Structure-based engineering of anti-GFP nanobody tandems as ultra-high-affinity reagents for purification

Green fluorescent proteins (GFPs) are widely used in biological research. Although GFP can be visualized easily, its precise manipulation through binding partners is still burdensome because of the limited availability of high-affinity binding partners and related structural information. Here, we report the crystal structure of GFPuv in complex with the anti-GFP nanobody LaG16 at 1.67 Å resolution, revealing the details of the binding between GFPuv and LaG16. The LaG16 binding site was on the opposite side of the GFP β-barrel from the binding site of the GFP-enhancer, another anti-GFP nanobody, indicating that the GFP-enhancer and LaG16 can bind to GFP together. Thus, we further designed 3 linkers of different lengths to fuse LaG16 and GFP-enhancer together, and the GFP binding of the three constructs was further tested by ITC. The construct with the (GGGGS)4 linker had the highest affinity with a KD of 0.5 nM. The GFP-enhancer-(GGGGS)4-LaG16 chimeric nanobody was further covalently linked to NHS-activated agarose and then used in the purification of a GFP-tagged membrane protein, GFP-tagged zebrafish P2X4, resulting in higher yield than purification with the GFP-enhancer nanobody alone. This work provides a proof of concept for the design of ultra-high-affinity binders of target proteins through dimerized nanobody chimaeras, and this strategy may also be applied to link interesting target protein nanobodies without overlapping binding surfaces.

LaG16 and GFP-enhancer can bind to GFPuv at the same time. To confirm that LaG16 and GFP-enhancer can bind to GFPuv noncompetitively in vitro, we used the FSEC method 22 . After the addition of only one kind of extra nanobody (LaG16 or GFP-enhancer) to GFPuv, the peak representing GFPuv emission exhibited an obvious shift compared to the peak of GFPuv alone, proving that either LaG16 or GFP-enhancer can bind to GFPuv (Fig. 2). In the sample with both LaG16 and GFP-enhancer added, all the GFPuv was incorporated into the LaG16-GFPuv-GFP-enhancer triple complex, whose peak shows a larger shift than that of the GFPuv-nanobody dimer. Therefore, the FSEC method confirmed that LaG16 and GFP-enhancer can bind to GFPuv at the same time.
Design of fusion nanobody based on the triple structure model. As nanobodies are powerful tools for the purification of GFP-tagged proteins and there are many commercialized nanobody resin products used for purification, we attempted to produce a fusion nanobody with heightened affinity to GFPuv for use in purifying protein with improved yield. Repeated (GGGGS) amino sequences can form a flexible linker between two proteins, and one turn of (GGGGS) has been found to be 19 Å long 23 . Based on the modelled structure of the LaG16-GFPuv-GFP-enhancer triple complex (Fig. 3A), we calculated that the distance from the N terminus of LaG16 to the C terminus of the GFP-enhancer (65.5 Å) is shorter than the distance from the N terminus of the GFP-enhancer to the C terminus of LaG16 (78.4 Å). Thus, we decided to add several (GGGGS) repeats between the N terminus of LaG16 and the C terminus of the GFP-enhancer. Too short a linker will cause tension when the fusion nanobody binds to GFPuv, while too long a linker will decrease the stability of the fusion nanobody. We added 4/5/6 (GGGGS) repeats between the two nanobodies ( Fig. 3B) and used the ITC method to select the best fusion nanobody with the most suitable linker.

Determination of the affinity constant between GFPuv and anti-GFP nanobody tandems.
