environment the BamA_POTRA:TamA_BARREL cannot form the oligomeric BAM complex. The same constructs were transformed in to a strain lacking the chromosomal copy of tamA (i.e. ΔtamA strain) and Western blots of cell extracts from cultures grown in parallel revealed equivalent steady-state levels of either the BamA, TamA, BamA_POTRA:TamA_BARREL or TamA_POTRA:BamA_BARREL chimeric proteins (Supplementary Fig. S2A), and confirmed consistent steady-state levels of the partner proteins of BamA and TamA (Supplementary Fig. S2B).

Analysed by BN-PAGE, the TamA_POTRA:BamA_BARREL was confirmed to be capable of interacting with TamB to form the TAM (Fig. 2C).

Dissection of BamA and TamA substrate selection

Previous work suggested that both the BAM complex and the TAM participate in the folding of FimD into its membrane-assembled form (Stubenrauch et al., 2016a). Those same experiments suggest that BamA and TamA are not equivalent in that TamA promotes more efficient folding of FimD and that the folding mediated by TamA is directional, starting from the C-terminal end of the barrel domain and leading towards the N-terminus. Pulse chase analysis of FimD folding can be characterized by proteolytic cleavage of a major loop once it appears on the bacterial cell surface. To test the function of the BamA_POTRA:TamA_BARREL and TamA_POTRA:BamA_BARREL constructs to complement a ΔtamA strain, plasmids encoding the native or chimeric proteins were transformed into a ΔtamA strain that also contains a plasmid for the controlled expression of ³⁵S-labelled FimD. The diagnostic fragments of natively-folded FimD (“A” and “C”) were only observed in the ΔtamA strain complemented using the TamA construct (Fig. 3). Despite the fact that the TamA_POTRA:BamA_BARREL is capable of binding to TamB to form a version of the translocation assembly module (Fig. 2C), the BamA barrel domain in this context cannot replace the function of the TamA barrel domain.

Figure 3. Effects on substrate assembly in ΔtamA complemented with BamA-TamA chimeras.

FimD assembly was monitored over time in ΔtamA cells harbouring the indicated plasmids, as assessed by pulse chase analysis. Aliquots were taken at 10 sec, 2, 4, 8, 16 and 32 min, treated with proteinase K (+/− PK). Total protein was analyzed by SDS-PAGE and storage phosphor-imaging. The time increment is indicated as a graded triangle above the autoradiogram and a cartoon of the relevant β-barrel species expressed are indicated on the right. The position of FimD and its fragments A, B and C are indicated at right. A is the N-terminal 50 kDa fragment, C is the C-terminal 40kDa fragment, while B is a central fragment of an assembly intermediate of FimD that accumulates in the absence of the TAM (Stubenrauch et al., 2016b).

To address this point further, a peptide blot assay was established to measure substrate recognition activity. FimD is an usher protein, with a complex topology (Phan et al., 2011) that includes a periplasmic “plug” domain inserted between β-strands 6 and 7, as well as additional extra-membrane domains at the N- and C-termini (Fig. 4A). Peptides were synthesized on a solid, nitrocelluose support to produce a β-strand peptide display presenting each of the 24 β-strand elements from the FimD structure (Table 1). To probe this β-strand peptide display, seven proteins were purified: OmpF (as a negative control), TamA, TamA_POTRA, TamA_BARREL, BamA, BamA_POTRA or BamA_BARREL (see Methods). Each of these proteins was incubated with the β-strand peptide array to assess binding activity, and controls to verify the stripping of proteins between assay replicated as shown in Supplementary Fig. S3. Protein binding was assessed using antibodies to OmpF, BamA or TamA as appropriate. The faint binding of OmpF to the filter sets a benchmark for non-specific interactions in the assay (Fig. 4B).

Figure 4. Dissection of BamA and TamA substrate selection.

A. Topology map representation of the 24 β-strands (Table 1) and extramembrane domains of the outer membrane usher protein FimD. The structure of FimD (pdb 3RFZ, chain B) is color-coded accordingly, with N- and C-termini indicated. B. FimD peptide blot assay containing individual peptides that represent each β-strand of the barrel domain (1 through to 24). This β-strand peptide display was probed with recombinant OmpF to guage any non-specific binding (see Methods). C. Recombinant BamA or TamA were used to probe the β-strand peptide display. D. FimD peptide blots were also probed with truncated, domain-only versions of BamA and TamA as indicated.

Table 1.

FimD peptides

	Peptide	Residue	β-strand
1	INAGLLNYNFSGNSVQNRSA	182–199	1
2	GGNSHYAYLNLQSGLNIGAA	201–219	2
3	AWRLRDNTTWSYNSSDRSSA	219–236	3
4	KNKWQHINTWLERDIIPLSA	241–256	4
5	PLRSRLTLGDGYTQGDIFSA	256–273	5
6	FDGINFRGAQLASDDNMLSA	273–290	6
7	GHTRYSITAGEYRSGNAQSA	372–389	7
8	KTRFFQSTLLHGLPAGWTSA	392–409	8
9	AGWTIYGGTQLADRYRAFSA	406–423	9
10	DRYRAFNFGIGKNMGALGSA	419–433	10
11	ALGALSVDMTQANSTLPDSA	433–450	11
12	DSQHDGQSVRFLYNKSLNSA	451–468	12
13	ESGTNIQLVGYRYSTSGYSA	469–486	13
14	AYNKRGKLQLTVTQQLGRSA	522–539	14
15	RTSTLYLSGSHQTYWGTSSA	539–556	15
16	NVDEQFQAGLNTAFEDINSA	557–574	16
17	FEDINWTLSYSLTKNAWQSA	570–587	17
18	QKGRDQMLALNVNIPFSHSA	587–604	18
19	WRHASASYSMSHDLNGRMSA	614–631	19
20	NGRMTNLAGVYGTLLEDNSA	628–645	20
21	NNLSYSVQTGYAGGGDGNSA	645–662	21
22	GNSGSTGYATLNYRGGYGSA	661–678	22
23	GGYGNANIGYSHSDDIKQSA	675–692	23
24	DIKQLYYGVSGGVLAHANSA	689–706	24

