Introduction

The 2021 Nobel Prize in Physiology was awarded for discovering temperature and touch receptors, in which ion channels play vital roles in the conduction of electrical signals1,2,3. Controlling and regulating electrical signals are essential for normal physiological functions, including nerve activity, heartbeat, exercise process, and immune response. The initiation, conduction, and termination voltages of action potentials (AP) in cells originate from the active cooperation of various ion channels and transporters. Minor abnormalities in voltage-gated ion channels (VGICs) can lead to severe diseases, including arrhythmia, epilepsy, and paralysis. Many references have pointed out that these ion channel proteins could be modified with various N- and O-linked glycans with sialic acid as the terminal, which accounts for approximately 30% of the mass consisting of VGIC4,5. SG on the VGIC senses external stimuli and transmits information through electrostatic interactions to realize the regulation of the VGIC. Typical examples include NaV1.4 sodium ion channels6, KV1.1 potassium ion channels7, CaV1.2 calcium ion channels8, and renal ion channels9, which can all be regulated by SGs.

The most famous example is the infection of influenza A virus8. As illustrated in Scheme 1A, the influenza A virus first binds to α2-6-linked SG on the CaV1.2 ion channel, which opens VGIC and allows calcium ions to enter the cell. The increase in the intracellular calcium ion concentration accelerates the endocytosis and absorption of the virus by the cell, which highlights the significance of SG in viral infection. From the perspective of glycochemistry, sialic acid usually binds adjacent monosaccharides via α-2,3- or α-2,6-linkages10,11. Different sialylated glycan linkage isomers can lead to completely different physiological effects. However, the structures of these linkage isomers are slightly different and difficult to distinguish. The main SG of the human upper respiratory tract is the α-2,6-linked type, and the hemagglutinin of influenza A H1N1 virus binds to this kind of SG. In contrast, the avian influenza subtype H5N1 binds to α-2,3-linked SG12. Therefore, H5N1 avian influenza has a poor ability to infect humans. In addition, the abnormal sialic acid expression on the cell surface is closely related to the occurrence of various cancers13,14. For example, prostate cancer is characterized by increased expression of α-2,3-linked SG, while lung cancer patients and breast cancer patients show overexpression of α-2,6-linked SG15,16. Recent research indicated that SG may also be related to the progression of Alzheimer’s disease17. Therefore, accurately identifying various SGs and distinguishing and determining the connection types of the SG chains are of great significance for the early diagnosis and targeted treatment of diseases.

Scheme 1
scheme 1

A Schematic diagram of the infection process of influenza A virus entering the cell by first binding to α2-6-linked SG on voltage-dependent Ca2+ channel CaV1.2. B Chemical structures of α2-3/α2-6-linked sialyllactose and lactose. C Molecular design of a functional polymer capable of recognizing sialyllactose; maltose is grafted onto the polyethyleneimine main chain. D α2-6 sialyllactose-triggered globule-to-coil transition of the polymer chains immobilized on the inner surface of the conical nanochannel, which impacts ionic transport passing through the nanochannel.

Full size image

On the other hand, inspired by the great significance of biological ion channels, various biomimetic ion nanochannels based on the electric field, ion potential, temperature, pH, and other stimulus-response modes have been developed in recent years, displaying excellent gating performance18,19,20,21,22. For example, Xie et al. modified nanochannels with azobenzene derivative-based polymers, and ionic transmission through the nanochannels could be adjusted by light and electric fields23. Sun et al. constructed a NO-regulated nanochannel based on a spiroring opening-closing reaction strategy24. Compared with artificial ion nanochannels regulated by various physical or small molecule stimuli, only a few works report biomimetic ion channels modulated by biomolecules, which are ubiquitous and play crucial roles in life processes. SG-modulated VGIC is a typical example; however, the development of glycan-specific affinity material and integration of the material with nanochannels are quite challenging.

