We analyzed a large number of deposited camelid nanobody structures and found that different lengths and arrangements of the CDR3 results in three groups of interaction surfaces, which can be described as concave, loop and convex. In order to mimic the surface complementarity repertoire of the camelid immune system and its capacity to efficiently target membrane proteins, we designed three sybody libraries based on prototypical camelid nanobody structures representing these three binding modes. The first template nanobody is a GFP-binder with a short CDR3 (six amino acids (aa) according to [Sircar et al., 2011]) that binds via a concave surface (PDB: 3K1K) (Kirchhofer et al., 2010). The second template is a β₂-adrenergic receptor binder that inserts a medium length CDR3 (12 aa) as an extended loop into a receptor cavity (PDB: 3P0G) (Rasmussen et al., 2011a). The third template is a lysozyme binder displaying a long CDR3 (16 aa), which is tethered via an extended hydrophobic core, and binds via a convex surface (PDB: 1ZVH) (De Genst et al., 2006) (Figure 1, Figure 1—figure supplement 1, Table 1). The resulting sybody libraries were therefore dubbed ‘concave’, ‘loop’ and ‘convex’, accordingly.

Figure 1 with 6 supplements see all

Download asset Open asset

Apart from the CDR regions, 3K1K and 3P0G share high sequence identities and therefore the concave and the loop library share the same framework (Figure 1—figure supplement 2). In contrast, 1ZVH contains an extended hydrophobic core and the convex library was therefore built on a different scaffold (Figure 1—figure supplement 2). A single, conserved disulfide bond at the center of the immunoglobulin domain is common to all three scaffolds. Three non-randomized scaffold sybodies representing the concave, the loop and the convex library were generated by gene synthesis. They contain serines and threonines at positions to be randomized in the libraries (Table 2, Figure 1—figure supplement 2). The corresponding purified proteins eluted as a single species from a size exclusion chromatography column (Figure 1—figure supplement 3A). They exhibited high melting temperatures of 74, 75 and 95°C for the concave, loop and convex scaffold, respectively, which corresponds to a stability increase of 21 to 35°C compared to their natural precursors (Figure 1—figure supplement 3B). Thermal stability of the convex sybody is particularly high. We attributed this increased stability to the extended hydrophobic core to tether CDR3 and the V51L substitution introduced into the framework prior to the CDR2 region.

Based on these scaffolds, the three sybody libraries were constructed by randomizing all three CDRs using defined mixtures of trinucleotides, thereby obtaining an optimal balance between charged, polar, aromatic and apolar amino acids to achieve an overall moderate hydrophobicity of the randomized surface (Figure 1—figure supplement 1). Decreased surface hydrophobicity was previously demonstrated in a DARPin library to be of paramount importance to counteract the enrichment of sticky binders when selecting against membrane proteins (Seeger et al., 2013). Three trinucleotide mixes were used for randomized residues placed (i) in loops, (ii) at the transitions from loops to β-sheets and (iii) in the middle of β-sheets (Figure 1—figure supplement 2C). Cysteines and prolines were generally excluded in any of the three mixes. Mix one is enriched by the residues A, S, T, N, Y (10.6%, each), contains D, E, Q, R, K, H, W at 5% frequency each and harbors only few of the apolar amino acids F, M, V, I, L, G (2%, each). Mix two lacks amino acids D and A, because these two residues are underrepresented at the end of β-sheets (Bhattacharjee and Biswas, 2010). Mix three is devoid of D, N, Q, G, S and M, because these amino acids are found less frequently in the middle of β-sheets (Bhattacharjee and Biswas, 2010). The theoretical diversity of the libraries amounts to 8.3 × 10¹⁷, 4.3 × 10¹⁹, and 2.8 × 10²² for the concave, loop and convex library, respectively. Three DNA fragments each containing one CDR of each of the three libraries were generated by assembly PCR. The resulting fragments were ligated in two subsequent steps using Type IIS restriction enzymes analogous to the assembly of designed ankyrin repeat proteins (Seeger et al., 2013). Finally, the three sybody libraries were flanked with the required sequence elements for in vitro transcription and ribosome display (Table 2). We determined an experimental library diversity of 9 × 10¹² for each of the three libraries.

Provided that libraries of high quality are used, binder affinities obtained by in vitro selections largely depend on the displayed library size in the initial selection round, because other than in the animal, no affinity maturation is performed. In contrast to the widely used phage and yeast display systems, which have maximal library sizes of 10⁹–10¹⁰ members, ribosome display offers the advantage of displaying 10¹² different library members with minor experimental effort. In ribosome display, a stable ternary complex between an encoding mRNA, the ribosome and the folded nascent polypeptide chain is formed. However, ribosome display is not widely used, because it used to require the preparation of home-made in vitro translation reagents and is associated with variable levels of unfavorable RNase acitivity (Zahnd et al., 2007). To overcome this technical hurdle which prevented non-expert labs from using this efficient display method, we implemented the commercial in vitro translation kit PUREfrexSS (GeneFrontier) for ribosome display. The kit is devoid of reducing agents and contains oxidized glutathione (GSSG) and the disulfide bond isomerase DsbC and is thus suited to support the folding of disulfide-containing proteins such as nanobodies and sybodies. We experimentally tested display efficiency in two independent assays. In a first assay, we fused a 3xFLAG tag followed by 3C protease cleavage site to the C-terminus of the non-randomized loop sybody in a construct containing the flanking regions for transcription and ribosome display (Figure 1—figure supplement 4A). The mRNA of this construct was displayed on ribosomes using the PUREfrexSS kit, sybody-3xFLAG was cleaved from the nascent polypeptide chain by 3C protease and the entire protein mixtures was analyzed by Western blotting using an anti-FLAG antibody. A purified non-randomized convex sybody containing exactly the same 3xFLAG sequence and 3C protease cleavage site at the C-terminus served as standard for protein quantification by Western blotting. The analysis revealed that more than 70% of the input mRNA was translated. In a second assay, we examined whether ribosome display produces correctly folded binders. To this end, we displayed 10⁶ mRNA encoding the 3K1K nanobody spiked into 10¹² mRNA encoding the non-randomized convex sybody using PUREfrexSS kit and assessed binding to immobilized GFP (target of 3K1K) and MBP (negative control) (Figure 1—figure supplement 4B). Quantitative PCR on reverse transcribed cDNA was used to determine the amount of mRNA which could be retrieved after binding and washing in comparison to the input mRNA added to the display reaction. For 3K1K panned against GFP, mRNA recovery was 84.6 ± 3.5%, while mRNA retrieved from 3K1K panning against MBP was not detectable. Background binding of the non-randomized convex sybody towards GFP and MBP was minimal (recovery fractions below 0.001%). These two independent experiments clearly demonstrated that ribosome display of single domain antibodies works efficiently using the commercial PUREfrexSS kit.