By contrast, BamA and TamA bind avidly to the β-strand peptide array (Fig. 4C). While the two Omp85 proteins show a somewhat similar overall binding profile, several distinct binding events were observed. For example, BamA binds better to the N-terminal β-strands 1 and 2 than TamA. Conversely, TamA binds β-strands 6 and 23 more avidly than BamA does (Fig. 4C).

The same assay was then used to determine the relative binding preferences for the BamA_POTRA, BamA_BARREL, TamA_POTRA and TamA_BARREL proteins (Fig. 4D). Transient interactions of POTRA domains and substrates have been suggested (Knowles et al., 2008), but little binding activity could be ascribed to the POTRA domains in this assay, particularly in the case of the BamA_POTRA protein. The assay suggested instead that the binding of the β-strand peptides is mediated by the barrel domain of each of the Omp85 proteins (Fig. 4D). However, the substrate recognition profile for TamA appears somewhat more restricted than that of the TamA_BARREL protein (Fig. 4C, 4D), suggesting an unexpected inhibitory impact of the POTRA domains on substrate binding to the barrel domain. Three pieces of evidence support the conclusion that there is some specificity in the binding being measured. First, the relative differences seen for BamA and TamA argue against a general, non-specific peptide binding activity (Fig. 4C). Second, that the same selectivity is observed in the TamA and TamA_BARREL proteins, and in the BamA and BamA_BARREL proteins (Fig. 4D). Third, using the control β-barrel protein OmpF revealed very little peptide binding (Fig. 4B).

Mechanism of substrate binding by the TamA barrel-domain

In order to further test the interactions between the FimD β-strand 6 and TamA, a cross-linking approach was taken. Peptide 6 of FimD was synthesized with an N-terminal biotin tag and C-terminal lysine: Biotin-FDGINFRGAQLASDDNMLK. The purified FimD peptide was incubated with TamA in the presence or absence of the cross-linker BS3. Fifteen minutes of incubation with the peptide was sufficient to see a small but quantitative mobility shift in the purified TamA that corresponds in size to peptide cross-linking (Fig. 5A, left panel). Probing with streptavidin to detect the biotin tag attached to the FimD peptide confirmed that cross-linking had succeeded (Fig. 5A, right panel). This cross-link could be formed with TamA or the TamA_BARREL domain while, under the same assay conditions, no cross-link was observed to the control protein OmpF (Supplementary Fig. S4, panel A). The TamA-peptide cross-linked sample was subject to trypsin digest and mass spectrometry revealing that the C-terminal lysine residue in the peptide had cross-linked to the side chain of Lys²⁸³ in TamA (Table 2, supplementary Fig. S4, panel B). The Lys²⁸³ residue sits in β-strand 2 of TamA, with the side-chain projecting into the lumen of the barrel domain (Fig. 5B). Allowing for the length of the lysine side-chains, the positioning of the ε–amine in the side-chain and the size of the BS3 cross-linker (length of ~1 nm), the cross-link would be consistent with the FimD peptide sitting between the first and last β-strand of the TamA barrel domain.

Figure 5. FimD peptide binding to TamA.

A. The FimD peptide was incubated with purified TamA and binding assessed by chemical crosslinking with BS3. Coomassie stained SDS-PAGE shows TamA purified protein cross-linking and the western blot is probed with anti-strep to confirm it cross-links to the FimD Peptide. B. Structure of TamA (pdb 4c00) mapping the position of the tryptic digest peptide identified as cross-linked to FimD peptide (silver). C–E. The cross-linked Lys²⁸³ in TamA is depicted by representing its sidechain that points into the lumen of the barrel domain. Selected states of the barrel domain of TamA (purple) from MD simulations, with the FimD peptide shown in yellow. C. TamA with free peptide at the point of closest approach between the C-terminal Lys of the peptide and Lys²⁸³ of TamA (see also Supplementary Fig. S5). D. TamA with peptide in the lateral gate at the point of closest approach (see also Supplementary Fig. S6). E. TamA with the peptide cross-linked to Lys²⁸³ via BS3. Three hydrogen bonds between the cross-linker and TamA’s first β-strand are shown in light blue (see also Supplementary Fig. S8).

Table 2.

Validated Mass Spec BS3 mediated cross-link hits (two independent runs).

Sequence	Score	Calc_M	Delta_M3
GAQLASDDNMLK(12)-VKATWK(2):0	9.68E-13	2131.098234	0.00659
		1399.6653	869.5011
GAQLASDDNMLK(12)-VKATWK(2):0	1.19E-14	2131.098234	0.00366
		1399.6653	943.5379

In order to test this hypothesis, molecular dynamics simulations of three different peptide-protein arrangements were carried out.

