Figures and data
Experimental principle and nanofabrication.
(A) Sketch of the experimental principle. Nanopores in a metal membrane block light from traversing if the pore diameter is small compared to the wavelength of light. The selectivity of Nsp1-coated metal nanopores is probed by measuring the translocation rate of fluorescently labeled proteins from the top reservoir (cis) to the detection (trans) side, where they rapidly diffuse out of the laser focus. Measurements on open pores (top) serve as a control where both the NTR Kap95 and the inert protein probe BSA pass unhindered. Nsp1-coated pores are expected to block the translocation of BSA while still allowing Kap95 to translocate. Zooms at bottom right illustrate the passivation of open pores with MUHEG (top) and functionalization of the palladium surface with the FG-nucleoporin Nsp1 and 350 Da SH-PEG (bottom), achieved via thiol-palladium chemistry. (B) Fabrication of nanopores in a free-standing palladium membrane was performed by physical vapor deposition of palladium onto silicon nitride (SiNx), reactive ion etching (RIE), and focused-ion-beam (FIB) milling of the nanopores. The palladium surface was then cleaned either with H2O2 or ethanol to remove contaminants before the functionalization step.
Experimental setup.
(A) A free-standing Pd membrane containing the nanopores was mounted on a confocal microscope using a PDMS flow cell with a reservoir of ≈3 μL on the dark (cis) side and a flow channel on the detection (trans) side that faced the objective lens. A constant flow in the channel avoided the accumulation of analytes on the detection side. The lasers were focused onto the nanopore by a high NA objective lens and the fluorescence signal was detected on single-photon avalanche photodiodes. From the recorded photon arrival times, fluorescence bursts were detected using a change point detection algorithm. (B) A total of eight pores were milled into the Pd membrane, each surrounded by partially milled markers that facilitate the localization of the nanopores in a bright-field image (top). An additional marker was added such that the individual pores in the array could be identified. A scanning electron microscope image of a single pore with markers is shown below. The size and shape of each ZMW pore used in this study was determined using transmission electron microscopy (bottom right). (C) Simulated electric field intensity distributions in the xz plane near a freestanding ZMW for pore diameters of 30 (top) and 150 nm (bottom). A Gaussian laser beam was focused on the pore at a wavelength of 488 nm, polarized in the x-direction. See Appendix 2—Figure 1 for different pore sizes and excitation at 640 nm. (D) Electric field intensity |E2| (top) and total detected signal S(z) (bottom) as a function of the z-position along the center of the pore for an excitation wavelength of 488 nm and an emission wavelength of 525 nm, corresponding to the blue detection channel, for pores of 50, 100, and 150 nm diameter. The palladium membrane is indicated by the gray shaded area. See Appendix 2—Figure 4 for the corresponding plots for the red channel.
Experimental workflow for measuring the selectivity of Nsp1-coated nanopores.
(A,B) Two pores of approximately 48 nm diameter were coated with MUHEG (A) or functionalized with Nsp1 (B). Pore dimensions were measured from TEM micrographs. (C,D) Fluorescence time traces recorded for the open and Nsp1-coated pores that are shown in A and B for Kap95–Alexa647 at 100 nM (red) and BSA–Alexa488 at 250 nM (blue). Both proteins were present at the same time. Whereas the Kap95 signal is comparable between the open and Nsp1-coated pores, a clear decrease of the BSA event rate is evident for the Nsp1-coated pore. (E,F) Measured event rates (top panel) resulting from the analyte concentrations for the different conditions probed sequentially during the experiment (bottom panel; see Appendix 4 for details). The concentrations used in the time traces shown in C and D are indicated with an asterisk. (G) In order to compare the different conditions, the obtained event rates are normalized to the respective protein concentration and corrected for the labeling degree (white dots in G). Bars indicate the average normalized event rates kKap95 or kBSA of the pore. Error bars represent the standard error of the mean. (H) The selectivity was calculated as the ratio of the average normalized event rates 
Dependence of translocation rates and selectivity on pore diameter
(A–D) Concentration-normalized event rate of Kap95 (A) and BSA (B) as a function of pore diameter for open pores (cyan squares) and Nsp1-coated pores (purple circles). While the normalized event rate for Kap95 did not change significantly between open and Nsp1-coated pores, a clear reduction was observed for BSA, which was most pronounced at small pore sizes. Solid lines are fits to a quadratic function given in Equation 2. To model the size dependence of BSA translocations through Nsp1 pores, an offset was introduced that shifts the onset of the quadratic curve to higher diameters (dashed line, Equation 3). (C, D) Zoom-ins of the indicated regions in A,B. In A–D, the error bars represent the standard error of the mean of the normalized event rates obtained at different protein concentrations. (E) Apparent selectivity versus pore diameter. The data show that selectivity was lost for Nsp1-coated pores with increasing diameter. The average selectivity for open pores of 0.70 ± 0.01 is shown by the dashed cyan line. Error bars indicate the propagated error from the normalized event rates shown in A–D.
