- Email: [email protected]
Evolution of a carbohydrate binding module into a protein-specific binder
Biomolecular Engineering 23 (2006) 111–117 www.elsevier.com/locate/geneanabioeng
Evolution of a carbohydrate binding module into a protein-specific binder Lavinia Cicortas Gunnarsson a, Linda Dexlin a, Eva Nordberg Karlsson b, Olle Holst b, Mats Ohlin a,* a
Department of Immunotechnology, Lund University, P.O. Box 7031, S-220 07 Lund, Sweden b Department of Biotechnology, Lund University, P.O. Box 124, S-221 00 Lund, Sweden Received in revised form 9 November 2005; accepted 5 December 2005
Abstract A carbohydrate binding module, CBM4-2, derived from the xylanase (Xyn 10A) of Rhodothermus marinus has been used as a scaffold for molecular diversification. Its binding specificity has been evolved to recognise a quite different target, a human monoclonal IgG4. In order to understand the basis for this drastic change in specificity we have further investigated the target recognition of the IgG4-specific CBMs. Firstly, we defined that the structure target recognised by the selected CBM-variants was the protein and not the carbohydrates attached to the glycoprotein. We also identified key residues involved in the new specificity and/or responsible for the swap in specificity, from xylan to human IgG4. Specific changes present in all these CBMs included mutations not introduced in the design of the library from which the specific clones were selected. Reversion of such mutations led to a complete loss of binding to the target molecule, suggesting that they are critical for the recognition of human IgG4. Together with the mutations introduced at will, they had transformed the CBM scaffold into a protein binder. We have thus shown that the scaffold of CBM4-2 is able to harbour molecular recognition for either carbohydrate or protein structures. # 2005 Elsevier B.V. All rights reserved. Keywords: Binding specificity; Combinatorial library; Mutant; Phage-display; Molecular engineering; Protein scaffold
1. Introduction In vitro generation of proteins capable of specific target recognition has always been a challenge for the molecular biologists. The development of combinatorial library approaches in combination with in vitro selection methods, such as phage (Smith, 1985), ribosomal (Mattheakis et al., 1994) and cell-surface (Daugherty et al., 1998) display, has facilitated the creation and identification of such proteins. For a long time, antibodies, a group of naturally diverse binders well suited for therapeutic applications, have been a primary choice of proteins for use in this type of studies. In the growing field of biotechnology, there is a constant need of novel binders and the
Abbreviations: AE, affinity electrophoresis; BSA, bovine serum albumin; CBM, carbohydrate binding module; ConA, concanavalin A; ELISA, enzymelinked immunosorbent assay; HRP, horseradish peroxidase; IPTG, isopropyl-bD-thiogalactoside; PBS, phosphate-buffered saline; PCR, polymerase chain reaction; WGA, wheat germ agglutinin * Corresponding author. Tel.: +46 46 222 4322; fax: +46 46 222 4200. E-mail address: [email protected] (M. Ohlin). 1389-0344/$ – see front matter # 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.bioeng.2005.12.002
search for such proteins has been extended to include alternative scaffolds to the antibody framework (Nygren and Skerra, 2004). Recent reports have described the successful use of proteins such as lipocalins (Skerra, 2000), a single protein A domain (Nord et al., 1997) and the fibronectin type III domain (Koide et al., 1998), for generation of specific protein binders. Depending on their additional properties, such as size, stability and production yields in for example Escherichia coli, these molecules have potential for use in diagnostics and bioseparation applications. The carbohydrate binding modules (CBMs) are noncatalytic modules able to recognise several types of target structures. They are connected to a variety of glycoside hydrolysing or modifying enzymes, showing specificity towards one or several carbohydrate ligands. The CBMs are divided into different groups depending on the topology of their binding site, which in turn is related to the type of ligand they recognise, just like antibodies. Flat binding surfaces, found on antibodies binding mostly to proteins (MacCallum et al., 1996) (http://www.unizh.ch/antibody/Structures/ AgContact/), are also present on CBMs binding to insoluble
112
L.C. Gunnarsson et al. / Biomolecular Engineering 23 (2006) 111–117
Fig. 1. Amino acid sequence alignment of intentionally mutated regions of the primary structure of human IgG4-specific CBM-variants. The amino acid numbering in this figure is according to (Simpson et al., 2002). The genes encoding these proteins can be accessed in GenBank (accession numbers AY534556–AY534560, AY842281–AY842282). Other modifications, characteristic for IgG4-specific protein modules, are also shown.
crystalline surfaces (Linder et al., 1996; Simpson and Barras, 1999). CBMs that recognise free single carbohydrate chains display a binding cleft (Simpson et al., 2002), which is also the case for antibodies binding to peptides. Small ligands such as haptens in the case of antibodies (MacCallum et al., 1996) (http://www.unizh.ch/antibody/Structures/AgContact/), and mono- and disaccharides in the case of CBMs bind into solvent exposed binding slots (Boraston et al., 2001). Altering the binding ability of a single carbohydrate binding module is a relatively unexplored area. Except for our work, there are, to our knowledge, only two reports (Lehtio¨ et al., 2000; Smith et al., 1998) where molecular engineering of one and the same CBM scaffold has been tested. In both those cases CBM variants specific towards a protein target have successfully been engineered. As described in our previous publication (Cicortas Gunnarsson et al., 2004), we have constructed a combinatorial library by limited substitutions of 12 residues around the binding site of CBM4-2, a thermostable CBM from the modular xylanase Xyn 10A, originating from Rhodothermus marinus (Nordberg Karlsson et al., 1997). This CBM binds to its original substrate, xylan, both by three dimensionally fitting the helical structured polysaccharide into the binding groove and by establishing specific interactions between the substrate and some of the surface exposed residues at the binding site (Simpson et al., 2002). Successful selections were performed using phage display on two other polysaccharides, Avicel (50% amorphous and 50% crystalline cellulose) and ivory nut mannan, that like xylan are suitable for binding into clefts as the one displayed on the CBM4-2. Despite the small size of the library we were also able to select CBM-variants specific towards the glycoprotein human IgG4 (Fig. 1). While the CBMs selected on carbohydrates retained some of the wild type specificity towards xylan, the changes undergone by the IgG4-specific CBMs led to a complete loss of their ability to bind to xylan. In this study we have further analysed how the CBM can evolve to specifically bind a quite different target like the human IgG4 molecule. One important issue has been to demonstrate that these CBMs recognise the protein itself and not the carbohydrates of the target. We have also identified some of the key residues involved in the new specificity and/or responsible for the swap in specificity of the CBM from xylan to human IgG4 and we have shown that the critical modifications extend beyond the surface directly involved in the recognition of the original carbohydrate substrate.
2. Materials and methods 2.1. Strains and vectors E. coli strain Top10F0 (Invitrogen, Paisley, UK) was used as host for construction of single mutants and for phage display work. The phagemid used was a variant of pFab5c.His (Engberg et al., 1996) carrying the part of gene III encoding only the final C-terminal domain of M13 protein 3. Production of soluble recombinant protein was achieved using E. coli strain BL21(DE3) (Novagen, Madison, WI, USA) and the expression vector pET-22b(+) (Novagen).
2.2. Phage stocks Phage stocks, each displaying a single CBM variant, were created by infecting exponentially growing cultures of a specific clone (2 YT, 1% (w/v) glucose, 100 mg/ml ampicillin, 10 mg/ml tetracycline at OD600 0.5) with helper phages VCSM13 (multiplicity of infection: approximately 20) (Stratagene, La Jolla, CA, USA) for 30 min at 37 8C. The medium was changed (2 YT, 100 mg/ml ampicillin, 10 mg/ml tetracycline, 50 mg/ml kanamycin and 0.25 mM isopropyl-b-D-thiogalactoside (IPTG)) and the cultures were grown at 30 8C overnight. After centrifugation, the supernatants containing the phage particles were filtered through 0.45 mm filters and stored at 4 8C.
