Abstract

The homotrimeric molecular chaperone Skp of Gram-negative bacteria facilitates the transport of outer membrane proteins across the periplasm. It has been unclear how its activity is modulated during its functional cycle. Here, we report an atomic-resolution characterization of the Escherichia coli Skp monomer-trimer transition. We find that the monomeric state of Skp is intrinsically disordered and that formation of the oligomerization interface initiates folding of the -helical coiled-coil arms via a unique stapling mechanism, resulting in the formation of active trimeric Skp. Native client proteins contact all three Skp subunits simultaneously, and accordingly, their binding shifts the Skp population toward the active trimer. This activation mechanism is shown to be essential for Salmonella fitness in a mouse infection model. The coupled mechanism is a unique example of how an ATP-independent chaperone can modulate its activity as a function of the presence of client proteins.

Molecular chaperones are central for the survival of the cell in all kingdoms of life (12). They are involved in many cellular processes such as helping proteins to fold, preventing protein aggregation, and reducing cellular stress (3). Some chaperones can use adenosine triphosphate (ATP) binding and hydrolysis to trigger conformational changes that, in turn, regulate their functional cycle, including their interaction with client proteins (4). ATP-independent chaperones, in turn, lack this possibility. Nonetheless, some ATP-independent chaperones were found to be regulated by major conformational changes, and the transition mechanisms for the activation of ATP-independent chaperones have been classified into three categories (5): oligomer disassembly [small heat shock protein (sHSP) (6) and trigger factor (TF) (79)], order-to-disorder transition {Hsp33 (10), HdeA [HNS (histone-like nucleoid structuring)dependent expression A] (11), and HdeB (12)}, and lack of major conformational change [spheroplast protein Y (Spy) (13, 14), seventeen kilodalton protein (Skp) (15), HSP40 (16), SecB (17), and survival factor A (SurA) (18)]. These mechanisms of activation are of major biological importance, because constitutively active chaperones can interfere with protein folding processes and proteostasis due to their high affinity and low specificity for client proteins, thus representing a potential hazard to cells (1922). An example of these detrimental effects has been reported for a constitutively active variant of the chaperone Hsp33, which lead to accumulation of large amounts of insoluble aggregates and severe growth disadvantages (20).

Representative of the first category, binding of chaperone sHSP to its client proteins is regulated via a shift from an inactive oligomeric ensemble toward an ensemble of smaller multimers, representing the active species (6). The monomeric species exposes a binding motif that is shielded within the oligomeric structure, making the large oligomeric state an inactive storage form that can be activated upon dissociation (23, 24). Similarly, it has been shown that binding of TF to client proteins is accompanied by a shift from the inactive dimeric state toward the active monomeric state (79). By contrast, the order-to-disorder activation is found for chaperones where the active form is intrinsically disordered. Thereby, to shift from the folded inactive chaperone to the unfolded active chaperone, not only the oligomeric state but also the secondary structure of the chaperone is undergoing change, triggered either by a pH drop to acidic conditions (HdeA and HdeB) or by a redox transition (Hsp33). Once stress factors are reduced, these chaperones can return to their folded/oligomeric inactive state with a release of the client (25). The third category contains chaperones for which only one conformational state is known, and therefore, these are assumed to require no major conformational changes for their activation, as well as chaperones for which activation requires only minor conformational changes. One such example is provided by the chaperone Hsp40, which has minimal structural differences between its client-bound and apo state (16). Another example is given by the chaperone SecB, for which high-resolution structures of client-bound states showed only a minor conformational change to the inactive client-free state (17). In the client-free form, helix 2 acts as a lid of the client protein binding site. Upon client binding, this helix swings outward, thereby allowing access to the client binding groove. Similarly, the chaperone SurA has been shown to have a dynamic mechanism of activation where a domain connected to the chaperone core by linkers assists client protein recognition, binding, and release (18).

The periplasmic chaperone Skp is an integral part of outer membrane protein (Omp) biogenesis, on a parallel pathway with the chaperone SurA. Skp transports Omps in their unfolded state across the periplasm toward their insertion point into the outer membrane (2628). Yersinia skp and Salmonella skp mutants show compromised virulence in rodent infection models, indicating a crucial role of Skp in vivo (29, 30). Skp is structurally characterized by a trimeric oligomeric state with a jellyfish-like architecture (31, 32). Each protomer contributes three -strands toward a nine-stranded -barrel in the trimerization interface and a long, -helical arm, made of two -helices in coiled-coil arrangement (31, 32). The combination of three arms from the individual subunits leads to formation of a cavity that can accommodate and bind unfolded Omps (15, 33).

The elongated arms of Skp are highly flexible in the apo state, and a recent molecular dynamics study has identified a pivot element to act as a hinge, allowing Skp to adapt to clients of different sizes (15, 34). Upon binding, the Skp arms undergo a rigidification and keep the bound Omps inside the cavity in the fluid globule state (15, 35). While Skp can accommodate differently sized protein clients, all functional complexes observed so far feature an Omp:Skp stoichiometry of 1:3 or 1:6, depending on the size of the client, suggesting that Skp binds clients always as a trimer (36). A recent study has emphasized that at physiological concentrations, Skp exists as an equilibrium between a trimeric and a monomeric form (37). The equilibrium was quantified by analytical ultracentrifugation (AUC), showing that the monomeric form is strongly dominant at 2 M Skp, the concentration found in Escherichia coli stationary phase (38, 39). The monomeric form of Skp has been proposed to be well folded based on indirect evidence (37); however, it has so far not been possible to directly analyze its structure, because at the high concentration required for most biophysical methods, the protein is mostly trimeric. Consequently, the structural features of the Skp monomeric state and the Skp activation mechanism remain poorly understood.

Here, we bypass this analytical challenge by introducing several weakly and non-oligomerizing mutants of Skp. We characterize their monomeric states by solution nuclear magnetic resonance (NMR) spectroscopy at the atomic level. The emerging reference data can then be used to fruitfully understand monomeric Skp(WT). The data show that monomeric Skp is intrinsically disordered and inactive and that binding of a client protein triggers Skp trimerization and activation. Last, we demonstrate that this mechanism is essential for bacterial virulence under in vivo conditions in a mouse infection model. The data thus reveal an essential mechanism regulating Skp chaperone activity by a combined disorder-to-order and oligomerization transition.

To prepare samples of monomeric Skp at concentrations sufficient for structural characterization, we set out to design mutants that would destabilize the oligomerization interface to shift the oligomerization equilibrium toward the monomeric form. The structure of trimeric Skp is stabilized by a network of three -sheets per subunit that together form the trimerization interface in the head of the molecule (Fig. 1A). We identified the conserved alanine-103 and alanine-108 as promising candidates, because they are located at the oligomerization interface with limited space for their side chains. Their replacement by a bulkier side chain such as leucine or arginine should introduce steric clashes, leading to destabilization of the trimer (Fig. 1, A and B). In addition, we designed the mutant V117P to insert a proline residue, which is a known secondary structure breaker, into the trimerization -sheet (2). The oligomerization state of each of the Skp mutants Skp(A103L), Skp(A103R), Skp(A108L), Skp(A108R), and Skp(V117P) was determined by SEC-MALS (size-exclusion chromatography coupled to multi-angle light scattering) experiments at an elution concentration of 80 M. At this concentration, the wild-type (WT) protein is mostly trimeric with a monomeric fraction lower than 4%. The mutant A103L was hardly distinguishable from WT, but the other Skp variants featured a gradually increased monomeric fraction, as evidenced by a smaller apparent mass, in the order A103R, A108L, V117P, and A108R (Fig. 1C). Thereby, mutants A108R and V117P were fully monomeric, and the others had effective molecular weights in between monomer and trimer, suggesting the presence of dynamic equilibria. We quantified the concentration dependence of these equilibria for Skp(A103L), Skp(A103R), and Skp(A108L) by solution NMR spectroscopy and SEC-MALS experiments (Fig. 1, C to E; Table 1; and fig. S1). Skp(WT) followed an equilibrium with C0.5 = 1.5 M, the protein concentration at which half of the molecules are in the monomeric form, in agreement with published data (37). Skp(A103L) showed a trimer-monomer equilibrium that was essentially identical to WT (Fig. 1E and fig. S1), whereas Skp(A103R) had C0.5 = 7 2 M and Skp(A108L) had C0.5 = 80 20 M, indicating that these mutations shifted the equilibrium by about one to two orders of magnitude toward the monomer (Fig. 1, D and E, and fig. S1). The two mutants Skp(A108R) and Skp(V117P) were found to be monomeric at concentrations of even up to 1 mM (Fig. 1E and fig. S1).

(A) Location of the mutation sites [red boxes (I), (II), and (III)], displayed on the Skp crystal structure (Protein Data Bank: 1SG2). Secondary structure elements and termini are indicated. (B) Close-up of the interface between Skp subunits, highlighting the position of the five mutations. See text for details. (C) SEC elution profiles (solid lines, left axis) and MALS apparent molecular mass (MM) (dotted lines, right axis) at elution concentrations of 80 M and a temperature of 25C. Dark gray, Skp(WT); brown, Skp(A103L); green, Skp(A103R); blue, Skp(A108L); magenta, Skp(A108R); purple, Skp(V117P). Gray horizontal lines indicate the molecular masses of monomers, dimers, and trimers of Skp. (D) Experiment as in (C) for Skp(A108L) as a function of the elution concentrations: 5, 20, and 79 M. (E) Fractional populations f of monomers in the monomer-trimer equilibrium as a function of total Skp concentration. Experimental data points from NMR and SEC-MALS are indicated by filled and open circles, respectively. These have been fitted by Eq. 4 for mutants A103R and A108L (solid lines). The corresponding fractional populations of the trimeric state, 1 f, are shown by dashed lines. Note that the concentration of Skp trimers equals one-third of the concentration of Skp molecules in the trimeric state, i.e., [Skp]trimer = 1/3 (1 f) [Skp]tot. For Skp(WT) and Skp(A103L), the data follow the WT association constant published by Sandlin et al. (37).

Error estimates have been omitted for clarity. n.d., not determined.

We then characterized the structural integrity of Skp(A103L), Skp(A103R), and Skp(A108L) in their trimeric forms by NMR spectroscopy. For each of these proteins, two-dimensional (2D) [15N,1H]-TROSY (transverse relaxation-optimized spectroscopy) fingerprint spectra show the presence of two species in slow exchange on the NMR time scale, i.e., with kinetic exchange rate constants kex 10 s1 (fig. S1). For each of the mutants, the overlay of the NMR spectra at 25C (fig. S2) shows a high degree of similarity with the WT protein for most resonances, with considerable chemical shift perturbations only for some residues. Those residues are all located in spatial vicinity of the mutation site, in full agreement with the expected local distortion effects of single point mutations (fig. S2). The signals of residues located in the arms are not affected by the mutations, suggesting that symmetry and structural integrity of the trimeric form of the protein are maintained in the mutant. Oligomeric states other than the monomer and the trimer were not detected. The mutations thus shift the oligomerization equilibrium while leaving the trimeric form largely intact.

The mutant Skp(A108L) with a C0.5 of 80 20 M at 25C allowed us to prepare the monomeric state at concentrations of 100 M and above, which is required for solution NMR spectroscopy. The NMR spectra of monomeric Skp(A108L) are completely overlapping with the monomeric, but low-abundant conformation of Skp(WT) (Fig. 2A), indicating that the conformations are essentially identical and thus validating the further analysis. Increasing the temperature from 25 to 37C shifted the equilibrium of Skp(A108L) further toward the monomer, resulting in around 95% monomer at a concentration of 1 mM and thus further increasing the NMR signal intensity (fig. S2). A primary classification of the type of conformational state of monomeric Skp was obtained from the observation of a narrow chemical shift dispersion of backbone amide NMR signals, which is characteristic for proteins with low structural propensity (fig. S3). To quantify the secondary structure elements, we established complete sequence-specific resonance assignments of the monomeric state (fig. S3) and determined backbone 13C and 13C secondary chemical shifts (Fig. 2B). These show that the three -sheets that constitute the oligomerization interface in the trimeric form are in random-coil conformation in the monomeric state. Furthermore, the four -helices forming the arms of Skp are in a fast conformational exchange between folded and unfolded conformations, as evidenced by the observation of a single set of resonances in fast exchange. Taking the fully denatured form of Skp in 8 M urea solution and the folded trimer as reference points, the residual helicity can be quantified for each residue (Fig. 2, B and C). The analysis shows that the helices 1, 3.B, and 4, which are closest to the trimerization interface, feature a residual helicity of <20%, while the helices 2.A, 2.B, and 3.A located at the tip of the arms display a helical population of 20 to 30% (Fig. 2, C and D). Overall, these data show that a small amount of residual -helical structure is present in the disordered Skp monomers, but that the complete formation of the helices requires the trimerization interface. Overall, these data demonstrate that the monomeric state of Skp is intrinsically disordered with some residual helical propensity located at the tip of the arms. In the trimeric structure, the circular-barrel interface, connecting the N- and C-terminal part of the protein, brings helices 2 and 3 close together in space and thus stabilizes their secondary structure (Fig. 2E). This unique mechanism resembles a stapling of the coiled-coil helices to the barrel in the head domain.

(A) Sections of 2D [15N,1H]-TROSY spectra of [U-2H,15N] Skp(WT) (dark gray) and Skp(A108L) (blue) at a concentration of 1 mM and 37C in NMR buffer (20 mM MES, pH 6.5, and 150 mM NaCl). NMR signals of the monomeric state of Skp(WT) are overlaying with the one from Skp(A108L). The assignments of the overlapping NMR signals of the monomeric state are indicated in the panel. (B) Residue-specific secondary backbone chemical shifts of Skp(WT) in 8 M urea solution, Skp(A108L) in its monomeric form, and Skp(WT) in its trimeric form. Positive and negative values indicate -helical and -sheet secondary structure elements, respectively. The gray-shaded area indicates the positions of helices in the Skp trimer. (C) Percentage of helical population in the conformational ensemble of the Skp monomer. Helical regions with 10 to 20% helicity or 20 to 30% helicity are highlighted with light or dark yellow, respectively. (D) Structural model of the Skp monomer. On a configuration of Skp with -helices formed, the degree of residual helical population present in the conformational ensemble is indicated. The large majority of monomeric Skp is disordered. (E) Schematic model of coupled oligomerization and folding mechanism of Skp. Monomeric Skp explores an ensemble of conformations with a low propensity for the formation of the arm -helices. The formation of the oligomerization interface brings the N and C termini together (red arrows), thus stabilizing the coiled-coil structure of the -helical arms.