Coarse-grained modeling of Nsp1-coated pores.
(A) One-bead-per-residue representation of an Nsp1-coated nanopore. (B) Top views of the Nsp1 meshwork in pores of 50 nm, 60 nm and 80 nm diameter (for a grafting density of 1 Nsp1 per 300 nm2). For pores with a diameter below 60 nm the Nsp1 meshwork was closed throughout the entire simulation, while for pores with diameters ≈60 nm transient openings were observed in the center of the pore, that were persistent for diameters ≥80 nm. (C) Calculated event rate of BSA as a function of pore diameter for open pores (cyan filled dots) and Nsp1-coated pores (purple dots) for a grafting density of 1 Nsp1 per 300 nm2. The calculated event rates are in good agreement with the experimental event rates (open squares). Solid lines are fits to the calculated event rates using the quadratic function given in (2) for open pores and (3) for Nsp1-coated pores. (D) Calculated selectivity versus pore diameter, showing that the selectivity is lost for Nsp1-coated pores with increasing diameter. The average selectivity for open pores is shown as a green dashed line. The effect of grafting density on the apparent selectivity is indicated by the purple dotted lines that depict the results for grafting densities of 1 Nsp1 per 200 nm2 and 1 Nsp1 per 400 nm2 (see Appendix 10—Figure 3). (E) Axi-radial and time-averaged protein density distributions inside Nsp1-coated pores of 50 nm, 60 nm and 80 nm diameter at a grafting density of 1 Nsp1 per 300 nm2. For pore diameters below 60 nm, we observed the highest protein density along the pore axis, while for diameters ≥60 nm, the highest density was found near the pore walls. This observation was valid for each of the probed Nsp1 grafting densities (see Appendix 10—Figure 4). (F) Axi-radial and time-averaged void distributions inside Nsp1-coated pores of 50 nm, 60 nm and 80 nm diameter at a grafting density of 1 Nsp1 per 300 nm2. The void distributions suggest that for pores of 50 nm diameter there is no preferred pathway for the BSA proteins, while for larger pores the translocations happen mostly along the central axis.
Figure 5—video 1. Video of the top view of the pores presented in B for 40 ns of the simulation.
BSA permeation depends on Kap95 concentration.
(A) BSA event rates measured at the different Kap95 concentrations of 0 nM, 100 nM, and 1000 nM were averaged to obtain the Kap95-concentration dependent event rates kBSA,0, kBSA,100 and kBSA,1000. Data shown were obtained using an Nsp1-coated pore with a diameter of 116 nm. (B,C) Plots of the relative change of the BSA event rate measured at 100 nM and 1000 nM Kap95 compared to no Kap95, i.e.
MUHEG grafting established on QCM-D.
(A,B) The frequency (F1-F7) and dissipation (D1-D7) response of of the different harmonics (numbers) for two QCM-D sensors versus time upon grafting of 250 μM MUHEG in ethanol (see Cleaning and surface grafting of Pd). The blue vertical lines show when the solution was switched in the flow cell. The grey shaded regions show the time when the solution was flowed through the flow cell. Upon MUHEG binding a decreasing frequency can be observed, which shows that mass attaches to the sensor’s surfaces.
MUHEG passivation established on QCM-D.
(A,B) Same chips and setup as in Appendix 1—Figure 1 The frequency response of flushing 500 nM of BSA and 500 nM of Kap95 over the MUHEG passivated surface of the QCM-D chips only shows a minor frequency shift of less than 5 Hz (KCl level before and after flushing the proteins).This suggests that the Pd surface can be effectively passivated against adhering proteins by a MUHEG coating.
Nsp1 binding on QCM-D
The frequency (F1-F7) and dissipation (D1-D7) response of of the different harmonics (numbers) for a Pd coated QCM-D sensors versus time upon grafting of 1.14 μM Nsp1 in PBS. The QCM-D chip was cleaned by boiling ethanol (see Cleaning and surface grafting of Pd). The frequency shift from before the Nsp1 coating to after was approximately 60 Hz which shows an acceptable coating efficiency. Nsp1 was flushed in the grey shaded area.