Here, we report a biomimetic ion nanochannel regulated by SG based on a carbohydrate–carbohydrate interaction strategy. Different from conventional saccharide recognition design, such as lectin affinity, phenylboronic acid, oligopeptides, macrocyclic framework, and others25,26,27, in this study, carbohydrate–carbohydrate interaction is introduced to recognize the SG. Maltose consists of two glucose units and has plentiful hydroxyls capable of forming multiple hydrogen-bonding interactions with the target glycan. Furthermore, maltose is grafted to a polyethyleneimine (PEI) chain through a five-step reaction, generating a glycan-affinity polymer (Mal-PEI, Scheme 2). Owing to the excellent ionic rectification effect, chemically etched polyethylene terephthalate (PET) film featuring conical nanochannels works as a substrate. Integration of Mal-PEI and PET film constructs an SG-sensitive nanochannel device. Selective binding between maltose and α2-6 sialyllactose initiates globule-to-coil transition of the polymer chain, which further determines the open and closed state of the nanochannel (Scheme 1D). Taking advantage of the sensitive changes in ionic current, sialyllactose linkage isomers, SG and neutral saccharides can be discriminated precisely. In addition, the nanochannel device can monitor the sialylation process of lactose catalyzed by sialyltransferases, showing the potential for real-time determination of enzyme activity.

Scheme 2
scheme 2

Reaction route of acetylated Mal-PEI (4a) and Mal-PEI (4b).

Full size image

Results and discussion

Characterization of PET conical nanochannels modified by a functional polymer

Because of the excellent rectification effect, PET conical nanochannel membranes were chosen as substrates28 and were prepared using an asymmetric ion track-etching technique based on a customized electrochemical device (Fig. S6 in Supporting Information (SI))29. Scanning electron microscopy (SEM) observations indicated that the diameter of the base side of the nanopore was approximately 500 nm (Fig. 1C), while the diameter of the tip side of the nanopore was approximately 30 nm (Fig. 1D). Hydrogen nuclear magnetic resonance (1H NMR, Fig. 1A, Fig. S4) and infrared (IR, Fig. S5) were used to verify the formation of the Mal-PEI polymer. According to the calculation of the peak area ratio in 1H NMR, the grafting ratio of maltose on PEI was approximately 22%. Then, the Mal-PEI polymer was anchored to the inner wall of the PET conical nanochannels through a one-step coupling reaction (Fig. 1E). X-ray photoelectron spectroscopy (XPS) of the PET film before and after polymer modification indicated that the nitrogen (N) element content in the modified film significantly increased (Fig. 1F). Since the PET film itself did not contain N, the increase should be derived from the Mal-PEI polymer. In addition, the C1s (Fig. 1G), N1s (Fig. 1H), and O1s (Fig. 1I) core-level spectra of the polymer-modified PET film display the formation of C–N bonds and C=N bonds and an increase in the proportion of C–O bonds. These results confirmed the successful modification of the polymer on the PET film. Due to the hydrophilicity of maltose appended in the polymer, the static water contact angle (CA) of the PET film decreased from 75.7° to 65.2° after polymer modification (Fig. 1J, K). Then, the modified PET film was installed in a homemade electrochemical device, and a current–voltage test was performed. Figure 2A shows the current–voltage curves of the nanochannels before and after polymer modification. Because of the presence of abundant –COOH groups, the inner surface of the nanochannels was negatively charged. After polymer modification, a large number of amines in PEI were introduced, which endowed the inner surface with a positive charge. Therefore, a distinct upward current–voltage curve was observed, which further confirmed the successful grafting of the Mal-PEI polymer.

Fig. 1: Characterization of Mal-PEI-modified PET film with conical nanochannels.
figure 1

A 1H NMR spectrum of Mal-PEI in D2O at 25 °C. B Large-scale SEM image of an etched PET membrane featuring a large number of nanopores. C, D Amplified SEM images showing the morphology of the base side (C) and tip side D ends of the nanopore. E Grafting method of Mal-PEI onto the conical nanochannel. F XPS spectra of the etched PET film before (black) and after (red) the Mal-PEI modification. G–I C1s G, N1s H, and O1s I core-level spectra of the polymer-modified PET film. J, K Water droplet profiles displaying the surface water contact angle (CA) of the PET film before (J) and after K polymer modification.

