VA is a PLOS ONE Editorial Board member. MTL is an employee of OBodies Ltd. VA, MB and JDS are named inventors on an international patent application owned by OBodies Ltd (entitled “OB FOLD DOMAINS”; PCT/NZ2007/000125) which covers part of this work. This does not alter the authors’ adherence to all the PLOS ONE policies on sharing data and materials.
Conceived and designed the experiments: MTL JDS VA MB. Performed the experiments: MTL JDS MB. Analyzed the data: MTL JDS MB VA JR. Contributed reagents/materials/analysis tools: JR. Wrote the paper: MTL JDS VA.
The OB-fold is a small, versatile single-domain protein binding module that occurs in all forms of life, where it binds protein, carbohydrate, nucleic acid and small-molecule ligands. We have exploited this natural plasticity to engineer a new class of non-immunoglobulin alternatives to antibodies with unique structural and biophysical characteristics. We present here the engineering of the OB-fold anticodon recognition domain from aspartyl tRNA synthetase taken from the thermophile
Molecular recognition is a crucial aspect of a successful biological system. High-affinity interactions govern immune function, while relatively transient events are seen in signal transduction. Each type of interaction is dependent on the adaptability of protein folds during evolution. We have explored this phenomenon using a widely-distributed protein domain, the OB-fold, originally named for observed functions of oligosaccharide and oligonucleotide binding
The OB-fold is a 5-stranded β-barrel domain that presents a concave binding face. In virtually all cases where the domain is present, this same face is used for binding ligand. A survey of the SCOP database reveals many OB-folds which are heavily modified with additional loops or entirely new domains inserted
Antibodies and antibody fragments are currently the dominant class of engineered proteins for molecular recognition; very large combinatorial libraries (in excess of 1011 individual members) have been made and selected
In principle, OB-folds possess features granting potential benefits to an engineered OB-fold domain over and above antibodies based on the immunoglobulin fold; they are small, single-domain binding modules and generally lack disulphide bonds. Due to their ubiquitous nature, potential OB-fold-based scaffolds may be sourced from widely diverse organisms, including human proteins for therapeutic and
In this study we have designed and tested the properties of a new protein scaffold based on the OB-fold of the aspartyl tRNA synthetase (AspRS) from the thermophile
To generate a naïve library, 17 specific sites on the surface of the domain were selected based on surface accessibility and their interaction with the native tRNA substrate, using a homology model built by the Swiss Model automated server
(a) Model of
Initial solid-phase phage display selections were done over five rounds, using immobilised HEL. Monoclonal phage isolates were screened for binding by phage ELISA, and positive clones were expressed in
NL8 | AM1L10 | AM2EP06 | AM3L15 | AM3L09 | |
Space Group | p41212 | p212121 | p212121 | p21 | p1 |
Cell dimensions | |||||
a, b, c (Å) | 76.759,76.759, 166.344 | 60.54, 186.26, 245.70 | 50.43, 58.33, 81.82 | 51.99, 58.83, 95.18 | 59.00, 69.38, 76.36 |
α, β, γ (°) | 90 | 90 | 90 | 90, 95.43, 90 | 72.17, 69.46, 77.55 |
Wavelength (Å) | 0.95666 | 0.95666 | 1.54179 | 1.54179 | 0.95369 |
Resolution (Å) | 34.9 - 2.69 (2.76 - 2.69) |
29.76 - 1.95 (2.05 - 1.95 | 34.58 - 1.86 (1.96 - 1.86) | 35.09 - 1.86 (1.96 - 1.86) | 51.50 - 2.57 (2.71 - 2.57) |
Measured Reflections | 144772 | 2883973 | 143392 | 130001 | 558386 |
Unique Reflections | 14295 (908) | 201772 (27775) | 20797 (2819) | 46187 (6490) | 32670 (4634) |
Multiplicity | 9 | 14.3 (12.4) | 6.9 (6.5) | 2.8 (2.7) | 7.8 (7.7) |
Completeness (%) | 99.15 (92.8) | 99 (95.0) | 99.4 (94.1) | 95.7 (92.3) | 96.2 (93.6) |