In the first arrangement, the FimD peptide was placed randomly below the membrane. Initially six systems were created with the disordered peptide placed 1–2 nm below the lumenal opening of the TamA barrel domain (see Methods). Each of these systems was equilibrated for at least 20 ns. Two systems in which the peptide remained close to the periplasmic lumen were run for an additional 100 ns each. In one of the extended trajectories, Lys²⁰ of the peptide initially approached the lateral gate of TamA, coming to within about 7.5 Å of Lys²⁸³ (in β-strand 2) of TamA (Fig. 5C). This distance is well within crosslinking distance for BS3. This approach is mediated by forces between Lys²⁰ in the peptide and Glu²⁶⁹ (in β-strand 1) of TamA. Later in the simulation, Lys²⁰ forms a stable interaction with Asp⁴⁴¹ (in periplasmic turn 5) on the opposite side of the barrel. In the second 100 ns simulation, Lys²⁰ only comes to within 14 Å of Lys²⁸³ (in β-strand 2), frequently interacting with Glu²⁶⁵ (in the periplasmic linker between POTRA and barrel domains) as it approaches (plots of distance vs. time are provided in Supplementary Fig. S5). Asp³ of the peptide forms a stable interaction with Lys⁵¹³ (periplasmic turn 7), but otherwise forms no stable interactions with the periplasmic region.

In the second arrangement, separation of the first and last strands in the barrel domain of TamA was induced, after which the FimD peptide was placed such that it contributes to a β-sheet with the N-terminal half of the TamA β-barrel (see Methods). Two 100-ns simulations were run from this starting scenario. In the first simulation, the backbone of Gly⁹-Ala¹³ in the FimD peptide forms stable hydrogen bonds with Thr²⁷⁰-Tyr²⁷⁴ in β-strand 1 of TamA. Lys²⁰ comes to within ~8 Å of Lys²⁸³ (in β-strand 2) after 50 ns of simulation, although this state is short-lived (Fig. 5D, Supplementary Fig. S6). Also, Asp³ of the FimD peptide forms a stable interaction with Arg⁴¹⁰ (in β-strand 9) TamA, bridging the two sides of the barrel. In the second simulation, the same backbone and Asp³-Arg⁴¹⁰ interactions are stably formed. In the final stage of the simulation, Lys²⁰ begins to interact with Glu²⁶⁹ of TamA in order to approach within 8 Å of Lys²⁸³ (in β-strand 2). For comparison, we repeated the simulations with the FimD peptide in an open lateral gate of BamA. In both 100-ns simulations, backbone hydrogen bonds were also formed between the FimD peptide and residues on the N-terminal strand of the BamA barrel domain (Phe⁴²⁸ to Tyr⁴³²). For both BamA and TamA, we observed two dominant (>50% of the time) salt bridges between the protein and peptide. In BamA, Arg⁸ of the peptide formed a salt bridge with Glu⁴³⁵ of BamA 90% of the time in both runs, while Asp³ (peptide) and Arg⁶³⁴ (BamA) also formed one for 30% of the time in one run and for 88% in the other. For TamA, Asp³ interacted with Arg⁴¹⁰ for practically the entirety of both runs (97%; 100%), while Arg⁸ interacted with Asp⁴⁷³ (65%; 84%). Arg⁴¹⁰ (in β-strand 9) and Asp⁴⁷³ (in extracellular loop 6) are both deeper in the TamA barrel than Glu⁴³⁵ (in extracellular loop 1) and Arg⁶³⁴ (in β-strand 11) are in BamA (Supplementary Fig. S7), which, when combined with the more highly charged interior in TamA, may lead to better recruitment of substrate peptide strands with multiple charged residues such as the FimD peptide.

A third arrangement was created whereby the FimD peptide was cross-linked to Lys²⁸³ (in β-strand 2) of TamA in silico. In this tethered scenario we observed up to four backbone hydrogen bonds form between the FimD peptide and β-strand 1 in the gap created by the tucked β-strand 16 over the course of the simulations (Supplementary Fig. S8). However, the majority of the FimD peptide remained in the periplasm where it adopted a random coil or α-helical conformation (see Fig. 5E).

Lateral gate opening energetics of TamA

To directly test the existence of a lateral gate in TamA, the potential of mean force (PMF) was calculated for separation of β-strands 1 and 16 and compared to those PMFs for BamA (pdb 4K3B) as well as the homologous protein FhaC (pdb 4QKY). The energy required to open TamA laterally is similar to that for BamA, around 5 to 10 kcal/mol at large (15 Å) separations (Fig. 6). This energy is significantly less (by about 20 kcal/mol) than that for lateral opening of FhaC, an Omp85 protein that is homologous to BamA but has an unrelated function as a protein secretion pore (Guerin et al., 2014, Guerin et al., 2017); it has been used previously as a control in structural analyses of BamA (Noinaj et al., 2013). The functional relevance of lateral gating in BamA has been established through mutagenesis studies and molecular dynamics simulations (Noinaj et al., 2013, Noinaj et al., 2014). That the PMF displays similar energetic barriers to lateral gate opening for TamA and BamA further support the existence of a functional lateral gate in TamA.

Figure 6. PMFs of lateral gate separation.

Potential of mean force (PMF) as a function of separation between β-strands 1 and 16 for TamA (green), BamA (blue), and FhaC (red). The energy required to separate the strands is similar for BamA and TamA (5 to 10 kcal/mol), and both are much lower than the control protein FhaC (25 to 35 kcal/mol).

Taken together with the peptide binding data from the β-strand peptide display and the suggestions from molecular dynamics simulations that this substrate peptide can approach TamA from the periplasmic side, and that it can interact with β-strand 1 of the TamA_BARREL domain, we are led to conclude that a lateral gate is present in TamA, which could interact with substrate proteins to convert them to β-strand structured elements.

