Introduction

Templating is an ancient technology which was already applied by the Muiscas—the original population of South America—before the arrival of the Spanish conquistadores (1575) for the preparation of golden sculptures, e.g., the El Dorado Raft. Using the templating concept, the Russian scientist Polyakov proposed the first polymer with molecular memory in 1931 [1]. The template, functional monomers, and crosslinkers are polymerized together in these processes. Removal of the template forms template-complementary binding sites resembling the “lock and key principle.” However, a real breakthrough has been only achieved ~ 50 years later [2, 3]. Initially, imprinted polymers have been developed exclusively for low-molecular weight substances and these synthetic sorbents coined the name of molecular imprinted polymers (MIPs). Imprinting of large biological molecules and particles (e.g., viruses and cells) faces challenges due to their complex nature, fragility, solvent compatibility, high cost for pure templates, distribution of heterogeneous binding sites, and difficulties in highly reproducible mass production [4, 5].

Antibodies—the biological counterparts of MIPs—bind only to a small area of large antigens—the immunogenic determinant, so-called epitope. Mimicking this principle Rachkov and Minoura [6] applied the three amino acids on the N-terminal of the nonapeptide oxytocin as the template for a MIP which could subsequently bind the whole peptide. Later, this approach was successfully extended to proteins and cells and termed as epitope imprinting to highlight the similarity between epitope-MIPs and antibodies [6]. Such epitope-MIPs can recognize the template fragment as well as the respective whole protein [6,7,8,9,10,11,12]. A generally claimed advantage of the epitope imprinting concept is the lower cost of the template in the MIP synthesis resulting a 10- to 50-fold lower cost per sensing spot (refer to SI1 in the supporting information).

On the other hand, the differences in terms of target binding characteristics of both types of MIPs are still not well understood. Only a few reports compare the analytical performance of MIPs, which use an epitope or the whole protein as the template. Our group developed MIPs either using cytochrome c adsorbed on a mercaptoundecanoic acetic acid (MUA) layer or the cysteine extended C-terminal nonapeptide as the template [9, 13]. Sellergren’s and Piletsky’s groups developed MIPs for the recognition of immunoglobulin G (IgG) using the whole antibody molecule, the Fc domain and the C-terminal 10-mer peptide, respectively [10]. All three MIPs recognized both human and goat IgGs.

In this paper, we compare the analytical performance in terms of measuring range, KD, and interference by HSA of MIPs generated by using a peptide epitope or the whole protein as the template for the recognition of alpha-fetoprotein (AFP) and the receptor binding domain (RBD) of SARS-CoV-2. For mechanistic insights, the density of binding pockets for whole-protein MIPs and epitope-MIPs has been analyzed by conductive AFM (C-AFM).

Both targets are of high diagnostic relevance: AFP is the most frequently determined biomarker for the early detection of the neoplastic disease hepatocellular carcinoma (HCC) which is the 7th most common cancer [14, 15]. The 29.5-kDa receptor binding domain is a globular structure protein which anchors the SARS-CoV-2 to the angiotensin-converting enzyme (ACE2) of the host cell and it is frequently used as the target structure in SARS-CoV-2 immunoassays and lateral flow assays [16]. Mimicking the biological counterpart, we have chosen “the non-terminal epitope” GFNCYFP derived from the exposed binding site of the RBD to the ACE. The heptapeptide contains a central cysteine (C488) which can bind the peptide to the gold surface [17]. In case of the MIPs targeting AFP beside the whole AFP glycoprotein (MW 69 kDa), the two peptides AALGVC and SKTRAALGVC of the exposed C-terminus containing an additional C-terminal cysteine for immobilization were used as template.