7.3 (54.3) | 6.4 (48.4) | 3.9 (22.6) | 8.2 (53.3) | 11.3 (42.2) | |
32.5 (3.6) | 24.5 (5.0) | 29.2 (7.5) | 8.1 (1.8) | 14 (4.9) | |
Wilson B (Å2) | 65 | 28.19 | 26.4 | 22.9 | 29.7 |
Mosaicity (°) | 0.6 | 0.2 | 0.5 | 0.7 | 0.7 |
Resolution (Å) | 27.5–2.75 (2.82 - 2.75) | 29.76- 1.95 (2.01 - 1.95) | 34.58 - 1.86 (1.91 - 1.86) | 34.58 - 1.86 (1.91 - 1.86) | 44.83 - 2.57 (2.67 - 2.57) |
22.9 (26.5) | 18.8 (21.34) | 17.18 (20.8) | 17.32 (28.10) | 20.14 (23.63) | |
Reflections | 12,789 (908) | 191,052 (12,500) | 19654 (1229) | 43800 (3090) | 31848 (2736) |
29.64 (37.6) | 22.97 (27.03) | 21.5 (29.4) | 20.24 (30.8) | 24.48 (31.43) | |
Free reflections | 667 (55) | 10136 (674) | 1061 (57) | 2336 (165) | 1656 (116) |
Refined Atoms | 3379 | 18,959 | 2101 | 4381 | 8155 |
Protein Chains | 4 | 18 | 2 | 4 | 8 |
r.m.s.d bond lengths (Å) | 0.013 | 0.007 | 0.021 | 0.023 | 0.004 |
r.m.s.d bond angles (°) | 1.452 | 1.052 | 1.854 | 1.863 | 0.762 |
B factor (mean) | 51.15 | 30.20 | 37.77 | 26.34 | 26.8 |
a Values in parentheses are for highest-resolution shell.
†
‡R = Σ||
Examination of the NL8-HEL structure showed a poorly-ordered L4, with electron density visible only for the backbone atoms, suggesting it made little contribution to binding. This loop was therefore targeted for randomisation and also the region randomised was extended by two residues to improve its chances of making contact with HEL.
At the β-sheet face, non-contacting residues S29 and A56 were identified as being small and close to the interface, suggesting that selection of larger residues at these positions may enable additional interactions with HEL. Residues K37 and P51 were identified as making possible negative contributions to binding; P51 because it is a β-breaking residue located on a β-strand, and K37 because of proximity to the guanidinyl group of R61 from HEL, leading to a like-charge clash.
A total of 10 residues in NL8 were therefore randomised (
(a) OBody sequences from the first round of affinity maturation, showing only the mutations at the targeted residues and labelled according to their sequence position in the library. For comparison, the equivalent positions from NL8 are also shown. Residues are coloured by polarity or charge (yellows = non-polar, greens = polar, red = acidic, blue = basic). (b) AM1L10 (pale blue Cα trace) in complex with HEL (pale green), superposed on to NL8 (dark blue, r.m.s.d. 0.45 Å) with the NL8-bound HEL also visible (dark green). Substitution A56Y introduces a hydrogen bond with HEL T47 and the subsequent relative shift in HEL position is evident, with a 2.5 Å shift in Cα position at that residue (indicated by red dashed line). (c) AM1L10 (pale blue Cα trace) in complex with HEL (pale green), superposed on to NL8 (dark blue, r.m.s.d. 0.45 Å) with the NL8-bound HEL also visible (dark green). At the top of the binding face, E95 accommodates the shift with a conformational change, maintaining close contact with a lysine and arginine from HEL. Dashed yellow lines are potential hydrogen or electrostatic bonds, labelled with distances in angstroms.
Analysis of the AM1L10 structure gave a clear rationale for the observed improvement in affinity: the substitution A56Y introduced a new hydrogen bond and increased buried surface area. However, analysis by the Robetta
For the next round of affinity maturation, AM1L200 was chosen as the parental clone over AM1L10. This was done for two main reasons: firstly, AM1L10 appeared only once in sequencing of individual phage clones from outputs, compared to multiple instances for all other unique AM1 sequences, suggesting possible problems with AM1L10 display or expression. Secondly, AM1L200 appeared to satisfy better the rational basis for selection of the randomised residues; S29H and A56R introduced large, polar residues and K37M removed the like-charge clash.