2.3. Denaturation and deglycosylation of human IgG4 The monoclonal antibody previously used in the selection of the IgG4specific variants of CBM (Cicortas Gunnarsson et al., 2004) is a human IgG4l antibody (Dynal A/S, Oslo, Norway) specific for an epitope shared by all mouse IgG subclasses. This human IgG4 in soluble form (Dynal) was denatured by incubation at 70 8C in 8 M urea for 30 min prior to coating on a 96 well plate. Enzyme-linked immunosorbent assay (ELISA) using phage-displayed CBMs (see below) was performed to investigate whether the IgG4-specific CBMs could still bind to the IgG4. The presence of carbohydrates in the denatured protein was detected by using the horseradish peroxidase (HRP) conjugated lectins wheat germ agglutinin (WGA) and concanavalin A (ConA) (Sigma– Aldrich Inc., St. Louis, MO, USA). Deglycosylation of 4 mg of human IgG4 (Dynal) with 0.5 units of Nglycosidase F (PNGaseF; Roche Diagnostic Corporation, IN, USA) was performed overnight at 37 8C prior to coating. The HRP conjugated lectin ConA was used to monitor the removal of N-linked carbohydrates by ELISA. In both cases, 96 well ELISA plates were coated at 4 8C overnight with 1 mg/ml of native, denatured or de-N-glycosylated human IgG4 (Dynal) in phosphate-buffered saline (PBS). After washing the plates with washing buffer (154 mM NaCl and 0.05% (v/v) Tween-20 in PBS) the phages diluted in a blocking buffer (PBS containing 1% bovine serum albumin (BSA) and 0.05% (v/v) Tween-20) were added and the plates were incubated for 2 h at room temperature. A second washing step was followed by addition of rat anti-M13 protein 8 antibody diluted in blocking buffer for 1 h at room temperature. Plates were washed and HRP conjugated rabbit-anti-mouse immunoglobulins that cross-react with rat IgG (DAKO A/S, Glostrup, Denmark) diluted in blocking buffer were added to the wells and incubated for 1 h at room temperature. For detection, substrate (0.67 mg ortophenylene-diamine/ml 35 mM citrate and 67 mM phosphate buffer pH 5 and 0.012% H2O2) was added. The reaction
L.C. Gunnarsson et al. / Biomolecular Engineering 23 (2006) 111–117 was stopped by addition of H2SO4 to a final concentration of 0.6 M and the absorbance was recorded at 490 nm.
2.4. Construction of mutants Site-directed mutagenesis was used to construct mutants originating both from the wild type CBM4-2 (wt E138A, wt E138G, wt D136G and wt Q108R) and from G-4, a human IgG4-specific variant of CBM (G4 A138E). In order to introduce these point mutations, polymerase chain reactions (PCRs) were performed using a set of primers (Table 1) and the phagemid vector encoding the wild type or the G-4 variant (human IgG4-specific) of CBM4-2 as templates. Initially two fragments, one harbouring the desired mutation, were amplified for each mutant using one gene- and one vector-specific primer (Table 1) and platinum Pfx DNA polymerase with proofreading (Invitrogen) (94 8C for 2 min; 20 or 25 cycles: 94 8C for 30 s, 55 8C for 30 s, 68 8C for 2 min; 68 8C for 10 min). The fragments were then assembled in an overlap-extension PCR (94 8C for 2 min; 8 cycles: 94 8C for 30 s, 52 8C or 55 8C for 30 s, 68 8C for 2 min; 68 8C for 10 min) and the resulting PCR-products were amplified (94 8C for 2 min; 25 cycles: 94 8C for 30 s, 55 8C for 30 s, 68 8C for 2 min; 68 8C for 10 min) using the pFab 5C vector-specific primers. The PCR-products were digested with SfiI and NotI before they were inserted into the phagemid vector, digested with the same restriction enzymes, and transformed into competent E. coli Top10F0. Another set of mutants from G-4 (G4 L76A and G4 L110A) and from the IgG4-specific CBM-variant G-11 (G11 G136D) were constructed as described above but instead using AmpliTaq DNA polymerase (Applied Biosystems,
Table 1 Oligonucleotide primers used to introduce site-directed mutationsa Mutant
Gene-specific primersb
wt E138A,G
Fragment 1: F 50 -AGTTTACGGTCA GTGATCAGGSGACGGTCATTCc Fragment 2: R 50 -CTGATCACTG ACCGTAAACTCGAACGTGAA Fragment 1: F 50 -CGTTCGAGTTTA CGGTCAGTGGTCAGGAGACGG Fragment 2: R 50 -ACTGACCGTAAA CTCGAACGTGAACGGCTG Fragment 1: F 50 -TCAGCTTCACG GTGGGGAACCGGTCGTTCCAGG Fragment 2: R 50 -GTTCCCCACCGT GAAGCTGACCACCGCCCC Fragment 1: F 50 -AGTTTACGGTCA GTGATCAGGAGACGGTCATTA Fragment 2: R 50 -CTGATCACTGA CCGTAAACTCGAACGTGAA Fragment 1: F 50 -TCGACATCGATGC GACGGCCGCCCCGGTGAACG Fragment 2: R 50 -GGCCGTCGCATCG ATGTCGAAGGGGTTGTT Fragment 1: F 50 -TCACGGTGGGGAA CCAGTCGGCTGAGCATTACG Fragment 2: R 50 -CGACTGGTTCCC CACCGTGAAGCTGACCAC Fragment 1: F 50 -CGTTCGAGTTTAC GGTCAGTGATCAGGAGACGG Fragment 2: R 50 -ACTGACCGTAAA CTCGAACGTGAACGGCTG
wt D136G
wt Q108R
G4 A138E
G4 L76A
G4 L110A
G11 G136D
Vector-specific primers pFab5c.His pFab5c.His pET-22b(+) pET-22b(+) a b c
F 50 -TTTCACACAGGAAACAGCTATG R 50 -GCCTTTAGCGTCAGACTGTAG F 50 - TAATACGACTCACTATAGGG R 50 -CTAGTTATTGCTCAGCGGT
Prefixes F and R stand for the forward and reverse primers, respectively. Mutated codons are underlined. S = G or C.
113
Foster City, CA, USA) and the pET-22b(+) vector-specific primers. The PCRproducts in these cases were digested with NdeI and XhoI and the genes were inserted into the pET-22b(+) vector and transformed into competent E. coli BL21(DE3). All constructs were validated by DNA sequencing.
2.5. Binding assays using phage-displayed proteins The binding of phage-displayed mutants to xylan, the wild type substrate, was studied by ELISA using plates coated with the soluble part of birchwood xylan (Sigma–Aldrich). After incubation for 2 h at 37 8C with the phages in blocking buffer followed by a washing step, the bound phages were detected by incubation with HRP conjugated mouse anti-M13 antibody (Amersham Bioscienses, Piscataway, NJ, USA) for 1 h at 37 8C. The detection was then carried out as described above. Human IgG4 binding of the mutants was assayed by incubating 10 ml of phage-stock with 5 ml (2 106) Dynabeads coated with human monoclonal IgG4 (Pan Mouse IgG; Dynal) in 1 ml of PBS containing 1% BSA and 0.05% (v/v) Tween-20. After 1 h of incubation on an end-to-end rotor at room temperature, the Dynabeads were washed four times with PBS containing 1% BSA and 0.05% (v/ v) Tween-20 and twice with PBS. Bound phages were eluted using 100 ml trypsin (0.5 mg/ml) (Invitrogen) by incubation for 30 min at room temperature with shaking (Johansen et al., 1995). After removing the substrate, the activity of trypsin was neutralised by addition of 100 ml aprotinin (0.1 mg/ml) (Roche Diagnostic Corporation, IN, USA). Both input and output phages were titrated on exponentially growing E. coli Top10F0 and the colony forming units were counted and used to calculate the binding ratio as bound phages per input phages.
2.6. Protein production and purification Genes encoding CBM variants cloned into the pFab5c.His vector were first re-cloned in-between the NdeI and XhoI sites found in the pET-22b(+) vector. These CBMs and other mutated variants encoded by genes directly cloned into the pET-22b(+) vector (see above), each carrying a C-terminal hexa-histidine tag were produced in E. coli BL21(DE3). Soluble proteins were purified using metal-ion-affinity chromatography as described earlier (Abou Hachem et al., 2000). The only differences in this work were that 20 mM sodium phosphate buffer (pH 7.4)/20 mM imidazole/0.75 M NaCl was used as binding buffer and the chromatography columns were packed with Ni-NTA (Qiagen, Hilden, Germany). The concentration of each purified protein was determined spectrophotometrically at A280 nm using extinction coefficients individually calculated for each CBM-variant.