Materials and methods

Chemicals and reagents

Sulfuric acid (H2SO4) (96%), absolute ethanol (99.9%), sodium chloride (NaCl), potassium chloride (KCl), disodium hydrogen phosphate dihydrate (Na2HPO4•2H2O), potassium dihydrogen phosphate (KH2PO4), potassium hydroxide (KOH), potassium hexacyanoferrate (II) trihydrate (K4[Fe(CN)6]•3H2O), and potassium hexacyanoferrate (III) (K3[Fe(CN)6]) were from ROTH (Karlsruhe, Germany). One molar hydrochloric acid (HCl), phosphate-buffered saline (PBS), scopoletin (7-hydroxy-6-methoxycoumarin), the receptor binding domain from SARS-CoV-2 (RBD, 25.9 kDa), and albumin from human serum (HSA, 66.5 kDa) were purchased from Sigma-Aldrich (Merck, Darmstadt, Germany). Human alpha-fetoprotein (AFP, 63 kDa) from Bio-Rad (Munich, Germany). HAuCl4•3H2O was obtained from Chinoin Zrt. (Budapest, Hungary). The solutions throughout this work were prepared using deionized and filtered water obtained from a water purification system Milli-Q from Sartorius (Göttingen, Germany).

All peptides listed in Table 1 were purchased from Biosyntan (Berlin, Germany) and possess an amide group at the C-terminus.

Table 1 Overview of the peptides utilized in this work
Full size table

Apparatus and electrochemical experiments

All the electrochemical measurements were conducted at room temperature using a CHI 440 electrochemical workstation (CH Instruments, Austin, TX, USA) and a three-electrode system. Gold wires (diameter: 0.5 mm; active surface: 20 mm2) (Goodfellow, Germany), a platinum coil, and an Ag/AgCl system were utilized as the working, counter, and reference electrodes, respectively.

SWV measurements and electrochemical oxidation for template removal were carried out using a one-compartment polymethylmethacrylate (PMMA) cell with a volume of 1 mL with Ag/AgCl (1 M KCl) as the reference electrode. SWV experiments were carried out in 5 mM ferri/ferrocyanide in phosphate-buffered saline (PBS) (137 mM NaCl, 10 mM Na2HPO4, 1.8 mM KH2PO4, and 2.7 mM KCl, pH 7.4) by scanning the potentials from − 0.2 to + 0.5 V, with a frequency of 10 Hz, amplitude of 50 mV, and a step height of 3 mV.

Electropolymerization of scopoletin was conducted in a one-compartment polychlorotrifluoroethylene (PCTFE) cell with a volume of 2 mL with Ag/AgCl (3 M KCl) (CH Instruments) as the reference electrode.

Cleaning of working electrodes

Gold wires were used as working electrodes for the preparation of MIPs. They were initially treated in 2.5 M KOH for 4 h and then stored overnight in 96% H2SO4. The purity of the electrodes was verified by cyclic voltammetry in 0.5 M H2SO4. Finally, anodic oxidation was performed in PBS at 1.1 V for 30 s.

Electrosynthesis of MIP, template removal, and binding assays

Table 2 shows the preparation details of the different MIPs used in this study with either the whole protein or peptide templates. Generally (Scheme 1), to create a MIP, a gold wire electrode was initially incubated in template in PBS at room temperature. To follow the adsorption of the respective template, the SWV signal of ferri/ferrocyanide was measured. The adsorption step was finished when the suppression of the SWV signal reached around 50%. Then, the polymer was synthesized around the adsorbed template by electropolymerizing 0.5 mM scopoletin in 10 mM NaCl containing 5% ethanol using between 20 and 40 pulse cycles, each starting with 0 V for 5 s and followed by 0.7 V for 1 s. The synthesis of non-imprinted polymers (NIPs) was conducted following the same procedure, but without template. After polymerization, the electrodes were rinsed with ddH2O and dried using a gentle nitrogen stream. The adsorbed template was then removed from the polymer by anodic oxidation in PBS (applied potential and length of the pulses are summarized in Table 2). Subsequently, the electrodes were washed with deionized water and dried carefully with nitrogen. The electrodes were then immersed in 1 mL of stirred ferri/ferrocyanide solution. For the binding assays, 2 µL of the analyte stock solution at a specific concentration was introduced into the cell. Both template removal and binding were characterized by measuring the SWV signal of ferri/ferrocyanide every 2 min without stirring until steady-state was reached.