To examine whether the fusion nanobodies had higher affinity for GFP than the individual ones, we measured the binding affinity of LaG16, GFP-enhancer, GGGGS 4 , GGGGS 5 and GGGGS 6 to GFPuv (GGGGS 4 , GGGGS 5 , and GGGGS 6 are the abbreviations of the fusion nanobodies GFP-enhancer-(GGGGS) 4 -LaG16, GFP-enhancer-(GGGGS) 5 -LaG16, and GFP-enhancer-(GGGGS) 6 -LaG16, respectively) ( Fig. 4, Table 1). GFPuv exhibits a Kd of 6.7 nM with LaG16 and a Kd of 24.3 nM with GFP-enhancer. All fusion nanobodies showed a greater affinity to GFPuv than the single GFP-enhancer or LaG16 nanobody. The Kd values of GGGGS 4 , GGGGS 5 and GGGGS 6 to GFPuv were 0.5 nM, 0.6 nM, and 1.2 nM, respectively. When the linker was too long, the LaG16 and GFP-enhancer in the tandem nanobodies could be treated as two separate and unrelated molecules and would not affect each other. When the linker length was properly optimized, as one of the nanobodies bound to GFP antigen, the linker restricted the movement of the tandem-linked nanobody to rotation and twisting in a small range. When the second nanobody's GFP binding site was nearby, there was a greater chance to simultaneously bind two nanobodies to one GFP molecule. As the fusion nanobody with the shortest linker, GGGGS 4 , showed the highest affinity with GFPuv, we chose GGGGS 4 for the nanobody-coupled resin application.
www.nature.com/scientificreports www.nature.com/scientificreports/ Application of the GGGGS 4 nanobody for membrane protein purification. We coupled GGGGS 4 or GFP-enhancer to NHS-activated Sepharose4 Fast Flow resin and used the resin to purify GFP-tagged zebrafish P2X4 receptor 24 , a membrane protein, from pelleted SF9 cell membrane. The eluted protein was analysed by SDS-PAGE (Fig. 5). The solubilized cell membrane showed a very weak band of GFP-P2X4 in the gel, while the eluted solution showed a strong band of GFP-zfP2X4, which means that both GGGGS 4 -coupled resin and GFP-enhancer-coupled resin can catch the GFP-tagged protein with high specificity. However, the GGGGS 4 -coupled resin had a higher yield, as the intensity of GFP-zfP2X4 analysed by ImageJ software was at about 1.5x that obtained with the GFP-enhancer (Table 2). We also performed and compared the purifications by the anti-GFP resins and by the TALON his-tag purification resin, which was previously employed for P2X4 purification 24 . The results showed that the anti-GFP resins yielded a much higher purity than the TALON resin ( Fig. 5, Table 2).

Discussion
In this work, we determined the structure of the GFPuv-LaG16 complex and revealed the interaction between the CDR regions of LaG16 and GFPuv. The model of the GFP-enhancer-GFPuv-LaG16 triple complex and FSEC testing confirmed that GFP-enhancer and LaG16 can bind to GFPuv at the same time. More importantly, we designed the fusion nanobody GGGGS 4 (GFP-enhancer-(GGGGS) 4 -LaG16) and tested it for purification of a GFP-tagged protein, obtaining a higher yield than the original GFP-enhancer. www.nature.com/scientificreports www.nature.com/scientificreports/ As GFP fusion expression screening techniques such as FSEC 24 . have been widely used in membrane protein structural biology, affinity purification using anti-GFP nanobodies has also become increasingly popular. In particular, after the Cryo-EM revolution, FSEC screening of functional membrane proteins suitable for single-particle Cryo-EM by fusion with GFP became the general strategy. However, the contents of important GFP-tagged membrane protein complexes in cultured mammalian cells are relatively low, and the yield of affinity purification by crosslinking a single GFP nanobody affinity resin with nanomole-scale affinity is not sufficient for Cryo-EM. Tandem nanobody binding to GFP with subnanomolar affinity significantly improved the yield and overcame this problem. The fusion nanobody GGGGS 4 in our study may provide a better choice for the purification of GFP-tagged proteins, particularly those with very low expression.
Additionally, direct manipulation of the in vivo target protein level is gradually becoming popular because DNA-and RNA-level manipulation, including knockout, knockdown and gene editing, is indirect, and unwanted side effects may cause incorrect results. Since GFP has been widely used to generate cell lines and animal models, controlling the expression level of target proteins fused with GFP may also simplify in vivo manipulation. Successful attempts have included directed protein degradation through anti-GFP nanobodies fused to E3 ligase. Several groups 25,26 have proven the usefulness of the nanobody-controlled degradation of specific nuclear proteins in mammalian cells and zebrafish embryos. With ultra-high-affinity nanobody chimaeras, the efficiency of this approach may be further improved.