2.7. Binding assays using purified proteins Once the mutants were purified as soluble proteins (see above), affinity electrophoresis (AE) was performed to confirm the results on binding to xylan obtained in the phage-display system. The AE method was performed in the BioRad (Hercules, CA, USA) mini-gel apparatus as earlier described for the wild type CBM4-2 (Abou Hachem et al., 2000). Purified CBM-variants (3 mg per gel) were separated at room temperature and 90 V on two different native gels, with or without 0.2% (w/v) oat spelt xylan (Sigma–Aldrich). A Kaleidoscope prestained standard (Bio-Rad) was used as a negative, non-interacting, control and the proteins were detected by staining with simply blue safe stain (Invitrogen). The studies on human IgG4-binding using phage-display, were confirmed by ELISA using purified CBM-variants (5 mg/ml in PBS) coated onto plates at room temperature overnight. After washing the plates human IgG4, 1 mg/ml diluted in blocking buffer, was added and the plates were incubated for 2 h at room temperature. The bound human IgG4 molecules were detected by adding HRP conjugated goat anti-human IgG antibody (Zymed, South San Francisco, CA, USA) for 1 h at room temperature and using the detection method for ELISA described above.
3. Results 3.1. Specificity studies of IgG4-binding CBMs As reported previously (Cicortas Gunnarsson et al., 2004), variants of CBM4-2 (Fig. 1) specific towards monoclonal
114
L.C. Gunnarsson et al. / Biomolecular Engineering 23 (2006) 111–117
human IgG4 carrying a l light chain have been selected from a combinatorial library. The binding specificity of these molecules was tested by ELISA using plates coated with the selection target IgG4, another monoclonal human IgG4l and polyclonal human IgG. The selected CBMs showed binding specificity towards the IgG4 used for selection while none of the other molecules gave raise to any detectable binding in this assay (data not shown). These results suggested that the CBMs recognise a protein or carbohydrate structure specific for the target IgG4-molecule and ruled out recognition of conserved structures found in the constant part of IgG4l. 3.2. Target epitope In order to determine whether the IgG4-specific CBMs recognise a protein or a carbohydrate structure on the glycoprotein target, ELISA was performed on plates coated with denatured human IgG4. Denaturation of the human IgG4 molecule using urea destroys the total protein structure, hardly affecting the carbohydrate structures present. Human IgG4specific CBMs showed no binding to these plates while the lectins WGA and ConA still bound, indicating a remained presence of carbohydrates (Fig. 2A). This suggests that the IgG4-specific CBM-variants actually recognise a protein structure of the target.
Fig. 2. Binding of phage-displayed wild type CBM4-2 and CBM variants selected on human IgG4 (G-1, G-4, G-11 and G-15) and lectins to native (black) and denatured (striped) human IgG4 (A) or native (black) and de-N-glycosylated (grey) human IgG4 and only PNGaseF (white), as a control coating, (B) determined by ELISA. The error bars represent the standard deviation of experiments performed in quadruplicates (A) or triplicates (B).
The finding that the IgG4-specific CBMs recognise the protein structure was also confirmed by using the deglycosylation enzyme PNGaseF to remove the N-linked carbohydrates (the major type of carbohydrates on immunoglobulins) from the human IgG4 molecule. This time ELISA-plates were coated with both native and deglycosylated IgG4 and the lectin ConA served as a control molecule demonstrating the effectiveness of the deglycosylation procedure. In agreement with our findings above the human IgG4-specific CBMs, in contrast to ConA, retained most of its binding to IgG4 even after deglycosylation, when the substrate was lacking the carbohydrates (Fig. 2B). This proves once again that the selected CBMs recognise the protein itself and not the carbohydrates of IgG4. Thus the CBM4-2 scaffold can be evolved to recognise a protein structure, which is quite different from the xylan structure that is recognised by the unmodified module. 3.3. Binding studies of the single-mutants Specific sequence features of the CBM-clones selected on human IgG4 raised issues on how the specificity of these modules was established. Additional to the mutations introduced by the library design, all binders (24 sequenced clones, 7 unique sequences) selected on the human IgG4 monoclonal antibody (Fig. 1) had one of the following spontaneous mutations Q108R, D136G or E138A/G, all located in proximity to residue 110 (Fig. 3), one of the residues believed to be important in xylan binding (Simpson et al., 2002). These mutations were not found among selected carbohydratebinding variants (Cicortas Gunnarsson et al., 2004) suggesting that they were important for the IgG4-specificity. A number of rationally designed mutants were hence constructed to investigate the importance of the observed additional mutations for the specific CBM-binding to IgG4. All mutants cloned into the pFab5c.His vector were successfully
Fig. 3. The structure of CBM4-2 (PDB 1k45) showing the residues not included in the design of the library but additionally mutated in the human IgG4-specific clones (Q108, D136 and E138) in red. In addition, two residues (69 in green and 110 in yellow) important for the binding of the wild type CBM4-2 to xylan and situated on the loops defining the binding cleft, are also highlighted.
L.C. Gunnarsson et al. / Biomolecular Engineering 23 (2006) 111–117
displayed as functional proteins on phages (data not shown), allowing binding assays to be performed. The phage-ELISA using plates coated with soluble birchwood xylan as a substrate indicated that the individual mutations Q108R, D136G or E138A/G introduced in the wild type CBM did not influence its binding to xylan (Fig. 4A). Furthermore, these mutations were not sufficient to change the specificity of CBM4-2 towards recognition of human IgG4 (Fig. 4B). However, reversion of the mutation in positions 138 of the G-4 variant, the most commonly modified position among the IgG4-specific variants, to the wild type encoded residue glutamate caused a complete loss of binding specificity for IgG4 (Fig. 4B) without regaining binding ability to xylan (Fig. 4A). Thus mutations such as E138A observed in the G4-variant are important for the modified module’s ability to recognise human IgG4 but are not critically involved in the xylan binding surface of CBM4-2. To confirm these results and extend the study of residues responsible for the evolved human IgG4-specificity, the binding properties of CBM-variants were also assayed using purified proteins in soluble form. Two additional mutants created from the G-4 variant (G4 L76A and G4 L110A) and a single mutant created from the G-11 variant (G11 G136D) were also included at this stage. The human IgG4-specific CBMs (G-4 and G-11) and mutants created from these variants had no binding specificity for xylan, as judged from affinity electrophoresis
115
Fig. 5. Binding studies using soluble variants of CBM. Non-denaturing affinity electrophoresis in the presence (A) or in the absence of oat spelt xylan (B). Proteins: lanes 1 and 2, two previously selected xylan specific clones (X-6 and X-2, respectively); lane 3, wild type CBM4-2; lane 4, a human IgG4-specific clone (G-11); lane 5, single mutant of G-11 (G11 G138D); lane 6, Kaleidoscope prestained standard; lane 7, a human IgG4-specific clone (G-4); lanes 8–10, single mutants of the G-4 (G4 A138E, G4 L76A and G4 L110A, respectively). (C) Addition of human IgG4 molecules to microtiter plates coated with different CBM variants showed binding to the IgG4-specific clones (G-4 and G-11) but not to the mutants created by reversion of their additional mutations (G4 A138E or G11 G136D) or to the wild type CBM4-2. Substitution of L110, but not L76, for alanine in G-4 abrogated binding to the human IgG4.
(Fig. 5A and B). ELISA, using plates coated with soluble CBMs, clearly showed that elimination of the additional mutations found in the G-4 and G-11 variants, those in positions 138 and 136, respectively, led to a complete loss of binding to IgG4 (Fig. 5C). The same consequence was seen when the leucin in position 110 of the G-4 variant was replaced with an alanine, whereas this type of mutation seemed to be well tolerated in position 76. Both residues, L76 and L110 were conserved in all sequences of human IgG4-specific CBMs (Fig. 1). In conclusion, this means that the transformation of CBM4-2 into IgG4-binding variants required specific changes not only inside but also outside the carbohydrate binding cleft and they together pushed the CBM into becoming a proteinbinding module. 4. Discussion
Fig. 4. Binding studies on birchwood xylan (A) performed by ELISA using CBM variants, such as the wild type CBM4-2, the IgG4-specific clone (G-4) and single mutants of these two CBMs, displayed on phages. The error bars represent the standard deviation of experiments performed in quadruplicates. (B) The same phage-displayed clones were assayed for binding to human IgG4 when the substrate was immobilised on Dynabead particles.