It was hypothesised that sparse random sampling of a wider sequence space throughout the whole AM1L200 OBody might yield improvements in affinity by introducing beneficial substitutions at sites other than the natural binding face or L4. Accordingly, full-length AM1L200 was subjected to error-prone (EP)-PCR. The resulting library was cloned into phagemid pRpsp2 and comprised 108 independent members; sequencing showed that there were an average of 4 nucleotide substitutions per 1000 bp. Solid-phase selections using immobilised HEL resulted in the isolation of clone AM2EP06, with an improved KD of 250 nM, as determined by SPR (
(a) Top hits from the third round of affinity maturation ranked by phage ELISA signal and showing the corresponding L4 sequence. (b) SPR sensorgrams for OBodies AM3L09 (blue line), AM3L15 (orange line) and AM2EP06 (green line) binding to HEL. For visual comparison, sensorgrams are shown only for the highest single concentration of each OBody analysed (AM3L09 at 32 nM, AM3L15 at 128 nM and AM2EP06 at 800 nM). These data were produced on a single chip and are representative of multiple independent analyses, performed with separate chips and protein samples. (c) Kinetic data for each OBody. The AM3L09 and AM3L15 kinetic data were calculated using Graphpad Prism association/dissociation modelling, whereas the affinity of AM2EP06 was calculated using an equilibrium model of maximum response. While the kon for both AM3L15 and AM3L09 are essentially the same, the difference in dissociation constant can be attributed to a substantial decrease (10-fold) in koff. Separate on and off rates could not be determined from the AM2EP06 data.
The crystal structure of AM2EP06 was solved in complex with HEL to a resolution of 1.86 Å. With three amino acid substitutions compared to the parent AM1L200 (T19S, M37K and K86E), it showed a binding face that was very similar to AM1L10 and NL8, with differences at the periphery. In a manner very similar to the AM1L10 structure, a small conformational change in R61 from HEL moved its guanidinyl group away from the interface.
In contrast to the previous two HEL-complex structures, L4, with an entirely new sequence, was moderately well-defined as electron density in the structure. The loop showed some contact with HEL, increasing buried surface area by ∼100 Å2, although no additional hydrogen bonds or other polar contacts were noted. The contribution of L4 to ΔG of binding to HEL appeard to be improving, which was suggested by the crystallographic B-factors. In the previous structures, atoms in the interface had higher B-factors than the average B-factor across the structure, whereas in AM2EP06, while L4 still showed B-factors consistent with a flexible surface loop (well above the average atomic B-factor), B-factors were slightly reduced compared to that seen for L4 in NL8 and AM1L10(
HEL is shown in green as a surface representation in identical orientations in each image and the L4 residues from each OBody variant, coloured according to B-factor relative to the average in each individual case; blue is low B-factor, graduating through green, yellow then red as it increases. The lower-affinity variants NL8 and AM1L10 both show poorly-ordered L4 residues. Note that residues D89, M90, H91 and N92 are missing from the model of AM1L10 L4, as they could not be resolved. Relative stabilisation is evident in L4 of the higher-affinity variants AM3L15 and AM3L09, compared to the parent clone AM2EP06, implying increased involvement in binding. Although the L4 structures of AM3L09 and AM2EP06 have superficially similar configurations, with Y88 and W90 binding in similar positions, W90 L4 from AM3L09 makes a greater number of contacts and is packed more closely. The alpha carbons of N- and C-terminal L4 residues have been labelled with ‘N’ and ‘C’ respectively, to denote the anchor points for the loop in each case.