Methods
Vector construction. The ORFs of LaG16, GFP-enhancer nanobodies and GFPuv were synthesized and inserted into the pET-28b vector between the NdeI and BamHI restriction sites by GENEWIZ, Inc. For the construction of fusion tandem nanobodies, (GGGGS) 4 , (GGGGS) 5 and (GGGGS) 6 were inserted between the C terminus of the GFP-enhancer and the N terminus of LaG16 by GENEWIZ, Inc. (Table 3).

Expression and purification. The plasmid was transformed into E. coli Rosetta (DE3) cells and plated
on Luria Bertani (LB) medium with 1.25% agar, 30 μg/ml kanamycin and 30 μg/ml chloramphenicol. Colonies of transformed Rosetta (DE3) cells were inoculated into LB medium. The next day, 1% of the cells cultured overnight were added to LB medium with 30 μg/ml kanamycin and incubated with shaking at 37 °C until the OD 600 nm reached approximately 0.6. Protein expression was induced by adding 0.5 mM isopropyl-b-D -1-thiogalactopyranoside (IPTG), and the cells were grown at 18 °C with shaking (220 rpm). Cells were harvested after 16 hours by centrifugation at 4000 × g for 10 min. Cell pellets were suspended in TBS (50 mM Tris pH 8.0, 150 mM NaCl) containing 1 mM phenylmethylsulfonyl fluoride (PMSF) and lysed using a High Pressure Homogenizer (JN-3000 PLUS, JNBIO, China) at 1,000 bar 5 times. The cell debris and inclusion bodies were removed by centrifugation at 35000 × g for 30 min. The supernatant was applied to a Ni-NTA (Qiagen) column pre-equilibrated with buffer A (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 30 mM imidazole). The mixture was rotated at 4 °C for 1 hour, the beads were washed to remove unbound protein with 10 CV of buffer A, and the protein was eluted with elution buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 300 mM imidazole). The eluted protein's His8 tag was removed in a 3.5 kD dialysis membrane (spectra/Por 7) by HRV3C protease at a mass ratio of target protein:HRV3C = 10:2 overnight at 4 °C. Then, 500 ml of dialysis buffer was added to remove imidazole (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 15 mM imidazole). The dialysis buffer was exchanged again during dialysis. On the next day, the digested protein was applied to a column equilibrated with dialysis buffer. Then, the column was rotated at 4 °C for 1 hour, and the flow-through fraction was collected and concentrated to www.nature.com/scientificreports www.nature.com/scientificreports/ 10 mg/ml using an Amicon Ultra 10 K filter (Millipore). Next, the protein was applied to a Superdex 75 Increase size-exclusion column (GE Healthcare) equilibrated with SEC buffer (20 mM HEPES pH 7.0, 150 mM NaCl). The target recombinant proteins with the tag removed were collected and concentrated to 10 mg/ml. Crystallization. LaG16 nanobodies and GFPuv (GFPuv: LaG16 = 1: 1.2; GFPuv: LaG16: GFP-enhancer= 1: 1.2: 1.2) were mixed and rotated at 4 °C for 1 hour. Then, the mixture was centrifuged at 41600 × g for 20 min, and the supernatant was applied to a Superdex 75 Increase size-exclusion column (GE Healthcare) equilibrated with SEC buffer (20 mM HEPES pH 7.0, 150 mM NaCl). The fractions containing the dimer/triple complex were collected and concentrated to 10 mg/ml. The crystals were obtained by vapour diffusion over a solution containing 0.3 M NaCl, 0.01M Tris-HCl 8.0, 27.5% w/v PEG4000 (for GFPuv-LaG16 complex).