Full size image
Fig. 2: Transmembrane ionic current tests.
figure 2

A Current–voltage curves of the etched PET film before (black) and after (red) Mal-PEI modification. B, C Current–voltage curves of Mal-PEI-modified PET film before and after additions of α2-6 (B) or α2-3 C sialyllactose solutions with different concentrations. D Comparison of the current reduction ratio of the films treated with α2-6 (black), α2-3 (red) sialyllactose and Neu5Ac (green) solutions (0.1 μM). E, F Concentration-dependent transmembrane ionic current decrease ratio (ΔI/I0) of the films upon treatment with α2-6, α2-3 sialyllactose, and Neu5Ac solutions (E) or α2-6 sialyllactose, galactose and lactose solutions F. G–I Concentration-dependent transmembrane ionic current decrease ratio (ΔI/I0) of bare PET film G, PEI-modified PET film (H), or acetylated Mal-PEI-modified PET film I treated with α2-6 (black) or α2-3 (red) sialyllactose solutions. All tests were conducted at 20 °C using 0.01 M NaCl solution as the electrolyte and repeated three times to obtain the average change ratios of the ionic current.

Full size image

Transmembrane ionic current measurements

Then, the Mal-PEI-modified PET film was installed in the customized electrochemical device, and the transmembrane ionic current was recorded 10 min after the addition of a NaCl electrolyte containing different concentrations of saccharides. First, sialylated trisaccharides were chosen to perform the tests, namely, Neu5Ac-α2-3Galβ-1-4Glc (abbreviated to α2-3 sialyllactose) and Neu5Ac-α2-6Galβ-1-4Glc (abbreviated to α2-6 sialyllactose) sodium salts. A pair of model sialyllactose linkage isomers with similar composition to the SGs were found in the influenza A virus receptor. After adding α2-6 sialyllactose to the electrolyte, the current–voltage curves showed a remarkable change. With the increase in the concentrations of α2-6 sialyllactose, the ionic current flowing through the nanochannels at +2 V decreased (Fig. 2B) from the initial 1.10 to 0.57 μA. With the increase in the concentration of α2-6 sialyllactose from 10–11 to 10–7 M, a linear relationship between the decrease in the current and the concentration of α2-6 sialyllactose could be built. The limit of detection (LOD) of the nanochannels for α2-6 sialyllactose was 0.59 pM, calculated through the formula LOD = 3*(SD/m)30, where SD is the standard deviation of the blank signal and m is the slope of the calibration curve. The corresponding ionic current reduction ratio (defined as [I–I0]/I0) was 48% (the concentration of α2-6 sialyllactose was 0.1 μM, which was similar to that below), indicating that the functionalized nanochannels had a good response to α2-6 sialyllactose.

Under the same conditions, the response of the nanochannels to α2-3 sialyllactose was weaker. The final ionic current reduction ratio was 27% when 0.1 μM α2-3 sialyllactose was added (Fig. 2C). To determine a possible binding site in sialyllactose, Neu5Ac was evaluated as the terminal monosaccharide in the glycan. However, the final decrease rate in the ionic current was only 15% (Fig. 2D). Saccharide concentration-dependent ionic current variation further confirmed this difference (Fig. 2E), showing that the Mal-PEI-modified nanochannels had more remarkable responsiveness to α2-6 and α2-3 sialyllactose than Neu5Ac. This indicated that the Mal-PEI polymer interacted with all SGs rather than the individual Neu5Ac unit.

Furthermore, lactose and galactose (the components of sialyllactose) were used to evaluate the responsiveness of the device. Twelve percent and 10% ionic current changes were detected for lactose and galactose (Fig. 2F), respectively, which confirmed the participation of these saccharide units in the complexation. Then, a series of control experiments were carried out. First, the bare PET conical nanochannel membranes were tested (Fig. 2G), and no evidential change in the current curve was detected for this pair of glycans. Then, the PET nanochannel membrane modified by the individual PEI was evaluated; similarly, neither α2-6 nor α2-3 sialyllactose induced an ionic current change (Fig. 2H), indicating that maltose in Mal-PEI was an indispensable binding molecule for the recognition of sialyllactose. There are abundant hydroxyls in maltose; thus, we assumed that multiple hydrogen bonding interactions between maltose and sialyllactose played a key role in the selective responsiveness of the device. To prove this assumption, acetylated protected maltose was grafted onto PEI (4a in Scheme 2), and the graft polymer was immobilized on the inner surface of the PET nanochannels. As shown in Fig. 2I, the acetylated Mal-PEI-modified nanochannels had no response to α2-6 or α2-3 sialyllactose, and this effect can be reasonably attributed to the largely weakened hydrogen bonding interactions induced by acetyl protection, highlighting the crucial role of carbohydrate–carbohydrate interactions. Two test solutions were prepared with maltose and α2-6 or α2-3 sialyllactose at a 2:1 molar ratio and added to the nanochannel device for competition assays (Fig. S7). The results showed that the nanochannel device showed no evidential change in response to the two mixed solutions. We presumed that the binding of free maltose with sialyllactoses blocked the recognition of sialyllactoses by the maltose on PEI immobilized on the inner surface of the nanochannel. The carbohydrate–carbohydrate interactions between maltose on the Mal-PEI polymer and sialyllactoses were confirmed.

