Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Jan 23.
Published in final edited form as: ACS Chem Biol. 2022 Dec 16;18(1):70–80. doi: 10.1021/acschembio.2c00683

Engineered Glycan-Binding Proteins for Recognition of the TF Antigen and Structurally-Related Disaccharides

Elizabeth M Ward 1,2, Cristina Y Zamora 1,3, Nathaniel S Schocker 1,3, Soumi Ghosh 1,3, Megan E Kizer 1,3, Barbara Imperiali 1,3,*
PMCID: PMC9868099  NIHMSID: NIHMS1859207  PMID: 36525666

Abstract

Glycan-binding proteins (GBPs) are widely used reagents for basic research and clinical applications. These reagents allow for the identification and manipulation of glycan determinants without specialized equipment or time-consuming experimental methods. Existing GBPs, mainly antibodies and lectins, are limited and discovery or creation of reagents with novel specificities is time consuming and difficult. Here we detail the generation of GBPs from a small, hyper-thermostable DNA binding protein by directed evolution. Yeast surface display of a variable library of rcSso7d proteins was screened to find variants with selectivity toward the cancer-associated glycan Galβ1–3GalNAcα, or Thomsen-Friedenreich antigen and various relevant disaccharides. Characterization of these proteins show them to have specificities and affinities on par with currently available lectins. The proteins can be readily functionalized with fluorophores or biotin using sortase-mediated ligation to create reagents that prove useful for glycoprotein blotting and cell staining applications. The presented methods for the development of GBPs toward specific saccharides of interest will have great impact on both biomedical and glycobiological research.

Complex carbohydrate polymers that can be secreted or appended to other biomolecules are biosynthesized in all domains of life. These oligosaccharides, or glycans, are directly involved in diverse processes such as protein stabilization, cell adhesion, and host-cell interactions.1 Cell-surface glycans can also serve as important disease markers, as glycosylation is often found to be aberrant in malignant tissues.2 Despite the importance of these glycans, there are significant difficulties in their study. This is due in part to the massive diversity of glycans, both at the monosaccharide level and at the glycosidic linkage level. Another major barrier to study of carbohydrates is that they are not encoded in the genome, so genetic techniques to manipulate and modify them cannot be employed. Many glycan structure analysis methods have disadvantages such as large sample requirements, expensive equipment, and difficult, time-consuming protocols.

Glycan-binding proteins (GBPs) are often employed as reagents to bind specific glycan epitopes and allow for structure determination and glycoconjugate manipulation without the extensive processing or specialized equipment that other techniques require. Because of this, well-characterized GBPs are crucial for glycobiological research. The predominant GBP reagents currently in use consist of fungal, plant, and animal derived lectins or monoclonal antibodies. Lectins are popular as they recognize monosaccharide determinants, can specifically bind different glycosidic linkages, and are relatively cheap to isolate or purify. These proteins are useful; however the specificities of known lectins do not cover the large diversity of glycans found across biology. Antibodies against sugars can be directly elicited in animals by inoculation with free glycans, glycan-protein conjugates, or whole cells.3 Anti-glycan antibody elicitation in hosts via immunization is unpredictable due to the poor immunogenicity of carbohydrates and similarities to the host species’ own glycoforms.4 Antibodies are also expensive to produce, can have poor stability, and require large quantities of difficult-to-obtain glycan for antibody generation.

Due to the drawbacks to antibody elicitation and lectin discovery, there have been numerous attempts to generate novel GBPs with higher affinities or different specificities by engineering existing GBPs.5, 6 These attempts have employed techniques such as mutation of a single amino acid to increase affinity to the native ligand, or more complicated directed-evolution approaches to pivot the specificity of the GBP to structurally-distinct sugars.7–10 Recently, the type B lamprey variable lymphocyte receptors (VLRBs) have shown promise as glycan recognition reagents.11, 12 This scaffold, referred to as a “lambody”, is a leucine-rich repeat protein that is part of the lamprey adaptive immune response, and glycan-binding properties have been engineered by directed evolution of naïve libraries of VLRBs using yeast surface display, or by elicitation in lampreys.11–14 Recently, nanobody (see Figure 1) generation using a camelid immunization strategy in alpaca to elicit GBPs against the Globo-H glycan antigen has also established a complementary approach for the development of GBPs.15

Figure 1. Approximate sizes of common glycan-binding proteins.

Figure 1.

Open in a new tab

Phenobarbitol-specific IgG1 antibody (PDB: 1IGY), ConA lectin (PDB: 1VAM), TF specific variable lymphocyte receptor of lamprey or “lambody” (PDB 5UFC), anti-SARS-CoV-2 spike nanobody mNb6 (PDB: 7KKJ) and the DNA binding protein Sso7d (PDB: 1SSO) compared to TF disaccharide. Inset shows sequence of rcSso7d with residues mutated from Lys (in the native sequence) colored in red and variable residues colored in maroon, dark blue, and light blue. The Sso7d structure shows variable residues color coded to those in the sequence as well as dimensions of the binding face.

However, many of the current GBPs in use have non-ideal properties, such as large size, compulsory multimeric structure, and difficulties in expression and purification in bacterial systems. Non-antibody alternate scaffolds such as DARPins, knottins, affibodies, and anticalins have shown great utility in the development of proteins for molecular recognition that are free from the drawbacks of antibodies: large size, multiple polypeptide chains, glycosylation sites, and difficulties in recombinant protein expression and purification.16, 17 Despite these advantages, none of these scaffolds have been applied to glycan recognition. Another alternate scaffold is the DNA-binding protein Sso7d of the hyperthermophilic archaea Sulfolobus solfataricus. This protein, at only 63 residues (Fig. 1), has a very high melting temperature even after mutagenesis of the binding site and surrounding residues, can be produced in E. coli, and has been evolved to bind small molecules, peptides, and proteins using yeast surface display (YSD).18 Due to these properties, evolved Sso7d variants are promising affinity reagents relative to antibodies with respect to both cost of production and protein stability.19, 20