Examination of the AM2EP06-HEL complex structure showed that few strong contacts with HEL were made by L4 residues. The third affinity maturation library therefore randomised six residues in L4 of AM2EP06 using a modified version of this sequence as the template, in which a stop codon was introduced by site-directed mutagenesis at position 90 in L4 (AM2EP06-stop). This construct ensured that only clones with a randomised L4 sequence would be functional in the resulting library. The pRpsp2 phagemid was also modified to remove a 2-residue duplication in the pelB leader and revert it to wild-type sequence, creating phagemid pAS1. Residues S85, E86, Q87, Y88, G89 and S90 of AM2EP06 were then randomised using oligonucleotides containing NNS codons by Kunkel mutagenesis
Three rounds of solution-phase selection were done using decreasing concentrations of biotinylated HEL (1 µM, 10 nM and 1 nM across rounds 1–3, respectively). A total of 192 randomly-picked variants were screened by phage ELISA from rounds two and three of selection. Sixteen unique clones were identified with an ELISA signal at least 2-fold higher than the parent clone AM2EP06 (AM3L01-16;
The interface between AM3L09 (KD = 3 nM) and HEL buries a total 1800 Å2 solvent-accessible surface area (SASA; 923 Å2 from the OBody) and maintains the intimate association with the HEL active site and substrate-binding groove as first established in the NL8-HEL complex. In a general sense, the primary interface residues are the same: a hydrophobic patch at the centre of the interface composed of Y33, V36 and I38, surrounded by an inter-molecular hydrogen bond network and electrostatic interactions (
(a) AM3L09 is coloured blue with interface residues shown as stick models. HEL residues calculated to make a hydrogen bond with AM3L09 are shown as green stick models. Potential hydrogen bonds are indicated by a dashed yellow line. (b) HEL electrostatic surface. The highly electronegative HEL active site (AS) is filled by R35 from AM3L09. (c) AM3L09 electrostatic surface, shown in the same orientation as panel a. The negatively charged patch containing D91 associates with a complementary positively charged patch on the HEL interface. (d) Comparative binding positions of AM3L09 (blue, thick Cα trace) and NL8 (red, narrow Cα trace) to HEL (green surface). (e) The AM3L09-HEL interface, shown in wall-eye stereo. Bridging water molecules between AM3L09 (blue) and HEL (green) are shown as red spheres. Potential hydrogen bonds are indicated by a dashed yellow line labelled with the length in angstroms.
In AM3L15 (KD = 35 nM) L4 has adopted a helical character, with no similarity to L4 in AM2EP06. The N-terminus of the helix in L4 is capped by a tryptophan cis-prolyl peptide bond (W86, P87). Intriguingly, this type of bond has been highly conserved across the proteins in which it has appeared during evolution (across all families of protein fold), underscoring its vital functional role
In both AM3L09 and AM3L15, complementary electrostatic charges are concentrated in two major patches. One consists solely of R35, which associates directly with the negatively charged active site of HEL. The other involves D91, which despiteL4 adopting completely different conformations in these two variants, occupies the same position in both structures, becoming one of three acidic residues (E83, E95 and D91) arranged in a line across the top of the interface, creating a negatively charged patch which binds to a lysine/arginine pair from HEL (
In addition to the protein atoms which interact directly, the AM3L09-HEL interface has a complement of highly-ordered water molecules which mediate both intra- and inter-domain interactions. In particular, two clusters of ordered solvent are closely involved with interface residues, with one anchoring the C-terminal end of L4, and another larger cluster arranged between critical binding residues Y33 and R35 (
Differential scanning fluorimetry
(a) Thermal denaturation by differential scanning fluorimetry of HEL-binding OBodies and a control non-HEL-binding OBody U81. Calculated Tm values are shown alongside the number of amino acid mutations as compared to the wild-type AspRS OB-fold domain from
HEL was originally chosen simply as a model protein for selection of OBody binders. However, the similarity of the binding to known HEL inhibitors with respect to interface statistics (
Molecule | Buried solvent-accessible area (A2) | H-bonds | Salt bridges | KD |
NL8 | 821 | 10 | 5 | 35 µM |
AM1L10 | 834 | 10 | 4 | 5 µM |
AM2EP06 | 945 | 10 | 3 | 250 nM |
AM3L15 | 974 | 14 | 6 | 30 nM |
AM3L09 | 923 | 16 | 6 | 3 nM |
Fab′ (1FDL) | 680 | 14 | 0 | 22 nM |
Camelid VHH (1JTP) | 772 | 8 | 0 | 50 nM |
YkfE (1GPQ) | 768 | 15 | 3 | 1 nM |