To validate the sybody libraries, first selections encompassing three consecutive rounds of ribosome display were carried out against the soluble maltose binding protein (MBP), which can be considered as easy target (Figure 2—figure supplement 1). Sybody pools were found to be strongly enriched after the third selection round as monitored by qPCR. Using FX cloning (Geertsma and Dutzler, 2011), single sybodies were introduced into expression vector pSb_init, which directs the protein into the periplasm by virtue of a PelB leader sequence and adds a Myc- and a His-tag to the C-terminus of the sybody for detection by ELISA (Table 3, Figure 1—figure supplement 5). Of note, pSb_init contains BspQI restriction sites, which permits to release sybodies for sub-cloning into expression plasmids pBXNPH3 and pBXNPHM3 for the production of tag-less binders for crystallization purposes. ELISA analysis of the selections of the concave, loop and convex library revealed about 20, 50 and 30% of all wells as strong and specific hits against MBP. Crystal structures of three convex MBP binders (Sb_MBP#1–3) in complex with MBP were solved at resolutions ranging from 1.4 to 1.9 Å (Figure 2, Table 4). The structures of the crystallized sybodies were highly similar to their natural precursor (e.g. RMSD of 1.02 Å comparing Sb_MBP#1 and 1ZVH), thereby validating our library design, which kept selected residues of the CDRs constant to assure folding of an extended hydrophobic core (Figure 2—figure supplement 2). Sb_MBP#1–3 have binding affinities ranging from 24 to 500 nM as determined by surface plasmon resonance (SPR). They bind into the cleft between the two lobes of MBP and thereby trap the target in its ligand-free conformation (Figure 2A, Figure 2—figure supplement 3) (Spurlino et al., 1991). In support of this notion, SPR measurements revealed decreasing sybody binding affinities at increasing maltose concentrations. A Schild plot analysis revealed 58-fold decreased sybody binding affinity at saturating maltose concentration and an affinity of 1.0 µM of MBP for maltose (Figure 2B, Figure 2—figure supplement 4), which is in close agreement with the literature (Telmer and Shilton, 2003). An analysis of the binding interface highlighted CDR3 residues W101, Q104, S105 and W110, which are identical among the three binders (Figure 2—figure supplement 3). Further contacts are mediated by variable randomized residues of CDR1, CDR2 and CDR3 as well as several invariant framework residues. Of note, the sybody selections against MBP generated a highly variable set of binders from all three libraries exhibiting affinities down to 0.5 nM (Figure 2—figure supplement 1). These sybodies are expected to bind to a variety of epitopes on MBP, but were not further analyzed by crystallization trials. In summary, the crystallized sybodies bind to MBP in an analogous fashion as camelid nanobodies, namely via interactions predominantly mediated by CDR3 residues (De Genst et al., 2006; Desmyter et al., 2001).

Figure 2 with 4 supplements see all

Download asset Open asset

Three consecutive rounds of ribosome display (analogous to the successful selections against MBP) were insufficient to obtain sybodies against membrane proteins, as we demonstrated at the example of the ABC transporter TM287/288 (Figure 1—figure supplement 6A). In order to analyze the problem, we used qPCR to quantify the cDNA corresponding to the eluted mRNA after each selection round. We were thus able to follow the enrichment of binders against the target compared to control proteins. Increasing amounts of mRNA were pulled down in subsequent selection rounds, but to a similar extent for the control proteins and the target, thus indicating selection bias. As every display system provides selective advantage for a particular subset of background binders, we hypothesized that alteration of display systems will allow us to eradicate a major source of selection bias. Furthermore, our qPCR analyses revealed, that the output from the initial ribosome display selection round yields 10⁶–5 × 10⁶ different sybodies, which can be used to generate a focused phage display library covering 10⁷–5 × 10⁷ sybodies. A first test selection using one round of ribosome display followed by two rounds of phage display against TM287/288 resulted in moderate binder enrichment and gave rise to only a few positive ELISA hits (Figure 1—figure supplement 6B). In the context of a separate study, we selected sybodies against the heterodimeric ABC transporter IrtAB, which is a homologue of TM287/288 (sequence identity 27%). We consider TM287/288 and IrtAB to represent similar difficulty levels for binder selections, because they exhibit a highly similar shape and thus a similar number of available epitopes. To further suppress accumulation of background binders, solution panning was performed, that is ribosome or phage display particles were first incubated with IrtAB in solution, followed by a target pull-down via streptavidin/neutravidin-coated surfaces. In addition, surface chemistries were altered in every selection round, namely magnetic Dynabeads Myone Streptavidin T1 for ribosome display, neutravidin-coated Maxisorp microtiter plates for the first phage display round and magnetic Dynabeads Myone Streptavidin C1 for the second phage display round. These alterations at the level of target immobilization together with the combination of ribosome and phage display resulted in a favorable selection outcome, as manifested by a high number of positive ELISA hits obtained after sybody selection (Figure 1—figure supplement 6C). However, only 25% of the sequenced ELISA hits were unique, hinting at diversity bottlenecks in the selection protocol. In addition, the strongest binder exhibited a disappointingly low affinity of only 238 nM. We suspected two potential bottlenecks in our selection cascade, namely the (i) PCR amplification of cDNA to recover the output of the initial ribosome display round and (ii) phage infection of E. coli to recover the output of the first phage display selection round. To remove the first bottleneck, we used Taq polymerase instead of proof-reading polymerases for cDNA amplification, since the 3’ to 5’ exonuclease activity of proof-reading polymerases degrades single stranded DNA. To examine the second bottleneck, we measured the infection rate of the M13 phages and found that only 2–5% of the eluted phages after the first panning round were infecting cells, which resulted in a substantial loss of sybody diversity. To compensate for low infection rates, we increased the volume in the first round of phage display from 100 µl to 4.8 ml. When the combined improvements were applied in a selection against TM287/288, excellent enrichment, high number of positive ELISA hits and high affinities down to the single digit nanomolar range were obtained (Figure 1—figure supplement 6D).

In its final form, the sybody platform is built as a selection cascade, starting with one round of ribosome display, followed by two rounds of phage display (Figure 1). Furthermore, immobilization surface chemistries are changed in every selection round. Our selection cascade thus introduces maximal changes in each selection round at the level of binder display and target immobilization and proved highly effective in enriching sybodies against challenging membrane proteins as shown below.

ABC transporters harness the energy of ATP binding and hydrolysis to transport substrates against a concentration gradient. In the absence of ATP, the bacterial ABC transporter TM287/288 almost exclusively adopts an inward-facing (IF) state, and two crystal structures were solved in this conformation (Hohl et al., 2012; Hohl et al., 2014). In contrast, a structure of outward-facing (OF) TM287/288 is still missing due to difficulties in stabilizing this alternate conformation. The transition to the OF state requires ATP binding (Figure 3A) (Timachi et al., 2017), but ATP hydrolysis constantly reverts the transporter back to its IF state. In order to populate the OF state, a glutamate to alanine substitution (E517A) in the ATP-binding cassette was introduced, which blocks ATP hydrolysis without impairing ATP binding (Timachi et al., 2017). Using the sybody platform (Figure 1), binders were selected in vitro against TM287/288(E517A) in the presence of ATP (Figure 3). Using qPCR and AcrB as background control (Seeger et al., 2013), we observed strong sybody enrichment of 170, 220 and 25 fold for the concave, loop and convex library, respectively, after the second round of phage display. For each library, 190 clones were analyzed for binding against TM287/288(E517A) in the presence of ATP by ELISA, of which 60% were ELISA positive. Of 48 sequenced ELISA hits, 40 were unique, indicating high diversity after binder selection (Figure 4, Figure 4—figure supplements 1–3). The unique sybodies were named Sb_TM#1–40, and 37 thereof could be purified at sufficient yields and quality for further analyses. State specificity of these binders was assessed by SPR using wildtype and E517A mutant of TM287/288 as ligands and the sybodies as analytes in the presence and absence of ATP (Figure 3B). SPR data of 31 binders could be quantified and revealed 11 sybodies specific for OF TM287/288(E517A) as defined by an affinity increase of at least ten-fold upon addition of ATP (Figure 4). Sb_TM#26 exhibited a particularly strong state-specificity as binding was exclusively detected in the presence of ATP. Therefore, this binder was analyzed for its capacity to inhibit ATP hydrolysis of TM287/288, revealing an IC₅₀ of ATP hydrolysis of 62 nM (Figure 3C). In summary, sybody selections against OF TM287/288 resulted in a large number of high affinity binders trapping the transporter in its ATP-bound state and thereby strongly inhibiting ATPase activities.