It has been previously proposed that the monomeric state of Skp would be well folded rather than disordered (37). That conclusion was obtained from indirect measurements of the molar heat capacity change Cp of trimer formation by a vant Hoff analysis of temperature-dependent AUC data, which indicated a value of Cp = 0.62 0.11 kcal mol1 K1 for the Skp monomer-trimer transition. Because the authors expected a value for a coupled folding and oligomerization of Cp = 8.01 3.3 kcal mol1 K1, they concluded that only trimerization, but not folding, would take place during oligomerization. To resolve these different views, we determined Cp of Skp(WT) directly by differential scanning calorimetry (DSC) to Cp = 2.9 0.4 kcal mol1 K1 at 37C (fig. S3). Considering the average residual helicity of 21% in the monomer, this corresponds to a value of 1.1 kcal mol1 K1 for folding of one monomer subunit, which is a similar value to proteins of the same size (40, 41). We note that Cp is strongly temperature dependent (fig. S3), which may have perturbed the precision of the vant Hoff analysis by Sandlin et al. (37).

Having established that Skp activation comprises an equilibrium between a folded trimer and a disordered monomer, it appears relevant to understand how this equilibrium contributes to Skp chaperone activity. As a model client, we use the native client protein tOmpA, an eight-stranded transmembrane domain of OmpA. tOmpA, when bound to Skp, adopts a conformational ensemble of rapidly reorienting conformers (15). To investigate whether tOmpA binds to the trimeric or the monomeric state of Skp, or to both, we used an activity assay with all mutants. In a first step, we measured the chaperone activity by quantifying the amount of Skp-bound tOmpA. Intriguingly, the activity correlated with the concentration of the trimer for all Skp variants, such that, e.g., Skp(A108L) has around 50% of the Skp(WT) activity and that no chaperone activity could be detected for Skp(V117P) and Skp(A108R) (Fig. 3, A and B, and fig. S4).

(A) Holdase activity of Skp variants as determined by the amount of aggregation-prone tOmpA solubilized in equilibrium. Values are normalized to the activity of Skp(WT). Error bars represent the SD of 15 individual signals of tOmpA. (B) 2D [15N,1H]-TROSY fingerprint spectra of [U-2H,15N]-tOmpA bound to unlabeled Skp(WT) or Skp(A108L). Spectra were recorded at a temperature of 37C in NMR buffer (20 mM MES, pH 6.5, and 150 mM NaCl). A 1D 1H cross section shows the intensity of alanine-176. (C) Combined amide chemical shift differences between [U-2H,15N]-Skp(WT) and [U-2H,15N]-Skp(A108L) with bound unlabeled tOmpA. The magnitude of 2 SDs [0.053 parts per million (ppm)] is indicated by a dashed line. (D) Structural model of Skp(108L) with bound tOmpA. Amide groups with chemical shift changes larger than 2 SDs upon binding of tOmpA to Skp(A108L) are marked in light blue. The position of A108 is indicated by a blue circle. (E and F) 2D [15N,1H]-TROSY fingerprint spectra of [U-2H,15N] Skp(A108L) in the absence (E) and presence (F) of unlabeled tOmpA. Spectra were recorded at 37C in NMR buffer. The spectral area 7.5 to 8.5 ppm in 1H, corresponding to disordered protein states, is indicated by gray lines. 1D 1H cross sections of lysine-141 in the monomeric (M) and trimeric (T) state of Skp are shown, and the relative fractions are indicated.

We then selected Skp(A108L) to characterize structure and arrangement of the tOmpA-Skp(A108L) complex. First, addition of tOmpA to Skp(A108L) increases the apparent molecular mass in SEC-MALS experiments (fig. S4). Second, the 2D [15N,1H]-TROSY NMR spectra of isotope-labeled tOmpA bound to unlabeled Skp(A103L), Skp(A103R), Skp(A108L), or unlabeled Skp(WT) are highly similar (Fig. 3B and fig. S4). Because the chemical shift is a population-weighted average over the individual conformers in the tOmpA ensemble, this observation indicates that the client conformational ensemble inside the chaperone is essentially unperturbed by the local structural adaptations, resulting from the mutation A103L, A103R, or A108L. Third, a direct spectral comparison showed that the chemical shift perturbations that occur on the Skp trimeric state upon tOmpA binding are highly similar for Skp(WT), Skp(A103L), Skp(A103R), and Skp(A108L) (Fig. 3, C and D, and fig. S4). Identically to the apo state, only one set of NMR signals is present for the trimeric state, showing that the complex with tOmpA does not involve other stable oligomeric states (Fig. 4, D to G). Furthermore, for all mutants with a considerable population of the trimeric state, binding of tOmpA induces similar chemical shift perturbations, confirming a similar mode of binding (Fig. 3, C and D, and fig. S3). As a consequence, the structural description that was previously established for the Skp-tOmpA complex (15) can be assumed in good first-order approximation also for Skp(A108L), although the thermodynamics and kinetic of the ensemble are somewhat different (Fig. 3, B to D).

(A) Fitness of Salmonella strains with various chromosomal skp mutations in rich lysogeny broth. Data for individual cultures and means are shown. (B) Fitness of Salmonella strains in a mouse infection model. Each circle represents data for one mouse from a total of two independent infection experiments (****P < 0.0001 and ***P < 0.001; statistical significance of difference to values for WT based on t test with Holm-dk correction for multiple comparisons). Corresponding competitive index data are shown in fig. S4. (C) Functional cycle of Skp. In the absence of client proteins, Skp populates the periplasm in monomeric form up to low micromolar concentrations. These partially disordered monomers are functionally inactive. An emerging Omp client at the inner membrane recruits an active trimeric chaperone from the ensemble equilibrium. Upon release of the client, trimeric Skp dissociates and the monomers enter the pool of inactive disordered conformations. See text for details.

Then, we investigated the effect of tOmpA binding on the Skp monomer-trimer equilibrium at a temperature of 37C, where Skp(A108L) is more than 80% in its monomeric state and Skp(A108R) and Skp(V117P) are completely monomeric (Fig. 3, E and F, and fig. S4). For Skp(A108L), binding of tOmpA resulted in a strong shift of the population levels from the monomeric toward the trimeric state, while no change was observed for Skp(A108R) and Skp(V117P) (Fig. 3, E and F, and fig. S4). Furthermore, for all Skp variants with considerable population of the monomeric state, the NMR signal positions of the monomeric state were not perturbed by the addition of tOmpA, confirming that there is no detectable interaction between monomeric Skp and the Omp client (fig. S3). This is an additional proof that only the structured trimer, but not the disordered monomer, has chaperone activity.

Because a bound tOmpA client is in direct contact with all three arms of Skp simultaneously (15), client binding contributes by avidity to the thermodynamic stability of the trimeric state of the chaperone. We quantified the difference in free energy of apo-Skp(WT) in comparison to tOmpA-Skp(WT) by a denaturation titration (fig. S4). Binding of tOmpA to Skp(WT) increased its stability by 1.7 kJ mol1, corroborating the stabilization effect of the trimeric state by the binding of its client protein. Overall, the data show that monomeric, disordered Skp does not interact with the Omp client and that client binding increases the stability of the Skp trimer by avidity, thus shifting the conformational equilibrium toward the trimeric state.

Skp is dispensable for growth of various bacterial species under rich laboratory conditions. However, bacterial pathogens such as Yersinia and Salmonella require Skp for growth in hostile host tissue. To determine whether the Skp activation mechanism that we identified is important under these physiologically relevant conditions, we engineered analogous point mutants in Salmonella enterica serovar Typhimurium. Salmonella Skp is highly homologous to E. coli Skp, with 91% identity (fig. S4). We selected three of the mutations for these experiments, the two mildest ones A103R and A103L, as well as V117P, and also engineered a strain with complete genetic deletion of the skp gene (skp). As expected, neither the point mutants nor a full skp deletion affects Salmonella fitness in rich lysogeny broth (Fig. 4A and Table 1). We then tested the same mutants in competitive infections in a mouse typhoid fever model. In competitive infections, mice are infected with a mixture of WT and mutant strains. Plating of bacteria retrieved from spleen of these mice yields the fitness of mutants relative to the WT bacteria in each mouse. This approach reduces interindividual variance and offers higher statistical power with limited numbers of experimental animals compared to single-strain infections. The data reveal a slight but significant fitness defect of Salmonella skp(A103L) compared to WT and strong fitness defects for mutants skp(A103R) and skp(V117P), which are comparable to the full skp deletion (Fig. 4B and Table 1; competitive index data in fig. S4). These results show that already subtle perturbations of the Skp monomer-trimer equilibrium diminish Skp function in vivo and that perturbation of this equilibrium by less than an order of magnitude in C0.5 completely abolishes Skp function, rendering bacteria nonvirulent.

In this work, we have elucidated the activation mechanism of the molecular chaperone Skp at atomic resolution. The monomer state of Skp is intrinsically disordered, with a limited residual propensity of -helicity in the coiled-coil tentacle arms. This low inherent stability of helices 2 and 3 is particularly interesting, because they are not involved in inter-subunit contacts in the trimer structure. The formation of the head domain trimer merely fixes the positions of the end points of the -helices in space, thus stabilizing them by reducing the conformational entropy of the unfolded state. This unique mechanism resembles a stapling of the coiled-coil helices to the barrel in the head domain. A directly related effect is being exploited in peptide chemistry to stabilize helical conformation of small peptides by a suitably chosen covalent circularization, the so-called stapled peptides (42). Furthermore, because the tOmpA client is in simultaneous direct contact with all three Skp subunits, its binding stabilizes and shifts the oligomer equilibrium of Skp toward the trimeric state. Last, the disordered Skp monomer does not exhibit chaperone activity.

These mechanistic insights integrate into an improved picture of the functional cycle of Skp in the bacterial periplasm (Fig. 4C). Monomeric, disordered Skp molecules populate the periplasmic space. As soon as a client protein emerges from the Sec translocase, the inactive monomers fold and assemble into a trimeric state around the unfolded client protein. Skp directly or indirectly transports the chain to the Bam complex for folding and insertion in the membrane and possibly also to DegP for degradation (27, 43). The exact mechanism of client release is not understood, but besides direct migration to a higher-affinity target, one exciting possibility to enhance the release kinetics could be a destabilization of the oligomeric state of Skp or a stabilization of the monomeric state of Skp in the vicinity of the downstream receptor of the substrate. This may include negative charges on membranes or BamA (36, 44, 45). After client release, the disordered Skp monomers enter the periplasmic reservoir of individually inactive chaperones. The absence of a chaperoning activity of the monomer ensures that only Skp molecules with complete cavity bind clients, providing maximal chaperoning effect in an all-or-none fashion. At the same time, it introduces a directionality of the chaperoning effect toward the center of the cavity, avoiding spurious binding effects that would not be directed into the Skp cavity. These could potentially destabilize periplasmic proteins that are not intended client proteins. Last, the disordered nature of monomeric Skp might facilitate its import into the periplasm through the Sec complex upon its own biogenesis. Additional impact for this type of activation mechanism comes from a direct comparison to the activation mechanism of the chaperone SurA (18). SurA is constitutively active with just a dynamic modulation of its activity upon rotation of a domain connected by linkers to its chaperone core, i.e., its activity is only weakly regulated (18). Skp activity, in turn, is strongly regulated, with a switch between a completely inactive and an active state, as shown in this work. This stark difference matches a fundamental difference in function of these two periplasmic chaperones. Skp has high affinity for its client proteins and a strong tendency to prevent their folding and therefore presumably requires to be tightly regulated to avoid unspecific chaperone activity under no-stress conditions, whereas SurA binds unfolded OMPs with lower affinity while promoting their folding and therefore presumably does not require a strong regulation of its chaperone activity (15, 4649).

The Skp activation mechanism provides an elegant example how a chaperone can regulate its functional cycle in an environment depleted of any source of energy. For ATP-independent chaperones, only three types of activation mechanisms have so far been described: an order-to-disorder transition [Hsp33 (10), HdeA (11), and HdeB (12)], oligomer disassembly [sHSP (6) and TF (79)], and no or minor conformational change [Spy (13, 14), HSP40 (16), SecB (17), and SurA (18)]. Skp is the first chaperone found to feature these activation mechanisms in the opposite direction and even combine them, i.e., by a disorder-to-order transition that is coupled to oligomerization. The high (nM)affinity Skp has for its client proteins and the strong tendency to prevent their folding could represent a potential hazard to the cell (15, 49). The coupled folding and oligomerization mechanism ensures that holdase function is only present in the trimer where it is geometrically oriented only toward the chaperone cavity. Under nonstressed conditions, Skp exists as an inactive disordered monomer with a minor population of active folded trimer to avoid detrimental effect for the cells. At the opposite, under stress conditions, up-regulation of the Skp concentration and binding to client proteins shift the equilibrium toward the trimeric folded active state, protecting the cells by preventing aggregation of unfolded protein. While most chaperones use strategies to cover a preexisting client binding site in their inactive state, Skp has thus evolved a more extreme mechanism where the client binding area exists only in the active state. This strong regulation allows the tight control of Skp activity while providing at the same time a fast mechanism for client release upon dissociation into the disordered monomeric state. The chaperone activity of Skp is thus regulated in dynamic response to chaperone concentration and client availability.

