Design of a ligand-dependent fluorescent biosensor, based on an engineered lipocalin (anticalin), for the sensitive detection of the Alzheimer β-amyloid peptide

Abstract

Edited by: Robert E. Campbell

Based on the Anticalin H1GA which tightly binds Aβ₄₀ and Aβ₄₂ peptides – both established biomarkers of Alzheimer's disease – we describe the design of a protein–dye conjugate as analytical reagent that shows strongly elevated fluorescence upon Aβ binding. An unpaired Cys residue was introduced at seven positions within the four loop segments that shape the ligand pocket of the engineered lipocalin. Five of these mutants were purified in the monomeric state and allowed the site-specific conjugation with IANBD amide as a solvatochromic fluorophore. Three conjugates showed ligand-dependent fluorescence and one of these, derived from H1GA(D45C), exhibited sixfold higher emission at 546 nm upon complex formation with the peptide while revealing a low K_D value of 1.2 ± 0.8 nM, even in the presence of 5% (w/v) albumin. This NBD-conjugated Anticalin offers a novel biosensor with potential for the detection of Aβ peptides in biochemical assays or human body fluid samples.

Introduction

The amyloid-β (Aβ) peptide, which is formed by abnormal processing of the amyloid precursor protein (APP) via β- and γ-secretases in the brain, stands at the beginning of the amyloid cascade, culminating in the extracellular self-aggregation of predominantly Aβ₄₀ and Aβ₄₂ peptides to yield neurotoxic oligomers, fibrils and, finally, senile amyloid plaques (Decourt et al., 2021; Hardy and Higgins, 1992). Due to the increasing prevalence of Alzheimer’s disease (AD) and the devastating nature of this form of dementia – both in the aging western population and also in developing countries (Bagga and Kumar, 2022; DeTure and Dickson, 2019) – there are ongoing efforts not only to develop therapies (Decourt et al., 2021; Golde and Levey, 2023; Karran and De Strooper, 2022), in order to deplete Aβ or to dissolve plaques, but also to provide early stage diagnostic procedures relying on the detection of Aβ peptides in biological samples (Cheng et al., 2022; Mattsson-Carlgren et al., 2022). Indeed, several attempts were made to develop biosensors that allow quantification of monomeric and aggregated forms of Aβ peptides – as well as other biomarkers – in patient samples, including electrochemical and immuno-infrared sensors (Beyer et al., 2023; Reja et al., 2024; Wang et al., 2025).

To offer an alternative to conventional antibody-based affinity reagents, we have developed a set of engineered lipocalins, so-called Anticalin® proteins, that specifically target a central epitope shared by both Aβ₄₀ and Aβ₄₂ peptides, comprising the amino acid sequence (V)FFAED (Rauth et al., 2016). These artificial binding proteins were derived from the human lipocalin 2 (Lcn2). Lcn2 is a small single-chain natural plasma protein (also known as neutrophil gelatinase-associated protein, NGAL, or as siderocalin, Scn) with less than 200 residues whose fold is dominated by an eight-stranded β-barrel which is open to the solvent and provides a ligand pocket at one end. This pocket is formed by four circularly arranged loops that are structurally variable among the lipocalin family of proteins (Schiefner and Skerra, 2015; Skerra, 2000) and which can be reshaped by combinatorial protein design and library selection techniques to provide novel ligand specificities (Achatz et al., 2022; Deuschle et al., 2021; Richter et al., 2014).

Among three different Aβ-specific Anticalins that were selected in our previous work one candidate, dubbed H1GA, revealed strong sub-nanomolar affinity towards the Aβ₄₀ peptide and exhibited both Aβ aggregation-blocking activity in vitro as well as protection against neurotoxic effects of aged Aβ₄₂ in cell culture assays (Rauth et al., 2016). Structural studies indicated that, in contrast to solanezumab and some other anti-Aβ antibodies with clinical potential, this Anticalin does not exhibit cross-reactivity with sequence-related human plasma proteins (Eichinger et al., 2022). Recently, this Anticalin was demonstrated to prevent neuronal hyperactivity in a mouse model of AD by intracerebral scavenging of the monomeric Aβ peptide in vivo (Zott et al., 2024). Together, its beneficial biochemical properties render this engineered protein scaffold not only an interesting drug candidate for therapeutic intervention of AD but also make it a potential diagnostic reagent for the specific and sensitive detection of Aβ peptides in samples from blood plasma or cerebrospinal fluid (CSF).