Laser scanning confocal microscopy (LSCM) observation

To validate the binding events in the nanochannels, LSCM was used to observe the adsorption of α2-3 and α2-6 sialyllactoses onto the Mal-PEI-modified nanochannels. For the convenience of fluorescent tracing, both α2-3 and α2-6 sialyllactoses were labeled with fluorescein by an amination reaction at the reducing end of fluoresceinamine (Fig. 3A, B) in the presence of sodium cyanoborohydride. The crude products were purified by high-performance liquid chromatography (HPLC) on a C18 semipreparative column (Fig. 3C). The final product was confirmed by high-resolution mass spectrometry (Fig. 3D, HRMS). The Mal-PEI-modified PET film was immersed in an aqueous solution of fluorescein-labeled α2-3 or α2-6 sialyllactoses (10–7 M) for 10 min, and then the film was washed with water twice and dried under N2 flows. Three-dimensional (3D) reconstructed LSCM images by layer-by-layer scanning recorded the morphology of the PET films. As shown in Fig. 3E, a large number of bright green cones were observed on the PET film upon treatment with the fluorescein-labeled α2-6 sialyllactose solution. After zooming in on the overall picture and observing the fluorescence picture of a single cone (right panel of Fig. 3E), the shape and pore size of a single cone were approximately consistent with those observed by SEM (Fig. S8), which can be attributed to the conical nanochannel. Both the Mal-PEI polymer and PET film are nonfluorescent, and the green fluorescent signals should originate from the adsorption of fluorescein-labeled α2-6 sialyllactose on the nanochannels. By comparison, when the film was treated with a fluorescein-labeled α2-3 sialyllactose solution, the observed green cones were not clear, and their fluorescent intensities were weaker, accompanied by evident background interference (Fig. 3F). From the perspective of 3D LCSM images, we presumed that α2-6 sialyllactose had a stronger adsorption capability on the nanochannel than α2-3 sialyllactose.

Fig. 3: Morphological observation of conical nanochannels by LSCM.
figure 3

A, B Chemical structures of fluorescein-labeled α2-6 (A) and α2-3 sialyllactose (B). C, D HPLC spectrum (C) for characterization of the purity of fluorescein-labeled α2-6 sialyllactose and its HRMS (D). E, F LSCM 3D images of the Mal-PEI-modified PET membrane treated with fluorescein-labeled α2-6 (E) or α2-3 (F) sialyllactose. The right panel shows the amplified morphology of a single conical nanochannel. Excitation wavelength: 465 nm.

Full size image

Adsorption dynamics on the polymer film

Mal-PEI was grafted onto the QCM resonator sensor surface to study the dynamic adsorption behavior of α2-3 or α2-6 sialyllactose on the polymer film by a quartz crystal microbalance with dissipation monitoring (QCM-D). The overall frequency shifts (∆F) are dependent on the absorption quality of the analyte on the QCM sensor surface; the greater the mass of the analyte absorbed, the greater the degree of binding of the polymer toward the analyte31. As shown in Fig. 4A, upon injection of an α2-6 sialyllactose solution passing through the sensor, the ∆F value decreased gradually and reached equilibrium after 11 min, and the final ∆F value was approximately 34 Hz. By comparison, the α2-3 sialyllactose-induced ∆F variation was only 10 Hz, which was substantially smaller than that induced by α2-6 sialyllactose. The slow adsorption dynamics (~10 min) of α2-6 or α2-3 sialyllactose on the polymeric film also revealed that chemical adsorption rather than physical adsorption dominated the process; correspondingly, hydrogen bonding interactions between the Mal-PEI polymer and sialyllactose played a crucial role in the complexation and were superior to electrostatic adsorption. To confirm our speculation, pure PEI-modified sensors were also used to study the dynamic adsorption behavior by QCM-D. The adsorption process of pure PEI-modified sensors to two kinds of sialyllactoses was fast, and the amount of adsorption was small (Fig. S9). In this respect, the adsorption of sialyllactoses on sensors was physical adsorption caused by electrostatic interactions between sialyllactoses and PEI polymer. In addition, the ∆F value difference caused by sialyllactoses was too small to be distinguished. These results further verified the critical role of carbohydrate–carbohydrate interactions in the response of the Mal-PEI polymer to sialyllactoses.

