HTML
-
To achieve specificity of detection, functionalization of plasmonic optical fiber to effectively discriminate the E2–streptavidin (STV) was indispensable. The well-established strategy to functionalize functional groups/molecules onto gold film-coated SPR surface is via Au-S bond by using thiol-end organic compounds to form self-assembled monolayers (SAMs) spontaneously30. Desthiobiotin–polyethylene glycol–thiol (DTB-PEG-SH) was used to form SAMs on gold surface, which offered the exposed DTB to bind the released E2–STV conjugate efficiently, while the bovine serum albumin (BSA) solution as a blocking agent was incubated to block the nonspecific binding sites and enhance the specific STV–DTB interaction, and hence increase the signal/noise ratio. Moreover, the PEG linker between -SH and DTB offered better water solubility and reduced the steric hindrance to enhance the reactivity of DTB.
The functionalization of the Au-coated TFBG fiber followed the procedures below, as shown in Fig. 1a. First, the probe was immersed in the deionized (DI) water for 66 min to stabilize this sensor, which can eliminate environmental interference to verify the stability of sensing system. Second, bare gold was functionalized with DTB-PEG-SH crosslinker and then coated with the BSA blocker for 67 and 33 min, respectively. Notably, the DI water was used to rinse the probe for 10 min to remove the residuals at the end of both procedures of DTB-PEG-SH and BSA modification. Figure 1b shows the variable spectra recorded every 20 s under the above three different conditions: pure gold film-coated TFBG in water (black curve); gold film-coated TFBG modified with DTB-PEG-SH (red curve) and then blocked with BSA solution (blue curve). Three modes with the highest sensitivity were selected, respectively, and marked as ①, ②, and ③. Based on the above results, the intensity variations of gold film-coated TFBG as a function of time during the water stability and the DTB-PEG-SH and BSA adsorption on gold surface, respectively, are demonstrated at each selected mode and compared in Fig. 1c. Its inset shows the atomic force microscopic (AFM) images of the morphology of bare and DTB-PEG-SH-modified plasmonic optical fiber. Extensive analysis of the AFM topography cross-sections (Fig. S2) showed that the surface thickness of the chemically modified fiber was slightly increased than that of the bare fiber considering the small size of the DTB-PEG-SH. The response curve versus time also clearly demonstrated that the intensity was stable at the selected highly sensitive mode ① in water while changed when the TFBG fiber was modified by DTB-PEG-SH and BSA.
Fig. 1 Schematic illustration and characterization of the surface chemical modification process.
a Schematic illustration of immobilizing the DTB-PEG-SH and BSA molecules on the fiber surface; b reflected spectra collected every 20 s for stability in water (66 min) and for real-time monitoring in DTB-PEG-SH (67 min) and BSA (33 min) solution; c the response curves of gold film-coated TFBG for the marked modes ①, ②, and ③, i.e., intensity variation as a function of time during the water stability, DTB-PEG-SH, and BSA functionalization processes. Inset shows 3-D AFM topographic images of (a) bare and (b) DTB-modified plasmonic optical fiber. -
The length and structure of the linker between the E2 and STV are critical factors affecting the binding performance of the E2-STV conjugate and hERα ligand-binding domain (LBD). Although the flexible and rigid linkers in the design of estradiol derivatives were considered, the ligands with rigid linkers showed unavoidable atomic position conflict with hERα LBD, indicating that the rigid structure was not suitable for the linker design for E2-STV conjugates (Fig. S3). Among estradiol derivatives with flexible linkers, estradiol derivative 4 was docked to hER-LBD as shown in Fig. S4, and the -COOH terminal was blocked in the hERα LBD, showing that the linker was too short to meet the requirement of connecting STV, so that no further simulation will be performed in the follow-up.
Figure 2 summarizes the molecular dynamics (MD) simulation results of carbonylated estradiol derivatives, named 8, 11, 16, and 20 with different numbers of flexible carbon skeleton as shown in Table 1. The root-mean-square deviation (RMSD) is an accepted index to measure the conformational changes of the MD simulation systems. Figure 2a displays the time evolution of RMSDs for hERα LBD in the simulations with four carbonylated estradiol derivatives, respectively. As expected, a plateau was reached with the RMSDs of all complexes < 1.0 nm within 10 ns. The representative snapshots of complexes at the end of simulations are shown as in Fig. 2a, and the complexes of estradiol derivatives (shown in green) and hERα LBD were compared with the original E2-hERα LBD crystal structure, where E2 is shown in magenta. The comparison confirmed that the estradiol derivatives maintained the binding conformation with hERα LBD in the MD simulation with the tail of the linking arm fully exposed. It also indicated that adding a flexible linker with appropriate length was an effective way not only to retain the capability being recognized by hERα LBD but also to avoid the possible steric hindrance caused by STV. The root-mean-square fluctuation (RMSF) is a measure of the atomic position deviation in a given length of time defined as the atomic position deviation during MD simulations31. The RMSF results further indicated that H12 (residues 531–549) was the most fluctuating part of hERα LBD, while H1–H11 and the ligand-binding pocket usually remained unaffected (Fig. S5). It could be another explanation for why the addition of linkers in the estradiol derivatives had a slight impact on the binding pocket of hERα LBD for estrogen identification.