In this regard, the crystallographic analysis of the complex between H1GA and the Aβ₄₀ peptide has provided important mechanistic insight (Eichinger et al., 2022). It was found that the central peptide moiety (residues P17–P27: LVFFAEDVGSN, minimal epitope typed bold) is tightly buried within the ligand pocket of the Anticalin where it adopts a compact zigzag-bend conformation, whereas the unbound N- and C-terminal segments on both sides are conformationally disordered. This structural notion has prompted us to design a biosynthetic sensor protein by site-specifically incorporating a fluorescent group into the binding site of the engineered lipocalin – at a peripheral position – such that its fluorescence emission is strongly modulated upon complex formation with the Aβ peptide without compromising the binding activity.

Methods

Bacterial strains, plasmids and DNA manipulation

DNA manipulation and handling of transformed bacteria was performed according to standard procedures (Sambrook et al., 1989). E. coli strains XL1-Blue (Achatz and Skerra, 2024) and Origami B (Prinz et al., 1997; Ritz et al., 2001) were used for cloning/mutagenesis and recombinant protein production, respectively. For site-directed mutagenesis and protein biosynthesis in the bacterial periplasm the plasmid pNGAL98-H1GA was used, which encodes the Anticalin with an N-terminal OmpA signal peptide and the C-terminal Strep-tag II (Gebauer and Skerra, 2012; Schmidt and Skerra, 2007). For cytoplasmatic expression, the coding regions were subcloned onto pASK75-T7RBS-2 (Eggenstein et al., 2020; Jerschke et al., 2023), thereby replacing the signal peptide with a start Met residue. Amino acid substitutions were introduced into the Anticalin coding region using the QuikChange site-directed mutagenesis method (Agilent, Santa Clara, CA) with suitable synthetic oligodeoxynucleotide pairs (Eurofins Genomics, Ebersberg, Germany). Double-stranded DNA sequencing of all mutated structural genes was performed using the SmartSeq Kit (Eurofins Genomics).

Recombinant protein preparation

For the initial functional evaluation of amino acid substitutions, the lipocalin variants were produced via secretion into the periplasm of the transformed E. coli strain JM83 (Achatz and Skerra, 2024) following published procedures (Gebauer and Skerra, 2012). In the resulting purified proteins the single Cys residue was found to be modified via disulfide crosslink with glutathione, as revealed by ESI-MS, a phenomenon known from other published studies (Eser et al., 2009). Lipocalin mutants intended for NBD conjugation were subsequently produced by direct expression in the cytoplasm of Origami B (Jerschke et al., 2023). In this E. coli strain, which is a derivative of E. coli BL21 (Daegelen et al., 2009), the genes for thioredoxin reductase (trxB) and glutathione reductase (gor) were inactivated to provide a more oxidizing milieu in the cytoplasm (Prinz et al., 1997). This physiological environment promotes formation of the single structural disulfide bond in the Lcn2 scaffold (Eggenstein et al., 2020) while preserving additional Cys residues with their unique thiol groups that were introduced for the purpose of dye conjugation in the present study. Bacterial cultures of transformed Origami B were grown and harvested as previously described (Eggenstein et al., 2020). After preparation of a whole cell extract using mechanical lysis in a French Pressure cell (SLM Aminco, Urbana, IL), all mutated lipocalin proteins were purified via the C-terminal Strep-tag II using StrepTactin affinity chromatography (Schmidt and Skerra, 2007). The proteins were finally polished via size exclusion chromatography (SEC) on an ÄKTApurifier system (GE Healthcare, Freiburg, Germany), equipped with a S75 10/300 GL column (GE Healthcare) at room temperature, using 50 mM Tris/HCl, 115 mM NaCl pH 7.5 as running buffer at a flow rate of 0.5 ml/min. SDS-PAGE was performed using a high-molarity Tris buffer system (Fling and Gregerson, 1986), followed by staining with Coomassie brilliant blue. In the case of dye-labeled proteins, the SDS polyacrylamide gel was imaged on an Ettan DIGE fluorescence scanner (GE Healthcare) using the Cy2 channel. Protein concentrations were determined by measuring the UV absorbance at 280 nm using a calculated extinction coefficient of 28 545 M⁻¹ cm⁻¹ for all unlabelled lipocalin mutants as well as the dye conjugates. Electrospray ionization mass spectrometry (ESI-MS) was performed on a maXis quadrupole time-of-flight (Q-TOF) instrument (Bruker Daltonics, Bremen, Germany) in the positive ion mode. The purified protein was first dialyzed against 10 mM ammonium acetate (pH 6.6) and then 20% (v/v) acetonitrile as well as 0.1% (v/v) formic acid (both LC–MS grade from Sigma-Aldrich, Steinheim, Germany) were added. Samples were directly applied to the ESI source using a syringe pump operated at 180 μl/h (KD Scientific, Holliston, MA).

