Ratiometric Bioluminescent Sensor Proteins Based on Intramolecular Split Luciferase Complementation.

Bioluminescent sensor proteins provide attractive tools for applications ranging from in vivo imaging to point-of-care testing. Here we introduce a new class of ratiometric bioluminescent sensor proteins that do not rely on direct modulation of BRET efficiency, but are based on competitive intramolecular complementation of split NanoLuc luciferase. Proof of concept for the feasibility of this sensor principle was provided by developing a blue-red light emitting sensor protein for the detection of anti-HIV1-p17 antibodies with a 500% change in emission ratio and a Kd of 10 pM. The new sensor design also improved the dynamic response of a sensor for the therapeutic antibody cetuximab 4-fold, allowing the direct quantification of this anti-EGFR antibody in undiluted blood plasma. The modular sensor architecture allows easy and systematic tuning of a sensor's dynamic range and should be generally applicable to allow rational engineering of bioluminescent sensor proteins.

Optical sensor proteins allow real-time monitoring of molecular events in complex biological environments, with applications ranging from intracellular imaging to point-of-care diagnostics. While fluorescent sensors have become attractive tools for intracellular imaging, their dependence on external illumination has hampered applications in strongly absorbing or scattering media such as blood.[1,2] This issue has been successfully addressed by employing bioluminescent sensor proteins that use a luciferase to generate light by catalyzing the oxidation of luciferin substrates.[3,4] The development of bioluminescent sensor proteins that display a robust signal change upon ligand binding remains an important molecular engineering challenge, however. Intensiometric reporter luciferases have been constructed by insertion of a ligand binding domain within a conformationally sensitive part of the enzyme or by ligand-induced complementation of a split luciferase.[5−10] The first approach can be cumbersome because of its reliance on subtle conformational changes, whereas the more modular split luciferase approach requires at least two sensor parts, rendering its performance concentration dependent. More importantly, both approaches yield intensiometric sensors that are not ideal for quantitative measurements, as their signal intensity depends on many factors including sensor, substrate and product concentration, pH, temperature, and ionic strength.

Ratiometric bioluminescent sensor proteins based on the modulation of bioluminescence resonance energy transfer (BRET) between a luciferase and an acceptor fluorophore are more attractive from this perspective, as in these sensors the ratio of bioluminescence at two different wavelengths is used as a measure of ligand binding. A crucial aspect in the design of robust BRET sensors is the ability to control ligand-induced switching between a high BRET-state, in which luciferase and the fluorescent acceptor are in close proximity, and a low BRET-state in which they are well-separated. Johnsson and co-workers introduced the use of semisynthetic BRET sensor proteins (LUCIDS) in which intramolecular complex formation between a fluorescently labeled ligand analogue and a ligand binding domain is used to create a high BRET state in the absence of analyte.[11−14] Displacement of the tethered fluorescently labeled ligand by the analyte of interest increases the distance to the luciferase domain fused to the ligand binding, switching the sensor to a low BRET state. Despite the intrinsic modularity of this design, each new receptor domain still requires extensive optimization to ensure sufficient BRET in the off-state of the sensor. An alternative approach reported by our group is the use of so-called helper domains to keep a blue light emitting luciferase domain and a green fluorescent acceptor in close proximity in the off-state.[15−17] This approach was employed in the development of a sensor platform for antibody detection (LUMABS), in which bivalent binding of a target antibody disrupts the interaction of the helper domains, switching the luminescence from green to blue. However, the helper domains and the fluorescent acceptor domain increase the distance between donor luciferase and the acceptor fluorophore, resulting in suboptimal BRET efficiency, in particular for more red-shifted fluorescent acceptors.

Here we introduce a new class of ratiometric bioluminescent sensor proteins that do not rely on modulation of BRET efficiency but are based on competitive intramolecular complementation of split NanoLuc luciferase. NanoLuc is a highly stable and bright, blue light producing luciferase that can be split into two fragments, an 18 kDa large BiT (LB) and a 1.3 kDa small BiT (SB), which were designed as a binary complementation reporter to study protein interactions (NanoBiT or NB).[21,22] In our new sensor format (NB-LUMABS) a single copy of the large fragment (LB) is fused to two copies of the small fragment (SB) yielding a protein switch that can exist in two conformations, where either the N- or the C-terminal SB binds to the LB and reconstitutes luciferase activity (Figure ). A red-emitting fluorophore is conjugated next to one of the SBs, permitting efficient BRET and consequently the emission of red light in only one of the two conformations. Bivalent binding of an antibody disrupts the intramolecular interaction of the fluorescently labeled SB, which allows reconstitution of NanoLuc by the nonfluorescently labeled SB, resulting in a color transition from red to blue. An important property of the split NanoLuc system is that the affinity between the LB and SB can be systematically varied between Kd = 0.7 nM and Kd = 190 μM,[22] allowing precise tuning of the relative thermodynamic stabilities of the two alternative conformations.