Data collection and structure determination. All data sets were collected at SPring-8 BL32-XU (Hyogo, Japan). The data sets were processed with XDS programs 27 . The structure of the GFPuv-LaG16 complex was determined by molecular replacement using the Phaser program from the CCP4 crystallography package 28,29 with PDB ID code 6IR6 for GFPuv and the LaG16 model built based on chain C of 3K1K as the search models. www.nature.com/scientificreports www.nature.com/scientificreports/ The refinement was performed by Refmac 30 and Phenix 31 , and the model was further adjusted by COOT 32 . The related figures were drawn using PyMOL 33 . The structure refinement statistics are summarized in Table 4.
Isothermal titration calorimetry. The binding of nanobodies to GFPuv was measured using a Microcal ITC2000 microcolorimeter (GE Healthcare) at 20 °C. GFPuv and related nanobodies were purified as described above. We injected 280 μl of 5 μM GFPuv into the cell, and the ligand solution was 75 µM nanobody. The ligand was injected 20 times (0.4 μl for injection 1, 2 μl for injections 2-20), with 120 s intervals between injections. The baseline was obtained by adding ligand to SEC buffer. Before analysis, the baseline determined from  www.nature.com/scientificreports www.nature.com/scientificreports/ GFPuv-nanobody samples was subtracted. The data were analysed by the Origin7 software package (MicroCal). Measurements were repeated two times, and similar results were obtained.
Coupling nanobodies to NHS-activated sepharose4 fast flow beads. Since the activated NHS resin will form a covalent band with Tris buffer, we used HEPES instead of Tris during purification. Nanobodies were expressed as described above. Cells were harvested by centrifugation at 4000 × g for 10 min. Cell pellets were suspended in HBS (20 mM HEPES pH 7.0, 150 mM NaCl) containing 1 mM PMSF and lysed using a High Pressure Homogenizer (JN-3000 PLUS, JNBIO, China) at 1,000 bar 5 times. The cell debris was removed by centrifugation at 35000 × g for 30 min. The supernatant was applied to a Ni-NTA (Qiagen) column pre-equilibrated with buffer A (20 mM HEPES pH 7.0, 150 mM NaCl, 30 mM imidazole), and the mixture was rotated at 4 °C for 1 hour. Then, the beads were washed with 10 CV of buffer A, and the protein was eluted with elution buffer (20 mM HEPES pH 7.0, 150 mM NaCl, 300 mM imidazole). The eluate was placed in a dialysis membrane (spectra/Por 7) to remove extra imidazole using dialysis buffer (20 mM HEPES pH 7.0, 150 mM NaCl). Then, the digested protein was concentrated to 10 mg/ml. Anti-GFP and TALON resins were used to purify GFP-tagged zfP2X4. The expression and cell disruption of zfP2X4 were performed as described previously 24 . One hundred and eighty microlitres of pelleted membrane (presumably containing approximately 60 μg of GFP-tagged zfP2X4) was solubilized with 180 µl of S buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 30% glycerol, 4% DDM, 1 mM PMSF, 5.2 μg/ml aprotinin, 2 μg/ ml pepstatin A, 2 μg/ml leupeptin, and 0.5 U/m apyrase. Then, the unsolubilized membrane was removed by ultracentrifugation at 41600 × g for 20 min at 4 °C. The supernatant was divided evenly into three 1.5 ml EP tubes and incubated with 50 μl of anti-GFP resin (GFP-enhancer or GGGGS 4 tagged resin) equilibrated with wash buffer Ι (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 15% glycerol, 0.05% DDM) or 50 μl of TALON resin (Takara) equilibrated with wash buffer ΙΙ (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 15% glycerol, 0.05% DDM, 25 mM imidazole). The mixture was rotated at 4 °C for 1 hour, and then the resin was centrifuged at 200 × g for 2 min to remove the unbound protein. Then, 100 μl of wash buffer was added to the resin and centrifuged at 200 × g for 2 min to remove the supernatant. This washing step was repeated 5 times. Finally, the resin was applied to a spin column (Micro Bio-Spin Columns, BIO-RAD), and     Table 4. Data collection, phasing and refinement statistics. *The highest resolution shell is shown in parentheses.