Real-time SPR measurements

Surface plasmon resonance (SPR) spectroscopy was performed on a BIAcore 2000 Instrument (GE Healthcare) essentially as previously described (Rauth et al., 2016) using phosphate-buffered saline (PBS) supplemented with 0.005% (v/v) Tween 20 as running buffer. The purified recombinant fusion protein MBP-Aβ₄₀ (Rauth et al., 2016) was covalently immobilized on a CM5 sensorchip (GE Healthcare) using amine coupling chemistry (∼4500 RU for the unconjugated lipocalins and ∼500 RU for the dye conjugates). Dilution series of the purified lipocalin mutants were applied at a flow rate of 30 μl/min, and data analysis was performed as published (Rauth et al., 2016).

Dye labeling of lipocalin variants

A solution of the purified lipocalin mutant at a concentration of 25–30 μM in PBS was mixed with an appropriate aliquot of a stock solution of 20 mM N,N′-dimethyl-N-(iodoacetyl)-N′-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)ethylenediamine (IANBD amide; Molecular Probes/Life Technologies, Darmstadt, Germany) in dry dimethyl sulfoxide (DMSO) to achieve a 5-fold molar ratio. After incubation at 4°C for 13–14 h, excess uncoupled dye was removed by passage through a PD-10 column (GE Healthcare), which was accompanied by a buffer exchange against 50 mM MES/NaOH pH 5.7 (mutants V40C, A127C, T132C) or pH 6.0 (mutants D45C and D72C; due to their slightly higher pI, in preparation for the subsequent ion exchange chromatography). Protein dimers that had formed by oxidative intermolecular disulfide formation between the exposed thiol groups were removed by cation exchange chromatography (CEX) on a Resource S column (GE Healthcare), 1 ml bed size, via elution in 50 mM MES/NaOH pH 5.7/6.0 with a concentration gradient of 0–1 M NaCl using the same chromatographic setup as described for SEC above. The eluted monomeric lipocalin-dye conjugate was finally dialyzed against 50 mM Tris/HCl, 100 mM NaCl, 1 mM EDTA pH 7.5.

Spectroscopic measurements

Absorption spectra of the purified lipocalin-dye conjugates were recorded using an Ultrospec 2100 pro spectrophotometer (Amersham Biosciences, Freiburg, Germany), having a fixed spectral bandwidth <3 nm, using a quartz cuvette with 10 mm path length (Hellma, Mühlheim, Germany). Fluorescence spectra were analyzed with a FluoroMax-3 spectrofluorimeter (HORIBA Jobin Yvon, Unterhaching, Germany) controlled by FluorEssence software v3.1 using a quartz cuvette with 100 μl sample volume, thermostatted at 25°C. Fluorescence emission was excited at 485 nm (slit width 5 nm) and recorded between 490 and 650 nm (slit width 20 nm), averaged over 2 scans (scan speed 100 nm/min).

Fluorescence titration of Anticalin-dye conjugates

Fluorescence titration was performed following a published procedure (Vogt and Skerra, 2001) using a FluoroMax-3 spectrofluorimeter, as described above, using a wavelength of 485 nm for excitation. At each titration point an emission spectrum from 490 to 650 nm was recorded and the intensity at the average maximum wavelength (see Table I) was measured. To this end, the purified lipocalin-dye conjugate was mixed (at 500 nM final concentration), in a total volume of 150 μl 100 mM Tris/HCl, 100 mM NaCl, 1 mM EDTA pH 7.5, with varying concentrations of the synthetic Aβ peptides Aβ₄₀ (Keck Biotechnology Resource Laboratory, Yale University, New Haven, CT) or Aβ₄₂ (EZBiolab, Carmel, IN), or the MBP-Aβ₄₀ fusion protein (Rauth et al., 2016), from concentrated stock solutions in the same buffer. Each titration was performed in duplicate and the mean of the two intensity values was calculated. These data were normalized to an initial fluorescence of 100% and fitted by non-linear least-squares regression using KaleidaGraph software ver. 5.0.4 (Synergy Software, Reading, PA) as previously described (Vogt and Skerra, 2001).