Figure 3

Download asset Open asset

Figure 4 with 3 supplements see all

Download asset Open asset

There are only a small number of approved drugs or drugs in development, which therapeutically target human SLC transporters, indicating untapped potential (Lin et al., 2015). A main reason behind these shortcomings is the intricate architecture and low thermal stability that makes human SLC transporters notoriously difficult to work with in early drug discovery stages (César-Razquin et al., 2015). Here we focus on two transporters with a high need for conformation-specific binders for the screening of small molecule therapeutics, namely the equilibrative nucleoside transporter 1 (ENT1, SLC29A1) that is involved in ischemia and acts as a biomarker in pancreatic cancer (Yang and Leung, 2015), as well as on the glycine transporter 1 (GlyT1, SLC6A9) that plays an important role in diseases of the central and peripheral nervous system (Harvey and Yee, 2013) (Figure 5A, Figure 6A). Multiple attempts to raise mouse antibodies or nanobodies against these targets by immunizations failed in our hands, presumably due to low thermal target stability and the limited number of accessible epitopes.

Figure 5 with 1 supplement see all

Download asset Open asset

Figure 6 with 2 supplements see all

Download asset Open asset

In order to obtain conformation-specific binders against ENT1 and GlyT1, we performed selections at 4°C in the presence of the inhibitors S-(4-Nitrobenzyl)−6-thioinosine (NBTI) and a Bitopertin-like molecule named Cmpd1, respectively. For ENT1, the concave but not the convex and loop library was enriched 4-fold over background after binder selection. One concave sybody called Sb_ENT1#1 was identified by ELISA and purified as monodisperse protein (Figure 5—figure supplement 1). SPR measurements revealed an affinity of 40 nM (Figure 5B, Table 5). To further characterize Sb_ENT1#1, a thermal shift scintillation proximity assay (SPA-TS) was established. ENT1 in complex with the sybody was incubated at varying temperatures in the absence of inhibitor, followed by measuring binding of tritiated NBTI. Sb_ENT1#1 binding led to a sharper transition trajectory and an increase in melting temperature by 6.1°C (Figure 5C), indicating that sybody binding increases the population of the inhibited conformation of ENT1. Supporting this notion, the absolute binding signal for NBTI was increased by more than seven-fold in the presence of Sb_ENT1#1 at temperatures well below T_m (Figure 5D).

For GlyT1, seven sybodies from the concave (Sb_GlyT1#1–4) and loop (Sb_GlyT1#5–7) library were identified by ELISA and complex formation was confirmed by size-exclusion chromatography (SEC) (Figure 6B, Figure 6—figure supplement 1). Binding kinetics were determined by SPR revealing a wide range of affinities from 494 pM to 2.52 µM (Figure 6C and D, Table 5, Figure 6—figure supplement 2). Competition binding SPR analysis using GlyT1 pre-saturated with Sb_GlyT1#6 revealed binding of concave sybodies Sb_GlyT1#1–4 but not of loop sybodies Sb_GlyT1#5 and Sb_GlyT1#7 (Figure 6E), demonstrating that at least two non-overlapping epitopes on GlyT1 are recognized. The large differences in affinities correlated well with SPA-TS analysis using the commercially available tritiated inhibitor Org24598 that addresses the same binding site as Cmpd1 (Alberati et al., 2012). Of the seven sybodies, Sb_GlyT1#1–5 increased the T_m by 0.9–2.6°C, whereas Sb_GlyT1#6 and Sb_GlyT1#7 stabilized the transporter by 8.8 and 10°C, respectively (Figure 6F). Absolute SPA binding signals obtained at 19°C increased up to 1.8-fold (Sb_GlyT1#7), suggesting that all sybodies stabilize the inhibited conformation of GlyT1 (Figure 6G).

In conclusion, our selection at 4°C and in the presence of non-covalent inhibitors enabled the rapid identification of sybodies against instable and previously intractable human membrane proteins. The identified binders trap inhibited conformations of ENT1 and GlyT1 and thereby enhance ligand binding. Hence, these binders increase assay sensitivity for inhibitor screening and can serve as crystallization chaperones for structure-based drug design.

An FX cloning vector for the periplasmic production of VHHs preceded by an N-terminal decaHis-tag and an HRV 3C protease cleavage side, designated pBXNPH3 (Figure 1—figure supplement 5), was constructed by polymerase chain reaction (PCR) using Phusion polymerase and pBXNH3 (Geertsma and Dutzler, 2011) as template in combination with the 5’-phosphorylated primers pBXNPH3_#1 and pBXNPH3_#2. The resulting 5848 bp product was DpnI-digested, column-purified, ligated and transformed to chemically-competent ccdB-resistant E. coli DB3.1. A variant designated pBXNPHM3 (Figure 1—figure supplement 5), which is compatible with the periplasmic production of VHHs as a fusion to an N-terminal decaHis-tag, maltose binding protein (MBP) and HRV 3C protease site, was constructed from fragments of pBXNH3 (Geertsma and Dutzler, 2011) amplified using primer pair pBXNPHM3_#1 (holding an NotI restriction site) and pBXNPHM3_#2 (5’-phosphorylated) and pET26FX (Bertozzi et al., 2016) amplified using primer pair pBXNPHM3_#3 (5’-phosphorylated) and pBXNPHM3_#4 (holding an NotI restriction site). The resulting products of 4241 bp (vector backbone) and 2732 bp (insert holding mbp and ccdB) bp, respectively were DpnI-digested, gel-purified, cut with NotI, ligated and transformed into E. coli DB3.1. To obtain pSb_init, the ampicillin (amp) resistance gene of pBXNPH3 was replaced by a chloramphenicol (cat) marker. Because the kill cassette of pBXNPH3 contained a chloramphenicol marker as well, pBXNPH3 containing nanobody 1ZVH was used as template to amplify the vector without the amp gene with the primer pair pBXNPH3_blunt_for/pBXNPH3_EcoRI_rev. The cat gene was amplified from pINIT_cat (Geertsma and Dutzler, 2011) (Addgene #46858) using primers Cm_EcoRI_for and Cm_blunt_rev (5’-phosphorylated). The PCR products were column purified, digested with EcoRI, purified by gel extraction, ligated and transformed into E. coli MC1061. The resulting plasmid was amplified by primers Nb_init_for (5’ phosphorylated) and Nb_init_rev and circularized by ligation. The resulting vector was cut with SapI and the kill cassette, excised from pINIT_cat using the same restriction enzyme, was inserted resulting in vector pSb_init. The FX cloning vector for phage display, named pDX_init, was constructed based on pMESy4 (Pardon et al., 2014). The internal SapI site preceding the lac promoter in pMESy4 was removed by generating a short 500 bp PCR fragment with a mutation in the SapI recognition site using the primer pair pDX_init_#1/pDX_init_#2. The NcoI/SapI-digested PCR product was gel-purified and ligated into NcoI/SapI-digested and dephosphorylated pMESy4 and transformed into E. coli MC1061. The amber stop codon on the resulting vector was replaced by a glutamine residue using Quickchange mutagenesis using the primer pair pDX_init_#3/pDX_init_#4. The resulting vector backbone was amplified using primer pair pDX_init_#5/pDX_init_#6, thereby introducing SapI sites as part of the open reading frame, digested with SapI and ligated with a SapI-digested PCR-fragment holding the counterselection marker sacB amplified from pINIT (Geertsma and Dutzler, 2011) using primer pair pDX_init_#7/pDX_init_#8.