It is well known that the field of biotechnology is in need of new binding proteins with well-defined specificity. The antibody scaffold has, for many years, served as a starting point when searching for new binders (So¨derlind et al., 2001). In addition, methods for creating molecular libraries and a series of selection procedures have contributed to the successful evolution of specific binders towards different target molecules such as proteins, peptides, nucleic acids and haptens (Nygren
116
L.C. Gunnarsson et al. / Biomolecular Engineering 23 (2006) 111–117
and Skerra, 2004; Lipovsek and Plu¨ckthun, 2004). However, raising specific antibodies towards carbohydrates has proven to be a challenge, presumably because binding to this kind of targets requires features difficult to harbour by antibodies. For this purpose the naturally occurring carbohydrate binding modules could be utilised and we have recently shown that a CBM can be successfully used as a scaffold to develop variants with new carbohydrate binding specificities (Cicortas Gunnarsson et al., 2004). In addition, we have proven that the specificity of this scaffold can be extended also to target protein structures, showing the versatility of the CBM-scaffold, as a scaffold complementing the antibodies in the evolution of novel recognition molecules. The present report shows that the IgG4-specific CBMs bind to the protein itself and not to its carbohydrates, suggesting that the CBM4-2 molecule can also be used as scaffold to evolve specific binders towards substrates other than carbohydrates, a possibility rarely investigated for CBMs. There are in fact only two other reports where one and the same cellulose-specific CBM (the CBM of Cel 7A, from Trichoderma reesei, which is originally a type A CBM with a flat binding surface) has been successfully modified to either bind to bovine alkaline phosphatase (Smith et al., 1998) or to inhibit the activity of porcine alpha-amylase (Lehtio¨ et al., 2000). In this investigation we have successfully converted a type B, chain binding CBM, into a protein binder. Thus, the CBMs, just as antibodies, have the potential to recognise quite different molecular targets and are not confined to the recognition of one type of molecular species, namely carbohydrates. Preliminary studies demonstrated rapid binding kinetics of the IgG4-specific variants with their target (data not shown). As the library from which these variants were selected was small and since the requirement for rare additional mutations was strong, further limiting the functional size of the library in relation to this target, we cannot conclude that binders based on this scaffold inherently display a low affinity. Further studies using new library constructs will have to be carried out to resolve this issue properly. Sequence studies of the human IgG4-specific CBMs revealed a strong selection for leucin, an aliphatic amino acid, in the positions 76 and 110 where the wild type CBM harbours the aromatic residue phenylalanine. In this work we chose to replace these residues in the G-4 variant with alanine, which is a smaller aliphatic amino acid, and study how this influences the target binding. We found that leucin in position 110 plays an important role in the CBM’s binding to IgG4 since its mutation to alanine led to a complete loss of binding. Position 76 on the other hand seemed to tolerate the replacement of one aliphatic amino acid with another. Leucin was however the only amino acid of that character allowed in position 76 in the combinatorial library of CBM4-2 (Cicortas Gunnarsson et al., 2004), most likely explaining its dominance among the IgG4-specific variants selected from the library. The evolutionary pattern chosen by the IgG4-specific CBMs not only involved preference of specific amino acids in some of the mutated positions at the binding site but also the
introduction of one additional, undesigned mutation. This mutation was found in either positions 108, 136 or 138, located on the edge and outside the wild type binding site and in proximity to a loop harbouring residue 110, one of the key residues for the xylan binding of the wild type (Simpson et al., 2002) and also important for the IgG4specificity, as discussed above. The presence of this additional mutation in all the IgG4-specific CBMs substantiated its relevance for the target recognition since these CBMs were selected out of a library where most members lacked this kind of mutation. The importance of these mutations was also confirmed by the complete loss in binding to human IgG4 when the additional mutations in the positions 138 and 136 of the G-4 and G-11 variant, respectively, were eliminated. Mutations in these two positions were found in six out of seven sequences and presumably the other additional mutation found in the G15 variant, Q108R, is in the same way critical for the target recognition. The amino acids into which these residues have been mutated, especially the arginine in position 108, which is a large and charged amino acid, may be important contact residues in the interaction between the CBM and the IgG4 molecule. However, most of the mutations led to introduction of amino acids with small side chains not capable of creating any strong interactions. We can only speculate on whether these mutations allow larger conformational changes of the loop in their proximity to occur, resulting in a more open binding site. Since ligand specificity of most types of binders is largely determined by the three-dimensional shape of the ligand, and binding to proteins in general requires flat binding surfaces, this is a striking possibility. Structure determination of one IgG4-specific CBM is of course needed to confirm this hypothesis. 5. Conclusion We have shown that not only carbohydrate but also protein binders were selected from a rather small library, focused on diversification of residues at the carbohydrate binding site of the CBM4-2. The human IgG4-specific CBMs recognising protein structures required, however, an additional mutation outside the carbohydrate binding site. These findings let us believe that by extending the diversity introduced in the first generation library based on CBM4-2 we can efficiently create CBM-variants able to recognise both carbohydrate and/or protein structures on various target molecules. The ability to adopt its binding site to fit a quite diverse type of targets and its previously described properties, like high thermal stability and high productivity yield in E. coli, suggest that the CBMscaffold has a great potential in biomolecular applications. Acknowledgements We would like to thank Dr. Mats Andersson for fruitful discussions and Ann-Charlott Olsson for laboratory assistance. This study was supported by a grant from the Swedish Research Council.
L.C. Gunnarsson et al. / Biomolecular Engineering 23 (2006) 111–117
References Abou Hachem, M., Nordberg Karlsson, E., Bartonek-Roxa˚, E., Raghothama, S., Simpson, P.J., Gilbert, H.J., et al., 2000. Biochem. J. 345, 53–60. Boraston, A.B., Creagh, A.L., Alam, M.M., Kormos, J.M., Tomme, P., Haynes, C.A., et al., 2001. Biochemistry 40, 6240–6247. Cicortas Gunnarsson, L., Nordberg Karlsson, E., Albrekt, A.S., Andersson, M., Holst, O., Ohlin, M., 2004. Protein Eng. Des. Sel. 17, 213–221. Daugherty, P.S., Chen, G., Olsen, M.J., Iverson, B.L., Georgiou, G., 1998. Protein Eng. 11, 825–832. Engberg, J., Andersen, P.S., Nielsen, L.K., Dziegiel, M., Johansen, L.K., Albrechtsen, B., 1996. Mol. Biotechnol. 6, 287–310. Johansen, L.K., Albrechtsen, B., Andersen, H.W., Engberg, J., 1995. Protein Eng. 8, 1063–1067. Koide, A., Bailey, C.W., Huang, X., Koide, S., 1998. J. Mol. Biol. 284, 1141– 1151. ˚ ., 2000. Proteins 41, 316–322. Lehtio¨, J., Teeri, T.T., Nygren, P-A Linder, M., Salovuori, I., Ruohonen, L., Teeri, T.T., 1996. J. Biol. Chem. 271, 21268–21272.
117
Lipovsek, D., Plu¨ckthun, A., 2004. J. Immunol. Methods 290, 51–67. MacCallum, R.M., Martin, A.C., Thornton, J.M., 1996. J. Mol. Biol. 262, 732– 745. Mattheakis, L.C., Bhatt, R.R., Dower, W.J., 1994. Proc. Natl. Acad. Sci. U.S.A. 91, 9022–9026. ˚ ., Nord, K., Gunneriusson, E., Ringdahl, J., Sta˚hl, S., Uhle´n, M., Nygren, P.-A 1997. Nat. Biotechnol. 15, 772–777. Nordberg Karlsson, E., Bartonek-Roxa˚, E., Holst, O., 1997. Biochim. Biophys. Acta 1353, 118–124. ˚ ., Skerra, A., 2004. J. Immunol. Methods 290, 3–28. Nygren, P-A Simpson, H.D., Barras, F., 1999. J. Bacteriol. 181, 4611–4616. Simpson, P.J., Jamieson, S.J., Abou-Hachem, M., Karlsson, E.N., Gilbert, H.J., Holst, O., et al., 2002. Biochemistry 41, 5712–5719. Skerra, A., 2000. Biochim. Biophys. Acta 1482, 337–350. Smith, G.P., 1985. Science 228, 1315–1317. Smith, G.P., Patel, S.U., Windass, J.D., Thornton, J.M., Winter, G., Griffiths, A.D., 1998. J. Mol. Biol. 277, 317–332. So¨derlind, E., Carlsson, R., Borrebaeck CAK, Ohlin, M., 2001. Comb. Chem. High Throughput Screen 4, 409–416.