Table I.

Properties of Cys mutants of H1GA(F83Y) and their NBD conjugates as well as affinities towards MBP-Aβ₄₀.

	Cysteine mutant a					Anticalin-NBD conjugate a
	M calc [Da]	pI	Oligomeric status	Yield [mg/l]	[nM] b	M calc [Da]	M obs [Da]	Yield [mg/l]	[nm]	[nm]	[nM]
H1GA (F83Y)c	21567.2	7.08	monomer	7.1	0.19	21567.2	21568.0	4.3	496 ± 8	—	0.31
V40C	21571.2	7.07	monomer	3.2	0.24	21862.5	21862.6	1.1	497 ± 1	547 ± 1	0.23
R43C	21514.2	6.48	dimer, disulfide-bond isomers	0.5	2.33	21805.5	—	—	—	—	—
D45C	21555.3	7.86	monomer	2.8	0.90	21846.6	21846.6	0.5	498 ± 2	546 ± 1	0.88
D72C	21555.3	7.86	monomer	3.3	6.03	21846.6	21847.3	0.5	498 ± 2	552 ± 1	1.59
D73C	21555.3	7.86	dimer, disulfide-bond isomers	1.4	1.29	21846.6	—	—	—	—	—
A127C	21599.3	7.07	monomer	1.9	0.66	21890.6	21891.6	0.2	499 ± 1	549 ± 1	0.43
T132C	21569.3	7.07	monomer	4.8	2.18	21860.6	21860.6	0.6	496 ± 2	550 ± 1	0.34

^aAll values (except for K_D) correspond to Lcn2 variants produced in the bacterial cytoplasm, which led to higher yields (these proteins carried an additional start Met residue at the N-terminus, as confirmed by ESI-MS, corresponding to a mass difference of 131.2 Da).

^bK_D values determined for protein preparations initially obtained by periplasmic secretion (in these proteins the single Cys residue was modified via disulfide crosslink with glutathione, as revealed by ESI-MS).

^cInitial Anticalin version, basis for the mutants listed below.

Results

The binding site of Anticalins is shaped by four structurally variable loops that emerge at the open end of an eight-stranded β-barrel, where each loop connects two neighbouring anti-parallel β-strands (Achatz et al., 2022; Skerra, 2000). In the crystal structure of the Anticalin H1GA (Eichinger et al., 2022) the Aβ₄₀ peptide is bound at the centre of the resulting ligand pocket, with 11 residues in the middle of its sequence fully defined, but fading electron density for the two flanking segments, thus indicating elevated flexibility and conformational disorder on both sides (Fig. 1). Based on this structural information – and inspired by a previously published approach for engineered antibodies as fluorescent immunosensors for other antigens (Renard et al., 2002) – we set out to identify exposed positions within the four protruding loops of the Anticalin to introduce a fluorescent group in a manner that it can engage in close contacts with the bound Aβ peptide ligand without sterically hindering its complex formation.

Figure 1.

Anticalin structure with mutation sites. (A) Positions considered for the introduction of a free Cys residue in the crystal structure of the H1GA·Aβ complex (PDB ID: 4MVL) (Eichinger et al., 2022). Cβ atoms of residues V40, R43, D45, D72, D73, A127, T132 are highlighted with green spheres. The Anticalin is shown as ribbon in grey with the four structurally variable loops in cyan. The bound Aβ peptide is depicted as a stick model with its structurally defined portion in atom colouring (C: light pink; O: red; N: blue) and the modelled flexible N- and C-terminal segments in pale blue. (B) Structural model of the H1GA·Aβ complex (rotated vertically by 90°) with the fluorescent NBD group (stick model; C: green; O: red; N: blue; cf.  Scheme 1) coupled to a Cys side chain introduced at position D45.