Skp, lacking its signal sequence, was cloned from genomic DNA through Nde I and Xho I into the pET28b expression vector (Novagen) containing a thrombin-cleavable N-terminal His6-tag (15). Skp was expressed in BL21-( DE3)-Lemo cells [New England Biolabs (NEB)] transformed with the Skp plasmid and grown at 37C in M9 minimal medium containing kanamycin (30 mg/ml) to OD600 (optical density at 600 nm) = 0.6, and then the expression was induced by adding 0.4 mM isopropyl--d-thiogalactopyranoside (IPTG) at 25 for 12 hours. Uniformly [2H, 13C, 15N]-labeled protein was prepared by growing cells in D2O-based M9 minimal medium, with 1 g of 15NH4Cl and 2 g of [U-13C,2H] glucose per liter of medium. Cells were harvested by centrifugation at 5000g for 20 min. The pellet was resuspended in 20 ml of lysis buffer A per liter of culture [20 mM tris (pH 7.5), 500 mM NaCl, deoxyribonuclease (DNase) (0.01 mg/ml), ribonuclease (RNase) (0.02 mg/ml), and inhibitor cocktail (cOmplete EDTA-free protease inhibitor; Roche)]. Cell lysis was performed using a microfluidizer (Microfluidics) for three cycles at 4C. The soluble bacterial lysate was separated from cell debris and other components by centrifugation at 14,000g for 60 min and loaded onto a Ni-NTA (nitrilotriacetic acid) column (Qiagen). Skp eluted at 250 mM imidazole concentration and was dialyzed against buffer [20 mM tris (pH 7.5) and 500 mM NaCl] overnight to remove the imidazole. In a final step, a size exclusion chromatography (Superdex-200 16/600 PG) step was applied to further purify the proteins and adjust the protein to its final buffer [20 mM MES (pH 6.5) and 150 mM NaCl]. Note that the His6-tag was consistently not cleaved from all Skp constructs, because in both our hands and published work by others (37), the presence of the His6-tag was found to not change the monomer-trimer equilibrium constant and because monomeric, disordered Skp was found to be sensitive to proteolytic degradation. Afterward, Skp was concentrated by ultrafiltration and stored at 20C until use. Final yield of purified protein was 25 mg for Skp(WT) and mutants per liter of deuterated M9 minimal medium.

The transmembrane domain of OmpA (residues 1 to 177) was cloned through Nco I and Xho I into the pET28b expression vector without any affinity tag and lacking its signal sequence (15). BL21-( DE3)-Lemo cells (NEB) were transformed with the tOmpA expression plasmid and grown at 37C in medium containing kanamycin (30 g/ml) to OD600 = 0.8. Expression was induced by 1 mM IPTG. Cells were harvested 4 hours after induction and resuspended in 20 ml of buffer B per liter of culture (20 mM tris-HCl and 5 mM EDTA, pH 8.5). Cell lysis was performed using a microfluidizer (Microfluidics) for three cycles at 4C. Purification from inclusion bodies was done as described (50). The ion-exchange elution fractions containing tOmpA were pooled and dialyzed against buffer B. The precipitate was resuspended in 6 M Gdm/HCl and stored at 20C until usage. Final yield of purified protein was 50 mg of tOmpA per liter of deuterated M9 minimal medium.

The QuikChange II mutagenesis protocol (Stratagene) was used to introduce the mutations A108L, A108R, A103L, A103R, or V117P into Skp. Polymerase chain reaction (PCR) primers (Table 2) were obtained from Microsynth. The expression and purification of the mutant proteins was performed as described for the WT proteins. The final yield of purified mutants was similar to WT.

Salmonella strains used in this study were based on S. enterica serovar Typhimurium SL1344 hisG xyl (51, 52). Salmonella mutants with gene deletions were obtained by two consecutive single crossovers with positive selection for resistance to kanamycin and negative selection for levansucrase-mediated sensitivity to sucrose. Salmonella was grown in lysogeny broth containing NaCl (5 g/liter; Lennox LB). Each strain was transformed with a low-copy plasmid expressing a distinct fluorescent protein (mtagBFP2, mNeonGreen, YPet, or mCherry). These plasmids have no impact on in vivo fitness (53, 54). All animal experiments were approved (license 2239, Kantonales Veterinramt Basel) and performed according to local guidelines (Tierschutz-Verordnung, Basel) and the Swiss animal protection law (Tierschutz-Gesetz). Eight 10- to 16-week-old female BALB/c mice (Charles River Laboratories) were infected by tail vein injection of mixtures containing WT Salmonella and different combinations of three mutants with about 1000 colony-forming units (CFU) each per strain. The exact inoculum size for each strain was determined by plating. After 4 days, mice were euthanized with carbon dioxide and Triton X-100 detergenttreated spleen homogenates were prepared as described previously (55). Total Salmonella loads were determined by plating dilution series on agar plates. Mutant-to-WT ratios were determined by flow cytometry counting of bacterial cells falling into gates indicative for the various fluorescent proteins using optical filters (55). Fitness was calculated as log2(FI), with FI corresponding to the fold increase starting from the initial spleen colonization [around 20% of the inoculum (56)] to the final spleen load for each strain. The relative fitness value of co-administered WT Salmonella was set to 100%. We also determined the more commonly used readout competitive index by dividing the output ratio (mutant/WT) by the inoculum ratio (mutant/WT).

Complex assembly was carried out following a modified version of the protocol published by Burmann et al. (15). A 1.5 M excess of denatured tOmpA was added to Skp(WT) or mutants in 20 ml of assembly buffer [20 mM MES (pH 6.5) and 150 mM NaCl] in a dropwise fashion under continuous stirring. The solution was then stirred for another 1 hour to ensure saturation of the chaperones. After centrifugation at 10,000g for 30 min, the supernatant fraction, containing the Skp-tOmpA complexes, was separated from the pellet, containing the precipitated tOmpA. The supernatant was exchanged by ultrafiltration to NMR buffer [20 mM MES (pH 6.5) and 150 mM NaCl], and after concentration, the volume was adjusted to 250 l. The chaperone activity of Skp(WT) and mutant was determined by quantifying the NMR signals in 2D [15N,1H]-TROSY spectra of [U-2H,15N]-tOmpA bound to unlabeled Skp. Control sample of [U-2H,15N]-tOmpA in NMR buffer was prepared following the reference protocol, showing that, in the absence of the functional Skp(WT), less than 2% of [U-2H,15N]-tOmpA signals were observed in comparison to [U-2H,15N]-tOmpA bound to Skp(WT).

All NMR experiments for Skp-Omp complexes were performed in NMR buffer [20 mM MES (pH 6.5) and 150 mM NaCl]. The experiments were recorded at the specified temperature on a Bruker AscendII 700 MHz or Avance 800 MHz spectrometer running Topspin 3.0 and equipped with a cryogenically cooled triple-resonance probe. For the sequence-specific backbone resonance assignment of [U-99% 2H, 13C, 15N]-Skp(A108L), the following NMR experiments were recorded at 37C: 2D [15N,1H]-TROSY, 3D TROSY-HNCA, 3D TROSY-HNCACB, 3D TROSY-HNCO, and 3D TROSY-HN(CA)CO. NMR data were processed with nmrPipe (57) and analyzed with CARA and ccpnmr (58). Secondary chemical shifts were calculated relative to the random-coil values of Kjaergaard and Poulsen (59). For the backbone assignment of the unfolded [U-2H,15N,13C]-Skp(WT), automated projection spectroscopy (APSY) experiments were recorded in NMR buffer [20 mM MES (pH 6.5) and 150 mM NaCl] containing 8 M urea at 15C. The 5D APSY-HNCOCACB (60) was recorded with 54 transients for Skp, two scans per transient, 0.7-s recycle delay, and 1024 150 complex points in the direct and indirect dimensions. The 4D APSY-HNCACB (60) was recorded with 46 transients, two scans per transient, 0.7-s recycle delay, and 1024 180 complex points in the direct and indirect dimensions, respectively. The GAPRO (geometric analysis of projections) (60) analysis of the projection spectra was carried out with = 5.0 Hz, Rmin = 15.0 Hz, S/N = 7.0, and Smin,1 = Smin,2 = 8 for the 5D APSY-HNCOCACB and with = 5.0 Hz, Rmin = 15.0 Hz, S/N = 10.0, and Smin,1 = Smin,2 = 15 for the 4D APSY-HNCACB. As the signals for glycine residues within the 4D APSY-HNCACB and the signals of residues succeeding glycines within the 5D APSY-HNCOCAB have a different sign than the other resonances, the GAPRO algorithm was run twice for positive and negative peaks, respectively, and the two resulting peak lists were combined. The combined peak lists were assigned by using the newest version of the MATCH algorithm within the UNIO10 software package, yielding a 65% complete assignment for Skp. By using a conventional 3D TROSY-HNCACB experiment, complete backbone assignment for Skp could be obtained. NMR data were processed using PROSA (61) and analyzed with CARA and XEASY. Combined chemical shift differences of the amide resonances in 2D [15N,1H]-TROSY spectra were calculated asHN=((H1))2+(0.2(N15))2(1)

SEC-MALS measurements of Skp were performed at 25C in NMR buffer [20 mM MES (pH 6.5) and 150 mM NaCl] using a GE Healthcare Superdex-200 Increase 10/300 GL column on an Agilent 1260 high-performance liquid chromatography. Elution was monitored using an Agilent multi-wavelength absorbance detector (data collected at 280 and 254 nm), a Wyatt Heleos II 8+ multiangle light-scattering detector, and a Wyatt Optilab rEX differential refractive index detector. The column was equilibrated overnight in the running buffer to obtain stable baseline signals from the detectors before data collection. Inter-detector delay volumes, band broadening corrections, and light-scattering detector normalization were calibrated using an injection of bovine serum albumin solution (2 mg/ml; ThermoPierce) and standard protocols in ASTRA 6. Weight-averaged molar mass, elution concentration, and mass distributions of the samples were calculated using the ASTRA 6 software (Wyatt Technology).

DSC data were acquired using a Microcal VP-Capillary DSC instrument (Malvern Panalytical, Malvern UK) at a Skp trimer concentration of 24.4 M (i.e., 73 M concentration in terms of monomer). After centrifugation, protein concentration was determined by ultraviolet spectrophotometry using a molar extinction coefficient of 4470 M1 cm1 at 280 nm for the trimer and correcting for minor scattering contributions to apparent absorbance. The Skp sample was scanned from 15 to 105C at a scan rate of 1C/min, and data points were acquired at 0.1C increments. Multiple buffer versus buffer scans, performed before the sample scan to establish the instrumental heat capacity baseline, were averaged and subtracted from the sample scan data, which were then normalized to excess molar heat capacity using the trimer concentration. Attempts to fit the complex thermogram with standard models of oligomer dissociation and denaturation proved unsuccessful, so Cp for folding was estimated from the difference between the slopes of the excess molar heat capacity in low- and high-temperature regions (the apparent pre- and post-transition baselines), fitted by linear regression, and extrapolated to the temperature of interest.

The chemical equilibrium between trimeric and monomeric Skp can be described by the reaction3SmSt(2)where the equilibrium constant L13 is given byL13=[St][Sm]3(3)where [St] and [Sm] are the molar concentrations of Skp trimers and free Skp monomers, respectively, and L13 has units of M2. For this equilibrium, the concentration of trimer [St] as a function of total Skp [S0] is given by Sandlin et al. (37)[St]([S0])=[S0]3+(23+)13+(23+)13(4)where , , and are given by=9L13[S0]2+1L13(5)=[S0]2981(6)=[S0]318[S0]162(7)

The fraction of total Skp protein that is trimeric at any total Skp concentration equalsfSt=3[St][S0](8)and the fraction of total Skp protein that is monomeric equalsfSm=1fSt(9)

In SEC-MALS experiments in equilibrium situations, the detected molar mass represents the concentration-weighted average mass of the species involvedMw=(ciMi)(ci)(10)where ci is the mass concentration and Mi is the molar mass of the ith species. Therefore, for the monomer-trimer equilibrium, by comparison with the limits for the completely monomeric or trimeric species, the weight-averaged mass reports directly on the fractional populations asfSm=MobsMStMSmMSt(11)where Mobs is the detected weight-averaged mass, and MSm and MSt are the detected masses of the completely monomeric and trimeric state, respectively.

For the estimation of the population of monomeric and trimeric states for the WT and mutants, the residue lysine-141 was chosen, because its signals are well resolved in each state and it is located in a nonstructured, locally flexible region in the trimer. The fractions were estimated by calculating the ratio of the intensity of the signals in the monomeric and trimeric state according to the equationfSm=ISmISm+ISt(12)where ISm and ISt are the intensity of the residue lysine-141 in the monomeric and trimeric state, respectively. Similarly, for the denaturation titration, the fractions of folded and unfolded Skp were determined from the signals of residue lysine-141, and for each titration point, G was calculated assuming a two-state model according to the equationG=RTlnfStfSm(13)

The data were fitted by linear regression, and G was extrapolated to a concentration of 0 M urea.

Acknowledgments: We thank C. Johnson for help in setting up the DSC experiments and the Biophysics Facility of the MRC Laboratory of Molecular Biology, Cambridge, for access to the DSC instrument. Funding: This work was supported by the Swiss National Science Foundation (grants 310030B_185388 and 407240_167125 to S.H. and 310030_182315 to D.B.). Author contributions: G.M., S.H., and D.B. designed the study, analyzed the data, discussed the results, and wrote the paper. G.M. and T.S. conducted the SEC-MALS experiments. T.S. conducted the DSC experiment. B.M.B. conducted the assignment of urea-unfolded Skp(WT). B.C. engineered the Salmonella mutants and conducted the mouse experiments. G.M. conducted all other experimental work. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors. Sequence-specific resonance assignments have been submitted to the Biological Magnetic Resonance Data Bank under the following accession codes: Skp(WT) in 8 M urea, 26613; monomeric Skp(A108L), 50195.