Due to these favorable characteristics, Sso7d was selected as a scaffold for development of new GBPs. In addition to having much smaller dimensions than other GBPs and a monomeric structure, Sso7d also shares many topological features found in native GBPs, further supporting its selection as a potential glycan-binding scaffold. First, the binding face of the Sso7d protein is a flat, solvent exposed surface comprising three β-sheets and shallow, solvent-exposed binding surfaces are common in lectins.21 Second, the evolved Sso7d protein can be highly enriched in aromatic amino acids in its binding face without a concomitant loss in thermostability, expression level, or solubility.22 This is important as aromatic amino acids are often enriched in carbohydrate-protein interactions, as they afford opportunities for CH-π interactions between the electron-rich π-systems and the polarized C-H bonds of carbohydrates.23–25

One carbohydrate epitope that has been the subject of several GBP engineering campaigns is the Thomsen-Friedenreich antigen, also known as the T or TF antigen.7, 11, 26, 27 The TF antigen is the core 1 structure of O-linked mucin-type glycans, consisting of the disaccharide Galβ1–3GalNAcα linked to Ser or Thr residues. In healthy tissue, core 1 structures are further modified by elaboration with sialic acids, sulfates, or other glycans, However, the motif is often unsialylated in cancerous and pre-cancerous tissue due to abnormal expression of glycosyl transferases and glycosidases in tumor cells. Thus, the epitope is a tumor-associated carbohydrate antigen found in 90% of carcinomas, making it an attractive target for cancer diagnostics and immunotherapies.28, 29

In this work, we report the screening of an Sso7d YSD library for binding to the TF antigen. Our method generated a TF-binding GBP with binding affinities and specificities that rival commercial lectins. The evolved GBPs were site-specifically labeled and utilized in glycoprotein blotting and cancer cell labeling, demonstrating their utility as potential reagent for the study of glycans in their native environments.

Results and Discussion

Enrichment of Glycan-Binding Variants from Naïve Library

The Sso7d library used in this study is a 1.4 × 109 member YSD library of a reduced charge variant of Sso7d (rcSso7d) created by Traxlmayr and coworkers (Fig. 1).22 Briefly, this combinatorial library features nine variable amino acids in the β-sheet binding-face of the protein (paratope), randomly mutated to all amino acids except Pro and Cys. The rcSso7d constructs comprised an N-terminal genetic fusion to the yeast surface protein Aga2P containing N-terminal HA and C-terminal Myc tags to allow for quantification of expression of full-length proteins on the yeast surface. Yeast displaying Sso7d variants capable of binding an antigen of interest were isolated from the library manually using glycan-functionalized magnetic beads or by fluorescence-activated cell sorting (FACS) with fluorescently labeled glycans (Fig. 2A). It is important that the initial steps in the selection system mimic the multimeric display of glycoconjugates to increase the opportunity to capture the low affinity variants with glycan-binding characteristics. Carbohydrate-protein interactions tend to display KD values in the μM-mM range, and in biological contexts GBPs will often oligomerize to form multivalent interactions with the dense glycans of the cell surface.30, 31 This increased binding strength, or avidity, can be orders of magnitude stronger than a monovalent interaction. The yeast surface display platform allows for multivalent display of the Sso7d because many copies of the protein are densely presented on the yeast cell surface. Multivalent ligands were used for selections, consisting of either high avidity magnetic beads functionalized with the glycan of interest or commercially available polyacrylamide-based fluorescent glycopolymers (Fig. 2B).32 The combination of multivalent display and multivalent ligand presentation allowed for low affinity Sso7d-carbohydrate interactions to be captured.

Figure 2. Enrichment of TF binding variants from naïve library.

Figure 2.

Open in a new tab

A) Cartoon of YSD construct. B) Structures of TF glycosides utilized in magnetic bead sorting (TF-S-NH2) and in FACS sorts (TF-PAA-FITC). TF-PAA-FITC contains 20% disaccharide conjugated by weight. C) Dot plot and bar graph showing enrichment of cells binding to the TF-PAA-FITC polymer (measured in median fluorescence intensity or MFI, y-axis) over the course of the three selections. Purple boxes labeled “binders” indicate cells with FITC fluorescence over PAA-FITC control, and bar graph shows median y-axis fluorescence intensity of the dot plots, n=1.

Selections with the large naïve library began with five rounds of enrichment using tosyl-functionalized magnetic Dynabeads conjugated to a monomeric TF glycan (TF-S-NH2, Fig. 2B). To deplete the library of variants exhibiting nonspecific binding or linker-binding, three rounds of negative selections using magnetic beads functionalized with the acyl thioether amine (S-NH2, Fig. S10) were interspersed between the positive selections with TF-modified beads. The resulting yeast population, enriched in carbohydrate-binding variants, was carried through three rounds of selection by FACS using a commercially available polyacrylamide glycopolymer bearing TF and FITC moieties (TF-PAA-FITC, Fig 2B). After three rounds of FACS, selecting for the top 1% of binding cells, the resulting enriched population was found to exhibit a 14-fold increase in binding to TF, noted as y-median fluorescence intensity (MFI) (Fig. 2C.) This binding was glycan mediated, as no secondary reagent or non-glycosylated polyacrylamide polymer binding was observed. Sequencing analysis of 96 plasmids isolated from the final enriched population revealed multiple inserts encoding for the same variant 0.8.F (Fig. 3A). Of the nine variable amino acids in the paratope region, 0.8.F has six aromatic amino acids with two Trp residues at positions 25 and 30, and four Tyr residues at positions 21, 28, 42, and 44. Other variants present in the final population contained two to six aromatic residues, primarily Trp and Tyr. Eleven variants were tested by growing clonal cultures and all were capable of binding to TF-PAA-FITC with no binding shown to PAA-FITC alone (Fig. S1).

Figure 3. Variants isolated from affinity matured libraries exhibit selectivity to the target ligand.

Figure 3.

Open in a new tab

A) Sequences of rcSso7d parent and selected non-sugar-binding and sugar-binding variants. Dots indicate the residue is the same as the rcSso7d reference sequence. Amino acids are colored according to ClustalW coloring scheme with blue indicating aromatic residues, green indicating nonpolar residues, red indicating acidic residues, and yellow indicating polar residues. B) Structures of tested monosaccharides. C) Structures of disaccharide moieties on glycopolymers showing negligible binding to evolved GBPs. D) Structures of binding di- and trisaccharides moieties showing significant binding to evolved GBPs E-H) Mono-, di-, and trisaccharide glycopolymers with clonal yeast expressing E) 0.8.F, F) 1.3.D, and G) 2.4.I, H) unrelated variant 83H. Star designates data not available. Error bars indicating standard deviation of n=3 replicates.