Threedimensional finite-difference time-domain simulations of the electric field intensity distribution |E|2 in units of V 2/m2 in the proximity of the ZMW for excitation by a diffraction-limited focused Gaussian beam with wavelengths of 488 nm (A) and 640 nm (B). The lower side of the 100 nm thick palladium membrane is placed at z = 0 nm. The source is located at the bottom and the electric field is polarized in the x-direction. The electric field intensity distributions are shown for pores with a diameter of 50, 100, and 200 nm (from top to bottom) in the xz (left) and yz (middle) planes passing through the center of the pore, and the xy (right) plane at the entrance to the pore at z = 0 nm. See Appendix 2—Figure 2 for a zoomed out representation of the intensity distribution.
Threedimensional finite-difference time-domain simulations of the electric field intensity distribution of the excitation spot in the presence of a freestanding ZMW (compare Appendix 2—Figure 1), at excitation wavelengths of 488 nm (A-D) and 640 nm (E-H). Shown are the intensity distributions of the focused Gaussian beam in the absence (A,E) and presence (B-C,F-G) of the freestanding palladium ZMW. (D,H) The z-profiles of the intensity distribution along the center of the pore. The position of the palladium membrane is indicated as a gray shaded area.
Threedimensional finite-difference time-domain simulations of the electric field intensity distribution |E|2 in units of V 2/m2 in the proximity of the ZMW for excitation by a plane wave with wavelengths of 488 nm (A) and 640 nm (B). The lower side of the 100 nm thick palladium membrane is place at z = 0 nm. The source is located at the bottom and the electric field is polarized in the x-direction. The electric field intensity distributions are shown for pores with a diameter of 50, 100, and 200 nm (from top to bottom) in the xz (left) and yz (middle) planes passing through the center of the pore, and the xy (right) plane at the entrance to the pore at z = 0 nm.
Finite-difference time-domain simulations of the dipole emission in the proximity of the freestanding ZMW. (A) A scheme of the simulation setup. The dipole is placed in the center of the pore in the xy plane at varying z-positions. The detected signal is monitored on the detection (i.e., lower) side. (B-C) From top to bottom: The z-profiles of the excitation probability, the detection efficiency η, the emitter quantum yield Φ, and the total detected signal, along the center of the nanopore are shown for the blue (B, λex =488 nm, λem =525 nm) and red (C, λex =640 nm, λem =670 nm) channels. The total detected signal S(z) is defined as the product of the excitation intensity, detection efficiency, and quantum yield. (D-E) Predicted fluorescence lifetimes τ of BSA–Alexa488 and Kap95–Alexa647. The position of the palladium membrane is indicated as a gray shaded area. The weighted averages of the fluorescence lifetime based on the detected signal S(z), ⟨τ⟩S, are shown as colored horizontal dashed lines. The gray dashed line indicates the measured fluorescence lifetime τ0 in the absence of the ZMW. The predicted signal-averaged lifetimes ⟨τ⟩S are 1.98, 1.88, and 1.88 ns for BSA–Alexa488, and 1.22, 1.95, and 1.15 ns for Kap95–Alexa647, for pore diameters of 50, 100, and 150 nm, respectively (see eq. 13). The quantum yields and fluorescence lifetimes were estimated based on a literature values of Φlit = 0.8 and τlit = 4.0 ns for Alexa488 (Sanabria et al., 2020), and Φlit = 0.33 and τlit = 1.17 ns for Alexa647 (Hellenkamp et al., 2018), and measured lifetimes in the absence of the ZMW of τ0 =2.3 ns for BSA–Alexa488 and τ0 = 1.37 ns for Kap95–Alexa647 (compare Appendix 7—Figure 1).
Z-profiles of the radiative (emission) and non-radiative (loss) rates in the proximity of the freestanding ZMW obtained from FDTD simulations for the blue (A, λex =488 nm, λem =525 nm) and red (B, λex =640 nm, λem =670 nm) channels. The position of the palladium membrane is indicated as a gray shaded area. The z-axis is defined as in Appendix 2–Figure 4 A. For the radiative rate (bottom), the rate of emission directed towards the objective lens is displayed in addition as a dashed line. The normalized loss rate 
Free fluorophore translocations
(A) Event rate versus pore diameter for Alexa 647. (B) Event rate versus pore diameter for Alexa 488. The event rate of both fluorophores does barely change between Nsp1-coated pores and open pores.