Sybodies were expressed in E. coli MC1061, which were grown in terrific broth containing 25 µg/ml chloramphenicol (in case of pSb_init) or 100 µg/ml ampicillin (in case of pBXNPH3 as well as pBXNPHM3) to an OD₆₀₀ of 0.7 at 37°C. Then the temperature was lowered to 22°C and cells were induced with 0.02% (w/v) L-arabinose for 15 hr. Cells were disrupted using a microfluidizer processor (Microfluidics, Westwood, MA, United States) at 25’000 lb/in² in TBS (20 mM Tris-HCl pH 7.5, 150 mM NaCl) supplement with 2 mM MgCl₂ and 25 μg/ml DNAse I. Cell debris was removed by centrifugation at 8’000 g for 20 min and 15 mM imidazole pH 7.5 was added prior to loading onto a gravity flow Ni-NTA superflow column of 2 ml bed volume (Qiagen, Venlo, The Netherlands). The column was washed with 25 ml TBS containing 30 mM imidazole pH 7.5 in case of pSb_init (50 mM in case of pBXNPH3 or pBXNPHM3) and the sybody was eluted with 5 ml TBS containing 300 mM imidazole pH 7.5. If expressed in pBXNPH3 or pBXNPHM3, the Ni-NTA purified sybodies were dialyzed against TBS in the presence of 3C protease for 3 hr and loaded onto Ni-NTA columns to remove the His-tag or the His-tagged MBP as well as the 3C protease. Tag-free sybodies were eluted from the Ni-NTA column using TBS containing 30 mM imidazole. Sybodies were concentrated using centrifugal filters with a 3 kDa cut-off (Amicon Ultra-4) and separated by size exclusion chromatography (SEC) using Superdex 200 Increase 10/300 GL (GE Healthcare, Glattbrugg, Switzerland) or Sepax-SRT10C SEC-300 (Sepax Technologies, Newark, DE, United States) in TBS.

Synthetic genes encoding for three non-randomized scaffold sybodies (convex, loop and concave) were ordered at DNA2.0 (Table 2). These scaffold sybodies contain serines and threonines in the positions to be randomized in the respective libraries and served as PCR templates for library assembly. Their sequences harbored within expression vector pBXNPHM3 were made available through Addgene (Table 3). Primers (Table 6) were added to the PCR reaction at a concentration of 0.8 µM if not specified differently. Primers for the sybody assembly were ordered in PAGE-purified form (Microsynth, Balgach, Switzerland). Randomized primers were synthetized using trimer phosphoramidites (Ella Biotech, Martinsried, Germany). If not otherwise mentioned, Phusion High-Fidelity DNA Polymerase (NEB, Ipswich, MA, United States) was used for PCR amplification. Five double-stranded PCR products, which served as megaprimers in the assembly of the library were first amplified from the genes encoding for the frameworks of the concave/loop and convex library. The gene of the concave/loop sybody framework was amplified with primer pairs FW1_a_b_for/FW1_a_b_rev (megaprimer 1), FW3_a_b_for/Link2_a_rev (megaprimer 2) or FW3_a_b_for/Link2_b_rev (megaprimer 3). The gene of the convex sybody framework was amplified with primer pairs FW1_c_for/FW1_c_rev (megaprimer 4) and FW3_c_for/Link2_c_rev (megaprimer 5). Megaprimers were gel-purified. In a second step, the individual CDR regions of the libraries were assembled by overlap extension PCR using Vent DNA Polymerase (NEB), applying 35 cycles and an annealing temperature of 60°C. 100 µl of the PCR reactions contained 10 µl 10 x Vent buffer, 1 µl Vent DNA polymerase, 5 µl DMSO, 0.4 mM dNTPs, 1 µM outer primers, 50 nM randomized primer, 25 nM megaprimers (if applicable) and 25 nM internal assembly primer (if applicable). Table 7 lists the primers used for the assembly of the CDRs of the respective libraries. The assembly PCR reactions yielded single DNA species of the expected size, which was purified by PCR purification kit (Qiagen). Fragments containing CDR1 were digested with BsaI and fragments containing CDR2 with BpiI. Digestion with these two Type IIS restriction enzymes resulted in complementary sticky ends of 4 base pairs. Digested DNA was purified by PCR purification kit and CDR1-CDR2 pairs belonging to the respective library were ligated with T4 ligase. Separation of the ligation product by DNA gel revealed almost complete ligation of the fragments. The three ligation products consisting of the respective CDR1-CDR2 pairs of the three libraries were purified by gel extraction and yielded around 800 ng of DNA. The purified ligation product served as template to amplify the CDR1-CDR2 region using primer pairs FW1_a_b_for/Link2_a_rev, FW1_a_b_for/Link2_b_rev and FW1_c_for/Link2_c_rev for the concave, loop and convex library, respectively. The resulting PCR product was cleaned up by PCR purification kit and digested with BsaI. The purified CDR3 regions were digested with BpiI. The resulting compatible overhangs were ligated and the ligation product corresponding to the final assembled library was purified by gel extraction. The DNA yields were 6.8, 8.6 and 9.4 µg for the concave, loop and convex library, respectively.

3.42, 3.6 and 3.77 µg of the ligated concave, loop and convex library, respectively, were used as template for PCR amplification using primer pairs Med_FX_for/Med_FX_rev for the concave and the loop library and Long_FX_for/Long_FX_for for the convex library. The amount of ligated libraries used as template for PCR amplification represented the bottleneck of the library construction with diversities corresponding to 9 × 10¹² for each of the three sybody libraries. The resulting PCR product was cleaned up by PCR purification kit (Macherey-Nagel, Oensingen, Switzerland), digested by BspQI, and column purified again. The pRDV vector (Binz et al., 2004) was made compatible with FX cloning by amplifying the vector backbone using the primer pair RD_FX_pRDV_for/RD_FX_pRDV_rev1. The resulting PCR product was digested with BspQI, ligated with the ccdB kill cassette excised from the vector pBXCA3GH (Bukowska et al., 2015) (Addgene #47071) with the same enzyme and transformed into E. coli DB3.1. The DNA region between the stem loop and the ribosome binding site was shortened by amplifying the resulting vector with the 5’-phosphorylated primer pair pRDV_SL_for/pRDV_SL_rev and ligating the obtained PCR product. The resulting vector was called pRDV_FX5 and served as template to amplify the 5’ and 3’ nucleotide sequences required for in vitro translation of mRNA for ribosome display. This was performed by PCR amplification with primer pairs 5’_flank_for/5’_flank_rev and 3’_flank_for/tolAk_rev, respectively. The resulting PCR products were column purified, digested with BspQI and column purified again. The digested flanking regions (60 and 80 µg of the 5’ and 3’ flank, respectively) were ligated with 75 µg of the respective sybody library in a volume of 4 ml and using 80 µl T4 ligase (5 U/µl, ThermoFisher). After ligation, the DNA fragments were gel-purified and yielded 13.8, 22 and 20.5 µg ligation products corresponding to the flanked concave, loop and convex libraries, respectively. 10 µg of each ligation was amplified by PCR using primers 5’_flank_for and tolAk_2 in a 96 well plate and a total of 5 ml PCR reaction and the resulting PCR product was column purified, yielding 180, 188 and 192 for the concave, loop and convex libraries, respectively. 10 µg of each amplified library was in vitro transcribed using RiboMAX Large Scale RNA Production System (Promega) and yielded 1.5 mg of mRNA for each library.