Figure 1

Design of NB-LUMABS. (A) Schematic representation of sensor concept. One large BiT (LB) was fused to two small BiTs (SBs) and a red-emitting fluorophore (red star) is conjugated next to one of the SBs, enabling efficient BRET in the antibody-free state. Bivalent binding of an antibody to the epitopes in the semiflexible linker disrupts the interaction with SB2, changing the color of emission from red to blue. (B) Illustration of split NanoLuc showing the position of SB and the N-terminal Cy3 labeling site. The yellow spheres indicate the proposed binding site[18] of the furimazine substrate which was docked into the crystal structure of NanoLuc (PDB ID: 5B0U) using AutoDock Vina[19] and visualized by PyMOL.[20] (C) Tunability of split NanoLuc interactions. Kd values are not limited to the listed values.

To establish proof-of-concept and explore the design constraints for this new sensor concept, we started by developing NB-LUMABS targeting the anti-HIV1-p17 antibody, a well-characterized antibody that binds linear epitopes with high affinity. In our initial design we combined SB1 with a Kd of 190 μM and SB2 with a Kd of 2.5 μM. The higher affinity for SB2 was chosen to favor formation of the high-BRET LB-SB2 complex in the absence of antibody, but not too high to prevent activation of the sensor by bivalent antibody binding. SB1 was connected to the N-terminus of LB via a flexible (GGS)10 linker, while SB2 was connected to the C-terminus of LB via a semiflexible linker with the antibody-targeting epitopes at each end of the linker to enable bivalent binding to the target antibody.[15] The synthetic fluorophore Cy3 was chosen as the BRET acceptor as its emission peak at 565 nm is well separated from that of NanoLuc. Since BRET efficiency is strongly distance dependent, three constructs were designed with the fluorophore conjugated at different positions via cysteine-maleimide labeling. HIV-NB-LUMABS-1 containing one Cy3 molecule next to the N-terminus of SB2 showed efficient BRET in the initial antibody-free state with the Cy3 peak higher than the NanoLuc emission (Figure A), suggesting Cy3 was kept in close proximity to the substrate binding site of NanoLuc. Sensor variants labeled with either one Cy3 molecule next to C-terminus of SB2 (HIV-NB-LUMABS-2) or two Cy3 molecules at both the N- and C-termini of SB2 (HIV-NB-LUMABS-3) showed a substantially lower Cy3 emission peak in the absence of antibody (Figure B and C). The decreased BRET efficiency observed for HIV-NB-LUMABS-3 compared to HIV-NB-LUMABS-1 is remarkable and suggests that conjugation of Cy3 at the C-terminal position may affect the interaction of SB2 with LB. Titration of HIV-NB-LUMABS-1 with increasing concentrations of anti-HIV1-p17 antibody revealed a large dynamic range (DR) of 218 ± 12% and an overall dissociation constant (Kd) of 10.0 ± 0.5 pM (Figure D). Since the dissociation constant for the monovalent antibody–peptide interaction is 42 nM,[15] the bivalent interaction between antibody and the sensor thus resulted in a more than 1000-fold increase in overall affinity.

Figure 2

NB-LUMABS for anti-HIV1-p17 antibody. (A–C) Luminescence spectra of 1 pM HIV-NB-LUMABS-1 (A), -2 (B), and -3 (C) in the absence (red line) and presence (blue line) of 1.25 nM anti-HIV1-p17 antibody. (D) Antibody titration to 1 pM of HIV-NB-LUMABS-1 in PBS buffer (pH 7.4, 1 mg/mL BSA). Error bars represent mean ± SD (n = 3). The red line in (d) represents the best fit to eq 1 (SI).

We next tested whether the already robust ratiometric response could be further enhanced by varying the relative affinities of the split luciferase interactions (Table ). Increasing the affinity of SB2 from 2.5 μM to 0.18 μM in HIV-NB-LUMABS-4 resulted in a high emission ratio in the absence of antibody but a moderate ratio change upon addition of antibody (Figure A). This result shows that the increased affinity of SB2 further pushes the equilibrium to the high BRET state in the absence of antibody, yet at the same time also hinders bivalent antibody binding to form the low BRET state. Sensor variants in which SB1 and SB2 had the same affinity (HIV-NB-LUMABS-5 and 6, with Kd’s of 190 μM and 2.5 μM, respectively) showed a strong attenuation of the emission ratio in the absence of antibody (Figure B,C), suggesting that in this case a substantial fraction resides in the low BRET, SB1-bound state. The optimal combination of SB1 and SB2 affinities was found for HIV-NB-LUMABS-7 (SB1: Kd = 2.5 μM; SB2: Kd = 0.18 μM), which combined an increased emission ratio of 1.8 in the absence of antibody, with efficient formation of the low-BRET, bivalent antibody-bound state, yielding a 493 ± 13% change in emission ratio (Figure D).

