Structural Basis for Blocked Excited State Proton Transfer in a Fluorescent, Photoacidic Non-Canonical Amino Acid-Containing Antibody Fragment

Abstract

The fluorescent non-canonical amino acid (fNCAA) L-(7-hydroxycoumarin-4-yl)ethylglycine (7-HCAA) contains a photoacidic 7-hydroxycoumarin (7-HC) side chain whose fluorescence properties can be tuned by its environment. In proteins, many alterations to 7-HCAA’s fluorescence spectra have been reported including increases and decreases in intensity and red- and blue-shifted emission maxima. The ability to rationally design protein environments that alter 7-HCAA’s fluorescence properties in predictable ways could lead to novel protein-based sensors of biological function. However, these efforts are likely limited by a lack of structural characterization of 7-HCAA-containing proteins. Here, we report the steady-state spectroscopic and x-ray crystallographic characterization of a 7-HCAA-containing antibody fragment (in the apo and antigen-bound forms) in which a substantially blue-shifted 7-HCAA emission maximum (~70 nm) is observed relative to the free amino acid. Our structural characterization of these proteins provides evidence that the blue shift is a consequence of the fact that excited state proton transfer (ESPT) from the 7-HC phenol has been almost completely blocked by interactions with the protein backbone. Furthermore, a direct interaction between a residue in the antigen and the fluorophore served to further block proton transfer relative to the apoprotein. The structural basis of the unprecedented blue shift in 7-HCAA emission reported here provides a framework for the development of new fluorescent protein-based sensors.

Introduction

Genetically encoded noncanonical amino acids (NCAAs)^1–3 possess chemical functionalities not found in naturally occurring proteins. The side chains of certain NCAAs contain fluorophores with spectroscopic properties that could be useful for studying biological systems.^4–7 Because such “fluorescent non-canonical amino acids” (fNCAAs) are incorporated within the protein backbone, their fluorescence properties can often be tuned by the surrounding protein environments.^5,6,8,9 This immediately suggests that this class of molecules could be used to develop new, protein-based fluorescent sensors of important protein functions; however, these efforts have likely been hampered by a relative lack of structurally characterized proteins that contain fNCAAs.

Among the fNCAAs that have been genetically encoded to date, L-(7-hydroxycoumarin-4-yl)ethyl glycine (7-HCAA)⁴ represents an attractive starting point for the rational design of new fluorescent, protein-based sensors. The 7-hydroxycoumarin (7-HC) moiety that comprises the side chain of 7-HCAA contains a photoacidic phenol whose pKa drops from ~7.8 in the ground state to ~0.4 in the excited state.¹⁰ In aqueous environments, a proton is rapidly transferred from 7-HC to a nearby water molecule in a process known as excited state proton transfer (ESPT).¹¹ Furthermore, the two protonation states of 7-HC emit light at different wavelengths; the neutral form has a λ_max-em at ~390 nm while the λ_max-em of the phenolate anion is ~450 nm.¹² Thus, the ability to block and unblock ESPT in a way that’s directly tied to protein function could lead to the development of protein-based sensors that exhibit dramatic changes in emission. Although we¹³ and others⁹ have reported 7-HCAA-containing proteins in which blue shifted emission spectra are observed, none of these proteins approach the ~390 nm emission maximum that would be expected for a protein in which ESPT was blocked. In the absence of structurally characterized proteins in which the ability to carry out ESPT has been reduced, it is difficult to imagine how protein-based sensors that take advantage of this unique feature of 7-HCAA could be generated.

Here, we report the identification of a mutant of a 7-HCAA-containing Fab antibody fragment variant whose fluorescence spectrum is consistent with one in which ESPT from 7-HCAA has been substantially attenuated. Structural characterization of this mutant provided direct evidence that a strong hydrogen bond between the 7-HCAA phenol and a backbone carbonyl precludes proton transfer from the excited state of the fluorophore. A second crystal structure of the antigen-bound form of the Fab also indicated that a hydrogen bonding interaction is formed between the 7-HCAA phenol and a lysine residue in the antigen that further blocks ESPT. To our knowledge, this represents the first report of a 7-HCAA-containing protein in which ESPT is impeded due to the surrounding protein environment. The structural information reported in this study should prove useful in future efforts aimed at rationally designing new 7-HCAA-containing proteins that serve as sensors of protein function.