Fig. 4: Adsorption dynamics, morphological observation, and EIS tests.
figure 4

A, B Dynamic frequency (A ∆F) and energy dissipation (B ∆D) curves of Mal-PEI-modified QCM sensors upon injections of α2-6 (black) or α2-3 (red) sialyllactose solution. The inset in (B) shows different expansions of the polymer in response to the adsorption of α2-6 and α2-3 sialyllactose. C, D Adhesion images of Mal-PEI-modified QCM sensors before (C) and after (D) treatment with α2-6 sialyllactose solution (1 mM). E The corresponding adhesion profiles along the green lines before (top) and after (bottom) treatment with α2-6 sialyllactose solution. F–H EIS measurements of the Mal-PEI-modified gold electrode in 0.1 M KCl solution containing 5 mM Fe(CN)63−/4− upon additions of α2-6 (F), α2-3 (G) sialyllactose and galactose (H) solution with different concentrations for 10 min at 20 °C. The inset in H represents the equivalent circuit for fitting the impedance spectrum to offer the electron transfer resistance (Ret). CPE: constant phase element; Rs: solution resistance; Zw: Warburg impedance. I Concentration-dependent Ret increase of the Mal-PEI-modified gold electrode treated with α2-6 (black), α2-3 (red) sialyllactose and galactose (green) electrolyte solutions.

Full size image

Moreover, QCM-D simultaneously provided an energy dissipation shift (∆D), corresponding to the variation in conformation, thickness and viscoelasticity of the polymer. An upward curve often represents the swelling of the polymer film, whereas a downward curve represents the shrinkage behavior32. As shown in Fig. 4B, both α2-6 and α2-3 sialyllactoses displayed upward dissipation curves, revealing the remarkable swelling of the polymer film. ∆D induced by the adsorption of α2-6 sialyllactose (3.7 × 10−6) was substantially larger than that induced by α2-3 sialyllactose (5 × 10−7). QCM-D data clearly indicated that α2-6 sialyllactose had stronger adsorption on the Mal-PEI thin film than α2-3 sialyllactose, accompanied by more remarkable polymer swelling. Based on this knowledge, we presumed that the expansion of the polymer chain in response to sialyllactose adsorption might obstruct the conical nanochannels and that the smaller pore size of nanochannels would lead to a decrease in the transmembrane ionic current33,34.

Atomic force microscopy (AFM)35 was used to observe morphological changes in the Mal-PEI-modified QCM sensors before and after immersion in α2-3 or α2-6 sialyllactose solution (1 mM) for 10 min, respectively. As shown in Fig. 4C, D, remarkable variation in surface adhesion was detected when the sensor surface was treated with the α2-6 sialyllactose solution, and the average surface adhesion value increased from 12 nN to 24 nN (Fig. 4E). Moreover, the surface also became rougher, as detected from the height images (Fig. S10A, B). This result was consistent with the results recorded by the dissipation curves, both of which indicated that the polymer film became softer, corresponding to the expansion of the polymer chains. By comparison, no evidential change in adhesion and roughness was detected when the sensor surface was treated with α2-3 sialyllactose solution (Fig. S10C, D).