Evolution of a carbohydrate binding module into a protein-specific binder Lavinia Cicortas Gunnarsson a, Linda Dexlin a, Eva Nordberg Karlsson b, Olle Holst b, Mats Ohlin a,* a
Department of Immunotechnology, Lund University, P.O. Box 7031, S-220 07 Lund, Sweden b Department of Biotechnology, Lund University, P.O. Box 124, S-221 00 Lund, Sweden Received in revised form 9 November 2005; accepted 5 December 2005
Abstract A carbohydrate binding module, CBM4-2, derived from the xylanase (Xyn 10A) of Rhodothermus marinus has been used as a scaffold for molecular diversification. Its binding specificity has been evolved to recognise a quite different target, a human monoclonal IgG4. In order to understand the basis for this drastic change in specificity we have further investigated the target recognition of the IgG4-specific CBMs. Firstly, we defined that the structure target recognised by the selected CBM-variants was the protein and not the carbohydrates attached to the glycoprotein. We also identified key residues involved in the new specificity and/or responsible for the swap in specificity, from xylan to human IgG4. Specific changes present in all these CBMs included mutations not introduced in the design of the library from which the specific clones were selected. Reversion of such mutations led to a complete loss of binding to the target molecule, suggesting that they are critical for the recognition of human IgG4. Together with the mutations introduced at will, they had transformed the CBM scaffold into a protein binder. We have thus shown that the scaffold of CBM4-2 is able to harbour molecular recognition for either carbohydrate or protein structures. # 2005 Elsevier B.V. All rights reserved. Keywords: Binding specificity; Combinatorial library; Mutant; Phage-display; Molecular engineering; Protein scaffold
1. Introduction In vitro generation of proteins capable of specific target recognition has always been a challenge for the molecular biologists. The development of combinatorial library approaches in combination with in vitro selection methods, such as phage (Smith, 1985), ribosomal (Mattheakis et al., 1994) and cell-surface (Daugherty et al., 1998) display, has facilitated the creation and identification of such proteins. For a long time, antibodies, a group of naturally diverse binders well suited for therapeutic applications, have been a primary choice of proteins for use in this type of studies. In the growing field of biotechnology, there is a constant need of novel binders and the
Abbreviations: AE, affinity electrophoresis; BSA, bovine serum albumin; CBM, carbohydrate binding module; ConA, concanavalin A; ELISA, enzymelinked immunosorbent assay; HRP, horseradish peroxidase; IPTG, isopropyl-bD-thiogalactoside; PBS, phosphate-buffered saline; PCR, polymerase chain reaction; WGA, wheat germ agglutinin * Corresponding author. Tel.: +46 46 222 4322; fax: +46 46 222 4200. E-mail address: [email protected] (M. Ohlin). 1389-0344/$ – see front matter # 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.bioeng.2005.12.002
search for such proteins has been extended to include alternative scaffolds to the antibody framework (Nygren and Skerra, 2004). Recent reports have described the successful use of proteins such as lipocalins (Skerra, 2000), a single protein A domain (Nord et al., 1997) and the fibronectin type III domain (Koide et al., 1998), for generation of specific protein binders. Depending on their additional properties, such as size, stability and production yields in for example Escherichia coli, these molecules have potential for use in diagnostics and bioseparation applications. The carbohydrate binding modules (CBMs) are noncatalytic modules able to recognise several types of target structures. They are connected to a variety of glycoside hydrolysing or modifying enzymes, showing specificity towards one or several carbohydrate ligands. The CBMs are divided into different groups depending on the topology of their binding site, which in turn is related to the type of ligand they recognise, just like antibodies. Flat binding surfaces, found on antibodies binding mostly to proteins (MacCallum et al., 1996) (http://www.unizh.ch/antibody/Structures/ AgContact/), are also present on CBMs binding to insoluble
112
L.C. Gunnarsson et al. / Biomolecular Engineering 23 (2006) 111–117
Fig. 1. Amino acid sequence alignment of intentionally mutated regions of the primary structure of human IgG4-specific CBM-variants. The amino acid numbering in this figure is according to (Simpson et al., 2002). The genes encoding these proteins can be accessed in GenBank (accession numbers AY534556–AY534560, AY842281–AY842282). Other modifications, characteristic for IgG4-specific protein modules, are also shown.
crystalline surfaces (Linder et al., 1996; Simpson and Barras, 1999). CBMs that recognise free single carbohydrate chains display a binding cleft (Simpson et al., 2002), which is also the case for antibodies binding to peptides. Small ligands such as haptens in the case of antibodies (MacCallum et al., 1996) (http://www.unizh.ch/antibody/Structures/AgContact/), and mono- and disaccharides in the case of CBMs bind into solvent exposed binding slots (Boraston et al., 2001). Altering the binding ability of a single carbohydrate binding module is a relatively unexplored area. Except for our work, there are, to our knowledge, only two reports (Lehtio¨ et al., 2000; Smith et al., 1998) where molecular engineering of one and the same CBM scaffold has been tested. In both those cases CBM variants specific towards a protein target have successfully been engineered. As described in our previous publication (Cicortas Gunnarsson et al., 2004), we have constructed a combinatorial library by limited substitutions of 12 residues around the binding site of CBM4-2, a thermostable CBM from the modular xylanase Xyn 10A, originating from Rhodothermus marinus (Nordberg Karlsson et al., 1997). This CBM binds to its original substrate, xylan, both by three dimensionally fitting the helical structured polysaccharide into the binding groove and by establishing specific interactions between the substrate and some of the surface exposed residues at the binding site (Simpson et al., 2002). Successful selections were performed using phage display on two other polysaccharides, Avicel (50% amorphous and 50% crystalline cellulose) and ivory nut mannan, that like xylan are suitable for binding into clefts as the one displayed on the CBM4-2. Despite the small size of the library we were also able to select CBM-variants specific towards the glycoprotein human IgG4 (Fig. 1). While the CBMs selected on carbohydrates retained some of the wild type specificity towards xylan, the changes undergone by the IgG4-specific CBMs led to a complete loss of their ability to bind to xylan. In this study we have further analysed how the CBM can evolve to specifically bind a quite different target like the human IgG4 molecule. One important issue has been to demonstrate that these CBMs recognise the protein itself and not the carbohydrates of the target. We have also identified some of the key residues involved in the new specificity and/or responsible for the swap in specificity of the CBM from xylan to human IgG4 and we have shown that the critical modifications extend beyond the surface directly involved in the recognition of the original carbohydrate substrate.
2. Materials and methods 2.1. Strains and vectors E. coli strain Top10F0 (Invitrogen, Paisley, UK) was used as host for construction of single mutants and for phage display work. The phagemid used was a variant of pFab5c.His (Engberg et al., 1996) carrying the part of gene III encoding only the final C-terminal domain of M13 protein 3. Production of soluble recombinant protein was achieved using E. coli strain BL21(DE3) (Novagen, Madison, WI, USA) and the expression vector pET-22b(+) (Novagen).
2.2. Phage stocks Phage stocks, each displaying a single CBM variant, were created by infecting exponentially growing cultures of a specific clone (2 YT, 1% (w/v) glucose, 100 mg/ml ampicillin, 10 mg/ml tetracycline at OD600 0.5) with helper phages VCSM13 (multiplicity of infection: approximately 20) (Stratagene, La Jolla, CA, USA) for 30 min at 37 8C. The medium was changed (2 YT, 100 mg/ml ampicillin, 10 mg/ml tetracycline, 50 mg/ml kanamycin and 0.25 mM isopropyl-b-D-thiogalactoside (IPTG)) and the cultures were grown at 30 8C overnight. After centrifugation, the supernatants containing the phage particles were filtered through 0.45 mm filters and stored at 4 8C.