Table 2 Synthesis details of the MIPs prepared for the comparative study
Full size table
Scheme 1
scheme 1

Workflow of MIP synthesis and application shown for the preparation of AFP-binding MIP by using peptide epitope imprinting: (1) the template peptide is immobilized on the gold surface via terminal cysteine (blue dots), (2) the scopoletin monomer is electropolymerized, (3) the template is removed by anodic stripping, (4) binding of the whole AFP protein or the peptide by the liberated binding sites

Full size image

Conductive atomic force microscopy measurements

To ensure planar MIP nanofilm specimens for C-AFM measurements gold slides were used instead of gold wires as working electrodes, which required slight changes in the surface pretreatment and MIP synthesis. Prior to use, the gold slides were rinsed successively with acetone and ethanol, dried under N2 stream, and then subjected to 10 min UV/ozone treatment (PSD Pro Series Digital UV Ozone System, Novascan, USA). For MIP preparation, a conical electrochemical cell made of methacrylate was used (CFLWCL-CONIC, Metrohm DropSens) connected to an Autolab PGSTAT12 Potentiostat/Galvanostat. A 3-electrode system was used with the gold slide, a platinum wire, and an Ag/AgCl (3 M KCl) (CH Instruments) as working, counter, and reference electrodes, respectively. For template immobilization, the conical cell was filled with 5 µM GFNCYFP peptide or 450 nM RBD in PBS and it was incubated at room temperature for 15 min by shaking at 350 rpm. After rinsing with deionized water, the electropolymerization was performed by cyclic voltammetry in a solution containing 0.5 mM scopoletin, 0.1 M NaCl, and 4 mM K4[Fe(CN)6]. Fifty cycles were made between 0.0 and 0.75 V at a scan rate of 50 mV s−1. Template removal was performed in PBS by chronoamperometry applying subsequently 0.9 V for 30 s, then 0 V for 5 s and finally again 0.9 V for 30 s. Gold deposition was performed in 0.2 mM HAuCl4 solution containing 0.1 M HCl and 2 mM sodium citrate by chronoamperometry using − 0.9 V for 30 s. C-AFM measurements were performed with a FlexAFM instrument (Nanosurf, Switzerland) at a setpoint of 20 nN and tip voltage of − 10 mV using an electrically conductive tip, PPP-ContPt (Nanosensors, Switzerland) with a tip radius of ca. 25 nm and a force constant of 0.2 N/m.

Results and discussion

MIPs for AFP using peptide epitope imprinting

Based on the prediction from modeling showing that already tripeptides could serve as templates for MIPs with µM measuring range [18], we applied the hexapeptide AALGVC (AC-MIP) as the template. However, the binding experiments for the AC-MIP revealed only negligible affinity towards the target AALGV (Figure S1). Already, the interaction of the AC-MIP with the unrelated peptide YFPLQS (Y-peptide from SARS-CoV-2 Spike protein) resulted in a considerably larger (non-specific) binding as revealed by the larger current suppression (Figure S1). Obviously, the template peptide (AALGVC) was too short for efficient rebinding of the imprinted peptide. To increase the binding strength of the MIP cavities, we applied a template which was extended at the N-terminus of the AC-peptide with four amino acids, resulting in SKTRAALGVC (SK-MIP). The elongation of the imprinted peptide template resulted in increased binding efficiency. The concentration dependence for the binding of the derived septa-peptide TRAALGV (TR-peptide) to the SK-MIP (Fig. 1) gave an estimated KD value of 14 µM. Imprinting with the longer SK-peptide also improved the discrimination of non-related C-terminal peptides, i.e., the pentapeptide VVSTQ from BSA (BSA-peptide), as shown by the significantly smaller relative current suppression as compared with the specific TR-peptide (Fig. 1b).

Fig. 1
figure 1

a SWVs curves characterizing the steps of MIP synthesis and the rebinding of the target TR-peptide (TRAALGV) to SK-MIP (SKTRAALGVC); b concentration dependent relative suppression of the SWV current signal for the SK-MIP upon binding the target protein AFP, the target TR-peptide (TRAALGV), and the non-related BSA-peptide (VVSTQ)

Full size image

The parent protein AFP binding to the SK-MIP resulted in an even larger current suppression than the TR-peptide. Evaluation of the concentration dependence (Figs. 1b and 2) gave a KD value of 13.2 nM (Table 3). Since the non-imprinted polymer (NIP) is electrically insulating, this detection method based on the target-binding gated redox marker current is not applicable to measure the nonspecific AFP binding to the NIP. Still, it is a compulsory control experiment that indeed revealed insignificant signal change upon interacting the NIP with AFP as represented by a “formal” imprinting factor (IF) of 6.1 in the linear measuring range (Fig. 2). The selectivity of the SK-MIP was tested using HSA a protein of very similar size to AFP. An excellent discrimination discrimination was observed, i.e. the signal level was close to the background level measured on NIPs (Fig. 2).