Affinity Maturation of Enriched Variants and Characterization of Binding Specificity

To improve the binding affinity for carbohydrate epitopes, selected rcSso7d variants were subjected to affinity maturation. Shotgun mutagenesis via error-prone PCR was performed on variants enriched in the first rounds of selection to introduce diversity across the entire rcSso7d sequence in the insert. This process afforded library 1.0 and three rounds of selection by FACS were performed on this affinity maturated library using 8-fold less TF-PAA-FITC for increased selection stringency. As before, the library was enriched in binding variants over the course of the selections. Plasmids isolated from a subset of the enriched yeast population were sequenced, and hierarchical clustering was performed to curate representatives of variant families based on the paratope sequence (Fig. S2). Of three clusters, the largest cluster contained variants that are descendants of 0.8.F, all containing two or more mutations outside of the binding face.

Several representative sequences from each cluster were chosen for further analysis by flow cytometry to assess binding affinities and cross-reactivity by using a panel of glycopolymers containing TF-PAA-FITC and both highly similar and distinct mono-, di-, and trisaccharides (Fig. 3B–D). TF-binding variant 1.3.D, a descendent of 0.8.F, was selected as it showed the highest selectivity for TF relative to the other tested variants (Fig. 3F). Variant 1.3.D has four mutations compared to 0.8.F: Q8R, K48T, K52E, and E53K all outside of the β-sheet binding face (Fig. 3A). Variant 1.3.D showed selectivity toward TF over the monosaccharide components Galβ and GalNAcα, differentiated between β vs α linkages of the Gal to GalNAcα glycoside bond (see Fig. 3D; core 8 O-glycan Galα1–3GalNAcα), reversed directionality of the monosaccharide components (see Fig. 3C:Adi, GalNAcα-Galβ), 1,3 vs 1,4 linkages between highly similar monosaccharides (see LacNAc, Galβ1–4GlcNAcβ), and the inclusion of a charged sulfate group (sulfo-TF). The variant also differentiated between a terminal and internal TF motif as seen with the H3 trisaccharide (Fucα1–2Galβ1–3GalNAcα).

Some off-target binding to other disaccharides was observed, including Lewis c (Lec) antigen bearing a single stereochemical difference of GlcNAc vs GalNAc (see Fig. 3D; Lec, Galβ1–3GlcNAcβ) and the Core 5 O-glycan (GalNAcα1–3GalNAcα) containing an additional N-acetyl and an α-glycosidic linkage. Surprisingly, binding was also observed to more structurally distinct glycans like the negatively charged polysialic acid disaccharide (Neu5Acα2–8Neu5Acα), or chitobiose (GlcNAcβ1–4GlcNAcβ) which contains an additional N-acetyl group, two stereochemical differences, and a 1,4 linkage. This broad glycan-binding profile is not unlike many naturally-occurring lectins.33 For example, the TF-binding lectin, Jacalin, predominantly binds TF but additionally binds C3-substituted GalNAcα epitopes containing residues such as GalNAc, GlcNAc, Gal, or more extended oligosaccharides with an α- or β-linkage.33

Compared to variant 0.8.F (Figure 3E), 1.3.D gained affinity for TF and the other binding glycans. We then hypothesized that further affinity maturation and selection could potentially afford higher affinity variants, and subjected variant 1.3.D to an additional round of error-prone PCR to generate library 2.0. During the course of these experimental efforts, TF-PAA-FITC was depleted, and supply chain issues made it impossible to obtain additional polymer. To circumvent this problem, the glycopolymer with the highest degree of similarity that showed strong binding behavior, Lec (Galβ1–3GlcNAcβ), was chosen for further selections to validate the affinity maturation hypothesis. Four rounds of selection with Lec-PAA-FITC were performed with decreasing concentrations of polymer at each round.

From this library, variant 2.4.I was identified as having a high preference for Lec over every other glycopolymer (Figure 3G). This variant has an additional three mutations compared to 1.3.D: W25R in the binding face, and D15A and D49A outside the binding face. The overall binding capacity (assessed from the MFI) for GalNAcα, Core 8, Core 5, H3 trisaccharide, and Lec increased for 2.4.I, while chitobiose, and polysialic acid decreased in MFI. Unfortunately, TF-PAA-FITC was not available for testing. However, it is expected to be bound by variant 2.4.I to some degree due to the observed binding to the H3 trisaccharide containing a TF motif.

An rcSso7d variant that weakly binds biotin, called 83H, was also evaluated with the glycopolymer panel to act as a negative control. The paratope region of this variant is distinct from the sugar-binding variants, as it has two acidic Asp residues while the sugar-binding variants contain 1–3 basic Arg residues (Fig. 3A). The 83H variant also has only two aromatic residues compared to the 5–6 in the sugar-binding variants, and importantly it shows negligible binding to PAA alone (Fig. 3H). Overall, the directed evolution of rcSso7d-based GPBs was successful, enriching in GBPs that recognize TF and several similar disaccharides. The affinity maturation process successfully generated GBPs with increased binding capacities as measured as yeast surface fusions.