The development of OBodies has followed a similar path to other engineered non-antibody scaffolds, with the initial library design and selection producing binders of moderate affinity, in the micromolar range
Through examination of these structures, we were able to propose a rationale for the effect of the mutations selected at each step, from fine- to broad-scale changes. Selections from the naïve library isolated OBody NL8 (KD = 35 µM); the crystal structure of the NL8-HEL complex at 2.75 Å showed an OBody bound to the active site and substrate-binding groove of HEL, with a central hydrophobic patch made up of Y33, V36 and I38, surrounding polar contacts (11 hydrogen bonds) and three inter-domain complementary electrostatic charges. This structure yielded two critical pieces of information. Firstly, binding to HEL was mediated by residues targeted in the library, showing that the library design was successful, even though it was based on a scaffold for which there was no available structural data. This is unusual for a new scaffold, as crystallographic structural data is usually considered vital as a starting point. The NL8-HEL interface made no apparent use of the four randomised positions in L4 of NL8 (out of a total of 17 distributed throughout the domain), which immediately suggested that the recruitment of L4 into binding interactions might be a viable strategy for affinity maturation. Secondly, the structure of the complex showed that the beta-sheet binding face had statistics which were not excessively different from other HEL-binding proteins, most notably antibodies, strongly suggesting that with appropriate maturation steps the affinity could be improved.
Selections from the first affinity maturation library yielded the NL8 variants AM1L10 (KD = 5 µM) and AM1L200 (KD = 1 µM), representing 6- and 35-fold improvements in affinity over NL8, respectively. The crystal structure of the AM1L10-HEL complex at 1.95 Å revealed two major factors in the improvement in binding: the introduction of additional inter-domain contacts at Y56, and a consequent moderate shift in binding orientation compared to the NL8-HEL complex. To some extent the pattern-recognition function of hydrogen bond networks is similar in both intra- and inter-molecular interactions
For the second phase of affinity maturation, AM1L200 was used as the template for an EP-PCR library, selections from which produced variant AM2EP06 (KD = 250 nM), representing a 4-fold improvement in affinity over AM1L200. The crystal structure of the AM2EP06-HEL complex was solved at 1.86 Å. In this case the exact molecular determinants of the improvement in affinity were less clear, but the most likely explanation seemed to be a change in L4 dynamics through the K86E substitution. Residue E86 from AM2EP06 made limited contacts with HEL, and its sidechain was poorly-ordered. Although some role in long-range electrostatic attraction may be hypothesised, we think it most likely that this mutation increased the occupation of binding-favourable conformations of L4 by removing competing interactions with HEL. Indeed, the NL8 and AM1L10 structures both showed K86 in a HEL-contacting position which could not be accommodated at the same time as the L4 arrangement following the K86E substitution in AM2EP06.
In the third and final phase of affinity maturation, six residues in L4 of AM2EP06 were randomised and selections from this library resulted in the isolation of AM3L15 (KD = 30 nM) and AM3L09 (KD = 3 nM), constituting 8- and 83-fold improvements in affinity over AM2EP06, respectively. In this case, selections were performed in solution using biotinylated target and trypsin-sensitive helper phage KM13
While the naïve library selections were clearly successful, indicating that the library design was sufficient, the structures solved subsequently showed an important property of the designed interface; interactions with HEL relied heavily on, and were consequently limited by, the involvement of wild-type residues. This is, at least partially, a consequence of the way NL8 bound its first β-strand into the HEL substrate binding groove, which forced non-randomised residues on the underside of the β-strand to make contact with HEL residues. In addition, the two regions targeted, namely L4 and the β-sheet binding face, did not form a continuous surface, which resulted in non-randomised residues from β-strands 4 and 5 also being forced into contact. Both of these non-randomised regions contain charged residues; especially important is E95 on β-strand 5, positioned directly between L4 and the β-sheet interface. While this may or may not be an issue for binding epitopes on a given target, we may be able use this most recent structural information to design an improved binding surface, by repositioning the scaffold binding region and removing charged groups which might otherwise interfere.