See the article here:
Regulation of chaperone function by coupled folding and oligomerization - Science Advances

Abstract

The serpinopathies are among a diverse set of conformational diseases that involve the aberrant self-association of proteins into ordered aggregates. 1-Antitrypsin deficiency is the archetypal serpinopathy and results from the formation and deposition of mutant forms of 1-antitrypsin as polymer chains in liver tissue. No detailed structural analysis has been performed of this material. Moreover, there is little information on the relevance of well-studied artificially induced polymers to these disease-associated molecules. We have isolated polymers from the liver tissue of Z 1-antitrypsin homozygotes (E342K) who have undergone transplantation, labeled them using a Fab fragment, and performed single-particle analysis of negative-stain electron micrographs. The data show structural equivalence between heat-induced and ex vivo polymers and that the intersubunit linkage is best explained by a carboxyl-terminal domain swap between molecules of 1-antitrypsin.

The misfolding of proteins and their spontaneous ordered aggregation underlie the pathology of Alzheimers, Huntingtons, and Parkinsons diseases; amyloidoses; and serpinopathiesthe latter involving self-association of mutant members of the serine protease inhibitor (serpin) superfamily. 1-Antitrypsin is a 52-kDa serpin expressed and secreted predominantly by hepatocytes and is the most abundant circulating protease inhibitor. The primary physiological role of 1-antitrypsin is the inhibition of neutrophil elastase, a protease whose production is increased during the acute phase inflammatory response (fig. S1, A and B). However, genetic variants such as the severe Z (E342K) allele of 1-antitrypsin promote proteasomal degradation and the formation of ordered linear polymers (1, 2). Despite the pronounced retention in the endoplasmic reticulum (ER), 1-antitrypsin polymers do not typically initiate the unfolded protein response. Instead, these ordered aggregates can be sequestered into ER-derived inclusion bodies that are associated with liver disease. The lack of circulating 1-antitrypsin results in dysregulation of neutrophil elastase and hence tissue destruction and emphysema (2).

The structure of the pathological polymers that accumulate in patients has not been demonstrated. The observation that 1-antitrypsin polymers show a similar degree of stabilization to the cleaved form (3) (fig. S1B, EI) and that peptide analogs of the inserted portion of the reactive center loop (RCL) could similarly stabilize the protein (4) and prevent polymerization (1, 3) suggested that polymers were the product of an interaction between the RCL of one molecule and sheet A of the next (1). This loop-sheet model (Fig. 1A, hypotheses H1 and H2) is consistent with nuclear magnetic resonance and H/D (hydrogen-deuterium) exchange data showing that polymerization proceeds via a compact, rather than an expanded, intermediate (5, 6). The subsequently proposed -hairpin hypothesis (Fig. 1A, H3) was based on the crystal structure of a self-terminating dimer of a homologous protein, generated artificially at low pH, and extrapolated to 1-antitrypsin using limited proteolysis and recombinant mutants with stabilizing disulfide bonds (7). The C-terminal model (Fig. 1A, hypothesis H4) posits that the C terminus fails to form properly in the donor molecule and is instead incorporated into an acceptor molecule, with latent-like self-insertion of the RCL providing the extreme stability found in polymers (8). This model is based on a crystal structure of a denaturant-induced circular trimer of recombinant disulfide-bonded 1-antitrypsin. The circular arrangement of subunits provides a rigid structure that is tractable for crystallography but reflects a minor component of the source sample that is not generally enriched in polymer preparations (1), although there is an absence of the latent conformation in humans that would be predicted to be a by-product of this mechanism (9).

(A) Different linkages hypothesized for the pathological polymer, H1 to H4, with the intermolecular interface proposed between one monomeric subunit and the next shown in black. (B) (i) Analysis of polymers isolated from intrahepatic inclusion bodies (denoted as ZZ) by 4-12 (w/v) acrylamide SDS-PAGE in comparison with the monomeric wild-type (M) variant purified from human plasma and visualized by Coomassie blue R stain. (ii, iv, and v) Western blots of ex vivo polymers (ZZ), polymers of the M variant induced by heating (H), and monomeric M variant (M) separated by denaturing SDS-PAGE (top) and nondenaturing native PAGE (bottom) and probed with a conformation-insensitive rabbit polyclonal antibody (pAb AAT, left) or a mouse monoclonal selective for polymeric 1-antitrypsin (mAb 2C1, right). No monomer is visible by native PAGE in the heat or ZZ preparation. (iii) Sensitivity of ex vivo Z 1-antitrypsin to PNGase F (+P) or EndoH (+E), the latter preferentially cleaving high-mannose glycans. (C) Representative micrograph of polymers isolated from ex vivo liver tissue, visualized by 2% (w/v) uranyl acetate negative stain using a Tecnai 120-keV transmission electron microscope at a magnification of 92,000. The image has been low-passfiltered to 30 . Black scale bar, 50 nm. Details of some polymers are shown at the right. (D) Same material, labeled with the Fab fragment of the 4B12 monoclonal antibody (Fab4B12), and visualized under the same conditions. Scale bar, 50 nm. Details from micrographs are shown at the right; readily discernible Fab protrusions are highlighted by arrows.

The question remains unresolved as to which polymerization model, if any, describes a realistic organization of the pathological polymer. To address this issue, we have performed a structural characterization of polymers from explant liver tissue of individuals homozygous for the Z allele who had undergone orthotopic transplantation. This has allowed us to define structural limits on the pathological polymer and to critically evaluate the proposed models in this pathological context.

Tissue samples were obtained from the explanted livers of individuals homozygous for the Z allele of 1-antitrypsin. After isolation of inclusion bodies, polymers released by sonication were found to contain a major component that resolved at ~50 kDa when dissociated and visualized by denaturing SDSpolyacrylamide gel electrophoresis (SDS-PAGE) (Fig. 1B, i). It was confirmed to be 1-antitrypsin by Western blot analysis (Fig. 1B, ii). The difference in migration with respect to monomeric material purified from human plasma (Fig. 1B, i and ii) was no longer observed following treatment with PNGase F or EndoH (Fig. 1B, iii). This is diagnostic for glycosylated material that has not undergone maturation in the trans-Golgi network and therefore has been retained by the cell. When visualized by nondenaturing PAGE, the protein migrated with a broad size profile with some discrete bands visible, it was reactive with the polymer-specific (10) monoclonal antibody mAb2C1, and it was free of detectable monomer (Fig. 1B, iv and v).

The liver-derived polymers were applied to carbon-coated copper grids and negatively stained with 2% (w/v) uranyl acetate; polymers could easily be distinguished in the resultant electron microscopy (EM) images by a beads-on-a-string appearance (1), with a curvature of the chain and an absence of branching (Fig. 1C). While some circular forms were present, in contrast to a small-angle x-ray scattering (SAXS) analysis of polymeric material produced in the cytoplasm of Pichia pastoris (11), most (~80%) were nonself-terminating with clearly separated termini.

Polymer subunits are ~50 kDa in size, their ellipsoidal shape has few distinct features that would aid orientation, and they are connected by linkages that appear flexible. These properties provide confounding factors to processing by single-particle analysis. To facilitate subsequent image processing, we doubled the effective size of the polymer subunits and introduced an orienting feature by labeling polymers with the antigen-binding fragment of the 4B12 monoclonal antibody (Fab4B12) (12). This antibody was selected as it recognizes all folded forms of 1-antitrypsin including the polymer, and the location of its epitope is well established (1214).

Following the addition of Fab4B12 at a stoichiometric excess to the 1-antitrypsin subunits and removal of unbound material, the polymer sample was visualized using negative-stain EM (NS-EM) (Fig. 1D). Fab4B12-labeled polymer subunits demonstrated additional density visible as tooth-like protrusions (Fig. 1D, insets). On consecutive subunits, Fabs were, in general, present on the same side of the polymer chain, potentially indicating a preference of the angular relationship around the polymer axis. Conversely, opposing 1-antitrypsinFab4B12 orientations, which would report substantial orientational freedom around the intersubunit linkage, were observed only infrequently.

The heterogeneity and flexibility of ex vivo polymers make them unsuitable for crystallography. Modern protocols for single-particle reconstruction of three-dimensional (3D) objects using EM images enable us to explicitly address heterogeneity in samples, and we therefore sought to structurally characterize the pathological polymers using this technique. A NS-EM image dataset of Fab4B12-labeled polymers was compiled from 100 30-frame movies that had been collected using a DE-20 direct detector and a Tecnai 200-keV transmission electron microscope. Preliminary experiments indicated that polymer flexibility would represent a challenge for a single-particle reconstruction approach. Thus, a minimal segment required to investigate the linkage between monomersa dimer of adjacent subunitswas chosen for the subsequent structural analysis.

The processing pathway for single-particle reconstruction is described in more detail in the Supplementary Materials and in fig. S2 and is summarized here. Initially, images of dimer particles were manually selected from regions of polymers that appeared by eye to be side views with relatively little curvature (fig. S2b) and divided into classes using the Class2D function of RELION (15). The class sums included dimers in which the subunits appeared as adjacent ellipses, and many subunits exhibited a protuberance with the characteristic narrow midriff present in Fab structures (fig. S2d). In some classes, these Fab4B12 subunits were poorly resolved, suggesting variability in rotation between adjacent subunits. Seven classes with well-defined Fab4B12 components were used as references for autopicking from the same set of micrographs; after removal of poorly defined components, this yielded ~100,000 230 230 particle images. This dataset, DA,100K, was found by 2D classification to be more diverse and less dominated by long-axis dimer views (fig. S2f). Later in the course of processing, a subset of 69,000 dimer images (DB,69K) was extracted from a 2D reclassification of the same dataset (fig. S2k).

One class in particular showed two well-resolved Fab subunits (fig. S2h). To generate an initial model-agnostic reference map for 3D classification, we converted this 2D image to a 3D surface representation (fig. S2h, right) with the height (along z in both directions) at each x,y coordinate proportional to the grayscale value of the corresponding pixel in the image (fig. S2h, right). This map was used as a reference for 3D classification of the DA,100K dataset (fig. S2i). In two of eight resulting maps, both 1-antitrypsin molecules exhibited Fab4B12 protrusions. The best-defined map was divided in half, and one subunit was used as a monomer input reference in a reclassification of DA,100K (fig. S2j). Following several iterations of 3D classification, five of eight classes exhibited either one or two well-defined 1-antitrypsinFab4B12 subunits (fig. S2n). These maps were divided in half, and the monomer subunits were individually superimposed and averaged together, providing a consensus density for the 1-antitrypsinFab4B12 monomer subunit Monav (fig. S2o, left). Monav was used as the reference map in successive rounds of 3D classification. Eventually, two classes were identified that showed connected 1-antitrypsin molecules with clear Fab4B12 subunits, comprising 9200 and 6200 particle images, respectively (fig. S2, p and q).

These 3D classes differed in the angles between the two 1-antitrypsinFab4B12 subunitsapproximately 60 and 90and were accordingly termed Dim60 and Dim90 (Fig. 2A). Both showed clear Fab4B12 protuberances and connectivity between the volumes representing the 1-antitrypsin molecules. 3D refinement using gold-standard FSC (Fourier shell correlation) analysis provided estimated resolutions of 19.1 and 24.8 , respectively (at a FSC threshold of 0.33) (fig. S3). Other attempts to obtain dimer reconstructions using variations of the processing pathway described above also converged on these two forms and no others.

(A) Orthogonal views of the reconstruction of Dim60 (left) and Dim90 (right) contoured at 3.9 105 3. In this orientation, the connected 1-antitrypsin density is situated at the bottom, and the Fab domains are at the top. Calculated resolutions (using FSC = 0.33) are 19.1 and 24.8 , respectively (fig. S3). (B) Particle images, clustered by view and averaged, that are the basis for the reconstructions. The relative support for each cluster, calculated from the sum of the weights of the constituent images, is shown as circles colored according to a heatmap, highlighting the enrichment of views orthogonal to the dimer axis.

A summary of the constituent particle images, clustered by orientation relative to the 3D reconstructions, can be seen in Fig. 2B. In both cases, the assigned views show that the datasets contain a larger number of side-on views of the dimers, consistent with the observed alignment of most polymers in the micrographs.

Polymers artificially induced at an elevated temperature have often been used to study the process of polymerization (3, 6, 12, 1618). It has been shown that this form shares a common epitope with ex vivo polymers in the vicinity of helices E and F (fig. S1A) (14); the epitope is not recognized when polymerization is artificially induced using a denaturant (10, 16). The lack of discrimination between heat and liver polymers does not, however, demonstrate structural equivalence, and a means of direct comparison between the two has been lacking.

Heat-induced polymers of the plasma-purified M variant were induced and purified, labeled with Fab4B12, and visualized by NS-EM using 2% (w/v) uranyl acetate stain. The resulting images showed the same flexible beads-on-a-string appearance (Fig. 3A), with a greater proportion exhibiting a circularized morphology. The Fab domains once again appeared as teeth-like protuberances with a general preferred orientation on the same side of the polymer axis in adjacent subunits with an occasional apparent ~90 to 180 inversion (Fig. 3B). A new dataset comprising 169 micrograph images was obtained, compiled from 30-frame movies collected using the DE-20 direct detector and the Tecnai 200-keV transmission electron microscope.

(A) Representative micrograph of polymers of M 1-antitrypsin induced at 55C for 48 hours, visualized by 2% (w/v) uranyl acetate negative stain using the Tecnai 120-keV transmission electron microscope at a magnification of 92,000. The image has been low-passfiltered to 30 . Black scale bar, 50 nm. Details of selected polymers are shown at the right. (B) Heat-induced polymers labeled with Fab4B12 and visualized in the same manner. Details from micrographs are shown at the right; discernible Fab protrusions are highlighted by arrows. (C) Orthogonal views of the reconstruction of a Dim60-like structure, with a calculated resolution of 26.4 (FSC = 0.33) (fig. S3). (D) Particles upon which the reconstruction is based, clustered by imputed orientation and with the relative sum of their weights shown as a spectrum. (E) Orthogonal projections of the aligned and contoured Dim60 (blue) and Dim60H (red) structures, with axes shown; overlapping regions appear as magenta. (F) 2D class sums from the liver and heat-induced polymer particle datasets arranged in pairs with columns denoted by L and H, respectively. For each liver polymer class, the most similar heat-induced polymer class by cross-correlation coefficient is shown; gray vertical lines through the images denote identified intensity peaks. (G) Distribution of the interpeak distances for the liver (blue) and heat (red) polymer distances. Dashed lines indicate the means of both sets of data.