Conformational transition disclosed by electrochemical impedance spectroscopic (EIS) tests

The expansion of the polymer might influence the electrochemical process on the surface of the polymer, and this effect was investigated by EIS36. A gold electrode was modified with Mal-PEI through the same method as in QCM-D. The Mal-PEI-modified gold electrode was immersed in the saccharide electrolyte solution (5 mM Fe(CN)63−/4−) for 10 min, and the EIS spectrum was recorded. Impedance spectra were plotted in the form of Nyquist plots (Fig. 4F–I) and fitted using an electronic equivalent circuit in order to derive the electron-transfer resistance (Ret) values by virtue of ZView software (version 2.1c, Fig. 4H inset). The semicircle diameter in the impedance spectra corresponds to Ret of the Mal-PEI layer37. After being immersed in a series of α2-6 sialyllactose solutions, the diameters of the semicircles grew gradually when the concentration of α2-6 sialyllactose increased, and the calculated Ret value increased from the initial 157 Ω to 967 Ω when 10−4 M α2-6 sialyllactose was tested (Fig. 4F). The increase in the Ret value can be explained by the fact that the Mal-PEI polymer chains changed from an initially contracted state to an expanded state after interacting with α2-6 sialyllactose, which blocked the mass transport of Fe(CN)63−/4− from the bulk solution to the surface of the electrode through the polymer film.

Compared with α2-6 sialyllactose, the increase in semicircle diameter in the impedance spectra of α2-3 sialyllactose was smaller (Fig. 4G), which indicated that the degree of expansion caused by α2-3 sialyllactose was less than that caused by α2-6 sialyllactose. Galactose is a neutral monosaccharide that exists in these two sialyllactoses, and an EIS test of galactose was also carried out. No evidential change was detected in the impedance spectra (Fig. 4H), which revealed that galactose had a weak influence on the polymer conformation. This in turn highlights the importance of the sialic acid unit. Concentration-dependent Ret increase curves further validated the remarkable difference among α2-6, α2-3 sialyllactose and galactose (Fig. 4I) when they interacted with the polymers. Furthermore, the pure PEI-modified gold electrode was also tested by EIS. Neither α2-6 nor α2-3 sialyllactose produced an evidential response in Nyquist plots (Fig. S11). Therefore, the recognition of sialyllactoses by the Mal-PEI polymer can be reasonably attributed to the carbohydrate–carbohydrate interactions between maltose and sialyllactoses.

Dynamic Light Scattering (DLS) Tests

The swelling of the polymer chains was revealed by DLS tests in solution38. The average particle size of Mal-PEI in H2O (1 mg mL–1) was measured to be 10.3 nm (Fig. 5A). After mixing with α2-3 or α2-6 sialyllactose, the average particle size of the polymer increased to 20.5 nm (Fig. 5B) or 29.1 nm (Fig. 5C), respectively. Therefore, the DLS test provided solid evidence for the expansion of the polymer chain.

Fig. 5: DLS tests and binding affinity analysis.
figure 5

A–C Hydrodynamic diameter distribution of the Mal-PEI polymer (1 mg mL–1) before (A) and after additions of 10–6 M α2-3 (B) and α2-6 (C) sialyllactose in H2O, determined by DLS. D ITC data recorded for titration of Mal-PEI solution (0.08 mg·mL–1) with additions of different equivalents of α2-6 sialyllactose. E Nonlinear fitting curve of Mal-PEI interacting with α2-6 sialyllactose using a sequential binding site model to calculate binding affinity (Ka). F Comparison of Ka values of Mal-PEI with α2-6 (black) and α2-3 (red) sialyllactose.

Full size image

Binding affinity analysis

Isothermal titration microcalorimetry (ITC) experiments were performed to measure the binding affinity (Ka) of the polymer with sialyllactoses in solution39. α2-6 or α2-3 sialyllactoses solution was dropped into the Mal-PEI solution, and the exothermic value was recorded (Fig. 5D). According to a series of exothermic amounts, Ka could be obtained by nonlinear fitting (Fig. 5E). The Ka of the polymer with α2-6 sialyllactose was 3.05 × 105 M–1, which was larger than that with α2-3 sialyllactose (2.18 × 105 M−1, Fig. 5F). The ITC result was consistent with the previous data collected in LSCM and QCM-D, which all proved that Mal-PEI had stronger binding affinity with α2-6 sialyllactose than with α2-3 sialyllactose.