Fig. 2
figure 2

Concentration dependence of the binding of AFP and HSA proteins to the SK-MIP, AFP-MIP, and NIP

Full size image
Table 3 Dissociation constant KD and binding capacity Bmax for AFP and RBD to the respective epitope and protein templated MIPs
Full size table

MIPs for RBD using peptide epitope imprinting

We turned to confirm the conclusions of the AFP-MIP studies to another model system, i.e., RBD. Similar to the AFP case where, as shown in the “MIPs for AFP using peptide epitope imprinting” section (Figure S1), only a very weak binding of the pentapeptide AALGV to the AALGVC-based MIP was observed, the tetrapeptide EGFN showed no measurable binding to the G-MIP.

The MIP using the G-peptide as the template showed almost fourfold higher affinity towards the whole RBD (KD of 14.7 ± 0.9 nM) as compared with the G-peptide (58 nM) (Fig. 3b). The KD value for RBD is comparable with that determined by SPRi in a similar system, but with the epitope-imprinted nanofilms prepared on a planar chip by a combination of peptide microspotting and electrosynthesis, i.e., 2.2 ± 0.4 nM [19].

Fig. 3
figure 3

a SWVs characterizing the synthetic steps of the RBD-MIP; b binding of RBD and HSA to the RBD-MIP and G-MIP

Full size image

MIPs using the pre-adsorbed (whole) proteins as template

A sufficiently strong adsorption of the proteins to the gold substrate to withhold them during the electrosynthesis of the polymer is the prerequisite for the MIP synthesis around the adsorbed protein template. We found that both AFP and RBD bound spontaneously to the gold electrode (Figure S2), but could be effectively desorbed after a 30-s duration anodic pulse of 950 mV for AFP and 970 mV for RBD.

The MIPs prepared by using the whole protein as the template had affinities in the lower nM range for the binding of the protein templates, i.e., the KD values for AFP and RBD were 14.5 nM and 16.8 nM, respectively (Table 3). As shown in the “MIPs for AFP using peptide epitope imprinting” and “MIPs for RBD using peptide epitope imprinting” sections, the respective epitope-MIPs showed very similar affinities towards the target proteins. On the other hand, the KD values reported earlier for MIPs either using cytochrome c adsorbed on a MUA layer (10 µM) or the cysteine extended C-terminal nonapeptide as the template (8.5 µM) are almost three orders of magnitude larger [9, 13].

For both proteins, the magnitude of the relative suppression of the SWV signal at saturation was higher for the whole protein imprinted polymers than that of the epitope-MIPs (Figs. 2b and 3). In case of AFP, the difference was less significant than for RBD.

C-AFM investigation of the MIP nanofilms

To better understand and reveal contingent differences in the morphology of the MIP nanofilms prepared by using peptide or protein templates, we used AFM measurements. Molecular size cavities are difficult to be identified in polymer films [20]; therefore, we attempted to reveal them by electroplating gold in the imprinted cavities. Since during gold deposition gold nanoparticles may also form nonspecifically on the polymer film surface, we used C-AFM to discriminate between the gold grown through the imprinted cavities and nonspecific gold deposition. For this purpose, a small potential difference (− 10 mV) is applied between the gold substrate underneath the MIP and NIP nanofilm and the conducting AFM tip and as such the current map will reveal only the gold spots that are electrically connected to the gold electrode, i.e., the gold was grown through the imprinted cavities.