Results and Discussion

We recently reported the apo and antigen-bound structures of an α-CD40L Fab antibody fragment (called 5c8) that had been substituted with 7-HCAA at position Ile^L98.¹⁴ In the original study,⁸ Ile^L98 was selected as a site of 7-HCAA substitution primarily due to its proximity to the Fab’s combining site and the fact that Ile^L98 did not appear to make direct interactions with the antigen. The 5c8Ile^L98 → 7-HCAA Fab (hereafter referred to as 5c8*) exhibits an ~1.7 fold increase in 450 nm emission upon binding its antigen.⁸ Two specific interactions between the Fab and the 7-HCAA side chain were observed in the apo crystal structure (PDB ID: 6bjz) that were not formed in the antigen bound structure. First, an unusually short (~2.5 Å) hydrogen bond was observed between the 7-HCAA phenol and the carboxylate of Glu^H50 (Figure S1). Second, a close packing interaction between 7-HCAA and Trp^H47 was also present that placed these two residues in van der Waals contact (Figures S1 and S2). Structural characterization of the 5c8*—CD40L complex (PDB ID: 6w9g) indicated that these interactions are dissolved upon antigen binding, which subsequently led to an increase in the 7-HCAA quantum yield. The ease with which 5c8* crystallized in numerous conditions in our previous studies suggested that this Fab would lend itself to additional studies aimed at elucidating how protein environments altered 7-HCAA’s fluorescence properties. Although the role that Glu^H50 might play in altering 7-HCAA fluorescence was previously studied,¹⁴ the potential contribution of Trp^H47 has yet to be explored.

Our previous structural characterization of 5c8*¹⁴ indicated that Trp^H47 adopts a T-shaped π-stacking interaction with the side chain of Tyr^H35 and a quasi-T-stack with the 7-HCAA^L98 coumarin ring (Figure S2). The closest distances between non-hydrogen atoms of the coumarin and Trp^H47 indole ring are ~3.7 Å. We developed a hypothesis that this close contact between 7-HCAA and Trp^H47 could result in fluorescence quenching, potentially through a mechanism such as photoinduced electron transfer.¹⁵ However, to our knowledge, no reports of quenching of umbelliferone or related 7-hydroxycoumarins by tryptophan (Trp) or indole have been communicated.

To test the hypothesis that Trp can quench 7-HCAA fluorescence, we mutated Trp^H47 to leucine, which was chosen because it is a bulky aliphatic residue that would not be predicted to be competent for photoinduced electron transfer. Absorbance spectra collected for the W^H47L mutant indicated that 7-HCAA^L98 is largely protonated in the W^H47L variant at pH 7.5 (Figure 1(a)). This suggests that the W^H47L mutation shifts the equilibrium towards the neutral form of 7-HCAA^L98 in both the apo and CD40L-bound forms.

Figure 1.

(a) Normalized absorbance (a) spectra for apo 5c8* W^H47L pH 7.5 (dark green), the 5c8* W^H47L/CD40L complex pH 7.5 (black) and 7-HCAA at pHs 5 (dashed red), 7.5 (dashed green) and 10 (dashed blue). (b) Normalized fluorescence emission spectra, λ_ex = 323nm, of apo 5c8* W^H47L (dark green), the 5c8* W^H47L/CD40L complex (black) and 7-HCAA, each at pH 7.5.

The emission spectrum of the W^H47L variant (Figure 1 (b)) shows a substantial blue shift to ~384 nm, although a shoulder corresponding to residual 450 nm emission is also clearly visible; this shoulder is significantly diminished upon CD40L binding. These data are consistent with emission occurring predominantly from the excited state of a protonated 7-HCAA species,¹² which suggests that W^H47L mutation causes 7-HCAA^L98 to occupy a position that substantially decreases the ESPT efficiency. Notably, in stark contrast to 5c8*—which emits in the blue range of the visible spectrum—the 384 nm emission of the W^H47L mutant is hardly visible by eye (Figure S3).