In addition, these five structures provide an empirical basis for design of OBody libraries in the future, enabling precise selection of target residues for mutations and assessment of resulting hypothetical binding surfaces without the need for further low-throughput structural studies of the kind employed here.
OBodies can also be produced easily by expression in
We have shown that an OBody scaffold is capable of binding in monomeric form with nanomolar affinity to HEL, a model target ligand. Detailed structural information at each step, from naïve clone and through the affinity maturation process showed an iterative improvement in interface statistics, and demonstrated the scaffold’s ability to maintain highly diverse sequences (mutating up to 22% of the protein) with no appreciable change in core residue arrangement (Cα r.m.s.d. across all of the structures of 0.2 Å, excluding surface loops ) and maintenance of the thermophilic character of the ancestral OB-fold domain (Tm 72°C for AM3L09). Due to their small size (13-fold smaller than IgGs), lack of disulfides and superior biophysical characteristics, OBodies derived from thermophilic bacterial OB-fold domains may be useful for extended-shelf-life
These libraries were constructed by PCR dissection of the ancestor gene into overlapping fragments followed by overlap-extension assembly incorporating degenerate oligonucleotides. The naïve OBody library based on the OB-fold from
Error-prone (EP)-PCR
The third affinity maturation library was made using Kunkel mutagenesis with NNS randomisation in the pAS1 phagemid, essentially as described elsewhere
Generation of heteroduplex DNA was performed using degenerate oligonucleotide 247 (
Purified phage were generated using VCSM13 helper phage (Agilent Technologies, USA), packaging the pRpsp2 phagemid
Eluted phage from a single selection were added to 10 mL log-phase TG1
Purified phage were generated using KM13 helper phage
To generate monoclonal phage samples and perform phage ELISAs for screening selections from the third round of affinity maturation, individual clones were grown in 100 µL 2YT with 100 µg/mL carbenicillin in a 96-well tissue culture plate overnight at 37°C, 150 rpm. The resulting saturated cultures were seeded into 500 µL 2YT with 100 µg/mL carbenicillin in 96-deepwell (2 mL) blocks and incubated for approximately 5 h at 37°C, 250 rpm, or until turbidity was evident. The cultures were infected with KM13 helper phage at ∼109 pfu and incubated at 37°C, 150 rpm, for 1 h. Media was exchanged by centrifugation for 10 min at 3,000
From the deepwell plate, 50 µL of the prepared phage solution was transferred in duplicate to the ligand and control plates then incubated at room temperature for 1 h. Anti-HEL human antibody (HuCAL Clone 11397, AbD Serotec, Oxford, UK) 50 µL in PBS-3M at 1 µg/mL was used as a positive control. A blank negative control well with 50 µL PBS-3M only was also used. The wells were washed three times with 300 µL PBS-T, then probed for 1 h with 50 µL monoclonal HRP-conjugated anti-M13 antibody (dilution 1∶5000, GE Healthcare), or HRP-conjugated polyclonal goat anti-human Fc (dilution 1∶1000, AbD Serotec) in the case of the positive control. The plates were washed 3 times as above. For visualisation, 50 µL 3,3′,5,5′-tetramethylbenzidine (TMB) peroxidase substrate (Sigma) per well was added and incubated for 30 min at room temperature, then the reaction stopped with 50 µL 2 M H2SO4 per well and read in in a spectrophotometer at 450 nM.