To this end, we chose IANBD amide, a chemically activated fluorescent dye that readily reacts with the thiol side chains of Cys residues (Scheme 1), which can be easily introduced into the recombinant lipocalin by protein engineering. 7-nitrobenzo-2-oxa-1,3-diazole (NBD) constitutes a well known solvatochromic fluorophore with a brightness of 263 cm⁻¹ M⁻¹ at 478 nm, as published for the unconjugated dye (Uchiyama et al., 2003). Its emission is strongly sensitive to the local microenvironment such that higher hydrophobicity leads to an increase in fluorescence intensity, also accompanied by a shift of emission towards shorter wavelengths. While the IANBD ester has been repeatedly used in previous efforts to construct so-called "reagentless" biosensors (Brient-Litzler et al., 2010; Renard et al., 2003; Renard et al., 2002) we employed the IANBD amide because of its higher chemical stability against hydrolysis in aqueous buffers. In fact, when using the ester derivative, preliminary experiments in our laboratory indicated progressive loss in fluorescence activity during storage or handling of corresponding protein conjugates.

Scheme 1.

The NBD fluorescent dye, its Cys-reactive ester and amide acetyl derivatives, and the thioester bond formation with a thiol side chain.

Based on the recently published crystal structure of the Aβ-specific Anticalin (Eichinger et al., 2022), we first introduced a couple of rational side chain substitutions in order to improve the shape complementarity within the H1GA^.Aβ complex (Fig. 1). Among seven mutations that were tested, we identified two amino acid exchanges, V40M and F83Y, that led to significantly improved affinities, with K_D values of 0.159 nM and 0.176 nM, respectively, compared to 0.273 nM as measured for the original H1GA protein via SPR spectroscopy. For these experiments, a purified recombinant Aβ fusion protein with the maltose binding protein (MBP) of E. coli (Rauth et al., 2016) was employed, which constitutes a better soluble and stably monomeric reagent (MBP-Aβ₄₀), in contrast to the Aβ peptides themselves. The new Tyr side chain of the F83Y mutant likely forms an energetically favourable hydrogen bond to the side chain carboxylate group of residue Asp P23 within the bound Aβ peptide. Due to its somewhat better folding and purification behaviour upon expression in E. coli – and in the light of the oxidation sensitivity of exposed Met residues, as it might be expected for the V40M mutant – we chose the version H1GA(F83Y) as a starting point to introduce an unpaired Cys residue at seven different positions distributed across the four loop segments that surround the bound peptide: V40, R43, D45, D72, D73, A127, T132 (Fig. 1).

All seven mutated proteins, each equipped with the C-terminal Strep-tag II (Schmidt and Skerra, 2007), were produced in the cytoplasm of an E. coli strain that provides a less reducing intracellular milieu (see Methods) and purified to homogeneity from the whole cell extract using Strep-Tactin affinity chromatography, followed by size-exclusion chromatography (SEC). This switch in the expression strategy from the conventional periplasmic secretion (Gebauer and Skerra, 2012), as it was applied for the preceding experiments, led to higher protein yields and a defined reduced state of the free thiol group that was introduced with the additional Cys residue. The resulting recombinant proteins were biochemically characterized and quantified prior to conjugation with IANBD (Table I). This step was important as the Lcn2-based Anticalins carry a structural disulfide bridge connecting C76 and C175, which is partially exposed to the solvent, such that non-physiological cross-links may be formed with the newly introduced third Cys residue during protein folding – or in the course of subsequent cell lysis and purification.

Indeed, two of the seven mutants, R43C and D73C, led to disulfide isomers as well as pronounced dimers (from oxidative cross-linking between two protein molecules involving their free thiol groups) as detectable by SDS-PAGE without reducing agent and by SEC (not shown). Therefore, these mutants were excluded from further analysis. The other five candidates appeared monomeric both in SDS-PAGE and in SEC and showed decent yields in the range of 1.9–7.1 mg purified protein per liter bacterial culture. However, when analyzing the affinity towards MBP-Aβ₄₀, only the mutant V40C revealed a dissociation conctant (K_D) similar to the unmodified Anticalin H1GA(F83Y), with K_D = 240 pM versus 180 pM, respectively, whereas the other mutants showed elevated K_D values, up to 6.03 nM for D72C (see Table I).