The coding sequence of GFP was cloned using primer set GFP_FX_FV/GFP_FX_RV into FX vector pBXNH3. GFP was purified by Ni-NTA, followed by HRV 3C protease cleavage, rebinding on Ni-NTA and SEC in PBS. Chemical biotinylation of GFP was carried out in PBS using EZ-Link Sulfo-NHS-LC-Biotin (Thermo Fisher) and mass spectrometry analysis revealed that the reaction contained predominantly GFP having one biotin moiety attached. The coding sequence of MBP was cloned using primer sets MBP_FX_FW/MBP_FX_FW into FX vector pBXNH3CA (Bukowska et al., 2015), which results in a fusion protein consisting of an N-terminal His10-tag, a 3C protease cleavage site, MBP and a C-terminal Avi-tag. In order to produce Avi-tagged TM287/288, a GS-linker was first introduced into pBXNH3CA between the 3C cleavage site and the adjacent SapI site by amplifying the vector with the primer pair GS-Linker_FW (5’ phosphorylated) and GS-Linker_RV, each containing half of the GS-linker as overhang, and blunt-end ligation of the resulting PCR product. The resulting expression vector was called pBXNH3LCA. TM287/288 was cloned into pBXNH3LCA by FX cloning, which attaches a cleavable His10-tag to the N-terminus of TM287 separated by a GS-linker and an Avi-tag to the C-terminus of TM288. To produce Avi-tagged IrtAB, the coding sequence of irtAB was amplified from genomic DNA of Mycobacterium thermoresistibile DSM44167 using the primer set IrtAB_FX_FW/IrtAB_FX_RV and cloned into the expression vector pBXCA3GH (Bukowska et al., 2015). GFP, MBP-Avi, TM287/288-Avi and IrtAB-Avi were expressed in and purified from E. coli by Ni-NTA chromatography as described previously (Binz et al., 2004; Hohl et al., 2012). The enzymatic site-specific biotinylation of the Avi-tag was carried out at 4°C overnight using purified BirA in 20 mM imidazole pH 7.5, 200 mM NaCl, 10% glycerol, 10 mM magnesium acetate, 0.03% β-DDM in case of TM287/288 and IrtAB and a two-fold excess of biotin over the Avi-tag concentration. 3C protease was added as well to this reaction mixture to cleave off the His10-tag. Next day, the mixture was loaded onto Ni-NTA columns to remove the His-tag, BirA and the 3C protease. Biotinylated target proteins were eluted from the Ni-NTA column using TBS containing 30 mM imidazole and in case of TM287/288 and IrtAB 0.03% β-DDM. Finally, biotinylated target proteins were purified by SEC using a Superdex 200 Increase 10/300 GL (GE Healthcare) in TBS containing in case of TM287/288 and IrtAB 0.03% β-DDM.

Human ENT1(2–456) was expressed by transient transfection in HEK293 freestyle cells as wild-type, full length protein using a pCDNA3.1(+) base vector (Invitrogen) and synthesized, codon optimized genes cloned by Genewiz. The protein was designed as N-terminal fusion of His-GFP-3C-Avi with His being a 10-fold repeat of histidine, GFP is the enhanced green fluorescent protein, 3C is the 3C precision protease cleavage site (LEVLFQGP) and Avi the sequence corresponding to the Avi-tag (GLNDIFEAQKIEWHE). Biotinylation was performed in vivo during protein production by co-transfection of cells with two different pCDNA3.1(+) plasmids coding for the ENT1 construct and E.coli BirA ligase and by supplementing the medium with 50 µM biotin during fermentation. Cells were harvested in all cases 65 hr post transfection and were flash frozen at −80°C. For purification, cell pellets were thawed and resuspended in solubilization buffer containing 50 mM Tris-HCl pH 7.5, 300 mM NaCl and 1 Roche protein inhibitors complete tab per 50 ml of buffer in a 1 to 3 ratio (3 ml of buffer for 1 g of cells) under gentle agitation for 30 min until homogeneity. Subsequently, 1% (w/v) of lauryl maltose neopentyl glycol (LMNG) and 10 µM S-(4-Nitrobenzyl)−6-thioinosine (NBTI) were added to the suspension and incubated for 1 hr at 4°C. Supernatant was cleared by ultracentrifugation at 100’000xg using a Beckmann Ti45 rotor for 45 min and incubated overnight at 4°C with 15 ml of preconditioned TALON affinity resin under gentle stirring. Resin was collected by low speed centrifugation and washed with a total of 15 column volumes 20 mM Tris-HCl pH 7.5, 0.003% (w/v) LMNG, 20 mM imidazole, 300 mM NaCl and 10 µM NBTI several times. Using an empty pharmacia XK16 column, resin was collected and further washed with a buffer containing 20 mM Tris-HCl pH 7.5, 0.003% (w/v) LMNG, 20 mM imidazole, 300 mM NaCl, 10 µM NBTI and 15 µM dioleoyl-sn-glycero-3-phospho-L-serine (DOPS) lipid and washed a second time using the same buffer containing 40 mM imidazole. Protein was eluted at 300 mM imidazole and subjected to a desalt step on a 53 ml GE Hi Prep 26/10 desalting column using desalting buffer consisting of 20 mM Tris-HCl pH 7.5, 0.003% (w/v) LMNG, 300 mM NaCl, 10 µM NBTI and 15 µM DOPS. To remove the GFP-His tag and reduce protein glycosylation, 3C-Prescission protease was added at a concentration of 1 unit per 50 µg of protein together with 10 µg/ml of PNgaseF and 10 µg/ml Endo-alpha-N-acetylgalactosaminidase. The mixture was incubated overnight at 4°C and His-GFP tag removal was monitored by fluorescence-detection size-exclusion chromatography (FSEC) analysis. To completely remove GFP, protein was purified by an affinity purification step using a HiTrap TALON column collecting the flow-through. Subsequently, protein was concentrated using an Amicon filter unit at 30 kDa molecular weight cut-off to about 0.5–1 ml volume and further purified via size–exclusion chromatography using a Superdex 200 10/300 GL increase column equilibrated in the SEC buffer containing 20 mM Tris-HCl pH 7.5, 0.003% (w/v) LMNG, 300 mM NaCl, 10 µM NBTI and 15 µM DOPS. SEC fractions corresponding to the biotinylated ENT1 protein were concentrated on an Amicon filter unit with a cut-off of 30 kDa to a final concentration of 1.35 mg/ml and stored at −80 upon flash freezing in liquid nitrogen. Quality and biotin modification of the protein were analysed by LC-MS revealing close to complete biotinylation.