Molecular modeling was applied to provide insight into the protein determinants that could be involved in disaccharide binding (Figure S3). The HADDOCK webserver was used to dock TF, Lec, and Core 5 disaccharides onto an AlphaFold model of variant 2.4.I.34, 35 All three disaccharides are predicted to fit into the aromatic groove made up of Tyr28, Trp30, and Tyr44 and make both hydrogen-bonding and CH-π interactions with surrounding residues as are commonly seen in protein-carbohydrate interactions.21 Based on these models, Tyr44 is in proximity to form CH-π interactions, while Arg25 and Trp30 form hydrogen-bonding interactions with the docked disaccharides. Notably, Arg25 is a mutation gained during affinity maturation of 1.3.D and based on these models could, in part, explain the increased binding affinity gained by 2.4.I.

rcSso7d-based GBPs Exhibit Binding Characteristics that Rival Commercial Lectins

To characterize the binding affinities of GBP 2.4.I isolated from the directed evolution campaigns, an insert encoding for 2.4.I was cloned into an expression vector containing a C-terminal hexahistidine tag (His6) and purified. The functional affinity binding constants, presented as an apparent KD (Kapp), were experimentally determined using bio-layer interferometry (BLI) utilizing the Lec-PAA-FITC and Core 5-PAA-FITC glycopolymers used in flow-based selections and screenings (Fig. 4A–B). Although the resulting values were not 1:1 stoichiometry KD measurements as the interacting binding partners were both multivalent, these apparent KD values are functionally relevant as they mimic the avidity context of cell-surface glycans being displayed in high density. The functional affinity of 2.4.I measured for Lec and Core 5 correlate with the YSD experiments, having a Kapp of 38.7 ± 1.4 nM for Lec and a higher Kapp of 53.8 ± 2.9 nM for Core 5 (Fig. 4E). The functional affinity for TF-PAA-FITC could not be determined due to unavailability of glycopolymer sample. However, we propose that the functional affinity would be in the range of Lec and Core 5 due to the sequence similarity between 1.3.D and 2.4.I, and because the affinity for the H3 trisaccharide containing the TF antigen with an additional fucose increased from 1.3.D to 2.4.I.

Figure 4. Binding affinity characterization of GBP 2.4.I using biolayer interferometry.

Figure 4.

Open in a new tab

A) BLI trace of immobilized GBP 2.4.I with 62.5–1000 nM Lec-PAA-FITC in solution. B) BLI trace of immobilized GBP 2.4.I with 31.25–1000 nM Core 5-PAA-FITC in solution. C) Structure of TF-sp-biotin. D) BLI trace of immobilized GBP 2.4.I with 25–75 μM TF-sp-biotin in solution E) Table of KD, kon, and koff value obtained from curves fit with one-phase exponential association and decay functions.

Although the functional affinity for TF was not determined with the multivalent glycopolymer, the KD value could be measured for TF using a commercially available monovalent TF sugar (TF-sp-biotin, Fig. 4C). This TF disaccharide is conjugated to biotin by a short linker (sp), with the reducing-end sugar in the closed pyranose form. GBP 2.4.I was immobilized and then dipped into a solution of TF-sp-biotin (Fig. 4D).36 The monovalent KD was determined to be 16.4 μM ± 4.6 μM (Fig. 4E). This binding affinity is in line with commercially available oligomeric lectins that typically have KD values of 1–10 μM for complex glycan epitopes, and it is much stronger than the mM KD values typically observed for oligomeric lectins with monosaccharides or for single domain GBPs with monosaccharides.3, 37, 38 This establishes that, even though the Lec glycopolymer was used for selections, TF-binding was not eliminated. These BLI data serve as a further solution-state confirmation of matured affinity of 2.4.I as compared to 1.3.D, as the TF binding affinity of variant 1.3.D was of too low affinity to be amenable to KD measurement in this way.

Sortase-Mediated Ligation Installs Handles for Reagent-Grade Glycan-Binding Proteins

To demonstrate the use of the evolved GBPs as affinity reagents, we first validated methods for protein functionalization with useful handles such as biotin and fluorophores. Sortase-mediated ligation (SML) is a powerful protein engineering method that allows for the site-specific incorporation of various biochemical and biophysical probes to the N- or C-terminus of proteins. The SrtA enzyme of S. aureus catalyzes the transpeptidase reaction between a C-terminal LPXTG motif and an N-terminal oligo-glycine sequence (Fig. 5A).39, 40

Figure 5. Functionalization of GBP 2.4.I by sortase-mediated ligation.

Figure 5.

Open in a new tab

A) Generalized scheme of SML reaction to modify any rcSso7d GBP. Sso7d variant shown in grey with C-terminal sortase motif and His6 affinity tag reacting with GGG containing peptide. Yellow star denotes desired modification, such as biotin or a fluorophore. B) Coomassie-stained gel showing representative reaction of GBP 2.4.I with biotin-functionalized GGG peptide. Lanes: (M) Molecular weight standard; (1) Unreacted 2.4.I; (2) SML reaction mix after 30 minutes at room temperature; (3) Purified labeled 2.4.I-biotin. C) Western blot probed with Streptavidin, showing presence of biotin-labeled 2.4.I from reaction shown in panel B. Lanes: (M) Molecular weight standard; (1): Purified 2.4.I-biotin.

We labeled the C-terminus of the evolved GBPs with biotin and FITC respectively by cloning an insert encoding for GBP 2.4.I followed by a C-terminal LPETGG motif and a His6 tag into an expression vector. Oligo-glycine-containing peptides appended to a biotin or FITC handle were prepared with the sequence GGGYK[K-biotin]T-amide or GGGYK[K-FITC]T-amide. Upon ligation to the peptide, the His6 tag was displaced. The reaction mixture was then passed over Ni-NTA resin to remove unreacted starting material and the His-labeled SrtA enzyme, followed by desalting to remove excess peptide (Fig. 5B). SML to generate 2.4.I-biotin or 2.4.I-FITC was confirmed by western blot analysis with streptavidin (Fig. 5C) or by fluorescence-based gel imaging respectively, validating SML as a robust means of generating reagent-level rcSso7d GBPs.

Engineered GBP 2.4.I Binds TF-Containing Mucin Glycoproteins

Mucins are a family of large, heavily O-glycosylated proteins that are abundant in mucus. Up to 80% of the mucin mass can made up of highly heterogeneous O-glycans.41 MUC2, MUC5AC, and MUC5B are all secreted gel-forming mucins expressed in the gastrointestinal tract (MUC2 and MUC5AC) and in the airway (MUC5AC and MUC5B).42 Mucin O-glycans consist of Core 1 (Galβ1–3GalNAc), 2 (Galβ1–3(GlcNAcβ1–6)GalNAc), 3 (GlcNAcβ1–3GalNAc), and 4 (GlcNAcβ1–3(GlcNAcβ1–6)GalNAc), O-glycans (Fig. 6A) that may be further elaborated with sulfate or with additional monosaccharides. Analysis of human MUC2 shows over 100 unique glycan structures with a high degree of sialylation, sulfation, and fucosylation, and primarily includes Core 3 and 4 structures.43–45 Alternatively, the gastric and airway mucins MUC5AC and MUC5B include primarily Core 1 and 2 structures, and like MUC2 are highly sialylated, sulfated, and fucosylated.46–48

Figure 6. Variant 2.4.I binds TF antigen-containing glycoproteins by dot blot.