Carbohydrate–carbohydrate interactions between maltose and sialyllactose

The transmembrane ionic current tests (Fig. 2H, I) indicated that maltose played crucial role in the complexation between Mal-PEI and sialyllactose. ITC experiments revealed that the Ka of maltose and α2-6 sialyllactose was approximately 5.10 × 105 M–1 based on a 1:1 binding mode, while the Ka of maltose and α2-3 sialyllactose was 2.97 × 105 M–1 (Fig. S12). A clear difference in Ka indicated that maltose had different interactions with α2-6 and α2-3 sialyllactose, which endowed Mal-PEI with the capacity to distinguish α2-3 from α2-6 sialyllactose. An additional ITC test indicated that maltose also had moderate affinity with galactose (Ka: 2.41 × 104 M−1), which provided extra binding sites with the glycans.

The binding details between maltose and α2-6 sialyllactose or α2-3 sialyllactose were investigated by 1H NMR spectra. Figure 6A shows the number of each hydrogen proton in maltose and sialyllactoses. When maltose interacted with equimolar α2-6 sialyllactose (Fig. 6B–D), remarkable changes were detected. For example, the dOH2 and eOH2 protons in maltose displayed downfield shifts, and the multiplet peaks converged into broad peaks, as indicated by the blue regions. bOH2-4, cOH2,4,6, and aOH7-9 in α2-6 sialyllactose also exhibited a downfield shift, revealing intensive multiple hydrogen bonding interactions between maltose and α2-6 sialyllactose. Similar chemical shift changes were also observed when maltose interacted with α2-3sialyllactose (Fig. 6E–G), but the variations in the dOH2 and eOH2 protons in maltose were smaller. From the comparison of the partial 1H-1H COSY spectrum of 2-6 sialyllactose, 2-3 sialyllactose, maltose, and their mixtures, it could also be concluded that maltose has a stronger binding ability to 2-6 sialyllactose than to 2-3 sialyllactose (Fig. S13).

Fig. 6 : Carbohydrate–carbohydrate interactions and possible explanation for the conformational transition of polymer.
figure 6

A Chemical structures of maltose and α2-6/α2-3 sialyllactose. For ease of clear assignments, “e” and “d” denote Glc, “a”, “b”, and “c” denote the “Neu5Ac”, “Gal”, and “Glc” units of α2-6 or α2-3 sialyllactose, respectively. B–G Partial 1H NMR spectra of maltose (B, E), α2-6 sialyllactose (C) and its equimolar mixture with maltose (D), α2-3 sialyllactose (F) and its equimolar mixture with maltose (G) in DMSO-d6 at 20 °C. H, I Optimized binding models of maltose with α2-6 (H) or α2-3 sialyllactose (I), calculated by quantum chemistry calculations (Gaussian, density functional theory at the 6–31 g level). J Schematic illustration of a potential conformational transition mechanism of the Mal-PEI polymer in response to α2-6 or α2-3 sialyllactose.

Full size image

In addition, possible binding models between maltose and α2-6 or α2-3 sialyllactose were obtained through quantum chemistry based on the Gaussian 09 software package. Figure 6H shows the interaction model of maltose with α2-6 sialyllactose. Five sets of intermolecular hydrogen bonds formed, and the lengths were 1.69, 2.38, 1.62, 1.65, and 1.66 Å, respectively. The short bond lengths corresponded to the strong complexation. By comparison, six sets of intermolecular hydrogen bonds with lengths of 1.88, 2.12, 2.06, 2.40, 3.06, and 2.48 Å formed between maltose and α2-3 sialyllactose (Fig. 6I), and the longer bond lengths indicated that the complexation of maltose with α2-3 sialyllactose was weaker than that with α2-6 sialyllactose. It is worth noting that the binding models shown here are only two possible models. Considering the complexity of carbohydrate–carbohydrate interactions, a more detailed structural analysis should be performed in the future.

Possible explanation for conformational transition of the polymer

Based on the experimental results and interaction models proposed above, a potential conformational change mechanism is proposed to explain the effect observed in the nanochannels. Initially, maltose interacts with neighboring maltose molecules or the secondary amines in PEI through hydrogen bonding interactions, which results in a contracted conformation of the polymeric chain (central panel in Fig. 6J). Strong binding between maltose and α2-6 sialyllactose destroys the initial polymer network, promoting the transition of the polymeric chains from the contracted state to a swollen state (left panel in Fig. 6J). By comparison, the interaction of maltose with α2-3 sialyllactose is weaker than that with α2-6 sialyllactose, and only part of the initial polymer network is broken; thus, the polymeric chains exhibit slight swelling (right panel in Fig. 6J). Remarkable expansion of the polymeric film decreases the diameters of the nanochannels, which blocks transmembrane ionic transport and reduces the ionic current. The Mal-PEI polymers have different binding affinities with α2-6 and α2-3 sialyllactoses. Therefore, the binding-induced expansion degrees of the polymeric chains are different, which is reflected in different variations in the ionic current.