DISCUSSION

The apparent homology of TamA to BamA led to the conjecture that they share functional similarities as well. One of the most striking features of BamA is its lateral gate, an opening between the first and last β-strands of its β-barrel, observed in high-resolution structures (Bakelar et al., 2016, Gu et al., 2016, Iadanza et al., 2016), simulations (Noinaj et al., 2013, Noinaj et al., 2014), and experiments (Noinaj et al., 2014). It was suspected that TamA also possesses a lateral gate, in part because its β-strand 16 was seen to be kinked at Gly⁵⁷⁴, tucking a portion of the strand into the barrel (Gruss et al., 2013). To test this hypothesis, we sought to identify where a substrate, namely β-strand 6 of the usher protein FimD, would preferentially cross-link to TamA. We found the cross-link formed with Lys²⁸³ in β-strand 2 of TamA, near the presumed lateral gate. Molecular dynamics simulations were then used to examine the plausibility of the substrate FimD peptide inserting into the lateral gate. Simulations of the FimD peptide in a forced-open TamA gate demonstrate that a cross-link between Lys²⁰ of the peptide and Lys²⁸³ (in β-strand 2) could form, with the two residues coming well within the ~1-nm BS3 cross-linker size (see Fig. 5E). Conversely, if the cross-link is already formed in silico, we observed new hydrogen bonds forming between BS3 and β-strand 1 of TamA in the gate. Finally, calculations of the free energy of gate opening show similarly low barriers to opening for TamA and BamA. Taken together, these results suggest that TamA possesses a dynamic lateral gate equivalent to that proposed for BamA.

While contacts have been observed in crystal forms of BamA between residues in the POTRA domains and inter-strand turns in the barrel domain of BamA (Noinaj et al., 2013, Gu et al., 2016, Bakelar et al., 2016), mutagenesis studies show that these contacts are not important to BamA function (reviewed by Noinaj et al., 2015). Chimeric proteins were therefore constructed to delineate functions of the POTRA domains and barrel domains in BamA and TamA. The TamA_POTRA and BamA_POTRA domains were shown to be necessary and sufficient for docking TamB and BamB, respectively. By contrast, the BamCDE module requires both the BamA_BARREL domain and the BamA_POTRA domain in order to dock to form the BAM complex. This is consistent with BamC spanning the membrane and is suggestive of protein:protein contacts between BamC and the BamA barrel domain in the plane of the membrane. Despite having partially (BAM complex) or fully (the TAM) restored the structure of the respective complexes, in no case did the chimeric constructs have sufficient activity to complement the phenotypes in loss-of-function mutants of E. coli.

In the context of the biochemical assay used in this study, the POTRA domains had little binding activity towards the transmembrane β-strands of an outer membrane protein. We do not, however, rule out a role for the POTRA domains in more transient interactions or participating in interactions with other elements of a substrate protein, such as inter-strand loops and turns. While it remains unexplained, we noted also a potential inhibitory impact of the POTRA domains on the binding of some substrate peptides to the barrel domain of TamA. In terms of stable interactions with the transmembrane β-strand peptides, the barrel domains of both BamA and TamA are capable of mediating interactions, and they do so with high selectivity for their substrate. This observation begins to explain the rapid, and directional, folding of proteins like FimD mediated by the TAM (Stubenrauch et al., 2016a). TamA demonstrated a selectivity for elements of FimD over and above any binding observed for BamA. Conversely, BamA showed preferential binding for strands such as peptide 2 i.e., towards the N-terminal region of FimD, that were not bound by TamA in these assays. It has been established in vivo that FimD folds from the C-terminus, and that this directionality depends on the activity of the TAM, underscoring that the two proteins interact differently with the same substrate (Stubenrauch et al., 2016a).

While mutations in the C-terminal β-strand of some substrate proteins can prevent their folding by the BAM complex (de Cock et al., 1997, Robert et al., 2006, Lehr et al., 2010, Gessmann et al., 2014) the C-terminal β-strand of FimD is not a prefered substrate for either BamA or TamA. The results from this substrate binding assay suggest instead other peptides would have a dominant effect on determining substrate engagement with the lateral gates in BamA and TamA. Sequence based comparisons of the peptide-binding data did not reveal any obvious sequence features (such as charge profile or hydropathy index values) that could be used to predict the features in a peptide that would drive a preferential binding to TamA or BamA.

That the BAM complex is essential for cell viability, and that repression of bamA expression impacts on the steady-state levels of essentially all β-barrel proteins, presents a conundrum concerning the value of the TAM. What is the evolutionary pressure that causes a species of bacteria like E. coli to maintain the TAM? The comparison of the lateral gate and other features of TamA with BamA suggests three possible explanations, which are not mutually exclusive.

First, the differences in structural elements in the barrel domains of BamA and TamA could have an influence on the mechanics of the lateral gate. The opening of this gate in a lipid environment is a considerable feat of thermodynamics. The different shear numbers in TamA and BamA influence the length and hydrophobic environment around the β-strand 1 and 16. In addition, differences in the charged residues in the lumenal surface close to the first strand would influence the approach of certain substrate elements, as evidenced in the molecular dynamics simulations and cross-linking studies herein. Another striking feature distinguishing TamA is the C-terminal residue of β-strand 16 is never an aromatic residue as it is in all species of BamA, but rather a leucine or isoleucine. We suggest that these various features might influence lateral gate mechanics, and thereby provide a capacity for engaging a broader suite of outer membrane protein substrates by maintaining a coding capacity for both BamA and TamA in the genome.