Figure 6.

Open in a new tab

A) Structure of core O-glycans 1–4 found in mucins. B) Dot blot of 0.5–2.0 μg MUC2, MUC5AC, and MUCB blotted with 2.4.I-biotin, Jacalin-biotin, and 83H-biotin at a concentration of 1.0 μM, probed with alkaline phosphatase-conjugated streptavidin. MUC samples were spotted on the membrane, incubated with biotinylated protein, then washed in buffer with (bottom) or without (top) 800 mM D-(+)-Galactose as a competitor.

GBP 2.4.I-biotin was applied in dot blotting applications for detection of mucins MUC2, MUC5AC, and MUC5B which are known to carry the TF epitope and similar O-glycans (Fig. 6B). In these analyses, the commercially available TF-binding lectin Jacalin-biotin was used as a positive control, and the non-sugar binding biotinylated Sso7d variant 83H was used as a negative control. After spotting of the membrane with mucins and incubating with the biotinylated proteins, the blots were washed in buffer with and without 800 mM soluble D-(+)-galactose as a competitor to disrupt the protein-carbohydrate binding interactions. GBP 2.4.I-biotin was very specific for MUC2, showing no binding to MUC5AC or MUC5B at any concentration of mucin. The galactose treatment decreased the binding signal, as shown by densitometry analysis (Fig. S4) indicating that the binding is glycan dependent. The Jacalin conjugate showed strong binding to all three mucins, and the intensity of the binding signal decreased slightly upon galactose washing. There was also noted a low level of binding to MUC2 with 83H-biotin but this could not be competed away with galactose, indicating it is not glycan-mediated. Treatment of the mucin proteins with neuraminidase and O-glycosidase was attempted but the deglycosylation was incomplete, likely due to the quantity and density of glycans on the mucin proteins (Fig. S5).

Engineered GBPs can be utilized as cell-surface glycan staining reagents

To demonstrate the utility of engineered rcSso7d GBPs in flow cytometry, the evolved GBP 2.4.I-FITC was evaluated for its ability to detect cell lines in vitro. MCF7 cells, an immortalized human breast cancer epithelial cell line, were used for this purpose as they express high levels of TF antigen on the surface.49 To disrupt GBP binding, MCF7 cells were enzymatically deglycosylated and treated with D-(+)-galactose as a soluble competitor. A protein deglycosylation cocktail containing PNGase F (N-glycans), α2–3,6,8,9 neuraminidase A (sialic acids), O-glycosidase (Core 1 and 3 O-glycans), β1–4 galactosidase S (galactosides), β-N-acetylhexosamindase (GalNAc and GlcNAc) was used to remove N-glycans and more complex O-glycans (Fig. 7A). The non-sugar binding variant 83H-FITC showed no binding to MCF7 cells (Fig. 7B). Treatment with 2.4.I-FITC showed some binding to MCF7 cells, although this binding could not be completely removed with glycosidase treatment and soluble competitor (Fig. 7C). Jacalin-FITC showed a high degree of binding to the MCF7 cells, and binding could be partially diminished with glycosidase treatment and competition (Fig. 7D). Jacalin shows higher affinity binding relative to 2.4.I, however, we note that the Jacalin a tetrameric protein with four carbohydrate-binding sites. When the concentration of Jacalin-FITC is reduced four-fold compared to 2.4.I-FITC, it shows a very similar binding profile (Fig. S6).

Figure 7. Mammalian cell staining with engineered GBP 2.4.I-FITC.

Figure 7.

Open in a new tab

A) Glycan structures removed by the deglycosylation enzyme cocktail. B)-D) Flow cytometry analysis of MCF7 cells unstained and untreated (red, unstained), stained with FITC labeled protein and untreated (orange), unstained and treated with deglycosidase enzyme mix and galactose (yellow), or stained with FITC labeled protein and treated with deglycosidase enzyme mix and galactose (green). B) 8 μM 83H-FITC. C) 8 μM 2.4.I-FITC. D) 8 μM Jacalin-FITC.

Conclusions

Herein we report a method for the evolution of GBPs from an archaeal DNA-binding protein using yeast surface display. The evolved GBPs rival commercial lectins with low nM functional affinities to Lec and Core 5 disaccharides, and low μM binding affinity to TF in a 1:1 binding context. This binding affinity was achieved after two rounds of affinity maturation, which introduced mutations both inside and outside the binding face. The evolved GBPs can bind multiple carbohydrate structures, similar to commercial TF-binding lectins such as Jacalin. Unlike Jacalin, the Sso7d-based GBPs are monomeric, consist of a single polypeptide chain, and require no post-translational modifications making them easier to manipulate for directed evolution campaigns and easier to express in E. coli.50 The Sso7d-based GBPs can also be strategically labeled using sortase-mediated ligation to introduce a single handle to create reagent-grade tools for glycan study. This method eliminates the need for cysteine or lysine conjugation, which can produce heterogeneous labeling and also potentially occlude the binding region. The evolved GBP 2.4.I was shown to to stain intact mucin glycoprotein MUC2 in dot blot applications and whole cancer cells by flow cytometry.

This method can be used to elicit Sso7d-based GBPs to carbohydrate epitopes of choice, and especially those with no current reagents for their detection such as many bacterial and archaeal oligosaccharides. This directed evolution method requires small amounts of glycan, can quickly produce proteins with carbohydrate-binding ability, and does not require specialty knowledge to perform. However, compared to proteins and peptides, the carbohydrate epitope is generally much smaller with fewer chemical functional groups that elicit strong binding interactions with a protein binder. As such, it is crucial to carefully design the glycan ligands used during selections. To be successful, multivalent presentation of the glycan of interest is critical, and other components of the directed evolution platform with strong chemical functionality that could lead to off target interactions must be minimized. In our experience, biotinylated glycan ligands produced undesirable off target interactions, and use of a fluorophore-conjugated glycan ligand was the key to success. Multivalent polyacrylamide-based glycopolymers are commercially available for many mammalian glycans like those used in this work, but in-house synthesis of multivalent ligands will be necessary for unique or unusual glycans. This method, coupled with glycan isolation and multivalent synthetic procedures, enables engineering of GBPs that can specifically recognize a vast array of user-defined glycans of interest. This will have great impact on glycobiological research, allowing for the study of important glycans with no existing affinity tools.

