Zero-Mode Waveguide Nanowells for Single-Molecule Detection in Living Cells

Single-molecule fluorescence imaging experiments generally require sub-nanomolar protein concentrations to isolate single protein molecules, which makes such experiments challenging in live cells due to high intracellular protein concentrations. Here, we show that single-molecule observations can be achieved in live cells through a drastic reduction in the observation volume using overmilled zero-mode waveguides (ZMWs- subwavelength-size holes in a metal film). Overmilling of the ZMW in a palladium film creates a nanowell of tunable size in the glass layer below the aperture, which cells can penetrate. We present a thorough theoretical and experimental characterization of the optical properties of these nanowells over a wide range of ZMW diameters and overmilling depths, showing an excellent signal confinement and a 5-fold fluorescence enhancement of fluorescent molecules inside nanowells. ZMW nanowells facilitate live-cell imaging as cells form stable protrusions into the nanowells. Importantly, the nanowells greatly reduce the cytoplasmic background fluorescence, enabling the detection of individual membrane-bound fluorophores in the presence of high cytoplasmic expression levels, which could not be achieved with TIRF microscopy. Zero-mode waveguide nanowells thus provide great potential to study individual proteins in living cells.

• Supplementary Figure to Figure 1 -Additional SEM images and layouts of version 1 and 2 arrays.
• Supplementary Figures to Figure 2 -Excitation field intensity distributions from FDTD simulations under widefield excitation at λ ex = 488 nm.
-Excitation field intensity distributions from FDTD simulations under widefield excitation at λ ex = 640 nm.

S-1
-Excitation field intensity distributions from FDTD simulations under TIRF illumination at an angle of 70°at λ ex = 488 nm.
-Excitation field intensity distributions from FDTD simulations under TIRF illumination at an angle of 70°at λ ex = 640 nm.
-Excitation field intensity distributions from FDTD simulations under excitation by a focused Gaussian beam at NA = 1.2 and a wavelength of λ ex = 488 nm.
-Excitation field intensity distributions obtained from FDTD simulations under excitation by a focused Gaussian beam at NA = 1.2 and λ ex = 640 nm.
-FDTD simulations of fluorescence emission and detected signal from overmilled ZMWs for Alexa488.
-FDTD simulations of fluorescence emission and detected signal from overmilled ZMWs for JFX650.
-Overview of radiative and non-radiative rates obtained from FDTD simulations of fluorescence emission within overmilled ZMWs.
-FDTD simulations of dipole emission as a function of the distance to the pore walls.
-Estimation of background signal from FDTD simulations for the dyes Alexa488 and JFX650 under the different excitation modes.
• Supplementary Figures to Figure 3 -Experimental characterization of photophysics and diffusion within ZMWs.
-Extracted parameters for the dye Alexa488 in overmilled Pd ZMWs.
-Extracted parameters for the dye JFX650 in overmilled Pd ZMWs.
-Quantification of the observation volume in overmilled ZMWs.
-Comparison of extracted parameters for the dyes Alexa488 and JFX650 in ZMW.