We performed autopicking in RELION from the new micrographs using the same 2D references as with the ex vivo dataset (fig. S2d, right) because the heat-induced polymer subunits were of a similar size. Following rounds of 2D classification and cleaning of the image dataset, 25,000 dimer particles were extracted for further image analysis. In 3D classification, the monomeric subunit Monav (fig. S2o, left), obtained from the ex vivo dataset, was used as the reference; monomer rather than dimer was chosen to avoid introducing bias in the relative rotation and translation between subunits. At the final step of classification, a Dim60-type class was identified (Dim60H; Fig. 3C), comprising 6750 particles and with a nominal resolution of 26.4 (at FSC = 0.33; fig. S3). Clustering of particles by their orientation relative to the 3D volume again showed a preference for side views (Fig. 3D). Attempts at reclassification of the residual 18,000 particles failed to reveal further well-defined 3D classes.

In a preliminary model-free analysis, the 1-antitrypsinFab4B12 dimer structure identified from the heat polymer data exhibited a somewhat different intersubunit distance and Fab4B12 orientation to that seen with the liver-derived dataset (Fig. 3, C and E): Translations and rotations of 64 /57 and 69 /65, respectively, were required to superimpose a subunit volume onto the adjacent one. The correspondence more generally between the two datasets was therefore investigated. A comparison was made between all 2D classes obtained from the liver-derived polymer dataset against those calculated from the heat-induced polymer dataset by optimally aligning every possible pair and recording those with the highest correlation coefficient. Most pairs showed good visual correspondence (representative comparisons of class averages are shown in Fig. 3F). Positions of subunits were identified from peaks in the intensity profile of each image. The distribution of distances between these peaks in the aligned classes was almost identical, with a mean of 65 12 and 64 11 (SD) for liver-derived and heat-induced polymer 2D classes, respectively (Fig. 3G). The putative distinction between the dimer volumes is therefore likely accommodated within the observed geometric relationships between subunits in both samples rather than supporting separate linkage mechanisms.

The 3D reconstructions of adjacent subunits reflect the asymmetric character of the Fab4B12-bound subunits and polarity of 1-antitrypsin within the polymer and embody shape, intersubunit distance, and rotational information. Accordingly, they could be used to challenge the different hypotheses regarding the structure of the pathological 1-antitrypsin polymer (Fig. 1A). As the foundation of this analysis, an atomic model of the Fab-antigen complex was required. Protein crystallization trials of Fab4B12 were successful and yielded a 1.9 structure, with the crystallographic parameters summarized in table S1. The asymmetric unit contained two molecules, one of which exhibited fully defined variable loop regions. Despite extensive efforts, it was not possible to obtain a crystal structure of the 1-antitrypsinFab4B12 complex; SAXS data were collected instead. The atomic model of the 1-antitrypsinFab4B12 subunit was then constructed using five sets of experimental data:

1) a consensus density map of the monomer generated by aligning and averaging the individual subunits of the Dim60 and Dim90 reconstructions from the liver polymer dataset (Mon60,90; shown in Fig. 4A, left);

(A) Left: Density for an 1-antitrypsinFab4B12 subunit calculated as the average of the Dim60 and Dim90 subunits, contoured at 1.9 105 3 with a nominal resolution (at FSC = 0.33) of 15.2 (fig. S3). Middle: Result of modeling trials in which complexes between 1-antitrypsin and Fab4B12 molecules with random starting orientations were optimized with respect to the antibody epitope and the subunit density. The resulting structures were evaluated according to their correspondence with the experimental SAXS profile recorded for the complex. A cluster of structures maximizing both parameters are highlighted in red and circled. Right: Superposition on the 1-antitrypsin chain of these five structures showing a consistent relationship between the two components, with the heavy chain in blue and light chain in red. (B) Left: Final model of the subunit shown in the context of the experimental density, with the heavy chain in blue, the light chain in dark green, and 1-antitrypsin sheets A, B, and C in red, pink, and yellow, respectively. The orientations are according to the axes shown in Fig. 3E. Right: Correspondence between the observed SAXS data (black) and the profile calculated from the coordinates of the final subunit model (red). (C) Top: Various polymer images extracted from NS-EM micrographs are shown in red, and 2D projections of polymer models that have been refined against these images are shown in black. Bottom: Mean relative correlations (SD) between each model and the experimental density are shown. Values were calculated for each oligomer relative to the best score observed for that oligomer. Significance was determined by one-way analysis of variance (ANOVA) and Tukeys multiple comparisons test (n = 18); ***P < 0.001 and ****P < 0.0001.

2) the experimentally determined epitope of Fab4B12 (13, 14) at 1-antitrypsin residues 32, 36, 43, 266, and 306 incorporated as a collection of distance constraints on the crystal structures of the individual components;

3) the Fab4B12 crystal structure;

4) the SAXS profile of the complex (Fig. 4B, right); and

5) the structure of cleaved 1-antitrypsin [Protein Data Bank (PDB): 1EZX (19)], as all extant polymer models propose a six-stranded sheet A configuration (Fig. 1A).

Integration of these data during modeling was performed using PyRosetta (20). One thousand randomized starting orientations for 1-antitrypsin and Fab4B12 were subjected to rigid-body energy optimization with reference to these constraints and the Mon60,90 subunit map and scored according to both the cross-correlation coefficient (CCC) with the density and their correspondence with the SAXS profile (Fig. 4A, middle). Backbone and side-chain flexibility was conferred on regions of the Fab likely to contribute to the interface (heavy chain: 27 to 33, 51 to 57, 71 to 76, and 94 to 102; light chain: 27 to 32, 49 to 54, 66 to 70, and 91 to 94) and 1-antitrypsin side chains within the boundaries of the epitope.

The five models that maximized these metrics showed an unambiguous polarity (Fig. 4A, right). One model was selected that best represented this cluster by root mean square distance comparison with the others. This showed the heavy-light chain partition to be oriented off-center along helix A, with the variable-constant domain axis perpendicular to the long axis of the serpin [Fig. 4, A (right) and B (left)]. The cleft between the variable and constant domains of Fab4B12 aligned closely with a central dimple exhibited by the monomer density (denoted by an asterisk in the figure), and the complex corresponded well with the experimentally determined SAXS profile (Fig. 4B, right).

Initial models of the C-terminal (8), loop-sheet (1), and -hairpin (7) polymer configurations (Fig. 1A) were built using the 1-antitrypsinFab4B12 subunit structure (representations of these can be seen in the left column of Fig. 5), differing most substantially in the linker regions connecting adjacent subunits in the polymer chain (detailed in Materials and Methods).

(Top) Different polymer configurations were randomly perturbed by rotation of the subunits with respect to one another and their conformations optimized against Dim60, Dim90, and Dim60H reconstructions. The correlation coefficient after perturbation and before optimization is shown on the x axis, while that after optimization is shown on the y axis. Values are expressed relative to subunits optimized into the density without restriction by a connecting linker. Flexible regions encompassed residues 357 to 368 in all models as well as 340 to 349 (H1), 340 to 352 (H2), and 309 to 328 (H3). (Bottom) The best-fitting model for each polymer configuration and for each of the three dimer EM structures is shown (1-antitrypsin in blue and Fab4B12 in dark green) with respect to the fit of unconstrained subunits (shown in pink). Regions treated as flexible linkers during the optimization are highlighted in light green. For all three reconstructions, the C-terminal model corresponds with the optimum arrangement of subunits.

From an examination of the representative micrographs shown in Fig. 1 (C and D), the intersubunit angular relationships along the polymer chains are not solely accounted for by the Dim60 and Dim90 configurations. Instead, these structures likely correspond to more highly populated species along a continuum of intermediate states. To investigate the compatibility of the loop-sheet, C-terminal, and -hairpin linkages with the arrangement of polymers seen in the micrographs, we used a method that optimized the 3D models to maximize their correspondence with the 2D polymer images. Stretches of residues connecting the dimer subunits were treated as flexible (as specified in Materials and Methods), while the 1-antitrypsinFab4B12 cores behaved as rigid bodies. A selection of 20 oligomers was chosen with different degrees of curvature and subunit orientation (Fig. 4C). Despite a lack of information along the z axis, this approach was able to discriminate between the models on the basis of their ability to adopt the shapes seen in the 2D polymer images: The highly constrained loop-sheet eight-residue insertion model (H1) performed significantly worse than the others (P < 0.0001). The flexibility of the C-terminal domain swap (H4) provided a better fit than the loop-sheet four-residue insertion model (H2) (P < 0.001), and the -hairpin (H3) and C-terminal models (H4) were not distinguishable by this analysis (Fig. 4D).

Next, the compatibility of loop-sheet, C-terminal, and -hairpin configurations with the 3D Dim60, Dim90, and Dim60H reconstructions was evaluated. Each model was repeatedly randomly perturbed by rotation around the dimer long axis (through the 1-antitrypsin subunits) and energy minimized with respect to the EM structures and default stereochemical restraints using PyRosetta (20). This process was undertaken 1000 times for each combination of model and map. As before, the 1-antitrypsinFab4B12 subunits were treated as rigid bodies connected by a flexible linker region. The correspondence between each model and the target map was assessed by the cross-correlation function. These CCC values were denoted as ccperturbed and ccrefined for each perturbed model before and after energy minimization, respectively. Benchmark maximum CCC values were obtained by performing model-free alignments of 1-antitrypsinFab4B12 subunits into each map in the absence of a linker region and reported as ccoptimal, denoted by red shaded models in the bottom panels of Fig. 5.

The result of this analysis is shown in Fig. 5 (top, color-coded by hypothesis). The random rotational perturbations applied to each model resulted in a spread of preminimization CCC values along the horizontal axis, and minimization of these models generally showed a convergence over a narrow range of CCC values on the vertical axis. The minimized structure giving the highest ccrefined/ccoptimal score for each polymer configuration (in rows) with respect to each map (in columns) is shown in Fig. 5 (bottom). By this analysis, the best-scoring C-terminal polymers (H4) exhibited a value close to one, indicating that the linkage-restrained models were essentially indistinguishable from the unrestrained ones, and this was reflected by an almost direct superimposition of the model over the aligned linker-free subunits (top row). In contrast, the translational and rotational restrictions imposed by the linkers of the other models (H13) prevented them, to varying degrees, from adopting the preferred orientation inherent with respect to the data (bottom three rows).

All models entail a connection between strand 4A of one 1-antitrypsin subunit and strand 1C of the next. A distinguishing characteristic of hypotheses H13, with respect to the C-terminal model (H4), is that they involve a second unique intermolecular linkage. Having dual intermolecular constraints might be expected to reduce conformational flexibility, and this may contribute to their lesser compatibility with the density. To explore this, we performed a variation on the experiment in Fig. 5 in which the dual-linkage models were converted to single linkage by breaking the peptide bond between residues 358 and 359 of the strand 4A-1C connection, leaving the unique second linker that each model embodies intact. During iterative rounds of optimization, displacement between residues adjacent to the site of cleavage confirmed that this modification allowed additional freedom of movement of the subunits. At the conclusion of the experiment, the scores obtained were very similar to those obtained with the intact models (fig. S4, top). We also performed the converse experiment, in which the strand 4A-1C connection was kept intact, and the second unique linker of each model was broken (between residues 344 and 345 for H12 and 324 and 325 for H3). This provided comparable results to the single-linkage C-terminal model (H4) (fig. S4, middle).

These results demonstrate that the head-to-tail orientation of 1-antitrypsin subunits, with the base of sheet A and the top of sheet C in proximity to one another, is an intrinsic feature of the dimer density. Therefore, for the dual-linker models, it is not the reduced flexibility that distinguishes them but the inconsistency of their second linkage with this subunit orientation.

Thus, the orientation provided by the C-terminal model is most compatible with the Dim60 and Dim90 structures present in liver-derived polymers. In the final structure, there are translations of 71 and 73 between the centers of mass of the 1-antitrypsin molecules and a final calculated rotation around the dimer axis of 65 and 81, respectively (Fig. 6A, top and middle). The same analysis, performed using the Dim60H model derived from the heat-induced dataset, gave the same conclusion: The C-terminal model (H4) provided a fit consistent with the model-free aligned subunits (Fig. 6A, bottom). While there was a relative improvement in the fit of the loop-sheet 4 dimer, this model remained unable to adopt an optimal alignment to the experimental data (Fig. 5, right, and fig. S4, top right).

(A) Best-fitting C-terminal model (H4) displayed against the Dim60 (top), Dim90 (middle), and Dim60H (bottom) density, annotated with intersubunit translations and rotations. Dashed lines represent vectors passing through the centers of mass of the 1-antitrypsin and Fab molecules. (B) Electrophoretic mobility shift assay comparing the affinity of the polymer-specific mAb2C1 for polymers of different origin. Binding of the antibody results in a cathodal shift of 1-antitrypsin polymers. Arrows highlight that cleavage-induced polymers, which are structurally analogous to C-terminal polymers, are readily recognized by the antibody with respect to denaturant-induced polymers. A schematic representation of P9-cleavageinduced polymers is shown at the left, with the domain-swapped peptide in black, based on PDB 1D5S (21). (C) Results of sandwich ELISA experiments showing the relative affinity of mAb2C1 for liver-derived, cleavage-induced, and denaturant-induced polymers, normalized to the half-maximal effective concentration (EC50) of the interaction with heat-induced polymers. The affinity of monomeric M and Z, denoted by open circles, was outside the maximum antigen concentration used in the experiment and, correspondingly, not less than two orders of magnitude worse than that of heat-induced polymers. Independent experiments are denoted by the markers, and the means SD are indicated by the bars (liver-derived and denaturant-induced, n = 3; cleaved, n = 6); heat-induced by definition is 1, represented by the dotted line; w.r.t, with respect to.