Human GlyT1 was cloned as a codon-optimized gene into a modified pOET1 base vector (Oxford Expression) by Genewiz and expressed either as a C-terminal 3C-GFP-His or C-terminal Avi-3C-GFP-His fusion in Spodoptera frugiperda (Sf9) cells. Cells were grown to 2 million cells/ml in Sf9III medium in 50 liter wave bags (Sartorius, 25 l maximal volume) and infected using 0.25–0.5% (v/v) of virus. 72 hr post infection at viabilities higher than 85%, cells were harvested by centrifugation at 3000 x g for 10 min and 4°C, washed in PBS and re-centrifuged for 20 min. Cells were filled in plastic bags and frozen by putting the bag in a −80°C freezer. Thawed biomass was further washed twice by resuspension and centrifugation for 20 min at 5000 x g in a buffer containing 50 mM Tris-HCl pH 7.5, 150 mM NaCl and 1 Roche Complete Tablet per 50 ml volume. Washed cells were resuspended for solubilization for 30 min at 4°C while stirring in a buffer containing 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 30 mM imidazole pH 7.5, 1% (w/v) LMNG, 0.1% CHS, 15 µM DOPS and 100 µM Cmpd1, a more soluble analogon of the glycine transporter 1 reuptake inhibitor Bitopertin. The suspension was centrifuged at 100’000xg in a Ti45 rotor for 20 min at 4°C to collect supernatant. Protein was purified by batch purification using TALON affinity resin (GE Healthcare), incubated with resin for 16 hr under stirring at 300 rpm and then centrifuged at 500xg for 2 min in a 50 ml Falcon tube. Four wash steps with a buffer containing 50 mM Tris-HCl, 150 mM NaCl, 30 mM imidazole pH 7.5, 0.05% (w/v) LMNG, 0.005% (w/v) CHS, 15 µM DOPS, 50 µM Cmpd1 and 1 Roche Complete Tablet per 50 ml volume were followed by loading the resin into a XK26 column (GE Healthcare), washed again with four column volumes to finally elute the protein with the same buffer that contained 300 mM imidazole. The Avi-3C-GFP-His but not the 3C-GFP-His protein was treated with HRV-3C protease (Novagen) and in-house produced PNGase F (F. meningosepticum) to cleave the GFP-His tag and trim existing glycosylations. Subsequently, the Avi-tagged GlyT1 was desalted into a buffer optimal for enzymatic biotinylation consisting of 20 mM bicine pH 8.3, 150 mM potassium-glutamate pH 7.5, 0.05% (w/v) LMNG, 0.005% CHS, 15 µM DOPS and 50 µM Cmpd1 to remove imidazole and subjected to another round of TALON affinity purification to remove the cleaved GFP-His tag in the same buffer. The flow through containing the GlyT1 protein was concentrated to 1.1 mg/ml concentration with an Amicon Ultra four filter unit (Millipore) with a molecular weight limit of 30 kDa for complete biotinylation using the BirA-500 biotinylation kit (Avidity) according to the protocol. Biotinylation was monitored by liquid-chromatography-coupled mass spectroscopy. Both the Avi fusion as well as the 3C-GFP-His fusion protein were further purified via size-exclusion chromatography using a Superdex 200 10/300 GL increase column and a buffer containing 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.05% (w/v) LMNG, 0.005% (w/v) CHS, 15 µM DOPS and 50 µM Cmpd1.

To display sybody libraries on ribosomes, 10 µl of the PUREfrexSS translation mix was prepared (GeneFrontier Corporation, Kashiwa, Japan). The kit components were mixed to a total volume of 9 µl and incubated at 37°C for 5 min. 1 µl of 2 µM library RNA was added to the translation mix and incubated at 37°C for 30 min. The ribosomal complexes were diluted in 100 µl ice cold WTB buffer (50 mM Tris-acetate pH7.5, 150 mM NaCl, 50 mM magnesium acetate) supplemented with 0.05% Tween 20, 0.5% BSA and 5 mg/ml Heparin. 10 µl Dynabeads MyOne Streptavidin T1 (Life Technologies) were washed 3 times with 500 µl WTB and blocked with 500 µl WTB 0.5% BSA for 1 hr followed by 3 washes 500 µl of WTB-T-BSA (0.05% Tween 20, 0.5% BSA). The magnetic beads were coated in 100 µl WTB-T-BSA containing 50 nM biotinylated MBP for 1 hr followed by 3 washes of 500 µl WTB-T-BSA. The ribosomal complexes were incubated with the beads for 20 min followed by 3 washes of 500 µl WTB-T. During the last wash step, the beads were placed in a fresh tube. The RNA was eluted by resuspending the beads in 100 µl TBS supplemented with 50 mM EDTA pH 8.0 and 100 µg/ml yeast RNA and incubated for 10 min at room temperature. The eluted RNA was purified using the RNeasy micro kit (Qiagen) and eluted in 14 µl RNase-free water. Reverse transcription was performed by mixing 14 µl of the eluted RNA with 2 µl of RT_Primer at 100 µM and 4 µl of 10 mM dNTPs. The mixture was heated to 65°C for 5 min, and then cooled on ice. Using this mixture, a 40 µl RT reaction was assembled according to the manual (Affinity Script, Agilent) and incubated 1 hr at 37°C, followed by 5 min at 95°C. The cDNA was purified using the PCR purification kit (Macherey Nagel) and eluted in 30 µl elution buffer. 25 µl of the purified cDNA was amplified by PCR using Q5 High-Fidelity DNA Polymerase (NEB) and primers Medium_ORF_for and Medium_ORF_rev for the concave and loop library, and Long_ORF_for and Long_ORF_rev for the convex library, respectively. The PCR product was purified via gel and used as template in an assembly PCR to add the flanking regions for in vitro transcription using megaprimers. Megaprimers to flank the concave and loop sybodies were obtained by amplifying pRDV_FX5 containing the non-randomized loop sybody using primer pairs 5'_flank_for/Medium_ORF_5’_rev and Medium_ORF_3’_for/tolAk_rev. Megaprimers to flank the convex sybodies were obtained by amplifying pRDV_FX5 containing the non-randomized convex sybody using primer pairs 5'_flank_for/Long_ORF_5’_rev and Long_ORF_3’_for/tolAk_rev. Flanking was performed by assembly PCR using 200 ng sybody pool obtained from RT-PCR, 120 ng of 5’-flank, 360 ng of 3’-flank and 5 µM of outer primers 5'_flank_for and tolAk_2 in a volume of 100 µl. The resulting PCR product was separated on gel, purified and used as input material for 10 µl reaction of the RiboMAX Large Scale RNA Production System (Promega). The resulting RNA was purified using the RNeasy kit (Qiagen) and used as input RNA of the next round. The second round was performed according to the first round. In the third round, the PCR product of the amplified cDNA was amplified using primers Med_FX_for/Med_FX_rev (concave and loop library) or Long_FX_for/Long_FX_for (convex library) to add FX overhangs to the DNA and subsequently cloned into the pSb_init vector by FX cloning for expression and ELISA.