To generate monoclonal phage samples and perform phage ELISAs for screening all other selections, 96-well ELISA plates were coated with 5 µg/ml hen egg white lysozyme or 1% BSA, in PBS at 4°C overnight. After two washes with PBS, plates were blocked with PBS containing 5% (w/v) skim milk powder (PBS-5M) for 1 h at RT before helper phage (109 per well; ΔgIII helper phage VCSM13d3
Measurement of changes in protein stability was performed using the general method for fluorescent detection with SYPRO Orange and an RT-PCR instrument, outlined elsewhere
Dissociation constants of the various Obodies for HEL were measured using a Biacore 3000 surface plasmon resonance instrument (GE Healthcare). HEL was immobilised on to a CM5 sensor chip in small steps to give approximately 100 response units (RU). Obodies were measured in 2-fold dilution series, beginning with at least 10-times the KD, where known. Analysis runs were performed at a standard 50 µL/min using HBS-EP running buffer (GE Healthcare). Chip surfaces were regenerated after each analyte injection with two 10 µL injections of 1 M NaCl. Where possible, kinetic data was modelled to obtain the kon/koff using global fitting to a 1∶1 Langmuir binding model in BIAevaluation software version 3.2, otherwise steady-state dissociation constants were modelled by plotting the maximum response at each concentration vs. OBody concentration in mol/L and fitting a Langmuir saturation binding curve to the data, using GraphPad Prism with the following equation, where Rmax is the maximum response, KD is the dissociation constant and for the linear component, m is the gradient and c is the y-intercept of the linear portion:
Assays were carried out in triplicate, using flat-bottomed 96-well untreated ELISA plates (Greiner). OBodies suspended in PBS pH 7.4 were first diluted 1∶2 serially over 11 steps, starting with a final assay concentration of 50 µM. A 25 µL aliquot of each OBody dilution was either added to 25 µL hen egg-white lysozyme (Sigma) at 648 nM (162 nM final concentration in assay) or 25 µL PBS pH 7.4 and allowed to equilibrate at room temperature for 15 min. To each OBody dilution, a 50 µL aliquot of substrate solution containing inactivated
All crystal were grown using the sitting-drop vapour-diffusion method. Conditions were as follows: NL8, 0.2 M HEPES pH 7.3, 7% MPEG 5,000; AM1L10, 0.2 M HEPES pH 7.4, 9% MPEG 5,000; AM2EP06, 0.2 M HEPES pH 7.0, 13% MPEG 5,000; AM3L15, 0.2 M HEPES pH 7.4, 5% MPEG 5,000; AM3L09, 0.2 M HEPES pH 7.8, 9% MPEG 5,000. Purified, concentrated OBody in PBS pH 7.4 was combined with equimolar HEL in 10 mM sodium acetate pH 5.0 to yield concentrations of 40 mg/mL (OBody) and 45 mg/mL (HEL). Diffraction data was collected using the home source at the Maurice Wilkins Centre, University of Auckland, the Stanford Synchrotron Radiation Lightsource, or the Australian Synchrotron MX2 beamline. Data were integrated using MOSFLM
Biotinylated HEL was produced using an amine-reactive biotinylation reagent from Qanta Biodesign, which attaches a biotin with a polyethylene linker via an N-hydroxysuccinimide group. The reagent was dissolved in dry dimethyl sulfoxide (DMSO) to make a 100 mM stock solution and stored at −20°C. The HEL was obtained from Roche and purified on a Superose 75 10/300 size exclusion column and incubated with a 5-fold molar excess of the biotinylation reagent for 1 h. The reaction was halted with the addition of ethanolamine to 1 mM. The sample was dialysed into ultrapure water and analysed for biotinylation levels by MALDI-TOF mass spectrometry.
Selected OBodies were subcloned by restriction/ligation into expression vector pRoEx Htb (Life Technologies) using generic OBody primers 005 and 006 (
Coordinates and structure factors for the reported crystal structures have been deposited at the Protein Data Bank under accession codes 4GLA (NL8-HEL), 4GN3 (AM1L10-HEL), 4GN4 (AM2EP06-HEL), 4GLV (AM3L09-HEL) and 4GN5 (AM3L15-HEL).
(TIF)
(TIF)
(TIF)
(TIF)
(TIF)
(XLSX)
The authors would like to thank Abby Sharrock for help with construction of pAS1. We thank David Goldstone for assistance with X-ray data collection and Meredith Ross for critical reading of the manuscript. We also thank Ted Baker and the Maurice Wilkins Centre for Molecular Biodiscovery for the use of their X-ray facilities.