Pressure-dependent reduction of translocation rates.
A) A solution of the free fluorophores Alexa488 and Alexa647 at a concentration of 100 nM was placed in the reservoir. The signal count rate was monitored at the exit of a pore with a diameter of 74 nm as a function of the applied pressure to the flow channel. Count rates were normalized to the values obtained in the absence of a pressure difference. Due to the pressure-induced hydrodynamic flow against the concentration gradient, the translocation rates decrease linearly with the applied pressure. B) Event rates for Kap95 at a concentration of 1 μM acquired for three different open pores with diameters in the range of 67-84 nm. The event rates decrease markedly at a high pressure of 200 mbar, but remain approximately constant in the range below 100 mbar.
Signal-to-background ratios of single molecule events
Distributions of the signal-to-background ratio for BSA (left) and Kap95, defined as the ratio of the event signal to the average background signal. The background level was estimated as the average photon rate in the absence of fluorescently-labeled analytes. The dot represents the mean and the bar the standard deviation of the distribution.
(A,B) Normalized and averaged event rate vs. pore diameter as in Figure 4. Solid lines are fits of a diffusion model as described in this section. A single fit for BSA through Nsp1 would deviate both for small and large pore diameters. Therefore the dataset was split into two regimes: Below 50 nm and above 60 nm in order to fit the data. This threshold was inspired by the results of the CGMD simulations. The dashed line shows a fit of the shifted quadratic function to the BSA data through Nsp1 pores as described in the main text. (C,D) Zoom ins of (A,B). E: Selectivity of open pores, predicted by the diffusion model. The value is pore size dependent because of the different sizes of Kap95 and BSA.
(A,B) Normalized event rate vs. pore diameter of the individual conditions underlying the points shown in Figure 4. Solid lines are fits of a quadratic function as given in Equation 16. The BSA event rates through Nsp1-coated pores were not well described by the quadratic function. Therefore, a shift parameter was introduced as given in eq. Equation 17 (dashed line). (C,D) Zoom ins of (A,B).
(A,B) Normalized and averaged event rate vs. pore diameter as in Figure 4. Solid lines are fits of a quadratic function as given in Equation 16. The fit for BSA through Nsp1 deviates both for small and large pore diameters. Therefore a shift parameter was introduced as given in Equation 17 to fit the data (dashed line).(C,D) Zoom ins of (A,B). (E) Open pore event rates obtained from the quadratic fit divided by Nsp1 pore event rates obtained from the quadratic fit vs. pore diameter. While there is barely any decrease of the event rate for Kap95 when Nsp1 is present, BSA experiences an approximately 10-fold decrease for small pores of 35 nm diameter. This ratio increases with pore diameter and approaches a value of 1 in the limit to very large pores.
(A,B) Scatter plots of the events detected for Kap95 (A) and BSA (B) indicating the distributions of event duration and the amount of photons within an event for the data shown in Figure 3 C,D. Both distributions overlap, showing that the spike detection works equally for coated and open pores. (C,D) Plots of the diffusion coefficient vs. pore diameter for Kap95 and BSA, respectively. The diffusion coefficient is estimated by FCS analysis of the time traces obtained in the nanopore experiments at the highest protein concentration. The horizontal lines indicates the average diffusion coefficient and the width corresponds to twice the standard error of the mean. The average diffusion coefficient of both Kap95 and BSA shows no significant difference between open pores and Nsp1-coated pores. This indicates that interactions of the proteins with the Nsp1 mesh do not obstruct the diffusion. Alternatively, it is possible that the bound fraction is not detected in our experiments if it is close to the metal surface due to metal-induced quenching of the fluorescence signal. (E,F) Plots of the fluorescence lifetime vs. pore diameter for Kap95 and BSA, respectively. The fluorescence lifetime is calculated based on the individual time traces of the highest protein concentration. The mean (horizontal lines) with twice the standard error of the mean (width of the lines) gives an estimate of the spread. For both Kap95 and BSA, the average fluorescence lifetime is significantly lower than what is measured in open solution (black lines). This can be attributed to the influence of the nearby metal nanostructure on the radiative and non-radiative rates. The predicted lifetimes based on FDTD simulations are shown as red crosses (compare Appendix 2—Figure 4). Whereas for BSA the lifetime in Nsp1-coated pores and open pores does not differ significantly, there is a significant decrease of the fluorescence lifetime of Kap95 in Nsp1-coated pores compared to open pores. This suggests that Kap95 remains within the proximity of the pore for a longer time when Nsp1 is present.