Methods

General Yeast Growth/Induction Conditions

Use of yeast-surface display for directed evolution of glycan binders was carried out according to established protocols published by Wittrup and coworkers as described in the supporting methods.51

Selection of TF-binding Sso7d variants

Sso7d variants capable of binding carbohydrates were isolated using yeast-surface display.51 Selections were carried out using magnetic bead sorts and fluorescent-activated cell sorting as described in the supporting methods.

Affinity maturation of variants of interest and improved library generation

To improve the affinities of enriched variants, plasmids from heterogeneous populations of variants or individual variants of interest were subjected to one or more rounds of affinity maturation. Plasmids were isolated and error-prone PCR was performed as described in the supporting methods.

Recombinant expression and affinity purification of Sso7d constructs in E. coli.

Sso7d variants of interest were recombinantly expressed from E. coli as LPETGG-His6 constructs for biophysical characterization, sortase-mediated ligation and mammalian cell staining as described in the supporting methods.

Biophysical characterization of Sso7d-LPETG-His6 constructs by bio-layer interferometry

Binding affinities and of Sso7d variant 2.4.I were determined using bio-layer inferometry. All BLI measurements were performed in PBSA supplemented with 0.05% Tween-20 at 22 °C on an Octet RED96 instrument (Pall ForteBio, Fremont, CA.) as described in the supporting methods.

Sortase-mediated ligation

Sortase-mediated ligation was used to label the evolved GBP 2.4.I with FITC or biotin for use with glycoprotein blotting and cell labeling as described in the supporting methods.

Dot blot analysis of glycoproteins

Human MUC2, MUC5AC, and MUC5B glycoproteins were a very generous gift from the Ribbeck Lab at MIT. Enzymatic treatment of the mucin glycoproteins and dot blotting protocols are described in the supporting methods.

Flow cytometry staining of mammalian cell line with FITC-labeled proteins

The binding of evolved and functionalized 2.4.I proteins to mammalian cell-surface glycans was quantified by flow cytometry as described in the supporting methods.

Supplementary Material

Supp Info

Acknowledgements

We thank the lab of K. D. Wittrup for providing the rcSso7d yeast library and the lab of K. Ribbeck for providing samples of MUC2, MUC5AC, and MUC5B mucin proteins. We also thank A. Tisdale for advice on directed evolution, G. J. Dodge for advice and assistance with molecular modeling, and C. A. Arbour for assistance with characterization of the TF-S-NH2. The financial support from the NIH (U01CA231079 to B.I., T32GM0083334 to E.M.W.) and the MIT Biophysical Instrumentation Facility (BIF), the MIT Flow Cytometry Core, and the MIT Chemistry Department DCIF for BLI, FACS, NMR and HR-MS instrumentation are also gratefully acknowledged.

Footnotes

Supporting Information

Sequences and binding characteristics of variants from naïve library; tree and sequences of variants from affinity matured library; molecular modeling and docking; dot blot densitometry; mucin dot blots after enzyme treatment; flow cytometry analysis of MCF7 cells with decreased Jacalin concentration; synthesis and analysis of glycoconjugate TF-S-NH2 and control linker S-NH2.

Conflict of Interest

A patent application has been filed presenting the approach for the development of glycan-binding Sso7d scaffolds.