Table 1

Properties of NB-LUMABS Sensorsa

Target	Sensor name	SB1 affinityb	SB2 affinityb	Cy3 position	DR (%)	Sensor Kd
Anti-HIV1-p17	HIV-NB-LUMABS-1	Kd = 190 μM	Kd = 2.5 μM	before SB2	218 ± 12	10.0 ± 0.5 pM
	HIV-NB-LUMABS-2	Kd = 190 μM	Kd = 2.5 μM	after SB2	177 ± 4	13.7 ± 1.3 pM
	HIV-NB-LUMABS-3	Kd = 190 μM	Kd = 2.5 μM	before and after SB2	182 ± 18	12.1 ± 0.6 pM
	HIV-NB-LUMABS-4	Kd = 190 μM	Kd = 0.18 μM	before SB2	138 ± 5	14.2 ± 4.7 pM
	HIV-NB-LUMABS-5	Kd = 190 μM	Kd = 190 μM	before SB2	252 ± 15	11.7 ± 3.7 pM
	HIV-NB-LUMABS-6	Kd = 2.5 μM	Kd = 2.5 μM	before SB2	160 ± 3	15.2 ± 1.0 pM
	HIV-NB-LUMABS-7	Kd = 2.5 μM	Kd = 0.18 μM	before SB2	493 ± 13	11.8 ± 0.5 pM
Cetuximab	CTX-NB-LUMABS-1c	Kd = 190 μM	Kd = 2.5 μM	before SB2	233 ± 12	34.7 ± 3.7 nM
	CTX-NB-LUMABS-2c	Kd = 190 μM	Kd = 2.5 μM	after SB2	110 ± 3	20.7 ± 3.4 nM
	CTX-NB-LUMABS-3d	Kd = 190 μM	Kd = 2.5 μM	before SB2	88 ± 2	189 ± 16 nM

The measurements were performed in PBS buffer (pH 7.4, 1 mg/mL BSA) with sensor concentration of 1 pM for anti-HIV-p17 and 100 pM for cetuximab.

See ref (22).

Meditope sequence CVFDLGTRRLRC (monovalent Kd = 61 nM).

Meditope sequence CQFDLSTRRLKC (monovalent Kd = 270 nM).

Figure 3

Optimization of sensor response by tuning split NanoLuc interactions. (A–D) Luminescence spectra of 1 pM HIV-NB-LUMABS-4 (A), -5 (B), -6 (C) and -7 (D) in the absence (red solid line) and presence (blue solid line) of 1.25 nM anti-HIV1-p17 antibody. Dashed lines are HIV-NB-LUMABS-1 in the absence (red) and presence (blue) of antibody.