As shown by the current map in Fig. 4, without imprinting the electrically insulating polyscopoletin film (NIP) proved to be largely pinhole-free after the gold electrodeposition. In turn, both the peptide and the protein imprinted films revealed locations with elevated current indicative of the electrodeposited gold through the imprinted cavities. While at this stage we cannot claim revealing all imprinted cavities, the results clearly suggest a higher surface density of the imprinted cavities in case of the peptide imprinting. However, apparently this higher density of binding cavities in case of peptide epitope imprinting is not beneficial for the much larger size proteins when using the target gated redox mediator current for detection as revealed by the larger Bmax values for whole protein imprinting shown in Table 3.

Fig. 4
figure 4

Topographic images and current maps as revealed by C-AFM after electrodeposition of gold on NIP, G-peptide epitope (G-MIP), and RBD (RBD-MIP) imprinted polyscopoletin nanofilms

Full size image

Since the purpose of these MIPs is ultimately the detection of the proteins, their analytical performance towards proteins is most relevant. Such KD values in the lower nanomolar range can be considered rather competitive given that the KD values for protein-MIPs are generally in the µM to nM range [13, 21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46] including AFP and RBD (Table S1).

On the other hand, the KD values for binding the template (or truncated) peptides to the epitope-MIPs for the G-peptide (58 nM) and TR-peptide (14 µM) are larger than for the respective parent protein. This suggests an apparently lower affinity for the peptide binding. However, considering only the binding to the imprinted cavity with 1:1 stoichiometry, the affinity of an epitope-MIP towards the parent protein should not exceed the affinity for the peptide. This relation was found earlier for the epitope-MIP templated with a cysteine extended C-terminal nonapeptide of Cyt c and direct readout by fluorescence, i.e., a lower affinity was found for the parent protein (KD of 8.54 µM) as compared with the target peptide (KD of 2.5 µM). On the other hand, when the “non-epitope” surface of the protein target nonspecifically interacts with the non-imprinted part of the MIP surface, an apparently stronger binding of the protein can be found. The MIP using the G-peptide as the template showed almost fourfold higher affinity towards the whole RBD as compared with the G-peptide and even higher towards the spike protein.

The higher (apparent) binding affinity of the MIPs determined by the indirect method of redox marker gating for both protein targets as compared with the respective template peptides  might be caused by the lager “footprint”of the proteins. Thus, upon binding to the cavities of the epitope-MIP, the larger surface covered by the protein will disturb the non-linear diffusion of the redox marker to the pores (microelectrode behavior) [21] and can block multiple adjacent binding pockets while the peptide occupies only one pocket. This effect could explain the decrease of the KD by one to two orders of magnitude for the binding of the parent protein as compared with that of the peptide.

However, the benefit of using peptide imprinting is remarkable in terms of selectivity, i.e., both epitope-imprinted polymers (G-MIP and SK-MIP) show an excellent discrimination of the unrelated HSA (Figs. 2 and 3). It seems that the presumably smaller binding cavities generated by peptide imprinting allow the binding of the target protein via the exposed epitope but restrain the binding of unrelated proteins. In spite of the smaller size of RBD (MW 25.9 kD), the RBD-MIP binds the larger HSA. We earlier found this behavior for a MIP using the pre-adsorbed globular protein transferrin as the template [22] and concluded that this effect might be caused by the larger cavities in the protein-MIPs formed by partially unfolded template molecules at the gold surface [47]. Obviously, these lager cavities can accommodate protein molecules of comparable and even larger size than the template, i.e., facilitating the nonspecific binding of HSA.

Conclusion

The affinities as expressed by KD of the peptide and whole protein imprinted polymers are very similar for both model proteins, RBD and AFP. For a cautious extrapolation of this conclusion, it should be noted that we used at least 7 amino acid long peptides and the extremely thin polymer layers (ca. 10 nm) provided by the self-regulated electrosynthesis of insulating polymer films. For both AFP- and RBD-MIPs, the epitope-imprinted films had an apparently lower affinity for the peptide than for the parent protein that is most likely an artifact of the detection method due to the larger footprint of the protein on the MIP surface. The main advantage of the epitope imprinting in terms of analytical performance is the significantly better discrimination of nonrelated proteins as exemplified for HSA as compared to whole protein imprinted polymer films. Furthermore, for mass production the lower reagent cost is a clear advantage, to which adds the high purity of peptides stemming from a controlled synthetic process undergoing purification and quality control by HPLC and mass spectrometry.