References

  • 1.Varki A (2017) Biological Roles of Glycans, Glycobiology 27, 3–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Fuster MM, and Esko JD (2005) The Sweet and Sour of Cancer: Glycans as Novel Therapeutic Targets, Nat. Rev. Cancer 5, 526–542. [DOI] [PubMed] [Google Scholar]
  • 3.Cummings RD, Etzler M, Hahn MG, Darvill A, Godula K, Woods RJ, and Mahal LK (2022) Glycan-Recognizing Probes as Tools, In Essentials of Glycobiology 4th Edition, Cold Spring Harbor Laboratory Press, Cold Spring Harbor (NY). [Google Scholar]
  • 4.Kappler K, and Hennet T (2020) Emergence and Significance of Carbohydrate-Specific Antibodies, Genes & Immunity 21, 224–239. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Ward EM, Kizer ME, and Imperiali B (2021) Strategies and Tactics for the Development of Selective Glycan-Binding Proteins, ACS Chem. Biol 16, 1795–1813. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Warkentin R, and Kwan DH (2021) Resources and Methods for Engineering “Designer” Glycan-Binding Proteins, Molecules 26, 380. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Adhikari P, Bachhawat-Sikder K, Thomas CJ, Ravishankar R, Jeyaprakash AA, Sharma V, Vijayan M, and Surolia A (2001) Mutational Analysis at Asn-41 in Peanut Agglutinin: A Residue Critical for the Binding of the Tumor-Associated Thomsen-Friedenreich Antigen, J. Biol. Chem 276, 40734–40739. [DOI] [PubMed] [Google Scholar]
  • 8.Yabe R, Suzuki R, Kuno A, Fujimoto Z, Jigami Y, and Hirabayashi J (2007) Tailoring a Novel Sialic Acid-Binding Lectin from a Ricin-B Chain-Like Galactose-Binding Protein by Natural Evolution-Mimicry, J. Biochem 141, 389–399. [DOI] [PubMed] [Google Scholar]
  • 9.Cicortas Gunnarsson L, Nordberg Karlsson E, Albrekt AS, Andersson M, Holst O, and Ohlin M (2004) A Carbohydrate Binding Module as a Diversity‐Carrying Scaffold, Protein Eng. Des. Sel 17, 213–221. [DOI] [PubMed] [Google Scholar]
  • 10.Lu Z, Kamat K, Johnson BP, Yin CC, Scholler N, and Abbott KL (2019) Generation of a Fully Human scFv That Binds Tumor-Specific Glycoforms, Sci. Rep 9, 5101. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Hong X, Ma MZ, Gildersleeve JC, Chowdhury S, Barchi JJ Jr., Mariuzza RA, Murphy MB, Mao L, and Pancer Z (2013) Sugar-Binding Proteins from Fish: Selection of High Affinity “Lambodies” That Recognize Biomedically Relevant Glycans, ACS Chem. Biol 8, 152–160. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Tasumi S, Velikovsky CA, Xu G, Gai SA, Wittrup KD, Flajnik MF, Mariuzza RA, and Pancer Z (2009) High-Affinity Lamprey VLRA and VLRB Monoclonal Antibodies, Proc. Natl. Acad. Sci 106, 12891. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.McKitrick TR, Goth CK, Rosenberg CS, Nakahara H, Heimburg-Molinaro J, McQuillan AM, Falco R, Rivers NJ, Herrin BR, Cooper MD, and Cummings RD (2020) Development of Smart Anti-Glycan Reagents Using Immunized Lampreys, Commun. Biol 3, 91. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.McKitrick TR, Bernard SM, Noll AJ, Collins BC, Goth CK, McQuillan AM, Heimburg-Molinaro J, Herrin BR, Wilson IA, Cooper MD, and Cummings RD (2021) Novel Lamprey Antibody Recognizes Terminal Sulfated Galactose Epitopes on Mammalian Glycoproteins, Commun. Biol 4, 674. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Khilji SK, Goerdeler F, Frensemeier K, Warschkau D, Lühle J, Fandi Z, Schirmeister F, Chen ZA, Turak O, Mallagaray A, Boerno S, Timmermann B, Rappsilber J, Seeberger PH, and Moscovitz O (2022) Generation of Glycan-Specific Nanobodies, Cell Chemical Biology 29, 1353–1361.e1356. [DOI] [PubMed] [Google Scholar]
  • 16.Könning D, and Kolmar H (2018) Beyond Antibody Engineering: Directed Evolution of Alternative Binding Scaffolds and Enzymes Using Yeast Surface Display, Microb. Cell Factories 17, 32. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Skerra A (2007) Alternative Non-Antibody Scaffolds for Molecular Recognition, Curr. Opin. Biotechnol 18, 295–304. [DOI] [PubMed] [Google Scholar]
  • 18.Gera N, Hussain M, Wright RC, and Rao BM (2011) Highly Stable Binding Proteins Derived from the Hyperthermophilic Sso7d Scaffold, J. Mol. Biol 409, 601–616. [DOI] [PubMed] [Google Scholar]
  • 19.Miller EA, Baniya S, Osorio D, Al Maalouf YJ, and Sikes HD (2018) Paper-Based Diagnostics in the Antigen-Depletion Regime: High-Density Immobilization of rcSso7d-Cellulose-Binding Domain Fusion Proteins for Efficient Target Capture, Biosens. Bioelectron 102, 456–463. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Miller EA, Traxlmayr MW, Shen J, and Sikes HD (2016) Activity-Based Assessment of an Engineered Hyperthermophilic Protein as a Capture Agent in Paper-Based Diagnostic Tests, Mol. Syst. Des. Eng 1, 377–381. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Weis WI, and Drickamer K (1996) Structural Basis of Lectin-Carbohydrate Recognition, Annu. Rev. Biochem 65, 441–473. [DOI] [PubMed] [Google Scholar]
  • 22.Traxlmayr MW, Kiefer JD, Srinivas RR, Lobner E, Tisdale AW, Mehta NK, Yang NJ, Tidor B, and Wittrup KD (2016) Strong Enrichment of Aromatic Residues in Binding Sites from a Charge-Neutralized Hyperthermostable Sso7d Scaffold Library, J. Biol. Chem 291, 22496–22508. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Hudson KL, Bartlett GJ, Diehl RC, Agirre J, Gallagher T, Kiessling LL, and Woolfson DN (2015) Carbohydrate–Aromatic Interactions in Proteins, J. Am. Chem. Soc 137, 15152–15160. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Kiessling LL, and Diehl RC (2021) CH−π Interactions in Glycan Recognition, ACS Chem. Biol 16, 1884–1893. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Asensio JL, Ardá A, Cañada FJ, and Jiménez-Barbero J (2013) Carbohydrate–Aromatic Interactions, Acc. Chem. Res 46, 946–954. [DOI] [PubMed] [Google Scholar]
  • 26.Ravn P, Danielczyk A, Bak Jensen K, Kristensen P, Astrup Christensen P, Larsen M, Karsten U, and Goletz S (2004) Multivalent scFv Display of Phagemid Repertoires for the Selection of Carbohydrate-Specific Antibodies and Its Application to the Thomsen–Friedenreich Antigen, J. Mol. Biol 343, 985–996. [DOI] [PubMed] [Google Scholar]