Having established the feasibility of using split-NanoLuc complementation as a new sensor principle to generate robust ratiometric bioluminescent proteins with low pM affinity, we also explored whether the same design principle could be used to improve the performance of LUMABS targeting the clinically important anticancer therapeutic antibody cetuximab. Since cetuximab binds to a discontinuous conformational epitope on the cancer marker EGFR, no linear epitope sequences are available. Nevertheless, we recently identified disulfide-linked cyclic meditope peptides with sufficient affinity to cetuximab and used them to construct a blue-green emitting LUMABS sensor for cetuximab with a relatively modest change in emission ratio (DR ∼ 60%).[17,23] As the presence of cysteine residues in the meditope peptides precludes the use of cysteine-maleimide chemistry to introduce the Cy3 dye, we instead introduced the noncanonical amino acid para-azidophenylalanine (pAzF) to allow site-specific conjugation with a DBCO-functionalized fluorophore via strain-promoted azide–alkyne click chemistry (SPAAC). In order to incorporate pAzF, a TAG amber stop codon was introduced either before the N-terminus (CTX-NB-LUMABS-1) or after the C-terminus of SB2 (CTX-NB-LUMABS-2) and the anti-HIV1-p17 epitope sequences were replaced by the cyclic cetuximab meditope sequences (monovalent Kd of 61 nM). Coexpression with the orthogonal tRNA synthetase/tRNA pair for pAzF allowed successful incorporation of pAzF into the cetuximab sensor variants at the desired position. Q-ToF LC-MS analysis of the purified proteins (Figure S2) showed partial reduction of pAzF to 4-aminophenyl alanine, which limited the degree of Cy3 labeling to 50–60% (Table S1). Nonetheless, CTX-NB-LUMABS-1 still showed bright Cy3 emission in the absence of cetuximab and a significant decrease in emission ratio upon antibody binding (Figure A). A dynamic range of 233 ± 12% was achieved, representing a 4-fold improvement compared to the blue-green CTX-LUMABS sensors,[17] while an overall Kd of 34.7 ± 3.7 nM obtained from fitting the titration data was found to be similar (Figure B). Introduction of Cy3 at the C-terminus of SB2 in CTX-NB-LUMABS-2 substantially decreased the BRET efficiency in the antibody-free state (Figure S4), which is consistent with the results obtained for HIV-NB-LUMABS-2. We also constructed a sensor variant (CTX-NB-LUMABS-3) containing a cyclic cetuximab meditope with a weaker affinity (monovalent Kd of 270 nM). Titration experiments with CTX-NB-LUMABS-3 showed a Kd of 189 ± 16 nM, enabling this sensor to reliably measure high cetuximab concentrations (Figure B). The lower affinity of CTX-NB-LUMABS-3 for cetuximab also resulted in a more rapid response compared to that of CTX-NB-LUMABS-1 (Figure S5). These results demonstrate that the NB-LUMABS sensor format can be reconfigured to detect antibodies recognizing disulfide constrained cyclic peptides, yielding two cetuximab sensors whose combined responsive regime covers the clinically relevant cetuximab concentration range (25 nM – 2.3 μM).[24]

Figure 4

NB-LUMABS for cetuximab. (A) Luminescence spectra of 100 pM CTX-NB-LUMABS-1 in the absence (red line) and presence (blue line) of 8.3 μM cetuximab. (B) Cetuximab titration to 100 pM of CTX-NB-LUMABS-1 (■) and CTX-NB-LUMABS-3 (●) in PBS buffer (pH 7.4, 1 mg/mL BSA). The emission ratios were obtained using a plate reader. (C) Linear regression fit to compare obtained emission ratios of CTX-LUMABS-1 in response to cetuximab addition using plate reader and camera measurement. (D) Cetuximab titration to 5 nM of CTX-NB-LUMABS-1 (■) and CTX-NB-LUMABS-3 (●) in undiluted blood plasma. The emission ratios were obtained using a plate reader. Error bars represent mean ± SD (n = 3). Inset: pictures of CTX-NB-LUMABS-1 (top) and CTX-NB-LUMABS-3 (bottom) with varying cetuximab concentration taken by a digital camera. The red lines in B and D represent the best fits to eq 1 (SI). The shaded area represents the physiologically relevant concentration range.

As an important step toward point-of-care application, we tested whether the bioluminescent signal could be detected by a simple digital camera instead of a dedicated plate reader. Due to the lower enzymatic activity of NanoBiT compared to intact NanoLuc the bioluminescent intensity of NB-LUMABS is less than that of LUMABS (Figure S6), but its brightness was still sufficient to be detected by a digital camera using only 100 pM of sensor. Moreover, the change in color from red-to-blue is more easily recognized compared to the green-to-blue transition of CTX-LUMABS (Figures B and S7). Image analysis software (ImageJ) was used to determine the average intensity in the red and blue color channels and the calculated red/blue ratio was plotted as function of antibody concentration. The Kd values determined by camera-based measurements matched well with those obtained using the plate reader (Figure C, Table S2). Finally, the performance of NB-LUMABS was tested in undiluted blood plasma. Although the absolute intensity of the emitted light was lower in blood plasma than that in buffer, the dynamic ranges and apparent affinities were found to be comparable to those obtained in buffer (Figure D, Figure S8, Table S2). Detection of the antibody-dependent color changes using a digital camera again yielded comparable results with plate-reader-based recording (Table S2).

In conclusion, a new class of ratiometric bioluminescent sensors has been developed that uses competitive intramolecular complementation of a split luciferase as an attractive alternative to the classical BRET sensor design. The new sensor format significantly improved the performance of previously reported LUMABS antibody sensors, allowing the use of a red-shifted acceptor and showing a large change in emission ratio to detect picomolar concentrations of antibody directly in solution. The intrinsic modularity of the sensor design not only allows easy and systematic tuning of the sensor’s dynamic range and antibody specificity, but also represents a generic design to construct ratiometric bioluminescent sensor proteins for applications ranging from in vivo imaging to developing cheap paper-based point-of-care diagnostics[25] for a wide range of molecular analytes.