Monitoring of enzymatic reactions

Sialyltransferases are a family of glycosyltransferases that play an integral role in the biosynthesis of Neu5Ac-containing oligosaccharides and glycoconjugates, which are closely related to the occurrence of cancers. To display the application of Mal-PEI nanochannels in monitoring enzymatic sialylation reactions, α2,6-sialyltransferase was introduced to catalyze the transfer of Neu5Ac from cytidine 5′-monophosphate-N-acetylneuraminic acid (CMP-NeuAc) to a lactose substrate40. Briefly, lactose (0.1 μM) and CMP-Neu5Ac (150 μM) were prepared based on a buffer solution (100 mM Tris-HCl, pH 7.5). Then, 2 mL lactose solution and 20 μL CMP-Neu5Ac were injected into the current measurement apparatus mounted with a piece of Mal-PEI-modified nanochannel membrane, and the temperature of the entire apparatus was maintained at ~37 °C. Then, α2,6-sialyltransferase (5 mU) was added to activate the enzymatic reaction (Fig. 7A, top), while the ionic current was continuously recorded. The recorded ionic current value (at +2 V) decreased with the extension of reaction time (Fig. 7B), from 0.29 μA at the beginning to 0.19 μA at 16 min (Fig. 7C). HRMS data further confirmed the formation of α2-6 sialyllactose (Fig. 7A, bottom) in the resulting product mixture. This test was repeated three times with three pieces of Mal-PEI-modified nanochannel membrane. As shown in Fig. 7D, the obtained mean current values along with the standard deviation were plotted versus the reaction time. The current reduction ratio reached a maximum value of approximately 32% at 16 min.

Fig. 7: Monitoring of enzymatic reactions.
figure 7

A Enzymatic sialylation reaction scheme of CMP-Neu5Ac with lactose (top) catalyzed by α2,6-sialyltransferase and HRMS of the product α2-6 sialyllactose (bottom). B Time-dependent current–voltage curves of Mal-PEI-modified PET nanochannels for the enzymatic sialylation reaction. C Time-dependent current at +2 V. D Time-dependent transmembrane ionic current decrease ratio (at +2 V). Error bars represent the standard deviations obtained from three ionic current measurements.

Full size image

Furthermore, the sialylation processes of trisaccharide (i.e., 2′-fucosyllactose, Fig. S14) and glycopeptide (obtained from tryptic digests of IgG, Fig. S15A,B) were monitored. The recorded ionic current value (at +2 V) decreased with the extension of reaction time, and current reduction ratios were approximately 26 and 15% at 16 min for 2′-fucosyllactose and tryptic digests of IgG, respectively. These results revealed the good potential of nanochannel devices for monitoring enzymatic sialylation reactions in real time. It is worth pointing out that the sialylation process of glycoprotein (IgG as an example, Fig. S15C) was difficult to monitor due to the more complex structure and the hidden glycosylation sites of glycoprotein.

Conclusions

Inspired by the SG-regulated VGIC that plays crucial role in influenza A virus infection, we developed a biomimetic ion nanochannel device that had sensitive and selective responsiveness to α2-6-linked sialyllactose, which further realizes the “ON–OFF” ion flux change of the nanochannel. Compared with conventional nanochannel devices modulated by physical or small molecule stimulation, the development of glycan-regulated nanochannels is more challenging due to the complicated composition and structure of glycans. This work indicated that a glycan recognition system and corresponding nanochannel device could be constructed through smart polymer design, and the remarkable globule-to-coil transition of the polymer chain driven by carbohydrate–carbohydrate interactions paved a new route for the future development of glycan-responsive biochips and biodevices. Our future planned work focuses on optimizing functionalized nanochannels and combining functional polymers with a single nanochannel for the single-molecule analysis of glycans. The present study displays the good potential of the nanochannel device in monitoring enzymatic sialylation reactions. This real-time monitoring method avoids the use of high-cost antibodies and complex pretreatment processes, which may promote the development of inhibitor target SGs.