Fig. 2 Design of E2–STV conjugate based on the MD simulations.
a Time evolution of RMSD for the backbone atoms of hERα LBD and estradiol derivatives 8, 11, 16, and 20 with different flexible carbon skeleton as shown in Table 1 (left) and the snapshot of the estradiol derivative–hERα LBD complex conformation at the end of MD simulation (right). In the right figure, E2 in magenta is used as a position reference for conformations; b total non-bond energy of estradiol derivative and hERα LBD complexes; c Hbond number of estradiol derivatives and hERα complexes.Transducer used Quantitative range (ng ml−1) LOD (ng ml−1) Ref. Electrochemical / / 6 Fluorescent / / 13 Fluorescent 20.8–476.7 1.05 14 Fluorescent 0.1–20 0.1 48 Piezoelectric 2.72–27.2 2.12 18 Nonisotopic / 0.2 10 Surface plasmon resonance 2–6 0.2 15 Surface plasmon resonance / 1 49 Surface plasmon resonance / 1.4 50 Surface plasmon resonance 0.01–100 0.0015 This work Note: "/" means undetectable or not available. Table 1. erformance comparison of this work with other reported nER-based biosensors for EE detection.
Binding energy in the simulations was calculated to predict the binding affinity of hERα LBD and estradiol derivatives. As shown in Fig. 2b, the complex of estradiol derivative 8 and hERα LBD exhibited the lowest average binding energy during the MD simulations, while the simulated binding energies of other derivatives–hERα LBD complex structures were much higher. The results of Hbond formation further confirmed that estradiol derivative 8 was able to form more hydrogen bonds to hERα LBD compared with other derivatives, including 11, 16, and 20 (Fig. 2c), therefore resulting in the higher binding affinity of the complex of estradiol derivative 8 and hERα LBD. Based on the above MD simulation results, estradiol derivative 8 with the most stable binding capability to hERα LBD was chosen for the synthesis of the E2-STV conjugate.
-
To achieve the best sensing performance of plasmonic optical fiber, the most sensitive plasmonic mode located in the SPR envelope to monitor the interactions of molecules was investigated and validated both experimentally and in silico, the mechanism of which is shown in Fig. S6. Figure 3a shows the cladding mode change of plasmonic optical fiber when modified with the DTB-PEG-SH on the gold film in 1 h. As the SPR envelope has a red shift associated with increased RIs, i.e., more molecules attached, the amplitudes of the plasmonic cladding modes (marked as 1, 2, 3, 4) decreased while the other plasmonic modes (marked as 5, 6, 7, 8) increased, and all modes exhibited different relative intensity sensitivities. Among them, the two most sensitive modes 4 and 5 were identified and their zoomed spectral changes are shown in the inset. Figure 3b shows the relative intensity changes of the modes 1–8 during the whole reaction process. The mode 4 closely located at left shoulder of the SPR envelope exhibited the maximum relative intensity response for the DTB-PEG-SH adsorption. Figure 3c shows the total intensity changes of modes 1–8 after 1 h reaction, which exhibited a quantification curve similar to the profile of the SPR envelope.
Fig. 3 Optimization of plasmonic sensing characteristics.
a Experimental SPR spectrum response of the plasmonic optical fiber when modified with the DTB-PEG-SH on the gold film in 1 h (inset: zoomed spectral changes of the two most sensitive modes 4 and 5); b Relative intensity change of the modes 1-8 during the whole DTB-PEG-SH modification; c Total intensity change of modes 1-8 after 1 h reaction; d–f The simulated results corresponding to the conditions in panels a–c, in good agreement with the experimental results.For better comparison with the experimental data, the simulated results corresponding to the above experimental conditions were conducted. The simulation of the transmission spectrum of the fabricated gold film-coated TFBG was carried out by first solving for the modes of the fiber structure (inclusive of core, cladding, metal layers, high refractive index molecular, and water) with a complex vectorial finite-difference algorithm and then using the coupled mode theory as described previously32 for the transmission of TFBG with the aid of MATLAB. The parameters used in the simulation were as follows: a core radius of 4.1 μm with a RI of 1.4545, a cladding radius of 62.5 μm with a RI of 1.4467, a gold coating with a thickness of 50 nm and complex RI of 0.58−i×11, and a RI index of 1.3154 for water. The effective RI of surface molecular over gold coating is 1.48 (Fig. S7)33. Figure 3d shows simulated transmission spectra of the fabricated gold film-coated TFBG in DI water. It is worth mentioning that there was a sharp decrease in the amplitude of cladding mode resonances in the vicinity of 1550 nm, where those cladding modes had transferred energy to a lossy plasmon wave at the gold–water interface. It means that the gold–water interface is highly sensitive to the molecular interactions. The intensity changes of eight selected cladding modes responding to the thickness of the small molecular modification layer were demonstrated and exhibited the same trends with the experimental results, i.e., the decreased amplitudes of modes from 1 to 4 and the increased ones from 5 to 8. All modes showed different sensitivities and the most sensitive modes were also identified to be 4 and 5 and their zoomed spectral changes are shown in the inset. Figure 3e shows the responses of different SPR cladding modes to the thickness of small molecular modification layer from 1 to 3 nm. The mode 4 closely located at left shoulder of the SPR envelope was further confirmed with maximum relative intensity response, which accorded well with the above experimental results. The simulated relative intensities of each cladding modes to 3 nm modification layer (Fig. 3f) also exhibited the same pattern as the SPR envelope, as the experiment revealed.