A neoepitope is recognized by the mAb2C1 antibody that is present in liver-derived and heat-induced polymers but not in those induced in the presence of a denaturant. Thus, the latter conditions produce a polymer structure not representative of pathological material (14, 16). Cleavage of the RCL of 1-antitrypsin in a noncognate position can also induce polymerization (3), and the atomic details of the resulting polymer linkage, defined by crystallography (21, 22), show that it produces a molecule that mimics a noncircular form of the C-terminal trimer (8). To determine whether mAb2C1 recognizes the open C-terminal configuration identified from the EM analyses, polymers mimicking this structure were produced by limited proteolysis of a recombinant Ala350Arg 1-antitrypsin mutant by thrombin. This material was readily recognized by mAb2C1 as demonstrated in a mobility shift experiment (Fig. 6B). The relative affinity of mAb2C1 for the different forms was then determined by enzyme-linked immunosorbent assay (ELISA). These experiments exhibited comparable recognition of liver, heat-induced, and C-terminalmimicking cleaved polymers by the antibody, with a markedly lower affinity for denaturant-induced polymers and monomer (Fig. 6C).

1-Antitrypsin deficiency is characterized by the accumulation of mutant protein as inclusions within hepatocytes. Extraction and disruption of these inclusions release chains of unbranched polymers, which, when isolated, exhibit pronounced flexibility and apparently lack higher-order interactions. Several models have been proposed for the molecular basis of the formation and properties of these polymers from in vitro experiments. On the basis of the observation that polymers are extremely stable and that artificially induced polymerization can be prevented by peptide mimics of the RCL, the first proposed loop-sheet molecular mechanism posited that the RCL of one molecule would incorporate into a sheet of the adjacent molecule (H1 and H2 in Figs. 1A and 5) (1). Since that time, while biophysical studies have attempted to address the question of mechanism, the only crystal structures that have been obtained of 1-antitrypsin oligomers are of forms produced artificially from recombinant nonglycosylated material: a chain of molecules spontaneously assembled following fortuitous cleavage by a contaminating protease (21, 22) and a circular trimer of a disulfide mutant produced by heating (H4) (8). Hence, there has been no direct evidence of the structure of the pathological polymers that deposit in the livers of patients with 1-antitrypsin deficiency.

The in vivo mechanism of 1-antitrypsin polymerization and accumulation in the liver has important consequences for the development of therapeutics that interfere with this process. The loop-sheet hypothesis (H1 and H2) involves relatively minor and reversible perturbations with respect to the native conformation to adopt a polymerization-prone state (1), the C-terminal model (H4) predicates a preceding substantial and irreversible conformational change (8), and the -hairpin model (H3) lies somewhere between the two (7). This has implications for the nature of the site and mode of ligand binding capable of blocking polymerization and, indeed, for the question as to whether the process can be reversed at all.

Polymer material obtained from liver tissue is heterogeneous in size, glycosylated, and difficult to obtain in substantial quantity, making it unsuitable for crystallography. Without the requirement to form a crystal lattice, single-particle reconstruction using EM images represents an excellent option to obtain structural information. The negative-stain approach used here for the analysis of small protein complexes provided a strong contrast between protein and background and, in conjunction with decoration by Fab moieties, made angular information easier to retrieve, revealing the interactions between the components of the flexible polymer chains present in explant liver tissue.

Interrogation of the extant models of polymerization revealed that the loop-sheet dimer model (H1), despite its general compatibility with many biophysical observations, was unable to adopt the intersubunit translation or rotation observed in the 2D and 3D data (Figs. 4 and 5). A less stringent test of this model, a four-residue insertion loop-sheet configuration (H2) with an interchain interface analogous to one binding site of a tetrameric peptide blocker of polymerization (23), still provided an incomplete fit to the data. The -hairpin domain swap model (H3), based on the structure of a self-terminating dimer of antithrombin, has been proposed to extend to 1-antitrypsin polymerization by limited proteolysis and the stability of a disulfide mutant against polymerization (7), a conclusion that has been questioned (16, 24) and not supported by peptide fragment folding data (25). Owing to its longer predicted linking regions, the fit to the Dim60 and Dim90 data was better than that seen with the loop-sheet models (Fig. 5), but it required 20 residues to lose their native structure with respect to the antithrombin crystal structure from which this model is derived. While the crystal structure unequivocally demonstrates the ability of this form to adopt a 180 inversion orthogonal to the dimer axis, there was no evidence in the micrographseither Fab-bound or unboundof a chain inversion of this magnitude.

In contrast, the NS-EM data were best explained by the location, length, and flexibility of the C-terminal linkage (H4). The C-terminal mechanism involves displacement (or delayed formation) of the C-terminal 4-kDa fragment of 1-antitrypsin comprising strands 1C, 4B, and 5B (fig. S1) and self-insertion of the RCL, which results in a monomeric latent-like intermediate conformation (8). The open, nonself-terminating arrangement of the subunits (Fig. 6A) contrasts with the observation that oligomeric components of recombinant material purified from P. pastoris were circular (11).

The data obtained, including the intersubunit orientation and distance (Figs. 3, F and G, 5, and 6A) and the presence of the mAb2C1 epitope (Fig. 6B), support a structural equivalence of heat-induced and liver-derived polymers. Hence, it follows that there will be components shared between their respective polymerization pathways; it should accordingly be possible to extend mechanistic observations made in vitro to the mechanisms that produce polymers in vivo, and here, we draw on observations made in the literature regarding the role of strands 5A, 1C, 4B, and 5B and the breach region (Fig. 7). The ability to induce polymers from folded native 1-antitrypsin by displacement of the C-terminal region at modestly elevated temperatures in the Z variant implies that core packing interactions are readily destabilized when the molecule is in a five-stranded sheet A configuration. In the native conformation (Fig. 7, i), the Z variant has been noted to increase the mobility of strand 5A (26) and the solvation and rotational freedom (27) of the solvent-accessible (28) Trp194 residue that is situated in the breach region (Fig. 7, ii, bottom). The breach is bounded by a hydrophobic cluster of residues including some contributed by strands 5A as well as C-terminal 4B and 5B, on which solvation (as reported by Trp194) would be expected to exert destabilizing effects. This is supported by sequential polypeptide folding experiments, suggesting that engagement of ~36 residues at the C terminus is predicated on a properly formed strand 5A (25). A related process likely occurs on the opposing side of the molecule: Helices A, G, and H form a trihelix clamp over this region, and disruption of stabilizing interactions by the S (Glu264Val) and I (Arg39Cys) mutations (Fig. 7, ii, top) also leads to an increased tendency to polymerize upon the application of heat. Moreover, the fact that S, I, and Z are able to copolymerize (29, 30) indicates that this occurs by a common mechanism and supports the mutual destabilization of the C-terminal region that is situated between them (Fig. 7, iii). This process is consistent with the site of polymerization-prone latch mutations clustered near the end of the polypeptide chain (31).

From the native state (i), the evidence suggests that during heating, decreased affinity for the C terminus can be induced by destabilization of the adjacent breach region with increased solvation of the hydrophobic core (26, 27), destabilization in the adjacent trihelix region (as in the S and I variants), and associated loss of strand 1C native interactions (ii and iii) (6, 24, 32). Upon dissociation of the C terminus, the molecule is equivalent to a final stage of folding of the nascent polypeptide chain (iv) (25). This (reversible) displacement is unable to immediately lead to self-insertion and generate the hyper-stable six-stranded sheet A (25) despite delayed folding (34) (v), but such a change is able to proceed rapidly and irreversibly upon incorporation of the C terminus of another molecule (vi) (25, 33). Under appropriate conditions, the latent conformation is generated as an off-pathway species (vii) that is expected to be inaccessible once full RCL insertion has taken place (v) (17, 36). Asterisks denote Trp194 (blue) and Glu264/Arg39 (red), regions colored as black and yellow arrows highlight structural changes, and symbols indicate the application of heat (triangle) or a hypothesized point of convergence with the nascent chain folding pathway (R).

The early (6) and necessary (24, 32) loss of native strand 1C contacts is consistent with the displacement of the C-terminal region (Fig. 7, iv). In this state, current evidence indicates that the molecule is equivalent to a final stage of the folding pathway (25). While the displaced C terminus (Fig. 7, iv) is relatively hydrophobic, in isolation, the equivalent C36 peptide has been found to be soluble, albeit fibrillogenic over a period of hours, and readily incorporated into native 1-antitrypsin at room temperature, inducing an increase in thermal stability consistent with transition to a self-inserted form (33). This suggests that displacement of this region even at ambient temperature is possible. While, by analogy with release of the RCL by proteolytic cleavage (fig. S1), it might be expected that the release of the C terminus would immediately give rise to self-insertion of the untethered RCL as strand 4A, there is evidence that the absence of an engaged C terminus will prevent this from occurring (25). This is congruent with the preferential folding of the protein to the kinetically stabilized five-stranded sheet A conformation rather than the loop-inserted six-stranded thermodynamically favored state (25) despite the adoption of the hyperstable form upon administration of exogenous C-terminal peptide (33) and the fact that some material does fold correctly to the active form even with the delayed folding of the Z variant (34).

Upon incorporation of the C terminus of another molecule (Fig. 7, v), self-insertion of strand 4A would be expected to follow (Fig. 7, vi) (33). The RCL of 1-antitrypsin is shorter than those of serpins known to undergo latency as a competing process to polymerization (35); once insertion has proceeded beyond a molecular decision point near the center of sheet A (17, 36), the molecule would no longer be able to (re-) incorporate its own C-terminal fragment (Fig. 7, vii), and it would effectively become irreversibly activated for oligomerization (Fig. 7, v). This mechanism is consistent with the suppression of polymerization in cells by a single-chain antibody fragment that alters the behavior of sheet A in the vicinity of the helix F (12, 13) and mutations that inhibit loop self-insertion (17).

Thus, of the proposed polymerization linkage models, our data most strongly support the C-terminal domain swap as the structural basis for pathological polymers of Z 1-antitrypsin. It remains to be determined how common or rare the exceptions are to this mechanism among other members of the serpin family. Serpins share a highly conserved core structure and exhibit common folding behaviors, and mutations that are associated with instability and deficiency tend to cluster within defined structural regions (37, 38). These factors likely place constraints on the mechanism by which mutations can induce polymerization. It is difficult to overlook the central role of the C terminus in both latency and the C-terminal domain swap, with the former essentially a monomeric self-terminating form of the latter (Fig. 7, v to vii). While a shorter RCL likely renders these two states mutually exclusive in 1-antitrypsin, it has been suggested that the greater tendency of plasminogen activator inhibitor-1 (PAI-1) to adopt the latent conformation is due to a common origin in the polymerogenic intermediate (35). In support of this, PAI-1 and the neuroserpin L49P variant can form polymers from the latent state (35, 39), a notable observation given the high stability of this conformation and inconsistent with the loop-sheet polymerization mechanism (which is predicated on a five-stranded native-like molecule) and the intermolecular strand 5A/4A linkage of the -hairpin model.

On the other hand, it has been shown that distinct alternative polymerization pathways are accessible in vitro depending on the nature of the destabilizing conditions used. The crystal structure of a -hairpinswapped self-terminating dimer of antithrombin (7) produced by incubation of this protein in vitro at low pH provides evidence of this. Similarly, induction of polymerization at acidic pH or with denaturants causes 1-antitrypsin to adopt a polymer form inconsistent with that seen upon heating or with pathological specimens from ZZ homozygotes (16). Biochemical evidence indicates that this may reflect the conformation of the rare 1-antitrypsin Trento variant (14).

From the data presented, here we expect the C-terminal domain swap to reflect the basis of pathological polymers in carriers of the Z 1-antitrypsin alleleand by extension, the S and I variantsand therefore account for more than 95% of cases of severe 1-antitrypsin deficiency. Because of its intimate association with the folding pathway and relationship with the latent structure more readily adopted by other serpins, it is probable that this form will be relevant to other serpin pathologies. Whether the same linkage underlies the shutter region mutants of 1-antitrypsin [such as Siiyama, Mmalton, and Kings (2, 10)] that also cause polymer formation and severe plasma deficiency remains to be determined.

Human M and Z 1-antitrypsin were purified from donor plasma, and recombinant 1-antitrypsin was purified from Escherichia coli as previously described (24, 40). Monoclonal antibodies were purified from hybridomas according to published methods (12) and stored in phosphate-buffered saline (PBS) with 0.02% (w/v) sodium azide. Fab fragments were generated by limited proteolysis using ficin or papain as appropriate with commercial kits according to the manufacturers instructions (Thermo Fisher Scientific) with the subsequent addition of 1 mM E-64 inhibitor.

Explanted liver tissue (5 to 10 g) from individuals homozygous for the Z allele was homogenized and incubated at 37C for 1 hour in 10 ml of Hanks modified balanced salt solution with 5 mg of Clostridium histolyticum collagenase, and fibrous tissue was removed from the resultant suspension by filtration through BioPrepNylon synthetic cheesecloth with a 50-m pore size (Biodesign). The filtrate was centrifuged at 3000g at 4C for 15 min, the pellet was resuspended in 3 ml of 0.25 M sucrose in buffer E [5 mM EDTA, 50 mM NaCl, and 50 mM tris (pH 7.4)], and the sample was layered onto the top of two 14-ml centrifuge tubes (Beckman Coulter) containing a preformed 0.3 to 1.3 M sucrose gradient in buffer E and centrifuged at 25,000g for 2 hours at 4C. The supernatant was discarded, and the pelleted inclusion bodies were washed with buffer E. Previous approaches to polymer extraction (41) have made use of detergents and denaturants, compounds that have been shown, under certain conditions, to induce conformational change in 1-antitrypsin (3), and therefore, we omitted their use. Soluble polymers were extracted by sonication on ice using a SoniPrep 150 with a nominal amplitude of 2.5 m (giving a probe displacement of 17.5 m) in bursts of 15 s and 15-s rest for a total of 6 min. The solution was repeatedly centrifuged for 5 min at 13,000g in a benchtop centrifuge to remove insoluble material. Purity of the soluble component was assessed by SDS- and nondenaturing PAGE.