2.3. Denaturation and deglycosylation of human IgG4 The monoclonal antibody previously used in the selection of the IgG4specific variants of CBM (Cicortas Gunnarsson et al., 2004) is a human IgG4l antibody (Dynal A/S, Oslo, Norway) specific for an epitope shared by all mouse IgG subclasses. This human IgG4 in soluble form (Dynal) was denatured by incubation at 70 8C in 8 M urea for 30 min prior to coating on a 96 well plate. Enzyme-linked immunosorbent assay (ELISA) using phage-displayed CBMs (see below) was performed to investigate whether the IgG4-specific CBMs could still bind to the IgG4. The presence of carbohydrates in the denatured protein was detected by using the horseradish peroxidase (HRP) conjugated lectins wheat germ agglutinin (WGA) and concanavalin A (ConA) (Sigma– Aldrich Inc., St. Louis, MO, USA). Deglycosylation of 4 mg of human IgG4 (Dynal) with 0.5 units of Nglycosidase F (PNGaseF; Roche Diagnostic Corporation, IN, USA) was performed overnight at 37 8C prior to coating. The HRP conjugated lectin ConA was used to monitor the removal of N-linked carbohydrates by ELISA. In both cases, 96 well ELISA plates were coated at 4 8C overnight with 1 mg/ml of native, denatured or de-N-glycosylated human IgG4 (Dynal) in phosphate-buffered saline (PBS). After washing the plates with washing buffer (154 mM NaCl and 0.05% (v/v) Tween-20 in PBS) the phages diluted in a blocking buffer (PBS containing 1% bovine serum albumin (BSA) and 0.05% (v/v) Tween-20) were added and the plates were incubated for 2 h at room temperature. A second washing step was followed by addition of rat anti-M13 protein 8 antibody diluted in blocking buffer for 1 h at room temperature. Plates were washed and HRP conjugated rabbit-anti-mouse immunoglobulins that cross-react with rat IgG (DAKO A/S, Glostrup, Denmark) diluted in blocking buffer were added to the wells and incubated for 1 h at room temperature. For detection, substrate (0.67 mg ortophenylene-diamine/ml 35 mM citrate and 67 mM phosphate buffer pH 5 and 0.012% H2O2) was added. The reaction
L.C. Gunnarsson et al. / Biomolecular Engineering 23 (2006) 111–117 was stopped by addition of H2SO4 to a final concentration of 0.6 M and the absorbance was recorded at 490 nm.
2.4. Construction of mutants Site-directed mutagenesis was used to construct mutants originating both from the wild type CBM4-2 (wt E138A, wt E138G, wt D136G and wt Q108R) and from G-4, a human IgG4-specific variant of CBM (G4 A138E). In order to introduce these point mutations, polymerase chain reactions (PCRs) were performed using a set of primers (Table 1) and the phagemid vector encoding the wild type or the G-4 variant (human IgG4-specific) of CBM4-2 as templates. Initially two fragments, one harbouring the desired mutation, were amplified for each mutant using one gene- and one vector-specific primer (Table 1) and platinum Pfx DNA polymerase with proofreading (Invitrogen) (94 8C for 2 min; 20 or 25 cycles: 94 8C for 30 s, 55 8C for 30 s, 68 8C for 2 min; 68 8C for 10 min). The fragments were then assembled in an overlap-extension PCR (94 8C for 2 min; 8 cycles: 94 8C for 30 s, 52 8C or 55 8C for 30 s, 68 8C for 2 min; 68 8C for 10 min) and the resulting PCR-products were amplified (94 8C for 2 min; 25 cycles: 94 8C for 30 s, 55 8C for 30 s, 68 8C for 2 min; 68 8C for 10 min) using the pFab 5C vector-specific primers. The PCR-products were digested with SfiI and NotI before they were inserted into the phagemid vector, digested with the same restriction enzymes, and transformed into competent E. coli Top10F0. Another set of mutants from G-4 (G4 L76A and G4 L110A) and from the IgG4-specific CBM-variant G-11 (G11 G136D) were constructed as described above but instead using AmpliTaq DNA polymerase (Applied Biosystems,
Table 1 Oligonucleotide primers used to introduce site-directed mutationsa Mutant
Gene-specific primersb
wt E138A,G
Fragment 1: F 50 -AGTTTACGGTCA GTGATCAGGSGACGGTCATTCc Fragment 2: R 50 -CTGATCACTG ACCGTAAACTCGAACGTGAA Fragment 1: F 50 -CGTTCGAGTTTA CGGTCAGTGGTCAGGAGACGG Fragment 2: R 50 -ACTGACCGTAAA CTCGAACGTGAACGGCTG Fragment 1: F 50 -TCAGCTTCACG GTGGGGAACCGGTCGTTCCAGG Fragment 2: R 50 -GTTCCCCACCGT GAAGCTGACCACCGCCCC Fragment 1: F 50 -AGTTTACGGTCA GTGATCAGGAGACGGTCATTA Fragment 2: R 50 -CTGATCACTGA CCGTAAACTCGAACGTGAA Fragment 1: F 50 -TCGACATCGATGC GACGGCCGCCCCGGTGAACG Fragment 2: R 50 -GGCCGTCGCATCG ATGTCGAAGGGGTTGTT Fragment 1: F 50 -TCACGGTGGGGAA CCAGTCGGCTGAGCATTACG Fragment 2: R 50 -CGACTGGTTCCC CACCGTGAAGCTGACCAC Fragment 1: F 50 -CGTTCGAGTTTAC GGTCAGTGATCAGGAGACGG Fragment 2: R 50 -ACTGACCGTAAA CTCGAACGTGAACGGCTG
wt D136G
wt Q108R
G4 A138E
G4 L76A
G4 L110A
G11 G136D
Vector-specific primers pFab5c.His pFab5c.His pET-22b(+) pET-22b(+) a b c
F 50 -TTTCACACAGGAAACAGCTATG R 50 -GCCTTTAGCGTCAGACTGTAG F 50 - TAATACGACTCACTATAGGG R 50 -CTAGTTATTGCTCAGCGGT
Prefixes F and R stand for the forward and reverse primers, respectively. Mutated codons are underlined. S = G or C.
113
Foster City, CA, USA) and the pET-22b(+) vector-specific primers. The PCRproducts in these cases were digested with NdeI and XhoI and the genes were inserted into the pET-22b(+) vector and transformed into competent E. coli BL21(DE3). All constructs were validated by DNA sequencing.
2.5. Binding assays using phage-displayed proteins The binding of phage-displayed mutants to xylan, the wild type substrate, was studied by ELISA using plates coated with the soluble part of birchwood xylan (Sigma–Aldrich). After incubation for 2 h at 37 8C with the phages in blocking buffer followed by a washing step, the bound phages were detected by incubation with HRP conjugated mouse anti-M13 antibody (Amersham Bioscienses, Piscataway, NJ, USA) for 1 h at 37 8C. The detection was then carried out as described above. Human IgG4 binding of the mutants was assayed by incubating 10 ml of phage-stock with 5 ml (2 106) Dynabeads coated with human monoclonal IgG4 (Pan Mouse IgG; Dynal) in 1 ml of PBS containing 1% BSA and 0.05% (v/v) Tween-20. After 1 h of incubation on an end-to-end rotor at room temperature, the Dynabeads were washed four times with PBS containing 1% BSA and 0.05% (v/ v) Tween-20 and twice with PBS. Bound phages were eluted using 100 ml trypsin (0.5 mg/ml) (Invitrogen) by incubation for 30 min at room temperature with shaking (Johansen et al., 1995). After removing the substrate, the activity of trypsin was neutralised by addition of 100 ml aprotinin (0.1 mg/ml) (Roche Diagnostic Corporation, IN, USA). Both input and output phages were titrated on exponentially growing E. coli Top10F0 and the colony forming units were counted and used to calculate the binding ratio as bound phages per input phages.
2.6. Protein production and purification Genes encoding CBM variants cloned into the pFab5c.His vector were first re-cloned in-between the NdeI and XhoI sites found in the pET-22b(+) vector. These CBMs and other mutated variants encoded by genes directly cloned into the pET-22b(+) vector (see above), each carrying a C-terminal hexa-histidine tag were produced in E. coli BL21(DE3). Soluble proteins were purified using metal-ion-affinity chromatography as described earlier (Abou Hachem et al., 2000). The only differences in this work were that 20 mM sodium phosphate buffer (pH 7.4)/20 mM imidazole/0.75 M NaCl was used as binding buffer and the chromatography columns were packed with Ni-NTA (Qiagen, Hilden, Germany). The concentration of each purified protein was determined spectrophotometrically at A280 nm using extinction coefficients individually calculated for each CBM-variant.