To summarize, both the above experimental and simulated results identified that the highest sensitivity plasmonic mode of the fabricated gold film-coated TFBG to surface RI was mode 4, which was located at the left shoulder of the SPR envelope.
-
As revealed above, the SPR envelope will shift in wavelength with the changed surface RI and then modulate the other phase-matched cladding modes. The wavelength shift of the SPR envelop was also observed under the different bulk solutions used in the analysis. As a consequence, the intensity response of the biosensor arose from both the biomolecule amount attached on the surface and the selected mode. The mode with the highest sensitivities was changing under different bulk solution conditions, hence possible to cover up the response caused by the target, especially at low concentrations. To mitigate the negative impact caused by the fluctuated bulk RI fluctuations of different samples, we calibrated and tested the reflected spectra under different E2 concentrations in the DI water. Under the same bulk solution conditions, the surface RI change caused by the EEs can be quantified with the intensity response of TFBG under the selected specific mode. More EEs competed with more E2–STV conjugates released from the resin-hERα LBD, which induced their more attachment and hence higher RIs on the TFBG surface. As a consequence, the stronger red shifted SPR envelope caused less attenuation of our selected resonance mode, hence resulting in the stronger relative intensity modulation of narrow cladding mode. All the above provides the theoretical basis for the ultrasensitive quantification of EEs.
Based on the optimized assay conditions (Fig. 3 and Figs. S8 and S9), the performance of TFBG-based SPR biosensor platform was evaluated by using E2 as the model analyte for EEs. Although 1% (w/v) sodium dodecyl sulfate (SDS) solution in phosphate-buffered saline (PBS) was reported to regenerate the antibody–antigen interaction on the nanocoated fiber surface34, the DTB-modified plasmonic optic fiber surface is expected to be stably regenerated by using the washing buffer (0.5% SDS, pH 1.9) in the light of the binding between STV and DTB31. We first tested the stability of temporal spectrum responses for different concentrations of E2 (Fig. 4). The representative real-time intensity curve revealed the stable and target-dependent optical response to E2 with the concentration varying from 0.1 ng ml−1 gradually increasing to 10, 000 ng ml−1, with a dynamic range of more than five orders of magnitude. The signal was reproducible with attenuation < 0.02 dB after the plasmonic optic fiber surface was regenerated.
Fig. 4 Stability of gold film-coated TFBG-based SPR biosensor using hERα LBD as a recognition element for EE detection.
Representative real-time intensity curve responding to different concentrations of E2 and reproducibility by washing buffer.We further evaluated the feasibility of ultrasensitive detection for EEs via the TFBG-based SPR biosensor platform. The spectrum response of monitored SPR mode under different concentrations of E2 is recorded in Fig. 5a. It was observed that higher E2 concentrations were accompanied with lower optical intensities. To better quantify the targets, we use the optical intensity change (ΔIntensity, which is defined as the intensity difference between the base intensity in DI water before reaction and the intensity after certain reactions) to test the sensitivity of this biosensor for EEs.