Nevertheless, all five viable Cys mutants were subjected to coupling with IANBD, directly after protein purification from the bacterial whole cell extract, and the resulting conjugates were further isolated via cation exchange chromatography (CEX) in order to remove any oligomeric protein species that might have formed as side products. For each mutant, the dye conjugate was successfully prepared in this manner, as evident from the visible absorption around 500 nm as well as bright fluorescence emission around 550 nm (Figs 2 and 3, Table I). Molecular masses were confirmed by ESI-MS, and affinities towards the Aβ peptide target were initially assessed via SPR measurements using essentially the same setup as before. Interestingly, all five dye conjugates showed K_D values equal or even better than those measured for the protein preparations that were produced in the bacterial periplasm (see Methods) before conjugation with the dye (Table I). An improvement in affinity was particularly evident for the dye conjugates of the mutants D72C and T132C, with factors of approximately 4 and 6.5, respectively. This phenomenon may be explained by a partial obstruction of the binding site due to conjugate formation with glutathione, or related physiological thiol compounds, during secretion into the bacterial periplasm (Eser et al., 2009).

Figure 2.

Biochemical analysis of Anticalin-dye conjugates. (A) SDS-PAGE of NBD-conjugated Cys mutants of H1GA(F83Y) (= H1GA^*): first, a fluorescence image was recorded on an Odyssey scanner (right), then the gel was stained with Coomassie brilliant blue (left). Small variations in the electrophoretic mobility – also referred to as “gel shifting anomaly” (Shi et al., 2012) – are within the range normally observed for point mutations in proteins. (B) Exemplary analytical SEC profile of the Anticalin-dye conjugate H1GA(F83Y/D45C)-NBD isolated by CEX. (C) Exemplary deconvoluted ESI mass spectrum of H1GA(F83Y/D72C)-NBD (calculated mass 21846.6 Da). The small difference of 0.7 Da between measured and calculated mass is within the typical experimental error of ±1 Da for ESI-MS measurements with intact proteins of this size.

Figure 3.

Absorption and fluorescence emission spectra of promising Anticalin-dye conjugates. (A) Absorption spectrum of unconjugated H1GA(F83Y) (black) and of the H1GA(F83Y/D45C)-NBD conjugate (blue) in 50 mM Tris/HCl, 100 mM NaCl, 1 mM EDTA pH 7.5. (B) Fluorescence spectra of the H1GA(F83Y/D45C)-NBD (black) and H1GA(F83Y/D72C)-NBD conjugates (blue) upon excitation at 485 nm. (C) Absorption (black) versus fluorescence emission (blue; λ_Ex = 485 nm) spectra for the H1GA(F83Y/D45C)-NBD conjugate.

In the next step, the five conjugates were investigated in greater detail with regard to the modulation of their fluorescence emission upon ligand binding and, thus, their suitability as an Anticalin-based Aβ biosensor. Main criteria were (i) the spectroscopic quality, i.e. change in fluorescence intensity and λ_max during complex formation with either the MBP-Aβ₄₀ fusion protein or the synthetic Aβ₄₀ peptide, and (ii) retention of high target affinity, this time measuring the K_D value for complex formation in solution between the Anticalin-NBD conjugate and MBP-Aβ₄₀, or the Aβ₄₀ peptide, using the method of fluorescence titration (Table II). In addition, K_D values and changes in fluorescence intensity as well as λ_max were determined for the most promising mutants by applying the Aβ₄₀ peptide in the presence of 5% w/v human serum albumin (HSA) and also investigating the slightly longer synthetic Aβ₄₂ peptide.

Table II.

Fluorescence properties of H1GA(F83Y)-dye conjugates determined via fluorescence titration.

H1GAa mutant	ΔI [%]				K D  [nM]				Δλ max a
	MBP-Aβ 40	Aβ 40	Aβ 40  / HSA	Aβ 42	MBP-Aβ 40	Aβ 40	Aβ 40  / HSA	Aβ 42
V40C	± 0	± 0	–	–	–	–	–	–	- 5
D45C	+ 488	+ 506	+ 509	+ 514	4.2 ± 1.8	2.6 ± 1.3	1.2 ± 0.8	6.8 ± 1.9	± 0
D72C	+ 491	+ 392	+ 413	–	4.3 ± 1.7	1.6 ± 0.8	5.8 ± 1.7	–	- 14
A127C	± 0	± 0	–	–	–	–	–	–	± 0
T132C	+ 264	+ 255	–	–	0.1 ± 0.2	0.7 ± 0.4	–	–	- 7

^aShifts in the maximum intensity wavelength were identical (within ±1 nm) between MBP-Aβ₄₀ and the isolated peptides.