For clarity of discussion, we describe here only the final form of the selection protocol and do not specify how we gradually evolved the selection method as outlined in Figure 1—figure supplement 6. One round of ribosome display was performed as in the MBP section with the following exceptions. Tween 20 was replaced by 0.1% β-DDM for selections against TM287/288(E517A), 0.1% β-DDM and 0.005% CHS against ENT1, or by 0.05% LMNG and 0.005% CHS against GlyT1. Selections against TM287/288(E517A), ENT1 and GlyT1 were carried out in the presence of 1 mM ATP, 10 µM NBTI and 50 nM Cmpd1, respectively. Solution panning was performed by incubating ribosomal complexes and 50 nM biotinylated target protein for 30 min prior to the pulldown via streptavidin coated magnetic beads (Dynabeads MyOne Streptavidin T1). The cDNA was amplified using GoTaq G2 DNA Polymerase (Promega) and primers Med_FX_for/Med_FX_rev for concave/loop sybodies or Long_FX_for/Long_FX_rv for the convex sybodies. The resulting PCR product was cloned into the phagemid vector pDX_init. To this end, amplified sybody pools and pDX_init were digested with BspQI, gel-purified and 500 ng sybody insert and 1 µg pDX_init backbone were ligated in 50 µl using 5 units of T4 DNA Ligase (Thermo Scientific, Reinach, Switzerland). The ligation reaction was mixed with 350 µl electrocompetent E.coli SS320 on ice. After electroporation (Bio Rad Gene Pulser, 2.5 kV, 200 Ω, 25pF), the cells were immediately resuspended in 25 ml SOC and shaken at 37°C for 30 min for recovery. Subsequently, the cells were diluted in 250 ml 2YT, 2% glucose, 100 µg/ml ampicillin and grown overnight shaking at 37°C. For phage production, this phagemid-containing E.coli SS320 overnight culture was inoculated 1:50 in 50 ml 2YT, 2% glucose, 100 µg/ml ampicillin and grown to OD₆₀₀ of 0.5. 10 ml of this culture was superinfected with 3 × 10¹¹ plaque forming units M13KO7 Helper Phage at 37°C without shaking for 30 min. Cells were collected by centrifugation and resuspended in 50 ml 2YT, 100 µg/ml ampicillin, 25 µg/ml kanamycin and incubated at 37°C shaking overnight for phage production. Next day, cells were pelleted by centrifugation and 40 ml of the culture supernatant were mixed with 10 ml 20% PEG6000 (v/v), 2.5 M NaCl and incubated on ice for 30 min to precipitate the phages, which were subsequently pelleted by centrifugation. The pellet was resuspended in phosphate-buffered saline (PBS) and cleared twice by centrifugation in a tabletop centrifuge at full speed. Phage concentration was determined by UV-Vis spectroscopy. The first round of phage display was performed in a Maxi-Sorp plate (Nunc) using 48 wells per target. The plate was coated overnight with 100 µl per well of 60 nM neutravidin in TBS. Next day, the plate was washed three times with TBS and blocked with TBS containing 0.5% BSA for 1 hr. The phages were diluted to 10¹² phages per ml in ice cold TBS-D-BSA (containing 0.1% β-DDM for TM287/288(E517A), 0.1% β-DDM and 0.005% CHS and ENT1 or 0.05% LMNG and 0.005% CHS for GlyT1, and 0.5% BSA for all targets). 4.8 ml of this phage stock was incubated in solution for 20 min with 50 nM biotinylated target protein (in the presence of 1 mM ATP +2 mM MgCl₂ for TM287/288(E517A), 10 µM NBTI for ENT1 or 50 nM Cmpd1 for GlyT1). The plate was prepared by washing three times with ice cold TBS-D-BSA and 100 µl of the phage/target mix was added per well (4.8 ml in total added to 48 wells) and incubated for 10 min at 4°C. The plate was subsequently washed 3 times with 250 µl/well ice cold TBS-D (devoid of ligands and BSA). Phages were eluted by adding 100 µl TBS with 0.25 mg/ml trypsin per well and incubation at room temperature for 30 min (total elution volume of 4.8 ml). Trypsin was inhibited by adding 0.125 mg/ml 4-(2-Aminoethyl)benzenesulfonyl fluoride (AEBSF) to the elution. For infection, a culture of E.coli SS320 was grown in 2YT to an OD₆₀₀ of 0.5. 4.8 ml of eluted phages was added to 50 ml of the culture and incubated at 37°C without shaking for 30 min. The cells were then diluted 1:10 in 500 ml 2YT, 2% glucose, 100 µg/ml ampicillin and grown overnight shaking at 37°C resulting in a preculture for phage production for the next round. The second round of phage display was performed according to the first one except that 10 μl magnetic beads (Dynabeads MyOne Streptavidin C1) were used to pull down target-phage complexes. The total working volume was reduced from 4.8 ml to 100 μl and beads were washed three times with 500 μl TBS-D. Phages were eluted in 100 μl and used to infect 1.4 ml E.coli SS320. After infection, cells were diluted 1:10 in 15 ml 2YT, 2% glucose, 100 µg/ml ampicillin and grown overnight shaking at 37°C. The overnight culture was used to purify the phagemids (MiniPrep, Qiagen) containing the selected sybodies. Sybody sequences were subcloned into pSb_init using FX cloning and transformed into E. coli MC1061 for ELISA analysis and protein purification.

Single sybody clones were picked and expressed in 1 ml terrific broth containing 25 µg/ml chloramphenicol in a 2 ml 96 deep well plate. After expression, the cells were pelleted by centrifugation and resuspended in 50 µl B-PER II for lysis. The lysate was diluted with 950 µl TBS and centrifuged to pellet cell debris. ELISAs were carried out in Maxi-Sorp plates (Nunc) coated overnight with 100 µl/well of 5 µg/ml Protein A in TBS. The plate was washed 3 times with 250 µl TBS and blocked with 250 µl TBS-BSA. All washing steps were performed using 3 times 250 µl TBS containing detergent (0.05% Tween-20 for MBP; 0.03% β-DDM for TM287/288; 0.03% β-DDM/0.003% CHS for ENT1; 0.005% LMNG/0,0005% CHS for GlyT1) between all incubation steps, which were carried out in 100 µl TBS-D-BSA for 20 min. These steps were anti-myc antibody 1:2000 (Sigma Aldrich, Buchs, Switzerland, M4439) followed by five fold diluted sybody lysate, then 50 nM of the biotinylated target protein or biotinylated control protein and finally streptavidin-HRP 1:5000 (Sigma Aldrich, S2438) (Figure 1). The ELISA was developed by adding 100 µl of 0.1 mg/ml TMB (Sigma Aldrich 860336) in 50 mM Na₂HPO₄, 25 mM citric acid and 0.006% H₂O₂. ELISA signals were measured at an absorbance of 650 nm.

With the term ‘enrichment’, we refer to the experimentally determined fold excess of polynucleotides eluted from a selection round against a target of choice versus an analogous selection round against another immobilized protein, which was not used for selections in preceding rounds. In order to determine enrichments, qPCR was performed in a 10 µl reaction containing SYBR select Master Mix (Thermo Fischer Scientific), 300 nM of each primer, 5% DMSO and 2 µl cDNA (ribosome display) or phages (phage display) diluted 10 fold in H₂O. Standard curves for each primer pair were determined using a dilution series of the phagemid pDX_init or a PCR product corresponding to the sequence of the cDNA obtained after ribosome display. Their initial concentration was determined by UV-Vis spectroscopy. PCR efficiencies for all primer pairs were between 95 and 98%. Cycling conditions were: 2 min 95°C initially for polymerase activation, followed by 10 s 95°C and 30 s 63°C for 45 cycles. The runs were performed in a 7500 fast qPCR machine (Thermo Fischer Scientific). The following primer pairs were used: qPCR_RD_5’_for in combination with qPCR_ RD_S and M_5’_rev for the concave and loop library or qPCR_ RD_L_5’_rev for the convex library to determine the amount of full length cDNA, qPCR_ RD_tolA_3’_for/qPCR_ RD_tolA_3’_rev to determine the total amount of cDNA, qPCR_PD_pDX_for/qPCR_PD_pDX_rev to determine the amount of phages and qPCR_3K1K_for/qPCR_3K1K_rev to determine the amount of 3K1K cDNA. The qPCR reactions were performed as technical triplicates.

Thermofluor was performed in a 25 µl reaction of PBS containing 100x SYPRO Orange (Life Technologies) and 0.5 mg/ml sybody. The reaction was heated with a 1% ramp from 25 to 99°C in a 7500 fast qPCR machine (Thermo Fischer Scientific) while the fluorescence intensity was measured through a ROX filter. The raw data were extracted and fitted as described previously (Ericsson et al., 2006). Two technical replicates were performed for each protein and one representative dataset was fitted.