Crystallographic analysis of 5c8* W^H47L

To elucidate the molecular basis for the shift in the emission maximum of 5c8* W^H47L relative to the parent protein, we solved crystal structures of this variant in both the apo and CD40L-bound states (see Table S1 for data collection and refinement statistics). F_o-F_c omit maps generated in the vicinity of 7-HCAA indicate clear difference density for the 7-HCAA^L98 residue in both the 5c8* W^H47L apo and CD40L complexed data sets (Figure S4). Superimposition of the variable domains of apo 5c8* and 5c8* W^H47L reveals a void left by mutation of Trp^H47 to leucine that allows the 7-HCAA^L98 ring to pack more tightly against the heavy chain (Figure 2). Additionally, our structural characterization of the W^H47L mutant revealed a ~132° rotation about χ₃, which orients the phenol such that it is located at the opposite end of the coumarin binding pocket relative to the position seen in apo 5c8*. In this orientation, the 7-HCAA^L98 phenol hydrogen bonds to the main chain amide oxygen of Arg^H102 while the carbonyl oxygen of the pyrone ring packs into space made by the W47^HL substitution (Figure 2). It appears that this interaction alone accounts for the decrease in ESPT efficiency given that the pK_a of a backbone carbonyl should be significantly lower than the excited state pK_a of 7-HCAA (~0.4).¹⁰ Additional structural rearrangements include the side chain of Tyr^H35 rotating nearly 90° about χ ₂ to fill the space created by the W^H47L mutation. Absent in the apo 5c8* W^H47L structure is a buried water molecule that was observed to directly hydrogen bond with the Tyr^H35 and 7-HCAA^L98 side chains in the parent structure (Figure 2). In the W^H47L variant, this water is occluded by up-puckering of the Pro^L100 side chain, which moves to this conformation in response to the large change in the 7-HCAA^L98 side chain position. The salt bridge between Glu^H50 and Arg^H102 that was observed in a previous structure¹⁴ is maintained in the W^H47L variant; however, subtle conformational rearrangements within these residues relative to 5c8* place atoms of the Arg^H102 guanidium moiety in a significantly closer (3.4–3.9 Å) cation-π interaction¹⁶ with the 7-HCAA^L98 ring (Figure 2).

Figure 2.

Stereo image showing the superimposition of apo 5c8* W^H47L (green) with the apo 5c8* structure (transparent blue). Black dashed lines depict hydrogen bonds within the W^H47L structure. The dashed red line illustrates how the conformation of the W^H47L structure Pro^L100 displaces a bound water molecule from the position observed in the 5c8* structure.

Unlike in 5c8*, binding of the W^H47L variant to CD40L does not induce a large conformational change in the 7-HCAA^L98 side chain position (Figures 3 and S5). Repositioning of 7-HCAA^L98 into the void created by the W^H47L mutation allows the side chains of Glu^H50 and Arg^H102 to pack more closely with 7-HCAA^L98 (Figure S5). Moreover, the hydrogen bond observed in the apo W^H47L structure between the 7-HCAA^L98 phenol and the main chain carbonyl oxygen of Arg^H102 is maintained in the complex structure (Figure 3). Interaction between cD40L and 5c8* W^H47L is mediated by hydrogen bonds from the s-amino group of Lys143 in CD40L to the mainchain carbonyl of Ser^L95, the amide oxygen of the Asn^H103 side chain and the phenol of 7-HCAA^L98 (Figure 3). It is likely that the diminishment of the 450 nm emission shoulder that is observed upon CD40L binding (Figure 1(b)) can be explained by the interaction of the ε-amine of Lys143 in CD40L and the 7-HCAA phenol oxygen. Namely, this additional hydrogen bond would further decrease possibility of ESPT by locking the 7-HCAA phenol in a conformation in which proton transfer is highly disfavored

Figure 3.

Stereo image showing the superimposition of the 5c8* W^H47L structure (transparent green) with its CD40L complex (Fab in black; CD40L in grey). Black dashed lines depict hydrogen bonds within the complex structure.