2.7. Binding assays using purified proteins Once the mutants were purified as soluble proteins (see above), affinity electrophoresis (AE) was performed to confirm the results on binding to xylan obtained in the phage-display system. The AE method was performed in the BioRad (Hercules, CA, USA) mini-gel apparatus as earlier described for the wild type CBM4-2 (Abou Hachem et al., 2000). Purified CBM-variants (3 mg per gel) were separated at room temperature and 90 V on two different native gels, with or without 0.2% (w/v) oat spelt xylan (Sigma–Aldrich). A Kaleidoscope prestained standard (Bio-Rad) was used as a negative, non-interacting, control and the proteins were detected by staining with simply blue safe stain (Invitrogen). The studies on human IgG4-binding using phage-display, were confirmed by ELISA using purified CBM-variants (5 mg/ml in PBS) coated onto plates at room temperature overnight. After washing the plates human IgG4, 1 mg/ml diluted in blocking buffer, was added and the plates were incubated for 2 h at room temperature. The bound human IgG4 molecules were detected by adding HRP conjugated goat anti-human IgG antibody (Zymed, South San Francisco, CA, USA) for 1 h at room temperature and using the detection method for ELISA described above.
3. Results 3.1. Specificity studies of IgG4-binding CBMs As reported previously (Cicortas Gunnarsson et al., 2004), variants of CBM4-2 (Fig. 1) specific towards monoclonal
114
L.C. Gunnarsson et al. / Biomolecular Engineering 23 (2006) 111–117
human IgG4 carrying a l light chain have been selected from a combinatorial library. The binding specificity of these molecules was tested by ELISA using plates coated with the selection target IgG4, another monoclonal human IgG4l and polyclonal human IgG. The selected CBMs showed binding specificity towards the IgG4 used for selection while none of the other molecules gave raise to any detectable binding in this assay (data not shown). These results suggested that the CBMs recognise a protein or carbohydrate structure specific for the target IgG4-molecule and ruled out recognition of conserved structures found in the constant part of IgG4l. 3.2. Target epitope In order to determine whether the IgG4-specific CBMs recognise a protein or a carbohydrate structure on the glycoprotein target, ELISA was performed on plates coated with denatured human IgG4. Denaturation of the human IgG4 molecule using urea destroys the total protein structure, hardly affecting the carbohydrate structures present. Human IgG4specific CBMs showed no binding to these plates while the lectins WGA and ConA still bound, indicating a remained presence of carbohydrates (Fig. 2A). This suggests that the IgG4-specific CBM-variants actually recognise a protein structure of the target.
Fig. 2. Binding of phage-displayed wild type CBM4-2 and CBM variants selected on human IgG4 (G-1, G-4, G-11 and G-15) and lectins to native (black) and denatured (striped) human IgG4 (A) or native (black) and de-N-glycosylated (grey) human IgG4 and only PNGaseF (white), as a control coating, (B) determined by ELISA. The error bars represent the standard deviation of experiments performed in quadruplicates (A) or triplicates (B).
The finding that the IgG4-specific CBMs recognise the protein structure was also confirmed by using the deglycosylation enzyme PNGaseF to remove the N-linked carbohydrates (the major type of carbohydrates on immunoglobulins) from the human IgG4 molecule. This time ELISA-plates were coated with both native and deglycosylated IgG4 and the lectin ConA served as a control molecule demonstrating the effectiveness of the deglycosylation procedure. In agreement with our findings above the human IgG4-specific CBMs, in contrast to ConA, retained most of its binding to IgG4 even after deglycosylation, when the substrate was lacking the carbohydrates (Fig. 2B). This proves once again that the selected CBMs recognise the protein itself and not the carbohydrates of IgG4. Thus the CBM4-2 scaffold can be evolved to recognise a protein structure, which is quite different from the xylan structure that is recognised by the unmodified module. 3.3. Binding studies of the single-mutants Specific sequence features of the CBM-clones selected on human IgG4 raised issues on how the specificity of these modules was established. Additional to the mutations introduced by the library design, all binders (24 sequenced clones, 7 unique sequences) selected on the human IgG4 monoclonal antibody (Fig. 1) had one of the following spontaneous mutations Q108R, D136G or E138A/G, all located in proximity to residue 110 (Fig. 3), one of the residues believed to be important in xylan binding (Simpson et al., 2002). These mutations were not found among selected carbohydratebinding variants (Cicortas Gunnarsson et al., 2004) suggesting that they were important for the IgG4-specificity. A number of rationally designed mutants were hence constructed to investigate the importance of the observed additional mutations for the specific CBM-binding to IgG4. All mutants cloned into the pFab5c.His vector were successfully
Fig. 3. The structure of CBM4-2 (PDB 1k45) showing the residues not included in the design of the library but additionally mutated in the human IgG4-specific clones (Q108, D136 and E138) in red. In addition, two residues (69 in green and 110 in yellow) important for the binding of the wild type CBM4-2 to xylan and situated on the loops defining the binding cleft, are also highlighted.
L.C. Gunnarsson et al. / Biomolecular Engineering 23 (2006) 111–117
displayed as functional proteins on phages (data not shown), allowing binding assays to be performed. The phage-ELISA using plates coated with soluble birchwood xylan as a substrate indicated that the individual mutations Q108R, D136G or E138A/G introduced in the wild type CBM did not influence its binding to xylan (Fig. 4A). Furthermore, these mutations were not sufficient to change the specificity of CBM4-2 towards recognition of human IgG4 (Fig. 4B). However, reversion of the mutation in positions 138 of the G-4 variant, the most commonly modified position among the IgG4-specific variants, to the wild type encoded residue glutamate caused a complete loss of binding specificity for IgG4 (Fig. 4B) without regaining binding ability to xylan (Fig. 4A). Thus mutations such as E138A observed in the G4-variant are important for the modified module’s ability to recognise human IgG4 but are not critically involved in the xylan binding surface of CBM4-2. To confirm these results and extend the study of residues responsible for the evolved human IgG4-specificity, the binding properties of CBM-variants were also assayed using purified proteins in soluble form. Two additional mutants created from the G-4 variant (G4 L76A and G4 L110A) and a single mutant created from the G-11 variant (G11 G136D) were also included at this stage. The human IgG4-specific CBMs (G-4 and G-11) and mutants created from these variants had no binding specificity for xylan, as judged from affinity electrophoresis
115
Fig. 5. Binding studies using soluble variants of CBM. Non-denaturing affinity electrophoresis in the presence (A) or in the absence of oat spelt xylan (B). Proteins: lanes 1 and 2, two previously selected xylan specific clones (X-6 and X-2, respectively); lane 3, wild type CBM4-2; lane 4, a human IgG4-specific clone (G-11); lane 5, single mutant of G-11 (G11 G138D); lane 6, Kaleidoscope prestained standard; lane 7, a human IgG4-specific clone (G-4); lanes 8–10, single mutants of the G-4 (G4 A138E, G4 L76A and G4 L110A, respectively). (C) Addition of human IgG4 molecules to microtiter plates coated with different CBM variants showed binding to the IgG4-specific clones (G-4 and G-11) but not to the mutants created by reversion of their additional mutations (G4 A138E or G11 G136D) or to the wild type CBM4-2. Substitution of L110, but not L76, for alanine in G-4 abrogated binding to the human IgG4.
(Fig. 5A and B). ELISA, using plates coated with soluble CBMs, clearly showed that elimination of the additional mutations found in the G-4 and G-11 variants, those in positions 138 and 136, respectively, led to a complete loss of binding to IgG4 (Fig. 5C). The same consequence was seen when the leucin in position 110 of the G-4 variant was replaced with an alanine, whereas this type of mutation seemed to be well tolerated in position 76. Both residues, L76 and L110 were conserved in all sequences of human IgG4-specific CBMs (Fig. 1). In conclusion, this means that the transformation of CBM4-2 into IgG4-binding variants required specific changes not only inside but also outside the carbohydrate binding cleft and they together pushed the CBM into becoming a proteinbinding module. 4. Discussion
Fig. 4. Binding studies on birchwood xylan (A) performed by ELISA using CBM variants, such as the wild type CBM4-2, the IgG4-specific clone (G-4) and single mutants of these two CBMs, displayed on phages. The error bars represent the standard deviation of experiments performed in quadruplicates. (B) The same phage-displayed clones were assayed for binding to human IgG4 when the substrate was immobilised on Dynabead particles.