Binding affinities were determined using surface plasmon resonance (SPR). MBP binders were analyzed using a Biacore X100 machine (GE healthcare). 570 response units (RU) of biotinylated MBP were immobilized on a streptavidin coated SPR chip (Sensor Chip SA). Sybodies Sb_MBP#1–3 were purified in TBS, 0.05% Tween-20 as described above. SPR measurements were carried out as technical triplicates for each sybody concentration in the same buffer. Affinities of Sb_MBP#1 were in addition determined in the same buffer supplemented with increasing maltose (4-O-α-D-Glucopyranosyl-D-glucose) concentrations (0, 5, 10, 25, 50 and 100 µM). In these analyses, 860 response units (RU) of biotinylated MBP were immobilized, and traces for each sybody concentration was measured once. SPR data were fitted with a 1:1 interaction model using the Biacore X100 evaluation software and further analyzed by plotting sybody affinity ratios determined in the presence (K_D’) and absence (K_D) of maltose against the maltose concentration and fitting the data with the following equation:

B corresponds to the binding affinity of maltose (K_D,maltose) and α corresponds to the allosteric constant.

Affinities of sybodies directed against TM287/288 were measured using a ProteOn XPR36 Protein Interaction Array System (Biorad). Biotinylated TM287/288 and TM287/288(E517A) mutant were immobilized on a ProteOn NLC Sensor Chip at a density of 1500 RU. Sybodies expressed in pSb_init were SEC-purified in TBS, and SPR analysis was carried out in the same buffer containing 0.015% β-DDM and either 1 mM MgCl₂ or 1 mM MgCl₂ +0.5 mM ATP, to measure binding affinities in the presence or absence of ATP. Traces for each sybody concentration were recorded once and the data were fitted with a 1:1 interaction model using the BioRad Proteon Analysis Software.

All ENT1 and GlyT1 SPR experiments were performed on a Biacore T200 (GE Healthcare, Uppsala, Sweden) instrument at 18°C in running buffers containing either 20 mM HEPES pH 7.5, 150 mM NaCl, 0.001% (w/v) LMNG, 1.25 µM DOPS (ENT1) or 20 mM Citrate pH 6.4, 150 mM NaCl, 0.004% (w/v) LMNG (GlyT1) at 50 µl/min flow rate. Running buffers were freshly prepared, filtered with ExpressPlus steritop filters with 0.22 µm cut off (Millipore, Billerica, MA, USA) and degassed prior the SPR analysis. Biotinylated ENT1 or GlyT1 were captured on streptavidin pre-coated SA sensors (GE Healthcare BR-1000–32). First, streptavidin sensors were conditioned with 3 consecutive 1 min injections of high salt solution in sodium hydroxide (50 mM NaOH, 1 M NaCl). Next, biotinylated protein samples were applied to a streptavidin sensor surface for protein immobilization levels of about 1000 RU and 300 RU for ENT1 and GlyT1, respectively. Finally, free biotin solution (10 µM in running buffer) was injected once (1 × 1 min) over the sensor surface to block remaining streptavidin binding sites. Dose-response experiments were performed at sybody concentrations up to 475 nM (ENT1) and up to 2 µM (GlyT1). All monitored resonance signals were single referenced, i. e. signals monitored on the binding active channel were subtracted with signals from a reference channel. SPR measurements on ENT1 and GlyT1 were carried out using two independently coated SPR chips and were highly reproducible. One of the two datasets is shown. Data fitting was performed with a 1:1 interaction model using the Biacore T200 Evaluation (v2.0) software.

The Octet RED96 System (FortéBio, Pall Inc.) uses disposable sensors with an optical coating layer immobilized with streptavidin at the tip of the sensor. Sensors were decorated with biotinylated MBP to reach a stable baseline, arbitrarily set to 0 nm (Figure 2—figure supplement 4). Sensors were dipped in a well containing 500 nM Sb_MBP#1 which leads to the formation of the Sybody-MBP complex. Sensors containing the complex were sequentially dipped in a row of wells containing 500 nM Sb_MBP#1 and increasing concentrations of maltose (0.1, 1, 10, 100, 1000 µM). The reversibility of the competition was shown by decreasing maltose concentrations (1000, 100, 10, 1, 0.1, 0 µM) again in the presence of 500 nM Sybody Sb_MBP#1. The experiments were done in technical duplicates.

Ligand binding assays were performed in a buffer containing either 50 mM Tris-HCl pH 7.5, 150 nM NaCl and 0.004% (w/v) LMNG or 20 mM Citrate pH 6.4, 150 mM NaCl and 0.004% (w/v) LMNG, respectively. For each analysis, 140 µl protein solution (7 nM for ENT1 and 10 nM for GlyT1) was added to each of 12 wells of a 96-well Eppendorf PCR plate at 4°C and incubated subsequently for 10 min with a temperature gradient from 30–60°C (ENT1) or 23–53°C (GlyT1) across twelve wells in a Techne Prime Elite thermocycler. Subsequently, the plate was centrifuged at 2250 x g for 3 min at 4°C and 135 µl of protein solution transferred to a 96-well Optiplate (Perkin Elmer) preloaded with 15 µl Copper SPA beads (20 mg/ml) per well (PerkinElmer) to obtain a final bead concentration of 0.3 mg/well. After 15 min of incubation at 4°C and 1000 rpm on a BioShake iQ, a final concentration of 6 nM tritiated [³H]-NBTI (Perkin-Elmer) or [³H]-Org24598 compound (1.2mCi/ml specific activity, 50 µl of a 24 nM stock solution) was added and incubated for 45 min at 4°C and 1000 rpm on a Bioshake iQ. Scintillation analysis was performed using a TopCount Microplate Scintillation Counter. Ent1 measurements were performed with three technical replicates for each temperature, whereas single measurement points for each temperature were determined for GlyT1. Apparent melting temperatures (T_ms) were determined in GraphPad Prism 6.07 using a non-linear fit to a Boltzmann sigmoidal function.

ATPase activity was measured as described previously (Hohl et al., 2014) in TBS containing 0.03% β-DDM and 10 mM MgSO₄ at increasing concentrations of sybody Sb_TM#26. ATP concentration was 50 µM, TM287/288 concentration was 25 nM, assay temperature was 25°C and incubation time was 20 min. Each data point was determined as technical triplicate. The data were fitted to a hyperbolic decay curve with the following function (SigmaPlot):

in which y corresponds to the ATPase activity at the respective sybody concentration divided by the ATPase activity in the absence of inhibitor normalized to 100%, IC₅₀ corresponds to the sybody concentration for half-maximal inhibition, y₀ corresponds to the residual activity at infinite sybody concentration, a corresponds to the maximal degree of inhibition, and x corresponds to the sybody concentration.

The sybody-MBP structures have been deposited at the Protein Data Bank under the accession codes 5M13, 5M14 and 5M15. Vectors have been deposited at Addgene. All other data generated or analyzed during this study are included in this published article.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We thank all members of the Seeger lab for stimulating discussions. We acknowledge Beat Blattmann and Céline Stutz-Ducommun of the Protein Crystallization Center UZH for performing the crystallization screening, and the staff of the SLS beamlines X06SA and X06DA for their support during X-ray data collection. We thank Jean-Marie Vonach, Marcello Foggetta, Martin Weber, Christian Miscenic and Georg Schmid for technical help during expression and fermentation, Johannes Ernie and Jörg Hörnschemeyer for mass spectrometry investigations and Alain Gast, Natalie Grozinger, David Bissegger and Luca Minissale for help with assay development and performance. We are grateful to Ralf Thoma, Armin Ruf and Michael Hennig for valuable discussions and organizational help in the early phase of the project. The Institute of Medical Microbiology and the University of Zurich are acknowledged for financial support. This work was funded by a grant of the Commission for Technology and Innovation CTI (16003.1 PFLS-LS) and a SNF Professorship of the Swiss National Science Foundation (PP00P3_144823, to MAS).

© 2018, Zimmermann et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.