  • 27.Sharma V, Vijayan M, and Surolia A (1996) Imparting Exquisite Specificity to Peanut Agglutinin for the Tumor-Associated Thomsen-Friedenreich Antigen by Redesign of Its Combining Site J. Biol. Chem 271, 21209–21213. [DOI] [PubMed] [Google Scholar]
  • 28.Yu L-G (2007) The Oncofetal Thomsen–Friedenreich Carbohydrate Antigen in Cancer Progression, Glycoconjugate J. 24, 411–420. [DOI] [PubMed] [Google Scholar]
  • 29.Feng D, Shaikh AS, and Wang F (2016) Recent Advance in Tumor-Associated Carbohydrate Antigens (Tacas)-Based Antitumor Vaccines, ACS Chem. Biol 11, 850–863. [DOI] [PubMed] [Google Scholar]
  • 30.Monsigny M, Mayer R, Roche AC (1991) Sugar-Lectin Interactions: Sugar Cluster, Lectin Multivalency and Avidity, Carbohydr. Lett 4, 35–52. [PubMed] [Google Scholar]
  • 31.Kiessling LL, Young T, Gruber TD, and Mortell KH (2008) Multivalency in Protein–Carbohydrate Recognition, In Glycoscience: Chemistry and Chemical Biology (Fraser-Reid BO, Tatsuta K, and Thiem J, Eds.), pp 2483–2523, Springer Berlin Heidelberg, Berlin, Heidelberg. [Google Scholar]
  • 32.Ryckaert S, Callewaert N, Jacobs PP, Dewaele S, Dewerte I, and Contreras R (2007) Fishing for Lectins from Diverse Sequence Libraries by Yeast Surface Display – an Exploratory Study, Glycobiology 18, 137–144. [DOI] [PubMed] [Google Scholar]
  • 33.Bojar D, Meche L, Meng G, Eng W, Smith DF, Cummings RD, and Mahal LK (2022) A Useful Guide to Lectin Binding: Machine-Learning Directed Annotation of 57 Unique Lectin Specificities, ACS Chem. Biol [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.van Zundert GCP, Rodrigues JPGLM, Trellet M, Schmitz C, Kastritis PL, Karaca E, Melquiond ASJ, van Dijk M, de Vries SJ, and Bonvin AMJJ (2016) The HADDOCK2.2 Web Server: User-Friendly Integrative Modeling of Biomolecular Complexes, J. Mol. Biol 428, 720–725. [DOI] [PubMed] [Google Scholar]
  • 35.Jumper J, Evans R, Pritzel A, Green T, Figurnov M, Ronneberger O, Tunyasuvunakool K, Bates R, Žídek A, Potapenko A, Bridgland A, Meyer C, Kohl SAA, Ballard AJ, Cowie A, Romera-Paredes B, Nikolov S, Jain R, Adler J, Back T, Petersen S, Reiman D, Clancy E, Zielinski M, Steinegger M, Pacholska M, Berghammer T, Bodenstein S, Silver D, Vinyals O, Senior AW, Kavukcuoglu K, Kohli P, and Hassabis D (2021) Highly Accurate Protein Structure Prediction with Alphafold, Nature 596, 583–589. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Wartchow CA, Podlaski F, Li S, Rowan K, Zhang X, Mark D, and Huang K-S (2011) Biosensor-Based Small Molecule Fragment Screening with Biolayer Interferometry, J. Comput. Aided Mol.. Des 25, 669–676. [DOI] [PubMed] [Google Scholar]
  • 37.Lee RT, and Lee YC (2000) Affinity Enhancement by Multivalent Lectin–Carbohydrate Interaction, Glycoconjugate J. 17, 543–551. [DOI] [PubMed] [Google Scholar]
  • 38.Hirabayashi J, Tateno H, Shikanai T, Aoki-Kinoshita KF, and Narimatsu H (2015) The Lectin Frontier Database (Lfdb), and Data Generation Based on Frontal Affinity Chromatography, Molecules 20. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Mao H, Hart SA, Schink A, and Pollok BA (2004) Sortase-Mediated Protein Ligation:  A New Method for Protein Engineering, J. Am. Chem. Soc 126, 2670–2671. [DOI] [PubMed] [Google Scholar]
  • 40.Antos JM, Ingram J, Fang T, Pishesha N, Truttmann MC, and Ploegh HL (2017) Site-Specific Protein Labeling Via Sortase-Mediated Transpeptidation, Curr. Protoc. Protein Sci 89, 15.13.11–15.13.19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Dolan B, and Hansson GC (2021) Mucins, In Reference Module in Life Sciences, Elsevier. [Google Scholar]
  • 42.Corfield AP (2015) Mucins: A Biologically Relevant Glycan Barrier in Mucosal Protection, Biochim. Biophys. Acta 1850, 236–252. [DOI] [PubMed] [Google Scholar]
  • 43.Holmén Larsson JM, Karlsson H, Sjövall H, and Hansson GC (2009) A Complex, but Uniform O-Glycosylation of the Human MUC2 Mucin from Colonic Biopsies Analyzed by Nanolc/Msn, Glycobiology 19, 756–766. [DOI] [PubMed] [Google Scholar]
  • 44.Capon C, Maes E, Michalski JC, Leffler H, and Kim YS (2001) Sd(a)-Antigen-Like Structures Carried on Core 3 Are Prominent Features of Glycans from the Mucin of Normal Human Descending Colon, Biochem. J 358, 657–664. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Robbe C, Capon C, Coddeville B, and Michalski JC (2004) Structural Diversity and Specific Distribution of O-Glycans in Normal Human Mucins Along the Intestinal Tract, Biochem. J 384, 307–316. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Jin C, Kenny DT, Skoog EC, Padra M, Adamczyk B, Vitizeva V, Thorell A, Venkatakrishnan V, Lindén SK, and Karlsson NG (2017) Structural Diversity of Human Gastric Mucin Glycans Mol. Cell. Proteomics 16, 743–758. [DOI] [PubMed] [Google Scholar]
  • 47.Thomsson KA, Prakobphol A, Leffler H, Reddy MS, Levine MJ, Fisher SJ, and Hansson GC (2002) The Salivary Mucin Mg1 (MUC5B) Carries a Repertoire of Unique Oligosaccharides That Is Large and Diverse, Glycobiology 12, 1–14. [DOI] [PubMed] [Google Scholar]
  • 48.Xia B, Royall JA, Damera G, Sachdev GP, and Cummings RD (2005) Altered O-Glycosylation and Sulfation of Airway Mucins Associated with Cystic Fibrosis, Glycobiology 15, 747–775. [DOI] [PubMed] [Google Scholar]
  • 49.Geiger P, Mayer B, Wiest I, Schulze S, Jeschke U, and Weissenbacher T (2016) Binding of Galectin-1 to Breast Cancer Cells MCF7 Induces Apoptosis and Inhibition of Proliferation in vitro in a 2D- and 3D- Cell Culture Model, BMC Cancer 16, 870. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Arockia Jeyaprakash A, Katiyar S, Swaminathan CP, Sekar K, Surolia A, and Vijayan M (2003) Structural Basis of the Carbohydrate Specificities of Jacalin: An X-Ray and Modeling Study, J. Mol. Biol 332, 217–228. [DOI] [PubMed] [Google Scholar]
  • 51.Chao G, Lau WL, Hackel BJ, Sazinsky SL, Lippow SM, and Wittrup KD (2006) Isolating and Engineering Human Antibodies Using Yeast Surface Display, Nat. Protoc 1, 755–768. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supp Info
NIHMS1859207-supplement-Supp_Info.pdf (1.3MB, pdf)

RESOURCES