For heat-induced polymers, purified plasma M 1-antitrypsin was buffer-exchanged into PBS to 0.2 mg/ml and polymerization induced by heating at 55C for 48 hours. Denaturant-induced polymers were formed by incubation at 0.4 mg/ml and 25C for 48 hours in 3 M guanidine hydrochloride and 40 mM tris-HCl (pH 8) buffer. Following dialysis, anion exchange chromatography using a HiTrap Q Sepharose column with a 0 to 0.5 M NaCl gradient in 20 mM tris (pH 8.0) was used to remove residual monomer, as confirmed by native PAGE.

An arginine residue was introduced at the P9 position (residue 350) of 1-antitrypsin in a pQE-30based (Qiagen) expression system (17) using the QuikChange mutagenesis kit according to the manufacturers instructions (Agilent). Following purification from E. coli, the protein was subjected to limited proteolysis by a 50-fold substoichiometric concentration of bovine thrombin (Merck) at 37C overnight and polymer isolated by anion exchange chromatography using a HiTrap Q Sepharose column with a 0 to 0.5 M NaCl gradient in 20 mM tris (pH 8.0).

Polymers were incubated with a threefold molar excess (with respect to subunit concentration) of Fab4B12 (12) for 2.5 hours at room temperature and repurified by anion exchange chromatography as described above or dialyzed overnight at 4C into buffer E using a 300-kDa molecular weight cutoff membrane (Spectrum). Copper grids (300 mesh, Electron Microscopy Services) were covered with a continuous carbon film of thickness ~50 m and glow discharged for 30 s. Three microliters of the prepared sample at ~0.05 to 0.1 mg/ml concentration was applied to the prepared grids for 1 min before blotting. Samples were negatively stained for 1 min using 5 l of 2% (w/v) uranyl acetate and blotted, and the staining step was repeated. For single-frame high-contrast micrographs, grids were visualized using an FEI Tecnai T12 BioTWIN LaB6 microscope operating at 120 keV, and images were recorded on an FEI Eagle 4K 4K charge-coupled device camera under low-dose conditions (25 electrons 2) at an effective magnification of 91,463 (1.64 per pixel) and a defocus range of 0.8 to 3.5 m. Micrographs for single-particle reconstruction were recorded as averages of 30-frame, 30-frames/s movies using a Tecnai F20 field emission gun transmission electron microscope at 200 keV with a Direct Electron DE-20 direct detector at a calibrated 41,470 magnification (1.54 per pixel) under low-dose conditions (~1 electron 2 per frame). Frames were motion-corrected using MotionCorr (42). Resulting images were corrected for the effects of the contrast transfer function of the microscope using CTFFIND3 (43). Micrographs with greater than 5% astigmatism were discarded. Manual particle picking was undertaken using EMAN (44). General processing scripts in Python made use of the EMAN2 (44), NumPy, SciPy, OpenCV, and Matplotlib libraries.

RELION v2.1 and v3.0.6 (15) were used for single-particle reconstruction including automated particle picking, 2D and 3D classification, and 3D refinement, with the final processing path described in detail in Results and fig. S2. In general, classification in RELION used a regularization parameter T = 2 and 25 iterations or 50 iterations where convergence of statistics was not observed to have occurred. Image boxes were 230 230 in size; for 2D processing, a mask diameter of 180 was used, and alignment was performed using an initial 7.5 interval with an offset search range of five pixels; for 3D processing, the mask diameter was 195 with a sampling of 15 and eight pixels; and 3D refinement used 195 , 7.5, and five pixels, respectively. Masks were generated for 3D dimer references by contouring at ~3.8 105 3 (or at noise), for monomer references at ~1.9 105 3, and both with the addition of a 7-voxel/7-voxel hard and soft edge. A 30- low-pass filter was applied to the resulting masked volumes before classification or refinement. After obtaining the Dim60 and Dim90 structures, the subsets of particle images on which they were based were subjected to a reference-free stochastic gradient-driven de novo reconstruction in RELION (sampling 15 and two-pixel increments; 50 initial, 200 in-between, and 50 final iterations from 40 down to 20 ). An equivalent model was returned in each case. Similarly, combining the two particle sets together and performing a 3D reclassification using the monomeric Monav reference (fig. S2o, left) effectively returned the same two models.

Proteins were resolved under denaturing conditions by NuPAGE 4 to 12% (w/v) acrylamide bis-tris SDS-PAGE gels and under nondenaturing conditions using NativePAGE 3 to 12% (w/v) acrylamide bis-tris gels (Thermo Fisher Scientific). For visualization by Coomassie dye, typical loading was 1 to 4 and 0.1 to 0.4 g for Western blot. Western blot transfer to a polyvinylidene difluoride membrane was undertaken using the iBlot system (Thermo Fisher Scientific) or by wet transfer (Bio-Rad), followed by these steps: soaking in PBS for 10 min; blocking for 1 hour at room temperature with 5% (w/v) nonfat milk powder in PBS; incubation with primary antibody (rabbit polyclonal at 0.8 g/ml or mouse monoclonal at 0.2 g/ml) overnight at 4C in PBS with 0.1% Tween (PBST), 5% (w/v) bovine serum albumin, and 0.1% sodium azide; washing with PBST; incubation with secondary antibodies at 1:5000 to 1:10,000 in PBST with 5% (w/v) bovine serum albumin and 0.1% sodium azide; and development by Pierce enhanced chemiluminesence (Thermo Fisher Scientific) or fluorescence (LiCor).

High-binding enzyme immunoassay microplates (Sigma-Aldrich) were coated with 50 l per well of anti-polymer mAb2C1 (2 g/ml) in PBS with incubation overnight at room temperature, washed once with distilled water and twice with wash buffer [0.9% (w/v) sodium chloride and 0.025% (v/v) Tween 20], and blocked for 1 hour with 300 l per well of PBST buffer [PBS, 0.025% (v/v) Tween 20, and 0.1% (w/v) sodium azide] supplemented with 0.25% (w/v) bovine serum albumin at room temperature (PBSTB). After washing the plates, antigens in PBSTB were applied by 1:1 serial dilution at a final volume of 50 l across the plate, incubated for 2 hours at room temperature, and washed. Fifty microliters of rabbit anti-human 1-antitrypsin polyclonal antibody (1 g/ml) (DAKO) in PBSTB was added to each well, the plates were incubated for 2 hours at room temperature and washed, 50 l of a 1:2000 dilution of goat anti-rabbit horseradish peroxidase antibody in PBSTB (without sodium azide) was added to each well, and the plates were incubated in the dark for 75 min at room temperature and then washed again. For detection, 3,3,5,5-tetramethylbenzidine substrate solution (Sigma-Aldrich) was added at 50 l per well, the plates were incubated for ~7 min in the dark, the reaction stopped by adding 50 l per well of 1 M H2SO4, and the absorbance was promptly measured at 450 nm in a SpectraMax M5 plate reader (Molecular Devices).

For crystallization trials, protein was buffer-exchanged into buffer C [10 mM tris (pH 7.4), 50 mM NaCl, and 0.02% (w/v) sodium azide] and concentrated to 10 mg/ml. Broad-screen sitting drop approaches against commercially available buffer formulations (Molecular Dimensions and Hampton Research) were performed with 100-nl protein:100-nl buffer drops dispensed using a Mosquito robot (TTP LabTech) and equilibrated against 75 l of buffer at 16C with automatic image acquisition by a CrystalMation system (Rigaku). Hanging-drop screens were performed at 20C with 1 l of protein:1 l of buffer equilibrated against 250 l of buffer. Crystals mounted on nylon loops were briefly soaked in the respective crystallization buffer supplemented by 10% (v/v) glycerol ethoxylate or 10% (v/v) ethylene glycol before plunge-freezing into liquid nitrogen. Data collection was undertaken at the European Synchrotron Radiation Facility (ESRF) ID30B beamline (with enabling work at the Diamond I03 beamline). Data reduction, integration, scaling, and merging were performed using autoPROC (45); the structures were solved by molecular replacement using Phaser (46); model refinement was undertaken with PHENIX (47); and model visualization and building were performed with Coot (48).

Recombinant 1-antitrypsin was incubated at a substoichiometric ratio to Fab4B12 for an hour at room temperature, and excess Fab was removed by anion exchange as described above. After concentration of the complex to 10 mg/ml, 50 l was applied to a Superdex 200 Increase 5/150 column (GE Life Sciences) at a rate of 0.3 ml/min in 30 mM NaCl and 50 mM tris (pH 7.4) buffer at the P12 BioSAXS beamline, European Molecular Biology Laboratory (EMBL) Hamburg (49). The x-ray scatter ( = 1.24 ) was recorded on a Pilatus 6M detector at 1 frame/s. The buffer baseline-corrected scatter profile was produced by integration over time points corresponding with elution of the complex from the size exclusion column using the ATSAS software package (50).

For initial working subunit and dimer models, Coot (48) and PyMOL (Schrdinger Software) were used to position crystal structures of 1-antitrypsin [PDB: cleaved, 1EZX (19); cleaved polymer, 1D5S (21)] or mAb4B12 (PDB: 6QU9) and modify chain boundaries, repair gaps, and improve stereochemistry of intermolecular segments. The initial -hairpin and loop-sheet models (Fig. 1A, H13) were further optimized in PyRosetta (Fig. 1A) (20). Superposition of the model of the 1-antitrypsinFab4B12 complex onto the dimer was undertaken using PyMOL. Modifications had to be made to each model to reconcile observations made here and in recent studies:

H1 and H2. Loop-sheet models have been represented with various degrees of insertion of the donor RCL into the site of strand 4A in the acceptor molecule. To explore the compatibility of this parameter with the flexibility and periodicity of the polymers visualized here, two forms were generated, one with a substantial eight-residue insertion (loop-sheet 8, H1) and one with a marginal interaction at the base of sheet A based on the observation that tetrameric peptides are able to block polymerization and induce stabilization of 1-antitrypsin (loop-sheet 4, H2) (18, 23). The loop insertion site is permissive of noncognate peptide residues; however, such out-of-register insertion has not been observed crystallographically for intra- or interprotein loop insertion. For the arrangements used here, inserted residues were maintained in register at their cognate positions as observed for the structures of the cleaved protein, cleavage-induced polymer (21), and the self-terminating dimer (7) and trimer (8).

H3. The hypothesized unwinding of helix I in the -hairpin polymer has been challenged (16) and is inconsistent with the role of this element in the 4B12 epitope (13). The ability of Fab4B12 to bind to the ex vivo polymers is unequivocal from the images recorded here; thus, if the pathological polymer is reflected by the -hairpin model, then helix I must remain intact.

H4. Contrary to a proposal that circular polymers are the predominant species (8, 11), most of those extracted from liver tissue were observed to be linear. Accordingly, the C-terminal dimer was arranged in an open configuration through redefinition of the chain boundaries in the crystal structure of a cleavage-generated polymer (21).

During optimization of Fab-bound 1-antitrypsin dimer models, the constituent subunits were treated as rigid bodies connected by flexible linker regions. As much intersubunit linker flexibility was allowed as possible while maintaining the integrity of the core 1-antitrypsin fold, consistent with serpin monomer and oligomer crystal structures and with the high stability of the polymer. Divergence from the canonical structure was permitted where this accorded with the characteristics of the model being tested and other experimental data. Specifically:

1) Although crystal structures of cleaved antitrypsin polymers (21, 22), an antithrombin dimer (7), and antitrypsin trimer (8) all have an intact strand 1C, it has been shown that during the process of (heat-induced) polymerization, this element is labile (24, 32). Accordingly, we allowed the residues of this element (362 to 368) to move in all models.

2) All models of polymerization, either structurally defined or modeled, propose a connection between the C terminus of strand 4A and the N terminus of strand 1C (residues 357 to 362). The evidence is that this is a region that lacks secondary structure: In the cleaved form, it is not part of strand 4A or strand 1C; in the native structure, it does not form polar contacts with the body of the molecule; and it forms an extended chain in the latent conformation (36). Thus, this was treated as a flexible region.

3) The -hairpin model (H3) involves a connection between helix I of the donor subunit and strand 5A of the acceptor. Limited proteolysis data were interpreted to support the unraveling of helix I in this polymer linkage, yet this is not a feature observed in the crystal structure of the antithrombin dimer on which the model is based (7), and this conclusion has been disputed (16). If the -hairpin model is indeed representative of the polymers considered in this study, then helix I should be intact as it is integral to the epitope of the nonconformation-selective Fab4B12 that decorates them (13). Hence, the region 309 to 328 between helix I and strand 5A was provided with full flexibility, which maintains the integrity of elements seen in the original crystal structure but allows all other linker residues to move.

4) All crystal structures exhibit an intact strand 5A, and while there is evidence of some lability of this structural element in the native conformation of a Z-like Glu342Ala mutant, this is not shared by the wild-type protein (26). For the loop-sheet models (H12) that propose connections between strand 5A of the donor subunit and strand 4A embedded in the acceptor, all connecting residues between residues 340 to 348 (H1) and 340 to 352 (H2) were provided full torsional freedom during refinement.