As result, three mutants emerged whose dye conjugates revealed impressive fluorescence response to Aβ peptide binding, D45C, D72C and T132C, whereas two mutants, V40C and A127C, showed no or marginal change in their spectroscopic properties. In particular, the mutants D45C and D72C both exhibited strong increase in fluorescence intensity, 6fold and 5fold, respectively, upon complex formation, regardless of the presence of HSA at high concentration (even exceeding the one in human plasma). While for D72C the gain in fluorescence was noticeably higher (6fold versus 5fold) when applying the MBP-Aβ₄₀ fusion protein instead of the isolated peptide, there was no corresponding difference for D45C (Figs 3 and 4). Furthermore, D72C showed a shift in λ_max by −14 nm whereas there was no relevant change in λ_max for the version D45C.

Figure 4.

Fluorescence titration of Aβ-specific Anticalin-dye conjugates with the target peptide. (A, B) Titration of various dye-labeled Cys mutants of the Anticalin H1GA(F83Y) with the MBP-Aβ₄₀ fusion protein (A) or the monomeric Aβ₄₀ peptide (B). (C, D) Titration of the D45C-NBD (C) and D72C-NBD (D) conjugates in the presence or absence of 5% w/v HSA, respectively. The concentration of the Anticalin protein was always 400 nM. The excitation wavelength was 485 nm while emission was detected at according to Table I.

These findings indicate that the NBD fluorophore is ideally positioned at amino acid position 45 within loop #1 of the Anticalin (see Fig. 1B) to allow formation of a tight hydrophobic contact with the bound Aβ peptide such that the quantum yield of its fluorescence emission is enhanced in comparison to the solvated state in the absence of the peptide ligand. As expected, the spectroscopic properties did not change when using the Aβ₄₂ peptide instead of the slightly shorter Aβ₄₀, since in the latter only the two C-terminal residues are lacking which are positioned ouside the ligand pocket of the Anticalin (see Fig. 1). Finally, the high target affinity was almost fully retained for the dye conjugate of the mutant D45C, which revealed a K_D value of 0.88 nM towards MBP-Aβ₄₀ in the SPR measurement (Table I). This affinity is in a suitable range regarding the typical Aβ₄₀ concentration in CSF of ~2000 pg/ml (Tapiola et al., 2000), which corresponds to ~0.46 nM. Most importantly, a very similar K_D value of 1.2 nM was determined when titrating the protein with the plain Aβ₄₀ peptide in the presence of 5% w/v HSA (Fig. 4, Table II). This last experiment demonstrates the potential of the NBD-labelled anti-Aβ Anticalin H1GA(F83Y/D45C) as a biosensor to sensitively detect Alzheimer amyloid peptides in biologicaly samples by transducing the binding event into a strongly increased fluorescence signal within the visible spectral region.

Discussion

A biosensor is a device which converts biological activity into a quantifiable signal, typically by combining a receptor protein that recognizes the target analyte and a transducer that translates this event into a detectable signal, for example a fluorescence signal (Reja et al., 2024). The preparation of a dye conjugate with a recombinant antibody fragment that responds to antigen binding by variation of its fluorescence was initially demonstrated by H. Bedouelle and coworkers (Renard et al., 2002). These authors introduced a Cys residue into the paratope region of the single-chain variable fragment (scFv) of the antibody D1.3, which recognizes hen egg-white lysozyme (HEL) (Skerra, 1994), and labeled it with the IANBD ester. One of 17 conjugated Cys mutants showed an increase in fluorescence by up to 90%. This signal was proportional to the concentration of the antigen lysozyme, which was applied in an optimized buffer, revealing a K_D value of 2 nM as measured by fluorescence titration. In following studies, this concept was extended to alternative binding proteins, i.e. a Designed Ankyrin Repeat Protein (DARPin) selected against the maltose-binding protein (MalE) from E. coli and a Nanofitin/Affitin, again with binding activity towards HEL (Brient-Litzler et al., 2010; Miranda et al., 2011).

Subsequent work was directed at antigens of more biomedical relevance, for example, with an scFv fragment that recognizes the envelope protein of dengue virus (Renard et al., 2003) or with a DARPin selected against the phosphorylated extracellular signal-regulated kinase (pERK) (Kummer et al., 2013). Further attempts focused at other alternative binding proteins such as derived from the Fn3 domain or the Sso7d/Sac7d scaffold (de Picciotto et al., 2016), also employing merocyanine (Kummer et al., 2013) and other solvatochromatic fluorescent dyes (de Picciotto et al., 2016).