Second, the outer membrane is far from typical energy sources, namely ATP and the transmembrane proton motive force that are available in the inner compartments of the bacterial cell. How the BAM complex is energized for its activity in integrating its substrate proteins is unclear, but it is a reasonable assumption that conformational switching to open the lateral gate might be constrained, and perhaps induced, by the interactions of BamA with its lipoprotein partners (Gu et al., 2016, Noinaj et al., 2014, Han et al., 2016, Bakelar et al., 2016, Iadanza et al., 2016). Herein lies a major difference between the BAM complex and the TAM. TamA makes contact with a partner protein that is braced against the turgor of the inner membrane. TamB is an elongated protein of ~165 Å in length (Shen et al., 2014), integrated by an N-terminal signal anchor in the inner membrane (Selkrig et al., 2012, Selkrig et al., 2015, Shen et al., 2014). When a substrate engages the lateral gate of TamA, the POTRA domains move more than 30 Å away from the outer membrane, movements that could be resisted or transformed by TamB (Shen et al., 2014). In this regard, the TAM coupling the inner and outer membranes is reminiscent of the energetics in other intermembrane systems, such as that used by processes of active nutrient acquisition through TonB-dependent receptors, delivery of virulence factors through secretion systems and directional motility through flagellar motors (Costa et al., 2015, Noinaj et al., 2010). We suggest that these two distinct mechanisms for activating the BAM complex and the TAM might provide a selective advantage under specific environmental conditions or for handling complex substrate proteins with difficult folding pathways.

Third, lateral gene transfer is an important aspect of bacterial evolution, and it means that a given species of bacteria will often be presented with an “alien” outer membrane protein sequence. Having more than one Omp85 protein system able to work with newly acquired protein sequences is a belts-and-braces approach to outer membrane biogenesis. Usher proteins and autotransporters are two examples of virulence factors acquired by lateral gene transfer. The sporadic distribution of the genes encoding these adhesin systems seen in comparative genomic studies show that various strains and pathotypes of E. coli have acquired diverse examples of adhesins in order to rapidly adapt to diverse niches (Wurpel et al., 2013, Celik et al., 2012). We suggest that having both a BAM complex and the TAM provides an embellished protein assembly machinery, better able to work with protein sequences that did not co-evolve with the species’ own assembly machinery.

METHODS

Strains and Plasmids

The bacterial strains and plasmids used in this study are listed in Table S1. E. coli strains (Lehr et al., 2010) were routinely incubated in LB (1 % w/v tryptone, 0.5 % w/v yeast extract and 0.5 % w/v NaCl) medium, and unless otherwise indicated, incubation was performed at 37 °C at 200 strokes per minute (25 mm orbit). Additionally, E. coli MC4100 bamA depletion strains were grown in the presence of 0.2 % w/v D-glucose (for depletion) or 0.2 % w/v L-arabinose (for repletion) with appropriate antibiotics (100 μg mL⁻¹ ampicillin, 30 μg mL⁻¹ kanamycin and 50 μg mL⁻¹ chloramphenicol) unless otherwise stated. E. coli BL21 Star™ (DE3) strains were routinely incubated in the presence of ampicillin (100 μg mL⁻¹) and chloramphenicol (34 μg mL⁻¹) to maintain pTnT-based vectors and pCJS71, respectively. For growth experiments, the bamA shutdown strain harboring pTnT-BamA, pTnT-BamA_POTRA:TamA_BARREL, pTnT-TamA_POTRA:BamA_BARREL, pTnT-TamA or pTnT (empty vector control) were incubated overnight in the presence of arabinose and diluted 1:100 in LB containing glucose then incubated at 37 °C. After 4 h of incubation, cultures were normalized to OD₆₀₀ of 0.03 in LB containing glucose to leap cells into exponential growth phase.

Membrane Isolation

Cells were harvested and resuspended in sonication buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl and 5 mM EDTA) and then disrupted by sonication. The cell extracts were subjected to centrifugation (3,000 ×g, 4 °C, 5 min) to remove unbroken cells and cell debris, and supernatants were collected. The heavy membrane fraction was collected by centrifugation (20,000 ×g, 4°C, 10 min), and then resuspended with SEM buffer (250 mM sucrose, 1 mM EDTA and 10 mM MOPS-KOH pH 7.2). Membrane fractions were snap-frozen by liquid nitrogen and stored −80°C until needed.

Analysis by SDS-PAGE and BN-PAGE were performed as described previously (Shen et al., 2014).

Peptide Dot-Blots

Twenty-four peptides of 18 residues each to cover the sequence of the transmembrane segments of FimD (Table 1) were spotted onto hardened cellulose membranes according the method of Hilpert et al. (2007). Peptides were extended at the C-terminus by a serine-alanine spacer to minimize membrane charge interferences. The membranes were incubated at room temperature for 2 min in methanol, washed for 2 min with H₂O and equilibrated for 20 min with binding buffer (TBS-T). The membrane was blocked with 5 % w/v BSA in TBS-T for 60 min at room temperature. Thereafter the membrane was incubated with binding buffer containing 0.1–0.3 μM purified His-tagged protein for 45 min at room temperature. After washing four times for ten min with TBS-T, the His-tagged recombinant protein was probed with anti-His-HRP (R&D Systems, MAB050H), at 1/20000 dilution in 1.5 % w/v BSA in TBS-T for 30 min. Bound antibody was detected using ECL.

Dot blots were stripped with 2 % w/v SDS, 50 mM Tris pH 6.7, 0.5 % w/v β-mercaptoethanol at 45 °C for 30 min. Washed thoroughly with TBS-T, and then milliQ water and stored in milliQ water between uses.