The selection of polymers was performed manually by visual inspection of micrographs, followed by automatic thresholding and excision of regions of interest from the individual polymer images. Where a region of interest contained more than one chain, the image was postprocessed to remove density not related to the polymer of interest. Starting models of each polymer configuration at an appropriate length were generated by permutation of a seed dimer structure according to the number of subunits in an oligomer. The PyRosetta application programming interface (20) was then used, in which the 1-antitrypsinFab4B12 subunits were treated as rigid bodies connected by flexible linker regions; a full-backbone centroid model was used in which each side chain was represented by a single pseudoatom. Following an initial rigid-body step to approximately align the model with the image, loose positional constraints were applied to subunits according to the polymer path determined during the manual selections from the micrographs. Angular relationships with respect to the underlying substrate plane were inferred according to the extent of the orthogonal Fab protrusion observable, from 90 (evidence of increased density along the z axis only) to 0 (full-length protrusion in the XY plane). A necessary simplification, resulting in an implicit minimization of the magnitude of the angular displacement between subunits, was that these would tend to orient away from the underlying carbon substrate. Refinement of these models used an energy term that sought to increase the correlation between the experimental reference image and a 2D projection of the target 3D molecule. Standard stereochemical, repulsive, and attractive terms, and loose positional restraints, were maintained throughout. Iterative refinement proceeded for a minimum of 10 steps of 25 iterations, following which convergence was deemed to have occurred when the root mean square deviation between prerefined and postrefined model was less than 0.05 . The score for a given model-oligomer pair was calculated as the ratio of the best correlation coefficient observed during the optimization of the model against the oligomer relative to the best score observed for any model against that oligomer image.

For each dimer configurationloop-sheet 8 (H1) or 4 (H2), -hairpin (H3), and C-terminal (H4)repeated (1000) rounds of optimization were undertaken from a starting model randomly perturbed by rotation around the dimer axis. Full-atom models were represented as rigid subunits connected by flexible linkers. Optimization (using PyRosetta) involved an alternating sequence of whole-dimer rigid body shift and torsional optimization into the experimental density. The scoring scheme used to steer the process involved default internal stereochemical, attractive, and repulsive terms as well as the correlation of the atomic configuration with the EM density, with relative weighting of these terms progressively adjusted during the iterative procedure. To avoid any contribution of the linker regions to the scores obtained, only the rigid core subunits were used in the calculation of the correlation coefficient with respect to the electron density. The van der Waals scoring term was monitored to exclude models where unresolvable clashes occurred. Structures were visualized using Chimera (51) and PyMOL (Schrdinger Software).

Statistical analyses were performed using Prism 6 software (GraphPad, La Jolla, CA, USA). The significance of the difference in correlation between the 2D projections of the different polymer models and the polymer images in Fig. 4 was determined by a one-way analysis of variance (ANOVA) and Tukeys multiple comparisons test; ***P < 0.001 and ****P < 0.0001. Mean values are reported throughout the text with SD or SEM, as indicated.

Tissue was used with the informed consent of donors and in accordance with local Institutional Review Boards.

Acknowledgments: We are indebted to M. Carroni (now at SciLifeLab) for collection of EM micrographs and training, and we would like to thank N. Lukoyanova and S. Chen at the ISMB Birkbeck EM Laboratory for support, training, and facility access (as well as D. Clare and Y. Chaban, now at eBIC, for antecedent enabling work) and N. Pinotis at the ISMB X-Ray Crystallography Laboratory for logistical support and facility access. We acknowledge the ESRF (Grenoble) for provision of synchrotron radiation facilities, and we would like to thank G. Leonard for assistance in using beamline ID30B; enabling work was performed on beamline I03 at the Diamond Light Source (proposal mx17201), and we would like to thank the staff for facility provision and technical support. The synchrotron SAXS data were collected at beamline P12 operated by EMBL Hamburg at the PETRA III storage ring (DESY, Hamburg, Germany), and we would like to thank M. Graewert and D. Franke for assistance. We acknowledge the contribution to this publication made by the University of Birminghams Human Biomaterials Resource Centre, which has been supported through the Birmingham Science CityExperimental Medicine Network of Excellence project. We acknowledge the use of the UCL Grace High Performance Computing Facility (Grace@UCL) and the UCL Legion High Performance Computing Facility (Legion@UCL), and associated support services, in the completion of this work. Funding: This work was supported by a grant from the Medical Research Council (UK) to D.A.L. (MR/N024842/1, also supporting J.A.I. as RCo-I and B.G. as Co-I) and the NIHR UCLH Biomedical Research Centre. D.A.L. is an NIHR Senior Investigator. S.V.F. was the recipient of an EPSRC/GSK CASE studentship. E.L.K.E. was the recipient of a Wellcome Trust Biomedical Research Studentship to the ISMB. B.G. was supported for this work by a Wellcome Trust Intermediate Clinical Fellowship and is currently supported by the NIHR Leicester Biomedical Research Centre. This work was funded, in part, by an Alpha-1 Foundation grant to J.A.I. The equipment used at the ISMB/Birkbeck EM Laboratory was funded by the Wellcome Trust (grants 101488 and 058736). Author contributions: S.V.F., E.L.K.E., A.R., and B.G. collected EM data. J.A.I., E.L.K.E., S.V.F., B.G., E.V.O., and M.B. analyzed EM data. A.M.J. and J.A.I. collected and analyzed crystallography data. A.M.J. and J.A.I. collected and analyzed SAXS data. E.L.K.E., S.V.F., I.A., and J.A.I. collected and analyzed biochemical data. J.A.I. performed modeling and wrote the computer code. S.V.F., E.L.K.E., N.H.-C., A.M.J., I.A., A.R., E.M., and J.A.I. prepared reagents. S.T.R., G.M.R., and D.H.A. provided reagents. E.M. provided advice and training. J.A.I., E.V.O., B.G., and A.R. supervised data collection and analysis. J.A.I., D.A.L., and E.V.O. supervised the project. J.A.I., S.V.F., E.L.K.E., and D.A.L. drafted the manuscript. All authors contributed to and approved the final manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. The Dim60, Dim90, and Dim60H maps have been deposited in the EMDB with accessions EMD-4632, EMD-4631, and EMD-4620. The crystal structure of Fab4B12 has been deposited as PDB accession 6QU9. Additional data related to this paper may be requested from the authors.

View original post here:
The structural basis for Z 1-antitrypsin polymerization in the liver - Science Advances

RNA folding may help explain how the coronavirus became so hard to stop after it spilled over from wildlife to humans.

We know that the coronavirus behind the COVID-19 crisis lived harmlessly in bats and other wildlife before it jumped the species barrier and spilled over to humans.

Now, researchers at Duke University have identified a number of silent mutations in the roughly 30,000 letters of the viruss genetic code that helped it thrive once it made the leap and possibly helped set the stage for the global pandemic. The subtle changes involved how the virus folded its RNA molecules within human cells.

For the study, published October 16, 2020, in the journal PeerJ, the researchers used statistical methods they developed to identify adaptive changes that arose in the SARS-CoV-2 genome in humans, but not in closely related coronaviruses found in bats and pangolins.

Were trying to figure out what made this virus so unique, said lead author Alejandro Berrio, a postdoctoral associate in biologist Greg Wrays lab at Duke.

Previous research detected fingerprints of positive selection within a gene that encodes the spike proteins studding the coronaviruss surface, which play a key role in its ability to infect new cells.

The new study likewise flagged mutations that altered the spike proteins, suggesting that viral strains carrying these mutations were more likely to thrive. But with their approach, study authors Berrio, Wray and Duke Ph.D. student Valerie Gartner also identified additional culprits that previous studies failed to detect.

The researchers report that so-called silent mutations in two other regions of the SARS-CoV-2 genome, dubbed Nsp4 and Nsp16, appear to have given the virus a biological edge over previous strains without altering the proteins they encode.

Instead of affecting proteins, Berrio said, the changes likely affected how the viruss genetic material which is made of RNA folds up into 3-D shapes and functions inside human cells.

What these changes in RNA structure might have done to set the SARS-CoV-2 virus in humans apart from other coronaviruses is still unknown, Berrio said. But they may have contributed to the viruss ability to spread before people even know they have it a crucial difference that made the current situation so much more difficult to control than the SARS coronavirus outbreak of 2003.

The research could lead to new molecular targets for treating or preventing COVID-19, Berrio said.

Nsp4 and Nsp16 are among the first RNA molecules that are produced when the virus infects a new person, Berrio said. The spike protein doesnt get expressed until later. So they could make a better therapeutic target because they appear earlier in the viral life cycle.

More generally, by pinpointing the genetic changes that enabled the new coronavirus to thrive in human hosts, scientists hope to better predict future zoonotic disease outbreaks before they happen.

Viruses are constantly mutating and evolving, Berrio said. So its possible that a new strain of coronavirus capable of infecting other animals may come along that also has the potential to spread to people, like SARS-CoV-2 did. Well need to be able to recognize it and make efforts to contain it early.

Reference: Positive selection within the genomes of SARS-CoV-2 and other Coronaviruses independent of impact on protein function by Alejandro Berrio1, Valerie Gartner and Gregory A. Wray, 16 October 2020, PeerJ.DOI: 10.7717/peerj.10234

Read the original post:
Silent Mutations Identified That Give the COVID-19 Coronavirus an Evolutionary Edge - SciTechDaily

Chemists studying how life started often focus on how modern biopolymers like peptides and nucleic acids contributed, but modern biopolymers don't form easily without help from living organisms. A possible solution to this paradox is that life started using different components, and many non-biological chemicals were likely abundant in the environment. A new survey conducted by an international team of chemists from the Earth-Life Science Institute (ELSI) at Tokyo Institute of Technology and other institutes from Malaysia, the Czech Republic, the US and India, has found that a diverse set of such compounds easily form polymers under primitive environmental conditions, and some even spontaneously form cell-like structures.

Understanding how life started on Earth is one of the most challenging questions modern science attempts to explain. Scientists presently study modern organisms and try to see what aspects of their biochemistry are universal, and thus were probably present in the organisms from which they descended. The best guess is that life has thrived on Earth for at least 3.5 billion of Earth's 4.5 billion year history since the planet formed, and most scientists would say life likely began before there is good evidence for its existence. Problematically, since Earth's surface is dynamic, the earliest traces of life on Earth have not been preserved in the geological record. However, the earliest evidence for life on Earth tells us little about what the earliest organisms were made of, or what was going on inside their cells. "There is clearly a lot left to learn from prebiotic chemistry about how life may have arisen," says the study's co-author Jim Cleaves.

A hallmark of life is evolution, and the mechanisms of evolution suggest that common traits can suddenly be displaced by rare and novel mutations which allow mutant organisms to survive better and proliferate, often replacing previously common organisms very rapidly. Paleontological, ecological and laboratory evidence suggests this occurs commonly and quickly. One example is an invasive organism like the dandelion, which was introduced to the Americas from Europe and is now a commo weed causing lawn-concerned homeowners to spend countless hours of effort and dollars to eradicate. Another less whimsical example is COVID-19, a virus (technically not living, but technically an organism) which was probably confined to a small population of bats for years, but suddenly spread among humans around the world. Organisms which reproduce faster than their competitors, even only slightly faster, quickly send their competitors to what Leon Trotsky termed the "ash heap of history." As most organisms which have ever existed are extinct, co-author Tony Z. Jia suggests that "to understand how modern biology emerged, it is important to study plausible non-biological chemistries or structures not currently present in modern biology which potentially went extinct as life complexified."

This idea of evolutionary replacement is pushed to an extreme when scientists try to understand the origins of life. All modern organisms have a few core commonalities: all life is cellular, life uses DNA as an information storage molecule, and uses DNA to make ribonucleic RNA as an intermediary way to make proteins. Proteins perform most of the catalysis in modern biochemistry, and they are created using a very nearly universal "code" to make them from RNA. How this code came to be is in itself enigmatic, but these deep questions point to their possibly having been a very murky period in early biological evolution ~ 4 billion years ago during which almost none of the molecular features observed in modern biochemistry were present, and few if any of the ones that were present have been carried forward.

Proteins are linear polymers of amino acids. These floppy strings of polymerised amino acids fold into unique three-dimensional shapes, forming extremely efficient catalysts which foster precise chemical reactions. In principle, many types of polymerised molecules could form similar strings and fold to form similar catalytic shapes, and synthetic chemists have already discovered many examples. "The point of this kind of study is finding functional polymers in plausibly prebiotic systems without the assistance of biology, including grad students," says co-author Irena Mamajanov.

Scientists have found many ways to make biological organic compounds without the intervention of biology, and these mechanisms help explain these compounds' presence in samples like carbonaceous meteorites, which are relics of the early solar system, and which scientists don't think ever hosted life. These primordial meteorite samples also contain many other types of molecules which could have formed complex folded polymers like proteins, which could have helped steer primitive chemistry. Proteins, by virtue of their folding and catalysis mediate much of the complex biochemical evolution observed in living systems. The ELSI team reasoned that alternative polymers could have helped this occur before the coding between DNA and protein evolved. "Perhaps we cannot reverse-engineer the origin of life; it may be more productive to try and build it from scratch, and not necessarily using modern biomolecules. There were large reservoirs of non-biological chemicals that existed on the primeval Earth. How they helped in the formation of life-as-we-know-it is what we are interested in," says co-author Kuhan Chandru.

The ELSI team did something simple yet profound: they took a large set of structurally diverse small organic molecules which could plausibly be made by prebiotic processes and tried to see if they could form polymers when evaporated from dilute solution. To their surprise, they found many of the primitive compounds could, though they also found some of them decomposed rapidly. This simple criterion, whether a compound is able to be dried without decomposing, may have been one of the earliest evolutionary selection pressures for primordial molecules.

The team conducted one further simple test. They took these dried reactions, added water and looked at them under a microscope. To their surprise, some of the products of these reaction formed cell-sized compartments. That simple starting materials containing 10 to 20 atoms can be converted to self-organised cell-like aggregates containing millions of atoms provides startling insight into how simple chemistry may have led to complex chemistry bordering on the kind of complexity associated with living systems, while not using modern biochemicals.

"We didn't test every possible compound, but we tested a lot of possible compounds. The diversity of chemical behaviors we found was surprising, and suggests this kind of small-molecule to functional-aggregate behavior is a common feature of organic chemistry, which may make the origin of life a more common phenomenon than previously thought," concludes co-author Niraja Bapat.

See the article here:
Scientists discover new organic compounds that could have helped form the first cells - Science Codex