Correlation between event rates for Kap95 and BSA. (A) Normalized event rates of Kap95 vs. BSA for small (purple) and large (pink) Nsp1-coated pores and open pores (cyan). The event rates show a high degree of correlation with correlation coefficients of 0.72, 0.93, and 0.98 respectively. Additionally pores with diameters of 55 nm to 56 nm are shown in grey. (B) zoom in of (A).
(A, B) Normalized event rates for different open pore and Nsp1 pore experiments, respectively. The different data sets were highlighted according to their experimental day. A label for each experimental realisation is given in the legend. We don’t observe any striking day-to-day variation. (C,D) Selectivity for different open pore and Nsp1 pore experiments. Also in the selectivity we don’t observe a striking difference between experimental repetitions.
Potentials of mean force along the pore axis of open pores with various diameters derived from void analysis using a probe radius of rprobe of 34 Å for BSA and 40 Å for Kap95. The numbers next to the dashed lines indicate the permeability barriers ΔE.
Potentials of mean force along the pore axis of Nsp1-coated pores with various diameters and grafting densities using a probe radius of rprobe = 34 Å (BSA) derived from void analysis. The height of the PMF barrier, ΔE given on the right of the plots, is obtained by averaging the PMF between −25 nm ≤ z ≤ 25 nm (gray region). The location of the peaks in the PMF curves are at the same z-coordinates as the anchoring points of the Nsp1 proteins.
Effect of grafting density on the selectivity of Nsp1 pores. (A) Calculated BSA event rates for Nsp1 pores for various grafting densities. (B) Zoom-in of A. (C) Apparent selectivity versus pore diameter, where the selectivity is calculated as the ratio of Kap95 to BSA event rates. For Kap95 we assumed that the event rate is the same for open pores and Nsp1 pores.
Time-averaged r–z density distribution of Nsp1-coated nanopores for various diameters and grafting densities. Although there is a significant variation in the central channel densities for small diameter pores, the range of diameters at which the structural transition of the Nsp1 mesh takes place is largely independent of the Nsp1 grafting density in the range that was tested.
(A) Rate change of Kap95 vs. pore diameter. Event rate of Kap95 at 1000 nM (kKap95,1000) divided by the Kap95 event rate at 100 nM (kKap95,100). At high Kap95 concentrations there is an increase in normalized event rate for both open and Nsp1 pores. It is largest for small Nsp1 pores. (B) The averages indicated in the plots by horizontal lines given with standard deviation estimated from fitting horizontal lines. (C,D) Influence of BSA concentration on Kap95 event rate. When dividing the Kap95 event rate with 250 nM of BSA being present kKap95,B250 by the Kap95 event rate with 0 nM of BSA being present kKap95,B0 (C) this shows that the Kap95 event rate is barely influenced by the presence of BSA, as expected. The same holds for the Kap95 event rate change with 500 nM of BSA to 0 nM being present. (E,F) Selectivity vs. pore diameter for different Kap95 concentrations. When switching from 100 nM of Kap95 to 1000 nM of Kap95, the selectivity of small Nsp1 pores increases by a factor of 3, whereas the selectivity of large pore decreases.
(A,B) Molecular brightness per time trace versus pore diameter on Nsp1 pores. We see two populations for Kap95, labeled in purple and yellow. The yellow data sets showed a decreased molecular brightness and therefore these data sets were removed from further analysis. (C,D) Comparison of molecular brightness between open pores (cyan), Nsp1 pores (purple), and free diffusion (black horizontal line). For Kap95 we found barely any difference of the data distribution. For BSA there we observed a decrease in molecular brightness for small pores of Nsp1. This can be explained by the lower event rate in these pores. Different markers show different experiments.
(A,B) Event rates of discarded pores of Nsp1. For completeness, we show the event rates of pores that were not further considered in the analysis (yellow) due to their lower molecular brightness, as shown in the previous figure. As expected, these pores show a much decreased normalized event rate both for Kap9 BSA compared to the Nsp1 pores that were kept in the analysis (purple). (C,D) Zoom in of (A,B). (E) Selectivity of discarded pores. The selectivity of discarded pores (yellow) deviates from the selectivity of the kept pores (purple). This can be explained when taking into account that a changed molecular brightness influences the event detection in each channel differently.