Peptide cross-linking and Mass Spectrometry

A synthetic peptide corresponding to peptide 6 of FimD was synthesised by Mimotopes with an N-terminal Biotin and C-terminal lysine added to facilitate cross-linking. The sequence of the peptide is therefore: Biotin-FDGINFRGAQLASDDNMLK. The C-terminal lysine residue was added to allow cross-linking mediated by BS3 between free amine groups, i.e. lysines or the N-terminus of a protein (Paramelle et al., 2013). The biotin was used to detect cross-linked peptide via Western blot, in which a sample of TamA exposed to BS3 in the presence and absence of peptide 6 was run on SDS-PAGE, transferred to nitrocellulose and probed with Precision Protein™ StrepTactin-HRP Conjugate (Bio-Rad, 1610381). A band at the approximate molecular weight of TamA was detected in the TamA plus peptide and BS3 sample, but not in the samples without peptide nor the TamA plus peptide without BS3.

Peptide and protein were mixed at a 1:1 ration (at 5 μM each) in sodium phosphate buffer at pH 8 with 50 mM NaCl. A negative control containing TamA alone was also subjected to the same cross-linking protocol and analysed. BS3 was added at a 10:1 molar excess and the cross-linking reaction stopped at 10 min by the addition of Tris-HCl pH 7.5 to 100 μM. Protein was TCA precipitated, and resuspended in ammonium bicarbonate. Protein was reduced and alkylated prior to an in-solution overnight trypsin digestion. Digestion was ended by addition of formic acid, and peptides desalted using OMIX C18 100 μL tips (Agilent). Using a Dionex UltiMate 3000 RSLCnano system equipped with a Dionex UltiMate 3000 RS autosampler, an Acclaim PepMap RSLC analytical column (75 μm × 50 cm, nanoViper, C18, 2 μm, 100 Å; Thermo Scientific) and an Acclaim PepMap 100 trap column (100 μm × 2 cm, nanoViper, C18, 5 μm, 100 Å; Thermo Scientific), the tryptic peptides were loaded onto the trap column at 15 μL min⁻¹ in 2 % v/v acetonitrile, 0.1 % v/v trifluoroacetic acid and were separated at a flow rate of 250 nL min⁻¹ over a 60 min gradient from 7.5 % to 42.5 % solution B (solution A: 0.1 % formic acid, solution B: 80 % acetonitrile, 0.1 % formic acid) and analyzed with a Orbitrap Fusion mass spectrometer (Thermo Scientific) employing a top speed approach with a 3 s cycle time. A 120k resolution MS scan (300–2000 m/z) was performed followed by data-dependent orbitrap HCD MSMS scans (60k res, 32 % collision energy). Dynamic exclusion was employed after 1 occurrence for 15 s. pLink (Yang et al., 2012) was used to interrogate the mass spectrometer raw output files for the faithful identification of cross-linked peptides using an eValue cutoff of 0.1. All pLink results were then manually confirmed.

Expression and purification of recombinant TamA and BamA, and domains thereof

In the cases of His-tagged BamA and TamA, and His-tagged TamA_BARREL domain, the appropriate plasmid was transformed into C41(DE3) OverExpressTM (Lucigen), grown in Terrific Broth at 37 °C and, at an optical density (600 nm) of 0.8, temperature was decreased to 20 °C and IPTG was added to a final concentration of 0.2 mM. Cells were harvested 18 hours after IPTG induction by centrifugation, resuspended in 10 mL buffer per 10 g (wet weight) cells (20 mM Tris-HCl pH 8.0, 300 mM NaCl, 5 % glycerol) and lysed using an Emulsiflex (3 passes at 15,000 psi). The cell lysate was clarified by centrifugation at 20,000 ×g for 15 min and the supernatant was centrifuged for a further 60 min at 38,000 ×g in a Type 45 Ti rotor to pellet membranes. The purified membranes were resuspended in 100 mL Elugent buffer (20 mM Tris-HCl pH 8.0, 300 mM NaCl, 5 % glycerol, 5 % Elugent, 20 mM imidazole). Soluble membrane fraction was applied to a HisTrapFF (5 mL) column. The protein was eluted using a gradient, as per manufacturers instruction, with the wash and elution buffers containing (20 mM Tris-HCl pH 8.0, 300 mM NaCl, 20 mM Imidazole, 0.05 % DDM) and (20 mM Tris-HCl pH 8.0, 300 mM NaCl, 1 M Imidazole, 0.05% DDM) respectively. After elution, the BamA protein was concentrated in a centrifugal concentrator and applied directly onto a SuperdexTM-200 (16/60) (GE-Lifesciences) column which had been equilibrated with buffer (20 mM Tris-HCl pH 8.0, 250 mM NaCl, 5 % glycerol, 0.03 % DDM). Monomeric protein peak pooled, concentrated and purity assessed by SDS-PAGE.

Initial experiments suggested that expression of the BamA_BARREL protein was toxic to E. coli. In order to generate sufficient quantities of the barrel domain, a TEV-cleavage site was engineered between the POTRA and barrel domains and this modified form of BamA was expressed. Post-purification the protein was cleaved to liberate a His-tagged BamA_BARREL protein for further study. The cleaved protein was then applied directly onto a SuperdexTM-75 (16/60) (GE-Lifesciences) column which had been equilibrated with buffer (20 mM Tris-HCl pH 8.0, 250 mM NaCl, 5 % glycerol, 0.03 % DDM). A monomeric BamA_BARREL protein peak was observed and the corresponding fractions pooled, concentrated and assessed for purity by SDS-PAGE.