Comparison of the strikingly different 7-HCAA conformations observed between 5c8* and the W^H47L variant when complexed with CD40L (Figure S5) led us to suspect that tighter packing of the 7-HCAA^l98 side chain along with a direct hydrogen bonding interaction with the antigen might manifest in a lower K_d for CD40L binding in the case of W^H47L. To test this, we measured the CD40L binding affinity of wild type 5c8, 5c8* and the 5c8* W^H47L variant using surface plasmon resonance (Figure S6 and Table 1). We obtained K_ds of 3 nM, 42 nM and 16 nM for the wild type, 5c8* and 5c8* W^H47L variants, respectively, which are in reasonably good agreement with previously reported data for wt 5c8 and 5c8*.⁸ Although neither 5c8* nor the 5c8* W^H47L variant binds CD40L as tightly as the wt protein, the W^H47L binds CD40L ~3-fold more tightly than 5c8* – a remarkable result considering that the sole difference is a point mutation in the core of the Fab. Moreover, this shows that fluorescent NCAAs such as 7-HCAA can be used not only as passive sensors to detect analyte binding but can directly participate in molecular recognition, contributing to affinity in ways that should be amenable to rational design.

Table 1.

K_d values of wild type 5c8 and mutants for CD40L

Analyte	Kd (nm)
Wild type 5c8	2.6
5c8*	41.6
5c8* WH47L	15.7

Exploration of the potential of 7-HCAA quenching by Trp

Given that the goal of this study was to assess whether Trp^H47 was responsible for quenching 7-HCAA fluorescence in 5c8*, the conformation adopted by 7-HCAA in the W^H47L mutant was somewhat disappointing. Specifically, the new environment in which 7-HCAA is found (Figure S5) makes a direct comparison of the quantum yields of 7-HCAA in 5c8* and 5c8* W^H47L difficult. Nonetheless, when un-normalized emission spectra of 5c8* and 5c8* W^H47L are compared (Figure S7), a higher absolute intensity is observed in the W^H47L variant, which supports the notion that Trp may contribute to quenching in apo 5c8*. In an effort to further explore the potential that Trp can quench 7-HCAA fluorescence, we also carried out a Stern-Volmer analysis in which 7-HCAA fluorescence was assayed in increasing concentrations of Trp (Figure S8). A decrease in 7-HCAA emission intensity was observed upon addition of Trp, which further supports the possibility that tryptophan can quench 7-HCAA in protein environments.

Conclusions

Here, we report the spectroscopic and structural characterization of a 7-HCAA-containing Fab antibody fragment in which excited state proton transfer from the 7-HC phenol is blocked by the surrounding protein environment. The 5c8* W^H47L Fab was generated in an effort to explore whether tryptophan residues can quench 7-HCAA fluorescence. Rather than providing direct insight into this question, our mutant Fab exhibited absorbance and fluorescence spectra that were consistent with blocked proton transfer from the 7-HC phenol. Structural characterization of the Fab in the apo form confirmed that an alternative conformation of 7-HCAA was favored in the 5c8* W^H47L Fab that led to a hydrogen bond between the 7-HC phenol and the protein backbone. A second structure of the 5c8* W^H47L mutant in bound form suggested that the mutant Fab likely bound its substrate with a higher affinity than the parent Fab; this was confirmed using SPR analysis. Given the large difference in emission wavelengths between the phenol and phenolate forms of 7-HCAA, these results provide a direct path to the development of novel protein-based fluorescent sensors of biological activities that possess differences in fluorescence emission maxima that can be visually assessed without the need for sophisticated instruments (Figure S3).

Experimental Methods

Mutagenesis, protein expression and purification

Site-directed mutagenesis, expression and purification of 5c8 variants and CD40L were carried out as previously described.¹⁴ Briefly, site-directed mutagenesis was performed using a modified QuikChange method.¹⁷ The 5c8 variants were expressed in the Top10 Escherichia coli strain from the pBLN200 vector while CD40L was expressed in Pichia pastoris from the pPic9K vector. 5c8 variants were purified sequentially over Protein G resin, cation exchange (MonoS) and gel filtration. Purifiec protein was characterized by SDS-PAGE, which confirmed the presence of a fluorescent protein o the correct size (Figure S9(A; mass spectrometric analysis further confirmed > 95% incorporation o 7-HCAA (Figure S9(B)). CD40L secreted into the Pichia expression media was concentrated, dia-lyzed into low salt buffer, then purified via anion exchange followed by gel filtration.