More recently, the environmentally sensitive fluorescent unnatural amino acid 3-(6-acetylnaphthalen-2-ylamino)-2-aminopropanoic acid (Anap) was site-specifically introduced into scFv fragments directed against the epidermal growth factor receptor (EGFR) or against cardiac troponin I, leading to a dose-dependent blue shift in emission wavelength – but only minute change in fluorescence intensity – upon binding of the target protein (Islam et al., 2023; Islam et al., 2019). Furthermore, building on small protein binders (such as nanobody variants) and genetically encodable fluorogenic amino acids (FgAAs), a pipeline to rapidly discover, evolve, and produce protein-based nanosensors has been described (Kuru et al., 2024). Nevertheless, even though several design principles were proposed in order to develop protein-based biosensors that exhibit maximally enhanced fluorescence upon ligand binding, for example to study dynamic fluctuations of protein levels in living cells and tissues (de Picciotto et al., 2016; Kuru et al., 2024; Renard et al., 2003; Renard et al., 2002), the actual number of published examples for targets of true biomedical or diagnostic relevance has remained astonishingly low up to now.

In the present study, we have designed a conditionally fluorescent biosensor protein based on the engineered lipocalin scaffold which is specific for the Aβ peptide, an important biomarker of Alzheimer's disease. The molecular design started from a previously developed anti-Aβ Anticalin with exquisite specificity and high affinity towards this peptide target (Rauth et al., 2016), including the experimentally elucidated three-dimensional structure of the ligand complex (Eichinger et al., 2022). Compared with the macromolecular targets of previously published examples of ligand-sensitive fluorescent protein conjugates, as discussed above, the Aβ peptide represents a particularly small target molecule, especially if considering that only 11 of its residues are conformationally defined in the crystal structure of the complex. In so far, the high success rate, 3 of 7 Cys mutants, in the search for suitable positions to introduce the fluorophore in a ligand-sensitive manner is surprising. In fact, this indicates that the complicated design rules that were previously proposed for such binding-activated fluorescent nanosensors (de Picciotto et al., 2016; Kuru et al., 2024; Renard et al., 2003; Renard et al., 2002) may not always be relevant.

Among these, the postulate of a specific binding pocket for the fluorophore on the analyte (de Picciotto et al., 2016) obviously is not fulfilled in the case of the slim Aβ peptide target. Indeed, the exposed position of the NBD dye at the tip of loop #1 in the most active Anticalin mutant D45C (see Fig. 1B) suggests an interaction with the loose ends of the bound peptide rather than with the central part that gets buried within the binding pocket of the Anticalin upon complex formation. This also explains the minute influence of the attached fluorophore on the ligand affinity of the Anticalin, whereas the enhancement of the fluorescence emission of the protein–dye conjugate by a factor 6 as a consequence of ligand binding is remarkable. Notably, this effect remained unchanged when MBP was attached to the Aβ peptide, which implies that the local environment of the Aβ epitope remains essentially unchanged in this modified molecular format – with a bulky fusion partner compared to the free peptide.

Most interestingly, both the ligand-sensitive fluorescence behaviour and the tight ligand affinity are unaffected in the presence of 5% w/v albumin, which is comparable to the conditions in human plasma. Hence, the fluorescent Aβ biosensor protein developed here bears the potential to be applied for the inexpensive and effort-less detection of the amyloid peptide in samples of extracellular body fluids, such as blood plasma or CSF, e.g. for purposes of initial screening of patients prior to the precise quantification and more elaborate analytical dissection of individual Aβ species (Cheng et al., 2022).

Author contributions

A.F. (Investigation [lead], Methodology [lead], Writing—review & editing [equal]), A.S. (Conceptualization [lead], Writing—original draft [lead], Writing—review & editing [equal]).

Conflict of interest

A.S. is co-inventor on a patent covering the anti-Aβ Anticalin that was used in this study. Otherwise, the authors declare no conflicts of interest regarding this article. Anticalin® is a registered trademark of Pieris Pharmaceuticals GmbH, Germany.

Funding

None declared.

Data availability

The data underlying this article are available in the article and its Tables.