S-2
-Correlation between fluorescence lifetime and molecular brightness in ZMW for Alexa488 and JFX650.
-FDTD simulations of excitation field and fluorescence emission in the absence of a Pd layer.
-Comparison of experimental and predicted enhancement factors in overmilled Pd ZMWs.
• Supplementary Figures to Figure 4 -Stability of the BFP signal during the experiments.
-Correlated imaging of flipped coverslips confirms the colocalization of pores with BFP signal with the presence of cells on the ZMW array.
• Supplementary Figures to Figure 5 -Brightfield and BFP-fluorescence images of array version 2.
-BFP fluorescence intensity switching between high and low signal levels.
• Supplementary Figures to Figure 6 -BFP and JFX650-HaloTag signal of pores showing low BFP signal.Supplementary Figure 11: FDTD simulations of dipole emission as a function of the distance to the pore walls.A: Schematic of the simulation setup.The dipole was placed at varying distances ∆x from the pore walls and the emission was monitored as a function of the z-position.B,C: Z-profiles of the normalized radiative and loss rates, the detection efficiency η, the quantum yield Φ, and the product of the detection efficiency and quantum yield, Φ • η for Alexa488 (B) and JFX650 (C) at the indicated excitation and emission wavelengths.The quantum yield of the free dye is shown as a dashed line.The dipole emission within the overmilled volume in the glass was found to be approximately independent of the lateral displacement within the pore at distance of ≈ 25 nm away from the ZMW.Within the ZMW, the non-radiative rate is strongly increased as the dipole approaches the pore wall, with significant non-radiative losses occurring at distances below 25 nm that result in a reduction of the quantum yield.Note that the loss rate is given on a log scale.The position of the palladium layer is indicated as a gray shaded area.
Supplementary Figure 12: Estimation of background signal from FDTD simulations for the dyes Alexa488 (A-F) and JFX650 (G-L) under the different excitation modes.A-C, G-I: The detected signal S(z) obtained for widefield (WF), TIRF, and focused excitation (solid lines) is extrapolated by fitting the signal profile in the z-range from 100 nm to 200 nm to an exponential decay (dashed lines).D-F, J-L: From the extrapolated signal profiles, the amount of signal detected from the upper side (above a z position of 100 nm) is estimated for different pore diameters and milling depths.The bars for 100 nm are barely visible due to their small height.S-15 Supplementary Figures to Figure 3 time  |, normalized radiative rate γ r /γ 0 r , normalized loss rate γ loss /γ 0 r quantum yield Φ, detection efficiency η, and total fluorescence signal S(z) as a function of the z position in the presence (D) and absence (E) of the Pd layer.In D, the position of the metal membrane is indicated as a gray shaded area.In E, the position of the SiO 2 layer is indicated as a blue shaded area (except for the free diffusion case).The detection efficiency was set to 0.5 in the absence of the Pd layer.A small radiative rate enhancement arises even in the absence of the Pd layer when the dipole is placed in the SiO 2 nanocavity due to the Purcell effect S6 .

1 :
Additional SEM images and layouts of version 1 and 2 arrays.A,B: SEM images of overmilled Pd ZMWs, imaged under 52 °.Pd shows up as a bright layer and glass as a dark layer.C: An SEM image of a Pd ZMW array of version 1 containing five regions of different milling depths (h), each made from nine rows of pores with varying diameter.D: Zoom in of C. E: Taper angle vs. milling depth.For deeper pores, the edges were more perpendicular with an average of 22 °(horizontal line).F: Diameters (d) and depths (h) in nm of ZMW arrays for arrays of version 2 together with their uncertainty (estimated from measuring several pores).S-4 Supplementary Figures to Figure2

