Fishing for lectins from diverse sequence libraries by yeast surface display – An exploratory study

Stefan Ryckaert 0 3 Nico Callewaert 1 2 Pieter P. Jacobs 0 3 Sylviane Dewaele 0 3 Isabelle Dewerte 1 2 Roland Contreras 0 3 0 Department for Molecular Biomedical Research, Unit for Fundamental and Applied Molecular Biology , VIB 1 Department of Biochemistry, Physiology and Microbiology, Ghent University , B-9052 Ghent , Belgium 2 Department for Molecular Biomedical Research , Unit for Molecular Glycobiology, VIB 3 Department of Molecular Biology, Ghent University The establishment of a robust technology platform for the expression cloning of carbohydrate-binding proteins remains a key challenge in glycomics. Here we explore the utility of using yeast surface display (YSD) technology in the interaction-based lectin cloning from complete cDNA libraries. This should pave the way for more detailed studies of protein-carbohydrate interactions. To evaluate the performance of this system, lectins representing three different subfamilies (galectins, siglecs, and C-type lectins) were successfully displayed on the surface of Saccharomyces cerevisiae and Pichia pastoris as a-agglutinin and/or agglutinin fusions. The predicted carbohydrate-binding activity could be detected for three out of five lectins tested (galectin-1, galectin-3, and siaoadhesin). For galectin-4 and E-selectin, no specific carbohydrate-binding activity could be detected. We also demonstrate that proteins with carbohydrate affinity can be specifically isolated from complex metazoan cDNA libraries through multiple rounds of FACS sorting, employing multivalent, fluorescent-labeled polyacrylamide-based glycoconjugates. Introduction Lectins are carbohydrate-binding proteins involved in numerous biological processes (Gabius et al. 2002). Moreover, they are essential tools in the emerging field of glycomics, allowing specific detection of target analytes, hence fulfilling functions analogous to that of antibodies in proteomics. Phage display represents a major advance in the antibodyengineering field, but is of limited use in the interaction cloning of proteins with carbohydrate-binding activity (Yamamoto et al. 1999; Ravn et al. 2004). First, because protein-carbohydrate interactions depend largely on multivalency (Ravn et al. 2004), 1To whom correspondence should be addressed: Tel: +32-9-331-3617; Fax: +32-9-331-3502; e-mail: 2Both first authors contributed equally. the monovalent display format, which is applied in most cases, may not be optimal for dealing with carbohydrate antigens. Second, a prokaryotic expression host has an unpredictable but frequently strong expression bias against many eukaryotic proteins. To complement the lectin cloning tools offered by phage display with a eukaryotic counterpart, we rely on yeast surface display (YSD) (Schreuder et al. 1993; Boder and Wittrup 1997). This technology was originally developed to enhance secretion efficiency, stability, and affinity of proteins (Boder et al. 2000; Holler et al. 2000; Shusta et al. 2000), but has not been used to identify proteins with carbohydrate affinity. The multivalent YSD format intrinsically mimics the natural multivalent presentation of lectins on cells (Figure 1A), and could largely overcome the technical problems observed with other (prokaryotic) protein display platforms that have been used for lectin cloning (Ravn et al. 2004). We build on the successes of the YSD format that were obtained in the antibody-engineering field. Here it proved successful in both the isolation of scFvs from large nonimmune libraries (Feldhaus et al. 2003) and the maturation of these scFvs toward higher affinities. Currently, the yeast display method has yielded the highest affinity (48 fM) for any antibody (Boder et al. 2000). We foresee similar application for YSD in the cloning and engineering of lectins. Results and discussion To evaluate the possibility of using YSD to study proteincarbohydrate interactions, we displayed on the surface of Saccharomyces cerevisiae three galectins from different subfamilies and possessing different natural oligomerization properties (galectin-1: dimeric, galectin-3 and -4: monomeric) (system outline: Figure 1A). Galectins are soluble -galactosidebinding lectins (Barondes et al. 1994) that are expressed as surface receptors when introduced in the YSD system. Likewise, when expressing cDNA libraries, all soluble receptors become cell surface localized. Hereby, the AGA1-AGA2 complex links these proteins to the cell wall. Using fluorescenceactivated cell sorting (FACS) analysis with a fluorescein isothiocyanate (FITC)-labeled anti-V5-tag mAb, we determined that a large fraction of cells displayed high levels of the fullsize proteins (mean fluorescence intensity is up to 100-fold higher than that of nonexpressing cells) (Figure 2AC). Subsequently, by FACS analysis we demonstrated the interaction between the surface-displayed galectins and multivalent, LacNAc-containing, fluorescent-labeled polyacrylamide-based glycoconjugates (LacNAc-PAA-FITC) (Galanina et al. 1998). We demonstrated that galectin-1 and -3 interact specifically with LacNAc-PAA-FITC (Figure 2F and G). However, only a few cells displaying galectin-4 showed binding to this conjugate (Figure 2H), which concurs with the notion that galectin-4 has fusion (O). a weak affinity for LacNAc (instead it binds specifically to 3 -O-sulfated Gal1,3-GalNAc) (Ideo et al. 2002). However, the lower intrinsic surface expressibility, compared to galectin1 and -3, could also contribute to this. As can be deduced from Figure 1B, the minimum expression threshold that allows detection of carbohydrate binding is not reached for galectin-4 (Figure 2). To broaden the scope of our technology, we displayed the carbohydrate recognition domain (CRD) of sialoadhesin, which belongs to the siglec family of animal lectins. Despite high levels of display (Figure 2D), no lectin activity could be detected on the yeast surface (Figure 2I). Therefore, we switched to Pichia pastoris, which is an outstanding host for the production of heterologous proteins and often gives higher yields than V5 detectiona Sugar-PAA-FITC binding V5 detectiona S. cerevisiae. Moreover, the availability of glycoengineered Pichia strains that produce hybrid and complex N-glycans (Vervecken et al. 2004; Wildt and Gerngross 2005) should allow the detection of mannooligosaccharide-binding lectins, which might otherwise become quenched and undetectable in this yeast-based approach due to the presence of large amounts of cell wall mannan. These lectins are instead expected to agglutinate the yeast cells, which produce high-mannose glycans. Strikingly, in a Pichia strain with abolished hypermannosylation (Vervecken et al. 2004), surface display levels of both galectin1 and sialoadhesin a-agglutinin fusions were lower than that in S. cerevisiae (compare Figure 2K and L with Figure 2A and D) and no sugar-PAA-FITC binding could be detected (Figure 2N and O). In contrast, in Pichia, surface display of sialoadhesin -aggl (...truncated)