Plasmids encoding the recombinant TamA_POTRA and BamA_POTRA proteins were transformed into E. coli BL21 Star™ (DE3) (Novagen), grown in Terrific Broth at 30 °C and, at an optical density (600 nm) of 0.8, IPTG was added to a final concentration of 0.2 mM. Cells were harvested 6 hours after IPTG induction by centrifugation, resuspended in 10 mL buffer per gram (wet weight) of cells (20 mM Tris-HCl pH 8.0, 500 mM NaCl, 30 mM Imidazole) and lysed using an Emulsiflex (3 passes at 15,000 psi). The cell lysate was clarified by centrifugation at 20,000 ×g for 15 min and the supernatant was purified by Ni²⁺ affinity chromatography using a HisTrapFF (5 mL) column. The TamA_POTRA or BamA_POTRA protein was then eluted using a gradient, as per manufacturers instruction, with the wash and elution buffers containing (20 mM Tris-HCl pH 8.0, 500 mM NaCl, 20 mM Imidazole) and (20 mM Tris-HCl pH 8.0, 500 mM NaCl, 1 M Imidazole) respectively. After elution, the TamA_POTRA or BamA_POTRA protein was concentrated in a centrifugal concentrator and applied directly onto a SuperdexTM-75 (16/60) (GE-Lifesciences) column which had been equilibrated with buffer (20 mM Tris-HCl pH 8.0, 250 mM NaCl, 5 % glycerol). A monomeric protein peak was observed in each case and the corresponding fractions pooled, concentrated and assessed for purity by SDS-PAGE.

A plasmid encoding His-tagged OmpF was transformed into E. coli BL21 Star™ (DE3) (Novagen), grown in Terrific Broth at 37 °C and, at an optical density (600 nm) of 0.8, IPTG was added to a final concentration of 0.2 mM. Cells were harvested 4 hours after IPTG induction by centrifugation, resuspended in 10 mL buffer per gram (wet weight) of cells (50 mM Tris.Cl pH 8.0) and lysed using an Emulsiflex (3 passes at 15,000 psi). Inclusion bodies were purified and protein refolded according to (Saleem et al., 2012). Refolded His-tagged OmpF was purified by Ni²⁺ affinity chromatography using a HisTrapFF (5 mL) column. The protein was eluted using a gradient, as per manufacturers instruction, with the wash and elution buffers containing (20 mM Tris.Cl pH 8.0, 500 mM NaCl, 20 mM Imidazole) and (20 mM Tris.Cl pH 8.0, 500 mM NaCl, 1 M Imidazole) respectively. After elution, OmpF protein was concentrated in a centrifugal concentrator and applied directly onto a SuperdexTM-200 (16/60) (GE Lifesciences) column which had been equilibrated with buffer (20 mM Tris pH 8, 250 mM NaCl, 5 % glycerol). The protein migrates as a trimer on semi-native SDS-PAGE (Supplementary Fig. S9) indicative of natively folded OmpF.

Pulse chase assays

Pulse chase analyses were performed as described previously (Stubenrauch et al., 2016a). Briefly, E. coli BL21 Star™ (DE3) wildtype or ΔtamA strains were incubated to mid-log phase, then transferred to M9-S media. Following a 30-minute incubation, cells were treated for 1 hour with rifampicin (200 μg mL⁻¹, 37 °C, 400 rpm [3 mm orbit]) and induced for 5 min with IPTG (0.2 mM, 30 °C, static). Cells were then ‘pulse’-labelled for 40 seconds with EXPRE³⁵S³⁵S, [³⁵S]-Protein Labeling Mix (22 μCi mL⁻¹, 30 °C, static), containing 73% [³⁵S]-methionine and 22% [³⁵S]-cysteine (NEG072, Perkin Elmer), and then immediately subjected to centrifugation (5 min, 3000 ×g, 4 °C) and resuspended in M9+S media. Cells were then ‘chased’ for up to 32 min (30 °C, static) and aliquots were taken at appropriate timepoints. Aliquots were treated with exogenous proteinase K (50 μg mL⁻¹; PK solution, Promega) for 10 min on ice, before total proteins were TCA-precipitated, washed with acetone and resuspended in SDS loading dye. Samples were analysed by 12% SDS-PAGE and transferred onto 0.45 μm nitrocellulose membranes. Radiation was captured overnight using a storage phosphor screen (GE Health Sciences) and detected using Typhoon Trio (320 nm).

Molecular Dynamics Simulations

The TamA simulation system was built using a crystal structure of TamA from E. coli (pdb 4c00), excluding the three POTRA domains, i.e., residues 25 – 264. The protein was placed in a 1,2-dilauroyl-sn-glycero-3-phosphocoline (DLPC) bilayer, solvated with TIP3P water, and neutralized with sodium and chloride ions to a concentration of 150 mM. DLPC, a short-tailed lipid (dodecanoic acyl chains), was chosen to enhance the lability of the protein. In the case of BamA, the structure from E. coli (pdb 4n75) was used, which is already lacking the POTRA domains (Ni et al., 2014). Each system was first equilibrated for 100 ns, after which a Biotinylated-peptide derived from FimD (Biotin-FDGINFRGAQLASDDNMLK) was added in three different ways as described in the main text. For simulations in which the peptide was cross-linked to Lys²⁸³, a BS3 cross-linker was inserted.

To produce the potential of mean force (PMF) for TamA, first the adaptive biasing force (ABF) method (Comer et al., 2015) was applied to induce a separation between β-strands 1 and 16, thus generating an open state. Targeted molecular dynamics (TMD) was then applied to interpolate between closed and the open states. Replica exchange umbrella sampling (REUS) was used to carry out the production simulations and the weighted histogram analysis method (WHAM) was used to calculate the PMF (Sugita et al., 2000). The center-of-mass distance between the Cα and Hα atoms of β-strand 1 (residues 270 to 276) and β-strand 16 (residues 569 to 575) was used as a reaction coordinate for the REUS simulations (Fiorin, 2013), which was divided into 20 windows, spaced 0.5 Å apart from 5.5 Å to 15.0 Å. REUS simulations were carried out for 10.0 ns per window for all systems to ensure convergence.