Steady state spectroscopy

Absorbance spectra for 5c8* W^H47L proteir preparations (400–600 μg/mL in 100 mM HEPES-NaOH pH 7.5 with 150 mM NaCl) were recorded with a SpectraMax M5 microplate reade (Molecular Devices) in a quartz cuvette (Starne Cell, Inc.) at room temperature. Protein fluorescence spectra (60–130 μg/mL upon dilution with 100 mM HEPES-NaOH pH 7.5 and 150 mM NaCl) were recorded on a NanoLoc spectrofluorometer (Horiba Scientific) in a quartz fluorometer cuvette (Starna Cell, Inc) at 25 °C.

Protein crystallization

Apo crystals of 5c8* W^H47L and its complexes with CD40L were obtained at room temperature by sitting-drop vapor diffusion. For apo 5c8* W^H47L crystals grew from 27–33% PEG 3350 0.1 M Citric acid/NaCitrate pH 3.0–4.0 Crystallization drops were set up at a two to one ratio of protein (2.0–4.5 mg/mL) to reservoii solution with a total volume of 3 μL. Diffraction quality crystals of the 5c8* W^H47L/CD40L complex grew in 3 μL drops with a two to one protein (1.2–2.5 mg/mL) to reservoir (16–24% PEG MME 2000, 0.1 M Trimethylamine N-oxide (TMANO), 0.1 M Tris-HCl pH 7.5–9.0) ratio Before flash freezing in liquid N2, crystals were equilibrated for 2 hours in either 40% PEG 3350 0.1 M HEPES-NaOH pH 7.5 for the apo 5c8* W^H47L or 40% PEG MME 2000, 0.1 M Tris-HCl pH 7.5, 0.1 M TMANO for the 5c8* W^H47L/CD40L complex.

Data collection and structure determination

Diffraction data from the 5c8* W^H47L apo crystals were collected at the Berkeley Center for Structura Biology (BCSB) from the Advanced Light Source (beamline 8.2.2) on an ADSC Q315R detector and data for the 5c8* W^H47L/CD40L complex were obtained at the Argonne National Laboratory Advanced Photon Source (beamline 19-ID) on a Pilatus 6 M detector. Crystals were flash frozen in liquid nitrogen prior to data collection at 100 K Diffraction data were indexed, refined, integrated and scaled using the HKL2000¹⁸ or MosFLM¹⁹ packages. Structures were solved by molecula replacement using Phaser²⁰ with search models comprising apo 5c8* (RCSB Protein Data Bank ID 6bjz) for apo 5c8* W^H47L and the 5c8*/CD40L complex (RCSB Protein Data Bank ID 6w9g) for the 5c8* W^H47L/CD40L complex. Model refinement was carried out using cycles of Refmac5²¹ interspersed with model building using the program Coot.²² All structural figures were generated with PyMOL.²³

Surface Plasmon Resonance

The interaction between 5c8 variants and recombinant CD40L were analysed by SPR using a Biacore T200 (GE Healthcare Life Sciences, Marlborough, MA) at 25 °C. Following pH scouting, CD40L was immobilized onto a CM5 sensor chip surface using standard amine coupling chemistry. A 1:100 dilution of 0.352 mg/mL CD40L into 10 mM sodium acetate pH 5.0 immobilized to a level of ~2600 RU. HBS-P (10 mM HEPES pH 7.4, 150 mM NaCl, 0.05% surfactant P20) was used as the immobilization running buffer. Kinetic experiments were performed in the presence of HBS-P. The flow rate of all analyte solutions was maintained at 50 μL/min. Analyte concentrations were 0 nM, 25 nM, 50 nM, 100 nM, 200 nM, and 400 nM. One 12 s pulse of 10 mM glycine at pH 2.5 was used for regeneration. To assess the effect of glycine treatment on the K_d values, 5c8-WT was injected in the beginning and end of the run. Sensorgrams from the overnight screening were evaluated using 1:1 kinetics model fitting.