Fig. 5 Sensitivities of gold film-coated TFBG-based SPR biosensor using hERα LBD as a recognition element for EEs detection.
a Spectral response of SPR mode versus different E2 concentrations; b variations of ΔIntensity with the concentration of E2 and logarithmic calibration plot; c linear range and corresponding calibration plot to detect E2. Inset: intensity fluctuation of a blank sample. The error bars correspond to the standard deviation of the data points in triplicate experiments.We recorded the ΔIntensity of both spectra from 1545 to 1556 nm and mode 4 (inset of Fig. 5a) and related them to the concentrations of E2 (Fig. 5b). A logistic function-fitted calibration curve revealed the linear range of 0.01–100 ng ml−1 between the intensity and E2 concentration with a slope of 0.433 and linearity of 0.99 (Fig. 5c, inset: intensity fluctuation of the blank sample). A dramatically linear dynamic range of four orders of magnitude was achieved with a linear fitting function of y = 1.234 + 0.433*x. The limit of detection (LOD) of this biosensor toward E2, defined as the concentration in which the signal to noise ratio is 3, was calculated to be 1.5 × 10−3 ng ml−1, considering that the standard deviation of blank was 0.003 dB (inset of Fig. 5c)35. It was one order lower than that of the defined maximal E2 level in drinking water, i.e., 8.0 × 10−2 ng ml−1 set by the Japanese government. The LOD of this biosensor for EE detection was compared with other reported nER-based biosensors (Table 2), and the sensitivity superiority of this biosensor is impressive.
Sample Spiked E2 (ng ml−1) Detected E2 (ng ml−1) Recovery (%) Coefficient variation (%) Tap water 0 / / / 1 0.91 ± 0.06 91 6 10 11.23 ± 0.49 112 5 Pond water 0 / / / 1 1.06 ± 0.10 106 10 10 9.80 ± 0.69 98 7 Note: "/" means undetectable or not available. Table 2. Recovery of E2 in real-water samples using the TFBG-based SPR biosensor (n = 3).
We attributed the sensitivity superiority of this biosensor to the following reasons. Through compensating the red shift of the SPR envelope by using the same bulk solution for testing, the gold–water interface of this biosensor was proven to be highly sensitive to a small change of RIs. Moreover, the EE-induced sensing interface change was characterized by the biological macromolecule of E2-STV conjugate through the rational sensing design, which greatly amplified the signals.
The broad sensing capability of this biosensor arises from the specific binding capability between EEs and nERs. To demonstrate its capability for detecting broad ranges of EEs, we selected six phthalate esters, which were potential EEs widely reported by numerous studies31, 36, for investigation. The normalized signal variations responding to 10, 000 ng ml−1 phthalate esters, including butyl benzyl phthalate (BBP), diisopentyl phthalate (DIPP), diethyl phthalate (DEP), dimethyl phthalate (DMP), and dioctyl phthalate (DOP), compared with that caused by 10 ng ml−1 E2 are summarized in Fig. S10. The signal decreased in the order BBP > DIPP > DEP > DMP, indicating the relatively weaker estrogen-agonist potencies; however, DOP showed a comparable signal as the control sample, indicating no evidence of estrogenic activities measured by this hERα-based biosensor. The results obtained above are consistent with previous studies31, 36.
Moreover, the interaction of hERα LBD with the investigated phthalate esters was simulated by molecular docking and MD. The complex of hERα LBD and four phthalate esters exhibited stable docking poses with the molecular docking scores (Dscores) in the order BBP < DIPP < DEP < DMP (Fig. S11a). The MD simulations also revealed that the binding energy of EE–hERα LBD complex structures showed the same changing trend as the Dscore (Fig. S11b). However, no stable docking pose or Dscore was generated for the complex of DOP and hERα LBD, indicating that DOP was not suitable for hERa LBD binding. The in silico studies that predicted the estrogen-agonist potencies of the phthalate esters are given as follows: BBP > DIPP > DEP > DMP, while the hERa-active potencies of DOP was negligible. The above in silico simulation results accorded well with the experimental results, which further confirmed the biosensing capability of this technique towards broad ranges of EEs.
To demonstrate its capability for practical and real-water samples, two types of real-water samples including laboratory tap water and pond water from the campus of Tsinghua University spiked with different concentrations of E2 were detected and evaluated by using the biosensing platform. Only the pond sample was filtered through a 0.45 μm filter (Millipore Corp., Bedford, MA) before spiking. Two spiked concentrations (1 ng ml−1, 10 ng ml−1) of E2 were chosen according to the sensitivity of our developed biosensor. The concentrations measured by this biosensing platform were compared with spiked concentrations, and results are listed in Table 3. When the non-spiked samples were pumped into the biosensor, no significant change in signal was captured, indicating that EEs were non-detectable in any samples by using this technique. For spiked samples, both for the case of low and high E2 concentrations, the calculated concentrations were close to original spiked values. Notably, the recovery rates in the pond water were more satisfactory than in the laboratory tap samples, considering that the tap water comes from deep groundwater sources and the hardness of this tap water is in the range of 290–350 μg ml−1 for calcium carbonate37. The high hardness could be one of the possible reasons affecting the interaction between hERα and EEs, hence the biosensor signal. In summary, the recovery rates of E2 ranged from 91 to 112%, demonstrating the satisfactory accuracy of this biosensor and indicating the application potential in real-water samples with a simple pretreatment.
Table 3. Eleven estradiol derivatives used in the simulation.