NAMD (Phillips et al., 2005) was used for all molecular dynamics simulations along with the CHARMM36 all-atom force field for protein (Best et al., 2012), ions, and phospholipids (Klauda et al., 2010). Force-field parameters for both biotin and BS3 were obtained using CGenFF (Vanommeslaeghe et al., 2010). Simulations were run at constant temperature (310 K) and pressure (1 atm). A 2-fs time step was used, with short-range and long-range non-bonded interactions evaluated every time step and every other time step, respectively. Van der Waals interactions were cutoff at 12 Å with a force-based switching function used from 10–12 Å as recommended (Klauda et al., 2010). Long-range electrostatics were calculated using the particle-mesh Ewald method (Darden et al., 1993).

All protein structure representations were prepared using either UCSF Chimera (Pettersen et al., 2004) or PyMOL (Schrodinger, 2010).

Supplementary Material

Supp FigS1–9 & TableS1

Supplementary Figure S1. Characterization of chimeric Omp85 proteins after bamA repression.

A. The bamA shutdown strain was transformed with each pTnT-based vector (see Methods), and growth was measured in LB media supplemented with glucose to test bamA complementation. Membranes were then isolated 0 and 3 hours after sub-culture (i.e. the 4- and 7-hour total timepoints), analysed by SDS-PAGE and probed with B. Antibodies raised to the POTRA domains of either BamA or TamA, or C. The partner proteins BamB, BamC, BamD and BamE, or the periplasmic control, BepA. Triangles indicate that samples equivalent to 50 μg total protein and 100 μg total protein were analysed.

Supplementary Figure S2. Characterization of chimeric Omp85 proteins in ΔtamA cells.

A. The ΔtamA deletion strain was transformed with each plasmid and steady state levels of BamA, BamA_POTRA:TamA_BARREL, TamA_POTRA:BamA_BARREL or TamA was assessed by SDS-PAGE and immunoblotting by probing with BamA and TamA antibodies. B. Steady state levels of other relevant proteins were also assessed by western blotting for TamB, BamB, BamC, BamD and BamE. C. The TAM and the BAM complex were assessed after the expression of BamA, BamA_POTRA:TamA_BARREL, TamA_POTRA:BamA_BARREL or TamA. Total membranes were analysed by BN-PAGE and probed with anti-sera against TamA.

Supplementary Figure S3. Controls for stripping and re-probing of peptide blots.

The panels show probing of the nitrocellulose dot blots with ECL alone (following the stripping process), with anti-His Ab alone (following the stripping process). Both display minimal luminescence after lengthy exposures (5 minutes, compared to the standard 30 sec exposures for the protein binding shown in Figure 4) Also shown is a UV exposure of the newly synthesized nitrocellulose blot showing peptide deposition in all positions probed.

Supplementary Figure S4. Peptide cross-linking to TamA and TamA_BARREL.

A. The panels show cross-linking experiments with equivalent amounts (5 μM protein) of TamA, OmpF and TamA_BARREL incubated with BS3 in the presence of 1:1 molar ratio of peptide 6 (Biotin- FDGINFRGAQLASDDNMLK). Protein cross-linked to the labelled peptide was detected in both the TamA and TamA_BARREL experiments. B. Mass spectra of cross-linked peptide detected in TamA + peptide 6 + BS3 tryptic digest.

Supplementary Figure S5. Free peptide approach towards Lys²⁸³ of TamA.

The plot shows the distance between Lys²⁰ of the FimD peptide and Lys²⁸³ of TamA as a function of time for two independent trajectories (red or black) of the free peptide.

Supplementary Figure S6. Peptide in the TamA lateral gate approach towards Lys²⁸³.

The plot shows the distance between Lys²⁰ of the FimD peptide and Lys²⁸³ of TamA as a function of time for two independent trajectories (red or black) of the peptide placed in the TamA lateral gate.

Supplementary Figure S7. Comparisons of TamA and BamA.

Predominant salt bridges between the FimD peptide in the lateral gate of (A) TamA, or (B) BamA. The TamA simulation system was built using a crystal structure of TamA from E. coli (pdb 4c00), excluding the three POTRA domains, i.e., residues 25 – 264. The BamA simulation system was built using a crystal structure from E. coli (pdb 4n75), which lacks the POTRA domains.

Supplementary Figure S8. Peptide hydrogen bonding with TamA.

Hydrogen bonds vs. time between the peptide or BS3 cross-linker and the backbone of the β-strand 1 of TamA for two independent runs (red or black).

Supplementary Figure S9. Purified OmpF is a natively folded β-barrel.

Purified OmpF (5 μg protein) was suspended in SDS-PAGE sample buffer and incubated for 10 min at the indicated temperature, before analysis by semi-native SDS-PAGE. The migration positions of the trimer and denatured monomer are indicated and molecular weight markers are shown.

Supplementary Table S1. Strains and plasmids.

Plain language summary.

The assembly of proteins into the outer membranes is a key process in bacteria. The folding and assembly of these proteins is mediated by a system referred to as the beta-barrel assembly machinery (BAM) which includes two related proteins: BamA in the BAM complex, TamA in the TAM. We sought to distinguish these two, ostensibly similar proteins, and to understand why both have had to be conserved through evolution.