It is well known that the field of biotechnology is in need of new binding proteins with well-defined specificity. The antibody scaffold has, for many years, served as a starting point when searching for new binders (So¨derlind et al., 2001). In addition, methods for creating molecular libraries and a series of selection procedures have contributed to the successful evolution of specific binders towards different target molecules such as proteins, peptides, nucleic acids and haptens (Nygren
116
L.C. Gunnarsson et al. / Biomolecular Engineering 23 (2006) 111–117
and Skerra, 2004; Lipovsek and Plu¨ckthun, 2004). However, raising specific antibodies towards carbohydrates has proven to be a challenge, presumably because binding to this kind of targets requires features difficult to harbour by antibodies. For this purpose the naturally occurring carbohydrate binding modules could be utilised and we have recently shown that a CBM can be successfully used as a scaffold to develop variants with new carbohydrate binding specificities (Cicortas Gunnarsson et al., 2004). In addition, we have proven that the specificity of this scaffold can be extended also to target protein structures, showing the versatility of the CBM-scaffold, as a scaffold complementing the antibodies in the evolution of novel recognition molecules. The present report shows that the IgG4-specific CBMs bind to the protein itself and not to its carbohydrates, suggesting that the CBM4-2 molecule can also be used as scaffold to evolve specific binders towards substrates other than carbohydrates, a possibility rarely investigated for CBMs. There are in fact only two other reports where one and the same cellulose-specific CBM (the CBM of Cel 7A, from Trichoderma reesei, which is originally a type A CBM with a flat binding surface) has been successfully modified to either bind to bovine alkaline phosphatase (Smith et al., 1998) or to inhibit the activity of porcine alpha-amylase (Lehtio¨ et al., 2000). In this investigation we have successfully converted a type B, chain binding CBM, into a protein binder. Thus, the CBMs, just as antibodies, have the potential to recognise quite different molecular targets and are not confined to the recognition of one type of molecular species, namely carbohydrates. Preliminary studies demonstrated rapid binding kinetics of the IgG4-specific variants with their target (data not shown). As the library from which these variants were selected was small and since the requirement for rare additional mutations was strong, further limiting the functional size of the library in relation to this target, we cannot conclude that binders based on this scaffold inherently display a low affinity. Further studies using new library constructs will have to be carried out to resolve this issue properly. Sequence studies of the human IgG4-specific CBMs revealed a strong selection for leucin, an aliphatic amino acid, in the positions 76 and 110 where the wild type CBM harbours the aromatic residue phenylalanine. In this work we chose to replace these residues in the G-4 variant with alanine, which is a smaller aliphatic amino acid, and study how this influences the target binding. We found that leucin in position 110 plays an important role in the CBM’s binding to IgG4 since its mutation to alanine led to a complete loss of binding. Position 76 on the other hand seemed to tolerate the replacement of one aliphatic amino acid with another. Leucin was however the only amino acid of that character allowed in position 76 in the combinatorial library of CBM4-2 (Cicortas Gunnarsson et al., 2004), most likely explaining its dominance among the IgG4-specific variants selected from the library. The evolutionary pattern chosen by the IgG4-specific CBMs not only involved preference of specific amino acids in some of the mutated positions at the binding site but also the
introduction of one additional, undesigned mutation. This mutation was found in either positions 108, 136 or 138, located on the edge and outside the wild type binding site and in proximity to a loop harbouring residue 110, one of the key residues for the xylan binding of the wild type (Simpson et al., 2002) and also important for the IgG4specificity, as discussed above. The presence of this additional mutation in all the IgG4-specific CBMs substantiated its relevance for the target recognition since these CBMs were selected out of a library where most members lacked this kind of mutation. The importance of these mutations was also confirmed by the complete loss in binding to human IgG4 when the additional mutations in the positions 138 and 136 of the G-4 and G-11 variant, respectively, were eliminated. Mutations in these two positions were found in six out of seven sequences and presumably the other additional mutation found in the G15 variant, Q108R, is in the same way critical for the target recognition. The amino acids into which these residues have been mutated, especially the arginine in position 108, which is a large and charged amino acid, may be important contact residues in the interaction between the CBM and the IgG4 molecule. However, most of the mutations led to introduction of amino acids with small side chains not capable of creating any strong interactions. We can only speculate on whether these mutations allow larger conformational changes of the loop in their proximity to occur, resulting in a more open binding site. Since ligand specificity of most types of binders is largely determined by the three-dimensional shape of the ligand, and binding to proteins in general requires flat binding surfaces, this is a striking possibility. Structure determination of one IgG4-specific CBM is of course needed to confirm this hypothesis. 5. Conclusion We have shown that not only carbohydrate but also protein binders were selected from a rather small library, focused on diversification of residues at the carbohydrate binding site of the CBM4-2. The human IgG4-specific CBMs recognising protein structures required, however, an additional mutation outside the carbohydrate binding site. These findings let us believe that by extending the diversity introduced in the first generation library based on CBM4-2 we can efficiently create CBM-variants able to recognise both carbohydrate and/or protein structures on various target molecules. The ability to adopt its binding site to fit a quite diverse type of targets and its previously described properties, like high thermal stability and high productivity yield in E. coli, suggest that the CBMscaffold has a great potential in biomolecular applications. Acknowledgements We would like to thank Dr. Mats Andersson for fruitful discussions and Ann-Charlott Olsson for laboratory assistance. This study was supported by a grant from the Swedish Research Council.
L.C. Gunnarsson et al. / Biomolecular Engineering 23 (2006) 111–117
References Abou Hachem, M., Nordberg Karlsson, E., Bartonek-Roxa˚, E., Raghothama, S., Simpson, P.J., Gilbert, H.J., et al., 2000. Biochem. J. 345, 53–60. Boraston, A.B., Creagh, A.L., Alam, M.M., Kormos, J.M., Tomme, P., Haynes, C.A., et al., 2001. Biochemistry 40, 6240–6247. Cicortas Gunnarsson, L., Nordberg Karlsson, E., Albrekt, A.S., Andersson, M., Holst, O., Ohlin, M., 2004. Protein Eng. Des. Sel. 17, 213–221. Daugherty, P.S., Chen, G., Olsen, M.J., Iverson, B.L., Georgiou, G., 1998. Protein Eng. 11, 825–832. Engberg, J., Andersen, P.S., Nielsen, L.K., Dziegiel, M., Johansen, L.K., Albrechtsen, B., 1996. Mol. Biotechnol. 6, 287–310. Johansen, L.K., Albrechtsen, B., Andersen, H.W., Engberg, J., 1995. Protein Eng. 8, 1063–1067. Koide, A., Bailey, C.W., Huang, X., Koide, S., 1998. J. Mol. Biol. 284, 1141– 1151. ˚ ., 2000. Proteins 41, 316–322. Lehtio¨, J., Teeri, T.T., Nygren, P-A Linder, M., Salovuori, I., Ruohonen, L., Teeri, T.T., 1996. J. Biol. Chem. 271, 21268–21272.
117
Lipovsek, D., Plu¨ckthun, A., 2004. J. Immunol. Methods 290, 51–67. MacCallum, R.M., Martin, A.C., Thornton, J.M., 1996. J. Mol. Biol. 262, 732– 745. Mattheakis, L.C., Bhatt, R.R., Dower, W.J., 1994. Proc. Natl. Acad. Sci. U.S.A. 91, 9022–9026. ˚ ., Nord, K., Gunneriusson, E., Ringdahl, J., Sta˚hl, S., Uhle´n, M., Nygren, P.-A 1997. Nat. Biotechnol. 15, 772–777. Nordberg Karlsson, E., Bartonek-Roxa˚, E., Holst, O., 1997. Biochim. Biophys. Acta 1353, 118–124. ˚ ., Skerra, A., 2004. J. Immunol. Methods 290, 3–28. Nygren, P-A Simpson, H.D., Barras, F., 1999. J. Bacteriol. 181, 4611–4616. Simpson, P.J., Jamieson, S.J., Abou-Hachem, M., Karlsson, E.N., Gilbert, H.J., Holst, O., et al., 2002. Biochemistry 41, 5712–5719. Skerra, A., 2000. Biochim. Biophys. Acta 1482, 337–350. Smith, G.P., 1985. Science 228, 1315–1317. Smith, G.P., Patel, S.U., Windass, J.D., Thornton, J.M., Winter, G., Griffiths, A.D., 1998. J. Mol. Biol. 277, 317–332. So¨derlind, E., Carlsson, R., Borrebaeck CAK, Ohlin, M., 2001. Comb. Chem. High Throughput Screen 4, 409–416.