2 : 3 : 5 : 6 : 7 : 8 : 9 :
Excitation field intensity distributions from FDTD simulations under widefield excitation at λ ex = 488 nm.A: Schematic of the simulation setup.B-E: Excitation field intensity distributions |E| 2 in V 2 /m 2 in the x-z (left), y-z (middle) and x-y plane at the entrance to the ZMW (right) at pore diameters d of 50 nm (B), 100 nm (C), 200 nm (D), and 300 nm (E) and milling depths h of 0 nm (top), 100 nm (middle), and 200 nm (bottom).The electric field is polarized along the x-axis.Excitation field intensity distributions from FDTD simulations under widefield excitation at λ ex = 640 nm.A: Schematic of the simulation setup.B-E: Excitation field intensity distributions |E| 2 in V 2 /m 2 in the x-z (left), y-z (middle) and x-y plane at the entrance to the ZMW (right) at pore diameters d of 50 nm (B), 100 nm (C), 200 nm (D), and 300 nm (E) and milling depths h of 0 nm (top), 100 nm (middle), and 200 nm (bottom).The electric field is polarized along the x-axis.Excitation field intensity distributions from FDTD simulations under TIRF illumination at an angle of 70°at λ ex = 488 nm.A: Schematic of the simulation setup.B-E: Excitation field intensity distributions |E| in V 2 /m 2 in the x-z (left), y-z (middle) and x-y plane at the entrance to the ZMW (right) at pore diameters d of 50 nm (B), 100 nm (C), 200 nm (D), and 300 nm (E) and milling depths h of 0 nm (top), 100 nm (middle), and 200 nm (bottom).The electric field is polarized along the x-axis.Excitation field intensity distributions from FDTD simulations under TIRF illumination at an angle of 70°at λ ex = 640 nm.A: Schematic of the simulation setup.B-E: Excitation field intensity distributions |E| 2 in V 2 /m 2 in the x-z (left), y-z (middle) and x-y plane at the entrance to the ZMW (right) at pore diameters d of 50 nm (B), 100 nm (C), 200 nm (D), and 300 nm (E) and milling depths h of 0 nm (top), 100 nm (middle), and 200 nm (bottom).The electric field is polarized along the x-axis.Excitation field intensity distributions from FDTD simulations under excitation by a focused Gaussian beam at NA = 1.2 and a wavelength of λ ex = 488 nm.A: Schematic of the simulation setup.B-E: Excitation field intensity distributions |E| 2 in V 2 /m 2 in the x-z (left), y-z (middle) and x-y plane at the entrance to the ZMW (right) at pore diameters d of 50 nm (B), 100 nm (C), 200 nm (D), and 300 nm (E) and milling depths h of 0 nm (top), 100 nm (middle), and 200 nm (bottom).The electric field is polarized along the x-axis.Excitation field intensity distributions obtained from FDTD simulations under excitation by a focused Gaussian beam at NA = 1.2 and λ ex = 640 nm.A: Schematic of the simulation setup.B-E: Excitation field intensity distributions |E| 2 in V 2 /m 2 in the x-z (left), y-z (middle) and x-y plane at the entrance to the ZMW (right) at pore diameters d of 50 nm (B), 100 nm (C), 200 nm (D), and 300 nm (E) and milling depths h of 0 nm (top), 100 nm (middle), and 200 nm (bottom).The electric field is polarized along the x-axis.FDTD simulations of fluorescence emission and detected signal from overmilled ZMWs for Alexa488.A: Schematic of the simulation setup.B-D: Computed quantum yield Φ, detection efficiency η, and fluorescence lifetime τ profiles of the dye Alexa488 as a function of the z position.Dashed lines indicate the values in the absence of a ZMW.E-G: Z-profiles of the excitation intensity profiles along the central pore axis (top) and the total detected signal S(z) (bottom) under excitation by a focused Gaussian beam (E), plane wave (i.e., widefield) (F) or excitation under TIRF angle (G).The position of the metal membrane is indicated as a gray shaded area.FDTD simulations of fluorescence emission and detected signal from overmilled ZMWs for JFX650.A: Schematic of the simulation setup.B-D: Computed quantum yield Φ, detection efficiency η, and fluorescence lifetime τ profiles of the dye JFX650 as a function of the z position.Dashed lines indicate the values in the absence of a ZMW.E-G: Z-profiles of the excitation intensity profiles along the central pore axis (top) and the total detected signal S(z) (bottom) under excitation by a focused Gaussian beam (E), plane wave (i.e., widefield) (F) or excitation under TIRF angle (G).The position of the metal membrane is indicated as a gray shaded area.Overview of radiative and non-radiative rates obtained from FDTD simulations of fluorescence emission within overmilled ZMWs.Shown are the non-radiative loss rate (top) and radiative rate (bottom) in the presence of the metal nanostructure for the dyes Alexa488 (A) and JFX650 (B).The radiative rate towards the detection side is given as a dashed line, from which the detection efficiency is computed.Note that the given rates are normalized to the rates in the absence of the ZMW as described in the methods.

13 :
Experimental characterization of photophysics and diffusion within ZMWs.A: Jablonski scheme of the photophysics in the ZMW.Dyes are radiatively excited from the electronic ground state S 0 to the first excited state S 1 with rate k ex , from where relaxation can occur radiatively (k r ) or non-radiatively (k nr ).In the waveguide, all displayed rates change due to excitation field enhancement, plasmonic coupling, and metal-induced quenching.The excited state lifetime reports on the sum of the radiative and non-radiative rates.B,C: Fluorescence decays acquired at different pore diameters for a constant milling depth of 100 nm (B) and at different milling depths for a constant pore diameter of 100 nm (C).D: Autocorrelation function of the fluorescence time trace shown in Figure3C.The FCS analysis informs on the average number of particles (N ) and the residence time of molecules within the ZMW (t D ).E,F: FCS curves of the data shown in Figure3D-F.In B,C,E, and F, the curves for the free dye obtained from a free diffusion experiment are shown in gray.FDTD simulations of excitation field and fluorescence emission in the absence of a Pd layer.A,B: Schematic of the simulation setup with (A), without (B) the Pd layer and for the free solution case (C).D,E: Computed excitation field intensity |E 2